Nucl - Phys.B v.771

NUCLEAR PHYSICS B
Journal devoted to the experimental and theoretical study of the fundamental

constituents of matter and their interactions
Editorial Board
Supervisory Editors: G. ALTARELLI, Geneva, Switzerland; W. BARTEL, Hamburg, Germany; C. BECCHI, Genova,
Italy; R.H. DIJKGRAAF, Amsterdam, The Netherlands; D. KUTASOV, Chicago, IL, USA; L. MAIANI, Rome, Italy;
H. MURAYAMA, Berkeley, CA, USA; T. NAKADA, Geneva, Switzerland; H. OOGURI, CalTech, Pasadena, CA, USA;
H. SALEUR, USC, Los Angeles, CA, USA; A. SCHWIMMER, Rehovot, Israel
Associate Editors for Particle Physics: S.J. Brodsky, SLAC; M. Dine, UC, Santa Cruz; G. Dvali, New York; G.W.
Gibbons, Cambridge, UK; D. Gross, St. Barbara; S.S. Gubser, Princeton; L. Ibanez, Madrid; R.L. Jaffe, MIT; K. Kajantie,
Helsinki; M. Luescher, CERN; E. Rabinovici, Jerusalem; L. Randall, Princeton; A. Sen, Bombay; G. Steigman, Ohio
State; G. t Hooft, Utrecht; Y. Totsuka, KEK; P. van Baal, Leiden; S. Weinberg, Texas; J. Zinn-Justin, Saclay
Associate Editors for Field Theory and Statistical Systems: B. Duplantier, Saclay; A.W.W. Ludwig, Santa Barbara;
S. Lukyanov, Rutgers, Piscataway; T. Miwa, Kyoto; G. Mussardo, Trieste; M. Oshikawa, Tokyo; G. Parisi, Frascati;
A.M. Polyakov, Princeton; R. Sasaki, Kyoto; M. Zirnbauer, Cologne
Aims and Scope
Nuclear Physics B focuses on the domain of high energy physics and quantum field theory, and includes three main sections, i.e. particle physics, field theory and statistical systems and physical mathematics. The emphasis is on original
research papers. Conference Proceedings on the topics covered by Nuclear Physics B are published in the (separately
available) journal Nuclear Physics B-Proceedings Supplements.
Abstracted/Indexed in:
Chemical Abstracts/Current Contents: Physical, Chemical & Earth Sciences. Also covered in the abstract and citation
database SCOPUS . Full text available on ScienceDirect .
Subscription Information 2007
Volumes 760787 are scheduled for publication in 84 issues. Publication frequency: weekly. Subscription prices are
available upon request from the Publisher or from the Regional Sales Office nearest you or from this journals website
(http://www.elsevier.com/locate/nuclphysb). A combined subscription to Nuclear Physics A ISSN 0375-9474 (volumes
781797), Nuclear Physics B ISSN 0550-3213 (volumes 760787) and Nuclear Physics B Proceedings Supplements
ISSN 0920-5632 (volumes 164174) is available at a reduced rate. Subscriptions are accepted on a prepaid basis only
and are entered on a calendar year basis. Issues are sent by standard mail (surface within Europe, air delivery outside
Europe). Priority rates are available upon request. Claims for missing issues should be made within six months of the
date of dispatch.
Orders, claims, and journal enquiries: please contact the Customer Service Department at the Regional Sales Office
nearest you:
Orlando: Elsevier, Customer Service Department, 6277 Sea Harbor Drive, Orlando, FL 32887-4800, USA; phone: (+1)
(877) 8397126 [toll free number for US customers], or (+1) (407) 3454020 [customers outside US]; fax: (+1) (407)
3631354; e-mail: usjcs@elsevier.com
Amsterdam: Elsevier, Customer Service Department, PO Box 211, 1000 AE Amsterdam, The Netherlands; phone:
(+31) (20) 4853757; fax: (+31) (20) 4853432; e-mail: nlinfo-f@elsevier.com
Tokyo: Elsevier, Customer Service Department, 4F Higashi-Azabu, 1-Chome Bldg, 1-9-15 Higashi-Azabu, Minato-ku,
Tokyo 106-0044, Japan; phone: (+81) (3) 5561 5037; fax: (+81) (3) 5561 5047; e-mail: jp.info@elsevier.com
Singapore: Elsevier, Customer Service Department, 3 Killiney Road, #08-01 Winsland House I, Singapore 239519;
phone: (+65) 63490222; fax: (+65) 67331510; e-mail: asiainfo@elsevier.com
Advertising information
Europe, USA, Canada and ROW: Advertising orders and enquiries can be sent to: Miss Katrina Barton, phone: (+44)
(0) 20 7611 4117; fax: (+44) (0) 20 7611 4463; e-mail: commercialsales@elsevier.com.
South America: Advertising orders and enquiries can be sent to: Mr Tino DeCarlo, The Advertising Department,
Elsevier Inc., 360 Park Avenue South, New York, NY 10010-1710, USA; phone: (+1) (212) 633 3815; fax: (+1) (212)
633 3820; e-mail: t.decarlo@elsevier.com.
US mailing notice Nuclear Physics B (ISSN 0550-3213) is published weekly by Elsevier B.V., P.O. Box 211, 1000
AE Amsterdam, The Netherlands. The annual institutional subscription price in the USA is US$ 17015 (valid in North,
Central and South America), including air speed delivery. Periodical postage paid at Rahway NJ and additional mailing
offices. USA Postmaster: Send change of address: Nuclear Physics B, Elsevier, 6277 Sea Harbor Drive, Orlando, FL
32887-4800. Airfreight and mailing in the USA by Mercury International Limited, 365, Blair Road, Avenel, NJ 07001.
Nuclear Physics B 771 (2007) 127
Neutrino mass hierarchy and 13 with a magic baseline

beta-beam experiment
Sanjib Kumar Agarwalla a,b , Sandhya Choubey a, ,
Amitava Raychaudhuri a,b
a Harish-Chandra Research Institute, Chhatnag Road, Jhunsi, Allahabad 211019, India
b Department of Physics, University of Calcutta, 92 Acharya Prafulla Chandra Road, Kolkata 700009, India
Received 1 November 2006; accepted 1 February 2007

Available online 23 February 2007
Abstract
We underscore the physics advantage of an experiment where neutrinos produced in a beta-beam facility
at CERN are observed in a large magnetized iron calorimeter (ICAL) at the India-based Neutrino Observatory (INO). The CERN-INO distance is close to the so-called magic baseline which helps evade some of
the parameter degeneracies and allows for a better measurement of the neutrino mass hierarchy and 13 . We
expound the possibility of using radioactive 8 B and 8 Li as the source isotopes for the e and e beta-beam,
respectively, and show that very good sensitivity to both the mass hierarchy and 13 is possible with a boost
in the 250500 ballpark.
2007 Elsevier B.V. All rights reserved.
PACS: 14.60.Pq; 14.60.Lm
1. Introduction
Spectacular results from a series of experiments involving solar [1], atmospheric [2], reactor
[3], and accelerator [4,5] neutrinos have firmly established neutrino oscillations and heralded
the precision era in neutrino physics. Next generation experiments have been planned/proposed
world-wide to further pin down the values of the oscillation parameters, m221 1 and 12 for solar
* Corresponding author.
E-mail addresses: sanjib@mri.ernet.in (S.K. Agarwalla), sandhya@mri.ernet.in (S. Choubey),

raychaud@mri.ernet.in (A. Raychaudhuri).
1 We define m2 = m2 m2 .
ij
i
j
0550-3213/$ see front matter 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2007.02.012
S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127
neutrinos and m231 and 23 for atmospheric neutrinos. However, a vital task to be undertaken in
the immediate future is the determination of the hitherto unknown mixing angle 13 .2 Discovery
of a non-zero value for 13 would open up the possibility of observing CP-violation in the lepton
sector. Another outstanding problem concerns the ordering of the neutrino mass states, aka, the
neutrino mass hierarchy. If Nature entertains a large enough mixing angle 13 , it could be possible
to ascertain sgn(m231 ) and hence the neutrino mass hierarchy.3
The neutrino mass hierarchy can be probed in terrestrial experiments through matter effects
[810]. Large matter effects are known to exist in the oscillation probability e (Pe ).
Since it also provides the best way of determining the mixing angle 13 and the CP phase, CP ,
it has been popularly hailed as the golden channel [11]. The Pe channel can be studied in experiments which use an initial e (or e ) beam and a detector which can efficiently see muons.4
An absolutely pure and intense e (or e ) flux can be produced using beta beams (-beam)
[12]. There is a possibility of such a facility coming up at the CERN accelerator complex and
an international R&D effort is underway to check the feasibility of this proposal [13]. The Tevatron at FNAL has also been discussed as a potential accelerator for producing -beams. Most
studies involving beta-beams have concentrated on using megaton water Cerenkov

detectors for
observing oscillations through muons produced via (or ). The reason is that these analyses
consider low to medium values for the acceleration of the unstable ions which create a relatively low energy (anti)neutrino beam. The detector therefore should be capable of observing the
resultant low energy muons efficiently and unambiguously. Water Cerenkov

detectors are particularly well suited for this. However, in contrast to most of the previous work involving -beams,
which have mainly shown the physics potential of these kinds of experiments in determining 13
and CP , in this paper we are especially interested in deciphering the neutrino mass hierarchy
and 13 simultaneously. To that end, we suggest a -beam with (anti)neutrino energies where
we expect the largest earth matter effects, that is, energies in the multi-GeV regime. These high
energy (or ) can be very efficiently observed in large magnetized iron detectors, such as the
envisaged ICAL detector at the India-based Neutrino Observatory [14].
A serious complication with all long baseline experiments involving the golden Pe channel
arises from degeneracies which manifest in three forms:
the (13 , CP ) intrinsic degeneracy [15],
the (sgn(m231 ), CP ) degeneracy [16],
the (23 , /2 23 ) degeneracy [17].
This leads to an eight-fold degeneracy [18], with several spurious or clone solutions in addition
to the true one and severely deteriorates the sensitivity of any experiment. It has been shown [18,
19] that the problem of clone solutions due the first two types of degeneracies can be evaded by
choosing the baseline of the experiment equal to the characteristic refraction length due to the
2 Currently, only an upper bound (sin2 2 < 0.17 at 3 [6,7]) exists.
13
3 The neutrino mass hierarchy is termed normal (inverted) if m2 is positive (negative). Note that the terms
31
normal and inverted hierarchy in this article refers to the neutrino mass ordering only, i.e., whether 3 is heavier or
lighter than 2 , and our discussions are valid even for a quasi-degenerate neutrino mass spectrum.
4 This is complementary to the standard accelerator beams where the initial (anti)neutrino flux consists of (or
) and where the relevant channel is the probability Pe . Produced from decay of accelerated pions, these conventional
(anti)neutrino beams suffer from an additional hurdle of an intrinsic e ( e ) contamination, which poses a serious problem
of backgrounds. A -beam is comprised of pure e (or e ).
matter inside earth [8,1820]. This special value goes by the colloquial name magic baseline
[19]. As we will discuss in detail later, at this baseline the sensitivity to the mass hierarchy and,
more importantly, 13 , goes up significantly [19,21], while the sensitivity to CP is absent.
Interestingly, the CERN-INO distance of 7152 km happens to be tantalizingly close to the
magic baseline. This large baseline also enhances the matter effect and requires traversal through
denser regions of the Earth. Thus, for neutrinos (antineutrinos) with energies in the range 3
8 GeV sizable matter effects are induced if the mass hierarchy is normal (inverted). A unique
aspect of this set-up is the possibility of observing near-resonant matter effects in the e
channel. In fact, to our knowledge, what we propose here is the only experimental situation where
near-resonant matter effects can be effectively used in a long baseline experiment to study the
neutrino mass matrix. We show in this paper that the presence of this near-maximal earth matter
effect not only maximizes the sensitivity to the neutrino mass hierarchy, it also gives the experiment an edge in the determination of the mixing angle 13 . The increase in the probability Pe
due to near-resonant matter effects, compensates for the fall in the -beam flux due to the very
long baseline, so that one can achieve sensitivity to 13 and mass hierarchy which is comparable,
even better, than most other proposed experimental set-ups. Therefore a -beam experiment with
its source at CERN and the detector at INO could emerge as a powerful tool for a simultaneous
determination of the neutrino mass hierarchy and 13 .
In [22] such an experimental set-up was considered for the first time and the physics potential
explored. It was demonstrated that both the neutrino mass hierarchy as well as 13 may be probed
through such a set-up. The ions considered for the -beam in that work were the most commonly
used 18 Ne (for e ) and 6 He (for e ). One requires very high values of the Lorentz boost for these
ions ( 103 ) because the energy E of the beam has to be in the few GeV range to enable
detection in the ICAL detector, which is expected to have a threshold of about 1 GeV.5 Such high
values of , although possible in principle, might turn out to be very difficult to realize. Very
recently two other ions, 8 B (for e ) and 8 Li (for e ), have been projected as viable options for the
-beam source [23]. The main advantage that these ions offer is their substantially higher endpoint energy, E0 . This allows one to access E few GeV very easily with medium values of ,
that could be possible to achieve with either the existing CERN technology, or with the projected
upgrades. The physics potential of these alternative ions to study standard neutrino oscillation
parameters in the context of a CERN to Gran Sasso -beam experiment has been expounded in
[24], while in [25] they have been studied to test R-parity violating SUSY.
In this paper we will work with 8 B and 8 Li as the candidate ions for the -beam source at
CERN and ICAL at INO as the far detector. We perform a 2 analysis of the projected data in
a future CERN-INO -beam experiment and present our results for three plausible benchmark
values of the Lorentz boost factor . We begin in Section 2 with a brief description of the beam produced at the source. In Section 3 we discuss the oscillation probability and highlight
the importance of the magic baseline and the near-resonant matter effects in the Pe channel.
Section 4 contains our results for the event rate expected in the ICAL detector. In Section 5 we
present the details of the statistical analysis used. The sensitivity of the results on the mixing
angle 13 are presented in Section 6, while Section 7 focuses on the sensitivity to the neutrino
mass hierarchy. In Section 8 we consider exploiting the potential of muon charge identification
at ICAL at INO through a -beam running simultaneously in both the e and e mode. We end
in Section 9 with our conclusions.
5 Most studies on the physics reach of -beam experiments have energies in the few hundred MeV range.
2. The -beam fluxes

The concept of a -beam was proposed by Zucchelli [12,26]. The idea is to produce a pure,
intense, collimated beam of e or e through the beta decay of highly accelerated and completely
ionized radioactive ions circulating in a storage ring [27]. The resultant beam therefore comprises
only of e (or e ) and has definite characteristics, determined by the well known beta decay
spectrum and the Lorentz boost factor. The former feature results in very low backgrounds while
the latter ensures that there are essentially negligible flux systematics. This proposal is being
studied in depth and fully exploits the existing CERN accelerator complex. The main future
challenge lies in manufacturing an intense proton driver and a hippodrome-shaped decay ring
which are at the core of this program.
2.1. The beta decay spectrum
In the rest frame of the radioactive ion the beta decay spectrum is given by

1
d 2
(E
=
E
)E
(E0 E )2 m2e ,
0
d dE 4 m5e f
(1)
where me is the electron mass, E0 the electron total end-point energy, E and are the neutrino
energy and lifetime of the decaying ion respectively in the latters rest frame and f is a function
of me /E0 . In the lab frame, the flux of the unoscillated -beam at the detector is given by

1 N
1
2
(E0 E )2 m2e ,
(E, ) =
(2)
(E0 E )E
4L2 m5e f (1 cos )
where L is the distance between the source and detector, N is the number of useful decays in the
storage ring per unit time, is the angle between the neutrino flight direction and the direction
in which the ions are boosted6 and is the Lorentz boost such that E = E(1 cos ),
E being the neutrino energy in the lab frame. The maximum energy of neutrinos produced by a
-beam with a Lorentz factor is given by
Emax =
(E0 me )
.
(1 cos )
(3)
2.2. Candidate ions for the -beam

From Eq. (2) it is seen that the total flux and energy of the -beam at the far detector depends
mainly on the end point energy E0 of the beta decay ion and the Lorentz boost factor . The
flux increases as both E0 and increase. Larger results in better collimation of the impinging
flux, thereby increasing the statistics. The spectrum also shifts to larger energies as E0 and
increase. The neutrino cross section in the detector increases with energy, so for the same total
flux a harder spectrum further enhances the statistics. Since the flux falls as 1/L2 as the sourcedetector distance increases, high values of either E0 or or both are needed to have sizable
number of events at the far detector. E0 is an intrinsic property of the decaying ion while is
restricted by the nature of the accelerators and the machine design. The other desired properties
which the ion should have include high production yield, large decay fraction, and a life time that
6 We work with the on-axis flux for which = 0.
Table 1
Beta decay parameters: lifetime , electron total end-point energy E0 , f -value and decay fraction for various ions [29]
Ion
18 Ne
10
6 He
2
8B
5
8 Li
3
(s)
E0 (MeV)
Decay fraction
Beam
2.41
3.92
820.37
92.1%
1.17
4.02
934.53
100%
1.11
14.43
600684.26
100%
1.20
13.47
425355.16
100%
Fig. 1. The unoscillated e spectrum at INO-ICAL from the 8 B (left-hand panel) and 18 Ne (right-hand panel) -beam
source at CERN, for three different choices of the boost factor = 250 (black dot-dashed curve), 350 (blue dashed curve),
and 500 (red solid curve). (For interpretation of the references to colour in this figure legend, the reader is referred to the
web version of this article.)
is long enough to allow the ions to be adequately accelerated. It is also easier to store larger number of lower-Z isotopes in the storage ring [28]. In Table 1 we show the characteristic features of
the four different ions which have been discussed in the literature as possible candidates for the
-beam. While 18 Ne and 8 B are + emitters (producing a e beam), 6 He and 8 Li are emitters (producing a e beam). They all have comparable life times, conducive to the requirements
necessary for the -beam, very high (or maximal) decay fraction and very low A/Z ratio. Note
that the end-point energies, E0 , for 8 B and 8 Li are much larger than those for 18 Ne and 6 He.
In this paper we are interested in the physics potential of an experimental set-up with a beam at CERN and a large magnetized iron detector in India. For such a large baseline, one needs
a very high value of , both to cross the detector energy threshold as well as to get reasonable
statistics in the detector. In particular, with 6 He (or 18 Ne) as the source ion, one needs 1000.
Such high values of can only be achieved by using the LHC itself. As noted above, E0 for
the alternative options for the radioactive ion source (8 B and 8 Li) is higher by a factor of about
four. This means that it should be possible to produce high intensity, high energy beams with 8 B
and 8 Li for a much lower boost factor.7 We show in Fig. 1 the 8 B (left-hand panel) and 18 Ne
(right-hand panel) -beam flux expected at INO as a function of the e energy, for three different
7 The loss in collimation is not significant as we show later.
Fig. 2. Same as in Fig. 1, but for the e spectrum from 8 Li (left-hand panel) and 6 He (right-hand panel) source.
benchmark values of . Fig. 2 shows the corresponding spectra for the 8 Li (left-hand panel) and
6 He (right-hand panel) -beam. Note that even though apparently it might seem from the
e
figures that the e ( e ) flux is larger for 18 Ne (6 He), in reality, for a given , the total flux is given
by the area under the respective curves. One can easily check that for a fixed this is same for
both the ions as expected, since we have assumed equal number of decays for both 18 Ne and 8 B
for e (6 He and 8 Li for e ).8 Also note that even though the total e ( e ) flux remains the same for
both the ions for a given , the number of events produced in the detector is much enhanced for
the 8 B (8 Li) beam since the energy of the beam is larger and the charged current cross sections
increase with the neutrino energy. Studies have shown that it is possible to accelerate 18 Ne to
250 with the existing facilities at CERN, while = 250600 should be accessible with
the super-SPS, an upgraded version of the SPS with super-conducting magnets [30,31]. The
Tevatron at FNAL could in principle also be used to produce a -beam with < 600.
In the low design of beta-beams 2.9 1018 6 He and 1.1 1018 18 Ne useful decays per year
should be possible to achieve. Earlier, only these ions were considered because it was believed
that 8 B could not be produced with the standard ISOLDE techniques. Since most exercises focused on observation of CP-violation, it was necessary to have both e and e beams with similar
spectra, so 8 Li (though considered in [28]) was also generally ignored. Interest in both these ions
have been rekindled recently [23,24], as it appears that having intense 8 B and 8 Li fluxes should
be feasible using the ionization cooling technique [23]. In what follows, we will vary the value of
to test the physics potential of the CERN -beam INO-ICAL set-up and, following the current
practice, assume that it is possible to get 2.9 1018 useful decays per year for 8 Li and 1.1 1018
for 8 B for all values of .
3. Oscillation phenomenologythe magic baseline
In this section we will very briefly review some issues related to the neutrino oscillation
probability. We consider the standard form of the neutrino mixing matrix
8 Of course, the on-axis flux increases with because of better collimation of the beam.
Table 2
Chosen benchmark values of oscillation parameters, except sin2 213
|m231 | = 2.5 103 eV2
sin2 223 = 1.0
m221 = 8.0 105 eV2
sin2 12 = 0.31
CP = 0

c12 c13
s12 c13
s13 eiCP
i
i
U = s12 c23 c12 s23 s13 e CP
(4)
,
c12 c23 s12 s23 s13 e CP
s23 c13
s12 s23 c12 c23 s13 eiCP
c12 s23 s12 c23 s13 eiCP
c23 c13
where cij = cos ij , sij = sin ij and we ignore the Majorana phases. From the analyses of the
current solar, atmospheric, reactor, and accelerator data set one has the following 3 constraints
on the oscillation parameters [7]:
7.2 105 eV2 < m221 < 9.2 105 eV2 and 0.25 < sin2 12 < 0.39,

2.0 103 eV2 < m231 < 3.2 103 eV2 and sin2 223 > 0.9,
2
sin 213 < 0.17,
(5)
(6)
(7)
Throughout this paper we show all our results assuming certain true values for the oscillation
parameters |m231 |, sin2 23 , m221 , sin2 12 and CP given in Table 2. We will let sin2 213 take
all possible true values (which we will call sin2 213 (true)) and present our result as a function of
sin2 213 (true).
Since both sin2 213 and m221 /m231 are very small parameters, as can be seen from
the current constraints, we can expand the so-called golden channel probability in the constant
matter density approximation as9 [11,32]
Pe sin2 23 sin2 213
sin2 [(1 A)]

2
(1 A)
+ sin 213 sin 212 sin 223 sin CP sin()
sin[(1 A)]
sin(A)
A
(1 A)
+ sin 213 sin 212 sin 223 cos CP cos()

+ 2 cos2 23 sin2 212
sin2 (A)
,
A 2
sin[(1 A)]
sin(A)
A
(1 A)
(8)
where
m231 L
,
4E
A
,
A
m231

(9)
(10)
9 This particular low order expansion of the transition probability is valid only in the range of L and E where the
resonance condition A = 1 is never reached. For the E and L range that we consider in this paper, this condition is
satisfied and Eq. (8) fails, as we discuss later in this section. Nonetheless, we present this expression to illustrate the
effect of the magic baseline.
A = 2 2GF Ne E being the matter potential, where Ne is the electron number density inside the
Earth and we have expanded the probability in and sin 213 , keeping only up to second order
terms in both these parameters. The + () sign in Eq. (10) refers to neutrinos (antineutrinos).
The first term of Eq. (8) can be used to extract information about the value of 13 . This is also
the term which has the largest earth effect and this effect of matter can be used to determine
the sign of m231 . The second (CP violating) and third (CP conserving) terms depend on the
CP phase CP and can be used to find it. The last term is independent of both 13 and CP and
depends mainly on the solar parameters m221 and 12 . In very long baseline experiments, this
term has sizable matter effects and if the true value of 13 turns out to be zero (or nearly zero),
this would be the only surviving term in Pe which could still be used to study matter enhanced
oscillations [33].
Despite its advantage in determining the most interesting oscillation parameters, this channel
is, however, rife with the problem of degeneraciesthe 13 CP intrinsic degeneracy [15],
the sgn(m231 ) degeneracy [16] and the octant of 23 degeneracy [17]leading to an overall
eight-fold degeneracy in the parameter values [18]. Various schemes to tackle some or all of
these problems have been studied in great detail in the literature. These can be broadly categorized into ones where data from experiments at different energies and different baselines are
combined together [15,3438], where data from accelerator experiments observing different oscillation channels are considered together [39,40], and where data from atmospheric [41,42] or
reactor experiments [43] are combined with the golden channel measurements in an accelerator
experiment.
A particularly interesting scenario arises when the condition
=0
sin(A)
(11)
is satisfied. In such an event, the last three terms in Eq. (8) drop out and the Pe channel enables
a clean determination of 13 and sgn(m231 ).
= AL/4E by definition, the condition (11) reduces to L = 2/GF Ye , where

Since A
Ye is the electron fraction inside the earth. This gives
L
32725,
[g/cc] [km]
(12)
which for the PREM [44] density profile of the earth is satisfied for the magic baseline [19]
Lmagic 7690 km.
(13)
We refer the reader to [20] for a very recent and enlightening discussion on the physical meaning
of the magic baseline.
The CERN-INO distance corresponds to L = 7152 km, which is tantalizingly close to the
magic baseline. We therefore expect that the -beam experiment we consider here should give
an essentially degeneracy-free measurement of both 13 and sgn(m231 ).
The large CERN-INO baseline, of course, results in very significant earth matter effects in
the Pe channel. In fact, for the baseline of 7152 km, the average earth matter density calculated
using the PREM earth matter density profile is av = 4.13 gm/cc, for which the resonance energy
|m231 | cos 213
2 2GF Ne
= 6.1 GeV,
Eres
(14)
(15)
Fig. 3. The transition probability Pe as a function of E for four values of the baseline L. The band reflects the effect
of the unknown CP . The dark (red) shaded band is for the normal hierarchy (NH) while the light (cyan) shaded band
is for the inverted hierarchy (IH). We have taken sin2 213 = 0.1 and for all other oscillation parameters we assume the
benchmark values given in Table 2. (For interpretation of the references to colour in this figure legend, the reader is
referred to the web version of this article.)
for |m231 | = 2.5 103 eV2 and sin2 213 = 0.1. One can check (cf. Figs. 1 and 2) that this is
roughly in the ballpark where we expect the maximum -beam flux for 8 B and 8 Li ions with the
benchmark values used in this paper.
Note that the low order expansion of the probability Pe given by Eq. (8) is valid only for
values of E and earth matter density (and hence L) where flavor oscillations are far from resonance, i.e., A 1. In the limit A 1, one can check that even though the analytic expression
for Pe given by Eq. (8) remains finite, the resultant probability obtained is incorrect [45]. We
reiterate that Eq. (8) was presented only in order to elucidate the importance of the magic baseline. For all the numerical results presented in this paper, we calculate the exact three generation
oscillation probability using the realistic PREM [44] profile for the earth matter density.
The exact neutrino transition probability using the PREM density profile is given in Fig. 3 as
a function of the neutrino energy, for four different baselines (four panels). We allow CP to take
on all possible values between 0360 degrees and the resultant probability is shown as a band,
with the thickness of the band reflecting the effect of CP on Pe . The figure is drawn assuming
the benchmark values of the oscillation parameters given in Table 2 and sin2 213 = 0.1. We show
the probability for both the normal hierarchy (dark band) as well as the inverted hierarchy (light
band). As discussed in detail above, for L = 7500 km, which is close to the magic baseline, the
effect of the CP phase is seen to be almost negligible, while for all other cases the impact of CP
on Pe is seen to be appreciable. In fact, for L = 1000 km, the probability corresponding to the
normal and inverted hierarchy become almost indistinguishable due to the uncertainty arising
from the unknown value of CP . As the baseline is increased, earth matter density increases,
enhancing the impact of matter effects. The probability for normal hierarchy is hugely enhanced
10
Fig. 4. The dark (red) shaded band is the same as in Fig. 3. The light (green) shaded band shows the corresponding Pe
for sin2 213 = 0.05. Values of all the other oscillation parameters are same as in Fig. 3 and the hierarchy is assumed to
be normal. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of
this article.)
for the neutrinos, while for the inverted hierarchy matter effects do not bring any significant
change. This difference in the predicted probability, evident in the panels corresponding to L =
4000, 7500 and 10 000 km of Fig. 3, can be used to determine the neutrino mass hierarchy.
In Fig. 4 we display the dependence of the neutrino probability Pe on the mixing angle 13 for
four different baselines. The dark bands, as in Fig. 3, are for normal hierarchy and sin2 213 = 0.1
with full variation of CP , while the light bands are for normal hierarchy and sin2 213 = 0.05.
The impact of matter effect in increasing the 13 sensitivity of a given experimental set-up is
evident from the figure. The 13 sensitivity for L = 1000 km can be seen to be much weaker than
for the other cases, since matter effects are smaller. However, the most striking feature seen in
Fig. 4 is the effect of the magic baseline in enhancing the sensitivity of the experiment to 13 .
The figure clearly shows that the difference in the predicted Pe for the two values of sin2 213 is
largest for L = 7500 km (since effect of CP is the least) and thus an experiment at this baseline
is most suitable for probing 13 . Figs. 3 and 4 therefore reinforce our choice of the near-magic
baseline as one of the best options for determining the neutrino mass hierarchy and 13 , since
both these parameters are directly related to large matter effects and the uncertainty of CP could
prove to be a hindrance in their measurement at non-magic baselines.
4. Event rates in INO-ICAL
4.1. The ICAL detector at INO
The proposed large magnetized iron calorimeter at the India-based Neutrino Observatory [14]
is planned to have a total mass of 50 kton at startup, which might be later upgraded to 100 kton.
11
Table 3
Detector characteristics used in the simulations
Total mass
Energy threshold
Detection efficiency ()
Charge identification efficiency (fI D )
50 kton
1.5 GeV
60%
95%
The INO facility is expected to come up at PUSHEP (lat. North 11.5 , long. East 76.6 ), situated
close to Bangalore in southern India. This constitutes a baseline of 7152 km from CERN. The
ICAL detector will have a modular structure with a total lateral size of 48 m 16 m, divided
into three modules of 16 m 16 m each. Each of these modules will have 140 horizontal layers
of 6 cm thick iron plates, separated from each other by a gap of 2.5 cm to hold the active
detector material, giving a total height of 12 m for the full detector. The active detector elements
will be resistive plate chambers (RPC), made from either glass or Bakelite and will be filled with
a suitable gas mixture, which will be recycled with approximately one volume change per day.
An external magnetic field of 1.3 Tesla would be applied over the entire detector. The detector
will be surrounded by an external layer of scintillator or proportional gas counters which will
act both as veto to identify external muon backgrounds as well as to identify partially contained
events.
According to the detector simulation performed by the INO Collaboration, the detector energy
threshold for is expected to be around 1 GeV and charge identification efficiency will be
about 95%. In what follows, we will present our numerical results assuming an energy threshold of 1.5 GeV, detector charge identification efficiency as 95% and unless stated otherwise,
detection efficiency as 60% for (cf. Table 3). The 60% detection efficiency we have used is
an extremely conservative estimate and it should be possible to increase it to at least 80%. We
shall therefore also present our main results for this more optimistic estimate for the detector
efficiency. However, unless it is mentioned otherwise, all our results correspond to the detector
characteristics given in Table 3. We have explicitly checked that our results remain unaffected
if the energy threshold is raised to 2 GeV for the entire range of assumed Lorentz boost factor
= 250500. For > 350 the threshold can be even 3 GeV, while for > 500 one can work
with an energy threshold of 4 GeV, without changing the final results. This can be seen from
Figs. 1 and 2; for = 350(500), the majority of neutrinos arriving at INO-ICAL would have
E > 3(4) GeV.
4.2. Oscillation signal
The total number of (+ ) events collected in INO-ICAL from a e (or e ) -beam exposure over a period of T years is given by,
E
max
N = T n n fI D
dE
0
dEA (E) (E)R(E, EA )Pe (E)
(16)
th
EA
where nn are the number of target nucleons, fI D is the charge identification efficiency, is
the detection efficiency, (E) (given by Eq. (2)) is the -beam flux at INO-ICAL in units of
m2 yr1 GeV1 , is the detection cross section for in units of m2 , E is the true energy
12
Fig. 5. The expected number of events in 5 years running time, as a function of sin2 213 . The value of and the hierarchy
chosen corresponding to each curve is shown in the figure legend.
of the incoming neutrino and EA is the measured energy,10,11 R(E, EA ) is the detector energy
resolution function,12 E max is the maximum energy of the neutrinos for a given Lorentz factor
(given by Eq. (3)) and Pe is the e oscillation probability. The expression for the +
signal in the detector from a e -beam flux is given by replacing e by e , by and Pe
by Pe . For the neutrinonucleon interaction we consider quasi-elastic scattering, single-pion
production and deep inelastic scattering and use the cross sections given in the Globes package
[46] which are taken from [47,48].
We show in Fig. 5 the number of events expected in INO-ICAL from 5 years exposure of
an 8 B (for e , shown in the left-hand panel) or 8 Li (for e , shown in the right-hand panel) beam from CERN. The expected number of events is presented as a function of sin2 213 for
both normal (NH) and inverted (IH) hierarchy for three benchmark values of the boost factor .
Large resonant matter effects in the neutrino channel for normal hierarchy drives the number of
expected events to very large values, compared to what would be expected for inverted hierarchy. Similarly, in the antineutrino channel we have resonant matter effects for inverted hierarchy
and the number of predicted events is many times larger than for normal hierarchy.13 This difference in the number of events is seen to increase with sin2 213 up to a certain (large) value,
beyond which it starts to decrease. This limiting value of sin2 213 at which the number of events
in the neutrino (antineutrino) channel peaks for the normal (inverted) hierarchy comes from an
interplay of two effects which we now discuss. In the approximation where m221 can be neM ) and the
glected, the probability Pe would be largest if the mixing angle in matter (sin2 213
2
2
M
mass squared difference driven oscillatory term in matter (sin [(m31 ) L/4E]) are maximum
simultaneously. In other words, largest matter effects come when the resonance energy Eres is
10 The measured energy of the neutrino can be reconstructed from the total measured energy of the muon and the
accompanying hadron shower.
11 The integration over the reconstructed (measured) neutrino energy is done from the threshold energy E th (taken as
A
1.5 GeV in this paper) to infinity.
12 We assume a Gaussian resolution function with = 0.15E.
13 In fact, matter effects are seen to suppress the oscillation probability for neutrinos (antineutrinos) compared to that in
vacuum when the hierarchy is inverted (normal).
13
almost equal to the energy at which the oscillatory factor goes to one [4951]. For a given L (and
hence matter density, ), one can calculate the value of sin2 213 for which this condition can be
satisfied and it is given by [50,51]
tan 213 =
16.18 103
.
(gm/cc)L(km)
(17)
For L = 7152 km, it turns out that this value is sin2 213 0.23. Therefore, matter effects and
hence the event rate in the detector keeps increasing until this value of sin2 213 is reached.
Beyond this limiting value of sin2 213 , effect of earth matter in Pe falls and the event rate for
the neutrinos, for normal hierarchy, and antineutrinos, for inverted hierarchy, decreases. However,
note that this value of sin2 213 is already disfavored from the CHOOZ data [6,7].
From the left-hand panel of Fig. 5 we note that for = 500 and sin2 213 = 0.05, the predicted
number of neutrino events for normal hierarchy is 101, while that for inverted hierarchy is only 3.
This implies that if the normal hierarchy was true, we could comprehensively rule out the wrong
inverted hierarchy. For smaller boost factors , the difference between the number of events
decreases; however, we could still determine the neutrino mass hierarchy if sin2 213 (true) is not
very small. We will perform a statistical analysis of projected data in the following sections and
present our results on the hierarchy sensitivity in Section 7. Instead of the neutrino, one could
work with the e -beam and achieve similar sensitivity to the neutrino mass hierarchy. Note
that while the interaction cross section for the e s are much smaller, the flux itself is larger owing
to the (assumed) larger number of decays per year for 8 Li. Thus the statistics expected in both
the neutrino as well as the antineutrino channel is comparable and, as we will see, the hierarchy
sensitivities for the e and e -beam are hence similar.
As discussed above, resonant matter effects in the neutrino (antineutrino) channel for the
normal (inverted) hierarchy, results in substantial enhancement in the observed number of events.
In particular, we note from Fig. 5 that the number of events in the near-resonant channels depend
strongly on the value of sin2 213 , since the extent of matter effects is dictated directly by 13 . We
can see from the figure that the dependence of the event rate on sin2 213 is much enhanced due to
matter effects. Therefore, if the true hierarchy was normal (inverted), we could use the neutrino
(antineutrino) -beam in the proposed experiment to measure/constrain the mixing angle 13 .
In Fig. 6 we show the effect of CP on the variation of the measured rate with the value of the
mixing angle 13 , for four different baselines 732 km, 4000 km, 7152 km and 10 000 km. The
left-hand panel shows the results for the e -beam assuming a normal mass hierarchy, while the
right-hand panel gives the same for the e -beam flux with inverted hierarchy. For all cases in
any panel, we have considered a 50 kton magnetized iron calorimeter as the far detector and the
same flux created at the source. Each curve gives the sets of values of {sin2 213 , CP } which give
the same observed rate in the detector as the set {sin2 213 = 0.04, CP = 0 }. In other words, if
the true value of sin2 213 and CP were 0.04 and 0 respectively, then every point on a given
curve would also be a solution for that experiment. It is clear from this figure that combining
results from experiments at different baselines helps solve/reduce the problem. However, the
most important issue exemplified here is the fact that for the baseline 7152 km, which is the
CERN-INO distance, the effect of the unknown value of CP on the measurement of 13 and the
mass hierarchy, is very small. This happens because this distance corresponds to a near-magic
baseline for which, as noted earlier, the CP dependent terms are almost vanishing.
14
Fig. 6. The iso-event curves in the CP - sin2 213 plane for four baselines are shown in the figure. The true values of
CP and sin2 213 are assumed to be 0 and 0.043 respectively. Left- (right-) hand panel is for the e ( e ) -beam. The
assumed hierarchy is mentioned in the figure.
4.3. Backgrounds
The possible backgrounds in a e -beam experiment14 using as an oscillation signal
come from neutral current events such as
x + d(u) x + d(u),
(18)
x + d(u) x + d(u) + q q,
(19)
and e charged current events

e + d e + u
or c (Cabibbo suppressed),
e + d e + u + q q.
(20)
(21)
The quarks in the final state could produce mesons as a part of the hadronic junk. These mesons
may then decay producing secondary muons, giving rise to a signal which might constitute a
possible background.
As discussed earlier, INO-ICAL will have 6 cm thick iron plates. Such a dense tracking detector will have excellent muon/pion and muon/electron separation capability in the energy range
we are working with. The simulations carried out by the INO Collaboration have shown that after
the standard kinematical cuts are imposed, the electrons do not give any signal at all, while in
99% of the cases, the pions and kaons get absorbed very quickly in the iron for the energy range
of interest to us and therefore do not hit enough RPCs to give a signal. At the energies of the
-beams considered here, production cross section of D mesons (also Cabibbo suppressed) is
small and they do not constitute a problem for the experiment. The associated strange or charm
production is also highly suppressed at these energies. In addition, the fact that the detector will
have a charge identification capability means that secondary + produced can be safely dis14 The discussion concerning the -beam is similar and hence is not repeated here.
e
15
carded, reducing the background even further. Therefore, in this analysis we do not consider any
backgrounds coming from either the neutral current events or charged current events of e .15
Since the oscillation probability e is about the same as that for e , we expect
almost as many arriving at the detector as . The produced through charged current
interaction may decay producing secondary with a branching ratio of 17.36%. But, the
threshold (3.5 GeV) is high and the production cross section suppressed compared to . So
we do not expect any significant background from this source either. We have estimated the
number of secondary muons produced from the component of the beam. For sin2 213 = 0.01,
we expect 0.008, 0.061 and 0.2 muon events per year respectively, for = 250, 350 and 500.
In addition, these secondary muons will be severely degraded in energy and therefore can be
eliminated through energy cuts. We therefore neglect the backgrounds from this source as well.
5. Details of the statistical method
In order to quantify the sensitivity of this -beam experimental set-up to the mixing angle 13
and sgn(m231 ), we perform a statistical analysis of the data generated in INO-ICAL, assuming
certain true values of the parameters. Since the number of events expected in the detector might
be very small, depending on how small the true value of sin2 213 is, we define a 2 assuming
Poissonian distribution for the error as,

N th
k2 .
2 {} = min 2 N th N ex N ex ln ex +
(22)
k
N
k
is the observed number of events, while

In Eq. (22), {} is a set of oscillation parameters,
the systematic errors in the data and the theory are accounted for through the set of pulls {k },
where k runs over the different systematic uncertainties involved. The pulls are defined in such a
way that the number of expected events, N th , corresponds to k = 0. N th is the number of events
when the systematic errors are included in a manner such that the effect of the kth uncertainty at
the 1 level is picked up when k = 1:

K

th
th
k
k + O k2 .
N {}, {k } = N {} 1 +
(23)
N ex
k=1
N th ({}, {
In Eq. (23)
k }) has been expanded in powers of k , keeping only linear terms. The
quantities k give the fractional rate of change of N th ({}) due to the kth systematic error.
In our analysis, we include systematic uncertainties in the normalization of the -beam flux at
the source, the error in the cross section and the detector systematic uncertainty. For the flux
normalization error we adopt a total uncertainty of 2%. The neutrinonucleon cross sections,
especially at large energies and for massive target nuclei, are known to have large uncertainties.
We include a 10% error coming from the uncertainty in the interaction cross section. A total
detector systematic uncertainty of 2% is also included. We do not include any systematic error
related to the shape of the energy spectrum since we work only with the energy integrated total
rates, for which such types of errors would be negligible.
15 Mesons produced in neutral current processes are degraded in energy. Note that backgrounds from these mesons are
very important in the case of the neutrino factory. However, since our relevant energy range is lower, the mesons produced
in each event are much lower in energy and hence can be easily rejected by putting suitable cuts. In our analysis we have
assumed cuts that are stringent enough to completely reject these backgrounds.
16
Unless otherwise stated, the data are always generated at the benchmark values of the oscillation parameters |m231 |, m221 , sin2 223 , sin2 12 and CP given in Table 2. The true value of 13
as well as sgn(m231 ) will be allowed to change, since these are the parameters under study in
this experimental set-up of interest. Throughout the paper we will use the notation sin2 213 (true)
as the true value of this parameter chosen by Nature, and sin2 213 as the fitted value. The data,
simulated with the (assumed) true values of the oscillation parameters, are fitted by calculating
N th ({}) for any set of values for the oscillation parameters. We next calculate the chi-square
function 2 ({}, {k }) for every possible N th ({}, {k }), obtained by varying the oscillation parameters and k . We first minimize this 2 ({}, {k }) function with respect to the pulls k . This
2 ({}) (given by Eq. (22)) is next minimized with respect to all the oscillation parameters,16
which are permitted to assume all allowed values in calculating N th ({}).
It is expected that the parameters |m231 | and sin2 223 would be fairly well determined by
the long baseline experiments T2K [53] and NOA [54]. In order to take into account this information, which should be available by the time the -beam facility comes up at CERN, we add to
our 2 the prior function such that
2
2
= 2 + prior
,
total

2

2
|m31 | |m231 (true)| 2
sin 223 sin2 223 (true) 2
2
prior
=
+
,
(m231 )
(sin2 223 )
(24)
(25)
where m231 (true) and sin2 223 (true) are taken as the benchmark values in Table 2. We will
assume that the 1 error on these parameters would be reduced to (m231 ) = 1.5% and
2
with respect to the oscillation
(sin2 223 ) = 1.0% [55]. We minimize the total function total
parameters and present the C.L. results assuming as a convention that n corresponds to
2
total
= n2 ,
2
total
(26)
2
total
where
is the minimum
obtained in the analysis for a given value of the parameter
2
sin 213 (true) and/or the mass hierarchy.
Note that we have not taken into account any error coming from the uncertainty in the matter
density. Fig. 7 shows our justification for this. This figure shows the expected event rate for the
e -beam as a function of sin2 213 . The solid curve corresponds to the PREM profile, while the
dashed and dot-dashed curves are for PREM densities scaled by 5% and +5%, respectively.
We can see that the effect of the 5% uncertainty in the matter density profile shows up in the event
rate only for values of sin2 213 already disfavored by the CHOOZ experiment. In particular, we
note that for sin2 213 0.1 the effect of the matter density uncertainty is negligible.
6. Determining sin2 213
We begin by presenting results on how well sin2 213 can be determined by the set-up we are
proposing. We divide our results into two subcategories. First we assume that sin2 213 (true) = 0
and expound the possibility of improving the upper bound on this parameter. This is what we
16 In order to save computer time, we keep m2 and sin2 fixed in our analysis. There is two-fold motivation for
12
21
this approximation. Firstly, these parameters, especially m221 , is expected to be measured with a high precision in the
proposed solar and reactor neutrino experiments [52]. Secondly, since we are working at a near-magic baseline, we do
not expect any significant impact of these parameters on our results.
17
Fig. 7. Number of events versus sin2 213 for the e -beam with = 500, showing the impact of uncertainty in the
earth matter density profile. The solid curve corresponds to the PREM matter density profile, the dashed (dot-dashed)
curve is for the PREM density scaled by 5% (+5%).
would call the sensitivity of the experiment to sin2 213 . Next we will assume that sin2 213 (true)
is large enough to give a positive signal in the experiment. In that case, we will probe quantitatively how precisely this parameter could then be determined. All results are presented after
marginalizing over |m231 |, sin2 223 , and CP , as described in Section 5.
6.1. Sensitivity to sin2 213
We define the sensitivity to 13 as the minimum value of sin2 213 which this experiment
would be able to distinguish from sin2 213 (true) = 0. The results of our analysis are presented
in Fig. 8, where we show the upper bounds on sin2 213 that this experiment can impose at the
90% C.L. (solid curve for 5 years running and dotted curve for 10 years running) and 3 C.L.
(long dashed curve for 5 years running and dashed-dotted curve for 10 years running). The lefthand panel is for a e -beam assuming that normal hierarchy is the true hierarchy, while the
right-hand panel is for an e -beam with the assumption that the inverted hierarchy is true. We
show our results as a function of the Lorentz boost . The 90% and 3 C.L. upper bounds that
we can impose on sin2 213 are displayed in Table 4 for the three benchmark values of and
with five and ten years of running of the experiment in the neutrino (columns 3 and 4) and in
the antineutrino (columns 5 and 6) mode. In the table we show the sin2 213 sensitivity limits
for the conservative detector efficiency of 60% for as well as for 80% efficiency. This can
be compared with the sensitivity we expect from the other forthcoming/planned experiments.
At the 90% C.L., the T2K and NOA experiments are expected to constrain sin2 213 < 2.3
102 and sin2 213 < 2.4 102 , respectively [55]. The Double-Chooz experiment is likely
to push the limit down to sin2 213 < 3.2 102 , at 90% C.L., which could be improved to
sin2 213 < 9 103 by the Reactor-II set-up [55]. The discovery potential17 of the CERN17 The authors define the discovery potential as the minimum value of sin2 2 (true) for which sin2 2 = 0 gives a
13
13
2 = 9.
18
Fig. 8. The upper limit on sin2 213 that can be imposed by the CERN-INO -beam set-up with 5 yr and 10 yr of running.
The left-hand panel shows the sin2 213 sensitivity expected for the neutrino mode assuming normal hierarchy to be true,
while the right-hand panel shows the sensitivity for the antineutrino mode and an inverted hierarchy.
Table 4
Upper limit on sin2 213 which can be imposed with 5 yr and 10 yr data respectively, at the 90% and 3 C.L., for different
benchmark values of . Columns 3 and 4 show the sensitivity using the e beam and assuming normal hierarchy as true
and columns 5 and 6 are the sensitivity possible with the e beam and assuming inverted hierarchy as true. Results are
shown for 60% and 80% detection efficiency
Detection
efficiency
60%
Years
5 yr
10 yr
80%
5 yr
10 yr
Neutrino beam (NH true)
Antineutrino beam (IH true)
90% C.L.
3 C.L.
90% C.L.
3 C.L.
250
350
500
250
350
500
8.1 103
1.6 103
6.7 104
4.2 103
8.7 104
3.7 104
2.8 102
5.2 103
2.1 103
1.4 102
2.7 103
1.1 103
8.3 103
1.3 103
4.5 104
4.1 103
6.7 104
2.4 104
3.1 102
4.3 103
1.5 103
1.4 102
2.2 103
7.4 104
250
350
500
250
350
500
6.3 103
1.3 103
5.3 104
3.4 103
6.9 104
2.9 104
2.1 102
4.0 103
1.6 103
1.0 102
2.1 103
8.4 104
6.2 103
9.8 104
3.5 104
3.1 103
5.1 104
1.8 104
2.2 102
3.2 103
1.1 103
1.1 102
1.6 103
5.7 104
MEMPHYS project for five years of running of both the -beam and the SPL super-beam set-ups
considered alone is sin2 213 5 103 at 3 [42]. With a combined data set in MEMPHYS
for five years running in -beam and five years in the SPL super-beam, the discovery limit could
be improved to sin2 213 3.5 103 at 3 [42]. The best limit, obtained after a thorough
optimization of the -beam experimental options for the 13 sensitivity reach in [56] (see also
19
[57]), is sin2 213 1.5 103 at 3 , for what the authors label as set-up 118 and set-up 2.19
Therefore, for the experimental set-up with the -beam source in CERN and INO-ICAL as the far
detector the projected sin2 213 sensitivity is comparable to, if not better, than most of the other
planned schemes. In fact, the sensitivity reach of this experiment is even better than that expected
for an entry level Neutrino Factory (NuFact-I) [58]. Only a NuFact-II with L = 7500 km [19]
corresponding to the magic baseline, could give a sin2 213 sensitivity significantly better than
what one can achieve with the CERN-INO -beam set-up [19,21].
We reiterate that the tremendous sin2 213 sensitivity of the CERN-INO -beam set-up comes
not really from high -beam flux, but from the fact that one has near-resonant matter effects
at the near-magic baseline. While the latter feature smothers the problems arising from clone
solutions, the former lends the experiment an extra sensitivity due to the dependence of matter
effects on sin2 213 . This is why the sin2 213 sensitivity depends on the true hierarchy. We have
checked that if we run the experiment for 10 years in the neutrino mode alone and with inverted
hierarchy assumed as true, then the sensitivity we expect is a meagre sin2 213 > 1.8 102 at the
90% C.L. Therefore, a hierarchy independent sin2 213 sensitivity can in principle be obtained
after runs in both the neutrino and the antineutrino mode, unless one is very lucky. We will return
to this issue again in Section 8.
6.2. Precision in measurement of sin2 213
In this subsection we assume that sin2 213 (true) is large enough such that, for a given boost ,
we have a positive signal in the ICAL detector. For each value of sin2 213 (true), varied between
the limiting value obtained in the previous subsection and the current upper limit, we generate the
data and fit it back using the 2 analysis to give us the range of measured values of sin2 213 .
We plot in the upper panels of Figs. 9 and 10, the 90% and 3 C.L. bands of sin2 213 as a function of sin2 213 (true) for 5 years and 10 years run, respectively. The left, middle and right-hand
panels are for = 250, 350 and 500, respectively. Plots are shown for the e -beam assuming
normal hierarchy as true. In order to quantify the precision which can be obtained on sin2 213 ,
we define, for any value of sin2 213 (true),
Precision =
sin2 213 (max) sin2 213 (min)

sin2 213 (max) + sin2 213 (min)
100%,
(27)
where sin2 213 (max) and sin2 213 (min) are respectively the largest and smallest values of
sin2 213 allowed at the given C.L. We show in the lower panels of Figs. 9 and 10, the variation
of this precision function with sin2 213 (true) for the 90% and 3 C.L. for 5 years and 10 years
of running respectively, in the neutrino channel assuming normal hierarchy. Similar results are
expected if we run the experiment in the antineutrino mode for inverted hierarchy.
As expected, the precision in the sin2 213 measurement improves as sin2 213 (true) increases.
However, a very interesting feature in the plots of Fig. 9 is the emergence of minima in the precision, thereby implying that for values of sin2 213 (true) beyond this limiting value, the precision
actually deteriorates.
18 This corresponds to successive 8 yr runs of a -beam in the neutrino and antineutrino mode (total 16 yr) with = 200,
L = 520 km and a water Cerenkov

detector with 500 kton fiducial mass.
19 This corresponds to successive 8 yr runs of a -beam in the neutrino and antineutrino mode (total 16 yr) with = 500,
L = 650 km and 50 kton of totally active scintillator detector.
20
Fig. 9. Precision in the measurement of sin2 213 expected in the CERN-INO -beam experiment with 5 yr running
in the neutrino channel assuming normal hierarchy. Upper panels show the 90% and 3 band of measured values of
sin2 213 for every sin2 213 (true). Lower panels show the corresponding 90% and 3 C.L. precision as defined in the
text, as a function of sin2 213 (true).
Fig. 10. Same as Fig. 9 but for 10 yr of running.
21
Fig. 11. The minimum value of sin2 213 (true) as a function of the boost factor at which the wrong hierarchy can be
disfavored at the 90% and 3 C.L. Results are presented for 5 yr and 10 yr of running in the e (left-hand panel) and e
(right-hand panel) mode, if the true hierarchy was normal and inverted respectively.
7. Sensitivity to Sgn(m231 )
We next turn our attention to the determination of the neutrino mass hierarchy. If the true
value of 13 has indeed been chosen to be large by Nature, we can expect sizable matter effects
in this experimental set-up, giving us a handle on the mass hierarchy. In our analysis, we assume a
certain true mass hierarchy and generate the prospective data in INO-ICAL for the benchmark oscillation parameter values given in Table 2 and for different values of sin2 213 (true). We next fit
this data with the wrong hierarchy and calculate a 2 , after marginalizing over |m231 |, sin2 223 ,
CP as well as sin2 213 , as described in Section 5. The results of this analysis are presented in
Fig. 11. The figure shows, as a function of , the minimum value of sin2 213 (true) for which the
wrong hierarchy can be disfavored at the 90% C.L. (solid curve for 5 yr running and dotted curve
for 10 yr running) and 3 C.L. (long dashed curve for 5 yr running and dashed-dotted curve for
10 yr running). The left-hand panel is for a e -beam and assuming that normal hierarchy is
the true hierarchy, while the right-hand panel is for an e -beam with the assumption that the
inverted hierarchy is true. Of course, since large matter effects appear in only either the neutrino
or the antineutrino channel for a given hierarchy, the sensitivity of the experiment to sgn(m231 )
would be zero if we used a e beam when the true hierarchy was normal or an e beam when the
true hierarchy was inverted.
In Table 5 we show the limiting values of sin2 213 (true), for which the wrong hierarchy can
be disfavored at the 90% and 3 C.L. for three specific values of . We show the results for 5 yr
and 10 yr of running of the experiment in the neutrino and the antineutrino mode. Table 5 shows
the results for both 60% and 80% detection efficiency. This can be directly compared with the
expected potential of the other planned/proposed long baseline experiments in discriminating between the right and the wrong hierarchy. The T2K experiment [53] has a baseline of only 295 km
which entails almost negligible matter effects. The NOA experiment [54] on its own might find
it hard to say anything conclusive about the neutrino mass hierarchy, unless Nature conspires to
arrange oscillation parameters in a conducive way. The so-called T2KK experimental set-up with
the 4 MW beam from the JPARC facility and two detectors, one in Japan and another in Korea
22
Table 5
Minimum value of sin2 213 (true) for which a CERN-INO beta-beam experiment can rule out the wrong hierarchy at
the 90% and 3 C.L. after 5 yr and 10 yr of running, for different benchmark values of . Columns 3 and 4 show the
sensitivity using the e beam assuming normal hierarchy as true and columns 5 and 6 are for the e beam assuming the
inverted hierarchy as true. Results are shown for 60% and 80% detection efficiency
Detection
efficiency
60%
Years
5 yr
10 yr
80%
5 yr
10 yr
Neutrino beam (NH true)
Antineutrino beam (IH true)
90% C.L.
3 C.L.
90% C.L.
3 C.L.
250
350
500
250
350
500
3.7 102
9.7 103
7.7 103
3.0 102
8.3 103
6.6 103
6.6 102
4.7 102
9.2 103
5.9 103
3.8 102
8.1 103
5.4 103
8.7 102
1.4 102
8.2 103
6.0 102
1.1 102
7.0 103
250
350
500
250
350
500
3.4 102
9.3 103
7.4 103
2.8 102
8.1 103
6.6 103
5.7 102
1.3 102
9.8 103
4.2 102
1.1 102
8.5 103
4.3 102
8.9 103
5.9 103
3.5 102
7.8 103
5.4 103
7.3 102
1.3 102
8.2 103
5.4 102
1.0 102
6.9 103
1.5 102
1.0 102
4.7 102
1.2 102
8.7 103
[35],20 is expected to rule out the wrong hierarchy at the 3 C.L. if sin2 213 (true) > 5.5 102
for a total of 8 yr of running, 4 yr in the e and 4 years in the e mode. The CERN-MEMPHYS
-beam and SPL superbeam proposals [42] again concern a baseline of only 130 km, which
is too short to enable these beam experiments to decipher the mass hierarchy on their own. A
higher -beam from CERN to (say) Gran Sasso, with a baseline of 732 km could have sensitivity to the mass hierarchy [24,57]. It was shown in [56] that for = 200 and L = 520 km,
the wrong hierarchy could be disfavored at 3 C.L. if sin2 213 (true) > 2 102 after 8 yr of
simultaneous running in the neutrino and antineutrino mode and with a 500 kton water detector.
Therefore we see that our sensitivity to the neutrino mass hierarchy can be well surpassed only
by a Neutrino Factory, where the right normal hierarchy could be determined at 3 C.L. for
sin2 213 (true) > 3 104 , if the experiment is done at the magic baseline L = 7500 km.
In Fig. 12 we show the minimum sin2 213 (true) as a function of the running time T (in years)
needed in this experimental set-up to discriminate between the wrong and the right hierarchy at
the 90% and 3 C.L. The left-hand panel is for a run in the e mode with true normal hierarchy, while the right-hand panel shows the results for running in the e mode with true inverted
hierarchy. We show the results for the three chosen values of . We see, for example, that if
sin2 213 (true) = 5 102 and the normal hierarchy was true, then the wrong inverted hierarchy could be disfavored at the 3 C.L. after just 9 (26) months of running in the e mode with
= 500 (350).
Fig. 13 shows the impact of CP on the hierarchy sensitivity of this experiment. The left-hand
panel is for the neutrino channel with normal hierarchy as true, while the right-hand panel is
for the antineutrino channel with inverted hierarchy as true. We show the 90% (dashed curves)
and 3 (solid curves) C.L. curves for 5 yr (thick lines) and 10 yr (thin lines) running of the beam with = 350. We note that loss in hierarchy sensitivity due to the uncertainty in CP is
very marginal, owing to the fact that we are proposing this set-up at a near-magic baseline. The
20 For an alternate T2KK proposal, see [36].
23
Fig. 12. The minimum value of sin2 213 (true) as a function of the running time necessary to disfavor the wrong hierarchy
at the 90% and 3 C.L. in the neutrino mode assuming normal hierarchy as true (left-hand panel) and in the antineutrino
mode assuming inverted hierarchy as true (right-hand panel). Results are shown for three values of .
Fig. 13. The effect of CP uncertainty on the limiting value of sin2 213 (true) for which the hierarchy determination will
be feasible for = 350. For all values to the right of the solid curves (dashed curves) it would be possible to rule out
the wrong hierarchy at the 3 (90%) C.L. Thick (thin) lines are for 5 (10) years of running. Left-hand panel is for e
-beam assuming normal hierarchy is true, while the right-hand panel is for e -beam assuming true inverted hierarchy.
best sensitivity to hierarchy comes for CP 125 in the neutrino mode and CP 300 in the
antineutrino mode.
8. Running the e and e -beam simultaneously
An exquisite feature of magnetized iron detectors is their charge identification capability.
Thus, if both the e and the e -beam were produced simultaneously at CERN, the oscillated
24
could be measured distinctly from the oscillated at INO.21 Since near-resonant matter
effects appear only in either the neutrino or the antineutrino channel, this gives the experiment
an added edge to unravel both the mass hierarchy as well as 13 simultaneously, within a much
shorter time. The argument goes as follows.
Throughout the paper so far, we had been tacitly working in a scenario where we have just
one type of -beam from CERN and the hierarchy was conducive to having large matter effect,
the key behind the excellent 13 as well as mass hierarchy sensitivity. However, we have to bear
in mind that in the absence of any prior knowledge of the true value of 13 and/or the mass
hierarchy, one would not know which of the two modesneutrino or the antineutrinowould
be preferable for extracting maximal matter effects. With luck, it might so happen that the beam
is run in the neutrino mode first and the true hierarchy turns out to be normal and sin2 213 (true)
not too small. In that case, we would see large number of events in the detector and know without
ambiguity both the mass hierarchy as well as the value of 13 . However, if the experiment is run
in say only, the neutrino mode and if we do not see enough number of events (as expected from
large matter effects), it would be difficult to tell if the non-observation should be attributed to
the true hierarchy being inverted or whether the mixing angle 13 is very small. We would find
ourselves in a similar impasse if a e -beam is used and one observed a very small number of
events. This deadlock would then be broken only if we run the -beam in both the neutrino and
the antineutrino mode.
Recall that a -beam facility has the useful feature that it can be operated simultaneously in the
e as well as e mode, since both the 8 B and 8 Li ions are positively charged, and therefore move
in the same direction in the storage ring. The only constraint is that, since the magnetic field
would be the same for both channels, the different charge-to-mass ratio for the two candidate
ions would fix the boost for the antineutrino channel to be 0.6 times that in the neutrino channel.
Therefore, we could work within a set-up where we have simultaneous e and e beams with
say = 500 and 300, respectively for the two channels. Since INO-ICAL would have excellent
charge identification capability, it would distinguish the produced from oscillated and +
produced from oscillated , when the e and e beams are run simultaneously. In this scenario,
unless 13 is extremely small, we would observe an enhancement in the number of events due
to earth matter effects in one of the channels, irrespective of the true mass hierarchy. Therefore,
a direct comparison of the number of and + events would immediately tell us the mass
hierarchy for 13 not small. If the true value of 13 is indeed chosen to be very small by Nature,
then we would see no enhancement in the event rate in either the neutrino or the antineutrino
channel and one would be able to place an upper limit on the value of sin2 213 unambiguously.
We have checked that with e and e beams running simultaneously with = 500 and 300,
respectively for the two channels, we could obtain a 13 sensitivity of sin2 213 < 4.5 103 at
3 , irrespective of the true mass hierarchy, after 10 yr of running. We would have sensitivity to
the neutrino mass hierarchy for at least sin2 213 > 2.1 102 at 3 after 10 years of simultaneous running, again irrespective of whether the true hierarchy was normal or inverted. One
could argue that running the beam successively with = 500, first in the e mode followed by
the e mode would give an unambiguous 13 sensitivity limit of sin2 213 < 1.5 103 at 3
and the mass hierarchy sensitivity limit of sin2 213 > 8.2 103 at 3 . However, note that it
21 The and -beam fluxes could be produced in distinct bunches from the storage ring. Therefore, in principle,
e
e
it should be straight-forward to distinguish the neutrino from the antineutrino events at any detector using the nanosecond time resolution that most detectors possess. However, since ICAL will be magnetized, it could directly distinguish
between the two channels irrespective of the timing information.
25
is not possible to run the e beam with the same as the e beam without any accelerator upgrade. Running the -beam simultaneously in both polarities has the additional advantage that if
sin2 213 (true) is indeed large in Nature and within reach of the CERN-INO -beam set-up, then
one would have a surplus of either or + events in INO-ICAL. This would directly tell us
the neutrino mass hierarchy unambiguously.
9. Conclusions
We explored the physics potential of an experimental program with the -beam source (e or
e ) at CERN and a 50 kton magnetized iron detector at INO. The CERN-INO distance is very
close to the magic baseline, which is ideal for performing a degeneracy-free measurement of the
mixing angle 13 and the neutrino mass hierarchy. The large baseline also opens up the possibility
for large matter effects. For the CERN-INO baseline, the largest earth effects corresponding to
resonant conversion of (anti)neutrinos in matter, happen for (anti)neutrino energy of E 6 GeV.
or produced at these energies through oscillations can be effectively detected in magnetized
iron calorimeters via the they produce, the detection energy threshold being about 1 GeV.
For the most common ions used in -beam studies, this range of (anti)neutrino energies can
be achieved for Lorentz factor 103 . In this paper we have considered the new possibility
of using the 8 B and 8 Li as source ions for e and e -beam respectively. Owing to the large
end-point energy of these ions, the resultant energy spectrum of the Lorentz boosted -beam for
these ions is harder by more than a factor of 3, compared to the standard alternative ions 18 Ne and
6 He. Therefore, with these ions we would get a -beam flux peaked at about 6 GeV, for plausible
values of the Lorentz boost in the range 250500. We showed that for this -beam flux from
CERN, it will be possible to get an essentially background free measurement of the golden
channel probability Pe , through observation of (or + ) events in the ICAL detector at
INO. We argued that the number of events in this experimental set-up is dictated directly by the
extent of near-resonant earth matter effect, which significantly enhances the transition probability
Pe . The extent of earth matter effect in turn is governed by the neutrino mass hierarchy and the
value of sin2 213 . The CERN-INO -beam experiment that we propose therefore emerges as a
powerful tool to pin down the neutrino mass hierarchy and 13 .
We simulated the prospective data in this experimental set-up for 5 yr and 10 yr of running of
the -beam in either the neutrino or the antineutrino mode and presented the sensitivity results
from a rigorous 2 analysis, after marginalizing over all the oscillation parameters and taking
into account the systematic uncertainties coming from the -beam source, the detector, as well
as the theoretical calculation of the interaction cross sections. For = 500, if true value of 13
was zero, this experiment with 60% detection efficiency and 5 yr of running would put a limit
of sin2 213 < 2.1 103 (sin2 213 < 1.5 103 ) at 3 C.L. with the e ( e ) -beam for the
normal (inverted) mass hierarchy. With 80% detection efficiency and 10 yr of running the corresponding limits would be sin2 213 < 8.4 104 (sin2 213 < 5.7 104 ) at 3 C.L. with the
e ( e ) -beam for the normal (inverted) mass hierarchy. With 60% detection efficiency and 5 yr
of running the wrong hierarchy could be ruled out at the 3 C.L. if sin2 213 (true) > 1.0 102
(sin2 213 (true) > 8.2 103 ) with the e ( e ) -beam for the true normal (inverted) mass hierarchy. If the detection efficiency was 80% and running period 10 years the corresponding sensitivity
would be enhanced to sin2 213 (true) > 8.5 103 (sin2 213 (true) > 6.9 103 ) respectively at
the 3 C.L. We showed that if sin2 213 (true) = 0.05, = 500 and the true hierarchy was normal,
then the inverted hierarchy could be disfavored at 3 C.L. with a e -beam within just 9 months
of running. Finally, we considered the scenario where the e and e flux could be produced si-
26
multaneously at the CERN -beam facility. Since ICAL at INO would be magnetized and hence
would possess charge identification capability, we argued that an unambiguous information on
13 and the mass hierarchy could be obtained within a much shorter time scale if sin2 213 (true)
is large. If 13 is very small, one would not be able to determine the mass hierarchy, but could
still save time in putting an upper bound on the mixing angle 13 .
In conclusion, the sensitivity for sin2 213 and the neutrino mass hierarchy that is expected
in the CERN-INO -beam experiment appears to be better than most of the other experimental
proposals involving superbeams and -beam and can only be surpassed by the NuFact-II experiment.
Acknowledgements
The authors acknowledge help from the INO Collaboration and thank F. Terranova for a useful
communication. S.K.A. is grateful to Abhijit Samanta for discussions. S.C. wishes to thank Walter Winter, Andrea Donini and Pasquale Migliozzi for helpful discussions. S.C. acknowledges
support from the University of Oxford and PPARC during the initial part of this work.
References
[1] B.T. Cleveland, et al., Astrophys. J. 496 (1998) 505;
J.N. Abdurashitov, et al., SAGE Collaboration, J. Exp. Theor. Phys. 95 (2002) 181, Zh. Eksp. Teor. Fiz. 122 (2002)
211;
W. Hampel, et al., GALLEX Collaboration, Phys. Lett. B 447 (1999) 127;
S. Fukuda, et al., Super-Kamiokande Collaboration, Phys. Lett. B 539 (2002) 179;
B. Aharmim, et al., SNO Collaboration, Phys. Rev. C 72 (2005) 055502.
[2] Y. Ashie, et al., Super-Kamiokande Collaboration, Phys. Rev. D 71 (2005) 112005.
[3] K. Eguchi, et al., KamLAND Collaboration, Phys. Rev. Lett. 90 (2003) 021802;
T. Araki, et al., KamLAND Collaboration, Phys. Rev. Lett. 94 (2005) 081801.
[4] E. Aliu, et al., K2K Collaboration, Phys. Rev. Lett. 94 (2005) 081802.
[5] D.G. Michael, et al., MINOS Collaboration, hep-ex/0607088.
[6] M. Apollonio, et al., Eur. Phys. J. C 27 (2003) 331.
[7] M. Maltoni, et al., New J. Phys. 6 (2004) 122, hep-ph/0405172;
S. Choubey, hep-ph/0509217;
S. Goswami, Int. J. Mod. Phys. A 21 (2006) 1901;
A. Bandyopadhyay, et al., Phys. Lett. B 608 (2005) 115;
G.L. Fogli, et al., Prog. Part. Nucl. Phys. 57 (2006) 742.
[8] L. Wolfenstein, Phys. Rev. D 17 (1978) 2369.
[9] S.P. Mikheev, A.Y. Smirnov, Sov. J. Nucl. Phys. 42 (1985) 913, Yad. Fiz. 42 (1985) 1441;
S.P. Mikheev, A.Y. Smirnov, Nuovo Cimento C 9 (1986) 17.
[10] V.D. Barger, K. Whisnant, S. Pakvasa, R.J.N. Phillips, Phys. Rev. D 22 (1980) 2718.
[11] A. Cervera, et al., Nucl. Phys. B 579 (2000) 17;
A. Cervera, et al., Nucl. Phys. B 593 (2001) 731, Erratum.
[12] P. Zucchelli, Phys. Lett. B 532 (2002) 166.
[13] http://www.hep.ph.ic.ac.uk/iss/.
[14] M.S. Athar, et al., INO Collaboration, A Report of the INO Feasibility Study, http://www.imsc.res.in/~ino/
OpenReports/INOReport.pdf.
[15] J. Burguet-Castell, et al., Nucl. Phys. B 608 (2001) 301.
[16] H. Minakata, H. Nunokawa, JHEP 0110 (2001) 001.
[17] G.L. Fogli, E. Lisi, Phys. Rev. D 54 (1996) 3667.
[18] V. Barger, D. Marfatia, K. Whisnant, Phys. Rev. D 65 (2002) 073023.
[19] P. Huber, W. Winter, Phys. Rev. D 68 (2003) 037301.
[20] A.Y. Smirnov, hep-ph/0610198.
[21] P. Huber, M. Lindner, M. Rolinec, W. Winter, Phys. Rev. D 74 (2006) 073003.
27
[22] S.K. Agarwalla, A. Raychaudhuri, A. Samanta, Phys. Lett. B 629 (2005) 33.
[23] C. Rubbia, A. Ferrari, Y. Kadi, V. Vlachoudis, Nucl. Instrum. Methods A 568 (2006) 475;
C. Rubbia, hep-ph/0609235.
[24] A. Donini, E. Fernandez-Martinez, Phys. Lett. B 641 (2006) 432.
[25] R. Adhikari, S.K. Agarwalla, A. Raychaudhuri, Phys. Lett. B 642 (2006) 111;
S.K. Agarwalla, S. Rakshit, A. Raychaudhuri, hep-ph/0609252.
[26] For a recent review see C. Volpe, J. Phys. G 34 (2007) R1.
[27] J. Bouchez, M. Lindroos, M. Mezzetto, AIP Conf. Proc. 721 (2004) 37.
[28] B. Autin, et al., J. Phys. G 29 (2003) 1785.
[29] L.P. Ekstrom, R.B. Firestone, WWW Table of Radioactive Isotopes, database version 2/28/99 from URL
http://ie.lbl.gov/toi/.
[30] J. Burget-Castell, et al., Nucl. Phys. B 695 (2004) 217.
[31] A. Donini, et al., hep-ph/0511134;
A. Donini, et al., hep-ph/0604229.
[32] M. Freund, P. Huber, M. Lindner, Nucl. Phys. B 615 (2001) 331.
[33] W. Winter, Phys. Lett. B 613 (2005) 67.
[34] H. Minakata, H. Nunokawa, S.J. Parke, Phys. Rev. D 68 (2003) 013010;
P. Huber, M. Lindner, W. Winter, Nucl. Phys. B 654 (2003) 3;
O. Mena, S.J. Parke, Phys. Rev. D 70 (2004) 093011;
O. Mena, H. Nunokawa, S.J. Parke, hep-ph/0609011.
[35] M. Ishitsuka, et al., Phys. Rev. D 72 (2005) 033003.
[36] K. Hagiwara, N. Okamura, K.i. Senda, Phys. Lett. B 637 (2006) 266;
K. Hagiwara, N. Okamura, K.i. Senda, Phys. Lett. B 641 (2006) 486, Erratum;
K. Hagiwara, N. Okamura, K.-i. Senda, hep-ph/0607255.
[37] O. Mena Requejo, S. Palomares-Ruiz, S. Pascoli, Phys. Rev. D 72 (2005) 053002;
O. Mena, S. Palomares-Ruiz, S. Pascoli, Phys. Rev. D 73 (2006) 073007.
[38] H. Minakata, H. Nunokawa, Phys. Lett. B 413 (1997) 369;
J. Burguet-Castell, et al., Nucl. Phys. B 608 (2001) 301;
V. Barger, D. Marfatia, K. Whisnant, Phys. Lett. B 560 (2003) 75;
V. Barger, D. Marfatia, K. Whisnant, Phys. Rev. D 66 (2002) 053007.
[39] D. Autiero, et al., Eur. Phys. J. C 33 (2004) 243;
A. Donini, D. Meloni, P. Migliozzi, Nucl. Phys. B 646 (2002) 321.
[40] A. Bueno, M. Campanelli, S. Navas-Concha, A. Rubbia, Nucl. Phys. B 631 (2002) 239.
[41] P. Huber, M. Maltoni, T. Schwetz, Phys. Rev. D 71 (2005) 053006.
[42] J.E. Campagne, M. Maltoni, M. Mezzetto, T. Schwetz, hep-ph/0603172.
[43] P. Huber, M. Lindner, T. Schwetz, W. Winter, Nucl. Phys. B 665 (2003) 487.
[44] A.M. Dziewonski, D.L. Anderson, Phys. Earth Planet. Interiors 25 (1981) 297;
S.V. Panasyuk, Reference Earth Model (REM) webpage, http://cfauves5.harvrd.edu/lana/rem/index.html.
[45] A. Takamura, K. Kimura, JHEP 0601 (2006) 053.
[46] P. Huber, M. Lindner, W. Winter, Comput. Phys. Commun. 167 (2005) 195.
[47] M.D. Messier, PhD thesis, UMI-99-23965.
[48] E.A. Paschos, J.Y. Yu, Phys. Rev. D 65 (2002) 033002.
[49] M.C. Banuls, G. Barenboim, J. Bernabeu, Phys. Lett. B 513 (2001) 391.
[50] R. Gandhi, P. Ghoshal, S. Goswami, P. Mehta, S. Uma Sankar, Phys. Rev. Lett. 94 (2005) 051801.
[51] R. Gandhi, P. Ghoshal, S. Goswami, P. Mehta, S. Uma Sankar, Phys. Rev. D 73 (2006) 053001.
[52] See for example A. Bandyopadhyay, et al., Phys. Rev. D 72 (2005) 033013;
J.N. Bahcall, C. Pena-Garay, JHEP 0311 (2003) 004.
[53] Y. Itow, et al., hep-ex/0106019.
[54] D.S. Ayres, et al., NOvA Collaboration, hep-ex/0503053.
[55] P. Huber, et al., Phys. Rev. D 70 (2004) 073014.
[56] P. Huber, M. Lindner, M. Rolinec, W. Winter, Phys. Rev. D 73 (2006) 053002.
[57] J. Burguet-Castell, D. Casper, E. Couce, J.J. Gomez-Cadenas, P. Hernandez, Nucl. Phys. B 725 (2005) 306.
[58] P. Huber, M. Lindner, W. Winter, Nucl. Phys. B 645 (2002) 3.
CP violation in neutrino mass matrix

Utpal Sarkar , Santosh K. Singh
Physical Research Laboratory, Ahmedabad 380009, India
Received 3 August 2006; received in revised form 8 February 2007; accepted 8 February 2007
Abstract
We constructed rephasing invariant measures of CP violation with elements of the neutrino mass matrix,
in the basis in which the charged lepton mass matrix is diagonal. We discuss some examples of neutrino mass matrices with texture zeroes, where the present approach is applicable and demonstrate how
it simplifies an analysis of CP violation. We applied our approach to study CP violation in all the phenomenologically acceptable 3-generation two-zero texture neutrino mass matrices and shown that in any of
these cases there is only one CP phase which contributes to the neutrino oscillation experiment and there
are no Majorana phases.
1. Introduction
In the Standard Model there is only one source of CP violation, which is in the charged
current mixing matrix in the quark sector. The chargedcurrent mixing matrix in the quark sector
contains one CP phase, which has been observed. It is not possible to identify the position of the
CP phase, since it is possible to make any phase transformations to the quarks. However, it is
possible to define a rephasing invariant quantity as product of elements of the mixing matrix that
remains invariant under any rephasing of the quarks [1,2]. This is known as Jarlskog invariant.
In the leptonic sector, standard model does not allow any CP violation. If one considers extensions of the Standard Model to accommodate the observed neutrino masses, then there can
be several CP phases [36]. In the simplest scenario of three generations, there could be one
CP phase in the mixing matrix in the leptonic sector, similar to the quark sector. In addition, if
neutrinos are Majorana particles they can have two more Majorana CP phases [4]. In this case
E-mail address: utpal.sarkar@cern.ch (U. Sarkar).

doi:10.1016/j.nuclphysb.2007.02.009
U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839
29
it is possible to work in a parametrization, in which all the three CP phases could be in the
chargedcurrent mixing matrix in the leptonic sector. One of these CP phase will contribute to
the neutrino oscillation experiments, while the other two will contribute to lepton number violating process like neutrinoless double beta decay. A natural explanation for the smallness of
the neutrino masses comes from the see-saw mechanism [7]. The origin of small neutrino mass
then relates to a large lepton number violating scale. It is quite natural that this lepton number
violation at the high scale would also explain the baryon asymmetry of the universe through leptogenesis [8,9]. This connection between the neutrino mass and leptogenesis makes the question
of CP violation in the leptonic sector more interesting [5,6].
The CP phases in the leptonic sector has been studied and rephasing invariants for both lepton
number conserving as well as lepton number violating processes have been constructed [3]. In
this article we try to study this question only in terms of neutrino masses. Since neutrinos are
produced only through weak interactions, it is possible to work in the weak interaction basis, in
which the charged lepton mass matrix is diagonal. The neutrino mass matrix in this basis will then
contain all the information about CP violation. We try to find rephasing invariant combinations
of the neutrino mass elements, so that with those invariants some general comments can be made
about CP violation in the model without deriving the structure of the chargedcurrent mixing
matrix.
2. CP violation in the quark sector
We briefly review the rephasing invariants in terms of the mixing matrices and then show how
the same results can be obtained from the mass matrix without taking the trouble of diagonalizing
them in the leptonic sector. Consider first the quark sector, where the up and the down quark
mass matrices are diagonalized by the bi-unitary transformations. We write the corresponding
unitary matrices that relates the left-handed and right-handed physical (with definite masses) up
and down quarks fields to their weak (diagonal charged current) basis as: UL , DL , UR and DR .
Then the charged current interactions in terms of the physical fields will contain the Kobayashi
CabibboMaskawa mixing matrix
V = DL UL .
Since the right-handed fields are singlets under the standard model interactions, they do not enter
in the charged current interactions. In any physical processes, only this CKM mixing matrix
would appear and hence the matrices UR and DR becomes redundant. So, the up and down
quark masses have much more freedom and the physical observables that can determine the Vi
cannot infer about the up and down quark masses uniquely.
For the CP violation, one needs to further consider the rephasing of the left-handed fields.
Any phase transformation to the up and down quarks will also transform the CKM matrix
Vi ei(d ui ) .
However, if there is any CP phase in the CKM matrix, which cannot be removed by any phase
transformations of the up and the down quarks, should be present in the Jarlskog invariant [1,2]

Jij = Im Vi Vj Vj
(1)
Vi .
Thus if the Jarlskog invariant is non-vanishing, that would imply CP violation in the quark mixing. It is apparent from the definition that any phase transformations to the up and down quarks
30
cannot change Jij . In a three generation scenario there can be only one such invariant and
hence the CKM matrix can have only one CP phase, which is invariant under rephasing of the up
and the down quarks.
3. CP violation in the leptonic sector
In the leptonic sector, the charged lepton mass matrix can in general be diagonalized by a biunitary transformation. In addition, the neutrinos are produced in weak interactions, so the flavour
of the neutrinos at the time of production is always same as the flavour of the charged lepton.
The chargedcurrent interaction is given by
g
Lcc = iL iL W
(2)
2
in the basis (iL , i = e, , ) in which charged lepton mass matrix is diagonal, i.e., the states e,
, correspond to physical states. Without loss of generality we further assume that elements of
the diagonal charged lepton mass matrix are real and positive.
The neutrino mass term can be written as
T 1
C j L ,
LM = Mij c iR j L = Mij iL
(3)
so that the neutrino mass matrix can be diagonalized by a single unitary matrix U through
U T M U = K 2 M ,
(4)
where M = diag[m1 , m2 , m3 ] is a real diagonal matrix and K is a diagonal phase matrix. The
unitary matrix Uia (with i, j = e, , and a, b = 1, 2, 3) relates the physical neutrino states a
(with masses ma ) to the weak states
2
a = Uai
i + Kaa
Uai ic ,
(5)
so that the physical neutrinos satisfy the Majorana condition

= K 2c.
(6)
The unitary matrix U thus gives the mixing of the neutrinos and hence neutrino oscillations,
which is known as the PMNS mixing matrix [10] and K is the Majorana phase matrix containing
the Majorana phases, which are the new sources of CP violation entering due to the Majorana
nature of the neutrinos. The PMNS matrix U also can contain CP violating phases, which should
be observed in the neutrino oscillation experiments. We call these phases in the PMNS mixing
matrix U as Dirac phases to distinguish them from the Majorana phases. The main difference
between a Majorana phase and a Dirac phase is that the Majorana phases do not affect any lepton
number conserving process like neutrino oscillations. On the other hand, the Dirac phases may
contribute to both lepton number conserving as well as lepton number violating processes.
From the above discussions it is apparent that the information about the CP phases can be
obtained from either U and K or only from the mass matrix M . In the literature the question
of CP violation is usually discussed by studying U and K. In this article we point out that it
is possible to study the question of CP phases only by studying the neutrino mass matrix M .
In particular, the information about CP violation is conveniently obtained from the rephasing
invariant combinations of neutrino mass elements. When the neutrino masses originate from seesaw mechanism, the question of CP violation has been studied in details and similar invariants
have been constructed [6,11]. The CP violation entering in leptogenesis has been studied in [6]
31
and an interesting connection between the unremovable phases and complex rephasing invariants
has been addressed in [11] for the see-saw model of neutrino masses. Our approach is different
in the sense that we are working with only the effective low energy neutrino mass matrix without
restricting the analysis to any specific origin of the neutrino masses. Our results are general and
applicable to any models of neutrino masses.
Consider the transformation of different quantities under the rephasing of the neutrinos
a eia a ,
i eii i ,
Uia ei(i a ) Uia ,
Ka eia Ka .
(7)
From these transformations it is possible to construct the rephasing invariants [3]

Ka Kb .
siab = Uia Uib
(8)
In the three generation case there will be three independent rephasing invariant measures. There
is another rephasing invariant which is similar to the Jarlskog invariant in the quark sector,

Uj a ,
tiaj b = Uia Uj b Uib
(9)
so that Tiaj b = Im tiaj b and Siab = Im siab becomes the independent measures of CP violation.
Tiaj b contains the information about the Dirac phase, while siab contains information about both
Dirac as well as Majorana phases.
In general, there should be at least one measure corresponding to any CP phase although the
reverse may not be true. However, it is possible to choose a set of independent measures, all of
which correspond to one CP phase in the mass matrix. Thus we shall try to restrict our discussions
to only the set of independent measures so that by constructing this set of independent measures
one can predict the numbers of CP phases in the model. In the present one can then use the
relation
tiaj b = siab siba
to eliminate the invariants T s or else keep the T s as independent measures and reduce the
number of independent Ss. One convenient choice for the independent measures is the independent tiaj b s and s1ab s. In the three generation case there is only one tiaj b and two s1ab s.
The advantage of this parametrization is that the measure Tiaj b gives the CP violation in any
neutrino oscillation experiment, while the measures S1ab corresponds to CP violation in lepton
number violating interactions like the neutrinoless double beta decay or scattering processes like
W + W
i + j also.
4. Rephasing invariants with neutrino masses
We shall now proceed to construct such measures of CP violation in terms of the mass matrix
itself. The rephasing invariant measures with the mixing matrix can allow all the rephasing invariants non-vanishing even when there is only one Dirac phase. However, in the present formalism,
the number of rephasing invariants is same as the number of CP phases. So, we can find out if
there is any Majorana phase or not. Since the neutrino mass matrix is diagonalized by a single
unitary matrix, the mass matrix contains all the information about the PMNS mixing matrix and
32
also the mass eigenstates. However, this is not obvious with the CP phase. When the neutrinos
are given a phase transformation, the mass matrix will be transformed the same way. Since the
we are working in the weak basis, any transformation to the charged leptons can be transformed
to the mixing matrix and in turn to the neutrino masses. Thus the phase transformation to the
mass matrix will become
i eii i ,
i eii i ,
Mij ei(i +j i j ) Mij
ei(i +j ) Mij ,
(10)
where i = i i .
Consider the transformation E XE, where X is the phase transformation to the charged
leptons. The mixing matrix will transform as U X U . However, in Eq. (4) this transformation
can be interpreted as a transformation to the mass matrix, M X MX . Thus any rephasing invariant measure constructed with only the mass matrix will contain the information about
CP violation.
Unlike the mixing matrices, the mass matrix is not unitary and instead it is symmetric. We
write the elements of the mass matrix M as mij and try to construct the rephasing invariants in
terms of mij . This analysis do not depend on the origin of neutrino masses. We work with the
neutrino mass matrix after integrating out any heavier degrees of freedom and in the weak basis.
Any quadratic terms that can be constructed from the elements of the neutrino mass matrix are
all real, mij mij = |mij |2 , as expected. Let us next consider the quartic terms
Iij kl = mij mkl mil mkj .
(11)
It is easy to check that any three factors of the above quartic invariant can be made real by appropriate rephasing, but fourth one will remain complex. Since there are n rephasing phases (i ),
one can get n number of linear equations to make mass elements of the mass matrix to be real.
So n number of entries (excluding symmetric elements) of the mass matrix can be made real, but
positions of the mass entries cannot be chosen randomly. That is the reason why all the above
rephasing quartic invariants cannot be made real in general. An n n symmetric matrix has
n(n + 1)/2 independent entries and so it has the same number of phases. By appropriate rephasing, as argued above, n independent phases can be removed. Then, one is left with n(n 1)/2
number of independent phases.
To find out the minimal set of rephasing invariants we list some of the transitive and conjugation properties of the invariants:
Iijpl Ipj kl = |mpj mpl |2 Iij kl ,
Iij kp Iipkl = |mip mkp |2 Iij kl
and
= Iklij = Ikj
Iij kl = Iilkj
il .
(12)
Using these relations it can be shown that all the Iij kl are not independent and they can be
expressed in terms of a subset of these invariants Iij and the quadratic invariants as
Iij kl =
I
Iij Ikl Ili
kj
|m |4 |mi mj mk ml |2
(13)
33
where i, j = and = 1, 2, . . . , n, where n is the number of generations. On the other hand, any
quartics of the form Iij can be expressed in terms of I as
Im[Iij ] = Im[Iij ] =
Im[Iii Ijj Iiijj ]

.
Re[Iij ](|mii |2 |mjj |2 )
(14)
Thus we can express all other invariants in terms of Iiijj and hence consider them to be of
fundamental importance. However, when there are texture zeroes in the neutrino mass matrix,
some or all of these invariants Iiijj could be vanishing. In that case, it is convenient to use
the Iij as the measure of CP violation. For the present we shall concentrate on the more
general case with neutrino mass matrices without any texture zeroes, when the simplest rephasing
invariants are Iiijj .
We can thus define the independent CP violating measures as

Iij = Im[Iiijj ] = Im mii mjj mij mj i (i < j ).
(15)
These are the minimal set of CP violating measures one can construct and this gives the independent CP violating quantities. Since Iij satisfies
Iij = Ij i
and
Iii = 0,
there are n(n 1)/2 independent measures for n generations.

We elaborate with some examples starting with a 2-generation scenario. There are three Iij kl ,
two of which are real: I1211 = |m11 m12 |2 ; and I1222 = |m12 m22 |2 . The third one can have imaginary phase, which is I12 = Im[I1122 ] = Im[m11 m22 m12 m21 ]. In the 3-generation case there are
thus three independent measures I12 , I13 , I23 . Imaginary phases in all other quartics Iij kl are
related to only these three independent measures. For example,
2
I1223
=
I I
I12
23 13
.
|m11 |2 |m33 |2
Similarly, for 4-generations there will be six rephasing invariant independent phases, which
are I12 , I23 , I31 , I14 , I24 , I34 .
The above arguments have been stated without considering any texture zeroes in the mass
matrix. If any element of the mass matrix is zero, then these discussions have to be generalized.
It is because some quartic invariants can become undefined because of vanishing denominator of
the right-hand side of the expression (13) and (14). In that case one needs to consider all possible
invariants Iij kl , which could be non-vanishing. In addition, even if all the quartic invariants
vanish, the product of six mass matrix elements of the form
Iij klpq = mij mkl mpq mil mkq mpj
could be non-vanishing and can contribute to CP violation. When there are no texture zeroes,
the product of six mass elements do not contain any new information about CP phases, they are
related to the quartic invariants
Iij klpq =
Iij kl Ipqkj
.
|mkj |2
(16)
Other products of six mass elements are of the form, mij mkl mpq mil mkj mpq = |mpq |2 Iij kl or
|mij mkl mpq |2 .
34
We summarize this section by restating that when all elements of the neutrino mass matrix are
non-vanishing, Iij (i < j ) gives the total number of Dirac and Majorana phases. If some of the
elements of the mass matrix vanishes, then either Iij kl or Iij klpq could also represent some of
the independent phases.
5. CP violation in lepton number conserving processes
The rephasing invariant independent phases contained in Iij , i < j , are inclusive of the Dirac
phases as well as the Majorana phases. We shall now identify the rephasing invariant measures,
which is independent of the Majorana phases, which would enter in the neutrino oscillation
experiments. The mass matrix (M ) in terms of the diagonal mass matrix (M ) can be expressed
following Eq. (4) as
M = U K 2 M U .
Thus the products

M = M M = M M = U M 2 U
(17)
are independent of the Majorana phases K and any rephasing invariant measure constructed with
elements m
ij of M will contain only the Dirac phases and hence should contribute to any lepton
number conserving processes.
The mass-squared elements m
ij transforms under rephasing of the neutrinos and charged
leptons as
m
ij ei(i j ) m
ij .
(18)
Since the mass-squared matrix M is Hermitian, M = M , the mass elements satisfy

m
ij = m
j i .
(19)
Thus the simplest rephasing invariant that can be constructed from the mass-squared matrix M
is just m
11 . However, from Eq. (19) it is obvious that this is a real quantity. The next possible
rephasing invariant would be a quadratic term, but even that is also real
m
ij m
ji = m
ij m
ij = |m
ij |2 .
Thus the simplest rephasing invariant combination that can contain the complex CP phase is of
the form
Jij k = m
ij m
jkm
ki
(i = j = k).
(20)
Im[Jij k ] are antisymmetric under interchange of any two indices and hence vanishes when any
two of the indices are same. We can express Jij k in terms of M matrix elements as,

ij m
jkm
ki =
mi mj
mj mk
mk mi ,
Jij k = m
(21)

where mi mj can be interpreted as scalar product of ith and j th row. A similar invariant
was constructed in the case of see-saw model of neutrino masses in Refs. [6,11], although the
approaches to the problem are completely different. In this expression, if any one scalar product
35
vanishes then number of independent rephasing measures Im[Jij k ] which are independent of the
Majorana phases will be reduced by one.
It is possible to express all the rephasing invariants containing the Dirac phases Jij k in terms
of a minimal set of (n 1)(n 2)/2 invariants Jij n (i < j < n) as
Jij k =
Jij n Jj kn Jkin
|min ||mj n ||mkn |
(22)
where n is the index corresponding to the number of generations. Thus we define the measures
of CP violation in lepton number conserving processes as
Jij n = Im[Jij n ]
(i < j < n).
(23)
These invariants Im[Jij k ] are not independent of the invariants Iij kl and can be expressed as
Jij k =
Iiaj b Ikaic
.
|mia |2
(24)
a,b,c
So, the independent measures Iij include these independent measures of Dirac CP phases
Im[Jij n ] (i < j < n).
There are n(n 1)/2 phases present in M for n generations, but all of them are not independent. (n 1) of these phases can be removed by redefining the phases of the leptons. That
This is the number of Dirac
n = (n1)(n2)
=(n1) C2 independent phases in M.
leaves n(n1)
2
2
phases and may be observed in neutrino oscillation experiments. Let us assume that some particular n 1 entries are made real by appropriate rephasing. We can take all possible pair-product
of these real entries. To have non-real rephasing invariant Jij k , one will have to multiply pairproduct with some complex entry. For each real pair-product there correspond only one complex
entry so that their product is a complex rephasing invariant defined as in Eq. (20). So number of
all possible pair of real entries will give the number of non-vanishing rephasing measures independent of Majorana phases which is (n1) C2 = (n 1)(n 2)/2. This number is same as the
number of physical phases present in M as it has been analyzed earlier.
In the 2-generation case there is only one CP phase which is a Majorana phase. Which implies
there should not be any non-vanishing Jij k , which is trivial to check. In the 3-generation case
there is only one Dirac CP measure, which is

ma1 ma2 mb2 mb3 mc3 mc1 .
12 m
23 m
31 =
J123 = m
(25)
a,b,c
Thus given a neutrino mass matrix one can readily say if this mass matrix will imply CP violation
in the neutrino oscillation experiments.
In the 4-generation case there are three CP phases in the PMNS mixing matrix and three
Majorana phases. The independent rephasing invariants of Dirac phases will be given as J124 ,
J134 and J234 . One dependent rephasing invariant is J123 which can be expressed as
J123 =
J124 J234 J134

.
|m
14 m
24 m
34 |2
In general, these invariants satisfy
Jij k Jikl
Jilj = |m
ij m
ik m
il |2 Jj kl
for n generations, where i, j, k = 1, 2, . . . , n.
(26)
36
We summarize this section by restating for n-generation neutrino mass matrix without any
texture zeroes, the rephasing invariants corresponding to the Dirac phase are Jij n (i < j < n). If
there are texture zeroes, then some of the Jij k (i < j < k, k = n) could also be independent.
6. Texture zeroes
In case neutrino mass matrix contains zero entries in all the columns, it is convenient to define
basic independent quartic invariants to the rephasing of charged leptons in slightly different form
as,
Rij nn =
lim
|min |,|mj n |,|mnn |0
mij min mj n mnn

|min ||mj n ||mnn |
(i j and i, j = n).
(27)
Limit has to be taken for all nth column elements. Any other such quartic invariant can be expressed in terms of these independent rephasing invariants Rij nn as,
Rij kl = Rij nn Rklnn Rlinn Rkj nn .
(28)
Advantage of defining the independent rephasing (to the rephasing of charged lepton) invariants Rij nn as the limiting case is that the expressions do not become undefined due to presence
of vanishing denominators. Let us write the invariants in a different form as,
Rij nn = |mij |ei(ij +nn in j n )
(i, j = n and i j ),
where kn is the phase present at (k, n) entry of the mass matrix. If there are some zero entries present in nth column, then the corresponding phases present in expression of Rij nn must
be unphysical. So a subset of the set of these basic independent invariants will be having the
unphysical phases associated with the zero entries of the nth column. We cannot define CP measures corresponding to the invariants having unphysical phases by extracting the imaginary part
of the invariants, although we can define CP measure as usual for rest of the invariants. All the
invariants defined above are invariants corresponding to the rephasing of charged leptons but
not rephasing of the unphysical phase corresponding to zero entries present in the nth column.
We will have to construct the full rephasing invariants to all unphysical phases in terms above
invariants. This can be done by multiplying two or more invariants having unphysical phases
in such a way that all of the unphysical phases cancel out. It turns out that eliminating one of
these unphysical phases corresponding to the zero entries in nth column reduces the number of
full rephasing invariants by one. Also, one independent full rephasing invariant (and so one CP
measure) vanishes corresponding to the zero entry present in other than nth column (or nth row).
Thus the number of independent CP measures NCP defined as the imaginary part of these full
rephasing invariants for neutrino mass matrix having p zero entries and q zero rows (all the row
entries are vanishing) for n generations is given by
n(n 1)
p + q.
(29)
2
This is equal to the number of physical phases present in the matrix which cannot be removed.
In the same way we can study the mass-squared matrices M and write down the number of
rephasing invariant measures independent of Majorana phases N CP is given as
NCP =
(n 1)(n 2)
N CP =
r + s,
2
37
where r is the number of zero entries in M and s is the number of those rows whose all the
entries excluding diagonal one are zero. It should be noticed that above relation of N CP is only
valid if NCP is not zero.
7. Application to texture two-zero mass matrices
With our present formalism, we shall now study a class of 3-generation neutrino mass matrices
with two-zero textures, which has been listed in Ref. [12]. There are seven such mass matrices
that are consistent with present information about neutrino masses:

0 0 X
0 X 0
(2 3)
A2 :
A1 :
0 X X ;
X X X ;
X X X
0 X X

X X 0
X 0 X
(2 3)
B2 :
B1 :
X 0 X ;
0 X X ;
0 X X
X X 0

X 0 X
X X 0
B3 :
(2 3)
B4 :
0 0 X ;
X X X ;
X X X
0 X 0

X X X
C:
X 0 X .
X X 0
From our discussions in the previous section, there can be only one CP phase in all these cases.
We shall now identify the rephasing invariants in all the cases. Although all these matrices differ
in phenomenology, as far as CP violation is concerned, the interchange of the indices (2 3)
will not change any discussion. So, we shall not explicitly discuss the models A2 , B2 , B4 , which
can be obtained by changing the indices (2 3) from the matrices A1 , B1 , B3 respectively.
7.1. Case A1
There is only one non-vanishing Iij , which is I23 . The lepton number conserving rephasing
invariant measure J123 is given by

[J123 ] = Im m31 m32 m22 m23 + m32 m33 m33 m31 = |m31 |2 I23 .
(30)
Thus there is only one Dirac CP phase in this case, which will contribute to the lepton number
conserving processes. The same result is valid for A2 .
7.2. Case B1
In this case all the measures Iij are vanishing. Even the invariants of the form Iij are all
vanishing. However, there is one CP phase as discussed in the previous section. The invariant
I122133 is non-vanishing, which cannot be related to the lower invariants by I122133 = I1221
I3322 /|m22 |2 , since m22 = 0. The lepton number conserving invariant is related to this invariant
by

[J123 ] = Im m11 m12 m32 m33 m23 m21 = Im[I122133 ].
(31)
Again there are no Majorana CP phase. The analysis is same for the case B2 .
38
7.3. Case B3
There is only one non-vanishing CP violating measure I13 , which is related to the lepton
number conserving measure by

[J123 ] = Im m31 m32 m32 m33 m13 m11 + m33 m31 = |m32 |2 I13 .
(32)
There are no more CP phase left in addition to the one entering in lepton number conserving
processes. Replacing the indices (2 3) we get for the case B4 a similar relation [J123 ] =
|m32 |2 I12 .
7.4. Case C
This is the most interesting case. There are no CP violating measures of the form Iij , although
the invariant I1123 is non-vanishing. So, there is one CP phase in this case, as expected. This is
related to the CP violating measure that affects lepton number conserving processes by

J123 = Im m11 m12 + m31 m32 m12 m13 m13 m11 + m23 m21
.
= |m12 |2 I1123 + |m13 |2 I1123
Although this shows that the phase is a Dirac phase, in the special case of m12 = m13 , there will
not be any CP violation in the neutrino oscillation experiments. This can be verified from the fact
that for m12 = m13 the third mixing angle and hence U13 vanishes. In this case the CP violation
can originate from a Majorana phase, since J123 vanishes even when Im[I1123 ] is non-vanishing.
Another way to understand this is to write the mass matrix in a different basis. When
m12 = m13 , we can write the mass matrix C as

X X 0
X X 0 .
0 0 X
In this case the third generation decouples from the rest and we know that for two generation
there is only a Majorana phase, which corresponds to non-vanishing I12 and there is no Dirac
phase, as we stated above. This is the only example of two-zero texture mass matrices where the
CP violating phase could be a Majorana phase, but this mass matrix is not allowed phenomenologically.
Thus there are no phenomenologically acceptable two-zero texture neutrino mass matrices,
which has any Majorana phase. The only CP phase possible in any two-zero texture 3-generation
mass matrix is of Dirac type and should allow CP violation in neutrino oscillation experiments.
8. Summary
In summary, we constructed rephasing invariant measures of CP violation with elements of
the neutrino masses in the weak basis. For an n-generation scenario, in the absence of any texture
zeroes there are n(n 1)/2 independent measures of CP violation, given by

Iij = Im mii mjj mij mj i (i < j )
which corresponds to n(n1)/2 independent CP violating phases. Only (n1)(n2)/2 of these
phases of CP violation can contribute to the neutrino oscillation experiments and are independent
39
of the Majorana phases for which the rephasing invariant measures of CP violation can be defined
as

(i < j < n).
Im mia mj a mj b mnb mnc mic
Jij n =
a,b,c
We then defined invariants for mass matrices with texture zeroes and elaborated with some examples. We studied all the phenomenologically acceptable 3-generation two-zero texture neutrino
mass matrices. We show that there are no Majorana phases in any of the allowed cases.
References
[1] C. Jarlskog, Phys. Rev. Lett. 55 (1985) 1039;
C. Jarlskog, Z. Phys. C 29 (1985) 491.
[2] O.W. Greenberg, Phys. Rev. D 32 (1985) 1841;
D. Wu, Phys. Rev. D 33 (1986) 860;
I. Dunietz, O.W. Greenberg, D. Wu, Phys. Rev. Lett. 55 (1985) 2935.
[3] J.F. Nieves, P.B. Pal, Phys. Rev. D 36 (1987) 315.
[4] B. Kayser, Phys. Rev. D 30 (1984) 1023.
[5] J.R. Ellis, M. Raidal, Nucl. Phys. B 643 (2002) 229.
[6] G.C. Branco, T. Morozumi, B.M. Nobre, M.N. Rebelo, Nucl. Phys. B 617 (2001) 475;
G.C. Branco, R. Gonzalez Felipe, F.R. Joaquim, I. Masina, M.N. Rebelo, C.A. Savoy, Phys. Rev. D 67 (2003)
073025.
[7] P. Minkowski, Phys. Lett. B 67 (1977) 421;
M. Gell-Mann, P. Ramond, R. Slansky, in: P. van Niewenhuizen, D. Freedman (Eds.), Supergravity, North-Holland,
Amsterdam, 1979;
T. Yanagida, in: O. Sawada, A. Sugamoto (Eds.), Workshop on Unified Theory and Baryon Number in the Universe,
KEK, Japan, 1979;
R.N. Mohapatra, G. Senjanovic, Phys. Rev. Lett. 44 (1980) 912;
J. Schechter, J.W.F. Valle, Phys. Rev. D 22 (1980) 2227.
[8] M. Fukugita, T. Yanagida, Phys. Lett. B 174 (1986) 45.
[9] M.A. Luty, Phys. Rev. D 45 (1992) 455;
R.N. Mohapatra, X. Zhang, Phys. Rev. D 46 (1992) 5331;
A. Acker, H. Kikuchi, E. Ma, U. Sarkar, Phys. Rev. D 48 (1993) 5006;
M. Flanz, E.A. Paschos, U. Sarkar, Phys. Lett. B 345 (1995) 248;
M. Flanz, E.A. Paschos, U. Sarkar, J. Weiss, Phys. Lett. B 389 (1996) 693;
E. Ma, U. Sarkar, Phys. Rev. Lett. 80 (1998) 5716;
T. Hambye, E. Ma, U. Sarkar, Nucl. Phys. B 602 (2001) 23;
G.F. Giudice, A. Notari, M. Raidal, A. Riotto, A. Strumia, Nucl. Phys. B 685 (2004) 89;
W. Buchmuller, P. Di Bari, M. Plumacher, Nucl. Phys. B 665 (2003) 445;
W. Buchmuller, P. Di Bari, M. Plumacher, Ann. Phys. 315 (2005) 305;
W. Buchmuller, R.D. Peccei, T. Yanagida, Annu. Rev. Nucl. Part. Sci. 55 (2005) 311.
[10] Z. Maki, M. Nakagawa, S. Sakata, Prog. Theor. Phys. 28 (1962) 870;
B. Pontecorvo, Sov. Phys. JETP 7 (1958) 172, Zh. Eksp. Teor. Fiz. 34 (1957) 247;
B. Pontecorvo, Sov. Phys. JETP 6 (1957) 429, Zh. Eksp. Teor. Fiz. 33 (1957) 549.
[11] A. Kusenko, R. Shrock, Phys. Lett. B 323 (1994) 18, hep-ph/9403315.
[12] P.H. Frampton, S.L. Glashow, D. Marfatia, Phys. Lett. B 536 (2002) 79;
W. Grimus, A.S. Joshipura, L. Lavoura, M. Tanimoto, Eur. Phys. J. C 36 (2004) 227;
W. Grimus, PoS (HEP 2005) 186, hep-ph/0511078.
N = 4 topological amplitudes and string effective action

I. Antoniadis a,1 , S. Hohenegger a, , K.S. Narain b
a Department of Physics, CERNTheory Division, CH-1211 Geneva 23, Switzerland
b High Energy Section, The Abdus Salam International Center for Theoretical Physics,
Strada Costiera, 11-34014 Trieste, Italy

Received 21 December 2006; accepted 12 February 2007
Abstract
Certain scattering amplitudes in the gravitational sector of type II string theory on K3 T 2 are found
to be computed by correlation functions of the N = 4 topological string. This analysis extends the already
known results for K3 by Berkovits and Vafa, which correspond to six-dimensional terms in the effective action, involving four Riemann tensors and 4g 4 graviphotons, R 4 T 4g4 , at genus g. We find two additional
classes of topological amplitudes that use the full internal SCFT of K3 T 2 . One of these string amplitudes
is mapped to a 1-loop contribution in the heterotic theory, and is studied explicitly. It corresponds to the
four-dimensional term R 2 (dd)2 T 2g4 , with a KaluzaKlein graviscalar from T 2 . Finally, the generalization of the harmonicity relation for its moduli dependent coupling coefficient is obtained and shown to
contain an anomaly, generalizing the holomorphic anomaly of the N = 2 topological partition function Fg .
1. Introduction
It is now known that certain physical superstring amplitudes corresponding to BPS-type couplings can be computed within a topological theory, obtained by twisting the underlying N = 2
superconformal field theory (SCFT) that describes the internal compactification space [13]. The
better studied example is the topological CalabiYau (CY) -model describing N = 2 supersymmetric compactifications of type II string in four dimensions. Its partition function computes a
E-mail addresses: ignatios.antoniadis@cern.ch (I. Antoniadis), stefan.hohenegger@cern.ch (S. Hohenegger),

narain@ictp.trieste.it (K.S. Narain).
1 On leave from CPHT (UMR CNRS 7644) Ecole Polytechnique, F-91128 Palaiseau, France.
doi:10.1016/j.nuclphysb.2007.02.011
I. Antoniadis et al. / Nuclear Physics B 771 (2007) 4092
41
series of higher derivative F-terms of the form W 2g , where W is the N = 2 Weyl superfield;
its lowest component is the self-dual graviphoton field strength T+ while its next upper bosonic
component is the self-dual Riemann tensor, so that W 2g generates in particular the amplitude
R 2 T 2g2 , receiving contributions only from the g-loop order [2]. The corresponding coupling
coefficients Fg are functions of the moduli that belong to N = 2 vector multiplets. Although
navely these functions should be analytic, there is a holomorphic anomaly expressed by a differential equation providing a recursion relation for the non-analytic part [35].
The topological partition function has several physical applications. It provides the corrections
to the entropy formula of N = 2 black holes [69]. Moreover, by applying the N = 1 involution
that introduces boundaries along the a-cycles, one obtains a similar series of higher dimensional
N = 1 F-terms involving powers of the gauge superfield W [1012]. In particular, (Tr W 2 )2
gives rise to gaugino masses upon turning on expectation values for the D-auxiliary component,
generated for instance by internal magnetic fields, or equivalently by D-branes at angles [12,13].
The N = 2 W 2g F-terms appear also in the heterotic string compactified in four dimensions
on K3 T 2 . Under string duality, all Fg s are generated already at one-loop order and can be explicitly computed in the appropriate perturbative limit [14,15]. This allows in particular the study
of several properties, such as their exact behavior around the conifold singularity [14,1620].
In this work, we perform a systematic study of N = 4 topological amplitudes, associated to
type II string compactifications on K3 T 2 , or heterotic compactifications on T 6 . The N = 4
topological -model on K3, describing the six-dimensional (6d) type II compactifications was
studied in Refs. [21,22]. Although in principle simpler than the N = 2 case, in practice the
analysis presents a serious complication due to the trivial vanishing of the corresponding partition
function.
The topological twist consists of redefining the energymomentum tensor TB of the internal
SCFT by TB 12 J , with J the U (1) current of the N = 2 subalgebra of N = 4. The conformal weights h are then shifted by the U (1) charges q according to h h q/2. As a result,
the N = 2 supercurrents G+ and G , of U (1) charge q = +1 and q = 1, acquire conformal
dimensions one and two, respectively; G+ becomes the BRST operator of the topological theory, while G plays the role of the reparametrization anti-ghost b used to be folded with the
Beltrami differentials (when combined from left- and right-moving sectors) and define the integration measure over the moduli space of the Riemann surface. Thus, the genus g partition
function of the topological theory involves (3g 3) G insertions. On the other hand, there is a
U (1) background charge given by c(g
1), with c the central charge of the N = 2 SCFT before
the topological twist. In the N = 2 case, the critical (complex) dimension is c = 3, corresponding to compactifications on a CalabiYau threefold, and the total charge of the (3g 3) G
insertions matches precisely with the required U (1) charge to give a non-vanishing result. In the
N = 4 case however, the critical dimension is c = 2, corresponding to K3 compactifications, and
the U (1) charge is not balanced. In Ref. [21], it was then proposed to consider correlation func are the other two supercurrents of
+ insertions, where G
tions by adding (g 1) (integrated) G
2
the N = 4 SCFT.
It was also shown [21], using the GreenSchwarz (GS) formalism, that the above topological correlation functions compute a particular class of higher derivative interactions in the 6d
4g
effective field theory of the form W++ , where W++ is an N = 4 chiral superfield (N = 2 in
2 Actually, an additional insertion of the (integrated) U (1) current J is required in order to obtain a non-vanishing
result.
42
six dimensions) containing a graviphoton field strength as lower component and the Riemann
tensor as the next bosonic upper component. Moreover, an effective 6d self-duality is defined by
projecting antisymmetric vector indices, say A , using the Lorentz generators in the reduced
4g
(chiral) spinor representation ( )a b , with a, b = 1, . . . , 4. As a result, W++ generates in par4
4g4
ticular an effective action term of the form R T
, receiving contributions only from g-loop
order. These terms can be covariantized in terms of the full SU(2) R-symmetry of the extended
supersymmetry algebra and the corresponding coefficients were shown to satisfy a harmonicity
relation, which is a generalization of the holomorphicity to the harmonic superspace and follows
as a consequence of the BPS character of the effective action term integrated over half of the full
N = 2 6d superspace.
It turns out that upon string duality, these amplitudes are mapped on the heterotic side to
contributions starting from genus g + 1, and thus they cannot be studied similar to the N = 2 4d
couplings Fg at one loop. In this work, we propose and study alternative definitions of the N = 4
topological amplitudes and their string theory interpretations, that avoid multiple point insertions
and have tractable heterotic representations.
Our starting point is to consider type II compactifications in four dimensions on K3 T 2 , associated to an SCFT which in addition to the N = 4 part of K3 has an N = 2 piece describing T 2 .
Since the central charge is now c = 3, the U (1) charge of the (3g 3) G insertions, with G
being the sum of K3 and T 2 parts, defining the genus g topological partition function, balances
the corresponding background charge deficit, avoiding the trivial vanishing of the critical case.
However, the partition function still vanishes due to the N = 4 extended supersymmetry of K3.
In order to obtain a non-vanishing result, one needs to add just two insertions. One possibility is
to insert the (integrated) U (1) current of K3, JK3 , and the (integrated) U (1) current of T 2 , JT 2 :
2

3g3

(1.1)
G (a ) JK3 JT 2

2
Mg
a=1
K3T
where a are the Beltrami differentials. Another possible definition of a non-trivial topological
quantity is
2
3g4

G (a )JK3 (3g3 )(z)

(1.2)

2
Mg
a=1
K3T
where J is part of the SU(2) currents of the N = 4 superconformal algebra (J , J, J ++ ),

whose dimension is
is the fermionic partner of the (complex) T 2 coordinate (JT 2 = ),
zero and charge +1 in the twisted theory, and z is an arbitrary point on the genus g Riemann
surface. Furthermore g 2 and a complete antisymmetrization of the Beltrami differentials is
understood. It turns out that both definitions of the topological partition function correspond to
actual physical string amplitudes in four dimensions. The former corresponds in particular to
2 R 2 T 2g2 , with the subscripts () + denoting the (anti-) self-dual part of the corresponding
R+
+
2 (dd )2 T 2g4 , where is the
field strength,3 while the latter corresponds to the term R+
+
+
+
complex graviscalar from the 6d self-dual graviton with positive charge under the N = 2 U (1)
current JT 2 ; it corresponds to the Khler class modulus of T 2 . Moreover, on the heterotic side,
3 The subscript will be mostly dropped in the following, for notational simplicity.
43
Fig. 1. Overview over the topological N = 4 amplitudes. R+ (R ) denotes the (anti-) self-dual part of the Riemann
tensor, T+ the self-dual part of the graviphoton field strength and + ( ) stands for a graviscalar with one positive
(negative) unit of U (1)-charge with respect to the current JT 2 .
the first starts appearing at two loops for every g, while the second at one loop and can therefore
be studied in a way similar to the N = 2 couplings Fg s.
We also obtain the generalization of the harmonicity relation for the full SU(4) R-symmetry
using known results of 4d N = 4 supersymmetry. The moduli in this case belong to vector multiplets and transform in the two-index antisymmetric representation of SU(4). Although we cannot
check this relation on the type II side because it involves RR (RamondRamond) field backgrounds, we can also derive it in the heterotic theory at the string level, where the whole SO(6) is
manifest, for the minimal series that appears already at one loop. The resulting equation reveals
the presence of an anomalous term, in analogy with the holomorphic anomaly of the N = 2 Fg s.
The structure of this paper is the following. We start in Section 2 by a brief review of the N = 4
superconformal algebra and its topological twist. We then proceed with our strategy in finding
topological amplitudes, depicted in Fig. 1. From the known R 4 T 4g4 coupling of 4 gravitons and
(4g 4) graviphotons on K3, we redo the computation (which was initially performed in [21])
in Section 3 using the RNS formalism. In Section 4, we compactify two additional dimensions
on a 2-torus and study various couplings of (2g 2) graviphotons and a number of gravitons and
44
KaluzaKlein (KK) graviscalars from the four 6d Riemann tensors. In Section 5, we consider
further modifications by looking at an odd number of vertex insertions and modifying the genus
of the corresponding world-sheet. In Section 6, we translate the type IIA scattering amplitudes
to their duals in the heterotic theory compactified on T 4 and T 6 , respectively, and in Section 7
we explicitly compute one of them as a 1-loop amplitude. In Section 8, we give a brief review
of the holomorphicity of N = 2 Fg s and we present a qualitative derivation of the harmonicity
relation for the topological amplitudes on K3. This allows us in Section 9 to generalize it in
four dimensions for the new topological amplitudes on K3 T 2 , by introducing appropriate
harmonic-like variables that restore the SU(4) symmetry. Indeed, in Section 10, we derive
this relation on the heterotic side on T 6 and find also an anomaly arising from boundary terms.
Finally, in Section 11, we present our conclusions. Some basic material and part of our notation
is presented in a number of appendices. Appendix A contains a representation of -matrices and
Lorentz generators, while in Appendix B we define the notion of self-duality in six dimensions.
Appendix C contains the definition of -functions, prime forms and the Riemann vanishing
theorem. In Appendix D, we review the N = 4 superconformal algebra for K3 orbifolds and
K3 T 2 . Finally, in Appendix E, we present some new amplitudes with insertions of Neveu
Schwarz (NS) graviphotons.
2. N = 4 superconformal algebra and its topological twist
A detailed review can be found in Appendix D. Here, we summarize the main expressions and
properties of its generators, the algebra and the topological twist, that will be used throughout
this paper. The N = 4 algebra associated to type II compactifications on K3 involves besides the
B , two complex spin-3/2 super currents G+ , G
+ (and their comenergymomentum tensor TK3
K3
K3
), and three spin-1 currents forming an SU(2) R-symmetry algebra

plex conjugates GK3 , G
K3
++
JK3
, its complex conjugate JK3
and JK3 . In this notation, the superscripts +/ count the units
of charge with respect to the U (1) generator JK3 , while the supercurrents form two SU(2) dou
+
blets (G+
K3 , GK3 ) and (GK3 , GK3 ). It follows that the non-trivial operator product expansions
(OPEs) among them are:
JK3
(0)
,
z
G
K3 (0)
JK3 (z)G
(0)
,
K3
z
(0)
G
K3
JK3
(z)G+
(0)
,
K3
z
+ (0)
G
++
K3
JK3
(z)G
,
K3 (0)
z
++
J ++ (0) JK3
(0)
+ (z)G+ (0) 2 K3
+
,
G
K3
K3
2
z
z
JK3 (0)
+
+
G
K3 (z)GK3 (0) GK3 (z)GK3 (0)
z2
JK3 (z)JK3
(0) 2
JK3 (0)
,
z
(0)
G
(0) K3 ,
JK3 (z)G
K3
z
+
G
(0)
++
(0) K3 ,
(z)G
JK3
K3
z
G (0)
+ (0) K3 ,
(z)G
JK3
K3
z
J (0) JK3
(0)
(z)G (0) 2 K3
+
G
,
K3
K3
2
z
z
T B (0) 12 JK3 (0)
+ K3
.
(2.1)
z
++
(z)JK3
(0)
JK3
In the case of K3 T 2 , the R-symmetry group of the N = 4 algebra is extended to SU(2)

2 and
U (1) and the above operators are supplemented by those referring to the torus: G
,G
T2
T
45
JT 2 satisfying the non-trivial OPEs:

(0)
JT 2 (z)G
T2
G
(0)
T2
z
2 (0)
JT 2 (z)G
T
2 (z)G
+2 (0) JT 2 (0) +
(z)G+
(0) G
G
T2
T2
T
T
z2
2 (0)
G
T
z
(2.2)
TTB2 (0) 12 JT 2 (0)

z
(2.3)
+
Note that G+ = G+
K3 + GT 2 and J = JK3 + JT 2 are the supercurrent and the U (1) current
of the N = 2 subalgebra, while TF = G+ + G is the internal part of the N = 1 world-sheet
supercurrent.
In this work for simplicity, we will consider only K3 orbifolds, but the final results will
be true for generic K3. In the orbifold case, the N = 4 generators are given in terms of free
fields. Starting with the ten real bosonic and fermionic spacetime coordinates, Z M and M
respectively, with M = 0, 1, . . . , 9, we introduce a (Euclidean) complex basis4 :

1
X 1 = Z 0 iZ 1 ,
2

1
X 2 = Z 2 iZ 3 ,
2

1
X 3 = Z 4 iZ 5 ,
2

1
X 4 = Z 6 iZ 7 ,
2

1
X 5 = Z 8 iZ 9 ,
2

1
1 = 0 i 1 ,
2

1
2 = 2 i 3 ,
2

1
3 = 4 i 5 ,
2

1
4 = 6 i 7 ,
2

1
5 = 8 i 9 .
2
(2.4)
(2.5)
(2.6)
(2.7)
(2.8)
In our conventions, X 1,2 , X 3 and X 4,5 correspond to the (complex) coordinates of the noncompact 4d spacetime, the 2-torus T 2 , and K3 orbifold, respectively. In this basis, the expressions of the N = 4 generators are:
G+
K3 = 4 X4 + 5 X5 ,
+ = 5 X4 + 4 X5 ,
G
K3
G
K3 = 4 X4 + 5 X5 ,
= 5 X 4 + 4 X 5 ,
G
K3
++
JK3
= 4 5 ,
JK3 = 4 4 + 5 5 ,
G+
T2
G
T2
= 3 X 3 ,
= 3 X3 ,
+2
G
T
2
G
T
JK3
= 4 5 ,
(2.9)
= 3 X3 ,
= 3 X 3 ,
JT 2 = 3 3 .
4 In order to avoid confusion with the indices, we adopt the following convention:
real spacetime indices: , = 0, . . . , 5,

2, 2,
3, 3,
complex spacetime indices: A, B = 1, 1,

where a bar means complex conjugation.
(2.10)
46
The topological twist is defined as usual by shifting the energymomentum tensor TB with
(half of) the derivative of the U (1) current of the N = 2 subalgebra of the full N = 4 SCFT:
1
TB TB J.
(2.11)
2
As a result, the conformal dimensions h of the operators are shifted by (half of) their U (1)
charges, h h q/2. Thus, the new conformal weights are:

2
2
= dim G
= 1,
dim G
= dim G
= 2,
dim G+
T 2 ,K3
T ,K3
T 2 ,K3
T ,K3
++
dim JK3 = 0,
dim JK3 = 2,
dim[A ] = 0,
(2.12)
dim[ A ] = 1.
Moreover, the last of the OPE relations (2.1) and (2.3) is changed to:
(z)G
+ (0) J (0) + TB (0) .
G (z)G+ (0) G
(2.13)
z
z2
The physical Hilbert space of the topological theory is defined by the cohomology of the op + , while G (or G
) can be used to define the integration measure over the
erators G+ and G
Riemann surfaces. Thus, the physical states are all primary chiral states of the N = 2 SCFT
satisfying h = q/2, while antichiral fields are BRST exact and should decouple from physical
amplitudes (in the absence of anomalies). Indeed, they can be written as contour integrals of G+
and upon contour deformation and the OPE relation (2.13), one gets insertions of the energy
momentum tensor TB leading to total derivatives (see e.g. [3]).
3. Review of N = 4 topological amplitudes in 6 dimensions
In [21] g-loop scattering amplitudes of type II string theory compactified on K3 involving four
gravitons and (4g 4) graviphotons were found to be topological by using the GreenSchwarz
formalism. The goal of this section is to repeat this computation using the RNS formalism for
orbifold compactifications in 6 dimensions.
3.1. The supergravity setup
The superfield used to write down the 6-dimensional couplings is the following subsector of
the 6d, N = 2 Weyl superfield
b ij b ij b i j
L R R + ,
=
T +
Wa
(3.1)
a
a
with T the graviphoton field strength and R the Riemann tensor in 6 dimensions. The space
time indices run over , , , = 0, 1, 2, 3, 4, 5 and the spinor indices are Weyl of the same
6-dimensional chirality and take values a, b = 1, 2, 3, 4. For the SU(2)L SU(2)R R-symmetry
indices i, j , we will in most applications choose one special component (say i = j = 1 as in [21])
and hence neglect them in the following for notational simplicity. Finally, the matrices ( )a b
are in the reduced spinor representation of the 6-dimensional Lorentz group (for a precise definition and useful identities see Appendix A).
This multiplet has been formally introduced in [21] to reproduce the topological amplitudes,
and its full expression must contain additional fields of the 6d N = 2 gravity multiplet, such
as the two-form B , which are not relevant for our analysis. Moreover in [21], the superspace
47
Lagrangian

g
4 1
d L d 4 R1 Fg(6d) Wa1 b1 Wa2 b2 Wa3 b3 Wa4 b4 a1 a2 a3 a4 b1 b2 b3 b4 ,
(3.2)
1
was considered, with L/R
a special choice for the SU(2)L/R indices, and was shown using the
GS formalism that correlation functions computed from these couplings are topological. Note
that the integral is extended only over half the superspace and thus corresponds to an N = 4
F-type (BPS) term.5
In order to repeat the computation of [21] in the RNS formalism the precise knowledge of
the helicities and kinematics of the involved fields is essential. As far as the gravitons are concerned they can easily be extracted by considering the case g = 1 and performing the superspace
(6d)
integrations (we take the lowest component of Fg which is just a function of moduli scalars)

4 1
d L d 4 R1 Wa1 b1 Wa2 b2 Wa3 b3 Wa4 b4 a1 a2 a3 a4 b1 b2 b3 b4

= 1 1 a b1 2 2 a b2 3 3 a b3 4 4 a b4 a1 a2 a3 a4 b1 b2 b3 b4 c1 c2 c3 c4 d1 d2 d3 d4
1
2
3
4

1 1 d c1 2 2 d c2 3 3 d c3 4 4 d c4
1
R1 1 1 1 R2 2 2 2 R3 3 3 3 R4 4 4 4 .
(3.3)
The antisymmetrization identity

[d1 d2 d3 d4 ]
c1 c2 c3 c4 d1 d2 d3 d4 = [c
,
1 c2 c3 c 4 ]
(3.4)
finally leads to contractions of -matrices of the form

= 24 tr 1 1 2 2 tr 3 3 4 4 tr 1 1 2 2 3 3 4 4

tr 4 4 3 3 2 2 1 1 + tr 1 1 3 3 tr 2 2 4 4

tr 1 1 3 3 2 2 4 4 tr 4 4 2 2 3 3 1 1

+ tr 1 1 4 4 tr 3 3 2 2 tr 1 1 4 4 3 3 2 2

tr 2 2 3 3 4 4 1 1 1 1 a b1 2 2 a b2 3 3 a
1
b1 b2 b3 b4 a1 a2 a3 a4 R1 1 1 1 R2 2 2 2 R3 3 3 3 R4 4 4 4 .
b3
4 4
b4
a4
(3.5)
Using the relations (A.4) and (A.5) of Appendix A, this may also be written in terms of the full
Lorentz generators , as:

= 24 4 tr 1 1 2 2 tr 3 3 4 4 tr 1 1 2 2 3 3 4 4

+ 4 tr 1 1 3 3 tr 2 2 4 4 tr 1 1 3 3 2 2 4 4

+ tr 1 1 4 4 tr 3 3 2 2 tr 1 1 4 4 3 3 2 2

1 1 a b1 2 2 a b2 3 3 a b3 4 4 a b4 b1 b2 b3 b4 a1 a2 a3 a4
1
R1 1 1 1 R2 2 2 2 R3 3 3 3 R4 4 4 4 .
5 In this paper, we count supersymmetries in 4 dimensions, unless otherwise stated.
(3.6)
48
Defining the trace over the self-dual6 part of the Riemann tensor in 6 dimensions as (for an analog
see for example [23])

4
tr R+
= R1 1 1 1 R2 2 2 2 R3 3 3 3 R4 4 4 4 tr 1 1 2 2 3 3 4 4

1 1 a b1 2 2 a b2 3 3 a b3 4 4 a b4 a1 a2 a3 a4 b1 b2 b3 b4 ,
(3.7)
1
2
3
4

2 2
tr R+ = R1 1 1 1 R2 2 2 2 R3 3 3 3 R4 4 4 4 tr 1 1 2 2 tr 3 3 4 4

1 1 a b1 2 2 a b2 3 3 a b3 4 4 a b4 a1 a2 a3 a4 b1 b2 b3 b4 ,
(3.8)
1
the result of the superspace integration can be written as

4 1
d L d 4 R1 Wa1 b1 Wa2 b2 Wa3 b3 Wa4 b4 a1 a2 a3 a4 b1 b2 b3 b4

2 2
4
.
= 24 4 tr R+
tr R+
(3.9)
In order to uncover the index structure of the Riemann tensors involved in these traces, we use
the complex basis for the bosonic and fermionic coordinates (2.4)(2.8). With these definitions,
Eq. (3.9) becomes

1 2 2
4
4 tr R+ tr R+
8
= 32(R1212 R121
2 R12
3 R13
3 R23
R1 2 1 2 + R1313 R131
R1 3 1 3 + R2323 R232
R2 3 2 3 )
12
13
23
+ 2(R1313 R131
3 R12
2 R13
2 R232
3 R1 2 1 2
12
R1 2 1 2 + R1212 R121
13
R1 3 1 3 + R2323 R121
+ R2323 R131
3 R23
3 R13
R1 3 1 3 + R1313 R232
R2 3 2 3 + R1212 R12
R23
R2 3 2 3 )
23
13
12
23
+ ,
(3.10)
where the dots indicate terms where the two pairs of indices are not the same. Two examples for
such terms are
R2312 R121
2 R23 12
R1 2 1 2 ,
(3.11)
R1223 R12 23
R122
3 R1 2 2 3 ,
(3.12)
or
and will be neglected, since they do not lead to any different results (see for example Appendix E
for a later application).
3.2. Computation of the scattering amplitude
3.2.1. Basic tools
The analysis of the superspace expression (3.2) entails computing a scattering amplitude on a
compact genus g Riemann surface with the insertion of 4 gravitons with helicities corresponding
to one of the terms of Eq. (3.10), and (4g 4) graviphotons (see Fig. 2).
We choose the gravitons to be inserted in the 0-ghost picture and the graviphotons in the
(1/2)-ghost picture. Therefore, the graviton vertex is given by

+ ip eipZ :,
Vg(0) (p, h) = :h Z + ip Z
(3.13)
6 For a more detailed view on the definition of generalized self duality in 6 dimensions, see Appendix B.
49
Fig. 2. Scattering of 4 gravitons (1) and (4g 4) graviphotons (!) on a genus g compact Riemann surface.
with the convention, that a tilde denotes a field from the right-moving sector. Furthermore, h is
symmetric, traceless and obeys p h = 0. The graviphoton is an RR state and its vertex reads
(1/2)
VT

a
(p, ) = :e 2 (+)
p S a a b Sb S S + Sa b S b S S eipZ :,
(3.14)
where the polarization fulfills p = 0. is a free 2d scalar bosonizing the super-ghost system,
S a and Sa are (6-dimensional) spacetime spin-fields of opposite chirality and S is a spin-field
of the internal N = 4 sector (and S its right-moving counterpart). In the present case of chiral
graviphotons, only the first term in (3.14) is taken into account (see also Appendix B in this
respect).
Counting the charges of the super-ghost system, there is a surplus of (2g 2) which has
to be canceled by inserting (2g 2) picture changing operators (PCO) at random points of the
surface. Another (2g 2) insertions are due to the integration of super-moduli making a total of
(4g 4) PCOs.
The first question is, which spin fields to choose for the graviphotons and which parts of the
PCO to contract them with. For the moment we only discuss the left-moving sector, since the
right movers follow exactly in the same way. Concerning the internal part we adopt the approach
of [21,22] and choose all graviphotons to have the same internal spin-field, namely
i
S = e 2 (4 +5 ) ,
(3.15)
where I bosonize the fermionic coordinates I defined in Eqs. (2.4)(2.8), I = eiI . This
creates a charge surplus of (2g 2) in the 4 and 5 plane which can only be balanced by the
picture changing operators. In other words we have (2g 2) PCO contributing
e ei4 X4 ,
and another (2g 2)
(3.16)
providing7
e ei5 X5 .
(3.17)
Finally we are left to deal with the spacetime part and since we are considering chiral
graviphotons, the following spin-fields are possible
i
S1 = e 2 (1 +2 +3 ) ,
S2 = e
i
2 (1 2 3 )
7 An antisymmetric sum over all insertions is implicitly assumed.
(3.18)
(3.19)
50

i
S3 = e 2 (1 +2 3 ) ,
S4 = e
i
2 (1 2 +3 )
(3.20)
(3.21)
Let us assume that ai graviphotons are inserted with the spin field Si , where the ai are subject to
the set of linear equations
a1 + a2 a3 a4 = 0,
a1 a2 + a3 a4 = 0,
a1 a2 a3 + a4 = 0,
a1 + a2 + a3 + a4 = 4g 4,
which reflect the constraints of charge cancellation in each of the three planes and the fact that
there is a total of (4g 4) graviphotons. The unique solution of this system is a1 = a2 = a3 =
a4 = g 1.8
Summarizing, until now we have inserted the following fields, where the columns i denote
the charges of the given field with respect to the various planes
Insertion
Number
Position
+ 12
+ 12
12
12
+ 12
12
+ 12
12
+ 12
12
12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
0
0
0
0
0
0
1
0
0
1
Graviphoton
g1
g1
g1
g1
xi
yi
ui
vi
PCO
2g 2
2g 2
{s1 }
{s2 }
and {si } stands for a collection of (a priori arbitrary) points on the Riemann surface.
In order to completely determine the amplitude still the graviton helicities are missing. To
this end, one has to pick one of the terms in (3.10). Since most of them are related by simply
exchanging two planes, it suffices to look at two generic cases.
3.2.2. Graviton combination I
When focusing on the fermionic part of the vertex operator in (3.13), one possible setup of
charges is (only the left-moving part is displayed):
Insertion
Number
Position
Graviton
1
1
1
1
z1
z2
z3
z4
0
+1
0
1
+1
1
+1
1
+1
0
1
0
0
0
0
0
0
0
0
0
8 Of course there would still be the possibility of leaving a surplus charge from the graviphotons, which is then canceled
by the graviton insertions. However these amplitudes would stem from terms of the form (T RT )4 T 4g12 in the effective
action which also arise when performing the full superspace integration in (3.2) and we will thus neglect them in the
remainder of the paper.
51
The correlation function can now be written in the form9

Fg(6d)
g1

e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
i=1
i
2 (1 +2 3 +4 +5 )
(ui )
i=1
e 2 e 2 (1 2 3 +4 +5 ) (yi )
i=1
g1

g1

e 2 e 2 (1 2 +3 +4 +5 ) (vi )
i=1
i(2 +3 )
{s1 }

g1

(z1 ) e
i(1 2 )
e ei4 X4 (ra )
{s2 }

(z2 ) e
i(2 3 )
(z3 ) ei(1 +2 ) (z4 )

e ei5 X5 (ra ) .
(3.22)
Computing all possible contractions for a fixed spin structure, the result for the amplitude is
Fg(6d)

+ yi ui vi ) + z2 z4
= g1
4g4
4g4
4g4
s 12 i=1 (xi + yi + ui + vi ) a=1 ra + 2 a<b=1 E(ra , rb ) a=1 2 (ra )
s
1 g1
2
i=1 (xi
1 g1
2

h4 ,s

h5 ,s

1 g1
i=1 (xi yi + ui vi ) + z1 z2 + z3 z4 s 2
i=1 (xi yi ui + vi ) + z1
g1
E(x
,
z
)E(y
,
z
)E(u
,
z
)E(v
,
z
)E(z
,
z
)E(z
i 4
i 1
i 2
i 3
1 2
1 , z4 )E(z2 , z3 )E(z3 , z4 )
j =1
{s }
{s }
1

1
(xi + yi + ui + vi )
ra
2
a
g1
i=1
2

1
ra
2
a
g1
z3
i=1
g1

(xi ) (yi ) (ui ) (vi )
{s2 }

a<b
E(xi , xj )E(yi , yj )E(ui , uj )E(vi , vj )
i<j
i=1
g1

E(ra , rb )
g1

j =1
E(xi , z1 )E(yi , z2 )E(ui , z3 )E(vi , z4 )
{s1 }

E(ra , rb )
a<b
{s1 }

X4 (ra )
{s2 }

X5 (ra ).
Here hI ,s stands for the Jacobi theta-function [24] with the spin structure s and twist hI , E(x, y)
is the prime form, is the g2 -differential with no zeros or poles and stands for the Riemann
theta constant (see also Appendix C). We should also mention again, that in this computation
(and also throughout the paper), phase factors and numerical constants of the amplitude, as well
as factors of the chiral determinant Z1 of the (1, 0) system are dropped. The latter can most
easily be restored by comparing with the N = 2 case at any intermediate step [2] and disappear
completely in the final result.
9 In the following, we drop g-dependent overall factors, which can be easily restored as in Ref. [2].
52
In order to simplify the expression the a priori arbitrary PCO insertion points ra are fixed in
the following way10
g1
g1
4g4

1
1
(xi + yi ui vi ) + z2 z4 =
ra + 2
2
2
i=1
i=1
4g4

a=1
g1

ra =
(ui + vi ) z2 + z4 + 2,
a=1
(3.23)
i=1
which results in a cancellation of one -function in the numerator against the one in the denominator. The next step is the sum over the different spin structures, for which the relevant terms are
the four remaining -functions:
g1

1
(xi yi + ui vi ) + z1 z2 + z3 z4 ,
s
2
i=1
g1

1
(xi yi ui + vi ) + z1 z3 ,
s
2
i=1
g1

{s1 }

1
ra ,
h4 ,s
2
a
i=1
g1

{s2 }

1
ra .
h5 ,s
(3.24)
2
a
i=1
To this end, the Riemann identity is used (see Appendix C.2). In the special case of (3.24) the
summed arguments read respectively
T++++ :
g1

xi +
i=1
g1
4g4
g1

1
1
1
(ui + vi ) + z1 (z2 + z4 )
ra =
xi + z1 z4 ,
2
2
2
i=1
a=1
i=1
T++ :
g1

yi +
i=1
g1
4g4
g1

1
1
1
(ui + vi ) z1 + (z2 + z4 )
ra =
yi z1 + z2 ,
2
2
2
i=1
a=1
i=1
T++ :
{s }
{s }
1
2
1
1
1
1
(ui vi ) z3 + (z2 + z4 ) +
ra
ra
2
2
2 a
2 a
g1
i=1
{s1 }

a=1
ra
g1

ui + z2 z3 ,
i=1
10 In the remainder of this section we will refer to this choice as the gauge choice or gauge condition.
53
T++ :
{s }
{s }
1
2
1
1
1
1
(ui vi ) z3 + (z2 + z4 )
ra +
ra
2
2
2 a
2 a
g1
i=1
{s2 }

ra
a=1
g1

ui + z2 z3 ,
i=1
where the gauge condition (3.23) was also used. With the further relation11 h4 + h5 = 0, the
correlation function takes the form
Fg(6d)
g1
g1

xi + z1 z4
i=1 yi z1 + z2
= 4g4
4g4 2
g1
a<b=1 E(ra , rb ) a=1 (ra ) j =1 E(xi , z4 )E(yi , z1 )E(ui , z2 )E(vi , z3 )
{s1 }
{s2 }

g1
g1
h4
i=1 ui + z2 z3 h5
i=1 ui + z2 z3
a=1 ra
a=1 ra
E(z1 , z2 )E(z1 , z4 )E(z2 , z3 )E(z3 , z4 )
g1

i=1
g1

i=1
(xi ) (yi ) (ui ) (vi )
{s1 }

E(ra , rb )
a<b
{s2 }

E(ra , rb )
a<b
E(xi , xj )E(yi , yj )E(ui , uj )E(vi , vj )
i<j
g1

E(xi , z1 )E(yi , z2 )E(ui , z3 )E(vi , z4 )
i=1
{s1 }

a
X4 (ra )
{s2 }

X5 (ra ).
(3.25)
Now we can in two occasions use the so-called bosonization formulas [25] to write
g1
g1
g1
g1

i=1 xi + z1 z4
E(xi , xj )
E(xi , z1 )
(xi ) (z1 )
g1
i=1
i=1
i=1 E(xi , z4 )E(z1 , z4 ) (z4 ) i<j
= Z1 det i (x1 , x2 , . . . , xg1 , z1 ),
(3.26)
g1
g1
g1
g1

i=1 yi z1 + z2
E(y
,
y
)
E(y
,
z
)
(yi ) (z2 )
i j
i 2
g1
E(y
,
z
)E(z
,
z
)
(z
)
i
1
1
2
1
i<j
i=1
i=1
i=1
= Z1 det i (y1 , y2 , . . . , yg1 , z2 ),
(3.27)
where are the Abelian 1-differentials, of which there are g on a compact Riemann surface of
genus g. With the help of the gauge condition (3.23) and upon multiplying (3.25) with
g1

i=1 vi + z4 z3
1 = 4g4
(3.28)
,
g1
a=1 ra
i=1 ui + z2 z3 3
11 The fact that the sum over all twists vanishes is a consequence of spacetime supersymmetry.
54
also a third time the bosonization identity can be used

g1
g1
g1
g1

i=1 vi + z4 z3
E(vi , vj )
E(vi , z4 )
(vi ) (z4 )
g1
i=1
i=1
i=1 E(vi , z3 )E(z4 , z3 ) (z3 ) i<j
= Z1 det i (v1 , v2 , . . . , vg1 , z4 ).
(3.29)
Taking into account all these identities and plugging them in (3.25), we obtain the result:
Fg(6d)

{s2 }
g1
ui + z2 z3 h5
i=1 ui + z2 z3
a=1 ra
=
4g4
g1
4g4
i=1 ui + z2 z3 3 a<b=1 E(ra , rb ) a=1 2 (ra )
g1
{s1 }
{s2 }
g1
i=1 (ui ) a<b E(ra , rb ) a<b E(ra , rb ) i<j E(ui , uj ) (z3 )
g1
j =1 E(ui , z2 )E(z2 , z3 ) (z2 )
h4
{s1 }
g1
a=1 ra
4g4
a=1 ra
g1

E(ui , z3 )
i=1
{s1 }

X4 (ra )
i=1
{s2 }

X5 (ra ) det i (xi , z1 ) det i (yi , z2 ) det i (vi , z4 ).
With the help of further identities of [25], this can be rewritten in terms of correlation functions
of the following kind
Fg(6d)
=
{s1 }
a
{s1 }

4 (ra ) 4 (z2 )
X4 (ra )
{s2 }

2 }

g1
4 (ui )4 (z3 ) {s
a 5 (ra )5 (z2 ) i=1 5 (ui )5 (z3 )

4g4
g1
a=1 b(ra )b(z2 ) i=1 c(ui )c(z3 )
g1
i=1
X5 (ra ) det i (xi , z1 ) det i (yi , z2 ) det i (vi , z4 ),
(3.30)
where the following relations were used

{s1 }
{s1 }
g1

{s1 }
h4
a ra
a<b E(ra , rb ) i<j E(ui , uj ) a E(ra , z2 )
i=1 ui + z2 z3
{s1 }
g1
g1

a,i E(ra , ui ) a E(ra , z3 ) i=1 E(ui , z2 )E(z2 , z3 ) i=1 (ui ) (z3 )
{s }

g1
g1
{s1 }
1

E(ui , z3 )
(ra ) (z2 ) =
4 (ui )4 (z3 ) ,
4 (ra ) 4 (z2 )
i=1
i=1
together with a similar definition for the correlator of 5 and

4g4
4g4
g1

4g4
a<b E(ra , rb ) i<j E(ui , uj ) a=1 E(ra , z2 )

a=1 ra
i=1 ui + z2 z3 3

4g4
g1
g1
E(ra , z3 ) i=1 E(ui , z2 )E(z2 , z3 ) i=1 3 (ui ) 3 (z3 )
a,i E(ra , ui ) a
4g4

g1
4g4
g1

3
3
E(ui , z3 )
(ra ) (z2 ) =
b(ra )b(z2 )
c(ui )c(z3 ) .
i=1
a=1
i=1
Rewriting the expression (3.30) using the operators of the N = 4 superconformal algebra described in Section 2, one finds
55
4g4
Fg(6d)

g1 ++
++
G
K3 (ra ) i=1 JK3 (ui )JK3 (z2 )JK3 (z3 )
=
4g4

g1
a=1 b(ra )b(z2 ) i=1 c(ui )c(z3 )
det i (xi , z1 ) det i (yi , z2 ) det i (vi , z4 ).
a=1
In the next step, using the fact that the gauge condition (3.23) still allows the freedom to move
g PCO points, we let the positions of (g 1) of the PCOs collapse with the ui and another
+ (the
one with z3 , which with the help of the OPEs in Section 2 and Appendix D yields G
K3
fact that there is no singularity in this procedure owes to the bc ghost system correlator in the
denominator):
3g4

g1 +
(ui )J (z2 )G
+ (z3 )
GK3 (ra ) i=1 G
a
K3
K3
(6d)
Fg =
3g4

a=1 b(ra )b(z2 )
det i (xi , z1 ) det i (yi , z2 ) det i (vi , z4 ).
(3.31)
This amplitude is to be multiplied by the (3g 3) Beltrami differentials folded with the b ghosts
to provide the correct measure for the integration over the genus g moduli space. Note that the
ratio of the correlators appearing in (3.31) is independent of the (3g 3) positions ra and z2 .
These positions can therefore be transmuted to the positions of the operators b that are folded
with the Beltrami differentials; the correlation functions of the latter then cancel with the denominator of (3.31). Including now the right-moving sector, we can integrate out the graviton and
graviphoton insertion points giving finally
2 g

3g4

2

G
+ det(Im ) 3 ,
Fg(6d) =
(3.32)
G

K3 (i )JK3 (3g3 )
K3

i
Mg
i=1
where it is understood that the Beltrami differentials are totally antisymmetrized and the absolute
square indicates, that also the right-moving contributions have been restored. The additional
factor of (det(Im ))3 cancels exactly the zero-mode contribution of the spacetime bosons and
the remaining expression is topological and exactly the same
in [21] (Eq. (2.15)). Actually,
as
+ s as the contour integral G
+ (z3 ) = G
+ JK3 (z3 ) (see OPEs (2.1) and
by writing one of G
K3
K3
weights (2.12)) and pulling off the contour integral, it can only encircle JK3
which converts it
to GK3 . Eq. (3.32) can then be expressed in a more symmetric form:

2 g1

3g3

2

G
+
|JK3 |2 ,
G
Fg(6d) =
(3.33)

K3 (i )
K3

i
Mg
i=1
int
where the subscript int means that the correlator is restricted to the internal (K3) part.
3.2.3. Graviton combination II
The other graviton combination different from the one considered previously in Section 3.2.2
is given
Insertion
Number
Position
Graviton
1
1
1
1
z1
z2
z3
z4
+1
+1
1
1
+1
1
+1
1
0
0
0
0
0
0
0
0
0
0
0
0
56
The amplitude can now be expressed as

g1
g1
i
i
(6d)
e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
e 2 e 2 (1 2 3 +4 +5 ) (yi )
Fg =
i=1
g1

i=1
i=1
i(1 +2 )
{s1 }

e 2 e 2 (1 +2 3 +4 +5 ) (ui )
(z1 ) e
i(1 2 )
e ei4 X4 (ra )
{s2 }

(z2 ) e
g1

e 2 e 2 (1 2 +3 +4 +5 ) (vi )
i=1
i(1 +2 )
(z3 ) e(i1 +2 ) (z4 )

e ei5 X5 (ra ) .
(3.34)
Going through similar steps as in the previous case, the result after performing the spin-structure
sum is
Fg(6d)
=

{s2 }
g1
g1
i=1 ui z3 + z2 h5
i=1 ui z3 + z2
a=1 ra
a=1 ra
4g4
4g4
g1
4g4 2
a=1 ra
i=1 ui z3 + z2 3
a<b=1 E(ra , rb ) a=1 (ra )
g1
{s1 }
{s2 }
g1
i=1 (ui ) a<b E(ra , rb ) a<b E(ra , rb ) i<j E(ui , uj ) (z3 )
g1
j =1 E(ui , z2 )E(z2 , z3 ) (z2 )
h4
{s1 }
g1

E(ui , z3 )
{s1 }

a
i=1
X4 (ra )
{s2 }

X5 (ra ) det i (xi , z1 ) det i (yi , z2 ) det i (vi , z4 ).
Note, that this is exactly the same expression as (3.30), such that the final answer can immediately
be written down
2 g

3g4

+ 2
(6d)
Fg =
(3.35)
GK3 (i )JK3 (3g3 )
.
GK3

Mg
i=1
int
Thus, we conclude that both generic helicity combinations of the superfield expression in (3.2)
lead to the same topological result, which is in perfect agreement with [21].
4. Type IIA on K3 T 2
After having established the results of [21] in the RNS formalism, the question arises whether
similar topological amplitudes can be found for N = 4 supersymmetry in 4 dimensions as well.
Our basic strategy in approaching this question will be to toroidally compactify two additional
dimensions and study again couplings on a compact genus g Riemann surface. A major difference compared to the above cases is that in addition to the 4-dimensional gravity multiplet there
are additional KaluzaKlein fields which can take part in the scattering process.
The easiest way to start out is to begin with the 6-dimensional graviton insertions of (3.10),
compactify one additional plane and supplement the arising fields with (2g 2) graviphotons
from the 4-dimensional Weyl multiplet. Note, that now we consider a coupling proportional to
2g2
4g4
T+
instead of T+ . This is because of the cancellation of U (1) charge, which is similar the
57
4g4
critical case of N = 2 Fg s, as explained in the introduction. Moreover, if we started with a T+

coupling in 4 dimensions, there would be an additional factor of det(Im ) not being canceled
by the spacetime bosons, that would spoil the topological nature of the resulting amplitude.
Considering only the case that the left-moving sector is symmetric to the right-moving one there
are 3 a priori different cases to analyze12 :
all of the 6-dimensional gravitons remain gravitons in the 4-dimensional amplitude,
2 out of 4 gravitons get converted into graviscalars in the process of compactification,
all gravitons become graviscalars.
Here by graviscalars we mean the moduli insertions corresponding to either the complex structure
U or the Khler modulus T of the torus T 2 .
In the following we will show that all three cases lead to the same topological expression. The
result remains the same even if we replace graviscalar insertions with NS KK graviphotons that
we present in Appendix E. A brief review of some basic tools which will be needed throughout
the computation, namely the N = 4 superconformal algebra for compactification on K3 T 2
and its topological twist, is given in Section 2.
4.1. 4 graviton case
We study the graviton setup of Section 3.2.3 but regard the 3rd plane as the additionally
compactified torus. Since we have now only (2g 2) graviphoton insertions in the (1/2)-ghost
picture, the number of PCO has also diminished to (3g 3). The way to distribute the charges
of the graviphotons and PCO is
Insertion
Number
g1
g1
g1
g1
g1
Graviphoton
PCO
Position
xi
yi
{s3 }
{s4 }
{s5 }
+ 12
12
+ 12
12
+ 12
+ 12
+ 12
+ 12
0
0
0
0
0
0
+ 12
+ 12
0
0
1
1
0
0
0
1
0
The amplitude in this setup is given by13

Fg(1)
g1

e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
i=1
e 2 e 2 (1 2 +3 +4 +5 ) (yi )
i=1
i(1 +2 )
{s3 }

g1

(z1 ) e
i3
e e
i(1 2 )
X3 (ra )
{s4 }

a
(z2 ) e
i(1 +2 )
i4
e e
(z3 ) ei(1 +2 ) (z4 )
X4 (ra )
{s5 }

i5
e e
X5 (ra ) ,
12 For some remarks concerning asymmetric distributions of the left- and right-moving sector see Appendix E.
13 As before, only the left-moving part is considered. Furthermore, we introduce the superscript (1) to distinguish this
expression from others that will be computed in subsequent sections of the paper.
58
which, after performing the contractions and the spin structure sum, together with some manipulations, using the gauge choice
3g3

ra =
a=1
reads:
g1

yi z1 z2 + z3 + z4 + 2,
(4.1)
i=1
{s3 }
Fg(1)
4 ,s5 }

++
GK3 (ra )JK3
(z2 )JK3
(z3 )
3 X3 (ra ) 3 (z2 )3 () {s
a
=
3g3

b(ra )b(z2 )c(z3 )
a
det i (xi , z1 ) det i (yi , z4 ),
a
where is an arbitrary position on the Riemann surface and

{s3 }
{s3 }
{s3 }
{s3 }
a ra + z2 z3
a<b E(ra , rb ) a E(ra , z2 ) a (ra ) (z2 )
{s3 }
a E(ra , z3 )E(z2 , z3 ) (z3 )

{s }
3

=
3 (ra ) 3 (z2 )3 () .
a
Now one may collapse one of the ra (for simplicity the last one r3g3 is chosen) with z3 . The
resulting expression is:
{s3 }
3g4

a 3 X3 (ra )3 (z2 )3 ()
a=g GK3 (ra )JK3 (z2 )GK3 (z3 )
(1)
Fg =

3g4
b(ra )b(z2 )
a
det i (xi , z1 ) det i (yi , z4 ).
This can also be written as (for simplicity we consider = z2 )

3g4
a=1 G (ra )JK3 (z2 )3 3 (z2 )GK3 (z3 )

(1)
det i (xi , z1 ) det i (yi , z4 ).
Fg =
3g4

b(ra )b(z2 )
a
One can easily check that the first term in the above expression does not contain any poles
with non-vanishing residue. One can therefore transport all G and the JK3
to the Beltrami
differentials as before and including the right-moving sector and the integration over xi , yi , z1
and z4 , one finds:
2

3g4

2
2

+
G
3 3 (z2 )
(z3 )
G (a )JK3 (3g3 )
,
Fg(1) =
(4.2)

K3

Mg
a=1
int
where antisymmetrization of all the (3g 3) Beltrami differentials is understood.

Compared to the results of Section 3.2, we note the following formal14 differences:
+
The g insertions of G
K3 have been reduced to only one, which is due to the reduced
number of graviphotons present in the new amplitude. Note that the G s above are given
14 The comparison between (3.32) and (4.2) is only formal because the operators involved are from different superconformal algebras.
59
2
by the sum of the corresponding operators in K3 (G
K3 ) and the ones in T (GT 2 ). The
only non-zero contribution comes from terms where (g 1) of them come from the T 2 part,
which together with JT 2 = 3 3 soak all the g zero modes of the 1-differential 3 and one
zero mode of the zero-differential 3 in the twisted theory. The remaining (2g 4) G come
precisely account for the U (1) charge anomaly (2g 2) of

from K3 and together with JK3
the twisted K3 SCFT.
The power of det(Im ) coming from the integration of xi , yi , z1 and z4 has changed from
3 to 2, exactly canceling the zero mode contribution of the spacetime bosons which is now
4-dimensional as compared to 6-dimensional in Section 3.2.
+
JK3 (z3 ) and pulling off the contour integral which only converts
+ (z3 ) = G
By writing G
K3
K3
JK3 to GK3 , Eq. (4.2) can be expressed in a more symmetric form:

2

3g3

2
2

JK3 (z3 )
J 2 (z2 )
G (a )
,
Fg(1) =
(4.3)

T

a=1
Mg
int
that can be compared to Eq. (3.33) of the 6d case.

4.2. 2 graviton2 scalar case
We study the graviton setup of Section 3.2.2 and again compactify the 3rd plane to a torus.
Besides that, we keep the rest of the setup as in Section 4.1:
Insertion
Number
Position
Graviscalar
Graviton
Graviscalar
Graviton
1
1
1
1
z1
z2
z3
z4
0
+1
0
1
+1
1
+1
1
+1
0
1
0
0
0
0
0
0
0
0
0
Graviphoton
g1
xi
+ 12
+ 12
+ 12
+ 12
+ 12
g1
yi
12
12
+ 12
+ 12
+ 12
g1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
PCO
In this setup, the amplitude takes the form15

g1
g1
i
i
(1)
Fg =
e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
e 2 e 2 (1 2 +3 +4 +5 ) (yi )
i=1
i=1
ei(2 +3 ) (z1 ) ei(1 2 ) (z2 ) ei(2 3 ) (z3 ) ei(1 +2 ) (z4 )
{s3 }

a
i3
e e
X3 (ra )
{s4 }

a
i4
e e
X4 (ra )
{s5 }

i5
e e
X5 (ra ) .
15 We use the same symbol F (1) , since we find that this amplitude is actually identical to (4.3).
g
(4.4)
60
Since the calculations are quite similar to the case just discussed, we restrain from reporting the
details of the computation, but content ourselves by writing down the final result:
2

3g4

+
2
2

(1)
G
J 2 (z2 )
(z3 )
Fg =
G (a )JK3 (3g3 )

T
K3

Mg
a=1
Mg
a=1
int
2

3g3

2
2

JK3 (z3 )
J 2 (z2 )
=
G (a )
.

T

(4.5)
int
4.3. 4 scalar case

Finally, the 4 graviscalar case is obtained from the 4 graviton configuration of Section 4.1, by
exchanging one of the spacetime planes with the one of the internal tori for the configuration in
Section 3.2.3. The setup of charges is hence as follows:
Insertion
Number
Position
Graviscalar
1
1
1
1
Graviphoton
PCO
z1
z2
z3
z4
0
0
0
0
+1
+1
1
1
+1
1
+1
1
0
0
0
0
0
0
0
0
g1
xi
+ 12
+ 12
+ 12
+ 12
+ 12
g1
yi
12
12
+ 12
+ 12
+ 12
g1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
The corresponding amplitude reads

g1
g1
i
i
(1)
e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
e 2 e 2 (1 2 +3 +4 +5 ) (yi )
Fg =
i=1
i(2 +3 )
{s3 }

(z1 ) e
i3
e e
i(2 3 )
X3 (ra )
{s4 }

a
(z2 ) e
i=1
i(2 +3 )
i4
e e
(z3 ) ei(2 +3 ) (z4 )
X4 (ra )
{s5 }

i5
e e
X5 (ra ) ,
which, by similar steps as before, is reduced to the same expression as in Eq. (4.5).
4.4. Comparison of the results
Comparing the final results of Sections 4.14.3, one encounters that indeed all three are the
same. This implies, that the topological amplitude does not depend on the precise details of
the field content which is considered, but is only related to more general structures such as the
compactification and the number of inserted fields. It therefore seems an interesting question,
whether one can find different topological expressions if one varies some of these main properties
of the amplitude. This question will be studied in the next section.
61
5. Different correlators
The simplest way to find different correlators than the above ones is to consider different
numbers of vertex insertions. The two feasible options are:
Changing the number of scalar and graviton insertions to 3: As we will demonstrate, although
this will lead to a new topological expression one can show that it is identically vanishing.
Changing the number of graviphotons: Since a change in the number of graviphoton insertions implies (via the balancing of the super-ghost charges) also a change in the number
of PCO insertions, it is more convenient to consider just a different loop order, instead of
altering the number of fields; we will be more precise below.
We will exploit both possibilities in this section.
5.1. 1 graviton2 scalar amplitude
Tampering with the number of scalar and graviton insertions raises the main problem of how to
cancel the charges in the spacetime part of the amplitude. One setup, which makes this possible
is the following:
Insertion
Number
Position
Graviton
Graviscalar
1
1
1
z1
z2
z3
+1
1
0
+1
0
1
0
+1
1
0
0
0
0
0
0
Graviphoton
g1
xi
+ 12
+ 12
+ 12
+ 12
+ 12
g1
yi
12
12
+ 12
+ 12
+ 12
g1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
PCO
Although all such amplitudes turn out to vanish, we still present here the computation for pedagogical reasons, showing in particular in a concrete example the gauge independence of gauge
choice for the PCO positions. The amplitude written down in this setup is
g1
g1
i
i
(2)
e 2 e 2 (1 +2 +3 +4 +5 ) (xi )
e 2 e 2 (1 2 +3 +4 +5 ) (yi )
Fg =
i=1
i(1 +2 )
{s3 }

a
(z1 ) e
i3
e e
i(1 +3 )
X3 (ra )
{s4 }

a
i=1
i(2 +3 )
(z2 ) e
i4
e e
X4 (ra )
(z3 )
{s
5}

i5
e e

X5 (ra ) .
(5.1)
This correlation function can be treated in quite the same way as the other ones considered
above, however, there is a small subtlety concerning the gauge condition. Choosing the gauge
3g3

a=1
ra =
g1

i=1
yi z1 + z2 + 2,
(5.2)
62
the result of the amplitude, following the same steps as before, is given by
2

3g3

2

J 2 (z3 )
Fg(2) =
G (a )
.

T

Mg
a=1
(5.3)
int
However, choosing for example

3g3

ra =
a=1
g1

yi z1 + z3 + 2,
(5.4)
i=1
the spin structure sum gives a vanishing result, implying that the correlation function is zero.
Therefore, the topological expression (5.3) has to vanish as well, which can indeed be shown as
follows. In order to soak all the zero modes of the 3rd plane, the G of (5.3) have to be distributed
in the following way among T 2 and K3
2

g1
3g3

2

(2)

JT 2 (z3 )
Fg =
(5.5)
GT 2 (a )
GK3 (a )
.

a=g
Mg
a=1
int
Using the OPE relations (2.1), one can express one of the G
K3 s as the contour integral GK3 =
+
J . Deforming the contour, the only possible operators that can be encircled are G
G
K3 K3
K3
yielding however zero residue. This entails
Fg(2) = 0,
(5.6)
consistently with the alternative vanishing through the different gauge choice (5.4).
5.2. Amplitudes at genus g + 1
As explained already before, instead of altering the number of graviphotons taking part in the
scattering process it is equivalent to change the genus of the world-sheet. In other words, the
same vertex insertions as in Section 4 are considered, with the only difference, that an additional
handle is attached to the surface. Of course, the kinematics of all fields involved has to be altered,
as well:
Insertion
Number
Position
Graviscalar
1
1
1
1
z1
z2
z3
z4
0
0
+1
1
+1
1
+1
1
+1
+1
0
0
0
0
0
0
0
0
0
0
g1
g1
g+1
g1
g1
xi
yi
{s3 }
{s4 }
{s5 }
+ 12
12
0
0
0
+ 12
12
0
0
0
+ 12
+ 12
1
0
0
+ 12
+ 12
0
1
0
+ 12
+ 12
0
0
1
Graviton
Graviphoton
PCO
The two additional PCO are present because of the integration over the super-moduli of the new
surface, and are used to balance the 3 charge from the vertex insertions. Then, the amplitude
63
can be written as
Fg(3)
g1

i
2 (1 +2 +3 +4 +5 )
(xi )
i=1
g1

e 2 e 2 (1 2 +3 +4 +5 ) (yi )
i=1
i(2 +3 )
{s3 }

(z1 ) e
i(2 +3 )
e ei3 X3 (ra )
{s4 }

(z2 ) ei(1 +2 ) (z3 ) ei(1 +2 ) (z4 )
e ei4 X4 (ra )
{s5 }

e ei5 X5 (ra ) .
(5.7)
Since this amplitude is a bit non-standard because the number of graviphotons is 2g 2 on a

genus g + 1 surface, we give some details of the calculation below. By taking all the contractions
we find
Fg(3)
s
1 g1
2

s 12

1 g1
i=1 (xi yi ) + z3 z4 s 2
i=1 (xi yi ) + z1 z2 + z3 z4
3g1
g1
3g1
3g1 2
i=1 (xi + yi )
a=1 ra + 2
a<b=1 E(ra , rb ) a=1 (ra )
g1

g1

s 1
(xi + yi ) + z1 + z2 a{s3 } ra h4 ,s 12 i=1 (xi + yi ) a{s4 } ra
{s 2} i=1
g1
3
2
a E(ra , z1 )E(ra , z2 ) j =1 E(xi , z4 )E(yi , z3 )E (z3 , z4 )E(z1 , z4 )E(z2 , z3 )

h5 ,s

g1
g1
{s5 }

1
(xi + yi )
ra E(z1 , z3 )E(z2 , z4 )
(xi ) (yi )
2
a
i=1
{s3 }

E(ra , rb )
a<b
g1

i=1
{s4 }

E(ra , rb )
a<b
{s5 }

E(ra , rb )
a<b
g1

E(xi , xj )E(yi , yj )
i<j
E(xi , z1 )E(xi , z3 )E(yi , z2 )E(yi , z4 )
{s3 }

j =1
X3 (ra )
{s4 }

a
X4 (ra )
{s5 }

X5 (ra ).
In this case, the gauge condition takes the form

g1
g1
3g1

1
1
(xi yi ) + z3 z4 =
(xi + yi )
ra + 2
2
2
i=1
i=1
3g1

a=1
ra =
g1

yi z3 + z4 + 2,
a=1
(5.8)
i=1
which results in a cancellation of the first -function in the numerator against the one in the
denominator. After performing the spin structure sum and deploying bosonization identities, the
result is given by ( is again an arbitrary position)
64
{s3 }
Fg(3)

3 X3 (ra )3 () a{s4 } 4 (ra )X4 4 (z2 ) a{s5 } 5 X5 (ra ) 5 (z2 )
=
3g1

a=1 b(ra )b(z2 )
det i (xi , z1 , z3 ) det i (yi , z2 , z4 ).
(5.9)
a
Using the expressions for the N = 4 superconformal algebra, this may also be written as
{3g1}

G (ra )3 ()J (z2 )
a
(3)
Fg =
(5.10)
det i (xi , z1 , z3 ) det i (yi , z2 , z4 ).
3g1

a=1 b(ra )b(z2 )
One immediately sees, that there are no poles or zeros if one of the ra approaches z2 , which
means, that the above expression is independent of z2 . This entails, that one can transport the
(3g 1) G and the J to the 3g Beltrami differentials. The integral over xi , yi , z1 , z2 , z3 , z4
can be performed without obstacles (including the right-moving sector) yielding a (det(Im ))2
which is needed to cancel the bosonic spacetime part. The remaining internal part can then be
written as (a complete antisymmetrization of the Beltrami differentials is understood):
2
3g1

(3)
Fg =
(5.11)
G (a )JK3 (3g )3 ()
.

a=1
Mg+1
int
This amplitude seems to be the minimal topological amplitude, in the sense, that it contains the
minimum number of insertions of additional SCFT-operators, while still being non-trivial and
topological. As we will see in the next section, this amplitude comprises even more interesting
and pleasant features.
6. Duality mapping
After having established a number of topological amplitudes on the type IIA side at arbitrary
loop order, the question is to which order they are mapped on the heterotic side if the appropriate
duality mapping is applied. Indeed, we will find that most amplitudes, especially the 6d couplings
of [21] are mapped to higher loop orders, whose computation seems to be out of reach.
For simplicity, we will consider in this chapter also for the 4d amplitudes the decompactification limit (T 2 C), so that the duality map stays in 6 dimensions. This strategy has the
further advantage, that we do not have to deal with moduli insertions, but only with gravitons
and graviphotons. According to [26] the only relevant information in this case is the behavior of
the 6d string coupling constant gs and the metric G , which are given by
1
,
gsHET
gsIIA =

GIIA
(GHET )
.
(gsHET )2
(6.1)
(6.2)
For further convenience, we then compute the following building blocks of the amplitudes.
First, every amplitude contains the corresponding integration measure

6

det GIIA = gsHET
det GHET ,
since the determinant of the metric in 6 dimensions behaves like the 6th power of G . Furthermore, the behavior of powers of the Riemann tensor is given by
65
8 4

4
R|IIA
= gsHET R|HET
,
6 3

3
R|IIA
= gsHET R|HET
,
since the Riemann tensor scales in the same way as the metric and there are 8 (6) inverse metrics
necessary to contract all indices of four (three) Riemann tensors. Similarly, the behavior of the
graviphoton field strength reads:
2 2

2
= gsHET T|HET
.
T|IIA
Using the above relations, the mapping of the amplitudes of the previous chapters on the
heterotic side is given by:

(gsIIA )2g2
det GIIA (R 4 T 4g4 )|IIA = (gsHET )2g
det GHET (R 4 T 4g4 )|HET .
(6d)
This means, that the type II g-loop amplitude Fg (which we studied in Section 3) is
mapped to a (g
+ 1)-loop amplitude on the heteroticside.

(gsIIA )2g2
det GIIA (R 4 T 2g2 )|IIA = (gsHET )2
(1)
This entails that the type II amplitude Fg of Section 4 is mapped to a 2-loop amplitude on
the heterotic
side.

(gsIIA )2g2
det GIIA (R 3 T 2g2 )|IIA =
(2)
In other words, the trivially vanishing Fg found in Section 5.1 get mapped to a 1-loop
computation
in the heterotic theory.

IIA
2g
(gs )
det GIIA (R 4 T 2g2 )|IIA =
(3)
Finally, the Fg of Section 5.2 get mapped to 1-loop in the heterotic side.
Since higher order corrections than 1-loop calculations do not seem feasible and the amplitude
with R 3 was already found to be trivially vanishing in the type IIA theory, we will restrict
(3)
ourselves to the heterotic computation of the Fg , which will be performed in the next section.
7. Heterotic on T 6
The aim of this section isas advertisedto compute the heterotic dual Fg(3,HET) of the
(3)
amplitude Fg (5.11). To this end the right-moving sector has also to be taken into account,
which we have disregarded up to now because we considered it to behave in the same way as
the left moving one.16 The computation on the type II side revealed, that the amplitude is topological for the left and right sectors separately. This entails, that there are various possibilities
of how the left-moving computation can be completed by adding right-moving components.
For concreteness, let us complete the amplitudes in such a way, that the scattering occurs between two gravitons and two graviscalars (from a 4-dimensional point of view). Still there are
two different possibilities corresponding to type IIA and type IIB superstring theory. Since in
6 dimensions, type IIB is not dual to heterotic, in the following we will concentrate only on
type IIA.
16 This is only up to sign changes of the internal theory because of the chiralities in the type IIA theory.
66
7.1. Type IIA setting

1
z1
z2
0
0
+1
1
+1
+1
0
0
0
0
0
0
+1
1
1
1
0
0
0
0
1
1
z3
z4
+1
1
+1
1
0
0
0
0
0
0
+1
1
+1
1
0
0
0
0
0
0
Graviphoton
g1
g1
xi
yi
+ 12
12
+ 12
12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
12
+ 12
12
12
12
12
12
12
12
PCO
g+1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
0
0
+1
0
0
0
+1
0
0
0
+1
Field
Number
Position
Scalar
1
1
Graviton
In order to figure out the vertex combination for the heterotic side, especially the scalar insertions have to be precisely identified and the duality map has to be applied:
Scalars and gravitons.
In terms of 0-ghost picture vertex operators of the type II theory, the scalar and graviton
insertions read

2
X 3 ip2 2 3 X 3 ip2 2 3 eip2 X (z1 ),
3

2
X i p 2 2 3 X 3 i p 2 2 3 ei p2 X (z2 ),
2
2

1
X ip1 1 2 X
ip1 1 2 eip1 X (z3 ),
2

2
X i p 2 2 1 X 2 i p 2 2 1 ei p2 X (z4 ).
The following corresponding momenta and helicities can then be extracted:
Momentum
Helicity
0
0
+
+
+
0
0
0
0
+
0
0
p2
p 2
p1
p 2
h3,3
h3,3
h2,2
h1,
1
The last two vertices are self-dual gravitons of the 4d supergravity multiplet, which is still
true after the duality map. The first two insertions are however moduli of the compact torus.
Since the first part of these vertices without momenta reads X 3 X 3 , which corresponds
to (1, 1) form, it is clear that these graviscalars are the T 2 Khler moduli. Upon type IIA
heterotic duality, they are mapped to the dilaton on the heterotic side [26]. The corresponding
heterotic vertices therefore read
Field
67
Position Momentum Helicity Vertex
Dilaton
z1
z2
Graviton z3
z4
eip2 X2
(Z ip2 2 )Z
ei p2 X 2
(Z i p 2 2 )Z
2 eip1 X1
(X 2 ip1 1 2 )X
2
( X 1 i p 2 2 1 ) X 1 ei p2 X
h2,2
h1,
1
p2
p 2
p1
p 2
RR graviphotons
Label Number 1
g1
g1
+
+
+
+
+
+
xi
yi
The relevant part of the graviphoton vertex operator in 4 dimensions is given by

:e
+
2

a
p S a a b S b S S + Sa b S b S S eipZ :.
In terms of the helicity combinations, the spin fields Sa are given by

i (1 +2 +3 )
e2
e 2i (1 2 3 )
Sa =
e 2i (1 +2 3 ) ,
i
e 2 (1 2 +3 )
i (1 +2 +3 )
e 2
e 2i (1 +2 +3 )
Sa =
e 2i (1 2 +3 ) .
i
e 2 (1 +2 3 )
(7.1)
(7.2)
(7.3)
One can now analyze the two different vertices entering in the setting of the table:
vertex 1:
The matrix p ( )a b has to read
0
b 0
p
a
0
0
0
0
0
0
0
0
0
0
1
0
.
0
0
(7.4)
The relevant (reduced) Lorentz generators are those which have a non-vanishing (1, 4)
entry, namely:
0 0 0 1
0
0 0 1
1 0 0 1 0
i 0
0 1 0
03 =
02 =
(7.5)
,
,
0 1 0 0
0 1 0 0
2
2
1 0 0 0
1 0 0 0
0 0 0 1
0 0 0 1
i 0 0 1 0
1 0 0 1 0
12 =
(7.6)
13 =
.
0
1
0
0
2
2 0 1 0 0
1 0 0 0
1 0 0 0
68
We thus have the linear equation
0
0
02
03
12
02
a + b + c + d =
0
0
0
0
0
0
0
0
0
0
1
0
.
0
0
(7.7)
Solving for the coefficients (a, b, c, d), one finds:

1 02
+ i 03 + i 12 + i 13
p a b
2

1 20
+ i 30 + i 21 + i 31 ,
=
2
which, upon switching to the complex basis given in (2.4)(2.6) entails the two possibilities

(p 1 , 2 ),
(pA , B ) =
(7.8)
(p 2 , 1 ).
vertex 2:
Here the matrix p ( )a b has to read
0 0 0 0

0 0 0 0
p a b
.
0 0 0 0
1 0 0 0
(7.9)
The relevant (reduced) Lorentz generators are the same as in the case above, but the determining equation (7.7) changes to:
0 0 0 0
0 0 0 0
a 02 + b 03 + c 12 + d 02 =
(7.10)
.
0 0 0 0
1 0 0 0
Solving now for the coefficients (a, b, c, d) reveals

1 02
i 03 i 12 i 13
p a b
2

1 20
i 30 i 21 i 31 ,
=
2
which upon the same complexifications as above entails

(p1 , 2 ),
(pA , B ) =
(p2 , 1 ).
For concreteness we then choose the following momenta and helicities

1 1
(pA , B )
(p 1 , 2 )
.
=
2 , 2 )
(p2 , 1 )
(pA
B
(7.11)
(7.12)
Finally, under the R-symmetry SO(4) SU(2)L SU(2)R (from the 6d point of view which
we are using here to determine the map from IIA to heterotic), the graviphotons transform in
the (2, 2) representation. The specific vertices we have chosen correspond to graviphotons
69
carrying charge (+1/2, 1/2) with respect to the Cartan generators of the two SU(2)s. On
the heterotic side, the R-symmetry group SO(4) acts on the tangent space of T 4 . Let us
choose (real) coordinates Z on T 4 with = 6, 7, 8, 9 normalized so that their two point
functions

,
Z (z)Z (w) =
(7.13)
(z w)2
then the above choice of graviphotons entails choosing a complex direction say X 4 =
1 (Z 6 iZ 7 ) (see also (2.7)). The fermionic partner of X A (A takes now the values
2
5, 5)
will be denoted as before by A in the following.
4, 4,
Plugging the momenta and helicities of (7.12) into the 0-ghost picture vertex operator

B ipZ
e
,
V (0),A = B X A ip A X
(7.14)
the two different vertices appearing in the amplitude take the following form (see also [14])

2
VFA = X A i p 2 2 A X 1 ei p2 X ,
(7.15)

1
2 eip1 X .
VFA = X A ip1 1 A X
(7.16)
7.2. Contractions
We summarize the relevant vertex operators on the heterotic side:

eip2 X2 ,
V (p2 ) = Z ip2 2 Z

ei p2 X 2 ,
V (p 2 ) = Z i p 2 2 Z
2 ip X1

e 1 ,
Vh (p1 ) = X 2 ip1 1 2 X

1
2
Vh (p 2 ) = X i p 2 2 1 X 1 ei p2 X ,
2 ip X1

e 1 ,
VFA (p1 ) = X A ip1 1 A X

2
VFA (p 2 ) = X A i p 2 2 A X 1 ei p2 X .
The next question is which parts of the vertex operators have to be considered. This will be
determined below for each type of vertex separately:
gravitons: Since we are considering couplings of the Riemann tensor, there have to be two
derivatives from every vertex, implying that each graviton has to contribute two momenta.
Thus, only the following terms of the vertices do contribute to the amplitude

2 1 + ip1 X 1 + ,
Vh (p1 ) = X 2 ip1 1 2 X

Vh (p 2 ) = X 1 i p 2 2 1 X 1 1 + i p 2 X 2 + ;
graviphotons: Since the graviphoton vertex holds only one momentum, and s cannot contract among themselves, it is clear that only the following terms can contribute

2 1 + ip1 X 1 + ,
VFA (p1 ) = X A ip1 1 A X

VFA (p 2 ) = X A i p 2 2 A X 1 1 + i p 2 X 2 + ;
70
dilaton: Comparing with the computation on the type II side, it is obvious that there have to
be 2 momenta for each vertex. Furthermore, it is necessary that the fermionic part contributes
so that the spin structure sum gives a non-vanishing result:

1 + ip2 X 2 + ,
V (p2 ) = Z ip2 2 Z

1 + i p 2 X 2 + .
V (p 2 ) = Z i p 2 2 Z
Furthermore, from the trace part only the following term leads to non-vanishing contractions

1 1 + ip2 X 2 + ,
V (p2 ) = X 2 ip2 2 1 X

V (p 2 ) = X 2 i p 2 2 1 X 1 1 + i p 2 X 2 + .
Note that if X 1 and X 1 are replaced by X 2 and X 2 , respectively, in the two scalar vertices above,
2 )2 and (
X 2 )2 ) which yields
the contribution of right movers is reduced to a total derivative ((X
zero result. Thus, the above terms have been chosen in such a way that all contractions are a priori
non-vanishing.
7.3. Computation of the spacetime correlator
Taking into account the momentum structure of the above vertices, one finds

g1
(i) g1
(j )
Ag = Vh (p1 )Vh (p 2 )V (p2 )V (p 2 )
VF p1
VF p 2
i=1
(p1 )4 (p 2 )4
g1

j =1
(i) (j )
p1 p 2
2
(g + 1)! Fg(3,HET) .
(7.17)
i,j =1
The contractions of the above vertices lead to the following (rather formal) expression for the
amplitude17

g+1
2
g+1

d 1
1
(3,HET)
2
1 2
2
2 1
d xi X X (xi )
d yj X X (yj )
Fg
=
((g + 1)!)2
23 24 i=1
j =1

A 2g2 1 P 2 1 P 2
(7.18)
PL
q 2 L q 2 R ,
(PL ,PR ) (6,22)
where = 1 + i2 is the Teichmller parameter of the world-sheet torus with q = e2i , and
PL , PR are the left and right momenta of the (6,22) compactification lattice. Possible additional
constant prefactors were dropped since they are not relevant for our analysis. Note that X A s being the same complex field (e.g. 1 (Z 6 iZ 7 )) do not have any singular OPE among themselves
2
and therefore they contribute only through their zero modes. In this expression, still the (normalized) correlator of the spacetime fields X 1,2 has to be computed. To this end, the following
17 Similar as in [14], one can show that the contributions of even and odd spin structures are the same.
71
generating functional can be introduced:

2h
h
h

1
2
1 2
d
d 2 yj X 2 X 1 (yj )
x
X
(x
)
X
G(, , ) =
i
i
(h!)2 2
h=0
i=1
j =1

2
2 2
2
2
2
d w1 X X (w1 ) d w2 X X (w2 )
=
2h Gh (, ).
(7.19)
h=1
In [14], this generating functional was calculated and shown to be

2
2i 3 2
exp
,
G(, , ) =
2
1 (, )
where 1 is the usual odd theta-function defined by
1
1 2
1
1
q 2 (n 2 ) e2i(z 2 )(n 2 ) .
1 (z, ) =
(7.20)
(7.21)
nZ
It can be obtained from (C.7) by the choice (1 , 2 ) = ( 12 , 12 ). We recall here some properties
satisfied by Gg (, ):

a + b a + b
Gg
,
= 2g Gg (, ),
c + d c + d
i
(7.22)
Gg = 2 Gg1 .
22
With this spacetime correlator, the final expression for the amplitude takes the form
2

A 2g2 1 P 2 1 P 2
d 1 2g+2
(3,HET)
PL
=
G
q 2 L q 2 R .
Fg
g+1
2
3
24
2
(6,22)
(7.23)
(PL ,PR )
8. Harmonicity relation for type II compactified on K3

8.1. Holomorphicity for type II compactified on CY3
After having established different topological amplitudes in various configurations, we will
now look for the analog of the harmonicity relation discussed in [21,27], that generalizes holomorphicity of N = 2 F-terms involving vector multiplets. Before plunging into the computation
for the new N = 4 topological amplitudes on K3 T 2 , we first review the known results for
N = 2 compactifications on a CalabiYau threefold CY3 and for N = 4 on K3.
Type II theory compactified on a CY3 exhibits N = 2 supersymmetry and the relevant mul(4d)
tiplets are the 4d, N = 2 supergravity Weyl multiplet W and the (chiral) YangMills multiplet V . The former contains as bosonic components the graviton and the graviphoton (see
[28,29]),18

(4d)
W
= T + L R R + fermions,
(8.1)
18 In our superfield expressions, numerical factors are dropped.
72
while in a chiral basis, the later (which has a complex scalar and a vector-field as bosonic components) can be written as the following ultrashort superfield
V = (x) + L R F + fermions.
(8.2)
Note that V depends only on half of the Grassmann coordinates, namely only the chiral ones
L , R but not on the anti-chiral ones L , R .
As shown in [2], topological amplitudes at the g-loop level correspond to the coupling

(4d) 2 g
4
2
2
d x d L d R Fg (V ) W
(8.3)
= d 4 x R 2 T 2g2 Fg (x) + ,
where Fg is captured by the partition function of the N = 2 topological string and the dots stand
for additional terms arising from the superspace integral. The right-hand side exhibits the fact,
that Fg depends only on the scalars of the YangMills multiplets , but not on their complex
Navely, one therefore can write down the following holomorphicity equation
conjugates .
Fg = 0.
(8.4)
However in [3], it was shown that this equation is spoiled by the so-called holomorphic anomaly,
which stems from additional boundary contributions when taking the derivative in (8.4). From
the effective field theory point of view, the anomaly is due to the propagation of massless states
in the loops [2].
8.2. 6d N = 2 harmonicity equation
Recalling the topological amplitudes of Section 3 for type II string theory compactified on K3,
they correspond to the following N = 2 supersymmetric term of the six-dimensional effective
action:

g
6
4 1
d x d L d 4 R1 Fg(6d) q ij Wa1 b1 Wa2 b2 Wa3 b3 Wa4 b4 a1 a2 a3 a4 b1 b2 b3 b4 + ,
(8.5)
where the dots stand for possible additional terms needed for complete supersymmetrization,
following from integration over harmonic superspace (see discussion below and [30]). Here Wa b
is the 6d N = 2 Weyl multiplet already defined in (3.1) with the SU(2)L SU(2)R indices fixed
to i = 1, j = 1 [22], and we included a (a priori arbitrary) function of the 6d N = 2 matter
multiplet q ij with the following component expansion
q ij = ij + Li R + R Li + Li R + R Li

j +
j
+ Li R F
+ Li R F
+ ,
j
(8.6)
with the pair of indices (i, j ) transforming under SU(2)L SU(2)R and F + (F ) denoting the
(anti-) self-dual part of the field strength. The scalars ij satisfy the hermiticity relation
ij = ik j l kl .
(8.7)
In passing from the 4d case discussed in the previous section to the 6d one it is necessary to
introduce harmonic superspace in order to obtain a short version of the multiplet q ij [31]. Moreover, one has to restore the R-symmetry indices and rewrite (8.5) in a covariant way. Since the
73
R-symmetry group is
SU(2) SU(2),
(8.8)
the harmonic coordinates should parameterize the coset manifold

SU(2)L SU(2)R
(8.9)
.
U (1)L
U (1)R
The BPS character of the (covariantized) effective action term (8.5) should then restrict the
moduli dependence q ij of Fg(6d) s, generalizing the holomorphicity equation (8.4). Such an equation was found in Ref. [21] by covariantizing the corresponding string amplitudes (3.22) and
(3.34). This can be done by introducing two constant doublets uL and uR , under SU(2)L and
SU(2)R respectively, satisfying |uL |2 = |uR |2 = 1,
L
uL,i ij uL
j = ui ,
(8.10)
b
b 11
b ij L R
and similar for uR
i , and replace the Weyl superfields Wa (Wa ) in (8.5) by (Wa ) ui uj .
(6d)
The moduli dependent coefficient Fg is then promoted to a function of uL and uR , as well.

The moduli ij couple to the string world sheet action as:

+ +
+ G
+
SS+
G
22 G+ G
12

+
+
+
+
+ 21 G
(8.11)
G + 11 G
G .
+ (and their right moving counterparts) are dimension one operators.
In the twisted theory G+ , G
+ and using N = 4 superconformal algebra, it was
By deforming the contours of G+ and G
(6d)
then shown [21] that Fg satisfy the following equation modulo possible contact terms and
contributions from the boundaries of the world-sheet moduli space:
D
F (6d) = 0,
Dij g
uL
i
(8.12)
for fixed j , which by the hermiticity relation satisfied by can also be written as:
ik
D
Fg(6d) = 0.
L
ui D kj
(8.13)
Note that since the index j is free there are actually two equations which transform as a doublet
of SU(2)R (there is actually another doublet of equation transforming under SU(2)L obtained
by exchanging uL and uR ). It turned out however, as realized in [27], that besides boundary
anomalous contributions, there are contact terms that spoil each of the two equations for given j ,
and only a certain linear combination of the two which is SU(2) invariant is free from both
boundary and contact terms. The correct equation was found to be:
2

uR
k ij
i,j,k=1
d
D
F (6d) = 0.
g
D
duL
jk
i
(8.14)
There is also another equation obtained by exchanging uL and uR :

2

i,j,k=1
uL
k ij
d
D
Fg(6d) = 0.
R
dui D kj
(8.15)
74
9. Harmonicity relation for type II compactified on K3 T 2

9.1. Supergravity considerations
After having reviewed in the previous section the known computations and results, we now
turn to our main objective, which is to establish a similar equation to (8.12) or (8.14) for the compactification on K3 T 2 . The vector multiplets in the resulting 4-dimensional N = 4 theory are
ij
described by the field strength multiplet YA where i, j transform in the fundamental representation of the R-symmetry group SU(4) and index A labels the vector multiplet (in the context of
ij
string theory A = 1, . . . , 22). YA is antisymmetric and therefore transforms in the antisymmetric
6-dimensional representation of SU(4) and satisfies the reality condition
1
ij
YA = ij kl Ykl,A ,
(9.1)
Ykl,A = YAkl .
2
As in the 6d N = 2 case discussed before, one has to introduce harmonic superspace variables
ij
in order to shorten in particular the YA multiplet. In this case, the harmonic coordinates should
parametrize generically the coset manifold [32]:
G
SU(4)
.
H
U (1) U (1) U (1)
(9.2)
Here we have chosen H to be the maximal Abelian subgroup of G as in [31]. Hence, we have
a set of four harmonic coordinates uIi , I = 1, . . . , 4 (together with their conjugates uiI ), each
of SU(4), which differ by their charges with respect to the three
transforming as a 4 (and (4))
U (1)s:
(1,0,1)
u1i ui
(1,0,1)
u2i ui
(0,1,1)
u3i ui
(0,1,1)
u4i ui
(9.3)
They satisfy the conditions

j
uIi uI = i ,
uiI uJi = IJ ,
i1 i4 u1i1 u4i4 = 1,
(9.4)
where an implicit summation over repeated indices is understood. The derivatives DJI which
preserve this condition are given by

DII = 0.
DJI = uIi J uiJ i ,
(9.5)
ui
uI
I
These derivatives satisfy an SU(4) algebra. The diagonal ones, namely for I = J , modulo the
trace are just the charge operators of the subgroup H . The remaining generators for I < J and
their conjugates for I > J define the subalgebras L+ and L (corresponding to positive and
negative roots respectively) in the Cartan decomposition of G = H + L+ + L . The highest
weight in a given irreducible representation is defined by the condition (called H-analyticity
condition) that it is annihilated by L+ . One must further impose G-analyticity condition which
ensures that the superfield depends only on half of the Grassmann variables and hence gives a
short multiplet. The details can be found in [31]. The resulting superfield is19
ij
YA12 = u1i u2j A + 3 4 F+,,A + 1 2 F,,A + ,
19 We thank E. Sokatchev for showing us the precise form of the superfields Y and K.
(9.6)
75
where I and I are harmonic projected Grassmann coordinates I = uiI i and I = uIi i , F,
are self-dual and anti-self-dual parts of the gauge field strength and dots refer to terms involving
fermions and derivatives of bosonic fields. By choosing different Cartan decompositions (i.e.
different choices of positive roots) we can get different superfields Y I J but here we will focus on
(I J ) = (12).
In the present case, we also have the gravitational multiplet which comes with graviphotons
that transform in the 6-dimensional representation of SU(4). Unfortunately the shortening of the
gravitational multiplet using harmonic variables is not known at present. However, as we will see
below, we will be able to reproduce the string amplitudes constructed in this paper by postulating
the superfield

ij
12
K
= u1i u2j T+, + 3 4 R+, + 1,a 2,b ( )ab a a bb + + ,

j
12
kl
K
(9.7)
= ui3 u4 ij kl T,
+ 1 2 R, + 3a 4b ( )a b a a bb + ,
ij
where T are the self-dual and anti-self-dual parts of the graviphoton field strengths. More
I J (and K
I J ) obtained by different harmonic projections

generally there should be superfields K
ij
ij
whose lowest component will be given by uIi uJj T+, (and uIi uJj T, ) but here we will focus
on the effective action terms involving the fields with (I J ) = (12), in a way similar to the 6d case
[21]. In the absence of a rigorous N = 4 supergravity construction, Eq. (9.7) should be thought
of as a convenient way of summarizing the results of string amplitudes constructed in this paper.
Actually, the above expressions can be obtained by taking two supercovariant derivatives of the
I J
DI DJ W
I J = I J MN K

,MN .
4d N = 4 Weyl multiplet WN=4 [29]: K
N=4 and K
Using these superfields we can write down the following couplings:

12 ,12 g 12 ,12
K K
,
K
S1 = d 4 x d 2 3 d 2 4 d 2 1 d 2 2 Fg(1) K
(9.8)

12 ,12 g+1
S3 = d 4 x d 2 3 d 2 4 d 2 1 d 2 2 Fg(3) K
(9.9)
K
,
(1)
(3)
where Fg and Fg depend on the vector multiplets and harmonic variables.20 Whether such
terms give rise to consistent supergravity couplings is not clear at present and needs to be understood. By performing now the integration over the Grassmann coordinates i , one finds that the
two expressions (9.8) and (9.9) give rise to the following effective action terms

2 2 2g2
2g2
R T+
+ (+ )2 ( )2 T+
S1 = d 4 x Fg(1) R+
2g2
+ ,
+ (+ )( )R+ R T+

2g2
2
S3 = d 4 x Fg(3) R+
(+ )2 T+
+ .
(9.10)
It follows that the two couplings in (9.8) and (9.9) correspond to the two classes of topological
(1)
(3)
amplitudes Fg and Fg discussed in Sections 4 and 5, respectively. They reproduce in particu20 Here by a slight abuse of notation we are using the same symbols as the ones appearing in the topological string
(1)
(3)
amplitudes, however we will see shortly the relation Fg and Fg appearing in the above equations to the corresponding
quantities appearing in the string amplitudes.
76

2g2
2g2
2 R2 T
2 (dd )2 T
lar the corresponding amplitudes R+
at genus g and R+
at genus g + 1
+
+
+
found above.
In order to find the corresponding generalized harmonicity equation, we will follow the same
procedure as before by going back to string theory and covariantize the respective string amplitudes under SU(4). As mentioned earlier there are several different projections labelled by (I J ).
ij
In order to pick out (I J ) = (12) we can choose the graviphoton field strength T+, to be proj]
portional to u[i1 u2 times some self-dual field strength. Due to the orthogonality relations (9.4)
12 and the resulting amplitudes will be
this will precisely pick out the lowest component of K
(1)
(3)
given by Fg and Fg appearing in (9.8) and (9.9) respectively. On the other hand, choosing
such graviphoton field strength means that the vertex operators for the graviphoton field strengths
j]
in the string computations should be folded with u[i1 u2 . This will render the topological string
(1)
amplitudes a dependence on the harmonic variables and it is those quantities that compute Fg
(3)
and Fg appearing in (9.8) and (9.9). In the next section we will show, by going to the heterotic
duals of these amplitudes, that Fg(3) satisfies the equation:
ij kl Di1 Dj2
D
F (3) = 0,
D kl g
(9.11)
up to an anomaly boundary term, very much in analogy to the N = 2 Fg s. Here the derivatives
DiI are defined as:
DiI uJi DJI = uJi uIk
i
uJk
uI
(9.12)
and we have relaxed the determinant condition and the related trace condition in Eqs. (9.4) and
(9.5). This can always be done by introducing another variable say and let uIi uIi and
uiI uiI /. One can check that as a result [DiI , DjJ ] = 0. Note that acting on functions only of
ui1 and ui2 , as will be the case in the following, only the second term on the right-hand side of
Eq. (9.12) will be relevant.
Eq. (9.11) is SU(4) invariant analogous to the correct equations (8.14) and (8.15) in the
6-dimensional case, which were singlets under the corresponding R-symmetry group SU(2)L
SU(2)R . It turns out that there are stronger harmonicity equations with only one harmonic derivative (Di1 or Di2 ), that transform covariantly under SU(4) and are similar to the six-dimensional
ones discussed in the previous section. However, here, we will restrict ourselves to the weaker
(3)
version (9.11).21 Moreover, by considering the decompactification limit in six dimensions, Fg
(6d)
is mapped to a semi-topological quantity similar to Fg , that satisfies the same (6d) harmonicity equations as (8.14) and (8.15). This supports the claim that all such Eqs. (8.14), (8.15) and
(9.11) should be a consequence of the BPS character of these couplings generalizing the N = 2
holomorphicity to extended supersymmetries.
21 We thank E. Sokatchev for pointing this out to us. The stronger version of the harmonicity equation will appear
elsewhere [33].
77
9.2. Further investigation

The next logical step would be to test this relation using directly the Fg s computed in Sections 4 and 5 for type II theory and by this procedure determine a possible harmonicity equation.
However in doing so, one encounters several problems:
The topological expression cannot be used, because some of the operator insertions necessary for testing (9.11) are from the RR sector,22 for which no representation in terms of the
superconformal algebra exists. One could therefore try to check the equation only perturbatively in RR field insertions.
The corresponding correlation functions with additional RR insertions lead to rather difficult
expressions which are hard to compute.
For these reasons, we change our strategy and switch to the heterotic side. As we have shown,
(3)
the Fg can be computed upon duality mapping by a one-loop torus amplitude on the heterotic
side, which is easy to obtain. In this way, the harmonicity relation can be tested and we can also
verify whether it contains any anomalous terms.
10. Harmonicity relation for heterotic on T 6
10.1. Derivation of the harmonicity relation
(3,HET)
To discuss the harmonicity relation on the heterotic side we first need to express the Fg
in an SO(6) SU(4) covariant way. In computing Fg(3,HET) we used a particular combination
of graviphotons
which was labeled say by G, with X G appearing in the vertex being (Z 6
7
iZ )/ 2. Since Z 4 , Z 5 , . . . , Z 9 transform as vector of SO(6), this particular choice of the
graviphoton is not SO(6) invariant. Using the fact that the vector of SO(6) can be expressed as
antisymmetric tensor product of two SU(4) fundamentals, we can define
Xij = Z (C )ij ,
(10.1)
where are the (4 4) part of the SO(6) gamma matrices acting on the chiral spinor (i.e.
SU(4) fundamental representation 4) and C is the charge conjugation matrix such that (C )T =
C . It is easy to see that Xij are complex and satisfy the reality condition Xij = 12 ij kl Xkl .
Furthermore we normalize them so that their 2-point function is

2 ij kl
.
Xij (z)Xkl (w) =
(z w)2
(10.2)
In real notation, this essentially means that X ij = Z + iZ for some and different from
each other, where Z are canonically normalized so that their kinetic term comes with the factor
1/2. As a result the zero mode part of the Xij is just given by 2PijL and the lattice part of the
1
left-moving partition function is q 2 PL where (P L )2 = 18 ij kl PijL PklL .

22 Recall that in type II compactified on K3 T 2 only SU(2) SU(2) U (1) subgroup of the R-symmetry group
L
R
SU(4) is manifest where U (1) is the rotation group acting on the tangent space of T 2 . In particular the full SU(4) group
mixes NSNS and RR sectors. Since Eq. (9.11) is SU(4) invariant it involves deforming by both NSNS and RR
moduli.
78

12 we must choose the graviphoton
As explained at the end of Section 8, in order to pick K
j]
field strength to be proportional to u[i1 u2 . This means that the graviphoton vertex must be folded
j]
with u[i1 u2 :

1
j
eip.Z (z, z ),
d 2 z ui1 u2 (Xij ip . ij )Z
VF (u1 , u2 ) =
(10.3)
2
(3,HET)
and define Fg
(u1 , u2 ) as in (7.17) with all the 2g 2 VF replaced by VF (u1 , u2 ). An
important point to notice is that the singular term in the OPE is
j

2ui u uk ul ij kl
j
= 0.
ui1 u2 Xij (z)uk1 ul2 Xkl (w) = 1 2 1 2 2
(z w)
(10.4)
This implies that Xij are replaced by their lattice momenta 2PijL inside the correlation function. The final result can be read off from Eq. (7.23)

Fg(3,HET) (u1 , u2 ) =
d 2 1 2g+2
Gg+1
23 24 2
ui1 u2 PijL
2g2
q 2 (P
L )2
(P L ,P R ) (6,22)
q 2 (P
R )2
(10.5)
where JA
The marginal deformations are given by the (1, 1) operators
are the right-moving currents for A = 1, . . . , 22 normalized so that the 2-point function satisfies
1
Xij JA ,
Vij,A = 2

JA (z)JB (w)
=
AB
.
(z w)
2
(10.6)
Let us denote by Dij,A the covariant derivative with respect to moduli which corresponds to
inserting the marginal operator Vij,A in various correlation functions. It is easy to see that the
resulting variation in the lattice momenta are given by
Dij,A PklL = ij kl PAR ,
Dij,A PBR = AB PijL .
(10.7)
This definition of the covariant derivative naturally follows from the insertion of the marginal
operators as we will see below. However what this covariant derivative precisely means geometrically in N = 4 supergravity needs to be understood.
(3,HET)
The resulting variation in the topological amplitude Fg
(u, v) is given by
ij kl Dkl,A Fg(3,HET) (u1 , u2 )
2
d 1 2g+2
Gg+1
=2
23 24 2
u1 P L u2
2g3
(P L ,P R ) (6,22)

1
i
L 2 1
R 2
j
j
( ) ij kl P L kl u1 P L u2 PAR q 2 (P ) q 2 (P ) ,
(2g 2) ui1 u2 u1 ui2 +
2
(10.8)
where (u1 P L u2 ) uk1 ul2 PklL . This equation can also be obtained directly by inserting the marginal operator Vij,M in the definition of Fg(3,HET) (u, v) and using the following formula for the
correlation function:
(Xkl JA )(w, w)
2g2

79
u1 X(za )u2
a=1

2g2
= (2)2g PklL PAR u1 P L u2
+2
2g2

2g3 R

n
2g2
w2 log 1 (w za ) klmn um
u1 P L u2
PA .
1 u2 (2)
(10.9)
a=1
Now we integrate w in the second term by partial integration using the formula
!

"

Im(w za )
2
2
2
d w w log 1 (w za ) = d w w w log 1 (w za ) + 2i
2
2
= ,
(10.10)
a)
) is
where in the second equality we have used the fact that (w log 1 (w za ) + 2i Im(wz
2
periodic on the torus. Using Eqs. (10.10) and (10.9) we obtain Eq. (10.8).
Now we can apply the derivative operators Di1 and Dj2 defined in Eq. (9.12) with the result

2g3 i j

2g3
j
Dj2 u1 P L u2
u1 u2 u1 ui2 = 2gui1 u1 P L u2
,

2g3
2g3
j
j
Di1 Dj2 u1 P L u2
ui1 u2 u1 ui2 = 2g(2g + 1) u1 P L u2
,

2g2
2g3
L
Dj2 ij kl PklL u1 P L u2
= (2g 2) ij kl PklL u1 P L u2
um
1 Pmj ,

2g2
Di1 Dj2 ij kl PklL u1 P L u2

2g3
2g4 ij kl L m L L n

= (2g 2) 8PL2 u1 P L u2
+ (2g 3) u1 P L u2
Pkl u1 Pmj Pin u2

2g3
2
= 2(2g 2)(2g + 1) u1 P L u2
(10.11)
PL ,
where in the last equality we have used the relation

L n

2
L
L
ij kl PklL um
1 Pmj Pin u2 = 2 u1 P u2 PL ,
(10.12)
n
which can be proven by comparing the coefficients of um
1 u2 on both sides. Using Eqs. (10.11) in
(10.8) we obtain
ij kl Di1 Dj2 Dkl,A Fg(3,HET) (u1 , u2 )

1
= 4i(2g 2)(2g + 1) d 2 24 Gg+1

2g3 R 1 (P L )2 1 (P R )2
2g
u1 P L u2
PA q 2
q 2
2
.
L R
(6,22)
(10.13)
(P ,P )
Note that apart from the factor Gg+1 , the total derivative with respect to can also be understood from the requirement of world-sheet modular invariance. We can now carry out a partial
integration with respect to . Since the boundary term is modular invariant as follows from the
second equation of (7.22) and in the infrared limit 2 there is exponential suppression due
to the presence of PL (for g > 1) we conclude that boundary terms vanish. The only contribution
therefore comes when the -derivative acts on Gg+1 . Using the second equation of (7.22) we
get:
80
ij kl Di1 Dj2 Dkl,A Fg(3,HET) (u1 , u2 )

1
2g2
= 2(2g 2)(2g + 1) d 2 24 Gg 2

2g3 R 1 (P L )2 1 (P R )2
u1 P L u2
PA q 2
q 2
.
(10.14)
(P L ,P R ) (6,22)
By using Eq. (10.8) we can rewrite the right-hand side of the above equation to obtain the following recursion relation:
j
(3,HET)
ij kl Di1 Dj2 Dkl,A Fg(3,HET) (u1 , u2 ) = (2g 2)(2g + 1)ui1 u2 Dij,A Fg1
(u1 , u2 ).
(10.15)
(3,HET)
This equation is satisfied also for g = 1 but in a trivial way since by construction F1
does not depend on u and v. As a result, the left-hand side vanishes and similarly the righthand side is zero due to the presence of the (2g 2) factor. However, there may be a non-trivial
(3,HET)
analogous to the t t equation in the N = 2 case, but we will not
equation satisfied by F1
attempt to analyze it here.
(3,HET)
with a lower value of g suggests that on
The fact that the right-hand side involves Fg
the type II side there must be boundary contributions coming from degeneration limits of the
Riemann surface, as in the holomorphicity equation (8.4) of the N = 2 Fg s. It will be interesting
to study the harmonicity condition in the type II side and check whether that relation is mapped to
the heterotic relation proven above under the duality map in the appropriate weak coupling limit.
However, as mentioned above, studying such relation in the type II side involves introducing RR
moduli fields which complete the SO(6) representation of the K3 moduli. Even though turning
on RR moduli in the Polyakov action is difficult, one may study this relation in the first order
perturbation with respect to RR moduli.
10.2. Decompactification to T 4 and connection to the harmonicity relation for type II on K3
Focusing on the heterotic 1-loop expression on T 6

Fg(3,HET) =
2g+2
2g2 1 (P L )2 1 (P R )2

d 2 2
Gg+1 (, ) u1 P L u2
q2
q 2
,
3
24
2
(10.16)
we want to study the behavior of the harmonicity equation (10.15) in the decompactification limit
for one of the T 2 . If harmonicity equations are a consequence of supersymmetry, in this limit,
it is expected that (8.14) and (8.15) are recovered. More precisely, by heterotictype II duality
(3,HET)
(3)
is mapped to the T 2 decompactification limit of Fg . From its
studied in Section 6, Fg
explicit form computed in Section 5.2, the resulting expression is not anymore topological on
K3, since there are leftover det(Im ) factors from the non-compact spacetime coordinates.
Indeed, from Eq. (5.10), in the limit where the torus coordinate X3 is non-compact, the X3 from
the left-moving sector and X 3 from the right-movers give rise to contact terms X3 X 3
b , where a, b = 1, . . . , g + 1 where g + 1 is the genus of the surface and a are the
a (Im )1
ab
normalized holomorphic Abelian differentials. This leads to:

Fg(3)
Mg+1
81

g+1

1
i ai bi (Im )1
i ci di
a1 ag+1 d1 dg+1
bi ci
det(Im )
i=1
2
3g1

GK3 (a )JK3 (3g ) ,

(10.17)
a=g+2
where a complete antisymmetrization of the Beltramis is understood. Despite its non-topological

nature, this quantity still satisfies the 6d harmonicity equations (8.14) and (8.15), as we show
below by considering the decompactification limit of the heterotic expression (10.16).
In (10.16), P L are the lattice vectors of a (6,22) . If we split T 6 into T 4 T 2 , this lattice gets
broken to
(6,22) (4,20) (2,2) .
(10.18)
(6,22)
to be entirely in (2,2)
Let us further denote by P13 and its complex conjugate P24 of
and the remaining four P12 , P14 and their complex conjugates P34 , P23 to be in (4,20) . SU(4)
now splits into SU(2)L SU(2)R where SU(2)L acts on the indices (1, 3) and SU(2)R on the
indices (2, 4), respectively. In the decompactification limit P13 and P24 decouple, so the relevant
part of the SU(4) harmonics uiI can be assembled into the harmonics of SU(2)L SU(2)R with
the identification ui1 , ui3 v1a , v2a and ui2 , ui4 v1a , v2a , where we have used dotted and undotted
indices a, a = 1, 2 to indicate that they are respectively SU(2)R and SU(2)L harmonics. In this
notation, (4,20) lattice vectors are denoted by:
(4,20),L
Pa b
(4,20),R
PA
with a, b = 1, 2,
(10.19)
and now the vector multiplet label A = 1, . . . , 20. Moreover

(4,20),L 2 1 (4,20),L (4,20),L a a b b
P
Pa b
Pa b
1 2 1 2.
2 2
2 1 1
The SU(2)L SU(2)R harmonics satisfy:
v1a = ab v1,b = va1 = v 2a ,
2a
1
v1a = a b v1,
b = va = v ,
as well as |v1 |2 = |v2 |2 = |v1 |2 = |v2 |2 = 1, where |v1 |2 = v1a va1 , etc.
Expression (10.16) in the decompactification limit of T 2 then goes into23
2 2g+2

a L b 2g2 1 P 2 1 P 2
d 2
(3,HET)
v1 Pa b v1
Fg
G
(,
)
q 2 L q 2 R .
g+1
4
24
2
(4,20)
(10.20)
(10.21)
(10.22)
(PL ,PR )
Introducing the differential operator similar to (9.12)

2
2
2

I
J I
J
I
b
v a DJ
va vb J uJ b =
vaJ vbI J a ,
Da =
v
vb
vI
vb
I
J =1
J =1
J =1
(10.23)
where we have used

2

vaI vIb = ab .
I =1
23 The (2,2) contribution leads to a 1 factor, together with the T 2 volume which is dropped.
2
(10.24)
82
involves only v1a and v1a , Da1 reduces essentially to the second term on
(3,HET)
Note that since Fg
the right-hand side namely v a . Now we can apply the operator

I
D
ab Da1
DbAa
(3,HET)
onto Fg
ab Da1
(10.25)
, with the following result:

D
DbAa
Fg(3,HET)
(2g 2) a b v1b

2g+2
d 2 2
Gg+1 (, )
24 24

2g3 1 P 2 1 P 2
(2g 1) 22 PL2 PAR v1c PcLc v1c
q 2 L q 2 R
(PL ,PR ) (4,20)
= (2g 2) a b v1b
2g1
2
24

d 2 Gg+1 (, )

PAR
2g3 1 P 2 1 P 2
v1c PcLc v1c
q 2 L q 2 R
(PL ,PR ) (4,20)

D
(3,HET)
(2g 2) a b v1b v1c v1c
Fg1 .
Dcc
(10.26)
Note that the last expression is again an anomalous term coming from a lower genus contribution. Multiplying now by v1a and summing over a we see that the right-hand side vanishes. This
equation resembles exactly the harmonicity relation (8.14) of [27], without any anomaly. Moreover, by exchanging v1 with v1 , one finds also a second equation which has precisely the same
form as (8.15).
11. Conclusions
In summary, we have studied N = 4 topological amplitudes in string theory. Unlike the N = 2
case of the topological CalabiYau -model, the N = 4 topological partition function vanishes
trivially and one has to consider multiple correlation functions. In the critical case of K3
-model, the definition is essentially unique and the number of additional vertex insertions increases with the genus g of the Riemann surface [21]. Furthermore, it computes a physical string
amplitude in the 6d effective action with four gravitons and (4g 4) graviphotons, of the type
R 4 T 4g4 . However, we found that contrary to the N = 2 case, on the heterotic side these amplitudes are not simplified because they start receiving contributions at higher loops (g + 1). On
the other hand, it turns out that on K3 T 2 one has a lot of freedom. In this work, we made a
systematic study of possible definitions with the following properties: (1) they are economic,
involving a minimum number of additional vertices; (2) they compute some (gravitational) couplings in the low energy 4d string effective action; (3) they have interesting heterotic duals that
allow their study by explicit computations.
Indeed, we found two such non-trivial definitions. The first uses two additional integrated
vertices of the U (1) currents of K3 and T 2 , JK3 and JT 2 (in left and right movers separately)
2 R 2 T 2g2 (and its
and computes the physical coupling of the 4d string effective action term R+
+
83
supersymmetric completion), where the +/ indices refer to self-dual/anti-self-dual parts of the

corresponding field-strength. Unfortunately, on the heterotic side, they start receiving contributions at two loops. In the second definition, the T 2 integrated current JT 2 = 3 3 is replaced
by just the fermionic coordinate 3 of dimension zero in the topological theory. Taking into
account the right-moving sector, this amounts to differentiation with respect to the T 2 Khler
modulus, which is needed for obtaining a non-vanishing result. It turns out that this is the min2 (dd )2 T 2g2 , with a
imal definition that computes the physical coupling of the term R+
+
+
+
KK graviscalar corresponding to T 2 Khler modulus. In the heterotic theory compactified on T 6 ,
+ is mapped to the dilaton and this amplitude is mapped at one loop in the appropriate limit,
and thus, it can be studied as the N = 2 Fg s.
We also studied the dependence of the above couplings, that we call N = 4 Fg , on the compactification moduli which belong to 4d N = 4 supermultiplets. Being BPS-type, they satisfy a
harmonicity equation upon introducing appropriate harmonic superspace variables, that generalizes holomorphicity of N = 2 Fg s and the harmonicity of the 6d case found in [21,22,27]. We
derived this equation on the heterotic side, where the full SU(4) R-symmetry is manifest, and
we uncovered an anomaly due to boundary contributions, analog to the holomorphic anomaly
of N = 2. It will be interesting to further study this equation on the type II side, compute ex(3)
plicit examples of the minimal N = 4 Fg s and work out possible applications of the N = 4
topological amplitudes, such as in the entropy of N = 4 black holes.
Harmonicity equation in the topological theory side is essentially a statement of decoupling
of the BRS exact states although in the N = 4 theory it appears in a more complicated way.
In the untwisted theory this BRS operator is translated to certain spacetime supersymmetry
generators (this statement can be made precise at least in N = 2 theories). Therefore decoupling
of BRS exact states should imply that the corresponding terms in the effective action must be BPS
operators. Therefore this equation must be a consequence of supersymmetry and the shortening
of the multiplets. An interesting open question is to understand what this harmonicity equation
precisely means in the context of N = 4 supergravity.
Acknowledgements
We thank S. Ferrara, W. Lerche, T. Maillard, M. Mario, M. Weiss and especially E. Sokatchev
for enlightening discussions. This work was supported in part by the European Commission under the RTN contract MRTN-CT-2004-503369 and in part by the INTAS contract 03-51-6346.
The work of S.H. was supported by the Austrian Bundesministerium fr Bildung, Wissenschaft
und Kultur.
Appendix A. Gamma matrices and Lorentz generators
In any number of spacetime dimensions, the Lorentz generators are defined as commutators
of the -matrices
b
i
a = , a b .
(A.1)
4
Upon choosing a suitable basis, the Lorentz generators can be brought into diagonal form, comprising the used in Section 3
b

b
( )a
0
a =
(A.2)
.
0
( )a b
84
The superspace integrals in Section 3 involve traces over the reduced , which are related to
traces over the full using the following identities
tr = tr = 0,
(A.3)

dspin

tr
(A.4)
= 2 tr
=
,
4

tr 1 1 2 2 3 3 4 4 = tr 1 1 2 2 3 3 4 4 + tr 4 4 3 3 2 2 1 1 ,
(A.5)
with24
= diag(1, . . . , 1),
(A.6)
and dspin the spinor dimension, which in the case of 6 spacetime dimensions is 8.
For completeness we also list all the
01 =
03 =
05 =
13 =
15 =
24 =
34 =
1 0 0
0
1 0 1 0
0
,
0
0
1
0
2
0 0 0 1
0
0 0 i
1 0
0 i 0
,
0 i 0 0
2
i 0 0 0
0 0 i 0
1 0 0
0 i
,
0 0
2 i 0
0 i 0 0
0 0 0 1
1 0 0 1 0
,
2 0 1 0 0
1 0 0 0
0 0 1 0
1 0 0 0 1
,
2 1 0 0 0
0 1 0 0
0 i 0 0
1 i 0 0 0
,
0 0 0 i
2
0 0 i 0
0 1 0
0
1 1 0 0
0
,
2 0 0 0 1
0 0 1 0
02 =
04 =
12 =
14 =
23 =
25 =
35 =
0 0 0 1
1 0 0 1 0
,
0 1 0 0
2
1 0 0 0
0 0 1 0
1 0 0 0 1
,
2 1 0 0 0
0 1 0 0
0 0 0 i
1 0 0 i 0
,
0 i 0 0
2
i 0 0 0
0 0 i 0
1 0 0 0 i
,
0
2 i 0 0
0 i 0
0
1 0 0 0
1 0 1 0 0
,
2 0 0 1 0
0 0 0 1
0 1 0 0
1 1 0 0 0
,
2 0 0 0 1
0 0 1 0
0 i 0 0
1i 0 0 0
,
2 0 0 0 i
0 0 i 0
24 Note, that the transition to Minkowski space is easily achieved by replacing 0 i 0 .
1
1 0
45
=
2 0
0
0
1
0
0
0
0
1
0
85
0
0
.
0
1
Appendix B. Generalized self-duality in 6 dimensions

The standard notion of (anti-)self-duality
F = F,
or
F = F,
(B.1)
does not make sense in 6 dimensions. However, there is another possibility to generalize the
notion of self-duality. Considering for example the vertex of an RR field in the (1/2)-ghost
picture (3.14)
#
(

1
(1/2)
z ) eipZ :,
z ) + Sa a S b S S(z,
(p, ) = :e 2 (+)
p S a a b Sb S S(z,
VT
b
%&
'
$
%&
' $
anti-self-dual
self-dual
(B.2)
the two terms corresponding by definition to self-dual or anti-self-dual terms of the field strength
tensor. This entails that it is possible to project to (anti-) self-dual parts of a 2-index antisymmetric
tensor by just contracting with ( )a b (( )a b ):

F + a b a b F ,
a

a
F b b F .
(B.3)
(B.4)
Since the Lorentz-generators can be defined in any number of dimensions, it is clear, that this notion of self-duality applies also in 6 dimensions. Note however that although the 4-dimensional
( )a b and ( )a b project to linearly independent sub-spaces, this is no longer true in 6 dimensions.
Appendix C. Theta functions and Riemann identity
C.1. Theta functions, spin structures and prime forms
In this appendix, we basically review Chapter 3 of [25]. Consider a compact Riemann surface
of genus g and choose a canonical basis of homology cycles ai , bi according to Fig. 3.
Furthermore, there is a normalized basis of holomorphic 1-forms i , with i = 1, . . . , g, whose
integrals along the homology cycles read
Fig. 3. A basis of homology cycles on a genus g compact Riemann surface .
86
j = ij ,
j = ij ,
ai
(C.1)
bi
where ij is a complex, symmetric g g matrix which is called period matrix.

One can then associate coordinates with points on the Riemann surface. For this, we choose
a base point P0 and define for each point P the Jacobi map25
P
I : P zi (P ) =
(C.2)
i .
P0
In this respect, z is an element of the complex torus

J (Mg ) = Cg / Zg + Zg .
(C.3)
One can now define the Riemann theta function [24] for z J (Mg )

(z, ) =
eini ij nj +2ini zi .
(C.4)
nZg
Shifting z by a vector of the lattice Zg + Zg , the Riemann theta function transforms as

(z + n + m, ) = ein n2imz (z, ).
(C.5)
We also use the Riemann vanishing theorem: There exists a divisor class of degree g 1 such
that (z, ) = 0 if and only if there are g 1 points p1 , . . . , pg1 on Mg such that
z=
g1

(C.6)
pi .
i=1
Furthermore, to every spin structure , one can associate a theta function with characteristics
(1 , 2 ) ( 12 Zg /Zg ) defined by
(z, ) = ei1 1 +2i1 (z+2 ) (z + 1 + 2 , ).
(C.7)
Besides -functions, the prime forms E(x, y) enter also in our computations. One can view
the prime form as a generalization of the holomorphic function x y on the Riemann sphere.
The precise definition is
E(x, y) =
f (x, y)
h (x)h (y)
(C.8)
where h is a holomorphic 12 -differential and f is given by

y
f (x, y) =
(C.9)
.
x
One can show, that E is independent of . Moreover, it is antisymmetric in x, y exchange and

has a simple root at x = y:
E(x, y) = E(y, x),
and
lim E(x, y) = O(x y).
xy
(C.10)
25 In computations, for notational simplicity, we drop the distinction between the point on the Riemann surface and the
corresponding Jacobi map.
87
C.2. Riemann addition theorem

The sum over all different spin structures for the product of four -functions gives:

s (a)s (b)s (c)s (d)
s

a+b+c+d
a b + c + d
=
2
2
$
$
%&
'
%&
'

=T++++
=T++

a + b c + d
a + b + c d
,
2
2
$
$
%&
'
%&
'
=T++
(C.11)
=T++
where we omitted overall phase factors which are irrelevant for our computations.
Appendix D. The N = 4 superconformal algebra
In this appendix, we review the N = 4 superconformal algebra and explain our notation. We
examine both four- and six-dimensional compactifications, or equivalently operators on K3 and
K3 T 2 . We remind the reader however, that for the sake of simplicity of our calculations, we
treat K3 only at its orbifold (free field) realization.
D.1. Superconformal algebra for compactifications on K3
The starting point is the spin-field of the internal theory which after bosonization of the fermionic coordinates A = eiA , can be written with the help of two free 2d scalar fields
i
S = e 2 (4 +5 ) ,
S e 2 (4 +5 ) .
(D.1)
According to [34] the current algebra (conformal dimension 1 operators) of the SCFT can be
uncovered by considering OPE of the spin fields among themselves:
S(z)S (w)
1
(z w)
++
(w)
JK3
3
4
+ (z w) 4 JK3 + ,
+ ,
1
(z w) 4
J (w)
+ ,
S (z)S (w) K3
1
(z w) 4
where the currents are given by
S(z)S(w)
JK3 = 4 4 + 5 5 ,
++
JK3
= 4 5 ,
(D.2)
(D.3)
(D.4)
JK3
= 4 5
(D.5)
JK3 JK3
= 2JK3
.
(D.6)
and satisfy an SU(2) current algebra

++
JK3 JK3 ,
JK3
++
++
JK3 JK3
2JK3
,
On the other hand, the N = 1 world-sheet supercurrent is given by
TF,K3 = 4 X 4 + 5 X 5 + 4 X4 + 5 X5 = G+
K3 + GK3 ,
88
where G
K3 are the two (complex conjugate) supercurrents of the internal N = 2 SCFT:
G+
K3 = 4 X4 + 5 X5 ,
G = 4 X4 + 5 X5 .
(D.7)
(D.8)
K3
G
K3
Indeed,
and the U (1) current JK3 form an N = 2 superconformal algebra with the energy
momentum tensor
1
TB,K3 = X4 X 4 + X5 X 5 + (4 4 + 5 5 ).
2
In order to generate the N = 4 superconformal algebra, one uses the SU(2) currents (D.6)
acting into (D.7) and (D.8) to find two additional supercurrents:
+ = 5 X4 + 4 X5 ,
G
K3
G = 5 X 4 + 4 X 5 .
(D.9)
(D.10)
K3
+
Note that the four supercurrents form two SU(2) doublets (G+
K3 , GK3 ) and (GK3 , GK3 ), while
the superscripts denote 1 unit of U (1) charge. Schematically, one can move within one doublet
with the help of the SU(2) currents JK3
++
JK3
G+
K3
G
K3
JK3
JK3
G
K3
+
G
K3
++
JK3
The OPE among the supercurrents read

G+
K3
JK3
z2
G
K3
B 1 J
TK3
K3
2
z
2JK3
z2
JzK3
2 +
G
K3
JK3
z
B 1 J
TK3
K3
2
z
+
G
K3
Summarizing, the full N = 4 superconformal algebra is given by the following OPEs:
++
(z)JK3
(0)
JK3
JK3 (0)
,
z
(0)
G
K3
,
z
+ (0)
G
++
K3
JK3
(z)G
(0)
,
K3
z
JK3
(z)G
K3 (0) 0,
(0) 0,
J (z)G
G+
K3 (0)
,
z
G (0)
+ (0) K3 ,
(z)G
JK3
K3
z
++
JK3 (z)G+
(0)
0,
K3
++
J (z)G+ (0) 0,
G+
K3 (z)GK3 (0) 0,
+
G
K3 (z)GK3 (0) 0,
(z)G+ (0)
JK3
K3
K3
++
(0)
(z)G
JK3
K3
K3
K3
89
B (0) 1 J (0)
JK3 (0) TK3
K3
2
,
z
z2
T B (0) 12 JK3 (0)
(0) JK3 (0) K3
+ (z)G
G
,
K3
K3
z
z2
++
2J ++ (0) JK3
(0)
+ (z)G+ (0) K3
+
,
G
K3
K3
2
z
z
2J (0) JK3
(0)
(z)G (0) K3
+
,
G
K3
K3
2
z
z
G+
K3 (z)GK3 (0)
while for any operator OK3 of U (1) charge q, one has:

q
OK3 (0)
q
.
JK3 (z)OK3 (0) q
z
D.2. Superconformal algebra for compactifications on K3 T 2

Following the procedure of the N = 4 case in 6 dimensions (D.2)(D.4), the starting point is
given by the internal spin-fields. In the case of K3 T 2 , the number of spin-fields is doubled
i
S1 = e 2 (3 +4 +5 ) ,
S2 = e
i
2 (3 4 5 )
S1 = e 2 (3 +4 +5 ) ,
S2 = e
2i (3 4 5 )
(D.11)
(D.12)
The conformal dimension 1 generators are again obtained according to [34] by studying the OPE
of the spin fields
SI (z)SJ (w)
SI (z)SJ (w)
SI (z)SJ (w)
I J
+ (z w) 4 J I J + ,
(D.13)
+ ,
(D.14)
+ ,
(D.15)
3
4
(z w)
I J (w)
(z w) 4
I J (w)
(z w) 4
which can be done in a straight-forward way with the result:

1
1
(3 3 + 4 4 + 5 5 )(z w) 4 + ,
2
(z w)
1
1
1
(3 3 4 4 5 5 )(z w) 4 + ,
S2 (z)S2 (w) =
3
2
(z w) 4
S1 (z)S1 (w) =
3
4
S1 (z)S2 (w) = (z w) 4 4 5 + ,
S2 (z)S1 (w) = (z w) 4 4 5 + ,
1
S1 (z)S1 (w) = 3 4 5 (z w) 4 + ,
S2 (z)S2 (w) = 3 4 5 (z w) 4 +
3
S1 (z)S2 (w) =
+ ,
1
(z w) 4
3
90
S1 (z)S2 (w) =
3
1
(z w) 4
+ .
Upon taking suitable linear combinations of these expressions, one finds besides the SU(2) currents (D.5), the following operators:
JT 2 = 3 3 ,
(D.16)
12 = 3 ,
(D.17)
= 3 ,
(D.18)
11 = 22 = 0,
(D.19)
12
where JT 2 is an additional U (1) current JT 2 associated to T 2 . Thus, we now have an SU(2)

U (1) R-symmetry group.
The construction of the dimension-3/2 supercurrents is not as simple as in the previous case,
because we have first to decouple the U (1) from SU(2) U (1). We start again with the N = 1
internal supercurrent energymomentum tensor for the case of 6 compact dimensions
TF = 3 X 3 + 4 X 4 + 5 X 5 + 3 X3 + 4 X4 + 5 X5 .
Instead of splitting its expression directly into two parts as above, we follow [34] and make the
following ansatz

TF =
eiq3 3 eiq4 4 eiq5 5 TF
(q3 ,q4 ,q5 )
(D.20)
q3 ,q4 ,q5
where only the following combinations of charges are allowed

q3
q4
q5
*

0
0
1
.
0 , 1 , 0
1
0
0
(D.21)
It is now easy to see that

TF(1,0,0) = X 3 ,
TF(1,0,0) = X3 ,
(0,1,0)
TF
(0,0,1)
TF
(0,1,0)
TF
(0,0,1)
TF
= X 4 ,
= X 5 ,
(D.22)
= X4 ,
(D.23)
= X5 .
(D.24)
At this point, it is possible, to separate the U (1) part by setting

G
= 3 X3 ,
T2
G+
= 3 X 3 ,
T2
JT 2 = 3 3 ,
(D.25)
while the K3 operators are the same as before. For further convenience, we also introduce the
complex supercurrent of the N = 2 subalgebra
+ G
G = G
K3 = 3 X3 + 4 X4 + 5 X5 ,
T2
which is used in several computations in the text.
(D.26)
91
Appendix E. Amplitudes with NSNS graviphotons

In this appendix, we present two possibilities for insertions of graviphotons from the NS sector
in the type II amplitudes discussed in Section 4. For instance, the 6d helicity combination (3.11)
for the R 4 gives rise to the following setup of left and right-moving components:
1
z1
z2
0
0
+1
+1
+1
1
0
0
0
0
+1
1
+1
+1
0
0
0
0
0
0
1
1
z3
z4
+1
1
1
1
0
0
0
0
0
0
+1
1
1
1
0
0
0
0
0
0
Graviphoton
g1
g1
xi
yi
+ 12
12
+ 12
12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
12
+ 12
12
12
12
12
12
12
12
PCO
g+1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
0
0
+1
0
0
0
+1
0
0
0
+1
Insertion
Number
Position
Graviphoton
1
1
Graviton
where the graviphotons at z1 and z2 are KaluzaKlein graviphotons and the graviton at z4 (z3 ) a
(anti-) self-dual graviton from the Weyl multiplet.
Similarly, the helicity setup of (3.12) corresponds to the insertion of 4 KK graviphotons at the
positions zi :
Insertion
Number
Position
Graviphoton
1
1
1
1
z1
z2
z3
z4
+1
+1
1
1
+1
1
+1
1
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
+1
1
+1
1
+1
+1
1
1
0
0
0
0
0
0
0
0
Graviphoton
g1
g1
xi
yi
+ 12
12
+ 12
12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
+ 12
12
+ 12
12
12
12
12
12
12
12
PCO
g+1
g1
g1
{s3 }
{s4 }
{s5 }
0
0
0
0
0
0
1
0
0
0
1
0
0
0
1
0
0
0
0
0
0
+1
0
0
0
+1
0
0
0
+1
In fact, one can show that these amplitudes lead also to the same result (4.2).
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
E. Witten, Prog. Math. 133 (1995) 637, hep-th/9207094.

I. Antoniadis, E. Gava, K.S. Narain, T.R. Taylor, Nucl. Phys. B 413 (1994) 162, hep-th/9307158.
M. Bershadsky, S. Cecotti, H. Ooguri, C. Vafa, Commun. Math. Phys. 165 (1994) 311, hep-th/9309140.
S. Cecotti, P. Fendley, K.A. Intriligator, C. Vafa, Nucl. Phys. B 386 (1992) 405, hep-th/9204102.
M. Bershadsky, S. Cecotti, H. Ooguri, C. Vafa, Nucl. Phys. B 405 (1993) 279, hep-th/9302103.
B. de Wit, Nucl. Phys. B (Proc. Suppl.) 49 (1996) 191, hep-th/9602060.
G. Lopes Cardoso, B. de Wit, T. Mohaupt, Phys. Lett. B 451 (1999) 309, hep-th/9812082.
G. Lopes Cardoso, B. de Wit, T. Mohaupt, Fortschr. Phys. 48 (2000) 49, hep-th/9904005.
H. Ooguri, A. Strominger, C. Vafa, Phys. Rev. D 70 (2004) 106007, hep-th/0405146.
92
[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]
C. Beasley, E. Witten, JHEP 0602 (2006) 060, hep-th/0512039.
I. Antoniadis, K.S. Narain, T.R. Taylor, Nucl. Phys. B 729 (2005) 235, hep-th/0507244.
I. Antoniadis, T.R. Taylor, Nucl. Phys. B 695 (2004) 103, hep-th/0403293.
M. Marino, G.W. Moore, Nucl. Phys. B 543 (1999) 592, hep-th/9808131.
A. Strominger, Nucl. Phys. B 451 (1995) 96, hep-th/9504090.
D. Ghoshal, C. Vafa, Nucl. Phys. B 453 (1995) 121, hep-th/9506122.
C. Vafa, Nucl. Phys. B 447 (1995) 252, hep-th/9505023.
R. Gopakumar, C. Vafa, hep-th/9809187.
R. Gopakumar, C. Vafa, hep-th/9812127.
N. Berkovits, C. Vafa, Nucl. Phys. B 433 (1995) 123, hep-th/9407190.
N. Berkovits, C. Vafa, Nucl. Phys. B 533 (1998) 181, hep-th/9803145.
K. Foerger, S. Stieberger, Nucl. Phys. B 559 (1999) 277, hep-th/9901020.
J.D. Fay, Theta Functions in Riemann Surfaces, Springer, 1973.
E.P. Verlinde, H.L. Verlinde, Nucl. Phys. B 288 (1987) 357.
See e.g., E. Kiritsis, N.A. Obers, B. Pioline, JHEP 0001 (2000) 029, hep-th/0001083.
H. Ooguri, C. Vafa, Nucl. Phys. B 451 (1995) 121, hep-th/9505183.
M. de Roo, J.W. van Holten, B. de Wit, A. Van Proeyen, Nucl. Phys. B 173 (1980) 175.
E. Bergshoeff, M. de Roo, B. de Wit, Nucl. Phys. B 182 (1981) 173.
A. Galperin, E. Ivanov, S. Kalitsyn, V. Ogievetsky, E. Sokatchev, Class. Quantum Grav. 1 (1984) 469;
A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Harmonic Superspace, Cambridge Univ. Press, 2001.
L. Andrianopoli, S. Ferrara, E. Sokatchev, B. Zupnik, Adv. Theor. Math. Phys. 4 (2000) 1149, hep-th/9912007.
E. Ivanov, S. Kalitsyn, A.V. Nguyen, V. Ogievetsky, J. Phys. A 18 (1985) 3433.
I. Antoniadis, S. Ferrara, S. Hohenegger, K.S. Narain, E. Sokatchev, in preparation.
T. Banks, L.J. Dixon, Nucl. Phys. B 307 (1988) 93.
Gauged supergravity from type IIB string theory

on Y p,q manifolds
Alex Buchel a,b, , James T. Liu c
a Perimeter Institute for Theoretical Physics, Waterloo, Ontario N2J 2W9, Canada
b Department of Applied Mathematics, University of Western Ontario, London, Ontario N6A 5B7, Canada
c Michigan Center for Theoretical Physics, Randall Laboratory of Physics, The University of Michigan,
Ann Arbor, MI 48109-1040, USA

Received 14 September 2006; received in revised form 26 February 2007; accepted 2 March 2007
Available online 12 March 2007
Abstract
We first construct a consistent KaluzaKlein reduction ansatz for type IIB theory compactified on Sasaki
Einstein manifolds Y p,q with FreundRubin 5-form flux giving rise to minimal N = 2 gauged supergravity
in five dimensions. We then investigate the R-charged black hole solution in this gauged supergravity, and
in particular study its thermodynamics. Based on the gauge theory/string theory correspondence, this nonextremal geometry is dual to finite temperature strongly coupled four-dimensional conformal gauge theory
plasma with a U (1)R -symmetry charge chemical potential. We study transport properties of the gauge
theory plasma and show that the ratio of shear viscosity to entropy density in this plasma is universal. We
further conjecture that the universality of shear viscosity of strongly coupled gauge theory plasma extends
to non-zero R-charge chemical potential.
1. Introduction
The gauge theory/string theory correspondence of Maldacena [14] provides a valuable insight into the non-perturbative dynamics of strongly coupled gauge theory plasma. However,
on the string theory side, this duality is tractable only for weakly curved backgrounds and with
small string coupling, i.e., in the supergravity approximation. Following the Maldacena corre* Corresponding author at: Department of Applied Mathematics, University of Western Ontario, London, Ontario N6A
5B7, Canada.
E-mail address: abuchel@uwo.ca (A. Buchel).
doi:10.1016/j.nuclphysb.2007.03.001
94
A. Buchel, J.T. Liu / Nuclear Physics B 771 (2007) 93112
spondence, this implies that the dual four-dimensional gauge theory must be strongly coupled at
all energy scales. Unfortunately, this latter condition excludes a dual supergravity description of
real QCD. Nonetheless, one expects that certain universal properties of strongly coupled gauge
theory plasma derived in the planar limit and for large t Hooft coupling would persist for real
QCD as well.
A noteworthy example of such a universal property is the ratio of shear viscosity to the entropy
density. In [57] a theorem was proven that, in the absence of chemical potentials, a gauge theory
plasma in the planar limit and for infinitely large t Hooft coupling1 in theories that admit a dual
supergravity description has a universal ratio of shear viscosity to the entropy density s:
1
(1.1)
=
.
s
4
Furthermore, other transport properties of the strongly coupled non-conformal gauge theory
plasma, like the speed of sound cs and the bulk viscosity , while not universal, have a certain
generic behavior in the near-conformal limit [1012]

1
1 2
2
2
= cs
+ O cs
, 2, . . . , 5.
(1.2)
3
3
Introduction of a non-zero chemical potential in the strongly coupled gauge theory plasma
violates the sufficient condition under which the universal result (1.1) has been derived [5,7].
On the other hand, it was recently shown [1315] that the universality (1.1) persists in N = 4
supersymmetric YangMills plasma even with non-zero R-charge chemical potentials.2 Thus it
appears that the shear viscosity relation (1.1) is more robust than originally anticipated in [57].
However, further exploration of shear viscosity with chemical potential in strongly coupled fourdimensional gauge theory plasma has been hampered by the fact that, until recently, few models
exist where the dual supergravity solutions are easily accessible.
The most well understood system is of course N = 4 YangMills plasma, which is holographically dual to IIB on AdS5 S 5 . From a five-dimensional point of view, this dual is given
by N = 8 gauged supergravity, which may be obtained by reduction of the ten-dimensional theory on S 5 . The N = 4 YangMills theory has an SO(6) R-symmetry which may correspondingly
be identified with the isometries of S 5 . It is then possible to turn on up to three commuting Rcharges under the maximum Abelian subgroup U (1)3 SO(6). At the same time, this subsector
of the full gauged supergravity admits a consistent truncation to N = 2 supergravity coupled to
two Abelian vectors, commonly referred to as the STU model. As a result, explicit investigations
of the non-zero R-charge sector of thermal N = 4 YangMills can be easily performed in the
context of the STU model. In particular, non-extremal R-charged black holes were constructed in
[17], and their thermodynamics were investigated in [1821]. In addition, these five-dimensional
solutions can be uplifted to the full IIB theory and investigated from the ten-dimensional point
of view [22,23].
The uplifting of five-dimensional gauged supergravity solutions to ten dimensions is closely
related to the existence of a full non-linear KaluzaKlein reduction ansatz of the original IIB
theory. As it turns out, a complete proof of the consistency of the reduction of IIB theory on S 5
has yet to be given. Nevertheless, it is generally assumed to be consistent based on the linearized
1 Finite t Hooft coupling corrections to shear viscosity of the N = 4 supersymmetric SU(N ) YangMills theory
plasma were discussed in [8,9].
2 The shear viscosity of the M2-brane plasma also appears to satisfy the universal relation (1.1) [16].
95
construction as well as explicit reductions to subsets of the full N = 8 theory in five dimensions.
In particular, the reduction to the STU model is given in [22], while reduction to N = 4 in five
dimensions was given in [24], and reduction of the metric and self-dual 5-form on S 5 was given
in [25].
Although N = 4 YangMills has proven to be a fruitful system to investigate, it would be
highly desirable to explore models with reduced supersymmetry. In this context, the explicitly
constructed SasakiEinstein metrics Y p,q [26,27] have led to rapid new developments in understanding their holographically dual N = 1 YangMills theories. At the same time, however,
much of this progress has been restricted to either zero temperature or zero chemical potential
(or both). What we aim to do below is to make progress towards the study of N = 1 Yang
Mills plasma at finite R-charge chemical potential. To do so, we first construct the consistent
non-linear KaluzaKlein reduction of IIB on Y p,q giving rise to gauged N = 2 supergravity in
five dimensions, and then turn to the thermodynamics and hydrodynamics of the corresponding
R-charged black holes.
In general, the KaluzaKlein reduction of a higher-dimensional theory on an internal manifold Y generally involves both massless lower-dimensional fields (related to zero modes on Y )
and the KaluzaKlein tower of massive states. This distinction, however, is not entirely clear,
except perhaps in the case where Y is a torus. Consistency of the reduction to the massless sector then depends on the decoupling of the KaluzaKlein tower, and this is a highly non-trivial
condition. For T n reductions, truncation to the zero modes is guaranteed to be consistent, as this
is a truncation to the charge singlet sector of U (1)n . However, for spheres or more complicated
manifolds Y , it is often the case that the states that are considered massless (such as states
in the lower-dimensional gauged supergravity multiplet) carry non-trivial internal charge. These
charged states then have the potential of acting as sources to the KaluzaKlein tower, thus rendering the truncation inconsistent unless some appropriate symmetry or conspiracy among fields
is in place.
As an example of this inconsistency, consider for example the compactification of IIB on T 1,1 .
The resulting five-dimensional theory is N = 2 gauged supergravity with a full KaluzaKlein
spectrum which was obtained in [28]. Because the isometry of T 1,1 is SU(2) SU(2) U (1),
the massless gauge bosons transform under the identical group. Since the U (1) gauge boson
is clearly the N = 2 graviphoton coupling to U (1)R , the massless sector may be described as
N = 2 gauged supergravity coupled to SU(2) SU(2) vector multiplets. It turns out, however,
that it is inconsistent to retain these SU(2) SU(2) vector multiplets in any truncation to the
massless sector [29]; the only consistent truncation is to pure N = 2 gauged supergravity. This,
however, is sufficient for our needs, as the pure supergravity sector is all that is needed for uplifting the R-charged configurations of interest.
The more general case of IIB reduced on Y p,q is similar. For generic Y p,q , its isometry is
SU(2) U (1) U (1), and the resulting KaluzaKlein reduction yields N = 2 gauged supergravity coupled to SU(2) U (1) vector multiples in the massless sector. Only the truncation to
minimal N = 2 gauged supergravity is consistent, and the resulting geometries are holographically dual to certain strongly coupled superconformal quiver gauge theories [30,31]. Working
at finite temperature and R-charge in these gauge theories corresponds to taking an R-charged
black hole in the supergravity description. Such black holes carrying U (1)R graviphoton charge
have been discussed in [17], and simply correspond to STU black holes with three equal charges.
Given these black holes and the KaluzaKlein reduction ansatz on Y p,q , we may then study
the thermodynamics and hydrodynamics of such objects. In particular, we examine the shear
viscosity and the speed of sound in the strongly coupled superconformal plasma which is holo-
96
graphically dual to the black hole background. We find that the shear viscosity and entropy
density in the quiver gauge theories [30,31] continues to satisfy the universality relation (1.1),
even in the presence of a non-zero chemical potential. Although we do not provide a proof, we
believe that this universality (1.1) is likely to be true in any gauge theory plasma with R-charge
chemical potentials turned on.
This paper is organized as follows. In the following section, we present the KaluzaKlein
reduction ansatz for IIB on Y p,q giving rise to minimal N = 2 gauged supergravity in five dimensions. We then take up the thermodynamics of R-charged black hole solutions of this N = 2
gauged supergravity in Section 3 and the hydrodynamics in Section 4. Finally, we present our
conclusions in Section 5.
2. N = 2 gauged SUGRA from type IIB string theory on Y p,q manifolds with 5-form
fluxes
A five-dimensional SasakiEinstein manifold Y is an Einstein manifold preserving some fraction of supersymmetry such that the cone over Y is (non-compact) CalabiYau. It is well known
that a SasakiEinstein manifold Y admits a constant norm Killing vector field, known as the
Reeb vector. The existence of the Reeb vector allows the metric on Y to be written as a U (1)
bundle over a four-dimensional KhlerEinstein base B
1
ds 2 (Y ) = ds 2 (B) + (d + A)2 ,
(2.1)
9
where dA = 6J and J is the Khler form on B.
The FreundRubin compactification of IIB on S 5 admits a straightforward generalization
where S 5 is replaced by the SasakiEinstein manifold Y . Taking only the metric and self-dual
5-form, the only ten-dimensional IIB equations of motion that we are concerned with are the
Einstein and 5-form equations
1 1
dF(5) = 0,
FMP QRS FN P QRS ,
2 2 4!
In addition, the IIB gravitino variation is given by

i
FN P QRS N P QRS M .
M = M +
16 5!
RMN =
F(5) = F(5) .
(2.2)
(2.3)
The resulting FreundRubin ansatz is then of the form

2
= ds 2 (AdS5 ) +
ds10
1 2
ds (Y ),
g2
F(5) = (1 + )G(5) ,
G(5) = 4g (5) ,
(2.4)
where (5) is the five-dimensional volume form of AdS5 , and g is the coupling constant of
the five-dimensional gauged supergravity (corresponding to the inverse AdS5 radius). The fivedimensional theory is described by N = 2 gauged supergravity, and is holographically dual to an
N = 1 superconformal gauge theory.
Since Y has the form of (2.1), there is a natural symmetry corresponding to the U (1) fiber.
On the gravity side, the gauge boson under this U (1) is the N = 2 graviphoton, while on the
SCFT side this symmetry is just the U (1)R symmetry. In order to retain the graviphoton, the
FreundRubin metric ansatz (2.4) may be extended in the obvious manner

1
1
2
2
2
,
ds10 = g dx dx + 2 ds (B) + d + A + A dx
(2.5)
9
g
97
where g is the five-dimensional metric and A is the graviphoton. The reduction of F(5) has
the form
F(5) = (1 + )G(5) ,
G(5) = 4g (5)
1
J(2) 5 F(2) ,
3g 2
(2.6)
where J(2) is the Khler form on B and F(2) = dA(1) with A(1) = A dx . At a linearized order
in gauge potential A , and without the backreaction of its field strength on the metric, the ansatz
(2.5), (2.6) appeared previously in [32]. We claim here that (2.5) and (2.6) are in fact consistent
at a non-linear level.
The above KaluzaKlein reduction ansatz, (2.5) and (2.6), gives rise to equations of motion
that may be obtained from the five-dimensional Lagrangian
1 2
1
(2.7)
F F A .
F +

12 108
This is the bosonic sector of minimal N = 2 gauged supergravity in five dimensions. As written, the graviphoton is non-canonically
normalized. However canonical normalization may be
achieved by the rescaling A 3A . Note, also, that the five-dimensional Newtons constant
G5 of this N = 2 gauged supergravity is related to the ten-dimensional Newtons constant G10
according to
e1 L5 = R + 12g 2
G5 =
G10
.
volume(Y )
(2.8)
Although the above results hold in general, much of the interest in this system is due to the
availability of explicit SasakiEinstein metrics Y p,q that were originally constructed in [26,27].
When IIB is compactified on Y p,q , the resulting AdS5 backgrounds are holographically dual to
explicit N = 1 superconformal quiver gauge theories SCFTp,q [30,31]. The special case of Y 1,0
corresponds to T 1,1 , and the dual gauge theory was investigated in [33]. To make the above more
explicit, we will first work out the T 1,1 reduction in detail, and then follow with the extension to
general Y p,q manifolds.
2.1. T 1,1 reduction of IIB
The well-known case of IIB on AdS5 T 1,1 was originally investigated in [33], and this
has led to numerous important variations. Furthermore, this was one of only a few explicitly
known examples of reduced supersymmetry systems, at least until the recent construction of
an infinite family of Y p,q manifolds. As indicated in the introduction, although T 1,1 admits an
SU(2) SU(2) U (1) isometry, it is only consistent to truncate to minimal N = 2 supergravity
with the U (1) gauged by the N = 2 graviphoton. Following (2.5) and (2.6), and expressing T 1,1
as U (1) bundled over S 2 S 2 , the reduction ansatz is as follows

1

1 1 2
2
= g dx dx + 2
d1 + sin2 1 d12 + d22 + sin2 2 d22
ds10
6
g 6

1
+ (d + cos 1 d1 + cos 2 d2 + gA(1) )2 ,
9

1
(sin
d
+
sin
d
)
F
F(5) = (1 + ) 4g (5) +
(2.9)
1
1
1
2
2
2
5 (2) .
18g 2
98
Before proceeding with the reduction, it is natural to introduce a vielbein basis for the T 1,1
coordinates
1
1
1
e1 = d1 ,
e2 = sin 1 d1 ,
e3 = d2 ,
6
6
6
1
e5 = (d + cos 1 d1 + cos 2 d2 + gA(1) ).
3
1
e4 = sin 2 d2 ,
6
(2.10)
Note that the self-dual Khler form on the S 2 S 2 base is given by J(2) = e1 e2 e2 e4 ,
where the sign is chosen to match (2.1) and (2.6). Using the relations
de1 = 0,
de2 = 6 cot 1 e1 e2 ,
de3 = 0,
de4 = 6 cot 2 e3 e4 ,
1
de5 = 2J(2) + gF(2) ,
(2.11)
3
we may obtain the spin-connections
34 = 6 cot 2 e4 + e5 ,
12 = 6 cot e2 + e5 ,
15 = e2 ,
25 = e1 ,
35 = e4 ,
45 = e3 ,
(2.12)
on T 1,1 , as well as
= (5)
1 5
F e ,
6g
1
5 = F e ,
6
(2.13)
for the mixed components.

The above vielbeins are naturally given for the unit radius T 1,1 . Hence the rewriting of (2.9)
in terms of the above vielbeins introduces several factors of g as follows:
2
ds10
= g dx dx +
1 i 2
e ,
g2
i

1
F(5) = (1 + ) 4g (5) 2 J(2) 5 F(2)
3g

1 1
1
1
= 4 g (5) + 4 e e2 e3 e4 e5 2 J(2) 5 F(2) + e5 F(2) .
g
g
3g
(2.14)
Since the Khler form is closed, dJ(2) = 0, it is now easy to see that the ten-dimensional Bianchi
identity dF(5) = 0 yields the five-dimensional Bianchi identity and equation of motion for F(2)
1
d 5 F(2) = F(2) F(2) .
(2.15)
3
Note that, to verify the Bianchi identity, we also need the fact that J(2) J(2) gives twice the
volume form on the EinsteinKhler base, namely J(2) J(2) = 2e1 e2 e3 e4 .
Turning next to the ten-dimensional Einstein equation, we note that it splits into several components, , ij , and i. We find that the components gives the five-dimensional Einstein
equation

1
1
(5)
R
(2.16)
= 4g 2 g +
F F g F 2 ,
6
6
dF(2) = 0,
99
while the internal ij components are identically satisfied. In addition, the mixed 5 component
yields the 2-form equation of motion
1
(2.17)
F F ,
12
which is identical to the one given in (2.15) in form notation. As indicated at the beginning
of this section, the combined equations of motion (2.15) and (2.16) may be obtained from the
Lagrangian (2.7) for minimal gauged supergravity in five dimensions.
Although we have only focused on the bosonic fields of the reduction, it ought to be apparent
that the full KaluzaKlein reduction onto N = 2 gauged supergravity including the gravitino is
a consistent one. To see this, we may highlight the reduction of the IIB gravitino variation (2.3).
In order to proceed, we first introduce a decomposition of the ten-dimensional Dirac matrices

A = 1 1 , 1 a 2 .
(2.18)
F =
By convention, we take the product of spacetime Dirac matrices to be 0 1 2 3 4 = i and

internal Dirac matrices to be 1 2 3 4 5 = 1. In addition, the IIB gravitino and supersymmetry parameter are ten-dimensional Weyl spinors satisfying the chirality projection 11 =
where 11 = 0 9 = 1 1 3 . As a result, such chiral IIB spinors may be written as

1
=
.
(2.19)
0
Given this decomposition of the Dirac matrices, we now proceed with the reduction. The
gravitino variation (2.3) is perhaps most straightforwardly investigated in tangent space. Using
(2.14) for the 5-form, and taking chirality into account, we note that

FBCDEF BCDEF = 40i 24g 12 + 34 F ,

FBCDEF BCDEF a = 40 24ga 12 + 34 F a .
(2.20)
Combining this with the covariant derivatives formed from the spin connections (2.12) and
(2.13), we obtain the spacetime variation

1
i
1
= (5) gA + F 5 F 12 + 34 + g , (2.21)
12
48
2
as well as the T 1,1 variations

i

i
1 =
61 F 1 12 + 34 + 1 1 i 34 ,
48
2

i

i
6 csc 1 1 F 2 12 + 34 + 2 1 i 34
2 =
48
2

1 12
,
6 cot 1 +
2

i 3

i
3 12
34
12
+ 1 i
,
62 F
3 =
48
2

i 4

i
4 12
34
12
+ 1 i

6 csc 2 2 F
4 =
48
2

1
6 cot 2 + 34 ,
2
100

1 12

1
12
34
34
5
2 i i
+ + + i
.
5 = 3 F
48
2
(2.22)
The internal Dirac matrices provide a spinor representation of the SO(5) tangent space of
T 1,1 . This 4 of SO(5) corresponds to taking the independent weights i 12 = 1 and i 34 = 1.
Examining the variations (2.22), we see that the Killing spinor must take the weights i 12 =
i 34 = 1. Furthermore, 5 is not independent, and must have corresponding eigenvalue 5 = 1.
Thus this T 1,1 background preserves 1/4 of the supersymmetries, and the Killing spinor of T 1,1
satisfies the projections
i 12 = i 34 = 5 = .
The
T 1,1
(2.23)
Killing spinor equation is now trivial to solve, and the resulting solution is
i
= e 2 0 ,
(2.24)
with 0 satisfying the above projections. This demonstrates that the Killing spinor is charged
along the U (1) fiber. This charge is of course expected, and gives rise to a charged gravitino in
five dimensions as appropriate for the gauged supergravity theory. Inserting this Killing spinor
into (2.21), and rewriting this variation using curved indices, we finally obtain

i
i
1
(5)

4 F + g ,
= gA +
(2.25)
2
24
2
which is the appropriate gravitino variation for minimal N = 2 gauged supergravity in five dimensions.
2.2. Y p,q reduction of IIB
We now turn to the case of Y p,q . As expected, the reduction of IIB on Y p,q is almost identical
to that for T 1,1 . We start with the metric on Y p,q written as [27,30]

1
1 1y 2
wv
2
ds10 = g dx dx + 2
d + sin2 d 2 +
dy 2 +
(d + cos d)2
6
wv
36
g

2
1
+ d cos d + y(d + cos d) + gA(1) ,
(2.26)
9
where
w=
2(a y 2 )
,
1y
v=
a 3y 2 + 2y 3
,
a y2
and introduce the natural vielbein basis

1y
1y
1
2
d,
e =
sin d,
e =
6
6
1
H
dy,
e4 = (d + cos d),
e3 =
6H
6

1
e5 = d cos d + y(d + cos d) + gA(1) .
3
(2.27)
(2.28)
For later convenience, we have defined the functions

wv
a 3y 2 + 2y 3
a 3y 2 + 2y 3
H
H=
,
=
,
K=
=
6
3(1 y)
2(1 y)
12(1 y)3
101
(2.29)
which satisfy the relation

y
dH
=K .
dy
H
(2.30)
In terms of the vielbeins, the 5-form ansatz takes the same form as that given above for the
T 1,1 case. In particular

1
F(5) = (1 + ) 4g (5) 2 J(2) 5 F(2) ,
(2.31)
3g
where the Khler form is given by J(2) = e1 e2 + e3 e4 . Using the relations

6
1
1
3
2
2
3
de = 6Ke e ,
de = 6Ke e +
cot e1 e2 ,
1y

y 3
de4 = 6 K
e e4 2 6Ke1 e2 ,
de3 = 0,
H

1
de5 = 2 e1 e2 + e3 e4 + F(2) ,
3
we see that

d e3 e4 = 2 6Ke1 e2 e3 .
d e1 e2 = 2 6Ke1 e2 e3 ,
(2.32)
(2.33)
As a result, d(e1 e2 + e3 e4 ) = 0, giving an explicit demonstration that the Khler form

is closed, namely dJ(2) = 0. As a result, the ten-dimensional Bianchi identity dF(5) = 0 again
reduces to the five-dimensional Bianchi identity and equation of motion (2.15). Similarly, the
reduction of the ten-dimensional Einstein equation follows just as in the T 1,1 case. The resulting
bosonic fields g and A are then described by the N = 2 Lagrangian (2.7).
Turning to the IIB gravitino variation, we first compute the spin connections from (2.32)

6
12 =
13 = 6Ke1 ,
14 = 6Ke2 ,
cot e2 + 6Ke4 e5 ,
1y

y 4
24 = 6Ke1 ,
34 = 6 K
e e5 ,
23 = 6Ke2 ,
H
15 = e2 ,
25 = e1 ,
35 = e4 ,
45 = e3 ,
= (5)
1 5
F e ,
6g
1
5 = F e .
6
(2.34)
and
(2.35)
Since the 5-form reduction (2.31) has the same form as for the T 1,1 case, the reduction of F A
results in an expression identical to (2.20), however with the replacement F F due to
the opposite sign convention for the Khler form J(2) .
102
The IIB gravitino variation (2.3) then breaks up into the spacetime component

1
i
1
= (5) gA + F 5 + F 12 + 34 + g ,
12
48
2
and Y p,q components

1
6(1 y)1/2 +
6K 23 12 34 +
1 =
2

i
F 1 12 34 ,
48

6(1 y)1/2 csc cot + cot

2 =
(2.36)

i 1
1 + i 34
2

1
12
2

1
i
i
6K 24 12 34 + 2 1 + i 34 F 2 12 34 ,
2
2
48

i
i
6H y + 3 1 + i 12 + F 3 12 34 ,
3 =
2
48

1
i

1
1 12
6H 6yH 1 34 +
6K 34 + 4 1 + i 12
4 =
2
2
2

i
+ F 4 12 34 ,
48

i
1
5 = 3 12 + 34 i 5 F 12 + 34 2i .
(2.37)
2
48
It is clear from the above that the Killing spinor on Y p,q is given by
i
= e 2 0 ,
i 12 0 = i 34 0 = 5 0 = 0 ,
(2.38)
and hence the spacetime component (2.36) reduces to the expected N = 2 gravitino variation
(2.25).
Although the derivation of this reduction ansatz is straightforward (since in essence we are
only performing a KaluzaKlein reduction along a U (1) fiber), the exact KaluzaKlein ansatz
is extremely useful in that it enables us to greatly expand the investigations of gauge theories at
finite R-charge chemical potential. This is what we now turn to in the following sections.
3. R-charged black holes and their thermodynamics
Working in the Poincar patch of AdS5 , we may consider the following black hole ansatz in
the minimal gauged supergravity described by (2.7):
ds52 = c12 (dt)2 + c22 (d x
)2 + c32 (dr)2 .
(3.1)
Here the functions only depend on the radial coordinate r, namely ci = ci (r). In order to work
in a given R-charge sector, we must also consider turning on a gauge field potential
At = a(r).
(3.2)
Solving the equations of motion (2.15), (2.16) within the ansatz (3.1), (3.2) yields
c1 = rf 1/2 ,
c2 = r,
c3 = c11 ,
a =
2Qc1 c3
c23
103
Q2
+ 6,
(3.3)
4
r
9r
where the prime denotes a derivative with respect to r, is the non-extremality parameter and
the parameter Q is related to the R charge of the black hole.
Let us compute the renormalized (in the sense of [34]) Euclidean gravitational action IE of
(2.7). First, we regularize (2.7) by introducing a boundary M5 at fixed (large) r with the unit
orthonormal space-like vector n r

f =1
S5r
1
=
16G5
1
16G5
V3
16G5
r

dr
M5
r+
r
dr
r+
d g E LE =
4
2c22 c1 c2
c3

2c22 c1 c2 r

1
16G5

dr
r+
d 4 gL
M5
d 4
M5
(3.4)
,
c3
r
r+
where the subscript E indicates that all the quantities are to be computed in Euclidean signature,
and r+ is the outer black hole horizon, i.e., the largest positive root of
f (r+ ) = 0.
(3.5)
The black hole temperature and entropy are given by

TH =
3
r+
Q2
5
6r+
SBH =
3V
r+
3
.
4G5
(3.6)
As usual, to have a well-defined variational problem in the presence of a boundary requires

the inclusion of the GibbonsHawking SGH term

1
V3 (c1 c23 )
SGH =
(3.7)
d 4 hE n =
,
8G5
8G5
c3
M5
where h is the induced metric on M5

h g n n .
(3.8)
Finally, as in [34], we supplement the combined regularized action (S5r + SGH ) by the appropriate
boundary counterterms constructed out of metric invariants on the boundary M5

1
S counter =
d 4 6 hE .
(3.9)
16G5
M5
In the present case the boundary curvature vanishes, so there are no extra counterterms. The
renormalized Euclidean action IE defined as

IE lim S5r + SGH + S counter ,
(3.10)
r
is finite.
104
We now proceed to the computation of the ADM mass for the background (3.1). Following
[34] we define

M = d 3 N ,
(3.11)
where S 3 is a spacelike hypersurface in M5 with a timelike unit normal u , N is the

norm of the timelike Killing vector in (3.1), is the determinant of the induced metric on ,
and is the proper energy density
= u u T .
(3.12)
The quasilocal stress tensor T for our background is obtained from the variation of the full
action
Stot = S5r + SGH + S counter ,
(3.13)
with respect to the boundary metric h

2 Stot
T =
.
h h
(3.14)
An explicit computation yields

T =

1
+ h 3h ,
8G5
(3.15)
where

1
(3.16)
= Tr .
n + n ,
2
Again, the renormalized stress energy tensor is finite.
Inserting the black hole solution into the above action (3.13) and renormalized stress tensor
(3.15), we obtain the action and mass
=
IE =
V3
,
16G5
M=
3V3
.
16G5
(3.17)
Notice that, by using (3.6) and (3.17), we find the expected thermodynamical relation
IE = (M q q)
SBH ,
(3.18)
where q is the chemical potential conjugate to the physical black hole charge density
Q
q = ,
3
(3.19)
related to the gauge potential At (3.2) at the horizon, r = r+

Q
V3 At
V3
q =
=
(3.20)
2.
8G5 3 r=r+ 8G5 3r+
The factors of 3 in (3.19) and (3.20) come from the non-canonical normalization of the gauge
field in (2.7).
105
We may now identify the thermodynamical quantities of the charged black hole {IE , M, SBH ;
TH , q } with the appropriate gauge theory quantities {, E, S, T , J }
{IE , M, SBH ; TH , q } {/T , E, S; T , J },
(3.21)
where the thermodynamic potential is related to the Helmholtz free energy F in the standard
way
= F J J = E T S J J.
(3.22)
On can explicitly verify that with the identification (3.21) the first law of thermodynamics for the
grand canonical ensemble with {T , J } as independent variables
d = S dT J dJ
is satisfied automatically. To check (3.23) it is useful to use

d f (r+ ) = 0,
(3.23)
(3.24)
which is simply the statements that given {, Q}, the radius of the outer horizon of the BH is
given by (3.5).
4. Hydrodynamics
In this section, we examine the hydrodynamics of the N = 2 YangMills plasma at finite
R-charge which is dual to the black hole solution of (3.1), (3.2) and (3.3). In particular, using
prescription [35], we compute the retarded Greens function of the boundary stress-energy tensor
T (t, x ) ( = {t, x }) at zero spatial momentum, and in the low-energy limit 0:

R
G12,12 (, 0) = i dt d 3 x eit (t) T12 t, x , T12 (0, 0) .
(4.1)
Computation of this Greens function allows a determination of the shear viscosity through the
Kubo relation

1 A
= lim
(4.2)
G12,12 (, 0) GR
12,12 (, 0) ,
0 2i
where the advanced Greens function is given by GA (, 0) = (GR (, 0)) .
Although the extraction of the retarded Greens function from the R-charged black hole background is somewhat involved, the result turns out to be independent of charge and chemical
potential. As demonstrated below, we find

is
GR
(4.3)
(,
0)
=
1
+
O
,
12,12
4
T
where
3
r+
(4.4)
4G5
is the BekensteinHawking entropy density of the black hole. Inserting this expression into (4.2)
then yields the universal ratio
s=
=
.
s
4
(4.5)
106
4.1. Computation of the retarded Greens function

We begin the computation of (4.1) by recalling that the coupling between the boundary value
of the graviton and the stress-energy tensor of a gauge theory is given by g21 T12 /2. According to
the gauge/gravity prescription, in order to compute the retarded thermal two-point function (4.1),
we should add a small bulk perturbation g12 (t, y) to the metric (3.1), and compute the on-shell
b (t). Symmetry arguments [36] guarantee that for
action as a functional of its boundary value g12
a perturbation of this type and metric and gauge potential of the form (3.1) and (3.2), all other
components of a generic perturbation g along with the gauge potential perturbations A can
be consistently set to zero.
Instead of working directly with g12 , we find it convenient to introduce the field = (t, r)
according to
1
1
= g 11 g12 = c22 g12 .
2
2
(4.6)
The retarded correlation function GR

12,12 (, 0) can be extracted from the (quadratic) boundary
effective action Sboundary for the metric fluctuations b given by

Sboundary b =
where

() =
b
M5

d 4k b
b

()F
(,
r)
()
,

4
(2)
horizon

d 4 k it
e
(t, r)
.
4
(2)
M5
(4.7)
(4.8)
In particular, the Greens function is given simply by

GR
12,12 (, 0) =
2F r (, r),
(4.9)
where F is the kernel of (4.7). The boundary metric functional is defined as

r

Sboundary b =
lim
Sbulk [] + SGH [] + S counter [] ,
r
(4.10)
lim
Mr5 M5
M5 M5
r
where Sbulk
is the bulk Minkowski-space effective supergravity action (2.7) on a cut-off space
r
M5 (where M5 in (3.1) is regularized by the compact manifold Mr5 with a boundary Mr5 ).
Also, SGH is the standard GibbonsHawking term over the regularized boundary Mr5 . The
r
is evaluated on-shell for the bulk metric fluctuations (t, r) subject
regularized bulk action Sbulk
to the following boundary conditions:
(a):
(b):
lim
Mr5 M5
(t, r) = b (t),
(t, r) is an incoming wave at the horizon.
(4.11)
The purpose of the boundary counterterm S counter is to remove divergent (as Mr5 M5 ) and
-independent contributions from the kernel F of (4.7).
We find that the effective bulk action for (t, r) in the supergravity background (3.1), (3.2)
takes the form
107

1
1
5
d xL5 =
d 5x
Sbulk []
16G5
16G5

1
1
3
2
2
(t ) 2 (r )
c1 c 2 c 3
2c12
2c3

3
3

c1 c22 c2 2
2c2 c3
2c2 c1
+ t
t + r
r +
.
c1
c3
c3
(4.12)
The second line in (4.12) is the effective action for a minimally coupled scalar in the geometry
(3.1), while the third line is a total derivative. Thus the bulk equation of motion for is that of
a minimally coupled scalar in (3.1). The latter equation is simplified by introducing a new radial
coordinate3
c1
x ,
(4.13)
c2
which using (3.3) can be inverted to give the near horizon expansion
c2 (x) = r+
7
9r+
6)
2(Q2 18r+
x2 +
13 (11Q2 90r 6 )
81r+
+
6 )3
8(Q2 18r+

x4 + O x6 .
(4.14)
Decomposing as
(t, x) = eit (x),
(4.15)
we find that the equation of motion reduces to

0 =
(18c28 c22 Q2 + 6xc2 c2 Q2 + 12x 2 (c2 )2 Q2 )

xc22 (Q2 18c26 )
92 c22 c2 (c2 + 2c2 x)

x 3 (Q2 18c26 )
w ,
(4.16)
where primes denote derivatives with respect to x.
A low-frequency solution of (4.16) which is an incoming wave at the horizon, and which near
the boundary satisfies
lim (x) = 1,
(4.17)
x1
can be written as

(x) = x iw F0 (x) + iwF (x) + O w2 ,
(4.18)
where w
(4.19)
.
2T
The functions {F0 , F }, which are smooth at the horizon, satisfy the following differential equations:
w=
0 = F0
(18c28 c22 Q2 + 6xc2 c2 Q2 + 12x 2 (c2 )2 Q2 )

xc22 (Q2 18c26 )
F0 ,
3 With such a definition, x has a range x [0, 1] with x 0 being the horizon and x 1 the boundary.
+
108
0 = F
(18c28 c22 Q2 + 6xc2 c2 Q2 + 12x 2 (c2 )2 Q2 )

xc22 (Q2 18c26 )
6c Q2 (c2 + 2xc2 )
2
F0 + 2 2
F0 .
x
xc2 (Q2 18c26 )
(4.20)
Notice that for Q = 0 the dependence on the specific background geometry (c2 (x) dependence)
drops out and (4.20) coincides with universal equations derived in [7]. In the present case, however, since the Q dependence appears highly non-trivial, universality is far from assured.
The general solution of the first equation in (4.20) takes the form

F0 = C1 + C2
x

18c28 c22 Q2 + 6yc2 c2 Q2 + 12y 2 (c2 )2 Q2
dx exp
dy
,
yc22 (Q2 18c26 )
(4.21)
where Ci are integration constants. Using the near horizon expansion (4.14), we find as x 0+

6
4
27Q2 r+
2
.
F0 = C1 + C2 ln x
x
+
O
x
(4.22)
6 )2
2(Q2 18r+
Thus non-singularity of F0 at the horizon requires C2 = 0. The boundary condition (4.17) further
specifies
F0 (x) = 1.
(4.23)
The solution of the second equation in (4.20) is a bit more complicated and will be discussed
in Appendix A. We note here that unlike the universal case [7], given (4.23) and the boundary
conditions, this solution is non-trivial. Altogether we have

(t, x) = eit x iw 1 + iwF (x) + O w2 .
(4.24)
Once the bulk fluctuations are on-shell (i.e., satisfy equations of motion) the bulk gravitational
Lagrangian becomes a total derivative. From (4.12) we find (without dropping any terms)
L5 = t J t + r J r ,
(4.25)
where
Jt =
3c23 c3
t ,
2c1
Jr =
3c23 c1
c 2 c1 c
r + 2 2 2 .
2c3
c3
(4.26)
Additionally, the GibbonsHawking term provides an extra contribution so that

Jr Jr
2c23 c1
(c1 c23 ) 2
r
.
c3
c3
(4.27)
We are now ready to extract the kernel F of (4.7). The regularized boundary effective action
for is
3

c c1
1
Sboundary []r =
(4.28)
dt d 3 x 2 r + c.t.,
16G5
2c3
Mr5
where as prescribed in [35], we need only keep the boundary contribution. In (4.28) c.t. stands
for (finite) contact terms that will not be important for computations. Substituting (4.24) into
109
(4.28) we can obtain F r (, r):

c23 (r)c1 (r) c1 (r)
iw
dF (x)
1 + O(w)
1
,
F (, r) =
32G5
c3 (r)
c2 (r)
dx
r
where we have recalled the definition of x in (4.13). Thus, to order O(w2 ), we have

iw
dF (x)
r
lim F (, r) =
lim 1
,
r
32G5 x1
dx
(4.29)
(4.30)
where we used (3.3). Now, given (A.1) and (A.8), we finally obtain
iw
lim F (x)
32G5 x1

3
3
T r+
is
i r+
iw
=
=
,
=
16G5
8 4G5
8
lim F r (, r) =
(4.31)
where we have used the expression for the entropy density (4.4) as well as the definition (4.19).
Using (4.9) to extract the Greens function from F r then gives
is
(4.32)
,
4
at least in the low frequency limit 0. This is the result claimed in (4.3), giving rise to the
universal ratio of shear viscosity to entropy density (4.5).
GR
12,12 (, 0)
4.2. The sound speed

We conclude this section by computing the speed of sound in the SCFTp,q plasma. Using the
fact that the thermodynamic potential is
= P V3 ,
(4.33)
where P is the pressure, we find from (3.21) and (3.17) that

P=
1E
,
3 V3
(4.34)
which further implies that the speed of sound is

cs2 =
1 P
1
= ,
V3 E 3
(4.35)
independent of the value for the chemical potential.

5. Conclusion
In the first half of this paper, we have derived explicit KaluzaKlein reduction anstze of IIB
on T 1,1 Y 1,0 and Y p,q , yielding in both cases minimal N = 2 gauged supergravity in five
dimensions. Although the KaluzaKlein spectra of these reductions include additional vector
multiplets in the massless sector (SU(2) SU(2) for T 1,1 or SU(2) U (1) for generic Y p,q ),
these vectors cannot be retained in a consistent truncation [29]. In fact, this inconsistency even
precludes the retention of the vectors in the U (1)2 subgroups of the above groups. As a result,
we see that it is in fact not possible to realize the N = 2 STU model from T 1,1 or Y p,q reduction,
110
even though this attractive possibility is otherwise suggested from the linearized KaluzaKlein
analysis.
Our main result is a demonstration that the shear viscosity of SCFTp,q plasma with non-zero
U (1)R symmetry charge chemical potential is universally related to the entropy density (1.1).
For p = q, a cone over Y p,p is a Z2p orbifold of C3 , and hence the dual superconformal quiver
plasma is just a Z2p orbifold of the N = 4 YangMills theory. It is probably not surprising that
the orbifold quivers of N = 4 YangMills plasma have a universal shear viscosity with nonzero chemical potential, much like the parent gauge theory [1316]. What is rather unexpected,
however, is that the universality (1.1) is also true for the Y 1,0 T 1,1 superconformal gauge
quiver, which arises as a nontrivial superconformal infrared fixed point of the renormalization
group flow from the Z2 orbifold of N = 4 supersymmetric YangMills in the ultraviolet [33].
Thus it is natural to conjecture that the ratio /s is a constant along the renormalization group
flow, and as such, must be true for any strongly coupled gauge theory plasma with non-zero
chemical potential which allows for a (gauged) supergravity dual. Needless to say, it would be
very interesting to prove this conjecture, and if it is true, understand its applications for the
charged black holes in string theory. In particular, one should try to understandthe shear viscosity
of the gauge theory plasma with a finite chemical potential but with cs = 1/ 3. An example of
exactly such a model would be the cascading gauge theory plasma [3740].
Acknowledgements
We would like to thank Rob Myers, Andrei Starinets and Leo Pando Zayas for valuable discussions. A.B. would like to thank the Benasque Center for Science and the European Centre
for Theoretical Studies in Nuclear Physics and Related Areas for hospitality during the final
stages of this work. J.T.L. would like to thank the National Center for Theoretical Sciences (Taiwan) and the National Taiwan University Department of Physics for hospitality. A.B. research
at Perimeter Institute is supported in part by the Government of Canada through NSERC and
by the Province of Ontario through MEDT. A.B. gratefully acknowledges further support by an
NSERC Discovery grant. J.T.L. is supported in party by the US Department of Energy under
grant DE-FG02-95ER40899.
Appendix A. Solution for F
As is clear from (4.30), what we really need for the computation of the shear viscosity is not
F , but rather its derivative. We thus introduce
1 dF (x)
.
F (x) = +
x
dx
Given the analytical solution for F (x), F can be found by integrating (A.1)
(A.1)
1
F =
dy F (y) + ln x.
(A.2)
We now note that non-singularity of Fw at the horizon x 0+ implies that

F=
1
+ O(1).
x
(A.3)
111
Using the definition (A.1), one finds from the second equation in (4.20)
0=
18c28 c22 Q2 + 6xc2 c2 Q2 + 12x 2 (c2 )2 Q2

dF
F.
dx
xc22 (Q2 18c26 )
(A.4)
In terms of the r coordinate [see (4.13)] (A.4) takes form

0=
(5Q4 + 18r 6 Q2 54Q2 r 2 + 1082 r 4 )

dF
F,
dr
r(Q2 6r 2 )(9r 6 + Q2 9r 2 )
(A.5)
where we used the explicit background solution (3.3). A general solution of (A.5) takes the form
F=
Cr 5
,
(9r 6 + Q2 9r 2 )1/2 (Q2 6r 2 )
(A.6)
where C is an arbitrary integration constant. Using r = r(x) = c2 (x) given by (4.14), the boundary condition (A.2) determines
C=
6 Q2
18r+
2
r+
3
= 18T r+
,
(A.7)
where we used the expression for the temperature (3.6) and (3.5). Using (A.6) and (A.7), we can
finally evaluate
lim F (x) = lim F (r) =
x1
3
T r+
C
=
.
18
(A.8)
References
[1] J.M. Maldacena, The large N limit of superconformal field theories and supergravity, Adv. Theor. Math. Phys. 2
(1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200.
[2] S.S. Gubser, I.R. Klebanov, A.M. Polyakov, Gauge theory correlators from non-critical string theory, Phys. Lett.
B 428 (1998) 105, hep-th/9802109.
[3] E. Witten, Anti-de Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150.
[4] O. Aharony, S.S. Gubser, J.M. Maldacena, H. Ooguri, Y. Oz, Large N field theories, string theory and gravity, Phys.
Rep. 323 (2000) 183, hep-th/9905111.
[5] A. Buchel, J.T. Liu, Universality of the shear viscosity in supergravity, Phys. Rev. Lett. 93 (2004) 090602, hepth/0311175.
[6] P. Kovtun, D.T. Son, A.O. Starinets, Viscosity in strongly interacting quantum field theories from black hole physics,
Phys. Rev. Lett. 94 (2005) 111601, hep-th/0405231.
[7] A. Buchel, On universality of stress-energy tensor correlation functions in supergravity, Phys. Lett. B 609 (2005)
392, hep-th/0408095.
[8] A. Buchel, J.T. Liu, A.O. Starinets, Coupling constant dependence of the shear viscosity in N = 4 supersymmetric
YangMills theory, Nucl. Phys. B 707 (2005) 56, hep-th/0406264.
[9] P. Benincasa, A. Buchel, Transport properties of N = 4 supersymmetric YangMills theory at finite coupling,
JHEP 0601 (2006) 103, hep-th/0510041.
[10] P. Benincasa, A. Buchel, A.O. Starinets, Sound waves in strongly coupled non-conformal gauge theory plasma,
Nucl. Phys. B 733 (2006) 160, hep-th/0507026.
[11] A. Buchel, Transport properties of cascading gauge theories, Phys. Rev. D 72 (2005) 106002, hep-th/0509083.
[12] P. Benincasa, A. Buchel, Hydrodynamics of SakaiSugimoto model in the quenched approximation, hepth/0605076.
[13] D.T. Son, A.O. Starinets, Hydrodynamics of R-charged black holes, JHEP 0603 (2006) 052, hep-th/0601157.
[14] J. Mas, Shear viscosity from R-charged AdS black holes, JHEP 0603 (2006) 016, hep-th/0601144.
[15] K. Maeda, M. Natsuume, T. Okamura, Viscosity of gauge theory plasma with a chemical potential from AdS/CFT,
Phys. Rev. D 73 (2006) 066013, hep-th/0602010.
112
[16] O. Saremi, The viscosity bound conjecture and hydrodynamics of M2-brane theory at finite chemical potential,
hep-th/0601159.
[17] K. Behrndt, M. Cvetic, W.A. Sabra, Non-extreme black holes of five-dimensional N = 2 AdS supergravity, Nucl.
Phys. B 553 (1999) 317, hep-th/9810227.
[18] A. Chamblin, R. Emparan, C.V. Johnson, R.C. Myers, Charged AdS black holes and catastrophic holography, Phys.
Rev. D 60 (1999) 064018, hep-th/9902170.
[19] M. Cvetic, S.S. Gubser, Phases of R-charged black holes, spinning branes and strongly coupled gauge theories,
JHEP 9904 (1999) 024, hep-th/9902195.
[20] A. Buchel, L.A. Pando Zayas, Hagedorn vs. HawkingPage transition in string theory, Phys. Rev. D 68 (2003)
066012, hep-th/0305179.
[21] J.T. Liu, W.A. Sabra, Mass in anti-de Sitter spaces, Phys. Rev. D 72 (2005) 064021, hep-th/0405171.
[22] M. Cvetic, et al., Embedding AdS black holes in ten and eleven dimensions, Nucl. Phys. B 558 (1999) 96, hepth/9903214.
[23] A. Buchel, Higher derivative corrections to near-extremal black holes in type IIB supergravity, hep-th/0604167.
[24] H. L, C.N. Pope, T.A. Tran, Five-dimensional N = 4, SU (2) U (1) gauged supergravity from type IIB, Phys.
Lett. B 475 (2000) 261, hep-th/9909203.
[25] M. Cvetic, H. L, C.N. Pope, A. Sadrzadeh, T. A Tran, Consistent SO(6) reduction of type IIB supergravity on S 5 ,
Nucl. Phys. B 586 (2000) 275, hep-th/0003103.
[26] J.P. Gauntlett, D. Martelli, J. Sparks, D. Waldram, Supersymmetric AdS5 solutions of M-theory, Class. Quantum
Grav. 21 (2004) 4335, hep-th/0402153.
[27] J.P. Gauntlett, D. Martelli, J. Sparks, D. Waldram, SasakiEinstein metrics on S 2 S 3 , Adv. Theor. Math. Phys. 8
(2004) 711, hep-th/0403002.
[28] A. Ceresole, G. DallAgata, R. DAuria, S. Ferrara, Spectrum of type IIB supergravity on AdS5 T 11 : Predictions
on N = 1 SCFTs, Phys. Rev. D 61 (2000) 066001, hep-th/9905226.
[29] P. Hoxha, R.R. Martinez-Acosta, C.N. Pope, KaluzaKlein consistency, Killing vectors and Khler spaces, Class.
Quantum Grav. 17 (2000) 4207, hep-th/0005172.
[30] D. Martelli, J. Sparks, Toric geometry, SasakiEinstein manifolds and a new infinite class of AdS/CFT duals, Commun. Math. Phys. 262 (2006) 51, hep-th/0411238.
[31] S. Benvenuti, S. Franco, A. Hanany, D. Martelli, J. Sparks, An infinite family of superconformal quiver gauge
theories with SasakiEinstein duals, JHEP 0506 (2005) 064, hep-th/0411264.
[32] D. Berenstein, C.P. Herzog, I.R. Klebanov, Baryon spectra and AdS/CFT correspondence, JHEP 0206 (2002) 047,
hep-th/0202150.
[33] I.R. Klebanov, E. Witten, Superconformal field theory on threebranes at a CalabiYau singularity, Nucl. Phys. B 536
(1998) 199, hep-th/9807080.
[34] V. Balasubramanian, P. Kraus, A stress tensor for anti-de Sitter gravity, Commun. Math. Phys. 208 (1999) 413,
hep-th/9902121.
[35] D.T. Son, A.O. Starinets, Minkowski-space correlators in AdS/CFT correspondence: Recipe and applications,
JHEP 0209 (2002) 042, hep-th/0205051.
[36] G. Policastro, D.T. Son, A.O. Starinets, From AdS/CFT correspondence to hydrodynamics, JHEP 0209 (2002) 043,
hep-th/0205052.
[37] A. Buchel, Finite temperature resolution of the KlebanovTseytlin singularity, Nucl. Phys. B 600 (2001) 219, hepth/0011146.
[38] A. Buchel, C.P. Herzog, I.R. Klebanov, L.A. Pando Zayas, A.A. Tseytlin, Non-extremal gravity duals for fractional
D3-branes on the conifold, JHEP 0104 (2001) 033, hep-th/0102105.
[39] S.S. Gubser, C.P. Herzog, I.R. Klebanov, A.A. Tseytlin, Restoration of chiral symmetry: A supergravity perspective,
JHEP 0105 (2001) 028, hep-th/0102172.
[40] O. Aharony, A. Buchel, A. Yarom, Holographic renormalization of cascading gauge theories, Phys. Rev. D 72
(2005) 066003, hep-th/0506002.
Spacetime instanton corrections in 4D string vacua:

The seesaw mechanism for D-brane models
Ralph Blumenhagen a,c , Mirjam Cvetic b,c , Timo Weigand b,
a Max-Planck-Institut fr Physik, Fhringer Ring 6, 80805 Mnchen, Germany
b Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104-6396, USA
c Kavli Institute for Theoretical Physics, University of California, Santa Barbara, CA 93106-4030, USA
Received 8 November 2006; accepted 26 February 2007

Available online 6 March 2007
Abstract
We systematically investigate instanton corrections from wrapped Euclidean D-branes to the matter
field superpotential of various classes of N = 1 supersymmetric D-brane models in four dimensions. Both
gauge invariance and the counting of fermionic zero modes provide strong constraints on the allowed nonperturbative superpotential couplings. We outline how the complete instanton computation boils down to
the computation of open string disc diagrams for boundary changing operators multiplied by a one-loop
vacuum diagram. For concreteness we focus on E2-instanton effects in type IIA vacua with intersecting
D6-branes, however the same structure emerges for type IIB and heterotic vacua. The instantons wrapping
rigid cycles can potentially destabilise the vacuum or generate perturbatively absent matter couplings such
as proton decay operators, -parameter or right-handed neutrino Majorana mass terms. The latter allow the
realisation of the seesaw mechanism for MSSM like intersecting D-brane models.
1. Introduction
Recent years have witnessed the emergence of a new picture of the space of string vacua. On
the one hand, it is by now widely appreciated that the introduction of non-trivial background
fluxes induces a computable scalar potential. The resulting plethora of stable minima of this
potential has given rise to the idea of a landscape of string vacua. On the other hand, progress in
E-mail addresses: blumenha@mppmu.mpg.de (R. Blumenhagen), cvetic@cvetic.hep.upenn.edu (M. Cvetic),

timo@sas.upenn.edu (T. Weigand).
doi:10.1016/j.nuclphysb.2007.02.016
114
R. Blumenhagen et al. / Nuclear Physics B 771 (2007) 113142
string model building has taught us that also the gauge sector of the perturbative four-dimensional
string vacua including both heterotic and D-brane constructions is much richer than anticipated.
As it stands, however, this picture is not complete due to the existence of important nonperturbative effects in string theory, which are well known to have dramatic consequences. In
particular, in the absence of any other stabilising effects like fluxes they may destabilise the entire string vacuum or generate new couplings which have to be taken into account both for the
analysis of the scalar potential and for string phenomenology. Attempts in the literature to confront such issues in string-motivated constructions include the KKLT scenario, where precisely
these non-perturbative effects play a starring role, namely in freezing all moduli in a type IIB
orientifold compactification.
Let us stress that the non-perturbative effects are uniquely determined for each concrete string
vacuumthey cannot simply be chosen at our will to add suitable terms to the potential which
freeze some of the moduli. One might argue that for small string coupling they can be neglected
as compared to the perturbative or tree-level contributions. However, there do exist important
couplings which are only generated by instanton effects and therefore constitute the leading
order terms which determine the physics. In particular in addressing the cosmological constant
problem, instanton effects cannot be neglected. Therefore, it is of utmost importance for a better
understanding of the string landscape to study these non-perturbative effects for all known classes
of four-dimensional string vacua. In the context of heterotic string models this problem was taken
seriously from the very beginning, but for D-brane models it has apparently not received the full
attention it deserves.
The aspects of instanton corrections discussed in the literature include world-sheet instanton
corrections for the heterotic string [16] and type IIA string [79], M2- and M5-brane instantons
in M-theory [1012], D-instanton effects in the D3D(1) system [1315] in type IIB string theory, inducing e.g. corrections to R 4 terms, and the effect of background fluxes on E2-instantons
[16], to name but a few. A most prominent role alluded to already is played by E3-instantons in
type IIB theory due to their dependence on the Khler moduli [17].
It is the aim of this paper to systematically investigate spacetime instanton effects with special
emphasis on D-brane models, where for concreteness we start our investigations in the framework of type IIA intersecting D-branes (see [18] for a review). Explicit examples of semi-realistic
four-dimensional string vacua have been constructed in this framework on various background
geometries. Whereas in the framework of general CalabiYau manifolds the description of
special Lagrangian (sLag) submanifolds wrapped by the D6-branes is a notoriously difficult
problem, it is among the virtues of toroidal orbifolds that they contain a large class of tractable
supersymmetric three-cycles. In addition, the exactly known underlying conformal field theory
(CFT) allows for the computation of some of the perturbative low energy couplings at string tree
level.
A distinctive feature of intersecting brane worlds is the appearance of U (Na ) gauge factors on the stacks of D6-branes, whose Abelian U (1) part is often anomalous. In this case, the
GreenSchwarz mechanism effectively converts them into global U (1) symmetries in that they
are, while broken as a gauge symmetry, still respected by all perturbative correlation functions.
These global U (1)s are often a nuisance for string phenomenology as they forbid some highly
desirable charged matter couplings such as certain Yukawa couplings for SU(5) models or Majorana masses for right-handed neutrinos.
115
In this paper we address the question whether charged matter superpotential couplings
W
M
i eSinst ,
(1)
i=1
violating the global U (1) symmetries, can be generated non-perturbatively. In fact, the global
U (1)s and the associated gauged global shift symmetries of the axions can be broken by Euclidean D2-brane instantons, called E2-instantons in the sequel. The possible appearance of a coupling of the form (1) was first considered in [2] in the context of the heterotic string. As already
mentioned above, non-perturbative effects also play a crucial role in the context of moduli stabilisation. As one of their most dramatic effects, they have the potential to destabilise the vacuum
completely or, often in combination with other ingredients like fluxes, to drive it to an isolated
minimum in the string landscape.
We outline the general structure of such non-perturbative contributions to the superpotential of
the form (1), to a future publication. The key ingredients, analysed in Sections 2 and 3, are given
by the issues of charge conservation and the counting of fermionic zero modes. The latter yield
strong constraints on the possible couplings. We also discuss how the instanton computation
works at the level of conformal field theory and find a structure analogous to the one derived
in more spacetime oriented approaches to the problem. In addition we make some comments
on multi-instanton corrections and instanton corrections to the D-terms and the gauge kinetic
function.
While for concreteness we will explain our methods in the setup of type IIA D-brane models,
we will point out in Section 4 how the same structure found in the type IIA context emerges for
type IIB orientifolds and for heterotic string models. In fact, our analysis is related, via a chain
of dualities, to the investigation of world-sheet instanton effects for the Spin(32)/Z2 heterotic
string [4]. We will see that some of the more abstract notions appearing in the heterotic computation have a clear geometric and conformal field theoretic interpretation for E2-instantons in
type IIA. Moreover, our analysis clarifies that for dual type IIB orientifolds, the naive uplifting
mechanism in the KKLT scenario by magnetised D7-branes does not work due to the appearance
of extra fermionic zero modes spoiling the generation of the superpotential term for Euclidean
D3-brane instantons. Only by multiplying the exponential factor by appropriate charged matter
fields the uplifting might eventually work.
Eventually, Section 5 analyses some E2-instanton generated effects of immediate phenomenological interest for MSSM-like type IIA intersecting D-brane models. These involve destabilisation mechanisms on the one hand and the possible appearance of perturbatively forbidden matter
couplings on the other. The latter include in principle dangerous dimension-four proton decay
operators, but also hierarchically small -terms and Majorana type masses for right-handed neutrinos.
Let us summarise our main findings:
The global U (1) charges formally carried (due to the GreenSchwarz mechanism) by the
exponential factor in the instanton amplitude have their microscopic origin in extra fermionic
zero modes living at the intersections of the instanton with the spacetime filling D-branes
present in the model.
Instantons wrapping rigid supersymmetric cycles can generate charged matter couplings in
the superpotential, which can potentially destabilise the vacuum or lead to new effects in the
116
four-dimensional action which are absent in perturbation theory. In this case their contribution, though exponentially suppressed, yields the leading order terms.
The technical treatment of such instanton amplitudes boils down to the computation of
conformal field theory disc and one-loop diagrams with boundary changing operators inserted along the boundary (see also [15]). These diagrams are multiplied by a non-vanishing
one-loop vacuum diagram, in which only the open string sectors preserving precisely two
supercharges contribute. We present the CFT analogue of the one-loop Pfaffian/determinant
known to generally appear in instanton computations [4].
By T-duality, E2-instantons in type IIA orientifolds are mapped to E1/E5- and E3-instantons
in type IIB orientifolds. S-duality maps the E1-instantons to world-sheet instantons for the
heterotic string. Many of the features of the latter have a corresponding geometric or open
string interpretation in the type IIA framework.
Instantons with appropriate intersections with the two D-branes supporting the right-handed
neutrinos on their intersection can give rise to Majorana mass-terms. The scale of these
masses is
Mm = xMs e
2
3s gs
VolE2
(2)
with x expected to be of order O(1). These can easily generate a hierarchy between the
string and the Majorana mass scale.
Note. We thank the authors of [19] and [20] for pointing out to us their work (see also [21]),
which likewise deals with instanton corrections in type II D-brane models.
2. The GS mechanism and global U (1)s
As our prototype class, we consider the type IIA string compactified on a CalabiYau (CY)
threefold X modded out by an orientifold projection (1)FL . The world-sheet parity transformation is combined with an anti-holomorphic involution of X .
Following mainly [22], on the threefold we introduce in the usual way a symplectic basis
(AI , B I ), I = 0, 1, . . . , h21 , of homological three-cycles with the topological intersection numbers
AI B J = IJ .
(3)
Upon introducing the corresponding three-forms satisfying

J = I J ,
J = I J ,
AI
(4)
BI
it follows that the Poincar dual of the three-cycles (AI , B I ) are the three-forms (I , I ).
Let us denote by O6 the homology class of the orientifold plane, which is the homology class
of the fixed point locus of the anti-holomorphic involution . The latter is a special Lagrangian
three-cycle in X . For tadpole cancellation, one introduces stacks of Na D6-branes wrapping
special Lagrangian three-cycles of homology class a . Their orientifold images are denoted
by a . Expanding such three-cycles into the symplectic basis
a = MaI AI + Na,I B I
(5)
117
Table 1
Chiral spectrum for intersecting D6-branes
Non-Abelian reps.
(2a )
a
a
(
(
U (1) charges
a,
a,
b)
b)
(2a )
(1a , 1b )
(1a , 1b )
Multiplicity
1 ( +
a
O6 a )
a
2
1 (
a
O6 a )
a
2
a b
a b
with M I , NI Z, one obtains for the topological intersection number between two three-cycles
a , b
a b = MaI Nb,I Na,I MbI .
(6)
The tadpole cancellation condition for an intersecting D-brane model of this type reads
K
Na (a + a ) = 4O6 .
(7)
a=1

For the generic gauge group G = K
a=1 U (Na ) the chiral massless spectrum is given in Table 1.
Note that this is just the chiral spectrum. For the non-chiral part one has to examine the
intersections more carefully. For this purpose, one introduces the physical intersection number
between two branes, which splits into the number of positive and negative intersections
a b = [a b ]+ + [a b ] .
(8)
The topological intersection number is then given by

a b = [a b ]+ [a b ] .
(9)
In general, the Abelian subgroup U (1)a U (Na ) is anomalous. It is known that these anomalies
are cancelled by a generalised GreenSchwarz mechanism involving the four-dimensional axions
coming from dimensional reduction of the ten-dimensional RR three- and five-form [23]. If we
expand C (5) in the symplectic basis
(2)

C (5) = 3s CI I D (2),I I ,
(10)
dimensional reduction of the ChernSimons couplings in the WZW action gives rise to the following mixing term between the four-dimensional axions and the U (1) gauge fields Fa = dAa

I

Na
(2)

Ma Ma I Fa CI + (Na,I Na,I
Smix = 2
)Fa D (2),I .
(11)
s
R1,3
The minus sign in front of the contribution from the image branes come from the fact that under
the U (1)a field strength transforms as Fa Fa . Expanding similarly the Hodge dual threeform C (3)
(0)

C (3) = 3s CI I D (0),I I ,
(12)
(2)
(0)
one notices that in four dimensions (CI , D (2),I ) is Hodge dual to (D (0),I , CI ). The Green
Schwarz mechanism now implies that under a U (1)a gauge transformation
Aa, Aa, + a
(13)
118

(0)
the axions (CI , D (0),I ) transform as

CI(0) CI(0) + QaI a ,
D (0),I D (0),I + P a,I a
(14)
with

Na
Na I

),
P a,I =
(Na,I Na,I
Ma Ma I .
(15)
2
2
Therefore, the U (1) gauge potentials and axions which remain massless have to lie in the left
and right kernels, respectively, of the (K, 2h21 + 2) matrix Q = (Q, P ). The U (1) symmetries
corresponding to the massive gauge potentials still survive as perturbative global symmetries in
that all perturbative correlation functions respect them.
Since some of the axions become the longitudinal modes of the massive U (1) gauge fields,
also their superpartners must become massive for supersymmetry preserving configurations.
These mass terms for the corresponding complex structure moduli arise from the generated
FayetIliopoulos terms

a Arg 3 .
(16)
QaI =
From dimensional reduction of the DBI action one can deduce the SU(Na ) gauge kinetic
functions to be given by1

1
1
(3)
fa =
(3 ) i C
(17)
.
(2)3s gs
a
With the normalisation tr(Ta Tb ) = 12 ab , one obtains for the physical non-Abelian gauge coupling [24,25]
4
1
=
VolD6a .
2
ga
gs 3s
(18)
3. E2-instantons
The question we address in this section concerns the effects of spacetime instantons on
the type of models just described. Such non-perturbative effects are generated in principle by
wrapped E0-, E2- and E4-branes. However, since a CalabiYau manifold does not have any continuous one and five-cycles, the only relevant2 instantons are E2-branes wrapped on three-cycles
of the CY threefold. Additional non-perturbative effects arise from world-sheet instantons and
in principle there could also be NS5-brane instantons wrapping the entire CY. The latter would
be magnetically charged under the NSNS two-form B2 with both legs along R(1,3) and are
therefore modded out by the orientifold projection.
1 For supersymmetry, = 0 and clearly

a
a (3 ) = | a 3 |.
2 We will not explore here the possibility of E4-instantons wrapping homologically trivial, but fluxed five-cycles being
the Euclidean version of the coisotropic eight-branes proposed in the intersecting brane context recently [26]. Their
effect, if relevant, is expected to be qualitatively of the same structure as the one of E2-instantons investigated in this
article.
119
For ease of presentation our detailed discussion will be restricted to single-instanton effects,
whereas we will indicate the relevant modifications required for the multi-instanton case in Section 3.4.
3.1. Global U (1) charges
The E2-branes couple to the RR form C (3) and therefore to the axions whose shift symmetries have been gauged for the cancellation of the U (1) anomalies. Consequently, such an
instanton transforms non-trivially under Abelian gauge symmetries.
Let us consider a single E2-brane wrapping a homological three-cycle
= E I AI + FI B I .
(19)
The orientifold action requires the introduction of an E2-brane wrapping also the (1)FL
image cycle . Since we are interested in contributions to the N = 1 superpotential, the threecycle should preserve supersymmetry, i.e. it should be an sLag cycle. Even more, in the internal
CY space this cycle should preserve the same supersymmetry as the orientifold plane and the
D6-branes do.
Since the E2-brane is localised in the flat four dimensions spanned by the D6-branes the whole
configuration breaks one-half of the supersymmetry so that two supersymmetries are conserved
and two are broken. It follows that the E2-brane world-volume accommodates at least the two
fermionic zero modes i corresponding to the broken supersymmetries, as required for a contribution to the superpotential.
Any correlation function in this single-instanton background contains the classical factor

2 1
SE2
(3)
e
(20)
= exp 3
(3 ) i C
,
s gs
which depends exponentially on the complex structure moduli U . In fact, the emergence of this
factor follows directly from the CFT formalism to be presented in Section 3.3 (see also [27]).
The expansion of C (3) in Eq. (12) and the gauged shift symmetries (14) allow us to compute
the transformation properties of this instanton factor under the U (1)a gauge symmetries to be
eSE2 eiQa (E2)a eSE2
(21)
Qa (E2) = Na (a a ).
(22)
with
Therefore, the U (1) charge of this term is given by the topological intersection number of the
three-cycle wrapped by the E2-instanton and the three-cycles wrapped by the D6-branes. The
relative minus sign between a and its image brane a is just right to take care of the fact that
for invariant D6-branes there is no U (1)a gauge field.
If any of the charges Qa (E2) is non-vanishing, it is clear that a purely exponential superpotential of the form W = c exp(SE2 ) cannot be generated due to gauge invariance. Only if the
exponential factor gets multiplied by a suitable combination of matter superfields also charged
under U (1)a can a superpotential

i eSE2
W=
(23)
i
120
be gauge invariant. Here appropriate contractions of indices are understood so that the matter
field product is invariant under all the non-Abelian gauge symmetries.
Note that while the product (23) is still invariant under the U (1)a global symmetries, of
course, the induced coupling of the superfields i clearly violates these Abelian symmetries
after all we have to evaluate eSE2 at its classical value which basically involves just the volume
of the three-cycle . In this sense the U (1)a global symmetries in the four-dimensional effective
action are broken by E2-instantons.
So far we have merely investigated selection rules resulting from the Abelian gauge symmetries for the generation of a non-perturbative superpotential. However, these are merely necessary
conditions for the appearance of an instanton induced coupling. To determine whether these
terms are really present requires a genuine instanton computation in which further issues such as
the number of fermionic zero modes play an important and restrictive role.
3.2. Zero modes
Of prime importance for the computation of an n-point function in an instanton background
are the fermionic zero modes. As is well known and will be reviewed in detail momentarily,
these have to be integrated over and therefore they have to be absorbed by the vertex operators
appearing in the n-point correlator. In the present context, it is important to distinguish between
two very different kinds of fermionic zero modes depending on whether or not they are charged
under the Abelian gauge groups.
We have already pointed out the presence of the usual two fermionic zero modes i corresponding to the broken supersymmetries of the half-BPS instanton which enable it to contribute
to the superpotential. These zero modes arise from the Ramond sector of an open string with both
ends attached to the E2-instanton. From the CFT point of view they are described by the vertex
operator (in the (1/2)-ghost picture)
1/2
V
(z) = i e
(z)
2
E2,E2
3/8
(z)S i (z),
(24)
where i , i = 1, 2, is the polarisation and S i denotes the 4D spin field of SO(1, 3). This is a Weyl
E2,E2
is essentially the spectral flow
spinor of conformal dimension h = 1/4. The twist field 3/8
operator of the N = 2 superconformal field theory describing the internal CY manifold. Since
the E2-brane wraps a three-cycle on the CY manifold, the 4D spacetime is transversal so that
there appears no 4D momentum factor in the vertex operator.
If the three-cycle has deformations counted by b1 ( ), then there will be extra fermionic
zero modes arising from open strings from the E2 to itself. Another source of zero-modes are
the intersections of with its image brane . Since on a single brane there are no fields transforming in the anti-symmetric representation of U (1)E2 and the projection changes sign for
the E2-branes relative to the D6-branes, the condition for the absence of extra uncharged zero
modes reads
[ ] = [O6 ] = 0
(25)
for single instantons.

In the sequel we call the zero modes of the type just described uncharged as they carry no
charge under any of the gauge symmetries localised on the D6-branes. In order to lead to a nonvanishing F-term, all these extra uncharged zero modes have to be absorbed by additional closed
121
or open string fields.3 As a result, if we want to generate a contribution to the superpotential of

only the charged matter fields, we require these uncharged zero modes to be absent. In particular,
this means that the sLag cycle needs to be rigid. The construction of such rigid cycles in
toroidal orbifolds has been explored in [29,30].
In view of the discussion of the previous section, one might suspect that the formal U (1)a
charges (22) of the instanton have a microscopic origin in the zero mode structure on the worldvolume of the E2-brane. Indeed, at the intersection of E2 with a stack of D6-branes, there do
appear extra charged states transforming in bi-fundamental representations. To determine these
states, we first note that generically for two intersecting D6-branes at non-trivial angles the
NeveuSchwarz and Ramond physical ground state energies for open strings stretching between
them are ENS > 12 and ER = 0 [31]. Moreover, at a supersymmetric intersection there lives one
4D chiral superfield, i.e. two real bosonic and two fermionic on-shell degrees of freedom. Now, if
one exchanges one of the D6-branes by an E2-brane wrapping the same sLag three-cycle, the four
flat NeumannNeumann boundary conditions for the open string become NeumannDirichlet
boundary conditions. This has three important effects for the open string partition function. First,
as is known from #ND = 4 systems, the GSO projection changes sign with respect to the GSO
projection in the D6D6 intersection. Second, the ground energies are now ENS + 12 > 0 and
ER = 0 so that there are no massless bosons at a non-trivial intersection. Third, the number of
massless fermions is halved, so that one obtains only one real or one-half complex fermionic
on-shell degree of freedom.4
As is well familiar from the intersecting brane context, for specific choices of the three-cycles
wrapped by the D6- and E2-branes, one gets further extended supersymmetry at the intersection
so that, as in [32], a pair of fermionic zero modes (a a ) appears. Only if the E2-instanton
wraps the same three-cycle as the D6-brane or at special limiting points in the complex structure
moduli space, also complex bosonic zero modes show up. It will become clear, however, that
these situations are of minor relevance for the computation of the superpotential terms we are
primarily interested in.
Let us call the extra fermionic zero modes arising from open strings stretched between the
E2-brane and a stack of Na D6-branes charged zero modes. The corresponding Ramond sector
open string vertex operators are of the form
V
(z)
1/2
a,E2
(z) = ia,I e 2 h=3/8
(z)h=1/4 (z),
ia,I
(26)
where I = 1, . . . , [ a ] runs over the number of intersection points and i = 1, . . . , Na

a,E2
is the ChanPaton (CP) index. Here h=3/8
denotes a spin field in the R-sector depending on
the internal intersection angles between the E2-brane and the D6-branes and is essentially the
same operator as it appears for vertex operators describing the fermionic zero modes for two
intersecting D6-branes [33,34]. The 4D spin field h=1/4 arises from the twisted 4D world-sheet
3 In this case, we also have to integrate over the bosonic components of the b ( ) superfields describing the instan1
ton moduli. We expect, however, that such configurations do not contribute to the superpotential, but rather to higher
derivative F-terms, see [28] for the analogous effect in the context of heterotic (0, 2) models.
4 As we will see in Section 4.2, these configurations are mapped by a chain of dualities to world-sheet instantons for
(0, 2) heterotic string models. Here these 1/2 complex fermionic zero modes are the familiar left-moving zero modes
discussed in [3,4]. These sit in Fermi supermultiplets of the (0, 2) supersymmetry, which indeed do not contain any
dynamical bosonic superpartners.
122
Table 2
Zero modes on E2, D6 intersections
Zero modes
Reps.
Number
a,I
a,I
(1E , a )
(1E , a )
I = 1, . . . , [ a ]+
I = 1, . . . , [ a ]
a ,I
a ,I
(1E , a )
(1E , a )
I = 1, . . . , [ a ]+
I = 1, . . . , [ a ]
bosons carrying half-integer modes. Note that also these zero modes carry no momentum along
the flat 4D directions.
The total number of such charged fermionic zero modes is displayed in Table 2.
From the table it is clear that the total charge of all the fermionic zero modes on the intersection of and (a + a ) is precisely Qa (E2) = Na (a a ).
To conclude, the microscopic origin of the U (1)a charges of the instanton factor is the appearance of extra fermionic zero modes living on the intersection of the instanton with the D6-branes.
Therefore they are charged under the gauge symmetry living on the D6-branes.
3.3. Outline of instanton calculus
In this section, we outline how one technically proceeds in computing the instanton corrections to the superpotential with special focus on charged matter couplings. As both the D6-branes
and the E2-instantons have a CFT description in terms of open strings, the computation can be
carried out in the CFT framework. Of course, since we are still lacking a string field theory
framework, one cannot derive the rules for instanton computations from first principles but has
to guess them by invoking analogies with what we experience in field theory (see however [35]
for an account of non-perturbative effects directly in ten dimensions). In the following we intend
to focus on the general structure of such a computation, making extra combinatorical factors and
normalisations explicit only where they are essential. Moreover, in order to avoid overloading the
notation with too many indices, we suppress the intersection indices I and the CP-indices i. It is
understood that all expressions we write down have also to include these labels in a consistent
manner.
Our aim is to derive the contribution of an E2-instanton to the superpotential involving powers
of charged chiral superfields
W
M
ai ,bi eSE2 ,
(27)
i=1
where the ai ,bi = ai ,bi + ai ,bi are superfields localised at the intersection of the D6-brane
ai with the D6-brane bi . If a D6-brane has open string moduli, then this includes the case
ai = bi . To make this distinction more transparent, we will call these latter superfields ai =
ai ,ai with bosonic components a in the sequel.
The vertex operators for the bosons in the (1)-ghost picture are given by
ab
(z) = e(z) h=1/2
(z)eip X
V1
a,b
(z)
(28)
ab
is a twist operator of conformal dimension h = 1/2 depending on the intersection
where h=1/2
angles between the two D6-branes in the internal CY directions (see [34] for the explicit form
123
of these states for toroidal models). The corresponding vertex operators for the fermions in the
(1/2)-ghost picture are
1/2
Va,b (z) = e
(z)
2
ab
h=3/8
(z)S (z)eip X
(z)
(29)
where S is a spin field of SO(1, 3) of conformal dimension h = 1/4. For a D6-brane modulus
or rather its bosonic component the corresponding boundary deformation vertex operator in the
(0)-ghost picture reads
aa
(z)eip X
V0a (z) = h=1
(z)
(30)
Note again that extra intersection and CP-labels have been suppressed. For ease of notation we
will not take care of the image branes explicitly, but it is understood that some of the matter
fields in (27) can also arise from intersections with the .
A coupling of the type (27) is detected by computing an appropriate matter field correlation
function in the E2-instanton background like

a1 ,b1 (p1 ) . . . aM2 ,bM2 (pM2 ) aM1 ,bM1 (pM1 ) aM ,bM (pM ) E2-inst .
(31)
In the amplitude we have to integrate over all fermionic and bosonic zero modes of the instanton,
thus essentially taking into account all possible instanton configurations. Since the correlation
function (31) we are probing for does not include any uncharged fields besides the two broken
supersymmetries, there may be no uncharged fermionic zero modes other than i present. Integration over them would make the amplitude vanish. As we explained already, this requires that
the E2-brane wrap a rigid sLag three-cycle and that there appear no extra massless states on
the intersection of with , i.e. = O6 = 0.
On the other hand, we also have to integrate over all charged fermionic zero modes, a,I
and a,I from the intersection of the E2-instanton with the D6-branes. By the same argument
as above, for an amplitude to be non-vanishing it has to involve each of these charged fermionic zero modes precisely once. We therefore propose that the complete instanton amplitude
(31) is given by the combinatorics of all possible non-vanishing CFT correlators involving the
E2-boundary

[ ]

+

a ]
a
[
g
4
2
d xd
akk ,bk . . . akk ,bk k .
da,I
d a,I
conf. a
I =1
I =1
(32)
Here the sum over all configurations means that we sum over all possible ways of assembling the matter fields into sets (labelled by k), for which in the integrand one computes the
open string CFT correlators of genus gk with all possible insertions of charged fermionic zero
modes. Note that in the CFT amplitudes one integrates over the world-sheet positions of the
CFT vertex operators so that the correlator only depends on the spacetime polarisation of the
fields.
For getting the 1PI contribution we do not allow any disc correlator without any zero modes
attached, i.e. those discs without any boundary on the E2-instanton. Even though these products
of CFT disc correlators with fermionic zero modes attached are 1PI reducible from the worldsheet point of view, they are nevertheless 1PI irreducible from the spacetime point of view, as we
integrate over the zero modes.
Recall that for precisely two of the ai ,bi we have to choose the fermionic components ai ,bi
and for the remaining ones the bosonic components ai ,bi . This is coupled to the question where
124
we insert the vertex operators of the i , as described below. To keep the notation as simple as
possible, we will not make this issue explicit in the following formulas.
The evaluation of (32) seems to be a daunting combinatorial exercise, but fortunately most
configurations do not contribute to the superpotential. Of prime importance is the standard fact
that the superpotential has to be a holomorphic function of the chiral superfields. As is well
known, this implies that the only dependence on the string coupling gs = e is via an exponential
factor of the type (20) involving also the axionic fields in the complex structure superfields.
Any other dependence on these axions and therefore on gs can be excluded due to the axionic
shift symmetries (U (1) charges) [1,2]. Note that in the CFT the matter fields are canonically
normalised so that the higher order corrections in gs reflect higher order corrections to the Khler
potential.
To conclude, counting of factors of gs together with the need to insert all fermionic zero
modes including the two uncharged i gives the terms which can, if non-zero, appear in the
superpotential. For this purpose, it is crucial to keep in mind the following normalisation factors
occurring:
Each disc diagram carries an overall normalisation factor proportional to gs1 . Each annulus
or Mbius diagram comes with an additional factor of gs . From the counting of gs it is
therefore clear that only world-sheets with the topology of a disc or of an annulus or Mbius
strip, respectively, can contribute to the superpotential. Moreover, at least one boundary has
to be the E2-instanton.
The vertex operators for the fields arising from D6-brane intersections do not carry any gs
normalisation, as otherwise they would decouple in the weak coupling limit gs 0.
On the contrary, the charged fermionic zero modes can in principle carry a normalisation
involving a positive power of gs , as in the weak coupling limit the instanton decouples. In
fact, if we do not assign any scaling to these zero modes, all disc amplitudes would be of
order gs1 and could therefore not appear in the superpotential. The minimal attachment of
fermionic zero modes involves two per disc, which we take as a strong indication that we
should assign a factor gs to each vertex operator a .5 As an immediate consequence, as

far as the superpotential is concerned, these fermionic zero modes can only contribute to the
disc amplitudes, whereas the 1-loop amplitudes have to be uncharged.
Disc amplitudes
Let us begin with the possible disc amplitudes which may contribute to the superpotential.
First, observe that one can always factor off the term ( 1disc )n for all powers n, as explained
1
from
in [15,27]. We have to sum over all such contributions, taking into account the factor n!
the combinatorics of boundaries. This leads to the familiar exp(SE2 ) factor in the instanton
amplitude.
The action of the E2-instanton depends holomorphically on the (1)FL invariant complex
structure moduli U . The non-renormalisation theorem of [36] for any polynomial dependence of
A-type disc diagrams on the complex structure moduli still applies, reflecting the fact that the
imaginary parts of the U are axionic states with a shift symmetry.
5 In the paper [15], the authors came to exactly the same conclusion, requiring that in the D3D(1) brane system the
zero modes localised on the intersection of these two branes should be identified with the gauge instanton moduli in the
ADHM construction.
125
Fig. 1. Standard disc tadpole.
Let us now turn to the non-trivial matter insertions, assuming for the time being that the D6branes are rigid. The basic building block of order (gs )0 involves just one factor of ai ,bi per
disc. Depending on the number of -modes inserted, we have the following options: As shown
in Fig. 1, the simplest process is described by a disc diagram with three boundary changing
operators, the bosonic vertex operator corresponding to ai ,bi being in the (1) picture and the
two fermionic vertex operators for ai and bi in the (1/2)-picture. With the above scaling
of the involved vertex operators and the normalisation of the disc amplitude, this amplitude is
indeed of the order (gs )0 , as required for terms in the superpotential. If we are dealing with one
of the fermionic modes ai ,bi , the disc also has to involve the vertex for one of the two i in order
that the ghost number of all vertex operators can add up to 2, i.e. all fermions can be chosen in
the (1/2)-picture.
Alternatively, the two i modes could of course be absorbed in a disc involving the auxiliary
field Fai ,bi and two charged fermion modes . This is what happens, for example, when we probe
for tadpoles of ai ,bi . Note that due to the Spin(1, 3) spinors present in the vertex operators of i
these are indeed the only ways to contract them with the vertex operators of the component fields
of ab . This is the well-known CFT explanation why half-BPS instantons can only contribute to
the superpotential.
All these diagrams can in principle be computed using conformal field theory techniques and
will subsequently be denoted as ai ,bi ai ,bi , where a consistent assignment of intersection
and CP-indices as well as an appropriate insertion of -modes and the corresponding choice of
components of ai ,bi is understood.
As an illustration, imagine the situation of four cycles a , b , c , d , each intersecting an
instanton E2 precisely once. The induced coupling for ab and cd is given by the product of
the two discs in Fig. 2, whereas the one-disc amplitude is of order gs and therefore does not
contribute to the superpotential.
As a second example, Fig. 3 shows a situation where a coupling ab can exist by soaking
up four fermionic zero modes, which is only possible in the presence of both d and d zero
1/2
modes. However, due to the gs factor for the fermionic zero modes, this coupling is likewise
of a higher power in gs and so does not contribute to the superpotential.
By contrast, as depicted in Fig. 4, it can also happen that the zero mode structure is such that
for instance a coupling a,x1 x1 ,x2 xn ,b can be generated by soaking up only two fermionic
zero modes.
We now consider the possibility of non-rigid D6-branes, i.e. b1 (a ) > 0. As anticipated, this
leads to the presence of additional b1 (a ) chiral superfields in the adjoint representation of
U (Na ), the open string moduli a [37]. The real part of their bosonic component parameterises
the deformations of the sLag cycle a and its imaginary part correspond to the Wilson lines of
126
Fig. 2. Leading and subleading disc diagrams for ab cd coupling. Only the leading order two-disc amplitude contributes
to the superpotential.
Fig. 3. Coupling ab absorbing four fermionic zero modes.
Fig. 4. Coupling a,x1 x1 ,x2 xn ,b and disc tadpole with D6-brane moduli insertions.
the U (1) gauge field on a single D6-brane. Again, the superpotential has to be a holomorphic
function of the full chiral superfield a , and the only possibility consistent with discrete U (1)
shifts in its bosonic imaginary part is an exponential dependence [7] of the form

a
tr exp
(33)
,
where the trace is over the CP-indices. Such a term is due to the fact that the standard disc tadpole
diagram can also have insertions of these moduli fields a (see Fig. 4). In this case one has to
sum over all these contributions inducing a consistent open string moduli dependence of the
superpotential couplings. As shown in [38], these integrable boundary deformations lead to the
expected exponential dependence.
127
To conclude, the only discs which can contribute to W are those with precisely two charged
fermionic zero modes per disc. It is useful to introduce the short-hand notation
ak ,bk [
xk ] = ak ,xk,1 xk,1 ,xk,2 xk,2 ,xk,3 . . . xk,n1 ,xk,n xk,nk ,bk
(34)
for the product of open string vertex operators from the set {a1 ,b1 , . . . , aM ,bM }, where each
field is allowed to only appear once, and again consistent assignment of CP-labels is tacitly
= ak ,bk . If we have L such chain-products, their lengths in an
assumed. We define ak ,bk [0]

M-point function have to add up to L
k=1 (nk + 1) = M Mloop , where Mloop is the number of
fields attached to one-loop diagrams (see below).
The resulting disc contribution to the instanton correlator therefore takes the comparably simple schematic form

[ ]

+

a ]
a
[
4
2
da,I
d a,I
d xd

I =1
conf. a

a1 ,b1 [
x1 ]
b
a1 ,
1
I =1

. . . aL ,bL [
xL ]
a
L ,bL
eSE2 .
(35)
Here, the sum over configurations only involves the possible ways of attaching the respective
fermionic zero modes to the matter field discs. Note that as in [33,39] each disc amplitude of
boundary changing operators contains by itself a whole sum of holomorphic disc instantons
ending on the respective
branes. This introduces a dependence of the instanton amplitude on the

Khler moduli Ti = i (J + iB) of the form
2

1
exp Ti ,
(36)
where 2i denotes a basis of the h11 cohomological two-cycles in the CY.

1-loop amplitudes
Having discussed the disc contribution to the instanton couplings, it only remains to clarify
the role of the 1-loop diagrams. Let us discuss the vacuum amplitudes first, where the story is
quite similar to the vacuum ( 1disc ) contribution.
The relevant 1-loop diagrams are the vacuum annulus and Mbius strip diagrams involving at
least one E2 boundary. Here for instance an annulus vacuum diagram Z A (E2, D6a ) is given by

Z (E2, D6a ) = c
A

dt
TrE2,D6a e2tL0 ,
t
(37)
where the trace is over all open strings stretching between the two branes.
Clearly supersymmetry implies that the amplitudes Z A (E2, E2) and Z A (E2, E2 ) vanish.
However, the open strings stretched between the E2-instanton and a stack of D6-branes preserve only two supercharges. As we have already seen for the zero mode counting, the amplitude
is not BoseFermi degenerate and we do not expect Z A (E2, D6a ), Z A (E2, D6a ) and the Mbius
amplitude Z M (E2, O6) to vanish. Following the same logic as for the ( 1disc )n terms in the
instanton amplitude, we can have all possible one-loop contributions, giving rise to a series
n

1 A
A

M
(38)
Z (E2, D6a ) + Z (D6a , E2) + Z (E2, O6) = exp(Z0 ),
n! a
n=0
128
Fig. 5. Series of 1-loop contributions for the example of only two D6-branes.
with
Z0 =

Z A (D6a , E2) + Z A (D6a , E2) + Z M (E2, O6).
(39)
This is also shown in Fig. 5.

In order to interpret this result, let us discuss the behaviour of this expression in the limits
t 0 and t . We expect the possible t 0 divergences to cancel by the tadpole cancellation condition for the D6a branes. The possible t divergence is due to zero modes in
the various open string sectors. If we have in the NS-sector a spacetime bosonic zero mode,
Z0 + and exp(Z0 ) diverges. On the other hand, in the case of a spacetime fermionic zero
mode in the R-sector, Z0 and exp(Z0 ) vanishes. Moreover, if Z0 vanishes identically in
a supersymmetric case, then exp(Z0 ) = 1. All this is precisely the behaviour characteristic of the
one-loop Pfaffians/determinants of the kinetic operators for the quantum fluctuation around the
E2-background. From the saddle point approximation such terms are expected to appear in the
instanton amplitudes.6 Moreover, these one-loop determinants have been argued to be exact for
the contribution to the superpotential, which is confirmed by the CFT approach.
6 In the dual heterotic model with a world-sheet instanton wrapping a holomorphic curve C, the uncharged contribution
to the superpotential was argued to be [4]
W=
Pfaff(V )
(det O )2 (det O(1) )2
exp(Sinst ).
(40)
This is consistent with our picture, as these determinants are only over the left-moving fluctuations. These involve the
vector bundle, whose role is played in our case by the D6-branes. The right-moving heterotic fluctuations cancel due to
supersymmetry and should therefore be identified with open strings between the E2-instanton and itself. Moreover, we
propose that the (det O )2 factor arises from the Mbius strip amplitude. Finally, let us point out that the connection
between the Pfaffian and the exponential of a suitable trace over the spectrum as in (38) is standard in field theory, of
course.
129
Fig. 6. Moduli dependence of 1-loop amplitudes.
As in [4], a zero or a divergence in the one-loop amplitude (38) signals that one has integrated
out a fermionic or bosonic zero mode, respectively. Since in our amplitudes we carry out this
integration over zero modes explicitly, the one-loop term should be regularised such that these
zero modes are removed from the traces. Therefore, we propose the following intriguing relation
between the spacetime notion of 1-loop Pfaffians/determinants for quantum fluctuations around
the instanton background and the exponential of a world-sheet open string partition function

Pfaff (DF )

(41)
Z A (E2, D6a ) + Z A (E2, D6a ) + Z M (E2, O6) .
= exp
det (DB )
a
One also has the freedom to attach the D-brane moduli fields a to the D6c -brane boundary
of the one-loop diagrams. After taking the sum over all possible insertions, this leads to the open
string moduli dependence of exp A0 (see Fig. 6).
This is similar to the case of world-sheet instanton corrections for the heterotic string, where
these 1-loop determinants do also depend on the bundle and complex structure moduli [4]. Finally, we stress once more that the one-loop determinant is manifestly uncharged under the
anomalous U (1) gauge factors. This direct consequence of the gs -assignment for the vertex
operators of the a -modes is crucial for the generation of a superpotential of the form (23)any
U (1) transformation of the one-loop prefactor would spoil the gauge invariance of (23).
To summarise, the single E2-instanton contribution to the charged matter superpotential can
be determined by evaluating the following zero mode integral over disc and 1-loop open string
CFT amplitudes

a1 ,b1 (p1 ) . . . aM ,bM (pM ) E2-inst

[ ]

+

a ]
a
[
4
2
i
i
= d xd
da
d a exp(SE2 ) exp(Z0 )
conf. a

x1 ]
a1 ,b1 [

b
a1 ,
1
i=1

. . . aL ,bL [
xL ]
i=1
aL , bL

loop
xk ]
ck ,ck [
k
A(E2,D6ck )
This formula contains from the disc the exponential instanton action and the combinatorics of
disc tadpole diagrams with two charged fermionic zero modes attached to each disc. From the
1-loop diagrams, i.e. the annulus and Mbius strip amplitudes, there also arises the exponential
vacuum contribution, i.e. the one-loop Pfaffian/determinant, and the combinatorics of attaching
D6-brane neutral products of vertex operators to the boundaries of annulus diagrams.
The schematic form of the superpotential, making explicit the moduli dependence, is

T
W=
(42)
eSE2 (U ) f exp , tr exp , ab ,
E2
130
where the exponential dependences are enforced by the fact that the imaginary parts of the complex scalars are either axions or Wilson lines having a shift symmetry.
In practice, to really carry out this CFT program, one needs the explicit form of the boundary
changing operators including their correct normalisation, which are only available in case we
really have a CFT description of the models. We will provide more details and examples for
such instanton computations in a future publication. For now it is a good working hypothesis that
every term which can in principle be generated, i.e. for which the zero mode counting and the
combinatorics of open strings works, is indeed non-vanishing.
3.4. Comments on multi-instanton effects
In general, the non-perturbative sector may not only involve a single E2-brane, but also multiinstantons. Of course, one expects that all these non-perturbative sectors allow a CFT description
along the lines presented so far. This includes essentially two types of generalisations.
First, the three-cycle may be wrapped by a stack of N coincident E2-branes. Clearly, in
such a situation the exponential suppression (20) of all resulting amplitudes is enhanced by a
factor of N as compared to the single-instanton case. The fermionic zero modes between the
D6- and the E2-branes carry an additional CP-index, accounting for the fact that the number
of such zero modes is likewise multiplied by N . Since again each fermionic zero mode has
to be soaked up in the computation of correlators and, as we just derived, each superfield ab
located at the intersection of two D6-branes can absorb at most two of them for contributions to
the superpotential, these N -instanton sectors soon become irrelevant for the most interesting
lowest order n-point amplitudes as N increases. Note also that the intersection between the
cycle and its orientifold image now gives rise to fermionic zero modes in the symmetric
and in the antisymmetric representation of the U (N ) gauge group located on the stack of E2branes. In analogy to Eq. (25) we have to require that
= 0 = O6
(43)
for the absence of these fermionic zero modes.

The most general instanton sector with possible contributions to the superpotential involves
several stacks of mutually supersymmetric E2-branes wrapping various rigid cycles i , each
with multiplicity Ni . A novelty occurs since the E2-branes will generically intersect not only
the D6-branes, but also among each other. Special care is required for the string modes located at
this latter type of intersection, i.e. between all pairs of i and j . This time the spectrum is indeed BoseFermi degenerate due to the mutual extended supersymmetry between the E2-branes,
which is a consequence of the DirichletDirichlet boundary conditions in the non-compact four
dimensions. In particular, the massless modes now include both a pair of fermions and a complex
boson and have to be integrated over. Note furthermore that, due to the BoseFermi degeneracy
between the E2-branes, the total Pfaffian simply factorises into the product of the Pfaffians resulting form the one-loop diagrams between each E2-instanton and the D6-branes. In order to
eventually extract superpotential couplings, one must only include 1PI diagrams.
3.5. E2-instanton corrections to D-terms and the gauge kinetic function
So far we have focused on E2-instanton corrections to the superpotential. From the general arguments of [7,8] one also expects non-perturbative corrections to the D-terms, which will depend
on the complex structure moduli. As a consequence there should appear E2-instanton corrections
131
to the FayetIliopoulos terms for the U (1)s living on the D6-branes. For all known examples
perturbative and non-perturbative corrections to the D-terms and the gauge couplings are by no
means independent. In fact, the latter two quantities constitute the argument and the absolute
value of a complex valued central charge, see Eqs. (16) and (17). Therefore, one likewise expects
E2-instanton corrections to the gauge kinetic functions. While we are not intending to compute
these terms explicitly here, we will nonetheless point our their origin and leave a more thorough
analysis for future work.
Given an intersecting D6-brane configuration, we know that the perturbative FI terms do
depend on the complex structure moduli, which is readily understood from the corresponding
dependence of the special Lagrangian condition on these moduli. Recall from Eq. (16) that the
leading order (supergravity) expression for the FI-term is proportional to the calibration condition for the three-cycle wrapped by the E2-brane. Assume now that the D6-branes are indeed
supersymmetric on a D-dimensional subspace MD MC of the naive complex structure moduli
space and consider the effect of a single E2-instanton. For the generation of F-terms we require,
as described, that this instanton wraps a special Lagrangian preserving the same supersymmetry
as the D6-branes. In general, this instanton is only supersymmetric on a (D 1)-dimensional
E2 of M . If one now moves off continuously from M E2 in the direction where
subspace MD1
C
D1
the D6-branes are still supersymmetric, then the instanton no longer preserves the 4D N = 1 supersymmetry. However, the number of fermionic zero modes between the D6- and the E2-brane
remains unchanged since the zero point energy in the R-sector vanishes, ER = 0, irrespective of
the intersection angle. Secondly, and in contrast to intersecting D6-branes, the bosonic modes
still have positive mass square. This is a consequence of the ENS 0 zero point energy in the
NS-sector and guarantees that the D6E2-brane configuration is still stable against brane recombination even away from the supersymmetric locus.
Due to the broken supersymmetry, the instanton world-volume now accommodates four
fermionic zero modes i and i as required for the generation of a D-term. We conclude that
this breaking of supersymmetry from the 4D point of view corresponds to a D-term breaking and
gives rise to the expected E2-instanton corrections to the FI terms on those D6-branes which are
not any longer relatively supersymmetric with the E2-instanton.
It follows that E2-instantons wrapping sLag three-cycles which preserve a different N = 1
subalgebra than the D6-branes induce instanton corrections to the D-terms. Let us be a bit more
precise and ask which further conditions the SUSY breaking instanton sector has to meet in order
to generate an FI-term. Recall that the actual FI-term arises in the four-dimensional effective
supergravity action as the term

SFI = d 4 x d 2 d 2 VU (1) ,
(44)
where VU (1) denotes the vector superfield containing the corresponding U (1) gauge potential.
What we have to compute in the CFT to probe for such a coupling are therefore diagrams with
an insertion of the vertex operator for the auxiliary field DU (1) on the D6-boundary in the disc
or the annulus diagram accompanied by an even number of boundary changing operators on the
disc corresponding to the fermionic zero modes between the D6-brane and the SUSY-breaking
instanton.
For single instantons, this implies that the intersection between the D6- and the E2-branes has
to be non-chiral and the E2-instanton has to be uncharged, in agreement with the fact that the
FI-term carries no U (1) charge. The combinatorics of multi-instanton contributions is richer as
now also diagrams involving several mutually non-supersymmetric instantons contribute.
132
The actual computation of the diagram parallels the story for the F-terms. The and vertex
operators match with the four-dimensional spinor polarisation of the D-term component in V ,
and all fermionic zero modes are to be integrated over. Again, the amplitude is exponentially
suppressed by the instanton action.
Similarly, inserting the vertex operators for the gauge fields on the D6-boundaries detects the
instanton corrections to the holomorphic gauge kinetic functions. In this case, only supersymmetric configurations contribute due to the necessary appearance of precisely two modes on
the instanton, and again the instantons have to be uncharged under the Abelian gauge group.
Clearly, it would be interesting to work this out in more detail. These corrections generalise
the familiar structure known from -stability [40] for B-type D-branes to also include spacetime
instanton corrections. Recall that the notion of -stability applies to the gs 0 limit and adds
up the world-sheet instanton contributions to the FI-terms, which by mirror symmetry are exact
at tree level for A-type D-branes. However, the spacetime instanton corrections are truly nonperturbative on both sides and finding a way to compute them would be a major step forward.
3.6. Example
After this amusing interlude, let us go back to the superpotential and give one simple example showing explicitly what kind of matter couplings can be generated and how the CP-factor
combinatorics is taken care of. Consider two stacks of in both cases N D6-branes wrapping
the three-cycles a and b with intersection number a b = 1. One gets one chiral super j = 1, . . . , N ) localised at this intersection point and transforming in the (N , N )
field i,j
(i,
representation of the U (N ) U (N ) gauge group.
Now, imagine there exists an E2-brane wrapping a three-cycle with non-vanishing intersection numbers
[ a ]+ = 1,
[ b ] = 1.
(45)
These generate in each case N zero modes i and j . Taking into account that one chiral superfield absorbs two fermionic zero modes, exactly one i and one j , it is clear that only a
term N has the potential to absorb all zero modes. Looking more closely, and observing that
interchanging the order of two fermionic zero modes gives a relative minus sign, the coupling
introduced by this instanton has the form
SE2
W det(i,j
)e
(46)
which is reminiscent of the instanton term la Affleck, Dine, Seiberg [41]. Of course, here
the gauge symmetry is different from the SU(NC ) case of [41] with Nf = NC 1 non-chiral
flavours. It shows that in contrast to perturbative string couplings, which only involve traces over
ChanPaton factors, instantons can also generate couplings like determinants of matter fields
involving anti-symmetrisations of indices. We will give more examples of possible couplings in
Section 5.
4. Type IIB/heterotic dual formulations
So far, mainly to profit from the illustrative geometric picture we have focused on instanton
effects for type IIA intersecting D6-brane models, but a very similar story applies to the various
dual D-brane models. In this section we would like to summarise aspects of instanton effects in
133
these dual formulations. Application of mirror symmetry to intersecting D6-brane models leads
to type IIB orientifolds with either O9- and O5-planes or O7- and O3-planes.
4.1. Type IIB orientifolds with O9/O5-planes
These are nothing else than compactifications of the type I string. Here for tadpole cancellation, one introduces D9-branes equipped with stable U (N ) vector bundles Va and D5-branes
wrapping effective curves of the CY threefold. As has been summarised in [42] (see also [43]),
i
the massless
2 spectrum is given by various cohomology classes such as H (X , Va Vb ),
i
Va ), etc. The GreenSchwarz mechanism works in complete analogy, now gauging
H (X ,
the shift symmetries of the RR two-form C (2) . More concretely there are two types of axions.
The first kind of axions are given by dimensional reduction of C (2) on the two-cycles of the
CalabiYau and are called Khler axions, being the imaginary parts of the superfields T . An
additional axion arises from reducing C (6) on the complete CY, which leads to the so-called
universal axion sitting in the dilaton superfield S.
The instantons which break these shift symmetries and therefore the global U (1)s are E1branes wrapping holomorphic two-cycles on the CY and E5-branes wrapping the entire CY. In
general these E5-branes can also carry stable holomorphic vector bundles so that their Chern
Simons action involves both the Khler axions and the universal axion. The dependence of the
superpotential on T can therefore only be exponential. By contrast, the complex structure U and
bundle moduli B do not contain any axions or Wilson-lines.
The necessary condition for E1-branes to generate a superpotential of the type described is
that they wrap rigid holomorphic curves, i.e. isolated C = P1 s. For such an E1-brane wrapping
a rigid curve C, the additional charged fermionic zero modes are counted by the cohomology
1/2
classes H i (C, Va |C KC ) with i = 0, 1 and KP1 = O(2).
For E5-instantons equipped with a gauge bundle VE , the number of extra uncharged zero
modes is given by h1 (X , VE VE ). For higher rank bundles this is generally non-vanishing,
but for line bundles L it is given by h1,0 (X ) = 0. The absence of zero modes from
intersections yields in this case the additional constraint H i (X , S2 L) = 0 for i = 1, 2, with
appropriate generalisations for non-Abelian bundles. The charged zero modes are counted by
H i (X , VE Va ) and H i (X , VE Va ) for i = 1, 2.
Again the actual instanton computation for charged matter contributions to the superpotential boils down to the computation of disc amplitudes for boundary changing operators. The
schematic form of the superpotential, making explicit the moduli dependence, is

eSE1 (T ) f (U, B, ab ) +
eSE5 (S,T ) f (U, B, ab ).
W=
(47)
E1
E5
Notice that since the complex structure moduli do not contain any axion and the bundle moduli B
no Wilson lines, the dependence on these variables does not have to be via an exponential factor.
In particular, there are no additional world-sheet instanton corrections in an E1- or E5-instanton
sector.
4.2. Heterotic models
The type I setup sketched in the previous section is S-dual to the Spin(32)/Z2 heterotic
string [44], more concretely to a specific embedding of products of U (N ) gauge bundles into
Spin(32)/Z2 [45,46]. A very similar story is likewise expected in the context of E8 E8
134
heterotic compactifications with anomalous U (1) factors as they generically appear upon embedding line or U (N ) vector bundles into E8 E8 [3,47]. In the heterotic picture one has
world-sheet instantons and Euclidean 5-brane instantons with the general moduli dependence
being as in (47). The concrete computation is different, of course, since there are no D-branes
and therefore no disc diagrams to be computed. In particular for the heterotic 5-brane instantons, it is not clear how a concrete computation can be performed, as the vertex operators are not
known.
World-sheet instantons have been discussed in a couple of papers [15]. For certain kinds of
bundles, namely those which are described in the (half) linear sigma model approach, the authors
of [28,4850] have been able to prove that the various contributions to the superpotential exactly
cancel. It remains to be seen whether such strong statements can also be made for certain classes
of for instance intersecting D6-brane models.
4.3. Type IIB orientifolds with O7/O3-planes
These orientifolds with D7- and D3-branes have extensively been studied recently, in particular as they allow a very clean introduction of flux compactifications with frozen complex
structure and dilaton moduli. Most importantly, E3-brane instantons can also freeze the Khler
moduli [17], as applied e.g. in the KKLT scenario [51]. Before we come back to this latter point,
let us first discuss the general framework of instanton effects.
Similarly to the type I case, the D7-branes wrap holomorphic divisors, i.e. four-cycles a
in X and also carry non-trivial bundles Va . Here the dimensionally reduced RR four-form C4
provides the four-dimensional axions to be gauged under the U (1)a gauge groups living on the
stacks of D7-branes.
Therefore, the role of the instantons generating the charged matter couplings is played by
E3-branes wrapping holomorphic four-cycles E on the CY threefold. Again in general these E3branes can be equipped with holomorphic stable vector bundles VE . To avoid neutral extra zero
modes, the four-cycles E should be rigid, i.e. h0 (E , N) = 0, which means h(2,0) (E ) = 0. In
addition there can be bundle moduli, which for line bundles VE = L are absent if h(1,0) (E ) = 0.
There can also be 4D spacetime filling D3-branes and E(1)-instantons, but these lead to nonchiral matter and only give rise to massless modes if the branes are localised at the same point in
the transversal space.
Additional charged zero modes are given by non-trivial intersections of the E3-branes with
the D7-branes, which can then be compensated by extra matter fields in the instanton correlator.
It has already been anticipated in [32] that if a D3-brane comes together with an E3-brane, there
appears a pair 3 , 3 of fermionic zero modes. As has been pointed out in the literature [52],
for the generalised KKLT [51] scenario proposed in [53] the extra fermionic zero modes pose a
problem with the up-lifting mechanism by magnetised D7-branes. Since in this simplified model
the CY only has one four-cycle, both the E3-instanton as well as the uplifting D7-brane are
wrapping the same four-cycle, leading necessarily to extra charged zero modes on the instanton
world-volume. Therefore, in this model the E3-brane does not generate a purely exponential
term in the superpotential; the induced non-perturbative terms rather involve appropriate extra
charged matter fields (see [20] for a similar discussion about a non-perturbative superpotential
via gaugino condensation on D7-branes).
To conclude, not only the background fluxes have the potential to change the zero mode
structure on the E3-instanton [5456], but likewise possibly present D7- and D3-branes.
135
5. Phenomenological implications
Building upon our analysis in Section 3, we now point out some immediate phenomenological
consequences of these instanton generated terms in the superpotential. The first one concerning
a new destabilisation mechanism is rather worry-some. Our second observation on the generation of perturbatively absent couplings is relevant for proton decay, -terms in the Higgs sector
and right-handed neutrino masses in MSSM-like intersecting D6-brane models. In the latter two
cases, instantons provide a stringy mechanism to explain hierarchies.
5.1. Vacuum destabilisation
As explained, for all known classes of N = 1 supersymmetric four-dimensional string vacua
spacetime instanton effects have the potential to not only generate new couplings in the theory
which break the global U (1) symmetries, but they can also generate a superpotential leading
(without any other stabilising effects) to a runaway behaviour for some of the scalar modes. That
means that such terms generate a tadpole for an auxiliary F -field and therefore break supersymmetry.
Clearly, those terms can be generated if the E2-instanton does not carry any additional uncharged and charged fermionic zero modes, i.e. if it is rigid and does not intersect any of the
D6-branes. This is the analogy of the condition for the heterotic string of Distler/Greene [3] that
there be no additional left-moving fermionic zero modes on the world-sheet instanton wrapping
an isolated curve in the CY. However, we would like to investigate the question whether instantons which carry additional zero modes cannot destabilise the vacuum as well.
First, if there are the right number of charged fermionic zero modes with just two U (1) stacks
of D6-branes, a tadpole of the form
W = a,b eSE2
(48)
for a charged matter field can be generated. Expanding around a,b = 0 we see that even though
W vanishes there, W does not and therefore supersymmetry is broken.
Second, one can imagine the situation depicted in Fig. 7, where a pair of charged fermionic
zero modes is soaked up by for instance an open string modulus a of a D6-brane. All these
terms lead to an instability for the D6-brane modulus such that the D6-branes gets pulled away
from the E2-brane in order to minimise the energy of the disc spanned between them. Here
again a W does not vanish. The effect of the resulting non-perturbative F-term for the open
Fig. 7. Disc instability.
136
string moduli is expected to be comparable to the one analysed in [57] on the mirror symmetric
type IIB side.
From these results we draw the conclusion that the presence of additional zero modes alone
is not a sufficient condition for the stability of the background. Additional zero modes can be
soaked up in a way that still leads to a destabilisation of the model. This picture is consistent
with the explicitly computed heterotic superpotential in [5]. There the one-loop determinants are
indeed (vector bundle) moduli dependent and their zeros are precisely at the locations where
extra massless modes appear. However, due to the moduli dependence, W could nevertheless
be non-vanishing, and it is precisely the derivatives of the superpotential which our correlation
functions probe for.
In general one expects such destabilising terms not to be completely absent and it is far from
obvious that summing up all these terms lead to unexpected cancellations like for world-sheet
instantons in the context of certain heterotic vacua [28,48,50]. At least, for toroidal models or the
Z2 Z2 orbifold in [58,59] one seems to be on the safe side as all sLag three-cycles known to
us are non-rigid.
5.2. Instanton generated matter couplings
To be a little bit more concrete, let us consider MSSM like models of intersecting D6-branes.
It is known that one economic way to realise such a model is to start with four stacks of D-branes
with initial gauge symmetry [60]
U (3) Sp(2) U (1) U (1).
(49)
The Standard Model matter fields can be realised by bifundamental fermions as shown in Table 3. Note that in this case there are no chiral fermions transforming in the symmetric or
anti-symmetric representations of the unitary gauge symmetries. We also assume the existence
of one vector-like pair of weak Higgs doublet H from the (bc) sector. In such a model one has
three U (1) gauge symmetries, of which the linear combination
1
U (1)Y = U (1)a U (1)c + U (1)d
(50)
3
is required to stay massless, i.e. it is in the kernel of the matrix Q in (15).
Since we assume that we get just the MSSM spectrum from such a brane configuration, we
require furthermore that all brane volumes are equal. This means that our vacuum really exhibits
gauge coupling unification at the GUT scale, which in type IIA we can identify with the string
scale by tuning gs accordingly. Moreover, we can also choose the KaluzaKlein scale to be the
string scale. Suppose now that we can find a completely perfect intersecting D-brane model.
Table 3
Chiral massless spectrum MSSM four stack model. ()c denotes the charge conjugated field
Intersection
Matter
Rep.
(a, b)
(a, c)
(a , c)
(b, d)
(c, d)
(c , d)
QL
(UR )c
(DR )c
LL
(ER )c
(NR )c
3 (3, 2)(1,0,0)
3 (3, 1)(1,1,0)
3 (3, 1)(1,1,0)
3 (1, 2)(0,0,1)
3 (1, 1)(0,1,1)
3 (1, 1)(0,1,1)
1/3
4/3
2/3
1
2
0
137
Since U (1)Y is required to stay massless, the net hypercharge of all fermionic zero modes is
guaranteed to vanish. This is simply a consequence of the fact that the linear combination (50)
is in the kernel of the matrix Q in (15). Therefore it is clear that all instanton-generated chiral
matter couplings in the superpotential respect the gauge symmetries of the Standard Model. Only
the global symmetries like baryon number QB = Qa or lepton number QL = Qd can be broken.
For couplings which are not forbidden in perturbation theory, i.e. those respecting all global
U (1) symmetries, the non-perturbative effects are expected to give merely small corrections to
the perturbative values. However, for couplings generated only non-perturbatively, these are the
leading order effects and might have interesting phenomenological consequences.
For instance, if the zero modes match, a dangerous tadpole can only be generated for the
right-handed neutrino. More generally, all neutrino couplings

n
W (NR )c eSE2 ,
n = 0, 1, 2, . . . ,
(51)
can in principle be generated, including also a Majorana mass term. We will discuss such terms
in more detail in Section 5.4.
Another important coupling which is absent in perturbation theory is a -term H + H
coupling, which involves the Standard Model Higgs pairs H that originate in an N = 1 chiral sector, e.g. as in systematic constructions of three-family supersymmetric Standard Models
in [61]. Here, suitable instantons can likewise induce exponentially suppressed couplings and
therefore explain the hierarchically small -term for appropriate choices of the three-cycle volumes.
We would also like to point out that, however, in other situations the restrictive fermionic
zero-mode structure prohibits the generation of perturbatively forbidden, but phenomenologically desirable terms such as the tri-linear Yukawa couplings 10 10 5H in certain intersecting
brane grand unified SU(5) models.
Since exemplifying these two latter points would require some discussion of the explicit realisation of the involved matter fields in the various setups, we prefer to return to our prototype
MSSM spectrum in Table 3 and analyse the possibility of generating non-perturbative potentially dangerous dimension-four proton decay couplings and Majorana masses in more detail.
The discussion of the other couplings in slightly different settings will be exactly along these
lines.
5.3. Proton decay
In this paper we do not intend to give a complete discussion about proton decay, but only
discuss the dimension-four couplings

W4 = QL (DR )c LL + (UR )c (DR )c (DR )c + LL LL (ER )c ,
(52)
which are usually absent in the MSSM due to R-parity. Taking into account that each matter field
soaks up two charged fermionic zero modes and that the number of these zero modes scales with
the number of branes Na in a stack, one can see that the first coupling [QL (DR )c LL ] cannot
be generated by E2-instantons. The second coupling for instance can be generated by an E2instanton wrapping a three-cycle with the non-vanishing intersections
[ a ]+ = 1,
[ c ] = 1,
[ c ] = 2
(53)
138
giving rise to six fermionic zero modes. Of course, the last two couplings are suppressed by the
instanton action
e
2
3s gs
VolE2
(54)
However, proton decay due to the dimension four couplings above always involves a contribution from the first coupling. In fact, the phenomenological bounds from these couplings always
involve the product [62] and therefore, even after taking the E2-instanton corrections into
account, the proton is stable in the MSSM intersecting D-brane model. Other phenomenological
bounds for baryon and lepton number violating processes induced by the dimension four couplings and require a rather mild exponential suppression [63], which should be naturally
generated by appropriate instanton actions.
5.4. Neutrino masses
The other sector where non-perturbative effects are important is the neutrino sector. As we
have seen, these MSSM-like intersecting D6-brane models automatically contain right-handed
neutrinos so that one can contemplate implementing the seesaw mechanism in this setup.7
First the Dirac mass terms are generated perturbatively via the weak Higgs effect
WH = H + LL (NR )c ,
(55)
so that as usual one expects these mass terms to be of the order of the quark/lepton masses. Since
LL carries non-vanishing U (1)Y charge, there are no E2-instanton contributions to the Majorana
mass of the left-handed neutrinos. There might be higher-dimensional couplings but these will
be further suppressed with Ms .
On the contrary, as we discussed, E2-instanton effects can generate Majorana type masses for
the right-handed neutrinos
Wm = Mm (NR )c (NR )c
(56)
with the mass scale

Mm = xMs e
2
3s gs
VolE2
(57)
where x summarises the contributions from the Khler potential (normalisation of kinetic terms)
and the one-loop determinant. One does not expect these contributions to be exponentially suppressed or enhanced so we assume that x is of order O(1). Such a mass-term satisfies all selection
rules if the intersection numbers of the rigid E2-instanton with the MSSM branes are
a = a = b = b = c = d = 0,
[ c ]+ = 2,
[ c ] = 0,
[ d ] = 2,
[ d ]+ = 0.
(58)
Note that even though instanton effects are genuinely small, they are small compared to the
natural scale in the problem, which here is the stringy mass scale, i.e. Ms .8
7 For an alternative proposal for small left-handed neutrino Majorana-like masses within SU(5) grand unified models
with intersecting D6-branes, see [64].

8 Our whole argument here relies on the assumption that the overall mass scale of this term is M . However, we only
s
see one other stringy mass scale in the problem and that is the KK scale, which was also assumed to be of order Ms .
Due to the purely stringy origin of this term, the weak scale is not a natural scale to appear here, as it only appears after
supersymmetry breaking.
139
Thus, we have now all ingredients together for the seesaw mechanism to work. The exponential suppression from the instanton action allows one, without any fine-tuning, to explain a
possible hierarchy between the string or GUT scale and the Majorana mass scale, which phenomenologically lies in the range 1011 GeV < Mm < 1015 GeV.
To get a feeling of the scales involved let us assume that one really found a four-stack intersecting D-brane model, which gives just the MSSM and features gauge coupling unification at
the usual GUT scale of MGUT = 2 1016 GeV, i.e. all the volumes of the three-cycle wrapped
by the MSSM branes are equal. Then the gauge coupling at the GUT scale is given by
1
GUT
1
3s gs
(59)
VolD6
which implies the following relation for the neutrino masses

Mm = xMs e
VolE2
GUT VolD6
(60)
Assuming x = 1 and isotropic three-cycles, the phenomenological bounds on the Majorana mass
scale implies the following moderate bounds for the length scales of the three-cycles wrapped by
the MSSM branes and the instanton
0.4 RD6 > RE2 > 0.27 RD6 .
(61)
6. Conclusions
For essentially all known phenomenologically interesting classes of supersymmetric four-dimensional string compactifications we have investigated the effects of non-perturbative spacetime instantons on the superpotential couplings for the charged matter superfields. We pointed
out that the formal global U (1) charges carried by the exponential instanton factor are due to
additional fermionic zero modes on the world-volume of the instanton which arise from its intersection with the other spacetime filling D-branes in the model. We provided an outline of the
concrete computation of such instanton amplitudes and argued that eventually it boils down to the
computation of open string disc diagrams for boundary changing operators. These are multiplied
by the one-loop vacuum diagram, where only those open string sectors preserving exactly two
supercharges contribute. We briefly pointed out some further applications of these methods including the computation of multi-instanton contributions and the important spacetime instanton
effects modifying the D-terms and the gauge kinetic functions. Whereas we argued that these latter non-perturbative corrections to the supersymmetry conditions are not expected to destabilise
the vacuum drastically, it is nonetheless important to analyse their effects more quantitatively.
In particular, it remains for future work to investigate possible consequences on the statistical
distribution of supersymmetric brane vacua in the type IIA geometric regime [43,65,66] or at the
small radius Gepner point [67,68].
We explained how E2-instanton effects in type IIA orientifolds are mapped via T-duality to
essentially E1/E5- and E3-instanton effects in type IIB orientifolds. These latter effects have
been important for recent discussions on moduli stabilisation and we pointed out that our general
investigation forbids the D-term uplifting mechanism in the KKLT scenario.
Finally, we applied this framework to MSSM like intersecting D-brane models and argued
that besides destabilising terms also new couplings violating the global U (1) symmetries can be
generated. Such couplings can induce Majorana mass terms for the right-handed neutrinos, thus
providing the essential ingredient for the seesaw mechanism.
140
In this paper we focused mainly on single instantons which do not have any additional bosonic
zero modes since they are wrapping rigid cycles in the internal CY manifold. Instantons carrying
additional zero modes can result in more general superpotential couplings presumably mixing
charged matter and closed string moduli fields. Moreover, here we have only presented the general description of how such instanton computations can be carried out in principle. It would be
interesting to apply these methods to concrete exactly solvable conformal field theory models
like toroidal orbifold and Gepner models.
It is important to see whether the heterotic instanton sum rules of [50] can be generalised
to D-brane models, thus ensuring that they are not destabilised by instantons. Of course combining these non-perturbative superpotential terms with flux generated terms can also lead to a
stabilisation of (all) moduli in certain string models.
The F- and D-term instanton corrections discussed in this paper point towards a finite gs generalisation of N = 1 mirror symmetry including not only world-sheet but also spacetime instanton
corrections. Such a mathematical structure, involving a spacetime non-perturbative extension of
Fukaya/derived categories, seems to underly a rigorous approach to the string vacuum problem
and eventually the discipline of what should really be called string phenomenology.
Acknowledgements
We gratefully acknowledge discussions with Ron Donagi, Michael Haack, Shamit Kachru,
Paul Langacker, Tao Liu, Dieter Lst, Tony Pantev, Michael Plmacher, Robert Richter, Maximilian Schmidt-Sommerfeld and Marco Zagermann. R.B. and M.C. would like to thank the KITP
at the University of California at Santa Barbara for hospitality. This research was supported in
part by the National Science Foundation under Grant No. PHY99-07949, the Department of Energy Grant DOE-EY-76-02-3071, the Fay R. and Eugene L. Langberg Endowed Chair and the
National Science Foundation Grant INT02-03585 (M.C. and T.W.).
References
[1] M. Dine, N. Seiberg, X.G. Wen, E. Witten, Nonperturbative effects on the string world sheet, Nucl. Phys. B 278
(1986) 769.
[2] M. Dine, N. Seiberg, X.G. Wen, E. Witten, Nonperturbative effects on the string world sheet (II), Nucl. Phys. B 289
(1987) 319.
[3] J. Distler, B.R. Greene, Aspects of (2, 0) compactifications, Nucl. Phys. B 304 (1988) 1.
[4] E. Witten, World-sheet corrections via D-instantons, JHEP 0002 (2000) 030, hep-th/9907041.
[5] E.I. Buchbinder, R. Donagi, B.A. Ovrut, Superpotentials for vector bundle moduli, Nucl. Phys. B 653 (2003) 400
420, hep-th/0205190.
[6] E.I. Buchbinder, R. Donagi, B.A. Ovrut, Vector bundle moduli superpotentials in heterotic superstrings and Mtheory, JHEP 0207 (2002) 066, hep-th/0206203.
[7] S. Kachru, S. Katz, A.E. Lawrence, J. McGreevy, Open string instantons and superpotentials, Phys. Rev. D 62
(2000) 026001, hep-th/9912151.
[8] S. Kachru, S. Katz, A.E. Lawrence, J. McGreevy, Mirror symmetry for open strings, Phys. Rev. D 62 (2000) 126005,
hep-th/0006047.
[9] M. Aganagic, C. Vafa, Mirror symmetry, D-branes and counting holomorphic discs, hep-th/0012041.
[10] K. Becker, M. Becker, A. Strominger, Five-branes, membranes and nonperturbative string theory, Nucl. Phys. B 456
(1995) 130152, hep-th/9507158.
[11] J.A. Harvey, G.W. Moore, Superpotentials and membrane instantons, hep-th/9907026.
[12] K. Belani, P. Kaura, A. Misra, Supersymmetry of noncompact MQCD-like membrane instantons and heat kernel
asymptotics, hep-th/0603063.
[13] M.B. Green, M. Gutperle, Effects of D-instantons, Nucl. Phys. B 498 (1997) 195227, hep-th/9701093.
[14] M. Gutperle, Aspects of D-instantons, hep-th/9712156.
141
[15] M. Billo, et al., Classical gauge instantons from open strings, JHEP 0302 (2003) 045, hep-th/0211250.
[16] A.-K. Kashani-Poor, A. Tomasiello, A stringy test of flux-induced isometry gauging, Nucl. Phys. B 728 (2005)
135147, hep-th/0505208.
[17] E. Witten, Non-perturbative superpotentials in string theory, Nucl. Phys. B 474 (1996) 343360, hep-th/9604030.
[18] R. Blumenhagen, M. Cvetic, P. Langacker, G. Shiu, Toward realistic intersecting D-brane models, hep-th/0502005.
[19] B. Florea, S. Kachru, J. McGreevy, N. Saulina, Stringy instantons and quiver gauge theories, hep-th/0610003.
[20] M. Haack, D. Krefl, D. Lust, A. Van Proeyen, M. Zagermann, Gaugino condensates and D-terms from D7-branes,
hep-th/0609211.
[21] J. McGreevy, On the capture of runaway quivers, in: Talk given at Strings, Beijing, 2006, http://strings06.itp.ac.cn/
talk-files/mcgreevy.pdf.
[22] R. Blumenhagen, V. Braun, B. Krs, D. Lst, Orientifolds of K3 and CalabiYau manifolds with intersecting Dbranes, JHEP 0207 (2002) 026, hep-th/0206038.
[23] G. Aldazabal, S. Franco, L.E. Ibanez, R. Rabadan, A.M. Uranga, D = 4 chiral string compactifications from intersecting branes, J. Math. Phys. 42 (2001) 31033126, hep-th/0011073.
[24] I.R. Klebanov, E. Witten, Proton decay in intersecting D-brane models, Nucl. Phys. B 664 (2003) 320, hepth/0304079.
[25] R. Blumenhagen, D. Lst, S. Stieberger, Gauge unification in supersymmetric intersecting brane worlds, JHEP 0307
(2003) 036, hep-th/0305146.
[26] A. Font, L.E. Ibanez, F. Marchesano, Coisotropic D8-branes and model-building, hep-th/0607219.
[27] J. Polchinski, Combinatorics of boundaries in string theory, Phys. Rev. D 50 (1994) 60416045, hep-th/9407031.
[28] C. Beasley, E. Witten, New instanton effects in string theory, JHEP 0602 (2006) 060, hep-th/0512039.
[29] R. Blumenhagen, M. Cvetic, F. Marchesano, G. Shiu, Chiral D-brane models with frozen open string moduli,
JHEP 0503 (2005) 050, hep-th/0502095.
[30] R. Blumenhagen, E. Plauschinn, Intersecting D-branes on shift Z(2) Z(2) orientifolds, JHEP 0608 (2006) 031,
hep-th/0604033.
[31] M. Berkooz, M.R. Douglas, R.G. Leigh, Branes intersecting at angles, Nucl. Phys. B 480 (1996) 265278, hepth/9606139.
[32] O.J. Ganor, A note on zeroes of superpotentials in F-theory, Nucl. Phys. B 499 (1997) 5566, hep-th/9612077.
[33] M. Cvetic, I. Papadimitriou, Conformal field theory couplings for intersecting D-branes on orientifolds, Phys. Rev.
D 68 (2003) 046001, hep-th/0303083.
[34] M. Cvetic, R. Richter, Proton decay via dimension-six operators in intersecting D6-brane models, hep-th/0606001.
[35] S.B. Giddings, A. Maharana, Dynamics of warped compactifications and the shape of the warped landscape, Phys.
Rev. D 73 (2006) 126003, hep-th/0507158.
[36] I. Brunner, M.R. Douglas, A.E. Lawrence, C. Romelsberger, D-branes on the quintic, JHEP 0008 (2000) 015, hepth/9906200.
[37] R. Mc Lean, Deformations of calibrated submanifolds, Commun. Anal. Geom. 6 (1998) 705.
[38] A. Recknagel, V. Schomerus, Boundary deformation theory and moduli spaces of D-branes, Nucl. Phys. B 545
(1999) 233282, hep-th/9811237.
[39] D. Cremades, L.E. Ibanez, F. Marchesano, Yukawa couplings in intersecting D-brane models, JHEP 0307 (2003)
038, hep-th/0302105.
[40] M.R. Douglas, B. Fiol, C. Romelsberger, Stability and BPS branes, hep-th/0002037.
[41] I. Affleck, M. Dine, N. Seiberg, Dynamical supersymmetry breaking in supersymmetric QCD, Nucl. Phys. B 241
(1984) 493534.
[42] R. Blumenhagen, G. Honecker, T. Weigand, Non-Abelian brane worlds: The open string story, hep-th/0510050.
[43] M.R. Douglas, W. Taylor, The landscape of intersecting brane models, hep-th/0606109.
[44] E. Witten, Some properties of O(32) superstrings, Phys. Lett. B 149 (1984) 351356.
[45] R. Blumenhagen, G. Honecker, T. Weigand, Supersymmetric (non-)Abelian bundles in the type I and SO(32) heterotic string, JHEP 0508 (2005) 009, hep-th/0507041.
[46] R. Blumenhagen, G. Honecker, T. Weigand, Non-Abelian brane worlds: The heterotic string story, JHEP 0510
(2005) 086, hep-th/0510049.
[47] R. Blumenhagen, G. Honecker, T. Weigand, Loop-corrected compactifications of the heterotic string with line bundles, JHEP 0506 (2005) 020, hep-th/0504232.
[48] E. Silverstein, E. Witten, Criteria for conformal invariance of (0, 2) models, Nucl. Phys. B 444 (1995) 161190,
hep-th/9503212.
[49] A. Basu, S. Sethi, World-sheet stability of (0, 2) linear sigma models, Phys. Rev. D 68 (2003) 025003, hepth/0303066.
142
[50] C. Beasley, E. Witten, Residues and world-sheet instantons, JHEP 0310 (2003) 065, hep-th/0304115.
[51] S. Kachru, R. Kallosh, A. Linde, S.P. Trivedi, De Sitter vacua in string theory, Phys. Rev. D 68 (2003) 046005,
hep-th/0301240.
[52] A. Achucarro, B. de Carlos, J.A. Casas, L. Doplicher, de Sitter vacua from uplifting D-terms in effective supergravities from realistic strings, JHEP 0606 (2006) 014, hep-th/0601190.
[53] C.P. Burgess, R. Kallosh, F. Quevedo, De Sitter string vacua from supersymmetric D-terms, JHEP 0310 (2003) 056,
hep-th/0309187.
[54] E. Bergshoeff, R. Kallosh, A.-K. Kashani-Poor, D. Sorokin, A. Tomasiello, An index for the Dirac operator on D3
branes with background fluxes, JHEP 0510 (2005) 102, hep-th/0507069.
[55] D. Lst, P. Mayr, S. Reffert, S. Stieberger, F-theory flux, destabilization of orientifolds and soft terms on D7-branes,
Nucl. Phys. B 732 (2006) 243290, hep-th/0501139.
[56] D. Lst, S. Reffert, W. Schulgin, P.K. Tripathy, Fermion zero modes in the presence of fluxes and a non-perturbative
superpotential, hep-th/0509082.
[57] D. Baumann, et al., On D3-brane potentials in compactifications with fluxes and wrapped D-branes, hep-th/0607050.
[58] M. Cvetic, G. Shiu, A.M. Uranga, Chiral four-dimensional N = 1 supersymmetric type IIA orientifolds from intersecting D6-branes, Nucl. Phys. B 615 (2001) 332, hep-th/0107166.
[59] M. Cvetic, G. Shiu, A.M. Uranga, Three-family supersymmetric standard like models from intersecting brane
worlds, Phys. Rev. Lett. 87 (2001) 201801, hep-th/0107143.
[60] D. Cremades, L.E. Ibanez, F. Marchesano, Towards a theory of quark masses, mixings and CP-violation, hepph/0212064.
[61] M. Cvetic, T. Li, T. Liu, Supersymmetric PatiSalam models from intersecting D6-branes: A road to the standard
model, Nucl. Phys. B 698 (2004) 163201, hep-th/0403061.
[62] P. Nath, P.F. Perez, Proton stability in grand unified theories, in strings, and in branes, hep-ph/0601023.
[63] B.C. Allanach, A. Dedes, H.K. Dreiner, The R parity violating minimal supergravity model, Phys. Rev. D 69 (2004)
115002, hep-ph/0309196.
[64] M. Cvetic, P. Langacker, New grand unified models with intersecting D6-branes, neutrino masses, and flipped
SU(5), hep-th/0607238.
[65] R. Blumenhagen, F. Gmeiner, G. Honecker, D. Lust, T. Weigand, The statistics of supersymmetric D-brane models,
Nucl. Phys. B 713 (2005) 83135, hep-th/0411173.
[66] F. Gmeiner, R. Blumenhagen, G. Honecker, D. Lust, T. Weigand, One in a billion: MSSM-like D-brane statistics,
JHEP 0601 (2006) 004, hep-th/0510170.
[67] T.P.T. Dijkstra, L.R. Huiszoon, A.N. Schellekens, Supersymmetric standard model spectra from RCFT orientifolds,
Nucl. Phys. B 710 (2005) 357, hep-th/0411129.
[68] P. Anastasopoulos, T.P.T. Dijkstra, E. Kiritsis, A.N. Schellekens, Orientifolds, hypercharge embeddings and the
standard model, hep-th/0605226.
Nuclear Physics B 771 [PM] (2007) 143166
The Langlands program and string modular K3 surfaces

Rolf Schimmrigk
Indiana University South Bend, 1700 Mishawaka Ave., South Bend, IN 46634, USA
Received 4 May 2006; accepted 5 January 2007
Available online 31 January 2007
Abstract
A number theoretic approach to string compactification is developed for CalabiYau hypersurfaces in
arbitrary dimensions. The motivic strategy involved is illustrated by showing that the Hecke eigenforms
derived from Galois group orbits of the holomorphic two-form of a particular type of K3 surface can be
expressed in terms of modular forms constructed from the worldsheet theory. The process of deriving string
physics from spacetime geometry can be reversed, allowing the construction of K3 surface geometry from
the string characters of the partition function. A general argument for K3 modularity is given by combining
mirror symmetry with the proof of the ShimuraTaniyama conjecture.
PACS: 11.25.-w; 11.25.Hf; 11.25.Mj
MSC: 11G25; 11G40; 14G10; 14G40
Keywords: Varieties over finite fields; L-functions; Zeta functions; Arithmetic varieties; Fundamental strings;
Conformal field theory; Compactification
1. Introduction
One of the intriguing aspects of string theory is the possibility of understanding the structure of
spacetime in terms of the physics of the worldsheet. In the past, a number of different techniques
have been used to address this question, e.g. LandauGinzburg theories and non-linear sigma
models. The aim of the present paper is to continue a different program that uses methods from
arithmetic geometry to understand this problem in the context of exactly solvable CalabiYau
varieties. The structure of higher dimensional CalabiYau varieties is quite intricate, and there
are several aspects of these theories that lend themselves to an arithmetic analysis. The focus
E-mail addresses: netahu@yahoo.com, rschimmr@iusb.edu.
doi:10.1016/j.nuclphysb.2007.01.027
144
R. Schimmrigk / Nuclear Physics B 771 [PM] (2007) 143166
in this paper will be on the formulation of a framework that combines methods from algebraic
number theory and arithmetic geometry in the context of CalabiYau hypersurfaces of arbitrary
dimensions. This general approach is then applied to a special class of K3 surfaces by showing
that basic building blocks of the underlying string partition functions can be derived from the
geometry of these K3s.
The simplest case in which the idea of using arithmetic geometry to derive conformal field
theoretic information from geometry can be tested is provided by the framework of toroidal
compactifications. Elliptic curves provide useful examples because of the proof of the Shimura
Taniyama conjecture. This modularity theorem states that all elliptic curves defined over the
rational number field are modular in the sense that the inverse Mellin transform of their Hasse
Weil L-series defines a modular form of weight two with respect to some congruence subgroup of
SL(2, Z) [14]. In the context of exactly solvable elliptic curves the issue then becomes whether
the modular forms derived from these curves can be expressed in terms of modular forms derived
from the underlying superconformal field theory. It was shown in Refs. [57] that this is possible
in terms of cusp forms constructed from the string functions associated to the affine Lie algebra
(1)
A1 of the N = 2 superconformal minimal models.
The generalization of this result to higher genus curves and higher dimensional varieties is
made difficult by the fact that no analog of the elliptic modularity theorem is known, even conjecturally. There is a general expectation, often summarized as part of the Langlands program
[8], that many varieties will lead to modular forms, but there is no known systematic procedure
which provides guidance on how this should be accomplished. There is in particular no known
higher dimensional analog of Weils experimental observation [9] concerning the geometric interpretation of the level of a modular form in the context of elliptic curves, an ingredient that
galvanized interest in the ShimuraTaniyama conjecture, culminating in Wiles breakthrough.
For higher genus curves of BrieskornPham type the situation is a little better because one can
use a known factorization procedure to decompose the associated Jacobian variety into simple
Abelian factors, which then can be tested for modularity [6].
In general the L-function of a variety will not be an interesting object in the context of a search
for string theoretic modular forms. Except for particularly simple varieties, such as rigid Calabi
Yau threefolds, for which many modular forms have been identified (Ref. [10] contains articles
with many further references), the L-function itself does not lead to a modular form. It is known,
from Grothendiecks proof [11] of part of the Weil conjectures [12], that Artins congruence zeta
function factorizes in a way that is determined by the cohomology groups of the variety. The zeta
function at a prime p decomposes into the quotient
Z(X/Fp , t) =
Pp1 (t)Pp3 (t) Pp2n1 (t)

Pp0 (t)Pp2 (t) Pp2n (t)
(1)
where dimC X = n, and Ppi (t) is a polynomial

i
Ppi (t) =
b
ji (p)t j
(2)
j =0
associated to the ith cohomology group, with degree bi = dim Hi (X) given by the ith Betti number. This result motivates the introduction of L-functions associated to the individual cohomology groups, thereby reducing the complexity of the zeta function. Even though this factorization
145
provides an important simplification, it is not enough for string theory. The individual cohomology groups can be quite complicated because the Betti numbers of CalabiYau varieties tend
to be large. The idea in this paper is to factorize these polynomials further, and to consider Lfunctions associated to the resulting factors. The problem that arises is that it is unclear a priori
which type of factorization leads to a physically meaningful L-function.
In the absence of a clear understanding of what the conditions are in higher dimensions that
can lead to string theoretic modular forms on the worldsheet, it is useful to identify selection
rules that can guide the factorization of the L-functions. There are several ways to think about
this problem, and in the present paper the following point of view will be adopted. The first idea
is utilitarian in nature, guided by the expectation that the results of [57], or some not too radical
modification thereof, will generalize to higher dimensions. It follows from those results that the
string theoretic modular forms relevant for the arithmetic approach are forms with coefficients
that are rational integers. The primitive factors of the polynomials Ppi (t) that arise from the
complete factorization
i
Ppi (t) =
b

1 ji (p)t ,
(3)
j =1
lead to L-function factors with coefficients ji (p) that are algebraic integers in number fields,
defined by extensions of the rational number field Q. This shows that a complete factorization is
not useful in the present context, and that from a practical point of view one should be guided by
the idea of identifying pieces of the cohomology that lead to forms with coefficients in Z.
A more conceptual way of thinking about this problem is representation theoretic. Associated
to a CalabiYau hypersurface in a toric variety is a cyclotomic number field, which for Brieskorn
Pham spaces admits an interpretation as the fusion field of the underlying conformal field theory
[13]. The Galois group of this field is a finite cyclic group which acts on the cohomology of
the variety. The action of this group is reducible in general, and therefore one can use it to
decompose the cohomology group into pieces defined by the irreducible representations of the
group. A possible strategy therefore is to focus on the L-functions associated to these irreducible
representations of the Galois group. A distinguished element in the cohomology ring of any ndimensional CalabiYau variety is the holomorphic n-form, denoted by , leading to the concept
of an -motive of a CalabiYau variety. The notion of such a motive is general, and the question
addressed here is whether this motive is string modular in some sense. The approach formulated
therefore provides a general method to gain a better understanding of the relation between the
geometry of spacetime in string theory and the physics on the worldsheet.
In the present paper the strategy just described is illustrated by considering the class of extremal K3 surfaces of BrieskornPham type, i.e. surfaces S which over the complex field C have
maximal Picard number (S) = 20. These surfaces are of the form

S 4 = (z0 : :z3 ) P3 z04 + z14 + z24 + z34 = 0 ,

S 6A = (z0 : :z3 ) P(1,1,1,3) z06 + z16 + z26 + z32 = 0 ,

S 6B = (z0 : :z3 ) P(1,1,2,2) z06 + z16 + z23 + z33 = 0 .
(4)
In order
the results of this analysis some
is needed. With q = e2i , let
to state
notation
n
2
n
f (q) = n an q be a cusp form, and f (q) = n bn q be the product given by bp = ap2
2p. The motivation for this definition will become clear below. Define the Hecke congruence
146
subgroup 0 (N ) as

a
0 (N ) =
c
b
d

a
SL(2, Z)
c
b
d
(mod N ) ,
(5)
and denote the Galois group of the cyclotomic number field Q(d ) by Gal(Q(d )/Q), where d
is the cyclic groupgenerated by a primitive dth root of unity. The Dedekind eta function is given
k
k
n
3
by (q) = q 1/24
n=1 (1 q ), while the theta functions ,m ( ) = ( )c,m ( ) are Hecke
k
indefinite modular forms associated to the KacPeterson string functions c,m ( ) of the affine
(1)
algebra A1 at level k. Finally, the quadratic characters determined by the Legendre symbol are
written as n (p) = ( pn ). (q) is a modular form of weight one described further below. The
following results will be shown.
Theorem 1. Let M H 2 (S d ) be the irreducible representation of Gal(Q(d )/Q) associated
to the holomorphic 2-form H 2,0 (S d ) of the K3 surface S d , where d = 4, 6A, 6B. Then the
inverse Mellin transforms f (S d , q) of the L-functions L (S d , s) associated to M are given
by

f S 4 , q = 6 q 4 ,

f S 6A , q = q 3 2 q 3 2 q 9 ,

f S 6B , q = 3 q 2 3 q 6 3 .
(6)
These functions are cusp forms of weight three with respect to 0 (N ) with levels 16, 27 and 48,
respectively. For S 4 and S 6A the L-functions can be written as L (S d , s) = L(2 fd , s), where
fd (q) are cusp forms of weight two and levels 64 and 27, respectively, given by
2
(4 )2 2 ,
f4 ( ) = 1,1
1
1
f6A ( ) = 1,1
(3 )1,1
(9 ).
(7)
For S 6B the L-series is given by L (S 6B , s) = L(2 f6B 3 , s) in terms of the cusp form of
level 144
1
f6B ( ) = 1,1
(6 )2 3 .
(8)
Physically, this result proves a string theoretic interpretation of the motivic L-function associ(1)
ated to the holomorphic -form of K3 surfaces in terms of the affine Lie algebra A1 on the
worldsheet. It generalizes results in lower dimensions for the HasseWeil L-function of elliptic
BrieskornPham curves obtained in [57]. Mathematically, it can be viewed as providing a motivic interpretation of modular forms derived from KacMoody algebras, i.e. it provides an affine
Lie-theoretic origin of modular motives for a class of K3s.
Corollary. The -motivic modular forms of extremal K3 surfaces of BrieskornPham type are
determined by twisted products of Hecke indefinite modular forms that arise from KacPeterson
string functions.
The modular forms of weight two that appear in the theorem as building blocks of the arithmetic structure of extremal BrieskornPham K3 surfaces are all elliptic, i.e. the inverse Mellin
147
transforms of HasseWeil L-series of elliptic curves. Consider the class of elliptic Brieskorn
Pham curves given by

E 3 = (z0 : z1 : z2 ) P2 z03 + z13 + z23 = 0 ,

E 4 = (z0 : z1 : z2 ) P(1,1,2) z04 + z14 + z22 = 0 ,

E 6 = (z0 : z1 : z2 ) P(1,2,3) z06 + z13 + z22 = 0 .
(9)
All three curves are string modular in the following sense [7].
Theorem 2. The inverse Mellin transforms fHW (E d , q) of the HasseWeil L-functions
LHW (E d , s) of the curves E d , d = 3, 4, 6, are modular forms fHW (E d , q) S2 (0 (N )), with
N = 27, 64, 144, respectively. These forms factor into products of Hecke indefinite modular
forms as follows
3 1 9

1
q 1,1 q ,
f E 3 , q = 1,1
4 2
4
2
2 ,
f E , q = 1,1 q
6

2
1
f E , q = 1,1
(10)
q 6 3 .
This shows that the string theoretic nature of the modular forms of extremal BrieskornPham
K3 surfaces is induced by the string theoretic modularity of elliptic BrieskornPham curves.
Given the central nature of modularity in string theory, it is natural to ask how general modularity is for K3 surfaces. An argument that points to modularity as a common property can be
made by combining mirror symmetry with the elliptic modularity theorem of [4]. The original
explicit observation of mirror symmetry [14,15] has been interpreted in a number of different
ways, e.g. in terms of toric geometry by Batyrev [16], in a homological context by Kontsevich
[17], and in terms of fibrations by Strominger et al. [18]. It is the latter framework that is most
useful for K3 modularity. The idea of the SYZ conjecture is based on a toroidal fibration structure of general CalabiYau varieties that is suggested by D-branes. For complex dimension two
this conjecture implies that mirror pairs of K3 surfaces are characterized by fibrations in terms
of elliptic curves. For elliptic curves defined over Q the modularity theorem proven by Breuil,
Conrad, Diamond and Taylor shows that every elliptic curve is modular in the sense that the
inverse Mellin transform of its associated HasseWeil L-function is a modular form of weight 2
with respect to a congruence group that is determined by the conductor of the curve. The SYZ
conjecture and the modularity theorem therefore imply that mirror pairs of K3 surfaces defined
over Q are modular.
The outline of the paper is as follows. Section 2 contains the arguments for considering irreducible Galois representations as the modular building blocks of CalabiYau varieties. These
arguments are general, not restricted to K3 surfaces. Section 3 computes the L-series of the extremal K3 surfaces of BrieskornPham type. Section 4 leads to the identification of the modular
forms derived from these K3 surfaces with modular forms derived from the string worldsheet,
completing the identifications of Theorem 1. Section 5 complements the previous computations
by constructing the K3 surfaces via the twist map [19]. The philosophy adopted here is similar
to the one used in [6] in the context of higher genus curves. Section 6 describes the identification of the modular form associated to a singular K3 surface in terms of string theoretic modular
forms. Section 7 generalizes to K3 surfaces aspects of complex multiplication and arithmetic
148
moonshine, discussed in [7] in the context of elliptic curves. Appendix A proves the identification of the geometric and the modular L-series to all orders. The proof is based on the method
of FaltingsSerreLivn, and uses results from the representation theory of the absolute Galois
group Gal(Q/Q)
of the rational numbers.
2. -motivic L-functions of CalabiYau varieties
2.1. Counting and Jacobi sums
L-functions of projective varieties can be computed in several ways. The most direct procedure starts from the rational form (1) of Artins congruence zeta function

tr
Z(X/Fp , t) = exp
(11)
#(X/Fpr )
,
r
r1
defined in terms of the degree r extensions Fpr of the finite field Fp for any prime p. This
factorization of the zeta function leads to the definition of cohomological L-functions

1
L(i) (X, s) =
(12)
Ppi (p s )
p prime
associated to the ith cohomology group H i (X).

The simplest way of computing this function is via a direct count of the number of solutions Nr,p = #(X/Fpr ). For the case of K3 surfaces the cohomological form of the zeta function
simplifies to
Z(X/Fp , t) =
1
(1 t)Pp2 (t)(1 p 2 t)
(13)
Expanding this quotient as

Z(X/Fp , t) = 1 + p 2 + 1 1 t + p 4 + p 2 (1 1 ) + 12 1 2 + 1 t 2 + ,
(14)
and comparing it with the expansion of the exponential, leads to
12 (p) = 1 + p 2 N1,p ,

1 2
22 (p) = 1 + p 2 + p 4 3 1 + p 2 N1,p + N1,p
N2,p ,
2
..
.
(15)
2
2
i
for the polynomial Pp (t) = i i (p)t .
For weighted Fermat varieties a second method to obtain both the cardinalities and the Lfunction was introduced by Weil [12]. This formulation is particularly useful because it allows
to disentangle the complicated cohomological structure of higher dimensional varieties in a systematic way, at least for this special type [20].
Theorem 3. For a smooth weighted projective surface with degree vector n = (n0 , . . . , ns )
n

X n = z0 0 + z1n1 + + zsns = 0 P(k0 ,k1 ,...,ks )
(16)
149
q,n
defined over the finite field Fq define the set As of rational vectors = (0 , 1 , . . . , s ) as

s

q,n
As = Qs+1 | 0 < i < 1, di = (ni , q 1), di i = 0 mod 1,
i = 0 (mod 1) .
i=0
(17)
For each (s + 1)-tuple define the Jacobi sum

1
0 (u0 )1 (u1 ) s (us ),
jq (0 , 1 , . . . , s ) =
q 1
(18)
ui Fq
u0 +u1 ++us =0
where i (ui ) = e2ii mi with integers mi determined via ui = g mi , where g Fq is a generator.

Then the cardinality of X n /Fq is given by

jq ().
# X n /Fq = N1,q X n = 1 + q + + q s1 +
(19)
q,n
As
With these ingredients the L-function associated to H s1 (X n ) takes the form (12) with the
polynomials Pps1 (t) expressed in terms of the Jacobi-sums variables as

1/f
Pps1 (t) =
(20)
1 (1)s1 jpf ()t f
n
As
where
n
As
= Q
s+1

s

0 < i < 1, ni i = 0 (mod 1),
i = 0 (mod 1)

(21)
i=0
and f is determined via

f
p 1 i = 0 (mod 1),
i.
(22)
The main point now is to identify an appropriate factor of the L-function of H s1 (X n ).

2.2. Galois representations and -motives of CalabiYau varieties
This section defines the notion of -motives for the class of CalabiYau hypersurfaces in
arbitrary dimensions. The Jacobi sums jpf () are algebraic numbers in the cyclotomic field
Q(d ), generated by primitive dth roots of unity. The factors (1 jpf ()t f )1/f therefore do
not themselves define string theoretic L-series of the type considered in [57] because the latter
have integral coefficients. The simplest way to produce real coefficients in the L-function is to
combine dual pairs of Jacobi sums, i.e. sums parametrized by (,
) such that +
= 1.
This sometimes leads to success, but in general the L-function of such pairs has coefficients
that are elements of the maximal real subfield of the cyclotomic field. The idea in this paper
is to achieve the necessary integrality of the coefficients by considering orbits of Jacobi sums
defined by the action of the multiplicative group (Z/dZ) , where d = ki ni , i, is the degree of
the hypersurface. The fact that these orbits lead to integral coefficients in the L-function can be
derived from a number theoretic result about finite extensions of number fields by noting that the
multiplicative groups (Z/dZ) can be interpreted as Galois groups of cyclotomic fields Q(d ).
150
This can be seen as follows (the relevant number theoretic concepts can be found in [21] and
[22]).
Jacobi sum orbits for the Galois group Gal(Q(d )/Q) = (Z/dZ) are obtained by defining
an action
(Z/dZ) As As
(23)
(t, ) = t (mod 1),
(24)
as
where t = (t0 , . . . , ts ). For any As one can therefore consider its orbit

n
O = As = t , t (Z/dZ) ,
(25)
and decompose the set As into a set of orbits

n
O ,
As =
(26)
where is a representative in each orbit.

To see that Galois orbits lead to L-series with rationally integral coefficients one observes that
the Jacobi sums jp () transform under the action of t (Z/dZ) as

jp (t ) = t jp () ,
(27)
where t Gal(Q(d )/Q) is defined as t (d ) = dt . Hence the Galois orbit leads to coefficients
ap of the terms with prime exponents that are of the form

jp () .
ap () =
(28)
Gal(Q(d )/Q)
The argument can now be completed by noting that the sum ap () is the trace of an algebraic
integer and by using a result from the theory of finite extensions. For any element x L in a
finite extension L/K over a number field K Dedekind defined the map
Tx : L L,
y xy,
(29)
which can be used to define a trace map

TrL/K (x) := tr Tx .
(30)
It turns out that this map takes values in the base field K, and that when restricted to the ring OL
of algebraic integers of L it maps integers to integers
TrL/K : OL OK .
(31)
The link to the above considerations of Jacobi sums is made by noting that for separable extensions L/K one can show that

TrL/K (x) =
(32)
(x)
: LK
151
where K denotes the closure of K. The cyclotomic fields of interest here are finite extensions
of the rational field Q, which is a perfect field, hence all its finite extensions are separable. This
means that the trace maps the ring of algebraic integers of any cyclotomic field into the rational
integers. It follows that the coefficients ap in the L-series are elements in Z. Cyclotomic fields
are normal, therefore the embeddings define the Galois group of automorphisms which leave Q
invariant.
The action defined above induces an action on the middle cohomology of any diagonal hypersurface X d of degree d embedded in a weighted projective space P(k0 ,...,ks ) with weights ki
n
because the set As parametrizes the (untwisted) cohomology classes in H s1 (X d ). The strategy
to compute the L-function of the orbits O determined by the factorization
Pps1 (t) =
Pps1 (O , t),
(33)
where
Pps1 (O , t) =
1 jpf ()t f
1/f
(34)
therefore implies the computation of the L-function of the irreducible representations of the
Galois group of the cyclotomic field associated to the variety on the intermediate cohomology.
From now on the superscript of the polynomial Pps1 (t), indicating the cohomology group, will
be dropped.
Of particular importance for CalabiYau spaces is the orbit O of the element :=
( kd0 , . . . , kds ) associated to the holomorphic (s 1)-form. This motivates the introduction of the
following two definitions.
n
Definition. The Galois orbit O of the element As with respect to the action of the Galois
group Gal(Q(d )/Q) is called the -representation of the Galois group Gal(Q(d )/Q).
It is possible to associate to an -representation defined by O a projection operator which
can be used to define a motive in the sense of Shioda.
Definition. The motive associated to the -representation defined by the orbit O will be called
the -motive M of the variety.
In this paper the focus will be on the computation of the L-function L (X, s) := L(M , s)
of the -motive of hypersurfaces X d

L X d , s =
p O
n
1
,
(1 (1)s1 jpf ()p f s )1/f
(35)
where O As denotes the orbit of the element that corresponds to the -form. More precisely, the aim is to check whether the -motives of extremal BrieskornPham K3 surfaces admit
a modular interpretation and if so, whether the resulting forms admit an affine Lie algebraic construction. Higher-dimensional varieties are considered in [23].
152
Table 1
The coefficients 1 (p) = 1 + p2 N1,p (S 4 ) of the HasseWeil modular form of the quartic Fermat surface S 4 in terms
of the cardinalities N1,p for small primes
Prime p
11
13
17
19
23
29
31
37
N1,p (S 4 )
1 (p)
16
6
0
26
64
14
144
22
128
42
600
310
400
38
576
46
768
74
1024
62
1152
218
3. Examples
The general framework formulated above is applied in this paper to extremal K3 surfaces
of BrieskornPham type. A K3 surface S defined over C is called extremal if its Picard number (S) = rk NS(S), defined as the rank of the NronSeveri group NS(S), is maximal, i.e.
(S) = 20. Such surfaces were originally called singular [24], and more recently have been
called attractive [25]. The set of extremal BrieskornPham K3 surfaces is given in Eq. (4).
3.1. The quartic Fermat K3 surface S 4
A summary of cardinality results Nr,p (S 4 ) for small primes p for the Fermat K3 surface of
degree four is contained in Table 1.
The results in Table 1 lead to the L-series of S 4

.
6
26 14 117
22
42
156 310
38
L S4, s = 1 + s s + s + s + s s s + s + s
3
5
7
9
11
13
15
17
19
84
46
+ s + s + ,
(36)
21
23
.
where the symbol = means that a finite number of Euler factors have been omitted, here the bad
prime p = 2. Associated to this L-series is the q-expansion

.
f S 4 , q = q + 6q 3 26q 5 + 14q 7 + 117q 9 + 22q 11 42q 13 156q 15 + 310q 17
+ 38q 19 + 84q 21 + 46q 23 + .
(37)
The product structure of L(S 4 , s) can be obtained from the Jacobi-sum formulation by enun
merating the set A3 of the surface S 4 . Replacing the degree vector n by the degree itself leads
to

1 1 1 1
1 1 1 1
3 3 3 3
1 1 1 3
1 1 3 3
4
, , ,
,
, , ,
,
, , ,
,
, , ,
,
, , ,
A3 =
4 4 4 4
2 2 2 2
4 4 4 4
4 2 2 4
4 4 4 4
+ permutations.
(38)
The Jacobi sums of the K3 surface S 4 at low primes are collected in Table 2. In this table the
permutations and the complex conjugates of the sums listed are suppressed.
The L-function of S 4 therefore factorizes as

L S 4 , s = LI S 4 , s LII S 4 , s LIII S 4 , s LIV S 4 , s ,
(39)
where the individual factors correspond to the orbits of the different Jacobi sums of Table 2.
The first factor describes the L-function of the Gal(Q(i)/Q)-orbit of jp ( 14 , 14 , 14 , 14 ), which
corresponds to the holomorphic 2-form of S 4 , and hence is the -motivic L-function of S 4 ,
153
Table 2
Jacobi-sums of the quartic K3 surface S 4 P3
Type
jq ( 14 ,
jq ( 12 ,
jq ( 14 ,
jq ( 34 ,
II
III
IV
1,
4
1,
2
1,
4
1,
4
1,
4
1,
2
3,
4
1,
2
1)
4
1)
2
3)
4
1)
2
13
17
29
37
3 + 4i
5 12i
15 8i
21 + 20i
35 12i
13
17
29
37
13
17
29
37
13
17
29
37
Table 3
Jacobi sums for the K3 surface S 6A P(1,1,1,3) for small primes
Prime p
jp ( 16 , 16 , 16 , 12 )
3
13
2 2
3i
13
12 + 15
2
19
3i
11 21
2
2
3i
31
23 + 12 3i
37
47 33
2
2
3i
L (S 4 , s) = LI (S 4 , s) of S 4 ,

1
L S 4 , s =
1 1 1 1
f
s
1/f
(1 jpf ( 4 , 4 , 4 , 4 ) p
) (1 jpf ( 34 , 34 , 34 , 34 ) p f s )1/f
p=2
9
10
30
11
42
70
6
.
= 1 s + s + s s + s + s s + .
5
9
13
17
25
29
37
(40)
The factor L (S 4 , s) of the complete L-function of S 4 is the only one with Jacobi sum characters
in an algebraic number field.
3.2. The degree six weighted K3 surface S 6A
The L-series of the -motive M of the double cover S 6A of the projective plane branched
over a degree six plane curve is determined by the Jacobi sums

1 1 1 1
5 5 5 1
jp , , ,
(41)
,
jp , , ,
,
6 6 6 2
6 6 6 2
which are complex conjugates. The values for these sums for low primes are collected in Table 3.
The resulting L-series of the -orbit is given by

.
13
1
11
46
47
L S 6A , s = 1 s s + s s + s +
7
13
19
31
37
up to a finite number of Euler factors.
(42)
3.3. The degree six weighted K3 surface S 6B

The -motivic L-series of the degree six BrieskornPham type surface embedded in P(1,1,2,2)
can be computed via the Jacobi sums

1 1 1 1
5 5 2 2
,
jp , , ,
jp , , ,
(43)
6 6 3 3
6 6 3 3
whose values at low primes are collected in Table 4.
154
Table 4
Jacobi-sums for the K3 surface S 6B P(1,1,2,2)
7
13
1 4 3
jp ( 16 , 16 , 13 , 13 )
11 4 3
19
13 8 3
31
23 12 3
37
13 + 20 3
The orbit of the -form with respect to the Galois group Q(6 )/Q leads to the L-series

.
2
22
26
46
26
22
L S 6B , s = 1 s s s + s + s + s + .
(44)
7
13
19
31
37
43
The question that arises now is whether these expansions can be given a string theoretic meaning, similar to the case of elliptic BrieskornPham curves.
4. CFT modularity of -motives
4.1. Modular forms
One of the important ingredients in the analysis of modular L-functions is their product structure, hence it is of interest to consider forms which admit Euler products, i.e. Hecke eigenforms.
Definition. A modular form of weight w, level N , and character withrespect
to 0 (N ) is a
map f : H C on the upper half-plane such that for any H and = ac db 0 (N )
f ( ) = (d)(c + d)w f ( ).
(45)
A cusp form can be characterized by the condition that the general q-expansion f ( ) =

n
n=0 an q starts with a1 . It is normalized if a1 = 1.
Hecke defined on the space of cusp forms Sw (0 (N ), ) operators which for prime p take the
form

Tpw f (q) =
(46)
anp q n + (p)p w1
an q np .
n=1
n=1
Eigenforms of these operators are particularly important in a geometric context.

Theorem 4 (Hecke). The space Sw (0 (N ), ) of cusp forms of weight w, level N , and character
is the orthogonal sum of the spaces of equivalent eigenforms. Each space of such forms has a
member that is an eigenvector of all Hecke operators Tw (n). A form f Sw (0 (N ), ) that is
an eigenvector for all Tw (n) can be normalized and its coefficients satisfy
apn+1 = apn ap (p)p w1 apn1 ,
a
pn
= (ap ) ,
n
amn = am an ,
p/|N,
p|N,
(m, n) = 1.
(47)
Moreover, the L-function L(f, s) has an Euler product of the form

1
1
L(f, s) =
s
s
1 ap p
1 ap p + p w12s
p prime
p|N
which is convergent for Re s >
p prime
p /|N
w
2
+ 1.
(48)
155
4.2. CFT ingredients

Quantities that have proven useful in the string modularity analysis of elliptic BrieskornPham
curves [57] are the theta functions

2
2
k
,m
(49)
( ) =
sign(x)e2i ((k+2)x ky ) ,
|x|<y|x|
(x,y) or ( 12 x, 12 +y)
+1
m
, 2k
)
Z2 +( 2(k+2)
(1)
defined by Kac and Peterson. These are related to the string functions of the affine algebra A1
at level k as
k
( ) =
c,m
k ( )
,m
3 ( )
(50)
The string functions, together with the classical theta functions

2
n,m (, z, u) = e2imu
e2im +2iz ,
(51)
n
Z+ 2m
k
( ) of the partition funcare the building blocks of the N = 2 supersymmetric characters ,q,s
tions of the N = 2 minimal theories
c
k
(, z, u) = e2iu trHk q (L0 24 ) e2izJ0
,q,s
,q,s

c

k
k
= e2iu
mult k,q,s , Qk,q,s e2i (,q,s 24 )+2izQ,q,s ,
(52)
Qk,q,s ,k,q,s
in terms of the conformal dimensions k,q,s and the charges Qk,q,s , leading to [26]

k
k
,q,s
(, z, u) =
c,q+4j
s ( )2q+(4j s)(k+2),2k(k+2) (, z, u).
(53)
It will become clear below that the modular forms that explain the modularity of extremal K3
surfaces of BrieskornPham type in a string theoretic way are given by

1
1,1
(q) = q 1/12 1 2q q 2 + 2q 3 + q 4 + 2q 5 2q 6 2q 8 2q 9 + q 10 + ,

2
(q) = q 1/8 1 q 2q 2 + q 3 + 2q 5 + q 6 2q 9 + q 10 + .
1,1
(54)
4.3. Modularity of -motives
If an -motive of K3 type is modular the associated modular form is expected to be of weight
three. A useful guide in the search for such forms is provided by the speculation that exactly
solvable models lead to motives which admit complex multiplication (CM) in the classical sense
[27] (see [28]for an alternate discussion of CM). The quartic Fermat surface admits CM
with
respect to Q( 1 ), while the degree six surfaces have CM with respect to the field Q( 3 ).
The goal therefore is to find modular forms of weight three which admit complex multiplication
with respect to these fields.
156
A further guide is provided by the bad primes of these surfaces. It is expected that the level
of a geometrically derived modular form is divisible by the bad primes of the underlying variety.
The only bad prime for the quartic Fermat K3 surface is p = 2, hence the level should be some
power of two. For the degree six surfaces the bad primes are p = 2, 3, hence the level should be
of the form 2a 3b for some non-negative integers a, b.
Combining these considerations leads to the candidate cusp forms

S3 0 (16), 1 6 (4 ) = q 6q 5 + 9q 9 + 10q 13 30q 17 + 11q 25 + ,

S3 0 (12), 3 3 (2 )3 (6 ) = q 3q 3 + 2q 7 + 9q 9 22q 13 + 26q 19 + , (55)
for the K3 surfaces considered above. Comparing the q-expansions of these two forms with
those of the -motivic L-series of the K3 surfaces shows that while the first form describes the
L-function of the quartic K3 surface, the latter describes the corresponding L-function of S 6B
only up to sign changes. This indicates that a twist is involved, and it turns out that this twist
can be provided by the Legendre character 3 . This leaves the surface S 6A . A candidate modular
form can be obtained by lifting a complex multiplication modular
form of weight two to weight
three via an Eisenstein series associated to the CM field Q( 3 ). In the present case a useful
modular form turns out to come from a class of theta series considered in [29]. Hecke
associates
to each element of an integral ideal a in an imaginary quadratic field K = Q( D ) the theta

series defined as

; , a, Q D =
(56)
q Nz/Q|D|Na ,
z(mod aQ D )
where Nz and Na denote the norms of z OK and a respectively, and OK is the ring of integers
of K. Relevant for the L-series L (S 6A
, s) is the special case given by = 0, Q = 1, and
a = OK for the Eisenstein field K = Q( 3 ), renamed here as

(q) =
(57)
q Nz .
zOK
In summary, the results above lead to the identification of the respective inverse Mellin transforms f (S d , q) of the motivic L-functions L (S d , s) for d = 4, 6A, 6B as noted in Eq. (6) of
Theorem 1. The proof of these relations to all orders is postponed to Appendix A.
The modularity of the -motives of extremal BrieskornPham K3s leaves the question
whether the resulting modular forms admit a string theoretic interpretation in terms of forms
derived from the associated conformal field theory. The basic idea in the following is to consider
the arithmetic building blocks of the K3 surfaces considered here. One way to identify these
structures is via the twist map. In this section the focus will remain on the arithmetic aspects,
postponing the identification of the basic irreducible geometric structure to Section 5.
4.4. Elliptic curves from K3 surfaces
The structure of the quartic K3 surface S 4 is very simple. It is shown below that it can be
constructed directly in terms of E 4 via the twist map [19]. Alternatively, one can use the Shioda
Katsura decomposition [30,31] to reconstruct the cohomology of S 4 from the cohomology of the
quartic plane curve

C4 = (z0 : z1 : z2 ) P2 z04 + z14 + z24 = 0 .
(58)
157
This is a genus three curve whose L-function factors into the triple product of the L-function of
the weighted elliptic curve E 4 . The L-function of this curve was computed in [6], where it was
shown that its inverse Mellin transform is determined by a twist of the Hecke indefinite modular
2 (q), as described in Theorem 2. Therefore the geometry and arithmetic structure of the
form 1,1
2 (q) of weight
quartic Fermat K3 suggests a relation between the modular forms 6 (q) and 1,1
three and two, respectively.
The structure of the degree six K3 surface S 6B can be recovered from the twist map in a
way similar to the quartic surface by constructing it from two copies of the elliptic curve E 6 , as
described below. Alternatively, for both degree six surfaces one can again consider the reduction
of the cohomology via ShiodaKatsura. There are two curves to consider, the weighted plane
curve

C6 = (z0 : z1 : z2 ) P(1,1,2) z06 + z16 + z23 = 0
(59)
and the elliptic curve E 3 . The latter was analyzed in detail in [5], with the result that its associated
1 (q) at conformal level k = 1. The Jacobian
modular form is determined by the theta series 1,1
of C6 factors into three different types of elliptic curves, E 3 just discussed, the degree six elliptic
curve E 6 , and a third curve of conductor 432. The curve E 6 has been shown to lead to the string
1 (q). These results then suggest a relation between
theoretic modular form given in terms of 1,1
k (q) for levels k = 1 or k = 2, or both.
the modular form 3 (2 )3 (6 ) and 1,1
4.5. From K3 to CFT forms
For the degree six surface S 6A the relation between the weight three form and the string
theoretic form is immediate because this form is the lift of a product of Hecke indefinite modular
forms. In general, the expected relation between forms of different weight cannot be established
for the forms themselves, but should proceed in terms of their associated L-series. Guidance for
such constructions is provided by the theory of convolutions of L-functions. It turns out that for
all three modular forms of weight three determined above the relation is of similar type. Denote
by
fw (q) =
an(w) q n
(60)
n=1
the q-expansion of the weight w form. Then the relation between the pairs of weight three and
weight two forms for the surfaces

2 4 2
S4:
q
(f3 , f2 ) = 6 q 4 , 1,1
2 ,
1 3 1 9
q 1,1 q
S 6A : (f3 , f2 ) = q 3 2 q 3 2 q 9 , 1,1
(61)
is of the form
2

ap(3) = ap(2) 2p
(62)
for rational primes p, while for third surface

1 6 2
S 6B : (f3 , f2 ) = 3 q 2 3 q 6 , 1,1
q
3 ,
(63)
a twist is necessary

2
ap(3) = ap(2) 2p 3 (p).
(64)
158
The origin of these relations, and the motivation for the various twists, will become clear in the
next two sections.
k ( ) introduced above suffice
This result shows that the string theoretic modular forms ,m
to explain the arithmetic structure of spacetime, represented here by the -motive carrying an
irreducible representation of the Galois group of the cyclotomic field, determined by the symmetry group of the K3 surfaces. Reading these relations in reverse explains string theoretic modular
forms in terms of the arithmetic geometry of the topologically non-trivial part of spacetime.
5. K3 surfaces from elliptic curves
This section describes the elliptic building blocks used in the previous section to identify
the string theoretic structure of extremal BrieskornPham K3s. A direct construction of the two
surfaces S 4 and S 6B in terms of elliptic curves can be obtained via the twist map [19]. This
construction was originally considered in the context of string dualities [32] (see also the later
Ref. [33]), and will also be useful below for the proof of modularity to all orders for the motivic L-series of these K3 surfaces.
5.1. The twist map construction
In the notation of [19] consider the map
: P(w0 ,...,wm ) P(v0 ,...,vn ) P(v0 w1 ,...,v0 wm ,w0 v1 ,...,w0 vn )
(65)
defined as

w /w

w /w
v /v
v /v
(x0 , . . . , xm ), (y0 , . . . , yn ) y0 1 0 x1 , . . . , y0 m 0 xm , x0 1 0 y1 , . . . , x0 n 0 yn
=: (z1 , . . . , zm , t1 , . . . , tn ).
(66)
This map restricts on the subvarieties

X1 = x0 + p(xi ) = 0 P(w0 ,...,wm ) ,

X2 = y0 + q(yj ) = 0 P(v0 ,...,vn ) ,
(67)
defined by transverse polynomials p(xi ) and q(yj ), to the hypersurface

X = p(zi ) q(tj ) = 0 P(v0 w1 ,...,v0 wm ,w0 v1 ,...,w0 vn )
(68)
as a finite map. The degrees of the hypersurfaces Xi are given by

deg X1 = w0 ,
deg X2 = v0 ,
(69)
leading to the degree deg X = v0 w0 .

5.2. Examples
Applying the twist construction to the quartic elliptic curve E 4 P(1,1,2) leads first to the
map of ambient spaces
: P(2,1,1) P(2,1,1) P3
(70)
159
defined as

1/2
1/2
1/2
1/2
(x0 , x1 , x2 ), (y0 , y1 , y2 ) y0 x1 , y0 x2 , x0 y1 , x0 y2 .
(71)
4 E4
Restricting the map to E+

4
E
= x02 x14 + x24 = 0 P(2,1,1) ,
(72)
relates this product of elliptic curves to the quartic K3 surface S 4 P3 .

The degree six surface S 6B can be constructed similarly by considering the elliptic curve
6
E P(1,2,3) . The ambient space map
: P(3,2,1) P(3,2,1) P(1,1,2,2)
(73)
is now defined by

2/3
1/3
2/3
1/3
(x0 , x1 , x2 ), (y0 , y1 , y2 ) y0 x1 , y0 x2 , x0 y1 , x0 y2 .
(74)
6 E 6 , where
Its restriction to E+

6
= x02 x13 + x26 = 0 P(3,2,1) ,
E
(75)
relates these curves to the surface S 6B P(1,1,2,2) .

It will become clear below that the surface S 6A is determined by the elliptic curve E 3 . Therefore all surfaces considered here can be understood in terms of the arithmetic structure of the
three BrieskornPham type elliptic curves described in [7].
5.3. K3 geometry from string theory
One of the fundamental problems in string theory is to derive the structure of spacetime solely
from the field theory on the worldsheet. In Refs. [5,7] it was shown that it is possible to construct
the elliptic geometry of Gepner models at c = 3 directly from the conformal field theory, providing a string theoretic derivation of the extra dimensional geometry for a simple class of exact
models. The question arises whether such a derivation can be generalized to the K3 surfaces analyzed here. In contrast to elliptic curves, the cohomology of K3 surfaces is more complicated, and
it is not obvious how a direct construction should proceed. The twist map described above does,
however, lead directly from the modular forms on the worldsheet to the geometry of spacetime
in this higher dimensional case as well. This can be seen as follows.
The first step is the construction, described above, of these K3 surfaces in terms of the elliptic curves of BrieskornPham type via the twist map. This reduces the modular construction
from two dimensions to one. The second step is to use the criteria formulated in [7], which
uniquely identify the modular forms for these elliptic curves. This leads to the modular forms
of weight 2 described in the previous section. This two-step procedure can be reversed by using
the EichlerShimura construction, which allows to construct elliptic curves from their modular
forms. Combining these modularity arguments with the twist map therefore leads to the construction of the K3 surfaces S 4 and S 6B directly from the conformal field theories on the worldsheet.
An alternative reduction strategy proceeds via complex multiplication. This approach is less direct, but allows to discuss all three surfaces in a unified manner, as shown below.
160
6. Modularity of a phase transition

The arithmetic structure of a projective variety is sensitive to deformations, changing as the
coefficients of its defining polynomials are varied. It is therefore of interest to consider the behavior of families of varieties and analyze their arithmetic structure as their moduli change. This line
of thought has already been followed in recent work [3436]. These interesting papers address,
among other issues, the question of what happens to the reduced variety at the conifold locus.
Ref. [36] in particular considers the lower dimensional analog of the conifold transition in the
context of the following family of quartic K3 surfaces

3

4
4
S () = (z0 : :z3 ) P3
(76)
zi 4
zi = 0 .
i=0
The result is that at the singular locus = 1 the congruence zeta functions Z(S 4 ()/Fp , t)
degenerate, as expected. It turns out that they become singular in a way that preserves modularity. The question therefore arises whether there is a cohomological L-function whose Mellin
transform admits a conformal field theoretic interpretation.
The zeta function computations of Ref. [36] indicate that at the singular point a modular form
emerges which can be given a string theoretic interpretation as
2 ( ) 2 (4 ))2
(1,1
1,1
2 (2 )
1,1
2 .
(77)
What is interesting about this form is that it is written in terms of cusp forms determined by
precisely the same conformal field theory that leads to the L-function at the Fermat locus in
the moduli space. Furthermore, the twist character 2 which
appears in this expression, is the
Legendre character of the field of quantum dimensions Q( 2 ) of that same theory.
This suggests that at the analog of the conifold point the arithmetic structure points to a conformal field theory at level k = 2, precisely the same structure that describes the situation for the
Fermat surface.
7. Symmetries
7.1. Complex multiplication
It was conjectured in [27] that exactly solvable CalabiYau varieties can be characterized by
a complex multiplication symmetry in the sense of [37]. In the context of elliptic curves this coincides with the classical notion of CM, and it was shown in [7] that all elliptic BrieskornPham
curves and their associated

modular forms admit CM by either the Eisenstein field Q( 3 ) or
the Gauss field Q( 1 ). For higher dimensional CalabiYau varieties the concept of complex
multiplication must be modified. The idea of Ref. [37] is to define the CM property of a general
CalabiYau variety in terms of the complex multiplication properties of the associated Jacobians
of the curves embedded in the variety. This notion can be formulated more generally in terms of
the CM properties of motives defined by the Galois representations, as described above.
A natural problem therefore is to determine the CM nature of the forms derived above from
the -motives of extremal
BrieskornPham K3 surfaces. In terms of the coefficients of the q
expansion, f (q) = n an q n , CM implies the existence of quadratic extension fields K/Q such
that the coefficients ap vanish at all primes p that are inert in K. In [7] this condition was
161
Table 5
CM modular forms with geometric interpretation
Form
Weight
Level
1 (q 3 ) 1 (q 9 )
1,1
1,1
1 (q 6 ))2
(1,1
2 (q 4 ))2
(1,1
2
6
(q 4 )
27
36
64
16
(q 3 )2 (q 3 )2 (q 9 )
27
3 (q 2 )3 (q 6 ) 3
48
2 (q) 2 (q 4 ))2
(1,1
1,1
2 (q 2 )
1,1
CM Field
Q( 3 )
Q( 3 )
Q( 1 )
Q( 1 )
Q( 3 )
Q( 3 )
Q( 2 )
Geometry
Elliptic curve
Elliptic curve
Elliptic curve
K3 -motive
K3 -motive
K3 -motive
Singular quartic K3
checked for the modular forms associated to elliptic BrieskornPham curves. Similarly one can
show that the forms of weight 3 of Theorem 1 admit CM. Table 5 summarizes the results for the
CM fields of all the forms encountered so far. The results indicate that the relevance of complex
multiplication generalizes from exactly solvable elliptic curves to higher dimensions.
The complex multiplication property of the K3 motivic modular forms provides an alternative
approach to the notion of dimensional reduction of K3 surfaces to elliptic curves discussed in
Sections 4 and 5. CM modular forms can be obtained from L-series
(a)

1
L(, s) =
(78)
=
(p)
Nas
1 s
pSpec OK
Np
aI (OK )
associated to Hecke characters of number fields K, where I(OK ) (Spec OK ) describes the set
of (prime) ideals in the ring of algebraic integers
OK , and Na is the norm of the ideal a [38]. In the
of
Q(
1
),
and two characters 6A and 6B associated to
present
case
a
Hecke
characters
Q( 3 ), all of weight two, determine the L-series of S 4 , S 6A and S 6B by considering the square
of the basic characters. More precisely, the arithmetic of the -motive of elliptic Brieskorn
Pham K3s can be obtained as follows. Define the character 4 at the prime ideals p of Q( 1 )
dividing the rational prime p as

2
mod (2 + 2i) ,
4 (p) = p , with p
(79)
p
where p = (p ). Define further the characters 6A and 6B on the prime ideals p|p of Q( 3 )
dividing the rational primes p as
6A (p) = p ,
with p 1 (mod 3),

3
mod (2 + 43 ) .
6B (p) = p , with p
(80)
p
These characters provide the Hecke theoretic interpretation of the inverse Mellin transforms of
the HasseWeil L-functions of the elliptic curves E 4 , E 3 and E 6 respectively. Renaming E 3 and
E 6 as E 6A := E 3 and E 6B := E 6 , one can check that

LHW E d , s = L(d , s), d = 4, 6A, 6B.
(81)
The motivic L-series of S 4 , S 6A and S 6B can then be described as

L S d , s = L d2 , s , d = 4, 6A
(82)
162
Table 6
Conway classes for the forms of Table 5
Form
1 (q 3 ) 1 (q 9 )
1,1
1,1
1 (q 6 )2
1,1
1 (q 6 )2
1,1
6 (q 4 )
3 (q 2 )3 (q 6 ) 3
Conway class
(3D, 3B)
6I
8F |T
4F
6G|T
2 (q) 2 (q 4 ))2
(1,1
1,1
2 (q 2 )
1,1
8E
for the first two surfaces, and as

2

3 , s
L S 6B , s = L 6B
(83)
for the third example. These results provide an alternative demonstration that the elliptic curves
E 4 , E 3 and E 6 can be viewed as the modular building blocks of the K3 surfaces considered here.
They also explain the relations (62) and (64) by rewriting the coefficients of the Hecke L-series
of the weight three forms in terms of the real coefficients of the weight two forms.
The Hecke interpretation of the -motivic L-series of elliptic curves and K3-surfaces provides a way to explicitly prove the modularity of these series to all order of their q-expansions
by using some results from Hecke and Shimura. Alternatively, it is possible to point to a general modularity result for extremal K3 surfaces defined over number fields [24,39]. It is useful,
however, to see modularity explicitly, and a proof is contained in Appendix A.
7.2. K3 arithmetic moonshine
It was pointed out in [7] that the modular forms given by the inverse Mellin transform of
the HasseWeil L-functions of elliptic BrieskornPham curves can be interpreted as generalized
McKayThompson series associated to the largest Mathieu group M24 , leading to the notion of
what might be called arithmetic moonshine. It is therefore natural, as an aside, to ask whether the
modular forms that appear in the context of extremal BrieskornPham K3 surfaces can also be
interpreted as such series. This is the case if the group considered is enlarged from the Mathieu
group to the Conway group, defined as the automorphism group of the Leech lattice.
The specific group elements associated to the forms encountered in the present paper are listed
in Table 6, where the notation of Ref. [40] is adopted, which in turn follows the atlas [41]. The
numerical part of the group element denotes its order while the letter part separates different
elements of the same order.
It is intriguing to see that all geometrically induced conformal field theoretic modular forms
obtained so far admit a sporadic group theoretic interpretation. It is an open question whether
this relation can help to provide a conceptual foundation of the relations established so far.
Acknowledgements
It is a pleasure to thank Monika Lynker for discussions, and John Stroyls for generously making available his library. This work was supported in part by an Incentive Grant for Scholarship
at KSU, and while the author was a Scholar at the Kavli Institute for Theoretical Physics in Santa
Barbara. It is a pleasure to thank the KITP for hospitality. This work was supported in part by
the National Science Foundation under Grant No. PHY99-07949.
163
Appendix A. Modularity proof

A.1. FaltingsSerreLivn strategy
Modularity of the L-series considered here can be shown in several ways, e.g. by using their
complex multiplication origin and results of Hecke and Shimura. It is also possible to refer to
some general results of Livn. CalabiYau modularity is not restricted to manifolds with CM,
as illustrated by many examples reviewed in [10]. For completeness, this appendix contains a
brief modularity proof of the -motives of extremal K3 surfaces of BrieskornPham type by
using techniques essentially developed by Faltings and Serre. It was first observed by Faltings
[42] that different Galois representations can be shown to be identical if they agree at a finite
number of primes. This observation has been developed by Serre [43] and Livn [44], and Livn
in particular made it into a practical tool by specifying precisely the set of primes that has to
be tested in order to guarantee agreement of two representations. The varieties discussed here
are defined over Q, hence Livns more general theorem can be reduced to the following form,
previously considered by Verrill [45] (see [10] for more references in this direction). Let Q be
the -adic field and denote by Fp the Frobenius element at the prime p.
Theorem 5. Let S be a finite set of rational primes, and denote by QS the compositum of all
quadratic extensions of Q unramified outside of S. Suppose
GL(2, Q2 )
1 , 2 : Gal(Q/Q)
(A.1)
are 2-adic continuous representations, unramified outside of S, which further satisfy the following conditions:
(1) tr 1 tr 2 0 mod 2.
(2) There exists a finite set T = of rational primes, disjoint from S, for which
(a) the image of the set {Fp }pT in Gal(QS /Q) is surjective,
(b) tr 1 (Fp ) = tr 2 (Fp ) for all p T ,
(c) det 1 (Fp ) = det 2 (Fp ) for all p T .
Then 1 and 2 have isomorphic semi-simplification.
The idea therefore is to translate the geometric and modular computations into two Galois
representations, and to show agreement by satisfying the requirements of Livns result. The
geometric part is given by the representations of the absolute Galois group Gal(Q/Q)
on the
-adic cohomology group

Q ) ,
1 = i : Gal(Q/Q)
(A.2)
Aut Hit (X,
where is a prime different from the prime of reduction, X/Q is assumed to be smooth projective
and Hi indicates the ith tale cohomology group.
variety, X = X Q Q,
t
In the absolute Galois group there exists a distinguished element, the geometric Frobenius
Given the representation i these Frobenius elements act on

endomorphism Fp Gal(Q/Q).
the tale cohomology, and it turns out that the polynomials Ppi (t) that define the geometric Lfunction associated to these cohomology groups can be determined in terms of these Frobenii
as

Ppi (t) = det 1 i (Fp )|Hi (X,Q
(A.3)
)t .
t
164
The second representation is the 2-dimensional Galois representation 2 = f associated by

Deligne [46] to Hecke eigenforms of arbitrary weight, where
GL(2, Q )
f : Gal(Q/Q)
(A.4)
is a representation that is unramified outside of and the prime divisors of the level N of the
modular form. This -adic representation can be realized in the cohomology of certain -adic
sheaves over a modular curve. These can be defined over Q and therefore Gal(Q/Q)
acts on the
-adic cohomology. Deligne has shown that these representations have the properties
tr f (Fp ) = ap ,
det f (Fp ) = p w1 (p)
(A.5)
for all primes p different from . In the present discussion of K3 surfaces the focus is on modular
forms of weight three and the goal is to use Theorem 5 to prove the identity of the geometric
representation and the modular representation.
A.2. Examples
The application of the strategy outlined above to the K3 surface S 4 P3 starts with the determination of the set T of primes considered in Livns theorem. This is obtained by considering
a representation of the Galois group of the composite field QS . For S 4 the bad prime is p = 2,
which is also the only divisor of the level of modular
form 6 (4 ) S3 (0 (16), 1 ). Hence
S = {2}, and the composite field is Q{2} = Q(i, 2 ) = Q(8 ). Therefore the set T of primes can
be chosen as
T = {3, 5, 7, 17}.
(A.6)
This set of primes is a subset of the primes for which agreement of the modular and the geometric
L-series is shown by the computations in the previous subsections. It remains to establish the
conditions formulated in the theorem.
The fact that tr 1 = 0 mod 2 follows by noting that the quartic surface can be constructed via
the twist map, as explained above. The elliptic curve E 4 involved in the construction of S 4 admits
complex multiplication, and it can be shown that its coefficients ap satisfy ap = 0 mod 2 (see
e.g. [47], or [7]). Hence the same follows for 1 . Alternatively, this follows from the Jacobi-sum
formulation of the L-function of the -motive. The same congruence holds for 6 (4 ), which
can be seen e.g. via Jacobis expansion for 3 ( ) [48]. The determinant condition, finally, follows
from Weils result for Jacobi sums in the geometric case, and Delignes result in the case of the
modular form.
The case of the degree six surface S 6B P(1,1,2,2) is similar to the discussion of the quartic.
The set of bad primes is S = {2, 3}, which gives the set of divisors of the level of the modular
form. This leads to the composite field Q{2,3} = Q(24 ). The set of primes representing the Galois
group of this field is given by
T = {5, 7, 11, 13, 17, 19, 23, 73}.
(A.7)
This set of primes is again a subset of the primes considered above for the surface S 6 . Comparison
between the geometric and modular L-series shows agreement on the set of primes given by T .
The fact that tr 1 = 0 mod 2 can again be seen via the twist construction. The curve E 6 , the
building block of S 6 , has complex multiplication, and the coefficients of its HasseWeil L-series
165
are zero mod 2. Hence it follows that this holds also for 1 . The remaining assumptions of the
theorem follow in the same way as in the case of the quartic surface in combination with results
proven in [45].
The surface S 6A does not satisfy the conditions of Livns theorem since the coefficients can
be odd. The following result, proven in [49], can be used to complete the proof.
Proposition. Let 1 , 2 be two 2-adic Galois representations with the same determinant and
even trace at Fr11 or Fr13 , which are unramified outside {2, 3}. Then they have isomorphic semisimplifications if and only if for any p {5, 7, 11, 13, 17, 19, 23, 31, 37} the traces of the Frobenii
are identical, tr 1 (Frp ) = tr 2 (Frp ).
A discussion of the notion of semi-simplification can be found in [50]. Important here is the
implication that the two L-functions associated to the Galois representations are identical. It follows from the Jacobi sum representation of the L-series and Delignes result that the determinants
of the geometric representation and the modular representations are identical. The computations
collected in Table 1 furthermore show agreement for the traces at the primes determined by the
proposition. Hence the proof follows.
References
[1] A. Wiles, Modular elliptic curves and Fermats last theorem, Ann. Math. 141 (1995) 443551;
R. Taylor, A. Wiles, Ring-theoretic properties of certain Hecke algebras, Ann. Math. 141 (1995) 553572.
[2] F. Diamond, On deformation rings and Hecke rings, Ann. Math. 144 (1996) 137166.
[3] B. Conrad, F. Diamond, R. Taylor, Modularity of certain potentially BarsottiTate galois representations, J. Am.
Math. Soc. 12 (1999) 521567.
[4] C. Breuil, B. Conrad, F. Diamond, R. Taylor, On the modularity of elliptic curves over Q or wild 3-adic exercises,
J. Am. Math. Soc. 14 (2001) 843939.
[5] R. Schimmrigk, S. Underwood, The ShimuraTaniyama conjecture and conformal field theories, J. Geom. Phys. 48
(2003) 169189, hep-th/0211284.
[6] M. Lynker, R. Schimmrigk, Geometric KacMoody modularity, J. Geom. Phys. 56 (2006) 843863, hepth/0410189.
[7] R. Schimmrigk, Arithmetic spacetime geometry from string theory, hep-th/0510091.
[8] R. Langlands, L-Functions and Automorphic Representations, ICM, 1978.
[9] A. Weil, ber die Bestimmung Dirichletscher Reihen durch Funktionalgleichungen, Math. Ann. 168 (1967) 149
156.
[10] N. Yui, Arithmetic of certain CalabiYau varieties and mirror symmetry, in: Arithmetic Algebraic Geometry, American Mathematical Society, 2001, pp. 507569;
N. Yui, Update on the modularity of CalabiYau varieties, with an appendix by H. Verrill, Fields Inst. Commun. 38
(2003) 307362.
[11] A. Grothendieck, Formul de Lefshetz t Rationalit de Fonction de L, Sminaire Bourbaki 279 (1964/1965) 115.
[12] A. Weil, Number of solutions of equations in finite fields, Bull. Am. Math. Soc. 55 (1949) 497508.
[13] R. Schimmrigk, CalabiYau arithmetic and rational conformal field theories, J. Geom. Phys. 44 (2003) 555569,
hep-th/0111226.
[14] P. Candelas, M. Lynker, R. Schimmrigk, CalabiYau manifolds in weighted projective P4 , Nucl. Phys. B 341 (1990)
383402.
[15] B.R. Greene, R. Plesser, Duality in CalabiYau moduli space, Nucl. Phys. B 338 (1990) 1537.
[16] V.V. Batyrev, Dual polyhedra and mirror symmetry for CalabiYau hypersurfaces in toric varieties, J. Alg. Geom. 3
(1994) 493535.
[17] M. Kontsevich, Homological algebra of mirror symmetry, in: Proceedings of the International Congress of Mathematicians 1994, Birkhuser, 1995, pp. 120139.
[18] A. Strominger, S.-T. Yau, E. Zaslow, Mirror symmetry is T-duality, Nucl. Phys. B 479 (1996) 243259.
[19] B. Hunt, R. Schimmrigk, K3 fibered CalabiYau threefolds: The twist map I, Int. J. Math. 10 (1999) 333366,
math.AG/9904059.
166
[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]
A. Weil, Jacobi Sums as Grssencharaktere, Trans. Am. Math. Soc. 73 (1952) 487495.
J. Neukirch, Algebraic Number Theory, Springer, 1999.
W. Narkiewicz, Elementary and Analytic Theory of Algebraic Numbers, Springer, 2004.
R. Schimmrigk, in preparation.
T. Shioda, H. Inose, On singular K3 surfaces, in: Complex Analysis and Algebraic Geometry, 1977, pp. 119136.
G. Moore, Arithmetic and attractors, hep-th/9807087.
D. Gepner, Spacetime supersymmetry in compactified string theory and superconformal models, Nucl. Phys. B 296
(1988) 757778.
M. Lynker, R. Schimmrigk, S. Stewart, Complex multiplication of exactly solvable CalabiYau varieties, Nucl.
Phys. B 700 (2004) 463489, hep-th/0312319.
N. Yui, The L-series of CalabiYau orbifolds of CM type, in preparation.
E. Hecke, ber einen neuen Zusammenhang zwischen elliptischen Modulfunktionen und indefiniten quadratischen
Formen, in: Nachrichten der K. Gesellschaft der Wissenschaften zu Gttingen, Mathematisch-physikalische Klasse,
1925, pp. 3544.
T. Shioda, T. Katsura, On Fermat varieties, Tohoku Math. J. 31 (1979) 97115.
P. Deligne, Hodge cycles on Abelian varieties, in: P. Deligne, J.S. Milne, A. Ogus, K.-y. Shen (Eds.), Hodge Cycles,
Motives, and Shimura Varieties, in: Lecture Notes in Mathematics, vol. 900, Springer, 1982.
B. Hunt, R. Schimmrigk, Heterotic gauge structure of K3-fibered CalabiYau manifolds, Phys. Lett. B 381 (1996)
427436, hep-th/9512138.
C. Borcea, K3 surfaces with involution and mirror pairs of CalabiYau manifolds, in: Mirror Symmetry, II, American Mathematical Society, 1997.
P. Candelas, X. de la Ossa, F. Rodriguez-Villegas, CalabiYau manifolds over finite fields I, hep-th/0012233;
P. Candelas, X. de la Ossa, F. Rodriguez-Villegas, CalabiYau manifolds over finite fields II, Fields Inst. Commun. 38 (2003) 121157, hep-th/0402133.
F. Rodriguez-Villegas, Hypergeometric families of CalabiYau manifolds, Fields Inst. Commun. 38 (2003) 223
231.
S. Kadir, The arithmetic of CalabiYau manifolds and mirror symmetry, hep-th/04109202.
M. Lynker, V. Periwal, R. Schimmrigk, Complex multiplication of black hole attractor varieties, Nucl. Phys. B 667
(2003) 484504, hep-th/0303111.
K.A. Ribet, Galois representations attached to eigenforms with nebentypus, in: J.P. Serre, D. Zagier (Eds.), Modular
Functions of One Variable, V, in: Lecture Notes in Mathematics, vol. 601, Springer, 1977.
R. Livn, Motivic orthogonal two-dimensional representations of Gal(Q/Q),

Israel J. Math. 92 (1995) 149156.
Y. Martin, Multiplicative eta quotients, Trans. Am. Math. Soc. 348 (1996) 48254856.
J.H. Conway, R.T. Curtis, S.P. Norton, R.A. Parker, R.A. Wilson, Atlas of Finite Groups, Clarendon, 1985.
G. Faltings, Endlichkeitsstze fr abelsche Varietten ber Zahlkrpern, Invent. Math. 73 (1983) 349366.
J.-P. Serre, Resum des cours de 19841985, Annu. Collge de France (1985) 8590.
R. Livn, Cubic exponential sums and Galois representations, in: K. Ribet (Ed.), Current Trends in Arithmetical
Algebraic Geometry, in: Contemp. Math., vol. 67, 1987, pp. 247261.
H. Verrill, The L-series of certain rigid CalabiYau threefolds, J. Number Theor. 81 (2000) 310334.
P. Deligne, Formes modulaires t reprsentations -adiques, Sem. Bourbaki 355 (1968/1969), Lecture Notes in
Mathematics, vol. 179, Springer, 1971.
D. Zagier, Aspects of Complex Multiplication, Lecture Notes, Berkeley, 2000.
C.G.J. Jacobi, Fundamenta nova theoriae functionum ellipticarum, 1829, in: Mathematische Werke I, Chelsea, 1969,
pp. 49239.
M. Schtt, On the modularity of three CalabiYau threefolds with bad reduction at 11, math.AG/0405450.
R. Taylor, Galois representations, Ann. Fac. Sci. Toulouse Math. 13 (2004) 73119.
Topology change from quantum instability of gauge

theory on fuzzy CP2
Djamel Dou a,b , Badis Ydri c,
a Department of Physics and Astronomy, King Saud University P.O. Box 2455, Riyadh 11451, Saudi Arabia
b Institute of Exact Science and Technology, University Center of Eloued, Eloued, Algeria
c Institut fr Physik, Humboldt-Universitt zu Berlin, Newtonstreet 15, D-12489 Berlin, Germany
Received 29 January 2007; accepted 6 February 2007

Abstract
Many gauge theory models on fuzzy complex projective spaces will contain a strong instability in the
quantum field theory leading to topology change. This can be thought of as due to the interaction between
spacetime via its noncommutativity and the fields (matrices) and it is related to the perturbative UVIR
mixing. We work out in detail the example of fuzzy CP2 and discuss at the level of the phase diagram the
quantum transitions between the 3 spaces (spacetimes) CP2 , S2 and the 0-dimensional space consisting of
a single point {0}.
0. Introduction
The approach of fuzzy physics [1,2] (1) to quantum geometry and (2) to non-perturbative
field theory insists on the use of finite dimensional matrix algebras with suitable Laplacians
(metrics) to describe the geometry [3]. Fields will be described by the same matrix algebras or
more precisely by the corresponding projective modules [4] and the action functionals will be
given by finite dimensional matrix models similar to the IKKT models [5,6].
As it turns out a topology change can occur naturally if we try to unify spacetime and fields
using this language of finite dimensional matrices. This is precisely the picture which emerges
from the perturbative and non-perturbative studies of non-commutative gauge theory on the fuzzy
sphere [7,8]. Indeed we found in the one-loop calculation [9] as well as in numerical simulations
E-mail addresses: ddou@ictp.it (D. Dou), ydri@physik.hu-berlin.de (B. Ydri).

doi:10.1016/j.nuclphysb.2007.02.010
168
D. Dou, B. Ydri / Nuclear Physics B 771 [PM] (2007) 167189
[10,11] and the large N analysis [12] that the non-commutative gauge model on the fuzzy sphere
written in [13] (which is obtained in the zero-slope limit of string theory) and its generalizations
undergo first order phase transitions from the fuzzy sphere phase to a matrix phase where the
fuzzy sphere vacuum collapses under quantum fluctuation. The matrix phase is the space consisting of a single point. This topology change from the two dimensional sphere to a single point
and vice versa is intrinsically a quantum mechanical process and it is related to the perturbative
UVIR mixing phenomena. This result was extended to the case of fuzzy S2 S2 in [14].
In this article we will go one step further and generalize this result to higher fuzzy complex
projective spaces. In particular we will work out the case of fuzzy CP2 in detail. We find the
possibility of first order phase transitions between CP2 and S2 , between CP2 and a matrix phase
and between S2 and a matrix phase. This richer structure of topology changes is due to the fact
that SU(3) contains also SU(2) as a subgroup besides the trivial Abelian subgroups U (1). As
we will explain generalization of our calculation to higher fuzzy complex projective spaces is
obvious and straightforward.
For other approaches to topology change using finite dimensional matrix algebra and fuzzy
physics see [15,16].
1. Fuzzy CP2
In this section we will follow [1719]. Let Ta , a = 1, . . . , 8, be the generators of SU(3) in the
symmetric irreducible representation (n, 0) of dimension N = 12 (n + 1)(n + 2). They satisfy
[Ta , Tb ] = ifabc Tc
(1.1)
and
1
Ta2 = n(n + 3) |n|2 ,
3
dabc Ta Tb =
2n + 3
Tc .
6
(1.2)
Let ta = 2a (where a are the usual Gell-Mann matrices) be the generators of SU(3) in the
fundamental representation (1, 0) of dimension N = 3. They also satisfy
1
2ta tb = ab + (dabc + ifabc )tc ,
3
1
1
tr3 ta tb = ab ,
(1.3)
tr3 ta tb tc = (dabc + ifabc ).
2
4
The N -dimensional generator Ta can be obtained by taking the symmetric product of n copies
of the fundamental 3-dimensional generator ta , viz.
Ta = (ta 1 1 + 1 ta 1 + + 1 1 ta )symmetric .
(1.4)
In the continuum CP2 is the space of all unit vectors | in C3 modulo the phase. Thus ei |,
for all [0, 2[, define the same point on CP2 . It is obvious that all these vectors ei |
correspond to the same projector P = ||. Hence CP2 is the space of all projection operators
of rank one on C3 . Let HN and H3 be the Hilbert spaces of the SU(3) representations (n, 0) and
(1, 0), respectively. We will define fuzzy CP2 through the canonical SU(3) coherent states as
follows. Let n be a vector in R8 , then we define the projector
1
P3 = 1 + na ta .
3
(1.5)
169
The requirement P32 = P3 leads to the condition that n is a point on CP2 satisfying the equations
4
n2a = ,
3
[na , nb ] = 0,
2
dabc na nb = nc .
3
(1.6)
We can write
n, 33, n|.
P3 = |
(1.7)
We think of |
n, 3 as the coherent state in H3 (level 3 3 matrices) which is localized at the point
n of CP2 . Therefore the coherent state |
n, N in HN (level N N matrices) which is localized
around the point n of CP2 is defined by the projector
PN = |
n, N N, n| = (P3 P3 P3 )symmetric .
(1.8)
We compute that
1
n, 3|ta |
n, 3 = na ,
tr3 ta P3 =
2
n
trN Ta PN =
n, N |Ta |
n, N = na .
2
(1.9)
Hence it is natural to identify fuzzy CP2 at level N = 12 (n+1)(n+2) (or CP2n ) by the coordinates
operators
2
x a = Ta .
n
(1.10)
They satisfy
[xa , xb ] =
2i
fabc xc ,
n
xa2 =

3
4
1+
,
3
n
dabc xa xb =

3
2
1+
xc .
3
2n
(1.11)
Therefore in the large N limit we can see that the algebra of xa reduces to the continuum algebra
of na . Hence xa na in the continuum limit N .
The algebra of functions on fuzzy CP2n is identified with the algebra of N N matrices MatN
generated by all polynomials in the coordinates operators xa . Recall that N = 12 (n + 1)(n + 2).
The left action of SU(3) on this algebra is generated by (n, 0) whereas the right action is generated by (0, n). Thus the algebra MatN decomposes under the action of SU(3) as
n

(n, 0) (0, n) =
(p, p).
(1.12)
p=0
A general function on fuzzy CP2n is therefore written as

F=
n

(p)
FI 2 ,I
3 ,Y
(p,p)
,
3 ,Y
TI 2 ,I
(1.13)
p=0
(p,p)
TI 2 ,I ,Y are SU(3) polarization tensors in the irreducible representation (p, p). I 2 , I3 and Y
3
are the square of the isospin, the third component of the isospin and the hypercharge quantum
numbers which characterize SU(3) representations.
The derivations on fuzzy CP2n are defined by the commutators [Ta , . . .]. The Laplacian is then
obviously given by N = [Ta , [Ta , . . .]]. Fuzzy CP2n is completely determined by the spectral
170
triple CP2n = (MatN , N , HN ). Now we can compute

n, N|F |
n, N = FN (
n) =
trN F PN =
n

(p)
FI 2 ,I
3 ,Y
(p,p)
(
n),
3 ,Y
YI 2 ,I
(1.14)
p=0
(p,p)
YI 2 ,I ,Y (
n)
3
are SU(3) polarization tensors defined by
(p,p)
(p,p)
(
n) =
n, N|TI 2 ,I ,Y |
n, N.
3 ,Y
3
(1.15)
YI 2 ,I
Furthermore we can compute

n, N|[Ta , F ]|
n, N = (La FN )(
n),
trN [Ta , F ]PN =
La = ifabc nb c .
(1.16)
And
n, N |F G|
n, N = FN GN (
n).
trN F GPN =
(1.17)
The star product on fuzzy CP2n is found to be given by [19]

n) =
FN GN (
n

(n p)!
p=0
p!n!
n)b1 bp GN (
n),
Ka1 b1 Kap bp a1 ap FN (
2
Kab = ab na nb + (dabc + ifabc )nc .
3
(1.18)
2. Fuzzy gauge fields on CP2n

We will introduce fuzzy gauge fields Aa , a = 1, . . . , 8, through the covariant derivatives Da ,
a = 1, . . . , 8, as follows
Da = Ta + Aa .
(2.1)
Da are N N matrices which transform under the action of U (N ) as follows Da U Da U +

where U U (N ). Hence Aa are N N matrices which transform as Aa U Aa U + +
U [Ta , U + ]. In order that the field A be a U (1) gauge field on fuzzy CP2n it must satisfies some
additional constraints so that only four of its components are non-zero. These are the tangent
components to CP2n . The other four components of A are normal to CP2n and in general they will
be projected out from the model.
Let us go back to the continuum CP2 and let us consider a gauge field Aa ,1 a = 1, . . . , 8,
which is strictly tangent to CP2 . By construction this gauge field must satisfy
T
Ab ,
Aa = Pab
P T = (na A dta )2 .
(2.2)
2
is the projector which defines the tangent bundle over CP . The normal bundle over CP2
will be defined by the projector P N = 1 P T . Explicitly these are given by
PT
T
Pab
= nc nd (A dtc )ae (A dtd )eb = nc nd fcae fdbe ,
N
Pab
= ab nc nd fcae fdbe .
(2.3)
1 Remark that we are using the same symbol as in the fuzzy case. However this A is a function on continuum CP2 as
a
opposed to the Aa in the fuzzy setting which is an N N matrix.
171
In above we have used the fact that the generators in the adjoint representation (1, 1) satisfy
T = n P T = 0. Hence the condition
(A dta )bc = ifabc . Remark that we have the identities na Pab
b ab
(2.2) takes the natural form
na Aa = 0.
(2.4)
This is one condition which allows us to reduce the number of independent components of Aa by
one. We know that there must be three more independent constraints which the tangent field Aa
must satisfy since it has only 4 independent components. To find them we start from the identity
[20]
1
dabk dcdk = [ac bd + bc ad ab cd + fcak fdbk + fdak fcbk ].
3
Thus
nc nd dabk dcdk =
(2.5)

2
2
na nb ab + nc nd fcak fdbk .
3
3
(2.6)
By using the fact that dcdk nc nd = 23 nk we obtain

2
dabk nk = na nb ab + nc nd fcak fdbk .
(2.7)
3
Hence it is a straightforward calculation to find that the gauge field Aa must also satisfy the
conditions
1
dabk nk Ab = Aa .
(2.8)
3
T = n n and hence P N =
In the case of S2 the projector P T takes the simpler form Pab
ab
a b
ab
na nb . From Eq. (2.7) we have on CP2
2
T
Pab
= dabc nc na nb + ab ,
3
1
N
Pab
= dabc nc + na nb + ab .
3
(2.9)
If we choose to sit on the north pole of CP2 , i.e., n = (0, 0, 0, 0, 0, 0, 0, 2 ) then we can find
3
that P T = diag(0, 0, 0, 1, 1, 1, 1, 0) and as a consequence P N = (1, 1, 1, 0, 0, 0, 0, 1). So A dta ,

a = 1, 2, 3, 8 correspond to the normal directions while A dta , a = 4, 5, 6, 7 correspond to the
tangent directions.
Indeed by substituting n = (0, 0, 0, 0, 0, 0, 0, 2 ) in Eq. (2.8) and using d8ij = 1 ij where
3
1
i, j = 1, 2, 3 and d8 =
where = 4, 5, 6, 7 and d888 = 1 we get A1 = A2 = A3 =
2 3
3
A8 = 0 which is what we want. In fact (2.8) already contains (2.4). In other words it contains
exactly the correct number of equations needed to project out the gauge field Aa onto the tangent
bundle of CP2 .
Let us finally say that given any continuum gauge field Aa which does not satisfy the constraints (2.4) and (2.8) we can always make it tangent by applying the projector P T . Thus we
will have the tangent gauge field
2
T
Ab = dabc nc Ab na (nb Ab ) + Aa .
ATa = Pab
(2.10)
3
Similarly the fuzzy gauge field must satisfy some conditions which should reduce to (2.4) and
(2.8). As it turns out constructing a tangent fuzzy gauge field using an expression like (2.2) is a
172
highly non-trivial task due to (1) gauge covariance problems and (2) operator ordering problems.
However implementing (2.4) and (2.8) in the fuzzy setting is quite easy since we will only need
to return to the covariant derivatives Da and require them to satisfy the SU (3) identities (1.2),
viz.
1
2n + 3
Da2 = n(n + 3),
(2.11)
dabc Da Db =
Dc .
3
6
So Da are almost the SU(3) generators except that they fail to satisfy the fundamental commutation relations of SU(3) given by Eq. (1.1). This failure is precisely measured by the curvature
of the gauge field Aa , namely
Fab = i[Da , Db ] + fabc Dc = i[Ta , Ab ] i[Tb , Aa ] + fabc Ac + i[Aa , Ab ].
(2.12)
The continuum limit of this object is clearly given by the usual curvature on CP2 , viz. Fab =
iLa Ab iLb Aa + fabc Ac + i[Aa , Ab ]. To check that this fuzzy gauge field Aa has the correct
degrees of freedom we need to check that the identities (2.11) reduce to (2.4) and (2.8) in the
continuum limit n . This fact is quite straightforward to verify and we leave it as an exercise.
Next we need to write down actions on fuzzy CP2n . The first piece is the usual YangMills
action
SYM =
1
2
TrN Fab
.
4g 2
(2.13)
By construction it has the correct continuum limit. TrN is the normalized trace TrN 1 = 1.
The second piece in the action is a potential term which has to implement the constraints
(2.11) in some limit. Indeed we will not impose these constraints rigidly on the path integral
but we will include their effects by adding to the action a very special potential term. In other
words we will not assume that Da satisfy (2.11). To the end of writing this potential term we will
introduce the four normal scalar fields on fuzzy CP2n by the equations (see Eqs. (2.11))

1
1
1
1
1
Da2 n(n + 3) = xa Aa + Aa xa + A2a na Aa
=
(2.14)
n
3
2
2
n
and

1
2n + 3
dabc Da Db
Dc
n
6
1
1
2n + 3
1
1
= dabc xa Ab + dabc Aa xb
Ac + dabc Aa Ab dabc na Ab Ac . (2.15)
2
2
6n
n
3
We add to the YangMills action the potential term
c =
V0 = TrN 2 + M 2 TrN a2 .
(2.16)
In the limit where the parameters and M 2 are taken to be very large positive numbers we can
see that only configurations Aa (or equivalently Da ) such that = 0 and c = 0 dominate the
path integral which is precisely what we want. This is the region of the phase space of most
interest. This is the classical prediction.
However in the quantum theory we will find that the parameter must be related to M 2 in
some specific way in order to kill exactly the normal components of Aa . This result (which we
will show shortly in the one-loop quantum fuzzy theory) is the quantum analogue of the classical
continuum statement that Eq. (2.8) contains already (2.4).
173
3. The classical and one-loop quantum actions on CP2n

The total action is then given by
1
2
TrN Fab
+ TrN 2 + M 2 TrN c2
2g 2

1
1
3n
= 2 TrN [Da , Db ]2 + ifabc Da Db Dc + 2 TrN + TrN 2
4
g
4g
S1 =
+ M 2 TrN c2 .
(3.1)
This is essentially the same action considered in [20]. However this action is different from the
action considered in [10] which is of the form

1
2i
1
S0 = 2 TrN [Da , Db ]2 + fabc Da Db Dc .
(3.2)
4
3
g
The first difference is between the cubic terms which come with different coefficients. The second
more crucial difference is the presence of the potential term in our case. The linear term in is
actually a part of the YangMills action.
The equations of motion derived from the action S0 are
[Da , Fab ] = 0.
(3.3)
These are solved by the fuzzy CP2n configurations

Da = T a
(3.4)
and also by the diagonal matrices

Da = diag da1 , da2 , . . . , daN .
(3.5)
We think of these diagonal matrices (including the zero matrix) as describing a single point in
accordance with the IKKT model [5].
More interestingly is the fact that these equations of motion are also solved by the fuzzy S2N
configurations
Di = Li ,
i = 1, 2, 3 and D = 0,
= 4, 5, 6, 7, 8.
(3.6)
Indeed in this case [Da , Fab ] = [Di , Fib ]. For b = j this is equal to 0 because fij k = ij k whereas
for b = this is equal to zero because fij = 0. In above Li are the generators of SU(2) in the
irreducible representation N1
2 .
The equations of motion derived from the action S1 are on the other hand given by

1
2n + 3
i
2
[Db , Fab ] + 2 fabc Fbc + 2{, Da } + M 2dabc {c , Db }
a = 0.
3
g2
2g
(3.7)
Now the only solutions of these equations of motion are the CP2n configurations (3.4). Thus
the potential term has eliminated the diagonal matrices (3.5) as possible solutions. In fact this
classical observation will not hold in the quantum theory for all values of the parameter M 2
since there will always be a region in the phase space of the theory where the vacuum solution
is not Da = Ta but Da = 0. However when we take M 2 to be very large positive number then
we can see that Da = Ta becomes quantum mechanically more stable. Hence by neglecting the
174
potential term we cannot at all speak of the space CP2n since it will collapse rather quickly under
quantum fluctuations to a single point.
The potential term has also eliminated the fuzzy S2N configurations (3.6) as possible solutions.
In fact even if we set M = = 0 in the above equations of motion the fuzzy S2N configurations
(3.6) are not solutions.
The other major difference between S1 and S0 is that if we expand around the fuzzy CP2n
solution Da = Ta by writing Da = Ta + Aa and then substitute back in S1 and S0 we find that
S0 does not yield in the continuum limit the usual pure gauge theory on CP2 . It contains an extra
piece which resembles the ChernSimons action (although it is strictly real). We skip here the
corresponding elementary proof (see Appendix A). S1 will yield on the other hand the desired
pure gauge theory on CP2n in the limit M 2 and hence it has the correct continuum limit.
If we do not take the limit M 2 then S1 will describe a gauge theory coupled to 4 adjoint
scalar fields which are the normal components of Aa .
The only motivation for S0 as far as we can seeis its similarity to the fuzzy S2 action
which looks precisely like S0 with the replacement fabc abc . This fuzzy sphere action was
obtained in string theory in the limit 0 when we have open strings moving in a curved
background with an S3 metric in the presence of a NeveuSchwarz B-field. However it is found
that perturbation theory with S0 is simpler than perturbation theory with S1 . More importantly it
is found that S0 allows for some new topology change which does not occur with S1 . In particular
the transitions CP2n S2N and S2N CP2n are possible in the quantum theory of S0 .
Using the background field method we find that the one-loop effective action in the gauge
2
1
= 1 + 2gn2i is given by (see Appendix A)
i [D] = Si [D] +
1
i
T R log Da2 .
Tr8 T R log ab
2
(3.8)
Where
Si [D] =

1
1
2
2
Tr
,
D
]
+
i
f
D
D
D
[D
N
a
b
i abc a b c + i TrN + i TrN
4
g2
+ Mi2 TrN c2 .
(3.9)
And

3i
2g 2 i
i
= Dc2 ab 2iFab + 2i 1
fabc Dc +
ab
ab
2
n

2g 2 Mi2
2g 2 i
+
(4D
D
+
2n
)
+
daa c dbb c Da Db + 4daa c dbb c Da Db
a b
ab
n2
n2

2n + 3
2n + 3 2
2
ab .
(3.10)
dabc Dc + 2ndabc c +
3
6
The trace T R corresponds to the left and right actions of operators on matrices whereas Tr8 is
the trace associated with 8-dimensional rotations. Given a matrix O the operator O is given by
O(. . .) = [O, . . .], for example Da (Qa ) = [Da , Qa ]. For S0 we have 0 = 23 , 0 = 0 = M0 = 0
3n
while for S1 we have 1 = 1, 1 = 4g
2 , 1 = and M1 = M.
175
4. Fuzzy CP2 phase and a stable fuzzy sphere phase

Fuzzy CP2 phase
Let us first neglect the potential term in Si , i.e., we will set i = Mi = 0 or equivalently V0 = 0
in S1 . The effective potential is given by the formula (3.8), viz.
1
i
T R log Da2 ,
Tr8 T R log ab
2
where the background field is chosen such that
i [D] = Si [D] +
(4.1)
Da = Ta .
(4.2)
vacuum Da = Ta
The reason is simply because we want to study the stability of the fuzzy
against quantum fluctuations. Hence is an order parameter which measures in a well defined
obvious sense the radius of CP2n . Let us compute the classical potential in this configuration, we
have

3|n|2 1 4 i 3 g 2 i 2
+
1 .
Si [D] = 2
(4.3)
4
2
3n
g
CP2n
The main quantum correction is equal to the trace of the logarithm of the Laplacian i which is
given by the simple formula

2g 2 i
3i
i
2 2
ab + 2i
fabc Lc ,
ab = Lc +
(4.4)
n
2
where La (. . .) = [Ta , . . .]. There are two cases to consider. For both S0 and S1 we obtain the
quantum correction (see Appendix A)
i [D] =

1
Tr8 T R log 2 L2a 18 T R log 2 L2a = +6N 2 log + const.
2
Case 1. For S0 we have 0 = 2/3 and 0 = 0. Thus the quantum effective potential is

0 [D]
2
1 4 1 3
= 2 2 + log + const.
Veff =
3
6N 2
3n g 4
(4.5)
(4.6)
The quantum minimum of the model is given by the value of which solves the equation
= 0. It is not difficult to convince ourselves that this equation of motion will admit a soVeff
lution only up to an upper critical value g of the gauge coupling constant g beyond which the
configuration Da = Ta collapses. At this value g the potential Veff becomes unbounded from
below. The conditions which will yield the critical value g are therefore

=
Veff
1
2
3
2 + = 0,
2
2
3n g

=
Veff

2
2
1
3 2 2 = 0.
2
2
3n g
(4.7)
We find immediately
3
= ,
4
n2 g2

2 3 4
=
= 0.0703.
9 4
(4.8)
Above the value g we do not have a fuzzy CP2n , in other words the space CP2n evaporates at this
point. This critical point separates two distinct phases of the model, in the region above g we
176
have a matrix phase while in the region below g we have a fuzzy CP2n phase in which the
model admits the interpretation of being a U (1) gauge theory on CP2 .
By going from small values of g (g g corresponding to the fuzzy CP2n phase) towards
large values of g we get through the value g where the space CP2n decays. Looking at this
process the other way around we can see that starting from large values of g (g > g corresponding to the matrix phase) and going through g we generate the space CP2n dynamically.
It seems therefore that we have generated quantum mechanically the spectral triple which defines
the space CP2n .
Furthermore we note that in the matrix phase we have a U (N ) gauge theory reduced to a
point where N is the size of the matrices since the minimum there is given by diagonal matrices.
The important point is that in this phase the gauge group is certainly not U (1). Let us recall that
in the fuzzy CP2n phase we had a U (1) gauge theory. Hence across the transition line between
the fuzzy CP2n phase and the matrix phase the structure of the gauge group also changes.
Thus we obtain in this model in correlation with the topology change across the critical line a
novel spontaneous symmetry breaking mechanism.
For the purpose of comparing with the numerical results of [10] we define the coupling constant such that 4 N = 1/g 2 . Then the critical value (4.8) is seen to occur at
= 2.309.
(4.9)
This is precisely the result of the Monte Carlo simulation reported in Eq. (3.2) of [10].
3n
Case 2. For S1 we have 1 = 1 and 1 = 4g
2 and hence the effective potential is given by

1 [D]
2
1 4 1 3 1 2
Veff =
(4.10)
=
+ log + const.
2
4
6N 2
3n2 g 2 4
A direct calculation yields the critical values

9 + 17
2
2
,
n2 g2 =
= 0.02552.
=
16
4
3
(4.11)
This g is smaller than the g obtained in (4.8) and hence the fuzzy CP2n is more stable in the
model S0 than it is in the model S1 which is largely due to the linear term proportional to
in S1 . In other words attempting to put true gauge theory on fuzzy CP2n causes the space to decay
more rapidly. However for S0 the true vacuum is the fuzzy sphere S2N and not the fuzzy CP2n as
we will now discuss.
A stable fuzzy sphere phase
We know that there is also a fuzzy sphere solution (3.6) for the model S0 . We consider then
the background field
Di = Ti ,
i = 1, 2, 3,
D = 0,
= 4, 5, 6, 7, 8.
(4.12)
We want now to study the stability of this vacuum against quantum fluctuations. The is now
an order parameter which measures the radius of the fuzzy sphere S2N . The classical potential in
this configuration is

2c2 1 4 1 3
S0 [D] = 2
(4.13)
.
3
g 4
177
In above c2 =
N 2 1
4
is the Casimir of SU(2) in the irreducible representation

|n|2
N1
2
(N = 12 (n +
1)(n + 2)). It is clear that 2c2 3|n|2 and c2 1 in the large n limit. Hence the action (4.13)
around the classical minimum = 1 is much smaller than the classical action (4.3). In other
words the fuzzy sphere is more stable than the fuzzy CP2n in this case.
The quantum corrections are given in this case by
1
1
Tr3 T R log ij + Tr5 T R log T R log 2 L2i .
2
2
In above

2g 2
3
ij + 2i
ij k Lk ,
ij = 2 L2k +
n
2

2g 2
2 2
3ifi Li .
= Lk +
n
(4.14)
(4.15)
Following the same arguments of the previous section (only now it is SU(2) representation theory
which is involved) we have in the large n limit
1
1
Tr3 T R log ij T R log 2 L2i = Tr3 T R log 2 ij T R log 2 +
2
2
= N 2 log + .
(4.16)
We can also argue that we have

1
(4.17)
Tr5 T R log + = 5N 2 log + .
2
In other words the configurations (4.12) although they are fuzzy sphere configurations they know
(through their quantum interactions) about the other SU(3) structure present in the model. Classically this SU(3) structure is not detected at all by these configurations in the classical potential
(4.13). The effective potential becomes in this case

0 [D]
1
1 4 1 3
=
+ log + const.
Veff =
(4.18)
3
6N 2
12g 2 4
A direct calculation yields the critical value

3
1 3 4
2
= ,
= 0.0087875.
g =
4
36 4
(4.19)
In terms of the coupling define by 4 = 1/g 2 the critical value g reads

= 3.26.
(4.20)
This is again what is measured in the Monte Carlo simulation of the model S0 as it is reported in
Eq. (4.2) of [10]. Therefore we have a fuzzy sphere phase above and a matrix phase below .
The model S0 can also be in a fuzzy CP2n phase for n2 g2 below the second value of (4.8) which
for large enough n is much smaller than the value n2 g2 with g2 given by the second equation of
(4.19). However we have seen in the previous paragraph that this CP2n will decay rather quickly
to a single point which (by the discussion of the present section) can only happen by going first
across a fuzzy sphere phase. We have then the transition pattern CP2n S2N {0}. In the limit
where 4 = 2/n2 g 2 is kept fixed we can see that the above critical value (4.19) is infinitely large
178
which means that the model S0 is mostly in the fuzzy sphere phase. The matrix phase shrinks to
zero and the fuzzy sphere is completely stable in this limit since the fuzzy CP2n phase can occur
only at very small values of the coupling constant n2 g 2 .
5. The large mass limit and the transition CP2 S2
The large mass limit
Now we include the effect of the potential term V0 . The relevant model is given by the action S1 . Naturally the calculation becomes more complicated in this case. The classical potential
in the configuration Da = Ta is

2 g 2 M 2 2
2
S1 [D]
2
1 4 1 3 1 2 g2 2
(5.1)
=
+
+
1
+
.
2
4
9
27
6N 2
3n2 g 2 4
The most important quantum correction is given by the determinant of

3
3
1
2 2
ab = Lc +
ab 2
(A dtc Lc )ab + 2g 2 1 ab
2
2
+ 2g 2 M 2 2 ab .
(5.2)
By using the identities (1.2) and (2.5) we find that the extra contributions are given explicitly by
the expressions

1
2
4 Ta Tb + 2 2 1 Tc2 ab ,
n2

2
2n + 3
1
2n + 3 2
ab + 3
dabc Tc
2 ab = 2
6
3
n

4 2
1
1 2
+
Tc2 +
A dtc Tc
3
16
4
ba

2
2
1
1
L2c +
A dtc Lc
.
3
16
4
ba
1 ab =
(5.3)
(5.4)
In the continuum large N limit the first extra correction behaves as

2 2
1 ab .
(5.5)
3
Let us introduce the projector Pab = 3 na nb . This is a rank one normal projector which projects
1 ab = 2 na nb +
T =
vector fields along the normal direction A dt8 . Recall the rank four tangent projector Pab
1
N = d
dabc nc na nb + 23 ab and the rank four normal projector Pab
abc nc + na nb + 3 ab .
N
Then we must necessarily have Pab = Pab + Pab where Pab is a rank three normal projector which projects vector fields along the normal directions A dti , i = 1, 2, 3. It is given by
Pab = dabc nc + 14 na nb + 13 ab . We have the decomposition 1 = P T + P N = P T + P + P .
Hence

4
2
1 ab = 2 Pab + 2 1 ab
3
3

T

2 2
1
2
2
= 1 Pab + 2
(5.6)
Pab + 2 1 Pab .
3
3
3
179
Similarly in the continuum large N limit the first three terms of 2 ab takes the form (by using
T 2 P + 2 P ).
also the identity dabc nc = 13 Pab
3 ab
3 ab
First 3 terms of 2 ab =
=

2 T
1 + 4 2
1
Pab +
ab + 2 3 dabc nc
3
9
3
1 + 2 2 3 T
1 + 6 2 6
Pab +
Pab
9
9
+
1 + 2 2 + 6
Pab .
9
(5.7)
Let us remark that the coefficients in front of the projectors P and P are the masses of the
normal components of the gauge field and hence they must be positive. For example the mass of
2 2
2 = 2g 9M (2(1 + 3 ) 2 + 6 + 1 6 ) where
the normal components Pab Ab is given by M 2 m
= M2 . This is positive definite for all values 0 of the radius of CP2n if is such that
1 + 3 0 and 1 6 > 0. Thus must be in the range 13 < 16 . Since must be positive
we obtain the condition
0 <
M2
.
6
(5.8)
2
2 = 2g 9M (6(1 + 3 ) 2 6 + 1
The mass of the normal component Pab Ab is given by M 2 m
6 ). The requirement that this mass must be positive definite gives now the condition that the
radius can only be in the range

1 12 13 + 16
,
<

6 13 +

1 + 1 12 13 + 16
.
> +

6 13 +
1
(5.9)
We remark that for all allowed values of we have > 0 and + < 1 so we can still access the
limits 1 and 0+ although there is now a forbidden gap between these two important
regions.
2 2
T A is given by M 2 m2 = 4g M (1+3 ) ( 1)(
The mass of the tangent components Pab
b
T
9
16
).
This
is
not
always
positive
in
the
range
(5.9).
However
this
mass
formally
vanishes in
2+6
the limit M where the most probable value of the radius of CP2n is 1. Finally the last
correction of the inverse propagator coming from the addition of the potential V0 (which is given
explicitly by the last term in (5.4)) is also negative. Remark that this correction is proportional to
2 and as a consequence we will not need to compute it explicitly (see below).
We are now ready to compute the determinant. We have
1
T
ab
= ab + M 2 m2T Pab
+ M 2m
2 Pab + M 2 m
2 Pab ,
(5.10)
180
where

3
3
ab 2
(A dtc Lc )ab
ab = 2 L2c +
2
2

2g 2 M 2 2
1
1 2
2
+
L
A
dt
L
c
c
c
16
4
3n2
ba
Thus
1
Tr8 T R log 1 = log
2
(5.11)
dAa exp TrN Aa ab Ab
2

M 2 m2T TrN ATa M 2 m
2 TrN (A a )2 M 2 m
2 TrN (A a )2 . (5.12)
From the last two terms we get in the large M limit the two delta functions (A a ) and (A a ) and
as a consequence the determinant reduces to

1
1
Tr8 T R log 1 = Tr8 T R log P T + M 2 m2T P T ,
2
2
M .
(5.13)
In above it is consistent to neglect the mass term M 2 m2T P T since in the large mass limit M
this term is subleading as we have discussed. The eigenvalues of the operator A dtc Lc were
computed in the appendix. We found that the second term in (which is proportional to ab )
and the third term (which is proportional to (A dtc Lc )ab ) can be neglected in the large n limit
2
compared to L2c ab . For example the eigenvalues of L2c are given by p +2p
with p = 0, . . . , n
3
whereas the eigenvalues of A dtc Lc are found to be at most linear in p and hence in the large n
limit (where large values of p which are of the order of n are expected to contribute the most)
we can make the approximation

2g 2 M 2
1
1 2
2
ab 2 L2c ab
(5.14)
+
L
A
dt
L
+ .
c c
c
16
4
3n2
ba
Thus the quantum effective potential is

2
2 T
1 [D] S1 [D]
1
1
VM
+ . (5.15)
=
+
Tr8 T R log P T R log
6N 2
6N 2
6N 2 2
The last term comes from the ghost contribution. The final result is

2 g 2 M 2 2
2
2
1 4 1 3 1 2 g2 2
1
1 +

+ log .
VM = 2 2 + +
4
2
4
9
27
3
3n g
(5.16)
The calculation of the critical values in terms of the mass parameters M 2 = g 2 M 2 2 and is
done in the same way as before and it yields the following equations. The critical radius occurs
at the solutions of the equation

4M 2
9
1 M 2
1
4M 2
1
1+
(5.17)
+
2 1 +
+
2
= 0.
9
3
8
27
4
9
3
2 This combination is the correct definition of the mass parameter in these models which should be used from the start.
In the limit M we get the solution
9+ 81 + 64(1 + 3 )(1 6 )

,
16(1 + 3 )
M .
181
(5.18)
The choice of the plus sign instead of the minus sign is so that when goes to zero (in other
words 0) this solution will reduce to the first equation of (4.11). This agreement is due to
the fact that the limit 0 is formally equivalent to the limit M 0 (since = M 2 ). Indeed
for very small values of M we get the potential

2 g 2 M 2 2
2
2
1
1
g2 2
1
1 +

VM0 = 2 2 4 3 + 2 +
2
4
9
27
3n g 4
+ log .
(5.19)
This will also lead to Eq. (5.17) which for M 0 admits the solution given by the first equation
of (4.11).
The critical value of the coupling constant g (or equivalently ) is given on the other hand
by the equation

n2 g2
1 2 3
1 2M 2
1
4M 2
1
= 4 =
1+
+
2
.
2
2
4
27
2
9
3

(5.20)
Hence in the limit M we get the behavior

4 =
18
18

9

.
2
2
2
2
2
2
M + 4 3 + 4 (3 2)
M + 4 23
The equation of motion
VM
(5.21)
= 0 could admit in general four real solutions where the one
with the least energy can be identified with the radius of fuzzy CP2n . This solution is found to be
very close to 1. However this is only true up to an upper value of the gauge coupling constant g
(or equivalently a lower bound of )
for every fixed value of M beyond which the equation of
motion ceases to have any real solutions. At this value the fuzzy CP2n collapses under the effect
of quantum fluctuations and we cross to a pure matrix phase. As the mass M is sent to infinity it
is more difficult to reach the matrix phase and hence the presence of the mass makes the fuzzy
CP2n solution Da = Ta more stable. In fact when M 2 the critical value approaches
zero.
The transition CP2 S2
We repeat the large mass analysis for the model S0 . In other words we add the potential V0
to the action S0 and study the effective potential when M, . The interest in this action
lies in the fact that it admits (at least for M = = 0) a fuzzy sphere solution and hence we can
contemplate a transition (at the level of the phase diagram) between fuzzy CP2n and fuzzy S2N
when we take the limit M, 0. As before we consider fuzzy CP2n configurations Da = Ta ,
a = 1, . . . , 8. For S0 + V0 (in other words non-zero values of M and ) these configurations are
in fact the true vacuum as we have discussed previously. When V0 = 0 the fuzzy S2N configurations become the true minimum. The calculation of the quantum corrections with non-zero V0 is
exactly identical to what we have done in the previous paragraphs and thus we end up with the
182
effective potential

2 g 2 M 2 2
2
2
1 4 1 3 g2 2
1
VM = 2 2 +
1 +

+ log .
3
9
27
3
3n g 4
(5.22)
In the large M limit we get the same critical value (5.18). The critical value of g (or equivalently
)
is found on the other hand to be given by
4 =
2
18
=
.

n2 g2 M 2 2 + 4 2 + 9 3
3
2
(5.23)
So again in the large mass limit the fuzzy CP2n phase is stable even for the model S0 .
However we know from our previous discussion that in the limit M 0 the minimum of the
model S0 should tend to the fuzzy sphere solutions. Thus it is important to consider also the fuzzy
sphere configurations Di = Ti , i = 1, 2, 3, D = 0, = 4, 5, 6, 7, 8. The classical potential in
these configurations becomes

S0 [D]
1
1 4 1 3 M 2 c2 2 |n|2 2
=
3
c2
6N 2
12g 2 4
2n2

2
2
M
c2
2n + 3
+
2 + 4 4 .
(5.24)
2
2
6n
3n
The quantum corrections in the limit M 0 should be given by (4.16) and (4.17). We get then
the effective potential

0 [D]
1
1 4 1 3 M 2 c2 2 |n|2 2
VM0 =
=
3
c2
6N 2
12g 2 4
2n2

2
2
M
c2
2n + 3
+
2 + 4 4 + log .
(5.25)
2
2
6n
3n
The critical values are

3 4M 2 9 c2
1 2n + 3 2
3
3c2
=
1+
+ 4+
+ O M 4 ,
2
2
4
3
3
n
2
8n
8n
6n

256 2
1
2
96
M + O M 4 .
4 = 2 2 = 2 3

2
2
3
n g
n 1 + 4M 1 + 3 3 2n+3
3
(5.26)
(5.27)
6n2
So when M 0 this goes to zero which is consistent with the result (4.19). This equation tell
us how we actually approach this critical value = 0.
The intersection of this equation with (5.23) gives a one-loop estimation of the value M T at
which the vacuum of the model S0 goes from a fuzzy sphere S2N to a fuzzy CP2n as we increase
Equivalently the intersection point occurs at the value M T at which the
the mass parameter M.
vacuum of the model S0 goes from a fuzzy CP2n to a fuzzy sphere S2N as we decrease M.
6. Conclusion
In this article we have studied the one-loop effective potential for two models of U (1) gauge
theory on fuzzy CP2n . The first model is given by the action S0 + V0 and the second model is given
by the action S1 . Each model is characterized by 3 parameters. (i) The gauge coupling constant g 2
183
or equivalently 4 =
2
, (ii) the mass M of the normal components of the 8-dimensional gauge
g 2 n2
parameter = M 2 which gives an extra mass for the normal component in
field and (iii) the

the direction A dt8 . The term in the action proportional to is not needed in the classical theory
while in the quantum theory the parameter must be in the range (5.8). The order parameter
(the variable) of the effective potential is the radius of the fuzzy CP2n and thus by studying the
stability of this potential we can test the stability of the space as a whole against the effect of
quantum fluctuations of the gauge field theory. The second term in the effective potential (the
log term) is not convex which implies that there is a competition between the classical potential
and the logarithmic term which depends on the values of M and g. The parameter plays no
further role at this stage. We found that there exists values of the gauge coupling constant g and
the mass M for which the fuzzy CP2n solutions are not stable. This instability is believed to be
related to (or is a reflection of) the perturbative UVIR mixing phenomena of the quantum gauge
field theory. The connection between the two effects was established explicitly for the case of the
lower dimensional co-adjoint orbit SU (2)/U (1) which is the case of the fuzzy sphere [9]. See
also [14].
The phase structure of the U (1) models on fuzzy CP2n which are studied in this article reads
as follows.
The model S1
This is the correct model which describes U (1) gauge theory in the continuum limit at least
classically. The minimum of the model (for non-zero potential) can only be fuzzy CP2n . There
are two phases. In the fuzzy CP2n phase we have a U (1) gauge theory on fuzzy CP2n whereas in
the matrix phase the fuzzy CP2n configurations Da = Ta collapse and we end up with a U (N )
gauge theory on a single point. We have described in this article the qualitative behavior of a first
order phase transition which occurs between these two regions of the phase space. However it is
obvious from the critical line (5.23) that when the mass M of the four normal scalar components
of the 8-dimensional gauge field on fuzzy CP2n goes to infinity it is more difficult to reach the
transition line. In this limit the fuzzy CP2n phase dominates while the matrix phase shrinks to
zero. Therefore we can say that we have a non-perturbative regularization of U (1) gauge theory
on fuzzy CP2n .
The model S0 + V0
This is a string-theory-inspired gauge model which does not go in the continuum limit to the
usual U (1) gauge theory on CP2 even classically. Indeed it can be shown that it contains in
the continuum limit (in addition to the usual YangMills term) a ChernSimons-like term (see
Appendix A). However this model has a more interesting phase structure since it allows for the
(quantum) transitions between fuzzy S2N and fuzzy CP2n . The main reason behind this remarkable
feature lies in the fact that when M the absolute minimum of the model is the fuzzy CP2n
configurations Da = Ta whereas in the limit M 0 the absolute minimum of the model is
the fuzzy sphere configurations Di = Ti , i = 1, 2, 3 and D = 0, = 4, 5, 6, 7, 8. The phase
1
diagram of this model with the particular value = 12
is plotted in Fig. 1 for illustration. The
phase diagram consists of 3 phases.
(1) The fuzzy CP2n phase: This is the region with M M T and above the line (5.23) where the
absolute minimum of the model is the fuzzy CP2n configurations Da = Ta and where the
184
1 . In this case = 1 and hence the large mass expansion (5.23)

Fig. 1. The phase diagram for the model S0 with = 12
108
4
2
while the small mass expansion (5.27) is 4 = 64

becomes = 2
9 m with m = M. The intersection point occurs
4m +27
at M T . This looks like a triple point.
field theory is some U (1) gauge theory on fuzzy CP2n . Recall that M T is the value of the
mass parameter M at which the two curves (5.23) and (5.27) intersect. The fuzzy CP2n phase
dominates the phase diagram when M .
(2) The matrix phase: This phase shrinks to zero when M . This occurs at the points of the
phase diagram which are below the two lines (5.23) and (5.27).
(3) The fuzzy S2N phase: These are the points which have M M T and which are above the line
(5.27) where the absolute minimum of the model is the fuzzy S2N configurations Di = Ti ,
i = 1, 2, 3, D = 0, = 4, 5, 6, 7, 8 and where the field theory is a U (1) gauge theory on
fuzzy S2N with complicated coupling to 6 adjoint scalars.
Generalization to higher gauge groups U (k) with and without fermions should be straightforward if we are only interested in the effective potential and topology change. Similarly
generalization to higher co-adjoint orbits CPd = SU(N )/U (N 1) with d = N 1 should also
be straightforward since the corresponding actions will be exactly of the same form as S0 and S1
and only we need to work with the group theory of SU(N )s instead of SU(3). In particular we
expect that there will more possibilities for topology change in higher co-adjoint orbits which
relate to the fact that the group SU(N ) contains besides SU(2) the groups SU(3), SU(4) and
many others (for high enough N ) as subgroups. Thus we may see transitions like CPd S2 ,
CP d CP 2 , CP d CP3 , . . . as well as transitions between the subspaces S2 , CP2 , CP3 , . . .
and transitions from and to the matrix (single point) phase.
Acknowledgements
The work of D.D. is supported by the College of Science-Research Center Project No:
phys/2006/04. The work of D. Dou is also supported in part by the associate scheme of Abdus
185
Salam ICTP. The work of Badis Ydri is supported by a Marie Curie Fellowship from the Commission of the European Communities (the Research Directorate-General) under contract number
MIF1-CT-2006-021797. B.Y. would like also to thank the staff at Humboldt-Universitt zu Berlin
for their help and support. In particular he would like to thank Michael Muller-Preussker and
Wolfgang Bietenholz.
Appendix A. The one-loop effective action and effective potential for zero mass
First we can compute explicitly the following classical actions

2
TrN Fab
= TrN [Da , Db ]2 + 4ifabc Da Db Dc + 3Da2 ,

2 2 2
1
1 2
2
2
2
TrN = 2 TrN Da n(n + 3)Da + n (n + 3) ,
3
9
n

1
1
1
TrN a2 = 2 TrN Da Db Da Db + faa c fbb c {Da , Db }{Da , Db }
3
6
n
2

2n + 3
2n + 3
Da2
dabc Da Db Dc .
+
6
3
(A.1)
(A.2)
(A.3)
In above we have used the identity (2.5) and the identity

fabc fabd = 3cd .
(A.4)
(0)
We can compute (with Fab = Fab i[Aa , Ab ])

S0 =
1
1
1
(0)
2
TrN Fab
2 fabc TrN Ac Fab
fabc TrN Ac Fab .
2g 2
6g
12g 2
(A.5)
Next we will study the quantization of the action

Si [D, J ] = Si [D] + TrN Ja Da ,
where
Si [D] = Si [D] + Vi [D],

1
1
2
Si [D] = 2 TrN [Da , Db ] + ii fabc Da Db Dc ,

4
g
Vi [D] = i TrN + Vi = i TrN + i TrN 2 + Mi2 TrN c2 .
(A.6)
(A.7)
For S0 = S0 + V0 we have 0 = 23 , 0 = 0 = M0 = 0 while for S1 = S1 + V1 we have 1 = 1,

3n
1 = 4g
2 , 1 = and M1 = M. Ja is a source.
We adopt the background field method to the problem of quantization of this model. We write
Da = Ba + Qa where Ba is the background field and Qa is the fluctuation field. We will fix the
gauge by adding to the action the gauge-fixing and FadeevPopov terms, viz
Sg.fixing + Sgh =

1 1
TrN [Ba , Qa ]2 TrN Ba , b+ [Ba , b].
2
2g
(A.8)
We compute

TrN [Da , Db ]2 = TrN [Ba , Bb ]2 + 2[Ba , Bb ][Qa , Qb ] + 2[Ba , Qb ]2

+ 2[Ba , Qb ][Qa , Bb ] + 4[Ba , Bb ][Ba , Qb ] + O Q3 .
(A.9)
186
And
ifabc TrN Da Db Dc = ifabc TrN Ba Bb Bc + 3ifabc TrN Ba Bb Qc
+ 3ifabc TrN Qa Qb Bc .
(A.10)
We find (by using the identity TrN [Ba , Qb ][Bb , Qa ] = TrN [Ba , Qa ]2 TrN [Ba , Bb ][Qa , Qb ])
the following expression

i
3i
1
fabc Bc , Ba Qb 2 TrN [Ba , Qb ]2

Si [D] = Si [B] + 2 TrN Bab 1
2
g
2g

3i
fabc Bc , Qb + O Q3
[Ba , Qa ]2 + 2iQa Bab 1
2

i
3
i
= Si [B] + 2 TrN Bab 1
fabc Bc , Ba Qb
2
g

3i
1
fabc Bc Qb + O Q3 .
+ 2 TrN Qa Bc2 ab Ba Bb 2iBab + 2i 1
2
2g
(A.11)
Bab is the curvature of the background curvature Ba , in other words Bab = i[Ba , Bb ] + fabc Bc .
Remark also how this action simplifies for S0 , i.e., for 0 = 2/3. This technical simplification
is a major advantage in considering S0 instead of S1 . Given a matrix O the operator O is given
by O(. . .) = [O, . . .], for example Ba (Qa ) = [Ba , Qa ].
We can also compute
2i
i
(A.12)
TrN Bb Qb + TrN Q2a ,
n
n
2i
i TrN 2 = i TrN 2 +
TrN {Bb , }Qb
n

i
(A.13)
+ 2 TrN Qa (Ba Bb + 4Ba Bb + 2nab )Qb + O Q3 ,
n

2Mi2
2n + 3
2
2
2
2
TrN dabc {Bb , c }
b Qb
Mi TrN c = Mi TrN c +
n
6

M2
+ 2i TrN Qa daa c dbb c Ba Bb + 4daa c dbb c Ba Bb
n

2n + 3
2n + 3 2
dabc Bc + 2ndabc c +
2
ab Qb + O Q3 .
3
6
(A.14)
i TrN = i TrN +
and c are the normal scalar fields corresponding to the background covariant derivative Ba ,
1
2
viz. = n1 (Ba2 |n|2 ) and c = n1 (dabc Ba Bb 2n+3
6 Bc ) where |n| = 3 n(n + 3).
Let us now introduce the actions Si [D, J ] = Si [D] + Vi [D] + TrN Ja Da and Si [B, J ] =
Si [B] + Vi [B] + TrN Ja Ba . By using the above ingredients we have immediately the result
Si [D, J ] + Sg.fixing + Sgh = Si [B, J ] + TrN Jbi Qb +
+ TrN b+ Ba2 b.

1
i
TrN Qa ab
Qb + O Q3
2
2g
(A.15)

i are given respectively by
In above Jai and ab

i
3i
2i
2i
i
Jb = Jb + 2 Bab 1
fabc Bc , Ba +
Bb +
{Bb , }
2
n
n
g

2Mi2
2n + 3
+
dabc {Bb , c }
b ,
n
6

1
3i
2g 2 i
i
= Bc2 ab +
1 Ba Bb 2iBab + 2i 1
fabc Bc +
ab
ab
2
n
2g 2 i
(Ba Bb + 4Ba Bb + 2nab )
n2

2g 2 Mi2
2n + 3

dabc Bc
daa c dbb c Ba Bb + 4daa c dbb c Ba Bb 2
+
3
n2

2n + 3 2
+ 2ndabc c +
ab .
6
187
(A.16)
(A.17)
In the following we will assume that the background fields Ba satisfy the equations of motion
2
Jbi = 0 and we will choose the gauge 1 = 1 + 2gn2i . We then obtain

1
i
TrN Qa ab
Qb + O Q3 + TrN b+ Ba2 b.
2
2g
(A.18)
The fluctuation fields Qa and the ghosts can be integrated out since they are Gaussian and one
obtains therefore the effective action
1
i
T R log Ba2 .
i [B, J ] = Si [B, J ] + Tr8 T R log ab
(A.19)
2
where

3i
i
= Bc2 ab 2iBab + 2i 1
ab
fabc Bc
2

2g 2 Mi2
2g 2 i
2g 2 i
+
(4Ba Bb + 2nab ) +
ab +
daa c dbb c Ba Bb
n
n2
n2

2n + 3
2n + 3 2
ab .
dabc Bc + 2ndabc c +
+ 4daa c dbb c Ba Bb 2
3
6
(A.20)
Si [D, J ] + Sg.fixing + Sgh = Si [B, J ] +
Now we compute the effective potential on fuzzy CP2 for V0 = 0. For the configuration (4.2)
we consider the following two cases.
Case 1. For S0 we have 0 = 23 and 0 = 0 and hence

3|n|2 1 4 1 3
1
0
0 [D] = 2
T R log 2 L2a ,
+ Tr8 T R log ab
4
3
2
g

0
ab
= 2 L2c ab ( 1) Jc2 L2c (A dtc )2 ab , J = L + A d t.
(A.21)
The total SU(3) angular momentum J corresponds to the tensor product of the irreducible repre with the adjoint representation (1, 1)
sentations (p, p) where p = 0, . . . , n (corresponding to L)
188
(corresponding to A d t ). By using Young tableaux we obtain the decomposition

(p, p) (1, 1) = (p + 1, p + 1) (p 1, p 1) (p, p) (p, p) (p 1, p + 2)
(p + 2, p 1) (p 2, p + 1) (p + 1, p 2).
(A.22)
The dimension of an irreducible representation (n1 , n2 ) of SU(3) is d(n1 , n2 ) =

+ 1)
1 2
(n2 + 1)(n1 + n2 + 2) and the quadratic Casimir is c2 (n1 , n2 ) = 3 (n1 + n1 n2 + 3n1 + 3n2 + n22 ).
Thus we can immediately compute

Jc2 L2c (A dtc )2 = Jc2 p 2 + 2p 3

= diag 2p, (2p + 4), 3, 0, p 3 .
(A.23)
1
2 (n1
In this diagonal matrix the dimensions of the first block is d(p + 1, p + 1) = (p + 2)3 , the
second block is d(p 1, p 1) = p 3 , the third block is 2d(p, p) = 2(p + 1)3 , the 4th block is
d(p 1, p + 2) + d(p + 2, p 1) = 2d(p 1, p + 2) = p(p + 3)(2p + 3) and the 5th block is
d(p 2, p + 1) + d(p + 1, p 2) = 2d(p 2, p + 1) = (p 1)(p + 2)(2p + 1).
It is obvious from Eq. (A.23) that the second term in is at most linear in p while the first
term is quadratic and hence in the large n limit (where large values of p which are of the order
of n are expected to contribute the most) we can make the approximation ab = 2 L2c ab .
Thus the quantum effective potential is

1
0 [D] = S0 [D] + Tr8 T R log 2 L2a 18 T R log 2 L2a
2

3|n|2 1 4 1 3
1
+ (8) N 2 log 2 N 2 log 2 + const
= 2
4
3
2
g

3|n|2 1 4 1 3
= 2
+ 6N 2 log + const.
4
3
g
Recall that N = 12 (n + 1)(n + 2) and |n|2 = 13 n(n + 3) and hence

[B]
2
1 4 1 3
= 2 2 + log + const.
Veff =
3
6N 2
3n g 4
Case 2. For S1 we have 1 = 1 and 1 =
3n
4g 2
(A.24)
(A.25)
and hence

3
3
1
= 2 L2c +
ab + 2i
fabc Lc .
ab
2
2
(A.26)
The only difference (as far as this quantum correction is concerned) with Case 1 is that we have
now a shifted Laplacian 2 L2c + 32 so the result for the determinant already obtained will not be
altered. We end up thus with the effective potential

[B]
2
1 4 1 3 1 2
= 2 2 + + log + const.
Veff =
(A.27)
2
4
6N 2
3n g 4
References
[1] B. Ydri, Fuzzy Physics, PhD thesis, 2001, hep-th/0110006.
[2] H. Grosse, C. Klimcik, P. Prenajder, Commun. Math. Phys. 180 (1996) 429;
H. Grosse, C. Klimcik, P. Prenajder, Int. J. Theor. Phys. 35 (1996) 231;
D. OConnor, Mod. Phys. Lett. A 18 (2003) 2423;
C. Klimcik, Commun. Math. Phys. 199 (1998) 257;
A.P. Balachandran, S. Kurkcuoglu, S. Vaidya, Lectures on fuzzy and fuzzy SUSY physics, hep-th/0511114.
[3] A. Connes, Noncommutative Geometry, Academic Press, London, 1994;
G. Landi, An Introduction to Noncommutative Spaces and their Geometry, Springer, 1997;
J.M. Gracia-Bondia, J.C. Varilly, H. Figueroa, Elements of Noncommutative Geometry, Birkhuser, 2000.
[4] B.P. Dolan, I. Huet, S. Murray, D. OConnor, hep-th/0611209.
[5] N. Ishibashi, H. Kawai, Y. Kitazawa, A. Tsuchiya, Nucl. Phys. B 498 (1997) 467491.
[6] H. Aoki, S. Iso, H. Kawai, Y. Kitazawa, T. Tada, Prog. Theor. Phys. 99 (1998) 713746;
H. Aoki, S. Iso, H. Kawai, Y. Kitazawa, T. Tada, A. Tsuchiya, Prog. Theor. Phys. Suppl. 134 (1999) 4783;
H. Aoki, N. Ishibashi, S. Iso, H. Kawai, Y. Kitazawa, T. Tada, Nucl. Phys. B 565 (2000) 176192.
[7] J. Madore, Class. Quantum Grav. 9 (1992) 6988;
J. Hoppe, MIT PhD thesis, 1982;
J. Hoppe, S.T. Yau, Commun. Math. Phys. 195 (1998) 6777.
[8] S. Iso, Y. Kimura, K. Tanaka, K. Wakatsuki, Nucl. Phys. B 604 (2001) 121.
[9] P. Castro-Villarreal, R. Delgadillo-Blando, B. Ydri, Nucl. Phys. B 704 (2004) 111153.
[10] T. Azuma, S. Bal, K. Nagao, J. Nishimura, hep-th/0405277.
[11] D. OConnor, B. Ydri, JHEP 0611 (2006) 016, hep-lat/0606013.
[12] B. Ydri, Quantum equivalence of NC and YM gauge theories in 2D and matrix theory, hep-th/0701057;
B. Ydri, The one-plaquette model limit of NC gauge theory in 2D, hep-th/0606206, Nucl. Phys. B, in press.
[13] A.Y. Alekseev, A. Recknagel, V. Schomerus, hep-th/0003187;
A.Y. Alekseev, A. Recknagel, V. Schomerus, hep-th/9812193.
[14] P. Castro-Villarreal, R. Delgadillo-Blando, B. Ydri, JHEP 0509 (2005) 066;
R. Delgadillo-Blando, B. Ydri, Towards non-commutative fuzzy QED, hep-th/0611177.
[15] A.P. Balachandran, S. Kurkcuoglu, Int. J. Mod. Phys. A 19 (2004) 33953408;
L.C. de Albuquerque, P. Teotonio-Sobrinho, S. Vaidya, JHEP 0410 (2004) 024.
[16] J. Arnlind, M. Bordemann, L. Hofer, J. Hoppe, H. Shimada, hep-th/0602290.
[17] G. Alexanian, A.P. Balachandran, G. Immirzi, B. Ydri, J. Geom. Phys. 42 (2002) 2853.
[18] H. Grosse, A. Strohmaier, Lett. Math. Phys. 48 (1999) 163179.
[19] A.P. Balachandran, B.P. Dolan, J. Lee, X. Martin, D. OConnor, hep-th/0107099.
[20] H. Grosse, H. Steinacker, hep-th/0407089.
189
Reducible gauge algebra of BRST-invariant constraints

I.A. Batalin a , K. Bering b,
a I.E. Tamm Theory Division, P.N. Lebedev Physics Institute, Russian Academy of Sciences,
53 Leninisky Prospect, Moscow 119991, Russia

b Institute for Theoretical Physics & Astrophysics, Masaryk University, Kotlrsk 2, CZ-611 37 Brno, Czech Republic
Received 12 January 2007; accepted 16 February 2007

Abstract
We show that it is possible to formulate the most general first-class gauge algebra of the operator formalism by only using BRST-invariant constraints. In particular, we extend a previous construction for
irreducible gauge algebras to the reducible case. The gauge algebra induces two nilpotent, Grassmann-odd,
mutually anti-commuting BRST operators that bear structural similarities with BRST/anti-BRST theories
but with shifted ghost number assignments. In both cases we show how the extended BRST algebra can be
encoded into an operator master equation. A unitarizing Hamiltonian that respects the two BRST symmetries is constructed with the help of a gauge-fixing boson. Abelian reducible theories are shown explicitly
in full detail, while non-Abelian theories are worked out for the lowest reducibility stages and ghost momentum ranks.
PACS: 04.60.Ds; 11.10.-z; 11.15.-q
Keywords: BFV-BRST quantization; Extended BRST symmetry; Reducible gauge algebra; Antibracket
1. Introduction
In the quantization of a classical system with first-class constraints T ,
{T , T }PB = U T ,
E-mail addresses: batalin@lpi.ru (I.A. Batalin), bering@physics.muni.cz (K. Bering).

doi:10.1016/j.nuclphysb.2007.02.013
(1.1)
I.A. Batalin, K. Bering / Nuclear Physics B 771 [PM] (2007) 190233
191
it is well known that there may appear quantum anomalies in the expression for the operator
commutator [T , T ] at quadratic (or higher) orders in h . A useful approach to circumvent this
obstacle is to replace the initial constraints T with BRST-invariant constraints T . In practice,
it turns out that the BRST invariance of T protects the commutator [T , T ] against anomalies.
Our goal is to investigate this fact systematically in a general setting. For irreducible first class
constraint such an investigation is undertaken in Ref. [1]. Here we will develop that construction
and extend it to reducible gauge algebras.
The use of BRST-invariant constraints is best illustrated by a motivating example. Consider
the critical open bosonic string theory in D = 26 dimensions with an intercept/normal-ordering
constant a = 1, cf. Refs. [2,3] and Appendix A. The matter Virasoro generators

1
0
Tm =
(1.2)
:mn n : ha
m
2
n
form a Virasoro algebra with non-zero central extension
0
[Tm , Tn ] = h (m n)Tm+n + h 2 Am m+n
,
Am
Dm(m2
1)
12
(1.3)
+ 2ma.
(1.4)
(c)
The remedy is well known: When one adds the appropriate ghost contributions Tm to the Virasoro constraints,

(m n):bm+n cn :,
Tm = Tm + Tm(c) , Tm(c) =
(1.5)
n
the new constraints Tm obey a Virasoro algebra with no central extension, also known as a Witt
algebra, precisely for D = 26 and a = 1,
0
[Tm , Tn ] = h (m n)Tm+n + h 2 Am m+n
,
(1.6)
+ (24a + 2 D)m
(1.7)
= 0.
12
How does the BRST formulation [4,5] fit into this? Surprisingly, the new anomaly-free constraints Tm happen to be BRST-superpartners of the ghosts momenta bm ,
26)m3
Am
(D
Tm =
1
[bm , 0 ],
h
(1.8)
where 0 denotes the BRST operator. As a simple consequence of Eq. (1.8), the new constraints
Tm are BRST-invariant,

1
1
bm , [0 , 0 ] = 0.
[bm , 0 ], 0 =
(1.9)
2h
h
Here we argue that Eq. (1.8) should be viewed as the rule rather than the exception. First of all,
Eq. (1.8) respects ghost number conservation, since 0 , Tm and bm carry ghost number 1, 0
and 1, respectively. In general, ghost number conservation severely restricts what can happen.
Secondly, it follows from Eq. (1.8), using quite broad assumptions, that the [Tm , Tn ] commutator
cannot develop a quantum anomaly, which is a strong hint. For instances, guided by ghost number
conservation, it is reasonable to expect that the commutator [Tm , bn ] between the BRST-invariant
constraints Tm and the ghost momenta bn again is proportional to the ghost momenta. In detail,
[Tm , 0 ] =
192
let us assume that

[Tm , bn ] = h
Umn p bp
(1.10)
for some BRST-invariant structure functions Umn p . Then the proof goes as follows:

1
1
Umn p [bp , 0 ]
Tm , [bn , 0 ] = [Tm , bn ], 0 =
h
h
p

p
Umn Tp ,
= h
[Tm , Tn ] =
(1.11)
i.e. there are no terms of order O(h 2 ). Therefore the Eqs. (1.8) and (1.10) imply that the [Tm , Tn ]
commutator is anomaly-free. Such an assurance is a rare commodity in a full-fledged quantum
theory, and this is why we would like to systematically seek for relationships
of the form (1.8).

(In string theory much more is known: The ghost momentum b(z) = m bm zm2 is a primary
field of conformal weight 2, which means that [Tm , bn ] = h (m n)bm+n , thereby confirming
p
the Ansatz (1.10) with Umn p = (m n)n+m . However, we emphasize the versatility of the
argument.) Later in the construction a new generation of ghosts C A is introduced, and the BRSTdoublets
TA = {Tm ; bm }
(1.12)
are turned into twice as many first-class constraints for a new BRST operator called 1 =
C A TA + , cf. Eq. (3.41) below.
We stress that choosing BRST-invariant constraints T is not a miraculous cure that turns
anomalous theories into anomaly-free ones. Rather it is a useful tool in overcoming a poor initial
choice of constraint basis T for otherwise sound theories. For instance, in the case of critical
open bosonic strings, the BRST formulation requiresfor startersthat the BRST charge 0 is
nilpotent,

0 = [0 , 0 ] =
(1.13)
[Tm , Tn ] h (m n)Tm+n cm cn = h 2
Am cm cm ,
m,n
which by itself implies the familiar D = 26 and a = 1, cf. Eqs. (1.7) and (A.10). In general, we
shall simply assume as a starting point that an anomaly-free nilpotent BRST operator 0 has
been given. (For a typical physical model with infinitely many degrees of freedom, it takes a fair
amount of mathematical analysis to rigorously establish this beyond the formal level [68].)
The paper is organized as follows. In Section 2 the standard Hamiltonian BRST construction
[610] is reviewed. In Section 3 a new generation of ghosts and BRST symmetry is introduced.
An anti-BRST operator is used to infer cohomologically the existence of an S operator, which
satisfies an operator master equation [11],

S, [S, ] = (i h)[S,
],
(1.14)
cf. Eq. (3.63) below. The S operator contains the so-called tilde constraints TA , which are the
analogues of the above string ghost momenta bm , and which descend via the relation
i
TA0 = VA0 B0 TB0 + [TA0 , 0 ],
h
(1.15)
193
to the new TA constraints (1.12), cf. Eq. (4.1) below. Eq. (1.15) generalizes Eq. (1.8) and is
the heart of the construction. It guarantees that the new TA0 constraints are BRST-invariant up
to the first term on the left-hand side, which depends on ghost momenta and vanishes in the
unitary limit. In short, Section 3 presents the various pieces and definitions that go into the construction. In Section 4 we assemble all the pieces and show how to gauge-fix, i.e. construct the
unitarizing Hamiltonian. One of our main new points is that the construction shares strikingly
many similarities with BRST/anti-BRST symmetric models. In particular, it has two nilpotent,
Grassmann-odd, mutually anti-commuting BRST operators, although their ghost number assignments are different. Here the first BRST operator is a ghost-deformed version of the ordinary
BRST operator 0 , and the second BRST operator is the new BRST operator 1 , associated
with a next generation of ghosts C A , cf. Eq. (1.12). Similar to BRST/anti-BRST theories, the
gauge-fixing will depend on a gauge-fixing boson, and the unitarizing Hamiltonian will respect
both BRST symmetries.
In Section 5 the Abelian case is treated in full details. The Abelian case, which besides being
an important example in its own right, establishes at the theoretical level the existence of the
whole construction. Later, in Section 6 we discuss algebraic properties of the constraints. Section 7 briefly states our conclusions. The paper also includes five appendices: Appendix A gives
an elementary derivation of the conformal anomaly for the critical open bosonic string. Appendices B and C discuss an optional superfield and matrix formulation, respectively. The superfield
formulation explores a (perfectly consistent) ghost number deficit among the two superpartners
of new constraints TA , as already evident from Eq. (1.12). Finally, Appendix D analyzes possible
candidates for the operator master Eq. (1.14), and Appendix E reformulates certain additional
involution relations as a nilpotency condition for a BRST operator.
1.1. Operator ordering
String theory is most often formulated using elementary Fock space creation and annihilation
operators, m , bm and cm , also referred to as Wick operators. The ordering prescription is Wick
(= normal) ordering, cf. Eqs. (A.2) and (A.3); and the Hermitian conjugation is implemented
through opposite modes m m,

bm
= bm ,
cm
= cm .
m = m ,
(1.16)
However for a general theory, one does not want to make unnecessary assumptions about the index structure of the fundamental operators, and one would therefore have to write the Hermitian
conjugation explicitly, thereby producing quite lengthy formulas. To avoid this, we take from
now on the fundamental operators to be Hermitian or anti-Hermitian, and the ordering prescription to be qp-ordering, i.e. the coordinates q ordered to the left of the momenta p. (In the sectors
of the theory that is not written out explicitly, one is of course free to assume any ordering.)
For more on the Wick formalism and the string example, see Subsection 6.B of Ref. [12] and
Subsection 6.2 of Ref. [1].
1.2. Notation
Square brackets
[A, B] AB (1)A B BA,
[A, B]+ AB + (1)A B BA
(1.17)
194
denote the supercommutator and the anti-supercommutator of two operators A and B of Grassmann parities A and B . As we already saw in the Introduction, it becomes quite tedious to write
out every h , so from now on we shall let a curly bracket
1
(1.18)
[A, B]
i h
denote a normalized supercommutator. As usual, the normalized commutator (1.18) becomes a
Poisson bracket {A, B}PB of commuting functions A and B in the classical limit h 0 by the
well known correspondence principle of quantum mechanics. However, we stress that the normalized commutator (1.18) is an operator for h = 0. The notation (1.18) enables us to (1) almost
eliminate appearances of i h,
(2) provide a full-fledged quantum operator formalism and (3) at
the same time give classical expressions. Note that {A; B} separated by a semicolon ; instead
of a comma , will denote a set of two elements A and B.
As a rule of thumb, calligraphic letters are associated with the new ghost sector, cf. Section 3.1. A bar over a quantity refers to negative ghost number, while a tilde (respectively breve
) over a quantity means that it is associated with S1 (respectively S2 ), cf.
Section 3.7 (respectively 3.8).
{A, B}
2. Old BRST algebra

In this section we review the construction of a general BRST operator algebra [610,1316]
with the operator ordering prescription taken to be qp-ordering. The review is partially to fix
notation and partially to motivate later constructions. The BRST algebra will only serve as a
starting point, and we shall refer to it as the old or the ordinary system to distinguish it from
newer sectors to be added in later sections.
2.1. Old ghosts
Besides the original phase variables (q i , pj ), which almost always appear only implicitly in
formulas, the ordinary phase space consists of L + 1 stages of ordinary ghosts

C = C s s = 0, . . . , L ,
(2.1)
and their momenta
P = {Ps | s = 0, . . . , L},
with canonical commutation relations

C , C = 0,
C , P = = (1) P , C ,
(2.2)
{P , P } = 0.
(2.3)
Here the index s runs over the set {1, . . . , ms }, the subscript s is associated with stage s
{0, . . . , L}, while
the index is a shorthand for all the stages 0 , . . . , L , i.e. the index runs
over {1, . . . , L
s=0 ms }, and so forth. The Grassmann parity and ghost number of the ordinary
ghosts are

C = + 1 = (P ),
C s = s + s + 1 = (Ps ),

gh C s = s + 1 = gh(Ps ).
(2.4)
195
2.2. Old BRST operator

The ordinary BRST operator 0 = 0 (q, p; C, P ) is Grassmann-odd, nilpotent and with
ghost number 1,
{0 , 0 } = 0,
(0 ) = 1,
gh(0 ) = 1.
(2.5)
It is a power series expansion in the ordinary ghosts C and P , and ordered according to the
C P -ordering prescription,
0 =
L
C s Ts +
s=0
1
2
1
4
1
C s C r Ur s r+s Pr+s (1)s +r+s +r
2
r,s0
r+sL
C r+s+2 Ur+s+2 s r Pr Ps (1)r +r
r,s0
r+s+2L

C s C r Ur s t u Pu Pt (1)s +s+u +u + O C 3 P , C P 3 ,
(2.6)
0r,s,t,uL
r+s=t+u+1
where U... ... = U... ... (q, p) denote structure functions of the original phase space. The 0
operator starts with the original first-class constraints T0 = T0 (q, p), and in the reducible case
one also introduces higher-stage constraints, which depend linearly on the ghost momenta,
Ts+1 Zs+1 s Ps (1)s +s ,
s {0, . . . , L 1}.
Note that the higher-stage constraints T do not necessarily commute with the
Grassmann parity and ghost number are
(T ) = ,
(Ts ) = s + s,
gh(Ts ) = s,
(2.7)
C
ghosts. The
(2.8)
which are opposite of the corresponding ordinary C and P ghosts. The reducible structure functions Zs s1 = Zs s1 (q, p) have Grassmann parity and ghost number given by

Zs s1 = s + s1 ,
(2.9)
gh Zs s1 = 0.
To be systematic, let m1 = 1 + 0 denote half the number of original phase variables (q, p),
i.e. the original phase space consists effectively of 21 physical degrees of freedom, 0 irreducible gauge-generating constraints, and 0 irreducible gauge-fixing constraints. The logic
behind reducible gauge algebras is that due to requirements of locality and symmetries, such as,
e.g. Lorentz symmetry, it is impossible to pick an irreducible set of constraints. At each stage s
(except the last stage) one keeps overshooting the number of remaining gauge symmetries. In
detail, the sth stage consists of ms = s + s+1 ghosts, which corresponds to s independent
gauge symmetries, and s+1 redundant symmetries, which in turn become gauge symmetries for
the next stage s + 1. If L < is finite, the process stops at the Lth stage with L+1 = 0.
0 = rank(T0 ) > 0,
s =
L
(1)rs mr =
r=s
(2.10)

0
for s L + 1,
rank(Zs s1 ) > 0 otherwise.
(2.11)
196
This leads to a rank condition on the 0 operator. It should contain as many non-trivial constraints as possible, i.e., a quarter of the total number of non-physical degrees of freedom,

s=0
s =
s=0
r=0
r=0

1
ms =
m2r =
m2r+1 .
2
(2.12)
Equivalently, the antibracket (,)0 should have half rank in the non-physical sector. Here the
(normalized, operator-valued) antibracket1 is defined as [1720]
1
1
(2.13)
A, {0 , B} + {A, 0 }, B ,
2
2
where A and B are arbitrary operators. We remark that this antibracket always satisfies the pertinent Jacobi identity modulo BRST-exact terms, and that the Jacobi identity holds strongly within
a Dirac subalgebra [20], which is by definition a maximal subalgebra that is (1) Abelian with
respect to the commutator and (2) stabile under the antibracket (2.13).
The rank R of a theory is defined as the highest power of ghost momenta P in 0 . Some of
the consequences of the 0 -nilpotency (2.5) are
(A, B)0
Z1 0 T0 = 0,
(2.14)
{T0 , T0 } = U0 0 T0 ,

T0 , Zs+1 s = U0 s+1 s+1 Zs+1 s + (1)0 (s+1 +s) Zs+1 s Us 0 s

U0 s+1 s 0 T0 + O(h ) terms, if R 2 ,
0
Zs+2 s+1 Zs+1 s = Us+2 s 0 T0
s
i h
(2.15)
(2.16)
Us+2 sr r Ur sr s
r=0

+ O h 2 terms, if R 3 and s 1 .
(2.17)
One recognizes the involution relation (2.15) for the first class constraints T0 , and the reducibility relations (2.14) and (2.17). Eq. (2.14) shows that the zeroth-stage constraints T0 are reducible
if rank(Z1 0 ) = 0. One can re-express Eq. (2.17) in the s = 0 case as
1
Z2 1 Z1 0 = U2 0 0 Z0 0 0 ,
2
0
where Z0 0 is defined as
(2.18)
i h
(2.19)
U 0 (1)0 0 (0 0 ).
2 0 0
If one takes the classical limit h 0 and goes on-shell with respect to the constraints T0 , the
quantities Z0 0 0 vanish, so that the structure functions Z2 1 become left zero-eigenvectors for
the matrix Z1 0 , cf. Eq. (2.18). At the quantum level and off-shell, the quantities Z0 0 0 fulfill
Z0 0 0 T0 00
Z0 0 0 T0 = [T0 , T0 ] i hU
0 0 0 T0 = 0,
(2.20)
due to Eq. (2.15).

1 The antibracket in Ref. [1] is based on the bare commutator (1.17), while we here use the normalized commutator
here
(1.18), so that (,)there
)2 (,)0 .
(i h
197
2.3. BRST-improved Hamiltonian

We mention for completeness that the original Hamiltonian Hor = Hor (q, p), which commutes weakly with the first-class constraints
{T0 , Hor } = V0 0 T0 ,
(2.21)
is supposed to be BRST-improved,

H0 = Hor + C V P (1) + O C 2 P , C P 2 ,
(2.22)
by letting the improved Hamiltonian H0 = H0 (q, p; C, P ) depend on the ghosts

such a way that it becomes BRST-invariant,
{0 , H0 } = 0,
(H0 ) = 0,
gh(H0 ) = 0.
and P in
(2.23)
Contrary to the standard approach [610] one does not introduce non-minimal variables in the old
sector. Instead, the idea is roughly speaking to supply the gauge-generating constraints T = 0
with more gauge-generating constraints, which kill the ordinary ghost momenta P = 0, and
supply the gauge-fixing constraints = 0 with more gauge-fixing constraints, which kill the
ordinary ghosts C = 0. The complete gauge-fixing procedure will be mediated through nonminimal variables in a new ghost sector, see Section 4.2.
2.4. Anti-BRST algebra
At this point we mention an interesting possibility to include an anti-BRST operator 0 =
0 (q, p; C, P ). Although anti-BRST algebras are not an essential part of this paper (i.e. one
is free to simply let 0 = 0), they nevertheless constitute an important topic that is worthwhile
mentioning. In general, the anti-BRST operator 0 is required to satisfy
{ 0 , 0 } = 0,
{0 , 0 } = 0,
( 0 ) = 1,
gh( 0 ) = 1.
(2.24)
We caution that the pertinent anti-BRST algebra [21] will here be different from the traditional
notion of an anti-BRST algebra [2224]. Recall that in the traditional setting, the anti-BRST
algebra explores an Sp(2) duality among the FaddeevPopov ghost and antighost C C .
Roughly speaking, the traditional anti-BRST operator (which are closely related to a co-BRST
operator [25]) amounts to substitute C C in 0 , while keeping the constraints T the same.
However in our current setup, there is no FaddeevPopov antighost within the minimal framework; only a ghost momenta P . (Instead the FaddeevPopov antighost typically belongs to a
non-minimal sector.) The crucial difference is that the traditional Sp(2) doublets (C ; C ) of
FaddeevPopov ghost pairs commute [C , C ] = 0, while the canonical pairs (C ; P ) do not
commute. Also the anti-constraints T 0 in 0 will in general be different from the first-class
constraints T0 , cf. Eq. (2.38) below. In detail, the 0 operator is a power series expansion in
ordinary ghosts C and P , and ordered according to the C P -ordering prescription,
0 =
L

s=0
1
T s Ps (1)s +s +
C r+s U r+s s r Pr Ps (1)r +r
2
1
2

r,s0
r+s+2L
r,s0
r+sL
C s C r U r s r+s+2 Pr+s+2 (1)s +r+s+2 +r
198
1
4
C s C r U r s t u Pu Pt (1)s +s+u +u
0r,s,t,uL
r+s+1=t+u

3
+ O C P , CP .
(2.25)
T 0
= T 0 (q, p),
The 0 operator starts with the anti-constraints

and in the reducible case one
also introduces higher-stage anti-constraints, which depend linearly on the ghosts,
T s C s1 Z s1 s ,
s {1, . . . , L}.
(2.26)
The anti-constraints T carry the same Grassmann parity as the corresponding constraints T ,

T s = s + s,
gh T s = s.
T = ,
(2.27)
The reducible structure functions Z s s+1 = Z s s+1 (q, p) have Grassmann parity and ghost
number given by

gh Z s s+1 = 0.
Z s s+1 = s + s+1 ,
(2.28)
Some of the consequences of the 0 -nilpotency (2.24) are
T 0 Z 0 1 = 0,

T 0 , T 0 = T 0 U 0 0 0 ,

Z s s+1 , T 0 = Z s s+1 U s+1 s+1 0 + (1)0 (s+1 +s) U s 0 s Z s s+1

T 0 U 0 s s+1 0 + O(h ) terms, if R 2 ,
s
i h sr r
Z s s+1 Z s+1 s+2 = T 0 U 0 s s+2
Us
Ur sr s+2
2
r=0

(2.29)
(2.30)
(2.31)
(2.32)
Some of the consequences of the compatibility (2.24) of the BRST and the anti-BRST operator,
i.e. the fact that they commute, are
T 0 T0 = 0,
(2.33)

i
h
T0 , T 0 + Z 0 1 Z1 0 = T 0 U0 0 0 + U 0 0 0 T0 U 0 0 0 U0 0 0 + O h 2
2

1
= T 0 U0 0 0 + U 0 0 0 Z0 0 0 + O h 2 ,
(2.34)
2
s
i h sr r
Ur sr s
Us
Zs s1 Z s1 s + Z s s+1 Zs+1 s = T 0 U0 s s + U s s 0 T0
2
r=0
i h
2
s2

Us s2r r U r s2r s + O h 2 .
(2.35)
r=0
2.5. The S2 operator

We have up until now mostly reviewed the notions of BRST and anti-BRST symmetry in the
Hamiltonian setting. In Sections 2.52.7 we shall make a series of new observations concerning
199
the cohomology of 0 , which lead to an operator master equation, cf. Eq. (2.42) below. There is
a parallel operator master Eq. (3.63) for the main construction, to be presented in Section 3.
Normally it is assumed that the cohomology of {0 , } is trivial in sectors of non-vanishing
ghost number. Then the 0 -closeness (2.24) of 0 implies that 0 is 0 -exact, i.e. there exists
a bosonic operator S2 such that
0 = {0 , S2 },
(S2 ) = 0,
gh(S2 ) = 2.
(2.36)
In detail, the S2 operator is a power series expansion in the ordinary ghosts C and P ,
1
S2 = A0 0 P0 P0 (1)0 + A1 P1 (1)1 + O(C P ),
2
(2.37)
where A0 0 = (1)0 0 A0 0 and A1 are arbitrary operators. For instance, the complete
T 0 solution to Eq. (2.33) alone is
1
T 0 = A0 0 Z0 0 0 + A1 Z1 0 ,
2
where Z0 0 0 is defined in Eq. (2.19).
(2.38)

Similarly, one may define an S4 operator as follows. First note that the antibracket
(S2 , S2 )0 of S2 with itself can be written as the commutator of S2 and 0 ,

(S2 , S2 )0 = S2 , {0 , S2 } = {S2 , 0 }.
(2.39)
The antibracket (S2 , S2 )0 is 0 -closed, as the following calculation shows,

0 , (S2 , S2 )0 = 0 , {S2 , 0 } = {0 , S2 }, 0 = { 0 , 0 } = 0,
(2.40)
since 0 is 0 -closed and nilpotent, cf. Eq. (2.24). Hence the antibracket (S2 , S2 )0 is 0 exact, i.e. there exists a quantity S4 such that
1
(S2 , S2 )0 = {0 , S4 },
2
(S4 ) = 0,
gh(S4 ) = 4.
(2.41)
2.7. Operator master equation

The process of finding higher and higher cohomology relations may be automated by introducing an operator master equation

(S, S)0 S, {S, 0 } = 0 .

(2.42)
To define the operator S, one first lets a quantity S0 be the ghost operator,
S0 = G,
(2.43)
see Eq. (3.10) below. Furthermore, one may show by mathematical induction that there exist
quantities S2k for each positive integer k 1, and with quantum numbers
(S2k ) = 0,
gh(S2k ) = 2k
(2.44)
200
such that the sum

S
S2k ,
(S) = 0,
(2.45)
k=0
of indefinite ghost number, satisfies the above operator master Eq. (2.42), cf. Appendix D.
The master Eq. (2.42) is really an infinite tower of equations. The first non-trivial equation is
Eq. (2.41).
Although we shall not pursue this here, let us mention that a stronger formulation of the
anti-BRST algebra would impose rank conditions on 0 (i.e. the analogues of conditions (2.10)
(2.12) for 0 ), and the anti-BRST algebra would contain non-trivial information, which allows
for a complexification the whole BRST algebra. For starters, the dimension ms must then be even
at each stage s {0, . . . , L}. In contrast, there are no such requirements for a traditional Sp(2)
algebra.
3. New BRST algebra
In this section we present all the ingredients of the main construction.
3.1. New ghosts
The new BRST algebra depends on L + 1 stages of new ghosts

C A = C As s = 0, . . . , L ,
(3.1)
and their momenta

P A = {P As | s = 0, . . . , L},
with canonical commutation relations
A B
A
C , C = 0,
C , P B = BA = (1)B P B , C A ,
(3.2)
{P A , P B } = 0.
(3.3)
In addition there are non-minimal variables, which we will often not write explicitly in formulas
for the sake of simplicity, cf. Section 4.2. Technically, the various expansions will be kept under
control by a bi-graded mesh of integer-graded conservation laws, which (besides the usual ghost
number conservation) includes a new ghost number grading, denoted ngh. The Grassmann
parity and new ghost number of the new ghosts are

C As = As + s + 1 = (P As ),
C A = A + 1 = (P A ),

ngh C As = s + 1 = ngh(P As ).
(3.4)
In detail, each of the new ghosts consists of two superpartners

C A C0 ; C1 B ; (1) +1 ,
C As B s ; (1)s +s+1 s ,

P A P 0 ; P 1 ; B ,
P As s ; Bs .
(3.5)
+1
Note that there are two equivalent notations C0 B , C1 (1)

, P and
P 1 B for each of the superpartners. The latter notation stresses a relationship with the antifield formalism, cf. Eq. (5.26) below. The canonical commutation relations read
0

C0 , P = B , = ,
(3.6)
1

C1 , P = B , = ,
201
(3.7)
with the remaining canonical commutation relations being zero. There are twice as many new
ghosts (C A , P A ) as old ghosts (C , P ). In other words, the index As corresponds to two copies
of index s , s {0, . . . , L}. Moreover, the index A is a shorthand for all stages A0 , . . . , AL . The
Grassmann parity and the old ghost number are shifted among the ghost superpartners as follows:

C0s = s + s + 1 = P 0s ,
C0 = + 1 = P 0 ,

C1 = = P 1 ,
C1s = s + s = P 1s ,

gh C0s = s + 1 = gh P 0s ,

gh C1s = s + 2 = gh P 1s ,

B = + 1 = ( ),
B s = s + s + 1 = (s ),

B = = ,
Bs = s + s = s ,

gh B s = s + 1 = gh(s ),

gh s = s + 2 = gh Bs .
(3.8)
The shifts (3.8) in Grassmann and ghost statistics will play a crucial role in the following. It
is compelling that the 2 2 ghost superpartners (3.5) can be incorporated into two N = 1 superfields of definite Grassmann and ghost statistics, where the offsets in statistics have been
compensated by the appearance of a -parameter; see Appendix B for details.
3.2. Ghost operators
The old (respectively new) ghost number gh(A) (respectively ngh(A)) of an arbitrary
operator A may be implemented through a corresponding ghost operator G (respectively G) as
follows2 :
{G, A} = gh(A)A,
{G, A} = ngh(A)A.
(3.9)
Here the ghost operators read

G GC + G
GC
1 1 s
1 s
Ps , C1 + = GC + G +
, Bs + ,
2
2
L
s=0
s=0
L

1
(s + 1) Ps , C s + ,
2
(3.10)
(3.11)
s=0
L
L

1
(s + 1) P As , C As + =
(s + 1)P As C As
2
s=0
s=0
L

1
=
(s + 1) s , Bs + s , B s + .
2
(3.12)
s=0
2 The new ghost number should not be confused with the notion of new ghost number in Sp(2) theories [24], although
the underlying motivation is basically the same: namely, to tame the ghost expansions. Ref. [1] calls the new ghost
number for degree: degthere nghhere .
202
The anti-supercommutators (1.17) in the above formulas (3.10)(3.12) ensure the Hermiticity of
the ghost operators. Whereas the Eqs. (3.11) and (3.12) contain no surprises, note that the last
term in Eq. (3.10) implements a crucial additional shift in the ghost number assignments for the
antifields B and , cf. the last line of Eq. (3.8).
3.3. Improved BRST operator
The next step is to improve the old BRST charge 0 = 0 (q, p; C, P ) so that it also probes
This is needed for covariance of the theory under (C, P)-dependent
the new ghost sector (C, P).
unitary transformations. In detail, one introduces an improved charge = (q, p; C, P ; C, P)
as a C P-ordered power series expansion in the new ghosts C and P. First of all, should meet
the boundary condition
= 0 + O(C P).
(3.13)
Moreover, the quantum numbers of should be the same as for 0 . Therefore should satisfy
{, } = 0,
() = 1,
gh() = 1,
In more detail, the operator has the

= 0 +
L
ngh() = 0.
(3.14)
form3
C As VAs Bs P Bs (1)Bs +s
s=0
1
+
2
+
1
2
1
4
C Bs C Ar VAr Bs Cr+s+1 P Cr+s+1 (1)Bs +Cr+s+1 +r+1
r,s0
r+s+1L
C Cr+s+1 VCr+s+1 Bs Ar P Ar P Bs (1)Ar +r
r,s0
r+s+1L

C P 3
C Bs C Ar VAr Bs Ct Du P Du P Ct (1)Bs +s+Du +u + O C 3 P,
0r,s,t,uL
r+s=t+u
= 0 + B (1)
+1
0
V0
1
V0
V1
V1

(1)
C P 2 ,
+ O C 2 P,
+1
(1)
B
(3.15)
where VA... B... = VA... B... (q, p; C, P ) denote structure functions of the old phase space. Finally,
should contain as many non-trivial constraints as possible, i.e. a quarter of the total number of
non-physical phase variables. One implements this by demanding that the southwestern quadrant
1
V0 (i.e. the block below the diagonal) of the matrices VA B is invertible,
1
rank V0 = m
L
ms .
(3.16)
s=0
3 Ref. [1] uses a different notation for the various BRST operators. In general, there here , there here ,
0
there here and 1there 1here .
203
Equivalently, the antibracket (,) should have maximal rank in the BB -sector, because

1
V0 = (1) +1 B , B ,
(3.17)
cf. Eq. (2.13). (Again, we have for simplicity notationally suppressed that is supposed to have
quadratic dependence on half the non-minimal variables to allow for gauge-fixing, cf. Eq. (4.7)
below.) Some of the consequences of the -nilpotency (3.14) are
{0 , 0 } = 0,
(3.18)
s1
i h
VAs Bs , 0 (1)Bs +s = VAs Cs VCs Bs +
VAs Dsr1 Cr VCr Dsr1 Bs
2
r=0

+ O h 2 terms, if s 2 .
(3.19)
3.4. Improved anti-BRST operator

There is also an improved version
= 0 + O(C P),
(3.20)
of the anti-BRST operator 0 such that

}
= 0,
{,
= 1,
()
= 1,
gh()
= 0.
ngh()
(3.21)
Perhaps surprisingly, one does not demand that the improved BRST and anti-BRST operators
commute. Instead the normalized commutator of the two improved charges is assumed to be
equal to the new ghost operator G,
= G.
{, }
(3.22)
In this way, the charge becomes a homotopy operator for the BRST complex of . As a result
the cohomology of {, } is trivial in sectors of non-zero new ghost number. Proof : If A is a
-closed operator, i.e. {, A} = 0, and if A has non-zero new ghost number, it is possible to
rewrite A as an -exact operator,

A = g 1 , {,
A} = , {,
g 1 A} ,
A = g 1 {G, A} = g 1 {, },
(3.23)
}.
because the operator g {G, } commutes with both {, } and {,
On the other hand, since ultimately one requires that the physical model in question is invariant, the homotopy Eq. (3.22) implies that one can no longer maintain the anti-BRST
symmetry of the model. In this sense, the anti-BRST algebra plays here only a secondary rle.
Similarly, to be consistent, we will not require that the BRST-improved Hamiltonian obeys the
anti-BRST symmetry, cf. Section 3.6 below.
3.5. New BRST operator 1
The operator has no net new ghost charge, ngh() = 0, and is merely a deformation of the
old BRST operator 0 . In general, it is not adequate to control expansions in the new ghosts. To
this end, one introduces a new BRST generator 1 that is charged with respect to the new ghost
has properties
number. In detail, the new BRST charge 1 = 1 (q, p; C, P ; C, P)
{1 , 1 } = 0,
(3.24)
204
(1 ) = 1,
gh(1 ) = 1,
ngh(1 ) = 1.
(3.25)
It should also respect the symmetries of the improved charge , i.e. 1 should be -closed,
{, 1 } = 0.
(3.26)
The new BRST operator 1 is a C P-ordered

power series expansion in the new ghosts C and P.
1 =
L
C As TAs +
s=0
r,s0
r+sL
1
2
1 Bs Ar
C C UAr Bs Cr+s P Cr+s (1)Bs +Cr+s +r
2
C Cr+s+2 UCr+s+2 Bs Ar P Ar P Bs (1)Ar +r
r,s0
r+s+2L
1
4

C P 3
C Bs C Ar UAr Bs Ct Du P Du P Ct (1)Bs +s+Du +u + O C 3 P,
0r,s,t,uL
r+s=t+u+1
L

C P 2 .
B s Ts s Xs + O C 2 P,
(3.27)
s=0
It starts with the new constraints TA0 = TA0 (q, p; C, P ), and in the reducible case one also introduces higher-stage constraints
TAs+1 ZAs+1 Bs P Bs (1)Bs +s ,
s {0, . . . , L 1}.
(3.28)
The Grassmann parity and new ghost number are

(TA ) = A ,
(TAs ) = As + s,
ngh(TAs ) = s.
Bs1 (q, p; C, P )
The reducible structure functions ZAs

= ZAs
new ghost number given by

ZAs Bs1 = As + Bs1 ,
ngh ZAs Bs1 = 0.
Bs1
(3.29)
have Grassmann parity and
(3.30)
Altogether, the construction is governed by the two mutually commuting Grassmann-odd nilpotent BRST operators and 1 , which form an Sp(2)-like algebra,
{, } = 0,
{, 1 } = 0,
{1 , 1 } = 0,
(3.31)
cf. Eqs. (3.14), (3.24) and (3.26). Some of the consequences of the -closeness condition (3.26)
for 1 read
{TA0 , 0 } = VA0 B0 TB0 ,

ZAs+1 Bs , 0 (1)Bs +s = ZAs+1 Cs VCs Bs + VAs+1 Cs+1 ZCs+1 Bs VAs+1 Bs C0 TC0
(3.32)
i h
UAs+1 Dsr1 Cr VCr Dsr1 Bs
2
s1
r=0
s
i h
VAs+1 Dsr Cr UCr Dsr Bs
2
r=0
2

+ O h terms, if R 3 and s 1 ,
+
(3.33)
205
1

UA0 B0 C0 , 0 (1)C0 = TA0 , VB0 C0 + UA0 B0 D0 VD0 C0 (1)B0 VA0 D0 UD0 B0 C0
2
1
D 0 E0
VA0 B0
ZE0 D0 C0 (1)A0 B0 (A0 B0 ),
(3.34)
4

UA0 Bs Cs , 0 (1)Cs +s = TA0 , VBs Cs + UA0 Bs Ds VDs Cs
+ (1)A0 (Bs +s+1) VBs Ds UDs A0 Cs

VA0 Bs Cs D0 TD0 + O(h ) terms, if R 2 ,
s = 0, (3.35)
where ZA0 B0 C0 is defined as

i h
(3.36)
UA0 B0 C0 (1)A0 B0 (A0 B0 ).
2
Here R continues to denote the rank of the theory. In particular, Eq. (3.32) shows that the new
zeroth-stage constraints TA0 are on-shell BRST-invariant. Some of the consequences of the 1 nilpotency (3.24) are
C
ZA0 B0 C0 TA0 B00
ZA1 B0 TB0 = 0,
(3.37)
{TA0 , TB0 } = UA0 B0 TC0 ,

TA0 , ZBs+1 Cs = UA0 Bs+1 Ds+1 ZDs+1 Cs + (1)A0 (Bs +s+1) ZBs+1 Ds UDs A0 Cs

UA0 Bs+1 Cs D0 TD0 + O(h ) terms, if R 2 ,
C0
ZAs+2 Cs+1 ZCs+1 Bs = UAs+2 Bs C0 TC0
s
i h
(3.38)
(3.39)
UAs+2 Dsr Cr UCr Dsr Bs
r=0

(3.40)
The analogy with the old sector Eqs. (2.14)(2.17) is evident. If the original theory is of rank R,
one may choose and 1 to have at most R powers of new ghost momenta P A . In the case of
a rank R = 1 theory, some of the consequences of Eq. (3.31) can neatly be recast as a nilpotency
condition for a matrix A B with operator-valued entries, cf. Appendix C.
The new constraints TAs consists of two superpartners4

TAs Ts ; (1)s +s Xs ,
TA T0 ; T1 T ; (1) X ,
(3.41)
with boundary conditions
T0 = 0 0 T0 + O(C P),
(3.42)
where 0 0 = 0 0 (q, p) is an invertible matrix. The zero-stage constraints T00 T0 are the
new BRST-invariant constraints that one is seeking for. In fact, the zero-stage components T00
and T10 are the analogues of the BRST-invariant Virasoro constraints Tm and the string ghost
momenta bm mentioned in the Introduction. The Grassmann parity and the old ghost number are
shifted among the two superpartner constraints

T0 = = T1 + 1,
T0s = s + s = T1s + 1,

gh T0s = s = gh T1s + 1,
4 Ref. [1] uses a different notation: X there (1)0 +1 X here .
206
(T ) = = (X ) + 1,
(Ts ) = s + s = (Xs ) + 1,
(3.43)
gh(Ts ) = s = gh(Xs ) + 1.
Note that the new constraints TA have non-positive ghost numbers gh(TA ) 0 and ngh(TA ) 0,
and they vanish if the old constraints T and the ghost momenta P and P A are put to zero,
TA = O(T , P , P).
(3.44)
In other words, if one takes the unitary limit, where the new ghost momenta P A 0 vanish, then
the new zero-stage constraints TA0 0 imply that both the old zero-stage constraints T 0
and the old zero-stage ghost momenta P0 0 vanish, at least within the naive path integral
formulation, cf. Eq. (4.25) below.
3.6. BRST-improved Hamiltonian
The old BRST-improved Hamiltonian H0 = H0 (q, p; C, P ) is once again BRST-improved
(this time with respect to the new BRST structures),
H = H0 + O(C P),
(3.45)
depend on the new ghosts C A

by letting the improved Hamiltonian H = H(q, p; C, P ; C, P)
and PB ) in such a way that it becomes BRST-invariant with respect to both and 1 ,
{, H} = 0,
{1 , H} = 0,
(H) = 0,
ngh(H) = 0,
gh(H) = 0. (3.46)
The reason why H can be chosen to commute simultaneously with both BRST operators and
1 is that and 1 , after all, convey the same BRST symmetry, originally encoded in the 0
operator.
It is useful to think of the new BRST operators and 1 as analogues of the old BRST/antiBRST operators 0 and 0 , in particular because of the Sp(2)-like algebra (3.31) although
the ghost number assignments are different, cf. Table 1. We shall now widen this analogy by
introducing a counterpart S1 of the old S2 of Section 2.5. To this end, note that 1 is -closed
(3.26) and has non-zero new ghost number. Hence, it must be -exact, cf. Eq. (3.22), i.e. there
Table 1
A dictionary between notions in the old and the new sector
Notion
Sector
Ghosts
Ghost momenta
Ghost operator
Ghost number
1st BRST operator
2nd BRST operator
Homotopy operator
Operator master equation
Old
C
P
G = S0
gh = {G, }
0
0 = {0 , S2 }
(Not introduced)
S2k
S=
k=0 S2k
{S, {S, 0 }} = 0
New
C A {B ; (1) +1 }
P A { ; B }
G = S0
ngh = {G, }
1 = {, S1 }
Sk
S=
k=0 Sk
{S, {S, }} = {S, }
207
exists a quantity S1 with quantum numbers

(S1 ) = 0,
ngh(S1 ) = 1,
gh(S1 ) = 0
(3.47)
such that
1 = {, S1 }.
(3.48)
In detail, one may expand S1 as follows:

S1 =
L
C As TAs +
s=0
1
2
1 Bs Ar
C C UAr Bs Cr+s P Cr+s (1)Bs +Cr+s +r
2
r,s0
r+sL

C 2 P 2 , C P 3
C Cr+s+2 UCr+s+2 Bs Ar P Ar P Bs (1)Ar +r + O C 3 P,
r,s0
r+s+2L
L

C P 2 .
B s Xs s Y s + O C 2 P,
(3.49)
s=0
It starts with the so-called tilde constraints TA0 = TA0 (q, p; C, P ), and in the reducible case one
also introduces higher-stage tilde constraints
s = 0, . . . , L 1.
(3.50)
or equivalently,
The TA constraints carry the same Grassmann parity as the new ghosts C or P,
the opposite Grassmann parity of TA ,
(TA ) = A + 1,
(TAs ) = As + s + 1,
ngh(TAs ) = s.
(3.51)
The reducible structure functions ZAs Bs1 = ZAs Bs1 (q, p; C, P ) have Grassmann parity and
new ghost number given by

ngh ZAs Bs1 = 0.
ZAs Bs1 = As + Bs1 + 1,
(3.52)
The tilde constraints consist of two superpartners

TAs Xs ; (1)s +s Y s .
TA T0 ; T1 X ; (1) Y ,
(3.53)
To lowest order, the zeroth-stage tilde constraints TA0 {T00 ; T10 } {X0 ; (1)0 Y 0 } are just
linear combinations of the old ghost momenta {P0 ; P1 },
X0 = X0 0 P0 (1)0 +1 + O(C P),
Y 0 = Y0 1 P1 (1)1 + O(C P).
(3.54)
The Grassmann parity and the old ghost number are shifted among the two superpartner constraints

T0s + 1 = s + s = T1s ,
T0 + 1 = = T1 ,

gh T0s = s 1 = gh T1s + 1,
(X ) + 1 = = (Y ),
(Xs ) + 1 = s + s = (Y s ),
gh(Xs ) = s 1 = gh(Y s ) + 1.
(3.55)
208
The tilde constraints TA have non-positive ghost numbers gh(TA ) < 0 and ngh(TA ) 0, and they
vanish if the ghost momenta P and P A are put to zero,
TA = O(P , P).
(3.56)
Eq. (3.56) shows that tilde constraints TA indeed can be viewed as bona-fide constraints in the
unitary limit, where the new ghost momenta P A 0 vanish, cf. Eq. (4.25) below. The vanishing
of the old zero-stage ghost momenta P0 0 is enforced via the superpartners T10 of the new
zero-stage constraints TA0 {T00 ; T10 }, cf. Eq. (3.41).
Similarly, one may define a S2 quantity as follows. First note that the antibracket (S1 , S1 )
of S1 with itself can be written as a commutator of S1 and 1 ,

(S1 , S1 ) = S1 , {, S1 } = {S1 , 1 }.
(3.57)
The antibracket (S1 , S1 ) is -closed, as the following calculation shows,

, (S1 , S1 ) = , {S1 , 1 } = {, S1 }, 1 = {1 , 1 } = 0,
(3.58)
since 1 is -closed and nilpotent, cf. Eqs. (3.24) and (3.26). Hence there exists a quantity S2
with quantum numbers
(S2 ) = 0,
ngh(S2 ) = 2,
gh(S2 ) = 0
(3.59)
such that
(S1 , S1 ) = {S2 , }.
(3.60)
One may expand S2 as follows:

S2 =
L

s=1
1
C As TAs + C B0 C A0 UA0 B0 (1)B0
2
1
2
1
2
C Bs C Ar UAr Bs Cr+s1 P Cr+s1 (1)Bs +Cr+s1 +r1
r,s0
0r+s1L

C 2 P 2 , C P 3 .
C Cr+s+3 UCr+s+3 Bs Ar P Ar P Bs (1)Ar +r + O C 3 P,
r,s0
r+s+3L
(3.61)
It starts with the breve constraints TA1 = TA1 (q, p; C, P ), and in the L 2 case one also introduces higher-stage breve constraints
s = 0, . . . , L 2.
(3.62)
209
3.9. Operator master equation

The process of finding higher and higher cohomology relations may be automated by introducing an operator master equation [11]

(S, S) {S, }, S = {, S}.

(3.63)
The operator master equation (3.63) expresses that {S, } acts as an idempotent on , cf.
Ref. [26]. To define the operator S, one first let a quantity S0 be the new ghost operator,
S0 = G.
(3.64)
Furthermore, one may show by mathematical induction that there exist quantities Sk for each
positive integer k 1, and with quantum numbers
(Sk ) = 0,
ngh(Sk ) = k,
gh(Sk ) = 0
(3.65)
such that the sum

S
Sk ,
(S) = 0,
gh(S) = 0,
(3.66)
k=0
of indefinite new ghost number, satisfies the above operator master equation (3.63), cf.
Appendix D. The master equation (3.63) is really an infinite tower of equations
k1

(Sj , Skj ) = (k 1){Sk , },
k {0, 1, 2, . . .}.
(3.67)
j =1
The first non-trivial equation (with k = 2) is just Eq. (3.60). The maximal arbitrariness of solutions S to the operator master equation (3.63) with the boundary conditions (3.48) and (3.64) is
given by canonical transformations of the form [19]
i
S
= e h {, } Se h {, } = e{{, },} S,
(3.68)
where the generating fermionic operator
k ,
( ) = 1,
gh( ) = 1,
(3.69)
k=1
has indefinite new ghost number,

(k ) = 1,
ngh(k ) = k,
gh(k ) = 1.
(3.70)
The first few components of the transformed solution (3.68) read

S0
= S0 = G,
(3.71)
S1
S2
(3.72)
= S1 {, 1 },

= S2 + {, 1 }, S1 2{, 2 },
..
.
(3.73)
210
It is instructive to mention that the new ghost operator G by itself is a trivial solution for S to the
operator master equation (3.63) and to the first boundary condition (3.64), but it fails to meet the
second boundary condition (3.48).
Proof of properties related to Eq. (3.68): It follows from the BakerCampbellHausdorff Theorem that the transformations (3.68) form a group. It is also clear that the transformations (3.68)
preserve the operator master equation (3.63) with the boundary conditions (3.48) and (3.64), so
it only remains to prove that all solutions are connected this way. Let there be given two arbi(2)
(1)
trary solutions S (1) and S (2) . For all k 1, call the difference for Xk Sk Sk . Consider
successively, for each k 1, starting with the smallest k and going up, whether the difference
Xk is -exact. If this is the case, perform a suitable transformation (3.68) of the second solution S (2) , generated by some k , such that the new difference Xk = 0 becomes identically zero.
One now proceeds by indirect reasoning: Assume that this process stops at some step k, i.e.
X1 = X2 = = Xk1 = 0, but Xk is not -exact. On the other hand, the two equations (3.48)
and (3.67) now guarantee that the difference Xk is -closed in the two cases k = 1 and k 2, respectively. Since Xk has non-vanishing new ghost number, ngh(Xk ) = k = 0, one then concludes
from cohomology considerations that Xk is -exact, cf. Eq. (3.22). This is in contradiction with
the above, and hence the process did not stop after all.
3.10. An inverse relation to 1 = {, S1 }
For any given solution S, it is possible to create an interesting primed solution S
by a canon
ically transformation (3.68) that has an -exact
fermionic generator of the form
S1 },
1 = {,
(3.74)
and where the higher generators k2 remain unspecified. Then an application of the Jacobi
S1 }} = {{, },
S1 } {,
{, S1 }} yields
identity {, {,

{, S1 } = {,
1 },
S1
= S1 {, 1 } = S1 {G, S1 } + ,
(3.75)
cf. Eqs. (3.22), (3.48) and (3.72). Eq. (3.75) can be thought of as an inverse relation to Eq. (3.48).
Whereas Eq. (3.48) gives the new TA0 constraints in terms of the tilde constraints TA0 , cf.
Eq. (4.1) below, the inverse relation (3.75) gives the primed tilde constraints TA
0 in terms of
the TA0 constraints.
Finally note that if one successively repeats the above transformation (3.74) and transforms
the primed solution S
with a fermionic generator

{,
1 } = 0,
S1
= ,
1
= ,
(3.76)
nothing happens, because of the nilpotency (3.21). So the primed solution S
is stabile in this
sense.
4. Assembling the pieces
Let us briefly summarize the construction so far. We are considering a physical gauge system,
which is ordinarily described by a BRST-improved Hamiltonian H0 , an ordinary BRST charge
0 and an ordinary ghost number gh. We introduced twice as many new ghosts C A and an
improved BRST operator , which is the old BRST charge 0 adapted to the new ghost sector.
We then collected the old constraints T0 and the old ghost momenta P0 into new constraints
211
TA0 and introduced a new BRST charge 1 = C A TA + , and a new ghost number called
ngh.
We then lifted cohomologically the new BRST operator 1 to a hierarchy of quantities
S
k=0 Sk , which satisfies an operator master equation (3.63). In this section we will stress
the main features and show how to gauge-fix the Hamiltonian.
4.1. The main relation
We now display some of the consequences of the relation 1 = {, S1 } in more detail,
cf. (3.48).
TA0 = VA0 B0 TB0 {TA0 , 0 },
(4.1)

ZAs+1 Bs = VAs+1 Cs+1 ZCs+1 Bs ZAs+1 Cs VCs Bs + ZAs+1 Bs , 0 (1)Bs +s
i h
+ VAs+1 Bs C0 TC0 (1)Bs +s +
VAs+1 Cr Dsr UDsr Cr Bs
2
s
r=0
s1
i h

UAs+1 Cr Ds1r VDs1r Cr Bs + O h 2 terms, if R 3 and s 1 ,
2
r=0
(4.2)
1
1
UA0 B0 C0 = VA0 D0 UD0 B0 C0 (1)B0 UA0 B0 D0 VD0 C0 + VA0 B0 D1 ZD1 C0
2
2

1
+ UA0 B0 C0 , 0 (1)C0 TA0 , VB0 C0 (1)B0
2
1
i h
+ VA0 B0 C0 D0 TD0 (1)C0 + VA0 B0 D0 E0 UE0 D0 C0 (1)A0 B0 (A0 B0 ).
2
4
(4.3)
From Eq. (4.1) follows that the new zeroth-stage constraints TA0 are BRST-invariant up to the
first term on the left-hand side, which depends on ghost momenta and vanishes in the unitary
limit, cf. Eq. (4.25) below. Eq. (4.1) is the heart of the construction. It may be regarded as a
generalization of Eq. (1.8) in the Introduction, where the tilde constraints TA0 generalize the
string ghost momenta bm .
4.2. Non-minimal sector
In this subsection we show how to gauge-fix the extended BRST symmetries. We shall for
simplicity concentrate on the irreducible case L = 0. The reducible case L > 0 will be discussed
elsewhere. One first introduces new non-minimal variables (P A0 , CB0 ; A0 , B0 ) with canonical
commutation relations
A
A
A
A
P 0 , CB0 = B00 ,
(4.4)
0 , B0 = B00 ,
and where the remaining canonical commutation relations are zero. (Pay attention to the perhaps
confusing but commonly used convention that P A0 is a coordinate, while the FaddeevPopov
antighost CA0 is a momentum.) The Grassmann parity and new ghost number are

P A0 = A0 + 1 = (CA0 ),
ngh P A0 = 1 = ngh(CA0 ),

(4.5)
A0 = A = (A ),
ngh A0 = 0 = ngh(A ).
0
212
The 4 new non-minimal variables (P A0 , CB0 ; A0 , B0 ) are naturally divided into 2 4 = 8

superpartner fields. The Grassmann parity and the old ghost number are shifted among the superpartners as follows:

gh P0 0 = 1 = gh C00 ,
P0 0 = 0 + 1 = C00 ,

P1 0 = 0 = C10 ,
gh P1 0 = 2 = gh C10 ,

0 0 = 0 = 00 ,
gh 0 0 = 0 = gh 00 ,

1 0 = 0 + 1 = 10 ,
gh 1 0 = 1 = gh 10 .
(4.6)
The non-minimal extensions of the two BRST operators and 1 are
= min + P1 0 C00 1 0 00 ,
(4.7)
1 = 1,min + P
(4.8)
A0
A0 .
This extension is consistent with the Sp(2)-like algebra (3.31). Similarly, the non-minimal extensions to the ghost operators G and G and the anti-BRST operator read

1 0
1

C0 , P0 0 + C10 , P1 0 + 10 , 1 0 + ,
2
2

1
A0
A0
G = Gmin CA0 , P + = Gmin CA0 P ,

2
= min P0 0 C10 .
G = Gmin
The gauge-fixed (or unitarizing) Hamiltonian

H = H + 1 , {, } = H + {1 , }
(4.9)
(4.10)
(4.11)
(4.12)
depends on a gauge boson through a -exact gauge fermion of the form

{, }.
(4.13)
Here the BRST-improved Hamiltonian H should be invariant with respect to both the BRST
operators and 1 , cf. Eq. (3.46). The quantum numbers for and are
() = 0,
( ) = 1,
ngh() = 1,
ngh( ) = 1,
gh() = 2,
gh( ) = 1.
(4.14)
A simple gauge choice is
= C10 0 0 + P 10 0 0 ,
where the FaddeevPopov matrix

rank 0 0 , T00 = m0 .
(4.15)
{0 0 , T00 }
carries maximal rank

(4.16)
Then the corresponding gauge fermion becomes

= {1 , } = C00 0 0 C10 0 0 , 0 + P 10 1 0 + , P 10 0 0 .
(4.17)
This is consistent with a gauge fermion of the form

= CA0 A0 + XA0 A0 + O C P 2 ,
(4.18)
213
where A0 {0 0 ; 1 0 } are gauge-fixing conditions with

ngh A0 = 0,
A0 = A0 ,

0 0 = 0 ,
gh 0 0 = 0,

1 0 = 0 + 1,
gh 1 0 = 1.
(4.19)
The superpartners 1 0 are of the form

1 0 = 0 0 , 0 = 0 0 , T0 C 0 + O C 2 P = O(C).
(4.20)
They have positive ghost number, and they vanish if and only if the old ghosts C 0 are put to
zero.
The XA0 {X00 ; X10 } are superpartners to the new constraints TA0 , i.e. one may introduce
a next generation of superpartners TA0 {TA0 ; XA0 } {T00 , T10 ; X00 , X10 }. (This leads naturally to a third generation of BRST operators and ghost numbers. Here we count the old BRST
operator 0 and the old ghost number gh as belonging to the first generation, and the new
BRST operator 1 and the new ghost number ngh as the second generation. Therefore in this
terminology the next generation is the third generation. In principle, one may introduce infinitely
many generations.) One has
(XA0 ) = A0 + 1,
ngh(XA0 ) = 1.
(4.21)
In fact,
0
X00 = V10 0 P 00 (1)0 ,
X10
= P 10 ,
(4.22)
(4.23)
where V10 0 0 is the invertible matrix from Eq. (3.16). The constraints XA0 have negative ghost
numbers gh(XA0 ) < 0 and ngh(XA0 ) < 0, and they vanish if and only if the new ghost momenta
P A0 are put to zero,
XA0 = O(P).
(4.24)
Let us mention for completeness that the unitary limit is obtained through the substitution
1
(4.25)
A0 A0 ,

and letting the parameter 0. In the naive path integral, the kinetic terms for the non-original
variables
CA0 CA0 ,
A0 A0 ,
P A0 CA0 + CA0 P A0 + A0 A0 0
(4.26)
vanish in this limit. The unitarizing Hamiltonian (4.12) becomes

H = H + A0 A0 + TA0 C A0 , XB0 B0 + XA0 P A0 + CA0 A0 , TB0 C B0

+ higher order terms.
(4.27)
The second (respectively third) term on the right-hand side is the delta-function term for the
gauge-fixing (respectively gauge-generating) constraints A0 (respectively TA0 ), where the integration is over the A0 (respectively A0 ) variables. The fourth term is the delta-function term
214
for the next generation of superpartners XA0 to the gauge-generating constraints, where the integration is now over the P A0 variables. (Recall that XA0 are the new ghost momenta P A0 in
disguise, cf. Eqs. (4.22) and (4.23). This fact implies that the first kinetic term (4.26) drops out.
The last two kinetic terms (4.26) are suppressed by a factor of , cf. Eq. (4.25).) And finally, the
fifth term in Eq. (4.27) is the FaddeevPopov determinant term.
5. Abelian case
In this section we provide details for the construction in the Abelian reducible case. Recall
that one may in principle Abelianize any BRST operator 0 by unitary similarity transformations
(although locality and symmetries, such as, e.g. Lorentz symmetry, may be lost in the process).
Therefore, from a theoretical perspective, it is enough to consider the Abelian case.
5.1. Old BRST operator 0
The reducible Abelian Ansatz for 0 is defined as
0 = C T = C 0 T0 +
L1
C s+1 Zs+1 s Ps (1)s +s .
(5.1)
s=0
The 0 nilpotency (2.5) is equivalent to the following set of equations:

Z1 0 T0 = 0,
Zs+2
s+1
Zs+1
(5.2)
s
= 0,
(5.3)
{T0 , T0 } = 0,

T0 , Zs+1 s = 0,
(5.4)
(s+1 +s)(s +s+1)
Zs+1 , Zs+1
(1)

Zr+1 r , Zs+1 s = 0, r = s.
(5.5)
= (1)
(s +s)(s +s)
(s s ),
(5.6)
(5.7)
The 0 nilpotency (2.5) does not guarantee by itself that all the structure functions Zs+1 s
commute, cf. Eq. (5.6), i.e. there could be a small non-Abelian remnant left in the Abelian
Table 2
The rectangular matrix Zs+1 s subjected to the ultimate Abelianization consists entirely of blocks of zero-matrices and
unit-matrices
1
2
.
..
L1
L
0
0
0
L2
L1
1
0
0
0
1
0
..
.
0
0
1
0
0
215
Ansatz (5.1), despite the name. Nevertheless, it is always possible to ensure that all the structure functions Zs+1 s commute

Zr+1 r , Zs+1 s = 0,
(5.8)
with the help of a rotation that preserves the Ansatz (5.1), cf. Table 2.
5.2. Anti-BRST operator 0
Similarly, the anti-BRST operator 0 is in the reducible Abelian case given as
0 = T P (1) = T 0 P0 (1)0 +
L
C s1 Z s1 s Ps (1)s +s .
(5.9)
s=1
The 0 nilpotency (2.24) is equivalent to the following set of equations:

T 0 Z 0 1 = 0,
(5.10)
= 0,
Z s
Z s+1

T 0 , T 0 = 0,

T 0 , Z s s+1 = 0,
(5.11)
s+1
s+2
(5.12)
+s)
(1)(s +s+1)(s+1
Z s s+1 , Z s s+1 = (1)(s +s)(s +s) (s s ),

Z r r+1 , Z s s+1 = 0, r = s.
(5.13)
(5.14)
(5.15)
The compatibility (2.24) of the BRST and the anti-BRST operator, i.e. the fact that they commute,
yields
T 0 T0 = 0,
(5.16)
Zs+1
Zs
Z s1 + Z s

T0 , T 0 + Z 0 1 Z1 0 = 0,

T0 , Z s s+1 = 0,

Zs+1 s , T 0 = 0,

Zr+1 r , Z s s+1 = 0.
s1
s+1
= 0,
(5.17)
(5.18)
(5.19)
(5.20)
(5.21)
In view of the non-Abelian remnant in Eqs. (5.6) and (5.14), it is a small miracle that the structure
functions Zr+1 r and Z s s+1 commute, cf. Eq. (5.21). The complete T 0 solution to Eq. (5.16)
alone is
T 0 = A0 0 T0 + A1 Z1 0 ,
(5.22)
where A0 0 = (1)0 0 A0 0 and A1 are arbitrary operators, cf. Eq. (2.38). Moreover, one
may show that there exists an Abelian maximal rank solution for T 0 and Z s s+1 to all of the
Eqs. (5.10)(5.21) if all the integers ms are even, where s {0, . . . , L}.
216
5.3. Improved BRST operators and
The new ghost sector is Abelianized with (C, P)-dependent

unitary transformations. The
Abelian Anstze for the improved BRST and anti-BRST operators are
= 0 +
L
C1s P 0s = 0
s=0
L

(1)s +s s s ,
(5.23)
s=0
L
L

s 1
= 0
(s + 1)C0 Ps = 0
(s + 1)B s Bs ,
s=0
(5.24)
s=0
respectively. The two improved charges and are nilpotent, and their normalized commutator
yields the new ghost operator G,
=
{, }
L
L

(s + 1) Bs s (1)s +s s B s =
(s + 1)P As C As = G, (5.25)
s=0
s=0
cf. Eqs. (3.12) and (3.22). The antibracket

B , B =
is non-degenerated in the
BB -sector,
(5.26)
as it should be, cf. Eq. (3.17).

The Abelian Ansatz for S1 is
S1 = C A TA = 0 Y 0 +
L
B s Xs
s=0
= 0 Y0 1 P1 (1)1 +1 + P0 B 0 +
L
B s Zs s1 Bs1 ,
(5.27)
s=1
where the constraints TA {X ; (1) Y } are on the form

Y 0 = Y0 1 P1 (1)1 ,
X0 = (1)0 +1 P0 ,
Xs = Zs s1 Bs1 , Y s = 0, s {1, . . . , L}.
(5.28)
The higher-stage tilde constraints can be rephrased as TAs+1 ZAs+1 Bs P Bs (1)Bs +s , where the
tilde structure functions ZAs+1 Bs are given by the old structure functions Zs+1 s ,

0 Zs+1 s (1)s +s+1

Bs
Bs
TAs+1 , C
(5.29)
= ZAs+1 =
, s {0, . . . , L 1},
0
0
and are therefore nilpotent. The rectangular matrix Y0 1 = Y0 1 (q, p) in Eq. (5.28) is the first
in a sequence of rectangular matrices Ys1 s = Ys1 s (q, p), s {1, . . . , L}, which are chosen
as a right inverse for the Zs s1 matrix,
Zs s1 Ys1 s = ss Ys s+1 Zs+1 s ,
s {1, . . . , L 1},
(5.30)
217
modulo matrices Ys s+1 and Zs+1 s associated with the next reducibility stage. One also assumes that the matrices Ys1 s commute with (1) the constraints T0 , (2) the structure functions
Zs+1 s and (3) among themselves, to avoid higher-order terms in expansions later-on. (This is
legitimate because after all, one could just have chosen all the structure functions as constants.)

T0 , Ys1 s = 0,

Zr r1 , Ys1 s = 0,

Yr1 r , Ys1 s = 0,
(5.31)
(5.32)
r, s {1, . . . , L}.
(5.33)
The first two conditions (5.31) and (5.32) are equivalent to

0 , Ys1 s = 0,
s {1, . . . , L}.
(5.34)
Notice that the tilde constraints TA in Eq. (5.28) are Abelian, i.e. they commute,
{TA , TB } = 0.
(5.35)
The involution (5.35) is not a consequence of any of the nilpotency relations that we have encountered so far. (However, see Appendix E.) The tilde constraints TA are also Abelian in the
antibracket sense, i.e.
(TA , TB )0 = 0,
(TA , TB ) = 0.
(5.36)
The antibracket involution (5.36) is closely related to the master equation (3.60), cf. Eq. (5.49)
below.
5.5. New BRST operator 1
The new BRST operator 1 may be calculated as the commutator of and S1 , cf. Eq. (3.48).
One finds

1 = {, S1 } = , C A TA C A {TA , }
= 0 {Y 0 , }
=B
T0 + 0
L1
L

B s {Xs , } + s Xs
s=0

00 Y0 1 Z1 0 P0 (1)0
B s+1 Zs+1 s s (1)s +s s+1 Zs+1 s Bs ,
(5.37)
s=0
where use is made of Eqs. (5.8) and (5.34) among others. Evidently the 1 operator has the
Abelian form
1 = C A TA =
L

B s Ts s Xs ,
s=0
(5.38)
218
with the new constraints TA {T ; (1) X } given by

T0 = {X0 , 0 } = T0 ,
X0 = X0 {Y 0 , 0 }

= 00 Y0 1 Z1 0 P0 (1)0 +1 ,
Ts = {Xs , }
Xs = Xs = Zs s1 Bs1 ,
= Zs s1 s1 (1)s1 +s1 ,
s {1, . . . , L},
(5.39)
where in general only the new constraints T in the first column of Eq. (5.39) are BRST-invariant.
The new constraints T0 are just the old constraints, since they are already BRST-invariant in the
Abelian case. The rle of the Y0 1 matrix is to appropriately reduce the number of independent
constraints among the zeroth-stage superpartners X0 . We also mention that the new higherstage constraints can be rephrased as TAs+1 ZAs+1 Bs P Bs (1)Bs +s , where the new structure
functions ZAs+1 Bs are two copies of the old structure functions Zs+1 s ,

Zs+1 s
0
Bs
Bs
= ZAs+1 =
TAs+1 , C
, s {0, . . . , L 1}.
0
(1)s+1 +s Zs+1 s
(5.40)
It is easy to check that the new constraints (5.39) do commute with the tilde constraints (5.28),
{TA , TB } = 0,
(5.41)
do commute among themselves,

{TA , TB } = 0,
(5.42)
and that the new BRST operator
1 = C A TA
is nilpotent.

The {S1 , 1 } commutator is -exact
{S1 , 1 } = (S1 , S1 ) = {S2 , },
(5.43)
according to the general theory, cf. Eqs. (3.57) and (3.60). When the commutator of S1 and 1
is calculated using the Abelian Anstze (5.27) and (5.37), one indeed finds that it is -exact:

{S1 , 1 } = B 1 Z1 0 200 Y0 1 Z1 0 P0 (1)0 + 1 Z1 0 Y0 1 P1 (1)1 +1
= B 1 Z1 0 Y0 1 Z1 0 P0 (1)0 + 1 Z1 0 Y0 1 P1 (1)1 +1

= , B 1 Z1 0 Y0 1 P1 (1)1 .
(5.44)
Here use is made of Eq. (5.30) in the second equality. In the irreducible case L = 0, one can
just let S2 = 0 be zero, but this is no longer true in the reducible case L > 0. (However, for an
alternative solution with S2 = 0 even for the reducible case L > 0, see Section 5.8 below.) In the
general Abelian case, it is consistent to define S2 as
L1

C As+1 ZAs+1 Bs TBs

S2 = C A TA = C A TA , C B TB =
s=0
= B Z1 Y0
1

P1 (1)1 = B 1 11 Y1 2 Z2 1 P1 (1)1 .
(5.45)
In particular, the TA constraints are linear combinations of the tilde constraints TA ,

TAs+1 = ZAs+1 Bs TBs .

TA = TA , C B TB ,
219
(5.46)
It is easy to check that they commute

{TA , TB } = 0.
(5.47)
One may wonder if there is a broader derivation of the fact that the Abelian Ansatz S1 = C A TA
fulfills the master equation (3.60)? It turns out that the question hangs on a somewhat artificially
looking formula,
(TA , C B ) TB = {TA , C B }{TB , },
(5.48)
which is essentially Eq. (5.30) in disguise. One calculates

(S1 , S1 ) = {C A TA , }, C B TB
= 2C A (TA , C B ) TB C A {TA , C B }{TB , } {, C A }{TA , C B }TB
= C A {TA , C B }{TB , } {, C A }{TA , C B }TB

= C A {TA , C B }TB , = {S2 , },

where use has been made of Eqs. (5.35), (5.36), (5.48) and

{TA , C B }, = 0.
(5.49)
(5.50)
Not surprisingly, there is a complete one-to-one correspondence between this derivation (5.49)
and the first derivation (5.44).
5.7. The higher Sk operators
There are no further non-trivial antibrackets,
1
1
(S1 , S2 ) = {S2 , 1 } + S1 , {1 , S1 } = 0,
2
2
(S2 , S2 ) = S2 , {1 , S1 } = 0,
(5.51)
(5.52)
cf. Eq. (5.43) and the explicit expressions (5.27), (5.44) and (5.45) for S1 , {S1 , 1 } and S2 .
Therefore it is consistent to put all the higher Sk operators to zero,
Sk = 0,
k {3, 4, 5, . . .},
(5.53)
cf. Eq. (3.67).

5.8. Another Abelian solution
We now apply the Abelian Ansatz (5.27) to the primed solution S
from Section 3.10. Recall
that S
is mediated by an -exact
fermionic generator of the form

S1 } = B 0 Y0 1 Z 0 1 P1 (1)1 0 Y0 1 Z 1 2 P2 (1)2 .
1 = {,
(5.54)
The change in S1 is -exact,
220

S1 S1
= {, 1 } = 0 Y0 1 00 Y0 1 Z1 0 Z 0 1 P1 (1)1 +1

B 0 Y0 1 Z 0 1 Z1 0 P0 (1)0 ,
(5.55)
cf. Eqs. (3.72) and (5.17). Then the primed solution S1

reads

S1
= 0 00 Y0 1 Z1 0 Z 0 1 P1 (1)1 +1
L

B 0 00 Y0 1 Z 0 1 Z1 0 P0 (1)0 +
B s Zs s1 Bs1 .
(5.56)
s=1
Derived from a canonical transformation (3.72), the primed solution (5.56) is guaranteed to meet
the correct boundary condition (3.48). Moreover, the primed solution has the remarkable property
that the S2
operator vanishes identically,
S2
= 0,
(5.57)
if one chooses 2 = 0. This is because the change in S2 is given by

S2
S2 = {, 1 }, S1 = B 1 Z1 0 Y0 1 P1 (1)1 = S2 ,
(5.58)
cf. Eqs. (3.73), (5.30) and (5.45). Hence it is consistent to choose all the higher operators
= 0 equal to zero. To simplify, let us specialize to the case of trivial anti-BRST symSk2
metry 0 = 0. In this case, the structure functions Z 0 1 = 0 vanish, and the S1
operator (5.56)
becomes
L

B s Zs s1 Bs1 .
S1
= B 0 00 Y0 1 Z1 0 P0 (1)0 +
(5.59)
s=1
Clearly the primed solution (5.59) (or its alter ego (5.56)) is too complicated to serve as a first
principle. For instance, the projection inside the B 0 P0 term looks rather artificial if postulated
from scratch. Also the Hessian of the primed solution (5.59) has a smaller rank than its unprimed
counterpart, and hence it represents a solution that is not proper.
6. Algebras of constraints
We now return to the fully non-Abelian case. An interesting question is what type of algebra
do the new constraints TA and tilde constraints TA in general obey? We will address this topic in
this section.
6.1. Unitary transformations
To study the question of constraint algebras, it is useful to observe how the new structure
functions behave under unitary transformations
i
A = e h G A(0) e h G = e{G,} A(0)
(6.1)
in the new ghost sector, where A(0) (respectively A) denotes any operator before (respectively after) a unitary transformation. Here G = G is a finite Hermitian generator of the transformation,
G = G0 +
L
221
C As GAs Bs P Bs (1)Bs +s
s=0
1
+
2
+
1
2
C Bs C Ar GAr Bs Cr+s+1 P Cr+s+1 (1)Bs +Cr+s+1 +r+1
r,s0
r+s+1L

C 2 P 2 , C P 3 , (6.2)
C Cr+s+1 GCr+s+1 Bs Ar P Ar P Bs (1)Ar +r + O C 3 P,
r,s0
r+s+1L
and where GA... B... = GA... B... (q, p; C, P ) depends on the old phase space variables. See
Ref. [27] for a related discussion. The quantum numbers for G are
(G) = 0,
gh(G) = 0 = ngh(G).
(6.3)
Hermiticity of G = G imposes non-trivial conditions on the GA... B... structure functions [10].
The G0 generates canonical transformations and rotations within the old sector, and it plays only
a relatively minor rle in the new ghost sector, so we shall assume that G0 = 0 is zero in this
section to keep the formulas as simple as possible. Then the BRST operator
0 = 0(0)
(6.4)
is invariant under such unitary transformations (6.1). Similarly, the constraints TA0 , TA0 and TA0
transform covariantly,
(0)
TA0 = A0 B0 TB0 ,
(6.5)
(0)
TA0 = A0 B0 TB0 ,
(6.6)
,
TA1 = A1 B1 TB(0)
1
(6.7)
where A B denotes the exponential of the matrix GA B , i.e.

1
1
B
A B A
+ GA B + GA C GC B + GA C GC D GD B + .
2
6
(6.8)
One deduces via Abelianization that there exists a set of new breve structure functions ZA1 B0
such that
TA1 = ZA1 B0 TB0 = O(T ),
(6.9)
because this holds in the Abelian case, cf. Eq. (5.46), and because both types of constraints TA0
and TA0 transform covariantly, cf. Eqs. (6.6) and (6.7). (For the purely rotational case, where
the generator G = C A GA B P B (1)B is linear in both C A and P A , the breve structure functions
ZA1 B0 = ZA1 B0 match their tilde counterparts as we shall soon see, cf. Eq. (6.11) below.) In
detail, let ( )A B denote the th power of the A B matrix, i.e. the exponential of the matrix
GA B . Then the structure functions ZAs+1 Bs , ZAs+1 Bs and VAs Bs transform as
1
ZAs+1
Bs

d 1 A
s+1
Cs+1

GCs+1 Ds E0 E

(0)
+ As+1 Cs+1 ZCs+1 Ds 1 D
Bs
s
F0
0

(0)
TF0 1 D
Bs
s

+ O(h ) terms, if R 2 ,
(6.10)
222
ZAs+1 Bs =
1

d 1 A
Cs+1
s+1

GCs+1 Ds E0 E

(0)
+ As+1 Cs+1 ZCs+1 Ds 1 D
F0
0

(0)
TF0 1 D
Bs
s
(1)Ds +s

+ O(h ) terms, if R 2 ,
(6.11)

(0)
VAs Ds = As Bs VBs Cs 1 C Ds + (1)As +s 0 , As Bs 1 B Ds
s
s

+ O(h ) terms, if R 2 and s 1 .
(6.12)
Bs
s
Note that Eq. (6.12) resembles the transformation law for a connection one-form, if one identifies
VAs Bs with the connection one-form and the BRST transformation {0 , } with the de Rham
exterior derivative. In this interpretation, the right-hand side of Eq. (4.1) behaves as a covariant
derivative, so that the new constraints TA0 on the corresponding left-hand side can transform
covariantly, cf. Eq. (6.5). The structure functions UA0 B0 inside S2 transform as
(0)
UA0 B0 = A0 B0 C0 D0 UC0 D0 +
1

d 1 A
C0 D0
0 B0

GC0 D0 E1 E
F1
1
(0)
TF1 .
(6.13)
Here A0 B0
C0 D0
denotes the exponential of the matrix

A0 B0 C0 D0 = 1 GA0 C0 D0 (1)A0 B0 (A0 B0 ) (1)C0 D0 (C0 D0 ).
G
(6.14)
B0
2
The antisymmetrization in the above equation (6.14) ensures the antisymmetry of the structure
(0)
= 0 in the Abelian case, there exist breve
functions UA B = (1)A0 B0 UB A . Since U
0 0
structure functions UA0 B0 C1 such that
A0 B0
UA0 B0 = UA0 B0 C1 TC1 = O(T ).
(6.15)
In the purely rotational case, the breve structure functions UA0 B0 = 0 vanish.
6.2. Algebra of constraints
The TA0 and TA0 constraints are in weak involution,
{TA0 , TB0 } = UA0 B0 C0 TC0 ,
{TA0 , TB0 } = EA0 B0 C0 TC0 + EA0 B0 C0 TC0 ,
(6.16)
{TA0 , TB0 } = FA0 B0 C0 TC0 ,
(6.18)
(6.17)
where UA0 B0 C0 is part of 1 , cf. Eq. (3.27), and where EA0 B0 C0 , EA0 B0 C0 and FA0 B0 C0 are some
new structure functions. The first involution (6.16) follows from the 1 nilpotency (3.24). The
involutions (6.17) and (6.18) are not consequences of any of the nilpotency relations that we have
encountered so far (However, see Appendix E), but follows for instance because of Abelianization. Namely, recall that there exist unitarily equivalent Abelian constraints in strong involution,
(0) (0)
TA0 , TB0 = 0,
(6.19)
(0) (0)
TA0 , TB0 = 0,
(6.20)
(0) (0)
= 0,
T , T
A0
223
(6.21)
B0
cf. Eqs. (5.35), (5.41) and (5.42), which imply that the TA0 and TA0 constraints are in general in
weak involution as displayed in Eqs. (6.16), (6.17) and (6.18). In the purely rotational case, the
structure functions read

(0) 1 E
B0 D0
D0
D0
C0
UA0 B0 E0 = A0 C0 TC(0)
+
T
,
(1)
D 0
B0
A0
B0
C0
0
0
2
A0 B0
(1)
(A0 B0 ),
(6.22)

(0)
FA0 B0 E0 = A0 C0 TC0 , B0 D0

(0) 1 E
1
+ (1)(B0 +1)(D0 +1) A0 D0 , B0 C0 TC0
D 0
0
2
(1)(A0 +1)(B0 +1) (A0 B0 ),
(6.23)
EA0 B0
E0
(0)
= (1)(A0 +1)B0 B0 C0 TC0 , A0 D0
(0) 1 E
1
B0 (D0 +1)
D0
C0
A0 , B0 TC0
+ (1)
D 0,
0
2
(6.24)

(0)
EA0 B0 E0 = A0 C0 TC0 , B0 D0

1 E
1
(1)(A0 +1)(B0 +D0 ) B0 D0 , A0 C0 TC(0)
D 0.
0
0
2
(6.25)
6.3. Antibracket algebra of constraints

The complete set of TA0 and TA0 constraints form a closed antibracket algebra,
(TA0 , TB0 )0 = QA0 B0 C0 TC0 ,
A B C0 TC ,
(TA0 , TB0 )0 = RA0 B0 C0 TC0 + R
0 0
0
(TA , TB ) = SA B C0 TC + SA B C0 TC ,
0
0 0
0 0
(6.26)
(6.27)
(6.28)
A B C0 , SA B C0 and SA B C0 are given by

where the structure functions QA0 B0 C0 , RA0 B0 C0 , R
0 0
0 0
0 0

B0
C0
C0
D0
C0
2QA0 B0 = TA0 , VB0 (1) VA0 UD0 B0
(1)(A0 +1)(B0 +1) (A0 B0 ),
(6.29)

2RA0 B0 C0 = TB0 , VA0 C0 (1)A0 B0 + VA0 D0 ED0 B0 C0
+ VB0 D0 EA0 D0 C0 (1)(A0 +1)(B0 +D0 +1)+B0 ,

A B C0 = TA , VB C0 (1)B0 + VA D0 ED B C0
2R
0 0
0
0
0
0 0
(A0 +1)(B0 +D0 +1)+B0
D0
C0
+ VB0 EA0 D0 (1)
UA0 B0 C0 ,
(6.30)
(6.31)
224
2SA0 B0 C0 = VA0 D0 FD0 B0 C0 +

TA0 , VB0 C0 EA0 B0 C0 (1)B0
(1)A0 B0 (A0 B0 ),
(6.32)
2SA0 B0 C0 = EA0 B0 C0 (1)B0 (1)A0 B0 (A0 B0 ),
(6.33)
cf. Eqs. (2.13), (3.32), (4.1), (6.16), (6.17) and (6.18).

It is an important observation that the new TA0 constraints by themselves form a closed
antibracket algebra (6.26). On the other hand, when considering the tilde constraints TA0 by
themselves, they are not necessarily in involution with respect to the antibracket (,)0 not
(0)
even at the classical level. This is despite the fact that the Abelian constraints TA0 satisfy
(T (0) , T (0) )0 = 0, cf. Eq. (5.36). Classically and on-shell with respect to the constraints TA0 ,
A0
B0
the antibracket (TA0 , TB0 )0 contains non-vanishing contributions
(0)
(0)
1
(TA0 , TB0 )0 A0 C0 TC0 , B0 D0 PB TD0 , 0 PB (1)B0
2
+ O(T ) (1)A0 B0 (A0 B0 ) for h 0.
(6.34)
B0 (q, p; C, P ),
and
There are practically no conditions on the rotation matrix A0 = A0
hence the right-hand side of Eq. (6.34) does not always vanish. The crucial difference between
Eq. (6.34) and the above commutator involutions (6.16), (6.17) and (6.18), is, that the operator
antibracket (2.13) does not satisfy the pertinent Leibniz rule, while the commutator does. One
can easily check with the help of Eq. (4.1) that the emerging extra contributions are proportional
to the new TA0 constraints, cf. last term on the right-hand side of Eq. (6.28).
B0
7. Conclusion
We have extended the construction of BRST-invariant constraints in Ref. [1] to include
reducible gauge algebras. We have also stressed a deep relationship with BRST/anti-BRST
symmetric models. Here the two nilpotent, Grassmann-odd, mutually anti-commuting BRST
operators come from a deformed version of the ordinary BRST operator 0 , and a new BRST
operator 1 = C A TA + , which encodes the new constraints TA . Note however, that all three
charges 0 , and 1 have ordinary ghost number +1, and only the latter operator is charges
with respect to the new ghost number, ngh(1 ) = 1, which is different from the usual BRST/antiBRST formulation. Nevertheless, one finds many similarities at the algebraic level. Some of them
is exposed in Table 1. In particular, we have constructed a unitarizing Hamiltonian that respects
the two BRST
operators and 1 with the help of a gauge boson.
The S
k=0 Sk operator, which satisfies an operator master equation (3.63), plays a prominent rle in the construction. For instance, the operator S1 = C A TA + contains the tilde
constraints TA , which decent through Eq. (1.15) to the BRST-invariant constraints TA . The existence of the S operator is deduced from cohomological considerations of the pertinent BRST
operators. In Appendix D we have considered various candidates to the operator master equation. In particular, we have analyzed the simplest cases, which are likely to become important for
practical calculations.
We have also investigated the algebras of new constraints TA and tilde constraints TA , cf.
Section 6. It is found that the corresponding commutator algebras are closed, but only the former
follows (with the machinery introduced in the main text excluding Appendix E) from a BRST
nilpotency relation. The full antibracket algebra of TA and TA constraints is also closed. So is the
225
antibracket algebra of TA constraints. However, the antibracket algebra of tilde constraints TA is,
on the other hand, in general an open algebra.
Acknowledgements
I.A.B. thanks R. Marnelius for numerous discussions. I.A.B. also thanks R. von Unge and the
Masaryk University for the warm hospitality extended to him in Brno. The work of I.A.B. and
K.B. is supported by the Ministry of Education of the Czech Republic under the project MSM
0021622409. The work of I.A.B. is also supported by grants RFBR 05-01-00996 and RFBR
05-02-17217.
Appendix A. Critical open bosonic string theory
For completeness, we here list the standard definitions that go into critical open bosonic
string theory. Let = be a constant invertible target space metric. There are bosonic
matter fields m and fermionic ghost variables cm and bm , where {1, . . . , D} and m
{. . . , 2, 1, 0, 1, 2, . . .} is an integer. The canonical commutation relations are

|m|
0
|m| 0
mn+m
,
[cm , bn ] = h
n+m ,
m , n = h
(A.1)
and zero in all the remaining sectors. Here 1 is a regularization parameter, which at the
end of the calculations should be set to 1. From a world-sheet perspective, the -regularization
smears the singularity of the operator product expansion (i.e. the OPE relation). The normal
ordering : :, also known as Wick ordering, is

n : = (n m)m
n + (m n)n m
,
:m
(A.2)
:bn cm : = :cm bn : = (n m)cm bn (m n)bn cm ,
(A.3)
where denotes the Heaviside step function with (0) = 12 .

Here we outline a simple (and we think compelling) proof of the conformal anomaly, which
does not rely on zeta function regularization or a choice of vacuum. (In reality, one should make
sure that the choice of normal ordering prescription can be accompanied with a compatible choice
of bra and ket vacuum states. This of course is the case.) One first rewrites the commutators of
Virasoro constraints (1.2) and (1.5) in terms of anti-supercommutators of the elementary modes
by straightforward algebraic manipulation, which uses the commutation relations (A.1),
[Tm , Tn ] =

h
i |i| mi , n+i
,
+
2
(A.4)
h
=
(m + i)(2n + i) |i| [bmi , cn+i ]+ (m n).
2
(A.5)
Tm(c) , Tn(c)
When one normal-orders the anti-supercommutators, cf. Eq. (1.17), one gets two terms: one
quadratic and one constant in the elementary modes,

0
m , n + = 2:m
(A.6)
n : + h |m| |m|n+m
,
0
[bm , cn ]+ = 2:bm cn : + h |m| sgn(m)n+m
.
(A.7)
226
Here sgn(m) denotes the sign of a number m with sgn(0) = 0. In the 1 limit, the quadratic
(c)
0
) and h(m
pieces in Eqs. (A.4) and (A.5) become h (m n)(Tm+n + h am+n
n)Tm+n , respectively. On the other hand, for any with || < 1, the constant pieces in Eqs. (A.4) and (A.5) are
() 0
(c) 0
and h 2 Am m+n
, respectively, where
of the form h 2 Am m+n
A()
m =D
i|j | |i|+|j | =
i,j
i+j =m
A(c)
m =
Dm(m2 1) |m|
,
12
(m + i)(m + j ) sgn(j ) |i|+|j | =
i,j
i+j =m
m(1 13m2 ) |m|

.
6
(A.8)
(A.9)
A few remarks are in order. In the restricted double summations (A.8) and (A.9), which are
absolutely and unconditionally convergent for || < 1, note that the (i, j )th term is antisymmetric
under an (i j ) exchange if the summation variables i and j have opposite signs. Therefore
one only has to consider is and j s with weakly the same sign. (The word weakly refers to
that i or j could be 0.) Since at the same time the sum i + j = m of i and j is held fixed, the
restricted (i, j ) double sum contains only finitely many terms, which may be readily summed to
give the familiar expression for the conformal anomaly, cf. Eqs. (A.8) and (A.9). In retrospect,
the -regularization has picked a particular (although very natural) summation ordering for two
infinite sums, which are conditionally convergent.
The BRST charge 0 and the ghost operator Gc read [2]

1
1 (c)
Tm cm +
(m n):bm+n cn cm : =
: Tm + Tm cm :
0 =
2 m,n
2
m
m

1
=
(A.10)
Tm cm
(m n)bm+n cn cm ,
2 m,n
m
Gc =

:cm bm :.
(A.11)
The last expression in Eq. (A.10), which has only normal ordering inside Tm , is useful when
proving the nilpotency relation (1.13). The Hermitian conjugate is defined in Eq. (1.16),
which leads to
Tm = Tm ,
0 = 0 ,
Gc = Gc .
(A.12)
Appendix B. Superfield formulation

A peculiar (although absolutely consistent) feature of the new ghosts sector is that the new
ghosts can be organized in superpartners that carry shifted ghost numbers. It is tempting to rewrite
the superpartner fields as N = 1 superfields by introducing a fermionic -coordinate, and absorb
the ghost number deficit into this ,
gh( ) = 1,
ngh( ) = 0,
( ) = 1.
In detail, one may rewrite the construction as follows

C ( ) C0 + C1 = B ,
C A C0 ; C1 B ; (1) +1
(B.1)

P A P 0 ; P 1 ; B
0 1

TA T ; T T ; (1) X
227
P ( ) P 0 + P 1 = + B ,
T ( ) T0 + T1 = T + X (1) .
(B.2)
As a result, the above superfields carry definite ghost number,

gh C s ( ) = s + 1 = gh Ts ( ) = gh P s ( ) 1.
(B.3)
A similar trick may be applied to the non-minimal variables (4.6) and the gauge-fixing condition
(4.19). The superfield transcription (B.2) basically amounts to rewrite all previous capital
indices,
A; B; C; . . . , in the corresponding superfield indices, , ; ,
; C,
; . . . . A sum A over a
repeated dummy
index, which one normally does not write explicitly, now also involve a Berezin
integration d , and so forth. In detail, we use the following superconventions,

d 1, () .
(B.4)
Then the canonical commutation relations (2.3) read

C ( ), P (
) = (
) = P ( ), C (
) .
(B.5)
Likewise, one gets

d C ( )T ( ) = B T X = C A TA ,
(B.6)

= 0 + d C ( )V (,
)P (
) d
+ ,
(B.7)

1
1 = d C ( )T ( ) +
d d
C (
)C ( )U (,
,
)P (
) d
+ ,
2
(B.8)
and

T ( ), T (
) = U (,
,
) d
T (
).
(B.9)
While aesthetically nice, the super-transcription unfortunately tend to increase the formula size,
which is why the superfield formulation is not used in the main part of the paper. We also stress
that this superfield formulation only affects the new ghost sector (C A , P B ), while the old phase
variables (q i , pj ; C , P ) remain in a non-supersymmetric formulation.
Appendix C. Matrix formulation
Here we give a matrix formulation of certain aspects of a rank 1 theory, i.e. a theory with
no terms of the form O(C P 2 ) in the power series expansions for and 1 , cf. Eqs. (3.15) and
(3.27). Some of the consequences of the -nilpotency (3.14) then read
[0 , 0 ] = 0,

B
C
VAs Cs , 0 (1)Cs +s = i hV
As s VBs s ,

(1)(Bs +s+1)(Cs +s+1) VAs Cs , VBs Ds = (1)(As +s)(Bs +s) (As Bs ),

VAr Cr , VBs Ds = 0, r = s,
(C.1)
(C.2)
(C.3)
(C.4)
228
where r, s {0, . . . , L}. Similarly, some of the consequences of the -closeness (3.26) for 1
read

B
C
B
C
ZAs+1 Cs , 0 (1)Cs +s = i hZ
(C.5)
As+1 s VBs s + i hV
As+1 s+1 ZBs+1 s ,

(Bs +s+1)(Cs +s+1)
Cs
Ds
(Cs +s)(Ds +s)
(1)
(C.6)
ZAs+1 , VBs
= (1)
(Cs Ds ),

(1)(Bs +s+1)(Cs +s+1) VAs Cs , ZBs Ds1 = (1)(As +s)(Bs +s) (As Bs ), s = 0, (C.7)

0,
ZAs+1 Cs , VBr Dr = 0, r s =
(C.8)
1,
where r {0, . . . , L} and s {0, . . . , L 1}. And finally, some of the consequences of the 1 nilpotency (3.24) read
ZAs+1 Bs ZBs Cs1 = 0,
(Bs+1 +s)(Cs +s+1)
s = 0,
ZAs+1 , ZBs+1
(1)

ZAr+1 Cr , ZBs+1 Ds = 0, r = s,
Cs
Ds
(C.9)
(Cs +s)(Ds +s)
= (1)
(Cs Ds ),
(C.10)
(C.11)
where r, s {0, . . . , L 1}. All these relations (C.1)(C.11) can precisely be recast into a nilpotency condition
]
=0
[,
(C.12)
for an operator
= C A A B P B (1)B = 0
L
C As P As (1)As +s i h
s=0
i h
L1
L
C As VAs Bs P Bs (1)Bs +s
s=0
C As+1 ZAs+1 Bs P Bs (1)Bs +s ,
(C.13)
s=0
cf. Table 3. The second and third term on the right-hand side of Eq. (C.13) contain the parts of
the and 1 operator that are linear in both C A and P A . The operator (C.13) is Grassmann = 1, has indefinite new ghost number (either 0 or 1), and is not
odd, has ghost number gh()
necessarily Hermitian. In turn, the nilpotency (C.12) is equivalent to following conditions for
the matrix elements A B = A B (q, p, C, P ),
A B B C = 0,
(1)(B +1)(C

+1)
A C , B

D
(C.14)
(1)A B (A B),
(1)C D (C D).
(C.15)
In particular, one sees that the operator-valued matrix element A B are nilpotent in a mixed
operator/matrix sense, cf. Eq. (C.14). The two possible right-hand sides of Eq. (C.15) are one
and the interplay
and the same condition written twice. A similar story is true for ,
yields an interesting canonical commutation relation in the operator/matrix sense,
B C = i h C ,
C +
A B
A
B
A
B
cf. Eq. (5.11) of Ref. [1].
(C.16)
229
Table 3
The operator-valued matrix A B
B
B0
A0
A0
(1)
A1
i h ZA1
B1
B
0 A0 i h VA0 B0
0
B0
A1
(1)
i h ZA2
A2
A3
..
.
B2
0 A1 i h VA1 B1
1
B1
A2
(1)
0 A2 i h VA2 B2
i h ZA3 B2
Appendix D. Operator master equations

In this appendix we consider candidates to the operator master equations for S and S,
cf. Eqs. (2.42) and (3.63), respectively. We start with candidates M = 0 to the master equation
for S, and we assume that M takes the form
M
N

n=0
n
n
(S, S, . . . , S ),

N = 0,
(D.1)
n entries
n (S, S, . . . , S) are higher antibrackets [17,20]. In general the nth (normalized, operator)
where
antibracket of n operators A1 , A2 , . . . , An , is defined as
n
(A1 , A2 , . . . , An )

1
(1),A . . . {, A(1) }, . . . , A(n) ,
n!

Sn
n normalized commutators
(D.2)
The sign-factor ,A arises from permuting the operators A1 , A2 , . . . , An under the permutation
Sn , see Ref. [20] for details, and see Refs. [28,29] for early field-antifield formulations of
higher antibrackets. The Ansatz (D.1) is partly motivated by the fact that the M operator has a
well-defined classical limit. This is because the multiple nested normalized commutators (1.18)
inside the higher antibracket (D.2) simply reduce to multiple nested Poisson brackets {,}PB
when h 0. Let us also assume that S0 is a linear combination of the old and the new ghost
operator,
S0 = G + G.
(D.3)

We are interested in which S0 and M that would guarantee a solution for S k=0 Sk . In
other words,
which coefficients , , 0 , 1 , 2 , . . . , would provide the existence of S? Note that

M=
M
k has quantum numbers
k=0
(M) = (Mk ) = 1,
ngh(Mk ) = k,
gh(M) = gh(Mk ) = 1.
(D.4)
Let us for the sake of simplicity restrict attention to the second-order case with N = 2, cf.
Eq. (D.1). We will show in this case that for each choice of S0 (more precisely, for each choice
of and = 0), there exists a unique M up to an over-all normalization factor. The proof goes
as follows:
230
Mk = 2
k
2
1
0
(Sj , Skj ) + 1
(Sk ) + k0 0
(D.5)
j =0

2
1
2 k1
j =1 (Sj , Skj ) + (1 2 (2 + k)) (Sk )
= (1 2 (2 + ))1
2
( 2 1 + 0 )
for k 1,
for k = 1,
for k = 0.
(D.6)
The equations M0 = 0 and M1 = 0 yield two conditions,

1
0
= ( + ),
= 2 + , 2 = 0.
2
2
The equation M2 = 0 then becomes
(D.7)
2
1
(S1 , S1 ) =
(S2 ) {, S2 }.
(S1 , S1 )
(D.8)
One knows that the antibracket (S1 , S1 ) = 0 does not always vanish, see for instance Eq. (5.44),
so one must demand that
= 0.
(D.9)
One next seeks for an integrability condition I = 0, where

m

n
M (S, S, . . . , S ), S, S, . . . , S ,
nm
I

m,n
m+nN
m entries
(I) = 0,
gh(I) = 2.
n1 entries
(D.10)
The coefficients nm are to be adjusted such that Eq. (D.10) is a non-trivial linear combination
of generalized Jacobi identities for the antibracket hierarchy. In the N = 2 case, the relevant
generalized Jacobi identities are [20]
0
1
(D.11)
= 0,
1

1
(S) = 0,
(D.12)
0
2
, S = 0,
(D.13)

1
2
1
1
(S, S) + 2
(S), S = 0,
(D.14)

2
2
1
3
6 (S, S), S (S, S, S) = 0.
(D.15)
They are all consequences of the nilpotency (3.14). In the N = 2 case, the integrability condition (D.10) reads
0
1
0

1
1
2
M + 11
M (S) + 20
M , S
I 10

1
2
1
0
2
(S, S) + 1
(S) + 0
= 10

1
3
2
1
2
(S, S, S) + 1
(S, S) + 0
(S)
+ 11

2
2
1
0
(S, S) + 1
(S) + 0
, S)
2
+ 20
2
3

1
1
(S, S) + 11 2
(S, S, S)
= (10 2 + 11 1 )

2
1
2
2
(S, S), S + 20 1
(S), S .
+ 20 2
(D.16)
The nm coefficients are then tuned so that I = 0 is automatically satisfied. This yield two conditions:
20
10
1
(D.17)
= 6,
= 4 = 4(2 + ), 11 = 0.
11
11
2
I=
k=0 Ik
231
is itself a sum of terms Ik with quantum numbers,
(Ik ) = 0,
ngh(Ik ) = k,
gh(Ik ) = 2.
(D.18)
One can now show the existence of a solution S to the master equation M = 0. Recall that the
existence of S1 and S2 follows from the -closeness (3.26) and the nilpotency (3.24) of 1 , cf.
Sections 3.7 and 3.8, respectively. One now proceeds by mathematical induction in the integer
k 3. Assume that there exist S1 , S2 , . . . , Sk1 such that M0 = M1 = M2 = = Mk1 = 0.
One wants to prove that there exists Sk such that Mk = 0. The kth integrability condition reads

20
Ik = 10 {, Mk } + 11 , {Mk , S0 } +
{, Mk }, S0 + {, S0 }, Mk
2
= 2(k 2)11 {, Mk },
(D.19)
cf. Eq. (D.16). Ik = 0 implies that Mk is -closed, and hence -exact, cf. statement after
Eq. (3.22). Therefore the 2 -term in Eq. (D.6) must be -exact, and hence there exists an Sk
that makes Mk = 0 vanish. The factor (k 2) in Eq. (D.19) indicates that non-trivial information
must be added to the construction at the second induction step, k = 2, namely the 1 nilpotency
(3.24).
The master equation (D.1) simplifies further if 0 = 0, i.e. if is either equal to 0 or , cf.
Eq. (D.7). In the main part of this paper we choose5 S0 = G corresponding to (, ) = (0, 1).
Now let us drop the use of calligraphic letters and consider candidates M = 0 to the master
equation for S in the old sector, cf. Eq. (2.42). The analysis is very similar to the above case, so
we shall be brief. Let us assume that S0 is proportional to the old ghost operator,
S0 = G.

Note that M =
k=0 M12k has quantum numbers
(M) = (M12k ) = 1,
gh(M12k ) = 1 2k.
(D.20)
(D.21)
One finds
M12k = 2
k

j =0
2
1
0
(S2j , S2(j k) ) + 1
(S2k ) + k0 0
0
0
0
k1
2 (S
1
2 j =1
2j , S2(j k) ) + (1 + 2(k 1)2 )0 (S2k )
0
= 1 0
2
( 2 1 + 0 )0
(D.22)
for k 1,
for k = 1,
for k = 0.
(D.23)
The equations M1 = 0 and M1 = 0 yield two conditions,

0
= 2 ,
2
1
= 0,
2
2 = 0.
(D.24)
The equation M3 = 0 then becomes

2
1
(S2 , S2 ) = 2
(S4 ) 2{0 , S4 }.
(S2 , S2 )0
0
0
(D.25)
5 Note that Ref. [1] has S = G G corresponding to (, ) = (1, 1). The operator master equation (3.63) remains
0
the same, while there is a change of sign in Eq. (3.60); see Eqs. (5.22) and (5.19) in Ref. [1].
232
One knows that the antibracket (S2 , S2 )0 = 0 does not always vanish, so one must demand
that
= 0.
(D.26)
As before one adjusts the nm coefficients so that the integrability condition (D.16) is automatically satisfied. This yield two conditions:
20
10
1
= 6,
= 4 = 0, 11 = 0.
11
11
2

I = k=0 I2(1k) is itself a sum of terms I2(1k) with quantum numbers,
(I2(1k) ) = 0,
gh(I2(1k) ) = 2(1 k).
(D.27)
(D.28)
The kth integrability condition reads

I2(1k) = 10 {0 , M12k } + 11 0 , {M12k , S0 }

20
{0 , M12k }, S0 + {0 , S0 }, M12k
+
2
= 4(2 k)11 {0 , M12k }.
(D.29)
In the main text we have chosen S0 = G corresponding to = 1.

Appendix E. A new tilde BRST operator 1
In this appendix we briefly outline how the involution relations (6.17) and (6.18) could be
reformulated as nilpotency requirement for a new tilde BRST operator 1 ,
{ 1 , 1 } = 0,
( 1 ) = 1,
gh( 1 ) = 1,
ngh( 1 ) = 1.
(E.1)
The tilde BRST operator 1 is a deformation of 1 ,

1 = 1 +
L

s=0
1 Bs Ar
CAs TAs +
C C FAr Bs Cr+s P Cr+s (1)Bs +Cr+s +r
2
r,s0
r+sL

C Bs CAr EAr Bs Cr+s P Cr+s EAr Bs Cr+s P Cr+s (1)Bs +Cr+s +r + .
r,s0
r+sL
(E.2)
Here we have introduced a new set of tilde ghosts CA {C0 , C1 } and P A {P 0 , P 1 } with
canonical commutation relations
A
A B
C , P B = BA = (1)B P B , CA ,
{P A , P B } = 0,
C , C = 0,
(E.3)
which have Grassmann parity and new ghost number given by

CA = A = (P A ),

CAs = As + s = (P As ),

ngh CAs = s + 1 = ngh(P As ).
(E.4)
233
The Grassmann parity and the old ghost number are shifted among the tilde ghost superpartners
as follows

C0s = s + s = P 0s ,
C0 = = P 0 ,

C1 = + 1 = P 1 ,
C1s = s + s + 1 = P 1s ,

gh C0s = s + 2 = gh P 0s ,

(E.5)
gh C1s = s + 3 = gh P 1s .
It is now straightforward to check that the weak involution relations (6.16)(6.18) follow from
the 1 nilpotency (E.1). An aesthetically drawback of the approach, is, that the tilde constraints
TA appear inside two different generating expansions, namely S1 and 1 .
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]
I.A. Batalin, I.V. Tyutin, Theor. Math. Phys. 138 (2004) 1.

S. Hwang, Phys. Rev. D 28 (1983) 2614.
M.B. Green, J.H. Schwarz, E. Witten, Superstring Theory, vol. 1, Cambridge Univ. Press, 1987.
C. Becchi, A. Rouet, R. Stora, Phys. Lett. B 54 (1974) 344;
C. Becchi, A. Rouet, R. Stora, Commun. Math. Phys. 42 (1975) 127;
C. Becchi, A. Rouet, R. Stora, Ann. Phys. (N.Y.) 98 (1976) 287.
I.V. Tyutin, Lebedev Institute preprint 39, 1975.
E.S. Fradkin, G.A. Vilkovisky, Phys. Lett. B 55 (1975) 224;
E.S. Fradkin, G.A. Vilkovisky, CERN preprint TH 2332-CERN, 1977.
I.A. Batalin, G.A. Vilkovisky, Phys. Lett. B 69 (1977) 309.
E.S. Fradkin, T.E. Fradkina, Phys. Lett. B 72 (1978) 343.
I.A. Batalin, E.S. Fradkin, Phys. Lett. B 122 (1983) 157.
I.A. Batalin, E.S. Fradkin, Phys. Lett. B 128 (1983) 303;
I.A. Batalin, E.S. Fradkin, Riv. Nuovo Cimento 9 (1986) 1.
I.A. Batalin, R. Marnelius, Phys. Lett. B 441 (1998) 243.
I.A. Batalin, E.S. Fradkin, Ann. Inst. Henri Poincar 49 (1988) 145.
T. Kugo, I. Ojima, Progr. Theor. Phys. Suppl. 66 (1979) 1.
K. Nishijima, Nucl. Phys. B 238 (1984) 601;
K. Nishijima, Progr. Theor. Phys. 80 (1988) 897;
K. Nishijima, Progr. Theor. Phys. 80 (1988) 905.
M. Henneaux, Phys. Rep. 126 (1985) 1.
M. Henneaux, C. Teitelboim, Quantization of Gauge Systems, Princeton Univ. Press, 1992.
I.A. Batalin, R. Marnelius, Phys. Lett. B 434 (1998) 312.
I.A. Batalin, R. Marnelius, Int. J. Mod. Phys. A 14 (1999) 5049.
I.A. Batalin, R. Marnelius, Theor. Math. Phys. 120 (1999) 1115.
K. Bering, hep-th/0603116.
I.A. Batalin, M. Grigoriev, Int. J. Mod. Phys. A 18 (2003) 4485.
G. Curci, R. Ferrari, Phys. Lett. B 63 (1976) 91.
I. Ojima, Progr. Theor. Phys. 64 (1980) 625.
I.A. Batalin, P.M. Lavrov, I.V. Tyutin, J. Math. Phys. 31 (1990) 6;
I.A. Batalin, P.M. Lavrov, I.V. Tyutin, J. Math. Phys. 31 (1990) 2708;
I.A. Batalin, P.M. Lavrov, I.V. Tyutin, Int. J. Mod. Phys. A 6 (1991) 3599.
W. Kalau, J.W. van Holten, Nucl. Phys. B 361 (1991) 233.
I.A. Batalin, S. Lyakhovich, R. Marnelius, Phys. Lett. B 534 (2002) 201.
I.A. Batalin, I.V. Tyutin, Int. J. Mod. Phys. A 20 (2005) 895.
K. Bering, P.H. Damgaard, J. Alfaro, Nucl. Phys. B 478 (1996) 459.
I.A. Batalin, K. Bering, P.H. Damgaard, Phys. Lett. B 389 (1996) 673.

Nucl - Phys.B v.771

Transféré par

Informations du document

Titre original

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

Nucl - Phys.B v.771

Transféré par

Droits d'auteur :

Formats disponibles

NUCLEAR PHYSICS B

Journal devoted to the experimental and theoretical study of the fundamental

Nuclear Physics B 771 (2007) 127

Neutrino mass hierarchy and 13 with a magic baseline

Received 1 November 2006; accepted 1 February 2007

E-mail addresses: sanjib@mri.ernet.in (S.K. Agarwalla), sandhya@mri.ernet.in (S. Choubey),

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

studies involving beta-beams have concentrated on using megaton water Cerenkov

resultant low energy muons efficiently and unambiguously. Water Cerenkov

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

2. The -beam fluxes

2.2. Candidate ions for the -beam

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

sin 213 < 0.17,

sin2 [(1 A)]

+ sin 213 sin 212 sin 223 sin CP sin()

+ sin 213 sin 212 sin 223 cos CP cos()

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

= AL/4E by definition, the condition (11) reduces to L = 2/GF Ye , where

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

dEA (E) (E)R(E, EA )Pe (E)

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

and e charged current events

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

is the observed number of events, while

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

Neutrino beam (NH true)

Antineutrino beam (IH true)

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

sin2 213 (max) sin2 213 (min)

L = 520 km and a water Cerenkov

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

Fig. 10. Same as Fig. 9 but for 10 yr of running.

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

Neutrino beam (NH true)

Antineutrino beam (IH true)

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

S.K. Agarwalla et al. / Nuclear Physics B 771 (2007) 127

Nuclear Physics B 771 (2007) 2839

CP violation in neutrino mass matrix

E-mail address: utpal.sarkar@cern.ch (U. Sarkar).

U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839

U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839

so that the physical neutrinos satisfy the Majorana condition

U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839

From these transformations it is possible to construct the rephasing invariants [3]

U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839

U. Sarkar, S.K. Singh / Nuclear Physics B 771 (2007) 2839

Im[Iii Ijj Iiijj ]

there are n(n 1)/2 independent measures for n generations.

sin2 [(1 A)]

+ sin 213 sin 212 sin 223 sin CP sin()

+ sin 213 sin 212 sin 223 cos CP cos()