Nucl - Phys.B v.745 PDF

Nuclear Physics B 745 (2006) 128
Thermodynamics of noncritical M-theory

and the topological A-model
Petr Horava, Cynthia A. Keeler
Berkeley Center for Theoretical Physics and Department of Physics, University of California,
Berkeley, CA 94720-7300, USA
Theoretical Physics Group, Lawrence Berkeley National Laboratory Berkeley, CA 94720-8162, USA
Received 31 December 2005; accepted 27 February 2006
Available online 24 March 2006
Abstract
In [P. Horava, C.A. Keeler, Noncritical M-theory in 2 + 1 dimensions as a nonrelativistic Fermi liquid,
hep-th/0508024], noncritical M-theory for two-dimensional type 0A and 0B strings was defined in terms
of a double-scaled theory of nonrelativistic fermions in 2 + 1 dimensions. Here we study this noncritical
M-theory at finite temperature. We derive the exact expression for the free energy of its vacuum solution, as a
function of a coupling constant gM and the radius R of the thermal circle. We show that at high temperature,
the theory is effectively described by another M-theory solution, whose effective loop-counting coupling
scales in a novel way characteristic of M-theory, as T 3 . Our calculations further suggest that noncritical
M-theory is dual to the closed string theory of the topological A-model on a CalabiYau, with the radius R
of the Euclidean time circle in M-theory playing the role of the string coupling constant of the A-model. In
this correspondence, T-duality on the Euclidean time circle of noncritical M-theory implies an S-duality for
the topological A-model.
2006 Elsevier B.V. All rights reserved.
1. Introduction and summary

Noncritical string theories in two spacetime dimensions (see [28] for reviews) provide a
useful laboratory in which many features of full string theory can be studied in a simpler, exactly
solvable setting. One aspect of the full theory which continues to be shrouded in mystery is its
M-theory regime, where fundamental strings are no longer the correct degrees of freedom. In this
regime, the dynamics of the theory can be described at low energies by an effective supergravity
E-mail addresses: horava@berkeley.edu (P. Horava), ckeeler@berkeley.edu (C.A. Keeler).
0550-3213/$ see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.nuclphysb.2006.02.039
P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128
theory, with M2-branes and M5-branes as solitons. When combined with anticipated dualities to
various string vacua, this rather incomplete picture still leads to a wealth of useful information
about M-theory. However, a more direct understanding of the M-theory degrees of freedom, in
which one would not have to resort to a nonperturbative duality, is clearly desirable.
This problem appears to be difficult in full eleven-dimensional M-theory, but it is precisely
the type of question that could be addressed first in the simplified setting of two-dimensional
string theory. Type 0A and 0B strings in two dimensions [9,10] seem particularly suitable for
this purpose. The spectrum of two-dimensional type 0A string theory contains stable D0-branes,
charged under a U (1) RR symmetry. The D0-brane charge can be naturally interpreted as the
KaluzaKlein momentum along an extra S 1 dimension of space, which we interpret as the extra S 1 of noncritical M-theory [1]. It is intriguing that in the Fermi-liquid formulation of the
eigenvalues of the matrix model, the extra S 1 simply plays the role of the angular variable on
a flat two-dimensional plane populated by the fermions. Noncritical M-theory is thus defined
in terms of a large N , double-scaling limit of a system of N nonrelativistic fermions in the inverted harmonic oscillator potential on this eigenvalue plane. We will describe the fermions
in the second-quantized framework, as quanta of a spinless fermionic field (t, 1 , 2 ), whose
dynamics is governed in the large N limit by the nonrelativistic action

1
1 2 2
1
2

2
S = N dt d i
(1.1)
+ + 2 .
t
2 1 1 2 2 2 2 0 1
Here (1 , 2 ) are Cartesian coordinates on the flat eigenvalue plane R2 , and 0 is a fundamental
frequency of the theory. We will set 0 = 1 throughout the paper.
In noncritical M-theory so defined, one can find two-dimensional type 0A and 0B string vacua
as solutions (as we will review briefly in Section 2.1). In addition, the space of all solutions also
includes a true ground state of the system, in which the lowest N single-particle states have
been filled by the N available fermions, up to some distance from the top of the potential. In
the double-scaling limit, which is the semiclassical limit of the Fermi liquid described by (1.1),
N is sent to infinity while the Fermi energy is held fixed. gM = 1/ plays the role of a coupling
constant in this vacuum. Some properties of this vacuum of noncritical M-theory were studied
in [1].
In this paper, we intend to probe this solution further, with the hope of learning more about the
nature of its collective excitations. For any given physical system, useful information about its
underlying degrees of freedom can be revealed by exposing the system to extreme physical conditions, such as high temperatures, high-energy scattering, or strong fields. This proven strategy
was used, for example, in the early days of superstring theory: A series of gedanken experiments
was designed (see, e.g., [1115]) to reveal the underlying degrees of freedom of the theory, or its
hypothetical unbroken phase in which large underlying symmetries could become manifest. In
this paper, we shall apply this strategy to the vacuum state of noncritical M-theory, and examine
it at finite temperature.1
In quantum field theory, it is well known that the high-temperature behavior is formally governed by a field theory in one fewer spacetime dimension, with the effective loop-counting coupling that scales as T . Frequently, this naive picture is strongly modified by renormalization effects. In string theory at finite temperature [14,15,1721], on the other hand, the high-temperature
behavior of one string vacuum is formally mapped by T-duality to the low-temperature behavior
1 Aspects of the high-energy behavior of noncritical M-theory will be studied elsewhere [16].
of another string vacuum in the same spacetime dimension. The effective loop-counting coupling
of this dual string theory scales as T 2 [14]. In most string vacua, this naive picture is modified by
the existence of the Hagedorn phase transition, where the canonical ensemble description breaks
down, and one needs to resort to the microcanonical ensemble.
One naturally wonders what is the high-temperature behavior of M-theory; in particular, is it
more akin to field theory, or string theory, or neither? In eleven dimensions, this question again
seems very difficult to answer, since we lack the required control over the nonsupersymmetric compactifications of M-theory that would formally define its thermodynamic ensemble in
the regime of interest.2 On the other hand, the same question can be addressed in noncritical
M-theory in 2 + 1 dimensions. In Section 2, the exact partition function of the vacuum solution
of noncritical M-theory at finite temperature will be evaluated, and its high temperature behavior
studied. We shall find the following:
The behavior of the M-theory vacuum state at high temperatures is governed by another,
dual solution of noncritical M-theory, related to the original vacuum by a symmetry similar to T-duality, which noncritical M-theory inherits from the underlying T-duality between
type 0A and 0B strings on a compact Euclidean time circle.
In the high-energy limit, the free energy of the system scales as T 3 , i.e., in a way characteristic of a massless relativistic field theory in 2 + 1 dimensions.
In this dual M-theory solution governing the high-temperature regime, the effective loopcounting coupling scales at high temperature as T 3 . This is a novel scaling behavior, unlike
in field theory ( T ) or string theory ( T 2 ).
Our results in Section 2 will be obtained using the grand-canonical partition function of
noncritical M-theory at finite temperature T . The grand-canonical ensemble turns out to be
equivalent, albeit in a somewhat subtle way involving intricacies of the double-scaling limit,
to the canonical ensemble. Once this equivalence is established, one can ignore the thermodynamic interpretation of the partition function, and simply interpret it as the partition function of
noncritical M-theory in Euclidean signature, with the Euclidean time dimension compactified on
a circle of radius R = 1/(2T ).
This Euclidean compactification of noncritical M-theory now has two U (1) isometries: the rotation on the spatial eigenvalue plane, and the Euclidean time translation. The angular dimension
on the eigenvalue plane has been interpreted as the M-theory dimension for two-dimensional
type 0A strings, with the angular momentum on the plane playing the role of the D0-brane charge
in type 0A string theory. In the next step, it is natural to ask whether the Euclidean time dimension can also be interpreted as the M-theory dimension of some string theory. This question
will be analyzed in Section 3. Somewhat surprisingly, we find evidence that this question has
a simple answer: The string theory associated with the reduction of noncritical M-theory along
the Euclidean time circle is the closed string theory of the topological A-model on the resolved
conifold!
The thermal ensemble of noncritical M-theory naturally corresponds to the path integral with
the antiperiodic condition on the Fermi field around the Euclidean time tE . In the study of the
relation to the A-model, it is more natural to relax this condition, and study noncritical M-theory
2 Some papers discussing critical M-theory at finite temperature include [22,23]; see also [24].
with periodic up to an arbitrary phase b,

(tE + 2R, 1 , 2 ) = eib (tE , 1 , 2 ).
(1.2)
We find that this entire class of Euclidean compactifications of noncritical M-theory is related to
the topological A-model. In this correspondence, the radius R of the Euclidean time circle, the
coupling constant gM and the phase b of noncritical M-theory are related to the A-model string
coupling gA and the Khler modulus tA of the resolved conifold by a simple duality relation,
gA = 2iR,
tA =
2R
+ ib.
gM
(1.3)
Note that what is a coupling constant on one side of the duality becomes a geometric parameter
on the other side of the duality, and vice versa.
This duality between noncritical M-theory and the topological A-model can be expected to
have a wide range of interesting implications. Some of them are:
The relation between the string coupling gA and the radius R of the M-theory circle suggests that noncritical M-theory can play the role of the topological M-theory [25,26] for the
topological strings of the A-model.
Noncritical M-theory provides a nonperturbative completion of the topological closed string
theory of the A-model.
T-duality of noncritical M-theory (inherited from the underlying T-duality of two-dimensional type 0A and 0B strings [27]) implies an S-duality for the topological A-model.
In combination with the GopakumarVafa duality [28,29], our duality relates noncritical
M-theory to ChernSimons gauge theory on S 3 .
In the melting crystal interpretation of the quantum foam phase of CalabiYau [30], the
temperature of the crystal turns out to be equal to the temperature of the noncritical M-theory
vacuum. The constituents of the crystal and the quantum foam of the CalabiYau appear
intimately related to the constituent fermions of noncritical M-theory.
2. Noncritical M-theory at finite temperature
Noncritical M-theory for two-dimensional type 0A and 0B strings is defined by the nonrelativistic Lagrangian (1.1) only formally. The proper definition, discussed in detail in [1], begins
with a finite number N of fermions, and with the inverted harmonic oscillator potential replaced with a more general regulating potential V (1 , 2 ). This potential includes stabilizing
anharmonic terms which ensure that the spectrum is bounded from below. The simplest way of
mimicking such terms is to place an infinite wall at some distance from the origin, which cuts off
the single-particle energy spectrum from below at some ; this will be the regulator assumed
throughout this paper. In the large N limit, the anharmonic terms are scaled away (or is taken
to infinity), and the only relevant piece of the potential in this double-scaling limit is its inverted
harmonic oscillator part.
2.1. Noncritical M-theory for type 0A and 0B strings in two dimensions
Both type 0A and 0B string vacua in two dimensions are solutions of this theory. In order to
see that, note that the Hamiltonian associated with (1.1) can be viewed from two complementary
perspectives.
In the polar coordinates (, ) on the eigenvalue plane, the Fermi field (and its conjugate
momentum ) can be decomposed
into eigenfunctions of the single-particle angular momen
tum operator J , (, ) = qZ eiq q (). Consequently, the second-quantized Hamiltonian
decomposes into an infinite sum,

Hq q , q .
H , =
(2.1)
qZ
Each individual term Hq is equivalent to the Hamiltonian of the type 0A string theory in the
linear dilaton vacuum with RR flux q.
In the Cartesian coordinates (1 , 2 ), the single-particle Hamiltonian h = h1 + h2 describes
two decoupled oscillators. The energy h2 of (say) the second oscillator is a conserved quantity.
We can decompose into a complete basis of eigenfunctions (2 ) of h2 . Since the eigenvalues
of h2 are continuous, we obtain an integral3

(1 , 2 ) = d (1 ) (2 ).
(2.2)
As a result, the second-quantized Hamiltonian H decomposes into an integral over of a oneparameter family of decoupled Hamiltonians H , each of which depends only on a single secondquantized field (1 ). Each member of this family of Hamiltonians is essentially equivalent to
the Hamiltonian of two-dimensional type 0B string theory in Fermi liquid representation.
Having understood these facts about noncritical M-theory, it is now clear how the type 0A and
0B vacua can be constructed as exact solutions of this theory. The type 0A linear dilaton vacuum
with RR flux q is given simply by the state of the (2 + 1)-dimensional noncritical M-theory in
which all single-particle states with angular momentum J = q are occupied up to some (double
scaled) Fermi energy = N F , while all fermion states with J = q are empty. The physics
of this solution is exactly equivalent to the physics of the linear dilaton ground state of the type
0A theory in the Fermi liquid representation. In particular, since the Fermi sea of this state is
empty in all sectors with J = q, all excitations in those sectors are infinitely energetic from
the perspective of the type 0A ground state, and they decouple in the double-scaling limit. The
physics of the remaining excitations is then isomorphic to that of type 0A string theory in the
Fermi liquid picture.
Similarly, the type 0B linear dilaton vacuum corresponds to the Fermi sea in which the N
fermions occupy (up to the Fermi energy ) only the single-particle states whose eigenvalue of
h2 is equal to some fixed , while all single-particle states whose eigenvalue h2 = are empty.
In the double-scaling limit, all excitations with h2 = are infinitely energetic and decouple. The
physics of the remaining finite-energy excitations of this solution is precisely equivalent to the
Fermi liquid representation of the type 0B string theory. In the type 0Aand 0B string vacua of
noncritical M-theory, the fundamental frequency 0 plays the role of 1/ 2 [1].
In addition to reproducing such known two-dimensional string vacua, noncritical M-theory
has other interesting solutions. The one of primary interest to us is the true ground state of
the theory, in which the N available fermions simply occupy the lowest N single-particle states
of the theory, irrespective of what other quantum numbers they may carry (and with the Fermi
energy held fixed in the large N limit).
3 More precisely, for each eigenvalue of h there are two eigenstates, one for each parity under . We
2
2
2
implicitly include the sum over parity in our definition of the integral over .
Fig. 1. The ground state of noncritical M-theory in the Fermi liquid representation. The inverted harmonic oscillator
potential is rotationally invariant on the eigenvalue plane, with the second dimension suppressed in the figure. All single-particle states are occupied from the cutoff up to the top of the Fermi sea, some distance below the top of the
potential.
In this simplified context of noncritical M-theory, it is this natural vacuum state which is
perhaps the closest analog of the eleven-dimensional vacuum of full M-theory. Various properties
of this vacuum state of noncritical M-theory were studied in [1], and we will review some of them
in Section 2.5 below.
It is also worthwhile to point out that families of static solutions interpolating between the
M-theory vacuum and the type 0A or 0B string theory can be constructed. Starting from type
0A (or 0B) solutions as described above, one can raise the Fermi surface in sectors with J = q
(or h2 = ) to some finite distance below the top of the potential, making the excitations in
those sectors finitely energetic. This change in the Fermi surface represents a hidden deformation
parameter of two-dimensional type 0A and 0B string theory.
2.2. The grand canonical ensemble
We find it convenient to analyze the thermodynamics of noncritical M-theory using the grand
canonical ensemble of the fermions. In this ensemble, the central object of our interest will be the
thermodynamic potential (, ), which is a function of the inverse temperature = 1/T and
the (double-scaled) chemical potential associated with the conserved number of fermions N .
(, ) is defined in terms of the grand canonical partition function ZM (, ) of the noncritical
M-theory vacuum,
1
1
(, ) = log ZM =

d ( ) log 1 + e( ) .
(2.3)
Here log ZM has been related to the single-particle density of states ( ) at zero temperature and
at energy , a quantity studied in detail in [1].
In the thermodynamic limit, which is a part of the double-scaling limit, this ensemble is
equivalent to the canonical ensemble provided the fluctuations in the number of particles are
suppressed, a condition that we will return to in Section 2.4 below. Potentially, the canonical ensemble in string theory is known to suffer from several instabilities, including the Jeans instability
common to all gravitating systems, and the stringy instability due to the Hagedorn phase transition. These destabilize the canonical ensemble and require the full power of the microcanonical
ensemble. Such instabilities do not occur in type 0A and 0B string theories in two dimensions,
Fig. 2. The density of states (2.4) as a function of the single-particle energy of the fermions. The density is invariant
under . Notice also the importance of the nonuniversal -independent constant in (2.4): if we ignored this
-dependent term, the density of states would be negative.
and we will see that they do not occur in noncritical M-theory either. The microcanonical, canonical and grand canonical ensembles are all equivalent, but the argument is surprisingly subtle and
involves the double-scaling limit in a nontrivial way.
In [1], an exact formula for the density of states ( ) in the M-theory vacuum was derived,
1
( ) =
Re
4
d ei
1
2
sinh (/2)
+ .
2 tanh( )
2
(2.4)
The integral representation for the density of states in (2.4) is divergent, but the divergence is
rather mild, with the divergent piece independent of and resulting in a simple dependence of
( ) on the cutoff . In the double scaling limit, is scaled to infinity with N . As in [1], we
only keep the leading dependence on , dropping all terms O(1/).
Similarly, the formal expression (2.3) for is divergent. In most of our arguments, we will
concentrate only on the universal part of all the thermodynamic variables. Thus, in order to
eliminate all dependence on and isolate the universal information in , we use a trick standard
in noncritical string theory, and evaluate
3
= 2
3

d
( )
1
,
2
4 cosh [( )/2]
(2.5)
which is cutoff-independent. We will return to the dependence of the thermodynamic quantities

on the cutoff in the next subsection.
Using the integral representation (2.4) for ( ) and performing the integral over by residues,
we get
3
1
=
3
4

d ei
d Re
1
=
Im
2

d ei
0
2
2
sinh (/2)
/2
1
1 + e2R( )
/(2R)
.
sinh (/2) sinh[/(2R)]
2
(2.6)
This integral representation for the thermodynamic potential of the M-theory vacuum is similar to
the corresponding expressions in noncritical string theory [31] (see also [2,4] for recent reviews).
Note thatunlike for example in noncritical bosonic string theorythe M-theory formula does
not exhibit any manifest form of self T-duality. However, as we will see later, M-theory does
inherit a form of T-duality from the underlying T-duality relating type 0A and 0B strings with
RR flux.
2.3. Terms of low order in
The integral formula (2.6) is convergent, but it determines (, ) only up to a second order
polynomial in ,
2
(2.7)
.
2
The coefficients i () are cutoff-dependent, and divergent in the double-scaling limit. However,
before we dismiss them as nonuniversal, we should note that they do contain some valuable
physical information. Imagine, for example, that one is interested in the behavior of our system
at finite temperature, and for zero chemical potential = 0. In such circumstances, the thermodynamic potential would be given entirely by the lowest term 0 () in (2.7) (at least if we
assume analyticity at = 0).
With such applications in mind, it is useful to separate the universal information in i () from
the nonuniversal terms. Even though i () are all divergent as we take the cutoff to infinity, we
can take additional derivatives with respect to , and obtain
0 () + 1 () + 2 ()
2 (0 ) 1
=
4
2
d ( )
=
2
cosh (/2) 2 8
3
tanh( ) cosh2 (/2)
(2.8)
In the last step, we have used the expression (2.4) for ( ), and evaluated the convergent integral
that multiplies the term. The remaining integral is now also convergent, and (2.8) determines
the dependence of 0 () on up to a (possibly divergent) nonuniversal polynomial linear in .
Similarly,
1
1
=

d ( )
cosh (/2)
= 0.
(2.9)
This integral vanishes because ( ) is an even function of , implying that 1 is independent of

, and equal to a (possibly divergent) nonuniversal constant.
As to the 2 term, its divergence comes completely from the linear divergence of , and we
can evaluate it without taking any additional derivatives,

2 =
2 2
d ( )
=
2
3 2
cosh2 (/2)
1

d
tanh( ) cosh2 (/2)
(2.10)
Despite being nonuniversal, the leading -dependent terms in (2.8) and (2.10) will play an important role in our discussion of the equivalence of various thermodynamic ensembles in the next
subsection.
2.4. Remarks on thermodynamic stability
In order to substantiate the discussion of the grand canonical ensemble, we must examine
whether this ensemble is well-defined, and if so, whether it is equivalent to the canonical and
the microcanonical ensemble. In order to test that, we will test the validity of the fluctuation
dissipation theorem. This theorem relates the fluctuations in the mean energy E
and the mean
number of particles N
to thermodynamic quantities: the specific heat and the isothermal compressibility. In terms of , we have

1 2
N 2 N
2 =
.
2
(2.11)
Similarly, the specific heat CV is the measure of the energy fluctuations,

CV
2 ( )
E 2 E
2 = 2
.
(2.12)
First of all, in order for the grand canonical ensemble to be well-defined, both of these quantities
need to be positive. Using (2.11) and (2.12) together with our evaluation of the leading divergent
terms in (2.8) and (2.10), we deduce that

1
N N
=
2 8
2
tanh( ) cosh2 [( )/2]
(2.13)
and

2
2
3
1
2
2
+
E E
=
.
2 3 2
8
tanh( ) cosh2 [( )/2]
(2.14)
As we see, if we discarded the nonuniversal -dependent pieces in (2.13) and (2.14) and kept
only the universal terms dependent on , both quantities would be negative, suggesting a possible
thermodynamic instability of the canonical as well as the grand canonical ensemble. The proper
behavior of the ensembles is restored by the leading nonuniversal term proportional to the cutoff
in (2.13) and (2.14). These terms are always positive, and stabilize both ensembles in the
double-scaling limit.
Having established the stability of the ensembles, it is also straightforward to verify that the
fluctuations of the mean values N
and E
are small in the thermodynamic limit, i.e., that
N 2
N
2
N
2
and
E 2
E
2
E
2
(2.15)
vanish. In the thermodynamic limit, we take while keeping and fixed, the denominators in (2.15) indeed diverge faster than (2.13) and (2.14), and all three ensembles are equivalent.
(In view of this, we can now drop the brackets in N
and E
, and denote them simply by N
and E.)
2.5. Expansion in the powers of the coupling 1/
In the matrix models of two-dimensional string theory, the inverse Fermi energy 1/ plays
the role of the string coupling, gs = 1/. The exact partition function can be expanded as an
10
asymptotic series4 in the powers of gs ,

Fstring
Fh (R)gs2h2
(2.16)
h=0
and compared to the perturbative string-theory calculation.

In the vacuum of noncritical M-theory, similarly, the inverse Fermi energy plays the role of
an M-theory coupling constant, which we denote by5
gM = 1/.
(2.17)
One can expand physical quantities in the powers of gM , and contrast the behavior of this expansion with the expansion (2.16) in two-dimensional string theory. In (2.16), the higher genus
coefficients Fh , with h > 1, are universal functions of the geometric moduli, such as the radius
R of the Euclidean time circle. In addition, the lowest genus terms F0 and F1 also depend logarithmically on and on the cutoff . Moreover, if open strings are present, odd powers of gs can
also appear in (2.16). In this sense, the string perturbation expansion is naturally an expansion
in halves of loops. More precisely, we wish to interpret (2.16) as a semiclassical expansion
of an effective action Seff /h in powers of h . The power of h counts by definition the number of
loops. The leading-order term in (2.16) is by definition classical, and therefore of order 1/h . The
one-loop term in Seff /h is independent of h . Thus, we see the well-known fact that the role of
h (or the loop-counting parameter) in string theory is effectively played by gs2 . Since the disk
diagram would be proportional to 1/gs , this diagram effectively contributes at one-half-loop
order compared to the classical term originating from the sphere 1/gs2 .
In the case of the vacuum energy of the M-theory vacuum state at zero temperature, this
expansion was studied in detail in [1], and found to take the form
FM
n3
Fn g M
.
(2.18)
n=0
This expansion exhibits several intriguing features [1]:

3 . In the semiclassical expansion, this will be the
The leading term in (2.18) is of order 1/gM
leading, classical term, implying that the role of the Planck constant in noncritical M-theory
3 . Since the subleading corrections found in [1] are integer powers of g ,
is played by h gM
M
the natural expansion is in thirds of loops, or h 1/3 gM . This is intriguingly reminiscent
of heterotic M-theory [32] in eleven dimensions.
Unlike in two-dimensional type 0A/0B string theory, where all orders in (2.16) are nontrivial,
it turns out that the perturbative expansion (2.18)in noncritical M-theory terminates at oneloop. All coefficients Fn with n > 3, i.e., all terms in (2.18) proportional to positive powers
of gM , are identically zero!
The instanton corrections to (2.18) were also determined in [1]. They have characteristic
nonperturbative weights
4 Throughout this paper, we use the symbol to denote exact asymptotic expansions of various formulas.
5 Folklore says that M-theory in eleven dimensions contains no dimensionless couplings. However, on a nontrivial
background geometry with a characteristic length scale R, a natural dimensionless coupling emerges as R measured in
Planck units. We believe that gM should be thought of as a dimensionless coupling of such a geometric origin. In fact, a
geometric interpretation for gM will indeed be found in Section 3.

2n
2n
exp
exp 1/3 ,
gM
h
11
(2.19)
with n any positive integer. According to a classic argument [33], nonperturbative effects
in string theory are stronger than in field theory, since they scale as exp(C/h 1/2 ) compared to the field-theory behavior exp(C/h ). By the same argument, (2.19) indicates that
nonperturbative effects in noncritical M-theory are even stronger than in string theory.
For each n, the instanton term (2.19) is multiplied by a perturbative expression similar to
(2.18). This expression starts at one-third loop order h 1/3 (i.e., the leading, classical term is
absent), and is also one-loop exact.
It is easy to show that the one-loop exactness of the asymptotic expansion (2.18) persists at
finite temperature as well. The expression (2.6) for the (third derivative of) the thermodynamic
potential can be expanded in the powers of 1/ at fixed R (after changing the integration variable
to = as in [1]). This yields

3
d
1
2n
(2.20)
fn (R) 2n e
.
sin( ) 1 +
n=1
We do not even need to determine the precise form of the R-dependent coefficients fn (R), because the integrals that they multiply all vanish identically, with the exception of the leading
constant term. Thus, the 1/ expansion (2.20) terminates at the lowest order,
3
1
+ nonperturbative terms,
3
2
(2.21)
which in turn implies that

1 3
+ F1 (R)2 + F2 (R) + F3 (R) + nonperturbative terms.
(2.22)
12
In addition to their dependence on the temperature, the subleading terms Fi , i = 1, 2, 3 may also
depend on the nonuniversal cutoff .
We have not checked explicitly the behavior of the perturbative expansions near the instanton
corrections, but we expect them to continue to be one-loop exact at finite temperature as well.
For the special case of R = 1, this will be explicitly verified in Section 2.9.
(, R)
2.6. The low-temperature expansion

We shall now analyze the thermodynamic behavior of noncritical M-theory in the regimes of
low and high temperature.
First, we expand as an asymptotic expansion at low temperature T = 1/(2R) at fixed :

3
2(22k1 1)B2k 2k
1
d sin( )
1
4
(2k)!
2R
3
sinh2 (/2)
0
k=1
1 1 (1 212k )(1)k1 B2k 2k
=
d
sin( ).
2
4
(2k)!
R 2k
2k
sinh (/2)
k=0
0
(2.23)
12
Bm are the Bernoulli numbers (see [1] for our conventions). The remaining integral in (2.23) can
be evaluated and the series formally resummed into the following compact expression,
1 2
R 2R
3
1
,
3
sin 2R tanh()
(2.24)
which should be interpreted in the sense of an asymptotic expansion in 1/R. Integrating (2.23)
once, we obtain
1 1 (1 212k )(1)k1 B2k 2k
+ .
2
2k
2k
2
(2k)!
2
R
tanh()
(2.25)
k=0
This result is in accord with the following general observation. The number of particles N is
related to via
N=
Taking a derivative of both sides, we obtain

eff (, )
N
2
,
=
where eff (, ) is the effective density of states at finite temperature. It is reassuring to see that
the leading term in (2.25) is indeed the zero-temperature density of states (), and reproduces
the zero-temperature result of [1].
Perhaps the most interesting feature of the low-temperature expansion (2.25) is the fact that all
dependence on T in (2.25) falls off rapidly as > 1. This suggests a radical reduction of the effective number of degrees of freedom in the theory at small values of the M-theory coupling gM .
This is compatible with the one-loop exactness of the weak-coupling expansion observed in [1]
and in the previous subsection. Both of these features suggest that the theory becomes effectively
topological for small values of gM , with a smooth crossover to a more dynamical regime at strong
coupling.
2.7. The high-temperature expansion
At high temperature, we can expand the thermodynamic potential in the powers of R,
3
R
3
2

0
d (1)m (2k 1)B2k 2m+1

( R)2m+2k .
sinh(/2)
(2m + 1)!
(2k)!
m=0
(2.26)
k=0
First, we wish to isolate the behavior of the system as T . At high temperature, the system
can be expected to become effectively classical, and we must find the correct way of identifying
this classical limit. Holding while taking the high-temperature limit R 0 in (2.26) would
lead to

R
3
+ O R3 .
2
3
(2.27)
13
Hence, in this naive limit of infinite temperature with fixed, 3 /3 would vanish. In order
to obtain a sensible classical limit at high temperatures, we must instead scale with R such that
R
(2.28)
is held fixed as R 0. This leads to a nontrivial high-temperature behavior,

3
1
3
2

2
1
d sin(
)
+O R
sinh(/2)

1
= tanh( )
(2.29)
+ O R2 ,
2
in which all orders in contribute at the leading order in T . Thus, the proper classical limit of
noncritical M-theory at high temperature involves holding fixed as T goes to infinity.
We are now ready to compare the high-temperature behavior of noncritical M-theory to the
high-temperature behavior in field and string theory.6
In quantum field theory, consider for illustration the case of YangMills gauge theory. The
high-temperature limit is formally governed by an effective field theory in one fewer spacetime
2
2 T . In order to obtain
dimension, with effective loop-counting coupling given by gYM,eff
= gYM
2
fixed while taking T to
the proper classical limit at high temperatures, we must keep gYM,eff
infinity. Of course, in most field theories, this naive picture will be strongly modified by renormalization effects and infrared divergences.
String theory at high temperature does not undergo an effective-dimensional reduction, and is
instead formally governed by a T-dual solution in the same dimension at the dual temperature.
The effective loop-counting coupling of the dual theory is given in terms of the original string
coupling by gs2 T 2 . In order to obtain a consistent classical limit at high temperatures, gs2 T 2 must
be held fixed as T goes to infinity. In this limit, as was pointed out in [14], the free energy of the
system scales as T 2 , i.e., as in a conformal field theory in 1 + 1 spacetime dimensions. In critical
string theory, this formal picture is strongly modified by the Hagedorn phase transition.
With this picture in mind, it is natural to ask the following questions in noncritical M-theory:
(1) What is the effective theory that governs the high-temperature behavior of the noncritical
M-theory vacuum?
(2) What is the leading behavior of the thermodynamic potential (or the free energy) at high
temperature?
(3) In the effective theory governing the high-temperature behavior, how does the effective coupling scale with T ?
We start answering these questions by isolating the leading high-temperature behavior of the
thermodynamic potential . As we have seen, is held fixed in the high-temperature limit. Upon
rewriting (2.29) in terms of and T , we obtain

3
1
3
(2.30)
= (2T ) tanh( )
+ ,
2
3
6 A nice discussion of the high-temperature classical limit in string theory and field theory can be found, for example,
in the classic paper by Atick and Witten [14].
14
where the denote all terms subleading in T . Thus, at high temperatures, the most divergent
term in the thermodynamic potential (and, consequently, also in the free energy) scales as
(T , )
T 3.
(2.31)
This high-temperature scaling of as T 3 is a behavior characteristic of a relativistic massless

field theory in 2 + 1 spacetime dimensions. This observed high-temperature behavior of the
noncritical M-theory vacuum lends further support to the arguments of [1] that the collective
excitations of the Fermi surface in this vacuum are effectively described by a bosonic collective
field in 2 + 1 dimensions.
At very high temperatures, the physics of the noncritical M-theory vacuum is governed by a
dual effective theory. We claim that this dual theory is another solution of noncritical M-theory
in 2 + 1 dimensions, effectively at zero temperature, and with the effective M-theory coupling
constant given by
g M = 1/.
(2.32)
This claim will be further substantiated by a duality argument in the next subsection. Here,
we present first evidence for this claim. (2.30) implies that the partition function ()
of the
effective dual theory that governs the classical limit at high temperature should satisfy
1
3
tanh( ).
2
3
(2.33)
This can be asymptotically expanded in the powers of g M ,

1
3
+ nonperturbative terms.
2
3
(2.34)
As a result, the leading term in the asymptotic expansion of ()

in the powers of 1/ is
proportional to 3 , and the asymptotic expansion is one-loop exact. As we saw in [1], and again
in Section 2.5 of this paper, this behavior is indeed a hallmark of noncritical M-theory in 2 + 1
dimensions.
We can gain further insight into the nature of this dual solution of M-theory as follows. As in
any other solution of the double-scaled Fermi liquid theory, the right-hand side of (2.33) should
be equal to the derivative of the effective density of states (
)
of this solution. Thus, integrating (2.29) once, we obtainup to a nonuniversal integration constantthe effective density
of states of the dual solution that represents the high-temperature limit of the M-theory vacuum,
(
)

1
log cosh( ).
(2.35)
Thus, even though this density of states exhibits the leading asymptotic behavior characteristic of
noncritical M-theory, it is strictly distinct from the density of states (2.4) of the original vacuum
at zero temperature, demonstrating that noncritical M-theory is not self-dual under the exchange
of the high and low temperature regions.
Since the effective dual theory at high temperatures is another solution of noncritical
M-theory, our arguments from Section 2.5 imply that the effective loop-counting coupling in
15
3 . This is related to the coupling g 3 of the original vacuum by

this solution is given by g M
M
3
3 3
g M
= (2)3 gM
T .
(2.36)
Thus, the effective loop-counting coupling in the high-temperature limit scales as T 3 . This behavior seems to be a novel signature of M-theory, to be contrasted with the behavior in quantum
field theory as well as string theory, as reviewed above. It would be very interesting to see whether
a similar T 3 scaling can also be found in full eleven-dimensional M-theory and its compactifications.
To summarize, the answers to the three questions about the high temperature behavior of
noncritical M-theory are as follows:
(1) In the high temperature limit, the M-theory vacuum is effectively described by another solution of noncritical M-theory, effectively at zero temperature, and with the density of states
given by (2.35).
(2) The free energy in the noncritical M-theory vacuum scales at high temperatures as T 3 , i.e.,
as in a massless field theory in 2 + 1 dimensions.
(3) The scaling (2.36) of the effective coupling constant with T is unlike in field or string theory,
suggesting that in this phase, noncritical M-theory is not (manifestly) equivalent to either.
2.8. Effective M-theory at high temperature from T-duality
In the previous subsection, we found signs indicating that the high-temperature limit of the
noncritical M-theory vacuum is effectively described by another solution of noncritical M-theory
in 2 + 1 dimensions. However, the exact nature of that solution was left somewhat obscure. Now
we will use T-duality of type 0A and 0B string theories to identify this dual solution of noncritical
M-theory.
As was reviewed in Section 2.1, the Hamiltonian of noncritical M-theory in the secondquantized fermionic representation can be expressed as an infinite sum of sectors with all possible
integer values q of the angular momentum J on the plane. For each fixed q, the Hamiltonian is
equivalent to that of the linear dilaton vacuum of type 0A string theory with RR flux equal to q.
In the vacuum state of noncritical M-theory, all available states are filled by fermions up to some
common value of the Fermi energy in all sectors independently of the value of q. Using this
decomposition, many physical quantities of the noncritical M-theory vacuum can be formally
evaluated as infinite sums of contributions from type 0A vacua with all possible values of the
RR flux q. For example, the vacuum energy of the vacuum solution equals the sum of vacuum
energies of type 0A vacua of all integer values of q, all filled up to the common value of the
Fermi energy . Similarly, the density of states (2.4) in M-theory can be evaluated as a sum over
q of densities of states in type 0A vacua at fixed q,
1
Re
( ) =
4

d e

1
1
Re d ei
e|q| .
=
2
2
sinh(
)
sinh (/2) qZ
1
(2.37)
Consider now noncritical M-theory at finite temperature T . Using (2.37), the thermodynamic
potential can be written as an infinite sum of contributions from sectors of fixed angular
momentum q. Recalling the integral formula (2.6), we can write
16
1
2R = Re
4

0
d i
1
1
e
sinh2 (/2) sinh[/(2R)]
1
=
Re
2
qZ
d i 1
1
e
e|q| .
sinh( ) sinh[/(2R)]
(2.38)
Each contribution of fixed q in (2.38) is equivalent to 2R evaluated in type 0A theory with

RR flux q, at temperature T .
In the canonical ensemble, type 0A theory at temperature T is described by the solution with
the Euclidean time compactified on a circle of radius R = 1/(2T ). It was shown in [27] that
this Euclidean solution of type 0A string theory is T-dual to a type 0B solution at the dual values
of the radius and of the inverse string coupling,7
1
R =
(2.39)
,
= R.
2R
In addition, this dual type 0B solution exhibits a nonzero value of the one-form RR flux, given
by [27]
=
iq
.
2R
(2.40)
the unit of the type 0B RR flux is quantized, and the flux becomes continuous in the
At finite R,
decompactification limit R .
In our expression (2.38) for the thermodynamic potential, we can now simultaneously perform
T-duality on all the type 0A string theories parametrized by the value of q. (R, ) is thus
mapped to

1
1
d i
1
2R =
e|q|/(2R)
Re
e
sinh(/2) sinh[/(2R)]
qZ

1
1
d i
1
.
= Re
(2.41)
e
sinh(/2) sinh2 [/(4R)]

In this way, the thermodynamic potential has been rewritten as an infinite sum over q of contri Indeed, with the help of [9] and
butions from type 0B vacua with RR flux equal to = iq/(2R).
[27], the individual terms in the sum in (2.41) can indeed be recognized as the partition functions
of the corresponding type 0B vacua with flux .
As we pointed out in Section 2.1, the Hamiltonian of noncritical M-theory can also be written
as an integral over a one-parameter family of type 0B theories parametrized by the continuous
eigenvalue of the single-particle Hamiltonian h2 of (say) the second of the two one-dimensional
the eigenvalues
oscillators. When we put the theory on a Euclidean time circle of radius R,
and the integral over turns into
become purely imaginary and quantized in the units of 1/R,
a discrete sum. This suggests that the eigenvalue of h2 in the Cartesian representation of the
Fermi liquid describing noncritical M-theory should be identified with the type 0B RR flux,
much like the type 0A RR flux q was identified with the eigenvalue of the angular momentum
J in the polar-coordinate representation of the theory. It also implies that noncritical M-theory
7 For some earlier work on T-duality in two-dimensional type 0A and 0B string vacua, see [3436].
17
inherits a T-duality symmmetry from the underlying T-duality between type 0A and 0B strings.
In noncritical M-theory, this T-duality swaps the Cartesian and polar-coordinate representations,
in which the theory is decomposed into a collection of type 0A or type 0B string theories will all
possible values of the corresponding RR flux.
In the infinite temperature limit, R goes to infinity, the type 0B flux becomes continuous,
and the sum in (2.41) turns into an integral. In this limit, the density of states of the dual M-theory
solution can be evaluated, and we obtain
(
)

2
1
log cosh( )
+ .
2
2
2

(2.42)
This reproduces our prediction (2.35) for the density of states of the effective high-temperature
theory.
Thus, our conclusions are as follows:
Noncritical M-theory inherits a T-duality symmetry from the underlying T-duality between
the type 0A and 0B strings.
The high-temperature regime of the M-theory vacuum is described by the T-dual solution of
noncritical M-theory, at the dual value of the temperature.
Having established that the high-temperature regime is related to the low-temperature regime
by T-duality, we can now return to the original task of Section 2.7 and study the high-temperature
expansion of the thermodynamic potential of the original M-theory vacuum. This is an expansion in the powers of R with fixed. From the perspective of the T-dual theory, this is a
low-temperature expansion in the powers of 1/R with fixed, analogous to the low-temperature
expansion of the original vacuum discussed in Section 2.6. We obtain
3
1
=
8R
3
1
2R 3

d 2 sin(
)
0

k=0
1
sinh2 ( R/2) sinh(/2)
(2k 1)B2k R 2k
(2k)!

d sin(
)
2k
sinh(/2)
2k
1
1 (2k 1)B2k R 2k
k
d sin(
)
(1)
3
2k
(2k)!
sinh(/2)
2R

k=0
1
2R

k=0
(2k 1)B2k
(2k)!
R 2k2
(1)k
2k
2k
tanh( ).
(2.43)
This can be summarized in a compact formula analogous to our low-temperature expansion formula (2.24),
2
3

(2.44)

tanh( ).
3
8R sin2 R2
The astute reader may recognize in (2.43) the strong resemblance to the partition function of
the A-model topological closed string theory on the resolved conifold (see, e.g., [37,38] for
18
reviews), with R essentially playing the role of the string coupling of the A-model, and related
to the Khler modulus of the conifold. This is our first hint of a much deeper relation between
noncritical M-theory compactified on a Euclidean time circle, and the strings of the topological
A-model. This relation will be the topic of Section 3 below.
2.9. Noncritical M-theory at the Debye temperature
It was noticed in [1] that the exact vacuum energy E0 () of the vacuum solution in noncritical
M-theory at zero temperature is mathematically equivalent to the partition function of Debye
phonons in a Debye crystal at temperature TD = 1/(2),
1
E0 () =
2

d D ( )

1
+
,
e2 1 2
1
(2.45)
with the Debye density of states D ( ) characteristic of a (2 + 1)-dimensional system. In a

slight detour from the main theme of this paper, we shall now study the partition function of the
noncritical M-theory vacuum at this Debye temperature.
The Debye temperature corresponds to the compactification radius R = 1 of the Euclidean
time dimension. In noncritical string theory, this particular value of R plays a special role, since
the theory at this radius is related to the Penner model [39].
At R = 1, the first derivative of the thermodynamic potential (2.3) of our noncritical M-theory
vacuum can be evaluated by residues, leading to

1 2 1
2
()
= +
tanh() +
+
.
(2.46)
4
4
2
2
This can be further integrated, yielding the following exact expression for (),
2

1
1 3 2
+ C()
() = + +
12
4
2
16

1
1
1
1
1 2
k
(1)
+ 2 2+
+ 3 3 e2k .
+
4
k
4k
k
2k
(2.47)
k=1
This formula nicely illustrates all of the features of the asymptotic expansion in the powers of
gM that were discussed in Section 2.5 above. In particular, () contains a perturbative series
3 . This
which is one loop exact and starts off with the classical contribution at order 3 = 1/gM
series is followed by an infinite series of instanton-like contributions. In each of the instanton
terms, the perturbative quantum corrections are also one-loop exact, with the leading classical
2 .
term vanishing, and the lowest nontrivial quantum correction starting at order 2 = 1/gM
Using (2.46), and dropping the nonuniversal terms, () can also be rewritten in another
interesting form, as

1
1
1
1
d 2 +
.
() =
(2.48)
2
4
e2 + 1 2
In this form, the universal part of the thermodynamic potential of noncritical M-theory at the
Debye temperature is mathematically equivalent to the energy of a system of fermions in 2 + 1
dimensions at the Debye temperature, and with the effective density of states D () 2 +
1/2. It is intriguing that this effective density of states is self-dual under an inversion of .
19
3. Duality to the closed string theory of the topological A-model

Our analysis of the partition function of noncritical M-theory at finite temperature in Section 2
has revealed a surprising connection with the amplitudes of the topological closed string of the
A-model on the resolved conifold. In this section, we will attempt to make this duality more
precise, and analyze some of its possible implications.
3.1. First comparison of the partition functions
In Section 2.8, we derived the asymptotic high-temperature expansion of the partition function
ZM of the noncritical M-theory vacuum,

(2k 1)B2k (iR)2k2 2k
3 log ZM
tanh( )
.
(2k)!
3
2k
(3.1)
k=0
This can be rewritten in terms of polylogarithms as

3
2

2
1
3 log ZM
2k2 (2) B2k
(2iR)
+
Li
Li2k e2 .
0
3
2
2k(2k 2)!

(iR) 2
k=1
(3.2)
We now wish to integrate this expression to get an expansion of log ZM . (3.2) determines log ZM
up to a second-order polynomial in ,
whose coefficients could be functions of R containing
nonuniversal dependence on the cutoff . The universal dependence of this polynomial on R
can be determined by taking derivatives of log ZM with respect to , in exact parallel with our
analysis of the R dependence of the polynomial (2.7) in Section 2.3. In the end, the nonuniversal
cutoff dependence afflicts only a few lowest-order terms in the double expansion in R and ,
and
we obtain

2
1
(2 )
3
e
log ZM
Li
p(
,
)
+
3
12
(2iR)2

1
+ C()
log 1 + e2
12
12

B2k
(2iR)2k2
Li32k e2 .
+
(3.3)
2k(2k 2)!
k=2
Here p(,
) is a nonuniversal polynomial of second order in ,
and C() is a nonuniversal
constant; both p(,
) and C() are independent of R.
In this form, the close similarity to the partition function of the closed topological string of
the A-model on the resolved conifold is apparent. Indeed, recall (or see in [28,29,37,38]) that
the partition function of the A-model on the resolved conifold can be written as an asymptotic
expansion in the powers of the string coupling constant gA , with coefficients being functions of
the Khler modulus tA ,

t

tA3
1
tA
1
tA
A
log ZA 2 pA (tA ) +
+ CA
Li3 e
log 1 e
12
24 12
gA

t
B2k
B2k B2k2
2h2
A
.
gA
+
Li32k e
+
(3.4)
2k(2k 2)(2k 2)! 2k(2k 2)!
h=2
20
Here pA (tA ) is a nonuniversal second-order polynomial in tA , and CA is a nonuniversal constant.

The real part of the Khler modulus tA measures the size of the S 2 of the resolved conifold,
and its imaginary part is given by the B-field flux through this S 2 . The tA -independent terms
in (3.4) come from constant maps from the string worldsheet to the target conifold, and the
polylogarithms come from worldsheet instantons.
A first comparison of (3.3) and (3.4) suggests the identification
gA = 2iR,
tA = 2 + i.
(3.5)
The matching between ZM and ZA is close, but there is an apparent difference: The constant
term in (3.4) at each order in the A-model string coupling is absent in the partition function of
the vacuum of noncritical M-theory.
We claim that this difference is purely due to the fact that the two partition functions ZA and
ZM are normalized differently. Indeed, notice that the partition function of the A-model comes
out naturally normalized so that the genus-h term in ZA diverges as tA22h , all subleading terms
going to zero as tA 0. This is in accord with the GopakumarVafa duality [28,29], which maps
the A-model to U (Nc ) ChernSimons gauge theory on S 3 . In order to check this behavior, we
can expand the polylogarithms around tA = 0, using the asymptotic expansion formula

(1)n (3 2k n) n
Li32k etA (2k 2)tA22k +
tA
n!
n=0
B2k2
(2k 3)!tA22k
+ O(tA ),
2k 2
k = 2, 3, . . . .
(3.6)
Thus, the polylogarithm terms in (3.4), which come from the sum over worldsheet instantons
of genus k, have a leading divergence tA22k , followed by a subleading constant term as tA
goes to zero. From the point of view of the GopakumarVafa duality, the leading divergence is
the nonperturbative term (discussed for example in [40]) which is not captured in the t Hooft
expansion of the dual ChernSimons theory. At each order in gA , the term originating from
the constant maps in (3.4) is precisely such that it subtracts the subleading constant from the
instanton polylogarithms, ensuring thatorder by order in gA log ZA vanishes at tA = 0 after
the leading tA22h divergence has been subtracted.
On the other hand, the partition function of noncritical M-theory is naturally normalized so
that it simplifies in a different limit, of . If it were not for the divergence in (2.3), setting
= in this expression would lead to ZM = 1. This is again intuitively clear, since sending
to infinity corresponds to emptying the Fermi sea; in the absence of any fermions, the partition
function should be equal to one. The divergence of log ZM makes the precise realization of
this formal expectation slightly subtle: One must empty the Fermi sea simultaneously with the
double-scaling limit (i.e., to set as ) in order to ensure that ZM = 1 as .
When we normalize ZM and ZA so that they are equal to one for the same (but otherwise
arbitrarily chosen) fixed value tA,fix of tA , they turn out to be equal as asymptotic series in tA . In
this sense, ZA and ZM carry precisely the same information, and we can summarize our result
in the following duality relation between the partition functions:
ZA (gA , tA )
ZM (R, )
=
,
ZA (gA , tA,fix ) ZM (R, fix )
with the parameters on the two sides related by (3.5) (in particular, tA,fix = 2 fix + i).
(3.7)
21
It is worth stressing again that since the A-model is only known as an asymptotic series in gA ,
the duality relation (3.7) is to be only interpreted as an equality between two asymptotic expansions. However, within the framework of noncritical M-theory, ZM is defined nonperturbatively
in R, and therefore represents a possible nonperturbative completion of the topological A-model.
3.2. A more general form of the duality
In the thermodynamic ensemble, the partition function corresponds to fermions antiperiodic
around the Euclidean time dimension tE . Since the thermodynamic interpretation is no longer
important in this section, we shall relax this condition, and allow for an arbitrary phase b in the
periodicity of the fermions,
(tE + 2R, 1 , 2 ) = eib (tE , 1 , 2 ).
(3.8)
The case of the thermal ensemble studied in Section 2 corresponds to b = .

How does the general phase appear in the partition function? In the first-quantized framework
that we used to calculate log Z in (2.3), allowing for the general phase in (3.8) amounts to the
insertion of an extra factor of (eib )F , with F the fermion number. (The minus sign is due to
the fact that (2.3) calculates log Z for the thermal case of b = .) Hence, the partition function
with arbitrary periodicity of the fermions becomes

log ZM (R, ,
b) =

.
d ( ) log 1 eib e2 2R
(3.9)
Expanding this in the asymptotic series in R, we obtain

2 ib

1
(2 + ib)3
log ZM
e
Li
p(
,
b,
)
+
3
12
(2iR)2

2 + ib
1
+ C(b, ) +
+
log 1 e2 ib
24
12

B2k
.
(2iR)2k2
Li32k e2 ib
2k(2k 2)!
(3.10)
k=2
This again matches log ZA modulo normalization, but now the entire range of complex values of
the Khler modulus tA is allowed, with the phase b playing the role of the imaginary part of tA .
Recalling that in the original variables is equal to R, and that gM = 1/ plays the role of
the coupling constant in the noncritical M-theory vacuum, we find the duality relations
gA = 2iR,
tA = 2 + ib =
2R
+ ib
gM
(3.11)
advertized in the introduction. With this matching of the parameters, the partition function are
related by a generalization of (3.7) to the case of arbitrary b,
ZA (gA , tA )
b)
ZM (R, ,
=
.
ZA (gA , tA,fix ) ZM (R, fix , bfix )
(3.12)
We have again chosen an arbitrary normalization point tA,fix , with fix and bfix related to tA,fix
via (3.11).
22
Fig. 3. The solution of noncritical M-theory with a universal bottom of the Fermi sea. In this solution, all single-particle
states are occupied between the top of the Fermi sea at + and the bottom of the sea at . (As in Fig. 1, the angular
dimension of the spatial plane has been suppressed.)
3.3. Solutions with the universal bottom of the Fermi sea

As we have just seen, the partition functions of the vacuum of noncritical M-theory and of
the topological A-model on the resolved conifold are virtually identical, differing only in their
normalization properties. Here we wish to point out the existence of a larger family of solution
of noncritical M-theory, whose partition functions are naturally normalized in a way similar to
that of the partition function of the A-model.
The conventional vacuum of noncritical M-theory corresponds to a large N limit of the system of N fermions occupying all single-particle states from the cutoff up to the top of the
Fermi sea at the double-scaled Fermi energy . Recall that in Section 7.4 of [1], we studied a
broader class of noncritical M-theory solutions, characterized by two Fermi surfaces: The universal top of the sea at some = + , and the universal bottom of the sea at some > +
(see Fig. 3). This construction gives a two-parameter family of static solutions of noncritical
M-theory, parametrized by .
Consider now this family of solution in Euclidean signature, with the Euclidean time compactified on a circle of radius R, and define = 2R . The special case of particular interest
to us is
+ = ,
= 0.
(3.13)
Thus, the Fermi sea is filled from the top of the potential at zero, to some Fermi energy above
the top of the potential.
The partition function ZM (R, )
of this solution of noncritical M-theory no longer approaches
one at . Instead, it simplifies in the limit of 0, which corresponds to emptying the
Fermi sea. It would be interesting to study this extended class of solutions of noncritical M-theory
for general values of , and identify their relation to the topological A-model. This is, however,
beyond the scope of this paper.
4. Consequences of the duality
The proposed duality between noncritical M-theory and closed string theory of the topological
A-model exhibits several notable features, and can be expected to have interesting implications
for a wide range of phenomena traditionally related to the topological A-model. Here we present
some comments and raise some questions about the possible implications of this duality.
23
As we saw in Section 3.2, the parameters on the two sides of the duality are related as follows,
gA = 2iR,
tA =
2R
+ ib.
gM
(4.1)
In particular, the duality relates the coupling constant gM of noncritical M-theory to the Khler
modulus of the resolved conifold on the dual side. Thus, the dimensionless coupling constant of
noncritical M-theory acquires a geometric interpretation, as promised in Section 2.5.
From the point of view of the dual A-model, the M-theory coupling constant gM plays the role
of a worldsheet coupling. In [1] (and again in Section 2.5), we found the characteristic leading
3 of the free energy of the noncritical M-theory vacuum at weak coupling. Our
behavior 1/gM
3 N 3 in the
duality (4.1) provides an intriguing explanation for this leading power of 1/gM
large-N behavior of noncritical M-theory, by relating it to the triple intersection form of the
CalabiYau, which gives a tree-level contribution in the dual A-model [37,38].
Topological M-theory for the A-model:
The duality (4.1) relates the string coupling constant gA of the A-model to a geometric parameterthe radius of the Euclidean time dimensionon the dual side of noncritical
M-theory. Hence, the A-model string coupling gA acquires a geometric interpretation, as the size
of the M-theory circle. Such a relation gA R between the string coupling and the radius of
the M-theory circle is one of the hallmarks of the string theory/M-theory duality in critical string
theory. This suggests that noncritical M-theory could be a realization of the idea of topological
M-theory, as proposed for closed topological string theory of the A-model on CalabiYau in
[25] (see also [26]).
In this context, it is intriguing that the role of the M-theory dimension is played by the
Euclidean time dimension of noncritical M-theory. One motivation for introducing topological
M-theory was to explain the apparent behavior of the partition function of the A-model as a wave
function of a system evolving in an extra time dimension. Perhaps, the factor of i in the relation
(4.1) between the string coupling gA and the radius R is an indication that the duality should be
more naturally interpreted after the Wick rotation to real time, effectively replacing R with iR
and reinterpreting R as the time lapse in real time evolution.
Relation to ChernSimons theory on S 3 :
GopakumarVafa duality [28,29] relates the A-model on the resolved conifold to the U (Nc )
ChernSimons gauge theory on S 3 . The GopakumarVafa relations are
gA =
2i
,
k + Nc
tA =
2iNc
,
k + Nc
(4.2)
where k is the level of the ChernSimons theory.

When combined with the GopakumarVafa duality, our duality implies a relation between
noncritical M-theory and U (Nc ) ChernSimons gauge theory on S 3 . For the quantities in noncritical M-theory, this implies
R=
1
,
k + Nc
ib
= iNc .
2R
(4.3)
These relations have a very natural interpretation, with R being the parameter of the large-Nc
2 . Moreover, if we choose to perform the Wick
expansion, i.e., the gauge coupling constant gCS
rotation to real time, (4.3) implies an intriguing relation gM = 1/Nc , identifying the M-theory
coupling constant with the inverse number of colors in the dual ChernSimons gauge theory.
24
Note also that since the ChernSimons theory can be reinterpreted as a matrix model [41], this
connection leads to a matrix model description of noncritical M-theory.
Nonperturbative completion of the A-model:
Since the partition function of the topological strings of the A-model is defined only as an asymptotic expansion in gA , relations between partition functions implied by (3.7) are understood
as equalities between asymptotic expansions. However, the partition function ZM of noncritical
M-theory is nonperturbatively defined, and therefore (3.7) can be interpreted as a nonperturbative
completion of the asymptotic series for the A-model.
Another nonperturbative completion of the A-model has emerged recently in the context of
the OSV conjecture [42]. This conjecture relates the entropy of certain supersymmetric black
holes to the absolute value squared of a topological string partition function, and its precise form
is still being developed. It should be interesting to investigate the relation between the nonperturbative completions of the A-model via the OSV conjecture and via the duality to noncritical
M-theory. In particular, the OSV conjecture has been tested in [43,44] for the topological string
on a noncompact CalabiYau given by the O(p) O(p + 2g 2) fibration over a genus-g
Riemann surface g . The partition functions are related to the large Nc , q-deformed version of
YangMills on g . In turn, two-dimensional large-Nc YangMills theory on g is known to
have an interpretation in terms of a string theory [45], whose worldsheet description is given by
the topological rigid string [46] (see also [47]). The case of p = 1 and g = 0 is rather singular,
and somewhat outside of the scope of [44]. In particular, the large-Nc YangMills theory on S 2
is known to exhibit the DouglasKazakov phase transition [48].8
However, p = 1 and g = 0 is the resolved conifold, for which the duality to noncritical M-theory represents a nonperturbative completion of the A-model, and could therefore be
complementary to the methods of [44]. Note in particular that the partition function (2.3) of
noncritical M-theory exhibits a singular behavior at b = 0, with each term of the asymptotic expansion (3.10) divergent at 2 + ib = 0. We expect this singular behavior to be related to the
DouglasKazakov phase transition of YangMills theory on S 2 .
S-duality of the A-model from T-duality of noncritical M-theory:
As we saw in Section 2, noncritical M-theory can be formally decomposed into an infinite
number of sectors equivalent to type 0A or 0B theory with all possible values of their RR flux.
When compactified on the thermal circle, noncritical M-theory inherits a duality symmetry from
the T-duality between the type 0A and 0B string theories. This duality was used in Section 2.8 to
shed light on the high-temperature behavior of noncritical M-theory. It acts on the parameters of
the vacuum by
1
.
R =
(4.4)
2R
When combined with the duality (4.1) to the topological A-model, this T-duality predicts the
existence of an S-duality for the A-model. Indeed, in terms of the A-model variables, (4.4) acts
by inverting the string coupling constant and rescaling the Khler modulus,
= R,
gA
1
,
gA
tA tA gA ,
i.e., as an S-duality transformation.

8 In the context of the OSV conjecture, this phase transition has been recently studied in [49].
(4.5)
25
An S-duality for the A-model on a CalabiYau has been proposed in the literature [50,51]. It
also acts by (4.5), and relates the A-model to the B-model on the same target and with its string
coupling given by gB 1/gA . It should be interesting to investigate the relationship between the
S-duality suggested by noncritical M-theory and the S-duality proposed in [50,51]. The latter has
been related to T-duality in critical string theory in [52].
The melting crystal and quantum foam:
At infinite tA , string theory of the topological A-model has been related to a classical crystal
melting problem [30]. The inverse temperature of that crystal is proportional to gA . In combination with our duality, (4.1) shows that the temperature of the melting crystal is essentially the
temperature of the thermal ensemble in the noncritical M-theory vacuum. Moreover, in [1], we
evaluated the exact vacuum energy of the noncritical M-theory vacuum, and found a surprising
similarity with the partition function of phonons in a Debye crystal, with related to the size of
this crystal. (4.1) relates to the Khler modulus of the conifold, which was set to infinity in
[30], corresponding to an infinite crystal in [30]. The case of finite tA was studied in [53], with
indeed providing a scale related to the size of the crystal. Given this matching of the physical
interpretation of the parameters, it is natural to suspect that the constituents of the spacetime
foam crystal of [30] are intimately related to the constituent fermions of noncritical M-theory.
Embedding into M-theory:
The relation between noncritical M-theory and the topological A-model was found in Section 3 by comparing the partition functions of the two theories. It would clearly be desirable
to have a more systematic derivation of such a duality, by embedding noncritical M-theory to
superstring theory or eleven-dimensional M-theory.
In the related case of the GopakumarVafa duality, such a derivation exists [54]. It takes advantage of a relation between topological strings on CalabiYau and M-theory on a noncompact
G2 holonomy manifold. From the eleven-dimensional vantage point of M-theory, the duality is
a simple consequence of the flop transition. In addition, G2 holonomy manifolds are also central ingredients in the conjectured relation of topological strings on CalabiYau to topological
M-theory in seven dimensions.
We expect M-theory on G2 holonomy manifolds to be relevant for the duality between noncritical M-theory and the topological A-model as well. Indeed, in the case of the Gopakumar
Vafa duality [54], the G2 holonomy 7-manifold in question is given by a quadratic equation
in C4 ,
|z1 |2 + |z2 |2 |z3 |2 |z4 |2 = 2.
(4.6)
This should be compared to the Fermi surface of the vacuum state in noncritical M-theory. In
the double scaling limit, the Fermi liquid describing noncritical M-theory in 2 + 1 dimensions
becomes semiclassical. The semiclassical Fermi surface is a hypersurface in the classical phase
space parameterized by coordinates 1 , 2 and momenta p1 , p2 . In particular, the vacuum state
of noncritical M-theory (at zero temperature) is described by the Fermi surface satisfying9
p12 + p22 21 22 = 2.
(4.7)
Thus, the Fermi surface of the noncritical M-theory vacuum corresponds to the intersection of
the G2 holonomy 7-manifold (4.6) with the real section R4 C4 ! This is very similar to the
9 For other solutions of the classical equations of motion of the Fermi surface in noncritical M-theory, see [1].
26
embedding of two-dimensional noncritical string theories into full string theory [55,56]. Interestingly, the flop transition of the G2 holonomy manifold is realized by , which in (4.7)
corresponds precisely to the particle-hole duality in noncritical M-theory.
5. Conclusions
Noncritical M-theory in 2 + 1 dimensions represents a unifying framework, in which various
string theories are related by dualities, in a controlled setting of an exactly solvable model. In this
paper, we have seen further evidence supporting this picture. In particular, we have found that
noncritical M-theory is related to string theories in at least two different ways, via two distinct
forms of M-theory/string theory duality.
When compactified on the Euclidean time circle, noncritical M-theory in 2 + 1 dimensions
has two U (1) symmetries: the rotations of the spatial eigenvalue plane, and the translations of
the Euclidean time circle. The reduction on the angular S 1 on the eigenvalue plane is related
to noncritical c = 1 type 0A theory. In this correspondence, the KaluzaKlein charge of the
reduction (i.e., the angular momentum on the eigenvalue plane) is identified with the RR charge
of stable D0-branes in type 0A string theory. This was the original picture that led to the definition
of noncritical M-theory in terms of a Fermi liquid in [1].
In Section 3 of this paper, we have seen evidence that the reduction of noncritical M-theory
along the Euclidean time circle is also related to a string theory, namely the closed string theory
of the topological A-model on the resolved conifold. In this correspondence, the radius R of the
Euclidean time S 1 is interpreted as the string coupling gA of the A-model.
Among the most intriguing features of noncritical M-theory (and of its cousins, matrix models
of noncritical string theories) is the fact that the physical spacetime emerges as a derived concept, associated to the existence and geometrical properties of the Fermi surface. In this emergent
picture of spacetime, the elementary fermions of the Fermi liquidoriginating from D0-branes
in the underlying string theoryare the fundamental constituents, and the smooth macroscopic
effective geometry of spacetime is a collective phenomenon. The rigid eigenvalue plane populated by the fermions is an auxiliary structure, only rather indirectly related to the physical space.
Thus, the exactly solvable setting of noncritical M-theory seems particularly suitable for extracting more lessons about the emergence and microscopic constituents of spacetime in string and
M-theory.
Acknowledgements
We wish to thank Mina Aganagic, Ofer Aharony, Sumit Das, Eric Gimon and Aleksey Mints
for useful discussions. This material is based upon work supported by NSF grant PHY-0244900,
DOE grant DE-AC02-05CH11231, an NSF Graduate Research Fellowship, and the Berkeley
Center for Theoretical Physics.
References
[1] P. Horava, C.A. Keeler, Noncritical M-theory in 2 + 1 dimensions as a nonrelativistic Fermi liquid, hep-th/0508024.
[2] Y. Nakayama, Liouville field theoryA decade after the revolution, hep-th/0402009.
[3] E.J. Martinec, The annular report on non-critical string theory, hep-th/0305148;
E.J. Martinec, Matrix models and 2D string theory, hep-th/0410136.
[4] S. Alexandrov, Matrix quantum mechanics and two-dimensional string theory in non-trivial backgrounds, hepth/0311273.
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
27
S.R. Das, Non-trivial 2d spacetimes from matrices, hep-th/0503002.

J. Polchinski, What is string theory?, hep-th/9411028.
I. Klebanov, String theory in two dimensions, hep-th/9108019.
P. Ginsparg, G. Moore, Lectures on 2D gravity and 2D string theory, hep-th/9304011.
M.R. Douglas, I.R. Klebanov, D. Kutasov, J. Maldacena, E. Martinec, A new hat for the c = 1 matrix model,
hep-th/0307195.
T. Takayanagi, N. Toumbas, A matrix model dual of type 0B string theory in two dimensions, JHEP 0307 (2003)
064, hep-th/0307083.
D.J. Gross, P. Mende, The high-energy behavior of string scattering amplitudes, Phys. Lett. B 197 (1987) 129;
D.J. Gross, P. Mende, String theory beyond the Planck scale, Nucl. Phys. B 303 (1988) 407.
D. Amati, M. Ciafaloni, G. Veneziano, Superstring collisions at Planckian energies, Phys. Lett. B 197 (1987) 81;
D. Amati, M. Ciafaloni, G. Veneziano, Classical and quantum gravity effects from Planckian energy superstring
collisions, Int. J. Mod. Phys. A 3 (1988) 1615.
D.J. Gross, High-energy symmetries of string theory, Phys. Rev. Lett. 60 (1988) 1229;
E. Witten, Spacetime and topological orbifolds, Phys. Rev. Lett. 61 (1988) 670;
D. Amati, M. Ciafaloni, G. Veneziano, Can spacetime be probed below the string size?, Phys. Lett. B 216 (1989)
41.
J.J. Atick, E. Witten, The Hagedorn transition and the number of degrees of freedom of string theory, Nucl. Phys.
B 310 (1988) 334.
G.G. Athanasiu, J.J. Atick, Remarks on thermodynamics of strings, in: S.J. Gates, et al. (Eds.), Strings 88: Proceedings, World Scientific, 1989.
P. Horava, C.A. Keeler, Strings on AdS2 and the high-energy limit of noncritical M-theory, in preparation.
E. Alvarez, M.A.R. Osorio, Superstrings at finite temperature, Phys. Rev. D 36 (1987) 1175.
M. McGuigan, Finite-temperature string theory and twisted tori, Phys. Rev. D 38 (1988) 552.
M.J. Bowick, L.C.R. Wijewardhana, Superstrings at high temperature, Phys. Rev. Lett. 54 (1985) 2485;
M.J. Bowick, S.B. Giddings, High-temperature strings, Nucl. Phys. B 325 (1989) 631.
N. Deo, S. Jain, C.-I. Tan, Strings at high-energy densities and complex temperature, Phys. Lett. B 220 (1989) 125;
N. Deo, S. Jain, C.-I. Tan, The ideal gas of strings, in: S. Das, et al. (Eds.), Modern Quantum Field Theory: Proceedings, World Scientific, 1991.
K.L. Kowalski, et al. (Eds.), Thermal Field Theories and Their Applications, Physica A 158 (1) (1989) 1.
J.G. Russo, Free energy and critical temperature in eleven dimensions, hep-th/0101132.
R. Easther, B.R. Greene, M.G. Jackson, D. Kabat, Brane gases in the early universe: Thermodynamics and cosmology, hep-th/0307233.
O. Bergman, M. Gaberdiel, Dualities of type 0 strings, JHEP 9907 (1999) 022, hep-th/9906055.
R. Dijkgraaf, S. Gukov, A. Neitzke, C. Vafa, Topological M-theory as unification of form theories of gravity, hepth/0411073.
N. Nekrasov, Z-theory, hep-th/0412021.
J. Maldacena, N. Seiberg, Flux-vacua in two-dimensional string theory, hep-th/0506141.
R. Gopakumar, C. Vafa, On the gauge theory/geometry correspondence, Adv. Theor. Math. Phys. 3 (1999) 1415,
hep-th/9811131.
R. Gopakumar, C. Vafa, M-theory and topological strings, I, hep-th/9809187;
R. Gopakumar, C. Vafa, M-theory and topological strings, II, hep-th/9812127.
A. Okounkov, N. Reshetikhin, C. Vafa, Quantum CalabiYau and classical crystals, hep-th/0309208;
A. Iqbal, N. Nekrasov, A. Okounkov, C. Vafa, Quantum foam and topological strings, hep-th/0312022.
D.J. Gross, I.R. Klebanov, One-dimensional string theory on a circle, Nucl. Phys. B 344 (1990) 475.
P. Horava, E. Witten, Heterotic and type I string dynamics from eleven dimensions, Nucl. Phys. B 460 (1996) 5,
hep-th/9510209;
P. Horava, E. Witten, Eleven-dimensional supergravity on a manifold with boundary, Nucl. Phys. B 475 (1996) 94,
hep-th/9603142.
S. Shenker, The strength of nonperturbative effects in string theory, in: Random Surfaces and Quantum Gravity,
Cargse Proceedings, 1990.
D.J. Gross, J. Walcher, Non-perturbative RR potentials in the c = 1 matrix model, hep-th/0312021.
S. Gukov, T. Takayanagi, N. Toumbas, Flux backgrounds in 2D string theory, hep-th/0312208.
U.H. Danielsson, M.E. Olsson, M. Vonk, Matrix models, 4D black holes and topological strings on non-compact
CalabiYau manifolds, hep-th/0410141.
28
[37] M. Mario, ChernSimons theory and topological strings, Rev. Mod. Phys. 77 (2005) 675, hep-th/0406005;
M. Mario, Enumerative geometry and knot invariants, hep-th/0210145;
M. Mario, Les Houches lectures on matrix models and topological strings, hep-th/0410165.
[38] A. Neitzke, C. Vafa, Topological strings and their physical applications, hep-th/0410178.
[39] J. Distler, C. Vafa, A critical matrix model at c = 1, Mod. Phys. Lett. A 6 (1991) 259.
[40] H. Ooguri, C. Vafa, Worldsheet derivation of a large N duality, hep-th/0205297.
[41] M. Mario, ChernSimons theory, matrix integrals, and perturbative three-manifold invariants, Commun. Math.
Phys. 253 (2004) 25, hep-th/0207096;
M. Aganagic, A. Klemm, M. Mario, C. Vafa, Matrix model as a mirror of ChernSimons theory, JHEP 0402
(2004) 010, hep-th/0211098.
[42] H. Ooguri, A. Strominger, C. Vafa, Black hole attractors and the topological string, Phys. Rev. D 70 (2004) 106007,
hep-th/0405146.
[43] C. Vafa, Two-dimensional YangMills, black holes and topological strings, hep-th/0406058.
[44] M. Aganagic, H. Ooguri, N. Saulina, C. Vafa, Black holes, q-deformed 2d YangMills, and non-perturbative topological strings, hep-th/0411280.
[45] D.J. Gross, Two-dimensional QCD as a string theory, Nucl. Phys. B 400 (1993) 161, hep-th/9212149;
D.J. Gross, W. Taylor, Two-dimensional QCD is a string theory, Nucl. Phys. B 400 (1993) 181, hep-th/9301068;
D.J. Gross, W. Taylor, Twists and Wilson loops in the string theory of two-dimensional QCD, Nucl. Phys. B 403
(1993) 395, hep-th/9303046.
[46] P. Horava, Topological strings and QCD in two dimensions, in: Quantum Field Theory and String Theory, Cargse
Proceedings, 1993, hep-th/9311156;
P. Horava, Topological rigid string theory and two-dimensional QCD, Nucl. Phys. B 463 (1996) 238, hepth/9507060;
P. Horava, On QCD string theory and AdS dynamics, JHEP 9901 (1999) 016, hep-th/9811028.
[47] S. Cordes, G. Moore, S. Ramgoolam, Large N 2D YangMills theory and topological string theory, Commun. Math.
Phys. 185 (1997) 543, hep-th/9402107.
[48] M.R. Douglas, V.A. Kazakov, Large N phase transition in continuum QCD2 , Phys. Lett. B 319 (1993) 219, hepth/9305047.
[49] X. Arsiwalla, R. Boels, M. Mario, A. Sinkovics, Phase transitions in q-deformed 2d YangMills theory and topological strings, hep-th/0509002;
D. Jafferis, J. Marsano, A DK phase transition in q-deformed YangMills on S 2 and topological strings, hepth/0509004;
N. Caporaso, M. Cirafici, L. Griguolo, S. Pasquetti, D. Seminara, R.J. Szabo, Topological strings and large N phase
transitions I, hep-th/0509041;
transitions II, hep-th/0511043.
[50] A. Neitzke, C. Vafa, N = 2 strings and the twistorial CalabiYau, hep-th/0402128.
[51] N. Nekrasov, H. Ooguri, C. Vafa, S-duality and topological strings, JHEP 0410 (2004) 009, hep-th/0403167.
[52] A. Kapustin, Gauge theory, topological strings, and S-duality, JHEP 0409 (2004) 034, hep-th/0404041.
[53] T. Okuda, Derivation of CalabiYau crystals from ChernSimons gauge theory, JHEP 0503 (2005) 047, hepth/0409270.
[54] M. Atiyah, J. Maldacena, C. Vafa, An M-theory flop as a large N duality, J. Math. Phys. 42 (2001) 3209, hepth/0011256.
[55] D. Ghoshal, C. Vafa, c = 1 string and the topological theory of the conifold, Nucl. Phys. B 453 (1995) 121, hepth/9506122.
[56] M. Aganagic, R. Dijkgraaf, A. Klemm, M. Mario, C. Vafa, Topological strings and integrable hierarchies, hepth/0312085.
Minimally fine-tuned supersymmetric standard models

with intermediate-scale supersymmetry breaking
Yasunori Nomura, David Poland, Brock Tweedie
Department of Physics, University of California, Berkeley, CA 94720, USA
Theoretical Physics Group, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA
Received 28 September 2005; received in revised form 25 February 2006; accepted 31 March 2006
Available online 18 April 2006
Abstract
We construct realistic supersymmetric theories in which the correct scale for electroweak symmetry
breaking is obtained without significant fine-tuning. We consider two classes of models. In one class supersymmetry breaking is transmitted to the supersymmetric standard model sector through Dirac gaugino
mass terms generated by a D-term vacuum expectation value of a U (1) gauge field. In the other class
the supersymmetry breaking sector is separated from the supersymmetric standard model sector in an
extra dimension, and the transmission of supersymmetry breaking occurs through gauge mediation. In
both these theories the Higgs sector contains two Higgs doublets and a singlet, but unlike the case for
the next-to-minimal supersymmetric standard model the singlet field is not responsible for generating the
supersymmetric or supersymmetry breaking mass for the Higgs doublets. These masses, as well as the mass
for the singlet, are generated through gravitational-strength interactions. The scale at which the squark and
slepton masses are generated is of order 1100 TeV, and the generated masses do not respect the unified
mass relations. We find that electroweak symmetry breaking in these theories is caused by an interplay between the topstop radiative correction and the holomorphic supersymmetry breaking mass for the Higgs
doublets and that the fine-tuning can be reduced to the level of 20%. The theories have rich phenomenology,
including a variety of possibilities for the lightest supersymmetric particle.
* Corresponding author.
E-mail address: brock@berkeley.edu (B. Tweedie).

doi:10.1016/j.nuclphysb.2006.03.034
30
Y. Nomura et al. / Nuclear Physics B 745 (2006) 2948
1. Introduction
Weak scale supersymmetry is an attractive candidate for physics beyond the standard model.
It not only stabilizes the Higgs potential against potentially large radiative corrections, but also
leads to an elegant picture of gauge coupling unification at a scale MX 1016 GeV [1]. Nondiscovery of the superparticles or the Higgs boson at LEP II, however, puts strong constraints on
how supersymmetry can be realized at the weak scale. In the minimal supersymmetric standard
model (MSSM), evading the lower bound on the physical Higgs boson mass typically requires
a large top squark mass. This in turn gives a large radiative correction to the Higgs boson masssquared parameter, leading to fine-tuning in electroweak symmetry breaking. In fact, the problem
of fine-tuning is somewhat generic in theories with weak scale supersymmetry, and is called the
supersymmetric fine-tuning problem.
Recently, a general framework has been discussed for supersymmetric theories that avoid
fine-tuning while preserving the successful features of supersymmetry [2,3]. The key point is
to lower both the top squark masses and the mediation scale of supersymmetry breaking, by
violating simple unified mass relations.1 This makes radiative corrections to the Higgs masssquared parameter small, and thus reduces fine-tuning. A simple way to accommodate such
light top squarks is to introduce an additional contribution to the Higgs quartic couplings other
than that from the SU(2)L U (1)Y D-terms of the MSSM (for theories giving such a contribution, see, e.g., [411]).2 In Refs. [2,3] these were achieved by adopting the mechanism of
[14]: the dynamical supersymmetry breaking sector has a global SU(5) symmetry, of which the
SU(3) SU(2) U (1) subgroup is explicitly gauged and identified as the standard model gauge
group, but this SU(5) is spontaneously broken to the gauged subgroup at the dynamical scale of
10100 TeV. This structure allows us to accommodate the successful prediction associated
with gauge coupling unification at the leading-log level, while the unwanted unified mass relations among the squark and slepton masses are avoided. The mediation scale of supersymmetry
breaking is very low and of order , which is the scale of dynamical supersymmetry breaking itself. The Higgs sector superpotential is generated from interactions between the Higgs and
dynamical supersymmetry breaking sectors through marginal operators.
In this paper we construct classes of explicit supersymmetric standard models
in which the
fundamental scale of supersymmetry breaking is an intermediate scale, MI = mweak MPl , and
yet the supersymmetric fine-tuning problem is ameliorated. An advantage of intermediate-scale
supersymmetry breaking is that the Higgs sector superpotential is obtained relatively easily
through non-renormalizable interactions suppressed by the Planck scale [15]. To implement this
mechanism without introducing the supersymmetric flavor problem, we consider two classes
of theories. In one class supersymmetry breaking is transmitted to the supersymmetric standard model sector through a D-term vacuum expectation value (VEV) of a U (1) gauge field
[16]. In the other class the supersymmetry breaking sector is separated from the supersymmetric
standard model sector in an extra dimension, and the transmission of supersymmetry breaking
1 The unified mass relations need not be violated if the top quark and Higgs boson are both rather heavy, m
t
180182 GeV and MHiggs 200250 GeV [3]. We do not consider this scenario in this paper.
2 A solution to the supersymmetric fine-tuning problem that does not require an extension of the Higgs sector at the
weak scale has recently been presented in [12], where the supersymmetry breaking mass for the up-type Higgs boson
is suppressed by a cancellation between two different contributions [13], and a large stop mixing parameter and a small
holomorphic supersymmetry breaking Higgs mass ensure successful electroweak symmetry breaking with relatively
small top squark masses.
31
occurs through gauge mediation [17]. In both these theories the Higgs sector contains two Higgs
doublets and a singlet. Our Higgs sector superpotential, however, differs from that of the nextto-minimal supersymmetric standard model: it contains weak-scale mass parameters which are
naturally generated through gravitational-strength interactions. The scale at which the squark and
slepton masses are generated is of order 1100 TeV, and the generated masses do not respect the
unified mass relations. These features nicely meet the general criteria discussed above. (Note
that the relevant scale for the fine-tuning argument is the scale at which the squark and slepton
masses are generated, and not the one at which the gaugino masses are generated.)
Electroweak symmetry breaking in our theories occurs because of an interplay between the
topstop radiative correction and the holomorphic supersymmetry breaking mass squared for the
Higgs doublets, the B term. The dynamics of the singlet field S almost decouples from the
electroweak symmetry breaking physics due to a relatively large supersymmetric mass for S.
These theories, therefore, naturally realize the scenario II of B-driven electroweak symmetry
breaking discussed in Ref. [18]. This has an advantage, compared with theories based on the nextto-minimal supersymmetric standard model (NMSSM), that there is no strict requirement on the
potential that the correct supersymmetric Higgs mass term ( term) should be reproduced by the
VEV of the singlet field, thus opening up a larger region of parameter space that correctly breaks
the electroweak symmetry. The and B terms of order the weak scale are naturally generated
in our theories through gravitational-strength interactions. (For attempts of reducing fine-tuning
in the context of the NMSSM, see, e.g., [19].) In the present scheme, the amount of fine-tuning
is essentially determined by the ratio of the lightest neutral Higgs-boson and the charged Higgsboson squared masses. We find that the fine-tuning in these theories can be reduced to the level
of 20%.
The organization of the paper is as follows. In Section 2 we discuss the first class of theories. We study electroweak symmetry breaking and the superparticle spectrum, identifying some
characteristic features of the theories. In Section 3 we discuss the second class of theories and
perform a similar analysis of electroweak symmetry breaking and the superparticle spectrum.
Conclusions are given in Section 4.
2. Models with D-type supersymmetry breaking
In this section we present the first class of models, in which supersymmetry breaking is transmitted to the supersymmetric standard model sector through a D-term VEV of a U (1) gauge
field. We find that the fine-tuning is reduced to the level of 20%.
2.1. Supersymmetry breaking from a D-term VEV
The supersymmetric standard model sector of our theories contains, as usual, the SU(3)C
SU(2)L U (1)Y (321) gauge multiplet, Vi (i = 1, 2, 3 for U (1)Y , SU(2)L and SU(3)C ), and
three generations of matter fields, Q, U , D, L and E. We also introduce a gauge singlet chiral
superfield S as well as two Higgs doublets Hu and Hd , with the standard Yukawa couplings in
the superpotential W = yu QU Hu + yd QDHd + ye LEHd .
Following Ref. [16], we consider that supersymmetry breaking is transmitted to the supersymmetric standard model sector through a D-term VEV of a U (1) gauge interaction, U (1) :
1
V = 2 2 D ,
2
(1)
32
where V is the vector superfield for U (1) . Introducing chiral superfields that transform as
adjoints under 321, A1 (1, 1)0 , A2 (1, 3)0 and A3 (8, 1)0 , supersymmetry breaking in Eq. (1) is
transmitted to the supersymmetric standard model sector through the following operators:

i
d 2
W W Ai + h.c.,
L=
(2)
M i
i=1,2,3
where i are coefficients of O(1), M is a mass parameter of order the Planck scale, and Wi and
W are the field-strength superfields for 321 and U (1) , respectively.3 These operators generate
Dirac masses for the gauginos of order D /M , which in turn generate flavor-universal squark and
slepton squared masses of order (1/16 2 )(D /M )2 at one loop. With M of order the Planck
scale, D (1010 1011 GeV)2 . An important property of this transmission is that the squark
and slepton masses are generated at the scale of Dirac gaugino masses D /M 110 TeV,
although the gaugino masses are present at the scale M . This reduces the logarithm associated
with the top-Yukawa induced radiative correction to the Higgs soft supersymmetry breaking
mass, and thus helps the reduction of fine-tuning.
We take the coefficients i in Eq. (2) to be free parameters. In particular, we do not impose
any unified relations on the three coefficients 1 , 2 and 3 . This is necessary to break unwanted
unified mass relations for the squarks and sleptons, such as m2t /m2e (4g34 /3)/(3g14 /5), and to
reduce
(For 1 = 2 = 3 in the basis where the gauge kinetic terms are given by

fine-tuning.
L = i d 2 (1/4gi2 )Wi Wi + h.c., which is expected to be the case in naive unified theories,

f
the squarks and sleptons obey unwanted unified mass relations m2 i g 4 C , where f =
f
l,
e and C f are the group-theoretical factors.)
q,
u,
d,
i
The introduction of the Ai fields destroys the successful supersymmetric prediction for gauge
coupling unification. Unification of the couplings can be recovered by the introduction of arbitrary vector-like matter fields, but at the price of losing the predictivity for the low-energy gauge
couplings. A possibility of recovering the prediction is to use the trinification idea, which has
been discussed in [16]. (The SU(5) case leads to a Landau pole for the gauge couplings much
below the unification scale at two loops.) This issue does not occur in the model presented in
Section 3, and we do not discuss it further in the context of the present model.
2.2. Masses for the A fields
For D (1010 1011 GeV)2 , the gravitino mass is roughly of the order of the weak scale:
m3/2 D /MPl , where MPl is the reduced Planck scale. The precise value of m3/2 depends on
various unknown parameters, for example, on M /MPl , so here we take m3/2 to be a free parameter of order m3/2 (100 GeV1 TeV). With these values of m3/2 , supersymmetric masses of
order the weak scale can be naturally generated
through the Khler potential. For example, if the

Khler potential contains the term K = i Ai A2i /2 + h.c. (i = 1, 2, 3), where Ai are dimen3 We assume that D is much larger than the largest F -type VEV, F , in the theory, i.e., D F , so that the contributions
to the supersymmetry breaking masses from F are negligible. For a discussion on how to obtain D F , see, e.g., [20].
Alternatively, one can separate the field giving the largest F from the supersymmetric standard model sector in an extra
dimension; see, e.g., [21]. (Mediation of supersymmetry breaking by Eq. (2) was also considered in [3,22] in a slightly
different context, in which the Ai fields arise as composites.)
33
sionless coefficients, the supergravity Lagrangian produces the effective superpotential term
1
Weff =
(3)
mA,i A2i ,
2
i=1,2,3
where mA,i = Ai m3/2 [15]. This can be understood easily in the compensator formalism (see,
e.g., [23]), in which the above Khler potential term can be written as

Ai 2
A + h.c.,
L = d 4
(4)
2 i
i=1,2,3
in the normalization where Ai are canonically normalized. Here, is the compensator field,
mass
which takes the value = 1 + 2 m3/2 . We then find that Eq. (4) gives the supersymmetric

term of Eq. (3) as well as the soft supersymmetry breaking term Lsoft = i mA,i m3/2 ai2 /2 +
h.c., where ai are the lowest components of Ai . Note that, assuming real couplings, we still have
a freedom of choosing the signs of mA,i (in the phase convention that i and m3/2 are real and
positive). Soft supersymmetry breaking terms could also receive contributions from the operators

i
d 2
L=
(5)
W W A2i + h.c.
2M2
i=1,2,3
Together with the contributions from the Khler potential of Eq. (4), we obtain

1
i D 2 2
Lsoft =
mA,i m3/2
ai + h.c.
2
M2
i=1,2,3
1
bA,i ai2 + h.c.
(6)
i=1,2,3
Here, as in the case of i , we do not impose any unified relations on Ai or i . The conditions
for quadratic stability of the ai -field origin are given by

2
|mA,i |4 + 4gi2 |mD,i |2 |mA,i |2 |bA,i |2 + 2gi2 bA,i m2
(7)
D,i + bA,i mD,i > 0,
where mD,i ii D / 2M . We assume that Eq. (7) is satisfied for all i = 1, 2, 3.

The introduction of a gauge singlet A1 has a potential danger of destabilizing the gauge hierarchy. Specifically, the operator L d 4 M A1 , together with Eq. (3), could lead to a large
VEV for A1 , and thus to a large FayetIliopoulos D term for U (1)Y through Eq. (2).4 Even
if absent at tree level, this operator would be generated at radiative level under the presence of
general non-renormalizable operators. We avoid this by imposing a symmetry
V V ,
Ai Ai ,
(8)
on the interactions of the observable sector. This symmetry is broken by the D and physics
generating it. However, if the breaking appears sufficiently soft in the observable sector, the
dangerous operator linear in A1 is sufficiently suppressed. Such a setup can naturally arise, for
example, by generating D on the infrared brane in warped space (with the infrared-brane scale
set to D ) and transmitting it to the observable sector on the ultraviolet brane through a bulk
U (1) . In the following, we assume that the operator linear in A1 is sufficiently suppressed.
4 The direct kinetic mixing between U (1) and U (1) , L =
Y
2
d W1 W , is assumed to be absent throughout.
34
Table 1
Z4,R charges for the fields
Z4,R
Vi
Ai
Hu
Hd
V
2.3. The Higgs sector

The Higgs sector of our theory consists of three chiral superfields S(1, 1)0 , Hu (1, 2)1/2 and
Hd (1, 2)1/2 . There are some variations on possible interactions in the Higgs sector. Here, to
demonstrate our point, we adopt a particular setup that uses a discrete Z4,R symmetry to constrain
the form of these interactions.
We consider a discrete R symmetry, Z4,R , under which fields transform as in Table 1. This
charge assignment allows all the interactions discussed so far, including the Yukawa couplings
and Eq. (2).5 In the absence of supersymmetry breaking, the Higgs sector superpotential consistent with Z4,R is W0 = SHu Hd + (/3)S 3 . (We assume that the possible term linear in
S is absent.) In addition, we have terms arising from the Khler potential K = H Hu Hd +
(S /2)S 2 + h.c. Adding these together, the superpotential of our Higgs sector is given by
WH = SHu Hd + Hu Hd +
MS 2 3
S + S ,
2
3
(9)
where = H m3/2 and MS = S m3/2 are mass parameters of order the weak scale.6 Soft supersymmetry
breaking parameters arise from the Khler potential terms as well as from the operators

L = d 2 (H Hu Hd + S S 2 /2)W W /M2 + h.c., giving

H D 2
1
S D 2 2
LH,soft = m3/2
MS m3/2
Hu Hd
S + h.c.
2
M2
M2
bS
bH Hu Hd S 2 + h.c.,
2
(10)
where we have used the same symbol for a chiral superfield and its scalar component for Hu , Hd
and S. The Higgs doublets also obtain non-holomorphic supersymmetry breaking masses at one
loop through 321 gauge interactions.
2.4. Parameters at the weak scale
Contributions to the gaugino masses arise from the operators in Eqs. (2), (3). The masses
of adjoint scalars also come from Eq. (6). Defining component fields as V = A

2
2
2
2
i + i + (1/2) D and A = a + 2 + F , these operators give
5 This Z
4,R symmetry forbids dangerous dimension four and five proton decay operators, as well as a large tree-level
supersymmetric mass for the Higgs doublets. It is broken by the VEV of the compensator field (a constant term in the
superpotential needed to cancel the cosmological constant) to the Z2,R subgroup, which is nothing but the standard R
parity.
6 The superpotential of Eq. (9) can also be written in the form of W = SH H + f (S), where f (S) is a general
u d
H
cubic function of S, by shifting the S field as S S /.
35
1
1
2imD Da 2imD Da mA mA
2
2
1
1 2

2
2
+ mA aF + mA a F bA a bA a ma a a,
(11)
2
2
for each gauge group factor SU(3)C , SU(2)L and U (1)Y (we have suppressed the index
i = 1, 2, 3). Here, we have added non-holomorphic supersymmetry breaking masses for as
(the last
term), although they are small in the present theory. The mass parameters mD
(ii / 2 )D /M , mA mA,i and bA bA,i are of order the weak scale or somewhat (an
order of magnitude) larger. The normalizations for A , and D are such that the inverse squares
of the 4D gauge couplings, 1/gi2 , appear in front of the kinetic terms.
We assume that the parameters in Eq. (11) are real. There are two gauginos, mixtures of and
, for each gauge group factor, and their masses are given by diagonalizing Eq. (11) as

1
2 2
2
2
m = 2g mD + mA 4g 2 m2D m2A + m4A ,
(12)
2
where we have suppressed the index i = 1, 2, 3 for m , g, mD and mA . The squark and slepton
masses arise from finite one-loop diagrams as
L = mD mD +
m2f
g 4 C f
i i 2
=
M ,
4 2 i
(13)
i=1,2,3
where (C1 , C2 , C3 ) = (1/60, 3/4, 4/3), (4/15, 0, 4/3), (1/15, 0, 4/3), (3/20, 3/4, 0) and
l and e,
(3/5, 0, 0) for f = q,
u,
d,
respectively, and M i2 are given by

4g 2 m2D + m2A bA + m2a
M i2 = m2D ln
g 2 m2D
2 2

4g mD + m2A + mA
mA

(14)
ln
,
4g 2 m2D + m2A mA
4g 2 m2D + m2A
which are positive in the entire parameter region [16]. Here, we have suppressed the index i =
1, 2, 3 for g, mD , mA , bA and m2a .
As we will see later, the relevant parameter region for us is where tan Hu /Hd is not
large, e.g., tan 3, so the only important Yukawa coupling is the top Yukawa coupling. The
soft supersymmetry breaking masses for the Higgs doublets are then given by

mD,3
3yt2 2
2
2
2
mHu ml
(15)
m + mu ln
,
m2Hd m2l ,
mq
8 2 q
where we have used the fact that the mediation scale for the squark masses is of order the Dirac
gluino mass mD,3 , and we have approximated mq mu inside the logarithm. A small soft mass
squared for S, m2S , is also generated at one loop through , picking up m2Hu and m2Hd . Eq. (15)
explicitly demonstrates that the effective messenger scale for this theory is very low
3 2
mq ,
Mmess mD,3
(16)
g3
so that larger squark masses can, in principle, be obtained for a given fine-tuning and Higgs
boson mass, compared with gauge-mediation-type models such as the ones considered in [2,3].
36
This is because the squark squared masses are suppressed by a one-loop factor compared with
the squared messenger scale in the present theory, while they are suppressed by a two-loop factor
in gauge-mediation-type models.
Weak-scale values for the couplings and are subject to the constraint that they do not hit the
Landau pole below the unification scale. In our theory, the 321 gauge couplings are large at the
ultraviolet due to the introduction of the A fields and any additional fields, e.g., needed to recover
coupling unification, which significantly weakens these constraints. We find that 0.8 can be
obtained for tan 1.8 for sufficiently large matter content, while the bound becomes somewhat
stronger for smaller tan , e.g., 0.7 for tan 1.4 [2,24]. Note that, with the strong 321
gauge couplings at the ultraviolet, tan as small as 1.2 is allowed because yt receives a strong
asymptotically non-free contribution from a large SU(3)C coupling at the ultraviolet. The bound
on is given by 0.2 (0.3) for 0.8 (0.7).
2.5. Suppression of the D-term potential and a constraint from the parameter
The operators in Eq. (2) give mixings between the auxiliary D fields and the scalar components of A (the third and fourth terms of Eq. (11)). As a consequence, the SU(2)L and U (1)Y
D-term contributions to the Higgs quartic couplings are suppressed [16]. Denoting the suppression factors by (2 and 1 for SU(2)L and U (1)Y , respectively), they are given by
=
m2A bA + m2a
4g 2 m2D + m2A bA + m2a
(17)
where, again, we have suppressed the index i = 1, 2. The D-term contributions to the Higgs
potential are given by times the standard contributions.
The suppression of the D-term potential can also be seen before integrating out the A fields.
Focusing on the T 3 direction of SU(2)L , the corresponding D-term potential is given, after
integrating out the D field, by

2
g22 1 2 1 2
h h + 2mD,2 2 ,
V=
(18)
2 2 u 2 d
where 2 is the imaginary part of the T 3 component of the SU(2)L adjoint field, a2 = i2 / 2 +
, and we have retained only the components for the Higgs doublets that obtain VEVs, Hu =
(0, hu )T and Hd = (hd , 0)T . The potential of Eq. (18) forces 2 to have a VEV
2 = (1 2 )
cos(2) 2 cos(2) 2
v
v ,
4mD,2
4mD,2
(19)
where v 2 hu 2 + hd 2 and 2 is given by Eq. (17). This mostly cancels the D-term potential.
(We find V = (2 g22 /2)(hu 2 /2 hd 2 /2) by substituting 2 back to V .)
The size of any SU(2)L triplet VEV is subject to a stringent constraint from electroweak data
(the parameter). This gives the upper bound on the value of 2 , and thus the lower bound on
mD,2 . Requiring | 1| cos2 (2)v 2 /8m2D,2 0.002 [25], we find mD,2 1 TeV for tan 2.
This bound is easily satisfied in the parameter region considered in the next subsection.
The imaginary part of the singlet field a1 also receives a small VEV of order v 2 /mD,1 , analogously to 2 . This VEV, however, does not affect phenomenology except that it is responsible
for the suppression of the U (1)Y D-term potential.
37
2.6. Electroweak symmetry breaking

We are now ready to discuss electroweak symmetry breaking. Our Higgs potential is given by
V = VF + VD + Vsoft ,
where VF , VD and Vsoft are given by

2
VF = Hu Hd + MS S + S 2 + |SHu + Hu |2 + |SHd + Hd |2 ,

2
2
3
a
a
g22
3g12 1
1
Hu + Hd
Hd + 1
H Hu Hd Hd ,
Hu
VD = 2
2
2
2
10 2 u
2
a=1

bS
Vsoft = m2Hu |Hu |2 + m2Hd |Hd |2 + bH Hu Hd + S 2 + h.c. .
2
(20)
(21)
(22)
(23)
Here, 2 and 1 are given by Eq. (17), and m2Hu and m2Hd by Eq. (15). The holomorphic supersymmetry breaking masses bH and bS , which we write as
bH = m3/2 bH,0 ,
(24)
bS = MS m3/2 bS,0 ,
(25)
are given by Eq. (10). Other supersymmetry breaking parameters are also generated at higher
loop orders.
The results of the potential minimization are given in Table 2 for four sample points A, B, C
and D, which lead to realistic phenomenology.
The parameters mD,i , mA,i , and bA,i (i = 1, 2, 3) are Dirac gaugino masses, supersymmetric
masses for Ai , and holomorphic supersymmetry breaking masses for Ai , respectively, and defined below Eq. (7), below Eq. (3), and in Eq. (6). The square bracket in the table is defined as
[X]n sgn(X) |X|n , and all masses are given in units of GeV. The effective and B parameters are defined by eff + S and (B)eff bH + (MS S + S2 ), and MHiggs is the
1 defined in Ref. [2], following [26], as a
lightest Higgs boson mass. We also list the parameter
measure of fine-tuning in our theory. All the parameters in the Higgs potential are taken to be real.
Our procedure to obtain these numbers is as follows. The input parameters of the analysis are
, , , MS , m3/2 , bH,0 , bS,0 , mD,i , mA,i , and bA,i . Using these, we can derive m2Hu and m2Hd ,
assuming some initial value for tan (which will be determined in the end by iteration). This
determines the Higgs potential of Eq. (20). We also add the one-loop contribution from topstop
loops to the Higgs quartic coupling in our analysis. (A precise calculation of this contribution
requires a knowledge of S, determined by iteration, but the effect of S = 0 is negligible.)
Corrections from higher loops are not so large for the values of top squark masses considered
here, only giving an additional negative contribution to the lightest Higgs boson mass of order
a few GeV. By minimizing the potential, we obtain Hu , Hd and S. These VEVs do not
2 H 2 + H 2 = (174 GeV)2 , so we iterate the entire process again
in general satisfy vH
u
d
using the input parameters appropriately rescaled by powers of (174 GeV/vH ) according to their
dimensions. In this process we use the derived value of tan , tan = Hu /Hd to determine
m2Hu . By iterating this several times, we obtain the final values for the parameters, which gives
Hu 2 + Hd 2 = (174 GeV)2 . The convergence of the whole procedure is fairly quick.
As is seen in the table, the fine-tuning required in our theory is very mild and only of order
20%. We find that the electroweak scale v is mainly sensitive to the values of , m3/2 , mD,2 ,
38
Table 2
Values for the parameters of the model for four sample points, A, B, C and D. The
resulting soft supersymmetry breaking masses for squarks and sleptons as well as
the quantities in the Higgs sector are also listed. Here, [X]n sgn(X) |X|n , and all
1 is defined in Ref. [2]
masses are given in units of GeV. The fine-tuning parameter
A
MS
m3/2
[bH,0 ]1/2
[bS,0 ]1/2
0.8
0.2
177
355
183
100
133
0.8
0.2
226
452
234
93
177
0.7
0.2
186
373
225
41
86
0.7
0.2
236
473
285
111
110
mD,1
mD,2
mD,3
mA,1
mA,2
mA,3
[bA,1 ]1/2
[bA,2 ]1/2
[bA,3 ]1/2
5076
2201
2018
385
269
459
92
110
318
6935
2688
2571
491
343
584
117
140
405
5331
2312
2120
2242
2415
2312
225
308
334
8112
3424
2690
2853
2527
2935
285
391
424
1.8
175
204
1.8
224
246
1.7
185
199
1.6
236
282
123
159
137
117
170
134
213
138
153
128
244
135
2.8
0.0063
0.0073
2.4
0.0054
0.0080
2.1
0.17
0.39
1.3
0.12
0.24
538
683
519
672
527
673
512
657
519
661
504
642
170
213
153
244
159
217
156
242
24%
16%
20%
13%
tan
eff
1/2

(B)eff
2 1/2
m
u
H
1/2
m2H
d
MHiggs
S
1
2
2 1/2
m
q2 1/2
mu
2 1/2
m
d2 1/2
m
l2 1/2
me
1

mD,3 , g2 , g3 and yt , and the fine-tuning parameter is determined by the sensitivity to , m3/2 ,
mD,3 , g3 and yt in most of the parameter region. The reduction of the fine-tuning occurs mainly
because the restoring force of the Higgs potential arises from the F -term potential, which is
stronger than the one from the SU(2) U (1)Y D-term potential [18]. A very small effective
messenger scale of Eq. (16) then allows squark masses as large as 500700 GeV. As can be seen
from the table, electroweak symmetry breaking in our theory is caused by an interplay between
the B term and the topstop radiative correction to m2Hu the diagonal entries in the Higgs
boson mass-squared matrix, {|eff |2 + m2Hu , |eff |2 + m2Hd }, are both positive, and one of the
eigenvalues becomes negative because of a non-zero value of (B)eff .
39
Table 3
The masses for the superparticles, Higgs bosons and adjoint scalars for
the four sample points A, B, C and D given in Table 2. All masses are
given in units of GeV
g 1
g 2
1
2
1982
2441
2526
3110
1427
3740
1812
4747
175
224
189
238
1305
1586
726
1299
1574
1929
3136
3823
169
219
186
236
207
250
208
255
10
20
30
40
50
60
70
392
482
400
494
1305
1586
726
1299
1574
1929
1599
2605
2175
2987
3136
3823
2560
3478
3840
5458
u L
u R
dL
dR
eL
eR
L
538
527
538
519
170
159
170
683
673
683
661
213
217
213
519
512
520
504
155
157
151
672
657
672
642
244
242
243
t1
t2
521
561
654
696
505
545
634
686
H10
134
138
128
135
H20
285
357
284
405
H30
P10
P20
H
476
600
493
618
173
220
177
253
365
280
430
353
354
281
453
403
aY,1
aY,2
aL,1
aL,2
aC,1
aC,2
396
4733
291
2872
558
4412
505
6464
370
3509
711
5622
2253
5434
2435
3844
2336
5156
2868
8057
2557
5104
2965
6544
183
234
225
285
2.7. Superparticle, Higgs boson and adjoint scalar spectrum

The masses for the superparticles, the Higgs bosons, and the adjoint scalars are calculated at
, and
tree level for the four sample points in Table 2, which are listed in Table 3. The g 1,2 , 1,2,3
0
17 represent the two gluinos, three charginos, and seven neutralinos, respectively, which come
40
from the linear combinations of the original gauginos, i , and the fermionic components of Ai ,
Hu and Hd . The u L,R , dL,R , eL,R , and L represent the left- and right-handed up-type squarks,
down-type squarks, charged sleptons, and the (left-handed) sneutrinos, respectively. The masses
for the top squarks, t1,2 , are listed separately because they split from the other squark masses
0
,
appreciably. The neutral scalar, pseudo-scalar, and charged Higgs bosons are labeled as H1,2,3
0
P1,2 , and H , respectively, which arise from the scalar components of S, Hu and Hd . There are
two adjoint scalar fields for each gauge group factor, which are denoted by aY , aL , and aC for
U (1)Y , SU(2)L , and SU(3)C , respectively. The gravitino is denoted by G.

In the first two points, A and B, the parameters are chosen such that mA,i mD,i , so that the
two gauginos for each gauge group factor are relatively close in masses: {g 1 , g 2 } for SU(3)C ,
{2 , 3 } and {40 , 50 } for SU(2)L , and {60 , 70 } for U (1)Y . Because of small values for mA,i ,
one of the two adjoint scalars for each gauge group factor, aY,1 , aL,1 and aC,1 , are relatively
light (below a TeV). For the other two points, C and D, the parameters are chosen such that
mA,i mD,i . Thus the two gaugino masses are not necessarily close, and the adjoint scalars are
all heavy, with masses above 2 TeV.
An interesting feature of the present model is that the effects of the gauge D-terms are suppressed because of mixings between the auxiliary D fields and the a fields (see Eq. (11)). This
also affects the spectrum of superparticles. For points A and B, the SU(2)L and U (1)Y D-terms
receive large suppressions, 1 , 2 1 (see Table 2). As a consequence, the squarks and sleptons
that are in the same SU(2)L multiplet are almost completely degenerate in mass. (Mass splittings
of order a few hundreds of MeV are generated from radiative corrections.) For points C and D,
the suppressions are not as strong as the case of points A and B, because of relatively large values
of mA,i , but the squarks and sleptons in the same SU(2)L multiplet are still quite degenerate.
We find that the lightest supersymmetric particle (LSP) in our theory can either be the third
generation right-handed slepton eR , the third generation sneutrino L , the lightest neutralino 10 ,
In either case, the mass of the LSP is naturally in the range 100300 GeV.
or the gravitino G.
Because of rather small values for tan and Mmess , the masses of the third generation eR and
L are almost degenerate with those of the corresponding first-two generation particles. For the
case of the L LSP, the left-handed selectrons eL will also be very close in mass, with the mass
difference to L only of order a few hundreds of MeV to a few GeV. For the 10 LSP, it is almost
purely the Higgsino, so that the lightest chargino 1 will be close in mass to 10 with the mass
difference of order a few GeV.
The lightest Higgs boson in our theory cannot be heavier than about 140 GeV. The mass of
the charged Higgs boson is also bounded by mH 450 GeV, as the amount of fine-tuning is
correlated with the charged Higgs boson mass [18]. This may have some implications on the rate
of the b s process. While the current theoretical estimates for this process still have some
uncertainties [27], the positive sign for the effective parameter seems to be preferred over
the other one, with which a partial cancellation between the charged Higgs boson and chargino
contributions is possible.
3. Models with sequestered gauge mediation
In this section we present the second class of models. We find that the fine-tuning is reduced
to the level of 1020% in these models.
41
3.1. Models
Our basic idea here is the following. We consider gauge mediation models, in which superparticle masses are generated by loops of messenger fields [28,29]. In particular, we consider a
model in which supersymmetric and supersymmetry breaking masses for the messenger fields
do not possess any particular unified relations (this requires multiple singlets in the messenger sector) [2,30]. Now, suppose that all the MSSM fields together with the messenger fields
are localized on a (3 + 1)-dimensional subspace in some higher-dimensional spacetime, and that
supersymmetry breaking occurs at some other subspace, which is transmitted to the messenger
sector through some bulk interactions. In this case, we can push up the fundamental scale of
supersymmetry breaking to an intermediate scale without affecting the gauge mediated spectrum
for the MSSM superparticles. On the other hand, supersymmetric masses of order the weak scale
can be generated in the Higgs sector from the Khler potential terms, as was the case in Section 2.3. In fact, this structure was used in Ref. [17] to generate the term in gauge mediation
models, where the coincidence of the scales for the term and for the superparticle masses was
also naturally obtained. Here we adopt the basic construction of this model to demonstrate our
point.
Let us consider (4 + 1)-dimensional spacetime with the extra dimension compactified on an
S 1 /Z2 orbifold, y : [0, 2], where y is the coordinate for the fifth dimension. The size of the extra
dimension we consider is small, only one or two orders of magnitude larger than the inverse of
the fundamental scale, which is of order the Planck scale. We consider that supersymmetry is
dynamically broken on the y = R brane at the scale , and (some of) the fields participating in
this dynamics are charged under a U (1)m gauge multiplet located in the bulk [17]. Our messenger
sector is localized on the y = 0 brane. Let us first consider only a single vector-like messenger
1)1/3 , where the numbers represent the 321 gauge quantum numbers. The
D(3 , 1)1/3 + D(3,
superpotential interactions in the messenger sector are then given by

f 3
2
L = (y) d kE XE E + X + kD XDD + h.c.,

(26)
3
where X, E and E are singlets under 321, and the U (1)m charges for these fields are cho
sen as E(+1), E(1),

X(0), D(0) and D(0).
Supersymmetry breaking is mediated from the
y = R brane to the y = 0 brane through U (1)m gauge interactions, generating positive super2 /16 2 )2 2 for E and E.
Here, gm is the
symmetry breaking squared masses of order (gm
4D U (1)m gauge coupling, which is naturally suppressed by the volume of the extra dimension. These positive squared masses in turn generate a negative mass squared for X through
the coupling kE , triggering the VEVs for the lowest and highest components of the X chiral
superfield: X = 0 and FX = 0. Note that, while the superpotential interactions of Eq. (26)
possess a U (1)R symmetry, it is explicitly broken by the trilinear scalar interactions arising from anomaly mediation [23,31], so that the dangerous Goldstone boson does not arise.
These VEVs then provide the supersymmetric and supersymmetry breaking masses for the
MD = kD X and FD = kD FX . For kE f kD O(1), the
messenger fields D and D:
2 F O(g 4 2 /(16 2 )3 ), so that we can naturally obtain
sizes of these masses are MD
D
m
6
2 (103 102 ), which is consistent with the volMD / FD / 10 105 for gm
ume suppression of gm . We thus take 1010 GeV and MD FD 10100 TeV in our
analysis. As we have seen, this requires some coincidence of the scales but does not require
fine-tuning.
42
In parOur messenger sector also contains vector-like messenger fields other than D and D.
ticular, to preserve the successful prediction for gauge coupling unification at the leading-log
level, we introduce messenger fields in complete SU(5) multiplets. Specifically, we introduce n5
L)
and n10 pairs of (Q, U, E) + (Q,
U , E),
where the 321 gauge quantum
pairs of (D, L) + (D,
numbers of Q, U, D, L and E are the same as the corresponding MSSM fields. The numbers n5
and n10 are bounded by n5 + 3n10 5 due to the Landau pole consideration for the 321 gauge
couplings. We consider that each component of the messenger fields has independent supersymmetric and supersymmetry breaking masses M and F : for example, we treat MD , FD , ML and
FL to be all independent for (n5 , n10 ) = (1, 0). There are a number of ways to achieve this. The
easiest way is to introduce E, E and X fields as well as the interactions of Eq. (26) for each
messenger field. Such a structure can naturally arise if we introduce a discrete Z3 symmetry for
each component of the messenger fields.7 In any event, with these most general Ms and F s, the
gaugino masses, Ma , and the sfermion masses, mf , at the messenger scale are written as
gi2
G,i ,
16 2
g 2 2 f
i
m2f = 2
Ci 2S,i ,
16 2
Mi =
(27)
(28)
i=1,2,3
where i = 1, 2, 3 represents U (1)Y , SU(2)L and SU(3)C , and Ci are the group theoretical factors. The parameters G,i and S,i are of order 10100 TeV, which can be explicitly calculated
in terms of the Ms and F s once the field content for the messengers is specified.
The Higgs sector of the present model is essentially the same as the one in the previous
model (see Section 2.3). The field content is given by S(1, 1)0 , Hu (1, 2)1/2 and Hd (1, 2)1/2 .
Imposing the discrete Z4,R symmetry of Table 1, the effective superpotential arises both from
W0 = SHu Hd + (/3)S 3 and K = H Hu Hd + (S /2)S 2 + h.c. as
MS 2 3
(29)
S + S ,
2
3
where and MS are parameters of order the weak scale generated via the mechanism of [15].
Note that this mechanism works even if the supersymmetry breaking sector (at the y = R brane)
and the Higgs sector (at the y = 0 brane) are geometrically separated [17]. The holomorphic
supersymmetry breaking terms
WH = SHu Hd + Hu Hd +
bS 2
S + h.c.
2
are also generated from the Khler potential terms as
LH,soft bH Hu Hd
(30)
bH = m3/2 ,
(31)
bS = MS m3/2 ,
(32)
where m3/2 2 /MPl is the gravitino mass of order the weak scale. The Yukawa couplings for
the quark and lepton superfields are given by W = yu QU Hu + yd QDHd + ye LEHd .
We here note that a theory having essentially the same properties can also be formulated
in warped spacetime of [33]. We can simply make our S 1 /Z2 extra dimension warped, with
7 These structures are consistent with gauge unification if the unified symmetry is realized in higher dimensions [32].
43
the scales on the ultraviolet and infrared branes set to be around the 4D Planck scale and the
intermediate scale, respectively. The MSSM fields, the singlet field S, and fields in the messenger
sector are all localized on the ultraviolet brane, while the U (1)m gauge multiplet propagates in
the bulk. Supersymmetry breaking occurs on the infrared brane, which is transmitted to the E
and E fields on the ultraviolet brane through bulk U (1)m gauge interactions, as in the models
of [34,35]. This theory allows a purely 4D interpretation through the AdS/CFT correspondence
[36,37], in which the separation of supersymmetry breaking and the other fields occurs through
conformal sequestering effects [38].
3.2. Electroweak symmetry breaking and particle spectrum
Now we study electroweak symmetry breaking in our model. The Higgs potential is given
by Eq. (20) with Eqs. (21)(23), but with both 1 and 2 set to 1 in Eq. (22). The holomorphic
supersymmetry breaking masses, bH and bS , are given by Eqs. (31), (32) rather than Eqs. (24),
(25). For smaller number of messenger fields, the 321 gauge couplings are not very strong at the
unification scale, so that the value of should be somewhat smaller than 0.8 to avoid the Landau
pole.
The results of the potential minimization are given in Table 4 for three sample points A,
B and C, which lead to realistic phenomenology. The square bracket in the table is defined as
[X]n sgn(X) |X|n , and all masses are given in units of GeV except for Mmess , G,i and S,i ,
which are given in units of TeV. The parameters G,i and S,i are defined in Eqs. (27), (28). The
quantity Mmess represents the scale at which the gaugino and sfermion masses of Eqs. (27), (28)
are given, which we take as a single scale of order s for simplicity. The sensitivity of physical
quantities to this parameter is rather weak. The effective and B parameters are defined by
eff + S and (B)eff bH + (MS S + S2 ), and MHiggs is the lightest Higgs boson
1 defined in [2]. All the parameters in the Higgs
mass. We also list the fine-tuning parameter
potential are taken to be real. The procedure to obtain these numbers is analogous to that in
Section 2.6.
As is seen in the table, we find that the fine-tuning in this theory is at the level of 1020%.
A difference from the previous model is that the logarithm ln(Mmess /mt) appearing in the top
stop correction to the Higgs mass-squared parameter is now not as small as the previous one.
(Mmess is several tens of TeV in the present model while it is a few TeV in the previous model.)
We find that for most of the parameter region the total mass-squared parameters for the up-type
and down-type Higgs doublets are both positive, and electroweak symmetry breaking is triggered
by a nonzero value of the effective B term, although the point C has a negative squared mass
for the up-type Higgs field, 2eff + m2Hu < 0. While the reduction of fine-tuning in the present
model is not as large as the previous one, the situation is still much better than in conventional
models of supersymmetry breaking, which typically require fine-tuning of order a few percent or
even worse.
We have listed the masses for the superparticles and the Higgs bosons in Table 5 for the
three sample points of Table 4. Here, we have included one-loop threshold corrections to obtain
these masses, as they are relevant for colored particles especially if the masses are close to the
and
experimental bounds. The meaning of the symbols is the same as that in Table 5: g,
1,2
0
15 denote the gluino, charginos and neutralinos, respectively, u L,R , dL,R , eL,R and L the
0
0 and H the neutral-scalar, pseudo-scalar, and charged
, P1,2
squarks and sleptons, and H1,2,3
is the gravitino.
Higgs bosons, respectively. The top squarks, t1,2 , are listed separately, and G
44
Table 4
Values for the parameters of the model for three sample points, A, B
and C. The resulting soft supersymmetry breaking masses for the gauginos, squarks and sleptons, as well as the quantities in the Higgs sector,
are also listed. Here, [X]n sgn(X) |X|n . All masses are given in units
of GeV except for Mmess , G,i and S,i (i = 1, 2, 3), which are given
1 is defined in Ref. [2]
in units of TeV. The fine-tuning parameter
A
MS
m3/2
n5
n10
Mmess
G,1
G,2
G,3
S,1
S,2
S,3
tan
1/2
eff
(B)eff
2 1/2
m
u
H
1/2
m2H
d
S
MHiggs
M1
M2
M3
2 1/2
m
q2 1/2
m
u2 1/2
m
d2 1/2
m
l2 1/2
me
1

0.65
0.2
202
598
268
0.75
0.2
135
580
362
0.75
0.2
175
616
127
1
0
50
53
68
30
56
68
30
4
0
50
100
140
39
57
72
19
1
1
50
150
82
48
98
49
24
1.9
204
234
1.8
139
224
3.6
175
149
143
17
177
238
1.6
123
250
4.4
129
187
0.7
120
74
185
258
450
142
391
333
406
213
224
411
434
391
335
416
387
331
406
238
250
187
94
95
165
13%
19%
12%
As is seen in the table, the theory accommodates various possibilities for the LSP. We find that
the LSP in this theory can be either the lightest neutralino 10 , the lightest chargino 1 , the third
We
generation right-handed slepton eR , the third generation sneutrino L , or the gravitino G.
note that the theory allows the relative signs between G,i s to be negative, although we did not
adopt such a case in our sample points A, B and C. In the case that 10 is the LSP, its thermal
relics may provide the dark matter of the universe [39]. In general, the neutral LSPs, 10 and G,
can be the dark matter if they are non-thermally produced. The cases with charged LSPs require
non-standard cosmology or small R-parity violation.
45
Table 5
The masses for the superparticles and the Higgs
bosons for the three sample points A, B and C given
in Table 4. All masses are given in units of GeV
g
1
2
10
20
30
40
50
306
376
450
183
148
165
235
404
257
84
158
187
204
158
207
204
159
207
216
365
216
616
606
641
u L
u R
dL
dR
eL
eR
L
448
390
452
388
240
100
232
404
334
408
331
253
100
246
431
415
438
407
192
170
178
t1
t2
358
451
310
415
384
448
H10
123
129
120
H20
H30
P10
P20
H
358
335
265
730
754
695
309
239
243
503
358
268
460
332
362
588
271
127
4. Conclusions
We have constructed two classes of realistic supersymmetric models in which no significant
fine-tuning is required to reproduce the correct scale for electroweak symmetry breaking. Both
classes of models accomplish this by incorporating (i) a means to increase the effective Higgs
quartic coupling beyond what is available in the MSSM, (ii) a low scale for the generation of
squark and slepton masses, and (iii) a degree of independent adjustability for the size of the
squark masses, such that they can be set not very far from their experimental bound of 300 GeV.
The first feature allows us to evade the LEP II bound of MHiggs 114 GeV without relying on
large radiative corrections from topstop loops, and thus to have small top squark masses. The
second and third features then facilitate a softening of the topstop loop correction to the up-type
Higgs mass-squared parameter to a level commensurable with the Higgs field VEV v 174 GeV,
thereby eliminating the need for delicate cancellations. Specifically, the low mass-generation
scale leads to a modest logarithm, 25, in the correction, and the free adjustability of the squark
masses moderates the overall mass scale of the correction.
46
In the first class of models, the additional Higgs quartic contribution is provided by F exchange of a singlet chiral superfield. A fundamental intermediate-scale supersymmetry breaking is communicated to the MSSM fields primarily via two means. First, a new U (1) gauge
superfield acquires a D-term VEV at the intermediate scale and marries the 321 gauginos with
adjoint chiral fermions via non-renormalizable interactions. This leads to a consequence that
the radiative generation of the sfermion masses occurs at a scale only a factor of O(4) higher
than the electroweak scale. This accomplishes the low scale of sfermion mass generation. The
adjustability of squark masses is accomplished simply by making the non-renormalizable couplings responsible for the gaugino masses independent for SU(3)C , SU(2)L , and U (1)Y . The
second means for transmitting supersymmetry breaking is via supergravity effects. These naturally generate weak-scale supersymmetric and supersymmetry breaking masses for the Higgs
and singlet fields, producing an appropriate Higgs sector superpotential. Working together, these
characteristic features of the models allow for the coexistence of the electroweak VEV with a
variety of superparticle spectra, with a very mild tuning of about 20%.
The second class of models also relies on F -exchange of a singlet to obtain the additional
Higgs quartic coupling. However, communication of supersymmetry breaking to the MSSM sector more closely follows traditional gauge mediation. This leads to sfermion mass-generation
scales O(16 2 ) higher than the electroweak scale, which are larger than the mass-generation
scales for the first class of models, but still provides only modest logarithmic enhancement in
the Higgs mass correction. Supersymmetry is fundamentally broken at an intermediate scale, at
a location physically separated from the MSSM fields in an extra dimension. The breaking is
then communicated between the two locations by a bulk U (1), which ultimately gives supersymmetric and supersymmetry breaking masses of O(10100 TeV) to the messenger fields of gauge
mediation. We assume no specific relation between any of the mass parameters for the messengers, ensuring the adjustability of the squark masses. Weak-scale values for supersymmetric and
supersymmetry breaking masses for the Higgs and singlet fields are again generated from supergravity effects. Within this class of models the reduction of tunings to 1020% can be obtained,
again with a variety of configurations for the superparticle spectrum.
In searching out parameter points of reduced tuning in the above models, it quickly becomes
apparent that the weakest tunings and largest Higgs boson masses are generally obtained when
the Higgs potential is made stable along both the Hu and Hd axes, and destabilized in an intermediate direction via the B term. The reason is simply that the Higgs quartic coupling due
to F -exchange is strongest when tan 1. Thus, to most successfully utilize the new quartic
coupling, smaller values for tan are preferred, and electroweak symmetry breaking must be
dominantly caused by the B term. This clearly demonstrates the framework introduced in [18].
It is very encouraging that this very generic idea has been successfully demonstrated in two
distinct classes of realistic models.
Acknowledgements
This work was supported in part by the Director, Office of Science, Office of High Energy and
Nuclear Physics, of the US Department of Energy under Contract DE-AC02-05CH11231. The
work of Y.N. was also supported by the National Science Foundation under grant PHY-0403380,
by a DOE Outstanding Junior Investigator award, and by an Alfred P. Sloan Research Fellowship.
47
References
[1] S. Dimopoulos, H. Georgi, Nucl. Phys. B 193 (1981) 150;
N. Sakai, Z. Phys. C 11 (1981) 153;
S. Dimopoulos, S. Raby, F. Wilczek, Phys. Rev. D 24 (1981) 1681.
[2] Z. Chacko, Y. Nomura, D. Tucker-Smith, Nucl. Phys. B 725 (2005) 207, hep-ph/0504095.
[3] Y. Nomura, B. Tweedie, Phys. Rev. D 72 (2005) 015006, hep-ph/0504246.
[4] J.R. Ellis, J.F. Gunion, H.E. Haber, L. Roszkowski, F. Zwirner, Phys. Rev. D 39 (1989) 844.
[5] M. Cvetic, D.A. Demir, J.R. Espinosa, L.L. Everett, P. Langacker, Phys. Rev. D 56 (1997) 2861, hep-ph/9703317;
M. Cvetic, D.A. Demir, J.R. Espinosa, L.L. Everett, P. Langacker, Phys. Rev. D 58 (1998) 119905.
[6] J.R. Espinosa, M. Quiros, Phys. Rev. Lett. 81 (1998) 516, hep-ph/9804235.
[7] P. Batra, A. Delgado, D.E. Kaplan, T.M.P. Tait, JHEP 0402 (2004) 043, hep-ph/0309149;
A. Maloney, A. Pierce, J.G. Wacker, hep-ph/0409127.
[8] J.A. Casas, J.R. Espinosa, I. Hidalgo, JHEP 0401 (2004) 008, hep-ph/0310137;
A. Brignole, J.A. Casas, J.R. Espinosa, I. Navarro, Nucl. Phys. B 666 (2003) 105, hep-ph/0301121.
[9] R. Harnik, G.D. Kribs, D.T. Larson, H. Murayama, Phys. Rev. D 70 (2004) 015002, hep-ph/0311349;
S. Chang, C. Kilic, R. Mahbubani, Phys. Rev. D 71 (2005) 015003, hep-ph/0405267;
A. Delgado, T.M.P. Tait, hep-ph/0504224.
[10] A. Birkedal, Z. Chacko, Y. Nomura, Phys. Rev. D 71 (2005) 015006, hep-ph/0408329.
[11] K.S. Babu, I. Gogoladze, C. Kolda, hep-ph/0410085.
[12] R. Kitano, Y. Nomura, hep-ph/0509039.
[13] K. Choi, K.S. Jeong, K. Okumura, hep-ph/0504037;
K. Choi, K.S. Jeong, T. Kobayashi, K. Okumura, hep-ph/0508029.
[14] Y. Nomura, D. Tucker-Smith, B. Tweedie, Phys. Rev. D 71 (2005) 075004, hep-ph/0403170.
[15] G.F. Giudice, A. Masiero, Phys. Lett. B 206 (1988) 480.
[16] P.J. Fox, A.E. Nelson, N. Weiner, JHEP 0208 (2002) 035, hep-ph/0206096.
[17] Y. Nomura, T. Yanagida, Phys. Lett. B 487 (2000) 140, hep-ph/0005211.
[18] Y. Nomura, D. Poland, B. Tweedie, hep-ph/0509244.
[19] M. Bastero-Gil, C. Hugonie, S.F. King, D.P. Roy, S. Vempati, Phys. Lett. B 489 (2000) 359, hep-ph/0006198;
R. Dermisek, J.F. Gunion, Phys. Rev. Lett. 95 (2005) 041801, hep-ph/0502105.
[20] T. Gregoire, R. Rattazzi, C.A. Scrucca, hep-ph/0505126.
[21] L. Carpenter, P.J. Fox, D.E. Kaplan, hep-ph/0503093.
[22] Z. Chacko, P.J. Fox, H. Murayama, Nucl. Phys. B 706 (2005) 53, hep-ph/0406142.
[23] L. Randall, R. Sundrum, Nucl. Phys. B 557 (1999) 79, hep-th/9810155.
[24] M. Masip, R. Munoz-Tapia, A. Pomarol, Phys. Rev. D 57 (1998) 5340, hep-ph/9801437.
[25] S. Eidelman, et al., Particle Data Group, Phys. Lett. B 592 (2004) 1.
[26] G.W. Anderson, D.J. Castano, Phys. Lett. B 347 (1995) 300, hep-ph/9409419;
See also: R. Barbieri, G.F. Giudice, Nucl. Phys. B 306 (1988) 63.
[27] See, e.g., F.M. Borzumati, C. Greub, Phys. Rev. D 58 (1998) 074004, hep-ph/9802391;
F.M. Borzumati, C. Greub, Phys. Rev. D 59 (1999) 057501, hep-ph/9809438;
M. Ciuchini, G. Degrassi, P. Gambino, G.F. Giudice, Nucl. Phys. B 527 (1998) 21, hep-ph/9710335;
P. Gambino, M. Misiak, Nucl. Phys. B 611 (2001) 338, hep-ph/0104034;
M. Neubert, Eur. Phys. J. C 40 (2005) 165, hep-ph/0408179.
[28] M. Dine, W. Fischler, Phys. Lett. B 110 (1982) 227;
M. Dine, W. Fischler, Nucl. Phys. B 204 (1982) 346;
L. Alvarez-Gaume, M. Claudson, M.B. Wise, Nucl. Phys. B 207 (1982) 96;
S. Dimopoulos, S. Raby, Nucl. Phys. B 219 (1983) 479.
[29] M. Dine, A.E. Nelson, Y. Shirman, Phys. Rev. D 51 (1995) 1362, hep-ph/9408384;
M. Dine, A.E. Nelson, Y. Nir, Y. Shirman, Phys. Rev. D 53 (1996) 2658, hep-ph/9507378.
[30] K. Agashe, M. Graesser, Nucl. Phys. B 507 (1997) 3, hep-ph/9704206;
S.P. Martin, Phys. Rev. D 55 (1997) 3177, hep-ph/9608224.
[31] G.F. Giudice, M.A. Luty, H. Murayama, R. Rattazzi, JHEP 9812 (1998) 027, hep-ph/9810442.
[32] Y. Kawamura, Prog. Theor. Phys. 105 (2001) 999, hep-ph/0012125;
L.J. Hall, Y. Nomura, Phys. Rev. D 64 (2001) 055003, hep-ph/0103125;
L.J. Hall, Y. Nomura, Ann. Phys. 306 (2003) 132, hep-ph/0212134.
[33] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hep-ph/9905221.
48
[34]
[35]
[36]
[37]
T. Gherghetta, A. Pomarol, Nucl. Phys. B 586 (2000) 141, hep-ph/0003129.

W.D. Goldberger, Y. Nomura, D.R. Smith, Phys. Rev. D 67 (2003) 075021, hep-ph/0209158.
J.M. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200.
N. Arkani-Hamed, M. Porrati, L. Randall, JHEP 0108 (2001) 017, hep-th/0012148;
R. Rattazzi, A. Zaffaroni, JHEP 0104 (2001) 021, hep-th/0012248.
[38] M.A. Luty, R. Sundrum, Phys. Rev. D 65 (2002) 066004, hep-th/0105137.
[39] Y. Nomura, K. Suzuki, Phys. Rev. D 68 (2003) 075005, hep-ph/0110040.
Sneutrino inflation in GaussBonnet brane-world

cosmology, the gravitino problem and leptogenesis
G. Panotopoulos
Department of Physics, University of Crete, Heraklion, Crete, Greece
Received 11 November 2005; accepted 4 April 2006
Abstract
We discuss sneutrino inflation in the brane-world scenario. We work in the RandallSundrum type II
brane-world, generalized with the introduction of the GaussBonnet term, a correction to the effective
action in string theories. We find that a viable inflationary model is obtained with a reheating temperature
appropriate to lead to the right baryon asymmetry and render the gravitino safe for cosmology. In specific
realizations we satisfy all the observational constaints without the unnaturally small Yukawa couplings
required in other related approaches.
1. Introduction
Inflation [1] has become the standard paradigm for the early Universe, because it solves some
outstanding problems present in the standard hot big-bang cosmology, like the flatness and horizon problems, the problem of unwanted relics, such as magnetic monopoles, and produces the
cosmological fluctuations for the formation of the structure that we observe today. The recent
spectacular CMB data from the WMAP satellite [2,3] have strengthen the inflationary idea, since
the observations indicate an almost scale-free spectrum of Gaussian adiabatic density fluctuations, just as predicted by simple models of inflation. According to chaotic inflation with a
potential for the inflaton field of the form V = (1/2)m2 2 , the WMAP normalization condition requires for the inflaton mass m that m = 1.8 1013 GeV [4]. However, a yet unsolved
problem about inflation is that we do not know how to integrate it with ideas in particle physics.
E-mail address: panotop@physics.uoc.gr (G. Panotopoulos).

doi:10.1016/j.nuclphysb.2006.04.002
50
G. Panotopoulos / Nuclear Physics B 745 (2006) 4961
For example, we would like to identify the inflaton, the scalar field that drives inflation, with one
of the known fields of particle physics.
One of the most exciting experimental results in the last years has been the discovery of
neutrino oscillations [5]. These results are nicely explained if neutrinos have a small but finite
mass [6]. The simplest models of neutrino masses invoke heavy gauge-singlet neutrinos that give
masses to the light neutrinos via the seesaw mechanism [7]. If we require that light neutrino
masses 101 to 103 eV, as indicated by the neutrino oscillations data, we find that the heavy
singlet neutrinos weight 1010 to 1015 GeV [8], a range that includes the value of the inflaton
mass compatible with WMAP. On the other hand, the hierarchy problem of particle physics is
elegantly solved by supersymmetry (see e.g. [9]), according to which every known particle comes
with its superpartner, the sparticle. In supersymmetric models the heavy singlet neutrinos have
scalar partners with similar masses, the sneutrinos, whose properties are ideal for playing the role
of the inflaton [4,10].
Superstring theory includes, apart from the fundamental string, other extended objects
called p-branes. A special class of p-branes are D(irichlet)p-branes, where open strings can
end. D-brane physics has motivated the brane-world idea, which has attracted a lot of interest
over the last years. In a brane-world scenario our universe is modeled by a 3-brane embedded
in a five-dimensional bulk spacetime. In the simplest cases, all the standard model fields (open
string sector) are confined on the brane, while gravity (closed string sector) propagates in the
bulk. The brane is a hypersurface that splits the five-dimensional manifold into two parts and
plays the role of a boundary of spacetime. Usually the brane is considered to be infinitely thin
and the matching conditions can be used to relate the bulk dynamics to what we observe on the
brane. The model first proposed by Randall and Sundrum (RS II) [11] offers a viable alternative
to the standard KaluzaKlein treatment of the extra dimensions and together with various extensions has been intensively investigated for its cosmological consequences (see e.g. [12] and for
reviews [13]).
In four dimensions, the Einstein tensor is the only second-rank tensor that (i) is symmetric,
(ii) is divergence free, (iii) it depends only on the metric and its first derivatives, and (iv) is linear
in second derivatives of the metric. However, in D > 4 dimensions more complicated tensors
with the above properties exist. For example, in five dimensions the second order Lovelock tensor
reads
Hab = RRab 2Rac Rbc 2R cd Racbd + Racde Rbcde

1
gab R 2 4Rcd R cd + R cdes Rcdes
4
and can be obtained from an action containing the GaussBonnet (GB) term [14]
LGB = R 2 4Rab R ab + R abcd Rabcd .
(1)
(2)
Higher order curvature terms appear also in the low-energy effective field equations arising in
string theory. Brane-worlds are string-inspired and so it is natural to include such terms in the
five-dimensional field equations.
It is important to note that in the context of extra dimensions and the brane-world idea one
obtains on the brane a generalized Friedmann equation, which is different from the usual one of
conventional four-dimensional cosmology. This means that the rate of expansion of the universe
in this novel cosmology is altered and accordingly the description of the physics in the early
universe can be different from the standard treatment. So it would be very interesting to study the
cosmological implications of these new ideas about extra dimensions and brane-worlds. Perhaps
51
the best laboratory for such a study is inflation, which has become the standard paradigm in the
big-bang cosmology and which is favoured by the recent observational data. The Friedmann-like
equation for a GaussBonnet brane-world has been derived in [14,15].
In [16] the authors first introduced the GB term in the RandallSundrum setup and also
they briefly discussed cosmology. Sneutrino inflation in the context of RandallSundrum type II
model has been analyzed in [17]. However, it would be interesting to study the effect of the
GB term. After all, this term is a high energy modification to general relativity and as such it is
expected to be important in the early universe. Furthermore, as it has been shown in [18], the
quadratic potential V 2 for the inflaton is observationally more favoured when the GB term
is present. The purpose of the present work is to discuss sneutrino inflation in the context of a
GaussBonnet brane-world.
Our work is organized as follows. There are five sections of which this introduction is the first.
In Section 2 we describe sneutrino inflation in a GaussBonnet brane-world. Section 3 contains
the discussion of reheating, gravitino production and baryogenesis through leptogenesis. Our
results are summarized in Section 4 and we conclude with a discussion Section 5.
2. Sneutrino inflation in a GaussBonnet brane-world
2.1. GaussBonnet brane-world
Here we review GaussBonnet brane-world, following essentially [18]. The five-dimensional
bulk action for the GaussBonnet brane-world scenario is given by

1
S = 2 d 5 x (5) g 25 + R + R 2 4Rab R ab + Rabcd R abcd
25
(3)
d 4 x g + Smat ,
brane
where > 0 is the GaussBonnet (GB) coupling, which has dimensions of length2 , > 0 is the
brane tension, 5 < 0 is the bulk cosmological constant and Smat denotes the matter action. The
fundamental energy scale of gravity is the five-dimensional scale M5 with 52 = 8/M53 . For the
discussion to follow we define a new mass scale through the relation = 1/M2 .
The GB term may be viewed as the lowest-order stringy correction to the five-dimensional
EinsteinHilbert action with 1/2 , where 1/ is the bulk curvature scale, |R| 2 . The
RandallSundrum type models are recovered for = 0. Moreover, for an anti-de Sitter bulk, it
follows that 5 = 32 (2 ), where
42 1.
(4)
Imposing a Z2 reflection symmetry across the brane in an anti-de Sitter bulk and assuming
that a perfect fluid matter source is confined on the brane, one obtains the modified Friedmann
equation

H2
H2
2
5 ( + ) = 2 1 + 2 3 + 2 2 .
(5)
This can be rewritten in the useful form

2
2
2
(1 ) cosh
1 ,
H =
(6)
52
where is a dimensionless measure of the energy density on the brane defined by

+ = m4 sinh
with

m =
82 (1 )3
54
(7)
1/8
(8)
the characteristic GB energy scale.

The requirement that one should recover general relativity at low energies leads to the relation
2
42 =
(9)
,
1+ 5
2 and M is the four-dimensional Planck scale. Since 1, we have
where 42 = 8/Mpl
pl
2 . Furthermore, the brane tension is fine-tuned to zero effective cosmological constant on
M53 /Mpl
the brane
52 = 2(3 ).
(10)
1/4 ,
The GB energy scale m is larger than the RS energy scale

since we consider that the GB
term is a correction to RS gravity. Using (10) this implies [19] 0.15, which is consistent with
Eq. (4).
Expanding Eq. (6) in , we find three regimes for the dynamical history of the brane universe
2 2
2/3
5
m4 H 2
(11)
(GB),
4
m4 H 2
42 2
(RS),
(12)
42
(GR).
(13)
3
Eqs. (11)(13) are considerably simpler than the full Friedmann equation and for inflation we
shall assume the first one (GB).
H2
2.2. Chaotic inflation in a GaussBonnet brane-world

We will consider the case in which the energymomentum on the brane is dominated by
the sneutrino inflaton field confined on the brane with a self-interaction potential V () =
(1/2)M 2 2 , where M is the mass of the sneutrino field. The field is a function of time only, as
dictated by the isotropy and homogeneity of the observed four-dimensional universe. A homogeneous scalar field behaves like a perfect fluid with pressure p = (1/2) 2 V and energy density
= (1/2) 2 + V . We shall assume that there is no energy exchange between the brane and the
bulk, so the energy-momentum tensor T of the scalar field is conserved, that is T = 0. In
terms of the pressure p and the energy density the continuity equation takes the form
+ 3H (p + ) = 0,
(14)
where H is the Hubble parameter H = a/a.
This is equivalent to the equation of motion for the

scalar field
+ 3H + V () = 0
(15)
53
the KleinGordon equation for in a RobertsonWalker background. The equation that governs
the dynamics of the expansion of the universe is the Friedmann-like equation of the previous
subsection. Inflation takes place in the early stages of the evolution of the universe, so we suppose
that inflation takes place in the GB high energy regime
2 2 2/3
5
H2 =
(16)
.
4
In the slow-roll approximation the slope and the curvature of the potential must satisfy the two
constraints 1 and || 1, where and are the two slow-roll parameters which are defined
by
H
,
H2
V
.
3H 2
In this approximation the equation of motion for the scalar field takes the form

V

3H
(17)
(18)
(19)
while the generalized Friedmann equation becomes (V 2 )

2 2 2/3
5
H2
.
V
4
(20)
The number of e-folds during inflation is given by

af
N ln
=
ai
tf
H dt.
(21)
ti
Before presenting all the formulae, it would perhaps be useful at this point to describe what
follows. Any model of inflation should (i) solve the flatness and horizon problems, (ii) reproduce the amplitude for density perturbations (COBE normalization), (iii) predict a nearly
scale-invariant spectrum, and (iv) predict very small tensor perturbations. For a strong enough
inflation we take N = 70, which is enough to solve the horizon and flatness problems. Using the
equations of motion we shall compute the spectral index, as well as the scalar and tensor perturbations. We will then fix the remaining parameters by requiring that the amplitude of scalar
perturbations is reproduced. This will lead to a prediction of the spectral index and the tensor-toscalar ratio.
According to a recent analysis [20], at 1
As 2 105 ,
(22)
0.048 < ns 1 < 0.016
(23)
with As the amplitude of the density perturbations and ns the spectral index. On large cosmological scales, data [20] give for the tensor perturbations
r < 0.47 95% c.l.,
(24)
54
with r the tensor-to-scalar ratio defined as r = 16A2t /A2s (consistent with the normalization of
Ref. [3] in the low energy limit), where At is the amplitude of the tensor perturbations.
In the slow-roll approximation the number of e-folds and the slow-roll parameters are given
by the formulae

V 2
4 2/3
,

(25)
9V 5/3 2 52

4 2/3
V
,

(26)
3V 2/3 2 52

2 52
N 3
4
2/3 end
V 2/3
d,
V
(27)
where end is the value of the inflaton at the end of inflation, which is determined from the
condition that the maximum of , || equals unity, and the denotes the point at which observable quantities are computed. The main cosmological constraint (normalization condition) comes
from the amplitude of the scalar perturbations [21]
As =
4
H2
,
2 |H ()|
5 Mpl
(28)
where the right-hand side is evaluated at the horizon-crossing when the comoving scale equals
the Hubble radius during inflation and Mpl = 1.22 1019 GeV is the four-dimensional Planck
mass. In the present context the amplitude of the scalar perturbations is given by

144V 8/3 2 52 2/3
2
As =
(29)
.
4 V 2
4
25Mpl
The spectral index for the scalar perturbations ns is given in terms of the slow-roll parameters
d ln A2s
= 2 6
d ln k
and is found to be
2N 3
= 0.98
ns =
2N
while the tensor-to-scalar ratio r is given by
ns 1
r=
2
3M2 Mpl
(30)
(31)
(32)
2N 3/2 MM53
with At the amplitude of the tensor perturbations [21]

2 H
At =
,
5 Mpl
(33)
where again the right-hand side is evaluated at the horizon-crossing. Taking the normalization
condition into account we obtain for M
2/3
M = 3.4 105
4/3
M Mpl
M5
(34)
55
and for the tensor-to-scalar ratio

4/3
r = 75.3
2/3
M Mpl
M52
(35)
3. Reheating, gravitino production and leptogenesis

3.1. Reheating
We start by introducing three heavy right-handed neutrinos Ni which only interact with leptons and Higgs. The superpotential that describes their interactions is [22]
W = fia Ni La Hu ,
(36)
where fia is the matrix for the Yukawa couplings, Hu is the superfield of the Higgs doublet
that couples to up-type quarks and La (a = e, , ) is the superfield of the lepton doublets. We
assume that the scalar partner of the lightest right-handed neutrino plays the role of the inflaton.
After inflation the inflaton decays into normal particles which quickly thermalize. This is the
way the universe reenters the radiation dominated era. The sneutrino inflaton decays into leptons
and Higgs and their antiparticles according to the superpotential (36) and the decay rate is given
by [22]
1 2
f M
=
(37)
4

with M the sneutrino mass and f 2 a |f1a |2 . The reheating temperature after inflation is defined by assuming instantaneous conversion of the inflaton energy into radiation, when the decay
rate of the inflaton equals the expansion rate H . In GaussBonnet braneworld cosmology H
is given by
2 1/3
5
1/3
H=
(38)
16
and in the radiation dominated era the energy density of the universe is given by
2 4
(39)
T ,
30
with geff = 228.75 the effective number of relativistic degrees of freedom in the MSSM for
T 1 TeV. Thus we obtain

2 1/3
5
2 4 1/3
.
geff T
H=
(40)
16
30
The condition H (TR ) = gives for the reheating temperature

15M 3 M53 6 1/4
f
.
TR =
(41)
16 6 geff M2
After inflation, the direct out-of-equilibrium decays of the sneutrino inflaton generate the lepton
asymmetry which is partially converted into a baryon asymmetry via sphaleron effects. This
requires that TR < M or that

16 6 geff MM2 1/3
2
f <
(42)
.
15M53
= R = geff
56
3.2. Gravitino production

Any viable inflationary model should avoid the gravitino problem [23]. This means that for
unstable gravitinos that decay after big-bang nucleosynthesis (BBN), their decay products should
not alter the abundances of the light elements in the universe that BBN predicts. This requirement
sets an upper bound for the gravitino abundance
3/2
n3/2
max
<
n
m3/2
(43)
with m3/2 100 GeV1 TeV the gravitino mass, n the photon number density and max a parameter related to the maximum gravitino abundance allowed by the BBN predictions. According
to the analysis of the authors of [24], max = 5 1012 GeV for m3/2 = 100 GeV. To find the
gravitino abundance one has to integrate Boltzmann equation
dn3/2
+ 3H n3/2 = C3/2 (T )
dt
(44)
with C3/2 (T ) the collision term responsible for the thermal production of gravitinos as a function of the temperature T < TR . The rate for the thermal production of gravitinos is dominated by
QCD processes since the strong coupling is considerably larger than the electroweak couplings.
Taking into account 10 two-body processes involving left-handed quarks, squarks, gluons and
gluinos, the authors of [25] computed the collision term C3/2 (T ) in the framework of supersymmetric QCD. They obtained

m2g T 6
C3/2 (T ) = a(T ) 1 + b(T ) 2
2
m3/2 Mpl
(45)
where mg 1 TeV is the gluino mass and a(T ), b(T ) are two slowly-varying functions of the
temperature, estimated to be [17]
a(TR ) = 2.38,
b(TR ) = 0.13.
(46)
If we assume that the quantity sa 3 is constant during the expansion of the universe, where a is the
scale factor and s is the entropy density s = heff (2 2 T 3 )/45, then the integration of Boltzmann
equation gives
3/2 (T ) =
heff (T ) C3/2 (TR )

,
heff (TR ) H (TR )n (TR )
(47)
with heff the effective number of relativistic degrees of freedom. For T 1 TeV all particles are
relativistic and for the MSSM heff (TR ) geff (TR ) = 915/4 = 228.75, while heff (T ) = 43/11 for
T < 1 MeV. Thus, using (43) with m3/2 = 100 GeV one is led to the following upper bound for
the reheating temperature
6/5
TR 1.63 10
2/5
Mpl M
3/5
M5
T0 .
(48)
At this point we should also check whether the contribution of the gravitinos to the energy density
of the universe is compatible with the observed matter density of the universe, m h2 < 0.143 [2],
57
s
where h = (H /100) Mpc
km . From the gravitino abundance we can calculate their normalized density
1
3/2 h2 = m3/2 3/2 n 0 h2 cr
(49)
with n 0 = 3.15 1039 GeV3 the photon density today and cr = 8.07 1047 h2 GeV4 the
critical density. For m3/2 = 100 GeV we obtain
5/3
3/2 h2 = 1.86 109
M5 T R
2/3
2M
Mpl
(50)
Using the WMAP bound on the matter density of the universe, m h2 < 0.143 we get the following relation between TR , M5 and M
6/5
TR < 8.54 107
2/5
Mpl M
3/5
(51)
M5
which is less stringent than the constraint (48) coming from BBN.
3.3. Direct leptogenesis from sneutrino decay
Any lepton asymmetry YL nL /s produced before the electroweak phase transition is partially converted into a baryon asymmetry YB nB /s via sphaleron effects [26]. The resulting YB
is
YB = CYL
(52)
with the fraction C computed to be C = 8/15 in the MSSM [27]. The lepton asymmetry, in
turn, is generated by the direct out-of-equilibrium decays of the sneutrino inflaton after inflation
and is given by [22]
3 TR

(53)
4M
with the CP asymmetry in the sneutrino decays. For convenience we parametrize the CP asymmetry in the form
YL =
= max sin L ,
(54)
where L is an effective leptogenesis phase and max is the maximum asymmetry which is given
by [28]

2
3 M matm
max =
(55)
8 v 2 sin2
with v = 174 GeV the electroweak scale, tan the ratio of the vevs of the two Higgs doublets of
the MSSM and m2atm = 2.6 103 eV2 the mass squared difference measured in atmospheric
neutrino oscillation experiments. For simplicity we shall take sin 1 (large tan regime), in
which case the maximum CP asymmetry is given by

M
max
10

(56)
= 2 10
.
106 GeV
58
Combining the above formulae we obtain

TR
YB = 8 1011 | sin L |
.
106 GeV
(57)
From the WMAP data [2] we know that

nB
= 6.1 1010 .
B
n
(58)
2
3
If we recall that the entropy density for relativistic degrees of freedom is s = heff 2
45 T and that
2 (3) 3
the number density for photons is n = 2 T , one easily obtains for today that s = 7.04n .
Thus, using (57) we have
1.08 106
GeV
| sin L |
from which we get a lower bound for the reheating temperature
TR =
(59)
TR 1.08 106 GeV.
(60)
4. Results
Let us summarize the results obtained above. We take M5 and M to be two independent mass
scales, in principle anywhere between the four-dimensional Planck mass Mpl and the electroweak scale,
v 200 GeV. First we present all the constraints that have to be satisfied. We have mentioned that 1
and that in the GB regime m4 . These lead to the constraints
M
2M53
2
Mpl
(61)
and
M M,
(62)
respectively. On the other hand, the sneutrino drives inflation and simultaneously produces the lepton asymmetry through its direct out-of-equilibrium decay after the inflationary era. This requires the reheating
temperature to be smaller than the sneutrino inflaton mass, namely TR < M. Furthermore, the gravitino
abundance constraint requires TR T0 . So we see that the reheating temperature has to be lower than
both M and T0 . Now the question arises, whether M is larger than T0 or vise versa. We have checked that
for M5 and M in their allowed range, M is always larger than T0 . Thus, the requirement that TR T0 also
guarantees that TR < M. Hence, for given M5 and M , the reheating temperature is bounded both from
below and from above as follows
1.08 106 GeV TR T0 .
(63)
Of course, T0 should not be lower than the minimum of the reheating temperature
T0 1.08 106 GeV.
(64)
Combining all the constraints mentioned above we find an upper bound for M
M 3 1011 GeV.
(65)
Then, for a given value for M , M5 has to range between a maximum and a minimum value. If M5 gets
too small, the tensor-to-scalar ratio gets larger than the observed value, while if M5 gets too large, then the
constraint (61) or (62) is not satisfied. For example, for the extremum values of M :
59
For M = 3 1011 GeV

1.31 1015 GeV M5 2.42 1015 GeV.
(66)
While for M = 200 GeV

9.97 108 GeV M5 5.3 1012 GeV.
(67)
We see that M5 can be very close to the unification scale MGUT 1016 GeV (but remains lower than
that) and not lower than 108 GeV. Interestingly, our findings are compatible with experiments to probe
deviations from Newtons law, which currently imply that M5 108 GeV [19]. Finally, for all the allowed
values of M5 and M , we find that the constraint (64) is always satisfied and that the tensor perturbations
are always negligible.
So far we have treated M as a phenomenological parameter of the model. However, the GB coupling
2
2
) [29]. Thus, M2 = 8Mstring
.
is related to the string mass scale Mstring and it is defined to be = 1/(8Mstring
M-theory seems to allow arbitrary values for the string scale. Experimental limits imply that is not lower
than O(TeV). If the string scale is around a few TeV [30], observation of novel effects in forthcoming
experiments becomes a realistic possibility (see e.g. [31]). For the special case Mstring = 7 TeV or M =
19.81 TeV we obtain
2.13 1010 GeV M5 2.45 1013 GeV.
(68)
For the minimum value M5 = 2.13 1010 GeV we obtain for M, tensor-to-scalar ratio and reheating temperature the following
r = marginal,
(69)
M = 3.28 1013 GeV
(70)
1.08 106 GeV TR 4.34 1010 GeV
(71)
and
while for the maximum value M5 = 2.45 1013 GeV we obtain

r = 3.56 107 ,
(72)
M = 2.85 1010 GeV
(73)
1.08 106 GeV TR 6.33 108 GeV.
(74)
and
Finally, for the Yukawa coupling f 2 we find

for M5 = 2.13 1010 GeV, f 2 < 6.79 102 ,
while for M5 = 2.45 1013 GeV, f 2 < 5.63 106 .
However, phenomenological issues such as neutrino masses and axion scale, seem more natural if Mstring
is in the range of 1010 1014 GeV [32] centered around 1012 GeV. For the case Mstring 1011 GeV or
M = 3 1011 GeV as mentioned already we obtain
1.31 1015 GeV M5 1.42 1015 GeV.
(75)
For the minimum value M5 = 1.31 1015 GeV we obtain for M, tensor-to-scalar ratio and reheating temperature the following
r = marginal,
(76)
60
M = 3.27 1013 GeV
(77)
1.08 106 GeV TR 4.33 1010 GeV
(78)
and
while for the maximum value M5 = 1.42 1015 GeV we obtain

r = 0.4,
(79)
M = 3.01 1013 GeV
(80)
1.08 106 GeV TR 4.12 1010 GeV.
(81)
and
Finally, for the Yukawa coupling f 2 we find

for M5 = 1.31 1015 GeV, f 2 < 0.07,
while for M5 = 1.42 1015 GeV, f 2 < 0.06.
Note that in contrast to the standard four-dimensional [4] or to the RandallSundrum sneutrino inflation [17]
scenarios, in all cases treated above, the Yukawa coupling f 2 in the presence of the GB term need not be
unnaturally small.
5. Conclusions
In the present work we have examined sneutrino inflation in the GaussBonnet brane-world.
The GaussBonnet term appears in the low-energy effective field equations of string theories and
it is the lowest order stringy correction to the five-dimensional Einstein gravity. Inflation is driven
by the sneutrino inflaton, which is the scalar superpartner of the lightest of the heavy singlet neutrinos, that might explain in a natural way the tiny neutrino masses via the seesaw mechanism.
The sneutrino inflaton, apart from driving inflation, also produces the lepton asymmetry that
partially is converted to the baryon asymmetry via sphaleron effects. We find that we can get
a viable inflationary model that reproduces the correct amplitude for density perturbations and
predicts a nearly scale-invariant spectrum and negligible tensor perturbations. Furthermore, the
reheating temperature after inflation is such that the gravitino does not upset the BBN results
and the required lepton asymmetry is generated. Our analysis shows that all these are simultaneously achieved for a wide range of values of the five-dimensional Planck mass M5 and the mass
scale M set by the GaussBonnet coupling.
Acknowledgements
We are grateful to T.N. Tomaras for critical comments on the manuscript. This work was
supported in part by the Greek Ministry of education research program Heraklitos and by the
EU grant MRTN-CT-2004-512194.
References
[1] For a review see D.H. Lyth, A. Riotto, Phys. Rep. 314 (1999) 1, hep-ph/9807278, and references therein.
[2] C.L. Bennett, et al., Astrophys. J. Suppl. 148 (2003) 1, astro-ph/0302207;
WMAP Collaboration, D.N. Spergel, et al., Astrophys. J. Suppl. 148 (2003) 175, astro-ph/0302209.
61
[3] H.V. Peiris, et al., Astrophys. J. Suppl. 148 (2003) 213, astro-ph/0302225.
[4] J.R. Ellis, M. Raidal, T. Yanagida, Phys. Lett. B 581 (2004) 9, hep-ph/0303242.
[5] Super-Kamiokande Collaboration, Y. Fukuda, et al., Phys. Rev. Lett. 81 (1998) 1562, hep-ex/9807003;
Super-Kamiokande Collaboration, Y. Fukuda, et al., Phys. Rev. Lett. 82 (1999) 1810, hep-ex/9812009;
Super-Kamiokande Collaboration, Y. Fukuda, et al., Phys. Rev. Lett. 82 (1999) 2430, hep-ex/9812011.
[6] SNO Collaboration, Q.R. Ahmad, et al., Phys. Rev. Lett. 89 (2002) 011301, nucl-ex/0204008;
KamLAND Collaboration, K. Eguchi, et al., Phys. Rev. Lett. 90 (2003) 021802, hep-ex/0212021.
[7] M. Gell-Mann, P. Ramond, R. Slansky, in: P. van Nieuwenhuizen, D. Freedman (Eds.), Proceedings of the Supergravity Stony Brook Workshop, North-Holland, Amsterdam, 1979;
T. Yanagida, in: A. Sawada, A. Sugamoto (Eds.), Proceedings of the Workshop on Unified Theories and Baryon
Number in the Universe, Tsukuba, Japan, 1979, KEK Report No. 79-18, Tsukuba.
[8] J.R. Ellis, Nucl. Phys. B (Proc. Suppl.) 137 (2004) 190, hep-ph/0403247.
[9] S.P. Martin, hep-ph/9709356.
[10] H. Murayama, H. Suzuki, T. Yanagida, J. Yokoyama, Phys. Rev. Lett. 70 (1993) 1912;
H. Murayama, H. Suzuki, T. Yanagida, J. Yokoyama, Phys. Rev. D 50 (1994) 2356, hep-ph/9311326.
[11] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, hep-th/9906064.
[12] P. Binetruy, C. Deffayet, D. Langlois, Nucl. Phys. B 565 (2000) 269, hep-th/9905012;
P. Binetruy, C. Deffayet, U. Ellwanger, D. Langlois, Phys. Lett. B 477 (2000) 285, hep-th/9910219;
E. Kiritsis, N. Tetradis, T.N. Tomaras, JHEP 0203 (2002) 019, hep-th/0202037;
E. Kiritsis, G. Kofinas, N. Tetradis, T.N. Tomaras, V. Zarikas, JHEP 0302 (2003) 035, hep-th/0207060.
[13] D. Langlois, Prog. Theor. Phys. Suppl. 148 (2003) 181, hep-th/0209261;
R. Maartens, Living Rev. Relativ. 7 (2004) 7, gr-qc/0312059.
[14] S.C. Davis, Phys. Rev. D 67 (2003) 024030, hep-th/0208205.
[15] C. Charmousis, J.-F. Dufaux, Class. Quantum Grav. 19 (2002) 4671, hep-th/0202107;
J.E. Lidsey, N.J. Nunes, Phys. Rev. D 67 (2003) 103510, astro-ph/0303168.
[16] J.E. Kim, B. Kyae, H.M. Lee, Phys. Rev. D 62 (2000) 045013, hep-ph/9912344;
J.E. Kim, B. Kyae, H.M. Lee, Nucl. Phys. B 582 (2000) 296, hep-th/0004005;
J.E. Kim, B. Kyae, H.M. Lee, Nucl. Phys. B 591 (2000) 587, Erratum.
[17] M.C. Bento, R. Gonzalez Felipe, N.M.C. Santos, Phys. Rev. D 69 (2004) 123513, hep-ph/0402276.
[18] S. Tsujikawa, M. Sami, R. Maartens, Phys. Rev. D 70 (2004) 063525, astro-ph/0406078.
[19] J.-F. Dufaux, J.E. Lidsey, R. Maartens, M. Sami, Phys. Rev. D 70 (2004) 083525, hep-th/0404161.
[20] SDSS Collaboration, M. Tegmark, et al., Phys. Rev. D 69 (2004) 103501, astro-ph/0310723.
[21] N. Straumann, hep-ph/0505249.
[22] K. Hamaguchi, H. Murayama, T. Yanagida, Phys. Rev. D 65 (2002) 043512, hep-ph/0109030.
[23] J.R. Ellis, A.D. Linde, D.V. Nanopoulos, Phys. Lett. B 118 (1982) 59;
M.Y. Khlopov, A.D. Linde, Phys. Lett. B 138 (1984) 265.
[24] R.H. Cyburt, J.R. Ellis, B.D. Fields, K.A. Olive, Phys. Rev. D 67 (2003) 103521, astro-ph/0211258.
[25] M. Bolz, A. Brandenburg, W. Buchmuller, Nucl. Phys. B 606 (2001) 518, hep-ph/0012052.
[26] V.A. Kuzmin, V.A. Rubakov, M.E. Shaposhnikov, Phys. Lett. B 155 (1985) 36.
[27] J.A. Harvey, M.S. Turner, Phys. Rev. D 42 (1990) 3344.
[28] S. Davidson, A. Ibarra, Phys. Lett. B 535 (2002) 25, hep-ph/0202239.
[29] G. Kofinas, Class. Quantum Grav. 22 (2005) L47, hep-th/0412299.
[30] I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G.R. Dvali, Phys. Lett. B 436 (1998) 257, hep-ph/9804398;
I. Antoniadis, E. Kiritsis, T.N. Tomaras, Phys. Lett. B 486 (2000) 186, hep-ph/0004214.
[31] S. Dimopoulos, G. Landsberg, Phys. Rev. Lett. 87 (2001) 161602, hep-ph/0106295;
E. Kiritsis, P. Anastasopoulos, JHEP 0205 (2002) 054, hep-ph/0201295;
L.A. Anchordoqui, J.L. Feng, H. Goldberg, A.D. Shapere, Phys. Rev. D 68 (2003) 104025, hep-ph/0307228;
A. Mironov, A. Morozov, T.N. Tomaras, hep-ph/0311318, Nucl. Phys. (Russian), in press;
A. Cafarella, C. Coriano, T.N. Tomaras, JHEP 0506 (2005) 065, hep-ph/0410358.
[32] K. Benakli, Phys. Rev. D 60 (1999) 104002, hep-ph/9809582.
Tri-linear couplings in an heterotic minimal

supersymmetric Standard Model
Vincent Bouchard a , Mirjam Cvetic b, , Ron Donagi c
a The Mathematical Sciences Research Institute, 17 Gauss Way, Berkeley, CA 94720-5070, USA
b Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104-6396, USA
c Department of Mathematics, University of Pennsylvania, Philadelphia, PA 19104-6395, USA
Received 26 March 2006; accepted 31 March 2006

Abstract
We calculate, at the classical level, the superpotential tri-linear couplings of the only known globally
consistent heterotic minimal supersymmetric Standard Model [V. Bouchard, R. Donagi, An SU(5) heterotic
standard model, Phys. Lett. B 633 (2006) 783791, hep-th/0512149]. This recently constructed model is
based on a compactification of the E8 E8 heterotic string theory on a CalabiYau threefold with Z2
fundamental group, coupled with a slope-stable holomorphic SU(5) vector bundle. In the observable sector
the massless particle content is that of the three-family supersymmetric Standard Model with n = 0, 1, 2
massless Higgs pairs, depending on the location in the vector bundle moduli space, and no exotic particles.
We obtain non-zero Yukawa couplings for the three up-sector quarks, and vanishing R-parity violating
terms. In particular, the proton is stable. Another interesting feature is the existence of tri-linear couplings,
on the loci with massless Higgs pairs, generating -mass parameters for the Higgs pairs and neutrino mass
terms, with specific vector bundle moduli playing the role of right-handed neutrinos.
1. Introduction
Consistent four-dimensional solutions of string theory play an important role as a link between
string theory and particle physics. One particular goal of this program is to derive consistent supersymmetric solutions of string theory with the massless particle content of the three-family
E-mail addresses: vincentb@msri.org (V. Bouchard), cvetic@physics.upenn.edu, cvetic@cvetic.hep.upenn.edu

(M. Cvetic), donagi@math.upenn.edu (R. Donagi).
doi:10.1016/j.nuclphysb.2006.03.032
V. Bouchard et al. / Nuclear Physics B 745 (2006) 6283
63
Standard Model. On the heterotic string theory side, compactifications of the E8 E8 heterotic
string theory on CalabiYau threefolds with stable holomorphic vector bundles yielded large
classes of supersymmetric three-family standard-like models (see [2] and references therein).
For the minimal supersymmetric standard-like model constructions with heterotic conformal
field theory orbifolds, see [3]. On the other hand, within the framework of type II string theory, compactifications with intersecting D-branes on toroidal orbifolds also provide a wealth of
three-family supersymmetric standard-like model constructions where non-Abelian gauge symmetry, chirality and family replication have a geometric origin (see [4] for a review and references
therein). For standard model-like constructions with rational conformal field theory orientifolds,
see [5] and references therein.
In spite of significant progress made in developing the techniques for these constructions and
a proliferation of models with semi-realistic particle physics features, most of the constructions
suffer from some phenomenological deficiencies. In addition to the minimal supersymmetric
Standard Model (MSSM) particle content, the models typically possess more than one Higgs
doublet pair and additional Standard Model chiral exotics. Furthermore, consistent constructions
with the supersymmetric Standard Model particle content and stabilized moduli remain elusive,
though there has been progress made on the type IIB side (see, e.g., [4] and references therein).
Another important test of these constructions is at the level of couplings. In particular, the
tri-linear couplings of the matter chiral superfields test the string theory predictions for Yukawa
couplings of the Standard Model. For the heterotic string theory compactifications on Calabi
Yau threefolds such couplings can be calculated, in the classical limit, by determining the nonzero triple pairings of the cohomology group elements which determine the massless particle
spectrum. There has also been progress made in the calculation of quantum contributions to such
couplings, i.e., non-perturbative worldsheet instanton contributions, which involve the pairing of
quantum cohomology groups (see, e.g., [6] and references therein). Within this framework there
remains an outstanding problem of determining the moduli dependence of the Khler potential
for the matter chiral superfields, which in turn determines the normalization of the kinetic energy
terms for the matter fields. On the type IIA side, for toroidal orbifold compactifications with
intersecting D6-branes, conformal field theory techniques allow for the full tree-level calculations
of such couplings, including the superpotential and Khler potential contributions to Yukawa
couplings (see [7] and references therein).
Recently, major progress has been achieved by constructing a specific, globally consistent
supersymmetric solution of heterotic string theory, that yields a massless spectrum with minimal
supersymmetric Standard Model particle content and no exotic particles [1]. The construction
is obtained by compactifying the E8 E8 heterotic string theory on a CalabiYau threefold
X with Z2 fundamental group, coupled with a stable holomorphic SU(5) vector bundle V . In
the observable sector the massless particle content is that of the three-family supersymmetric
Standard Model (MSSM) with n = 0, 1, 2 massless Higgs pairs. The value of n depends on the
location in the vector bundle moduli space. The model also possesses a number of Khler and
complex structure moduli of the CalabiYau threefold, and a number of vector bundle moduli.
The GreenSchwarz anomaly cancellation (global consistency) condition requires that the
difference of the second Chern classes c2 (T X) c2 (V ) c2 (U ) = [W ], where V is the visible
bundle and U the hidden bundle, be the effective class [W ] of an holomorphic curve, around
which M5-branes are wrapped. In the construction of [1] the hidden bundle is chosen to be
the trivial bundle, which implies that the hidden sector has the spectrum of N = 1 E8 superYangMills theory. In this case, the anomaly condition requires that c2 (T X) c2 (V ) = [W ] be
64
effective, which is indeed satisfied in this particular construction. In conclusion, the model of [1]
is a manifestly supersymmetric, globally consistent solution of heterotic string theory.
Another heterotic string theory construction [8] that yields the massless spectrum of the
MSSM (with an extra U (1) factor), based on a slope-stable visible vector bundle V [9], has
been claimed. However, as it stands, this construction does not satisfy the GreenSchwarz anomaly cancellation condition, and thus it is not globally consistent. More precisely, in this model
c2 (T X) c2 (V ) = [W ] is not an effective class. Therefore, the hidden bundle U cannot be
trivial, and one must find a slope-stable hidden vector bundle U such that c2 (T X) c2 (V )
c2 (U ) = [W ] is effective. Currently, there are no known examples of such slope-stable vector
bundles; it was shown in [10] that the proposed hidden bundles U that satisfy the GreenSchwarz
anomaly cancellation condition [8,11], are not slope-stable. It was further suggested in [9] that
the introduction of anti-M5-branes may cancel the GreenSchwarz anomaly. However, in this
context the introduction of anti-M5-branes should render the solution physically unstable, leading to annihilation processes with M5-branes and/or vector bundle states.
The main purpose of this paper is to confront the predictions of the globally consistent model
of [1] at the level of tri-linear superpotential couplings for the chiral matter superfields and to
study the implications of these couplings for particle phenomenology. The calculation of the
tri-linear superpotential couplings for the matter superfields is performed in the classical limit.
The model has non-zero Yukawa couplings for all three up-sector quarks, and depending on the
location in the moduli space a suitable mass hierarchy for the up-sector quarks may be achieved.
At the classical level the couplings to the down-quarks are zero, though the expectation is that
such couplings would be non-zero at the quantum level. Moreover, all R-parity violating couplings vanish at tree level. In particular, baryon number and lepton number violating processes
that could lead to physically unacceptable rapid proton decay are absent; thus, at tree level the
proton is stable (see [12]). The requirement that R-parity violating couplings also vanish at the
quantum level should further constrain phenomenologically viable solutions by restricting the
allowed subspace of vector bundle moduli.
Another interesting feature of these model is the existence of tri-linear couplings of the upHiggs fields, down-Higgs or lepton doublet fields, and vector bundle moduli. We calculate these
couplings on the loci with n = 1 and n = 2 massless Higgs pairs. Our results demonstrate that
specific directions in moduli space, perpendicular to the locus with n massless Higgs pairs, generate -terms for the Higgs pairs. In the effective field theory, this is demonstrated by giving
a non-zero vacuum expectation value to a specific linear combination of vector bundle moduli
which in some tri-linear couplings generates the parameters for the n Higgs pairs. There are
also additional couplings of the down-Higgs fields to one lepton doublet and specific vector bundle moduli, which can be interpreted as right-handed neutrinos. These couplings in turn provide
mass terms for neutrinos.
The paper is organized as follows. In Section 2 we summarize the construction of [1]. In
Section 3 we briefly review the features of the MSSM, and the low-energy physics obtained
in the minimal supersymmetric Standard Model of the heterotic string. Section 4 is devoted
to the computation of the vector bundle moduli of the model. This section provides important
prerequisite results for the superpotential coupling calculations. In Section 5 we compute the
tri-linear superpotential couplings. We give a phenomenological interpretation of the effective
theory couplings in Section 6.
65
2. Construction
In this section we briefly review the SU(5) heterotic standard model introduced in [1]. More
details can be found in [1], and in previous papers [2,1315] where different bundles on the same
manifold were constructed.
In this model we compactify the E8 E8 heterotic string theory on a CalabiYau threefold
X with Z2 fundamental group. Moreover, we construct a stable SU(5) bundle on X which is
twisted by a Z2 Wilson line to break the visible E8 gauge group to the standard model gauge
group SU(3)C SU(2)L U (1)Y .
2.1. The manifold
The non-simply connected CalabiYau threefold X is constructed by considering a simply
elliptically fibered over a rational elliptic surface B, that
connected CalabiYau threefold X,
is a CalabiYau
admits a free F = Z2 action preserving the fibration. The quotient X = X/F
threefold, has fundamental group F and is a genus-one fibration.
Let B be a rational elliptic surface, and let X be a CalabiYau threefold with an elliptic fibration : X B (we also require that has a section). The manifold X also admits a description
as a fiber product B P1 B of two rational elliptic surfaces B and B over P1 :

X = (p, p ) B B | (p ) = (p) ,
(2.1)
where : B P1 and : B P1 are the elliptic fibrations of the rational elliptic surfaces B
and B .
Thus X can be described by the following commuting diagram
X
B
(2.2)
P1
The two rational elliptic surfaces B and B are chosen such that they lie in the four-parameter
family of rational elliptic surfaces described in [2,13]. Both of them admit a Z2 involution B
and B respectively, which lift to a free Z2 involution := B P1 B on X.

2.2. The bundle
To get an SU(5) bundle V on X, we construct an SU(5) bundle V on X together with an
action of the involution on V .
Instead of working directly with the bundle V , in the following we will consider its dual V ,
since in that case we can apply directly the results of [2,14]. The bundle V is constructed as an
extension
0 V2 V V3 0,
where V2 and V3 are rank 2 and 3 bundles, respectively.
(2.3)
66
The bundles Vi are given by

Vi = Wi Li ,
(2.4)
where the Li are some line bundles on B and the Wi are rank i bundles on B given by the
FourierMukai transforms Wi = FM B (Ci , Ni ): as usual in the spectral cover construction, the
Ci B are curves in B and the Ni Pic(Ci ) are line bundles over Ci .
We choose the following data:

C 2 OB (2e9 + 2f ),

C3 OB (3e9 + 3f ),
C2 = C 2 + f ,
N2 Pic3,1 (C2 ),
N3 Pic7 (C3 ),
L2 = OB (3r ),
L3 = OB (2r ).
f is the smooth fiber of at containing the four fixed points of B , and
line bundles of degree 3 over C 2 and degree 1 on f . Finally, r is given by1
r = e1 + e4 e5 + e9 + f .
(2.5)
Pic3,1 (C
2)
denotes
(2.6)
It was shown in [1] that V is slope-stable and invariant under the Z2 involution. Its cohomology was computed in [1], and it leads to exactly the MSSM massless particle spectrum, with no
c2 (V )
exotic particles. Furthermore, it satisfies the anomaly cancellation condition: c2 (T X)
is an effective class around which M5-branes wrap to cancel the anomaly. This means that we
are in the strongly coupled regime of the heterotic string. It may also be possible to add a gauge
instanton U of small rank in the hidden sector such that c2 (U ) = 2f pt + 6pt f , which
would give a weak coupling vacuum of our model.
In conclusion, the manifold X with the Z2 -invariant stable SU(5) bundle V is a good candidate for a realistic compactification of the heterotic string, at least at the level of the massless
particle spectrum. The aim of this paper is to compute the tri-linear couplings in the low-energy
superpotential of this model.
3. Massless spectrum and couplings of the effective theory
In the next subsection we briefly summarize salient features of the MSSM and compare them
with those of the effective theory for the heterotic supersymmetric Standard Model. We further
discuss the massless spectrum and classical superpotential couplings for the specific heterotic
string model and confront it with the features of the MSSM in the subsequent subsections.
3.1. Minimal supersymmetric Standard Model
The MSSM is the minimal N = 1 supersymmetric extension of the Standard Model (SM).
It has gauge group SU(3)C SU(2)L U (1)Y . The massless spectrum of N = 1 superfields and
1 For an explanation of the notation see [1].
67
Table 1
The particle spectrum of the MSSM. Only the left-chiral fields are shown. The right-chiral fields have conjugate representations under the gauge group
Superfield
Symbol
Representation
Quarks
Anti-up
Anti-down
Leptons
Anti-leptons
Up Higgs
Down Higgs
Color gauge fields
Weak gauge fields
Q
u
d
L
e
H
H
G
W , Z,
(3, 2)1/3
1)4/3
(3,
1)2/3
(3,
(1, 2)1
(1, 1)2
(1, 2)1
(1, 2)1
(8, 1)0
(1, 3 + 1)0
their representations of the SM gauge group are given in Table 1. In particular, the matter chiral
superfields consist of three families of quarks and leptons and one Higgs doublet pair.
The four-dimensional N = 1 supersymmetric Lagrangian is fully specified by a Khler potential K, a superpotential W and gauge kinetic functions fi . The Khler potential is a real function
of chiral superfields, while the superpotential and the gauge kinetic functions are in general
holomorphic functions of chiral superfields. The Khler potential, among others, determines the
kinetic energy terms for chiral superfields, the superpotential carries information on the Yukawa
couplings of these superfields and the gauge kinetic functions determine gauge couplings for the
super-YangMills sector of the theory.
In the MSSM, the matter particle content is that of three families of quarks and leptons and one
Higgs doublet pair, supplemented by the supersymmetric partners to form chiral superfields. The
Khler potential for these chiral superfields is chosen so that their kinetic energy is canonically
normalized. On the other hand, in string theory the Khler potential for matter chiral superfields
depends on moduli fields. It also receives corrections at the higher genus level. In the heterotic
string context it may be possible to determine these couplings in the classical limit, by obtaining
leading contributions in the limit of large Khler and complex structure CalabiYau moduli.
We postpone the study of moduli dependence of the Khler potential.
In the MSSM the gauge kinetic functions are fixed to specific constant values that match
the experimental values for gauge couplings. On the other hand, in string theory gauge kinetic
functions are holomorphic functions of the dilaton (string coupling modulus) and moduli fields.
In the heterotic string theory, at the tree level, the gauge functions are universal and proportional
to the dilaton field.
In the MSSM the superpotential of the matter chiral superfields can have the following trilinear (renormalizable) couplings:
W = W1 + W2 ,
(3.1)
where
ij
ij
ij
W1 = l ei Lj H + d Qi dj H + u Qi uj H + H H ,
ij k
ij k
ij k
W2 = 1 Li Lj ek + 2 Li Qj dk + 3 ui dj dk ,
(3.2)
i, j, k being generation indices. The terms in W1 conserve baryon and lepton numbers, while
those in W2 do not. The latter couplings can be set to zero by imposing R-parity symmetry. In
this case, the lepton and baryon violating terms are absent and the lightest superparticle (LSP)
68
of the MSSM is stable, and is a weakly interacting massive particle (WIMP); thus a good dark
matter candidate. The three first terms of W1 determine Yukawa couplings between quarks and
Higgs fields, and leptons and Higgs fields. After electroweak symmetry breaking, i.e., when
Higgs doublets H and H acquire non-zero vacuum expectation values (VEVs), these terms give
masses to quarks and leptons. One of the main goal of this paper is to determine the superpotential
Yukawa couplings for quarks and leptons.
The last term of W1 is sometimes referred to as a bare Higgs -term. Since in string theory
all the couplings are field dependent, such bare -mass term is absent.
In the original formulation of the MSSM there are no right-handed neutrinos, and the lefthanded neutrinos are massless. However, there is now considerable experimental evidence for
neutrino oscillations, which require non-zero neutrino masses. There are various ways to modify
the MSSM in order to give a mass to the neutrinos. Perhaps the simplest way is to add a Majorana
mass term for the left-handed neutrinos. In string theory this is typically hard to achievefor a
recent study of these issues within heterotic string orbifolds see [16].
Another possibility is to introduce right-handed neutrinos Ri , associated with the chiral superfields that are gauge singlets, i.e., transform in the (1, 1)0 representation of the SM gauge
group. These fields can couple via tri-linear couplings to lepton doublets Lj and the H Higgs
field:
W = ij
Ri Lj H ,
(3.3)
which, after electroweak symmetry breaking generate non-zero neutrino masses.

In fact, we can extend further the MSSM by adding more superfields i which are singlets of
the SM gauge group, with additional superpotential couplings:
W = i i H H + ij k i j k .
(3.4)
If the i acquire non-zero VEVs the first term induces an effective parameter for the Higgs
doublet pairs: such extensions of the MSSM are often referred to as next to minimal supersymmetric Standard Model (NMSSM).
In string theory the role of chiral superfields which are singlets of the SM gauge group, such
as the fields Ri and i , can potentially be played by moduli fields. In particular, in our specific
heterotic string theory construction we shall show that the tri-linear couplings of W and the
first coupling in W do exist, and the fields Ri and j are identified with specific vector bundle
moduli. On the other hand, the second term in W , i.e., the tri-linear couplings of vector bundle
moduli fields i should be zero. In perturbation theory the moduli do not have self-interactions.
It is non-perturbative effects, such a gauge instantons in the E8 gauge sector, that are expected to
introduce non-perturbative superpotential for moduli fields, thus allowing their stabilization. But
this is a topic beyond the scope of the paper.
At energies well below the string scale, supersymmetry is broken. At low energies supersymmetry breaking effects manifest themselves as soft supersymmetry breaking mass terms for the
SM matter fields. In string theory, supersymmetry breaking and moduli stabilization can in principle be addressed by studying the strong gauge dynamics associated with hidden sector gauge
instantons or M5-branes wrapping the effective curves. We do not address this difficult task in
this paper; we shall only focus on the string construction, which at the string energy scale represents the (stable) four-dimensional supersymmetric solution of string theory.
69
3.2. The MSSM from heterotic string theory

We now review the calculation of the massless spectrum of the heterotic MSSM construction.
More details may be found in [17].
The particle spectrum of the E8 E8 heterotic string consists in the zero-modes of the tendimensional Dirac operator, ker(D).
/ Define

Spec =
H q (X, ad V );
(3.5)
q=1,3
/ is given by adding the duals to Spec [2]. In fact, Spec gives the left-chiral superfields,
then ker(D)
while its dual gives the right-chiral superfields.
We compactify the E8 E8 heterotic string theory on a CalabiYau threefold X with fundamental group Z2 . We also construct an SU(5) bundle V on X, which breaks the visible E8 gauge
group to an SU(5) gauge group. Then, using a Z2 Wilson line, we break the SU(5) grand unified
gauge group down to the MSSM gauge group SU(3)C SU(2)L U (1)Y .
The resulting low-energy superfields are given by the decomposition of Spec under the above
symmetry breaking pattern. In particular, the multiplicity of the representations of the low-energy
MSSM gauge group are given by the dimensions of the invariant and anti-invariant parts of
some cohomology groups. For the case under consideration, the decomposition of Spec and the
associated cohomology groups has been worked out in [2]. The resulting low-energy spectrum is
shown in Table 2.
We computed the required cohomology groups in [1], and found the multiplicities also presented in Table 2. Notice that the low-energy spectrum has three generations of quarks and
leptons, no exotic particles, 0, 1 or 2 pairs of Higgs, and 51 vector bundle moduli fields.2
Table 2
The particle spectrum of the low-energy SU(3)C SU(2)L U (1)Y theory. Notice that all exotic particles come with 0
multiplicity, and that the spectrum include n copies of Higgs conjugate pairs, where n = 0, 1, 2
Multiplicity
Representation
Superfield
O )+
1 = h3 (X,
X
V )+
3 = h1 (X,
(8, 1)0 (1, 3)0 (1, 1)0

1)4/3 (1, 1)2
(3,
G, W , Z,
(3, 2)1/3
(3, 1)4/3 (1, 1)2

2)1/3
(3,
exotic
V )
3 = h1 (X,
1
V )+
0 = h (X,
V )
0 = h1 (X,
1
2 V )+
3 = h (X,
2 V )
3 + n = h1 (X,
1
2
V )+
0 = h (X,
2 V )
n = h1 (X,
1
ad V )+
51 = h (X,
u, e
exotic
1)2/3
(3,
1
(1, 2)
(3, 1)2/3
exotic
(1, 2)1
(1, 1)0
L, H
2 In fact we did not compute the dimension of H 1 (X,

ad V ) in [1]; we simply gave an estimate of its dimension from
simple parameter counting. We will study this cohomology group in more detail in Section 4.
70
3.3. Superpotential
The main focus of this paper is to compute the superpotential W of the model. More precisely,
we want to determine which terms in W1 , W2 , W and W are non-vanishing. Computing the
exact numerical coefficients and their explicit dependence on the (vacuum expectation) values of
the moduli is harder and we shall not do that explicitly. In addition our calculation is done only
in the classical limit, and thus the quantum (world-sheet instanton) effects are not included.
We have seen that massless superfields correspond to equivalence classes in some cohomology
groups of some bundles over our CalabiYau threefold X. In other words, we can associate to
each superfield a -closed

(0, 1)-form i taking values in some bundle over X. Compactifying
heterotic string theory on X yields cubic terms in the superpotential of the four-dimensional
effective action. The coefficients of these terms are given by the unique way of extracting a
complex number out of the three associated (0, 1)-forms i , that is, by wedging the three (0, 1)forms and the holomorphic volume (3, 0)-form of X and integrating over X:

ij k i j k .
(3.6)
X
Note that this is only a term determined at the tree level of sigma model perturbation, that is in
the large volume limit; these coefficients can receive corrections due to worldsheet instantons
but we will not compute them in this paper.
In cohomological language, the coefficients are given by the images in C of some triple pairings of cohomology groups. Thus, to compute the coefficients in the superpotential we must first
find all possible triple pairings of cohomology groups3 in Table 2 mapping into C; these are the
cubic terms that may appear in the superpotential of the four-dimensional effective action. Then
we must show whether these pairings vanish or not.
Using the multiplicities given in Table 2, in particular the fact that all exotic particles have
multiplicity zero, we find the following allowed triple pairings4 :

(3,3+n)

(3,3+n)
(3,3)

(1,0)
H 1 2 V
H 1 V
H 3 5 V
C,
(d) H 1 2 V

(3,0)
(0,3)
(0,n)
(u) H 1 V + H 1 V H 1 2 V H 3 (O)(1,0) C,
(0,n)

(0,3+n)

(51,0)
() H 1 (ad V )+
H 1 2 V
H 1 2 V H 3 (O)(1,0) C,
()
(51,0)
(51,0)
(51,0)
H 1 (ad V )+
H 1 (ad V )+
H 1 (ad V )+
H 3 (O)(1,0) C,
(3.7)
where the pairings are given by cup product and wedge product, and n = 0, 1, 2 (note that for
n = 0 the (u) and () pairings vanish identically since there is no Higgs pair). The superscripts
(x, y) mean that the invariant part of the cohomology group has dimension x, while the antiinvariant part has dimension y. Each of these pairings correspond to various cubic couplings in
the superpotential; they can be read off from the associated superfields presented in Table 2. The
names we gave to the triple pairings are of physical significance, as will be explained in more
detail in Section 6. Jumping ahead a little, let us simply say that the (d) pairing corresponds
to couplings of down sector quarks and charged lepton sector to the down-Higgs doublet, and
3 We do not consider the group H 3 (X,
O ), since it corresponds to the gauge connections and thus determine nonX
Abelian gauge couplings; our focus is on the superpotential tri-linear couplings of the matter chiral superfields.
4 In the following we will sometimes suppress the X
in the cohomology groups for clarity.
71
also to the potential R-parity violating couplings of W2 in (3.2). The (u) pairing is related to the
Yukawa couplings of the up-sector quarks to the up-Higgs doublet. The () pairing corresponds
to the moduli-dependent Higgs -terms (the first term of W in (3.4)) and potential neutrino
mass terms (W in (3.3)). Finally, the () pairing corresponds to the tri-linear couplings of the
vector bundle moduli (the second term of W in (3.4)).
4. Vector bundle moduli
In this section we study the vector bundle moduli space H 1 (ad V ), which enters into the
calculation of the tri-linear couplings.
In fact, we only need its invariant subspace, which we denote by H 1 (ad V )+ , as shown in
Table 2. Note that ad V is defined to be the traceless part of V V . But since H 1 (O) = 0, we
can identify

H 1 (ad V )+ H 1 V V + .
(4.1)
Before doing anything, we must understand the cohomology H (V2 V3 )+ . We computed
in [1] that5

h1 V2 V3 + = 45.
(4.2)
Using similar techniques, it is straightforward to show that

h0 V2 V3 + = h2 V2 V3 + = h3 V2 V3 + = 0.
(4.3)
We are now ready to attack the computation of H 1 (V V )+ . Recall that

0 V2 V V3 0
(4.4)
implies the long exact sequence in cohomology

H 0 (V3 ) H 1 (V2 ) H 1 (V ) H 1 (V3 ) H 2 (V2 ) .
(4.5)
Tensoring (4.4) with V3 , taking the long exact sequence in cohomology (keeping only the invariant subspaces of the cohomology groups) and using the results above we obtain the exact
sequence

d1

0 H 0 V V3 + H 0 V3 V3 + H 1 V2 V3 +

H 1 V V3 + H 1 V3 V3 + 0.
(4.6)
Only the trace part of V3 V3 contributes to H 0 , and so h0 (V3 V3 )+ = h0 (O)+ = 1. Thus, the
map d1 that we identified in the long exact sequence is simply multiplication of constant sections
by the invariant extension class of our bundle in H 1 (V2 V3 )+ . Since our bundle is a non-trivial
extension, the map d1 is non-zero, and therefore must have rank 1. Hence

H 0 V V3 + = 0,
(4.7)
5 Note that in [1] we stated that the invariant subspace of the 90-dimensional cohomology space is 50-dimensional, but
a careful analysis shows that it is rather 45-dimensional.
72
and the sequence reduces to

0 H 0 V3 V3 + H 1 V2 V3 + H 1 V V3 + H 1 V3 V3 + 0.
(4.8)
From (4.8) we see that
V3 )+ and
F1 =
H 1 (V
V3 )+
has a filtration
H 1 (V2 V3 )+
.
H 0 (V3 V3 )+
F0
F1
{0}, with
F0
= H 1 (V
(4.9)
Its associated graded vector space is

H 1 (V V3 )+
H 1 (V2 V3 )+
H 0 (V3 V3 )+ H 1 (V2 V3 )+ /H 0 (V3 V3 )+

H 1 (V2 V3 )+
H 1 V3 V3 + .
H 0 (V3 V3 )+
(4.10)
Let us now tensor the exact sequence (4.4) with V2 and take the long exact sequence in
cohomology (keeping again only the invariant subspaces). Using Serre duality, we obtain

0 H i V2 V2 + H i V V2 + 0,
(4.11)
for i = 0, 1, which implies that

H i V V2 + H i V2 V2 + ,
(4.12)
for i = 0, 1.
Consider now the short exact sequence dual to (4.4), and tensor it with V . Taking the associated long exact sequence in cohomology we obtain

d2

0 H 0 V V + H 0 V V2 + H 1 V V3 +

H 1 V V + H 1 V V2 + H 2 V V3 + .
(4.13)
We know that h0 (V V2 )+ = h0 (V2 V2 )+ = 1. Thus h0 (V V )+ is 0 or 1. But it cannot
be 0, since V V = O (V V )traceless and h0 (O)+ = 1. Thus h0 (V V )+ = 1, which
implies that rank(d2 ) = 0 and the sequence splits.
Now we must understand the coboundary map . It is given by cup product with the invariant
extension class

H 1 V V2 + H 1 V2 V3 + H 2 V V3 + .
(4.14)
But since H 1 (V V2 )+ H 1 (V2 V2 )+ , the image must lie in H 2 (V2 V3 )+ , which is zero.
Thus = 0, and we obtain the short exact sequence

0 H 1 V V3 + H 1 V V + H 1 V V2 + 0.
(4.15)
Using all these results, we conclude that the space of vector bundle moduli H 1 (V V )+ has
a filtration G0 G1 G2 {0}, with G0 = H 1 (V V )+ , G1 = H 1 (V V3 )+ and
G2 =
H 1 (V2 V3 )+
.
H 0 (V3 V3 )+
(4.16)
73
Its associated graded vector space is given by

H 1 (V2 V3 )+
H 1 V3 V3 + H 1 V2 V2 + ,
0
H (V3 V3 )+
(4.17)
where we used the fact that

H 1 (V V )+
H 1 V2 V2 +
.
H 1 (V V3 )+
(4.18)
This result is easy to understand. The first factor of (4.17) is the invariant extension space,
quotiented by a one-dimensional space. This simply means that we are free to choose any invariant extension class for our bundle V , but rescaling does not change the bundle. The second factor
corresponds to moduli coming from the vector bundle V3 , while the third factor corresponds to
moduli coming from the vector bundle V2 .
4.1. Dimension
In order to get the dimension of the space of vector bundle moduli H 1 (V V )+ we simply
have to add up the dimensions of the three spaces in the associated graded vector space (4.17).
First, we computed in [1] that the invariant subspace of H 1 (V2 V3 ) is 45-dimensional. Thus
the first factor of (4.17) is 44-dimensional. We now compute the dimension of the two other
cohomology spaces in (4.17). Let us first study H 1 (V3 V3 )+ .
To start with, we can use a Leray spectral sequence to show that

V3 V3 H 1 B, W3 W3 .
H 1 X,
(4.19)
Hence we must compute the cohomology H (B, W3 W3 ) on the rational elliptic surface B.
Only the trace part contributes to H 0 , and we obtain

h0 B, W3 W3 = 1.
(4.20)
Using Serre duality and a Leray spectral sequence, we also find

H 2 B, W3 W3 H 0 B, W3 W3 O(f )

H 0 P1 , O(1) W3 W3
= 0.
(4.21)
To compute the remaining cohomology group H 1 (B, W3 W3 ), we use the Hirzebruch

RiemannRoch theorem (or index theorem). For a rational elliptic surface B, we have that
td(B) = 1 + f + pt.
(4.22)
The Chern character of W3 was computed in [1], which gives

ch W3 W3 = ch(W3 ) ch W3 = (3 + f 3pt) (3 f 3pt) = 9 18pt.
(4.23)
Then, HirzebruchRiemannRoch tells us that

h0 h1 + h2 = 1 h1 = deg (1 + f + pt) (9 18pt) 2 = 9,
(4.24)
which implies that

V3 V3 = h1 B, W3 W3 = 10.
h1 X,
(4.25)
74
We can do exactly the same computation for V2 V2 , and we obtain

V2 V2 = h1 B, W2 W2 = 9.
h1 X,
(4.26)
Now we must find the dimension of the invariant subspaces. In order to do so, let us give a
geometrical description of these cohomology groups. H 1 (B, W3 W3 ) (and similarly for W2 )
is the space of vector bundle moduli for W3 . From the spectral cover construction, we see that it
splits into two components; the dimension of the (projectivization) of the linear system of which
the spectral curve C3 is an element, and the genus of the curve C3 . More precisely, in this case
the linear system is [14]

|3e + 3f | = H 0 B, O(3e + 3f ) H 0 P1 , O(3) H 0 P1 , O(1) H 0 P1 , O ,
(4.27)
which has dimension 7. The arithmetic genus of C3 is 4, and thus

h1 B, W3 W3 = 4 + 7 1 = 10,
(4.28)
which is indeed correct. In the case of W2 , the linear system is given by

|2e + 3f | = H 0 B, O(2e + 3f ) H 0 P1 , O(3) H 0 P1 , O(1) ,
(4.29)
which has dimension 6, and the curve C2 has arithmetic genus 4. Hence

h1 B, W2 W2 = 4 + 6 1 = 9.
(4.30)
Using the decomposition of the linear systems in terms of cohomology groups on P1 ,

we can find the dimension of the invariant and anti-invariant parts. The 7-dimensional space
H 0 (B, O(3e + 3f )) breaks into a 4-dimensional invariant subspace and a 3-dimensional
anti-invariant subspace, while the 6-dimensional space H 0 (B, O(2e + 3f )) breaks into a
3-dimensional invariant subspace and a 3-dimensional anti-invariant subspace. The remaining
step is to find the genus of the quotient curves.
As described in [14], the involution B on B fixes the fiber f0 above 0 P1 pointwise, and
has 4 isolated fixed points in the fiber f above P1 . An invariant curve C3 |3e + 3f |+
intersects f at three of these four fixed points, and intersects f0 in three points. Thus the action
of B on C3 has 6 fixed points. Using Hurwitz theorem, we find
2g(C3 /B ) 2 =
1
1
2g(C3 ) 2 6 = (8 2 6) = 0,
2
2
(4.31)
which implies
g(C3 /B ) = 1.
(4.32)
In the case of C2 |2e + 3f |+ , a generic curve consists in a reducible curve which contains
f plus a curve C 2 |2e + 2f |+ . The quotient of such a curve is an elliptic curve plus a rational
curve attached to it at one point, which has arithmetic genus 1. In other words, the action of B
on C2 has 4 fixed points in f , and two fixed points at the intersection points of C2 and f0 . Thus
by Hurwitz theorem
2g(C2 /B ) 2 =
1
1
2g(C2 ) 2 6 = (8 2 6) = 0,
2
2
(4.33)
which again implies

g(C2 /B ) = 1.
(4.34)
75
Putting all these results together, we find that the invariant parts of the cohomology groups
have dimensions

h1 V3 V3 + = 1 + 4 1 = 4,

h1 V2 V2 + = 1 + 3 1 = 3.
(4.35)
Therefore, we conclude that the space of vector bundle moduli has dimension

h1 V V + = 45 1 + 4 + 3 = 51.
(4.36)
5. Computation of the triple pairings

We now compute the tri-linear couplings given by (3.7).
5.1. (d) triple pairing
Let us start by analyzing the (d) triple pairing, given by

(3,3+n)

(3,3+n)
(3,3)
(d) H 1 2 V
H 1 2 V
H 1 V
C.
(5.1)
First, recall from [1] that H 1 (V ) H 1 (V2 ), and that we have a long exact sequence

MT
0 H 1 2 V2 H 1 2 V H 1 (V2 V3 ) H 2 2 V2 H 2 2 V 0. (5.2)
Thus H 1 (2 V ) has a filtration F 0 F 1 {0}, with F 0 = H 1 (2 V ) and F 1 = H 1 (2 V2 ). Its
associated graded vector space is

H 1 (2 V )

H 1 2 V2 ker M T .
H 1 2 V2 1 2
H ( V2 )
(5.3)
If we restrict both H 1 (2 V ) in (5.1) to their subspaces H 1 (2 V2 ), the (d) pairing becomes

H 1 2 V2 H 1 2 V2 H 1 (V2 ) H 3 5 V2 = 0,
(5.4)
which vanishes since V2 has rank 2 and thus i V2 = 0 for i > 2.
Suppose now that one H 1 (2 V ) factor lives in its H 1 (2 V2 ) subspace and that the other one
lives in the quotient space
H 1 (2 V )
ker M T H 1 (V2 V3 ).
H 1 (2 V2 )
Then, the (d) pairing becomes

H 1 2 V2 H 1 (V2 V3 ) H 1 (V2 ) H 3 4 V2 V3 = 0.
Thus both factors must be in their quotient spaces. But even then, we find

H 1 (V2 V3 ) H 1 (V2 V3 ) H 1 (V2 ) H 3 3 V2 2 V3 = 0,
from which we conclude that the (d) pairing vanishes identically for n = 0, 1, 2.
(5.5)
(5.6)
(5.7)
76
5.2. (u) triple pairing

We now turn to the pairing
(3,0)
(0,3)

(0,n)
(u) H 1 V + H 1 V H 1 2 V C,
or equivalently
(3,0)
(0,3)

(0,n)
H 1 V
H 1 V
H 2 2 V
.
+
(5.8)
(5.9)
Recall that H 1 (V ) H 1 (V2 ) , and that from the long exact sequence (5.2) the map

(0,8)

(0,n)
: H 2 2 V2 H 2 2 V
(5.10)
is surjective. In fact, we can identify H 2 (2 V ) coker M T . Thus let us first analyze the
pairing

(0,8)
(3,0)
(0,3)
H 1 (V2 )+ H 1 (V2 ) H 2 2 V2
(5.11)
(0,n)
and then use to project to H 2 (2 V ) , i.e., mod out by the anti-invariant part of Im M T ,
T
which we denote by (Im M ) .
As explained in [1,2] there is a natural identification H 1 (V2 ) Sx4 ySx1 and H 2 (2 V2 )
6
Sx ySx4 (St1 ) . Thus we can write an explicit basis for the vector spaces involved in the
pairing (5.11)6 :
3

x0 , x0 x12 ; yx0 ,
H 1 (V2 )+ :

2
H 1 (V2 ) :
x0 x1 , x13 ; yx1 ,

H 2 2 V2 : x05 x1 , x03 x13 , x01 x15 ; yt0 x03 x1 , yt0 x0 x13 ; yt1 x04 , yt1 x02 x12 , yt1 x14 .
(5.12)
For instance, an element of H 1 (V2 )+ is given by a polynomial

a30 x03 + a12 x0 x12 + a10 yx0 ,
(5.13)
where the subscripts denote the powers of x0 and x1 respectively, or simply by a vector
(a30 , a12 , a10 ).
(5.14)
Using this description of the cohomology groups we can write down the map (5.11) explicitly.
It is given by
(a30 , a12 , a10 ) (a21 , a03 , a01 ) (b51 , b33 , b15 , b31 , b13 , b40 , b22 , b04 )
(5.15)
= (a30 a21 , a30 a03 + a12 a21 , a12 a03 , a30 a01 + a10 a21 , a10 a03 + a12 a01 , 0, 0, 0).
The image lives in the five-dimensional subspace of H 2 (2 V2 ) given by evaluating at [t0 : t1 ] =
[1 : 0].
(0,n)
Now to get the (u) pairing, we must project to H 2 (2 V ) , that is we must quotient by
T
(Im M ) . In other words, we must determine whether the above five-dimensional subspace (or
any subspace thereof) lies in (Im M T ) which would imply that (some of) the (u) pairings
vanishor not.
6 Recall that [1] the Z action is given by t t , t t , x x , x x , y y.
2
0
0 1
1 0
0 1
1
77
Recall from [1] that the map

M T : H 1 (V2 V3 ) H 2 2 V2 ,
(5.16)
given by evaluating at the invariant extension class of V , can be expressed in matrix form. More
precisely, the 18 17 matrix M has the form
A
0
0
,
(5.17)
A
B
where 0 means a column of zeroes, the top-left block has dimension 9 9 and the bottom-right
block has dimension 9 8. In order to get n pairs of Higgs (for n = 0, 1, 2), we require that
the matrix M has rank 17 n. From simple dimension counting this implies that Im M T has
codimension n as required. In fact, since we specifically want the anti-invariant part of Im M T to
have codimension n, we need a slightly stronger condition on M. We demand that the bottomright 9 8 block has rank 8 n, while the top-left 9 9 block has rank 9. It was shown in
[1] that on the loci in moduli space where the bottom-right block has rank 8 n for n = 1, 2,
requiring that the top-left block has rank 9 is an open condition, but not empty. So, on these loci,
the solution space is a dense open subset. Thus we can get almost any subspace of codimension n
(0,8)
(for n = 1, 2) in H 2 (2 V2 )
as (Im M T ) .
In particular, for a generic choice of invariant extension class satisfying the above condition
(0,8)
on the rank of M, the five-dimensional subspace of H 2 (2 V2 ) described in (5.15) (or any
T
subspace thereof) will not lie in (Im M ) . Hence, after quotienting by (Im M T ) we obtain
non-zero values, and the (u) pairings (5.9) are non-zero.
To make things more transparent, let us look first at the one Higgs case. Then, we can express
the (u) pairing (5.9) between the two 3-dimensional spaces as a 3 3 matrix. From the explicit
description of (5.15), we see that we obtain a symmetric 3 3 matrix with the bottom-right entry
being zero:

a
b
c
b
d
e

c
e .
0
(5.18)
To obtain this matrix, consider elements of the first 3-dimensional space in the pairing (5.11)
as column vectors, elements of the second 3-dimensional space as two vectors, and multiply
them to get a 3 3 matrix. Then set the bottom-right entry to zero and symmetrize the matrix.
This reformulation of (5.15) is an explicit version of the map (5.11). Here each of the entries
a, b, c, d, e is an element of H 2 (2 V2 ), which by (5.12) can be written as a linear combination
of the 8 basic monomials. By reading off the coefficients, the 3 3 matrix of polynomials (5.18)
can therefore be interpreted as a set of eight 3 3 matrices with scalar entries. The conclusion of
the previous discussion is that our original coupling (5.9) is given by a generic linear combination
of these eight matrices.
In other words, what we obtained above is that generically, the coefficients in the matrix (5.18)
are non-zero, and so the matrix has rank 3.
For the n = 2 case, we obtain two 3 3 matrices of the form (5.18) (one for each pair of
Higgs). Generically, they both have rank 3.
78
5.3. () triple pairing

The () pairing is given by

(0,3+n)

(0,n)
(51,0)
() H 1 (ad V )+
H 1 2 V
H 1 2 V C,
which can be rewritten as

(51,0)

(0,3+n)

(0,n)
H 1 2 V
H 2 2 V .
H 1 V V +
(5.19)
(5.20)
Recall that H 1 (2 V ) has a filtration F 0 F 1 {0}, with F 0 = H 1 (2 V ) , F 1 =

and F 0 /F 1 ker M T .
Recall also that H 1 (V V )+ has a filtration G0 G1 G2 {0}, with G0 = H 1 (V
V )+ , G1 = H 1 (V V3 )+ ,
H 1 (2 V2 )
G2 =
H 1 (V2 V3 )+
,
H 0 (V3 V3 )+
(5.21)
G1 /G2 H 1 (V3 V3 )+ and G0 /G1 H 1 (V2 V2 )+ .

Finally, recall that H 2 (2 V ) coker M T .
Let us first restrict H 1 (2 V ) to its F 1 = H 1 (2 V2 ) subspace. Since (Im M T )
H 2 (2 V2 ) and H 2 (2 V ) coker M T , we can replace H 2 (2 V ) in the pairing (5.20)
by H 2 (2 V2 ) , keeping in mind that we want to quotient by (Im M T ) afterwards. After rearrangement, the pairing (5.20) becomes

H 1 2 V2 H 1 2 V2 H 2 V V .
(5.22)
But 2 V2 and 2 V2 are dual line bundles, and so the pairing must be

H 1 2 V2 H 1 2 V2 H 2 Tr V V = H 2 (O) = 0,
(5.23)
which vanishes.
Hence, only the quotient space F 0 /F 1 ker M T of F 0 = H 1 (2 V ) may give non-zero
pairings. Restricting to the subspace
G2 =

H 1 (V2 V3 )+
H 1 V V +
0
H (V3 V3 )+
we get the remaining () pairings

1

(0,1+n)
(0,n)

H (V2 V3 )+ (44,0)
ker M T
coker M T
,
0
H (V3 V3 )+
(5.24)
(5.25)
which are generically non-zero.

We conclude that in the one Higgs case, only 2 of the vector bundle moduli have non-vanishing
() couplings, while in the two Higgs case only 6 of the vector bundle moduli have non-vanishing
() couplings. For n = 1 (respectively n = 2), one coupling (respectively four) involves a down
Higgs and a up Higgs, and one coupling (respectively two) involves a lepton doublet and a up
Higgs.
These pairings afford a nice geometrical description. The n = 2 locus has codimension 6 in
the invariant extension moduli space H 1 (V2 V3 )+ . The 6-dimensional normal space to it at
an n = 2 point has a pretty picture: it can be identified with the space of 2 3 matrices, or
79
linear maps f from a fixed three-dimensional space A to a fixed two-dimensional space B . The
number n of Higgs pairs that corresponds to such an f is n = 2 rank(f ). The n-dimensional
space H 1 (2 V ) of up Higgs is given by the dual of the cokernel of f , which is of course a
subspace of B, the vector space dual to B . It is two-dimensional at the origin, zero-dimensional
generically, and one-dimensional on the determinantal locus, i.e. for those maps f whose rank
is 1. The space H 1 (2 V ) of leptons and down Higgs consists everywhere of the two light
generations of leptons (the image of H 1 (2 V2 ) ) plus the (1 + n)-dimensional kernel of f .
Generically this kernel is just the heavy lepton doublet; on the locus where n = 1 (respectively
at the origin where n = 2) it has dimension 2 (respectively 3) and there is no consistent way to
say which of these is the heavy lepton and which are the Higgs.
One more piece of geometrical interpretation. As explained in [1], generically, the matrix M T
has rank 17, and we obtain a model with no massless Higgs fields. To get n Higgs pairs, the
rank of M T must decrease to 17 n. As mentioned above, this is the case in a codimension 2
region for n = 1 (and codimension 6 for n = 2) in the invariant extension moduli space. Thus,
the 42 tangent directions for n = 1 (the 38 tangent directions for n = 2)7 to the special locus
where we obtain n Higgs pairs should not develop moduli-dependent Higgs -terms, exactly as
we obtained in (5.25). However, if we move in the normal directions to the n Higgs pair locus,
the rank of M T increases, and the pairs of Higgs doublets should acquire a mass.
In the effective theory we have an analogous interpretation that precisely parallels the above
discussion. In the effective theory, the vector bundle moduli fields normal to the n Higgs pair
locus possess the tri-linear couplings to the Higgs fields given by (5.25). When these moduli
fields acquire non-zero vacuum expectation values (VEVs), we move to a point in moduli space
where the effective theory is deformed in a direction perpendicular to n Higgs pair locus. In this
case, these tri-linear couplings generate non-zero -terms for the Higgs fields, i.e., at this new
point in moduli space the Higgs field pairs become massive.
5.4. () triple pairing
The triple pairing, involving only moduli fields, is given by
()
(51,0)
(51,0)
(51,0)
H 1 (ad V )+
H 1 (ad V )+
H 1 (ad V )+
C.
(5.26)
Although we have not computed it explicitly, this pairing must be zero. Namely, to all orders
in string perturbation the self coupling of moduli fields is zero and thus these fields can acquire
non-zero VEVs, which in the effective theory parameterize deformations from a chosen locus in
moduli space. However, it is expected that non-perturbative effects introduce moduli superpotential which would fix VEVs of these fields; this topic is beyond the scope of this paper.
6. Physics implications
In this section we discuss physics implications of the tri-linear superpotential couplings. In order to interpret the triple pairings (3.7)computed in the previous sectionas Yukawa couplings
of the massless chiral superfields, one should always refer to Table 2, and associate particles to
their respective cohomology groups.
7 These numbers take into account that we quotient the invariant extension space by a one-dimensional space corresponding to rescaling the bundle.
80
Although the triple pairing calculation was performed for the loci with n = 0, 1, 2 massless
Higgs pairs, we shall primarily focus on the interpretation of the results for the locus with n = 1
massless Higgs pair.
6.1. (u) triple pairing: up-sector Yukawa couplings
ij
For the n = 1 Higgs pair locus, these couplings are the Yukawa couplings u for the up-sector
quarks (the third term in the superpotential W1 in (3.2)). The results of the previous section reveal
ij
that the 3 3 matrix u is in general of rank 3 and has the following form:

a b c
u = b d e .
(6.1)
c e 0
The coefficients in the u matrix are holomorphic functions of moduli (tangent to the n = 1
Higgs pair locus). The physical Yukawa matrix depends on the normalization of the kinetic energy terms for the quark fields, which can depend on CalabiYau and/or vector bundle moduli.
Nevertheless since the rank of (6.1) is in general 3, the physical Yukawa matrix should also have
rank 3, thus yielding non-zero masses for all the three up-sector quarks; at special points on the
n = 1 locus one expects to obtain a fully realistic up-quark sector mass hierarchy.
The superpotential Yukawa couplings for the n = 2 massless Higgs pair locus involve two
matrices of the type (6.1), giving the couplings with the two up Higgs fields. In this case as well,
all three up-sector quark masses are generically non-zero, and at special points on the n = 2 locus
one expects to obtain a fully realistic up-sector mass hierarchy.
6.2. (d) triple pairing: down-sector and R-parity violating Yukawa couplings
The tri-linear couplings of the (d)-triple pairing determine the Yukawa couplings of the downsector quarks (the second term of W1 in (3.2)) and of the charged-sector leptons (the first terms
of W1 in (3.2)), as well as the R-parity violating terms of W2 in (3.2). As it was shown in the
previous section these couplings are all zero. The expectation is that some of these couplings will
become non-zero due to quantum, worldsheet instanton effects. The calculation of such effects
is beyond the scope of this paper.
Experimental constraints essentially require the absence of R-parity violating couplings.
Therefore, phenomenological viability of the model should eventually be tested by calculations
of the couplings at the quantum level. In an optimistic scenario, R-violating couplings could
remain zero at the quantum level for a restricted subspace of the moduli space.
On the other hand, for the model to be phenomenologically viable it has to have non-zero
down-sector quark and charged-sector lepton Yukawa couplings, generated at the quantum level.
A non-zero Yukawa matrix of the down-sector quarks, along with the Yukawa matrix of the upsector quarks, also determines the CabibboKobayashiMaskawa (CKM) matrix, which specifies
the CP violating effects in the quark sector of the model.
6.3. () triple pairing: -terms and neutrino Yukawa couplings
The () pairing corresponds to the moduli-dependent Higgs -term, i.e., the first term of W
in (3.4), as well as the tri-linear couplings in the neutrino sector, i.e., the terms of W in (3.3),
where the role of the right-handed neutrinos Ri is played by vector bundle moduli i . This
81
model therefore provides an interesting mechanism for generating a -term as well as a term
responsible for giving a Dirac mass to a neutrino.
These couplings were determined in the previous section on both the loci with n = 1 and n = 2
massless Higgs pairs. Let us first focus on the results for the n = 1 locus. The moduli space,
perpendicular to the n = 1 locus, is two-dimensional. In the effective theory it is specified by two
vector bundle moduli fields, say 1 and 2 . These two moduli fields have tri-linear couplings,
say, to the up-Higgs H and the down-Higgs H , and to the up-Higgs H and one lepton doublet L,
respectively:
W 1 H H + 2 LH .
(6.2)
Of course, this is a specific choice for the assignment of H and L fields. (Since H and L are
in the same representation of the SM gauge group, their role can be interchanged, as will be
explained below.)
The first term therefore plays the role of a moduli-dependent Higgs -term. Namely, after 1
acquires a non-zero VEV, the Higgs pair becomes massive. In order to generate an acceptable
-term, the non-zero VEV has to be proportional to the electro-weak scale (O(1 TeV)), which is
much smaller than the string scale (O(1017 GeV)). Thus the VEV has to be fine-tuned, specifying
a deformation in the moduli space that is extremely close to the n = 1 locus. Thus, we have
a way to technically obtain a phenomenologically acceptable -term, however with some finetuning.
The second term plays the role of a neutrino tri-linear coupling where the role of the righthanded neutrino is played by the vector bundle modulus 2 . After electro-weak symmetry breaking, i.e., when H acquires a non-zero VEV, this term generates a Dirac mass for one neutrino
species.
Note that in principle one can choose any other linear combination of the 1 and 2 fields
to acquire a non-zero VEV. In this case the down-Higgs becomes a specific combination of the
H and L fields, and the non-zero VEV generates a -term for the Higgs pair. A combination of
1 and 2 fields, orthogonal to the one that acquires a non-zero VEV, in turn corresponds to the
right-handed neutrino field. It couples to the up-Higgs and a specific combination of the H and L
fields that are orthogonal to the down-Higgs field. This is now a tri-linear coupling which, after
electro-weak symmetry breaking, again generates a mass for one neutrino species.8
Note also that one can remain within the n = 1 locus where both 1 and 2 have zero VEVs.
In this case, 1 and 2 can be interpreted as two right-handed neutrinos, and (L, H ) as two lepton
doublets. Eq. (6.2) then generates masses for two neutrino species. In this case the model has no
parameter.
The n = 2 locus also has a very interesting structure for these tri-linear couplings. As determined in the previous section there are now six moduli fields ij (i = 1, 2, 3 j = 1, 2), transverse
to the n = 2 locus, that couple via tri-linear couplings to the two up-Higgs fields H j (j = 1, 2)
and the three fields Li = {H1 , H2 , L} (i = 1, 2, 3). Here the three Li fields can be interpreted as
two down-Higgs and one lepton doublet. The tri-linear couplings are schematically of the form:
W
2
3

ij Li H j .
(6.3)
i=1 j =1
8 Which combination of H and L fields is interpreted as a down-Higgs field and which one as a lepton field may
be further constrained at the quantum level: the requirement that R-parity violating couplings be absent may dictate a
specific combination of H and L fields to be a down-Higgs.
82
Note that these terms indeed provide the moduli dependent -terms for n = 2 Higgs doublet
pairs as well as two candidates for the right-handed neutrinos. For example, choosing specific
non-zero VEVs for 11 and 22 moduli fields generates the two -mass terms for both Higgs
pairs, (H1 , H 1 ) and (H2 , H 2 ), respectively. On the other hand, 31 and 32 play the role of
two right-handed neutrinos.9 They couple to a single lepton doublet L, so after the electroweak
symmetry breaking (and assuming that the kinetic energy terms do not have off-diagonal mixing
terms for the lepton doublet fields) there is only one massive neutrino. There is also a possibility
that with non-zero Yukawa couplings for the charged-sector leptons (obtained at the quantum
level), the model could possess a non-trivial CKM matrix in the lepton sector.10
One can also remain within the n = 2 locus where all i,j s have zero VEVs. Now the three
Li can be interpreted as lepton doublets, and the i,j s as right-handed neutrinos. Eq. (6.3) now
generates masses for all three neutrinos.
The physics implications of the Yukawa couplings, calculated in this paper at the classical
level, are encouraging. We have demonstrated that in the up-quark sector one can in principle
obtain a realistic mass hierarchy. We also demonstrated that parameter(s) for the Higgs pairs(s)
can be generated, and that at least one of the neutrinos can obtain a non-zero Dirac mass. At the
classical level, down sector quarks and charged sector leptons have zero masses. In addition, all
R-parity violating terms vanish. Thus all baryon number violating processes and lepton number
violating processes are absent. The phenomenological viability of the model should be further
tested at the quantum level where the down-sector and charged-lepton sector Yukawa couplings
are expected to be generated. It is also at the quantum level that the absence of R-violating
couplings is expected to impose strong constraints on the allowed moduli subspace of the model.
Acknowledgements
We would like to thank Volker Braun, Paul Langacker, Pran Nath and Tony Pantev for valuable discussions. The work of R.D. is supported by an NSF grant DMS 0104354 and by an
NSF Focused Research Grant DMS 0139799. The work of M.C. is supported by the DOE grant
DE-FG03-95ER40917 and by the Fay R. and Eugene L. Langberg Chair. The work of V.B. is
supported by an MSRI Postdoctoral Fellowship for the New Topological Structures in Physics
program and an NSERC Postdoctoral Fellowship. V.B. would also like to thank the Department
of Mathematics of University of Pennsylvania for hospitality while part of this work was conducted.
References
[1] V. Bouchard, R. Donagi, An SU(5) heterotic standard model, Phys. Lett. B 633 (2006) 783791, hep-th/0512149.
[2] R. Donagi, Y.-H. He, B.A. Ovrut, R. Reinbacher, The spectra of heterotic standard model vacua, JHEP 0506 (2005)
070, hep-th/0411156.
[3] G.B. Cleaver, A.E. Faraggi, D.V. Nanopoulos, A minimal superstring standard model, I: Flat directions, Int. J. Mod.
Phys. A 16 (2001) 425482, hep-ph/9904301;
W. Buchmuller, K. Hamaguchi, O. Lebedev, M. Ratz, The supersymmetric standard model from the heterotic string,
hep-ph/0511035.
9 There are also additional SM singlet fields and that couple to the Higgs fields.
12
21
10 Of course, there is a wealth of other moduli directions, transverse to the n = 2 locus, that can yield -mass parameters
for Higgs pairs, and one massive neutrino. In general, the two down-Higgs fields and the lepton doublet will correspond
to specific linear combinations of (H1 , H2 , L) fields.
83
[4] 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;
R. Blumenhagen, M. Cvetic, P. Langacker, G. Shiu, Toward realistic intersecting D-brane models, hep-th/0502005.
[5] 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.
[6] S. Katz, E. Sharpe, Notes on certain (0, 2) correlation functions, Commun. Math. Phys. 262 (2006) 611, hepth/0406226;
E. Sharpe, Notes on correlation functions in (0, 2) theories, hep-th/0502064.
[7] M. Cvetic, I. Papadimitriou, Conformal field theory couplings for intersecting D-branes on orientifolds, Phys. Rev.
D 68 (2003) 046001, hep-th/0303083;
D. Cremades, L.E. Ibez, F. Marchesano, Yukawa couplings in intersecting D-brane models, JHEP 0307 (2003)
038, hep-th/0302105;
D. Lst, P. Mayr, R. Richter, S. Stieberger, Scattering of gauge, matter, and moduli fields from intersecting branes,
Nucl. Phys. B 696 (2004) 205250, hep-th/0404134;
M. Bertolini, M. Billo, A. Lerda, J.F. Morales, R. Russo, Brane world effective actions for D-branes with fluxes,
hep-th/0512067.
[8] V. Braun, Y.-H. He, B.A. Ovrut, T. Pantev, The exact MSSM spectrum from string theory, hep-th/0512177.
[9] V. Braun, Y.-H. He, B.A. Ovrut, Stability of the minimal heterotic standard model bundle, hep-th/0602073.
[10] T.L. Gomez, S. Lukic, I. Sols, Constraining the Khler moduli in the heterotic standard model, hep-th/0512205.
[11] V. Braun, Y.-H. He, B.A. Ovrut, T. Pantev, A heterotic standard model, hep-th/0501070.
[12] P. Nath, P.F. Perez, Proton stability in grand unified theories, in strings, and in branes, hep-ph/0601023.
[13] R. Donagi, B. Ovrut, T. Pantev, D. Waldram, Spectral involutions on rational elliptic surfaces, Adv. Theor. Math.
Phys. 5 (2002) 499, math.AG/0008011.
[14] R. Donagi, B. Ovrut, T. Pantev, D. Waldram, Standard-model bundles, Adv. Theor. Math. Phys. 5 (2002) 563,
math.AG/0008010.
[15] R. Donagi, B.A. Ovrut, T. Pantev, D. Waldram, Standard-model bundles on non-simply connected CalabiYau
threefolds, JHEP 0108 (2001) 053, hep-th/0008008.
[16] J. Giedt, G.L. Kane, P. Langacker, B.D. Nelson, Massive neutrinos and (heterotic) string theory, Phys. Rev. D 71
(2005) 115013, hep-th/0502032.
[17] M.B. Green, J.H. Schwarz, E. Witten, Superstring Theory, vol. 2: Loop Amplitudes, Anomalies and Phenomenology, Cambridge Monographs on Mathematical Physics, vol. 5, Cambridge Univ. Press, Cambridge, 1987.
Revisiting + at low energies

J. Gasser a, , M.A. Ivanov b , M.E. Sainio c,d
a Institute for Theoretical Physics, University of Bern, Sidlerstrasse 5, CH-3012 Bern, Switzerland
b Laboratory of Theoretical Physics, Joint Institute for Nuclear Research, 141980 Dubna, Moscow region, Russia
c Helsinki Institute of Physics, P.O. Box 64, 00014 University of Helsinki, Finland
d Department of Physical Sciences, University of Helsinki, Finland
Received 28 February 2006; accepted 24 March 2006

Abstract
We complete the recalculation of the available two-loop expressions for the reaction in the
framework of chiral perturbation theory. Here, we present the results for charged pions. The cross section
and the values of the dipole polarizabilities agree very well with the earlier calculation, provided the same
set of low-energy constants (LECs) is used. With updated values for the LECs at order p4 , we find for the
dipole polarizabilities (1 1 ) = (5.7 1.0) 104 fm3 , which is in conflict with the experimental
result recently reported by the A2 Collaboration at MAMI.
PACS: 11.30.Rd; 12.38.Aw; 12.39.Fe; 13.60.Fz
Keywords: Chiral perturbation theory; Two-loop diagrams; Pion polarizabilities; Compton scattering
1. Introduction
We evaluate the amplitude for + in the framework of chiral perturbation theory
(ChPT) [13] at two-loop order, and compare the result with the only previous calculation performed at this accuracy [4]. We employ the calculational techniques outlined in our previous
work [5], which allow us to provide a rather compact and easy to use integral representation for
the full amplitude. We find that the cross section for the reaction + and the dipole
polarizabilities agree very well with the results reported in Ref. [4], provided that the same set
E-mail address: gasser@itp.unibe.ch (J. Gasser).

doi:10.1016/j.nuclphysb.2006.03.022
J. Gasser et al. / Nuclear Physics B 745 (2006) 84108
85
of LECs is used. With updated LECs at order p 4 [6,7], we find for the dipole polarizabilities the
value
(1 1 ) = (5.7 1.0) 104 fm3 .
(1.1)
The A2 Collaboration at MAMI [8] has recently reported the experimental result
(1 1 ) = (11.6 1.5stat 3.0syst 0.5mod ) 104 fm3 .
(1.2)
The index mod denotes the uncertainty generated by the theoretical models used to analyze the
data. The ChPT calculation is in conflict with this prediction, see also [9] for a recent discussion.
There are good reasons to believe that the chiral prediction Eq. (1.1) is rather stable against
contributions from still higher orders in the chiral expansion. The main point is that the Bornterm subtracted amplitudein terms of which the polarizabilities are defined is regular at the
Compton threshold. Therefore, the influence of chiral logarithms is suppressed as compared to
the amplitude at, say, the threshold for production, where the amplitude generates a branch
point. Further, we have analyzed the potential influence of resonance exchange and could not
identify any significant contribution from nearby resonances to the combination (1 1 ) .
We conclude that the low-energy constants at order p 6 are expected to play a negligible effect
here. In view of these observations, the discrepancy between the chiral prediction and the recent
MAMI data cannot be explained.
The article is organized as follows. In Section 2 we spell out the kinematics of the process
+ . To make the article selfcontained, we summarize in Section 3 the necessary ingredients of the effective Lagrangian framework. Here, we also fix the LECs to be used in numerical
calculations. In Section 4, we display the Feynman diagrams and discuss shortly their evaluation.
Section 5 contains a concise representation of the two Lorentz-invariant amplitudes that describe
the scattering matrix element. Section 6 contains explicit expressions for the dipole and quadrupole polarizabilities valid at next-to-next-to-leading order in the chiral expansion, together with
a detailed numerical analysis and a comparison with the recent MAMI data [8], and with an
evaluation from data on + [10]. The summary and an outlook are given in Section 7.
A detailed comparison with the earlier calculation [4] is provided in Appendix A, and several
technical aspects of the calculation are relegated to additional Appendices BD.
2. Kinematics
The amplitude describing the process + may be extracted from the matrix element

+
+
(p1 ) (p2 ) out (q1 ) (q2 ) in = i(2)4 4 (Pf Pi )T ,
(2.1)
where
T

= e2 1 2 W
,

+
= i dx ei(q1 x+q2 y) + (p1 ) (p2 ) outTj (x)j (y)|0.
W
(2.2)
Here j denotes the electromagnetic current and = e2 /4 1/137 is the electromagnetic

coupling. It is convenient to change the pion coordinates according to ( , 0 ) ( 1 , 2 , 3 )
and instead of + production, we consider in the following the process 1 1 , with

W

= W
1 1
.
= V ,
(2.3)
86
where the relative minus sign stems from the CondonShortly phase convention. (We use the
same sign convention as Ref. [4].) The decomposition of the correlator V into Lorentz-invariant
amplitudes reads
V = A(s, t, u)T1 + B(s, t, u)T2 + C(s, t, u)T3 + D(s, t, u)T4 ,
1
T1 = sg q1 q2 ,
2
T2 = 2s 2 g 2(q1 q2 ),
T3 = q1 q2 ,
T4 = s(q1 q2 ) (q1 q1 + q2 q2 ),
= (p1 p2 ) ,
(2.4)
where
s = (q1 + q2 )2 ,
t = (p1 q1 )2 ,
u = (p2 q1 )2 ,
=t u
(2.5)
are the standard Mandelstam variables. The tensor V satisfies the Ward identities
q1 V = q2 V = 0.
(2.6)
The amplitudes A and B are analytic functions of the variables s, t and u, symmetric under
crossing (t, u) (u, t). The amplitudes C and D do not contribute to the process considered
here, because i qi = 0.
It is useful to introduce in addition the helicity amplitudes

H ++ = A + 2 4M2 s B,
8(M4 tu)
(2.7)
H + =
B.
s
The helicity components H ++ and H + correspond to photon helicity differences = 0, 2, respectively. With our normalization of states p1 |p2 = 2(2)3 p10 (3) (p1 p2 ), the differential
cross section for unpolarized photons in the centre-of-mass system is
+
d
2s
(2.8)
=
(s)H (s, t),
H (s, t) = |H ++ |2 + |H + |2 ,
d
32

with (s) = 1 4M2 /s. The relation between the helicity amplitudes M+ in Ref. [10] and
the amplitudes used here is
M++ (s, t) = 2 H ++ (s, t),
M+ (s, t) = 16B(s, t).
(2.9)
q1 ||p1 | cos , where is the scatIn the centre-of-mass system, q1 + q2 = 0, one has q1 p1 = |
tering angle. Then the Mandelstam variables are given by

s
s = 4|
q |2 ,
(2.10)
t = M2 1 (s) cos .
2
For comparison with experimental data, it is convenient to present also the total cross section for
the case having | cos | less than some fixed value Z,

2
s; | cos | < Z =
8
t+
dt H (s, t)
t
with t = M2 (s/2)(1 (s)Z).
(2.11)
87
3. The effective Lagrangian and its low-energy constants

The effective Lagrangian consists of a string of terms. Here, we consider QCD with two
flavours, in the isospin symmetry limit mu = md = m.
At next-to-next-to-leading order (NNLO),
one has [2]
Leff = L2 + L4 + L6 .
(3.1)
The subscripts refer to the chiral order. The expression for L2 is

L2 =

F2
D U D U + M 2 U + U ,
4
e
(3.2)
diag(1, 1),
2
where e is the electric charge, and A denotes the electromagnetic field. The quantity F denotes
the pion decay constant in the chiral limit, and M 2 is the leading term in the quark mass expansion
of the pion (mass)2 , M2 = M 2 (1 + O(m)).
Further, the brackets denote a trace in flavour
space. In Eq. (3.2), we have retained only the terms relevant for the present application, i.e., we
have dropped additional external fields. We choose the unitary 2 2 matrix U in the form
+

2
0
2
2
U = + i/F,
+ 2 = 122 ,
(3.3)
=
.
2 0
F
D U = U i(QU U Q)A ,
Q=
The Lagrangian at NLO has the structure [2]

L4 =
li Ki +
i=1
3

i=1
hi K i =
2
l1
D U D U + ,
4
(3.4)
where li , hi denote low-energy couplings, not fixed by chiral symmetry. At NNLO, one has
[1113]
L6 =
57

ci Pi .
(3.5)
i=1
For the explicit expressions of the polynomials Ki , K i and Pi , we refer the reader to
Refs. [2,1113]. The vertices relevant for + involve l1 , . . . , l6 from L4 and several ci s from L6 , see below.
The couplings li and ci absorb the divergences at order p 4 and p 6 , respectively,

li = (c)d4 lir (, d) + i ,
(1)

(c)2(d4) r
(2)
(L)
ci (, d) i 2 i + i (, d) ,
F2

1
1
,
ln c = ln 4 +
(1) + 1 .
=
2
2
16 (d 4)
ci =
(3.6)
The physical couplings are lir (, 4) and cir (, 4), denoted by lir , cir in the following. The coef(1,2,L)
are tabulated in [12]. In order to compare the present
ficients i are given in [2], and i
calculation with the result of [4], we shall use the scale independent quantities li introduced
88
in [2],
lir =
i
(li + l),
32 2
(3.7)
where the chiral logarithm is l = ln(M2 /2 ). We shall use [6]

l1 = 0.4 0.6,
l2 = 4.3 0.1,
l3 = 2.9 2.4,
l4 = 4.4 0.2,
(3.8)
and
.
l = l6 l5 = 3.0 0.3
(3.9)
obtained from radiative pion decay to two loop accuracy [7,14].

The constants cir occur in the combinations

r
r
r
r
r
r
r
a1r = 4096 4 6c6r + c29
,
c30
3c34
+ c35
+ 2c46
4c47
+ c50
r

r
4
r
r
r
r
r
r
r
a2 = 256 8c29 8c30 + c31 + c32 2c33 + 4c44 + 8c50 4c51 ,
r

r
r
r
br = 128 4 c31
.
+ c32
2c33
4c44
Their values have been estimated by resonance exchange, e.g., in Ref. [4]. We have repeated that
analysis. Taking into account , a1 and b1 exchange which contribute with a definite sign, we
obtain
r r r
a1 , a2 , b = (3.2, 0.7, 0.4) (present work).
(3.10)
Unless stated otherwise, we will use these estimates at the scale = M . Contributions from
scalar and tensor exchange are of a similar order of magnitude (see Table 2 in Ref. [15]). On the
other hand, the sign of these contribution is not fixed. In Ref. [16], a large NC framework and
the ENJL model were used to pin down these constants, with the result
r r r
a1 , a2 , b = (8.7, 5.9, 0.38)
(3.11)
(Ref. [16]).
Only br agrees in the two approaches. We have checked that scalar and tensor exchange, taken
r that are not in disagreement with Eq. (3.11)as has
with the proper sign, generate values for a1,2
been foreseen in the comments made in Ref. [16] concerning these two approaches. It would be
very useful to recalculate these couplings, by minimizing the amount of information used which
goes beyond what is known from QCD, e.g., along the lines outlined in [17]. Finally, we note
r has been determined from a chiral sum rule.
that in Ref. [18], c34
We now shortly discuss the uncertainties that we shall attach in the following to these couplings. In the case of br , we shall use
br = 0.4 0.4.
(3.12)
and determined
As far as the polarizabilities are concerned, (1 1 ) is independent of
precisely by the chiral expansion to two loops, once a1r is fixed. We will then simply display this
quantity as a function of a1r the result turns out to be rather independent of its exact value, see
Section 6.2 for a detailed discussion, and the uncertainty to be attached to it does not, therefore,
matter here. On the other hand, those polarizabilities which depend on a2r cannot be determined
precisely from a calculation to two loops for reasons explained in Section 6.2we do not, therefore, worry here about the precise value and uncertainty for a2r .
To complete this discussion, we note that we shall use F = 92.4 MeV [19] (see [20] for a
recent update of this value), and M = 139.57 MeV in numerical calculations.
a2r
89
Fig. 1. A set of two-loop diagrams generated by L2 and one-loop diagrams generated by L4 .
Fig. 2. Construction scheme for the diagrams in Fig. 1.
4. Evaluation of the diagrams

The lowest-order contributions to the scattering amplitude are described by tree- and one-loop
diagrams. These contributions were calculated in [21]. The two-loop diagrams are displayed in
the Figs. 1, 3 and 4. The two-loop diagrams in Fig. 1 may be generated according to the scheme
indicated in Fig. 2, where the filled in blob denotes the d-dimensional elastic -scattering
amplitude at one-loop accuracy, with two pions off-shell.
90
Fig. 3. A class of one-particle reducible diagrams. The filled in circles summarize self-energy and vertex corrections.
Fig. 4. Acnode and butterfly diagrams.
Fig. 5. The remaining diagrams at order p6 : one-loop graphs generated by L4 , and counterterm contributions from L6 .
The diagrams shown in Fig. 3 may be reduced to tree-diagrams by using Ward identities [22].
They sum up to the expression

1
2Z g (2p1 q1 ) (2p2 q2 )
(4.1)
Z
R(t)
+
crossed
,
M2 t
where Z is the pion renormalization constant. The function R(t) starts at order 1/F4 and can
be obtained from the full pion propagator [22].
Two further diagrams are displayed in Fig. 4. The first onecalled acnode in the
literaturemay again be evaluated by use of a dispersion relation, see [5]. The second one
is trivial to evaluate, because it is a product of one-loop diagrams. The remaining diagrams at
order p 6 are shown in Fig. 5.
The evaluation of the diagrams was done in the manner described in [5,23] and invoking
FORM [24]. In particular, we have verified that the counterterms from the Lagrangian L6 [12]
91
remove all ultraviolet divergences, which is a very non-trivial check on our calculation. Furthermore, we have checked that the (ultraviolet finite) amplitude so obtained is scale-independent.
5. The two-loop amplitudes
We give the expressions for the amplitudes A and B using the same notation as in [4]. This
results in

1
2
l
1
+ 2
+ 2 G
(s)
+
+ UA + PA + O p 4 .
A= 2
(5.1)
2
M t M u F
48
The unitary part UA contains s, t and u-channel cuts, and PA is a linear polynomial in s. Explicitly we find
UA =

1 4
l
s J(s)
G(s) 2M 4M2 s + 3s 2 J(s) + C(s, li ) +
4
sF
48 2 F4

2

=
=
l1 43
s 4M2 H (s) + 4 s G(s)

+ 2M2 G (s) 3 J (s) d00
+
288 2 sF4

2

=
=
l2 56
s 4M2 H (s) + 4 s G(s)

+ 2M2 G (s) 3 J (s) d00
+
2
4
96 sF
+ A (s, t, u),
with
C(s, li ) =
(5.2)

1
4 2
1
l1
16s 56M2 s + 64M4
2
3
48 3

2
5
8s 24M2 s + 32M4
+ l2
6

4
2
2
4
12M l3 + 12M s l4 12M s + 12M ,

1
(5.3)
3 cos2 1 .
2
(s) in Eq. (5.1) stands for G(s)
The loop functions J, etc., are displayed in Appendix B, and G

2
evaluated with the physical mass. The term proportional to d00 in UA contributes to D-waves
only. For A see below. The polynomial part is
2
d00
=

1
a1 M2 + a2 s ,
(16 2 F2 )2

1
337
5
r
2
4l + l 10l1 + 18l2 12l +

l1 5l2 + 12l4 l + 4 ,
a1 = a1 +
9
6
3

1 2
1
3
127
5
21
5
r
l + l l1 + l2 + 3l +
l 1 l 2 + 3l +
.
a2 = a 2
9
2
2
24
12
4
2
PA =
The result for B reads

1
1
1
+
B=
+ UB + PB + O p 2 ,
2
2
2s M t M u
(5.4)
(5.5)
92
Fig. 6. The + cross section (s; | cos | Z = 0.6) as a function of

from [25].
s. The experimental data are taken
with the unitary part

1
4
5
1
UB =
l1
+ l2
H (s) + B (s, t, u).
3
6
192 2 sF4 3
(5.6)
For the polynomial part we find

PB =
,
(16 2 F2 )2

1 2
1
3
53
1
7
1
l + l l1 + l2
l1 l2 +
.
b = br
18
2
2
24
12
4
2
(5.7)
The integrals A,B (s, t, u) contain contributions from the two-loop box, vertex and acnode
graphs and also from the reducible diagrams. The explicit expressions for these quantities are
given in Appendices C and D. 1
As an application of the above, we plot the total cross section in Fig. 6, using the LECs from
Eqs. (3.8)(3.10). The data are taken from [25]. It is seen that the two-loop corrections are tiny
in this kinematical region.
In order not to interrupt the argument, a detailed comparison of our result with the previous
calculation performed by Burgi [4] is relegated to Appendix A.
6. Pion polarizabilities: dipole and quadrupole
The dipole and quadrupole polarizabilities are defined [26,27] through the expansion of the
helicity amplitudes at fixed t = M2 ,

s
(6.1)
H+ s, t = M2 = (1 1 ) + + (2 2 ) + + O s 2 ,
M
12
1 The corresponding F ORTRAN codes are available upon request from the authors.
93
where H+ denote the helicity amplitudes H + with Born term subtracted. Because we have
at our disposal the helicity amplitudes at two-loop order, we can work out the polarizabilities to
the same accuracy. It turns out that all relevant integrals can be performed in closed form. We
discuss the results in the remaining part of this section.
6.1. Chiral expansion
Using the same notation as in [4], we find for the dipole polarizabilities
(1 1 )

4
M2 d1
=
+ O M ,
c1 +
16 2 F2 M
16 2 F2
(6.2)
where
2
c1 = l ,
3

4
3
3
91
1
1
53
r
d1+ = 8b
l l + l1 + l2 l + l1 + l2 +
+ + ,
9
2
2
24
2
2
72

4
65
1
1
187
1
l l1 l2 + l
l1 l2 + l3 l l4 +
+ ,
d1 = a1r + 8br
3
12
3
3
4
108
(6.3)
c1+ = 0,
with
+ =
8105 135 2
,
576
64
41
53
2.
432 64
(6.4)
For the quadrupole polarizabilities, we obtain

(2 2 ) +

4
M2 d2
=
+ O M ,
c2 +
16 2 F2 M3
16 2 F2
(6.5)
with
c2+ = 0,
c2 = 2,
2062 10817 2
8
8
d2+ =
+
+ l1 + l2 ,
27
1440
45
15
8
218
238
l1
l2
d2 = 12a2r 24br l(10 + 4l ) l3 + 4l4 4l
15
45
45
56 1199 2
.
45 1920
(6.6)
We end this subsection by evaluating the polarizabilities, using the above expressions and the
central values for the LECs in Eqs. (3.8)(3.10). The results are shown in Table 1.2 The numbers
in brackets correspond to the order p 6 LECs in Eq. (3.11). The uncertainties are discussed in the
following subsection.
2 Dipole (quadrupole) polarizabilities are given in units of 104 fm3 (104 fm5 ).
94
Table 1
The dipole and quadrupole polarizabilities. The numbers in brackets correspond to the order p6 LECs in Eq. (3.11)
To one loop
(1 1 ) +
(1 + 1 ) +
6.0
0
(2 2 ) +
(2 + 2 ) +
11.9
0
To two-loops
5.7 [5.5]
0.16 [0.16]
16.2 [21.6]
0.001 [0.001]
Table 2
The contribution of the LECs at order p6 to the polarizabilities, according to Eq. (3.10). The numbers in brackets correspond to Eq. (3.11)
(1 1 )
(1 + 1 )
(2 2 )
(2 + 2 )
a1r
a2r
br
Total
0.14 [0.37]
0
0
0
0
0
0.72 [6.09]
0
0.14 [0.13]
0.14 [0.13]
0.83 [0.78]
0
0 [0.24]
0.14 [0.13]
0.10 [5.31]
0
Table 3
Resonance saturation induces a scale dependence in the amplitudes. Displayed are the values of the polarizabilities in case that saturation is assumed
at = 500 MeV or at = 1 GeV
(1 1 )
(1 + 1 )
(2 2 )
(2 + 2 )
= 500 MeV
= 1 GeV
6.1
0.20
14.6
0.001
5.5
0.13
17.2
0.001
6.2. Estimating the uncertainties

The uncertainty in the prediction for the polarizability has two sources. First, the low-energy
constants are not known precisely. Second, we are dealing here with an expansion in powers of
the momenta and of the quark masses. We carried out this expansion up to and including terms
of order p 6 . Higher order terms will influence the resultby which amount?
We start the discussion by considering the uncertainties in the LECs. We shall use the order p 4
LECs displayed in Eqs. (3.8) and (3.9). The LECs at order p 6 are not well known, see the
discussion in Section 3. In Table 2, we display the contributions from the LECs at order p 6 to
the four polarizabilities, using the values from Eq. (3.10). The ones corresponding to Ref. [16]
displayed in Eq. (3.11)are given in square brackets. The only significant difference occurs in
the value of the difference of the quadrupole polarizabilities (2 2 ) .
Resonance saturation generates a second inherent uncertaintyshould one saturate at =
500 MeV or at = 1 GeV? In Table 3 we display the polarizabilities for saturation at these two
scales, using Eq. (3.10). It is seen that the induced change (which is independent of the values
of the LECs at order p6 ) is quite substantial for (2 2 ) . This is related to the fact that this
quantity contains a rather substantial chiral logarithm, see below.
95
Fig. 7. The helicity non-flip amplitude H++ in units of M2 as a function of s, at t = u, with Born term subtracted. For
s 4M2 the quantity shown is 102 M2 H++ , and for s 4M2 we display 102 M2 |H++ |. The solid (dashed) line is the
expression to two loops (to one loop). The Compton threshold in and the threshold in + are
denoted by the encircled letters C and A, respectively. It is clearly seen that two-loop corrections are suppressed at the
Compton threshold.
We now discuss the second source of uncertainties, the truncation of the chiral expansion
itself. It is clear that, to have an idea of higher order terms, one needs at least the first two terms
in the expansion. This makes it already clear that it is difficult to make reliable predictions for the
polarizabilities connected with the helicity flip amplitude, from which we have determined here
the leading order contribution only. So, let us concentrate first on the helicity non-flip case H++ .
The Born-term subtracted helicity amplitude H++ does not have branch points at the Compton threshold. This is why it can be expanded there into an ordinary Taylor series, e.g., in the
variable s and = (t u). One then expects that the amplitude at the Compton threshold is less
affected by chiral logarithms than its counterparts at the threshold for , where unitarity
cusps do occur. This is illustrated in Fig. 7, where we display the quantity 102 M2 H++ (s, t = u)
as a function of s at t = u. Above the threshold s = 4M2 , the modulus is shown. The solid
(dashed) line is the expression to two loops (to one loop). It is clearly seen that the corrections
at the Compton threshold are much smaller than the ones at the threshold for . To
identify the chiral logarithms, we note that according to Eq. (3.7), the quantities li diverge in
the chiral limit, li = l + ri , where ri is quark mass-independent. We decompose the li in
the amplitude H++ in this manner and display the contributions from the chiral logarithms l in
Table 4. The numbers illustrate that, indeed, chiral logarithms generate a smaller contribution at
the Compton threshold than at threshold for pion pair production.
As for the quadrupole polarizabilities (2 2 ) , also connected with the non-flip amplitude, it is seen from Table 1 that there is a substantial two-loop correction to the one-loop result.
This can be again understood from the behavior of H++ . As its value is considerably increased
at the threshold , its slope at the Compton threshold receives a substantial correction
as well, in order to make up that change. Indeed, the chiral expansion generates chiral logarithms that are responsible for the major part of the increase. If we decompose (2 2 ) in
a manner analogous the amplitude H++ above, we find that chiral logarithms contribute with
4.5 104 fm5 at two-loop order. These logarithms are, of course, independent of the LECs
at order p 6 .
96
Table 4
The amplitude 102 M2 H++ (s, t = u) at the Compton threshold s = 0 and at
s = 4M2 . The contribution from chiral logarithms, listed in the fourth column,
is included in the two-loop result quoted in column three. The normalization is
NH = 102 M2 . The LECs at order p6 are the ones from Eq. (3.10)
NH H++ (s, t = u)
To 1 loop
To 2 loops
Chiral logarithms
s=0
s = 4M2
2.89
7.13
2.77
8.80
0.35
1.28
Finally, we have checked whether there are potentially large contribution to H++ at order p 8 .
Using the same procedure as in [5], we found that all contributions from resonance exchange
with masses below 1 GeV have a negligible effectwe do not quote the corresponding numbers
here.
We now turn to the helicity flip amplitude H+ , which starts out at order p 6 : we have determined here only its leading order term in the chiral expansion. We checked whether there are
potentially large contribution to H+ at order p 8 , as is the case in 0 0 [5]. Whereas,
there is substantial contribution from -exchange in the neutral case, this resonance does not
contribute here, and the remaining contributions from -exchange are very small, except for the
contribution to the slope, parametrized by (2 + 2 ) , which is affected by 0.04 104 fm5 .
On the other hand, there are also contributions from one-loop graphs at order p 8 , where each
vertex is generated by an anomalous contribution from the WessZuminoWitten Lagrangian at order p 4 . We see no reason why these should be small compared to the leading term
at order p 6 . We, therefore, do not consider the chiral prediction for these quantities particularly
reliable.
6.3. Values of the polarizabilities
We have now all ingredients to provide a value for the polarizabilities in the chiral expansion, and start the discussion with (1 1 ) . We add the uncertainties from the couplings at
order p 4 , from br and from the scale dependence introduced by the resonance scheme in quadrature and obtain
= 0.80 104 fm3
(6.7)
for the so generated uncertainty. We display in Fig. 8 the result as a function of the rather inaccurately known coupling a1r . We indicate with a filled square (filled circle) the two-loop result,
evaluated with the LECs given in Eq. (3.10) (Eq. (3.11)). The width of the band is twice the uncertainty displayed in Eq. (6.7). Let us note that the two-loop prediction differs only slightly from
the one-loop calculation, see Table 1. This again shows that the value for the dipole polarizability
is rather reliablethere is no sign of any large, uncontrolled correction to the two-loop result.
We use the maximum deviation 1.0 from the central value 5.7 as the final theoretical uncertainty
for the dipole polarizability, and obtain
(1 1 ) = (5.7 1.0) 104 fm3 .
(6.8)
In the same figure, we also show the result of the recent measurement at MAMI [8] of this
quantity. It is seen that the chiral expansion is in conflict with that measurement, independent
of any reasonable value chosen for a1r . The discrepancy with the recent dispersive analysis by
Filkov and Kashevarov [10], displayed in Table 5, is even more pronounced.
97
Fig. 8. The polarizabilities (1 1 ) at two-loop order, as a function of a1r . The filled in square (filled in circle) denotes
the value obtained by using for a1r the value from Eq. (3.10) (Eq. (3.11)). The width of the ChPT band is calculated in the
manner described in the text. MAMI data refers to the recent MAMI experiment [8]. We have added the errors quoted
there in quadrature.
Next, we discuss the quadrupole case (2 2 ) . Here, the situation differs. First, as we
discussed above, chiral logarithms at order p 6 generate now a rather large contribution. As a
result of this, the scale dependence of the saturation scheme is pronounced as well, see Table 3.
Second, the LEC a2r enters with weight 12. As a result of this, it now matters which value of a2r is
used, even though its contribution is suppressed by the factor M2 (16 2 F2 )1 0.014. Although
the effect of using the [16] value for a2r goes in the right direction and brings the result in closer
agreement with the analysis of Filkov and Kashevarov [10],

4
fm5 ,
(2 2 ) = 25.0+0.8
0.3 10
(6.9)
we cannot take the outcome as a reliable two-loop prediction of ChPT: compared to the one-loop
result, the two-loop contribution generates nearly a 100 percent contribution in this case, see
Table 1.
As we discussed in the previous subsection, the situation is no better for the polarizabilities
(1,2 + 1,2 ) associated with the helicity flip amplitude: considerably more work is required
to put the chiral prediction on a firm basis. Orders of magnitudes can be read off from Table 1.
In Table 5, we collect experimental information on the dipole polarizabilities of the charged
pions. In view of the large scatter of the results, which are partly inconsistent among each other,
it is interesting to note that the COMPASS Collaboration at CERN is presently remeasuring
pion and kaon polarizabilities with high statistics, using the Primakoff effect [42]. It is also
worth mentioning that a reanalysis of the MAMI data [8], using a chiral-invariant framework, is
underway [43].
98
Table 5
Experimental information on (1 1 ) , in units of 104 fm3 . We indicate the reaction and the data used. In [28]
and [29], 1 was determined, using as a constraint 1 = 1 . To obtain (1 1 ) , we multiplied the results by a
factor of 2
(1 1 )
Experiments
p + n
11.61.5stat 3.0syst 0.5mod
Mainz (2005) [8]
13.0+2.6
1.9
L. Filkov, V. Kashevarov (2005) [10]

+
MARK II [25], TPC/2 [30],
CELLO [31], VENUS [32],
ALEPH [33], BELLE [34]
A. Kaloshin, V. Serebryakov (1994) [35]
MARK II [25]
+
Crystal Ball [36]
5.25 0.95
J.F. Donoghue, B. Holstein (1993) [28]

+
MARK II [25]
5.4
D. Babusci et al. (1992) [29]

+
PLUTO [37]
DM 1 [38]
DM 2 [39]
MARK II [25]
38.2 9.6 11.4

34.4 9.2
52.6 14.8
4.4 3.2
p + n
Lebedev Inst. (1986) [40]
40 24
Z Z
Serpukhov (1983) [41]
15.6 6.4stat 4.4syst
Table 6
Comparison of the polarizabilities with the previous calculation by Burgi [4]. The second column displays Burgis result,
the third one our evaluation, using Eq. (A.1). The fourth column displays the polarizabilities evaluated with the LECs
used in the present work. See text in Appendix A
(1 1 ) +
(1 + 1 ) +
Burgi [4]
Present work LECs from Eq. (A.1)
Present work LECs from Eqs. (3.8)(3.10)
4.42
0.31
4.39
0.28
5.72
0.16
7. Summary and outlook

1. We have recalculated the two-loop expression for the amplitude + in the framework of chiral perturbation theory. We have made use of the techniques developed in
Ref. [5,23], and of the effective Lagrangian L6 constructed in [11,12].
2. The two Lorentz-invariant amplitudes A and B are presented as a sum over multiple Feynman parameter integrals, whose numerical evaluation poses no difficulty. This is in contrast
to Ref. [4], where part of the amplitudes, denoted by A,B , could be published in numerical
form only. Further, the method has allowed us to evaluate the dipole and quadrupole polarizabilities in closed form. As far as we are aware, the quadrupole polarizabilities for charged
pions have never been calculated in ChPT before.
3. Our result agrees with the earlier calculation [4] up to the remainder A,B . The induced
changes in the cross section and in the dipole polarizabilities are far below the uncertainties
generated by the (not precisely known) values of the low-energy constants. For the cross
99
section below 500 MeV, the change is less than 1 percent. For the polarizabilities, the change
is given in columns two and three of Table 6.
4. We have investigated the uncertainties in the polarizabilities due to higher order corrections,
and due to the uncertainties in the LECs at order p 6 . According to our analysis, the twoloop result Eq. (6.8) for the dipole polarizability (1 1 ) is particularly reliable. It is
in conflict with the recent experimental result obtained at MAMI [8], or with the dispersive
analysis performed in [10].
Acknowledgements
This work was completed while M.A.I. visited the University of Bern. It is a pleasure to
thank Lev Filkov for useful correspondence. This work was supported by the Swiss National
Science Foundation, by RTN, BBW-Contract No. 01.0357, and EC-Contract HPRN-CT200200311 (EURIDICE). M.A.I. also appreciates the partial support by the Russian Fund of Basic
Research under Grant No. 04-02-17370. Also, partial support by the Academy of Finland, grant
54038, is acknowledged.
Appendix A. Comparison with the previous calculation
In this appendix, we compare the amplitudes A, B and the dipole polarizabilities with the
previous calculation presented in Ref. [4]. In that reference, the amplitudes were evaluated with
a different technique. Furthermore, the Lagrangian L6 was not available in those days, and an
important ingredient to check the final result was, therefore, missing. We can make the following
observations.
(1) The amplitudes A and B consist of a part with explicit analytic expressions, and additional
terms A,B , that are displayed in Appendices C and D of the present work in the form of
integrals over Feynman parameters. A,B were given only in numerical form in [4].
(2) We find that our explicit analytic expressions agree with the ones in [4]. To compare A,B ,
we made two checks. First, we evaluated the cross section for the reaction +
below a centre-of-mass energy of 500 MeV, using the same values for the LECs as in [4],
(l1 , l2 , l3 , l4 , l ) = (1.7, 6.1, 2.9, 4.3, 2.7),
r r r
a1 , a2 , b = (3.3, 0.75, 0.45).
(A.1)
Our result agrees within a fraction of a percent with the two-loop result displayed in Fig. 9
of Ref. [4].
(3) Second, we evaluated the dipole polarizabilities by using the values of the LECs in Eq. (A.1).
The result is displayed in the third column of Table 6. In the second column, we display the
values obtained in [4]. It is seen that the two results agree very well. The small difference is
entirely due to the slightly different values of the quantities A,B .
The following comments concerning the central values of the polarizabilities are in order. We
display in the last column in Table 6 the values of polarizabilities obtained by using the LECs
from Eqs. (3.8)(3.10). It is seen that the combination (1 1 ) differs considerably from
the one obtained with Eq. (A.1). To identify the source of this difference, we first note that the
polarizabilities depend linearly on the LECs, except the term l l4 in d1 , as can be seen from
100
the Eqs. (6.2)(6.4). We then expand (1 1 ) in the LECs around their central values, drop
quadratic terms and obtain at = M the decomposition
(1 1 ) = 5.719 + 0.215(l1 + 0.4) 0.177(l2 4.3) 0.014(l3 2.9)

+ 0.172(l4 4.4) + 2.44(l 3) + 0.04 a1r + 8br .
(A.2)
Inserting in this expansion, e.g., the values in Eq. (A.1) generates the result in column 3 of
Table 6. It turns out that all changes in the order p4 LECs have conspired to modify Burgis
result for (1 1 ) towards positive values. The contribution from the LECs at order p 6 is
negligible.
In summary, we conclude that our calculation nicely confirms Burgis result [4], provided that
the same set of LECs is used.
Appendix B. One-loop functions
In order to simplify the expressions, we set the pion mass equal to one in the following appendices,
M = 1.
(B.1)
1. The loop-integral G(s)

is

1

1
dx
2
G(s)
=
ln 1 sx(1 x) .
1+
s
x
16 2
(B.2)
is analytic in the complex s-plane, cut along the positive real axis for Re s 4. At small s,
G
G(s)
=
(n!)2
1 n
s
.
2
(n + 1)(2n + 1)!
16
(B.3)
n=1
The absorptive part is
Im G(s)
=
and
1+
1
ln
,
8s 1
s > 4,
1
2
1
,
1 + s ln 1+ + i

16 2 G(s)
= 1 4 arctg2 s 1/2 ,
s
4s
1 + 1s ln2 1
+1 ,
1 4/s,
(B.4)
4 s,
0 s 4,
s 0.
(B.5)
In the text we also need

=

(0).
sG
G (s) = G(s)
(B.6)
2. The loop-integral J(s) is

1
J(s) =
16 2
1
0

dx ln 1 sx(1 x) .
(B.7)
101
J is analytic in the complex s-plane, cut along the positive real axis for Re s 4. At small s,
J(s) =
1 n (n!)2
.
s
n(2n + 1)!
16 2
(B.8)
n=1
The absorptive part is
Im J(s) =
,
16
(B.9)
s > 4.
Explicitly,
1

+ i + 2,
ln 1+

1/2
s 1/2
16 2 J(s) = 2 2 4s
arctg 4s
,
s
+
2,
ln 1
+1
4 s,
0 s 4,
s 0.
(B.10)
In the text we also need

=
J (s) = J(s) s J
(0).
(B.11)
and J,
3. The loop-function H is defined in terms of G
H (s) = (s 10)J(s) + 6G(s).
(B.12)
Appendix C. The quantities A and B

Here we display the expressions for the quantities A(B) .

35947 2
1130291
1
A (s, t, u) =
+
s
S(t)
S(u)
8100
3
6350400
(4F )4

1
1
FAacn (s, t, u) + FAver (s) + FAbox (s, t, u) ,
+
4
(4F ) 288

492197
81101 2 1
1
1
1
+ S(t) + S(u)
B (s, t, u) =
4
70560
6
2
2
s
(4F ) 1411200

1 acn
1
F (s, t, u) + FBver (s) + FBbox (s, t, u) .
+
(4F )4 576 B
(C.1)
(C.2)
Here, the function S(t) is defined by

hF (t)
,
6(t 1)2

1

hF (t) = d ( ) dx 6t + 1 12x + 18x 2 t 2
S(t) =

z2 (x, t) (t 1)x(1 x)
+
,
ln
z2 (x, 1)
z2 (x, 1)

( ) = 1 4/ ,
z2 (x, t) = x 2 + (1 x) (t 1)x(1 x).

(C.3)
102
The loop function hF (t) stems from the sunset diagram. The functions FI are stemming from
the acnode, vertex and box diagrams and are defined by

FIacn
1
d
1
FIver =

6
PI ;acn(1) PI ;acn(i)
d x
+
,
2
yzacn(1)
zacn(i)
i=2
d 2x
0

FIbox
1
d 3x
0
PI ;ver
,
zver
2

(n)
(n)
(n)
PI(n)
;box+ Dbox+ + PI ;box Dbox ,
I = A, B,
(C.4)
n=1
and
(n)
Dbox =
1
1
n
.
n
zbox;t
zbox;u
Here PI are polynomials in s, = t u and in xi . Their explicit expressions are given in Appendix D. The arguments of the functions are defined by
zacn(i) = y ai x3 (1 x3 )
(i = 1, . . . , 6),
y = x32 + (1 x3 ),
a1 = a2 = x1 x2 s + x1 (t 1) + x2 (u 1),
a3 = t 1,
a4 = x1 (t 1),
zver = (1 x3 ) + x32 y2 ,
a5 = u 1,
a6 = x2 (u 1),
y2 = 1 sx2 (1 x2 ),
zbox;t = Bt At x1 ,

At = x2 x3 s(1 x2 )x3 + (1 t)(1 x3 ) ,
zbox;u = zbox;t |tu .
Bt = At + zver ,
(C.5)
The acnode integrals are easy to evaluate numerically in the physical region for the reaction
, because branch points occur at t = 4, u = 4 only. On the other hand, the vertex and
box integrals contain branch points at s = 4. In order to evaluate these integrals at s 4, we
invoke dispersion relations in the manner described in [23].
Appendix D. The polynomials PA and PB
Here, we display the polynomials PA(B) that occur in the expressions A(B) in Appendix C.
We use the abbreviations
x+ = x1 + x2 2x1 x2 ,
x = x 1 x2 ,
x123 = (1 + x3 2x2 x3 )(1 x3 + 2x1 x2 x3 ).

D.1. The polynomials PA
PA;acn(1) = 144x3 (1 x3 )(sx+ x ),
(D.1)
103

2
2
2 2
PA;acn(2) = 12s 2 (1 x3 )2 x+ 5x+ + 6x
1 5x+
x3 6x
x3
2
3
+ x (9 6x+ ) x+ (2 7x+ ) x3

2
+ 6x
(x+ 1) + (3 7x+ )x+ x34

2
14x+ + 10x+ x3
+ 12s(1 x3 )2 x 1 6x

2

2
2
2
+ 6 1 + 2x x+ x3 2 6x
+ 5x+ 1 x33 + 6x
+ 8x+ 3 x34

12 2 (1 x3 )2 6x+ (1 2x+ )x32

2
9 5x3 12x+ x32 (6 18x+ )x33 + (5 12x+ )x34
x

+ 24(1 x3 )x 1 + 2x+ (25 + 2x+ )x3 + 3(7 8x+ )x32

+ (100x+ 41)x33 + (51 118x+ )x34 + (42x+ 19)x35

2
2
x3
17x+ 2x+
12s(1 x3 ) 2x+ (1 + 2x+ ) + 2 2 12x
2
3

2
2
2
+ 2 13 6x 11x+ x3 2 41 18x 89x+ + 8x+ x3
4
5

2
2
2
2
x3 2 15 6x
x3
+ 82 36x
194x+ + 28x+
35x+ + 6x+

+ 48x3 (1 x3 ) 2 4x+ + (35 50x+ )x3

+ 71(2x+ 1)x32 + (61 130x+ )x33 + 2(21x+ 10)x34 ,
PA;acn(3) = 3(s )2 (1 x3 )5 (1 + x3 )

+ 6(s )(1 x3 )3 3 + 5x3 + 2x32 2x33

+ 12x3 (1 x3 )(2 x3 ) 3 + 3x3 + x32 x33 ,
PA;acn(4) = 24(s )2 x12 (1 2x1 )(1 x3 )4 (2 + x3 )

+ 48(s )x1 (1 2x1 )(1 x3 )2 2 + 4x3 3x32

+ 96(1 2x1 )x3 (1 x3 )(2 x3 ) 2 + 2x3 x32 ,

PA;acn(5) = PA;acn(3) tu ,

PA;acn(6) = PA;acn(4) tu,x x ,
1
2

PA;ver = 8s 2 (1 2x2 )x22 x34 54 + 72(3 + 4x2 )x3 + 330 1134x2 811x22 x32

108 5 15x22 4x23 x33 45x2 18 75x2 + 32(3 x2 )x22 x34

+ 32s(1 2x2 )x22 x34 33 + 8(27 7x2 )x3 3 222 30x2 + 55x22 x32

540 1 3x2 + x22 x33 + 1215(1 x2 )2 x34 ,

(1)
PA;box;+ = 4s 2 x22 x33 6x1 9 2(23 + 8x2 )x3 67 405x2 + 31x22 x32

+ 70 + 39x2 808x22 + 20x23 x33

+ 9 6 61x2 + 53x22 + 60x23 x34 81x2 3 10x2 + 8x22 x35

+ 3x12 x2 x3 92 171x3 + 592x2 x3 + 231 + 4(194 339x2 )x2 x32

+ 9(1 2x2 ) 71 22x2 60x22 x33 + 27(1 2x2 )2 (11 16x2 )x34

2x13 x22 x32 245 + (1 2x2 )x3 470 + 27(1 2x2 )x3 15 + 8(1 2x2 )x3

+ 6x3 19 + 24x3 + 35x32 36x33 + x22 x3 25 + 40x3 + 27(5 6x3 )x32
104

+ x2 10 + 44x3 199x32 + 9x33 (8 + 9x3 )

+ 12 2 x23 x34 20 16x1 (2 + 7x1 )

+ 48 + 50x2 + 310x1 + 101x12 + 30x1 3 + 15x1 5x12 x2 x3

+ 78 + 2x1 + 411x12 20 9 + 63x1 + 57x12 + 13x13 x2

+ 120x1 (1 + x1 + 5x12 )x22 x32

3 24 + 3x1 (94 + 73x1 ) 2 69 + 351x1 + 381x12 + 19x13 x2

+ 4x1 24 + 117x1 + 34x12 x22 x33

+ 9(1 2x2 ) 18 + 54x1 36x1 x2 + x12 33 2(9 + 8x1 )x2 x34

48sx2 x32 6 + [18 + 20x2 31x1 x2 ]x3

29 + 49x2 + 4x22 + 4x1 x2 9 (22 8x1 )x2 x32

+ 6 + 202x2 54x22 + 164x1 x2

+ 114x1 x22 4x1 x23 x12 x22 (149 58x2 ) x33

+ 65 333x2 40x22 + 189x1 x2 4x1 (137 + 75x2 )x22

+ 5x12 x22 23 + 12x2 (5 + x2 ) x34

9 4 46x22 x12 x22 (1 2x2 )(41 + 8x2 ) + x1 x2 25 + 18x2 76x22 x35

+ 81x2 (1 2x2 ) 2 3x1 + x1 (2 + x1 )x2 2x12 x22 x36

+ 96x2 x32 21 + 36 + 65(1 x1 )x2 x3

+ 23 8(23 x1 )x2 + 176x1 x22 x32

2 6 (83 + 37x1 )x2 + 120x1 x22 x33

29 + 10(17 7x1 )x2 100x1 x22 x34

9 8 (31 + x1 )x2 + 32x1 x22 x35 + 81(1 2x2 )(1 2x1 x2 )x36 ,

(1)
PA;box; = 8sx22 x33 27x1 57 + 6x1 13 + (26 + 51x1 )x2 x3

+ 72 129x1 + 3 68 + x1 (456 + 25x1 ) x2

+ x1 282 + x1 (1383 470x1 ) x22 x32

+ 105 + 330x1 714x2 3x1 (200 489x1 )x2

2 75 + x1 (1557 + 912x1 + 470x12 ) x22 120x1 3 + 9x1 16x12 x23 x33

+ 9 2x13 x22 (1 2x2 )(13 + 34x2 ) + 3x12 x2 (1 2x2 ) 1 23x2 + 4x22

6x1 (1 2x2 ) 3 + 36x2 + 17x22 6 2 2x2 19x22 x34

+ 27x2 (1 2x2 ) 18 + 18x1 x2 33x12 + 2x12 x2 33 9x2 + 8x1 (1 2x2 ) x35

+ 48x22 x33 18 29x1 21 + 4(10x1 + 12x2 19(2 x1 )x1 x2 ) x3

+ 42 + 106x2 + 252x1 23x12 x2 (7 10x2 ) 6x1 x2 (3 + 20x2 ) x32

83 + 230x2 41x1 + 5x1 x2 134 20x2 3x1 (9 + 10x2 ) x33

9 12 54x2 + x1 (1 2x2 ) 53 (24 + 41x1 )x2 x34
105

+ 81(1 2x2 ) 2 + 3x1 x1 (4 + x1 )x2 x35 ,

(2)
PA;box;+
= 6s 3 1 + x1 (1 2x2 ) x23 x36 x123

8 5x2 + 8x1 x1 (6 + 5x1 )x2 6x3

2
6x1 (1 2x2 )x3 + 9x2 1 + x1 (1 2x2 ) x32
6s 2 (1 x1 )2 x23 x36 x123

8 6x3 + x2 1 + x1 (1 2x2 ) 5 3x3 (4 9x3 )

+ 12s 2 x22 x34 x123 2 2 + x2 + 2x1 (6 x1 )x1 x2 12 1 + x1 (1 2x2 ) x3

+ 13(2 x2 ) + 2x1 (1 x2 )(13 16x2 ) x12 x2 13 32x2 + 32x22 x32

12 1 + x1 + x2 + x12 (1 2x2 )2 x2 4x1 x22 x33

2
+ 27x2 1 + x1 (1 2x2 ) x34

12 2 (1 x1 )2 x23 x34 x123 2 + 12x3 17x32 + 24x33 27x34

+ 24sx2 x34 x123 6(2 x3 )(1 x3 )x3

+ 2x1 x22 23 24x3 + x32 (2 3x3 )(2 9x3 )

+ (1 + x1 )x2 23 + 24x3 4x32 + 3(8 9x3 )x33

+ 144x2 (1 x3 )2 x34 x123 5 + x3 (2 + 3x3 ) ,

(2)
PA;box; = 6 3 (1 x1 )3 x24 x36 x123 5 6x3 + 9x32

+ 6s 2 (1 x1 )x23 x36 x123 16 + 16x1 5x2 22x1 x2 5x12 x2

6 2 + (1 + x1 )x2 2x1 x22 1 + x1 (1 2x2 ) x3

2
+ 27x2 1 + x1 (1 2x2 ) x32

24s(1 x1 )x22 x34 x123 2 6x3 + (13 6x3 )x32

x2 1 + x1 (1 2x2 ) 6 2x3 + 9x32 (2 3x3 ) x3

+ 72(1 x1 )(1 x3 )x22 x34 x123 5 7x3 + (1 9x3 )x32 .
D.2. The polynomials PB

PB;acn(1) = 144x3 (1 x3 ) x+ x ,
s
2

2

12
PB;acn(2) =
(1 x3 )2 3 5x3 x34 x
+ 6x32 x+
s

24x
(1 x3 )2 7 12x3 + 3x32 + 4x33 5x34
+
s

48
x3 (1 x3 ) 14 + 19x3 43x32 + 29x33 10x34
s

2
2
+ 6x32 x
+ 12s(1 x3 )2 x+ (1 + 3x+ ) 5x3 x+
2
2

x33 9x
x+ (2 + 3x+ ) + x34 6x
x+ (3 + x+ )

12x (1 x3 )2 1 + 6x+ 10x3 x+ + 6x32 (1 + x+ )
106

+ 2x33 (1 3x+ ) x34 (3 4x+ )

24(1 x3 )2 7x+ 2(1 + 6x+ )x3

(1 x+ )x32 (15 26x3 ) (15 13x+ )x34 ,
3(s )2
(1 x3 )5 (1 + x3 )
s

6(s )
+
(1 x3 )3 3 + 5x3 + 2x32 (1 x3 )
s

12
+ x3 (1 x3 )(2 x3 ) 3(1 + x3 ) + x32 (1 x3 ) ,
s
24(s )2 2
PB;acn(4) =
x1 (1 x3 )4 (2 + x3 )
s

48(s )
x1 (1 x3 )2 2 + x3 (4 3x3 )
s

96
x3 (1 x3 )(2 x3 ) 2 + (2 x3 )x3 ,
s

PB;acn(5) = PB;acn(3) tu ,

PB;acn(6) = PB;acn(4) tu,x x ,
1
2

PB;ver = 8sx22 (1 2x2 )x34 54 8(69 + 50x2 )x3 + 15 70 + x2 (94 + 77x2 ) x32

180 3 + x2 (4 + 23x2 ) x33 405x2 (2 9x2 )x34

96x22 (1 2x2 )x34 25 144x3 + 500x32 780x33 + 405x34 ,
PB;acn(3) =

12 2 3
x2 (1 x3 )x34 20 + 16x1 (7 + 9x1 ) 98x3 x1 (574 + 573x1 )x3
s

+ 6 32 + x1 (166 + 117x1 ) x32 27 6 + x1 (18 + 11x1 ) x33

288
x2 (1 x3 )2 x32 7 17x3 + 14x32 17x33 + 27x34
s

12sx22 x33 18x1 + 38 + 4 51x1 5x2 + 45x12 x2 x3

156 50x2 + 322x1 + 586x1 x2 + 5x12 x2 (85 + 74x2 ) x32

+ 5 38 + 18x2 + 4x1 + 278x1 x2 + 76x1 x22 3x12 x2 7 4x2 (17 + 7x2 ) x33

3 24 + 94x2 36x1 + 130x1 x2 + 424x1 x22

3x12 x2 (1 2x2 )(37 + 76x2 ) x34

+ 27x2 6 x1 (1 2x2 ) 18 11x1 (1 2x2 ) x35

+ 48x2 x32 6 (34 + 25x1 x2 )x3 + 97 x2 7 16x1 (7 + 4x2 ) x32

2 79 + x2 13 + x1 (191 + 118x2 ) x33

+ 5 25 + x2 45 + x1 (101 + 110x2 ) x34

3 12 + x2 118 x1 (7 284x2 ) x35 + 81x2 2 3x1 (1 2x2 ) x36 ,

48 2
(1)
x (1 x3 )x33 14 + 11x1 9(7 + 9x1 )x3
PB;box; =
s 2

+ (127 + 373x1 )x32 12(19 + 46x1 )x33 + 81(2 + 3x1 )x34
(1)
PB;box;+ =

+ 24x22 x33 9x1 + 19 + 2x1 41 + (38 + 81x1 )x2 x3

78 + 34x2 + 189x1 + x1 x2 524 + 84x2 + 3x1 (193 + 59x2 ) x32

+ 5 19 + 38x2 + 34x1 + x1 x2 224 + 78x2 + 9x1 (9 + 20x2 ) x33

3 2(6 + 53x2 ) + 18x1 + x1 x2 (240 111x1 + 248x2 + 447x1 x2 ) x34

+ 27x2 6 + 18x1 x2 11x12 (1 2x2 ) x35 ,
(2)

12 2
(1 x1 )2 x23 (1 x3 )2 x34 2 + 12x3 17x32 + 24x33 27x34
s

144
+
x2 (1 x3 )4 x34 5 + x3 (2 + 3x3 )
s

+ 12sx22 (1 x3 )2 x34 2 2 + x2 + x1 2 (6 x1 )x2

12 1 + x1 (1 2x2 ) x3

+ 13(2 x2 ) + 26x1 x1 x2 58 32x2 + x1 13 32(1 x2 )x2 x32

12 1 + x1 + x2 + x12 (1 2x2 )2 x2 4x1 x22 x33

2
+ 27x2 1 + x1 (1 2x2 ) x34

6s 2 x23 1 + x1 (1 2x2 ) (1 x3 )2 x36

8 6x3 x2 5 9x32 x12 x2 5 9(1 2x2 )2 x32

+ 2x1 4 3x3 18x22 x32 3x2 1 2x3 3x32
PB;box;+ =
6 2 (1 x1 )2 x23 (1 x3 )2 x36

8 6x3 + x2 1 + x1 (1 2x2 ) 5 3x3 (4 9x3 )

+ 6x2 (1 x3 )2 x34 24(2 x3 )(1 x3 )x3

4(1 + x1 )x2 23 x3 24 x3 (2 3x3 )(2 9x3 )

+ 8x1 x22 23 x3 24 x3 (2 3x3 )(2 9x3 ) ,
(2)
PB;box; =

6 3
(1 x1 )3 x24 (1 x3 )2 x36 5 6x3 + 9x32
s

72
+
(1 x1 )x22 (1 x3 )3 x34 5 x3 7 x3 (1 9x3 )
s

+ 6s(1 x1 )x23 (1 x3 )2 x36 16(1 + x1 ) 5x2 22x1 x2 5x12 x2

6 2 + (1 + x1 )x2 2x1 x22 1 + x1 (1 2x2 ) x3

2
+ 27x2 1 + x1 (1 2x2 ) x32

24(1 x1 )x22 (1 x3 )2 x34 2 x3 6 x3 (13 6x3 )

+ (1 + x1 )x2 6 x3 2 9(2 3x3 )x3

2x1 x22 6 x3 2 9x3 (2 3x3 )
.
References
[1] S. Weinberg, Physica A 96 (1979) 327.
[2] J. Gasser, H. Leutwyler, Ann. Phys. 158 (1984) 142.
107
108
[3] J. Gasser, H. Leutwyler, Nucl. Phys. B 250 (1985) 465.

[4] U. Burgi, Nucl. Phys. B 479 (1996) 392, hep-ph/9602429;
U. Burgi, Phys. Lett. B 377 (1996) 147, hep-ph/9602421.
[5] J. Gasser, M.A. Ivanov, M.E. Sainio, Nucl. Phys. B 728 (2005) 31, hep-ph/0506265.
[6] G. Colangelo, J. Gasser, H. Leutwyler, Nucl. Phys. B 603 (2001) 125, hep-ph/0103088.
[7] J. Bijnens, P. Talavera, Nucl. Phys. B 489 (1997) 387, hep-ph/9610269.
[8] J. Ahrens, et al., Eur. Phys. J. A 23 (2005) 113, nucl-ex/0407011.
[9] S. Scherer, hep-ph/0512291.
[10] L.V. Filkov, V.L. Kashevarov, Phys. Rev. C 73 (2006) 035210, nucl-th/0512047.
[11] J. Bijnens, G. Colangelo, G. Ecker, JHEP 9902 (1999) 020, hep-ph/9902437.
[12] J. Bijnens, G. Colangelo, G. Ecker, Ann. Phys. 280 (2000) 100, hep-ph/9907333.
[13] H.W. Fearing, S. Scherer, Phys. Rev. D 53 (1996) 315, hep-ph/9408346].
[14] C.Q. Geng, I.L. Ho, T.H. Wu, Nucl. Phys. B 684 (2004) 281, hep-ph/0306165.
[15] S. Bellucci, J. Gasser, M.E. Sainio, Nucl. Phys. B 423 (1994) 80, hep-ph/9401206;
S. Bellucci, J. Gasser, M.E. Sainio, Nucl. Phys. B 431 (1994) 413, Erratum.
[16] J. Bijnens, J. Prades, Nucl. Phys. B 490 (1997) 239, hep-ph/9610360.
[17] G. Ecker, J. Gasser, H. Leutwyler, A. Pich, E. de Rafael, Phys. Lett. B 223 (1989) 425;
G. Ecker, J. Gasser, A. Pich, E. de Rafael, Nucl. Phys. B 321 (1989) 311;
M. Knecht, A. Nyffeler, Eur. Phys. J. C 21 (2001) 659, hep-ph/0106034;
J. Bijnens, E. Gamiz, E. Lipartia, J. Prades, JHEP 0304 (2003) 055, hep-ph/0304222;
V. Cirigliano, G. Ecker, M. Eidemuller, R. Kaiser, A. Pich, J. Portoles, JHEP 0504 (2005) 006, hep-ph/0503108.
[18] M. Knecht, B. Moussallam, J. Stern, Nucl. Phys. B 429 (1994) 125, hep-ph/9402318.
[19] B.R. Holstein, Phys. Lett. B 244 (1990) 83.
[20] S. Descotes-Genon, B. Moussallam, Eur. Phys. J. C 42 (2005) 403, hep-ph/0505077.
[21] J. Bijnens, F. Cornet, Nucl. Phys. B 296 (1988) 557.
[22] U. Burgi, Ph.D. thesis, University of Bern, 1996.
[23] J. Gasser, M.E. Sainio, Eur. Phys. J. C 6 (1999) 297, hep-ph/9803251.
[24] J.A.M. Vermaseren, math-ph/0010025.
[25] J. Boyer, et al., Phys. Rev. D 42 (1990) 1350.
[26] I. Guiasu, E.E. Radescu, Ann. Phys. 120 (1979) 145;
I. Guiasu, E.E. Radescu, Ann. Phys. 122 (1979) 436.
[27] L.V. Filkov, V.L. Kashevarov, Phys. Rev. C 72 (2005) 035211, nucl-th/0505058.
[28] J.F. Donoghue, B.R. Holstein, Phys. Rev. D 48 (1993) 137, hep-ph/9302203.
[29] D. Babusci, S. Bellucci, G. Giordano, G. Matone, A.M. Sandorfi, M.A. Moinester, Phys. Lett. B 277 (1992) 158.
[30] H. Aihara, et al., TPC/Two-Gamma Collaboration, Phys. Rev. Lett. 57 (1986) 404.
[31] H.J. Behrend, et al., CELLO Collaboration, Z. Phys. C 56 (1992) 381.
[32] F. Yabuki, et al., VENUS Collaboration, J. Phys. Soc. Jap. 64 (1995) 435.
[33] A. Heister, et al., ALEPH Collaboration, Phys. Lett. B 569 (2003) 140.
[34] H. Nakazawa, et al., BELLE Collaboration, Phys. Lett. B 615 (2005) 39, hep-ex/0412058.
[35] A.E. Kaloshin, V.V. Serebryakov, Z. Phys. C 64 (1994) 689, hep-ph/9306224.
[36] H. Marsiske, et al., Crystal Ball Collaboration, Phys. Rev. D 41 (1990) 3324.
[37] C. Berger, et al., PLUTO Collaboration, Z. Phys. C 26 (1984) 199.
[38] A. Courau, et al., Nucl. Phys. B 271 (1986) 1.
[39] Z. Ajaltouni, et al., DM1DM2 Collaboration, Phys. Lett. B 194 (1987) 573;
Z. Ajaltouni, et al., DM1DM2 Collaboration, Phys. Lett. B 197 (1987) 565, Erratum.
[40] T.A. Aibergenov, et al., Czech. J. Phys. B 36 (1986) 948.
[41] Y.M. Antipov, et al., Z. Phys. C 26 (1985) 495;
Y.M. Antipov, et al., Phys. Lett. B 121 (1983) 445.
[42] S. Paul, AIP Conf. Proc. 814 (2006) 163, hep-ex/0511008.
[43] S. Scherer, private communication.
Nuclear Physics B 745 [PM] (2006) 109122
Conformal sigma models on supercoset targets

David Kagan a, , Charles A.S. Young b
a DAMTP, Centre for Mathematical Sciences, University of Cambridge, Wilberforce Road, Cambridge CB3 0WA, UK
b Department of Mathematics, University of York, Heslington Lane, York YO10 5DD, UK
Received 12 January 2006; accepted 21 February 2006

Abstract
We investigate the quantum behaviour of sigma models on coset superspaces G/H defined by Z2n gradings of G. We find that, whenever G has vanishing Killing form, there is a choice of WZ term which renders
the model quantum conformal, at least to one loop. The choice coincides with that for which the model is
known to be classically integrable. This generalizes results for models associated to Z4 gradings, including
IIB superstrings in AdS5 S 5 .
1. Introduction and overview

Non-linear sigma models with supermanifolds as targets are of importance both in string
theory and condensed matter physics. In condensed matter they are relevant in a variety of applications [1], notably the theory of the quantum Hall effect [2]. There are interesting connections
to Lagrangian formulations of logarithmic conformal field theories [3].
In string theory, the proposal of the AdS/CFT correspondence [4] prompted renewed interest
in superstrings on curved RamondRamond backgrounds.1 The GreenSchwarz action for IIB
strings in AdS5 S 5 , given by Metsaev and Tseytlin in [6] (see also [7]), takes the form of a
sigma model on the coset superspace PSU(2, 2|4)/SO(1, 4) SO(5). To gain insight into how
this theory might be quantized, sigma models on simpler targets, the supergroups PSL(n|n), were
considered in [8,9] (see also [10]). As we recall below, these supergroups have vanishing Killing
E-mail addresses: d.kagan@damtp.cam.ac.uk, dk285@cam.ac.uk (D. Kagan), charlesyoung@cantab.net

(C.A.S. Young).
1 For more on the technical issues of formulating string theories on supermanifold targets see [5].
doi:10.1016/j.nuclphysb.2006.02.027
110
D. Kagan, C.A.S. Young / Nuclear Physics B 745 [PM] (2006) 109122
form and are therefore Ricci-flat, which guarantees that sigma models on them are conformal to
one-loop; the result of Bershadsky, Zhukov and Vaintrob in [8] is that they are exactly conformal. (The representation theory of the relevant superalgebras [11] thus becomes important.) This
remarkable fact should be contrasted with the more familiar sigma models on bosonic groups,
where a correctly normalized WZ term (which may be thought of as a parallelizing torsion [12])
must be added to the action if the theory is to be quantum conformal [13].
It was subsequently shown by Berkovits, Bershadsky, Hauer, Zhukov and Zwiebach [14] that
sigma models on certain quotients G/H of Ricci-flat supergroups G by bosonic subgroups are
again conformal to one loopand can be expected to be so exactlygiven a suitable WZ
term. These targets include the coset superspace PSU(2, 2|4)/SO(1, 4) SO(5), as well as
PSU(1, 1|2)/U (1) U (1), whose bosonic geometry is AdS2 S 2 . The vital property in each
case is that the isotropy subgroup H is the fixed point set of a Z4 grading of G. This, in particular, allows the addition of the required WZ term.
What is striking is that the Z4 grading was also a key element in the demonstration by Bena,
Polchinski and Roiban [15] that the sigma model on PSU(2, 2|4)/SO(1, 4) SO(5) is classically integrable. (See also [25], which contains references to the extensive recent work on this
subject.) It was shown by one of the present authors [16] that a similar construction holds, more
generally, for sigma models on spaces G/H defined by Zm gradings: for any m, the grading permits the addition of a certain preferred WZ term, and with this WZ term the equations of motion
may be put into Lax form, ensuring integrability, just as in the Z4 case.
Given this result, and the dual role played by the Z4 -grading, it is natural to hope that conformal invariance is also present in sigma models whose targets are quotients of (again, Ricci-flat)
supergroups G by subgroups H defined by gradings of order greater than 4. In this paper, we
show that this is indeed the case, at least at one loop.
The structure of this paper is as follows: in Section 2 we introduce some notation and discuss
supergroups with vanishing Killing form and related coset superspaces. We also write down
the sigma model action with WZ term, and recall the one-loop beta-functions. In Section 3 we
compute the Ricci curvature of a torsion connection on a reductive homogeneous superspace, and
then in Section 4 we demonstrate that, for the connection whose torsion is given by the preferred
WZ term, this curvature vanishes, completing our argument. Some conclusions, and comments
about the consequences of this result, are given in Section 5.
2. Sigma models on graded coset superspaces
We shall consider non-linear sigma models on certain homogeneous superspaces. We begin
by summarizing the relevant facts about Lie supergroups and establishing some notation. For full
details see [17].
2.1. Supergroups
Let G be a Lie supergroup and let g be the complexification of its Lie superalgebra. Write
g = g(0) + g(1)
(2.1)
111
for the decomposition of g into its even and odd parts, and pick a basis {ta } of g consisting of the
disjoint union of bases of g(0) and g(1) . For every ta g(0) , let |a| be an even number; for every
ta g(1) , let |a| be an odd number.2
We assume for simplicity that a suitable representation has been chosen and the ta are concrete
matrices. Elements of G near the identity are then of the form
1 + X a ta + ,
(2.2)
where, for each a, the parameter

is Grassmann even (i.e. a c-number) if |a| is even and
Grassmann odd if |a| is odd.3 Since G is a group, the commutator
a

X ta , Y b tb = X a ta Y b tb Y b tb X a ta
(2.3)
Xa
closes onto some Z c tc , and, given the statistics of the X a and Y b , this implies that g closes under
the graded commutator:
[ta , tb ] = ta tb ()|a||b| tb ta = f c ab tc ,
where we define also the structure constants f a bc . It follows from

0 = Z, [X, Y ] + X, [Y, Z] + Y, [Z, X]
(2.4)
(2.5)
for X = X a ta , Y = Y a ta and Z = Z a ta that the f a bc satisfy the graded Jacobi identity:

0 = ()|a||c| f e ad f d bc + ()|b||a| f e bd f d ca + ()|c||b| f e cd f d ab .
(2.6)
The Killing form K(X, Y ) = Kab X a Y b on g is defined by

Kab = ()|d| f d ac f c bd
(2.7)
and is an invariant (i.e. K([X, Z], Y ) + K(X, [Y, Z]) = 0), graded-symmetric, bilinear form. The
unusual feature of Lie superalgebras, in comparision with Lie algebras, is that there exist simple
Lie superalgebras whose Killing forms vanish identically. As we recall in Section 3 below, the
vanishing of the Killing form implies that G is Ricci-flat. This will be crucial for conformal
invariance, so we will focus on these cases. However, even if Kab is zero, g may possess a nondegenerate, invariant, graded-symmetric bilinear form. We assume that such a form exists and
denote it by
, .
(2.8)
(In the examples in Appendix A, , is given by the supertrace in the defining representation,
rather than the adjoint.)
2.2. Graded coset superspaces
We are concerned with models whose targets are coset superspaces G/H with the special
property that g is Z2n -graded and h, the complexified Lie algebra of the isotropy subgroup H , is
2 |a| will be defined fully in (2.12) below; only |a| modulo 2 is relevant here.
3 These parameters will also satisfy some reality conditions, which depend on the choice of real form of g. (For
example, if the real form is the real span of the ta , the conditions are simply that the Xa are real.)
112
the subspace of grade zero. That is, we suppose that g may be written as a direct sum
g = g0 + g1 + + g2n1
(2.9)
of vector subspaces (where g0 = h), and that this decomposition respects the graded Lie bracket
(2.4):
[gr , gs ] gr+s mod 2n.
(2.10)
We suppose further that this grading is compatible with the splitting (2.1) of g into its even
and odd graded subspaces, in the sense that
g2s g(0) ,
g2s+1 g(1) ,
for s = 0, 1, . . . , n 1.
(2.11)
The basis {ta } can be chosen to be a disjoint union of bases of the gk , and one can then consistently define
|a| = s ta gs .
(2.12)
It is also useful to adopt the following conventions for the naming of indices:
a, b, . . .
generators of g,
, , . . .
generators of h,
i, j, . . .
generators of g \ h.
(2.13)
Finally, we assume that the inner product , respects the grading, in the sense that if X gr
and Y gs then X, Y = 0 unless r + s = 0 mod 2n. Non-degeneracy implies that if X gr and
X, Y = 0 for all Y g2nr then X = 0.
One family of examples of such coset superspaces is given in Appendix A.
2.3. The action
Sigma models on G/H are most conveniently expressed in terms of a dynamical field g(x )
G (where x are worldsheet coordinates, and will be the worldsheet metric). We write
j = g 1 g g
(2.14)
for the current invariant under the global left action

g Ug,
U G
(2.15)
of G on itself. This decomposes into currents of definite grade:

j = j t + ji ti
(2.16)
and under the local right action

g gh with h(x ) H,
corresponding to the redundancy in the choice of representative g from the coset gH ,

i
ji h1 j h .
j h1 j h + h1 h ,
(2.17)
(2.18)
It is now possible to write down the most general local action for a sigma model with B-field
constructed using , and with the required symmetries. Global G symmetry is ensured by
113
writing the action in terms of j ; local H symmetry requires that it takes the form

S = d 2 x p|i| + q|i| ji jj ti , tj ,
(2.19)
where ps = p2ns and qs = q2ns are arbitrary real coefficients. (In this expression the summation convention applies as usual over the indices i, j . The point is that since we require only
H -invariance we can rescale each pair (s, 2n s) of subspaces separately, both in the kinetic and
WZ terms.)
However, a certain choice of the coefficients qs is privileged. It was shown in [16] that
the sigma model becomes classically integrable if4
s
qs = 1 .
(2.20)
n
The goal of the present paper is to show that, for supergroups G with vanishing Killing form,
the same choice of qs also ensures conformal invariance, at least at one loop order. To see this,
we first recall some well-known [18] facts about the renormalization-group behaviour of sigma
models.
2.4. Sigma model -functions to one-loop
We begin by quoting the -functions for the general non-linear sigma model with B-field. Let
m, n, . . . denote coordinate-induced tangent-vector indices on G/H . Then the classical action is

1
d 2 x ( gmn + Bmn ) m n ,
S=
(2.21)
2
where m (x ) is the sigma-model field, gmn is the metric on the target space and Bmn = Bnm
can be regarded as the potential for the torsion 3-form H = dB. Under changes in the energy
scale , these quantities are renormalized according to

gmn = (mn) and
Bmn = [mn] ,
with, to one-loop order [18],

h
mn = R mn + O h 2 ,
(2.22)
2
where R is the generalized Ricci curvature of the metric connection for g with torsion H . To
connect the notation in (2.19) with the general sigma model action (2.21), define

a
a
em
(2.23)
= g 1 m g ,
j
a m and the metric and B-field in (2.19) are g

i
i
so that ja = em
mn = em en gij , Bmn = em en Bij

with
gij = ti , tj p|i| ,
Bij = ti , tj q|i| .
(2.24)
The task is now purely geometrical: we must show that the choice of WessZumino (i.e. Bfield) term given by the coefficients qs in (2.20) renders G/H Ricci-flat. In what follows it is
easiest to work in the non-coordinate-induced basis (indices i, j, . . .) of tangent vectors on G/H
i .
defined by the vielbeins em
1
4 As in that paper, we specify set p = = p
1
n1 = 1, pn = 2 , which is the natural choice in the sense that the kinetic
term is then simply 12 (j A) (j A) . This gives us the sigma model for the Berkovits action in the Z4 case, rather
than the GreenSchwarz action which would correspond to the choice p1 = 0, p2 = 1/2.
114
3. Ricci curvature of G/H

3.1. The torsion-free connection
The identity de + e e = 0 satisfied by the (super-)MaurerCartan form
g 1 dg = e = ea ta = e t + ei ti
(3.1)
yields the structure equations

1
dei = f i j k ej ek f i j ej e ,
2
1
1
de = f ij ei ej f g e eg .
2
2
(3.2)
(3.3)
From the first of these one reads off the components of the torsion-free (dei + i j ej = 0)
connection 1-form
1
i j = ()|j ||k| f i j k ek f i j e ,
(3.4)
2
whose curvature 2-form is
Ri j = di j + i p p j

1
= ()|j ||p| f i jp f p kl + ()|k||p|+|j ||l| f i pk f p j l + 2f i j f kl ek el .
4
(3.5)
(The graded Jacobi identity (2.6) ensures that the components proportional to ek e and e eg
vanish, as they must to maintain H -gauge-covariance.) The curvature tensor is then
1
1
R i j kl = ()|j ||p|+|k||l| f i jp f p lk + ()|k||p|+|j ||l| f i pk f p j l
2
4
1
+ ()|j ||k|+|k||l| f i lp f p j k + f i j f kl
4
(3.6)
and the Ricci tensor, which is graded-symmetric in i j given the Jacobi identity, is5
1
Rij = ()|k| R k j ki = ()|k| f k ip f p j k f ip f p j .
(3.7)
4
Note in particular that the Ricci curvature of G itself is proportional to the Killing form (to see
this, suppose H were the trivial subgroup). Now, from the definition (2.7),
Kij = 2f ip f p j + ()k f k ip f p j k
(3.8)
and hence
1
1
Rij = ()|k| f k ip f p j k Kij ,
4
2
so that, finally, if the Killing form vanishes then
1
Rij = ()|k| f k ip f p j k .
4
5 Note that f k = 0 by virtue of the gradation of g.
kp
(3.9)
(3.10)
115
3.2. The torsion connection

We now turn to the generalized Ricci tensor in (2.22). The connection with torsion H = dB
is
1
i j = i j + H i j ,
2
(3.11)
where H i j = H i kj ek , for then

dei + i j ej = H i ,
(3.12)
where H i = 12 H i j k ej ek is the torsion 2-form. The curvature 2-form is

R i j = d i j + i p p j
1
1
1
1
= Ri j + dH i j + H i p p j + i p H p j + H i p H p j
2
2
2
4
1 i
i
p
=R j + H p H j
4

+ k H i lj H i pj p lk ()|k||l| H i lp p j k + i pk H p lj ek el
and we write the expression in parentheses here as k
Hi
lj .
(3.13)
Then
1
1
1
1
R i j kl = R i j kl + H i kp H p lj ()|k||l| H i lp H p kj + k H i lj ()|k||l| l H i kj ,
4
4
2
2
(3.14)
and the generalized Ricci tensor is
1
1
R ij = ()|k| R k j ki = Rij + ()|k| H k ip H p j k + ()|k| k H k ij .
(3.15)
4
2
It now remains to show that this quantity vanishes, implying one-loop conformal invariance of
the sigma model.
4. Vanishing of the -functions
We now return to the particular B-field specified by (2.20). As noted in [16], this choice of qs
leads to a particularly simple form for the torsion H = dB. Using the structure equations (3.3)
one finds that

+fij k , |i| + |j | + |k| = 2n,
Hij k = fij k , |i| + |j | + |k| = 4n,
(4.1)
0,
otherwise
(where the third index on f is lowered using the metric , ) and hence that the torsion tensor
H i j k , with indices in their natural positions, is
+f i j k , |j | + |k| = |i|,
i
H j k = f i j k , |j | + |k| = 2n + |i|,
(4.2)
0,
otherwise.
Now since the components H i j k are constant (l H i j k = 0) and invariant under the action of the
isotropy subgroup H , we have, from the definition of in (3.13),
116
l H k ij = H p ij k pl H k pj p il ()|i||l| H k ip p j l
= H p ij f k lp H k pj f p li ()|i||l| H k ip f p lj
(4.3)
and hence
()|k| k H k ij = ()|k| H k pj f p ki ()|k||i|+|k| H k ip f p kj

= ()|k| H k ip f p j k ()|i||j | H k jp f p ik .
(4.4)
(4.5)
(To reach the final line note that f i j k and H i j k are non-zero only when |i| = |j | + |k| mod 2n.)
The generalized Ricci curvature is therefore

1

1
R ij = ()|k| f k ip f p j k + H k ip H p j k + ()|k| H k ip f p j k ()|i||j | H k jp f p ik .
4
2
(4.6)
It follows almost immediately that the metric -function vanishes: the graded-symmetric part
of (4.6) is proportional to

()|k| f k ip f p j k + H k ip H p j k
(4.7)
and if, in any given term in the summation on the right, |j | + |k| = |p| then, since |i| = 2n |j |,
|i| + |p| = 2n + |k|; conversely if |j | + |k| = 2n + |p| then |i| + |p| = |k|. So in each term
precisely one minus sign appears, and the H H and ff summations cancel perfectly.
We now turn to demonstrating that the B-field -function vanishes. This argument is considerably more subtle. The graded-antisymmetric part of (4.6) is

1
R [ij ] = ()|k| H k ip f p j k ()|i||j | H k jp f p ik .
(4.8)
2
We shall treat this grade by grade. Suppose |i| = r and |j | = 2n r with r {1, 2, . . . , 2n 1}.
Then, writing kt for an index of the grade t subspace, we find
R [ir j2nr ] = H p1 ir k2nr+1 f k2nr+1 j2nr p1 + H p2 ir k2nr+2 f k2nr+2 j2nr p2
()r H pr1 ir k2n1 f k2n1 j2nr pr1 ()r H pr+1 ir k1 f k1 j2nr pr+1 +
H p2n1 ir k2nr1 f k2nr1 j2nr p2n1
= f p1 ir k2nr+1 f k2nr+1 j2nr p1 f p2 ir k2nr+2 f k2nr+2 j2nr p2
+ ()r f pr1 ir k2n1 f k2n1 j2nr pr1 ()r f pr+1 ir k1 f k1 j2nr pr+1 +
f p2n1 ir k2nr1 f k2nr1 j2nr p2n1 ,
(4.9)
where the factor 12 disappears because the definition of H ensured that the two terms in (4.8)
simply reinforced one another (or, equivalently, that the first term automatically had the necessary
graded-antisymmetry).
The plan is to show that (4.9) can be re-written as a certain sum of terms involving f kt kt . To
do this we use the vanishing of the ij th component of the Killing form, which it is useful to write
as a sum of separate summations restricted to subspaces of definite grade:
0 = f ir k2nr f k2nr j2nr f p1 ir k2nr+1 f k2nr+1 j2nr p1 + f p2 ir k2nr+2 f k2nr+2 j2nr p2
()r f pr1 ir k2n1 f k2n1 j2nr pr1 + ()r f pr ir k f k j2nr pr
()r f pr+1 ir k1 f k1 j2nr pr+1 + f p2n1 ir k2nr1 f k2nr1 j2nr p2n1 .
(4.10)
117
We require also the graded Jacobi identity (2.6), which, for the components of interest, takes the
form
0 = ()r|e| f a ir b f b j2nr e + ()r f a j2nr b f b eir + ()r|e| f a eb f b ir j2nr .
(4.11)
The key idea is that this, since it holds for all indices a and e, holds in particular when a = e.
One can sum the resulting identities over any set of indices a one chooses. The sums relevant
here are those over the (bases of the) individual subspaces of grades r = 1, 2, . . . , 2n 2, 2n 1:
0 = f p1 ir k2nr+1 f k2nr+1 j2nr p1 + f p1 j2nr kr+1 f kr+1 p1 ir + f p1 p1 f ir j2nr ,
0 = f p2 ir k2nr+2 f k2nr+2 j2nr p2 + ()r f p2 j2nr kr+2 f kr+2 p2 ir + f p2 p2 f ir j2nr ,
..
.
0 = f p2n2 ir k2nr2 f k2nr2 j2nr p2n2 + ()r f p2n2 j2nr kr2 f kr2 p2n2 ir
+ f p2n2 p2n2 f ir j2nr ,
0 = f p2n1 ir k2nr1 f k2nr1 j2nr p2n1 + f p2n1 j2nr kr1 f kr1 p2n1 ir
+ f p2n1 p2n1 f ir j2nr .
(4.12)
Consider now subtracting from Eq. (4.9) a linear combination (with coefficients, say, a1 ,
a2 , . . . , a2n2 , a2n1 ) of these identities and a multiple b of the identity (4.10). The goal
is to reduce the right-hand side of (4.9) to an expression involving only terms of the type
f pt pt f ir j2nr . This demand produces a set of linear equations for the coefficients ar and b.
First, the terms f ir k2nr f k2nr j2nr and f pr ir k f k j2nr pr do not appear if and only if
0 = ()r b + ar ,
0 = b ()r a2nr ,
(4.13)
respectively. And the remaining termsthe terms which are present in (4.9) as it standsare
cancelled if and only if
1 = a1 ()r a2nr+1 b,
1 = a2 ()r a2nr+2 + b,
1 = a3 ()r a2nr+3 b,
..
.
()r 1 = ar1 ()r a2n1 ()r b,
()r 1 = ar+1 ()r a2n+1 ()r b,
..
.
1 = a2n3 ()r a2nr3 b,
1 = a2n2 ()r a2nr2 + b,
1 = a2n1 ()r a2nr1 b.
(4.14)
118
These equations are soluble: a solution6 is

r
t
t
at = () 1
,
b=1 ,
n
n
and one is left, after these subtractions, with

2n1

t
t
() 1
f kt .
fir j2nr
n kt
(4.15)
(4.16)
t=1
Since the summation here does not depend on r, we have shown that
R [ij ] = f ij
2n1

()
t=1

t
f kt .
1
n kt
(4.17)
But now, just as for the Z4 case in [14], the two-form with these components is exact, and in
fact zero if H is semisimple. The argument for exactness is as follows: call the sum on the right
V , and note that the graded Jacobi identity implies that V f vanishes.7 It then follows from
the structure equations (3.3) that
R [ij ] ei ej dV e
(4.19)
and hence that the B-field is renormalized only by a total derivative, which has no physical effect
since it does not modify the torsion H = dB.
Furthermore, if the isotropy subgroup H is semisimple then every generator T of H can be
expressed in terms of commutators of other generators and so the vanishing of V f actually
implies the vanishing of V . The two-form B is then not merely exact but actually zero.
5. Conclusions and outlook
In this paper we have calculated the generalized Ricci curvature of coset superspaces G/H
defined by Z2n gradings of Ricci-flat supergroups. We have shown that these supercosets admit
a Ricci-flat torsion connection, and that, consequently, sigma models with such spaces as targets
are quantum conformal to one loop for an appropriate choice of WZ term. As in the Z4 case in
[14] we expect that conformal invariance is exact, but further work is needed to establish this.
This result provides strong evidence to support the suggestion, made in [8], that quotienting
Ricci-flat supergroupsspecifically products of PSL(n|n)by suitable subgroups might be a
fruitful way to find new Lagrangian conformal field theories. Theories of this type have been
vital in the study of the gauge/string correspondence, and we hope that a more general class of
such theories might be used to shed further light on this remarkable subject [29,30].
Another reason for interest in these backgrounds is their possible relevance to mirror symmetry. Some years ago, it was argued [1921] that certain Ricci-flat supergroups could profitably
6 The solution is not always unique. In particular when r = n the situation is simplified since b = 0 and there is some
freedom in the choice of the ai .
7 For each grade t , we have
f kt kt f + f kt pt f pt kt + f kt pt f pt kt = 0
and, on renaming the indices, it is clear that the last two terms cancel.
(4.18)
119
be regarded as CalabiYau supermanifolds, and that they actually arise as mirrors of rigid
CalabiYau manifolds. This prompted work on Ricci-flat supermanifolds [2224]. Our targets
are Ricci-flat only in the generalized sense, but one can speculate that there are links to be
made.
Finally, it is interesting that the same WZ term needed for classical integrability, as evidenced
by the existence of one-parameter families of flat currents, is also necessary for generalized Ricci
flatness of the target. One can hope that there are deeper connections to be made between Ricciflat targets and classical integrability.
Acknowledgements
We thank Jonathan Evans for his helpful comments. D.K. is grateful for an NSF graduate
research fellowship and to the University of Pennsylvania Physics Department, where some of
this work was done. C.A.S.Y. thanks Nicolas Cramp, Niall MacKay, Thomas Quella, and Andrei Babichenko for useful discussions, and gratefully acknowledges the financial support of
PPARC.
Appendix A. Example of graded coset superspaces
There are two classical families of simple Lie superalgebras for which the Killing form vanishes identically [17]: A(r 1|r 1) = psl(r|r) and D(r + 1, r) = osp(2r + 2|2r). Consider the
former. In its defining representation, sl(r|r) consists of matrices (even super-matrices) of the
block form8

A X
M=
,
(A.1)
Y B
where A, B are r r matrices with bosonic entries and X, Y are r r matrices with fermionic
entries with the condition that the supertrace vanishes:
str M := tr A tr B = 0.
(A.2)
The set of such matrices forms a Lie algebra under the matrix commutator, but this algebra is not
simple because the element

1rr
(A.3)
1rr
is central and so generates a one-dimensional ideal. One reaches psl(r|r) by quotienting out by
this ideal, so we can regard psl(r|r) as the set of matrices M as above but with tr A = tr B = 0,
and the Lie product as the matrix commutator composed with the obvious projection back onto
this subspace.
Finding a Z2n -grading of g = psl(r|r) is equivalent to finding an automorphism of order
2n [27,28].9 For, given such an automorphism, the grade k subspace is the eigenspace with
eigenvalue eik/n , and conversely given such a grading one can define an automorphism by
specifing its action on the subspaces of definite grade. We thus seek automorphisms of psl(r|r)
8 More precisely, this is the form of a general linear combination of the superalgebra generators with coefficients of
appropriate Grassmann grade.
9 That is, a map : g g such that [ M, L] = [M, L], 2n = 1 and k = 1 for all k < 2n.
120
of even order such that the even eigenspaces of are bosonic and the odd eigenspaces
are fermionic. That is to say: we want the eigenspace with eigenvalue epi/n to be block on1
diagonal, and the eigenspace with eigenvalue e(p 2 )i/n to be block off-diagonal, for every
p {1, 2, . . . , n}. One class of such automorphisms is as follows. Let
: A N 1 AN
(A.4)
be an (inner) automorphism of sl(r) of order n (so N n = 1 and hence, up to a similarity transformation [26] and an irrelevant overall phase,
1
aa
N =
2i
n
1bb
..
.
e
2i(n1)
n
(A.5)
1cc
where a + b + + c = r). Consider the order 2n automorphism of psl(r|r) defined by

: M g 1 Mg

with g =
i
n
e N
(A.6)
This sends
A N 1 AN,
X e
i/n
B N 1 BN,
XN,
Y e
i/n
(A.7)
N
Y N;
(A.8)
and it is clear that the even eigenspace of with eigenvalue e2ip/n consists of supermatrices
of the form

A
,
(A.9)
B
where A and B are both eigenvectors of with eigenvalue e2ip/n , while the odd eigenspace
of with eigenvalue e2ip/n ei/n consists of supermatrices of the form

(A.10)
where X and Y are eigenvectors of with eigenvalues e2ip/n and e2i(p+1)/n , respectively.
(If one specializes to n = 2 the resulting spaces G/H are PSL(a + b|a + b)/(SL(a) SL(b)
GL(1))2 . When a = b = 1 this is PSL(2|2)/(GL(1))2 and one real form is PSU(1, 1|2)/(U (1)2 ),
whose bosonic geometry is AdS2 S 2 [14]. Another possible class of automorphisms involve
the outer automorphism

A X
Y B
A
X
Y
B

(A.11)
of psl(n|n), where A is the transpose of A. The automorphism of PSL(4|4) which defines

AdS5 S 5 is of this type.)
121
References
[1] N. Read, H. Saleur, Exact spectra of conformal supersymmetric nonlinear sigma models in two dimensions, Nucl.
Phys. B 613 (2001) 409, hep-th/0106124;
S. Guruswamy, A. LeClair, A.W.W. Ludwig, gl(N |N ) super-current algebras for disordered Dirac fermions in two
dimensions, Nucl. Phys. B 583 (2000) 475, cond-mat/9909143.
[2] M.R. Zirnbauer, Conformal field theory of the integer quantum Hall plateau transition, hep-th/9905054, and references therein;
H.A. Weidenmuller, M.R. Zirnbauer, Instanton approximation to the graded nonlinear sigma model for the integer
quantum Hall effect, Nucl. Phys. B 305 (1988) 339;
K.B. Efetov, Supersymmetry and theory of disordered metals, Adv. Phys. 32 (1983) 53.
[3] V. Schomerus, H. Saleur, The GL(1|1) WZW model: From supergeometry to logarithmic CFT, hep-th/0510032,
and references therein.
[4] 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;
E. Witten, Anti-de Sitter space and holography, Adv. Theor. Math. Phys. 2 (1998) 253, hep-th/9802150;
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.
[5] P.A. Grassi, G. Policastro, Super-ChernSimons theory as superstring theory, hep-th/0412272.
[6] R.R. Metsaev, A.A. Tseytlin, Type IIB superstring action in AdS(5) S(5) background, Nucl. Phys. B 533 (1998)
109, hep-th/9805028.
[7] R. Roiban, W. Siegel, Superstrings on AdS(5) S(5) supertwistor space, JHEP 0011 (2000) 024, hep-th/0010104.
[8] M. Bershadsky, S. Zhukov, A. Vaintrob, PSL(n|n) sigma model as a conformal field theory, Nucl. Phys. B 559
(1999) 205, hep-th/9902180.
[9] N. Berkovits, C. Vafa, E. Witten, Conformal field theory of AdS background with RamondRamond flux,
JHEP 9903 (1999) 018, hep-th/9902098.
[10] J. de Boer, S.L. Shatashvili, Two-dimensional conformal field theories on AdS(2d + 1) backgrounds, JHEP 9906
(1999) 013, hep-th/9905032.
[11] G. Gotz, T. Quella, V. Schomerus, Tensor products of psl(2|2) representations, hep-th/0506072.
[12] T.L. Curtright, C.K. Zachos, Geometry, topology and supersymmetry in nonlinear models, Phys. Rev. Lett. 53
(1984) 1799;
E. Braaten, T.L. Curtright, C.K. Zachos, Torsion and geometrostasis in nonlinear sigma models, Nucl. Phys. B 260
(1985) 630;
S. Mukhi, Finiteness of nonlinear sigma models with parallelizing torsion, Phys. Lett. B 162 (1985) 345.
[13] E. Witten, Non-Abelian bosonization in two dimensions, Commun. Math. Phys. 92 (1984) 455.
[14] N. Berkovits, M. Bershadsky, T. Hauer, S. Zhukov, B. Zwiebach, Superstring theory on AdS(2) S(2) as a coset
supermanifold, Nucl. Phys. B 567 (2000) 61, hep-th/9907200.
[15] I. Bena, J. Polchinski, R. Roiban, Hidden symmetries of the AdS(5) S(5) superstring, Phys. Rev. D 69 (2004)
046002, hep-th/0305116.
[16] C.A.S. Young, Flat currents, Zm gradings and coset space actions, Phys. Lett. B 632 (2006) 559, hep-th/0503008.
[17] L. Frappat, A. Sciarrino, P. Sorba, Dictionary on Lie Algebras and Superalgebras, Elsevier, ISBN 0-12-265340-8,
2000.
[18] S.V. Ketov, Quantum Nonlinear Sigma Models: From Quantum Field Theory to Supersymmetry, Conformal Field
Theory, Black Holes and Strings, Springer-Verlag, ISBN 3540674616, 2000, and references therein.
[19] S. Sethi, Supermanifolds, rigid manifolds and mirror symmetry, Nucl. Phys. B 430 (1994) 31, hep-th/9404186.
[20] A. Schwarz, Sigma models having supermanifolds as target spaces, Lett. Math. Phys. 38 (1996) 91, hep-th/9506070.
[21] M. Aganagic, C. Vafa, Mirror symmetry and supermanifolds, hep-th/0403192.
[22] C.G. Zhou, On Ricci flat supermanifolds, JHEP 0502 (2005) 004, hep-th/0410047.
[23] M. Rocek, N. Wadhwa, On CalabiYau supermanifolds, hep-th/0408188;
M. Rocek, N. Wadhwa, On CalabiYau supermanifolds. II, hep-th/0410081.
[24] U. Lindstrom, M. Rocek, R. von Unge, Ricci-flat supertwistor spaces, hep-th/0509211.
[25] N. Mann, J. Polchinski, Bethe ansatz for a quantum supercoset sigma model, Phys. Rev. D 72 (2005) 086002,
hep-th/0508232.
[26] V.G. Kac, Infinite-Dimensional Lie Algebras, Cambridge Univ. Press, 1990, Chapter 8: Twisted affine algebras and
finite order automorphisms.
122
[27] P. Goddard, D.I. Olive, KacMoody and Virasoro algebras in relation to quantum physics, Int. J. Mod. Phys. A 1
(1986) 303.
[28] V.V. Serganova, Automorphisms of simple Lie superalgebras, Math. USSR, Izv. 24 (1985) 539551, Izv. Akad.
Nauk SSSR, Ser. Mat. 48 (3) (1984) 585589 (in Russian).
[29] A.M. Polyakov, Supermagnets and sigma models, hep-th/0512310.
[30] A.M. Polyakov, Conformal fixed points of unidentified gauge theories, Mod. Phys. Lett. A 19 (2004) 1649, hepth/0405106.
Noncritical osp(1|2, R) M-theory matrix model

with an arbitrary time-dependent cosmological constant
Jeong-Hyuck Park
Max-Planck-Institut fr Physik, Fhringer Ring 6, D-80805 Mnchen, Germany
Received 16 January 2006; received in revised form 3 March 2006; accepted 7 March 2006
Abstract
Dimensional reduction of the D = 2 minimal super-YangMills to the D = 1 matrix quantum mechanics is shown to double the number of dynamical supersymmetries, from N = 1 to N = 2. We analyze the
most general supersymmetric deformations of the latter, in order to construct the noncritical 3D M-theory
matrix model on generic supersymmetric backgrounds. It amounts to adding quadratic and linear potentials with arbitrary time-dependent coefficients, namely, a cosmological constant, (t), and an electric
flux background, (t), respectively. The resulting matrix model enjoys, irrespective of (t) and (t), two
dynamical supersymmetries which further reveal three hidden so(1, 2) symmetries. All together they form
the supersymmetry algebra, osp(1|2, R). Each so(1, 2) multiplet in the Hilbert space visualizes a dynamics
constrained on either Euclidean or Minkowskian dS2 /AdS2 space, depending on its Casimir. In particular, all the unitary multiplets have the Euclidean dS2 /AdS2 geometry. We conjecture that the matrix model
provides holographic duals to the 2D superstring theories on various backgrounds having the spacetime
signature Minkowskian if (t) > 0, or Euclidean if (t) < 0.
PACS: 11.25.Yb; 11.25.Pm
Keywords: M-theory; AdS/CFT; Supersymmetry
1. Introduction
String or M-theory dress all the known supersymmetric gauge theories with the insightful
geometrical pictures by the notion of holography or AdS/CFT correspondence [1,2]. In particular, the symmetry group of a gauge theory is identified as the isometry of the corresponding
E-mail address: park@mppmu.mpg.de (J.-H. Park).
doi:10.1016/j.nuclphysb.2006.03.004
124
J.-H. Park / Nuclear Physics B 745 [PM] (2006) 123141
higher-dimensional string/M-theory background. Conversely, different string theoriesbosonic

or supersymmetric, critical or noncriticalon various backgrounds are expected to have holographic dual gauge theories.
However, despite of some progress [310], the conformal dual description of the noncritical
2D superstring is a yet unresolved problem. In the present paper, we attempt to address the issue
from the M-theory point of view [1113]. The spacetime dimension two is singular in the sense
that the holographic dual of 2D superstring theory, whatever its concrete form is, should share
many common features with the corresponding noncritical M-theory matrix model.
As is well known, superstring lives in 2, 3, 4, 6 and 10 dimensions, while the supermembrane
exits in dimensions one higher, i.e. 3, 4, 5, 7 and 11, since only in those spacetime dimensions
the relevant Fierz identities hold. Although the pioneering work on super-p-branes [14] excludes
the possibility of the space-filling p-branes i.e. p-branes propagating in (p + 1)-dimensional
target spacetime, supermembrane does exit in three dimensions, since the Fierz identity for the
supermembrane works manifestly, from 012 = 1,

(d d ) d d = (d d )(d d ) = 0,
(1.1)
where d is a bosonic spinor. The matrix regularization [15,16] of the supermembrane prescribes
the replacement of the Poisson bracket appearing in the light cone gauged membrane action by
a matrix commutator. For 3D supermembrane action, it leads to a supersymmetric and gauged
version of a one matrix model, where the local gauge symmetry originates from the area preserving diffeomorphism for the Poisson bracket, and the appearance of only one matrix is due to the
light cone gauge, i.e. 3 2 = 1.
The resulting N N matrix model, at least for the flat 3D background, can be also obtained
by the dimensional reduction of the 2D minimal super-YangMills1 to D = 1, and it is supposed
to describe exactly the D0-brane dynamics of the discrete light cone momentum sector, p =
N/R, in M-theory compactified on a light-like circle, x x + 2R, as initially proposed by
Banks, Fischler, Shenker and Susskind for the critical M-theory [18,19]. As for the D0-branes,
the local gauge symmetry is required to reflect the identical nature of the N D-particles [20].
Also for the noncritical 2D superstring, almost by definition, its holographic dual should
be one-dimensional, supersymmetric and gauged theories. In the presence of RR electric field,
F , the low energy effective action of 2D string theory typically reads, neglecting the massless
tachyon and putting 1 [2123],

S2D = d2 x g e2 8 + R + 4()2 12 F 2 ,
(1.2)
where 12 F 2 plays the role of the negative cosmological constant, and the solutions are characterized by the AdS2 -like geometries.2 Indeed, switching off the dilaton completely we have the AdS2
solution, while turning on , one has static extremal black hole-like solutions [7,24,25]. In the
1 Recently all the minimal noncritical super-YangMills (except D = 3) have been identified in the noncritical superstring theories [17].
2 In two dimensions the geometries of AdS and dS coincide, and we will distinguish them by the sign of the cos2
2
mological constant. Also it is to be reminded that
k02 k12 k22 = R 2 > 0
:Euclidean AdS2 /dS2 (hyperboloid of two sheets),
k02 k12 k22 = R 2 < 0
:Minkowskian AdS2 /dS2 (hyperboloid of one sheet).
125
asymptotic region the latter becomes the usual linear dilaton vacuum, and in the near-horizon
region it approaches to AdS2 with the dilaton reaching the critical value, c = ln( 14 F ).
However, the effective action, (1.2), cannot be thoroughly trusted due to the corrections as
well as the tachyon tadpoles. Necessarily one has to work on the full sigma model (e.g. [26]) with
the difficulty of dealing with background fluxes. Hence to find the exact nontrivial superstring
background is not an easy task. And also for the M-theory, the matrix regularization of the
supermembrane action is not always straightforward for generic nontrivial backgrounds.
In this work, we take supersymmetry itself as the principal guideline to tackle the problem
of constructing the noncritical 3D M-theory matrix model on generic supersymmetric backgrounds. Namely after the dimensional reduction of the D = 2 super-YangMills to the D = 1
matrix quantum mechanics, we analyze all the possible deformations of the latter without breaking any supersymmetry. We show that the most general supersymmetric deformations simply
amount to adding quadratic and linear potentials with arbitrary time-dependent coefficients,
namely, a cosmological constant, (t), and an electric flux background, (t), respectively.
The latter couples to the u(1) sector or the center of mass only, while the su(N ) sector and
the u(1) sector are completely decoupled. Remarkably we find that, irrespective of (t) and
(t), the resulting matrix model always enjoys two dynamical supersymmetries, not just one as
in the 2D minimal super-YangMills. Namely after the dimensional reduction, the number of
supersymmetry is doubled, from N = 1 to N = 2. Furthermore, again for arbitrary (t) and
(t), these two supersymmetries reveal three hidden nontrivial bosonic symmetries. All together
the five symmetries form the super-Lie algebra, osp(1|2, R), where the even part corresponds
to so(1, 2) i.e. the isometry of the Euclidean or Minkowskian dS 2 /AdS 2 . We introduce a projection map from the phase space to a three-dimensional so(1, 2) hyperspace associated with
the bosonic symmetries. The dynamics therein is always constrained on a two-dimensional rigid
surface, Euclidean dS2 /AdS2 or Minkowskian dS2 /AdS2 , depending on the sign of the so(1, 2)
Casimir for each multiplet in the Hilbert space. The richness of the matrix model comes from the
arbitrariness of the time-dependent coefficients, (t), (t), and the vast amount of supermultiplets existing in the Hilbert space each of which has its own two-dimensional geometries.
The organization of the present paper is as follows. In Section 2, we analyze the most general
supersymmetric deformations of the matrix model having the 2D super-YangMills origin. We
discuss its symmetries, Hamiltonian dynamics and the BPS configurations. We also comment
on the relation to the matrix cosmology. Section 3 is devoted to the detailed analysis on the underlying supersymmetry algebra, osp(1|2, R), both from the kinematical and dynamical point of
view. In particular, we show that all the unitary multiplets correspond to the Euclidean dS2 /AdS2
geometry, rather than the Minkowskian one. The last section, Section 4 contains our conjecture
that the matrix model with different choices of (t) and (t) may provide holographic duals to
various 2D superstring or superconformal theories.
2. Noncritical osp(1|2, R) M-theory matrix model
2.1. Derivation of the matrix model and SUSY enhancement
In two-dimensional Minkowskian spacetime the fermion satisfies the MajoranaWeyl condition, resulting in only one component real spinor. After the dimensional reduction to D = 1, the
2D super-YangMills leads to the following supersymmetric matrix model, which can be also
obtained by the matrix regularization of the 3D supermembrane action in the light cone gauge,

L = tr 12 Dt XDt X + i 12 Dt + X ,
(2.1)
126
where X, are respectively bosonic or fermionic N N Hermitian matrices. With a gauge

potential, A0 = A0 , the covariant time derivative reads, in our convention,
Dt = t i[A0 , ].
(2.2)
Bosons, X, A0 , have the mass dimension 1, while the fermion, , has the mass dimension 32 , so
that the Lagrangian has the mass dimension, 4.
The supersymmetry transformation, YM , descending from the 2D super-YangMills theory
is, with a constant supersymmetry parameter, ,
YM A0 = YM X = i,
YM = Dt X.
(2.3)
Now we look for the generalizations of the above Lagrangian as well as the supersymmetry
transformations. First of all, we note from

tr i 12 Dt + X = tr i 12 t + (X A0 ) ,
(2.4)
that in order to cancel the cubic terms of in any possible supersymmetry variation which will
transform the bosons, (X A0 ) to the fermion, it is inevitable to impose3
A0 = X.
(2.5)
Hence, introducing a time-dependent function, f (t), we let the generalized supersymmetry transformation be

= f (t)Dt X + ,
A0 = X = if (t),
(2.6)
where is a bosonic quantity having the mass dimension 2, and its explicit form is to be determined shortly. After some straightforward manipulation, we obtain

L = tr i Dt (fX + ) fX + i[X, ] + t K,
(2.7)
where the total derivative term is given by

K = tr Dt XX i 12 .
(2.8)
Of course, the simplest case where f (t) = 1 and = 0 reduces to the supersymmetry of the
original 2D super-YangMills, (2.3). For the generic cases, we are obliged to set
= fX 1,
and obtain the following supersymmetry invariance,

)X = t K.
L + tr 12 (f/f )X 2 + (/f
(2.9)
(2.10)
This essentially leads to a novel supersymmetric matrix model with two arbitrary time-dependent
functions, (t), f (t), as spelled out in Eq. (2.12).
For given functions, (t) and (t), there exit two sets of solutions given by f (t), (t) to
satisfy the following second order differential equations,4
= f/f ,
= /f
.
(2.11)
3 Essentially this rigidity corresponds to the Fierz identity, tr(

[ , ]) = 0, relevant to the existence of the
minimal super-YangMills in 2, 3, 4, 6, 10 dimensions.
4 The integral constant for (t) corresponds to the kinematical supersymmetry.
127
Thus, surprisingly, there are two dynamical supersymmetries in the matrix model, even for the
case,5 = = 0. This will further reveal three nontrivial bosonic symmetries as we discuss in
the next subsection.
Rather than taking (2.9) one might attempt to close the supersymmetry invariance by adding
other terms to L. However, since there exits only one component spinor, there cannot be any
mass term for the fermion6 as tr() = 0. Thus, as long as we restrict on the nonderivative
corrections, the above generalization is the most generic one.
2.2. Noncritical osp(1|2, R) M-theory matrix model: Final form
With an arbitrary time-dependent cosmological constant, (t), having the mass dimension
two and an arbitrary time-dependent electric flux background, (t), having the mass dimension
three, the generic form of the noncritical 3D M-theory matrix model reads

Losp(1|2,R) = tr 12 (Dt X)2 + i 12 Dt + X + 12 (t)X 2 + (t)X .
(2.12)
The Lagrangian corresponds to the most general supersymmetric deformations of the N = 2
matrix quantum mechanics of the 2D super-YangMills origin. The matrix model is to describe
the noncritical 3D supermembrane in a controllable manner through the matrix regularization,
and our claim is further that it also provides holographic duals to 2D superstring theories, as
discussed in the last section.
The matrix model is equipped with the standard local gauge symmetry,
X gXg 1 ,
gg 1 ,
A0 gA0 g 1 it gg 1 ,
g U(N ),
(2.13)
and enjoys two dynamical supersymmetries,

A0 = X = if (t) ,

= f (t)Dt X f (t)X (t)1 ,
(2.14)
where + , , are two real supersymmetry parameters, and f (t), (t) are the two different
solutions of the second order differential equations,
f (t) = f (t)(t),
t
(t) :=
dt (t )f (t ).
(2.15)
t0
5 This kind of supersymmetry enhancement after the dimensional reduction can be also noticed elsewhere. For example, in the earlier works [27,28], we derived the effective worldvolume gauge theories for the longitudinal D5- and
D2-branes on the maximally supersymmetric 11D pp-wave background. After the dimensional reductions to D = 1,
both of them lead to a matrix quantum mechanics which is equivalent to the BMN M-theory matrix model [29] up
to field redefinitions. The formers have only four dynamical supersymmetries, while the BMN model has 32 supersymmetries, 16 dynamical and 16 kinematical. The physical reason for the enhancement is that the D-branes which
4 .
the higher-dimensional gauge theories describe preserve only the fraction of the full M-theory supersymmetries, 32
The same reasoning also holds for the present osp(1|2, R) M-theory matrix model having three supersymmetries, two
dynamical and one kinematical. As we see shortly, all the BPS states preserve only one supersymmetry, breaking the
other two. Hence, the minimal 2D super-YangMills can be interpreted as the worldvolume action of the longitudinal
M2-brane which preserves only one supersymmetry. However, it remains somewhat mysterious that the total number of
supersymmetries is three, a rather unusual odd number.
6 This is a special feature only present in the matrix quantum mechanics of the 2D super-YangMills origin. In fact, in
the higher-dimensional cases one needs to add the fermion mass term for the supersymmetry invariance as in the BMN
matrix model [29] or [30].
128
The above two dynamical supersymmetries further reveal three hidden nontrivial bosonic symmetries,7 which we denote by ++ , , {+,} , in order to indicate the anti-commutator origin
from the two supersymmetries,
++ A0 = ++ X = f+ (f+ Dt X f+ X + 1),
A0 = X = f (f Dt X f X 1),
++ = 0,
= 0,
{+,} A0 = {+,} X = 2f+ f Dt X (f+ f + f f+ )X (f+ + f + )1,

{+,} = 0.
Since
dt (f+ f
(2.16)
f f+ ) = 0, if we set a nonzero constant,
c := f+ (t)f (t) f (t)f+ (t) = 0,
(2.17)
and define

1 2
1 2
1
J1 := i
J2 := i f+ f t , (2.18)
f+ + f2 t ,
f+ f2 t ,
2c
2c
c
then the isometry of AdS2 or the global conformal algebra, sp(2, R) so(1, 2) sl(2, R) follows
in the standard form,
J0 := i
[J0 , J1 ] = +iJ2 ,
[J1 , J2 ] = iJ0 ,
[J2 , J0 ] = +iJ1 .
(2.19)
Now the above three bosonic symmetries (2.16) can be identified as the conformal transformations of X having the conformal weight 12 ,
X = tDt X 12 (t t)X + 1,
(2.20)
where the conformal diffeomorphism, t, is generated by J0 , J1 , J2 above and the inhomogeneous term, , satisfies
= 0.
+ 32 t t + t
(2.21)
Furthermore, as we show in the next section, all the five symmetries form the osp(1|2, R) superalgebra, where the three bosonic symmetries correspond to its even part, so(1, 2).
Apart from the dynamical supersymmetries, there is the usual kinematical supersymmetry,
corresponding to the integral constant of (t),
A0 = X = 0,
= 1.
(2.22)
In parallel to this, there exit two extra bosonic symmetries given by8
X = f+ (t)1,
= A0 = 0 or
X = f (t)1,
= A0 = 0,
(2.23)
which can be also identified as the special case of (2.20), (2.21), with the choice, t 0.
Note that the su(N ) sector and the u(1) sector are completely decoupled, while (t) couples
to the u(1) sector or the center of mass only.
7 It is worth to note that the three bosonic symmetries are still valid in the bosonic matrix model obtained after putting
0,

Lso(1,2) = tr 12 (Dt X)2 + 12 (t)X 2 + (t)X .
8 We thank Gordon Semenoff for pointing out the extra bosonic symmetries.
129
2.3. Hamiltonian and the Dirac bracket

The EulerLagrangian equations read
Dt Dt X (t)X (t)1 = 0,
Dt + i[X, ] = 0,
[Dt X, X] + i = 0
:Gauss constraint.
(2.24)
Up to the Gauss constraint or the first-class constraint, the cubic vertex term vanishes,
tr(X)
0, so that the Hamiltonian becomes simply a harmonic oscillator type, being free
of the fermion,

H = tr 12 P 2 12 (t)X 2 (t)X , P := Dt X.
(2.25)
In fact, for any gauge invariant object,

F = tr F (X, P , , t) ,
(2.26)
the EulerLagrangian equations, (2.24), imply

F
F
dF
= tr P
+ (t)X + (t)1
F i tr [X, F ] +
= [F , H }D.B. +
.
dt
X
P
t
t
(2.27)
The Dirac bracket for our matrix model is given by, after taking care of the primary second-class
constraint for the fermion [31,32],
[F , G}D.B. =
F G
F G
F G
+ i(1)#F
,
X a b P b a P a b X b a
a b b a
(2.28)
where a, b are the N N matrix indices, while #F = 0 or 1, depending on the spin statistics of
F , i.e. 0 for the boson and 1 for the fermion.
Due to the five symmetries of the action, there are five conserved quantities given by the
Noether charges. For their explicit expressions we refer (3.29) and (3.30).
Since the su(N ) and u(1) sectors are completely decoupled, it is convenient to introduce the
trace over either the su(N ) or u(1) sector only,

trsu(N ) F (X, P , ) := tr F X N 1 tr(X), P N 1 tr(P ), N 1 tr() ,

tru(1) F (X, P , ) := tr F N 1 tr(X), N 1 tr(P ), N 1 tr() .
(2.29)
Accordingly the quadratic Hamiltonian decomposes into the two distinct pieces,
H = Hsu(N ) + Hu(1) ,
(2.30)
where
Hsu(N ) = trsu(N )
1
2P

12 (t)X 2 ,
Hu(1) = tru(1)
1
2P

12 (t)X 2 (t)X .
(2.31)
2.4. BPS states and the cosmological principle

From the supersymmetry transformations of the fermion, the BPS equations are
f (t)Dt X = f (t)X + (t)1,
(2.32)
130
so that the generic BPS configurations decompose into the traceless and u(1) parts,
X(t) = f+ (t)X + h+ (t)1
or X(t) = f (t)X + h (t)1,
(2.33)
where X is an arbitrary traceless constant matrix, and h (t) are the solutions of the first order differential equation, f h = f h + , corresponding to the center of mass position,
N 1 tr X(t) = h (t).
Since f+ (t) = f (t), the BPS state preserves only one supersymmetry out of three supersymmetries (two dynamical and one kinematical). It is interesting to note that for arbitrary
time-dependent functions, say f+ (t) and h+ (t), there exits a supersymmetric matrix model
where X(t) = f+ (t)X + h+ (t)1 corresponds to a BPS state, and furthermore there exits always
its twin BPS state given by + .
Utilizing the gauge symmetry (2.13), one can diagonalize X in order to show the positions of
the N D-particles in the BPS sector,

X(t) = diag x1 (t), x2 (t), . . . , xN (t) = f (t) diag(x1 , x2 , . . . , xN ) + h (t)1.
(2.34)
A remarkable fact is that all D-particles have precisely the same relative movement, same position, same velocity, same acceleration, etc. up to the constant scaling factors which entirely
depend on their initial positions or so-called the co-moving coordinates. This matches precisely
with the homothetic ansatz adopted in the cosmology literature in order to incorporate the cosmological principle [33,34]. In fact, the second order differential equation, f = f , (2.15)
can be identified as the Raychaudhuris equation in cosmology, where is indeed the timedependent cosmological constant. Also, in the matrix approach to the cosmology [3436], it
is natural to associate as the nonrelativistic cosmological constant, and associate > 0 and
< 0 with the de Sitter and anti-de Sitter space respectively accounting the repulsive and attractive potential. Thus, although the geometries of dS2 and AdS2 coincide, we distinguish them
by the sign of , throughout the paper.
3. osp(1|2, R) superalgebra
After the standard quantization, [F , G}D.B. i[F , G}, the present osp(1|2, R) matrix
model leads to the following Heisenberg Clifford algebra,

a
a
b , c d = a d c b .
X b , P c d = i, a d c b ,
(3.1)
In Section 3.1, utilizing the above algebra alone, especially from the su(N ) sector only, we construct explicitly the generators of the osp(1|2, R) superalgebra.9 The number of odd generators
is two, and this is consistent with the fact that there are two dynamical supersymmetries in the
matrix model, rather than one. Section 3.2 is devoted to the analysis on the unitary irreducible
representations of the superalgebra, osp(1|2, R). Further analysis on the superalgebra from the
dynamical point of view is given in Section 3.3.
3.1. osp(1|2, R) superalgebrakinematical point of view
There are five real generators in osp(1|2, R) which we take as
QP := trsu(N ) (P ),
QX := trsu(N ) (X),
9 For the construction of other various algebras, see [37].
(3.2)
131
and

K0 := 12 trsu(N ) P 2 + X 2 ,

K1 := 12 trsu(N ) P 2 X 2 ,
K2 := 12 trsu(N ) (XP + P X).
(3.3)
Alternatively we may construct the generators out of the full u(N ) matrices, including the u(1)
part and using the ordinary trace, i.e. trsu(N ) tr. All the results below will remain identical,
but the resulting so(1, 2) Casimir will not be a conserved time-independent operator, which is
not what we want. In order to account for the kinematical supersymmetry, (2.22), we may also
include one more odd generator, Qkinematical = tr(). However this commutes with any generator
above in the su(N ) sector.
All the super-commutator relations10 of the osp(1|2, R) superalgebra follow simply from the
Heisenberg Clifford algebra, (3.1),
Q2P = 12 (K0 + K1 ),
Q2X = 12 (K0 K1 ),
{QP , QX } = K2 ,
[K0 , QP ] = +iQX ,
[K0 , QX ] = iQP ,
[K1 , QP ] = iQX ,
[K1 , QX ] = iQP ,
[K2 , QP ] = +iQP ,
[K2 , QX ] = iQX ,
[K1 , K2 ] = 2iK0 ,
[K0 , K1 ] = +2iK2 ,
[K2 , K0 ] = +2iK1 .
(3.4)
The Casimir of the osp(1|2, R) superalgebra reads

Cosp(1|2,R) = Cso(1,2) + i[QP , QX ],
[Cosp(1|2,R) , anything] = 0,
(3.5)
where the so(1, 2) Casimir is given by

2
Cso(1,2) = K02 K12 K22 = 12 trsu(N ) P 2 , trsu(N ) X 2 14 trsu(N ) (XP + P X)

= 2 Q2X , Q2P {QX , QP }2 .
(3.6)
The root structure of the osp(1|2, R) superalgebra can be identified by complexifying the
generators as11
Q+ := QP + iQX ,
Q := QP iQX = Q+ ,
K+ := K1 + iK2 ,
K := K1 iK2 = K+
.
(3.7)
The Cartan subalgebra has only one element, K0 , and all others are either raising, Q+ , K+ , or
lowering, Q , K , operators to satisfy
Q+2 = K+ ,
Q2 = K ,
{Q , Q+ } = 2K0 ,
[K0 , Q+ ] = +Q+ ,
[K+ , Q+ ] = 0,
[K , Q+ ] = +2Q ,
[K0 , Q ] = Q ,
[K+ , Q ] = 2Q+ ,
[K , Q ] = 0,
[K0 , K+ ] = +2K+ ,
[K0 , K ] = 2K ,
[K , K+ ] = 4K0 .
(3.8)
10 Although the osp(1, |2, R) super-commutator relations above are direct consequences of the Heisenberg Clifford
algebra, the way to express the generators in terms of X, P and is not unique. In fact, so(1, 2) algebra was identified
thirty years ago [38] using a nonpolynomial basis in the conformal matrix model having the inverse square potential,
and based on the observation, Strominger proposed that the conformal matrix model is dual to 2D type 0A string theory
on AdS2 [23] (see also [39,40]).
11 For further analysis by us on the root structures of super-Lie algebras, see [41,42].
132
In the Cartan basis, the Casimir operators, (3.5), (3.6), read

Cosp(1|2,R) = Cso(1,2) + 12 [Q , Q+ ],
Cso(1,2) = K02 12 {K , K+ }.
(3.9)
The osp(1|2, R) superalgebra can be represented by (2 + 1) (2 + 1) real supermatrices, M,

satisfying

0 1 0
T
M J + J M = 0, J = 1 0 0 ,
(3.10)
0 0 1
so that its generic form reads, with the even and odd real Grassmann entries, x, ,

x2 x+ 1
M = x x2 2 .
2
1
0
(3.11)
Note that the 2 2 bosonic part corresponds to sp(2, R) so(1, 2) sl(2, R), as it corresponds
to x , where x = x0 x1 , and is the so(1, 2) gamma matrix,

0 1
0 1
1 0
,
1 :=
,
2 :=
,
0 :=
1 0
1 0
0 1
+ = 2 , = diag( + +).
(3.12)
In fact, with the notion of a (1 + 2)-dimensional two component Majorana spinor and its
charge conjugate,

QP
Q1
:= QT 0 = (QX QP ),
=
,
Q
Q=
(3.13)
Q2
QX
the osp(1|2, R) superalgebra, (3.4), can be rewritten in a compact form,
= K ,
{Q, Q}
[K , Q] = i Q,
[K , K ] = 2i K ,
(3.14)
= [QX , QP ], that the

where is the usual three form with 012 1. Note also, from QQ
osp(1|2, R) Casimir operator, Cosp(1|2,R) , (3.5) is indeed manifestly SL(2, R) invariant.
In a similar fashion to above, one can also equip the bosonic operators with the SL(2, R)
covariant structure,

P
V1
V.
:=
,
K = 12 V
V=
(3.15)
V2
X
3.2. Unitary irreducible representations of osp(1|2, R)
In order to analyze the unitary irreducible representations or unitary supermultiplets of the
osp(1|2, R) superalgebra spanned by the five real generators, (3.2), (3.3), one needs to take K0
as the good quantum number operator to diagonalize it. Different choice of the good quantum
number operator, e.g. K2 , is not compatible with the unitarity, as it would lead to the raising and
lowering operators with the pure imaginary unit, such as [K2 , (K1 K0 )] = 2i(K1 K0 ).
Any osp(1|2, R) supermultiplet decomposes into so(1, 2) multiplets. We first review briefly
the general properties of the latter or the unitary irreducible representations of so(1, 2).12
12 For further analysis see e.g. [43].
133
From (3.9) and the commutator relations, we get

Cso(1,2) + 1 + K K = (K0 1)2 .
(3.16)
From the Hermitian conjugacy property, K+ = K

, the third term on the left-hand side, K K ,
is positive semi-definite, while the possible minimum value of the right-hand side for the states
in a unitary irreducible representation may lie

0 min (K0 1)2 1,
(3.17)
if the raising or lowering operators act nontrivially ever. But, this is impossible when Cso(1,2) > 0.
In this case, the unitary representation is infinite-dimensional and characterized by the existence
of either the lowest weight state obeying
K |l, l = 0,
K0 |l, l = l|l, l ,
Cso(1,2) = l(l 2) > 0,
l > 2,
(3.18)
or the highest weight state obeying

K+ |h, h = 0,
K0 |h, h = h|h, h ,
Cso(1,2) = h(h + 2) > 0,
h < 2.
(3.19)
When Cso(1,2) = 0, there exists only one trivial state, |0, 0 , satisfying
K |0, 0 = 0,
K0 |0, 0 = 0.
(3.20)
When 1 Cso(1,2) < 0, the representation is called the continuous principal series. It is infinite-dimensional, and the lowest or highest weight state may or may not
exit. If there is
a lowest or highest weight state, then its good quantum number is +1 Cso(1,2) + 1 or

1 Cso(1,2) + 1, respectively but not simultaneously. When Cso(1,2) < 1, there must be
neither lowest nor highest weight state, and the representation is called the continuous supplementary series.
As for the present osp(1|2, R) matrix model, K0 is positive definite for the unitary multiplets
as

K0 = 12 trsu(N ) P 2 + X 2 = trsu(N ) A A + 12 N 2 1 12 N 2 1 ,

a
A b , Ac d = a d c b .
A := 1 (P iX),
(3.21)
2
Thus there exits always a lowest weight state in any so(1, 2) multiplet, and from (3.16), the
so(1, 2) Casimir is bounded below13

Cso(1,2) 14 N 2 1 N 2 5
for N 3,
Cso(1,2) > 0 or
Cso(1,2) = 34
for N = 2.
(3.22)
Now as for the osp(1|2, R) unitary supermultiplet, we first note that the odd roots, Q ,
shift the good quantum number by one unit, half of what K do. Hence the odd roots move
one so(1, 2) multiplet to another inside a osp(1|2, R) supermultiplet, but at most once due to
13 In fact, from (3.6), expressing C
so(1,2) in terms of the odd generator, the trace of Cso(1,2) also formally shows the
positiveness,

Tr Cso(1,2) Tr [QX , QP ]2 0.
The subtlety is due to the infinite sum over the infinite-dimensional so(1, 2) multiplet.
134
Q2 = K . Similar to (3.16), we also have

Cosp(1|2,R) + K K Q Q = K0 (K0 1).
(3.23)
After all, utilizing all the facts above, we conclude that any unitary irreducible representation
of the osp(1|2, R) superalgebra satisfying the positiveness, (3.21), is infinite-dimensional and
characterized by the existence of the super-lowest weight state obeying

Q |ls , ls = 0,
K0 |ls , ls = ls |ls , ls ,
Cosp(1|2,R) = ls (ls 1), ls 12 N 2 1 .
(3.24)
Furthermore, the osp(1|2, R) unitary supermultiplet always decomposes into two so(1, 2) multiplets whose lowest weight states are given by
|ls , ls
1
and |ls + 1, ls + 1 = Q+ |ls , ls .
2ls
(3.25)
3.3. osp(1|2, R) superalgebradynamical point of view

The Noether charges corresponding to the two dynamical supersymmetries, (2.14), decompose into the su(N ) and u(1) parts,

tr(i ) = if (t) Q
su(N ) + Qu(1) ,

Q
su(N ) := trsu(N ) P g (t)X ,

Q
(3.26)
u(1) := tru(1) P g X ( /f )1 ,
where we put
g (t) :=
f (t)
,
f (t)
2
= (t).
g + g
(3.27)
Because the Hamiltonian as well as the above two supercharges in the su(N ) sector can be
expressed in terms of the previous kinematical basis, QX , QP , K0 , K1 , K2 , (3.2), (3.3), the
underlying supersymmetry algebra must correspond to osp(1|2, R), no matter what the dynamics
is. However, the use of the above supercharges, Qsu(N ) , will not lead to simple expressions for
the superalgebra. For example, from the conservation of the Noether charge and Eq. (2.27), the
commutator relation between the Hamiltonian and the supercharge reads in a less economic
manner,

g +
Qsu(N ) Q
H, Q
(3.28)
su(N ) = ig Qsu(N ) + i g g
su(N ) .
+
Henceforth, in order to analyze the underlying osp(1|2, R) superalgebra in a simple fashion

but still to keep track of the dynamical properties, we slightly modify the basis of the odd generators and keep the Hamiltonian explicitly as a so(1, 2) generator. Note that the change of basis
requires the time-dependent coefficients due to (t), as Q
su(N ) = QP g (t)QX . Hence, only
with specific time-dependent coefficients we can write down the time-independent conserved
quantities, as one can expect from (2.27). All together there are five conserved true Noether
charges corresponding to the five symmetries, (2.14), (2.16). Namely we have the two fermionic
conserved Noether charges for the two dynamical supersymmetries,
f Q
su(N ) = f QP f QX ,
(3.29)
135
and three bosonic conserved Noether charges for the so(1, 2) symmetries (2.16),
2 1 2

2

1
2
2
f Q
su(N ) = 2 f + f K0 + 2 f f K1 f f K2 ,
f+ Qsu(N ) , f Qsu(N ) = (f+ f + f+ f )K0 + (f+ f f+ f )K1

(f+ f + f f+ )K2 .
(3.30)
Apart from the above five Noether charges, both of the osp(1|2, R) and so(1, 2) Casimir operators, Cosp(1|2,R) (3.5) and Cso(1,2) (3.6), are also conserved time-independent quantities, since
they do not include any explicit time dependency and they commute with the Hamiltonian, for
sure.
3.3.1. osp(1|2, R) superalgebra when (t) = 0
In a similar fashion to the standard harmonic oscillator analysis, we first set a pair of operators,14
P (t)X
A (t) :=
(3.31)
,
2
and define a pair of even generators in osp(1|2, R) by

J (t) := trsu(N ) A (t)2 ,
(3.32)
as well as a pair of odd generators,

Q (t) := trsu(N ) A (t) .
(3.33)
Note that Q coincide with the actual supercharges, Q

su(N ) , (3.26), provided that (t) is constant.
The Hamiltonian for the su(N ) sector is then

Hsu(N ) = 12 trsu(N ) A+ (t)A (t) + A (t)A+ (t)

= trsu(N ) A (t)A (t) 12 i (t) N 2 1 ,
(3.34)
and from the quantization relation,

A (t)a b , A+ (t)c d = i (t) a d c b ,
(3.35)
we obtain such as

[Hsu(N ) , A ] = i (t)A ,

[J , A + ] = 2i (t)A ,

[J+ , A ] = +2i (t)A + ,
(3.36)
where we set
A := A N 1 tr(A )1.
(3.37)
Now, the osp(1|2, R) superalgebra reads in terms of Q , J , Hsu(N ) ,

Q+2 = 12 J+ ,
Q2 = 12 J ,
[J , Q+ ] = 2i (t)Q ,
[J+ , Q ] = +2 (t)Q+ ,
[J , J+ ] = 4i (t)Hsu(N ) , [Hsu(N ) , J ] = 2i (t)J ,
{Q+ , Q } = Hsu(N ) ,
[Hsu(N ) , Q ] = i (t)Q ,
[J , Q ] = 0.
14 In fact, when (t) is constant, A correspond to the generators of W algebra [44].
(3.38)
136
Especially, the so(1, 2) Casimir operator, (3.6), can be reexpressed as

1 1
2
{J+ , J } Hsu(N
Cso(1,2) = K02 K12 K22 =
) .
2
(t)
(3.39)
From J = J for > 0 and J+ = J for < 0, there exits a SO(1, 2) rotation which transforms Hsu(N ) to K1 if > 0 or K0 if < 0.
3.3.2. osp(1|2, R) superalgebra when (t) = 0
If = 0, A+ coincides with A , and the above super-commutator relations (3.38) do not
faithfully represent the super-Lie algebra, osp(1|2, R). In order to do so, one needs to define
the generators differently. When = 0 we have f+ = 1, f = t , and the corresponding two
supercharges, (3.26), in the su(N ) sector are

1
Q+
(3.40)
Q
=0 = trsu(N ) [P ],
=0 = trsu(N ) P t X ,
while the Hamiltonian is given by

Hsu(N ) = 12 trsu(N ) P 2 = 12 (K0 + K1 ).
(3.41)
Rather than Q
=0 , we adopt the kinematical odd generators, (3.2),

QP = trsu(N ) (P ) = Q+
QX = trsu(N ) (X) = t Q+
=0 ,
=0 Q=0 ,
(3.42)
and write the osp(1|2, R) superalgebra in terms of the real basis,

Q2P = Hsu(N ) ,
Q2X = Vsu(N ) := 12 trsu(N ) X 2 , {QP , QX } = K2 ,
[Hsu(N ) , QX ] = iQP ,
[Hsu(N ) , QP ] = 0,
[Vsu(N ) , QX ] = 0,
[Vsu(N ) , QP ] = +iQX ,
[K2 , QX ] = iQX ,
[K2 , QP ] = +iQP ,
[Hsu(N ) , Vsu(N ) ] = iK2 ,
[K2 , Hsu(N ) ] = +2iHsu(N ) ,
[K2 , Vsu(N ) ] = 2iVsu(N ) .

(3.43)
In particular, the so(1, 2) Casimir operator, (3.6), reads

Cso(1,2) = K02 K12 K22 = 2{Hsu(N ) , Vsu(N ) } K22 .
(3.44)
4. Discussion and conclusion

We have derived a N = 2 supersymmetric matrix model, (2.12), with quadratic and linear potentials whose coefficients are arbitrary time-dependent cosmological constant, (t),
and electric flux background, (t). The matrix model corresponds to the most general supersymmetric deformations of the matrix quantum mechanics having the 2D super-YangMills
origin. We have shown that, for arbitrary (t) and (t), the matrix model enjoys two dynamical supersymmetries, Q1 , Q2 , and three bosonic symmetries, K0 , K1 , K2 , which amount to the
superalgebra, osp(1|2, R), (3.14),
= K ,
{Q, Q}
[K , Q] = i Q,
[K , K ] = 2i K .
(4.1)
If the matrix model had only one supersymmetry as in the 2D minimal super-YangMills, the
osp(1|2, R) structure would be absent.
The matrix model is to describe the noncritical 3D M-theory on generic supersymmetric
backgrounds in a controllable manner through the matrix regularization, and our claim is further
that, with the arbitrariness of (t) and (t), it also provides holographic duals to various twodimensional superstring theories, as we argue below.
137
4.1. Normalizable and nonnormalizable wave functions

At the quantum level, the wave function satisfies

tr 12 P 2 12 (t)X 2 (t)X (t) = i (t) .
(4.2)
t
As is well known, when (t) is negative, the potential is bounded below and the normalizable or
unitary wave functions have the discrete spectrum.
Namely, from (3.36), the lowering operator,
A lowers the eigenvalue of Hsu(N ) by the unit ||. However, since Hsu(N ) should be positive
definite for the unitary representations, there must be a ground state which is annihilated by A ,
A |0 = 0,
|0 = e 2
|| tr X 2
|P 0 .
(4.3)
All other excited states are then constructed by acting the raising operator, A + to the ground
state. Due to the Gauss constraint, one needs to restrict on the gauge singlets, which can be simply done by taking the trace of the u(N ) indices in all the possible ways [45]. The quantum
states in the Hilbert space then form the unitary irreducible representations of the superalgebra,
osp(1|2, R), and in particular, their so(1, 2)Casimir is positive definite for N 3, (3.22). The
energy spectrum is discretized by the unit |(t)|, and the zero point vacuum energy is, from
2
(3.34), 12 |(t)|(N 2 1). The vacuum has the degeneracy, 2[N /2] , due to the fermions. The
nonvanishing zero point energy refers to the existing two other bosonic charges in the superalgebra apart from the Hamiltonian.
On the other hand, when (t) is positive, the wave functions cannot be normalizable,
as the
raising and lowering operators shift the energy spectrum by the imaginary unit, i , while
Hsu(N ) should have real eigenvalues for the normalizable states. Physically, this amounts to the
fact that the matrix model describes the Fermi sea (see e.g. [36,46]).
Therefore, in order to have a unifying description for arbitrary (t), the full Hilbert space of
the M-theory matrix model should include not only the normalizable states but also the nonnormalizable states, allowing both the unitary and the nonunitary representations of osp(1|2, R).
The former is relevant only to the case (t) < 0. Without the concern about the normalizability,
the following Schrdinger equation has always solutions for arbitrary energy, E(t),

2 1
+ 2 (t)X 2 + (t)X (X, t) = E(t) (X, t).
tr 12
(4.4)
X
In particular, the momentum operator P = i X

is no longer necessarily real. Again the rasing
and lowering operators, A , generate new solutions with the shifted energy, E(t) i (t).
The reason to consider the phase space over the complex planes rather than the real lines is
manifest in the path integral formalism, since when (t) > 0, the local minima of the Hamiltonian are located on the genuine complex planes rather than the real lines so that one should
take P Hermitian and X anti-Hermitian, or vice versa.
All the BPS states are nonnormalizable or nonunitary: From its defining property,

QBPS |BPS = 0,
(4.5)
QBPS = tr (f P fX 1) = QBPS ,
and the positive definite property of Q2BPS = 12 tr(f P fX 1)2 for the unitary states,
if the BPS state were normalizable, it would mean f P |BPS = (fX + 1)|BPS so that
BPS|f XP |BPS = BPS|(fX 2 + X)|BPS = BPS|f P X|BPS . But this clearly contradicts
with the quantization, [X, P ] = 0.
138
4.2. Projection to the so(1, 2) hyperspace

We consider a projection map from the full phase space to the so(1, 2) hyperspace given by
the coordinates, K0 , K1 , K2 , (3.3). The induced dynamics therein is subject to

T := 12 (1 ), 12 (1 + ), 0 ,
K = i[Hsu(N ) , K ] = 2 T K ,
T K = Hsu(N ) ,
T T = .
(4.6)
Naturally, as seen from (3.12), the so(1, 2) hyperspace is equipped with the so(1, 2) metric,
= diag( + +). The so(1, 2) Casimir, Cso(1,2) , (3.6) highlights the geometrical picture,
Cso(1,2) = K02 K12 K22 .
(4.7)
Cso(1,2) is a conserved time-independent operator, since it does not include any explicit time
dependency and it commutes with the Hamiltonian, just like the osp(1|2, R) Casimir. Classically,
this can be also seen, from (4.6), as K K = 0. Therefore, we observe that for each so(1, 2)
multiplet in the Hilbert space, the corresponding so(1, 2) hyperspace dynamics is constrained
on a two-dimensional rigid surface such that
Euclidean dS2 /AdS2
if Cso(1,2) > 0,
Minkowskian dS2 /AdS2
if Cso(1,2) < 0,
Null cone
if Cso(1,2) = 0.
(4.8)
Surely the specific value of the Casimir for each multiplet is to be superselected just like any
boundary condition in quantum field theories. This also fits into the 3D M-theory picture, to
include or provide holographic dual descriptions to all the superstring theories. The richness of
the osp(1|2, R) M-theory matrix model originates from the arbitrariness of the cosmological
constant, (t) and the electric flux background, (t) as well as the vast amount of existing
so(1, 2) multiplets in the Hilbert space each of which has its own two-dimensional geometry.
However, if we restrict on the unitary irreducible representations, i.e. the normalizable sector
relevant to the case (t) < 0, we have the bound for the Casimir, (3.22),

Cso(1,2) 14 N 2 1 N 2 5 .
(4.9)
Thus, the corresponding geometry is always Euclidean dS2 /AdS2 if N 3. As for the nonnormalizable or nonunitary sector, the above bound does not hold.
The bound can be also understood classically as

2
Cso(1,2) = trsu(N ) P 2 trsu(N ) X 2 trsu(N ) (P X)

2
trsu(N ) (P X) 2
P .
= trsu(N ) P trsu(N ) X
(4.10)
trsu(N ) (P 2 )
This is positive semi-definite if both X and P are Hermitian, as is the case for the expectation
values of the unitary states. Otherwise, of course, not. Especially, when P is Hermitian and X
is anti-Hermitian or vice versa, as in the path integral formalism for > 0, Cso(1,2) is negative
semi-definite, implying the Minkowskian dS2 /AdS2 geometry. From (3.27), among the on-shell
configurations, only the BPS configurations saturate the bound, Cso(1,2) = 0.
From (3.39), (4.6), a dispersion relation follows

2

2
= 4 Hsu(N
K K = 4 Hsu(N
(4.11)
) K K
) + (t)Cso(1,2) = 2 J+ (t), J (t) ,
139
which shows that the mass is conserved if (t) is constant. Furthermore, we have the positive
semi-definite bound both for the unitary states,
For (t) = 0,
For (t) > 0,
For (t) < 0,
K K = 4H (t)2 0,

K K = 8 Q+ Q+ , Q Q 0,

K K = 2 J+ , J+ 0,
(4.12)
and also for the nonunitary states for which > 0 and P , iX being Hermitian (or antiHermitian),

2

2
K K = trsu(N ) P 2 + X 2 trsu(N ) (P X + XP ) 0.
(4.13)
The equality holds only for the trivial case, X = P = 0. Thus, the velocity vector, K is always space-like, which is natural for the Euclidean geometry of the unitary states. But in the
Minkowskian space, as for the nonunitary states with > 0, it implies the superluminar behavior, i.e. tachyon. As shown above, all the BPS configurations correspond to the null geometry,
and hence not tachyonic.
To summarize, the normalizable or the unitary sector in the Hilbert space relevant to the case
< 0 is characterized by the Euclidean dS2 /AdS2 geometry, while the nonnormalizable or the
nonunitary sector relevant to the case > 0 has the Minkowskian geometry. All the BPS states
always correspond to the null geometry, i.e. Cso(1,2) = 0. When (t) > 0 and Cso(1,2) < 0, i.e.
the Minkowskian de Sitter geometry, from (4.12), the particles in the so(1, 2) hyperspace are
tachyonic and cannot be supersymmetric.
4.3. Holographic dual to 2D superstring
Various matrix models with potentials having a single maximum have been proposed as dual
candidates of 2D string theories on AdS2 -type backgrounds with the rolling tachyon or the linear dilaton [9,21,23,4750]. The continuum or so-called the double scaling limit in the matrix
models zoom in on the maximum of the potential, effectively leaving a single upside down harmonic potential [5153], precisely the same feature as our osp(1|2, R) matrix model shares when
> 0. Furthermore, the Hermitian matrix itself is supposed to represent the non-Abelian open
string tachyon [47], and this is manifest in our dispersion relation, (4.12), for the case of > 0
and Cso(1,2) < 0. Thus, we conclude that when (t) is positive, the osp(1|2, R) M-theory matrix model provides holographic duals to the two-dimensional Minkowskian superstring theories.
The relevant sector in the matrix model Hilbert space is then the nonnormalizable or nonunitary
one satisfying Cso(1,2) < 0. From (4.12), the choice of the decreasing (t), like (t) = et/to ,
seems appropriate for the description of the tachyon condensation [60] or the D-brane decay [47].
Further investigation is to be required.
On the other hand, when = 0, for the constant positive the generic BPS configurations
(2.33) are given by the hyperbolic functions,

X(t) = cosh t X(0) + sinh t 1,

(4.14)
while for the constant negative they are the usual harmonic oscillators,

sin ||t 1.
X(t) = cos ||t X(0) +
||
(4.15)
140
The latter is also consistent with the Euclidean 2D superstring theory or the N = 2 superLiouville theory results [5459]. The classical shape of the so-called FZZT brane (falling Euclidean D0-brane), which is given as time-dependent boundary state, precisely matches with (4.15).
Thus, we expect that when (t) is negative, the osp(1|2, R) M-theory matrix model provides
holographic dual description of 2D Euclidean superstring theories or superconformal theories.
In particular, if is negative constant, it corresponds to the N = 2 super-Liouville theory, with
the relation to the Liouville background charge, QLiouville = 2||.
Acknowledgements
The author wishes to thank Xavier Bekaert, Nakwoo Kim, Nikita Nekrasov, Gordon Semenoff, Jan Troost, Satoshi Yamaguchi for valuable comments, and the organizers of RTN Corfu
Summer Institute 2005 for the enlightening workshop where the present paper was initiated.
The work was partially supported by the European Research Training Network contract 005104
ForcesUniverse.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
J.M. Maldacena, Adv. Theor. Math. Phys. 2 (1998) 231, Int. J. Theor. Phys. 38 (1999) 1113, hep-th/9711200.
O. Aharony, S.S. Gubser, J.M. Maldacena, H. Ooguri, Y. Oz, Phys. Rep. 323 (2000) 183, hep-th/9905111.
E. Marinari, G. Parisi, Phys. Lett. B 240 (1990) 375.
A. Jevicki, T. Yoneya, Nucl. Phys. B 411 (1994) 64, hep-th/9305109.
J. McGreevy, S. Murthy, H.L. Verlinde, JHEP 0404 (2004) 015, hep-th/0308105.
A. Dabholkar, Nucl. Phys. B 368 (1992) 293.
S. Gukov, T. Takayanagi, N. Toumbas, JHEP 0403 (2004) 017, hep-th/0312208.
T. Takayanagi, JHEP 0411 (2004) 030, hep-th/0408086.
H.L. Verlinde, Superstrings on AdS2 and superconformal matrix quantum mechanics, hep-th/0403024.
A. Jevicki, J.P. Rodrigues, Phys. Lett. B 268 (1991) 53;
J.P. Rodrigues, A.J. van Tonder, Int. J. Mod. Phys. A 8 (1993) 2517;
A.J. van Tonder, A continuum description of superCalogero models, hep-th/9204034.
P.K. Townsend, The M(atrix) model/AdS2 correspondence, hep-th/9903043.
P. Horava, C.A. Keeler, Noncritical M-theory in 2 + 1 dimensions as a nonrelativistic Fermi liquid, hep-th/0508024.
M. McGuigan, Noncritical M-theory: Three dimensions, hep-th/0408041.
A. Achucarro, J.M. Evans, P.K. Townsend, D.L. Wiltshire, Phys. Lett. B 198 (1987) 441.
B. de Wit, J. Hoppe, H. Nicolai, Nucl. Phys. B 305 (1988) 545.
J. Hoppe, Membranes and matrix models, hep-th/0206192.
S.K. Ashok, S. Murthy, J. Troost, D-branes in non-critical superstrings and minimal super-YangMills in various
dimensions, hep-th/0504079.
T. Banks, W. Fischler, S.H. Shenker, L. Susskind, Phys. Rev. D 55 (1997) 5112, hep-th/9610043.
L. Susskind, Another conjecture about M(atrix) theory, hep-th/9704080.
J.-H. Park, Phys. Lett. A 307 (2003) 183, hep-th/0203017.
M.R. Douglas, I.R. Klebanov, D. Kutasov, J. Maldacena, E. Martinec, N. Seiberg, A new hat for the c = 1 matrix
model, hep-th/0307195.
D.M. Thompson, Phys. Rev. D 70 (2004) 106001, hep-th/0312156.
A. Strominger, JHEP 0403 (2004) 066, hep-th/0312194.
T. Banks, M. OLoughlin, Phys. Rev. D 48 (1993) 698, hep-th/9212136.
N. Berkovits, S. Gukov, B.C. Vallilo, Nucl. Phys. B 614 (2001) 195, hep-th/0107140.
Y. Kazama, H. Suzuki, Nucl. Phys. B 321 (1989) 232.
S. Hyun, J.-H. Park, JHEP 0211 (2002) 001, hep-th/0209219.
S. Hyun, J.-H. Park, S.H. Yi, JHEP 0303 (2003) 004, hep-th/0301090.
D. Berenstein, J.M. Maldacena, H. Nastase, JHEP 0204 (2002) 013, hep-th/0202021.
N. Kim, J.-H. Park, Noncritical M-theory matrix models in 3, 4, 5, 7 dimensions, in preparation.
P.A.M. Dirac, Lectures on Quantum Mechanics, Yeshiva University, New York, 1964.
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]
[57]
[58]
[59]
[60]
141
M. Henneaux, C. Teitelboim, Quantization of Gauge Systems, Princeton Univ. Press, Princeton, 1992.
E. Alvarez, P. Meessen, Phys. Lett. B 426 (1998) 282, hep-th/9712136.
D.Z. Freedman, G.W. Gibbons, M. Schnabl, AIP Conf. Proc. 743 (2005) 286, hep-th/0411119.
G.W. Gibbons, C.E. Patricot, Class. Quantum Grav. 20 (2003) 5225, hep-th/0308200.
S.R. Das, J.L. Davis, F. Larsen, P. Mukhopadhyay, Phys. Rev. D 70 (2004) 044017, hep-th/0403275.
M. Gunaydin, C. Saclioglu, Commun. Math. Phys. 87 (1982) 159;
I. Bars, M. Gunaydin, Commun. Math. Phys. 91 (1983) 31;
M. Gunaydin, S. Hyun, J. Math. Phys. 29 (1988) 2367;
M. Gunaydin, R.J. Scalise, J. Math. Phys. 32 (1991) 599;
M. Gunaydin, J. Math. Phys. 29 (1988) 1275;
S. Fernando, M. Gunaydin, S. Hyun, Oscillator construction of spectra of PP-wave superalgebras in eleven dimensions, hep-th/0411281.
V. de Alfaro, S. Fubini, G. Furlan, Nuovo Cimento A 34 (1976) 569.
P.M. Ho, JHEP 0405 (2004) 008, hep-th/0401167.
O. Aharony, A. Patir, The conformal limit of the 0A matrix model and string theory on AdS2 , hep-th/0509221.
N. Kim, J.-H. Park, Phys. Rev. D 66 (2002) 106007, hep-th/0207061.
S. Lee, J.-H. Park, JHEP 0406 (2004) 038, hep-th/0404051.
L.J. Dixon, M.E. Peskin, J. Lykken, Nucl. Phys. B 325 (1989) 329;
M.S. Plyushchay, J. Math. Phys. 34 (1993) 3954;
H. Nicolai, D.I. Olive, Lett. Math. Phys. 58 (2001) 141, hep-th/0107146;
J.M. Maldacena, H. Ooguri, J. Math. Phys. 42 (2001) 2929;
B. Morariu, A.P. Polychronakos, Quantum mechanics on noncommutative Riemann surfaces, hep-th/0201070.
J. Avan, A. Jevicki, Phys. Lett. B 266 (1991) 35;
J. Avan, A. Jevicki, Phys. Lett. B 272 (1991) 17;
S.R. Das, A. Dhar, G. Mandal, S.R. Wadia, Int. J. Mod. Phys. A 7 (1992) 5165;
S.R. Das, A. Dhar, G. Mandal, S.R. Wadia, Mod. Phys. Lett. A 7 (1992) 71;
E. Witten, Nucl. Phys. B 373 (1992) 187.
J.-H. Park, Class. Quantum Grav. 19 (2002) L11, hep-th/0108145.
J.L. Karczmarek, A. Strominger, JHEP 0404 (2004) 055, hep-th/0309138.
J. McGreevy, H.L. Verlinde, JHEP 0312 (2003) 054, hep-th/0304224.
E.J. Martinec, The annular report on non-critical string theory, hep-th/0305148.
I.R. Klebanov, J. Maldacena, N. Seiberg, JHEP 0307 (2003) 045, hep-th/0305159.
T. Takayanagi, N. Toumbas, JHEP 0307 (2003) 064, hep-th/0307083.
D.J. Gross, N. Miljkovic, Phys. Lett. B 238 (1990) 217.
E. Brezin, V.A. Kazakov, A.B. Zamolodchikov, Nucl. Phys. B 338 (1990) 673.
P.H. Ginsparg, J. Zinn-Justin, Phys. Lett. B 240 (1990) 333.
T. Eguchi, Y. Sugawara, Modular bootstrap for boundary N = 2 Liouville theory, JHEP 0401 (2004) 025, hepth/0311141.
C. Ahn, M. Stanishkov, M. Yamamoto, One-point functions of N = 2 super-Liouville theory with boundary, Nucl.
Phys. B 683 (2004) 177, hep-th/0311169.
D. Kutasov, D-brane dynamics near NS5-branes, hep-th/0405058.
C. Ahn, M. Stanishkov, M. Yamamoto, ZZ-branes of N = 2 super-Liouville theory, JHEP 0407 (2004) 057, hepth/0405274.
Y. Nakayama, Y. Sugawara, H. Takayanagi, JHEP 0407 (2004) 020, hep-th/0406173.
J.M. Lapan, W. Li, Falling D0-branes in 2D superstring theory, hep-th/0501054.
A. Sen, JHEP 0405 (2004) 076, hep-th/0402157;
A. Sen, Int. J. Mod. Phys. A 18 (2003) 4869, hep-th/0209122;
A. Sen, Mod. Phys. Lett. A 17 (2002) 1797, hep-th/0204143;
A. Sen, JHEP 0207 (2002) 065, hep-th/0203265;
A. Sen, JHEP 0204 (2002) 048, hep-th/0203211.
Tunneling in a quantum field theory on a compact

one-dimensional space
Jrgen Baacke , Nina Kevlishvili
Institut fr Physik, Universitt Dortmund, D-44221 Dortmund, Germany
Abstract
We compute tunneling in a quantum field theory in 1 + 1 dimensions for a field potential U () of the
asymmetric double well type. The system is localized initially in the false vacuum. We consider the case
of a compact space (S1 ) and study global tunneling. The process is studied in real-time simulations. The
computation is based on the time-dependent HartreeFock variational principle with a product ansatz for the
wave functions of the various normal modes. While the wave functions of the nonzero momentum modes
are treated within the Gaussian approximation, the wave function of the zero mode that tunnels between the
two wells evolves according to a standard Schrdinger equation. We find that in general tunneling occurs in
a resonant way. If the nonzero momentum modes are excited efficiently, they react back onto the zero mode
causing an effective dissipation. In some region of parameter space this back-reaction causes the tunneling
being replaced by a sliding of the wave function.
PACS: 03.70.+k; 03.65.-w; 11.10.-z; 98.80.Qc
Keywords: Quantum field theory; Tunneling; Hartree approximation; Real time approach; Numerical simulations
1. Introduction
Tunneling is one of the important elementary processes that may happen in a quantum mechanical system. There is a vast literature on the subject and the WKB approach is discussed in
textbooks on quantum mechanics. In quantum field theory tunneling has been mostly discussed
* Corresponding author. Tel: +49 231 7553573; fax: +49 231 7555025.
E-mail addresses: baacke@physik.uni-dortmund.de (J. Baacke), nina.kevlishvili@het.physik.uni-dortmund.de

(N. Kevlishvili).
doi:10.1016/j.nuclphysb.2006.03.031
J. Baacke, N. Kevlishvili / Nuclear Physics B 745 [PM] (2006) 142164
143
as a local process proceeding via bounce or bubble solutions [14] for the simple reason that
in an infinite space a global transition of a mean field through a barrier would have an infinite
action. However, if space is finite such global transitions are possible.
Tunneling with compact spaces can be of interest in cosmology. Such transitions in de Sitter
space have been evoked in order to describe the quantum creation of a universe [5,6]. In a less
specific way one may think of transitions within a finite volume of a chaotic initial state of the
universe [7]. Tunneling in a finite volume may also have some relation to local tunneling in an
infinite volume as an alternative to bounce transitions. This point has been elaborated in Ref. [8].
Tunneling with compact space also occurs in models for the quantum creation of a universe
[9,10], which in the minisuperspace space approximation is a closed 3 sphere and whose radius
a is a simple quantum mechanical degree of freedom. This effect has been widely discussed,
and most authors use the WKB approximation [1116]. Our analysis is not directly related to
quantum cosmology because the rle of time is quite different there and here. However, there are
similarities concerning the rle of particle production and their back-reaction which still are an
open issue there [13,1719].
Tunneling is of course a process that occurs in real time and the continuation to imaginary time
is a technical tool which is widely used and whose application is based in a controlled way on the
eikonal expansion. Its application to quantum field theory becomes cumbersome whenever one
is interested in the physical evolution of the system during and after the tunneling process, the
matching of the wave functions that is well understood in quantum mechanics now encompasses
an infinite system of modes and becomes rather involved. It therefore seems to be of interest to
look for a description of the system entirely in real time.
Such an approach was taken for the case of quantum mechanics in Ref. [20]. These authors
have studied in detail the case of the double-well potential. They find that tunneling is characterized by the occurrence of resonances between degenerate (approximate) levels in the spectrum
of the separate wells.
For quantum field theory the real time approach to quantum tunneling, in the same way as
it is used here, has recently been discussed by Hirota [8]. His analysis, as ours, is based on the
time-dependent HartreeFock approximation. He uses analytic approximations for the zero mode
wave functions. While quantum back reaction and renormalization are mentioned, they are not
considered in detail and the author does not present any numerical results.
We do not solve the field theoretical problem exactly but use two approximations: for the
wave function of the entire system we make a product ansatz, i.e., a Hartree approximation.
The wave functions of the fluctuation modes are taken to be Gaussian; for the zero mode wave
function, however, we numerically solve the Schrdinger equation. In this respect we differ from
the widely used out-of-equilibrium simulations [2125], where a Gaussian wave function, shifted
by a classical field (t), is used for the zero mode as well. By using the Hartree approximation we
go beyond the one-loop back-reaction by taking into account the back-reaction of the nonzero
momentum modes onto the zero mode and onto themselves. The latter back-reaction avoids a
possible instability of the system whenever the effective mass of the nonzero momentum modes
gets imaginary.
The plan of the paper is as follows: in Section 2 we present the model, the decomposition into
a discrete set of degrees of freedom, and the Hamiltonian in the Schrdinger representation; in
Section 3 we formulate the time-dependent Hartree approximation [2628] using our ansatz for
the product wave function; renormalization is discussed in Section 4 and the initial conditions in
Section 5; numerical results for some parameter sets are presented and discussed in Section 6;
we end with conclusions and an outlook in Section 7.
144
2. The model
We consider a scalar quantum field theory in 1 + 1 dimensions with a compact space, i.e., on
a manifold R S1 . As ultimately we want to use our approach on R S3 we only introduce
those self-couplings that would lead to a renormalizable theory in 3 + 1 dimensions. The action
is given by

2a
1 2 1 2
dx dt
U () ,

S=
2
2 x
(2.1)
with
1
U () = m2 2 3 + 4
(2.2)
2
8
and periodic boundary conditions in x. With positive values of the couplings the potential has
the asymmetric double well form with a minimum at = = 0 and a second one at

3
8
1+ 1 ,
+ =
(2.3)
9
with = m2 /42 . The parameter is restricted to the range 0 < < 1. The field is dimensionless, and and have the dimension mass2 . We introduce the expansion into normal
modes

n (t)eikn x ,
(x, t) = 0 (t) +
(2.4)
n=
with
n
kn = .
a
Then the action takes the form

1 2 1
1 2
S = 2a dt
0 +
n n
n n
2
2
2 n n
+
(2.5)
n=0
nn
n
nn

n n n nn n ,
8
(2.6)
nn n
with
n2 = m2 +
n2
.
a2
(2.7)
As the field is real we have the condition n = n . We therefore introduce, for n = 0 real
fields via
1
n = (n1 in2 ), n > 0.
(2.8)
2
In the Schrdinger representation the Hamiltonian is given by
2

2 1

1
2 2
2 2
+ m 0 +
H = 2a 2 2
+
n nj
2
2
8 a 02
nj
n>0,j
n>0,j

nn

n n nn +
n n n nn n .
8
145
(2.9)
nn n
In order to get rid of factors 2a we introduce the rescaling
k = k / 2a,
(2.10)
so that the Hamiltonian takes the form

2 1

1 2
2 2
2 2
H =
+
m
+
n nj
0
2
2 02
2
nj
n>0,j
n>0,j

n n nn +
nn

n n n nn n ,
8
(2.11)
nn n
with = / 2a and = /2a. Note that in the cubic and quartic parts we still have retained
the complex fields
1
n = (n1 in2 ).
2
(2.12)
3. The time dependent HartreeFock approximation

We now make a variational ansatz for the wave function and impose a variational principle,
which is known under the name of time-dependent HartreeFock approach. The ansatz for the
wave function is

(0 , n , t) = 0 (0 , t)
(3.1)
n (nj , t).
n>0
Furthermore we will restrict the ansatz for the modes with n = 0 to a Gaussian wave function

sn (t)
ein (t)
1
1
i
n (nj , t) =
(3.2)
exp
2 ,
2 2n2 (t)
n (t) nj
[2n2 (t)]1/4
while we do not further specify 0 . The time dependent variational principle [26,29], now imposes the condition

dnj (0 , nj , t)(it H ) (0 , nj , t) = 0.
d0
(3.3)
n>0,j
We will find later that the dynamics is independent of j , which therefore has already been suppressed in the index for the wave functions. Before we write down the resulting Schrdinger
equations for the various degrees of freedom we rewrite the Hamiltonian in an appropriate way.
The part of H which exclusively contains the zero mode is given by
1 2
+ U (0 ),
2 02
(3.4)
1

U (0 ) = m2 02 03 + 04 .
2
8
(3.5)
H00 =
with
146
The parts bilinear in the quantum modes lead to the Hamiltonian

1 2
1 2 2
H0n =
+
2
2 nj
2 n nj
(3.6)
n>0,j
Due to the Gaussian ansatz for the n = 0 modes the expectation values of any odd powers of
n = 0 fluctuations vanish. So if we consider the interaction between zero and nonzero modes we
need to retain only terms bilinear in the n = 0 modes. We then have for the interaction between
zero and nonzero modes

2
(3.7)
n n nn
= 30
n n = 30
nj
I 0n
n,n
and

n=0

= 602
n
n
n
nn n
I 0n
n,n ,n
n>0,j

n=0
n n = 602
2
nj
.
(3.8)
n>0,j
So with the scaled couplings and the interaction Hamiltonian coupling zero and nonzero
modes is given, in the approximation considered here, by

3 2 2

nj .
HI 0n = 3 0 +
(3.9)
4 0
n>0,j
The part of the Hamiltonian cubic in the nonzero modes has a vanishing expectation value. The
quartic term has the expectation value

2
(3.10)
nj
n2 j ,
n n n nn n = 3
n,n ,n
nj,n j
where we have used the fact that the wave functions for n = 0 are Gaussian. This part yields the
Hamiltonian for the mutual and self interaction of the nonzero modes. It is given by

3 2 2 3 2
HI nn =
(3.11)
nj
= F .
8
2
nj
Applying now the variation principle we get for 0 the Schrdinger equation

1 2
3 2

F 0 (0 , t).
it 0 (0 , t) =
+ U (0 ) + 6 0 +
2 02
2 0
Here we have introduced the fluctuation integral

1 2
F=
nj ,
2
(3.12)
(3.13)
n>0,j
which will be specified more explicitly later. For the modes with n = 0 we find

2
1 2
1 2
it n (n , t) =
+ n + W (t) nj n (n , t).
2
2 nj
2
(3.14)
147
Here

W (t) = U (0 ) m2 + 3 F,
with

U (0 ) m2 =

3 2
d0 0 (0 , t) 6 0 +
0 0 (0 , t).
2
(3.15)
(3.16)
We also introduce n2 (t) = n2 + W (t). The Schrdinger equation for the quantum mode wave
function is seen to be independent of the index j . Therefore the previous summations over j just
lead to a degeneracy factor 2, so that for the mode functions the index j can be suppressed.
With the Gaussian ansatz (3.2) the Schrdinger equation for the n = 0 modes implies
n (t) = sn (t),
(3.17)
sn (t) = n2 (t)n (t) +

n (t) =
1
,
4n3 (t)
1
.
4n2 (t)
(3.18)
(3.19)
The first two of these equations can be related to mode functions fn (t) satisfying
fn (t) + n2 (t)fn (t) = 0,
(3.20)
as they arise from a KleinGordon equation for a field with the effective mass m2 + W (t). We
have, with n0 = n (0),
|fn (t)|
,
n (t) =
2n0

d
1
sn (t) =
fn (t),
2n0 dt
while the wave function is given by

1/4

i fn (t) 2
2n0
in (t)
exp
.
n (n , t) = e
2 fn (t) n
2|fn (t)|2
(3.21)
(3.22)
(3.23)
In deriving these relation one uses repeatedly the Wronskian relation

fn (t)fn (t) fn (t)fn (t) = 2in0 ,
(3.24)
which corresponds to an initial condition

fn (0) = 1,
fn (0) = in0 .
For this wave function the expectation value of n2 is given by

2
|fn (t)|2
dn n (n , t) n2 =
,
2n0
(3.25)
(3.26)
so that the fluctuation integral becomes

F (t) =
|fn (t)|2
n>0
2n0
(3.27)
148
The system of equations we have presented here is consistent with an conserved energy
E = H00 + H0n + HI 0n + HI nn ,
with

1 2
H00 =
+ U (0 ) 0 (0 , t),
2 02

1 1

fn (t)2 + 1 2 fn (t)2 ,
H0n = 2
2n0 2
2 n
n>0

HI 0n = U (0 ) m2 F,
3
HI nn = F 2 .
2
d0 0 (0 , t)
(3.28)
(3.29)
(3.30)
(3.31)
(3.32)
2 in H , so that H becomes the free Hamiltonian for

It is convenient to replace n2 by n0
0n
0n
the initial fluctuation wave functions. Then the term HI 0n receives an additional contribution
W (0)F .
A particle number for the n = 0 modes may be defined in various ways. A Fock space is
defined by the mode decomposition (2.4). To the operators nj and their conjugate momenta
nj = /nj we can associate creation and annihilation operators referring to an oscillator of
frequency n0 via

n0
1
cnj =
(3.33)
nj +
,
2
n0 nj

n0
1
cnj
(3.34)
=
nj
2
n0 nj
and a particle number

Nnj

= cnj
cnj =

1 2 2
1
1 2
+ n0 nj ;
2
2
2 nj
2
n0
(3.35)
computing the expectation value with the wave function n (nj ), Eq. (3.23), one finds
Nnj =

2 1
1 2
2
.
f
(t)
+
(t)
f
n
n
n0
2
2
4n0
(3.36)
In the present context the quantum

excitations do not describe free particles; we merely use the

total particle number N = nj Nnj as an indicator for the excitations of the n > 0 oscillators.
4. Renormalization
The fluctuation integral and the energy density, as introduced in the previous section, are divergent quantities. In 1 + 1 dimensions the renormalization is in principle rather straightforward.
However, here it has to be done for a nonequilibrium system with discrete momenta, and for
the Hartree approximation, i.e., for a resummed perturbation series. We here limit ourselves to
present the main steps needed to derive the formulas that enter the numerical codes.
Though we treat the system in a nonperturbative approach, the divergences are related exactly
to those of standard perturbation theory. The mode functions can be expanded perturbatively with
149
respect to the potential V (t) which contains the couplings and ; so such an expansion is at the
same time an expansion with respect to these couplings.
As discussed in previous publications on nonequilibrium dynamics [24] we write the mode
functions as

fn (t) = ein0 t 1 + hn (t)
(4.1)
and convert the differential equation for the fn (t) into an integral equation for the functions hn (t)
as
t

i

hn (t) =
(4.2)
dt e2in0 (tt ) 1 V (t ) 1 + hn (t ) .
2n0
0
Here V (t) = W (t) W (0), see Eq. (3.15).

We further have

fn (t)2 = 1 + 2 Re hn (t) + hn (t)2 .
One easily finds that Re hn (t) behaves as

V (t)
1
Re hn (t) 2 + O
,
3
4n0
n0
for large n. For the fluctuation integral one finds

1
1
V (t)
F=
1 2 +O
,
3
2n0
2n0
n0
n
(4.3)
(4.4)
(4.5)
so that we can separate it into a divergent and a subtracted finite part as

F (0) =
Fsub =

n=0

n=0
1
,
2n0
(4.6)

2
1
2 Re hn (t) + hn (t) .
2n0
(4.7)
The divergent part has to be regularized and to be separated into the standard renormalization
part and a finite contribution. The sum over discrete momenta here and below can be done in the
same way as the Matsubara sums in finite temperature quantum field theory, or using the Plana
formula [30]. One finds

1
dk
1
2
(4.8)
= a
1+
,
2n0
(2)20
exp(2a0 ) 1 a0
n=1
with 0 = k 2 + m20 . The first term in the bracket is obviously the divergent part we would
obtain for infinite space and goes into the renormalization of the various couplings. The second
term in the bracket arises from the periodic boundary conditions. The third part arises from the
fact that F contains the nonzero modes only and no subtraction has been applied to the zero
mode. The first term can be regularized in a standard way [24]; we have

dk
1
i
d 2 k
(4.9)
=
=
L0 ,
(2)20 reg
(2)2 k02 k 2 m20 + io 4
150
with L0 = 2/ + ln(4/m20 ) using dimensional regularization. m20 depends on the initial conditions, and the renormalization condition should not; we therefore write L0 = L +
ln m2 /m20 with L = 2/ + ln(4/m2 ) and include the second term into the finite part. So
we define the remaining finite part as
(0)
Ffin

= 2a
dk
1
a m2
1
+ ln 2
(2)20 exp(2a0 ) 1 4m0 4 m0
(4.10)
and the finite part of the fluctuation integral as

(0)
Ffin = Ffin + Fsub .
(4.11)
We similarly decompose the fluctuation energy as

1 1
2 1 2
2

H0n = 2
fn (t) + n fn (t)
2n0 2
2
n>0
1
= Efl, sub m0 + a
2

dk 0 2a
with
Efl, sub = 2

n>0
k 2 dk
,
0 (exp(2a0 ) 1)

2
1
1 2 1 2
2

fn (t) + n fn (t) 0 .
2n0 2
2
(4.12)
(4.13)
Integrating by part the last (thermal) integral in Eq. (4.12) can be recast into the form of a free
energy

2a
0
k 2 dk
=
0 (exp(2a0 ) 1)

dk
ln 1 exp(2a0 ) .
2
(4.14)
The integral over 0 is given, in dimensional regularization, as

a 2
m2
a dk 0
= m0 L + ln 2 + 1
4
m0
reg
(4.15)
and we define

dk
a 2
m2
Efl,fin = Efl,sub + m0 ln 2 + 1 +
ln 1 exp(2a0 ) .
4
2
m0
(4.16)
The renormalization is done in analogy to the case of nonequilibrium dynamics in [31], following the scheme of Ref. [32], by adding a counterterm

CM2 = C m2 + U (0 ) m2 + 3 Ffin .
(4.17)
Here one has used already the finite gap equation

M2 = m2 + U (0 ) m2 + 3 Ffin ,
(4.18)
151
which determines the effective mass of the fluctuations. Choosing

a
C = L ,
4
and defining the initial mass m0 by the initial gap equation

(0)
m20 = m2 + U (0 ) m2 + 3 Ffin ,
(4.19)
(4.20)
the gap equation (4.18) is satisfied for all t . It can be written as

M2 (t) = m20 + V (t),
(4.21)
with the previously introduced potential

V (t) = U (0 ) m20 + 3 Ffin (t),
(4.22)
which now is well-defined.

For the energy we obtain

1
3
2
E = H00 (0 ) + Efl,fin m0 + U m20 Ffin + Fsub
2
2

dk
a
m2
+ m20 ln 2 + 1 +
ln 1 exp(2a0 ) .
4
2
m0
(4.23)
The last two terms are, for a fixed radius a, independent of time; we leave them out when presenting the energy conservation. The term m0 /2 marks the absence of the zero mode in the sum
over fluctuations. If m0 = m this term exactly cancels, at t = 0, the expectation value H00 (0 ).
Likewise, the term 1/4m0 in the initial gap equation (4.20) is cancelled, for m0 = m, by the
term U (0 ) m2 . Indeed if a is sufficiently large, am > 1, corresponding to a temperature Teff /m < 1/2 , the thermal integrals become very small; then the solution m0 of the gap
equation indeed is close to m and E 0.
5. Initial conditions
As already discussed in Section 3 the system is started for the nonzero modes with
fn (t) en0 t ,
which is equivalent to initial wave functions

n0 1/4 n0 n2 /2
e
.
(n , 0) =

Here n0 = m20 + n2 /a 2 and m0 is determined by the gap equation (4.20).
For the zero mode we likewise start with a Gaussian wave function
1/4
m
2
(0 , 0) =
emn /2 .
(5.1)
(5.2)
(5.3)
This would be the ground state wave function of an oscillator with frequency = m, i.e., the
ground state wave function for a potential with = = 0.
152
The choice of these initial conditions is of course quite arbitrary. Here we try to start in an
approximate ground state for the system in the left well of the double well potential. We find
the total energy of the system to be close to zero, as one would expect for such an approximate
ground state (it is understood that the zero point energies of the nonzero modes are not included).
A different choice would in general lead to a higher total energy and make tunneling easier, or
make the transition to a large part an over the barrier transition.
6. Numerical results
6.1. Numerical details
We have implemented the formulas of the previous sections numerically. This is essentially
straightforward. The Schrdinger equation for the zero mode becomes a system of first order
differential equations for the values 0 (0,k ) where 0,k are the equidistant discrete values of
the variable 0 , and the coupling arises from the discretized second derivative
(0,k+1 ) + (0,k1 ) 2(0,k )
2
0 (0,k )
,
2
(0 )2
0
(6.1)
where 0 is the step width for the 0,k . The Schrdinger equation for the n = 0 modes is
converted into the second order mode equations, which are coupled to the zero mode and the other
nonzero modes by V (t). This discretization leads to instabilities in the time evolution unless the
time intervals are chosen of the order 02 . We used 4000 grid points for the zero mode 0 , which
typically extends over a region of 10 < 0 < 50. So 0 is of the order 102 . The RungeKutta
time step was chosen t = 0.00002. This choice makes the codes very slow, much slower than
those of nonequilibrium dynamics in the Gaussian approximation. As we do not use the Gaussian
approximation for the zero mode wave function, we have to compute, at each RungeKutta step,
not only the sums over the quantum modes n , but also the averages of various observables in the
ground state wave function. The excited quantum modes were taken into account up to n = 200.
A simulation of the time evolution until t 100 takes a few hours on a standard PC.
The accuracy of the computations was checked by computing Wronskians and energy conservation. The relative accuracy obtained was better than six significant digits. The energy
conservation also checks the correct implementation of the basic equations. We show a typical example, for = = 1 (set IV below) and a = 1.2. The total energy is E = (2.39 0.001) 105
throughout the total time interval, the single components take values up to 35.
6.2. Parametrization and parameter sets
The parameter space, encompassing , and a is quite large, so before presenting numerical
results we try to get some qualitative insight into the physics to be expected in certain ranges of
their numerical values. The parameter m sets the overall scales and is put equal to m = 1.
While in Section 2, we introduced a scaling of the fields, (2.10), which was suitable for the
canonical formalism on the basis of the Hamiltonian (2.11), the discussion of the results is more
transparent if we introduce the parametrization widely used in bounce computations [4,
3335]. One introduces the rescaling X = mx and = m2 /2.

Then for infinite space the
classical action takes the form

m4
1
2
2
Scl = 2 d X (X ) + U () = Scl (),

(6.2)
2
4
153
with = m4 /42 , = /m2 and

1
1
= 2 3 + 4 .
U ()
2
2
8
(6.3)
While multiplies the classical action the one-loop effective action is a function of only. So
for large the system essentially becomes classical and the quantum effects only lead to small
corrections while for small the quantum effects become large. The parameter determines the
shape of the potential: for 1 we get a symmetric double-well potential, for small the right
hand well becomes much deeper than the left hand one, the barrier between the wells becomes
shallow. For bubble nucleation 1 is the thin-wall limit, where the bubble size becomes very
large. The case > 1 is of no interest here, as the minimum at = 0 then becomes the global
minimum.
We will in the following consider parameter sets with fixed and , i.e., fixed and , and
study the dependence on a.
For finite space extension the semiclassical tunneling action of a spatially homogeneous
bounce is given by
0

Sbounce = 2a d U (),
(6.4)
where
0 =
2
(1 1 )
(6.5)
is the zero at the right-hand side of the potential barrier. For fixed , the tunneling via a homogeneous bounce should shut off with increasing a. The tunneling will then occur via local
bounces, as considered recently in the Hartree approximation in [36,37].
The transition rate obtained from the homogeneous bounce, Eq. (6.4), will only have a qualitative relation to the observed tunneling phenomena which will be characterized by the occurrence
of resonances. For quantum mechanical tunneling this was already observed in Ref. [20]. In order to estimate the separation of the resonances we consider the approximate spectra of the left
and right wells, treating them as separate oscillators. The resonances can then be thought of as
arising from the degeneracy of levels in the left and in the right wells. From the Hamiltonian and
the potential written in canonical variables, Eqs. (3.4) and (3.5), we see that the energies of the
left hand oscillator are Enl = (n + 1/2)m. The right hand oscillator potential has its minimum at
0+ =
2
2a+ = 2a + ,
+ =

3
8
1+ 1 ,
2
9
with
and the energy levels are given by

1
+ + U 0+ .
Enr = n +
2
(6.6)
(6.7)
(6.8)
154
Here
2
+

= U 0+ = 2m2

9
8
1+ 1 1
8
9
(6.9)
and

U 0+ = 2a U ( + ) < 0.
(6.10)
So the condition for a degeneracy of a level in the spectrum of the right hand well with the ground
state of the left hand well (our initial state) is

1
1
+ = m.
2aU (+ ) + n +
(6.11)
2
2
We present here the dependence of the tunneling phenomena on the spatial length scale a. If
the degeneracy holds for some value a and an integer n, it will hold again for a = a + a,
n = n n with
+
a =
(6.12)
n.
2U (+ )
The constant multiplying n on the right-hand side determines the spacing of resonance levels
as a function of a at fixed and . Of course we may also have resonances between excited states
of the left well and those of the right hand well, but, as our initial state will roughly correspond
to the ground state of the left hand well, these will be less important. Another, more essential
feature is the excitation of field quanta of the nonzero modes. These will have a dissipative effect
on the dynamics of the zero mode and broaden the resonances. The interaction consists in a
deformation of the potential in which the zero mode is moving, which takes the form

U0 (0 , t) = U (0 ) + U (0 ) m2 F(t).
(6.13)
If F(t) is negligible the evolution of the system proceeds like in the quantum mechanics of the
zero mode. If F(t) remains small this time-dependent modulation of the potential will allow a
few other approximate eigenstates of the zero mode to mix in, the resonant oscillations develop
higher harmonics and become irregular. If F(t) gets large then the potential can be deformed
in such a way that the potential barrier disappears entirely, in such cases the zero mode may
slide into the new minimum. This happens if F(t) is positive; if F(t) is negative the potential
is tilted counterclockwise and the barrier is enhanced.
6.3. Results of the numerical simulations
We have performed a study of tunneling as a function of the radius of the space manifold
S1 , for fixed values of the parameters and , or and ; the mass is chosen to be unity,
which determines length and
time scales. We have considered four parametersets: set I: = 0.8,
= 0.5, i.e., = 1.6, = 1/ 2; set II: = 0.6, = 2, i.e., = 0.3, = 1/ 8; set III: = 0.4,
= 1, i.e., = 0.4, = 1/2; set IV: = 0.25, = 0.25, i.e., = = 1. For set II we expect
the quantum corrections to be small, for sets I and IV we expect large quantum corrections, and
moderate ones for set III.
We have studied in general the behavior of the average of the zero mode 0 (t), of the fluctuations (fluctuation integrals, particle number, and the various energies), and of the wave function
0 (0 ) of the zero mode, as functions of time. As the main indicator of the general behavior
155
Fig. 1. Energy conservation; parameters m = = = 1, a = 6; dotted line: classical energy, Eq. (3.29); dash-dotted line:
fluctuation energy, Eq. (3.30); dashed line: interaction energy, Eq. (3.31); solid line: self-interaction of the fluctuations,
Eq. (3.32). The total energy (not displayed) is E = (2.39 0.001) 105 .
Fig. 2. The maximum 0 of the expectation value 0 (t) for set I, = 0.8, = 0.5, as a function of a.
we use the maximal value attained by the expectation value of the zero mode during the time
evolution. We denote this value as 0 . In those cases where we have effective tunneling, this zero
mode average settles at a value of 0 beyond the potential barrier, the late time average being
typically 20% smaller than the maximal value.
We display in Figs. 1 to 4 the dependence of the maximal values 0 of 0 (t), as functions
of a for the four parameter sets. The horizontal lines labelled by m and 0 indicate the position of the maximum of U () and the zero of the position between the two minima (the end
of the tunnel). We see in all cases that an effective tunneling occurs as a resonance phenomenon. According to our estimate in the previous subsection, Eq. (6.12), the spacing of different
156
Fig. 3. The maximum 0 of the expectation value 0 (t) for set II, = 0.6, = 2, as a function of a.
Fig. 4. The maximum 0 of the expectation value 0 (t) for set III, = 0.4, = 1, as a function of a.
resonances as a function of a would be given by a = 0.529 for set I, a = 0.0338 for set II,
a = 0.0167 for set III and a = 0.0166 for set IV. The observed spacings roughly correspond
to these estimates. We note that also the excited states of the left hand oscillator or of the nonzero
modes can come into play, so if we observe some irregular spacings this may be due to such
effects. When compared to similar figures obtained in quantum mechanical simulations [20] the
resonances seen in our simulations are broader. We observe that even on resonance the maximal
value of 0 never reaches the second minimum (the true vacuum). In part this is due to the
fact that the effective potential seen by the zero mode is modified by the quantum fluctuations,
moreover, even at late times the wave function generally retains a finite probability density near
the false vacuum 0 = 0. This will be discussed in more detail below.
157
Fig. 5. The maximum 0 of the expectation value 0 (t) for set IV, = 0.25, = 0.25, as a function of a.
For the parameter sets IIII the tunneling shuts off entirely at higher values of a as expected
from the bounce action (6.4) being proportional to a. As large implies a large classical action
and, therefore, a small semiclassical tunneling rate, this shutting-off already happens at small
values, a 0.5, for set II. For set I on still finds resonant tunneling for a > 3. However, the
transition time at the resonances increases substantially (t 1000 at the last peak in Fig. 1),
between the resonances the amplitude of the zero mode oscillations decays exponentially with
a, as for sets II and III, and the zero mode wave function displays no tunneling at all. The case
of set IV is quite different. Here at large a the system always tunnels. This effect is due to the
fact that the quantum fluctuations deform the potential seen by the zero mode, Eq. (6.13) into a
potential without barrier, the late time wave function extending over a range from 5 to +20,
we call this kind of transition a sliding transition. We will discuss this below.
The results for the maximal value 0 of the zero mode average display a global picture of
the tunneling in the asymmetric double well potential in quantum field theory. They look quite
different from what one usually assumes to happen in false vacuum decay, when one just
considers the homogeneous bounce action (6.4) without or with quantum corrections. We will
now discuss in some more detail the evolution of the zero mode wave function.
For all parameter sets and all values of a the wave function of the zero mode initially evolves
slowly; it becomes slightly asymmetric and develops a tail into the region of the potential barrier.
Here already the fact that we do not restrict it to be Gaussian is of prime importance. The further
behavior then depends on the parameter sets.
On the resonances the wave function, which initially is essentially the ground state wave
function of the left well first connects to a particular wave function in the right well, obviously
the one of the degenerate level. This is seen in Fig. 6, the well-defined number of peaks within the
right hand well indicate a specific approximate level within the right hand well. If the fluctuations
of the quantum field, i.e., the amplitudes of the n , remain small, then one observes an oscillation
forth and back between these two wave functions, as expected in a purely quantum mechanical
system. If the nonzero field modes get excited, then these oscillations of the zero mode are
disturbed, other wave functions mix in, as seen by the increasing number of peaks, in some cases
the wave function looks quite chaotic.
158
Fig. 6. Probability density |0 (0 )|2 at intermediate times, here t = 10, for = 0.8, = 0.5, and a = 1.35.
Fig. 7. Evolution of the expectation value of the zero mode 0 and of the particle number N ; parameters = 0.6,
= 2, a = 0.5.
Off resonance the expectation value of 0 oscillates regularly with a small amplitude. It still
develops a tail into the right hand well, but this part of the wave functions remains small.
Beyond the resonance region there is efficient tunneling if the fluctuations of the nonzero
momentum modes are large. In this case, while the wave function penetrates into the barrier
region, the fluctuation integral becomes positive, the barrier disappears and the wave function
slides down the hill retaining almost its initial Gaussian form.
The simulations of set II, = 0.6, = 2 are expected to resemble most closely the case of
quantum mechanics. The nonzero momentum quantum fluctations indeed remain small and we
have the resonances expected from the qualitative picture of degenerate levels of the individual
wells. The results for the expectation value of the zero mode and for the evolution of the wave
159
Fig. 8. The zero mode wave function for = 0.6, = 2, a = 0.5; dashed lines: the wave functions at t = 0 and 109;
solid lines: the wave functions at t = 50 and t = 168.
Fig. 9. Evolution of the expectation value 0 (t) and of the fluctuation integral F (t) for = = 0.25 (set IV) and
a = 1.2.
functions correspond to these expectations. We present the regular oscillations of the zero mode
on the resonance at a = 0.5 in Fig. 7. We plot, in the same figure, the particle number N defined
below Eq. (3.36), multiplied by 1000. The wave function, shown at four different times in Fig. 8,
is seen to return to itself almost exactly after half a period of t 55.
Set III is intermediate in the sense that with = 1 we expect the fluctuations of the nonzero
modes to be sizeable but not really large. We find that tunneling is again characterized by the
occurrence of resonances, and again shuts off at larger values of a, here around a = 2. Around
this value the resonant transitions occur on longer and longer time scales. So for a > 2 we cannot
exclude further resonances if we run the simulations for times longer than t 300.
160
Fig. 10. The wave function after tunneling, = = 0.25 (set IV) and a = 1.2 and time t = 30.
Fig. 11. Evolution of the wave function during the barrier transition: parameters m = = = 1; a = 6; long-dashed line:
t = 7; dash-dotted line: t = 7.4; short-dashed line: t = 7.8; dotted line: t = 8.2; straight line: t = 8.6.
For set IV the zero mode shows resonant behaviour at small a, but for a > 0.5 the system
always ends up in the right well. The actual evolution is quite involve here. Once the fluctuations
set in, they tilt the potential in such a way that the barrier disappears and that the wave function
can start to move right. Then the fluctuation integral F gets small again or negative, so the
potential barrier appears again. This process repeats itself in an oscillatory way, and the wave
function gradually shifts towards the deeper well. This is displayed in Fig. 9. At late times the
quantum modes still oscillate with a sizeable amplitude, so that the potential seen by the zero
mode keeps on changing and can neither be considered to be double or single well. A typical
wave function at late times is shown in Fig. 10 for a = 1.2 and t = 30. One sees no trace of a
161
Fig. 12. Evolution of the expectation value 0 (t) and of the fluctuation integral for = 0.8, = 0.2, and a = 5.
Fig. 13. Evolution of the effective potential U0 (0 , t); solid line: t = 0; dashed line: t = 11; dash-doubledotted line:
t = 13.
double-well structure. In spite of the complexity of the wave function the energy is still conserved
better than one part in 108 (see Fig. 1).
The situation becomes more transparent for larger values of a.1 We show results for a = 6.
Here again the potential is tilted by the fluctuations, but the fluctuation integral remains positive on the average at late times. This implies that the barrier has disappeared definitively. The
transition can be described as a sliding of the wave function towards the new minimum. This is
displayed in Fig. 11.
1 These values do not appear in Fig. 5 as otherwise the structure in the resonance region cannot be resolved.
162
For the case under consideration, set IV with = 0.25 the potential is very asymmetric. So
one may infer that the tilting of the potential is most effective here, and that this effect may in
fact be limited to small values of . We therefore have performed simulations for = 0.8 and
= 0.2, i.e., for a more symmetric potential, but again with large quantum corrections. We find
that for values of a > 1 the transition again occurs by sliding, but that at late times the fluctuation
integral becomes negative and the potential is tilted in such a way that the barrier reappears. This
is presented in Figs. 12 and 13. The wave function at late times is then found to have sizeable
amplitudes in both wells.
So it seems that sliding occurs universally for small and large a.
7. Conclusions and outlook
We have presented here an analysis of global tunneling in quantum field theory on a compact
space. This analysis was based on a real-time formulation, using the time-dependent Hartree approximation. We have performed numerical simulations of a system where the wave function of
the zero momentum mode evolves as a solution of the Schrdinger equation, while the nonzero
momentum modes are treated in the Gaussian approximation. The approximation includes the
back-reaction of the nonzero momentum modes onto the zero momentum mode and onto themselves. The renormalized equations were obtained using the 2PPI formalism, adopted here to a
system with finite space extension.
We have found that tunneling for such systems occurs in a variety of different ways. For large
, implying small field fluctations, the system behaves like a quantum mechanical system with
a single degree of freedom. Tunneling is effective whenever it connects degenerate modes, so
varying the parameters, here the radius a, one finds resonance enhancements. For large spatial
extension a this resonant tunneling shuts off as expected from the fact that the WKB integral
over the barrier is proportional to a. For smaller values of the nonzero momentum modes,
the modes which make up the quantum field, are enhanced and modify the quantum mechanical
behavior. If this enhancement is weak it can be considered as a kind of dissipation; the mean
value of the zero mode shows regular periodic oscillations, on and off resonance. On resonance
this again involves two or a few more modes in the two separate wells, off resonance the wave
function essentially remains in the left well, with some tail in the right hand well. For parameter
sets where the excitation of the nonzero momentum modes becomes sizeable, these regular oscillations are disturbed, the oscillations of the mean value of the zero mode become irregular and
so do the wave functions of the zero mode. Off resonance at late times the system again remains
concentrated in the left hand well. On resonance at late times the wave function extends over the
entire region allowed at the given energy. While the system may then be considered as having
tunneled, it certainly ends up in a rather complex excited state.
If the quantum corrections are large the system exhibits another phenomenon: the quantum
fluctuations tilt the effective potential for the zero mode in such a way that the barrier disappears
entirely. In such cases the system wave function can be observed to slide into the new minimum.
The disappearance ot the barrier may be a temporary effect. It is due to the substantial quantum
excitations once the wave function enters the barrier, leading to negative squared masses for the
fluctations. At late times the potential may become be a double-well potential again, or the barrier
may disappear definitively.
This variety of phenomena observed here in a real-time analysis cannot be expected to be
described by a simple transition rate formula. A better approximation than the classical bounce
action, Eq. (6.4), could be obtained by fitting together WKB patches within the allowed and for-
163
bidden regions or by applying a time-dependent WKB approximation to the k = 0 mode itself,

as done for quantum mechanics in Ref. [27]. However, the advantage of dealing with analytic or
semi-analytic formulas is lost at the latest if one includes the nonzero momentum fluctuations.
In most cases they cannot be evaluated analytically and, as implied by the present analysis, they
cannot be expected to give reliable estimates. For homogeneous tunneling they exhibit the phenomenon of multiple unstable modes: if the unstable mode for k = 0 has the squared frequency
| |2 < 0 then the k = 0 modes will have an eigenmode with the eigenvalue k 2 /a 2 | |2
which is positive for small a but will become negative whenever a > ak = k/| |. The interpretation of these modes is unclear, there is no trace of them in the numerical simulations.
With increasing a one expects the system to become unstable with respect to inhomogeneous,
local bounces. This instability may be closely related to the sliding phenomenon for which a
strong excitation of k = 0 modes is crucial: the k = 0 modes are of course spatially inhomogeneous; if they are excited substantially they may be interpreted as classical fluctuations. For
k = n they have the general form of a chain of n bubbles, much in the way as they are depicted
in Refs. [3,4] for the decay of the false vacuum at finite temperature. It could be interesting to
investigate this relation between inhomogeneous and homogeneous tunneling in a more quantitative way. Clearly, in view of this instability with respect to local bounces, our present analysis
becomes unreliable at values a 1.
The restriction to one compact space dimension was done in order to reduce technical complications to a minimum. Similar phenomena are expected for compact three-dimensional spaces.
The discrete spectrum of momenta then becomes a discrete spectrum of angular momenta, whose
excitation will be small for small a and become more effective with increasing a. The potential
seen by the homogeneous mode is again a double-well potential, so again resonances are expected to dominate the low a regime. If gravity is included, like for transitions in de Sitter space,
renormalization becomes problematic, especially if quantum backreaction is included. Still it
would be interesting to investigate along these lines the relation between homogeneous transitions like the HawkingMoss instanton and inhomogeneous ones like the ColemanDe Luccia
bounce.
Of course it would be even more interesting to follow, in real time, the local creation of real
vacuum bubbles including the associated excitation of quantum modes with back-reaction. This
seems to be out of scope with the presently available numerical methods and computer facilities.
Our calculations may elucidate in a qualitative way effects that are missed in the WKB approach
to local transitions, even when quantum corrections are taken into account [3437].
Acknowledgements
One of us (N.K.) thanks the Deutsche Forschungsgemeinschaft for support in the framework
of the Graduiertenkolleg 841: The physics of elementary particles at accelerators and in the
universe.
References
[1]
[2]
[3]
[4]
[5]
[6]
S.R. Coleman, The fate of the false vacuum: Semiclassical theory, Phys. Rev. D 15 (1977) 29292936.
S.R. Coleman, F. De Luccia, Gravitational effects on and of vacuum decay, Phys. Rev. D 21 (1980) 3305.
A.D. Linde, Fate of the false vacuum at finite temperature: Theory and applications, Phys. Lett. B 100 (1981) 37.
A.D. Linde, Decay of the false vacuum at finite temperature, Nucl. Phys. B 216 (1983) 421.
S.W. Hawking, I.G. Moss, Supercooled phase transitions in the very early universe, Phys. Lett. B 110 (1982) 35.
S.W. Hawking, N. Turok, Open inflation without false vacua, Phys. Lett. B 425 (1998) 2532.
164
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
A.D. Linde, Chaotic inflation, Phys. Lett. B 129 (1983) 177181.

K. Hirota, Nonequilibrium dynamics of quantum tunneling, Phys. Rev. D 61 (2000) 123502.
J.B. Hartle, S.W. Hawking, Wave function of the universe, Phys. Rev. D 28 (1983) 29602975.
T. Vachaspati, A. Vilenkin, On the uniqueness of the tunneling wave function of the universe, Phys. Rev. D 37
(1988) 898.
A. Vilenkin, Approaches to quantum cosmology, Phys. Rev. D 50 (1994) 25812594.
A. Vilenkin, Eternal inflation and the present universe, Nucl. Phys. B (Proc. Suppl.) 88 (2000) 6774.
J.-y. Hong, A. Vilenkin, S. Winitzki, Particle creation in a tunneling universe, Phys. Rev. D 68 (2003) 023520.
D.H. Coule, J. Martin, Quantum cosmology and open universes, Phys. Rev. D 61 (2000) 063501.
M. Bouhmadi-Lopez, L.J. Garay, P.F. Gonzalez-Diaz, Quantum behavior of FRW radiation-filled universes, Phys.
Rev. D 66 (2002) 083504.
S.P. Kim, Quantum cosmology for tunneling universes, J. Korean Phys. Soc. 45 (2004) S172S180.
V.A. Rubakov, Particle creation during vacuum decay, Nucl. Phys. B 245 (1984) 481.
D. Levkov, C. Rebbi, V.A. Rubakov, Tunneling in quantum cosmology: Numerical study of particle creation, Phys.
Rev. D 66 (2002) 083516.
J. Hong, A. Vilenkin, S. Winitzki, Particle creation in a tunneling universe. II. Massive particles, Phys. Rev. D 68
(2003) 023521.
M.M. Nieto, V.P. Gutschick, F. Cooper, D. Strottman, C.M. Bender, Resonances in quantum mechanical tunneling,
Phys. Lett. B 163 (1985) 336342.
O.J.P. Eboli, R. Jackiw, S.-Y. Pi, Quantum fields out of thermal equilibrium, Phys. Rev. D 37 (1988) 3557.
F. Cooper, S. Habib, Y. Kluger, E. Mottola, Nonequilibrium dynamics of symmetry breaking in 4 theory, Phys.
Rev. D 55 (1997) 64716503.
D. Boyanovsky, H.J. de Vega, R. Holman, J.F.J. Salgado, Analytic and numerical study of preheating dynamics,
Phys. Rev. D 54 (1996) 75707598.
J. Baacke, K. Heitmann, C. Patzold, Nonequilibrium dynamics: A renormalized computation scheme, Phys. Rev.
D 55 (1997) 23202330.
S.A. Ramsey, B.L. Hu, Nonequilibrium inflaton dynamics and reheating: Back reaction of parametric particle creation and curved spacetime effects, Phys. Rev. D 56 (1997) 678705.
R. Jackiw, A. Kerman, Time dependent variational principle and the effective action, Phys. Lett. A 71 (1979) 158
162.
F. Cooper, S.-Y. Pi, P.N. Stancioff, Quantum dynamics in a time dependent variational approximation, Phys. Rev.
D 34 (1986) 3831.
C. Destri, E. Manfredini, An improved time-dependent HartreeFock approach for scalar 4 QFT, Phys. Rev. D 62
(2000) 025008.
P. Dirac, Proc. Cambridge Philos. Soc. 26 (1930) 376.
G.H. Hardy, Divergent Series, Oxford Univ. Press, Oxford, 1949, Section 13.14.
J. Baacke, S. Michalski, Nonequilibrium evolution in scalar O(N ) models with spontaneous symmetry breaking,
Phys. Rev. D 65 (2002) 065019.
Y. Nemoto, K. Naito, M. Oka, Effective potential of O(N ) linear sigma model at finite temperature, Eur. Phys. J.
A 9 (2000) 245259.
M. Dine, R.G. Leigh, P.Y. Huet, A.D. Linde, D.A. Linde, Towards the theory of the electroweak phase transition,
Phys. Rev. D 46 (1992) 550571.
J. Baacke, V.G. Kiselev, One loop corrections to the bubble nucleation rate at finite temperature, Phys. Rev. D 48
(1993) 56485654.
J. Baacke, G. Lavrelashvili, One-loop corrections to the metastable vacuum decay, Phys. Rev. D 69 (2004) 025009.
Y. Bergner, L.M.A. Bettencourt, The self-consistent bounce: An improved nucleation rate, Phys. Rev. D 69 (2004)
045012.
J. Baacke, N. Kevlishvili, Self-consistent bounces in two dimensions, Phys. Rev. D 71 (2005) 025008.
Universal character and large N factorization

in topological gauge/string theory
Hiroaki Kanno
Graduate School of Mathematics, Nagoya University, Nagoya 464-8602, Japan
Abstract
We establish a formula of the large N factorization of the modular S-matrix for the coupled representations in U (N ) ChernSimons theory. The formula was proposed by Aganagic, Neitzke and Vafa, based
on computations involving the conifold transition. We present a more rigorous proof that relies on the
universal character for rational representations and an expression of the modular S-matrix in terms of the
specialization of characters.
1. Introduction
In [1] a remarkable relation
ZBPS |Ztop |2 ,
(1.1)
was proposed. The proposal (1.1) relates the partition function ZBPS of counting BPS black holes
formed by D-branes and topological string amplitudes Ztop . While several attempts have been
made at making (1.1) more precise, the conjecture has been tested for non-compact Calabi
Yau 3-folds in [25], where BPS microstates arise from D4-branes wrapping over a line bundle
O(n) on a Riemann surface . The (electric) charges of black hole are given by the
numbers of D0 and D2 branes bound to D4 branes. When we have N D4 branes, the partition
function ZBPS is computed from U (N ) topological YangMills theory on the world volume
of D4 branes and it has been shown that it is reduced to the q-deformed YangMills theory
on [3]. The large N factorization of two-dimensional YangMills theory and its interpretation
E-mail address: kanno@math.nagoya-u.ac.jp (H. Kanno).
doi:10.1016/j.nuclphysb.2006.03.014
166
H. Kanno / Nuclear Physics B 745 [PM] (2006) 165175
by a closed string theory were given in [6,7]. The group theory of coupled representation labeled
by a pair of Young diagrams was employed there. In the q-deformed case, the key to the large N
factorization of the partition function is the following formula given in [4]:
1
q N + 24 SPQ (q, N ) = M(q)(q)N q 2 (Q+ +Q ) q 2 (|P+ |+|P |+|Q+ |+|Q |)

(1)|R| q N |R| CP+ Qt+ R (q)CP Qt R t (q),

2
(1.2)
for an extension SPQ of the modular matrix of the U (N ) ChernSimons theory to the coupled
representations P, Q. In view of the conjecture (1.1), it is important that in (1.2) we have the
topological vertex CP QR (q), which is a building block of all genus topological string amplitudes
on local toric CalabiYau 3-folds [810]. On the other hand, the modular S-matrix SPQ appears
as a building block for ZBPS [3,5]. For example, for D4 branes wrapping over a non-compact
four cycle O(p) P1 , the BPS partition function is

p
(SR )2 q 2 C2 (R) eiC1 (R) ,
Z P1 , p =
(1.3)
R
where C1 (R) and C2 (R) are the Casimir of irreducible U (N ) representation R and is the
angle of the q-deformed YM 2 . In this paper we will use for the trivial representation. We note
that SR is related to the quantum dimension by SR = S dimq R. Up to the normalization
associated with S , the formula (1.2) gives

N
(1)|Q| q N |Q| CR+ Q (q)CR Qt (q).
SR (q, N ) = q 2 (|R+ |+|R |)
(1.4)
Q
Since only the trivial representation Q = survives in the large N limit, we obtain1

2 p
+
CR (q) q 2 C2 (R ) eiC1 (R ) .
Z P1 , p Ztop
Ztop ,
Ztop
:=
(1.5)
The factor Ztop

is the topological string amplitude on the local toric CalabiYau manifold:
O(p 2) O(p) P1 , with an appropriate (complexified) Khler parameter.2 Thus we find
the topological string amplitudes in the factorization of ZBPS at large N .
In [4] the validity of the formula (1.2) has been argued by comparing the open/closed topological string amplitudes related by the conifold transition. The main purpose of this article is to
give a more rigorous proof based on the universal character of the coupled representation labeled
by a pair of partitions (, ) or Young diagrams. The Schur function s (x) is nothing but the universal character of irreducible (polynomial) representation corresponding to a single partition .
The coupled representation used in [4] is what is called rational representation in mathematics
literature and the universal character s[,] (x, y) of the rational representation is a generalization
of the Schur function defined, for example, in [12]. This article is organized as follows. In Section 2 we introduce the universal character for rational representation following [12]. We will see
that s[,] (x, y) is expanded in terms of (skew) Schur functions. In Section 3 we review how the
Hopf link invariants, or the normalized modular S-matrices of the ChernSimonsWZW theory,
are expressed in terms of specialization of Schur functions [13]. These facts are basic ingredients
1 We use C (R) = C (R ) C (R ) and C (R) = C (R ) + C (R ) + O(1/N ).

1
1 +
1
2
2 +
2
2 See also [11] for the relation of topological string amplitudes to the q-deformed 2D YangMills theory.
167
in our proof of (1.2), since it is known that the topological vertex has an expression in terms
of specialization of skew Schur functions [14]. Finally, we prove (1.2) in Section 4 by using
the formulas for the specialization of skew Schur function. In might be interesting to investigate
some applications of the formula (1.2) to the large N limit of the discrete matrix model and the
generalized two-dimensional YangMills theory [15].
2. Universal character for rational representation
A finite-dimensional representation X (X) of GL(N, C)3 is called polynomial or rational,
if the matrix components of (X) are polynomial or rational functions of the components of X,
respectively. For example, the determinant (X) = det X is a one-dimensional polynomial representation. The Cartan subalgebra of GL(N, C) is the set of diagonal matrices hN := {X; X =
diag(x1 , x2 , . . . , xN )}. The weight lattice is generated by i : hN C, i (X) := xi . It is well
known that the irreducible polynomial representations of GL(N, C) are labeled by partitions :
1 2 () > 0 of length () N . The highest weight of the representation
labeled by is 1 1 + 2 2 + + N N and the character is known as the Schur polynomial Tr X = s (x1 , . . . , xN ), where (x1 , . . . , xN ) are eigenvalues of X. For a fixed partition ,
the projective limit s (x) = lim
s (x1 , . . . , xN ) has the meaning in the ring of symmetric functions x , which is the projective limit of the algebra of symmetric polynomials in N variables. In
this sense the Schur function s (x) x is sometimes called universal character of polynomial
representation.
The irreducible rational representations of GL(N, C) were classified by Schur. He showed
any rational representation is of the form X (det X)r (X), where r Z and is a polynomial
representation. That is, the denominator of rational representation is a power of the determinant. We may identify the integer r with the U (1)
charge of the rational representation. The
highest weight of the determinant representation is N
i=1 i . Thus a complete set of inequivalent rational representations of GL(N,
C)
is
indexed
by
N -tuples ZN : 1 2 N
N
and the highest weight is given by i=1 i i . The polynomial representations are characterized by the condition N 0. Separating the (strictly) positive and negative parts of , we
can also label them in terms of a pair of partitions (+ , ) with (+ ) + ( ) N so that
(1 , 2 , . . . , N ) = (+,1 , . . . , +, (+ ) , 0, . . . , 0, , ( ) , . . . , ,1 ). In SU(N ) gauge theory the difference of polynomial representations and rational representations is irrelevant, since
the determinant representation is trivial. However, for U (N ) gauge theory the difference is important. In fact, Gross and Taylor employed the rational representation, which they called composite
representation, to work out the large N factorization of the partition function of two-dimensional
YangMills theory [6,7].
( )
Let be a partition of | | := i=1 ,i . Then it is known that the rational representation R labeled by (+ , ) occurs in the tensor product of |+ | copies of the defining
representation CN and | | copies of the contragradient representation (CN ) :
|+ |
| |

N
N
.
T|+ |,| | :=
(2.1)
C
C
The universal character of the rational representation has been introduced in [12]. Let us denote
the universal character of the rational representation R by s[+ , ] (x, y), where the variables x
3 U (N ) and SU(N ) are the compact real forms of GL(N, C) and SL(N, C), respectively.
168
Fig. 1. Rational representation as a tensor product of the determinant (the shaded rectangle) and a polynomial representation.
and y are associated with the defining representation CN and the contragradient representation
(CN ) , respectively. Consequently, s[+ , ] (x, y) is in the tensor product of universal character rings x y and it can be expanded in terms of the products of Schur functions. Let us
introduce the LittlewoodRichardson coefficients c, defined by the relation

s (x)s (x) =
(2.2)
c,
s (x).
The coefficients
c,
are non-negative integers. In [12] the expansion

+
s[+ , ] (x, y) =
(1)| | c,
c, t s (x)s (y),
(2.3)
,,
and its inverse relation

s (x)s (y) =
c+ , c , s[+ , ] (x, y),
(2.4)
,+ ,
were proved. The partition t in (2.3) is defined by the transpose of the Young diagram. Using
the skew Schur function defined by
s/ (x) =
(2.5)
c,
s (x),
we can write the expansion (2.3) as

(1)| | s+ / (x)s / t (y).
s[+ , ] (x, y) =
(2.6)
Finally, we note that there is the embedding GL(N, C)

g (g, (g t )1 ) GL(N, C)
GL(N, C), which is a group homomorphism. We can use the induced homomorphism x
y x obtained by putting y = x 1 to send the above formulas to relations in the universal
character ring x of GL(N, C).
3. Hopf link invariants and specialization of Schur function
Both the elementary symmetric functions e1 (x), . . . , en (x), . . . and the complete symmetric
functions h1 (x), . . . , hn (x), . . . are Z-basis of the ring x of the symmetric functions. They are
defined by the following generating functions:
E(t, x) := 1 +

k
ek (x)t k =

(1 + xi t),
i=1
(3.1)
169
H (t, x) := 1 +

hk (x)t = (1 xi t)1 ,
k
(3.2)
i=1
that satisfy E(t, x)H (t, x) = 1. We note that the power sum functions p1 (x), . . . , pn (x), . . .
are not Z-basis, but Q-basis of the symmetric functions. The generating function is

d
P (t, x) :=
(3.3)
pk (x)t k1 =
log H (t, x).
dt
k
Any symmetric function f x can be written as a polynomial f (e1 , . . . , en , . . .) in the elementary symmetric functions. For example, the JacobiTrudi formula gives the Schur function
in terms of {ei (x)}:

s (x) = det et i+j (x) .
(3.4)
i
We can define a specialization

of the elementary symmetric functions by taking the generating

n , to be any formal power series with the leading coefficient 1. We
function E(t) = 1 +
e
t
n=1 n
denote f (e1 , . . . , en , . . .) as f (E(t)), which gives a specialization of the symmetric function f .
A basic example of such a specialization of symmetric functions is given in [16, Examples I-2.5
and I-3.3]. When we take the generating functions
H (t) =

1 bq i t
i=0
1 aq i t
E(t) =

1 + aq i t
i=0
1 + bq i t
(3.5)
we have
hn (q) =
n

a bq i1
1 qi
i=1
en (q) =
n

aq i1 b
i=1
1 qi
(3.6)
In this specialization the power sum function is

pn (q) =
a n bn
,
1 qn
(3.7)
and the Schur function is given by

a bq c(i,j )
,
1 q h(i,j )
(3.8)

1
1 + c(i, j ) h(i, j ) .
2
(3.9)
s (q) = q n()
(i,j )
where n() is
n() =
(i,j )
We have introduced the standard notation for the content c(i, j ) := j i and the hook length
h(i, j ) := i + tj i j + 1 of the box at (i, j ) in the Young diagram.
Morton and Lukac used this kind of specialization of Schur functions to express the quantum
dimension and Hopf link invariants [13]. For example, when a = 1 , b = 1,
1
s (q) = q 2
(i,j ) (h(i,j )c(i,j )1)
1 q c(i,j )
|| [c(i, j )]
= (q) 2
. (3.10)
h(i,j
)
[h(i, j )]
1q
(i,j )
(i,j )
170
We find the quantum dimension dimq R =
[c(i,j )]
(i,j ) [h(i,j )]
of the representation R defined by
2i
) and = q N . Thus
the partition . In the SU(N ) ChernSimons theory at level k, q = exp( N+k
we have

||
W (q, ) := dimq R = 2 s E (t; q, ) ,
(3.11)
where
E (t) := 1 +

q 2 t
n
n
1 1 q i1
n=1
qi 1
i=1

1 + 1 q i 2 t
1
i=1
1 + q i 2 t
(3.12)
||
We have absorbed the overall factor q 2 in (3.10) by the shift of the exponent q i q i 2 . In
this paper we will take the convention4 q = exp(gs ) with gs being string coupling (the parameter
of genus expansion of topological string theory). Thus the relevant region of q in our convention
is |q| > 1. Since the expression in (3.12) is that for |q| < 1, we make an analytic continuation to
the region |q| > 1 [17,18], to obtain
E (t) =
1 + q i+ 2 t
i=1
1 + 1 q i+ 2 t
(3.13)
We note that SU(N ) specialization = q N gives

E(N ) (t) =

1 + q i+ 2 t
i=1
1+q
N i+ 12
N

1
1 + q i+ 2 t ,
(3.14)
i=1
1
which is the same as the generating function (3.1) for the specialization xi = q i+ 2 (1 i N ),
xj = 0 (N < j ). Similarly, the Hopf link invariants are also expressed in terms of an appropriate
specialization of Schur function as follows:

W (q, ) = W (q, )||/2 s E (t : q, ) ,
(3.15)
where
E (t) :=
()

1 + q i i+ 2 t
i=1
1 + q i+ 2 t
E (t) =

1 + q i i+ 2 t
1
i=1
1 + 1 q i+ 2 t
(3.16)
When N , 1 = q N 0 for |q| > 1 and consequently

1
W (q) 2 (||+||) s q s q + ,
where q and q + mean the specialization xi = q
i+ 12
and xi = q
(3.17)
i i+ 12
, respectively.
4. Large N factorization of the modular matrix of U (N) CS theory

The modular S-matrix SPQ of the U (N ) ChernSimons theory for coupled representations P, Q is defined by

(1)w q w(P +N )(Q+N ) ,
SPQ (q, N ) =
(4.1)
wSN
4 This is the same as [9], but different from [35], where q = exp(g ).
s
171
where the symmetric group SN is the Weyl group of U (N ) and N is the Weyl vector with the
components 12 (N 2i + 1), i = 1, . . . , N . Note that the same notation P, Q is used for the
highest weight. When P and Q are integrable representations P , Q of SU(N ) affine Lie algebra at level k, (4.1) gives the matrix element of the S-transformation on the space of conformal
blocks, which is the physical Hilbert space of the ChernSimons theory on T 2 R. The Hopf
link invariants are obtained as the normalized S-matrix elements: WP Q = SP Q /S . The definition (4.1) is a formal extension of the modular S-matrix to rational representations of U (N ).
2i
) = exp(gs ). In [4] the following formula for SPQ was claimed:
Recall that q = exp( N+k
1
q N + 24 SPQ (q, N ) = M(q)(q)N q 2 (Q+ +Q ) q 2 (|P+ |+|P |+|Q+ |+|Q |)

(1)|R| q N|R| CP+ Qt+ R (q)CP Qt R t (q),

2
(4.2)
where
P = [P+ , P ],
Q = [Q+ , Q ],
(4.3)
(q) is the topological vertex.5 For the representation R

are two coupled representations and CP QR
corresponding to a partition R , R := 2 (i,j )R c(i, j ). In this section we identify the representations, the partitions and the Young diagrams and often use the same notation for them. The
MacMahon function M(q) and the eta function (q) defined by

1
1/24
1 qn ,
,
(q)
=
q
M(q) =
(4.4)
n
n
(1 q )
n
n
appear as overall factors and they are related to the normalization of the Hopf link invariants (see
Appendix A).
Let us look at some special cases of the formula (4.2). When P = Q = , S is nothing
but the partition function of U (N ) ChernSimons theory on S 3 and (4.2) gives

N
2
(1)|R| q N |R| CR (q)CR t (q)
q N + 24 S = M(q)(q)N
R
= M(q)(q)N
1 1 q n
n=1
= M(q)(q) exp
N

n=1
n

etn
.
n(q n/2 q n/2 )2
(4.5)
If we identify the t Hooft coupling t = Ngs with the Khler parameter of the resolved conifold
geometry, we recover the basic example of gauge/geometry correspondence of GopakumarVafa
[19]. When P = Q = , the left-hand side is the usual modular matrix of U (N ) theory and
(4.2) implies
WP Q (q, N ) =
SP Q (q, N )
S (q, N )
1
= q 2 Q + 2 (|P |+|Q|)
1 1 q n
n=1
n
(1)|R| q N|R| CP Qt R CR t ,
R
(4.6)
5 In [4] q := exp(g ), which is different from ours.
s
172
which has been derived in [17]. When N only the trivial representation survives in the sum
and
1
WP Q (q) = q 2 Q + 2 (|P |+|Q|) CP Qt (q).
(4.7)
We recover the formula (3.17) for WP Q (q).

Topological vertex is expressed in terms of the skew Schur function [14]:

R2t
R2
sR1 /Q q + sR t /Q q + .
CR1 R2 R3 (q) = q R3 /2 sR2 q
3
(4.8)
Using the Cauchy formula

sR/R1 (x)sR t /R2 (y) =
(1 + xi yj )
sR t /Q (x)sR t /Qt (y),
2
i,j 1
(4.9)
and the relation

t
sP /Q q R+ = (1)|P ||Q| sP t /Qt q R ,
(4.10)
we can compute6
1
q 2 (Q+ +Q )

(1)|R| q N |R| CRP+ Qt+ (q)CR t P Qt (q)
R

t
t
1 1 q P+,i +P,j ij +1
= sP+ q sP q
1i,j

()|V | sQ+ /V 1 q P , q P+ + sQ /V t 1 q P+ , q P + ,
(4.11)
where the Schur function with two variables sR (x, y) is defined by

cPR Q sP (x)sQ (y) =
sR/Q (x)sQ (y).
sR (x, y) =
P ,Q
(4.12)
The corresponding generating function of the elementary symmetric functions is

E(t) =

i=1
(1 + xi t)
(1 + yj t).
(4.13)
j =1
The cyclic symmetry of the topological vertex,

(1)|R| q N|R| CRP+ (q)CR t P (q)
R

(1)|R| q N|R| CP+ R (q)CP R t (q),
(4.14)
implies
6 We use the cyclic symmetry of C
R1 R2 R3 . This type of computation appears in confirming the flop invariance of
topological string amplitudes [17,20].
173

t
t
1 1 q P+,i +P,j ij +1
sP+ q sP q
1i,j

n

1 1 q n |P |
(1)|V | sP+ /V t 1 q , q sP /V q , q .
(4.15)
n=1
Hence, we finally obtain

1
q 2 (Q+ +Q )

(1)|R| q N|R| CRP+ Qt+ (q)CR t P Qt (q)
R

n

1 1 q n (|P |+|Q |)
(1)|V | sP+ /V t 1 q , q sP /V q , q
n=1

(1)|V | sQ+ /V 1 q P , q P+ + sQ /V t q P+ , q P + .
(4.16)
On the other hand, the result reviewed in Section 3 formally implies an expression of SPQ in
terms of the specialization the Schur function sR (x):
WPQ (q, N ) :=

1
SPQ (q, N )
= 2 (|P |+|Q|) sP E (t; q, ) sQ EP (t; q, ) .
S (q, N )
(4.17)
But, we have to make it precise what sR (x) and ER (t; q, ) mean for the rational representation R. Firstly, the Schur function sR (x) should be regarded as the universal character of rational
representation defined by the pair of partitions (R+ , R ). The generating function ER (t) that
defines the specialization can be obtained by looking at the Young diagram of the rational representation R = [R+ , R ] (see Fig. 2). Assuming (R+ ) + (R ) N , we find
ER (t) =
(R
+ )
1 + q i+ 2 t
i=1
j =1
1 + q N+j 2 t
E (t)
1
1

1 + q R+,i i+ 2 t 1 + 1 q R,j +j 2 t
1
1 + q i+ 2 t
i=1
1 (R )
1
1 + q R+,i i+ 2 t 1 + q NR,j +j 2 t
1 + q R+,i i+ 2 t
i=1
1 + 1 q j 2 t
j =1
1 + q k+ 2 t
1
k=1
1 + 1 q k+ 2 t

1
1 + 1 q R,j +j 2 t .
j =1
Fig. 2. Extended Young diagram for rational representation labeled by (R+ , R ).
(4.18)
174
Since the universal character of the rational representation R is given by

(1)|Q| sR+ /Q (x)sR /Qt x 1 ,
sR (x) =
(4.19)
and the specialization of sR (x) is defined by ER (t) of (4.18), we obtain

1
(1)|U | sP+ /U 1 q , q sP /U t q , q
WPQ (q, N ) = 2 (|P |+|Q|)
U

(1)|V | sQ+ /V 1 q P , q P+ + sQ /V t q P + , q P+ .
V
(4.20)
Comparing (4.16) with (4.20), we see that the identification |P| = |P+ | |P |, |Q| = |Q+ |
|Q | completes the proof of (4.2).
Acknowledgements
We would like to thank T. Eguchi for helpful comments on the manuscript. We also thank
Y. Konishi, S. Matsuura, K. Ohta, S. Okada and S. Minabe for discussions.
Appendix A. Normalization of Hopf link invariants
The normalization factor of the Hopf link invariants S (q, N ) is the partition function of the
ChernSimons theory on S 3 . In topological string theory it is the contribution of constant maps
to topological string amplitudes [19]. For SU(N ) theory it is given by
1

1
q 2 (j i) q 2 (j i)
S (q, N ) =
1i<j N
= exp
1i<j N

j i
log q log 1 q j i
2
Using the strange formula for SU(N )

1
1
2
(j i) = N N 2 1 = N
2
12
(A.1)
(A.2)
1i<j N
and

log 1 q (j i) =

1i<j N
m=1

q m q m(N+1)
N qm
,
m(1 q m )
m(1 q m )2
(A.3)
we find
N
q N + 24 S (q, N ) = M(q)(q)N N0 (q, ),

2
where
= qN
and
N0 (q, ) = exp

n=1

qn
n
,
n(1 q n )2
is regarded as non-perturbative correction to the constant map contributions.
(A.4)
(A.5)
175
References
[1] H. Ooguri, A. Strominger, C. Vafa, Black hole attractors and the topological strings, Phys. Rev. D 70 (2004) 106007,
hep-th/0405146.
[2] C. Vafa, Two-dimensional YangMills, black holes, and topological strings, hep-th/0406058.
[3] M. Aganagic, H. Ooguri, N. Saulina, C. Vafa, Black holes, q-deformed 2d YangMills and non-perturbative topological strings, Nucl. Phys. B 715 (2005) 304348, hep-th/0411280.
[4] M. Aganagic, A. Neitzke, C. Vafa, BPS microstates and the open topological string wave function, hep-th/0504054.
[5] M. Aganagic, D. Jafferis, N. Saulina, Branes, black holes and topological strings on toric CalabiYau manifolds,
hep-th/0512245.
[6] D. Gross, W. Taylor, Two-dimensional QCD is a string theory, Nucl. Phys. B 400 (1993) 181210, hep-th/9301068.
[7] D. Gross, W. Taylor, Twists and Wilson loops in the string theory of two-dimensional QCD, Nucl. Phys. B 400
(1993) 395452, hep-th/9303046.
[8] M. Aganagic, M. Mario, C. Vafa, All loop topological string amplitudes from ChernSimons theory, Commun.
Math. Phys. 247 (2004) 467, hep-th/0206164.
[9] M. Aganagic, A. Klemm, M. Mario, C. Vafa, The topological vertex, Commun. Math. Phys. 254 (2005) 425478,
hep-th/0305132.
[10] J. Li, C.-C.M. Liu, K. Liu, J. Zhou, A mathematical theory of the topological vertex, math.AG/0408426.
[11] N. Caporaso, M. Cirafici, L. Griguolo, S. Pasquetti, D. Seminara, R.J. Szabo, Topological strings and large N phase
transitions I: Nonchiral expansion of q-deformed YangMills theory, JHEP 0601 (2006) 035, hep-th/0509041;
transitions II: Chiral expansion of q-deformed YangMills theory, JHEP 0601 (2006) 036, hep-th/0511043.
[12] K. Koike, On the decomposition of tensor products of the representation of the classical groups by means of the
universal characters, Adv. Math. 74 (1989) 5786.
[13] H.R. Morton, S.G. Lukac, The HOMFLY polynomial of the decorated Hopf link, J. Knot Theory Ramif. 12 (2003)
395416, math.GT/0108011.
[14] A. Okounkov, N. Reshetikhin, C. Vafa, Quantum CalabiYau and classical crystals, hep-th/0309208.
[15] S. Matsuura, K. Ohta, Localization on D-brane and gauge theory/matrix model, hep-th/0504176.
[16] I.G. Macdonald, Symmetric Functions and Hall Polynomials, second ed., Oxford Univ. Press, Oxford, 1995.
[17] A. Iqbal, A.-K. Kashani-Poor, The vertex on a strip, hep-th/0410174.
[18] J. Zhou, On a deformed topological vertex, math.AG/0504460.
[19] R. Gopakumar, C. Vafa, On the gauge theory/geometry correspondence, Adv. Theor. Math. Phys. 3 (1999) 1415
1443, hep-th/9811131.
[20] Y. Konishi, S. Minabe, Flop invariance of the topological vertex, math.AG/0601352.
On compactified harmonic/projective superspace,

5D superconformal theories, and all that
Sergei M. Kuzenko
School of Physics M013, The University of Western Australia, 35 Stirling Highway, Crawley WA 6009, Australia
Abstract
Within the supertwistor approach, we analyse the superconformal structure of 4D N = 2 compactified
harmonic/projective superspace. In the case of 5D superconformal symmetry, we derive the superconformal
Killing vectors and related building blocks which emerge in the transformation laws of primary superfields.
Various off-shell superconformal multiplets are presented both in 5D harmonic and projective superspaces,
including the so-called tropical (vector) multiplet and polar (hyper)multiplet. Families of superconformal
actions are described both in the 5D harmonic and projective superspace settings. We also present examples
of 5D superconformal theories with gauged central charge.
1. Introduction
According to Nahms classification [1], superconformal algebras exist in spacetime dimensions D 6. Among the dimensions included, the case of D = 5 is truly exceptional, for it
allows the existence of the unique superconformal algebra F (4) [2]. This is in drastic contrast
to the other dimensions which are known to be compatible with series of superconformal algebras (say, 4D N -extended or 6D (N , 0) superconformal symmetry). Even on formal grounds,
the exceptional feature of the five-dimensional case is interesting enough for studying in some
depth the properties of 5D superconformal theories. On the other hand, such rigid superconformal theories are important prerequisites in the construction, within the superconformal tensor
calculus [3,4], of 5D supergravity-matter dynamical systems which are of primary importance,
for example, in the context of bulk-plus-brane scenarios.
E-mail address: kuzenko@cyllene.uwa.edu.au (S.M. Kuzenko).
doi:10.1016/j.nuclphysb.2006.03.019
S.M. Kuzenko / Nuclear Physics B 745 [PM] (2006) 176207
177
The main motivation for the present work was the desire to develop a systematic setting to
build 5D superconformal theories, and clearly this is hardly possible without employing superspace techniques. The superconformal algebra in five dimensions includes 5D simple (or N = 1)
supersymmetry algebra1 as its super Poincar subalgebra. As is well known, for supersymmetric theories in various dimensions with eight supercharges (including the three important cases:
(i) 4D N = 2, (ii) 5D N = 1 and (iii) 6D N = (1, 0)) a powerful formalism to generate offshell formulations is the harmonic superspace approach that was originally developed for the 4D
N = 2 supersymmetric YangMills theories and supergravity [5,6]. There also exists a somewhat
different, but related, formalismthe so-called projective superspace approach [710], first introduced soon after the harmonic superspace had appeared. Developed originally for describing
the general self-couplings of 4D N = 2 tensor multiplets, this approach has been extended to include some other interesting multiplets. Both the harmonic and projective approaches make use
of the same superspace, RD|8 S 2 , which first emerged, for D = 4, in a work of Rosly [11] (see
also [12]) who built on earlier ideas due to Witten [13]. In harmonic superspace, one deals with
so-called Grassmann analytic superfields required to be smooth tensor fields on S 2 . In projective
superspace, one also deals with Grassmann analytic superfields required, however, to be holomorphic on an open subset of S 2 (typically, the latter is chosen to be C = C \ {0} in the Riemann
sphere realisation S 2 = C {}). In many respects, the harmonic and projective superspaces are
equivalent and complementary to each other [14], although harmonic superspace is obviously
more fundamental. Keeping in mind potential applications to the brane-world physics, the projective superspace setting seems to be more useful, since the 5D projective supermultiplets [15]
are easy to reduce to 4D N = 1 superfields.
To our knowledge, no comprehensive discussion of the superconformal group and superconformal multiplets in projective superspace has been given, apart from the analysis of SU(2)
invariance in [7] and the semi-component consideration of tensor multiplets in [16]. On the contrary, a realisation of the superconformal symmetry in 4D N = 2 harmonic superspace2 has been
known for almost twenty years [6,17]. But some nuances of this realisation still appear to be
quite mysterious (at least to newcomers) and call for a different interpretation. Specifically, one
deals with superfields depending on harmonic variables u
i subject to the two constraints
u+i u
i = 1,
u+i = u
i ,
i = 1, 2,
(1.1)
when describing general 4D N = 2 super-Poincar invariant theories in harmonic superspace [5,6]. In the case of superconformal theories, on the other hand, only the first constraint in
(1.1) has to be imposed [6,17]. Since any superconformal theory is, at the same time, a superPoincar invariant one, some consistency issues seem to arise, such as that of the functional
spaces used to describe the theory. It is quite remarkable that these issues simply do not occur if
one pursues the twistor approach to harmonic superspace [1921] (inspired by earlier constructions due to Manin [22]). In such a setting, the constraints (1.1) can be shown to appear only
as possible gauge conditions and therefore they have no intrinsic significance, with the only
structural condition being u+i u
i = 0. In our opinion, the supertwistor construction sketched
in [19,20] and further analysed in [21] is quite illuminating, for it allows a unified treatment
of the harmonic and projective superspace formalisms. That is why it is reviewed and developed further in the present paper. Unlike [19,21] and standard texts on Penroses twistor theory
1 On historic grounds, 5D simple supersymmetry is often labeled N = 2, see e.g. [4].
2 See [18] for an extension to six dimensions.
178
[23], see e.g. [24], we avoid considering compactified complexified Minkowski space and its
super-extensions, for these concepts are not relevant from the point of view of superconformal model-building we are interested in. Our 4D consideration is directly based on the use of
(conformally) compactified Minkowski space S 1 S 3 and its super-extensions. Compactified
Minkowski space is quite interesting in its own right (see, e.g. [25]), and its universal covering space (i) possesses a unique causal structure compatible with conformal symmetry [26], and
(ii) coincides with the boundary of five-dimensional anti-de Sitter space, which is crucial in the
context of the AdS/CFT duality [27].
In the case of 5D superconformal symmetry, one can also pursue a supertwistor approach.
However, since we are aiming at future applications to brane-world physics, a more pragmatic
course is chosen here, which is based on the introduction of the relevant superconformal Killing
vectors and elaborating associated building blocks. The concept of superconformal Killing vectors [21,2833], has proved to be extremely useful for various studies of superconformal theories
in four and six dimensions, see e.g. [3436].
This paper is organized as follows. In Section 2 we review the construction [37] of com 4 = S 1 S 3 as the set of null two-planes in the twistor space. In
pactified Minkowski space M
4|4N and introduce
Section 3 we discuss N -extended compactified Minkowski superspace M
the corresponding superconformal Killing vectors. In Section 4 we develop different aspects
of 4D N = 2 compactified harmonic/projective superspace. Section 5 is devoted to the 5D superconformal formalism. Here we introduce the 5D superconformal Killing vectors and related
building blocks, and also introduce several off-shell superconformal multiplets, both in the harmonic and projective superspaces. Section 6 introduces the zoo of 5D superconformal theories.
Three technical appendices are also included at the end of the paper. In Appendix A, a nonstandard realisation for S 2 is given. Appendix B is devoted to the projective superspace action
according to [8]. Some aspects of the reduction [15] from 5D projective supermultiplets to 4D
N = 1, 2 superfields are collected in Appendix C.
2. Compactified Minkowski space
4=
We start by recalling a remarkable realisation3 of compactified Minkowski space M
3
4
S as the set of null two-dimensional subspaces in the twistor space which is a copy
of C4 . The twistor space is defined to be equipped with the scalar product

12
0
,
T , S = T S, =
(2.1)
0 12
S1
for any twistors T , S C4 . By construction, this scalar product is invariant under the action of
the group SU(2, 2) to be identified with the conformal group. The elements of SU(2, 2) will be
represented by block matrices

A B
g=
SL(4, C), g g = ,
(2.2)
C D
3 This realisation is known in the physics literature since the early 1960s [25,26,37,38], and it can be related (see,
e.g. [26]) to the WeylDirac construction [39,40] of compactified Minkowski space S 1 S 3 /Z2 as the set of straight
lines through the origin of the cone in R4,2 . In the mathematics literature, its roots go back to Cartans classification of
the irreducible homogeneous bounded symmetric domains [41,42].

4 In the literature, the term twistor space is often used for CP 3 . In this paper we stick to the original Penrose terminonology [23].
179
where A, B, C and D are 2 2 matrices.

4 the space of null two-planes through the origin in C4 . Given a twoWe will denote by M
plane, it is generated by two linearly independent twistors T , with = 1, 2, such that

T , T = 0, , = 1, 2.
(2.3)
Obviously, the basis chosen, {T }, is defined only modulo the equivalence relation

T T , T = T R , R GL(2, C).
4 consists of 4 2 matrices of rank two,
Equivalently, the space M

F
, F F = G G,
T =
G
where F and G are 2 2 matrices defined modulo the equivalence relation

F
FR
, R GL(2, C).
GR
G
(2.4)
(2.5)
(2.6)
In order for the two twistors T in (2.5) to generate a two-plane, the 2 2 matrices F and G
must be non-singular,
det F = 0,
det G = 0.
(2.7)
Indeed, let us suppose the opposite. Then, the non-negative Hermitian matrix F F has a zero
eigenvalue. Applying an equivalence transformation of the form

F
FV
, V U (2),
G
GV
and therefore

F F V 1 F F V,

G G V 1 G G V,
we can arrive at the following situation

0 0
, R.
F F =G G=
0 2
In terms of the twistors T , the conditions obtained imply that T 1 = 0 and T 2 = 0. But this
contradicts the assumption that the two vectors T generate a two-plane.
Because of (2.7), we have

F
h
, h = F G1 U (2).
(2.8)
G
1
4 can be identified with the group manifold U (2) = S 1 S 3 .
It is seen that the space M
The conformal group acts by linear transformations on the twistor space: associated with
the group element (2.2) is the transformation T gT , for any twistor T C4 . This group
4 . It is defined as follows:
representation induces an action of SU(2, 2) on M
h g h = (Ah + B)(Ch + D)1 U (2).
(2.9)
4 is a homogeneous space of the group SU(2, 2), and therefore it

One can readily see that M
4
can be represented as M = SU(2, 2)/Hh0 , where Hh0 is the isotropy group at a fixed unitary
180
4 . With the choice

matrix h0 M
h0 = 1,
(2.10)
4 can be chosen as follows (see,

a coset representative s(h) SU(2, 2) that maps h0 to h M
e.g. [43]):

h 0
,
s(h) h0 = h U (2).
s(h) = (det h)1/4
(2.11)
0 1
The isotropy group corresponding to h0 consists of those SU(2, 2) group elements (2.2) which
obey the requirement
A + C = B + D.
(2.12)
This subgroup proves to be isomorphic to a group generated by the Lorentz transformations,

dilatations and special conformal transformations. To visualise this, it is useful to implement a
special similarity transformation for both the group SU(2, 2) and the twistor space.
We introduce a special 4 4 matrix ,

1 12 12
=
(2.13)
, = 14 ,
2 12 12
and associate with it the following similarity transformation:
g g = g 1 ,
g SU(2, 2);
T T = T ,
T C4 .
The elements of SU(2, 2) are now represented by block matrices

A B
g=
SL(4, C), g g = ,
C D
where
=

=
0
12
12
0
(2.14)
(2.15)

(2.16)
The 2 2 matrices in (2.15) are related to those in (2.2) as follows:

1
A = (A + D B C),
2
1
C = (A + C B D),
2
1
B = (A + B C D),
2
1
D = (A + B + C + D).
2
(2.17)
Now, by comparing these expressions with (2.12) it is seen that the stability group Hh0 1
consists of upper block-triangular matrices,
C = 0.
(2.18)
4,
M
transformation5
the effect of the similarity

When applied to

1 h1
1
h
h
,
=
ix
1
1
2 h+1
is
x = x m ( m )
,
5 We follow the two-component spinor notation of Wess and Bagger [44].
(2.19)
181
where
h+1
, x = x.
h1
The inverse expression for h in terms of x is given by the so-called Cayley transform:
ix =
1 ix
.
1 + ix
It is seen that
h =
h0 = 1
(2.20)
(2.21)
x0 = 0.
(2.22)
Unlike the original twistor representation, Eqs. (2.5) and (2.7), the 2 2 matrices h 1 in (2.19)
may be singular at some points. This means that the variables x m (2.20) are well-defined local
4 which is specified by det(h 1) = 0 and, as will become
coordinates in the open subset of M
clear soon, can be identified with the ordinary Minkowski space.
As follows from (2.19), in terms of the variables x m the conformal group acts by fractional
linear transformations
iB x)
1 .
ix ix = (C iD x)(A
(2.23)
These transformations can be brought to a more familiar form if one takes into account the
explicit structure of the elements of SU(2, 2):
1

2
ib
L
g=e , L=
(2.24)
, L = L.
ia
+ 12
,
Here the matrix elements correspond to a Lorentz transformation ( , ), translation a
special conformal transformation b and dilatation . In accordance with (2.18), the isotropy
group at x0 = 0 is spanned by the Lorentz transformations, special conformal boosts and scale
transformations.
3. Compactified Minkowski superspace

The construction reviewed in the previous section can be immediately generalised to the case
of N -extended conformal supersymmetry [22], by making use of the supertwistor space C4|N
introduced by Ferber [45], with N = 1, 2, 3 (the case N = 4 is known to be somewhat special,
and will not be discussed here). The supertwistor space is equipped with scalar product

12
0
T , S = T S, =
(3.1)
,
12
0
1N
for any supertwistors T , S C4|N . The N -extended superconformal group acting on the supertwistor space is SU(2, 2|N ). It is spanned by supermatrices of the form
g SL(4|N ),
g g = .
(3.2)
In complete analogy with the bosonic construction, compactified Minkowski superspace

4|4N is defined to be the space of null two-planes through the origin in C4|N . Given such
M
a two-plane, it is generated by two supertwistors T such that (i) their bodies are linearly inde 4|4N consists of rank-two
pendent; (ii) they obey Eqs. (2.3) and (2.4). Equivalently, the space M
182
supermatrices of the form

F

T = G , F F = G G + ,
(3.3)
defined modulo the equivalence relation

F
FR
G GR , R GL(2, C).
(3.4)
Here F and G are 2 2 bosonic matrices, and is a N 2 fermionic matrix. As in the bosonic
case, we have

h
F
(3.5)
G 1 , h h = 1 + .
Introduce the supermatrix

12 12
0
1
=
,
12 12 0
2 0
21N
0
= 14+N ,
(3.6)
and associate with it the following similarity transformation:

g g = g 1 ,
g SU(2, 2|N );
T T = T ,
T C4|N .
The supertwistor metric becomes

0 12
0
= 1 = 12 0
.
0
0 0 1N
(3.7)
(3.8)
4|4N , the similarity transformation results in

When implemented on the superspace M

1
h
h
h1
1
1 1 =
h
+ 1 ix+ = ix+
2
2
2
2i
(3.9)
where
ix+ =
h+1
,
h1
2 = (h 1)1 .
The bosonic x+ and fermionic variables obey the reality condition

x+ x = 4i , x = x+ .
(3.10)
(3.11)
It is solved by

x = x
2i i
i ,
i
= i ,
x = x,
(3.12)
with zA = (x a , i , i ) the coordinates of N -extended flat global superspace R4|4N . We therefore

a and which are the
see that the supertwistors in (3.9) are parametrized by the variables x+
i
coordinates in the chiral subspace. Since the superconformal group acts by linear transformations
183
on C4|2N , we can immediately conclude that it acts by holomorphic transformations on the chiral
subspace.
To describe the action of SU(2, 2|N ) on the chiral subspace, let us consider a generic group
element:

ib
2 j
,
+
2 j
g = eL , L = ia
(3.13)
2
j
j
2i
2 i
(
)
+
i
i
N
where
=

N
1
+i
,
2
N 4
= ,
tr = 0.
(3.14)
Here the matrix elements, which are not present in (2.24), correspond to a Q-supersymmetry
), S-supersymmetry (i , ), combined scale and chiral transformation , and chiral
(i , i
i
SU(N ) transformation i j . Now, one can check that the coordinates of the chiral subspace
transform as follows:
x+ = a + ( + )x+ x+ x+ + x+ bx+ + 4i 4x+ ,

1
(N 2) + 2 + + bx+ i x+ 4 .
= +
N
Expressions (3.15) can be rewritten in a more compact form,
a
x+
= +a (x+ , ),
i = i (x+ , ),
(3.15)
(3.16)
where
i
+a = a + i a i , a = a .
8
Here the parameters a and i are components of the superconformal Killing vector
= = a (z)a + i (z)Di + i (z)D i ,
(3.17)
(3.18)
which generates the infinitesimal transformation in the full superspace, zA zA + zA , and is

defined to satisfy

, D i D j ,
(3.19)
and therefore

D i = 4i i .
(3.20)
All information about the superconformal algebra is encoded in the superconformal Killing
vectors. From Eq. (3.20) it follows that

1
j
, Di = Di j D = Di
(N 2) + 2 Di j i Dj .
N
Here the parameters of local Lorentz and scale-chiral transformations are

1 i
1
1
(z) = D(
)i ,
(z) =
(N 2)Di i D i i
N
N (N 4) 2
(3.21)
(3.22)
184
and turn out to be chiral

D i
= 0.

= 0,
D i

(3.23)
The parameters j i

1 i k
i i
D
, = ,
,
D
tr = 0
D , Dj
32
N j
correspond to local SU(N ) transformations. One can readily check the identity

1 i k
k i
k i
D j = 2 j D j D .
N
j i (z) =
(3.24)
(3.25)
4. Compactified harmonic/projective superspace

For Ferbers supertwistors used in the previous section, a more appropriate name seems to be
even supertwistors. Being elements of C4|N , these objects have four bosonic components and
N fermionic components. One can also consider odd supertwistors [20]. By definition, these are
4 + N vector-columns such that their top four entries are fermionic, and the rest N components
are bosonic. In other words, the odd supertwistors are elements of CN |4 . It is natural to treat the
even and odd supertwistors as the even and odd elements, respectively, of a supervector space6
of dimension (4|N ) on which the superconformal group SU(2, 2|N ) acts. Both even and odd
supertwistors should be used [19,20] in order to define harmonic-like superspaces in extended
supersymmetry.
Throughout this section, our consideration is restricted to the case N = 2. Then, ij = ik k j
is symmetric, ij = j i , and Eq. (3.25) implies
D(i j k) = D j k) = 0.
(i
(4.1)
4.1. Projective realisation

Following [20], we accompany the two even null supertwistors T , which occur in the
4|8 , by an odd supertwistor with nonconstruction of the compactified N = 2 superspace M
vanishing body (in particular, the body of , is non-zero). These supertwistors are required
to obey

T , T = T , = 0, , = 1, 2,
(4.2)
and are defined modulo the equivalence relation

c
0
c 0
GL(1|2),
, T , T
,
R
R
(4.3)
4|8
with anticommuting complex parameters. The superspace obtained can be seen to be M
2
S . Indeed, using the above freedom in the definition of T and , we can choose them to be of
the form

h
0
T 1 ,
(4.4)
v , h h = 1 + , v = 0.
v
6 See, e.g. [32,46] for reviews on supervector spaces.
185
Here the non-zero two-vector v C2 is still defined modulo re-scalings v cv, with c C .
A natural name for the supermanifold obtained is projective superspace.
4.2. Harmonic realisation
4|8 S 2
Now, we would like to present a somewhat different, but equivalent, realisation for M
inspired by the exotic realisation for the two-sphere described in Appendix A. We will consider
a space of quadruples {T , + , } consisting of two even supertwistors T and two odd
supertwistors such that (i) the bodies of T are linearly independent four-vectors; (ii) the
bodies of are linearly independent two-vectors. These supertwistors are further required to
obey the relations
+
T , T = T , = T , = 0, , = 1, 2,
(4.5)
and are defined modulo the equivalence relation
0
+ + a
, ,T , ,T
b
c
+
0
0
with anticommuting complex parameters. Using the gauge

and , these supertwistors can be chosen to have the form
h
1
0
v
v
a
b
0
0
R

GL(2|2),
(4.6)
freedom in the definition of T
0
c
+
h h = 1 + ,

det v v + = 0.
(4.7)
Here the complex harmonics v are still defined modulo arbitrary transformations of the form
(A.9). Given a 2 2 matrix v = (v v + ) GL(2, C), there always exists a lower triangular matrix
R such that vR SU(2). The latter implies that v is uniquely determined in terms of v + , and
4|8 S 2 . In accordance with the
therefore the supermanifold under consideration is indeed M
construction given, a natural name for this supermanifold is harmonic superspace.
4.3. Embedding of R4|8 S 2 : Harmonic realisation
We can now analyse the structure of superconformal transformations on the flat global super 4|8 S 2 .
space R4|8 S 2 embedded in M
Upon implementing the similarity transformation, Eq. (3.7), we have

1
0
0

2 = 2
T ix+ = ix+ ,
(4.8)
2
u
u
2i
i
with

+
+i
det u
i ui = u ui = 0,
u+i = ij u+
j .
m and fermionic variables are related to each other by the reality condiHere the bosonic x+
i
tion (3.11). The orthogonality conditions T , = 0 imply
+
+ = i
ui ,

= i
ui .
(4.9)
186
The complex harmonic variables u

i in (4.8) are still defined modulo arbitrary transformations
of the form

+

a 0
+
ui ui u
(4.10)
R,
R
=
GL(2, C).
u
i i
b c
The gauge freedom (4.10) can be reduced by imposing the gauge condition
u+i u
i = 1.
(4.11)
It can be further reduced by choosing the harmonics to obey the reality condition
u+i = u
i .
(4.12)
Both requirements (4.11) and (4.12) have no fundamental significance, and represent themselves
possible gauge conditions only. It is worth pointing out that the reality condition (4.12) implies
, + = 0. If both equations (4.11) and (4.12) hold, then we have in addition + , + =
, = 1.
In what follows, the harmonics will be assumed to obey Eq. (4.11) only. As explained in the
appendix, the gauge freedom (4.10) allows one to represent any infinitesimal transformation of
the harmonics as follows:
u
i = 0,
++
u+
(u)u
i =
i ,
for some parameter ++ which is determined by the transformation under consideration.

In the case of an infinitesimal superconformal transformation (3.13), one derives
u
i = 0,
++ u ,
u+
i =
i
+
++ = ij u+
i uj ,
(4.13)
with the parameter ij given by (3.24). Due to (4.1), we have (using the notation D = Di u
i
and D = D i u )
++ = 0,
D+ ++ = D +

D ++ ++ = 0.
(4.14)
Here and below, we make use of the harmonic derivatives [5]

D ++ = u+i
,
ui
D = ui
.
u+i
(4.15)
It is not difficult to express ++ in terms of the parameters in (3.13) and superspace coordinates:

+ +
+
+ +
+ +
++ = ij u+
(4.16)
i uj + 4i b .
The transformation (4.13) coincides with the one originally given in [17].
For the superconformal variations of + and +
, one finds
+
i
i +
++ i u ,
+ = i u+
i + ui = ui
i
(4.17)
and similarly for +

. From Eqs. (3.21) and (4.14) one then deduces
D+ q+ = D + + = 0,
(4.18)
+
+
and similarly for +
. The superconformal variations q and can be seen to coincide
with those originally given in [17]. One can also check that the superconformal variation of the
187
analytic bosonic coordinates
y a = x a 2i (i a j ) u+
i uj ,
D+ y a = D + y a = 0,
is analytic. This actually follows from the transformation

+
D+ ++ D , D+ = D+ + ij u+
i uj D ,
(4.19)
(4.20)
and similarly for D +

.
We conclude that the analytic subspace parametrized by the variables

+
+
+
= y a , + , +
, ui , uj , D = D = 0,
is invariant under the superconformal group. The superconformal variations of these coordinates
coincide with those given in [17]. No consistency clash occurs between the SU(2)-type constraints (1.1) and the superconformal transformation law (4.13), because the construction does
not require imposing either of the constraints (1.1).
Using Eq. (3.25) one can show that the following descendant of the superconformal Killing
vector
+
= ij u+
i uj +
(4.21)
possesses the properties

D+ = D + = 0,
D ++ = ++ .
(4.22)
It turns out that the objects , ++ and determine the superconformal transformations of
primary analytic superfields [6].
4.4. Embedding of R4|8 S 2 : Projective realisation
Now, let us try to exploit the realisation of S 2 as the Riemann sphere CP 1 . The superspace
can be covered by two open setsthe north chart and the south chartthat are specified by the
conditions: (i) u+1 = 0; and (ii) u+2 = 0.
In the north chart, the gauge freedom (4.10) can be completely fixed by choosing
u+i (1, w) w i ,
u+
i (w, 1) = wi ,
ui (0, 1),
u
i (1, 0).
(4.23)
Here w is the complex coordinate parametrizing the north chart. Then the transformation
law (4.13) turns into
w = ++ (w),
++ (w) = ij wi+ wj+ .
(4.24)
It is seen that the superconformal group acts by holomorphic transformations.

The south chart is defined by
u+i (y, 1) y i ,
u+
i (1, y) = yi ,
ui (1, 0),
u
i (0, 1),
(4.25)
with y the local complex coordinate. The transformation law (4.13) becomes
y = ++ (y),
++ (y) = ij yi+ yj+ .
(4.26)
188
In the overlap of the north and south charts, the corresponding complex coordinates are related
to each other in the standard way:
y=
1
.
w
(4.27)
5. 5D superconformal formalism
As we have seen, modulo some global topological issues, all information about the superconformal structures in a superspace is encoded in the corresponding superconformal Killing
vectors. In developing the 5D superconformal formalism below, we will not pursue global aspects, and simply base our consideration upon elaborating the superconformal Killing vectors
and related concepts. Our 5D notation and conventions follow [15].
5.1. 5D superconformal Killing vectors
In 5D simple superspace R5|8 parametrized by coordinates zA = (x a , i ), we introduce an

infinitesimal coordinate transformation
zA zA = zA + zA
(5.1)
generated by a real vector field

= = a (z)a + i (z)Di ,
(5.2)
with DA = (a , Di ) the flat covariant derivatives. The transformation is said to be superconforj
mal if [, Di ] D , or more precisely

j
, Di = Di j D .
(5.3)
The latter equation is equivalent to

Di b = 2i b i = 2i b i .
(5.4)
It follows from here

j
ij (a ) a b = b D i + b Di j .
(5.5)
This equation implies that a = a (x, ) is an ordinary conformal Killing vector,

2
a b + b a = a b c c ,
5
depending parametrically on the Grassmann superspace coordinates,
a (x, ) = ba ( ) + 2 ()x a + a ( )x b + k a ( )x 2 2x a xb k b ( ),
b
(5.6)
(5.7)
with a b = a b .
From (5.4) one can derive a closed equation on the vector components = ( b ) b :
1 i
D(i )
= D (
)
.
5
(5.8)
189
One can also deduce closed equations on the spinor components i :

1
j)
(i j )
D = D (i ,
4
b

i = 0.
D i
(5.9)
(5.10)
+i
At this stage it is useful to let harmonics u
i , such that u ui = 0, enter the scene for the first
time. With the definitions D = Di ui and = i ui , Eq. (5.9) is equivalent to
1
D+
+ = D + +
D+
D + + = 0.

4
The above results lead to

j
, Di = D i Di ij D ,
(5.11)
(5.12)
where
1
1
1 (i
)
= Dk k ,
ij = D j ) .
= D k( k ,
2
8
4
The parameters on the right of (5.12) are related to each other as follows

Di = 2 Di + D i ,

j
Di j k = 3 ik D + ij Dk .
(5.13)
(5.14)
The superconformal transformation of the superspace integration measure involves

a a Di i = 2 .
(5.15)
5.2. Primary superfields

Here we give a few examples of 5D primary superfields, without Lorentz indices.
Consider a completely symmetric iso-tensor superfield H i1 ...in = H (i1 ...in ) with the superconformal transformation law
i2 ...in )k
1
H i1 ...in = H i1 ...in p H i1 ...in (i
,
k H
(5.16)
with p a constant parameter being equal to half the conformal weight of H i1 ...in . It turns out that
this parameter is equal to 3n if H i1 ...in is constrained by
(j
D H i1 ...in ) = 0
p = 3n.
(5.17)
The vector multiplet strength transforms as

W = W 2 W.
(5.18)
The conformal weight of W is uniquely fixed by the Bianchi identity

1
j)
j)
D(i D W = D (i D W
4
obeyed by W .
(5.19)
190
5.3. Analytic building blocks

In what follows we make use of the harmonics u
i subject to Eq. (4.11). As in the 4D N = 2
case, Eq. (4.11) has no intrinsic significance, with the only essential condition being (u+ u )
u+i u
i = 0. Eq. (4.11) is nevertheless handy, for it allows one to get rid of numerous annoying
factors of (u+ u ).
Introduce
= ij u+
,
i uj + 3
+
++ = D ++ = ij u+
i uj .
(5.20)
] = 0, that and ++ are analytic superfields,

It follows from (5.14) and the identity [D ++ , D+
= 0,
D+
D+
++ = 0.
(5.21)
Representing = a a + D
+ D+
, one can now check that

++ D , D + = D + ( 2 )D + .
(5.22)
This relation implies that the operator ++ D maps every analytic superfield into an analytic one. It is worth pointing out that the superconformal transformation of the analytic subspace
measure involves
a a + D
+ D ++ = 2.
(5.23)
5.4. Harmonic superconformal multiplets

We present here several superconformal multiplets that are globally defined over the harmonic superspace. Such a multiplet is described by a smooth Grassmann analytic superfields
(n)
(z, u+ , u ),
(n) = 0,
D+

which is endowed with the following superconformal transformation law

(n) = ++ D (n) (n) .
(5.24)
(5.25)
The parameter is related to the conformal weight of (n) . We will call (n) an analytic density
(n)
(n)
of weight . When n is even, one can define real superfields, = , with respect to the
analyticity-preserving conjugation [5,6] (also known as smile-conjugation).
Let V ++ be a real analytic gauge potential describing a U (1) vector multiplet. Its superconformal transformation is

V ++ = ++ D V ++ .
(5.26)
Associated with the gauge potential is the field strength

2
2
i
W=
D
= D D
du D V ++ ,
(5.27)
which is known to be invariant under the gauge transformation V ++ = D ++ , where the gauge
parameter is a real analytic superfield. The superconformal transformation of W ,

2

i
du D + D ++ V ++ ,
W =
(5.28)
8
191
can be shown to coincide with (5.18).

There are many ways to describe a hypermultiplet. In particular, one can use an analytic
superfield q + (z, u) and its smile-conjugate q + (z, u) [5,6]. They transform as follows:

q + = ++ D q + q + . (5.29)
q + = ++ D q + q + ,
One has = n in (5.25), if the superfield is annihilated by D ++ ,
+
H (n) = D ++ H (n) = 0 H (n) (z, u) = H i1 ...in (z)u+
D+
i1 uin ,

H (n) = ++ D H (n) nH (n) .
(5.30)
Here the harmonic-independent superfield H i1 ...in transforms according to (5.16) with p = 3n.
5.5. Projective superconformal multiplets
In the projective superspace approach, one deals only with superfields (n) (z, u+ ) obeying
the constraints
D+
(n) = D ++ (n) = 0,
n 0.
(5.31)
Here the first constraint means that (n) is Grassmann analytic, while the second constraint demands independence of u . Unlike the harmonic superspace approach, however, (n) (z, u+ ) is
not required to be well-defined over the two-sphere, that is, (n) may have singularities (say,
poles) at some points of S 2 . The presence of singularities turns out to be perfectly OK since the
projective-superspace action involves a contour integral in S 2 , see below.
We assume that (p) (z, u) is non-singular outside the north and south poles of S 2 . In the north
chart, we can represent
= u+1 (w),
D+
(w) = Di wi ,
wi = (w, 1).
(5.32)
Then, Eq. (5.31) are equivalent to

(z, w) =
+

n (z)w n ,
(w)(z, w) = 0,
(5.33)
n=
with the holomorphic superfield (z, w) (n) (z, u+ ). These relations define a projective multiplet, following the four-dimensional terminology [9]. Associated with (z, w) is its smileconjugate
w) =
(z,
+

(1)n n (z)w n ,
w) = 0,
(w)(z,
(5.34)
n=
w) = (z, w), the projective superfield is called real.

which is also a projective multiplet. If (z,
Below we present several superconformal multiplets as defined in the north chart. The corresponding transformations laws involve the two analytic building blocks:
++ (w) = ij wi+ wj+ = 11 w 2 2 12 w + 22 ,
(w) = 1i wi = 11 w + 12 .
Similar structures occur in the south chart, that is

++ (y) = ij yi+ yj+ = 11 2 12 y + 22 y 2 ,
(y) = 2i yi = 12 + 22 y.
192
In the overlap of the two charts, we have

1 ++
(w) ++ (y)y = ++ (w)w ,
w2
1
(y) = (w) + ++ (w).
(5.35)
w
To realise a massless vector multiplet, one uses the so-called tropical multiplet described by
++ (y) =
V (z, w) =
+

Vn (z)w n ,
Vn = (1)n Vn .
(5.36)
n=
Its superconformal transformation

V = + ++ (w)w V .
(5.37)
The field strength of the vector multiplet7 is

2
dw
1
1
dw D V (z, w) =
P(w)V (z, w),
W (z) =
(5.38)
16i
4i
w
where
w 1 2
1 2
(D1 ) + 5
D .
P(w) =
(5.39)
4w
4
The superconformal transformation of W can be shown to coincide with (5.18). The field
strength (5.38) is invariant under the gauge transformation

w) (z, w) ,
V (z, w) = i (z,
(5.40)
with (z, w) an arbitrary arctic multiplet, see below.
To describe a massless off-shell hypermultiplet, one can use the so-called arctic multiplet
(z, w):
q + (z, u) = u+1 (z, w) (z, w),
(z, w) =
n (z)w n .
(5.41)
n=0
The smile-conjugation of q + leads to the so-called the antarctic multiplet (z, w):
q + (z, u) = u+2 (z, w) w (z, w),
(z, w) =

1
(1)n n (z) n .
w
(5.42)
n=0
Their superconformal transformations are

= + ++ (w)w (w),

1
= + ++ (w)w (w ) (w) .
w
In the south chart, these transformations take the form

1
= ++ (y)y (y ) (y),
y
7 A more general form for the field strength (5.38) is given in Appendix B.
(5.43)

= ++ (y)y (y) .
193
(5.44)
Both (z, w) and (z, w) constitute the so-called polar multiplet.

Since the product of two arctic superfields is again arctic, from (5.43) we obtain more general
transformation laws

= + ++ (w)w (w) ,

1
= + ++ (w)w w (w) ,
(5.45)
w
for some parameter . The case = 1 corresponds to free hypermultiplet dynamics, see below.
Since the product U = is a tropical multiplet, we obtain more general transformation
laws than the one defined by Eq. (5.37):

1
U = + ++ (w)w w U 2(w)U .
(5.46)
w
Finally, let us consider the projective-superspace reformulation of the multiplets (5.30) with an
even superscript,
n

H (2n) (z, u) = iu+1 u+2 H [2n] (z, w) (iw)n H [2n] (z, w),
H [2n] (z, w) =
n

Hk (z)w n ,
H k = (1)k Hk .
(5.47)
k=n
The projective superfield H [2n] (z, w) is often called a real O(2n) multiplet [9]. Its superconformal transformation in the north chart is

1
H [2n] = n + ++ (w)w w n H [2n] 2n(w)H [2n] .
(5.48)
w
In a similar way one can introduce complex O(2n + 1) multiplets. In what follows, we will use
the same name O(n) multiplet for both harmonic multiplets (5.30) and the projective ones just
introduced.
Among the projective superconformal multiplets considered, it is only the O(n) multiplets
which can be lifted to well-defined representations of the superconformal group on a compactified 5D harmonic superspace. The other multiplets realise the superconformal algebra only.
6. 5D superconformal theories
With the tools developed, we are prepared to constructing 5D superconformal theories. Superfield formulations for 5D N = 1 rigid supersymmetric theories were earlier elaborated in the
harmonic [15,47] and projective [15] superspace settings.8
6.1. Models in harmonic superspace
Let L(+4) be an analytic density of weight +2. Its superconformal transformation is a total
derivative,

L(+4) = ++ D L(+4) 2L(+4)
8 In the case of 6D N = (1, 0) rigid supersymmetric theories, superfield formulations have been developed in the
conventional [48], harmonic [49] and projective [50] superspace settings.
194

+ (+4)

L
+ D ++ L(+4) .
= a a L(+4) D
Therefore, such a superfield generates a superconformal invariant of the form

d (4) L(+4) ,
where

du
(6.2)
1 2 2
D
.
D
(6.3)
32
This is the harmonic superspace action [6] as applied to the five-dimensional case.
Let V ++ be the gauge potential of an Abelian vector multiplet. Given a real O(2) multiplet L++ ,

D+
(6.4)
L++ = D ++ L++ = 0,
L++ = ++ D L++ 2L++ ,
d (4) :=
4
d5 x D ,
(6.1)
we can generate the following superconformal invariant

d (4) V ++ L++ .
(6.5)
Because of the constraint D ++ L++ = 0, the integral is invariant under the vector multiplet gauge
transformation V ++ = D ++ , with a real analytic gauge parameter.
The field strength of the vector multiplet, W , is a primary superfield with the transformation (5.18). Using W , one can construct the following analytic superfield [15]
1 2
iG++ = D + W D+
W + W D + W, D+
G++ = D ++ G++ = 0,
2
which transforms as a harmonic superfield weight 2,

G++ = ++ D G++ 2G++ .
(6.6)
(6.7)
In other words, G++ is a real O(2) multiplet. As a result, the supersymmetric ChernSimons
action9 [15]

1
d (4) V ++ G++
SCS V ++ =
(6.8)
12
is superconformally invariant.
Super-ChernSimons theory (6.8) is quite remarkable as compared with the superconformal
models of a single vector multiplet in four and six dimensions. In the 4D N = 2 case, the ana + W , with W the chiral field strength,
logue of G++ in (6.8) is known to be D + D+ W = D +
D
and therefore the model is free. In the case 6D N = (1, 0), the analogue of G++ in (6.8) is
(D + )4 D
W , see [18] for more details, and therefore the models is not only free but also has
higher derivatives. It is only in five dimensions that the requirement of superconformal invariance
leads to a nontrivial dynamical system.
The model (6.8) admits interesting generalisations. In particular, give several Abelian vector
multiplets VI++ , where I = 1, . . . , n, the composite superfield (6.6) is generalised as follows:

+ 2
1
++
++
+
++
+
WJ ) ,
G GI J = G(I J ) = i D WI D WJ + W(I D
2
9 A different form for this action was given in [47].
D+
G++ = D ++ G++
I J = 0.
IJ
195
(6.9)
The gauge-invariant and superconformal action (6.8) turns into

1
d (4) VI++ cI,J K G++
SCS =
J K , cI,J K = cI,KJ ,
12
(6.10)
for some constant parameters cI,J K . One can also generalise the super-ChernSimons theory
(6.8) to the non-Abelian case.
In harmonic superspace, some superconformal transformation laws are effectively independent (if properly understood) of the dimension of spacetime. As a result, some 4D N = 2
superconformal theories can be trivially extended to five dimensions. In particular, the model for
a massless U (1) charged hypermultiplet [5]

Shyper = d (4) q + D ++ + ieV ++ q +
(6.11)
can be seen to be superconformal. This follows from Eqs. (5.26) and (5.29), in conjunction with
the observation that the transformation laws of q + and D ++ q + are identical.
The dynamical system SCS + Shyper can be chosen to describe the supergravity compensator
sector (vector multiplet plus hypermultiplet) when describing 5D simple supergravity within
the superconformal tensor calculus [3,4]. Then, the hypermultiplet charge e is equivalent to the
presence of a non-vanishing cosmological constant, similar to the 4D N = 2 case [6].
Our next example is a naive 5D generalisation of the 4D N = 2 improved tensor multiplet
[7,51,52] which was described in the harmonic superspace approach in [6,53]. Let us consider
the action

Stensor H ++ = d (4) L(+4) H ++ , u ,
(6.12)
where

L(+4) H ++ , u = 3
H++
1 + 1 + H++ c
2
,
H++ = H ++ c++ ,
(6.13)
with a constant parameter of unit mass dimension, and c++ a (spacetime) independent holomorphic vector field on S 2 ,
c (u) = cij u
i uj ,
cij cij = 2,
cij = const.
(6.14)
Here H ++ (z, u) is a real O(2) multiplet possessing the superconformal transformation

law (5.30) with n = 2. The superconformal invariance of (6.12) can be proved in complete
analogy to the detailed consideration given [6,53].
Now, let us couple the vector multiplet to the real O(2) multiplet by putting forward the action

Svectortensor V ++ , H ++ = SCS V ++ + d (4) V ++ H ++ + Stensor H ++ , (6.15)
with a coupling constant. This action is both gauge-invariant and superconformal. It is a fivedimensional generalisation of the 4D N = 2 model for massive tensor multiplet introduced
in [54].
The dynamical system Svectortensor can be chosen to describe the supergravity compensator
sector (vector multiplet plus tensor multiplet) when describing 5D simple supergravity within the
196
superconformal tensor calculus [3,4]. Then, the coupling constant is equivalent to a cosmological constant, similar to the 4D N = 2 case [10].
Finally, consider the vector multiplet model

SCS V ++ + Stensor G++ /3 ,
(6.16)
with G++ the composite superfield (6.6). The second term here turns out to be a unique superconformal extension of the F 4 -term, where F is the field strength of the component gauge field.
In this respect, it is instructive to recall its 4D N = 2 analogue [55]

d4 x d8 ln W ln W .
(6.17)
The latter can be shown
[56] to be a unique N = 2 superconformal invariant in the family of

actions of the form d4 x d8 H (W, W ) introduced for the first time
in [57]. In five spacetime
dimensions, if one looks for a superconformal invariant of the form d5 x d8 H (W ), the general
solution is H (W ) W , as follows from (5.15) and (5.18), and this choice corresponds to a total
derivative.
6.2. Models in projective superspace
Let L(z, w) be an analytic superfield transforming according to (5.46) with = 1. This transformation law can be rewritten as

wL = + ++ w (wL) 2wL

+

wL w ++ wL .
= a a wL D
(6.18)
Such a superfield turns out to generate a superconformal invariant of the form

4
dw
d5 x D wL(z, w),
I=
(6.19)
2i

where dw is a (model-dependent) contour integral in CP 1 . Indeed, it follows from (6.18) that
this functional does not change under the superconformal transformations. Eq. (6.19) generalises
the projective superspace action [7,8] to the five-dimensional case. A more general form for
this action, which does not imply the projective gauge conditions (4.23) and is based on the
construction in [8], is given in Appendix B.
It is possible to bring the action (6.19) to a somewhat simpler form if one exploits the fact
that L is Grassmann analytic. Using the considerations outlined in Appendix C gives

4
1
1 1 2 2
d5 x D L = 2 d5 x D 4 L, D 4 =
(6.20)
D (D1 ) .
16
w

Here D 4 is the Grassmann part of the integration measure of 4D N = 1 superspace, d4 = D 4 .
Then, functional (6.19) turns into

dw
dw
d5 x D 4 L =
d5 x d4 L.
I=
(6.21)
2iw
2iw
Our first example is the tropical multiplet formulation for the super-ChernSimons theory
[15]

dw
1
d5 x d4 V G,
SCS =
(6.22)
12
2iw
197
with the contour around the origin. Here G(w) is the composite O(2) multiplet (6.6) constructed
out of the tropical gauge potential V (w),

G++ = iu+1 u+2 G(w) iwG(w),
1
(6.23)
+ K + w .
w
The explicit expressions for the superfields and K can be found in [15]. The above consideration and the transformation laws (5.37) and (6.7) imply that the action (6.22) is superconformal.
Next, let us generalise to five dimensions the charged -hypermultiplet model of [9]:

dw
Shyper =
(6.24)
d5 x d4 eqV ,
2iw
G(w) =
with q the hypermultiplet charge, and the integration contour around the origin. This action is
superconformal, in accordance with the transformation laws (5.37) and (5.43). It is also invariant
under gauge transformations
= iq,
V = i( ),
(6.25)
with an arctic superfield.

Now, let us couple the vector multiplet to a real O(2) multiplet H (w)

H ++ = iu+1 u+2 H (w) iwH (w),
H (w) =
+ L + w .
w
We introduce the vectortensor system

dw
1
dw
d 5 x d4 V
G + H + 3
d5 x d4 H ln H,
S=
2iw
12
2iw
(6.26)
(6.27)
where the first term on the right involves a contour around the origin, while the second comes
with a contour turning clockwise and anticlockwise around the roots of the quadratic equation
wH (w) = 0. The second term in (6.27) is a minimal 5D extension of the 4D N = 2 improved
tensor multiplet [7]. It should be pointed out that the component superfields in (6.26) obey the
constraints [15]
1
(6.28)
D 2 L = 5 .
4
It should be also remarked that the real linear superfield L can always be dualised into a chiral
scalar and its conjugate [15], which generates a special chiral superpotential.
Given several O(2) multiplets H I (w), where I = 1, . . . , n, superconformal dynamics is generated by the action

dw
d5 x d4 F H I , I = 1, . . . , n,
S=
(6.29)
2iw
D = 0,
where F (H ) is a weakly homogeneous function of first degree in the variables H ,

dw
5
4
I F (H )
d xd H
F(H ) = 0.
2iw
H I
(6.30)
This is completely analogous to the four-dimensional case [7,10,16] where the component structure of such sigma-models has been studied in detail [60].
198
A great many superconformal models can be obtained if one considers -hypermultiplet actions of the form

dw

d5 x d4 K I , J , I, J = 1, . . . , n,
S=
(6.31)
2iw
with the contour around the origin. Let us first assume that the superconformal transformations
of all s and s have the form (5.43). Then, in accordance with general principles, the action is
superconformal if K(, ) is a weakly homogeneous function of first degree in the variables ,

dw
5
4
I K(, )
K(, ) = 0.
d xd
(6.32)
2iw
I
This homogeneity condition is compatible with the Khler invariance
),
K(, ) K(, ) + ( ) + (
(6.33)
which the model (6.31) possesses [14,15,58].

Unlike the O(n) multiplets, the superconformal transformations of and are not fixed
uniquely by the constraints, as directly follows from (5.45). Therefore, one can consider superconformal sigma-models of the form (6.31) in which the dynamical variables s consist of
several subsets with different values for the weight in (5.45), and then K(, ) should obey
weaker conditions than Eq. (6.32). Such a situation occurs, for instance, if one starts with a
gauged linear sigma-model and then integrates out the gauge multiplet, in the spirit of [16,52].
As an example, consider

dw
d5 x d4 eV + eV ,
S=
(6.34)
2iw
where and are constant diagonal metrics, = 1, . . . , m and = 1, . . . , n. Integrating
out the tropical multiplet gives the gauge-invariant action

dw
d5 x d4 .
S=2
(6.35)
2iw
The gauge freedom can be completely fixed by setting, say, one of the superfields to be unity.
Then, the action becomes

dw
d5 x d4 nn + ,
S=2
(6.36)
2iw
where , = 1, . . . , n 1. This action is still superconformal, but now and transform
according to (5.45) with = 2 and = 0, respectively.
is the Khler
Sigma-models (6.31) have an interesting geometric interpretation if K(, )
potential
of
a
Khler
manifold
M
[14,15,58].
Among
the
component
superfields
of (z, w) =

n , the leading components = | and = | considered as 4D N = 1 super
(z)w
n
0
1
n=0
fields, are constrained:
1
(6.37)
D 2 = 5 .
4
The and can be regarded as a complex coordinate of the Khler manifold and a tangent
vector at point of the same manifold, and therefore they parametrize the tangent bundle T M of
the Khler manifold. The other components, 2 , 3 , . . . , are complex unconstrained superfields.
These superfields are auxiliary since they appear in the action without derivatives. The auxiliary
D = 0,
199
superfields 2 , 3 , . . . , and their conjugates, can be eliminated with the aid of the corresponding
algebraic equations of motion

K(, )
= 0, n 2.
dw wn1
(6.38)
I
Their elimination can be carried out using the ansatz
nI =

p=o
UJI
1 ...L p
1 ...Jn+p L
J1 Jn+p L1 Lp ,
(, )
n 2.
(6.39)
It can be shown that the coefficient functions U s are uniquely determined by equations (6.38)
in perturbation theory. Upon elimination of the auxiliary superfields, the action (6.31) takes the
form

gI J (, )
I J
d5 x d4 K(, )
, ] =
S[, ,
RI1 ...Ip J1 ...Jp (, )
I1
Ip
J1
Jp
(6.40)
p=2
and its cowhere the tensors RI1 ...Ip J1 ...Jp are functions of the Riemann curvature RI JK L (, )
variant derivatives. Each term in the action contains equal powers of and , since the original
model (6.31) is invariant under rigid U (1) transformations

n (z) ein n (z).
(w) ei w
(6.41)
The complex linear superfields I can be dualised into chiral superfields10 I which can be
interpreted as a one-form at the point M [15,58]. Upon elimination of and , the action
, ]. Its target space is (an open neighborhood of the zero section) of the
turns into S[, ,
cotangent bundle T M of the Khler manifold M. Since supersymmetry requires this target
space to be hyper-Khler, our consideration is in accord with recent mathematical results [61]
about the existence of hyper-Khler structures on cotangent bundles of Khler manifolds.
6.3. Models with intrinsic central charge
We have so far considered only superconformal multiplets without central charge. As is
known, there is no clash between superconformal symmetry and the presence of a central charge
provided the latter is gauged. Here we sketch a 5D superspace setting for supersymmetric theories
with gauged central charge, which is a natural generalisation of the 4D N = 2 formulation [59].
To start with, one introduces an Abelian vector multiplet, which is destined to gauge the
central charge , by defining gauge-covariant derivatives

DA = Da , Di = DA + VA (z), [, DA ] = 0.
(6.42)
Here the gauge connection VA is inert under the central charge transformations, [, VA ] = 0.
The gauge-covariant derivatives are required to obey the algebra
i

i
j
D , = 0,
D , D = 2iij (D + W),
10 This is accompanied by the appearance of a special chiral superpotential [15].
200

Di , D = ia Di W + 2i Di Di W,
(6.43)
where the real field strength W(z) obeys the Bianchi identity (5.19). The field strength should
possess a non-vanishing expectation value, W = 0, corresponding to the case of rigid central
charge. By applying a harmonic-dependent gauge transformation, one can choose a frame in
which
D+
,
D+
D ++ D ++ + V ++ ,
D D + V ,
(6.44)
with V ++ a real analytic prepotential, see [59] for more details. This frame is called the -frame,
and the original representation is known as the -frame [5].
To generate a supersymmetric action, it is sufficient to construct a real superfield L(ij ) (z) with
the properties
(i
D Lj k) = 0,
which for
(6.45)
+
L++ (z, u) = Lij (z)u+
i uj
L++ = 0,
D+
take the form
D ++ L++ = 0.
(6.46)
In the -frame, the latter properties become

++

D+
L ++ = 0,
D + V ++ L ++ = 0.
(6.47)
Associated with L ++ is the supersymmetric action

d (4) V ++ L ++
(6.48)
which invariant under the central charge gauge transformations V ++ = D ++ and L ++ =

L ++ , with an arbitrary analytic parameter .
Let us give a few examples of off-shell supermultiplets with intrinsic central charge. The
simplest is the 5D extension of the FayetSohnius hypermultiplet. It is described by an iso-spinor
superfield q i (z) and its conjugate q i (z) subject to the constraint
(i
D q j ) = 0.
(6.49)
This multiplet becomes on-shell if = const. With the notation q + (z, u) = q j (z)u+
i , the hypermultiplet dynamics is dictated by the Lagrangian
1 + +
q imq + q + ,
L++
(6.50)
FS = 2 q
with m the hypermultiplet mass/charge. This Lagrangian generates a superconformal theory.
Our second example is an off-shell gauge two-form multiplet called in [3] the massless tensor
multiplet. It is Poincar dual to the 5D vector multiplet. Similarly to the 4D N = 2 vectortensor
multipet [59], it is described by a constrained real superfield L(z) coupled to the central charge
vector multiplet. By analogy with the four-dimensional case [59], admissible constraints must
obey some nontrivial consistency conditions. In particular, the harmonic-independence of L (in
the -frame) implies
2 2
0 = D + D + D L
2 2
+ 2
= D D + D + L 4D D+
D
L + 8iD D+
D+
2
2

8i L D + W + W D + L + 4D+ WD+
L .
201
(6.51)
Let us assume that L obeys the constraint

2
1
D+ L = D + L D+
D+ D+ L = 0
D+
(6.52)

4
which in the case = 0 coincides with the Bianchi identity for an Abelian vector multiplet.
Then, Eq. (6.51) gives
2

2
L = 0.
L D + W + W D + L + 4D+ WD+
(6.53)
The consistency condition is satisfied if L is constrained as

2

1 + 2
W + 4D+ WD+
L .
L D
D + L =
W
The corresponding Lagrangian is

i
1 + 2
+
++
+
L = D LD L + L D
L .
4
2

(6.54)
(6.55)
The theory generated by this Lagrangian is superconformal.

Another solution to (6.53) describes a ChernSimons coupling of the two-form multiplet to
an external YangMills supermultiplets:

+ 2
1 + 2
++
L=
W + 4D+ WD+
L +
L D
G ,
D
W
W
where

1 + 2
iG++ = tr D+ WD+
W
+
W
.
W,
D
(6.56)
(6.57)
Here is a coupling constant, and W is the gauge-covariant field strength of the YangMills
supermultiplet, see [15] for more details. As the corresponding supersymmetric Lagrangian one
can again choose L++ given by Eq. (6.55).
A plain-dimensional reduction 5D 4D can be shown to reduce the constraints (6.52) and
(6.56) to those describing the so-called linear vectortensor multiplet11 with ChernSimons couplings [59].
When this paper was ready for submission to the hep-th archive, there appeared an interesting
work [62] in which 4D and 5D supersymmetric nonlinear sigma models with eight supercharges
were formulated in N = 1 superspace.
Acknowledgements
It is a pleasure to thank Ian McArthur for reading the manuscript. The author is grateful to
the Max Planck Institute for Gravitational Physics (Albert Einstein Institute) in Golm and the
Institute for Theoretical Physics at the University of Heidelberg for hospitality during the course
of the work. This work is supported by the Australian Research Council and by a UWA research
grant.
11 Ref. [59] contains an extensive list of publications on the linear and nonlinear vectortensor multiplets and their
couplings.
202
Appendix A. Non-standard realisation for S 2

Let us consider a quantum-mechanical spin-1/2 Hilbert space, i.e. the complex space C2
endowed with the standard positive definite scalar product | defined by

u1
i
|u = (ui ) =
,
u| = u i , u i = ui .
u|v = u v = u vi ,
(A.1)
u2
Two-sphere S 2 can be identified with the space of rays in C2 . A ray is represented by a normalized
state,

+i
u = u ,
(A.2)
u u = 1,
u = u , u+i = u
i
i ,
defined modulo the equivalence relation

i
u
ui , ei = 1.
i e
Associated with
+ +
u = u ,
i
|u
(A.3)
|u+ ,
is another normalized state

+ +
+j
u+
u u = 1,
i = ij u ,
(A.4)
which is orthogonal to |u ,
+
u u = 0.
(A.5)
The states |u and |u+ generate the unimodular unitary matrix

+
u = u , u+ = u
i , ui SU(2).
(A.6)
In terms of this matrix, the equivalence relation (A.3) becomes

i

e
0
uu
.
0
ei
(A.7)
This gives the well-known realisation S 2 = SU(2)/U (1).

The above unitary realisation for S 2 is ideal if one is interested in the action of SU(2), or
its subgroups, on the two-sphere. But it is hardly convenient if one considers, for instance, the
action of SL(2, C) on S 2 . There exists, however, a universal realisation. Instead of dealing with
the orthonormal basis (|u , |u+ ) introduced, one can work with an arbitrary basis for C2 :

v = v , v + = vi , vi+ GL(2, C), det v = v +i vi .
(A.8)
The two-sphere is then obtained by factorisation with respect to the equivalence relation

a 0
v vR, R =
GL(2, C).
b c
(A.9)
Given an arbitrary matrix v GL(2, C), there always exists a lower triangular matrix R such
that vR SU(2), and then we are back to the unitary realisation. One can also consider an
intermediate realisation for S 2 given in terms of unimodular matrices of the form

w = w , w + = wi , wi+ SL(2, C) w +i wi = 1,
(A.10)
and the matrix R in (A.9) should be restricted to be unimodular. The harmonics w are complex
in the sense that wi and w +i are not related by complex conjugation.
203
Let us consider a left group transformation acting on S 2

w gw = vi , vi+ v.
(A.11)
If g is a small transformation, i.e. if it belongs to a sufficiently small neighbourhood of the

identity, then there exists a matrix R of the type (A.9) such that

gwR = wi , w i+ SL(2, C).
(A.12)
Since
w +i wi = 1,
w +i wi = 1,
for the transformed harmonic we thus obtain

w i+ = wi+ + ++ (w)wi .
(A.13)
All information about the group transformation g is now encoded in ++ .

Appendix B. Projective superspace action
Following [8], consider
+ +i
4 ++ +
ui d u
1
5
d
I=
z, u ,
x
D
L
2i
(u+ u )4
(B.1)
where the Lagrangian enjoys the properties

D+
L++ z, cu+ = c2 L++ z, u+ ,
L++ z, u+ = 0,
c C .
(B.2)
The functional (B.1) is invariant under arbitrary projective transformations (4.10). Choosing the
projective gauge (4.23) gives the supersymmetric action (6.19).
It is worth pointing out that the vector multiplet field strength (5.38) can be rewritten in the
projective-invariant form
+ +i
ui du 2 +
1
W (z) =
(B.3)
D
V z, u ,
16i
(u+ u )2
where the gauge potential enjoys the properties

D+
V z, cu+ = V z, u+ ,
V z, u+ = 0,
c C .
(B.4)
Appendix C. From 5D projective supermultiplets to 4D N = 1, 2 superfields
The conventional 5D simple superspace R5|8 is parametrized by coordinates zA = (x a , i ),

with i = 1, 2. Any hypersurface x 5 = const in R5|8 can be identified with the 4D, N = 2 super . The Grassmann coordinates
space R4|8 parametrized by zA = (x a , i , i ), where (i ) = i
5|8
4|8
of R and R are related to each other as follows:
i

i
i = i , i
(C.1)
,
=
.
204
Interpreting x 5 as a central charge variable, one can view R5|8 as a 4D, N = 2 central charge
superspace. One can relate the 5D spinor covariant derivatives (see [15] for more details)
i

Di = i
(C.2)
= Di = i b i b , Di = Di , D i
D
i
to the 4D, N = 2 covariant derivatives DA = (a , Di , D i ) where
Di =
+ i b i b + i 5 ,

D i
ii b b i
=
5 .
i
These operators obey the anti-commutation relations

i
j
= 2ij 5 ,
D i
D , D = 2 ij 5 ,
,D

j

c
i
i
= 2ij c ,
D , Dj
(C.3)
(C.4)
which correspond to the 4D, N = 2 supersymmetry algebra with the central charge 5 .
Consider a 5D projective superfield (5.33). Representing the differential operators (w),
Eq. (5.32), as

(w)
1
2
, (w) wD D , (w) D 1 + w D 2 ,
(w) =
(C.5)
(w)
the constraints (5.34) can be rewritten in the component form
2
D 2 n = D 1 n+1 .
D n = D n1 ,
(C.6)
The relations (C.6) imply that the dependence of the component superfields n on 2 and is
1
uniquely determined in terms of their dependence on 1 and . In other words, the projective
superfields depend effectively on half the Grassmann variables which can be chosen to be the
spinor coordinates of 4D N = 1 superspace
2
= 1 ,
1
= .
(C.7)
2
Then, one deals with reduced superfields |, D |, D 2 |, . . . (of which not all are usually independent) and 4D N = 1 spinor covariant derivatives D and D defined in the obvious way:

1
| = x, i , i = 2 =0 ,
(C.8)
D = D = 2 =0 ,
D = D 1 = 2 =0 .
2
References
[1] W. Nahm, Supersymmetries and their representations, Nucl. Phys. B 135 (1978) 149.
[2] V.G. Kac, Lie superalgebras, Adv. Math. 26 (1977) 8.
[3] T. Kugo, K. Ohashi, Off-shell d = 5 supergravity coupled to matterYangMills system, Prog. Theor. Phys. 105
(2001) 323, hep-ph/0010288;
T. Fujita, K. Ohashi, Superconformal tensor calculus in five dimensions, Prog. Theor. Phys. 106 (2001) 221, hepth/0104130;
T. Fujita, T. Kugo, K. Ohashi, Off-shell formulation of supergravity on orbifold, Prog. Theor. Phys. 106 (2001) 671,
hep-th/0106051;
T. Kugo, K. Ohashi, Superconformal tensor calculus on orbifold in 5D, Prog. Theor. Phys. 108 (2002) 203, hepth/0203276.
205
[4] E. Bergshoeff, S. Cucu, T. De Wit, J. Gheerardyn, R. Halbersma, S. Vandoren, A. Van Proeyen, Superconformal
N = 2, D = 5 matter with and without actions, JHEP 0210 (2002) 045, hep-th/0205230;
E. Bergshoeff, S. Cucu, T. de Wit, J. Gheerardyn, S. Vandoren, A. Van Proeyen, N = 2 supergravity in five dimensions revisited, Class. Quantum Grav. 21 (2004) 3015, hep-th/0403045.
[5] A. Galperin, E. Ivanov, S. Kalitsyn, V. Ogievetsky, E. Sokatchev, Unconstrained N = 2 matter, YangMills and
supergravity theories in harmonic superspace, Class. Quantum Grav. 1 (1984) 469.
[6] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Harmonic Superspace, Cambridge Univ. Press, Cambridge, 2001.
[7] A. Karlhede, U. Lindstrm, M. Rocek, Self-interacting tensor multiplets in N = 2 superspace, Phys. Lett. B 147
(1984) 297.
[8] W. Siegel, Chiral actions for N = 2 supersymmetric tensor multiplets, Phys. Lett. B 153 (1985) 51.
[9] U. Lindstrm, M. Rocek, New hyper-Khler metrics and new supermultiplets, Commun. Math. Phys. 115 (1988)
21;
U. Lindstrm, M. Rocek, N = 2 super-YangMills theory in projective superspace, Commun. Math. Phys. 128
(1990) 191;
F. Gonzalez-Rey, M. Rocek, S. Wiles, U. Lindstrm, R. von Unge, Feynman rules in N = 2 projective superspace.
I: Massless hypermultiplets, Nucl. Phys. B 516 (1998) 426, hep-th/9710250.
[10] N. Berkovits, W. Siegel, Superspace effective actions for 4D compactifications of heterotic and type II superstrings,
Nucl. Phys. B 462 (1996) 213, hep-th/9510106;
N. Berkovits, Conformal compensators and manifest type IIB S-duality, Phys. Lett. B 423 (1998) 265, hepth/9801009.
[11] A.A. Rosly, Super-YangMills constraints as integrability conditions, in: M.A. Markov (Ed.), Group Theoretical
Methods in Physics, Nauka, Moscow, 1983, p. 263.
[12] A.A. Rosly, A.S. Schwarz, Supersymmetry in a space with auxiliary dimensions, Commun. Math. Phys. 105 (1986)
645.
[13] E. Witten, An interpretation of classical YangMills theory, Phys. Lett. B 77 (1978) 394.
[14] S.M. Kuzenko, Projective superspace as a double-punctured harmonic superspace, Int. J. Mod. Phys. A 14 (1999)
1737, hep-th/9806147.
[15] S.M. Kuzenko, W.D. Linch, On five-dimensional superspaces, hep-th/0507176.
[16] B. de Wit, M. Rocek, S. Vandoren, Hypermultiplets, hyper-Khler cones and quaternion-Khler geometry,
JHEP 0102 (2001) 039, hep-th/0101161.
[17] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Conformal invariance in harmonic superspace, in: I.
Batalin, C.J. Isham, G. Vilkovisky (Eds.), Quantum Field Theory and Quantum Statistics, vol. 2, Adam Hilger,
Bristol, 1987.
[18] E.A. Ivanov, A.V. Smilga, B.M. Zupnik, Renormalizable supersymmetric gauge theory in six dimensions, Nucl.
Phys. B 726 (2005) 131, hep-th/0505082.
[19] A.A. Rosly, Gauge fields in superspace and twistors, Class. Quantum Grav. 2 (1985) 693.
[20] J. Lukierski, A. Nowicki, General superspaces from supertwistors, Phys. Lett. B 211 (1988) 276.
[21] P.S. Howe, G.G. Hartwell, A superspace survey, Class. Quantum Grav. 12 (1995) 1823.
[22] Yu.I. Manin, Gauge Field Theory and Complex Geometry, Springer, Berlin, 1988.
[23] R. Penrose, Twistor algebra, J. Math. Phys. 8 (1967) 345;
R. Penrose, M.A.H. MacCallum, Twistor theory: An approach to the quantization of fields and spacetime, Phys.
Rep. 6 (1972) 241.
[24] R.S. Ward, R.O. Wells, Twistor Geometry and Field Theory, Cambridge Univ. Press, Cambridge, 1991.
[25] G.W. Gibbons, A.R. Steif, Sphalerons and conformally compactified Minkowski spacetime, Phys. Lett. B 346
(1995) 255, hep-ph/9412210.
[26] I.E. Segal, Mathematical Cosmology and Extragalactic Astronomy, Academic Press, New York, 1976.
[27] 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.
[28] M.F. Sohnius, The conformal group in superspace, in: L. Castell, M. Drieschner, C.F. von Weizscker (Eds.), Quantum Theory and the Structures of Time and Space, vol. 2, Carl Hanser Verlag, Mnchen, 1977, p. 241.
[29] W. Lang, Construction of the minimal superspace translation tensor and the derivation of the supercurrent, Nucl.
Phys. B 179 (1981) 106.
[30] L. Bonora, P. Pasti, M. Tonin, Cohomologies and anomalies in supersymmetric theories, Nucl. Phys. B 252 (1985)
458.
[31] K.i. Shizuya, Supercurrents and superconformal symmetry, Phys. Rev. D 35 (1987) 1848.
206
[32] I.L. Buchbinder, S.M. Kuzenko, Ideas and Methods of Supersymmetry and Supergravity or a Walk Through Superspace, IOP, Bristol, 1998.
[33] P.C. West, Introduction to rigid supersymmetric theories, hep-th/9805055.
[34] H. Osborn, N = 1 superconformal symmetry in four-dimensional quantum field theory, Ann. Phys. 272 (1999) 243,
hep-th/9808041.
[35] J.H. Park, Superconformal symmetry in six dimensions and its reduction to four, Nucl. Phys. B 539 (1999) 599,
hep-th/9807186;
J.H. Park, Superconformal symmetry and correlation functions, Nucl. Phys. B 559 (1999) 455, hep-th/9903230.
[36] S.M. Kuzenko, S. Theisen, Correlation functions of conserved currents in N = 2 superconformal theory, Class.
Quantum Grav. 17 (2000) 665, hep-th/9907107.
[37] A. Uhlmann, The closure of Minkowski space, Acta Phys. Pol. 24 (1963) 295.
[38] I.T. Todorov, M.C. Mintchev, V.P. Petkova, Conformal Invariance in Quantum Field Theory, Pisa, Scuola Normale
Superiore, 1978;
I.T. Todorov, Conformal Description of Spinning Particles, Springer, Berlin, 1986.
[39] H. Weyl, SpaceTimeMatter, Dover, New York, 1952.
[40] P.A.M. Dirac, Wave equations in conformal space, Ann. Math. 37 (1936) 429.
[41] . Cartan, Sur les domaines borns homognes de lespace de n variables complexes, Abh. Math. Sem. Univ.
Hamburg 11 (1935) 116.
[42] L.K. Hua, Harmonic Analysis of Functions of Several Complex Variables in the Classical Domains, American
Mathematical Society, Providence, 1963.
[43] S.M. Paneitz, I.E. Segal, Analysis in spacetime bundles. I. General considerations and the scalar bundle, J. Funct.
Anal. 47 (1982) 78.
[44] J. Wess, J. Bagger, Supersymmetry and Supergravity, Princeton Univ. Press, 1992.
[45] A. Ferber, Supertwistors and conformal supersymmetry, Nucl. Phys. B 132 (1978) 55.
[46] B.S. DeWitt, Supermanifolds, Cambridge Univ. Press, Cambridge, 1992.
[47] B. Zupnik, Harmonic superpotentials and symmetries in gauge theories with eight supercharges, Nucl. Phys. B 554
(1999) 365, hep-th/9902038.
[48] P.S. Howe, G. Sierra, P.K. Townsend, Supersymmetry in six dimensions, Nucl. Phys. B 221 (1983) 331.
[49] P.S. Howe, K.S. Stelle, P.C. West, N = 1 D = 6 harmonic superspace, Class. Quantum Grav. 2 (1985) 815;
B.M. Zupnik, Sov. J. Nucl. Phys. 44 (1986) 512.
[50] J. Grundberg, U. Lindstrom, Actions for linear multiplets in six dimensions, Class. Quantum Grav. 2 (1985) L33;
S.J. Gates, S. Penati, G. Tartaglino-Mazzucchelli, 6D supersymmetry, projective superspace and 4D, N = 1 superfields, hep-th/0508187.
[51] B. de Wit, R. Philippe, A. Van Proeyen, The improved tensor multiplet in N = 2 supergravity, Nucl. Phys. B 219
(1983) 143.
[52] U. Lindstrom, M. Rocek, Scalar tensor duality and N = 1, 2 nonlinear sigma models, Nucl. Phys. B 222 (1983)
285.
[53] A. Galperin, E. Ivanov, V. Ogievetsky, Superspace actions and duality transformations for N = 2 tensor multiplets,
Sov. J. Nucl. Phys. 45 (1987) 157;
A. Galperin, E. Ivanov, V. Ogievetsky, Duality transformations and most general matter self-coupling in N = 2
supersymmetry, Nucl. Phys. B 282 (1987) 74.
[54] S.M. Kuzenko, On massive tensor multiplets, JHEP 0501 (2005) 041, hep-th/0412190.
[55] B. de Wit, M.T. Grisaru, M. Rocek, Nonholomorphic corrections to the one-loop N = 2 super-YangMills action,
Phys. Lett. B 374 (1996) 297, hep-th/9601115.
[56] I.L. Buchbinder, S.M. Kuzenko, A.A. Tseytlin, On low-energy effective actions in N = 2, 4 superconformal theories
in four dimensions, Phys. Rev. D 62 (2000) 045001, hep-th/9911221.
[57] M. Henningson, Extended superspace, higher derivatives and SL(2, Z) duality, Nucl. Phys. B 458 (1996) 445, hepth/9507135.
[58] S.J. Gates, S.M. Kuzenko, The CNM-hypermultiplet nexus, Nucl. Phys. B 543 (1999) 122, hep-th/9810137;
S.J. Gates, S.M. Kuzenko, 4D N = 2 supersymmetric off-shell sigma models on the cotangent bundles of Khler
manifolds, Fortschr. Phys. 48 (2000) 115, hep-th/9903013.
[59] N. Dragon, E. Ivanov, S. Kuzenko, E. Sokatchev, U. Theis, N = 2 rigid supersymmetry with gauged central charge,
Nucl. Phys. B 538 (1999) 411, hep-th/9805152.
[60] B. de Wit, B. Kleijn, S. Vandoren, Rigid N = 2 superconformal hypermultiplets, in: J. Wess, E.A. Ivanov (Eds.),
Supersymmetries and Quantum Symmetries, Springer, Berlin, 1999, p. 37, hep-th/9808160;
B. de Wit, B. Kleijn, S. Vandoren, Superconformal hypermultiplets, Nucl. Phys. B 568 (2000) 475, hep-th/9909228.
207
[61] D. Kaledin, Hyper-Khler structures on total spaces of holomorphic cotangent bundles, in: D. Kaledin, M. Verbitsky
(Eds.), Hyper-Khler Manifolds, International Press, Cambridge, MA, 1999, alg-geom/9710026;
B. Feix, Hyper-Khler metrics on cotangent bundles, Cambridge PhD thesis, 1999;
B. Feix, J. Reine Angew. Math. 532 (2001) 33.
[62] J. Bagger, C. Xiong, N = 2 nonlinear sigma models in N = 1 superspace: Four and five dimensions, hepth/0601165.
Hyper-Khler sigma models on (co)tangent bundles

with SO(n) isometry
Masato Arai a, , Muneto Nitta b
a High Energy Physics Division, Department of Physical Sciences, University of Helsinki
and Helsinki Institute of Physics, P.O. Box 64, FIN-00014, Finland

b Department of Physics, Tokyo Institute of Technology, Tokyo 152-8551, Japan
Received 9 March 2006; received in revised form 31 March 2006; accepted 31 March 2006
Abstract
We construct N = 2 supersymmetric nonlinear sigma models whose target spaces are tangent as well
as cotangent bundles over the quadric surface Qn2 = SO(n)/[SO(n 2) U (1)]. We use the projective
superspace framework, which is an off-shell formalism of N = 2 supersymmetry.
1. Introduction
Hyper-Khler manifolds provide a fruitful relation between physics and mathematics. One of
ingredients to study them may be solitons. For instance, it is well-known that the moduli space
for YangMills (YM) instantons on R4 [1] and one of BogomolnyiPrasadSommerfield (BPS)
monopoles [2,3] are both hyper-Khler. Gravitational instantons [4,5] and the moduli space of
YM instantons on gravitational instantons [6] are also hyper-Khler. More direct connection related to these would be given by supersymmetric nonlinear sigma models with eight supercharges
(like N = 2 supersymmetry in d = 4 and N = (4, 4) supersymmetry in d = 2): scalar fields in
these models belong to hypermultiplets, parametrizing target spaces which must be hyper-Khler
from the requirement of N = 2 supersymmetry [7]. Conversely there exists the unique (massless)
nonlinear sigma model for arbitrary hyper-Khler manifold. Hyper-Khler structure on soliton
moduli spaces can be understood in terms of nonlinear sigma models as follows. Instantons and
E-mail addresses: masato.arai@helsinki.fi (M. Arai), nitta@th.phys.titech.ac.jp (M. Nitta).

doi:10.1016/j.nuclphysb.2006.03.033
M. Arai, M. Nitta / Nuclear Physics B 745 [PM] (2006) 208235
209
BPS monopoles can be naturally embedded into supersymmetric gauge theories with sixteen supercharges, and they preserve/break half of supersymmetry. Their dynamics can be described by
nonlinear sigma models with eight supersymmetry preserved by them. The hyper-Khler quotient was discovered in the context of nonlinear sigma models [8,9]. Since then it has become an
important tool: YM instantons, gravitational instantons and YM instantons on gravitational instantons can be obtained by certain hyper-Khler quotients [1,5,6]. Hyper-Khler sigma models
also give low energy effective action on the Higgs branch of vacua in N = 2 supersymmetric
gauge theories [10,11], where target space is obtained as the hyper-Khler quotient.
Construction of explicit metrics on hyper-Khler manifolds is an important problem. Compact hyper-Khler manifolds are difficult to construct whereas known ones are all non-compact.
In general an isometry of manifolds often restricts the form of their metrics. An important class of hyper-Khler manifolds is given by toric hyper-Khler (hypertoric) manifolds
[8,1215], namely 4n-dimensional hyper-Khler manifolds admitting mutually commuting n triholomorphic isometries. This class of manifolds was firstly found in construction of the general
action of N = 2 supersymmetric tensor multiplets [8], which can be dualized by the Legendre
transformation to hypermultiplets with toric hyper-Khler manifolds. Thus N = 2 supersymmetric models are obviously useful to study hyper-Khler manifolds. Any toric hyper-Khler
manifold (with dimension 4n) can be obtained as hyper-Khler quotient of flat space Hn+m by
U (1)m [14]. Hyper-Khler manifolds
k which are not toric were studied with hyper-Khler quotient by a product of gauge group i=1 U (Ni ) in Ref. [16].
An isometry of non-Abelian group G is further restrictive for metrics of manifolds. Homogeneous Khler manifolds G/H were completely classified [17,18] and their Khler potentials were
systematically constructed [19]. Hyper-Khler manifolds cannot become homogeneous, so we
may consider in a slightly different way. Let us remember that homogeneous Khler manifolds
can be formulated as co-adjoint orbits of Lie algebra G with the so-called KirillovKostant
Souriau symplectic structure [20]. Then, co-adjoint orbits of complex Lie algebra GC become
cotangent bundles GC /H C T (G/H ) over homogeneous Khler manifolds G/H , and they
are known to admit hyper-Khler metrics [21].1 More explicit analysis was performed for cotangent bundle over Hermitian symmetric spaces [25]. Later it was shown that cotangent bundles
over any Khler manifolds M (without any isometry) admit hyper-Khler metrics at least in
neighbour of M [26,27].
When one would like to construct arbitrary hyper-Khler manifold, fully off-shell N = 2
supersymmetric formalisms should be useful. The harmonic superspace provides such a fully
off-shell N = 2 superspace [28,29]. It can provide the most general action for hypermultiplets
which induces the most general hyper-Khler manifolds if one can eliminate infinite number of
auxiliary fields. The projective superspace [3037] is another fully off-shell N = 2 superspace.
Its equivalence to harmonic superspace was discussed [38]. Recently the projective superspace
in five- and six-dimensions has also been studied [3941]. In the six-dimensional case, the projective superspace was first introduced in [42] to construct self-coupling of N = (1, 0) tensor
multiplets.
It was shown by Gates and Kuzenko [43,44] that the particular multiplets in the projective
superspace, called the polar multiplets, are suitable to describe N = 2 supersymmetric nonlinear
1 Stenzel constructed the Ricci-flat Khler metric on complexification of Riemann symmetric spaces GC /H C
T (G/H ) [22]. See [23,24] for explicit metric in the case of S N 1 SO(N )/SO(N 1).
210
sigma models on tangent (but not cotangent) bundles T M over Khler manifolds M.2 The polar
multiplets i contain chiral superfields i and complex linear (nonminimal) superfields i [46]
in terms of N = 1 superfields, where i parametrize the base Khler manifold M and i are regarded as components of a tangent vector on M. Complex linear superfields i can be dualized
by the Legendre transformation to chiral superfields i which can be regarded as components
of a cotangent vector. Then, the nonlinear sigma models on cotangent bundles T M in terms of
solely chiral superfields ( i , i ) are obtained. Once the Khler potential K(, ) of N = 1
supersymmetric model on any Khler manifold M is given, one can easily obtain its N = 2 supersymmetric extension on (co)tangent bundle T () M with replacing chiral superfields i by
polar multiplets i . This is nicely conforming to the mathematical result [26,27].3 However,
the main problem to obtain explicit action in terms of component fields (or N = 1 superfields)
is that one has to eliminate infinite number of auxiliary N = 1 superfields contained in the polar multiplets. The authors in [43,44] explicitly constructed nonlinear sigma models on tangent
and cotangent bundles over the complex projective space CP n1 = SU(n)/[SU(n 1) U (1)],
which is one of the Hermitian symmetric spaces, by eliminating auxiliary fields with the help
of the isometry SU(n) on CP n1 . The cotangent bundle action recovers the T CP n1 sigma
model constructed by the hyper-Khler quotients [47]. The purpose of the present paper is to
construct N = 2 supersymmetric nonlinear sigma models on (co)tangent bundle on another Hermitian symmetric space, the so-called quadric surface Qn2 = SO(n)/[SO(n 2) U (1)], with
following their work.
This paper is organized as follows. We give a review of the Khler quotient construction of
Qn [48,49] in the rest of introduction. In Section 2, we give a brief review of the projective
superspace. In Section 3, we review how to construct nonlinear sigma models with tangent and
cotangent bundles over the projective space CP n1 . We consider a sigma model with tangent
bundle T Qn2 in Section 4. In Section 4.1, by using the isomorphism Q2 CP 1 CP 1 , we
construct the sigma model with tangent bundle T Q2 . In Section 4.2, we solve the equations of
motion for auxiliary fields and derive the T Qn2 action. In Section 5, we derive the nonlinear
sigma action with cotangent bundle T Q2 via the Legendre transformation. Further we propose
the cotangent bundle action for T Qn2 . Section 6 is devoted to discussion. In Appendix A, we
review an another method to eliminate infinite set of auxiliary fields based on the duality between
polar and O(2) multiplet with some examples. In Appendix B, we show the detailed derivation
of identities between metric and Riemann tensor. In Appendix C, we derive the sigma model with
cotangent bundle T Q2 with the isomorphism T Q2 T CP 1 T CP 1 . We discuss the n = 3
case in T () Qn2 sigma model, and show that the solution for and the (co)tangent bundle
action is T () CP 1 s one in Appendix D.
Before closing introduction we review how to construct nonlinear sigma models on the
quadric surface Qn2 in terms of the N = 1 superfields [48,49]. Let i (x, , ) (i = 1, . . . , n)
be chiral superfields, D i = 0 belonging to the vector representation of SO(n). Introducing
an auxiliary vector superfield V (x, , ) (= V ) and an auxiliary chiral superfield (x, , )

(D = 0), being a singlet representation of SO(n), the Lagrangian can be written as

L=

d 4 i i eV r 2 V +

d 2 i i + c.c.
2 They also studied hyper-Khler manifolds by using c-map in the projective superspace framework [45].
3 The work of Gates and Kuzenko was done independently from [26,27]. In fact it was earlier than [27].
(1.1)
211
with summation over repeated index i implied, and r 2 a real positive constant called Fayet
Iliopoulos parameter. This Lagrangian possesses gauge invariance
V V ,
i e i ,
e2
(1.2)
with arbitrary chiral superfield (x, , ). Equation of motion of V read

= 0,
which can be solved as V = log( i i /r 2 ). When the superpotential is absent in the Lagrangian (1.1), we obtain the Khler potential of CP n1 by substituting the solution back into
the Khler potential of (1.1) as

| i |2
K = r 2 log 1 + 2
(1.3)
r
r2
i i eV
with a gauge fixing = ( i , r) (i = 1, . . . , n 1). But now there exists the superpotential in
the Lagrangian (1.1). Decomposing i in the representation of the SO(n 2) U (1) group of
SO(n) as i = (x, y j , z) (j = 1, . . . , n 2), the SO(n) transformation law is given by [48]
i
j
0
x
i = i ij i y j ,
(1.4)
j
z
0

i
where ij = 2i kl (T kl )ij and (T ij )kl = 1i ( ik j l kj il ). We take the rank-2 invariant tensor as
0
0
1
J = 0 1n2 0 .
(1.5)
1
The equation of motion of gives the constraint

2 = T J = 2xz + y 2 = 0.
(1.6)
This can be solved to give
= y j .
(1.7)
y2
2x
With a gauge fixing x = r, T = (r, i , 2r1 2 ), we obtain the Khler potential of the quadric
surface [50,48,49,51], given by

i j
| i |2 ( i )2 ( j )2
2
K , = r ln 1 + 2 +
(1.8)
.
r
4r 4
The Khler metric can be calculated to give
gi j =
=
2K
i j
i j
1+
| k |2
r2
( l )2 ( m )2
4r 4
2 i j | k |2 i j ( k )2 i j ( k )2
2r 4
l 2
m 2 n 2
(1 + |r 2| + ( )4r(4 ) )2
i j j i
r2
(1.9)
212
2. Projective superspace
The projective superspace [3037] consists of a complex projective coordinate , which is an
inhomogeneous coordinate of CP 1 , in addition to N = 2 global superspace R4|8 parameterized
by

zM = x , i , i ,
(2.1)
i = i
, i = 1, 2,
where the index i labels the fundamental representation of the automorphism group SU(2)R .
Superfields on the projective superspace are functions of this projective superspace with the
constraints
(z, ) = (z, ) = 0,
(2.2)
where and are linear combination of N = 2 supercovariant derivatives in four dimensions,

given by
( ) = D1 + D2 ,
( ) = D 2 D 1 .
(2.3)
algebra4
Here the supercovariant derivatives satisfy the following

j
j
{Di , Dj } = {D i , D j } = 0,
Di , D = 2ii .
(2.4)
Notice that is the conjugate of under the composition of complex conjugation with the
antipodal map on the Riemann sphere, 1/ , and multiplication by an appropriate factor.
For example,

1
( ) = ( )( ) .
(2.5)
In the following, all conjugate of fields and operators in projective superspace are defined in this
sense.
The constraints (2.2) for superfields are analogous to one for a chiral superfield in N = 1 superspace formalism where the chiral subspace is defined. The constraints (2.2) define a subspace
of the full N = 2 superspace (2.1). Since a function K(, ) of superfields is independent of
some (a half) of the Grassmann coordinates of N = 2 superspace by definition (2.2), its integration over the orthogonal operators for (2.3).
= 1 D1 D2 ,
= 1 D 2 + D 1

(2.6)
is invariant under N = 2 supersymmetry. This leads to the following N = 2 invariant action:

2
1
2
d 4 x d
S=
(2.7)
K(, , ).
32i
16
C
Here the integration contour C in the -plane is supposed to be chosen to make the action (2.7)
nontrivial (i.e. not equal to zero). In the following, we take the contour to surround the origin in
the -plane.
and
is given by
The algebra for , ,
= {, } = {, }
= {, } = 0,
{, } = {, }
4 We take the normalization as D 2 = 1 D D .
(2.8)
} = { , } = 4i .
{ ,
213
(2.9)
Using Eqs. (2.2), (2.8) and (2.9) with the identities

= 1 (2D ),
= 2D + 1 ,

(2.10)
the manifestly N = 2 supersymmetric action (2.7) can be reduced to the action in terms of N = 1
superfields,

d
d D 2 D 2
1
1
S = d 4x
(2.11)
K |, |, = d 8 z
K |, |,
2i
16
2i
with N = 1 superfield covariant derivative defined by D D1 , and d 8 z d 4 x D 2 D 2 /16 =

d 4 x d 2 d 2 . Here | indicates the 2 and 2 independent part of a superfield . In the following, we will simply write it as .
The superfields obeying the constraints (2.2) are classified into (i) real/complex O(k) multiplets [31], (ii) rational multiplets [32], and (iii) analytic multiplets [32]. Furthermore analytic
multiplets contain the so-called polar multiplets and the real tropical multiplets, which describe
charged N = 2 hypermultiplets and vector multiplets, respectively [36,37].5 In what follows, we
focus on the polar multiplets to consider N = 2 supersymmetric nonlinear sigma models. The
polar multiplets and their conjugation can be expanded in terms of as

1 n
n
n (z) ,
(z, ) =
,
n (z)
(z, ) =
(2.12)
n=0
n=0
Here all n (and n ) are N = 1 superfields: 0 is a chiral superfield, 1 a complex

linear (or nonminimal) superfield [46], satisfying the N = 1 constraints
respectively.6
D 0 = 0,
D 2 1 = 0,
(2.13)
respectively, due to the constraints (2.2). The rests of fields 2 , 3 , . . . , are complex unconstrained superfields, which are always auxiliary once the action is given.
The free action obeying hermiticity and N = 2 supersymmetry is given by

d
.
Sfree = d 8 x
(2.14)
2i
C
On the other hand, the action (2.11) with the polar multiplets is the most general action for
N = 2 supersymmetric nonlinear sigma models on the tangent (but not cotangent) bundles over
Khler manifolds. For convenience let us rewrite physical N = 1 superfields in the projective
superfields i (with i labelling projective superfields) as

d i ( )
i
i
i

( ) =0 ,
(2.15)

.
d =0
Then, i and i are regarded as coordinates of the base Khler manifold and components of a
tangent vector, respectively, as explained as follows. The action (2.11) respects all the geometric
5 Cutting off the power series in (2.12) at some finite k(> 2), one results in the complex O(k) multiplet. The case k = 1
corresponds to the on-shell hypermultiplet, while for k = 2 we obtain two tensor multiplets.
6 The projective superfields and are called arctic and antarctic [36,37], respectively.
214
features which N = 1 supersymmetric nonlinear sigma model on Khler manifolds possesses.

For instance, the action of N = 1 supersymmetric nonlinear sigma model

S = d 8 z K i , i
(2.16)
is invariant under the Khler transformation

K(, )
+ () + (
)
.
K(, )
(2.17)
This invariance can be promoted to

) ,
K(, ) K(, ) + ( ) + (
(2.18)
for the action (2.11). A holomorphic field redefinition i f i ( j ) of the chiral superfields
in the action (2.16) gives a holomorphic coordinate transformation. This is promoted for the
action (2.11) to

i fi j
(2.19)
deducing the transformation laws of i and i as holomorphic coordinates of the base Khler
f i
j
manifold, i f i ( j ), and components of a holomorphic tangent vector, i
j () .
Thus, the set of fields ( i , i ) parameterizes the tangent bundles of the Khler manifolds. Note
that the action (2.11) is invariant under the rigid U (1) transformations

( ) ei n (z) ein n (z),
(2.20)
which can be regarded as chiral rotations of the fermionic coordinates of the N = 2 superspace
(a diagonal group of automorphism group SU(2)R ). This U (1) action is precisely the one in
Refs. [26,27] acting on fiber.
In Ref. [43], it was claimed that there exists a minimal extension of every four-dimensional
N = 1 supersymmetric nonlinear sigma model described by (2.16) to four-dimensional N = 2
supersymmetric nonlinear sigma model described by (2.11). Indeed, it is easy to see that the
action (2.11) involves N = 1 Khler potential and can be regarded as an N = 2 extension.
Representing ( ) in the form = + + A( ) where A( ) contains all the auxiliary
superfields, the action (2.11) can be rewritten as

1
d
.
S = d 8z
exp
K(, )
exp A
+ A

2i

(2.21)
One can see that an extension of N = 1 model into N = 2 nonlinear sigma model can be obtained via (2.21) with corresponding N = 1 Khler potential. However, since this action still
includes the infinite tower of auxiliary superfields A, we have to eliminate them by their equations of motion, in order to obtain the action in terms of physical superfields and only. Their
equations of motion read

d n
1
(2.22)
K( , ) = 0, n 2
2i
i
with ( ) denoting a solution.
In general, it is difficult to solve (2.22) exactly and the auxiliary fields can be eliminated
at most perturbatively [43]. However, it was claimed in Ref. [43] that one can exactly solve
Eq. (2.22) if the following conditions are satisfied:
215
The Khler manifold is a homogeneous space, a coset space G/H , with an isometry G.
The Khler potential is invariant under the N = 1 U (1)R symmetry, defined by ei
and ei .
The authors in Ref. [43] showed how to solve Eq. (2.22) for the CP 1 base manifold explicitly
and constructed N = 2 supersymmetric nonlinear sigma models on the tangent T CP 1 and the
cotangent T CP 1 bundles over CP 1 . They also wrote down the T CP n1 model in Ref. [44].
In the following, we give a comprehensive review of how to obtain the nonlinear sigma model
with T CP n1 . Then, we consider the action with the tangent T Qn2 and cotangent T Qn2
bundles over quadric surface Qn2 = SO(n)/[SO(n 2) U (1)].
3. (Co)tangent bundle over CP n1
The Khler potential of CP n1 nonlinear sigma model is given by7

i j
| i |2
2
K , = r ln 1 + 2 ,
r
(3.1)
where i (i = 1, . . . , n 1) are chiral superfields and r is a real constant with mass dimension
one. The potential for its N = 2 extension can be obtained by replacing i in (3.1) by the
superfield i . The equations of motion to the infinite auxiliary fields read from (2.22):
1
2i
d m r 2 i
= 0,
r 2 + | |2
m 2.
(3.2)
It is difficult to find a solution of Eq. (3.2) at arbitrary point of the base manifold CP n1 . However, one can readily find the solution at the origin = 0 as
0i = 0i ,
0i =
0i
.
(3.3)
Here 0i is a tangent vector at the origin of . We need a solution i at

= 0. In order to obtain
it, one should take into account that Eq. (3.2) is invariant under the SU(n) transformation because
CP n1 is a homogeneous space with SU(n) isometry and the Khler potential (3.1) is invariant
under SU(n) transformation up to a Khler transformation. Then, applying SU(n) transformation
to the curve 0i ( ), one can obtain the solution i ( ) at
= 0 (see Fig. 1).
Let i = (x, y j )T (j = 1, . . . , n 1) be homogeneous coordinates of CP n1 . Infinitesimal
SU(n) transformations for i can be decomposed into [48]

i = iT + i A TA + j E j + j E j i

i 2(n1)
j
x
n
=
(3.4)
,

j
2
y
i
i A (T )ij i
ij
A
n(n1)
7 We take the notation that repeated indices mean the summation with respect to them even if they are at the same

upper(down) positions. For instance, Aij B j j Aij B j . Throughout this paper, we respect upper and down indices as
vector and covector, respectively.
216
Fig. 1. Base manifold CP n1 with its tangent spaces at = 0 and

= 0. One can arrive at any point
= 0 from
0 = 0 by SU(n) transformation with parameters and on the manifold. By using the transformation law for ,
the tangent vector at
= 0 is also obtained.
where T is the U (1) generator, E i (E i = (E i ) ) is n 1 raising (lowering) operators represented

by upper (lower) triangle matrices, and (TA ) is an n 1 by n 1 matrix for the fundamental
representation of SU(n 1). From Eq. (3.4), one can obtain the action of the finite transformation
on the homogeneous coordinates

x cos A + y Asin A
x
i =
(3.5)
,
j
j
yj
x sin A + y j y (1 cos A)
A2
i i
= .
Here we take = A = 0 since such a parameter choice is sufficient to
with A
consider the point
= 0. Using Eq. (3.5), one obtains the transformation law for inhomogeneous
coordinates i ry i /x (i = 1, . . . , n 1) of the CP n1 as
=f
i
j
0 =
r i sin A
A cos A
0i
cos A
1+
i ( 0 )
(1 cos A)
A2 cos A
.
0 sin A
rA cos A
Eq. (3.6) tells us a transformation law for the tangent vector i

f i ()
j
i
=
.
j =0 0
(3.6)
(3.7)
Substituting 0 = 0 into (3.6) and (3.7), one obtains the transformation law from the point 0 =
0 to
= 0 as
r i sin A
,
i = f i (0 = 0) =
A cos A

ij
f i ()
i j
j
j
j
i
(cos A 1) 0 V i j (, )
=
0.
=
j =0 0
cos A A2 cos2 A
Here V i j can be written in terms of i with (3.8) as

j
i
k |2
m |2
|
V ij = 1 + 2
ij
1 1+
,
r
| l |2
r2
(3.8)
(3.9)
(3.10)
217
and its inverse is given by

1 i
=
i j
1

1
.
| l |2
(1 + | m |2 )/r 2
ij
(1 + | k |2 )/r 2
(3.11)
Taking that i follows the same transformation law with (3.6) into account, one finds
i
= f (0 ) =
i
r i sin A
A cos A
0i
cos A
1+
i ( 0 )
(1 cos A)
A2 cos A
.
0 sin A
rA cos A
(3.12)
By using (3.3) and (3.8), Eq. (3.12) can be rewritten in terms of i and 0i as
j
i + D i j 0
=
,
1 ( 0 ) /r 2
with
(3.13)

D

| k |2 ij i j
i j
1+ 2
.

+
r
| l |2
| k |2
(3.14)
Using (3.9) and (3.11), we obtain a simple form of the solution

i =
i (1 + | j |2 /r 2 ) + { i (1 + | j |2 /r 2 ) i ( )/r 2 }
.
1 + | k |2 /r 2 ( )/r 2
(3.15)
It can be easily checked that this is actually the solution of (3.2) by substituting (3.15) into (3.2).
Replacing i in the Khler potential (3.1) by i in Eq. (3.15), we obtain

| i |2
2
K = + + r ln 1 + 2
r

2
i 2
1
| |
2

+ r ln 1 2
(3.16)
+ 2
,
r (1 + | j |2 /r 2 )
r (1 + | i |2 /r 2 )
with defined by

r 2 ln 1 2
.
r 1 + | i |2 /r 2
(3.17)
After integrating over the projective coordinate and introducing the Khler metric of CP n1
gi j =
r 2 i j
r 2 + | k |2
r 2 i j
,
2
(r + | k |2 )2
we finally obtain the following action

| i |2
1
8
2
2
i j
S = d z r ln 1 + 2
+ r ln 1 2 gi j
.
r
r
(3.18)
(3.19)
This is the action of the nonlinear sigma model with the tangent bundle T CP n1 . Since the
second term vanishes when = 0 holds, the first term is the Khler potential of the base CP n1
and the second term represents the tangent space. This form was obtained for T CP 1 in Ref. [43]
and for T CP n1 in Ref. [44].
218
In order to obtain the cotangent bundle, we need dualize the complex linear superfields i
in the action (3.19) into chiral superfields i being components of a cotangent vector via the
Legendre transformation. The resultant action can be identified with a hyper-Khler potential.
The action can be dualized if we replace the action (3.19) by the following one:

| i |2
1
i
8
2
2
i j
i
+ r ln 1 2 gi j U U + U i + U i ,
S = d z r ln 1 + 2
(3.20)
r
r
where U i is the complex unconstrained auxiliary superfields. By the construction, U i is a tangent
vector at point of the manifold and therefore i is a one-form at the same point. Eliminating
the auxiliary variables U i and U j , with the aid of the equations of motion
0=
gi j U j
S
=
+ i ,
U i
1 gi j U i U j /r 2
(3.21)
0=
gj i U j
S
=
+ i
U i
1 gi j U i U j /r 2
(3.22)
results in a purely chiral sigma model which is dually equivalent to the N = 2 supersymmetric
model (3.19) and is defined on the cotangent bundle T CP n1 . The final result is [47,52]

| i |2
8
2
2
2
,
r ln f () + 2r
S = d z r ln 1 + 2
f ()
r

1
f () = 1 + 1 + 4 ,
(3.23)
2
where g i j i j /r 2 and g i j the inverse metric of gi j .

Here we solved the equation of motion (3.2) to remove the auxiliary fields. There is an another
method to eliminate the infinite auxiliary fields, which is briefly reviewed in Appendix A.
4. Construction of tangent bundle T Qn2

In this section we construct the nonlinear sigma model with tangent bundle T Qn2
SO(n)
T [ SO(n2)U
(1) ]. Here we use a strategy different from the last section, because it is difficult to solve the equations of motion (3.2) for elimination of auxiliary fields in this case.
First we consider the n = 4 case of Q2 (called the Klein quadric) by noting the isomorphism
Q2 SO(4)/[SO(2) U (1)] [SU(2) SU(2)]/[U (1) U (1)] CP 1 CP 1 . We then solve
the equations of motion for auxiliary fields in this T Q2 case. We can show that the solution can
be extended to the T Qn2 case, and then construct the T Qn2 model.
4.1. First look: T Q2
Let us start in the base manifold. Considering the isomorphism mentioned above, the Khler
potential of the nonlinear sigma model with Q2 can be described by

i j
| 1 |2
| 2 |2
2
2
K , = r ln 1 + 2
+ r ln 1 + 2
r
r

i
2
1
2
2
2
| |
| | | |
, i = 1, 2.
= r 2 ln 1 + 2 +
(4.1)
r
r4
Using the unitary transformation

1

1
1 1 i
2
2 1 i
219
(4.2)
we arrive at the Khler potential (1.8) with i = 1, 2 for Q2 . The Khler metric is then given by
Eq. (1.9) with i = 1, 2.
Next we consider its tangent bundle. Replacing the chiral superfield i in Eq. (4.1) by the
scalar multiplet i , we arrive at the action of the tangent bundle:

1
d 1 1
S = d 8z
(4.3)
K , + K 2 , 2
2i
with

| i |2
K i , i = r 2 ln 1 + 2
(4.4)
(no sum).
r
Since the calculation for the CP 1 case in the previous section can be done independently for
each term in (4.3), the resultant action is just a sum of the action (3.19) with a different kind of
chiral and complex linear superfields:

| i |2 | 1 |2 | 2 |2
8
S = d z r 2 ln 1 + 2 +
r
r4

2

1 2 2 2
| i |2
1
2

Gi 2 + 4 G1 G2
,
+ r ln 1
(4.5)
r
r
i=1
where Gi is the metric of CP 1 given by

Gi =
1
| i |2 2
)
r2
(1 +
(4.6)
(no sum).
In Eq. (4.5), the first term is the Khler potential of the base manifold Q2 and the second one is
of tangent one. This form of the action is in particular coordinates and so it is better to rewrite it
by geometrical quantities such as Khler potential, metric and Riemann tensor. Indeed, as shown
in (1.8), the first term is written by Khler potential after performing unitarity transformation.
Let us focus on the tangent vector sector (second logarithm). Taking that the tangent vector
i follows the same unitary transformation law with one of i into account, we find that the

term 2i=1 Gi | i |2 /r 2 is the form of gi j i j , where gi j is a metric of quadric surface given

in Eq. (1.9). The last term in the second logarithm, G1 G2 | 1 |2 | 2 |2 , can be rewritten by the
covariant quantity by noting that it is in forth order in or . First, we calculate (gi j i j )2

in the frame taken in (4.5).
4
4
2 2

2
gi j i j = G21 1 + G22 2 + 2G1 G2 1 2 .
(4.7)
Another forth order term is
Ri jkl i j k l =
2| 1 |4
| 1 |2
2| 2 |4
r 2 (1 + r 2 )4 r 2 (1 + |r 2| )4
4
2 4
= 2 G21 1 + G22 2 ,
r
2 2
(4.8)
220
where Ri jkl is Riemann tensor defined by Ri jkl = k lgi j g mn m gi j n gkl. Using (4.7)
and (4.8), we get the relation

2 2 1
r2
2
G1 G2 1 2 =
(4.9)
gi j i j + Ri jkl i j k l .
2
2
Putting all together, and performing the unitary transformation, we finally obtain the action of
the tangent bundle T Q2 over quadric surface Q2

| i |2 ( i )2 ( j )2
S = d z r ln 1 + 2 +
r
4r 4

1
1
r2
2
+ r 2 ln 1 2 gi j i j + 4 gi j i j + Ri jkl i j k l
.
2
r
2r
(4.10)
There appear the fourth order terms of the tangent vector. This is a particular form of the general expression of the hyper-Khler sigma models suggested in (2.47) in Ref. [43]. The method
which we use is simpler and easier than solving the equation of motion (3.2) as performed in
the previous section, in which one may suffer from rewriting the transformation parameters i
and i in terms of i as has been done in (3.15) and also from finding the form of the tangent
bundle action (4.10). In the following subsection, we solve the equations of motion for auxiliary fields for T Qn2 case as the same with the T CP n1 case. Then, we show that the tangent
bundle action (4.10) is also valid in T Qn2 with simply extending the index i to run from 1
to n 2.
4.2. Solving the equations of motion and deriving T Qn2 action
We start with N = 2 extension of the Khler potential of Qn2

| i |2 ( i )2 ( j )2
j
2
K , = r ln 1 + 2 +
.
r
4r 4
(4.11)
The equation of motion for auxiliary fields of the polar multiplet reads
1
2i

C
j
r 2 i + 12 i ( )2
d m
= 0,
r 4 + r 2 |k |2 + 1 (k )2 (l)2
m 2.
(4.12)
It is easy to check that the same solution with (3.3) also satisfies this equation. We can obtain the
solution at
= 0 by applying the finite SO(n) transformation to the curve (3.3) since (4.12) is
SO(n) invariant.
In the following we again focus on the n = 4 case, T Q2 , for a while. Using Eq. (1.4) leads to
the finite SO(4) transformation law for the homogeneous coordinates as
1
+ (+ y0 ) sin + ( y0 ) sin +
x = (cos + + cos )x0 +
2
2+

( i )2 (cos + cos )

z0 ,
2 ( i )2
(4.13)
yi =
where
221

1
(cos cos + ) i ( y0 ) y0i ( ) + i ( y0 ) + (cos + + cos )y0i

2
+ +i sin + i sin +
+ +i sin + i sin +
x0
z0 ,
(4.14)
2+
2+

2 2
i
j ,
+ ,
+ ,
2
2
i j i + i .
+i j i + i ,

(4.15)
(4.16)
(4.17)
Using the definition of the inhomogeneous coordinates of quadric surface i ry i /x, the transformation laws for i and i are given by

+ +i sin + i sin +
y i
i
=r
(4.18)
= r
,
x 0 =0
+ (cos + + cos )

(cos cos + )( i j ij ( ) + i j )
i j
j
i
ij
+ +
0
=
(cos + + cos )
r2
j
(VQ )i j 0 .
(4.19)
Here, in the first line in Eq. (4.19), we have replaced the third term written by and with i
and j defined in (4.18). Since the transformation law for i is the same with one of i , we
can obtain the i at non-zero value of i from (4.13) and (4.14) as

2 i i ( )} + cos { i (
2 i + i ( )}
cos + { i ( 0 ) +
0 ) +
0
0
0
0
i =
(cos + + cos )

+ +i sin + i sin + + +i sin + i sin + l 2

r
+
+ (cos + + cos )
2r+ (cos + + cos ) 0

+ (+ 0 ) sin + ( 0 ) sin +
1+
r+ (cos + + cos )

( i )2 (cos + cos ) k 2 1

+
(4.20)
.
0
2r 2 ( j )2 (cos + + cos )
Let us rewrite Eq. (4.20) in terms of i and i . In order to do that, we need some formulas.
Using the expression (4.18) we have
i 2 2r 2 ( i )2 (cos + cos )
=
,
(cos + + cos )
(4.21)
i 2 2r 2 ( i )2 (cos + cos )
=
.
(cos + + cos )
(4.22)
Multiplying (4.21) and (4.22), we get

2 j 2 4r 4 (cos + cos )2
=
.
(cos + + cos )2
(4.23)
222
Dividing (4.21) by (4.22), we find

2
i
( )
( i )2

=
.
( j )2
( j )2
(4.24)
Using (4.18), (4.23) and (4.24), Eq. (4.20) can be rewritten as

2 i i ( )} + cos { i (
2 i + i ( )}
cos + { i ( 0 ) +
0 ) +
0
0
0
0
(cos + + cos )

1 j 2
0 ( i )2 k 2 1
+ i 2 i 0
+
.
(4.25)
1
0
2r
r2
4r 4
Further, using (4.19) with 0i = 0i , we obtain a simple form
i =
1 ))
i t 1 t 1
i ((V
Q
2 ( (VQ ) VQ )
r2
2r 2
1 )
k )2 ( t (V 1 )t V 1 )
(V
(
Q
Q
Q
+ 2
r2
4r 4

i + i
1
(4.26)
ij
To complete the work, we have to find the expression of VQ in terms of i . The result is

VQ =
| k |2
2r 2
1
2
1 2
2r2
1+

VQ1
=L
1 2 1 2
2r 2
| k |2
2r 2
1 2 1 2
2r 2
1+
1+
| k |2
2r 2
1
2 1 2
1+
(4.27)
2r 2
| k |2
2r 2

(4.28)
with
| i |2 ( i )2 ( j )2
.
L1+ 2 +
r
4r 4
(4.29)
Eq. (4.28) leads (VQ1t VQ1 )ij = L1 ij and we obtain the final form of the solution

j 2
i j 2
i + i i
+ ()(2 )
2 2r(2 L)
2
i
r L
2r
=
.

k 2 l 2
k 2
1 2 + ()(2 ) + 2 ( ) 4( )
r L
2r
(4.30)
4r L
This solution satisfies the equation of motion (4.12) for the T Q2 case of n = 4.
Let us turn back to general n. When one tries to derive the solution i for higher-dimensional
case (n > 4), it might be problematic to convert the transformation parameters i and i in VQ
and VQ1 into the base coordinates and eventually to find the final form of the solution . But
fortunately if one respects the solution (3.12) of n = 4 by making the index i run from 1 to n 2,
it becomes a solution for general n. Indeed, one can straightforwardly check that it satisfies the
equations of motion (4.12) by substituting Eq. (4.30) into (4.12):
1
2i

C
223
r 2 i + 12 i ( )2
d m
r 4 + r 2 |k |2 + 1 (k )2 (l)2
4
1
2i

d m2 8r 4 ! 4 i 2 i
r 2 + j 2 Y i
4r Y 8r 2 ( )
g(, )
2
2 4

+ j k Y i + 8r 4 2 r 2 + j ( ) Y i
2

i + 2 j 2 Y i + 2r 2 Y i + j 2 Y i 2 ( )Y i
+ 2 j r 2 2( )Y
2 "
+ 2 j Y i ,
(4.31)
2
where

i j 2
i
( )( j )2
2 ( )
.
Y = L + L 2 +
r
2r 2
2r 2
i
(4.32)
Here the denominator in the integrand can be written as 3 g(, ) and the function of and
, g(, ) can be factored out from the integral. Recalling that m 2, the numerator is a
polynomial starting from an order of 0 in the m = 2 case. In this case, the terms greater than
the first order of vanish because of the residue theorem and the terms in the order of 0 only
remain. However, they cancel and the integral becomes zero. In the case m 3, the integral are
trivially zero since all terms in the numerator starts from first order of .
Substituting the solution (4.30), with the index i running from 1 to n 2, into the Khler
potential (4.11), we can derive the nonlinear sigma model action with T Qn2 space:

1
| i |2 ( i )2 ( j )2
d
2
+ + r ln 1 + 2 +
S= d z
2i
r
r4

1
1
r2
2
+ r 2 ln 1 2 gi j i j + 2 gi j i j + Ri jkl i j k l
,
2
r
2r
(4.33)
with defined by

i 2 j 2
( )( i )2
2 ( ) ( )
= r 2 ln 1 2
+
+
.
r L
2r 2
4r 4 L
(4.34)
After performing the -integration, we arrive at the same form with (4.10) with the index i
running from 1 to n 2.
In Appendix D, we show this action in n = 3 coincides with the result in Section 3 due to
isomorphism Q1 CP 1 .
5. Cotangent bundle T Qn2
Here we consider the nonlinear sigma model action with the cotangent bundle over Qn2 ,
(hyper-Khler potential). To obtain the cotangent bundle, we need dualize the tangent
vector i into cotangent one i . In order to do that, let us consider the following Lagrangian
instead of the tangent bundle action (4.10):
T Qn2
224

| i |2 ( i )2 ( j )2
S = d z r ln 1 + 2 +
r
4r 4

1
1
r2
2
i j
i j 2
i j k l
+ Ri jklU U U U
+ r ln 1 2 gi j U U + 4 gi j U U
2
r
2r

+ U i i + U i i ,
(5.1)
where U i and U j are unconstrained superfields, regarded as a tangent vector, and i and j
are chiral superfields, regarded as a cotangent vector. Equation of motion of brings back to
original Lagrangian (4.10). On the other hand, varying with respect to U i and U j , we obtain
i =
j =
gi j U j
gi j U i
1
g U j 12 Ri jklU j U k U l
r 2 i j
,
2
1 r2 + 2r
4 + 4r 2
1
g U i 12 Ri jklU i U k U l
r 2 i j
,
2
1 r2 + 2r
4 + 4r 2
(5.2)
(5.3)
where
gi j U i U j ,
Ri jklU i U j U k U l .
(5.4)
Multiplying (5.2) and (5.3) by U i and U i , respectively, we get
U i = U i =
i
2
2
r2
+ 2r
4 + 4r 2
r2
(5.5)
Substituting Eqs. (5.2), (5.3) and (5.5) into the action (5.1), one can see that the action (5.1) is
a function of the covariant quantities and along with base manifold coordinates i . In the
following, we show that and are written by quantities
g i j i j ,
R ij kl i j k l ,
(5.6)
which are possible scalars in terms of cotangent vectors i and i . In order to do that, first we
substitute (5.2) and (5.3) into Eq. (5.6), and show that and are written in terms of and
with the help of the identities between metric and Riemann tensor. Then, solving them with
respect to and , and substituting the solution into the action (5.1), we obtain the cotangent
bundle action. In the following, we give an explicit calculation on the T Q2 case. The detailed
derivation of the identities in this case is given in Appendix B. The resultant equations after
substituting (5.2) and (5.3) into (5.6) is

2
2
3
2 2
i j
2 + 4 + 2 ,
= g i j = 1 2 + 4 + 2
(5.7)
r
2r
4r
r
2r
4r

= R ij kl i j k l

2
4
6 2
2 3
4
3
2 2
2
= 1 2 + 4 + 2
+
+ 2 4 + 6 + 8 +
r
2r
4r
r
r
r
4r
16r 4
4r 6

4
5
2
2
3
6
2
3
4
6 + 8 2 + 4 + 4 .
(5.8)
r
r
2r
2r
r
225
Solving Eqs. (5.7) and (5.8) with respect to and we obtain
= 2r 2
1
#
2
r2
1+ 1+
2
= 4r 1 +
#
1+ 1+
#
1+
1+
2
r2

2
+ 2 r 4
2
2
r2
r2
2 r 4
2
.

2
r4
#
1+
2
r2
2
r2
1+
1+
1+
r2
(5.9)
2
r 4
2
r2

2
+ 2 r 4
2
r2
(5.10)
r2
Substituting back (5.9) and (5.10) into the action (5.1), we obtain the action for the contangent
bundle T Q2

| i |2 ( i )2 ( j )2
8
2
S = d z r ln 1 + 2 +
r
4r 2
+

,
,
2
2

1 1 r 2 ln +
1+ 2 +2 4 2
2 2
r
r
r
+

,
,
2
2

1 11+ 2 2 4 2
r 2 ln +
2 2
r
r
r
+ r2
+
1
2
1
2
1+
2
r2
2
r 4
r2

2
+ 2 r 4
r2
r2
1
2
1
2
1+
2
r2
2
r 4
r2

2
2 r 4
r2
(5.11)
This action is of T Q2 . When one tries to derive the cotangent bundle action of T Qn2 case,
one comes across difficulty in deriving the identities for cotangent bundle action for T Qn2 , as
shown in Appendix B. However, we believe that this form of the action would be also valid in
T Qn2 case with extending the index to run from 1 to n 2, since the action is written in geometric quantities such as Khler potential, metric and Riemann tensor as in the tangent bundle case.
One might consider a possibility that invariant quantities higher in such like (g i j i j )3

k l m n ) together with and appear in the cotangent bundle action.
and (g i j i j )(R kl mn
However, one can confirm by deriving the cotangent bundle action in another way that such terms
do not appear. We derive the action (5.11) with the similar method used in Section 4.1. Using
the isomorphism T Q2 T CP 1 T CP 1 , we can represent the T Q2 action as a sum of two
T CP 1 action with different coordinate variables. Although naive summation is not written by
geometrical quantities, similarly to the tangent bundle case, we can write the action in terms of
metric and Riemann tensor. Finally we can reach the same form with (5.11) which is written in
226
terms on and with base manifold coordinates. It is shown in Appendix C. Discussion on

n = 3 case is given in Appendix D.
6. Discussion
We would like to propose possible future works. Potential terms in N = 2 hyper-Khler nonlinear sigma models are known to be written in the form of the square of a Killing vector of
the target space [53,54]. Recently these potentials in the framework of the projective superspace
has been established [55] by gauging the isometry to obtain the potential and then by freezing
it. Using this method we could construct the potential of our T () Qn nonlinear sigma model.
Massive extension of T CP n [56] and T Grassmann [57] nonlinear sigma models were formulated by the hyper-Khler quotient in the harmonic superspace formalism as well as components
and N = 1 superfields. These models are known to admit various (composite) BPS solitons, like
domain walls [58,56,59,15], (Q-)lump-strings [60] (vortex-strings [61]), domain wall junctions
[62], strings stretched between walls [63,64], intersecting vortex-strings [65,66], and other solitons [67]. See [6870] for a review in this subject. So constructing the massive T Qn model and
investigating BPS solitons in it are interesting future directions.
We have seen in the Lagrangian (1.1) in the introduction that the N = 1 Qn model can be
constructed by the Khler quotient. It has been found [59,15,61,62,65] that the hyper-Khler
quotient construction is crucial in solving the BPS equations to construct all the exact soliton solutions. Therefore, hyper-Khler quotient construction of the (massive) T Qn model is awaited.
N = 1 superfield formalism is difficult to perform it because the superpotential exists even in
the massless case (1.1). The projective superspace formalism should be useful to construct the
massive T Qn model in hyper-Khler quotient.
The cotangent bundle over the projective space CP n1 can be locally written as

SU(n)
SU(n)
n1
=T
R
.
T CP
(6.1)
SU(n 1) U (1)
SU(n 2)
Therefore, this is in cohomogeneity one. It was proved [71] that this is the unique cohomogeneity
one hyper-Khler manifold. On the other hand, T Qn2 can be locally written as

SO(n)
SO(n)
R2
T Qn2 = T
(6.2)
SO(n 2) U (1)
SO(n 4)
which is in cohomogeneity two. Cohomogeneity two hyper-Khler manifolds were discussed in
[72] but complete classification is not yet known at least to our knowledge. We hope our example
is useful to explore hyper-Khler manifolds with higher cohomogeneity.
There exist other Hermitian symmetric spaces, Gn,m = SU(n)/[SU(n m) SU(m) U (1)],
SO(2n)/U (n), Sp(n)/U (n), E6 /SO(10)U (1) and E7 /E6 U (1). A Khler quotient construction of these manifolds was given in [48].8 It should be possible to construct (co)tangent bundles
over Hermitian symmetric spaces [25] in the framework of the projective superspace. Extension
to (co)tangent bundle over arbitrary homogeneous Khler manifold G/H is one goal of this subject. To this end, tangent bundles are in more compact form than cotangent bundles as seen in
the cases of CP n (3.19) and Qn (4.10). General form of the Khler potential of tangent bundle
over arbitrary Khler manifold [26,27] expanded in terms of a tangent vector was proposed
8 The CalabiYau metrics on line bundles over Hermitian symmetric spaces were constructed [51,73], which are all in
cohomogeneity one.
227
in [43,44]. There, each coefficient should be written in terms of the metric, curvature, covariant derivative like the cases of CP n (3.19) and Qn (4.10), but explicit expression is not known
in general. This expansion is very similar with the Khler normal coordinate expansion [74].
We hope that Khler normal coordinates are useful toward the construction of general action of
hyper-Khler nonlinear sigma models on (co)tangent bundle over arbitrary Khler manifold.
Acknowledgements
We would like to thank Rikard von Unge for useful discussions and comments, especially, on
duality between polar and O(2) multiplets. We thank Kiyoshi Higashijima and Ulf Lindstrm for
discussions. M.A. acknowledges the Institute of Physics in Prague and the Masaryk university
for their hospitality while M.N. is grateful to the Helsinki Institute of Physics for their hospitality.
The work of M.A. is supported by the bilateral program of Japan Society for the Promotion of
Science and Academy of Finland, Scientist Exchanges while the work of M.N. is supported by
Japan Society for the Promotion of Science under the Post-doctoral Research Program.
Appendix A. Duality between polar and O(2) multiplets
Here we show the other way to eliminate infinite number of auxiliary fields, which is not
mentioned in the text. The point is to perform a duality transformation [32,75] between the polar
multiplet and real O(2) multiplet defined by
= + 2 ,
= .
(A.1)
The duality transformation is possible only if the polar multiplet appears as the linear combination such that + in an action. After the duality transformation, the action is described
by O(2) multiplet and there is thus no auxiliary fields.9 Let us illustrate this with a couple of
examples.
Firstly we consider the flat space. Let us start with an action depending on the arctic projective
multiplet :

1
d
8
d z
S=
(A.2)
( + )2 .
2
2i
This is a sigma model in flat space since the Lagrangian consists of the Khler potential of
flat space and a Khler transformation. Now we instead introduce an O(2) multiplet and an
unconstrained projective superfield X

d
1 2
8
S= d z
(A.3)
X X
.
2i 2
Varying with respect to tells us that X can be written as a sum of an arctic and antarctic field
X = + and we are back to the original model. If we on the other hand vary with respect to
X we get an equation
X= .
(A.4)
9 As more general case, there is duality between the real O(2p) and polar multiplets [36], which involves the case we
will discuss below.
228
Inserting this back into the action we get

d 2
,
S = d 8z
(A.5)
2i 2
which is the action for a sigma model in flat space written with an O(2) multiplet.
In the second example, we start from another Khler potential in flat space:

d +
e
S = d 8z
(A.6)
.
2i
In the same way, we introduce an O(2) multiplet and a polar multiplet X to get the action

d
8
X
S= d z
(A.7)
e X
.
2i
Again, integrating out tells us that X can be written as a sum over an arctic and an antarctic
multiplet while integrating out X gives the action

d
8
1 ln
.
S= d z
(A.8)
2i
This is also a known form of the action for a sigma model in flat space. There is one unresolved
issue here with the integration contour. Because of the nonsingle validness of the log, the integration contour will not be simply a contour around the origin but rather a shaped curve around
the zeroes of the quadratic polynomial ( ).
This method can be applied to any Khler potential which can be written as a function of
+ . Namely

d
d
8
8
f ( + ) = d z
f (X) X
.
S= d z
(A.9)
2i
2i
Integrating out X one gets
f (X) = ,
(A.10)
which can be formally inverted to give

X=g
(A.11)
.
Inserting this back into the action gives

d
8
S= d z
(A.12)
f g
g
.
2i
Technically we see that the new action is the Legendre transform of the old one. For example,
starting with the Khler potential of T CP 1 ln(1 + e + ), we get

d

ln 1 + e +
S = d 8z
2i

d
ln
1
ln 1
= d 8z
(A.13)
2i
which is again a known form for the nonlinear sigma model with the EguchiHanson metric. The
contour is still an issue similarly to the flat case e + . As be seen, one sees how one does not
have to solve the infinite number of equations although it is not applicable to the quadric surface.
229
Appendix B. Deriving equations (5.7) and (5.8)

In this appendix, we show how to derive Eqs. (5.7) and (5.8). Substituting (5.2) and (5.3) into
and leads to, after some algebra
= g i j i j

2
2
= 1 2 + 4 + 2
r
2r
4r

2
2
3
gij
l p q k m n
2 + 4 + 2 +
R R U U U U U U ,
4 i lpq k j mn
r
r
r
(B.1)
= R ij lk i j lk

2
4 ji lk
= 1 2 + 4 + 2
R
r
2r
4r

4 6 2 4 3 4
n m
a b
gnj gi m ga lgk b U U U U 1 2 + 4 6 + 8
r
r
r
r

2
2 3
6 6
n m
a b c d
2 + 2 4 + 6
+ gnj gi m gka Rblc
d U U U U U U
r
r
r

3 3 2
m
n p q a b c d 3
U
U
U
U
+
+ gi m gka Rnjpq Rblc

U
U
U
U
d
2
r2
2r 4

1
x m
n p a b c y z w
1 + 2
+ gk x Ri mn
w U U U U U U U U U U
p Ra jbc Ry lz
2
r

1
m
n p a b c x y z d e f
+ Ri mn
.
f U U U U U U U U U U U U
p Ra jbc Rk xy
z Rd le
16
(B.2)
It is seen that these equations are not closed with respect to and . However, using iden
tities between metric and Riemann tensor, one can rewrite terms such like g i j Ri lp
q Rk jmn
p q k m n
l
U U U U U U in the right-hand side in (B.1) in terms of and .

First we consider Eq. (B.1). Calculating last term in the right-hand side in (B.1) in the frame
taken in (4.5), we find
lU p U q U k U m U n =
g i j Ri lp
q Rk jmn U
4|U 1 |6
4|U 2 |6
+
.
r 4 (1 + |1 |2 /r 2 )6 r 4 (1 + |2 |2 /r 2 )6
(B.3)
On the other hand, it is shown that

2|U 1 |6
2|U 2 |6
r 2 (1 + |1 |2 /r 2 )6 r 2 (1 + |2 |2 /r 2 )6
2 2 2
2
2
2 G1 G2 U 1 U 2 G1 U 1 + G2 U 2 .
r
(B.4)
Using (B.3), (B.4) and (4.9) with = gi j U i U j , we get the following identity
lU p U q U k U m U n = 3 2 3 .
g i j Ri lp
q Rk jmn U
r2
r4
(B.5)
230
Since this identify is written by geometric quantities, it holds in arbitrary frame. Substituting
(B.5) into (B.1), we obtain (5.7).
Next we consider Eq. (B.2). Once again choosing the frame taken in (4.5), we find the following identities:
4
2
2

m U n U p U q U a U b U c U d = 4 + 4 ,
R ij kl gnj gi m gi m gka Rnjpq Rblc
d U
r6
r2
r4
kl
ij
x
n
p
a
b
c
y
z
w
U U U U U U U U U
R g gi m gk x Ri mn
p R R U
nj
a j bc
(B.6)
y lzw
5 2 4 5
= 4 8 ,
r
r
kl
ij
m U n U p U a U b U c U x U y U z U d U e U f
R gnj gi m Ri mn
f U
p Ra jbc Rk xy
z Rd le
3 12 4
12 2 2
.
r4
r8
r6
Substituting (B.5) and (B.6)(B.8) into (B.2), we find Eq. (5.8).
=
(B.7)
(B.8)
Appendix C. Derivation of cotangent bundle without using Legendre transformation

Instead of using the Legendre transformation, we can directly obtain the cotangent bundle
action (5.11) for T Q2 case. Similarly to the tangent bundle case, with the isomorphism T Q2 =
T CP 1 T CP 1 , the action for T Q2 can be written as a direct sum of two actions of cotangent
bundle over CP 1 . Using the action (3.23), we have

| i |2 | 1 |2 | 2 |2
S = d 8 z r 2 ln 1 + 2 +
r
r4

1
2
r 2 ln f (1 ) + 2r 2
(C.1)
r 2 ln f (2 ) + 2r 2
,
f (1 )
f (2 )
where

1
1
(C.2)
i i (no sum).
1 + 1 + 4i , i = 2 G1
2
r i
Here Gi is defined in (4.6). In the following, we rewrite the action (C.1) in terms of geometric
quantities similarly to the tangent bundle case. We consider the two quantities = g i j i j

and = R ij kl i j k l as possible invariant forms written by i and j as the same with in
Section 5. In the frame taken in (4.5), one finds
f (i ) =
1
1
|1 |2 +
|2 |2 ,
G1
G2
2
2
= 2 |1 |4 2 |2 |4 .
2
G1 r
G2 r 2
=
(C.3)
(C.4)
Using (C.3) and (C.4) leads to

2 =
|1 |4 |2 |4
2
r2
2
+
+
|1 |2 |2 |2 = R ij kl i j k l +
|1 |2 |2 |2 ,
2
2
G1 G2
2
G1 G2
G1
G2
2
1
2
2
|
|
|
|
=
+ 2.
1
2
2
4
r G1 G2
2r
4r
(C.5)
231
From (C.3), (C.4) and (C.5), we get two equations

1 + 2 =
,
r2
(C.6)
+ 2.
4
2r
4r
Solving (C.6) and (C.7) with respect to 1 and 2 , we obtain

2

1
2
(1 , 2 ) =
+
4 2, 2
4 2 .
2
2r 2 2
r
r 2r
r
r
1 2 =
(C.7)
(C.8)
Substituting back this solution into the action (C.1) and performing the unitary transformation
(4.2), we arrive at the action (5.11).
Appendix D. T () Q1 case
Here we consider the n = 3 case in the sigma model with T ()Qn2 , say, T () [SO(3)/SO(2)]
T () CP 1 . In this case, everything is drastically simplified. The Khler potential (1.8) in this case
reduces to the CP 1 s one:

2

||2 ()2 ()
||2
K i , j = ln 1 + 2 +
(D.1)
=
2
ln
1
+
.
r
4r 4
2r 2
Eqs. (4.15)(4.17) become
= ,
+ = 0,
+ = 2,
(D.2)
2 ,
= 0.
Now the transformation factor VQ in Eq. (4.19) can be simply written as

||2
= VQ 0 = 1 +
0 .
2r 2
(D.3)
(D.4)
(D.5)
Substituting these expressions into (4.26), we find

=
||2
) +
2r 2
.
2
||
2r 2
2r 2
(1 +
1+
(D.6)
This result coincides with one of (3.15) in the n = 2 case with rescaling r r/ 2.
We can also see that the actions (4.10) and (5.11) reduce to ones of T CP 1 and T CP 1 in the
n = 3 case. In this case, the covariant quantities , , and become
1 2
G ||4 ,
r2
1
G1 ||2 ,
2 G2 ||4 ,
r
and Eq. (C.8) and f (i ) are
G||2 ,
1 , 2
||2
,
2r 2 G
(D.7)
(D.8)
(D.9)
232
f (1 ), f (2 ) f (),
(D.10)

2
). Using these formulas, we find
where G = (1 + ||2 /(2r 2 ))1 and f () = 12 (1 + 1 + 2||
r2G
that the actions for tangent and cotangent
bundles
coincide
with
(3.19) and (3.23) in the n = 2
case, respectively, with rescaling r r/ 2.

References
[1] M.F. Atiyah, N.J. Hitchin, V.G. Drinfeld, Y.I. Manin, Construction of instantons, Phys. Lett. A 65 (1978) 185.
[2] W. Nahm, A simple formalism for the BPS monopole, Phys. Lett. B 90 (1980) 413.
[3] M.F. Atiyah, N.J. Hitchin, Low-energy scattering of non-Abelian monopoles, Phys. Lett. A 107 (1985) 21;
M.F. Atiyah, N.J. Hitchin, Low-energy scattering of non-Abelian magnetic monopoles, Philos. Trans. R. Soc. London, Ser. A 315 (1985) 459.
[4] T. Eguchi, A.J. Hanson, Asymptotically flat selfdual solutions to Euclidean gravity, Phys. Lett. B 74 (1978) 249;
T. Eguchi, P.B. Gilkey, A.J. Hanson, Gravitation, gauge theories and differential geometry, Phys. Rep. 66 (1980)
213;
G.W. Gibbons, S.W. Hawking, Gravitational multi-instantons, Phys. Lett. B 78 (1978) 430;
N.J. Hitchin, Polygons and gravitons, in: G.W. Gibbons, S.W. Hawking (Eds.), Euclidean Quantum Gravity, pp.
527538.
[5] P.B. Kronheimer, The construction of ALE spaces as hyper-Khler quotients, J. Diff. Geom. 29 (1989) 665.
[6] P.B. Kronheimer, H. Nakajima, YangMills instantons on ALE gravitational instantons, Math. Ann. 288 (1990)
263;
H. Nakajima, Instantons on ALE spaces, quiver varieties, and KacMoody algebras, Duke Math. J. 76 (1994) 365.
[7] L. Alvarez-Gaume, D.Z. Freedman, Geometrical structure and ultraviolet finiteness in the supersymmetric sigma
model, Commun. Math. Phys. 80 (1981) 443.
[8] U. Lindstrm, M. Rocek, Scalar tensor duality and N = 1, N = 2 nonlinear sigma models, Nucl. Phys. B 222 (1983)
285.
[9] N.J. Hitchin, A. Karlhede, U. Lindstrm, M. Rocek, Hyper-Khler metrics and supersymmetry, Commun. Math.
Phys. 108 (1987) 535.
[10] P.C. Argyres, M.R. Plesser, N. Seiberg, The moduli space of N = 2 SUSY QCD and duality in N = 1 SUSY QCD,
Nucl. Phys. B 471 (1996) 159, hep-th/9603042.
[11] I. Antoniadis, B. Pioline, Higgs branch, hyper-Khler quotient and duality in SUSY N = 2 YangMills theories,
Int. J. Mod. Phys. A 12 (1997) 4907, hep-th/9607058.
[12] H. Pedersen, S.P. Yat, Hyper-Khler metrics and a generalization of the Bogomolny equations, Commun. Math.
Phys. 117 (1988) 569.
[13] G.W. Gibbons, R. Goto, P. Rychenkova, Hyper-Khler quotient construction of BPS monopole moduli spaces,
Commun. Math. Phys. 186 (1997) 585, hep-th/9608085.
[14] R. Bielawski, Complete hyper-Khler 4n-manifolds with n commuting tri-Hamiltonian vector fields, Math.
Ann. 314 (1999) 505, math.DG/9808134.
[15] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, K. Ohta, N. Sakai, Y. Tachikawa, Global structure of moduli space for
BPS walls, Phys. Rev. D 71 (2005) 105009, hep-th/0503033.
[16] U. Lindstrm, M. Rocek, R. von Unge, Hyper-Khler quotients and algebraic curves, JHEP 0001 (2000) 022, hep-th/
9908082.
[17] A. Borel, Khlerian coset spaces of semisimple Lie groups, Proc. Natl. Acad. Sci. 40 (1954) 1147.
[18] M. Bordemann, M. Forger, H. Romer, Homogeneous Khler manifolds: Paving the way towards new supersymmetric sigma models, Commun. Math. Phys. 102 (1986) 605.
[19] K. Itoh, T. Kugo, H. Kunitomo, Supersymmetric nonlinear realization for arbitrary Khlerian coset space G/H,
Nucl. Phys. B 263 (1986) 295.
[20] A.L. Besse, Einstein Manifolds, Springer-Verlag, Berlin, 1987.
[21] P.B. Kronheimer, A hyper-Khlerian structure on coadjoint orbits of a semisimple complex group, J. London Math.
Soc. 42 (2) (1990) 193.
[22] M.B. Stenzel, Ricci-flat metrics on the complexification of a compact rank one symmetric space, Manuscripta
Math. 80 (1993) 151.
[23] M. Cvetic, G.W. Gibbons, H. Lu, C.N. Pope, Ricci-flat metrics, harmonic forms and brane resolutions, Commun.
Math. Phys. 232 (2003) 457, hep-th/0012011.
233
[24] K. Higashijima, T. Kimura, M. Nitta, Supersymmetric nonlinear sigma models on Ricci-flat Khler manifolds with
O(N ) symmetry, Phys. Lett. B 515 (2001) 421, hep-th/0104184.
[25] O. Biquard, P. Gauduchon, Hyper-Khler metrics on cotangent bundles of Hermitian symmetric spaces, in: Geometry and Physics, Aarhus, 1995, in: Lecture Notes in Pure and Applied Mathematics, vol. 184, Dekker, New York,
1997, pp. 287298;
O. Biquard, P. Gauduchon, La mtrique hyperkhlrienne des orbites coadjointes de type symtrique dun groupe de
Lie complexe semi-simple, The hyper-Khler metric of coadjoint orbits of symmetric type of a complex semisimple
Lie group, C. R. Acad. Sci. Paris Sr. I Math. 323 (1996) 12591264.
[26] D. Kaledin, A canonical hyper-Khler metric on the total space of a cotangent bundle, Quaternionic Structures in
Mathematics and Physics, Univ. Studi Roma La Sapienza, Rome, 1999, pp. 195230, alg-geom/9710026.
[27] B. Feix, Hyper-Khler metrics on cotangent bundles, J. Reine Angew. Math. 532 (2001) 3346.
[28] A. Galperin, E. Ivanov, S. Kalitsyn, V. Ogievetsky, E. Sokatchev, Unconstrained N = 2 matter, YangMills and
supergravity theories in harmonic superspace, Class. Quantum Grav. 1 (1984) 469.
[29] A.S. Galperin, E.A. Ivanov, V.I. Ogievetsky, E.S. Sokatchev, Harmonic Superspace, Cambridge Monographs on
Mathematical Physics, Cambridge Univ. Press, 2001.
[30] A. Karlhede, U. Lindstrm, M. Rocek, Selfinteracting tensor multiplets in N = 2 superspace, Phys. Lett. B 147
(1984) 297.
[31] S.V. Ketov, B.B. Lokhvitsky, I.V. Tyutin, Hyper-Khler sigma models in extended superspace, Theor. Math. Phys. 71
(1987) 496, Teor. Mat. Fiz. 71 (1987) 226.
[32] U. Lindstrm, M. Rocek, New hyper-Khler metrics and new supermultiplets, Commun. Math. Phys. 115 (1988)
21.
[33] U. Lindstrm, M. Rocek, N = 2 super-YangMills theory in projective superspace, Commun. Math. Phys. 128
(1990) 191.
[34] S.V. Ketov, New N = 2 matter coupling in superspace, Int. J. Mod. Phys. A 3 (1988) 703.
[35] S.V. Ketov, Quantum Non-linear Sigma-Models: From Quantum Field Theory to Supersymmetry, Conformal Field
Theory, Black Holes and Strings, Springer, 2000.
[36] F. Gonzalez-Rey, M. Rocek, S. Wiles, U. Lindstrm, R. von Unge, Feynman rules in N = 2 projective superspace.
I: Massless hypermultiplets, Nucl. Phys. B 516 (1998) 426, hep-th/9710250.
[37] F. Gonzalez-Rey, R. von Unge, Feynman rules in N = 2 projective superspace. II: Massive hypermultiplets, Nucl.
Phys. B 516 (1998) 449, hep-th/9711135;
F. Gonzalez-Rey, Feynman rules in N = 2 projective superspace. III: YangMills multiplet, hep-th/9712128.
[38] S.M. Kuzenko, Projective superspace as a double-punctured harmonic superspace, Int. J. Mod. Phys. A 14 (1999)
1737, hep-th/9806147.
[39] S.M. Kuzenko, W.D. Linch, On five-dimensional superspaces, hep-th/0507176.
[40] S.J. Gates, S. Penati, G. Tartaglino-Mazzucchelli, 6D supersymmetry, projective superspace and 4D, N = 1 superfields, hep-th/0508187.
[41] S.M. Kuzenko, On compactified harmonic/projective superspace, 5D superconformal theories, and all that, hepth/0601177.
[42] J. Grundberg, U. Lindstrm, Actions for linear multiplets in six dimensions, Class. Quantum Grav. 2 (1985) L33.
[43] S.J. Gates, S.M. Kuzenko, The CNM-hypermultiplet nexus, Nucl. Phys. B 543 (1999) 122, hep-th/9810137.
[44] S.J. Gates, S.M. Kuzenko, 4D N = 2 supersymmetric off-shell sigma models on the cotangent bundles of Khler
manifolds, Fortschr. Phys. 48 (2000) 115, hep-th/9903013.
[45] S.J. Gates, T. Hubsch, S.M. Kuzenko, CNM models, holomorphic functions and projective superspace C-maps,
Nucl. Phys. B 557 (1999) 443, hep-th/9902211.
[46] S.J. Gates, M.T. Griaru, M. Rocek, W. Siegel, Superspace, or one thousand and one lessons in supersymmetry,
Front. Phys. 58 (1983) 1, hep-th/0108200.
[47] T.L. Curtright, D.Z. Freedman, Nonlinear sigma models with extended supersymmetry in four-dimensions, Phys.
Lett. B 90 (1980) 71;
T.L. Curtright, D.Z. Freedman, Phys. Lett. B 91 (1980) 487, Erratum;
L. Alvarez-Gaume, D.Z. Freedman, Ricci flat Khler manifolds and supersymmetry, Phys. Lett. B 94 (1980) 171;
M. Rocek, P.K. Townsend, Three loop finiteness of the N = 4 supersymmetric nonlinear sigma model, Phys. Lett.
B 96 (1980) 72.
[48] K. Higashijima, M. Nitta, Supersymmetric nonlinear sigma models as gauge theories, Prog. Theor. Phys. 103 (2000)
635, hep-th/9911139.
[49] K. Higashijima, T. Kimura, M. Nitta, M. Tsuzuki, Large-N limit of N = 2 supersymmetric QN model in two
dimensions, Prog. Theor. Phys. 105 (2001) 261, hep-th/0010272.
234
[50] F. Delduc, G. Valent, Classical and quantum structure of the compact Khlerian sigma models, Nucl. Phys. B 253
(1985) 494;
U. Ellwanger, Supersymmetric sigma models in four-dimensions as quantum theories, Nucl. Phys. B 281 (1987)
489.
[51] K. Higashijima, T. Kimura, M. Nitta, A note on conifolds, Phys. Lett. B 518 (2001) 301, hep-th/0107100.
[52] A.M. Perelomov, Chiral models: Geometrical aspects, Phys. Rep. 146 (1987) 135.
[53] L. Alvarez-Gaume, D.Z. Freedman, Potentials for the supersymmetric nonlinear sigma model, Commun. Math.
Phys. 91 (1983) 87.
[54] J. Bagger, C. Xiong, N = 2 nonlinear sigma models in N = 1 superspace: Four and five dimensions, hepth/0601165.
[55] S.M. Kuzenko, On superpotentials for nonlinear sigma-models with eight supercharges, hep-th/0602050.
[56] M. Arai, M. Naganuma, M. Nitta, N. Sakai, Manifest supersymmetry for BPS walls in N = 2 nonlinear sigma
models, Nucl. Phys. B 652 (2003) 35, hep-th/0211103;
M. Arai, M. Naganuma, M. Nitta, N. Sakai, BPS wall in N = 2 SUSY nonlinear sigma model with EguchiHanson
manifold, in: J. Arafune, et al. (Eds.), Garden of QuantaIn Honor of Hiroshi Ezawa, World Scientific, Singapore,
2003, pp. 299325, hep-th/0302028.
[57] M. Arai, M. Nitta, N. Sakai, Vacua of massive hyper-Khler sigma models of non-Abelian quotient, Prog. Theor.
Phys. 113 (2005) 657, hep-th/0307274;
M. Arai, M. Nitta, N. Sakai, Massive hyper-Khler sigma models and BPS domain walls, Phys. At. Nucl. 68 (2005)
1634, Yad. Fiz. 68 (2005) 1698, hep-th/0401102.
[58] E.R.C. Abraham, P.K. Townsend, Q kinks, Phys. Lett. B 291 (1992) 85;
E.R.C. Abraham, P.K. Townsend, More on Q kinks: A (1 + 1)-dimensional analog of dyons, Phys. Lett. B 295
(1992) 225;
J.P. Gauntlett, D. Tong, P.K. Townsend, Multi-domain walls in massive supersymmetric sigma-models, Phys. Rev.
D 64 (2001) 025010, hep-th/0012178;
D. Tong, The moduli space of BPS domain walls, Phys. Rev. D 66 (2002) 025013, hep-th/0202012.
[59] Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Construction of non-Abelian walls and their complete moduli space,
Phys. Rev. Lett. 93 (2004) 161601, hep-th/0404198;
Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Non-Abelian walls in supersymmetric gauge theories, Phys. Rev. D 70
(2004) 125014, hep-th/0405194;
M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, K. Ohta, N. Sakai, D-brane construction for non-Abelian walls, Phys. Rev.
D 71 (2005) 125006, hep-th/0412024.
[60] R.S. Ward, Slowly moving lumps in the CP 1 model in (2 + 1)-dimensions, Phys. Lett. B 158 (1985) 424;
I. Stokoe, W.J. Zakrzewski, Dynamics of solutions of the CP 1 and CP 2 models in (2 + 1)-dimensions, Z. Phys.
C 34 (1987) 491;
R.A. Leese, Q lumps and their interactions, Nucl. Phys. B 366 (1991) 283;
E. Abraham, Nonlinear sigma models and their Q lump solutions, Phys. Lett. B 278 (1992) 291.
[61] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Moduli space of non-Abelian vortices, hep-th/0511088;
M. Eto, T. Fujimori, Y. Isozumi, M. Nitta, K. Ohashi, K. Ohta, N. Sakai, Non-Abelian vortices on cylinder: Duality
between vortices and walls, hep-th/0601181.
[62] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Webs of walls, Phys. Rev. D 72 (2005) 085004, hep-th/0506135;
M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Non-Abelian webs of walls, Phys. Lett. B 632 (2006) 384,
hep-th/0508241;
M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, K. Ohta, N. Sakai, D-brane configurations for domain walls and their
webs, AIP Conf. Proc. 805 (2005) 354, hep-th/0509127.
[63] Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, All exact solutions of a 1/4 BogomolnyiPrasadSommerfield equation,
Phys. Rev. D 71 (2005) 065018, hep-th/0405129.
[64] J.P. Gauntlett, R. Portugues, D. Tong, P.K. Townsend, D-brane solitons in supersymmetric sigma-models, Phys.
Rev. D 63 (2001) 085002, hep-th/0008221;
N. Sakai, D. Tong, Monopoles, vortices, domain walls and D-branes: The rules of interaction, JHEP 0503 (2005)
019, hep-th/0501207.
[65] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Instantons in the Higgs phase, Phys. Rev. D 72 (2005) 025011,
hep-th/0412048.
[66] M. Naganuma, M. Nitta, N. Sakai, BPS lumps and their intersections in N = 2 SUSY nonlinear sigma models,
Grav. Cosmol. 8 (2002) 129, hep-th/0108133;
R. Portugues, P.K. Townsend, Sigma-model soliton intersections from exceptional calibrations, JHEP 0204 (2002)
039, hep-th/0203181.
235
[67] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, 1/2, 1/4 and 1/8 BPS equations in SUSY YangMillsHiggs systems:
Field theoretical brane configurations, hep-th/0506257;
M. Eto, M. Nitta, K. Ohashi, D. Tong, Skyrmions from instantons inside domain walls, Phys. Rev. Lett. 95 (2005)
252003, hep-th/0508130;
J.P. Gauntlett, D. Tong, P.K. Townsend, Supersymmetric intersecting domain walls in massive hyper-Khler sigma
models, Phys. Rev. D 63 (2001) 085001, hep-th/0007124.
[68] D. Tong, TASI lectures on solitons, hep-th/0509216.
[69] M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Solitons in the Higgs phasethe moduli matrix approach,
hep-th/0602170.
[70] Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Walls and vortices in supersymmetric non-Abelian gauge theories,
hep-th/0410150, Proceedings of NathFest at PASCOS Conference Northeastern University, Boston, MA, August
2004, in press;
Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, Non-Abelian walls and vortices in supersymmetric theories, in: K. Hagiwara, et al. (Eds.), Proceedings of 12th International Conference on Supersymmetry and Unification of Fundamental
Interactions (SUSY 04), Tsukuba, Japan, 1723 June 2004, KEK, 2004, pp. 116, hep-th/0409110;
M. Eto, Y. Isozumi, M. Nitta, K. Ohashi, N. Sakai, AIP Conf. Proc. 805 (2005) 266, hep-th/0508017.
[71] A. Dancer, A. Swann, Hyper-Khler metrics of cohomogeneity one, J. Geom. Phys. 21 (1997) 218230.
[72] A. Swann, P. Kobak, Hyper-Khler potentials in cohomogeneity two, J. Reine Angew. Math. 531 (2001) 121139,
math.DG/0001024.
[73] K. Higashijima, T. Kimura, M. Nitta, Ricci-flat Khler manifolds from supersymmetric gauge theories, Nucl. Phys.
B 623 (2002) 133, hep-th/0108084;
K. Higashijima, T. Kimura, M. Nitta, Gauge theoretical construction of non-compact CalabiYau manifolds, Ann.
Phys. 296 (2002) 347, hep-th/0110216;
K. Higashijima, T. Kimura, M. Nitta, CalabiYau manifolds of cohomogeneity one as complex line bundles, Nucl.
Phys. B 645 (2002) 438, hep-th/0202064;
K. Higashijima, T. Kimura, M. Nitta, Construction of supersymmetric nonlinear sigma models on noncompact
CalabiYau manifolds with isometry, Nucl. Phys. B (Proc. Suppl.) 117 (2003) 867, hep-th/0210034.
[74] K. Higashijima, M. Nitta, Khler normal coordinate expansion in supersymmetric theories, Prog. Theor. Phys. 105
(2001) 243, hep-th/0006027;
K. Higashijima, E. Itou, M. Nitta, Normal coordinates in Khler manifolds and the background field method, Prog.
Theor. Phys. 108 (2002) 185, hep-th/0203081.
[75] B. de Wit, M. Rocek, S. Vandoren, Hypermultiplets, hyper-Khler cones and quaternion-Khler geometry,
JHEP 0102 (2001) 039, hep-th/0101161.
Non-constant non-commutativity in 2d field theories

and a new look at fuzzy monopoles
A. Stern
Department of Physics, University of Alabama, Tuscaloosa, AL 35487, USA
Received 16 February 2006; accepted 4 April 2006
Abstract
We write down scalar field theory and gauge theory on two-dimensional non-commutative spaces M
with non-vanishing curvature and non-constant non-commutativity. Usual dynamics results upon taking the
limit of M going to (i) a commutative manifold M0 having non-vanishing curvature and (ii) the noncommutative plane. Our procedure does not require introducing singular algebraic maps or frame fields.
Rather, we exploit the Khler structure in the limit (i) and identify the symplectic two-form with the volume two-form. As an example, we take M to be the stereographically projected fuzzy sphere, and find
magnetic monopole solutions to the non-commutative Maxwell equations. Although the magnetic charges
are conserved, the classical theory does not require that they be quantized. The non-commutative gauge
field strength transforms in the usual manner, but the same is not, in general, true for the associated potentials. We develop a perturbation scheme to obtain the expression for gauge transformations about limits (i)
and (ii). We also obtain the lowest order SeibergWitten map to write down corrections to the commutative
field equations and show that solutions to Maxwell theory on M0 are stable under inclusion of lowest order
non-commutative corrections. The results are applied to the example of non-commutative AdS2 .
1. Introduction
Much work has been carried out for field theories on the non-commutative plane. This is the
case of constant non-commutativity. On the other hand, not much is known for field theories on
spaces with non-constant non-commutativity. Exceptional cases are when the non-commutativity
is associated with certain Lie-algebra structures, such as the case with fuzzy spheres and fuzzy
E-mail address: astern@bama.ua.edu (A. Stern).

doi:10.1016/j.nuclphysb.2006.04.001
A. Stern / Nuclear Physics B 745 [PM] (2006) 236259
237
CPn models (for a review, see [1]). Other cases have been discussed in [2,3]. Among the obstacles
to constructing field theories on general non-commutative spaces are problems in defining differentiation, integration and a Dirac operator.1 Recently, scalar field theory [5] and gauge theory [6]
have been formulated on general curved non-commutative manifolds. Although the procedures
in [5] and [6] differ, both make use of frame fields induced by a non-constant metric. They were
associated with algebraic maps to the non-commutative plane in [5], and appear in the definition
of derivations in [6]. A loss of generality may result in the non-commutative theory, however, as
frame fields are only defined on local coordinate patches. In [5] this resulted in the possibility of
singular algebraic maps in the non-commutative theory. Below we shall develop an alternative
approach for two-dimensional non-commutative scalar and gauge theories which avoids the introduction of frame fields. We also avoid the problem of defining non-commutative derivatives
by writing the theory algebraically from the start, expressing the underlying commutative theory
in terms of a Poisson bracket algebra.
In writing down field theories on some non-commutative space M, we shall require that the
results be consistent with deformations of known theories. In particular, we insist that the field
theories reduce to the standard form in the limit that M reduces (i) to a commutative manifold
with non-vanishing curvature and (ii) to the non-commutative plane. As stated above, we restrict
to two-dimensional field theories. In that case we can exploit the Khler structure of the commutative space and identify the symplectic two-form with the volume two-form. The Lagrangian
densities for the commutative theories can then be expressed in terms of Poisson brackets. In
passing to the non-commutative theory we simply replace the Poisson bracket by an appropriate
commutator and the classical measure by its non-commutative counterpart. This can be done
without spoiling the symmetries of the underlying commutative space M0 , if there are any, since
the Poisson brackets can be constructed to preserve these symmetries. The resulting free noncommutative scalar field and Maxwell equations have a simple form. Concerning the latter, there
are no propagating degrees of freedom, just as with commutative electrodynamics in two dimension, and solutions are characterized by a constant flux per unit area and action per unit area. The
non-commutative field strength is covariant with respect to gauge transformations. However, the
corresponding transformations of the potentials are nontrivial and geometry dependent.
We shall apply the results to find magnetic monopole solutions on the fuzzy sphere. Electrodynamics on the fuzzy sphere has been of considerable interest [714]. Anstse for magnetic
monopoles on the fuzzy sphere have been proposed [15], although they were not required to be
deformations of monopoles on the commutative sphere. Non-commutative magnetic monopoles
were expressed using the analogue of embedding coordinates in [12], and had the correct commutative limit. Here we obtain magnetic monopoles as solutions to the non-commutative Maxwell
equations on the analogue of the projective plane. A nonsingular map from the fuzzy sphere to the
non-commutative projective plane was given in [16]. (The coordinate singularity appears only in
the commutative limit.) Since it is nonsingular, one can express the non-commutative potentials
free of Dirac-string singularities. We find that the associated magnetic charges are conserved,
although not for topological reasons, although they need not be quantized, at least at the classical
level. Alternatively, we can get charge quantization, upon imposing additional constraints, but
these charges have a singular commutative limit.
1 On the other hand, such problems do not arise if one is only interested in doing particle mechanics on these spaces,
as for example is done in [4].
238
It is common to realize the non-commutative algebra with a star product. The Groenewold
Moyal star product is often used, but since it is associated with constant non-commutativity, it
is not very convenient to realize the algebra on M. More appropriate is Konsevichs formality
map [17] which was utilized in [6]. Alternatively, we shall rely upon the star product developed
in [16] which is based on a non-linear deformation of coherent states on the complex plane [18].
An exact integral expression for this star product was given, which can be expanded about either
limit (i) or (ii). An expansion for the measure can also be given by simply demanding that the
result satisfies the usual properties of a trace. By applying these expansions we get corrections to
the scalar and Maxwell actions about the two limits. Although approximation schemes for these
actions have been given previously [5,6], the one presented here has the advantage of simplicity.
Concerning the scalar field action, we get that the lowest order effects of non-commutativity are
obtained by just replacing derivatives on M0 by covariant ones. We also compute lowest order
corrections to the commutative expression for gauge transformations of the potentials, and show
how to SeibergWitten map [19] these corrections away. Using the SeibergWitten map we can
also compute corrections to the commutative flux through any region, as well as to the Maxwell
equations and its solutions. As expected from the exact theory, the flux per unit area and action
per unit area are constants for the solutions, but their values are shifted from the commutative
results. Because the shift is small (i.e. of the order of the non-commutativity parameter) we say
that solutions to the commutative theory are stable under inclusion of the non-commutative corrections. As an example, we apply the techniques to the case where M is the non-commutative
analogue of the Lobachevsky plane. This space is defined by a projection from non-commutative
AdS2 . Here we show how to obtain corrections to the solutions to the commutative free scalar
field theory. The solutions to the commutative Maxwell equations receive no first order corrections.
We review scalar field theory and gauge field theory in the commutative limit (i) in Section 2 and the non-commutative plane limit (ii) in Section 3. Field theories on spaces with
non-vanishing curvature and non-constant non-commutativity are described in Section 4. In Section 5 we apply the results to the fuzzy sphere and obtain the analogue of magnetic monopole
solutions. The first order corrections away from the two limits are computed in Section 6. In
Section 7, we apply the results to the example of non-commutative Lobachevsky plane. Brief
remarks are made in Section 8.
2. Curved spacecommutative theory
Here we take advantage of the Khler structure of any two-dimensional commutative manifold
M0 to express scalar field and gauge field Lagrangians on any coordinate patch P0 of M0 in
terms of Poisson brackets. Let g denote the metric tensor associated with P0 , parameterized
by real coordinates x , = 1, 2. Alternatively, we can define complex coordinates z = x 1 + ix 2
and z = x 1 ix 2 . We introduce a function 0 (z, z ), which we assume is non-singular, and the
commutative measure d0 (z, z ) on P0 by writing the area two form as
g d 2x =
i dz d z
2 d0 (z, z ).
0 (z, z )
(2.1)
This can be identified with a symplectic two-form, with a corresponding Poisson bracket { , }. So
if and are functions of z and z their Poisson bracket is
),
{, } = i0 (z, z )(
(2.2)
239

where = z
, = z . Its integral over P0 with respect to the measure d0 (z, z ) vanishes provided and vanish sufficiently rapidly as the boundary of P0 is approached. More generally,
for some region in P0 , the integral is equal to boundary terms:

(2.3)
d0 (z, z ) {, } = dz + d z = dz d z ,
where is the boundary of .

In writing down scalar field theory we shall choose the conformal gauge. In that case, the free
action for a massless scalar field is
S0 = i dz d z ,
(2.4)
which can then be re-expressed in terms of Poisson brackets

S0 = 2 d0 (z, z ) 0 (z, z )1 {z, }{z, }.
(2.5)
It is not necessary to choose a gauge restricting the metic tensor in the case of gauge theories.
For this introduce a potential one form a = dz a + d z a on P0 which gauge transforms as
a a + d.
(2.6)
The invariant field strength two-form is

a)
f = if dz d z = (a
dz d z .
(2.7)
Using (2.3) the flux 0 through any region can be expressed as an integral of Poisson brackets
of a and a

0
= f = a = 2 d0 (z, z ) {z, a} + {z, a}
(2.8)
,
having no dependence on the metric, since 0 appearing in the measure cancels with 0 appearing
in the Poisson brackets. In two dimensions the standard quadratic field action depends on the
metric only through its determinant, and like the flux, its integrand can be expressed solely in
terms of the Poisson brackets of a and a

2
d x 2 i
Sf0 =
(2.9)
dz d z 0 (z, z )f 2
f =
g
4

2
=
d0 (z, z ) {z, a} + {z, a}
.
(2.10)
2
In comparing with the free scalar field action (2.5), here we have not specified a gauge for the
metric and 0 does not appear explicitly in the integrand, despite its implicit dependence. There
are no propagating degrees of freedom for this system. Rather, the equations of motion imply
that
f=
C0
,
0 (z, z )
(2.11)
240
where C0 is the constant associated with the flux per unit area
C0 =
0
.
2 d0 (z, z )

(2.12)
The action per unit area of the solution is also a constant, namely C02 /4.
3. Flat spacenon-commutative theory
Now we review field theory on the non-commutative plane. This is the case of constant noncommutativity and no curvature. The algebra is generated by the operator z and its Hermitean
conjugate z , satisfying

z, z = k ,
(3.1)
where k denotes the non-commutativity parameter. It is standardly realized using the GroenewoldMoyal star product M

k
M = exp
(3.2)
,
2 z z z z
where the complex coordinates z and z are now symbols of z and z , respectively. Upon defining
the star commutator of any two functions A and B on the complex plane according to [A, B]M
A M B B M A one has, for example, [z, z ]M = k . The standard convention for the integration
measure is
dM (z, z ) =
i
dz d z
2 k
(3.3)
k has units of length-squared and hence the measure is dimensionless, unlike the commutative
measure in (2.1). A well known identity results from the star product M

dM (z, z ) A M B = dM (z, z )AB,
(3.4)
where A and B vanish sufficiently rapidly at infinity, and from this, the cyclic property of trace
easily follows.
The free scalar field action on the non-commutative plane is well known

2
dM (z, z )[z, ]M M [z, ]M
SM =
k

i
= 2 dz d z [z, ]M [z, ]M ,
(3.5)
k
and from the fact that derivatives in z and z are realized by = k1 [z, ]M and = k1 [z, ]M ,
respectively, (3.5) is identical to the free scalar field action on the commutative plane.
For gauge theories on the non-commutative plane we replace the potential one form a by
Infinitesimal gauge variations by of A and A are given by
dz A + d z A.
1
A = [z, ]M i[A, ]M ,
k
1
]M .
A = [z, ]M i[A,
k
(3.6)
241
The field strength two-form is iFM dz d z , where

1
1
M ,
M + i[A, A]
iFM = [z, A]M + [z, A]
k
k
which transforms covariantly under gauge transformations,
FM = i[FM , ]M ,
(3.7)
(3.8)
FM can be also be expressed as

M + 1 [z, z ]M ,
FM = [Z, Z]
k 2
where
i
i
Z = z + A,
Z = z + A,
k
k
which also transform covariantly,
Z = i[Z, ]M ,
]M .
Z = i[Z,
The standard gauge theory action on the non-commutative plane is

k
M
dM (z, z )FM M FM
Sf =
2

i
2
dz d z FM
=
.
4
(3.9)
(3.10)
(3.11)
(3.12)
When k 0, the star commutator goes to ik times the Poisson bracket (2.2), with 0 (z, z ) equal
to one, and so (3.12) reduces to Sf0 with the flat metric. The free field equations following from
variations of A and A are
FM ]M = 0.
[Z, FM ]M = [Z,
(3.13)
They are solved for FM proportional to the identity. For the case of a pure gauge solutions
(FM = 0), Z and Z are given by
i
Zpg = U M z M U,
k
i
Zpg = U M z M U,
k
(3.14)
where U are unitary functions on the complex plane with regard to the GroenewoldMoyal star
product, U M U = U M U = 1.
It is straightforward to couple the scalar field to gauge theories on the non-commutative plane.
For this replace (3.5) by

M
]M M [Z, ]M
=
2
k
dM (z, z )[Z,
S,
Z

]M [Z, ]M .
= i dz d z [Z,
(3.15)
242
Gauge invariance follows from (3.11) and = i[, ]M . The coupled field equations resultM are
ing from variations of the combined action SfM + S,
Z

[Z, FM ]M + 2 , [Z, ]M = 0,
M

]M
FM ]M 2 , [Z,
= 0,
[Z,
M

]M
[Z, ]M
Z, [Z,
(3.16)
+ Z,
= 0.

M
4. Curved spacenon-commutative theory

In the previous section, field theories on the non-commutative plane can be expressed in terms
of commuting inner derivatives = k1 [z, ]M and = k1 [z, ]M , satisfying the usual Leibniz
rule. This, however, is not in general possible for non-constant non-commutativity. Fortunately,
two dimensional non-commutative field theories can be expressed purely algebraically, without
relying on the notion of derivatives. Non-constant non-commutativity in two dimensions means
we replace (3.1) by

z, z = z, z
(4.1)
for some function (z, z ). (4.1) defines an algebra associated with some non-commutative manifold M. In addition to being a function of the generators z and z of the algebra, depends
on an additional parameter, the non-commutativity parameter, which we again denote by k . The
GroenewoldMoyal star product is not very convenient to realize this algebra since then z and
z appearing in its definition (3.2) cannot be symbols of z and z . A more convenient associative
star product was developed in [16] which has an exact integral expression, and will be reviewed
in Subsection 6.1. Here we denote it by , and so if z and z C are symbols of z and z , respectively, then
[z, z ] z z z z = (z, z ),
(4.2)
where (z, z ) denotes the symbol of

Now in general we will not have the analogue
of the identity (3.4). On the other hand, the appropriate integration measure d(z, z ) will be
required to satisfy

d(z, z ) [A, B] = 0
(4.3)
(z, z ).
for any functions A and B that fall off sufficiently rapidly at infinity. (4.3) corresponds to the
cyclic property of the trace. In Subsection 6.2 we shall use this property and the definition of
the to perturbatively construct the measure. We assume that like the measure dM (z, z ) on the
non-commutative plane, d(z, z ) is dimensionless.
To recover the systems of the previous two sections we will need to examine two limits:
(i) The commutative limit. This is k 0. We assume that (z, z ), and consequently (z, z ),
is linear in k in this limit,
(z, z ) k 0 (z, z ),
(4.4)
where 0 (z, z ) is a dimensionless function which is independent of k . We shall identify it with

0 (z, z ) appearing in (2.1) and (2.2). As is usual, we require that the star product goes to the
point-wise product, and the star commutator goes to i times the Poisson bracket in this limit. The
243
commutative limit of the measure d(z, z ) is d0 (z, z )/k , where d0 (z, z ) was given in (2.1)
(k is introduced since d0 (z, z ) has units of length-squared).
(ii) The non-commutative plane limit. This is
(z, z ) k .
(4.5)
The combination of both of the above limits gives the commutative plane.
Ordering ambiguities occur in deforming the free scalar actions (2.4) and (3.5) of the previous
two sections to this general case. Moreover, here we need that (z, z ) is nonsingular. We can
choose the ordering such that the general scalar field action is

z, ] ,
S = 2 d(z, z ) (z, z )1
(4.6)
[z, ] [
1
1
where (z, z )1
is defined by (z, z ) (z, z ) = (z, z ) (z, z ) = 1. One easily recovers
(2.4) from (4.6) in limit (i), and (3.5) in limit (ii). The field equation following from variations of
in (4.6) read

[z, ] 1 , z + 1 [z, ] , z = 0.
(4.7)
It has the trivial solutions = z and = z , as well as the constant solution, but the analytic and
anti-analytic solutions of the commutative theory, = + (z) and = (z), are not in general
present in the non-commutative theory.
The situation is more straightforward for gauge theories. Since the commutative action (2.10)
could be expressed without explicit reference to 0 the above ordering ambiguity does not arise,
and, moreover, no spacetime gauge was necessary in writing down (2.10). Upon again introducing potentials A and A we can define the field strength by
1
1
,
+ i[A, A]
iF = [z, A] + [z, A]
k
k
and generalize the commutative flux (2.8) in some region on the complex plane to

= 2 k d(z, z )F.
(4.8)
(4.9)
From (4.3), it is zero if A and A vanish on the boundary of . The action (3.12) on the
non-commutative plane can be generalized to

k
d(z, z ) F F.
Sf =
(4.10)
2
In the commutative limit (i), F 0 f , and (4.10) reduces to the commutative Maxwell action
(2.10). In the non-commutative limit (ii), the field strength (4.8) becomes (3.7) and we recover
(3.12) from (4.10). The non-commutative Maxwell equations following from variations of A and
A in (4.10) (ignoring boundary terms) are
F ] = 0,
[Z, F ] = [Z,
(4.11)
where Z and Z were defined in (3.10). They are again solved for F proportional to the identity.
Then from (4.9) and (4.10), respectively, the flux per unit area and action per unit area for any
such solutions are constants, just
as in the commutative case. Now the area of any region on
the complex plane is given by d(z, z ). We have not found a simple expression for pure gauge
244
solutions, analogous to those on the non-commutative plane (3.14), although an expansion about
the commutative answer can be obtained. We do this in Subsection 6.4.
The issue of gauge invariance of the action (4.10) is more complicated than it was for the
previous two limits. Applying (4.3), gauge invariance of the action follows if the field strength
transforms covariantly, i.e. variations are of the form
F = i[F , ] ,
(4.12)
for infinitesimal . Although the field strength transforms in a simple manner, the same is not,
For example, something analogous to (3.6) will not
in general, true for the potentials A and A.
work because does not have zero star-commutator with . The gauge symmetry of the action
is therefore hidden. Here it does not help to introduce the quantities Z and Z defined in (3.10)
and express F according to
+ 1 (z, z ),
F = [Z, Z]
k 2
(4.13)
In Subsection 6.4 we
in analogy to (3.9). Since (z, z ) is not covariant, neither can be Z and Z.
or equivalently Z and Z,
gauge
develop a perturbation scheme for determining how A and A,
transform away from limits (i) and (ii).
The scalar field action coupled to gauge theories on the non-commutative plane (3.15) can be
generalized to arbitrary (z, z ) by
S,Z = 2 k 2 d(z, z ) (z, z )1

(4.14)
[Z, ] [Z, ] .
For gauge invariance we need that

i
[Z, ] = [Z, ] , ,
k

] = i [Z,
] , .
[Z,

k
(4.15)
It is then clear that the scalar field cannot, in general, transform covariantly. After developing
a perturbation scheme for the gauge transformation of the potentials, one can then use (4.15) to
do the same for . The coupled fields equations (3.16) are then generalized to

[Z, F ] + 2k , [Z, ] 1 = 0,

F ] 2k , 1 [Z,
] = 0,
[Z,

1

1
[Z, ] , Z + [Z, ] , Z = 0.
(4.16)

5. Magnetic monopoles on the fuzzy sphere

Fuzzy spaces are standardly defined to be non-commutative theories with finite dimensional
matrix representations. So in that case the generators z and z of the algebra in (4.1) are represented N N matrices. This also applies to the fields , Z and Z which are polynomials
functions of z and z . The star can be replaced by ordinary matrix multiplication, and so the
star commutator can be replaced by the matrix commutator. Integration corresponds to taking
the trace. Specializing to gauge theories, one gets that the field strength is traceless and the total
flux vanishes Tr F = 0. Furthermore, the constant solution, i.e. F proportional to the identity, to
245
the non-commutative Maxwell equations (4.11) collapses to the trivial solution, i.e. F = 0. This
indicates the absence of any magnetic monopole solutions in a fuzzy physics. In deriving (4.11)
one assumed arbitrary variations of the gauge fields in the Maxwell action (4.10). The negative
result for monopoles can be avoided if we restrict variations of Z and Z to block diagonal matrices. In that case F has solutions which are block diagonal matrices, with the individual blocks
being proportional to identity matrices, and their combined trace equal to zero. We shall use this
technique to construct fuzzy magnetic monopole solutions in Subsection 5.2.2 In Subsection 5.1
we review the commutative case.
5.1. Commutative case
We shall fix the radius of the sphere to be 1, so in terms of embedding coordinates xi , i =
1, 2, 3, x12 + x22 + x32 = 1. Poisson brackets which preserve the SO(3) symmetry are
{xi , xj } = ij k xk .
(5.1)
In defining gauge theory, one can introduce 3-potentials ai , i = 1, 2, 3, which gauge transform
like [20]
ai ai + {xi , },
(5.2)
for some function on the sphere. A constraint should be imposed on ai as there are only two
independent gauge potentials on the surface. It should not restrict the gauge transformations
(otherwise it would be a gauge condition), and be invariant under rotations. This is the case for
ai xi = 0.
(5.3)
From this one gets the identity

xj {xi , aj } = ij k xj ak ,
(5.4)
in addition to
xi {xi , aj } = 0.
(5.5)
A gauge invariant scalar is

b = ij k xi {xj , ak },
(5.6)
which can be interpreted as the magnetic flux density normal to the surface. The Maxwell action
on the sphere is

1
d b2 ,
Sf0 =
(5.7)
4
where the integral is over the solid angle .
To recover the formalism of Section 2, we can stereographically project to the complex plane,
where the north pole is mapped to infinity thus corresponding to a coordinate singularity
z=
x1 ix2
,
1 x3
z =
x1 + ix2
.
1 x3
(5.8)
2 Although the procedures differ, reducible representations were also necessary for describing monopoles in [12].
246
The Poisson structure is then projected to

{z, z } = i0 |z|2 ,
1
2
0 |z|2 = 1 + |z|2 ,
2
while the potentials ai are mapped to the one form a = dz a + d z a,
where
2ia = (1 x3 )(a1 + ia2 ) + (x1 + ix2 )a3 ,
and a is the complex conjugate of a. The inverse map is

a1 ia2 = i a + z2 a ,
a1 + ia2 = i a + z 2 a ,
(5.9)
(5.10)
a3 = i(za z a).
(5.11)
From (5.2) one recovers the gauge transformations (2.6). The magnetic flux density (5.6) is
mapped to3
b = {z, a} + {z, a}.
(5.12)
Thus, by doing the sterographic projection of the Maxwell action (5.7) we are able to recover the
expression (2.10).
g0
The magnetic monopole solutions are b = C0 = 4
, or
f=
1
g0
,
2 (1 + |z|2 )2
(5.13)
g0 2
) .
where g0 is the magnetic charge. The Maxwell action (5.7) evaluated for this solution is ( 4
Potentials can be given after removing the point at infinity, the location of the Dirac string,
a=
ig0 z dz z d z
.
4 1 + |z|2
(5.14)
5.2. Non-commutative case

In going to the fuzzy sphere, we replace real coordinates xi by Hermitean operators xi , satisfying commutation relations:
[xi , xj ] = ik ij k xk ,
(5.15)
as well as xi xi =
1, 1 being the unit operator. When the non-commutativity parameter k has
values k J = 1/ J (J + 1), J = 12 , 1, 32 , . . . , xi have finite dimensional representations. For a
given J , one can set xi = k J Ji , Ji being the angular momentum matrices associated with the
(2J + 1)-dimensional irreducible representation J acting on Hilbert space H J with states
|J, M, M = J, J + 1, . . . , J . The commutative limit is J corresponding to infinitedimensional representations.
A non-singular fuzzy stereographic projection was given in [16].4 It is defined up to an operator ordering ambiguity to be
z = (x1 ix2 )(1 x3 )1 ,
z = (1 x3 )1 (x1 + ix2 ).
(5.16)
3 To prove this substitute (5.10) in (5.12) to get (1 x )({z, a} + {z, a})

= b + ij 3 {xi , aj } a3 and use the identity
3
x3 b = a3 ij 3 {xi , aj }, which follows from (5.4) and (5.5).
4 A singular one was given in [21].
247
This is a non-singular map because 1 is excluded from the spectrumof x3 , except in the commutative limit J . For the top state M = J , x3 has eigenvalue J / J (J + 1), which approaches
1 in the limit, and we thereby recover the coordinate singularity. Using properties of angular
momentum matrices, (z, z ) is represented by a diagonal matrix

z, z |J, M = J,M |J, M,
(5.17)
with elements
J (J + 1) M(M + 1)
J (J + 1) M(M 1)

.
J,M =
(5.18)
2
( J (J + 1) M 1)
( J (J + 1) M)2
When evaluated on the top state one has J,J = 2J /( J (J + 1) J )2 , which goes like 8J
in the limit J .
In constructing gauge theories, as stated previously, the field strength expressed in terms of
potentials, (4.8) or (4.13), is traceless, and as a result the total flux vanishes, i.e. Tr F = 0. This
implies that there can be no magnetic monopoles in this formalism, and furthermore that the constant solution to the non-commutative Maxwell equations (4.11) collapses to the trivial solution,
i.e. F = 0. This is not surprising since also in the commutative theory, if we insist on writing the
field strength in terms of potentials globally, there can be no magnetic monopoles. The monopole potential in (5.14) is defined only after removing the point at infinity from the domain of the
commutative theory. In the non-commutative theory, this point is approached by the top state as
J . So let us similarly remove it from the domain of the non-commutative theory. Equivalently, we can restrict variations of the fields Z and Z to be block diagonal matrices, one block
being 2J 2J and the other being 1 1, the latter associated with the top state. Then the equations of motion (4.11) will only hold for the diagonal blocks. Solutions to the non-commutative
Maxwell equations (4.11) for F will then also be block diagonal matrices F2J 2J and F11 ,
which are proportional to the identity, and since Tr F = 0,
g
12J 2J ,
F2J 2J =
(5.19)
2J
F11 = g,
(5.20)
12J 2J being the 2J 2J identity matrix. The trace over only F2J 2J is g, which by analogy
with the commutative theory defines the magnetic charge, while F11 corresponds to the compensating flux of the Dirac string. So (5.20) is the non-commutative analogue of the Dirac string.
The action (4.10) evaluated for this solution gives
g 2 (2J + 1)
.
4J J (J + 1)
g0 2
) , we need that g grows like
In comparing with the commutative answer of ( 4
commutative limit, i.e.
J g0
, as J ,
g
(2)3/2
(5.21)
J in the
(5.22)
in order to recover a finite charge g0 in the commutative theory.

By equating (5.20) with the last row and column of the matrix associated with (4.13) we get
g=
2

J, M|Z|J, J 2
J, M|Z|J,
J 2 + 2J (J + 1)
.
( J (J + 1) J )2
M=J
(5.23)
248
Note that only off-block diagonal matrix elements of Z and Z are present in the result. These
are non-dynamical fields (they were not varied in obtaining the field equations), and so the magnetic flux is a constant of the motion. This is a result of the dynamics, and not topology. In the
commutative limit, the last term in (5.23) goes to infinite like 8J 3 , and thus the sum must go like
8J 3 in order for g to have the limit in (5.22). Alternatively, we can use (4.8) to write the charge
in terms of matrix elements of A and A
g=

J, M|A|J, J 2
J, M|A|J,
J 2
M=J

J 2(J + 1)
J .
J, J |A|J, J 1
J, J 1|A|J,
i
J (J + 1) J
(5.24)
Again, only the (non-dynamical) off-block diagonal matrix elements appear, and so g is a constant of the motion.
In the above, although the charge is a constant of motion, we get no quantization, at least at the
classical level.5 On the other hand, quantization does occur if we impose the stronger condition
and not just their variations, are block diagonal. Then the sum in (5.23)
that the fields Z and Z,
vanishes and g is just equal to the last term. We can also allow for more general block diagonal
matrices. So let Z and Z be reducible to (2J + 1 N ) (2J + 1 N ) and N N matrices,
1 N 2J . Solutions to the non-commutative Maxwell equations (4.11) for F will then be
block diagonal matrices F(2J +1N )(2J +1N ) and FNN , where
F(2J +1N )(2J +1N ) =
g
1(2J +1N )(2J +1N ) ,
2J + 1 N
g
1NN .
(5.25)
N
So now the non-commutative analogue of the Dirac string includes N states. By equating the
trace of either of the matrices in (5.25) with the corresponding trace of (4.13) we get
FNN =
N
1
g= 2
J,J +1n .
k J n=1
(5.26)
Using (5.18) we thereby get quantized magnetic charges, depending on J and N . Examples are
1
J= ,
2
N = 1,
J = 1,
N = 1,
N = 2,
3
J= ,
2
N = 1,
N = 2,
3
,
g=
( 3 1)2
4
,
g=
( 2 1)2
g = 2,
45
,
g=
( 15 3)2
30(8 15 31)
,
g=
95 15 368
5 Quantized magnetic charges were obtained in [12] upon requiring the gauge fields to be associated with reducible
SU(2) representations.
N = 3,
45
g=
.
( 15 + 1)2
249
(5.27)
These charges do not have a well defined limit when J . Here it does not help that we have
an additional parameter N . For N equals one, or close to one, g goes like 8J 3 as J , which
is too divergent when compared to (5.22). On the other side is N = 2J , where g = J,J /k 2 . So
when N equals 2J , or close to 2J , g goes like J /2 as J , which is also too divergent when
compared to (5.22). It follows that off-block diagonal matrices must be present for Z and Z in
order to recover the usual commutative limit, even though these matrices are non-dynamical.
6. Non-commutative corrections
In this section we do an expansion in k to obtain non-commutative corrections to the commutative scalar and gauge field actions. We also obtain corrections to the corresponding actions on
the non-commutative plane. For the latter we do an expansion in derivatives of (z, z ). The two
expansions are not independent as is explained below. We will also give an expansion for gauge
transformations of A and A about the two limits. The expression in general depends on the star
product and (z, z ), which in the commutative limit is related to the determinant of the metric.
So for this gauge theory, motion along a fibre depend on the geometry of the base manifold.
The non-commutative gauge theory can be SeibergWitten [19] (also see [2224]) mapped to
the commutative theory, leading to corrections to the commutative flux (2.8) and the Maxwell
action (2.10). We then get the corrections to the commutative solution (2.11). In Section 7 we
apply the techniques to the example of non-commutative AdS2 . There are a number of obstacles
in using the approximation scheme developed here for analyzing magnetic monopoles in fuzzy
gauge theories, which we comment on in Section 8.
6.1. Star product
We now review the star product in [16] which can be expressed in terms of the symbols z and
z C of operators z and z , respectively, and which easily reproduces the star commutator (4.2).
It is based on an overcomplete set of unit vectors {|z} spanning an infinite-dimensional Hilbert
space. The states |z are, in general, non-linear deformations of standard coherent states on the
complex plane. They diagonalize z,6
z|z = z|z.
(6.1)
The covariant symbols of operators A, B, . . . are given by A(z, z ) =

z|A|z, B(z, z ) =
z|B|z, . . . , and their star product by [A B](z, z ) =

z|AB|z. The expression for the star product was obtained in [16]. Given the symbols A(z, z ) and B(z, z ), then [A B](z, z ) can be written
as

( z):
z| :exp( z ) :B(z, z ),
A(z, z ) d(, ):exp
(6.2)
z
z
where
d(z, z ) is the appropriate measure on the complex plane satisfying the partition of unity

d(z, z )|z
z| = 1. The colons in (6.2) denote an ordered exponential, with the derivatives
6 It is problematic to construct such states for fuzzy manifolds. Alternatively, for the case of the fuzzy sphere we found
a set of states {|z} where the difference of z|z and z|z was proportional to the top state.[16] As a result, the expressions
which follow for the star product and measure get modified for the fuzzy sphere.
250
ordered to the right in each term in the Taylor expansion of exp( z) z

, and to the left in each
term in the Taylor expansion of exp ( z); i.e.
2
1
:exp( z ) : = 1 + ( z ) + ( z )2 2 + ,
z
z 2
z
1 2
:exp ( z): = 1 + ( z) +
( z)2 + .
z
z
2 z2
(6.3)
When (z, z ) =
z|(z, z )|z equals k , |z are the standard coherent states and the star product
reduces to the Voros product, which can be transformed to (3.2). A derivative expansion of the
above star product was performed in [25]. There one obtained the following leading three terms
acting between functions of z and z :

1 2
2
2
= 1 + (z, z ) +
(z, z )
+ (z, z )
+ .
z
z 4 z2 z
z z
z z 2
(6.4)
At lowest order in k , we assume that (z, z ) is given by (4.4). Since we interpret k as the
non-commutativity parameter, then the lowest order in the derivative expansion (6.4) gives the
commutative product, and from the first order term, one gets that the star commutator goes to i
times the Poisson bracket. Moreover, after expanding (z, z ) in k ,
(z, z ) = k 0 (z, z ) + k 2 1 (z, z ) + ,
(6.5)
the derivative expansion of the star can also be regarded as an expansion in k . So one can use
(6.4) to expand about the commutative field theory or the field theory on the non-commutative
plane.
6.2. Measure
We next expand the integration measure about its (i) commutative limit d0 (z, z )/k and (ii)
the non-commutative plane limit dM (z, z ). For this we require that the trace property (4.3)
holds order-by-order for functions A and B that fall off sufficiently rapidly at infinity. Using
(6.4) we then find

i dz d z
1
d(z, z ) =
(6.6)
1 +
(z, z ) + .
2 (z, z )
2
We can regard this as a derivative expansion, and thus a perturbation about the non-commutative
plane limit (ii). Eq. (6.6) can also be regarded as an expansion in k , and thus a perturbation about
the commutative limit (i). For the latter, apply (6.5) to get
d(z, z ) =

d0 (z, z )
1
1
0
+ .
1 + k
2
0
k
(6.7)
More generally, if real functions A and B and their derivatives are non-vanishing on the
boundary of some region . Then the integral of their star commutator can be expressed
on the boundary. The generalization of (2.3) gives
i
d(z, z ) [A, B] =
2
251
+ 1
AB
dz d z AB
2

1 2 2
A B + 2 A 2 B + (A B)
+
4

i
[A,B] J[A,B] )
dz d z (J
=
2

i
=
(dz J[A,B] + d z J[A,B] ),
2
(6.8)
with
1
J[A,B] = BA +
2
1
+
J[A,B] = B A
2

1
+ (A B),
A)
2 B A
B( A
4

1

A) 2 BA + (A B).
B( A
4
(6.9)
6.3. Field theory

We can now apply the previous expansions about limit (i) and (ii) to the scalar field and gauge
theory actions. Starting with the real scalar field, we get

+ 1 2 + ,
[z, ] =
2

1 2
+ .
[z, ] = +
(6.10)
2
After substituting into (4.6), and dropping boundary terms, we can write S , up to first order, by
simply replacing the ordinary derivatives in the commutative action (2.4) with covariant ones

S = i dz d z D D ,
(6.11)
where D and D are defined by7

1
D = 1 1
2

1 1 1
D =
2
1 2
+ ,
2
1 2
+ .
2
(6.12)
The result is quite simple when compared to previous approaches [5]. From the lowest order term
we recover the scalar field action in the conformal gauge, S0 . Conformal invariance is broken
by the non-commutative corrections. The lowest order corrections can be expressed in terms of
the determinant of the classical metric using (2.1) and (6.5). The field equation following from
variations of in (6.11) gives the conservation law
+ I = 0,
I
(6.13)
7 The geometric meaning of these derivatives is not immediately obvious. They may be associated with an automorphism between the star product and the commutative product [26].
252
with currents

1 2
+ (
) + ,
I = 1 1
2

1 2

1
1
+ (
)
+ .
I =
(6.14)
2
For gauge theories we can use (6.4) to compute lowest order corrections to the field strength
F in (4.8)

i 2
2
F=
(6.15)
FM A A + ,
2
k
where FM is the MoyalWeyl field strength. As a result we have corrections to the noncommutative plane limit, as well as the to commutative limit. For the latter, note that there
are terms of order k in FM , as well as in (z, z ) and its derivatives. The lowest order noncommutative corrections to the flux can be read off from (4.9). Using (6.8) and (6.9) one gets the
boundary terms

1
=
+
dz A ik A A + A
2

1
+ d z A + ik AA + A + .
(6.16)
2
Corrections to the gauge field action are

i
1
2
2 A
dz d z FM
iFM 2 A
Sf =
1 +
4 k
2

FM ) + .
+ 1 (FM )(
(6.17)
When (z, z ) = k we easily recover the expression for the non-commutative plane. A Seiberg
Witten map should be utilized to compare with the commutative case, which we do in Subsection 5.5.
6.4. Gauge transformations
Although the field strength transforms covariantly under gauge transformations, this is not
the case the potentials. Here we compute corrections to gauge transformations of A and A from
(i) the commutative limit (2.6) and (ii) the non-commutative plane limit (3.6). For this write
infinitesimal gauge variations as
A = + ,
+ ,
A =
and substitute into (4.12) using (6.15) to get

1
2
+ FM + ik A
+
ik A
2

1
+ FM
ik A 2 = 0,
+
+ ik A
2
(6.18)
(6.19)
253
up to the divergence of some arbitrary

up to first order in k . Then we can write down and ,
function

1
2 + ,
FM ik A
= ik A
2

1
FM
+ ik A 2 + .
= + ik A
(6.20)
2
The divergence of can be absorb in a re-definition of the gauge parameter = + ,
yielding
1
M + (k )FM + ,
A = i[A, ]
(6.21)
2
1
]
M
+ (k )FM + .
A = i[A,
(6.22)
2
From this we easily recover the expressions for limit (i) by setting (z, z ) = k = 0, and limit
(ii) after setting (z, z ) = k with = or = 0. Gauge transformations close after including the
first order corrections. If one goes beyond the leading order, the closure of gauge transformations
should put restrictions on . The corresponding gauge variations of Z and Z are
1
M + (k )FM + ,
Z = i[Z, ]
(6.23)
2
1
]
M
+ (k )FM + .
Z = i[Z,
(6.24)
2
Setting A and A equal to zero in (6.21) and (6.22) gives a first order expression for pure gauge
potentials
i
1
Zpg = z + + ,
2
k
i
1
+ ,
Zpg = z +
(6.25)
2
k
which generalizes the infinitesimal version of pure gauge solutions (3.14) to the non-commutative
plane.
6.5. SeibergWitten map
We now construct the lowest order map from the gauge theory written on a coordinate patch
P0 of commutative manifold M0 to the non-commutative gauge theory. (A more geometrical
treatment can be found in [26].) We write the non-commutative potentials and gauge parameter
a],
a, a],
as functions of the commutative ones, A[a, a],
A[a,
[,
and require that the right + , a +
A[a, a]
hand sides of (6.21) and (6.22) be equal to A[a + , a + ]
and A[a
a],
A[a,
]
respectively. The first order equation is then solved by
ik
(a a aa)
2
ik
a]
a a)

A[a,
= a(1
f ) (a a
2
ik
a, a]
a),
[,
= + (a
2
A[a, a]
= a(1 f ) +
1
a,
2
1
a,
2
(6.26)
254
a.
where f is the commutative curvature, if = a
Eq. (6.26) reduces to the standard Seiberg
Witten map to the Moyal plane when = k . The map to the non-commutative curvature up to
first order in k is

F [a, a]
=
(6.27)
.
f 1 + i (f a) i(f a)
2
k
Substituting into the non-commutative flux (6.16) gives, up to first order in k ,

dz a(1 f + ) + d z a(1
f + )
=
= 0 + 1 + ,
where
(6.28)
is the commutative flux (2.8) and

1
= k 0 f (dz a + d z a).
(6.29)
After substituting (6.27) into the action (4.10) we get the lowest order correction to the commutative action (2.10)

i
1
2
dz d z f 1 f + (f )(f )
Sf =
4 k
2
= Sf0 + Sf1 + ,
where Sf0 is the commutative Maxwell action (2.10) and

ik
1
1
2
2
Sf =
dz d z f 1 0 0 0 f + (0 f )(0 f ) ,
4
2
(6.30)
(6.31)
and we have dropped boundary terms. As in the commutative theory, there are no propagating
degrees of freedom. The field equations resulting from variations of a and a in (6.30) imply that

3
C
1
)= ,
f 1 f (f
(6.32)
2
2
C being a constant. If we apply the expansion (6.5), then (6.32) gives the solution for the corrected field strength
f = f 0 + f1 + ,
where the lowest order agrees with the commutative solution f0 = C0 /0 and

k C0 1
1
f1 =
0
,
0 2
0
(6.33)
(6.34)
and we also expanded the constant

3
C = C0 k C02 k 2 + .
2
(6.35)
255
From (2.12), C0 was equal to the flux per unit area of the solution in the commutative theory.
The flux per unit area remains a constant, but its value gets shifted from the commutative result
= C0 (1 k C0 + ),
2 k d(z, z )

(6.36)
where we used the corrected flux (6.28) and measure (6.6). The action per unit area of the solution, Sf = Sf00 +f1 + Sf10 , also remains a constant, its value being shifted from the commutative
result by the same factor appearing in (6.36),
C02
(1 k C0 + ).
(6.37)
4
Since these shifts are small for small k we can say that the solutions are stable under the inclusion
of non-commutative corrections.
7. Non-commutative AdS2
We now apply the results of the previous section to obtain the lowest order non-commutative
corrections to the scalar and gauge theory actions on the Lobachevsky plane and show that the
solution to the commutative Maxwell action receives no such corrections. For other approaches
to the non-commutative AdS2 and the Lobachevsky plane, see [27,28].
7.1. Lobachevsky plane
We first review the commutative theory. Here we write down the AdS2 measure on the
disc, which defines the Lobachevsky plane. We start with AdS2 embedded in three-dimensional
Euclidean space with coordinates xi , i = 0, 1, 2, and the constraint
x02 x12 x22 = 1,
x0 1.
(7.1)
A natural Poisson structure on it is

{x0 , x1 } = x2 ,
{x2 , x0 } = x1 ,
{x1 , x2 } = x0 ,
(7.2)
as it preserves the SO(2, 1) symmetry. We parameterize the Lobachevsky plane by complex

coordinates z and z , with 0 |z|2 < 1. The projection from AdS2 to the disc is given by
z=
x1 ix2
,
x0 + 1
z =
x1 + ix2
.
x0 + 1
(7.3)
The Poisson brackets (7.2) are projected to

{z, z } = i0 |z|2 ,
1
2
0 |z|2 = 1 |z|2 ,
(7.4)
2
with
the associated measure given by (2.1). The boundary |z| = 1 corresponds to x0 and
x12 + x22 going to infinity in the AdS2 space. From (7.4) it follows that the non-commutativity
will tend to zero as the boundary is approached in the non-commutative version of the theory.
This is fortunate because of the known difficulties associated with defining boundaries in noncommutative field theory [29,30]. We note that the Lobachevsky plane differs from the disc, and
hence the non-commutative version differs from the fuzzy disc [31], since the metric and {z, z }
are not constants in the interior.
256
The commutative Maxwell action (2.10) is

2
i
dz d z 1 |z|2 f 2 ,
Sf0 =
8
(7.5)
and the solutions of the corresponding free field equations are of the form
f=
2C0
,
(1 |z|2 )2
(7.6)
which diverges as the boundary |z|2 = 1 is approached. A possible gauge choice for the potential
one-form is
a = iC0
z dz z d z
.
1 |z|2
(7.7)
The flux going through a disc of radius r = |z| < 1 is r = 4C0 /(1
1
).
r2
7.2. Non-commutative case

For non-commutative AdS2 we replace the embedding coordinates xi , i = 0, 1, 2, by hermitean operators xi , i = 0, 1, 2, satisfying
x20 x21 x22 = 1,
(7.8)
with commutation relations

[x0 , x1 ] = ik x2 ,
[x2 , x0 ] = ik x1 ,
[x1 , x2 ] = ik x0 .
(7.9)
Thus the SO(2, 1) symmetry of the commutative theory is undeformed. Here, of course, there are
no finite-dimensional unitary representations of the algebra.
Next we write down an operator analogue of the projection (7.3) to the Lobachevsky plane.
Up to operator ordering ambiguities we have
z = (x1 ix2 )(x0 + 1)1 ,
z = (x0 + 1)1 (x1 + ix2 ),
(7.10)
which is non-singular if 1 is not in the spectrum of x0 . We can find the commutation relations
for z and z in a manner similar to what was done for the projective coordinates of the fuzzy
sphere [16]. For this define 1 = 12 (1 + x0 ), which from (7.9) commutes with |z|2 = zz . After
some algebra one can show that

k 2
2
z, z = k 1 |z|
(7.11)
1 + |z|
,
2
2
and from the constraint (7.8) one also has

z 2 z + 1 z z 1 2 1 1 1 = 0.
(7.12)
Using the commutation relations it is then possible to solve for as a function of |z|2 , and
hence write (z, z ) as a function of just |z|2 , and of course, the commutativity parameter k . We
will not write down the exact expression, but instead give the expansion up to first order in k .
Actually, we only need the zeroth order expression for (z, z ), its symbol being the classical
answer (7.4) times k , to write down the first order corrections to the real scalar field action (6.11)
257
and the conserved currents I and I in (6.14). Substituting (7.4) in the latter gives
I0 = ,
I = I0 + I1 + ,

k
0 + 1 |z|2 2 I
0 ,
I1 =
6|z|2 1 I0 1 |z|2 zI0 + z I
2
(7.13)
and I is the complex conjugate. It follows that the solutions to the commutative theory are not
preserved in the first order noncommutative theory. If 0 is a solution to the commutative theory,
0 = 0, then one can set
i.e.
= 0 + 1 ,
(7.14)
and, in principle, iteratively solve for the non-commutative corrections. Substituting into (6.13)
gives the following for the first order correction 1 :

k
1 2 2
2 2
3(z0 + z 0 ) + z 0 + z 0 .
1 =
(7.15)
2
2
To obtain the leading non-commutative corrections to gauge theory we need to compute
beyond zeroth order. This is since 1 appears in the lowest order corrections to the Maxwell
action (6.31). We first expand and substitute in (7.12) to get

2
k
2
2
= 1 |z|
(7.16)
1 + 1 3|z| + O k
.
4
Then from (7.11)
z, z

2
k
k 2
2 2
= 1 |z|
,
1 |z| + O k
2
2
(7.17)
whose symbol (z, z ) is

2
k
k
2
,
(z, z ) = (1 z z ) 1 z z + O k
2
2
(7.18)
where 2 = . If we apply the expansion of the star product, given in (6.4), up to order k this
leads to
2 k 2
3

k
1 |z|2
1 |z|2 + O k 3 ,
2
2
and so 1 in (6.5) is
(z, z ) =

3
1
1 |z|2 = 1 |z|2 .
2
This leads to
1 1
1
= ,
0
2
0 2
(7.19)
(7.20)
(7.21)
and as a result the commutative measure only gets corrected at first order by an overall constant
factor

2
i dz d z
k
d (z, z ) =
(7.22)
.
+
O
k
1
+
k (1 |z|2 )2
2
258
If one assumes a non-commutative analogue of the relationship (2.1) between the measure and
the metric, then the components of the metric tensor also scale by the factor 1 + k /2, up to
first order. From (7.21) it also follows that the solution (7.6) to the commutative gauge theory is
unchanged at this order; i.e. (7.6) satisfies the first order corrected field equation (6.32), although
the constant C0 is modified from its commutative value. In terms of this shifted constant, the flux
per unit area in (6.36) then goes to C0 [1 k (C0 + 12 )], and the action per unit area is shifted by
the same small factor.
8. Concluding remarks
In Section 6 we found that the flux per unit area of solutions to non-commutative Maxwell
theory gets shifted from the commutative answer. If the result can be applied to the example
of monopoles on the fuzzy sphere that would imply a shift in the magnetic monopole charge.
However, it is not straightforward to apply the techniques of Section 6 to the case of the fuzzy
sphere. Among the problems are: (a) As mentioned earlier, non-linear coherent states |z satisfying (6.1) are not readily available for fuzzy manifolds. Alternatives states have been found for
the fuzzy sphere, and they lead to a modified star product and integration measure [5,16]. (b) The
SeibergWitten map is problematic in the case of fuzzy gauge theories since mapping from
the non-commutative theory to the commutative one would mean connecting finite-dimensional
spaces with an infinite dimensional one [23]. (c) A constraint must be introduced to insure that
off block-diagonal matrix components of the potential remain non-dynamical when k is small.
This constraint was necessary in the full non-commutative theory in order to obtain solutions
with non-vanishing magnetic charge. We also found that the off block-diagonal matrix components should not vanish if we were to recover a finite value for the magnetic charge g0 in the
limit. If a proper treatment does reveal a shift in the magnetic charge due to non-commutative
corrections, one expects that there should also be a shift in the Dirac quantization condition. For
this it would be useful to analyze the quantum mechanics of a particle in a general monopole
background (5.25) (similar to what was carried out in [12]). This would require finding a gauge
invariant coupling to the non-commutative potentials. Constraints should then be found on the
charges which reduce to the Dirac quantization condition in the commutative limit.
In the approach taken here we took advantage of the Khler structure in two dimensions to
construct the fundamental commutator (4.1) from the determinant of the metric tensor on P0 .
As the commutative gauge theory action depends on the metric only through its determinant,
no other ingredient was needed to write down the non-commutative version of the theory. This
was not the case for the scalar field theory. There we chose a spacetime gauge to get rid of
additional degrees of freedom in the metric. More constraints will have to be imposed in order to
generalize this approach to other field theories. In particular, frame fields and spin connections
must be dealt with in the case of fermions, and also for gravity. Concerning the latter, the noncommutativity was taken to be non-dynamical, and in fact a constant, in most previous treatments
of non-commutative gravity [3235]. However, here since the non-commutativity is connected to
the metric tensor, a consistent approach would mean that (z, z ) is elevated to a dynamical field,
which should lead to a more realistic model for quantum gravity (at least in two dimensions).
Acknowledgement
The author is very grateful to A. Pinzul for discussions.
259
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]
[14]
[15]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
[23]
[24]
[25]
[26]
[27]
[28]
[29]
[30]
[31]
[32]
[33]
[34]
[35]
A.P. Balachandran, S. Kurkcuoglu, S. Vaidya, hep-th/0511114.

C.D. Fosco, G. Torroba, Phys. Rev. D 71 (2005) 065012.
V. Gayral, J.M. Gracia-Bondia, F. Ruiz Ruiz, Nucl. Phys. B 727 (2005) 513.
A.P. Balachandran, G. Marmo, B.S. Skagerstam, A. Stern, Lecture Notes in Physicss, vol. 188, 1983, p. 1;
A. Berard, H. Mohrbach, Phys. Rev. D 69 (2004) 127701;
P.A. Horvathy, L. Martina, P.C. Stichel, Phys. Lett. B 615 (2005) 87.
A. Pinzul, A. Stern, Nucl. Phys. B 718 (2005) 371.
W. Behr, A. Sykora, Nucl. Phys. B 698 (2004) 473.
H. Grosse, J. Madore, Phys. Lett. B 283 (1992) 218.
C. Klimcik, Commun. Math. Phys. 199 (1998) 257.
H. Grosse, P. Presnajder, Lett. Math. Phys. 46 (1998) 61.
P. Presnajder, J. Math. Phys. 41 (2000) 2789;
P. Presnajder, Mod. Phys. Lett. A 18 (2003) 2431.
S. Iso, Y. Kimura, K. Tanaka, K. Wakatsuki, Nucl. Phys. B 604 (2001) 121.
D. Karabali, V.P. Nair, A.P. Polychronakos, Nucl. Phys. B 627 (2002) 565.
A.P. Balachandran, G. Immirzi, Phys. Rev. D 68 (2003) 065023.
H. Steinacker, Nucl. Phys. B 679 (2004) 66.
S. Baez, A.P. Balachandran, B. Ydri, S. Vaidya, Commun. Math. Phys. 208 (2000) 787.
G. Alexanian, A. Pinzul, A. Stern, Nucl. Phys. B 600 (2001) 531.
M. Kontsevich, Lett. Math. Phys. 66 (2003) 157.
V.I. Manko, G. Marmo, E.C.G. Sudarshan, F. Zaccaria, Phys. Scr. 55 (1997) 528.
N. Seiberg, E. Witten, JHEP 9909 (1999) 032.
C. Jayewardena, Helv. Phys. Acta 61 (1988) 636.
B. Morariu, A.P. Polychronakos, Phys. Rev. D 72 (2005) 125002.
T. Asakawa, I. Kishimoto, JHEP 9911 (1999) 024;
B. Jurco, L. Moller, S. Schraml, P. Schupp, J. Wess, Eur. Phys. J. C 21 (2001) 383;
D. Brace, B.L. Cerchiai, A.F. Pasqua, U. Varadarajan, B. Zumino, JHEP 0106 (2001) 047;
D. Brace, B.L. Cerchiai, B. Zumino, Int. J. Mod. Phys. A 17 (S1) (2002) 205;
G. Barnich, F. Brandt, M. Grigoriev, JHEP 0208 (2002) 023.
J.M. Grimstrup, T. Jonsson, L. Thorlacius, JHEP 0312 (2003) 001;
J.M. Grimstrup, Acta Phys. Pol. B 34 (2003) 4855.
A. Pinzul, A. Stern, Int. J. Mod. Phys. A 20 (2005) 5871.
A. Pinzul, A. Stern, JHEP 0203 (2002) 039.
B. Jurco, P. Schupp, Eur. Phys. J. C 14 (2000) 367.
H. Fakhri, A. Imaanpur, JHEP 0303 (2003) 003.
J. Madore, H. Steinacker, J. Phys. A 33 (2000) 327.
A. Pinzul, A. Stern, JHEP 0111 (2001) 023.
A.P. Balachandran, K.S. Gupta, S. Kurkcuoglu, JHEP 0309 (2003) 007.
F. Lizzi, P. Vitale, A. Zampini, JHEP 0308 (2003) 057;
F. Lizzi, P. Vitale, A. Zampini, Mod. Phys. Lett. A 18 (2003) 2381;
F. Lizzi, P. Vitale, A. Zampini, JHEP 0509 (2005) 080.
P. Aschieri, C. Blohmann, M. Dimitrijevic, F. Meyer, P. Schupp, J. Wess, Class. Quantum Grav. 22 (2005) 3511;
P. Aschieri, M. Dimitrijevic, F. Meyer, J. Wess, hep-th/0510059.
M. Banados, O. Chandia, N.E. Grandi, F.A. Schaposnik, G.A. Silva, Phys. Rev. D 64 (2001) 084012.
S. Cacciatori, D. Klemm, L. Martucci, D. Zanon, Phys. Lett. B 536 (2002) 101.
A. Pinzul, A. Stern, Class. Quantum Grav. 23 (2006) 1009.
CUMULATIVE AUTHOR INDEX B741B745
Abe, H.
Agashe, K.
Ahl Laamara, R.
Altarelli, G.
Altarelli, G.
Antoniadis, I.
Apolloni, A.
Aprea, G.
Arai, M.
Arianos, S.
Arkani-Hamed, N.
B742 (2006) 187

B742 (2006) 59
B743 (2006) 333
B741 (2006) 215
B742 (2006) 1
B744 (2006) 156
B744 (2006) 340
B744 (2006) 277
B745 (2006) 208
B744 (2006) 340
B741 (2006) 108
DAdda, A.
de Forcrand, Ph.
Delgado, A.
Delgado, A.
Desrosiers, P.
Di Castro, C.
Donagi, R.
Donini, A.
Dorey, P.
Dreyer, O.
Drissi, L.B.
B744 (2006) 340

B742 (2006) 124
B741 (2006) 108
B744 (2006) 156
B743 (2006) 307
B744 (2006) 277
B745 (2006) 62
B743 (2006) 41
B744 (2006) 239
B744 (2006) 1
B743 (2006) 333
Baacke, J.
Ball, R.D.
Ball, R.D.
Balog, J.
Barrau, A.
Becker, K.
Bellucci, S.
Benakli, K.
Bergman, O.
Bernabu, J.
Bertolini, M.
Bi, X.-J.
Bill, M.
Bolognesi, S.
Bouchard, V.
Burrington, B.A.
B745 (2006) 142

B742 (2006) 1
B742 (2006) 158
B741 (2006) 390
B742 (2006) 253
B741 (2006) 162
B741 (2006) 297
B744 (2006) 156
B744 (2006) 136
B744 (2006) 180
B743 (2006) 1
B741 (2006) 83
B743 (2006) 1
B741 (2006) 1
B745 (2006) 62
B742 (2006) 230
Eberle, H.
Elias, V.
B741 (2006) 441

B743 (2006) 104
Faisst, M.
Fernndez-Martnez, E.
Feruglio, F.
Flohr, M.
Flohr, M.
Forrester, P.J.
Forte, S.
Forte, S.
Fr, P.
Fr, P.
Froggatt, C.
B742 (2006) 208

B743 (2006) 41
B741 (2006) 215
B741 (2006) 441
B743 (2006) 276
B743 (2006) 307
B742 (2006) 1
B742 (2006) 158
B741 (2006) 42
B742 (2006) 86
B743 (2006) 133
Caracciolo, S.
Casteill, P.-Y.
Chen, B.
Chetyrkin, K.G.
Chetyrkin, K.G.
Chishtie, F.A.
Colangelo, G.
Contino, R.
Cove, H.C.D.
Cvetic, M.
B741 (2006) 421

B741 (2006) 297
B741 (2006) 269
B742 (2006) 208
B744 (2006) 121
B743 (2006) 104
B744 (2006) 14
B742 (2006) 59
B741 (2006) 313
B745 (2006) 62
Gargiulo, F.
Gasser, J.
Giudice, G.F.
Gomshi Nobary, M.A.
Grain, J.
Grange, P.
Grilli, M.
Gudnason, S.B.
B741 (2006) 42
B745 (2006) 84
B741 (2006) 108
B741 (2006) 34
B742 (2006) 253
B741 (2006) 199
B744 (2006) 277
B741 (2006) 1
Haefeli, C.
Hamanaka, M.
B744 (2006) 14
B741 (2006) 368
0550-3213/2006 Published by Elsevier B.V.

doi:10.1016/S0550-3213(06)00387-7
Harmark, T.
He, Y.-L.
Higaki, T.
Hirano, S.
Horava, P.
Hoyos, C.
B742 (2006) 41
B741 (2006) 269
B742 (2006) 187
B744 (2006) 136
B745 (2006) 1
B744 (2006) 96
Ilderton, A.
Itou, E.
Ivanov, M.A.
B742 (2006) 176

B743 (2006) 74
B745 (2006) 84
Jacobsen, J.L.
Jacobsen, J.L.
Jaffe, R.L.
Jankiewicz, M.
B743 (2006) 153

B743 (2006) 207
B743 (2006) 249
B744 (2006) 380
Kagan, D.
Kajiyama, Y.
Kanno, H.
Keeler, C.A.
Kephart, T.W.
Kevlishvili, N.
Kishimoto, I.
Kobayashi, T.
Koike, Y.
Koike, Y.
Konishi, K.
Kormos, M.
Korthals Altes, C.P.
Krohn, M.
Kubo, J.
Khn, J.H.
Kuzenko, S.M.
B745 (2006) 109

B743 (2006) 74
B745 (2006) 165
B745 (2006) 1
B744 (2006) 380
B745 (2006) 142
B744 (2006) 221
B742 (2006) 187
B742 (2006) 312
B744 (2006) 59
B741 (2006) 180
B744 (2006) 358
B742 (2006) 124
B743 (2006) 276
B743 (2006) 74
B744 (2006) 121
B745 (2006) 176
Lerda, A.
Lishman, A.
Liu, J.T.
Livine, E.R.
Lorenzana, J.
L, H.
Lukyanov, S.L.
B743 (2006) 1
B744 (2006) 239
B742 (2006) 230
B741 (2006) 131
B744 (2006) 277
B741 (2006) 17
B744 (2006) 295
Mann, R.B.
Marino, E.C.
Markopoulou, F.
Marmorini, G.
Mavromatos, N.E.
McKeon, D.G.C.
Meloni, D.
Merlatti, P.
Minasian, R.
Mognetti, B.M.
Morales, J.F.
Moriyama, S.
B743 (2006) 104

B741 (2006) 404
B744 (2006) 1
B741 (2006) 180
B744 (2006) 180
B743 (2006) 104
B743 (2006) 41
B744 (2006) 207
B741 (2006) 199
B741 (2006) 421
B743 (2006) 1
B744 (2006) 221
Naculich, S.G.
Nagashima, J.
B742 (2006) 295

B742 (2006) 312
261
Nagashima, J.
Nevzorov, R.
Niedermayer, F.
Nielsen, H.B.
Nitta, M.
Nomura, Y.
Nunes, L.H.C.M.
B744 (2006) 59
B743 (2006) 133
B741 (2006) 390
B743 (2006) 133
B745 (2006) 208
B745 (2006) 29
B741 (2006) 404
Obers, N.A.
Oota, T.
B742 (2006) 41
B742 (2006) 275
Panotopoulos, G.
Papavassiliou, J.
Park, J.-H.
Pelissetto, A.
Peschanski, R.
Philipsen, O.
Poland, D.
Pope, C.N.
B745 (2006) 49
B744 (2006) 180
B745 (2006) 123
B741 (2006) 421
B744 (2006) 80
B742 (2006) 124
B745 (2006) 29
B741 (2006) 17
Quirs, M.
B744 (2006) 156
Randjbar-Daemi, S.
Reyes, S.A.
Richard, J.-F.
Rigolin, S.
Rim, C.
Rulik, K.
Russo, R.
B741 (2006) 236

B744 (2006) 330
B743 (2006) 153
B743 (2006) 41
B744 (2006) 239
B741 (2006) 42
B743 (2006) 1
Saidi, E.H.
Sainio, M.E.
Salas, J.
Saleur, H.
Salvio, A.
Scardicchio, A.
Schnitzer, H.J.
Sepahvand, R.
Shaposhnikov, M.
Shore, G.M.
Smolin, L.
Smolin, L.
Steele, T.G.
Stelle, K.S.
Stern, A.
Sturm, C.
Sturm, C.
Szabo, R.J.
B743 (2006) 333

B745 (2006) 84
B743 (2006) 153
B743 (2006) 207
B741 (2006) 236
B743 (2006) 249
B742 (2006) 295
B741 (2006) 34
B741 (2006) 236
B744 (2006) 34
B742 (2006) 142
B744 (2006) 1
B743 (2006) 104
B741 (2006) 17
B745 (2006) 236
B742 (2006) 208
B744 (2006) 121
B741 (2006) 313
Takcs, G.
Tateo, R.
Tentyukov, M.
Teraguchi, S.
Terno, D.R.
Trigiante, M.
Tseng, L.-S.
Tsvelik, A.M.
B741 (2006) 353

B744 (2006) 239
B742 (2006) 208
B744 (2006) 221
B741 (2006) 131
B741 (2006) 42
B741 (2006) 162
B744 (2006) 330
262
Tuckmantel, M.
Tweedie, B.
B744 (2006) 156

B745 (2006) 29
B744 (2006) 59
Yang, W.-L.
Yasui, Y.
Yokoi, N.
Young, C.A.S.
B744 (2006) 312

B742 (2006) 275
B741 (2006) 180
B745 (2006) 109
Vogelsang, W.
Wgner, F.
Waldron-Lauda, A.
Weisz, P.
B741 (2006) 353

B744 (2006) 180
B741 (2006) 390
Zamolodchikov, A.B.
Zhang, P.
Zhang, Y.-Z.
B744 (2006) 295

B741 (2006) 269
B744 (2006) 312

Nucl - Phys.B v.745 PDF

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Nucl - Phys.B v.745 PDF

Загружено:

Авторское право:

Доступные форматы

Nuclear Physics B 745 (2006) 128

Thermodynamics of noncritical M-theory

1. Introduction and summary

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

with periodic up to an arbitrary phase b,

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

which is cutoff-independent. We will return to the dependence of the thermodynamic quantities

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

This integral vanishes because ( ) is an even function of , implying that 1 is independent of

tanh( ) cosh2 (/2)

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

Similarly, the specific heat CV is the measure of the energy fluctuations,

tanh( ) cosh2 [( )/2]

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

asymptotic series4 in the powers of gs ,

and compared to the perturbative string-theory calculation.

This expansion exhibits several intriguing features [1]:

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

which in turn implies that

2.6. The low-temperature expansion

1  1 (1 212k )(1)k1 B2k 2k

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

1  1 (1 212k )(1)k1 B2k 2k

Taking a derivative of both sides, we obtain

d  (1)m  (2k 1)B2k 2m+1

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

is held fixed as R 0. This leads to a nontrivial high-temperature behavior,

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

This high-temperature scaling of as T 3 is a behavior characteristic of a relativistic massless

This can be asymptotically expanded in the powers of g M ,

As a result, the leading term in the asymptotic expansion of ()

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

3 . This is related to the coupling g 3 of the original vacuum by

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

sinh2 (/2) sinh[/(2R)]

Each contribution of fixed q in (2.38) is equivalent to 2R evaluated in type 0A theory with

sinh(/2) sinh2 [/(4R)]

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

with the Debye density of states D ( ) characteristic of a (2 + 1)-dimensional system. In a

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

3. Duality to the closed string theory of the topological A-model

This can be rewritten in terms of polylogarithms as

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

Here pA (tA ) is a nonuniversal second-order polynomial in tA , and CA is a nonuniversal constant.

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

The case of the thermal ensemble studied in Section 2 corresponds to b = .

Expanding this in the asymptotic series in R, we obtain

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

3.3. Solutions with the universal bottom of the Fermi sea

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

where k is the level of the ChernSimons theory.

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

i.e., as an S-duality transformation.

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

S.R. Das, Non-trivial 2d spacetimes from matrices, hep-th/0503002.

P. Horava, C.A. Keeler / Nuclear Physics B 745 (2006) 128

1 1 (1 212k )(1)k1 B2k 2k

1 1 (1 212k )(1)k1 B2k 2k

d (1)m (2k 1)B2k 2m+1

where mD,i ii D / 2M . We assume that Eq. (7) is satisfied for all i = 1, 2, 3.

while the generalized Friedmann equation becomes (V 2 )