UT-919

hep-th/0012210
December, 2000

Brane-Antibrane Action
arXiv:hep-th/0012210 v3 22 Jan 2001

from Boundary String Field Theory

Tadashi Takayanagi1 , Seiji Terashima2 and Tadaoki Uesugi3

Department of Physics, Faculty of Science

University of Tokyo
Tokyo 113-0033, Japan

Abstract
In this paper we give the boundary string field theory description of brane-
antibrane systems. From the world-sheet action of brane-antibrane systems we
obtain the tachyon potential and discuss the tachyon condensation exactly. We also
find the world-volume action including the gauge fields. Moreover we determine
RR-couplings exactly for non-BPS branes and brane-antibranes. These couplings
are written by superconnections and correspond to K 1 (M ) and K 0 (M ) for the non-
BPS branes and brane-antibranes, respectively. We also show that Myers terms
appear if we include the transverse scalars in the boundary sigma model action.

1
takayana@hep-th.phys.s.u-tokyo.ac.jp
2
seiji@hep-th.phys.s.u-tokyo.ac.jp
3
uesugi@hep-th.phys.s.u-tokyo.ac.jp
1 Introduction
In recent years there has been a lot of work on tachyon physics [1]. In string theory
tachyon fields naturally appear on the non-BPS branes [2, 3, 4, 5] and the brane-antibrane
systems [6, 7, 8]. These studies of the dynamical aspects of non-BPS systems are very
important to understand the vacuum structure of open string theory because we can
always see the process of the vacuum transition from an unstable one to a stable one
via the tachyon condensation. In the special case this process can be analyzed via the
marginal deformation of conformal field theory [3, 4, 9, 10, 11]. However in the general
situation the tachyon condensation is an off shell phenomenon. Thus we should use string
field theories.
Historically the most famous string field theory-Witten’s cubic string field theory-
has been mainly used to compute the tachyon potential of bosonic branes, non BPS
branes and brane-antibranes by the approximation which is called the level truncation
(for example see [12]). This approximation is needed because generally in the process of
the tachyon condensation many higher massive modes on a D-brane are excited, and we
can not consider the infinite numbers of modes at the same time.
However quite recently another string field theory has been applied to the tachyon
condensation. Some exact tachyon potentials and effective actions including both the
tachyon and gauge fields were calculated [13, 14, 15, 16, 17, 18]. This is called background
independent open string field theory (BIOSFT) or boundary string field theory (BSFT),
which was first formulated by Witten [19]. This string field theory is based on the world-
sheet sigma model action which is perturbed by the relevant operators on the boundary
of disk. The strategy for the exact analysis is the following. If we put the profile of the
tachyon field to the special form which makes the world-sheet sigma model action become
free, then the massive modes on branes are not excited due to the renormalization of the
world-sheet theory [13, 14]. Therefore the calculations including only the tachyon field
are exact.
The boundary string field theory was first formulated for bosonic open string. Its
superstring version is not known. However from the argument of the world-sheet super-
symmetry and the boundary entropy the authors of [15] conjectured that the string field
theory action is equal to the partition function. If this conjecture is correct, we can cal-
culate more easily the string field theory action than in the bosonic case. In [15] by using
this conjecture the effective action of non-BPS brane was calculated and this result was
equal to the proposed form in several papers [20, 21, 22, 23] if we assume that the tachyon
field is constant.
On the other hand only limited results for the effective action of the brane-antibrane
have been obtained. For the results from on-shell scattering amplitude, see [24]. However
the explicit form of the effective action is very important to know the dynamical aspects

1
of the tachyon condensation in the brane-antibrane systems. For example, if we want to
discuss the noncommutative tachyon on the brane-antibrane, then the detailed form of the
effective action is required [25]. These dynamical aspects of the brane-antibrane system
such as its effective action can only be obtained by the off-shell calculations. Especially
the boundary string field theory is suitable for investigating the general structure of the
effective action exactly.
Therefore one of the purpose of this paper is to study the effective action of the brane-
antibrane system in boundary string field theory. Indeed, we obtain exact results for
the tachyon condensation. Especially we show that by considering the special profile of
tachyons (kink or vortex) lower dimensional D-branes are produced and these tensions
are equal to the known values exactly. We also discuss the general non-abelian cases and
show that so called Atiyah-Bott-Shapiro (ABS) construction [26] naturally appears in the
boundary string field theory.
However if we include the gauge field, the tachyon on the brane-antibrane couples to
two kinds of the gauge fields in the bi-fundamental representation. Therefore there is no
choice of the profiles of the tachyon and the gauge field to make the world-sheet action
free. And it is difficult to obtain the exact effective action for the tachyons and gauge
fields. This is different from abelian non-BPS case [15] (in this case the tachyon does
not couple to the gauge field because the tachyon is in the U(1) adjoint representation).
However it is possible to calculate several lower terms in the α′ expansion. And we can
also discuss the general form of the effective action. Indeed this general form is consistent
with the argument on the noncommutative solitons [25].
Above arguments are limited to the effective action for NSNS sector. However the
boundary string field theory has the remarkable property that the on-shell RR closed
vertex can be inserted, while it is difficult for the cubic string field theory. Moreover in
the boundary string field theory we can formally incorporate the gauge fields at any order,
while in the cubic theory we can introduce the gauge fields only by the perturbation.
If one notices that the boundary interactions for the gauge field strengthes are similar
to the RR-couplings of BPS D-branes represented by the Chern character, one expects
that in the boundary string field theory RR-couplings of the non-BPS branes and brane-
antibranes is computable exactly. Indeed this expectation is true. Therefore in this paper
we give the most general coupling forms in the case of non-BPS branes and of brane-
antibrane systems. These are represented by the so called superconnection [27], which
was conjectured in [28, 29] in the case of brane-antibranes. We can also show that the RR
coupling of non-BPS branes has the structure of the superconnection. In mathematics it
is known that the charge which is represented by Chern character of a superconnection
is equivalent to K-theory charge. This means that the RR-coupling of brane-antibranes
and non-BPS branes corresponds to K 0 (M) and K 1 (M) respectively. Therefore this gives
another evidence of K-theory classification of D-brane charges [30, 28, 31].

2
 The plan of the paper is the following.
In section 2 we review the boundary string field theory for non-BPS branes and present
the world-sheet action for the brane-antibrane system. We justify this world-sheet action
by showing that with putting the tachyon field to zero the partition function becomes the
sum of two DBI actions, which is one for a brane and the other for an antibrane.
In section 3 we study exact tachyon condensations for special profiles of the tachyon
and show that the tensions of lower D-branes which are produced after the tachyon con-
densation are equal to known values in general situations and we relate these tachyon
profiles to Atiyah-Bott-Shapiro construction.
In section 4 we calculate RR couplings for brane-antibranes and non-BPS D-branes
by boundary string field theory. We also discuss that these forms are written by the
superconnections, and we relate these to K-theory groups. In the last subsections we
generalize these couplings to the couplings including noncommutative transverse scalars,
which is called Myers term [32].
In section 5 we calculate the effective action for NSNS sector. We show that the
form of the action is the sum of the DBI actions multiplied by the tachyon potential and
that this action is consistent with the argument of the noncommutative soliton. We also
calculate several lower terms in the α′ expansion.
In appendix we summarize the notations and spinor formulas mainly for the calculation
of RR-couplings.

2 World-Sheet Action of Brane-Antibrane System

Recently using background independent open string field theory several effective actions
has been calculated exactly in a certain sense [13, 14, 15]. This string field theory (from
now on we call this BSFT) was first considered by Witten in the bosonic open string field
theory [19]. In that paper the open string field action was defined by extending Batalin-
Vilkovisky (BV) formalism to that for open-string fields. The solution of this string field
master equation was given by [33]:

∂
S = (β i (λ) + 1)Z, (2.1)
∂λi
where S is the string field action, Z is the partition function, λi is one dimensional coupling
of sigma model (i.e. target space field) and β i (λ) is beta function of it.
This is for bosonic open string field theory. The BV-like formulation of background
independent superstring field theory has not been found until now. However the relation
between S and Z in (2.1) can be generalized to the supersymmetric version. Some years
ago Tseytlin et.al.[34, 35] calculated several partition functions including only the gauge
fields and they confirmed that partition functions were equal to the effective actions

3
constructed by calculating S-matrix perturbatively in supersymmetric case (not in bosonic
case). Moreover they conjectured that this partition function can be identified with off
shell string field action. In [15] they extended these interpretations to the full open string
field theory including tachyons.
Therefore in this paper we expect that the same relation holds not only for non-BPS
branes but also for brane-antibrane systems:

S = Z. (2.2)

Below we propose a brane-antibrane sigma model action and calculate the string field
action. This sigma model action is the extension of non-BPS brane’s one, thus before
giving this we first review non-BPS brane’s one[15, 36].
The partition function is defined by:
Z
Z = DXDψDη exp[−I(X, ψ, η)]. (2.3)

In this definition the action of the σ model I is:

I = I0 + IB , (2.4)
1 Z 2
I0 = d z[∂z X µ ∂z̄ Xµ + ψ µ ∂z̄ ψµ + ψ̃ µ ∂z ψ̃µ ], (2.5)
4π Σ
1
Z
IB = dτ dθ[−ΓDθ Γ + √ T (X)Γ − iDθ Xµ Aµ (X)]. (2.6)
∂Σ 2π
4
where the superspace representation in the boundary theory is defined by:



 Xµ = X µ + 2iθψ µ ,
Γ = η + θF, (2.7)

 D ∂ ∂
= ∂θ + θ ∂τ .

θ

If one writes IB in the component form and integrate out the auxiliary field F ,
then it becomes:
s
Z
1 2 µ
IB = dτ [ T (X)2 + η η̇ + i ψ η∂µ T − iẊ µ Aµ + 2iFµν ψ µ ψ ν ]. (2.8)
∂Σ 8π π
This is the world-sheet action for a non-BPS brane.
The superfield Γ corresponds to the internal degrees of the freedom of non-BPS branes
which is equal to 2×2 matrices 1, σ2 (Pauli matrix) [5, 1]. This Γ field description is first
given by Witten [28] and Harvey et.al. proposed that this action (2.8) describes non-BPS
branes in [37].
4
In this paper we set α′ to 2.

4
 The tachyon field T (X) is gauge-transformed in the U(1) adjoint representation (that
is equal to the gauge singlet), thus this action is gauge invariant without Γ being gauge
transformed. This makes T (X), Γ decoupled from the gauge field, which fact appears in
the (2.6). Therefore if one exponentializes the action I and performs the path-integration
in the approximation of neglecting the third term of (2.8), the partition function Z(=S)
becomes the simple structure which is the product of the DBI action and tachyon potential
exp(− 14 T 2 ) [20, 22, 23, 14, 15].
However if we consider non-abelian non-BPS D-branes, the tachyon field couples to
the gauge fields, its action is more complicated than U(1) case and the calculation of
effective action is difficult. Non-abelian action was proposed in [36]:
Z
¯ θ ∆ − ΓDθ Γ + ∆{−
¯ 1
IB = dτ dθ[−∆D √ ΓT (X) + iAµ (X)DXµ}∆], (2.9)
∂Σ 2π
where T(X) is N × N matrix in the case of N non-BPS D-branes, ∆ is complex fermionic
superfield which couples to the gauge field in the fundamental representation.
Next, we want to extend this action to the brane-antibrane system 5 , which contains
tachyons and is unstable. Before considering the world-sheet action, we should be re-
minded of the characteristic properties of this system (For a review, see [1]). First, a
brane-antibrane system has two kinds of vector multiplets (gauge fields and GSO even
fermions). One lives in D-brane, another in anti D-brane. Second, the tachyons and GSO
odd fermions come from the open strings between a D-brane and an anti D-brane. In
Type II theories open strings have the orientation, thus the D-brane and the anti D-brane
have two kinds of real tachyon fields and we can represent these by a complex tachyon
field which belongs to the bi-fundamental representation. Third, this system is essentially
non-abelian(this contains at least two branes in one pair system) and this contains the
Chan-Paton factors. Especially in the one pair case, the Chan-Paton factors are repre-
sented by 2 × 2 matrices (identity matrix and Pauli matrices (σ1 , σ2 , σ3 )). The identity
matrix represents the freedom of the total sum of gauge fields of the system. The matrix
σ3 represents the freedom of the relative difference of gauge fields and σ1 , σ2 the ones of
tachyons. These are the main properties of brane-antibrane systems.
It is famous that these are related to non-BPS D-branes by the “descent relation”
conjectured by Sen [1, 4, 38]. Therefore we expect that the world-sheet action is very
similar to the one of non-BPS branes. Naively the real tachyon field T (X) in (2.6) is
extended to the complex field T (X), T̄ (X) which is gauge-transformed as follows:

T (X) → eiλ1 (X) T (X)e−iλ2 (X) , (2.10)

where λ1 , λ2 are arbitrary functions of X. The sigma model action for the non-BPS D-
5
In this paper we mainly consider one pair brane-antibrane case. About the generic configuration of
brane-antibrane system (m D-brane + n anti D-brane) we comment at several points.

5
brane (2.6) respects gauge-symmetry6 and world-sheet supersymmetry, thus in the case of
the brane-antibrane it is natural to require these symmetries. From these considerations
we propose that the following action defines the D9−D9 action in BSFT (One for Dp−Dp
is simply obtained by T-duality).

I = I0 + IB , (2.11)
1 Z 2
I0 = d z[∂z X µ ∂z̄ Xµ + ψ µ ∂z̄ ψµ + ψ̃ µ ∂z ψ̃µ ], (2.12)
4π Σ
1
Z
IB = dτ dθ[−Γ̄(Dθ − iA(−) µ
µ (X)Dθ X )Γ + √ Γ̄T (X)
∂Σ 2π
1 i
+ √ T̄ (X)Γ − Dθ Xµ A(+) µ (X)]. (2.13)
2π 2

If we write IB in the component form and integrate out the auxiliary fields F, F̄ in Γ
and Γ̄:
Z
IB = (−)
dτ [η̄η̇ + 2iη̄ηψ µ ψ ν Fµν − iη̄η X˙ µ A(−)
µ
∂Σ
s s
2 µ 2 µ 1 i
−i η̄ψ Dµ T + i ψ ηDµ T + T̄ T − Ẋ µ A(+) µ ν (+)
µ + iψ ψ Fµν ], (2.14)
π π 2π 2
where we have employed the following definition:



 Aµ(±) = A(1)
µ ± Aµ ,
(2)


Dµ T = ∂µ T − iA(−)
µ T, (2.15)
 F (1),(2) = ∂ A(1),(2) − ∂ A(1),(2) .

µν µ ν ν µ

The field Γ is gauge-transformed in the bi-fundamental representation which is same

as (2.10). This fact forces the first term in (2.13) to be gauged. This is crucial difference
from the non-BPS brane case (2.6). This prescription is usual in Type I or Heterotic
non-linear sigma model action (for example see the section 12.3 in [39]) .
On the other hand, if we construct the most general one-dimensional renormalizable
action which is written by superfields Γ, Γ̄ and X, then this action coincides with (2.13)
up to the arbitrary real function g(X) in front of the first term of (2.13). This may be
confusing, because it looks as if the new real scalar field g(X) appeared on the brane-
antibrane. However, this g(X) can be eliminated by the redefinition of Γ, T (X) and
A±µ (X). Therefore this g(X) is a redundant two-dimensional sigma model coupling, and
even if we set g(X) to 1, the renormalizability is respected. At any rate eq.(2.13) is the
renormalizable action. We will use this fact in section 5.
6
Strictly speaking this symmetry is not two dimensional gauge symmetry but non-linear global sym-
metry. However it is famous in usual sigma model(Type I, Heterotic) that world-sheet global symmetry
corresponds to target space gauge symmetry

6
 From this action, we can calculate the effective action of the brane-antibrane system.
Especially we are interested in the form of NSNS effective action of the brane-antibrane
system, which corresponds to the Dirac-Born-Infeld (DBI) part of BPS D-brane action,
because the explicit form of this effective action is not known as much as non-BPS one (In
the non-BPS case the action is more familiar than the brane-antibrane. This is proposed
in several papers [20, 22, 23]).
Since above arguments of constructing brane-antibrane action are too heuristic, we
have to confirm that this action describes the brane-antibrane system correctly from
several point of view. In this paper, we confirm three nontrivial checks before calculating
the full NSNS action of the brane-antibrane as follows:

• With setting T (X) to 0 in (2.13) we reproduce the sum of the DBI action of two
kinds of gauge fields.

• By considering the tachyon condensation without the gauge fields by BSFT we check
the descent relation between the non-BPS D-brane and the brane-antibrane.

• We reproduce exactly the RR couplings of the brane-antibranes conjectured in [29]

which are represented by superconnection formula [27].

At first sight the first fact looks false because even if we set T (X) = 0, it is likely that
massive modes which fly between a D-brane and an antiD-brane modify the sum of DBI
actions. However this effect comes from open string one loop effect (cylinder amplitude)
and in the disk amplitude this effect does not cause any modification.
Now we check the first fact. The second and the third fact will be checked in the
section 3 and 4, respectively, and finally in the section 5 we calculate the NSNS action.
The path integral representation of the partition function is given by (2.3). First we split
Z into two dimensional part (internal of disk) and one dimensional part (boundary of the
disk):
Z Z
Z= DXDψ exp[−I0 (X, ψ)] DηD η̄ exp[−IB (X, ψ, η, η̄)]. (2.16)

In open string NS sector ψ(τ ) obeys the anti-periodic boundary condition so that ψ(τ )
has half-integer Fourier modes. Then η(τ ) and η̄(τ ) should also obey the anti-periodic
boundary condition in order for (2.14) to be locally well defined.
Here we integrate η(τ ) and η̄(τ ) first. The path integral of η(τ ) and η̄(τ ) is defined
on the circle, which corresponds to one loop partition function. Therefore transforming
this path integral to Hamiltonian formalism, η and η̄ are quantized and from (2.14) these
obey the usual canonical quantization condition:

{η, η̄} = 1. (2.17)

7
 By using canonical quantization method of η and η̄, the partition function Z becomes
as follows:
Z
Z = DXDψ exp[−I0 (X, ψ)]
s
[η̄, η] µ (−)
π [η̄, η] µ ν (−) 2 µ
Z 
× Tr P exp dτ i Ẋ Aµ (X) − 2i ψ ψ Fµν (X) + i η̄ψ Dµ T (X)
−π 2 2 π
s
2 µ 1 i

−i ψ ηDµ T (X) − T̄ (X)T (X) + Ẋ µ A(+) µ ν (+)
µ (X) − iψ ψ Fµν (X) ,(2.18)
π 2π 2
where “P” represents the path ordering and “Tr”(trace) implies that we should sum
expectation values in two state Hilbert space:
η| ↓i = 0 , η̄| ↓i = | ↑i,
η| ↑i = | ↓i , η̄| ↑i = 0. (2.19)
When we construct Hamiltonian from Lagrangian, we set the operator ordering by
antisymmetrization of η and η̄.
This is a consequence from the quantum mechanics, but at a first glance it is strange.
We said that classical fields η, η̄ obey anti-periodic boundary conditions and have half-
integer modes. Therefore they do not have zero-modes. However if we use Hamiltonian
formalism, τ dependence of η, η̄ drops out and it looks like that only zero modes remain.
This is confusing. Yet if we perform the path integral simply by the perturbation using
η, η̄ Green function with anti-periodic boundary condition:
1 1 X sin{r(τ − τ ′ )}
hη(τ )η̄(τ ′ )i = ǫ(τ − τ ′ ) = , (2.20)
2 π r∈Z+ 1 >0 r
2



 1 (τ > 0)
ǫ(τ ) ≡ 0 (τ = 0) , (2.21)

 −1 (τ < 0)


then we can check order by order that path integral representation (2.16) is equal to
(2.18). However note that this identity holds only for ǫ(0) = 0 regularization.
Here we can check that the operator commutation and anti-commutation relation of
η̄, η, [η̄, η] are same as that of Pauli matrix, σ+ , σ− , σ3 (where σ± ≡ 21 (σ1 ±iσ2 )). Therefore
we can replace η̄, η, [η̄, η] by σ+ , σ− , σ3 , respectively.
In that form Z becomes:
Z Z π
Z = DXDψ exp[−I0 (X, ψ)] × Tr P exp dτ M(τ ) , (2.22)
−π
where
 q 
1 2 µ
iẊ µ A(1) µ ν (1)
µ − 2iψ ψ Fµν − 2π
T T̄ i π
ψ Dµ T
M(τ ) =  q
2 µ 1
.
−i π
ψ Dµ T iẊ µ A(2)
µ −
(2)
2iψ µ ψ ν Fµν − 2π
T̄ T
(2.23)

8
 This is one expression of the brane-antibrane partition function. Furthermore this form
is able to be extended to the generic configuration of the brane-antibrane system (m D-
branes and n antiD-branes) because in that case we have only to replace A(1) (2)
µ (X), Aµ (X)
and T (X) in (2.23) with m×m, n×n and m×n matrices, respectively, while the expression
(2.22) has one fault that the gauge symmetry and the world-sheet supersymmetry cannot
be seen explicitly. We could not find (2.13) type action in the case of the non-abelian
brane-antibrane system.
Then we go back to the original question. That is “With setting T (X) to 0 in (2.13)
can we reproduce the sum of the DBI actions ?” If one sets T (X) to 0 in (2.23), the
off-diagonal part vanishes and the diagonal part remains. In this case Z becomes as
follows:
Z
Z = DXDψ exp[−I0 (X, ψ)]
" #
π iẊ µ A(1) µ ν (1)
µ − 2iψ ψ Fµν 0
Z
× Tr P exp dτ µ (2) (2)
−π 0 iẊ Aµ − 2iψ µ ψ ν Fµν
Z
= DXDψ exp[−I0 (X, ψ)]
2 Z π
dτ [iẊ µ A(k) µ ν (k)
X
× Tr P exp µ − 2iψ ψ Fµν ]. (2.24)
k=1 −π

This defines the sum of the NSNS action for A(1) (2)
µ (X), Aµ (X). In general form of
Aµ (X) this integration is only perturbatively possible, while in the approximation that
Fµν and the metric gµν are constant the integration is exact [40, 35, 34] and Z becomes
as follows:
Z q q 
10 (1) (2)
Z = T9 d x − det{gµν + 4π(Bµν + Fµν )} + − det{gµν + 4π(Bµν + Fµν )} ,(2.25)

where T9 is tension of a BPS D9-brane. Here we have replaced F (i) with F (i) + B using
(i) (i)
the Λ symmetry, Fµν → Fµν + (dΛ)µν and Bµν → Bµν − (dΛ)µν for later convenience.
This is the desired result (Dp − Dp case is obtained by T-duality).

3 Exact Tachyon Condensation on Brane-Antibrane

Systems
Here we investigate the tachyon condensation on brane-antibrane systems. It is obvi-
ous that the condensation of a constant tachyon field leads to the decay into the closed
string vacuum S = 0. The more interesting process is the condensation of topologically
nontrivial configurations of the tachyon field. According to the Sen’s conjecture [3, 28]
a codimension n configuration on Dp-branes generally produces D(p − n)-branes. As we

9
will see below, from the tractable free field calculations in BSFT we can describe such
a tachyon condensation exactly. We can perform explicit computations in parallel with
that for non-BPS D-branes discussed in [15]. A crucial difference is that we can allow
the vortex-type configurations since the tachyon field on a brane-antibrane system is a
complex scalar field.
Note also that even though our regularization throughout in this paper is based on
“ǫ -prescription” used in [34, 35, 36, 41], the result does not change if we use the point
splitting regularization as in [19, 15].
First let us consider the condensation of the vortex-type tachyon field on a single
Dp − Dp. From the viewpoint of the boundary conformal field theory (BCFT) one can
describe the condensation as a marginal deformation [10, 11] only. In BSFT defined by
the boundary action (2.14), we can also handle a relevant perturbation with two real
parameters ui (i = 1, 2) :
1
T (X) = (iu1 X1 + u2 X2 ). (3.1)
2
Here we have set the gauge fields to zero. One can show that any tachyon fields of the
form T (X) = a + pµ=0 bµ X µ , (a, bµ ∈ C) can be put into the form (3.1) by a Poincaré
P

transformation and the U(1) gauge transformation.

Note also that this perturbation can be treated as a free boundary interaction and
therefore the mixings with the other open string modes are avoided for all values of ui
(see also [14, 15]).
For non-zero ui this represents a vortex which is a codimension two configuration along
the x1 -x2 plane. If u1 = 0 or u2 = 0, then this corresponds to a kink configuration. This
fact is easy to see if one notes that the bosonic zero mode structure of BSFT on the
Dp − Dp behaves as
2 1 2 − 1 (u x )2
S ∼ e−|T (x1 ,x2 )| = e− 4 (u1 x1 ) 4 2 2
, (3.2)
which shows the tachyon condensation leads to the desired localized configuration for
large u1 and u2 .
In the presence of this boundary interaction the correlation functions in the NS-sector
are given by [15]
1 ′
GB (τ − τ ′ , yi , ǫ) =< X i (τ )X i (τ ′ ) >= 2 eim(τ −τ )−ǫ|m| ,
X
(3.3)
m∈Z |m| + yi
i X r ′
GF (τ − τ ′ , yi , ǫ) =< ψ i (τ )ψ i (τ ′ ) >= − eir(τ −τ )−ǫ|r| ,
2 r∈Z+ 1 |r| + yi
2

(i : no sum) (3.4)
where we have defined
yi = u2i . (3.5)

10
Note that in the above expression we have used “ǫ -regularization” discussed in [34, 35,
36, 41].
On the other hand the BSFT action S(= Z) including the boundary interaction can
be computed by differentiating with respect to y1 and y2 , respectively
∂ 1 2π 1
Z
log S = − dτ < X i (τ )X i (τ ) − 4ψ i (τ ) ψ i (τ ) >
∂yi 8π 0 ∂τ
1
= − [GB (0, yi , ǫ) − GB (0, 2yi, ǫ/2)]. (3.6)
2
The difference of the correlators GB (0, yi, ǫ) − GB (0, 2yi, ǫ/2) is given in the limit of
ǫ → 0 by
!
1 1 ǫ 2 1
e−ǫm − e− 2 m + ( − )
X
lim[GB (0, y, ǫ) − GB (0, 2yi, ǫ/2)] = 4
m≥1 m + y m + 2y y y
ǫ→0

X 1 ǫ 1 1 1 1 2 1
(e−ǫm − e− 2 m )] + 4
X X
= lim[4 ( − )−4 ( − )+( − )
m≥1 m m≥1 m + y m m≥1 m + 2y m y y
ǫ→0

d 2 d 1
= −4 log 2 − (4 log Γ(y) + + 4γ) + (2 log Γ(2y) + + 4γ), (3.7)
dy y dy y
where we have used the following formulae:
X 1 −ǫm
e = − log ǫ + O(ǫ), (3.8)
m≥1 m
d 1 X y
log Γ(y) = − + −γ (γ: Euler’s constant). (3.9)
dy y m≥1 m(m + y)
Then it is easy to integrate eq.(3.6) and we obtain S up to the overall normalization S0 :
S(y1 , y, 2) = S0 Z(y1 )Z(y2 ),
Z1 (y)2
Z(y) = 4y , (3.10)
Z1 (2y)
where Z1 is a function peculiar to BSFT [19],
√
Z1 (y) = yeγy Γ(y). (3.11)
The original Dp − Dp corresponds to ui = 0 and at this value the action is divergent since
the world-volume of the brane is non-compact :
2
S(Dp − Dp) → S0 √ (yi → 0). (3.12)
y1 y2
In the above computation the parameters yi play the role of cutoffs and are equivalent to
a regularization by compactification X i ∼ X i + Ri as
R1 R2 Z ∞ dx1 dx2 − 1 y1 (x1 )2 − 1 y2 (x2 )2 2
∼ e 4 4 =√ . (3.13)
2π 0 2π y1 y2

11
Let us now condense the tachyon field. Only when the tachyon is infinitely condensed
ui = ∞, the conformal invariance is restored, which implies that the equation of motion is
satisfied, because Z(y) is a monotonically decreasing function of y. Therefore there exist
three decay modes (u1 , u2 ) = (∞, 0), (0, ∞), or (∞, ∞). The first two cases represent
the kink configurations and we expect a non-BPS D(p − 1)-brane will be generated at
x1 = 0 or x2 = 0, respectively. This speculation is verified if one computes the tension
(for (u1 , u2) = (∞, 0)) and see that the correct value7 is reproduced as follows
TDp−D̄p S(0, 0) · (R1 )−1 1
= = , (3.14)
TD(p−1) S(∞, 0) 2π

where TDp−D̄p and TD(p−1) denotes the tension of a Dp−Dp and a non-BPS D(p−1)-brane,
respectively; we have also used the fact
√
Z(y) → 2π (y → ∞). (3.15)

More intuitive way to see the generation of a non-BPS D(p − 1)-brane is to discuss
the boundary interaction (2.14). Let us shift the original tachyon field by a real constant
T0 along x1 as follows:
1 i
T (X) = T0 + u1 X1 . (3.16)
2 2
Then the boundary interaction (2.14) after the condensation of the tachyon field (3.16)
becomes
Z
1 1 1
IB = dτ [η̄ η̇ + √ u1 ψ 1 (η − η̄) + T02 + u1 (X 1 )2 + · · ·], (3.17)
∂Σ 2π 8π 8π
where the new tachyon field T0 depends only on X a (a = 0, 2, ···, p). From this expression
it is easy to see that in the limit of u1 → ∞ we can set η = η̄ after we perform the path
integral of the fermion ψ 1 . Then the term in · · · which depends on A(−)
µν vanishes because
it is proportional to η̄η ∼ 0. On the other hand, the gauge field A(+) µ is not sensitive to
(+)
the tachyon condensation except that the element A1 is no longer a gauge field but a
transverse scalar field since the boundary condition along x1 becomes Dirichlet. Thus the
final boundary action after integrating out the fields X 1 , ψ 1 is identified with that of a
non-BPS D(p − 1)-brane (2.8).
Next we turn to the last case (u1 , u2 ) = (∞, ∞). This corresponds to the vortex-type
configuration and a BPS D(p − 2)-brane is expected to be generated at (x1 , x2 ) = (0, 0).
This fact is also checked by comparing the tension as follows:
TDp−D̄p S(0, 0) · (R1 R2 )−1 1
= = 2, (3.18)
TD(p−2) S(∞, ∞) 4π
7
Note that we set α√′ = 2 and that the tension of a non-BPS Dp-brane is larger than that of a BPS
D-brane by the factor 2.

12
matching with the known result. Also note that this configuration has no tachyonic modes
as desired. Indeed constant shifts of the original tachyon field (3.1) are equivalent to the
shift of the position where the D-brane is generated.
It is also interesting to consider multiple branes and antibranes. This can be rep-
resented by the Chan-Paton factors. Following [28] let us consider 2k−1 pairs of brane-
antibranes and condense the tachyon field
2k
uX
T (X) = i Γ µ Xµ , (3.19)
2 µ=1

where Γµ denote 2k−1 × 2k−1 Γ-matrices in 2k dimension and the extra factor i is due to
our convention of Γ-matrices.
These configurations carry K-theory charges known as Atiyah-Bott-Shapiro construc-
tion [26] and a BPS D(p − 2k)-brane is expected to be generated. This fact will be more
explicit by investigating the RR couplings in the next section. The verification of the
correct tension is the same as in the previous cases if one notes that the additional fac-
tor 2k−1 from the Chan-Paton factor should be included. Similarly one can also see the
condensation of the tachyon field

u 2k−1
Γ µ Xµ ,
X
T (X) = i (3.20)
2 µ=1

on 2k−1 pairs of brane-antibranes produces a non-BPS D(p − 2k + 1)-brane.

In this way we have obtained all decay modes which can be handled in BSFT by
free field calculations and these are all consistent with the BCFT results and K-theoretic
arguments. The incorporation of B-field (or equally F (+) ) can also be performed by free
field calculations and in the same way as in [42, 41, 43, 44, 45] we have only to replace
the parameters as follows
y
y ≡ diag(y1 , y2 ) → . (3.21)
1 + 2πB
q
This explains the extra factor det(1 + 2πα′B) of the Dp − Dp tension in the presence
of the B-field.

4 RR Couplings and Superconnection

In this section we compute the RR couplings on non-BPS Dp-branes and brane-antibrane
systems (Dp − Dp) in a flat space within the framework of BSFT. Since the backgrounds
of the closed string should be on-shell in BSFT, we can only consider RR-fields Cµ1 ···µp
which obey the equation of motion. This is sufficient to determine the RR couplings of
D-branes.

13
 Some of the RR couplings from the considerations of the descent relations [4, 1] and
from the calculations of on-shell scattering amplitudes were already obtained in the lit-
erature [46] for non-BPS D-branes and [29] for brane-antibrane systems. However our
off-shell calculations in BSFT give a more powerful and unified viewpoint as we will see.
For example our method determines all the unknown coefficients of the higher order terms
with respect to dT for non-BPS D-branes. Furthermore the resulting expressions in both
systems can be identified with an intriguing mathematical structure known as supercon-
nection [27]. This fact was already conjectured in [28, 29] for brane-antibrane systems.
Here we find the explicit proof of this in BSFT and we point out that this structure can
also be found in the RR couplings on non-BPS D-branes. These results give another
evidence of the K-theory classification of D-brane charges [30, 28, 31].
In the first two subsections we assume p = 9. In the last subsection we determine
the RR couplings for any p including the effects of non-abelian transverse scalars. As a
result we obtain the complete forms of Myers terms [32] for both non-BPS D-branes and
brane-antibrane systems.

4.1 RR Couplings on Non-BPS D-branes and Brane-Antibrane

System in BCFT
We regard the small shifts of the RR backgrounds as the perturbations. These shifts are
realized in BSFT as the insertions of the RR vertex operators in the disk Σ. The picture
[47] of the vertex operators8 should be (− 12 , − 23 ). This is because here we consider only
one insertion of them and because the total picture number on the disk should be −2.
We mainly follow the conventions in [49], where the scattering amplitudes of closed string
from D-branes were computed. The RR vertex operators are given by
1 3 1 3
V (− 2 ,− 2 ) = e− 2 φ− 2 φ̃ (P− Ĉ)AB SA S̃B , (4.1)
1 µ1 ···µp AB
Ĉ AB = (Γ ) Cµ1 ···µp , (4.2)
p!

where φ, φ̃ denote the bosonized superconformal ghost of left-moving and right-moving

sector, respectively; we define SA , S̃B as the spin-fields of world-sheet fermions and P−
denotes the projection of the chirality. For more details of the notation for spinors see
the appendix.
In the definition S = Z of BSFT on non-BPS D-branes and brane-antibrane systems
the ghost sectors are neglected in the same way as in the case of the BPS D-branes
discussed in [35, 34]. Therefore it is difficult to handle the ghost parts of the above RR
vertex operators explicitly. However it is natural to consider that the ghost parts and
the matter parts are decoupled and that the ghost parts give the trivial contribution in
8
For some subtleties, we recommend the readers to refer to [48].

14
the present calculations on the disk. Thus we can compute RR couplings taking only the
matter parts into consideration.
Next we discuss the supersymmetry in the boundary interactions. The supersymmetry
is completely preserved in the one dimensional boundary theory since all fermions at the
boundary of the disk obey periodic boundary conditions due to the cut generated by the
RR vertex. Therefore one can believe that the contributions from fermions and bosons
are canceled with each other for nonzero-modes and that the boundary theory becomes
topological in the sense of [50]. Note that in this paper we consider only D-branes in a
flat space and we have no corrections from world-sheet instantons.
First we turn to non-BPS D9-branes and determine the RR couplings up to the overall
normalization. To see the bose-fermi cancellation explicitly let us assume that the tachyon
field T (X) is a linear function as T (X) = T0 +uµ X µ and the field strength Fµν is constant.
Then the boundary interactions (2.8) are described as a free theory. Furthermore in the
R-sector the zero-modes and nonzero-modes are completely decoupled and it is easy to see
the bose-fermi cancellation for nonzero-modes because of the supersymmetry as follows
∂ 1 2π 1 µ
Z
2
log Z = − dτ < ξ µ (τ )ξ µ (τ ) − 4ψ µ (τ ) ψ (τ ) >= 0. (µ : no sum) (4.3)
∂(uµ ) 8π 0 ∂τ

Note that this property in the R-sector is in strikingly contrast with the results (3.6) in
the NS-sector.
Thus we have only to discuss the bosonic and fermionic zero-modes. The path integral
of the former is written as an integral over the world-volume coordinates x0 , · · ·, x9 . The
latters are divided into that of the world-sheet fermions ψ µ and of the boundary fermion
η. The integral of the zero modes of ψ µ in the action (2.8) can be replaced with the trace
over Γ-matrices in Hamiltonian formalism as follows:
1 1
ψ µ → √ i 2 Γµ , (4.4)
2
1
where the factor i 2 is due to the conformal map from the open string picture to the
closed string picture9 . Furthermore, we can compute the contribution from the boundary
fermion in Hamiltonian formalism and its quantization is given by η̂ 2 = 14 . Notice that
we should assume Γµ and η̂ do anti-commute because in eq.(4.4) we have not included a
cocycle factor. Then the result is given as follows including the RR vertex operator :
1 3√
Z
S = µ̃ Tr d10 x[ : exp[− T 2 − 2i 2 πΓµ η̂∂µ T + 2πFµν Γµ Γν ] : (P− Ĉ) η̂ ], (4.5)
4
where µ̃ represents the overall normalization and Tr denotes a trace with respect to
both the Γ-matrices and the boundary fermion η̂; the symbol : : means that Γ-matrices
9
See for example [51].

15
are antisymmetrized because any operators should be normal-ordered in Hamiltonian
formalisms. Note that an extra zero mode of η is inserted due to its periodic boundary
condition.
After we take the trace using the famous relation (A.11) between the Clifford algebra
(A.10) and the differential forms (A.9) and recover α′ = 2, we easily obtain the final
expression of the RR coupling on a non-BPS D9-brane10 as we will show in the appendix.
Its non-abelian generalization is also straightforward using the expression (2.9) and one
has only to add the trace of the Chan-Paton factor in front of the above expression. Thus
the result is given up to the overall factor µ′11 by:
s
1 πα′ 3
Z
S = µ′ Trσ [ C ∧ exp [− T 2 − i 2 DT σ1 + 2πα′ F ]σ1 ],
4 2
s
′
Z
1 2 πα′ 1
= iµ Tr C ∧ exp [− T − i 2 DT + 2πα′ F ] |odd , (4.6)
4 2
where the Pauli matrix σ1 is equivalent to the boundary fermion as σ1 ≃ 2η̂ and the
trace Trσ in the first equation also involves this freedom; the covariant derivative of the
Hermitian tachyon field on the non-BPS D-branes is denoted by DT = dT − i[A, T ].
Also note that in the second expression only the terms which include the odd powers of
DT should be remained because of the trace with respect to the boundary fermion and
therefore we have represented this prescription by |odd . From the above arguments, we
can see that the boundary fermion η plays a crucial role in the computations of the RR
couplings.
Next let us discuss the RR couplings of a D9 − D9 in BSFT. In this case decoupling
of the zero-modes and nonzero-modes should also occur. Even though it is not so easy to
give the explicit proof of the bose-fermi cancellation in this case, it is natural to assume
this cancellation. The path integral of the boundary fermions η, η̄ can be represented by
the RR-sector analog of the important formula (2.22):
Z Z π
DηD η̄ e−IB = Tr P (−1)F exp dτ M(τ ), (4.7)
−π
 q 
1 2 µ
iẊ µ A(1) µ ν (1)
µ − 2iψ ψ Fµν − 2π
T T̄ i π
ψ Dµ T
M(τ ) =  q
2 µ 1
,
−i π
ψ Dµ T iẊ µ A(2)
µ −
(2)
2iψ µ ψ ν Fµν − 2π
T̄ T

where the insertion of (−1)F (= [η̄, η]) is due to the periodic boundary condition of η, η̄
and can be replaced with the Pauli matrix σ3 . Since we have only to take the zero modes
into account, we can regard the path-ordered trace Tr P as the ordinary trace and thus
1 2
10
R
The leading term ∼ C ∧ e− 4 T dT was already pointed out in [15].
11 ′ − 32
√ action is real if only and only if µ is proportional to i . Later we will determine
Note that the above
′ − 32
this as µ = −i T9 / 2.

16
we obtain
3√
(1) µ ν
!
2πFµν Γ Γ − T T̄ 2(i) 2 πΓµ Dµ T
Z
−I0 −IB
DηD η̄DXDψ e = Str exp 3√
(2) µ ν
, (4.8)
−2(i) 2 πΓµ Dµ T 2πFµν Γ Γ − T̄ T

where a supertrace Str is defined to be a trace with the insertion of σ3 .

After we insert the RR vertex and again replace products of Γ-matrices with differential
forms, we obtain the following RR couplings on a D9 − D9 :
3√
2πα′ F (1) − T T̄ (i) 2 2πα′ DT
Z !
′′
S = µ Str C ∧ exp 3√ , (4.9)
−(i) 2 2πα′ DT 2πα′F (2) − T̄ T

where µ′′ is a real constant12 . This result coincides with the proposal in [29] as we will
see in the next subsection.
The non-abelian generalization is also straightforward if the above abelian supertrace
is replaced with the non-abelian one:
N
X M
X
Str diag(a1 , a2 , · · ·, aN , b1 , b2 , · · ·, bM ) = ai − bj , (4.10)
i=1 j=1

where we assume that there are N D9-branes and M antiD9-branes. This result coin-
cides with the proposal in [29] including the numerical factors as we will see in the next
subsection.
In this way we have derived the explicit RR couplings on a non-BPS D9-brane and
a D9 − D9 system in BSFT. The point is that one can read off the RR couplings if one
extracts the fermionic zero modes from the boundary action IB . This may be said as a
boundary topological model which can naturally lead to the notion of superconnection as
we will see in the next subsection. Note also that the above results can be applied for
general p-brane if the transverse scalars are set to zero.

4.2 Superconnection and K-theory Charge

Here we discuss the interpretation of the RR couplings on non-BPS D-branes and brane-
antibrane systems in BSFT as superconnections [27]. For brane-antibrane systems this
fact was first suggested in [28]. A definite relation between the RR couplings and the Chern
character of the superconnection was proposed in the paper [29]. Our calculations in the
previous subsection show that this interpretation indeed holds within the framework of
BSFT as we will see below. Moreover we argue that such an interpretation can be applied
to non-BPS D-branes and our previous calculations give an evidence for this.
3
12
Even though one may think the factor i 2 strange at first sight, it is an easy task to show that the
action is indeed real by expanding the exponential. Later we will determine this as µ′′ = T9

17
 Let us first review the definition and properties of superconnection following [27].
There are two kinds of superconnections: one is for even-cohomology and the other is for
odd-cohomology. In the K-theoretic language the former is related to K 0 (M) and the
latter to K 1 (M), where M is a manifold regarded as the D-brane world-volume. Both are
defined as follows13 :

Superconnection for K 0 (M)

In this case we consider the Z2 -graded vector bundle E = E (0) ⊕E (1) , which can be directly
applied to a brane-antibrane system if one identifies E (0) and E (1) as the vector bundle
on the branes and antibranes, respectively. Then the endomorphism of this superbundle
X ∈ End E has the following Z2 -grading:

0 if X : E (0) → E (0) or E (1) → E (1) ,

(
deg(X) = (4.11)
1 if X : E (0) → E (1) or E (1) → E (0) .

In addition, there is also a natural Z-grading p if one considers the algebra of the
differential forms Ω(M) = ⊕Ωp (M), where Ωp (M) denotes the algebra of p-forms on M.
The crucial observation is to mix these two gradings and to define the Z2 -grading for
α ∈ Ωp (M, End E) = Ωp (M) ⊗ Ω0 (M, End E) as follows:

α = ω ⊗ X ∈ Ωp (M) ⊗ Ω0 (M, End E), deg(α) ≡ p + deg(X), (4.12)

where Ω0 (M, End E) denotes the space of sections of End E. Then the superalgebra is
defined by the following rule:

(ω ⊗ X)(η ⊗ Y ) = (−1)deg(X)·deg(η) (ωη ⊗ XY ), (4.13)

and the supercommutator can be defined as:

[α, β] = αβ − (−1)deg(α)·deg(β) βα. (4.14)

An element of Ωp (M, End E) can be written as a 2 × 2 matrix, where the diagonal ele-
ments and off-diagonal elements have even and odd degree of Ω0 (M, End E), respectively.
We also define the supertrace as
!
α1 α2
α ∈ Ω(M, End E) = , Str(α) = Tr(α1 ) − Tr(α4 ) ∈ Ω(M), (4.15)
α3 α4

where Tr denotes the ordinary trace of vector bundles. Note that the supertrace vanishes
on supercommutators.
13 i
We include the explicit factor 2π in front of the field strength which was omitted for simplicity in
1
the original paper [27]. This is the reason why the factor i 2 does appear in the expressions below.

18
 Let us now define a superconnection on E to be an operator D = d + A on Ω(M, E)
of odd degree satisfying the derivation property:

D(ωφ) = (dω)φ + (−1)deg(ω) ω(Dφ), ω ∈ Ω(M), φ ∈ Ω(M, E). (4.16)

For local calculations familiar for physicists one can regard A as a degree odd element
of Ω(M, End E):
√ !
d + A(1) 2πiT
D =d+A= √ , (4.17)
2πiT̄ d + A(2)
√
where the factor 2πi has been included for later convenience. The diagonal parts
d + A(1) , d + A(2) denote the ordinary gauge connections of vector bundles E (1) , E (2) ,
respectively. T denotes a odd degree endomorphism of E. Notice that in this definition
the exterior derivative d does anti-commute with any odd element in End E.
Then the curvature F of a superconnection D is defined to be an even degree element
of Ω(M, End E):

F = D 2 = dA + A2
√ !
F (1) + 2πiT T̄ 2πiDT
= √ , (4.18)
2πiD T̄ F (2) + 2πiT̄ T

where we have defined DT = dT + A(1) T + T A(2)14 .

The “Chern character” of this superconnection is given by
i 2 i
Str exp( D ) = Str exp F . (4.19)
2π 2π
It is easy to see that this is closed because

d (StrD 2n ) = Str[D, D 2n ] = 0. (4.20)

Furthermore as shown in the main theorem in [27], its cohomology class does not depend
on the choice of T . In other words, this Chern character defines the same element of
K 0 (M) irrespective of T :
i 2
Str exp( D ) ≃ ch(E1 ) − ch(E2 ) ∈ H even (M, Q) ∼
= K 0 (M), (4.21)
2π
where ch(E) denotes the ordinary Chern character and we have applied the Chern iso-
morphism, which states that the even cohomology and the K-group K 0 (M) are equivalent
if Q is tensored.
14
Note that since T is an odd element, it anti-commutes with any one form. Therefore we can say that
T does couple to the relative gauge field A(1) − A(2) .

19
Superconnection for K 1 (M)

The first step to define the second superconnection is to regard a bundle E as a module
over the Clifford algebra C1 = C ⊕ Cσ1 . In other words, we define the endomorphism
of this superbundle as Endσ E = End E ⊗ C1 . Let us call all elements which include σ1
degree odd and the others degree even. In the physical context these correspond to the
fields on non-BPS D-branes which belong to GSO odd and even sectors, respectively. The
supertrace on Endσ E is defined as follows:

Trσ (X + Y σ1 ) = 2Tr(Y ), (4.22)

where X, Y ∈ Endσ E are degree even elements. Further we mix the degree of differential
forms in the same way as in the previous case eq.(4.12),(4.13) and (4.14).
A superconnection on E is defined locally to be an odd element as follows
s
iπ
D =d+A=d+A− T σ1 , (4.23)
2
where A is an ordinary connection and T is a self-adjoint endomorphism. The curvature
of this is also defined as

F = D 2 =sdA + A2
iπ iπ
= F− DT σ1 − T 2 , (4.24)
2 2
where we have defined DT = dT − i[A, T ]. Then the “odd Chern character” is given by
i 2 i
Trσ exp( D ) = Trσ exp F . (4.25)
2π 2π
The main theorem in [27] again tells us that this character is closed and its cohomology
class does not depend on the choice of T . Further we can regard this as an element of
K-theory group K 1 (M):
i 2
Trσ exp( D ) ∈ H odd (M, Q) ∼
= K 1 (M). (4.26)
2π
Physical interpretations

It is a well-known fact that the RR couplings on a BPS D9-brane are written by using
Chern characters [52, 53, 54]
Z
S = T9 Tr C ∧ exp 2πα′ F , (4.27)
M

20
where we have assumed that the world-volume (=spacetime) M is flat. As can be deduced
from this [30], the D-brane charges in type IIB were proposed to be regarded as an
element of K-theory group K 0 (M), which is equivalent to the Chern character up to
torsion via the Chern isomorphism. This proposal was strongly convinced in the study of
tachyon condensation on brane-antibrane systems [28]. The original definition of K 0 (M)
is given by considering the equivalence class of a pair of vector bundles (E1 , E2 ). This
definition can be naturally seen as a mathematical description of brane-antibrane systems.
Moreover it was pointed out that the other K-theory group K 1 (M) is related to the
tachyon condensation on non-BPS D9-branes [31]. This leads to the classification of the
D-brane charges in type IIA. At first sight, there are two different physical observations
about the generation of K-theory charges: the K-theory charges from RR couplings on a
BPS D-brane and those from the tachyon condensation. Then it is natural to ask if we
can directly fill this gap in string field theories. The answers to this question is yes in
BSFT and the key is superconnections as we see below.
The role of superconnections in the D-brane physics is explicit if one notes that the RR
couplings on D9-D9 systems and non-BPS D9-branes can be expressed as wedge products
of RR-fields and the Chern characters of superconnections:

i
Z
′′
S = µ Str C ∧ exp( F ) (for a D9-D9 ), (4.28)
M 2π
i
Z
S = µ′ Trσ C ∧ exp( F ) (for a non-BPS D9-brane), (4.29)
M 2π
where the curvature F in the first equation represents the superconnection for K 0 (M) and
in the second for K 1 (M). One of the expression (4.28) was already proposed in [29]. One
can indeed transform these mathematical expressions eq.(4.18),(4.24) into the physical
ones eq.(4.9),(4.6) by following the prescription15 :
√ √
D = d + A → 2π α′ D = 2π α′ (d − iA). (4.30)

Note that if one assumes the descent relation [4, 1], one can formally obtain the coupling
(4.29) from (4.28). Here the descent relation argues that one can reduce the degree of
freedom on a brane-antibrane to that on a non-BPS D-brane if one projects the Chan-
Paton factor Λ on a brane-antibrane by the following action:

(−1)FL : Λ → σ1 Λσ1 , (4.31)

15
Also note that for a brane-antibrane we need an additional minus sign in front of DT . This occurs
due to the following reason. The mathematical definition of the superconnection for K 0 (M ) assumes
that an odd form anti-commutes with an odd degree endomorphism as in eq.(4.13). On the other hand,
in the physical expression eq.(4.9) it does commute.

21
where FL denotes the spacetime fermion number in the left-moving sector. Thus we have
proved that the proposal in [29] is correct if we consider brane-antibranes in BSFT. Also
the second new expression (4.29) is interesting because this explicitly shows that we can
obtain the odd forms which correspond to K 1 (M) by including the Hermitian tachyon
field.
Then let us turn to the first question. In the expression (4.28) we can smoothly
connect the following two regions through the process of a tachyon condensation. Before
the condensation the RR charge in (4.28) comes only from the gauge field-strengthes. On
the other hand when the tachyon maximally condenses, the contribution from the tachyon
field dominates. For a trivial example, in [27] it was shown that if the tachyon field T
is invertible at some regions in M then the Chern character (4.19) does locally vanish
there. Physically this is natural since the condensation of a constant tachyon leads to
the decay into the vacuum and the lower dimensional charges are generated only at the
regions where T is not invertible. The same theorem also holds for the odd case (4.25).
A nontrivial example of the tachyon field which is not invertible is given by the Atiyah-
Bott-Shapiro construction (3.19). The important point is that RR charges or equally
K-theory charges do not change globally during the tachyon condensation as is shown in
eq.(4.21) and thus the charges are quantized in off-shell regions. To give a more concrete
picture, let us remind the calculation of the tachyon condensation in the previous section.
If one calculates the RR coupling (4.28) for the vortex-type tachyon configuration (3.1)
and integrates it over the world-volume, then it is easy to see that the RR charges are
independent of the parameters u1 , u2 except the “singular points” u1 = 0 or u2 = 0. This
3
analysis of the tachyon condensation determines the values of µ′ and µ′′ as µ′ = −i− 2 √T92
and µ′′ = T9 . Notice that the topology of the tachyon field T becomes trivial at the
points u1 = 0 or u2 = 0 and therefore one can not regard T as an element in the desirable
endomorphism. In this way we can relate D-brane charges in the RR-couplings to D-brane
charges due to tachyon condensations directly in BSFT.

4.3 RR Couplings of Transverse Scalars: Myers Terms

As we have seen above, the BSFT calculations determine the RR couplings on the various
systems of 9-branes exactly. If one wants to obtain the RR couplings for a single p(6= 9)-
brane, one has only to interpret the RR-fields as their pull-backs (see eq.(4.36)) to the
D-brane world-volume.
However their non-abelian generalizations (4.27),(4.28) and (4.29) are incomplete for
p(6= 9)-branes from the viewpoint of T-duality. As pointed out in [32] for BPS Dp-
branes, in order to recover the T-duality symmetry we must take the transverse scalars
Φi (i = p + 1, · · ·, 9) into account. For example if a pair of scalars is noncommutative
[Φi , Φj ] 6= 0, then D(p + 2)-brane charges emerge and therefore we should include this

22
effect. Those terms which represent such an effect are called Myers terms and their
structures were investigated in [32, 55, 56, 57, 58]. Here we argue that if one would like
to determine all of the Myers terms for any D-branes, then one has only to compute RR
couplings in BSFT including the transverse scalars. Here we set the value of B-field to
zero.
First let us determine the Myers terms for non-BPS D-branes. We use the non-abelian
boundary action (2.9) with an additional term due to the transverse scalars
Z
−i ¯ i (X)Dn Xi ∆,
dτ dθ ∆Φ (4.32)
∂Σ

where we have defined the T-dualized covariant derivative as Dn = ∂/∂θ + iθ∂/∂σ; σ

is the world-sheet coordinate transverse to the boundary ∂Σ. Note that if one wants to
discuss BPS D-branes, one has only to set T = 0, Γ = 0 in eq.(2.9). After we integrate
out the auxiliary fields and extract the zero modes as was done in (4.5), we obtain the
additional terms:
1√
: exp[· · · + 2i 2 π[Φi , T ]Γi η̂ − 2iπ[Φi , Φj ]Γi Γj − 4πDµ Φi Γµ Γi ] : , (4.33)

where · · · denotes the contribution from (4.5). When one estimates the trace of Γ-
matrices and is reminded of the calculations in the appendix, note that the matrices
Γi (i = p + 1, · · ·, 9) contract the indices of RR-fields Cµ1 ,···,µq in contrast with the
matrices Γµ (µ = 0, · · ·, p). Then we obtain the following additional RR couplings (Myers
terms) to eq.(4.6):
s
πα′ 1
Z
S = µ′ Tr [exp[ · · · + i 2 [iΦ , T ]σ1 − 2πα′ i iΦ iΦ − 2πα′ iDΦ ] σ1 ∧ C], (4.34)
2
where Tr denotes both the trace with respect to Chan-Paton factors and the trace defined
by eq.(4.22); iΦ and iDΦ denote the interior product by Φ and DΦ:
1
A= Aν1 ,ν2 ,··,νr dxν1 dxν2 · ·dxνr , (4.35)
r!
1 1
→ iΦ A = Φi Ai,ν2 ,···,νr dxν2 · ·dxνp , iDΦ A = Dµ Φi Ai,ν2 ,···,νr dxµ dxν2 · ·dxνr .
(r − 1)! (r − 1)!

In the above RR coupling the first term is peculiar to non-BPS D-branes. The second
corresponds to the generation of higher dimensional D-brane charges due to the noncom-
mutative transverse scalars. The last term changes the RR fields C into their covariantized
expression P̃ [C] of the pull-backs P [C] :

∂y ν1 ∂y νq
P [C]µ1 ,···,µq = Cν1 ,···,νp ( )···( ), (4.36)
∂xµ1 ∂xµq

23
where y µ = xµ (µ = 0, · · ·, p) denote the coordinates of the p-brane world-volume and
we also define y i = −2πα′ Φi (i = p + 1, · · ·, 9). In addition “covariantized” means
that all derivatives ∂µ Φi in the above definition (4.36) should be replaced with covariant
derivatives Dµ Φi . Then the total RR-couplings are given by
q q
πα′ πα′ i
− 14 T 2 −
Z
−2πα′ i iΦ iΦ + [iΦ ,T ] DT +2πα′ F
S = iµ′ Tr[ P̃ [e 2i
∧ C] ∧ e 2
] |odd , (4.37)

where the trace Tr is a symmetric trace with respect to [Φi , Φj ], [Φi , T ], T 2 , DT, F and
DΦi . For example the term proportional to [iΦ , T ] was already pointed out in [56]. On
the other hand if we set T to zero and neglect the restriction to odd forms, then one gets
the RR couplings for BPS Dp-branes, matching with the results in [32]. In this way we
have determined the complete form of Myers terms for non-BPS D-branes in BSFT and
these include new terms which are higher powers of [iΦ , T ]. Note also that our calculations
explicitly preserve the T-duality symmetry Aµ ↔ −Φi .
Then let us turn to the final task in this section: Myers terms in brane-antibrane
systems. In the same way as before we have only to add the extra terms which involve
transverse scalars. As a result the matrix in the exponential of eq.(4.9) includes Myers
terms as follows:
3√ !
2πα′ (F (1) − iiΦ(1) iΦ(1) − iDΦ(1) ) − T T̄ (i) 2 2πα′{DT + i(iΦ(1) T − iT Φ(2) )}
3√ ,(4.38)
−(i) 2 2πα′ {DT − i(iT̄ Φ(1) − iΦ(2) T̄ )} 2πα′ (F (2) − iiΦ(2) iΦ(2) − iDΦ(2) ) − T̄ T

where we interpreted Str in eq.(4.9) as both the symmetric trace with respect to Chan-
Paton factors and the original supertrace. Note that if one requires that the branes and
the antibranes always have the common world-volume, then we get Φ(1) = Φ(2) (= Φ). In
this case we can find intriguing terms in the RR couplings:
Z
(p+3)
S∼ Ci,j,··· Tr[Φi , T ][Φj , T̄ ]e−T T̄ . (4.39)
p-brane

5 The World-Volume Action for Brane-Antibrane Sys-

tem
In this section we compute the BSFT action for tachyons and gauge fields in the brane-
antibrane system perturbatively.
First, we will show below that the action computed from (2.18) can be written as
Z q q
(1) (2)
Z = T9 d10 xe−T T̄ (x) − det{gµν + 4πFµν (x)} + − det{gµν + 4πFµν (x)}
∞
#
(i)
X
+ G2n (Fµν , T, T̄ , Dρ ) , (5.1)
n=1

24
 (i)
where G2n is a 2n-derivative term constructed from Fµν , T, T̄ and 2n covariant derivatives
(−)
Dµ . Note that, for example, Fµν T is regarded as a 2-derivative term and should be
(−)
included in G2 because of the identity [Dµ , Dν ]T = −iFµν T . These ambiguities are
reminiscent of the case of the non-Abelian Born-Infeld action [59] in which [Fµν , Fγσ ] =
i[Dµ , Dν ]Fγσ was regarded as a derivative term.
What does this action represent? We first answer the question. As shown in [40], this
action (5.1) is regarded as an on-shell effective action for T and Aµ , in which the massive
modes are integrated out, or an off-shell BSFT action with other modes than T and Aµ
setting to zero. However, if the one-point function of the massive fields vanishes in all the
off-shell region, then we can regard this action as the off-shell BSFT action, in which the
massive fields are integrated out.
The explanation is as follows. First we expand the full string field action by the power
series of the massive fields (λi ):
(1) (2)
S[T, Aµ , λi] = S (0) [T, Aµ ] + λi Si [T, Aµ ] + λi λj Sij [T, Aµ ] + · · · · · · . (5.2)

If we want to obtain the effective string field action including only the tachyon and the
gauge fields, then we integrate out the massive fields. Note that “integrate out” means
that we only insert the solution of the equation of motion for λi because we want to obtain
the tree level effective string field action. The general solution of the equation of motion
(1)
for λi is very complicated. However, if Si [T, Aµ ] (the one-point function of λi ) vanishes,
then we can easily find one solution of the equation of motion, that is λi = 0. Therefore
in this case we can regard the S (0) [T, Aµ ] as the effective action, in which the massive
fields are integrated out. S (0) [T, Aµ ] is just the action we are able to calculate from the
renormalizable sigma model action where the massive fields are set to zero. Therefore
(1)
the crucial point to obtain the effective action is whether Si [T, Aµ ] does vanish or not.
(1)
From the argument of [14, 15], Si [T, Aµ ] vanishes at least at the conformal fixed point
of the renormalization group flow (on shell point), however in the off shell region it is
non-trivial. In [14] they say that it is correct for the free sigma model action because
from the relation of BSFT:
∂S
= β j Gij (λ), (5.3)
∂λi
∂S
(where Gij (λ) is some positive definite metric) ∂λ i |λ=0 vanishes. However, in this ar-
gument, the only ambiguity is whether the non-diagonal elements of Gij (λ), where j
corresponds to T, Aµ and i to the massive fields, does vanish or not (note that β i |λ=0 = 0
(i : the massive fields) and β T,A |λ=0 6= 0.). However, from the success and correctness of
the tachyon condensation in [14, 15] we can consider that above facts hold in the case of
the free sigma model action. Therefore if the above facts hold not only for the free sigma
model but also for our renormalizable one (eq.(2.13)), S (1) [T, Aµ ] vanishes in off-shell re-

25
gions and we can regard eq.(5.1) as the effective action, in which the massive modes are
integrated out. However, we can not assert that it is true.
Now we prove eq.(5.1). We use the matrix form of the world-sheet action
" #
π M11 M12
Z Z
Z= DXDψexp[−I0 (X, ψ)]Tr P exp dτ , (5.4)
−π M21 M22

where
 q 
1 2 µ
iẊ µ A(1) µ ν (1)
" #
M11 M12 µ − 2iψ ψ Fµν − 2π
T T̄ i π
ψ Dµ T
= q
2 µ 1
.
M21 M22 −i π
ψ Dµ T iẊ µ A(2)
µ −
(2)
2iψ µ ψ ν Fµν − 2π
T̄ T
(5.5)

Now we expand X µ from zero mode xµ as X µ = xµ + ξ µ and path integrate ξ µ and ψ µ

(ψ does not include the zero mode in this case.). Then from the expansion T T̄ (X) =
T T̄ (x) + ξ µ (∂µ (T T̄ ))(x) + · · · = T T̄ (x) + ξ µ (Dµ (T T̄ ))(x) + · · ·, we can replace T T̄ (X) in
(i)
Z by T T̄ (x) if we remove the derivative terms which can be included in G2n . The Fµν (X)
(i)
terms can also be replaced by Fµν (x). Furthermore we see that the contributions from
the off diagonal part of the matrix Mab have the form of G2n since they can be expanded
as Dµ T (X) = Dµ T (x) + ξ ν (∂ν (Dµ T ))(x) + · · ·. The terms iẊ µ A(i) µ (X) in the diagonal
part are combined with the other non-gauge covariant terms to give gauge covariant
ones. Therefore we conclude that the action for the brane-antibrane system becomes
(5.1). Similarly, we can show that the action for a system of n branes and m anti-branes
becomes
Z   q q 
10 −T T̄ (1) −T̄ T (2)
Z = T9 d x SymTr e − det{gµν + 4πFµν } +e − det{gµν + 4πFµν }
∞
!
(i)
X
+ G2n (F , T, T̄ , Dµ ) , (5.6)
n=1

(i)
where SymTr denotes the symmetrized trace for T T̄ , T̄ T and Fµν [59].
Now we will use this action in order to investigate the non-commutative soliton in the
(1)
brane-antibrane system [60, 61, 25]. Here we simply assume that Si [T, Aµ ] is zero or
derivative terms, then the action (5.6) is exact even for the off shell fields. According to
the argument in [62] if we include the background constant B-field, then the propagator
is modified and the action can be written as the non-commutative field theory when we
use the point splitting regularization for the world-sheet theory. This will be true at least
for the on-shell fields. We assume here that this is also true for the off-shell action since
the evidences for this have been obtained [44, 42, 41]. Then the non-commutative action
for the brane-antibrane system with background constant B-field becomes the same form
as (5.6) where the product is ∗-product and closed string metric gµν and coupling gs are
replaced by open string metric Gµν and Gs , respectively. We should also replace the

26
 (i) (i)
field strength Fµν by F̂µν + Φµν , where Φµν represents a freedom to relate closed string
quantities to open string quantities. Below we take Φµν = −Bµν for simplicity [62, 25].
In [25] the exact non-commutative solitons for the string field theories were obtained
using the technique called solution generating technique, which is also useful for the BPS
case [63] [64]. For the brane-antibrane system, they assumed the form of action which does
not vanishes at the closed string vacuum T = T0 . Our action (5.6), however, vanishes
at the minimum T = T0 = ∞ and does not have the form assumed in [25]. Thus we
should confirm whether their construction of the exact soliton works for a noncommutative
version of our brane-antibrane action (5.1) or not. In order to use the solution generating
technique, we regard the fields on the non-commutative field theory as operators on Fock
space. In [25] an almost gauge transformation was defined as

Dµ(i) → S (i) Dµ(i) S (i)† ,

T → S (1) T S (2)† , (5.7)

where D (i) is a covariant derivative operator D (i) = d − iA(i) and i = 1, 2. Here S (i) is an
almost unitary operator which satisfies S (i)† S (i) = 1 and S (i) S (i)† = 1 − P (i) where P (i) is
a projection operator. First, we start from the trivial vacuum A(i) µ = 0 and T = T0 (= ∞),
which is a solution of the equations of motion for the noncommutative version of the
action (5.1). Then the configuration constructed by the above transformation becomes a
nontrivial exact solution of the equations of motion from the argument in [25].
We can see that the tension of this soliton is correct value. The process of its calculation
is almost same as in [25]. The only difference from [25] is the form of the action, especially
the explicit form of terms of field strengths without covariant derivatives. These terms in
our action are the sum of Born-Infeld actions multiplied by the function of the tachyon,
(i)
exp(−T̄ T ), or exp(−T T̄ ). We note that the tachyon T̄ T, T T̄ and the field strength F̂µν +
(i) (i) 2 (2) 2 (1) (i) (i) (i) (i)†
Φµν ∼ [Dµ , Dν ] are transformed to |T0 | (1−P ), |T0 | (1−P ) and S [Dµ , Dν ]S ∼
Φµν (1 − P (i) ) respectively. Here we can obtain

V (T̄ T )[Dµ(2) , Dν(2) ] = 0, (5.8)

for the soliton configuration from the equation V (T̄ T ) = V (T̄ T ) − V (0) + V (0) =
(V (T¯0 T0 ) −V (0))(1 −P (2)) + V (0) = V (0)P (2) where V (T̄ T ) ∼ exp(−T̄ T ) and V (T̄ T ) = 0
at T = T0 = ∞. Therefore for this soliton configuration constructed from the vacuum
where the action vanishes, the sum of Born-Infeld actions remains to vanishes except the
gauge fields independent term, i.e. the tachyon potential. Hence the soliton constructed
in [25] is an exact solution of the brane-antibrane action which represents N1 Dp-brane
and N2 anti Dp-brane, where Ni = dim(Ker(1 − P(i) )).
Note that the action evaluated at this soliton configuration has a non-zero value, while
the action evaluated at the closed string vacuum T = ∞ vanishes. In [25] they argued

27
that the action can not vanish even at the closed string vacuum from an observation that
the BPS brane after the tachyon condensation has non-zero mass. In fact, our action has
nonzero value for the soliton solution and is consistent.
The properties of vanishing kinetic terms may be required from the observation that
at the closed string vacuum in order to solve the U(1) problem the strong coupling effects
should be important and the vanishing kinetic terms almost mean the strong coupling
physics [65, 20, 66, 67].
Next we compute the brane-antibrane effective action as a sigma model partition
function perturbatively up to α′2 . Hereafter we will restore the dimension-full parameter
α′ by including a factor α′ /2. Here we use the regularized correlation function
1 im(τ −τ ′ )−ǫ|m| X 1
hξ µ (τ )ξ µ (τ ′ )i = α′ = 2α′ cos(m(τ − τ ′ ))e−ǫ|m| ,(5.9)
X
e
m∈Z6=0 |m| m∈Z>0 m
i X r ir(τ −τ ′ )−ǫ|r|
hψ µ (τ )ψ µ (τ ′ )i = − sin(r(τ − τ ′ ))e−ǫ|r|.
X
e = (5.10)
2 r∈Z+ 1 |r| m∈Z+ 1 >0
2 2

This regularization keeps world-sheet supersymmetry and the spacetime gauge invariance
which corresponds to world-sheet global symmetry [35, 34]. First we expand fields in
(2.18) as
1 σ ρ ν µ
T T̄ (X) = T T̄ (x) + ξ µ (Dµ (T T̄ ))(x) + · · · + ξ ξ ξ ξ (Dσ Dρ Dν Dµ (T T̄ ))(x) · · · ,
4!
Aµ(i) (X) = Aµ(i) (x) + ξ ν ∂ν A(i)
µ (x) + · · · ,
(i)
Fµν (i)
(X) = Fµν (i)
(x) + ξ ρ(Dρ Fµν )(x) + · · · ,
1
Dµ T (X) = Dµ T (x) + ξ ν (∂ν (Dµ T ))(x) + ξ ρ ξ ν (∂ρ ∂ν (Dµ T ))(x) · · · . (5.11)
2
Then we can compute the partition function usually by the perturbation in α′ . Since
the actual computations are somewhat complicate, we will only show the outline of the
computation and the result below.
The gauge invariance of the effective action can be checked by replacing the ∂µ by
Dµ and picking the terms which depend on A(i) µ . For example, the coefficient of the term
(−) µ
Aρ (D T )(∂ρ Dµ T ) is proportional to

1 1 1 X 1 1 1 X 2
   
e−(r+m)ǫ + e−(r+m)ǫ − e−(r+m)ǫ ,
X
+ +
r,m>0 m r + m r−m r,m>0 r r+m m−r r,m>0 rm
(5.12)
1
which is indeed zero, where r ∈ Z + 2 > 0 and m ∈ Z > 0. The other terms can be
calculated explicitly by the formulae in [34] except a finite constant
 
1 1 1
 
e−(r+m)ǫ − (log ǫ)2  .
X
γ0 = lim  + (5.13)
ǫ→0
r,m>0 m r + m r − m

28
The result up to α′2 can be rearranged in a rather simple form:
  
−T¯R TR (1) 2 (2) 2
  
′ ′2 2
S = Z = T9 e 2 + 8α log 2Dµ TR Dµ TR + α π FµνR + FµνR
(−)
+4α′2 γ0 Dν Dµ TR Dν Dµ TR + 32α′2 i(log 2)2 FµνR Dµ TR Dν TR
1 2 2
 
′2 2
+2α 8(log 2) − π (Dµ TR Dµ TR )2 − α′2 π 2 (Dµ TR )2 (Dν TR )2
3 3
2
π  
+ α′2 (Dµ Dν TR )TR + TR (Dµ Dν TR )
6 i
× Dµ Dν (TR TR ) + Dν TR Dµ TR + Dµ TR Dν TR , (5.14)

where we renormalized the tachyon field only as

1
T = TR + α′ log ǫDµ Dµ TR + α′2 (log ǫ)2 Dν Dν Dµ Dµ TR
2
(−)
+ iα (log ǫ)2 Dν FνµR Dµ TR ,
′2

(−) (−)
Aµ(−) = AµR + α′ log ǫDν FνµR (5.15)

Note that we obtain the form (5.14) by a field redefinition which corresponds to a renor-
malization (5.15) from the two dimensional point of view.

6 Conclusions and Future Directions

In this paper we described the tachyon dynamics on the brane-antibrane by the method
of the boundary string field theory. We constructed the world-sheet boundary action
of the brane-antibrane system by using the boundary fermions. The remarkable point
of this formalism is that for the special profile of the tachyon field the calculation of
the tachyon potential and the tensions of the lower branes which are produced after the
tachyon condensation can be performed exactly. From these calculations we confirmed
the ‘descent relation‘ of Non-BPS systems.
On the other hand in the case of including the all gauge fields it was difficult to calculate
the exact effective potential except RR couplings. We found that the explicit form of the
RR coupling in boundary string field theory can be represented by the superconnection.
Furthermore we also took the transverse scalars into account and showed that Myers
terms appear naturally.
We have also calculated several lower terms of the action of BSFT in the α′ expansion.
Further we discussed the general structure of the action and the result was consistent with
the arguments on noncommutative solitons.
Our work raises some interesting questions and we hope to return to these in future
work.

29
 • As we have seen, the incorporation of one of the gauge field strength F (+) on a
brane-antibrane system can be treated as a free theory and leads to the familiar
noncommutative theory if the point-splitting regularization is employed. On the
other hand, the exact treatment of the other field strength F (−) is found to be
difficult. Then for brane-antibrane systems it seem to be essential to ask whether
we can express this effect as a sort of a noncommutative theory.

• Our calculation here is performed assuming that the target space is flat. Then it is
natural to ask what will happen to the tachyon physics if one considers a non-BPS
D-brane system wrapping on a more complicated manifold such as a Calabi-Yau
manifold. In such a case one should take world-sheet instantons into account. For
example the discussion on their RR couplings will be modified and couplings may be
expressed by some “stringy” Chern characters. A related question is how much the
world-sheet supersymmetry has effects on the dynamics of tachyon condensation.

• One more interesting question is the physical meaning of the Myers terms which we
have found for brane-antibrane systems and non-BPS D-branes in the framework of
BSFT.

Acknowledgements
We would like to thank Y. Matsuo and K. Ohmori for useful discussions. The works of
S.T. and T.T. were supported in part by JSPS Research Fellowships for Young Scientists.

Note added:
While preparing this paper for publication, we received the preprint [69] in which the
world-sheet action for brane-antibrane system was given and the preprint [70] in which
the BSFT for brane-antibrane system was discussed.

A Notation for Spinors and Some Formulas

Here we summarize our notation for spinors following [49] and after that we show some
calculations which is needed for the derivation of the RR coupling in section 4.
The ten dimensional Γ-matrices ΓA B , (A, B = 1, 2, ···, 32) are defined by the following
Clifford algebra:
{Γµ , Γν } = −2η µν , η µν = diag(−1, 1, · · ·, 1). (A.1)
Note that we distinguish spinor and adjoint spinor indices as subscripts and superscripts,
−1
respectively. The charge conjugation matrix C AB , CAB , which obey the relations:
CΓµ C −1 = −(Γµ )T , (A.2)

30
can raise or lower these spinor indices. Therefore we can omit the matrix C as
−1
(Γµ )AB = CBC (Γµ )A C , (Γµ )AB = C AC (Γµ )C B
. (A.3)

Note also from the above equations it is easy to see

CAB = −CBA , (Γµ )AB = (Γµ )BA . (A.4)

B
We also define (Γ11 )A as

Γ11 = Γ0 Γ1 · · · Γ9 , (A.5)

and the chirality projection matrix P± is defined by

1
P± = (1 ± Γ11 ). (A.6)
2
The matrix Γ11 satisfies the following identities

(Γ11 )AB = (Γ11 )BA , (Γ11 )2 = 1, {Γ11 , Γµ } = 0. (A.7)

Before we will discuss the calculation of the RR couplings, let us now show some useful
formulae. The first one is about the trace of Γ-matrices:
p(p−1)
Tr[Γµ0 µ1 ···µp Γ01···p ] = 32 (−1) 2 ǫµ0 µ1 ···µp (0 ≤ µi ≤ p), (A.8)

where Γµ0 µ1 ···µp = 1/p! (Γµ0 Γµ1 · ·Γµp − Γµ1 Γµ0 · ·Γµp + · · ·) denotes the antisymmetrized
Γ-matrices. The second one is the famous relation between the Γ-matrices and the differ-
ential forms. More explicitly, a r-form in ten dimension:
1
C= Cµ µ ···µ dxµ1 dxµ2 · · · dxµr , (A.9)
r! 1 2 r
corresponds to the following 32 × 32 matrix:
1
Ĉ = Cµ µ ···µ Γµ1 µ2 ···µr . (A.10)
r! 1 2 r
This correspondence preserves the multiplication as
1
: Ĉ1 Ĉ2 := (C1 ∧ C2 )µ1 µ2 ···µ(r1 +r2 ) Γµ1 µ2 ···µ(r1 +r2 ) , (A.11)
(r1 + r2 )!
where : : denotes the antisymmetrization.
Let us now turn to the derivation of the RR couplings. It involves the computations
of the correlation functions on a disk whose boundary is on a Dp-brane. We assume
its world-volume extends in the direction x0 , x1 , · · ·, xp . Then it is easier to calculate

31
the correlation functions by performing T-duality transformation. This transformation is
given with respect to the spin operators by

SA → SA , S̃A → MA B S̃B ,
±iΓ0 Γ1 · · · Γp (p = even)
(
B
MA = , (A.12)
±Γ0 Γ1 · · · Γp Γ11 (p = odd)

where SA and S̃B denote the left-moving and right-moving spin operators, respectively;
the sign ambiguity ± depends on the conventions and we choose the the plus sign. The
above rule can be derived by requiring that the OPE’s of the left-moving fermions ψ µ and
spin operators SA have the same structure as those of right-moving ones ψ̃ µ , S̃B after
one performs the T-duality transformation. Using these facts, the RR couplings on both
a non-BPS D-brane and a brane-antibrane system are summarized as the following form
up to the overall normalization (see eq.(4.5) and (4.7)):
p+1
1
Kµ1 µ2 ···µr Tr[ P− ĈMΓµ1 µ2 ···µr ],
X
S= (A.13)
r=0 r!

where Kµ1 µ2 ···µr (0 ≤ µi ≤ p) is a r-form which depends on the field-strength and the
tachyon field; Ĉ denotes the RR-sector vertex as defined in eq.(4.2). If one takes the
transverse scalars into account, one should also discuss Kµ1 µ2 ···µr for µi ≥ p. However
such a case can also be treated similarly and we omit this. Then we can write down
the RR couplings explicitly up to the overall normalization which is independent of r as
follows:
p+1
1
δp+1,q+r ǫµ1 ···µq ν1 ···νr Cµ1 ···µq Kν1 ν2 ···νr ,
X
S= (A.14)
q,r=0 q!r!

where we have used the formula (A.8). Finally these couplings are written in the language
of the differential forms as
Z
S= C ∧ K, (A.15)
Σ(p+1)

where Σ(p+1) denotes the world-volume of the p-brane.

32
References
[1] For a review, see A. Sen, “Non-BPS States and Branes in String Theory,” hep-
th/9904207.

[2] O. Bergman and M.R. Gaberdiel, “Stable non-BPS D-particles,” Phys. Lett. B441
(1998) 133, hep-th/9806155.

[3] A. Sen, “SO(32) Spinors of Type I and Other Solitons on Brane-Antibrane Pair,”
JHEP 9809 (1998) 023, hep-th/9808141.

[4] A. Sen, “BPS D-branes on Non-supersymmetric Cycles,” JHEP 9812 (1998) 021,
hep-th/9812031.

[5] A. Sen, “Type I D-particle and its interactions,” JHEP 9810 (1998) 021, hep-
th/9809111.

[6] M. B. Green, “Pointlike states for type 2b superstrings,” Phys. Lett. B329 (1994)
435, hep-th/9403040.

[7] T. Banks and L. Susskind, “Brane-Antibrane Forces,” hep-th/9511194.

[8] A. Sen, “Tachyon Condensation on the Brane Antibrane System,” JHEP 9808 (1998)
012, hep-th/9805170.

[9] M. Frau, L. Gallot, A. Lerda and P. Strigazzi, “Stable non-BPS D-branes in type I
string theory,” Nucl. Phys. B564 (2000) 60, hep-th/9903123.

[10] J. Majumder and A. Sen, “Vortex Pair Creation on Brane-Antibrane Pair via
Marginal Deformation,” JHEP 0006 (2000) 010, hep-th/0003124.

[11] M. Naka, T. Takayanagi and T. Uesugi, “Boundary State Description of Tachyon

Codensation,” JHEP 0006 (2000) 007, hep-th/0005114.

[12] V.A. Kostelecky and S. Samuel, “The Static Tachyon Potential In The Open Bosonic
String Theory,” Phys. Lett. B207 (1988) 169; “On A Nonperturbative Vacuum For
The Open Bosonic String,” Nucl. Phys. B336 (1990) 263.
A. Sen and B. Zwiebach, “Tachyon condensation in string field theory,” JHEP 0003
(2000) 002, hep-th/9912249.
N. Berkovits, A. Sen and B. Zwiebach, “Tachyon condensation in superstring field
theory,” Nucl. Phys. B587 (2000) 147, hep-th/0002211.

[13] A.A. Gerasimov and S.L. Shatashvili, “On exact tachyon potential in open string
field theory,” JHEP 0010 (2000) 034, hep-th/0009103.

33
[14] D. Kutasov, M. Marino and G. Moore, “Some exact results on tachyon condensation
in string field theory,” JHEP 0010 (2000) 045, hep-th/0009148.

[15] D. Kutasov, M. Marino and G. Moore, “Remarks on tachyon condensation in super-

string field theory,” hep-th/0010108.

[16] S. Moriyama and S. Nakamura, “Descent relation of tachyon condensation from

boundary string field theory,” hep-th/0011002.

[17] J.R. David, “Tachyon condensation using the disc partition function,” hep-
th/0012089.

[18] A. Fujii and H. Itoyama, “Behavior of Boundary String Field Theory Associated with
Integrable Massless Flow,” hep-th/0012150.

[19] E. Witten, “On background independent open string field theory,” Phys. Rev. D46
(1992) 5467, hep-th/9208027; “Some computations in background independent off-
shell string theory,” Phys. Rev. D47 (1993) 3405, hep-th/9210065.

[20] A. Sen, “Supersymmetric World-volume Action for Non-BPS D-branes,” JHEP 9910
(1999) 008, hep-th/9909062.

[21] J.A. Minahan and B. Zwiebach, “Effective tachyon dynamics in superstring theory,”
hep-th/0009246.

[22] M.R. Garousi, “Tachyon couplings on non-BPS D-branes and Dirac-Born-Infeld ac-
tion,” Nucl. Phys. B584 (2000) 284, hep-th/0003122.

[23] E.A. Bergshoeff, M.de Roo, T.C. de Wit, E. Eyras and S. Panda, “T-duality and
Actions for Non-BPS D-branes,” JHEP 0005 (2000) 009, hep-th/0003221.

[24] I. Pesando, “On the effective potential of the Dp Dp-bar system in type II theories,”
Mod. Phys. Lett. A14 (1999) 1545 , hep-th/9902181.

[25] J.A. Harvey, P.Kraus and F. Larsen, “Exact Noncomuutative Solitons,” hep-
th/0010060.

[26] M.F. Atiyah, R. Bott and A. Shapiro, “Clifford Modules”, Topology 3 (1964) 3.

[27] D. Quillen, “Superconnection and the Chern character,” Topology 24 (1985) 89.

[28] E. Witten, “D-branes and K-theory,” JHEP 9812 (1998) 019, hep-th/9810188.

[29] C. Kennedy and A. Wilkins, “Ramond-Ramond Couplings on Brane-Antibrane Sys-

tems,” Phys. Lett. B464 (1999) 206, hep-th/9905195.

34
[30] R. Minasian and G.Moore, “K-theory and Ramond-Ramond charge,” JHEP 9711
(1997) 002, hep-th/9710230.

[31] P. Horava, “Type IIA D-branes, K-theory, and Matrix Theory,” Adv. Theor. Math.
Phys. 2 (1999) 1373, hep-th/9812135.

[32] R.C. Myers, “Dielectric-branes,” JHEP 9912 (1999) 022, hep-th/9910053.

[33] S. L. Shatashvili, “Comment on the background independent open string theory,”

Phys. Lett. B311 (1993) 83, hep-th/9303143; “On the problems with background
independence in string theory,” hep-th/9311177.

[34] O.D. Andreev and A.A. Tseytlin, “Partition Function Representation For The Open
Superstring Effective Action: Cancellation Of Mobius Infinities And Derivative Cor-
rections To Born-Infeld Lagrangian,” Nucl. Phys. B311 (1988) 205.

[35] A.A. Tseytlin, ”Renormalization Of Mobius Infinities And Partition Function Rep-
resentation For String Theory Effective Action,” Phys. Lett. B202 (1988) 81.

[36] A.A. Tseytlin, “Sigma model approach to string theory effective actions with
tachyons,” hep-th/0011033.

[37] J.A. Harvey, D. Kutasov and E. Martinec, “On the relevance of tachyons,” hep-
th/0003101.

[38] A. Sen, “Descent relations among bosonic D-branes,” Int. J. Mod. Phys. A14 (1999)
4061, hep-th/9902105

[39] J. Polchinski, “String theory I, II,” Cambridge university press.

[40] E.S. Fradkin and A.A. Tseytlin, “Non-Linear Electrodynamics From Quantized
Strings,” Phys. Lett B163 (1985) 123.

[41] O. Andreev, “Some computations of partition functions and tachyon potentials in

background independent off-shell string theory,” hep-th/0010218.

[42] K. Okuyama, “Noncommutative tachyon from background independent open string

field theory,” hep-th/0010028.

[43] D. Nemeschansky and V. Yasnov, “Background independent open string field theory
and constant B-field,” hep-th/0011108.

[44] L. Cornalba, “Tachyon condensation in large magnetic fields with background inde-
pendent string field theory,” hep-th/0010021.

35
[45] K. Li and E. Witten, “Role of short distance behavior in off-shell open string field
theory,” Phys. Rev. D 48 (1993) 853, hep-th/9303067.

[46] M. Billo’, B. Craps and F. Roose, “Ramond-Ramond couplings of non-BPS D-

branes,” JHEP 9906 (1999) 033, hep-th/9905157.

[47] D. Friedan, E. Martinec and S. Shenker, “Conformal Invariance, Supersymmetry And

String Theory,” Nucl. Phys. B271 (1986) 93.

[48] M. Billo, P. Di Vecchia, M. Frau, A. Lerda, I. Pesando, R. Russo and S. Sciuto,

“Microscopic string analysis of the D0-D8 brane system and dual R-R states,” Nucl.
Phys. B526 (1998) 199, hep-th/9802088.

[49] M.R. Garousi and R.C. Myers, “Superstring Scattering from D-Branes, ” Nucl.Phys.
B475 (1996)193, hep-th/9603194.

[50] H. Ooguri, Y. Oz and Z. Yin, “D-branes on Calabi-Yau spaces and their mirrors,”
Nucl. Phys. B477 (1996) 407, hep-th/9606112.

[51] C.G. Callan, C. Lovelace, C.R. Nappi and S.A. Yost, “Loop Corrections To Conformal
Invariance For Type 1 Superstrings,” Phys. Lett. B206 (1988) 41.

[52] M. Li, “Boundary States of D-Branes and Dy-Strings,” Nucl.Phys. B460 (1996) 351,
hep-th/9510161.

[53] M.R. Douglas, “Branes within branes,” hep-th/9512077.

[54] M.B. Green, J.A. Harvey and G. Moore, “I-brane inflow and anomalous couplings
on D-branes,” Class. Quant. Grav. 14 (1997) 47, hep-th/9605033.

[55] S.F. Hassan and R. Minasian, “D-brane couplings, RR fields and Clifford multipli-
cation,” hep-th/0008149.

[56] B. Janssen and P. Meessen, “A non-Abelian Chern-Simons term for non-BPS D-

branes,” hep-th/0009025.

[57] S. Mukhi and N.V. Suryanarayana, “Chern-Simons terms on noncommutative

branes,” JHEP 0011 (2000) 006, hep-th/0009101.

[58] M.R. Garousi and R.C. Myers, “World-volume potentials on D-branes,” JHEP0011,
(2000) 032, hep-th/0010122.

[59] A. A. Tseytlin, “On non-abelian generalisation of the Born-Infeld action in string

theory,” Nucl. Phys. B501 (1997) 41, hep-th/9701125.

36
[60] J.A. Harvey, P. Kraus, F. Larsen and E.J. Martinec, “ D-branes and Strings as
Non-commutative Solitons, ” JHEP 0007 (2000) 042, hep-th/0005031.

[61] E. Witten, “Noncommutative Tachyons and String Field Theory,” hep-th/0006071.

[62] N. Seiberg and E. Witten, “String Theory and Noncommutative Geometry,”

JHEP9909 (1999) 032, hep-th/9908142.

[63] M. Hamanaka and S. Terashima, “On exact noncommutative BPS solitons,” hep-
th/0010221.

[64] K. Hashimoto, “Fluxons and exact BPS solitons in non-commutative gauge theory,”
hep-th/0010251.

[65] P. Yi, “Membranes from five-branes and fundamental strings from Dp branes,” Nucl.
Phys. B550 (1999) 214, hep-th/9901159.

[66] O. Bergman, K. Hori and P. Yi, “Confinement on the brane,” Nucl. Phys. B580
(2000) 289, hep-th/0002223.

[67] M. Kleban, A. Lawrence and S. Shenker, “Closed strings from nothing,” hep-
th/0012081.

[68] For a review, see M.R. Gaberdiel, “Lectures on Non-BPS Dirichlet Branes,” Class.
Quant. Grav. 17 (2000) 3483, hep-th/0005029.

[69] K. Hori,“Linear Models of Supersymmetric D-Branes, ” hep-th/0012179.

[70] P. Kraus and F. Larsen, “Boundary String Field Theory of the DDbar System, ”
hep-th/0012198.

37