Nuclear Physics B 585 (2000) 3–27

www.elsevier.nl/locate/npe

Optimal-observable analysis of the angular and
energy distributions for top-quark decay products
at polarized linear colliders
Bohdan Grzadkowski a,∗ , Zenr¯o Hioki b
a Institute of Theoretical Physics, Warsaw University, Ho˙za 69, PL-00-681 Warsaw, Poland
b Institute of Theoretical Physics, University of Tokushima, Tokushima 770-8502, Japan

Received 27 April 2000; revised 6 June 2000; accepted 16 June 2000

Abstract
An optimal-observable analysis of the angular and energy distributions of the leptons and bottom
(−)

quarks in the process e+ e− → t t¯ → l ± / b · · · has been performed in order to measure the most
general top-quark couplings to gauge bosons at polarized linear colliders. The optimal beam
polarization for determination of each coupling has been found. A very sensitive test of CP violation
in t t¯ production and decay has been proposed.  2000 Elsevier Science B.V. All rights reserved.
PACS: 13.65.+i
Keywords: Top quark; CP violation; Anomalous top-quark interactions

1. Introduction
In spite of the fact that the top quark has been discovered already several years ago [1,2]
its interactions are still very weakly constrained. It remains an open question if top-quark
couplings obey the Standard Model (SM) scheme of the electroweak forces or there exists
a contribution from physics beyond the SM. We could interpret the great success of the 1loop precision tests of the SM as a strong indication that the third generation also obeys the
SM scheme. However, an independent and direct measurement of the top-quark couplings
is definitely necessary before drawing any definite conclusion concerning non-standard
physics.
∗ Corresponding author.

E-mail addresses: bohdan.grzadkowski@fuw.edu.pl (B. Grzadkowski), hioki@ias.tokushima-u.ac.jp
(Z. Hioki).
0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 3 8 5 - 0

4

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

Over the past several years there was a substantial effort devoted to a possibility of
determining top-quark couplings through measurements performed at the open top region 1
of future e+ e− linear colliders [4–29]. The existing studies focused mainly on tests of CP
violation in top-quark interactions. In this article we will construct some new tools which
could help to measure both CP violating and CP conserving top-quark couplings at linear
colliders and therefore reveal the structure of fundamental interactions beyond the SM.
The top quark decays immediately after being produced and its huge mass mt ' 174
GeV leads to a decay width Γt much larger than ΛQCD . Therefore the decay process is not
influenced by any fragmentation effects [30–32] and decay products will provide useful
information on top-quark properties. Here we will consider distributions of either l ± in
(−)

the inclusive process e+ e− → t t¯ → l ± · · · or bottom quarks from e+ e− → t t¯ → b · · · . It
turns out that the analysis of the leptonic and b-quark final states is similar and could be
presented simultaneously. Although t t¯ are also produced via W W fusion [33], we do not
consider here this mechanism since σ (e+ e− → t t¯ν ν¯ ) is expected to be much smaller than

σ (e+ e− → t t¯) for the energy of our interest ( s . 2 TeV) [34].
This paper is organized as follows. First in Section 2 we describe the basic framework of
our analysis, and then show the angular and energy distributions of the lepton and b-quark
in Section 3. In Section 4, after briefly reviewing the optimal-observable procedure [35]
(see also [15,36–38]), we estimate to what precision all the non-standard parameters can
be measured or constrained adjusting the initial beam polarizations. Finally, we summarize
our results in Section 5. In the appendix we collect several functions used in the main text
for completeness, though some of them could also be found in our previous papers [5,7].

2. Framework and formalism
We parameterize t t¯ couplings to the photon and the Z boson in the following way  

g
µ
¯ t ) γ µ Av + δAv − (Bv + δBv )γ5
Γvt t¯ = u(p
2 

(pt − pt¯)µ
(δCv − δDv γ5 ) v(pt¯),
(2.1)
+
2mt
where g denotes the SU (2) gauge coupling constant, v = γ , Z, and
4
sin θW ,
Bγ = 0,
3  

1
8
1 − sin2 θW ,
AZ =
2 cos θW
3

Aγ =

BZ =

1
2 cos θW

denote the SM contributions to the vertices. Among the above non-SM form factors, δAv ,
δBv , δCv describe CP -conserving while δDv parameterizes CP -violating interactions.
1 Recently an interesting and complementary analysis by Jezabek, Nagano and Sumino has been published [3]
where the authors discussed possibility of determining CP -violating production form factors at the t t¯ threshold
region.

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

5

Similarly, we adopt the following parameterization of the W tb vertex suitable for the t
and t¯ decays:   

iσ µν kν L 

g
µ
¯ b ) γ µ f1L PL + f1R PR −
f2 PL + f2R PR u(pt ),
ΓW t b = − √ Vt b u(p
MW
2    

g
iσ µν kν ¯L
µ
¯ t¯) γ µ f¯1L PL + f¯1R PR −
f2 PL + f¯2R PR v(pb¯ ),
Γ¯W t b = − √ Vt∗b v(p
MW
2
(2.2)
where PL/R = (1 ∓ γ5 )/2, Vt b is the (tb) element of the Kobayashi–Maskawa matrix and
k is the momentum of W . In the SM f1L = f¯1L = 1 and all the other form factors vanish.
On the other hand, it is assumed here that interactions of leptons with gauge bosons are
properly described by the SM. Throughout the calculations all fermions except the top
quark are considered as massless. We also neglect terms quadratic in the non-standard
form factors.
Using the technique developed by Kawasaki, Shirafuji and Tsai [39,40] one can derive
the following formula for the inclusive distributions of the top-quark decay product f in
the process e+ e− → t t¯ → f + · · · [5,6]:
Z
d 3 Γf
1

d3 σ
+ −
(e
(t → f + · · ·),
e

f
+
·
·
·)
=
4
dΩ
(n,
0)
t
dΩt
Γt dpf /(2pf0 )
dpf /(2pf0 )
(2.3)
where Γt is the total top-quark decay width and d3 Γf is the differential decay rate for the
process considered. dσ (n, 0)/dΩt is obtained from the angular distribution of t t¯ with spins
s+ and s− in e+ e− → t t¯, d σ (s+ , s− )/d Ωt , by the following replacement:  

PR
¯ + γ5 γ ν B
pt µ pt ν
dΦ BΛ
f
R
, s−µ → 0,
(2.4)
s+µ → nµ = − gµν −
P
¯ +B
dΦ BΛ
m2t
¯ t (pt , s+ ), Λ+ ≡
where the matrix element for t (s+ ) → f + · · · was expressed as Bu
P
denotes the appropriate
p
/ t + mt , dΦ is the relevant final-state phase-space element and
spin summation.

3. Angular/energy distributions
(−)

In this section we present d2 σ/d xf d cos θf for the top-quark decay product f (= l ± / b ),
energy of f defined in terms of its energy Ef and the
where xf denotes the normalized
q 

2
top-quark velocity β ≡ 1 − 4mt /s as
s
2Ef 1 − β
,
xf ≡
mt
1+β
and θf is the angle between the e− beam direction and the f momentum, all in the e+ e−
CM frame.

6

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

Direct calculations performed in presence of the general decay vertex (2.2) lead to the
f
following result for the nµ vector defined in Eq. (2.4):  

pt µ pt ν
mt ν
p ,
(3.1)
nfµ = α f gµν −
2
pt pf f
mt
where for a given final state f , α f is a calculable depolarization factor

for f = l + ,

1
√  

f
α = 2r − 1 

8 r(1 − r)

1+
Re f2R
for f = b,

2r + 1
(2r − 1)(2r + 1)

(3.2)

¯
with r ≡ (MW /mt )2 . Similarly we have α f = −α f with replacement f2R → f¯2L . It
should be emphasized here that the above result means that there are no corrections to
the “polarization vector” nlµ for the semileptonic top-quark decay. On the other hand, one
can see that the corrections to α b could be substantial as the kinematical suppression factor
in the leading term 2r − 1(= −0.56) could be canceled by the appropriate contribution
from the non-standard form factor f2R .
Applying the strategy described above and adopting the general formula for the t t¯
distribution dσ (s+ , s− )/dΩt from Refs. [7,41], one obtains the following result for the
double distribution of the angle and the rescaled energy of f for longitudinally polarized
e+ e− beams:
2  

f (∗)
3πβαEM
d2 σ (∗)
f (∗)
f (∗)
Bf Θ0 (xf ) + cos θf Θ1 (xf ) + cos2 θf Θ2 (xf ) ,
=
dxf d cos θf
2s
(3.3)

where αEM is the fine structure constant and Bf denotes the appropriate branching fraction.
f (∗)
The energy dependence is specified by the functions Θi (xf ), explicit forms of which
for unpolarized beams were shown in Refs. [42,43]. 2 They are parameterized both by the
production and the decay form factors.
The angular distribution for f could be easy obtained from Eq. (3.3) by the integration
over the energy of f :
dσ (∗)

d cos θf

Zx+
dxf
x−

d2 σ (∗)
dxf d cos θf

2 

3πβαEM
f (∗)
f (∗)
f (∗)
Bf Ω0 + Ω1 cos θf + Ω2 cos2 θf ,
(3.4)
2s
Rx
f (∗)
f (∗)
= x−+ dx Θi
are shown by Eq. (A.1) in the appendix and x± define
where Ωi
kinematical energy range of x:

=

r(1 − β)/(1 + β) 6 xl 6 1

and (1 − r)(1 − β)/(1 + β) 6 xb 6 1 − r.

(3.5)

2 The functions Θ f (∗) (x ) for polarized beams could be easily obtained from formulas for unpolarized beams
f
i
(∗)
(∗)
(∗)
(∗)
replacing DV ,A,VA , EV ,A,VA , F1∼4 , G1∼4 defined by Eq. (A.18) with DV ,A,VA , EV ,A,VA , F1∼4 , G1∼4 as in

Eq. (A.17) in the appendix.

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

7

The decay vertex is entering the double distribution, Eq. (3.3), through (i) the functions
f
F f (xf ), Gf (xf ) and H1,2(xf ) defined in the appendix, and (ii) the depolarization factor
f

α f . All the non-SM parts of F f , Gf and H1,2 disappear upon integration over the energy
xf both for l + and b, as it could be seen from the explicit forms for Ωi . Since α f = 1 for
the leptonic distribution, we observe that the total dependence of the lepton distribution on
non-standard structure of the top-quark decay vertex drops out through the integration over
the energy [42,43]. 3 However, one can expect substantial modifications for the bottomquark distribution since corrections to α b could be large.
The fact that the angular leptonic distribution is insensitive to corrections to the V –
A structure of the decay vertex allows for much more clear tests of the production
vertices through measurements of the distribution, since that way we can avoid a
contamination from a non-standard structure of the decay vertex. As an application of the
angular distribution let us consider the following CP -violating forward-backward charge
asymmetry: 4
R0
R +cm
dσ +(∗) (θf )
dσ −(∗) (θ )
d cos θf d cos θff
−cm d cos θf d cos θf − 0
f
ACP (Pe− , Pe+ ) = R
,
(3.6)
R +cm
dσ +(∗) (θf )
dσ −(∗) (θf )
0
d
cos
θ
+
d
cos
θ
f
f
0
−cm
d cos θf
d cos θf
f (∗)

where Pe− and Pe+ are the polarizations of e and e¯ beams, dσ +/−(∗) is referring to f and
f¯ distributions respectively, and cm expresses the experimental polar-angle cut. As θf →
π − θf¯ under CP , this asymmetry is a true measure of CP violation. Since dσ −(∗) /d cos θf
(∗)
is obtained from dσ +(∗)/d cos θf by reversing the sign of cos θf and F1,4
terms and
¯

f (∗)

replacing α f with −α f in Ω0,1,2 , the asymmetry is explicitly given by the following
formula
f
f f
(3.7)
ACP = NA DA
with (in the leading order)
f   

1 − β2 1 + β
(∗) 
(∗) 
2
ln
1 − cm
Re F1 + cm Re F4
1−

1−β  

2 f
R
L
¯
1 − β α Re f2 − f2

f

NA = 2cm α0

− cm
1  

(0,∗) 
(0,∗)
2
+ cm E A
× 2 1 − cm Re DVA   

1+β
(0,∗) 
(0,∗)
(0,∗)  1
2
ln
− 2 1 − cm
Re DVA + cm EV + EA
2β 1 − β       

2
2
f
(0,∗)
2
2
1 − β 2 DV(0,∗) − 2cm 1 − 2β 2 − cm
1 − β 2 DA
DA = 2cm 1 + cm
3
3 
f 

(0,∗) 
2
− 4cm 1 − cm
α0 1 − β 2 Re DVA
3 The same conclusion has also been reached through a different approach using the helicity formalism in
Ref. [44].
4 Which is an integrated version of the asymmetry we have considered in Refs. [42,43].

8

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27 

f 
(0,∗)
(0,∗) 
2
− 2cm
α0 1 − β 2 EA + 2 Re EVA
n 
(0,∗)
f
(0,∗)
(0,∗) 
2
DV + DA + 2α0 Re DVA
+ c m 1 − cm
o
2 
f (0,∗)
1+β
(0,∗) 
(0,∗)  1 − β
+ 2 Re EVA
ln
,
+ cm α0 EV + EA
β
1−β

(3.8)

where all the coefficients are specified in the appendix, the superscript (0) indicates the
f
f
SM contribution and we expressed α f as α0 + α1 Re(f2R ) with
f

α0 = 1,
f

α0 =

f

α1 = 0

2r − 1
,
2r + 1

(for f = l),

8 r(1 − r)
f
α1 =
(1 + 2r)2

(for f = b).

As one could have anticipated, the asymmetry for f = l is sensitive to CP violation
(∗)
that
originating exclusively from the production mechanism: It depends only on F1,4
contains contributions from the CP -violating form factors δDγ and δDZ while the
contributing decay-vertex part consists of SM CP -conserving couplings only. For bottom
quarks the effect of the modification of the decay vertex is contained in the corrections
¯
to b and b¯ depolarization factors, α b + α b = α1b Re(f2R − f¯2L ), with SM CP -conserving
contributions from the production process. 5
It will be instructive to give the following remark here: The asymmetry is defined for
various initial beam polarizations Pe± . For Pe− 6= Pe+ , the initial state seems not to be CP
invariant and therefore one might expect contributions to the asymmetry originating from
the CP -conserving part of the top-quark couplings. However, as it is seen from Eq. (3.8),
this is not the case. It turns out that even for Pe− 6= Pe+ the asymmetry is still proportional
only to the CP -violating couplings embedded in F1,2,3,4 . The explanation is the following:
Whatever the polarizations of the initial beams are, the electron (positron) beam consists
of e(±1)(e(±1))
¯
where ±1 indicates the helicity, and only e(±1) and e(∓1)
¯
can interact
non-trivially in the limit of me = 0 since they couple to vector bosons. Therefore, the
interacting initial states are always CP invariant.
Now, since we have observed in Refs. [42,43] that the differential version of the
asymmetry discussed here could be substantial for higher collider energy, in order to
illustrate the potential power of the asymmetry we present in Tables 1 and 2 (as a function

of s ) the expected statistical significance (NSD ) for the asymmetry:
s
f
|ACP |

f

Leff σtot
= ACP
,
(3.9)
NSD ≡
f
f
1ACP
1 − (ACP )2
where Leff ≡ L is an effective integrated luminosity for the tagging efficiency . Hereafter
we adopt the integrated luminosity L = 500 fb−1 and the efficiency  = 60% both for
5 One can show that f L,R = ±f¯L,R and f L,R = ±f¯R,L where upper (lower) signs are those for CP 1
1
2
2
conserving (-violating) contributions [45,46]. Therefore, when only linear terms in non-standard form factors
are kept, any CP -violating observable defined for the top-quark decay must be proportional to f1L,R − f¯1L,R or
f L,R − f¯R,L .
2

2

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

9

Table 1
The CP -violating asymmetry AlCP and the expected statistical significance NSD for Re(δDγ ,Z ) =
+0.05, and beam polarizations Pe− = Pe+ = (1) 0, (2) +0.8 and (3) −0.8 as an example
(1) Pe− = Pe+ = 0

s (GeV)
AlCP
NSD

500

700

1000

1500

−1.2 × 10−2
2.42

−2.6 × 10−2
3.87

−4.0 × 10−2
4.32

−5.4 × 10−2
3.94

500

700

1000

1500

−1.4 × 10−2
2.80

−2.6 × 10−2
4.09

−3.6 × 10−2
4.04

−4.2 × 10−2
3.25

700

1000

1500

−2.6 × 10−2
5.72

−4.1 × 10−2
6.57

−5.7 × 10−2
6.20

(2) Pe− = Pe+ = +0.8

s (GeV)
AlCP
NSD

(3) Pe− = Pe+ = −0.8

s (GeV)
500
AlCP
NSD

−1.2 × 10−2
3.53

lepton and b-quark detection. 6 In addition, to fit the typical detector shape [50] we impose
a polar-angle cut | cos θf | < 0.9, i.e., cm = 0.9 in Eq. (3.8), both for leptons and bottom
quarks. On the other hand, we will not impose any cut on the lepton/b-quark energy

since their kinematical lower bounds Elmin = 7.5 GeV and Ebmin = 27.5 GeV (for s =
500 GeV) are large enough to be detected. Perfect angular resolution will be assumed both
for lepton and b-quark final states. Also ideal leptonic-energy resolution will be used.
f
As it is seen from the tables, the asymmetry ACP turned out to be a very sensitive CP violating observable; even for unpolarized beams and CP -violating couplings of the order
of 0.05 one can expect 2.4σ ∼ 5.5σ effect both for lepton and b-quark asymmetries once
L = 500 fb−1 is achieved.
The CP -violating form factors discussed here could be also generated within the SM.
However, it is easy to notice that the first non-zero contribution to δDγ ,Z would require at
least two loops. For the top-quark decay process CP violation could appear at the one-loop
level, however it is strongly suppressed by the double GIM mechanism [51]. Therefore we
can conclude that an experimental detection of CP -violating form factors considered here
6 That low efficiency is supposed to take into account cuts necessary to suppress the background [47–49]. If the
b-tagging is applied then, as shown in the second paper of Ref. [48], the irreducible background to top events due
Therefore for
to W ± + 2b + 2j is negligible, provided that a vertex tagging efficiency b & 0.5 can be achieved. √
the b-tagging case the efficiency we have employed is definitely conservative. Since NSD scales as L it would
be easy to estimate the statistical significance for any given luminosity and efficiency.

10

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

Table 2
The CP -violating asymmetry AbCP and the expected statistical significance NSD for Re(δDγ ,Z ) =
Re(f2R − f¯2L ) = +0.05, and beam polarizations Pe− = Pe+ = (1) 0, (2) +0.8 and (3) −0.8 as an
example
(1) Pe− = Pe+ = 0

s (GeV)
AbCP
NSD

500

700

1000

1500

+1.2 × 10−2
5.10

+1.7 × 10−2
5.50

+2.2 × 10−2
5.03

+2.6 × 10−2
4.03

700

1000

1500

−4.6 × 10−3
1.52

+1.4 × 10−3
0.33

+7.8 × 10−3
1.27

700

1000

1500

+3.0 × 10−2
14.3

+3.3 × 10−2
11.2

+3.5 × 10−2
8.02

(2) Pe− = Pe+ = +0.8

s (GeV)
500
−9.4 × 10−3
4.04

AbCP
NSD

(3) Pe− = Pe+ = −0.8

s (GeV)
500
+2.6 × 10−2
16.0

AbCP
NSD

would be a clear indication for physics beyond the SM. In particular, non-vanishing AlCP
in the lepton distribution will strongly indicate some new-physics in t t¯γ /Z couplings.

4. Optimal-observable analysis
4.1. Optimal observables
Let us briefly recall the main points of the optimal-observable (OO) technique [35] (see
also [15,36–38]). Suppose we have a distribution 
X

≡ Σ(φ) =
ci fi (φ)
(4.1)

i

where fi (φ) are known functions of the location in final-state phase space φ and ci ’s are
model-dependent coefficients. The goal would be to determine
ci ’s. It can be done by
R
using appropriate weighting functions wi (φ) such that dφ wi (φ)Σ(φ) = ci . Generally,
different choices for wi (φ) are possible, but there is a unique choice so that the resultant
statistical error is minimized. Such functions are given by
X
Xij fj (φ)/Σ(φ),
(4.2)
wi (φ) =
j

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

11

where Xij is the inverse matrix of Mij which is defined as
Z
fi (φ)fj (φ)
.
Mij ≡ dφ
Σ(φ)

(4.3)

The statistical uncertainty of ci -determination through dσ/dφ measurement becomes
p
(4.4)
1ci = Xii σT /N ,
R
where σT ≡ dφ (dσ/dφ) and N is the total number of events.
It is clear from the definition of the matrix Mij , Eq. (4.3), that Mij has no inverse if
the functions fi (φ) are linearly dependent, and then we cannot perform any meaningful
analysis. One can see it more intuitively as follows: if fi (φ) = fj (φ) the splitting between
ci and cj would be totally arbitrary and only ci + cj could be determined.
4.2. For application
In order to apply the OO procedure to the processes under consideration, we have
to reexpress the distributions in the form shown in Eq. (4.1). The angular distribution,
Eq. (3.4), has already an appropriate form for this purpose, where fi (φ) = cos i θf (i =
f (∗)
are the coefficients to be determined. On the other hand, the double
0, 1, 2) and Ωi
angular and energy distribution Eq. (3.3) must be modified. We reexpress the distribution
in the following way, keeping only the SM contribution and terms linear in the non-standard
form factors:
3πβα 2
d2 σ (∗)
Bf Sf(∗) (xf , θf ),
=
dxf d cos θf
2s

(4.5)

where
Sf(∗) (xf , θf ) = Sf(0,∗) (xf , θf )
X 
f (∗)
f (∗)
Re(δAv )FAv (xf , θf ) + Re(δBv )FBv (xf , θf )
+
v=γ ,Z
f (∗)

f (∗)

+ Re(δCv )FCv (xf , θf ) + Re(δDv )FDv (xf , θf )  

f (∗)
+ Re f2R F2R (xf , θf ).
As it is seen from the above formula, the coefficients ci of Eq. (4.1) are just the anomalous form factors to be determined. The SM contribution reads:
f (0,∗)

Sf(0,∗) (xf , θf ) = Θ0

f (0,∗)

(xf ) + cos θf Θ1

with
f (0,∗)

Θ0

(x) =

f (0,∗)

(xf ) + cos2 θf Θ2 

(0,∗) 
(0,∗)
1
3 − β 2 DV − 1 − 3β 2 DA
2 

f
(0,∗)  f
− 2α0 1 − β 2 Re DVA
f (x) 

f
(0,∗) f
g (x)
+ 2α0 Re DVA

(xf )

(4.6)

12

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27 

1  (0,∗)
f
f
(0,∗)
(0,∗)  f
+ DV + DA + 2α0 Re DVA
(4.7)
2h1 (x) − h2 (x) ,
2    

f (0,∗)
f
(0,∗)
(0,∗) f
(x) = 2 2 Re EVA + α0 1 − β 2 EA
f (x)
Θ1
f
(0,∗)
(0,∗)  f
+ 2α0 EV + EA
g (x) 

f
(0,∗) 
(0,∗)  f
(4.8)
h1 (x),
− 2 2 Re EVA + α0 EV(0,∗) + EA     

1
f (0,∗)
f
(0,∗)
(0,∗)
(0,∗)
3 − β 2 DV + DA
(x) =
+ 6α0 1 − β 2 Re DVA
f f (x)
Θ2
2
f
(0,∗) 
+ 2α0 Re DVA g f (x) 

3
f
f
(0,∗)
(0,∗)  f
+ 2α0 Re DVA
(4.9)
− DV(0,∗) + DA
2h1 (x) − h2 (x) .
2
f (∗)

f (∗)

Explicit forms of the functions F{A,B,C,D}{γ ,Z} and F2R
together with the functions

f f (x),

f
and h1,2 (x).
(0,∗)
Eq. (4.5): Sf ,

are shown in the appendix

g f (x)

f (∗)

f (∗)

F{A,B,C,D}{γ ,Z} and F2R . As
There are ten functions entering
explained earlier, one cannot determine their coefficients separately if they are not
independent. As could be found from the appendix, for the double lepton distribution,
the first nine functions are linear combinations of
f l (x),

f l (x) cos θ,

f l (x) cos2 θ,

g l (x), g l (x) cos θ, g l (x) cos2 θ,
hl1,2 (x)(1 − 3 cos2 θ ), hl1 (x) cos θ,

(4.10)

l(∗)

while the last one, F2R , is a combination of δ{f l , g l , hl1,2 }(x) and cosn θ (n = 0, 1, 2).
Since there are ten coefficients to be measured, 7 it looks always possible to determine all
of them. However, it turns out not to be the case on some special occasions. Indeed the
possibility for the determination of all the ten form factors depends crucially on the chosen
beam polarization.
This can be understood considering the invariant amplitude for ee¯ → t t¯, which could be
expressed in terms of eight independent parameters as    

M(ee¯ → t t¯) = CVV v¯e γµ ue · u¯ t γ µ vt¯ + CVA v¯e γµ ue · u¯ t γ5 γ µ vt    

+ CAV v¯e γ5 γµ ue · u¯ t γ µ vt + CAA v¯e γ5 γµ ue · u¯ t γ5 γ µ vt    

+ CVS v¯eq/ ue · u¯ t vt + CVP v¯eq/ ue · u¯ t γ5 vt    

+ CAS v¯e γ5q/ ue · u¯ t vt + CAP v¯e γ5q/ ue · u¯ t γ5 vt .
However, if e or e¯ is perfectly polarized, contributions from [v¯e γµ ue ] and [v¯e γ5 γµ ue ] are
identical. For example, when e has fully left-handed polarization, ue is replaced with ueL ≡
(1 − γ5 )ue /2 and in that case they are changed as
v¯e γµ ue → v¯e γµ ueL ,

v¯e γ5 γµ ue → v¯e γµ ueL .

Therefore, the invariant amplitude becomes
7 Counting the SM coefficient in front of S (0,∗) which is normalized to 1.
f

B. Grzadkowski, Z. Hioki / Nuclear Physics B 585 (2000) 3–27

13

M(ee¯ → t t¯)    

= (CVV + CAV ) v¯e γµ ueL · u¯ t γ µ vt¯ + (CVA + CAA ) v¯e γµ ueL · u¯ t γ5 γ µ vt    

+ (CVS + CAS ) v¯eq/ ueL · u¯ t vt + (CVP + CAP ) v¯eq/ ueL · u¯ t γ5 vt ,

and one ends with just four independent functions and therefore only four coefficients
could be determined. More details could be found in the appendix below Eq. (A.16). Of
course, such singular configurations of polarization are not considered in our analyses.
As for b-quark distributions δf b (x) = δg b (x) = δhb1 (x) = δhb2 (x) = 0, instead of ten
functions φi (x) we have in that case only nine of them given by the b-quark version
Eq. (4.10). Therefore, at most nine couplings could be determined. Since b-quark energy
resolution is expected to be relatively poor, we will not apply OO procedure to the b-quark
double distribution.
4.3. Numerical analysis
Below, we will adjust beam polarizations to perform the best measurement of the
form factors. In order to gain some intuition we show in Figs. 1 and 2 the functions
l(∗)
l(∗)
(0,∗)
for unpolarized beams (Fig. 1) and for the beam
F{A,B,C,D}{γ ,Z} , F2R plus Sl
polarization Pe− = Pe+ = +0.5 (Fig. 2). The figures illustrate how much the polarization
could modify the functions and therefore influence the possibility for the determination of
the form factors.
4.3.1. Lepton angular distribution
Since we have only three independent functions {1, cos θ, cos2 θ }, M and its inverse X
are (3, 3) matrices. We have considered the following polarization set-ups: Pe− = Pe+ =
0, ±0.5 and ±1. Since 1 > | cos θ | > cos2 θ we observe that X11 < X22 < X33 , therefore
(∗)
(∗)
the statistical uncertainty for Ω0 measurement, 1Ω 0 , is always the smallest one.
Once we assume the detection efficiency  and the integrated luminosity L, we can
compute the statistical significance of measuring the non-SM part of Ωi(∗)

.

(i) NSD = .

∗).Ωi(∗) − Ωi(0.

X22 = 1.5. X11 = 0.5. M11 = 3./1Ωi(∗).3.06. M33 = 0. (2) • Pe− = Pe+ = +0. (0) NSD (1) NSD = 3.5 = 7.82. M22 = 0. NSD = 1.7.66.06. M22 = 0. For the efficiency and luminosity specified earlier we obtain • Pe− = Pe+ = 0 M11 = 2. = 2.55.09. X22 = 1. M33 = 0.49. X33 = 8.25. (0) NSD (1) NSD (2) NSD = 1.1. = 16.95. X11 = 1.9.40.83.27. X33 = 6. .

Z.Z} . 1.B.∗) Fig. Hioki / Nuclear Physics B 585 (2000) 3–27 l(∗) l(∗) (0. and Sl beams. F2R . for unpolarized .C. Grzadkowski. The shape of the coefficient functions F{A.14 B.D}{γ .

The shape of the coefficient functions F{A.Z} . Z. 2.D}{γ . and Sl Pe− = Pe+ = 0.5.∗) Fig. for . F2R .B.B.C. Hioki / Nuclear Physics B 585 (2000) 3–27 l(∗) l(∗) 15 (0. Grzadkowski.

X33 = 7.3.11) where we put all the non-SM parameters Re(δ{A.06. X33 = 7.3.6.2.8. positively-polarized beams give smaller Xii and this is independent of the choice of non-SM parameters.96.90.22.7.29. (0) NSD (1) NSD = 7. = 37. X11 = 0. Z.4. In fact. Therefore polarization of the initial beams should be carefully adjusted for each tested model in actual experimental analysis.5.16 B. (2) • Pe− = Pe+ = +1 = 9. X22 = 1.39.5. X33 = 7.2.0.9. M22 = 1. M22 = 0. b-quark angular distribution (i) We can compute M.5 = 26.45. (0) = 5.91. = 35. (2) • Pe− = Pe+ = −1 M11 = 0.65. X11 = 1. M33 = 0.31.8. • Pe− = Pe+ = +0. M33 = 0. (2) NSD = 1. 4.27.98. M22 = 0.0.11. Grzadkowski.8.63.35. X11 = 3. (2) (4. M33 = 0.21. X22 = 2.23. X11 = 1. M22 = 0. B.5. partly because of larger number of events.81. X22 = 2.2.00. X22 = 1.76.61. M11 = 1.0. NSD (1) NSD = 3.9.29. M33 = 0. M33 = 0. X33 = 22.85. (0) NSD (1) NSD = 17.4. the precision is better for negative beam polarization.63. NSD = 1. = 4.17.25. since NSD the non-SM parameters used in the computations. X11 = 1. However we cannot conclude that using negatively-polarized (i) strongly depends on beams is always more effective for new-physics search. (0) NSD (1) NSD NSD = 3. M11 = 1. (0) NSD (1) NSD (2) NSD = 2.39. .72. Hioki / Nuclear Physics B 585 (2000) 3–27 • Pe− = Pe+ = +1 M11 = 3.32. D}γ .38. M22 = 0. X and NSD in the same way as for the lepton distribution: • Pe− = Pe+ = 0 M11 = 2. X33 = 13. = 4. (2) NSD = 1. (0) NSD (1) NSD NSD = 4. X22 = 4.Z ) and Re(f2R ) to be +0. = 6.04. • Pe− = Pe+ = −0. X11 = 1.00. M22 = 0. X22 = 2. As one can see.3. M33 = 0. X33 = 10. C.3.2.20.5 = 15.05 as an example. M11 = 2.

3.09. (2) (4.5 and Pe+ = 0. our final goal is to determine each form factor separately.2 and Pe+ = 0. also for OO analysis. δDZ : 20.6. X11 = 1.43. Grzadkowski.Z ) = precision again.87.7.10.13. however. M33 = 0. = 85.1.49. Unfortunately.70 .12) = +0.16 for Pe− = 0.11. (0) NSD (1) NSD NSD = 3. X33 = 21.03 . = 29.47 .   δAZ : 0.09 for Pe− = 0. we can only constrain them. δDγ : 7.69. δDZ : 14.63.6.1. X22 = 3. the optimal polarizations for each form-factor measurement is different. (0) NSD (1) NSD (2) NSD = 2.2.02 1[Re(δBγ )] = 0.B. (4. The above results prove that the optimal observables utilizing the angular distributions f (∗) should be very efficient seeking for the non-SM parts of Ωi . δBZ : 0.17. M22 = 0. Below we present the smallest statistical uncertainties and the corresponding beam polarizations for each parameter: 1[Re(δAγ )] = 0.6.4.25.7 and Pe+ = 0. δCγ : 2. δAZ : 0.01. Re(f2R ) 4. δDγ : 27. As discussed earlier. δDZ : 53. it would be exciting if we found any signal of non-standard physics. C. B. D}γ . M33 = 0. Eq.11.05.0. Lepton angular and energy distribution Because of high precision of direction and energy determination of leptons we adopted the double energy and angular distributions.74. δBγ : 0.95.7. However. Hioki / Nuclear Physics B 585 (2000) 3–27 17 • Pe− = Pe+ = −0.02 1[Re(δAZ )] = 0.   δAγ : 0.Z ) and Re(f2R ).21.2. B.76.4. δCZ : 1.5.5). That is why we proceed to the next analysis using the double angular and energy distributions.42. Z. δDγ : 0. f2R : 0. We found that positive polarizations lead to a smaller 1ci for eight form factors in the production vertices. δCγ : 0. δBZ : 0.25. = 62. C.13. X33 = 12. X11 = 2. δBγ : 0. δCγ : 0. Negatively-polarized beams give better for Re(δ{A. in principle all nine form factors could be determined with the expected statistical uncertainties 1ci for ci = Re(δ{A. • Pe− = Pe+ = −1 M11 = 0.21 .36.   δAγ : 0. f2R : 0. however. = 39.27.23.3.2. D}γ . δBZ : 0.45. X22 = 5. Of course. The beam polarizations Pe− and Pe+ were adjusted to minimize the statistical error for determination of each form factor. since they are combinations of the form factors.5 M11 = 1.07 for Pe− = 0. δCZ : 4.3. M22 = 0. but the same remark as to the lepton angular distribution should be kept in mind also here. f2R : 0. δCZ : 1.

The results are as follows: . 9. δBγ : 0. δDγ : 1.43. .10.15. as well. f2R : 0.27 for Pe− = 0. The source of that sensitivity is hidden in the neutral-current structure with sin2 θW ' 0. δCγ : 0.13. f2R : 0. For example. δBZ : 0.   δAγ : 0.99 for Pe− = 0.18 B. Hioki / Nuclear Physics B 585 (2000) 3–27 1[Re(δBZ )] = 0. is O(1) for i. D}Z measurement would be very poor even for l(∗) .22. δCγ : 0. respectively. while the size of Mi7 (= M7i ) and Mi9 (= M9i ) is at most O(10−2 ). δCγ : 0. whereas even for the unpolarized beams we obtain 1[Re(f2R )] = 0.11.25.3/Pe+ = 0. More quantitatively.4.13.65 . δBγ : 0.6.65 .1 instead of 0.2 (1[Re(δDγ )] = 0.D}Z is illustrated in Figs.8 and Pe+ = −0. Indeed. a good determination (almost independently of the polarization) could be expected for f2R .56. δDZ : 14.23.22. δDγ : 1. Re(δBγ ).4 for Pe− = 0. As it is seen the precision of δ{C.1. δCZ : 1.01 for Pe− = −0.11. δBZ : 0.2 and Pe+ = 0. f2R : 0.13. δAZ : 0. δCZ : 1. δCZ : 1.25.03 (4.8. δBZ : 0.25. δBv and f2R are known and OO are used to determine δCv and δDv only (here we put δAv = δBv = f2R = 0 for simplicity). j 6= 7.08 for Pe− = 0. Z. δDZ : 14. the optimal polarization becomes Pe− = 0.07.09) for sin2 θW = 0.11 for Pe− = 0. f2R : 0. Below we provide an example of a combined analysis assuming δAv . In addition.10. Mij .   δAγ : 0.4 and Pe+ = 0. 1 and 2: these two functions are very small in a large area. δBγ : 0. .1 and Pe+ = 0. δAZ : 0.52. δBγ : 0.82.2 and Pe+ = 0. 1[Re(δDγ )] becomes 0. At the time a linear collider will be operating.1. .   δAγ : 0.13) where we also showed the expected precision of the other parameter measurements for the same beam polarizations. 0.76. δDZ : 18.76. the best precision is   (4. δAZ : 0.52.42 . for the same polarizations are 0.14) 1 Re f2R = 0.1.16 for Pe− = Pe+ = 0.13. Grzadkowski.6. δDγ : 0.08. we can expect 1[Re(δAγ )] = 0. which the optimal polarization. δAZ : 0.86 for Pe− = 0. .03 1[Re(δDγ )] = 0. data from Tevatron Run II and LHC will also provide independent constraints on top-quark couplings. δBZ : 0.4. Indeed. .1 and 0.11 for Pe− = 0. the size of the elements of M matrix.15.1/Pe+ = 0.1 and Pe+ = 0.56 . This result is independent of the choice of the non-SM parameters in contrast to the preceding results. determination of δDγ would be practically difficult.02 1[Re(δCγ )] = 0. . .09.42 .5.0. δDZ : 14.7 while the expected precisions of Re(δAZ ).0. δDγ : 5. δAZ : 0. f2R : 0. δBγ : 0. On the other hand. since its error varies rapidly with the polarization. For instance.   δAγ : 0.03 1[Re(δDZ )] = 14.03. δCZ : 1.09.10.82.13. δCγ : 0.   δAγ : 0. This is mainly a consequence of the size of F{C.03 1[Re(δCZ )] = 1.

It turned out that in the case of the bottom-quark angular distribution the statistical significance of the signal is in general higher than for leptons because of larger event rate and varies between 1.2 and Pe+ = 0. Therefore.97 for Pe− = 0. 5.1σ for the same non-standard form factors. However. Summary and conclusions (−) We have presented here the angular and energy distributions for f in the process (−) e+ e− → t t¯ → f · · ·. the asymmetry could be utilized for a pure test of CP -violation in the top-quark production process. 19 (4. The expected statistical significance NSD for the measurement of the asymmetry f has been calculated. The most general (CP -violating and CP -conserving) couplings for γ t t¯. However.2. the CP -sensitive asymmetry ACP would provide much stronger constraints on δDγ .2 and Pe+ = 0. final results for statistical significances considered here depend on the size of the non-standard parameters. as f we have seen in Section 3. However. Having both beams polarized at 80% the signal for bottom quarks (leptons) could reach even 16σ (6. When deriving the above results we have fixed all the non-SM parameters to be +0. one should stress that the beam polarizations should be carefully adjusted for each model to be tested in actual experimental analysis.8σ and 85. Zt t¯ and W tb have been assumed. where f = l or b quark in the form suitable for an application of the optimal observables (OO). f The most convenient beam polarizations for a measurement of the asymmetry ACP and for testing the angular distributions varies with the non-standard parameters. In the case of the lepton angular distribution.5σ assuming non-standard form factors of the order of 0.B.Z . in any case.1.9 and the integrated luminosity L = 500 fb−1 .6σ ). as well.23 for Pe− = 0.1. assuming CP -violating couplings of the order of 0. We have assumed the tagging efficiency at the level of 60% both for lepton and b quark detection. 1[Re(δCZ )] = 0.2 and Pe+ = 0.5σ (4.04 for Pe− = 0. 1[Re(δDZ )] = 2. f CP -violating charge forward-backward asymmetry ACP has been introduced as an efficient way for testing CP -violation in top-quark couplings.2 and Pe+ = 0. the range of the polar angle restricted by | cos θf | < 0.05. Since the angular distribution for leptons is insensitive to variations of the standard V –A structure of the W tb coupling. 1[Re(δDγ )] = 0.03 for Pe− = 0. Hioki / Nuclear Physics B 585 (2000) 3–27 1[Re(δCγ )] = 0. the expected statistical significance for signals of non-standard physics varies between 1. . We have found that it should be possible to detect ACP at 5. Z. Next.3σ and 35.05 as a reasonable example for the strength of beyond-the-SM physics. the OO procedure has been applied to the angular distributions. the above f results show that a measurement of ACP and OO analysis of the angular distributions are both very efficient for new-physics search.3σ ) level for bottom quarks (leptons) for unpolarized beams. All fermion masses except mt have been neglected and we have kept only terms linear in anomalous couplings.1.15) The error for δDZ became much smaller but still too large for practical use.05. Grzadkowski.

4) are the following:  (∗) f (∗) (∗) (∗)  Ω0 = DV − 1 − 2β 2 DA − 2 Re G1      (∗)  − Re F1(∗) + 3 − 2β 2 Re G(∗) − α f 2 1 − β 2 Re DVA 3   (∗) + 2 Re G(∗) + DV(∗) + DA 1 (∗) (∗)  1 − β (∗) + α f Re 2DVA − F1 + 3G3 2  2β ln 1+β . Concluding. 1−β (∗) (∗)  = 4 Re − Re F4 − G2 1−β n  o 1 − β 2 1 + β (∗)  (∗) ln . Appendix A f (∗) f (∗) in the angular distribution Integrals of Θi (x) denoted in the main text by Ωi Eq. Z. Nagano.4.) is grateful to Y. Acknowledgements We would like to thank K. 1−β (A. Fujii for useful discussions concerning details of lepton and bottom-quark detection and J. This result is independent of the choice of the non-SM parameters in contrast to the above two types of analyses. Ikematsu for stimulating discussions. One of us (Z. we have observed that the angular distributions and the angular and energy distributions of top-quark decay products both provide very efficient tools for studying top-quark couplings to gauge bosons at linear colliders. T.20 B. (3. Sumino. Hioki / Nuclear Physics B 585 (2000) 3–27 Then we have analyzed the angular and energy distribution of the lepton toward separate determinations of the anomalous form factors. Takahashi and K. Pliszka for his remarks on the statistical analysis. Grzadkowski. T. This work is supported in part by the State Committee for Scientific Research (Poland) under grant 2 P03B 014 14 and by Maria Skłodowska-Curie Joint Fund II (Poland–USA) under grant MEN/NSF-96-252.07 while the lowest precision is expected for Re(δDZ ) with 1[Re(δDZ )] = 14. We have found that at s = 500 GeV with the integrated luminosity L = 500 fb−1 the best determined coupling would be the axial coupling of the Z boson with the error 1[Re(δAZ )] = 0.1) . − Re F4(∗) − G(∗) + α f EV(∗) + EA − 2 Re EVA 2 β 1−β  (∗) f (∗) (∗) (∗)  Ω2 = 3 − 2β 2 DV + DA + 2 Re G1    (∗)  (∗)  (∗)  + 3α f 2 1 − β 2 Re DVA − Re F1 + 3 − 2β 2 Re G3  (∗) (∗) (∗)  − 3 DV + DA + 2 Re G1 f (∗) Ω1  (∗)  EVA + 2α f 2 (∗) EA (∗) − F1(∗) + 3G(∗) + α f Re 2DVA 3  1 − β 2 2β ln 1+β . In order to reach the highest precision we have been adjusting beam polarizations to minimize errors for each form √ factor.H.

θ ) and F2R (x. f (∗) FCv (x. f (∗) FDv (x. θ )  = −β 2 C(G1 : Cv )f f (x) 1 + cos2 θ   f f + 2α0 C(G2 : Cv ) f f (x) + g f (x) − h1 (x) cos θ h  f f − C(G1 : Cv ) + α0 2 − β 2 C(G3 : Cv ) f f (x) + α0 C(G3 : Cv )g f (x) f  f − 2C(G1 : Cv ) + 3α0 C(G3 : Cv ) h1 (x) f i   f (A. (A. Grzadkowski.5) F{A.B.2) f (∗) FBv (x. θ ) (f = l/b): f (∗) FAv (x. (4.B.Z}(x.5) . Z.3) − α0 C(EA : Bv ) + 2C(EVA : Bv ) h1 (x) cos θ.4) + C(G1 : Cv ) + α0 C(G3 : Cv ) h2 (x) 1 − 3 cos2 θ .D}{γ . θ ) takes different forms for f = l and f = b as while F2R l(∗) (x. Hioki / Nuclear Physics B 585 (2000) 3–27 21 Next we present explicit formulas of the coefficient functions for the nine anomalous f (∗) f (∗) form factors in Eq. θ )    1 f = (3 − β 2 )C(DV : Av )f f (x) + 2α0 C(DVA :Av )g f (x) 1 + cos2 θ 2  f − α0 (1 − β 2 )C(DVA : Av )f f (x)   f  1 f f − C(DV : Av ) + 2α0 C(DVA : Av ) 2h1 (x) − h2 (x) 1 − 3 cos2 θ 2 h   i f f f + 2 α0 C(EV : Av ) g f (x) − h1 (x) + 2C(EVA : Av ) f f (x) − h1 (x) × cos θ. θ )   1 f = β 2 C(DA : Bv )f f (x) 3 − cos2 θ + 2α0 C(DVA : Bv )g f (x) 1 + cos2 θ 2  1 h f C(DA : Bv ) + 2α0 1 − β 2 C(DVA : Av ) f f (x) − 2  f i   f f − C(DA : Bv ) + 2α0 C(DVA : Bv ) 2h1 (x) − h2 (x) 1 − 3 cos2 θ h  f f + 2 α0 1 − β 2 C(EA : Bv ) + 2C(EVA : Bv ) f f (x) + α0 C(EA :Bv )g f (x) f i  f (A. θ )    f f = α0 C(F1 : Dv ) f f (x) − h1 (x) 1 − 3 cos2 θ  f − α0 C(F1 : Dv )g f (x) 1 + cos2 θ   f f − 2α0 C(F4 : Dv ) f f (x) + g f (x) − h1 (x) cos θ. l(∗) (x. θ ) F2R (A.C.

2 (x) are defined as  F f (x) = f f (x) + Re f2R δf f (x). Gf (x) ≡ .  f f f H1. g f (x).∗)  l δf (x) cos θ + 2 1 − β 2 EA + 2 Re EVA (0.∗)  l 3 − β 2 DV − 1 − 3β 2 DA − 2 1 − β 2 Re DVA δf (x) 2   (0. δg f (x) and δh1.9) dω(1 − ω)2 x Γt dx dω Z dω and ω is defined as ω ≡ (pt − pf )2 /m2t .7) + 2 1 − β 2 EA f b (x) + EV + EA f f where the functions f f (x).∗)  (0. Hioki / Nuclear Physics B 585 (2000) 3–27 =  (0.43] F f (x) ≡ 1 Bf f 1 Bf f 1 Bf H1 (x) ≡ H2 (x) ≡   Z 1 1 d 2 Γf 1 + β 1 d 2 Γf . .2 (x) + Re f2R δh1.  Gf (x) = g f (x) + Re f2R δg f (x).∗)  (0. θ ) F2R n   (0.∗)  l 3 − β 2 DV + DA + 6 1 − β 2 Re DVA δf (x) cos2 θ. (A. Z.2(x) being given as follows [42. (A.∗) (0.∗)  b g (x) − hb1 (x) cos θ . dω 1 − x Γt dx dω Bf 1 − ω Γt dx dω Z 1−β 1 d 2 Γf .∗) + 2 Re DVA δg l (x) 1 + cos2 θ   1  (0.∗)   = α1b Re DVA − 1 − β 2 f b (x) − 2hb1 (x) + hb2 (x) 1 − 3 cos2 θ  + 2g b (x) 1 + cos2 θ o   (0. dω(1 − ω) x Γt dx dω  Z  1 d 2 Γf 1−β 2 . After performing the above integrations using     √ 1 + β 3Bl 3 1  R  − r  β W ω 1 + 2 Re f2 1 d 2 Γf 1 − ω 1 + 2r = Γt dx dω  1+β   δ(ω − r) 2β(1 − r) for f = l + . h1.∗)  (0.8) f with F f (x).∗)  l δh1 (x) cos θ − 2 EV + EA + 2 Re EVA    1 (0.2 (x).6) and b(∗) (x. for f = b. + 2 (A.2 (x) = h1.∗)  1 (0.2 (x).∗) (0.∗) (0. δf f (x).∗)  2δhl1 (x) − δhl2 (x) 1 − 3 cos2 θ + DV + DA + 2 Re DVA 2  (0. Grzadkowski.22 B. Gf (x) and H1.∗) (0.∗) (0.∗)  l δg (x) cos θ + 2 EV + EA  (0. (A.∗) (0.∗) (0.∗) (0.

ω− = 0 for Br 6 x < B for B 6 x < r for r 6 x 6 1 (A. ω− = 1 − x/B ω+ = 1 − x. ω− = 1 − x/B ω+ = 1 − x.2 (x) for leptonic and bottom-quark final states: • For f = l:    3(1 + β)  2 ω+ 3(1 + β) 2 2 ω ω− ≡ ω+ − ω− . g (x) = 1 − 1−r 2β(1 − r) f b (x) = . βW 1 + 2r ω− 6(1 + β)2 √ rx δg l (x) = δf l (x) − βW    ω+ 3 1 + . Z.11) for Br 6 x < r for r 6 x < B for B 6 x 6 1 (A. ω− = 0 ω+ = 1 − x. δf f (x). g l (x) = f l (x) + − βW 2 ω 1 − β 1 2 ω (3 − 2ω) ω+− . Grzadkowski.10) where ω± are given as follows: 2 /m2 and B ≡ (1 − β)/(1 + β)) For r > B (r ≡ MW t ω+ = 1 − r.2 (x). δg f (x) and f δh1.B. δh1 (x) = βW x 1 + 2r ω− δhl2 (x) = 1 (1 + β)(1 − β)2 2βW √    ω+ 3ω2 r × 2 2ω2 (3 − 2ω) − . 2β(1 − r)   1+β 1+β b x . h2 (x) = − 4βW x δf l (x) = −   3(1 + β) √ 3ω2 ω+ r 2ω + 2 ln |1 − ω| + . h1. ω + ln |1 − ω| × ln |1 − ω| + 1 − ω 1 + 2r ω− √    3(1 − β 2 ) r 2 3 − 2ω ω+ l ω 1− . f l (x) = 2βW 2βW ω 3(1 + β)2  x ω + ln |1 − ω| ω+ . Hioki / Nuclear Physics B 585 (2000) 3–27 23 f one obtains the following explicit forms of f f (x). ω− = 1 − x/B ω+ = 1 − r.12) For r < B ω+ = 1 − r. g f (x). hl1 (x) = 2βW x ω 1 1 l (1 + β)(1 − β)2 2 ω2 6 − 8ω + 3ω2 ω+ . ω− = 0 • For f = b: 1+β (= constant). 6 − 8ω + 3ω2 1 + 2r x ω− (A.

The coefficients C(X : Y ) employed in the definition of the coefficient functions have been introduced through the following formulas: (∗) (0.∗)  = Re DVA Re DVA X   C(DVA : Av ) Re(δAv ) + C(DVA : Bv ) Re(δBv ) . P⊗ ≡ 1 + Pe− Pe+ . C(DVA : Aγ ) = −C(P⊕ + ve P⊗ )d 0 BZ . (0.Z (∗)  (0.VA .14) v=γ .A.∗) and in the analogous manner for EV . (A.    (A. (∗) .A. Z.∗) (∗) G1∼4 could be obtained from Eq.∗) (0.17) below as a SM approximation of DV . v=γ . v=γ . hb2 (x) = 2βx 2 hb1 (x) = δf b (x) = δg b (x) = δhb1 (x) = δhb2 (x) = 0.A. (A. 2βx (1 − r)(1 + β)(1 − β)2 .Z (0. + (A. Hioki / Nuclear Physics B 585 (2000) 3–27 1 − β2 .   C(EV : Aγ ) = −2C P⊕ Aγ − (P⊗ + ve P⊕ )d 0 AZ . F1∼4 1∼4 .    C(DVA : AZ ) = C 2ve P⊕ + 1 + ve2 P⊗ d 02 BZ .     C(DV : AZ ) = −2C (P⊕ + ve P⊗ )d 0 Aγ − 2ve P⊕ + 1 + ve2 P⊗ d 02 AZ . F1∼4 .A. C(EVA : Aγ ) = C(P⊗ + ve P⊕ )d 0 BZ .13) where x is bounded as B(1 − r) 6 x 6 1 − r.∗) (0.∗) + DA X C(DA : Bv ) Re(δBv ).VA . G(∗) EV(∗). two polarization factors P⊕ and P⊗ are defined as P⊕ ≡ Pe− + Pe+ .∗) DV = DV + X C(DV : Av ) Re(δAv ).     C(EV : AZ ) = 2C (P⊗ + ve P⊕ )d 0 Aγ − 2ve P⊗ + 1 + ve2 P⊕ d 02 AZ . where ve = −1 + 4 sin2 θW . EV .Z (∗) = DA(0.VA . F1∼4 and G1∼4 . Explicit forms of the independent coefficients are given as   C(DV : Aγ ) = 2C P⊗ Aγ − (P⊕ + ve P⊗ )d 0 AZ .15) C(EVA : AZ ) = −C 2ve P⊗ + 1 + ve2 P⊕ d 02 BZ .VA . d 0 ≡ s/[4 sin θW cos θW (s − MZ2 )]. and the others are thereby given as .VA .A. Grzadkowski.24 B. DV .

2. C(G2 : Cv ) = C(EV : Av )/2.A.    EV ≡ 2C Aγ AZ d 0 − A2Z ve d 0 2 + AZ d 0 Re(δAγ ) + Aγ d 0 − 2AZ ve d 0 2 Re(δAZ ) . F1∼4 .3. Hioki / Nuclear Physics B 585 (2000) 3–27 C(DA : Bv ) = 2C(DVA : Av ). EVA } : Aγ . (∗) G1. C(G4 : Cv ) = C(EVA : Av ).16) As explained in the main text.     F1 ≡ C −(Aγ − AZ ve d 0 )δDγ + Aγ ve d 0 − AZ 1 + ve2 d 0 2 δDZ .A. we have   C {DV .2.3. DVA .     DA ≡ C BZ2 1 + ve2 d 0 2 − 2BZ ve d 0 Re(δBγ ) + 2BZ 1 + ve2 d 0 2 Re(δBZ ) .2.   DVA ≡ C −Aγ BZ ve d 0 + AZ BZ 1 + ve2 d 0 2 − BZ ve d 0 (δAγ )∗  + (Aγ − ve d 0 AZ )δBγ + BZ 1 + ve2 d 0 2 (δAZ )∗    − Aγ ve d 0 − AZ 1 + ve2 d 0 2 δBZ .B. D}Z .VA .4.17) for   DV ≡ C A2γ − 2Aγ AZ ve d 0 + A2Z 1 + ve2 d 0 2 + 2(Aγ − AZ ve d 0 ) Re(δAγ )    − 2 Aγ ve d 0 − AZ 1 + ve2 d 0 2 Re(δAZ ) .4 = P⊗ G1.2.. EVA } : AZ = ∓(1 ± ve )d 0 C {DV . θ ).A.  EVA ≡ C Aγ BZ d 0 − 2AZ BZ ve d 0 2 + BZ d 0 (δAγ )∗ + AZ d 0 δBγ   − 2BZ ve d 0 2 (δAZ )∗ + Aγ d 0 − 2AZ ve d 0 2 δBZ . C(G3 : Cv ) = C(DVA : Av ).B.1. DVA . (∗) (∗) (∗) (∗) Finally we present here formulas for DV . Pe− = Pe+ = ±1.1. f (∗) f (∗) In this case all we can determine (for the production form factors) are the following four combinations  Re δ{A. C(F1 : Dv ) = −C(DV : Av )/2.4 − P⊕ G2. EV .VA = P⊗ DV . i. Z.4 − P⊕ F2.B.3.A. 25 (A. When P⊕ = ±P⊗ .4. C(F4 : Dv ) = −C(EVA : Av ).D}Z (x.    F3 ≡ C BZ ve d 0 δDγ − BZ 1 + ve2 d 0 2 δDZ . C(G1 : Cv ) = C(DV : Av )/2.C.C.4 = P⊗ F1. (A. B. EV .VA − P⊕ EV .VA − P⊕ DV . Grzadkowski.VA . D}γ ∓ (1 ± ve )d 0 δ{A. they are not always independent of each other. θ ) = ∓(1 ± ve )d 0 F{A.e. C(EVA : Bv ) = C(EV : Av )/2.VA = P⊗ EV .    F2 ≡ C −AZ d 0 δDγ − Aγ d 0 − 2AZ ve d 0 2 δDZ .3.A. (∗) F1. EV(∗).D}γ (x.A. C.   EA ≡ 2C −BZ2 ve d 0 2 + BZ d 0 Re(δBγ ) − 2BZ ve d 0 2 Re(δBZ ) .A. C(DVA : Bv ) = C(DV : Av )/2. G1∼4 for completeness: (∗) DV .VA. B. C. C(F2 : Dv ) = −C(EV : Av )/2. EV .3 .VA . C(F3 : Dv ) = −C(DVA : Av ).A. C(EA : Bv ) = 2C(EVA : Av ). As a consequence of the above relations one gets F{A.3. .

Phys. C 72 (1996) 461. B 316 (1993) 137. . Grzadkowski. E. Kim. D 54 (1996) 4326. B 429 (1994) 255. Phys. Ladinsky. Yu. Rev. J. Int. Phys. Phys. Bigi. Jezabek. [3] M. [11] C. Ma. Grzadkowski. G. hep-ph/9803426. B 343 (1995) 333. J. Hioki / Nuclear Physics B 585 (2000) 3–27   F4 ≡ C −BZ d 0 δDγ + 2BZ ve d 0 2 δDZ . [10] M. J. Rindani. XXIV (1982) 203.A. Gajdosik. Z. Bernreuther. Poulose. Bernreuther.18) with C ≡ 1/(4 sin θW ). D 59 (1999) 077503. A 14 (1999) 1261. Rev. D 45 (1992) 2405. [26] M. Hioki. Rev. Rev. Lett. Nucl. Grzadkowski. Chang. D 50 (1994) 4372. Chang.H. Abachi et al. G. T.. hepph/9802352. hep-ph/9710358. Phys. Z. Phys. 74 (1995) 2632.D. H. Phys. 2 References [1] CDF Collaboration. Z. B 279 (1992) 389.. hep-ph/0001322. Rev. Bigi. Schröder. Krasemann. Phys.S. Phys. B 305 (1993) 384. S. G. A 2 (1987) 319. Phys. Phys. B 268 (1991) 424. Young. Overmann. W. hep-ph/9604301. Lett. Yuan. hep-ph/9712357.D. B 349 (1995) 379. 73 (1994) 225. Christova. [15] D.26 B.L. Erratum. [27] B. Bernreuther. Int. [16] G. hep-ph/9703463. Cuypers. D 56 (1997) 5907. Rev. Rev. C. Lett.B. [20] D.    G2 ≡ C AZ d 0 δCγ + Aγ d 0 − 2AZ ve d 0 2 δCZ .    G3 ≡ C −BZ ve d 0 δCγ + BZ 1 + ve2 d 0 2 δCZ . C 61 (1994) 599. T. Phys. Dokshitser. [29] B. Nachtmann. Phys. [30] I. L. Phys. Phys. [4] D. Lett. Lett. W. Rev. S. T. Phys. D 58 (1998) 035002. [24] P. Gavela et al. Valencia. Rev. Rev. Phillips. Phys.S. Lett. Nagano. Hioki. Phys. [13] W. Phys. Hioki. Rev. Phys. B 391 (1997) 172. V. Z. Donoghue. [7] L.-Y. W. Bartl.H.Y. T. Brzezi´nski. Y.M. B. Phys. Rev. Phys. Nucl. Zerwas. S. Rev. Z.-Y. Baek. Lett. Phys.-Y. [5] B.R. Valencia.-L. Bernreuther. hep-ph/9410357. B. [12] W. Phys. [32] I. Phys. Gunion. hep-ph/9509299. Phys. C. Lett. Kühn. Rev. B. Nelson.-P. Rindani. [17] W. F. Acta Phys. Z.M. [31] J. Choi. Phys. [22] W. hep-ph/9608306. B 484 (1997) 17. Rev. Phys. Arens. Z. [19] B. S. B 297 (1992) 318. P. Phys. I. [23] F. C 7 (1981) 127. Grzadkowski. [2] D0 Collaboration. [6] T. D 56 (1997) 6835. Lett. P. Austr. D 45 (1992) 124. Phys. Grzadkowski. D 39 (1989) 1870. Phys.A. 74 (1995) 2626. Nucl. J. Lett. Phillips.. Grzadkowski. Keung. Mod. Keung. B 408 (1993) 286. Szafranski. Mod. Holstein. Phys. Rev. B 181 (1986) 157. O. Majerotto. D 48 (1993) 3225. B. [21] G. B 383 (1996) 212. Lett.F. Phys. Lett. Grzadkowski. Erratum. J.   G4 ≡ C BZ d 0 δCγ − 2BZ ve d 0 2 δCZ (A. Pham. Lett.F.-P. Lett. HEPHY-PUB-686.F. D 50 (1994) 2966. Kane. hepph/9511256. I. A. Phys. hep-ph/9704312. Hioki. [25] J. T. D 58 (1998) 074007. [8] J.     G1 ≡ C (Aγ − AZ ve d 0 )δCγ − Aγ ve d 0 − AZ 1 + ve2 d 0 2 δCZ . Phys. Z. Cruz. Keung. hep-ph/9409243. [28] A. Soni. Lett. [14] R.A. Rev. Phys. Rev. Kühn. Lett. D 41 (1990) 2805. Phys. C. [9] J. Donoghue. W. Yang. Sehgal. Sumino. 58 (1987) 451. Rev. B 289 (1992) 440.P. Khoze. Suppl. Atwood. hep-ph/9301259. M. Phys. Yuan. Grzadkowski. Abe et al. D 49 (1994) 4415. hep-ph/9606356. [18] B.N. Ladinsky. Schröder. 60 (1988) 243.

Hioki. http://www-jlc. B 512 (1998) 19.kek. [40] S. Phys. Grzadkowski. J. Pilaftsis. [38] B. hep-ph/9501339. A. Z. [49] S. Phys. Phys. O. Le Diberder. hep-ph/0002006. Z. Phys. B 388 (1992) 53. B. Theor. Gunion. Z. D 61 (2000) 014013. Rev. D 41 (1990) 3343. Grzadkowski. Rindani. C 9 (1999) 229. [44] S. Phys. [51] B. Phys. 49 (1973) 1656. [39] Y. Tsai. Phys. Phys.F. P.P. A 9 (1994) 1097. Pizzio. hep-ph/ 9805318.jp/subg/cdc/lib/DOC/Seoul/fujiik-paper. [36] M. [34] R. hep-ph/9911505. Accomando. Grzadkowski. [41] B. Shirafuji. M. Diehl. Phys. Rev. Lett. hep-ph/0003294. Mod. He. Lett. B 350 (1995) 218. 77 (1996) 5172. Phys. C 62 (1994) 397. Z. [42] B. Schröder.Y. hep-ph/9901438.S. [45] W. J. Gunion. W. A 9 (1994) 5849. Phys. Rougé. Eur. F. Moretti. X-G. Davier. Hioki. Int. Moretti. hep-ph/9706201. Rev. Phys. Phys. B 319 (1993) 526. Nucl. hep/ph-9605326. Overmann. Grzadkowski. B 287 (1992) 237. Duflot. [47] S. L. B 476 (2000) 87. Nachtmann. T. Grzadkowski.-Y. Phys. Ballestrero. [35] J. Lett. T. Phys. Hioki / Nuclear Physics B 585 (2000) 3–27 27 [33] M. B 306 (1993) 411. J. Gunion. Lett. J. O. Z. [50] K. [48] E. [37] M. Kawasaki.B. Z. . A. Keung. Nowakowski. Rev. PRL-TH-00/1. [46] B. Grzadkowski.F. Grzadkowski. C 75 (1997) 465. Hioki. hep-ph/9611432. Nucl. Tsai. Lett. Mod.F.ps. Bernreuther. Prog. Int. Kauffman. Grzadkowski. A. D 4 (1971) 2821. S. Phys. Lett.gz. Fujii. Nachtmann. Phys. J. hep-ph/9301273. Erratum. [43] B.

Most notable among them. different from the standard νµ → ντ or νµ → νsterile interpretations. Italy b Dipartimento di Fisica dell’Università di Pisa and INFN.3 2 1 3 ( 0 0 ) 0 0 3 4 8 . the most natural candidate to propagate in some large extra dimension is the right handed neutrino. Paolo Creminelli a . 14. of ∗ Corresponding author.60. Always insisting on the graviton case. All rights reserved. Constraints from nucleosynthesis and supernova 1987a are discussed.V. in fact. is the possible disentanglement of the Planck scale from the scale where gravity becomes strong [1–4]. this has been and is being discussed both in the case of particle collisions at high energy [5–9] and of tests of classical gravity [10–13]. PACS: 11.5 . Alessandro Strumia b. we attempt to give a comprehensive description of the phenomenology of neutrino oscillations. Piazza dei Cavalieri 7 and INFN.∗ a Scuola Normale Superiore. Italy Received 6 March 2000.60.Pq. 14.elsevier. All rights reserved. With the relevant parameters at the boundary of the allowed region. I-56126 Pisa. 97. The constraints from SN1987a indicates a maximum radius of any extra dimension of about 1 Å.nl/locate/npe Neutrino oscillations and large extra dimensions Riccardo Barbieri a .60. accepted 25 May 2000 Abstract Assuming that right-handed neutrinos exist and propagate in some large extra dimensions. Introduction The hypothesis that Standard Model (SM) singlets propagate in some extra dimensions with relative large radii leads to striking consequences. PII: S 0 5 5 0 . After the graviton.unipi. Interestingly enough. may not be impossible. one can start asking if it will ever be possible to make any experimental observation related to these phenomena. I-56127 Pisa. In spite of this last remark. the smallness of the neutrino masses. It seems possible.it 0550-3213/00/$ – see front matter  2000 Elsevier Science B. A few alternative explanations of the atmospheric neutrino anomaly emerge.10. E-mail: astrumia@difi. the graviton.St.  2000 Elsevier Science B. although not easy.Nuclear Physics B 585 (2000) 28–44 www.Kk.V. if applied to the obvious candidate.Bw 1. such observations. that more surprises from the intense theoretical activity on this and related subjects have yet to come.

one per generation. . which. (2) after compactification of R theP5th dimension with suitable boundary conditions. λ can be made diagonal without loss of generality at the only price of introducing the usual unitary matrix V which describes flavour changes in the charged current neutrino interactions [17. ψjc ). the most significant cnstraints. ψj is one of the two independent Weyl spinors that compose Ψj = (ψj . would make impossible any observation of the oscillation phenomena discussed in this paper. Such constraint indicates an upper limit on the largest radius of any extra dimension close to 1 Å. τ.c. Barbieri et al. could in fact be a manifestation of this very hypothesis [14–18]. As manifest from (1). which we still consider as the relatively more likely description of nature at small distances. as discussed in Section 9. [19]. y). / Nuclear Physics B 585 (2000) 28–44 29 Dirac type. In Section 6 we show how the phenomenology of neutrino oscillations gets modified.) The purpose of this paper is to make a first tentative exploration of the related experimental consequences. In Section 5 we describe the oscillation amplitude of an interacting neutrino with its Kaluza–Klein tower. A strong constraint appears to come from the consideration of the neutrino luminosity of the supernova 1987a. . the action (1) becomes S = d 4 x 3i=1 Li . alternative to supersymmetric unification. including the flavour aspects. e. µ. 4}. The plan of the paper is as follows. if taken at face value. Redefining the νi in terms of the neutrino flavour eigenstates να(f ) = Vαi νi . nontrivial study. . The possible connection with gravity is briefly outlined in Section 3. the action which defines our framework involves 5-dimensional massless fermions Ψi (xµ . A specific suggestion along these lines has already been made [20]. In Section 2 we define the framework. 2. The relevant action is Z Z     4 A ¯ ∂ νi + νi λij ψj (xµ . with . 0)h + h. For an example see. Our conclusions are summarized in Section 10. (Alternatively. ‘ordinary’ neutrino masses can be obtained from extra-dimensional models.g. α = e.18]. Possible signatures in on-going or future experiments are summarized in Section 7. In Section 4 we diagonalize the neutrino mass matrix for arbitrary values of the compactification radius in presence of matter effects. . h is a normal Higgs doublet in four dimensions and λ is a matrix of Yukawa couplings with dimension of (mass)−1/2 . The framework defined Effectively. interacting on our brane with the standard left-handed neutrinos νi in a way that conserves lepton number. We view this work as a contribution to the present discussion about possibilities of physics beyond the SM.R. (1) S = d x dy Ψi ΓA i∂ Ψi + d 4 x ν¯ i i/ where A = {0. . without giving rise to a peculiar phenomenology. In Section 8 we show that big-bang nucleosynthesis is not manifestly inconsistent with the outlined phenomenology.. the ‘right-handed neutrinos’.

the relevant Hamiltonian is (one for any index i) H(i) = 1 Mi MiT 2Eν (5) (i) (i) . . νnL i m i  0 and Mi =   0 . (f ) (f ) Postponing the explicit calculation of λ2n and U0n . The elements U0n give the diagonalized by a matrix U (i) as U (i) H(i) U (i) T = Hdiag composition of the νi state in terms of the mass eigenstates corresponding to the 2 (i) eigenvalues λ(i)2 n /2Eν R of H . (6a) Aαβ (t) = i=1. t) = ∞ X   (i)2 2 U0n exp iλ(i)2 n t/2Eν R . the Dirac nature of the mass matrix in (3) is made explicit by an appropriate field redefinition. νR ν √ nR 2 mi 0 2/R .  NRi = √ 2 mi 1/R 0 . As shown in Appendix A. Barbieri et al.. for 1/R  mi . all of the Kaluza–Klein (KK) states ψin . Quite clearly. (6b) n=0 Note the formal identity of (6a) to the standard oscillation amplitude with the replacement Ai (t) → exp(im2i t/2Eν ). ..30 R. is. the amplitude Aαβ (t) = hνβ |να (t)i for finding at any time t a neutrino with flavour β. Otherwise. leading to the mass Lagrangian Lmass = 3 X T NLi Mi NRi + h. ψi−n with n > 1 in (3) decouple and . Within this framework. + R (3) n=1 the eigenvalues of the Yukawa matrix λ.  .c. born at t = 0 with flavour α. but only if 1/R is not too much bigger than all the mi ... R is the only extra parameter that will be introduced in the neutrino phenomenology other than the 3 masses mi and the mixing matrix V . i n > 1. . using (2). (4) To describe the evolution of a neutrino state of energy Eν . / Nuclear Physics B 585 (2000) 28–44 Li = ν¯i i/ ∂ νi + +∞ X  n  ∂ ψin + ψ¯ ic n i/ ∂ ψic n + mi νi ψin ψ¯ i i/ n=−∞ +∞ X  n n cn ψi ψi − ψi−n ψic −n + h.2. i=1 with  NLi =  ν .. ··· ···  ···.c. . X ∗ Vαi Vβi Ai (t). the neutrino phenomenology from (6) can be significantly different from the standard one. .3 where Ai (t) = A(νi → νi .

the asymmetry of the different radii.e. we can ignore this complication in the present work. with Ri>1  R1 .R. To be precise. we are in fact lead to consider several extra dimensions of different radii Ri . 1/R should not exceed 1 eV or so. let us discuss the possible connection with gravity [20]. Hereafter R is set to unity unless explicitly reintroduced. they satisfy the eigenvalue equation (see Appendix A) λ2 − µ − πλξ 2 cot πλ = 0. (9) where ξ ≡ mR and the index i is left understood. (8) Mi Mi + diag Hmatter = 2Eν R2 ± + where µi /R 2 = 2Eν2 (n± i − 1) depends on the refraction indices ni for νi (ni ) and ν¯ i − (ni ) in the relevant medium. R & 0. Diagonalization of the neutrino Hamiltonian Making reference for the details to appendix Appendix A. so that (3) is only an approximation where we neglect the heavier KK excitations in the extra dimensions with small radii Ri>1 [20].22]. we make the simplifying assumption that the graviton propagates in the same extra-dimensional space as the right handed neutrinos. for the heaviest neutrino mass.. 4.1 eV δ+2 26−15δ 1 Ri ' 10 and V ' mmax eVδ i mmax ' v (7) which shows. To maintain a simple connection with gravity. . . for any δ > 2. if we want to influence the atmospheric or the solar neutrino anomalies. the factorization of flavour as in Eqs. mmax ' v/(Mfδ V )1/2 .1 eV 3 MPl 10 TeV   Y 0. / Nuclear Physics B 585 (2000) 28–44 31 one remains with the standard Dirac masses mi νi νRi only. . 0. extended to include matter effects [21.1 µm. Furthermore the Planck mass MPl is related to the fundamental scale of the theory Mf and to the volume V of the δ compactified extra dimensions as MPl = Mf (Mfδ V )1/2 . we describe in this section the diagonalization of the Hamiltonian (5). Barbieri et al. At the same time the matrix elements U0n are given by . In this case Mf Mf ' 0. as explained. Connection with gravity Before going further. In turn. It is    1 µi (i) T . . 0. 3. essential at least to discuss nucleosynthesis. and.1 µm. i. Although not necessary. Since we will only study flavour mixing between µ and τ neutrinos in media that do not distinguish µ from τ flavour. if R1 & 0. (6) is no longer exact in presence of matter effects. 2 (i) Still calling λ(i)2 n the eigenvalues of 2Eν R Hmatter . as it follows from a simple generalization of (1) to δ extra dimensions [16].

+ (λn − µ/λn )2 /ξ 2 (10) Both (9) and (10) extend the results of [15] for µ = 0. so that the corresponding λ is imaginary. with small mixing. for large negative µ. the corresponding eigenstate is an almost pure ν. For ξ  1 the ν-state is spread over a larger √ 2 ≈ number of levels. (12) . 1/R  m. µ < 0 case The eigenvalues are as before. whose eigenvector tends to coincide with the ν-state up to mixings of order ξ/n. of order ξ . depends only on the combination ξ 2 /R = m2 R rather than on m and R independently. when the standard neutrino mixes with about πξ 2 KK states. 2 ≈ 2/(π 2 ξ 2 + n2 /ξ 2 ) we have U0n Z∞ A≈ 2 −n dn U0n e 2 (−iL/2E R 2 ) ν 0 where z = −i L 2Eν  πξ 2 R √  = ez 1 − erf z . This means that. When µ varies with the medium density or temperature and crosses a positive integer.32 R. one λ2 eigenvalue can be negative. For large ξ . Using also. 2 . mostly those with λ ≈ µ ± πξ 2 . for the component of the 2 ' 1 − πξ 2 /2√−µ. Matter effects are neglected for the time being. ν-state on the special eigenstate. Oscillation amplitudes for ξ & 1 For small ξ . By carefully expanding in ξ 2 / −µ. A qualitative description of λn and U0n is as follows. Note that. except for a special eigenvalue at λ2 = µ. depending on the sign of µ. For the probability Pνν = |A|2 one has at large L/Eν Z∞ Pνν (L → ∞) ≈ 4 dn U0n = 0 1 π 2ξ 2 . i. for negative µ. Barbieri et al. √ there is matter suppression. µ > 0 case For ξ → 0 the eigenvalues λ tend to the positive integers. from (10). / Nuclear Physics B 585 (2000) 28–44 2 U0n = 1 + π 2ξ 2 + µ/λ2n 2 . λn ≈ n and we can safely approximate the sum in (6b) with an integral. furthermore. one finds U00 5. Let us consider on the contrary the large ξ limit. each with a small mixing U0n 1/π 2 ξ 2 and eigenvalues close to semi-integers. In this case it is not difficult to obtain the neutrino oscillation amplitudes. Matter effects are not negligible if µ & (πξ 2 )2 and do not suppress oscillations when µ is large. In this limit. except for the special eigenvalue related to µ. (11) Note that the oscillation amplitude. obtained in the large ξ limit. only few of the KK states mix with the standard neutrino and. For µ < −ξ 2 the special eigenvalue λ20 becomes negative and tends to µ as µ  −ξ 2 .e.. a resonant MSW conversion can take place.

Barbieri et al.R. and on the Science Direct website: http://www. Oscillation probabilities (the colour version of the figure can be seen in the electronic publication of this paper.g.. 1/3 we also show the big ξ approximation (dotted lines). e. 3. on the Nuclear Physics Electronic website: http://www.nl/ locate/npe. For ξ = 1. 1. A good analytic approximation for Pνν at any L is Pνν s ".com).sciencedirect. (a) Pνν as function of L/Eν R 2 for ξ = 1/10. 1/3.elsevier. / Nuclear Physics B 585 (2000) 28–44 (a) 33 (b) Fig. 1. (b) Pνν as function of ξ for L  Eν R 2 (continuous line) and the small ξ and big ξ approximations (red dotted lines). 10 (continuous lines).

2 .

πξ L ≈ max .

.

1 − erf −i R 2E ν # .

2 .

.

1 . .

1 shows Pνν as function of L/Eν in units of R 2 for different values of ξ . atmospheric and LSND oscillations. Unconventional fits of atmospheric and solar neutrino anomalies In this section we look for a comprehensive description of the neutrino anomalies. Fig. The comparison of the numerical result with the analytic approximation at ξ = 1. They are characterized by the different ranges of 1/R. we consider a hierarchical neutrino spectrum. both atmospheric and solar. we have not been able to also account for the LSND result. A summary of the alternative possibilities that we have found is given in Table 1. in principle. with m3 > m2 > m1 . Different spectra are only possible if all ξ are small. as we now discuss. be associated with solar. 6. 1/3 shows the validity of the analytic expression even at moderate values of ξ . Matter effects on the oscillation amplitudes can be discussed along similar lines (see Appendix A). However. as in the standard phenomenology based on 3 neutrinos. π 2ξ 2 . . To keep things simple. the disappearance of the standard neutrino into the tower of KK states becomes increasingly important. When ξ gets sizable. The new scale 1/R of KK excitations allows in many different ways to have three different neutrino squared-mass splittings that could.

In absence of flavour mixing. as extracted from [26]. νKK with ξ 2 /R ≈ 0. An increasing portion of the atmospheric neutrinos starts oscillating into their KK towers at the expenses of the conventional νµ → ντ transition (case C in Table 1).tohoku.. (c) The latest SK data (848 days of exposure).e. Barbieri et al. These considerations are confirmed by an explicit fit of the data 1 .awa. i. ξ2  1 and neglecting matter effects A1 and A2 are accurately approximated by the standard phase Ai = exp(im2i L/2Eν ). To understand what happens when ξ3 increases. Atmospheric neutrinos When 1/R decreases from about 1 eV the standard picture of neutrino oscillation is progressively perturbed. 1/1m2ij  L/Eν  R 2 .01 eV νµ → ντ . available at the www address www. In Eqs. we can take A3 = 0. The situation in our case. (13) 3 45 Note that these formulae do not allow to interpret 1/R as the extra mass scale useful to account for the LSND result. On the other hand. (6) only A3 is modified from the conventional form. (6) with ξ1 . and for α 6= β Pαβ ' |Vα3Vβ3 |2 . with only one extra parameter R. using the unitarity of the V matrix. In particular we have used: (a) The prediction of atmospheric ν fluxes of [25]. is far more constrained than in the case where one adds one sterile neutrino with arbitrary mixing parameters.jp/ ˜etoh/atmnu/enu/index. In Eqs. From (6a) and above the SK threshold. for ξ3 > 1 and above the SK threshold for νµ disappearance. a fit may be possible for Ve3 ≈ 0 and 2 | ≈ 0. we have therefore 2 (14) Pαα ' 1 − |Vα3 |2 . using 30 bins of experimental . νKK or νµ → νKK Standard νµ → ντ 6. If ξ1. 1. shown in |Vµ3 1 The fit of SK atmospheric data is done as in [24]. corresponding to the various classes of events measured at SK as given by the Monte Carlo simulation of the SuperKamiokande detector.34 R.4. let us consider the case when ξ3 gets large.ac. which reduce to 1 at the L/Eν relevant to the atmospheric neutrinos.2  ξ3  1 approximate formulae for the transition probabilities in an intermediate L/Eν range. As shown in Fig.1. are 2π 2 7π 4 |Vα3 |2 ξ32 .5 and Pee ≈ 1. Pαα = 1 − Since the data suggest Pµµ ≈ 0. / Nuclear Physics B 585 (2000) 28–44 Table 1 List of the possible oscillation patterns ξ3 1/R (eV) A  1/3  3×10−3 ≈ 3 × 10−3  3×10−3 B ≈ 1/3 ≈ 10−1 Standard C  1/3 & 10−1 Standard Case Solar Atmospheric Only VO SAM νe → νKK Standard νµ → ντ . (b) The energy spectra of the parent atmospheric neutrinos. and for α 6= β Pαβ = |Vα3Vβ3 |2 ξ34 . this would certainly disagree with the SuperKamiokande (SK) results [23]. for ξ3 & 1 and for sufficiently large L/Eν the survival probability of ν3 drops below 10%. (d) The χ 2 function defined in [27].html.

Now ξ3 and 1/R. preliminary investigation. / Nuclear Physics B 585 (2000) 28–44 35 2 . This is case A in Table 1.5. 2. ξ 2 /R). 20. Fit of the SK atmospheric data in (Vµ3 ξ  1 and any R.e. In Fig. which can vary below about 10−2 eV since ξ > 1. data and including all various systematic uncertainties.R. in the intermediate region between A and C. 3a. the oscillation amplitudes depend on the combination ξ 2 /R.02 eV is selected by the shape of the angular distribution of the muon neutrino events. i. |Vµ3 |2 ) at fixed ξ3 = 1/2. whereas the one for large ξ does not because in this case it is difficult to compute neutrino propagation across few layers with different density. As mentioned in the previous section. no νµ → ντ oscillation is present in this case. since we set Ve3 = 0. The fit does not fix the value of 1/R. The best standard fit in terms of νµ → ντ gives min χ 2 ≈ 14. because at ξ = 1/2 the individual oscillation probability Pνν at large neutrinos oscillate into the KK states. for large ξ . the two minima of the χ 2 distribution are clear. 1 for colour version). . shown in Fig. or of the standard fit with ξ = 0. 2b. 3b we show the profile of the χ 2 again in the plane (ξ 2 /R. (see caption of Fig. The other minimum originates from the one encountered before at large ξ . these data could discriminate the non-standard L/Eν dependence of the oscillation probability from the standard one. 2b: ξ = 0. We can now ask what happens when ξ3 is decreased below 1. which becomes the only relevant parameter other than Vµ3 . Back to Fig. As a consequence a fit of the atmospheric data is possible with Vµ3 ≈ 1. If systematic errors are sufficiently low. 2a: Fig. We have just discussed the case Vµ3 ≈ 1. 2a. The fit at ξ = 1/2 includes earth matter effects. This should give an idea of what happens at various intermediate values of ξ . as well as the mixing angles. as illustrated in Fig.. independently influence the fit. We expect that the main features of solution A will remain unchanged. Unlike the case of very large ξ . m3 > 1/R. 3c. The value of ξ 2 /R around 0. In this case (B in Table 1) 1/R ranges around 10−2 eV. Barbieri et al. 3b. 25}. At this intermediate value of ξ3 about half of the interacting neutrinos oscillate into the KK states. Fig. The contour plot of the χ 2 in these parameters is shown in Fig. ‘Official SK fits’ employ unpublished data with finer energy and zenith-angle subdivisions of the neutrino-induced events. The contour lines correspond to χ 2 = {15. This completes our discussion of atmospheric neutrinos.

including only statistical uncertainties.004 eV and Vµ3 µ τ 23 23 6. 3c: standard fit (ν → ν : 1m2 = 3 × 10−3 eV2 . 3b: fit with ξ = 1/2 (ν → ν with ξ  1 (νµ → νKK .2. Solar neutrinos Given these alternative descriptions of the atmospheric neutrino anomaly. since they are the same in all rows. / Nuclear Physics B 585 (2000) 28–44 Fig. it is in fact simple to see that a standard description of solar neutrinos is possible in all cases with . Crosses denote experimental data. Each plot is drawn in the (cos ϑzenith . This being the case. (see caption of Fig.015 eV and Vµ3 µ KK : 2 = 1). 3c: as usual the most significant data are the shapes of the individual zenith-angle distributions. sin2 2θ = 1). Continuous lines denote fit predictions and gray bars denote the no-oscillation predictions. ξ 2 /R = 0. Best fits of the SK zenith-angle distribution of sub-GeV (first column) and multi-GeV (second column) νe . The systematic uncertainty in the overall number of neutrino events in each sample of events has been used to optimize the visual appearance of the standard fit in Fig. 3. number of events) plane. νµ data are plotted in blue. 1 for colour version). The arrows on the horizontal axes denote the direction of the scattered leptons. we should ask now how the solar neutrino deficit can be accounted for. νe data are plotted in red and only in the last row. We only required that ξ2 and ξ1 are both small.36 R. νµ events and of upward-going muons (third column). 3a: fit 2 = 1/2). So far m2 and m1 have not been fixed. ντ : ξ 2 /R = 0. Barbieri et al.

We give there.R. Continuous blue line: standard νµ → ντ fit. would be to indicate precise signatures of such solutions visible in appropriate neutrino experiments. so that a fit of SK atmospheric data requires ξ3 ∼ 2. For small values of 1/R there is however the possibility of a transition νe → νKK using the MSW effect. unlike what happens in the standard case. A few features of these plots might be relevant for an experimental discrimination of the various possibilities. Dashed green line: νµ → ντ . / Nuclear Physics B 585 (2000) 28–44 37 negligible interference of the KK towers. 4 is of interest. versus L/Eν . . The non-standard transition from unoscillated to oscillated atmospheric neutrinos requires a L/Eν -range longer than the standard one and even the one that would be produced by neutrino decay [28–31]. at clear variance with the shape of Pµµ in the standard νµ → ντ interpretation of the data. Dotted red line: νµ → νKK fit with intermediate ξ = 1/2. The Pµµ (a) and Pµτ (b) that give the best SK fits (see caption of Fig. or m2 ≈ 10−(4÷5) eV. 1 for colour version). which is compatible with the solar data [20]. 4. 3. Barbieri et al. small ξ ) takes place and suppresses the different components of the solar νe spectrum as possibly observed by the various solar neutrino experiments. 2. however. For example the on-going K2K experiment [32] should observe only 65% ÷ 85% of the events with respect to the no-oscillation case. Fig. a good fit of atmospheric data significantly constrains the outcome of νµ disappearance experiments. the probabilities Pµµ and Pµτ that correspond to the fits of the SK results shown in Fig.01. The crucial point. Special features of the proposed solutions Some alternative descriptions of the atmospheric neutrinos appear possible. νKK fit with large ξ . 7. the resonant MSW conversion mentioned there (µ positive. 1. It requires 1/R ≈ 3 × 10−3 eV and a mixing with the KK states determined by ξ2 ≈ 0. The absence of a first clear dip in the L/Eν -shape of Pµµ is a characteristic of the KK fits that we have discussed at intermediate and big ξ . Therefore. To this purpose Fig. When the parameter µ of Section 4 is specified for the electron neutrino and with the solar density profile.

01 mm on the scenario under study. At small L/Eν . 3. / Nuclear Physics B 585 (2000) 28–44 while the larger range 30% ÷ 90% is allowed by a standard fit of SK data. an analytic approximation for Pαα and for Pαβ is r .38 R. A related characteristic feature of the same νµ survival probability is a precocious disappearance of νµ (and appearance of ντ if Vτ 3 6= 0) at relatively low L/Eν . Hence K2K could put the constraint R < 0. The effect is not big since otherwise it would have made difficult the same SK fit or it could have been in conflict with the null result of CHORUS and NOMAD [33]. Barbieri et al.

2 .

ξ32 πL .

and for α 6= β Pαα ' 1 − 2. .

Two different production mechanisms have to be studied: (i) by incoherent scattering. the incoherent production is not dangerous provided a temperature T ∗ is assumed. Earth matter effects suppress the oscillations of a long-baseline beam of νµ with large energy Eν & 10 GeV. . On the contrary earth matter effects enhance the oscillations of a ν¯µ beam. (15) Pαβ ' 2π|Vα3Vβ3 | R Eν Note the linear dependence on L of Pαβ . The effects related to (15) are not without interest already for the downward going νµ in SK. 4.20]. (ii) by coherent oscillations. 2. A partial suppression of the ντ -events might occur in an appearance experiment trying to measure Pµτ at large L/Eν relative to the expectation in the case of the standard νµ → ντ interpretation of the data. Constraints from nucleosynthesis So far we have compared the phenomenology of neutrinos from extra dimensions with oscillation experiments. This argument does not apply in the present case. The SK collaboration [23] claims that the νµ → νsterile interpretation of the atmospheric neutrino anomaly is disfavoured because oscillations of the more energetic neutrinos should be suppressed by matter effects. the equilibrium rate of the crucial nucleosynthesis reactions at temperatures at about 1 MeV. at which the abundance of relic KK neutrinos is negligibly small. where there is a KK tower of sterile states. In short. 5. This agrees qualitatively with several statements in the literature [16. All these points deserve further quantitative investigation. 8.Vα3 R E  2 2 L 2 ξ3 . As well known. in view of the parameters shown in Fig. The case with coherent oscillations is more delicate since the evolution of an infinite number of different neutrino mass eigenstates is not easy to follow. rather than the quadratic one characteristic of the standard oscillation formula. especially in the case of a sizable ξ parameter. the danger for nucleosynthesis is that too many KK neutrino modes are produced before the time of nucleosynthesis. In this section we study the constraint arising from standard bigbang nucleosynthesis. although with a slower Eν dependence than in the usual case. T ∗ is in the GeV range for δ = 2 and can easily be much higher as δ increases.

Here we only check that standard big-bang nucleosynthesis is not grossly inconsistent with the picture described above. / Nuclear Physics B 585 (2000) 28–44 39 Fig. Eν is the neutrino energy. The opposite would have been a clear problem for nucleosynthesis since the typical temperatures are close to the MeV. Barbieri et al. Fig. and c is a numerical coefficient close to 0. As discussed in Section 4. a negative µ implies a MSW suppression of Γn (T ). one can estimate if any one of the mass eigenstates may reach equilibrium by comparing to the expansion rate H (T ) the effective interaction rate of the 2 Γ (T ). The µ parameter has a negative fixed sign both for neutrinos and antineutrinos.2 for µ and τ neutrinos. Values of Γn /H for ξ = 2 and 1/R = 2. Even though the energy density stored in any state is small because none gets even close to equilibrium. (16) where αem and GF are the fine structure and Fermi constants. The rise at small T is a manifestation of the T 3 -behavior of Γν /H . their total energy density might be significant. . Here Γ (T ) is the typical interaction rate of a standard n-th state Γn (T ) ≈ U0n ν ν 2 is given in (10) in terms of the matter potential µ. 5 shows the plot of Γn (T )/H (T ) versus T for ξ = 3 for a large number of mass eigenstates different from the special one.5 × 10−3 eV for n = 1. 5 is compatible with standard nucleosynthesis. which coincides with the interacting neutrino at large temperature. whereas the fall at large T is a 2 .R. In the T -dependent Hmatter (T ) . It is harder to say if the situation described in Fig. . 5. None of the states consequence of the MSW suppression of the effective mixing factor U0n is close to equilibrium at any temperature. A simple way to get an idea of what happens is the following. . Since the oscillation frequencies are large relative to the collision or the expansion rates at any temperature T close to nucleosynthesis. either original or generated by the neutrino evolution equations themselves [34] µ = −2c(Eν RGF )2 T 4 /αem . To answer this question requires solving the coupled system of infinite equations for the neutrino density matrix ρ. to be suitably averaged. 1 for colour version). 100 (see caption of Fig. . If neutrino with the medium and U0n one neglects any matter asymmetry in the primordial medium.

but do not directly affect the total neutrino P number N ≡ n Nn . A detailed and non trivial study could give significant constraints. the right-hand side of(19) is small at T . More precisely. An unambiguous problem. if present. the evolution equations for the neutrino densities Nn ≡ ρnn of all individual neutrino mass eigenstates. Constraints from supernova 1987a In this section we show that the SN1987a observation is likely to put a very severe constraint on R. 1 MeV and progressively increases. but not both. if taken at face value. Using this value of rapidly convert it into Nn (T ) = U0n ν νν n 0n 4 ≈ 1 − πξ 2 /√−µ valid at large µ. normalized to the equilibrium density. (18) d ln T H (T ) P 2 Nn is the number density of the interacting neutrino in the flavour where Nν ≡ n U0n basis. thus shortening the observed neutrino burst in an unacceptable way. Barbieri et al. In the present case where ν is a combination of νµ and ντ . Winv . neutrino propagation averages to zero the off-diagonal elements of ρ much faster than the typical interaction or expansion rates at T ∼ MeV. Setting to zero the offdiagonal elements ρnm . not compatible. d ln T H (T ) 0n (17) We have explicitly included annihilation processes and indicated with · · · additional terms due elastic scatterings. These terms have comparable rates and redistribute the total neutrino number N in its various components. d ln T H (T ) −µ(T ) (19) As for the individual rates Γn /H . Summing all the equations we find the evolution equation for the total neutrino density Γν (T ) dN =− (1 − Nν ). we demand that the energy loss rate in invisible channels. could not be avoided as usual by invoking a large lepton asymmetry which. unlike Γn /H . or large T . can only suppress ν or ν¯ oscillations. n 6= m.40 R. 9. give neutrino oscillations incompatible with standard big-bang nucleosynthesis. is less than . relevant to the atmospheric neutrino anomaly. with the phenomenology of neutrino oscillations as discussed in the previous sections. the numerical coefficient N is of order unity and has a large uncertainty. oscillations 2 (T ) so that N = P U 4 = P . we Nν (T ) and the approximation Pνν ≈ U00 estimate the right-hand side of (18) as Γν (T ) πξ 2 dN ≈− √ . are Γν (T ) 2 dNn =− U (T )(1 − Nn ) + · · · . and assuming that neutrinos are in kinetic equilibrium. it does not get to a maximum but rather flattens out to a horizontal asymptote at a level N (ξ 2 /R)/(10−2 eV). However. in the case under study. If neutrinos start with a normalcy distribution at some temperature T . This is because a large number of KK neutrinos would carry away from the supernova (SN) too much energy in invisible channels. Therefore we cannot conclude that values of ξ 2 /R ≈ 10−2 eV. / Nuclear Physics B 585 (2000) 28–44 eigenstate basis.

due to the small vacuum mixing angle between the nth KK state and the standard neutrino. about (RE)δ . As in the case of nucleosynthesis. the KK states interested in the resonant transitions have a mass mn ≈ n/R ≈ (V E)1/2 ≈ 104÷5 eV. the width of the nth resonance is smaller than the separation between two contiguous resonances. 10−8 gives (10−3 eV · R)(m2 /10−3 eV2 ) . for each Pn one can take an approximate formula describing the transition with initial and final densities respectively much larger and much smaller than the resonance density. Furthermore. with a number of states interested in the resonant transition of order nres ≈ R × 104÷5 eV. / Nuclear Physics B 585 (2000) 28–44 41 about 1019 erg/(gr · sec) for typical average conditions of a SN core with a density ρ ≈ 3×1014 gr/cm3 and a temperature T ≈ 30 MeV [35–39]. the increase in the number of KK states of energy less than E. that had we considered δ > 1 extra dimensions. (20) R < 0. As a consequence. MSW resonant oscillations will take place either in the neutrino or in the antineutrino channel. Requiring Winv /Wν . E ≈ 100 MeV is a typical neutrino energy and Γν ≈ G2F nN E 2 is the collision rate of a standard neutrino in the supernova core. approximately RE. This is so because in such a case.2 mm × 10−11(1−1/δ) 10−3 eV2 Let us now consider the effect of coherent oscillations of the active neutrino state associated with the mass m into its KK tower. The effect of these oscillations can be reliably estimated for small ξ = mR. θn ≈ ξ/n. where . one has to consider the production of KK neutrinos both by incoherent scatterings and by coherent oscillations. The rate of incoherent production of a single KK state is estimated as Γinc ≈ (m/E)2 Γν . in terms of the nucleon density nN . where m is the largest of the masses mi introduced in Section 2. Since the neutrino effective potential V in the core is of order 10 eV. Accounting for the number of KK states. Pνν ≈ Q n Pn [20]. Barbieri et al. the corresponding energy loss rate can be estimated as Winv ≈ (m/E)2 ER · Wν . the survival probability for a standard neutrino (or antineutrino) produced in the core can be computed as the product of the survival probabilities in each resonance crossed during a mean free path. 1 which is not incompatible with the values of m and R considered in the previous sections. Note.R. Depending on the sign of µ in the supernova core. all with comparable radii. however. where Wν ≈ 1027 erg/(gr · sec) is the energy loss rate produced by Γν if the standard neutrinos were not trapped. Pn ≈ e−πγn /2 . would have lead to a stronger constraint on R  −1/δ m2 .

V .

.

4ξ 2 γn ≈ 2 R E dV /dr .

Approximating V /(dV /dr) with the radius of the core. We do not even see how such a . γn is of order m2 /(10−3 eV2 ) and is approximately n-independent. which is a fortiori the ase for the overall probability Pνν due to the large number of resonances crossed. either escaping or trapped in the core. This is untenable no matter what the fate will be of the KK states produced in the oscillations. Hence the individual Pn are expected to deviate sensibly from unity.res is the adiabaticity parameter at the n-th resonance crossing. rcore ≈ 10 km. since in both cases the luminosity in standard neutrinos will be drastically reduced.

G.. making impossible any observation in neutrino oscillation phenomena. where right-handed neutrinos propagate. . Barbieri et al. Above this value no resonance is crossed anymore and the mixing angle of any KK state with the standard neutrinos becomes negligibly small. However ad hoc ways to evade it without affecting the signals discussed in Sections 6. This is so if there is a large extra-dimension that extends above 0. a maximum radius of any extra dimension is implied by these considerations close to 1 Å. Constraints arise in principle on this phenomenology both from standard big-bang nucleosynthesis and from SN1987a. ERBFMRX-CT96-0090. We thank A. we know of no observation. e. As we have discussed. particularly constraining appears to be the consideration of the neutrino luminosity from SN1987a. This bound is likely to be valid.D. of supersymmetric unification. Acknowledgements Work supported in part by the E. the existence is being considered of some large extra dimensions where SM singlets could propagate. In this paper we have shown that neutrinos offer a concrete possibility of testing the existence of large extra dimensions. even though we cannot perform a clear calculation in all the relevant range of the parameters. This bound. or ξ & 1. Rattazzi for useful discussions.Yu. alternative to those considered in high energy collisions or in gravity measurements. even though a precise calculation in this case is not easy to make due to the overlap of the different resonances. together with (20). In our view. 2 We thank G. Conclusions In the discussion about possible patterns of physics beyond the SM. A precise computation should also take into account Pauli blocking effects and concentrate on the cross section for reactions accompanied by a significant energy exchange between the neutrino and the matter in the supernova. We think that neutrino oscillation experiments can fully explore this range of distances. particularly acute. which undergo the resonant transitions in the supernova core. therefore. Hence. On the other hand. The problem of finding possible signals of large extra dimensions is. it is anyhow worth pursuing a direct comparison with neutrino oscillation data. direct or indirect. would seem to indicate that the largest of the radii cannot be bigger than about 1 Å independently from the number of dimensions. / Nuclear Physics B 585 (2000) 28–44 conclusion could be avoided by having 1/R approaching m.g. 10. The possibility of lowering the scale where gravity becomes strong is a striking consequence of this hypothesis. Ross and A. Dolgov and R. like suitable decays of the heavier KK neutrinos in the 100 keV range. the hierarchy problem is generally quoted as its main motivation. that supports this picture. under TMR contract No.42 R.1 µm. 2 We therefore believe that this sets an upper limit on R close to 10−4÷5 eV−1 . Smirnov for discussions about these issues. If taken at face value. 7 may exist. unlike the case.C.

(9) admits one λ2 < 0 with λ purely imaginary. using (9). the effective L/Eν above which neutrino oscillations are significant is reduced by a factor ρ 1/2 with respect to the µ = 0 value given by Eq. ρ ≡ µ/(πξ 2 )2  1. 2 λn (A. one has to separately include the contribution of the special eigenstate with 2 λ < 0. where the µ term parameterizes matter effects. k=1 In the limit kmax → ∞ we obtain the eigenvalue equation (9).1) In this way ψic 0 and the linear combinations orthogonal to νRn and νLn decouple and can be forgotten. (12) remains valid. Using the normalization condition P i ei ei = 1 we arrive at ∞ X 1 k2 1 2 = 1 + 2ξ = 2 (k 2 − λ2n )2 U0n e02 (λn ) k=1    π 2ξ 2 1 = 2 + π 2 ξ 2 1 + cot2 πλn − cot πλn . the eigenvalue equation can be written in the closed form   det X − λ2 1 = " X00 − λ 2  − kX max k=1 " = ξ + µ − λ − 2ξ 2 2 Xk0 X0k Xkk − λ2 2 kX max k=1 #k max Y Xkk − λ2  k=1 λ2 k 2 − λ2 #k max Y  k 2 − λ2 . Barbieri et al. when U00 µ > 0. . If µ < 0. (11). On the contrary. Calling λ2 the eigenvalues of X. when µ < 0 the squared mass matrix may have negative eigenvalues: for µ < ξ 2 . It is not difficult to obtain simple expressions that extend the large-ξ oscillation amplitudes (11) and (12) to include matter effects. when the matter potential is very large.2) which. Its diagonal elements are Xkk = k 2 for k > 1 and X00 = (2kmax + 1)ξ 2 + µ. Moreover. The resulting effective Hamiltonian (8) for the first kmax KK states is Heff = X/2Eν R 2 where the dimensionless matrix X has an ‘almost diagonal’ form: keeping only the √ first kmax KK excitations its only non-vanishing off-diagonal elements are Xk0 = X0k = k 2ξ (k > 1). 2  1 νLn = √ ψic n − ψic −n for n > 1 2 and νR = ψi0 . gives (10). For any eigenvectors ei (λ) the secular √ equation gives a relation between the various elements and the first one: ek /e0 = −k 2ξ/(k 2 − λ2 ). Explicit diagonalization of the Hamiltonian From (3) we arrive to the Dirac mass matrix (4) with the following definitions:  1 νRn = √ ψin + ψi−n . Eq.R. When µ  −(πξ 2 )2 this state has a small KK component and Pνν (L → ∞) ≈ 4 ≈ 1 − πξ 2 /√−µ so that matter effects suppress oscillations. in fact. / Nuclear Physics B 585 (2000) 28–44 43 Appendix A. (A. As already remarked.

Dimopoulos. B 458 (1999) 274.E.). Nucl. G. H. Phys. Arkani-Hamed. Y. Phys.D.G. hep-ph/9907520. N. NSF proposal.H. / Nuclear Physics B 585 (2000) 28–44 References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] I.. Dudas. P. hep-ph/9904245. M. Peskin. B 544 (1999) 3. hep-ph/9803315. astroph/9810121. hep-ph/9901350. J. Agrawal. Rev. Dvali. V. Talk given at the SUSY 1998 Conference. B 548 (1999) 180. JHEP 04 (1999) 026. China. hep-ex/9807003. Nucl. 81 (1998) 1562. S. Price. A. 1996. B 246 (1990) 377. Phys. DiLella. Lett. J. N. hep-ph/9912408.. Nucl. Rev. Lett. Nucl. Michelson (Ed. hep-ex/9908049. Mirabelli. 49 (1999) 163. World Scientific. Gaisser. Grossman. hep-ph/9904262. Neubert. S. E.R. T. B 436 (1998) 263. Phys. Proc. hep-ph/9811428. Lipari. hep-ph/9907421. Kapitulnik. Lett. Nucl. Oyama. Fogli. Z. Nandi. E. Phys. Nucl. Sigl. Rev. S. hep-ph/9811337. Marrone. Rev. B 358 (1991) 435. Lett. J.Yu. Dvali. Nötzold.D. G. Lykken. G. D 53 (1996) 1314. 82 (1999) 2644. S. B 467 (1999) 185. Antoniadis. Lett. B 466 (1999) 115. J. M. Scioscia. Kainulainen.. E. J. Perez-Lorenzana. A. as summarized by L. R. G.F. Price. L. J. Part. Rev. Lusignoli. 1997. G. S. J. Kakushadze. Singapore. 42 (1986) 913.. S. Lykken. hep-ex/9812014. B 307 (1988) 924. D 60 (1999) 013003.C. hep-ph/9910474. Lett. Phys. Dvali. Rev. Phys.K. Phys.W. Lett. hep-ex/9803014. G. . J. Giudice. B 557 (1999) 25.C. SuperKamiokande Collaboration. hep-ph/9902267. De Rújula et al. hep-ph/9811448. Phys. V.A.T. Wells.N. NOMAD Collaboration. Long. Phys. Talk given by Masato Shiozawa for the SK Collaboration at the 35th Rencontres de Moriond. Rattazzi. Lett. Phys. D 17 (1978) 2369. T. G. A. Tye. G. Price et al. Phys. Mohapatra. Int. Phys. Phys. Raffelt. R. B 462 (1999) 109. R. P. J. Dimopoulos. B 433 (1998) 9. hep-ph/9904211. 1988. Mikheyev. D 59 (1999) 105006. I. Nucl. Barger. Phys. T.J. B 429 (1998) 263. 82 (1999) 2236.L. hep-ex/9803006. astro-ph/9209005. T. 1 (1993) 165. Phys. Lipari. Phys. D 59 (1999) 117303. Shrock. G. D 54 (1996) 3693.H. Phys.P. A. Y. S. Lett. A. Astropart. Strumia. A. hep-ph/0001124. Mohapatra. Lisi. Phys. G. Rev. Astrophys. Lett. Lett. Han. Dimopoulos. Arkani-Hamed. K. hepph/9804398. Phys.N. 254 (1992) 121. S.. Phys. Phys. R. Gherghetta. Kenny. hep-ph/9811291. M. Rev. hep-th/9603133. A. Barger et al. Wolfenstein. Arkani-Hamed et al. G.. Weiler. Rev. N. Dienes. Rev. R. epex/9912010. Annu. D 56 (1997) 1704. Astron. NSF proposal. D. July 1998. Shiu.L. Sci. Phys. Phys. Antoniadis. CHORUS Collaboration. Raffelt. H. Pakvasa.C. on Experimental Gravitational Physics. J.Yu. Peltoniemi. T. Raffelt. Phys. Chan. Lett. in: P. K2K Collaboration. Rev. Maalampi. hep-ph/9811350. 82 (1999) 2640. Stanev. Phys. M.D. Symp. Rev. Guangzhou. A. Perelstein. V. J. hep-ph/9805217. Barbieri et al. hep-ph/9808205. J. Nucl. K. Zhang. Tye.44 R. Smirnov. hep-th/9809147. T.T. B 563 (1999) 63. Nunokawa et al. Learned. Perez-Lorenzana. Sov.C. Peltoniemi. Smirnov. Fogli et al. Phys. G.

Switzerland Received 18 August 1999. Abada ∗ .elsevier.ch (M. Imposing the constraints. Losada).V. 14. All rights reserved.cern. Colombia. Losada 1 Theory Division. Furthermore.3 2 1 3 ( 0 0 ) 0 0 3 9 7 . France. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.cern. All rights reserved. accepted 19 June 2000 Abstract We consider a general symmetric (3 × 3) mass matrix for three generations of neutrinos. oscillations of neutrinos imply masses for neutrinos and mixing angles. 91405 Orsay. CERN.ch (A. Université de Paris XI. ∗ Corresponding author.  2000 Elsevier Science B. The present experimental results on neutrinos from laboratories.τ ) R-parity-violating couplings or constrain combinations of products of these couplings. deficits of the solar electron neutrino and atmospheric muon neutrino fluxes compared to their theoretical predictions are strong evidence in favour of neutrino oscillations as an explanation of these deficits. Abada).µ.V. as is the case in the quark sector.nl/locate/npe Constraints on a general 3-generation neutrino mass matrix from neutrino data: application to the MSSM with R-parity violation A. Introduction The current experimental evidence strongly suggests the existence of a non-trivial structure of the mass matrix for 3 generations of neutrinos. cosmology and astrophysics are implemented to either put bounds on trilinear (λij k .Cq 1. E-mail addresses: abada@mail.70. Indeed. which relate the flavour basis to the mass eigenstate basis. λ0ij k ) and bilinear (µe. 1 On leave of absence from the Universidad Antonio Nariño.+g. on the mass squared differences and on the mixing angles. PACS: 13. We apply our results to Majorana left-handed neutrino masses generated at tree level and through fermion–sfermion loop diagrams in the MSSM with R-parity violation. On leave of absence from the Laboratoire de Physique Théorique. losada@mail. CH-1211 Geneva 23. PII: S 0 5 5 0 . we identify the ranges of allowed inputs for the 6 matrix elements. revised 6 June 2000.15.Bh.Nuclear Physics B 585 (2000) 45–78 www.7 . Santa Fe de Bogotá.30. bât 210. from the atmospheric neutrino and solar neutrino anomalies as well as from the CHOOZ experiment. M. 95.

1m2atm ∈ [10−3. M. namely 1m2sun  1m2atm  1m2LSND . νµ . which is applicable when the neutrinos are Majorana particles. 3 the deduced pattern of neutrino masses and mixing requires two different neutrino mass squared differences and three mixing angles. requires the presence of new physics. arises from neutrinoless double beta decay. see Refs.46 A. In addition. or other more exotic possibilities such as the violation of Lorentz invariance [13] or the equivalence principle [14].16] that neutrino decays or modification of special or general relativity do not fit the data. which calls for a fourth neutrino species or a sterile neutrino. . can involve flavourviolating interactions. other interesting bounds on neutrino masses arise from cosmological or astrophysical sources. ντ ).6]. Soudan [3]. it has been shown [15. In supersymmetric (SUSY) extensions. MACRO [11] and LSND [12]). SuperKamiokande [10]. [18–28] for more details. the atmospheric neutrino anomaly and the LSND results suggest three very different values of neutrino mass squared differences. which have to be added to those of other atmospheric neutrino experiments (IMB [2]. If the neutrino has a Majorana mass. if massive. 10−2 ] eV2 and 1m2LSND ∈ [0. In a general manner. However. hierarchical. Our approach differs from most of the analyses in that we do not specify the model initially. For reviews. which could accommodate the experimental neutrino data. the model must violate lepton number conservation. the strongest constraint. The results of this work can be applied to any class of model with three neutrino generations. SAGE [8]. Even barring the presence of a sterile neutrino. We also present an application of our results of the specific Minimal Supersymmetric Standard Model (MSSM ) with R-parity violation. The generalization of the MSSM. We will also try to elucidate if any of the different possible neutrino spectra (degenerate neutrinos. concerning models for neutrino masses and neutrino properties. The MiniBooNE at FNAL or MINOS longbaseline experiments could provide the answer. we will use the current relevant information on the mixing angles and mass squared differences to constrain the neutrino mass matrix elements in the flavour basis (νe . We will also analyze our results in view of the two latter bounds. Gallex [7]. 10] eV2 . such as having a neutrino component of hot dark matter (HDM). 10−4 eV2 . The neutrino is massless in the Standard Model and. Kamiokande [9]. Consequently this will constrain the parameters of the involved model. with 1m2sun . which can accommodate the data. low energy experiments can also constrain neutrino masses. and excluding the LSND results. Given the atmospheric and solar neutrino observations. we will include LSND results. The oscillation explanation 2 of the solar neutrino problem. 3 In a forthcoming analysis. in the case of atmospheric neutrinos.3. see [30–33] and references therein. Kamiokande [4]) and solar neutrino experiments (Homestake [5. which allows R-parity (RP ) violation 2 Other possible solutions. pseudo-Dirac [29]) are preferred for each solution of the solar anomaly combined with SuperKamiokande and CHOOZ constraints. Losada / Nuclear Physics B 585 (2000) 45–78 Observations in favour of the oscillation solutions came recently with the SuperKamiokande collaboration [1] results. The evidence for ν¯ µ → ν¯ e observed by LSND has not been confirmed or excluded by the KARMEN [17] experiment. Recent interest in this subject has produced a very large number of possible models. gauge invariance and renormalizability no longer ensure lepton number L (or baryon number B) conservation. Abada.

The paper is organized as follows. As a by-product. the relevant mass squared differences 1m2 and mixing angles sin2 2θ assuming the oscillation solution as an explanation of the solar and atmospheric neutrino deficits. S are the lepton and baryon number and the spin of the particle. We then check whether the upper bound coming from neutrinoless beta decay (ββ)0ν [35–37] and also the global cosmological upper bound on the sum of the neutrino masses [38] are satisfied. and defined by . at tree level through mixing with the neutralinos. We start by parametrizing the general neutrino eigenmasses and eigenstates and use the neutrino data to constrain the ranges of variation of the general inputs. In Section 5 we present our results for the general case and for the specific case of Majorana masses arising in the R-parity violating model. we compute the masses and mixing angles of the neutrino spectrum assuming three generations of neutrinos (this means that we are not taking the LSND result into account). In Section 4 we introduce our notation for an application of our general results to the MSSM with RP -violation. In Section 4. General inputs In this work. in Section 6 we summarize and conclude. respectively). where L. These inputs are from laboratories such as the CHOOZ experiment [34]. 2. In Section 2 we expand on the inputs used. We thus constrain all relevant RP conserving and non-conserving couplings by imposing the constraints from atmospheric and solar neutrino anomalies and the CHOOZ experiment. B. M. we see that in this model the constraints from neutrinoless double beta decay and cosmology are automatically satisfied. and through loop diagrams that violate lepton number (in two units).A. We apply our general results to the MSSM with R-parity violation at one-loop order. Losada / Nuclear Physics B 585 (2000) 45–78 47 (RP = (−1)L+3B+2S . allowing the presence of both bilinear and trilinear RP -violating couplings without imposing any hierarchy between the parameters. We also give the results for the tree-level neutrino mass matrix and the loop corrections to each matrix element. allows left-handed neutrinos to obtain a Majorana mass. The best limit on the effective mass meff appearing in (ββ)0ν .1 we introduce a toy-model that corresponds to several limiting cases of the MSSM without R-parity conservation. We then apply all relevant existing bounds on the neutrino masses and mixing angles to constrain the elements of our neutrino mass matrix in the flavour basis. Finally. Abada. In Section 3 we establish our notation for the general case of a 3-generation neutrino mass matrix.

X .

X .

2.

.

.

2.

.

. .

(1) mνi Uei .

6 mνk .

Uek |meff | = .

L. Another direct upper bound comes from the measurement of the high energy part of the 3 H β-decay spectrum: the upper limits on the electron neutrino mass obtained in the TROITSK [39] .5−1.h.0) eV (at 90% C. The CP phases that might appear in the mixing matrix Uek are not relevant in our derivation. term in Eq.). because we only use the second r. (1). i k has been derived in the Heidelberg–Moscow 76 Ge experiment [35]: |meff | < (0.s.

then 1m212 = m22 − m21 is the parameter relevant to the solar case and 1m213 = m23 − m21 is the dominant one for the atmospheric case with 1m212  1m213 and 1m223 ∼ 1m213 . 2.41] experiments are mν < 2.46.4 eV. M. 3. since a hot dark matter component is not a necessary ingredient in this case. 4 A recent analysis (observation of distant objects favouring a large cosmological constant instead of HDM) suggesting a non-zero cosmological constant [42] shows that the bound of a few eV for the sum of the neutrino masses is no longer imposed.5 eV and mν < 3. In view of the ranges of the mass squared in Table 1. such that m 1 6 m2 6 m3 . if the neutrino spectrum is hierarchical. Losada / Nuclear Physics B 585 (2000) 45–78 and MAINZ [40. respectively. MSW with a large mixing angle (MSW-LMA) and MSW with a small mixing angle (MSW-SMA). It is important to keep in mind that the neutrino oscillation scenario can only restrict mass squared differences but not fix the absolute neutrino mass scale. for this case with three generations. is between the two lightest mass eigenstates or the two heavier ones. which suggests P that i mνi = m1 + m2 + m3 < few eV.43–46]. The corresponding mixing angles (see the parametrization below) are sin2 2θ12 and sin2 2θ23 . The estimates required to fit the data for the neutrino mass squared differences and the mixing angles are shown in Table 1. 4 Global fits to the data show that neutrino oscillations among three neutrino flavours are sufficient to accommodate the solar and atmospheric data [29. it cannot distinguish whether the smallest mass splitting. one in which the three neutrinos are quasi-degenerate and another (called pseudo-Dirac) in which only two neutrinos are quasi-degenerate and the third is much lighter (or heavier). i = 1. To fix our notation. which corresponds to the solar data.47]. Three different types of solar experiments (using different detectors: chlorine.48]. Besides the hierarchical mass spectrum. There are also upper bounds on the neutrino masses coming from astrophysical and cosmological considerations such as the one from hot dark matter (HDM). these experiments suffer from some ambiguities referred to as “the negative mass squared problem”. .48 A. there exist two other alternative neutrino spectra [29. to accommodate them. However. (2) i=1 We hierarchically order the mass eigenvalues and denote them mi . which correspond to the vacuum oscillation solution. In Table 2 we show the different possible regimes and indicate which is the corresponding mass squared difference. the flavour states νl are expressed in terms of the mass eigenstates νi using the 3 × 3 mixing matrix U ν` = 3 X U`i∗ νi . three regions of oscillation parameter space are allowed [43. respectively. which is still not completely understood and we will therefore not use their bounds. gallium and Cherenkov) are sensitive to different solar neutrino energy ranges. moreover. Abada.

MSW-SMA and Vacuum stand for MSW large mixing angle.5−8)×10−10 CHOOZ > 3×10−3 0.A. (3) M13 M23 M33 We construct the mass squared differences of the eigenvalues of M and the mixing matrix. given in Table 1. the hierarchy m1 6 m2 6 m3 . these masses and also the mixing matrix would be given in terms of the parameters of the model. small mixing and vacuum oscillation solutions. which satisfies. We then apply the constraints. For any given model. Generic mass matrix We consider a general 3 × 3 symmetric mass matrix in the flavour basis for three generations of neutrinos (νe . respectively Experiment 1m2 (eV2 ) sin2 2θ Atmospheric (1−8)×10−3 0. sin2 2θ ) plane. Losada / Nuclear Physics B 585 (2000) 45–78 49 Table 1 MSW-LMA. We denote the three mass eigenvalues by mi for i = 1. In this way we can determine the region. 2. which relates the flavour basis to the mass eigenstates. as mentioned above. Abada.5−1 sin θ13 < 0. ντ ):   M11 M12 M13 M =  M12 M22 M23  . combined.1. M. in the (1m2. in a general way in terms of these input elements. The combined constraints are obtained by requiring the atmospheric. Computation of the mass and mixing matrix 3.22 Table 2 Different possible regimes and corresponding mass squared difference Spectrum Solar 1m212 Hierarchy Degenerate Pseudo-Dirac Atmospheric 1m223 1m223 or or 1m213 1m212 1m212 1m213 1m213 3. . νµ . CHOOZ and one of the solar solutions to be satisfied simultaneously.4−1)×10−5 (0. The answer is then translated into terms of ranges where the parameters of the model vary so as to fulfill neutrino data constraints. that satisfies each of the three possible combinations of constraints simultaneously.6−1 10−3 −10−2 0.85−1 Solar MSW-LMA MSW-SMA Vacuum (3−30)×10−5 (0. 3.

on |U`3 |2 .53] have established in a basis-invariant way that neutrino masses are always generated in these models even when universality of the soft SUSY-breaking terms is assumed at some high scale. which contains the relevant mixing angles. Application: MSSM with R-parity violation The specific RP -violating model to which we apply our general results has been extensively analyzed in the context of pure phenomenology. The other elements of the matrix are not constrained by any direct experimental observation. We use the Chau and Keung parametrization of a 3 × 3 matrix [49] given by   s12 c13 s13 e−iδ c12 c13  U =  −s12 c23 − c12 s23 s13 eiδ c12 c23 − s12 s23 s13 eiδ s23 c13  diag eiα1 . [29. As stated in the introduction here we do not bias our selection of RP -violating couplings. In Refs. our results present the allowed ranges for these parameters using the solar and atmospheric anomalies and the result from the CHOOZ experiment only. with ` = e. With this parametrization.2. 1 . depends only on the first row of the mixing matrix in Eq.50 A..50–59] different phenomenological aspects of this theory have been analyzed with respect to neutrino masses. s12 s23 − c12 c23 s13 eiδ −c12 s23 − s12 c23 s13 eiδ c23 c13 (4) where cij = cos(θij ) and sij = sin(θij ). as we will see. Refs. Previous studies have also tried to constrain the different RP -violating couplings that appear in the MSSM Lagrangian. which would be contrary to a common practice of keeping only one R-parity-violating parameter at a time. considering only the effect of bilinear terms [60–63] or only of trilinear couplings [64–66]. which we need for our general analysis. µ. .. τ .e. Both tree-level and one-loop effects have been considered. Abada. Finally. we obtain the rotation matrix. In the atmospheric case.e. or from both [68–74]. relevant to the case of solar fluxes.2 are the Majorana ones (we have two additional CP phases in the case of Majorana particles). In addition we do not impose any hierarchy between the trilinear couplings. 4. 2. As previously mentioned. we are going to work with a real mixing matrix and will ignore CP phases because we are interested in our derivation in the modulus of the matrix elements. Generic mixing matrix Computing the mass eigenstates of the mass matrix (3). [52. i. M. solar and atmospheric neutrino data. In particular. (4). with i = 1. the oscillation probability depends on the last column of (4). Losada / Nuclear Physics B 585 (2000) 45–78 3. as well as in trying to fit the data of neutrino anomalies. The survival probability Pνe →νe . eiα2 . δ is the Dirac CP phase and α1. 3. We also include all possible bilinear and trilinear parameters for the MSSM without R-parity. on |Uei |2 . i. we directly obtain the parameters sin2 2θ12 and sin2 2θ23 and also the relevant CHOOZ parameter sin θ13 . We will present our comparison to previous analyses in the last section.

Note that thanks to the additional degrees of freedom. which we take to be universal. M. (p) X M1 1 (p) (k) λqkp λmpk m` sin 2φ` ln (p) (mqm )loop = 16π 2 M k.2 and Mq1. matrix has only one non-zero eigenvalue. we can rotate in the flavour space of the “down-type” scalar fields to set the vacuum expectation values of the sleptons to be zero. Abada. which could provide two different mass splittings. The function f is defined to be 5 This can be done order by order in the loop expansion when one appropriately defines the mass matrices of the Higgs sector [70].A. (a. (9) = mq sin 2φq ln (p) = mq (p) 2 (p) 2 (p) (p) 2 Mq2 Mq1 − Mq2 Mq2 Mq2 (p) where m`. ν = 2 αe ατ αµ ατ ατ the corresponding expressions for the αi ’s are given in Appendix B. The loop corrections. 5 Henceforth. b) are SU(2) indices. Note that the λ couplings are antisymmetric in the first two indices. (5) we have 9 additional λ couplings and 27 new λ0 couplings compared to the R-parity-conserving case. λ0ik = hdik . we will work consistently in this basis. The usual R-paritypreserving Lagrangian is obtained when only µo . X = A + µ tan β [75]. from slepton–lepton and squark–quark loops to each element of the mass matrix are given by [51].and right-handed squarks and sleptons. shown in Fig. k) are flavour indices. β) are flavour indices running from 0 to 3. Losada / Nuclear Physics B 585 (2000) 45–78 51 The most general renormalizable superpotential for the supersymmetric Standard Model with lepton-number violation is   (5) W = ab µα Lˆ aα Hˆ ub + λαβk Lˆ aα Lˆ bβ Eˆ kC + huik Qˆ ai Hˆ ub Uˆ kC + λ0αik Lˆ bα Qˆ ai Dˆ kC . . (7) Mq2 (p) Here the angles φ` and φq are given in terms of the slepton and squark mass eigenstates for a given flavour (p) M1.p +3 2 X (p) λ0qkp λ0mpk m(k) q sin 2φq ln k. In the model of Eq.q are lepton and quark masses of flavour p. (8) = m` sin 2φ` ln (p) = m` (p) 2 (p) 2 (p) (p) 2 M2 M1 − M2 M2 M2 ! (p) (p) !! Mq1 Mq1 2X X (p) (p) (p) (p)  Ln f xq .q2 as ! !! (p) (p) M1 M1 2X X (p) (p) (p) (p)  Ln f x` . λ0i0k = hdik are non-zero and we can identify Lˆ o ≡ Hˆ d . The Lˆ α are the doublet superfields with hypercharge Y = −1. the (i. j. It is thus necessary to include additional corrections. We use the following parametrization for the tree-level mass matrix in the basis of charged-lepton eigenstates:  2  αe αµ αe αe ατ  αµ αe (6) Mtree αµ2 αµ ατ  . and the (α.p (p) (p) ! Mq1 (p) . and provides a mixing term between the left. 1.

and that the λ. however. One-loop Feynman diagrams contributing to the neutrino masses. the total mass matrix we consider is loop Mν = Mtree ν + Mν .p Mq2 k. In this model. for simplicity we do not introduce additional notation. 1. (10) Mq2 We consider here that we are in the down-quark mass eigenstate basis. (13) .52 A. M. in order to simplify. respectively.p (11) We use the following parametrization for the one-loop mass matrix written in the same basis as the tree-level mass matrix:   m11 m12 m13 loop (12) Mν =  m12 m22 m23  . Finally. where we have kept only the relevant contributions after employing the mass hierarchy selection from the chargedlepton and down-quark sector. m13 m23 m33 The loop mass matrix elements mij are given in Appendix C. We are then led to ! f (xq ) X 0 1 f (x` ) X (k) (p) 0 (k) (p) X λqkp λmpk m` m` + 3 2 λqkp λmpk mq mq . Losada / Nuclear Physics B 585 (2000) 45–78 Fig. (mqm )loop = 16π 2 M22 k. f (x) = − 1−x (p) (p) x` = M1 !2 (p) M2 (p) !2 and (p) xq = Mq1 (p) . we drop the slepton-flavour dependence from (p) (p) X X (p) 2 f (x` ) and (p) 2 f (xq ) and consider them to be universal in the slepton and M2 Mq2 squark sector. λ0 couplings have been redefined in terms of the couplings appearing in the superpotential and of the corresponding Cabibbo–Kobayashi–Maskawa matrix elements. Abada. ln x .

K2 + αe ατ K2 + αµ ατ K2 + ατ2 where X f (x` ) 2 2  λ mτ . or K2 = 0. • λ0 = 0 or. µµ and µτ .. 6 We especially thank S. Naively. In fact we are naively assuming in Eq. . (16) The different physical motivations (or limits) for these approximations correspond to the cases where: q • λ = 0 or f (x` ) = f (xq ) and M2 = M2 gives K1 = 0 due to the hierarchy mb  mτ . K2 = 3 16π 2 M q2 K1 = (15) 2 Generically this matrix has three non-zero eigenmasses. In this toy model.τ appearing in the tree-level mass with the neutrino data. K1 = 0.e.77]. m2 . if we take the limit in which the squarks decouple. or αµ = αe ). K2 = 0. and three bilinear couplings. m3 }. 6 An additional gaugino–sneutrino (Higgs) loop appears where the R-parity violation is now contained in the mass-mixing term between the sneutrino and antisneutrino. we put ourselves in the naive picture where all the λ couplings are equal and also the λ0 couplings. Abada. 4. such that αe ∼ αµ  ατ . our mass matrix can be written as follows   K1 + K2 + αe2 K1 + K2 + αe αµ K2 + αe ατ (14) Mν =  K1 + K2 + αe αµ K1 + K2 + αµ2 K2 + αµ ατ  . (11) (see Appendix C) that λ133 = λ233 = λ333 and λ0133 = λ0233 = λ0333 .µ. M. µe . λ and λ0 .A. it can be shown that this effect is small when the soft mass vector B = Bµα and the vacuum expectation value vector of the down-type scalar fields vα are aligned [76. a full treatment of this effect requires a one-loop calculation of the scalar mass matrices (Higgses and sleptons). 16π 2 M22 X f (xq ) 0 2 2  λ mb . Under several different approximations (i. However. • Assume some hierarchy between the bilinear terms in the Lagrangian. Losada / Nuclear Physics B 585 (2000) 45–78 53 An important remark is that there are additional loop corrections that can generate neutrino masses through the sneutrino mass splitting in these models. this mass matrix yields only two non-zero eigenvalues U t × M × U = diag{0. We thus have two universal trilinear R-parity couplings. which is beyond the scope of this paper. Davidson for discussions on this point. Toy model In order to have an idea of how to constrain the whole space of parameters given by the different λ and λ0 involved in the loop contribution to the mass and the µe.1.

in order to have a combined fit of the vacuum solution for solar anomaly. After several runs. M. SuperK constraints and one (or all) of the solar solutions is (are) possible. Abada. 2 that all three possible solutions for the solar anomaly. We also note that the density of points (solutions) for Vacuum + CHOOZ + SuperK is smaller than the one for MSW-SMA + CHOOZ + SuperK which in turn is much smaller . 2. 1m2 in units of eV2 as a function of sin2 2θ . The atmospheric and different solar solutions that are presented result from a combined fit with CHOOZ and SuperK constraints. Allowing only combined solutions of solar. CHOOZ and atmospheric SuperK bounds can be simultaneously satisfied. SuperK and CHOOZ data restricts the available region of parameter space. we choose the interval of variation of the matrix elements Mij in such a way that. For example. and our discussion of them is given in the next section. sin2 2θ23 ).1. Results 5. We also see in Fig. We present in Fig. sin2 2θ12 ). 5. combined to CHOOZ. which satisfy the combined constraints (CHOOZ. 2 the allowed points on the (1m2 .54 A. In this section we will simply present the results. SuperK and solar data) for the possible solutions of the solar and atmospheric anomalies.5 eV. The MSW-LMA and MSW-SMA solar solutions are represented by the parameters (1m212 or 1m223 and sin2 2θ12 ) while the vacuum solution is represented by the parameters (1m212 . Losada / Nuclear Physics B 585 (2000) 45–78 Fig.5 eV 6 Mij 6 0. (3). together with CHOOZ and SuperK constraints then the matrix elements have been varied in the interval −0. General results We have performed a general scan of parameter space made up by the six matrix elements that appear in Eq. sin2 2θ ) plane. The atmospheric solution is represented by the parameters (1m213 .

3 −3.1.4 8.0] eV Inputs M11 (10−2 M12 (10−2 M13 (10−2 M22 (10−2 M23 (10−2 M33 (10−2 eV) eV) eV) eV) eV) eV) Minimum Maximum −2. Table 3 Allowed ranges of the input parameters that satisfy MSW-LMA. For illustration. Fig. M. which are given in the next subsections. In Figs.0 7. Our results for the latter two quantities show that for the specific range that we have chosen to satisfy the combined fit of solar..9 −1. This bound is valid when we vary the mass matrix elements in the ranges that are specified below for each subcase. a slightly stronger bound on the effective mass can be placed by our analysis and will be given below in each subsection. 3 shows the available regions in parameter space from the combined constraint of SuperK data together with CHOOZ and the MSW-LMA solution. MSW-LMA For a given model. M33 (eV).9 8. we can cover most of the allowed region that satisfies the combined constraints mentioned above. M23 (eV). 4(a) and 4(b) we plot the results of the individual mass eigenvalues and the effective mass (only valid for Majorana masses) together with the sum of the eigenmasses. for Majorana neutrinos. They also fulfill neutrinoless double beta decay and cosmological constraints. with this scan of parameter space. SuperK and CHOOZ constraints simultaneously. CHOOZ and atmospheric constraints.9 . the present bounds are always satisfied. i.3 3. M13 in (eV). We also plot the effective mass relevant for neutrinoless double beta decay and the sum of the mass eigenvalues. In order to obtain these results we have initially allowed all six matrix elements to vary in the interval [−1. In fact.0] eV.A.0. M12 (eV).1. Mij ∈ [−1. which varies when we scan over parameter space. Tables 3.3 4. It is clear from Fig.1 −7.7 −4. To study the spectrum for each of the possible combined solutions we give samples of the different mass spectra for different values of the matrix elements Mij . imposing MSW-LMA. Abada. 4 and 5. present the allowed ranges of the input parameters (matrix elements) that satisfy the different solar oscillation solutions together with the CHOOZ and the atmospheric SuperK constraints. +1. This illustration is for a model where the allowed inputs vary initially in a given interval.0. atmospheric SuperK and CHOOZ constraints simultaneously determines the allowed ranges for the input parameters M11 (eV). 3 that. 5.e.3 7. Losada / Nuclear Physics B 585 (2000) 45–78 55 than the one corresponding to MSW-LMA + CHOOZ + SuperK. we have present our plots for the spectrum as a function of the matrix element M13 .4 −5. M22 (eV). +1. summarized in Table 3.

Here once again there is a bound . 1m2 in units of eV2 as a function of sin2 2θ . meff . this illustrative table is valid for a given model in which the mass matrix elements vary initially. We show in Fig. before applying the solar.1. 5 × 10−2 eV. 4(b). in the range [−1.e. Fig. 3). SuperK and CHOOZ constraints simultaneously gives the ranges for the input parameters summarized in Table 4. Losada / Nuclear Physics B 585 (2000) 45–78 Fig. MSW-SMA Imposing MSW-SMA. As in the previous section. We have noticed that with our scan of the parameter space the resulting density of points in Fig. the effective mass and the values of 6mi when we impose the constraints. CHOOZ and SuperK constraints. sin2 2θ23 ) while the MSW-LMA solar solutions are presented with the parameters (1m212 or 1m223 and sin2 2θ12 ).. We can see from Fig. +4] × 10−2 eV (while the other matrix elements vary in the intervals given in Table 3). The atmospheric and solar solutions that are presented result from a combined fit of MSW-LMA. Abada.2. 5(a) is much smaller than in the case of MSW-LMA (Fig. +1. 5. i. atmospheric and CHOOZ constraints.0. a strong bound can be placed on the neutrinoless double beta decay effective mass. for M13 ∈ [−4. 5(b) the results of the individual mass eigenvalues. 4(a) that the hierarchical spectrum is preferred for small values of our parameter M13 . 3.56 A. which is the relevant quantity for HDM when we impose the constraints simultaneously. The atmospheric solutions are represented with the parameters (1m213 . for the entire range of variation of the matrix element M13 .0] eV. 5(a) gives the combined constraint coming from SuperK data together with the CHOOZ bound and the MSW-SMA solution. M. Using the results presented in Fig.

Abada.A. M. Mass spectrum (a) which presents the three possible mass eigenvalues (eV). 4. 57 . The effective mass and the sum of eigenmasses are given in (b). Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. The spectrum presented results from a combined fit with CHOOZ and SuperK and MSW-LMA solar constraints.

M. In (b). The atmospheric solutions are represented with the parameters (1m213 . 5. In (a) we represent 1m2 in units of eV2 as a function of sin2 2θ . . the mass spectrum is represented together with meff and the sum of the eigenvalues (eV).58 A. The atmospheric and solar solutions that are presented result from a combined fit with CHOOZ and SuperK constraints. Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. sin2 2θ23 ) and the MSW-SMA solar solutions with the parameters (1m212 or 1m223 and sin2 2θ12 ). Abada.

Here we also have a strong bound on the effective mass . there are few solutions that satisfy simultaneously all of the constraints. 6(a) how the allowed points on the (1m2 .0 4. +0. where the spectrum of eigenmasses is represented together with the effective mass meff and the sum over the eigenmasses. The allowed ranges for the input parameters.98 −3.0] eV Inputs M11 (10−2 M12 (10−2 M13 (10−2 M22 (10−2 M23 (10−2 M33 (10−2 eV) eV) eV) eV) eV) eV) Minimum Maximum −4.0 +0.0 5.7 −4.A.5. (5(a)). SuperK and CHOOZ constraints simultaneously. This illustrative table is valid for a given model in which the mass matrix elements vary initially.9 Table 5 Allowed ranges of the input parameters that satisfy solar Vacuum. 5.1. M.3 4. One can see in Fig.0 2.0] × 10−2 eV (the other matrix elements take values in the ranges given in Table 4).5 +0. before applying the solar.3 8. 6(b).5.0 −4. few × 10−2 eV for M13 ∈ [−4.0 −1.0 −4.3 −8. (3)) and MSW-SMA (Fig.0 4. +4. In Fig.0. They also fulfill neutrinoless double beta decay and cosmological constraints.0 −4.3. Losada / Nuclear Physics B 585 (2000) 45–78 59 Table 4 Allowed ranges of the input parameters that satisfy MSW-SMA.5] eV. SuperK and CHOOZ constraints simultaneously.0 −4.98 Maximum 1. are summarized in Table 5. Abada.0 4. atmospheric and CHOOZ constraints. sin2 2θ ) plane are reduced compared to the density of points in the cases of MSW-LMA (Fig. Vacuum solar solution In this case.1 meff . after imposing the vacuum oscillation solution.1 4.0 8. +0.9 4. +1.0. They also fulfill neutrinoless double beta decay and cosmological constraints. a salient feature is the presence of two nearly degenerate eigenvalues (m1 ∼ m2 ). This illustration is for a model where the allowed inputs vary initially in the interval [−0. This illustration is for a model where the allowed inputs vary initially in the interval [−1. in the range [−0.5] eV Inputs M11 (10−2 M12 (10−2 M13 (10−2 M22 (10−2 M23 (10−2 M33 (10−2 Minimum eV) eV) eV) eV) eV) eV) −0.1 4. SuperK and CHOOZ constraints simultaneously.

2. 5. Applying the results of our general scan of parameter space. αe (eV1/2 ). 8(b) that in the range of parameters x ∈ [0. 7 shows the regions that satisfy the combined constraint coming from SuperK data together with CHOOZ and the MSW-LMA solution. 5. One can read from Fig.28 0.2. we show how the mass spectrum and the effective mass vary as a function of x = |αe (eV1/2 )|.280 −0.7 × 10−2 0. 0. which we have summarized in Table 6. They also fulfill neutrinoless beta decay and cosmological constraints Inputs Minimum Maximum K1 (eV) K2 (eV) αe (eV1/2 ) αµ (eV1/2 ) ατ (eV1/2 ) 0. Table 6 Allowed ranges of input parameters that satisfy MSW-LMA.0.0 × 10−2 2. few × 10−2 eV as we vary M13 in [−4. αµ (eV1/2 ). Three different possibilities can be considered: – the tree-level contributions are much larger than the loop contributions. 8(b).238 3. Losada / Nuclear Physics B 585 (2000) 45–78 meff . SuperK and CHOOZ constraints simultaneously defines the ranges for the input parameters K1 (eV) (which is proportional to λ2 ).154 0. we are in fact including all of the above-mentioned cases in our analysis.238 . As each combination of constraints defines a different allowed range of parameters we will present below the corresponding bounds on R-parity-violating couplings separately for each case. K2 (eV) (which is proportional to λ02 ). 4. Fig. 5 × 10−2 eV. For illustration. in the same region of the parameter space. Bounds on R-parity-violating couplings from neutrino data Our approach to the numerical analysis is to use low energy input without any reference to high-scale physics. 8(a) the results of the individual mass eigenvalues. We can also determine. that is meff . a strong limit on m1 + m2 + m3 .0 −0. and in Fig. M. we can put a strong bound on the effective mass. the effective mass meff together with the sum over the eigenmasses when we impose all the constraints simultaneously. ατ (eV1/2 ). We plot in Fig. – the tree-level contributions are of the same order as the loop contributions. SuperK and CHOOZ constraints simultaneously.0 0.2 eV. Abada.60 A.0] × 10−2 eV (the other matrix elements take values in the ranges given in Table 5).114 −0.1].1. – the tree-level contributions are much smaller than the loop contributions. 0. The loop contributions are proportional to different R-parity-violating parameters from those which control the tree-level terms. MSW-LMA Imposing MSW-LMA.

61 . In (a) we represent 1m2 in units of eV2 as function of sin2 2θ .A. the mass spectrum is represented together with meff and the sum of the eigenvalues (eV). Abada. sin2 2θ23 ) and the vacuum solar solution (1m212 . 6. In (b). Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. respectively. M. The square and the triangle refer to the atmospheric solution (1m213 . sin2 2θ12 ).

2 Xf (xl )mτ q2 16π 2 M2 3Xf (xq )m2b  × 2. respectively. sin2 2θ12 ).0 × 10−2 .154 eV−1/2 . 2 + tan θW M2 ) 2 4 det M × 0. 7. 1m2 in units of eV2 as a function of sin2 2θ . Applying our results from Table 6 matrix of Eq. (17) (18) Using the results of Appendix B we can also give a bound on the bilinear couplings. (14) we find that the following bounds on the RP -violating parameters exist for each solar solution combined with SuperKamiokande and CHOOZ data constraints.7 × 10−2 . µ2µ 6 2 2 2 mZ cos β(M1 + tan θW M2 ) 2 4 det M × 0. For the trilinear couplings we have λ2 6 02 λ 6  16π 2 M22 × 3.238 eV−1/2 . Losada / Nuclear Physics B 585 (2000) 45–78 Fig. The square and the circle refer to the atmospheric solution (1m213 .28 eV−1/2 . µ2τ 6 2 2 2 mZ cos β(M1 + tan θW M2 ) µ2e 6 m2Z cos2 β(M1 (19) (20) (21) . 2 4 det M × 0. sin2 2θ23 ) and to the MSW-LMA solar solution (1m212 .62 A. M. Abada.

Abada. The mass spectrum (eV) is represented in (a). This is when the MSW-LMA solution.A. the atmospheric one and CHOOZ constraint are imposed simultaneously. M. Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. 63 . 8. The effective mass and the sum of eigenmasses (eV) are represented in (b).

5.28 0. sin2 2θ12 ). M.28 −0. SuperK and CHOOZ constraints simultaneously. MSW-SMA Imposing MSW-SMA.2 × 10−2 0. 9. As in the previous subsection by applying our results from Table 7 to the specific mass .64 A.2.21 Fig.2 × 10−2 −0.0 × 10−3 4. As mentioned previously for our analysis of a generic mass matrix.0 × 10−3 3. 1m2 in units of eV2 as a function of sin2 2θ .21 3. 7). respectively. SuperK and CHOOZ constraints simultaneously determines the ranges for the input parameters given in Table 7. They also fulfill neutrinoless beta decay and cosmological constraints Inputs Minimum Maximum K1 (eV) K2 (eV) αe (eV1/2 ) αµ (eV1/2 ) ατ (eV1/2 ) 0. Abada. The square and the circle refer to the atmospheric solution (1m213 . sin2 2θ23 ) and the MSW-SMA solar solution (1m212 .0 0.0 −4. 9 gives the combined constraint coming from SuperK data together with CHOOZ and the MSW-SMA solution. the density of points is much smaller than in the case of the MSW-LMA solution (Fig. Losada / Nuclear Physics B 585 (2000) 45–78 Table 7 Allowed ranges of the input parameters that satisfy MSW-SMA.2. Fig.

µ. these are summarized in Table 8.0 × 10−3 . (23) 2 4 det M × 4. atmospheric and CHOOZ constraints simultaneously or vacuum oscillation and CHOOZ constraints simultaneously offers no solution of the solar and atmospheric anomalies. 2 4 det M × 0. They also fulfill neutrinoless beta decay and cosmological constraints Inputs Minimum Maximum K1 (eV) K2 (eV) αe (eV1/2 ) αµ (eV1/2 ) ατ (eV1/2 ) 0. (14) we find the following bounds on the RP -violating parameters for each solar solution combined with SuperKamiokande and CHOOZ data constraints: λ2 6 02 λ 6  16π 2 M22 × 3.2.35 eV−1/2 .0 × 10−3 . + tan2 θW M2 ) 2 4 det M × 0. we find that imposing vacuum oscillations.0 − 0.0 0. µ2τ 6 2 2 2 mZ cos β (M1 + tan θW M2 ) µ2e 6 m2Z cos2 β (M1 (24) (25) (26) 5. 2 Xf (xl )mτ q2 16π 2 M2 3Xf (xq )m2b (22)  × 3. These are λ2 6 02 λ 6 µ2i 6  16π 2 M22 × 6.35 0.35 − 0.0 × 10−3 .35 matrix of Eq. µ2µ 6 2 2 2 mZ cos β (M1 + tan θW M2 ) 2 4 det M × 0. (29) .28 eV−1/2 . We determine the bounds on the RP -violating parameters from the results from Table 8.0 × 10−3 1. Abada.35 6. Losada / Nuclear Physics B 585 (2000) 45–78 65 Table 8 Allowed ranges for input parameters that satisfy vacuum oscillation constraints. 2 + tan θW M2 ) m2Z cos2 β(M1 (28) i = e. Vacuum In our present application. We have determined which are the allowed ranges for the input parameters by imposing only vacuum solution constraints. τ.20 × 10−2 .2 × 10−2 eV−1/2 . Xf (xl )m2τ q2 16π 2 M2 3Xf (xq )m2b (27)  × 1.21 eV−1/2 .A.3.35 0.35 − 0. M.20 × 10−2 0.

0 × 10 2 −8 .3 × 10−4 GeV. 2 and 3. CHOOZ and MSW-SMA (Figs. 12(a) and 12(c)) as well as for Super. Fig.1 × 10−9 . CHOOZ and MSW-LMA (Figs. Abada. K2 ∝ λ0 2 . Here one can observe how the region of parameter space from the fit that satisfies MSW-LMA. 11).9 × 10−4 GeV. µµ 6 3. (31) • Vacuum µi 6 4. sin2 2θ ) solar and atmospheric parameters on the RP -violating parameters We now plot in Figs. In Figs. For the particular case where Msusy = 100 GeV.02 × 10−9 . Discussions and conclusions Our results from the general analysis show that there are many regions in parameter space that can accommodate simultaneously the MSW-LMA and atmospheric oscillation solutions together with the CHOOZ constraint. Dependence of the (1m2 . Our general results can be analyzed by considering the SUSY particle spectrum in detail.5 × 10−7 . 10) and the atmospheric variables 1m213 and sin2 2θ23 (Fig. 12 illustrates the dependence of the solar and atmospheric parameters on the sum of the squares of the RP -violating bilinear terms for the combined fits of SuperK. λ2 6 1. 02 λ 6 3.8 × 10−10. 12(b) and 12(d)).4 × 10−4 GeV. 10.4. λ 6 3.5 × 10−8 . Limits on λ. CHOOZ and SuperK simultaneously is considerably larger than the available region where we fit the MSW-SMA. and also α 2 = αe2 + αµ2 + ατ2 . M.2 × 10−5 GeV. i = e.6 × 10−4 GeV. see Figs.τ . SuperK and the MSW-LMA solution are satisfied simultaneously. τ. λ2 6 1.9 × 10−4 GeV. µτ 6 2.66 A. we will present these results in a forthcoming paper [78]. (30) • MSW-SMA + CHOOZ + SuperK µe 6 5. λ0 and µi Given our results for the bounds on the RP -violating couplings. 6. (32) 5.µ. µµ 6 3. µ. In comparison. αe. f (x) → 1 and tan β = 2. λ0 2 6 7. we . Msusy . 11 and 12 the different solar and atmospheric parameters as functions of the RP -violating parameters K1 ∝ λ2 . CHOOZ and SuperK atmospheric constraints. Losada / Nuclear Physics B 585 (2000) 45–78 5.3.4 × 10−4 GeV. where the constraints from CHOOZ. 10 and 11. the analysis in terms of RP -conserving SUSY parameters is straightforward. the bounds are • MSW-LMA + CHOOZ + SuperK µe 6 1. we have plotted the solar parameters 1m212 and sin2 2θ12 (Fig.2. µτ 6 2. λ0 2 6 7. A common assumption used to place bounds in this model is to take all RP -conserving mass parameters to be of the same order.

Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. 67 . The solar parameters 1m212 and sin2 2θ12 as functions of the different R-parity parameters αe (a).A. M. SuperK and CHOOZ constraints. 10. in the case of a combined fit with MSW-LMA. Abada. αµ /ατ (b).

The solar parameters 1m212 and sin2 2θ12 as functions of the different R-parity parameters K1 (∝ λ2 ) (c) and K2 (∝ λ02 ) (d). — Continued. in the case of a combined fit with MSW-LMA. Abada. SuperK and CHOOZ constraints. . Losada / Nuclear Physics B 585 (2000) 45–78 (c) (d) Fig. M. 10.68 A.

69 .A. Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) Fig. M. 11. The atmospheric parameters 1m213 and sin2 2θ23 as functions of the different RP parameters αe (a). αµ /ατ (b). in the case of a combined fit with MSW-LMA. SuperK and CHOOZ constraints. Abada.

— Continued. Abada. Losada / Nuclear Physics B 585 (2000) 45–78 (c) (d) Fig. in the case of a combined fit with MSW-LMA.70 A. . SuperK and CHOOZ constraints. M. 11. The atmospheric parameters 1m213 and sin2 2θ23 as functions of the different RP parameters K1 (∝ λ2 ) (c) and K2 (∝ λ0 2 ) (d).

A. Abada. The atmospheric (1m213 and sin2 2θ23 ) (in (a)) and solar parameters (1m212 and sin2 2θ12 ) (in (c)) as functions of α 2 = αe2 + αµ τ solution (a) and (c) and in the case of MSW-SMA (b) and (d). 71 . 12. SuperK and CHOOZ constraints are also imposed. Losada / Nuclear Physics B 585 (2000) 45–78 (a) (b) 2 + α 2 in the case of MSW-LMA Fig. M.

— Continued. 12. . Abada. M. Losada / Nuclear Physics B 585 (2000) 45–78 (d) (c) Fig.72 A.

the effective mass constraint and the cosmological constraint are always satisfied. We have also indicated previously when a stronger bound can be placed on these quantities. The bound is implemented by using the upper bound on the electron neutrino mass from the 3 H β-decay spectrum. for the combined vacuum oscillation solution. (30) agree with the previous bounds. In reference to Figs. It is clear that increasing the range of variation of our parameters changes the values of the eigenvalues of the mass matrix. There are also a few values of parameter space that present a degenerate spectrum. The overall conclusion is that the bounds are very strong on the R-parity violating parameters. Losada / Nuclear Physics B 585 (2000) 45–78 73 see from Fig. CHOOZ and SuperK constraints only pseudo-Dirac spectra are present. From our analysis we have seen that fitting only one of the neutrino anomalies is straightforward. Other authors have also studied the possibility of fitting both atmospheric and solar anomalies. (14). sin2 2θ ) plane is covered when MSW-SMA and the atmospheric oscillations solutions together with the CHOOZ constraints are imposed. Some of the previous work regarding neutrino masses in this specific model we consider have tried to fit only one of the 1m2 to either the atmospheric or one of the solar solutions.80] cited for the trilinear couplings λ133 and λ0133 are placed when only one of these two R-parity-violating couplings are considered to be present in the theory. What is much harder to obtain is a region in parameter space that satisfy the combined fits. the three possible spectra are very common. or exclusively trilinear contributions. eventually the same will occur with the MSWSMA and MSW-LMA solutions. We have shown that the simultaneous inclusion of many of these couplings can constrain them much more than other bounds arising from other . given by K1 or K2 depending on whether the trilinear coupling is λ133 or λ0133 . M. As far as the possible spectra are concerned. the region of parameter space for which the atmospheric. CHOOZ and SuperK constraints. As expected. It is really the combination of constraints that solve both atmospheric and solar anomalies which limit the allowed ranges. In contrast. to having a single non-zero matrix element in Eq. 5(a) that not all of the allowed region in the (1m2 . after obtaining a simultaneous solution for the solar and atmospheric anomalies and the CHOOZ constraint. CHOOZ and SuperK constraints both the hierarchical and pseudo-Dirac spectra are more commonly present. larger values of the inputs make it impossible to satisfy the vacuum oscillation constraints. They consider either supergravity-inspired models or exclusively bilinear contributions. of course. Having only one of the trilinears corresponds. respectively. Abada. For a combination of MSW-LMA. 4. 5(b) and 6(b) our conclusions hold and the bounds mentioned in the previous section are still maintained when we study the spectrum in terms of the other parameters Mij 6= M11 . Moreover. we clearly see that for the combined MSW-SMA. we find that for the specific case of the MSSM with R-parity violation there are no regions in parameter which can accommodate simultaneously the vacuum solar solution and all other constraints. CHOOZ and vacuum oscillation constraints are satisfied has very few solutions. although their approaches are somewhat different from ours. We see that in the special limits we have considered on R-parity-conserving SUSY parameters our bounds from Eq. An important conclusion is that. In fact. The usual bounds [79.A.

We also see. of course. ICARUS. Tree-level neutral colour-singlet mass matrix  Mtree N = M ξ ξT m0ν  . will provide additional decisive information to pin down the neutrino mysteries. BooNE. ψL2 3 ). 0    0 (A. that the dependence on a single parameter is complicated. 10–12 illustrate the different dependences of the solar and atmospheric parameters on the R-parity couplings. ψL2 0 . HERON with solar neutrinos as well as KAMLAND. and many others. Appendix B.80]. This is in agreement with the analysis we have done. . K2K. since many solutions can be found for the same value of the variable. the majority of the solar anomaly solutions we have that also satisfy both SuperK and CHOOZ are MSW-LMA. Losada / Nuclear Physics B 585 (2000) 45–78 low-energy or high-energy processes [79. Indeed. Appendix A. thus their effect on the eigenvalues and corresponding mass eigenstates will be tiny. BOREXINO. the matrix  M2  g2 v√cos β  2  Mtree c =  0  0 0 g2 v√sin β 2 0 0 µ0 0 0 µ1 µ2 µ3 m1 0 0 0 m2 0 0   0   . This is due to the large number of different parameters involved. M. Figs. The discreteness of the allowed values of the parameters is due in part to the step we use in our scan of the parameter space.1) m3 can be diagonalized explicitly to obtain the charginos and charged-lepton tree-level masses. It is a good approximation for small R-parity violation to neglect the RPV contribution to the masses in the charged leptons and charginos [82]. The overall conclusion is that the bounds are very strong on the R-parity violating parameters. . ψL2 1 . ψL2 2 .74 A. The future experiments SNO. However. MINOS. Acknowledgements We thank S. . It is clear that for large values of the parameters there are no solutions. Abada. our analysis of the neutrino spectrum requires the µi ’s to be very small. Charged colour-singlet mass matrix Written in the basis Ψ = (−iλ− . Davidson for very interesting and useful discussions. The general global interpretation of the experimental data is that the SuperK seriously challenges MSW-SMA and vacuum oscillations solutions [81].

1) m11 = s b 133 τ 133 b 123 132 q2 16π 2 M22 M2   X f (x` ) loop λ133 λ233 m2τ m12 = 2 2 16π M2    f (xq ) (C.3) loop where Mν = m0ν + mν . where hvi i is the vacuum expectation value of the slepton fields Li .  0   0   0  0 (B. (C. Losada / Nuclear Physics B 585 (2000) 45–78  M1   0   g v sin β  1√  2 =  g1 v√cos β − 2   0   0 0 M2 − g2 v√sin β 2 g1 v√sin β 2 0 0 0 0 g1 v√sin β 2 g2 v√sin β 2 − g1 v√cos β 0 0 0 0 0 µ0 −µ1 −µ2 −µ0 0 0 0 −µ1 −µ2 −µ3 0 0 0 0 0 0 0 0 0 2 g2 v√cos β − 2 75 0   0    −µ3   . (B.A.70]: Mν = −ξ M −1 ξ T + Mν loop loop . 7 = g22 Mνtree ij Appendix C. (B.1) In the limit where the µi ’s are small. M. v1 = vd = v cos β.2) The assumption we make so as to include the one-loop corrections is that only the corrections to m0ν need be considered.4) 4 det M as we are working in the basis where the slepton vacuum expectation value is zero. The basis-invariant expression corresponds to substituting µi µj v12 with (µ < vi > −µi v1 )(µhvj i − µj v1 ). The expression for mν is given in the next subsection. which is parametrized in terms of the αi ’s of Eq. loop (B.2) + 3 q2 λ0133 λ0233 m2b + λ0132 λ0223 + λ0123 λ0232 ms mb . The elements of the matrix Mνtree = ξ M −1 ξ T . [83]. see Ref. (6) are given by (M1 + tan2 θW M2 ) µi µj v12 . Abada. M2 7 For a general discussion of basis-independent parametrizations of R-parity violation. . the effective neutrino mass matrix has the “seesaw” structure −1 T 0 Mtree ν = −ξ M ξ + mν . (7) leads to the following symmetric matrix elements of which we keep only the relevant contributions due to the mass hierarchy in the chargedlepton sector me  mµ  mτ and in the down-quark sector md  ms  mb :     f (xq ) 0 2 2 X f (x` ) 2 loop 2 0 0 m m + λ λ m m λ + 3 λ . thus giving the one-loop-corrected expression for the 3 × 3 neutrino mass matrix [54. Loop-mass matrix Expliciting the indices in Eq.62.

[11] M. D. Lett. Phys. D 59 (1999) 116008.4) (C. Lett. Phys. Phys. Lett. E. Davis et al. Phys. Ramond. [10] Y. Phys... Anthanassopoulos et al. 77 (1999) 241. Lett. Phys. hep-ph/9905426. [13] S.6) 2 References [1] Y. Nucl. Rept. . Weiner. Lett. P. J.. KARMEN Collaboration. [25] J.T. Rev. L. Raby. Rev. Cleveland et al. Lett. Lisi. Pakvasa. Phys. Rev. [5] R. [20] T. B 391 (1997) 491. 68 (1992) 1984. 21 (1968) 1205. [24] G. Becker-Szendy et al. Strumia.. Fogli.. Prog. A. Lett.. Lett. [26] M. Freedman (Eds. [6] B. Rev. Phys. [22] G. in: A. LSND Collaboration. [9] K. Smith. Phys. Altarelli. Rev. B 335 (1994) 237. hep-ph/9907532. B 81 (1998) 1774. Marrone. A. [7] W. Prog. Phys. Kamiokande Collaboration. Ambrosio et al. Soudan-2 Collaboration. Gaillard. Vergados. Lett. hep-ph/9906362. Part. Phys. Slansky. J. [30] G. [15] S. Phys. Phys. Phys. Phys. Phys. A. 496 (1998) 505. Lett. S. 1979. Kim... Proceedings of the Supergravity Stony Brook Workshop. SuperKamiokande Collaboration. Phys. and references therein. Scioscia. Rev. hep-ph/9905301. Barbieri. [16] G. hep-ph/9906281. 58 (1995) 1207. Rev. Leontaris et al. Fukuda et al.J.D. Lett. [4] Y. 81 (1998) 1562. Ellis.B. Feruglio. Rossi. B 434 (1998) 451. Lett. Lett. Abdurashitov et al. Carena et al. 77 (1996) 1683.. Part. Nucl. A. Sawada..S. Phys. L. Halprin. Hirata et al. D 59 (1999) 117303.. Allison et al.. Phys. B 77 (1996) 4708. [17] B.).J. Nucl. Langacker. Lett. 81 (1998) 1158. [27] P. Lola. Zeitnitz et al. B 90 (1980) 249. E. Prog.W. Berezhiani. SuperKamiokande Collaboration.5) 2 loop m33 3X f (xq ) 0 2 2 = λ m . Fukuda et al. Hampel et al. JHEP 9903 (1999) 002. Hall. [12] C. Abada.. Gell-Mann.L. IMB Collaboration.K. JHEP 9812 (1998) 017. R. Astrophys. Phys. 1979. [3] W. H. B 388 (1996) 384. Losada / Nuclear Physics B 585 (2000) 45–78 3X f (xq ) 0 λ λ0 m2 .K. Suppl. Nucl.L.76 A. Yanagida. D 60 (1999) 053006..M. B 449 (1999) 137. and references therein. Roulet. 16π 2 M q2 333 b (C. Dimopoulos. MACRO Collaboration. Rev. Phys. S. 16π 2 M q2 133 333 b  2    f (xq ) 0 2 2 X f (x` ) 2 loop 2 0 0 λ233 mτ + 3 q2 λ233 mb + λ223 λ232 ms mb . D 53 (1996) 6381. Phys. [19] M.. Lett. [29] R.). Phys. Coleman. M. 16π 2 M q2 233 333 b (C.N. B (Proc. Rev. Proc. Fukuda et al. B 436 (1998) 33. Phys. KIAS-P99028.. 40 (1998) 169. F. Suppl. J.3) (C. Van Nieuwenhuizen. Hall. Barbieri. D. hep-ph/9905536. [28] Z. [21] S.) 38 (1995) 331. [23] S. Proceedings of the Workshop on Unified Theories and Baryon Number in the Universe. Gelmini. Sugamoto (Eds.A. astro-ph/9907131. Rev. Casas et al. Rev. [18] R. in: P. m22 = 16π 2 M22 M2 loop m13 = loop m23 = 3X f (xq ) 0 λ λ0 m2 . Glashow. N. Kamiokande Collaboration. M. [8] D. [2] R. 40 (1998) 71. G. [14] A. B 433 (1998) 9. hep-ph/9905381.

[44] V. South Africa. Petcov. D 55 (1997) 5772. P.L. hep-ph/9811201.T.A. South Africa. Phys. hep-ph/9905549. Phys. D 59 (1999) 33001. B 478 (1996) 3. Peres. Mukhopadhyaya. Krastev. Nir.J. to be published in the Proceedings of 17th International Workshop on Weak Interactions and Neutrinos (WIN 99).M.. Grossman. Phys. Nucl. Rev. S. P. Valle. Lett. astro-ph/9903335. Rev. Bednyakov. X. [64] S. Grimus. [55] F. Dress. [45] M. O. Phys. Croft. Phys. to be published in the Proceedings of 17th International Workshop on Weak Interactions and Neutrinos (WIN 99). Phys.J. N.Y. Phys. [46] J. Bahcall. Y. hep-ph/9812360. Pas. E. D 56 (1997) 6117. Hempfling.J. Nucl. Pakvasa. Cape Town. hep-ph/9907216. Mod. Phys.J. hep-ph/9809234. 40 (1998) 337. S. D 52 (1995) 5319. [52] E. Phys. E. hep-ph/9811363. Lee. H. Valle.C. S. [60] D.C. S. Choi. Singapore. Bilenkii. Phys. Ravindran. E. Lobashev. Phys. X. [71] A. Lee. Y. South Africa. Nardi. Roy. [70] O. CHOOZ Collaboration. Kang. . January 24–30. H. World Scientific. Bonn. B 384 (1996) 123.M. Nucl. Rev. Kim.G. T. and references therein. [61] E. Phys. Lett. C. Georgi. Raychadhuri. A. Keung. Nardi. Lee. J. [57] B.Y. Kong. to be published in the Proceedings of 17th International Workshop on Weak Interactions and Neutrinos (WIN 99). [42] G. W. [73] A. S. Rev. [47] N.W. Joshipura. Rev. Nir. Smirnov. [51] Y. [63] J. Günther et al. Nucl. Lisi. Romão.M. D 59 (1999) 091701. G. Chun.S. hep-ph/9904369. Lett. Polonsky. C. A 14 (1999) 903. Baudis et al. Davé. Cape Town. Nilles. D 55 (1997) 54.P. E. B 272 (1986) 1. Glashow. Rev. Godbole. 1999. [38] R. D 54 (1996) 3427. 1999. Nucl. [36] M. Gonzalez-Garcia. [62] V. A. Nucl. [54] M. B 407 (1997) 219. [59] H. Smirnov. Abada. [33] J. 53 (1984) 1802.L. 1999.. Nardi. [43] G. Chau. K. Losada / Nuclear Physics B 585 (2000) 45–78 77 [31] B. Chun. S. Giunti. Y. [37] L. 1989. Lett.F. Phys. H. Rev. January 24–30. P.Yu. J.K. The Physics of Massive Neutrinos. Prog. [74] A. January 24–30. Haber.V. Rev. F. E. Pilaftsis.E. South Africa. A. to be published in the Proceedings of 17th International Workshop on Weak Interactions and Neutrinos (WIN 99).C. B. Apollonio et al.E. Kaplan. Perrier. M. Vempati. Borzumati. Grossman. D 57 (1997) 5335. B 499 (1997) 33. Datta. [39] V.I. Nucl. Rev. [50] R. 1999. [68] E. R.K. F. Rakshit. Phys. Langacker. Phys. [48] S. Marrone.W. January 24–30. F. Phys. Kang. Y. [41] J. Whisnat. Part.E.C. ter Veldhuis. Phys. D 59 (1999) 093008. hep-ph/9907466. Roy. de Carlos.J. B 543 (1999) 3. Klapdor-Kleingrothaus. S. [40] H.W. Scioscia. Kovalenko. Phys. Phys. P. Vempati. Lett B 451 (1999) 98. [32] S.F. W. Tata. Chun. B 460 (1996) 37. Phys. V. Phys.L.A. [65] M.S. Nunokawa. W. B 440 (1998) 1. Nowakowshi. Phys. B 461 (1996) 19. [35] H. B 442 (1998) 203. Part. 40 (1998) 353. A. K. D 51 (1995) 2421. [56] R. D 58 (1998) 96016. Barger. Joshipura. Hwang.S. Nucl. [58] A. G. Phys.N.. Phys. Rev.W. H. [75] H. Nucl. T.L. G. Phys. Gelmini. Rev. Fogli. Rev. Faessler. A. Gibrat-Bebu. Prog. Phys. B 401 (1993) 67. B 544 (1999) 89. Phys. Choi. Banks. Cape Town. Phys. S. M. hep-ph/9907432. [49] L. Cape Town. Lett. [72] A. hep-ph/9903435. Chun. Kane. Grossman. Hata. Phys. [34] M. [67] G. hep-ph/9903465. Weiler. Vissani. Tata. hep-ph/9808293. B 420 (1998) 397. Barth. [66] S. A. Lett. Nowakowski. B 407 (1997) 219. Lett. Kayser. Nucl. [53] T. [69] K. White. Haber.Y. hep-ph/9901254. J. Nelson. Hu. Joshipura. Bhattacharyya. J. Phys. Bhattacharyya. Lett.

Davidson. Phys. A. M. Lett. . Y. Losada. King. Suppl. B 445 (1998) 191. J. Rev. Macenasu. G. Lett. A 52 (1997) 83. Davidson. [82] M. L. Bhattacharyya. Losada / Nuclear Physics B 585 (2000) 45–78 S. H.W. Abada. Phys. H. 78 (1997) 3438. Grossman. Phys. Ellis. S. C. Tampere.78 [76] [77] [78] [79] [80] [81] A.K. Abada.C. Finland. Lett. Nucl. hep-ph/9811498. Orr. hep-ph/9707435. O. Talk given at International Europhysics Conference on High-Energy Physics (EPSHEP 99). Phys. D 56 (1997) 4182. Rev. Haber. C. 1999. Dreiner. hep-ph/9709395. Kong. hep-ph/9906310. M. Jung. in preparation. Bisset. Proc. Phys. [83] S. July 15–21. B 39 (1997) 210.

PACS: 14. there have been numerous models of neutrino masses proposed in the literature. All rights reserved.edu (S. PII: S 0 5 5 0 . The essential ideas.60.nl/locate/npe A general classification of three-neutrino models and Ue3 S.elsevier. Models 1.Ff Keywords: Neutrinos.  2000 Elsevier Science B.4 . accepted 20 June 2000 Abstract A classification of models with three light neutrinos is given. and especially since the discovery of the large mixing of νµ seen in atmospheric neutrino experiments. attractive features. Oscillations. the parameter Ue3 will be very important in the future for distinguishing among different kinds of models and testing particular schemes. Introduction Over the years. One of the goals of this paper is to give a helpful classification of these models. Ilja Dorsner Bartol Research Institute.V. we present in Section 2 a general classification of threeneutrino models that have a hierarchical neutrino spectrum.udel. of which there are approximately one hundred. Such a classification is possible because in actuality there are only a few basic ideas that underlie the vast majority of published neutrino-mixing schemes.Nuclear Physics B 585 (2000) 79–104 www. In Section 3 we discuss the parameter Ue3 . as many as one hundred different models have been published. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. All rights reserved. 12. The classification is based principally on how large νµ −ντ mixing is obtained. This classification includes virtually all of the three-neutrino models proposed in the last few years. showing what values are to be expected for it in each type of model. Since theoretical models are constructed to account for the solar and atmospheric neutrino oscillation data. Barr).M. which tightly constrain the ‘1–2’ and ‘2–3’ mixings but not the ‘1–3’ mixing. USA Received 28 April 2000.M. Newark.bartol. Mixings. In the last two years alone.V.3 2 1 3 ( 0 0 ) 0 0 3 8 7 . DE 19716.15. After some preliminaries. and characteristic problems of the different classes of model are discussed. Masses. E-mail address: smbarr@bxclu. A general discussion of the mixing parameter Ue3 is then given. which describes the ‘1–3’ mixing of neutrinos.Pq. Barr. University of Delaware.

) (2) The atmospheric neutrino anomaly strongly implies that νµ is oscillating with nearly maximal angle into either ντ or a sterile neutrino. and (4) dark matter. We shall not focus on such models in our classification. The experimental reason is that KARMEN has failed to corroborate the discovery. Dorsner / Nuclear Physics B 585 (2000) 79–104 There are four indications of neutrino mass that have guided recent attemps to build models: (1) the solar-neutrino problem. In order to achieve then the small mass splittings needed to explain the solar and atmospheric anomalies one would have to assume that νe . Since the oscillations are caused by the small hierarchical piece.3−1 eV2 is regarded with more skepticism for two reasons. the large-angle MSW solution (LMA).) For these reasons. it is regarded as not very natural. There are many excellent reviews of the evidence for neutrino mass [1. (Again. to have a fourth sterile neutrino that is extremely light compared to the weak scale. These are the small-angle MSW solution (SMA). (There are some theoretical frameworks that can give light sterile particles. and δm2 ∼ 4.93. (3) The LSND result.0. Barr.6 × 10−5 eV2 . . we assume that the LSND results do not need to be explained by neutrino oscillations. with the data preferring the former possibility [5]. νµ . such models can be classified together with hierarchical models. and the vacuum-oscillation solution (VO). This significantly complicates the problem of modelbuilding. I. in most models with nearly degenerate masses. these are best-fit values from a recent analysis [4]. However. the neutrino-mass matrix consists of a dominant piece proportional to the identity matrix and a much smaller hierarchical piece.5 × 10−3 and δm2 ∼ 5.1 × 10−6 eV2 . If a significant amount of this dark matter is in neutrino mass.80 S. in general. which would indicate a mixing between νe and νµ with δm2 ∼ 0. All these solutions involve νe oscillating into some other type of neutrino — in the models we shall consider. and the classification we present includes only three-neutrino models. (The numbers are best-fit values from a recent analysis [3]. predominantly νµ .) And for the VO solution sin2 2θ ∼ 0. (3) the LSND experiment. νµ and ντ were nearly degenerate. the models we shall classify are those which assume (a) three flavors of neutrino that oscillate (νe . (1) The three most promising solutions to the solar-neutrino problem are based on neutrino mass. VO). although it is true that KARMEN has not excluded the whole LSND region. For the LMA solution one has sin2 2θ ∼ 0. it would imply a neutrino mass of order several eVs . but they tend to give many of them. (4) The fourth possible indication of neutrino mass is the existence of dark matter. not just one. In particular. and (c) the solar anomalies explained by νe oscillating primarily with νµ with either small angle (SMA) or large angle (LMA. and ντ ). In the SMA solution the mixing angle and mass-squared splitting between νe and the neutrino into which it oscillates are roughly sin2 2θ ∼ 5. which is primarily devoted to models with “hierarchical” neutrino masses. and δm2 ∼ 3.4 × 10−10 eV2 .2]. In sum. (2) the atmospheric-neutrino anomaly. and that can only be achieved with four species of neutrino.M.79. The theoretical reason is that to account for the LSND result and also for both the solar and atmospheric anomalies by neutrino oscillations would require three quite different mass-squared splittings. One has sin2 2θ ∼ 1. (b) the atmospheric anomaly explained by νµ −ντ oscillations with nearly maximal angle. and δm2 ∼ 3 × 10−3 eV2 .

U . There need not be three of them. Consequently. and are likely to shed great light on that problem. and perhaps even be the key to solving it.7]. the Dirac mass matrices of the quarks and charged leptons. And. In conventional see-saw models there are three right-handed neutrinos. one for each of the three families of quarks and leptons. and those in which the neutrino masses are generated directly at low energy. and (c) R-parityviolating terms in low-energy supersymmetry. Barr. the mass matrix of the left-handed neutrinos is given by Mν = −N T MR−1 N . there are only four mass matrices to consider. D. This field can couple to the lepton doublets Li as Li Lj h and to the Higgs doublets φa (if there is more than one) as φa φb h. U . L. (b) In the Zee model. Clearly it is not possible to assign a lepton number to h in such a way as to conserve it in both these terms. the Standard Model is supplemented with a scalar. Typically in such schemes Mν has nothing directly to do with the matrices U . After integrating out the superheavy right-handed neutrinos. I. In non-see-saw schemes. in the (1. and D. Consequently. (c) In supersymmetry the presence of such R-parity-violating terms in the superpotential as Li Lj Ekc and Qi Djc Lk . while MR is generally much larger than the weak scale. However. That means that in conventional see-saw models neutrino masses and mixings are just one aspect of the larger problem of quark and lepton masses. 1. and the Majorana mass matrix MR of the righthanded neutrinos. +1) representation and having weak-scale mass. In see-saw models. There are also what we shall refer to as unconventional see-saw models in which the fermions that play the role of the heavy right-handed partners of the neutrinos are not part of the ordinary three-family structure but are some other neutral fields. The four Dirac mass matrices are all roughly of the weak scale. h. Such unconventional see-saw models we classify together with non-see-saw models. (b) variants of the Zee model [8]. U . Dorsner / Nuclear Physics B 585 (2000) 79–104 81 There are several major divisions among models. and L. The freedom in MR is the major obstacle to making precise predictions of neutrino masses and mixings in most see-saw schemes. down quarks. respectively. in most see-saw models the Majorana matrix MR is either not related or is tenuously related to the Dirac mass matrices of the quarks and leptons. and allows one-loop diagrams . The three most popular possibilities in recent models for generating Mν at low energy in a non-see-saw way are (a) triplet Higgs. in such conventional see-saw models there is a close relationship between the 3 × 3 Dirac mass matrix N of the neutrinos and the other 3 × 3 Dirac mass matrices L. there are five fermion-mass matrices to explain: the four Dirac mass matrices. there are both leftand right-handed neutrinos. (a) In triplet-Higgs models. where HT is a Higgs field in the (1. and the Majorana mass matrix of the light left-handed neutrinos Mν . +1) representation of SU(3) × SU(2) × U (1). typically. The resulting lepton-number violation allows one-loop diagrams that generate a Majorana mass for the left-handed neutrinos. Mν arises from a renormalizable term of the form λij νi νj HT0 . and L. there are no right-handed neutrinos.S. causes lepton-number violation. and N of the up quarks. D. charged leptons. D.M. and the Dirac mass matrix of the neutrinos therefore need not be 3 × 3 nor need it have any particular connection to the other Dirac mass matrices L. but is generated at low-energy by completely different physics. U . 3. Usually the four Dirac matrices are related to each other by grand unification and/or flavor symmetries. and neutrinos. and D. One is between models in which the neutrino masses arise through the see-saw mechanism [6.

hfi i. On the other hand. for example by the appearance of Clebsch coefficients in certain entries of some of the mass matrices. while the righthanded charged leptons and left-handed down quarks are unified in a 10. In minimal SU(5) there is one relation among the Dirac mass matrices.M. While this allows more freedom to the neutrino masses. A drawback of such models compared to many grand unified models is that actual numerical predictions. as opposed to order of magnitude estimates. In SO(10). models based on flavor . It is clear that unified symmetries are so powerful that very predictive models are possible. different fermion multiplets can differ in charge under a U (1) flavor symmetry that is spontaneously broken by some “flavon” expectation value (or values). It is clear that in all of these schemes the couplings that give rise to neutrino masses have little to do with the physics that gives mass to the other quarks and leptons. This kind of scheme can explain small mass ratios and mixings in the sense of predicting them to arise at certain orders in the small quantities i . Another major division among models has to do with the kinds of symmetries that constrain the forms of mass matrices and that. Thus. Dorsner / Nuclear Physics B 585 (2000) 79–104 that give neutrino masses. where M is the scale of flavor physics. Models in which these operators do not arise from a see-saw or where their origin is left unspecified we classify as non-see-saw models. Non-abelian symmetries are useful for obtaining the equality of certain elements of the mass matrix. which will be discussed later. There are two main approaches: (a) grand unification. different elements of the fermion-mass matrices would be suppressed by different powers of hfi i/M ≡ i  1. Neutrino mass can also arise at tree level from R-parity-violating terms of the form Hu Li . Most of the published models which give sharp predictions for masses and mixings are unified models. In realistic models these relations are modified in various ways. Abelian symmetries are useful for explaining hierarchical mass matrices through the so-called Froggatt–Nielson mechanism [9]. are not possible. Barr. the minimal model gives the relations N = U ∝ D = L. which mix neutrino and Higgs superfields and lead to sneutrino vacuum expectation values. In particular. in some models. While such terms do arise in the conventional see-saw mechanism they can also arise in other ways. (b) Flavor symmetries can be either abelian or non-abelian. The idea is simply that different elements of the mass matrices arise at different orders in flavor symmetry breaking. relate different mass matrices to each other. In SU(5) there do not have to be right-handed neutrinos. though they may be introduced. as in models where the neutrino-masses are nearly degenerate. and (b) flavor symmetry. and in the so-called “flavor-democracy” schemes. Many models use both. I. it would from one point of view be very disappointing. (a) The simplest grand unified group is SU(5). coming from the fact that the left-handed charged leptons are unified with the right-handed down quarks in a 5. namely D = LT . which in several ways is a very attractive group for unification. It should also be mentioned that some models derive the neutrino-mass matrix Mν directly from non-renormalizable terms of the form νi νj Hu Hu /M without specifying where these terms come from.82 S. as it would mean that the observation of neutrino oscillations is almost irrelevant to the great question of the origin of quark and charged-lepton masses.

Unified models put more in but get more out than do abelian-flavor-symmetry models.7. and could therefore be considered more economical in a certain sense. In particular. (Especially noteworthy is the remarkably prophetic 1982 paper of Harvey. The first point of interest. [10] would predict Uµ3 to be only 0. The most significant new fact about neutrino mixing is the largeness of the mixing between νµ and ντ This comes as somewhat of a surprise from the point of view of both grand unification and flavor symmetry approaches. which is known to be 0. and that paper assumed it to be 25 GeV. The fact that the quark-mixing angles are small suggests that there is a family symmetry that is only weakly broken.12). It is true that even in the early 1980’s some 1/ third families of leptons: Uµ3 ∼ = = grand unified models predicted large neutrino-mixing angles. Ramond. in looking at any model of neutrino mixing is how it explains the large mixing of νµ and ντ . Of course this distinction only makes sense in some preferred basis. This mixing is described by the element Uµ3 ≡ sin θ23 of the so-called MNS matrix [11] (analogous to the CKM matrix for the quarks). Classification of three-neutrino models Virtually all three-neutrino models published in the last few years fit somewhere in the simple classification now to be described. 2. Barr. Since grand unification relates leptons to quarks. between models of what we shall call class I and class II. I. the model of Ref. one might expect lepton-mixing angles to be small like those of the quarks. The corresponding lepton mixing was further boosted by a Clebsch of 3. Dorsner / Nuclear Physics B 585 (2000) 79–104 83 symmetry involve less of a theoretical superstructure built on top of the Standard Model than do unified models. the mixing between the second and third family of quarks is given by Vcb . or L has large off-diagonal elements and Mν is nearly diagonal. there are two obvious ways of obtaining large θ23 : either Mν has large offdiagonal elements while L is nearly diagonal. The same puzzle exists in the context of flavor symmetry. (It is also possible that the large mixing is . in those days the top mass was expected to be light. The mixing angles of the neutrinos are the mismatch between the eigenstates of the neutrinos and those of the charged leptons. while the large mixings of some of the neutrinos would suggest that family symmetries are badly broken. which explicitly predicted and emphasized that there should be large νµ −ντ mixing. That gave Vcb to be about 0. therefore.04. But in almost every model there is some preferred basis given by the underlying symmetries of that model. and Reiss [10]. However. almost all of them are cited below. What makes the largeness of Uµ3 a puzzle in the present situation is the fact that we now know that both Vcb and mc /mt are exceedingly small. This will be the feature that we will use to organize the classification of models. Thus.S. This distinction gives the first major division in the classification. In fact. That is to be compared to the nearly maximal mixing of the second and √ ∼ 2 0. or in other words between the mass matrices L and Mν . With the actual value of mt that we now know.22. The main divisions of this classification are based on how the large νµ −ντ mixing arises.M.

we called class II. and right-handed for the quarks. 2. We therefore distinguish these cases as class I(1) and class I(2).M. C.) If the large θ23 is due to Mν (class I). We call this approach class II(2).84 S. I(1) large mixing from Mν . making Uµ3 large. II large mixing from L: (1) CKM small by cancellation. especially D (which is typically closely related to L). The main examples of this idea are the so-called “flavordemocracy models”. we call I(2C). so a further subdivision is possible: models in which the large mixing comes from large off-diagonal elements in MR we call I(2A). Vcb comes from Dirac mass matrices. models in which the large mixing comes from large off-diagonal elements in N we call I(2B). B. and models in which neither MR nor N have large off-diagonal elements but Mν = −N T MR−1 N nevertheless does. The other main class of models. where θ23 is due to large off-diagonal elements in L the mass matrix of the charged leptons. The other idea is that the matrices L and D T (related by unified or flavor symmetry) are “lopsided” in such a way that the large off-diagonal elements only affect the mixing of fermions of one handedness: left-handed for the leptons. One way involves the CKM angles being small due to a cancellation between large angles that are nearly equal in the up and down-quark sectors. respectively. giving examples from the literature. then it becomes important whether Mν arises from a non-see-saw mechanism or the see-saw mechanism. Barr. but this possibility seems to be realized in very few published models. there are not also large off-diagonal elements in the Dirac mass matrices of the other charged fermions. I (1) Now let us examine the different categories in more detail. Schematically. large mixing from MR . In the literature there seem to be two ways of answering this question. Dorsner / Nuclear Physics B 585 (2000) 79–104 due almost equally to large off-diagonal elements in L and Mν . which are all presumably nearly diagonal like L. given that L has large off-diagonal elements. In the see-saw models. We call this class II(1). whereas Uµ3 comes from the matrix Mν . We will put them into class II. causing large CKM mixing of the quarks. (2) lopsided L. leaving Vcb small. and since in non-see-saw models Mν comes from completely different physics than do the Dirac mass matrices it is not at all surprising that it has a very different form from the . I. large mixing from −N T MR−1 N.1. one then has: large mixing from Mν : (1) non-see-saw. large mixing from N. (2) see-saw: A. Mν is given by −N T MR−1 N . The question in these models is why. non-see-saw This kind of model gives a natural explanation of the discrepancy between the largeness of Uµ3 = sin θ23 and the smallness of Vcb .

and from R-parity and lepton-number-violating terms in supersymmetry in Refs. [33] the opposite is assumed. δM there is a hierarchy of mass eigenvalues. [12–16]. In most schemes of “flavor democracy”. The value of the solar angle depends on the . as we shall see later. containing some large off-diagonal elements. Thus. namely why the mass splitting seen in solar-neutrino oscillations (δm212 ) is much smaller than that seen in atmospheric oscillations (δm223 ). and there are published models of neutrino mixing corresponding to all these ideas [12–49]. since it comes from new physics unrelated to the origin of the other mass matrices. Dorsner / Nuclear Physics B 585 (2000) 79–104 85 others. there are few definite predictions. for masses and mixings in such schemes. c ≡ cos θ . in general. If all the elements of Mν were of the same order. and mij  M.M. but in Ref. from the Zee mechanism in Refs. [17–21].) Several other models in class I(1) exist in the literature [34–45]. On the other hand the atmospheric-neutrino angle θ23 . is of order one. While this basic idea is very simple and appealing. δ  1. there are a variety of attractive ideas for generating a non-see-saw Mν at low energy. [22–32]. then the question arises why the mixing angles are not also controlled by that small parameter. By a rotation in the 2–3 plane by an angle close to θ . Barr. in some schemes constraints can be put on the new physics responsible for Mν . (a) In the literature one finds that the majority of models of class I(1) (and.S. is highly unconstrained. these models have the drawback that in non-see-saw models the form of Mν . many models of other classes too) assume the following form for Mν :   m11 m12 m13 (2) Mν =  m12 s 2 M + O(δ)M scM + O(δ)M  . and that δm223 ∼ = M2 2 2 2 and δm12 = O(δ )M . and one would expect δm212 ∼ δm223 . the 2–3 block will be diagonalized and the matrix will take the form   m11 cm12 − sm13 cm13 + sm12 . giving models of class II(1). then indeed large mixing angles would be typical. We shall consider them in turn. I. In Ref. As we saw. if there is a small parameter in Mν that accounts for the ratios of mass splittings. as desired to explain the atmospheric neutrino oscillations. but the neutrino-mass ratios would then also be typically of order unity. A satisfactory answer to these questions requires that Mν have a special form. as has been pointed out in several analyses [46–49]. Three satisfactory forms are possible. (The democratic form is one in which all the elements of the matrix are equal or very nearly equal. However. (3) Mν0 ∼ O(δ)M 0 =  cm12 − sm13 0 M cm13 + sm12 It is clear that for mij . [33] a “democratic form” of Mν is enforced by a family symmetry. which is approximately given by θ . it is the charged leptonmass matrix L that is assumed to have a democratic form and Mν is assumed approximately diagonal. There is a basic question that has to be answered by any model of class I(1). Mν comes from triplet Higgs in Refs. θ ∼ π/4. as we shall see later. Conversely. m13 scM + O(δ)M c2 M + O(δ)M where s ≡ sin θ .

giving (Mν )ij ∼ = λi λj M. (2) in a technically natural way using flavor symmetries. where λi is an R-parity-violating and lepton-number-violating coupling of νi to a quark–squark (or lepton–slepton) pair in supersymmetry. in which an S2 × S2 permutation symmetry among four left-handed neutrinos is used to obtain the form   A B C D  B A D C   (4) Mν =   C D A B . A factorized form of Mν can also arise at tree-level by a nonstandard see-saw mechanism in which λi is a Dirac coupling of νi to a single heavy Majorana fermion that is integrated out. two ways of achieving the special form in Eq. since the part proportional to the identity does not contribute to oscillations. In particular θ12 ∼ (cm12 − sm13 )/(δM). [12]. (2) results. . One sees that in order to get the hierarchy among neutrino-mass splittings and at the same time a large atmospheric angle one has assumed a form for Mν in Eq. (2) is permutation symmetry. (i) The idea of factorization is that the neutrino-mass matrix arises from one-loop contributions that are dominated by a single diagram. Dorsner / Nuclear Physics B 585 (2000) 79–104 size of mij . Specifically. Again. and (ii) permutation symmetry. and factorized forms of Mν can result. Barr. F B A F B −B which in effect has the form in Eq. (2). D C B A √ Then by assuming that the linear combination of (ν1 − ν2 )/ 2 acquires a superlarge Majorana mass. [22]. A good example of this kind of model is Ref. [40–45]. the 2–3 block of Mν would have a determinant that was of O(δ) times its “natural” value. However. (2) in a technically natural way in models of class I(1) have been proposed in the literature: (i) factorization. 32. thus giving maximal mixing angles.) (ii) The other idea for achieving the form in Eq. the residual three light species of neutrino end up with a mass matrix     A+B F F 0 F F Mν0 =  F (5) A B  = (A + B)I +  F −B B  . The basic idea is to use non-abelian symmetry to relate different elements of Mν . either small-angle or large-angle solutions of the solar-neutrino problem can be naturally obtained. I. where λi is the coupling of νi to the particles in the loop. then the model would be “fine-tuned” to some extent. If such a relationship existed simply accidentally. either small or large solar angle can arise depending on the magnitude of F /B. If λ2 ∼ λ3 then the form in Eq. 23.M. Generally the relationship will be one of equality.86 S. From this one sees that it is possible to get the form in Eq. A good example is the model of Ref. This is the basic idea in the papers in Refs. Consequently. In these the neutralinos play the role of the heavy fermions in the see-saw. A number of the models in the literature that are of class I(1) are indeed fine-tuned in this way. and 33 elements. (A special case of this is supersymmetric models with R-parity-violating terms that mix neutrinos with other neutralinos. (2) that has a special relationship among the 22.

sM m23 m33 (6) where mij  M. [17]. Dorsner / Nuclear Physics B 585 (2000) 79–104 87 (b) Another form for Mν that is satisfactory is   m11 cM sM Mν =  cM m22 m23  . Barr. while the third neutrino will have smaller mass. If this is not to be a fine-tuning of parameters. (c) A third possible form for Mν is   M 0 m12 m13 Mν =  m12 m22 M  .20. A particularly interesting possibility [14–16. which if exact would allow only the large elements of order M in Eq. i. (6) is found in Ref. .S.e. large mixing angle for both atmospheric neutrinos and solar neutrinos. Such a form always gives bimaximal mixing.M. By a rotation in the 2–3 plane by angle θ one brings this to the form   m11 M 0 Mν0 =  M m022 m023  . it must be that M 0 ∼ = M to great accuracy. (2) which can give either large or small-angle solutions for the solar-neutrino problem. 1 0 −m/ 2 where 1  m. Thus δm223 ∼ = M 2 and δm212 ∼ mij M. In such a scheme. I. 21]. in order for the splitting δm212 to be even smaller. with θ ∼ π/4. The form in Eq. (6). νµ and ντ are nearly maximally mixed and nearly degenerate. This model is based on the Zee mechanism. There is some mild fine-tuning in this model in the sense that in order for the 12 and 13 elements of Mν to be nearly equal in magnitude (as must be so to have nearly maximal atmospheric angle) a relation among the couplings and masses of the Zee model must be satisfied that has no basis in symmetry. In particular it can give a matrix of the form  √ √  0√ m/ 2 −m/ 2   (8) Mν ∼ 0 1 =  m/ √2 . m13 M m33 (9) where mij  M. Therefore.21] is that the symmetry in question is Le − Lµ − Lτ . then it must be the consequence of some non-abelian flavor symmetry.20. 0 m023 m033 (7) This pseudo-Dirac form in the 1–2 block shows that ν1 and ν2 will be maximally mixed with nearly degenerate masses approximately equal to M. which gives a neutrinomass matrix Mν that is symmetric but has vanishing diagonal elements. (6) can easily be achieved using various family symmetries [13–16. with δm223 ∼ mij M  M 2 . An interesting and instructive model in which Mν is of the form given in Eq. in contrast to Eq. and as before s ≡ sin θ and c ≡ cos θ .

then obviously Mν is approximately diagonal and hierarchical. Another way of getting around the Dirac–Majorana conspiracy problem is to assume a special form for MR . . I(2A) see-saw Mν . The problem is that if N has a hierarchical and nearly diagonal form.88 S. Barr. For example. while the lepton mixings involve the Majorana matrix MR . the large atmospheric neutrino-mixing angle comes from Mν . A number of models in the literature exploit this idea [59–64]. An apparently simple solution is to take the 2–3 block of MR to be skew diagonal. suppose   0   0 M 0 0  . where N is a 3 × 3 matrix typically related by symmetry to L. and D. as needed to explain the atmospheric neutrino anomaly. as in class I(1). The quark mixings all come from Dirac matrices. The trouble is that to have such a relationship between the magnitudes of parameters in N and MR is usually unnatural. As with the models of class I(1). and D. with 1     0 . Mν therefore has the form −N T MR−1 N . the large off-diagonal elements in Mν are assumed to come from MR . (11) N∼ MR ∼  =  0 0 M . these models have the virtue of explaining in a natural way the difference between the lepton angle Uµ3 and the quark angle Vcb .2. it tends to communicate this property to Mν . large mixing from MR In models of class of I(2). Dorsner / Nuclear Physics B 585 (2000) 79–104 2.  0 a13 a23 a33 If all the nonvanishing elements aij are of the same order of magnitude. U . Then the matrix Mν will have the form  02   a11  0 a12  0 a13 (10) Mν ∝   0 a12  2 a22 a23  . and one has that νµ and ντ are nearly degenerate and maximally mixed. since these matrices have very different origins.M. We shall call it the Dirac–Majorana conspiracy problem. which it is quite reasonable to suppose might have a very different character. Many examples of models of class I(2A) exist in the literature [50–70]. 1)M. For example. suppose we take N = diag( 0 . while the Dirac neutrino matrix N is assumed to be nearly diagonal and hierarchical like the other Dirac matrices L. there is a general problem with models of this type. Then the 2–3 block of Mν = −N T MR−1 N is also approximately skew diagonal. The contribution to the leptonic angles coming from Mν would therefore typically be proportional to the small parameters  and  0 . This problem has been pointed out by various authors [53–55]. Example of such correlated hierarchies can be found in Refs. which however is now assumed to arise from the conventional see-saw mechanism. And suppose that the ij -th element of MR−1 is called aij . U . which not all the examples in the literature convincingly overcome. . I. = 1 0 M 0 where  0    1. This problem is avoided in models in which the hierarchies in N and MR are controlled by the same family symmetries and the same small parameters. One way that a θ23 of O(1) could arise is that the small parameter coming from N gets cancelled by a correspondingly large parameter from MR−1 [50–52]. However. [56–58]. In class I(2A). with large off-diagonal elements.

and x c are all of the same order. the Dirac and Majorana matrices of the neutrinos have the forms     2 0 0 A 0 0 x y MR =  0 1 0  mR . 0 2 z 0 2x In the 1–2 block this matrix has a Dirac form. x 0 1 0 0 M 0−1 where x  1. In the model of Ref. c ∼ 1. (13) is that it gives bimaximal mixing. Note that the matrix in Eq. . x b a +  (16) M 1 Here  ≡ M 0 2  1. giving nearly maximal mixing of νe . If one computes Mν = −N T MR−1 N one finds that   0 O(x 4 y/A) x 2 y/A (13) Mν = −  O(x 4 y/A) x2 x 2  m2D /mR . An interesting point about the form of Mν in Eq. This is easily seen by doing a rotation of π/4 in the 2–3 plane. N = 0 2 A 0 0 0 O(x ) 1 where x and y are small parameters. b. This example is generalized in the papers [67. Another interesting model that avoids the Dirac–Majorana conspiracy problem. bringing the matrix to the form   0 z z0 (14) Mν0 =  z 0 0  . for instance. 2 2 2 x y/A x x Observe that this gives a maximal mixing of the second and third families.S. Therefore. but requires a mild fine-tuning to get the hierarchy among neutrino-mass splittings. One has m2 ∼ = m3 ∼ = M and m1 ∼ =  02 M 0 . It is clear that if one seeks to avoid the Dirac–Majorana conspiracy problem and also to explain both solar and atmospheric neutrino oscillations. Barr. (12) x x  mD . Dorsner / Nuclear Physics B 585 (2000) 79–104 89 Unfortunately. MR−1 =  bM −1 cM −1 0 . This gives  2  x 0 x  Mν =  0 c b  m2 x 2 /M . The problem is that it is unnatural in such a scheme for the splitting δm212 to be smaller than δm223 . b. [66].68]. The Majorana and Dirac neutrino-mass matrices in that model have the form     aM −1 bM −1 0 0 0 x (15) N =  0 x 0  m. an even cleverer choice of the forms of N and MR must be found. (2). and a. [69]. this idea has a problem with solar neutrinos. I. The atmospheric-neutrino angle will be of order unity if a. M/M 0  x 2 . is given in Ref. and y) and the parameter in MR (namely A). (13) is of the general form given in Eq.M. but here it arises through the see-saw mechanism. as most of the papers [59–64] noted. which requires no fine-tuning or Dirac–Majorana conspiracy. Several papers have found such forms [65–69]. unless M 0 is tuned with great accuracy this scheme cannot give a satisfactory solution to the solarneutrino problem. without having to assume any special relationship between the small parameters in N (namely x.

large mixing from −N T MR−1 N In order for the see-saw mass matrix Mν = −N T MR−1 N to have large off-diagonal elements it is not necessary that either MR or N have large off-diagonal elements. In that model all of the quark and leptonmass matrices are given (in terms of relatively few parameters) by linear combinations of certain matrices that are hierarchical and nearly diagonal. Barr. [73] the difference between N and the other Dirac matrices is accounted for by a fine-tuning. U . this seems to be a less natural approach. The point is that if the large θ23 is due to large off-diagonal elements in N . In order that N have offdiagonal elements that are comparable to its diagonal elements. b. 2. [71] only attempts to describe the lepton sector and so does not resolve this problem. r1  02 . Models of class I(2A) can be constructed that predict either small or large values of the solar-neutrino angle θ12 . one right-handed neutrino dominates MR−1 . 0   1 0 0 r3 where it is assumed that r2  2  r3 . and L.M. In the interesting model of Ref. [76] (see also [77]). 1 1  0 / 0 It is easy to understand what is happening. to make δm212  δm223 requires that the condition ac − b2  a. giving large CKM angles. consider the matrices    0 0 0  r1 0 0    MR−1 =  0 r2 0  M −1 . Dorsner / Nuclear Physics B 585 (2000) 79–104 √ However. an accidental cancellation must occur that suppresses the diagonal elements.75]. In Ref.4. As a consequence one . I(2C) see-saw Mν . experience seems to show that this approach is harder to make work than the others.77]. large mixing from N We now turn to see-saw models in which the large atmospheric-neutrino angle comes from large off-diagonal elements in the Dirac neutrino-mass matrix N rather than in the Majorana matrix MR . There are ways to construct models of class I(2B) in which the difference between N and the other Dirac matrices is explained without fine-tuning [74. 2. I(2B) see-saw Mν . c.90 S. (17) N ∼   0    m. be satisfied. [72] it is assumed that N has large offdiagonal elements and L does not. Effectively. At least at first glance. The fact that r2 is much larger than r1 and r3 means that the right-handed neutrino of the second family is much lighter than the other two. would also have large off-diagonal elements. Then to leading order in small quantities Mν has the form  0 2 0  ( /)  /  0 /  (18) Mν ∼   0 / 1 1  r2  2 m2 /M . I. it might be expected that the other Dirac mass matrices. but the difference in character of these matrices is not accounted for. which does not arise from any symmetry. [76. then. as emphasized in Ref. D. However. Following Refs.3. The model in Ref.

However. 1 1 1 A Dirac mass matrix can have such a form as the result of separate S3 permutation symmetries acting on the left-handed and right-handed fermions. However. and if so why this does not lead to large CKM angles for the quarks. the large rotation angles nearly cancel between the up and down sectors. This difference in form is plausible. √ √ 0 −2/ 6 1/ 3 The reason why the CKM angles are small in flavor-democracy models is that both U and D are assumed to have approximately the democratic form. Thus. with Mν being nearly diagonal. The issue that arises in such models is whether the other Dirac mass matrices. have nearly identical forms. Of course. In fact. in which a single right-handed fermion also dominated. especially D and U . the matrix is  √ √ √  1/ √2 1/√6 1/√3   (20) UFD =  −1/ 2 1/ 6 1/ 3  . it would seem that under most reasonable assumptions the matrix L should have some relationship to U and D. Dorsner / Nuclear Physics B 585 (2000) 79–104 91 obtains an approximately “factorized” form for Mν . would imply that U and D. Barr. That. I. Some published models do not deal with this question since they are only models of the lepton sector and do not attempt to describe the quarks at all [78. and thus gives only one family a mass. which give rise to masses for the lighter families. even though highly nondiagonal. in realistic models based on flavor democracy it is assumed that the mass matrices also get small corrections that come from the breaking of the permutation symmetries. just as one did in the unconventional see-saw models considered in the papers [40–45]. It is clear that the flavor-democratic form is diagonalized by a unitary matrix with rotation angles that are large. [81].M. while it may in non-see-saw models make sense to discuss Mν apart from the other mass matrices. in turn. Why. 2. II(1) large mixing from L.79]. it is possible to have large neutrino-mixing angles if Mν is assumed to have not the democratic form but a nearly diagonal form. CKM small by cancellation We now turn to those models in which the large value of θ23 comes predominantly from the charged-lepton-mass matrix L. This form is of rank 1. This idea is realized in most of so-called “flavor democracy” models (for a review see [80]). then. also have large off-diagonal elements. Those unconventional see-saw models could also be considered as examples of class I(2C). are the CKM angles small? One possibility is that the CKM angles are small because of an almost exact cancellation between large angles needed to diagonalize U and D. given . A “flavor-democratic” mass matrix is one in which all the elements are equal:   1 1 1 (19) MFD ≡  1 1 1  m` . as first noted in Ref.5.S.

Barr. In some. therefore. In Ref. It is in this sense a narrower approach to the problem of fermion masses than some of the others we have discussed. Then the 1–2 angles from the diagonalization of both L and Mν will be close to π/4 and their difference can be small. I. that the small permutationsymmetry-breaking contributions to the mass matrices can lead to additional large mixings. it is possible for θ12 to be small. the angles required to diagonalize Mν would be small. There are many possibilities. but the splittings can be made hierarchical to accomodate the solar and atmospheric data. . And if Mν were exactly proportional to the identity matrix. the smallness of the quark angles.M. (19). given the matrix comes out to be very close to UFD degeneracies the the exact permutation-symmetric forms give. However. However. most published versions of the flavor-democracy idea 1 [82–89] assume that Mν is approximately proportional to the identity matrix. Dorsner / Nuclear Physics B 585 (2000) 79–104 that Mν is a Majorana matrix rather than a Dirac matrix like the others. it should be noted that Mν ∝ I is not the most general form consistent with permutation symmetry. There exists an extensive and growing literature in this area. and so to make this form technically natural some further symmetries must be invoked. [81]. is large. This can be close to π/4 as in the matrix UFD happen if the matrix Mν is such that the neutrinos ν1 and ν2 form a pseudo-Dirac pair. In such versions.) Small deviations from the identity matrix would arise from terms that break the flavor symmetries of the model. but allowed to differ in complex phase. it is assumed that Mν is diagonal and hierarchical in form. but on forms in which all the elements of the mass matrix are assumed to be approximately equal in magnitude. it is based on very special forms for the mass matrices which come from very specific symmetries. on the form of the small contributions to the mass matrices that break the permutation symmetries. it is not surprising. and the elements of Mν are assumed to arise entirely from the breaking of the permutation symmetries of the model. and in many cases it comes out to † . An exact flavor-democratic form of L would leave two charged leptons degenerate and therefore one of the neutrino-mixing angles undefined. based on different ways to break permutation symmetry. especially in that it uses one basic idea to explain the largeness of the leptonic angles. † . and the fact that one family is much heavier than the others. [81] therefore has the neutrino masses being hierarchical. all three neutrinos would be degenerate and all three neutrino-mixing angles would be undefined. the MNS † . However.92 S. There are also many models based not on the pure flavor-democratic form in Eq. The model of Ref. and the MNS matrix would come predominantly from diagonalizing L. The form Mν ∝ I is invariant under an S3 permutation of the left-handed neutrinos. In this case. (However. It is typical in flavorand to forms for the MNS matrix that depart significantly from UFD democracy models for the angle θ12 to come out large. The idea of flavor democracy is an elegant one. Exactly what angles are predicted for the neutrinos depends. This is sometimes called the “universal strength for Yukawa couplings” approach or USY [91–93]. 1 For an excellent review of flavor-democracy schemes of neutrino mass mixing see [90]. the three neutrino masses are nearly degenerate. The number of possible models. On the other hand.

in which the CKM angles are small by cancellations of large angles. while σ ∼ 1. In the case of the charged leptons  controls the mixing of the second and third families of right-handed fermions (which is not observable at low energies). II(2) large mixing from “lopsided” L We now come to an idea for explaining the largeness of θ23 that has great flexibility. The large θ23 arises from L while the neutrino-mass spectrum arises from Mν . as can be seen from Refs. (b) these off-diagonal elements appearing in a highly asymmetric or lopsided way.6. while the large O(σ ) mixing is in the right-handed sector. (21) L ∼  0 0   mD . The lopsided L idea next is seen in three papers that appeared almost simultaneously: [96–99]. by contrast. The basic idea of the “lopsided” L approach is that the charged-lepton and down-quark-mass matrices have the approximate forms     0 0 0 0 0 0 D ∼  0 0 σ  mD . The assumption is that   1. What makes this approach so flexible is that the problem of obtaining a realistic pattern of neutrino masses is decoupled from the problem of getting large θ23 . Thus one is freed from having very special textures for Mν as was the case in class I models. [94–114]. It is interesting that the same basic mechanism of lopsided L was arrived at . Dorsner / Nuclear Physics B 585 (2000) 79–104 93 It would be interesting to know whether simple models of class II(1). [95]. In this approach the three crucial elements are these: (a) large mixing of neutrinos (in particular of νµ and ντ ) caused by large off-diagonal elements in the charged-lepton mass matrix L. I. in the sense that it can be implemented in many different kinds of models: grand unified.S. 0  1 0 σ 1 The “∼” sign is used because in realistic models these σ and  entries could have additional factors of order unity. For the quarks the reverse is the case because of the “SU(5)” feature: the small O() mixing is in the left-handed sector. can be achieved in many ways. The fact that L is related closely in form to the transpose of D is a very natural feature from the point of view of SU(5) or related symmetries. where it cannot be observed and does no harm. Barr. there the necessity of near cancellation between upand down-quark angles forced a very particular kind of mass matrix texture and flavor symmetry. models with abelian or non-abelian flavor symmetries. The lopsided mass matrices. and so on. such as from Clebsch coefficients. The ideas of that paper were further explored in Ref. which contributes to θ23 and makes it large. can be constructed using ideas other than flavor democracy.M. accounting for the smallness of Vcb . This is also true of the flavor-democracy schemes. The first paper that has all three elements that define this approach seems to be Ref. see-saw or non-see-saw neutrino masses. and (c) L being similar to the transpose of D by SU(5) or a related symmetry. however. [94]. while σ controls the mixing of the second and third families of left-handed fermions. and is a crucial ingredient in this approach. 2. which proposed a very specific idea for generating the fermion-mass hierarchies.

In Refs. rather it emerged from the attempt to find a realistic model of the masses of the charged leptons and quarks in the context of SO(10). (21) above plays three crucial roles in this model that have nothing to do with neutrino mixing: (1) it is required to get the Georgi–Jarlskog [117] factor of 3 between mµ and ms . in both it is the SU(5) subgroup that plays the critical role in relating L to D T . [96] the model is based on E7 /SU(5) × U (1). I. In Ref.M. The reason this lopsided form was built into the model of Refs. the fact that L is related to D T follows ultimately from the requirement of anomaly cancellation for the various U (1) flavor symmetries of the model. the mass matrix entry that is denoted σ in Eq. 2. [96]. Rather. and the structure of the mass matrices is determined by the Froggatt–Nielson mechanism. [113. However. like that of Ref. was that it was necessary to account for certain well-known features of the mass spectrum of the quarks. and non-abelian discrete — are used in these models to constrain the forms of mass matrices. models based on lopsided L can give either large-angle or small-angle solutions to the solar-neutrino problem. The reason for the large mixing of νµ and ντ in this model is precisely the fact that the charged-lepton-mass matrix has a lopsided form.6.94 S. it is well known that anomaly cancellation typically enforces charge assignments that can be embedded in unified groups. A remarkable fact about this model is that it was not constructed to explain neutrino phenomenology. A predictive model with lopsided L We shall now briefly describe a particular model of class II(2). but is not based on a grand unified group. in which the implications for neutrino phenomenology were stressed [97]. Only subsequently was it noticed that the model actually predicts a large mixing of νµ with ντ and this led to a second paper. it turns out not only to perform these . non-abelian continuous.1. The model of Refs.99].116]. it could be said to be “SU(5)-like”. Remarkably. [97. The idea of the model was to take the Higgs sector of SO(10) to be as minimal as possible. In Refs. [96. Dorsner / Nuclear Physics B 585 (2000) 79–104 independently by these three groups of authors from completely different starting points. Rather. In particular. (2) it explains the value of Vcb . and does not use the Froggatt–Nielson approach. Barr. A variety of symmetries — abelian. Though both Refs. the constraints on the form of the mass matrices come from assuming a “minimal” set of Higgs for SO(10)breaking and choosing the smallest and simplest set of Yukawa operators that can give realistic mass matrices for the quarks and charged leptons.97] assume a unified symmetry larger than SU(5). [98. In fact. As pointed out in Ref. in the first paper proposing this model [116] no attention was paid to the neutrino mixings at all. and then to find what this implied for the mass matrices of the quarks and charged leptons.114] are papers that are not unified and do not discuss the quark-mass matrices. so that the third element of the approach (L being related to DT by a symmetry related to SU(5)) is not explicitly present. In Ref. [97] the model is based on SO(10). uses the Froggatt–Nielson idea. it is one of the most predictive models of quark and lepton masses that exists in the literature. [100–112] are listed numerous papers that have used the lopsided L approach in the context of grand unified theories. (3) it explains why mc /mt  ms /mb . In fact. So that even though the model does not contain an explicit SU(5). [115].

a pair of vectors (10). These gave the following mass matrices:     η 0 0 0 δ δ0 1  mU . a structure emerged that had just six effective Yukawa terms (just five if mu is allowed to vanish). The patterns that are evident in these matrices are due to = |δ | ∼ the SO(10) group-theoretical characteristics of the various Yukawa terms. with natural doublet–triplet splitting [119. Third.  ∼ = 0. then instead of the Georgi–Jarlskog factor of 3. . δ0 σ +  1 0  1 (The first papers [97.) Here σ ∼ = 1. by having σ ∼ 1 one ends up with Vcb and ms /mb being of the same order (O()) as is indeed observed. The adjoint. and some singlets. Dorsner / Nuclear Physics B 585 (2000) 79–104 95 three tasks. D0 =  δ U0 =  0 0 0 σ + 13   mD 3 1 0 1 0 −3 1 δ − 13  (22)    0  η 0 0 0 δ δ N 0 =  0 0 −  mU . Not often are four birds killed with one stone. 0 −5 δ∼ = 0. First. In constructing the model. Fourth. with this structure. the Clebsch of 13 that appears in D due to the generator B−L gets squared in computing ms . the ratio mµ /ms would be given by 9. it was required that the Georgi–Jarlskog factor arise in a simple and natural way. for several reasons that have nothing to do with neutrinos one is led naturally to exactly the lopsided form that is found to give an elegant explanation of the mixing seen in atmospheric-neutrino data. Third. L0 =  δ 0 −  mD .M. Thus. (That is. And since σ does not appear in U (for grouptheoretical reasons) the ratio mc /mt comes out much smaller. It has been shown [118] that the smallest set of Higgs that will allow a realistic breaking of SO(10) down to SU(3) × SU(2) × U (1).008. in order to give the doublet–triplet splitting. the mass of charm is predicted correctly to within the level of the uncertainties. while this was extended to the first family in a subsequent paper [121]. Second. it was required that the qualitative features of the quark and lepton spectrum should not arise by artificial cancellations or numerical accidents. Having imposed these conditions of economy and naturalness. of O( 2 ).116] gave only the structures of the second and third families. In fact.14. if σ were not present. Barr. as far as the two heavier families are concerned. several considerations played a part. η ∼ = 0. First.6 × 10 . This fact is an important constraint. and Vcb would vanish. Second.120]. it was required that the entries in the mass matrices should come from operators of low dimension that arise in simple ways from integrating out small representations of fermions.S. but also gives mixing of order 1 between νµ and ντ .) Since the large entry σ overpowers the small entries of order . must have a VEV proportional to the SO(10) generator B−L. U and D would be proportional. the correct Georgi–Jarlskog factor emerges. Notice several facts about the crucial parameter σ that is responsible for the lopsidedness of L and D. I. also as observed. consists of a single adjoint (45). two pairs of spinors (16 + 16 ).8. if σ were not present. a “minimal” set of Higgs for SO(10) was assumed.

In this table θSA and θLA stand for the value of θ12 in the small-angle and large-angle solutions of the solar-neutrino problem. I. and the mixing angles Ue2 Ue3 . If the solar angle is small. In this section we examine the other mixing. or by one√of the large-angle solutions (large-angle MSW or vacuum oscillation). Some of these are postdictions (including the highly nontrivial one for mc ). it is not surprising that many precise predictions result. A special case of00 this occurs in some flavor-democracy models.2. In fact there are altogether nine predictions [121]. which we label with the Greek letters α through δ. in which Ue2 ' 0. Dorsner / Nuclear Physics B 585 (2000) 79–104 From the very small number of Yukawa terms. what we find is that the great majority of viable models give one of four mixing patterns. the small-angle MSW solution. These patterns are indicated in Table 1. and they all aim to explain the solar-neutrino problem either by the small-angle MSW solution. one typically finds one of two results for Ue3 : either it is given by the relation Ue3 ∼ = Ue2 tan θatm (pattern γ ) or it is small compared to that value (pattern δ). and a priori could take values ranging from zero up to the present limit of about 0. respectively. and we call this α . Expectations for the parameter Ue3 All of the models that we have discussed aim to explain the atmospheric neutrino √ anomaly by saying that there is maximal mixing between νµ and ντ . However.05 p me /mµ sin θatm p √2 me /mµ β sin θLA 0 γ γ0 sin θSA p me /mµ cos θatm sin θSA tan θatm p me /mµ sin θatm δ sin θSA  sin θSA tan θatm α sin θLA α0 sin θLA 6 .e. i. some models predict it sharply. i. In most models that fit pattern α only the order of magnitude of Ue3 is predicted.e. Barr. But several predictions will allow the model to be tested in the future. and not merely order of magnitude estimates for them. including predictions for Vub .05 (pattern α) or that it will vanish (pattern β). where sin θatm = 2/ 6. A particularly interesting p prediction that arises in certain types of models is that |Ue3 | = me /mµ sin θatm . that Uµ3 ∼ = 1/ 2.05. In certain models with pattern γ there is the Table 1 Pattern Ue2 Ue3 α 00 sin θLA p O(mνµ /mντ ). Ue3 is independent of the other two mixings. which is described by Ue3 . O( me /mµ ) ∼ 0.96 S. What one sees is that if the solar angle is maximal one expects either that Ue3 will be of order 0. which 0 we distinguish with the name √ α . However. in which Ue2 ' 1/ 2.M. and from the fact that SO(10) symmetry gives the normalizations of these terms. 3.

Eq. and it leads to the same prediction for Ue3 that one has in pattern α 0 . i. for the practically interesting cases these expressions can be greatly simplified due to the smallness of certain angles. (23) L† L.M. From Eq. but s 12 is small in almost all models (except the flavor- . I. and Uν is where UL is the unitary matrix that diagonalizes diagonalizes Mν† Mν . Dorsner / Nuclear Physics B 585 (2000) 79–104 97 p further prediction for the solar angle that Ue2 ∼ = me /mµ cos θatm . The same is true in virtually all models for the charged-lepton sector. In other words. (25) tells us that Ue2 = c13 c12 c˜13 s˜12 − s 12 c23 c˜12 + s 12 s23 s˜13 s˜12 + s 13 c12 s23 c˜12 + s 13 c12 c23 s˜13 s˜12 . Barr. (26) As we shall now see.e. or “MNS matrix” has the form UMNS = UL† Uν . c ∼ 1. where m2 and m3 are the second and third eigenvalues of Mν (m3 ∼ = M and m2 = O(δ)M). Uµ3 = c12 s23 c˜13 + c13 s 12 s˜13 − s 13 s 12 c23 c˜13 . if we assume all quantities are real the MNS matrix can be written:     c13 c12 −s 12 −s 13 c12 c˜13 s˜12 s˜13 1 0 0 c˜13 c˜12 UMNS =  −c13 s 12 c12 −s 13 s 12  0 c23 s23  −˜s12 c˜12 0 . First let us consider the form in Eq. the angles θ ij and θ˜ij are given with sufficient accuracy very simply in terms of ratios of elements of the mass matrices. unless there is fine-tuning. Consequently. and most of the mass matrices that we shall have to deal with. s 13 ∼ (mµ /mτ )s 12  s 12 . 1 (24) where s ij ≡ sin θ ij .S. s 13 0 c13 0 −s23 c23 −˜s13 c˜12 −˜s13 s˜12 c˜13 (25) where s23 ≡ sin(θ˜23 − θ 23 ). One can write Uν in a similar way. while a tilde means that it comes from the neutrino sector. It is convenient to write UL in the form    1 0 0 c13 0 s 13 c12 s 12 UL =  0 c23 s 23   0 1 0   −s 12 c12 0 −s 23 c23 −s 13 0 c13 0 0 the unitary matrix that  0 0 . (3) one sees that s˜13 ∼ = ∼ (cm13 + sm12 )/m3 and s˜12 = (cm12 − sm13 )/m2 . and so on. showing why the four patterns we have described are the ones that arise in the great majority of published models. s˜12 may be large or small depending on which solution to the solarneutrino problem is assumed. with the corresponding angles being denoted θ˜ij . We will first give some general preliminaries and then proceed to analyze different kinds of models. We call this pattern γ 0 . From Eq. Since s. one expects that s˜13 ∼ (m2 /m3 )˜s12  s˜12 . Ue3 = −s 12 s23 c˜13 − s 13 c12 c23 c˜13 + c13 c12 s˜13 . The lepton-mixing matrix. one expects that cm13 + sm12 ∼ cm12 − sm13 . (2) one sees immediately that s˜23 ∼ = s. What makes these expressions useful is that for hierarchical mass matrices. It is also usually true in models (except the flavor-democracy models) that if the 1–2 mixing is large it is due to the angle θ˜12 being large rather than the angle θ 12 . (2). a bar over a quantity means that it comes from the charged-lepton sector. (Henceforth.) Consequently.

all one can say is that Ue3 = O(mνµ /mντ ) or me /mµ . (6) . Thus s˜12 ∼ 1 and s 12  1. It is angle MSW) Ue2 ≈ 0.05 and the formulas can be simplified to Ue3 ∼ = −s 12 s23 and Ue2 ∼ = −s 12 c23 + s˜12 . i. Barr. in virtually all published models the large solar angle comes from the neutrino sector.05. barring accidental cancellations. These facts allow one to write Ue2 ∼ = s˜12 − s 12 c23 c˜12 . In such models one can get the fairly sharp predictions for both Ue2 and Ue3 that we call pattern γ 0 . s˜12 . Consequently.05. that s˜13 ∼ (m2 /m3 )˜s12 ∼ m2 /m3 . but such a coincidence is not what one would typically expect.98 S.) So far we have been considering models of class I(1) in which the matrix Mν has the form given in Eq. In most models one has no sharp prediction for this.e.M. in which case the mixing pattern is α 0 . Next let us consider models of class I(1) with Mν of the form given in Eq. as seen above.07. in other words what we called mixing pattern δ. Ue3 ∼ = −s 12 s23 + s˜13 . then one has the predictions that Ue2 ∼ = −s 12 c23 and Ue3 ∼ = −s 12 s23 . [115. and therefore Ue3 ∼ = Ue2 tan θ23 ∼ = Ue2 tan θatm . one could have s 12 ∼ = s˜12 . (2). [69].121]. [97. s 12  s˜12 . Therefore Ue3 comes from the single term −s 12 s23 . (2) but with large-angle solar solution.or small angle solution to the solar-neutrino problem. as noted. (A good example of a model with pattern α 0 is the large-angle version of the model in Refs. This form only gives large-angle solutions to the solar-neutrino problem. If the solar-neutrino angle is predominantly from the charged-lepton sector. One typically finds in most models that s 12 ∼ me /mµ ∼ = 0.125]. If angle is small then one has that Ue2 ∼ 0. (2).7 and that (for small0. The other possibility in the small-solar-angle case is that the solar angle comes predominantly from the neutrino sector. In that case.121]. i. implying that s 13 is even smaller and in fact negligible. Then it is apparent that one would have Ue3  Ue2 . so that one has s˜13 ∼ 10−4 . However. we have the mixing pattern γ .05. The relation s 12 ∼ = me /mµ is what would be obtained in models where the 1–2 block of the charged-lepton-mass matrix has the Weinberg–Wilczek–Zee–Fritzsch form [122–124]. It is apparent by inspection of Eq. however. (27) Now let us turn to models of class I(1) that have Mν of the form given in Eq. In that case. One then expects. I. The very interesting p point that Ue2 = me /mµ cos θ23 can arise in a simple way and that it gives a good fit for the small-angle MSW solution was first emphasized in Ref. Dorsner / Nuclear Physics B 585 (2000) 79–104 democracy ones). s˜12  s 12 . Thus but not the two terms in the expression Ue3 ∼ = −s 12 s23 + s˜13 are typically of the same order p sharply predicted. (6). In other words. For large-angle MSW solution to the p this gives s˜13 ∼ 0. In other words.05. There are two cases to consider: either large. One of the models that gives the pattern γ 0 predictions is the small-angle case of the model of Refs.e. [97. This version is discussed in Refs. in some models having the WWZF form p for the 1–2 block of L it is predicted that s 12 ∼ = me /mµ . and therefore the mixing pattern is again α. However. one has what we called mixing pattern α. s˜13 can be significant if s˜12 ∼ 1. which is negligible. Now let us consider the form given in Eq. Thus s˜13  0. δm212 is much smaller. Similar results follow for the vacuum-oscillation solution with hierarchical neutrino masses. It is known experimentally that c23 ≈ 0. which for hierarchical models is solar-neutrino problem mνµ /mντ ∼ = (δm212 /δm223 )1/2 . Of course.07. Uµ3 ∼ = s23 + s 12 s˜13 . 0. and so these relations imply that |s 12 | ∼ = |Ue2 /c23 | ≈ p p quite interesting that this is numerically close to me /mµ . s 12 . and therefore.

it is assumed that the lepton-mass matrices have the following forms: L = MFD + 1L. (29) L0 ≡ UFD 0 0 3 Define (1L)ij ≡ δij m` . if the 2–3 block of the charged-lepton-mass matrix L is diagonalized. In Ref.) Thus typically mij /M ∼ δm212 /δm223 ∼ 10−3 or 10−7 for large-angle MSW and vacuum-oscillation solutions. Barr.) Let us consider. as can be seen from Eq. These models give the same results for Ue3 as do class I(1) models that have the form in Eq. (2). (13). unless there is some artificial tuning of the mij one can conclude that s˜13 . As can be seen from Eqs. these forms are convenient when the mass matrices have a hierarchy among their elements. which is not the case for the flavor-democratic form. while δm212 ∼ mij M. respectively. as typically is the case in class II(2). In class II(2) models. and 1L and 1Mν are small corrections that break the flavor-permutation symmetries of the model (generally S3 × S3 ). (6) gives s˜13 ∼ = (−scm22 + (c2 − s 2 )m23 + scm33 )/M. This brings us to models of class I(2). (19). But the more usual assumption is that mν 6= 0. Eq. it turns out that δm212 ∼ = 2(m11 + 2 c m22 + 2scm23 + s 2 m33 )M.S. 10−3 and hence −s s . However. Ue3 ∼ p = 12 23 s 12 is predicted to be me /mµ one has pattern α 0 . most models of this class that do not involve fine-tuning seem to yield the form for Mν given in Eq. I. (2). giving m3 ∼ = m2 ∼ = m1 . Almost all published models of this class are of the “flavor-democracy” type. (19). Dorsner / Nuclear Physics B 585 (2000) 79–104 99 that δm223 ∼ = M 2 . (This will be the case if the neutrino masses have the hierarchy m3  m2  m1 . It is true that in class II(2) models the large 2–3 mixing comes from the charged-lepton sector rather than from Mν . this means mixing pattern α. (24) and (25). The first step in diagonalizing L is to transform it by the orthogonal matrix UFD given in Eq. However. Up to now we have analyzed predictions for Ue3 using the forms given in Eqs. Consequently. (20):   0 0 0 † † LUFD = UFD (MFD + 1L)UFD =  0 0 0  m` + 1L0 . I is the identity matrix. [81] the parameter mν vanishes and Mν = 1Mν has a hierarchical form.M. Therefore we shall analyze the flavordemocracy models in a different way. (15) and (18). (25) it does not much matter in computing the MNS matrix where the 2–3 mixing originates. there is still assumed to be a hierarchy in 2 1Mν so as to get δm12  δm223 . It is straightforward to show that Eq. thus giving m3  m2  m1 for the three neutrino masses. Therefore. Mν = mν I + 1Mν . Then the δij0 are given by (1L0 )ij ≡ δij0 m` . finally. namely the models with lopsided L. (2). but where negligible. The same is also effectively true for most models of class II(2). However. (28) where MFD is the form in Eq. (More precisely. the matrix Mν generally goes over to the form in Eq. Generally. as we have seen. (30) . the models of class II(1). In flavor-democracy models.

δ12 2 3 1 0 δ13 = √ (δ11 − δ22 + δ12 − δ21 + δ13 − δ23 ). Dorsner / Nuclear Physics B 585 (2000) 79–104 1 0 δ11 = (δ11 + δ22 − δ12 − δ21 ). 23. Barr. a point to which we shall return presently. 2 1 1 2 0 δ22 = (δ11 + δ22 + δ12 + δ21 ) − (δ13 + δ31 + δ23 + δ32 ) + δ33 . where (33) (34) As emphasized in Ref. so that Uν ∼ = I and the MNS matrix is † given by UMNS = UL† Uν ∼ = UL . From Eq. and 32 elements of L0 as follows:     0 0 1 0 −δ13 1 0 δ31 /3 /3 L00 =  0 1 −δ 0 /3  L0  0 1 δ 0 /3  23 0 0 /3 δ23 /3 1 δ13   0 0 0 δ11 δ12 0 0 ∼  0  m` . δ23 3 2 i (31) The next step in the diagonalization is to rotate away the 13. = δ21 δ22 0 0 3 0 0 /3 −δ32 /3 −δ31 32 1 (32) Finally the 1–2 block of L00 is diagonalized:     cos θ 0 − sin θ 0 0 cos θ` sin θ` 0 Ldiag =  sin θ 0 cos θ 0 0  L00  − sin θ` cos θ` 0  .100 S. (35) one has then √ √ √ Ue2 ∼ Uµ3 ∼ = −2 cos θ` / 6 + O(δ). = − cos θ` / 2 − sin θ` / 6 + O(δ).M. 31. there is no reason a priori for this angle to be small. δ33 = 3 ij 1 0 = √ (δ11 − δ22 + δ12 − δ21 − 2δ13 + 2δ23). 6 3 3 1X 0 δij . Altogether. the matrix UL that diagonalizes L† L is given by  1  √ √1 √1    0 6 3 2 cos θ` sin θ` 0 1 0 δ31 /3   1 1 1 0 √ √ √  UL =  /3   − sin θ` cos θ` 0  . [127]. √ √ (36) Ue3 ∼ = 2 sin θ` / 6 − (sin θ` δ 0 − cos θ` δ 0 )/3 3. then. 32 31 . 6 X 1 0 = √ (δ1i + δ2i − 2δ3i ). 0 0 1 0 0 1 tan 2θ` = 0 δ0 + δ0 δ0 ) 2(δ11 12 21 22 02 + δ 02 − δ 02 − δ 02 ) (δ22 12 21 11 . 0 1 δ32 − 2 6 3  0 0 1 −δ31/3 −δ32 /3 0 0 1 0 − √2 √1 6 3 (35) The usual assumption is that Mν is nearly diagonal. I.

H.A. [8] A. 28 (1962) 870.N. Phys. hep-ph/9906469. de Holanda. H. Valencia.D. Harvey. Freedman (Eds. Prog. p In fact. Lett. Nucl.). S. B 199 (1982) 223. . Ramond. Slansky. Nussinov. If such a form is chosen. one can conclude that θ`  1. The exact value is evidently dependent on the scheme of symmetry breaking. M. hep-ph/9811265. since the parameters δij are involved in generating the interfamily hierarchy of of chargedlepton masses. B 147 (1979) 277. B 93 (1980) 389. [11] Z. There are many possible forms for 1L that give vanishing θ` . However. North-Holland. in: Lectures at the 1999 European School of High Energy Physics. [2] S. in: P. Slovakia. Nucl. Gibson. Workshop on Unified Theory and The Baryon Number of The Universe. Nielson. Sugamoto (Eds. [4] V.C.C. D 60 (1999) 013002.M. as is apparent from Eq. Phys. However. D. Froggatt. Mohapatra. Barger. one expects that Ue3 will closely related to small lepton-mass ratios. Valle.Z. Phys. [6] M. Supergravity. S. I. Valle. Lett.D. a posteriori we do know that the CKM angles are small.B.W. in other words pattern α. However. Nielson. Supergravity Workshop at Stony Brook. then one has √ √ Ue2 = −1/ 2 + O(δ). Phys. August 22–September 4. Proc.). Nakagawa. [7] T. Gonzalez-Garcia. hep-ph/9911224. Phys. In conclusion.W. Phys. Ue3 = O(δ). Reiss. in: O. it is clear that knowledge of Ue3 will give great insight into the possible underlying mechanisms that are responsible for neutrino mixing [130]. one would typically expect to find that large angles θu and θd were required in the diagonalization of U and D as well. Phys. KEK. Rev.M. Casta Papiernicka. C. Gonzalez-Garcia. there are also schemes of 6 symmetry breaking [128] where Ue3 = 0. Unless there were a conspiracy and θu ∼ = θd . which we have called pattern α 00 . [13] C. (34).F. Theor. 1999. Sakata. Barr. And although there is not a one-to-one correspondence between the type of model and the value of Ue3 . hep-ph/0001311. In a popular scheme of symmetry breaking [128. hep-ph/9903262. Spain. The point is that if a large angle θ` were required in the diagonalization of L. Zee. [5] M.129].C. Under the assumption that 1L has the same form (with different values of parameters) as 1U and 1D. and that strongly suggests that θ` is small. B 161 (1985) 141. |Ue3 | ∼ = p √2 me /mµ .S. Dorsner / Nuclear Physics B 585 (2000) 79–104 101 Since the angle θ` is very sensitive to the nine parameters δij and has no reason a priori to be small. Ramond. Froggatt. this is the case. D. Lett. talk at International Workshop on Particles in Astrophysics and Cosmology: From Theory to Observation. P. [12] R. Whisnant. J. May 3–8.C. 1979. [3] M. A. References [1] J. K. (37) Uµ3 = −2/ 6 + O(δ). for instance. it might seem that the flavor-democracy idea has no predictivity as far as the MNS matrix elements are concerned. Maki. Gell-Mann. Amsterdam. P. P. Yanagida. Bilenky. Peña-Garay. which we called pattern β. [9] C. Van Nieuwenhuizen. 1999. large CKM angles would result. [10] J. Proc. 1979.B. M. Sawada. hep-ph/9809415. and typically one finds that Ue3 ∼ me /mµ . R. B 446 (1999) 256. we see that there are a few patterns of neutrino mixing that tend to arise in the great majority of published models.

S. Jezabek. Leontaris. S.C.K. Kang. Hwang. Phys. C. S. hep-ph/9901254.W. [58] A. [17] C. D 59 (1999) 097702. R. D 57 (1998) 5335.A. hep-ph/9907452.A. Rev. Z. Barr. Pires. Shadmi. C. [41] S. Nir. B 465 (1999) 219. Perez-Lorenzana. Lett. hep-ph/9811252. Rev. Hwang. O. [47] R. M. Joshipura. [26] O. [51] H. Lee. hep-ph/9807353. M. Osaka Univ. F. D 60 (1999) 095009. [38] J. D 60 (1999) 111303. hep-ph/9811363. Y. hep-ph/9507376. Albright. hep-ph/9911275.W. Rev. Phys. Chun.Y. Phys. Lola. X. Fukuura. Phys. C. M. B 442 (1998) 238. Chun. Jarlskog. Lebed. hep-ph/9909318. Carone. [46] R. Rev. [24] A.W. Rizos.D. B 562 (1999) 57. B 440 (1998) 327. F. preprint. Phys. Vergados.D. [16] R. Yoshimura. [35] W. B 445 (1999) 399.K. [54] G. Lett. hep-ph/9911465. J. I. hep-ph/0002044. Smith. Nucl. S. L. Joshipura. Frampton. hep-ph/9808261. S.E. S. S.D. B 445 (1999) 407. Strumia. N. Kong. [27] K. [15] A.N. King. hep-ph/9902385. [39] U. Mohapatra. Lett. I. B 544 (1999) 89. [59] G. Lett. Nelson. [21] A. Yanagida. hep-ph/9806440. Altarelli.102 S. hep-ph/9808304. S. Joshipura. Vergados. [31] O. Lett. W. [57] R. Phys. hep-ph/9904366. F. Rindani. [22] M. Romao. Barbieri.J. Phys. Joshipura. King. Matsuda. [36] C. Büchmuller. Grossman. Gibson. D 53 (1996) 6381. . Froggatt. Phys. Mohapatra. B 439 (1998) 112. H. Dorsner / Nuclear Physics B 585 (2000) 79–104 [14] A. Leontaris.S. Vempati. [44] S. [23] E. Rev. de S. Phys. [55] G. hep-ph/9808355. [50] C. [30] J. [20] P. S. hep-ph/9811265. Stech.K. B 458 (1999) 93. Mahanta. Phys. B. B 443 (1998) 191. Lett.H. hep-ph 9808296.N. Phys. hep-ph/9903435. [34] G. Phys. A.L. D. B 449 (1999) 240. [53] M.K. [18] E. Feruglio. [32] E. hep-ph/9712392. [37] A. hep-ph/9905340. Lett. hep-ph/ 9807235.W. hep-ph/9907532. T. JHEP 9810 (1998) 007. Phys. X. Rev. T. D 53 (1996) 2699. Hirsch. OU-HET-326. Aranda. hep-ph/9907327. [19] K. Y. Altarelli. Rev. hep-ph/ 9812282. H. Vempati.C. S. terVeldhuis. Phys. Drees. Strumia. Babu.K. CERN-TH-99-268. [25] B.S.J. hep-ph/9904210. [49] C. Phys. Nucl. Kong. R. [28] D.D.E. K. [43] Q. Grimus. Hall. Phys. J. hep-ph/9909518.S. Lett. Neufeld.M. Chun. Phys. Phys. Vissani. M.P. Kim. Roy. Davidson. A. Phys. S.W.C. Lett. Phys. hep-ph/9905470. A. Scheich. hep-ph/9905440. D. Lett.S.G. hep-ph/9901243. J. S. hep-ph/9909415. Roy. Lett. Phys. hep-ph/9808333. Sumino.S. hep-ph/9810308. hep-ph/9811266. Joshipura. Choi. Diaz. [33] K. D 60 (1999) 013006. Nielson. B 464 (1999) 239. Shafi.J. Tanimoto. Ghosal. He. Glashow. Ma.D. Masina. hep-ph/9909206. Kim. Rindani. Kang. Kaplan. J.F. Mukhopadhyaya. Phys. E. Tata. Lett. B 461 (1999) 95.K. hep-ph/9907499. Hall. [29] A. A. T. B 451 (1999) 129. Dutta. Feruglio. hep-ph/9907390. M. J. Lett. Phys. JHEP 0001 (2000) 033. Phys.E. Rev. Lett. Pakvasa. Y. C. Cheung. Lee. Phys.J. [45] W. A. Skadhauge. B 445 (1998) 191. hep-ph/9912238. Kovalenko. B 446 (1999) 256. Fukuyama. Valle. Matsuda.S. King.S. hep-ph/ 9911395. Phys. S. hep-ph/9808232. Faessler.L. [52] K. Tavartkiladze. Phys.F. Weiner. B 474 (2000) 355. Lett. Porod. hep-ph/9807386. D 61 (2000) 075012. hep-ph/9906375. hep-ph/9807310. Dermí˘sek. S. Miura. E. Takasugi. [56] B. W. B 439 (1998) 350. Nandi. hep-ph/ 9509351. Barbieri. Chua. D 60 (1999) 031301. hep-ph/9909429. T. [40] S.J.D.D. Haug. L. Ma.K. hep-ph/9808265. Rev. Nishiura. Raby. M. JHEP 9812 (1998) 017. Lett. U. S. [42] E. [48] Y. K.

Phys. Nomura. hep-ph/9607419. Rev. Haba. Buccella. Yoshioki. hep-ph/9807234. S. Takasugi. 101 (1999) 707. Barr. 81 (1998) 1167. JHEP 9903 (1999) 002. Nucl. Tramontano. Phys. Hashida. Creminelli. Phys. A. Phys. Mohapatra. Yanagida. H. D. Lett. 1998. Rev. Babu. B. Rev. Part. hep-ph/9710516. hep-ph/9603376.Kh. Lett. hep-ph/9903460. Prog. Lett. . P. Phys. BA-99-39. D 59 (1999) 113016. Berezhiani. Phys. D 59 (1999) 017303. I. J. hep-ph/9809554. T. Xing. E.S. Costa. F. Petcov. 80 (1998) 3004. T. hep-ph/9807449.C. Lavignac. hep-ph/9809596. Phys. J. Lavignac. A. G. M.S. D 55 (1997) 1659. Roy. Elwood. hep-ph/9908262. Pati. JHEP 9811 (1998) 021. 17–21 May 1999. Y. Q.N. Yoshimura.I. Altarelli. G. Mod. Phys.C. G. M. Theor. Yanagida. Phys. Ramond. Eur. Lett. K. Fritzsch. F. Yanagida. H. hep-ph/9709271. R. D 58 (1998) 035003. Allanach. Rebelo. D 59 (1999) 091302.C. hep-ph/9802314. hep-ph/9906368. hep-ph/9808269. B 372 (1996) 265. Tavartkiladze. Lett. P. Sato. Kugo. G. Phys. Xing. P. Ramond. H. Rev. D 59 (1999) 055001. Barr. Purwanto. hep-ph/9812475. Masina. hep-ph/9905202. Rev. Branco. S. D 61 (2000) 035007. C. hep-ph/ 9808241. Phys. Tanimoto. Tanaka. hep-ph/9806294.I. Kang. N. Ricciardi. Tanimoto. Wilczek. D 61 (2000) 013006. Watari. Talk at Ringberg Euroconference on New Trends in Neutrino Physics. Rev. Malkawi. Fritzsch. Nucl. N. M. Ramond. D 59 (1999) 035011. Phys. Lett. Rev. Lett. E. A.P. M. B 439 (1998) 112. J. Tanimoto. Tanimoto. J. hep-ph/9907532. Phys. Z. Z. E. N. Phys. hep-ph/9802340. A 11 (1996) 2849. hep-ph/9511446. M. M. Rev. B 381 (1996) 202. Jezabek. F. Lett. F. B 430 (1998) 127. C 10 (1999) 491. Z. Nucl. Phys. Phys. S. hep-ph/9808272. K. Akhmedov. Koide. Hagiwara. Phys. hep-ph/ 9911238.M. Phys. DSF-T-99-36. hep-ph/9509389. Fukugita. T. S. Bando. K. Rev. Feruglio. Talk at 6th Topical Seminar on Neutrino AstroParticle Physics. hep-ph/9908294. Phys. Barr. D 57 (1998) 4429. Y.H. B 440 (1998) 313. Rebelo. hep-ph/9912358. M.K. Z. Feruglio.Z. T.L. Phys. C. Phys. hep-ph/9807283. hep-ph/9709388. Yanagida. Binetruy. hep-ph/9802334.N. B 461 (1999) 345. Fukugita. B 559 (1999) 17. B 451 (1999) 388. E. hep-ph/9812538. Lett.C. 40 (1998) 71. D. Lola. K. hep-ph/9811379. Rev. Lett. San Miniato. G. Lett. Irges. Rebelo. Altarelli. Rev. Phys. Kim. Y. hep-ph/9610481. hep-ph/9811495. Nussinov. T. Lunghi. J. B 496 (1997) 3. Irges. M.S. G. M. G. hep-ph/9810542. Feruglio. Phys. Ringberg. Sogami. Sumino. Rev. Falcone. Fritzsch. hep-ph/0001306. H. Prog. M.S. Morizumi. hep-ph/9909208. Phys. Phys. G. Yanagida. Italy. Lett. P. Tanimoto. Ray. Shinohara. Nuovo Cimento A 110 (1997) 549. S. hep-ph/9901245. Shafi. hep-ph/9904338. Eyal.M.N. B 450 (1999) 182. Feruglio. 81 (1998) 5064. Z. J. Branco.Z. Vergados. J. Silva-Marcos. Dorsner / Nuclear Physics B 585 (2000) 79–104 [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] [88] [89] [90] [91] [92] [93] [94] [95] [96] [97] [98] [99] [100] [101] [102] [103] [104] [105] [106] 103 Y. Lett. hep-ph/9807552. I. Babu. hep-ph/9807310. hep-ph/9811447. T. F. hep-ph/9911364. Oda. S. J. M. G. hep-ph/9807325. K. B 548 (1999) 60. hep-ph/9808301. K.N. B 441 (1998) 299. T. hep-ph/9710417.D.S. I. Rossi. H. Romanino. Rev. B 440 (1998) 327. and references therein.S. Nucl. Lett. S. Germany. Albright. F. D 59 (1999) 017304. Phys. Barr. M. Branco. T. Barbieri. M.M. Phys. B 441 (1998) 191. hep-ph/9807228. S. M. N. Sarkar. Phys. Phys. Altarelli. Tanaka. T. Phys.Z. Babu. Phys. Abud. R. M. Wu. G. Phys. Altarelli. Xing. Fritzsch.K. hep-ph/9807308. Okamura. S. Phys. Lett. hep-ph/9807353. Silva-Marcos. A. Rev. S. F.K. P. Lett.M. B 428 (1998) 136.

Yanagida. hep-ph/9903334. Branco. D 60 (1999) 115016. Phys. gives an analysis of Ue3 that is different from but not inconsistent with the one given here. H. Rev. C. Lett. M. Z. T. hep-ph/9912205. Phys. T. T. Lett. R. 79 (1998) 4748. hep-ph/9901228. Italy. hep-ph/9901318. S. Lett. G. K. D 48 (1993) 5354. NSF-ITP-82-07.M. Fritzsch.104 [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] S. S. Tanimoto. Barbieri. Phys.S. Dimopoulos. D 57 (1998) 4429. D 50 (1994) 3529.H. 101 (1999) 1313. I. 461 (1999) 218. Wilczek.C. Z.I. Albright. OUTP-9901-P. C. B 440 (1998) 313. Rev. 033012. hep-ph/9808272. B 86 (1979) 297. D 59 (1999) 017304. Y. Barr. Y. M. Lett. Ma. Hall. School of Subnuclear Physics. Y. Zichichi (Ed. K. Lett. F. 1983. Weinberg. Phys. S. hep-ph/9910314. hep-ph/ 9904303. Albright. H. Bando. Phys. in: A. Proc. H. L. Babu. S. Y. C. Kugo. NY Acad. A. Phys. Yanagida. Phys. Q. Phys. Nir. hep-ph/9807283. . Rev. B 70 (1977) 418.H. G. 1981. P. Erice. C. Barr. C. Rebelo. Phys. K. H. hep-ph/9909249.Kh. Kurosawa. Lett. D 61.M. A. hep-ph/9906296. Zee. Barr. Nomura. Rev. Dorsner / Nuclear Physics B 585 (2000) 79–104 M. hep-ph/9902293. Rev. Z.J. Trans. hep-ph/9910522. Phys. Frampton. D 58 (1998) 013002.N. New York. hep-ph/9712488. S. Plenum Press. Phys. Rev. 1981. S. Phys. 19th Course of the Int. Fritzsch. Jarlskog. E. Fukugita. hep-ph/9709388. Report No. hep-ph/0002155. Tavartkiladze.M. Xing. IFP-777-UNC. Izawa.M. The Unity of Fundamental Interactions. T. Prog. Fritzsch. M.H. B 70 (1977) 436. Phys.M. Phys. Shafi.M. Tanimoto. Rasin. Barr. Barr. Lett. S. hep-ph/9912358. S. F. Georgi. Nomura.Z. Xing. Albright. M.L. Barr. hep-ph/9902204. G. Rev. Akhmedov. BA-99-63. S.G. 38 (1977) 185. Ross. Theor.H. Rev. Barr. E. Kane. Albright.). Sugimoto. Wilczek.M.Z. Phys. Raby. B 452 (1999) 287. JHEP 9905 (1999) 023. Shadmi. Sci.

Lindner ∗ Theoretische Physik. James-Franck-Strasse. masses and mixings at a neutrino factory ✩ M. E-mail addresses: martin. 2000 Elsevier Science B. as well as the magnitude of sin2 2θ13 in very long baseline neutrino oscillation experiments. We study neutrino oscillation at a neutrino factory in the νµ → νµ disappearance and νe → νµ appearance channels with and without muon charge identification. which requires however very good muon charge identification. manfred. leptonic mixings and CP-violation. All rights reserved.V.de (P.60. PII: S 0 5 5 0 . Without charge identification it is still possible to operate the neutrino factory both with µ− and µ+ beams and to analyze the differences in the total neutrino event rate spectra. PACS: 12. Introduction Proposed neutrino factories [1–5] offer unique possibilities to improve the knowledge about neutrino masses. One possibility is to analyze the νe → νµ appearance channels leading to wrong sign muon events. Freund. P. We show that this leads already to a quite good sensitivity. 14. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.freund@physik. Huber. patrick. the value and the sign of 1m231 . accepted 6 June 2000 Abstract We discuss and quantify different possibilities to determine matter effects.de (M. All rights reserved.de (M. Germany Received 20 April 2000.2 . Freund).V.15.tu-muenchen. M.tu-muenchen. 14. D-85748 Garching.elsevier.nl/locate/npe Extracting matter effects.St 1. ∗ Corresponding author.lindner@physik.Pq. which may be important if right sign charge rejection capabilities are insufficient. Physik Department.60. The leptonic sector is not plagued by hadronic uncertainties such that vacuum neutrino oscillation allows to ✩ Work supported by “Sonderforschungsbereich 375 für Astro-Teilchenphysik” der Deutschen Forschungsgemeinschaft. Technische Universität München. The best method is finally achieved by combining all available information of the νµ → νµ disappearance and νe → νµ appearance channels with charge identification and we show the sensitivity which can be achieved. Lindner). Huber).tumuenchen. With muon charge identification one can study the νµ → νµ disappearance and the νe → νµ appearance channels independently.Nuclear Physics B 585 (2000) 105–123 www.Ff.huber@physik.3 2 1 3 ( 0 0 ) 0 0 3 6 9 .

1m2 = m2j − m2i and one mixing angle θ only.e. First. but one can operate a neutrino factory already both with µ− and µ+ beams and analyze the differences in the combined muon neutrino and antineutrino event rate spectrum. one can study the νµ → νµ disappearance and the νe → νµ appearance channels separately. mj .. with muon charge identification. We discuss and quantify in this paper further possibilities to determine the sign of 1m231 via matter effects and sin2 2θ13 by considering the appearance channels and/or the νµ → νµ and ν¯ µ → ν¯ µ disappearance channels in different scenarios with and without muon charge identification. The basic mechanism which allows the extraction of the sign of 1m2 via matter effects can. The point is that the total event rates decrease slower than the vacuum rates leading for a large range to a statistically constant result.106 M. this provides however not much extra information. The basic effect which allows to extract the sign of 1m231 comes from coherent forward scattering of electron neutrinos in matter which leads to effective masses and mixings different from vacuum. It has however been pointed out recently [6–8] that the sign of 1m231 can be determined from the νe → νµ and ν¯ e → ν¯µ appearance rates via matter effects [9. We show for our baseline of 7332 km the sensitivity which can be achieved. already be seen in a simplified 2 × 2 picture with two mass eigenvalues mi . sin2 2θm = sin2 2θ · C± where the neutrino energy E enters via the matter term (1) . We will discuss in this paper a full three neutrino framework in the limit where the small solar mass splitting can be neglected in vacuum. Last. −2 . We use therefore in this paper mostly a baseline of 7332 km since for this larger baseline there is further information in the disappearance channels. the appearance channels). the best method is achieved with muon charge identification by combining all available information of the νµ → νµ disappearance and νe → νµ appearance channels. Altogether there are four possibilities.e. i. 1m221 = 0. Next.10] in very long baseline neutrino oscillation experiments at neutrino factories. For a baseline between 2800 km and 7332 km we find that the significance of these result is essentially unchanged compared to 7332 km. Vacuum neutrino oscillation is however only sensitive to mass squared differences thus the sign of 1m231 can not be determined allowing therefore at the moment different scenarios for the ordering of mass eigenvalues. however. without charge identification one can not study wrong sign muon events (i. One obtains then the well known relations for the parameters in matter s 1m2m = 1m C± = 1m 2 2 + A ∓ cos 2θ 1m2 2 ± sin2 2θ . Freund et al.. This degeneracy is however destroyed in matter and we need therefore a full three neutrino description. A third option would be to use the disappearance channel alone. / Nuclear Physics B 585 (2000) 105–123 determine many parameters quite precisely. We show that this leads already to a quite good sensitivity since the effects in the combined appearance and disappearance channels are for a baseline of 7332 km about the same size and go in the same direction. This is useful if charge identification is not available or not operative in an initial stage of the experiment.

Matter effects modify the 2 × 2 vacuum transition probabilities  2  2 2 1m L and P (νi → νj ) = sin 2θ · sin E P (νi. τ.e. for cos 2θ = 1 roughly the relation     1m231 2. (1) becomes “resonant” for A = 1m2 cos 2θ. l = e. given A) clearly on the sign of 1m2 allowing thus in principle an extraction of this sign. Freund et al. One could. (4) leads already to a rough approximation for a “optimal mean neutrino energy”..8 g/cm3 .5 × 10−3 eV2 i. Analytic description of three neutrino mixing in matter The mixing of three neutrinos can be described via |νl i = 3 X k=1 Ulk |νk i. Our paper is organized as follows.2 g/cm3 .M. First we develop in Section 2 analytic expressions for neutrino oscillations in matter in a suitable approximation. for example. since the matter corrections go then in opposite direction inducing an asymmetry. (4) This condition can be fulfilled for a given sign of 1m2 either for neutrinos or for antineutrinos at a specific resonance energy.. In Section 3 we describe the experimental framework on which our numerical results are based.. (6) . 2. The matter corrections in Eq. Section 6 contains our results on the sensitivity to the magnitude and sign of 1m231 and for sin2 2θ13 . (5) · Eopt = 15 GeV ρ 3. i.e. The biggest matter effects (and therefore the best sensitivity to the sign of 1m2 ) occur when Eq. / Nuclear Physics B 585 (2000) 105–123 √ 2 2 GF YρE A ≡ 2EV = ± . Nevertheless Eq.j ) = 1 − P (νi → νj ).j → νi. µ. mn 107 (2) The sign of the matter term A is for electron neutrinos (antineutrinos) traveling inside the earth positive (negative) and leads thus to different corrections for neutrinos (C+ ) and antineutrinos (C− ). but optimization will be more complicated in a real experiment with an energy spectrum and other free parameters. compare neutrinos with antineutrinos.8 g/cm3 going down to about 10 GeV for paths crossing the earth at a distance of 7332 km with hρi ' 4. (1) depend for a given neutrino species (i. an optimal energy of about 15 GeV in the crust of the earth with an average effective density hρi ' 2. (4) gives for small θ .e. (3) since in matter one must replace sin2 2θ → sin2 2θm and 1m2 → 1m2m . Section 4 describes the effects of matter on the total rates and the used statistical methods are outlined in Section 5. In earth Eq.

U is a 3 × 3 unitary leptonic mixing matrix which we parameterize as   Ue1 Ue2 Ue3  Uµ1 Uµ2 Uµ3  Uτ 1 Uτ 2  Uτ 3 c12 c13 =  −s12 c23 − c12 s23 s13 eiδ s12 s23 − c12 c23 s13 eiδ s12 c13 c12 c23 − s12 s23 s13 eiδ −c12 s23 − s12 c23 s13 eiδ  s13 e−iδ s23 c13 . 2. The squared mass splittings 1m2 = m2j − m2i show the hierarchy 1m221  |1m231| as discussed in more detail below and allow for 1m221  10−4 eV2 . Alternatively the CP-phase could be included by making any of the three rotations complex. The angles are assumed to be in the ranges 0 6 θ12 . k 6= j = 1. We can assume therefore for the following discussion that there is only one nonvanishing quadratic mass splitting 1m232 = 1m231 = 1m2 in vacuum. This becomes useful when we write the full Hamiltonian in matter with the help of U in flavour basis 1 Fig. We have thus two almost degenerate mass eigenvalues and for 1m231 > 0 (1m231 < 0) these almost degenerate eigenvalues m1 and m2 are lighter (heavier) than m3 (see Fig.e. Two mass ordering schemes determined by the sign of 1m231 . To understand this case in matter we can write the mixing matrix Eq. i. Freund et al. 3. / Nuclear Physics B 585 (2000) 105–123 where |νl i corresponds to the neutrino flavour state νl and |νk i corresponds to the neutrino mass eigenstate νk with eigenvalues mk . The usual oscillation formulas depend only on 1m231 and it is thus impossible to discriminate between the two schemes. 0 6 δ < 2π .108 M. E > 1 GeV and L 6 104 km the approximation where 1m221 = 0 in vacuum. c23 c13 (7) where cij = cos θij and sij = sin θij . . by replacing s13 → s13 e−iδ . (7) by a complex rotation in 13-subspace. Matter effects and CP-violating effects can lift this degeneracy in future measurements in very long baseline experiments. 1 A constant common neutrino mass can be separated leading to an overall phase. θ23 . θ13 < π/2. mk 6= mj . 1).. (7) as a sequence of rotations U = R(θ23 ) · R(θ13 ) · R(θ12 ) where the CP-phase δ is included in Eq. 1. The mixing matrix contains in general further CP-violation phases which do not enter in neutrino oscillation and are therefore not specified [11–13].

We can readily use the standard two neutrino parameter mapping given in Eqs. Jijlm := Uli Ulj∗ Umi mj 2 and 1ij := 1mij L/4E the transition probabilities P (νl → νm ) from flavour l to flavour m in vacuum can be written as X . (8) inside the square brackets.M. Furthermore R(θ23 ) operates in the 23-subspace and commutes therefore with the matter term. which implies that matter affects in the 1m221 = 0 limit only the 13-subspace. however. which operates in 12-subspace. (9) 1m221. 2 1m2 (C± − 1) − A 00 . drops out in the limit where 1m221 = 0.m = 1m2 C± and sin2 2θm = sin2 2θ/C± 2 2 2 1m ≡ 1m31 enters into C+ defined in Eq. Since we could use a convention where the CP-phase would be defined by a complex rotation in 12-subspace we can immediately see that the CP-phase drops out as well. Freund et al. / Nuclear Physics B 585 (2000) 105–123   0 0 1   R(θ23 ) R(θ13 ) 0 0 H= 2E 0 0   0 A T   R(θ13 ) + 0 0 1m2 0 109  0 0 0 0  R(θ23 )T . Thus even if we can approximate 1m221 = 0 in vacuum we need for oscillation in matter the full three neutrino oscillation formulae. namely 1m2 (C± + 1) + A .m = 1m2 C± = 2 An important consequence is that the mass degeneracy between the first two eigenvalues is destroyed in matter. Note. (8) 0 0 Here use has been made of the fact R(θ12 ). (1) for the 2 . The only simplification left is that one can set in the mixing matrix θ12 = 0 and δ = 0. We see that matter effects are confined in Eq. (1). that 1m32 and 1m221 become in matter in this way also energy dependent. where 13-subspace and obtain in matter 1m231. Defining as shorthands Dab = e−iEa t δab . These matter induced parameter mappings must now be inserted into the oscillation for∗ U mulae for three neutrinos.

.

2 Re Jijlm sin2 1ij P (νl → νm ) = .

hl|U DU + |mi.

namely the shifted mass eigenvalues (1m231. (11a) P (νµ ↔ νµ ) = 1 − sin 2θ23 sin θ13 sin (121 ) − sin θ23 sin 2θ13 sin (131 ) 2 2 2 4 2 − sin2 2θ23 cos2 θ13 sin2 (132 ).m = 1m2 C± and Eqs. (10) i>j As described above we can set in matter θ12 = 0 and δ = 0 resulting in P (νµ ↔ νe ) = sin2 θ23 sin2 2θ13 sin2 (131 ). = δlm − 4 0 = 1m232. (9)) 2 .m = 1m2 C± i>j −2 X Im Jijlm sin 21ij .  P (νµ ↔ ντ ) = sin2 2θ23 sin2 θ13 sin2 (121 ) − 14 sin2 2θ13 sin2 (131 )  + cos2 θ13 sin2 (132 ) .m = sin2 2θ13/C± antineutrinos are only formally identical due to the opposite sign of the matter term A . sin2 2θ13. Note that the relations for and the shifted 13-mixing angle. 2 (11b) (11c) To describe oscillation in matter we must insert into these formulae the parameter mappings discussed above.

3. The detector with NkT kilotons is assumed to be able to measure muons above an assumed threshold of 5 GeV [19] with an efficiency µ . 2 We avoid the core of the earth and take in our study distances up to 7332 km. since they can be separated by kinematical means without loosing too much muon detection efficiency [6]. The inclusion of various other potential backgrounds and detector properties will have some influence on the analysis. In vacuum the relations (11a)–(11c) could be further simplified since the tight CHOOZ results [14] can be taken into account. Freund et al. These results can be studied in a simplified 2 × 2 oscillation picture where the mixing matrix contains only one mixing angle θ with the two flavour transition probabilities in vacuum given in Eq.1. Among these the LSND evidence [23] for oscillation is most controversial and we omit this result therefore in our analysis. while a constant density for all paths ignores even the mean density variations in the earth and has sizable errors. Experimental framework We use for our study a standard earth matter density profile [15. (1).0 × 10−3 eV2 6 |1m231| < 3. Most limitations due to backgrounds and statistics can however be overcome by increasing the number of muon decays and/or the size of the detector. while we take the atmospheric result (mostly from SuperKamiokande [24]) and the solar neutrino deficit (mostly from GALLEX [25]) as evidence for oscillation. For typically allowed 1m231 values one has sin2 2θ13 . We assume furthermore a neutrino factory which can operate with µ− or µ+ beams leading to νµ + ν¯ e or ν¯ µ + νe neutrino beams resulting from Nµ± muon decays.2 are allowed. We do not include a detailed simulation of background effects at the moment. (11a)–(11c) are potentially dangerous.0 × 10−3 eV2 where values of sin2 2θ13 ' 0. For three neutrinos there is only room for two independent quadratic mass splittings while there are three different results for oscillation.18] that this approximation works very well. It has been shown recently [17. The muons may come from the dominant νµ → νµ disappearance channel or from the subdominant νe → νµ appearance channel with or without charge identification. More detailed simulations of backgrounds are still in progress [19–22]. Muons coming from the subdominant νµ → ντ or νe → ντ channels with the τ decaying subsequently to a muon should not be a problem. while this upper limit is less stringent for 1. The dominant atmospheric νµ ↔ ντ oscillations imply in this picture a mass splitting [26] . 0.110 M. / Nuclear Physics B 585 (2000) 105–123 in Eq. (3). The varying matter profile of the earth implies that the average density is still path dependent. but with current understanding this would only result in moderate corrections.16] and we will treat neutrino oscillations in earth by a constant density approximation along each neutrino path. Note however that the effective sin2 2θ13 in matter can be much bigger when the resonance condition is fulfilled such that further simplifications of Eqs.

.

(12) 10−3 eV2 . .

1m231 .

Either one should avoid paths passing the core or one should take core effects into account. 8. .0 × 10−3 eV2 . Specific core effects become however small when the oscillation wavelength becomes bigger than the size of the core. 2 Note also that care should be taken about the core of the earth. .

all with |1m221|  |1m231 |. These are the solar vacuum oscillation (VO) solution and the large and small mixing angle (LMA and SMA) solar MSW solutions. We will see that our study depends only on 1m231 such that in a very good approximation 1m221 can be set to zero. / Nuclear Physics B 585 (2000) 105–123 111 while the solar neutrino deficit leads to different solutions [27. We use therefore as default initial parameters of our study .M. SMA MSW and LMA MSW solution of the solar neutrino problem.28] for 1m221 . Freund et al. while 1m221 is most relevant for the VO. The above assignment implies thus that 1m231 dominates the atmospheric νµ → ντ oscillation.

.

.

.

.

1m2 .

= 0. .

1m2 .

The second term is simply the product of flux times detection cross section which has to be folded with the third term. Our analysis is performed at the level of total and differential event rates. efficiency µ one obtains for the channel “i” (where i stands for a proper index uniquely given by the polarity of the muon beam and the considered oscillation channel) the contributions to the differential event rates    109NA Eµ dni = Nµi NkT µ 2 gi (E/Eµ )σi (E) [Pi (E)]. Nµ− = 2 × 1020µ− decays. = 3.5 × 10−3 . The second term grows initially like E 3 and reaches a maximum before it goes to zero at E = Eµ . (15) | {z } dE mµ π L2 {z } | {z } | oscillation normalization 109NA flux is the number of nucleons per kiloton in the detector and Pi stands for where the relevant oscillation probability in matter as discussed above. the oscillation probability in matter. An important feature of the second term is that the low energy part is not changed when the muon energy is increased. sin2 2θ23 = 1. detector size NkT . This implies in a good approximation that the increase in the total event rates by increasing Eµ comes . (15). In cases where different channels contribute the total differential rates are given by the sum of all individual terms.0 (13) 31 21 and we will discuss variations of these parameters within the allowed ranges. an NkT = 10 kt iron detector and a muon detection efficiency µ = 1. which is only important for the proper event rate normalization.67 × 10−38 E cm2 /GeV. Total rates are obtained by integrating these differential rates from the threshold at 5 GeV to the maximal possible neutrino energy Eµ . Unless otherwise mentioned we use a baseline of L = 7332 km. We include therefore the charged current neutrino cross sections per nucleon in the detector and the normalized νµ + ν¯ e and ν¯µ + νe beam spectra gνi of the neutrino factory [1] σνµ (E) = 0. σν¯ µ (E) = 0. gνe (x) = gν¯ e (x) = 12x 2(1 − x). (14b) where x = E/Eµ . (14a) gνµ (x) = gν¯ µ (x) = 2x 2 (3 − 2x). This corresponds for example to a setup with Nµ+ = 2 × 1020µ+ decays. To understand the event rates we look at the three terms in square brackets on the rhs of Eqs. The first term contains overall factors which are constant in energy. The event rates of our results depend only on the combination Nµ± NkT µ and we take for both polarities Nµ± NkT µ = 2 × 1021 .34 × 10−38 E cm2 /GeV. For a given number of useful muon decays Nµ± .

Since the event rates are dominated by the maximum of the flux factor close to Eµ we can set E ' Eµ and find no oscillation effects when Eµ is much larger than E1 . This has also the important consequence that experimental results for a lower muon beam energy Eµ0 < Eµ can in principle be obtained to a good approximation at higher energies by simply removing events with energies above Eµ0 . This means that there will essentially be no effects of oscillation in the total disappearance rates when E1 is much smaller than the neutrino energy E. one finds for the asymptotic appearance rates for small sin2 2θ13 3 Note that the mass splittings 1m2 behave nontrivial in the high energy limit such that the asymptotic ij properties of Eq. 4. depend then for fixed 1m231 = 1m2 only on E/L. (5) at roughly 10−15 GeV with E1 and Eµ ' E. and using the fact that V · L is numerically small. In order to understand how matter effects influence the disappearance rates one must compare the matter “resonance energy” Eopt in Eq. At neutrino energies which are not close to this resonance energy the mixing sin2 2θ13. 3 The total appearance rates are given for Eµ  E1 by the asymptotic behavior of Eq. For lower beam energies there are in principle further dips due to all other oscillation maxima. The effects of oscillation can be seen in the total rates as a dip in the overall Eµ2 scaling. but this is on our case below the threshold energy of 5 GeV.112 M. The oscillation becomes maximal for Eµ ' E1 while it vanishes again for Eµ ' E1 /2. (10). Total rates The analytic description above allows now an understanding of the event rates which we calculated numerically and which are presented in Fig. The increase in the rates with Eµ comes again predominantly from the increased flux at energies close to Eµ such that we need to look again in a good approximation at the asymptotic behaviour of Eµ2 P (νµ ↔ νe ). which is small. 2. To understand first the total disappearance rates without matter one must look at the folding of the vacuum oscillation probabilities with the flux factor in Eq. The first effects of vacuum oscillation show up for lower beam energies when E ' Eµ is decreased such that Eµ = O(E1 ). The overall rates will however still decrease with the muon energy Eµ according to the cross sections and fluxes approximately like Eµ2 as long as Eµ is sufficiently larger than the muon threshold energy of 5 GeV. / Nuclear Physics B 585 (2000) 105–123 essentially from the high energy tail of the spectrum only. The oscillatory terms in the oscillation probabilities Eqs. (15). Expanding Eµ2 P (νµ ↔ νe ) in powers of 1/E. Freund et al.m approaches quickly its value in vacuum. . One can see immediately from the Hamiltonian in Eq. (11a) weighted with the flux. namely the sin2 1ij factors. (11b) have to be evaluated carefully to reach the same conclusion. The oscillatory behavior vanishes for E/L → ∞ and the first maximum of the vacuum oscillation occurs for given L at an energy E1 = 21m2 L/π . (8) that the dominant νµ → ντ oscillation decouples in the limit where sin2 2θ13 = 0 from the matter effects coming from the first generation such that it is clear that matter effects should be rather localized around the resonance region.

5 × 10−3 eV2 and sin2 2θ13 = 0. / Nuclear Physics B 585 (2000) 105–123 113 Fig. Muon appearance rates νe → νµ and ν¯e → ν¯µ (dotted) and muon disappearance rates νµ → νµ and ν¯ µ → ν¯µ (solid) as a function of the muon beam energy for a baseline of 7332 km. .1. ⊕ and indicate positive and negative mass squared difference. Freund et al. 2. |1m231 | = 3.M.

|1m231| = 3. a fixed baseline of 7332 km..e. The following figures for the total rates in the different oscillation channels assume Nµ± NkT µ = 2 × 1021 . however. i.   (1m2 )2 L4 4(1m2)3 L4 V n ∝ sin2 θ23 sin2 2θ13(1m2 )2 L2 1 − . There are thus significant effects even for beam energies which are somewhat away from the resonance energy.5 × . + 3Eµ2 3Eµ (16) The result shows that the rates approach for Eµ  E1 in vacuum a constant which depends only on even powers of 1m2 . The leading corrections to this constant fall like 1/Eµ2 . on the sign of 1m231 . This understanding of matter effects can now be compared with the results of our full numerical calculations. that the discussion of effects is much more complicated for muon energies of the order or smaller than E1 . Turning on matter effects leads in Eq. This induces in our case a matter dependent splitting in the appearance rates which is less localized in energy than the matter effects in the disappearance channels. (16) to the third term which falls only like 1/Eµ and which depends on (1m2 )3 . indicates no oscillation rates. Note. • and ◦ label the neutrino and the antineutrino channels.

(14). 2 shows the cases with positive 1m231 (⊕). 2 shows the νe → νµ and ν¯ e → ν¯µ total appearance rates (dotted lines) as well as the νµ → νµ and ν¯ µ → ν¯ µ disappearance rates (solid lines) on a logarithmic scale for maximal sin2 2θ13 = 0.1. / Nuclear Physics B 585 (2000) 105–123 10−3 eV2 . The lines appear in pairs of neutrino (•) and antineutrino (◦) channels which differ roughly by a factor two coming from the cross sections Eqs.114 M. negative 1m231 ( ) and no oscillation (. Fig. sin2 2θ23 = 1 and different values of sin2 2θ13 . If the muon beam energy is fixed then we use always Eµ = 20 GeV. Freund et al. Fig. These rates are in agreement with the results obtained recently by Barger et al. [6].

0. Fig. 0. The combination of appearance and disappearance channels amplifies thus always the signal. Fig. The difference between the two signs of 1m231 influences the total rates up to a factor two or more.e. For negative 1m231 . muons with both charges. i. The effect comes partly from the appearance rates and partly from the disappearance rates. while the νe → νµ appearance rates are mostly unchanged. while matter corrections are moderate for the ν¯ µ → ν¯ µ disappearance channel. The resulting total muon rates of the combined channels are shown for both µ− and µ+ beams in Fig. We find that the different matter effects have no specific spectral features which could be used as a basis for cuts in order to amplify the signal relative to the oscillation signal. or in other words the combination of the two oscillation channels. which appear only to be smaller due to the logarithmic scale. 2 shows clearly the important feature that the matter effects in the combined appearance and disappearance muon rates go for a neutrino factory with either νµ + ν¯ e or ν¯ µ + νe beams in the same direction. The asymmetry between ⊕ and in the appearance rates. the disappearance channel ν¯ µ → ν¯ µ shows a resonant transition of νµ to νe resulting in a suppression of the total νµ + ν¯ µ rate. This is for positive 1m231 the neutrino component of the beam and for negative 1m231 the antineutrino component.04.1.01). on the other side. We will show how well these parameters can be re-extracted in an appropriate statistical way.. This shows that an analysis of the combined appearance and disappearance channels is interesting and it may be especially important if the separation of right and wrong sign muon events is not available. is clearly visible. We provide therefore in 4 For µ− beams holds the same for a reversed sign of 1m2 . This opens an interesting possibility if muon charge identification is not available to separate the appearance channels (wrong sign muon events) from the disappearance channels (right sign muon events). 31 . but always from the resonant channel.29]. We will fit therefore in the following differential event rate spectra of data simulated with certain initial parameters and with statistical noise added. where matter effects add up.1 (CHOOZ bound). Note that there are also comparable matter effects in the disappearance channels at these large baselines.). The figure shows the sizable matter induced splittings in the total muon-neutrino rates (νµ + ν¯ µ ) for the mixing angles sin2 2θ13 = (0. It is then still possible to measure all muons. 3 on a linear scale for µ− and µ+ muon beams. which has been used to extract the sign of 1m231 at smaller baseline [6. For µ+ beams the interplay of the two channels is as follows: 4 for positive 1m231 the νe → νµ appearance channel is resonant leading to an enhancement of the combined rate. 4 shows the corresponding differential event rate spectrum of the νµ + ν¯µ induced muon rate over twenty energy bins for a mixing angle sin2 2θ13 = 0.

.5 × 10−3 eV2 and Nµ± NkT µ = 2 × 1021 which corresponds typically to a 10 kt-year. The left plot was done for negative stored muons. Fig.M. Total muon rates νµ + ν¯ µ (i. the right plot for positive muons.1.e. / Nuclear Physics B 585 (2000) 105–123 115 Fig. 3 for sin2 2θ13 = 0.. We use 20 energy bins over the range from 5 GeV to 20 GeV. Spectral distribution of the muon rates νµ + ν¯µ according to Fig. |1m231 | = 3. Freund et al. The different line types stand for different mixing angles θ13 (see labels). ⊕ and indicate positive and negative mass squared difference. νµ → νµ combined with ν¯e → ν¯µ and ν¯ µ → ν¯ µ combined with νe → νµ ) as a function of the muon beam energy for a baseline of 7332 km. 3. 4.

With charge separation we calculate χµ2 ± separately for neutrinos and antineutrinos and use the sum χ 2 = χµ2 − . simulated) and theoretical predicted muon event rates. 99.ν + χµ2 − .e. 6. 2σ. 95. 1χ 2 = χ 2 − χmin 1χ 2 (20) χ2 obeys a distribution for k degrees of freedom (i. sin2 2θ23 .e.8). We subtract the minimum value of χ 2 and define confidence levels according to the values of 2 . / Nuclear Physics B 585 (2000) 105–123 the next section information about the used statistical procedures and we will study then in more detail the statistical significance of the matter effects in the appearance channels. (19) Next χ 2 is minimized as usual with respect to the parameters shown in the plots. Confidence levels (CL) are calculated by using  n  X   nobs i obs obs − n log + 2n 2 nth .. The analysis of these “data” is done according to the procedure proposed by the Particle Data Book [30] for Poisson distributed quantities.. The value of 1χCL 2 1χCL Z CL = k k 0 x x 2 −1 e− 2 2 2 Γ ( k2 ) dx.2. to 1χCL We use in our analysis for the interval between 5 GeV and 20 GeV a total of 20 bins. Finer binning does thus not improve the significance due to the statistical limitation by fluctuations. (18) If there is no charge identification then ni = (nµ+ )i + (nµ− )i is the total (indistinguishable) number of νµ and ν¯µ induced muon events in the ith energy bin and obs and th label the observed (i. µ+ and µ− beams together with a detector such that Nµ+ NkT µ = Nµ− NkT µ = 2 × 1021 .116 M.. 5.3. .e.ν + χµ2 + . Statistical methods and data evaluation We numerically simulate data for a given parameter set (Eµ . as usual.¯ν . the number of parameters This 2 which corresponds to a given confidence level CL is given by: fitted). In this sense it is best to use around 20 energy bins which should not be a problem with the usual energy resolution of detectors. 1m231 . i. the sign of 1m231 and to test matter effects. The total χ 2 is given by χ 2 = χµ2 + + χµ2 − . sin2 2θ13 ) and add Poisson distributed fluctuations to the energy bins. (21) For two degrees of freedom (1σ. We checked that the statistical significance of the results does not change much by changing the number of bins.5%. as long as one has enough bins to describe the spectral information sufficiently. We focus on the potential to extract θ13 . Freund et al. 11. disappearance channels and combinations thereof. (17) χ2 = i i i nth i i=1 We assume symmetric operation of the neutrino factory in both polarities.¯ν + χµ2 + .3%. 3σ ) corresponds to CL = (68.7%) and 2 = (2.

Fig. 5 the values of the global minima with the wrong sign. This allows a determination of |1m231 | and sin2 2θ23 which is essentially independent of the sign of 1m231 and sin2 2θ13 . Results As outlined above. This are in Fig. the precision in the determination of sin2 2θ13 gets worse. Instead of a global fit to all parameters one can first fit the magnitude of |1m231| and sin2 2θ23 with good precision (a few percent [3]) by analyzing the muon spectrum for the dominant νµ → νµ and ν¯ µ → ν¯ µ disappearance oscillation. Fig. We will see that this works very well if the mixing angle θ13 is not too small. We therefore restrict the fit for the wrong sign of 1m231 to the local minimum of χ 2 in the neighborhood of the best fit of |1m231|. however. 5 shows the 1σ and 2σ contour lines of the right sign fit for different |1m231| and different θ13 in the case with no charge separation (i.e. 5–7 next to the CL contour lines are for the wrong sign of 1m231 relative to the best fit with the correct sign of 1m231 . i.M. such that the matter dependence is almost removed.. For the right sign we can calculate confidence level contour lines from which we can read off the sensitivity for the determination of the relevant oscillation parameters. In this case. This is shown in Fig. 6. The determination of the sign of 1m231 is always better in the appearance channel. / Nuclear Physics B 585 (2000) 105–123 117 6. quite a good precision can be obtained also for |1m231|. muons of either charge coming from νµ or ν¯ µ ). we simulate a measurement with certain input parameters including statistical fluctuations for a given sign of 1m231 . Since the fitted spectrum includes the dominant disappearance channel. The value of 1χ 2 for the wrong sign of 1m231 will tell us the confidence level at which we can reject the incorrect sign of 1m231 . Charge separation capabilities improve the sensitivity for sin2 2θ13 . Freund et al.e. 5 One could combine the differential event rates coming from the µ+ and µ− beams. With decreasing θ13 . The precision for |1m231| can of course be improved in this case by a separate analysis of the disappearance spectrum [6]. but this method allows to determine sin2 2θ13 already down to values of order O(10−2 ) (see also Fig. we find that the sum-rates without charge identification provide at least an equally well suited possibility to extract the value of sin2 2θ13 . that for not too small mixing angles sin2 2θ13 of about one magnitude below the current CHOOZ constraint.. one looses of course precision in finding the right value of |1m231|. Then we try to fit this dataset. but note that the sum-rates give also quite good confidence levels for not too small mixing angles θ13 . 6 shows that charge identification improves the precision in the determination of sin2 2θ13 for small values of sin2 2θ13 and the sensitivity will go down to values of order O(10−3 −10−4 ). we re-extract the input parameters both with the right and the wrong sign of 1m231 . The 1χ 2 values printed in the following Figs. In the other two figures we used the information on |1m231| from the fit in the sin2 2θ23 -|1m231| plane to reject 1m231 values which are inconsistent with the |1m231| value obtained before. weighted to correct for the cross section differences. 5 The following plots assume that this analysis is done and we analyze then with sin2 2θ23 fixed the matter induced effects and extract sin2 2θ13 and the sign of 1m231 . . which shows the results of a fit to the appearance channel only (wrong sign muon events). which does not include the disappearance events in the analysis. Note. 8 and discussion).

31–35]. If charge identification is available then there is however an even better strategy by combining all available information. Now high precision is obtained both for θ13 and in the determination of 1m231 .e. The contour lines of our fits have to be extended into the third sin2 2θ23 dimension and the precision is roughly given by the results in Ref. there is no need to include further parameters like 1m221 and the CP-phase δ as long as the LMA MSW solution is not realized with a value of 1m221 close to its current upper limit of order O(10−4 ) [29. a global fit to the matter effects of both the appearance and disappearance channels. Such a scenario would however also enable a measurement of the CP-violating effects at . 5. Freund et al.e.. 7.118 M. If this scenario were however realized. the sum of unoscillated (νµ → νµ ) and oscillated (ν¯ e → ν¯µ ) muon neutrinos) for 1m231 > 0 at a baseline of 7332 km. i. The rectangles denote the parameter pair for which the data were generated and the stars denote the obtained best fit. Shown are the 1σ and 2σ contours. Fit to the muon neutrino spectrum without charge identification (i.o.f for the best fit with the wrong sign of 1m231 . such that CP-effects play a role. sin2 2θ13 and sin2 2θ23 . then this would be another reason to go for measurements based on matter effects to largest possible baselines like 7332 km. The results of this fit are shown in Fig. The point is that the relative magnitude of effects coming from 1m221 6= 0 and δ decrease for longer baselines [29]. [6]. As explained above in the analytic discussion. The numbers printed next to each case are the values of χ 2 per 2 d. / Nuclear Physics B 585 (2000) 105–123 Fig. The presented fits were all done with a fixed mixing angle sin2 2θ23 = 1 and it is in principle no problem to perform a full three parameter fit for 1m231 ..

The black rectangles denote the parameter pair for which the data were generated and the stars denote the best fit. Charge separation capabilities which allow to use the statistically favored pure appearance channel are thus very important if sin2 2θ13 turns out to be smaller than ' 10−2 .f values for the local minima of the same parameters. but with the wrong sign of 1m231 . Fit to the muon neutrino appearance (ν¯ e → ν¯ µ ) spectrum with charge identification for 1m231 > 0 at a baseline of 7332 km. Shown are the 1σ and 2σ contour lines. Thus in the case of positive 1m231 the confidence level at which the wrong sign can be excluded is slightly better. For positive 1m231 the cross section favored neutrino channels show resonant matter effects whereas for negative 1m231 the cross section disfavored antineutrino channels are resonant. The solid lines are based on an analysis of the combined channels without charge identification.o. 8 shows finally the region of the 1m231 -sin2 2θ13 parameter space where a determination of the sign of 1m231 will be possible. Shown are the 1σ . / Nuclear Physics B 585 (2000) 105–123 119 Fig. The global minimum is excluded in this case because its 1m231 is already excluded. shorter baselines such that two experiments. Freund et al. Fig. The dotted lines show the limits which can be obtained with charge identification from the spectral data of the appearance channels.M. The numbers next to the contour lines are χ 2 per 2 d. one at a large baseline (for matter effects) and another a shorter baseline (say 2800 km for CP-violating effects) would be ideal. 2σ and 3σ contour lines of a fit with the wrong sign of 1m231 for positive 1m231 (left plot) and negative 1m231 (right plot). 6. .

But there are other issues which may turn out to be equally important as the discussion of statistics and rates which was presented here. / Nuclear Physics B 585 (2000) 105–123 Fig. for example. but the wrong sign of 1m231 . The global minimum is excluded because its 1m231 value is out of range.. The rectangles denote the parameter pairs for which the data were generated and the stars denote the best fits. right sign muon rejection (i. Many of these issues are still under discussion [20–22]. by increasing Nµ± NkT µ ) since the signal to background ratio stays constant. charge identification) turns out to be less good than hoped..e. νµ → νµ and ν¯ e → ν¯ µ channels separated) for 1m231 > 0 at a baseline of 7332 km. The problem cannot be overcome by increasing statistics (i. From this discussion it seems clear that one should include charge identification.o. which would lead in average to 10 background events in the appearance channel which limits the ultimate sensitivity to sin2 2θ13 considerably. Freund et al.e. If. To illustrate the requirements for charge identification assume 100000 muon events in a detector (which corresponds roughly to a baseline of 3000 km and a beam energy of 50 GeV with the usual beam and detector parameters) and right sign charge rejection of 10−4 . Shown are the 1σ and 2σ contours.f with the same parameters. Any real detector may thus be limited by its right sign charge rejection capability and the number .120 M. or if there were background issues which make it hard or impossible to isolate correctly the wrong sign νe → νµ muon signal then our proposed method without charge identification should still work.e.. The numbers printed next to the contour lines are the local minima of χ 2 per 2 d. Fit to the muon neutrino spectrum with charge identification (i. 7.

Sensitivity to the sign of 1m2 . where the dominant non-oscillated neutrino rates are high. improving the signal to background ratio for longer baselines. Freund et al. even for the longest baselines. The amount of background is typically directly proportional to the disappearance rates. This would favour larger baselines. 2σ and 3σ contour lines of a fit with the wrong sign of 1m2 . The dotted lines are based on spectral data from the appearance channel only. Shown are the 1σ . We discussed the relevant appearance and disappearance rates analytically and we performed numerical simulations which were used to test the parameter extraction from event rates in a statistical reliable way. Studies of charge identification capabilities are presently performed [8. Our results show for large baselines like 7332 km that there are matter effects of comparable size in the event rates in the νe → νµ and . 8. |1m231 | = 3.20–22]. (7332 km. The solid lines use the combined data from the appearance and disappearance channels without separation of right and wrong sign muons. of right sign muons. since the disappearance rates drop faster than the appearance rates. / Nuclear Physics B 585 (2000) 105–123 121 Fig. The analysis rests on the detection of muons coming from muon-neutrinos or antineutrinos and we distinguish detectors with and without right sign muon charge rejection capabilities. The decreased statistics in the sin2 2θ23 and |1m231| determination (with a precision of the order of a few percent) is not a problem since this will be limited by systematics and not statistics.5 × 10−3 eV2 and Nµ± NkT µ = 2 × 1021 ). This is also connected to the question which baseline should be used. 7.M. This may be important since the νe → νµ and ν¯ e → ν¯µ appearance channels have rather small total event rates and they require very good efficiencies for the detection of wrong sign muon events. Conclusions We studied in this paper different possibilities to use matter effects to determine the value and the sign of 1m231 and the magnitude of sin2 2θ13 in very long baseline experiments with neutrino factories. Backgrounds due to charge misidentification play an especially important role for very small sin2 2θ13 and at too short baselines.

We mentioned however that a too short baseline could potentially also be dangerous if one neglects imperfect charge separation and other backgrounds. In this case the fit of 1m231 . / Nuclear Physics B 585 (2000) 105–123 ν¯ e → ν¯µ appearance and in the νµ → νµ and ν¯ µ → ν¯ µ disappearance channels. We restricted our analysis therefore to a baseline of 7332 km which is very well suited for our study. generated from the combined appearance and disappearance channels. Matter effects in the disappearance channels alone are however statistically somewhat disfavored.. In this case one can measure or limit the mixing down to sin2 2θ13 ' 10−4 .e. We gave exclusion plots which display the parameter range where a determination of the sign of the mass squared difference 1m231 will be possible. improvements on the knowledge of some parameters and the number of different baselines (i. add in a constructive way. With charge identification one can get better results from the appearance channels alone. . This new method allows thus to measure or limit the mixing sin2 2θ13 roughly one order of magnitude below the present CHOOZ limit. In the limit 1m221 = 0 CP-violating effects drop out and matter effects can be understood in an effective two neutrino scheme which is inserted into the full three neutrino oscillation formulae. This allows a determination of sin2 2θ13 and the sign of 1m231 even without charge identification down to mixings sin2 2θ13 = 10−2 and with a precision comparable to the appearance channels with perfect charge separation. It should however be stressed that there will be sizable corrections to this picture for shorter baselines if 1m221 is at its upper limit. such that larger baselines are safer without loosing detection capabilities for matter effects. It seems therefore that larger baselines above 3000 km are also preferred from this point of view. since they have to be extracted from the relatively large amount of un-oscillated events. Acknowledgements We wish to thank S.e. since it gives a spectrum centered around the resonance energy of matter effects in earth and which seems to be preferred by recent studies of entry level neutrino factories [19].122 M. Romanino for discussions on subjects related to this work. i. Without sufficient charge identification capabilities there is however still the advantage that matter effects in the combined muon rates resulting from νµ and ν¯µ . Our results were obtained in the approximation where 1m221 corrections are negligible which allowed an analytic description. sin2 2θ13 and sin2 2θ23 had to include in addition 1m221 and the CP-phase δ. beams and experiments) available. These extra 1m221 6= 0 effects become however much smaller for longer baselines. Petcov and A. We used a muon beam energy of 20 GeV.. Freund et al.T. We pointed however out that with charge identification the best results can be obtained by combining all information from the appearance and disappearance channels. We did not discuss in more detail baseline and beam energy optimization since it depends on the preferred quantities.

L. O. De Rujula. Talk presented at the FNAL twoday meeting: 17–18th February. A. 9910387. S. Campanelli. Barger. Raja. Geer. J. E. 77 (1999) 207. Nucl. hep-ph/9912457. A. hep-ph/0003184. Apollonio et al. http://lyopsr. Cervera. Nuovo Cimento C 9 (1986) 17.gov/). Barger. V. Nucl. A. hep-ph/9909469. Hernandez. http://www. Barger. K.K. Wolfenstein. hep-ph/9911524. S. Dick. Lyon. P. Phys.Yu. Gavela. K.gz. R. Steigman. P. Nufact ’99 Workshop. Eur. A. S.fr/TAUP99/). Rev. K. S. hep-ph/9912231. K. Rigolin. Bueno.M. S.F. hep-ph/0002108. Caso et al. Rev.J. Gavela. 1977. France (transparancies available at http://taup99. Phys. http://www. Nucl.gov/projects/muon_collider/nu/study/ study. Wiley. 77 (1999) 26. Phys. T. Freund.fnal. K. Mocioiu. LSND Collaboration. Fiz. D. Phys. Kuo. Phys. Valle. Rev.T. Schechter. A. Toshev.fnal. Geer. Langacker. A. Phys. A. F. Proc. L. J. M. Dziewonski. M.. hep-ph/9911258. Bilenky. Rigolin. hep-ph/9910554. S. Lett.D. Hernandez. Geer. http://www. J. Yad.html. Shrock. Scholberg. Romanino. Hosek. Koike. 1999. Freund. hep-ph/0002149. M. e. M. Stacey. July 5–9th. Donini. J. Anderson. New York. Lindner. Nucl.in2p3. hep-ex/9907037. V. Phys. Nucl. Gomez Cadenas. / Nuclear Physics B 585 (2000) 105–123 123 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] S. Rev.B.gov/projects/muon_collider/nu/study/ study. M.gov/projects/muon_collider/nu/study/ study. 2nd edn. hep-ph/9906487. Rev.html. Freund et al. Sov. 42 (1985) 1441. C. V. S. Super-Kamiokande Collaboration. Mena.fnal. A. to be published in Nucl. Nucl.M. Whisnant. B 547 (1999) 21.T. Palazzo. 76 (1999) 425. See. Talk presented at the FNAL twoday meeting: 17–18th February. 2000. Whisnant. D 57 (1998) 6989. M. S.B. J. Petkov. the Summary of Detector/Neutrino Beam Parameters by D. Lisi. and 1999 off-year partial update for the 2000 edition available on the Particle Data Group WWW pages (http://pdg. hep-ph/9909425. A.. M. B 562 (1999) 29. hep-ph/9909501. P. S. J.L.html.W. Paris. Nakahata. Phys. J.B. G. M. 20 (1979) 2634. J. G. Phys. Phys. Romanino. hep-ph/9909254. D. S. GNO Collaboration.. Harris. M.g. Raja. Smirnov. Petcov. M. Whisnant. D 59 (1999) 039903. Talk presented at the FNAL twoday meeting: 17–18th February. Sato. Proc. Hernandez. Phys. 2000. Romanino. CHOOZ collaboration. M.D. D 17 (1978) 2369. P.. hep-ph/9905240. Donini. 2000. Geer for the Neutrino Factory Physics Study Group. K. B 198 (1987) 406. P.fr/nufact99/talks/harrisconc. Fogli. http://www. Ohlsson. Physics of the Earth. D. Spentzouris for the Neutrino Factory Physics Study Group. A. Phys. Super-Kamiokande Collaboration. Talk presented at the FNAL twoday meeting: 17–18th February.fnal. Phys B. Nucl. B 282 (1987) 589. Phys. Lett. Phys. Montanino. Gavela. A. R. Rubbia.gov/projects/muon_collider/nu/study/ study. Petcov.html. Earth Planet Interiors 25 (1981) 297. Pantaleone.lbl. Freund. S. Harris for the Neutrino Factory Physics Study Group. Phys. Geer. McFarland for the Neutrino Factory Physics Study Group. Mikheyev. R. . GALLEX Collaboration. P. C 3 (1998) 1. Phys. 2000. Suppl. Suppl. B 94 (1980) 495. A.in2p3. G. Proc. Suppl.P. 42 (1985) 913.ps. hep-ex/9905016.T. I. VIth International Workshop on Topics in Astroparticle and Underground Physics. Lindner. Erratum. M. September 6–10. D 23 (1981) 1666. in: TAUP’99.

R. must be electrically neutral and weakly interacting. for mSUGRA at large tan β.e. College Station.Pb. Accomando. Nonuniversality 1. 12.e. (0. PACS: 12. Arnowitt. revised 22 May 2000 Abstract The neutralino–proton cross section is examined for supergravity models with R-parity invariance with universal and non-universal soft breaking.Jv Keywords: Supersymmetry. m0 can be quite large reducing the tension between proton decay bounds and dark matter analysis. Experimental bounds on exotic isotopes strongly imply that the lightest supersymmetric particle (LSP).V.elsevier. In gravity mediated supergravity (SUGRA) grand unified models (GUTs) ([1–4]. is examined. In such models. the allowed region in the supersymmetry (SUSY) parameter space where the ν˜ is the LSP is generally small. i.Nuclear Physics B 585 (2000) 124–142 www. B.60. 11. USA Received 25 April 2000.. The minimal supersymmetric model (MSSM) then has two possible candidates. The region of parameter space that dark matter detectors are currently (or will be shortly) sensitive.10. Texas A&M University. i. Department of Physics.60. For universal soft breaking (mSUGRA). for reviews see [2–4]). PII: S 0 5 5 0 .nl/locate/npe Neutralino–proton cross sections in supergravity models E. Y. All rights reserved.3 2 1 3 ( 0 0 ) 0 0 3 2 1 . which is absolutely stable by R-parity invariance. and for nonuniversal models with small tan β. Unification. the lightest neutralino (χ˜ 10 ) and the sneutrino (ν˜ ). Current relic density bounds restrict mχ˜ 0 6 120 1 GeV for the maximum cross sections. Introduction Supersymmetric models with R-parity invariance generally predict the existence of dark matter relics from the Big Bang. The absence of the decay Z → ν˜ + ν˜ at LEP implies mν˜ > 45 GeV and LEP bounds on the light Higgs is 0550-3213/00/$ – see front matter  2000 Elsevier Science B.  2000 Elsevier Science B. Santoso Center For Theoretical Physics. Nonuniversal soft breaking models can allow much larger cross sections and can sample the parameter space for tan β & 4. Universality. 12. Dutta. detectors with sensitivity σχ˜ 0 −p > 1 × 10−6 pb will be able to 1 sample parts of the parameter space for tan β & 25..7 .-g.V. All rights reserved. which is below where astronomical uncertainties about the Milky Way are relevant. TX 77843-4242.1−10) × 10−6 pb.-i.30. We note the existence of two new domains where coannihilation effects can enter.

It is an appealing feature then of SUGRA models that the predicted amount of relic density of neutralino CDM is consistent with what is observed astronomically for a significant part of the SUSY parameter space. and perhaps a factor of 10 improvement 1 may be expected in the near future. there has been a great deal of theoretical analysis and experimental activity concerning the detection of local CDM particles..g. the χ˜ 10 is the unique candidate for cold dark matter (CDM). not all authors include all the constraints on the parameter space described below. and ρc = 3H02/(8πGN ) (H0 = Hubble constant parameterized by H0 = (100 km s−1 Mpc−1 )h.8] and the subsequent suggestion that local χ˜ 10 in the Milky Way might be observed by terrestrial detectors [9]. Thus Refs.25. 1 mh > 95 GeV) and include the b → s + γ and Tevatron constraints.1 × 10−6 pb 6 σχ˜ 0 −p 6 10 × 10−6 pb. Refs. which allows one to extract from the data the spin independent neutralino–proton cross section σχ˜ 0 −p . Since the initial observation that the χ˜10 represented a possible CDM particle [7. [24. CDMS) is 1 approximately (1−10) × 10−6 pb for σχ˜ 0 −p [30]. . and no work has been done combining all types of non universal breaking terms for different values of tan β (small and large). The neutralino–nucleus scattering amplitude contains spin independent and spin dependent parts. we restrict the SUSY parameter space to be consistent with the 1 astronomical estimates of the amount of relic CDM. [26] and will be seen below. [26–28] include also nonuniversal effects in the third generation (but these analysis hold only for small tan β). Recent measurements of Ωm = ρm /ρc (ρm is the matter density).. DAMA. 1 (1) as one varies SUSY parameters (e. Our calculations are done within the 1 framework of SUGRA GUT models with non-universal soft breaking allowed in both the Higgs and third generation squark and slepton sectors. If one further includes cosmological constraints.E. This is conventionally measured by the quantity ΩCDM = ρCDM /ρc where ρCDM is the mean CDM mass density. It is the purpose of this paper to examine what part of the SUSY parameter space can be tested with such a sensitivity.g. and calculations in Refs. Further. In calculating σχ˜ 0 −p . [10– 14] assume the MSSM model. mχ˜ 0 ). [10–29] have made use of a number of different SUSY models. We find. in fact. Accomando et al.28] allow for nonuniversal soft breaking in the Higgs sector and Refs. / Nuclear Physics B 585 (2000) 124–142 125 sufficient to eliminate the ν˜ as the LSP for the minimal SUGRA model (mSUGRA) [5]. it is necessary to include both Higgs and third generation nonuniversalities as the two can have constructive interference. the neutron and proton scattering amplitudes are approximately equal.) We also update earlier analyses by including the latest LEP bounds on the light chargino (χ˜ 1± ) and light Higgs (h) mass (mχ˜ ± > 94 GeV. the spin independent part dominates. (As discussed in Ref. In these. that the additional nonuniversal effects in the third generation open significant additional parameter space. [15–23] are performed using mSUGRA GUT models (with universal soft breaking at the GUT scale MG ∼ = 2 × 1016 GeV). tan β. We do this by examining the maximum theoretical cross section that can lie in the domain 0. Thus for these models. GN = Newton constant). The sensitivity of current experiments (e. Recent theoretical calculations in Refs. However for detectors with heavy nuclei. the sneutrino is also excluded for the general MSSM [6].

though this is somewhat relaxed in the domain 0. 2. In Section 4 we discuss the nonuniversal models.34].) We examine tan β in the range 2 6 tan β 6 50. A brief qualitative discussion is also given there of the effect of these results on proton decay since the above nonuniversal results appear to releave some of the tension previously noted [27] between σχ˜ 0 −p in the 1 range of Eq. and include leading order (LO) corrections to the b → s + γ decay and correct approximately for NLO effects [21.33] (which will be discussed in a subsequent paper). and possible CP violating phases are set to zero. Naturalness constraints. We start with a set of parameters at MG . Recent measurements at the Hubble Space Telescope using a number of different 1 techniques has led to a combined average of [37] . / Nuclear Physics B 585 (2000) 124–142 H0 . and see that here Eq. Accomando et al. 1 1 In our analysis below we use the one loop renormalization group equations (RGE) [31] from MG to the t-quark mass mt = 175 GeV and impose the radiative breaking constraint at the electroweak scale. and use one loop corrections to the b-quark mass so that mb takes on its experimental value mb (mb ) = (4. Gaugino masses are assumed universal at MG . supernovae data. and iterate until a consistent spectrum is obtained. Conclusions are summarized in Section 5. (We find. One loop corrections are included in diagonalizing the Higgs mass matrix. the Cosmic Microwave Background (CMB). integrating out the heavy particles at each threshold. (1) and current Super Kamiokande proton lifetime bounds [36]. We will see below that both the upper and lower bounds on Ωχ˜ 0 h2 strongly affect the predicted values of σχ˜ 0 −p .1 × 10−6 pb 6 σχ˜ 0 −p 6 1 1 × 10−6 pb. In Section 2 we discuss the range of the astrophysical parameters that enter into the relic density analysis. etc. We require that the theoretical branching ratio lie in the range 1.. that this is a significant constraint with nonuniversal soft breaking even for low tan β. The SUSY mass spectrum expected for our domain of cross sections is also 1 examined. Astronomical and quark parameters The basic experimental quantity that controls the SUSY analysis of relic density is Ωχ˜ 0 h2 .1 − 4. (1) can be satisfied for relatively small tan β and large m0 . and also that σχ˜ 0 −p sustains for 1 large mχ˜ 0 .5 × 10−4 . in fact. and also the uncertainties of the quark content of the proton which affect our calculation of σχ˜ 0 −p . The loop correction is significant for large tan β † 0† H2 + h. Thus we do not treat here D-brane models [32. and L–R mixing is included in the sfermion mass matrices so that large tan β may be treated. In Section 3 we examine the mSUGRA model where it 1 is seen that σχ˜ 0 −p > 1 × 10−6 pb (the current experimental sensitivity) requires tan β to 1 be quite large. that the gluino mass obey mg˜ 6 1 TeV. indicate that ΩCDM is smaller then previously thought. (See Appendix A.5) GeV [29].126 E.c. The SUSY mass spectrum is also constrained so that coannihilation effects are negligible.9 × 10−4 6 BR(B → Xs γ ) 6 4.) We do not and stems from the part of the Lagrangian given by −µ∗ λb b˜L b˜R assume any particular GUT group constraints and do not impose b − τ Yukawa unification as done in [28] (since the latter is sensitive to possible unknown GUT physics). the scalar mass m0 6 1 TeV and |A0 /m0 | 6 5 are imposed.

In view of possible systematic errors. Recently. We follow the procedure of Ref.02 6 Ωχ˜ 0 h < 0. An analysis of combined data (excluding microlensing) yields [42] Ωm = 0. [21]. σπN (10) The quark mass ratios are fairly well known. (9) and y= hp|¯s s|pi 1 ¯ ¯ + dd|pi 2 hp|uu ≡1− σ0 . (3) 1 Combining errors in quadrature then yields Ωχ˜ 0 h2 = 0. (11) . and since the baryonic content is ΩB ∼ = 0.25.3 ± 0. [14].10.41]. mp (6) where σπN is the (π − N ) σ -term and is given by 1 ¯ ¯ + dd|pi. and these results are consistent with the supernovae data [40. They find 40 MeV .08. the uncertainties in σπN and σ0 have been analyzed in Ref.25 ± 0.126 ± 0. Measurements on clusters of galaxies yield Ωm < 0.5 [43]. 65 MeV. Accomando et al.23 ± 0. The fundamental SUSY Lagrangian allows one to calculate the neutralino–quark scattering amplitude. and we use in the following r = 24. the quark content of the proton.E. we take Ωχ˜ 0 = 0. 2 (4) 1 (As pointed out in Ref. around the mean: 1 0.05 [38. 30 MeV . σπN = (mu + md )hp|uu 2 f can be written as [14] 1 σπN . (5) fˆ = mp and f= hp|ms s¯ s|pi . 40 MeV.052. 127 (2) There is now sufficient data on the CMB anisotropies to show that Ωtot ' 1 and Ωm small is strongly favored [38]. / Nuclear Physics B 585 (2000) 124–142 H0 = (71 ± 3 ± 7) km s−1 Mpc−1 .05. the lower bound is the minimum amount of DM to account for the rotation curves of spiral galaxies. In the following we 1 will restrict the range of Ωχ˜ 0 h2 by what is approximately 2 std. σ0 . the two parameters that enter sensitively are σπN . 39].32 ± 0.4 ± 1. [14].1. σπN . [44].) Future measurements by the MAP and Planck sattelites will greatly reduce these errors. To obtain the χ˜ 10 −p cross section one needs to know in addition. we assume here Ωm = 0. In the notation of Ref. f = ry 2 mp (7) (8) where r= ms 1 2 (mu + md ) .

vrel the relative velocity. and (Tχ˜ 0 /Tγ )3 is the reheating factor. the sf 1 channel pole gives rise to a large amount of annihilation (which can reduce Ωχ˜ 0 h2 below 1 . It is convenient sometimes to replace m1/2 by the gluino mass mg˜ ∼ = (α3 /αG )m1/2 (αG ∼ = 1/24 is the GUT scale gauge coupling constant) or mχ˜ 0 1 which also scales with m1/2 . Set 2: σπN = 65 MeV.2 i gives rise to (down. From Fig. [14]. [14] gives considerably larger cross sections. Accomando et al. 1 The relic density decreases with increasing annihilation cross section. (4). and in order to understand some of the results obtained below.48 × 10 R xf 1 Tγ 2. but we will exhibit the difference between Set 1 and Set 2 in one case to illustrate some of the uncertainties that exist. we first discuss which parameters control σann .128 E. A convenient choice of parameters is m0 (the universal scalar mass at MG ). Nf is the number of 1 degrees of freedom at freezeout.) In calculating σχ˜ 0 −p . σ0 = 30 MeV. (15) Ωχ˜ 0 h = 2. fˆ = 0.455. f = 0. 3. 46]: dnχ˜ 0  R˙ 1 + 3 nχ˜ 0 = −hσann vrel i n2χ˜ 0 − n2eq .0693. the b → s + γ constraint eliminates mostly µ > 0. (12) Set 1 corresponds approximately to the original analysis of Ref. The mSUGRA model We begin our analysis by examining the minimal SUGRA model which depends on four parameters and the sign of the Higgs mixing parameter µ.73 0 dx hσann vrel i where xf = kTf /mχ˜ 0 ' 1/20. The final relic density is given by 1/2   T 0 3  Nf Tγ 3 χ˜ 1 2 −11 . mH or mA (but lies below). Our sign convention on the µ parameter is defined by the quadratic term in the superpotential  (13) W (2) = µH1 H2 = µ H10 H20 − H1− H2+ . fˆ = 0. / Nuclear Physics B 585 (2000) 124–142 In the following we will consider two possible choices for σπN and σ0 : Set 1: σπN = 40 MeV. The diagrams governing σann are shown in Fig. neq its equilibrium value.) In the following. A0 (the cubic soft breaking mass at MG ) and tan β = hH2 i/hH1 i (where hH1. However. 1 one expects σann to fall with increasing mχ˜ 0 and also increasing m0 (since 1 m2˜ increases with m20 ). if 2mχ˜ 0 is near mh . σann is the annihilation 1 cross section. up) quark masses). σ0 = 30 MeV. f = 0. and h i means thermal average. (14) dt R 1 1 where nχ˜ 0 is the number density of χ˜ 10 . 1. [44] (and is the most conservative possibility) while Set 2 is similar to the Set 2 of Ref. (With this convention. Tf is the freezeout temperature. one must impose the relic density constraint of Eq. This is 1 governed by the Boltzmann equation describing χ˜ 10 annihilation in the early universe [45. m1/2 (the universal gaugino mass at MG ).195. we will use mostly Set 2 in showing our result. (Set 3 of Ref.0480.

the t-channel annihilation will drive Ωχ˜ 0 h2 down [51]. 2. if one of the sleptons or squarks becomes light. the allowed minimum) and due to the thermal averaging. / Nuclear Physics B 585 (2000) 124–142 129 Fig. the maximum value of σχ˜ 0 −p for tan β = 20. i. H . Diagrams for neutralino–quark scattering. A) and Z poles and squark and slepton (f˜) poles. which is governed by the diagrams of Fig. Further. 1 1 In order now to see the sensitivity of current detectors to mSUGRA we plot in Fig. Fig. we will see that since H and A become light at large tan β. The LEP data. f We turn next to σχ˜ 0 −p . 30. However. effects of this type become significant in that regime.E. This effect can 1 again become significant at large tan β where large L–R mixing in the sfermion mass matrices reduces m2˜ .. Diagrams for early universe annihilation of χ˜ 10 through Higgs (h. 40 1 . These regions of parameter space are just the ones that reduce Ωχ˜ 0 h2 . ' 100 GeV. this effect can be significant when 2mχ˜ 0 is less than the Higgs mass and within five times the Higgs width of the 1 Higgs mass [47–50]. Accomando et al. We see here that the 1 cross section can become large for light (first generation squarks) and light Higgs bosons. has eliminated most of this effect for the light Higgs. and so there can be 1 a bound produced on σχ˜ 0 −p so that Ωχ˜ 0 h2 does not fall below its minimum. 3 (for Set 2 parameters of Eq. (12)).e. 1. 2.

130

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

Fig. 3. The maximum value of σχ˜ 0 −p vs. mχ˜ 0 for tan β = 20, 30, 40, and 50 for Set 2 parameters of
1
1
Eq. (12).

Fig. 4. mH vs. mχ˜ 0 when σχ˜ 0 −p takes on its maximum value. Top curve is for tan β = 30, bottom
1
1
curve for tan β = 50.

and 50 as a function of mχ˜ 0 (obtained by allowing all other parameters to vary subject
1
to the constraints listed in Sections 1 and 2). We see the expected fall off with increasing
mχ˜ 0 . The current DAMA experiment is thus sensitive to mSUGRA for tan β & 25 (i.e.,
1

σχ˜ 0 −p & 1.0 × 10−6 pb). We note that the fall off is less severe for tan β = 50, since at
1

this high value of tan β the H and A Higgs become relatively light enhancing the χ˜ 10 –p
cross section. This can be seen in Fig. 4 where we have plotted mH for tan β = 30 and
tan β = 50. We note also the importance of including the loop corrections to mb for large
tan β (e.g., tan β = 50) to obtain the correct results here.

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

131

Fig. 5. Maximum σχ˜ 0 −p vs. mχ˜ 0 for tan β = 30 for Set 2 parameters (solid), and Set 1 parameters
1
1
(dashed). See Eq. (12).

Fig. 6. Ωχ˜ 0 h2 vs. mχ˜ 0 for tan β = 30.
1

1

Fig. 5 shows the sensitivity of the calculations to the choice of particle physics
parameters. Set 1 gives cross sections about a factor of 2 smaller than Set 2. In the
following, we will use Set 2 in all our analysis.
Fig. 6 shows Ωχ˜ 0 h2 vs. mχ˜ 0 for tan β = 30. We see, as expected, Ωχ˜ 0 h2 is an increasing
1

1

1

function of mχ˜ 0 (since σann is a decreasing function). The upper bound on Ωχ˜ 0 h2 then
1
1
implies an upper bound of the neutralino mass of mχ˜ 0 ∼
= 120 GeV (i.e., mg˜ . 900 GeV)
1
as has been discussed previously [16–18]. Note that one can obtain cross sections within
the DAMA sensitivity range without going to the edges of the parameter space in Ωχ˜ 0 h2 .
1
Also, these cross sections all fall below the current experimental sensitivity before the
uncertainties in the Milky Way astronomical parameters discussed in [12,13] become
important.

132

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

Fig. 7. md˜ vs. mχ˜ 0 for tan β = 30 when σχ˜ 0 −p takes on its maximum value.
1

1

Fig. 8. mh vs. mχ˜ 0 for tan β = 30 when σχ˜ 0 −p takes on its maximum value.
1

1

Figs. (7)–(9) exhibit the particle spectrum expected for the example of tan β = 30, when
σχ˜ 0 −p takes on its maximum value. Fig. 7 shows that the d-squark is quite heavy (m0
1
is large). This arises from our constraint mτ˜R − mχ˜ 0 > 25 GeV to prevent coannihilation
1
effects from occurring. Thus the large tan β being considered here reduces mτ˜R (due to
L–R mixing) and m0 must be increased to prevent it from becoming degenerate with the
χ˜ 10 . These coannihilation effects are thus different from the ones that can occur at low
tan β [24], since they occur at low mχ˜ 0 where σχ˜ 0 −p is large enough to fall within the
1
1
range of Eq. (1). (They will be discussed elsewhere.) Fig. 8 shows the light Higgs mass,
which is relatively heavy due to the fact that m0 is large. Fig. 9 shows mχ˜ ± vs mχ˜ 0 for
1
1
tan β = 30. One sees that scaling is obeyed [52], i.e., mχ˜ ± ∼
= 2mχ˜ 0 since µ is relatively
large (µ2 /MZ2  1).

1

1

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

133

Fig. 9. mχ˜ ± vs. mχ˜ 0 for tan β = 30 when σχ˜ 0 −p takes on its maximum value.
1

1

1

4. Nonuniversal models
Nonuniversal soft breaking can arise in SUGRA models if in the Kahler potential, the
interactions between the fields of the hidden sector (that break supersymmetry) and the
physical sector are not universal. Nonuniversalities allow for a remarkable increase in the
neutralino–proton cross section.
In order to suppress flavor changing neutral currents (FCNC), we will assume the first
two generations of squarks and sleptons are universal at MG (with soft breaking mass m0 )
but allow for nonuniversalities in the Higgs and third generation. Thus we parameterize the
soft breaking mass at MG as follows:
mH21 = m20 (1 + δ1 ),
mu2R
mq2L

= m20 (1 + δ4 ),
= m20 (1 + δ3 ),

mH22 = m20 (1 + δ2 ),

(16)

= m20 (1 + δ5 ),
= m20 (1 + δ6 ),

(17)

me2R
md2L

mlL2 = m20 (1 + δ7 ),

(18)

where m0 is the universal mass of the first two generations, qL = (t˜L , b˜L ); lL = (˜νL , τ˜L );
uR = t˜R ; eR = τ˜R , etc. We take here the bounds
−1 6 δi 6 1.

(19)

An alternate way of satisfying the FCNC constraint is to make the first two generations
very heavy, and only the third generation light. This is essentially included in the above
parameterization by making m0 large, and taking the δi sufficiently close to −1.
One of the important parameters effected by the nonuniversal soft breaking masses is µ2 ,
which is determined by the radiative breaking condition. While the RGE must be solved
numerically, an analytic expression can be obtained for low and intermediate tan β [26]:  

 

1
t2
1 + D0
δ1
1 − 3D0
1 − D0
2
+ 2 +
(δ3 + δ4 ) −
δ2 + 2 m20
µ = 2
t −1
2
t
2
2
t
+ universal parts + loop corrections,
(20)

134

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

Fig. 10. Maximum value of σχ˜ 0 −p vs. mχ˜ 0 for tan β = 7 for nonuniversal soft breaking (upper
1
1
curve) and universal soft breaking (lower curve).

where t ≡ tan β and  

2
mt
.
D0 ∼
=1−
200 sin β

(21)

A similar expression holds for large tan β in the SO(10) limit, so Eq. (20) gives a qualitative
picture of the effects of nonuniversalities in general (a result borne out from detailed
numerical calculations).
We see first that in general D0 is small, i.e., for mt = 175 GeV, D0 6 0.2, and hence
the squark nonuniversality, δ3 and δ4 , produce comparable size effects as the Higgs
nonuniversalities δ1 and δ2 , so that both must be included for a full treatment [26]. Second,
one can choose the signs of δi such that either µ2 is reduced or µ2 is increased. The
significance of this is that in general, the χ˜ 10 is a mixture of higgsino and gaugino pieces
e3 + β B
e+ γH
e1 + δ H
e2 .
χ˜ 10 = α W

(22)

Now the spin independent part of χ˜ 10 −q scattering depends on interference between the
gaugino and higgsino parts of χ˜ 10 [44] (it would vanish for pure gaugino or pure higgsino)
and this interference increases if µ2 is decreased (increasing σχ˜ 0 −q ) and decreases if µ2
1
is increased (decreasing σχ˜ 0 −q ) [19]. Thus there are regions in the parameter space of
1
nonuniversal models where σχ˜ 0 −p is significantly increased compared to the universal
1
case.
The above effect can be seen in Fig. 10 where the maximum σχ˜ 0 −p are plotted for
1
tan β = 7 for the nonuniversal and universal cases. We see that nonuniversalities can
increase σχ˜ 0 −p by a factor of ' 10. Fig. 11 plots the nonuniversal curves for tan β = 3, 5,
1
and 7. One sees here that with nonuniversal soft breaking, the current DAMA sensitivity
requires tan β & 4 (compared to tan β & 25 in the universal case). For larger tan β one
can get very large nonuniversal cross sections. Fig. 12 shows the maximum σχ˜ 0 −p for
1
tan β = 15, which already lies in the region excluded by CDMS and DAMA.

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

135

(a)

(b)
Fig. 11. Maximum value of σχ˜ 0 −p vs. mχ˜ 0 for (a) tan β = 3, and (b) tan β = 5, 7, for nonuniversal
1
1
soft breaking. Note that the tan β = 3 curve terminates at mχ˜ 0 ∼
= 70 GeV.
1

For GUT groups containing an SU(5) subgroup (such as SU(5), SO(10), SU(6), etc.)
with matter in the usual 10 + 5¯ representations, the δi of Eqs. (17), (18) obey
δ3 = δ4 = δ5 ≡ δ10 ,

δ6 = δ7 ≡ δ5¯ .

(23)

We consider this case in more detail (where it is assumed that the gauge group breaks to
the Standard Model at MG ). Fig. 13 shows Ωχ˜ 0 h2 when σχ˜ 0 −p takes on its maximum
1

1

value for the characteristic example of tan β = 7. One sees that Ωχ˜ 0 h2 is generally small
1

since one has δ10 < 0 to obtain the maximum σχ˜ 0 −p (reducing µ2 in Eq. (20) and hence
1
increasing the cross section). This however reduces mτ˜R (from Eq. (17)) increasing the
annihilation rate as in the discussion of Fig. 1. (If the SU(5)-type constraint were relaxed
and δ5 left arbitrary, Ωχ˜ 0 h2 could be increased. For example, δ5 = 0 produces ≈ 50%
1

increase in Ωχ˜ 0 h2 .) The further fall off of Ωχ˜ 0 h2 for mχ˜ 0 & 110 GeV arises from the fact
1
1
1
that mH ' 300 GeV, and the nearness of the mH s-channel pole of Fig. 1 increases the
early universe annihilation. This can be seen explicitly in Fig. 14 where 2mχ˜ 0 is close to
1

136

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

Fig. 12. Maximum value of σχ˜ 0 −p vs. mχ˜ 0 for tan β = 15 for nonuniversal soft breaking.
1

1

Fig. 13. Ωχ˜ 0 h2 vs. mχ˜ 0 for maximum σχ˜ 0 −p for tan β = 7, nonuniversal soft breaking.
1

1

1

mH when mχ˜ 0 & 110 GeV. Fig. 15 shows that the light Higgs for this case is quite light
1
lying just above the LEP2 bounds. Particularly interesting is that the first two generations
of squarks, however, are relatively heavy. This is shown in Fig. 16 for the d-squark. The
reason for this can be seen from Eq. (20) where since δ3 = δ4 = δ10 < 0 (to lower µ2 and
hence increase σχ˜ 0 −p ) the nonuniversal terms produce a net negative m20 contribution to
1

µ2 , the lowering of µ2 being enhanced, then, the larger m0 is. Thus it is possible to get
heavy squarks in the first two generations at low tan β, which may have implications with
respect to proton decay as discussed in the next section.
5. Conclusions
If the dark matter of the Milky Way is indeed mainly neutralinos, then current detectors
are now sensitive to interesting parts of the SUSY parameter space. Thus either discovery

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

137

Fig. 14. mH vs. mχ˜ 0 for maximum σχ˜ 0 −p for tan β = 7, nonuniversal soft breaking.
1

1

Fig. 15. mh vs. mχ˜ 0 for maximum σχ˜ 0 −p for tan β = 7, nonuniversal soft breaking.
1

1

(or lack of discovery) will determine (or eliminate) parts of the parameter space, and this
analysis is complementary to what one may learn from accelerator experiments.
To examine what parts of the parameter can be tested with current detectors or in
the near future, we have considered σχ˜ 0 −p , the χ˜ 10 –p cross section, in the range 0.1 ×
1
10−6 pb 6 σχ˜ 0 −p 6 10 × 10−6 pb, and have plotted the maximum theoretical cross section
1
for different SUGRA models. There is a major difference between the universal and
nonuniversal soft breaking models. Thus the current DAMA experiment (with sensitivity
of σχ˜ 0 −p & 1 × 10−6 pb) is sensitive to tan β & 25 for universal soft breaking (Fig. 3)
1
while it is sensitive to tan β & 4 for the nonuniversal model (Fig. 11). Thus while dark
matter cross sections increase with tan β and hence detectors are more sensitive at higher
tan β, it is possible for current detectors to probe part of the low tan β parameter space for

138

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

Fig. 16. md˜ vs. mχ˜ 0 for maximum σχ˜ 0 −p for tan β = 7, nonuniversal soft breaking.
1

1

Fig. 17. Example of p-decay diagram. The major contribution comes from the second generation
loop (i = 2), the third generation contributing ≈ 30% correction with arbitrary relative phase.

the nonuniversal models.
For the mSUGRA model, we find that Ωχ˜ 0 h2 monotonically increases with mχ˜ 0 from
1
1
the minimum to the maximum bounds of Eq. (4) (Fig. 6), leading to the upper bound
mχ˜ 0 6 120 GeV (mg˜ . 900 GeV) which is below where astronomical uncertainties about
1

the Milky Way [12,13] become significant. In general µ2 is large (i.e., µ2 /MZ2  1)
leading to the usual gaugino scaling relations, e.g., Fig. 9, and the Higgs mass is relatively
heavy (Fig. 8). At the very largest tan β, e.g., tan β = 50, the loop corrections to λb at the
electroweak scale become very large (see Appendix A), requiring that λb , the b-Yukawa
coupling, be adjusted so that one obtains the experimental b-quark mass [35].
For the nonuniversal model, significantly increased χ˜ 10 –p cross sections can be obtained
by choosing δ3,4 < 0 and δ2 > 0 in Eq. (20). This reduces µ2 , increasing the higgsino
content of the χ˜ 10 , and hence increasing the higgsino–gaugino interference which enters in
σχ˜ 0 −p . (In the SU(5)-like models, this generally leads to a light τ˜R and hence a relatively
1

low Ωχ˜ 0 h2 (Fig. 13).) In this case the maximum cross sections arise with µ2 relatively
1

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

139

small, and so scaling no longer holds accurately, and the light Higgs lies close to the LEP2
bounds (Fig. 15).
While coannihilation effects have not been treated in this analysis, we have noted two
regions where such effects can occur. In mSUGRA models, due to the fact that tan β must
be large to obtain σχ˜ 0 −p in the range of Eq. (1), L–R mixing reduces mτ˜R making the τ˜R
1

near degenerate with the χ˜ 10 . In the nonuniversal case, where tan β is small or moderate,
large σχ˜ 0 −p are obtained by lowering µ2 which makes the χ˜ 1± nearly degenerate with the
1

χ˜ 10 . Both these domains of coannihilation are different that previously treated [23], and they
inhabit regions of parameter space with σχ˜ 0 −p within the reach of current detectors. We
1
have prevented coannihilation here from becoming significant by imposing the constraints
mτ˜R − mχ˜ 0 , mχ˜ ± − mχ˜ 0 > 25 GeV. Further study is required to see what occurs when these
1
1
1
constraints are removed.
It has for sometime been realized that tension exist in GUT theories that simultaneously
allow for dark matter and proton decay [27,53]. Thus for SU(5)-type models, minimal
e3 , the superheavy higgsino color triplet
SUGRA GUT proton decay proceeds through the H
components of the Higgs 5 and 5¯ representations. The basic diagram is shown in Fig. 17,
showing that the decay rate scales approximately by 
2 

1
1 mχ˜1±
,
(24)
Γ (p → ν¯ K) ∼ 2
M3 m2q˜ sin β cos β
e3 mass. In mSUGRA models, scaling is generally a good
where M3 = O(MG ) is the H

approximation and mχ˜ ± = 2mχ˜ 0 . Hence proton stability requires small mχ˜ 0 , large mq˜
1
1
1
and small tan β. We have seen, however, that if dark matter exists with the sensitivity
of the current DAMA experiment, while moderately heavy squark masses could exist in
mSUGRA (Fig. 7), tan β would have to be quite large, i.e., tan β & 25, which would be
sufficient to violate the current Super Kamiokande bounds on the proton lifetime [37].
However, this tension is releaved for the nonuniversal SUGRA GUT models. Thus we saw
in these cases, one could have a σχ˜ 0 −p in the range of the DAMA experiment for small
1
tan β, i.e., tan β & 4, and further such large cross sections also implied large squark masses,
Fig. 16. This would be expected to remove any disagreement between a large σχ˜ 0 −p and a
1
small proton decay rate.
Finally we mention that in this paper we have plotted the maximum χ˜ 10 −p cross sections
for each tan β and mχ˜ 0 . Of course nature may not chose SUSY parameters such that σχ˜ 0 −p
1
1
takes on its maximum value. However, by looking at the maximum σχ˜ 0 −p we are able to
1
see in a given model whether detection of dark matter at current detector sensitivities is
consistent with the predictions of the theoretical model.

Acknowledgements
This work was supported in part by National Science Foundation Grant No. PHY9722090.

140

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

(a)

(b)
Fig. 18. On-loop correction to b-mass.

Appendix A
The b-quark coupling to the down type Higgs field which gives rise to tree level bottom
mass is described by
LbbH = λb b¯L bR H10 + h.c.

(A.1)

There also exists a term in the Lagrangian where the bottom squarks are coupled to the up
type neutral Higgs (H20 ) and is given by:
∗ ˜ ˜† 0†
Lb˜ bH
˜ = −λb µ bL bR H2 + h.c.

(A.2)

The above interaction can give rise to a one loop contribution to the tree level bottom
mass [54,55]. We do the analysis in the mass insertion approximation which produces
errors of less than 10% in mb for the relevant parts of the parameter space. The loop
diagram arising from the above interaction, shown in Fig. 18(a), involves gluino, squark
fields, αs and tan β and hence can be large for large tan β. There also exists another one
loop contribution which involves the stop quarks and the chargino. This loop, shown in
Fig. 18(b), depends on λ2t and contributes less than the gluino loop. The net b-quark mass
generated from the above contributions is mb + δmb , where
δmb = λb v1 K tan β,

v1 = hH10 i,

(A.3)

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

K '−  

2αs
λ2t
mg˜ µG m2b˜ , m2b˜ , m2g˜ −
At µG m2t˜ , m2t˜ , µ2 ,
2
L
R
L
R
(3π)
(4π)

141

(A.4)

where
G(a, b, c) =

ab Log[a/b] + bc Log[b/c] + ac Log[c/a]
,
(a − b)(b − c)(a − c)

(A.5)

mg˜ is the gluino mass, mb˜L,R are the left and right handed sbottom masses and mt˜L,R are
the left- and right-handed stop masses.
The correction K is evaluated at the electroweak scale which we take here to mt (the
endpoint of running the RGE down from MG ). Using the RGE for λb , we then determine
λb (mt ) so that the total b-quark mass, mb = λb v1 + δmb , agrees with the experimental
value of mb (mb ) [36] at the b scale. This produces a significant change in λb for large
tan β.
References
[1] A.H. Chamseddine, R. Arnowitt, P. Nath, Phys. Rev. Lett. 49 (1982) 970.
[2] P. Nath, R. Arnowitt, A.H. Chamseddine, Applied N = 1 Supergravity, World Scientific,
Singapore, 1984.
[3] H.P. Nilles, Phys. Rep. 110 (1984) 1.
[4] R. Arnowitt, P. Nath, in: Proc. of VII J.A. Swieca Summer School, World Scientific, Singapore,
1994.
[5] R. Arnowitt, P. Nath, Phys. Rev. D 54 (1996) 2374.
[6] T. Falk, K.A. Olive, M. Srednicki, Phys. Lett. B 339 (1994) 248.
[7] H. Goldberg, Phys. Rev. Lett. 50 (1983) 1419.
[8] J. Ellis, J.S. Hagelin, D.V. Nanopoulos, K. Olive, M. Srednicki, Nucl. Phys. B 238 (1984) 453.
[9] M. Goodman, E. Witten, Phys. Rev. D 31 (1985) 3059.
[10] M. Drees, N. Nojiri, Phys. Rev. D 48 (1993) 3483.
[11] A. Bottino, F. Donato, N. Fornengo, S. Scopel, Phys. Rev. D 59 (1999) 095003.
[12] M. Brhlik, L. Roszkowski, hep-ph/9903468.
[13] P. Belli et al., hep-ph/9903501.
[14] A. Bottino, F. Donato, N. Fornengo, S. Scopel, hep-ph/9909228.
[15] S. Kelley, J. Lopez, D. Nanopoulos, H. Pois, K. Yuan, Phys. Rev. D 47 (1993) 246.
[16] R. Arnowitt, P. Nath, Phys. Rev. Lett. 70 (1994) 3696.
[17] G. Kane, C. Kolda, L. Roszkowski, J. Wells, Phys. Rev. D 49 (1994) 6173.
[18] H. Baer, M. Brhlik, Phys. Rev. D 53 (1996) 597.
[19] R. Arnowitt, P. Nath, Phys. Rev. D 54 (1996) 2374.
[20] H. Baer, M. Brhlik, Phys. Rev. D 55 (1997) 3201.
[21] H. Baer, M. Brhlik, Phys. Rev. D 57 (1998) 567.
[22] V. Barger, C. Kao, Phys. Rev. D 57 (1998) 3131.
[23] J. Ellis, T. Falk, K. Olive, M. Srednicki, hep-ph/9905481.
[24] V. Berezinsky, A. Bottino, J. Ellis, N. Fornengo, G. Mignola, S. Scopel, Astropart. Phys. 5
(1996) 1.
[25] V. Berezinsky, A. Bottino, J. Ellis, N. Fornengo, G. Mignola, S. Scopel, Astropart. Phys. 6
(1996) 333.
[26] P. Nath, R. Arnowitt, Phys. Rev. D 56 (1997) 2820.
[27] R. Arnowitt, P. Nath, Phys. Lett. B 437 (1998) 344.
[28] A. Bottino, F. Donato, N. Fornengo, S. Scopel, Phys. Rev. D 59 (1999) 095004.

142

E. Accomando et al. / Nuclear Physics B 585 (2000) 124–142

[29]
[30]
[31]
[32]
[33]
[34]
[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]

R. Arnowitt, P. Nath, Phys. Rev. D 60 (1999) 044002.
R. Bernabei et al., Phys. Lett. B 424 (1998) 195; Phys. Lett. B 450 (1999) 448.
V. Barger, M.S. Berger, P. Ohmann, Phys. Rev. D 49 (1994) 4908.
M. Brhlik, L. Everett, G. Kane, J. Lykken, hep-ph/9908326.
E. Accomando, R. Arnowitt, B. Dutta, hep-ph/9909333.
H. Anlauf, Nucl. Phys. B 430 (1994) 245.
Particle Data Group, Eur. Phys. J. C 3 (1998) 1.
Super Kamiokande Collaboration, Phys. Rev. Lett. 83 (1999) 1529.
W. Freedman, astro-ph/9909076.
S. Dodelson, L. Knox, astro-ph/9909454.
J. Mohr, B. Mathiesen, A.E. Evrad, Astrophys. J. 517 (1999) 627.
S. Perlmutter et al., astro-ph/9812133; Astrophys. J. (in press).
A.G. Riess et al., Astron. J. 116 (1998) 1009.
C. Lineweaver, astro-ph/9909301.
H. Leutwyler, Phys. Lett. B 374 (1996) 163.
J. Ellis, R. Flores, Phys. Lett. B 263 (1991) 259; Phys. Lett. B 300 (1993) 175.
G. Jungman, M. Kamionkowski, K. Greist, Phys. Rep. 267 (1995) 195.
E.W. Kolb, M.S. Turner, The Early Universe, Addison-Wesley, Reading, 1990.
K. Greist, D. Seckel, Phys. Rev. D 43 (1991) 3191.
P. Gondolo, G. Gelmini, Nucl. Phys. B 360 (1991) 145.
R. Arnowitt, P. Nath, Phys. Lett. B 299 (1993) 58; Phys. Lett. B 303 (1993) 403 (Erratum).
P. Nath, R. Arnowitt, Phys. Rev. Lett. 70 (1993) 3696.
L. Roszkowski, Phys. Lett. B 262 (1991) 59.
R. Arnowitt, P. Nath, Phys. Rev. Lett. 69 (1992) 725.
T. Goto, T. Nihei, Phys. Rev. D 59 (1999) 115009.
R. Rattazi, U. Sarid, Phys. Rev. D 53 (1996) 1553.
M. Carena, M. Olechowski, S. Pokorski, C. Wagner, Nucl. Phys. B 426 (1994) 269.

Nuclear Physics B 585 (2000) 143–170
www.elsevier.nl/locate/npe

General matter coupled N = 2, D = 5 gauged
supergravity
Anna Ceresole a , Gianguido Dall’Agata b
a Dipartimento di Fisica, Politecnico di Torino and Istituto Nazionale di Fisica Nucleare, Sezione di Torino,

C.so Duca degli Abruzzi, 24, I-10129 Torino, Italy
b Dipartimento di Fisica Teorica, Università di Torino and Istituto Nazionale di Fisica Nucleare,

Sezione di Torino, via P. Giuria 1, I-10125 Torino, Italy
Received 20 April 2000; accepted 24 May 2000

Abstract
We give the full Lagrangean and supersymmetry transformation rules for D = 5, N = 2
supergravity interacting with an arbitrary number of vector, tensor and hyper-multiplets, with gauging
of the R-symmetry group SU(2)R as well as a subgroup K of the isometries of the scalar manifold.
Among the many possible applications, this theory provides the setting where a supersymmetric
brane-world scenario could occur. We comment on the presence of AdS vacua and BPS solutions that
would be relevant towards a supersymmetric smooth realization of the Randall–Sundrum “alternative
to compactification”. We also add some remarks on the connection between this most general 5D
fully coupled supergravity model and type IIB theory on the T 11 manifold.  2000 Elsevier Science
B.V. All rights reserved.

1. Introduction
In the quest for a unified description of gravity and matter interactions, several higherdimensional theories have been proposed in the past. In this respect, gauged supergravities,
where the global isometries of the matter Lagrangean are promoted to local symmetries,
have been widely explored and by now almost all allowed models, for diverse spacetime
dimensions and number N of supersymmetries have been analysed (see for instance [1,2]).
The five-dimensional N = 2 supergravity theory, in particular, has been considered at
various stages, but despite the many papers on the subject [3–12], it still lacks a complete
description where all possible matter couplings are included and the most general gauging
is performed.
The present renewed interest in gauged supergravity theories is mainly due to their
prominent role within the AdS/CFT correspondence [13–15]. It is believed that the
E-mail addresses: ceresole@athena.polito.it (A. Ceresole), dallagat@to.infn.it (G. Dall’Agata).
0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 3 3 9 - 4

144

A. Ceresole, G. Dall’Agata / Nuclear Physics B 585 (2000) 143–170

5D gauged supergravities provide a consistent non-linear truncation of the lowest lying
Kaluza–Klein modes of type IIB supergravity on an AdS5 × X5 space, that should be
dual to some four-dimensional supersymmetric conformal field theory. In detail, the
truncation of the AdS5 × S 5 compactification [16–19] should be described by N = 8
gauged supergravity [20,21], while the remaining AdS5 × X5 models should correspond
to N < 8 gauged theories. A special attention is devoted to the X5 = T 11 case [22], maybe
the most tested non-trivial instance of the AdS/CFT correspondence [23–28], whose
low-energy action should be expressed in terms of an N = 2UR (1) gauged supergravity
theory [28].
Still in the AdS/CFT framework, gauged supergravities are very interesting because
they open the possibility of studying the renormalisation group (RG) flows of deformations
of Yang–Mills theories by looking only at their supergravity formulation [29–35].
Supersymmetric [29–31] and non supersymmetric flows [32–34] to other conformal or
non-conformal theories have been studied, mostly using N = 8 gauged supergravity.
In connection with the N = 1 CFT dual to the AdS5 × T 11 compactification, the N = 2
gauged theory could be useful to follow the RG flow associated with fractional branes [35],
as well as a supersymmetric deformation which breaks the SU(2) × SU(2) flavour group
to the diagonal SU(2), as we will describe later.
Another line of development providing a strong motivation to find the most general
N = 2 gauged supergravity in D = 5 is the increasing phenomenological interest in
brane-world scenarios 1 based on both heterotic M-theory compactifications [10,37] and
Randall–Sundrum type models [38,39]. More precisely, one can distinguish between two
setups. In the first (RS1) [38], meant to provide a solution to the hierarchy problem, there
are two membranes located at the orbifold fixed points of a fifth compact dimension. The
complete action includes 5D gravity and sources for the two membranes:
S = Sbulk + Sbranes .
In the second (RS2) [39], there is a single membrane source where gravity is confined
by a volcano potential given by two surrounding AdS spaces. Since in this case the
fifth dimension is really uncompactified, RS2 suggests an alternative to Kaluza–Klein
reduction, and one can aim at obtaining this model within a gravity theory, without singular
sources. In order to realize RS2, one must find a model yielding two different stable critical
points with equal values of the vacuum energy and a domain-wall solution interpolating
between them.
Although it is perhaps desirable to embed any of the above scenarios into a supersymmetric string or gravity theory, the supersymmetrisation of the RS2 “alternative to compactification” is obviously much more appealing in view of its theoretical implications.
Regrettably, Kallosh and Linde have shown [41] that none of the available 5D supergravity
theories allow for such RS2 construction, and a definite answer for the minimal supersymmetric extension can only derive from the study of the fully coupled 5D N = 2 theory.
1 The idea of embedding our universe into an uncompactified higher-dimensional spacetime can be traced back
to [36] and was further pursued in [70,71].

after a description of the generic matter couplings and SU R (2) and Yang–Mills gaugings. the addition of tensor multiplets obtained by dualising some of the vectors has been explored within the Einstein–Maxwell theory. hypermatter was later coupled to the Einstein–Maxwell theory [10]. The interaction with vector multiplets. where the RS1 scenario is considered within the gauged N = 2 pure supergravity [5] modified by the presence of singular sources. G. yielding the general Einstein–Maxwell ungauged theory.4].A. we write the general Lagrangean and complete supersymmetry transformation rules. was proposed in [6]. also interaction with an arbitrary number of hyper-multiplets and generic gauging of K ⊂ G and SU(2)R .12]. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 145 Notice that this is not in contrast with the result in [42]. The history of the couplings and gaugings of D = 5. The pure theory was developed long ago [3. the scalar potential of the theory can be completely expressed in terms of the shifts that appear in the supersymmetry rules of fermionic fields due to the gauging [59. the gauged and fully coupled D = 5. We find that the scalar field potential is modified by the inclusion of hypermultiplets in that a new term appears and the contribution due to the R-symmetry gauging is now given in terms of SU R(2) rather than UR(1) invariant objects. N = 2 supergravity is quite long. In particular. Aside from the supersymmetric brane worlds. As a byproduct of the above-mentioned heterotic M-theory compactifications. the study of possible non-BPS solutions surely deserves a deep investigation that we postpone to future work. Ceresole. The scalar fields belonging to the vector and hypermultiplets parametrise a manifold M that is the product of a very special [54] by a quaternionic manifold.60]. 2 The possibility of a coupling to hypermatter was foreseen in [53] in the context of M-theory compactifications over Calabi–Yau manifolds and a first description of such ungauged coupling can be found in [9]. Rather than the Noether method. We then turn in Section 5 to the investigation of the possible realization of the RS2 scenario and perform a very simple preliminary analysis of the scalar potential along the lines of [41]. Attempts to obtain any of these scenarios from a pure stringy perspective can be found in [44–46]. . 2 Very recently. with the gauging of a subgroup K ⊂ G of the isometries of the scalar manifold and of a U (1) subgroup of the R-symmetry group [11. that were often neglected in the past from both Lagrangeans and supersymmetry transformation rules. as its UR (1)-gauged version [5]. This paper completes the above work by adding to the coupling of vector and tensor multiplets. N = 2 theory is the starting point for the study of five-dimensional black holes along the lines of [47–52]. We seem to find that no new supersymmetric AdS vacua arise and no new BPS solutions are generated. In Sections 2–4. we use our past experience [55– 58] and construct the Lagrangean by a geometrical technique that yields quite naturally also all the higher order terms in the fermion fields. However. while some of its possible gaugings appeared in [7].

and their consistent solution determines the couplings and the dynamics of the fields. Once the constraints are imposed. . the Lorentz and graviphoton curvatures Ra b and F are defined by T a = Dea = dea + eb ωb a . In order to exemplify our technique and show how to analyze the results. DRa b = 0. the second is the use of superfields [61]. G. The reduction to the ordinary component formalism is trivial and it shows that one obtains naturally all the supercovariantized quantities. 2. 3 Our conventions are collected in the appendix. Aµ (2. The unphysical fields are eliminated by imposing constraints on the supercurvatures. the BI of the various superfields are no longer automatically satisfied. Thus superspace formalism is also a good tool to simplify computations involving higher order fermion terms.1) contains the graviton eµa .2) F = dA. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 As a further application of our results. Preliminaries: pure supergravity Rather than component formalism and Noether method. which are often disregarded in other treatments. two gravitini ψµαi and a vector field Aµ (the graviphoton). we chose to work with superspace language for two main reasons: the first is the use of differential forms. (2. The pure supergravity multiplet 3  a αi eµ . containing a large number of component fields. The torsion T a . dF = 0. that guarantee a natural way to achieve supersymmetry. (2. The promotion of differential forms to superforms yields the supersymmetry transformations and equations of motion. most of which are superfluous. we briefly revisit the pure N = 2 five-dimensional supergravity [3–5] in the superspace formalism [61]. ψ αi ). Ceresole.146 A. the Lorentz connection ωa b and the one-form A. that often simplify computations and make more transparent the geometric meaning of the various structures in the theory [2]. without the need of an action to start from. that are described in superspace by the supervielbeins ea = (ea . through the derivation of the equations of motion.3) It is well known that each tensor component represents a full superfield multiplet. Ra b = dωa b + ωa c ωc b . These fields satisfy the Bianchi identities (BI) DT a = eb Rb a . ψµ . we finish in Section 6 with some remarks concerning the explicit realization of the theory corresponding to the T 11 compactification of type IIB supergravity.

5) i√ 6 ij Cαβ . Only the definition of the gravitational covariant derivative differs by a sign.4) and (2. but the Taαi σ k component of the torsion is determined by the F -BI. and those imposing the dynamics: 4 Tαiβj σ k = 0. (2. the solution given in [5] would read Tab c ∼ ab cde Fde and Taαi βj ∼ Γ b  α β j δi Fab . Tαiβj 2 (2. one finds that the closure of the supersymmetry algebra still requires the fulfillment of the fundamental constraint (2.. that reflects in the opposite sign in the definitions of ω and R with respect to [6].63] based on group theoretical arguments leads to the fundamental constraint 1 a a = ij Γαβ . it turns out that also the (2.7c) 8 6 √ i 6 i 1 a b ψ¯ ψi . they cannot be solved separately.7). we mention that the general strategy [62.6) and their solution (2. translating all its non-zero components in the definition of Taαi βj . 4 All the conventions have been chosen such that the supersymmetry laws and the structure of the Lagrangean match the formulae in [6].A.4). Dall’Agata / Nuclear Physics B 585 (2000) 143–170 147 Remarkably. 4  1 i (2. One can now solve the BI and find the following parametrizations: 1 (2.g. (2. G.5) constraint can be preserved. in our formulation. the T a bc component of the torsion can always be annihilated by a shift of the Lorentz connection ωa b c . that highly reduce the number of effective degrees of freedom and simplify the solution of the BI. (2. . In particular. (2. Ceresole. Without entering the technical details.7b) T i = ea eb Tba i + √ ea γabc ψ i − 4ηab γc ψ i F bc .7a) T a = − ψ¯ i γ a ψi .and F -BI are coupled.6) 4 From the rheonomic approach point of view [2.7a) solution and the (2. but some of them can be absorbed in superfield redefinitions [62. First of all. E. Fαiβj = − (2. Moreover. 2|1) superalgebra.4)–(2.63]. since the T .6) are a result of the Maurer– Cartan’s equations dual to the SU(2. 2 4 6 1 1 1 Rab = ed ec Rcd ab + ec ψ¯ i γc Tab i − ec ψ¯ i γ[a Tb]c i 2 4 2  cd i i i cd cd + √ ψ¯ i γab ψ + 4ψ¯ i ψ ηab F .5].4) which is needed to preserve rigid supersymmetry. Although the coupling with matter multiplets could in principle change the constraints (2. not all the constraints are dynamical.7d) F = e e Fba + 2 8 This solution deserves some comments.

Dall’Agata / Nuclear Physics B 585 (2000) 143–170 which differs from (2.148 A. since by Dragon’s theorem the R-BI follow once solved the T ones [62. For the (2. G. Ceresole.63]. Finally. which does not correspond to any structure in our supergravity model and indeed it is set to zero by the same T -BI. the supersymmetry transformations of the component fields are given by covariantized superspace Lie derivatives of the corresponding superfields. evaluated at ϑ = 0 = dϑ: . εαi ) being the translation parameter. the (2.7). one sees that only its 4 irreps. 5 The same field redefinitions also tell that the only physical component in Taαi b is the 40.63]. using the vielbein and connection redefinitions found in [62.7) equations can be determined by solving only the F .5) constraint. The ordinary supersymmetry transformations can be easily read off from the superspace results (2.and T -BI. In fact. of SO(5) are physical and they are fixed by the solution of the lowest-dimensional T -BI which does not change in presence of matter.7) by a shift of abcde F de in the ωa b c definition. ε A = (0.

8) δε φ = (iε D + Diε )φ . (2.

λi . . . while the tensor multiplet  Bµν . and again an SU R (2) doublet of spin-1/2 fermions and one real scalar field. ω )εi + √ ea (γabc − 4ηab γc )εi F δε ψ i = D(b 4 6  i cd b cd δε ωa b = √ ψ¯ i γab cd εi + 4 ψ¯ i εi ηab F . Gauged supergravity with generic matter coupling Five-dimensional N = 2 supergravity allows. φ contains a tensor field of rank two.q there is a doublet of spin-1/2 fermions A = 1. 4.. λi . They are explicitly 1 δε ea = ε¯ i γ a ψi . δε A = 4 (2. . i. In the hypermultiplet  A X ζ . an SU R (2) doublet of spin-1/2 fermions and one real scalar field. The vector multiplet  Aµ . . where the hatted quantities refer to supercovariantized terms. φ contains a vector field.1). 4√ 6 i 6 ¯i ψ εi . . 2 i bbc . F [a 4 3. three kinds of matter multiplets: the vector. 5 We will see in the next section that this freedom can be used to couple the hypermatter.e. beside the supergravity multiplet (2. 2 and four real scalars X = 1.ϑ=0=dϑ .9) √ bab = Fab + i 6 ψ¯ i ψb]i . tensor and hypermultiplet.

. we have two types of flat indices A = 1.3) ˜ ˜ of an ambient space parametrized by nV + nT + 1 coordinates hI = hI (φ x˜ ). . . . The Q manifold is the target space of the q X scalars and X = 1. which are directly related to the computations we present in this and the next sections. to which we surely refer the reader for all the details. A complete classification of the allowed homogeneous manifolds has been given in [54] and a lot of their interesting properties. the theory we are going to describe has the following field content  a i I M i a˜ A x˜ X eµ . Bµν . The λai SO(nV + nT ) and a˜ = 1. Ceresole. φ . which we denote by I˜ = (I. nV is an index labeling the vector fields of the nV vector multiplets and the graviphoton. λ . The S target space manifold The scalar field target manifold of the vector and tensor multiplets S is a very special manifold [6] which can be described by an (nV + nT )-dimensional cubic hypersurface ˜ ˜ ˜ CI˜J˜K˜ hI hJ hK = 1 (3. . since they will mix in the interactions. nV + nT . . . . . . the scalars of the nV vector. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 149 To allow for self-interactions. As expected for a quaternionic manifold. . M = 1. 2.A. (3. . nT labels the tensor multiplets. where CI˜J˜K˜ is a completely symmetric constant tensor that determines also the Chern– Simons couplings of the vector fields. (3. especially when they are restricted to be a coset of the Jordan family. . We will shortly see that it is also useful to introduce a collective index for the vector and tensor fields. nV + nT is the corresponding flat index. as stated in [11]. ψµ . . Aµ . .1. x˜ = 1. 3. . . 4nH are the curved indices labeling the coordinates. corresponding to the fundamental representations of USp(2nH ) and USp(2) ' SU(2). . ζ . but.2) Here I = 0. Here we only collect some notions and results for their geometrical structures. . the only components that survive the gauging of a Yang–Mills group are the CI J K and CI MN . M). . . we note that when nT 6= 0 not all the CI˜J˜K˜ coefficients differ from zero. have been given in [6]. . . . . First. G. nT tensors and nH hypermultiplets parametrize a manifold Mscalar which is the direct product of a very special [54] and a quaternionic manifold M = S(nV + nT ) ⊗ Q(nH ). The scalars φ x˜ .1) with dimR S = nV + nT and dimQ Q = nH . parametrize the target space S and thus ˜ instead transform as vectors under the tangent space group x˜ is a curved index. It is known that the kinetic term for such scalars can be written in terms of the coordinates as ˜ ˜ ˜ CI˜J˜K˜ hI ∂µ hJ ∂ µ hK . q . In detail. 2nH and i = 1. 1.

hI˜x. fiC (3. ˜ ˜ hIx˜ hJy˜ aI˜J˜ = gx˜ y˜ . which can be given implicitly in terms of fx˜a˜ through the formula ˜ ˜ ˜ a˜ b b f[ax. as explained in the analogous sixdimensional case [55]. The quaternionic metric tensor will be denoted by gXY (q). while ωX i j (q) and ωX A B (q) will be the USp(2) and USp(2nH ) connections.64]. fx˜a˜ . that are essential to close the supersymmetry algebra [6]: 5 3 x˜ y˜ z˜ CI˜J˜K˜ = hI˜ hJ˜ hK˜ − a(I˜J˜ hK) ˜ + Tx˜ y˜ z˜ hI˜ hJ˜ hK˜ . ˜ y˜ = − ˜ y˜ = I 3 3 (3.5) 3. ˜ = 0. To obtain the coupling of these fields to supergravity when they are chargeless with respect to the Yang–Mills (Maxwell) fields. As a consequence of its quaternionic structure. X to pass to the flat indices iA ∈ This means that one can introduce the vielbeins fiA USp(2) ⊗ USp(2nH ). h . The coordinates of the ambient space ˜ hI have an index which is raised and lowered through the φ x˜ -dependent metric aI˜J˜ . Dall’Agata / Nuclear Physics B 585 (2000) 143–170 ˜ For what concerns the S manifold. ˜ y] ˜ + Ω[y˜ fx] The Riemann tensor is given by Kx˜ y˜ z˜ w˜ =  4 t˜ gx[ ˜ w˜ gz˜ ]y˜ + Tx[ ˜ w˜ Tz˜ ]y˜ t˜ .6) 1 XY g CAB . X Yi Y Xi fB + fiA fB = fiA .10. h x. These vielbeins obey the following relations: X Y fj B = ij CAB .150 A. the holonomy group of the manifold Q is a direct product of SU(2) and some subgroup of the symplectic group in 2nH dimensions. gXY fiA X YC Y XC fj + fiC fj = g XY ij .x˜ = 3 I x˜ 3 x˜ r r  2 2 ˜ ˜ z˜ I˜ gx˜ y˜ hI˜ + Tx˜ y˜ z˜ h ˜ . gx˜ y˜ hI + Tx˜ y˜ z˜ hz˜ I . one uses the freedom left by the solution of the T -BI of introducing some fermions in Tαiβj γ k .4) where Tx˜ y˜ z˜ is a completely symmetric function of φ x˜ . gx˜ y˜ and Ωxa˜˜ b denote its (nV + nT )-bein. the metric and the spin connection. The Q target manifold Self-interacting hypermultiplets in an N = 2. G. Ceresole. All these functions are subject to the following algebraic and differential constraints.x˜ = − hI˜. 3 (3. y˜ aI˜J˜ = hI˜ hJ˜ + hxI˜˜ h ˜ gx˜ y˜ . h . 2 2 ˜ hI hxI˜˜ = 0. D = 5 theory are known to live on a quaternionic Kähler manifold [9. nH where ij and CAB are the SU(2) and USp(2nH ) invariant tensors respectively. J r r 2 2 I˜ I˜ h˜ .2.

10) where Rj i ≡ dωj i + ωj k ωk i = 12 dq X dq Y RY X j i is the USp(2) curvature. nH = 2 i l . G. In the first case.9) = 0. # = {A. 3.11) RXY ij = κ fXiC fYj with κ a constant fixed by supersymmetry requirements to κ = −1/2.A.3. but this of course breaks USp(2) covariance. RXY W Z fiA (3. It is more convenient to proceed in a slightly different but equivalent way which does not break covariance. (3. Redefining the covariant derivative D and introducing the USp(2) ⊗ USp(2nH ) connections D = d + ωLorentz + dq X ωX# # . supersymmetry requires K to be a subgroup of the full M.8) one gets the new torsion definitions T a ≡ dea + eb ωb a . which will be dualised to tensor fields. T αi ≡ dψ αi + ψ βi ωβ α + ψ αj ωj i . As explained in [56–58]. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 151 Since the structure of the possible spinor component Tαiβj γ k is of the same form as the pullback of the USp(2) connection on the cotangent bundle basis of the superspace. (3. charged under K. and imposes again the constraint Tαiβj γk (3.12) RXY AB = κ fXiA fYi B − fY iA fXB + fXiC fYDi ΩABCD . W Z fj B = ij RXY AB + CAB RXY ij . Ceresole.13) that can easily be pulled back on superspace. one is left with up to nT = dim G − nV other ones. The gauging The gauging of matter coupled N = 2 supergravity theories is achieved by identifying the gauge group K as a subgroup of the isometries G of the M product space. Other useful identities regarding the definition of the Q-Riemann tensor and the USp(2nH ) curvature are  i (3. and the hypermultiplet space will generically split into X 1X P P ni Ri + nl Rl . two main cases can occur: K non-abelian and K = U (1)nV +1 . It is known [64] that a quaternionic manifold is maximally symmetric.7) and solve the BI. Here ΩABCD denotes the totally symmetric tensor of USp(2nH ). i}. This modifies also the T -BI according to DT αi = ψ βi Rβ α + ψ αj Rj i . If one chooses to gauge nV + 1 vector fields. one can write the new constraint Tαiβj γ k = δαγ ωβj i k + (αi ↔ βj ). (3. This fixes the USp(2) curvature to  C C − fY iC fXj . (3.

Dq X = Dq X + gAI KIX (q).Y fiXA fBY i . (3. ˜ (3.22) .16) ˜ Dλai˜ = Dλai˜ + gAI LI a˜ b (φ)λbi . for fIKJ the gauge group structure constants.2] first introduced in [65] for the N = 8.19) D q = gF KIX . The gauging now proceeds by introducing nV + 1 Killing vectors acting generically on M: φ x˜ → φ x˜ +  I KIx˜ (φ). (3. the S-manifold is not required to have any isometry and if the hypermultiplets are charged with respect to the nV + 1 U (1)’s.17) At the same time.18) This reflects into suitable changes in the definition of the gauged curvatures and BI: D2 φ x˜ = gF I KIx˜ . the connection Ωa˜ b˜ is replaced by its gauged counterpart Dφ x˜ Ωx˜ a˜ b˜ + gAI KI a. ˜ ˜ ˜ ˜ ˜ ˜ = Rj i λaj + K a˜ b λbi + gF I LI a˜ b λbi . LaI˜ b the G-transformation ˜ ˜ matrices of the gluinos LaI˜ b ≡ ∂ b KIa˜ and ωI B A ≡ KI X. D2 λai D2 ζ A = RB A ζ B + gF I ωI B A ζ B . The quaternionic structure of Q implies that KIX can be determined in terms of the Killing prepotential PI i j (q) [56–58]. ˜ b˜ . (3.21) (3.152 A.14) and. for an infinitesimal parameter  I . q X → q X +  I KIX (q). (3. (3. 1 1 gRXY i j KIX KJY + gR P[I i k PJ ] k j + g fIKJ PK i j = 0. In the abelian case. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 where Ri and RlP are a set of irreps of K (P = pseudoreal). Following the well-established general procedure [1. AI the gauge field one-forms. ˜ where g is the coupling constant. SO(nV + nT ) and USp(2) ⊗ USp(2nH ) connections by K-covariant derivatives Dφ x˜ = Dφ x˜ + gAI KIx˜ (φ). which satisfies dq X KIY RXY i j = DPI i j (3. G. the Q manifold should at least have nV + 1 abelian isometries. D = 4 case.15) 2 2 Gauging the supergravity theory is now done by gauging the composite connections of the underlying σ -model.20) 2 X where I (3. we proceed by replacing for the YM couplings the covariant derivatives on the scalar and fermion fields containing the Lorentz. Dζ A = Dζ A + gAI ωI B A (q)ζ B . Ceresole.

4).4. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 153 1 Ka˜ b˜ = Dφ x˜ Dφ y˜ Ky˜ x˜ a˜ b˜ . The solution of the Bianchi identities In order to solve the superspace BI in presence of gauging and to obtain the new susy rules.28) and that the SU R (2) curvature definition is replaced by bi j = 1 Dq X Dq Y RY X i j + gR F I PI i j .32) cannot be derived by the standard procedure of solving the BI for the superfield two-form B M . M fields satisfy a first order equation of motion of the type It is known [11] that the Bµν DB M = MM N ? B N .31) and 3.14) and (3.30) bj i . R 2 (3. G. .26) For the SU R (2) connection. being a first order equation of motion.32) contains the Hodge star product. one must first face a technical problem. the existence of a Killing vector prepotential PI i j (q) allows the following definition: ωi j → ωi j + gR AI PI i j (q). (3. DT αi = ψ βi Rβ α + ψ αj R (3. (3.32) where M is a mass matrix. (3. that is how to implement the tensor multiplet BI in superspace. which is not well defined in superspace.A. Moreover. with components defined in (3.24) RAB = Dq X Dq Y RY X AB . This leads to a further redefinition of the gaugino BI (3. after imposing some constraint on its field strength H M . The problem is that (3.23) 2 1 (3.21) and of the torsion one. Ceresole. (3.12) and 1 F I = dAI − gfJIK AJ AK 2 is the gauge field strength satisfying the BI DF I = 0. 2 are the S Riemann and USp(2nH ) curvatures.25) (3. (3.27) which replaces the SU(2)R connection with its gauged counterpart.29) provided PI i j satisfies the (3. (3. This implies that the new covariant derivative acting on the gravitino is T αi ≡ Dψ αi = dψ αi + ψ βi ωβ α + ψ αj ωj i + gR ψ αj AI PI j i (q). which now read ˜ ˜ ˜ ˜ ˜ ˜ bj i λaj =R + K a˜ b λbi + gF I LI a˜ b λbi D2 λai (3. (3.15) relations.

37a) (3. This allows the vector fields to be completely gauged away by taking ΛM = AM .34) where the superfield connection B M appears explicitly. from now on we will use HI to denote collectively the F I and B M fields. Closure of the supersymmetry algebra imposes that the vector field trivially under the gauge transformations of the tensor fields (3. depending on the value of I˜.35) AM transforms non- δb M = dΛM ⇒ δAM = −ΛM . one must also introduce some bM tensor fields with the usual invariance δb M = dΛM . ΛM IN (3. To obtain the gauging of K. These must be introduced in the Lagrangean and transformation rules in such a way that everywhere F M gets replaced by the combination B M = bM + F M . =− Hαiβj 4 Tαiβj γ k = 0. These BI now will also provide the first-order B M equations of motion at the level of the abαiβj sector. Tαiβj 2√ 6 ˜ I˜ iij Cαβ hI . transforming nontrivially under K. The solution to this problem lies in the origin of the tensor fields. since the massive B M has lost the gauge invariance under δB M = dΛ) and then solve consistently the H -BI. where now bM has “eaten” the AM degrees of freedom.36) leaving the (3. In perfect analogy with the six-dimensional case [66]. leaving M B = bM . 6 is the where This implies that the H -BI become I N DH M = gΛM IN F B .37b) (3. In addition. this can be seen as the need of imposing directly on B M the same constraints and F -BI solutions imposed on F M (which is now allowed.33) representation matrix. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 It must also be noted that the B M transform under the gauge group K and the correct definition of H M is [11] I N H M ≡ dB M + gΛM IN A B . before the gauging the B M degrees of freedom are described by AM vectors.35) combination gauge invariant. (3. From the superspace point of view. This also implies that B M is now massive. breaking covariance.37c) . (3.6). The constraints to be imposed on the curvatures are the straightforward generalization of those proposed for pure supergravity (2. (3.4)–(2. G. and obtained its longitudinal modes by a Higgs-type mechanism. Ceresole. ˜ Following [11]. one must fix the normalization of the fermion fields: 6 It has been shown [11] that they must lie in a symplectic representation of the gauge group K. and read 1 a a = ij Γαβ .154 A. and its supersymmetry variation acquires an additional term of the form dδsusyAM .

the scalar parametrization directly follows from (3. we obtain the new parametrizations of the curvatures as well as some algebraic and differential constraints on the geometric structures (e. (3.42) Regarding the Yang–Mills and tensor multiplets. these can be collectively written as √ 1 a b I˜ 1 a i i 6 i ˜ I˜ a˜ I˜ ¯ ψ¯ ψi hI . Dαi q X ≡ ifAi 155 (3. The torsion parametrization remains the same as in the pure gravity case: 1 (3. 2 (3. apart from the component fixed in (3. Since the B M parametrization must have the same form. has a new term involving the gluino fields which is fixed by the F -BI.g.38): i Dφ x˜ = ea Da φ x˜ + ψ¯ i λai˜ fa˜x˜ .40) T a = − ψ¯ i γ a ψi . Ceresole. G. those presented in (3.5)) and the equations of motion.39) definitions in the BI. (3.37b). Dall’Agata / Nuclear Physics B 585 (2000) 143–170 i ˜ Dαi φ x˜ ≡ − fa˜x˜ λaαi .39) Introducing the (3.41) where Pij ≡ hI PI ij .38) (3.A.38). 4 However.44) H = e e Hba − e ψ γa λi ha˜ + 2 2 8 The gluinos field strength (and supersymmetry variation) are i 1 I˜ Dλai˜ = ea Da λai˜ − fx˜a˜ γ a ψi Da φ x˜ + haI˜˜ γ ab ψi Hab 2 4   1  i ˜  ˜ ˜ − √ T a˜ b˜ c˜ −3ψ j λ¯ bi λcj˜ + γa ψ j λ¯ bi γ a λcj˜ + γab ψ j λ¯ bi γ ab λcj˜ 2 4 6 (3. 2 X A ζα ..43) The F I field-strength. the super-field-strength of the gravitino now contains many new terms involving couplings with fermions and a new gauging term. where ˜ PI ij Pija˜ ≡ haI (3.46) .45) + gR ψ j Pija˜ + gW a˜ ψi . which is fixed by the solution of the T -BI: 1 i ˜ Ti = ea eb Tba i + √ hI˜ ea (γabc ψi − 4ηab γc ψi )Hbc I 2 4 6 1 a 1 1 − e γab ψ j λ¯ ci˜ γ b λcj˜ + ea γabc ψ j λ¯ ai˜ γ bc λaj˜ + ea ψ j λ¯ ci˜ γa λcj˜ 12 48 6 1 κ i − ea γ b ψ j λ¯ ci˜ γab λcj˜ − ea γ bc ψi ζ¯A γabc ζ A + √ gR ea γa ψ j Pij . 12 8 6 (3.37) constraints and the (3. (3.

Dζ A = ea Da ζ A − γa ψ i Da q X fiX 2 (3.50) where it must be noted that in Habc one should substitute the B equations of motion. the first bilinear in the gluini and the R-symmetry gauging terms are fixed by closure of the supersymmetry algebra. hI KIx˜ = 0 if we assume that under an infinitesimal K-transformation with parameter η.54) . differently from the four-dimensional case. In addition. the Killing vectors get a new contribution coming from the K-transformation properties of the tensors and therefore now hI KIx˜ 6= 0.39) is imposed. G. namely 2 MP CNP I . which reads √ 1 a b c i b a ¯i 6 a ¯i M a˜ MN e ψ γa ψi Ω MN hN . the Yang–Mills gauging in the absence of tensor multiplets does not lead to any extra term. and is given by X . √ M aI  ˜ = 6 ΛI N h ˜ hN + hI haN . The Yang– Mills gauging term is then fixed by the H -BI. Ω MN hN = Ω MN haN˜ (3. (3. +g H = e e e Hcba + ge e ψ γab λi hN Ω 3! 8 16 (3. (3. This is due to the fact that. We point out that.47) and it reads √ 6 I x˜ a˜ a˜ h KI fx˜ . Dq X = ea Da q X − i ψ¯ i ζ A fiA (3.53) Closure of supersymmetry on q X and ζ A imposes then i A + gψ i NiA .49) W = 4 The solution of the H -BI fixes the H M parametrization. ΛM IN = √ Ω 6 (3.52) Turning to the hypermultiplets.45) parametrization contains the obvious terms needed to close the F -BI and the supersymmetry algebra on φ x˜ . This gives us back a Yang–Mills gauging term.12) of [11].51) which can be shown to be equivalent to condition (5. Ceresole. the vectors transform as a˜ δη AI ∼ fJIK AJ ηK . determined by closure of the supersymmetry algebra on φ x˜ .48) When the tensor fields are involved. since it appears only when tensor multiplets are involved [11]. (3. according to [7]. The closure of the H -BI also imposes some constraints on the representation matrix Λ: √ M I N 6 ΛI N h h .47) Ω W ≡− 8 The (3. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 and √ 6 MN a˜ hM hN . which must be equivalent to (3. the scalar parametrization is fixed once (3.156 A.

Ceresole. (3. One sees that. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 157 with the g-order shift defined as √ 6 fXAi KIX hI . (3. 2 √ 1 i 6 i ˜ ˜ ˜ ψ¯ εi hI .8). (3. where ϑ I ≡ − ea ε¯ i γa λai˜ haI˜ + 2 4 i I˜ ba φ x˜ − δε φ x˜ Ωx˜ a˜ b˜ λb˜ − δε q X ωX i j λa˜ + 1 ha˜ γ ab εi H bab δε λai˜ = − fx˜a˜ γ a εi D i j 2 4 I˜    i ˜  ˜ ˜ − √ T a˜ b˜c˜ −3εj λ¯ bi λcj˜ + γa εj λ¯ bi γ a λcj˜ + 12 γab εj λ¯ bi γ ab λcj˜ 4 6 (3.63) δε q X = −i ε¯ i ζ A fiA i ba q X f A − δε q X ωX B A ζ B + gεi N A . G.64) δε ζ A = − γ a ε i D iX i 2 The hatted quantities b are the supercovariantization of the unhatted ones. (3. as expected. 2 (3.56) 4 2 As shown in the previous section. (3.A.61) + gR εj Pija˜ + gW a˜ εi . we give here also the parametrization of the Riemann tensor:  I˜ 1 i cd Hcd hI˜ Rab = ed ec Rcd. the ordinary supersymmetry transformations of the fields are recovered from the above superspace results by using (2.62) δε B M = dϑ M + geb ea ε¯ i γab λai˜ hN Ω MN + g 8 8 X . Finally. √ i 6 a i e ψ¯ γa εi Ω MN hN . . (3.60) δε AI = ϑ I .57) i b bc I˜ − δε q X ωX i j ψj ω)εi + √ hI˜ ea (γabc εi − 4ηab γc εi )H δε ψi = D(b 4 6   1  1 a 1 − e γab εj λ¯ ci˜ γ b λcj˜ + ea γabc εj λ¯ ai˜ γ bc λaj˜ + ea εj λ¯ ci˜ γa λcj˜ 12 48 6  1  i 1 − ea γ b εj λ¯ ci˜ γab λcj˜ + ea γ bc εi ζ¯A γabc ζ A + √ gR ea γa εj Pij . they complete those given in [11] with the higher order Fermi terms and the new couplings: 1 δε ea = ε¯ i γ a ψi .(3.59) δε φ x˜ = ε¯ i λai˜ fa˜x˜ .ab + √ ψ¯ i γab cd ψ i + 4ψ¯ i ψ i ηab 2 8 6 1 1 + ψ¯ i γabc ψ j λ¯ i γ c λj + ψ¯ i γc ψ j λ¯ i γ abc λj 24 48 1 i j κ i + ψ¯ ψ λ¯ i γab λj + ψ¯ i γ c ψj ζ¯A γabc ζ A − √ ψ¯ i γab ψ j Pij 24 4 2 6 1 1 + ec ψ¯ i γc Tab i − ec ψ¯ i γ [a Tb]c i .58) 12 8 6 i (3.55) NiA ≡ 4 This term is due to the Yang–Mills charge of the hypermultiplets. for completeness.

ap b1 .158 A.1). the superspace differential d now becomes the ordinary differential. Lmass gives the mass of the fermions. Each form can be decomposed along the vielbeins ea = dx µ eµ a and the gravitino.ap ≡ − (5 − p)! √ In particular eˆ = −g d 5 x. it is also convenient to define the (5 − p)forms 1  a1 . The kinetic terms are 1 i 1 LKIN = R ab eˆab + ea eb ψ¯ i γab Dψi − λ¯ i a˜ γa Dλai˜ eˆ a − ζ¯ A γ a DζA eˆa 2  4 2   1 i i a˜ y˜ x˜ a y˜ a x˜ y˜ ¯ + gx˜ y˜ Qa Q eˆ − eˆ gx˜ y˜ Qa Dφ − ψ λi fa˜ 2 2    1 X aY a X Y i A Y ¯ + gXY Qa Q eˆ − eˆ gXY Qa Dq + i ψ ζ fiA 2 . with ωˆ given in (4. tensor and hypermultiplets. and add all the modifications that are needed in presence of hypermultiplets and SU R(2) gauging. augmented by the standard gravitino bilinears. we now mean those objects evaluated at ϑ = 0 = dϑ..b5−p eb1 · · · eb5−p .11]. (2. The Lagrangean for the gauged theory can be split as L = LKIN + LPauli + LCS + Lmass + Lpot + L4Fermi .1) This determines ωˆ as the metric connection.8) gives Z Z Z δε L = iε DL + Diε L and Diε L is a total derivative that we can discard. Lpot contains the typical potential of the gauged theories and L4Fermi contains the four-Fermi terms. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 4. (4. where LKIN contains the kinetic terms. whenever we use the same symbols as in Section 3. via Eq. (3.2) eˆ a1 . which reduces to ψ i = dx µ ψµi ≡ ea ψai . as 1 dea + eb ωˆ b a = − ψ¯ i γ a ψi . 4 (4.. Invariance of the action under supersymmetry can be checked by using the standard trick of lifting this action to superspace.40). performing the superspace differential and taking the interior product with εA . D = 5 gauged supergravity in interaction with vector. Since we write the Lagrangean as a five-form. We’ll exhibit our result at the component level. now evaluated at ϑ = 0 = dϑ.. It is obvious that Eq. LCS contains the Yang–Mills Chern–Simons term.. The action We now turn to the Lagrangean of the N = 2. Ceresole. and for brevity. In particular.. LPauli describes the couplings between the bosonic field-strengths and the fermions. The supercovariant curvature two-form becomes Ra b = d ωˆ a b + ωˆ a c ωˆ c b . The supercovariant connection one-form ωˆ a b = dx µ ωˆ µ a b is naturally introduced.. We take as a start the results of [6. G.

such that  1 ˜ ˜ √ λ¯ i a˜ γab λbi λ¯ j c˜ γ ab λdj Ta˜ b˜ c. G.4) which. (4.5) 2 where r   2 1 c˜ δ ˜ h ˜ + Ta˜ b˜ c˜ hI˜ .3) + ΩMN B M H N . ea QX a √ 1 c i i 6 i 1 a b I˜ ˜ I˜ a˜ I˜ ¯ e e Qba = H + e ψ γc λi ha˜ − ψ¯ ψi hI . upon substituting this back in LKIN . The Pauli-like couplings are described by: √ i i a˜ i 6 i ˜ I˜ a˜ a b ¯ ψ¯ γa ψi HI hI˜ ea − λ¯ ai˜ γab ψi Dφ x˜ fx˜a˜ eˆ ab LPauli = λ γab ψi H hI˜ e e − 4 8 4 i i ˜ ˜ ˜ b I I i a ˜ ab A − ΦI˜a˜ b˜ λ¯ γab λi H eˆ + √ ζ¯A γab ζ H hI˜ eˆ ab 4 4 6 i ¯i X + ψ γab ζ A fAi Dq Y gXY eˆ ab .A. 2 2 8 (4. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 159 √    1 1 c ¯i i 6 ¯i I˜ J˜ ab J˜ J˜ ab a˜ J˜ + aI˜J˜ Qab Q eˆ + 2eˆ ψ ψi h H + e ψ γc λi ha˜ − 4 2 8 1 (4.7) . ΦI˜a˜ b˜ ≡ 3 4 a˜ b I whereas the Chern–Simons couplings are fixed as usual to  2 3 K LCS = √ CI J K F I F J AK − gF I AJ AL AF fLF 4 3 6  3 J K I G H L F + g 2 fGH fLF A A A A A . 2 X X = Dq + i ψ¯ i ζ A fiA . Their equations of motion give i ea Qxa˜ = Dφ x˜ − ψ¯ i λai˜ fa˜x˜ . following L4Fermi = L4λ + L3λψ + Lother . QA a and Qab are auxiliary fields which have been introduced in order to write the Lagrangean as a five-form. Ceresole.6) The four Fermi terms can further be split. ˜ d˜ 48 6 1 1 ˜ ˜ ˜ ˜ ˜ ˜ + Ka˜ b˜ c˜d˜ 2λ¯ i a˜ λbi λ¯ j c˜ λdj + λ¯ i a˜ γa λbi λ¯ j c˜ γ a λdj − λ¯ i a˜ λj a˜ λ¯ bj λbi 24 12  1 ¯ i a˜ 1 ¯ i a˜ j a˜ ¯ b˜ a b˜ j a˜ ¯ b˜ ab b˜ − λ γa λ λj γ λi + λ γab λ λj γ λi 24 64 L4λ = eˆ (4. 20 (4. yields the usual super-covariantized kinetic terms. g ˜ I where Qxa˜ .

46). (3. and of a corresponding new term in the scalar potential. Lother contains the new four-Fermi interactions with the hypermultiplet spinors and reads:   i a i 1 1 i j ab A ab i A ¯ ¯ ¯ ¯ ¯ ¯ ψ ψi ζA γ ζ + ζA γ ψi ψ ζ Lother = − e ψ γa ψi ψ ψj + eˆab 32 32 4   1 ˜ c 1 ˜ 1 ai ˜ ˜ k ab ¯ λ¯ ˜ γab ψi λ¯ ak ψ ψk + λ¯ ai γ ψi λ¯ aj γabc ψj − λ¯ ai γab λai ψ + k eˆ ˜ 16 16 64  κ κ − eˆ ζ¯A γ ab ζ A ζ¯B γab ζ B + ΩABCD 5ζ¯ A ζ B ζ¯ C ζ D − ζ¯ A γ a ζ B ζ¯ C γa ζ D 16 4 κ abc A ai ˜ a ˜ (4.55). a first difference is the existence of tensor multiplets satisfying a first order equation of motion. (3. The first reads √   i 6 i i i a˜ j b˜ a˜ b˜ i a˜ j a˜ a j ab ¯ ¯ ¯ ψ γab ψ Pij eˆ Lmass = gR √ λ λ Pij eˆ − λ γa ψ Pij eˆ − 8 2 6  ˜ + g λ¯ i a˜ γa ψi W a˜ eˆa + λ¯ i a˜ λbj Wa˜ b˜ eˆ + 2ψ¯ i γa ζ A NiA eˆ a  ˜ − 2i ζ¯ Aλai MAi a˜ eˆ + ζ¯ A ζ B MAB eˆ . (4.X] hI .10) where Pija˜ . √ i 6 I a. (3. one ˜ ˜ gluino and one gravitino due to the orthogonality properties of hI and hI a˜ .12) W a˜ b ≡ ihI [a˜ KIb] + 4 (4.14) MAB ≡ 2 AX This mass term is of first order in the gauge coupling constants and has coefficients fixed by variations of the kinetic terms. the potential is (LPot = −V e):   (4. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 and L3λψ   2i 1 ¯ i a˜ i a˜ j ¯ b˜ c˜ j ¯ b˜ c˜ ¯ = √ Ta˜ b˜c˜ λ ψ λi γa λj − λ γa ψ λi λj eˆ a 2 3 6 (4. (4. (4.9) − eˆ ζ¯A γ ζ λ¯ γabc λi . Pij .13) MAi a˜ ≡ fAiX KIX hI a˜ . W a˜ and NiA were defined in the previous section by Eqs. √ i 6 i f fBiY KI[Y .47) and (3.11) Pija˜ b ≡ δ a˜ b Pij + 4T a˜ bc˜ Pijc˜ .8) exactly reproduce the terms of the Einstein–Maxwell theory presented in [6]. and the other mass matrices are defined as ˜ ˜ ˜ (4. The second lies in the presence of the . ˜ ˜ ˜ h KI˜ b . ˆ Finally. G.160 A. 16 We point out that there are no terms with three gravitinos and one gluino or two ζ . However.42). We can finally describe the mass and potential terms. one can remark many similarities with the analogous four-dimensional matter coupled theory [56–58]. Ceresole. As an outcome of this complete analysis.15) V = 2g 2 W a˜ W a˜ − gR2 2Pij P ij − Pija˜ P a˜ ij + 2g 2 NiA N iA .

7) √ √ 6 MN a˜ 6 I x˜ a˜ Ω h KI fx˜ .2) have different origins.4) (5. In detail. (5.9) N iA ≡ 4 As expected.3) (5. whereas those of order g 2 come from the gauging of the Yang– Mills group K. the Yang–Mills gauging in D = 5 does not give rise to any contribution to the scalar potential unless some of the ungauged vectors are dualised into tensor multiplets.6) ≡ h PI ij . 4g (5. Finally.5) (5. The four terms in the potential (5. G. 5. hM hN = (5. 2 4 i A µ ib X γ ε Dµ q + gεi NiA . Some comments on the scalar potential We have found that the bosonic sector of 5D. N = 2 supergravity is described by the Lagrangean: 1 1 1 ˜ I˜ N =2 = − R − aI˜J˜ Hµν HJ µν − gXY Dµ q X Dµ q Y eˆ−1 Lbosonic 2 4 2 −1 1 e ˆ I J − gx˜ y˜ Dµ φ x˜ Dµ φ y˜ + √ CI J K  µνρσ τ Fµν Fρσ AK τ 2 6 6 + eˆ −1 µνρσ τ M  ΩMN Bµν Dρ BσNτ − V(φ. (5.2) The bosonic part of the supersymmetry transformation rules is given by  νρ I˜ i i b ω)εi + √ hI˜ γµνρ εi − 4gµν γρ εi H + √ gR γµ εj Pij . and thus the U (1) Kähler connection does not exist and all the 4D structures deriving from its gauging are missing. δε ψµi = Dµ (b 4 6 6 i a˜ µ b x˜ 1 a˜ µν bI˜ a˜ j a˜ δε λi = − fx˜ γ εi Dµ φ + hI˜ γ εi Hµν + gR ε Pij + gW a˜ εi . one sees that the scalar potential of the gauged supergravity theory is constructed out of the squares of the fermion shifts that arise in the supersymmetry transformations due to the gauging. Pija˜ aI ˜ (5. Moreover. δε ζ A = − fiX 2 where Pij ≡ hI PI ij . as already remarked. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 161 Chern–Simons term. (5.8) W a˜ ≡ − 8 4 √ 6 I X Ai h KI fX . Ceresole. the geometry described by the scalars of vector multiplets is now “very special” rather than special Kähler. that as well known is a peculiar feature of odd-dimensional spacetimes. Those of order gR2 come from the R-symmetry gauging.1) where   V = 2g 2 W a˜ W a˜ − gR2 2Pij P ij − Pija˜ P a˜ ij + 2g 2 NiA N iA . q).A. the YM ones are given by the squares of the g order shifts in the .

12]. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 supersymmetry transformation laws of the λai˜ and ζ A fields. As in the four-dimensional case [56–58]. does not allow anti-de Sitter solutions. Their most general form is now given by PI i j = iξIr (σr )i j .15) condition on the prepotential to ensure it.13) which is the supersymmetry requirement of [11].15) condition becomes gfIKJ ξKr = gR  rst ξIs ξJt . Moreover. (5. as it simply removes the g 2 W 2 term. The nT = 0 case is trivial. G. φ ∗ ) = 0. . The above mechanism is the local analogue of the Fayet–Iliopoulos phenomenon occurring also in four dimensions [56–58]. this potential allows for the existence of anti-de Sitter vacua if V(q ∗ . all the previously studied examples must be found as peculiar subcases. due to the reality properties of the very special manifold S parametrised by the hI coordinates. one can still preserve the full SU r r R(2) gauging by the choice ξI = δI and thus identifying by the (5.11) where ξIr are three real constants generically 7 breaking SU(2) → U (1). the condition (5. q) of (5. q) and Pija˜ (φ. Thus it is straightforward to see that the only contribution which can allow for such solutions is the 2Pij P ij term. the absence of the quaternionic fields q X implies that the prepotentials are set to zero. all the results therein can be recovered by substituting PI ij = VI δij (5. VI . For our metric signature.12) If one makes the choice ξI = (0. or at most are SU(2)-valued constants.6) and (5.12) reduces to fIKJ VK = 0. 0).162 A. we are forced to put KIX = 0.10) where (σr )i j are the usual Pauli matrices. coming from the R-symmetry gauging of the gravitinos. where the gauge group was chosen as K = U (1)nV +1 and gR = g. and the Yang–Mills sector is reduced to the W 2 part due to the tensor multiplets. Since (5. In particular. but here we don’t need to use the (3.14) in the action and supersymmetry laws. This implies that a simple Yang–Mills gauging. N = 2 gauged supergravity coupled to all matter. 2.7) containing the prepotential PI ij ≡ iPIr (q)(σr )ij . while the R-symmetry ones are given by the square of the order gR shifts in the ψ i and λai˜ supersymmetry variations. φ ∗ ) < 0 for V 0 (q ∗ . We first analyze the choice nH = 0. Since in this case the Q-manifold disappears. 3. in place of the P0 δij and P a˜ δij of [7. (5. (5. this can be related to the existence of a “Ward identity” for the scalar potential [59. In presence of hypermultiplets. r = 1.2) is the most general potential of 5D.60]. This leaves us with the potential presented in [10]. The (3.11. 7 In some special cases [69]. one generically has in the fermionic shifts the SU R(2) valued quantities Pij (φ. (5. even in presence of both tensor and hypermultiplets. Ceresole.12) the SU R(2) structure constants with those of a SU(2) ⊂ K: gf r st = gR  r st .

Thus. However. the only non-trivial effect of the W a˜ and NiA terms can be a change in the shape of the critical point. the anti-de Sitter fivedimensional supergravity.18) W = Pij = h NiA i = 0.16) for the choice Pij = V δij (ξ = (0. This implies that there is only one vector: the graviphoton. they can never cancel unless they vanish.41]. q ∗ = gR2 Pij P ij φ ∗ . (5. .15) V = g 2 K X KX − 2gR2 Pij P ij . while if tensor multiplets are added one still excludes the presence of supersymmetric solutions but leaves open the possibility of having non-BPS ones.41]. As for the theory without hypermultiplets [12. Although we have not yet performed a complete analysis of (5. This means that the N = 2 supersymmetric critical points of the full N = 2. The above result can also be obtained by a different argument. the cosmological constant of an N = 2 supersymmetric vacuum is only given by   (5. An easy way to do it is to determine the vacua obtained from the full potential (5.17). The nV = nT = nH = 0 case gives back the pure gauged supergravity of [5]. 4 which in principle could admit anti-de Sitter vacua.17) V φ ∗ . 8 We are indebted to R. Therefore. being all positive squares. Kallosh for explaining this argument and for important discussions on the results below. h Pija˜ i and h NiA i they must compensate each other in the potential. in order to have non vanishing h W a˜ i.2).12. and becomes 3 (5. one can easily show that the requirement h δε λ i = h δε ζ i = 0 implies that an N = 2 supersymmetric ground state must satisfy a˜ a˜ (5. 0)). Using the orthogonality property W a˜ Pija˜ = 0. with the potential V = −4V 2 . of course. This is. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 163 More interesting is to take nV = 0.A. G. we can already make some comments on the supersymmetric vacua and BPS solutions of the full theory. D = 5 gauged supergravity theory have the same nature as those of the reduced theory analyzed in [41]. q ∗ . Ceresole. The potential is still non-vanishing. V .3) imposes that the vacuum expectation value is given by (5. 8 The integrability condition on the gravitino supersymmetry rule (5. It has been shown [41] that in absence of tensor multiplets there are no RS2 solutions at all.2) and check whether they are of the same nature as those already studied in [7. It is now relevant to consider the possible existence of smooth Randall–Sundrum domain-wall solutions of type RS2 providing an “alternative to compactification”.

164 A.21) where Qr a˜ indicates some real combination of the various y-dependent quantities Qr a˜ (y) ∼ ∂y φ a˜ P r + P r a˜ .20) where the operator matrix Aa˜ can be read off from (5. (5. this condition still has no solutions. with a metric of the form ds 2 = a(y) dx 2 + dy 2 . When tensor multiplets are added. imposes for each a˜ 2 2 (5. one tries to relax the conditions (5.4). the Aa˜ operator reduces to the form j Aa˜ i ∼ iQr a˜ (σr )i j + W a˜ δi . where scalars are not fixed anymore. that also the ˜ examples considered up to now with HI 6= 0 do not seem to admit solutions of the RS2 type [67].3) iγ y εi ∼ gR Pij εj . We now show along the lines of [41] that the same phenomenon occurs also in the most general case where hyper-multiplets are added and the full SU R(2) group is gauged. one has to find some Killing spinors which. i.4). it turns out [41] that relaxing the condition W a˜ = 0. It can be seen that even if with the full R-symmetry gauging the Pij and Pija˜ are SU R (2)valued matrices. and requiring that all bosonic functions of the scalars have the same y behaviour. To do this. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 There remains to examine the possibility of having lower supersymmetric BPS solutions.e.18) near the critical point. We leave the investigation of all these more general cases to future work. Ceresole. ˜ ˜ This seems to rule out the presence of BPS solutions. We choose the y dependence of all fields appearing in the gluino susy rule (5.19) also satisfy j δε λai˜ = Aa˜ i εj = 0 (5.19) into the gluino transformation rule (5. j (5. This amounts to finding a Killing spinor eigenvector of Aa˜ with zero eigenvalue. approaching asymptotically AdS5 . It must be noted. i. upon substituting (5. forbids any supersymmetric solution.4) as ∂y φ a˜ (y) ∼ W a˜ (φ) ∼ Pija˜ (φ.22) and the obvious notation Pij ≡ iP r (σr )ij . Pija˜ ≡ iP r a˜ (σr )ij . Requiring det Aa˜ = 0 (or equivalently for a projector A2 = A). . G.. at least when hI˜ HI = ha˜˜ HI = 0. Indeed. however. or non-BPS solutions. leading to solutions that preserve half supersymmetry. q).. (5. It has been verified that in absence of tensor and hyper-multiplets the derivative of scalars in the radial direction y can be chosen proportional to the derivative of the superpotential.e. which has no solutions except for W a˜ = 0 = Qa˜ and this takes us back to the cases of [41]. in addition to the usual constraint coming from the gravitino susy rule (5. the matrix Aa˜ must be degenerate. I The only open possibilities appear to be either BPS solutions with non-trivial electric ˜ or magnetic fields HI . For BPS solutions to exist.23) W a˜ + Qa˜ = 0.

a very interesting deformation of the four-dimensional conformal field theory [23–27] is given by the operator ik s j l δrs . 2) + (1. this implies that there should exist another N = 2 vacuum of five-dimensional supergravity with this symmetry group. 1) + (1. Once a specific scalar manifold M is chosen. There are two possible embeddings of G in the N = 8 gauge group SO(6) ' SU(4). (6. some comments about the structure of the theory corresponding to the type IIB compactification on AdS5 × T 11 [28]. This should correspond to a flux in the CFT side. The first family can be obtained by turning on simultaneously the two scalars λ and µ. where λ and µ are two coupling constants and the isometry group has now been broken from SU(2) × SU(2) to the diagonal SU(2). the operator Wab must have. 3). which cannot preserve N = 2. This study could reveal very useful for further analyzing the AdS/CFT correspondence in such non-trivial case. which break SO(6) → SO(4) × SO(2) and SO(6) → SO(5). we should find a stable vacuum with G = SU(2) × SU(2) isometry preserving N = 2 supersymmetry. If such solution would exist. g Wab = − . An important point is that the T 11 compactification does not seem to be a solution of the N = 8 theory [68]. The generic superpotential now is given by ik    (6. As pointed out by S.3) 4e 4 in the gravitinos supersymmetry transformations. The low-energy theory for the AdS5 × T 11 compactification We collect here. Ferrara. Lifting this solution to ten dimensions should give a metric which is the warped product of AdS5 and a deformation of T 11 with isometry reduced to SU(2)diag .1) σ Tr(Ai Bj Ak Bl ) σ r which has conformal dimension ∆ = 3 and therefore corresponds to a marginal deformation preserving supersymmetry. at the critical point. this gives a supersymmetry operator of the form  1 λ+µ + eµ−λ + e−λ−µ δab . Ceresole.2) W = λ ij  kl + µ σ r (σr )j l Tr Ai Bj Ak Bl . 1) + (1. leading to two families of vacua: (i) 6 → (2. as a further application of the matter coupled 5D gauged supergravity. in order to preserve N supersymmetries. N eigenvalues of the form s 3 (6. According to the analysis of [68]. (ii) 6 → (3. one can study the stationary points of the potential and look for solutions interpolating between them. (6. One is interested in the construction of the low-energy theory in order to study the deformations and RG fluxes of the dual field theory. G. 1).A. From the supergravity point of view.4) ± − 2 V|∗ . Dall’Agata / Nuclear Physics B 585 (2000) 143–170 165 6. In fact.

the Q manifold must contain at least the G = SU(2) × SU(2) isometries and two zero-modes: i. The result is that the polynomial (3. the theory should be described by the massless graviton multiplet. Ceresole. To build this model. . we must study the potential given by the three scalars respecting this invariance [68]. we expect that the low-energy gauged supergravity states correspond to the theory given by fields which are the products of two singletons. Bi B¯ j and Tr (Ai A¯ i + Bi B¯ i ). since they have to correspond to the moduli of the conformal field theory.3) describing the S target manifold is given by   (6. We find that the potential becomes in this case   3  V = − g 2 3f 2 (λ) α 2 + β 2 sinh2 λ + 3 cosh2 (2λ) − cosh(4λ) . α.4) one finds that eight supersymmetries are preserved.6) αξr3 + βξr ξA2 + ξB2 + γ ξr ξb2 + δξb ξA2 − ξB2 = 1. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 To obtain vacua of the type (ii). β. Using (6. the AdS/CFT foresees the value of the Chern–Simons couplings CI˜J˜K˜ through the computation of the anomalies in the boundary theory. As already claimed in [28]. We can say something more on the S manifold of the vector multiplet scalars. β are the three scalars and f (λ) ≡ e −e 3λ . we need all the predictive power of the AdS/CFT correspondence and the spectrum analysis performed in [28]. The correct manifold must anyhow fall in the classification of [54].166 A. 8 (6. Using the [28] results and notations. For example. 9 This means that their masses are less than or at most equal to zero. ξA and ξB denote the ambient coordinates corresponding to the R-symmetry. It is known that there always exists a point cI where the metric becomes flat (δI J ) and the cubic polynomial takes the standard form. D = 5 gauged theory corresponding to the T 11 low-energy model. γ and δ are constants and ξr . ξb . and six hypermultiplets corresponding to the Tr (AB) and W 2 operators.. In conclusion. At the extremum one finds the potential V = − 34 g 2 and the supersymmetry operator Wab = − 32 δab . As for the maximally symmetric cases of type IIB on AdS5 × S 5 or M-theory on AdS4/7 × S 7/4 . There are no tensor multiplets. and one retrieves the highest symmetric S 5 solution. two of its scalars must be massless. Although this is not enough to uniquely fix the scalar manifold M = S ⊗ Q. λ where α. corresponding to the stress-energy tensor Wα W¯ α˙ + · · · from the boundary point of view. since they should be described by the product of at least three singleton states Tr [Wα (AB)]. In our case this happens for cr = 1 and cA = cB = cb = 0 and fixes 9 We would like to remark here that the correct prescription can only be given by identifying which truncation of the supergravity spectrum is consistent with the ten-dimensional equations of motion. the seven massless vector multiplets corresponding to the SU A (2) × SU B (2) × UBetti(1) conserved currents Ai A¯ j . we can still extract some useful information. Betti and SU A (2) × SU B (2) symmetries. the aforementioned flux from the T 11 to a solution with residual SU(2)diag symmetry can only be studied within the N = 2. G.5) −2λ where λ.e.

(A. Anyway. At this point one can go farther and see whether this manifold corresponds to one of the homogeneous spaces classified in [54]. . The five-dimensional (γ a )α β matrices satisfy the Dirac algebra  a b γ . since the cubic surface is specified by (6.7) 2 whereas δ is still free. 2 which is raised and lowered with the invariant USp(2) tensor 12 =  12 = 1 as follows: ϑ mi =  ij ϑmj . Ceresole. D’Auria for his interest in our work. The vector ones are raised and lowered with the flat metric ηab = {− + + + +}. G. Unfortunately. This research is supported in part by EEC under TMR contract ERBFMRX-CT96-0045. Ferrara for helpful conversations and A. this does not spoil our hope to study the minima of the potential for such a model in the future. 4) are symplectic-Majorana spinors carrying the USp(2) doublet index i = 1. . Dall’Agata / Nuclear Physics B 585 (2000) 143–170 167 1 β =γ =− . (6. We have also profited greatly from discussions with R. θ mi ). which can be computed explicitly evaluating the three-point functions of the corresponding anomalies.2) where λ¯ is the usual Dirac conjugate and C is the charge conjugation matrix satisfying tC = −C = C −1 . (A.A. Appendix A. The symplectic-Majorana condition on a generic spinor λαi reads λ¯ i ≡ λ†i γ 0 = tλi C. . . Acknowledgements We are glad to thank S.7) up to the δ coefficient. tC = −C = C −1 . 4) are the ordinary space-time coordinates and θ mi (m = 1. . . ϑim = ϑ mj j i . the numerous valuable comments and for his continuous encouragement. . γ = 2 ηab . . van Proeyen also for pointing out some misprints. The spinorial indices can be naturally raised and lowered through the use of the charge conjugation matrix Cαβ .6) to the form of one of the three families classified in [54]. Notations and conventions The five-dimensional superspace is spanned by the supercoordinates Z µ = (x µ . α = 1. as there is no SO(7) rotation reducing the δ piece of (6. We are particularly grateful to R. it is easy to prove that the given polynomial cannot correspond to a homogeneous space.6) and (6. where x µ (µ = 0. .1) The flat superspace indices are a = (a α). Kallosh and we thank her for clarifying to us the peculiarities of the various braneworld scenarios.

Regge. van Nieuwenhuizen. Grav. Phys.K. [19] M. B 242 (1984) 244.3) where γ [n] means the antisymmetrized product of n γ matrices with weight one: γ [n] = γ [a1 · · · γ an ] . Math. Günaydin. Lett. 1989. Fre’. The vacua of 5D.a 1 . Math. North-Holland/World Scientific. Consistent SO(6) reduction of type IIB supergravity on S 5 . T. G. Ann. hep-th/0002028. Phys. Salam. [7] M. hep-th/0002228. [12] M. t (A. Klebanov. [18] H. Phys. Zagermann. Sadrzadeh. Roˇcek (Eds. hep-th/9806051. H. N. [5] R.S. Vaman. Phys. [10] A.M. G. .H. p! where the wedge product between forms is understood. E. Marcus. [8] M. Townsend. Sierra. [17] H. R. 135 (1981) 237. B 144 (1984) 41. Kim. [9] G. Phys. Stelle. Sezgin (Eds. Günaydin. hepth/9803235. Townsend. Romans. hep-th/0003103. Rev. Supergravity in Diverse Dimensions. Lett. Nucl. P. N = 2 Maxwell–Einstein supergravity theories coupled to tensor multiplets. 2 (1985) L11. M. Castellani. B 143 (1984) 103. Nucl. in: S.K. B 157 (1985) 379. Supergravity and Superstrings: a Geometric Perspective.   σ αβ  β Γ [n] ≡ C ασ γ [n] σ .A. K. Maina. Nucl. [20] M.S. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 To have matrices with fixed symmetry properties we define  β  β α  σ Γ [n] β ≡ C αρ γ [n] ρ Cσβ . Günaydin. Chamseddine. L. 53 (1984) 332. B 133 (1983) 72. Phys. D’Auria. Hawking. Lett. Günaydin. P. [11] M. References [1] A. [14] S. Lu. hep-th/9912027.. Maldacena. B. Phys. World Scientific. Phys. N = 2 gauged Yang–Mills/Einstein tensor supergravity: abelian case. D. Rev. Phys. 1991. [2] L. T.). [3] E. P. Waldram. t Γ ab = Γ ab t and Γ abc = Γ abc . B 253 (1985) 573. Cvetic. hep-th/9711200. M. G. and (A. Pope. Cambridge University Press. Witten. Fre’. H. G. Phys.R. Γ [n] α ≡ γ [n] α . Sierra. [6] M. L. Class. 2 (1998) 231. Rev. Phys. A. P.M. Phys. Sierra. Cremmer. 2 (1998) 253.K. Superspace and Supergravity. Theor. Warner. Günaydin. Adv. B 428 (1998) 105. Nastase. Thus Γ a = −Γ a . N. Quant. Nucl. Townsend. On the non-linear KK reductions on spheres of supergravity theories. C. Lett. Phys. [15] E. hep-th/9802150. P. M. P. hep-th/9802109. Polyakov. Lukas. Romans. Phys. Lett. Tran. Phys. Günaydin. The gauging of five-dimensional. B 272 (1986) 598. Ovrut. Theor. Lett. Sierra. [4] A. Günaydin. Lett. [16] M. Adv. D. Zagermann. Gubser.168 A. Phys. I. D’Auria. D 32 (1985) 389. 1980. E. B 96 (1980) 89. B 552 (1999) 246. Nicolai. A.4) We define the tangent space components of a generic p-form Φp according to [61] Φp = 1 a e 1 · · · eap Φa p . Ceresole. D 59 (1999) 086001. [13] J. Γ [n] αβ ≡ γ [n] α Cσβ .)..

hep-th/0004103. D’Auria. K. E. Nucl. D 61 (2000) 066001. Nucl. Entropy of black holes in D = 5. 1977. D. Pernici. M. D’Auria. Andrianopoli. Altendorfer. Lett. hep-th/9907216. B 359 (1991) 705. Rev. D. R. P. S. Lykken. Ferrara. Phys. D. B. hep-th 9909058. Oxford. hep-th/9906182. Phys. JHEP 11 (1999) 009. Phys. Math. hep-th/9905221. B 508 (1997) 147. A. A. S. Kallosh. G. K. Teitelboim. Bagger.A. B 357 (1995) 76. G. G. Phys. Klebanov. Tseytlin. D 59 (1999) 025006.S. Commun. M. A. Lett. Nucl. Pomarol. Ceresole. Nemeshansky. Gherghetta. hep-th/9912053. M. O. J. Dall’Agata. Chamseddine. N = 2 supergravity and AdS central charges. Karch. W. Nucl. D. Phys. B 411 (1997) 39. Supersymmetric Randall–Sundrum scenario. Phys. K. Behrndt. hep-th/9909047. Phys. Cacciatori. P. K. Ceresole. Sundrum. D 55 (1997) 3647. Grav. A. Romans. Zanon. H. Klebanov. JHEP 9905 (1999) 026. Jatkar. Phys. B 468 (1999) 46. hepth/9610155. Randjbar-Daemi. Linde. Cadavid. Ferrara. Shmakova. Verlinde. Klemm. hep-th/9807080. D’Auria.J. B 475 (1996) 94. Trieste. J. Chou. Nucl. D’Auria. Nucl. R. Renormalization group flows from holography. Phys. P. Ferrara. Nucl. Nucl. Kallosh. D’Auria. de Wit. Cvetic. A. Freedman. Warped compactifications in M and F theory. Lett. S. L. H. B 460 (1999) 281. Kallosh. Sabra.S. R. T. P. Van Proeyen. General relativity a la string: a progress report. hep-th/9903026. R. Zaffaroni. Lett. D. hep-th/9908076. Wong. Phys. JHEP 9812 (1998) 022. Shiu. A. Phys. p. Bertolini. Chan. Andrianopoli. Phys. Lett. hep-ph/0003129. B 460 (1996) 506. Ferrara. Gibbons. hep-th/9910117. hep-th 0002159. Rev. Fre’. S. A. Paul. L. L. Phys. hep-th/9510209. Nucl. Gubser. Townsend. Phys. M. Lett. S. M. L. hep-th/9603004.A. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] 169 M. T. Randall. R. G. hep-th/9904187. S. Class. D’Auria. Warner. S. Porrati. hep-th/9905104. Dall’Agata. B. Phys. hep-th/9705024. in: Proceedings Marcel Grossmann Meeting on General Relativity. DeWolfe.W. Gubser. G. B 153 (1985) 392. Randall. C.S. hep-th/9704142. Quant. Kallosh. B 556 (1999) 89. K. Girardello. JHEP 0001 (2000) 001. A. G. B 536 (1999) 199. R. Schalm. Phys. E. R. Klebanov. Randall. hep-th/9909070. Phys. hep-th/9112027. hep-th/9810126. B 511 (1998) 326. Ceresole. hep-th/9707236. hepth/9905226. 83 (23) (1999) 4690. S. Skenderis. B 259 (1985) 460. E. hepth/0003117. Greene. S. C. hep-th/9904017. hep-th/9807164. The shape of gravity. L. A. Pilch. Witten. 83 (17) (1999) 3370. Lett. R. Kallosh. Ferrara. JHEP 9911 (1999) 010. L. Rev. I. Rahmfeld. hep-th/9603142. hep-th/9506144. S. Petrini. 1975. 17 (2000) 1081. A note on warped string compactification. B 476 (1996) 397. S. Supersymmetric domain wall world from D = 5 simple gauged supergravity. supersymmetry and a c-theorem. Rev. Ferrara. R. . D. Witten. K. R. hep-th/0004077. ep-th/0003236. L. S. Rey. R. I. Regge.K. hep-th/9909134. Holography and compactification. Freedman. R.R. A. Hoˇrava. Dall’Agata. A. Ceresole. Phys.R. M. Gubser. T. hep-th/9910021. hep-th/9906064. P. Witten. R. van Nieuwenhuizen. Modeling the fifth dimension with scalars and gravity. I. hep-th/0001071. Bulk fields and supersymmetry in a slice of AdS. A. JHEP 0002 (2000) 005. A. A. R. Rev. Phys.S. Shmakova. 149 (1992) 307. Gubser.Z.P. N. S. Lechner. Gravity duals of supersymmetric SU(N) × SU(N + M) gauge theories. Linde.L. Pilch.A. Sundrum. Verlinde. Phys. P. Ceresole.R.K. 77. W. Fre’.

L. Phys. Phys. Bandos. Lechner. A. G.J. P. A 5 (1990) 461. Nucl. Int. Lechner. D’Auria. [64] J. . Mod. Nucl. Bandos. Phys. B 553 (1999) 317. Magri. Bertolini. M. Geom. J. B 222 (1983) 1. L. Mod. Phys. [65] B. Fre’. [63] A. de Wit. Ferrara. [68] M. Mod. R.J. Phys. Romans. [71] I. Klummer.170 A. Supersymmetry and Supergravity. B 208 (1982) 323. B 412 (1993) 479. World Scientific. [59] S. Zagermann. hep-th/9608093. N = 2 Yang–Mills/Einstein tensor supergravity. W. Phys. Phys.P. Cvetic. Günaydin. hep-th/9605032. A 12 (1997) 799. Porrati. Nucl. hep-th/9309143. M. [61] J. Bregola. hep-th/0004117. Witten. J.. Lett. [67] K. Maiani. 1986. J. J. Wess. Nucl. hep-th/9810227. P. B 268 (1986) 295. Phys. A 14 (1999) 4881. Dall’Agata / Nuclear Physics B 585 (2000) 143–170 [58] L. Cecotti. Nicolai. [70] I. Romans. K. Phys. B 154 (1985) 268. Ceresole. W. T. Ferrara. Proc. M. S. Phys. 23 (1997) 138. Tonin. Warner. M. Günaydin. H. [69] M. Princeton Series in Physics. 1982. [62] L. Andrianopoli. M. Lett. hep-th/9703099. Sabra. [60] S. Int. B 278 (1986) 598. Singapore. Candiello. Nucl. E. Phys. Girardello.A. Gauging the full R-symmetry group in five-dimensional. Nucl. Vth Silarg Symp. [66] L. N. B 269 (1986) 691. Pasti. Nucl. Behrndt. Bonora. Bagger. K. Bagger. Ceresole. Phys. L. M.

PII: S 0 5 5 0 .nl/locate/npe D-brane effective field theory from string field theory Washington Taylor Center for Theoretical Physics. Bldg. leading to E-mail address: wati@mit.2 . We find that the level truncation method converges for these terms within 2–4% when all massive fields up to level 20 are integrated out. The bosonic open string has a tachyonic instability around the usual string vacuum. although the convergence is slower than exponential. Some simple calculations are performed in open bosonic string field theory which validate this approach. MIT. Taylor). A particularly elegant formulation of covariant string field theory for an open bosonic string was provided by Witten [1]. The simplicity of this theory makes it feasible to perform interesting off-shell calculations. Aν ]2 for the zero momentum tachyon and gauge field on one or many bosonic D-branes. the entire classical action is contained in a pair of terms which are quadratic and cubic in the string field. Introduction String field theory provides an off-shell formulation of string theory which has the potential to address nonperturbative questions in a systematic fashion. 6-306. All rights reserved. We also describe a brane configuration in which tachyon condensation arises in both the gauge theory and string field theory pictures. revised 23 May 2000 Abstract Open string field theory is considered as a tool for deriving the effective action for the massless or tachyonic fields living on D-branes. 1.edu (W. We discuss the possibility of extending this work to determine the structure of the nonabelian Born–Infeld theory describing the gauge field on a system of many parallel bosonic or supersymmetric D-branes.elsevier.V.3 2 1 3 ( 0 0 ) 0 0 3 3 8 . Early work by Kostelecky and Samuel [2] and by Kostelecky and Potting [3] indicated that string field theory could be used to describe the condensation of the tachyon.V. In this string field theory. USA Received 9 March 2000. MA 02139. This provides a natural connection between recent work of Sen and Zwiebach on tachyon condensation in string field theory and unstable vacua in super Yang–Mills and Born–Infeld field theory. (Aµ Aµ )2 and [Aµ .Nuclear Physics B 585 (2000) 171–192 www. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. The level truncation method is used to calculate successive approximations to the quartic terms φ 4 . All rights reserved.  2000 Elsevier Science B. Cambridge.

Samuel and Potting suggest that string field theory may be a much more effective tool for asking nonperturbative and off-shell questions about string theory than was previously suspected. and we suggest that further insight into the process of tachyon condensation may be gleaned from consideration of simple D-brane models in which the tachyon. While Witten’s formulation of open string field theory 1 Another approach to describing the stable vacuum was taken in [4]. renewed interest in tachyonic instabilities related to D-branes led Sen to propose that the condensation of a tachyon in the open bosonic string should be understood as the decay of an unstable D-brane in the bosonic string theory. We perform a systematic truncation of string field theory. the convergence properties of the theory are poorly understood. the energy at the minimum of the potential differs from the energy in the unstable vacuum by 99% of the energy of the bosonic D-brane. These results increase significantly our confidence in the viability of string field theory as a tool for calculating higher order terms in the action for the massless/tachyonic fields on the D-brane. The work in this paper provides new evidence that string field theory provides a systematic and completely convergent framework in which to study many features of string theory. in which the authors demonstrated that when the string field theory is truncated to states of level 4. there are still many fundamental questions about the structure of these theories. We discuss briefly how these results may be extended to study the structure of the nonabelian Born–Infeld theory. however. While string field theory has been studied for quite some time. The main goal of this paper is to initiate a systematic study of how effective field theories on D-branes can be derived from open string field theory. the structure of the vacuum and the role of the tachyonic instability were poorly understood at that time. Even in the context of the simplest string field theory. namely the quartic terms φ 2 and (Aµ Aµ )2 in the abelian theory and [Aµ . The vacuum energy of the open bosonic string field theory in the “true” vacuum should thus differ from that in the unstable vacuum by precisely the mass energy of the unstable D-brane. Aν ]2 in the nonabelian theory. including all fields up to a fixed level n for various values of n 6 20. This conclusion was given very strong support by a recent paper by Sen and Zwiebach [6]. 1 Since. and Sen argued that this mass difference should be calculable using the bosonic open string field theory [5]. this paper primarily focuses on the open bosonic string field theory. Recently.172 W. the unstable vacuum and the stable vacuum can all be described in the language of supersymmetric Yang–Mills theory. Except for the discussion of tachyon condensation in the last section. . It would obviously be of great interest to develop an analogous systematic approach to computations in supersymmetric open [7] and closed [8] string field theory. We find that the contributions to each of the quartic terms from fields at level n decrease monotonically and appear to give completely convergent series with slower than exponential convergence. that of the open bosonic string. We focus on some of the simplest terms in the effective actions for the tachyon and massless vector field in the open bosonic string. Taylor / Nuclear Physics B 585 (2000) 171–192 another more stable vacuum. the significance of this result was not generally appreciated. The result of Sen and Zwiebach and the earlier work of Kostelecky.

this formalism may be problematic due to the necessity of considering higher order contact interactions [10]. The open bosonic string has a µ system of 26 matter oscillators αn and ghost oscillators bn . In Section 3 we discuss the possible extension of this work to studying the nonabelian Born–Infeld theory and tachyon condensation in D-brane systems.1 with a brief review of bosonic open string field theory and the Fock space representation of the Witten vertex. Open bosonic string field theory In this section we perform some simple examples of calculations in the open bosonic string field theory to demonstrate the effectiveness of this approach in finding terms in the effective action for massless or tachyonic fields on a D-brane.2 we calculate the terms in the effective action which are quartic in the tachyon by integrating out all non-tachyonic fields. Several alternative formulations of open superstring field theory have been suggested [11. in order to fix conventions and notation.1. Throughout this paper we use the conventions of Kostelecky and Samuel from [2]. which were carried out using the symbolic manipulation program M ATHEMATICA. cn .12]. 2.9]. for a more detailed review of open string field theory see [14. Taylor / Nuclear Physics B 585 (2000) 171–192 173 can be extended to the supersymmetric theory [7.15].W. We begin in Subsection 2. In Subsection 2. We summarize here briefly the basic formalism we will need.16–18]. Review of notation and formalism As formulated by Witten [1]. in which he finds a similar result to that of Sen and Zwiebach for tachyon condensation in a supersymmetric theory. following [2. In Subsection 2. In Section 2 we review the basic structure of open bosonic string field theory and describe the calculation of the quartic terms in the tachyon and vector field potential at zero momentum. (1) S = 0 Φ ? QΦ + 2α 3! where Q is the BRST operator and ? is the string field theory star product. Some details of the exact calculations. are described in the appendices. which were performed using the symbolic manipulation program M ATHEMATICA are given in the appendices. covariant open string field theory is described in terms of a string field Φ which contains a component field for every state in the first-quantized string Fock space. which act on the vacuum |0i through . Recent work of Berkovits [13]. indicates that his alternative formulation of the supersymmetric open string field theory [12] may present a viable framework for systematic calculations. For all these calculations we describe the first few terms analytically and summarize the results of the summation of higher level terms. The string field theory action is Z Z g 1 Φ ? Φ ? Φ.3 we calculate the terms quartic in the gauge field by integrating out the tachyon and all massive fields. Some details of the calculations. 2.

In the Feynman–Siegel gauge. This restricts us to states in which an equal number of b and c raising operators act on the vacuum |0i. R) invariant vacuum |i through |0i = c1 |i. p is the momentum of the state S. cm } = δn+m . etc. α−1 |0i are the allowed states at level 1 (of ghost number 1). From (3) we can write the kinetic terms in the string field action Z 1 S2 = 0 Φ ? QΦ 2α  Z  1 1 (4) d 26 x 0 −φ 2 + Bµν B µν − β 2 + · · · = 2 α  + ∂µ φ ∂ µ φ + ∂µ Aν ∂ µ Aν + ∂µ Bνλ ∂ µ B νλ − ∂µ β ∂ µ β + · · · for the terms of interest up to level 2. the vacuum is the unique state at level 0. given by the level of S minus 1. The string field Φ is restricted to a sum over states with ghost number 1 in the classical action. We will work in Feynman–Siegel gauge. These are the states associated with the tachyon and gauge field on the D-brane. Taylor / Nuclear Physics B 585 (2000) 171–192  αnµ |0i = 0. the leading terms of interest in an explicit expansion of the string field Φ are  µ µ ν + βb−1 c−1 + · · · |0i. bn |0i = 0. . {bn . cn |0i = 0. The vacuum |0i is related to the SL(2. hS| is the BPZ dual state to |Si produced by acting with the conformal transformation z → −1/z. The quadratic term evaluated on a state |Si is simply Z  (3) |Si ? Q|Si = hS|c0 α 0 p2 + 12 M 2 |Si. αm = nηµν δn+m . where 12 M 2 is the string mass operator. We denote the Fock space of physical states by H and its dual space by H∗ . n > 0. n > 0. n > 0. Thus. we will only be interested in fields which couple through the cubic string field vertex to two tachyons or two gauge fields. In our calculations. and the dual vacuum satisfies h0|c0 |0i = 1. where we can impose a further restriction to states |Si satisfying b0 |Si = 0. The level of a state is defined to be the sum of the oscillator numbers n used to produce µ the state from the vacuum |0i. (2) Φ = φ + Aµ α−1 + √1 Bµν α−1 α−1 2 It is possible to write both the quadratic kinetic term and the cubic interaction term from the string field action (1) in terms of the Fock space representation of the string field. we need only consider massive fields which have both an even level number and an even number of space–time indices.  µ ν αn . thus.174 W.

(6) × exp 12 αn(r)µ Nnm The indices r. − 3. 0. −1/ 3. (5) where V is a state in the tensor product space H∗ ⊗ H∗ ⊗ H∗ . An explicit representation of this state in the string Fock space was found in [16–18] and can be written as  hV | = δ(p(1) + p(2) + p(3) ) h0|c0(1) ⊗ h0|c0(2) ⊗ h0|c0(3)  rs (s)ν rs (s) ηµν αm + cn(r)Xnm bm . the interactions at zero momentum are √  5 11 24 5 · 23 2 µ Bµ Aν Aν S3 = κg φ 3 − √ Bµµ φ 2 − 2 βφ 2 + 2 φAµ Aµ − 3 3 35 32 2 √  28 2 µν 11 · 24 µ + 5 B Aµ Aν − βAµ A + · · · . The coefficients Nnm nm rs N¯ nm . X rs are given in terms of the 6-string Neumann functions act in. A table of the coefficients Nnm nm From (6) the cubic interaction terms in the string field theory can be written down for an arbitrary set of 3 fields. 1/ 3. which are given by I I (−1)r n (−1)s m 1 1 dz dw rs = (−1)n(r−1)+m(s−1) f (z) f (w) N¯ nm 2 nm 2πi 2πi (z − w) zr zs (8) with f (z) = z(z2 − 3) 3z2 − 1 and z1 . 2. . Taylor / Nuclear Physics B 585 (2000) 171–192 175 The cubic interaction terms in the string field action can be described in terms of Witten’s vertex operator V through Z Φ ? Φ ? Φ = hV |(Φ ⊗ Φ ⊗ Φ). s take values from 1. Nnm  rs rs r(s+3) = −n N¯ nm − N¯ nm . . (9) (10) rs .W. (7) Xnm  rs rs r(s+3) ¯ ¯ . m < 9 is given in Appendix A. r + 2. s 6 6 through  rs rs r(s+3) (r+3)s (r+3)(s+3) = 12 N¯ nm + N¯ nm + N¯ nm + N¯ nm . 3 and indicate which Fock space the oscillators rs . s = r + 1. X rs with n. Xnm = n Nnm − Nnm rs are fixed by using (7) and the cyclic symmetry of the For s < r the values of Xnm coefficients under r → (r mod 3) + 1. For the fields appearing in (2). z6 = √ √ √ √ 3. s = r. s → (s mod 3) + 1. (11) 3 35 where κ = 37/2/27 . . ∞. When all momenta are zero we only need the Neumann functions with n. m > 0. . . 1 6 r.

the µ contributions from the level 2 fields β. We denote by γˆ (n) the contribution to γ from fields at level n. (14) The quartic term produced by integrating out the level 2 fields (13) represents 72% of the total (14). and as a test of our algorithm for calculating contributions from fields at arbitrary level. Terms quartic in tachyon field The answers to many physical questions in string field theory may be very well approximated by truncating the string field action at a finite level. 2a The (tree-level) diagram for such a term is shown in Fig.02.2. The results of this calculation are summarized in Table 1. Kostelecky and Samuel calculated the contribution from level 4 fields. Thus. and by γ (n) the cumulative γ found by adding the contribution from all fields at levels m 6 n. Taylor / Nuclear Physics B 585 (2000) 171–192 Fig. 2.176 W. as was shown in [2. we find that the exact result from [20] .112 fields at levels 6 20. where they considered the effective quartic term in the tachyon field φ 4 which arises from integrating out all the massive scalar fields in the theory. where γ ≈ −1. 1. for example. In [2]. φ κ 2 g2 α0 2 32 2 32 2 27 Sφ4 = − The exact quartic term in the tachyon potential was determined in [20]. 11) to be 2     2   1 5 1 11 4 2 2 0 34 − 26 · √ = −κ g α (13) φ4. and is given by κ 2 g2 α0 γ φ 4 . 1. A more detailed description of the contributions from each field up to level 6 is given in Appendix B. Bµ to the φ 4 term in the effective tachyon potential are seen from (4. A calculation of this type was carried out by Kostelecky and Samuel in [2]. and showed that these produce an additional 12% of the total quartic term.75 ± 0. Tree diagram contributing to φ 4 term. we have used M ATHEMATICA to carry out this truncation exactly up to order 20. As a first test of how well the level truncation method does in reproducing quartic terms in the effective action on a D-brane. After summing the contributions from all 13. For each massive scalar field ψ whose quadratic term and coupling to φ 2 are given by a (12) Sψ = ψ 2 + cψφ 2 2 there is a term in the effective potential for φ of the form c2 4 φ .19].

96 ±0. Taylor / Nuclear Physics B 585 (2000) 171–192 177 Table 1 Contributions at each level to coefficient of φ 4 n # of fields 2 2 4 7 6 20 8 55 10 139 12 331 14 747 16 1618 18 3375 20 6818 γˆ (n) γ (n) γ (n) /γexact − 2·17 3 ≈ −1. the successive approximations seem to have slower than exponential falloff.646 0.04423 −1. There are several significant aspects of how the successive approximations to this term in the effective action behave as the level number increases.01 3 − 1399 38 2 ·7·643463 2 − 316 ·5 − 167·1846847 2·320 − 2·5·193·241·1341187 328 − 5·2912243·232794533 334 ·11 2 − 2 ·11·1775214529932629 337 ·13 − 298001292970739836603 2·345 ·5 2 − 2 ·7·21159416989·1463230224529 352 ·17 − 31·1061·1319·164809·26468211709651 357 ·19 Fig.72 ±0.93 ±0.677 0.01009 −1.84 ±0.628 0.01 ≈ −0.670 0.684 0.01 ≈ −0. It would be very interesting to have a better understanding of the convergence properties of these terms from a study of the asymptotic properties of the states and their couplings.556 0.00634 −1.94 ±0. the approximations seem to be completely convergent. Second. with no sign of bad behavior at high level as one might expect if the series were only asymptotic but not convergent. 2.02727 −1. A graph of the coefficient of the quartic term coming from the successive level truncation approximations is shown in Fig.95 ±0.660 0.01 ≈ −0. is reproduced to within approximately 4%.01 ≈ −0.25926 −1.01334 −1. Approximations to coefficient of φ 4 term.600 0.01 ≈ −0. 2.92 ±0.08371 −1.00789 −1.96 ±0.89 ±0.01 ≈ −0. First.W. but do not seem to follow a simple power law.01 ≈ −0.01 ≈ −0.259 0.01 ≈ −0.01848 −1. .472 0.21323 −1.96 ±0.

this term breaks translation invariance. As we will discuss in the next section. it seems most reasonable to integrate out the tachyon field along with the massive states when constructing an effective action for the vector field. Again. the kinetic term for the vector field in Feynman–Siegel gauge is simply 12 ∂µ Aν ∂ µ Aν . The quartic term in the T-dual picture analogous to (15) is of the form (Xµ Xµ )2 . Let us first consider the abelian U(1) theory associated with a single bosonic D-brane.3. The usual approach to deriving an effective action in a theory with fields at many energy scales is to integrate out the fields with very large masses keeping the light or massless fields. since by choosing Feynman–Siegel gauge we have broken the gauge invariance of the effective U(1) theory. Terms quartic in gauge field µ Now let us turn to the massless gauge field Aµ associated with the spin 1 states α−1 |0i. the gauge noninvariance of this term suggests that contributions to the quartic term from higher string modes should cancel this term. we expect that this effective action should take the form of a bosonic Born–Infeld action. A related argument for the conclusion that the A4 term should vanish in the abelian theory is that if we compactify the theory in some set of dimensions and then perform a series of T-duality symmetries which change the Neumann boundary conditions on the open string to Dirichlet boundary conditions in those directions. perhaps even a worse transgression than the breaking of gauge invariance by (15). so that in the true effective U(1) gauge theory there is no A4 term. For example. This gives a low-energy effective theory for the light fields in the theory which has a clear physical significance. It may seem strange that we have integrated out the tachyon field and kept the massless vector field. . as this term does not appear as the 0-momentum part of any expression which is invariant under the U(1) gauge symmetry we expect to have in the effective gauge theory found after integrating out the tachyonic and massive string modes. In this theory the only Lorentz invariant term which is quartic in Aµ is (Aµ Aµ )2 . It is less clear how to interpret the action for the vector field on a bosonic D-brane which arises when we integrate out the tachyon. Taylor / Nuclear Physics B 585 (2000) 171–192 2. This argument is not completely conclusive. From the vertices in (11) of the form φAµ Aµ we see that there is a term in the effective action of the U(1) theory    4 2   2 2 1 2 µ 2 2 0 128 A = κ g α (15) Aµ Aµ . Because the energy scale of the tachyon is the same as that of the massive string states. we have an effective theory of a bosonic D-brane of dimension p < 25 whose position in the transverse spatial directions is described by a set of coordinates Xµ which are T-dual to the constant gauge field components Aµ giving rise to nontrivial holonomies around the compact directions in original picture.178 W. which is the kinetic term for the massless vector field in Lorentz gauge ∂µ Aµ = 0. Indeed. which has negative mass squared. we are interested in finding quartic terms in the gauge field. κ 2 g2 α0 A µ 2 2 3 81 The appearance of this term in the effective action for the world-volume gauge field may seem rather surprising. however. however. Such a term is indeed induced by integrating out the tachyon at tree level.

3. 3.W. Let us consider in particular those terms with two distinct pairs of indices on the gauge field. such as Ai Ai Aj Aj with i 6= j . In any case. Diagrams contributing to terms quartic in vector field at level 2. 3). at least at quartic order. The diagrams associated with these contributions are depicted in Fig. there are terms arising from the field Bij . It is instructive to consider how this term is computed. At level 2. These additional terms come to . For terms of this type. the field β and the µ diagonal components Bµ with µ 6= i. j contribute in the same fashion as we have found previously through the couplings of the forms β(Aµ Aµ ) and Bνν (Aµ Aµ ) (diagrams (a) and (b) in Fig. 3. so we include some of the details in this particular case. =κ g α 59049 (18) In addition to these terms. Bµν . the effective action which appears when integrating out the tachyon as well as the massive fields in the theory seems to be quite well behaved. Taylor / Nuclear Physics B 585 (2000) 171–192 179 Fig. The total contribution from these diagrams to the quartic term in the gauge field is   2 2 2 0 71168 Aµ Aµ . A consideration of the effects of the fields of level 2 and higher bears out the conjecture that the term (15) is completely cancelled by the infinite tower of massive string states. associated with diagrams (c) and (d) in Fig. there are contributions to the quartic gauge field term from the fields β. giving terms in the effective action        1 5 · 27/2 2 1 11 · 24 2 − 24 · 2A2i A2j κ 2 g2 α0 5 5 2 3 2 3   22912 (2Ai Ai Aj Aj ). (17) = −κ 2 g 2 α 0 59049 The fields Bii and Bjj together contribute      (32 − 5) · 27/2 5 · 27/2 2 2 0 − A2i A2j κ g α −2 · 35 35   2 2 0 17280 (2Ai Ai Aj Aj ). (16) −κ g α 59049 This term cancels 76% of the spurious term generated by the tachyon.

and find the same result for the overall coefficient. We have performed a similar analysis for the terms of the form A4i . and the relevant coupling terms are simply √ √  27 2  5 · 23 2 µ ν Tr Bµ Aν A + 5 Tr B µν Aν Aµ + B µν Aµ Aν − 5 3 3  11 · 24 Tr βAµ Aµ . however. Now. (21) − 35 We can now go through the calculation of the terms quartic in Aµ for the nonabelian theory just as we did above in the abelian case. it is necessary to separately compute the coupling for each ordering of the fields in the vertex. however. We have thus explicitly shown that by including the fields at level 2. We have continued this calculation further using M ATHEMATICA. we find that the term (16) decomposes into a sum of two terms with distinct ordering of the A’s . Let us consider the nonabelian theory which arises when we add Chan–Paton factors to the strings in the original string field theory. 11) in the string field action are modified by simply taking the trace of each term. If string field theory is sufficiently well behaved. Before presenting the results of this calculation. If the Chan–Paton factors range from 1 through N then all the fields in the string field expansion (2) become N × N matrices. determining the exact contribution to the quartic term arising from all fields up to level 16. the quartic term in the effective gauge field action will take the form (20). we should hope that by keeping track of the nonabelian structure of the fields in the U(N) theory. we briefly discuss the modifications which arise in the nonabelian theory. For the terms describing the couplings of the level 2 fields to a pair of gauge fields. Indeed. Taylor / Nuclear Physics B 585 (2000) 171–192 √       1 2 · 28 2 2 2 2 2 0 65536 (2Ai Ai Aj Aj ). so that the term arising from the tachyon has been 97% cancelled at level 16. In the nonabelian theory with Chan–Paton indices. The gauge field. which in the 0-momentum theory reduces to a term of the form [Aµ . as is necessitated by Lorentz invariance. κ g α − (Ai Aj ) = −κ g α 2 35 59049 2 2 0 (19) Summing these. however. we find that the contributions at each level serve to decrease the total coefficient of the quartic term. we arrive at the total coefficient described by (16). becomes a U(N ) gauge field. this subtlety is not important. It is important to note. The term (15) produced by integrating out the tachyon is unchanged except for the addition of an overall trace. We expect that the effective action for this gauge field will have a leading term of the form F 2 . the quadratic and cubic terms (4. we should find that instead of vanishing identically. (20) Thus.180 W. we might hope to find that summing the contributions to this term from successive level truncation will give us a series of terms which converge to a finite value for the coefficient of the term (20). in particular. Because the Neumann coefficients Nnm reversal of order in the 3 strings at a string vertex. Aν ]2 . we cancel 76% of the spurious A4 term in the U(1) theory induced by the tachyon. that this trace must be taken before the order of the fields in the term is rs are not necessarily symmetric under a modified.

Aν .711    κ 2 g 2 α 0 γ1 Tr Aµ Aµ Aν Aν + γ2 Tr Aµ Aν Aµ Aν . From Table 2 and Fig.0212 −0. Thus.03878 ≈ −0. γ± = The successive approximations to γ± at levels n 6 16 are graphed in Fig.721 −0.02528 ≈ −0.717 −0.713 −0. (23) 2 The only way to form a gauge/translation invariant combination of the two quartic terms in (22) is to have a term of the form (20).70. we see that string field theory satisfies the important consistency check that the effective U(N) vector theory has a gauge-invariant quartic term for Aµ in the effective . γ− → ≈ 0.689 0.1875 −0. These results are summarized in Table 2.736 0. γ+ (n) γ1 γ− (22) where 38400 32768 . γ+ → 0.0344 0.08221 3 ≈ −0.761 0. γ2 = − .742 0. (22) can be rewritten in the form  2  (24) −κ 2 g 2 α 0 (γ− /2) Tr Aµ .65031 3 8 − 2 ·6229 ≈ −0.01157 ≈ −0.000 0.684 −0.00282 ≈ −0. 3 contribute to γ1 while diagram (d) contributes to γ2 .00414 (n) γˆ2 0 ≈ 0. 4.780 0. We have computed these coefficients from terms at all levels up to 16. 59049 59049 In this decomposition.58020 34 9 ·52 2 − 9 ≈ −0.749 0. γ2 . this will be the situation at every level.55493 3 17 − 213 ≈ −0.01933 ≈ −0.741 0.580 0.00771 ≈ −0. with nonzero contributions to both coefficients γ1 and γ2 . 4.819 0. A more detailed description of the contribution from each field up to level 6 is given in Appendix B.662 −0.930 −0. the diagrams (a)–(c) of Fig.0908 0.00409 ≈ −0.0434 0.728 −0.7901 −0. Taylor / Nuclear Physics B 585 (2000) 171–192 181 Table 2 Contributions at each level to coefficients of terms of order A4 (n) γˆ1 n 27 ≈ 1.W. we see that indeed as the contribution from all levels is considered.715 −0.712 −0.0285 0. When this is the case. An important feature of these results is the behavior of the sum and difference of the coefficients γ1 .637 −0.01150 ≈ −0.687 −0.0243 0.674 −0. (25) At level 16. In general.0588 0.00000 15 − 210 ≈ −0.680 −0. γ1 = − γ1 ± γ2 .00645 ≈ −0.732 −0. the γ+ term produced by the tachyon has been cancelled to within 3%.790 0. which appears when γ+ = 0. and the coefficient of γ− has converged to within approximately 2% of its apparent asymptotic value.11113 15 3 0 2 4 6 8 10 12 14 16 ≈ −0.555 0.00206 (n) (n) (n) γ2 1.00551 ≈ −0.

Taylor / Nuclear Physics B 585 (2000) 171–192 Fig. Approximations to coefficients of A4 terms. The results of this calculation seem to have nice convergence properties.234. This discrepancy does not depend on the details of the calculation at high level. Aν ]2 term with the kinetic and cubic terms for the vector field in the string field theory action √ √  25 2    1 2 0 (∂µ Aν ) − igκ α (26) √ ∂µ Aν Aµ . even though gauge invariance is apparently broken when only a subset of the fields are integrated out.182 W. Aν . with slower than exponential convergence just as the terms discussed previously for the tachyon. 3 2 3 3 From these terms. assuming that the results for γ2 from successive level truncation are monotonically decreasing as they seem to be. We leave a resolution of this puzzle as an outstanding problem. so it is difficult to see how the needed factor could arise. but in any case it seems likely that this discrepancy can be resolved without requiring dramatic changes in the picture we have presented here. (27) √ This result disagrees with (25) by a significant margin. indeed. This discrepancy is somewhat puzzling. A] or at order [A. but most of these would affect (27) by an overall factor of 2. It is interesting to compare our results for the coefficient of the [Aµ . There are numerous factors of 2 and 2 appearing in the calculation of (26). Another possible explanation of the discrepancy is that the gauge of the U(1) theory has been fixed in such a way as to introduce spurious terms either at order ∂A[A. perhaps by a factor of 3 or 2 2. One obvious possibility is that √ there is an error somewhere in the calculation. It would be nice to have a better theoretical understanding of the relative rates of convergence of these terms in the effective action. A]2 . 4. theory arising when all massive fields are integrated out. . we might expect to find a quartic term of the form (24) with coefficient γ−? = 29 /37 ≈ 0. the result (27) seems to already be ruled out from the explicit calculations of the quartic term at level 0 and level 2 given in detail above.

see [30]. For a review of recent work on the nonabelian Born–Infeld action. either in the abelian or nonabelian theories (for some discussion of these terms. At this point there is no systematic understanding of the α 0 corrections to the Born–Infeld action. Nonabelian Born–Infeld In supersymmetric string theories. In Subsection 3. In the abelian theory. When there are multiple parallel D-branes of the same type. These terms are not uniquely defined due to the possibility of changing the apparent form of the effective action through field redefinitions. Because the field strength components Fµν do not commute. 3.1 we discuss the calculation of higher-order terms in the gauge field effective action. In Subsection 3. (28) is corrected by terms of higher order in α 0 containing derivatives of F .24] and M(atrix) theory [25. Determining the exact structure of this nonabelian action is a long-standing problem in string theory. Tp is the brane tension and Fµν is the U(1) field on the world volume of the D-brane.22] Z p (28) S ∼ −Tp d p+1 ξ − det(ηµν + Fµν ). (29) S ∼ Tp 4 8 4 A similar action describes the dynamics of a D-brane in the bosonic theory. In the nonabelian theory. where p is the dimension of the D-brane. the ordering of the terms in the nonabelian generalization of the action (28) is not well-defined without further information.W. . In the supersymmetric case. this proposal agrees with calculations which have been performed in string theory [21. Expanding this action in F gives    Z  2 1 1 1 Fµν F νλ Fλσ F σ µ − Fµν F µν −1 − Fµν F µν + + · · · . it is not possible to separate these correction terms from the ordering ambiguity just mentioned since [Dµ Dν ]Fλσ ∼ Fµν Fλσ . which for the U(N) theory may give a form of the nonabelian Born–Infeld theory. see [31–33]). infinite flat D-branes are stable BPS objects whose fluctuations are described in flat space by a Born–Infeld action of the form [21.26] for the F 4 terms in (29). A particularly interesting set of corrections are those which appear for constant but noncommuting field strengths F . and then taking a symmetrized trace in which all possible orderings of the components Fµν are given an equal weighting. This prescription amounts to treating all commutators of F ’s as corrections of higher order in α 0 . Further directions We conclude with a brief discussion of several further issues. the Born–Infeld action should be extended to a nonabelian Born–Infeld (NBI) action. and for the terms linearly coupling the NBI action to supergravity background fields [27–29]. It was proposed by Tseytlin in [23] that the nonabelian action at leading order in α 0 should be given by first expanding in F as in the abelian theory.1. Taylor / Nuclear Physics B 585 (2000) 171–192 183 3. just as the leading U(1) Yang–Mills theory is extended to a nonabelian U(N) Yang–Mills theory.2 we discuss the extension of this approach to the supersymmetric theory and a class of configurations in which a direct connection can be made between descriptions of tachyon condensation in string theory and in D-brane world-volume gauge theories.

since the cubic vertices between an arbitrary set of 3 fields must be used in the calculation if we include all diagrams with fields below a fixed level.35]. It would be very interesting to see if the perturbative calculation of [36] can be correctly reproduced from open bosonic string field theory. there is no reason why the zero momentum part of the action must preserve gauge invariance. If this approach can be extended to the supersymmetric theory. If the action is indeed to take the form of the nonabelian Born–Infeld action. According to [21]. F 4 [21] and F 5 [37]. this is a somewhat more challenging calculation than the determination of the quartic terms performed in this paper. In the supersymmetric nonabelian Born–Infeld theory. If these terms can indeed be calculated in the bosonic theory it may help us to understand the structure of higher-derivative corrections in this theory. the next interesting question comes in at order A12 . If it were possible to use level truncation to convincingly demonstrate the structure of these terms. While only four topological types of diagrams contribute to these terms. In particular. there are corrections containing commutators of F ’s at orders F 3 [36]. F λσ . in the bosonic theory we should find in addition to these terms a correction of the form   Fµν Fλσ F µν . An interesting application of the level truncation method in string field theory would be to extend the analysis of this paper past the quartic terms in the gauge field and to derive the higher order terms in the effective action of the nonabelian gauge field on a bosonic or supersymmetric D-brane. A]2)2 /4) of the Born–Infeld action (29). The next interesting question is whether the terms of order A8 have the characteristic structure Tr([A. there are no corrections to (29) up to order F 5 . we expect that the terms at order A6 will all vanish in the supersymmetric theory. Whether a calculation at this level will be sufficient to fix the structure of the higher order terms in the action is an interesting open problem for future research. Taylor / Nuclear Physics B 585 (2000) 171–192 Such terms are necessary in order to have agreement between the Born–Infeld and string theory descriptions of certain brane configurations [34. F λµ . on the other hand. One subtlety which may complicate the interpretation of the higher order terms in the zero-momentum vector field action is the issue of gauge invariance. While these calculations would be fairly complicated. A]4 − ([A.184 W. including all fields at least up to level 8 or 10. Assuming that all these terms can be calculated and agree with the predictions of perturbative calculations. At order A10 we again expect a nonzero result containing commutator terms [37]. Since. however. they should in principle be quite tractable with the aid of automated symbolic manipulation tools. it might give valuable insight into the structure of the full supersymmetric nonabelian Born–Infeld theory. since the kinetic terms in the vector field action are not gauge invariant. it would represent a new piece of information about the nonabelian Born–Infeld action. corresponding to order F 6 in the nonabelian Born–Infeld action. In the bosonic theory. as discussed above the gauge invariance of the zero-momentum sector is T-dual to translation . while in the bosonic theory we should find a term of the form [36]   Fµν Fνλ .

we have a fairly good understanding of the nature of the ground state: a D-brane and anti D-brane of the same dimension should annihilate into the ground state of the theory. As mentioned in the previous subsection. we devote the last part of this paper to sketching part of this connection.2. In these cases. additional charges are added to the configuration representing lower-dimensional D-branes [40] which may persist after the . Taylor / Nuclear Physics B 585 (2000) 171–192 185 invariance. There are many other scenarios in which similar questions arise. In the low-energy limit. and if such a vacuum exists it is of interest to understand the dynamics of fluctuations around that true vacuum for the theory. and the simplification in diagrams may not make up for the added complications of including momentum. is that of understanding the mechanism of tachyon condensation. Aν ] for a Born–Infeld action. A parallel brane and antibrane of the same dimension which come close enough together develop a tachyonic instability [38]. which has been the subject of much study since the early days of the subject. 3. Only further calculations or a better understanding of the role of the Feynman– Siegel gauge in the effective vector field theory will resolve this question completely. this reduces to super Yang–Mills theory in flat space. An interesting aspect of some brane–antibrane systems is that they can be described in field theory as well as in string theory. it is still necessary to include a fairly general class of cubic vertices at this order. to see whether either of these approaches can give new information about the nonabelian Born–Infeld action. the mechanism of tachyon condensation can be seen explicitly in the field theory. Tachyon condensation A fundamental question about the open bosonic string field theory. It would be very interesting to make a more explicit connection between the string field theory and field theory descriptions of these configurations.W. so that the A12 term of interest. including in particular the tachyonic closed bosonic string theory and brane–antibrane systems in supersymmetric string theories. It is also more likely that the momentum-dependent terms of the F 6 part of the action will be difficult to interpret due to the issues of gauge-dependence mentioned above. it is believed that the resulting vacuum is the true vacuum of the type II string theory. It is also worth pointing out in this context that it may be simpler to approach the Born–Infeld action by including momentum dependence. we are fairly confident that all the terms in the action which do not have momentum dependence should have the appropriate dependence on [Aµ . In the case of brane–antibrane systems in type II string theory. representing a 12-string interaction. It is believed [39] that the brane and antibrane should annihilate in a process which may be understood as the tachyon and other string fields developing nonzero expectation values. the world-volume dynamics of a D-brane in type II superstring theory is described by a supersymmetric Born–Infeld type action. If the brane and antibrane are both free of additional structure. leading to a vacuum which is nonperturbative in the original open string field theory. While this may indeed be an easier approach. It is important to understand whether there is a stable vacuum in the theory which can be found by giving the fields expectation values. By placing fluxes on one or both of the D-branes. however. It will be quite interesting. can instead be determined by calculating a 6-string interaction with 6 powers of momentum of the form (∂A)6 .

A parallel brane and antibrane can be interpreted as two branes with opposite orientation. The spectrum of excitations around this background in the Yang– . A2 (x1 . it is possible in certain circumstances to describe the brane– antibrane annihilation process completely in the language of super Yang–Mills theory.42] that K-theory is the natural language for describing the topological quantum numbers associated with configurations of multiple D-branes. x2 ) = e2πi(x2/L2 )τ3 A1 (x1 . this configuration goes to a D2-brane and an anti D2-brane each with a unit of D0-brane charge.186 W. We place +1 unit of magnetic flux on one D-brane and −1 unit of magnetic flux on the other. x2 )e−2πi(x2 /L2 )τ3 +  2π L2  τ3 . (30) A02 = L1 L2 This corresponds to placing a unit of flux on the first D2-brane and a unit of anti flux on the second D2-brane. the two D2 ± D0 branes become a pair of intersecting diagonally wrapped D1-branes. Taylor / Nuclear Physics B 585 (2000) 171–192 initial brane and antibrane have annihilated. This corresponds to placing a D0-brane and an anti-D0-brane on the two D2-branes. 2π x 1 τ3 . L2 . This picture forms the basis for Witten’s suggestion [41. the spectrum contains a tachyon and an infinite tower of massive states with level spacing proportional to the angle θ between the D1branes. After the addition of further brane charges encoded in fluxes on the brane and antibrane. A particularly interesting aspect of this configuration is that the tachyonic instability as well as the mechanism of tachyon condensation can be completely understood within the framework of gauge theory. The spectrum of string excitations around this background was found in the intersecting D1-brane language by Berkooz. A1 (x1 . x2 ). and without the addition of further charges or fluxes cannot be given a simple description in terms of super Yang–Mills theory. x2 ) = e2πi(x2/L2 )τ3 A2 (x1 . The angle θ is related to the magnetic flux density F on the branes in the D2-brane picture through tan(θ/2) = 2πα 0 F. Although these boundary conditions look nontrivial. We now consider the gauge field background A01 = 0. Douglas and Leigh [43]. x2 ). A2 (x1 + L1 . x2 + L2 ) = A1 (x1 . they are equivalent to trivial boundary conditions through a gauge transformation. however. x2 + L2 ) = A2 (x1 . When the branes are coincident in the directions transverse to the torus. we can define a U(2) gauge field with boundary conditions A1 (x1 + L1 . Let us wrap two D2-branes around the torus with the same orientation. x2 )e−2πi(x2 /L2 )τ3 . Under a single T-duality. This is an example of the type of system mentioned above with a tachyonic instability leading to a vacuum with residual D-branes. Under a pair of T-dualities. Consider compactifying type IIA string theory on a 2-torus with sides of length L1 . A configuration of this type was discussed in [34]. In our original system with two D2-branes of the same orientation. For concreteness we will frame the following discussion in terms of the example discussed there.

At higher order. we should see the structure of the nonabelian Born–Infeld theory which allows the initial unstable but abelian vacuum to decay into the final stable vacuum. however.) This spectrum indeed contains a tachyon. Narain and Sarmadi [46]. the entire process by which the tachyonic mode condenses and the entire system of modes acquire expectation values. . (Van Baal worked on T 4 . The effective action of the set of string fields which correspond to the Yang– Mills fluctuations should then at quartic order precisely reproduce the Yang–Mills action in terms of the diagonal and theta function fluctuations on the D2-brane world-volume. The gap between initial and final energies in the Born–Infeld theory is precisely the difference in energies between a pair of bound D2 ± D0 brane systems and a pair of D2-branes. It should in principle be possible to go from the known spectrum of excitations on the string side to an open string field theory. Sloan Foundation and in part by the DOE through contract #DE-FC02-94ER40818. In the small F limit the Yang–Mills and string theory spectra coincide. in [34] it is shown that under certain circumstances the symmetrized trace formulation of nonabelian Born–Infeld theory produces combinatorial relations which transform F into θ ∼ tan−1 F . but it may be that the parallel presented here with tachyonic configurations in gauge theory will be a helpful clue to resolving some of these difficulties. In this configuration. but we can reduce to the T 2 case by setting the curvature F34 to vanish. The infinite tower of theta function modes in the gauge theory picture precisely corresponds to the infinite tower of string modes stretching between the branes in the D1-brane picture. as mentioned above there are technical challenges remaining to finding a consistent calculable formulation of open or closed superstring field theory. Some progress in understanding tachyon condensation in this system from the string theory side was made by Gava. The gauge field fluctuations around the background (30) can be written in terms of theta functions on the torus. comments and suggestions. leading to a nonperturbative vacuum. and the theta functions for all the other modes are given by Troost in [45]. Taylor / Nuclear Physics B 585 (2000) 171–192 187 Mills theory was originally studied by Van Baal [44]. discussions. the energy gap is on the same order as the energy differential between the unstable and stable brane configurations. as discussed in the preceding section. This work was supported in part by the A.W. We propose that this system may be an excellent model with which learn more about the mechanism of tachyon condensation in string field theory.P. So far. Both the initial and final configurations have diagonal gauge fields. particularly regarding the bosonic Born–Infeld theory. can be seen in the language of Yang–Mills theory. The tachyonic modes are given explicitly in [34]. so that the energies can be computed using abelian Born–Infeld. who showed that when the tachyon condenses to the minimum of the potential arising when quartic terms computed from string theory are included. this configuration is not completely understood in the language of string field theory. Acknowledgements I would like to thank Barton Zwiebach for interesting me in this subject and for numerous helpful explanations. as well as an infinite series of massive modes with level spacing proportional to F . Thanks also to Leonardo Rastelli for helpful conversations. Indeed.

computed from Eqs. (8) n m 11 Nnm 12 Nnm 13 Nnm 1 1 −5/27 1 2 2 1 0 0 16/27 √ 32 √3/243 −32 3/243 16/27 √ −32√ 3/243 32 3/243 1 2 3 3 2 1 32/729 13/486 32/729 1 2 3 4 4 3 2 1 0 0 0 0 −16/729 −64/243 −16/729 √ −64 √3/2187 −160√ 3/2187 160√ 3/2187 64 3/2187 −16/729 −64/243 −16/729 √ 64 √3/2187 160 √3/2187 −160√ 3/2187 −64 3/2187 1 2 3 4 5 5 4 3 2 1 −416/19683 −256/19683 −893/59049 −256/19683 −416/19683 1 2 3 4 5 6 6 5 4 3 2 1 0 0 0 0 0 0 208/19683 128/19683 10288/59049 128/19683 208/19683 √ 800 √3/59049 352 √3/19683 3136 √3/59049 −3136√ 3/59049 −352 √3/19683 −800 3/59049 208/19683 128/19683 10288/59049 128/19683 208/19683 √ −800 √3/59049 −352 √3/19683 −3136√ 3/59049 3136√ 3/59049 352 √3/19683 800 3/59049 1 2 3 4 5 6 7 7 6 5 4 3 2 1 2272/177147 1408/177147 1504/177147 5125/708588 1504/177147 1408/177147 2272/177147 1 2 3 4 5 6 7 8 8 7 6 5 4 3 2 1 0 0 0 0 0 0 0 0 −1136/177147 −704/177147 −752/177147 −22784/177147 −752/177147 −704/177147 −1136/177147 √ −38272 √3/4782969 −41312 √ 3/4782969 −203296 √3/14348907 −195136√ 3/4782969 195136√ 3/4782969 203296 √3/14348907 41312 √3/4782969 38272 3/4782969 −1136/177147 −704/177147 −752/177147 −22784/177147 −752/177147 −704/177147 −1136/177147 √ 38272 √3/4782969 41312 √ 3/4782969 203296 √3/14348907 195136 √3/4782969 −195136√ 3/4782969 −203296 √3/14348907 −41312 √3/4782969 −38272 3/4782969 .188 W. X rs Appendix A. Tables of coefficients Nnm nm Table 3 These are the coefficientsa appearing in the Witten vertex for the bosonic string (6). (7). Taylor / Nuclear Physics B 585 (2000) 171–192 rs .

. nm nm so we only give the coefficients for r = 1. Taylor / Nuclear Physics B 585 (2000) 171–192 189 Table 3 — continued n m 11 Xnm 12 Xnm 13 Xnm 1 1 11/27 1 2 2 1 0 0 8/27 √ 40 √3/243 −80 3/243 8/27 √ −40√ 3/243 80 3/243 1 2 3 3 2 1 −80/729 −19/243 −80/243 1 2 3 4 4 3 2 1 0 0 0 0 40/729 −112/243 40/243 √ −104 √3/2187 −304√ 3/2187 152√ 3/729 416 3/2187 40/729 −112/243 40/243 √ 104 √3/2187 304 √ 3/2187 −152√ 3/729 −416 3/2187 1 2 3 4 5 5 4 3 2 1 1136/19683 800/19683 2099/19683 1600/19683 5680/19683 1 2 3 4 5 6 6 5 4 3 2 1 0 0 0 0 0 0 −568/19683 −400/19683 8792/19683 −800/19683 −2840/19683 √ 1528√ 3/59049 688 √3/19683 3320 √ 3/19683 −13280√ 3/59049 −1720 √3/19683 −3056 3/19683 −568/19683 −400/19683 8792/19683 −800/19683 −2840/19683 √ −1528√ 3/59049 −688 √3/19683 −3320√ 3/19683 13280√ 3/59049 1720 √3/19683 3056 3/19683 1 2 3 4 5 6 7 7 6 5 4 3 2 1 −6640/177147 −4640/177147 −3568/59049 −8251/177147 −17840/177147 −4640/59049 −46480/177147 1 2 3 4 5 6 7 8 8 7 6 5 4 3 2 1 0 0 0 0 0 0 0 0 3320/177147 2320/177147 1784/59049 −84448/177147 8920/177147 2320/59049 23240/177147 √ −82040 √ 3/4782969 −86288 √3/4782969 −239720 √ 3/4782969 −763424√ 3/4782969 954280 √ 3/4782969 479440 √ 3/4782969 302008 √ 3/4782969 656320 3/4782969 3320/177147 2320/177147 1784/59049 −84448/177147 8920/177147 2320/59049 23240/177147 √ 82040 √3/4782969 86288 √3/4782969 239720 √ 3/4782969 763424 √3/4782969 −954280 √ 3/4782969 −479440 √ 3/4782969 −302008 √ 3/4782969 −656320 3/4782969 a The coefficients N rs .W. X rs have a cyclic symmetry under r → (r mod 3) + 1. s → (s mod 3) + 1.

0084 ≈ −0.0145 ≈ 0.1024 ≈ 0.5549 3 ≈ 0. many fields can be contracted in multiple ways with a pair of scalar or vector fields.0000 ≈ 0.0225 ≈ 0.2623 ≈ −0.0011 ≈ 0. Note that. Table 4 γˆ State γˆ1 |0i b−1 c−1 |0i µ ν |0i α−1 α−1 b−1 c−3 |0i b−2 c−2 |0i b−3 c−1 |0i µ ν |0i α−3 α−1 µ ν α−2 α−2 |0i µ ν b−1 c−1 |0i α−1 α−1 µ ν λ α ρ |0i α−1 α−1 α−1 −1 b−1 c−5 |0i 112 2·34 2 − 5 ·13 2·34 27 ·52 310 192 2·39 27 ·52 310 10 − 2 10·13 3 3 − 13 9 2·3 52 ·112 ·13 2·311 4 − 53 ·13 2 ·310 ≈ 0.0030 ≈ 0.0006 ≈ −0.0307 ≈ −0.5802 0 ≈ 0.0188 ≈ −0.0149 ≈ −0. Taylor / Nuclear Physics B 585 (2000) 171–192 Appendix B.2254 ≈ −0.0317 ≈ −0.0004 ν |0i α−4 α−2 µ ≈ −0.9126 0 15 − 210 ≈ −0.0523 ≈ −0.0053 0 b−2 b−1 c−2 c−1 |0i ≈ −0.0016 ≈ −0.0032 0 ≈ 0.0145 ≈ 0.0172 ≈ 0.0074 ≈ 0.0017 ≈ 0.0056 ≈ 0.0290 ≈ −0.0036 0 b−4 c−2 |0i ≈ 0.0092 ≈ 0.0066 ≈ 0.0025 ≈ 0. The contribution calculated includes interactions from all such contractions.0030 ≈ 0.0066 ≈ 0. like Bµν .0150 27 34 27 ·112 310 7 − 2 ·421 310 215 ·52 316 27 ·192 315 215 ·52 316 16 2 − 2 16·7 3 7 − 2 ·7·3739 17 3 27 ·112 ·421 317 7 ·52 ·17 2 − 312 γˆ2 ≈ 1.0746 ≈ −0.0015 ≈ −0.0162 ≈ 0.0019 ν |0i α−3 α−3 µ ≈ −0.0190 0 ≈ 0.7469 ≈ −2.0952 0 b−2 c−4 |0i ≈ 0.0259 ≈ 0.0102 ≈ 0.190 W.0009 ≈ −0.0040 ≈ 0.0558 ≈ 0.0053 0 16 − 216 3 21 − 217 3 215 ·112 317 15 3 2 − 16·5 3 ≈ −0.0542 ≈ 0.0542 ≈ −0.0026 ≈ 0.0241 ≈ −0.1110 ≈ −0.0190 ≈ −0.0040 µ ν b c |0i α−1 α−1 −1 −3 µ ν α−1 α−1 b−2 c−2 |0i µ ν α−1 α−1 b−3 c−1 |0i µ ν α−2 α−1 b−1 c−2 |0i µ ν α−2 α−1 b−2 c−1 |0i µ ν α−3 α−1 b−1 c−1 |0i µ ν α−2 α−2 b−1 c−1 |0i µ ν λ α ρ |0i α−3 α−1 α−1 −1 µ ν λ α ρ |0i α−2 α−2 α−1 −1 µ ν λ α ρ b c |0i α−1 α−1 α−1 −1 −1 −1 µ ν λ α ρ α σ α τ |0i α−1 α−1 α−1 −1 −1 −1 0 0 0 0 0 0 0 ≈ 0.0505 ≈ −0.0062 ≈ 0.0010 0 b−5 c−1 |0i ≈ 0.0091 ≈ −0.0150 ≈ 0.0007 ≈ 0.0002 ≈ 0.0070 ≈ −0.0010 0 b−3 c−3 |0i ≈ 0.0003 ν |0i α−5 α−1 µ ≈ −0.0102 ≈ 0.0095 ≈ −0.0105 ≈ −0.0157 .0173 ≈ −0.0003 ≈ −0.0085 ≈ 0. Some details of the calculations up to level 6 Table 4 describes the contributions of each state to the quartic terms in the effective tachyon and gauge field potential.

[17] E. B 276 (1986) 291. B 181 (1986) 255. [7] E. Schwimmer. [16] D. Non-commutative geometry and string field theory. A. hep-th/9701125. Operator formulation of interacting string field theory (III). B 317 (1989) 411–463. Yost. Phys. Kostelecky. Lorentz invariance. S. Quartic interactions in superstring field theory. Lett. On non-Abelian generalisation of Born–Infeld action in string theory. 2. [23] A. A. Brown preprint HET-672.J. A. [10] C. Tseytlin. Jevicki. Lett. Kostelecky. Tseytlin. Nucl. Nucl. B 336 (1990) 263–296. [25] I.J. Samuel. Nucl. Meng. [6] A. Witten. B 179 (1986) 57. R. B 450 (1995) 90. B 426 (1998) 297–305. hep-th/9911116. Gross.A. hep-th/9705120. Kabat. Thorn. Phys. The static tachyon potential in the open bosonic string. Jevicki. Nucl. Witten. B 287 (1987) 225. Nucl. Preitschopf. [11] C. B 337 (1990) 363. String field theory on the conformal plane (I). B. Wendt. Samuel. Zwiebach. [4] K. Nucl. Phys. A. B 505 (1997) 569. W. Nucl. B 268 (1986) 253. Phys. Closed string field theory: quantum action and the B–V master equation. Linearized supergravity from Matrix theory. G. Taylor. Peskin. . hep-th/9707051. Super-Poincaré invariant superstring field theory. hep-th/9605088. Phys. The tachyon potential in open Neveu–Schwarz string field theory.A. Expectation values. [13] N. Taylor / Nuclear Physics B 585 (2000) 171–192 191 References [1] E. Thorn. Gross.G. [21] A. Lett. (Erratum). Gross. Phys. Nucl. Zwiebach. Zwiebach.J. S.A. Nucl. B 277 (1986) 1.. hep-th/ 0001084. Phys. Nucl. M. Phys. [8] B. [26] D.B. [5] A.A. Nucl. Phys. Tseytlin. Vector field effective action in the open superstring theory. A 4 (1989) 2767. Leclair. Universality of the tachyon potential. B. Phys. Interacting field theory of open superstrings. Phys. Berkovits. [2] V. Operator formulation of interacting string field theory (I). Nucl. Chepelev. Tensor constructions of open string theories 1. Phys. Phys. hep-th/9503099. On a nonperturbative vacuum for the open bosonic string. B 293 (1988) 29. Phys. Phys.A. [19] A. [24] D. The vertex function in Witten’s formulation of string field theory.A. B 390 (1993) 33–152. B 501 (1997) 41. [9] D. Phys.. Samuel. Sen. Tachyon condensation in string field theory. Mod. Spontaneous symmetry breaking in the standard dual string model. Gaberdiel. B 410 (1997) 151. [12] N. B 276 (1986) 391. S. Sen. Phys. Superstring modification of Einstein equations. Jevicki. E. C. Phys. (II).R. B 207 (1988) 169. B 381 (1996) 89. [18] S. hep-th/9912249. Phys. hep-th/9705038. Berkovits.A.E. and CTP in the open bosonic string. [20] V. Lett. Lett. 1988. Nucl. Interactions of type IIB D-branes from D-instanton matrix model. [3] V. [22] R.R. Phys. Nucl. Bardakci. Cremmer. hep-th/9712185. B 133 (1978) 297. B 314 (1989) 209. Superstring field theory.W. Scattering amplitudes and contact interactions in Witten’s superstring field theory. Phys. Witten. C. Preitschopf. A. Nucl. The physical and ghost vertices in Witten’s string field theory.R. B 283 (1987) 1. Kostelecky. Phys. [15] M. Nucl. B 291 (1987) 876. Dirac–Born–Infeld action from Dirichlet sigma model. Phys. Potting. Lett. Bogojevic. Nucl. C. Lett. [14] A. Phys. Leigh.

2 (1999) 1373. M. B 311 (1988) 205. supersymmetry and string theory. Gava. [40] M. van Baal. hep-th/9909154. Myers. J. [31] O. Phys.A. Brane–antibrane forces. Dual models for nonhadrons. Kitazawa. Phys. Phys. Adv. [32] K. Susskind. Branes intersecting at angles. Hashimoto. [42] P. Horava. hep-th/9909187. [41] E. hep-th/9909027. Hashimoto. . Narain. [45] J. JHEP 9808 (1998) 012. Multiple D0-branes in weakly curved backgrounds. Douglas. K-theory. Lett. Corrections to D-brane action and generalized boundary state. Sen. Taylor / Nuclear Physics B 585 (2000) 171–192 [27] W. Taylor. Born–Infeld action. hep-th/9810188. hep-th/9909132. Witten. Constant field strengths on T 2n .and (p + 2)-branes. [44] P. hep-th/9512077. JHEP 9812 (1998) 019. Phys. B 81 (1974) 118. Commun. Sarmadi.R. [29] R. hep-th/ 9904095.A. [34] A. W.C. [35] P. Tseytlin. B 289 (1987) 599. hep-th/9812135. Effective Lagrangian for open superstring from five point function. M. [36] J. [33] Y. Dielectric-branes. Douglas. K. Schwarz. Van Raamsdonk. [30] A. Phys. A.192 W. M.H. [38] T. Troost. R. Fluctuation spectra of tilted and intersecting D-branes from the Born– Infeld action. Type IIA D-branes. On the bound states of p. Berkooz. hep-th/9511194. Taylor. Math. Nucl. Banks. hep-th/9703217. Theor. Leigh. Derivative corrections to Dirac–Born–Infeld Lagrangian and noncommutative gauge theory. Multiple Dp-branes in weak background fields. 94 (1984) 397–419. p. On the nonabelian Born–Infeld action. D-branes and K-theory. Partition function representation for the open superstring effective action: cancellation of Mobius infinities and derivative corrections to Born–Infeld Lagrangian. Math. Scherk.H. Taylor. [39] A.R. B 503 (1997) 193–219. Nucl. SU(N ) Yang–Mills solutions with constant field strength on T 4 . Van Raamsdonk. in: Cargese 97: Strings. L. Phys. Tseytlin.S. hepth/9704006.D. Branes within branes. 267. hepth/9910052. Andreev. Nucl. Okawa. Tachyon condensation on the brane antibrane system. [43] M. M. hep-th/9908105. hep-th/9910053. hepth/9805170. [28] W. hep-th/9606139. Lett. and matrix theory. [37] Y. B 480 (1996) 265. Bain. Phys. [46] E. Branes and Dualities.

unine. PII: S 0 5 5 0 .2]. CH-2000 Neuchâtel. with riemannian geometry being the appropriate mathematical framework. All rights reserved. The field theory limit is also considered in some examples at two loops and reproduces exactly the results of a noncommutative scalar field theory.6 . Sezione di Torino. rodolfo. For higher loops. Università di Torino. The idea of noncommutative space–time is not new. Therefore we can easily sew these vertices together and write the general expression for the multiloop N-reggeon vertex. This has been E-mail addresses: chong-sun. University of Neuchâtel. This kind of deformation is different from that of general relativity. the D-brane worldvolume becomes noncommutative. See [3–5] for a comprehensive introduction to noncommutative geometry. which contains any loop string amplitude.ch (C. it is necessary to use the so-called three-reggeon vertex. Rodolfo Russo a . It was realized recently that a constant B-field in string theory can also deform (D-brane worldvolume) space–time [1. Switzerland b Dipartimento di Fisica Teorica. Sciuto). The openstring Green function on the annulus is obtained from the one-loop scattering amplitude among N tachyons. unlike the early approaches.nl/locate/npe Multiloop string amplitudes with B-field and noncommutative QFT Chong-Sun Chu a .chu@iph. in presence of the B-field.V. 1. Via P. Stefano Sciuto b a Institute of Physics.unine. We find that the modifications to the three. INFN.ch (R. sciuto@to.V. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.-S. However. Chu).  2000 Elsevier Science B.it (S.infn. Russo). I-10125 Torino. which describes the emission from the open-string of another string and not simply of a tachyon.3 2 1 3 ( 0 0 ) 0 0 3 8 8 . Italy Received 9 May 2000. the current developments of string theory display noncommutativity as a result of the underlying string or M-theory and do not require any further input. Giuria 1. In this case.Nuclear Physics B 585 (2000) 193–218 www. All rights reserved. accepted 20 June 2000 Abstract The multiloop amplitudes for the bosonic string in presence of a constant B-field are built by using the basic commutation relations for the open-string zero modes and oscillators.(and multi-) reggeon vertex due to the B-field only affect the zero modes and can be written in a simple and elegant way.russo@iph.elsevier. Introduction It was Einstein’s remarkable idea that space–time is dynamical and can be deformed by energy–momentum tensor.

In fact. Another interesting direction leading by Alekseev. in all these works the 1-loop open-string Green function is obtained from the closed-string Green function by restriction to the boundary. Recknagel and Schomerus [57] is the study of fuzzy physics for D-branes in WZW models [58–60]. As we will review below. also in presence of B-field. The basic building blocks in this formalism are the BRST invariant open-string propagator and three-reggeon vertex.9). it can be written down in an elegant and compact form at all loops. Therefore the results of [9.15] for the charged string case. In [9. in [13] for the M-theory case and has also been extended in [14. One advantage of this approach is that the commutation relations for the modes of the string expansion are explicitly determined. we will find that. This observation allows us to construct the open-string vertex operators with B-field and determine the one-loop open-string amplitudes directly.194 C. Different directions of investigation have been undertaken since then. called N -reggeon vertex. at field-theory level. This approach is indirect and there is no convincing reason that the closed-string Green function should give rise to the open-string Green function in this manner. We show that a similar pattern is present also at string level.e. Chu et al.10] the quantization of an open-string in the presence of a B-field was carried out in the hamiltonian formulation.10] are particularly suitable for doing calculations in the operator formalism. / Nuclear Physics B 585 (2000) 193–218 clarified in [6–12] for the string theory case. One can uniformly construct higher-loop string amplitudes by putting together (i. However. One of the aims of the present paper is to settle this issue using only open-string theory without resorting to closed-string or comparison with fieldtheory calculations. The paper is organized as follows. we will build the generating functional for any N point amplitude. where these commutation relations are essential. There has also been a lot of interest recently [61–65] in exploring the relationships between noncommutative field theory and string theory. In Section 2. It is then straightforward to extract the one-loop open-string Green function. the interaction part can be summarized by the three-reggeon vertex. turning on a B-field has the effect of making the endpoints of open strings noncommutative. it is easy to see that the quadratic part of the action is unaffected by the noncommutative parameter and only the interaction terms depend on it. For higher loops. One direction is the study of the perturbative aspects [16–56] of noncommutative field theory. (2.5). These studies have been partially motivated by the suggestion of [18] on the stringy origin of the UV/IR mixing. However the power of the operator formalism remains.-S. In Section 3. In this case. with the important discovery of UV/IR mixing [17]. sewing) the basic building blocks. with the noncommutativity of the string coordinates obtained as a derived concept. By sewing a suitable number of three-reggeon vertices by means of propagators. we construct the open-string vertex operators with B-field and use them to compute the open-string one-loop amplitudes and extract the one-loop open-string Green function. the modification of this vertex is again determined by the noncommutativity of the zero modes. Intuitively it can be expected that turning on a B-field does not modify the openstring propagator. In terms of the commutation relations of the string expansion modes. the approach of vertex operators becomes less useful.. we explain how B-field affects . it is a simple modification to the zero-mode commutation relation (2.

in fact. then. In this background the open-string mode expansion is  µ µ Xµ (τ. 2. σ ) = x0 + 2α 0 p0 τ − p0ν Fν µ σ X e−inτ √  ianµ cos nσ − anν Fν µ sin nσ . this is the basic ingredient for deriving noncommutative Feynman diagrams from string theory [61–65]. As was recently shown.C. it is sufficient to deal with space–time filling D-branes.1. In Section 4. pˆ0 = pˆ0 .4) aˆ m . defined in terms of the hatted operators. aˆ n = mη δm+n. this simple modification of the commutation relations for the zero mode xˆ0 plays a fundamental role in giving the whole θ -dependence of the string amplitude.0 . in terms of these operators. / Nuclear Physics B 585 (2000) 193–218 195 the basic building blocks in the construction of multiloop string amplitudes and obtain the h-loop N -point reggeon vertex with B-field. (2. We also recall that the normal-ordered Virasoro generators. As we will see.2) In fact. 2. We finally conclude with some open problems. (2. we consider the noncommutative 83 field-theory in six dimensions at the two-loop level and show that the field-theory results agree precisely with the field-theory limit of the two-loop string amplitudes we computed with the reggeon vertex. In order to write the commutation relations of these modes. we will need only the open-string coordinates in the longitudinal directions of the D-branes. aˆ nµ = anν (1 − F )ν µ . for the purpose of deriving field-theory diagrams from string amplitudes.10]  µ ν  µ ν  µ ν (2. pˆ0 = iη .10] for an open-string-ending on a D-brane in presence of a constant B-field. In the last line we have introduced the dimensionful quantity θ := 2πα 0 F which directly measures the noncommutativity. xˆ0 = 2πiα F = iθ . Using these vertices. it is convenient to introduce the rescaled quantities µ xˆ0 = x0ν (1 − F )ν µ .  µ ν 0 µν µν (2. directly from this result. we extract the open-string Green function on the annulus. One-loop open-string Green function In this section. (2. . we introduce the vertex operators for open-string states in the presence of a constant B-field. the equations take the standard form. satisfy the standard Virasoro algebra [9] with a central charge unmodified by F . xˆ 0 = aˆ n .3) aˆ n . we can compute string amplitudes among tachyons states at tree and one-loop level.5) xˆ0 .  µ ν  µ ν µν µν xˆ0 . µ pˆ0 = p0ν (1 − F )ν µ .1) + 2α 0 n n6=0 where F = B − dA is the modified Born–Infeld field strength. Open-string vertex operator We first recall the results [9. pˆ0 = 0.-S. Chu et al. Here. except for that of the zero modes [9.

appearing in the expansion of X  µ ν xˆ0 . σ ). Vε (p. µ µ (2. X 0 0 n n6=0 where we have introduced yˆ0 := xˆ0 − 2α 0 π pˆ 0ν Fν µ . Note that (2.5). σ 0 ) = ±iθ . For example.7) bµ (τ.13) is equivalent to say that multiplication for functions defined on the D-brane worldvolume are done in opposite ordering (Section 5 of [9] and Section 6.10] where it was shown that  µν 0   µ bν (τ. for the emission at σ = π . It is straightforward to check that it satisfies the desired property. otherwise. satisfies The zero-mode yˆ0 .13) X 0. X (2. 0) = xˆ µ + 2α 0 pˆ µ τ + i 2α 0 aˆ µ . However. This difference in the commutation relation for xˆ0 and yˆ0 is not so important for tree-level calculations. the gluon-vertex operator is b ip·X dX b e . / Nuclear Physics B 585 (2000) 193–218 The vertex operators for open-string states emitted at the boundary are constructed using bµ evaluated at σ = 0 or σ = π : X X e−inτ √ bµ (τ. τ ) = : eip·X(τ. (2.6) X 0 0 n n n6=0 X (−e−iτ )n √ bµ (τ. 2α 0 p · = exp n n n=1 n=1 (2.196 C. τ ) = e dτ 2 Similarly.-S. τ ) = ε · bµ at equal time is entirely This means that the noncommutativity of the string coordinates X determined by the zero modes. the open-string tachyon vertex operator for emission at σ = 0 is b V (p. one can construct the vertex operators for higher open-string states. both for σ = 0 and σ = π :     1 2 d imτ −i + mp V (p. (2. τ ). σ = σ = 0 or π.10) while. This is in agreement with the result of [9. this difference becomes essential for one-loop diagrams: in fact. (2.9) has an opposite sign compared to (2.8)  µ yˆ0 .12) dτ We want to stress that the commutation relation (2. in this . b (τ. since one can choose to put all interactions at σ = 0. (2.11) Lm . π) = yˆ µ + 2α 0 pˆ µ τ + i 2α 0 aˆ nµ . V (p.9) For instance. π).0) : ! ! ∞ ∞ X X √ √ a ˆ a ˆ 0 −n n einτ eip·xˆ0 ei2α p·pˆ 0 τ exp − 2α 0 p · e−inτ .2 of [12]). we just have to replace xˆ0 by yˆ0 in the above. Chu et al.  (2. yˆ0ν = −iθ µν . yˆ0 = 0.

9) and of the fact that either the zero mode xˆ0 or yˆ0 appears in the vertex.-S. M. Chu et al.  2πα 0 p1 · p2 (τ1 − τ2 ) − p1 θ p2 . while the other particles s ∈ {M + 1. ρM+N ) k L0 . we remark that the vertex operator Vα (p. τ2 )Vα1 (p1 . In Fig. The 1-loop amplitude is given by A(1. . . vertices inserted on the σ = 0 (σ = π ) boundary are represented by a vertical line above (below) the horizontal line. 2. M + N) ∼ Z (2.2. . . moreover. we have chosen the radial ordering k 6 |ρM+N | · · · 6 |ρM+1 | 6 ρM · · · 6 ρ2 6 ρ1 6 1. The θ -dependent piece is a simple consequence of the zero-mode commutation relations (2. where some of the legs are emitted from the border σ = 0 and the others from σ = π . (2. . depending on whether it is on the σ = 0 or σ = π border. τ2 ) = Vα2 (p2 . . For instance.C. τ ) for a string state α satisfies Vα1 (p1 . . ρ1 ) · · · V (pM+N . . . φ = −2πα 0 p1 · p2 (τ1 − τ2 ) + p1 θ p2 . it is easy to work out the N -point amplitude. σ = 0 and σ = π . τ1 )Vα2 (p2 . . (2.15) is crucial. . and then to verify that the result is cyclically invariant: one has simply to use the commutation relation (2. / Nuclear Physics B 585 (2000) 193–218 197 Fig. 1. The difference of sign for the two borders in (2. Open-string Green function Now we compute the 1-loop open-string amplitude for tachyons using the vertex operators constructed above and extract the open-string Green function from it. 1. Cyclicity of the tree-level N -reggeon vertex.16)  dd p tr V (p1 .15) and (τ ) = τ/|τ |. M} := I1 are on the boundary σ = 0.8) and (2. if α1 and α2 are on the same border σ = π. M + N} := I2 are on the boundary σ = π . . Consider the (M + N )-point amplitude depicted in Fig.14) and momentum conservation. . case nonplanar diagrams require to put vertex operators at the two different borders.14) where  if α1 and α2 are on the same border σ = 0. . The external states n ∈ {1. 2.  0. . (2. Finally. τ1 )eiφ . . if α1 and α2 are on different borders. since it ensures that tree-level string amplitudes are invariant under cyclic permutations.5). M + 1. .

(2.198 C.10). Chu et al. / Nuclear Physics B 585 (2000) 193–218 Fig. one has a new term linear both in θ and in p which has its origin in the commutation relations of the zero modes xˆ0 and yˆ0 . One annulus in Schottky representation. Notice that. besides the usual p-dependent factors coming from the propagators. 2.-S. which is given by eip 1 xˆ 0 · · · eip M xˆ 0 eip (M+1) yˆ 0 · · · eip (M+N) yˆ 0 =e −ipˆ 0 θ P s∈I2 i ps − 2 P e r<s r.s∈I2 p r θp s . where V is the vertex introduced in (2.s∈I1 ∪I2 r∈I1 ∪I2 . First we look at the part containing xˆ0 and yˆ0 .s∈I1 p r θp s e i 2 P r<s r.17) Evaluating the trace turns pˆ 0 into the loop momentum p. 1 Globally the zero-mode part becomes   Z X  X pr ln |ρr | − iθ ps dd p exp α 0 p2 ln k + p · 2α 0 −α 0 X r<s r.

.

 r s .

ρs .

p · p ln.

.

in particular. as usual. the expression in the exponent is equal to " ! 2 X X 1 θ pr ln |ρr | − i 0 ps α 0 ln k p + ln k 2α r∈I1 ∪I2 s∈I2  2  2 . . by completing the square. ρr s∈I2 (2.18) The gaussian integration over p is performed.

.

 X X ln |ρs /ρr | θ 1 .

ρs .

− ln.

.

s∈I2 1 The importance of the commutation relations for the zero modes was emphasized.s∈I2 i X X ln |ρr |pr θps .s∈I1 ∪I2 + r∈I1 .20) r∈I1 . ln k (2. in [41]. ln k (2. + 0 + α0 pr · ps pr ps ln k ρ α ln k 2 r r<s r. There one has pˆ 0 = 0 and thus obtains the usual Filk phase. .19) r∈I1 ∪I2 s∈I2 It is easy to see that the last term in the above equation is equal to X i ln |ρr ρs |pr θps . in a tree-level example.

In the nonzero modes part. the extra minus sign b expansion has the effects to change the sign of ρ in front of the aˆ n oscillators in the X(π) in the oscillator sum for the nonplanar vertices. ρ 0 ) = I0NP ηµν + 8α 02 ln k 2α 0 ln k with the θ -independent piece I0 given by s .-S.ρs )pµ ν .s∈I2 where the planar and nonplanar open-string Green function are iθ µν (ρ − ρ 0 ). 4α 0 iθ µν ln |ρρ 0 | (θ 2 )µν 1 µν ± . Chu et al.ρs )pµ ν r<s r.19) contains two new terms arising in nonplanar diagrams. M + N)  Z1  Z dρM+N dk −2π d/2 dρ2 n 2−d (1 − k ) ··· ∼ 2 ln k ρ2 ρM+N k 0 × Y 0 µν e2α GP r ps (ρr . . We will see that the same pattern repeats for the general case of the h-loop reggeon vertex.s∈I2 r∈I1 .s∈I1 Y 0 µν (2. / Nuclear Physics B 585 (2000) 193–218 199 The momentum integral can now be evaluated and the second line of (2.ρs )pµ pν Y 0 µν e2α GP r ps (ρr . . . M + 1. . Finally we can summarize 1-loop amplitudes in the following compact equation: A(1.21) r s e2α GNP (ρr . (2. .C.22) GNP (ρ. . r<s r. . M. . It is worthwhile to remark again that the θ and θ 2 terms above arise as a result of the new linear p-shift in the gaussian integration.

.

r .

∞ .

.

ρ 2 0 Y .

(1 − k n ρ/ρ 0 )(1 − k n ρ 0 /ρ) .

ln ρ/ρ ρ 0 .

.

.

P 0 .

.

23) + ln. (2.

− I0 (ρ. ρ ) = .

+ ln .

.

.

ρ0 2 ln k ρ .

while for nonplanar contractions one has s s ! 2 0| 0| ln |ρ/ρ |ρ| |ρ + ln + I0NP (ρ. ρ 0 ) = 2 ln k |ρ 0 | |ρ| . ρ 0 ) = I0P ηµν − µν n=1 for the planar case. (1 − k n )2 GP (ρ.

.

∞ Y .

(1 + k n |ρ/ρ 0 |)(1 + k n |ρ 0 /ρ|) .

.

.

+ ln ..

.

reproduce the Feynman diagrams of diverse noncommutative field-theory. By using these Green functions. by introducing in (2. This follows directly from our result (2.20). −α 0 ln |ρk | = t2 + t3 + · · · + tk . This provides an independent field theory check of the correctness of the string computation (2.21) the Schwinger parameters ti (2 6 i 6 M + N ).24) n=1 In the last term of the nonplanar Green function in (2. the result is precisely the same and taking → 0 with θ as the corresponding noncommutative 83 field-theory amplitude. ρ < 0. derived by means of the noncommutative Feynman rules. in the limit α 0 → 0. ρ 0 < 0 and negative sign is taken for the opposite case: ρ 0 > 0. For example.22).25) and ti both kept fixed. positive sign is taken when ρ > 0. one can compute different string amplitudes that. (1 − k n )2 (2.21) and of the vertex operators we constructed . (2. α0 = 2πα 0 F 2 6 k 6 M + N.

one can also obtain from (2. / Nuclear Physics B 585 (2000) 193–218 above. We note that. in the tachyon amplitudes (2. by using momentum conservation.19) as a function of the ratio of the ρ’s over the two borders: ! X . to rewrite the last term of (2.21). By contracting some of the propagators [63].200 C. Chu et al.21) other field-theory limits.-S. for example noncommutative 84 . it is also possible. the Green functions are always contracted with the external momenta. Because of this.

.

ρs .

.

X .

.

ρr .

.

i r r0 s s0 ln.

.

p θp − ln.

.

one may shift this piece to the planar diagram and extract a different planar and nonplanar Green function .26). (2. and ρs → −λ/ρs . We also remark that written in the form (2.26) − ln k ρr 0 ρs 0 0 0 r<r r.r 0 ∈I1 s<s s.27) It is easy to generalize the manipulations here and to show that this is also true for the gluon amplitudes as well as the amplitudes for other higher string states.s 0 ∈I2 This shows explicitly that the tachyon amplitude is invariant under the transformation ρs → λρs . (2.p θp . λ > 0.

ρ.

iθ µν iθ µν .

.

0µν 0 ln (ρ − ρ ) ∓ GP = I0P ηµν − .

0 .

This object is the generator of all tree-level scattering amplitudes among arbitrary string states and will provide the first basic ingredient for constructing higher-loop “noncommutative” string diagrams. naturally derived also in the scalar computation. to introduce the reggeon vertex formalism.28) are not equivalent when one wants to deal with external massless states and derive. (2.22) and (2. in order to construct loop amplitudes. Then one has to rewrite the final amplitude in the form of the string master formula for gluons. (2. Reggeon vertex and multiloop-string amplitudes In this first section we construct the tree-level BRST invariant N -reggeon vertex for open bosonic string in presence of a constant B-field. the Feynman diagrams of the noncommutative gauge theory. For instance. 0µν GNP = I0NP ηµν + 3.28) 8α 02 ln k where ‘−’ sign is for the positive border and ‘+’ sign is for the negative border. without the need to refer to any field-theory results..22) and (2. and identify the Green function from it. for our purposes. for instance. these are the same as one obtained from restricting to the boundary of the 1-loop closed-string Green function [61]. by following the same procedure of this section.12) are used instead of (2. Incidentally. one can compute a general 1-loop gluon interaction if the vertex operator (2. as stressed in [63].23) above. one sews together pairs of external legs and sums . The result is indeed given by (2. It is quite clear why it is necessary. As usual. However.10). 4α 0 2α 0 ln k ρ (θ 2 )µν 1 . Indeed the necessity of shifting the result naturally obtained from the closed-string calculations is clear within string theory.

in the vertex operator formalism. As we will briefly review.67] the quantum numbers of the emitted string with a whole Hilbert space.1. q = 3| : exp dz −Xv (ζ + z) ∂z X(z) − cv (ζ + z)b(z) 0 + b (ζ + z)c(z) v )  :. The second fundamental ingredient that is necessary for constructing higher-loop interactions is the BRST invariant open-string propagator. However. also the coordinates of the external string appear directly. From the geometrical point of view. the commutations relations of these new modes are the usual ones (2. In the previous section this has been achieved by simply identifying the first (z = ∞) and the last (z = 0) leg and then taking the trace.C. As is clear by looking at the simplest case (2. while the label q = 3 specifies the ghost number |0. 3.2) Besides the fields of the virtual string denoted by v. / Nuclear Physics B 585 (2000) 193–218 201 over all contributions coming from the exchange of the different string states in the various loops.5). Thus the B-dependent N -reggeon vertex will be the only new building block we need to use for computing the “noncommutative” string amplitudes at loop level.-S. Formally our expression is identical to the one normally used in string calculation in the trivial background B = 0. however. (3. .10). one basically replaces [66.70].0 (ζ ) = h0. In the generalization of these usual vertices. considering also the ghost contribution. all these legs have been represented by vertex operators describing specific string states and thus it is very difficult to sum the contributions coming from the exchange of different states.3)–(2. the introduction of reggeon vertex solves this problem. the 3-reggeon vertex is [68] 2 (I θ V3. these vertices are built by means of the coordinates of the propagating virtual string and their explicit form is determined by the quantum numbers of the emitted states like momentum or polarizations. Tree-level reggeon vertex The starting point is a generalization of the usual vertices describing the emission of string states. It turns out that this ingredient is not modified by the presence of a constant B-field. in fact. Thus. (3. in the 3-reggeon vertex. (3. this propagator is a conformal transformation identifying two circles around the punctures sewn together. while the oscillators of the virtual and the external strings simply commute among them. 0a i ⊗ c−1 c0 c1 |q = 0i.1) where the bra indicates the vacuum of the emitted string for both the oscillators and the zero mode x0 . we need to identify further couples of legs that cannot be fixed at z = 0 and z = ∞. q = 3i = |x0 = 0. Chu et al. in order to repeat the sewing procedure. This approach is no longer adequate for constructing higher-loop amplitudes.1) depends 2 The ghost contribution to the symmetric vertex was given in [69.

The key property of the 3-reggeon vertex is that it can be saturated with a string state α and one recovers the usual vertex operator corresponding to the considered state θ (ζ )|p. Then. we have dropped the hat over the X fields. the phase arises because of the use of the commutation relation (2. in the direction flowing towards the 0 indicates the N -reggeon vertex derived for the usual “commutative” boundary. Chu et al. the (b. ζ ). which have to be used both for the virtual and the external string. in particular. Because of this. α1 | V3. we will consider the noncommutative field-theory limit α 0 → 0.2). but written in terms of the same propagating string v. again.0 = VN. the 3-reggeon vertex gives correctly the tree-level amplitude. (3.3) It is easy to check that when completely saturated with external states. in what follows. In particular.6) where momentum conservation has been used. one obtains the new phase factor eip 1x 0 · · · eip Nx 0 = e− 2 i PN r<s p r θp s .1). Notice that. we will employ the usual physical states having ghost number 1. we obtain the N -reggeon vertex with all legs emitted from the σ = 0 border ! N iX i j θ 0 p θp . F →∞ (3.5) (or (2. The new θ -dependent part comes when one collects together the zero-mode factors. α2 i |p3 .5) with α 0 F fixed so that a noncommutative field-theory is obtained.4) where. one can explicitly evaluate the vacuum expectation value in the Hilbert space v in order to keep only the dependence on the oscillators of the external states. c) system is not affected by the background field F and one recovers the well known results for ghosts. As a result. We remark that all the formula written in this section are general and hold for any value of α 0 .7) VN. / Nuclear Physics B 585 (2000) 193–218 on the value of the B-field through the mode expansion (2. The tree-level N -reggeon vertex can be constructed simply by multiplying N 3-reggeon vertices in different positions ζ . V3. In the next section.-S. (3. Here for convenience. the zero-mode x0 (or y0 if the interaction is at σ = π ) appears only in the expansion of the virtual string and is the only source of the non-trivial dependence on θ . including the phase factor θ v hp1 . αi = Vα (p. (3.0 |p2 .0 .6) and the new commutation relations.0 exp − 2 i<j where pi is the momentum operator of the ith leg. Together with the reggeon vertex (3. α3 iv i ∼ e∓ 2 p 1 θp 2 .0 (3. the upper (lower) sign refers to the emission of the particle α2 from the border σ = 0 (σ = π ) of the propagating string v. in terms of the rescaled coordinates. although it should be understood that all the X’s above contain the oscillators defined in (2.202 C.9)) for the zero modes. thus. we will no longer mention ghosts and focus only on the F -dependent modifications coming from the orbital part. if the external legs of all the original 3-reggeon vertices are emitted from the border σ = 0. Here VN.

/ Nuclear Physics B 585 (2000) 193–218 203 case F = 0 in [71.0 N Y i=1 N . like the 3-reggeon vertex.-S. In particular. This vertex is a bra in the direct product of the N distinct Fock spaces for the external strings. it gives the scattering amplitude when the external legs are saturated with physical states. VN. like the 3-reggeon 0 contains the vacuum of the zero modes vertex.72]. Chu et al.C.

Y x0i = 0.

= Z .

dpi pi .

p(2) → −p(N−1) . for the sake of simplicity. In the next section. there are in general many more redundant variables in the reggeon vertex than in the physical amplitudes. We will see that the momenta of the corresponding legs are to be identified and interpreted as the loop momentum. This is exactly the same kind of modification to the tree-level vertex in a noncommutative field-theory due to the Moyal ∗-product.1). Because of this arbitrariness in the choice of the Vi . (3.s. 3. z0 ). When saturated with external states.9) bn(2) → bn(N−1)†. in order to sew together leg 2 and leg (N − 1). For example. In terms of local coordinates Vi−1 around the puncture zi . We will see that this identification of the modifications in the string vertex and field-theory vertex is the basic reason which guarantees that the usual one–one correspondence between corners of string moduli space and field-theory Feynman diagrams continues to hold in the noncommutative case.. the N -reggeon vertex can be written for general conformal transformation Vi ’s and it is easy to check that the string amplitude is independent of the choice of them. 3 Vi (z) = z + zi .h.10) The next step in the sewing procedure is to identify the punctures of the legs that are going to be sewn together. −1/2) in (z. zj ) is of conformal weight (−1/2. a more convenient choice V (z) = z z + z is usually taken [75].7) to construct the multiloop reggeon vertex. this means we have to identify Vi−1 (z) and Vj−1 (z) so that a complex structure can still be defined. (3. This has the effect i i i of shifting the Green function by the term − 12 ln |zi zj | so that the resulting Green function G(zi . we identify the corresponding Fock spaces: an(2) → an(N−1)†. However. We also note that the only modification to the tree-level N -point vertex is the momentum-dependent phase factor.8) i=1 The momentum operator dependent phase factor in (3.7) is to be interpreted as an operator acting on the r. of (3.2.8) effectively instructs us to integrate over the loop momentum. Off-shell continuation of string results is more convenient using this Green function. . (3. This choice is also adopted in the recent analysis [63].8). (3. h-loop N -reggeon vertex The h-loop N -reggeon vertex in the presence of a constant F -field can be constructed by sewing together legs of a tree-level (N + 2h)-reggeon vertex. it gives a (numerical) phase factor as the pi ’s become the external momenta. In (3. It is therefore useful to make some convenient 3 In taking the field limit. we will sew together legs of the tree-level reggeon vertex (3. cn(2) → −cn(N−1)†. we have chosen.

. As we have already said the ghost system is not modified by the F -field.h = V N.11) dp exp VN. 1 to zi−1 . B.8). N .74]. . . / Nuclear Physics B 585 (2000) 193–218 Fig. Chu et al. zi . B. it is P1 = p1 + p2 .12) (3.16) .204 C. (3. P2 = 0. it is straightforward to show that the h-loop N -reggeon vertex in the presence of F -field is ! h Z h h Y X X 1 µν µ θ 0 I I J I e pµ AI J pν + BI pµ + C .J =1 I =1 e0 contains the ghost contribution and only the nonzero-mode piece of the orbital where V N. JI J is the intersection matrix for the internal loops. (3. h.h part of N -reggeon vertex in absence of background. Vi is a projective transformation which maps ∞. We obtain the following modifications to A. pi . I = 1. which is a function of L0 and L±1 and of the ghosts. (3. In this choice. Fig. Its explicit form can be found in [71. PI is the sum of external momentum entering the I th loop. (3. BI = B 0 I − iθ µν PI ν . the Lovelace choice [73] is useful. . i = 1. For example in Fig. What interests us is that all the dependence in θ is localized in the zero-modes loop momentum integration. zi+1 . 3. 0. we conclude that the string propagator P (x) and the sewing procedure are not modified by the presence of F -field. thus its contribution (both zero and nonzero modes) to the reggeon vertex is unchanged. for constructing higher-loop amplitudes through the sewing procedure.h 2 I =1 I.-S. C: µν µν µ µ AI J = A0 I J − iθ µν JI J . are the loop momenta.13) (3. C for θ = 0 are given in [74]: µν A0 I J = 2α 0 (2πiτI J )δ µν .14) i<j where pI . . A reggeon vertex with external momenta p1 and p2 crossing an internal momentum. Since Ln does not involve the zero modes x0 . The sewing is achieved by using the BRST invariant sewing operator P (x) [72]. Therefore the only new feature for θ 6= 0 is in the zero modes part of the reggeon vertex.7) and recalling (3. In particular. are the external momenta. 3. . . choices on the Vi ’s so as to obtain a simpler expression for the reggeon vertex in order to simplify intermediate calculations and manipulations. 3 shows how to construct a N -reggeon vertex with all legs emitted on the border σ = 0.15) The explicit form of A. . C = C0 − i 2 N X pi θpj . Using (3.

j =1 0 (3.C.20) bJ is the period matrix and E(z. One can easily carry out the integration over z in the above expressions: . Their explicit expressions in term of the Schottky parameters can be found in [74]. w) is the prime form. 2π i=1 0 1X 2 N C0 = − I 205 (3.-S. Chu et al. Vj (y)) 1 X (i) ∂X(j ) (y). ωI is the normalized abelian differential. dz dy ∂X (z) ln + 0 4α Vi (z) − Vj (y) i.19) aJ I 2πiτI J = ωI (3. / Nuclear Physics B 585 (2000) 193–218 µ B0I VZi (z) N I 1 X (i)µ = dz ∂X (z) ωI . I ωI = δI J and (3.17) z0 dz ∂X(i) (z)p0(i) ln Vi0 (z) i=1 0 + I N I   1 X dz dy ∂X(i) (z) ln Vi (z) − Vj (y) ∂X(j ) (y) 0 2α i<j 0 0   I N I E(Vi (z).18) 0 Here Vi (z) is chosen to satisfy Vi−1 (z) = 0 for z = zi .

I X an √ .

1 0 n 0 ∂ f (z).

dz ∂X(z)f (z) = 2α pf (0) + 2α
(3.21)

n!
n=1

0
µ

z=0

and explicitly rewrite B 0 I and C 0 in terms of modes. In particular, one finds here (and in
the following (3.24), (3.26) and (3.27)) terms containing only the zero modes p. In this
case, the arguments of the log’s must be taken in absolute value. This absolute value has
its origin in the matrix element D00 = (1/2) log |γ 0 (0)| of the real infinite-dimensional
representation of the projective group, relevant for the open string (see Ref. [72]).
A couple of remarks are in order. The modification in (3.14) is the universal phase factor
depending on the external momentum. It corresponds to the Filk phase [16] in the fieldtheory limit. The θ -dependent shift in (3.13) corresponds in the field-theory limit to the
shift due to the dipole mechanism [24,25,40]. The string theory interpretation of this shift
in terms of stretched string was recently discussed in [65]. The shift in (3.12) is new and,
among other effects, gives rise to a θ -dependent measure factor for the zero modes of the
orbital degrees of freedom. One obvious advantage of our approach is the clarity of the
modifications due to B-field to the multiloop-string amplitudes: it is summarized in the
zero-mode contributions in (3.22). As a result, apparently unrelated field-theory effects
could find a uniform string explanation.

206

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

Carrying out the loop-momentum integration, we obtain finally:  

1 T −1
1
θ
0
e
exp − B A B + C ,
VN;h = VN;h q
2
det −A
2

(3.22)

where the determinant is taken over the space of Lorentz and loop indices (µI ). When
θ = 0, it is
p 
d/2
det(−A/2) = det(−2πiα 0 τ )
.
(3.23)
It is remarkable that the multiloop-string amplitudes can again be written in terms of
geometrical quantities of the Riemann surface like the first abelian differentials, period
matrix and the prime form, together with the intersection matrix of the loops.
Finally we close this section by writing down explicitly the θ -dependences in the
µν
reggeon vertex for the one-loop case. First, there is no modification to A0 I J and the
modification to C 0 is simply the usual field-theory Filk phase. Thus we concentrate on
µ
the effects from the shift of B 0 I . The one-loop abelian differential is ω = dz/z. With the
convenient choice of Vi (ρ) = ρ + ρi , we have
X
µ
B r µ (ρ) ln(ρ + ρr ),
(3.24)
B0I =
r∈I1 ∪I2

where
B

1
(ρ) =

I
dρ ∂X(r)µ (ρ)

(3.25)

0

is an operator in which the integration in carried out on any function that multiply B on the
P
P
right. Since P = s∈I2 ps = s∈I2 B s for the only loop we have, it is
B 0µ − iθ µν Pν =

X

B r µ (ρ) ln(ρ + ρr ) −

r∈I1 ∪I2

iθ µν X s
Bµ .
2α 0

(3.26)

s∈I2

Substituting into (3.22), we have
1
−1 2
B = 2α 0 ·
2
0
2A
8α ln k

X
r<s
r,s∈I1 ∪I2

B r (ρ)B s (ρ 0 ) ln2 

ρ +ρ 
r

ρ 0 + ρs

!
X  

1
i X r
B (ρ)θ B s (ρ 0 ) ln (ρ + ρr )(ρ 0 + ρs ) + 02
B r (ρ)θ 2 B s (ρ 0 ) .
+ 0
α

r∈I1
s∈I2

r∈I1
s∈I2

(3.27)
The second line contains all the θ -dependence to the one-loop string amplitude. Putting
B r (ρ) → 2α 0 pr ,

(3.28)

we recover the previous expression for the tachyon amplitude and this is a consistency
check of the correctness of our results. However (3.27) allows us to determine the θ dependence to the string amplitudes for gluons as well as any higher mass state. The
advantage of the reggeon-vertex formalism is obvious.

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

207

4. Noncommutative field-theory limit
In this section, we consider the noncommutative 83 theory in 6 dimensions at two-loops.
We show that the 2-loop reggeon vertex we constructed in the previous section reproduces
exactly the field-theory results. One–one correspondence between Feynman diagrams and
corners of string moduli space is established.
4.1. Two-loop noncommutative 83 in six dimensions
Consider 83 interaction in d = 6: 

Z 
1
1 2 2 1
µ
∂µ φ ∂ φ − m φ + g3 φ ∗ φ ∗ φ dd x.
S3 =
2
2
3!

(4.1)

As an illustration, we will consider 2-loop nonplanar diagrams with 2 external legs. The
first type of diagrams have the two external legs attached to two different propagators.
Some of them are shown in Fig. 4.
For all these diagrams, we have (up to overall numerical coefficient):
Z
Z
A = dq dr Dα exp −α1 (q − p)2 − β1 q 2 − α2 (q − p − r)2 

− β2 (q − r)2 − t3 r 2 + ϕ ,
(4.2)
where we have introduced Schwinger parameters and labelled the momenta as in Fig. 5
with
t1 = α1 + β1 ,

t2 = α2 + β2 .

(4.3)

ϕ is the product of the phase factors arising at the various junctions A, B, C, D due to the
Moyal product and
Z∞

Z
Dα =

Z∞
dt3

0

Zt2
dt2

0

Z∞
dα2

0

Zt1
dt1

0

dα1 e−(t1 +t2 +t3 )m

2

0

Fig. 4. 2-loop diagrams with external legs attached to different propagators.

(4.4)

208

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

Fig. 5. Momentum flow and Schwinger parameterization.

is the integration over the Schwinger parameters. For convenience, we have grouped the
mass dependent factors here. In this section, we will be interested in displaying the precise
θ dependence and thus will ignore overall numerical coefficients.
Let us start with Fig. 4e. The phase factor is
ϕ4e = −ipθ r.

(4.5)

Carrying out the q integral, we obtain
Z
Z

A4e = dr
(t1 + t2 )d/2     

(α1 + α2 )2
1  2
× exp −p2 α1 + α2 −
r ∆ − 2r µ ξµ ,

t1 + t2
t1 + t2

(4.6)

where
∆ := t1 t2 + t2 t3 + t1 t3

and

(4.7)

i
(4.8)
ξµ := (α1 β2 − α2 β1 )pµ + (t1 + t2 )(θp)µ .
2
Completing the square and integrate over momentum r, we obtain
Z
1
(4.9)
A4e = Dα d/2
∆    

(α1 + α2 )2
(α1 β2 − α2 β1 )2
t1 + t2
− p2 α1 + α2 −
(θp)2 .
× exp p2

(t1 + t2 )∆
t1 + t2
4∆
The exponent can be simplified easily and we obtain finally:   

Z
α22 (t1 + t3 ) + α12 (t2 + t3 ) + 2α1 α2 t3
1
2
A4e = Dα d/2 exp −p α1 + α2 −

∆ 

t1 + t2
(θp)2 .

(4.10)
4∆
It is in this form that the field-theory results may be compared with string theory result
most conveniently.
For the diagram of Fig. 4b, we have the phase factor

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

ϕ4b = i(p + r)θ q.

209

(4.11)

Carrying out the r integral, we obtain    

Z
Z 

α22
1 µ

2
ν
µ

exp
−p
+
α


q

2q
ξ
q
α
A4b = dq
1
2
θ µν
µ , (4.12)
l
l
l d/2
where
l := t2 + t3 ,

(4.13)
i
(4.14)
ξµ := [α1 l + α2 t3 ]pµ + (α2 − l)(θp)µ and
2
2
θµν
.
(4.15)
(∆θ )µν := ηµν ∆ −
4
For these calculations it is convenient to choose θ µν to be a block diagonal matrix; in this
case ∆θ is a diagonal matrix in the Lorentz indices and can be easily inverted. Completing
the square and integrate over momentum q, we obtain   

Z
α22
1
1
2
−1 µν
.
(4.16)
exp ξµ (∆θ ) ξν − p α1 + α2 −
A4b = Dα √
l
l
det ∆θ
The exponent can again be simplified and we obtain finally 
 

Z
(t1 + t3 )α22 + (t2 + t3 )α12 + 2α1 α2 t3
1
exp −p α1 + α2 −
p
A4b = Dα √
∆θ
det ∆θ  

θ2
p .
(4.17)
× exp −(2α2 − l)p
4∆θ
Next we consider the type of diagrams (Fig. 6) with both external legs attached to the
same propagator. The momenta and Schwinger parameters are labelled as in Fig. 7.

Fig. 6. 2-loop diagrams with external legs attached to the same propagator.

Fig. 7. Momentum flow and Schwinger parameterization.

210

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

For these kind of diagrams, we have
Z
Z
A = dq dr Dα exp −t1 q 2 − t2 (q − r)2 − α1 r 2 − α2 (r + p)2 

− α3 r 2 + ϕ ,

(4.18)

where we have introduced
Z∞

Z
Dα =

Z∞
dt1

0

Z∞
dt2

0

dt3
0

t3Z−α3

Zt3

dα1 e−(t1 +t2 +t3 )m ,

dα3
0

2

t3 := α1 + α2 + α3 ,

0

(4.19)
and ϕ is a phase factor depending on the “twisting” at the various vertices. As an example,
we consider the diagram Fig. 6a. It is
ϕ6a = irθp.

(4.20)

Evaluating the momenta integral, we obtain    

Z
α2
1
t1 + t2
(θp)2 .
A6a = Dα d/2 exp −p2 α2 − 2 (t1 + t2 ) −


4∆

(4.21)

Now we will turn to the string amplitude computations and their field-theory limits.
4.2. Field-theory limit of string amplitudes
Now we turn to the string formula. Two-loop amplitudes in commutative scalar theory
has been considered in [76–79] where a precise one–one correspondence (including
numerical coefficients) between corners of string-moduli space and Feynman diagrams
had been established. Here we will be interested in reproducing the θ -dependent terms
of (4.10), (4.17) and (4.21) from (3.22). The relation between the open-string coupling
constant gop and the field-theory one g3 has been fixed at tree-level [63]:
g3 = 25/2 gop (2α 0 )

d−6
4

.

(4.22)

This equations allows us to write the string normalization in the field-theory quantities.
Within field-theory this normalization depends on the different combinatoric factors of the
various Feynman diagrams and its determination can be quite cumbersome, in particular
in the noncommutative case. Because of this, here we neglect all numerical factors also
on the string side and focus only on the integrand structure. However, at least in the string
derivation, numerical factors can be put back easily.
In the general case of h loops, the string worldsheet is a h-annulus, which in the Schottky
representation is given by the h generators SI :
SI (z) =

aI z + bI
,
cI z + d I

aI dI − bI cI = 1,

(4.23)

of the Schottky group. Equivalently, one can introduce for each generator the multiplier kI
and the fixed points ξI and ηI , defined by

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

211

Fig. 8. Two-annulus and the Schottky representation.

SI (z) − ηI
z − ηI
= kI
.
SI (z) − ξI
z − ξI

(4.24)

As usual one can use the projective invariance to fix 3 of the 2h fixed points and
inequivalent h-loop worldsheet is parametrized by 2h − 3 fixed points and h multipliers.
To arrive at a noncommutative field-theory, in addition to the double scaling limit
(3.5), one also has to go to corners of the string moduli space in order to get
a finite contribution to the string amplitude that can be identified with Feynman
diagrams.
For the case of two loops, the string worldsheet is a two-annulus which in the Schottky
representation is represented by the shaded region in Fig. 8. We can fix η2 = 0, ξ1 = 1 and
ξ2 = ∞ using the projective invariance and we are left with one fixed point η1 and two
multipliers k1 and k2 . One can verify that [76–78]


p
η1 − k1
η1 + k1
√ ,
√ ,
C=
D=
B = −A = k2 ,
1 − k1
1 + k1


1
1 − η1 k1
1 + η1 k1
0
0
0
0
√ ,
√ .
C =
D =
(4.25)
B = −A = √ ,
k2
1 − k1
1 + k1
The irreducible vacuum bubble represents the backbone for all field-theory diagram we are
interested in. It is obtained, from the string formula, in the limit α 0 → 0, k1 , k2 , η1 → 0
with the following ti ’s fixed:
−α 0 ln k1 = t1 + t3 ,

−α 0 ln k2 = t2 + t3 ,

−α 0 ln η1 = t3 ,

(4.26)

where, as usual, the ti ’s are identified with the field-theory Schwinger parameters
introduced in the previous section. In this limit, the matrix A0 and the 2-loop period matrix
is given by  

1 0
(t1 + t3 )1
t3 1
0
,
(4.27)
− A = −α 2πiτ 1 =
(t2 + t3 )1
t3 1
2

212

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

where we have written A0 as a 2 × 2 block (will be h × h block for h loops case) with 1
being the d × d identity matrix in the Minkowski space.
In the string world-sheet the external states are represented by punctures on the borders
of the surface: in our case, we introduce the first external state in z1 and the second one
in z2 . In the field-theory limit, these points are identified with remaining two Schwinger
parameters αi through a relation similar to that of (4.26). However, the exact form of
the identification between puncture and Schwinger parameters depends on the particular
corner of the z’s integration region one is looking at; technically, this is the reason why a
single string Green function can give rise to field-theory diagrams having different forms
in terms of Schwinger parameters. Following [76–79], we start looking at the region where
z2 → A0 and z1 → D 0 which is related to the field-theory diagram of Fig. 4e. In this case,
B 0 is given in terms of the Schwinger parameters αi ’s by
!  

µ
1 B01
α1 pµ
=
.
(4.28)
µ
−α2 pµ
2 B02
Now we are ready to reproduce the field-theory results from (3.22). The string
worldsheet that gives a nonzero contribution to the field-theory diagram Fig. 4e is given
by Fig. 9a. Since there is no intersection between the two loops, the only modification is
in BI . We have   

−1
1 (t1 + t3 )1 −t3 1
0
=
(4.29)
−A /2
−t3 1 (t2 + t3 )1

and
i
µ
µ
BI = B 0 I + (θp)µ
2 


1
.
1

(4.30)

Thus, after the integration over the momenta running in the loops, the string formula gives
the following θ dependent part
(t1 + t3 )α22 + (t2 + t3 )α12 + 2α1 α2 t3 2
1
1
p −
(θp)2 (t1 + t2 ). (4.31)
− B T A−1 B =
2

4∆
e0 and C0 , the
Together with measure factor ∆−d/2 and the other pieces coming from V
N;h
θ -independent piece of (3.22) is exactly the same as the field-theory integrand (4.10).
Moreover, the θ -dependence is also precisely reproduced.

Fig. 9. String worldsheets corresponds to field-theory diagrams in Figs. 4e, 6a and 4b.

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

213

A remark about the integration region is in order. As explained in [79], the corner of
zi ’s considered yields the correct functional form of Feynman diagram in Fig. 4e, but does
not reproduce the field-theory integration region over the Schwinger parameters. However,
we know that a field-theory diagram is identified with the sum of contributions from a
few different corners of the string-moduli space. This sum reconstructs the expected fieldtheory integration region, without modifying the integrand previously found. Since turning
on a B-field does not change this identification, the situation is therefore the same as in
[79]. It is easy to check that the complete field-theory result (4.10) is reproduced from
string theory.
Next we consider the string worldsheet configuration of Fig. 9b. In this case,


i
θ
+
t
)1
t
1
+
(t
1
3
3

2 
(4.32)
−A/2 = 
,
i
t3 1 − θ (t2 + t3 )1
2
where again we have written A as a 2 × 2 block with 1 and θ being d × d matrices. It is
easy to work out the inverse


i
θ
(t
+
t
)1
−t
1

3
1  2 3
2 
(4.33)
(−A/2)−1 =
.

i
∆θ −t 1 + θ (t + t )1
3
1
3
2
As for B, it is

  

i
µ
µ
1 B1µ
α1 p + (θp) 
.
(4.34)
2
µ =
2 B2
−α pµ
2

Therefore
(t1 + t3 )α22 + (t2 + t3 )α12 + 2α1 α2 t3
θ2
1
p + (t2 + t3 − 2α2 )p
p.
− B T A−1 B = p
2
∆θ
4∆θ
(4.35)
Again the complete field-theory result (4.17) including the region of integration is
reproduced.
Finally we consider the field-theory diagram Fig. 6a. As explained in [76–78], Fig. 9a
actually contributes to both types of field-theory diagrams: Fig. 4e with external legs in
distinct propagators and Fig. 6a with external legs on the same propagator. One has to be
careful in selecting the range of the punctures in order to get the desired correspondence.
For the present case of Fig. 6a, one has to consider the range z2 ∈ [−1, −η] and (z1 −
η)/(z1 − 1) ∈ [−1, −η]. According to (3.12)–(3.14), A is unmodified from A0 and is given
by (4.27) since there is no intersection of the loops. As for B, it is given by (4.30). The
only difference from the case of Fig. 4e is in the range of the punctures z1 , z2 and hence a
different expression of B 0 :
!  

µ
1 B01
α2 pµ
=
.
(4.36)
µ
α2 pµ
2 B02

214

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

For our purpose, we concentrate on the θ -independent terms. Precisely the θ 2 term in (4.21)
is reproduced and confirm the complete field-theory result.
We have also checked the other field-theory diagrams in Figs. 4 and 6 and they are all
reproduced correctly from the string master formula (3.22).

5. Conclusions and discussions
In this paper, we used the open-string operator formalism to construct multiloopstring amplitudes with B-field. The key observation is that when expressed in terms of
the commutation relations of the string modes, the effects of noncommutativity is very
simple: apart from an overall scaling of the operators, only the zero-mode commutation
relation is modified. This allowed us to repeat the basic steps in operator formalism
and construct multiloop-string amplitudes with B-fields. In particular, we constructed
the open-string vertex operators and determined directly the one-loop open-string Green
function. For higher-loop amplitudes, we constructed the multiloop-reggeon vertex by
sewing together legs of a tree-level reggeon vertex. It was thus possible to write down
explicitly string amplitudes with B-field with any number of loops. Moreover one–one
correspondence between corners of open-string moduli space and Feynman diagrams is
again established.
With the master formula at our disposal, it is in principle possible to tackle many fieldtheory questions, which may appear mysterious otherwise, from the string point of view.
This is one major advantage of the string approach. One such question is the UV/IR effect
in noncommutative field theory. Since the θ dependences in any higher-loop amplitude are
isolated explicitly in our master formula (3.22), by taking the field-theory limit of the string
result, one can do a systematic analysis of the UV/IR effects in a noncommutative fieldtheory at arbitrary loop levels. Thus this approach may provide an uniform string theory
understanding of the UV/IR effect.
Another possible application is in the understanding of a similar effect in 3 dimension.
It was found in [50] that a Majorana fermion induces a noncommutative Chern–Simons
action at one loop. The coefficient is a step function which is 1/2 when the theory is
noncommutative and is 0 when the theory is commutative. It would be interesting to
understand this as well as other perturbative aspects of this theory from the string point
of view.
Although we did not write down explicitly the open-string Green function for higherloop, it is easy to extract it from the master formula (3.22). It is also easy to generalize the
considerations of this paper to the superstring case and to put Chan–Paton factors to obtain
results for noncommutative field-theory with U (N) gauge group.
So far, noncommutative D-branes appear in string theory in two basic forms: Moyal
deformation in the case of flat space–time with B-field, and “fuzzy deformation”
in the case of D-branes in WZW model. In particular the latter is generically a
more general nonassociative deformation [57]. It would be interesting to explore other

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

215

possibility of noncommutative geometry and deformation that can arise in string and
M-theory.
We hope to return to some of these issues in the future.

Acknowledgements
This work was partially supported by the Swiss National Science Foundation, by the
European Union under TMR contract ERBFMRX-CT96-0045, by the Swiss Office for
Education and Science and by MURST (Italy).
References
[1] M.R. Douglas, C. Hull, D-branes and the non-commutative torus, J. High Energy Phys. 2 (1998)
8, hep-th/9711165.
[2] A. Connes, M.R. Douglas, A. Schwarz, Noncommutative geometry and matrix theory:
compactification on tori, J. High Energy Phys. 02 (1998) 003, hep-th/9711162.
[3] A. Connes, Noncommutative Geometry, Academic Press, New York, 1994.
[4] G. Landi, An Introduction to Noncommutative Spaces and Their Geometries, Springer-Verlag,
Berlin, 1997.
[5] J. Madore, An Introduction to Noncommutative Differential Geometry and Its Physical
Applications, Cambridge University Press, Cambridge, UK, 1999.
[6] Y.-K.E. Cheung, M. Krogh, Noncommutative geometry from 0-branes in a background B field,
Nucl. Phys. B 528 (1998) 185, hep-th/9803031.
[7] T. Kawano, K. Okuyama, Matrix theory on noncommutative torus, Phys. Lett. B 433 (1998) 29,
hep-th/9803044.
[8] F. Ardalan, H. Arfaei, M.M. Sheikh-Jabbari, Noncommutative geometry from strings and
branes, JHEP 9902 (1999) 016, hep-th/9810072.
[9] C.-S. Chu, P.-M. Ho, Noncommutative open string and D-brane, Nucl. Phys. B 550 (1999) 151,
hep-th/9812219.
[10] C.-S. Chu, P.-M. Ho, Constrained quantization of open string in background B field and
noncommutative D-brane, Nucl. Phys. B 568 (2000) 447, hep-th/9906192.
[11] V. Schomerus, D-branes and deformation quantization, JHEP 9906 (1999) 030, hep-th/9903205.
[12] N. Seiberg, E. Witten, String theory and noncommutative geometry, JHEP 9909 (1999) 032,
hep-th/9908142.
[13] C.-S. Chu, P.-M. Ho, M. Li, Matrix theory in a constant C field background, hep-th/9911153,
Nucl. Phys., in press.
[14] A. Abouelsaood, C.G. Callan, C.R. Nappi, S.A. Yost, Open strings in background gauge fields,
Nucl. Phys. B 280 (1987) 599.
[15] C.-S. Chu, Noncommutative open-string: neutral and charged, hep-th/0001144.
[16] T. Filk, Divergences in a field theory on quantum space, Phys. Lett. B 376 (1996) 53.
[17] S. Minwalla, M.V. Raamsdonk, N. Seiberg, Noncommutative perturbative dynamics, hep-th/
9912072.
[18] M.V. Raamsdonk, N. Seiberg, Comments on noncommutative perturbative dynamics, hep-th/
0002186.
[19] J.C. Varilly, J.M. Gracia-Bondia, On the ultraviolet behavior of quantum fields over noncommutative manifolds, Int. J. Mod. Phys. A 14 (1999) 1305, hep-th/9804001.
[20] M. Chaichian, A. Demichev, P. Presnajder, Quantum field theory on noncommutative space–
times and the persistence of ultraviolet divergences, hep-th/9812180.

216

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

[21] C.P. Martin, D. Sanchez-Ruiz, The one-loop UV divergent structure of U (1) Yang–Mills theory
on noncommutative R 4 , hep-th/9903077.
[22] M. Sheikh-Jabbari, One-loop renormalizability of supersymmetric Yang–Mills theories on
noncommutative torus, JHEP 06 (1999) 015, hep-th/9903107.
[23] T. Krajewski, R. Wulkenhaar, Perturbative quantum gauge fields on the noncommutative torus,
hep-th/9903187.
[24] D. Bigatti, L. Susskind, Magnetic fields, branes and noncommutative geometry, hep-th/
9908056.
[25] Z. Yin, A note on space noncommutativity, hep-th/9908152.
[26] M. Li, Strings from IIB matrices, Nucl. Phys. B 499 (1997) 149, hep-th/9612222.
[27] H. Aoki, N. Ishibashi, S. Iso, H. Kawai, Y. Kitazawa, T. Tada, Noncommutative Yang–Mills in
IIB matrix model, hep-th/9908141.
[28] N. Ishibashi, S. Iso, H. Kawai, Y. Kitazawa, Wilson loops in noncommutative Yang–Mills, hepth/9910004.
[29] S. Iso, H. Kawai, Y. Kitazawa, Bi-local fields in noncommutative field theory, hep-th/0001027.
[30] O. Andreev, A note on open strings in the presence of constant B-field, hep-th/0001118.
[31] H. Grosse, T. Krajewski, R. Wullkenhaar, Renormalization of noncommutative Yang–Mills
theories: a simple example, hep-th/0001182.
[32] J. Madore, S. Schraml, P. Schupp, J. Wess, Gauge theory on noncommutative spaces, hep-th/
0001203.
[33] I. Bars, D. Minic, Non-commutative geometry on a discrete periodic lattice and gauge theory,
hep-th/9910091.
[34] J. Ambjorn, Y.M. Makeenko, J. Nishimura, R.J. Szabo, Nonperturbative dynamics of noncommutative gauge theory, hep-th/0002158.
[35] I. Chepelev, R. Roiban, Renormalization of quantum field theories on noncommutative R d ,
I. Scalars, hep-th/9911098.
[36] I.Ya. Aref’eva, D.M. Belov, A.S. Koshelev, Two-loop diagrams in noncommutative φ44 theory,
hep-th/9912075.
[37] I.Ya. Aref’eva, D.M. Belov, A.S. Koshelev, A note on UV/IR for noncommutative complex
scalar field, hep-th/0001215.
[38] M. Hayakawa, Perturbative analysis on infrared aspects of noncommutative QED on R 4 , hep-th/
9912094.
[39] M. Hayakawa, Perturbative analysis on infrared and ultraviolet aspects of noncommutative QED
on R 4 , hep-th/9912167.
[40] A. Matusis, L. Susskind, N. Toumbas, The IR/UV connection in the non-commutative gauge
theories, hep-th/0002075.
[41] C.-S. Chu, F. Zamora, Manifest supersymmetry in non-commutative geometry, JHEP 02 (2000)
022, hep-th/9912153.
[42] A. Schwarz, Noncommutative supergeometry and duality, hep-th/9912212.
[43] S. Ferrara, M.A. Lledo, Some aspects of deformations of supersymmetric field theories, hep-th/
0002084.
[44] S. Terashima, A note on superfields and noncommutative geometry, hep-th/0002119.
[45] W. Fischler, J. Gomis, E. Gorbatov, A. Kashani-Poor, S. Paban, P. Pouliot, Evidence for winding
states in noncommutative quantum field theory, hep-th/0002067.
[46] G. Arcioni, J.L.F. Barbon, J. Gomis, M.A. Vazquez-Mozo, On the stringy nature of winding
modes in noncommutative thermal field theories, hep-th/0004080.
[47] F. Ardalan, N. Sadooghi, Axial anomaly in non-commutative QED on R 4 , hep-th/0002143.
[48] J.M. Gracia-Bondia, C.P. Martin, Chiral gauge anomalies on noncommutative R 4 , hep-th/
0002171.
[49] L. Bonora, M. Schnabl, A. Tomasiello, A note on consistent anomalies in noncommutative YM
theories, hep-th/0002210.

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

217

[50] C.-S. Chu, Induced Chern–Simons and WZW action in noncommutative space–time, hep-th/
0003007, Nucl. Phys., in press.
[51] R. Blumenhagen, L. Goerlich, B. Kors, D. Lust, Asymmetric orbifolds, noncommutative
geometry and type I string vacua, hep-th/0003024.
[52] R. Gopakumar, S. Minwalla, A. Strominger, Noncommutative solitons, hep-th/0003160.
[53] N. Ishibashi, S. Iso, H. Kawai, Y. Kitazawa, String scale in noncommutative Yang–Mills, hep-th/
0004038.
[54] F. Zamora, On the operator product expansion in noncommutative quantum field theory, hep-th/
0004085.
[55] A. Rajaraman, M. Rozali, Noncommutative gauge theory, divergences and closed strings, hepth/0003227.
[56] J. Gomis, K. Landsteiner, E. Lopez, Non-relativistic non-commutative field theory and UV/IR
mixing, hep-th/0004115.
[57] A.Yu. Alekseev, A. Recknagel, V. Schomerus, Non-commutative world-volume geometries:
branes on SU(2) and fuzzy spheres, JHEP 9909 (1999) 023, hep-th/9908040.
[58] A.Yu. Alekseev, A. Recknagel, V. Schomerus, Brane dynamics in background fluxes and noncommutative geometry, hep-th/0003187.
[59] C. Bachas, M. Douglas, C. Schweigert, Flux stabilization of D-branes, hep-th/0003037.
[60] J. Pawelczyk, SU(2) WZW D-branes and their noncommutative geometry from DBI action,
hep-th/0003057.
[61] O. Andreev, H. Dorn, Diagrams of noncommutative phi-three theory from string theory, hep-th/
0003113.
[62] Y. Kiem, S. Lee, UV/IR mixing in noncommutative field theory via open string loops, hep-th/
0003145.
[63] A. Bilal, C.-S. Chu, R. Russo, String theory and noncommutative field theories at one loop,
hep-th/0003180.
[64] J. Gomis, M. Kleban, T. Mehen, M. Rangamani, S. Shenker, Noncommutative gauge dynamics
from the string worldsheet, hep-th/0003215.
[65] H. Liu, J. Michelson, Stretched strings in noncommutative field theory, hep-th/0004013.
[66] S. Sciuto, The general vertex function in dual resonance models, Lett. Nuovo Cimento 2 (1969)
411.
[67] A. Della Selva, S. Saito, A simple expression for the Sciuto three-reggeon vertex generating
duality, Lett. Nuovo Cimento 4 (1970) 689.
[68] P. Di Vecchia, R. Nakayama, J.L. Petersen, S. Sciuto, Properties of the three-reggeon vertex in
string theories, Nucl. Phys. 282 (1987) 103.
[69] A. Neveu, P. West, Symmetries of the interacting gauge-covariant bosonic string, Nucl. Phys.
B 278 (1986) 601.
[70] A. Neveu, P. West, Group theoretic approach to the perturbative string S-matrix, Phys. Lett.
B 193 (1987) 187.
[71] P. Di Vecchia, M. Frau, A. Lerda, S. Sciuto, A simple expression for the multiloop amplitude in
the bosonic string, Phys. Lett. B 199 (1987) 49.
[72] P. Di Vecchia, M. Frau, A. Lerda, S. Sciuto, N -string vertex and loop calculation in the bosonic
string, Nucl. Phys. B 298 (1988) 526.
[73] C. Lovelace, Simple N-reggeon vertex, Phys. Lett. B 32 (1970) 490.
[74] P. Di Vecchia, F. Pezzella, M. Frau, K. Hornfeck, A. Lerda, S. Sciuto, N -point g-loop vertex for
a free bosonic theory with vacuum charge Q, Nucl. Phys. B 322 (1989) 317.
[75] P. Di Vecchia, A. Lerda, L. Magnea, R. Marotta, R. Russo, String techniques for the calculation
of renormalization constants in field theory, Nucl. Phys. B 469 (1996) 235, hep-th/9601143.
[76] P. Di Vecchia, A. Lerda, L. Magnea, R. Marotta, R. Russo, Two-loop scalar diagrams from
string theory, hep-th/9607141.

218

C.-S. Chu et al. / Nuclear Physics B 585 (2000) 193–218

[77] R. Marotta, F. Pezzella, Two-loop φ 4 -diagrams from string theory, Phys. Rev. D 61 (2000)
106006, hep-th/9912158.
[78] R. Marotta, F. Pezzella, String-generated quartic scalar interactions, hep-th/0003044.
[79] A. Frizzo, L. Magnea, R. Russo, Scalar field-theory limits of bosonic string amplitudes, hep-th/
9912183.

Japan b Departament ECM. Introduction Many aspects of string physics in ten and eleven dimensions can be understood from a world volume point of view using D-branes and M-branes. Spain Received 29 March 2000.25. World volume solitons 1.Nuclear Physics B 585 (2000) 219–252 www.es (J.nl/locate/npe T-duality covariance of SuperD-branes Kiyoshi Kamimura a. We conclude with some comments on extensions of our approach to arbitrary kappa symmetric backgrounds. In particular.  2000 Elsevier Science B. non-BPS D-branes and non-abelian SuperD-branes.elsevier. of the NS–NS and R–R coupling superfields.10. 11. and consequently. duality symmetries of the full string theory admit a field theory realization on the world volume effective actions describing the low energy dynamics of these branes. revised 7 June 2000. PACS: 11.∗ . 0550-3213/00/$ – see front matter  2000 Elsevier Science B. the expected ones in the T-dual theory. In particular. Universitat de Barcelona and Institut de Física d’Altes Energies.3 2 1 3 ( 0 0 ) 0 0 4 0 3 . Diagonal 647.V.-w Keywords: T-duality covariance. accepted 20 June 2000 Abstract T-duality realized on SuperD-brane effective actions probing in constant Gmn and bmn backgrounds is studied from a pure world volume point of view.ub. Joan Simón b a Department of Physics. Toho University.jp (K.toho-u. It is proved that requiring T-duality covariance of such actions “fixes” the T-duality transformations of the world volume dynamical fields. All rights reserved. Simón). The latter proof is generalized to arbitrary bosonic backgrounds. jsimon@ecm. E-08028 Barcelona. which allow us to prove that bosonic supersymmetric world volume solitons of the original theory generate. The analysis is extended to uncover the mapping of the symmetry structure associated with these SuperD-brane actions. All rights reserved. Facultat de Física. solutions of the classical equations of motion of the latter actions (world volume solitons) do admit a space time interpretation in terms of intersections of branes due to the gauge invariant character of the scalars describing these actions [1–3]. D-branes.sci.V. E-mail addresses: kamimura@ph. Kamimura). we determine the T-duality transformation properties of kappa symmetry and supersymmetry. through T-duality. Besides that.Kk.X . Funabashi 274-8510. Av. Kappa symmetry. In this paper we will study the realization of T-duality on SuperD-brane effective actions probing in constant ∗ Corresponding author. PII: S 0 5 5 0 .

Thus. we are left with an open string whose end points are constrained to move in a p-dimensional hyperplane... or viceversa [5]. Simón / Nuclear Physics B 585 (2000) 219–252 Gmn and bmn backgrounds. which requires the existence of an isometric direction.11] it was proved that the double dimensional reduction of a Dp-brane action yields the direct dimensional reduction of a D(p−1)-brane. Since under a longitudinal T-duality 2 . It is well known that T-duality admits a field theory realization in the zero slope limit of closed string theory giving rise to the T-duality rules among the NS–NS and R–R massless fields and mapping N = 2 D = 10 IIA Supergravity into N = 2 D = 10 IIB Supergravity. but this is beyond the scope of the present paper. giving a full detailed analysis of the commutative side. 3 Such a reduction consists of a partial gauge fixing of the world volume diffeomorphisms to fix in which direction the original D-brane is wrapping. J. and a functional truncation that discards the non-zero modes of the dynamical 2 By a longitudinal T-duality. The analysis done in [10–12] shows that the right way to realize a longitudinal T-duality on D-brane effective actions is to apply a double dimensional reduction. generalizing the approach followed in Matrix theory compactifications [15]. Since both of them are known to be of DBI + WZ type form [6–9]. our analysis can be seen as a first step towards the supersymmetric extension of such an equivalence.220 K. Their approach was based on the already known T-duality rules mapping type IIA/IIB backgrounds derived in [5]. the T-dual one fits into a vector supermultiplet in (1. Although the number of bosonic massless states in the open string spectrum remains invariant ( 8 ). T-duality covariance is the most natural requirement having in mind the conformal field theory description of D-branes in terms of open strings [13]. while the original bosonic massless open string spectrum fits into a vector supermultiplet in (1.c. One can ask whether such a realization exists in the same limit for the open string sector. Dp-branes appear as hyperplanes on which open strings can end. The dimensionality (p + 1) depends on the number of scalar fields satisfying Neumann boundary conditions (b. the number of bosonic scalar ones increases by one. Given the physical equivalence between bosonic commutative Dbrane gauge theory actions on such backgrounds and non-commutative gauge theory [4]. by requiring T-duality covariance of the Dirac–Born–Infeld (DBI) and the Wess–Zumino (WZ) terms appearing in the D-brane effective action. i. In [10. 3 It would also be interesting to study transverse T-duality on abelian D-brane effective actions. but without rewriting the ten dimensional background fields in terms of the nine dimensional ones. a Neumann b.). we mean a T-duality along a direction parallel to the hyperplane defined by the initial Dp-brane.e. p − 1) dimensions [14]. D(p−1)-brane. the requirement of T-duality covariance is certainly justified.c. In other words. any effective field theory description of the initial and T-dual open string sectors should be a field theory realization of such vector multiplets in the corresponding dimensions. It was later proved in [12] that the latter set of transformations could be derived from a pure world volume perspective.c. on their symmetry structures and on supersymmetric bosonic world volume solitons. p) dimensions. while the number of bosonic vectorial ones decreases by the same amount. which describe the low energy dynamics for the massless open string fields including their interactions with the massless closed string sector [6–9]. is transformed into a Dirichlet b. . Kamimura. This is answered by studying the T-duality properties of D-brane effective actions.

which in our case is a p-dimensional field theory. (1.1) which is from now on called kappa symmetry preserving condition [18. thus generalizing not only previous bosonic analysis but also the supersymmetric one [16] in which kappa gauge symmetry was fixed. the T-dual D(p−1)-brane effective action. We will show how the requirement of T-duality covariance fixes the necessary chirality changing mapping between the fermionic degrees of freedom describing type IIA/IIB D-branes in addition to the one for bosonic fields. and θi0 with the same chiralities in type IIB. this mapping of dynamical degrees of freedom indeed maps the original DBI and WZ terms into the T-dual ones. It is interesting to remark that such mapping of supersymmetric world volume solitons we have described is nothing but a world volume realization of a well known . generically. When describing SuperD-branes. we will just keep their zero modes along the direction of dualization. Here  is a linear combination of Killing spinors of the background and the number of supersymmetries preserved by the combined background/brane configuration is the number of linearly-independent solutions of (1. The supersymmetric projection conditions Pi0  0 = ± 0 are obtained from the initial ones Pi  = ± by rewriting them in terms of the T-dual Killing spinors  = ( 0 ). we will show how kappa symmetry and supersymmetry transformations of D-branes probing in constant Gmn and bmn are mapped under T-duality. Being world volume scalars. physics have a much more natural description in terms of the T-dual theory. For any super-brane action in any background compatible with kappa symmetry. Simón / Nuclear Physics B 585 (2000) 219–252 221 fields in the infinitely massive (R → 0) limit. a set of constraints among the excited dynamical fields or BPS equations and a set of supersymmetry projection conditions P i  = ± determining the type of branes described by the configuration. Furthermore. This is again consistent with the conformal field theory picture. 2) having different ten-dimensional chiralities in type IIA. We shall also study the effect of T-duality on bosonic supersymmetric world volume solitons. one must also include fermionic scalar fields θi (i = 1. Kamimura. thus generalizing the corresponding bosonic analysis done in [12]. we will find out the T-duality transformation properties of the Γκ matrix appearing in kappa symmetry transformations. In particular. In this way. Such an equation involves. such configurations must satisfy Γκ  = . Since we will always be concerned with T-duality performed along an isometric direction of the background it ensures the preservation of supersymmetry under the dualization [17]. We will argue that the functionally truncated and partially gauge fixed BPS equations are the corresponding BPS equations describing the supersymmetric T-dual configuration.1).19].K. This mapping having the same form as the one among dynamical fermionic fields θ = θ (θ 0 ) which would have already been fixed by T-duality covariance of the SuperD-brane effective action. because whenever the radius R of the√circle along which we are Tdualizing becomes much more smaller than the string scale α 0 . The extension of the analysis done in [12] to the supersymmetric case is conceptually straightforward. J. Our proof is not only concerned with effective actions but also uncovers their gauge and global symmetry structures.

altogether giving a ν = 1/4 threshold bound state. D-brane effective actions are supersymmetric field theories. we show the generating solution character of T-duality transformations in the low energy description of the open string sector. BIons are mapped among themselves. We will show that. which are known to be related by some transformation of their generators. Some details are given in appendices.222 K. non-BPS D-branes and nonabelian SuperD-branes. To begin with. Given such a relation. from which we can derive the T-duality transformation properties of the bosonic kappa matrix Γκ |θ=0 . in agreement with the conformal field theory picture. Kamimura. . while dyons are mapped to a non-threshold bound state of a D2-brane and a fundamental string parallel to it intersecting in a point with a D2-brane. Simón / Nuclear Physics B 585 (2000) 219–252 algebraic mapping. in close analogy with such generating character already known in type IIA/IIB supergravities describing the massless closed string sector. just as constant flux of magnetic field on the D-brane is seen as D-branes at angles in the T-dual picture. In particular. relying on the θ = 0 condition characterizing any bosonic configuration and the standard T-duality rules mapping bosonic backgrounds. Having proved the mapping of supersymmetric world volume solitons under T-duality for constant Gmn and bmn backgrounds we extend the proof for an arbitrary bosonic background. These BPS equations derived from the phase space formulation of D-branes are entirely equivalent to the resolution of (1. they provide a field theory realization of N = 2 D = 10 IIA(IIB) SuperPoincaré algebras for p even (odd). The mapping of BPS equations and supersymmetry projection conditions will be illustrated by some examples. They include T-duality of D-brane effective actions for arbitrary kappa symmetric backgrounds. generically. we concentrate on solitons in bmn 6= 0 constant backgrounds. constant electric field (induced by the electric components b0a ) boosts the configuration in the direction along which we are T-dualizing. BPS states in string theory admit different realizations. Such a bound is exactly the same one derived from a hamiltonian analysis of brane effective actions [3]. Later. In the last section we comment on possible generalizations and/or extensions of our present work. The extension to arbitrary bosonic background is examined in Section 6. In our case. In Section 5 the mapping of world volume solitons is discussed with some examples. this being the field theoretical description of such states. giving rise to some set of BPS equations. J. reminiscent of T-duality [21]. we will study the effect of T-duality on BIon and dyon solutions of D-brane effective actions probing SuperPoincaré (bmn = 0) background. being tendimensional target space covariant. One is purely algebraic and is based on the saturation of the BPS bound in the supersymmetry algebra.1) due to the connection between the supersymmetry algebra and the structure of the kappa symmetry projector [22–24]. In Sections 2–4 we discuss the T-duality covariance of the D-brane actions and their symmetry properties. we will study T-duality on tilted dyons and tilted BIons on non-threshold bound states of D-strings and D3-branes. whose group of global isometries contains the isometry supergroup of the background [20]. In this way.

The R–R field strength R is expressed in type IIA as . containing additional constant bosonic components (bmn )   1 1 1 M N a a ˜ ˜ ˜ ˜ B = dZ dZ BMN = −θ Γ11 Γa dθ dx˜ + θ Γ dθ + dx m dx n bmn . it is the p + 1 form part of a symbolic sum of differential forms LWZ [10.2) (2. Concerning the WZ term in (2. (2. LWZ = LWZ p+1 . H = dB = −E α CΓ11 Γa αβ E β E a ≡ − EΓ11ΠE (2.30] satisfying dLWZ = −Tp ReF . Simón / Nuclear Physics B 585 (2000) 219–252 223 2.5) which for the backgrounds considered in this paper take the form E a = dx˜ a + θ˜ Γ a dθ˜ ≡ Π a . Effective action and symmetry structure The effective Lagrangian density of a type IIA Dp-brane is a sum of DBI and WZ terms [25–29] L = LDBI + LWZ .3). Due to the fact that we are considering constant background all the dependence of the constant dilaton background is included in the string coupling constant through gs = eφ0 . (2. E α = dθ˜ α . x˜ a ≡ x m em a . (2. Kamimura.10) / = Γa Π a . θ α ) parametrize the target superspace. (2. (2.4) where Eµ a stand for the components of the supersymmetric invariant one forms E A ≡ dZ M eM A = dσ µ ∂µ Z M eM A ≡ dσ µ Eµ A . (2.6) (2. (2.3) where Tp is the Dp-brane tension scaling as Tp ∝ (gs α 0(p+1)/2 )−1 . The DBI term depends on the world volume induced metric Gµν = Eµ a Eν b ηab . θ˜ α ≡ θ α eα α . the dynamical fields will be indicated by where Π “primes” and Γ11 must be replaced by τ3 .11) where R is the field strength of the R–R gauge potential C.7) em a and eα α are constant components of the supervielbeins and Z M ≡ (x m .11. For type IIB D(p − 1)-branes. J. p LDBI = −Tp − det(Gµν + Fµν ).K.9) 2 2 2 but still satisfying the supergravity constraint   / .1) (2.   LWZ = −Tp CeF .8) where B stands for the NS–NS two form. It also depends on the supersymmetric invariant F = dV − B.

an abelian U (1) gauge symmetry (c) and kappa symmetry (κ). ρ (p) . Vµ . ρ (p) = SA (Π)e SA (Π) / = X Π / 2`+1 . we will decompose the infinitesimal transformations (˜s φ i ) leaving the effective action invariant. s θ˜ α = ξ µ ∂µ θ˜ α + δκ θ˜ α .20) `=0 while for type IIB D(p−1)-brane   0 / 0 )eF τ1 p . (Γ11 )`+1 (2` + 1)! (2. Simón / Nuclear Physics B 585 (2000) 219–252 R = ECA (Π)E.18) + δκ θ˜ Γ a θ˜ θ˜ Γ11Γa ∂µ θ˜ + δκ x m ∂µ x n bmn . we will not take into account the infinite number of non-trivial global symmetries existing for the D-string and D0-brane effective actions [31–33].17) ν ν (2. (τ3 )` (2` + 1)! (2. R0 = −E 0 SB (Π SB (Π / )= X Π / 0 2`+1 . (τ3 )`+1 (2`)! (2. into gauge (sφ i ) and global (1φ i ) ones.15) sVµ = ξ ∂ν Vµ + Vν ∂µ ξ + ∂µ c + δκ Vµ . γ (p) = √ − det(G + F ) (2. even though our conclusions also apply to them.16) where the kappa symmetry transformation for the gauge field δκ Vµ is determined by requiring the invariance of the gauge invariant tensor F in (2.13) `=0 Denoting the set of fields described by the SuperD-brane effective action (2. They are given by s x˜ a = ξ µ ∂µ x˜ a + δκ x˜ a = ξ µ ∂µ x˜ a − δκ θ˜ Γ a θ˜ . J. The set of gauge symmetries involves world volume diffeomorphisms (ξ µ ). 4 They act as follows: 4 Along the whole paper.224 K.8) as   1 a a ˜ ˜ ˜ ˜ δκ Vµ = −δκ θ Γ11 Γa θ ∂µ x˜ − θ Γ ∂µ θ 2 1 (2.14) {φ i } = Z M . Kamimura.12) `=0 whereas in type IIB as 0 0 / )τ1 E 0 .19) / )eF for type IIA theory. (2.21) `=0 The set of global symmetries involves supersymmetry (). (2. 2 while δκ θ˜ is fully determined by  δκ θ˜ = κ¯ 1 − γ (p) . bosonic translations (am ) and Lorentz transformations (ωmn ). ρ (p) is the (p + 1) world volume form coefficient of SA (Π   / F p+1 . . ρ (p−1) = − CB (Π 0 CB (Π / )= X Π / 0 2` .1) by  (2. / CA (Π) / = X Π / 2` (Γ11 )`+1 (2`)! (2.

It was argued in the introduction that Tk was conveniently realized on the world volume action as a kind of double dimensional reduction.2) realizing a longitudinal T-duality transformation (Tk ) on SuperD-brane effective actions.1) z = ρ. (3.  b0 b = x 0 mˆ . Gzz 5 The derivation is left to Appendix A. (3. Requiring T-duality covariance for the DBI action (2. Vµ (3. 4 1Vµ = δ Vµ + δa Vµ + δω Vµ  1 = ˜ Γ11 Γa θ˜ ∂µ x˜ a − ˜ Γ11 Γa θ˜ θ˜ Γ a ∂µ θ˜ + ˜ Γa θ˜ θ˜ Γ11 Γ a ∂µ θ˜ 6 + 1x m ∂µ x n bmn . T-duality covariance of SuperD-branes In this section. Z b0 0M Vµˆ = Vµˆ0 . As in the analysis of degrees of freedom done in [10–12]. ρ . we will find the constraints derived from the T-duality covariance requirement on the DBI term (2. Z Z p p p+1 0 σ − det(G + F ) −→ −Tp−1 dp σ − det(G 0 + F 0 ) −Tp d Tk (3.2) ∂ρ φˆı = 0. saying in which direction the D-brane is locally wrapping the circle of radius R. Kamimura. J. allow us to derive 5 a relation between brane tensions Z p 0 R ≡ dρ Gρρ (3. (3. Simón / Nuclear Physics B 585 (2000) 219–252 1x˜ a = δ x˜ a + δa x˜ a + δω x˜ a = ˜ Γ a θ˜ + aa + ωa b x˜ b .4) b0 where η and ΓNb M are some set of constants.22) (2. we will allow the following 0 relations among {φˆı } and {φ ˆı } Vρ = ηz0 . 225 (2. and a set of constraints among the background superfields η2 = G0zz .3) (3.1)–(3. whose solution is given in the appendix.2). Afterwards. θi0 α = Z N ΓNb M . θiα .K. that discards all but the zero modes of the rest of dynamical fields in the limit R → 0. θ . it will be proved that such a solution maps the WZ terms (2.24) 3. α 1 1θ˜ α = δ θ˜ α + δω θ˜ α = ˜ α + ωab Γab θ˜ .23) (2.7) . The latter consists of a partial gauge fixing of the world volume diffeomorphisms   σ µ ≡ σ µˆ . x m ≡ x mˆ .5) under the assumptions (3.3) of both theories.6) Tp−1 = Tp R. z . besides a functional truncation  ˆı  mˆ α φ = x .  b Z M = x mˆ .4).

10) b0 N Gzz Mz N z  1 b0 b0 0 M N (M+M 0 )N (−)MN BMz ΓM BM bN b− b GNz b − GMz b BN bz = (−) b ΓN b BM b0 . λ (3.11) can be interpreted as the T-duality rules for the NS–NS background superfields considered in this paper.13) (3. The second set of constraints. GM bN b− 1 log |Gzz |. due to the fact that these superfields admit an expansion in the fermionic variables θ . of a local SO(1.16) ˆ z).18) where Γ11 θ˜i = (−1)i+1 θ˜i .(3. equations (3.226 K.11) b0 N Gzz √ 0 ∝ (gs0 α 0p/2 )−1 since gs0 = gs α 0 /R.7)–(3. We have also made use where we have splitted the flat tangent space indices as a = (a. (3. The analysis is carried out in Appendix A to which we refer for further details.15) (3. Gzz (3. related to the fermionic components of the NS–NS superfields. Kamimura.14) (3.8) (3. ez0 a = λ0 δz a θ˜1 = a2 θ˜20 .(3. (3. Eq. emz ˆ m ˆ λ0 0 0 a ˆ emˆ aˆ = Γmˆ mˆ em ˆ0 .17) p √ with λ and λ0 being some constants standing for Gzz and Gz0 z0 . fixes the chirality change mapping of the target space spinor fields up to signs (3.6) is consistent with the T-dual tension. θ˜2 = −a1 Γz θ˜10 . Since we already know such superfields. (3. (3. respectively.4). (3. It is nevertheless natural to expect two different sets of constraints. Simón / Nuclear Physics B 585 (2000) 219–252 −η b0 0 M G b = ΓM BM b b0 z . 9) rotation in both type IIA/B tangent spaces to choose ez a = λδz a .9) 1  b0 b0 0 M N (M+M 0 )N ΓM G b G b − BMz b BN bz = (−) b ΓN b GM b0 .12) 2 Eqs.   b[0mˆ 0 z m ˆ0 nˆ 0 0 0 bmˆ nˆ = Γmˆ Γnˆ bmˆ 0 nˆ 0 − 0 enˆ 0 ]z . J. λ b0ˆ nˆ nz = Γ . The first set has to do with the bosonic components of such superfields. They are the usual T-duality rules for the bosonic NS–NS background fields expressed in terms of the vielbeins φ0 = φ − bmz ˆ 0 = Γmˆ nˆ enz ˆ .11) can be used to fix the set of constants introduced in (3.7)–(3. Gzz Mz −η b0 0 M BMz GM b = ΓM b b0 z . Tp−1 The latter is equivalent to the standard T-duality transformation for the dilaton background field. τ3 θ˜i0 = (−1)i+1 θ˜i0 .19) .

J.22) 4. From these equations and the algebra of Pauli matrices. The first one ensures that s˜ z will not move our configuration from the gauge slice defined by the partial gauge fixing. the following set of identities can be derived θ˜ Γ aˆ ∂µˆ θ˜ = θ˜ 0 Γ aˆ ∂µˆ θ˜ 0 .21) ReF = −a1 a2 λ R0 eF dρ. ˜ 11 Γ z ∂µˆ θ˜ = −θ˜ 0 Γ z ∂µˆ θ˜ 0 . The WZ action of IIB D-brane is obtained by integrating the IIA one over ρ if a1 a2 = −1. LWZ A −→ Tk LWZ B . . (3. T-duality and symmetry structure Let us define by A the subspace of the field configuration space defined by the partial gauge fixing (z = ρ) and functional truncation (∂ρ φˆı = 0). In general.20) ∂θ (∂θ 0 ) and are replaced by any The same form of formulas are also applied when IIA(IIB) spinors related by (3. θ˜ Γ z ∂µˆ θ˜ = −θ˜ 0 τ3 Γ z ∂µˆ θ˜ 0 . Kamimura.18). θ (θ 0 ) (3. θΓ θ˜ Γ11 Γaˆ ∂µˆ θ˜ = θ˜ 0 τ3 Γaˆ ∂µˆ θ˜ 0 .K. Simón / Nuclear Physics B 585 (2000) 219–252 227 and a1 2 = a2 2 = 1.  0 (3. so we must require two consistency conditions. A is not left invariant under s˜φ i .5) in Appendix B. The T-duality covariance of the WZ actions follows from (B.

.

(4.1) s˜z.

A = 0 H⇒ ξ ρ = −(δκ z + 1z).

A . The second consistency condition ensures that s˜φˆı will respect the functional truncation .

(4.2) ∂ρ s˜ φˆı .

ωzaˆ = 0. 8).1) modifying the transformation property of Vµˆ . The fixation of the diffeomorphism with respect to ρ will induce compensating transformations through (4.4) which is explicitly breaking the global symmetry group SO(1. A sufficient solution of this set of constraints is given by ξ µˆ = ξ µˆ (σ νˆ ). The gauge symmetries are given by 6 .3) (4. 9) into the SO(1. c(σ µˆ .A = 0 . constraining the gauge and global parameters. In this way. ρ) = c(σ ˆ µˆ ) + Aρ. κ = κ(σ νˆ ). the symmetry structure (˜s φˆı |A ) of the partially gauge fixed truncated action is found. (4.

˜ aˆ θ. ˜ s x˜ aˆ .

5) .A = ξ µˆ ∂µˆ x˜ aˆ − δκ θΓ (4.

α µ ˆ α α (4.6) s θ˜ .

.A = ξ ∂µˆ θ˜ + δκ θ˜ .

7) sVρ . ˜ 11 Γz θ˜ λ + δκ x mˆ bmz (4.

6 It should be understood that the transformations appearing in the right-hand side of the forthcoming equations must be computed in A.A = ξ µˆ ∂µˆ Vρ − δκ θΓ ˆ . .

Simón / Nuclear Physics B 585 (2000) 219–252 .228 K. Kamimura. J.

sVµˆ .

A = ξ νˆ ∂νˆ Vµˆ + Vνˆ ∂µˆ ξ νˆ + ∂µˆ c∗ + δκ∗ Vµˆ and the global symmetries are .

ˆ 1x˜ aˆ .

A = δ x˜ aˆ + δa x˜ aˆ + δω x˜ aˆ = ˜ Γ aˆ θ˜ + aaˆ + ωaˆ bˆ x˜ b . .

α 1 ˆ 1θ˜ α .

4 .A = δ θ˜ α + δω θ˜ α = ˜ α + ωaˆ b Γaˆ bˆ θ˜ .

1Vρ .

A = A + ˜ Γ11 Γz θ˜ λ + 1x mˆ bmz ˆ . .

∗ ∗ ∗ 1Vµˆ .

(4. (p − 1)-dimensional diffeomorphisms and U (1) gauge symmetry (ξ µˆ . it is proved that .1. a. and c ∗ = c + Vρ ξ ρ . To begin with. (4.11) ω (4. c∗ ) are trivially mapped as can be seen by inspection of Eqs.10) (4.7).13) where δi∗ Vµˆ = δi Vµˆ + δi z ∂µˆ Vρ . ω) (4. that is we will prove that the whole symmetry structure of these theories is properly mapped under T-duality.14) In the following we will see that these transformations give the right transformation properties of the T-dual variables. 4.9) (4. = δ Vµˆ + δ Vµˆ + δ Vµˆ . .5)–(4. Vρ becomes the new T-dual scalar. Kappa symmetry In Appendix C. A  a (4.8). As we already know from the bosonic analysis.12) i = (κ. as can be seen from its diffeomorphism transformation (4.8) (4.

δκ θ˜ α .

Once (4.20). (4.4). (4. using the identities (3. it is straightforward to extend the proof for δκ x˜ aˆ .A −→ δκ 0 θ˜0 α . .15) Tk where IIA and IIB kappa parameter functions are related by the same form of equations as θ ’s in (3. The first non-trivial check is the gauge symmetry analysis of the T-duality mapping Vρ |A = ηz0 in (3.16) It shows that kappa symmetry transformations of the fermionic sector of the theory are correctly mapped under T-duality. κ˜ 2 = −a1 Γz κ˜ 10 .15) is known.18) κ˜ 1 = a2 κ˜ 20 .

δκ Vρ .

A = −δκ θ˜ Γ11 Γz θ˜ λ + δκ x mˆ bmz ˆ −→ Tk 0 ηδκ z0 = λδκ 0 θ˜0 Γ z θ˜0 + λδκ 0 x 0 mˆ em ˆ . thus allowing us to write the kappa symmetry transformations of the bosonic scalar sector in the T-dual description in a fully ten-dimensional covariant way . z (4.17) It can be seen that δκ Vρ |A turns out to be the kappa symmetry transformation for the T-dual scalar δκ 0 z0 .

ˆ z . Kamimura.18) We are left with kappa transformations of the Vµˆ components.19) −→ δκ 0 Vµˆ0 .20) which finishes the proof of our claim. + δκ 0 θ˜ 0 Γ a θ˜ 0 θ˜ 0 τ3 Γa ∂µˆ θ˜ 0 + δκ 0 x 0 ∂µˆ x 0 bmn 2 κ0 (4. Supersymmetry It is natural to apply the T-duality transformation properties of the fermionic scalar fields (3.K. J. it can be shown that  (4. δκ∗ Vµˆ ≡ δκ Vµˆ + δκ z ∂µˆ Vρ Tk where δ κ0 Vµˆ0   1 ˜0 a ˜0 0 0a 0 ˜ ˜ = −δ θ τ3 Γa θ ∂µˆ x˜ − θ Γ ∂µˆ θ 2 1 m n 0 . δκ 0 x˜ 0a = −δκ 0 θ˜0 Γ a θ˜0 229 (4. There is an additional contribution to the kappa transformation from ξ ρ in (4.2.18) for the supersymmetry parameters ˜1 = a2 ˜20 .20). In this way . 4.1). Simón / Nuclear Physics B 585 (2000) 219–252  for a = a. Using the identities (3.

δ θ˜ α .

22) and the corresponding behaviour for δ x˜ aˆ follows immediately.9) and Vρ in (4. δ∗ Vµˆ ≡ δ Vµˆ + δ z ∂µˆ Vρ Tk 4.23) δ Vρ = ˜ Γ11 Γz θ˜ λ + δ x mˆ bmz ˆ Tk from which the ten-dimensional character of Vρ can be emphasized again and δ 0 x˜ 0a = δ 0 θ˜0 Γ a θ˜0  for a = a. −→ Tk (4.25) −→ δ 0 Vµˆ0 . ˆ z . (4. 8) symmetry group. 0z  −→ ηδ z0 = −λ ˜ 0 Γ z θ˜ 0 − ˜ 0 Γ aˆ θ˜ 0 ea0ˆmˆ emˆ (4.A ˜2 = −a1 Γz ˜10 . Poincaré bosonic global symmetries Let us concentrate on the manifest ISO(1.24) Concerning to Vµˆ the susy transformation modified by ξ ρ is mapped to the IIB one as in the kappa symmetry case. From the transformation of x˜ aˆ in (4.  (4. To begin with.3.11) it follows . We are thus left with the supersymmetry transformations of the gauge field.21) δ 0 θ˜0 α (4.

230

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252
ˆ

1x˜ 0 aˆ = aaˆ + ω aˆ bˆ x b ,

1x˜ 0z = Aλ0 ,

(4.26)

which allows us to interpret it as the corresponding ISO(1, 8) infinitesimal transformations
and a x˜ 0z coordinate translation in the T-dual target space, whenever we take the constant A
as A = a0z /λ0 , without loss of generality. It is worthwhile emphasizing, as in [12], that the
origin of the translational symmetry is the original U (1) gauge symmetry. Furthermore,
δa∗ Vµˆ + δω∗ Vµˆ −→ δa0 Vµˆ0 + δω0 Vµˆ0 + ∂µˆ c(1)

(4.27)

does describe the ISO(1, 8) transformations in the T-dual theory up to a U (1) transformation, which can be absorbed in a redefinition of the T-dual U (1) gauge parameter c∗
without loss of generality.
The latter analysis shows the existence of the ISO(1, 8) symmetry group, but we know
it should be enhanced to the full ISO(1, 9). In fact, there is no theorem guaranteeing the
equality of the full symmetry group of the T-dual effective action with the symmetry group
of the partially gauge fixed truncated action. What is indeed true is that the latter group
is a subgroup of the former. In other words, H −1,p+1 (s|d) ⊆ H −1,p (sA |dA ), H −1,n (s|d)
being the cohomological group at ghost number minus one characterizing the set of nontrivial global symmetries of any n-dimensional classical field theory [34,35]. There are
examples of such an enhancement in the literature. For instance, the D-string effective
action is known to have an infinite set of non-trivial global symmetries [31–33], while
such structure is not known to exist for the D2-brane effective action, even though they
are T-dual to each other. In the present case, it is certainly true that the T-dual theory is
invariant under the following set of rotations,
ˆ 0z
x˜ , 1x˜ 0z = ω0z aˆ x˜ 0 aˆ ,
1x˜ 0 aˆ = ω0 az 
α
1
ˆ
1θ˜ 0α = ω0zb Γzbˆ θ˜ 0 ,
2
0
1Vµˆ0 = 1x 0m ∂µˆ x 0n bmn

(4.28)
(4.29)
(4.30)

as is clear from the fact that the T-dual IIB action has manifest ISO(1, 9) invariance.

5. Supersymmetric world volume solitons
The goal of the present section is to prove that any bosonic supersymmetric world
volume soliton of a IIA Dp-brane in constant Gmn , bmn backgrounds is mapped under
T-duality into the corresponding bosonic supersymmetric world volume configuration for
the T-dual theory. This will be shown in two different, but complementary, ways. First of
all, the T-duality behaviour of the kappa symmetry preserving condition will be analyzed,
and after that, the same analysis will be carried in the hamiltonian formalism describing
D-branes [20], paying special attention into the hamiltonian constraint, giving rise to the
energy density of such BPS configurations.
It is known that any bosonic supersymmetric world volume configuration must satisfy
[18,19]
Γκ  = ,

(5.1)

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

231

where Γκ is the matrix appearing in the kappa symmetry transformations (δκ θ˜ α ) while 
is the Killing spinor of the corresponding background geometry, in our case a constant
spinor. It will be useful to determine Γκ explicitly, in terms of the γ (p) matrix appearing
in our previous discussions of kappa symmetry. Due to the fact that θ˜ = θ˜ t C, where C is
the antisymmetric charge conjugation matrix, it is straightforward to derive the type IIA
relation
"
#

1
/ 2`+1 (Γ11 )`+1 F
−1 (p)t
e
C=√
(5.2)
Γκ = C γ
(2` + 1)!
− det(G + F )
`=0

p+1

whereas for type IIB D(p−1)-brane, it reads as
"
#
0 2`
X (Π)
1
/
0
(τ3 )`+1 τ1 eF
.
Γκ0 0 = √
− det(G 0 + F 0 ) `=0 (2`)!

(5.3)

p

5.1. BPS equations and susy projectors
It will be proved that the kappa symmetry preserving condition (5.1), when projected
into the subspace A and applying a T-duality transformation on the background (em a , bmn ),
dynamical fields (φ i ) and Killing spinor (), is correctly mapped into the corresponding
kappa symmetry preserving condition in the T-dual theory Γκ 0  0 =  0 . Consider equation
(5.1) and split it into its different chiral components
"
#
X (−1)`+1Π
1
/ 2`+1 F
e 
2 , 
1 = √
− det(G + F ) `=0 (2` + 1)!
p+1
"
#
X Π
1
/ 2`+1 F
e 
1 .
(5.4) 
2 = √
− det(G + F ) `=0 (2` + 1)!
p+1

Using the T-duality relations 
1 = a2 20 , 2 = −a1 Γz 10 ,

(5.5)

and the T-duality transformation properties of the matrix Γκ , Eqs. (5.4) can be rewritten as
" 
# 

X
a1
/ 0 )2` F 0 0
0
` (Π
e 
1 ,
(5.6)
(−1)
a2 2 = √
(2`)!
− det(G 0 + F 0 ) `=0
p
" 
# 

X (Π
a2
/ 0 )2` F 0 0
0
e 
2 ,
(5.7)
−a1 1 = √
(2`)!
− det(G 0 + F 0 )
`=0

p

which are combined to give the final result
" 
# 

0 2`
X (Π
1
/
)
0
(τ3 )`+1 τ1 eF 
0 = Γκ0 0  0 . 
0 = √
(2`)!
− det(G 0 + F 0 )
`=0

(5.8)

p

It is important to stress that the latter proof applies to any configuration solving the kappa
symmetry preserving condition in the subspace A. Since any solution to such a condition

232

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

involves a set of BPS equations and a set of supersymmetry projection conditions, we
conclude that both sets of equations are mapped to the corresponding BPS equations and
supersymmetry projection conditions under T-duality, thus generating a supersymmetric
world volume soliton for the T-dual theory. This is nothing but the same phenomena
observed in supergravity theories describing the low energy dynamics of the massless
closed string spectrum. There, T-duality is a generating solution transformation. The above
proof, which will be examined in particular examples in next subsections, ensures the same
generating character for the low energy dynamics of the massless open string spectrum.
5.2. Hamiltonian analysis
Given any world volume brane theory, and for any bosonic supersymmetric configuration solving (5.1), one can always use its phase space formulation to compute its energy
density [3]. When the world volume theory is defined on a SuperPoincaré background, it
gives us a field theory realization of the SuperPoincaré algebra. Since BPS states are known
to saturate the BPS bound, it must always be possible to write the energy density as a sum
of squares,
X 
2
(5.9)
t i fi (φ j ) ,
E 2 = E02 + Z 2 +
i

for non-threshold BPS states and
X 
2
t i fi (φ j ) ,
E 2 = (E0 + Z)2 +

(5.10)

i

for threshold BPS states. Here E0 stands for the energy density of the vacuum configuration
and fi (φ j ) = 0 stand for the set of BPS equations derived from (5.1). They allow us to
define a natural lower bound on the energy or BPS bound, respectively,
q
(5.11)
E > E02 + Z 2 ,
E > E0 + |Z|

(5.12)

being saturated precisely when fi (φ j ) = 0 are satisfied, thus justifying their qualification
as BPS equations. 7
If our previous analysis was correct, it should be possible to prove that the Hamiltonian
constraint of the original theory is mapped to the T-dual hamiltonian constraint in the Tdual theory. To prove this we will study the phase space description of D-branes in constant
Gmn and bmn backgrounds by setting θ = 0. The phase space formulation is given by [20] 

a
L = Pm x˙ m + E a V˙a + Vt ∂a E a − s a P˜a Π a + E b Fab 

1 
(5.13)
− v P˜ 2 + E a E b gab + Tp2 det Gab + Fab ,
2
7 We have assumed the existence of a single Z charge in the above derivation, but the extension to more general
configurations is straightforward and completely analogous to the BPS bounds derived from a pure algebraic
approach.

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

233

where Gab stands for the world space induced metric. Pm and E a are conjugate momenta
of x m and Va , respectively, and Pm = em a P˜a − E a ∂a x n bmn . When computed in the subspace A,
Π aˆ = ∂aˆ x mˆ emˆ z ,
z


Π aˆ

z

Π ρ = λ,

(5.14)


Πρ

= 0,

(5.15)

P˜aˆ + emˆ z P˜z − E a ∂a x nˆ bmˆ nˆ − E ρ bmz
ˆ

(5.16)

= ∂aˆ x mˆ emˆ aˆ ,

Pmˆ = emˆ

Pz = λP˜z + E ∂aˆ x bmz
ˆ .

m
ˆ

(5.17)

Note that a, b stand for world space indices, while the underlined ones stand for
background tangent space indices.
It is straightforward to derive the T-duality properties of these objects from the rules that
we have already derived in the Lagrangian formulation:

Π aˆ −→ Π 0 aˆ ,

(5.18)

b0ˆ
z
,
Π aˆ −→ ∂aˆ x mˆ mz
λ0

Faρ
ˆ −→

det

(5.19)

η 0z
Π ,
λ0 aˆ

(5.20)

b0ˆ
z
Π0 ˆ
Faˆ bˆ −→ F 0 ˆ + ∂aˆ x mˆ mz
aˆ b
b
λ0  

Gab + Fab = λ2 det Ga0ˆ bˆ + Faˆ0 bˆ .

b0ˆ
z
− Π 0 aˆ ∂bˆ x mˆ mz
,
λ0

(5.21)
(5.22)

R
ˆ L0 = Tp−1 Lˆ 0 . It is
Note that L0 = dρ L. Let us be more specific and rewrite L = Tp L,
clear that under T-duality Lˆ = λLˆ0 . From these considerations, we can derive the following
relation between momenta
λTp nˆ 0
∂ Lˆ
=
Γ P ,
m
ˆ
Tp−1 mˆ nˆ
∂ x˙
λTp 0
Eρ =
P ,
ηTp−1 z˜
λTp 0 aˆ
E
E aˆ =
Tp−1

Pnˆ0 = Tp−1

Pmˆ = Tp

∂ Lˆ 0
∂ x˙ 0nˆ

(5.23)

p ˜0
Paˆ .
from which we can derive that P˜aˆ = Tp−1
The process of partial gauge fixing (z = ρ) in the Lagrangian formulation corresponds,
in the phase space formulation, to solve the equation

λT

δL
=0
δs ρ

ˆ

H⇒

P˜z =

E b Fbρ
ˆ
λ

.

(5.24)

Using the above information, one can show that the remaining diffeomorphism constraints
(δL/δs aˆ = 0) and the hamiltonian constraint (δL/δv = 0) are mapped to the corresponding
λTp
v,
T-dual constraints, by defining v 0 = Tp−1

234

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

δL0
δL0
δL
δL
−→
−→
,
,
δv
δv 0
δs aˆ
δs 0aˆ
thus indeed proving our initial claim.

(5.25)

5.3. Examples
The aim of this subsection is to show, explicitly, how the BPS equations and
supersymmetry projection conditions characterizing world volume solitons are mapped
into the corresponding ones under T-duality. We will, first of all, concentrate on orthogonal
BIon solutions that are common to all Dp-branes and on dyons in a D3-brane propagating
in SuperPoincaré background (bmn = 0). After that, we consider the more subtle effect
of T-duality on tilted BIons (bmn 6= 0). In the following we will be using the explicit
n which can always be done.
parametrization η = a2 = −a1 = 1 and Γmn = δm
5.3.1. BIons and dyons
Let us start our discussion with BIons. That is, we will look for classical solutions to
the Dp-brane equations of motion propagating in Minkowski space, corresponding to a
fundamental string ending on the brane. The latter configuration is known to be described
by the ansatz [1]
xµ = σ µ,

x p+1 = y(σ a ),

V0 = V0 (σ a ) ,

(5.26)

µ = 0, . . . , p, a = 1, . . . , p, the rest of bosonic scalar fields being constant and the
magnetic components of the gauge field have a pure gauge configuration, corresponding to
the array of branes
Dp : 1 2 . . p _ _ _ _
F1 : _ _ _ _ _ y _ _ _

probe,
soliton.

(5.27)

The solution to the kappa symmetry preserving condition when (5.26) is satisfied, is given
by
P1  = ,

(5.28)

P2  = ,

(5.29)

F0a = ∂a y,

(5.30)

where  is a constant Killing spinor. The first two conditions (5.28), (5.29) correspond
to the supersymmetry projection conditions telling us that we are describing a Dp-brane
and a fundamental string, P2 = Γ0y Γ11 in type IIA and P2 = Γ0y τ3 in type IIB, while
Eq. (5.30) is the usual BPS equation, which by using the Gauss’ law (∂a E a = ∂a δ ab F0b =
0) determines the harmonic character of the excited transverse scalar (δ ab ∂a ∂b y = 0).
Let us study the effect of T-duality along the world volume direction ρ = p. If, as
suggested by our analysis, we apply the partial gauge fixing plus functional truncation
on (5.30), the only non-trivial equation that we get is the corresponding BPS equation in
the T-dual description
0
= ∂a y 0 .
F0a

(5.31)

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

235

Concerning supersymmetry projections, take Eq. (5.28) for a D4-brane. In that case,
P1 = Γ01234 Γ11 and Eq. (5.28) is equivalent to
Γ01234 2 = −1 ,
Γ01234 1 = 2 ,

(5.32)

which when written in terms of the T-dual Killing spinors  0 look as
Γ0123 10 = −20 ,
Γ0123 20 = 10 ,

(5.33)

which is consistent with the projection Γ0123 iτ2  0 =  0 satisfied by a D3-brane. Concerning
the soliton projection (5.29), it can be splitted into
Γ0y 1 = −1 ,

(5.34)

Γ0y 2 = 2 ,

(5.35)

which are equivalent to
Γ0y 20 = 20 ,

(5.36)

Γ0y 10 = −10 ,

(5.37)

respectively. Eq. (5.37) can be joined into Γ0y τ3  0 = − 0 , 8 which is the soliton projection
for a fundamental string in type IIB. Analogous discussion applies for other values of p.
To sum up, we have indeed shown that BPS equations and susy projections are mapped,
under T-duality, to the corresponding BPS equations and susy projections describing the
T-dual configuration
D(p − 1) : 1 2 . p − 1 _ _ _ _ _
F1 :
_ __ _ _y _ __

probe,
soliton.

(5.38)

We would like to comment on the explicit solution of the harmonic equation δ ab ∂a ∂b y =
0. Of course, we could just restrict ourselves to a particular solution of this equation
independent on the world volume coordinate ρ along which we are T-dualizing, to be
consistent with the functional truncation we were discussing in previous sections. Another
possibility is to consider a superposition of BIons of the same mass and charge, located
periodically along the ρˆ axis with period a = 2πR [36],
X
1
, p > 3,
y = kp
|σ − na ρ|
ˆ p−2
n∈Z
X
log |σ − na ρ|,
ˆ
p = 2.
y = k2
n∈Z
In the limit R → 0, which is the one we have been studying along the whole paper, the
discrete sum is replaced by an integral,
8 The minus sign is related with the freedom of choosing as a BPS equation F 0 = −∂ y 0 , instead of (5.31).
a
0a

236

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

X
n∈Z

X
n∈Z

X

kp
−→
|σ − na ρ|
ˆ p−2
k3
−→
|σ − na ρ|
ˆ

Z∞

−∞
Z∞

−∞

(σˆ 2

kp dρ
k˜p
=
,
(σˆ 2 + ρ 2 )(p−2)/2 σˆ p−3
k3 dρ
= k˜2 log |σˆ |,
+ ρ 2 )1/2

Z∞
k2 log |σ − na ρ|
ˆ −→

n∈Z

p > 4,

p = 3,

q
k2 dρ log | σ 2 + ρ 2 | = k˜1 σ1 ,

p = 2,

−∞

which is effectively equal to ignoring all the heavy modes along the ρˆ direction, giving the
correct functional behaviours in the T-dual theory.
Let us now describe the effect of T-duality on dyons. We will look for classical solutions
to the D3-brane equations of motion propagating in Minkowski space, corresponding to a
(p, q) string ending on the brane. The latter configuration is known to be described by the
ansatz [3]
xµ = σ µ,

x p+1 = y(σ a ),

V0 = V0 (σ a ),

Vb = Vb (σ a ),

(5.39)

where µ = 0, . . . , 3, a, b = 1, . . . , 3, and the rest of bosonic fields are being constant,
corresponding to the array of brane
D3 : 1 2 3 _ _ _ _ _ _
F1 : _ _ _ 4 _ _ _ _ _
D1 : _ _ _ 4 _ _ _ _ _

probe,
soliton,
soliton.

(5.40)

The solution to the corresponding kappa symmetry preserving condition is
Γ0123 iτ2  = ,

(5.41)

Γ0y (cos α τ3 + sin α τ1 )  = ,

(5.42)

(5.43)
F0a = cos α ∂a y,
1 abc 
Fbc = sin α δ ab ∂b y.
(5.44)
2
Eqs. (5.41), (5.42) are the supersymmetry projection conditions for this configuration. The
first one describes a D3-brane along directions 123, as expected, while the second one
describes a (p, q)-string along the transverse direction y. Eqs. (5.43), (5.44) are the BPS
equations for this configuration [3].
The longitudinal T-dual configuration is known to be
D2 : 1 2 _ _ _ _ _ _ _
F1 : _ _ _ 4 _ _ _ _ _
D2 : _ _ 3 4 _ _ _ _ _

probe,
soliton,
soliton.

(5.45)

Proceeding as before, the truncated BPS equations one gets are
F00 aˆ = cos α ∂aˆ y 0 ,

(5.46) 

aˆ b ∂bˆ z0 = sin α δ aˆ b ∂bˆ y 0 ,

(5.47)

ˆ

ˆ

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

237

while the supersymmetry projections become
Γ012  0 =  0 , 

−Γ0y Γ11 cos α + Γ0yz sin α  0 =  0 .

(5.48)
(5.49)

Eqs. (5.46)–(5.49) describe a threshold bound state of a D2-brane and a fundamental
IIA string realized on the world volume of the first D2-brane. Note that studying the
particular limit, α = 0 we recover the BIon discussion, while for α = π/2, Eq. (5.47)
is equivalent to the Cauchy–Riemann equations (when written in terms of complex world
volume coordinates and the complex function U = y +iz) describing a D2 ⊥ D2(0), which
is the direct dimensional reduction of the M2 ⊥ M2(0) configuration [3,37].
5.3.2. World volume solitons in constant b fields
We will concentrate on world volume solitons on a D3-brane proving in Minkowski
space and a constant arbitrary bmn field. The kappa symmetry preserving condition looks
as
p 

1
− det(G + F ) =  µ1 ···µ4 γµ1 ···µ4 iτ2 + 6 Fµ1 µ2 γµ3 µ4 τ1 + 3Fµ1 µ2 Fµ3 µ4 iτ2 .
4!
(5.50)
We will describe two different configurations solving Eq. (5.50). First of all, we will
generalize the BPS equations (5.43)–(5.44) describing dyons in the absence of a bmn field.
Using the same ansatz as in (5.39), the solution to (5.50) involves the same supersymmetry
projectors (5.41) and (5.42), while the BPS equations are given by
(5.51)
F0a = cos α ∂a y,
1
(5.52)
B a =  abc Fbc = sin αδ ab ∂b y,
2
which are the straightforward generalization of the usual dyonic BPS equations in the
presence of a bmn field, this being the reason of the appearance of the gauge invariant
tensor F . If we T-dualize along the direction 3, the BPS equations that we obtain are
∂0 z0 = −G003 ,

(5.53)

0
= cos α ∂a y 0 ,
F0a

(5.54)

1 aˆ bˆ 
Faˆ bˆ = 0,
2

(5.55) 

ˆ
0 0
0 
aˆ b ∂bˆ z0 + G0b3
ˆ + ∂bˆ y Gby
ˆ = sin α∂aˆ y ,

(5.56)

where a,
ˆ bˆ = 1, 2. The most remarkable feature of this T-dual configuration is being nonstatic, see Eq. (5.53). Let us discuss in more detail Eqs. (5.51) and (5.52) when the
background is such that only the electric components of the bmn field along the world
volume are non-vanishing (b0a 6= 0), and α = 0, that is, we will be concerned with BIon
type solutions. In this case, Eqs. (5.51) and (5.52) besides the Gauss’ law are easily
integrated to give the solution

238

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

y(σ b ) = yh (σ b ) + da σ a ,

(5.57)

V0 (σ ) = −yh (σ ) + ca σ ,

(5.58)

da + ca = −b0a ,

(5.59)

b

b

a

where yh (σ b ) denotes the harmonic part of the solution whereas da , ca is some set of
constants constrained by (5.59). Notice that da are physical parameters, due to the gauge
invariant character of the excited scalar, determining the tilting of the BIon [38]. In other
words, due to the non-orthogonal character of the BIon, when we study T-duality along the
ρ direction, this is seen as a T-duality at angle from the BIon perspective. As a result of
that, one should expect the T-dual configuration to be one with a tilted BIon ending on a
D2-brane boosted in the direction of dualization, which is what we get from inspection of
Eq. (5.53). To sum up, constant electric field (F0a ) boosts the configuration in the direction
along we T-dualize , just as constant flux of magnetic field on the D-brane (Fab ) is seen as
D-branes at angles in the T-dual picture [9].
Due to the generating character of the T-duality transformation, we would like to
understand how the original static configuration give rise to the non-static solution in the
T-dual theory. First of all, due to the functional truncation defining the A subspace, we will
choose d3 = 0, thus avoiding the linear dependence in σ 3 on the gauge invariant quantity
y(σ b ). By (5.59), c3 = −b03 , or equivalently,
V0 (σ b ) = −yh (σ b ) + caˆ σ aˆ − b03σ 3 .

(5.60)

The latter has also a linear dependence in σ 3 , but this is certainly gauge dependent, since
we can find a gauge parameter c = b03τ σ 3 transforming the gauge field configuration
(5.60) into
V0 (σ b ) = −yh (σ b ) + caˆ σ aˆ ,

V3 (τ ) = b03 τ,

(5.61)

which is explicitly time dependent. The latter is the most natural higher dimensional
solution giving rise to the non-static T-dual configuration. 9
As a second example, we will consider a non-threshold bound state of a D-string inside
the D3-brane together with some BIon, which is generically tilted, due to the non-vanishing
of the b field. Using the same ansatz as in (5.39), the solution to the kappa symmetry
preserving condition is given by 

(5.62)
cos αΓ0123 iτ2 + sin αΓ01 τ1  = ,
Γ04 τ3  = ,

(5.63)

F23 = F = tan α,

(5.64)

F0a = −∂a y,

(5.65)

a = 2, 3,

F01 = ∂1 y = 0.

(5.66)

Eqs. (5.62), (5.64) are a straightforward generalization of the conditions satisfied by any
non-threshold bound state involving a D(p − 2)-brane inside a Dp-brane in the case of
9 J.S. would like to thank David Mateos for discussions related to this point.

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

239

non-vanishing b field, for p = 3. On the other hand, Eqs. (5.63) and (5.65) describe a
tilted BIon, this time being delocalized in the direction where the D-string lies along σ 1
direction, (5.66).
We can study two different T-duality transformations, since there are two inequivalent
world volume directions. Let us study T-duality along direction σ 1 . Proceeding as before,
the BPS equations in the T-dual configuration turn out to be
F 0 = tan α,

(5.67)

0

∂0 z = −G003 ,
F00 aˆ = −∂aˆ y 0 ,

(5.68)
(5.69)

which correspond to a non-threshold D0-D2 bound state with some (tilted) BIon ending on
it, boosted in the compact direction.
Instead, we could have T-dualized along the σ 3 direction. The truncation of the BPS
equations is given by
0
= 0 = ∂1 y 0 ,
F01

∂0 z = −G003 ,
0
= −∂2 y 0 ,
F02
∂2 z0 + G023 + ∂2 y 0 G03y

(5.70)

0

(5.71)
(5.72)
= tan α,

(5.73)

which describe a non-threshold D2-D2 bound state with some (tilted) BIon ending on it,
again, boosted in the compact direction.

6. Arbitrary bosonic background
In previous sections, we showed that in constant Gmn and bmn backgrounds, bosonic
configurations satisfying the kappa symmetry preserving condition (5.1) are mapped under
T-duality to the corresponding bosonic configurations in the T-dual picture. We would like
to extend that proof for an arbitrary bosonic background.
It is well known that D-branes are kappa symmetric whenever the background satisfies
the superspace constraints [25–27]. The structure of kappa symmetry transformations is
always given by the requirements
a

δκ Z M E M = 0,
1
α
α
δκ Z M EM = (1 + Γκ )β κ β ,
2
A

(6.1)
(6.2)

where EM are the different components of the supervielbeins, which should be thought of
as power expansions in the fermionic θ fields and
X
1
l+1
γ(2l+1) Γ11
∧ eF (type IIA),
(6.3)
Γκ = √
− det(G + F ) l=0
X
1
γ(2l)τ3l ∧ eF iτ2 (type IIB),
(6.4)
Γκ = √
− det(G + F ) l=0

240

K. Kamimura, J. Simón / Nuclear Physics B 585 (2000) 219–252

where
a

(6.5)
γ(1) = dσ µ γµ = dσ µ ∂µ Z M EM Γa
1
A
B
(6.6)
F = F − dZ M EM ∧ dZ N EN BAB .
2
It is nevertheless true that the condition for any bosonic configuration (θ = 0) to preserve
some supersymmetry is still given by Γκ  = . The reason is that when studying the θ = 0
limit of the supervielbein components [39–41],

α

Em

θ=0 = 0,
Bmn

θ=0 = bmn (x n ).7) Em a . (6.

θ=0 = em a (x n ). so that .

α 1 1 + Γκ .

2 which determines the universal condition .θ=0 β κ β .

Γκ .

8) (6.θ=0  = .  0 0 0 Gmn = G0mn − G0m˜z G0n˜z − bm˜ z bn˜z /Gz˜ z˜ .13) Under T-duality. (6. it can be checked that 0 b Π / =Π /b .9)  being the Killing spinor of the corresponding bosonic supergravity background. α δ κ θ α eα = (6. m m 2 1 (6. (6. (6. but since the T-duality rules for the bosonic sector of the supergravity fields are known [5] Gzz = 1/G0z˜ z˜ .12) (6. bnz = −G0n˜z /G0z˜ z˜ . γ(1) |θ=0 = dx m em a Γa = Π (6. 0 0 Gnz = −bn˜ z /Gz˜ z˜ .11) where b Π / ≡ Γaˆ dx m eamˆ (x n ). Dρ ≡ Π = λ dρ z z + dx m em (x n ).14) 0z F = F + DρΠ . Π 0z = λ0 dz0 + dx 0m em 10 In the following we are explicitly using the parametrization η = a = −a = 1 and Γ n = δ n . 0 (6.  0 0 0 0 0 0 − bm˜ bmn = bmn z Gn˜z − bn˜z Gm˜z /Gz˜ z˜ .16) . as we did previously.15) where 0 z . Γκ |θ=0 depends on the background geometry.10) one can indeed compute the behaviour of Γκ |θ=0 under T-duality. 10 Using the same notation as in previous section / =Π /b + Γz Dρ.

θ1|2 α β f−|+ (θ ). First of all. is mapped to the corresponding T-dual one satisfying Γκ 0 |θ 0 =0  0 =  0 . Arbitrary kappa symmetric backgrounds There has been some recent interest in the open problem related to T-duality in curved kappa symmetric backgrounds [42. the mapping should be T-duality covariant and satisfy δκ Z M E M = 0 −→ δκ 0 Z 0M E 0 M = 0 α 1 1 α α α δκ Z M EM = (1 + Γκ )β κ β −→ δκ 0 Z 0M E 0 M = 1 + Γκ0 0 β κ 0β .1) θ1α Eα = a2 θ 0α2 E 0 α . – Non-abelian D-branes. When trying to extend our approach to this general case. Kamimura. since we have used a Lorentz (gauge) rotation to set ez a = 0. (7. they will involve components of the spin connection. J. the latter should certainly satisfy some constraints. we will concentrate on three subjects: – T-duality realized on D-branes coupled to an arbitrary kappa symmetric background.2) (7.3) .43]. Secondly.15) is known.9) in type IIA. it seems rather natural to demand the relations α β α α α θ2α Eα = −a1 Γz β E 0 α θ 0α1 . (6. (7.1) deserve several remarks. so that when computing the T-duality transformation of the operators coupling to derivatives of these fermionic fields. they are reminiscent of the extension of the kappa symmetry transformations from the SuperPoincaré case to the arbitrary kappa 0α = symmetric background. – Non-BPS D-branes. Γz 2 = 10 .17) where the relation among Killing spinors is given by 1 = 20 . Discussion We would like to finish with some discussion concerning possible natural extensions of the results presented in this paper. since the supervielbeins appearing in both sides of them admit an expansion in the corresponding fermionic fields. the mapping between fermionic fields will be non-constant in general. 2 2 a a (7.K. that is. 7. In particular. 7. Simón / Nuclear Physics B 585 (2000) 219–252 241 Once (6. In other words.18) which is consistent with the transformations found in [42. First of all. it is straightforward to extend the techniques developed in Appendices B and C to show that any bosonic configuration solving (6. Irrespectively of which is the real solution. Eqs. Finally. it is not clear which is the solution to them.43]. it should be such that the T-duality rules for the closed string sector must map the supergravity constraints of type IIA to the ones of type IIB.1. This is equivalent to map the D-brane effective action and its kappa symmetry structure in type IIA to the corresponding ones in type IIB. (6.

would like to thank Eduardo Eyras for a discussion concerning this point. Indeed. . 7. we can not distinguish between 11 p − det(Gµν + Fµν + ∂µ T ∂ν T ) (7. G˜A ) is covariant. In the following.4). both being T-duality covariant due to the covariance µν of G˜S = (G + F )−1(µν) . This requirement does not fix f . it just constraints it. (7. When being concerned about the T-duality properties of (7. J. For instance. . For example. we will briefly comment on the extension of that result to non-abelian SuperD-branes propagating in SuperPoincaré.7) n for arbitrary constant coefficients an . . .5) SWZ = Mp+2 Due to the scalar character of the tachyonic field. T-duality covariance does not fix the effective dynamics of the open string sector by itself.242 K. the approach followed in this paper was used to determine the effect of nontrivial commutators among scalar fields in the non-abelian bosonic generalization of the DBI action. which means that µν µν f (T .4) must be T-duality covariant. . G˜S . G˜A plus a WZ term describing the coupling of the tachyonic scalar field T to the R–R sector [45].3. Kamimura. since the usual DBI square root is. to check that WZ terms (7. G˜S . it is natural to extend the functional truncation (∂ρ φˆı = 0) to it.6) and p − det(G + F ) X µν an G˜S ∂µ T ∂ν T n (7. we will assume a symmetrized 11 J. ∂µ T . See [46] for a discussion of T-duality properties of non-BPS D-brane effective actions. As in [47]. ∂ρ T = 0. using our previous analysis. it is straightforward.5) are indeed T-duality covariant. as they should be. Z C ∧ dT ∧ eF . we do appreciate an important characteristic of the T-duality covariance requirement. . Simón / Nuclear Physics B 585 (2000) 219–252 7. . . Non-abelian D-branes In [47]. .2. Non-BPS D-brane effective actions It has recently been argued that the effective action describing a non-BPS D-brane probing in SuperPoincaré should be splitted into a DBI term [44] Z p µν µν  (7. The main idea there was to assume that the trace over the U (N) gauge group indices was the symmetrized one (again T-duality does not fix this possibility) and study the dimensional reduction of the D9-brane field theory where world volume diffeomorphisms had been gauge fixed (since no covariant version is known for nonabelian D-brane effective actions).S. In this way. ∂µ T . (7.4) Snon-BPS = − dp+1 σ − det(G + F ) f T .

  Eµˆ i = −2iλ¯ Γµˆ [x i . ν} into T-dual world volume ones {µ. λ]. λ] + i[x i . The components of the tensor Eµν = Πµm Πνn ηmn + Fµν (7. GC 1998SGR (CIRIT).13) where we used the same reduction rules as those used in [47]. λ] ηmn − Dµˆ x i . λ] λΓ ¯ n [x j . one must also gauge fix kappa symmetry. − λΓ we can rewrite E j i and its inverse Eik as j j j E j i = δ j i + Qk − δk δ ki = Qk δ ki .18) thus generalizing the result presented in [47]. E ij = δ ij − λΓ (7. λ]δj k Qik = δki + i[x i .11) (7. λ] + i λ¯ Γ m Dµˆ λ λ¯ Γ n [x i . λ] ηmn δj k . Kamimura. + F ) = det A det Q Aµˆ ˆ ν = Eµˆ ˆ ν − Eµˆ Eik E νˆ . J. x j ]δj k − 2iλΓ   ¯ m [x i . λ] ηmn − 2iλΓ ¯ i [x j .14) (7. l Eik = δil Q−1 k .8) ensuring the vanishing of the WZ term.S. x j ]. Simón / Nuclear Physics B 585 (2000) 219–252 243 prescription for the trace and replace all partial derivatives by covariant derivatives. we can now compute the determinant of the ten-dimensional original matrix (notice that det E ij = det Qik ): det(G . λ] ηmn + Dµˆ x i .K. E i µˆ = −2λΓ   ¯ m [x i . θ2 = λ. Acknowledgements The authors would like to thank Joaquim Gomis for valuable discussions. This work was supported in part by AEN98-0431 (CICYT). J. ˆ ν} ˆ and transverse directions denoted by scalars {x i }.15) (7. and we splitted the initial world volume directions {µ. By introducing the matrix ¯ i [x j . . Since the new action includes fermions. 29]. we choose θ1 = 0.12) (7. λ] λΓ ¯ n [x j . (7.   ¯ i Dµˆ λ + i λΓ ¯ m Dµˆ λ λΓ ¯ n [x i . with the addition that Di λ = i[x i . so that we concentrate on the DBI term of the effective action.16) In this way. Following [28.10) (7.17) (7. is supported by a fellowship from Comissionat per a Universitats i Recerca de la Generalitat de Catalunya.9) can be written after the gauge fixing as   ¯ ¯ m ¯ n Eµˆ ˆ ν = ηµˆ νˆ − 2λΓµˆ Dνˆ λ + Fµˆ νˆ + λΓ Dµˆ λ λΓ Dνˆ λ ηmn . i k (7. (7.

Eq. (A. Simón / Nuclear Physics B 585 (2000) 219–252 Appendix A.2) gives the correct tension for the T-dual D-brane. Eq. J. On the other hand. (A. ˆ = ∂µˆ Vρ + ∂µˆ Z BzN 1 b b 0 b 0 0 0M 0N 0 0N Fµˆ bN b − ∂[µˆ z ∂νˆ ] Z BzN b. Proof of T-duality covariance In this appendix.2) which is derived from operators involving no derivative of the dynamical fields {φ 0ˆı }.6) Gµρ Fµˆ ˆν + ˆ Fνˆ ρ − Fµρ ˆ Gνˆ ρ = Fµˆ ˆ ν. Kamimura. ˆ ν = ∂[µˆ Vνˆ ] − 2 ∂[µˆ Z ∂νˆ ] Z BM (A.10) . R ≡ dρ Gzz Tp−1 (A. ˆ = ∂µˆ Z GMz b b Gµˆ νˆ = ∂µˆ Z M ∂νˆ Z N (−)MN GM bN b.244 K. (A. (A.3) are further split into their symmetric and antisymmetric parts  1 0 (A. and   1 (A. (−)MN = 1 for others.3) Gµρ Gνˆ ρ − Fνˆ ρ = Gµ0ˆ νˆ + Fµ0ˆ νˆ Gµˆ νˆ + Fµˆ ˆν − ˆ + Fµρ ˆ Gρρ from operators involving such derivatives {∂µˆ φ 0ˆı }.9) (A.8) (−)MN = −1 when both M and N are odd. while the one on the IIB D(p−1)-brane is just b b b 0M 0N MN 0 GM Gµ0ˆ νˆ = ∂µˆ z0 ∂νˆ z0 G0zz + ∂(µˆ Z 0M ∂νˆ ) z0 G0Mz b + ∂µˆ Z ∂νˆ Z (−) bN b.1)–(3.4) φ 0 = φ − log |Gzz |.4) defining Tk . Notice that it is equivalent to the usual T-duality transformation for the dilaton field 1 (A.1) −Tp dp+1 σ − det(G + F ) −→ −Tp−1 Tk For this mapping to be satisfied. it is sufficient to hold Z p 0 = Tp R. we will analyze the constraints derived from requiring T-duality covariance of the DBI term Z Z p p 0 dp σ − det (G 0 + F 0 ). Gρρ  1 0 (A. Gρρ The induced metric on the IIA Dp-brane is given by Gρρ = Gzz . 2 when the latter is constant. (A. b M Gµρ b .5) Gµρ Gµˆ νˆ − ˆ Gνˆ ρ − Fµρ ˆ Fνˆ ρ = Gµˆ νˆ .7) where we took into account the conditions (3. the components of the gauge invariant tensor F /(F 0 ) on the IIA/IIB D-branes can be decomposed as 1 b b M N Fµˆ bN b. ˆ ν = ∂[µˆ Vνˆ ] − ∂[µˆ Z ∂νˆ ] Z BM 2 b N Fµρ b.

jβ = θ˜ i Γa α θ˜ j Γ a β .17) In type IIB.11)–(A. We can read the components of the superspace metric GMN = eM a eN b ηab in type IIA from (2.6). They can be interpreted as the generalization of the usual bosonic T-duality rules for the kind of superfields we are considering along the whole paper. we find η2 = G0zz . Simón / Nuclear Physics B 585 (2000) 219–252 245 Using these decompositions in (A.12) (A.16) and those of the NS–NS superfield BMN from (2.11)–(A.18) . Gzz N z  1  b0 b0 M (M+M 0 )N ΓM ΓNbN G0M GMz GM bN b− b GN bz − BzM b b BzN b = (−) b0 . we will start analyzing Eqs.   Giα. (A. we obtain −η b 0 N G b = ΓM bz . G0mn = em   0 a 0 0 0 a ≡ θ˜ 0i Γa α em .m = (θ˜ i Γa )α eα α em a ≡ (θ˜ i Γa )α em a .14)   1  b0 b0 M (M+M 0 )N ΓM ΓNb N B 0 M (−)MN GMz b BzN bz = (−) b b0 . b − BzM b GN b0 N Gzz (A.13) Proceeding in the same way with (A.K. Kamimura. Before doing so.9) Bmn = bmn .m = −(−1)i θ˜ i Γa α em a .15) are the set of constraints derived from the T-duality covariance requirement. α m (A.jβ = θ˜ 0i Γa α θ˜ 0j Γ a β .15) Eqs.7) Gmn = em a en b ηab . b BN Gzz Mz BM bN b− (A. Bmn 0 = (−1)i θ˜ 0i Γa Biα.6). b0 N Gzz (A.m = θ˜ i Γa α eα0 α em   G0iα. In the following. Giα. Giα.jβ = (−1)i θ˜ i Γa α θ˜ j Γ a β . (2. and matching the coefficients of the different independent operators {∂µˆ φ 0 ıˆ ∂νˆ φ 0 ˆ } appearing in both sides. J. (A.11) (A. we must identify the different components of the superfields appearing in them. 0 as while the NS–NS superfield BMN 0 0 = bmn .   Biα.  Biα. Gzz −η b B b = ΓNbM G0zM b.5).m  e0 a . (A.15). G0MN is decomposed as 0 a 0b en ηab . (A.

for aˆ 6= z components that   s θ˜ i Γaˆ α = θ˜ 0j Γaˆ β Γiα jβ . (A.27) (A. λ s θ˜ i Γz where s ≡ take  −η λλ0 α semz = Γm n = −(−1)i Γiα jβ θ˜ 0j Γz  β 0 bnz . Kamimura. in addition to (A. (A. (A.30) 0 n0 . (A. we can θ˜ 1 = a2 θ˜ 02 .246 K.29) and (A.15) requires b[mz 0 0 0 en]z = Γm m Γn n bm (A. T-duality covariance of the DBI action fixes the chirality change mapping among spinor fields (A. G0z0 z0 Gzz = 1. λ0 (A. (A. It follows.20) where λ and λ0 are constants such that G0zz = λ0 . a2 ) from the fermionic components of the background superfields.22). (A.jβ = −(−1)i θ˜ 0i Γa  α θ˜ 0j Γ a  β .11) becomes η2 = λ2 λ0 .13) allow us to set some elements of ΓM N to zero Γm jβ = Γjβ m = Γ1β 1α = Γ2β 2α = 0 (A. bmn − .21) Eq.23) and s bmz 0 = Γm n enz . (A. ez0 a = λ0 δz a . 9) rotation in both IIA/IIB tangent spaces to set ez a = λδz a . and reproduces the well known transformations for their bosonic components (A.14) we get 0 0 aˆ em aˆ = sΓ m m em0 . (A. (A. (A.24) . 2 Gzz = λ2 . −sa1 Γz β e α α α = eβ0 β Γ2α 1β .29) Finally (A.25) = ±1 is a signature.12) and (A. (A. (A. Simón / Nuclear Physics B 585 (2000) 219–252 0 Biα. Assuming IIB spinors θj0 have positive chirality. Eqs. as we claimed in the introduction.26) and sa2 eα α = eβ0 α Γ1α 2β .19) It is always possible to make a local SO(1. J. θ˜ 2 = a1 θ˜ 01 Γz .24).25).30).22) which is the analogue of the usual T-duality rules relating the radius of the original circle with the radius of the T-dual circle. λ Thus.28) From Eq. 2 (A.26) up to constant factors (a1 .

Furthermore. we will next study the T-duality properties of /= the different supersymmetric invariant forms defined on D-branes.40) which are the form version of the identities (3.31)  b Π / ≡ Γaˆ dx m eamˆ + θ˜ Γ aˆ dθ˜ Dρ ≡ Π z = λ dρ z + dx m em (A. (A.26) and (A. Let us consider Π Γa Π a in type IIA. where (A.36) + a22 θ˜ 02 Γ aˆ dθ˜20 + a12 θ˜ 01 Γ aˆ dθ˜10 . (A. (A. Concerning the supersymmetric invariant two form F    1 1 F = dV + θ˜ Γ11 Γm dθ˜ dx˜ m + θ˜ Γ m dθ˜ − dx m dx n bmn . a12 = 1. in type IIB.35) b / in terms of type IIB variables by inserting (A. Simón / Nuclear Physics B 585 (2000) 219–252 247 Having solved such T-duality constraints. J.41) (A. Kamimura.39) θ˜ Γ11 Γ z dθ˜ = −θ˜0 Γ z dθ˜ 0 .K.38) The latter equation is telling us that the supersymmetric invariant one form in nine dimensions is T-duality covariant.26).  z Π 0z ≡ λ0 dz0 + dx 0m e0 m + θ˜ 0 Γ z dθ˜ 0 . a22 = 1 (A.33) Analogously.20). (A.32)   0 b Π / ≡ Γaˆ dx m eamˆ + θ˜ Γ aˆ dθ˜ = Γaˆ dx m sΓmm e0 amˆ 0  (A.42) . θ˜ Γ11 Γ aˆ dθ˜ = +θ˜ 0 τ3 Γ aˆ dθ˜ 0 . 2 2 it can be written in terms of type IIB variables as (A. by choosing s = +1. and ˜ z dθ˜ = λdρ + dx 0m Dρ = λ dρ + dx m em + θΓ z 0 bmz λ0 − θ˜ 0 τ3 Γ z dθ˜ 0 .28) and (A.34)  0 b Π / ≡ Γaˆ dx 0m e0 maˆ + θ˜0 Γ aˆ dθ˜ 0 .32) + θ˜ Γ z dθ˜ .37) we can identify both one forms  b0 b Π / = Γaˆ dx 0m e0maˆ + θ˜ 0 Γ aˆ dθ˜ 0 = Π /.28). when using (A. Thus. It can be splitted in terms of Π / =Π /b + Γz Dρ. θ˜ Γ z dθ˜ = −θ˜ 0 τ3 Γ z dθ˜ 0 . 0 b Π / 0 =Π / + Γz Π 0z . where (A. it follows that θ˜ Γ aˆ dθ˜ = +θ˜ 0 Γ aˆ dθ˜ 0 .29) into We can write Π (A. (A.37) in (A. (A.

X F n X (F 0 + DρΠ 0z )n  0 = = 1 + DρΠ 0z eF . F 0 ≡ dV − Appendix B. Π / = Γa dx˜ a + θΓ (B. whose relevance may become more clear in the discussion of the T-duality transformation of the WZ term in Appendix B. E β = dθ˜ α Cαβ . all the dependence of dρ in supersymmetric invariant forms Π Dρ = Π z .3) eF = n! n! n=0 n=0 b / and Γ z Dρ are commuting one forms. WZ terms of IIA D-branes are obtained from F dLWZ A = −Tp ECA Ee .38) besides that Π ECA E = dθ˜ X `=0 (Γ11 )`+1 b X Π / 2` ˜ (Π / + Γ z Dρ)2` ˜ dθ = dθ˜ dθ (Γ11 )`+1 (2`!) (2`!) `=0 b b Π / 2` ˜ Π / 2` ˜ dθ1 + dθ˜ 1 (−1)`+1 dθ2 = dθ˜ 2 (2`)! (2`)! b Π /b2`+1 ˜ Π / 2`+1 ˜ dθ1 + dθ˜1 (Γ z Dρ)(−1)` dθ2 +dθ˜ 2 (Γ z Dρ) (2` + 1)! (2` + 1)!  b b Π / 2` ˜ 0 Π / 2` dθ2 + dθ˜ 02 (−1)`+1 Γz dθ˜10 = a1 a2 dθ˜ 01 Γz (2`)! (2`)! .28) and (A. the supersymmetric one form along the original circle.2) The latter show that dLWZ A just depends on supersymmetric invariant forms. can be written as Π 0z . Simón / Nuclear Physics B 585 (2000) 219–252 F = F 0 + DρΠ 0z . F = F − + dρ ∧ i∂ρ F . using the results obtained in Appendix A. what T-duality does is to exchange both forms Π z ←→ Π 0z . Further/ and F is through more.44) 2 2 It is remarkable that all terms i∂ρ F appearing in the decomposition of the supersymmetric invariant form F under the double dimensional ansatz.43) and using Dρ ∧ Dρ = 0. (A. (B. Thus. CA Π / = (Γ11 )`+1 (2`!) (B. the supersymmetric invariant one form along the T-dual circle.248 K.43) where   1 0m 0n 0 1 0 a dx dx bmn + θ˜ 0 τ3 Γa dθ˜ 0 dx 0m em + θ˜ 0 Γ a dθ˜ 0 . T-duality transformation of WZ term In this appendix we prove that WZ terms of type IIA SuperD-branes are mapped to WZ terms of type IIB SuperD-branes under T-duality. This is the supersymmetric generalization of the corresponding phenomena observed in the bosonic case [12]. from (A. whose Tduality properties were determined in Appendix A. using (A. (A.  X Π / 2` . J. Next.1) `=0 where E α = dθ˜ α = dθ α eα α .  ˜ a dθ˜ . Kamimura. In particular.

26). J.26)  (C. WZ From dLWZ A in (B.5) WZ contributes to LWZ B .K. Simón / Nuclear Physics B 585 (2000) 219–252 249  b Π /b2`+1 ˜ 0 / 2`+1 ˜ 0 0 ` Π ˜ dθ + dθ 2 Dρ(−1) dθ (2` + 1)! 2 (2` + 1)! 1   Π /b2` Π /b2`+1 ` ˜ 0 Γz τ3` τ1 dθ˜ 0 + dθ˜ 0 Dρ τ3 τ1 dθ . Kamimura. The second term in the right-hand side of (C. ¯ + Γ = κ¯ 2 Γ κΓ ¯ Γ− = κΓ z z (C. (C.1) is (C.3) z ≡ a1 κ¯ 10 .4) τ3` τ1 ˜0 (B. = a1 a2 Tp dθ SB (Π = −a1 a2 Tp (Dρ) dθ # `=0 e ˜0 (B.2) Γ Γ− . by taking the p + 1 form part of LA on σ . Kappa symmetry In this appendix we prove that the infinitesimal kappa symmetry transformation δκ θ in type IIA is mapped to δκ 0 θ 0 in type IIB as claimed in (4. The kappa symmetry transformations for type IIB spinors are obtained from those of IIA using (A.3) and (B.1) a2 δ θ˜ 02 = δ θ˜ 1 = δ θ˜ Γ− = κ¯ 1 − γ (p) Γ−  z 0 z z z (p) (C.6) B = −Tp−1 E SB Π 0 is given in (A.5) .5) gives dLB  0   0 0 dLWZ / 0 τ1 E 0 e F .1) and (C.5) where dots stand for terms not depending on dρ.4) where we assumed that kappa symmetry parameters κj have the same T-duality transformation as the dynamical fields θ in (A. if a1 a2 = −1. The coefficient of dρ in (B. = a1 a2 dθ˜ 0 (2`)! (2` + 1)! + dθ˜ 01 Dρ Joining (B.4) dLWZ A ˜0 = −Tp a1 a2 dθ " b XΠ / 0 2` (2`!) Γ z τ3` τ1 `=0  F 0 0z ˜0 × dθ 1 + DρΠ " b X Π / 0 2`+1 ` τ τ1 + (Dρ) (2` + 1)! 3 X (Π / 0 )2`+1 # 0 dθ eF + · · · (2` + 1)! `=0  0  0 0 ˜ / )τ1 dθ˜ 0 eF (λ dρ) + · · · .2). written in terms of IIB WZ µ variables.2) are κΓ ¯ − = κ¯ 1 ≡ a2 κ¯ 20 . (B. It means that only the coefficient of dρ in (B. κ¯ 1 = a2 κ¯ 20 . IIB (p − 1) brane WZ term will be obtained by integrating it over ρ. where Tp−1 Appendix C.5) we can find the WZ Lagrangian LA . a1 δ θ˜ 1 = δ θ˜ 2 Γ = δ θ˜ Γ+ Γ = δ θ˜ Γ Γ− = κ¯ 1 − γ First terms in the right-hand side of (C. κ¯ 2 = a1 κ¯ 10 Γ z .15).

Simón / Nuclear Physics B 585 (2000) 219–252    −κ¯ SA Π −κ(γ ¯ (p) )Γ− = √ / eF p+1 Γ− − det(G + F ) " # X Π −κ¯ / 2`+1 F e =√ Γ− − det(G + F ) `=0 (2` + 1)! p+1 " # b X (Π  −κ¯ / + Γ z Dρ)2`+1 0 1 + DρΠz0 eF =√ Γ− (2` + 1)! − det(G + F ) `=0 p+1 "   X Π −κ¯ /b2`+1 Π /b2` z + (Γ Dρ) =√ − det(G + F ) `=0 (2` + 1)! (2`)! #  F0 0 Γ− + · · · × 1 − Πz Dρ e p+1 " b X (Π / 0 + Πz0 Γ z )2` −κ¯ 0 z DρeF Γ = √ 0 0 (2`)! λ − det(G + F ) `=0 "   # X (Π −κ¯ 2 / 0 )2` F 0 z e Γ =√ Γ− (2`)! − det(G 0 + F 0 ) `=0 p "   # X (Π −a1 κ¯ 10 / 0 )2` F 0 =√ Γ− . We have also used the relation of DBI term p p − det(G + F ) = λ − det(G 0 + F 0 ). In the last second line p + 1 form coefficient is replaced with p form coefficient (the coefficient of dσ 0 dσ 1 · · · dσ p−1 ) by dropping dρ. J.250 K. the coefficient of dσ 0 dσ 1 · · · dσ p after taking the pullback.7) Analogously for the second term in (C.6) p Here [· · ·]p+1 means p + 1 form coefficient of [· · ·]. (C. e (2`)! − det(G 0 + F 0 ) `=0 # Γ− p+1 (C.2).    −κ¯ SA Π / eF p+1 Γ z Γ− − det(G + F ) " # X Π −κ¯ / 2`+1 F e =√ Γ z (−1)`+1 Γ− − det(G + F ) `=0 (2` + 1)! p+1 " b X −κ¯ (Π / + Γ z Dρ)2`+1 =√ (−1)`+1 (2` + 1)! − det(G + F ) `=0 #  0 Γ− × 1 + DρΠz0 eF Γ z −κ(γ ¯ (p) )Γ z Γ− = √ −κ¯ 1 =√ − det(G 0 + F 0 ) " X `=0 p+1 (−1)`+1   # / 0 )2` F 0 (Π e Γ− (2`)! p . Kamimura.

Ortín. Mod. Tasi lectures on D-branes. Callan.R. Simón / Nuclear Physics B 585 (2000) 219–252 =√ " X −a2 κ¯ 20 − det(G 0 + F 0 ) `=0 (−1) / `+1 (Π 0 2` ) (2`)! 251 # e F0 Γ− .M. Bergshoeff. Phys. B 451 (1995) 547. B 514 (1998) 603. Bergshoeff. R. Lett. Green. Seiberg. M. E. Phys. J. Phys. Nucl. E. δ θ˜ 2 = κ¯ 2 − √ a2 − det(G 0 + F 0 ) (2`)! `=0 Since as a12 = a22 (C. Horava. B 456 (1995) 130. Phys. B 231 (1989) 251. C. Aganagic. P. Lett. (Eds. 1999. Phys. A. Townsend.8) p Thus. Strathdee. joining the partial results " # 0 2` 0 X κ ¯ a ( Π / ) 0 2 2 eF (−1)`+1 Γ− . J. B 502 (1997) 149. 141. Kluwer Academic. N. Bergshoeff. JHEP 9802 (1998) 003. B 380 (1996) 265. D 51 (1995) 3009. Nucl. E. Hull. P. Phys. G. J. Dine.G. Strings. Phys. Schwarz. Nucl. R. in: L. Witten.H. B 322 (1989) 301. P. Polchinski. Gibbons. M. Simón. Nucl.G. Phys. Gauntlett. C. Kamimura. Townsend. Ortín. Gomis. J. R. (C. JHEP 9909 (1999) 032. Becker. J. Douglas. de Roo. Lett. B 513 (1998) 198.K.B. Branes and Dualities.13) γ 0(p−1) = √ − det(G0 + F 0 ) References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] C. Polchinski. J. Strominger. Polchinski. Nucl. Kallosh. Hull. J. hep-th/9712004. E. P. N. B 531 (1998) 226. E. Phys. J. G. JHEP 9801 (1998) 003. Phys.K. A 4 (1989) 2073.11) p Thus we have shown that the kappa symmetry transformation of IIA spinor δκ θ˜ is mapped to that of IIB spinor δκ 0 θ˜ 0 under Tk . T. . Townsend. Phys. Bergshoeff. Townsend. B 495 (1997) 99. A. δ θ˜ 01 = κ¯ 10 − √ a1 − det(G 0 + F 0 ) (2`)! `=0 p "  # 0 2`  X (Π κ¯ 10 a1 / ) 0 0 0 F e Γ− . Rev. Nucl. Connes. Huet. Mod. Phys. Kallosh.    −1 0 b CB Π / 0 τ1 e F p . Leigh. Popescu. Nucl. Maldacena.). Ortín. Bergshoeff.M. P. M.K. Lett. Cargèse 1997. Rev.  (C. P. J. δ θ = κ¯ + √ (2`)! − det(G 0 + F 0 ) `=0 ˜0 (C. E. J. C.10) p = −a1 a2 = 1 we get a1 /a2 = −1. K. We can express them by using τ matrices " # b X κ¯ 0 / 0 )2` F 0 `+1 (Π e (τ3 ) τ1 Γ − . J. Phys. M.9) 0 (C. Papadopoulos. (C. M. A 2 (1987) 273. A. Baulieu et al. Lett. M. D 61 (2000) 047702. Schwarz. 75 (1995) 4724. Becker. p. T. Int. T. J.12) δ θ˜ 0 = κ¯ 0 1 − γ 0(p−1) . Phys. Dai. M-theory from its superalgebra.K. Phys. Phys. Rev.K. B 382 (1996) 65. hep-th/9611050. Nucl. Seiberg.

JHEP 9907 (1999) 016. Rev. Phys. Branes within branes. Perry. P. [43] S. Codirla. Math.C. [32] F. P.K. Gomis. Douglas. JHEP 0005 (2000) 009. Simón. Westerberg.W. Popescu. Billo.W. Phys. B 496 (1997) 215. [30] M. Lett. B 412 (1997) 265. Izquierdo. Stelle. J.E. J. [28] M. J. J. Craps. Phys. Phys. Barnich.P.E. B. Sundell. B 419 (1998) 148.A. Lü. M. Cederwall. F. F. [35] G. Phys. Lett. J. Cremmer. B 490 (1997) 179. Nucl. E. Schwarz. Phys. Lett.F. B 393 (1997) 311.H. Roose. [33] F. Phys. Nucl. hep-th/9512077. A. Brandt. B.N. Howe. [27] M. Pope. Nucl. Nucl. B 490 (1997) 145. [47] R.S. M. Henneaux. Brandt. C. [40] L. B. Townsend.K. Aganagic. Hassan. [26] M.K.W. Bergshoeff. [23] D. Phys.C. 174 (1995) 57. [42] M. Plefka. [37] G. Lett. Papadopoulos. de Wit. JHEP 9906 (1999) 033. [31] F. B 91 (1980) 384. P. Phys. B 561 (1999) 43. Rev. Phys. Henneaux. P. A. Cederwall. pp. S. Kamimura. Branes and Dualities. 81 (1998) 1770. [24] T. Nucl. [41] B. Phys. de Roo. 63 (1989) 2443. von Gussich. B 568 (2000) 145. Eyras. J. Lett. B 532 (1998) 99. [46] E. Peeters.A. [44] A. de Azcárraga. Ferrara. Phys. Barnich. Simón / Nuclear Physics B 585 (2000) 219–252 [22] J. 267– 275. Phys. K. Bergshoeff. B 91 (1980) 61. Nucl. Gomis. Nucl. 202 (1999) 593. Westerberg. J. B 490 (1997) 163. von Gussich. Commun. Math.H. [38] K. Cveti˜c. Simón. Commun. Nilsson. Hashimoto. JHEP 9912 (1999) 022. C. Townsend. B 548 (1999) 231.J. Phys. Gomis.R. Phys. Simón. F. B 573 (2000) 149. [36] C. A. Myers. B 443 (1998) 147. Nucl. Brink. Panda.252 K. Aganagic. Phys. G. Townsend. . Brandt. Mateos. 1997. M. A. J. H. de Wit. Local BRST cohomology in gauge theories. in: Strings. K. Sato.M. D. C. Phys. Popescu. J. [39] E. Brandt. [29] M. [45] M. Schwarz. P. Phys. Sorokin. Brandt. T. M. Nucl. JHEP 9910 (1999) 008. Nilsson. J. Lett. hepth/0002245. Cargèse. Lett. Sen. Phys. Gibbons. S. [34] G. J. Lett. [25] E. Gauntlett. J.

Slavnov–Taylor identity 1.  2000 Elsevier Science B. By further integrating out the fields with frequencies up to some lower scale Λ one obtains the so called Wilsonian effective action.e. In particular the breaking of the symmetry can be studied at the physical point Λ = 0 [5] or at the UV scale [4. at least in perturbation theory. In this formulation [1–3] quantum fluctuations at short distances are integrated out to give an effective action for longer distances. 11.F. All rights reserved. In this approach the bare action can be though as the result of integrating out the degrees of freedom with frequencies higher than the ultraviolet (UV) scale Λ0 of the underlying theory (eventually more fundamental). PII: S 0 5 5 0 .Rd Keywords: Renormalization group formalism.Nuclear Physics B 585 (2000) 253–274 www. viale delle Scienze.V. For a gauge theory. when all cutoffs are removed) satisfies the Slavnov–Taylor (ST) identity.-q.V.30.15. gauge theory..infn. PACS: 11.3 .elsevier. This fine-tuning provides some of the boundary conditions of the RG flow.Gh. All rights reserved.3 2 1 3 ( 0 0 ) 0 0 3 8 1 . E-mail address: bonini@pr. [4. To illustrate this method we systematically compute the non-invariant couplings of the ultraviolet action of the SU(2) pure Yang– Mills theory at one-loop order. revised 10 May 2000. Bonini).10.nl/locate/npe Fine-tuning and the Wilson renormalization group M. This problem as been investigated in Refs.it (M. E. Gruppo collegato di Parma. the gauge symmetry is completely broken at the intermediate scale Λ and one has to show that the physical theory (i.N. In this process one needs to introduce a scale which may conflict with symmetries. accepted 17 June 2000 Abstract We use the Wilson renormalization group (RG) formulation to solve the fine-tuning procedure needed in renormalization schemes breaking the gauge symmetry. Introduction The Wilsonian renormalization group (RG) provides the most physical framework to study general properties of quantum field theories. 43100 Parma.N.. 11.8] where it has been proved that by finely tuning the couplings in the Wilsonian effective action the usual ST identity is recovered at Λ = 0. In ∗ Corresponding author. Italy Received 22 February 2000.8]. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Tricarico Università degli Studi di Parma I. Bonini ∗ .

Bonini. In the latter the fine-tuning provides the couplings of the bare action. Dµ c = ∂µ c + g Aµ ∧ c and we have introduced the usual scalar and external SU(2) products. γ ] = SYM + d x uµ Dµ c −  v c c . E. respectively. The conflict between symmetries and regularization is a long-standing problem in quantum field theory: in presence of chiral fermions no consistent regularization is known to preserve chiral symmetry and all possible non-invariant counterterms must be added to the classical action in order to compensate the breaking of the symmetry generated by the regularization. Due to the non-locality of these vertices. which contains all possible relevant interactions. v a associated to the variations of Aaµ . 4 2 where the gauge stress tensor and the covariant derivative are given by Fµν = ∂µ Aν − ∂ν Aµ + g Aµ ∧ Aν . Although for non-anomalous theories this is an algebraic solvable problem. The paper is organized as follows. Technical details of this calculation are given in Appendix A. In this paper we will use the RG formulation and the fine-tuning at the UV scale to provide all non-invariant counterterms of the bare action at the one-loop order.254 M. after introducing some notation. Tricarico / Nuclear Physics B 585 (2000) 253–274 the former case one enforces the relevant part of the usual ST identity and determines the boundary conditions at the physical point Λ = 0 of the relevant non-invariant couplings of the Wilsonian effective action in terms of physical vertices evaluated at the subtraction point. 2. we give a brief description of the RG method for the gauge case. The inclusion of chiral fermions can be done following the same lines [8. [5] where the boundary conditions of six relevant non-invariant couplings of a pure gauge SU(2) theory at the first loop order were found in this way). ca . g δca = − 12 η  abc cb cc . In order to keep things as simple as possible we will restrict to a pure gauge SU(2) theory. In Section 5 we present some concluding remarks and hints for the generalization of this computation to higher loops.9]. 1 δ c¯a = − η ∂ µ Aaµ g with η a Grassmann parameter. the determination of these counterterms is quite difficult in practice. In Section 4 we determine the non-invariant couplings of the bare action at one-loop order by solving the fine-tuning at Λ = Λ0 . the fine-tuning is analytically difficult to solve (see Ref. In Section 2. In Section 3 we recall the effective ST identity for this theory. Renormalization group flow Consider the SU(2) gauge theory described by the classical Lagrangian (in the Feynman gauge)   Z 1 1 µ 2 4 µν µ SYM = d x − Fµν · F − ∂ Aµ − c¯ · ∂ Dµ c . g 2 . This action is invariant under the BRS transformation [10] δAaµ = 1 η Dµab cb . Introducing the sources uaµ . one has the BRS action [10]   Z 1 a ab b 1 abc a b c 4 (1) SBRS [Φ.

D −1 Φ)ΛΛ0 − (Φ. 2 2 µ p Z Z ≡ d 4p . γ . Λ. Notice that at Λ = Λ0 the Wilsonian effective action coincides with the bare action. In (2) we have also introduced the cutoff scalar products between fields and sources   Z 1 −1 −1 2 1 a a a a (Φ D Φ)ΛΛ0 ≡ KΛΛ0 (p) p A (−p) Aµ (p) − c¯ (−p) c (p) . Λ. According to Wilson [1–3] the fields with frequencies Λ2 < p2 < Λ20 are integrated in the path integral giving Z  Z[J. γ . p2 . γ ] and the functional Seff is the Wilsonian effective action which is the generator of the connected amputated cutoff Green functions (except the tree-level two-point function). Λ0 ] + (Φ. 2 2 (4) . (3) Λ∂Λ Π[Φ. v a }. Λ. Λ. The flow equation can be written as [11–14]   2 Z  Λ∂D −1 (q)  i δ Γ [Φ. Λ. ca . Λ0 ] −1 ΛΛ0 (−1)δA . (2π)4 p Z (J Φ)ΛΛ0 ≡ p    i −1 a a a a p KΛΛ (p) j (−p) A (p) + χ ¯ (−p) − u (−p) ca (p) µ µ µ µ 0 g  + c¯a (−p) χ a (p) . Λ20 and rapidly vanishes outside. Λ0] contributes to the quadratic part of Z[J. γ ] = N[J . Λ. γ . Λ. which is the generator of 1PI cutoff vertex functions and reduces to the physical quantum effective action in the limits Λ → 0 and Λ0 → ∞ . γ ] is independent of the infrared cutoff Λ can be translated into a flow equation for Seff . Λ. γi = {wµa .M. Λ0 ] = − 2 ∂Λ δΦA (q)δΦB (−q) AB q where δA = 1 if ΦA is a fermionic field and 0 otherwise and 1 1 Π[Φ. Seff contains cutoff propagators DΛΛ0 given by the free propagators D(p) multiplied by a cutoff function KΛΛ0 which is one for Λ2 . Λ0] DΦ exp i − 12 (Φ D −1 Φ)0Λ + (J Φ)0Λ + Seff [Φ. The requirement that Z[J. D −1 Φ)0Λ0 . c¯a }. and wµa = uaµ + g∂µ c¯a (no source is introduced for c¯a ). E. Bonini. γ . Tricarico / Nuclear Physics B 585 (2000) 253–274 255 where we have denoted by Φ and γ the fields and the BRS sources ΦI = { Aaµ . Λ0 ]. referred to as the exact RG equation. γ . Λ0 ] . γ . (2) where N[J . Λ0 ] = Γ [Φ. For our purposes it is convenient to perform a Legendre transform on Seff in order to obtain the so-called “cutoff effective action” Γ [Φ.

thus irrelevant vertices must vanish at Λ = Λ0 .5] the symmetry is ensured if the relevant couplings are properly fixed at some value of Λ. For the SU(2) theory there are nine relevant couplings which are the coefficients of the following monomials 1 Aµ (gµν ∂ 2 − ∂µ ∂ν ) · Aν . 1 In the following this limit will be taken and Γ [Φ. In the next section we briefly recall how this procedure can be implemented. In fact the cutoff Λ explicitly breaks gauge invariance and one has to show that at the physical point Λ = 0 and Λ0 → ∞ the Slavnov–Taylor identity can be recovered. Effective ST identity In the RG formulation the ST identity for the physical effective action Γ [Φ. Λ0 ] at Λ = Λ0 becomes the bare action and to ensure perturbative renormalizability it must be local. once this limit has been performed. as shown by Polchinski [3]. 2 1 1 (Aµ ∧ Aν ) · (Aµ ∧ Aν ). Λ. γ . 2 (∂ν Aµ ) · Aµ ∧ Aν . Eq. wµ · c ∧ Aµ . wµ · ∂µ c. Relevant couplings have non-negative mass dimension and are defined as the value of some vertices and their derivatives at a given normalization point (see [5] for their definition). . 2 1 v · c ∧ c. Once the boundary conditions are given. 4 4 (5) To set the boundary conditions. γ .256 M. [5]. Bonini. Λ]. In order to set them we distinguish between relevant couplings and irrelevant vertices according to their mass dimension. γ . (Aµ · Aν ) · (Aµ · Aν ). E. 2 1 (∂µ Aµ )2 . 3. 1 Aµ · Aµ . Λ. the boundary conditions must provide the physical coupling g(µ) and guarantee symmetry. one observes that the cutoff effective action Γ [Φ. (3) can be thought as an alternative definition of the theory which in principle is non-perturbative. ∞] is ensured if the cutoff effective action satisfies a modified ST [4–8]. Tricarico / Nuclear Physics B 585 (2000) 253–274 This equation can be integrated by giving boundary conditions in Λ. for instance. Λ] do not vanish at Λ = Λ0 but are only suppressed by inverse powers of Λ0 . it is convenient to remove from the functional Π given in (4) the gauge fixing term and introduce the Slavnov operator [15]  Z  δ Π0 δ δ δ Π0 0 + . γ . Φi = {Aµ . γ . c} SΠ = −i δΦi (−p) δγi (p) δγi (p) δΦi (−p) p with 1 For a generalization of this proof to gauge theories see. 0. As far as the relevant couplings. On the other hand the iterative solution of (3) reproduces the usual loop expansion and. ∞] will be denoted by Γ [Φ. Λ0 ] is finite for all values of Λ . As usual in studying ST identities. Notice that. γ ] ≡ Γ [Φ. the irrelevant vertices of Γ [Φ. As shown in Refs. γ . the Λ0 → ∞ limit of Γ [Φ. [4. Λ.

or more precisely. E. γ . Λ] + 1 2 257 Z pµ pν Aµ (−p) · Aν (p). Λ] = −h¯ Z p. due to the boundary conditions imposed on the irrelevant vertices of Π . in (6) one can choose Λ much bigger than all external momenta.M. In fact this system of equations overdetermines the UV couplings and in order to prove its solvability one must exploit the so-called consistency conditions which reduce the number of independent equations [4]. Λ] + ∆ˆ Γ (Λ) = 0 . and can be shown by a direct calculation for the functional ∆ˆ Γ (Λ0 ). At Λ = Λ0 the functional ∆Γ is local. γ . γ . One-loop computations At one-loop and at Λ = Λ0 . In the next section we will solve these fine-tuning equations at one-loop order tuning the couplings of the UV action at this order. namely Λ = Λ0 . This is obviously true for the functional SΠ 0 Π 0 (Λ0 ). Besides determining these couplings this calculation provides a direct check of the solvability of the fine-tuning problem. Therefore in the Λ0 → ∞ limit the effective ST identity (6) reduces to a finite set of equations (the so called fine-tuning equations) obtained by requiring the vanishing of relevant part of ∆Γ . Notice that at the physical point Λ = 0 the gauge symmetry condition (6) reduces to the usual ST identity. We will adopt the latter possibility in the present paper. Λ] = Π[Φ. Both these procedures provide the boundary conditions of the RG flow for the non-invariant relevant couplings of the cutoff effective action. γ . (8) . since Π becomes the physical effective action and ∆ˆ Γ vanishes. Λ] δΦL (−p)δΦJ (−q) ! Z 1 δ2 Π[Φ. one gets the boundary conditions for the relevant part of the effective action given in terms of the physical coupling g(µ). If one imposes (6) at Λ = 0. Bonini. its irrelevant contributions vanish as inverse powers of Λ0 and disappear in the Λ0 → ∞ limit (see [7] for a proof). Λ] − × uµ ∂µ c .q   −1 K0Λ (p)(−1)δL DΛ∞ (−p) Li  δ 2 Γ [Φ. Tricarico / Nuclear Physics B 585 (2000) 253–274 Π 0 [Φ. γ . the fine-tuning equation (6) reduces to SΠ (0) Π (1) (Λ0 ) = − ∆ˆ (1) Γ (Λ0 ). 4. δΦJ (q)δγi (p) g −1 (7) x The factor h¯ has been inserted to put into evidence that ∆ˆ Γ vanishes at tree-level. p The modified ST identity then reads ∆Γ (Λ) ≡ SΠ 0 Π 0 [Φ. Alternatively. (6) where ∆ˆ Γ [Φ. and determine the cutoff dependent couplings of UV action. It can be shown that if ∆Γ vanishes up to loop ` then at the next loop ∆Γ is Λ-independent and therefore can be analyzed at every value of Λ. γ .

. The explicit form of Πinv is    Z 1 1 1 wµ · Dµ c − v · c ∧ c .. The values of these couplings corresponding to three different choices of the cutoff function are collected in the tables 1–3 given below. zi (Λ = 0) = 1. E. the l.e. On the contrary the couplings zi are not involved in the finetuning. Πinv (Λ0 ) = d4 x − z1 Fµν · F µν + z2 z3 4 gz3 2 with Fµν = ∂µ Aν − ∂ν Aµ + g z3 Aµ ∧ Aν and the covariant derivative given by Dµ c = e which can be written as ∂µ c + g z3 Aµ ∧ c. σi = σi (g. Tricarico / Nuclear Physics B 585 (2000) 253–274 where the local functional Π (1) (Λ0 ) ≡ Π (1) [Φ. (10) 4 4 The functional Π(Λ0 ) is equal to the UV action SΛ0 [Φ. Notice that the fine-tuning equation (8) allows to compute the couplings in Π (1) (Λ0 ) since (1) ∆ˆ Γ (Λ0 ) depends only on Π (0) = SBRS .h. where 2 For the sake of simplicity the loop order index will be sometimes understood. Π(Λ0 ) = Πinv (Λ0 ) + Π(Λ (9) where Πinv satisfies the ST identity. In the standard language this corresponds to the renormalization prescriptions. of (8) becomes  Z (1) 4 SΠ (0) Π (Λ0 ) = d x δ1 Λ20 Aµ · ∂µ c + δ2 Aµ · ∂ 2 ∂µ c + δ3 Aµ · (∂ 2 Aµ ) ∧ c 1 1 +δ4 Aµ · (∂µ ∂ν Aν ) ∧ c + δ5 (∂µ wµ ) · c ∧ c + δ6 (wµ ∧ Aµ ) · (c ∧ c) 2 2    +δ7 (∂µ Aµ ) · Aν (Aν · c) + δ8 (∂µ Aµ ) · c (Aν · Aν ) + δ9 (∂ν Aµ ) · Aν (Aµ · c)    +δ10 (∂ν Aµ ) · Aµ (Aν · c) + δ11 (∂ν Aµ ) · c (Aµ · Aν ) . i. As a matter of fact only the couplings σi are tuned with this procedure. Using the parameterization given in (9) and (10). They are free parameters which can be fixed setting the field normalization and the gauge coupling at the subtraction point µ equal to their physical values at Λ = 0. SΠinv Πinv = 0.s. γ . Bonini.258 M. The remaining monomials contribute to Π  Z 1 1 1 e(Λ0 ) ≡ d4 x σ1 Λ20 A2µ + σ2 (∂µ Aµ )2 + σ3 wµ · c ∧ Aµ + σ4 v · c ∧ c Π 2 2 2  2 g2 g + σ5 (Aµ ∧ Aν )2 + σ6 (Aµ · Aν )2 . µ/Λ0 ). µ/Λ0 ). γ ] with the UV wave function constants and UV couplings given by zi = zi (g. . In the following subsections we will determine the expression for the couplings σi in term of the Yang–Mills coupling g and the cutoff function. i. Λ0 ] contains the relevant monomials given in (5) (as discussed in the previous section its irrelevant contributions vanish as Λ0 → ∞ and therefore can be omitted in the calculations performed in this section). This functional can be split into two contributions 2 e 0 ).e.

γ . (11) δ7 = δ9 = δ11 = ig(σ3 + σ6 − σ5 ). q) = (p − q)ρ gµν + (2q + p)µ gνρ − (2p + q)ν gµρ (14) .M. g 259 δ3 = −iσ3 . At (1) Λ = Λ0 . At one-loop order ∆ˆ Γ can be evaluated taking the integrand in (7) at tree level and noticing that Π (0) = SBRS and then the sum over ΦJ is restricted to the (A. q ) 2 LJ q0 IL int the interaction part of SBRS and with SBRS   δ δ int S int . c)-fields. −p. (13) DΛ0 ∞ (q ) Γ [−q . 2 The system of equations obtained by imposing δi to be equal to the coefficient of the corresponding monomials in −∆ˆ Γ overdetermines the couplings σi (eleven equations and five unknown couplings) therefore it can be solved if one has only five independent equations. −p. δ5 = ji p. −p) S Γ¯I(0) 0 BRS J   IJ  Z  0 ¯ (0) 0 int 0 − . Λ ] = (−) (−q. Λ0] ∆ˆ Γ (Λ0 ) = δ4 = −i(σ2 − σ3 ). as can be seen in (11). Bonini. Λ0 ] SBRS (−q. g g 1 δ8 = δ10 = ig(σ5 − σ3 ). E. Φ.γ . q) Acρ (−p − q) − g 2 tµνρσ SBRS Abν Aaµ Z × Acρ (k) Adσ (−p − q − k). q) δΦJ (q) δΦI (p) BRS IJ For instance. the direct calculation given in the following subsections will shown that this is true. = SBRS (p. The necessary six relations among the parameters δi are automatically obtained from the anticommutativity of the BRS operator SΠ . (12) x δ Γ [Φ. Though it is not evident that the same relations hold for the coefficients of ∆ˆ Γ . Tricarico / Nuclear Physics B 585 (2000) 253–274 i δ1 = − σ1 .Λ] is recursively where the functional Γ¯ (0) originating from the inversion of δΦ(−q)δΦ(−p) defined as   δJ int [−q. q) = −ig abc tµνρ (p.q δ2 1 SBRS − × δΦj (q)δγi (p) g ! Z uµ ∂µ c . Φ. when ΦI and ΦJ are vector fields one has   int abcd (p. k where tµνρ (p. 1 i δ6 = (σ3 − σ4 ). ∆ˆ Γ is given by   Z (1) K0Λ0 (p)(−)δi DΛ0 ∞ (q) Γ¯ (0) [−q. −p. g δ2 = i σ2 . γ .

Λ0 ) = −2ig Z tνµν (q. p) − 2qµ  KΛ0 ∞ (q) q2 K0Λ0 (q + p). Using the vertices of SBRS one obtains Z (Ac) Aµ (p) · c(−p)∆ˆ Γ µ (p.Φj = c) and (γi = v .260 M.e. Φj = c). Recalling that K0Λ0 (0) = 1 and that K0Λ0 (q) → 0 rapidly enough for q 2 → ∞. 6q 2 (16) q where K ≡ K0Λ0 (q) and the 0 denotes the derivative with respect to q 2 and we have used the identity KΛ0 ∞ (q) = 1 − K0Λ0 (q).Φj = Aν ). q2 (15) q and δˆ2 = −2g Z (1 − K)(18K 0 + 21K 00 q 2 + 4K 000 q 4 ) . 4. E. (q + p)2 . q 2 & Λ20 . Λ20 ) ensures that this vertex is quadratically divergent in the Λ0 → ∞ limit but without logarithmic divergences. Ac-vertex There are three contributions 3 to the Aµ · c-vertex which are obtained inserting in (12) the first term of (13) and setting (γi = wν . In the following subsections we compute the coefficients of the various field monomials in ∆ˆ Γ . In this limit one gets 4 2 ˆ 2 ˆ 2 ∆ˆ (Ac) Γ µ (p. The coefficients δˆ1 and δˆ2 are finite and cutoff function dependent. q The presence of the two cutoff functions having almost non-intersecting supports ( i. Tricarico / Nuclear Physics B 585 (2000) 253–274 and  a1 a2 c ca3 a4 4  −  a1 a4 c  ca2 a3 gµ1 µ3 gµ2 µ4 tµa11···a ···µ4 =   +  a1 a3 c  ca2 a4 −  a1 a4 c  ca3 a2 gµ1 µ2 gµ3 µ4  +  a1 a3 c  ca4 a2 −  a1 a2 c  ca3 a4 gµ1 µ4 gµ2 µ3 are the three and four-vector elementary vertices. The corresponding graphs are given in Fig 1. Bonini. respectively. Λ0 ) = i pµ [δ1 Λ0 − δ2 p + O(p /Λ0 )] with 2g δˆ1 = 2 Λ0 Z (1 − K)(3K + 2K 0 q 2 ) . (15) and (16) can be rewritten as Z 2g (1 − 2K)K ˆδ1 = (17) 2 q2 Λ0 q .1. (γi = wν . Λ0 ) p with ∆ˆ (Ac) Γ µ (p..

i.. . (18) q = i/(16π 2 ) (the factor i comes from the q-integration). 4. on the cutoff function. All momenta are incoming.e. The fine-tuning equation where r (8) for the Ac-vertex becomes δ1 = −δˆ1 . as shown in Table 1. Table 1 One-loop values of the couplings σ1 and σ2 for three different cutoff functions (r = i and n > 1) 16π 2 K0Λ0 (p) 2 2 e−p /Λ0 Λ40 (p 2 +Λ20 )2 Λ2n 0 p 2n +Λ2n 0 σ1 0 2 − 2ig3 r 2πig 2 r(n−2) n2 sin(π/n) σ2 ig 2 r 12 7ig 2 r 30 ig 2 r(5−3n) 6 and 5r −3 δˆ2 = g 6 ! Z K 02 . Tricarico / Nuclear Physics B 585 (2000) 253–274 261 Fig.2. σ1 = − 2 q2 Λ0  (1) σ2 = ig 2 q 5r −3 6 Z K 02 (19)  (20) q and their values depend. The curly and dashed line denotes the gluon and ghost field respectively. the double lines represent the BRS source associated to the field depicted by the top line. Bonini. Graphical contribution to the A-c-vertex of ∆ˆ Γ .M. AAc-vertex In order to find the AAc-vertex one has to insert in (12) the functional Γ¯ (0) obtained from the first iteration in (13). Then from (11) one finds that σ1 and σ2 are given by Z 2ig 2 (1 − 2K)K (1) . 3 This computation can be found in [8] and is repeated here for completeness. δ2 = −δˆ2 . 1. E.

where P is a combination of the external momenta p1 and p2 . p1 ) − (1 ↔ 2) . Λ0 ) 2 p1 p2 with ∆ˆ (AAc) Γ µ1 µ2 (p1 . Using the vertices of SBRS one obtains Z 1 Aµ1 (p1 ) · Aµ2 (p2 ) ∧ c(−p1 − p2 ) ∆ˆ (AAc) Γ µ1 µ2 (p1 . Φj = c) and one for (γi = v . Graphical contribution to the AAc-vertex of ∆ˆ Γ . −p)   int 0 SBRS (−q. (21) IL One finds one contribution for (γi = wν . Λ0 ) = gµ1 µ2 (p12 − p22 )δˆ3 + (p1µ1 p1µ2 − p2µ1 p2µ2 )δˆ4 + O(P 4 /Λ20 ). Φj = Aν ).  Z  δJ 0 int 0 − (−) DΛ0 ∞ (q ) SBRS (−q . q ) LJ q0 . In the Λ0 → ∞ limit one gets (AAc) ∆ˆ Γ µ1 µ2 (p1 . 2. Tricarico / Nuclear Physics B 585 (2000) 253–274 Fig. p2 ) =− 2 q 2 (q + p1 )2 q KΛ0 ∞ (q)KΛ0 ∞ (q + p1 + p2 ) K0Λ0 (q + p1 ) q 2 (q + p1 + p2 )2  KΛ0 ∞ (q)KΛ0 ∞ (q + p1 ) K0Λ0 (q − p2 ) + K0Λ0 (q + p1 + p2 ) − qµ1 (q + p1 )µ2 q 2 (q + p1 )2 # − (q + p1 + p2 )ν tνµ1 µ2 (q. p2 . E.262 M. Z (1 − K)  13K 2 − 2q 2(11K 0 + 8K 02 q 2 + K 00 q 2 ) δˆ3 = −g 2 6q 4 q + K(−13 + 15K 0 q 2 − 8K 00 q 4 ) and δˆ4 = g 2 Z  (22) (1 − K)  13K 2 + 2q 2 (−8K 0 + 10K 02 q 2 − K 00 q 2 ) 6q 4 q  + K(−13 + 45K 0 q 2 + 10K 00 q 4 ) . p2 . Bonini. The corresponding graphs are given in Fig. p2 . two for (γi = wν . Φj = c) (plus the terms obtained from the permutation of the two vectors). (23) . Λ0 ) Z " KΛ0 ∞ (q)KΛ0 ∞ (q + p1 ) g2 K0Λ0 (q + p1 + p2 ) tρµ1 ν (q. p1 ) tνµ2 ρ (q + p1 . 2.

i. AAAc-vertex The AAAc-vertex receives contribution from (21) and the term originating from the second iteration of (13). respectively. E. In order to check the consistency of our computation the fine-tuning equation δ4 ≡ −i(σ2 − σ3 ) = δˆ4 must be automatically satisfied with the couplings σ2 and σ3 previously determined and given in (20) and (24). (25) The corresponding graphs are shown in Figs. Λ0 ). Bonini. Tricarico / Nuclear Physics B 585 (2000) 253–274 Table 2 One-loop values of the coupling σ3 for three different cutoff functions (r = 2 2 e−p /Λ0 K0Λ0 (p) − σ3 ig 2 r(39 ln 43 +11) 18 263 i and n > 1) 16π 2 Λ40 (p 2 +Λ20 )2 Λ2n 0 p 2n +Λ2n 0 49igr 360 −37n+39) − igr(10n 36n 2 The fine-tuning equation (8) for the AAc-vertex allows to find the value of σ3 in term of g and the cutoff function. p2 .M. q 0)  IL . After performing the trace over the gauge indices this vertex can be written as Z   1 Aµ1 (p1 ) · Aµ2 (p2 ) Aµ3 (p3 ) · c(−p1 − p2 − p3 ) 3! p1 p2 p3 × ∆ˆ (AAAc) Γ µ1 µ2 µ3 (p1 . Z   δJ int int (−q 00 . 5–7 and the various contributions are computed in Appendix A. p3 . −p) H J DΛ0 ∞ (q 0 ) SBRS (−q 0 .e. one has to prove the consistency condition δˆ3 + δˆ4 = −g δˆ2 .3. In the Λ0 → ∞ limit one obtains . From (22) and (23) one finds Z 1−K 0 1 (K + 5KK 0 + 6K 02 q 2 + 3KK 00 q 2 ) − (δˆ4 + δˆ3 ) = −g g q2 q which becomes equal to (18) after integration by parts. 4. From δ3 = δˆ3 and (11) one finds ! Z 13K(1 − K)2 + 10q 4K 02 (1) 2 37r − (24) σ3 = −ig 36 6q 4 q and its value for different choices of the cutoff function is given in Table 2. Alternatively.. q 00 ) LH DΛ0 ∞ (q 00 ) SBRS (−) q 0 q 00 int × SBRS (−q.

the couplings σ5 and σ6 are given by ! Z K(1 − K)(23K 2 − 12K − 11) − 10K 02 q 4 (1) 2 3r + . E. p3 .264 M. Using the vertices of SBRS one gets Z 1 (wcc) wµ (−p1 − p2 ) · c(p1 ) ∧ c(p2 ) ∆ˆ Γ µ (p1 .4. Λ0 )  = −3i δˆ7 (p1µ1 gµ2 µ3 + p2µ2 gµ1 µ3 ) + 2δˆ8 p3µ3 gµ1 µ2 + δˆ9 (p1µ2 gµ1 µ3 + p2µ1 gµ2 µ3 )  (26) + δˆ10(p1µ3 + p2µ3 )gµ1 µ2 + δˆ11 (p3µ2 gµ1 µ3 + p3µ1 gµ2 µ3 ) + O(P 4 /Λ20 ). Bonini. 3. Tricarico / Nuclear Physics B 585 (2000) 253–274 Table 3 One-loop values of the couplings σ5 and σ6 for three different cutoff functions (r = i and n > 1) 16π 2 K0Λ0 (p) 2 2 e−p /Λ0 Λ40 (p 2 +Λ20 )2 Λ2n 0 p 2n +Λ2n 0 σ5 2 − ig r(13+9412ln2−70 ln 3) 47ig 2 r 630 ig 2 (5n2 −27n+28) 18n σ6 2 − ig r(2−63 3ln 3+99 ln 2) r − 58ig 105 2 − 2ig 3 2 (AAAc) ∆ˆ Γ µ1 µ2 µ3 (p1 . (28) 6q 4 q By using the fine-tuning equation (8). 4. wcc-vertex The wcc-vertex is obtained inserting (21) in (12) and is given by the three graphs depicted in Fig. where δˆ7 = δˆ9 = δˆ11 Z  (1 − K)  3 2 0 2 0 2 0 2 = −g 3 + K (124K q − 47) + K(16 − 111K q ) + 14K q 31K 6q 4 q (27) and 1 δˆ8 = δˆ10 2 Z = −g 3  (1 − K)  23K 3 + K 2 (56K 0 q 2 − 25) + K(2 − 48K 0 q 2 ) + K 0 q 2 . p2 . σ5 = −ig 2 6q 4 q ! Z (1 − K)(9K 3 + 12K 2 + 3K) (1) 2 2r + σ6 = −ig 3 q4 q and their values for different cutoff functions are given in Table 3. Λ0 ) 2 p1 p2 . p2 . together with the results (11) and (24).

p2 . Λ0 ) = gµν δ6 + O(P /Λ0 ) and. Λ0 ) = i(p1 + p2 )µ δˆ5 + O(P 2 /Λ20 ) . therefore.5. Thus from (11) the one-loop value of the coupling σ4 is given by (24). p1 p2 p3 (Accw) The complete expression of ∆ˆ Γ µν is not needed since 2 2 ˆ ∆ˆ (Accw) Γ µν (p3 . q 2 (q + p2 )2 In the Λ0 → ∞ limit one obtains   (wcc) ∆ˆ Γ µ (p1 .M. p1 . 4. All the terms can be collected in the following monomial Z   (Accw) Aµ (p3 ) · c(p2 ) wν (−p1 − p2 − p3 ) · c(p1 ) ∆ˆ Γ µν (p3 . where δˆ5 = g Z (−1 + 4K − 3K 2 )K 0 =0 q2 q as integration by parts can show. gµν δ6 = g q6 q which vanishes. p2 . p2 . thus this vertex must vanish. In this limit one obtains Z  (1 − K)3 K  2 2 ˆ −q gµν + qµ qν + qρ tρνµ (q. 0) . Graphical contribution to the wcc-vertex of ∆ˆ Γ . with (wcc) ∆ˆ Γ µ (p1 . 4. wccA-vertex The fine-tuning equation for this vertex is directly obtained form the the relation between δ5 and δ6 in (11). . Λ0 ). in the Λ0 → ∞ limit one can set to zero all the external momenta. Λ0 ) Z " KΛ0 ∞ (q)KΛ0 ∞ (q − p1 − p2 )K0Λ0 (q − p2 ) (q − p1 − p2 )µ =g q 2 (q − p1 − p2 )2 q # KΛ0 ∞ (q)KΛ0 ∞ (q + p2 )[K0Λ0 (q + p1 + p2 ) + K0Λ0 (q − p1 )] − qµ + (1 ↔ 2) . p3 . The graphs which contribute to this vertex are obtained inserting (25) in (12) and are given in fig. p3 . p2 . Bonini. Tricarico / Nuclear Physics B 585 (2000) 253–274 265 Fig. 3. E. p1 .

Bonini. This procedure systematically generalizes to higher orders. There are actually two crucial points which deserve mentioning. Conclusions The computation presented in this paper has shown that the requirement that the physical effective action satisfies the ST identity translates into a fine-tuning equation which allows to determine the couplings of the relevant. the problem of assigning the boundary conditions of the RG flow is finally solved. Since the integral in (7) is damped by these cutoff functions. 4. Once the field normalization and the gauge coupling are fixed at the physical point Λ = 0. where P is a combination of external momenta. Graphical contribution to the wccA-vertex of ∆ˆ Γ . Tricarico / Nuclear Physics B 585 (2000) 253–274 Fig. they turns out to be finite. non-invariant interactions in the bare action Π(Λ0 ). We explicitly compute the non-invariant couplings at one-loop. 5. At oneloop order this has been explicitly shown in Section 4 and is ensured by the fact that at tree δ 2 Γ −1 ) is given by either a relevant vertex or level and at the UV scale the functional ( δΦδΦ a sequence of relevant vertices joint by propagators with a cutoff function KΛ0 ∞ (q + P ). only the contributions with a restricted number of propagators survive in . First of all one must discuss the locality of ∆ˆ Γ .266 M. E.

which preserves BRS invariance and thus avoids the fine-tuning. However the RG formulation is general and can be applied to chiral gauge theories without anomalies along the same lines. Tricarico / Nuclear Physics B 585 (2000) 253–274 267 the Λ0 → ∞ limit.. Another issue to discuss is the finiteness of the non-invariant couplings at higher loops. In this case all the regularization procedures break the gauge symmetry and the fine-tuning is unavoidable. Nevertheless at higher loops one could worry about non-local terms which may arise from the full propagators. For the same reason. As a consequence higher derivative interactions (i. 4 Recently [16] these counterterms have been systematically computed at one-loop order in the dimensional regularization scheme. A similar argument holds for the possible non-local contributions coming from K0Λ0 (p). it is easy to realize that the terms of this expansion which ( δΦδΦ survive in the Λ0 → ∞ limit have less than three propagators. but a restricted number of fields). more elegant formulations exist for a pure gauge theory. one more than in the one-loop case. vertices with higher powers of the momenta. are also involved in the fine-tuning. In particular one has to consider irrelevant interactions involving the BRS sources. As a final remark we should mention that it has been suggested [17. makes the q-integration in (7) finite. Moreover in the dimensional regularization with minimal subtraction scheme the breaking of the symmetry requires the introduction of all the possible non-invariant counterterms 4 while in the RG approach only interactions which are invariant under global chiral transformation need to be considered in the fine-tuning procedure [8]. i. but they can be reabsorbed in field and gauge coupling redefinition. .. By dimensional analysis non-local parts of these vertices are proportional to negative powers of Λ0 . With a little thought. such as dimensional regularization with minimal subtraction.e. One can infer from power counting that they are of the relevant type. yet they can contribute to the relevant part of ∆ˆ Γ since in (7) the loop momentum q is of the order of Λ0 . This related to the fact that at higher loops the wc-vertex in Π(Λ0 ) δ2 Γ contributes to (7) and therefore all the fields of ∆ˆ Γ originate from the inversion of δΦδΦ (recall that the relevant part of the functional ∆Γ contains at most four fields). E. irrelevant proper vertices with at most six legs have to be taken into account. in order to compensate the powers of q in the irrelevant vertices. Therefore at higher loops the fine-tuning procedure is rather cumbersome but well defined since the irrelevant vertices are completely determined by the flow equation (3) while the relevant ones are given by the fine-tuning at the lower loop orders.M. having almost non-intersecting supports. The presence in (7) of the cutoff functions K0Λ0 (q + Pi ) and KΛ0 ∞ (q + Pj ).e. to overcome the difficulty of the fine-tuning. Bonini. However is not obvious how this procedure can be extended to chiral gauge theories. Clearly. Π(Λ0 ) and the proper vertices generated in the expansion of δ 2 Γ −1 ) . Divergent contributions to the non-invariant couplings σi can be found only when we consider in (7) the relevant vertices zi (Λ0 ). such as the Wilson loop.18] to use gauge invariant variables. Actually the cutoff function K0Λ0 (q) must fall off more rapidly than any negative power of q for q 2 > Λ20 .

p1 ) (q + p23 )2   KΛ0 ∞ (q + p13 ) + tµ1 µ2 µ3 (q + p13 .268 M. p3 . p3 ) + tµ2 µ3 µ1 (q + p12 . Graphical contribution to the AAAc-vertex of ∆ˆ Γ . Bonini. Appendix A In this appendix we compute the various contributions to the AAAc-vertex. Vian. Marchesini and F. E. p2 ) (q + p13 )2   KΛ0 ∞ (q + p23 ) + tµ2 µ1 µ3 (q + p23 . p1 ) − 2tµ3 µ1 µ2 (q + p23 . p2 . (q + p12 )2 For Λ0 → ∞ this graph gives the following contribution to the various coefficients δˆi in (26) Z (1 − K)(7K 2 − 7K 0 q 2 − K(7 − 10K 0 q 2 )) ˆδ (5a) = . p2 ) − 2tµ3 µ2 µ1 (q + p13 . 7 2q 4 q (a) (b) (c) Fig. Tricarico / Nuclear Physics B 585 (2000) 253–274 Acknowledgements We have greatly benefited from discussions with G. p2 and p3 are the momenta of the vector fields and that one has to consider the three permutations obtained by choosing the two external vectors which enters in the four-vector vertex) Z KΛ0 ∞ (q) 3 K0Λ0 (q + p1 + p2 + p3 ) 3ig q2 q  KΛ0 ∞ (q + p12 ) × 2gµ1 µ2 tµµ3 µ (q + p12 . p3 ) . p1 ) (q + p23 )2    KΛ0 ∞ (q + p12 ) − tµ1 µ3 µ2 (q + p12 . . p2 ) (q + p13 )2 KΛ0 ∞ (q + p23 ) +gµ2 µ3 tµµ1 µ (q + p23 . The (AAAc) contribution to ∆ˆ Γ µ1 µ2 µ3 (p1 . 5. 5. We first consider in (12) the terms with γi = wµ . Λ0 ) from the the graph in Fig. p3 ) (q + p12 )2 KΛ0 ∞ (q + p13 ) +gµ1 µ3 tµµ2 µ (q + p13 . The former involves the four-vector vertex given in (14) and the corresponding graphs are shown in Fig. In this case both the contributions to the functional Γ¯ (0) given in (21) and (25) are needed to compute this vertex. 5(a) is (recall that p1 .

p2 ) − tµ3 µ2 µ1 (q. p2 ) + gµ2 µ3 tµµ1 µ (q. 2q 4 q Finally from Fig. 2q 4 Z (1 − K)(2K 2 + 7K 0 q 2 − K(2 + 13K 0 q 2 )) . p3 )] . 5(b) gives Z KΛ0 ∞ (q) 3 K0Λ0 (q + p1 + p2 + p3 ) 3ig q2 q  KΛ0 ∞ (q + p3 ) × 2gµ1 µ2 tµµ3 µ (q. 7 2q 4 q δˆ8(5b) = Z (1 − K)(9K − 5)K 0 . p1 )] (q + p1 )2  KΛ0 ∞ (q + p3 ) −[tµ1 µ3 µ2 (q. Bonini. 2q 4 Z (1 − K)(9K 2 − 9K + 10K 0 q 2 − 11KK 0 q 2 ) . Tricarico / Nuclear Physics B 585 (2000) 253–274 δˆ8(5a) = Z q (5a) δˆ9 = − (5a) δˆ10 = Z (1 − K)(5K 2 − 5K 0 q 2 − K(5 − 6K 0 q 2 )) . 2q 4 q (5b) δˆ9 = Z q (5b) =− δˆ10 (5b) δˆ11 = Z (1 − K)(9K 2 − 9K − 7K 0 q 2 + 8KK 0 q 2 ) . (q + p3 )2 In the Λ0 → ∞ limit the various coefficients δˆi for this term are given by Z (1 − K)(11K − 7)K 0 ˆδ (5b) = .M. 2q 4 q (1 − K)(19K 2 − 10K 0 q 2 − 19K(1 − K 0 q 2 )) . E. p3 ) + tµ2 µ3 µ1 (q. 2q 4 The graph in Fig. 2q 4 q (5a) = δˆ11 Z q (1 − K)(7K 2 − 7K 0 q 2 − K(7 − 12K 0 q 2 )) . p1 ) (q + p2 )2 (q + p1 )2 KΛ0 ∞ (q + p2 ) −[2tµ1 µ2 µ3 (q. p1 ) − tµ3 µ1 µ2 (q. p2 )] (q + p2 )2 KΛ0 ∞ (q + p1 ) −[2tµ2 µ1 µ3 (q. p3 ) (q + p3 )2 KΛ0 ∞ (q + p2 ) KΛ0 ∞ (q + p1 ) +gµ1 µ3 tµµ2 µ (q. 2q 4 q (1 − K)(9K − 7)K 0 . 5c we get 269 .

p2 ) tρµ1 µ (q + p23 . p3 . 6(a) is Z KΛ0 ∞ (q) K0Λ0 (q + p1 + p2 + p3 ) −ig 3 q2 q  × tµµ1 ν (q. p2 ) tρµ3 µ (q + p12 . 6. For γi = wµ there are four different terms according to the position of the ghost field in (25) which are depicted in Fig. E. Graphical contribution to the AAAc vertex of ∆ˆ Γ . 2q 4 q (AAAc) All the remaining contributions to ∆ˆ Γ µ1 µ2 µ3 (p1 . 7 2q 4 q (5c) = δˆ8(5c) = −δˆ11 (5c) δˆ9 = − (5c) = δˆ10 Z Z Z (1 − K)(2K 2 − 2K + K 0 q 2 − 3KK 0 q 2 ) . p2 . (d) . p1 ) tνµ2 ρ (q + p1 . p3 ) KΛ0 ∞ (q + p1 )KΛ0 ∞ (q + p12 ) (q + p1 )2 (q + p12 )2 +tµµ3 ν (q. Λ0 ) are generated from the part of the functional Γ¯ (0) given in (25). p1 ) × (a) (b) (c) Fig. 6. 2q 4 q (1 − K)(K 2 − K + K 0 q 2 − 2KK 0 q 2 ) . q4 q (1 − K)(4K 2 − 4K + K 0 q 2 − 5KK 0 q 2 ) . The contribution from the graph in Fig.270 M. p3 ) tνµ2 ρ (q + p3 . + (q − p1 )2 The δˆi for this graph are Z (1 − K)(2K 2 − 2K − K 0 q 2 − KK 0 q 2 ) ˆδ (5c) = − . Tricarico / Nuclear Physics B 585 (2000) 253–274 Z −3ig 3 KΛ0 ∞ (q) K0Λ0 (q + p1 + p2 + p3 ) q2 q   KΛ0 ∞ (q − p3 )  2gµ1 µ2 qµ3 − gµ2 µ3 qµ1 − gµ1 µ3 qµ2 × 2 (q − p3 )  KΛ0 ∞ (q − p2 )  gµ1 µ3 qµ2 + gµ1 µ2 qµ3 − 2gµ2 µ3 qµ1 + (q − p2 )2   KΛ0 ∞ (q − p1 )  gµ2 µ3 qµ1 + gµ1 µ2 qµ3 − 2gµ1 µ3 qµ2 . Bonini.

E. 12q 4 q From Fig. q4 q The graph in Fig. 6(b) gives Z KΛ0 ∞ (q)KΛ0 ∞ (q + p1 ) qµ1 ig 3 q 2 (q + p1 )2 q  KΛ0 ∞ (q + p12 ) × (q + p1 )µ2 (q + p12 )µ3 K0Λ0 (q − p3 ) (q + p12 )2  KΛ0 ∞ (q + p13 ) K (q − p ) + permutations. 12q 4 Z (1 − K)2 (K 2 + K 0 q 2 − K(1 − 4K 0 q 2 )) . 12q 4 q (6b) =− δˆ10 Z (1 − K)2 (4K 2 + K 0 q 2 − K(1 − 4K 0 q 2 )) .M. 271  × where the permutations are performed among the vector fields. p2 ) −ig 3 q KΛ0 ∞ (q + p3 )KΛ0 ∞ (q − p12 ) (q + p3 )2 (q − p12 )2 × K0Λ0 (q − p1 ) + qµ1 (q + p1 )µ tµµ3 µ2 (q − p23 . + (q + p1 )µ3 (q + p13 )µ2 0Λ 2 0 (q + p13 )2 The δˆi for this graph are Z (1 − K)2 (5K 2 − K 0 q 2 − K(5 + 4K 0 q 2 )) ˆδ (6b) = . 12q 4 q (6b) δˆ11 = − Z (1 − K)2 (K 2 + 2K 0 q 2 − K) . Tricarico / Nuclear Physics B 585 (2000) 253–274 KΛ0 ∞ (q + p3 )KΛ0 ∞ (q + p23 ) (q + p3 )2 (q + p23 )2 +permutations. 6(c) we get the contribution Z ( qµ3 (q + p3 )µ tµµ2 µ1 (q − p12 . For Λ0 → ∞ the contribution of this graph to the various coefficients δˆi in (26) is (6a) (6a) (6a) (6a) (6a) δˆ7 = 2δˆ8 = δˆ9 = δˆ10 = δˆ11 Z (1 − K)2 (K 2 − 2K 0 q 2 − K(1 − 4K 0 q 2 )) = −5 . p3 ) KΛ0 ∞ (q + p1 )KΛ0 ∞ (q − p23 ) (q + p1 )2 (q − p23 )2 . 7 12q 4 q δˆ8(6b) = Z q (6b) δˆ9 = − (1 − K)2 (K 2 − K 0 q 2 − K) . Bonini.

Bonini. 12q 4 q (1 − K)2 (2K 2 + 11K 0 q 2 − 2K(1 + 14K 0 q 2 )) . p1 )tνµ2 µ3 (q − p23 . 2q 4 q (1 − K)2 (5K 2 − 4K 0 q 2 − K(5 − 8K 0 q 2 )) . p3 )tνµ2 µ1 (q − p12 . 12q 4 .272 M. q2 which gives δˆ7(6c) = − Z (1 − K)2 (4K 2 + K 0 q 2 − K(4 + 5K 0 q 2 )) . p2 ) ig q × KΛ0 ∞ (q − p23 ) K0Λ0 (q − p3 ) (q − p23 )2 + tµµ3 ν (q − p1 − p2 − p3 . 12q 4 q From Fig. 6(d) one has  Z 3 qµ tµµ1 ν (q − p1 − p2 − p3 . p2 )  KΛ0 ∞ (q − p12 ) K (q − p ) × 0Λ0 1 (q − p12 )2 KΛ0 ∞ (q)KΛ0 ∞ (q − p1 − p2 − p3 ) + permutations. Tricarico / Nuclear Physics B 585 (2000) 253–274 ) × K0Λ0 (q − p2 ) × KΛ0 ∞ (q) + permutations. × q 2 (q − p1 − p2 − p3 )2 and finds (6d) δˆ7 = Z q δˆ8(6d) = − (6d) δˆ9 = Z q (6d) δˆ10 = − (1 − K)2 (5K 2 − 7K 0 q 2 − K(5 + 22K 0 q 2 )) . E. 12q 4 q (1 − K)2 (19K 2 + 4K 0 q 2 − K(19 + 14K 0 q 2 )) . 6q 4 q (6c) (6c) δˆ8 = −δˆ11 = δˆ9(6c) = − (6c) δˆ10 = Z Z Z (1 − K)2 KK 0 . 12q 4 Z q (1 − K)2 (49K 2 + K 0 q 2 − K(49 + 26K 0 q 2 )) . 12q 4 Z (1 − K)2 (15K 2 + 7K 0 q 2 − K(15 + 22K 0 q 2 )) .

Phys. M. Field Theory and Statistical Mechanics. Λ0 ) for γi = v. The contribution to the δˆi of this graph δˆ7(7) = 2δˆ8(7) = δˆ9 = δˆ10 = δˆ11 = − (7) (7) (7) Z (1 − K)2 (K 2 + K 0 q 2 − K(1 + 2K 0 q 2 )) . Ann. Nucl. Nucl. Kogut.R. Rouet. Elementary Particles. B 378 (1996) 213. Bonini. B 4 (1971) 3184. Math. M.G. On the construction of renormalized quantum field theory using renormalization group techniques. Lett. which is depicted in Fig. Ellwanger. Lett. Bonini. Morris. Tricarico / Nuclear Physics B 585 (2000) 253–274 273 Fig. B 437 (1995) 163. Wilson. R. Marchesini. Parma University. Rep. Phys.G. Phys. Marchesini. M. Rev. A. Rev. 42 (1975) 127. B 231 (1984) 269. Bonini.). there is a contribution to ∆ˆ Γ µ1 µ2 µ3 (p1 . 12q 4 q Finally. B 335 (1994) 364. Phys. J. B 511 (1998) 479. . M. B 329 (1994) 249. B 346 (1995) 87. Nucl. in: M. G. U. D’Attanasio. Vian. 7. G.G. Phys. Becchi. Phys. Phys. Wilson. (6d) δˆ11 = − Z (1 − K)2 (3K 2 − 10K 0 q 2 − K(3 − 34K 0 q 2 )) . Lett. B 421 (1994) 429. Marchesini. Bonini. Commun. Onofri (Eds. 12 (1974) 75. Phys. F. Phys. Phys. 1993. C. Phys. T. D’Attanasio. M. p3 . References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] K. D’Attanasio. Becchi. Polchinski. J. C. B 4 (1971) 3174. Lett. 7 and gives Z KΛ0 ∞ (q) K0Λ0 (q + p1 + p2 + p3 ) −ig 3 q2 q  KΛ0 ∞ (q + p1 )KΛ0 ∞ (q + p12 ) × qµ1 (q + p1 )µ2 (q + p12 )µ3 (q + p1 )2 (q + p12 )2  KΛ0 ∞ (q + p3 )KΛ0 ∞ (q + p23 ) + qµ3 (q + p3 )µ2 (q + p23 )µ1 (q + p3 )2 (q + p23 )2 (AAAc) + permutations. (NY) 98 (1976) 287. Graphical contribution to the AAAc vertex of ∆ˆ Γ . 3q 4 q After putting together all these results one obtains (27) and (28). G. M. K. Bonini. Stora. E.M. p2 . Phys. Phys. Nucl. E.

Tricarico / Nuclear Physics B 585 (2000) 253–274 C. Exact renormalization group and loop equation.274 [11] [12] [13] [14] [15] M. E. Bonini. T. Elsevier Science. hep-th/9905076. Mod. 1984. M. [16] C. J. Bonini. Nucl. [17] T. [18] S. B 409 (1993) 441. D. Stora (Eds. Wetterich. B 417 (1994) 181. Morris. Marchesini. Morris. Lett. 1. hep-th/9910256. C. restoration of BRS symmetry and the renormalization group equation for chiral non-Abelian gauge theories in dimensional renormalization with a non-anticommuting Γ5 . D’Attanasio. Hirano. Int. Phys. hep-th/9910058. M. Nucl. Martin. . Lectures on the renormalization of gauge theories. in: B. Phys. Reuter. A 9 (1994) 2411. B 301 (1993) 90. Phys. Becchi. Groups and Topology II (Les Houches 1983).).P. DeWitt. Sanchez-Ruiz. A gauge invariant exact renormalization group. Action principles.S. G. Relativity. Wetterich. C. R. Phys. M.

etc.edu.3 2 1 3 ( 0 0 ) 0 0 3 7 8 . the two possible solutions of the 4 × 4 Cabibbo–Kobayashi–Maskawa factor Vt 0 s Vt 0 b are obtained as a function of the t 0 -quark mass. In this case the above-mentioned physical quantities can serve as efficient tools in search of the fourth generation. Savcı).edu.tr (T.3 .25. There is no any theoretical argument to restrict the SM to three known generations of the fermions. savci@metu.edu. is studied.M. 13. and the ratio ΓL /ΓT of the decay widths when K ∗ meson is longitudinally and transversally polarized. what E-mail addresses: taliev@metu. if ever exists. CKM matrix elements 1. Rare exclusive B meson decays. mass spectrum.60.-i. A. M. 13. lepton forward–backward asymmetry. Using the experimental results on the branching ratios for the B → Xs γ and semileptonic B → Xc `¯ν decays. Another one of the unsolved fundamental problems of SM is the number of generations. Aliev).  2000 Elsevier Science B.V. All rights reserved.elsevier. PII: S 0 5 5 0 . Aliev. PACS: 12.nl/locate/npe Fourth generation effects in rare exclusive B → K ∗`+`− decay T. ozpineci@metu. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. From the LEP result of the invisible partial decay width of Z boson it follows that the mass of the extra generation neutrino N should be larger than 45 GeV [1]. The other solution yields almost identical results with the SM.20. Savcı Physics Department.Hv Keywords: Fourth generation.tr (M. Turkey Received 6 April 2000. But from theoretical point of view SM is an incomplete theory.Nuclear Physics B 585 (2000) 275–289 www. Middle East Technical University.M.tr (A. It is observed that the results for the above-mentioned physical quantities are essentially different from the standard model predictions only for one solution of the CKM factor. In this connection there comes into mind the following question: If extra generations exist.V. Özpineci). revised 31 May 2000. Özpineci. 06531 Ankara. on the experimentally measurable quantities such as invariant dilepton mass distribution.-v. This theory contains many unsolved open problems. All rights reserved. such as the origin of CP violation. Introduction At present Standard Model (SM) describes very successfully all low energy experimental data. accepted 6 June 2000 Abstract Influence of the fourth generation.

M. in the model we consider in this work where the fourth generation is introduced in the same way the three generations are introduced in the SM. because it occurs only at loop level in the SM. Heff = √ Vt b Vt∗s 2 i=1 10 (1) where the full set of the operators Oi (µ) and the corresponding expressions for the Wilson coefficients Ci (µ) in the SM are given in [24.18]. the W boson and the top quark masses strongly excluded the existence of the new generations with all fermions heavier than the Z boson mass. for example. Section 3 is devoted to the numerical analysis and our conclusion. if one allows neutral leptons to have masses close to 50 GeV. 2.276 T. we present the necessary theoretical expressions for the B → K ∗ `+ `− decay in the SM with four generations. The fourth generation changes only the values . [2–8] and the references therein). It is also well known that in the SM B → K ∗ `+ `− decay has “large” branching ratio and it has experimentally clean signature because two leptons are present in the final state. For this reason this decay is one of the most probable candidates to be detected in these machines and in our view it is the right time for an investigation in this direction. The paper is organized as follows. As has been noted already. / Nuclear Physics B 585 (2000) 275–289 effect they would have in low energy physics? This problem was studied in many works (see. Aliev et al.25]. Theoretical results The matrix element of the B → K ∗ `+ `− decay at quark level is described by b → transition for whom the effective Hamiltonian at O(µ) scale can be written as s`+ `− X 4GF Ci (µ)Oi (µ). Note that the inclusive B → Xs γ and B → Xs `+ `− decays have already been studied with the inclusion of the fourth generation in the SM [6. One promising area in experimental search of the fourth generation. This year the upgraded B factories at SLAC and KEK will provide us with the first experimental data. Contributions of the new generation to the electroweak radiative corrections were considered in papers [9–15]. However. no new operators appear and clearly the full operator set is exactly the same as in SM. whose values can be different from the ones in the SM [17. New physics effects can manifest themselves through the Wilson coefficients. At the same time this decay is sensitive to the various extension of the SM. The most straightforward and economical generalization of the SM to the fourgeneration case is similar to the three generations present in SM [16]. It was shown in [15] that the existing electroweak data on the Z-boson parameters. which we consider in this work. In this work we study the contribution of the fourth generation in the rare B → K ∗ `+ `− decay. the same data allows few extra generations. In Section 2. as well as the expressions of the other physical observables such as forward–backward asymmetry and the ratio of the decay widths when K ∗ meson is polarized longitudinally and transversally.20–23]. as well as through the new operators [19]. via its indirect loop effects. is the rare B meson decays.

A perturbative calculation leads to the result [24. Vt∗b Vt s 7 V ∗0 Vt 0 s new C (µ). C10 Vt b Vt s 10 C7tot (µ) = C7SM (µ) + (2) where the last terms in these expressions describe the contributions of the t 0 quark to the Wilson coefficients and Vt 0 b and Vt 0 s are the two elements of the 4 × 4 Cabibbo– Kobayashi–Maskawa (CKM) matrix. (3) − 2C7tot 2 s¯ σµν q ν (1 + γ5 )b `γ q where q 2 = (p1 + p2 )2 and p1 and p2 are the final leptons four-momenta. s) and the values where m bc = mc /mb . C9 (µ) and C10 (µ). . Aliev et al. J /ψ.25]. g(1.25]. The explicit expressions for g(b of Ci in the SM can be found in [24. C9tot (µ) = C9SM (µ) + t ∗b Vt b Vt s 9 V ∗0 Vt 0 s new tot SM (µ) = C10 (µ) + t ∗b C (µ). Yper (s) receives also long distance contributions. .M.. (4) where s = q 2 /m2B and the function Y (s) contains the contributions from the one loop matrix element of the four quark operators. / Nuclear Physics B 585 (2000) 275–289 277 of the Wilson coefficients C7 (µ). The effective ¯ µ ` can be written in the following coefficient C9tot of the operator O9 = s¯ γµ (1 − γ5 )b `γ form C9tot = C9 + Y (s). which have their origin in the real cc¯ intermediate states.e. the above effective Hamiltonian leads to following matrix element for the b → s`+ `− decay  Gα ¯ µ ` + C tot s¯γµ (1 − γ5 )b `γ ¯ µ γ5 ` M = √ Vt b Vt∗s C9tot s¯ γµ (1 − γ5 )b `γ 10 2 2π  mb ¯ µ` . 2 9 (5) mc . via virtual exchange of the fourth generation up quark t 0 . . In addition to the short distance contribution. The J /ψ family is introduced by the Breit–Wigner distribution for the resonances through the replacement [27–29] . Neglecting the s quark mass. The explicit forms of the Cinew can easily be obtained from the corresponding Wilson coefficient expressions in SM by simply substituting mt → mt 0 (see [24. ψ 0 . s). s)(3C1 + C2 + 3C3 + C4 + 3C5 + C6 ) Yper (s) = g(b 1 − g(1. (1). (2) we factored out the term Vt∗b Vt s in the effective Hamiltonian given in Eq.T. s)(C3 + 3C4 ) + (3C3 + C4 + 3C5 + C6 ). s). s)(4C3 + 4C4 + 3C5 + C6 ) 2 2 1 − g(0.26]). g(0. In deriving Eq. i. The above mentioned Wilson coefficients can be written in the following form Vt∗0 b Vt 0 s new C (µ). . mc .

278 T. (3) that in order to calculate the decay width and other physical observables of the exclusive B → K ∗ `+ `− decay.007 −0. ε)|¯s γµ (1 − γ5 )b|B(pB ) C (0) 2V (q 2 ) − iεµ (mB + mK ∗ )A1 (q 2 ) mB + mK ∗ A2 (q 2 ) + i(pB + pK ∗ )µ (ε∗ q) mB + mK ∗  2mK ∗ ∗  + iqµ 2 (ε q) A3 (q 2 ) − A0 (q 2 ) . 2 α mVi − sm2B − imVi ΓVi V =ψ i (6) i = 3C1 + C2 + 3C3 + C4 + 3C5 + C6 .65 and κ = 2. It follows from Eq. In other words.362 C1 C2 −0. q = −µνρσ ε∗ν pρ q σ (7) where ε is the polarization vector of K ∗ meson and q = pB − pK ∗ is the momentum transfer. For the lowest resonances J /ψ nad ψ 0 we will use κ = 1. ε)|¯s iσµν q ν (1 + γ5 )b|B(pB )   = 4µνρσ ε∗ν pρ q σ T1 (q 2 ) + 2i εµ∗ (m2B − m2K ∗ ) − (pB + pK ∗ )µ (ε∗ q) T2 (q 2 ) . Aliev et al.313 4.669 Y (s) = Yper (s) + 3π (0) X mV Γ (Vi → `+ `− ) C κi 2 i .36. we demand that A3 (q 2 = 0) = A0 (q 2 = 0). The semileptonic form factors coming from the dipole operator σµν q ν (1 + γ5 )b are defined as ∗ K (p. can be parametrized in terms of form factors.107 C3 C4 C5 C6 C7SM C9SM SM C10 0. / Nuclear Physics B 585 (2000) 275–289 Table 1 The numerical values of the Wilson coefficients at µ = mb scale within the SM.026 0. where the data for the right hand side is given in [30].248 1.M. the form factor A3 (q 2 ) can be written in terms of the form factors A1 (q 2 ) A2 (q 2 ) as follows A3 = mB + mK ∗ mB − mK ∗ A1 − A2 . The phenomenological parameters κi where can be fixed from B(B → K ∗ Vi → K ∗ `+ `− ) = B(B → K ∗ Vi ) B(Vi → `+ `− ). (3) over meson states. 2mK ∗ 2mK ∗ (8) In order to ensure finiteness of (8) at q 2 = 0. The corresponding numerical value of C 0 is 0.344 −4.011 −0. For the vector meson K ∗ with polarization vector εµ the semileptonic form factors of the V −A current is defined as ∗ K (p. Using the equation of motion. respectively. the exclusive B → K ∗ `+ `− decay which is described in terms of the matrix elements of the quark operators given in Eq. In our numerical analysis we use the average of J /ψ and ψ 0 for the higher resonances ψ (i) (see [31]). the matrix elements hK ∗ |¯s γµ (1 − γ5 )b|Bi and hK ∗ |¯s iσµν q ν (1 + γ5)b|Bi have to be calculated.031 −0.

. T3 (q 2 ) . m2B − m2K ∗ 279 (9) Using the form factors. i.33] F (s) = F (0) . for example. s) = 1 + r 2 + s 2 − 2rs − 2r − 2s. aF and bF are listed in Table 3 for each form factor. In the present work we choose light cone QCD sum rules method predictions for the form factors. T1 (q 2 ). This problem is related to the nonperturbative sector of QCD and it can be solved only in framework a nonperturbative approach. for the B → K ∗ `+ `− decay. A0 (q 2 ). r. we get for the B → K ∗ `+ `− decay width G2 α 2 dΓ = |Vt b Vt∗s |2 vλ1/2 (1. / Nuclear Physics B 585 (2000) 275–289  + 2i(ε∗ q) qµ − (pB + pK ∗ )µ  q2 T3 (q 2 ). r. (11) λ(1.e. s) dq 2 dx 214 π 5 mB  +− 2 ++ 2 ++ 2 −+ 2 −+ 2 2 × |M+− + | + |M− | + |M+ | + |M− | + |M+ | + |M− | −− 2 ++ 2 +− 2 −+ 2 −− 2 2 + |M−− + | + |M− | + |M0 | + |M0 | + |M0 | + |M0 | . the matrix element of the B → K ∗ `+ `− decay takes the following form    Gα ∗ tot ¯ tot ¯ )`γµ (1 − γ5 )` + (C9tot + C10 )`γµ (1 + γ5 )` M = √ Vt b Vt s (C9tot − C10 4 2π  2V (q 2 ) ρ − iεµ∗ (mB + mK ∗ )A1 (q 2 ) × −µνρσ ε∗ν pK ∗ q σ mB + mK ∗    A2 (q 2 ) 2mK ∗ + iqµ 2 (ε∗ q) A3 (q 2 ) − A0 (q 2 ) + i(pB + pK ∗ )µ (ε∗ q) mB + mK ∗ q  tot C ρ − 4 72 mb 4µνρσ ε∗ν pK ∗ q σ T1 (q 2 ) q   + 2i εµ∗ (m2B − m2K ∗ ) + (pB + pK ∗ )µ (ε∗ q) T2 (q 2 )     q2 2 ¯ (q ) `γ ` . In Eq. we face the problem of computing the form factors. Aliev et al. T2 (q 2 ). (10) T + 2i(ε ∗ q) qµ − (pB + pK ∗ )µ 2 3 µ mB − m2K ∗ From Eqs.M. A1 (q 2 ).T.. Using this matrix element and the helicity amplitude formalism (see. In what follows we will use the results of the work [32–34] in which the form factors are described by a three-parameter fit where the radiative corrections up to leading twist contribution and SU(3)-breaking effects are taken into account. A3 (q 2 ). the q 2 -dependence of any of these form factors could be parametrized as [32. (7). (9) and (10) we observe that in calculating the physical observables at hadronic level. 1 − aF s + bF s 2 where the parameters F (0). (11) where superscripts denote helicities of the leptons and subscripts correspond to the helicity of the vector meson (in our case K ∗ meson). [35– 37]). Letting  F (q 2 ) ∈ V (q 2 ). A2 (q 2 ).

77 0.13 1.   −− tot tot mb tot 0 HS .36 −0.30 1.51 0. / Nuclear Physics B 585 (2000) 275–289 Table 2 The form factors for B → K ∗ `+ `− in a three-parameter fit [32] F (0) AB→K 0 AB→K 1 AB→K 2 V B→K ∗ ∗ ∗ ∗ ∗ T1B→K ∗ T2B→K ∗ T3B→K aF bF 0.08 0. x = cos θ (θ = angle between K ∗ and `+ ).M.07 0.50 0. r = m2K ∗ /m2B . q v = 1 − 4m2` /q 2 (velocity of the lepton).94 0.47 1. Aliev et al. H M+− ± ± ± 7 9 10 2 q2 s    q2 −+ tot tot tot mb 2 C9 − vC10 H± + 4C7 2 H± . M± = (1 ± cos θ ) 2 q  −− ++ M± = − M± .19 0. s) √ + (1 + r)A1 (q 2 ) . r. 1+ r   √ mB A2 (q 2 ) 2 −(1 − r − s)(1 + r )A1 (q ) + λ(1. M0 = 2m` cos θ 2C9 H0 − 4C7 2 H0 + 4m` C10 q   q  mb tot = − q 2 sin θ 2 C9tot + vC10 H0 − 4C7tot 2 H0 . M++ ± = ± 2 m` sin θ 2C9 H± + 4C7 q2 s    q2  tot tot tot mb 2 C = (−1 ± cos θ ) + vC + 4C H .64 0.49 0.54 −0.16 q 2 = (pB − pK ∗ )2 .   ++ tot tot mb tot 0 HS .02 0. The explicit forms of the helicity amplitude Mλλ`V λ` are as follows: √  tot tot mb H± .53 1.02 0.47 1. M0 = −2m` cos θ 2C9 H0 − 4C7 2 H0 + 4m` C10 q where   √ V (q 2 ) H± = mB ±λ1/2 (1. s) H0 = √ √ . 2 rs 1+ r (12) .280 T.35 0.19 1. r. M+− 0 q +− M−+ 0 = M0 (v → −v).

existence of the new physics can be confirmed by the shift in the position of the zero of the forward–backward asymmetry. s)T1 (q 2 ) + (1 − r)T2 (q 2 ) .T. (16) R0 dq 2 dx dΓ + dx dΓ 2 2 0 dq dx −1 dq dx To obtain quantitative results we need the value of the fourth generation CKM matrix element |Vt∗0 s Vt b |. Therefore in the present work the forward–backward asymmetry AFB is considered. respectively. we consider the following ratio R= B(B → Xs γ ) . we will use the experimental results of the decays B(B → Xs γ ) and B(B → Xc eν¯ e ) to determine the fourth generation CKM factor Vt∗0 s Vt 0 b . For this aim following [20]. From Eq. (iii) The lepton forward–backward asymmetry on q 2 and on the t 0 quark mass for the fixed values of the “new” CKM factor Vt∗0 s Vt 0 b . s) A0 (q 2 ). These quantities are measurable from the experiments. Here the subscripts 0. such as (i) Γ+ /Γ− .    m2B s 2 2 2 T3 (q ) . √ s   H± = 2m2B ±λ1/2 (1. (14) where the upper(lower) subscript in Γ corresponds to M+ (M− ) and Γ0 = G2 α 2 |Vt b Vt∗s |2 214 π 5 mB Z Z  ++ 2 −+ 2 −− 2 2 × dq 2 dx vλ1/2 |M+− 0 | + |M0 | + |M0 | + |M0 | . s) T2 (q ) + H0 = √ 1−r rs HS0 = − In the present paper. (13) (1 − r)(1 − r − s)T2 (q ) − λ(1. The normalized forward–backward asymmetry AFB is one of the most useful tools in search of new physics beyond SM. Aliev et al. L and T indicate the helicities of the K ∗ meson. (ii) ΓL /ΓT = Γ0 /(Γ+ + Γ− ). r. Indeed. B(B → Xc eν¯ e ) . r. / Nuclear Physics B 585 (2000) 275–289 281 mB λ1/2 (1. This shift of zero position can be used in looking for new physics.M. we study the dependence of the following measurable physical quantities. r. In order to reduce the uncertainties arising from b-quark mass. (14) and (15) the expressions for the ratios Γ+ /Γ− and ΓL /ΓT = Γ0 /(Γ+ + Γ− ) can easily be obtained. (15) From Eqs. we can easily obtain the explicit expressions for Γ+ . (11). Γ− and Γ0 as Γ± = G2 α 2 |Vt b Vt∗s |2 214 π 5 mB Z Z  ++ 2 −+ 2 −− 2 2 × dq 2 dx vλ1/2 |M+− ± | + |M± | + |M± | + |M± | . Especially the determination of the position of the zero of AFB can predict about new physics [38]. which defined in the following way R1 R0 dΓ dx dqdΓ 2 dx − −1 dx dq 2 dx d 0 2 AFB (q ) = R 1 .

we get " s # 2 f (b Vt∗s Vt b πR|V | m )κ(b m ) cb c c SM .M. (17) mc )κ(b mc ) |Vcb |2 πf (b mc ) [39] of b → c`¯ν where the phase factor f (b mc ) and O(αs ) QCD correction factor κ(b are given by R= m2c + 8b m6c − m b8c − 24b m4 ln(b m4c ). (17) for Vt∗0 s Vt 0 b and taking into account Eqs. (19) that Vt∗0 s Vt 0 b depends on the t 0 -quark mass and its values for different choices of the t 0 -quark mass and for the experimentally measured branching ratio B(B → Xs γ ) = 3.07 −6. but with opposite sign. (2) and (18). It should be noted that the LEP result on (B → Xs γ ) [41] coincide with the CLEO result within the error limits and for this reason we don’t present LEP result.80 1. Substituting the value of Vt∗0 s Vt 0 b from Table 3.282 T. (20) is about 7. Solving Eq. and our numerical analysis confirms this expectation.52 .36 1.69 1. f (b mc ) = 1 − 8b   c  2αs (mb ) 31 3 2 2 π − (1 − m bc ) + . The superscripts (+) and (−) correspond to the respective signs in front of the square root in Eq. / Nuclear Physics B 585 (2000) 275–289 In leading logarithmic approximation this ratio can be written as |Vt∗s Vt b |2 6α|C7tot (mb )|2 . whose value is about 10−3 and one order of magnitude smaller compared to the previous case. we observe that the sum of the four terms on the left-hand side of Eq. and the error in sum of first three terms in Eq.6 × 10−2 . On the other if we consider Vt∗0 s Vt(−) 0 b .61 1. (19) is about ±0. Moreover. Aliev et al.15 × 10−4 [40] (see also [20]). Table 3 The numerical values of Vt∗0 s Vt 0 b for different values of the t 0 quark mass. κ(b mc ) = 1 − 3π 4 2 where m bc = mc /mb . However it should be noted that the data for the CKM is not determined to a very high accuracy. since Vt∗0 s Vt 0 b is very close to the sum of the first three terms. From unitarity condition of the CKM matrix we have ∗ Vub + Vcs∗ Vcb + Vt∗s Vt b + Vt∗0 s Vt 0 b = 0. Vus (20) If the average values of the CKM matrix elements in the SM [30] are used.6 × 10−2 . (19) is closer to zero compared to the SM case. In summary both Vt∗0 s Vt 0 b ∗ (−) and Vt∗0 s Vt(−) 0 b satisfy the unitarity condition (19) of CKM.75 −6. − C (m ) Vt∗0 s Vt±0 b = ± b 7 C7new (mb ) 6α|Vt∗s Vt b |2 (18) (19) It is observed from Eq. It is easy to (−) (+) see then that the value of Vt∗0 s Vt 0 b is within this error range.55 −7. (19) mt 0 (GeV) Vt∗0 s Vt 0 b × 10−2 (+) (−) Vt∗0 s Vt 0 b × 10−3 200 250 300 400 −7. Vt∗0 s Vt 0 b contribution to the physical quantities should be practically indistinguishable from SM results. the sum of the (+) first three terms in Eq. are listed in Table 3 (here we present only the central value). since |Vt 0 s Vt 0 b | 6 (+) (−) 10−1 × |Vt∗0 s Vt 0 b |.

Fig.892 GeV. for the choice of Vt∗0 s Vt 0 b . For this reason. This result (−) can be explained by the fact that Vt∗0 s Vt 0 b is of the order of 10−3 and it is one order of smaller than Vt∗s Vt b = 0. 2. Similar conclusion for the same case can be drawn for the lepton forward–backward asymmetry (see Figs. 5 and 6 depict the invariant mass distribution for the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays for this case. 6 depicts that. Numerical analysis Having the explicit expressions for the physically measurable quantities. mB = 5. mc = 1. 1–8 the ordering of the lines with respect to different values of mt 0 are the same. the . as indicated in Fig. 3. the situation changes essentially and Figs. / Nuclear Physics B 585 (2000) 275–289 283 Fig.105 GeV. However. In Figs.35 GeV. Γ+ . mµ = 0. 3 and 4). The values of the main input parameters. For completeness. The invariant dilepton mass distribution for the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays. The behavior of the differential decay width between J /ψ and ψ 0 regions differ essentially in the two considered models. when we consider the Vt∗0 s Vt(+) 0 b case.8 GeV. mτ = 1. in this section we will study the influence of the fourth generation to these quantities. with and without the long distance effects are presented in Fig. which appear in the expression for the decay widths Γ0 . (−) respectively. with and without the long distance effects.M. and at mt 0 = 200 GeV. 1. for the B → K ∗ τ + τ − decay. the effect of the fourth generation in this case is small. From these figures we see that the predictions of the fourth generation model and SM on the differential branching ratio practically coincide in this case. 300 GeV and 400 GeV. 250 GeV. here and in rest of the figures. respectively. for the choice of Vt∗0 s Vt 0 b . The dependence of the invariant dilepton mass distribution for the B → K ∗ µ+ µ− decay on (−) q 2 at different values of mt 0 . mK ∗ = 0.T. Further. Aliev et al. 1.28 GeV.78 GeV. 1 and Fig. the SM prediction is also presented.038 [30] in the SM. Γ− and AFB are: mb = 4.

four generation model’s prediction differs from that of the SM result and this departure becomes exaggerated even further as mt 0 increases. 3. for the Vt∗0 s Vt(+) 0 b case. with and without the long distance effects. The dependence of the lepton forward–backward asymmetry AFB on q 2 of the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays are depicted in Figs. 7 that the zero position of AFB is shifted in the fourth generation case . 2. 1 but for the B → K ∗ τ + τ − decay. Fig. / Nuclear Physics B 585 (2000) 275–289 Fig. for the choice of Vt∗0 s Vt 0 b . 7 and 8. The dependence of the forward–backward asymmetry for the B → K ∗ µ+ µ− decay on q 2 at (−) different values of mt 0 . Aliev et al.M.284 T. Same as Fig. We see from Fig.

respectively. due to the smallness . 4. Therefore a careful analysis of the shift in zero position of the AFB for different values of mt 0 could be an indication of the unambiguous information about the existence of new physics [38]. 1 but for the choice of Vt∗0 s Vt 0 b .M. Same as Fig. 9 and 10 we investigate the the dependence of the ratios Γ+ /Γ− and ΓL /ΓT on q 2 for both roots of the Vt∗0 s Vt 0 b . Fig. Aliev et al. / Nuclear Physics B 585 (2000) 275–289 285 Fig. Same as Fig. (+) compared to that of the SM prediction. 3 but for the B → K ∗ τ + τ − decay. 5. at different values of the mt 0 for the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays. As have already been noted.T. In Figs.

the effect of the fourth generation in the ratio ΓL /ΓT is depicted in Fig. the new physics. (+) Fig. (+) of the Vt∗0 s Vt(−) 0 b . Same as Fig. On the other hand. 9 that for the Vt∗0 s Vt(+) 0b essential departure from the SM result is quite obvious in the B → K ∗ µ+ µ− decay for the ratio Γ+ /Γ− . 10.286 T. 3 but for the choice of Vt∗0 s Vt 0 b . So. Same as Fig. 2 but for the choice of Vt∗0 s Vt 0 b . / Nuclear Physics B 585 (2000) 275–289 Fig. these ratios practically display the same behavior as predicted by SM for this choice. whose behavior clearly shows a strong dependence on the value of . this case can yield no information about case. Aliev et al. Moreover. confirming our expectation. 7.M. it follows from Fig. 6.

Aliev et al. the terms proportional to the lepton mass can give considerable contribution. Finally we would like to note that 4 × 4 CKM matrix contains three CP-violating phases . / Nuclear Physics B 585 (2000) 275–289 287 Fig. the mt 0 . for the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays.M. 9. The dependence of the ratio Γ+ /Γ− on mt 0 for both roots of the Vt∗0 s Vt 0 b .T. (+) Fig. Such a dependence can be explained as an indication of the fact that for ΓL /ΓT . Same as Fig. 4 but for the choice of Vt∗0 s Vt 0 b . 8. especially in the B → K ∗ τ + τ − decay.

D 43 (1991) 3669. B-TeV and LHC-B machines and is very sensitive to the various extensions of the Standard Model. W. To summarize. Turke et al. Phys. while the measurement of the ratio ΓL /ΓT in the B → K ∗ τ + τ − decay seems to be very informative in searching new physics. τ ) decay has a clean experimental signature and will be measured at the present asymmetric B factories and future hadronic HERA-B. Stuart. C.288 T. In the present work this decay is studied in the SM with the four generation model. Phys.G. DELPHI Collaboration. Hewett. Anselm et al. Bamert.S. and hence CP violation might be sizeable. 10. 62 (1989) 719. Nucl. Phys. B 278 (1986) 905. We will discuss this problem elsewhere in one of the future works. X. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] P. Eilam. G.G. Hou. Rev. Z. It is found out that for the choice of the positive root of the factor Vt∗0 s Vt(+) 0b . Phys. R. Nucl.G. . / Nuclear Physics B 585 (2000) 275–289 Fig. Rev. N. Phys. Phys. Lett. Lett. the measurements of the invariant mass distribution. Phys. B 340 (1994) 81. Phys. for the B → K ∗ µ+ µ− and B → K ∗ τ + τ − decays. Rev. Bigi. Phys. Abreu et al. P. The two solutions of the fourth generation CKM factor Vt∗0 s Vt 0 b have been used. A. J. I.. T. A. Lett.L. Evans.M. Aliev et al. Eilam. Lett. Rev.. G. B 156 (1985) 103. S. C 66 (1995) 495. lepton forward–backward asymmetry and the ratio Γ+ /Γ− could be quite efficient in establishing the fourth generation in the B → K ∗ µ+ µ− decay. U. B 274 (1992) 23. B 258 (1985) 103. Burgess. D 41 (1990) 875.P. Z. the exclusive rare B → K ∗ `+ `− (` = µ. Paksava. Soni. Rizzo. The dependence of the ratio ΓL /ΓT on mt 0 for both roots of the Vt∗0 s Vt 0 b . C 27 (1985) 303.. B. Phys. Phys. Haeri. D 34 (1986) 2773. He.

B 393 (1993) 23. Nucl. Phys. [40] S. P. Misiak. Y. Demir. Phys. Kim. Soni. Phys. B 192 (1987) 441.M. Erler. / Nuclear Physics B 585 (2000) 275–289 289 [11] T. Mod. Kim. Prog. hep-ph/9806471.B. Burdman. B 327 (1989) 569. Pois. Handoko. [17] T.A. Wise. D 55 (1997) 7059. Grinstein. C 3 (1998) 90. Martin. [38] G. B 193 (1987) 327. Rozanov et al. Y. Phys. [20] C. V.M. A 11 (1996) 687 (Erratum). [41] R. Phys. Ball. Okun. M. Phys. Aliev. D 58 (1998) 094006. Novikov. J. Hewett.I. [14] J. Braun. W. D.L. Nucl. M. Buras. Phys.M. Goto. Burato et al. Mod. P. Phys. Lett.H. Ali. Lett. Phys. [15] M..T. G. Huo. Okada. [25] M. [13] V. Gunion. Y. Inami et al. M. D 58 (1998) 094016. Pak. Phys. L. [16] J. Wu. Vainshtein. A 14 (1999) 2453. Lett. Rev. [12] A. Lett. Phys. Phys. J. Lett. Deshpande. Rev. B 319 (1989) 271. D 62 (2000) 014026. Lett. Phys. Phys.. [34] T. Zhu. Ball. Phys. Aliev. G. D 52 (1995) 186. L. Phys. N. [28] C.I. [27] N.. Shimizu. M. M.S. Lett. D 61 (2000) 015011. Sov. [35] K.F.J. A 10 (1995) 1915.A. Wade. Özpineci. Okun. [30] C. D 61 (2000) 074024. M.N. CLEO Collaboration. 62 (1999) 1275. Schuler. [33] P. Rev. Rev.J. Phys. . Mod. [24] A. Buras. Savcı. H. B 389 (1996) 83. Sanda. Ponose. Nucl. Phys. [29] A.I. Eur. A 10 (1995) 1471.D.J.G.. [23] W. G. Zakharov. Nucl. A. Aliev. Y. [31] A. Phys. A.F. ALEPH Collaboration.A.. McKay. [22] J. B 439 (1995) 461 (Erratum). Nucl.M. hep-ex/9908022. Rev. J. Phys. B 218 (1989) 343.J. Phys. Phys. Rev. Aliev. Langacker.L. Rev. ˙Iltan. A.K. Lett. Körner. Hiller. C 3 (1998) 1. Lett. M.. [36] J. S. 25 (1999) 989. Morozumi. JHEP 9809 (1998) 005. C 46 (1990) 93. Maltoni et al.. D. Phys. Steger. [32] P. Mod. [19] T. [26] B. A. H. D 57 (1998) 4254. Caso et al. L. Lett. Savage. Phys. Tanaka. Trampetic. C. [18] C. Phys. Lim. Hou. A. Ball.G. Rep.G. Lett. J. [37] T. [39] A.M. Aliev et al. B 429 (1998) 169. Phys. Rev. Ahmed et al. D 39 (1989) 1461.B. Hagiwara. Rev.T. Huang. J. Phys. Eur. 24 (1976) 427. Masiero et al. Phys. Rev. V. Münz. B 476 (2000) 107.B.S. Phys. Huang. Z. K. [21] T. T. Phys. B 355 (1995) 329.M.S.S. D 51 (1995) 201. Phys.S. Shifman. E.

. pseudoscalar masses and π–π scattering. At low energies it is possible to obtain a family of theories within a low-energy expansion using only symmetry principles from QCD.∗ . P. All rights reserved. This family of low-energy theories is known as Chiral Perturbation theory (CHPT) [1]. Amorós a. in order to compare with experiment it is very important to have calculations performed to as high an order as possible. 14. J. 12.se (J.40.Yx.V. PACS: 12.nl/locate/npe K`4 form-factors and π–π scattering G. The perturbation expansion is in terms of the energy. Finland b Department of Theor. to one-loop [4. Chiral Perturbation theory. Bijnens).7 . collectively denoted as p. 1 is one of the several processes that can be calculated in the framework of CHPT. PII: S 0 5 5 0 . 0550-3213/00/$ – see front matter  2000 Elsevier Science B. accepted 30 May 2000 Abstract The F and G form-factors of K`4 and the quark condensates are calculated to O(p6 ) in Chiral Perturbation theory (CHPT).Fe.lu.b . We discuss the consequences for the vacuum expectation values. It has been calculated to tree level.Ff.39.elsevier. University of Helsinki. 1 The early history of this decay is reviewed in [3]. The Ke4 decay. Introduction The low-energy realization of QCD is still disconnected from the high-energy structure where perturbative QCD applications have made it clear beyond doubt that QCD is the theory of the strong interaction. 12.  2000 Elsevier Science B. Since this is a perturbative framework. Phys. decay constants. Box 9. S-22362 Lund. Bijnens b. Sölvegatan 14A. In addition Ke4 is the main input in determining three of the CHPT ∗ Corresponding author. Sweden Received 24 March 2000.Nuclear Physics B 585 (2000) 293–352 www. P. All rights reserved. More precise Ke4 experiments in progress will go beyond the precision of these calculation.40. momenta and meson masses. Full formulas are presented as much as possible. Kaon decay 1.3 2 1 3 ( 0 0 ) 0 0 3 6 6 .15. CHPT in the three flavour sector was firmly worked out in [2] where all parameters were determined to O(p4 ).-n Keywords: Chiral symmetry. E-mail address: bijnens@thep. University of Lund. FIN-00014 Helsinki. revised 26 May 2000.O.V. A full refit of most of the O(p4 ) CHPT parameters is done with a discussion of all inputs and underlying assumptions.5] and to improved oneloop [6]. Talavera b a Department of Physics.

1. The appendices provide explicit formulas for the form-factors. in other words. Using previous two-loop results [10] a complete three-flavour and. At present.4 and 6.5. In addition we also present the calculation of the vacuum expectation values to two-loops. but in this case the convergence is not so obvious and we can ask the question: is the Kaon mass small enough to allow a perturbative theory? To answer this we need a large spectrum of calculations to two-loop accuracy.1. the two-flavour effective Lagrangian can be applied in a small region of physical phenomena. it is the two-flavour sector where chiral symmetry predictions are expected to work best because of the clean feature of pseudo-Goldstone bosons for the pions. The comparison with a proposed experimental parametrization is done in Section 6.2 introduces the standard notation for the form-factors followed by a short overview of CHPT. Previous theoretical results are described in Section 3. Also. as far as possible. The precision of the prediction on the latter process is now such that these CHPT parameters are the main uncertainty thus requiring an extension of the calculation for Ke4 as well. [7–9].3 updates the π –π scattering threshold parameters with our fit for the LEC as well as comparisons with available low-energy data. . We also discuss all the inputs and underlying assumptions there and compare with previous results and a few models. On the other hand. there is some work to try to estimate the Ke4 two-loop contribution [6] and there exist several complete two-loop results [11– 18] for other processes. This two-flavour framework seems to be the best one to extract the low-energy parameters of the pion–pion scattering. with the other light flavour and the excited states contributing only through the local constants in the CHPT Lagrangian. present full results and discuss more implications. / Nuclear Physics B 585 (2000) 293–352 parameters. Section 6.1 we present the kinematics with the description of the process and the variables. masses and decay constants with the ratio of the quark masses in Sections 6. Details about the O(p4 ) constants (LEC).1. Here we extend the discussion. We presented a short summary of the main results with a slightly different input in [19]. we discuss the behaviour of the vacuum expectation values.2. A lot of work has gone into improving the calculation of π –π scattering in CHPT. Using the LEC fit.2. Section 6. We discuss several model dependent assumptions. Finally we shortly summarize the results.2. The resulting theoretical form-factors and partial waves are shown in Section 6. There are several papers where a complete two-loop calculation is done. Section 2. A more interesting Lagrangian has to include the third flavour. since the relevant quark mass is so small. mainly standard. Section 4 and in Section 5 some of the O(p6 ) constants in CHPT are estimated using resonance saturation. to increase confidence in the generality of the numerical results. method and checks. Section 2. This is the purpose of the present work: one further step in the perturbative CHPT series of Ke4 or.4. Most of them with the two-flavour chiral Lagrangian as the effective theory. fits and errors are in Section 6. Next we discuss the calculation. with other states starting to modify the behaviour at higher energies. besides results to tree and one-loop level. and a discussion of the available data in Section 2. Amorós et al. vacuum expectation values and the one-loop integrals.294 G. In Section 2. to obtain a two-loop result. a model independent result is obtained.3.

Amorós et al. / Nuclear Physics B 585 (2000) 293–352 295 2.1. Following the original work [20] we consider three reference frames. 1): (i) sπ . For the last two processes we keep the momentum notation of the first. the Kaon rest frame. K 0 (p) → π − (p+ )π 0 (p− )`+ (p` )ν` (pν ). (2. √ m2` 6 s` = (p` + pν )2 6 (mK − sπ )2 . the angle between the π + and the dipion line of flight with respect to the Kaon rest frame. (iii) θπ . related through Fig. T 00 and T −0 . 0 6 θπ . Kinematics We shortly review the variables that parametrize the processes K + (p) → π + (p+ )π − (p− )`+ (p` )ν` (pν ). (iv) θ` . The amplitudes for the three processes we denote by T +− . we also use tπ = (p+ − p)2 and uπ = (p− − p)2 . Definitions and the data 2. Kinematical variables in K`4 decays. These variables can vary inside the range 4m2π 6 sπ = (p+ + p− )2 6 (mK − m` )2 . (2.1) In brackets we have labelled the corresponding momentum.2) Instead of using the previous variables. the dipion center-of-mass system (π−π plane) and the dilepton center-of-mass system (e−ν plane). and besides sπ . 0 6 φ 6 2π.G. cos θ` and φ.3) . respectively. 1. (2. the squared effective mass of the dipion system. (ii) s` . θ` 6 π. (v) φ. the squared effective mass of the dilepton system. The K`4 decays can be parametrized in terms of five kinematical variables (see Fig. the angle between the `+ and the dilepton line of flight with respect to the Kaon rest frame. the angle between the π−π and e−ν planes with respect to the Kaon rest frame. K + (p) → π 0 (p+ )π 0 (p− )`+ (p` )ν` (pν ).

easier to obtain from the experiment. T +− = √ Vus 2 (2. The other two amplitudes. λ(m1 . m3K  i  (p+ + p− )µ F + (p+ − p− )µ G + (p` + pν )µ R . can be found in [22. The relations between the form-factors and the intensities. 2 q σπ = 1 − 4m2π /sπ . Here we are interested in the F and G form-factors. T −0 is antisymmetric under the interchange of the pion momenta while T 00 is symmetric. This also implies relations between the form-factors themselves. (2. Aµ = − mK (2.23] or in [6].9) with T ij the matrix element defined in Eq. s` and cos θπ only. are defined similarly. (2. or alternatively of sπ . (2. The main feature is the spontaneously broken chiral symmetry. T −0 and T 00 . The H form-factor is already known to O(p6 ) [21] and the R form-factor always appears with the m2` factor and is negligible for the electron case. Form-factors With the previous notation the amplitude for the decay K + → π + π − `+ ν` is GF ∗ u(p ¯ ν )γµ (1 − γ5 )v(p` )(V µ − Aµ ). Chiral perturbation theory CHPT is an expansion in the energy. The amplitudes for the three processes of Eq. sπ . m3 ) = m21 + m22 + m23 − 2 (m1m2 + m1 m3 + m2 m3 ). The pseudoscalar fields are the lowest mass excitations of the vacuum expectation values.8) Vus is the relevant CKM matrix-element. / Nuclear Physics B 585 (2000) 293–352 sπ + tπ + uπ = m2K + 2 m2π + s` .7) (2. R is known to O(p4 ) [6]. m2 . Several realizations . tπ − uπ = −2σπ X cos θπ . Observe the different phase convention in the isospin states compared to [6] where T 00 appears with a minus sign because of the Condon– Shortley phase convention.2. tπ and uπ . The form-factors are functions of sπ . 2.3.5) 2.6) with Vµ = − H µνρσ (p` + pν )ν (p+ + p− )ρ (p+ − p− )σ .1) are related using isospin by T −0 T +− = √ + T 00 2 (2. Amorós et al. (2.296 G. with the light mesons being pseudo-Goldstone bosons.6).4) with  1 X = λ1/2 m2K . s` .

energies or masses. collectively denoted as p. The pseudoscalar fields are included in √ (2.   (2. the non-linear realization is more useful with the power counting present explicitly. the Lagrangian starts as Leffective = L2 + L4 + L6 + · · · = L2 + 10 X i=1 Li O4i + 2 X i=1 ei + Hi O 4 90 X i=1 Ci O6i + 94 X ei + · · · . + L10 U † Fµν (2.13) and 2 L4 = L1 Dµ U † D µ U + L2 Dµ U † Dν U D µ U † D ν U + L3 D µ U † Dµ U D ν U † Dν U + L4 D µ U † Dµ U χ † U + χU † 2 + L5 D µ U † Dµ U (χ † U + U † χ) + L6 χ † U + χU † 2 + L7 χ † U − χU † + L8 χ † U χ † U + χU † χU † R µ L D U D ν U † + Fµν Dµ U † Dν U − iL9 Fµν R Rµν R L Lµν U F Lµν + H1 Fµν F + Fµν F + H2 χ † χ . / Nuclear Physics B 585 (2000) 293–352 297 are possible. 4 (2. The notation for the covariant derivative is Dµ U = ∂µ U − irµ U + iU lµ . In the following we include the basic notation and terms we are working on. It contains 90 + 4 terms compared to the 10 + 2 of L4 .10) U (φ) = u(φ)2 = exp(i 2 Φ/F0 ). Amorós et al. The expressions for the first two terms are (F0 is the pion decay constant in the chiral limit) L2 = F02  Dµ U † D µ U + χ † U + χU † .11) 2 η8 −√ 6 With the chiral symmetry constraints and the transformation properties under this symmetry as well as Lorentz invariance and charge conjugation invariance.15) . where  π0 η8 √ +√  2 6  E λ  E − Φ(x) ≡ √ φ =  π  2  K− π+ K+ η8 π0 −√ + √ 2 6 0 K¯ K0     .14) The next-to-next-to-leading order L6 is a rather long expression and can be found in [24]. Ci O 6 (2. (2.12) i=91 where the subscripts stand for the power of the low momenta.G. Explicit chiral symmetry breaking terms are included as perturbations of the symmetry. For an extensive discussion about CHPT we refer the reader to [1] and references therein.

µν FR = ∂ µ r ν − ∂ ν r µ − i[r µ . 9 and 10 in Section 6.16) Their field strength tensors are µν FL = ∂ µ l ν − ∂ ν l µ − i[l µ . M = diag(mu . For later use we define the following quantities µν µν  v µν = FR + FL /2. (2. G = geiδp + d-wave. Available K`4 experimental results Let us briefly comment on the status of the present data.20) 2. We confirm this using our calculation as shown in Figs. / Nuclear Physics B 585 (2000) 293–352 with lµ and rµ the left and right external currents related with the vector and axial-vector external currents rµ = vµ + aµ . Neither was there dependence on sπ found for the reduced form-factor g (2. Within the experimental sensitivity no dependence of the form-factors on s` nor on d-waves was observed.21) where fs .  uµ = i u† (∂µ − irµ )u − u(∂µ − ilµ )u† . 1 Γµ = u† (∂µ − irµ )u + u(∂µ − ilµ )u† . (2. (2. In the experiment [25] the data are analyzed with the partial wave expansion F = fs eiδs + fp eiδp cos θπ + d-wave. lµ = vµ − aµ .2. As we show below. l ν ]. ms ). 12.298 G. (2. (2.19) M stands for the diagonal quark mass matrix.18) where the explicit masses are included through the scalar current s s = M + ···.5]. r ν ]. (2.4. 2 χ± = u† χu† ± uχ † u. The form-factor fp was found to be compatible with zero and consequently neglected when the final value of g was derived. The expected errors in the planned experiments will allow to see the s` dependence. fp and g are assumed to be real. These two dominate in statistics and precision over previous ones. There are two experiments [25. Earlier experiments are within errors compatible using isospin relations with these two and are discussed in [4. Fig. md . fs . this is a borderline assumption even for the present sensitivity. In fact our main results rely entirely on [25] and [26] is used to give an idea on the experimental uncertainties involved. Amorós et al. ∇µ X = ∂µ X + Γµ X − XΓµ .17) Scalar (s) and pseudoscalar (p) external fields are introduced with χ = 2B0 (s + ip). µν µν  a µν = FR − FL /2. 26] on which our analysis is based.22) g¯ = .

.5] F and G come through the topologies shown in Fig. (2.77 ± 0. containing vertices of L2 in (b). As is clear. (a) One-particle irreducible tree level diagram. with the same slope as the fs form-factor.14. The assumption of constant g¯ constrains the g form-factor parametrization to be  (2.220 the values of [25] yield for the form-factors and slope at threshold fs (0) = 5.  q 2 = sπ − 4m2π /4m2π . 2 — both the F and G form-factors are equal and given by a single insertion of L2 [28] mK .50 ± 0. λ = 0. Dots refer to strong vertices or current insertions from L2 . Amorós et al. g(0) = 4. At one-loop [4. Using the PDG averaging methods [27] on both results we obtain g(0) = 4. 9 the theoretical slopes are different but the effect still remain within present errors.. (2. Furthermore fs (sπ ) was parametrized as fs (sπ ) = fs (0)(1 + λf q 2 ).50.27. (c) or a single vertex of L4 in (a). Internal lines are pseudoscalars only.26) thus both experiments are marginally compatible. λ = λf = λg . Using Vus = 0. Form-factors at next-to-leading order For completeness we start with the calculation up to one-loop precision. from Fig. (2. 2. 3.31. i. 2.27) that we use in fit 9.G.24) g(sπ ) = g(0) 1 + λg q 2 .25) From [26] the same assumptions and using isospin give g(0) = 5. L4 or L6 .59 ± 0.23) Notice that fs is assumed to have only a linear dependence in sπ . (b) One-particle irreducible one-loop diagrams. The result can be cast in the general form Fig. As shown in Fig. (3.02.e.08 ± 0. External legs stand for pseudoscalar or weak current.1) F =G= √ 2 Fπ where mK is the physical Kaon mass. The linear dependence can only be ensured for sπ < 0. 9 future experiments might detect curvature.93 ± 0. / Nuclear Physics B 585 (2000) 293–352 299 thus the latter is assumed to be constant. At tree level — see diagram (a) in Fig.11 GeV2 . (2.

uπ ) = √ 1+ + O(p6 ). (3. tπ ). Lr4 . tπ ) + (2m2π + m2K − uπ )/3B(m ¯ 2π . m2K . m2K . m2π . m2π . Amorós et al. m2η . (3.4) It is worthwhile to point out at this level some relevant features in Eqs. m2η . sπ ) − B¯ 22 (m2π . For instance they depend on the L4 constants Lr1 . m2K . tπ ) − 2B¯ 22 (m2π . uπ ) + (m2π + m2K − tπ )/4B(m  ¯ 2K . m2η . tπ ) + (−3m2π + 3m2K + tπ )/4B¯ 21 (m2K . sπ ) ¯ 2π ) + 6A(m + 7/24 A(m ¯ 2π . Lr2 . / Nuclear Physics B 585 (2000) 293–352   F1-loop mK F (sπ .3) and for the G form-factor to G1-loop = m2π 4(−Lr3 + Lr5 − Lr9 ) − 2m2K (2Lr3 + Lr9 ) + 2sπ Lr9 + 2tπ (4Lr2 + 2Lr3 + Lr9 )  ¯ 2π ) + 14A(m ¯ 2K ) − 3A(m ¯ 2η ) + 2uπ (−4Lr2 + Lr9 ) + 1/24 35A(m ¯ 2π . m2K . m2K . sπ ) + (3m2π − 3m2K − 5/3tπ )/4B¯ 21 (m2π . m2K . uπ ) + (m2π /3 − 7/3m2K + tπ )/4B¯ 1 (m2K . m2K . tπ ) − uπ /3B¯ 1 (m2π . (ii) A purely logarithmic piece coming from the tadpole. m2K .300 G. uπ ) + (−m2π − m2K + tπ )/4B(m + (−m2π − m2K /2 + 2tπ )/3B¯ 1 (m2π . sπ ) + (2m2π + sπ /3)B¯ 1 (m2π . Provided we . A straightforward calculation [4. 2 — given by the functions B¯ i — see Appendix D — some of which develop an imaginary part responsible for the I = 0 s-wave. tπ . (3. sπ ) − 8/3B¯ 22(m2π . making the full process scale-independent while the remaining finite piece is tuned to match some experimental observables. m2η . I = 1 p-wave π –π phase-shift. m2K . Fπ2 2 Fπ   G1-loop mK (3. m2K . tπ ) − 5/2B¯ 22 (m2K . m2π . Lr5 and Lr9 . uπ ) = √ Fπ2 2 Fπ F1-loop (G1-loop ) contain three possible contributions: (i) a unitary correction generated by the loop graphs — see diagram (c) in Fig. 2. G(sπ . Lr3 . tπ ) − 2/3sπ B¯ 21 (m2π . tπ )/2. tπ )/3 − B¯ 22 (m2K . tπ ) + (3m2π − 3m2K − tπ )/4B¯ 21 (m2K .3). m2η . sπ ) + sπ /2 B(m + (m2π + m2K /2 − 2tπ )/3B¯ 1 (m2π . sπ ) − B¯ 22 (m2K .2) 1+ + O(p6 ). (b) diagram in Fig. sπ ) + (−m2π /3 + 7m2K /3 − tπ )/4B¯1 (m2K . m2π . m2K . (3. m2K . m2π . (iii) A polynomial part which split in two pieces.5] leads to F1-loop = 4m2π (−16Lr1 + 4Lr2 − 3Lr3 + 8Lr4 + Lr5 − Lr9 ) + 2m2K (8Lr2 + 2Lr3 − Lr9 ) + 2sπ (16Lr1 + 4Lr3 + Lr9 ) − 2tπ (4Lr2 + 2Lr3 − Lr9 ) + 2uπ (−4Lr2 + Lr9 )  ¯ 2K ) + 3A(m ¯ 2η ) + (−3/2m2π + sπ )B(m ¯ 2π . The first one cures the scale dependence of the one-loop functions. tπ ) − 2/3B¯ 22 (m2π . m2K . m2η . m2η . m2K . tπ ) − uπ /3B¯ 1 (m2π . tπ ) + 3/4(−m2π + m2K + 5/9tπ )B¯ 21 (m2π .4). uπ ) + sπ /2 B¯ 1 (m2K . tπ . tπ ) + (2m2π + m2K − uπ )/3B(m ¯ 2π . sπ ) + B(m ¯ 2η . m2K .

For reasons explained later. Lr2 and Lr3 evaluation is to consider an Omnèsrepresentation for both form-factors. (3. This together with the discrepancy between the SU(2) CHPT constants l¯1 and l¯2 obtained from the K`4 fit [6] and Roy equation analysis at O(p4 ) [30] and O(p6 ) [31] motivated us to calculate the two-loop contribution to K`4 ..G.1) should be taken. Both showed that in this particular case the next order is expected to be quite sizeable. In fact these one-loop corrections are rather large making the low-energy constants evaluation sensitive to variations of the treatment of Fπ and pseudoscalar masses. (4. We summarize here the main steps and checks. / Nuclear Physics B 585 (2000) 293–352 301 make some assumptions (large Nc and neglecting d-waves) we can use the low-energy data together with Eqs. For any O(p6 ) contribution the masses m20 can be used since the difference is O(p8 ). (3. However for O(p4 ) contribution the masses with the first two terms of Eq.4) to obtain values for Lr1 . Amorós et al. in a one-loop calculation we can safely write all the masses as physical because the difference is in the next order. One has to be much more careful in precisely defining quantities.3). This is the reason why we have to keep in mind which are the masses to consider in the two-loop result. (3. to calculate the dispersive contribution from the next chiral order considering that no intermediate new channels are opened [6]. Lr2 and Lr3 . Form-factors at next-to-next-to-leading order 4.1) with m0 the lowest-order mass and where the superscripts (4) and (6) stand for O(p4 ) and O(p6 ). One-loop: We reproduce the one-loop result obtained some time ago [4.3). In CHPT lowest-order and renormalized masses are finite and then the result can in principle be obtained with lowestorder masses appearing inside the loops and in the chiral symmetry breaking terms. renormalization group analysis allows to fully calculate the double logarithm contribution [29].5].4) have also acquired a different energy dependence that brings the one-loop result nearer to the experimental values [25].g. (3. Even more. For the tree level the three terms are needed. In other words. General technique The calculation is done with dimensional regularization (DR) with the extended dimension d = 4 − 2. Once we have all the mathematical tools to calculate all the integrals. In this scheme the chiral symmetry is preserved and because of the mass independent regularization the divergences do not mix different orders in the chiral expansion. Masses and decay constants: A two-loop calculation is a step beyond and with some differences with the one-loop calculation.1. E. Eqs. The masses can be written as: m2 = m20 + m2(4) + m2(6) + · · · (4. in the expansion with the DR parameter  we keep the O( 1 ) term. 4. A possible improvement in the Lr1 . but we . the process to obtain the form-factors is quite general.

/ Nuclear Physics B 585 (2000) 293–352 have to use the full Eq. the double poles in  can be obtained from a one-loop calculation. Therefore the result should be a product of at most twopoint one-loop integrals and the naive three-point one-loop integrals from diagrams (e) and (g) must cancel when (c) is rewritten in terms of physical masses.36] and also useful to recalculate the sunset-integrals.12)). These can be calculated in general [29] or from the Li ×one-loop graphs for a particular process. in general products of one-loop contributions: one-loop× one-loop) where A(m2 ) is a one-loop integral. can not be expressed as contributions from terms in the effective Lagrangian. Once the Li constant is renormalized Li = Lri + Γi λ. They appear in the separate contributions and cancel in any quantum field theory. . We choose to write all the masses in terms of the physical ones..2) + C and with Γ and C real and finite numbers. the sunset-integrals are calculated using the results obtained in [33. so it taking place.π the renormalization constant for the fields including O(p6 ). 1 2 (4. Irreducible two-loop integrals: The diagrams in Fig. . with ZK. Another kind of divergences are non-local. This check is done below the three-particle thresholds. . A new feature of the two-loop calculations is that the expansion in the DR parameter until O( 1 ) needs to be kept. (2. Three-point one-loop integrals: The contribution from the first four diagrams in the Fig. Amorós et al.36] for a longer discussion. (4. For the vertex-integrals we use another technique explained in [35. (d)–(g). 4 generate sunset-integrals and vertex-integrals already known in the literature. The first one is pure polynomial. Li Bj . The reason for that is the existence of terms as Li A(m2 ) (or as Li Lj . That means the matrix element is multiplied by ZK Zπ1 Zπ2 . is a strong check of the result. These divergences were already found in [32]. 2.34]. we choose also to write the physical decay constant Fπ instead of the lowest-order one F0 . This is equivalent to include the one-particle-reducible diagrams. Additionally. We obtain the same result as [29] for the double logarithms and the products Li × Lj and Li × log. In a previous work [10] on the masses and decay constants of the pseudoscalar mesons to two-loops. Cancellation of divergences: There are two kind of divergences to cancel. there appears a new finite Double logarithms: Because the nonlocal divergences have to cancel. But this is a global cancellation.1) for the squared momenta.e. This provided another check on our integral calculations. We refer to [35. The same discussion can be done for the decay constants. could be obtained through the renormalization of the meson line in diagrams (b) and (c) of Fig. they cancel with the renormalization of the Ci constants (see Eq. Wave-function-renormalization (WFR): We have to include the WFR for each external 1/2 1/2 1/2 leg. . i.302 G. 3. in the vertex-integrals we obtain the same imaginary part from the two-particle discontinuities as using the Cutkosky rules. as is the case for our calculation. . where λ = term.

3. the correction to the O(p6 ) using the physical masses instead of the bare masses is O(p8 ). The result is a variation roughly of 10% of the tail of the O(p6 ) part of the form-factors in Fig. Amorós et al. 4. As explained in the previous paragraph about masses.5. 5. This particular effect is mainly due to terms containing (3m2η − 4m2K + m2π )/(uπ . / Nuclear Physics B 585 (2000) 293–352 303 Fig. Then. Satisfying these relations is another check of our result. 1+ F (sπ . to two-loop order it is safe to use the GMO relation 3m2η = 4m2K − m2π . In addition there are the diagrams (b) and (c) of Fig. sπ ) which appear when reducing the integrals to a more minimal basis. Gell-Mann–Okubo (GMO) relation: The last comment in this subsection is related to effects from O(p8 ).16 GeV2 ) is not modified. Results Up to O(p6 ) we split the contributions as follows   F1-loop mK 1 + 4 (FLL + FV + Fct ) + O(p8 ). One-particle irreducible O(p6 ) diagrams with irreducible two-loop integrals. We can calculate with and without this relation. One-particle irreducible O(p6 ) diagrams with only one-loop integrals.2. 2 with one of the dots replaced by a O(p4 ) vertex. The dots are O(p2 ) vertices except the top dot in (d) and (e) which is an O(p4 ) vertex. uπ ) = √ Fπ2 Fπ 2 Fπ . tπ . tπ . Isospin relations: Isospin symmetry relates the decays as discussed in Section 2. The plot in the kinematic regime accessible for the experiment (sπ < 0.4 and 6.G. 4.2. The dots are O(p2 ) vertices. Another example of this type of effects is discussed in Sections 6. Fig.

We present the results for F (G)LL and F (G)ct in Appendixes A. The diagrams in the Fig. Notice that in all cases we quote F and G as for K + → π + π − `+ ν` .2. 0.3) The two first terms have been explained in Section 3. 1+ LL V ct Fπ2 Fπ4 2 Fπ (4. Plotted are the fit with lines and the full calculation with +. in particular all dependence at O(p6 ) on the Lri is collected here. which are not likely to change significantly. A. a simple parametrization is sufficient. it contains the pure two-loop unitarity contribution. We observe a fairly strong numerical cancellation between the contributions from irreducible vertex-integrals and the rest of the terms in FV and GV . tπ . For the vertex diagram we refer to [35. F(G)ct is the part of the polynomial contribution depending on the couplings of L6 . 3 can be written as product of one-loop integrals and evaluated with the standard methods. 4 we profit from previous work with the sunset [10] and the vertex-integrals [35. For F (G)V a significantly longer expression is found. respectively. mK + and mπ 0 for K + → π 0 π 0 `+ ν` and mK 0 and mπ av = (mπ + + mπ 0 )/2 for K 0 → π − π 0 `+ ν` . Amorós et al.304 G. We present it for three sets of masses for the three possible decays.36] since we do not explicitly present any formulas containing them. 0. ±1 and s` = 0.0225 GeV2 . Since it only depends on the physical masses. The parametrization is given by . Using mK + and mπ + for K + → π + π − `+ ν` . F(G)LL is the remaining part.   G1-loop mK 1 G(sπ . uπ ) = √ + (G + G + G ) + O(p8 ). The method to manage the sunsetintegrals is extensively explained in [10] and we refer to this work. Taking the even or odd part in cos θπ then gives the form-factors for the other decays.01. In all cases we plotted cos θπ = 0. F (G)V depends only on the physical masses of the pseudoscalars.36]. The O(p6 ) part is split as follows: F(G)V the part from diagrams with vertices from only the lowest-order Lagrangian. Fit to the FV and GV contribution. / Nuclear Physics B 585 (2000) 293–352 Fig. For the diagrams in the Fig.1. 5.

9589 −i0.519 −0.29854 −i0.6231 +i7. The data are also not sufficient to determine all of them so we need to estimate using some other method. mπ + 1.6824 +i17.4861 −i0.5074 −i0.9431 −i0.659 −9.7479 3.5262 +i12.90367 +i0.14408 −i1.38] to the present case.23 0.6007 −2.5226 −206.5087 2.06771 3.51 −0.1682 −215.92331 1. The values of the parameters are given in Table 1. there is a large correlation expected between these effective constants.7286 −7.17035 −5. 5. Amorós et al.035941 +i14.803725 −3. p p where X = 1/2 (m2K − sπ − s` )2 − 4sπ s` and σπ = 1 − 4m2π /sπ depend on sπ and s` .5445 +i14.15502 −8.24452 8.3974 −30.78072 0.8875 −23.91997 −i0.041 −i0.7041 −3.4139 −5.00198 −i0.5732 −7.1122 −i0. QCD in our case.4238 0.5113 −0.28955 +i4.7514 −2.1053 8. 5 we show how well this parametrization fits the calculated expressions.6377 −0.1061 −1. We assume saturation by the lightest vector.000496 0.21235 +i11.3521 54.6368 3. Of course in principle they are calculable from the underlying theory but this is not possible at present.068565 −1. Their main contribution is expected to come from the exchange of higher mass resonances.254 0.4695 50.7268 −3.59204 −3.4592 −31.8042 −3.294 3.5278 22.901 −i0.060766 0.562 −i0. With higher orders in the chiral Lagrangian the number of constants to estimate increases.816 −0.6144 7. For these we use a Lagrangian with .1912 −7. mπav 1.1412 FV GV mK + .6143 −9. In Fig.88056 +i0.7610 FV GV mK + .0270 −1.17 −3.27801 +i6.4) + sπ3 (a7 + cos θπ a8 ) + σπ X(a9 + cos θπ a10 ).000468 0. axial-vector and scalar mesons. mπ 0 0.653 0.456 0.837 −30.7821 −0.0970 F (G) = (a1 + cos θπ a2 )(1 + s` a11 ) + sπ (a3 + cos θπ a4 ) + sπ2 (a5 + cos θπ a6 ) (4.946 +i2. extending the formalism used in [37.0277 −2.09106 −0. Estimates of some O(p6 ) constants In this section we estimate some of the O(p6 ) constants that appear in the results.26763 3.G. This is due to the fact that the chiral Lagrangian is constructed using only some general conditions and can describe all the possible theories compatible with these.037925 +i2.02308 +i6.54512 +i5.810 +i2.29305 −9.457 −i0.17864 −7.7289 −7.69997 −0.09514 −0.0812 +i4.0993 0. For any specific theory.46175 0.5832 −9. / Nuclear Physics B 585 (2000) 293–352 305 Table 1 The parameters in the fit to FV and GV for the three mass cases FV Masses 105 a1 106 a2 105 a3 107 a4 105 a5 107 a6 104 a7 106 a8 106 a9 105 a10 a11 GV mK + .078793 −i0.

χ− ] + iαV Vµ [uν . fields. For the spin-1 mesons it is immediate to derive the Lagrangian. uν u } LA = − (5. which we do not consider here. The rest of the notation is as in Section 1. Specifically we use for the vector matrix Vµ 1 1 fV µν Vµν V µν + m2V Vµ V µ − √ Vµν f+ 4 2 2 2 igV µν − √ Vµν [uµ . However that will increases the number of unknown constants to determine without improvement for the estimates.306 G. uν ] + fχ Vµ [uµ .3) The last two terms are needed for the calculation of the vector and axial-vector two-point functions and not interesting for the present discussion.2) with Vµν = ∇µ Vν − ∇ν Vµ . / Nuclear Physics B 585 (2000) 293–352 chiral symmetry incorporating extra. The result after the integration of resonances is for the vector fields   ifχ gV gV αV  ∇λ ([uλ . µν f± = u(v µν − a µν )u† ± u† (v µν + a µν )u. The scalar mesons are considered through LS = 1 µ dm 2 ∇ S∇µ S − MS2 S 2 + cd Suµ uµ + cm hSχ+ i + S χ+ 2 2 µν µν + cγ Sf+µν f+ + cγ0 Sf−µν f− . f−νλ ∇ µ [uµ . The precise input we use for this is described in this section. There are two ways to calculate the contribution from these Lagrangians: using the equations of motion to integrate out directly the heavy fields. For the spin-1 mesons we use the realization where the vector contribution to the chiral Lagrangian starts at O(p6 ). In this case it seems faster to use the equations of motion. however they are related to the anomalous sector. considering all the diagrams with explicit resonance fields and expanding the result to the O(p6 ). Amorós et al. uν ])[uν . or diagrammatically. We can construct other terms similar to the terms in the O(p4 ) Lagrangian. (5. f− ] 2 2 LV = − (5. higher mass. Vµ and Aµ are three-by-three matrices describing the full vector and axial-vector nonets. masses and decay constants.1) and for the axial-vector Aµ 1 1 fA µν (1) Aµν Aµν + m2A Aµ Aµ − √ Aµν f− + γA Aµ uν uµ uν 4 2 2 2 (2) µ ν + γA Aµ {u . There are other terms with the same order in momenta as in the previous Lagrangian. uν ] LV = − √ 2 2 2 MV 2 MV . χ− ] + √ uλ . We will only discuss terms relevant for K`4 . for the scalars also shifts in their vacuum expectation values have to be considered. It is however one source of error to keep in mind. We remark that these are not all the allowed contributions from the scalar mesons.

1. αV = −0.38]. For the work here we take the next term in this expansion to estimate the unknown O(p6 ) constants. cm = 42 MeV. However we remark that the spin-1 Lagrangians are directly of this order after the integration. The resulting resonance contribution is displayed in Appendix A. The combination of terms we are interested in. Some of the terms of Eqs.6) are already directly listed in the full list of the O(p6 ) Lagrangian terms of [24].77 GeV. can be more easily calculated directly from Eqs. where the use of finite sum rules could provide the values for the combination of the unknown constants appearing in the considered process.2. fA and cγ are taken from experiment [8. fχ . Together with these we take the couplings fV = 0. χ− ] .4) for the axial-vector fields (1) (1) fA γ fA γ µν  µν  LA = − √ A 2 ∇µ f− uλ uν uλ − √ A 2 ∇µ f− uν .7) and the masses are the experimental ones [27]. 39]. For the remainder we use the ENJL model estimates of [40]. mA = ma1 = 1.1. mV = mρ = 0. cm and cd are from comparison with the resonance saturation at O(p4 ) [37.09. fχ = −0. cγ0 cd = 32 MeV. (5.025.8) The values for fV . cγ = 19 × 10 γA(1) = 0. −√ 2 MV2 307 − (5.23 GeV. gV = 0. mS = 0.38]. uν ] − [uν .16. We test the influence of the scalar fields by also performing the fit with vector meson contributions only as discussed below (fit 6).4)–(5. + MS4 (5. uλ uλ . gV . −3 GeV −1 .014. fA = 0. χ− ] uλ . f−νλ 2 2 2MV MV  fχ fV λµ ∇λ f+ [uµ . 2 MA 2MA (5. γA(2) = −0. In all cases the obtained value is in reasonable agreement with the resonance estimate. The others can be rewritten in terms of them but the resulting list is rather long.6) The three Lagrangians are O(p6 ). (5. ∼ cγ .5) and for the scalar mesons LS = cd2 2  ν    cm µ λ u u ∇ u ∇ u + ∇ν χ+ ∇ ν χ+ ν µ λ 4 4 2MS 2MS   c d cm ∇ν uµ uµ ∇ ν χ+ . Besides these ways to estimate the constants it is in some cases possible to obtain the O(p6 ) couplings from experiment. For the scalar case the expansion of the scalar propagator produces O(p4 ) as leading contribution but that part is already included via the values of the Lri constants. It was also done for the pion formfactors [18]. This has been done in previous calculations [15.01.1.G. Amorós et al.98 GeV.20.37. (5.6).4)–(5. (5. / Nuclear Physics B 585 (2000) 293–352  iαV fχ    igV fV ∇λ f+λν ∇ µ [uµ . .

(5. in such a way that the η and η0 fields are a combination of the singlet and octet fields when SU(3) symmetry is broken but SU(2) is still respected.38] 1 1 iγ F0 Lη0 = ∂µ P1 ∂ µ P1 − Mη20 P12 + √ P1 hχ− i. The behaviour of the resonances is different from that for the Goldstone bosons: no perturbative theory as for the pseudoscalar mesons exists.42]. that are supposed to be strongly correlated if they give the low-energy behaviour of QCD. used to calculate the decay constants and the masses. (5. Again. However the above procedure is arguably the correct picture in the large Nc limit. some correlations among the constants are obtained. except for η properties.9) At O(p4 ) the η0 contribution is entirely encoded in Lr7 and dominates its value. 2 2 2 6 (5. O(p6 ).1. Amorós et al. The first one is to obtain a prediction for the form-factors to be compared with experiment and to use . (5. An alternative equivalent way is to introduce it as a separate pseudoscalar singlet field P1 described by the Lagrangian [37. 6. / Nuclear Physics B 585 (2000) 293–352 Before continuing with the next section. gives Lη0 = − 2 d˜m ∂µ hχ− i∂ µ hχ− i 2Mη40 with d˜m = 20 MeV.9) that will contribute as well to O(p6 ). affecting mainly the value for Lr7 with a small modification for Lr8 . are obtained with the external currents from the octet. with the same symmetries as the pseudoscalar one. Phenomenological applications There are two points to address with the two-loop result for K`4 . A more complete list of terms including their integrating out shows only small modifications to the phenomenological implications [41. the pole in the propagator of course corresponds to the real η and the decay constant is the coupling of the real η to the octet current.308 G. In order to include explicitly the η0 degrees of freedom we can enlarge the pseudoscalar multiplet including the singlet state. The η–η0 mixing which affects considerably the η properties is known to be fully contained at O(p4 ) in the low-energy constant Lr7 . There are 90 + 4 O(p6 ) constants in the most general three-flavour chiral Lagrangian. The next order expansion in the P1 propagator. The η0 field Another possible contribution from heavier states is given through the η0 meson that at low-energies does not play as essential a rôle as for instance the ρ. We remark that although the two-point functions.10) Similar to the scalar case we can add terms to Eq.10) into the terms of [24] leads to a long expression so we do not present it. The contribution is of relevance to our fit only via the η mass and the prediction for the η decay constant. 5. it is worth it to summarize some ideas. Considering the Lagrangian for heavier states. rewriting Eq.

The fit is performed by minimizing the χ 2 function defined by X X  xi − xiinput 2 2 2 χi = .1. so a similar rough estimate is expected to work for the Cir . In particular. Unfortunately. one phase-shift combination can be obtained independently from the same decay data. Lr2 . defined below. The comparison with the scattering quantities obtained directly is a consistency check of CHPT. The values are updated fitting the results obtained for the masses and decay constants to two-loops [10] and the result of the present work. They are B0 m. simultaneously. F0 and Lr1 –Lr10 . These values are obtained through partially independent fits. FK and three inputs from K`4 . Values of the low-energy constants The previous values of Lri are obtained from one-loop or one-loop improved calculations [2. In practice we thus need to fit 11 free parameters. We test the dependence by removing all the resonance contributions except for the vector ones.6]. m2η . The three additional inputs are the ˆ Lr4 and Lr6 . that are estimated in the previous section. This is presented below in Section 6. We therefore use below the known two-flavour two-loop result for the scattering [7. Lr9 is discussed below. Lr3 are obtained from K`4 taking fixed the other Lri . whose couplings are determined experimentally elsewhere. Amorós et al.1. see also [19]. The values of Lri can be reasonably well predicted from resonance exchange. and the η0 . Below we use as inputs m2π . In the future we hope to be able to include additional quark-mass ratio ms /m. Section 5. This allows us to fit everything since in all higher orders we have replaced quark masses and F0 by the relevant physical quantities. Changing the values of the Cir by 50% changes the fitted values within the fit errors. Fπ . 6. In the present work we avoid to fix them and perform the first global fit of most of the O(p4 ) low-energy parameters ˆ B0 ms . (6.6 with as weight 1xinput the error in the input values xiinput . m2K .2. We remark that in this two-loop calculation the only way to obtain an imaginary part for K`4 is through the intermediate states of two pions. We also present a reasonable variation of the other inputs. it is not possible to find at present 11 fully independent inputs from direct experimental quantities. fs (0).6 i=1. experimental input to constrain these as well. The total number of parameters up to O(p4 ) is 13. gp (0) and λf . / Nuclear Physics B 585 (2000) 293–352 309 our prediction to check the validity of assumptions made in the data analysis. The other free quantities in the full result are the Cir . The second one is to extract a fit for the low-energy constants Lri which values will be used to predict other quantities of interest. Lr10 does not contribute to any of the quantities considered here and is thus fully independent.1) χ = 1xiinput i=1.8]. Then we can use the mass and decay constant .G. However the π –π amplitude appearing here is oneloop. in particular Lr1 . and see if the results change. It is then not possible to obtain indirectly a three-flavour two-loop result for the rescattering quantities. they are needed to obtain predictions for low-energy π –π scattering.

The standard values are ms /m ˆ = 24.85 −2. Notice that Lr4 .18 0.37 ± 0.310 G.71 0.52 0.2 0. All Lri (µ) values quoted have been brought to the scale µ = 0. Lr6 and Lr9 are input.293 GeV Fit 5 Fit 6 0.27 1.95 −2.46 −0.9 × 10−3 .336 GeV and s` = 0 Main fit 103 Lr1 103 Lr2 103 Lr3 103 Lr5 103 Lr7 103 Lr8 Changed Quantity Unit [3.00 0.49 1. the Kaon masses mK + = 493. mπ 0 = 0.1.9 ± 0.677 MeV.64 −0.08 O (p 4 ) Fit 2 0.65 1.80 −2. Lr6 : The first main feature that we use is the behaviour under the large Nc limit.3 O (p 4 ) 0.23 −2. Moreover Lr9 appears with the factor s` in K`4 .73 −2. Fit 5 in Table 2 assumes them to be zero at a scale of m2η instead. For other fits relaxing this assumption we refer to Section 6. If we assume that O(p ) and O(p6 ) do not appreciably modify this ratio we can use this as input or alternatively the values from QCD sum rules and/or the ˆ = 24 and we check that ms /m ˆ = 26 does not lead lattice. the pion decay constant Fπ = 92.67 −3. As standard input we use ms /m Table 2 Results for Lri (µ) for the various fits described in the main text.51 0. / Nuclear Physics B 585 (2000) 293–352 expressions up to O(p6 ) to determine B0 m ˆ and F0 from the physical input afterwards.30 MeV.44 0.5 GeV µ 1.93 . Lr6 = 0 and Lr9 = 6.0 GeV g(0) 4.12 1. Lr4 . This is the case for Lr4 and Lr6 that are both set zero for the main fit (at mρ scale).90 × 10−3 .60 −0.15 −0. Errors are fitting errors described in the text. η0 only Fit 7 Fit 8 Fit 9 0.4 ± 0.58 −2.5. at the end six LEC’s to fit. The consequences for the Lri are discussed in [2] and for the Cir in [24]. mK + = 497.16 0.04 0.73 0.4 MeV and the Kaon decay constant FK /Fπ = 1. Lr4 = 0.53 0.73 0.26 0.26 ± 0.63 0.1349764 MeV. ˆ To O(p2 ) it is possible to relate the ratio of the strange The quark-mass ratio ms /m: ˆ = (mu + md )/2.20 0. Lr6 −0.5 −0.75 0.23 −0. Other Lri : The remaining constants will be taken as free parameters.46 1.52 ± 0.67 −0.5 ± 0.47 √ s` 0.35 ms /m ˆ 26 Fit 3 Fit 4 0.51 −2. with a combination quark mass ms over the isospin doublet quark mass m 4 of the meson masses.01 [27].77 GeV.52 0.26 0.65 ± 0.65 0.49 −3.48 µ 0.6] 0. −0.58 0.30 0.3.72 ± 0. Lr9 : Another point is the value for Lr9 .23 0. For the other fits we can safely use the value Lr9 = 6.1 GeV p sπ0 0.4 ± 0.67 0. p sπ0 = 0. such that it is not necessary for the main fit where we use s` = 0.70 ± 0. The value is completely saturated by the ρ resonance.62 −0.47 ± 0. and for this reason the contribution from the logarithms is negligible.50 0.672 MeV and η mass mη = 547.85 0. Let us now discuss the various inputs and χi2 . Physical masses and decay constants: For these quantities we use the pion masses mπ + = 139.24 1.49 0.26 0.99 −3.26 −0.45 0. This limit tells us that some of the constants are expected to vanish at some unknown scale.47 0.18 1.46 Lr4 .82 −0.22 ± 0.35 ± 0.44 0.2 10−3 Vµ .23 0.25 −0. This constant is related to the charge radius of the pion.74 −2.56995 MeV. Amorós et al.

using or not the η mass . Amorós et al.G. We remark that there are two possible ways to calculate this ratio from the lowest-order meson masses. / Nuclear Physics B 585 (2000) 293–352 311 to significantly different results (fit 2 in Table 2).

.

3 m20η − m20π 2 m20K − m20π ms .

.

ms .

.

2) m ˆ . = . = . (6.

1 m ˆ .

(6. For the Kaon mass we subtract the electromagnetic contribution through the formula  1 2 m + + m2K 0 − 1.8 is a modification due to the corrections to Dashen’s theorem [43–46]. Fπ and the Lri .8 m2π + − m2π 0 = (494. The expression of the bare masses in terms of the physical masses to two-loop is taken from [10]. For the pion mass we use the neutral pion mass. Whenever we fit or calculate masses we use m2K av and m2π 0 . We included both relations in the fit. The contribution to χ 2 is .2 m20π 2 m20π where m0 stands for the lowest-order mass that we calculate in terms of the physical masses.53 MeV)2 .3) 2 K where the factor 1.

.

!2 X mmˆs .

i − mmˆs .

input 2 2 .

. (6.4) χ 1 + χ2 = 0.1 ms .

contrary to what is stated in the text there. where fs (0) and gp (0) are the form-factors at threshold of the dipion system. cos θπ = 0). In [25] the s-wave fs and p-wave gp are fitted with the expressions  fs (q 2 ) = fs (0) 1 + λf q 2 .6) gp (q 2 ) = gp (0) 1 + λg q 2 .2 m ˆ input FK /Fπ : This quantity is calculated using the formulas of [10]. we use the following approximation fs (0) = F (sπ . (6.  (6.01 K`4 decay: The expression for the form-factors F and G obtained in the theoretical part is compared with the experimental analysis from [25]. In order to keep the computer time needed in reasonable limits. This is also confirmed in the expressions as discussed below. (6. 2 and enters in the fit as   (FK /Fπ ) − 1. The others contributing waves are not appreciable [25].2 the formulas in terms of lowest-order masses and decay constant.7) We use sπ = (2mπ + 1 MeV)2 to avoid numerical problems at threshold. s` = 0. m2Kav = i=1.5) χ3 = 0. . The s` √ dependence is very mild. Here we use the formulas in terms of physical masses. The results we are interested in are the extraction of the s-wave contribution from the F form-factor and the p-wave from the G form-factor.22 2 2 . we checked this by performing an alternative fit with s` = 2 The published version has in Appendix A. The slopes are taken to be the same λf = λg .

226 which showed no significant deviations from our main fit. χ52 = gp (0) − 4.10) is modified with   gp (0) − 4. This has both numerical problems and does not correspond to the experimental situation. cos θπ = 0)| −1 0 π (6.23. where we include additionally the gp (0) measurement in KL0 → π 0 π − e+ ν of Ref.312 G.59 0.12.47 ± 0.02 2 .77 0. Similarly gp (0) = G(sπ .11) χ5 = 0.15.27  2 . 103 Lr2 = 0. 103 Lr7 = −0. (6.5 and µ = 1 GeV.93 2 2 . We also included two fits (fit 7 and 8) with a different choice for the scale.52 ± 0. The rescaling of the Lri is considered such that in the table we quote the shifted value at µ = 0. but almost covered by the fitting errors.8) We do not make use of the derivative at threshold as the slope. The slope is then given by   4m2 |F (sπ0 . (6.18.72 ± 0.47]. These choices imply different scales for the resonance saturation on the Cir . The errors are those quoted by MINUIT and are those that give 1χ 2 = 1 assuming the quadratic approximation near the minimum. / Nuclear Physics B 585 (2000) 293–352 100 MeV (fit 3 in Table 2).12) In Table 2 we present the alternative fits discussed above. Instead we shift a bit from the threshold. Only using cos θπ = 0 corresponds to assume d and higher waves to be negligible. As can be noticed the changes due to the scale are quite considerable.08 0. Eq.26 ± 0. Amorós et al. s` = 0. This value is from using unitarity in the CKM matrix and Vud = 0. .65 ± 0. (6. χ62 = λf − 0. cos θπ = 0)| sπ − sπ with the value sπ = (2mπ + 1 MeV)2 and sπ0 = (336 MeV)2 . 103 Lr5 = 0. s` . These quantities enter in the fit via  χ42 = fs (0) − 5. In addition we have performed a fit where Vus was changed to 0. s` .24. this mainly affects the values of Lr7 .9740 [27. Combining errors using the PDG procedure [27].77 GeV.31 With the previous fit procedure we find for the values of the low-energy constants 103 Lr1 = 0. The difference with the results of [19] is due to the η0 inclusion everywhere. (6. fit 9. 103Lr8 = 0. As an alternative we also present a fit with sπ0 = 293 MeV (fit 4 in Table 2).10) We also present an alternative fit. This suggest that the errors quoted in the main fit are unlikely to be reduced if there is more control over the O(p6 ) constants. [26]. µ = 0.14  2 . cos θπ = 0). (6.99.9) λf = |F (sπ .70 ± 0. 103Lr3 = −2.

2. Similar plots can be made for all other combinations.6]. is negligibly small. All of them. Therefore we fix Lr5 . Not included are the errors due to the restriction to O(p6 ) and other theory uncertainties. Comparison with previous results and models In this section we want to give some insight about the sizeable corrections of the LEC. Lr7 and Lr8 to their O(p4 ) value. except the last one. masses and WFR is introduced. with respect to the unitary calculation. log × log and the resonance saturation piece. As examples of the correlations we show the plot of Lr1 . Li × log. All in all this amounts to determine Lr1 . Furthermore.12). 4 are kept. Notice that this only shows the experimental type of error. Eq. the dispersive approach misses the full two-loop renormalization. In Fig. Lr2 and Lr3 together with their projections on the two-variable planes and the plot of Lr7 and Lr8 in Fig. Lr2 and Lr3 from the slope (λ). Those are the errors quoted in the remainder since they are more relevant than those calculated from propagating the errors of Table 2. To make the analysis more transparent we remove first of all the constraints on the masses and decay constants. We then keep the sets that have a χ 2 less than the value corresponding to 68% confidence level.49]. shows that the results are quite similar to those quoted in Eq. only the one-loop renormalization of Fπ .1. comparing O(p4 ).. thus the constraints imposed by the masses and decay constants do not drive at all the differences. 7 we show this fact. We have thus generated a distribution of sets of Lri that have a correlated Gaussian distribution according to the χ 2 function defined above.12). we deal almost with the same assumptions as in previous calculations [2. the Omnès improvement and the O(p6 ) for the F formfactor using the results of [6] as input as well as Fπ = 93.G. Amorós et al. This has some influence on G. (6. (6. To disentangle the pieces from the full calculation and compare with [6] is extremely delicate. using the dispersive method only some pieces of diagrams (h). Besides these obvious remarks a less straightforward point needs clarification: in [6] the full relativistic resonance propagator is kept. The main point to stress is that already at threshold there is a significant shift of the form-factors with respect to the dispersive calculation. 6.2 MeV. Each point correspond to one set of Lri of the distribution described above. Pieces that can account for the difference at threshold are the Li × Lj . 7 we have compared the Omnès improved calculation with the full O(p6 ). 3 and of those depicted in Fig.1. the s-wave (fs ) and the p-wave (gp ) — see Eq. Errors and correlations The problem in determining errors on derived quantities is that the Lri values given above are very strongly correlated and that with 6 free parameters the chance that all of them are within their one sigma error. 6.6). To take into account . (6. Notice the considerable shift already at threshold. second column in Table 2. A close look. / Nuclear Physics B 585 (2000) 293–352 313 6. Projecting this distribution on the relevant variable leads to larger ranges than the errors quoted above. In Fig. (j) in Fig.1. For instance. thus resumming all chiral orders. are O(p6 ) contributions not considered in the dispersive approach [6]. The Omnès improved calculation for F had the right sign but the wrong magnitude.e. The rest has no new imaginary part and is therefore modeled by a pure multiplicative constant only. The results obtained after that are displayed in fit A of Table 3. i. and to show the differences with some existing models based on large Nc [48. Due to the high correlation between the various variables this is not a trivial affair.

6. The second plot shows the Lr7 –Lr8 correlations. Lr2 and Lr3 . Amorós et al.314 G. . shadow area. The sets of Lri within a 68% confidence level range of χ 2 . and as a consequence the Omnès calculation matches the experimental result. In the figure we have used the values of the low-energy constants obtained in [6]. (6. Lr1 . To some extent this assumption was successful at O(p4 ) [37. With those parameters the O(p6 ) result is outside the experimental errors. / Nuclear Physics B 585 (2000) 293–352 Fig. this difference the values of the LEC.12) are in good agreement with the resonance saturation hypothesis. In the second column of Table 3 we quote the values for this model-approach to some of the LECs. have to change substantially. In the first plot we show + the values of the first three Lri and × the projections on the coordinate planes.38]. In our calculation we have assumed a resonance saturation for the O(p6 ) coefficients. One can see that Lr1 and Lr3 in Eq.

52] 0. Second is the result from the Resonance Saturation model.58 −3. It is clear that our result is within the range as exemplified by the two examples discussed below.74 ± 1.17 0.38] 0.G. We also compare in Table 3 with two model-predictions of CHPT parameters.21 0. [48] the large Nc and chiral limits lead including the .24 0.24 1. one should compare between different models.6 1.25 −2.46 Ref.60 a Input.9a Ref. and violates the relation 2Lr1 = Lr2 by 20%. Table 3 Values of Lri for comparison with Eq.2 −3. Lr2 is off from this hypothesis. Third and fourth columns are values from other large Nc based models as described in the text Fit A 103 Lr1 103 Lr2 103 Lr3 103 Lr5 103 Lr8 0.53 ± 0.79 1.67 ± 0. In Ref. [49. The shaded band is the experimental result.2 MeV. [37. which is of the expected size of deviations from the large Nc limit.0 1. (6. 7. As is discussed in [50] there is some indication that the resonance saturation assumption fails in the scalar sector. Amorós et al. including also the values of Lr5 and Lr8 . / Nuclear Physics B 585 (2000) 293–352 315 Fig.43 0. Comparison of the Omnès improved estimate with the full O(p6 ) calculation using the same old set of values of Lri as input together with Fπ = 93.81 1.4a 0.12). [48] 0.62 −4. For this matter.09 fixed fixed Ref. The first column is the fit without the constraints from the masses and decay constants.

Lr3 and Lr8 . Amorós et al. 9 we show the absolute value of F √ and G for s` = 0. 4 5Nc MQ 96π 2 4L1 = 2L2 = (6. The resulting relations led to the proposal of a new model.12). . together with the experimental error band. 6. 8a and for G in Fig.23] F (sπ . s` ). uπ . [48] are sensitive to the precise regularization scheme chosen in the 5 8 chiral quark model. sl ) − σπ P L cos θπ G(sπ . and the various separate contributions. 6L1 = 3L2 = − L3 = 4L5 = 8L8 = 7 8 m2ρ 8 6 16π 2 (6.22. Let us now look at the dependence on s` . uπ .14) where the last equality is obtained using higher orders in QCD sum rules [52]. K`4 form-factors and partial wave expansions We now plot the form-factors F and G at cos θπ = 0 for the main fit. Comparing the third column in Table 3 with Eq. We show the full result for F in Fig.13) where MQ can be interpreted as a constituent quark mass. In Fig. 2 96π   α  Nc π 2 h πS GGi 6 + O 1/M 1 + L3 = − Q . LMD [49]. Lr2 .23].22. The lines for the higher s` -values stop at the kinematical limits. 8b. 8a and 8b. For G this is dominated by vector exchange with experimentally determined parameters so the uncertainty from this source is quite small. / Nuclear Physics B 585 (2000) 293–352 leading gluonic contribution to 3  Nc  6 1 + O 1/MQ . 4 The partial wave expansion is not simple for F and G since the components with a well defined L = 0.316 G. Eq. tπ . tπ .12) one can see an improvement in the central value of Lr1 . The model of [48] was extended to the full ENJL model with gluonic corrections in [51]. 100 MeV and 150 MeV. Its values are given in the fourth column of Table 3 and the same comments as in the previous model apply for Lr1 . s` ) = ∞ X l=0 Pl (cos θπ )fl (sπ . It mimics QCD by taking into account the first resonance plus the continuum obtaining the relations 8 3 Fπ2 15 1 = √ . Notice the change in scale with respect to Figs. went off suggesting the relevance of higher gluonic or large Nc corrections. X 3 The predictions for Lr and Lr of Ref. while the central value of Lr5 turns out to be worse. which essentially agrees with the O(p4 ) LEC’s. The form-factors F and G can be decomposed in partial waves 4 as [20.2. Given the accuracy of the present data we can safely neglect the s` dependence completely. Lr5 and Lr8 with respect to the one-loop result. (6. (6. Notice that the convergence of the series is reasonable for both form-factors and that for F the contribution from the O(p6 ) Lagrangian is fairly small. ±1 z need to be expanded [20. while the values for Lr2 and Lr3 .

The absolute value of the F and G form-factor at with cos θπ = 0. The shaded band is the result of [25]. gl (sπ . tπ . 100 MeV and 150 MeV. 9. / Nuclear Physics B 585 (2000) 293–352 317 Fig.15) l=1 where Pl0 (z) = d Pl (z). s` ) = ∞ X Pl0 (cos θπ )gl (sπ . The curves for Re G and |G| are very close since Im G is very small. G(sπ . dz (6. s` ). and the real part to lowest-order. this is dominated by the vector contribution.G. Shown are the full result for the absolute value. 8. s` ) have as . The F (a) and G (b) form-factor at s` = 0 and cos θπ = 0. uπ . s` ). All cases (6.16) Pl (x) are the Legendre polynomials and the functions fl (sπ . Fig. The shaded band is the result of [25]. Amorós et al. We also show the contribution from the O(p6 ) Lagrangian. O(p4 ) and O(p6 ). √ s` = 0.

 1 0 G = g˜p + g˜ p0 sπ + g˜1 s` eiδ1 (sπ ) + g˜d σπ X cos θπ eiδ2 (sπ ) . Amorós et al.318 G. 10. (a) The partial wave expansion of the combination F¯ = F + σπ P L cos θπ G/X. Similarly we see that for G the d-waves are very small. / Nuclear Physics B 585 (2000) 293–352 Fig. The imaginary part for the other waves is negligible. (b) The partial wave expansion of G. fs has an appreciable imaginary part.1. We can discuss one possible isospin breaking effect. (6. phase the phase-shifts δlI (sπ ) from π –π scattering with isospin I = 0 for even angularmomentum l. Plotted are the real part of gp and gd and the imaginary part of gp . We also show the fp contribution from the second term in F¯ . The d-waves are very small and the f -waves are not visible on this scale. The proposed parametrization was  0 1 F = f˜s + f˜s0 sπ + f˜s00 + f˜1 s` eiδ0 (sπ ) + f˜p σπ X cos θπ eiδ1 (sπ ) . The difference between fp and fp (G) is negligible with the present data as was also found by [25]. The effect this has on F and G is plotted in Fig. The gf and higher waves are not visible on this scale and only gp has a visible imaginary part. fp (G) is dominant. fp and fd . 10a and 10b. 6. and I = 1 for odd l. The contribution to fp coming from the term with G.2. As described for the parametrization in Section 4. The results are plotted in Figs. Here we check how well the approximations inherent in the parametrization are satisfied by our full two-loop calculation. We perform this expansion numerically for the full F and G expressions. A parametrization for K + → π + π − e+ ν In [53] two of us proposed a parametrization to be used to extract form-factors and phase-shifts from K`4 data. 11.2 we use slightly different masses for the three Ke4 decays. Plotted are the absolute value of fs and the real part of fs . The imaginary part of fp is also dominated by the part from G. It is obvious that with the present experimental accuracy this source of isospin breaking is fully negligible.17) .

(6. The absolute value form-factors calculated with the different masses used for the three possible Ke4 decays. s. u. t) = A(t. t). I =0 (6. u) + A(t.19) . T 2 (s. Amorós et al.3. and the absolute value of the p. We can now fit these parametrizations to the partial waves obtained in the previous section. 6. The fits are performed for sπ < 0. s. T 1 (s. t) = A(t. s) + A(u. s) + A(u. (b) |G|. / Nuclear Physics B 585 (2000) 293–352 319 Fig. t).G. For this purpose one expands the π –π amplitude as combinations of definite isospin amplitudes in the s-channel T 0 (s. t) = 3A(s. s. we can discuss how the threshold parameters and phase-shifts in π –π scattering are affected with respect to the analysis performed in [8] and [31].17). Partial wave expansion and threshold parameters for π –π scattering amplitudes With the main fit result for the O(p4 ) low-energy constants. t). s) − A(u. t) = 32π ∞ X (2l + 1)Pl (cos θ )tlI (s). u. As foreseen in [53] the parametrization works extremely well. We have plotted in Fig. since that is the region where most events of the data will be. The curves shown are with cos θπ = s` = 0.and p-waves of F . The tilde notation is because these quantities are related to the partial-wave expansion but not identical. t. u. 11.18) These isospin amplitudes are expanded into partial waves T I (s. f¯s and f¯p . The labels indicate the charges of the pion.and d-waves of G are parametrized by Eq. (6. (a) |F |.16 GeV2 only. 12 how well the absolute value of the s. The higher partial waves are completely negligible within the experimental precision expected in the near future. The experimental results are indicated with the grey bands.

0.12). (6. Notice that for comparison purposes. We have shown s` = 0.16 GeV2 only. The agreement of the parametrization with our full two-loop result. 24 1 νK + O(p6 ). the scattering lengths alI and the slopes blI . 0. / Nuclear Physics B 585 (2000) 293–352 Fig. However the two-loop relations between two. 12 .20) tlI (s) = 2 s − 4mπ 2i holds in the elastic region 4m2π 6 s 6 16m2π with real phase-shifts δlI (s).and three-flavour constants is unknown.01.320 G. This last expansion defines the threshold parameters. An expansion near threshold in Eq. and θ is the scattering angle in the center of mass frame. A full matching would need the knowledge of processes involving the relevant constants in both CHPT versions and to the same order. at present this is not the case. s = 4(m2π + q 2 ) and q is the center-of-mass three-momentum of the pions. (6.0225 GeV2 .20) gives    (6. the main parts have been strongly shifted downwards. 12. The fits are performed for sπ < 0. The π –π scattering amplitude is known up to O(p6 ) [8] — see also [9] for a dispersive calculation — in two-flavour CHPT. We can study the variations of these parameters with the Lri of this work. allows to obtain the threshold parameters in terms of the low-energy two-flavour O(p4 ) constants. Expanding this amplitude into partial waves and in q 2 . Unitarity implies that  1/2 s 1  2iδ I e l −1 (6. We use as a first estimate the relations obtained from an O(p4 ) matching [2] l1r = 4Lr1 + 2Lr3 − l2r = 4Lr2 − 1 νK + O(p6 ). The lines are the parametrization.21) Re tlI (s) = q 2l alI + q 2 blI + O(q 4 ) . with s. Eq. t and u the usual Mandelstam kinematical variables. The points are the full result. Amorós et al. lir .

22 – 0. that is due to the fact that the contribution from the double logarithms almost saturates the result to O(p6 ).5.24 0.48 – 0.4 .005 0.61+1.1 while set II is l 1 = −1.40 0.81 0. Bearing this last issue in mind.82 ± 0.12 0 0. The errors are given by projections on the relevant variable of the 68% confidence level domain.33 0.12) barely shift the results of set I [8] — with exception of b11 . The interesting result that comes out from Table 4 is that with the large difference in the input parameters. (6.28 input 0.22)  2     2  mP γi 1 ¯i + ln mπ = l and ν ln + 1 .159 0.23 0.32 0. Experimental values are taken from [54].423 0.378 ± 0.420 ± 0.03 0.90 ± 1.450 0.219 ± 0.36 ± 0.420 0.254 0.38 ± 0.62 ± 0.23) We want to emphasize that the relations in Eq.59 ± 0.10 0 0. P 32π 2 µ2 32π 2 µ2 (6. l¯i .03 0.83 0.56 – 0. However . l¯4 = 4.40.404 0. 2 with lir = (6.24) of Eq.203 0. As was shown in [8].303 0.12) O(p2 ) O(p4 ) O(p6 ) set I O(p6 ) set II Experiment a00 0.0.02 0 0.9 −2.17 ± 0.021 0.021 0.22) are affected by higher — O(p6 ) — chiral corrections which might be quite sizeable.5 and l 2 = 4.08 0.02 0.05 b00 a02 b02 a11 b11 a20 b20 a22 b22 a31 b31 0.65 0. Both sets have l 3 = 2.32 ± 0.09 0 0.011 0.256 0.16 0.28 ± 0. a20 and b20 . (6.182 0.729 0.3.755 0.11 0 0.20 0.356 0. Errors are from the 68% confidence level ranges only −10 −10 10 102 102 −103 103 −103 104 −104 Eq.38 0.22). / Nuclear Physics B 585 (2000) 293–352 321 Table 4 Threshold parameters in units of mπ + .29 ± 0.010 0.756 ± 0.12) and Eq.22 ± 0. Set I is l 1 = −1. 16 1 l4r = 8Lr4 + 4Lr5 − νK + O(p6 ). l¯3 = 2. (6. Amorós et al. the changes obtained with Eq.9 and l 4 = 4.25 ± 0.13 ± 0. The projection of the filled ellipse on the axis produces the errors in Table 4. using the Lri l¯2 = 4.04 0 0.10 0 0.G. in Table 4 we compare previous results and the experimental data [54] with the scattering lengths and slopes obtained with the values l¯1 = 0.24 input 0.05 ± 0.282 0.38 0.31 0.26 ± 0.279 ± 0.18 (6.454 0. (6. In Fig. 13 we plot the predictions of a00 and a02 for the the set of Lri distributed according to the χ 2 function as discussed earlier.908 0. (6.212 0.12 0.27 – 0.30 0.40 ± 2.222 0.7 and l 2 = 6.15 0.36 ± 0.37 – l3r = −8Lr4 − 4Lr5 + 16Lr6 + 8Lr8 − 1 νη + O(p6 ).

For completeness we also display in Table 5 the value of the variables bi defined in [7] as a finite combination of the two-flavour low-energy constants and which determines the polynomial piece in the π –π scattering at O(p6 ). Eq. Errors are the projections of the 68% confidence level domain only.65 8.33 7.25 2. Similar remarks apply to Figs. .03 −2.27 ± 3. 14 and 15.322 G.49 O(p6 ) set II −8.14 8.08 2. 13.83 −0. Only points within the 68% confidence level domain are plotted. The threshold parameters which were not listed in [8] were calculated by doing the relevant expansions in the formulas there to higher order. Amorós et al.23 O(p6 ) set I −9.16 ± 0. (6. Table 5 The bi quantities as defined in [7].20 1. The prediction for the a00 and a02 scattering lengths for the set of Lri distributed according to the correlations of our main fit.10 6. The results of set I and II differ from [8] because of the different value of Fπ and we have kept some O(p8 ) parts which were dropped in the numerics there. / Nuclear Physics B 585 (2000) 293–352 Fig.32 ± 1.32 1.12) 102 b1 102 b2 103 b3 103 b4 104 b5 104 b6 −6. see Appendix B.83 the new experiments are expected to have enough accuracy to distinguish small changes in the scattering lengths and slopes.20 5.88 −4.56 ± 2.50 ± 2.37 0.00 −1. Set I and set II are defined in the caption of Table 4.44 ± 1.64 4.

. Amorós et al. 15. Data points are taken from [25]. The shaded band is the 68% confidence level range. / Nuclear Physics B 585 (2000) 293–352 323 Fig. Difference of phase-shifts δ00 − δ11 plotted as a function of the center of mass energy Eππ .G. The shaded band indicates the 68% confidence level range. Data points are taken from [55]. Phase-shift δ02 as function of the center of mass energy Eππ . 14. Fig.

24) and set I is excellent and also with the available data. B0 . We write the full O(p6 ) result as  (4) (6) + h0|qq|0i ¯ + ··· . The procedure is the same as described in Section 4. the contributing diagrams form only a small subset of those entering in the decay constant evaluation [10]. however.324 G. In fact. I and II. It is. Amorós et al.. but involves a second low-energy constant. at the first order in perturbation theory Z . introduces an additional dependence on the high-energy constant H2r . In Fig. which is intimately related with the vacuum expectation values of the scalar densities in the chiral limit h0|q¯ i q j |0i = −F02 B0 δ ij .1.26) ¯ h0|qq|0i ¯ = −F02 B0 1 + h0|qq|0i (4) and h0|qq|0i (6) are scale independent. due ¯ Note that h0|qq|0i ¯ to the quark masses. The value of B0 depends on the way the scalar quark bilinear is defined in QCD. In particular it depends on the QCD renormalization scale µQCD . (6.24) just interpolates between both sets. 15 we plot the I = 2 s-wave phase-shift δ02 as function of the center of mass energy together with available experimental data. This is the combination measured in K`4 decays. (6. 14 we show the difference of phase-shifts δ00 − δ11 as a function of the center of mass energy.e.25) In the absence of external scalar and pseudoscalar fields it is only observable in combination with the quark masses. The phase-shift corresponding to Eq. The calculation of the one-point scalar Green function without any external pseudoscalar field is rather simple as it only involves one-loop×one-loop diagrams at O(p6 ). I. This ambiguity is forced by the need of a subtraction in the scalar two-point function at zero separation.4. 6. (6. (6. masses and decay constants. the conclusions that can be drawn from the π –π scattering analysis is that good agreement is obtained for all the threshold parameters and with δ00 − δ11 and a substantial improvement with the data in the δ02 phase-shift case. a quantity of theoretical interest and thus deserves study at the two-loop level as well. The soft SU(3) breaking.3. As can be appreciated the agreement between Eq. This amounts to consider diagrams with one vertex involving the external field s part of the matrix χ defined in Section 2. whose value also depends on the precise definition of the quark bilinear in QCD. All in all. / Nuclear Physics B 585 (2000) 293–352 In Fig. The results are displayed in Appendix C in terms of the renormalized quantities. Vacuum expectation values The lowest-order effective Lagrangian of QCD is not determined by the pion decay constant alone.

.

In contrast [56] the non-analytic contribution is unambiguous. − i d 4 xh0| :(qMq)(x)( ¯ qq)(0): ¯ |0i + O(M2 ). h0|qq|0i ¯ = h0|qq|0i ¯ chiral limit (6. As a last point we mention that as QCD predicts [57] the vacuum condensate increases once the quark masses are switched on.27) which blows up as x → 0. .

m2K /(m2K )phys = 0.12).247] . with the value of (6. The numerical results are h0|uu|0i ¯ = −B0 F02 [1 + 0. (6.1) with the higher orders expressed in terms of Fπ and the physical masses. O(p4 ) and O(p6 ) contribution. The difference now vary them at a fixed ratio ms /m with the results in Eq. We use the scalar dominance assumption that leads to H2r = 2Lr8 . (6. / Nuclear Physics B 585 (2000) 293–352 325 In order to quote results we need to make certain assumptions on the value of H2r .742 − 0. Eq.19]. Fπ = F0 1 + F + F (6.277 + 0. We believe that the results using the physical decay constant and the physical masses in the higher orders are numerically more reliable since they incorporate part of the higher order corrections that we know will appear.28) Lr8 that follows from Eq. The separate contributions for them are m2π /(m2π )phys = 0. All masses are written as in Eq.30) where the numbers are the lowest-order. m2η /(m2η )phys = 0. respectively. (6.31) (6.29) comes from higher orders.007 + 0. We can now also check how the pseudoscalar masses vary with the quark masses. 16 we the unrenormalized masses and decay constants and vary B0 m plotted the vacuum expectation values as a function of ms /(ms )phys . The result is shown in Fig.30) shows that the light meson masses at O(p6 ) have a large contribution which seems in conflict with the convergence of the chiral series. (6.286. and the higher order formulas expressed in terms of real pseudoscalar masses and Fπ . The O(p4 ) contribution is much smaller than the O(p6 ) one.12). 17a. O(p4 ) and O(p6 ).746 + 0. Eq.32) Plugging in the values following from the lowest-order contribution in Eq.247. (6. [10]. (4. Amorós et al. For this we can use for B0 m. ˆ B0 ms and F0 is defined through where (m2 )(i) is a function of B0 m.695 + 0.29) where the numbers correspond to lowest-order.G.832 + 0.298 (6.040 + 0.  (4) (6) .103] . h0|¯s s|0i = −B0 F02 [1 + 0.30) is due to higher orders.30) we can ˆ = 24. In Fig. Masses and decay constants In this section the pseudoscalar masses are studied using the new values for the LEC. We can also see how the vacuum expectation value changes as a function of the quark ˆ B0 ms and F0 the values derived using the main fit masses. respectively. As has been previously noticed [10. For this purpose we rewrite the pseudoscalar masses as m2phys = m20 + (m2 )(4) + (m2 )(6). (6.5. We then insert these values into the expressions for the vacuum expectation values in terms of ˆ and B0 ms . The difference with the values in Eq.019 + 0. 6. This is due to two . (6.

(b) The pseudoscalar decay constants as a function of the quark mass at fixed ratio m/m ˆ s . The curves are normalized to the chiral limit. The results up to O(p2 ). . Amorós et al. 17.326 G. 16. / Nuclear Physics B 585 (2000) 293–352 Fig. The vacuum expectation values (VEV) as a function of ms /(ms )phys . The results up to O(p4 ) and 6 O(p ) are shown for the three pseudoscalar decay constants. (a) The pseudoscalar masses as a function of the quark mass at fixed ratio m/m ˆ s . Fig. Plotted are the strange and light quark VEV for the ratio ms /m ˆ = 24 and the light quark VEV for m ˆ = 0. O(p4 ) and O(p6 ) are shown for the three pseudoscalar masses. For the η it is the coupling to the octet axial current.

33) which shows that the corrections to the ratios are much smaller. ˆ = 20 and ms /m ˆ = 30 but otherwise the same input We have performed fits with ms /m as the main fit. / Nuclear Physics B 585 (2000) 293–352 327 adding effects.5(p2). 13. Without new input we feel we cannot discuss further this question.3(p4 ). In particular we will focus on Lr4 and Lr6 . (6. (6. the partial contributions at O(p6 ). .g. And second and mainly. (6.3(p2).134 + 0.136 − 0. Zweig rule In this subsection we relax some of our assumptions to see whether the unusual relative size of the O(p6 ) versus the O(p4 ) contribution in Eq.30) can be expressed as m2K = 12. 12. Another possibility is to relax the accepted common wisdom about the large Nc [58– 60] behaviour of the low-energy constants. 15. The corrections to physically measurable quantities are in fact not so large.1. For the pseudoscalar decay constants we use the formulas of [10] and obtain using our main fit Fπ /F0 = 1 + 0. We therefore rewrite the full formulas in terms of F0 and the quark masses. one of these possibilities is discussed in Section 6. 6.35) Using this value of F0 and the lowest-order masses obtained above we can now plot the dependence of the decay constants on the quark masses.7(p4 ). Eq. for the Kaon mass the contribution is given by 0. The results is shown in Fig. Other options are ad hoc choosing the values of some of the Cir to diminish the large O(p6 ) corrections. there are cancellations that happen inside the O(p4 ) contribution. (6.086. the difference with Eq. (6. In other observables we typically had cancellations between these various contributions. Again.019(total p4 ). 16. The numerical result allows to obtain F0 = 87 MeV. 17b. Amorós et al.34) The η-decay constant here is the coupling of the physical η to the octet axial current.4(p6 ). Large Nc . Some arguments for this can be found in [50] where it is argued that resonance saturation seems to fail in the 0++ sector.5.066(Lri ) − 0. Li × Lj and pure two-loop. log × log.202 + 0.076. m2π m2η m2π = 16. Fη /Fπ = 1 + 0. (6. E.G.1. recapitulating some of the reasoning of [2]. all go in the same direction. This does not qualitatively change any of the predictions for decay constants and K`4 formfactors made here. We have checked that relaxing some of the constraints used in the fits can improve the above mentioned behaviour.. FK /Fπ = 1 + 0.34) is higher order. Li × log.30) can be avoided. First. Neither of these changes the relative size of the contributions to the masses qualitatively.047(loops) = 0.069.4(p6 ).5.

2L1 − L2 . but (ii) Lr4 and Lr6 are scale dependent. H1 and H2 . ¯ Fπ and m2π are small if both Lr4 and Lr6 ∼ 0. main question is now if higher chiral orders in those quantities are still protected against a strong ms dependence. after using some trace identities [2] O(Nc ) O(1) L1 . clearly one can not trust large Nc arguments in CHPT. (2. If the answer is no. L9 . Setting U = 1 in Eq. The behaviour of the constant L7 depends on the treatment of the singlet η1 [2. Large corrections would also cast doubt on CHPT beyond one-loop calculations in the three-flavour sector. L10 . Thus there is no reason to argue for large deviations from the Zweig rule. In the constants of O(1) there are two different behaviours: (i) the combination 2Lr1 − Lr2 is scale independent. L4 . L3 . As it is shown. / Nuclear Physics B 585 (2000) 293–352 In principle the main question is whether the corrections of the relative order ms and m2s to the ground state are small in comparison with the scale of the theory. This last dependence should cancel via the meson loops that explicitly violate large Nc .14) the large Nc counting comes automatically. L5 .61].328 G. and L6 . Amorós et al. In order to estimate possible sizes we can just vary the scale inside a reasonable range. The At O(p4 ) the effects of ms on h0|uu|0i. there is a range in Lr4 and Lr6 values that allows large Nc and CHPT at O(p6 ) to coexist without conflict. Using . L8 . L2 .

 .

.

.

.

Γi µ1 .

.

36) ln |1Lri | = . (6.

Lri (µ2 ) − Lri (µ1 ).

= .

.

2 µ2 .

30). leads to Lr4 = (0 ± 0. as shown in Eq.62] use O(p4 ) expressions so these ranges might well change when those calculations are performed to O(p6 ). gp (0). (6. nor fit 5 have the expected ordering of O(p4 ) and O(p6 ) contribution to the pseudoscalar masses.13 and on QCD sumrules [62] which allow a band 0. (6. i. Let us now check for which ranges of Lr4 and Lr6 we find results where all the quantities considered up to now have an expansion of the form X = X2 + X4 + X6 with |X2 | > |X4 | > |X6 |.38). considering that large Nc is reached at the ρ scale. Other constraints on Lr4 and Lr6 are from a recent analysis [50] based on the positivity of the fermionic measure giving a lower bound 103Lr6 (Mρ ) > 0.38) above as a convergent series.3) × 10−3 . The ratios of the pseudoscalar masses always have small corrections even if they do not always obey Eq. (6. This is for reasonable variations always the case for fs (0). (6. We will refer to Eq. λf .e. h0|¯s s|0i and Fη /Fπ . Neither Eq. 103 Lr6 (Mρ ) .37) The central value correspond to the main fit presented in Table 2. 0. (6.6.2 .. For h0|uu|0i ¯ H2r has only a small influence but it . 16π with the choice µ1 : µ2 = 2. Fit 5 corresponds to use large Nc at the η scale.4 with unknown errors.38) where the subscript refers to the chiral order. The values and bounds from [50.5) × 10−3 and Lr6 = (0 ± 0.12). The latter reference also quotes 103 Lr4 (Mρ ) ' 0. (6.

38) is satisfied for all quantities considered in this manuscript. For this quantity we thus allow |X2 | . ¯ In Fig. The problematic quantities are Fπ /F0 . negative value for both Lr4 and Lr6 is penalized for the mass and the vacuum expectation value. the decay constant has the broadest range (in both parameters) of convergence. 5 Quantities are only calculated on the points of the grid.38). / Nuclear Physics B 585 (2000) 293–352 329 Fig. . taking into account possible higher order corrections to the bounds. The mass behaviour is quite unexpected: large. is very large for h0|¯s s|0i. Notice that a large deviation from the Zweig rule is not needed thus the bounds.37). The dash-dotted line is the lower bound that follows from positivity constraints [50] and the thick dashed line the value of Lr4 of Ref. absolute values of Lr6 reorder the series making O(p4 ) > O(p6 ). while the mass seems to be the most restrictive. A low. 5 The allowed bottom right region extends further to the right but with (FK /Fπ )4 < (FK /Fπ )6 even though both are small. (6. Amorós et al. Eq. |X4 | since this can easily be changed by varying H2r . 18. The grey circles are the points where Eq.38) for all quantities. If we increase Lr4 fixing the value of Lr6 the vacuum expectation value recover once more the convergent pattern. For instance. the pseudoscalar masses separately and the vacuum expectation value h0|uu|0i. Let us mention some general features in the behaviour of these quantities as a function of Lr4 and Lr6 . (6. Fits were only performed for points on the grid. large Nc constraints and the other bounds are reasonably satisfied. (6. As can be seen there is a region where Eq. still hold. (6. [62]. but then O(p4 ) starts to compete with O(p2 ) contribution. 18 we show the allowed region in the plane Lr4 −Lr6 consistent with Eq.G. The Lr4 –Lr6 plane. FK /Fπ . The grey band shows the bounds on Lr6 from QCD sum-rules.

We mention briefly how to treat the errors and show one way of treating the correlations using a 68% confidence level distribution of the Lri . We have described the calculation to O(p6 ) in CHPT of K`4 decays and of the vacuum expectation values. no evidence of strong infrared singularities. To this end we made use of the existing K`4 data. as a well defined perturbative series. pointing out the source of the difference. The long expressions involving the vertexand sunset-type integrals we simply parametrized. 16 and 17 (at fixed ratio m ˆ : ms ).and three-flavour low-energy constants. involving the low-energy constants. pseudoscalar masses and decay constants. The rest of the expressions. . We varied the inputs over a fairly wide regime and discussed in detail the underlying assumptions and their consequences once some of them are relaxed. The results were then used to obtain predictions for the π –π threshold parameters and the π –π phase-shifts. testing the validity of the large Nc relation (Lr2 −2Lr1 )/Lr3 ∼ 0. Eq. 14 and 15. reflected in the unknown O(p6 ) corrections to the relation between the two. Summary Let us summarize our results.4).5. giving some general lines on the methods that we used and all the various checks on the final results. can exist without conflict with any of the known quantities up to O(p6 ). We also warned the reader about the uncertainty in the relation between the theoretical knowledge of K`4 decays and π –π scattering. where CHPT. Among the assumptions it is worth to mention the discussion on relaxing the large Nc behaviour of Lr4 and Lr6 . decay constants and masses in terms of the lowest order parameters were shown in Figs. and large Nc assumptions on Lr4 and Lr6 we updated most of the values of the O(p4 ) parameters. We also present the comparison with previous results and model estimates. The explicit formulas have been presented as much as possible. see Table 2. As it becomes immediately clear the main source of the theoretical uncertainties resides in the O(p6 ) constants. Section 6. With all this at hand. are fully listed in Appendix A. Lri . We review the experimental assumptions of [25] finding some of them only borderline compatible with the theoretical expression. up to O(p ). We do not expect large changes in the main results when this is done. Amorós et al.1.12). / Nuclear Physics B 585 (2000) 293–352 7. Eq. The variations of the vacuum expectation values. We also show the form-factors for different values of s` and present the partial-waves predicted from the full result. (6. This is the reason we did not include π –π scattering data in our fit as was done in [6]. Good agreement with the existing experimental results was obtained as is visible from Table 4 and Figs.330 G. They 6 show. This provides insight in the relevant assumptions that the new experimental analysis should make. We show the allowed region in the Lr4 –Lr6 plane. (4. We have assumed resonance saturation to estimate the new O(p6 ) parameters that appear. Once the low-energy constants are found we use them to give the phenomenological implications on the forthcoming K`4 experiments.

Gasser. was supported from the Swedish Natural Science Research Council (NFR). Knecht. Lri × Lrj and Lri × one-loop contributions The FLL part of F is   FLL = m4π (16π 2) 2/3Lr1 + 2/9Lr2 − 31/108Lr3  + m4π −256Lr1Lr5 + 64Lr2 Lr5 − 48Lr3 Lr5 + 256(Lr4 )2 + 416Lr4Lr5 − 512Lr4 Lr6 − 512Lr4Lr8 + 56(Lr5 )2 − 64Lr5 Lr6 − 64Lr5 Lr8 − 16Lr5 Lr9   + m2π m2K (16π 2 ) 8/9Lr2 + 8/27Lr3   + m2π m2K Lr5 256Lr1 + 64Lr3 − 64Lr4 − 32Lr5 − 128Lr6 + 8Lr9 + 512(Lr4 )2 − 1024Lr4Lr6   + m2π m2η (16π 2 ) −2Lr5 n + m2π Lr5 128sπ Lr1 + 32sπ Lr3 + 8sπ Lr9 − 32tπ Lr2 − 16tπ Lr3 + 8tπ Lr9  − 32uπ Lr2 + 8uπ Lr9  ¯ 2π ) −404/3Lr + 268/3Lr − 13/3Lr + 334/3Lr + 31/2Lr − 55/6Lr + A(m 1 2 3 4 5 9  ¯ 2K ) −48Lr1 + 124/3Lr2 − 53/3Lr3 + 32Lr4 + 25Lr5 − 11Lr9 + A(m o ¯ 2η ) 16Lr1 + 20/3Lr2 − 73/18Lr3 − 10Lr4 + 73/6Lr5 − 7/2Lr9 + A(m   + m4K (16π 2 ) −2/3Lr1 − 19/9Lr2 − 55/108Lr3  + m4K −64Lr2Lr5 − 16Lr3 Lr5 + 8(Lr5 )2 + 8Lr5 Lr9  + m2K m2η /(16π 2 ) 8Lr5 n + m2K Lr5 −128sπ Lr1 − 32sπ Lr3 − 8sπ Lr9 + 32tπ Lr2 + 16tπ Lr3 − 8tπ Lr9  + 32uπ Lr2 − 8uπ Lr9  ¯ 2π ) 54Lr + 27/2Lr + 4/3Lr − 14/3Lr − 55/12Lr + A(m 2 3 4 5 9  ¯ 2K ) −4/3Lr1 + 130/3Lr2 + 1/2Lr3 + 24Lr4 + 8Lr5 − 11/2Lr9 + A(m o ¯ 2η ) −32/3Lr + 10/3Lr − 49/18Lr + 4Lr − 8/3Lr − 7/4Lr + A(m 1 2 3 4 5 9 . Work supported in part by TMR. ERBFMRX– CT980169 (EURODAPHNE). Amorós et al. / Nuclear Physics B 585 (2000) 293–352 331 Acknowledgements We thank G. Moussallam and J.G. A. (4. Stern for discussions and comments. J. Colangelo. Lri .T.1. Appendix A. B.3). EC–Contract No. P. M. Some O(p6 ) contributions In this appendix we give the explicit formulas for the O(p6 ) part of the form-factors F and G of K`4 split in parts as in Eq.

m2K . m2K . sπ )m2π m2K −32Lr4 + 3Lr9 + B(m    ¯ 2π . tπ )m4K 8/3Lr4 + 4/3Lr5 − 40/3Lr6 − 20/3Lr8 − 1/2Lr9 + B(m  ¯ 2π . tπ )m2K 1/2sπ Lr9 + 4tπ Lr4 + 2tπ Lr5 + tπ Lr9 + 1/2uπ Lr9 + B(m  ¯ 2π . m2K . uπ )4/3m2π sπ Lr9 + tπ Lr9 − 10uπ Lr2 − 8uπ Lr4 − 4uπ Lr5 + 2uπ Lr9 + B(m  ¯ 2π . tπ )m2π m2K −8/3Lr2 − 16/3Lr3 + 8/3Lr4 + 5Lr5 + B(m − 20/3Lr6 − 3/2Lr9   ¯ 2π . tπ ) −1/2sπ tπ Lr9 − 1/2tπ2 Lr9 − 1/2tπ uπ Lr9 + B(m  ¯ 2π . m2K . uπ ) −2/3sπ uπ Lr9 − 2/3tπ uπ Lr9 − 2/3u2π Lr9 + B(m  ¯ 2K . tπ )m2π m2K 3 2/3Lr − 46/3Lr − 22/3Lr + 20Lr + 32Lr7 + B(m 3 4 5 6 r r + 16L8 − 2L9 . m2π . sπ )m2K sπ −80Lr1 + 16Lr2 − 20Lr3 + 28Lr4 − 2Lr9 + B(m  ¯ 2K . m2K . m2K . m2K . m2K . m2K . m2K . sπ )m2π m2K 96Lr1 + 24Lr3 − 96Lr4 − 24Lr5 + 96Lr6 + 48Lr8 + B(m  ¯ 2K . uπ )m4K −16/3Lr4 − 8/3Lr5 − 2/3Lr9 + B(m  ¯ 2π . sπ )m4π 80Lr1 + 32Lr3 − 80Lr4 − 46Lr5 + 48Lr6 + 24Lr8 + 6Lr9 + B(m  ¯ 2π . tπ )m2π sπ /2Lr + tπ 8/3Lr + 16/3Lr + 14/3Lr − 3Lr + 3/2Lr + B(m 9 2 3 4 5 9 + uπ /2Lr9  ¯ 2π . m2K . sπ ) 48sπ2 Lr1 + 48sπ2 Lr2 + 24sπ2 Lr3 + 2sπ2 Lr9 − 16/3sπ tπ Lr1 + B(m − 40/3sπ tπ Lr2 − 4sπ tπ Lr3 + 2sπ tπ Lr9 − 16/3sπ uπ Lr1 − 40/3sπ uπ Lr2 − 4sπ uπ Lr3 + 2sπ uπ Lr9  ¯ 2η . sπ )m2π 3 sπ −200Lr1 − 68Lr3 + 116Lr4 + 46Lr5 − 21Lr9 + B(m − 9tπ Lr9 − 9uπ Lr9   ¯ 2π . uπ )m2K 2/3sπ Lr9 + 2/3tπ Lr9 + 16/3uπ Lr4 + 8/3uπ Lr5 + B(m + 4/3uπ Lr9  ¯ 2π . m2K . m2K .332 G. m2π . m2K . sπ )m2π sπ −256/3Lr1 + 80/3Lr2 − 16Lr3 + 42Lr4 + 10Lr5 − 4Lr9 + B(m  ¯ 2K . m2K . tπ )m4π −4Lr − 28/3Lr − 16/9Lr + 8/3Lr + 32/3Lr7 + B(m 3 4 5 6 + 8Lr8 − 8/3Lr9   ¯ 2η . uπ )m4π 56/3Lr2 + 64/3Lr4 + 20/3Lr5 − 8/3Lr9 + B(m  ¯ 2π . m2π . m2K . sπ ) m2K 32/3sπ Lr4 − 2sπ Lr9 + 2sπ2 Lr9 + 2sπ tπ Lr9 + 2sπ uπ Lr9 + B(m  ¯ 2π . Amorós et al. m2π . m2K . uπ )m2π m2K 40/3Lr2 + 32Lr6 + 16Lr8 − 8/3Lr9 + B(m  ¯ 2π . m2K . tπ )m4π 8/3Lr2 − 8/3Lr3 + 14/3Lr4 + 3Lr5 − Lr9 + B(m  ¯ 2π . / Nuclear Physics B 585 (2000) 293–352   + m4η (16π 2) Lr2 + 1/2Lr3 − 6Lr5  ¯ 2η ) 8Lr + 2Lr + Lr + 3Lr + m2η A(m 1 2 3 5  ¯ 2π .

tπ )m2π 3 8sπ Lr1 + 4sπ Lr2 + 5sπ Lr3 + 2sπ Lr9 − 24tπ Lr1 + 12tπ Lr2 + 39tπ Lr3 + 22tπ Lr4 − 20tπ Lr5 + 10tπ Lr9 + 8uπ Lr1 + 4uπ Lr2 + 5uπ Lr3 + 2uπ Lr9   + B¯ 1 (m2π . m2K .G. m2π . m2K . m2K . sπ )m4π 56/9Lr4 + 10/3Lr5 − 16Lr6 + 64Lr7 + 24Lr8 − 2Lr9 + B(m  ¯ 2η . sπ ) 200/3sπ2 Lr1 + 12sπ2 Lr2 + 68/3sπ2 Lr3 + 2/3sπ2 Lr9 − 64/3sπ tπ Lr2 − 16/3sπ tπ Lr3 + 2/3sπ tπ Lr9 − 64/3sπ uπ Lr2 − 16/3sπ uπ Lr3 + 2/3sπ uπ Lr9  + B¯ 1 (m2π . m2K . m2K . m2K . m2K . m2K . tπ ) −4Lr m2K m2η − 2Lr m2π m2η − 2Lr m4η + B(m 4 4 5  ¯ 2η . tπ )m4π −8Lr1 − 20Lr2 − 18Lr3 − 46/3Lr4 + 16/3Lr5 − 4/3Lr9  + B¯ 1 (m2π . tπ )m2π m2K /3 16Lr1 + 8Lr2 − 2Lr3 + 56Lr4 + 36Lr5 − 76Lr6 − 48Lr8 − 4Lr9   + B¯ 1 (m2π . Amorós et al. m2K . sπ )m2π m2K −32Lr2 − 8Lr3 + 64Lr4 − 4Lr9  + B¯ 1 (m2π . m2π . m2η . m2π . m2η . sπ ) 16sπ Lr4 m2K + m2η −24sπ Lr1 − 4sπ Lr3 + B(m  + B¯ 1 (m2π . tπ )m4K 3 8Lr1 + 4Lr2 − 4Lr3 − 40Lr4 − 20Lr5 + 40Lr6 + 20Lr8 − Lr9   + B¯ 1 (m2π . tπ )mK 3 −3Lr3 − Lr9 (sπ + uπ ) + 47/3tπ Lr3 + 24tπ Lr4 − 4tπ Lr5 + 2tπ Lr9  ¯ 2η . m2π . m2η . tπ )m2π 3 3Lr3 + 4Lr9 (sπ + uπ ) + 43/3tπ Lr3 + 12tπ Lr4 + B(m − 2tπ Lr5 + 10tπ Lr9   ¯ 2η . sπ )m4π −8Lr3 + 32Lr4 + 40Lr5 + 64Lr6 + 32Lr8 − 8Lr9  + B¯ 1 (m2π . sπ )m2π −4sπ Lr4 − 2sπ Lr5 + sπ Lr9 + tπ Lr9 + uπ Lr9 + B(m   ¯ 2η . tπ )m2K 3 −8sπ Lr1 − 4sπ Lr2 + sπ Lr3 + sπ Lr9 − 8tπ Lr1 + 4tπ Lr2 + 17tπ Lr3 − 4tπ Lr4 − 2tπ Lr5 + 5tπ Lr9 − 8uπ Lr1 − 4uπ Lr2 + uπ Lr3 + uπ Lr9 . tπ ) −1/3sπ tπ Lr3 − sπ tπ Lr9 − 7/3tπ2 Lr3 − tπ2 Lr9 − 1/3tπ uπ Lr3 + B(m − tπ uπ Lr9  ¯ 2η . sπ )m2π m2η 112/3Lr + 32/9Lr − 24Lr − 4Lr + B(m 1 3 4 5  ¯ 2η . sπ )m2π −368/3sπ Lr1 + 80/3sπ Lr2 − 32sπ Lr3 + 24sπ Lr4 + 8/3sπ Lr9 + 16tπ Lr2 + 4tπ Lr3 + 4tπ Lr9 + 16uπ Lr2 + 4uπ Lr3 + 4uπ Lr9  + B¯ 1 (m2π . m2η . tπ )m4K 3 −14/3Lr + 20Lr + 38/3Lr + 8Lr − 64Lr7 + B(m 3 4 5 6 r r − 28L8 + L9    2 2 2 ¯ + B(mη . / Nuclear Physics B 585 (2000) 293–352 333    ¯ 2η . sπ )m2π m2K −224/9Lr4 − 16/3Lr5 + 64Lr6 − 64Lr7 − Lr9 + B(m  ¯ 2η . m2η . sπ )m2K 128/3sπ Lr2 + 32/3sπ Lr3 − 32/3sπ Lr4 − 2/3sπ Lr9  + B¯ 1 (m2π . mK . m2π .

sπ )m2π 256/3sπ Lr1 − 80/3sπ Lr2 + 16sπ Lr3 − 36sπ Lr4 − 2sπ Lr5 + 2sπ Lr9  + B¯ 1 (m2K . / Nuclear Physics B 585 (2000) 293–352  + B¯ 1 (m2π . uπ )1/3 −16sπ uπ Lr1 − 8sπ uπ Lr2 − 8sπ uπ Lr3 − 2sπ uπ Lr9 − 16tπ uπ Lr1 − 8tπ uπ Lr2 − 8tπ uπ Lr3 − 2tπ uπ Lr9 + 24u2π Lr1 + 52u2π Lr2 + 12u2π Lr3 − 2u2π Lr9  + B¯ 1 (m2K . tπ )m4π 25/3Lr3 + 55/3Lr4 + 16/9Lr5 − 8/3Lr6 − 32/3Lr7 − 8Lr8 + 17/3Lr9   + B¯ 1 (m2η . uπ )m4K 3 −40Lr1 − 20Lr2 − 20Lr3 + 24Lr4  + B¯ 1 (m2π . m2K . sπ )m2K 32/3sπ Lr1 − 64/3sπ Lr2 + 16sπ Lr4 + sπ Lr9  + B¯ 1 (m2K . tπ )m4π 1/3Lr4 + 1/3Lr5 − 8/3Lr6 − 8/3Lr8    + B¯ 1 (m2K . m2η . uπ )m2π m2K −32/3Lr1 − 16/3Lr2 − 16/3Lr3 − 16/3Lr4   + B¯ 1 (m2π . m2K . m2η . m2K . uπ )m2π 3 32sπ Lr1 + 16sπ Lr2 + 16sπ Lr3 + 32tπ Lr1 + 16tπ Lr2 + 16tπ Lr3 + 48uπ Lr1 − 48uπ Lr2 + 24uπ Lr3 + 8uπ Lr4 − 4uπ Lr5 + 4uπ Lr9   + B¯ 1 (m2π . m2K . m2K . tπ ) 8sπ tπ Lr1 + 4sπ tπ Lr2 − sπ tπ Lr3 − 4sπ tπ Lr9 − 8tπ2 Lr2 − 13tπ2 Lr3  − 4tπ2 Lr9 + 8tπ uπ Lr1 + 4tπ uπ Lr2 − tπ uπ Lr3 − 4tπ uπ Lr9 3  + B¯ 1 (m2π . sπ ) −80/3sπ2 Lr1 − 200/3sπ2 Lr2 − 22sπ2 Lr3 − sπ2 Lr9 + 16/3sπ tπ Lr1 + 40/3sπ tπ Lr2 + 4sπ tπ Lr3 − sπ tπ Lr9 + 16/3sπ uπ Lr1 + 40/3sπ uπ Lr2 + 4sπ uπ Lr3 − sπ uπ Lr9  + B¯ 1 (m2K . m2π . Amorós et al. tπ )m2π −19/9sπ Lr3 − 17/6sπ Lr9 − 79/9tπ Lr3 − 3tπ Lr4 + 2tπ Lr5 − 35/6tπ Lr9 − 19/9uπ Lr3 − 17/6uπ Lr9   + B¯ 1 (m2η . m2K . m2K . uπ )m2K 16/3sπ Lr1 + 8/3sπ Lr2 + 8/3sπ Lr3 + 16/3tπ Lr1 + 8/3tπ Lr2 + 8/3tπ Lr3 + 16/3uπ Lr1 − 32/3uπ Lr2 + 8/3uπ Lr3 − 56/3uπ Lr4 − 16/3uπ Lr5 + 2/3uπ Lr9  + B¯ 1 (m2π . m2K .334 G. tπ ) m4K 46/3Lr4 + 2/3Lr5 − 8/3Lr6 − 4/3Lr8 − 6tπ Lr4 m2K  + B¯ 1 (m2η . m2K . m2η . tπ )m4K −20/3Lr4 − 10/3Lr5 + 40/3Lr6 + 20/3Lr8  + B¯ 1 (m2K . m2K . m2K . tπ )m2π m2K −10/3Lr4 + 20/3Lr6  + B¯ 1 (m2K . tπ )m2π m2K 3 −56/3Lr3 + 73Lr4 + 22/3Lr5 − 20Lr6 − 32Lr7 − 16Lr8 − 5/2Lr9  + B¯ 1 (m2η . uπ )m4π −40Lr1 − 20Lr2 − 20Lr3 − 56/3Lr4  + B¯ 1 (m2π . tπ ) m2π m2K 25/3Lr4 + 7Lr5 − 20/3Lr6 + m2π −3tπ Lr4 − 3tπ Lr5   + B¯ 1 (m2K . tπ )m4K 3 −7/3Lr3 − 74Lr4 − 38/3Lr5 − 8Lr6 + 64Lr7 + 28Lr8 − 11/2Lr9 . m2K . m2K . m2π . m2K .

m2K . sπ ) m2π −32sπ Lr2 − 32/3sπ Lr4 − 40/3sπ Lr5 + 8/3sπ Lr9  − 16tπ Lr2 − 4tπ Lr3 − 16uπ Lr2 − 4uπ Lr3 o + m2K −64/3sπ Lr4 + 4/3sπ Lr9  + B¯ 21 (m2π . m2K . m2K . tπ ) −5/6sπ tπ Lr9 − 8tπ2 Lr1 − 44/3tπ2 Lr2 − 34/3tπ2 Lr3 − 5/6tπ2 Lr9 − 5/6tπ uπ Lr9    + B¯ 21 (m2π . m2K . m2K . Amorós et al. m2η . sπ ) m4π + m2π m2K 32Lr2 + 8Lr3 n + B¯ 21 (m2π . tπ )m2K 32/3sπ Lr1 + 16/3sπ Lr2 + 2/3sπ Lr3 − 3/2sπ Lr9 − 16/3tπ Lr4 − 20/3tπ Lr5 − 2/3tπ Lr9 + 32/3uπ Lr1 + 16/3uπ Lr2 + 2/3uπ Lr3 − 3/2uπ Lr9  + B¯ 21 (m2π . m2η . m2K . tπ ) m4π 9Lr4 + 9Lr5 + m2π m2K 9Lr4 − 9Lr5  + m2π −3tπ Lr4 − 3tπ Lr5 . tπ )m2π 3 −32sπ Lr1 − 16sπ Lr2 − 8sπ Lr3 + sπ /2Lr9 + 48tπ Lr2 + 36tπ Lr3 − 52tπ Lr4 − 23tπ Lr5 + 19/2tπ Lr9 − 32uπ Lr1 − 16uπ Lr2 − 8uπ Lr3 + uπ /2Lr9  + B¯ 21 (m2π . sπ )m2π −112/3sπ Lr1 − 32/9sπ Lr3 + 24sπ Lr4 + 4sπ Lr5 − 16/3tπ Lr2 − 4/9tπ Lr3 − 16/3uπ Lr2 − 4/9uπ Lr3  + B¯ 1 (m2η . m2η . m2K . tπ ) 1/3Lr3 + 3/2Lr9 (sπ + uπ )tπ + 4tπ2 Lr3 + 3/2tπ2 Lr9    + B¯ 1 (m2η . m2K . uπ ) m4K 40/3Lr1 + 20/3Lr2 + 20/3Lr3 − 8u2π Lr1 + 4/3u2π Lr2 − 4u2π Lr3  + B¯ 21 (m2K . m2K . tπ ) 2Lr4 m2π m2η + 4Lr4 m2K m2η + 2Lr5 m4η  + B¯ 1 (m2η . sπ ) m4π + m2π m2K 32/3Lr2 + 8/9Lr3  + B¯ 1 (m2η . uπ ) 16/3m2π + 8/3m2K (sπ + tπ ) −2Lr1 − Lr2 − Lr3   + B¯ 21 (m2π .G. tπ )m4π 40Lr1 + 20Lr2 + 10Lr3 + 16Lr4 + 9Lr5 − 3Lr9  + B¯ 21 (m2π . m2π . uπ ) 20m4π + 16/3m2π m2K 2Lr1 + Lr2 + Lr3    + B¯ 21 (m2π . tπ )m2π m2K −112/3Lr1 − 56/3Lr2 − 10/3Lr3 − 8Lr4 − 9Lr5 + 3/2Lr9   + B¯ 21 (m2π . sπ )4/3 44sπ2 Lr1 + 9sπ2 Lr2 + 16sπ2 Lr3 − sπ2 Lr9 − sπ tπ Lr9 − sπ uπ Lr9  + B¯ 21 (m2π . tπ )m4K −8/3Lr1 − 4/3Lr2 + 4/3Lr3 − 8Lr4 + 3/2Lr9  + B¯ 21 (m2π . m2K . m2η . m2K . m2K . m2π . m2π . tπ )m2K 25/9sπ Lr3 + 11/6sπ Lr9 − 41/9tπ Lr3 − 6tπ Lr4 + 4tπ Lr5 + 1/3tπ Lr9 + 25/9uπ Lr3 + 11/6uπ Lr9   + B¯ 1 (m2η . sπ ) 64/3sπ2 Lr1 − 56/3sπ2 Lr2 + 2sπ2 Lr3    + B¯ 21 (m2K . sπ ) 24sπ2 Lr1 + 12sπ2 Lr2 + 8sπ2 Lr3    + B¯ 21 (m2π . m2K . / Nuclear Physics B 585 (2000) 293–352 335  + B¯ 1 (m2η .

m2K .336 G. sπ ) −12sπ2 Lr2 − 4sπ2 Lr3  + B¯ 22 (m2π . sπ )m2π 16/3tπ Lr2 + 4/9tπ Lr3 + 16/3uπ Lr2 + 4/9uπ Lr3  + B¯ 21 (m2η . m2π . sπ ) m4π + m2π m2K 32/3Lr2 + 8/9Lr3  + B¯ 21 (m2η . Amorós et al. sπ ) m2π 32/3Lr1 + 368/3Lr2 + 32Lr3  + m2K 32/3Lr1 + 80/3Lr2 + 8Lr3  + B¯ 22 (m2K . sπ )m2K 128/3Lr2 + 32/3Lr3 − 64/3Lr4 + 4/3Lr9  + B¯ 22 (m2π . m2K . / Nuclear Physics B 585 (2000) 293–352    + B¯ 21 (m2K . m2K . m2η . tπ ) m2π −30Lr4 − 30Lr5 − 60m2K Lr4  + B¯ 22 (m2η . m2K . m2K . tπ ) m4π −13/3Lr3 − 9Lr4 − 3Lr9  + m2π m2K 56/9Lr3 − 9Lr4 + 3/2Lr9   + B¯ 21 (m2η . uπ ) m2π /3 160Lr1 + 80Lr2 + 80Lr3 + 16Lr4  + 16m2K 2Lr1 + Lr2 + Lr3   − B¯ 22 (m2π . m2η . m2η . tπ ) −16/3sπ Lr + 40/3sπ Lr − 16/3sπ Lr − 16/3sπ Lr 1 2 3 9 + 64/3tπ Lr1 − 24tπ Lr2 − 4tπ Lr3 − 16/3tπ Lr9 + 32/3uπ Lr1 + 40/3uπ Lr2 − 4/3uπ Lr3 − 16/3uπ Lr9   + B¯ 22 (m2π . m2K . m2K . tπ ) m4K −18Lr4 + m2K −6tπ Lr4   + B¯ 21 (m2η . m2η . tπ )m2K 8/9Lr3 + 52Lr4 − 8/3Lr5 + 5Lr9  + B¯ 22 (m2η . tπ ) m2π 10/9sπ + 4tπ + 10/9uπ Lr3 − tπ Lr4 − 4/3tπ Lr5    + 3sπ + 5tπ + 3uπ Lr9 /2 + m4K 7/9Lr3 + 18Lr4 + 3/2Lr9 + m2K −16/9sπ Lr3  − 3/2sπ Lr9 − 2tπ Lr4 − 8/3tπ Lr5 − tπ Lr9 − 16/9uπ Lr3 − 3/2uπ Lr9  + B¯ 21 (m2η . tπ ) −22/3sπ Lr3 − 5sπ Lr9 + 20/3tπ Lr3 − 5tπ Lr9 − 16/3uπ Lr3 − 5uπ Lr9   + B¯ 22 (m2η . m2K . tπ ) −1/2sπ tπ Lr9 − 5/3tπ2 Lr3 − 1/2tπ2 Lr9 − 1/2tπ uπ Lr9    − B¯ 21 (m2η . m2K . m2η . sπ )m2π 272/3Lr2 + 56/3Lr3 − 32/3Lr4 − 40/3Lr5 + 8/3Lr9  + B¯ 22 (m2π . m2K . m2η . m2K . m2π . uπ )8/3 2Lr1 + Lr2 + Lr3 (sπ + tπ ) + 10uπ Lr1 + 3uπ Lr2 + 5uπ Lr3   + B¯ 22 (m2K . tπ )m2K 4Lr3 − 88/3Lr4 − 20/3Lr5 + 16/3Lr9  + B¯ 22 (m2π . m2K . sπ ) 32/3sπ Lr1 − 208/3sπ Lr2 − 12sπ Lr3 − 16/3tπ Lr1 − 64/3tπ Lr2 − 6tπ Lr3 − 16/3uπ Lr1 − 16/3uπ Lr2 − 2uπ Lr3   + B¯ 22 (m2K . sπ ) 176sπ Lr1 − 128sπ Lr2 + 24sπ Lr3 − 4sπ Lr9 − 64tπ Lr2 − 4tπ Lr3 − 4tπ Lr9 − 64uπ Lr2 − 28uπ Lr3 − 4uπ Lr9 /3   + B¯ 22 (m2π . m2π . tπ )m2π 3 −32Lr1 − 112Lr2 + 88Lr3 − 196Lr4 − 104Lr5 + 32Lr9  + B¯ 22 (m2π . tπ )m2π 268/9Lr3 + 26Lr4 − 4/3Lr5 + 10Lr9  + B¯ 22 (m2η . sπ ) m2π 112/3Lr2 + 136/9Lr3 − 24sπ Lr2 − 8sπ Lr3 . m2K .

m2K . m2η . m2K . sπ )4/3 4Lr1 + 10Lr2 + 3Lr3 (tπ + uπ )sπ  + B¯ 31 (m2K . m2K . m2π . uπ ) m2π + m2K /3 −16uπ Lr1 − 8uπ Lr2 − 8uπ Lr3  + B¯ 31 (m2π . tπ ) −64/3sπ Lr − 32/3sπ Lr − 4/3sπ Lr + 112/3tπ Lr 1 2 3 1 r r r + 56/3tπ L2 + 10/3tπ L3 − 112/3uπ L1 − 56/3uπ Lr2 − 10/3uπ Lr3    − B¯ 32 (m2π . m2K . m2π . m2η . sπ ) m2π + m2K −128/3sπ Lr2 − 32/3sπ Lr3   + B¯ 31 (m2π . m2π . / Nuclear Physics B 585 (2000) 293–352 337  + B¯ 31 (m2π . sπ ) −32sπ Lr1 − 80sπ Lr2 − 24sπ Lr3 + 80tπ Lr1 + 104tπ Lr2 + 48tπ Lr3 − 16uπ Lr1 + 56uπ Lr2 /3  + B¯ 32 (m2K . m2K . m2K . (A. sπ )32/3 6sπ Lr1 + 7sπ Lr2 + 4sπ Lr3 − 4tπ Lr2 − tπ Lr3 − 4uπ Lr2 − uπ Lr3   + B¯ 32 (m2π . tπ )m2π 8Lr1 + 4Lr2 + Lr3 (sπ + uπ ) + 8tπ Lr1 + 4tπ Lr2 − tπ Lr3   + B¯ 31 (m2π .G. uπ ) 16/3 3m2π + m2K + sπ + tπ − uπ 2Lr1 + Lr2 + Lr3  + B¯ 32 (m2K . tπ ) m2π (−sπ − tπ − uπ )Lr3 + m2K (sπ − 1/3tπ + uπ )Lr3   + B¯ 31 (m2K . tπ )m2K −4(sπ − tπ /3 + uπ ) 2Lr1 + Lr2 − (sπ + tπ /3 + uπ )Lr3    + B¯ 31 (m2π . m2K . m2K . tπ ) m2π 96Lr1 + 48Lr2 + 8Lr3  − m2K 32/3Lr1 + 16/3Lr2 + 8/3Lr3  + B¯ 32 (m2π . tπ ) m4π −32Lr1 − 16Lr2 − 4Lr3 + m2π m2K 32Lr1 + 16Lr2 + 4Lr3   + B¯ 31 (m2π . tπ ) −8Lr1 − 4Lr2 + Lr3 (sπ + uπ )tπ /3 + 8Lr1 + 4Lr2 + Lr3 tπ2  + B¯ 31 (m2π . Amorós et al. m2K . sπ ) m2π 64Lr1 − 160/3Lr2 + 32/3Lr3  − m2K 256/3Lr2 + 64/3Lr3  − B¯ 32 (m2π . sπ )8/3 −6sπ2 Lr1 − 3sπ2 Lr2 − 3sπ2 Lr3 + 8Lr2 + 2Lr3 (uπ + tπ )sπ    + B¯ 31 (m2π . uπ )8/3 2sπ uπ Lr1 + sπ uπ Lr2 + sπ uπ Lr3 + 2tπ uπ Lr1 + tπ uπ Lr2 + tπ uπ Lr3  + B¯ 31 (m2K . m2K . m2K . tπ ) 4Lr3 m4π − m2π m2K + (1/3sπ tπ − tπ2 + 1/3tπ uπ )Lr3   + B¯ 32 (m2π .1) . m2K . sπ ) m2π + m2K −32/3sπ Lr1 − 80/3sπ Lr2 − 8sπ Lr3   + B¯ 31 (m2K . m2K . m2π . m2η . sπ ) m2π + m2K −64/3Lr1 − 160/3Lr2 − 16Lr3  + B¯ 32 (m2K . tπ )Lr3 −12m2π + 4/3m2K + 8/3sπ − 14/3tπ + 14/3uπ  ¯ 2π ) −4/3sπ Lr1 − 16sπ Lr2 − 19/3sπ Lr3 + 55/12sπ Lr9 − 27tπ Lr2 + A(m − 27/2tπ Lr3 + 55/12tπ Lr9 − 27uπ Lr2 + 55/12uπ Lr9  ¯ 2K ) −56/3sπ Lr1 − 32/3sπ Lr2 − 10sπ Lr3 + 11/2sπ Lr9 − 14/3tπ Lr2 + A(m + 17/3tπ Lr3 + 11/2tπ Lr9 − 110/3uπ Lr2 + 11/2uπ Lr9  ¯ 2η ) −12sπ Lr1 − sπ Lr3 − 3tπ Lr2 + 11/6tπ Lr3 − 3uπ Lr2 + A(m + 7/4(sπ + tπ + uπ )Lr9 .

m2K . m2K . uπ )m4π 56/3Lr + 64/3Lr + 20/3Lr − 8/3Lr + B(m 2 4 5 9  ¯ 2π . Amorós et al. tπ )m2π 2 sπ Lr9 + 2tπ 8/3Lr2 + 16/3Lr3 + 14/3Lr4 − 3Lr5 − B(m  + 3/2Lr9 + uπ Lr9  ¯ 2π . uπ )m4K −16/3Lr4 − 8/3Lr5 − 2/3Lr9 + B(m . m2K . uπ )4/3m2π sπ Lr + tπ Lr − 10uπ Lr − 8uπ Lr − 4uπ Lr + 2uπ Lr + B(m 9 9 2 4 5 9  ¯ 2π . tπ )m4K −8/3Lr4 − 4/3Lr5 + 40/3Lr6 + 20/3Lr8 + 1/2Lr9 + B(m  ¯ 2π . uπ )m2π m2K 40/3Lr2 + 32Lr6 + 16Lr8 − 8/3Lr9 + B(m  ¯ 2π . m2K . m2K . m2K . tπ )m2π m2K 8/3Lr + 16/3Lr − 8/3Lr − 5Lr + 20/3Lr + 3/2Lr + B(m 2 3 4 5 6 9   ¯ 2π . m2K . tπ )m2K −1/2sπ Lr − 4tπ Lr − 2tπ Lr − tπ Lr − 1/2uπ Lr + B(m 9 4 5 9 9  ¯ 2π . tπ ) 1/2sπ tπ Lr9 + 1/2tπ2 Lr9 + 1/2tπ uπ Lr9 + B(m  ¯ 2π . tπ )m4π −8/3Lr2 + 8/3Lr3 − 14/3Lr4 − 3Lr5 + Lr9 + B(m  ¯ 2π . m2K . m2K .338 G. / Nuclear Physics B 585 (2000) 293–352 For G form-factor the expression reads   GLL = m4π (16π 2 ) −2Lr1 − 10/9Lr2 − 49/108Lr3  + m4π −16Lr3Lr5 + 32Lr4 Lr5 + 56(Lr5 )2 − 64Lr5 Lr6 − 64Lr5 Lr8 − 16Lr5 Lr9   + m2π m2K (16π 2 ) 8/9Lr2 + 8/27Lr3  + m2π m2K 64Lr4 Lr5 − 32(Lr5 )2 − 128Lr5Lr6 + 8Lr5 Lr9   + m2π m2η (16π 2 ) −2/3Lr5 n + m2π 8sπ Lr5 Lr9 + 32tπ Lr2 Lr5 + 16tπ Lr3 Lr5 + 8tπ Lr5 Lr9 − 32uπ Lr2 Lr5 + 8uπ Lr5 Lr9  ¯ 2π ) −4Lr − 10Lr − 19Lr + 46/3Lr + 161/6Lr − 83/6Lr + A(m 1 2 3 4 5 9  ¯ 2K ) 32Lr2 + 17/3Lr3 − 8/3Lr4 + 25/3Lr5 − 19/3Lr9 + A(m o ¯ 2η ) 8/3Lr1 + 2/3Lr2 + 53/18Lr3 − 2Lr4 − 11/6Lr5 + 1/2Lr9 + A(m   + m4K (16π 2) 2Lr1 − 7/9Lr2 − 1/108Lr3  + m4K 16Lr3 Lr5 + 8(Lr5 )2 + 8Lr5 Lr9   + m2K m2η (16π 2) 8/3Lr5 n + m2K −8sπ Lr5 Lr9 − 32tπ Lr2 Lr5 − 16tπ Lr3 Lr5 − 8tπ Lr5 Lr9 + 32uπ Lr2 Lr5 − 8uπ Lr5 Lr9  ¯ 2π ) −27/2Lr3 − 4/3Lr4 − 2/3Lr5 − 83/12Lr9 + A(m ¯ 2K ) 4Lr1 + 26Lr2 + 5/6Lr3 + 32/3Lr4 − 8/3Lr5 − 19/6Lr9 + A(m  o ¯ 2η ) −32/3Lr − 8/3Lr − 47/18Lr − 4Lr − 8/3Lr + 1/4Lr + A(m 1 2 3 4 5 9  r  4 2 r r + mη (16π ) L2 + 1/6L3 − 2L5  ¯ 2η )Lr2 + 3A(m ¯ 2η )Lr3 + A(m ¯ 2η )Lr5 ¯ 2η )Lr1 + 2A(m + m2η 8A(m  ¯ 2π . m2K .

tπ )m2K 3 8sπ Lr1 + 4sπ Lr2 − sπ Lr3 − sπ Lr9 + 8tπ Lr1 − 4tπ Lr2 − 17tπ Lr3 + 4tπ Lr4 + 2tπ Lr5 − 5tπ Lr9 + 8uπ Lr1 + 4uπ Lr2 − uπ Lr3 − uπ Lr9   + B¯ 1 (m2π . uπ )m2π m2K −32/3Lr1 − 16/3Lr2 − 16/3Lr3 − 16/3Lr4 339 . m2K . m2K . tπ )m2π /3 (sπ + uπ ) 3Lr3 − 4Lr9 − B(m  + tπ −43/3Lr3 − 12Lr4 + 2Lr5 − 10Lr9  ¯ 2η . Amorós et al. tπ ) 1/3sπ tπ Lr3 + sπ tπ Lr9 + 7/3tπ2 Lr3 + tπ2 Lr9 + 1/3tπ uπ Lr3 + B(m + tπ uπ Lr9  + B¯ 1 (m2π . m2K . m2K . tπ ) m2K m2η 4Lr + 8Lr − 4Lr m4η + B(m 4 5 5    ¯ 2η . / Nuclear Physics B 585 (2000) 293–352  ¯ 2π . tπ )m4K 3 −8Lr1 − 4Lr2 + 4Lr3 + 40Lr4 + 20Lr5 − 40Lr6 − 20Lr8 + Lr9   + B¯ 1 (m2π . tπ )m4π 8Lr1 + 20Lr2 + 18Lr3 + 46/3Lr4 − 16/3Lr5 + 4/3Lr9   + B¯ 1 (m2π . m2K . m2K . tπ )m2K 3 3Lr + Lr (sπ + uπ ) + B(m 3 9  − tπ 47/3Lr3 + 24Lr4 − 4Lr5 + 2Lr9  ¯ 2η . m2K . m2K . m2K . tπ )m4π 4Lr3 + 28/3Lr4 + 16/9Lr5 − 8/3Lr6 − 32/3Lr7 − 8Lr8 + B(m + 8/3Lr9   ¯ 2η . uπ )m4π −40Lr1 − 20Lr2 − 20Lr3 − 56/3Lr4  + B¯ 1 (m2π . tπ )m2π m2K 3 −16Lr1 − 8Lr2 + 2Lr3 − 56Lr4 − 36Lr5 + 76Lr6 + 48Lr8 + 4Lr9   + B¯ 1 (m2π . m2K . uπ ) −2/3sπ uπ Lr9 − 2/3tπ uπ Lr9 − 2/3u2π Lr9 + B(m  ¯ 2η . m2K . m2K . tπ )m4K 14/9Lr − 20/3Lr − 38/9Lr − 8/3Lr + 64/3Lr7 + B(m 3 4 5 6 + 28/3Lr8 − 1/3Lr9   ¯ 2η . tπ ) (sπ + uπ )tπ −8Lr1 − 4Lr2 + Lr3 + 4Lr9  + tπ2 8Lr2 + 13Lr3 + 4Lr9  + B¯ 1 (m2π . m2K . tπ )m2π m2η 2Lr4 − 2Lr5 + B(m   ¯ 2η . tπ )m2π 3 −8sπ Lr1 − 4sπ Lr2 − 5sπ Lr3 − 2sπ Lr9 + 24tπ Lr1 − 12tπ Lr2 − 39tπ Lr3 − 22tπ Lr4 + 20tπ Lr5 − 10tπ Lr9 − 8uπ Lr1 − 4uπ Lr2 − 5uπ Lr3 − 2uπ Lr9   + B¯ 1 (m2π . tπ )m2π m2K 3 −2/3Lr3 + 46Lr4 + 22/3Lr5 − 20Lr6 − 32Lr7 + B(m − 16Lr8 + 2Lr9  ¯ 2η . uπ )m2K 2/3sπ Lr9 + 2/3tπ Lr9 + 16/3uπ Lr4 + 8/3uπ Lr5 + B(m + 4/3uπ Lr9  ¯ 2π . m2K . m2K . m2K .G. m2K . m2K .

tπ )m2π m2K 112/3Lr1 + 56/3Lr2 + 10/3Lr3 + 8Lr4 + 9Lr5 − 3/2Lr9  + B¯ 21 (m2π . m2K . uπ )8/3m2K 2sπ Lr1 + sπ Lr2 + sπ Lr3 + 2tπ Lr1 + tπ Lr2 + tπ Lr3 + 2uπ Lr1 − 4uπ Lr2 + uπ Lr3 − 7uπ Lr4 − 2uπ Lr5 + 1/4uπ Lr9  + B¯ 1 (m2π . m2π . tπ )m4K 3 7/3Lr3 + 74Lr4 + 38/3Lr5 + 8Lr6 − 64Lr7 − 28Lr8 + 11/2Lr9  + B¯ 1 (m2η . m2K . m2K . tπ ) m2π m2η 2Lr4 − 4/3Lr5 + m2K m2η 4Lr4 + 16/3Lr5 + 2m4η Lr5  + B¯ 1 (m2η . / Nuclear Physics B 585 (2000) 293–352   + B¯ 1 (m2π . tπ ) m2π m2K −25/3Lr4 − 7Lr5 + 20/3Lr6 + 6m2K tπ Lr4    + B¯ 1 (m2K . m2K . m2η . m2K . tπ )m4π −40Lr1 − 20Lr2 − 10Lr3 − 16Lr4 − 9Lr5 + 3Lr9  + B¯ 21 (m2π . tπ )m4K 8/3Lr1 + 4/3Lr2 − 4/3Lr3 + 8Lr4 − 3/2Lr9 . tπ )m2K −25/9sπ Lr3 − 11/6sπ Lr9 + 41/9tπ Lr3 + 6tπ Lr4 − 4tπ Lr5 − 1/3tπ Lr9 − 25/9uπ Lr3 − 11/6uπ Lr9   − B¯ 1 (m2η . uπ )m4K −40/3Lr1 − 20/3Lr2 − 20/3Lr3 + 8Lr4  + B¯ 1 (m2π . m2η . m2K . tπ )m2π m2K /3 56/3Lr3 − 73Lr4 − 22/3Lr5 + 20Lr6 + 32Lr7 + 16Lr8 + 5/2Lr9    − B¯ 1 (m2η . m2η . tπ ) 1/3Lr3 + 3/2Lr9 (sπ + uπ )tπ + 4tπ2 Lr3 + 3/2tπ2 Lr9  + B¯ 21 (m2π . m2K . Amorós et al. m2K . tπ ) 3m2π tπ Lr4 + Lr5 + m4K /3 −46Lr4 − 2Lr5 + 8Lr6 + 4Lr8   + B¯ 1 (m2η . tπ )m2π 32/3sπ Lr1 + 16/3sπ Lr2 + 8/3sπ Lr3 − 3/2sπ Lr9 − 16tπ Lr2 − 12tπ Lr3 + 52/3tπ Lr4 + 23/3tπ Lr5 − 19/6tπ Lr9 + 32/3uπ Lr1 + 16/3uπ Lr2 + 8/3uπ Lr3 − 3/2uπ Lr9  + B¯ 21 (m2π . m2K . m2K . tπ )m4π −1/3Lr4 − 1/3Lr5 + 8/3Lr6 + 8/3Lr8   + B¯ 1 (m2K . tπ )m2π 19/9sπ Lr3 + 17/6sπ Lr9 + 79/9tπ Lr3 + 3tπ Lr4 − 2tπ Lr5 + 35/6tπ Lr9 + 19/9uπ Lr3 + 17/6uπ Lr9   + B¯ 1 (m2η . m2K . m2K . m2K .340 G. tπ )10/3 m2π m2K Lr4 − 2Lr6 + m4K 2Lr4 + Lr5 − 4Lr6 − 2Lr8  + B¯ 1 (m2K . uπ )4/3m2π 8Lr1 + 4Lr2 + 4Lr3 (sπ + tπ )  + uπ 12Lr1 − 12Lr2 + 6Lr3 + 2Lr4 − Lr5 + Lr9  + B¯ 1 (m2π . m2K . uπ )1/3 −16sπ uπ Lr1 − 8sπ uπ Lr2 − 8sπ uπ Lr3 − 2sπ uπ Lr9 − 16tπ uπ Lr1 − 8tπ uπ Lr2 − 8tπ uπ Lr3 − 2tπ uπ Lr9 + 24u2π Lr1 + 52u2π Lr2 + 12u2π Lr3 − 2u2π Lr9    + B¯ 1 (m2K . m2K . tπ )m4π 3 −25Lr3 − 55Lr4 − 16/3Lr5 + 8Lr6 + 32Lr7 + 24Lr8 − 17Lr9  + B¯ 1 (m2η .

m2K . m2K . uπ )m2K 32Lr1 + 16Lr2 + 16Lr3 . m2K . tπ )m2K 16/9Lr3 + 3/2Lr9 (sπ + uπ ) + 2tπ Lr4 + 8/3tπ Lr5 + tπ Lr9  + B¯ 21 (m2η . m2K . uπ ) −8u2π Lr1 + 4/3u2π Lr2 − 4u2π Lr3    + B¯ 21 (m2K . uπ )16/3m2π −2sπ Lr1 − sπ Lr2 − sπ Lr3 − 2tπ Lr1 − tπ Lr2 − tπ Lr3 + uπ Lr4  + B¯ 21 (m2π . m2η . m2K . sπ ) m2π 8Lr3 − 32Lr4 − 24Lr5 + 8Lr9 + m2K 8Lr3 + 4Lr9  + B¯ 22 (m2π . tπ )m2π m2K −56/9Lr3 + 9Lr4 − 3/2Lr9   − B¯ 21 (m2η . tπ )m2π −64/3Lr1 − 80/3Lr2 − 64/3Lr3 + 4/3Lr4 − 4/3Lr5 + 4/3Lr9  + B¯ 22 (m2π . m2K . tπ )m2K 32Lr1 + 4Lr3 − 8/3Lr4 + 20/3Lr5 + 2/3Lr9  + B¯ 22 (m2π . m2K . / Nuclear Physics B 585 (2000) 293–352 341  + B¯ 21 (m2π . m2η . uπ )8/3m2K 2sπ Lr1 + sπ Lr2 + sπ Lr3 + 2tπ Lr1 + tπ Lr2 + tπ Lr3  + B¯ 21 (m2π . m2π . uπ )m4K 40/3Lr1 + 20/3Lr2 + 20/3Lr3  − B¯ 21 (m2π . m2K . tπ ) 18m4K Lr4 + 6m2K tπ Lr4 + m4π 13/3Lr3 + 9Lr4 + 3Lr9  + B¯ 21 (m2η . m2K . m2π . tπ )m4K −7/9Lr3 − 18Lr4 − 3/2Lr9   + B¯ 21 (m2η .G. uπ )m2π 160/3Lr1 + 80/3Lr2 + 80/3Lr3 + 16/3Lr4  + B¯ 22 (m2π . tπ )m2π 10/9Lr3 + 3/2Lr9 (sπ + uπ )  − tπ −4Lr3 + Lr4 + 4/3Lr5 − 5/2Lr9  + B¯ 21 (m2η . m2K . tπ )m2K −32/3sπ Lr1 − 16/3sπ Lr2 − 2/3sπ Lr3 + 3/2sπ Lr9 + 16/3tπ Lr4 + 20/3tπ Lr5 + 2/3tπ Lr9 − 32/3uπ Lr1 − 16/3uπ Lr2 − 2/3uπ Lr3 + 3/2uπ Lr9   + B¯ 21 (m2π . m2K . m2K . m2K . sπ )4 4sπ Lr1 − 2sπ Lr2 + 2sπ Lr3 − sπ Lr9 − tπ Lr3 − tπ Lr9 − uπ Lr3 − uπ Lr9  + B¯ 22 (m2π . m2K . m2K . tπ ) −9m4π Lr4 + Lr5 + 9m2π m2K −Lr4 + Lr5  + 3m2π tπ Lr4 + Lr5   + B¯ 21 (m2K . tπ ) 5/2sπ tπ Lr9 + tπ2 24Lr1 + 44Lr2 + 34Lr3 + 5/2Lr9  + 5/2tπ uπ Lr9 3    + B¯ 21 (m2π . tπ ) 16/3sπ Lr1 + 8/3sπ Lr2 + 4/3sπ Lr3 − 2/3sπ Lr9 + 32/3tπ Lr1 + 8tπ Lr2 + 16tπ Lr3 − 2/3tπ Lr9 − 32/3uπ Lr1 + 8/3uπ Lr2 − 8/3uπ Lr3 − 2/3uπ Lr9  + B¯ 22 (m2π . tπ ) 1/2sπ tπ Lr9 + 5/3tπ2 Lr3 + 1/2tπ2 Lr9 + 1/2tπ uπ Lr9    + B¯ 22 (m2π . Amorós et al. m2K . m2K . uπ ) 20m4π + 16/3m2π m2K 2Lr1 + Lr2 + Lr3  + B¯ 21 (m2π .

m2K . m2K . m2K . sπ ) m2π + m2K −32Lr1 − 16Lr2 − 16Lr3  + B¯ 32 (m2K . tπ )m2K 28/9Lr3 + 20Lr4 + 8/3Lr5 + Lr9  + B¯ 22 (m2η . / Nuclear Physics B 585 (2000) 293–352  + B¯ 22 (m2π . m2K . tπ ) −32sπ Lr1 − 16sπ Lr2 − 8sπ Lr3 − 112tπ Lr1 − 56tπ Lr2  − 10tπ Lr3 + 16uπ Lr1 + 8uπ Lr2 − 2uπ Lr3 3    − B¯ 32 (m2π . m2η . Amorós et al. uπ )8uπ m2π + m2K /3 2Lr1 + Lr2 + Lr3  + B¯ 31 (m2π . m2η . tπ ) −2/3sπ Lr3 − sπ Lr9 + 16/3tπ Lr3 − tπ Lr9 − 8/3uπ Lr3 − uπ Lr9    + B¯ 31 (m2π . tπ )m2π −16/9Lr3 + 10Lr4 + 4/3Lr5 + 2Lr9  + B¯ 22 (m2η . m2K . uπ ) −16/3sπ Lr1 − 8/3sπ Lr2 − 8/3sπ Lr3 − 16/3tπ Lr1 − 8/3tπ Lr2 − 8/3tπ Lr3 − 80/3uπ Lr1 − 8uπ Lr2 − 40/3uπ Lr3  + B¯ 22 (m2K . m2K . tπ )Lr3 m2π (sπ + tπ + uπ ) + m2K (−sπ + 1/3tπ − uπ )    + B¯ 32 (m2π . tπ ) m4π − m2π m2K 32Lr1 + 16Lr2 + 4Lr3   − B¯ 31 (m2π . m2η . tπ ) (sπ + uπ )tπ /3 8Lr1 + 4Lr2 − Lr3 − 8tπ2 Lr1 − 4tπ2 Lr2 − tπ2 Lr3    − B¯ 31 (m2π . m2K .342 G. m2K . m2K . sπ ) 16tπ Lr1 + 8tπ Lr2 + 8tπ Lr3 + 16uπ Lr1 + 8uπ Lr2 + 8uπ Lr3   + B¯ 32 (m2K . m2K . m2K . tπ )Lr3 4 −m4π + m2π m2K − 1/3sπ tπ + tπ2 − 1/3tπ uπ  + B¯ 31 (m2K . m2K . m2K . tπ ) m2π −6Lr4 − 6Lr5 − 12m2K Lr4  + B¯ 22 (m2η . sπ )m2K 32Lr1 + 12Lr3 − 16Lr4 + 2Lr9  + B¯ 22 (m2K . m2K . tπ ) m2π 64Lr1 + 32Lr2 + 12Lr3 − m2K /3 64Lr1 + 32Lr2 + 4Lr3  + B¯ 32 (m2π . (A. sπ )2 2sπ Lr3 − sπ Lr9 − 8tπ Lr1 − 3tπ Lr3 − tπ Lr9 − 8uπ Lr1 − 3uπ Lr3 − uπ Lr9   + B¯ 22 (m2K . sπ )m2π 32Lr1 + 12Lr3 − 12Lr5 + 4Lr9  + B¯ 22 (m2K . tπ ) 8 −m2π + m2K /3 Lr3 + 4/3sπ Lr3 + 14/3tπ Lr3 − 2/3uπ Lr3 ¯ 2π ) 83/12sπ Lr9 + 27tπ Lr2 + 27/2tπ Lr3 + 83/12tπ Lr9 − 27uπ Lr2 + A(m  + 83/12uπ Lr9 ¯ 2K ) 19/6sπ Lr9 + 14/3tπ Lr2 − 17/3tπ Lr3 + 19/6tπ Lr9 − 110/3uπ Lr2 + A(m  + 19/6uπ Lr9 ¯ 2η ) −1/4sπ Lr9 + 27tπ + 3tπ Lr2 − 11/6tπ Lr3 − 1/4tπ Lr9 − 3uπ Lr2 + A(m  − 1/4uπ Lr9 . m2η . uπ ) 16/3 3m2π + 3m2K − sπ − tπ + uπ 2Lr1 + Lr2 + Lr3  + B¯ 32 (m2K . tπ )m2π 4 2Lr1 + Lr2 (sπ + tπ + uπ ) + sπ Lr3 − tπ Lr3 + uπ Lr3   + B¯ 31 (m2π . m2K .2) . tπ )m2K (sπ + uπ ) 8Lr1 + 4Lr2 + Lr3 − 8/3tπ Lr1 − 4/3tπ Lr2 + 1/3tπ Lr3   + B¯ 31 (m2π . m2K . m2K . uπ )8/3 2sπ uπ Lr1 + sπ uπ Lr2 + sπ uπ Lr3 + 2tπ uπ Lr1 + tπ uπ Lr2 + tπ uπ Lr3   + B¯ 31 (m2K . m2K .

L6 contribution Here we collect the relevant contribution coming from the tree diagram involving a O(p6 ) Lagrangian vertex [24].2. Amorós et al. For the F form-factor we obtain  r r r r + 16C11 − 80C12 − 176C13 Fct = m4π −32C4r − 32C5r − 24C6r − 64C7r + 8C8r + 8C10 r r r r r r r r + 32C14 + 8C15 + 128C16 − 48C17 − 32C23 + 16C25 + 32C26 + 128C28 r r r r + 8C64 + 8C65 + 8C66 − 16C67 r r r r r + 32C68 + 16C69 + 16C83 + 32C84 + 8C90  r r r + 48C11 + 24C12 + m2π m2K −16C4r + 8C5r − 40C6r − 64C7r − 24C8r + 16C10 r r r r r r r r − 96C13 − 16C14 + 80C15 + 64C17 − 16C22 − 16C23 − 8C25 − 16C26 r r r r r r r r − 32C29 − 64C30 − 32C36 + 16C63 + 20C64 + 4C65 − 4C66 − 24C67 r r r r r + 16C68 + 8C69 + 8C83 + 16C84 + 12C90  r r + 8C22 + m2π sπ 20C1r + 64C2r + 8C3r + 8C4r + 16C5r + 16C6r + 32C7r + 16C13 r r r r r r r r + 8C25 − 4C64 − 4C65 − 12C66 + 8C67 − 32C68 − 16C69 − 8C83 r r − 16C84 − 4C90  r r r + 32C13 + 12C22 + m2π tπ −20C1r − 64C2r + 8C3r − 4C4r − 8C6r − 8C8r + 8C12 r r r r r r r r + 16C23 − 20C25 + 4C63 − 4C64 − 4C65 + 6C66 + 16C67 − 16C68 r r r r r − 10C69 − 12C83 − 16C84 + 8C88 − 10C90  r r r r − 16C11 + 8C12 + 80C13 + m2π uπ −12C1r − 64C2r − 8C3r + 36C4r − 8C10 r r r r r r r r − 12C22 + 16C23 + 4C25 − 4C63 − 4C64 − 4C65 − 6C66 + 16C67 r r r r r − 16C68 − 6C69 − 4C83 − 16C84 − 6C90  r r r r r r r + 32C11 − 8C12 + 8C22 + 16C23 − 8C34 + 8C63 + m4K 8C5r + 16C6r + 8C10 r r r r + 8C64 − 4C66 − 8C67 + 4C90  r r r + 32C13 − 16C23 + m2K sπ 4C1r + 8C3r + 16C4r + 32C6r + 32C7r + 16C8r + 8C12 r r r r r r r + 8C25 − 8C63 − 8C64 + 8C67 − 8C68 − 4C69 − 4C90  r r r r − 64C13 − 4C22 − 8C23 + m2K tπ −4C1r + 8C3r + 4C4r − 8C5r − 16C6r − 24C12 r r r r r r r r − 12C63 − 8C64 + 6C66 + 12C67 + 2C69 + 4C83 + 4C88 − 10C90  r r r r r r − 32C11 + 16C12 + 32C13 − 4C22 − 8C23 + m2K uπ 4C1r − 8C3r − 4C4r − 8C10 r r r r r r r − 4C63 − 8C64 + 6C66 + 12C67 − 2C69 − 4C83 − 2C90  r r r + 8C68 + 4C69 + sπ2 −8C1r − 32C2r + 8C4r + 4C66  r r r r r r r − 4C67 + 8C68 + 6C69 − 2C88 + 2C90 + sπ tπ 4C1 + 32C2r − 4C4r − 2C66  r r r r − 4C67 + 8C68 + 2C69 + sπ uπ 8C1r + 32C2r − 8C3r − 20C4r + 6C66 r r − 2C88 + 2C90 . for their numerical estimate one has to resort to models. at present. Each Cir is finite and as was explained in Section 5. / Nuclear Physics B 585 (2000) 293–352 343 A.G.

3) while for the G form-factor the contribution is  r r r r r r + 16C11 − 16C13 + 8C15 + 16C17 + 8C22 Gct = m4π −24C4r − 8C6r − 8C8r + 8C10 r r r r r r r r − 8C25 + 8C63 + 8C64 + 8C65 − 4C66 + 12C69 + 8C83 + 12C90  r r r r + 48C11 − 24C12 − 64C13 + m2π m2K −16C4r − 8C5r − 24C6r − 8C8r + 16C10 r r r r r r r + 16C14 + 16C15 − 16C22 + 16C25 + 16C26 − 32C29 + 16C63 r r r r r r + 20C64 + 4C65 + 4C66 + 8C69 + 8C83 + 12C90  r r r r r r − 16C13 − 8C63 − 4C64 − 4C65 + 4C66 + m2π sπ −4C1r + 8C3r + 8C4r − 16C12 r r r − 8C69 + 4C88 − 12C90  r r r r + 4C22 + 16C23 + 4C25 + m2π tπ 4C1r + 8C3r + 12C4r + 8C6r + 8C8r − 8C12 r r r r r r r − 4C63 − 4C64 − 4C65 − 2C66 − 4C67 − 10C69 − 4C83 r r + 4C88 − 10C90  r r r r r − 16C11 + 8C12 + 16C13 − 12C22 + m2π uπ −4C1r − 8C3r + 20C4r − 8C10 r r r r r r r r − 16C23 + 4C25 − 4C63 − 4C64 − 4C65 + 2C66 + 4C67 − 6C69 r r − 4C83 − 6C90  r r r r r r + 32C11 − 40C12 − 64C13 − 8C34 + 8C64 + m4K −8C4r − 8C5r − 16C6r + 8C10 r r + 4C69 + 8C83  r r r r r r r − 32C13 + 8C22 − 8C25 − 8C64 − 4C69 − 8C83 + m2K sπ −4C1r + 8C3r + 8C12 r r + 4C88 − 4C90  r r r r r + 4C22 + 8C23 − 4C63 − 8C64 + m2K tπ 4C1r + 8C3r + 4C4r + 8C5r + 16C6r + 8C12 r r r r − 2C66 − 6C69 − 4C83 − 2C90  r r r r r r − 32C11 + 16C12 + 32C13 − 4C22 − 8C23 + m2K uπ −4C1r − 8C3r + 12C4r − 8C10 r r r r r r r − 8C23 − 4C63 − 8C64 − 2C66 − 2C69 − 4C83 − 2C90  r r + 4C90 + sπ2 −4C88  r r r r r + 6C69 − 2C88 + 2C90 + sπ tπ 4C1 − 4C4r − 2C66  r r r r + 2C69 − 2C88 + 2C90 + sπ uπ −8C3r − 4C4r − 2C66  r r r r r + 2C67 + 2C69 − 2C88 + 2C90 + tπ2 −4C1r + 4C4r + 2C66  r r r r + 2C69 − 2C88 + 2C90 + tπ uπ 4C1r − 8C3r − 12C4r + 2C66  r .4) + u2π 8C3r − 2C67 A. Amorós et al.2. (A. of the previous combinations of O(p6 ) constants.344 G. / Nuclear Physics B 585 (2000) 293–352  r r r r r + tπ2 4C1r − 4C4r − 2C66 − 2C67 − 2C69 − 6C88 + 6C90  r r r r r − 12C67 + 2C69 − 2C88 + 2C90 + tπ uπ −4C1r − 8C3r + 4C4r − 6C66  r . based on resonance saturation as discussed in Section 5. The main drawback of this . + u2π 8C3r − 2C67 (A.1. Resonance contribution The aim of this section is to give an estimate.

5 and 1 GeV.5) F0 MA and for the G form-factor   1 n 2 2 2 mπ mK − m4K − 4cm cd m2π m2π − m2K − tπ GRS = 2 4 4cm F0 MS o − 2tπ cd2 m2π + m2K − tπ n   1 fV fχ −6m4π − 6m2π m2K + 4m2π sπ + 3m2π + m2K (uπ + tπ ) + 2√ 4 F0 2 MV + gV fχ 8m4π + 14m2π m2K − 2m2π tπ − 6m2π uπ + 2m4K + 4m2K sπ  − 2m2K uπ  + gV αV −2m4π + m2π m2K + 3sπ + tπ + uπ + m4K  − m2K (sπ + uπ ) − 2sπ tπ − sπ uπ . / Nuclear Physics B 585 (2000) 293–352 345 method is that it does not specify the scale at which it should be apply.G. (A. Amorós et al. A rough estimate of this uncertainty is obtained varying the scale µ. Using the equations of motion for integrating out the massive fields we get for the F form-factor  1 n 2 −8m4π + 12m2π m2K − 4m4K FRS = 2 4 cm F0 MS  + cm cd 4m4π − 4m2π m2K + 8m2π sπ − 4m2π tπ + cd2 −8m2π sπ + 2m2π tπ + 2m2K tπ + 4sπ2 − 2tπ2 o n  1 4 2 2 √ f + m + 2s + t + 3u f −4m −6m − 2m4K V χ π π π π π K 2 4 F0 2 MV  + m2K (2sπ + 3tπ + uπ ) + + gV fχ 12m4π + 10m2π m2K − 4m2π sπ + 10m2π tπ − 6m2π uπ + 2m4K − 4m2K sπ  − 2m2K uπ  + gV αV 2m4π − m2π m2K + sπ − tπ + uπ − m4K + m2K (sπ + 2tπ + uπ ) + sπ uπ  − 2tπ (sπ + uπ ) √   + 2 gV fV −2m2π tπ − m2K tπ + sπ tπ + sπ uπ + 3tπ2 + tπ uπ 2 √  √  + 2 gV2 5m2π tπ + m2K tπ − sπ tπ − sπ uπ − tπ uπ + 2 fχ2 32m4π + 16m2π m2K √ + 2 fχ αV 16m4π + 8m2π m2K − 8m2π sπ − 12m2π tπ − 4m2π uπ  + 4m2K tπ − 4m2K uπ √ o + 2 fχ2 32m4π + 16m2π m2K √  o 2 s` n + 2 2 fA γA(1) −m2π − m2K + uπ + fA γA(2) 3m2π − m2K − 2sπ + tπ . let’s say between 0. and comparing the results with the scale at the ρ mass.

b¯5. (A.    549221 41π 2 1 169 ¯ 958 ¯ 47 ¯ ¯ − − 5b1 − = b2 − b3 − 418b4 . Amorós et al.3.6 = (16π 2 )2 b5. 16π 2 Fπ2 b¯1.2) For the rest of the notation and the expression for the bi coefficients. − +x 15 5400 18 15 5 35280π 3Fπ4 m4π (B. . (B. / Nuclear Physics B 585 (2000) 293–352 + + √ √ √ 2 gV fV −m2π (sπ + tπ ) − m2K sπ 2 gV2   + sπ2 + sπ tπ + sπ uπ + tπ2 + tπ uπ 2   (2sπ + tπ ) m2π + m2K − sπ tπ − sπ uπ − tπ uπ 2 fχ αV 8m4π + 8m2π m2K − 4m2π (tπ + uπ ) + 8m4K  − 4m2K (2sπ + tπ + uπ ) o √ + 48 2 fχ2 m2π m2K + √  o 2s` n + 2 2 fA γA(1) −m2π − m2K + uπ + fA γA(2) m2π + m2K − tπ .6) the axial-vector contribution vanishes at s` = 0 and thus has no influence on the values quoted in Eq.4.5). (6. we refer to Appendix D in [8]. Scattering lengths Using the definitions of Eq. F0 MA (A.2.1) where x= m2π .3. b20 a31 = b31   4111 37π 2 1 1 17 151 ¯ 653 ¯ 11 + x + + b¯1 + b¯ 2 − b3 − b4 3 4 2 560π Fπ mπ 168 1134 840 14 84 126 126  + b¯5 + b¯6 .12). Appendix B. 3   193π 2 24218 1 ¯ 1 337 ¯ 157 ¯ 87 277 + x − − b1 − b2 + b3 + b¯4 − b22 = 240π 3Fπ4 m2π 840 7560 2835 7 420 126 14  − 3b5 + 5b6 .4 = 16π 2 b1. (A.21) is rather straightforward to evaluate the threshold parameters alI and blI not displayed in [8]:   1849 1583π 2 1 m−2 124 ¯ 179 ¯ 773 ¯ 481 π + x − − − b¯1 − b2 − b3 − b4 = − 3 4 80π Fπ 2520 17010 22680 84 315 378 210  17 − b¯5 + b¯ 6 .6) As can be seen from Eqs.346 G.6 . (6.2.

Amorós et al.  1 1 (6) m4 (47/3888π 2 + 47/648) h0|qq|0i = 4 Fπ (16π)2 π h0|qq|0i(4) = − m2π m2K (113/1944π 2 + 113/324) + m2π m2η (19/432π 2 + 19/72) + m4K (19/486π 2 + 19/81) − m2K m2η (13/216π 2 + 13/36) + m4η (5/216π 2 + 5/36) + 21/8 A¯ m2π 2      + 7/2A¯ m2π A¯ m2K + 7/12A¯ m2π A¯ m2η ¯ 2π ) m2π −24Lr − 12Lr + 112Lr + 68Lr + 10H r + A(m 4 5 6 8 2   2 + 112m2K Lr6 + A¯ m2K   2 − A¯ m2K A¯ m2η /6 + A¯ m2η /72   + A¯ m2K m2π 8Lr5 + 32Lr6 + 8Lr8 + 4H2r + m2K −32Lr4 − 16Lr5 + 128Lr6 + 32Lr8 + A¯ m2η   2  mπ /3 20/3Lr5 − 16Lr6 + 64Lr7 + 12Lr8 − 2H2r   − m2K /3 32/3Lr5 − 112Lr6 + 64Lr7 − 8m2η Lr4   2 + 64m4π Lr4 + Lr5 4Lr6 + 2Lr8 + H2r − 4 Lr6 − 6Lr6 Lr8 − Lr6 H2r − 2 Lr8 2 − Lr8 H2r     + 128m2π m2K Lr4 − Lr6 8Lr6 + 2Lr8 + H2r + 2Lr5 Lr6 . For the non-strange current we find     1 n 3/2A¯ m2π + A¯ m2K + A¯ m2η /6 + 4m2π 4Lr6 + 2Lr8 + H2r 2 Fπ o + 32m2K Lr6 . Vacuum expectation values at O(p6 ) Here we collect the relevant formulas for the vacuum condensate in the limit mu = md . As can be seen they can be cast in terms of the simplest one-loop integral A¯ defined in Appendix D.G. / Nuclear Physics B 585 (2000) 293–352 347 Appendix C.

348 G. Amorós et al. The results for the strange quark reads  2 n ¯ 2  ¯ 2 A mK + A mη /3 + m2π 8Lr6 − 4Lr8 − 2H2r Fπ2 o + 4m2K 4Lr6 + 2Lr8 + H2r  1 1 (6) (π 2 /6 + 1) −7/162m4π + 13/81m2π m2K + 1/36m2π m2η h0|ss|0i = 4 Fπ (16π)2  + 4/81m4K − 2/3m2K m2η + 17/36m4η h0|ss|0i(4) =     2 2 + 2/9 A¯ m2η + 4A¯ m2π A¯ m2K + 4/3A¯ m2π A¯ m2η + 2 A¯ m2K + A¯ m2π  2  mπ −24Lr4 + 88Lr6 − 20Lr8 − 10H2r + 16m2K 4Lr6 + 2Lr8 + H2r + A¯ m2K   2  4mπ 4Lr5 + 12Lr6 − 2Lr8 − H2r + 8m2K −4Lr4 − 4Lr5 + 20Lr6 + 10Lr8 + H2r + A¯ m2η   2  mπ /3 80/3Lr5 + 8Lr6 − 128Lr7 − 60Lr8 + 2H2r   + m2K /3 −128/3Lr5 + 160Lr6 + 128Lr7 + 144Lr8 + 8H2r − 8m2η Lr4 + m4π    2 Lr4 + Lr5 − Lr6 256Lr6 − 128Lr8 − 64H2r + 128 Lr8 + 64Lr8 H2r  + 64m2π m2K 16Lr4 Lr6 + 4Lr5 Lr6 + 2Lr5 Lr8 + Lr5 H2r − 16(Lr6 )2      + 64m4K 4 Lr4 + Lr5 4Lr6 + 2Lr8 + H2r − 4Lr6 4Lr6 + 4Lr8 + H2r   r 2 r r − 4 L8 − 2L8 H2 . / Nuclear Physics B 585 (2000) 293–352 + m4K 1024Lr4Lr6 + 256Lr5 Lr6 − 1024 Lr6 + 2 − 512Lr6Lr8     4 n 4 r r r r r r + 20C20 + 12C21 − C94 + 2C94 mπ 12C19 + m2π m2K 48C21 2 Fπ o r r + 48C21 . (C.1) + m4K 16C20 where h0|qq|0i stands for h0|uu|0i either h0|dd|0i.

An expansion in  leads to the following series Bij    m21 λ0 + A¯ m21 +  A¯  m21 + · · · . two and three point functions. m22 . We have not discussed these integrals in more detail here since we do not present any formulas involving them. Amorós et al. m22 . m22 . m22 . p2 +  B¯ ij m21 . p2 + gµν B22 m21 . All in all we only have to deal with the following set of functions — in the remainder we use d = 4 − 2. B m21 .G. A m21 = 2 16π    1 poleij + B¯ ij m21 . p2 = d 2 2 i (2π) (q − m1 )((q − p)2 − m22 )   = pµ pν B21 m21 . m22 . Z  1 qµ qν qα dd q Bµνα m21 . Z  1 1 dd q 2 . m22 . m22 . The two-loop integrals that appear we discussed in [10] and for the others we used the expressions of [35. p2 . p2 = d 2 2 i (2π) (q − m1 )((q − p)2 − m22 )  = pµ B1 m21 . p2 = 16π 2 ¯ B¯ ij defining finite quantities and where “poleij ” denotes the singular part of each with A. p2 = d 2 2 i (2π) (q − m1 )((q − p)2 − m22 )  = pµ pν pα B31 m21 . p2 + · · · . Through the calculation one has to use one-loop integrals of one. Loop integrals In this appendix we collect for completeness some familiar formulas for the one-loop integrals. p2 = i (2π)d (q 2 − m21 )((q − p)2 − m22 ) Z  1 qµ dd q Bµ m21 . p2 .2) m21 . + 192m4K C19 (C. p2 . / Nuclear Physics B 585 (2000) 293–352 + 349  1 n 4 r r r r + 4C20 + 12C21 + C94 4mπ 12C19 2 Fπ  r r r + C20 − 3C21 − 64m2π m2K 3C19 o r r r + C20 + C21 . .36]. of the Bij functions. (D. m22 . A m1 = d 2 i (2π) q − m21 Z  1 1 dd q . m22 . m22 . m22 . p2  (D. The latter disappears after mass renormalization and the use of some recursion relation. Z  1 qµ qν dd q Bµν m21 .2) Appendix D.1) + (pµ gνα + pν gµα + pα gµν )B32 m21 . m22 .

(D.3) with 1 + ln(4π) + 1 − γ . m22 . 24 (D. p2 2p  − 4B32 m21 . p     1  B21 m21 . / Nuclear Physics B 585 (2000) 293–352 pole = λ0 . p2 = 2 A m22 + m21 B m21 . p2 . m22 . p2 . p2 = − 16π 2 m21 − m22     m21 1 Σ ν (p2 + ν)2 − ∆2 ∆ + 2+ − 2 + ln 2 − 2 ln 2 . 4 3  λ0 2m1 + 4m22 − p2 . Σ = m21 + m22 and ν 2 = [p2 − (m1 + m2 )2 ][p2 − (m1 − m2 )2 ]. 3 pole22 =   λ0 p2 m21 + m22 − . λ0 = B32 m21 .5) one can obtain the expressions of the  terms in the expansion as functions of A¯  and B¯  . p2 = 2(d − 1)   − p2 + m21 − m22 B1 m21 . m22 .350 G. p2 B22 m21 . m22 . m22 .6) (32π 2 ) p ∆ (p − ν)2 − ∆2 m2 p with ∆ = m21 − m22 . the functions defined in Eq. 2p   (d − 2) A m21 . p2 ) through the identities     1  B31 m21 . (D. p     1 A m22 + 2m21 B m21 . m21 . m22 . m22 . p2 = 2 A m22 − m22 − m21 − p2 B21 m21 .2) and (D.5) B m21 . − m2 − m1 − p B22 m1 . m22 . A straightforward calculation leads to . p2 . p2 − dB22 m21 . m22 . m22 . (D. p2 = 2 A m22 − A m21 + m21 − m22 + p2 B m21 . (D. p2 . m22 . m22 . Similarly combining Eqs. 4 λ0 . Then the one-loop contribution is reduced to   m2 A¯ m21 = − 1 2 ln m21 . pole31 = pole1 = λ0 . m2 . (D. m22 .4)  After some simpler algebraic manipulation. p2    m2  m2 1 = 2 − 1 A m21 + 2 A m22 d d 2p    2 2 2 2 2 2 . Amorós et al. m22 . 2 pole32 = pole21 = λ0 .      1  B1 m21 . 0 = 2 2m1 Notice that the inclusion of the previous identities is only done at the final (numerical) level in order to avoid cancellations between different terms occurring in the form-factors.1) can be related to the basic integrals A(m21 ) and B1 (m21 . 16π  1 m21 ln(m21 ) − m22 ln(m22 ) B¯ m21 .

[28] S. hep-ph/9806303. Phys.. Nucl. Ecker. Nucl. Phys. B 479 (1996) 392. Talavera. Riggenbach et al. Rev. Sainio.7) with C = ln(4π) + 1 − γ and m2 = (1 − x)m21 + xm22 − x(1 − x)p2 . Bellucci.A. Chounet. Schilcher. J. Rev. Treiman. Post. Bijnens. Nucl. Bijnens. Makoff et al. Talavera. Phys. [3] L. B 457 (1995) 513.J. [19] G. Phys.. Phys.. A.. Lett. Phys. Burgi. Ametller et al. Gasser. Phys. Lectures given at Les Houches Summer School in Theoretical Physics. B 250 (1985) 465. G. Tausk. Nucl. Lett. Amorós. M. Phys. Nucl.. [26] G. M. Ecker. Bijnens et al. Phys. G. Lett. [34] J. Phys. Lett. Lett. B 337 (1990) 635. Ann. G. [35] A. P. [2] J. hep-ph/9912398. Nucl. Rev. Bijnens. Rev. D 58 (1998) 036004. Gaillard. hep-ph/9707291. G. [17] K. Phys. 168 (1968) 1858. [4] J. D 15 (1977) 574. J. Golowich. Maksymowicz. B 26 (1967) 109. Colangelo. hep-ph/9808421. B 303 (1993) 140. 17 (1966) 336. Rev. A. P. Colangelo. B 436 (1995) 30. Girlanda et al. Berends. [12] U. [10] G. (Erratum) 18 (1967) 1178. Amorós. J. G. Bijnens. Rev. Gasser. B 441 (1998) 437. Nucl. Colangelo. K.. p + dx ln2 (m2 ). C 6 (1999) 297. hep-ph/9302219. Les Houches. Phys. Phys. hep-ph/9710214.B. Gaillard. Weinberg.I. hep-ph/9805389. Rev. hep-ph/9907333. References [1] A. Talavera. hep-ph/9511397. [30] B. in: Probing the Standard Model of Particle Interactions. 16π 2 A¯  m21 = m21 2 2 12 2 Z1  C2 1 π 2  1 2 ¯ 2 2 2 2 2 2 − + + (C − 1)B¯ m1 . G. m2 . J. Phys. Rosselet et al. Bijnens. [15] E. [16] E. France. Nucl. [8] J. Session 68. Bijnens et al. Leutwyler. P. hep-ph/9501318. Phys. 1997.E. Nucl. Ghinculov. B 447 (1995) 373. Donnachie. J. Davydychev. Bijnens. Bijnens.. B 397 (1993) 123. [20] N. G. Phys. / Nuclear Physics B 585 (2000) 293–352 351  2    1 π 2 1 2 2 C + + + ln m1 − C ln m21 . J. Ecker. [5] C. J.. B 427 (1994) 427. Gasser. Phys.B. Lett. Bijnens. 4 (1972) 199. B 374 (1996) 210. Lett.. [24] J. [25] L. D 43 (1991) 127. [22] A. B 409 (1997) 461. Lett. (Erratum) 75 (1995) 2069. Nucl. JHEP 9805 (1998) 014. Rev. Phys. Rev. Colangelo. B 568 (2000) 319. (NY) 280 (2000) 100. Knecht et al. [6] J. Phys.C. JHEP 02 (1999) 020. Phys. [32] J. Phys. Lett. p = 2 2 12 2 0 (D. [14] P. Talavera. [7] J. Phys. G. hep-ph/9703448. [23] F. M.G. Eur. Phys. [33] A. Pich. B 508 (1997) 263. hep-ph/9602429. Phys. Amorós et al. 79 (1997) 4088. D 53 (1996) 2573. Rev. [29] J. Lett. J. Phys. Cabibbo. Pais. C 3 (1998) 1. hep-ph/9610269. 171 (1968) 1457. hep-ph/9403390. Caso et al. Oades. Eur. Rev. B 423 (1994) 80. 16π B m1 . [21] Ll. Phys. B 480 (2000) 71. Kambor. Nucl. C. [13] J. [18] J. 28 July– 5 September. hep-ph/9507319. B 489 (1997) 387. S. Rep. P. Colangelo. Rev. Lett. hep-ph/9401206. hep-ph/9405418. Sainio. G. Rev. [31] L. P. J. [11] S. Maltman. Phys. J. Phys. Nucl. B 137 (1965) 438. D 54 (1996) 1125. hep-ph/9803251. Phys. m2 .K. [27] Particle Data Group. J.E. Bijnens. Ananthanarayan. Phys. Phys. Kambor. Golowich. Phys. 70 (1993) 1591. Buttiker. H. Phys. . Phys. Phys. hep-ph/9601285. Gasser. van der Bij. [9] M. hep-ph/9907264. hep-ph/9902437.

de Rafael. B 72 (1974) 461. hep-ph/9701400. L. JHEP 0001 (2000) 041. hep-ph/9707341. Nucl. B 490 (1997) 239. Ecker et al. hep-ph/9910537. Espriu. Eur. J. JHEP 9805 (1998) 011. V. Nucl. Ghinculov. Nucl. de Rafael. M. Bijnens.F. Girlanda. B 126 (1977) 109. A. Stern. Holstein. O. Phys. Golowich. Dumbrajs et al. . E. N. Nucl. Yao. Phys. Phys. Ecker et al. Phys. J. hep-ph/9302217. B 216 (1983) 277. Zakharov. Rev. W. G. Phys. Phys. B 419 (1998) 326. Z. B 516 (1998) 385.A.. Phys.R. B 345 (1990) 22.352 [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] [61] [62] G. M. Phys. Lett. Phys. Phys. C 14 (2000) 111. Lett. Moussallam. hep-ph/0001188.I. P. Phys. hep-ph/9412343. J. G. Nucl.. B. J. J. Phys. de Rafael. Eur. G. B 160 (1979) 57. C. D 53 (1996) 2651. Nucl. hep-ph/9805442. hep-ph/9702266. B 223 (1989) 425. E. / Nuclear Physics B 585 (2000) 293–352 A. Lett. Peris. Bijnens. Knecht.. Vainshtein.. Phys. Phys. J. Phys. G 25 (1999) 1607.F. S. B 390 (1993) 501. J. M. Rev. B 504 (1997) 381.A. S. Amorós. Prades. Perrottet. Phys. Lett. Phys. E. B 306 (1993) 343. J. hep-ph/9902446. Fuchs. hep-ph/9710268. Rev. Phys. Hoogland et al. hep-ph/9206236. Rev. Phys. Phys. Peris. Shifman. Nucl. Veneziano. Lett. (Erratum) C 11 (1999) 571. J. Rev. hep-ph/9610360. S. Lett. hep-ph/9902463. Phys. ’t Hooft. Wyler. hep-ph/9908252. B 321 (1989) 311. hep-ph/9711256. Taron. S. Herrera-Siklody et al. Kambor.H. C 63 (1994) 491. Bruno. Prades. Descotes. M. hep-ph/9909292. Stern. Bijnens. Nucl. Golterman. J. B 117 (1976) 519. Nucl. Peris. hep-ph/9509304. B. B 147 (1979) 385. D 47 (1993) 2089. Phys. Herrera-Siklody. D 61 (2000) 034018. Phys. Y. Nucl. Phys. G. J. V. J. P. Witten. E. G. D 62 (2000) 033003. J. Donoghue. Phys.. Nucl. E. E.I. D. Novikov et al. 79 (1997) 4092. Nucl.. D. B 421 (1998) 319. B 191 (1981) 301. Moussallam. Phys. hep-ph/9302246. de Rafael. Phys. J. Nucl. B. Bijnens. B 348 (1995) 539. Amorós et al.

12.20.2 . 3. a new lab for such investigations is a doubly heavy long-lived quarkonium Bc recently observed by the CDF collaboration [1] for the first time. Then we can provide the study of heavy-quarks dynamics by testing the various conditions. A.3 2 1 3 ( 0 0 ) 0 0 3 8 6 . 142284. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.ihep.  2000 Elsevier Science B. Protvino. so that the NRQCD approach [2.He. the form factors for the semileptonic decays of Bc+ → Bs (Bs∗ )l + νl are calculated with account for the Coulomb-like αs /v-corrections in the heavy quarkonium. since it is composed by two nonrelativistic heavy quarks.20. 13. Moscow region. This meson is similar to the charmonium and bottomonium in the spectroscopy. E-mail addresses: kiselev@th1. NRQCD. Moscow Region.Nuclear Physics B 585 (2000) 353–382 www. Likhoded a a Russian State Research Center “Institute for High Energy Physics”. The nonleptonic decays are studied using the assumption on the factorization.3] ¯ levels is well justified to the system.39. Kovalsky b . Decays. Kiselev a . 11. So. All rights reserved. Introduction For better understanding and precise measuring the weak-action properties of heavy quarks governed by the QCD forces. The generalized relations due to the spin symmetry of HQET/NRQCD for the form factors are derived at the recoil momentum close to zero.su (A. Russia b Moscow Institute of Physics and Technology. [4–8] in the framework of potential models and lattice simulations.V.elsevier.V. Russia Received 2 March 2000.38.-k Keywords: QCD sum rules.-t.K.55.su (V. PACS: 12. A. Kiselev). Dolgoprudny.V. Likhoded). accepted 20 June 2000 Abstract In the framework of three-point QCD sum rules. determining the forming of bound states as well as the entering of strong interactions into the weak processes.ihep.Hg. 13. Heavy quarks.V. likhoded@mx.E. PII: S 0 5 5 0 . The modern predictions for the mass spectra of bc were obtained in Refs.Fc. The Bc -meson lifetime is estimated by summing up the dominating exclusive modes in the c → s transition combining the current calculations with the previous analysis of b → c decays in the sum rules of QCD and NRQCD. we need as wide as possible collection of snapshots with hadrons containing the heavy quarks.Hx. revised 16 May 2000.nl/locate/npe Bc decays and lifetime in QCD sum rules V. All rights reserved. HQET 1.25.K.

However. In hadron collisions. where the initial and recoil mesons are both heavy quarkonia.16 ± 0. v  1.354 V.40 ± 0. where the form factors are about 3 times less than the values expected in the potential-quark models. This can be calculated in the framework of factorization approach by a careful evaluation of complete set of diagrams in the given αs -order. The measured Bc lifetime τ [Bc ] = 0. [9–23]. in agreement with the theoretical expectations. The most simple picture takes place for the production in the e+ e− -annihilation.V. The authors of [33. The production mechanism for the Bc meson was studied in Refs. as the heavy quarks move nonrelativistically. gluons and leptons due to QCD and QED. where v1. the feature of Bc mesons is an absence of annihilation into light quarks. and at pT ∼ mBc the subleading terms in 1/pT or higher twists have to be taken into account. the calculations of Bc decays in the framework of QCD sum rules were performed in [33–36]. The measured value of Bc mass yet has a large uncertainty MBc = 6. In the strict limit of v1 = v2 .18 −0. The reason for a such disagreement was pointed out in [35] and studied in [36]: in the QCD sum rules for the heavy-quarkonia the Coulomb-like corrections are significant. which basically determine the theoretical uncertainties in OPE. The accurate measurement of Bc lifetime could allow one to distinguish various parameter dependencies such as the optimal heavy-quark masses. The Coulomb rescaling of quark–quarkonium vertex enhances the estimates of form factors in the QCD sum rules for the Bc+ → ψ(ηc )l + ν decays.13 GeV. which can be reliably described in the potential models. / Nuclear Physics B 585 (2000) 353–382 The arrangement of excitations is close to what was observed in the charmonium and bottomonium. P. where the universal perturbative fragmentation functions can be analytically calculated for the S-. In the framework of NRQCD at the recoil momentum close to zero one derives the spin-symmetry relations for the form factors of semileptonic Bc decays [36. . At present.PM [Bc ] = 0.03 ps agrees with the estimates obtained in the framework of both the OPE combined with the NRQCD evaluation of hadronic matrix elements [24–26] and potential-quark models.2 denote the four-velocities of initial and recoil mesons. Kiselev et al.46+0. the fragmentation regime takes place at the transverse momenta pT  mBc .and Dwave levels in the framework of factorization for the hard production of quarks and their soft binding into the hadron. since they correspond to summing up the ladder diagrams.37]. O(αs4 ). and the semileptonic and hadronic widths of Bc are one order of magnitude less than those in OPE. where one has to sum up the dominating exclusive modes to calculate the total Bc width [27–32] τOPE. where αs /v is not a small parameter.39 ± 0.55 ± 0. that implies the higher excitations decay into the low lying levels and ground state due to the emission of photons and pion pairs.34] obtained the results.15 ps. The nonfragmentational contributions dominate at pT ∼ mBc [9–23].

In the present paper we calculate the Bc decays due to the c → s weak transition in the framework of QCD sum rules. / Nuclear Physics B 585 (2000) 353–382 355 respectively. We demonstrate the significant role of the product term for the convergency of method and reevaluate the constants fB as well as fBs . and the generalized spin-symmetry relations were obtained for the Bc → ψ(ηc ) transitions: four equations. including that of [37].V.V. . The relations between the form factors of semileptonic decays as follows from the spin symmetry of HQET and NRQCD are 1 In Refs. the uncertainty in the values of heavy-quark masses is much less than in OPE. Summing up the dominating exclusive modes. Kiselev et al. We perform the numerical estimates of semileptonic Bc widths and use the factorization approach [41–43] to evaluate the hadronic modes. We stress that in the QCD sum rules to the given order in αs . In the semileptonic decays the hadronic final state is saturated by the pseudoscalar Bs and vector Bs∗ mesons. Section 2 is devoted to the general formulation of three-point sum rules for the Bc decays with account of Coulomb-like corrections. that shows a low contribution of next-to-leading 1/mQ -terms. taking into account the Coulomb-like αs /v-corrections for the heavy quarkonium in the initial state. which agree with the experimental data and the predictions of OPE and quark models. The relations are in a good agreement with the sumrules calculations up to the accuracy better than 10%. [38. where we also show the convergency of three-point sum rules with respect to a dependence on the threshold energy of continuum in the heavy–light system. we observe the stability of form factors versus Ec . The spin symmetries of leading terms in the Lagrangians of HQET [40] for the singly (∗) heavy hadrons (here Bs ) and NRQCD [2. 1 In [36] the soft limit v1 · v2 → 1 at v1 6= v2 was considered. We estimate (∗) the form factors of semileptonic Bc → Bs decays. the gluon condensate term was calculated in both QCD and NRQCD. Taking into account the dependence on the threshold energy ¯ system in both the value of fBs extracted from the twoEc of hadronic continuum in the bs point sum rules and the form factors in the three-point sum rules. This fact leads to a more definite prediction on the Bc lifetime. so that we need the values of their leptonic constants entering the sum rules and determining the normalization of form factors. [37]. we reanalyze the two-point sum rules for the B mesons to take into account the product of quark and gluon condensates in addition to the previous consideration of terms with the quark and mixed condensates. For this purpose.39] the relations were studied in the framework of potential models.3] for the doubly heavy mesons (here Bc ) result (∗) in the relations between the form factors of semileptonic Bc → Bs decays. We discuss the preferable prescription for the normalization point of nonleptonic weak Lagrangian for the charmed quark and present our optimal estimate of total Bc width. so that it enforced a convergency of the method. The paper is organized as follows. the authors of [37] found a single relation between the form factors. Moreover. which indicates the convergency of sum rules. We derive two generalized relations in the soft limit v1 · v2 → 1: one equation in addition to what was found previously in Ref. The analysis of two-point sum rules for the leptonic constant of singly heavy meson with the introduction of term allowing for the product of quark and gluon condensates is presented in Section 3. we calculate the lifetime of Bc .

In conclusion we summarize the results.356 V. From the two-point sum rules we extract the values for the leptonic constants of mesons in the initial and final states. / Nuclear Physics B 585 (2000) 353–382 derived in Section 4 in the soft limit of zero recoil. Kiselev et al. In our consideration we use the following notations: . 2. Three-point sum rules for the Bc meson In this paper the approach of three-point QCD sum rules [44–46] is used to study the form factors of semileptonic and nonleptonic decay rates for the c → s transition in decays of Bc meson.V. We discuss the optimal estimation of lifetimes for the heavy hadrons in Section 6. Section 5 contains the description how the nonleptonic decays modes are calculated and the Bc lifetime is evaluated.

fP MP2 . and (1) 0|q¯1iγ5 q2 .

P (p) = m1 + m2 .

(2) 0|q¯1iγµ q2 .

The hadronic matrix elements for the semileptonic c → s transition in the Bc decays can be written down as follows: . ) = iµ MV fV .V (p. where P and V represent the scalar and vector mesons. m2 are the quark masses. and m1 .

.

(3) Bs (p2 ).

Vµ .

.Bc (p1 ) = f+ (p1 + p2 )µ + f− qµ .

.

1 ∗ (4) B (p2 ).

Vµ .

i s .Bc (p1 ) = iFV µναβ  ∗ν (p1 + p2 )α q β .

.

1 ∗ Bs (p2 ).

Aµ .

FV . q 2 = i2 dx dy ei(p2 ·x−p1 ·y) . It should be noted that since the ¯ µ (1 + γ5 )νl is transversal in the limit of massless leptons. Following the standard procedure for the evaluation of form factors in the framework of QCD sum rules [47–49]. p2 . The form factors f± .Bc (p1 ) = F0A µ∗ + F+A ( ∗ · p1 )(p1 + p2 )µ + F−A ( ∗ · p1 )qµ . we consider the three-point functions: Z  Πµ p1 . F0A and F±A are functions of q 2 only. the leptonic current lµ = lγ probabilities of semileptonic decays are independent of f− and F−A (the τ + ντ mode is forbidden by the energy conservation). (5) i where qµ = (p1 − p2 )µ and  µ =  µ (p2 ) is the polarization vector of Bs∗ meson. Vµ and Aµ are the flavour-changing vector and axial electroweak currents.

 .

.

¯ × 0.

Z  V .A p1 . Vµ (0). p2 . b(y)γ (6) 5 c(y) 0 .T q¯1 (x)γ5q2 (x). q 2 = i2 dx dy ei(p2 ·x−p1 ·y) Πµν .

 .

.

¯ × 0.

. where q¯1 (x)γ5 q2 (x) and q¯1 (x)γν q2 (x) denote interpolating currents for Bs and Bs∗ .A (0).A are the currents Vµ and Aµ of relevance to the various cases. b(y)γ (7) 5 c(y) 0 .T q¯1 (x)γµ q2 (x). correspondingly. JνV . JµV .

µ p1. Πi = Πi (p12 . q = − Πi (2π)2 (s1 − p12 )(s2 − p22 ) where Q2 = −q 2 > 0. The leading QCD term is a triangle quark-loop diagram in Fig. ΠV . F0A and Π0A and Π±A = 12 (Π1A ± Π2A ). Q2 ) 1 pert 2 2 2 ds1 ds2 + subtractions. (13) Πi p1 . Now let us proceed with the physical part of three-point sum rules. In (8)–(10) the scalar amplitudes Πi are the functions of kinematical invariants.ν F±A are determined from the + Π4A p1. s2 . derived from a double dispersion relation at q 2 6 0. 1. Q2 ) are given in Appendix A. = Π0A gµν + Π1A p2.ν (8) (9) + Π2A p1. so that Fig. The triangle diagram. The form factors f± . / Nuclear Physics B 585 (2000) 353–382 357 The Lorentz structures in the correlators can be written down as Πµ = Π+ (p1 + p2 )µ + Π− qµ . Q2 ρi    (14) + θ s1 − s1th · θ s2 − s2th · ρicont s1 . i. Q2 .ν .µ p2. where the resonance term is expressed through the product of leptonic constant and form factor for the transition under consideration. . respectively.ν + Π3A p2.V. s1 . x2 .e. x3 ) = (x1 + x2 − x3 )2 − 4x1 x2 . q = − (2π)2 (s1 − p12 )(s2 − p22 ) Assuming that the dispersion relation (13) is well convergent. p22 . Q2 ) 1 2 2 2 ds1 ds2 + subtractions. s2 . s2 . s2 . s2 .µ p1. V Πµν A Πµν β = iΠV µναβ p2α p1 . p2 . The integration region in (11) is determined by the condition −1 < 2s1 s2 + (s1 + s2 − q 2 )(m2b − m2c − s1 ) λ1/2 (s1 . giving the leading perturbative term in the OPE expansion of three-point function. 1. (11) p1 . q 2 )λ1/2 (m2c . p2 . m2b ) < 1. The expressions for spectral densities pert ρi (s1 . the physical spectral functions are generally saturated by the ground hadronic states and a continuum starting at some effective thresholds s1th and s2th :   phys s1 .V. for which we can write down the double dispersion representation at q 2 6 0 Z pert  ρi (s1 . (12) where λ(x1 .µ p2. The connection to hadrons in the framework of QCD sum rules is obtained by matching the resulting QCD expressions of current correlators with the spectral representation. FV . s2 . q 2 ). (10) amplitudes Π± . s2 . Q2 = ρires s1 . Kiselev et al. Z phys  ρi (s1 .

Kiselev et al. The ladder diagram of the Coulomb-like interaction.358 V. 2.V. . / Nuclear Physics B 585 (2000) 353–382 Fig.

.

.

 .

¯ 5 (γµ )s .

Bs (Bs∗ ) Bs (Bs∗ ).

Fi (Q2 ).

s2 . Q2 = 0.Bc hBc |bγ ¯ 5 c|0i ρires s1 .

50–55]. an essential role is taken by the Coulomb-like αs /v-corrections. Then. s . In our case it leads to the finite renormalization for ρi [36]. which means the formulation of sum rules. shown in Fig.e.bγ   × (2π)2 δ s1 − M12 δ s2 − M22 + higher state contributions. 2. i. so that ρic = Cρi .(15) where M1. It is well known that an account for this corrections in two-point sum rules numerically leads to a double–triple multiplication of Born value of spectral density [44–46. They are caused by the ladder diagram. ¯ where the relative velocity of quark movement is small. the expressions for the form factors Fi can be derived by equating the representations for the three-point functions Πi in (11) and (13). For the heavy quarkonium bc.2 denote the masses of hadrons in the initial and final states. The continuum of higher states is modelled by the perturbative absorptive part of Πi . by ρi .

C .

   .

Ψ (0).

C = . 4παs −1 4παs ¯ bc 1 − exp − .

free .

= .

Ψ (0).

which corresponds to fBc = 400 MeV [50–55]. In our calculations ¯ bc Ec = 1. The requirement of stability in the sum rules including the . like corrections give αsc (bc) The quark masses are fixed by the calculations of leptonic constants fΨ and fΥ in the same order over αs . v= 1− 2 p1 − (mb − mc )2 (18) To the moment. Numerical results on the form factors and the semileptonic decay widths We evaluate the form factors in the scheme of spectral density moments. the procedure of calculations is completely described.2 GeV. 3v 3v (16) (17) ¯ bc ¯ where v is the relative velocity of quarks in the bc-system: s 4mb mc . This scheme ¯ is not strongly sensitive to the value of the bc-system threshold energy. 3.45. The two-point sum rules for the Bc meson with account for the Coulomb¯ = 0.

23 × GeV)3 . Using the Padé approximation. we find that higher-order corrections can be about 30%. Kiselev et al. For this factor we have the following expression [59. We expect it is large.60] have the following form: ¯ fB2 MB e−Λ(µ)τ 3 = K 2 C(µ) π   m2 τ 2 π 2 τ 4 D αs 2 E G . so that K can be put as the overall factor). The leptonic constant for the Bs meson is extracted from the two-point sum rules.27 GeV)3 (this variation corresponds to the renormalization group evolution and the insertion of αs -corrections to this term.25 GeV. we can extract the above values of αsc (bc) S (Q1 Q2 )-quarkonia constants obey the scaling relation [32. we hold the K-factor in conservative limits 1. h απs G2 i = 1. αsc (bb) exchange αs . ¯ and fBc .V.77 × 10−2 GeV4 as the central values. Note. depending on the quarkonium contents. The contribution of quark condensate term is not sensitive to the variation of hqqi ¯ in the limits from −(0.37. that the heavy In this way.8 GeV2 .03 GeV and mb = 4. m20 = 0.28 GeV.5 GeV.4 ÷ 1. The K-factor is not sensitive to Ec changing in the range 1. and MB = 5.60]: K2 = E τ  Zc  0 −1 E τ    Zc  ˜ 13 2π 2 2αs (Λ) + − ln z dz. which obey the renormalization group ¯ ' 0. The Borel improved sum rules for the B-meson leptonic constant [59.60. The K-factor is due to αs corrections. but we suppose the appearance of the same factor in evaluating the αs corrections to the heavy–light vertex in the triangle diagram. Mn 4m1 m2 n where n denotes the radial excitation number of nS-level.02 GeV. The values of leptonic constants fΨ . and c is independent of heavyquark flavors. 3. which are in a good agreement with the recent estimates in [56–58]. z2 e−z dz · z2 e−z 1 +  π 6 9 (21) 0 where we suppose that the scale Λ˜ is equal to 1.50–55]   fn2 Mn (m1 + m2 ) 2 c (19) = .23 GeV)3 to −(0. It is quite reasonable to suppose its cancellation in evaluating the semileptonic form factors due to the renormalization of heavy–light vertex in the triangle diagram. . fΥ linearly depend on the Coulomb ¯ ' 0. dω ω2 e−ωτ + hqqi ¯ 1− 0 + 16 288 π ωZ0 (µ) 2 0 (20) where we use hqqi ¯ = −(0. results in quite an accurate determination of masses mc = 1. So.40 ± 0. where the quark masses free off a renormalon ambiguity were introduced. The dependence of K-factor on the Borel parameter τ and the threshold energy Ec is shown in Fig. / Nuclear Physics B 585 (2000) 353–382 359 contributions of higher excitations. We find αsc (cc) evolution with the appropriate scale prescription.V.60 ± 0. We see that NLO corrections to the leptonic constant are about 40%.0 ÷ 1.7.

63 GeV.1 ÷ 1. whereas the contribution of condensate is numerically suppressed. The results of general Borel scheme calculations of fB MB ignoring the overall K-factor are presented in Fig. as expected from the √ semilocal duality. The K-factor dependence on the Borel parameter τ and the threshold energy Ec .5 GeV−1 . that the sum rules in (20) with the redefined ωren and Λ¯ ren . 4. we find that sum rules in Eq. We write down:   m2 τ 2 π 2 τm4 D αs 2 E 3 ¯ G . Demanding a low deviation of Λ¯ from Λ¯ ren = where Ec is already not equal to ω0.360 V. ω0.V. ¯ In the limit of semilocal duality [61–71] τ → 0 we get the relation: Λ(µ) = 34 ω0 (µ) (the contribution of the quark condensate term to this equation is about ∼ 15%).84 GeV.60]. Then the optimal value of Borel parameter τ = τm ' 6.dual . We expect. / Nuclear Physics B 585 (2000) 353–382 Fig. We observe two stability regions. which are in a good agreement with the semilocal duality if Ec = 1. + hqqi ¯ 1− 16 288 π (23) 0 ren . (22) ω0. ¯ 1− 0 m + (24) fB2 MB e−Λτm = K 2 2 REc3 + hqqi π 16 288 π √ 3/2 where R denotes the average value of e−ωτm .3 GeV (see Fig.dual ren we have Λ ¯ ren = MB − mb = 0. (23) can lead to the results. We introduce the renormalization invariant quantities 3 ren ren Λ¯ dual = ω0. 3. The stability region at τ = 2 ÷ 4 corresponds to that of considered in [59. 4 ren = C −1/3 (µ) ω0 (µ). and we obtain that in the semilocal For Λ¯ dual dual ren = 0. 4). in the general Borel scheme for fB we have to consider the stability at τ 6= 0 with the extended region of resonance contribution. Neglecting the quark condensate term in duality the threshold energy ω0. So. Kiselev et al. we find the Ec -dependence of fB MB . have a stability point at τ ∼ 1¯ : fB2 MB = K 2 Λ ¯ fB2 MB e−Λτ 3 =K 2 π ZEc 2 2 −ωτ dω ω e   m20 τ 2 π 2 τ 4 D αs 2 E + G .dual 0.dual 2 π Since in the three-point sum rules we use the scheme of moments and search for a stable region.dual the leptonic constant we have 3 3 ren . as mentioned.63 GeV. The results for the .

4 GeV. Note that the intermediate stability point τ ∼ 4 exhibit a low variation of Λ¯ close to 0. fB in the semilocal duality sum rules (pale curves) and in the general Borel scheme. where the optimal values of Λ¯ agree with the semilocal duality and the estimate Λ¯ ren = MB − mb .72]. The similar situation has been observed in the NRQCD sum rules for doubly heavy baryons [72]. the intermediate stability point was observed only. (24) from Fig.60. say. The calculated physical quantity should be independent of parameters in the sum-rule scheme. Kiselev et al. However.V. for instance.60] is about 1. The second region appears only when we introduce the term with the product of quark and gluon condensates. 5. dashed lines correspond to the accounting for the condensates contributions. 5. we truncate both the operator product expansion and the perturbative series for the Wilson coefficients by fixed orders in the operator dimension and αs . 4. Solid lines correspond to the sum rules without condensate terms. √ Fig. / Nuclear Physics B 585 (2000) 353–382 361 Fig.V. 4. respectively. 4). The product of quark and gluon condensates was not taken into account in [59. the . as shown in Fig. we can invert the sum rules to study the dependence of Λ¯ on τ . The dependence of Λ¯ on τ with the fixed values of fB MB . Fixing the optimal values of fB2 MB in Eq. [73]). and usually given by the potential models (see. This fact leads to that the results depend on the scheme parameters. and therefore. which correspond to the stability regions at τ = 6 ÷ 7 GeV−1 (the notations are the same as in Fig.60]. which was obtained in [59. leptonic constant fB obtained from this region [59.5 greater than the value obtained from the stability region at τ = 6 ÷ 7.

54 GeV. Moreover. the physical part of sum rules is modelled by the contributions of resonances and a continuum term starting at a threshold. we study the general Borel sum rules in the scheme. fB in the semilocal duality sum rules (dashed curve) and in the general Borel scheme (solid ¯ which improves the stability of result obtained in the semilocal line) with the corrected value of Λ. . Second. we expect that the estimate of leptonic constant should result in the same value obtained in the semilocal duality. which is not important for the current discussion. at first. In order to clarify this statement we present the results of semilocal sum rules for the leptonic constant of B meson in Fig. which requires the stability at τ → 0 and corresponds to the duality region containing the ground state. Then. that the value of leptonic constant is the same as we have found in the perturbative limit of semilocal duality. Let us.e. where we do not require the stability at τ → 0. Then. / Nuclear Physics B 585 (2000) 353–382 Borel variable. we present also the result for the other ratio of condensate values (the lower value of mixed condensate and the upper value of gluon condensate in the regions mentioned below) in Fig. The stability of semilocal sum rules can be improved by a variation of Λ¯ and ω0 . Kiselev et al. Certainly. We see that this can be reached at the same values of condensates involved as well as in the same region of τ . i. 6. duality. and the leading correction at τ → 0 is the term with the quark condensate. where we put Λ¯ = 0. consider the results on the leptonic constant in the limit of semilocal duality. The central values of condensates as mentioned above correspond to the results shown in Fig. the extension of duality Fig. Then. We see that the semilocal duality is completely broken at inappropriate choice of condensate values.V. 7. in the second point of local extremum in Fig. In order to confirm. 4. that introduces the dependence on the threshold value and suggests that the stability of results can be improved by the terms of excited resonances in addition to the ground state. 4. Note. which is again given by the second extremum (see Figs. we state that the value of leptonic constant obtained at τ → 0 agrees with the value corresponding to the stability at τ ≈ 7 GeV−1 . we expect that the region of stability will extend at τ > 0 if we will add the higher condensates with the appropriate ratios of their values. which means that we extend the duality region to incorporate possible excitation contributions. 6. The sum rules show that the condensate contributions are given by the polynomials over τ . only.362 V. 4 and 6).

The criterion on the convergency of both the semilocal duality and general Borel sum rules was ignored in [59. The physical meaning of gluon condensate as it stands in the operator product expansion taken between the observed states is independent of the scheme of calculations. 7. 5 and 8. that indicates the divergency of the method.60]. / Nuclear Physics B 585 (2000) 353–382 363 Fig. . 8. which is caused by additional terms given by the excitations beyond the control of the method.V. we can achieve the better stability in the general Borel scheme by adjusting the condensate values. Of course. 4 and 6. region leads to a failure of stability at low τ . corresponding to the inverted sum rules for the leptonic constant given in Figs. which correspond to the stability regions at τ = 6 ÷ 7 GeV−1 in Fig. Kiselev et al. fB in the semilocal duality sum rules (dashed curve) and in the general Borel scheme (solid line) with the condensate values which destroy the semilocal duality. In the √ Fig. but this will destroy the semilocal duality. present the results in Figs. 6. we The same notes can be done in the discussion on Λ. where the greater value of leptonic constant was obtained (see Fig. The dependence of Λ¯ on τ with the fixed values of fB MB . while the convergency demands an appropriate choice of condensate values as we put. At higher values of τ the stability would be reached by introduction of higher condensates contributing as the polynomials of higher powers. ¯ For the sake of comparison.V. 7).

In Table 1 we present the results of sum rules for the form factors in comparison with estimates in the framework of potential models [32.5 ÷ 2) × 10−2 GeV4 . which is in a good agreement with the recent lattice results [74] and the estimates in the QCD SR by other authors [75]. So.4 GeV. we can conclude that the 1/mb corrections are not valuable for fB . Thus. the semilocal duality means the duality for the region containing the ground state only. For the sake of completeness the quark-model expressions for the form factors are given in Appendix B.V. when the excitations are suppressed. i. In [33] the form factors were derived using the similar SR technique but without the ¯ system. results in the uncertainty less than 2%.11 (see [75. 12 and 13. As for the mixed condensate we have used the range m20 = (0. mc = 1. For the vector-B ∗ -meson constant fB ∗ we put fB ∗ /fB = 1. However. . The variation of these parameters does not change the qualitative picture for the lepton constant as it has been discussed. while the usual Borel (or moment) scheme explore the extended region containing several hadronic states: the ground state and its excitations. In Figs. by the K-factor we find the value fB = 140 ÷ 170 MeV. which agrees with the various estimates [77]. and we use ms = 0.2 GeV. Kiselev et al.364 V.16. The uncertainty of estimates is basically connected with the higher-orders in αs . Remember. In our consideration the quark masses are equal to mb = 4. So. its value is the same for both schemes used. which expresses the flavor SU(3)-symmetry violation for B mesons [71]. The definition of sum-rules scheme includes the parameter determining the threshold energy of continuum contribution in addition to the resonance term. in sum rules the heavy-quark masses are fixed by the two-point sum rules for bottomonia and charmonia with the precision of 20 MeV.78]. multiplying the result taken from Fig.72 ÷ 0.35 ÷ 1. ¯ threshold energy of We have investigated the dependence of form factors on the bs continuum in the range Ec = 1.88) GeV2 .45 GeV. For the leptonic constant of Bs meson we explore the following relation fBs /fB = 1. the gluon condensate essentially contributes at the scheme parameters. We see a good agreement of estimates in the QCD sum rules with the values in the quark model. Numerically.3 GeV.6 GeV. The characteristic forms of this dependence ¯ system threshold are shown in Figs. the condensate essentially determines the binding energy of quarks in the hadron containing the heavy quark. while the ground state corresponding to the current under consideration dominates.e. So. as we have found. 4. We see that the optimal choice for the bs energy is 1.14 ÷ 0. the difference between the schemes of semilocal duality and usual Borel representation is due to the variation of duality region. where the excited states are suppressed. The value of gluon condensate has been varied in the range h απs G2 i = (1. / Nuclear Physics B 585 (2000) 353–382 sum rules the gluon condensate contributes in the region. So.1 ÷ 1. for the semilocal duality and in the Borel scheme. The uncertainties in the values of form factors are basically determined by the variation of b-quark mass.16 GeV and mc = 1. while changing the other-quark masses in the ranges ms = 0. which enhance the form factors about three Coulomb-like corrections in the bc times. while the numerical uncertainty of its value is less than 7%. 9–11 we present the results in the scheme of spectral density moments.76]).15 GeV.

Kiselev et al. f+ in the scheme of the spectral density moments. n is the number of moment with respect ¯ channel. momentum in the bs Fig.V.2 GeV. n is the number of moment with respect ¯ channel. / Nuclear Physics B 585 (2000) 353–382 365 Fig. n is the number of moment with respect ¯ channel.V. of momentum in the bs . F0A in the scheme of the spectral density moments.2 GeV. F0A in the scheme of the spectral density moments. 11. m is the number of moment with respect to the square to the square of momentum in the bc ¯ channel at Ec = 1. 9. 10. The number of moment with respect to the square of to the square of momentum in the bc ¯ channel is equal to 1 at Ec = 1. momentum in the bs Fig. The number of moment with respect to the square of to the square of momentum in the bc ¯ channel is equal to 1 at Ec = 1.2 GeV.

The corresponding accuracy is about 10%. Indeed. ¯ threshold energy Ec .3 −5. So.366 V. Kiselev et al.2 0.38] with similar choices of parameters result in the form-factor values. Then.1 0. we expect that the potential models give good reference points for the appropriate numerical values.9 1. The dependence on the choice of threshold energy in the bs .1 8. determining the region of resonance Fig. the significant αs correction to the leptonic constant of Bs meson should cancel the same factor for the renormalization of quark–meson vertex in the ¯ channel can triangle diagram. 12.V. the potential models [27–31.1 8. The accuracy of sum rules under consideration is basically determined by the variation of heavy-quark masses. which are slightly model-dependent. determining the region of resonance Fig.2 1. 13.3 1.8 Potential model [32] 1. / Nuclear Physics B 585 (2000) 353–382 ¯ threshold energy Ec .1 −5. Table 1 The form factors of Bc decay modes into the Bs and Bs∗ mesons at q 2 = 0 Method f+ f− FV (GeV−1 ) F0A (GeV) A (GeV−1 ) F+ A (GeV−1 ) F− This paper 1. The dependence of f+ on the bs contribution.8 1.4 Let us discuss the uncertainties in the sum-rules and other approaches. The dependence of F0A on the bs contribution.

/ Nuclear Physics B 585 (2000) 353–382 367 ¯ channel be optimized and. We have calculated the form factors in the SR consistent with the values in the OPE and potential models. Of course. (25) 2 can be evaluated in the framework of sum so that the first derivative f 0 (0) = f (0)/Mpole rules to estimate the “pole” mass Mpole .25 GeV. and Mpole = 1. That is why we calculate the form factors as well as their Borel transforms at q 2 < 0. we have found Mpole = 1. MBs = 5. Γ (Bs∗ e+ νe ) = 7.V.9 GeV for the decay form factors with the Bs modes.36]. The semileptonic widths are presented in Table 2. Numerically. However.8 ÷ 1.37 GeV. which gives the same uncertainty for the form factors.7 × 10−14 GeV. These results agree with the values obtained in the framework of covariant quark model [32]: Γ (Bs e+ νe ) = 4. as we could expect looking at Table 1. their variations result in the most essential uncertainty.4 × 10−14 GeV. The variation of threshold energy in the bc give the error less than 1%.V. Naively. We use the three-point sum rules to determine the dependence on q 2 of form factors in vicinity of q 2 = 0. hence.0 Mode . Summing up all of mentioned variations we estimate δf/f ' 5%.41 GeV [79]. where the method works even in the approximation by a bare-quark loop [47–49]. minimized. we expect a simple pole form  f q2 = f (0) 1− q2 2 Mpole . The mesons masses are equal to MBc = 6. the bare-quark loop approximation cannot be justified at q 2 ∼ m2pole .4 GeV for the form factors with the Bs∗ decay modes.3 ÷ 1.2 5. potential models and the values in the SR. The enhancement of form factors in the decays of heavy quarkonium considered in the framework of sum rules was also found in the decays Bc → ψlν [35. and its variation is less than 2%.975 [77]. We have supposed the quark-mixingmatrix element |Vcs | = 0. The heavy-quark masses are determined by the twopoint sum rules for the heavy quarkonia. Table 2 The widths of semileptonic Bc decay modes and the branching fractions calculated at τBc = 0. while the three-point sum rules in the form of double-dispersion relation cannot be explored in the region of resonances at q 2 > 0 because of so-called “nonLandau” singularities indicating the presence of strongly bound states in the q 2 channel. The effective Coulomb constant is fixed from the two-point sum rules for the heavy quarkonium.46 ps Γ (10−14 GeV) BR (%) Bs e+ νe 5. It is important to stress that the current consideration removes the contradiction between the estimates in OPE.0 Bs∗ e+ νe 7. which may deviate from the value of physical mass of corresponding bound state. too.8 4. Kiselev et al. MBs∗ = 5.

368

V.V. Kiselev et al. / Nuclear Physics B 585 (2000) 353–382

4. The symmetry relations
At the recoil momentum close to zero, the heavy quarks in both the initial and final
states have small relative velocities inside the hadrons, so that the dynamics of heavy
quarks is essentially nonrelativistic. This allows us to use the combined NRQCD/HQET
approximation in the study of mesonic form factors. The expansion in the small relative
velocities to the leading order leads to various relations between the different form factors.
Solving these relations results in the introduction of an universal form factor (an analogue
2 .
of the Isgur–Wise function) at q 2 → qmax
We consider the soft limit
µ

µ

v1 6= v2 ,

w = v1 · v2 → 1,
(26)
q
µ
µ
2 are the four-velocities of heavy mesons in the initial and final
where v1,2 = p1,2 / p1,2
states. The
q study of region (26) is reasonable enough, because in the rest frame of Bc meson
µ
µ
(p = ( p2 , 0E )), the four-velocities differ only by a small value |pE2 | (p = (E2 , pE2 ),
1

1

2

whereas their scalar product
w deviates from unity only due to a term, proportional to the
q

square of |pE2 |: w = 1 + |pE2 |2 /p22 ∼ 1 + 12 |pE2 |2 /p22 . Thus, in the linear approximation
at |pE2 | → 0, relations (26) are valid and take place. Here we would like to note, that
(26) generalizes the investigation of [37], where the case of v1 = v2 was considered. This
condition severely restricts the relations of spin symmetry for the form factors and, as a
consequence, it provides a single connection between the form factors. In the soft limit of
zero recoil we find
µ
µ
µ
(27)
v˜3 = − 12 v1 + v2
for the four-velocity of spectator b quark, and
m3
µ
µ
(v1 − v2 )µ
v˜1 = v1 +
2m1

(28)

for the decaying c-quark. The matrix element of Jµ = Q1 Γµ q2 with the spin structure
Γµ = {γµ , γ5 γµ } has the form

  

µ 
ν

(29)
HQ1 Q
S3 Jµ Hq2 Q
S3 = tr Γµ 1 + v˜1 γµ Γ1 1 + v˜3 γν Γ2 ρlight · h,
S3 (in our case it is pseudoscalar,
where Γ1 determines the spin state in the heavy meson Q1 Q
so that Γ1 = γ5 ), Γ2 determines the spin wave function of quarkonium in the final state
Γ2 = {γ5 ,  µ γµ } for the pseudoscalar and vector states, respectively (H = P , V ). The
‘propagator of the light quark’ 2 is taken in a general form
ρlight = 1 + B(/v 2 − v/ 1 ) + C(/v 2 + v/1 ) + D/v 2v/1 ,

(30)

where B, C and D are the functions of w. The quantity h in (29) at w → 1 is an universal
factor independent of the spin state of meson. So, for the form factors, discussed in our
paper, we have:
2 More strictly, we determine the spin structure of matrix element and the term given by light degrees of
freedom.

V.V. Kiselev et al. / Nuclear Physics B 585 (2000) 353–382



S1 γ µ Q3

P S = cP · v µ + cP · v µ · h,
PQ1 Q
S3 Q
1
2
q2 Q3
1
2

S1 γ µ Q3

V S = icV ·  µναβ ν v1α v2β · h,
PQ1 Q
S3 Q
q2 Q3

 .

S1 γ5 γ µ Q3 .

0 and f− ' −8.8 and f (q 2 ) = −8. PQ1 Q S3 Q q2 Q3 1 2 369 (31) (32) (33) where e − m3 . for example.2 depend on the C and D parameters. F0A · cV − 2c · FV M1 M2 = 0. 5. Substituting B e in first and third relations. cV = −1 − B 2m1 m 3 e+ e − m3 . The deviation could increase at q 2 < q 2 q 2 = qmax max because of variations in the pole (∗) masses governing the evolution of form factors. while their ratios develop more slowly. We have the and B 1+C symmetry relations for the following form factors: 3   f+ c1P · M2 − c2P M1 − f− c1P · M2 + c2P · M1 = 0. c2P = 1 + B (34) c1P = 1 − B 2m1 2m1 e = B−2D . For the dominant nonleptonic decay modes Bc+ → Bs(∗) π + (ρ + ) the effective Hamiltonian can be written down as  GF ∗ C+ (µ)O+ + C− (µ)O− . (35) where M1 = m1 + m3 . Equating the second relation in (35). we obtain e = − 2m1 + m3 + 4m3 (m1 + m3 )FV ' 10. Nonleptonic decays and the lifetime The hadronic decay widths can be obtained on the basis of assumption on the factorization for the weak transition between the quarkonia and the final two-body hadronic states. (38) 3 To remove an error in [36] the analogous second relation for the B → J /Ψ (η ) transition should have the c c missed factor 2 in front of c . (36) B 2m1 F0A 2 . M2 = m2 + m3 . .3.0. we find that in the QCD sum rules. However. so that the changes of form factors are about 50%. (25)). where (37) Heff = √ Vcs Vud 2 2     O± = u¯ i γν (1 − γ5 )di s¯j γ ν (1 − γ5 )cj ± u¯ i γν (1 − γ5 )dj s¯i γ ν (1 − γ5 )cj . form factors obtained in the QCD sum rules: f+ (qmax − max where we suppose the pole like behaviour of form factors (see Eq. we where all form factors are taken at qmax get f+ ' 2. Thus.V S = c ·  µ + c1 · v µ ( · v1 ) + c2 · v µ ( · v1 ) · h. .1. These values have to be compared with the corresponding 2 ) = 1. c = −2. The rest coefficients c1. in Bc+ → Bs l + ν decays the phase space is restricted. relations (35) are valid with the accuracy better than 10% at 2 . F0A c1P + c · M1 (f+ + f− ) = 0.

for example.2 a12 20.0 a12 7. The results are collected in Table 3. The factors C± (µ) account for the strong corrections to the corresponding four-fermion operators caused by hard gluons. we insist Table 3 The widths of dominant nonleptonic Bc decay modes due to c → s transition and the branching fractions calculated at τBc = 0. In addition.26 Mode Γ (10−14 GeV) BR (%) Bs π + 15.9 22. / Nuclear Physics B 585 (2000) 353–382 where i.9 GeV. In our calculations we put the following light-meson parameters: mπ + = 0. mρ + = 0. Kiselev et al.2 .8 a12 17. The Bc+ → Bs π + amplitude.14 GeV. The review on the evaluation of C± (µ) can be found in [80]. In the present paper.5 Bs ρ + Bs∗ π + Bs∗ ρ + 6. takes the form  A Bc+ → Bs π + GF ¯ ν (1 − γ5 )d |0i hBs | s¯ γ ν (1 − γ5 )c |Bc i.132 GeV. The modern lattice estimates show that the ‘bag’ parameters are about 7% less than 1 [74].77 GeV. Concerning the comparison of hadronic width summing up the exclusive decay modes with the estimate based on the quark–hadron duality.7 a12 6. In the parton approximation we could expect      2 2 (µ) + C− (µ) Γ Bc+ → Bs(∗) e+ νe .V. we explore the C± (µ)-evolution to the leading log accuracy. It is worth noting that the sum of widths for transitions Bc+ → Bs (Bs∗ )π + (ρ + ) is 10% larger than the width for the transition Bc+ → Bs (Bs∗ ) + light hadrons.4 6.9 GeV.208 GeV. Γ Bc+ → Bs(∗) + light hadrons = 2C+ which results in the estimate very close to the value obtained as the sum of exclusive modes at µ > 0. dealing with the QCD sum rules in the leading order over αs . the deviation between these estimates can be caused by the corresponding ‘bag’ factor appearing in the formulation of factorization approach and vacuum saturation in the connection of leptonic form factors to the hadronic ones.46 ps. fπ + = 0.370 V. where we neglect the contributions given by the modes with the factor a2 instead of a1 . (39) = √ Vcs Vud a1 (µ)hπ + | uγ 2 where a1 (µ) = 12 (C+ (µ)(Nc + 1) + C− (µ)(Nc − 1))/Nc at Nc = 3 being the number of colors. j run over the colors. fρ + = 0. The deviation between these two estimates slightly increase at mc /2 < µ < 0. We put a1 = 1. which is calculated using the simple formula     Γ Bc+ → Bs (Bs∗ ) + light hadrons = Nc a12 (µ) Γ Bc+ → Bs (Bs∗ )e+ νe .

In order to estimate the contribution of nonresonant (∗) 3π modes of Bc decays into Bs we use the Bjorken’s technique. The width of the electroweak in [36]: Γ (Bc+ → cc annihilation is taken from [32] as 12 × 10−14 GeV. where the basic uncertainty is given by the variation of heavyquark masses. First. once we consider the Kπ .26]. which means that BR(Bc+ → Bs(∗) (3π)+ ) ≈ 0. 4 We stress that in the Bc → Bs decays caused by the weak decays of charmed quark the possible hadronic final states are the charged mesons π . while BR(Bc+ → (∗) Bs (2π)+ |non-resonant) ≈ 3%. we see that the nonresonant multiparticle states are suppressed in comparison with the resonance yields. b → c transitions with the Bs(∗) and J/ψ. Kππ and K ∗ π . Kρ branching fractions in the decays of D mesons as measured experimentally.2 ± 0. Kiselev et al. In Fig. 14. So.V. ρ and K or the multiparticle states like ππ . the states with the kaon are suppressed by the Cabibbo angle.V. as given by the Poisson distribution with the average value of pions determined by the energy release. The method for the calculation of multiparticle branching fractions was offered by Bjorken in his pioneering paper devoted to the decays of hadrons containing heavy quarks [81].e. The width of beauty decay in the sum rules was derived using the similar methods ¯ + X) = (28 ± 5) × 10−14 GeV. / Nuclear Physics B 585 (2000) 353–382 371 that the deviation between these two estimates is less that 10% and. We also show the results of the lifetime evaluation in the framework of Operator Product Expansion in NRQCD [25. We have chosen the following branching ratios: BR(D+ → K − π + π + ) = 9. . We see that the neglected modes contribute to the total width of Bc as a small fraction in the limits of uncertainty involved.2%. it cannot be treated as an essential argument against the validity of our calculations.85 given by Bjorken. Then. and we neglect their contributions in the total nonleptonic width of Bc (the corresponding error of estimates is about 4%). these parameters are fixed by the two-point sum rules for bottomonia and charmonia. ηc final states respectively.6% and BR(D + → K − π + π + π 0 |non-resonant) = 1. the Poisson distribution with the average value corrected to agree with the nonresonant 3π -modes in the decays of D mesons. i. where mc /2 < µ < mc and the dark shaded region 4 The b¯ → cc¯ ¯ s transition is negligibly small in the Bc decays because of destructive Pauli interference for the charmed quark in the initial state and the product of decay [25]. we check that the ratio of R2π = Γ (Bc+ → Bs(∗) π + )/Γ (Bc+ → Bs(∗) ρ + ) ≈ 0. The same fact can be found. The Bjorken’s model of multiparticle yields in the decays does not distinguish the resonant and continuum final states as well as Bs and Bs∗ . hence. In contrast to OPE. We estimate the lifetime using the fact that the dominant modes of the Bc -meson decays are the c → s. 14 we present the Bc -meson lifetime calculated in the QCD SR under consideration. we have found n¯ ≈ 18 . so that the accuracy of SR calculations for the total width of Bc is determined by the choice of scale µ for the hadronic weak Lagrangian in decays of charmed quark.0 ± 0. We show this dependence in Fig. This consideration confirms us that we take into account all of significant nonleptonic decay modes of Bc . and the electroweak annihilation. He supposed a simple relation for the yields of pions.6%. So. πππ or Kπ .82 calculated in the framework of (∗) (∗) sum rules is close to the estimate R2π = Γ (Bc+ → Bs π + )/Γ (Bc+ → Bs π + π 0 ) ≈ 0.

V. Kiselev et al. To the moment. / Nuclear Physics B 585 (2000) 353–382 Fig. 6. The dependence of Bc -meson lifetime on the scale µ in the effective Hamiltonian (37). wherein we can factorize the Wilson coefficients taken in the perturbative QCD and the matrix elements of operators over the hadronic states with µ usually posed to mQ . are also fixed to the same order (they have to be reevaluated in next-to-leading order). The shaded region shows the uncertainty of estimates. say. Therefore. The corresponding parameters. 14. the analysis of decays in the QCD sum rules is restricted by the leading order (LO) in αs (except the Coulomblike corrections in the heavy quarkonia). we can write down the following formulae: (1) The semileptonic branching fraction of D 0 meson is given by   1 . The discussion on the optimal choice of scale in hadronic decays is addressed in the next section. the heavy-quark masses. Therefore. we use the LO expressions. This prescription is in a qualitative agreement with the current data on the measured lifetimes of charmed and beauty hadrons and their branching ratios for the semileptonic decay modes. we can argue the operator product expansion in the inverse powers of mQ . is µ ' mQ . Let us consider this issue in more details. The motivation is the following: the characteristic scale is determined by the energy release given by the heavy-quark mass. for the sake of consistency. (40) BRsl D 0 = 2 2 (µ) 2 + 2C+ (µ) + C− .372 V. the dark shaded region is the preferable choice as given by the lifetimes of charmed mesons. In this way. Discussion on the lifetimes of heavy hadrons At present the ordinary prescription for the normalization point of Lagrangian generating the nonleptonic decays of heavy quark Q. corresponds to the scales preferred by data on the charmed-meson lifetimes. The dots represent the values in the OPE approach. which are given by the partonic approximation improved by 1/mQ -corrections in HQET or NRQCD.

• fD ≈ fB ≈ 200 MeV. the qualitative agreement of predictions with the measured differences of Γ [D + ] − Γ [D 0 ] and Γ [B + ] − e ≈ 1. In the same way. one usually puts: • µD = mc in decays of D mesons.22 + [2C+ − . fD is the leptonic constant of D meson. The similar expression can be derived for the beauty mesons [82]:     Γ B+ − Γ B0 = |Vbc |2 G2F  m3b fB2  2 2 C+ (µ) − C− (µ) B + 8π 1 3   2 2 e. where the authors of [86] found that the .22 + [2C+ − 2 (µ) + C 2 (µ)](1 + k) 2 + 0. First. A similar statement was obtained in the description of D ∗ -meson production at HERA.V. • B =B At k = 0. hD|(cΓ ¯ µ c)(qΓ (42) (43) 3 with Γµ = γµ (1 − γ5 ).e. Second. we note that the semileptonic width of D 0 is well described to the given order and at chosen value of mc . Voloshin [85] clearly drawn a conclusion that the naive picture of color structure as given by the potential models (i. and the ‘bag’ constants are defined by = cos4 θc G2F 2 ¯ µ c)|Di = fD2 MD B. (µ) + C− (µ) B C+ (44) (3) The semileptonic branching fraction of B 0 meson is given by   BRsl B 0 = 1 2 (µ) + C 2 (µ)](1 + k) 2 + 0. (46) As for numerical applications. (45) where the fraction of 0.22 is due to the τ ντ -contribution. hD|(cΓ ¯ µ q)(qΓ 1 2 2 e ¯ µ q)|Di = fD MD B. while its branching ratio is in a valuable contradiction with the data indicating a more higher enforcement of nonleptonic modes. while the value of k denotes the fraction of b → ccs ¯ transition in the nonleptonic decays. • µB = mb in decays of B mesons.22 + [2C+ − .4 [83. so that [82]     Γ D+ − Γ D0    m3c fD2  2 2 2 2 e . the average yield of charm in the decays of beauty mesons is equal to nc = 2 (µ) + C 2 (µ)](1 + 2k) 2 + 0.V. the purely antisymmetric colorcomposition of valence flavors) is significantly broken. this set (marked as SETMQ column) results in the estimates shown in Table 4 in comparison with the experimental data [79].84]. Recent consideration of charmedΓ [B 0 ] is mainly based on the assumption of B baryon lifetimes by M. / Nuclear Physics B 585 (2000) 353–382 373 (2) The difference between the total widths of charmed mesons is determined by the Pauli interference in decays of D + . Kiselev et al. and e = 1 naively motivated by nonrelativistic potential models. (41) C+ (µ) − C− (µ) B + 13 C+ (µ) + C− (µ) B 8π where θc is the Cabibbo angle.

2M Ψ (q) √ (50) (8) 3 (2π) 2 2 where the c¯ and q fields represent the rapid valence quarks.1 ± 1. O(8)[D]/O(1) [D] ' 1.017 −0. if we introduce the valuable contribution by the color-octet state in addition to the singlet one.043 ± 0.024 −0.05 1.4 8.05 1.00 1.05 1.20 1.00 ± 0.05 nc 1. So.53 Γ [D + ] − Γ [D 0 ] (ps−1 ) −1. 8.8)[D] can be obtained from the above expressions for D ∗ by the substitution of γ5 γµ for γµ and removing the transverse projector.e.1 SETMQ SETMQ2 SETH 15.00 1. We have hna nb i = δab in the production. can be represented as O(8) = |Ψ(8)(0)|2 .00 1.4 15. by the quark–gluon sea). (47) ¯ µ q)(qγ ¯ µ c)|D ∗ i · −gµν + O(1) = hD ∗ |(cγ p2 12M which is reduced to O(1) = |Ψ(1)(0)|2 in the framework of nonrelativistic potential model.20 1. We can analogously expect that in the D meson the ratio of four-quark-matrix elements is close to that in D ∗ . Then we straightforwardly find .044 Γ [B 0 ]/Γ [Bs ] 1.12 ± 0.12 ± 0.56 ± 0. So.21 14 14 10.19 −1.3.374 V.00 1.03 Γ [D 0 ]/Γ [Ds ] BRsl [B 0 ] (%) Γ [B + ] − Γ [B 0 ] (ps−1 ) 1.26 −0. in a similar way. where O(1.11 10.2 −0. and na is a random color-vector determined by soft degrees of freedom inside the meson (i. the four-quark singlet-operator results in  pµ pν  1 . / Nuclear Physics B 585 (2000) 353–382 Table 4 The comparison of theoretical estimates at various sets of parameters with the experimental data on the decays of heavy mesons Quantity BRsl [D 0 ] (%) Exp.00 0. The term of color-octet was parameterized by  pµ pν  1 .05 1.022 0. The data on the D ∗ production in DIS give O(8) /O(1) ' 1. Kiselev et al.45 ± 0.V. where Z √ δij 1 d3 q Ψ(1) (q) √ · √ c¯i /qj |0i. while hna nb i = 18 δab in decays. if we introduce the additional Fock state Z √ λaij d3 q 1 na · √ c¯i /qj |0i.81 O(αem αs3 )-calculations for the differential cross sections of cq-pair ¯ composing the meson. are able to reproduce the measured spectra.00 1. (48) |D ∗ i = 2M (2π)3 3 2 with α denoting the polarization vector and Ψ(1) (q) being the wave function.12 Γ [B 0 ]/Γ [Λb ] 0. (49) ¯ a γµ q)(qλ ¯ a γµ c)|D ∗ i · −gµν + O(8) = hD ∗ |(cλ p2 64M which is.6 −1.81 ± 0.

T determines the 2S–1S splitting and it is approximately independent of quark flavors.V. We see a good agreement with the data. we put µ2c¯s = µ2bs¯ = 2T Λ¯ s . ¯ where the ¯ form-factor behaviour versus the transferred momentum is determined by the c¯s and bc states. widths for the heavy mesons at the most reasonable value of B In this position we argue the following: there are other physical scales in the problem. ¯ We use more realistic values for k close to 0. to our opinion. (54) with mbc ' mb mc /(mb + mc ) ' mB mD /(mB + mD ) ≈ 1. Analogously. Putting B So. µ2Bs = µbs¯ · µcb¯ .5 GeV. Next. the choice of µ = mQ is in a deep contradiction with the observed differences of total e = 2. The same effect has to take place in the b-quark decays with two charmed quarks in the final state. (55) (56) e = 2. being the average squares of heavy-quark momentum hp2 i. We stress that the theoretical OPE prediction for the contribution of b → ccs ¯ mode. T ' 0. looking at the semileptonic decay with τ + ντ in the final state we see that at the same values of form factors as in modes with light leptons the heavy-lepton mode is suppressed because of the restricted phase space. which are characteristic for two hadronic systems in the decay process. / Nuclear Physics B 585 (2000) 353–382 e = 1 + O(8) [D] ≈ 2. (53) where Λ¯ s = Λ¯ + (mDs − mD ) = Λ¯ + (mBs − mB ) ' 0. that certainly leads to the overestimation of phase space in the decay of b → ccs. Indeed. Those hadronic systems have the following scales. OPE operates with the quark masses at so moderate release of energy.18 is represented by the SETH column in Table 4. µ2Ds = µ2c¯s . so that the reduction factor equals 0. 15. Kiselev et al. So. Nevertheless. B O(1) [D] 375 (51) e = 2 into the SETMQ we get the values given in the SETMQ2 column in Table 4. since the sum of D and K masses is greater than the τ mass we could expect even the greater suppression than for the heavy-lepton mode.4 GeV is the binding energy of heavy quark (i. The result of µB and k variations is also shown in Fig. which are phenomenologically equal to ¯ µ2cu¯ = µ2bd¯ = 2T Λ. The first system is the decaying hadron.45 GeV is the average kinetic energy in the system. k = 0. So.3 GeV.V. The second is the transition current c → s or b¯ → c.22. µ2B = µbu¯ · µcb¯ . the contribution of four-quark operators are suppressed in decays of Λb baryon. For the b → c current we put µ2cb¯ = 2T mbc . and k = 0. the constituent mass of light quark). this set of parameters with fB ∼ At B = fD ∼ = 175 MeV. (52) where according to the potential models Λ¯ ' 0. one expects that the deviation of its total width from that of B mesons is about . is significantly overestimated.e.4. Let us suppose that the decay scale is given by the following combinations: µ2D = µcu¯ · µc¯s .2.

The predictions on the semileptonic branching ratio of B 0 meson and the average yield of charmed quarks in its decays with the ±3%-variation of hadronic scale in the nonleptonic effective Lagrangian and the change of doubly charmed mode fraction k = 0. and. To the moment. Then we pose µ2Λb = µud · µcb¯ .V. To finish this discussion we draw the conclusion: a probable way to reach the agreement between the theoretical predictions and available experimental data on the lifetimes and inclusive decay widths of heavy hadrons is to suggest the different normalization points in the effective nonleptonic Lagrangian for the heavy-quark weak decays as dependent on the hadron.20 in the charmed-quark decays we predict τ [Bc ] = 0. second.05 ps. 15. (59) and at a1 (µBc ) = 1. (60) . the experimental result is far away from this expectation (see Table 4).85 GeV)2 . 2 (57) where T /2 corresponds to the half tension of color string inside the diquark. the appropriate choice of scale can allow us to decrease the scale-dependent higher-orders terms. Kiselev et al. We can reproduce the data on the Λb lifetime. the necessity to proceed with the higher-orders. first. which indicates. This assumption provide us with quite acceptable results to the leading order in αs .48 ± 0. we suppose that the preferable choice of scale in the c → s decays of Bc is equal to µ2Bc = µcb¯ · µc¯s ≈ (0. 2–3% [87].22 in comparison with the ARGUS and CLEO data shown by the dot with the error bars. The variation of normalization point shows the sensitivity of calculations to the higher orders in the QCD coupling constant. if we introduce the light-diquark scale T ¯ µ2ud = 2 Λ. (58) which result in a good agreement with the experimental data.376 V.15 ÷ 0. / Nuclear Physics B 585 (2000) 353–382 Fig. Thus.

Q2 ) we have the following expressions [36]:     3 k (∆1 + ∆2 ) − k m3 (m3 − m1 ) + m3 (m3 − m2 ) ρ+ s1 . Finally. Q2 = 3/2 2k 2   − 2(s2 ∆1 + s1 ∆2 ) − u(∆1 + ∆2 ) . The obtained dependence has the stability region.V. As in the case of twopoint sum rules. Appendix A For the perturbative spectral densities ρi (s1 . we have estimated the Bc -meson lifetime. This contribution essentially improves the stability of SR results for the leptonic constants of B mesons. and showed the dependence on the scale for the hadronic weak Lagrangian in decays of charmed quark: τ [Bc ] = 0. s2 . which determines the continuum ¯ system. Our estimates are in a good agreement with the theoretical predictions for the lifetime in both the potential models and OPE as well as with the experimental data.48 ± 0. Kiselev et al. which means that the corrections to the form factors ¯ quarkonium and the inverse heavyin both relative velocity of heavy quarks inside the bc quark masses are small within the accuracy of the method. The results on the widths and branching fractions for various decay modes of Bc are collected in the tables. The HQET two-point sum rules for the leptonic constant fBs and fBs∗ have been reanalyzed to introduce the term caused by the product of quark and gluon condensates. which allows us to derive the generalized relations due to the spin symmetry of effective Lagrangian. Acknowledgement This work was in part supported by the Russian Foundation for Basic Research. Next. We have found the important role played by the Coulomb-like αs /v-corrections. / Nuclear Physics B 585 (2000) 353–382 377 7. Conclusion We have investigated the semileptonic decays of Bc meson due to the weak decays of charmed quark in the framework of three-point sum rules in QCD. s2 . The relations are in a good agreement with the full QCD results. using the assumption on the factorization of the weak transition. grants 99-02-16558 and 00-15-96645. serving as the test region of bs of convergency for the sum-rule method. yielding: fB = 140 ÷ 170 MeV.V. we have studied the nonleptonic decays. the form factors are about three times enhanced due to the Coulomb renormalization of quark–meson vertex for the heavy quarkonium Bc .05 ps. We have studied the soft limit for the form factors in combined HQET/NRQCD technique at the recoil momentum close to zero. We have studied the dependence of form factors on the threshold energy.

s2 . / Nuclear Physics B 585 (2000) 353–382 h i u × m23 − + m1 m2 − m2 m3 − m1 m3 .V.   3  A s1 . s2 . Q2 = − 5/2 −2(m1 − m3 ) (k − 3us2 )∆21 + 6s12 ∆22 2k  + ku(m1 − m3 ) 2m23 + ∆2 + ku2 m3 . (63)      1 3 ρ0A s1 . s2 . Q2 = − 3/2 2k 2   + 2(s2 ∆1 − s1 ∆2 ) + u(∆1 − ∆2 ) i h u × m23 − + m1 m2 − m2 m3 − m1 m3 . Q2 = 3/2 (2s1 ∆2 − u∆1 )(m3 − m2 ) k + (2s2 ∆1 − u∆2 )(m3 − m1 ) + m3 k .378 V. Q     3 = 3/2 m1 2s2 ∆1 − u∆2 + 4∆1 ∆2 + 2∆22 2k + m1 m23 [4s2 − 2u] + m2 [2s1 ∆2 − u∆1 ]  − m3 2(3s2 ∆1 + s1 ∆2 ) − u(3∆2 + ∆1 ) + k + 4∆1 ∆2  + 2∆22 + m23 (4s2 − 2u)   6 + (m1 − m3 ) 4s1 s2 ∆1 ∆2 − u 2s2 ∆1 ∆2 + s1 ∆22 + s2 ∆21 k   (65) + 2s2 s1 ∆22 + s2 ∆21 . (64) A ρ+ 2 s1 . 2 (61) (62)  3  ρV s1 . 2     3 k (∆1 − ∆2 ) − k m3 (m3 − m1 ) − m3 (m3 − m2 ) ρ− s1 . Q2 = − 5/2 kum3 2m1 m3 − 2m23 + u + 12(m1 − m3 )s22 ∆21 ρ− 2k   + k∆2 (m1 + m3 )u − 2s1 (m2 − m3 ) + 2∆22 (k + 3us1 )(m1 − m3 )   + ∆1 ku(m2 − m3 ) + 2∆2 (k − 3u2 )(m1 − m3 )   + 2s2 (m1 − m3 ) 2km23 − k∆1 + 3u∆21 − 6u∆1∆2  (66) − 2s1 s2 km3 − 3∆22 (m1 − m3 ) . Kiselev et al. s2 . Q2 = 1/2 (m1 − m3 ) m23 + s1 ∆22 + s2 ∆21 − u∆1 ∆2 k k     ∆1 ∆2 − m1 m23 − − m2 m23 − 2 2   2 1  + m3 m3 − 2 (∆1 + ∆2 − u) + m1 m2 . s2 . s2 .    3  0 ρ+A s1 .

MBc MBs∗ ξ(w). ∆1 = s1 − m21 + m23 and ∆2 = s2 − m22 + m23 . Q2 = (67)   3  2(m1 − m3 ) (k + 3us2 )∆21 + 6s12 ∆22 2k 5/2  + ku(m1 − m3 ) 2m23 + ∆2 + ku2 m3  + ∆1 ku(−2m1 − m2 + m3 )  − 2(m1 − m3 ) ks2 − k∆2 + 3u2 ∆2   + 2s1 (m1 − m3 ) 2km23 − 6u∆1∆2 + 3u∆22  − 2s2 km3 − 3m1 ∆21 + 3m3 ∆21  + k∆2 (2m1 + m2 − m3 ) .V.V. s2 . MBc MB2 c + MB2 ∗ − q 2 + 2MBc (e ms − m eb ) F0A = F+A 1 − 2e mb /MBc =− 2e ms s 1 + 2e mb /MBc F−A = 2e ms where s 2e ms s s MBs∗ ξ(w). Appendix B Here we list the expression for the form factors of semileptonic decays Bc → Bs(∗) taken from the potential model [78]: s (e mc + m es ) MBs ξ(w). Kiselev et al. (69) f+ = 2e ms MBc (e mc − m es + 2e mb ) f− = − 2e ms MBs ξ(w). MBc (70) s MBs∗ ξ(w).  0 ρ−A s1 . see prescriptions in Fig. 1. (68) Here k = (s1 + s2 + Q2 )2 − 4s1 s2 . m1 . m2 and m3 are the masses of quark flavours relevant to the various decays. u = s1 + s2 + Q2 . / Nuclear Physics B 585 (2000) 353–382 379  + ∆1 ku(2m1 − m2 − 3m3 )  − 2(m1 − m3 ) ks2 − k∆2 + 3u2 ∆2   − 2s1 (m1 − m3 ) 2km23 − 6u∆1 ∆2 − 3u∆22  + 2s2 km3 + 3m1 ∆21 − 3m3 ∆21  + k∆2 (2m1 − m2 − 3m3 ) . MBc (71) (72) (73) .

/ Nuclear Physics B 585 (2000) 353–382 s 2ωωx ξ(w) = 2 ω + ωx2  ω = 2π   m e2b (w2 − 1) 2ωωx exp − . Phys.8 GeV. O. ! MB (∗) f˜2(∗) 1/3 s 12 Bs (74) . Likhoded. G. Bs (76) References [1] F.C. Yad. D 51 (1995) 3613. Phys. Buchalla. [20] C. D 50 (1994) 6013(E). Lusignoli.V. Chang et al. Shevlyagin. [10] E. 59 (1996) 742. B 348 (1995) 219. hep-ph/9912424. A. Likhoded.V. [4] E. Phys. Lett. Lett. V. Eichten.I.V. [26] A. Rev. Phys. Shevlyagin. Woloshyn.V.V.K. Yad. D 48 (1993) 4230. Yuan. m ec = 1. Berezhnoy. E. K. [13] K. [21] K. Lett. [3] T.V. Likhoded. Rev. Yuschenko.K. D 53 (1996) 4991. M. Rev.T. [8] H. Lett.V. Phys. B 364 (1995) 78. D 60 (1999) 014502. C 67 (1995) 159. Phys. C. [25] M. Phys. P. Berezhnoy.K.. ˜ f (∗) = 0. Rückl. 2000. Likhoded. Fiz.K.D.V. H. Lett.V. V. A. Braaten. Rev. Brambilla. Phys. R. D 49 (1994) 5845. A 356 (1996) 79. Fiz.V. Likhoded. [2] G. Shanahan. . Likhoded. Bigi. A 356 (1996) 89. Quigg. Z. Kiselev. Phys. [22] A. [23] A. Rev. [27] M.M. Cheung. Nucl. Z. Phys. D 53 (1996) 3591. Chang. Lett.. Berezhnoy. CDF collaboration. Kolodziej. A. [7] B. ω2 w2 + ωx2 ω2 w2 + ωx2  MBc f˜B2c 1/3 12 ωx = 2π .C.S. Likhoded. Phys. V.P. Berezhnoy. Berezhnoy. 81 (1998) 2432. [11] V. Phys.5 GeV. A. Berezhnoy. Kiselev. Newton. A. R. Schuller. Y. hep-ph/0002075. Leike.-H. Likhoded. Kiselev. Mannel.T.K. [17] A. Phys. Phys. Preprint CERN-TH-2000-036. B 355 (1995) 337. Z. Z. Davies. A. B 381 (1996) 341. [24] I. Onishchenko.K. C 51 (1991) 549. Vairo. 60 (1997) 1889. [6] N. Phys. B 342 (1995) 351. D 58 (1998) 112004. Rückl. D 50 (1994) 5664.-H. [14] A. Cheung. [12] T. Berezhnoy.V. M. Beneke. [19] A.. Phys. Phys. M. A. Kolodziej.P.380 V.V.P. A. Phys. Kiselev et al. Phys.A. B 453 (1999) 289. Rev. R. 58 (1995) 730. [5] S. Phys. The quark masses and the leptonic constants have the values generally used in the calculations in the framework of potential models: m eb = 4.17 GeV.V. Atom. (75) (∗) where w is the product of Bc and Bs four-velocities.H. Rev. D 46 (1992) 3845.V. C 63 (1994) 77. Masetti. [18] A. f˜Bc = 0. Yuan. A. Onishchenko.I. Chen. Lepage. Yad. Phys.47 GeV. Braaten. Gershtein et al. Z. Kiselev.-Q. A. Boyle. Lett. Bodwin. Shevlyagin.C. Fiz. G. 60 (1997) 1729. Rev. UKQCD collaboration. Rev. A. Leike. Rev.K. [9] C. Phys. Kiselev. T. T.55 GeV. Abe et al. Jones. C. [16] A. [15] K. Lett. A. m es = 0. Yuan. Phys. Phys. Rev. D 51 (1995) 1125. Rev. V. M. G.K. B 371 (1996) 105.V.V. Phys.

[64] K.V. G. Sov. Nucl. Kataev.A. B. E. [36] V. Phys. Chetyrkin. [53] S.A. Phys. [73] S. D 52 (1995) 3929. Hoang.A. A. Braun.6879/93. Narison. Nucl. B 390 (1993) 463. [31] D. B 267 (1991) 105. Ovchinnikov. Lett. P.G. Phys. [51] V. Pivovarov. Lett. V. de Rafael. Likhoded. Beneke. Phys. B 135 (1984) 457.M. Nucl. Lett. Phys. O.J. [55] V. Nucl. J.A. [69] E. [46] V. Lu. Wise. Manohar. A. B 325 (1994) 205. N. Krasnikov. Phys. Krasnikov.A.I. Nucl. Yad. Onishchenko. Heavy Ion Phys.I. Krasnikov. Gorishny. J. [29] A. Semay. Z. Launer.A. D 48 (1993) 5208.R. Pivovarov.M. Phys.V. Lett. Phys. [60] D. [70] R. [33] P. H. B 341 (1995) 391. B 406 (1993) 340. K. Nardulli. Isgur. V. Rep. Phys. Krasnikov.V. Scora.G. Moorhose. E. Chetyrkin. C 57 (1993) 43. [39] G.. [57] A. Yelkhovsky. S. B 389 (1993) 534. Phys. [42] M. Phys. [72] V.V. [34] E. Likhoded. Phys.G. Aliev.S. . Phys. Novikov et al. Phys. Neubert. Melnikov. A. Reinshagen. Rev. [67] A. Dosch. [50] S. Kulikov. A.S. 62 (1999) 1868. Vainshtein. Braun. B 76 (1978) 83. Bagan et al. A. D 45 (1992) 2451. [66] N.J.V. [47] P. B. C. Beneke. Shifman. 245 (1994) 259. Phys. Ball. Phys. Phys. A. Kiselev. Phys. 62 (1999) 1739. Rev. hep-ph/9905359. Savage. [38] P. G. Lett.G. Nucl. Phys. A. Phys.R. F. N. Larin. Phys.V.Yu. Kataev. Phys.V.A. Kiselev. Lett. Phys. B 250 (1985) 61. B 119 (1977) 174. D 39 (1989) 799. V. 9 (1999) 133. [61] R. Phys. A.G.V. I. Nucl. hep-ph/9911490. V. Kiselev. [40] M. Rev. Phys. I. Phys. A. [30] N. J. Nucl.L. 50 (1989) 1063. D 15 (1977) 755. Dugan. [62] R. de Rafael. Grozin. S. Rev. C 19 (1983) 301. Scora. A 11 (1996) 3689. B 210 (1988) 238. M. Gershtein et al. B 388 (1992) 346. Phys. [45] L. [58] K. Phys. Rev. Silvestre-Bra. N. Phys. [59] M. Mod. Z. Fiz.A. A. Pennington. MPI-Ph/93-88. Rev.I. Ter-Martirosian. Tavkhelidze. Yazaki.A. Grinstein. hep-ph/9811212. M. Preprint DESY 99-053. Kiselev. Narodetsky. Jenkins. Kiselev. A. Narison. hep-ph/9812514. R. Lett. B. Nucl. Kiselev. D 52 (1995) 2783.V. [41] M. Phys. D. B 123 (1983) 93. Nucl. Nuovo Cimento A 105 (1992) 827. Rep. Bertlmann.G. Colangelo. Rev. N.I. Chabab. Floratos. Int. B 452 (1995) 563. Yad. [43] B.I.. [44] M. Isgur.V. Reinders.H. A. Broadhurst. Neubert. Tkabladze. Lett. A. Shankar. Paver.N. 50 (1989) 891. B 255 (1991) 583. S. Onishchenko. hep-ph/9809249.B. D 48 (1993) 2110. C 41 (1978) 1. Ross. Grinstein. Rep. 50 (1989) 1433.G. B 124 (1977) 285. Ball. N. Luke. Yilmaz.V. N. Yang. hep-ph/9803433. Phys.V. D 61 (2000) 034012. A. M. Zakharov. Preprints CERN-TH. H. Phys.K.-D. Y. Phys. Rev.M. C 64 (1994) 57. Onishchenko. M. Slobodenyuk. [54] M. Phys. Phys.G. B.V. Phys. 1993. hep-ph/9905550. Nucl.N. A. D 59 (1999) 114009. Phys. / Nuclear Physics B 585 (2000) 353–382 381 [28] A. Tkabladze. M. H. Li. [56] M. Shifman. M.A. [37] E. Phys. [63] K. in press. B 155 (1979) 115. Narison.A. Lett. in press.Yu. Lett. Gershtein. Preprint IHEP 98-22. Nucl. B 147 (1979) 385. Kiselev et al. Atom. [65] N. Pivovarov.V. Anisimov.-G. Phys.. Anisimov. Lett. De Fazio.M. Rubinstein. Z. Krasnikov.Yu. Phys. [32] S. B. Colangelo.M. Shifman. [49] A. [71] S. [52] T. [35] V. Rückl. Nucl. Sov. hep-ph/9909337. Narodetskii. V. Phys. Rev. B 322 (1994) 247. Tavkhelidze. [68] S. [48] P.V. G.L. 127 (1985) 1. B 112 (1982) 397. B 452 (1999) 129.A. A.J.-B. Blok. Phys. Nucl.V. Nucl. Fiz.

2000.J. 289 (1997) 1. Invited talk at International Europhysics Conference on High-Energy Physics (HEP 97). M.fnal. Bigi. Bargiotti et al. Mod. V. Braun. Rev. hep-ph/9504313.J. Dornan. 1986. 1999. V.V. 1999.PDF. pp. M. in: Jerusalem 1997. A. Neubert.B. http://www-theory. V. Phys.382 [74] [75] [76] [77] [78] [79] [80] [81] [82] [83] [84] [85] [86] [87] V. A 10 (1995) 1049. hep-ph/9912322.B. 243–268. M. Buras. 1999. Kiselev et al. Phys. Likhoded. Mod. Voloshin. Berezhnoy. Kiselev. C. hep-ph/0004257.. Neubert. Phys. Bellini. Lett. hep-ph/9901445.E.V. Eur.V. hep-ph/9911206. 1998. Phys.D.. Preprint CERN-TH-98-2. August 1997. G.K. M. C. 1997. Lellouch. hep-ph/0001293. M. G. I. hep-ph/9908455.V. Rep. Israel. D 46 (1992) 1076. Buchalla. M. A. Preprint TPI-MINN-9907-T. Phys. A.gov/people/ellis/Bphysics/ bjorken. Preprint TPI-MINN-99-40-T. Rev.M. Draft report 07/22/86. C 3 (1998) 1. Bjorken. J. / Nuclear Physics B 585 (2000) 353–382 L. High energy physics. Kiselev. Lautenbacher. Neubert. M. Voloshin. Lin. Caso et al. Jerusalem. hep-ph/9905555. P. J.J. hep-ph/9801269. 68 (1996) 1125. .

accepted 4 July 2000 Abstract We compute the 4-dimensional cosmological constant in string compactifications in which the Standard Model fields live on a non-supersymmetric brane inside a supersymmetric bulk.g.V.  2000 Elsevier Science B. The cosmological constant receives contributions only from the vacuum energy of the bulk supergravity fields. All rights reserved. one finds in the case of a single-component bosonic field [1]: ρk = E-mail address: christof. the energy density of the vacuum and thus the effective value of λ changes by a large amount. Theory Division.. The latter is absorbed in a warp factor. Summing these contributions up to some large-momentum cutoff Λ  m. Schmidhuber). Switzerland Received 13 June 2000. Supersymmetry breaking on the brane at the TeV scale implies supersymmetry breaking in the bulk at the micrometer scale. when the Higgs field rolls down its potential.ch (C. This creates a tiny cosmological constant that agrees with experiment within a few orders of magnitude. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. They could be observed in short-distance tests of Einstein Gravity. All rights reserved. Our argument predicts superpartners of the graviton with masses of order 10−3 eV. 1. but not from the vacuum energy of the brane fields. Introduction The observed smallness of the cosmological constant λ in Einstein’s equations poses a fine-tuning problem already in classical field theory coupled to gravity. Other expected contributions to λ that are suspiciously absent include those from condensates in QCD.nl/locate/npe Micrometer gravitinos and the cosmological constant Christof Schmidhuber CERN. 2 of the oscillators with momentum k of the massless and light fields of the Standard Model.Nuclear Physics B 585 (2000) 385–394 www. E.elsevier.3 2 1 3 ( 0 0 ) 0 0 4 2 8 .schmidhuber@cern.V.4 . But perhaps the most mysterious aspect of the problem is that λ does not seem to receive contributions from the quantum mechanical ground state energies h¯ ω . ω 2 = k 2 + m2 . 1211 Genève 23. PII: S 0 5 5 0 .

then in order to explain such a small value of λ one would need a momentum-cutoff as small as 1 1 eV ∼ . Experimentally. Rather. Supersymmetry looks so much like the missing piece in the puzzle that it has been questioned whether supersymmetry is really broken [3]. the scale mBulk-Susy of supersymmetry breaking in the bulk sector and the Hubble expansion rate H0 . 1026 m is the order of magnitude of the curvature radius of the universe. By a simple argument. The recently revived suggestion that we live on a 4-dimensional brane that is embedded in a higher-dimensional bulk opens up a new perspective and a way out (under an assumption stated in Section 2). Based on the known values of mPlanck and H0 . Schmidhuber / Nuclear Physics B 585 (2000) 385–394 ZΛ λ = 8πG Λ4 2 k 2 dk l . because the vacuum energies of the superpartners cancel each other. ρ ∼ k 2π 2 2π P (1. the scale mBrane-Susy of supersymmetry breaking on the brane. The result is a relation between the Planck mass mPlanck . In this sense the observed smallness of the Hubble expansion parameter seems inconsistent even with what we can see with our bare eyes. But even if one considered only the contributions of the two helicity states of the massless photon to the cosmological constant (1. it presently seems that [2] 2 −2 (1.7]. Supersymmetry could explain a zero cosmological constant. it is of course hard to explain why we do not see superpartners of the Standard Model fields [4]. which carries the Standard Model fields. it is absorbed in the warp factor transverse to the brane. the vacuum energy of the brane fields is shown not to contribute to the 4d cosmological constant. It will be proposed below that supersymmetry is indeed unbroken up to micrometer scales — but only in the bulk supergravity theory.3) log mBrane-Susy 2 mBulk-Susy 4 H0 (a more detailed relation is given in the text). Due to a mechanism first proposed in [5] and re-invented in [6. Only the vacuum energy of the bulk supergravity fields is argued to contribute to the 4d cosmological constant. Λ∼ 100 20 µm So the minimum wavelength would have to be as large as 20 µm.1).7 × 10−35 m is the Planck length (setting c = 1).386 C. G is the Newton constant and lP = h¯ G ∼ 1. This vacuum energy is estimated. Already the wavelength of visible light is much smaller than 20 µm. But if it isn’t broken. which is roughly the inverse Hubble constant. supersymmetry breaking in the bulk at micrometer scale is derived from supersymmetry breaking at the TeV scale on the brane.1) 0 √ Here.   mPlanck mPlanck 1 1 mPlanck ∼ log ∼ log (1.2) λ ∼ 2 × 10−33 eV ∼ 1026 m (setting h¯ = G = 1). roughly. this relation predicts gravitinos or other superpartners of the supergravity multiplet with .

Let Vol(M) be the volume of the compactifi cation manifold. The setup is introduced in Section 2. 1. Similarly as in [10. the thickness) of the brane with lBrane-Susy ∼ m−1 Brane-Susy . In Section 4 it is shown that the quantum mechanical ground state energy of the supergravity sector produces a cosmological constant that is within a few orders of magnitude of its observed value. Section 6 contains conclusions. we will identify the size (i. The Standard Model fields are assumed to live only on the brane. The setup We consider a 3-brane soliton that is embedded in a (4 + n)-dimensional bulk spacetime (Fig. the scale of supersymmetry breaking in the Standard Model. Conversely.1) = ∼ lBrane-Susy mPlanck Vol(M) Fig. as explained in Section 5. . In Section 3 it is argued that the 4d cosmological constant is zero as long as the bulk supergravity theory is treated classically. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 387 masses of order 10−3 eV (which is inside experimental bounds [8]) and a supersymmetry breaking scale on the brane of several TeV. 1).e. Precise matching yields the predictions for supersymmetry breaking in the Standard Model and in the bulk sector. Vol(B) ∼ (lBrane-Susy )n . because the Einstein action is integrated over Vol(M) while the Standard Model action is integrated only over Vol(B). the 4-dimensional Planck length lPlanck is related to lBrane-Susy by 2    mBrane-Susy 2 Vol(B) lPlanck (2.C. 2. A 3-brane in a (4 + n)-dimensional embedding space. So roughly. We assume that the n extra dimensions are compactified on some manifold M. and let Vol(B) be the volume of the ball B inside M that intersects with the brane.g. the relation explains the observed small value of the cosmological constant... in [9]).11]. Since we are going to consider non-supersymmetric branes inside supersymmetric bulks (as. based on the assumption that supersymmetry is restored in the Standard Model at energies not too much above the weak scale. while gravity lives in the bulk. e.

Supersymmetry is then broken not only in the bulk theory in the vicinity of the brane. Although the bulk theory is treated classically. In string theory. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 (assuming a (4 + n)-dimensional Newton constant of order one). So we assume that we have a supersymmetric brane inside a supersymmetric compactification manifold. why it does not create a 4d cosmological constant) as long as the bulk supergravity theory is treated classically (see [5–7]). only a 4-dimensional supersymmetric effective theory of Standard Model fields plus their superpartners coupled to 4d supergravity is seen. at the level of classical supergravity. x µ are the spacetime coordinates parallel to the brane.. Supersymmetry of the effective 4d theory implies that the 4d metric gˆ µν is Ricci-flat (we are assuming that there are no 4-form gauge field strengths or expectation values of other supergravity fields). 3. and f (r). this is achieved by considering a compactification on a Calabi–Yau 3-manifold that involves branes parallel to the 4-dimensional spacetime. The basic assumption under which the arguments in the next section apply is that. . This will result in a brane vacuum energy of the order of (mBrane-Susy )4 . Corrections from loops of the bulk fields will be discussed in the next section.e. In other words.15]. For concreteness. The construction of explicit examples must be left for future work. the world-brane theory containing the Standard Model fields is assumed to be treated fully quantum mechanically. At distances much larger than the size of the compactification manifold. we can consistently do the following: we can replace the supersymmetric brane soliton solution by a stable non-supersymmetric one (perhaps of the type of [13]). where r denotes the distance from the brane. g(r) are some functions. Now suppose that we cut out a region of radius lBrane-Susy around the brane. we assume that there are consistent string compactifications that involve spacetime filling stable non-BPS branes. examples of supergravity solutions that smoothly interpolate between a supersymmetric and a non-supersymmetric region are the kink solutions of 5d gauged supergravity discussed in [14. the effective 4d cosmological constant is zero. such that 4-dimensional supersymmetry is unbroken away from the brane at least in the classical supergravity approximation.. In the case of one extra dimension. Classical supergravity approximation Let us first explain why the vacuum energy on the brane does not curve the 4d metric gˆµν parallel to the brane (i. Let us first consider the supersymmetric version of the story. as in [12].388 C.e. i. It is also broken in the world-brane theory that contains the Standard Model fields and lives in the non-supersymmetric gravitational background. such that the bulk fields smoothly connect to a solution at r > lBrane-Susy that does not break supersymmetry on the 4d slices parallel to the brane. we may assume a metric in the vicinity of the brane of the form ds 2 = dr 2 + f (r)gˆµν dx µ dx ν + g(r) dΩ52 . gˆ µν is the 4d metric parallel to the brane.

The issue of boundary conditions at r =  will be commented on below. there may be a singularity or horizon near the center that we know little about. the rest of the bulk M 4 × (M − B). we focus on the example of a single extra dimension. In this non-supersymmetric brane region. Plugging this into the Einstein equation for the 4-dimensional components (µ. Let us assume some distribution ρ(r) around r = 0 with width of the order lBrane-Susy . r) component to eliminate α. matching at r = lBrane-Susy to the solution in . labelled by the constant 4d curvature k 2 . The classical supergravity equations of motion can be solved separately for each region. where  is a cutoff that hides the region where the supergravity approximation breaks down. The 5-dimensional Ricci tensor can be written (cmp. assume a constant dilaton and neglect the other supergravity fields. ¨ 2 4(α¨ + α˙ 2 ) = − λˆ . and the supersymmetric region. Let us therefore restrict the discussion to the region  6 r 6 lBrane-Susy . i. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 389 The bulk has been separated into two regions: the non-supersymmetric neighborhood of the brane M 4 × B. We have also neglected corrections of second order in derivatives of the supergravity background with respect to r. ρ(r) enters the Einstein equations like an r-dependent cosmological constant: 1 Rmn − gmn R = −λˆ (r)gmn 2 with ˆ λ(r) = 8πGρ(r) − λflux (r). the 4d metric gˆ is taken to be r-independent. where k 2 = e2α λˆ + 3α˙ 2 Rˆ µν 2 is an integration constant that is by definition the 4d cosmological constant (k 2 and λˆ may also be negative). 3 we obtain (to lowest order in derivatives with respect to r):   1 (4) = k 2 gˆµν .e. For simplicity. with [7]): (5) (4) = Rˆ µν − gˆµν e2α(r)(α¨ + 4α˙ 2 ) Rµν (a “dot” means d/dr). In the bulk region. the 4d metric gˆ µν parallel to the brane must still be Ricci-flat because of supersymmetry on the 4d slices parallel to the brane. the generalization is straightforward. So the equations for α have a one-parameter family of solutions. However. We make the metric ansatz ds 2 = dr 2 + e2α(r)gˆµν dx µ dx ν . where M 4 is the Minkowski space parallel to the brane. As for the brane region. In this ansatz. and can then be matched at their interface at r = lBrane-Susy .. the brane is a source of vacuum energy ρ of order (mBrane-Susy )4 that arises from the world-brane fields. Here we have included another r-dependent contribution λflux (r) that arises when the brane is a source of electric or magnetic flux. ν) and using the equation for the (r.C.

obtained by integrating over the compactification manifold M. i. so it only remains to compute the vacuum energy produced by the bulk supergravity fields: the gravitino.e.18]. these vacuum energy contributions cancel. k = 0. the dilaton. where derivatives with respect to r are no longer small and the supergravity approximation breaks down. . Higher-order corrections will make the differential equations for α more complicated. etc. let us assume that conditions are favorable and solutions exist. For this solution. and there may be regions in parameter space where no solutions exist [17. However. 1 4. But because of the small overlap of the wave functions of the supergravity fields with the brane. it remains to construct explicit examples. So the vacuum energy does not lead to a 4d cosmological constant. the mass scale mBulk-Susy of supersymmetry breaking in the bulk sector of the 4d effective theory will be suppressed with respect to the scale of 1 I thank E. since we know little about the region behind it. recently rediscovered in [6. breaking supersymmetry in the region of the bulk near the brane also breaks supersymmetry in the effective 4d theory. Witten for comments in the context of the singularity.. This is a generalization of the suggestion in [16] of “supersymmetry on the Planck brane” in the context of the Randall–Sundrum model. We have not discussed boundary conditions for α at the cutoff r = . For n extra dimensions the discussion is similar. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 the supersymmetric region requires that we pick the solution that is Ricci-flat in 4d. Again. This is the main new step taken in this paper and it will lead to our main results. We have already demonstrated that the ground state energy of the Standard Model fields does not contribute to the 4d cosmological constant. 6 and therefore does not curve the 4d metric parallel to the brane. As long as supersymmetry is unbroken in the bulk. the ground state energies h¯ ω with ω2 = k 2 + m2 2 of modes of light fields with momentum k should give a quantum mechanical contribution to the cosmological constant.7].390 C. antisymmetric tensor fields. This is the mechanism of Rubakov and Shaposhnikov [5]. As mentioned in the introduction. Now. whatever boundary conditions must be imposed — the assumption that they can be satisfied is part of the assumption that we have already made in the previous section: that there are consistent string compactifications that involve stable non-BPS branes and leave 4-dimensional supersymmetry unbroken away from the brane at the classical level. Supergravity at one loop Let us now go beyond the classical supergravity approximation. the vacuum energy on the brane is completely absorbed by the warp factor 1ˆ α˙ 2 = − λ(r). We have supplemented it by a matching condition at r = lBrane-Susy that picks out the solution with vanishing 4d cosmological constant without fine-tuning.

in this case the mPlanc suppression factor is simply due to the smallness of Newton’s constant. 3 A more precise calculation amounts to making the following corrections: N must be replaced in Eqs. ln mBulk-Susy Str MBulk 2 mBrane-Susy where “Str” denotes the supertrace and MBulk is the mass matrix in the bulk supergravity sector after supersymmetry breaking.1) mBrane-Susy Vol(M) mPlanck One way of seeing this is to consider a scalar field Φ in the supergravity multiplet and assume that it has a large mass of order mBrane-Susy inside the region where supersymmetry is broken: we take its (4 + n)-dimensional Lagrangean to be of the form ∂m Φ∂ m Φ + θ (rBrane-Susy − r)m2Brane-SusyΦ 2 . Integrating this Lagrangean over the compactification manifold.C. Already in the introduction we have discussed the relation (1.4) below for supersymmetry breaking in the brane and bulk sectors to 2–6 TeV and 0. the kinetic term acquires a prefactor Vol(M) while the mass term only acquires a prefactor Vol(B). This changes the predictions (5.1) between the momentum cutoff Λ in the sum over vacuum energies and the value of the cosmological constant λ. 2 So the hierarchy between the scales of supersymmetry breaking in the bulk supergravity sector and supersymmetry breaking in the Standard Model that lives on the brane is the same as the hierarchy between the scale of supersymmetry breaking on the brane and the Planck scale. In the case of N massless bosonic propagating degrees of freedom.     mBrane-Susy 2 mBulk-Susy 2 Vol(B) ∼ ∼ .3) and (4.1). 2π P λ is related to the Hubble expansion rate H0 of the universe by λ∼N (4. In a first estimate we may identify the cutoff Λ with the scale mBulk-Susy of supersymmetry breaking in the bulk. (4. its mass is Vol(B) .1). 3 Then Eq. After normalizing Φ to have a standard kinetic term. θ is the step function: θ (x) = 0 for x < 0 and θ (x) = 1 for x > 0. more details).5–5×10−3 eV. m2Φ ∼ m2Brane-Susy Vol(M) This implies relation (4. respectively (see the hep-th version of this paper for 2 = 0. (4. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 391 supersymmetry breaking on the brane by the same volume factor that we already found in (2.4) below by   1 −4 MBulk 4 .3) and (5.2) implies (converting lPlanck ∼ m−1 Planck ): λ = 3ΩΛ H02 2 The relation m ∼ Φ with ΩΛ ∼ m2Brane-Susy could also have been derived without reference to branes. the relation changes to Λ4 2 l . (4. This calculation also reveals the requirement Str MBulk .2) 2 3 being the value suggested by observation [2].

in order to insure that the running coupling constants . Being optimistic about them. Schmidhuber / Nuclear Physics B 585 (2000) 385–394  mBulk-Susy mPlanck 2 6ΩΛ π ∼ N  H0 mBulk-Susy 2 . Remarkably. Similarly as in the case of millimeter-size extra dimensions [10. with (4. (4.11].g.1) this yields the relation claimed in the introduction (where we have set N 1/2 ) by 1): ΩΛ = 2/3 and approximated ( 4π s !     mPlanck mPlanck N mPlanck 1 1 ≡ L.4) may even be somewhat (but not much) larger.4 to 4. The lower experimental bound on the gravitino mass appears to be only 10−5 eV [8]. so r N ∼ 0. Their contributions to λ have opposite sign and depend on their masses after supersymmetry breaking.. etc. p √ λ = 3ΩΛ H0 ∼ 2 × 10−33 eV.4) The example with the intermediate value of L = 15. e.1) (5. the predicted scale of supersymmetry breaking in the Standard Model is roughly where it is expected to be.4) = log = log log mBrane-Susy 2 mBulk-Susy 4 6ΩΛ π H0 5. This corresponds to a Compton wavelegth of the order of a fraction of a millimeter.2 is plotted in Fig.3) Together.2) What should we insert for N ? Type IIB supergravity multiplets. Without going into details. The effects of these new fields might show up in short-distance measurements of gravity in the µm range in the near future [19]. 2. 2π Since there may be other hidden factors of π. So we expect a mass of order 10−3 –10−2 eV for the gravitino or at least for some members of the supergravity multiplet. The numbers Let us now plug in the numbers. This yields the predictions mBrane-Susy ∼ 4–10 TeV. it seems safe to assume that the effective N is somewhere between 1 and 128. the presence in the bulk of gravitinos or dilatons in the micrometer range is not ruled out by experiment: while the brane physics (the Standard Model) has been probed down to the weak scale. 1/2. We use mPlanck ∼ 1019 GeV. the bulk physics (gravity) has only been probed down to centimeter scales.392 C. (4. we infer that L ∼ 15.3) mBulk-Susy ∼ 10−3 –10−2 eV. (5. (5. the actual errors in the relation (4. have 128 bosonic and 128 fermionic degrees of freedom.2. that were missed by our crude analysis.2 ± 0. (5.

if we assume a probable scale of supersymmetry breaking between 1 and 100 TeV. a dilaton or other supergravity fields with masses of 10−3 –10−2 eV. Our explanation for the small cosmological constant can be tested by searching for signs of a gravitino. Note After this paper appeared. We can thus look forward to a number of surprises in future tests of Einstein gravity in the micrometer range. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 393 Fig. (Note however that we discuss four-dimensional supersymmetry and the four-dimensional cosmological constant on slices .22]. meet in supersymmetric Grand Unification. 2. This is very nontrivial. provided that the scale of supersymmetry breaking in the Standard Model is in the TeV range. a priori it could have come out many orders of magnitude off the mark.2. The relation between hierarchies for the intermediate value L = 15.C. I was notified by the authors of [23] that they mention on page 5 the possibility of using infinite volume extra dimensions to suppress supersymmetry breaking in the bulk and the bulk cosmological constant. 6. It remains to construct explicit examples of such compactifications and to show that they are consistent and stable. then we can predict the value of the cosmological constant within a few orders of magnitude of the value that seems to have been measured! Let us finally note that our derivation and results apply just as well to the case of a single extra dimension as in the Horava–Witten model [20] or in the Randall–Sundrum model [21. Reversing the logic. Conclusion It seems that the proposal that we live on a non-supersymmetric brane that is embedded in a supersymmetric higher-dimensional string compactification can explain the observed small value of the cosmological constant.

Non-BPS states and branes in string theory. Verlinde.L. Dimopoulos. hep-th/9912018. 83 (1999) 3370–3373. [14] L. J. hep-th/0004003. Science 284. where non-supersymmetric branes inside a supersymmetric bulk are constructed. Phys. Kaloper. S. [7] C. Dimopoulos. S. Novel local CFT and exact results on perturbations of N = 4 super Yang–Mills from ADS dynamics. Mod. Phys. R. A small cosmological constant from a large extra dimension. hep-th/0001197. Dvali. New dimensions at a millimeter to a Fermi and superstrings at a TeV. Lett. An alternative to compactification. B 436 (1998) 257. Phys. hep-ph/0002297. Arkani-Hamed. Schmidhuber / Nuclear Physics B 585 (2000) 385–394 parallel to the brane. A. Porrati. [19] J. [24]. G. [2] N. Dvali. Schmidhuber. Rubakov. H. This work is supported in part by a Heisenberg fellowship of the DFG.H. [13] A. Verlinde. Porrati. Raman Sundrum. [17] S. M. [21] Lisa Randall.Z. Nucl. Chan. hep-th/9904017. F. Ostriker.P. Dvali. hep-th/0001206. M. hep-th/9912156. Freedman. Silverstein. S. Brignole. [15] D. [11] I. Girardello. [12] H. Phys. hep-th/9510209. Dynamical compactification as a mechanism of spontaneous supersymmetry breaking. Perlmutter and P. M. Lerche and P. Signals of a superlight gravitino at hadron colliders when the other superparticles are heavy. Antoniadis. [5] V. Pilch. J. gravity and the cosmological constant. B 480 (2000) 193. A 10 (1995) 1247. Gabadadze. Bahcall.A. [6] E. B 125 (1983) 139. [8] A. RG flow. B 526 (1998) 136. Phys. hep-th/9906182. The cosmological constant problem.C. Dudas. 1481. [20] P. J. Holography and compactification. Extra space–time dimensions: towards a solution of the cosmological constant problem. B 429 (1998) 263. M. K. Self-tuning flat domain walls in 5d gravity and string theory. Zwirner. The hierarchy problem and new dimensions at a millimeter. Feruglio. Phys. Antoniadis. N. JHEP 9812 (1998) 022. [9] G. Experimental status of gravitational-strength forces in the sub-centimeter regime. S. Rev. G. E. .S. Mangano. N. Shaposhnikov. Nucl. Steinhardt. S. Phys. E. Warner. Schulz. Rev. Kachru. hep-th/0002190. [4] E. Petrini. Sen. [10] N. Supersymmetry at large distance scales.394 C. [3] E. Metastable gravitons and infinite volume extra dimensions.C. Phys. Brane supersymmetry breaking. B 539 (1999) 93.J. Lett. A large mass hierarchy from a small extra dimension. M. N. Arkani-Hamed. Verlinde. Phys. Lett. Raman Sundrum. Witten. Mayr for conversations. [16] H. Horava. Arkani-Hamed. B 464 (1999) 38. Gubser. Mod. References [1] S. E. Lett. Renormalization group flows from holography: supersymmetry and a C-theorem. AdS(5) and the 4d cosmological constant. Heterotic and type I string dynamics from eleven dimensions. Weinberg. Dvali. Witten. Is supersymmetry really broken?. Sundrum.P. hep-th/9906064. [24] I. hep-th/9904207. M. Phys. A. Dimopoulos. Phys.E. M. Lett. The cosmic triangle: revealing the state of the universe. [18] N. Shifman. [23] G. Int. Acknowledgements I thank W. F. Phys. Lett. Nucl. Lett. 61 (1989) 1.W.) I also became aware of Ref. Witten. B 460 (1996) 506. G. The cosmological constant from the viewpoint of string theory. Sagnotti. [22] Lisa Randall. B 504 (1997) 127. Zaffaroni. Price. Long.

Nuclear Physics B 585 (2000) 395–404 www. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Pittsburgh.3]. and/or a potential for the dilaton field.  2000 Elsevier Science B. Assuming a simple double-well potential.nl/locate/npe Naked singularity and Gauss–Bonnet term in brane world scenarios Ian Low a. Naked singularity 1. USA b Department of Physics.7 . we consider adding to the action higher derivative terms. Here we explore the question of how the solution of the cosmological constant problem might be modified by the addition of these terms. 04. accepted 6 June 2000 Abstract We add a Gauss–Bonnet term to the Einstein–Hilbert action and study the recent proposal to solve the cosmological constant problem. our 4-dimensional world is embedded in a 5-dimensional universe and remains flat in the presence of an arbitrary vacuum energy density V .V.-i Keywords: Extra dimension. PACS: 11. A simple anti-de Sitter solution and inflationary solutions to the action with Gauss–Bonnet term were studied in [10. in particular the Gauss–Bonnet term. In [2. Santa Barbara.elsevier. Recently. A. Introduction It was suggested long time ago [1] that the cosmological constant problem could be solved if one is willing to go to higher dimensional spacetime.10.V. we obtain first order Bogomol’nyi equation as a solution-generating method in our scenario. When the coefficient of the Gauss–Bonnet term is positive. we find the dilaton field to be a kink in the fifth dimension.3 2 1 3 ( 0 0 ) 0 0 3 5 2 . Models with gravity coupled to scalars are considered in [12–18]. USA Received 19 April 2000. In the absence of supersymmetry.50. revised 9 May 2000. CA 93106. Zee a a Institute for Theoretical Physics. University of California. All rights reserved.11]. All rights reserved. PII: S 0 5 5 0 . the dilaton potential is bounded below.Kk. We also consider the possibility of adding a dilaton potential to the action.+h. 12. In this note. a number of authors [2–6] have outlined possibilities for solving this long standing problem within the brane world scenario [7–9]. and in the context of five dimensional supergravity as well [19–24]. The price one pays for this remarkable phenomenon is the appearance of naked singularities in the 5-dimensional universe.60. Carnegie Mellon University. Gauss–Bonnet. PA 15213.b .

RMN = RMPN . . (7) Eq. Low. +). 3. The metric gµν = δµM δνN GMN (y = 0) is the 4-dimensional metric on the M = +0 M P M brane. 1. Here M. 1 4 0 2 ϕ + A00 = − V δ(y). (5) is the difference between the µµ and 55 components of Einstein’s equation. We will follow [2] which we will refer to as KSS.396 I. 3. The 5-dimensional Bianchi identity states that only . . A. +. . KSS showed. 2. ν = 0. 2. let us review the proposed solution reduced to its simplest version. 9 6 2 1 2 A0 − ϕ 0 = 0. N = 0. that for any V there exists a solution to (2) and (3) with the metric taking the form  (4) ds 2 = e2A(y) −dt 2 + dx12 + dx22 + dx32 + dy 2. Zee / Nuclear Physics B 585 (2000) 395–404 To introduce the subject and to set the stage for our discussion. +. (6) is the 55 component (which does not receive a brane contribution proportional to δ(y)). this is possible because the equations of motion are solved separately for y > 0 and y < 0 and we have enough arbitrary integration constants that we can adjust in order to get the flat 4-dimensional world (4) we want. respectively. KSS referred to this as a self-tuning solution of the cosmological constant problem. and (7) is the dilaton equation of motion. The equations of motion are. Arithmetically. In other words. The equation of motion for the dilaton field is given by ∇ 2 ϕ = 0. (We use the convention in which RNPQ NQ. 5 and µ.P . R = R M . +. (3) A theorist living in the 4-dimensional brane world would notice that the action contains a cosmological constant given by the vacuum energy density V .) Einstein’s equations read   1 4 1 2 RMN − GMN R − ∇M ϕ∇N ϕ − GMN (∇ϕ) 2 3 2 r 1 g µ ν V gµν δM δN δ(y) = 0 (2) + 2 G with g and G the determinant of gµν and GMN . where 0 denotes differentiation with respect to y. 9 (5) (6) and ϕ 00 + 4A0 ϕ 0 = 0. and the signature (−. 1. the 4-dimensional world is flat in spite of the presence of V . 3 Gravity and a dilaton field ϕ live in the 5-dimensional world and is coupled to a thin 4-dimensional brane whose position is taken to be at y = 0. The dilaton field adjusts itself so that this solution exists. The low energy effective action is taken to be   Z Z √ √ 4 (1) S = d 5 x −G R − (∇ϕ)2 + d 4 x −g (−V ). however.

for example the Ricci scalar R = −20(A0)2 − 8A00 diverge. of the form ds 2 = e2A(y)gµν (x)dx µ dx ν + dy 2 we can see easily that √ For metrics2A(y) √ −G R = e −g R(4) + · · · . (13) 4 y+ y− These two equations merely fix y+ and y− in terms of V . The heavy price that one pays in the KSS solution is the appearance of naked singularities at y+ and y− . Thus. KSS obtained a flat space solution. Low. Thus. Near y+ for example. y+ and y− are integration constants. where the 4-dimensional scalar curvature R(4) is constructed outR of gµν (x). for y+ > y > 0 3 (8) ϕ(y) = − log(y+ − y) + d+ 4 and 1 (9) A(y) = log(y+ − y) + a+ . the metric components g00 = gii = √ e2A(y) vanish like y+ − y.I. the cosmological constant problem is solved because we have a lot of integration constants. the framework of embedding our universe in a larger universe is of crucial importance. d+. 4 Here a+ . For example. Recently. 0 y− (14) . A. Various curvature invariants. 4 and for y− < y < 0 3 (10) ϕ(y) = log(y + y− ) + d− 4 and 1 (11) A(y) = log(y + y− ) + a− . Solving (6) for A0 and inserting into (7) we see immediately that ϕ is given by the logarithm of y. the solution may be chosen to be. type I in their classification. The continuity of ϕ and A at y = 0 determines d+ and a+ in terms of the other constants. Integrating (5) and (7) across y = 0 gives the jump conditions that     1 1 1 1 (12) − − =− V A0 y = 0+ − A0 y = 0− = 4 y+ y− 6 and     3 1 1 0 + 0 − − ϕ y =0 −ϕ y =0 = = 0. for any V . KSS were forced to choose y+ and y− positive. the effective 4-dimensional Planck mass squared is proportional to dy e2A(y) . in some sense. That some types of naked singularity are not acceptable is known as cosmic censorship. d− . In order to have a finite 4-dimensional Planck mass. (5) and (6) imply (7). They (rather arbitrarily) postulated that the universe is cutoff at y+ and y− and thus obtained a finite Planck mass squared proportional to Z dy e 2A(y) Zy+ Z0 2A(y) = dy e + dy e2A(y). a−. One can say that. Gubser [25] has studied in detail various singularities. These singularities are not clothed by an event horizon. Zee / Nuclear Physics B 585 (2000) 395–404 397 two of the three equations are independent. Thus. Of course.

one has to adjust V to have a value determined by Λ. this is not called fine tuning. For example. At this point. it does affect the 5-dimensional universe. in the sense that the 4-dimensional world remains flat regardless of the value of V . We should perhaps mention here that KSS showed that more generally we can write f (ϕ) instead of V and the same conclusion continues to hold. We replace the bulk action in (1) by . it may be useful to say a few words about fine tuning and self tuning. In particular. the 5-dimensional action (1) could perfectly well contain the term −Λ eaϕ corresponding to a bulk cosmological constant Λ. has already committed fine tuning according to the strong definition by excluding all sorts of possible terms from (1). starting with the action (1). which KSS set to 0 in their self-tuning solutions. in the language of this subject. A traditionalist would say that KSS. then we are said to have fine tuned. we would regard the KSS solution as not a fine tuning solution. it moves the naked singularities around. one is to write down in a field theory Lagrangian all terms with dimension less than or equal to 4 allowed by symmetry. The coefficients of all such terms are to be regarded as free parameters. V is replaced by f (ϕ) and in (7) the term + 38 δ(y) ∂ϕ bϕ A particularly popular choice is f (ϕ) = V e as inspired by string theory. We feel that it is necessary to formulate a weak definition of fine tuning. 2. Thus. If we arbitrarily set one coefficient equal to another or to zero without being able to invoke a symmetry principle. KSS called this fine tuning. the self tuning feature of their solution is lost. Zee / Nuclear Physics B 585 (2000) 395–404 Once these integrals are thus cutoff to give finite values. A traditionalist using the strong definition of fine tuning would definitely call setting Λ to 0 fine tuning. Presumably. Traditionally. KSS described their solution as self tuning. however. then as KSS themselves showed. this action is to be regarded as the low energy effective action of some more fundamental theory such as string theory. This is easy to see: in (5) ∂f is added to the right hand side. A. Low. At the same time. This completes a necessarily brief review of KSS to which we refer for further details. of course we still have the R0 R∞ highly non-trivial constraint that 0 dy e2A(y) and −∞ dy e2A(y) have to be finite. we can obtain any desired value for the 4-dimensional Planck mass squared simply by shifting the additive integration constant allowed by (9) and (11) in the solution for A(y). according to which one can exclude or include any sorts of terms from the higher dimensional action without being accused of fine tuning. Action with Gauss–Bonnet term We ask whether we could avoid naked singularities by adding higher derivative terms such as R 2 to the Einstein–Hilbert action. We refer to this as the strong definition of fine tuning. Notice that while changing V does not appear to affect our 4-dimensional universe. Rather. To obtain a flat 4-dimensional world. Here we are dealing with a non-renormalizable 5-dimensional action (1) which nobody would regard as fundamental.398 I. If Λ is not equal to 0. as we think anybody would. which remains flat whatever the value of V .

(16) agrees with [31].P . 3 (15) We have also included a potential V(ϕ) for the dilaton field. Varying Sbulk with respect to the metric tensor we obtain the equation of motion    1 1 RMN − GMN R − GMN aR 2 − 4bR PQRPQ + cR PQST RPQST 2 2 PQS + 2aRRMN − 4cRMP R PN + 2cRMPQSRN + 2(a − b)GMN R . For our purposes.) We will choose a = b = c = λ so that the added term λ(R 2 − 4R MN RMN + MNPQ RMNPQ ) is of the Gauss–Bonnet form. We will proceed in stages. (16)   4 1 1 2 = ∇M ϕ∇N ϕ − GMN (∇ϕ) − GMN V(ϕ). 9 and ∂f 3 (ϕ). 3 2 2 (17) These higher derivative terms naturally arise in the low energy effective action of string theory [26–31]. In higher-dimensional spacetime. One sees from the equation of motion that R all terms involving fourth derivative vanish in this case and (16) reduces to the result given in [32]. Zee / Nuclear Physics B 585 (2000) 395–404 Z Sbulk = d 5x 399  √ −G R + aR 2 − 4bR MN RMN  4 + cR MNPQ RMNPQ − (∇ϕ)2 − V(ϕ) . Then in Section 3 we will include V(ϕ). With the ansatz for the metric in (4). We proceed in an exploratory spirit we believe appropriate for this stage of development of this nascent subject and do not apologize further for this specific choice.I. We will first study the effect of the higher derivative terms with V(ϕ) = 0. (For a = b = 0.P − 2(a − 2b + c)R. y = 0. (20) ϕ 00 + 4A0 ϕ 0 = δ(y) 8 ∂ϕ The matching condition at the location of the 3-brane. (18) 9 6 2 1 2 4 (19) A0 − ϕ 0 − 2λ A0 = 0. Low.P = TMN . we do not inquire of their deeper origin but simply treat the action as “phenomenological”.P where TMN + 4(2b − c)R PQ RMPQN  .N − 4(b − c)RMN . we obtain 2  1 4 0 2  ϕ + 1 − 4λ A0 A00 = − f (ϕ) δ(y). A. it is not a topological invariant but nevertheless has particularly attractive properties. the Gauss–Bonnet combination is a topological invariant in 4-dimensional spacetime.M. As is well known. as discovered by Zwiebach [33] and explained by Zumino [34]. becomes .

+  ∂f 8 0 .

.

0 ϕ (y).

3 ∂ϕ 0− (21) . = ϕ(0) .

Zee / Nuclear Physics B 585 (2000) 395–404  . Low. A.400 I.

+  4 0 3 .

.

0 0 −6 A (y) − λA (y) .

for y > 0 say. For λ > 0. (30) 3 In this case. only the plus sign in (25) is allowed and we have      1 ψ ψ 2 tan−1 tanh − csch . For λ = −|λ| < 0. we must have e8A > (8d 2/9) λ and hence the 4-D Planck mass squared is infinite unless spacetime is cut-off by some singularity. The scalar equation (20) gives ϕ 0 = d e−4A(y) with d an integration constant. To solve (24). These four solutions correspond to the four different choices of signs in (25). In order to have finite 4-D Planck mass. = f ϕ(0) . Again. we change variable to κ(y) = e−4A = 1/Gt t 2 and (24) becomes s ! 2  4κ 8 d2 2 0 2 1± 1− λκ . After solving two equations we can use the third one as a convenient check. the naked singularity persists if we demand finite 4-D Planck mass. sinh ψ = . for λ > 0. 3 0− (22) and the continuity condition for ϕ(y) and A(y) at y = 0. In the limit λ → 0 these equations reduce to the ones studied in KSS of course. In conclusion. We solve these equations in the bulk. there are again naked singularities in the fifth dimension. (29) y(κ) = y0 ± √ 4 2 8|λ| where 2dp 2|λ| κ. y(κ) = y0 ± √ 4 2 8λ (26) (27) where sin θ = 2d√ 1 2d√ 2λ κ = 2λ 2 . the Bianchi identity assures us that only two out of three bulk equations are independent. y(κ) = y0 ± √ 4 2 8λ     θ θ 1 log tan − sec . we obtain     θ +π θ 1 log cot − csc . 3 3 Gt t (28) 0 6 θ 6 π/2 and y0 is an integration constant. Inserting this into (19) we obtain s !  1 8 d 2 −8A 0 2 1± 1− λe . A = 4λ 9 (23) (24) From (24) we see immediately that. while we still have the nice self tuning solution of the cosmological constant problem as in KSS we are also stuck with the naked singularities. (25) κ = λ 9 which can be readily solved.

A potential can arise in string theory from higher order corrections.36]. these equations have been studied previously and a solution-generating method inspired by supergravity was suggested in [18. It is shown that.  9 ∂W (ϕ) . A. if V(ϕ) takes the special form   27 ∂W (ϕ) 2 − 12W (ϕ)2 . The equations of motion are modified to 2  1 4 0 2  ϕ + 1 − 4λ A0 A00 = − f (ϕ) δ(y). by counting the number of integration constants one can show that the solution space of (34)–(36) coincides with the solution space of the equations of motion following from (31)–(33) for λ = 0 [18]. 4 ∂ϕ A0 = −W (ϕ). Zee / Nuclear Physics B 585 (2000) 395–404 401 3. (40) ϕ 0 = 1 − 4λ W (ϕ)2 4 ∂ϕ In terms of the superpotential W (ϕ). we obtain similar first order equations for nonzero λ:     2 3 3 27 ∂W (ϕ) 2 1 − 4λ W (ϕ)2 − . is now . 8 ∂ϕ 8 ∂ϕ (33) When λ = 0. (34) V(ϕ) = 4 ∂ϕ then a solution to 9 ∂W (ϕ) .35. (37) A00 = − 4 ∂ϕ From (31) and (32). y = 0. ϕ0 = (35) (36) is also a solution to the equations of motion. W (ϕ) is sometimes called the “superpotential” for obvious reason. staying in the bulk and thus ignoring δ(y) for now. for λ = 0.I. 9 12 (31) (32) and ϕ 00 + 4A0 ϕ 0 = ∂f 3 ∂V(ϕ) 3 + δ(y) (ϕ). 9 6    1 1 2 2 4 A0 − ϕ 0 − 2λ A0 = − V(ϕ). but again in the “phenomenological” spirit of this paper we do not concern ourselves with its origin. (38) + V(ϕ) = 4 ∂ϕ 2λ 2λ (39) A0 = −W (ϕ). though no supersymmetry is involved here. Low. Note that in this case   9 ∂W (ϕ) 2 6 0. Furthermore. the matching condition at the location of the 3-brane. Adding a dilaton potential We next explore what happens if in the action we add a potential V(ϕ) for the scalar field.

Low.402 I. A. Zee / Nuclear Physics B 585 (2000) 395–404  .

y=0+ .

 4 ∂ ∂f 3 .

W (ϕ(y)) − λ W (ϕ(y)) .

ϕ(0) . 6 = ∂ϕ 3 ∂ϕ y=0−  .

y=0+ .

 4 = f ϕ(0) . 6 W (ϕ(y)) − λW (ϕ(y))3 .

.

 ∂W (ϕ) ∂W (ϕ) ∂W (ϕ) 2 · . The reason that we are able to obtain the first order Bogomol’nyi equations in the presence of higher derivative terms is that we have chosen the particularly nice Gauss– Bonnet combination. For our purposes we regard the method as simply a method for solving coupled differential equations. in a specific model. We now choose W (ϕ) = s ϕ to be a linear function of ϕ. Generalization to n scalar fields ϕ = (ϕ1 . 3 y=0− (41) (42) If. we generate the solution for y > 0   y − y+ 3 . Note that if we did not use the Gauss–Bonnet combination we would have third and fourth derivatives of A appearing in (31). and (37) is used to prove a c-theorem [37]. In five dimensional gauged supergravity. The second derivative of A now becomes    9 ∂W (ϕ) 2 00 1 − 4λW (ϕ)2 (45) A =− 4 ∂ϕ and is no longer to be non-positive always. these equations reduces to (34)–(36). Note that the dilaton potential V(ϕ) is now bounded below if λ is positive. . Following the steps outlined above. (46) ϕ(y) = √ σ tanh l 2 2    y − y+ 1 + k+ (47) A(y) = − σ 2 log cosh 2 l and for y < 0   y − y− 3 . ϕ(y) = ε √ σ tanh l 2 2    y − y− 1 2 + k− . It is simple to check that (33) is satisfied automatically. Thus. A(y) = − σ log cosh 2 l (48) (49) . ϕn ) is achieved by the following replacement:  W (ϕ) → W (ϕ). It would be interesting to study the implications of (38)–(40) in the context of these models. → ∂ϕ ∂ϕ ∂ϕ (43) (44) In the limit λ → 0. the 3-brane tension f (ϕ) is given by 12(W − 43 λW 3 ). . (35) and (36) arise as conditions for unbroken supersymmetry. Note that s has√dimension of an inverse length. for a given choice of W we generate a solution for some V(ϕ). . We find it√convenient to define the length scale l = 2/(9 λ s 2 ) and the dimensionless ratio σ = (l/ λ )1/2 . these jump conditions can be satisfied identically. .

Low. with negative λ the hyperbolic functions in (46)–(49) turn into trigonometric functions and we again have naked singularities in the bulk. as can be seen from (45). In these two examples we constructed. take f (ϕ) = V for simplicity.L. in this particular example. The crucial point here is that we no longer have the freedom of including an additive constant in the solution for ϕ in this example. Steve Gubser. We find √ that. Acknowledgements We thank Lisa Randall and Eva Silverstein for informative discussions and for encouragement. in the presence of Gauss–Bonnet term. is not true in our case. also benefitted from discussions with Aki Hashimoto and Michael . The continuity of A fixes k+ = k− while (21) tells us that the jump in [1 − 43 λ(A0 )2 ]A0 across y = 0 is equal to − 16 V . the self tuning feature of KSS is also lost. in the scenario of [2. A00 6 0. and Joseph Polchinski for helpful comments. and David Gross. we would like to point out that a crucial ingredient of the proof given in [38] . we need either fine tuning to avoid the naked singularities or naked singularities to maintain the self tuning feature. the 4-D Planck mass is finite. A. by choosing λ < 0 and s = 2/(9|λ|). Unfortunately. Moreover. But this cannot be if V is not zero since A(y) = − 12 σ 2 log cosh 1l (y − y∗ ) is perfectly smooth across y = 0. are consistent with this no-go theorem. the dilaton potential is a constant V(ϕ) = 3 2|λ| (51) which acts like a bulk cosmological constant and is independent of r in the superpotential W (ϕ). The continuity of ϕ and ϕ 0 fixes ε = +1 and y+ = y− ≡ y∗ . The ϕ solution we have is a kink in the fifth dimension interpolating two vacua of the potential. Gary Horowitz. I. Thus r plays the role of an integration constant and we recover the self tuning feature in KSS. Although the two examples we constructed here. We can also choose a more general superpotential of the form W (ϕ) = s ϕ + r. Note added After this paper was submitted a paper [38] appeared in which the authors proved a no-go theorem which states that. one needs either fine-tuning or naked singularities to achieve the flatness of our universe. The spacetime is asymptotically AdS and there is no singularity at all. Zee / Nuclear Physics B 585 (2000) 395–404 The potential is a familiar double well    8 ϕ2 2 3 3 1 +1 1− − . To see this.I. Unfortunately. V(ϕ) = 2λ σ 2 9 σ2 2λ 403 (50) Note from (33) that we have freedom in choosing the sign of ϕ represented by ε = ±1 in (48).3]. Therefore there might still be hope of retaining self-tuning feature without invoking naked singularities.

M. M. hep-th/0001197. Irges. D.M. N. B 125 (1983) 139.J. E.C. D 60 (1999) 123506. Lett. R. hep-th/0002091.K. Kaloper. D 36 (1987) 392. Antoniadis. Meyers. JHEP 9911 (1999) 010. hep-th/9909058. hep-th/0002174. A. Wise. C. N. DeWolfe. B 277 (1986) 1. D.A. B 293 (1987) 385.E. S.R. Lee.B. Chambin. M. C. Kim. Grojean. Skenderis.G. S. S. J. B 311 (1988) 673. de Alwis. C. Dvali. hep-th/0004133. Lett. N. hep-th/9912060. hep-ph/9912344.L. hep-th/9910081. Gubser. de Boer. hep-th/0001206. Witten. J.S. Arkani-Hamed. Nucl. Phys.C. I. S.E. Lett. B 468 (1999) 46. Warner. H. Lukas. Tseytlin. L.G. hep-th/9904017. P.R. Callan. Perry. Dimopoulos. Linde. K. G. S. B 291 (1987) 41. L.J. Sundrum. O. Katz. 83 (1999) 4922. Grojean. 137 (1986) 109. E. Cline. Lett. C. Randall. Metsaev. Phys. B 156 (1985) 315.S. Chen. M. R. B 429 (1998) 263. R. J. Gubser. J. M. E. Zee / Nuclear Physics B 585 (2000) 395–404 Quist. hep-th/0003067. Rev. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] V. S. 55 (1985) 2656. 83 (1999) 4690. Rev. Kaloper. This work was supported in part by the NSF under grant number PHY 89-04035 at ITP and by Department of Energy under grant number DOE-ER-40682-143 at CMU. Stelle. J. W. A. Shmakova. A. hep-th/9909134. Cvetic. Brandhuber. Sundrum.W. Phys. Lett. Silverstein. M. C. J. S. Townsend. J.P. Nucl. A. R. Bakas.S. Phys. B. Gremm. Lett.P.404 I. Rev. Phys. Nucl. D 59 (1999) 086001. Phys. 83 (1999) 3370. Boulware. Karch. H. G. Phys. Goldberger. D. B. S. Schulz. Behrndt. Ovrut. Y.M. D. Metsaev. Csaki. hep-th/0004005. Phys. Gross. Rev. Rev. N. Dvali. Csaki. Phys. .J. Randall. E. Phys. K. Luty. Phys. S. I. Kallosh.Z. Lee. is supported in part by an ITP Graduate Fellowship. Hollowood. A. B. K. J. E. Sfetsos.J. N. de Alwis. Kyae. T.B. Waldram. Rev. Flournoy. M. Servant. hep-th/9909130. Phys. Lett. hep-th/9912132. B 436 (1998) 257. N. Rubakov. R. hep-th/0002160.Z. Phys.A. Erlich. Gubser. G. hep-th/0004125. Giddings.P. N. Freedman. Arkani-Hamed. Shaposhnikov. Zumino. Nucl. Rep. Phys. Zwiebach. Hollowood. H. D. Arkani-Hamed. hep-th/9912012. K. T. Erlich. Gibbons. Verlinde. A. D. Deser. Phys. Gross. Phys. Kachru. Ponton.D. Dimopoulos.T. Sloan. I. Meyers.H. Verlinde. Shirman. hep-th/0001033. Lett. Lett.A. Tseytlin. R. Phys. Rev. D 60 (1999) 107505. M. S. Freedman. B. Lett. Plich. Rev. S. B. Low. G. Dimopoulos. Phys. Kyae. M. K. R.W. Gremm. Kim. hep-th/0002040. A. Phys. A. A. B 191 (1987) 354.

An obvious physical reason is that in many cases a general field can be treated as a constant one to a good approximation. e ab s sinh(eas) sin(ebs) 0 E-mail address: schubert@lapp. Quantum electrodynamics.V. All rights reserved. PII: S 0 5 5 0 .20. BP 110. the constant field is distinguished by being one of the very few known field configurations for which the Dirac equation can be solved exactly. All rights reserved.fr (C. PACS: 12.  2000 Elsevier Science B.in2p3. In the first part of this series. The two-point cases are calculated explicitly.Nuclear Physics B 585 (2000) 407–428 www. Constant field. we give a detailed exposition of the method for the case of the QED one-loop N-photon amplitude in a general constant electromagnetic background field. Introduction: QED processes in constant electromagnetic fields Processes involving constant electromagnetic fields play a special role in quantum electrodynamics. In QED this is expected to be the case if the variation of the field is small on the scale of the electron Compton wavelength. leading to compact representations for the constant field vacuum polarisation tensors for both scalar and spinor QED. 1 L= 8π 2 Z∞ ds −ism2 2 cosh(eas) cos(ebs) e . Chemin de Bellevue.elsevier. accepted 26 June 2000 Abstract The string-inspired technique is used for the calculation of vacuum polarisation tensors in constant electromagnetic fields.15. France Received 31 January 2000.-m. String-inspired 1.3 2 1 3 ( 0 0 ) 0 0 4 2 3 . allowing one to obtain results which are nonperturbative in the field strength.V. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.Bt Keywords: Vacuum polarization. F-74941 Annecy-le-Vieux Cedex. An early and well-known example is the Euler–Heisenberg Lagrangian [1–3].1) . Schubert).nl/locate/npe Vacuum polarisation tensors in constant electromagnetic fields: Part I Christian Schubert Laboratoire d’Annecy-le-Vieux de Physique Théorique LAPTH. 11.5 (1. Mathematically. the one-loop QED vacuum amplitude in a constant field.

This subject was then over the years investigated by a number of authors. Moreover. Nevertheless. i. The recent [24] contains another recalculation of this quantity. in this formalism the inclusion of constant external fields turned out to require only relatively minor modifications [41–45]. Shortly later Toll [11] initiated the study of the effect of a background field on the one-loop photon propagator. For QED calculations in constant external fields it is possible and advantageous to take account of the field already at the level of the Feynman rules. Moreover the modification of the photon dispersation relation through the background field can. This Lagrangian encodes the information on the low energy limit of the one-loop photon S-matrix in a form which is convenient for the derivation of nonlinear QED effects such as photon–photon scattering and vacuum birefringence [4–6]. For the case of a magnetic field this process was calculated in the low photon energy limit in [25] and for general photon energies in [26]. for example. photon splitting may be observable in the laboratory in the near future [30. 33]. However.40]. Suitable formalisms have been developed decades ago [21. Schubert / Nuclear Physics B 585 (2000) 407–428 where a 2 − b2 ≡ B 2 − E 2 . beyond the simplest special cases they lead to exceedingly tedious and cumbersome calculations. the photon splitting process is believed to be of relevance for the physics of neutron stars which are known to have magnetic fields approaching.e. A parallel line of work led to the formulation of analogous computation rules for quantum gravity [36.10] allowed him to reproduce this result. the “critical” magnetic field strength Bcrit = m2e /e = 4. It seems also not impossible that.37]. Their work led to the formulation of new computation rules for QCD amplitudes which made it. as well as a detailed analysis of the implications for light propagation. first for the pure magnetic field [12–17] and crossed field cases [18. The vacuum polarisation tensor in a general electromagnetic field was first obtained in [20] and given in more explicit form in [21–23]. Later it was found that even in abelian gauge theory significant improvements over standard field theory methods can be obtained along these lines [39. the first measurement of this effect in the laboratory seems now imminent [8]. and even surpassing.408 C. The relation between the string-derived rules and ordinary Feynman rules was clarified in [38]. Schwinger’s ingenious use of Fock’s proper-time method in 1951 [9. to absorb it into the free electron propagator. ab ≡ E · B.. due to recent improvements in laser technology. with considerably less effort. feasible to perform a complete calculation of the one-loop five gluon amplitudes [35].41 × 1013 Gauss [27– 29]. Vacuum birefringence is a subject of actual interest [7] since.31]. as well as the analogous one for scalar QED. A different and more efficient formalism for such calculations has been developed during the last few years. Another consequence of the presence of a background field is the invalidation of Furry’s theorem. already the three-photon amplitude is non-vanishing in a constant field. depending on the photon polarisations.19].32. The idea of using string theory methods as a practical tool for calculations in ordinary quantum field theory was advocated by Bern and Kosower [34]. using the so-called “string-inspired” technique. For this reason it has been extensively applied to constant . The amplitude turns out to be very small for the magnetic field strengths presently attainable in the laboratory. with some further improvements in laser technology. lead to the opening up of phase space for the photon splitting process γ → γ + γ .

T → is. F This will allow us to arrive at explicit results in a manifestly Lorentz covariant way. The one-loop effective action due to a scalar loop for a Maxwell background can (in modern notation) be written as 1 Z∞ Z RT 1 2 dT −m2 T ˙ ))) e Γscal [A] = Dx(τ ) e− 0 dτ ( 4 x˙ +iex·A(x(τ . 2. This includes a recalculation of the photon-splitting amplitude [46] as well as high order calculations of the derivative expansion of the QED effective action [42.47–49]. F 4i → F 0i = Ei . F 4i = −iEi . 3. As a first step in any evaluation. For fixed T Dx denotes an integral over the space of all closed loops in spacetime with periodicity T . (2. Schubert / Nuclear Physics B 585 (2000) 407–428 409 field processes in QED. As a technical improvement on [45] we derive a decomposition of the generalised worldline Green’s functions GB. In Section 3 we extend this formalism to the inclusion of constant external fields along the lines of [45]. Section 5 contains our conclusions.e. 2. The QED one-loop N -photon amplitude in vacuum In [39] it was shown that. ε1234 → iε1230 . For other work on QED amplitudes similar in spirit to the string-inspired approach see [58–63].43. k 4 → −ik 0 . In Section 4 we apply the formalism to a calculation of the scalar and spinor QED vacuum polarisation tensors in a constant field. diag(− + + +).50–52] was applied to a calculation of the two-loop correction to the Euler–Heisenberg Lagrangian. for the QED case.. j = 1. F. F 2 with coefficients that are functions of the two standard Maxwell invariants. This is done by fixing the average position of the loop. in Section 2. 2. a detailed and self-contained exposition of the string-inspired technique for the calculation of one-loop scalar/spinor QED photon amplitudes in vacuum. i. F . This path integral can be used for the calculation of the effective action itself as well as for obtaining the corresponding scattering amplitudes. 1. one has to take care of the zero mode contained in it.C. its dual by F 2 with ε1234 = 1. A generalization to multiloop photonic amplitudes [40. the full content of the Bern–Kosower rules can be captured using an approach to quantum field theory based on first-quantized particle path integrals (“worldline path integrals”). using both proper-time [45] and dimensional regularisation [53–55].45. Scalar quantum electrodynamics For the case of scalar QED the basic formulas needed go back to Feynman [64].1. ε0123 = 1.1) T 0 x(T )=x(0) R Here T denotes the usual proper-time for the loop fermion. one writes 1 We work initially in the Euclidean with a positive definite metric g µν = diag(+ + + +). We will not discuss here the involved history of this subject but rather refer the interested reader to the review articles [56. The Euclidean eµν = 1 εµναβ F αβ field strength tensor is defined by F ij = εij k Bk . In the present paper we first give.57].F in terms of the matrices e. The corresponding Minkowski space amplitudes can be obtained by replacing gµν → ηµν = eµν → −iF eµν . i.

εN ] . εN ] Z∞ dT −m2 T N e (4πT )−D/2 VA0 [k1 . . Aµ (x) = N X εiµ eiki ·x . This immediately yields the following representation for the N -photon amplitude. .3) 0 The remaining y path integral is. given by Z RT 1 2 Dy e− 0 dτ 4 y˙ = (4πT )−D/2 . ε1 . τ2 ). performed using the Wick contraction rule µ (2.7) exp − dτ ieAµ x˙ = N! N=0 0 i=0 0 the individual terms correspond to Feynman diagrams describing a fixed number of interactions of the scalar loop with the external field.5) (2. # " ZT ∞ N ZT X   (−ie)N Y µ dτi x˙ µ (τi )Aµ (x(τi )) (2. (2. τ2 ) = |τ1 − τ2 | − (2. . Expanding the “interaction exponential”. in the “string-inspired formalism”. Γscal [k1 .410 C.4) y (τ1 ) y ν (τ2 ) = −g µν GB (τ1 . Z Z Z Dx = dx0 Dy. = (−ie) T 0 (2. in our conventions. The free Gaussian path integral determinants are. (2. GB (τ1 . T A “dot” always refers to a derivative in the first variable.9) .8) i=1 and picking out the term containing every εi once. . Although in this paper we will consider only the four-dimensional case in this factor D must be left variable in anticipation of dimensional regularization.2) where µ x0 1 ≡ T ZT dτ x µ (τ ). ε1 ] · · · VA0 [kN . kN . The corresponding N -photon scattering amplitude is then obtained by specializing to a background consisting of a sum of plane waves with definite polarizations. where (τ1 − τ2 )2 . Schubert / Nuclear Physics B 585 (2000) 407–428 µ x µ (τ ) = x0 + y µ (τ ). (2.6) Here D denotes the spacetime dimension. We can use this path integral for constructing the scalar QED N -photon amplitude as follows [39].

yielding the energymomentum conservation factor Z N X  Y eiki ·x0 = (2π)D δ (2. ε] ≡ dτ ε · x(τ ˙ )eikx(τ ).2) can be performed. dD x0 i=1 R The reduced path integral Dy is Gaussian.4).11) ki . using the correlator (2. writing . Its evaluation therefore amounts to Wick contracting the expression E D µ µ (2.g.10) 0 At this stage the zero-mode integration (2. [65]).C.12) y˙1 1 eik1 ·y1 · · · y˙NN eikN ·yN .. For the performance of the Wick contractions it is convenient to formally exponentiate all the y˙i ’s. (2. Schubert / Nuclear Physics B 585 (2000) 407–428 411 Here VA0 denotes the same photon vertex operator which is also used in string perturbation theory (see. ZT VA0 [k. e.

13) εi · y˙i eiki ·yi = eεi ·y˙i +iki ·yi . (2.

Then one needs only to “complete the square” to arrive at the following closed expression for the one-loop N -photon amplitude [39] Γscal [k1 . . i This allows one to rewrite the product of N -photon vertex operators as an exponential. . .j =1 0 i=1 0 1 ¨ Bij εi · εj ˙ Bij ki · εj + 1 G GBij ki · kj + iG 2 2 ). . εN ] N ZT X  Z∞ dT Y −D/2 −m2 T (4πT ) e dτi = (−ie) (2π) δ ki T N D ( × exp N  X i. ε1 . kN .lin(ε ) .

.

.

.

.

16) G . . . . Bij B i j ∂τ1 B 1 2 The expression (2. . ˙ B (τ1 . τ2 ) = sign(τ1 − τ2 ) − 2 (τ1 − τ2 ) . etc. G T 2 ¨ B (τ1 . G T (2.14) multi-linear Here it is understood that only the terms linear in all the ε1 . G ∂ G (τ . Let us consider explicitly the vacuum polarisation case. N = 2. For N = 2 the expansion of the exponential factor yields the following expression. Besides the Green’s function GB also its first and second deriatives appear. τ ).15) ˙ B (τ1 . τ2 ) = 2δ(τ1 − τ2 ) − .14) is identical with the corresponding special case of the “Bern– Kosower Master Formula” [34]. τ2 ) ≡ With “dots” we generally denote a derivative acting on the first variable. τ ). (2.  ˙ 2B12 ε1 · k2 ε2 · k1 eGB12 k1 ·k2 . and we abbreviate G ≡ G (τ . εN have to be taken. ¨ B12 ε1 · ε2 + G (2.

g. G (2. (2. ˙ B (τ1 . Schubert / Nuclear Physics B 585 (2000) 407–428 After performing a partial integration on the first term of Eq.1) can be generalized to the spinor QED case in various different ways. i = 1. k2 .22) ψ (τ1 )ψ ν (τ2 ) = g µν GF (τ1 .21) 0 Thus weRhave. T e dτi G k k −g k Πscal (k) = D/2 T (4π) 0 i=1 0 (2.19) After performing the trivial T -integration one arrives at µν Πscal (k) = e2 (k µ k ν − g µν k 2 )Γ (2 − D/2) (4π)D/2 Z1   D −2 × du (1 − 2u)2 m2 + u(1 − u)k 2 2 . (2. ×exp − dτ 4 2 (2. of course. 2 . The appropriate correlator is 1 µ (2.1).17) ε1 · ε2 k 2 − ε1 · kε2 · k G (k = k1 = −k2 ). ε1 . Thus we have Γscal [k1 . u2 = 0. We rescale to the unit circle. so that there is no new zero mode. τ2 ). Spinor quantum electrodynamics The worldline path integral representation (2. in addition to the same coordinate path integral as in (2.. the same as is found in the standard field theory calculation for the sum of the corresponding two Feynman diagrams (see. a Grassmann path integral Dψ representing the fermion spin.412 C. τ2 ) = 1 − 2u. We have then GB (τ1 . and use translation invariance in τ to fix the zero to be at the location of the second vertex operator.18) Note that the transversality of the vacuum polarization tensor is already manifest. 1 Γspin[A] = − 2 Z∞ Z Z dT −m2 T e Dx Dψ T 0 " ZT  # 1 2 1 ˙ x˙ + ψ · ψ + ieA · x − ieψ · F · ψ . [66]).16) in either τ1 or τ2 . 2. Z∞ 2 ZT  dT −D/2 −m2 T Y e2 µν µ ν µν 2 ˙ 2B12 e−GB12 k 2 . The formulation most suitable to the “stringy” approach uses Grassmann variables [67–69]. the integrand turns into  2 −G k 2 ˙ B12 e B12 (2. 2. e. The boundary conditions on the Grassmann path integral are antiperiodic.20) 0 The result of the final integration is. τ2 ) = T u(1 − u). τi = T ui .2. ψ(T ) = −ψ(0). ε2 ] = (2π)D δ(k1 + k2 )ε1 · Πscal (k) · ε2 . u1 = u.

(2. the explicit calculation of the Grassmann path integral can be circumvented. and replaced by the following simple pattern matching rule [34].14) one obtains an integrand " N # X . Schubert / Nuclear Physics B 585 (2000) 407–428 Our normalization for the free Grassmann path integral is Z RT 1 ˙ Dψ e− 0 dτ 2 ψ·ψ = 4.25) Taking the free Grassmann path integral normalization (2.C. (2.27) 0 Remarkably.24) 0 Looking again at the vacuum polarizationRcase. we need to Wick-contract two copies of the above vertex operator.23) The photon vertex operator (2. The calculation of Dx is identical with the scalar QED calculation.18) for the scalar QED vacuum polarisation tensor generalises to the spinor QED case as follows. Writing out the exponential in Eq. (2. ε] ≡ ZT dτ (ε · x˙ + 2iε · ψk · ψ)eikx . (2. (2.26) i=1 0 0 Proceeding as before one obtains e2 µν Πspin (k) = 8 D/2  k µ k ν − g µν k 2 Γ (2 − D/2) (4π) Z1   D −2 × du u(1 − u) m2 + u(1 − u)k 2 2 . R The additional contribution from Dψ is µ   (2i)2 ψ1 ψ1 · k1 ψ2ν ψ2 · k2 = −G2F 12 g µν k 2 − k µ k ν = − g µν k 2 − k µ k ν . Eq. 413 (2.23) into account.10) acquires an additional Grassmann piece 1/2 VA [k. e T dτi G × T (2. µν Πspin (k) = −2 e2 D/2 k µ k ν − g µν k 2  (4π) Z∞ 2 ZT  dT −D/2 −m2 T Y ˙ 2B12 − G2F 12 e−GB12 k 2 .

28) exp{· · ·}. G GBij ki · kj (2.  1 ¨ Bij exp ˙ Bij .

Now one removes all second derivatives GBij appearing in PN by suitable partial integrations in the variables τi . The result is an alternative integrand for the scalar QED ˙ B . ˙ Bij .int. up amplitude involving only GB and G .j =1 ˙ Bij .29) This is possible for any N [34].  1P 1P ¨ Bij e 2 GBij ki ·kj part. G ¨ Bij and on the kinematic with a certain polynomial PN depending on the various G ¨ invariants. ˙ Bij ) e 2 GBij ki ·kj .multi-linear = (−i)N PN G 2 i. G −→ QN (G PN G (2. The integrand for the spinor loop case can then.

(3. however the above replacement rule is valid for all possible results. non-trivial Lorentz matrices. Generalization of the Bern–Kosower master formula The presence of an additional constant external field. GF are. τ1 ). τ1 ). the Wick contraction rules Eqs. changes the path integral Lagrangian in Eq. G 44. τ1 ).4). 2 .. and thus functions of τ1 − τ2 . In [70] a certain standardized way was found for performing the partial integrations which leads to a canonical. These expressions should be understood as power series in the field strength matrix F . GF by ([42. τ2 ).414 C. τ2 ). replacing GB . (2. 1L = 12 iey µ Fµν y˙ ν − ieψ µ Fµν ψ ν . every closed cycle of G ˙ Bi2 i3 · · · G ˙ Bi2 i3 · · · G ˙ Bin i1 → G ˙ Bi1 i2 G ˙ Bin i1 − GF i1 i2 GF i2 i3 · · · GF in i1 . the vacuum worldline Green’s functions G This allows us to avoid making a case distinction between τ1 > τ2 and τ1 < τ2 that would become necessary otherwise [44]. GF (τ1 . GF we have left the τ -dependence implicit.21) only by a term quadratic in the fields. (2. G˙B (τ1 .3) where we have defined Z ≡ eF T . (2. (3. τ2 ) = −GFT (τ2 . (3.4) Since GB . The field can therefore be ˙ B . in general.6) ψ (τ1 )ψ ν (τ2 ) = GF (τ1 . τ2 ) = −G˙BT (τ2 . Schubert / Nuclear Physics B 585 (2000) 407–428 to the global factor of −2.1. τ2 ) = 2Z 2 sin(Z)   i Z ˙ e−iZ GB12 − 1 . cos(Z) (3. permutation symmetric and gauge invariant decomposition of the QED N -photon amplitudes. 3.g.5) y (τ1 )y ν (τ2 ) = −GB (τ1 . see also [63])   T Z ˙ B12 −iZ G ˙ e + iZ GB12 − 1 . Note that the generalized Green’s functions are still translationally invariant in τ .2) G˙B (τ1 . For N > 3 the result of the partial integration procedure is not unique. The replacement is done ˙ B12 G ˙ B12 G ˙ B (e. be obtained from the one for the scalar loop simply by replacing ˙ B ’s appearing in QN by its “worldline supersymmetrization”.22) have to be replaced by µ µν (3. (2.45]. Note also the symmetry properties GB (τ1 .30) ˙ Bi1 i2 G G Note that an expression is considered a cycle already if it can be put into cycle form ˙ B12 = −G ˙ B21 ). G using the antisymmetry of G simultaneously on all cycles. τ2 ) = Z sin(Z) ˙ GF (τ1 . τ2 ) = GF 12 e−iZ GB12 . By writing them as functions of ˙ B . taken in Fock–Schwinger gauge centered at x0 [41].1) GB (τ1 . τ2 ) = GBT (τ2 . µ 1 µν (3. absorbed by a change of the free worldline propagators. The QED N -photon amplitude in a constant field 3.

. and GF have non-vanishing coincidence limits.10) This is almost all we need to know for computing one-loop photon scattering amplitudes. (3. (3. ε1 . in a constant overall background field.1).12) (4πT )−D/2 → (4πT )−D/2 det−1/2 Z Since those determinants describe the vacuum amplitude in a constant field they turn out to be. just the proper-time integrands of the Euler–Heisenberg–Schwinger formulas (see (1. τ ) = (3.2). .3) using the rules that ˙ B (τ.14). (2. G˙B . 2 2Z i G˙B (τ.8) (3.11) → (4πT ) det (4πT ) Z   tan(Z) (Spinor QED). GB . τ ) = 1. kN . εN ] = (−ie)N (2π)D δ Y  Z∞ N ZT dT −D/2 −m2 T −1/2 sin(Z) e [4πT ] det dτi × T Z 0 ( × exp N  X i.7) (3. (3.C. This change is [41]   −D/2 −D/2 −1/2 sin(Z) (Scalar QED). τ ) = i cot(Z) − . (3. GB (τ.9) They are obtained from Eqs. Z GF (τ. . of course.45]: X  ki Γscal [k1 . G ˙ 2B (τ. τ ) = −i tan(Z). The only further information required at the one-loop level is the change in the free path integral determinants due to the external field. Retracing our above calculation of the N -photon path integral with the external field included we arrive at the following generalization of Eq. (3.j =1 i=1 0 1 1 ki · GBij · kj − iεi · G˙Bij · kj + εi · G¨Bij · εj 2 2 ). or the corresponding effective action. in contrast to their vacuum counterparts. Those are τ -independent:  T Z cot(Z) − 1 . representing the scalar QED N -photon scattering amplitude in a constant field [44.1)). . G (3. Schubert / Nuclear Physics B 585 (2000) 407–428 415 Another slight complication compared to the vacuum case is that. τ ) = 0.

.

.

.

.

The transition from scalar to spinor QED is done as in the vacuum case. (3.14) No such redefinition is possible for G˙B or GF . The spinor QED integrand for a given number of photon . We can use this fact to get rid of the coincidence limit of GB . again with only some minor modifications. instead of GB one can work with the equivalent Green’s function G¯B . τ2 ) ≡ GB (τ1 . (3. defined by SB (τ1 . τ ). namely.7). G (3. τ2 ) − GB (τ.13) multi-linear From this formula it is obvious that adding a constant matrix to GB will have no effect due to momentum conservation.

τi ) = sin(Z)cos(Z) Z 3. GF are. without specialisation of the Lorentz frame. . 3. This can be done in the following way. For example. . and taking the part linear in all ε1 . (3.17) g ≡ Fµν F 4 we have the relations . Defining the Maxwell invariants  1 1 f ≡ Fµν Fµν = B 2 − E 2 . GF . . it must be mentioned here that the cycle property is defined solely in terms of the τ -indices. τi ) → G˙B (τi . Schubert / Nuclear Physics B 585 (2000) 407–428 legs N is obtained from the scalar QED integrand by the following generalization of the Bern–Kosower algorithm: 1. non-trivial matrices in the Lorentz indices. Replacement rule: Apply to the resulting new integrand the replacement rule (2. GF in more explicit form. This would also be possible in the generic case. For this case the Green’s functions can be easily written out explicitly. (3.53] for a purely magnetic field B pointing along the z-direction the explicit matrix form of GB . irrespectively of what happens to the Lorentz indices. Partial integration: After expanding out the exponential in the master formula (3. (3. The scalar QED Euler–Heisenberg–Schwinger determinant factor must be replaced by its spinor QED equivalent. 2. Multiply by the usual factor of −2 for statistics and degrees of freedom.     −1/2 sin(Z) −1/2 tan(Z) → det . 4 2 1 eµν = iE · B. In the calculations performed in [45. The only other difference compared to the vacuum case is due to the non-vanishing coincidence limits of G˙B . where. Eqs.16) det Z Z 4.46. Those lead to an extension of the “cycle replacement rule” to include one-cycles [45]: i i − .30) ˙ B . in with G contrast to their vacuum counterparts. remove all second derivatives G¨B appearing in the result by suitable partial integrations in τ1 . one could use the Lorentz invariance to choose both E and B to point along the z-axis.8).9). . τi ) − GF (τi . τN . GF substituted by G˙B . (3. GF had been used. εN . .15) G˙B (τi .2. the expression ε1 · G˙B12 · k2 ε2 · G˙B23 · ε3 k3 · G˙B31 · k1 would have to be replaced by ε1 · G˙B12 · k2 ε2 · G˙B23 · ε3 k3 · G˙B31 · k1 −ε1 · GF 12 · k2 ε2 · GF 23 · ε3 k3 · GF 31 · k1 . . it is possible to directly express them in terms of Lorentz invariants. (3. . excepting the case E · B = 0. GF .13). Lorentz covariant decomposition of the generalized worldline Green’s functions For the result to be practically useful it will be necessary to know GB .416 C. Since the Green’s functions GB . . However.

F + AB. F+2 F−2 + f (iN ) even − (iN+ )2 (iN− )2 n  2  2  o 2 2 (iN ) N 1 + F (iN ) N 1 + F −f + f .F ) GB. r f ±g .F = SB.20) (3. (3. (3. GF   A− T A+ B12 b2 b−2 SB12 = Z+ + B12 Z 2 z+ z− (   +  ) 2 2 z+ z− + AB12 A− T − B12 AB12 − AB12 1 + − Z2 . (3. G˙B .23) F± = −N±2 F± .24) F+ F− = 0. feven (F ) = ∞ X c2n F 2n .F (AB. F 2 (3.27) n=0 Decomposing GB. even + even − − + 2 feven (F ) = feven (iN+ ) = N+2 1 − N− F+ F− + fodd (iN− ) iN+ iN− n  N− fodd (iN− ) − N+ fodd (iN+ ) F 2 fodd (F ) = fodd (iN+ ) = N+2 i − N−  o  e . (3. FF 417 (3. F2 + F e = −g1. n± ≡ 2 .26) where feven (fodd ) are arbitrary even (odd) functions in the field strength matrix regular at F = 0.F . + N− fodd (iN+ ) − N+ fodd (iN− ) F (3. = 2 − z2 ) z z z z 2(z+ + − + − − .F into their even (odd) parts SB.C.25) With the help of these relations one easily derives the following formulas.19) Define F± ≡ e N±2 F − N+ N− F N±2 − N∓2 N± ≡ n+ ± n− . (3. n=0 fodd (F ) = ∞ X c2n+1 F 2n+1 .18) (3.22) Then one has F = F+ + F− .21) (3. Schubert / Nuclear Physics B 585 (2000) 407–428 e2 = −2f 1.28) and applying the above formulas we obtain the following matrix decompositions of GB .

= 2 − + − + B12 B12 B12 B12 2 z+ − z− b+2 − S − Z b2 SF 12 = −SF+12 Z F 12 − n o  2 −   − 1 +  2 2 + = 2 S − z S − S z . dimensionless coefficient functions e = −z+ z− 1. 1 + S Z + F 12 − F 12 F 12 F 12 2 z+ − z−  + b  b AF 12 = −i A− F 12 Z− + AF 12 Z+ n o  i − +  + − e A − z A A − z A z Z + z Z . Z± ≡ eT F± = e ≡ eT F e. A± − . B12 = sinh(z± ) sinh(z± ) z± ˙ B12 ) ˙ B12 ) cosh(z± G sinh(z± G . Schubert / Nuclear Physics B 585 (2000) 407–428   b b + ˙ B12 Z+ + S − − G ˙ B12 Z− SB12 −G B12 z+ z− (  + iT −  SB12 − SB12 Z = 2 2 2(z+ − z− )  )   z− +  z+ − e . A± . The scalar. 2 z+ − z−  + b  b A˙ B12 = −i A− B12 Z− + AB12 Z+ n o  i − +  + − e A − z A A − z A z Z + z Z . Z Note that Z Z appearing in these formulas are given by ˙ B12 ) ˙ B12 ) 1 sinh(z± G cosh(z± G . those functions are the basic building blocks of parameter integrals for processes involving constant fields.29) Here we have further introduced z± ≡ eN± T . 2 − z2 z± ∓ b± ≡ Z± . Z z± (3.F .31) The non-vanishing coincidence limits are in A± B. Let us also write down the first few terms of the weak field expansions of these functions. = 2 − F 12 + F 12 − F 12 + F 12 2 z+ − z− (3. z± (3. Z 2Z −z z Z z± + −e . SF±12 = GF 12 F 12 = GF 12 cosh(z± ) cosh(z± ) ± = SB12 (3.32) In the string-inspired formalism. . 1 . ˙ B12 − ˙ B12 Z −G −G S S + z− B12 z+ B12 AB12 = iT 2 + b2 − b2 Z+ − SB12 Z− S˙B12 = −SB12 n o   − 1 +  2 2 − 2 + = 2 SB12 − z− SB12 1 + SB12 − SB12 z+ Z . A± Bii = coth(z± ) − A± F ii = tanh(z± ).30) b±3 = −Z b± .418 C.

34) det−1/2 Z tanh(z+ ) tanh(z− ) In Appendix A we write these formulas also for various special cases of interest.C.37) . (3. 0 0 0 0 0 0 Introducing also the projectors   1 0 0 0 0 1 0 0  g⊥ ≡  0 0 0 0. (3. n± = (B ± iE). where  0  −1 r⊥ ≡   0 0 1 0 0 0  0 0 . 2 zα α=⊥. F µν = F Using the above formulas we can obtain explicit results in a Lorentz covariant way. 0. 3 T 45 T 15 T 2      GB12 2 G2B12 3 1 1 ± 5 −2 z± + − + z± + O z± . 1 0  0 0  0 1 the matrix decompositions (3. (For this to be possible we have to assume that E ·B > 0. B). = det Z sinh(z+ ) sinh(z− )   z+ z− tan(Z) = . it will be useful to write down these formulas also for the Lorentz system where E and B are both pointing along the positive z-axis.36) (3. 2 F− = iErk .k µν AB12 = α −G ˙ B12 iT X SB12 rαµν . AB12 = 2 3 T 45 3 T      2 GB12 G2B12 4 GB12 2 6 z± + + + O z z SF±12 = GF 12 1 − 2 ± ± .35)  0 0 rk ≡  0 0  0 0 gk ≡  0 0 0 0 0 0 0 0 0 −1 0 0 0 0 0 0 1 0  0 0 . In the same way one finds for the (here we used the identity G B12 T determinant factors (3. the crossed field case f = g = 0. so that 1 N+ = B. Schubert / Nuclear Physics B 585 (2000) 407–428      2 GB12 2 G2B12 4 ± 2 6 ˙ B12 1 − 2 GB12 z± SB12 =G + + O z + z ± ± . 0 0 0 0 N− = iE.k (3. T 3 T T2      1 2 GB12 3 5 ˙ + G = G − + O z z z A± F 12 B12 ± ± ± F 12 3 3 T 419 (3. Nevertheless. and the self-dual field eµν . 2 zα α α=⊥. B = (0.) In this Lorentz system g = iEB. 0. E).29) can then be rewritten as follows: µν SB12 = − T X AαB12 µν g .12)   z+ z− −1/2 sin(Z) .33) ˙ 2 = 1 − 4 GB12 ). (3. Those are the magnetic field case E = 0. E = (0. F+ = Br⊥ .11).

For the 2-point case Eq. = det Z sinh(eBT ) sin(eET )   eBT eET tan(Z) .13) yields the following integrand . det−1/2 = Z tanh(eBT ) tan(eET ) (3. µν AF 12 α=⊥. The determinant factors specialize to   eBT eET −1/2 sin(Z) .420 C.38) α=⊥.39) 4. (3.k with  +  S A⊥ B/F ≡ S AB/F   k S AB/F ≡ S A− B/F (z+ = eBT ≡ z⊥ ).k = −i X AαF 12 rαµν . Schubert / Nuclear Physics B 585 (2000) 407–428 µν S˙B12 = X α SB12 gαµν . The scalar/spinor QED vacuum polarization tensors in a constant field We now apply this formalism to a calculation of the scalar and spinor QED vacuum polarisation tensors in a general constant field. (3.k α=⊥.k µν SF 12 = X X AαB12 rαµν . α=⊥. µν A˙ B12 = −i SFα 12 gαµν . (z− = ieET ≡ zk ).

}.1) exp{. .   ¯ (4. .

2) −ε1 · G˙B12 · ε2 k1 · G˙B1j · kj − ε1 · G˙B1i · ki ε2 · G˙B2j · kj ek1 ·GB12 ·k2 . where summation over i. We therefore decompose GB and GF into their parts symmetric and antisymmetric in the Lorentz indices as in (3. 2 is understood.28). We apply the “cycle replacement rule” to this expression and use momentum conservation.3) Next we would like to use the fact that this integrand contains many terms which integrate to zero due to antisymmetry under the exchange τ1 ↔ τ2 . where I µν = G˙B12 k · G˙B12 · k − GF 12 k · GF 12 · k µλ νκ   µλ + GF 12 GFνκ21 k κ k λ .4) k1 · G I µν turns. (4. Removing the second derivative in the first term by a partial integration in τ1 this becomes   ¯ (4. G˙B21 − G˙B22 + GF 22 − G˙B11 − GF 11 − G˙B12 µν µν (4. The content of the brackets then turns into ε1µ I µν ε2ν .multi-linear= ε1 · G¨B12 · ε2 − ε1 · G˙B1i · ki ε2 · G˙B2j · kj ek1 ·GB12 ·k2 .  SB12 · k2 = k1 · SB12 − SB11 · k2 ≡ −T k · Φ12 · k. First note that only the even part of GB contributes in the exponent. and deleting all τ -odd terms. j = 1. GF as above. into . after decomposing all factors of G˙B . k ≡ k1 = −k2 .

Note that again the transversality of the vacuum polarization tensors is manifest at the µν µν integrand level.e.4)).8) k 2 →0 F →0 Counterterms appropriate to this condition are easy to find from our above results for the ordinary vacuum polarisation tensors. we impose the following Sµν (k) (see. g → iG. N → ib (to be able to fix all signs we + − assume G > 0). kµ Iscal/spin = Iscal/spinkν = 0. e. In this way we find for the renormalised vacuum polarisation tensors Sµν (k) = Π µν (k) + Π scal scal  α µν 2 g k − kµkν 4π Z∞ dT −m2 T e T 0 Sµν (k) = Π µν (k) − Π spin spin  α µν 2 g k − kµkν 2π Z∞ 0 Z1 ˙ 2B12 .C.8) by subtracting those same expressions 2 with the last factor e−GB12 k deleted.5). condition on the renormalized vacuum polarization tensor Π S µν (k) = 0.. From the representations Eqs. (4.7) Πspin (k) = 2 T Z [4π]D/2 0 0 µν Here Iscal is obtained simply by deleting. in Eq. (2. The constant field vacuum polarisation tensors contain the UV divergences of the ordinary vacuum polarisation tensors (2.18).27). 2 For the Maxwell invariants this means f → F .9) 0 The remaining u1 -integral can be brought into a more standard form by a transformation of ˙ B12 = 1 − 2u1 . [5]).26) for these tensors it is obvious that we can implement (4.29) variables v = G and continuing to Minkoswki space 2 we obtain our final result for these amplitudes. In this way we obtain the following integral representations for the dimensionally regularised scalar/spinor QED vacuum polarisation tensors [71]. N → a. all quantities carrying a subscript ‘F’. i. Schubert / Nuclear Physics B 585 (2000) 407–428 µν Ispin ≡ 421 n  µν κλ µλ νκ  µν µλ − S˙B12 S˙B12 − SF 12 SFκλ12 − SF 12 SFνκ12 S˙B12 S˙B12 o µλ νκ µλ κ λ − AF 12 Aνκ A˙B12 − A˙ B22 + AF 22 + A˙B12 − A˙ B11 + AF 11 F 12 k k (4. Note also that r⊥ k → k˜⊥ . du1 G (4. and thus require renormalization. (2.6) T Z [4π]D/2 e2 0 0  Z1  Z∞ e2 dT 2−D/2 −m2 T µν µν −1/2 tan(Z) e T det du1 Ispin e−T k·Φ12 ·k . As is usual in this context we perform the renormalization on-shell. (2. Writing the integrands explicitly using the formulas (3. lim lim Π (4. . (4. µν Πscal (k) = −  Z1  Z∞ dT 2−D/2 −m2 T −1/2 sin(Z) µν T e det du1 Iscal e−T k·Φ12 ·k . (4.20). du1 G 0 dT −m2 T e T Z1  ˙ 2B12 − G2F 12 .5) (here we used (3..g. As usual we have rescaled to the unit circle and set u2 = 0. rk k → −ik˜k .

422 C. z+ = iesa. k 2 . In this system a = B. B).−  β  β β  β  − AαB12 − AαB11 + AαF 11 AB12 − AB22 + AF 22 − AαF 12AF 12 ) µ ν  µν 2 µ ν 2 bβ k bα k Z (4.−   β β  b µ b ν Zβ k − AαB12 − AαB11 AB12 − AB22 Z αk )  µν 2 µ ν 2 − g k −k k v .10) Z∞ ds −ism2 e s 0 ( " # z+ z− s X AαB12 − AαB11 b2 dv exp −i × k · Zα · k 2 tanh(z+ ) tanh(z− ) 2 α=+. the constant field vacuum polarization tensor was obtained before by various authors [20–24]. (4. For fermion QED.14) . Sµν (k) = Π spin α 2π (4. 0. k 2 . k⊥ = (0. 0). b = E. For the sake of comparison with their results. let us also specialize to the Lorentz system where E = (0. 0. Z Z 2 2 2 2 a +b a +b with a.11) − (g k − k k )(v − 1) . 0. 0. 0. G = E · B). k˜⊥ = (0.− zα −1 h X    i  β bβ2 · k − Z bα2 k µ Z bα2 µν k · Z bβ2 k ν Sα S Z × Z1 B12 B12 α. × Z where now z− = −esb. 0). Denoting kk = (k 0 . b≡ F 2 + G2 − F a≡ (4. k 3). k 0). 0. b denoting the standard “secular” invariants qp qp 2 2 F + G + F.13) (F = 12 (B 2 − E 2 ). −k 1 .β=+. k˜k = (k 3 . k 1 . e e b− = −i bF + a F b+ = aF − bF .12) (4. Schubert / Nuclear Physics B 585 (2000) 407–428 Sµν (k) = − Π scal α 4π Z∞ ds −ism2 e s 0 ( #  z+ z− s X AαB12 − AαB11 dv 2 bα · k exp −i × k·Z 2 sinh(z+ ) sinh(z− ) 2 α=+.β=+.− zα −1 h X    i β β  b2 µν α bβ2 · k − Z bα2 k µ Z bβ2 k ν SB12 − SFα 12 SF 12 Z k·Z SB12 × α Z1 α. E) and B = (0.

18) ) − aspin = sinh(z0 ) cosh2 (z0 ) In this form our integrand for the spinor QED case can easily be identified with the one given in [22. ascal = sinh(z0 ) ⊥k. sin(z) sinh(z0 ) = cos(zv) − cos(z) sin(z) !2 . Z∞ Z1 (   −2 zz0 cos(z) cosh(z0 ) −isΦ0 ds dv e 1 s 2 sin(z) sinh(z0 ) 1 0 −1  X  αβ  αβ × s spin  gαµν kβ2 − kαµ kβν + a spin  k˜αµ k˜βν α Π spin  (k) = − 4π scal Sµν  scal α.k⊥ − .k⊥ .C.k⊥ ascal (4.k⊥ 0 − tan(z) + tanh(z ) aspin = − sin(z) sinh(z0 ) sin(zv) sinh(z0 v) . (4. (4. sinh2 (z0 v) sinh2 (z0 ) sscal . Schubert / Nuclear Physics B 585 (2000) 407–428 423 our result can be written as follows.24] using the identities [24] ⊥⊥ = sspin − µν µ µ gk kk2 − kk kkν = −k˜k k˜kν . 2 z sin(z) 2 z0 sinh(z0 ) sin2 (zv) sin2 (z) ⊥k. sin(z) sinh(z0 )   cosh(z0 v) − cosh(z0 ) 2 kk . − cos(z) cosh(z0 )  2 cosh(z0 v) − cosh(z0 ) sinh2 (z0 v) kk 0 + tanh(z .17) cos2 (zv) .16) =− sin2 (zv) (4. −e v2 scal  (4.k ) v2 − 1 −ism2 µν 2 µ ν (g k − k k ) . ⊥⊥ ascal = sin(zv) sinh(z0 v) .19) . sspin = sin(z) sinh(z0 ) cos(z) cosh(z0 ) sinh2 (z0 v) cosh2 (z0 v) kk − . z0 = eEs. sspin = sinh2 (z0 ) cosh2 (z0 ) 2  cos(zv) − cos(z) sin2 (zv) ⊥⊥ − tan(z) − . Φ0 = m2 + ⊥⊥ = sscal kk sscal = 2 2 k⊥ cos(zv) − cos(z) kk cosh(z0 v) − cosh(z0 ) − . = aspin sin(z) cos2 (z)    cosh(z0 v) − cosh(z0 ) cos(zv) − cos(z) ⊥k. cos(zv) − cos(z) cosh(z0 v) − cosh(z0 ) .β=⊥. µν 2 µ ν µ ν g⊥ k⊥ − k⊥ k⊥ = k˜⊥ k˜⊥ .15) where z = eBs. cos2 (z) sin2 (z) sin(zv) sinh(z0 v) cos(zv) cosh(z0 v) ⊥k.

76]. Various other generalizations of this formalism would be of interest. the string-inspired technique combines all loop propagators into a single propagator from the beginning. Already the two-point case clearly shows the advantages of the method as compared to the standard field theory techniques used in previous calculations of these quantities. An obvious one is the extension to QED amplitudes involving external scalars or fermions. 3 Up to the global sign. (4. the vacuum polarization tensor in a constant field was obtained in [21] using the “operator diagram technique”. using the generalization of the path integral representation (2.1). Discussion We have described in detail the application of the “string-inspired” formalism to the calculation of QED photon amplitudes in a constant external field. See [51] (scalar QED) and [77. As is usual in the application of the string-inspired technique to QED processes.79.21) given in [75.73] rather than the more usual first-order one. 5. Due to our novel representation (3. 3 For the scalar QED case. Also.10). The two-point cases were studied explicitly.80].424 C. The close relationship between the scalar and spinor QED calculations is also useful. which differs from the one in [22. In a sequel paper [74] we will consider more generally the calculation of amplitudes involving both vectors and axial vectors in a general constant field. the relevant generalizations of Eqs. Schubert / Nuclear Physics B 585 (2000) 407–428 and trigonometric identities. Clearly one would expect these advantages to become more significant with increasing number of legs. (2. The partial integration procedure combined with the Bern– Kosower “replacement rule” effectively replaces the calculation of Dirac traces.24] due to a different definition of the vacuum polarization tensor.39. For the three-photon amplitude in a magnetic field this was already demonstrated in [46].29) of the generalized worldline Green’s functions here this was achieved without the use of explicit matrix representations. (Our field theory conventions follow [72]. There are several aspects to this. The resulting parameter integral representation agrees with our Eq. our calculation of the spinor QED vacuum polarisation tensor in a constant field has yielded the corresponding scalar QED quantity as a byproduct. We believe that in the present formalism even a calculation of photon-photon scattering in a constant field would not be excessively cumbersome. In both cases we have arrived at compact results written in terms of the two Lorentz invariants of the Maxwell field.78] (spinor QED) for relevant work for the vacuum case.) . and the global loop propertime variable is to be introduced at a later stage. (2. It is due to the fact that the “string-inspired” formalism is based on a second-order field theory formulation of fermion QED [38. while in a Feynman diagram calculation each propagator has to be parametrized separately. Interesting would also be the extension to the finite temperature case along the lines of [47. and is technically clearly preferable to the latter.21) have been known for a long time [67–69].

SB12 = GB12 1 + 3T T 3 2 ˙ B12 Z. AB12 = − G SSB12 = GB12 1 + B12 GB12 Z. 1 2 i ˙ G Z 2. AB12 = i GB12 GB12 Z . AF 12 = −iGF 12 z cosh(z) E = g = 0.3) (A. N− = 0.C. N+ = B. AF 12 = −iGF 12 G SF 12 = GF 12 1 + GF 12 GB12 Z 2 . We have subtracted from SB its coincidence limit. SF 12 = GF 12 1 + GF 12 2 − 2 z z cosh(z) ˙ B12 ) sinh(zG Z. Hernquist. analogous. A˙ B12 = i 2 − z z sinh(z)  ˙ B12 )  cosh(zG 1 Z 2. n+ = n− = . The purely electric case is. Worldline Green’s functions for special field configurations In this appendix we give the explicit form of the (Euclidean) generalized worldline Green’s functions and determinants for the following three special cases of a constant field: (i) the purely magnetic field case. SB12 = GB12 1 + 2 GB12 + T z 2z3 sinh(z)   ˙ B12 ) iT sinh(zG ˙ B12 Z. 1. Ritus for various helpful information.I.2) where z = eBT . 2. Dittrich. L.1) (A. (iii) the self-dual case. −G AB12 = 2 2z sinh(z)   ˙ ˙ B12 1 + 1 G ˙ B12 − sinh(zGB12 ) Z 2 . det−1/2   z sin(Z) = . Z tanh(z) (A. S˙B12 = G z2 sinh(z)   ˙ 1 cosh(zGB12 ) Z. M. T     −1/2 sin(Z) −1/2 tan(Z) = det = 1. det Z Z (A. Z sinh(z) det−1/2   z tan(Z) = . Reuter and V. 2 2  ˙ B12 ) − cosh(z)  2 1 cosh(zG S Z . (ii) the crossed-field case. Appendix A. Schubert / Nuclear Physics B 585 (2000) 407–428 425 Acknowledgements The author would like to thank W. Magnetic field case: B 1 f = B 2 .4) . Crossed field case: f = g = n± = N± = 0.E. 3T B12 3   2 ˙ 2 1 2 ˙ ˙ ˙ GB12 − Z. which is expressed by the “bar”. of course.

Phys. Ritus. T. Selsk. Phys. in: J. Rev. 1987. Naturwissenschaften 23 (1935) 246.6) References [1] H. Dover. Fock. Quantum Electrodynamics.). Phys. Nuovo Cimento 19 (1961) 847. Toll. Effective Lagrangians in Quantum Electrodynamics. K.E.5) (A. n− = 0. Z. B. New York. Euler. 1985. SF 12 = GF 12 cosh(Z) Z cosh(Z)     Z2 Z2 sin(Z) −1/2 tan(Z) . Rev.A. Acta Phys.426 C.. Heron Press. Phys. [8] R. Gies. 14 (1936) 1.). Springer. Nuovo Cimento 47 (1967) 261. 69 (1972) 555. Shabad. [11] J. New York. ˙ B12 ) T cosh(Z) − cosh(Z G 1. Sandansky 1998. Fy. 25 (1967) 212. Z. P. Phys. B 36 (1972) 121. Erber. Self dual case: e. Ann. 1952. Austr. det . Euler. 82 (1951) 664. Schwinger (Ed. Breitenlohner. hep-ph/9806417. Nova Science Publ. Cover. F =F f = g. Dittrich. n+ = N+ = N− = p f.. 98 (1936) 714.B. [9] V. [14] D.E. 1998. [18] N. Nucl. A. 29. Phys. 1992.-Y. Quantum Electrodynamics.. H. [17] A. Dittrich. Pengo et al. [16] R. (A. [5] W. Sov. [19] V. Reinhardt. Schwinger. H.I. [4] V. Heisenberg. Sowjetunion 12 (1937) 404. Nuovo Cimento 6 (1957) 501. AF 12 = −iGF 12 Z. G. Schubert / Nuclear Physics B 585 (2000) 407–428 3. Kalman. Lebedev Phys. Ginzburg (Ed. Springer. Heron Press. [13] R. [20] I. Phys. 168. SSB12 = 2 Z sinh(Z)   ˙ B12 ) T sinh(Z G ˙ − GB12 Z. Tsai. [15] W. Medd.A. Physik. p. JETP 33 (1971) 483. Greiner. JETP 28 (1969) 371. Proc. D 10 (1974) 492.I. AB12 = i 2 sinh(Z) 2Z   ˙ B12 ) ˙ B12 ) 1 sinh(Z G cosh(Z G 1. PhD Thesis. [3] V. Vacuum birefringence in strong magnetic fields. reprint. Kockel. Minguzzi. [2] W. Nuovo Cimento 9 (1958) 145. 1958. Baier. of Frontier Tests of Quantum Electrodynamics and Physics of the Vacuum. Narozhnyi. Sandansky 1998.H. p. A˙ B12 = −i − 2 Z. M. J. 33 (1974) 1113. Princeton University. . Sov. in: Proc. Reuter. Vol. The Lagrangian function of an intense electromagnetic field and quantum electrodynamics at short distances. Vidensk. 1998. Constantinescu. 90 (1975) 166. [7] W. [10] J. Phys. unpublished. [6] W. Vacuum birefringence in strong magnetic fields. Inst. S˙B12 = sinh(Z) Z sinh(Z) Z ˙ ˙ B12 ) cosh(Z GB12 ) sinh(Z G 1. Ann. Weisskopf. Mat. 59. in: Proc. in: V. = = det−1/2 2 Z Z sinh (Z) tanh2 (Z) √ where Z = f eT . Shabad. Batalin. Ritus. Phys.I. Lett. [12] A. Rev. of Frontier Tests of Quantum Electrodynamics and Physics of the Vacuum. Dan.

Fried. Kosower.C.G. Schubert / Nuclear Physics B 585 (2000) 407–428 427 [21] V. University of Chicago Press. B 433 (1995) 181. hep-ph/9401221. B 480 (1996) 99. Rev. M. Schubert. C. Polon. Nucl. Lett. D 10 (1974) 1342. M. Artimovich. D. hep-th/9307001. [47] I. C 6 (1999) 175. Nuovo Cimento B 64 (1981) 287. Schubert.C. [25] Z. 1998. Strassler. Nucl.G. Functional Methods and Models in Quantum Field Theory. Phys. Raffelt. D.G.M. Schmidt. 1972. D. Ann. Tübingen University. FERMILAB-PROPOSAL-P877A. Sumino. Bern. I. D 38 (1988) 1888. Bern. Phys. [27] G. [57] C. 1996. Bern.Y. Schubert. Schmidt. Theor. Nucl. Dunbar. Bialynicka-Birula.N. Rev. MIT Press. Baring. M. Batalin. [55] G. . Schubert. 113 (1997) 1442. [56] Z. hep-th/9806156. Dunbar. D 71 (1978) 1977. Phys. Phys. 1992.-T. Ann. C. [50] M. J. [22] L. Eur. Schmidt.P. [36] Z. hep-th/9803144. D 53 (1996) 4598. 471. Phys. B 385 (1992) 145.A. [43] V. Sov. B 331 (1994) 69. Schubert. Norridge. hepth/9704194. E. I. hep-ph/9302280. J. Bordi. R. Phys. Boulder. Lectures given at 36th Cracow School of Theoretical Physics. B 27 (1996) 3965.L. C. Rev. Daikouji. Acta Phys. Shimada. M. E. Cambridge. Kosower. Phys.A. [49] D. Ann. B 258 (1985) 435. Phys. Lett. V. Bialynicka-Birula. Reuter. B 318 (1993) 438. Rev. B 379 (1992) 562. Tsai. Phys. hep-th/9804143. (N. hep-th/9907190.M. Phys. 1996. hep-th/9605035. Phys. JETP 41 (1975) 190. Dixon. Schubert. INFN/AE-94/27. [34] Z.S. Rev. Lett. [39] M. Z. hep-th/9408014. L. Shvartsman. Nucl. Phys. P. Phys. Körs.G. hep-th/9403158. Bologna. Kaspi et al. Phys. [33] W. Phys. hep-th/9309055. M. JETP 70 (1990) 787. G. Measurement of the Magnetically-induced QED Birefringence of the Vacuum and an Improved Laboratory Search for Light Pseudoscalars. [53] D. Rev. Schmidt. Sov. Harding.) 67 (1971) 599. Strakhovenko. Géhéniau. [51] K. Katkov. hep-th/9604152.J.G. P. Nucl. V. [29] V. UCTP-106-98. JETP 40 (1975) 225. D 51 (1995) 2513. Fradkin. C. hep-th/9610191. Dunbar. Rev. Schmidt. Fliegner. Phys. Phys. (N. D. B 564 (2000) 591. [59] A. C 64 (1994) 111. Haberl. Barducci. hep-ph/9509383. Gusynin. [41] M. B 379 (1992) 451.S. Phys. D 53 (1996) 2150. 259 (1997) 313.M. Sov.V.K. A. 74 (1996) 282. Lett. Shovkovy.. Rev. Cangemi. 1999.A. Nucl. Shino. Lett.Y. H. Roland. Baier. Schubert. Nucl. hep-ph/9909500. Sato. Phys. [45] M. Phys. [35] Z. Shaisultanov. Lett. Physica 16 (1950) 822. D 2 (1970) 2341. Lett. [30] D.G. Reuter.A. [26] S.M. Phys.M.K. in: TASI Lectures. M.G. [23] G. Phys. PhD Thesis. [31] S.) 264 (1998) 51. C. Phys. Phys. C. AGN and the Diffuse Background. hep-th/9610108. Adler. Phys. Rev. [52] K. [46] S.F. Y. Lee et al. Schubert. Phys.Y. 1999. B 378 (1996) 354.G. C. Bakalov. [38] Z.. Bern. Zakopane. Rev. [42] D. C. F.C. Bern. T. Schmidt. p. Schmidt. hep-th/9410100. [37] D. Urrutia. Phys.G. Gies. Chicago. 66 (1991) 1669. [24] H. Schubert.C. B 312 (1993) 277. Can. [28] M. Casalbuoni.G. Dunne. Phys. Schmidt. of Stellar Endpoints.A. Dunne. hep-ph/9508377. Sh. [40] M. M.L. Phys. C. Fliegner. Schubert. astro-ph/9704210. Adler. Math. unpublished. B 441 (1998) 313. [54] B. Fliegner. 70 (1993) 2677. Shovkovy. D’Hoker. Lett. Stars as Laboratories for Fundamental Physics. to appear in the Proc. hep-ph/9304249. 77 (1996) 1695. [44] R. hepth/9707189. [58] H. [48] D. [60] I.A. Phys. hep-th/9512142. [32] J. Phys. Phys.

[67] E. D. Rev. Phys. [74] C. Phys.I. Nuovo Cimento A 33 (1976) 389. Nucl. [75] D. Berezin. [78] A. [70] C. [66] C. Karanikas. 1985.M.C. B 440 (1998) 101. Schubert / Nuclear Physics B 585 (2000) 407–428 [61] S. hep-ph/9502230. Phys. Quantum Electrodynamics with Unstable Vacuum. Ann. McKeon. M. Press.N. 104 (1977) 336. Ktorides. McKeon.428 C. [64] R. C.S. JHEP 9911. [62] E. Schubert. hep-th/9807072.C. Rebhan.H. Phys. Springer. Eur. Sato. J. Itzykson. [65] M. [68] R. Lett. [76] F. 80 (1950) 440. [79] M.S. McKeon.P. Phys. D.C. [77] D. Schubert. Feynman.A. E. Rev. Sherry. Witten. J. Fradkin. Casalbuoni.N. C.S. 1. hep-th/9806138. Phys. of Corfu Summer Institute on Elementary Particle Physics. LAPTH-Conf-729/99. Fradkin.G.B. A 9 (1994) 2167. Cambridge Univ. Ann. Shvartsman. Marinov. Phys. S. [80] H. Cambridge Univ. hep-th/9905027. Morgan.G. hep-ph/9905525. 033. J. Phys. B 585 (2000) 429. Lett.-T. Haack. C 7 (1999) 149.G. Phys. B 351 (1995) 249. Superstring Theory. Weinberg. 03 (1999) 022. 1995. . Press. 1998. A. Green. HD-THEP-98-41. Zuber. Quantum Field Theory.A. 76 (1966) 588. Eur. Dilkes. McGraw-Hill. [72] S. Gitman. Nucl. to appear in the Proc. J. 1987. C 5 (1998) 693. [71] C. High Energy Phys. hep-th/9710067.G. hepth/9812213. Schubert. Phys. M. D 48 (1993) 2891. Mod. J. Lett. C. next article in this issue.C. Vol. [63] D. 1991. Schwarz. Schubert.M. Rajeev. Schmidt. [69] F. The Quantum Theory of Fields. T. 173 (1987) 249. McKeon.G. Phys. hep-th/9809053. [73] A.G. Schubert.

BP 110. as well as a general constant electromagnetic background field.-y Keywords: Axial vacuum polarization. In vacuum. γ ν → γ ν and ν ν¯ → γ γ were considered already before the advent of the standard model using the Fermi theory [15]. Schubert). The vector–axialvector two-point function in a constant field is calculated explicitly. Introduction: standard model processes in constant electromagnetic fields In the first part of this series [1] we explained in detail how the “string-inspired” technique [2–5] can be used for the efficient calculation of one-loop scalar/spinor QED N -photon amplitudes in a constant external field [6–14]. Chemin de Bellevue.f (C. PACS: 11. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. In the standard model. F-74941 Annecy-le-Vieux. France Received 29 February 2000.3 2 1 3 ( 0 0 ) 0 0 4 2 4 .in2p3.elsevier. In the standard model this suppression E-mail address: schubert@lapp. as was noted by Gell-Mann [16] (for massless neutrinos.V. we extend this analysis to the case of mixed vector–axialvector amplitudes in a constant electromagnetic field. In the present sequel.15. Amplitudes of this type have.V. PII: S 0 5 5 0 .  2000 Elsevier Science B. in recent years. String-inspired 1.7 . a typical example would be the diagram shown in Fig. However in the Fermi limit they vanish due to the Landau–Yang theorem. Photon–neutrino. Constant field. and with both photons on-shell).Nuclear Physics B 585 (2000) 429–442 www.nl/locate/npe Vacuum polarisation tensors in constant electromagnetic fields: Part II Christian Schubert Laboratoire d’Annecy-le-Vieux de Physique Théorique LAPTH. 1. As an application we calculated the scalar and spinor QED vacuum polarisation tensors in a constant field. contributing to the process γ γ → ν ν¯ .Bt. 12. photon–neutrino interactions appear at the one-loop level.15. revised 19 June 2000. The 2 → 2 processes γ γ → ν ν¯ . All rights reserved. been much investigated in connection with photon–neutrino processes. All rights reserved. accepted 26 June 2000 Abstract In the second part of this series we apply the “string-inspired” technique to the calculation of one-loop amplitudes involving both vectors and axialvectors.

which has the advantage of avoiding the separation of these amplitudes into their real and imaginary parts. 1. Similarly one would expect the magnetic field to remove the Fermi limit suppression of the above 2 → 2 processes. Plasma effects must be taken into account. 2(c). which in vacuum are therefore more important at low energies than the four-leg processes [17–21]. Similarly the Cherenkov process ν → νγ becomes possible. Ioannisian for providing this figure.41 × 1013 Gauss [23].430 C. believed to be an important source for neutrino production in some types of stars [26]. This has recently been verified both for the weak [30. Secondly. have recently been found to surpass the “critical” magnetic field strength Bcrit = m2e /e = 4. Diagram contributing to γ γ → ν ν. We will use here the one proposed in [35] and elaborated in [36]. The double line represents the electron propagator in the presence of the B-field. Schubert / Nuclear Physics B 585 (2000) 429–442 Fig. An important example is the plasmon decay γ → ν ν¯ [24. where ω is the center-of-mass energy and MW the W boson mass. Firstly. The amplitude then effectively reduces to a photonic amplitude with one of the photons replaced by the neutrino current. by modifying the photon dispersion relations it opens up phase space for neutrino–photon reactions of the type 1 → 2 + 3. One is thus led to the study of vector – axialvector amplitudes in a constant field.25]. various different representations have been derived for axialvector couplings [33–35].N. it provides an effective photon–neutrino coupling via intermediate charged particles [27–29].27]. In the string-inspired formalism. 2(a) and 2(b). In astrophysical environments it is often not realistic to consider these processes as occuring in vacuum.31] and strong field cases [32]. as well as the presence of magnetic fields which. For processes of this type the magnetic field plays a double role. although it turns out to be of lesser astrophysical relevance [22. There is no such suppression for processes involving two neutrinos and more than two photons. at the surface of neutron stars. Many more photon–neutrino processes become possible if one admits neutrino masses or anomalous magnetic dipole moments [22]. Of particular interest are then processes which do not occur in vacuum but become possible in a medium or B-field. 1 In the limit of infinite gauge-boson masses both diagrams can be replaced by diagram Fig. Extending that work to the case where an additional constant electromagnetic background field is present will enable us to perform the first calculation of the vector – 1 I thank A. . In the standard model the effective coupling is provided by the diagrams shown in Figs. ¯ manifests itself by factors of ω/MW .

T → is. eµν → −i F eµν . ε0123 = 1.37] special cases. ε1234 → iε1230 . i. 2 Γ [A. F 4i = −iEi . A5 ] = ln Det[/ p + eA / + e5 γ5A / 5 − im] Z Z∞ RT 1 dT −m2 T e Dx Dψe− 0 dτ L(τ ).29. as well as possible generalizations. Schubert / Nuclear Physics B 585 (2000) 429–442 431 Fig. we shortly review the worldline representation of vector–axialvector amplitudes proposed in [35.36]. (b) Penguin diagram (only for νe ). k 4 → −ik 0 . The organization of the paper is the following. =− 2 T (2. Worldline representation of the vector–axialvector effective action In [35] the following first-quantized path integral representation was derived for the oneloop effective action induced by a Dirac fermion loop for a background vector field A and axial vector field A5 (both abelian).1) 0 2 We work initially in the Euclidean with a positive definite metric g µν = diag(+ + ++). Chapter 4 contains a detailed calculation of the vector–axialvector amplitude in a constant field. F 4i → F 0i = Ei . 2. j = 1. In the second chapter. its dual by F 2 ε1234 = 1.C. Neutrino–photon coupling in an external magnetic field. This amplitude is of relevance for various of the above processes. It has been obtained previously for the magnetic [22. (c) Effective coupling in the Fermi limit. To obtain the corresponding Minkowski space amplitudes replace gµν → ηµν = diag(− + ++). 3. 2. F . (a) Z–A-mixing. The Euclidean field eµν = 1 εµναβ F αβ with strength tensor is defined by F ij = εij k Bk . axialvector amplitude in a general such field. We then extend this formalism to the inclusion of constant external fields in Chapter 3. The result is discussed in chapter five. 2.37] and crossed field [27.

6) (2. the Grassmann path integral has to be taken R with antiperiodic boundary conditions. µ x µ (τ ) = x0 + y µ (τ ).7) (2. This zero mode RT µ we eliminated by fixing the average position x0 ≡ T1 0 dτ x µ (τ ) of the loop. Note that.8) . and Dψ a Grassmann path integral representing the spin of the loop fermion. The reduced path integrals will.5) d 4 ψ0 ψ0 ψ0ν ψ0κ ψ0λ = εµνκλ . 2 where (τ1 − τ2 ) . although we will consider only the four-dimensional case in the present work. determined by the power of γˆ5 appearing in a given term. The boundary conditions on the Grassmann path integral are. Z µ (2. they are periodic (antiperiodic) with period T if that power is odd (even). τ2 ). For the case of an even number of axial vectors the Wick contraction rules are the same as in the QED case treated in part I.4) 0 The zero mode integration then produces the ε-tensor expected for a fermion loop with an odd number of axial insertions. so that only Dx has a zero mode. Schubert / Nuclear Physics B 585 (2000) 429–442 1 1 L = x˙ 2 + ψ · ψ˙ + iex˙ · A − ieψ · F · ψ 4 2 +ie5 γˆ5 (−2x˙ · ψψ · A5 + ∂ · A5 ) + (D − 2)e52 A25 . τ2 ).432 C. Before evaluating this double path integral one must eliminate the zero mode(s) contained in it. the Lagrangian has an explicit dependence on the spacetime dimension D through dimensional regularisation [36]. hy µ (τ1 )y ν (τ2 )i = −g µν GB (τ1 . (2. For the N -photon amplitude. after expansion of the interaction exponential. τ2 ) = |τ1 − τ2 | − T GF (τ1 . the Grassmann path integral has a zero mode as well. (2. (2.2) Here T denotes the usual Schwinger proper-time for the loop fermion. This can be removed analogously. R (2. µ ψ µ (τ ) = ψ0 + ξ µ (τ ). R Dx the integral over the space of all closed loops in spacetime with periodicity T . be evaluated using Green’s functions in τ -space. After the boundary conditions are determined γˆ5 can be replaced by unity.3) When appearing with periodic boundary conditions. τ2 ) = sign(τ1 − τ2 ). following the recipes of the “string-inspired” formalism. GB (τ1 . 1 hψ µ (τ1 )ψ ν (τ2 )i = g µν GF (τ1 . Z Dx = dx0 Dy. ZT dτ ξ µ (τ ) = 0.

it is convenient to linearize this term through the introduction of an auxiliary path integration [36]: . The only slight complication compared to the QED case is due to the term quadratic in A5 appearing in the worldline Lagrangian (2. One-loop vector–axialvector amplitudes in a constant field In the same way as it was done in part I for the pure vector case. Schubert / Nuclear Physics B 585 (2000) 429–442 433 In the case of an odd number of axialvectors the Grassmann Wick contraction rule (2. in our conventions. hξ µ (τ1 )ξ ν (τ2 )i = g µν G 2 (2. Z RT 1 2 (2. 3. Aµ (x) = M X εiµ eiki ·x . leaving a D-dependence only for the free y-path integral (2.2). To be able to define an analogue of the photon vertex operator 1 2 ZT VA [k. we can obtain from (2. This anticipates the dimensional regularization. See [36] for the case of an arbitrary (even) spacetime dimension.9) (A “dot” always refers to a derivative in the first variable.2) 0 for the axial coupling. (2.7) has to be replaced by 1 ˙ B (τ1 .1) and picking out the term containing every εi .10). Z RT 1 ˙ Dψe− 0 dτ 2 ψ·ψ = 4. i=1 A5µ (x) = N X ε5iµ eik5i ·x i=1 (3. τ2 ).2) the M-vector–N -axialvector amplitude by specializing to backgrounds consisting of sums of plane waves with definite polarizations. (2.12) P In writing them we have specialized to the four-dimensional case.11) A Z Dξ e− RT 0 dτ 12 ξ ·ξ˙ = 1. ε5i once.10) Dye− 0 dτ 4 y˙ = (4πT )−D/2 .C. ε] = dτ (ε · x˙ + 2iε · ψk · ψ)eikx (3.) The free Gaussian path integral determinants are.

3) 0 The Wick contraction rule for this auxiliary field is simply hzµ (τ1 )zν (τ2 )i = 2g µν δ(τ1 − τ2 ). G˙B (τ1 . 1 2 1 2 (3. In Fock–Schwinger gauge centered at x0 [6] its contribution to the worldline Lagrangian (2. however the result is lengthy. εM ]VA5 [k51. see also [7])   T Z ˙ B12 −i Z G ˙ B12 − 1 . εi }. τ2 ) = GF 12 e−i Z GB12 . (3.9). depending on the boundary conditions.11].P (−i)M+N eM e5N 2 Z∞ dT (4πT )−D/2 T 0 × VA [k1 .7). (2. ε1 ] · · · VA [kM .8) . and either (2. cos(Z) (3.5) 0 We can then write the one-loop M-vector–N -axialvector amplitude in terms of Wick contractions of vertex operators: Γ [{ki .6) The Wick contractions are done using (2. This leads to a replacement of GB . ZT VA5 [k. (3.4). τ2 ) = Z sin(Z) ˙ GF (τ1 .P .434 C. and be taken into account by a ˙ B .7) or (2. {k5j . ε51 ] · · · VA5 [k5N . Dz exp − dτ 4 (3. G ([8. It can therefore be absorbed into the kinetic part of Lagrangian. Let us now introduce an additional background vector field A¯ µ (x) with constant field strength tensor F¯µν . e + iZ G GB (τ1 .6). τ2 ). ε] ≡ γˆ5 dτ (iε · k + 2ε · ψ x˙ · ψ + √ D − 2ε · z)eik·x .7) where we have defined Z ≡ eF T (omitting the “bar”). Schubert / Nuclear Physics B 585 (2000) 429–442 " # ZT exp −(D − 2)e52 dτ A25 0 Z = " ZT  # √ z2 + ie5 D − 2z · A5 . (2. (2. The presence of the background field thus modifies the Wick contraction rules Eqs. This can still be done in closed form [36].2) can be written as 1L = 12 iey µ F¯µν y˙ ν − ieψ µ F¯µν ψ ν . ε5j }] 1 = − NA. (3.4) This allows us to define an axialvector vertex operator as follows.9) to µν hy µ (τ1 )y ν (τ2 )i = −GB (τ1 . τ2 ) = 2Z 2 sin(Z)   i Z ˙ e−i Z GB12 − 1 . ε5N ] A. (3.10.6). GF by change of the free worldline propagators.

While in the present paper we will consider only the amplitude case. (3. F 2 . 4. however. F.36].13) Note.8. that for the case of periodic Grassmann boundary conditions this field dependence cancels out between the coordinate and Grassmann path integrals.38.C. Z (3.13. This cancellation can be understood as a consequence of the fact that the two path integrals are related by worldline supersymmetry [2.11]:   Z RT 1 2 1 D 1 sin(Z) µ ν .11). 2 435 (3.9) (3. τ2 ). τ2 ). we calculate the vector–axialvector two-point function in a constant field. In Section 3. and the two standard worldline Green’s functions and determinants in terms of 1. (3.11) Dye− 0 dτ ( 4 y˙ + 2 iey Fµν y˙ ) = (4πT )− 2 det− 2 Z Z h i RT 1 1 µ ν ˙ Dψe− 0 dτ ( 2 ψ·ψ−ieψ Fµν ψ ) = 4det 2 cos Z .4.12) A Z Dξ e − RT 0 µF ξ ν ) ˙ dτ ( 21 ξ ·ξ−ieξ µν P = det 1 2   sin(Z) . hAµ (k1 )A5ν (k2 )i Z Z∞ 1 dT −m2 T e Dx Dψ = 2 T 0 ( ZT  ) i 1 2 1 ˙ × exp − dτ x˙ + ψ · ψ + ex · F · x˙ − ieψ · F · ψ 4 2 2 0 .12) also become field dependent [6. axialvectors as well as a constant external (vector) field. It does not occur in the antiperiodic case since here the supersymmetry is broken by the boundary conditions.2 of part I we gave explicit expressions for the generalized e. (2.10) The free Gaussian path integral determinants (2.9. the formalism can be applied as well to the calculation of the effective action itself in the inverse higher derivative expansion (see [6.10). Worldline calculation of the vector–axialvector amplitude in a constant field As an explicit example. 2 1 µν µ ν hξ (τ1 )ξ (τ2 )i = G˙B (τ1 . (2. then. g = 4 F · F .35. Schubert / Nuclear Physics B 585 (2000) 429–442 1 µν hψ µ (τ1 )ψ ν (τ2 )i = GF (τ1 . According to the above we can represent this amplitude as follows. which we have to know to calculate one-loop processes involving any numbers of (abelian) vectors. The above trigonometric expressions should be understood as power series in the field strength tensor.39]). This is all. F 1 1 e Maxwell invariants f = 4 F · F .

(4. and the Grassmann zero mode integrated out using Eq. F F . If one multiplies the right-hand side by k1 then the integrand of the vector vertex operator becomes a total derivative in τ1 . In the next step.1). Nothing analogous holds for the axialvector vertex operator.8) and (3. both path integrations are performed using the field-dependent Wick contraction rules Eqs. (2. in the exponent of Eq. so that we can set D = 4 in its evaluation.3) d 4 ψ0 eieT ψ0 ·F ·ψ0 ψ0µ ψ0ν ψ0κ ψ0λ = εµνκλ . so that the integral vanishes by periodicity. G do not occur for the periodic case we delete the subscript “B” in the following. ZT ZT dτ ψ(τ ) · F · ψ(τ ) = T ψ0 · F · ψ0 + 0 dτ ξ(τ ) · F · ξ(τ ).e.10). Schubert / Nuclear Physics B 585 (2000) 429–442 ZT ×  dτ1 x˙µ (τ1 ) + 2iψµ (τ1 )k1 · ψ(τ1 ) eik1 ·x1 0 ZT ×  dτ2 ik2ν + 2ψν (τ2 )x(τ ˙ 2 ) · ψ(τ2 ) eik2 ·x2 .5). εµνκλ Fµν Fκλ = −(eT )2 F · F d 4 ψ0 eieT ψ0 ·F ·ψ0 = − 2 Z eµν . (3.436 C.2) 0 Thus for the case at hand the Grassmann zero mode integral can appear in the following three forms. i.4) we can rewrite. the zero-modes of both path integrals are separated out according to Eqs. (2. Z (eT )2 e. This amplitude is finite. (2. This results in the following parameter integral representation for the v