Nuclear Physics B 586 (2000) 3–38

www.elsevier.nl/locate/npe

Complete two-loop dominant corrections
to the mass of the lightest CP-even Higgs boson
in the minimal supersymmetric standard model
Jose Ramón Espinosa a , Ren-Jie Zhang b,1
a Instituto de Matemáticas y Física Fundamental (CSIC), Serrano 113 bis, 28006 Madrid, Spain
b Department of Physics, University of Wisconsin, 1150 University Avenue, Madison WI 53706, USA

Received 29 March 2000; accepted 26 June 2000

Abstract
Using an effective potential approach, we compute two-loop radiative corrections to the MSSM
lightest CP-even Higgs boson mass Mh0 to O(αt2 ) for arbitrary left-right top-squark mixing and
tan β. We find that these corrections can increase Mh0 by as much as 5 GeV; assuming a SUSY scale
of 1 TeV, the upper bound on the Higgs boson mass is Mh0 ≈ 129 ± 5 GeV for the top quark pole
mass 175 ± 5 GeV. We also derive an analytical approximation formula for Mh0 which approximates
our numerical results to a precision of . 0.5 GeV for most of the parameter space and suitable
to be further improved by including renormalization group resummation of leading and next-toleading order logarithmic terms. Our final compact formula admits a clear physical interpretation:
radiative corrections up to the two-loop level can be well approximated by a one-loop expression with
parameters evaluated at the appropriate scales, plus a smaller finite two-loop threshold correction
term.  2000 Elsevier Science B.V. All rights reserved.
PACS: 12.60.Jv; 14.80.Cp; 11.10.Gh
Keywords: Supersymmetric standard model; Higgs physics

1. Introduction
It is difficult to overestimate the importance of the experimental discovery of the Higgs
boson. It would not only help us to elucidate the dynamics responsible for electroweak
symmetry breaking but it will most probably offer also an important clue as to the nature
of the Physics beyond the Standard Model (SM).
The paradigm for this new Physics, granted that a fundamental scalar drives electroweak
symmetry breaking, is the Minimal Supersymmetric Standard Model (MSSM) (for reviews
1 After September 2000: School of Natural Science, Institute for Advanced Study, Princeton, NJ 08540, USA.

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 4 2 1 - 1

4

J.R. Espinosa, R.-J. Zhang / Nuclear Physics B 586 (2000) 3–38

see, e.g., [1,2]): the most economical extension of the Standard Model that incorporates
(softly-broken) Supersymmetry (SUSY). In spite of the uncertainties related to the
origin of supersymmetry breaking (and therefore of the masses of the so far undetected
supersymmetric particles), it is well known that the MSSM predicts the existence of a
light Higgs particle with mass below about 135 GeV (this bound depends sensitively on
the top quark mass one uses; our present calculation intends to set a precise and firm
bound). Unlike the case of the Standard Model (in which the mass of the Higgs boson
is an unknown parameter), the mass of the light CP-even Higgs boson of the MSSM is
calculable as a function of other masses of the model. A precise calculation of that mass is
of prime importance for Higgs searches at LEP, Tevatron and the LHC, and is the topic of
this paper.
We recall at this point that the Higgs sector of the MSSM consists of two SU(2) doublets,
H1 (which gives mass to down-type quarks and charged leptons) and H2 (which gives
mass to up-type quarks). The vacuum expectation values (v1,2 ) of these doublets break
the electroweak symmetry, after which, the Higgs spectrum contains two CP-even scalars
(h0 and H 0 ; with mh0 6 mH 0 ), one CP-odd pseudoscalar (A0 ) and a pair of charged
Higgses (H ± ). At tree-level, the masses, couplings and mixing angles of these particles
are determined by one unknown mass parameter (say, mA0 ) and the parameter β, which
measures the ratio v2 /v1 (≡ tan β). In the limit mA0  MZ all the Higgs particles except
h0 have masses ∼ mA0 and rearrange in a complete SU(2) doublet almost decoupled from
electroweak symmetry breaking, while h0 remains light with m2h0 6 MZ2 cos2 2β and has
SM properties. This bound (which applies for any value of mA0 and is saturated for mA0 
MZ ) is extremely important: it represents the limit that experimental bounds should reach
to falsify the MSSM. In fact, the present experimental bound [3] from LEP, including the

latest data with up to s = 202 GeV, is mh0 & 107.7 GeV (for large mA0 case in which the
SM limit is applicable; the limit falls to ∼ 91 GeV for smaller mA0 ), which is well above
this bound. This is not yet conclusive evidence against the MSSM because it does not take
into account the radiatively corrected form of the mass bound.
Radiative corrections to m2h0 have been computed using three different techniques (or
combinations of them): effective potential method [4–13], direct diagrammatic calculation
[14–22] and effective theory (or renormalization group) approach [10,13,23–30]. The full
one-loop radiative corrections to mh0 have been computed diagrammatically. The most
important of these corrections come from top quark/squark loops and are given by
1m2h0 =

m2t˜
3m4t
ln
,
2π 2 v 2 m2t

(1)

where mt is the top quark mass, mt˜ an average top-squark mass and v 2 ≡ v12 + v22 =
(246 GeV)2 . This correction can be very large if mt˜  mt , and in such case mh0 can evade
easily the current experimental lower bound. This important O(αt ) logarithmic correction
to the dimensionless ratio m2h0 /m2t [here αt ≡ h2t /(4π), where ht is the top-quark Yukawa
coupling] can be most easily reproduced using renormalization group (RG) techniques. In
addition, there is a finite (non-logarithmic) correction which may also be important, and

J.R. Espinosa, R.-J. Zhang / Nuclear Physics B 586 (2000) 3–38

5

which depends on the details of the top-squark spectrum. This correction is (assuming
again for simplicity degenerate soft masses for the top-squarks)
!
3m4t
Xt4
Xt2
2

,
(2)
1mh0 =
2π 2 v 2 m2˜
12m4t˜
t
where Xt = At + µ cot β is the top-squark mixing parameter, At the soft trilinear coupling
associated to the top-Yukawa term in the superpotential and µ the supersymmetric Higgs
mass parameter. Correction (2) is maximized for Xt2 = 6m2t˜ (the so-called “maximalmixing” case). When using one-loop equations like (1) and (2) to compute the Higgs mass
one has to decide whether to use on-shell (OS) or running values for the mass parameters
that enter such formulae (and if the latter, at which scale to evaluate them). The difference
between two such choices is of higher order, but can be non-negligible, especially because
of the m4t -dependence of 1m2h0 . Although RG techniques can be used to make an educated
guess of the scale at which those mass parameters should be evaluated (see, e.g., [28,29]),
a precise answer to such questions could only be unambiguously given by a two-loop
calculation like the one we perform in this paper.
At two loops, radiative corrections to m2h0 depend not only on the large top-Yukawa
coupling but also on the QCD coupling g3 . It is reasonable to expect that the dominant
two-loop corrections will be of order O(αs αt [ln(m2t˜ /m2t )]k ) and O(αt2 [ln(m2t˜ /m2t )]k ), k =
0, 1, 2. Terms with k = 2 are the two-loop leading logarithmic contributions and can be
obtained by RG techniques using one-loop RG equations; no true two-loop calculation
is required and RG resummation will take into account such leading-logarithmic (LL)
corrections to all loops. The k = 1 terms are the two-loop next-to-leading-logarithmic
(NTLL) corrections, which can be obtained (and resummed to all loops) with two-loop RG
equations. Finally, the two-loop non-logarithmic terms (k = 0) can be interpreted in the
effective theory language as threshold corrections (at the supersymmetric scale set by the
mass of the top-squarks) and require a genuine two-loop calculation; they simply cannot
be obtained from RG arguments.
The status of these higher-loop calculations of the radiatively corrected m2h0 is the
following. Higher-order logarithmic corrections were included in studies which used
RG techniques almost since the dramatic impact of radiative corrections on mh0 was
first recognized. Hempfling and Hoang [11] were the first to perform a genuine twoloop calculation of mh0 which also included non-logarithmic terms. They computed the
dominant two-loop radiative corrections [to O(αs αt ) and O(αt2 )] in the case tan β  1 and
zero top-squark mixing. Their computation also included the most important logarithmic
corrections, which could be alternatively incorporated by RG resummation from one-loop
results, as done, e.g., in Ref. [10]. In this last paper it was also pointed out that by a
judicious choice of the renormalization scale at which to evaluate one-loop corrections, the
higher order logarithmic corrections could be automatically taken into account. A similar
idea was later implemented in [28–30] to write down simple analytical approximations for
the radiatively corrected m2h0 , obtained by iterative integration of RG equations.
Besides being limited to a particularly simple value of tan β, the calculation in Ref. [11]
missed the sizable impact of non-zero top-squark mixing in two-loop effects, that is, higher

6

J.R. Espinosa, R.-J. Zhang / Nuclear Physics B 586 (2000) 3–38

order corrections to the contribution written down in Eq. (2). Such corrections were first
included to order O(αs αt ) in the diagrammatic calculation [21], and by the effective
potential method in Ref. [12]. The effect of these corrections is to shift the values of
Xt that give maximal mixing, change the corresponding Higgs mass by up to ∼ −10
GeV 2 and introduce an asymmetry in the dependence of mh0 with the sign of Xt . This
two-loop top-squark mixing dependent correction was also explicitly isolated recently
by the present authors in Ref. [13], which uses effective potential plus RG techniques.
Besides confirming the independent diagrammatic results of Ref. [21] we clarified the
relation of these calculations to previous ones (in particular matching results expressed
in different renormalization schemes; see also [31]). We also derived a compact formula
for the Higgs mass (in the spirit of [28–30]) which took into account the most important
radiative corrections, and used RG techniques to include in a compact way two-loop LL
and NTLL corrections. With the O(αs αt ) radiative corrections organized in this way, we
find that the mixing-dependent genuine two-loop threshold corrections are generally small
<
3 GeV).
(∼
Nevertheless, the large computing effort just reviewed did not exhaust the potentially
important radiative corrections: the two-loop O(αt2 ) top-squark-mixing-dependent corrections to mh0 remained unknown to this day, while it is clear that they could compete in
principle with the O(αs αt ) contributions. The purpose of this paper is to complete the calculation performed in [11–13] by using effective potential techniques (plus RG techniques)
to compute such O(αt2 ) contributions for general top-squark mixing parameters and any
value of tan β. The results in this paper can be considered the most complete and accurate
approximation to mh0 presented in the literature.
The structure of the paper is the following: Section 2 describes the strategy of our
calculation and presents some analytical formulae for mh0 , obtained in the limit of
mt˜  mt . Section 3 goes one step ahead implementing the RG-improvement of such
approximations and, in doing so, clarifies the organization of the higher order radiative
corrections. This procedure is not only important to provide a clearer physical picture
in connection with the effective field theory but also to classify those corrections
calculated in Section 2 into a numerically dominant and compact part plus smaller finite
threshold correction terms. In Section 4 we present our numerical results for the Higgs
mass, illustrate the size of the new corrections and check the validity of our analytical
approximation formulae. We draw some conclusions in Section 5.
Several appendices are devoted to technical details of different aspects of the calculation.
Appendix D is worth special mention as it contains the two-loop O(αt2 ) MSSM effective
potential used as starting point of our calculation and first computed in this paper.

2 This correction is relative to the one-loop mass using on-shell parameters. The size of the correction would
be much smaller if running parameters are used in the one-loop formula (1).

J.R. Espinosa, R.-J. Zhang / Nuclear Physics B 586 (2000) 3–38

7

2. CP-even Higgs boson masses to two-loop order
The momentum-dependent mass-squared matrix for the CP-even Higgs bosons of the
MSSM in the interaction eigenstate basis h1 , h2 is
"   

#
m2Z cβ2 + m2A0 sβ2 + 1M211 p2
− m2Z + m2A0 sβ cβ + 1M212 p2 

2 2
, (3)
Mh p =   

m2Z sβ2 + m2A0 cβ2 + 1M222 p2
− m2Z + m2A0 sβ cβ + 1M221 p2
where sβ ≡ sin β and cβ ≡ cos β. The mass parameters mZ and mA0 are the (scaledependent) running masses of the Z-boson and CP-odd Higgs boson A0 ; they are related to
the on-shell masses MZ and MA0 (we use capital letters to distinguish on-shell parameters
from running ones) by 

T
MZ2 ,
m2Z = MZ2 + ReΠZZ 

T1
T2
(4)
m2A0 = MA2 0 + ReΠAA MA2 0 − sβ2 − cβ2 ,
v1
v2
T is the transverse part of the Z-boson self-energy and Π
0
where ΠZZ
AA the A -boson selfenergy, T1 , T2 are the tadpoles of the CP-even (real) fields h1 , h2 . Their explicit one-loop
expressions can be found, e.g., in Ref. [22].
In (3), 1M2 stands for the contributions from radiative corrections. They are  

Ti
(5)
1M2ij p2 = −Πij p2 + δij , i, j = 1, 2,
vi

where Πij is the self-energy matrix of the Higgs fields h1 and h2 . The masses, mh0 , mH 0 ,
of the two CP-even Higgs bosons are then obtained from the real part of the poles of the
propagator matrix,  

(6)
Det m2h0 ,H 0 1 − M2h m2h0 ,H 0 = 0.
The radiatively corrected mixing angle α is obtained as the angle of that rotation which
diagonalizes M2h (for some choice of p2 , say p2 = m2h0 ): 

2 M2h 12  

.
(7)
tan 2α =
M2h 11 − M2h 22
Computing Higgs boson masses to a certain order of perturbation theory then requires
calculating the self-energies and tadpoles in Eqs. (4) and (5) to that order.
In the effective potential approach [4–13], self-energies and tadpoles can be calculated
as derivatives of the Higgs potential V according to:  

2 



∂V (h1 , h2 )

∂ V (h1 , h2 )

,
Πij (0) = −
.
(8)
Ti = −

h1 =v1

and its radiatively corrected mass has a very simple expression . In the limit MA0  MZ . α → β − π/2 + O(m2Z /m2A ). that is. h1 =v1 ∂hi ∂hi ∂hj h2 =v2 h2 =v2 Note that the Π ’s obtained from derivatives of V have zero external momentum. the lightest CP-even Higgs state lies along the direction of the breaking in field space [32].

Espinosa. related to the self-energies of h1. V is the projection of V (h1 . h2 → hsβ ). Πhh (p2 ) is the light Higgs self-energy at external momentum p. Then V (h) can be expressed as a function of mt V (h) = V (h1 √ using h → mt 2/(ht sβ ). h2 ) along the light Higgs h = h1 cβ + h2 sβ : → hcβ . (10) Notice that the self-energy difference in (9) involves non-zero external momentum and would require a diagrammatic two-loop calculation. In that case.-J. We decompose V in its nth-loop pieces Vn (explicitly given in Appendix D) as V = V0 + V1 + V2 . where V2s is the O(αs αt ) part and V2t the O(αt2 ) one. The tree-level part V0 is the only one in which we keep non-zero electroweak gauge couplings. realizing that at tree level mh0 depends only on electroweak gauge couplings while its dependence on ht appears only at one-loop. R.2 by Πhh (p2 ) ≡ Π11 (p2 )sα2 + Π22 (p2 )cα2 − 2Π12 (p2 )sα cα .8 J. throughout this paper we work in the approximation of neglecting in the radiative corrections all couplings other than ht or g3 .R. we can write . 2 v dmt (9) which is exact up to corrections of order O(m4Z /m2A0 ). We approximate the one-loop part V1 by its O(αt ) piece coming from top quark/squark loops. The two-loop part V2 is approximated by V2s + V2t . However. (9). 3 In Eq. Zhang / Nuclear Physics B 586 (2000) 3–38 Mh20    4m4t d 2 = 2 V − ReΠhh m2h0 + ReΠhh (0). Next.

.

(11) Πhh m2h0 − Πhh (0) ' m2h0 2 Πhh (p2 ).  d .

.

and at the same time simplifies the structure of the radiative corrections. we can easily express the partial derivatives of V in terms of the total derivative in (9). . and we can therefore use the Z A0 approximation 1m2 0 ' 1M211 cβ2 + 1M222 sβ2 + 21M212 sβ cβ . both for a better understanding of the numerical results and for practical applications. Observing h A that the potential V depends on the fields h1 and h2 only through (field-dependent) top quark mass and the off-diagonal elements of the top-squark mass-squared matrix and using (8). dp p 2 =0 which gives rise to O(αt2 ) contributions from the one-loop O(αt ) self-energy Πhh . In Section 4 we present the numerical results of such procedure for the two-loop potential with O(αs αt ) and O(αt2 ) corrections included. focusing on the particles relevant for the radiative 3 This formula can be proved as follows: If m2  m2 . up to higher order terms in m4Z /m2 0 . The general expression for the O(αs αt ) potential was first computed in [12] while the complete O(αt2 ) terms were still missing. the case one should be able to discard to rule out the MSSM). to derive an analytical expression for the light Higgs mass in the case of a large hierarchy between the supersymmetric scale and the electroweak scale (say when the SUSY scale is of order 1 TeV). We consequently assume now that all supersymmetric particles have roughly the same mass MS  MZ . avoiding the proliferation of a multitude of different supersymmetric thresholds. α → β − π/2. It is useful. Both contributions to V are given in Appendix D. A similar formula was already used in [11]. Such limit is interesting because it maximizes the radiative corrections to m2h0 (so that it corresponds to the most pessimistic scenario for Higgs searches. In more detail.

we find that.R. Zhang / Nuclear Physics B 586 (2000) 3–38 9 corrections to mh0 . The two eigenvalues and mixing angle of the top-squark squared-mass matrix are then m2t˜ = m2t˜ + mt Xt . Dm V 1 = ln 2 + Xt − 2π 2 v 2 12 mt ( M2 M2 αs m4 m2 m2 2 V2s = 3 2t ln2 2S − 2 ln2 S2 + 2 ln2 t2 + ln t2 − 1 Dm π v Q Q Q mt   M2 m2 + −1 + 2 ln S2 + 2 ln t2 Xˆ t Q Q )   ˆ t3  Xˆ t4 MS2 X 2 + 1 − 2 ln 2 − . (13) Dm v dm2t the different parts entering (9) are 2 V0 = m2Z cos2 2β. 1 m2t˜ = m2t˜ − mt Xt .J. Dm   MS2 3m4t Xˆ t4 2 2 ˆ . R. In this situation. using the operator   4m4 d 2 2 ≡ 2t . which broadly corresponds to the case of a common heavy SUSY scale. Xˆ t + Q 3 12 ( 2 4 2 MS2 3α m2t m t t 2 2 MS 2 mt V2t = − 6 ln ln − 3 ln 9 ln Dm 16π 3 v 2 Q2 Q2 Q2 Q2   M2 + 2 3f2 (µ) ˆ − 3f1 (µ) ˆ − 8 ln 2S mt   2 M π2 + 6µ ˆ 2 1 − ln S2 − 2(4 + µˆ 2 )f1 (µ) ˆ + 4f3 (µ) ˆ − Q 3   2 MS 2 2 2 ˆ + (4 − 6µˆ )f1 (µ) ˆ Xˆ t2 + (33 + 6µ ˆ ) ln 2 − 10 − 6µˆ − 4f2 (µ) Q   ˆ4 M2 Xt + −4(7 + µˆ 2 ) ln S2 + 23 + 4µˆ 2 + 2f2 (µ) ˆ − 2(1 − 2µˆ 2 )f1 (µ) ˆ 4 Q   2 M 1 + sβ2 Xˆ t6 ln S2 − 1 2 Q (14) (15) (16) . 2 (12) We also take the same mass MS for the gluino and the pseudoscalar Higgs [this means in particular that we can use Eq. In principle we admit the possibility that the µ parameter could be smaller than MS . (9) for the light Higgs boson]. in which case we expect that one chargino and two neutralinos will have masses ∼ |µ| below the common supersymmetric threshold.-J. Espinosa. we take equal soft masses MQ e = MU e = MS for the top-squarks (with diagonal masses m2t˜ ' MS2 + m2t ). 2 1 st2 = ct2 = .

It is in this respect convenient to write down the two-loop expressions just obtained in the particular case in which all mass parameters in the one-loop correction are the OS ones. The relationships between running and OS parameters are listed in Appendix C.R.1953256. the magnitude of the two-loop corrections will change. f2 (0) = 1. (9) is given by (see Appendix C)     h2t Xˆ t2 m2t 2 2 2 . Zhang / Nuclear Physics B 586 (2000) 3–38 " MS2 m2t 4π 2 − 4 ln − 3 + 60K + Q2 Q2 3 m2t     2  MS2 MS 2 ˆ + 12 − 24K − 18 ln 2 Xt − 3 + 16K − 3 ln 2 4Xˆ t Yˆt + Yˆt2 Q Q     2 2 M 11 MS ˆ 4 ln 2 Xt + 4 + 16K − 2 ln S2 Xˆ t3 Yˆt + −6 + 2 Q Q #)     2 MS MS2 1 14 19 2 2 4 2 + 24K − 3 ln 2 Xˆ t Yˆt − + 8K − ln 2 Xˆ t Yˆt + . (9). so that it should be possible to choose the scale in such a way that the bulk of the corrections is transferred to the one-loop terms (which depend on the scale implicitly). It is evident that for different values of the renormalization scale. Therefore.10 J. is renormalization-scale independent (up to two-loop order). This scale independence is at the root of the RG-resummation procedure discussed in the next section. we obtain for the two-loop correction to m2h0 : DR -scheme and . Finally. (19) m 0 s 3 ln 2 + 2 − Re −Πhh (mh0 ) + Πhh (0) = 2 16π 2 h β Q The parameters that appear in these expressions are running parameters. We also use the following non-singular functions of µˆ µˆ 2 ln µˆ 2 . Yˆt = Aˆ t − µˆ tan β. the magnitude and relevance of the two-loop corrections depends on the definition of the mass parameters that enter the one-loop corrections. This is also useful to compare with explicit diagrammatic calculations. and (see Appendix A) K ' −0. f3 (1) = −9/4. R. In fact. 3 Q 12 2 Q + cβ2 3 ln2 MS2 + 7 ln (17) The notations used are Xˆ t = Aˆ t + µˆ cot β. 1+ f2 (µ) 1 − µˆ 2 1 − µˆ 2   (−1 + 2µˆ 2 + 2µ π2 ˆ 4) 2 2 2 2 2 − µˆ ln µˆ . the correction for non-zero external momentum in Eq. it can be checked that the (physical) Higgs mass. Espinosa. 1 − µˆ 2   1 µˆ 2 2 ˆ = ln µˆ . ˆ = ln µˆ ln(1 − µˆ ) + Li2 (µˆ ) − f3 (µ) 6 (1 − µˆ 2 )2 ˆ = f1 (µ) (18) with f1 (0) = 0. f3 (0) = −π 2 /6 and f1 (1) = −1. evaluated in the satisfy MSSM RG equations.-J. with reduced parameters zˆ ≡ z/MS . given by Eq. f2 (1) = 1/2. Using them. as it should.

although roughly speaking the top-Yukawa correction has a small pre-factor 3/16 in comparison with the QCD correction. 3 12 2 3 11 (20) We emphasize that this expression gives the two-loop corrections when the one-loop contribution (15) is expressed in terms of OS parameters.-J. (20) are worth commenting. the result quoted for that finite term in Ref. 8. we will see that for two-loop topsquark-mixing-dependent corrections of (20). However. [11]. that term is sensitive to the value of the parameter µ. the O(αt2 ) finite term is different. (21) 1mh0 1-loop = 2 Mt˜ 12 Mt˜ 8π 2 MW Mt Several features of Eq. This dependence on Yt originates from the H−t˜−b˜ and H−t˜−t˜ diagrams of Fig. Zhang / Nuclear Physics B 586 (2000) 3–38 1m2h0 ( ) 2 MS2 MS2 2 3 4 αs m4t 2 MS ˆ ˆ ˆ = 3 2 −3 ln − 6 ln 2 + 6Xt − 3 ln 2 Xt − Xt π v 4 m2t mt mt (  2 MS2 2 7 π 2 3αt m4t 2 MS − 3µˆ 2 + 13 ln 3 ln sβ − − + 16π 3 v 2 2 3 m2t m2t  2  − 11 − µˆ 2 + 3µˆ 4 f1 (µ) ˆ − 3 1 − µˆ 2 ln 1 − µˆ 2 + 3f2 (µ) ˆ + 4f3 (µ) ˆ   2 13 4π + + cβ2 60K + 2 3    2 M 73 69 + 9µ ˆ 2 + f1 (µ) + 24K Xˆ t2 ˆ − 7f2 (µ) ˆ − cβ2 + 3sβ2 ln 2S + 2 2 mt   sβ2 1 61 + −26 − 9µˆ 2 + 3f1 (µ) ˆ + 3f2 (µ) ˆ + cβ2 Xˆ t4 + Xˆ t6 6 2 2     Xˆ 4 + 3(1 − µˆ 2 ) (2 − 3µˆ 2 )f1 (µ) ˆ − (1 + 3µˆ 2 ) ln(1 − µˆ 2 ) Xˆ t2 − t 6    √ 2π + cβ2 (3 − 16K − π 3)(4Xˆ t Yˆt + Yˆt2 ) + 16K + √ Xˆ t3 Yˆt 3     ) √ π 4 7 2 2 4 2 + 8K + √ Xˆ t Yˆt + − + 24K + π 3 Xˆ t Yˆt − . R. we see that radiative corrections no longer depend on At and µ in the combination Xt that appears through the off-diagonal entry of the top-squark mass matrix: besides the explicit dependence on the parameter µ already noticed. to compare with the result of Ref. (20) to tan β  1 and zero At . [11] is inside the range we would find by varying µˆ 2 from 0 to 1. First. the top Yukawa contributions have opposite . contrary to what is stated in Ref. "  OS 2   # Mt˜2   3g 2 Mt2 Xt 1 XtOS 4 2 OS − ln 2 + . In fact. and the impact of this µ-dependence on the final Higgs mass is quite small. Nevertheless. the quantity Yt also introduces a different combination of At and µ. In particular. Espinosa. that is.R. [11]. Third. this does not guarantee that the new contributions will be negligible compared to the QCD one. Second. we find the same logarithmic terms. if we restrict Eq.J.

The Standard Model MS top quark mass mt and the Higgs VEV v are related to the on-shell mass Mt and MSSM VEV v by [cf. Notice that. (B. (C. v (Q) = v (Q) 1 + (25) 32π 2 t We also use one-loop LL solutions of the SM RG equations to relate these parameters at different scales . [13] to the O(αs αt ) two-loop corrections. we show in the following how the use of renormalization group techniques [23–30] allows us to write the previous complicated corrections [see Eq. Eqs.17) and (B. (20). We use the following equations to relate supersymmetric running parameters at different scales [cf. Xt (Q) = Xt (Q ) − S t t t t β β 16π 2 3 3 Q2 0 (22) (23) where we have used At = Xt sβ2 + Yt cβ2 . Renormalization group resummation Before illustrating in Section 4 the impact of the newly computed corrections on the Higgs mass. with the most important corrections of Eq.18)]: m2t˜ (Q) ( )   02  1 3 16 Q g 2 − h2 Xˆ 2 s 2 + Yˆt2 cβ2 + cβ2 + 2 − 2µˆ 2 ln 2 . By a convenient (and physically well motivated) choice of the scale at which to evaluate running parameters in the one-loop mass correction one can absorb large logarithms in Eq.2) and (C. (20)] in a simpler and more transparent way.10). = m2t˜ (Q0 ) 1 + 16π 2 3 3 2 t t β Q   1 Q02 16 2 2 2 2 g M + 3h (X + X s + Y c ) ln . In the next section. R. Eqs. 6). (20) reshuffled in a RG-motivated one-loop formula.12 J.-J.R. it is sufficient to use these one-loop LL approximations to the full RG evolution because we are concerned with parameters that appear in a one-loop order term. we will follow RG methods and reorganize these corrections in the effective theory language. while at the same time it clarifies the connection to the RG programme. (24) 6π 2 Q 32π 2 Q " # h2t sβ2 2 2 2 Xˆ . We already applied this idea in Ref. The RG evolution of the parameters is given by the corresponding one-loop RG functions listed in Appendix B. 3. to the order we work. from which relevant terms can be easily identified] "    # h2t sβ2 g32 m2t m2t 2 2 m t (Q) = Mt 1 − 4 − 3 ln 2 + 8 − 3 ln 2 . which can be used to improve the precision of the mass formula by resummation of higher order corrections. Zhang / Nuclear Physics B 586 (2000) 3–38 signs as that of the QCD corrections and could be as much as 60% of the latter (see Fig. Espinosa. A2t + µ2 = Xt2 sβ2 + Yt2 cβ2 and m2H2 + µ2 = m2A0 cβ2 .

16π 2 2 Q " # 3h2t sβ2 Q02 v 2 (Q) = v 2 (Q0 ) 1 + ln 2 . Q∗1 = e−1/3 mt ' 0. Q∗2 = e1/3mt ' 1. Espinosa. (31) Qt˜ = Qth = mt˜. Qt = mt mt˜. . Zhang / Nuclear Physics B 586 (2000) 3–38 13 " #   1 3 2 2 Q02 2 1+ 8g3 − ht sβ ln 2 . that the uncertainty found there in the definition of the scales Q0t and Qt˜ is here resolved by the need of absorbing the new radiative corrections. 16π 2 Q m 2t (Q) = m 2t (Q0 ) (26) (27) Using the above equations. which is one of the main results of this paper Mh20 = MZ2 cos2 2β + 3m 4t (Qt ) 2π 2 v 2 (Q∗1 ) ln m2t˜ (Qt˜) m 2t (Q0t ) (2) 2 2 + 1(1) th mh0 + 1th mh0 . R. Xt (Qth ) − 1th mh0 = 12 2π 2 v 2 (Q∗2 ) (28) (29) and the two-loop threshold correction reads   αs m4 7 1 1 (2) 1th m2h0 = 3 2t −2Xˆ t − Xˆ t2 + Xˆ t3 + Xˆ t4 − Xˆ t5 π v 3 12 6 ( 4 1 3αt mt ˆ + R2 (µ) ˆ Xˆ t2 + R4 (µ) ˆ Xˆ t4 − sβ2 Xˆ t6 R0 (µ) + 3 2 2 16π v  2  9 4π ˆ t2 − (3 + 16K) 4Xˆ t Yˆt + Yˆt2 + (15 − 24K)X + cβ2 60K − + 2 3   25 ˆ 4 14 3ˆ ˆ + 24K Xˆ t2 Yˆt2 − Xt + (4 + 16K)Xt Yt + 4 3   ) 19 + 8K Xˆ t4 Yˆt2 . We see. (32) It is a non-trivial check of our calculation that the values of the scales (32) required to re-absorb the large ln(MS2 /m2t ) logarithms in the two-loop corrections are consistent with the ones obtained in [13] for the QCD corrections alone.R. − (30) 12 We have used the short-hand notation  9 π2 + 6µˆ 2 − 11 + 2µ ˆ =− − ˆ 2 f1 (µ) ˆ + 3f2 (µ) ˆ + 4f3 (µ). R2 (µ)   2 ˆ = 6 + µˆ 1 + f1 (µ) ˆ + f2 (µ) ˆ .J.4 mt . we find the following compact expression for the Higgs boson mass.-J. in particular. R4 (µ) The scales required in (28.29) are √ Q0t = (mt m2t˜ )1/3 .7 mt . ˆ R0 (µ) 2 3   ˆ = −11 − µˆ 2 6 + 6f1 (µ) ˆ + 10f2 (µ) ˆ . The one-loop threshold correction is   3m 4t (Qth ) ˆ 2 Xˆ t4 (Qth ) (1) 2 .

Espinosa. R4 (0) = 6. R2 (1) = −16. Ref.and two-loop contributions to βλ are approximated by [neglected all couplings other than the strong gauge coupling g3 and the SM top Yukawa coupling gt (≡ ht sβ )]  3gt2 −gt2 + λ . below that scale. Making a loop expansion of βλ in (35) and a further expansion around a particular value of Q (say the low energy limit of the running interval.g.. 3 2 and for µ ' MS : R0 (0) = R2 (0) = −11. The light Higgs quartic coupling λ at MS consists of a tree-level part plus higher-order threshold corrections which arise from the heavy decoupling supersymmetric particles. we can write R0 (1) = 7 − ZQt˜ λ(Qt ) = λ(Qt˜) − βλ d ln Q2 . [30]). say Q = mt . 3 2 It is perhaps convenient to make more explicit the connection between our results and those obtained in the RG approach (see. due to the fact that we have kept free the µ parameter. Explicitly. This is the natural definition in the RG approach. (34) R4 (1) = .e.14 J. using the SM RGEs. defining βλ ≡ dλ/d ln Q2 . at that scale λ is related to the physical Higgs mass. Zhang / Nuclear Physics B 586 (2000) 3–38 2 We still find a somewhat complicated expression for the threshold correction 1(2) th mh0 .. the effective theory is the SM.-J. the resulting threshold correction is much simpler than the general case (30) and contains no more logarithms. for µ  Ms we find π2 3 − . To be concrete. R. (35) Q=Qt We use a special notation for the high and low scales between which we run λ to distinguish them from other definitions of mt and mt˜ that appear in the paper. In both cases. These quantities are defined by: Qt ≡ mt (Qt ). Qt˜ ≡ mt˜(Qt˜). βλ = 2 2 (16π ) (1) βλ = (38) . (33) π2 13 . e. Expressions for the two limiting cases of heavy µ (µ ' MS ) and light µ (µ  Ms ) can be readily derived. (37) λ(Qt ) ' λ(Qt˜) − βλ(1) (Qt ) + βλ(2)(Qt ) ln t − 2 Qt 2 d ln Q Qt where the one.R. This procedure should reproduce all the logarithmic corrections we have found. we obtain to the two-loop order   Q ˜ 1 dβλ(1) Q˜ (Qt ) ln2 t + · · · . Qt ). it can be evolved down to the electroweak scale. they are the running masses evaluated at a scale equal to the corresponding mass. 2 8π  2gt4 (2) 15gt2 − 16g32 . let us assume |µ| = MS so that all supersymmetric particles (including charginos and neutralinos) have masses of order MS . More explicitly. (36) i.

R. we can combine both factors together using (25) to write the simpler formula Mh20 = λ(Qt )v 2 (Q∗1 ). affect the large logarithmic terms only through multiplication of λ(Qt ). dβλ /d ln Q2 can be calculated from the one-loop RG evolution of gt  gt2 dgt2 = 9gt2 − 16g32 . and rewrites them in a very convenient way. (39) d ln Q2 32π 2 Once λ(Qt ) is obtained from (37). and not at zero external momentum. (20). λ= t 2 Xt (Qth ) − t δth 12 8π (2) λ= δth 2 1(2) th mh0 . absorbing the large (logarithmic) two-loop effects in the one-loop correction and leaving behind two-loop threshold corrections which are numerically small. the large LL and NTLL corrections to Mh20 arise solely from λ(Qt ). According to (37). It is now straightforward to show perfect agreement of Mh0 as obtained from the above expression with our results (28)–(30). It is therefore clear that it is enough for our purposes to know these correction factors at one-loop order. The connection to the effective theory language clarifies the origin of the different terms in (28). the bulk of which is transferred to the RG-reshuffled one-loop threshold correction of Eq. as we will see in the next Section. in accordance with (32). . Zhang / Nuclear Physics B 586 (2000) 3–38 15 We note that for a correct two-loop computation it is necessary to retain also the λgt2 term (1) (1) in βλ because λ gets one-loop contributions proportional to gt4 . (29). Note that this applies in particular to the sizable top-squark-mixing-dependent corrections of Eq. Its physical content is therefore similar to the correction (11) in our effective potential approach. (42) λ(Qt˜) = g12 + g22 cos2 2β + δth 4 with   3g 4 (Qth ) ˆ 2 Xˆ 4 (Qth ) (1) . (43) v2 To summarize. we find full agreement between our approximate formula (28) for the Higgs boson mass and the RG-improved mass calculated in the RG (or effective theory) approach. coming from v 2 (Qt ) and the wave-function correction factor. All logarithmic corrections up to two-loops are exactly reproduced while the finite part agrees if one uses as boundary condition at the SUSY scale  1 (1) (2) λ + δth λ. Additional radiative contributions in (40). to two-loop order. Q∗1 (41) = e−1/3 Q with t . we extract the physical Higgs mass using the SM relation [33]:   g2 (40) λ(Qt )v 2 (Qt ) = Mh20 1 − t 2 . Based on this observation. Espinosa. 8π This correction arises from wave-function renormalization and takes into account the fact that the physical mass is defined on-shell.-J. R.J.

we have to calculate the one-loop self-energy diagrams for Higgses and top-squarks (the latter are collected in Appendix C). equipped with the corrections 1M2ij . g3 . (25)] to evaluate the parameters in (28) at their corresponding scales. we compute the lightest CP-even Higgs boson mass by solving Eq. note that our compact formula. To do this. the partial derivatives in these equations are replaced by finite differences in h1 . Eq. Zhang / Nuclear Physics B 586 (2000) 3–38 Knowing the boundary condition.118 to the MSSM DR running value. In this respect. Numerical results In this section we present numerical results from our two-loop study. From them we can determine the values of the corresponding running parameters at any renormalization scale Q. Next. [22]. we calculate in the MSSM the two-loop corrections to the CP-even Higgs mass matrix. Mt .16 J. 1M2ij . we shall concentrate on two-loop radiative corrections. from the twoloop potential (D. . (35) for λ] would correctly take into account the LL (but not the NTLL) terms to all loops. λ(Qt˜). Our results for the boundary condition λ(Qt˜) will still be useful in such a case. If MS turns out to be significantly larger than that (starting to be in conflict with naturalness criteria).5) and (D. one can integrate (35) numerically by solving a coupled set of differential equations (describing the two-loop evolution of λ. we can alternatively take as inputs the on-shell top-squark masses and mixing angle. 4. We also input tan β and µ parameters. 4 For the top-squark sector. and convert αs (MZ ) = 0. which is the following: we first take OS OS OS as inputs the on-shell mass parameters 4 MA0 . we vary the values of these fields by a finite amount and recalculate the field-dependent top-quark mass m2t = 12 h2t h22 and topsquark masses mt˜1 . the error made in neglecting logarithmic corrections of higher order is very small for SUSY scales of interest [below MS ∼ O(1) TeV]. evaluation of those parameters by means of a full numerical integration [similar to that in Eq.R.6) using Eq. mt˜2 . MQ e . Espinosa. (24). In what follows. gt ). find λ(Qt ) and use (41) to get the Higgs mass. MU e and At . (28). R. contains in fact logarithmic corrections of higher order. h2 .-J. It can be shown that these higher order logarithmic corrections do not match exactly the correct ones (obtained by the RG method) if we use simple one-loop approximations [like those given in Eqs. then one should revert to the numerical RG integration of λ to get a reliable estimate of the Higgs mass. Nevertheless. i. The final result will be the full RG-improved value of Mh0 and will resum LL and NTLL corrections to all loops [numerical integration includes all the terms from the expansion around Qt which were neglected in (37)]. Numerically. Finally. which has included complete radiative corrections from the dominant top quark/squark sector and the sub-dominant gauge/Higgs boson and neutralino/chargino sectors.5). the two-loop potential is reevaluated and their variations from the reference values [calculated at h21 + h22 = (246 GeV)2 ] are found. (8). However. as we will see in the next Section. mixing angle θt˜ from Eq. For the one-loop analysis we closely follow Ref.. (6). which has been found by requiring that logarithmic contributions are correctly reproduced up to two-loops only. We start by sketching the procedure for this analysis. With these new parameters.e. (B.

a complete two-loop calculation of mh0 would need ΠAA . and explicit two-loop calculation of the corresponding two-point functions are needed. We find good agreement (given the fact that they are two independent programs) between both one-loop (shown in dotted and dashed lines) and two-loop QCD corrected (shown in dot-dashed and solid lines) masses. Several approximations have been made to quantities in Eqs. corresponding results from our numerical analyses are shown in dashed and solid lines. dot-dashed lines show one-loop and two-loop O(αs αt ) results from the program FeynHiggs. This equivalence is easily understood since the effective potential. Zhang / Nuclear Physics B 586 (2000) 3–38 17 Fig. (a) MA0 = M3 = MQ S e e U OS = M OS = M = 1 TeV.6 and 20) are plotted. In this way we pick up the dominant radiative effects only. results for two values of tan β (1. 1.) Fig. (9) is correct for large mA0 and we can safely neglect the ΠAA corrections. Higgs boson mass Mh0 vs. [21]. Dotted. the on-shell top-squark mixing parameter Xˆ tOS . We choose two sets of parameters: OS = M OS = M = 500 GeV. Espinosa. what we term throughout leading corrections. (For mA0 ∼ mZ . 1 is used as calibration: we compare in it our numerical results for Mh0 including only up to two-loop O(αs αt ) corrections with the mass obtained by the program FeynHiggs [34] which uses the explicit two-loop diagrammatic results of Ref. (3)–(5). this shows numerically that the two approaches are equivalent to that order. the correction to mh0 from ΠAA is always numerically negligible for large mA0 as can be easily seen from the structure of the Higgs mass matrix (3). as a generating . we need to calculate ΠAA to the two-loop level since it has O(αs αt ) and O(αt2 ) corrections and in principle could contribute to (4) at the same order as 1M2ij .-J.J. µ = −500 GeV. That is.R. R. (b) MA0 = M3 = MQ S e e U For each case. in particular we neglect all dimensionless couplings other than the top-Yukawa coupling ht and the QCD gauge coupling g3 . respectively. We notice that the two-loop self-energy of the Zboson and the non-zero external momentum corrections to two-loop Higgs boson selfenergies can be neglected in our calculation since all these corrections are higher order effects in the leading approximation. µ = 200 GeV. It is not possible to obtain these self-energies in our current approach. However. Fortunately.

For Xˆ tOS = −2 and Mt = 175 GeV. we try to follow the rule that denser lines correspond to more precise approximations. We also show the Higgs boson masses for Mt = 180 and 170 GeV (including all two-loop dominant corrections) in dashed and dot-dashed lines. As is well know. Higgs boson mass Mh0 vs. In Fig. and the twoˆ OS loop O(αt2 ) corrections are generally small.R. All the physical masses MA0 .21]. OS OS MQ e . We show corrected massed to the two-loop O(αs αt ) and O(αs αt + αt2 ) orders in dot-dashed and solid lines for Xˆ tOS = 0. the masses are increased or decreased by ∼ 5 GeV. respectively. R. One-loop mass. We remark that this upper bound on Mh0 is asymmetric with respect to Xˆ tOS .18 J. we find the bound is about 5 GeV lower. Without two-loop O(αt2 ) corrections. we find that the two-loop O(αt2 ) corrections are sizable (' 5 GeV). Espinosa. functional [35. The dashed.36]. 2(b) (Xt = 2). The effective potential approach is more efficient for the purpose of calculating Mh0 and much simpler to implement in a Fortran program since it requires evaluating only one set of two-loop functions. for two top-squark mixing parameters Xˆ tOS = 0 and 2. two-loop masses to O(αs αt ) and O(αs αt + αt2 ) are shown in dashed. Zhang / Nuclear Physics B 586 (2000) 3–38 Fig. for two values of the top-squark mixing parameters Xˆ tOS (0 and 2). 2. 5 Fig. 3 we examine the upper limit on the Higgs boson mass Mh0 by including the dominant two-loop corrections. encompasses all tadpole and self-energy diagrams (as well as all other multi-point functions) which are calculated in [21]. which roughly corresponds to the maximal left-right top-squark mixing case. MU e and M3 have been set to MS (and the same will be done for all the following plots). < ∼ 2 GeV. We see that maximal values for Mh0 of ' 129 GeV can be reached for large tan β and left-right top-squark mixing parameter Xˆ tOS ' 2. In Fig. the upper bound of Mh0 would be at ' 124 GeV. respectively.-J. dot-dashed and solid lines in this figure correspond to masses Mh0 corrected to one-loop. dot-dashed and solid lines. 2 we show the Higgs boson mass Mh0 vs. 2(a) (Xˆ tOS = 0) corresponds to the case of minimal left-right top-squark mixing. the on-shell SUSY scale MS . 2 and the top quark pole mass Mt = 175 GeV. . 5 In general. the (on-shell) SUSY scale MS . two-loop O(αs αt ) and two-loop O(αs αt + αt2 ) order. For Fig. this asymmetry arises from the two-loop O(αs αt ) corrections [12.

The analytical approximation has a complicated dependence on the µparameter. Xˆ tOS . In Fig. R. Higgs boson masses Mh0 vs. Numerically this dependence is quite weak: varying µ from 100 GeV to 1 TeV . tan β for the top-squark mixing parameters Xˆ tOS = 0 and 2. Two-loop masses to O(αs αt + αt2 ) for Mt = 170 and 180 GeV are also shown in dotted and dashed lines. Espinosa.R. Fig. two-loop masses to O(αs αt + αt2 ) and their approximations are shown in dashed.J. 4. solid and dot-dashed lines. They are shown in dot-dashed and solid lines respectively.5 GeV for almost all the parameter space. Higgs boson mass Mh0 vs.-J. One-loop masses. 4 we compare results from our analytical approximation formula for Mh0 in Section 2 with those obtained by full numerical evaluations. Zhang / Nuclear Physics B 586 (2000) 3–38 19 Fig. Dot-dashed and solid lines correspond to two-loop Higgs boson masses to O(αs αt ) and O(αs αt + αt2 ) respectively for Mt = 175 GeV. The analytical approximation formula works remarkably well: it approximates our numerical results to a precision of < ∼ 0. 3.

Two-loop masses without the O(αt2 ) threshold correction and the complete two-loop results are shown in dashed and solid lines. and they are depicted in dashed lines. so one can easily locate the particles giving the biggest contributions. as we have shown in the second part of Section 3. Finally in Figs.R. We see clearly that these terms are sizable: for large mixing parameters. In Fig. Espinosa. 7 shows the effect of two-loop threshold corrections 1(2) th mh0 evaluated following the RG-inspired analysis of Section 3. we have also shown two-loop O(αs αt ) corrections and their RG-corrected results. their corresponding RG-corrected masses are shown in dot-dashed and solid lines. Higgs boson masses Mh0 vs. 6 and 7 we detail the size of the two-loop top-squark-mixing-dependent corrections in the OS-scheme and their corresponding finite threshold corrections in the RG approach. 5. 2 Fig. Xˆ tOS .20 J. Two-loop masses to O(αs αt ) and O(αs αt + αt2 ) are shown in dotted and dashed lines. (20). with those of the full numerical evaluation. Eqs.21] have already calculated the QCD corrections. Fig. respectively.-J. Refs. (28)–(30). the good agreement between these curves is an indication of the smallness of the logarithmic corrections beyond two-loops and illustrates the accuracy of our results. (3) the formula can significantly simplify the numerical evaluations of Mh0 to a good precision. they increase Mh0 by about 4 GeV and 2−3 GeV for small and large tan β. for a fixed Xˆ tOS changes the Higgs boson mass by less than 1 GeV.13. R. The difference of the solid and dashed lines is the two-loop O(αt2 ) terms which are calculated in this paper. 6 shows in dotted lines two-loop masses without including the topsquark-mixing-dependent corrections of Eq. Zhang / Nuclear Physics B 586 (2000) 3–38 Fig. (2) all terms can be traced back to the potential. As mentioned in Section 3. they have been studied previously in [13]. [12. The dotted lines show the Higgs boson mass neglecting these corrections. We emphasize that the analytical formula is useful for several reasons: (1) the logarithmic and finite corrections can be easily separated. 5 we further compare the results for our RG-corrected Higgs boson masses. . and one can weight the relative importance of these terms. For comparison. this would have been obtained by integrating two-loop RG equations (with the two-loop boundary threshold correction being set to zero).

Zhang / Nuclear Physics B 586 (2000) 3–38 21 Fig. 6. (20) are shown in dotted lines. Xˆ tOS .R. < ∼ 3 GeV. The corresponding masses without the O(αt2 ) threshold h corrections only and with the full numerical results are shown in dashed and solid lines. The corresponding masses without the O(αt2 ) corrections only and the full numerical results are shown in dashed and solid lines. 7. the remaining genuine two-loop threshold corrections (in the sense of the effective field theory) are generally small. Two-loop masses without the threshold correction (2) term 1th m2 0 are shown in dotted lines. respectively. Fig. (20) into the RG-corrected one-loop (1) term 1th m2h0 .-J. Two-loop masses without the top-squark-mixingdependent correction terms of Eq. Higgs boson masses Mh0 vs. Xˆ tOS . . respectively. Higgs boson masses Mh0 vs. respectively. R. Espinosa.J. The RG reshuffling of radiative corrections has absorbed the main part of the two-loop top-squark-mixing-dependent terms of Eq.

DE-FG02-95ER40896 and in part by the Wisconsin Alumni Research Foundation. In particular. Our final approximation formulae (20)–(21) and (28)–(30) have been shown to excellently agree with the full numerical results and can be easily implemented in precision numerical studies. which accurately reproduce results from the full numerical study. was supported in part by a DOE grant No. identifies the most important contributions and links our work to the effective theory or RG approach.-J. The remaining two-loop finite terms can be interpreted as threshold corrections.R. we have shown that two-loop top Yukawa corrections to Mh0 are sizable for the maximal top-squark mixing case. R. and are numerically less important. R.Z. the two-loop O(αt2 ) top-squark-mixingdependent corrections to Mh20 [see Eq. as we have shown in detail in Section 3. Although we have focussed in this paper on the Higgs mass. Our calculation includes effects which can have an impact on the final Higgs mass but were neglected by previous studies.21] using a simpler effective potential approach and provides the most complete and accurate calculation presented in the literature. The upper bound on Mh0 is 129 ± 5 GeV for Mt = 175 ± 5 GeV. Conclusions In this paper we calculate radiative corrections to the lightest MSSM CP-even Higgs boson mass to the two-loop O(αt2 ) order. (20)] are calculated for the first term in this paper and are numerically important. Zhang / Nuclear Physics B 586 (2000) 3–38 5. (28)]. Espinosa. Our analysis extends existing two-loop diagrammatic results [11–13. Acknowledgements We thank André Hoang for correspondence. In that expression all mass parameters are evaluated at appropriate renormalization scales chosen to reproduce the numerically most important leading and next-to-leading logarithmic corrections. This RG rewriting clarifies the structure of the two-loop corrections to Mh20 . To summarize our numerical results. We further simplify our analytical formula by reshuffling higher order logarithmic corrections (using RG techniques) in a compact one-loop expression [Eq. it is worth mentioning that we have also presented in Appendix D the MSSM two-loop effective potential including top-quark Yukawa contributions (for general top-squark mixing parameters and any tan β). This knowledge may well prove useful for other studies. They can increase the Higgs boson mass Mh0 by as much as 5 GeV (among which the top-squark-mixingdependent corrections account for about 4 GeV) for small tan β where ht is large. applicable when the supersymmetric particles are heavy. . We also derive useful analytical approximation formulae.-J.22 J.

m21 . m21 . − p 3      G p2 . Q m1 m1 m1 m1 . m21 . The A0 function is defined by the following momentum integral in d = 4 − 2 dimensions   Z m2 dd p 1 2 1 = m + 1 − ln . Q2 (A. m22 = p2 − m21 − m22 B0 p2 . (A. m21 .5) In all one-loop expressions of radiative corrections. m22 − A0 m21 − A0 m22 .and two-loop scalar functions A. m1 . m22  + m21 + m22 (A.4) (A. 0 = 2 − ln 2 − 1 − 2 ln 1 − 2 − 22 ln 22 . with γE the Euler constant. Espinosa.2) 0 The remaining functions can be related to A0 and B0 as follows     i 1 h B1 p2 . 2 + m21 − m22  2 . m22 6  p2 − B1 p . m21 .R. One-loop scalar functions In this subsection we define the scalar functions A0 .1) A0 (m2 ) = 16π 2 µ4−d i(2π)d p2 − m2 + iε  Q2 where Q2 = 4πµ2 e−γE is the renormalization scale. m2 (A. m2 .-J. The B0 function is Z  1 dd q 2 2 2 2 4−d B0 p .3)    1 = A0 m22 + 2m21 B0 p2 . m21 . m22 = 2 A0 m22 − A0 m21 + p2 + m21 − m22 B0 p2 . Some useful expressions for these functions in limiting cases are (after minimal subtraction)  m2 m2 m2 B0 0. m22 . which appear in one-loop self-energy calculations. 2p  2 B22 p2 .J. B22 and G. Zhang / Nuclear Physics B 586 (2000) 3–38 23 Appendix A.1. m21 . m2 = 16π µ i(2π)d [q 2 − m21 + iε][(q − p)2 − m22 + iε] 1 = −  Z1 dx ln (1 − x)m21 + xm22 − x(1 − x)p2 − iε . we adopt a (modified) minimal subtraction procedure to remove poles in  and keep only finite (real) parts of the above functions. Q m1 − m2 m1      m21 m22 m22 m2 m2 2 2 B0 m1 . R. m22 = 1 − ln 12 + 2 2 2 ln 22 . m21 . One. B1 . B0 .

m2 . m2 . 1 d B0 p2 .

7) (A.8) .6) (A.p2 =0 = . 2 dp 6m2 (A.

The momentum integrals appearing in a two-loop effective potential calculation can be reduced to the following two types of scalar functions [corresponding to the topologies of two distinct zero-point two-loop irreducible Feynman diagrams (the figure-8 and sunset diagrams)]: Z d d   1 d pd q 2 2 2 4−d 2 . m2 = − 16π µ (2π)2d [p2 − m21 + iε][q 2 − m22 + iε] and  2 I m21 . R = 0. + iε][(p + q)2 − m23 + iε] The function J is symmetric in m1 . Espinosa. any Feynman diagram in the two-loop effective potential is subtracted by all its possible one-loop sub-diagrams. They have been studied extensively by several groups using two different methods: a differential equation method [37–39] and an integral Mellin-Barnes transformation method [40–42]. (A. m23  −J m21 . The initial value of this function can be evaluated from (A. m23 2 ∂m3  2  m − m21 + m22 J m21 . m22 = A0 m21 A0 m22 . m2 and m3 . Q 3 (A. m22 .12) The function I satisfies the following first-order partial differential equation [39] R2    ∂ I m21 . m23 = 16π 2µ4−d × [p2 − m21 Z dd p dd q (2π)2d + iε][q 2 − m22 1 .-J.e. m22   .14) This differential equation can be used to solve for the I function. their results all agree. Two-loop scalar functions In this subsection we collect some useful formulae of zero-point two-loop scalar functions. m2 and I symmetric in m1 . m22 .11) (A. In our calculation. (A. m2 ..13) which reduces to a simple algebraic equation when m3 = m1 + m2 . m22 . R. m2 = − ln 2 + 2 − √ . [38].13) where R 2 = m41 + m42 + m43 − 2m21 m22 − 2m21 m23 − 2m22 m23 .9) A. The function J can be reduced to the product of one-loop scalar functions as    J m21 . i.R.24 J. (A. m23 = (d − 3) m23 − m21 − m22 I m21 . Here we mainly follow Ref. Zhang / Nuclear Physics B 586 (2000) 3–38  π m2 B0 m2 .10) J m1 . (A.2. m23 + (d − 2) 3 2 2m3 + m23 + m21 − m22 2m23 J m22 . this is done by replacing the I and J functions as follows [38]: .

17) (A. where (for R 2 > 0) ξ is given by [40]   m2 + m21 − m22 − R m23 − m21 + m22 − R ln ξ m21 .18) is most convenient for series expansions. (A. m23 = R 2 ln 3 2m23 2m23  2  m3 + m21 − m22 − R m2 m2 − ln 12 ln 22 − 2 Li2 m3 m3 2m23   2  m3 − m21 + m22 − R π2 .18) should be replaced by its analytical continuation. m22 . m22 .J. y (A. m22 + m21 A0 m22 + m22 A0 m21 . + − 2 Li2 3 2m23 (A.     1h  i 2 2 2 2 J m1 . Equivalent expressions for ξ also appear in [37] and [39]. m23  1h   i = I m21 .19) 0 In the region where R 2 < 0.-J.16) J m1 . m22 . m23 + 5 m21 + m22 + m23 . We also define a function L for future use . (A. we find that (A. (A. m2 = J m21 . m22 .15)  It is then straightforward to show     m21 m22 m21 m22 2 2 2 2 ˆ 1 − ln 2 . m22 . m23 → Iˆ m21 . m23 − A0 m21 + A0 m22 + A0 m23 . m2 → Jˆ m1 . R. m2 = − 2 + m1 m2 1 − ln 2  Q Q and with some effort    1 1 m21 + m22 + m23 Iˆ m21 . Zhang / Nuclear Physics B 586 (2000) 3–38 25   I m21 . m23 = 2 m21 + m22 + m23 − 2 2   m2 m2 1 −m21 + m22 + m23 ln 22 ln 32 − 2 Q Q  m2 m2 + m21 − m22 + m23 ln 12 ln 32 Q Q  m2 m2 + m21 + m22 − m23 ln 12 ln 22 Q Q   m2 m2 m2 − 4 m21 ln 12 + m22 ln 22 + m23 ln 32 Q Q Q    + ξ m21 . m22 .R.18) where Li2 (x) is the dilogarithm function Z1 Li2 (x) = − dy ln(1 − xy) . Espinosa.

21) + m21 − m22 − 6 m21   m2 5 π 2 1 2 m2 ln + .22) − 2 ln + I (m2 . Espinosa. m23 /m21 .26 J. m22 . (A. m22 .. 2 6 (A. some argument of the I function.-J. m22 . we collect expansion formulae for the function ξ which we use in the derivation of an analytical approximation formula for the two-loop Higgs boson mass corrections. 0 = −m21 ln 12 ln 22 − m21 − m22 ln 1 2 2 ln 12 Q Q Q m2     m2 m2 m2 1 5 + m21 − m22 ln2 12 + 2 m21 ln 12 + m22 ln 22 − m21 + m22 2 2 Q Q Q     π2 m22 + Li2 . The symmetry of the above expression for I (m21 . (A. tree-level Higgs boson mass mh0 .R.15) of I ] can be explicitly checked by using the identity Li2 (x) = − Li2 (x −1 ) − 1 2 π2 ln (−x) − . When computing the two-loop potential. Zhang / Nuclear Physics B 586 (2000) 3–38     L m21 . m23 = m21 ξ 1.g.23) Finally. m22 . m22 /m21 . R.17) that we use in our two-loop effective potential expressions. The ξ functions we find can be reduced to one of the different types we list below using the relation   (A.20) Performing a (modified) minimal subtraction (by removing the single and double poles in ). (1) For 0 6 r 6 1 and 0 6   1: ) (   π2 + ln  − 2 ln(1 − r) ln r − 2 Li2 (r) ξ(1. In (A. ) = (1 − r) 3 ( −  2 − 2 ln  + ln r   )  1 + r π2 + ln  − 2 ln(1 − r) − 1 ln r − 2 Li2 (r) + 1−r 3 . it is the finite (real) parts of (A. the bottom-quark mass mb . m23 . 0. m22 − J m21 . (A. 0) = −m2 2 6 Q2 Q2 2 where we have kept only the finite terms as explained before. m23 = J m22 . can be taken to be zero. m23   − m21 − m22 − m23 I m21 . m23 − J m21 .16) and (A. 0) in m1 and m2 [which obviously follows from the definition (A. We will also omit the carets of Iˆ and Jˆ to simplify the notation. e.24) ξ m21 . The function I is well-behaved in these limiting cases:   m2 − m2 m2 m2 m2 I m21 . r.21) we have implicitly assumed m1 > m2 . m22 .

Zhang / Nuclear Physics B 586 (2000) 3–38 2 + (1 − r)3 ( 27  3 2π 2 − ln  (1 − r 2 ) − r 2 3 )   − 2 ln  + r − 4 ln(1 − r) r ln r + 4r Li2 (r) + O( 3 ).J. 1 . 1/r.-J.29) . ) = rξ(1. 1 + 2 ) = 36K + (8K − 1)1 2 + 36 3      2 3 16 8 1 2 K− − K 14 + 12K1 + (1 − 8K)1 +  + 3 9 1 108 9    2     11 8 + K 13 2 + (1 ↔ 2 ) + O 1m 2n . /r) and the above expression. 0 (A.R. and (1b) For 0 6 |1 |. (A. 2 ) = 3 3   2 π + 2 ln 1 − ln 1 ln 2 1 + −2 − 3     3 2 − ln 1 1 + (1 ↔ 2 ) + O 1m 2n .1953256. + − 1 + 6 54 9 (A. Espinosa. 1 + 1 . 2  1:   8 1 − ln 2 22 ξ(1. Two particular cases of the previous expansion are: (1a) For 0 6 1 . one uses ξ(1. 2 ) = −2(4 + 1 − 2 ln 2 )2 + 9 3     1 7 + ln 2 2 12 + 2 − ln 2 + 18 6      1 1 2 1 3 − ln 2 14 + O 1m 2n . + 2 with m + n = 3. + − + ln 2 1 + 2 2 9 3 with m + 2n > 5. 1 + 1 . R. |2 |  1: (A.27)  8 5 − K 12 22 ξ(1. 2  1:   π2 π2 + ln 1 ln 2 − 2 −1 + + ln 1 ln 2 1 2 ξ(1. Finally we also give (2) For |1 |. r. In this expansion the constant number K is given by 1 K = −√ 3 Zπ/6 dx ln(2 cos x) ' −0.26) (A.28)  with m + n = 5.25) If r > 1.

h2 ) are 0 0 + + rotated into mass eigenstates G and A (G and H ) by the angle β. Zhang / Nuclear Physics B 586 (2000) 3–38 Appendix B. Masses and Feynman rules The Higgs sector scalar potential in the leading approximation is   VHiggs = m2H1 + µ2 |H1 |2 + m2H2 + µ2 |H2 |2 + Bµ (H1 H2 + H. useful to check the scale invariance of the potential. In this appendix. The fact that β diagonalizes M2− is obvious when the minimization conditions of the potential (B. m2H1 + µ2 = −Bµ tan β and m2H2 + µ2 = −Bµ cot β. which we of course include in our calculations. Espinosa. a2 (charged states h+ 1 . and sα = −cβ . sβ = sin β. hh1. with many different fields and field mixings.1. g2 related terms in (B. we summarize the necessary Feynman rules for computing the two-loop potential in this leading approximation and also some MSSM renormalization group equations. etc.2 .2 i = v1. The SU(2) doublet Higgs fields H1 and H2 can be written as follows: √ # " " # h+ (h1 + ia1 )/ 2 2 (B. B. Since we have neglected all g1 . Similarly. (B.4) . We call this the leading approximation and it greatly simplifies our two-loop effective potential calculation. h2 are rotated by the angle α into the mass + eigenstates H 0 and h0 . (B. they are responsible for the tree-level mass of the lightest Higgs boson.2) . respectively. the mass-squared matrices for CP-even and odd Higgs fields are ! m2H1 + µ2 ±Bµ 2 . and µ the supersymmetric Higgs-boson mass parameter. (h2 + ia2 )/ 2 h− 1 In our approximation.28 J. The charged Higgs fields have the same mass-squared matrix M2− as the CP-odd Higgses.1) where mH1 .R. the CP-odd states a1 . mH2 and Bµ [with dimensions of (mass)2 ] are the soft-breaking Higgs mass parameters. it is a good approximation to keep only those terms of the MSSM Lagrangian which depend of the SU(3) gauge coupling g3 and the top Yukawa ht (and neglect the electroweak gauge couplings g1 . This angle β is conventionally defined in terms of the CP-even Higgs field VEVs. H2 = H1 = √ . g2 and the rest of the Yukawa couplings).c.). The CP-even interaction eigenstates h1 .-J. which depend on the electroweak gauge coupling constants. in our approximation (we use shorthand notations cβ = cos β. Although we do not write the quartic Higgs couplings. MSSM in the leading approximation The general structure of the MSSM is quite complicated. R.3) M± = ±Bµ m2H2 + µ2 where the positive (negative) sign applies to the CP-even (odd) fields.1).1). This makes the computation of the complete potential prohibitive at two-loops. However. by tan β = v2 /v1 . are imposed and the soft parameters in the matrix are replaced by tan β and m2A0 = −Bµ (tan β + cot β).) cα = sβ . (B.

The (field-dependent) top squark mass-squared matrix in the leading approximation is   1 2 2 1 2 h M h + h (A h + µh ) √ t t 2 1 e   Q 2 t 2 2 .6) where ct = cos θt˜. with λ as listed below: λH + t˜1 b˜L = −ht (ct mt + st Yt )cβ . 2 1 λA0 t˜1 t˜2 = −λA0 t˜2 t˜1 = √ ht Yt cβ . 2 1 λG0 t˜1 t˜2 = −λG0 t˜2 t˜1 = √ ht Xt sβ . for µ < 0. 2 Other Feynman rules of O(g3 ) and O(ht ) vertices are exactly the same as in the general MSSM and we do not present them explicitly. c2t = cos 2θt˜ and Xt = At + µ cot β.R. √ λH 0 t˜2 t˜2 = − 2 ht (mt − st ct Yt )cβ .5) M2t˜ =   1  1 2 2 2 MUe + ht h2 √ ht (At h2 + µh1 ) 2 2 where MQ e . 1 λh0 t˜1 t˜2 = √ ht c2t Xt sβ .8) when µ > 0. with ht ah˜ 0 t t˜1 = −iah˜ 0 t t˜1 = bh˜ 0 t t˜2 = ibh˜ 0 t t˜2 = √ ct . 2 λG+ t˜2 b˜L = −ht (st mt − ct Xt )sβ . the Feynman rules for Higgs/Goldstone-boson-top squark trilinear coupling are simply −iλ. Espinosa. we only need to interchange a h˜ 0 t t˜i and a h˜ 0 t t˜i .7) The couplings of squarks to neutralinos and charginos are very simple in the leading approximation. Yt = At − µ tan β. 2 (B. mt˜2 and the mixing angle by θt˜.-J. 1 2 1 2 2 ht −ah˜ 0 t t˜2 = iah˜ 0 t t˜2 = bh˜ 0 t t˜1 = ibh˜ 0 t t˜1 = √ st . (B. MU e are soft-breaking mass parameters of the left-handed and right-handed e e. (B. Denoting the top-squarks Q and U mass eigenvalues of the matrix (B. The effect is numerically relevant for the tree level masses and we take it into account. 1 2 1 2 2 ah˜ + bt˜1 = −ht st . Zhang / Nuclear Physics B 586 (2000) 3–38 29 This approximation is excellent when MA0  MZ but would fail for MA0 ∼ MZ . √ λH 0 t˜1 t˜1 = − 2 ht (mt + st ct Yt )cβ . ah˜ + bt˜2 = −ht ct . and At the usual trilinear soft-breaking parameter. but it may be consistently neglected in the two-loop corrections. st = sin θt˜. ah˜ + t b˜ L = −ht . 1 λH 0 t˜1 t˜2 = − √ ht c2t Yt cβ . λH + t˜2 b˜L = ht (st mt − ct Yt )cβ .R are chiral projectors and C the charge-conjugation matrix). √ λh0 t˜1 t˜1 = 2 ht (mt + st ct Xt )sβ . λG+ t˜1 b˜L = −ht (ct mt + st Xt )sβ . . R. since the gaugino-Higgsino mixing can be neglected and the only interactions are Higgsino-squark interactions. √ λh0 t˜2 t˜2 = 2 ht (mt − st ct Xt )sβ .5) by mt˜1 . (B. as well as bh˜ 0 t t˜i and 1 2 1 bh˜ 0 t t˜i . The Feynman rules for the h˜ 0i t t˜j couplings can be written as −i(aPL + bPR ) and that of h˜ + t b˜L as iC −1 (aPL + bPR ) (PL.J.

R.11) − t (16π 2 )2 2 = 2 + M 2 + A2 .16) seem unfamiliar. (B.3). 2. cβ2 and En = s2α . sβ2 . we find one-loop RGEs for Xt and m2t˜ . the (arithmetic) average of the (squared) top squark masses. ∂ ln Q2 3    ∂m2t˜ 16g32 2 Xt 2 1 m = − + M − s m − M 16π 2 2t t 3 t 3 3 2 ∂ ln Q2    3Xt . it can be replaced by the β angle as in (B.5) and the one-loop MSSM RGEs of soft parameters entering those equations. (B.2.9) 2 2 2 4 8g3 ht ∂ ln h2 3ht 9 ht =− = + − .10) ∂ ln Q2 32π 2 (16π 2 )2 2 (16π 2 )2     16g32 h2t Bµ 3h2t Bµ + A + At µ − M3 µ = µ + t 16π 2 2 (16π 2)2 2   Bµ 9h4t + 2A µ . but they follow directly from (B. [Here we use the α angle to keep track the H 0 and h0 contributions. Using (B.14). H + for n = 1. 2. (B. Then we need one-loop RGEs for those masses where M2t = m2H2 + MQ t e e U entering in the one-loop potential   16g32 ∂m2t 2 2 + 3h = − (B.15) µ 16π 2 n n t t t ∂ ln Q2 2    ∂m2H + Bµ n 2 2 2 2 + At µ . (B. s2β for n = 1.14) + h2t 3m2t + (1 + ct2 )M2t − 3s2t mt At + 2    ∂m2H 0 Bµ 2 2 2 n + A = 3h + D M + E µ . G0 and A0 for n = 1. (B. R.13) + h2t 3m2t + (1 + st2 )M2t + 3s2t mt At + 2    ∂m2t˜ 16g32 2 Xt 2 2 m = − + M + s m − M 16π 2 2t t 3 t 3 3 2 ∂ ln Q2    3Xt .13)– (B. 4. Renormalization group equations The MSSM RGEs [46–49] that we will use to check the invariance of the potential to two-loop order under renormalization scale transformations are the following. t 2 2 2 2 2 16π (16π ) (16π ) (B. = 3ht µ + Dn+2 Mt + En+2 (B. 3. They are .13) and (B. First.16) 16π 2 ∂ ln Q2 where Dn = sα2 . 2.4) in the leading approximation. −s2β . h0 .-J. 3. −s2α . we need the two-loop RGEs for those parameters entering in the tree-level potential (B. Zhang / Nuclear Physics B 586 (2000) 3–38 B. 4 and Hn+ = G+ .] The ordering of the Higgs/Goldstone bosons are Hn0 = H 0 .1) ∂m2H2 ∂ ln Q2 ∂ ln µ ∂ ln Q2 ∂Bµ ∂ ln Q2   16g32 h2t 3h2t 18h4t 2 M + M2t + 2M32 − 2M3 At − M2t + A2t . Eqs.30 J. cα2 . (B.12) 16π 2 t mt . Espinosa. (B.

m2t˜ − mt 5 − 3 ln t2 1 2 3 Q o    − s2t mg˜ B0 p2 . Zhang / Nuclear Physics B 586 (2000) 3–38 31 16 ∂Xt = g32 M3 + 3h2t (At + Xt ).2) mt (Q) = Mt 1 − 32 5 − 3 ln t2 + ln S2 − Xˆ t 6π Q Q      1 MS2 m2t 2 8 − ln 2 + sβ − ln 2 2 Q 3 Q )   µˆ 2 µˆ 2 2 . m2A0 + B1 p2 . taken to be ∼ mZ . m2A0 2    + sβ2 2B1 p2 .J. t t 3 ∂ ln Q2 3 3 t 2 t 16π 2 (B. Espinosa. as explained in Appendix B.-J. m2b . their full form can be found in Ref. see Section 2) (   g2 M2 m2 2 2 (C. [22]. ln µ ˆ 1 + − 1 − µˆ 2 1 − µˆ 2 3h2t + 32π 2  1 + cβ2 . (B.17) (B. m2b .) – Top quark 16π 2 Σt (p2 ) =     4g32 n  m2 mt B1 p2 . m2t˜ + B1 p2 . m2Z   o + B1 p2 . m2g˜ .1) where we have assumed all heavy Higgs bosons have mass mA0 much larger than the masses of the light Higgs and W -boson. Eq. m2t . m2t˜ + B1 p2 .17) can also be derived from (B.7) and one-loop RGEs of At . m2t˜ − B0 p2 . µ2 . µ2 . m2g˜ . (See also [50] for top quark/squark self-energies. R. µ and tan β. Appendix C.18) these two equations are used in Section 3 for the RG discussion of the formula for the Higgs boson mass Mh0 . we collect formulae for those MSSM one-loop self-energies which are necessary for our analysis. which we follow for notation. 2 ∂ ln Q 3   2 ∂m  16 2 2 3 2 2 2 2 t˜ g M = − + M + m + h 16π 2 3m . m2Z + B1 p2 .1) we find the running top-quark mass at the scale Q (under the simplified assumptions of a common heavy SUSY scale MS while the µ parameter is left free. m2g˜ . h2t 1 2 L (C. m2t . µ2 . m2t˜ 1 2 n    + mt cβ2 2B1 p2 .R. m2t˜ + B1 p2 . We present these self-energies in the leading approximation of keeping only ht and g3 -dependent terms. From (C. m2g˜ . One-loop self-energies In this appendix. m2b˜ .

m2H 0 . λH 0 t˜R t˜1 = st λH 0 t˜1 t˜1 + ct λH 0 t˜2 t˜1 . m2H + . m2H + . m2t˜ . m2t˜ . µˆ ≡ µ/MS and Mt is the top quark pole mass (we use capital letters to denote on-shell mass parameters). 0 2  2 2   1 + h2t st2 A0 m2t˜ + ct2 A0 m2t˜ + 1 2 2 + 2 4 X X 4 X    Dn A0 m2H 0 + G p2 .7). (C. m2t˜ + n=1 i=1 + 2 X n  λHn+ t˜L b˜L λHn+ t˜R b˜L B0 p2 . m2t n n=1 2  X λ2H 0 t˜ t˜ B0 p2 . µ2 . m2b + n n=1 + 2 4 X X 2 X  λ2H 0 t˜ t˜ B0 p2 . m2t + G p2 . – Top squarks 16π 2 Πt˜L t˜L (p2 )      8g 2 n = 3 G p2 . We have used the reduced parameters Xˆ t ≡ Xt /MS . µ2 . 0 2 2 2 " 4    1X  Dn A0 m2H 0 + h2t ct2 A0 m2t˜ + st2 A0 m2t˜ + A0 m2b˜ + 1 2 L n 2 n=1 # 2 X    Dn+2 A0 m2H + + G p2 .3) and (A. 0 1 1 1 3 o    2 2 2 2 2 2 + st A0 mt˜ − p + mt˜ B0 p .n=1 n  p2 . 0 +2 p2 + m2t˜ B0 p2 . m2g˜ .6). 0 1 1 2 2 3  3    + 4mg˜ mt B0 p2 . m2H + . m2b˜ . n i. m2t˜ . m2b˜ . m2H 0 . m2t˜ .32 J. Zhang / Nuclear Physics B 586 (2000) 3–38 In this equation we have neglected the external momentum and used (A. µ2 .4) B0 R b˜ L n L     4g32  −2 p2 + m2t˜ B0 p2 .n=1 i L (C. m2t˜ + λ2H + t˜ n=1 i=1 16π 2 Πt˜L t˜R (p2 ) = n R i n i i. mt˜ .R. m2g˜ .-J. 0 1 1 1 3 o    + ct2 A0 m2t˜ − p2 + m2t˜ B0 p2 . . m2t + h2t s2t A0 m2t˜ − A0 m2t˜ 1 2 2 2 4 X X  λHn0 t˜L t˜i λHn0 t˜R t˜i B0 p2 .5) where λH 0 t˜L t˜1 = ct λH 0 t˜1 t˜1 − st λH 0 t˜2 t˜1 . m2t˜ . m2t + st2 A0 m2t˜ − p2 + m2t˜ B0 p2 . R. m2b˜ .3) 16π 2 Πt˜R t˜R (p2 )      8g 2 n = 3 G p2 . etc. m2t˜ + λ2H + t˜ n=1 i=1 n L i n i i. B0 L b˜ L n L (C. Espinosa. The symbols Dn are defined after (B. m2t + ct2 A0 m2t˜ − p2 + m2t˜ B0 p2 . m2g˜ .n=1 n  p2 . m2H 0 .

M = v 2 (Q) = 2 MW t W W g g2 32π 2 Q2 (C. described already for previous self-energies)   2  h2t sβ2  4MW 4 2 m2t T 2 2 ˆ + Re ΠW + 3 + X 1 − −6 ln . m2b 2    1  2 2 2 2 2 − 2ct B22 p . (C. (A.8) where we have used (A.6) Mt˜ = MQ e 2 2 2 2 2 we obtain (assuming again a common heavy SUSY scale MS and leaving free the µparameter): (   g32 MS2 2 2 2 − ln 2 mt˜ (Q) = Mt˜ 1 − 3π 2 Q      2  MS2 M2 π ˆ2 3ht ˆ t2 sβ2 + Yˆt2 cβ2 2 − ln S + cβ2 1 − √ Y X − ln + t 32π 2 Q2 Q2 3 )       MS2 4 2 2 2 2 2 ln 1 − µˆ 3 − 2 ln 2 − 1 − µˆ .7) + µˆ ln µˆ + 1 − µˆ Q 2 2 Mt˜2 = MQ e + mt + mt Xt − 1     g32 MS2 MS2 ˆ + mt MS 4 2 − ln 2 + 2Xt ln 2 12π 2 Q Q (   2 2  M π 3ht mt Xt sβ2 + Yt cβ2 2 − ln S2 − √ Yt cβ2 + 16π 2 Q 3 )   2   3 MS 1 2 4 2 4 2 + Xt 1 − ln 2 − 1 − µˆ + µˆ ln µˆ + 1 − µˆ ln 1 − µˆ Xt . Espinosa.J.3)–(C.7). − 2st B22 p . mb˜ − A0 mt˜ 1 1 L 4 # )   1   1 2 2 2 2 2 2 + A0 mb˜ . mt˜ . m2b + G p2 .-J.R.10) where we have neglected the external momentum in (C. R. mt˜ . mb˜ − A0 mt˜ 2 2 L L 4 2 T 2 2 ΠW W (p ) = 3g (C.5) we derive relations between running and on-shell top-squark masses and mixing parameters using the following one-loop relationships (for ct2 = st2 = 1/2):    1  Re Πt˜L t˜L Mt˜2 + Πt˜R t˜R Mt˜2 − Re Πt˜L t˜R Mt˜2 . – W boson 16π 2 (  1  2B22 p2 . .6). 2 Q 2 mt Xt (Q) = Mt XtOS (C. Zhang / Nuclear Physics B 586 (2000) 3–38 33 From (C.3)–(A. 1 1 1 2    1  2 2 2 + mt − mt Xt − Re Πt˜L t˜L Mt˜2 + Πt˜R t˜R Mt˜2 + Re Πt˜L t˜R Mt˜2 .9) and used (A. m2t .9) and the definition Yˆt ≡ (At − µ tan β)/MS . m2t . (C.9) This gives (under the assumption of the simplified SUSY spectrum of Section 2.

Espinosa.R.-J. Zhang / Nuclear Physics B 586 (2000) 3–38 – Higgs boson We need only the difference    16π 2 Πhh m2h0 − Πhh (0) " # .34 J. R.

.

d 2 2 2 .

mt . mt . B0 p .

− dp2 p 2 =0 .

X .

m2t˜ . m2t˜ .  d λ2h0 t˜ t˜ B0 p2 .

.

6) and (A. 3 for (s)quarks and 1 for (s)leptons. Following the leading approximation. h2 ) + V2 (h1 .6).1).3) f where f sums over all the (s)quarks and (s)leptons. 7 We adopt the (modified) DR -scheme of Refs. it is needed for the invariance of the potential under a RG transformation. we keep only the numerically 6 This term is a function of the soft-breaking parameters. i (D. The one-loop potential is well known and the O(αs αt ) part of the two-loop potential was computed in [12]. (D. h2 ) = Vvac + V0 (h1 . MSSM effective potential to the two-loop order In this appendix we present the MSSM effective potential for the (real) neutral components of the Higgs fields up to the two-loop level in the leading approximation (which neglects all dimensionless couplings except ht and g3 ).1) where Vvac is a field-independent vacuum energy term. h2 ) + V1 (h1 .11) i. we list them here for completeness and for future reference. Appendix D.7) we get (19).j where λht˜i t˜j are defined in (B. 6 The tree-level potential V0 is 1 2 1 (m + µ2 )h21 + (m2H2 + µ2 )h22 + Bµ h1 h2 . We first write the potential as V (h1 . . (D. h2 ) = Nc H mf˜ − 2H mf + 3H m2W + H m2Z i 2 V0 (h1 . . (B. h2 ).2) into the MSSM Higgs sector scalar potential (B.2 4 X n=1 H m2H 0 n  + 2 X H m2H + n  n=1 −2 2 X i=1 H m2χ˜ + i  − 4 X i=1  H m2χ˜ 0 . m2t  4m2t (C. The complete one-loop potential in Landau gauge is 7 # " X f X    3  2 2 2 16π V1 (h1 . m2t . [43–45]. Nc is the color factor.2) 2 H1 2 which simply follows from substituting Eq. Using (A. + 3m2h0 i j dp 2 i j p 2 =0 = 3h2t m2h0 sβ2 B0 0. h2 ) = f 1 + 2 i=1.

Hn0 represent H 0 . The direct way requires the computation of some one-loop quantities to order O(). those from top (s)quarks. and the function H is   m2 3 m4 ln 2 − . m2t˜ + mt˜ I m2t˜ . h2 ) = 8g3 J m2t . m2t˜ + 2st2 ct2 J m2t˜ .2 (≡ χ˜3. m2t − 2m2t I m2t . 0 16π + +  X   X 2 1 4 ct + st4 J m2t˜ . with tree-level mass given by the SU(3) gaugino soft mass M3 . 4.. The two-loop scalar functions I . To simplify the final result. Feynman diagrams for the two-loop effective potential of order O(αt2 ) in the MSSM. mg˜ . m2g˜ .4 h+ (≡ χ˜ 2+ ) have degenerate mass of |µ|. mt . Zhang / Nuclear Physics B 586 (2000) 3–38 35 important parts. (A. [11]. R.5) 1 2 i=1 where g˜ is the gluino. We have explicitly checked that. m2t . 4) represent charginos and neutralinos.R.20). 2) and χ˜ i0 (i = 1.16) and (A. we neglect left-right mixings in the bottom-squark sector and the gaugino–Higgsino mixings in the neutralino–chargino sector (under this assumption.5) are given in Appendix A. mt − I mt˜ . we exactly reproduce their unexpanded mass formula. m2t˜ . 2. (D. 3. h0 .3). (A. The relevant Feynman diagrams are shown in Fig. (D. in the particular limit studied in [11]. Fig. 2. which shows the equivalence of both methods.-J.17). 2. G0 and A0 for n = 1. H + for n = 1.J. 3. 8 The top Yukawa contribution to the two-loop potential is a new result of this paper. the Higgsino masses are simply |µ|). (D. Eqs. Espinosa. m2t i  )    2 2 2 2 2 2 − 4mg˜ mt st ct I mt˜ . Hn+ represent G+ . 8 The procedure we have followed of subtracting all possible one-loop subdivergences to define these functions is an alternative to the direct way used in Ref. i. perhaps we find the subtraction method simpler. Eq. 8. mg˜ . (D. . J and L in Eq.e. 0 i i i i i 1 2 2 2 X L 2 2 i=1 i=1 m2t˜ . The neutral and 0 ) and e charged Higgsinos e h01. χ˜ i+ (i = 1. 8. In Eq.4) H (m2 ) = 2 Q 2 The QCD contribution to the two-loop effective potential in the MSSM is (    2 2 2 V2s (h1 . (11) of [11].

(B. First. m22 . m2t L  + st2 L m2t˜ . m2b˜ . m2b 1  + ct2 L m2t˜ . µ2 . mt . The Higgs/Goldstone-squark couplings are defined in Eq. mt˜ − 3 λ2H + t˜ b˜ I m2H + . Using the derivatives of I . µ2 . when all soft-breaking parameters are taken to be zero). µ2 . mt ± 2mt I mH 0 .8) (D. m2t˜ + c4t J m2t˜ . − i j i n L n i j n n i L 2 i. R. (D. positive and negative signs apply to CP-even (H 0 . m2b˜ 1  L + ct2 J m2t˜ . mt˜ .6) while the Dn symbols are defined after Eq. m2H + + J m2b˜ . G0 ) Higgs/Goldstone bosons respectively. Zhang / Nuclear Physics B 586 (2000) 3–38 these simplifications are valid in the leading approximation. the potential should vanish in the supersymmetric limit (i. 2 ∂ ln Q     ∂L(m21 . m22 .j =1 n=1 i.j =1 n=1 (D. h0 ) and odd (A0 . Two tests can be applied to check the correctness of the effective potential V (h1 .9) .. 8 is of order h2t but does not contribute to mh0 ) 16π 2 2 V2t (h1 . we find (the last diagram of Fig. m2b 2  ) 4 2 2 2 X X   3 XX 2 2 2 2 λH 0 t˜ t˜ I mH 0 . m2b˜ 2  L 2   X  J m2t˜ . m2H + 2 X 2 n L  n + st2 J m2t˜ . Using the Feynman rules given in Appendix B. J and L functions with respect to the renormalization scale Q X   ∂I (m21 . m2t˜ + L m2t˜ . In the following we show the invariance of the two-loop potential under a RG transformation. m23 ) =− A0 m2i + m2i .7). up to the order of our perturbative calculation. m2t i i 1 i 2 i=1 +L n i=1 m2b˜ . m2t . (D.36 J. h2 ) should be invariant under changes of the renormalization scale. µ2 .R. the potential V (h1 . m22 ) = m21 A0 m22 + m22 A0 m21 . and second.6) where in the first line of Eq.6). Espinosa. mH 0 = 3ht i n n n 2 n=1 + 2 X    Dn+2 L m2H + .-J.7) i=1   ∂J (m21 . m2b + st2 J m2t˜ . The vanishing of the potential in the supersymmetric limit is proved by simple algebra. h2 ). mt + J mt˜ . 2 ∂ ln Q 3 (D. m2H + 1 n n=1 + ct2 J 2 +s2t i=1  m2t˜ . mt . (B. m23 ) = m21 − 2m22 − 2m23 A0 m21 − m22 A0 m22 − m23 A0 m23 ∂ ln Q2 2 + m41 − m22 + m23 . h2 ) " # ( 4 2 X Dn   X  2 2 2 2 2 2 2 2 2 2 L mH 0 . m2t˜ .e.

Phys. Kazakov. Haber. Z. Nucl. B 263 (1991) 233. Lett.M. Lett. (D. Papadopoulos. F. B 281 (1992) 284. Rev.6). Hempfling.cern. 66 (1991) 1815.2) and (B. B 140 (1984) 339. Drees. Lett. [13] J. Nojiri. [10] J. Zhang. H. Phys.H. J.9)–(B.V. B 436 (1995) 3. besides many straightforward expansions. T. Phys.ch/∼offline/physics− links/lepc.htm. Rev. Zwirner. A.R. G.11)]. [19] A. Burkart. Phys. Riotto. Rep. . Yanagida.I. B 423 (1994) 437. Espinosa. Phys. Phys.L.3). B 447 (1999) 89. Okada. M. D 46 (1992) 3086. Li. A.12)– (B. Theor. 85 (1991) 1. Zhang. Phys. Haber.J. B 274 (1992) 191. To derive the analytical expression of 1m2h0 in Section 2. M.E. Hoang. (D. B 331 (1994) 99. Rev. Phys. [9] A. Sher. S. Nilles. Ehret.A. Rosiek. we find that the RG variation −  8g32 h2t h22 ∂V2 (1) 2 2 2 2 − D V = MQ 1 e + MU e + 2M3 + Xt − 2M3 Xt ∂ ln Q2 (16π 2)2    9h4t h22 1 2 2 2 2 m + (At + Xt ) MQ − e + MU e+ (16π 2 )2 2 H2 (D. Rev. Prog. D 45 (1992) 2482. Pokorski. Lett. D.R. Casas. Espinosa. Phys. S. Lett. Phys. Nucl. http://www. Zhang / Nuclear Physics B 586 (2000) 3–38 37 and the one-loop MSSM RGEs for top-(s)quark and Higgs boson masses. A.10) modulo terms independent of the Higgs field h2 . B 498 (1997) 3. Chankowski. [17] P.11) Note that this is a nontrivial check that all ln Q2 terms cancel with each other between Eq. http://l3www. D 41 (1990) 225.M. with q˜ = t˜ or b. B 262 (1991) 477. Phys. [8] M. (D. (B. Haber. [18] A. d ln Q2 ∂ ln Q2 (D. Phys. Hempfling. This result agrees exactly with the two-loop RG variation of the tree-level potential D(2) V0 [cf.ch/Opal/PPwelcome. [3] LEP Experiments Committee Meeting.-J. [4] S. Ridolfi. R. so that ∂V2 d (V0 + V1 + V2 ) ≡ D(2) V0 + D(1) V1 + = 0. March 7th. Phys. we need to expand the twoloop potential V2 in powers of mt /MS and mt Xt /MS2 . Berger. Phys. R. 68 (1992) 3678.E. Quirós. Eqs.ch/ALPUB/seminar/lepc− mar2000/LEPC2000files/v3− document.cern. W. Lett.-J.A.cern. Lett. Phys. Yamada. M. [12] R. Eqs. J. Yamaguchi. Johnson. [11] R. Phys. Brignole. [7] D. [2] H. Lett. [6] Y.16). Phys. 110 (1984) 1.25)–(A.B. R. JHEP 0003 (2000) 026. we have used (A. [15] H.-J.10) and D(2) V0 . Lett. [14] M.html.ch/analysis/latestresults. Phys. Ellis. Phys.H. Rep. 117 (1985) 75. [16] M. C 61 (1994) 247.P. R.cern. Nucl.28) for the t − q˜ − h˜ and t˜ − q˜ − h diagrams of ˜ (D.R. Gladyshev. G.html. Lett. G. de Boer. B 257 (1991) 83. References [1] H.E. Rev. this cancellation guarantees the correct leading and next-to-leading order logarithmic behavior of the effective potential. [5] J. Phys. Díaz.P.S. Pierce. Espinosa. Here D(1) V1 represents the one-loop RG variation of the one-loop potential Eq.html. http://delphiwww. Kane. M. 2000: http://alephwww.

Jones.E. B 285 (1992) 399 (Erratum). Okada. Phys. Lett. A. R.P. Heinemeyer. Phys. [39] M. B 266 (1986) 389.R. E.R. Zucchini. B 274 (1992) 409. C 75 (1997) 539. Phys.T. Zhang. Tausk. Phys. Lett. Rev. J. Nucl.-J. Nuovo Cimento A 111 (1998) 365. Phys. J. W. Inoue. Comp. [50] A.R. [44] D. [46] K. D. Phys. Phys.E. [38] C. 68 (1982) 927. Jones. Nucl. Ford. Haber. Tausk. D 50 (1994) 5481.A. Lett.T.E.R. Rev. Phys. [24] Y. Barbieri. H. M. Jack.R. Bagger. Rev. M. Haber. Frigeni. B 491 (1997) 3. Nucl.A. D. D. Prog. Kakuto. J. Pierce. Lett. B 387 (1992) 373. Hollik. D 50 (1994) 3537.I. Phys. Phys.M. Quirós. Z.T. G.M. B 167 (1980) 479. Vaughn. [31] M. R. Komatsu. Nucl. B 84 (1979) 19. H. Capper. [27] H. B 461 (1996) 407. Nucl. M. [32] D. Carena. [43] W. [30] H. Jack. Ford. Carena. W. Quirós. Martin.R. M. M. Yamada.E.I. Jones. Berends.P. B 262 (1991) 54. Phys. [42] F. Rev. Phys. [47] S. Yamaguchi.R. Coleman. Yanagida. B 381 (1992) 66. [37] C. D. Phys. Wagner. Phys.E. Lett. Nucl. [48] Y. Phys.M. Theor. [33] A. H. I.E.M. J. 71 (1984) 413.T. Jones. . Davydychev. Lett. C. B 421 (1994) 456. Jackiw. Czy˙z. Weiglein. D 58 (1998) 091701. B 467 (1996) 3. Phys. Hollik. Zhang / Nuclear Physics B 586 (2000) 3–38 [20] A. Haber. Caffo. Phys. Phys. Espinosa. K. [23] R. C 67 (1995) 495. Siegel. Davydychev. Phys. S. J. Z. Nucl. Phys. R. Rev. Carena. J. [28] M.T. T. Lett. Phys. B 388 (1996) 793. C. Nucl. Phys. Laporta. B 397 (1993) 123. Phys. R.T. van Nieuwenhuizen. [36] R.M. Jack. Tausk. Dabelstein. C 9 (1999) 343.T. B 258 (1991) 167. B 410 (1993) 325. Weiglein. M. Espinosa. V. Wagner. Espinosa. Espinosa. Comelli. G. Martin. [35] S. Lett. Phys. Phys. [26] K.38 J. [22] D.-J. Jones. [29] M. D. Hollik. Weinberg.B. Commun. [34] S. D 7 (1973) 1888. Quirós.R. [41] A. D 9 (1974) 1686. [49] I. E. Lett. [45] I.A. Hempfling. [40] A. Nucl. Prog. Phys.B. Vaughn. B 333 (1994) 372. Remiddi. B 580 (2000) 29. Phys. D 50 (1994) 2282. Phys. B 504 (1997) 551 (Erratum). J. Matchev. Phys. S. Smirnov. Hoang. C. Rev. Weiglein. S. Caravaglios. Phys. Eur. Heinemeyer. 124 (2000) 76. Heinemeyer. Rev.R. W. Nucl. M. B 440 (1998) 296. Nucl. Takeshita. Sirlin. B 355 (1995) 209.T. P.B. Y. Sasaki. R. [21] S. Phys. M. Nucl. [25] J. M. S. Wagner. Carena. B 266 (1991) 389.E. Yamada.H. Theor. Phys. G. Lett. Phys. D 48 (1993) 4280. Phys.M. Donini. C. A. Phys. Hempfling. Wagner.

14. All rights reserved. accepted 26 June 2000 Abstract We compute the branching fractions of Bs.80. this decay mode is complementary to b → sγ and efficiently probes the large tan β region. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. and the LHC experiments ATLAS. We find that the parameters of the neutral Higgs sector of the two-Higgs-doublet model cancel in the result. Ulrich Nierste 1 Theoretical Physics Department. For tan β = 60. revised 9 May 2000. DE-AC02-76CH03000 with the US Department of Energy. Since the branching fraction increases like tan4 β. leptonic decay. For large values of tan β and a charged Higgs mass above the bound from b → sγ . 12. so that the branching fractions depend only on the charged Higgs mass and tan β.Nuclear Physics B 586 (2000) 39–55 www.Cp Keywords: B. ∗ Corresponding author. Logan ∗ .3 2 1 3 ( 0 0 ) 0 0 4 1 7 . IL 60510-0500.elsevier. Introduction The ongoing and forthcoming high-statistics B-physics experiments at BaBar. the Tevatron. All rights reserved. BELLE.20. PII: S 0 5 5 0 .He. multiplet. an integrated luminosity of 20 fb−1 in an extended Run II will probe charged Higgs masses up to 260 GeV. These experiments may reveal physics beyond the SM. under contract no. We have analyzed in detail the region of the (MH + .V. For the same value of tan β.d → `+`− in a two-Higgs-doublet model Heather E. E-mail: logan@fnal.60. HERA-B. Higgs particle.  2000 Elsevier Science B.d → `+ `− in the type-II two-Higgs-doublet model with large tan β. Fermilab. if the background to Bs → µ+ µ− is small. CMS and LHCb [1–7] will probe the flavor sector of the Standard Model (SM) with high precision. PACS: 13.V. and a substantial theoretical effort is devoted to calculating the observables that will be tested in various scenarios of new physics. tan β) plane to be probed by searches for Bs → µ+ µ− in Run II of the Tevatron. we find that the branching fractions can be enhanced by up to an order of magnitude or suppressed by up to a factor of two compared to the Standard Model result. USA 2 Received 17 April 2000.Fr. the LHC may be able to explore charged Higgs masses up to 1 TeV using this decay. We point out that previous calculations in the literature are gauge-dependent due to the omission of an important diagram.X .gov 1 nierste@fnal. mass 1. which gives the dominant contribution in the ’t Hooft–Feynman gauge.nl/locate/npe Bs.gov 2 Fermilab is operated by Universities Research Association Inc. Higgs particle. Batavia.

This paper is organized as follows. Appendix A contains a discussion of trilinear Higgs couplings. The model is reviewed in detail in Ref. where d 0 = d or s and ` = e. Φ2 = . the Yukawa coupling to b quarks is of order one and large effects on B decays are possible. ! ! φ1+ φ2+ . Yet the presented results differ analytically and numerically substantially from each other. Their ratio is parameterized by tan β = v2 /v1 .40 H. Earlier papers have already addressed the decay Bd 0 → `+ `− in the 2HDM or the full MSSM [11–17]. µ or τ . The two-Higgs-doublet model In this paper we study the type-II 2HDM. In Section 4 we compare our result with previous calculations. MW 1 2 2 1 2 2 2 2 g (v1 + v2 ) = 2 g vSM . because the theoretically predicted mass range increases with tan β. The leptonic decay Bd 0 → `+ `− . This helicity suppression factor numerically competes with the suppression factor of (m` /MW ) tan β stemming from the Higgs–Yukawa couplings to the final state leptons. In the SM the decay amplitude suffers from a helicity-suppression factor of m` /mb . 2. [18]. is especially well suited to the study of an enlarged Higgs sector with large tan β. The 2HDM contains two complex SU(2) doublet scalar fields. which has the same particle content and tree-level Yukawa couplings as the Higgs sector of the Minimal Supersymmetric Model (MSSM). so that one expects the new contributions in the 2HDM to be similar in size to those of the SM. (1) Φ1 = φ10 φ20 which acquire the vacuum expectation values (vevs) hφi0 i = vi and break the electroweak 2 = symmetry. We present our conclusions in Section 6. We finish Section 3 by combining the results for the 2HDM diagrams and giving compact expressions for the branching fractions.E. In this paper we study the type-II two-Higgs-doublet model (2HDM). which is absent in the Higgs-mediated contribution. . In Section 2 we give a brief review of the type-II 2HDM. If the ratio tan β of the two Higgs vacuum expectation values is large. Finally. Direct searches for the lightest neutral MSSM Higgs particle have begun to constrain the low tan β region in the MSSM. Nierste / Nuclear Physics B 586 (2000) 39–55 A common feature of all popular weakly-coupled extensions of the SM is an enlarged Higgs sector. corresponding to tan β of order 50. Logan. In Section 3 we review the SM calculation of the decay Bd 0 → `+ `− and describe our calculation of the relevant 2HDM diagrams. Hence observables allowing us to study the complementary region of large tan β are increasingly interesting. A further theoretical motivation to study the large tan β case is SO(10) grand unified theories [8– 10]: they unify the top and bottom Yukawa couplings at high energies. The Higgs vevs v1 and v2 are constrained by the W boson mass. U. In Section 5 we present a numerical analysis of our result and estimate the reach of future experiments. so that we have decided to perform a new analysis. where vSM = 174 GeV is the SM Higgs vev.

The CP-odd states are G0 = φ10.20]. H 0 = φ10.H. The CP-even states mix by an angle α giving two states. The Yukawa terms in the Lagrangian are: ¯ 1 d − Yu QΦ ¯ c u − Yl LΦ ¯ 1 e + h. In both the SM and 2HDM. the contributions with a top quark in the loop are dominant.i cos β + φ20. We impose the discrete symmetry Φ1 → −Φ1 . H + = −φ1+ sin β + φ2+ cos β.i sin β + φ20. In addition we must consider the trilinear H + H − H couplings (H = h0 . QP = mb bP ¯ µ PL d 0 `γ ¯ µ γ5 `. LYuk = −Yd QΦ 2 (5) where Φ c = iτ2 Φ ∗ . αEM is the electromagnetic fine structure constant and θW is the Weinberg angle. h0 = −φ10.E. The would-be Goldstone bosons G± and  + iφi0. H 0 ) which were first given in Ref.i sin β. QA = bγ (7) . The charged Higgs states are G+ = φ1+ cos β + φ2+ sin β. Φ2 → Φ2 (which is softly broken by dimension-two terms in the Higgs potential).r sin α. [13] and are discussed in Appendix A. U.i for the real G0 are eaten by the (3) and imaginary parts of W and Z bosons. Effective Hamiltonian for B → `+ `− The decay Bd 0 → `+ `− proceeds through loop diagrams and is of fourth order in the weak coupling. with the SU(2) singlet fermion fields transforming as d → −d. Nierste / Nuclear Physics B 586 (2000) 39–55 41 Since in this paper we are not interested in CP-violating quantities. we assume CP is conserved by the Higgs sector for simplicity. Most of the Higgs couplings needed in our calculation are given in Ref. These transformation rules define the type-II 2HDM and determine the Higgs-fermion Yukawa couplings. ¯ 0 `` bd αEM GF ξt [CS QS + CP QP + CA QA ]. [18]. where we use the notation φi0 = vi + √1 φi0. The CKM elements are contained in ξt = Vt∗b Vt d 0 . e → −e.r 2 φi0 . 3. The mass eigenstates are then given as follows. Down-type quarks and charged leptons (up-type quarks) are given mass by their couplings to Φ1 (Φ2 ).c.. Logan. u → u. so that one may describe the decay at low energies of order mb by a local ¯ coupling via the effective Hamiltonian. ¯ QS = mb bP ¯ L d 0 `γ ¯ 5 `.r sin α + φ20. (6) H=√ 2 2π sin2 θW Here GF is the Fermi constant. (2) and their Hermitian conjugates.i cos β. The operators in (6) are ¯ L d 0 ``.r cos α.r cos α + φ20. (4) In order to avoid large flavor-changing neutral Higgs interactions we require natural flavor conservation [19. A0 = −φ10.

E. γν ]/2 contribute to Bd 0 → `+ `− . mt . . because it mass.42 H. the dominant contributions to this decay come from the W box and Z penguin diagrams shown in Fig. Dominant SM diagrams. Hence the renormalization group evolution is trivial: scalar current bP if the bottom quark mass in QS and QP is normalized at a low scale µ = O(mb ). In the SM. are calculated at a low scale µ = O(mb ) characteristic of the Bd 0 decay. The calculation of the diagrams in the full high-energy theory gives the initial condition for the Wilson coefficients at a high renormalization scale µ on the order of the heavy masses in the loops. We have neglected the righthanded scalar quark operators because they give contributions only proportional to the d 0 ¯ µ ` does not contribute to Bd 0 → `+ `− . These diagrams were first calculated in [21] and give a nonnegligible contribution only to the Wilson coefficient CA . MH + . the operators QS and QP have zero anomalous dimension because the anomalous dimensions of the quark mass mb (µ) and the (chiral) ¯ L d 0 (µ) add to zero. which must be summed to all orders in perturbation theory with the help of renormalization group techniques. then no large logarithms appear in the effective Hamiltonian or in the decay rate. which is conserved in the limit of vanishing quark masses. Logan. Yet the operator QA has zero anomalous dimension because it is a (V − A) quark current. Because mb  MW . The hadronic matrix elements. Similarly. Finally. 1. no operators involving σµν = i [γµ . however. QS and QP . 1. There are Fig. The renormalization group evolution down to this low scale requires the solution of the renormalization group equations of the operators QA . Nierste / Nuclear Physics B 586 (2000) 39–55 where PL = (1 − γ5 )/2 is the left-handed projection operator. there are highly separated mass scales in the decay Bd 0 → `+ `− . Short-distance QCD corrections can therefore contain large logarithms like log(mb /MW ). The vector leptonic operator `γ gives zero when contracted with the Bd 0 momentum. There is no contribution from a photonic penguin because of the photon’s pure vector coupling to leptons. U.

giving mt (mt ) = 166 GeV. but 2 relative to the these contributions to the amplitude are suppressed by a factor of m2b /MW dominant contributions and can be ignored.  Y (xt ) = 0. Logan. Dominant diagrams in the 2HDM with large tan β. For compactness.E. We have calculated the individual diagrams explicitly in a general Rξ gauge. (9) where we have parameterized the dependence on the running top quark mass in the MS scheme. we give the results of the individual diagrams below in the ’t Hooft–Feynman gauge. The NLO corrections increase Y (xt ) by about 3%. Nierste / Nuclear Physics B 586 (2000) 39–55 43 also contributions to the Wilson coefficient CS from a SM Higgs penguin [22] and to the Wilson coefficient CP from the would-be neutral Goldstone boson penguin [23].27 ± 0. . (8) 2 = 4. for which the 2HDM contributions to this decay are significant. Although each diagram that involves a W ± or G± boson is gauge-dependent. 2. if mt is normalized at µ = mt . keeping only the terms proportional to tan2 β. There are no new contributions to CA in the 2HDM. In the large tan β limit. Fig.997 mt (mt ) 166 GeV 1. We limit our consideration to the case of large tan β. the Wilson coefficients CP and CS receive sizeable contributions from the box diagram involving W and H + and the penguins and fermion self-energy diagrams with neutral Higgs boson exchange shown in Fig. Numerically.H. which therefore retains its SM value.25]. U. The SM decay amplitude is then given by the Wilson coefficient [21] CA = 2Y (xt ).55 . their sum is gauge-independent.26 and m is evaluated in the MS scheme at µ = m . 2. where xt = m2t (mt )/MW t t αs Y1 (xt ). The function Y (xt ) is given by Y (xt ) = Y0 (xt ) + 4π where Y0 (xt ) gives the leading order (LO) contribution calculated in [21] and Y1 (xt ) incorporates the next-to-leading (NLO) QCD corrections and is given in [24.

1 CP = m` 1 tan2 β P+(xH + . U. (13) (14) . higherorder QCD corrections are small in leptonic decays. (11) B+ (x. This diagram gives m` peng. 3. Logan. 2) contributes  2  m` sin α cos2 α peng. experience with NLO calculations in the SM [24. y) = x−y y −1 x −1 B+ (xH + . xt ) = + . The effect of a nonzero charm quark mass is negligibly small.1. and the effect of nonzero charm quark mass is negligible. P+ (x. xt ) 2 .2 CP m` 1 tan2 β P+ (xH + . CSbox = CPbox = 2 2MW (10) 2 /M 2 and x was defined after (8) in terms of m (m ).1 2 tan β P+(xH + . Finally the loop function B+ in (10) reads   log x y log y − . Penguins The penguin diagram with H + and W + in the loop (see Fig.25] shows that with the definition of mt adopted here. Here xH + = MH t t t + W in a LO calculation like ours. 2 MA (12) 0 Here again all three quark flavors enter the result from the GIM mechanism. Box diagram The box diagram in Fig. because the coupling of G+ to quarks is proportional to either of the quark masses and we neglect ms . xt ) =− CS 2  2 2 2 2 2  sin α (MH + − Mh0 ) cos2 α (MH + − MH 0 ) + . However. Nierste / Nuclear Physics B 586 (2000) 39–55 3. By contrast. Strictly speaking.44 H. in the penguin diagram involving H + and G+ in the loop only the internal top quark contribution is relevant. one is not sensitive to the renormalization scheme and we could equally well use the top quark pole mass in xt . y) = x−y x −1 y −1 2 MW  . × 2 2 2 Mh20 MW MH MW 0 peng. xt ) 2 =− 2 M 0  2 − M2 MH + A0 A The results in (12) and (13) involve the loop function   x log x y log y y − . CS 2 2 Mh20 MH 0 peng.2. 2 gives the following contribution to CS and CP : m` tan2 β B+ (xH + . xt ) .E.2 tan2 β P+(xH + . xt ) also contains the contribution from internal up and charm quarks with mc = mu = 0 from the implementation of the GIM mechanism.

This is due to the fact that flavor-conserving self-energies come from the residue of the one-particle pole. and one obtains an incorrect result if one starts with ms = mb and regulates the propagator pole with an off-shell momentum p with p → 0. In these works flavor-changing self-energies have been discussed in a different context. FCNC transitions become meaningless for degenerate quark masses. j = d. ms → 0 (except in the tan βenhanced Higgs couplings. For the down-type quarks these counterterms form two 3 × 3 matrices in flavor space. we discuss how these contributions come into play. 2 differ from the situation with γ . b for our case of external down-type quarks. 2 to properly account for the quark propagator 1/(mb − ms ). where both methods give the same correct result.or Zexchanges.28] that the anti-Hermitian parts of the counterterm matrices for the left-handed fields can be absorbed into a renormalization of the CKM matrix. However. Hence we cannot absorb the anti-Hermitian parts of the flavorto bd changing self-energy matrices into counterterms and they do contribute to our calculation. In addition we find that the approach of [26–28] leads to an inconsistency in our calculation. A treatment of flavorchanging self-energies has been proposed in [26. while in our approach flavorchanging self-energies are part of the non-truncated Green’s function.27] and analyzed in some detail in [28]. the renormalization of the W -boson coupling to quarks. and similarly for the up-type quarks. of course) at the end.H. Self-energies Before we write down the result from the diagrams with self-energies in the external quark lines. it has also been argued [29] that the on-shell scheme of [27. In this respect the Higgs exchange diagrams in Fig. The absorption of the Hermitian parts into wave function counterterm matrices is optional.28] is not gauge invariant. s. 2. By close inspection of the formulae in [28] one also recovers this “factor of 1 rule” from the expressions for the wave function renormalization matrices derived in [28].E. and then take the limit mb .3. and the Hermitian parts of the matrices can be interpreted as wave function renormalization matrices ZijL and ZijR with i. It is crucial to note that one must start with ms 6= mb in the diagrams with external self-energies in Fig. It was argued in [27. . This calculation is straightforward because the internal b quark line is off-shell and therefore it does not contribute to the 1-particle pole of the s quark and needs not be truncated. because unlike in the case of the W coupling renormalization there is no tree-level coupling of a neutral scalar or vector boson ¯ 0 to be renormalized. In [26–28] counterterms have been chosen in such a way that the flavor-changing self-energies vanish if either of the involved quarks is on-shell. We cannot cancel the anti-Hermitian parts of the self-energies in the external lines with the counterterms for the CKM matrix. because the introduction of wave function counterterms only trivially shuffles self-energy contributions into vertex counterterms. It is most straightforward then to avoid the issue of counterterms altogether by simply calculating the fermion self-energy diagrams as they are shown in Fig. Nierste / Nuclear Physics B 586 (2000) 39–55 45 3. Further we note that for ms 6= mb one must include flavor-changing self-energies in external lines with a factor of 1 rather than of 1/2 as in the flavor-conserving case. Logan. one for the left-handed quark fields and one for the right-handed ones. U.

Logan. one with a would-be Goldstone boson G+ and one with the physical charged Higgs H + .4.E. tan2 β 2 r −1 2MW (16) 2 /m2 (m ). Note that the dependence on the masses of the neutral where r ≡ xH + /xt = MH t + t Higgs bosons from the penguin and fermion self-energy diagrams has dropped out in their sum without invoking any relation between the mixing angle α and the Higgs masses. The result depends on only two of the 2HDM parameters: tan β and MH + . Nierste / Nuclear Physics B 586 (2000) 39–55 There are two fermion self-energy diagrams that contribute in the 2HDM. 2 M 0 (15) A 3. This is different from the SM case. CSself = 2 2 Mh20 MH 0 CPself = m` 1 tan2 β (xH + − 1) P+ (xH + . (13) and (15) we obtain the 2HDM Wilson coefficients in (6): CS = CP = m` log r . in which the UV divergence of the G+ diagram cancels with the UV divergence of a SM Higgs vertex diagram involving a G+ and a top quark in the loop. The two hadronic matrix elements involved are related by the field equation of motion .46 H. Their sum is ultraviolet-finite. (12). xt ) 2 . xt ) + . 2HDM Wilson coefficients and branching ratios Adding (10). U. only the internal top quark contributions to the self-energy are relevant here. As in the penguin diagrams involving H + and G+ in the loop. The self-energy diagrams add the following term to the Wilson coefficients:  2  m` sin α cos2 α tan2 β (xH + − 1) P+ (xH + . because the coupling of H + or G+ to quarks is proportional to either of the quark masses and we neglect ms .

 .

µ ¯ γ5 d 0 (x).

0.Bd 0 PB 0 = ifB 0 P µ e−i PBd 0 ·x .

bγ B 0 d d d .

 .

¯ 5 d 0 (x).

Bd 0 PB 0 = −ifB 0 0.

bγ d d m2B d0 mb + md 0 e −i PB 0 ·x d . The resulting decay amplitude is .

 .

  .

 2m` GF αEM mBd 0 fBd 0 .

.

.

. ¯ ¯ C C − C ` m `` + m `γ ξ |A| = √ t B S B P A 5 0 0 .

.

the branching fractions are given by . (19) m Bd 0 mB 0 d where τBd 0 is the Bd 0 lifetime. d d m Bd 0 2 4π sin2 θW (17) (18) Here fBd 0 is the Bd 0 decay constant. The corresponding branching ratio is v u 2 2 α2 m3B 0 τBd 0 fB2 0 u G d d EM + − 2 F t1 − 4m` |ξ | B(Bd 0 → ` ` ) = t 8π m2B 0 32π 2 sin4 θW d " !  2 # 2 4m` 2m` 2 2 × 1− 2 CA mBd 0 CS + mBd 0 CP − . Numerically. normalized according to fπ = 132 MeV.

In the limit of large MH + .E. The analyses of He et al. 4. (20) 2 2 r −1 r −1 mB 8MW 8MW s It is a well known property of the 2HDM that there exists a limit in which the particles H 0 .54 ps 245 MeV 0. If tan β is between a few and 15 the 2HDM corrections are small and experimentally hard to resolve.008 m2B m2Bd d " ! 2   2 2 # mBd tan2 β log r mBd tan2 β log r 2 4m2` − Y (xt ) × 1− 2 + . if tan β = 50 the terms subleading in tan β give a correction only of O(2%). in the paper of Savage [12]. U. CP and CS fall as MH + . but they are suppressed by two powers of cot β compared to the SM terms in (16). Such terms indeed occur. the decays B → `+ `− . the decay B → µ+ µ− is considered in the 2HDM.. Hence the formulae above are sufficient for all purposes.54 ps 210 MeV 0. with and without tree-level flavor-changing neutral Higgs couplings.g. so that an error of order cot β is tolerable as well. and H + become very heavy and decouple from processes at the electroweak energy scale while h0 remains light and its couplings approach those of the SM Higgs −2 particle [30. B → K`+ `− and b → s`+ `− are analyzed in both type-I and type-II 2HDMs. the only diagrams considered are the box diagram involving two charged Higgs bosons and the A0 penguin involving H + and W + in the loop. Thus the deviation of −2 the branching fractions from their SM prediction falls as MH + in the large MH + limit. and the effects of the enlarged Higgs sector decouple. Nierste / Nuclear Physics B 586 (2000) 39–55 47 v   2  2 2 u u  4m2` m τ | f |V B B t d d d ` t B(Bd → `+ `− ) = 3. as is the case for the SM contribution. Comparison with other calculations 4. Logan. the A0 penguin involving H + and G+ in the loop is not considered.1. Although the calculations of [11] are performed in the ’t Hooft–Feynman gauge. . Similarly. 2 2 r −1 r −1 m Bd 8MW 8MW v  2  2 2 u  u  4m2` m fBs τB s |Vt s | ` t 1 − B(Bs → `+ `− ) = 1.1 × 10−5 1.040 m2B m2Bs s " ! 2  2 2 #  mBs tan2 β log r mBs tan2 β log r 2 4m2` − Y (xt ) × 1− 2 + .31]. Next we discuss the accuracy of the large tan β approximation. Only the contribution of the A0 penguin is considered. A0 .0 × 10−7 1 − 1. e. and of Savage In the paper of He. Nguyen and Volkas [11]. Subleading terms in tan β could be enhanced by powers of mt /mb compared to our result in (16).H. several diagrams that are important at large tan β and required in order to obtain a gauge-independent result are neglected. In [11]. In both [11] and [12].

Leaving out the box diagram. In [15–17] the Wilson coefficients in (6) are calculated for supersymmetric models with large tan β. Huang and Huang [14]. There are two remaining discrepancies. Nierste / Nuclear Physics B 586 (2000) 39–55 4. Considering only terms proportional to tan2 β. The analyses of Huang et al. This is partly due to a typographical error in [14]. with two important exceptions. We argue in Appendix A that the trilinear couplings should not be considered tan β enhanced.2. H 0 . We have neglected these contributions in our analysis. after leaving out the box diagram.E.48 H. Second. Therefore we conclude that the penguin diagram with two H + bosons in the loop should not be included in the O(tan2 β) calculation because it is of subleading order in tan β. and the box diagram with a charged Higgs boson and W boson is not included. If the diagrams involving supersymmetric particles are neglected. As in our calculation. our result for the A0 penguin diagram involving H + and G+ in the loop differs from that of [13] by a sign. In [13] additional penguin diagrams are considered that are not proportional to tan2 β. This diagram is included in [14] because it apparently receives a factor of tan β from the trilinear H + H − H couplings (H = h0 . the authors of [13] incorrectly conclude that the box diagram involving H + and W + is negligible and therefore neglect it. In the paper of Dai. A0 ). we find that the sum of the remaining contributions proportional to tan2 β is gauge-dependent. Our sign is required for the gauge-independence of CP . our results for the individual diagrams agree with those of [13]. H 0 ). This . These additional penguin diagrams can be important for small values of tan β and in regions of the parameter space where some of the Higgs quartic couplings are very large resulting in large trilinear H + H − H couplings (H = h0 . The analysis of Skiba and Kalinowski In the paper of Skiba and Kalinowski [13]. the authors of [14] consider only the penguin and fermion self-energy diagrams with neutral Higgs boson exchange and neglect the box diagram with a W and charged Higgs boson in the loop. only the diagrams proportional to tan2 β are considered. Again. and Choudhury et al. our results for CS and CP in the ’t Hooft–Feynman gauge do not agree with those of [14].3. If we neglect the box diagram. because they are not relevant for the interesting case of large tan β. in [14] a contribution from the h0 and H 0 penguin diagrams with two H + bosons in the loop is included. Logan. 4. Still. However. the Wilson coefficients in (6) are calculated in the type-II 2HDM at large tan β. but rather contain one or no powers of tan β. affecting the numerical result substantially. First. which is corrected in [15–17]. Second. First. this calculation reduces to that for the type-II 2HDM with parameters constrained by supersymmetric relations. our result for the A0 penguin diagram involving H + and W + in the loop differs from that of [14] by a sign. our result for CS and CP in the ’t Hooft–Feynman gauge does not agree with the non-SUSY part of that of [15–17]. in [15–17] only the diagrams with neutral Higgs boson exchange are considered. U. In the ’t Hooft–Feynman gauge employed in [13] the omitted diagram gives the dominant contribution. the decay Bs → τ + τ − is analyzed in the type-II 2HDM.

None of the results in [11–17] passes the test of gauge-independence and. Because the 2HDM Wilson coefficients in (16) depend on only two of the parameters of the 2HDM. B(Bs → `+ `− ) is significantly larger than the corresponding branching fraction for Bd . Nierste / Nuclear Physics B 586 (2000) 39–55 49 discrepancy arises because our result for the penguin diagrams involving H + and W + in the loop differs from that of [15–17] by a sign. They include an additional renormalization group factor to account for the running of the Wilson coefficients. Because of the suppression of the branching fractions by m2` /m2B . Once SUSY relations are imposed on the Higgs sector. 5. it is clear that the penguin diagrams with two H + bosons in the loop are not of order tan2 β. clearly the decay to τ pairs is the largest of the leptonic branching fractions in both the SM and the 2HDM. Therefore. In conclusion. Phenomenology As can be seen from the numerical coefficients in (20). For this reason we concentrate on the decays of Bs . (22) where we have taken the central values for all inputs in (20) and ignored the 2HDM contributions as well as the errors in the hadronic parameters. as we did. Yet these authors have overlooked that the running of the (chiral) scalar quark current in QS and QP (see (7)) is compensated by the running of the b-quark mass multiplying the currents as explained in Section 3. we may neglect Y (xt ) in .H. the behavior of the result in different parts of parameter space is easy to understand. this decay is very difficult to reconstruct experimentally (due to the two missing neutrinos).1 × 10−9 (SM). this event was consistent with the expected background and lay in the overlapping part of the search windows for Bd and Bs . This is due primarily to the relative sizes of |Vt s | and |Vt d |.6 × 10−6 (expt). However. Logan. where one candidate event for B → µ+ µ− has been reported. U. even though the production rate of Bs is three times smaller than that of Bd at the Tevatron. The best experimental bound comes from CDF [32]. and our calculation does not agree with any of them. each contains mistakes. the bounds on the leptonic branching fractions of Bs are much closer to the SM predictions than those of Bd [32]. This leads to an underestimate of the Wilson coefficients by roughly 23%. tan β and MH + . (21) The SM prediction for the branching fraction is B(Bs → µ+ µ− ) = 4. As a result. and the authors of [15–17] have omitted these diagrams. in our numerical analysis we focus on the decay Bs → µ+ µ− . In regions of the parameter space with a large 2HDM contribution to B(Bd 0 → `+ `− ) compared to the SM contribution. A final critical remark concerns the treatment of the renormalization group in the paper by Choudhury and Gaur [17]. for which the experimental bound is the closest to the SM prediction.E. The corresponding 95% confidence level upper bound on the Bs → µ+ µ− branching fraction is [32] B(Bs → µ+ µ− ) < 2. in our opinion. the papers in [11–13] and [14–17] disagree with each other. and as a result the experimental limits on B decays to τ pairs are very weak.

3. obtained from the current 95% confidence level experimental upper bound of B(b → sγ ) < 4. 75. The vertical line is the lower bound on MH + in the type-II 2HDM from b → sγ [33–37]. The effect of the destructive interference can clearly be seen in Fig. Then the result is particularly simple: the branching fractions are proportional to tan4 β log2 r/(r − 1)2 .5 × 10−4 from the CLEO experiment. For comparison we show the constraint on MH + from b → sγ [33–37]. for various values of tan β. In Fig. 50 and 25. If we ignore the kinematical factor of 1 − 4m2` /m2B 0 in front of d CS in (19) (which is a good approximation for Bd 0 → µ+ µ− but not for Bd 0 → τ + τ − ) then we can easily show that the branching fraction in the 2HDM crosses the SM value when m2Bs tan2 β log r = Y (xt ). the branching fraction drops. 3 we plot the predicted value of B(Bs → µ+ µ− ) in the 2HDM as a function of MH + .E. For very large tan β and relatively light H + . Logan. B(Bs → µ+ µ− ) in the 2HDM as a function of MH + for tan β = 100. We can see from (20) and the value of Y (xt ) given in (9) that the interference between the SM and 2HDM contributions to the branching fractions is destructive. the 2HDM contribution dominates and the branching fraction is significantly enhanced compared to its SM value. . For comparison we show the current experimental bound [32] and the SM prediction for the branching fraction. 2 r −1 8MW (23) and reaches a minimum of half the SM value when Fig. eventually falling below the SM prediction due to the  destructive interference.50 H. U. As the 2HDM contribution becomes smaller due to increasing MH + or decreasing tan β. 3. 60. Nierste / Nuclear Physics B 586 (2000) 39–55 (20).

4. 4. We also show two contours for the expected sensitivity with Fig. The vertical dashed line is the present lower bound on MH + in the type-II 2HDM from b → sγ [33–37]. shown by the solid line at the far left of Fig. taking tan β = 60 and MH + = 175 GeV (500 GeV). Nierste / Nuclear Physics B 586 (2000) 39–55 m2Bs tan2 β log r 1 = Y (xt ). Contours are chosen as described in the text. As a numerical example. to 6. The contours shown (from left to right) were chosen as follows. For the values of tan β that we consider. shown by the short dashes in Fig. this sensitivity will still only probe values of MH + already excluded by b → sγ .1 × 10−9 ). 4 show the regions that we expect to be probed at the Tevatron Run II and extended Run II with various amounts of integrated luminosity. With 2 fb−1 from each of the two detectors.6 × 10−6 from CDF [32] with about 100 pb−1 of integrated luminosity. 4 we plot the regions of MH + and tan β parameter space that will be probed as the sensitivity to the decay Bs → µ+ µ− improves at the Tevatron Run II. the sensitivity should improve by a factor of 40 over the current sensitivity. 2 8MW r − 1 2 51 (24) These correspond to tan2 β log r/(r − 1) = 1790 and 893.H.E. respectively. The rest of the contours in Fig. then the sensitivity will scale only with the square root of the luminosity. In this case the sensitivity to the branching fraction should scale with the luminosity. Such low H + masses are already excluded by the constraint from b → sγ [33–37].5 × 10−8. 4. Logan. assuming that the background for this process remains negligible. we find B(Bs → µ+ µ− ) = 1. In Fig. If there is background. The current upper bound on B(Bs → µ+ µ− ) is 2. however. This bound excludes a small region of parameter space with very high tan β and very light H + . Regions of MH + and tan β parameter space probed by measurements of B(Bs → µ+ µ− ).8 × 10−8 (2. U. .

U. and LHCb: 33 (10). With 10 fb−1 we expect the sensitivity to reach a branching fraction of 1. . These uncertainties will be reduced as the B physics experiments progress and lattice calculations improve.g. These correspond to an extended Run II of the Tevatron. Neither have we taken into account the theoretical uncertainty. e. a fit to the Z decay data in the 2HDM [39] puts weak constraints on the H + mass for very large tan β: 3 We do not consider charged Higgs masses larger than 1 TeV for naturalness reasons. Also QCD corrections to the 2HDM contribution will arise at NLO and require the calculation of two-loop diagrams. In particular.52 H. for tan β = 60. Since the suppression of this branching fraction in the 2HDM is at most a factor of two.5 ×10−9 . the NLO corrections increase the decay amplitude by about 3%. the measurement of b → sγ will improve as well. the branching fraction is suppressed by 30% or more compared to the SM. for tan β = 60 and 375 GeV < MH + < 1 TeV. primarily the B meson decay constants and CKM matrix elements in (20). a non-observation of Bs → µ+ µ− at this sensitivity would rule out H + lighter than 260 GeV. For. the 2HDM contributions to Bs → µ+ µ− depend sensitively on tan β. Logan. assuming the SM branching fraction [38]: ATLAS: 27 (93). allowing one to begin to probe H + masses above the current bound from b → sγ for tan β > 54. the branching fraction in the 2HDM is enhanced by 30% or more compared to the SM. respectively). CMS: 21 (3). While the 2HDM contributions to b → sγ are independent of tan β for tan β larger than a few. while b → sγ will remain more sensitive for moderate and small tan β. the LHC experiments will be able to observe this decay for any configuration of the 2HDM at large tan β. 3 In these regions. Finally. We expect the largest theoretical uncertainty to come from uncertainties in the input parameters. in which case our conclusions are not significantly modified. hence the LHC measurement will give powerful constraints on tan β in the large tan β region. less than a factor of two above the predicted SM branching fraction. In the region of large MH + . we must compare it to other measurements that constrain the 2HDM in the large tan β regime. the dependence on MH + is very weak. it must be sensitive to a range of parameter space not already explored by b → sγ at each integrated luminosity. tan β = 60 and MH + < 285 GeV. Similarly. Fortunately. Looking farther into the future. In the SM. the experiments at the CERN LHC expect to observe the following numbers of signal (background) events for Bs → µ+ µ− after three years of running at low luminosity. We have made no attempt to simulate the experimental background for this decay in order to obtain an accurate estimate of the reach of the Tevatron Run II.E. 4. If Bs → µ+ µ− is to be a useful probe of the 2HDM. As the statistics of B physics experiments improve.3 × 10−8 . In order to evaluate the usefulness of Bs → µ+ µ− as a probe of the 2HDM. Bs → µ+ µ− is complementary to b → sγ because of the different dependence on tan β. and therefore increase the branching fraction by about 6%. With 20 fb−1 we expect a reach of 6. we expect the LHC to be able to distinguish the 2HDM from the SM.. This would allow one to probe H + masses above the current bound from b → sγ for tan β > 47. Nierste / Nuclear Physics B 586 (2000) 39–55 10 fb−1 and 20 fb−1 of integrated luminosity per detector (dotted and dot-dashed lines in Fig. This makes Bs → µ+ µ− an especially sensitive probe of the 2HDM in the large tan β regime. We expect the NLO corrections to the 2HDM contribution to be of the same order.

The fit gives no constraint for tan β < 94. We showed that when all the contributions are properly included in the large tan β limit. Appendix A. 6. the resulting expressions for the branching fractions are quite simple and depend only on tan β and the charged Higgs mass. We would also like to thank the conveners of the B-Physics at Run II Workshop at Fermilab for facilitating useful interaction among theorists and experimentalists. it promises to be an experimentally and theoretically clean probe of new Higgs physics. Acknowledgements We are grateful to Jan Kalinowski and Witold Skiba for helpful discussions. U. These couplings depend strongly on the model considered. We find that for reasonable values of tan β up to 60 and charged Higgs masses above the lower bound set by b → sγ . Conclusions In this paper we have analyzed the decays Bd 0 → `+ `− in the type-II 2HDM with large tan β. H 0 ) for the non-supersymmetric 2HDM were first presented in [13]. For the most general CP-conserving 2HDM with natural flavor conservation. providing tantalizing search possibilities with the potential to probe large parts of the large tan β parameter space of the 2HDM. for which the the experimental sensitivity is best. the H + H − H couplings are given by igQH /MW .and 2HDM-induced contributions to the decay rate in an earlier version of this manuscript.E. where  2 sin(β − α)(2λ3 + λ4 ) − sin β cos β cos(α + β)λ5 Qh0 = −vSM  + 2 sin β cos β(− sin α sin βλ1 + cos α cos βλ2 ) . Nierste / Nuclear Physics B 586 (2000) 39–55 53 MH + > 40 GeV at 95% confidence level for tan β = 100. These 2HDM contributions can enhance or suppress the branching fractions by a significant amount compared to their SM values. Trilinear Higgs couplings The trilinear H + H − H couplings (H = h0 . B(Bs → µ+ µ− ) can be increased by up to a factor of five above its SM expectation or suppressed by up to a factor of two. Logan. Finally we owe a debt of gratitude to Piotr Chankowski and Łucja Sławianowska for pointing out an error in the relative sign between the SM. Although these decays have been studied in a 2HDM before. and to Jonathan Lewis for discussions on the reach of the Tevatron in Run II. and for confirming our result for the 2HDM Wilson coefficients. depending on the charged Higgs mass.H. We have focused in our numerical analysis on Bs → µ+ µ− . which is the dominant contribution at large tan β in the ’t Hooft–Feynman gauge and is needed for gauge independence. Although very high statistics will be needed to observe this decay. the previous analyses omitted the box diagram involving W and H + . .

Krizan et al. B 404 (1993) 3. A study of CP violation in B meson decays: Technical design report. ATLAS Technical Proposal. Lett. [13] W. B 266 (1991) 135. BaBar technical design report. Phys. D 38 (1988) 814. one must make an assumption to eliminate one of the independent λi . References [1] D.J. in: Proceedings 4th Workshop on Heavy Quarks at Fixed Target (HQ 98). A 351 (1994) 111. sin α ∼ cot β. Nierste / Nuclear Physics B 586 (2000) 39–55  2 QH 0 = −vSM cos(β − α)(2λ3 + λ4 ) − sin β cos β sin(α + β)λ5  + 2 sin β cos β(cos α sin βλ1 + sin α cos βλ2 ) . Minkowski.V. CMS Technical Proposal. [9] M. B 155 (1979) 52. SLAC-R-0457. Kalinowski.S. T. Ellis. Phys. [6] The CMS Collaboration. In [13]. Logan. Rev. Skiba. 2 Naively.. [12] M. Cheng et al.E. BELLE-TDR-3-95. D. 93 (1975) 193. Fritzsch.R.K. for Fermilab D0 and CDF Collaborations. (25) reduces to   2 cos α(2λ3 + λ4 ) 1 + O(cot β) . [3] K. To write these couplings in terms of Higgs masses and mixing angles. Nucl. Phys. (26) QH 0 ' −vSM These couplings are not enhanced at large tan β. Meth. Batavia. Savage. M. formulae are presented for the two cases λ1 = λ2 and λ5 = λ6 . CERN/LHCC/98-4. Georgi. Volkas.D. R. we find at large tan β. The formulae in (25) agree with [13] in these two cases. CERN/LHCC/94-43. Qh0 ' −vSM   2 sin α(2λ3 + λ4 ) 1 + O(cot β) . Phys. [2] M. U. Here vSM = 174 GeV is the SM Higgs vev. At large tan β. Phys. J. B 128 (1977) 506. Nanopoulos. ..   1 2 2 Qh0 ' − MH 0 tan β cos α sin α 1 + O(cot β) . Nucl. [11] X. one would conclude that these couplings are enhanced at large tan β. [10] H. Phys. Nucl. Nucl. HERA-B.54 H. LHCb Technical Proposal. Chanowitz. Nguyen. Pitts. At large tan β we have tan 2α =   2(4λ3 + λ5 ) cot β 1 + O(cot2 β) . and the tan β enhancement in (27) is cancelled. USA. an experiment to study CP violation at the HERA proton ring using an internal target. λ5 − 4(λ2 + λ3 ) (28) Thus for generic values of the λi . This is incorrect because the angle α depends on tan β. Belle Collaboration.G. Considering instead the case λ1 = λ2 and writing the trilinear couplings in terms of Higgs masses and mixing angles.T. He. Gaillard. Ann. [7] The LHCb Collaboration. Boutigny et al. Instrum. [4] P. (25) The H + H − A0 coupling is zero. J. 2   1 2 (27) QH 0 ' Mh0 tan β cos α sin2 α 1 + O(cot β) .. 1998. [8] H. [5] The ATLAS Collaboration. CERN/LHCC/94-38. P. and the λi are the scalar quartic couplings of the Higgs potential given in [18].

Pocsik (Eds. D 59 (1999) 011701. Rev. Abe et al.. and 1994. July 23–29. Sack.-S. B Decays. 1998. MA. H. Aoki. S.E. Lebedev. Phys. Addison-Wesley. J. Sirlin. Rev. S. (Erratum). 1990. Urban. U. F. A. Reading. Liao. 1995.A. Huang. Bertolini. Paschos. hep-ph/0003238. C. G. Huang. Kane. CDF collaboration. Krawczyk. B 347 (1990) 203. Rev. Phillips. Lett. Gaur. S. D 15 (1977) 1966. T. Phys. T.-W. B. 2nd edn.E. M. Misiak. hep-ph/9302272 (Errata). Canada.-S. pp. 65 (1981) 297. Yan. hep-ph/0002106. Lim. 1992. C 18 (1983) 43. M. Marciano. hep-ph/9501320. Nucl. Phys. 458–478. B 454 (1999) 98.J. P. Lett. 1. E. P.. K.I.L. C. World Scientific.-S. Rev. Phys. A. C. Nucl. Phys. B 400 (1993) 225. Buras. Csikor. Nir. Huang. Phys. Prog. Theor. Stone (Ed. hep-ph/9810240. Kawabe. Z. Muta. The Higgs Hunter’s Guide.. Nucl. Phys. Choudhury.-S. Prog. G. G. D 15 (1977) 1958. P. Z. Nierste / Nuclear Physics B 586 (2000) 39–55 [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] 55 Y.).-S. Phys. Lett.. Weinberg. M. Loinaz. Lett.M. Q. in: F.J. 73 (1982) 1. Konuma. Alam et al. Phys. Dai. N. Rev. S. Phys.R. Rev. Vancouver. F. Huang. Dawson. F. J.H. Phys. Krawczyk. Inami. H. 70 (1993) 1045. 620–643. B 451 (1999) 86. Z. Berger. Masiero. Haber. B 93 (1975) 303. P. Gambino. Phys. Borzumati. Suppl. Grz¸adkowski. Y. Denner. W. Singapore. Theor. D 57 (1998) 3811.N.E. S. CLEO collaboration. T. Ball et al. Glashow. Lett. Borzumati. pp. Phys. Probing new physics in FCNC B decays and oscillations. H.S. M. Prog. O. C. Buchalla. Lett. P. B 390 (1997) 257. Takeuchi. C. .-B. Madricardo. Lett. B 335 (1990) 363. Phys. CLEO CONF 98-17. F. B 442 (1998) 209. J. 70 (1993) 1368. Gunion. Haber. Phys. Greub. Phys. W. Grassi.). W. Singapore. in: S. R. Hewett. Hioki. Nucl. Budapest Electroweak 1994. Phys. R. Q..J. 1st edn.F. Yan. A. C 44 (1989) 509. Phys. B 451 (1999) 161.E. H. talk presented at XXIX International Conference on High Energy Physics. p. A. Haber. Logan. Phys. Phys. Theor. T. S.A. Barger.L. World Scientific. 65 (1981) 1772. V.

For energies sufficiently far above threshold it can be predicted by perturbative QCD. E-mail: rharlan@bnl. respectively (for a review see [7]). Upton. In this limit.gov 0550-3213/00/$ – see front matter  2000 Elsevier Science B. NY 11973.X .V.38. After expansion in m2 /s the quartic terms. i. Universität Karlsruhe.elsevier. All rights reserved.Nuclear Physics B 586 (2000) 56–72 www. in a number of interesting cases quark masses do play an important role [7]. All rights reserved. D-76128 Karlsruhe. For Z decays into bottom quarks the mass effects have to be included as a consequence of ∗ Corresponding author. accepted 20 June 2000 Abstract The total cross section for the production of massive quarks in electron–positron annihilation can be predicted in perturbative QCD.65.H. R. USA Received 16 May 2000.Bx.nl/locate/npe Quartic mass corrections to Rhad at O(αs3) K. at the same time. By combining these results with the prediction for massless case and the quadratic mass terms the cross section for massive quark production at electron–positron colliders is put under control in order αs3 from the high energy region down to fairly low energies. Kühn a a Institut für Theoretische Teilchenphysik. those proportional to m4 /s 2 . J.. 13. the relative size of the large order terms decreases by replacing the running mass m(µ) with the scheme independent invariant one m b. Harlander b. PII: S 0 5 5 0 .  2000 Elsevier Science B. The stability of the prediction is improved and.-t.∗ .+i The total cross section for hadron production in electron–positron annihilation is one of the most fundamental observables in particle physics.V.3 2 1 3 ( 0 0 ) 0 0 3 9 3 . 12. It allows for a precise determination of the strong coupling αs and. The αs3 corrections are shown to be comparable to terms of order αs and αs2 . As a consequence.G. However. the results of O(αs2 ) [1–3] and O(αs3 ) [4–6] were obtained more than two and nearly one decade ago. the predictions exhibit a sizeable dependence on the renormalization scale. Predictions relevant for charm. In many cases the cms energy is far larger than the quark masses which motivated the original calculations to be performed in the idealized case with the masses set to zero from the start.38. Germany b Brookhaven National Laboratory.e. and it is well accessible experimentally from threshold up to the highest energies of LEP and a future linear collider. Chetyrkin a .V. PACS: 12. for a test of its evolution dictated by the renormalization group equation. bottom and top quarks production are presented. are calculated up to order αs3 for vector and axial current induced rates. once precision measurements at different energies are available.

In fact. Other cases of interest [8. exploiting the optical theorem. from high energies down to values of 2m/ s = 0.G. They were obtained by reconstructing the logarithmic αs3 m2 /s terms of Π(q 2 ) from the full three-loop O(αs2 m2 /s) result of [17] with the help of the renormalization group equations. numerical studies [15] as well as qualitative arguments demonstrate that the sum of the first two or three terms of the expansion can be trusted even down to 3m or even. Compared to the case of the m2 /s terms an important difference arises: Even adopting the MS definition (at scale µ2 = s) for the quark mass. Bottom quark production at lower energies is affected from production threshold up to a few tens of GeV. logarithmic terms ∝ ln(s/m2 ) remain in order αs2 and above. In addition to the anomalous mass dimension and the β-function the anomalous dimensions of the operators of dimension four are required in appropriate order. To predict the R ratio. it is well justified to consider only one quark as massive and to neglect the masses of the lighter quarks. These are sufficient to calculate the m2 /s terms of the imaginary part in the time-like region. A generalization of this approach has been formulated for the quartic mass terms in [19. In three-loop approximation (O(αs2 )) the corresponding results were obtained only recently. This method also allows to determine the αs3 m4 /s 2 terms. The result is cast into a particularly compact form once expressed in terms of the √ running mass m(µ2 ).9] are charm production between roughly 5 and 10 GeV and. these logs may be also summed up (see Refs. last but not least. with some optimism. The calculation can be reduced to the evaluation of massless propagators and massive tadpole integrals. which provides an expansion of Π(q 2 ) in powers (m2 /q 2 )n .14] and references therein). however. 2. Nevertheless. / Nuclear Physics B 586 (2000) 56–72 57 the extremely precise measurements. The heavier quarks decouple (apart from the .20] and was originally adopted for the calculation of αs2 m4 /s 2 terms. Two methods have been used for this purpose: the first one is based on the evaluation of a large number of terms for the Taylor series of the polarization function Π(q 2 ) at q 2 = 0 and an appropriately chosen analytic continuation [12].65 − 0. both at most in three-loop approximation. From the comparison between full result and expansion one learns that the first few terms of the high energy expansion provide a remarkably good description of √ the full result. since this choice eliminates all terms ∝ ln(s/m2 ). Chetyrkin et al. The second one is based on the application of the large momentum expansion (see [13. with the ’t Hooft scale µ set to s throughout [18]. both real and imaginary parts are available [11]. top quark production at a future linear collider [10]. These considerations pave the way to a prediction of R(s) including the quark mass dependence to O(αs3 ) along the following route: In addition to the massless result the m2 /s terms of O(αs3 ) have been calculated for the absorptive part nearly a decade ago [16].75. Details of the calculation will be given elsewhere [21]. modulo logarithms [15]. [19.5m (where m stands for the pole mass). The power of these logarithms.22]). is reduced by one. It is based on the operator product expansion and the usage of the renormalization group equation to again construct the logarithmic terms of the polarization function.K. The existence √ of the four quark threshold at s = 4m and of a corresponding branching point for the √ polarization function suggests that the high energy expansion diverges for s below 4m. In two-loop approximation the full mass dependence of the cross section has been evaluated since long and.

the massive quark then only appears through its coupling to virtual gluons.n and rQ. For low energies only the electromagnetic current contributes. (v/a) rQ (v/a) (v/a) = r0 + rQ. Denoting nf the number of active flavors. vq/Q (aq/Q ) is the coupling constant to the light/heavy quarks in the vector (axialvector) case. receives additional contributions from the axial current. both the electromagnetic and neutral current pieces are relevant (v/a) represent and have to be included with appropriate weights (see.. Chetyrkin et al. (v/a) (2) r0 denotes the massless approximation. [23]).4 + · · · .58 K. The sum runs over all massless quark flavors. (1) R (v) = 3 q and similarly for the axial-vector part (v → a).00517836n2f . v → e and a → 0.g. They originate from two different types of diagrams: For rq the external current couples to massless quarks. Both vector and axial-vector will be considered separately in the following. Top quark production. In the energy region where quark mass effects are relevant. The massive quark will generically be denoted by Q in the following. e. For high energies.98571 − 0. If not stated otherwise.63694 − 1. rsing .2 + rq. the renormalization scale µ2 = s is adopted below.G. rq is the same for external vector and axial-vector currents. where s is the cms energy and mQ is Both rq and rQ the MS mass of the quark Q: rq = r0 + rq.n are the mass terms of order mnQ . comprises massless and massive singlet contributions. rq and rQ the non-singlet contributions arising from diagrams where the external currents are linked by a common quark line. The prediction for both of them is conveniently split up as follows: X  (v) 2 (v) vq2 rq + vQ rQ + rsing . while q refers to all the lighter quarks. the massless terms are given by (all the following formulæ are valid up to O(αs3 ) unless indicated otherwise)  2    αs αs 365 11 2 + − 11ζ3 + nf − + ζ3 r0 = 1 + π π 24 12 3  3  1103 275 87029 121 αs − ζ2 − ζ3 + ζ5 + π 288 8 4 6     1 262 25 19 7847 11 2 151 + ζ2 + ζ3 − ζ5 + nf − ζ2 − ζ3 + nf − 216 6 9 9 162 18 27  2 αs αs + (1. / Nuclear Physics B 586 (2000) 56–72 tiny axial-vector singlet contribution — see below) and can therefore be omitted in the calculations. charm and bottom production is solely induced by the electromagnetic vector current. where either of the external currents is coupled to a separate closed quark line. finally. rQ corresponds to diagrams where the external current couples to the (v/a) massive quark.115295nf ) ≈1+ π π  3  αs −6.2 + rQ. (v/a) are written as series in m2Q /s.20013nf − 0. (3) + π . while rq. however.4 + · · · . (v/a) On the other hand. which sets in above 350 GeV.

24]:      m2Q αs 3 32 8 − ζ3 −80 + 60ζ3 + nf rq. we present the quartic mass corrections up to O(αs3 ):  2 2  2   mQ αs 13 − lms − 4ζ3 rq.2 = s π 9 3   m2Q αs 3 [−7.4 = s π 3    3  50 9707 43 αs 2 + 2lms + 15ζ2 + 25ζ3 + ζ5 + lms − 22ζ3 − + π 144 3 12    22 457 2 13 4 − ζ2 − ζ3 + lms − + ζ3 + nf 108 3 9 18 3 . Chetyrkin et al.167nf + 1.14 − 104.21819n2f .K.3578nf − 0.33333nf ) 12 + ≈ s π π   3  αs 1032.2 = s π π 2 3  3  490 5225 855 αs ζ2 + ζ3 − ζ5 2442 − + π 2 3 6     2 466 1045 4846 2 125 + 34ζ2 − ζ3 + ζ5 + nf − ζ2 + nf − 27 27 27 54 3  2  m2Q αs αs (126.378270nf . Finally. separated according to (1).247 + 73. + π 59 (4) (5) (6) (7) The quadratic mass terms in rq contribute in O(αs3 ) and higher only [16]. + π   2  2 mQ αs αs 8221 (a) −6 − 22 + + 57ζ2 + 117ζ3 − rQ.48524nf ) −6 − 22 + ≈ s π π   3  αs 2 −409.5 − 4.140 + 4. / Nuclear Physics B 586 (2000) 56–72 The quadratic mass corrections.G.2 = s π π 24   151 − 2ζ2 − 4ζ3 + nf 12  3  121075 4613165 αs + 1340ζ2 + ζ3 − 1270ζ5 − + π 864 36   656 72197 209 − ζ2 − ζ3 + 5ζ4 + 55ζ5 + nf 162 2 3   26 13171 16 + ζ2 + ζ3 + n2f − 1944 9 9  2  2 mQ αs αs (−108.87659 + 0. read [16. ≈ s π  2    m2Q αs αs 253 13 (v) − nf 12 + rQ.35007nf ].

59784 − 22.95 − 244.612lms + 13lms + π  2 nf + −275.666667lms  +(4.4 = (8) m2Q (9) .8619lms + 2lms   + (0. Chetyrkin et al.84078 + 0. +  2  2  αs 13 αs 2977 + 162ζ2 + 108ζ3 − lms −6 − 22 + − s π π 12 2   8 143 1 + lms − 4ζ2 − ζ3 + nf 18 3 3  3  64123 13285 1274461 12099 αs + ζ2 + ζ3 − ζ5 − + π 432 4 18 9    1672 1309 130009 574 2 − 22ζ3 + 13lms − ζ2 − ζ3 + lms − + nf 6 648 3 9    440 199 4 2 2 ζ5 + lms + ζ3 − lms + 10ζ4 + 9 12 3 3   28 5 463 23 + ζ2 + ζ3 − lms + n2f − 972 9 27 27  2 2  mQ αs −6 − 22 ≈ s π  2  αs 148.218 − 6.898 + 18.196497 + 0.4 = m2Q  2  2  αs 75 αs 1147 − 162ζ2 − 224ζ3 + lms 6 + 10 + s π π 4 2   16 7 41 + nf − + 4ζ2 + ζ3 − lms 6 3 3  3  20147 2225 222421 10881 αs − ζ2 − ζ3 − ζ5 + π 48 4 6 3   4385 2 − 22ζ3 − 75lms +lms 6  5002 1040 200923 + 196ζ2 + ζ3 − 10ζ4 + ζ5 + nf − 648 27 27  (a) rQ.333333lms )nf + π  3  αs 2 4776.97396 − 0.88052lms )nf .5lms + (−1.474894 − lms ) 2 4.1861lms − 0.185185lms )n2f .  (v) rQ.G. / Nuclear Physics B 586 (2000) 56–72  ≈ m2Q s  2  αs π 3 αs π 2 (−0.60 K.

The separate piece from massless cuts is known in analytical form [25] and has to be subtracted if one wants to obtain the contribution from the final states with top quarks.G. one should take into account the full top–bottom doublet. the contributions where the top quark couples to one of the external currents are not suppressed by powers of 1/m2t .8121vQ vQ + vq +O . [10]. Considering the axial vector singlet contribution to top quark production.K. However. Chetyrkin et al.99 + 37. The complete result is given in Ref.15737 − 2.5lms + (6.851852lms )n2f .66667lms nf   + (−3. The singlet contributions in the axial-vector case are exceptional in the sense that for bottom quark production the top quark does not decouple. s s q rsing = αs π  (11) They do not receive m2 corrections [16]. on the other hand. Thus the singlet contribution to t t¯ production is given in expanded form by . They turn out to be small. i. this strategy induces completely massless final states without any top quarks.151 − 62..33333lms )nf + π  3 αs 2 −4646. so we will neglect them in the following.388lms − 75lms + π  2 + 264.10948 + 0.e. The corresponding corrections up to O(m2b αs3 ) have been computed in [25–29].9250lms + 4.22 + 704.413180 vQ + ≈ π q   2 3  X  m2Q 2  mQ + 17. At order αs2 the expansion for the contributions from the full top–bottom doublet starts with the m6 terms [10]. 61 (10) The singlet contributions in the vector case are numerically small:  (v) 3   X 2  55 5 − ζ3 vq vQ + 216 9 q   X  20 50  m2Q 2 vQ vQ + vq − + ζ3 + s 9 3 q    3  2 X αs vq −0. / Nuclear Physics B 586 (2000) 56–72    2323 4 14 2 + ζ3 + lms + lms − 36 3 3   29 20 23 2995 − ζ2 − ζ3 + lms + n2f 972 9 27 27  2 2  mQ αs 6 + 10 ≈ s π  2  αs −248. because the axial anomaly cancels in this combination.

0 9.153 3.0952 152 (4) (5) (6) αs (s) (6) √ mt ( s ) .80675lms − 0.756 0. In principle it should be subtracted from the following result in order to arrive at the production rate for top quarks:        1 αs 3 m2t 2 380 1520 160 40 0 ζ3 + ζ5 + nf ζ3 − ζ5 − rsing = 4 π s 3 3 9 9     1 αs 3 m2t 2 (373.784 0.769 0.0 460. Chetyrkin et al.5 0.097 154 0.829 0.803 0. The separate contribution from the massless cuts is not yet available in order αs3 .0 12.333333lms + s  + O αs3 .153377 + = 4 π s   2 2  m 2 3 − 0.0 10.G.0 αs (s) (5) √ mb ( s ) 0.174 0.161 3.74 √ s 11.171 3.0 7.0 0.212 0.745 0.174 3.0 16.2 0. (13) ≈ 4 π s The prime indicates that this expression still contains the contributions from purely massless final states.62 K. Matching for αs is performed at the values m(m) given in (14) √ s 5.18565 − 2.0918nf ). We explicitly denote the mass by mt here in Table 1 Running coupling and masses at different scales.0961 153 0.116 − 13.179 0.61 0.192 0.222222lms 1. / Nuclear Physics B 586 (2000) 56–72      4 20 1 αs 2 5 2 m2 2 2 − ζ2 + − + ζ2 + 4lms − lms rsing = 4 π 4 3 s 3 3 3     2 2  16 2 3 13 2 4 1 2 m − lms + ζ2 − ζ3 + lms −5 + ζ2 − lms + s 2 3 3 3 3 9  3 + O αs     m2 1 αs 2 2 9.44 0.666667lms 0.0 20.0 10.35 √ s 420.0 6. as mentioned before.5 αs (s) (4) √ mc ( s ) 0.63289 + 4lms − 0.0 8.172 0.165 3.185 0.0 500.0 14. (12) where the weak coupling at2 = ab2 = −at ab = 1/4 has been pulled out for clarity.57 0.

and 420– 450 GeV for top contributions. The charm. on the one hand.118. (b) bottom and (c) top production.5–6 GeV for charm.2 GeV. Let us now investigate the numerical significance of mass effects for the charm and bottom case. this result applies only to top production. we have to take into (n ) account the matching conditions for αs f when going from nf to nf ± 1. b) doublet has been taken into account.K.G. mc (mc ) = 1. Fig.19 GeV αs (MZ2 ) = 0. which justifies the choice for the cms energies in the figures presented below. and 6. (14) √ The running of the quark masses and the coupling constant to the scale s is performed with three-loop accuracy.2 GeV. that means below 5. As stated above. We perform the matching from nf = 5 to nf = 4 at mb (mb ) and from nf = 5 to nf = 6 at mt (mt ). mc ( s ). some values for αs (s). . and mt ( s ) are displayed in Table 1. mb (mb ) = 4. and top masses are defined for nf = 4. The massless approximation r0 in three different energy ranges.5 GeV for bottom. we will adopt the following input data: MZ = 91. Since we are working in the MS scheme. the full (t. relevant for (a) charm. bottom. mt (mt ) = 165 GeV. / Nuclear Physics B 586 (2000) 56–72 63 order to recall that. 1. mb ( s ). the approximation is expected to become unreliable in the threshold region. √ √ √ For illustration. 12–12. If not stated otherwise. 5. respectively. and on the other hand. Chetyrkin et al.

2nd and 3rd row. and t production (1st. 2. Left column: quadratic.64 K. arising from diagrams where the external current couples to massless quarks only.G. Chetyrkin et al. respectively). b. / Nuclear Physics B 586 (2000) 56–72 Fig. right column: quartic mass corrections. . Mass corrections to the non-singlet contribution of rq for c. starting to be non-zero in O(αs3 ). starting to be non-zero in O(αs2 ).

This is reflected in the stability of the result under variation of the renormalization scale µ by a factor between 1/2 and 2 as displayed in Fig. b. arising from diagrams where the external current couples directly to the massive quark. 5. Terms of order m4 αs3 and m4 αs2 are of comparable magnitude. As discussed already in earlier papers. (c). lower row: vector and axial currents for t quark. (e)) arise for the first time in order αs3 . this term is typically around 10−4 and thus irrelevant for all practical purposes. 1). (c). 6 (a). 3. / Nuclear Physics B 586 (2000) 56–72 65 Fig. whence the prediction for the m4 . In Figs. 2 (a).K. including successively higher orders in αs . Upper row: vector current for c and b quarks. The axial contribution is presented for the top quark only. and (f) contribute in O(αs2 ) and higher. 1) and for the m2 and m4 terms. and (e). Chetyrkin et al.G. t . 2 shows the effects of the diagrams with light quarks coupling to the external current. 1–4 we display separately the contributions for the massless case (Fig. Quadratic mass corrections (∝ m2 ) to the non-singlet contribution of rQ for Q = c. not only for the low energy case but even at the highest energy. The m2 terms in rq (Figs. While Fig. Figs. The quartic terms in rq which are displayed in Figs. However. the size of the higher order corrections decreases quickly with increasing order in αs . 2 (b). 5 to 8. As a consequence this prediction exhibits a strong µ dependence as displayed in Figs. (d). 3 and 4 correspond to the case where the massive quark couples directly to the external current. as far as the massless approximation is concerned (Fig. The variation of the prediction with the renormalization scale µ is shown in Figs.

b.001628 0.004581 0. t .05482 0.063 1. / Nuclear Physics B 586 (2000) 56–72 Fig. arising from diagrams where the external current couples directly to the massive quark. lower row: vector and axial currents for t quark.66 K. Table 2 √ (v) rQ for s = 10.G.000519 0.00002016 1.063 . (d).00002969 0. Upper row: vector currents for c and b quarks.001898 0. Chetyrkin et al. No reduction of this variation is observed when moving from the αs2 to the αs3 calculation. and (e).058 1. the prediction for the total cross section is not seriously affected by this instability since these terms are again of order 10−4 only and thus far below the foreseeable experimental precision. 6 (b). Nevertheless. Quartic mass corrections (∝ m4 ) to the non-singlet contribution of rQ for Q = c. This is reflected in the strong µ dependence of the result shown in Figs.0001477 0.5 GeV m0 m2 m4 αs0 αs1 αs2 αs3 Σ 1 0 −0.00001245 −0. 4.003264 −0. terms in rq must be considered as uncertain within a factor two.

4 at 6 GeV varies by ±0. b. 7). their instability does not affect the stability of the overall prediction for R resulting from the sum of the different terms. 4). The dominant quartic mass terms are obviously expected from rQ . the higher orders are small compared to the m4 Born terms and.4 the variation is negligible. Accepting as an estimate of the uncertainty √ √ of rQ.G.0016. 3 for c. For rt. However. This welcome behavior is reflected by the improved stability of higher orders under variations of µ (Fig. Depending on the choice of µ. even more important. / Nuclear Physics B 586 (2000) 56–72 67 Fig. the axial piece is shown for t quarks only. the relative size of the three corrections varies drastically. 5. The order αs -. αs2 -. The bulk of the result is given by the Born approximation. For comparison we also discuss the quadratic terms which are known since long and are shown in Fig. with increasing order they do not decrease but remain roughly comparable in magnitude. rb. the part where the massive quark is coupled to the external current. The abscissa ranges from µ = s/2 to µ = 2 s. and t quarks. √ √ and (c) top quarks. The behavior of the quartic terms which are the theme of this paper is more problematic (Fig. .K. 8. (b) bottom.4 its variation with µ between s/2 and 2 s. Chetyrkin et al. and αs3 -corrections are significantly smaller than the leading terms.4 at 14 GeV by ±0. s is fixed to values relevant for the production of (a) charm.0005. Variation of r0 with µ. 3. Nevertheless. This is reflected in the strong µ dependence displayed in Fig. The vector current induced piece is displayed in Fig. rc. Terms of increasing order in αs decrease in magnitude and apparent convergence is observed.

in particular of their QCD corrections. and top production.2 (left column) and rq. Variation of rq. These considerations demonstrate that a prediction for R has been obtained which is valid in order αs3 and includes mass terms in an expansion up to order m4 . From a pragmatic point of view the smallness of the m4 terms.G. respectively. one may also attempt to arrive at a more stable result for the quartic terms by replacing the running mass by the . The first. / Nuclear Physics B 586 (2000) 56–72 Fig. Chetyrkin et al.4 (right column) with µ.68 K. 6. and third row correspond to charm. second. bottom. allows to ignore their apparent instability. Nevertheless.

(14).G. (b) bottom production for vector case. and the resulting expression for R is reexpanded in αs up to the third power. second row: (c) vector and (d) axial-vector case for top production.8 GeV. β(a) and γm (a) are the renormalization group functions governing the running of the strong coupling constant and the quark mass: µ2 d a = aβ(a).20 × 10−2 with a √ central value of −0. dµ2 µ2 d m = mγm (a). invariant one (b m) through Z γm (a) .18 × 10−2 at µ = s. 9. For the scale invariant mass we assume the values bb = 15.K.32]. Variation of rQ.3 GeV. m b for µ = m at three-loop order and using the numerical values of Eq. First row: (a) charm. m w.10 × 10−2 and −0. √ The variation of the O(αs3 ) prediction with µ is reduced. dµ2 (16) Their perturbative expansion is known up to O(αs4 ) both for β(a) [30] and γm (a) [31. 7. m(µ) = m b exp da aβ(a) (15) where a ≡ αs (µ2 )/π . Chetyrkin et al.13 × 10−2 to .r.2 with µ. compared to a range from −0. / Nuclear Physics B 586 (2000) 56–72 69 (v/a) Fig. 8 varies between −0. The (v) result for the µ dependence of rQ.4 in this approach is shown in Fig. for s = 6 GeV the O(αs3 ) prediction in Fig. The integral in (15) is solved perturbatively. and m bt = 1076 GeV which corresponds to solving (15) m bc = 2. For example.t.

Variation of rQ. −0. After expansion in m2 /s the quartic terms. Summary The total cross section for the production of massive quarks in electron–positron annihilation can be predicted in perturbative QCD.e. 7. which gives additional confidence in the reliability of these numbers. an improved understanding of the origin of these large corrections would be highly desirable. bottom and top quark production were presented. Conventions as in Fig. those proportional to m4 /s 2 . Chetyrkin et al. . Obviously. at the same time. For µ = s the results within the two approaches are quite close.70 K. were calculated up to order αs3 for vector and axial current induced rates. 9. 8. At the same time one observes a reduction of the higher order terms compared to the Born approximation and the O(αs ) √ prediction.20×10−2 with a central value of −0. Adopting instead of the MS scheme a framework where the running b.19×10−2 in Fig. the relative size of the large order terms decreases. the stability of the prediction is improved mass m(µ) is replaced by the invariant mass m and.. / Nuclear Physics B 586 (2000) 56–72 (v/a) Fig. the predictions exhibit a sizeable dependence on the renormalization scale.G. Predictions relevant for charm. The αs3 corrections were shown to be comparable to terms of order αs and αs2 . As a consequence. i.4 with µ.

Tkachov. 9. R. using the invariant mass m b. Acknowledgements This work was supported by DFG-Forschergruppe “Quantenfeldtheorie. B 85 (1979) 277. Dine. The results of this paper are available in M ATHEMATICA format at http://wwwttp.physik. . J.L. Chetyrkin. F. 43 (1979) 668. Phys. / Nuclear Physics B 586 (2000) 56–72 71 (v/a) Fig. [2] M. Combining these results with the massless prediction and the quadratic mass terms we have demonstrated that the cross section for massive quark production at electron–positron colliders is under control in order αs3 from the high energy region down to fairly low energies. The conventions are the same as in Figs. Phys. Chetyrkin et al. A. Computeralgebra und Monte-Carlo-Simulation”. Rev. 7 and 8. Lett.de/Progdata/ttp00/ttp00-09/.G. References [1] K.G.uni-karlsruhe.V. Variation of rQ.K. Lett.4 with µ. Kataev.H. Sapirstein. acknowledges support by Landesgraduiertenförderung at the University of Karlsruhe and by Deutsche Forschungsgemeinschaft.

493. B 308 (1993) 127. J. J. M. Kwiatkowski. Lett.G.A. Chetyrkin. K.G.R. Lett. K. Phys. Phys. T. 277 (1997) 189. 66 (1991) 560. 1997. Chetyrkin. Phys. S. Lett. Basel.H.G.G.M.L. Phys. Lett. A 10 (1995) 1485. S. A.A. Phys. Phys. van Ritbergen. Fiz. Kataev. A. V. 15. Lett. . S. Nucl.G. Mod. / Nuclear Physics B 586 (2000) 56–72 W. Nucl. T. J. Rev. T. Phys. Larin. Chetyrkin. Kühn. B 505 (1997) 40. Lett. G. B 503 (1997) 339. Smirnov. R. Teubner. Sabry. B 406 (1997) 102. B 404 (1997) 153.V. K.H. Vermaseren. B 432 (1994) 337. 1991. Phys. B 404 (1997) 161.J. J. Heavy Flavours II. J. Lett. Kühn. K.G. Kühn. Chetyrkin. K. Kühn. S. J. Nucl. Kühn. Kniehl. Samuel.A. 47 (1988) 818.H. Phys.A. Generalis. J.H. A.H. Nucl. Renormalization and Asymptotic Expansion. J.H. B 320 (1994) 159. Vermaseren. K.H. Vermaseren. J. Surguladze. B 248 (1990) 359. Phys. Nuovo Cimento A 92 (1986) 119. Chetyrkin.G. OUT-4102-12. M.A. Nucl. S. Källen. R.G. Steinhauser. S. 1984. B 391 (1997) 402. K. Zerwas. Videnk. Chetyrkin. Kwiatkowski. Phys. Phys. Kühn. B 61 (1973) 455. Smirnov. Gorishny. Harlander. Phys. ’t Hooft.72 [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] K. Chetyrkin. Phys. Rev. O. B 305 (1993) 285. Sov.L. J. Chetyrkin et al. A. D. Phys. B. Chetyrkin. Lett. in preparation. Kühn. Harlander.A. Lett. Phys.H. Buras. Chetyrkin.G. T. Larin. Chetyrkin. S. K.A. Phys. Rev. B 329 (1990) 547. Phys. P. J. Phys. Spiridonov. Yad. Chetyrkin.G.A. p. B 259 (1991) 144. B 400 (1997) 379. Kühn. B 327 (1994) 114. Chetyrkin. Phys.J. Chetyrkin. J. Rev. Nucl. K.J.G. Kühn. Tarasov. Vol. Kühn. Celmaster. Larin. Broadhurst. S. B 224 (1990) 229. Lett. K. Chetyrkin.A.G. K. Lett.). V. Phys. Steinhauser. B 342 (1995) 356. Lett.M.-Fys. K. Mat. Advanced Series on Directions in High Energy Physics.G. J.G. 44 (1980) 560. Lett. Phys. Lett. K. Phys. Lett. J. C 2 (1998) 151. B 482 (1996) 213. Kühn.A. Lett. Phys. A. Kühn. Lett.H.A. Nucl.C. Harlander. Kataev. R.H. van Ritbergen. B 371 (1996) 93. Larin. Chetyrkin. K. J.G. Phys. 66 (1991) 2416.G. Birkhäuser. K. M.H. M. Larin. B 319 (1993) 307. Selsk. G. L. D 56 (1997) 3011.G. Rep. 29 (17) (1955). Open University preprint. Phys. Gorishny. Phys. Lett. Steinhauser.M. Phys. World Scientific. Lett. J. K. J. Eur.G. 47 (1988) 522. Phys.H. Medd.P. Dan. M. Gonsalves. in: A.A. V. Lindner (Eds. van Ritbergen. R.

hep.e. the term MSSM is a misnomer. Philadelphia.10. E-mail address: erler@langacker. the Minimal Supersymmetric Standard Model (MSSM) provides no explanation to the µ-problem [3]. However.10.60. Moreover.60. Proton decay 1. 12. PA 19104-6396. the prime motivation to consider low energy supersymmetry (SUSY) (for reviews see [1. all fields are protected by chirality and supersymmetry from acquiring high scale masses. accepted 3 July 2000 Abstract I discuss supersymmetric extensions of the Standard Model containing an extra U (1)0 gauge symmetry which provide a solution to the µ-problem and at the same time protect the proton from decaying via dimension 4 operators. i. to why the scales of SUSY breaking and of the supersymmetric bilinear Higgs µ-term. University of Pennsylvania. While there are several studies addressing subsets of these requirements.Nuclear Physics B 586 (2000) 73–91 www. All rights reserved.Dm Keywords: Supersymmetry. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. this work uncovers simultaneous solutions to all of them.2]) is the stabilization of the electroweak scale under radiative corrections. (1) are of comparable magnitude. Erler). I discuss the solutions to these constraints including some phenomenological implications.. In this sense. The MSSM should therefore be extended by at least some sector solving the µ-problem.Kt.Cn.upenn.nl/locate/npe Chiral models of weak scale supersymmetry Jens Erler Department of Physics and Astronomy. PII: S 0 5 5 0 .V.edu (J. especially as current precision data seem to offer a hint to the existence of its corresponding Z 0 boson.2 . It is also remarkable that there are many solutions where the U (1)0 charges of the known quarks and leptons are predicted to be identical to those E6 motivated Z 0 bosons which give the best fit to the data.Jv. It is surprising and encouraging that extending the Minimal Supersymmetric Standard Model by a simple U (1) factor solves its major drawbacks with respect to the non-supersymmetric Standard Model. 12.3 2 1 3 ( 0 0 ) 0 0 4 2 7 . USA Received 6 June 2000.  2000 Elsevier Science B. The additional requirements of anomaly cancellation and gauge coupling unification imply the existence of extra matter multiplets and that several fields participate in U (1)0 symmetry breaking simultaneously. 12. Introduction Besides its conceptual elegance. PACS: 12. µhd hu . All rights reserved.V.elsevier.

baryon number conservation by the principles of gauge invariance (and renormalizability) has been on the wish list of SUSY practitioners for almost two decades [6–8]. The predictions of the SM are generally in very good agreement with experiments provided the Higgs boson mass. Additional U (1)s have long been considered as very well motivated extensions of the MSSM (for a recent review see [9]).34]. I. whether any solutions exist. and all bilinear (mass) terms must be forbidden by gauge invariance. Moreover. so that the MSSM is likewise in agreement with observations [33. This is desirable since our confidence in the MSSM is boosted by the observation of approximate gauge coupling unification [10–19] (as predicted by the simplest and most economic GUT and string models) at a scale somewhat below the Planck scale. It has also been argued [4] that those discrete symmetries 1 which survive as remnants of some underlying string model. such as R-parity. that these requirements present a strong set of constraints. If such an extension is free of gauge and gravitational anomalies. it severely compromises the initial elegance and one prefers to save the successful features of the SM when exploring its alternatives.23]. but without consideration of gauge coupling unification. Indeed. An extra U (1)0 does not affect the renormalization group equations at the one-loop level. This is also predicted by the MSSM and many extensions (MH . Addition of 1 Continuous global symmetries are not expected to arise from string theory [5]. is imposed ad hoc. MH . In order to rigorously protect the electroweak scale from higher mass scales. unless an additional discrete symmetry. It is well known that either of the problems addressed in the previous two paragraphs can be solved by the introduction of an extra U (1)0 symmetry. Erler / Nuclear Physics B 586 (2000) 73–91 Moreover.. . might either be unsuited to forbid (at least) the dangerous dimension 4 operators. Note. A similar approach in the context of gauge mediated supersymmetry breaking. Examples for either case are reviewed in the next section using U (1) groups motivated by E6 Grand Unified Theories (GUTs). all superfields should transform non-trivially with respect to the low-energy SU(3) × SU(2) × U (1)Y × U (1)0 gauge symmetry.21]. 150 GeV) [25–32]. or arise only along some fine-tuned flat scalar field direction. I also demand that the field content of the models be completely chiral before gauge and supersymmetry breaking. Electroweak precision data imply further hints and constraints. in sharp contrast to the non-supersymmetric Standard Model (SM) where the stability of the proton can be understood entirely by gauge invariance. In this article. and it is not clear a priori. While this is a logical possibility. can be found in Refs. they are predicted in most GUTs and appear copiously in superstring theories. I address the question whether it is possible to find models in which an extra U (1)0 solves both problems simultaneously. [20. one can allow a grand desert scenario between the electroweak scale and a more fundamental high mass scale.e. The key ingredient necessary to arrive at a solution is that at least two MSSM singlet fields which are charged under the U (1)0 have to acquire vacuum expectation values (VEVs). is smaller than about 250 GeV [22.74 J. the superpartners and extra Higgs fields decouple from the precision observables over large parts of the MSSM parameter space. and it is conceivable that the quality of unification even improves at two-loop precision. the MSSM predicts rapid proton decay via dimension 4 operators.

but if they are non-chiral (with respect to the SM gauge group) and approximately mass degenerate their effects are rendered small. driven mostly by the parity violation experiments in Cs [36–42].J. in the bottom-up approach of this paper I do not assume an SU(5)-GUT symmetry. Section 4 shows that there are 33 solutions (family universal with respect to the ordinary fermions) if there are two singlets acquiring VEVs in the course of U (1)0 breaking. O (1 TeV) and continue to yield results on MH consistent with SUSY [35]. My conclusions are presented in Section 6. and U [46] (or 1 . lhu ν¯ . gauge coupling unification. as an illustrative example. If they arise in complete representations of SU(5). but no extra matter beyond that. Fits to the oblique parameters. I will also assume SU(5)-type charge quantization. Five of these models involve a number of 5 + 5¯ representations and singlets of SU(5). 2 but the results in Tl [43. and 3 [47]). To avoid potential problems with fractionally charged states. E6 inspired U (1)0 models Consider. In the next section. qhu u. However. hd and hu . they do not affect gauge coupling unification at the one-loop level. and chirality. I will also briefly address the prospects to derive these models from the E8 × E8 heterotic string theory. I will show that other matter configurations can preserve unification. S. W (2) . Note. About half of the resulting models predict a sufficiently long proton lifetime. nor its scale..e. T . I review the situation for the most popular U (1)0 extensions arising from SO(10) and E6 GUTs [48]. [12. i. ¯ lhd e+ .44]. I will classify the solutions to the conditions of anomaly cancellation. In Section 3. Allowing incomplete representations of SU(5) and the weaker requirements of gauge unification and absence of fractionally charged states yield the remaining 28 solutions. 2. To reiterate. 36–40]. ¯ qhd d. that these fits suggest a Z 0 boson mass MZ 0 . 2 . This implies restrictions on extra matter fields which can contribute significantly to S and T . Some of the phenomenological implications of the type of models obtained in Section 4 are discussed in Section 5. and from the Z lineshape measurements at LEP 1 [45] also play a role. yield results consistent with zero [23] (for Higgs masses in the MSSM range). as well. the case of an E6 gauge group which is free of anomalies in all its representations. allowing the Yukawa couplings. carrying the appropriate U (1)Y hypercharges to serve as the MSSM Higgs doublets. 2 Implications of a Z 0 for Q have been emphasized in Refs. Erler / Nuclear Physics B 586 (2000) 73–91 75 an extra Z 0 boson can even improve the global fit to all data [35]. I will assume that there is an additional U (1)0 symmetry with a mechanism built in to solve the SUSY µ-problem. A fundamental 27 representation contains a pair of SU(2) doublets. describing new physics contributions to vector boson self-energies. I introduce the anomaly conditions and show that it is not possible to find chiral models if only one MSSM singlet field develops a VEV.

6– 8. Models containing an anomaly free U (1)0 symmetry designed to stabilize the proton have been discussed by various authors [4. B. Moreover. Therefore. while additional fields will be capitalized. 4 Terms of this type can also trigger radiative U (1)0 symmetry breaking in analogy to electroweak symmetry breaking in the MSSM. One possibility is to ignore the µ-problem. d¯d¯u. The U (1)ψ symmetry also forbids all baryon number. A 27 also includes an MSSM singlet superfield. and lepton number. one predicts MZ 0 . Since it is assumed that electroweak symmetry breaking is triggered by SUSY breaking.76 J. and ν¯ refers to the right-handed neutrino superfield. and any term of the form (3) can only include a U (1)0 neutral singlet field. as well. (3) in the superpotential. which has the right U (1)0 charge to allow a trilinear Yukawa term. ¯ e+ D u. lle+ . S. 4 S SD D. ¯ (4) S Upon However. lhu . linking these scales. Su. Erler / Nuclear Physics B 586 (2000) 73–91 ¯ and u¯ denote the SU(2) Here l and q are the lepton and quark doublet superfields 3 . one can avoid this problem by restricting oneself to SO(10) and identifying the U (1)0 with the U (1)χ appearing in SO(10) → SU(5) × U (1)χ . As a result.50]. The ordinary fermions and the ν¯ complete anomaly free 16 representations (avoiding the extra singlet quarks). and one has to add a pair of Higgs doublets which by anomaly cancellation are required to have equal and opposite U (1)0 charges. violating terms in the renormalizable superpotential of the MSSM. non-anomalous U (1)0 s addressing the µ-problem without reference 3 I denote the MSSM multiplets by lower case letters. (7) S quarks. e+ . excluding an elementary µ-term. (4). a primordial µ-term is allowed. U (1)0 breaking they are expected to receive TeV scale masses through terms of the form. If the scalar component of S develops a VEV triggered by SUSY breaking. which likewise forbids the terms in Eq. provided the coupling strength is comparable to the top quark Yukawa coupling [50]. d. O (1 TeV) if one wants to avoid excessive fine-tuning. If one identifies the U (1)0 with the U (1)ψ defined by E6 → SO(10) × U (1)ψ . ¯ lq d. all scales are related and the µ-problem is solved [49. This is the reversed situation compared to the U (1)ψ case.52].51. Shd hu . while there is no problem with too rapid proton decay. it breaks the U (1)0 gauge symmetry and induces an effective µ-term. qqD. both aspects of the U (1)0 solution to the µ-problem (exclusion of an elementary µ-term and generation of an effective µ-term at the TeV scale) are lost. S. D and D. dimension 4 proton decay is reintroduced through the exchange of D and D Clearly. . then hd and hu have equal U (1)0 charges. Conversely. each 27 also contains an extra pair of SU(2) singlet quarks. (5) but they also allow new B and L violating (though B–L conserving) terms. d¯D ¯ (6) Together with the B and L conserving operators. S lq D. L. singlets. ¯ ν¯ D d.

their hadronic interactions give rise to rather large number densities of these stable hadrons [56..g. and hu . and where C2 is an even (odd) integer for vector (spinor) representations of SU(2). The present work is therefore devoted to models without fractionally charged states. respectively.59]. which requires the addition of an extra pair of SU(2) doublets. and (−5/3..e. ±1 for quarks and antiquarks). and forbids the dangerous terms in Eq. yet. to the U (1)χ . Similarly. C2 C3 C3 = QY + + ∈ Z. Lowering the reheating temperature does not solve the problem either. (a. (6). By virtue of anomaly cancellation these must be non-chiral and one would therefore need a mechanism preventing them from receiving high scale tree-level Q+ 5 I am indebted to Michael Plümacher for discussions on this point. and the U (1)Y (up to normalization)   ν¯ − e   u d  N E− +1/3   + E S N ν¯ e+ +5/3 +1/3 +a˜ −a˜ +2b˜ +b˜ +b˜ u¯ d¯ +1/3 −1 +a˜ −a˜ +b˜ −1 +2/3 +a˜ −b˜ D S D −2/3 +2/3 −2b˜ −b˜ −2/3 −a˜ −2b˜ S 0 +3b˜ to proton decay are discussed in Ref. since successful baryogenesis probably requires a reheating temperature after inflation at least of the order of the critical temperature of the electroweak phase transition of about 100 GeV (for a recent review and references see [60]). C3 is the triality class of SU(3) (e. it implies the existence of stable fractionally charged baryons with TeV scale masses. −4/3). 5 If the reheating temperature after inflation is of the order of the mass of these exotic particles or above. ˜ b) the U (1)ψ . and lead to serious cosmological problems. 0) correspond. Then a significant number density of TeV scale particles can still be produced by the high-energy tail of thermal particle distributions. orders of magnitude above experimental bounds [58. or during reheating itself [61]. (8) 3 2 3 where Q is electric charge. respectively. This does not alter the SM anomaly conditions. Erler / Nuclear Physics B 586 (2000) 73–91 77 Table 1 U (1)0 charges of a 27 representation for an arbitrary E6 boson [35]. Such states have not been observed.J. hd . All of the E6 inspired models discussed above have another unwanted feature in the context of gauge coupling unification. QY is hypercharge. The upper half of the table with b˜ = 0 corresponds to a family in SO(10) GUT.54]. The doublets on the left-hand side are the fields l. 0). However. The special cases ˜ = (0. An interesting solution to both problems S quarks relative to the is to flip the signs of the hypercharge assignments of the D and D E6 case [55]. i. SU(5) charge quantization is imposed. q. . (0. [53.57].

the extra doublets would then be massless. Anomaly cancellation in the presence of an extra U (1)0 Consider the U (1)0 charge assignment in Table 2. as well. the U (1)Y transforming in the adjoint representations of SU(2) and SU(3). Indicated are three generations of ordinary matter fields. Shown are also examples of extra matter fields. ν¯ . Baryon allows a Dirac mass term. respectively. Ti . while assuring that all fields receive large enough masses after symmetry breaking. hd and hu . . The role of d¯ is now played by D. S and Q S are their mirror partners. S lq D is the only type of operator involving it. as these models stand. However. and U (1)Y . but initially any field respecting Eq. . and trade its role with hd . (2). W and G are neutral fields under where E + . Hence. L. but in general this is not protected to be of electroweak size. one pair of Higgs doublets. and further singlet fields 6 . 3 (9) ¯ so that D and d¯ can be removed from the spectrum. a˜ + 2b˜ + 5 = 0. This paper is therefore devoted to stabilize electroweak scale physics by the principles of gauge and supersymmetry. (9) defines therefore an alternative set of SO(10) models within E6 .. where E − are charged lepton singlets. An alternative way to remove the dangerous terms in Eq. S. the new fields have identical U (1)0 charges in the cases of multiplicities greater than one. for U (1)0 symmetry breaking. This is not a problem in the context of string models which predict the absence of fundamental mass parameters in the low energy effective Lagrangian by conformal invariance. one singlet Higgs. MZ . D d. I assume family universality for the ordinary fermions and also the exotic fields. One could couple them to an overall singlet receiving a VEV. (6) would be to choose a particular linear combination of U (1)χ . (6) decouple. These are the fields which will play a role in the models surviving the analysis in Section 4. but it seems difficult to arrange that all chargino and neutralino masses are compatible with the experimental lower mass limits as some masses would be predicted to be strictly of order the Z boson mass. and S but since effects from the ones in Eq. U and D are up-type and down-type quark singlets. in analogy with the alternative left-right model [62–64]. However. Eq. Q are quark doublets. U (1)ψ . Table 1 shows that the condition. 6 I will refer to a singlet as a right-handed neutrino. i. Eq. D. one can assume that the extra doublets develop VEVs. (9) also implies that l and hu have equal and opposite U (1)0 charges and furthermore allows ν¯ to decouple from the spectrum. Erler / Nuclear Physics B 586 (2000) 73–91 masses. but this assumption is not crucial and can be relaxed easily. L are lepton doublets. 3. which are (initially) assumed not to receive VEVs.e. l must act as a Higgs superfield. (8) is allowed. appropriate to break electroweak symmetry.78 J. Alternatively. we recover the SO(10) case (with ν¯ replaced by S) and the µ-problem which comes with it. and S U S. number is now conserved because the trilinear terms in Eq. if it admits the Yukawa coupling in Eq. (4) are still forbidden.

Three generations of ordinary fermions are assumed and generation indices suppressed. Examples are the fields W and G. and has Q0 = −s/2. and d1 = −2b QY Q0 QY Q0 −1/2 a ν¯ e+ 0 +1 −(a + c2 ) −(a + c1 ) +1/6 b u¯ d¯ −2/3 +1/3 −(b + c2 ) −(b + c1 ) hd −1/2 c1 hu +1/2 c2 D L Q U E− W −1/3 −1/2 +1/6 +2/3 −1 0 d1 d2 d3 d4 d5 −s/2 S D S L S Q S U E+ G +1/3 +1/2 −1/6 −2/3 +1 0 −(s + d1 ) −(s + d2 ) −(s + d3 ) −(s + d4 ) −(s + d5 ) −s/2 S 0 s Ti 0 ti   ν − e   u d As usual. Since the exotic matter is vector-like with respect to the SM gauge group it does not spoil the anomaly cancellation present there. Triangle diagrams with two SM (i. or SU(2).e. r = r¯ . If an irreducible representation is real. I demand. s = −(c1 + c2 ) 6= 0..J. This way their contributions to oblique parameters decouple. all ordinary charged leptons and quarks are expected to become massive through the Yukawa couplings 7 in Eq. and in this case it suffices to add a single copy. either SU(3). Table 1 corresponds to a = −1 (normalization). (5). if the exotic matter consists of k5 copies of 5 + 5¯ representations of SU(5). it is not firmly ruled out. or U (1)Y ) and one U (1)0 gauge fields as external legs yield three conditions which have to be satisfied. are assumed to get vector-like (with respect to the SM group) masses induced by trilinear couplings with S and their respective SU(5) mirror partners. Finally. At first sight one would expect that there are many solutions given the many free parameters in Table 2. (2). (10) to guarantee a solution to both aspects of the µ-problem. On the other hand. but as it will turn out there are none. the mixed SU(3)/U (1)0 anomaly condition reads. c1 + c2 = −s = −(a + 3b). For example. u . (k5 − 3)(c1 + c2 ) = 0. the extra fields. such as D and E. and extra exotic fields. (11) 7 While m = 0 is strongly disfavored. as for example in Eq. it can acquire a supersymmetric Majorana mass through a trilinear coupling to S. a number of singlets S and Ti . Erler / Nuclear Physics B 586 (2000) 73–91 79 Table 2 U (1)Y and U (1)0 charge assignments for the MSSM matter fields including righthanded neutrinos and the two Higgs doublets (upper half). I will comment on this case later in this section.

(11)–(13) still apply. Table 2). Family non-universal U (1)0 charges can also induce significant flavor changing neutral current effects [66]. the index k5 . and the anomaly conditions remain unchanged. preservation of gauge coupling unification enforces the differences of the one-loop β-function coefficients to be the same as in the MSSM.  +1 −3 −2 34 3 10 3 k5 − 14 c1 + c2 However. (11) is independent of the di . which are. that this conclusion depends crucially on the additional 8 I will refer to such configurations as quasi-complete SU(5) representations. for example. This results in a zero eigenvalue for the up-type quark mass matrix. (15) in Eqs. but k5 is now defined as the index of N.    2a + a 0 + 9b 0 +1 k5 − 3    (16) k5 + 1   b − b0  = 0. where N is any (in general reducible) representation of SU(5). X nr dim r Cr . The setup discussed in this section does not seem to provide solutions to the µ-problem. and so are the mixed SU(2)/U (1)0 and U (1)2Y /U (1)0 conditions. 3(a + 3b) → 2a + a 0 + 3(2b + b0 ).  10 3 k5 − 14 (c1 + c2 ) − 6(a + 3b) = 0. is also independent of the assumption of family universality. (10) fixes k5 = 3. however. (12) (13) This is solved only by the SO(10)-type relation. c1 = −c2 . A more interesting case [65] can be constructed from the one in the previous paragraph ¯ and conversely by replacing −(b + c2 ) by −(b0 + c2 ) for the quark singlets u¯ and c. If one allows for this possibility. nothing new. Eqs. i. It should be stressed. or that a non-vanishing mu can be generated radiatively. (k5 + 1)(c1 + c2 ) + 3(a + 3b) = 0. Eqs.80 J. the determinant of this matrix is non-vanishing (= −4) independently of k5 . (14) k5 = 2 dim SU(5) r where the sum is over the irreducible representations. 8 This leaves only one adjustable parameter. However. and for equal and opposite Higgs charges.e. nonperturbatively. If. or by some other mechanism. and would either require mu = 0. Eq. c1 +c2 = 0. nr are the multiplicities (nr = 1/2 for Majorana types) and Cr is the second-order Casimir invariant of representation r. for t¯. The conclusion. (12) and (13). (11)–(13) are replaced by. r ∈ N. the U (1)0 charges of the third generation were different from the two lighter ones and associated with parameters a 0 and b0 (cf. One can also allow extra matter fields in incomplete representations of SU(5). Erler / Nuclear Physics B 586 (2000) 73–91 and together with Eq. This conclusion cannot be avoided even if we generalize to arbitrary extra matter fields of the form N + S N. respectively. . then this would merely result in the replacement. a = −3b. and there is only the trivial solution implying b = b 0 ..

Two Higgs singlet solutions Suppose now that besides S another MSSM singlet field. which are solved by. Suppose the exotic matter consists of the fields (Q + S + (U + U S).r. +2 10 3 (k5 − k1 ) − 14 10 3 k1 t P where k3 is the SU(3) index of M. the model in Ref. 1 1 5 k1 − k2 − k3 ≡ k0 ∈ Z. Extra singlets. M + M N. k2 . k3 = (k5 − 3)(k0 + k2 − k3 ). For example. (13) becomes equivalent to Q) Eq. T = T0 . (17) is nonvanishing. 4(c1 + c2 ) = −3(a + 3b). This model is chiral but the gauge couplings do not unify. The anomaly conditions in Eqs. provided the U (1)0 charge assignments S are changed 9 from di and −(s + di ) to ei and −(t + ei ). S ⊂N+S develops a VEV. implying c1 + c2 = 0. of the fields in M and M respectively. Erler / Nuclear Physics B 586 (2000) 73–91 81 assumptions of gauge coupling unification and chirality. but fails to produce unification. this explicitly violates the assumption of universality w. A subset of the exotic matter fields. this model mirrors the situation in Ref. If the determinant of the matrix in Eq. can be added to satisfy the cubic (pure U (1)0 ) and trace (mixed gravitational/U (1)0) anomaly conditions. τ≡ . t (a + 3b) . (19) Using the definitions. Another example can be constructed with the assignments in Table 2. Eq. with Q0 = t = t0 . and Eq. Ti . s s the solutions can be classified as follows: σ≡ (20) 9 In general. [55] is chiral. (11) is then satisfied (k5 = 3).    k3 3(a + 3b) 0 k5 − k3 − 3    (17) k5 − k2 + 1 k2   s  −1  = 0. The quadratic anomaly condition from triangle graphs with one hypercharge and two U (1)0 gauge bosons in the external legs depends on d3 and d4 and can easily be solved. [55] except that it has no fractionally charged baryons. Unification can be arranged S representation of SU(5). which implies the bound |k0 | 6 5k5/6. Thus. 0 6 kj 6 k5 for 1 6 j 6 3. The SU(5) charge quantization condition (8) implies. Similarly. and k1 = 6/5 TrM Q2Y . that k3 . 4. there is only the unwanted solution. . spoiling the chirality of the model. (11)–(13) now change to. k2 is the SU(2) index of M. the exotic fields which was made in Section 3. and k0 are integers. if unification is enforced by adding an extra pair of doublets it turns non-chiral. and that by definition. but it would have by adding an E ± pair to complete a 10 + 10 to carry equal and opposite U (1)0 charges.t. (18) 6 2 3 Notice.J. alternatively. The matrix becomes singular for. (12). k3 dim SU(3) = 2 r mr dim r Cr . could then acquire TeV scale masses through trilinear couplings to T .

k0 ) = (2. Eq. there are an infinite number of non-chiral solutions. (21) and (26). and gauge coupling unification. and therefore k0 = k1 = k2 = σ = τ = 0 independently of k5 . 1/5. k1 . σ = 5/3 − k2 /6 and τ = 1/2. 1) cannot be made consistent with the SU(5) quantization condition. For comparison.. 3σ = 3 + k0 /(k0 + k2 ) and τ = 1 − 1/(k0 + k2 ). Such solutions exist for k5 = 3 (e. and kj 6 k5 requires k3 = k5 − 3 and k0 + k2 = k5 − 2. In addition. (19) implies that k3 is an integer multiple of k5 − 3.82 J. Thus. but only 11 chiral ones. There are three solutions. in which case they are defined in Eqs. the non-chiral three generation E6 solution with an extra neutral (t = 0) Higgs singlet giving mass to an extra vector-like pair of doublets (k2 = 1) is also shown. This is the only possibility for k5 = 1. S representation these models are also non-chiral since the fields within the M + M are not tied to U (1)0 symmetry breaking. It is assumed that two fields participate in U (1)0 gauge symmetry breaking. k2 .e. which will be rejected because it corresponds essentially to the non-chiral situation in Section 3. k2 . All of them predict τ = t/s > 0. 19/5. but the case (k3 . 3.. i. −1) cannot be made consistent with the SU(5) quantization condition. k3 > k2 . 2(L + L) by the combination W + E ± . the three generation E6 model with an extra pair of doublets) and k5 = 4. Table 4 below). k0 ) = (2. 1. k0 + k2 = 4. implying that there are no D-flat Table 3 Classification of solutions to the conditions of anomaly cancellation. There are eight solutions of this type. respectively. which are collected in Table 3. SU(5) charge quantization. as well. • k5 = 2 implies k0 +k2 = 0. One has 3σ = (k5 − k2 + 1) and τ = 0. There are two solutions. The column before these specifies the fields receiving masses S fields can be replaced through trilinear couplings to T (cf. chirality. but they will be discarded. k1 . but the case (k3 . • k5 = 4 implies k3 = 2. Erler / Nuclear Physics B 586 (2000) 73–91 • The case k3 = k0 + k2 = 0 implies 5k1 = −3k2 .g. ω1 and ω3 are shown for 5 + 5¯ representations only. 3σ = 3 +k2/k3 and τ = 1 +1/k3 . • k5 = 3 implies k3 = 0. • If k5 > 5. where applicable Solution k5 k3 k2 k1 k0 σ τ I II III IV V 2 2 3 3 4 1 2 0 0 2 0 0 2 3 4 2/5 4/5 6/5 9/5 16/5 0 0 0 0 0 1 1 1 1 1 2 3/2 1/2 2/3 1/2 E6 3 0 1 3/5 0 1 0 VI VII VIII IX X XI 3 3 3 3 3 3 0 0 0 0 0 0 3 1 2 3 0 1 3/5 9/5 12/5 3 12/5 3 −1 1 1 1 2 2 5/6 7/6 10/9 13/12 4/3 11/9 1/2 1/2 2/3 3/4 1/2 2/3 Yukawa couplings to T ω1 ω3max S (D + D) S 2(D + D) 2(L + S L) 3(L + S L) S all but 2(D + D) 0 1/2 3/2 4/3 3/2 −3/2 23/16 3/2 31/27 9/16 2 2 (L + S L) W + (L + S L) S + E± (L + L) S + E± 2(L + L) S + E± 3(L + L) 2E ± S (L + L) + 2E ± .

Tr Q0 = 0.J. VI. unless noted otherwise. In the remainder of this section and in the next section I will focus on exotic matter in complete 5 + 5¯ representations of SU(5). but solution V requires S representation of SU(5) (k5 = 3) cannot S quarks. IV. II I. There are other configurations with k5 = 4. (22) which is familiar from the E6 assignment in Table 1. while k5 = 5 would yield non-perturbative unification (for a discussion see [67]). the 10 + 10 at least two pairs of D + D be used for any of the solutions in Table 3 and has been omitted Representation k5 S D+D S L+L S Q+Q S U +U E± W G Solution(s) 1 2 3 2 2 2 1 2 2 – – 2 – – – 1 – – – 1 – 1 1 – – – – I I. IV 10 11 12 13 14 15 16 4 4 4 4 4 4 4 2 2 3 3 4 4 4 – 1 – 2 – 2 4 – 1 – – – – – 2 – 1 1 – – – – 2 1 – 2 1 – 2 – 2 1 2 1 – – – – – – – – V V V V V V V scalar field directions. which is fixed by the trace condition. . II 4 5 6 7 8 9 3 3 3 3 3 3 – – 1 2 3 3 1 3 – 1 1 3 – – 1 – – – – – – 1 – – 2 1 2 – 1 – 1 – – 1 1 – 1 1 – – – – III. For 5 + 5¯ representations one has k0 = 0. as 1X ti = 5k5 + (τ − 1)(3k3 + 2k2 − 1) − 12. IV. s = −(c1 + c2 ) = (a + 3b). and also to the other solutions with k0 = 0. each solution in Table 3 can be realized with various exotic matter contents. which guarantees that the gauge couplings remain perturbative up to the unification scale. (10). These fields can be combined into quasi-complete SU(5) representations shown in Table 4. Thus. VI–XI III. and (18) imply the relation. Similarly. (21) ω1 ≡ s i=1 Interestingly. ω1 is completely determined within each model. In general. IV. (17). In this case one can write the sum over the other singlet charges. all of them predict k5 6 4. and then Eqs. Erler / Nuclear Physics B 586 (2000) 73–91 83 Table 4 Quasi-complete SU(5) representations (preserving gauge coupling unification) which can be used to realize the solutions in Table 3. when 3 right-handed neutrinos are included. Furthermore. VII III. most of the phenomenological studies of E6 inspired Z 0 bosons apply here. Observe that the configurations of ki in the table restrict the possible matter content to the fields displayed in Table 2. VII–IX X VI III.

a + 3b = c1 + c2 = s = t = 0. τ −1 (6τ + 3)(a − b) + (7τ + 2)(c1 − c2 ) 2[(k3 − k5 )d1 + (k2 − k5 )d2 ] = τ −1   + 2k5 − (k3 + k2 ) s. these solutions could be interesting if they 10 Tr Q Q0 2 also vanishes if s = 0. While they would yield non-chiral pairs of Higgs doublets. 4 (26) The inequality (26) is saturated when. U (1)χ˜ . one can construct another family of special solutions. Y . Tr QY Q0 = 0. Tr Q0 3 = 0.84 J. but the E6 gauge symmetry and Yukawa coupling relations are usually explicitly broken to avoid serious conflicts with observation (such as the proton lifetime). reads. + 3(k5 − k3 ) s s s s and one can show that. which can be relaxed at a later stage. Combined with the anomaly condition quadratic in the U (1)0 charges 10 . the cubic condition. and using the trivial solution to Eq. Recall that U (1)0 subgroups from E6 are frequently discussed because they represent manifestly anomaly free examples (and because they arise in popular GUT and string models). 2(k3 e1 + k2 e2 ) = 9 (23) (24) Similarly. 2 (27) which yields a class of Z 0 bosons which couple like the Zψ to the ordinary fermions and the Higgs doublets. Tr QY Q0 2 = 0. ω3 6   1 5(k5 − 1) + τ 3 − 1 (3k3 + 2k2) − τ 3 . where the MSSM fields have charges as in the U (1)χ case. (a − b) + (c1 − c2 ) − τ (k3 + k2 )s. Relaxing the orthogonality condition. I impose the orthogonality condition. 4 s c1 = c2 = d 1 = d 2 = − . Tr QY Q0 = 0. ω3 ≡  1 X 3 ti = 5k5 + τ 3 − 1 (3k3 + 2k2 − 1) (25) 3 s i=1       c1 − c2 2 a−b 2 3 +4 − 3+9 4 s s      e1 e1 e2 e2 + τ + 2τ k2 +τ + 3 3τ k3 s s s s     d1 d1 d2 d2 + 1 + 2(k5 − k2 ) +1 . Here we recover U (1)0 ≡ U (1)ψ˜ symmetries which are very similar to the U (1)ψ from a very different bottom-up approach. s a=b= . This is encouraging as the recent analysis [35] of precision data revealed an excellent fit to the Zψ assuming some amount of mixing with hypercharge. Erler / Nuclear Physics B 586 (2000) 73–91 To simplify the algebra. I find. (17). 2 τ e1 = e2 = − s.

e1 + 2/5 ∈ 2Z. But the D quarks have the sign of their U (1)ψ˜ charges reversed and the T field (Qχ˜ = 0) has half-integer instead of integer U (1)ψ˜ charge. (29) is respected by all fields. Examples are heterotic string constructions where the gauge group factors considered are realized at the same Kac–Moody level. P 2 = 5/8 for a 16 of SO(10). Assuming the √ normalization in Eq. The 10 of SU(5) is relevant here. P 2 = 2/3 for a 27 of E6 . 5 5 Qψ − Qχ ∈ Z. Qχ = 1 for ν¯ (a = −3/5). On the other hand in the level 2 models or Ref. gχ = 5/8 g and gψ = 2/3 g. Eqs. Notice. (29) one finds the well-known result. With the exception of two fields. 11 In Kac–Moody level 1 models SU(5) charge quantization can never be obtained unless a complete SU(5) GUT group survives [68]. (29). (29) one finds in a similar way. [69] the third Eq. Qψ = 1 for S (s = 1). It is not clear at present 11 whether E8 charge quantization. (28) implies for the U (1)χ˜ of solution II. one has P 2 = 3/5 for a 10 of SU(5). (29). (28) where solutions with c1 = a correspond to hypercharge and c1 = −2a/3 to the U (1)χ˜ . k. Referring again to the √ √ normalization fixed by Eq.J. and P 2 = 3/4 for a 56 of E7 . 2d2 + 3e1 + 2/5 = 0. In units in which an E8 root vector P has length P 2 = 1. and only solution I would survive. 4 4 Qφ − Qψ ∈ Z. that this argument needs no reference to a complete GUT group like SU(5). There are classes of string models where gauge coupling unification is predicted without the appearance of a full GUT group in the zero-slope limit. Note. g. The quadratic anomaly condition now reduces to (k3 − k5 )d1 + (k2 − k5 )d2 − τ (k3 e1 + k2 e2 ) + 2(a + c1 ) = 0. The E8 embedding also allows one to fix the overall normalization (coupling strength) of the U (1)0 . 3 (29) where the last relation refers to the U (1)φ defined by E7 → E6 × U (1)φ . The E8 × E8 heterotic string is particularly suited here. gY = 3/5 g. that there are other solutions to the anomaly constraints with s = 0. (29) is also satisfied.e. The reference E8 symmetry implies the relevant matching (unification) conditions. i. . the second Eq. Erler / Nuclear Physics B 586 (2000) 73–91 85 enter linear combinations with the U (1)ψ˜ or if the corresponding Zχ˜ coexists with the Zψ˜ . or when combined with the first Eq. in terms of one of the non-Abelian gauge couplings. The latter are given by 6 Qχ − QY ∈ Z. (28) can mix with the chiral U (1)0 s. and Qφ = 1 for one of the E6 singlets within the 56 of E7 .. The normalizations are chosen such that QY = 1 for e+ (a = −1/2). gY . are predicted in string models with gauge coupling unification and SU(5) charge quantization. but only those satisfying Eq. because it offers the prerequisite E6 subgroups. as well as quantization conditions for its U (1) subgroups. If so one would have to impose the condition τ ∈ Z. because it contains the SU(3) × SU(2) singlet with unit charge (the e+ ) and can therefore serve to properly normalize the hypercharge gauge coupling. but it applies only to string constructions in which the embedding into E8 is still traceable. The encounter of the U (1)ψ˜ and U (1)χ˜ groups is also encouraging from a top-down perspective. or open string models with the group factors arising from the same D-brane. Eq.

respectively. −(1/2)9. but with one copy of D + D charges of the other two copies altered. Erler / Nuclear Physics B 586 (2000) 73–91 The reference E8 root vector also allows to set lower limits on the possible Kac–Moody levels. .e. the SU(3) × SU(2) × U (1)Y one-loop β-function coefficients are β3 = k5 − 3. the theorem about fractionally charged states in level k = 1 models [68] implies the same conclusion. 3 In Eq. Now the fields with ti = ±s/4 (and some others) can form bilinear Dirac mass terms. are summarized in Table 5. For comparison. the singlet T of solution II has Qψ˜ = 3s/2. (1/4)9 ]. SU(5) singlets with U (1)0 charges. There are no singlets with charges −1 or −3/2 in this choice. The trace and cubic anomaly conditions.      β10 a−b 2 c1 − c2 2 2 = ω + 5k + τ − 1 (3k + 2k + 1) + 6 + 9 + 7 (30) 2 5 3 2 s2 s s      e1 e1 e2 e2 + τ + 2k2 +τ + 2 3k3 s s s s     d1 d1 d2 d2 + 1 + 2(k5 − k2 ) +1 . The quantum numbers for this case. (27) are satisfied. For example. s 2 = 2/3. −(3/4)4. I. resulting in a neutrino mass matrix with three zero eigenvalues. or in short-hand notation. (1/2)6 . S and T . 3/4. ω1 = 1/2 and ω3 = ω3max = 23/16. there are unacceptably large neutrino masses. β2 = k5 + 1. one predicts a Dirac mass matrix with three electroweak scale eigenvalues. in which case one finds. and t2i+3 = s/4. a 2 = 3/20).. t3 = 3s/4. Interestingly. where 1 6 i 6 9. k.86 J.. cannot acquire high . I now focus on the Zψ˜ of solution II which is reminiscent of an E6 S quarks removed and the U (1)0 model with three families of 27. ti ∼ [12 . For concreteness. −(1/4)4]. can be satisfied by adding. t1 = t2 = s. scale masses. n3 − 4(n − 1)3 + 6(n − 2)3 − 4(n − 3)3 + (n − 4)3 = 0.e. assuring that the crucial fields. 5. for the U (1)ψ˜ . A clean way to cure this problem is to add the set of singlets 12 . Phenomenological aspects The one-loop β-function coefficient for the U (1)0 is given by. while k > 2 would be allowed by this consideration. Z (II) ψ˜ There are a total of 12 fields with quantum numbers like the right-handed neutrinos. unless the Yukawa couplings are chosen zero or tiny. for example.   1 (31) β10 = 5(k5 + 2) + τ 2 − 1 (3k3 + 2k2 ) + 2 τ 2 + ω2 . which correspond predominantly to the left-handed neutrinos provided the bilinear masses are much larger than the Yukawa mass 12 By virtue of the identity. + 3(k5 − k3 ) s s s s P −2 2 0 where ω2 = s t =i ti . i. such a configuration does not alter ω1 or ω3 . and β1 = k5 + 33/5 (using SU(5) normalization. ti ∼ 3[1. which implies PT2 = 3/2 > 1 and means that T cannot be obtained in the massless sector if k = 1. t2i+2 = −s/2. Since these fields are protected from acquiring high scale Majorana masses. (31) I have chosen the E6 inspired normalization. β1 is minimized when Eqs.

As is well known. and the three ν¯ ).5% in Γinv . 14. (1/2)4 . Clearly. where the fields with Q0 = −5s/4 allow trilinear couplings with S and the ν¯ . such that the ν¯ are U (1)0 neutral and allowed to receive high scale Majorana masses. The U (1)0 normalization corresponds to s = 1 QY  ν − e   u d (II) ψ Q˜  (II) ψ QY Q˜ −1/2 +1/4 ν¯ e+ 0 +1 +1/4 +1/4 +1/6 +1/4 u¯ d¯ −2/3 +1/3 +1/4 +1/4 hd −1/2 −1/2 hu +1/2 −1/2 D1. in agreement with the experience that the third family Yukawa couplings are typically the largest. −(3/4)6. Erler / Nuclear Physics B 586 (2000) 73–91 87 Table 5 Charge assignment for the Zψ˜ of solution II realized with a 5 + 5¯ of SU(5).2 L T +3/4 T3 +1/4 T2i+3 terms. this also produces small but non-vanishing Majorana masses . ∼ [3/2. the LEP Collaborations [45] currently report a shortage of about 0. ti ∼ 3[−(5/4). and if the masses generated by S are an order of magnitude larger than the ones from the standard Yukawa terms.2 S −1/3 −1/2 0 0 0 −3/4 −1/2 +1 +1 −1/2 +1/3 +1/2 0 0 0 −3/4 −1/2 +3/2 +3/4 +1/4 +1 T1. 115 . ∼ [3/2. Furthermore. −(1/4)] to the ones in Table 5. (1/4)9].J. There are strong limits on the non-universality between the first two families. Thus the effect in Γinv should be expected to be dominated by the ντ sector. however. B(π→eνeµ) . these also correspond predominantly to the left-handed neutrinos. 16. Indeed. Γinv . T . GF . yielding the residual chiral set of singlets (now including S. Yet another solution to the problem of neutrino mass would be to implement the see-saw mechanism which can be achieved by choosing a linear combination of the U (1)ψ˜ and the U (1)χ˜ . (1/2)9 ]. significantly increasing the extracted Higgs boson mass from electroweak fits (the central value of which is currently below the lower search limit [23]) but potentially deteriorating the quality of the global fit to all data. the Fermi constant. (1/2)3 . such as a violation of weak charged current universality. This once again results in three zero eigenvalues. but the mixing with the massive states could induce a small amount of missing invisible Z width. extracted from µ decays could increase.2 D S1. There is another interesting solution which can be obtained by adding the fields. −(5/4)3. −(3/4)17. There can be other effects associated with this scenario.2 L1. the bilinear masses can be taken to infinity and the corresponding fields can be smoothly removed from the spectrum. −(3/4)11. This results in the chiral singlet set.2 −1/2 T2i+2 S1. driven by the B(π→µν ) well measured branching ratio.

In comparison. Therefore. S ν¯ u¯ d¯ D. Hd qqq. e+ DDD. (35) can be removed if there is an admixture of the U (1)χ˜ . However. ν¯ qqD. lqqq. ¯ ∗. Thus. +5/4 Ti SD. There one has to assume that its corresponding coupling strength is small. because at the renormalizable level the D quarks couple +1/2 ¯ S dD-term) via gauge interactions and the T D Donly (ignoring the possibility of a Ti term. both conserving “D-number”. 6. ¯ e+ u¯ u¯ d. observable proton decay is not expected.03 eV. qq d¯ ∗ . While this can be arranged. Alternatively. and phase space. and electric charge is quantized. M ∼ 2×1016 GeV. −3/4 Ti ¯ u¯ d¯d. qq D DD u¯ ∗ . the D-terms. Hu qDD. Lqqq. there is another dimension 5 F -term. Dimension 5 operators would yield the same kind of suppression as in the see-saw mechanism. the critical dimension 5 unless a large d¯D F -term in Eq. the µ-problem can be solved. and the unification scale. s/4. M 0 ≈ 800 GeV [35]. One might wonder whether this agreement not only signals the presence of a new physics scale close to the unification scale. the resulting proton lifetime can be acceptable. u¯ dD S ∗. 13 Employing the Z 0 scale. Dimension 5 operators can also lead to proton decay at rates conflicting with the experimental limits.88 J. in excellent agreement with the mass scale inferred from atmospheric masses of order MZ 0 neutrino oscillations [71]. and likewise for the F -terms. but higher order suppressions (by intermediate scales) are conceivable. but possibly even suggests the absence of further scales below that. Erler / Nuclear Physics B 586 (2000) 73–91 for the left-handed neutrinos. thus assuring proton stability at the renormalizable level. which is not the case for the scenarios in the previous two paragraphs. Conclusions I have presented a class of models with an additional non-anomalous U (1)0 in which gauge couplings unify. u¯ DD (32) are also absent by virtue of the U (1)ψ˜ . Cabibbo mixing. suggests neutrino Z 2 /M ∼ 0. choosing k3 = k5 (or more generally coupling all of the exotic quarks to singlets other than S) is an efficient strategy to achieve a realistic proton lifetime. S mixing effect can be generated. S∗ . (33) On the other hand. the F -terms. the MSSM. 13 As for baryon number violating interactions. . and taking into account suppressions from first generation Yukawa couplings. are guaranteed to be absent only when there are no MSSM singlets with or −3s/4. However. S e+ u¯ u¯ D. S LqDD. S u¯ D (34) Q0 = 5s/4. (6). the U (1)ψ˜ symmetry of solution II forbids the terms in Eq. the term (35) is less concerning. (35) ¯ present in the R-parity conserving version of which is similar to the operator e+ u¯ u¯ d. one should generally expect non-vanishing masses for the lefthanded neutrinos originating from non-renormalizable terms suppressed by powers of the high mass scale [70].

Furthermore. In this case an infinite number of solutions exist. in some of these models the U (1)0 is sufficient to avoid conflicts with experimental bounds on the proton lifetime. but is also conceivable. by having the smallest contribution to the U (1)Y β-function. Only a finite number of solutions exist since the anomaly matrix in Eq. I have further assumed that all exotic fields receive their masses through couplings to MSSM singlets acquiring VEVs upon U (1)0 breaking. If it is found that higher order or threshold effects tend to raise the prediction for αs . one can increase the unification scale. A preliminary study of this class reveals that unlike the solutions in Table 3 there are also models with D-flat scalar field directions. I assumed that two MSSM singlet fields participate in U (1)0 symmetry breaking. a large (but finite) number of which being consistent with perturbative gauge coupling unification. and 1β1 = 6/5. these choices yield SU(2) × U (1)Y unification slightly above the reduced Planck scale of 2. or choose a family plus a mirror-family S U +U S. S Both cases correspond to 1β3 = 4. Moreover.J. significantly higher than most observations and the unification scale at least one order of magnitude lower than the string scale. 1β1 = 11/5. This seems difficult but is a possible loophole which I have ignored in this work. and D + D. An implicit assumption has been made concerning the precise form of gauge coupling unification. The analysis generalizes straightforwardly if one allows three (or more) singlets to acquire VEVs. In the present work. In this context it is amusing to note that one would obtain a very good prediction for αs . Erler / Nuclear Physics B 586 (2000) 73–91 89 These models are chiral. The case 1β3 = 5 (and the same β-function differences) would yield unification in the non-perturbative domain. Within a deliberately traditional and conservative framework the analysis presented is rather general.4 × 1018 GeV. Q + Q. I have already relaxed this restriction and included quasi-complete SU(5) representations. For example. but I have not considered them in the analysis of this paper. Alternatively. respectively. Indeed. as well as unification . these scenarios may be of potential interest. The minimal choice is the set G + W + E ± . with the prediction for the strong coupling constant. It is the often made assertion that gauge coupling unification works satisfactorily in the MSSM and should not be altered in leading (one-loop) order. (17) has to be singular. gauge coupling unification works only approximately. a property which can be phenomenologically advantageous. 1β2 = 2. The prediction for αs (MZ ) is now below typical observed values but the discrepancy is in general neither much better nor much worse than in the MSSM. and of quarks. But exotic leptons participating in SU(2) × U (1)Y symmetry breaking will in general receive mass contributions of O(MZ ) which could conceivably by themselves be large enough to be in agreement with current limits. Extra matter multiplets are then usually restricted to appear exclusively in complete SU(5) representations. In any case. αs (MZ ). one could add another 5 + 5¯ of SU(5) to this set. but there are various directions which can be chosen to go beyond it. It is fair to ask whether some arbitrary (incomplete from the perspective of SU(5)) representation could improve the quality of unification. protecting a large desert between the electroweak (SUSY breaking) and unification (string) scales. 1β2 = 3. contributing 1β3 = 3. by including a set of fields which contributes stronger to the SU(3) β-function than to the one of SU(2) improves the prediction for αs (MZ ). There are also solutions simultaneously in accordance with E8 charge quantization (see Eq. However. (29)) and the proton lifetime.

C. Cape Town. Chankowski. Phys. B 543 (1999) 47. F. T. Lett. 105B (1981) 439. D. Lett. J. Phys. H. Phys. Polonsky. Lett. Nilles. R. Acknowledgement It is a pleasure to thank Paul Langacker and Michael Plümacher for many fruitful discussions. S. Having left the framework of E6 inspired U (1)0 models.G. 110 (1984) 1. M. K. D 24 (1981) 1681. Matchev. 260B (1991) 447. Matchev. Hempfling. Groom et al. Rep. Hall. Dimopoulos. Eur. Dobrescu. 40. Erler. January 1999. Quevedo. N. J.E. The charges of some of the extra fields typically differ. Phys. J. Ibáñez. Banks. P. D 52 (1995) 3081. 249B (1990) 441.H.E. WIN 99. K. Langacker. Raby. Phys. Phys. N. H. B 307 (1988) 93. H.L. Nanopoulos. Erler. D. Matchev. T. Pierce.A. B 197 (1982) 533. Zhang. P. B. Singapore. anomaly cancellation tends to provide U (1) groups which have charges to the ordinary fermions just as the U (1) subgroups of E6 . Carena. Z. K. Phys. Phys. J.T. sometimes forbidding fast proton decay mediating couplings.J. Sakai. Bagger.A. in Ref. Polonsky. 312. Langacker. H. S. Dixon. Langacker. Phys. Phys. Phys. Nucl. Fürstenau. H. L. Kane. D 36 (1987) 1385. D 44 (1991) 817.M.S. W. Lett. Langacker. Langacker. South Africa. Nucl. Luo.. Z. Cheng. P. Rev. [24]. U. in: G. Amaldi et al. Weinberg.E. Nucl. Phys. hep-ph/9904235. Department of Energy Grant EY76-02-3071. 112B (1982) 351. B 439 (1998) 301.C. Phys. World Scientific. Font. R. Phys. Perspectives in Supersymmetry. Even given these possible directions it is possible to draw some general conclusions from the kind of analysis introduced in this work. Lett. F. in: Proceedings of the 17th International Workshop on Weak Interactions and Neutrinos. Wilczek. 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] H. Rev. Implications of precision electroweak measurements for the Standard Model Higgs boson. Z 0 Physics and Supersymmetry. Phys. G. Nilles.P.J. Nucl. J. Lett. I.).T. Phys.M. Wagner. Yanagida. Erler / Nuclear Physics B 586 (2000) 73–91 very close to the string scale. D.H. Hoang. N. H. 1998. B 491 (1997) 3. Electroweak Model and Constraints on New Physics. . Haber. Rep. Nucl. J. Pierce. Ibáñez. Phys. M. B.L. This work was supported in part by U. D 47 (1993) 4028. P. Kane (Ed.-X.V. C 75 (1997) 539. B 461 (1996) 407. L. S. Ellis. A. Lett.90 J. B 439 (1995) 23. Phys. Phys. 138B (1984) 150. K. J. 228B (1989) 79. Rev. M. Rev. U. D. p. Bagger.E. p. Cvetiˇc. Pokorski. Phys. Płuciennik. Dobrescu. P. Rev. Haber. 348B (1995) 443. Hinchliffe. Quiros. 117 (1985) 75. G. A more thorough phenomenological investigation will be presented elsewhere.E. D 26 (1982) 287. L. A.. Kelley.E. Phys.P.A. Ross. Cheng. P. Nucl. Rev. L. Kim. Matchev. Amaldi. C 15 (2000) 1. Phys. S. de Boer. if these extra fields have intermediate scale masses near the geometric mean of the string and electroweak scales. Phys. M. S. C.E. Lett. hep-ph/9707451. Phys.

N. Phys. Chamseddine. D 58 (1998) 093017. Rev. P. Phys. D 36 (1987) 274. R. C 9 (1999) 343. Casalbuoni. Lett. Phys. 256B (1991) 277. Zhang. Masip. Y.C. Langacker. Eur. Rev.F. Rev. ALEPH. Plümacher. Part. Langacker. D 58 (1998) 091701. Rizzo. Bennett. DELPHI. Lewin. 65 (1990) 964. J. Riotto. Mahanthappa. Phys. Schellekens. Phys. D 61 (2000) 016006. Rev. Gatto. Int. Phys.. Lett. A 10 (1995) 4521. Rev. K. Ross. Phys. J. P. A 2 (1987) 1199. B 447 (1995) 195. R. Phys. Espinosa. Rev. S. Peskin. Nucl. the LEP Electroweak Working Group. Phys. Phys. Langacker. Rosner. E. Lett. D 36 (1987) 878. R. Dominici. Langacker. Nucl.K. M. Weiglein. J..J. G. M. Lett. L. Phys. R. H. Heinemeyer. 65 (1990) 2963. M. Hewett. Rev. Hollik. Marciano. Rev. E. P. 74 (1995) 2658. Phys. Rev. Int. W.L.E. Yamagishi. Chung. B 368 (1992) 3. Rev. Phys. Rev. R. Rev. Langacker.J. T. Weiglein.D. Wood et al. Ann. 253B (1990) 161. Rev. 237B (1990) 363. Walford. C. Phys.K. A. Lett. D. Everett. W.. Dreiner.A. Mod. P.H. J. Phys. Vetter et al.T. 447B (1999) 89. De Curtis. W. A. Nucl. 82 (1999) 2484. Phys. P. G. 351B (1995) 206. D 56 (1997) 2861. Fukuda et al. Barbieri. Takeuchi. Phys. Sci. Super-Kamiokande Collaboration.A. D 54 (1996) 3570.N. D 43 (1991) 3093. M. Phys. B 526 (1998) 53.R. P. Phys. Lett. Nucl. D 41 (1990) 2074. Roulet. Phys. Homer. D 62 (2000) 013006. P. Heinemeyer. Phys. Demir. Lett. Rev. Lett. Phys. Nucl.J. 74 (1995) 2654. M. D 58 (1998) 119905. T. Cvetiˇc. 82B (1979) 65. G. P. D. Rosner. Phys. Phys. S. J. K. A. Lett. Phys.F.J. Rev.S. Mohapatra. Phys. L3. Cvetiˇc. Phys. Lett. Phys. E. Lett.. A. Espinosa. G.. D 57 (1998) 5340. Suematsu. S.L. Langacker. Gunion et al. 440B (1998) 296. P. Erler / Nuclear Physics B 586 (2000) 73–91 [28] [29] [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] 91 M. P. OPAL. private communication. C.C. Rev. Pomarol. Lett. J. J. Edwards et al.E. Langacker. Hemmick et al. preprint CERN-EP-2000016. and the SLD Heavy Flavour and Electroweak Groups. J. Nardi. G. P. L. J. W. A Combination of Preliminary Electroweak Measurements and Constraints on the Standard Model. N. J. Phys. Phys.. Phys. The LEP Collaborations. G. M. Wolfram. 49 (1999) 35. Smith. Erler. Riotto. Phys. D. Lett. Cvetiˇc. S. 84 (2000) 5269..S. J.J. B 206 (1982) 333. Rev. Ma. Y. Trodden. Heinemeyer. Phys. 81 (1998) 1562. J. Bennett. S. Phys. Lett. T. Phys. Weiglein. Rev. Lett. W. Phys.R. Lett.A. Hagiwara. Babu et al. Muñoz-Tapia. Wieman. Lett. Hempfling. Kolb. G. Ibáñez. Erler. Phys. Rev. Hollik.L.H. Rev. Smith. J. Phys. Demir.W. Oshimo. K. Langacker. Mod. 245B (1990) 105. Aoki. D. B 475 (1996) 597. Altarelli.G. Pierce. Nucl. L. 183 (1989) 193.A. Rev. Erler. Rep. 460B (1999) 135. B 574 (2000) 623. Hollik. Nucl. Phys. Everett. 84 (2000) 212. D. R. Phys. Science 275 (1997) 1759.E. Rev. S. .R. E. D.G. J. D 60 (1999) 063504. H. A.M. Cho. M.

J. Within this minimal CP-violating framework.11]. Our analysis shows that sizeable radiative effects of CP violation in the Higgs sector of the MSSM may lead to significant modifications of previous studies for Higgs-boson searches at LEP 2. A. USA c High Energy Physics Division. We also include the potentially large two-loop non-logarithmic corrections induced by one-loop threshold effects on the top. due to the decoupling of the third-generation squarks. In addition ∗ apostolos.and bottom-quark Yukawa couplings.  2000 Elsevier Science B. the charged and neutral Higgs sectors become intimately related to one another and therefore require a unified treatment. 5640 Ellis Ave. Switzerland b Fermilab. We calculate the charged and neutral Higgs-boson masses and couplings. Pilaftsis a.E.nl/locate/npe Renormalization-group-improved effective potential for the MSSM Higgs sector with explicit CP violation M. [10. Box 500. including the two-loop logarithmic corrections that arise from QCD effects. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. 1.V. revised 26 May 2000. PII: S 0 5 5 0 . and CP violation in B-meson decays 1 is strongly suggested by recent experiments [12]. IL 60510. Chicago.d a Theory Division. CH-1211 Geneva 23.pilaftsis@cern. IL 60439. Wagner a.c.and bottom-quark Yukawa couplings.b. C. University of Chicago.3 2 1 3 ( 0 0 ) 0 0 3 5 8 ..8 ..elsevier.M. In particular. where a deviation from the superweak theory has recently been confirmed [7–9]. All rights reserved. Argonne. USA d Enrico Fermi Institute. Carena a. CP violation could enable a relatively light Higgs boson to escape detection at LEP 2. Ellis a . accepted 5 June 2000 Abstract We perform a systematic study of the one-loop renormalization-group-improved effective potential of the minimal supersymmetric extension of the Standard Model (MSSM).b .V. the Tevatron and the LHC. as well as those associated with the top.∗ . Batavia. In the limit of a large charged Higgs-boson mass. P. including CP violation induced radiatively by soft trilinear interactions related to squarks of the third generation. IL 60637.ch 1 For pedagogical introductions to CP violation in the B-meson system. Argonne National Lab. see Refs. CERN. Introduction The violation of CP was first observed in the neutral kaon system [1–6]. the lightest neutral Higgs boson resembles that in the Standard Model (SM). All rights reserved. MH ±  MZ . USA Received 3 April 2000.O.Nuclear Physics B 586 (2000) 92–140 www. and CP violation occurs only in the heavy Higgs sector.

in which the CP symmetry of the theory is broken spontaneously by the ground state of the Higgs potential [14–18]. the Higgs potential of the MSSM is invariant under CP at the tree level.M. a significant portion of the parameter space spanned by tan β and the CP-odd scalar mass MA can be tested for the CP-conserving case in this next round of experiments at LEP 2 [25]. Consequently. and any explicit or spontaneous breakdown of CP symmetry can arise only via radiative corrections. The case of purely spontaneous CP violation in the MSSM [19] leads to an unacceptably light CP-odd scalar [20–22]. Carena et al. even though the upper bound on its mass was found [25] to be very similar to that found previously in the CPconserving case [27–49. predict an extended Higgs sector. as a result of the Georgi–Pais theorem [23]. namely for explaining the underlying mechanism which caused matter to dominate over anti-matter in our observable universe [13]. experiments at LEP 2. √ running at center-of-mass energies s = 196–202 GeV. The earlier study of the renormalization-group (RG) improved effective potential of the MSSM with explicit CP violation in [25] was based on an expansion of the Higgs quartic couplings in inverse powers of the arithmetic average of stop and sbottom masses. CP violation is broken explicitly in the Standard Model (SM) by complex Yukawa couplings of the Higgs boson to quarks. Supersymmetric (SUSY) theories. On the other hand. Although a fundamental understanding of the origin of CP violation is still lacking. The mass expansion of the one-loop effective potential was truncated . most of the scenarios proposed in the existing literature indicate that Higgs interactions play a key role in mediating CP violation. which may lead to drastic modifications of the tree-level couplings of the Higgs particles to fermions [25. / Nuclear Physics B 586 (2000) 92–140 93 to its interest for particle physics.51–53]. It has recently been shown [24] that the tree-level CP invariance of the MSSM Higgs potential may be violated sizeably by loop effects involving soft CP-violating trilinear interactions of the Higgs bosons to top and bottom squarks. for stop masses smaller than 1 TeV. and therefore may realize either or both the above two schemes of explicit and spontaneous CP violation. the explorable region of parameters is smaller for larger amounts of stop mixing and/or larger CP-violating phases.and right-handed stop and sbottom masses are small. Another appealing scheme of CP violation occurs in models with an extended Higgs sector. A detailed study [25] has shown that significant CP-violating effects of level crossing in the Higgs sector can take place in such a minimal SUSY scenario of explicit CP violation. including the minimal supersymmetric extension of the Standard Model (MSSM). and hence such a scenario is ruled out experimentally. The latter can have important phenomenological consequences on the production rates of the lightest Higgs particle. The MSSM predicts an upper bound on the lightest Higgs boson mass of approximately 110 (130) GeV for small (large) values of the ratio of Higgs vacuum expectation values tan β ≈ 2 (> 15). LEP 2 is expected to √ run at center-of-mass energies up to s = 206 GeV during the year 2000. A decisive test of such a scenario can only be provided by the upgraded Tevatron collider and the LHC. However. For instance. have placed a severe lower bound of approximately 108 GeV on the mass of the SM Higgs boson [54]. CP non-conservation provides a key ingredient for cosmological baryogenesis. 26] and to the W ± and Z bosons [25]. under the assumption that the mass splittings of the left. However.

we also include the leading logarithms generated by one-loop gaugino and higgsino quantum effects [44]. it is important to consider the constraints on the low-energy CP-violating parameters of the MSSM that originate from experimental upper limits on the electron [55] and neutron [56] electric dipole moments (EDMs) [57–84]. Although the above approach captures the basic qualitative features of the underlying dynamics under study. u and d quark flavours [75. and through two-loop gaugino/higgsino-mediated EDM graphs [84]. The results of the analysis are compared with those obtained in the CP-conserving case [48. Finally. making the first two generations of squarks rather heavy. In this analysis.50] and also with the results obtained by truncating the one-loop RG-improved effective potential up to renormalizable operators [25]. third-generation squarks can also give rise by themselves to observable effects on the electron and neutron EDMs through the three-gluon operator [78. Since the dominant CP-violating loop contributions to the effective Higgsboson masses and mixing angles occur for large values of the third-generation squarkmixing parameters. it is known [25. through the effective coupling of the ‘CP-odd’ Higgs boson to the gauge bosons [80–83]. we consider here the two-loop leading logarithms induced by top.79]. Another interesting possibility for avoiding any possible CP crisis in the MSSM is to arrange for cancellations among the different EDM terms. we implement the potentially large two-loop corrections that are induced by the one-loop stop and sbottom thresholds in the top. one might require a specific form of nonuniversality in the soft trilinear Yukawa couplings [77. which may become particularly relevant in the large-tan β regime. 2 To this end. Alternatively.and bottom-quark Yukawa couplings as well as those associated with QCD corrections.76]. . In the calculation of the RG-improved effective potential. is a possibility that can drastically reduce one-loop contributions to the neutron EDM.94 M. it is necessary to provide a more complete one-loop computation of the effective MSSM Higgs potential with explicit radiative breaking of CP invariance.49] that such a mass expansion has limitations when the third-generation squark mixing is large. On the basis of the RG-improved effective potential.and bottom-quark Yukawa couplings. we present predictions for the Higgs-boson mass spectrum and the effective Higgs couplings to fermions and gauge bosons. In our phenomenological discussion. much above the TeV scale [66–71]. 48. whilst sbottom contributions and relevant D-term effects on the effective potential are not considered in [26]. However. without suppressing the CP-violating phases of the theory. Carena et al. / Nuclear Physics B 586 (2000) 92–140 up to renormalizable operators of dimension four. we take into 2 We recall that the diagrammatic computation of scalar–pseudoscalar transitions is the focus of [24]. Most of the EDM constraints affect the CP-violating couplings of the first two generations [57–65].80–83]. including non-renormalizable operators below the heavy scalar-quark scale and without resorting to any other kinematic approximations. Thus. either at the level of short-distance diagrams [72–74] or (for the neutron EDM) at the level of the strong-interaction matrix elements for operators with s. by means of RG methods. These different EDM contributions of the third generation can also have different signs and add destructively to the electron and neutron EDMs.

c. 2. Section 4 is devoted to the calculation of the effective top. These new CP-odd phases may reside in the following parameters: (i) the mass parameter µ describing the bilinear mixing of the two Higgs chiral superfields in the superpotential. + M e λW e + mB 2 2 e† e 2 e∗ e eU2 U e∗ U e+M eD eE2 E e∗ E e + m21 Φ eQ e† Φ e Q Q+M D D+M +M 1 1  e1T iτ2 Φ2 + h. including the quantum-number assignments of the fields under the SM gauge group. such as soft CP-violating trilinear interactions. (iii) the soft bilinear Higgs-mixing mass Bµ.2) b1 and τ2 is e1 = iτ2 Φ ∗ is the scalar component of the Higgs chiral superfield H where Φ 1 the usual Pauli matrix. . CP violation is introduced into the MSSM through the Higgs superpotential and the soft supersymmetry-breaking Lagrangian: b1T iτ2 Lˆ E b + hd H b1T iτ2 Q bD b + hu Q bT iτ2 H b2 U b − µH b1T iτ2 H b2 .2). SU(2)L and U(1)Y . after including the leading two-loop logarithms associated with Yukawa and QCD corrections. respectively. In Section 5. and derive the analytic expressions for the effective charged and neutral Higgs-boson mass matrices. we discuss the phenomenological implications of representative CP-violating scenarios compatible with EDM constraints for direct Higgs searches at LEP 2 and the upgraded Tevatron collider.M. mW e and mB e of the gauge groups SU(3)c . are displayed in Table 1. CP-violating one-loop effective potential In this section.and bottomquark Yukawa couplings. (ii) the soft supersymetrybreaking gaugino masses mg˜ .c. This paper is organized as follows: in Section 2 we calculate the complete one-loop CPviolating effective potential. Section 3 describes our approach to determining the RG-improved Higgs-boson mass matrices. As can be seen from (2. Carena et al. Then we calculate the general one-loop CP-violating effective potential. in Section 6 we summarize our conclusions. We also compare the results of our analysis with those obtained using the mass-expansion method [25]. (2. after implementing the minimization tadpole conditions related to the Higgs ground state. and (iv) the soft . Finally. Finally. + m22 Φ2† Φ2 − BµΦ  T e + hd Ad Φ † Q ee ee + hl Al Φ1† L˜ E 1 D − hu Au Φ2 iτ2 QU + h.c. / Nuclear Physics B 586 (2000) 92–140 95 account the relevant EDM constraints related to the CP-violating parameters of the stop and sbottom sectors. we first describe the basic low-energy structure of the MSSM that contains explicit CP-violating sources.1) and (2. Technical details are given in Appendices A and B. the MSSM includes additional complex parameters with new CP-odd phases that are absent in the SM. The conventions followed throughout this paper. we derive the effective charged and neutral Higgs-boson mass matrices. W = hl H  1 i i eL2 L˜ † L˜ −Lsoft = − mg˜ λag˜ λag˜ + mW e λW eλB eλB e + h. in which one-loop threshold effects of the third-generation squarks are implemented.1) (2.

1. 2. .. / Nuclear Physics B 586 (2000) 92–140 Table 1 The field content of the MSSM Superfields Bosons Fermions SU(3)c ⊗SU(2)L ⊗U(1)Y Gauge multiplets ba G b W Gaµ 21 λa g˜ a ei W (8. the θ coordinate of superspace carries charge 1. d)L (uR )C = (uC )L (dR )C = (d C )L 0 . We concentrate on the parameters which may have a dominant CP-violating effect on the MSSM Higgs potential. there may exist other large CP-odd phases. 3.e. 23 ) (1. 2. (ii) The U(1)R symmetry acting on the Grassmann-valued coordinates. Q(H b2 ) = −2. associated with flavor off-diagonal soft supersymmetry-breaking masses of squarks and sleptons. the different trilinear couplings Af may be considered to be all equal at MX . l)L (eR )C = (eC )L QT = (u. whilst the Higgs superfields are R-neutral.86]. i. Af ≡ A. 2. under the assumption of a common phase for the gauginos. 0) b B Matter multiplets Lˆ b E b Q b U b D b1 H b2 H Wµi 21 τi Bµ ˜L L˜ T = (˜νl . −φ − ) Φ 1 1 Φ2T = (φ2+ . 1. the gaugino masses will then have a common phase. 2. 0) LT = (νl . 2) (3. ˜ L Q ˜ d) ∗ e = u˜ U R e = d˜∗ D R eT = (φ 0∗ . −1) (1. −1) (1. 3 Two CP-odd phases may further be eliminated by employing the following two global symmetries that govern the dimension-four operators in the MSSM Lagrangian: ˆ = 0. Bµ. 1. the trilinear couplings Af and the parameter µ. Q(Q) b = Q(L) (i) The U(1)Q symmetry specified by Q(H b) = 2 and Q(D) b = Q(E) b = −1. 1) supersymmetry-breaking trilinear couplings Af of the Higgs bosons to sfermions. ψ− ) (ψ¯ H H1 1 + 0 ) (ψH . We assume that these offdiagonal masses are small and therefore do not give sizeable contributions to the effective Higgs potential [85. 1. l) e = l˜∗ E R eT = (u. This U(1)Q symmetry is broken by the µ Q(U parameter and the respective soft supersymmetry-breaking one. the matter superfields and gaugino fields carry charge 1.e. i. 0) e B (1. b1 ) = 1. The number of independent CP-odd phases may be reduced if one prescribes a universality condition for all gaugino masses at the unification scale MX . Under the R transformation. see [87]. 1. Correspondingly.. Carena et al.96 M. In this case. In addition. The R symmetry is violated by the gaugino masses. because of the different RG running of the phases of the trilinear couplings. − 43 ) (3. however. their values at low energies will be different. the latter is made less important by the fact that the one-loop gaugino corrections are 3 For a discussion of the RG effects. 13 ) (3. φ20 ) 1 (1. ψH 2 2 (1.

M. As has been mentioned above. / Nuclear Physics B 586 (2000) 92–140 97 subdominant compared to the ones induced by the third-generation squarks. QT = (tL . It is obvious from (2. bR .b ) are the only physical CP-violating phases in the MSSM which affect the Higgs sector in a relevant way. it is useful to give the interaction Lagrangians related to the F and D terms of the third generation: . A} may be removed by employing the global symmetries (i) and (ii). Specifically.1) and (2. b˜L ) and t˜R . one of the Higgs doublets and the common phase of the gaugino fields can be rephased in a way such that the gaugino masses and Bµ become real numbers. play the most significant role in radiative corrections to the Higgs sector. two CP-odd phases of the complex parameters {µ.2) that the Yukawa interactions of the third-generation eT = quarks. b˜R . Q (t˜L . As a consequence. arg(µ) and arg(At. mλ . Therefore. bL ) and tR . Carena et al. as well as their SUSY bosonic counterparts. m212 .

.

.

.

e.

2 + |ht |2 .

Φ2T iτ2 Q e.

2 −LF = |hb |2 .

c. − µh∗b Q   ∗ − ht Φ2 t˜R∗ .3) − h∗b b˜R Φ1T iτ2 + h∗t t˜R Φ2† hb iτ2 Φ1∗ b˜R 2 .Φ1+ Q  e† Φ2 b˜R + µh∗t Q e† iτ2 Φ1∗ t˜R + h. (2.

.

.

.

  g e† Q e.

2 + 2.

Φ † Q e.

2 − Q e Φ † Φ1 + Φ † Φ2 −LD = w 2.

L0V is the tree-level Lagrangian of the MSSM Higgs potential     2 † † L0V = µ21 Φ1† Φ1 + µ22 Φ2† Φ2 + m212 Φ1† Φ2 + m∗2 12 Φ2 Φ1 + λ1 Φ1 Φ1 2     + λ2 Φ2† Φ2 + λ3 Φ1† Φ1 Φ2† Φ2 + λ4 Φ1† Φ2 Φ2† Φ1 .6) In (2. . More explicitly.c. where ht and hb are the top and bottom Yukawa couplings. (2.8) .Φ1T iτ2 Q 2 1 2 4   02  1 †  4   g ∗ eQ e − t˜R t˜R∗ + 2 b˜R b˜R Φ2† Φ2 − Φ1† Φ1 Q . we now proceed with the calculation of the one-loop effective potential.3)–(2. . µ21 = −m21 − |µ|2 .5) −Lfermions = hb b¯R QT Φ1∗ + h.b i=1. −LV = −LV + 32π 2 Q 2 Q 2 q=t. (2. the one-loop CP-violating effective potential is determined by "  #   m e2qi m ¯ 2q 3 X X 4 3 3 0 4 m eqi ln 2 − − 2m ¯ q ln 2 − . Further. m212 = Bµ. respectively.7) with µ22 = −m22 − |µ|2 . With the help of the Lagrangians (2. the interaction Lagrangian of the Higgs bosons to the top and bottom quarks is given by   (2.4) + 4 3 3 3 where gw and g 0 are the usual SU(2)L and U(1)Y gauge couplings. − g 02 .5). in the MS scheme.6).   1 2 1 2 1 2 + g 02 . λ3 = − gw λ4 = gw λ1 = λ2 = − gw 8 4 2 (2.c. + ht t¯R QT iτ2 Φ2 + h.2 (2.

f2 e e = M M 1 2 DQ Q D f2 M f2 M f2 M  DU e2 e† Q e = MQ 12 Q  eU e∗ U  eD e∗ D  f2 e e∗ M UD  et2 + |ht |2 Φ † Φ2 + 1 g 02 Φ † Φ1 − Φ † Φ2 .11) M  (M UQ UU UD  . t2 . D = bR }. we denote by breaking masses of the third generation of squarks.6). which depend on the of the quark and squark mass matrices M† M and M † Higgs background fields. respectively. M Q et2 M e2 M b (2. f2 e e = M M 2 1 UQ Q U   † † † f2 e† e∗ = hb Ab Φ − hb µ∗ Φ . In the weak basis {Q eT = The (4 × 4) squark mass matrix M ∗ 2 ∗ f ˜ ˜ e e ˜ ˜ (tL . Carena et al. m ¯ 2i (with i = t. t2 . b2 ) of M Appendix A. − gw Φ1 Φ1 − Φ2 Φ2 + 2 4 w 12   f2 †e† e∗ = −ht At Φ T iτ2 + ht µ∗ Φ T iτ2 .98 M. (2. = M b 1 DU e2 . It is rather difficult to express the four f2 in a simple form. 2 2 2 f ) e e (M f ) e e∗ (M f ) e e∗ (M DQ with DD  + |hb |2 Φ1 Φ1† + |ht |2 Φ2† Φ2 12 − Φ2 Φ2†    1 02 1 2 1 2 † † g − g Φ1† Φ1 − Φ2† Φ2 12 . as we detail in eigenvalues m e2qk (qk = t1 . for φ1. b1 .3. |ht | |φ2 | and m f2 is more complicated. However. U = tR . M M reads   |ht |2 |Φ2 |2 ht h∗b Φ1T iτ2 Φ2 † .2 = 0. it is not necessary to know the analytic form of m e2qk in order to evaluate the ± Higgs-boson masses and mixing angles [88].2 0 2 0 2 2 2 2 ¯ b = |hb | |φ1 | . / Nuclear Physics B 586 (2000) 92–140 Further. =M 2 1 2 3  1 † eb2 + |hb |2 Φ Φ1 − g 02 Φ † Φ1 − Φ † Φ2 . Specifically. b) and m e2qk (with qk = t1 . b1 . in (2.10) simplifies to the known expressions: m ¯ 2t = It is easy to see that. the field-dependent squark eigenvalues simplify to  2 02 .10) ± = 0.12) and the soft supersymmetryHere and in the following. M may be cast in the form:  f2 f2 ) e† e∗ (M f2 ) e† e∗  (M )Q e† Q e (M Q U Q D f2 =  f2 ) e e (M f2 ) e e∗ (M f2 ) e e∗  (2. Of course. (2.9) M M= h∗t hb Φ1† iτ2 Φ2∗ |hb |2 |Φ1 |2 with eigenvalues 1h m ¯ 2t (b) = |hb |2 |Φ1 |2 + |ht |2 |Φ2 |2 2 q i 2 + (−) |hb |2 |Φ1 |2 + |ht |2 |Φ2 |2 − 4|ht |2 |hb |2 |Φ1† Φ2 |2 . =M 1 1 2 6  f2 ∗e e∗ = ht h∗ Φ T iτ2 Φ2 . bL ). (2. b2 ) denote the eigenvalues f2 . for φ1.

.

.

.

.

.

 1 e2 2 et2 + 2|ht |2 .

φ 0 .

2 + gw + g .

φ 0 .

2 − .

φ 0 .

2 m et1 (t2 ) = MQ + M 2 1 2 2 4  q .

.

 0 2 0 2 2 0 0 .

2 2 2 2 ∗ .

e e + 4|ht | At φ2 − µ φ1 . + (−) MQ − Mt + xt |φ1 | − |φ2 | .

M. / Nuclear Physics B 586 (2000) 92–140 m e2b1 (b2 ) = 99  2 02 . Carena et al.

.

.

.

.

.

 1 e2 eb2 + 2|hb |2 .

φ 0 .

2 − gw + g .

φ 0 .

2 − .

φ 0 .

2 MQ + M 1 1 2 2 4  q .

.

 e 2 − xb |φ 0 |2 − |φ 0 |2 2 + 4|hb |2 .

A∗ φ 0 − µφ 0 .

13) 2 − 5 g 02 ) and x = 1 (g 2 − 1 g 02 ). i.14) Φ1 = 1 √ (v1 + φ1 + ia1 ) √1 (v2 + φ2 + ia2 ) 2 2 where v1 and v2 are the moduli of the vacuum expectation values (VEVs) of the Higgs doublets and ξ is their relative phase.2 (2.b i=1. e2 − M + (−) M Q b 1 b 1 2 2 (2. m ln = +(−)v2(1) Im m212 eiξ − ∂a1(2) q˜i 16π 2 Q2 q=t. Moreover. from (A. Following [24.e. (2.1) and (A.3). h∂ m e2qk /∂φ1(2)i and h∂ m e2qk / where he m2qi i = m2q˜i . we consider the following linear expansion of the Higgs doublets Φ1 and Φ2 around the ground state:     φ1+ φ2+ . with tan β = v2 /v1 . Φ2 = eiξ .15) Q   ∂LV Ta1 (a2 ) ≡ ∂a1(2)     m2 e2qi  3 X X ∂m q˜ i 2 − 1 . and the tadpole derivatives h∂ m ∂a1(2)i are given in (A. we readily see that Ta1 = − tan βTa2 .17) = a2 a sin β cos β The Higgs potential then has a flat direction with respect to the G0 field. we require the vanishing of the total tadpole contributions   ∂LV Tφ1 (φ2 ) ≡ ∂φ1(2)    v1 v2 1 2 2 iξ 2 2 = v1(2) µ1(2) + 2 Re m12 e + λ1(2) v1(2) + (λ3 + λ4 )v2(1) 2 v1(2) "      m2  ∂m ¯ 2q e2qi 3 X X ∂m q˜ i 2 − 1 − 2 m ln m2 − ∂φ1(2) q˜i ∂φ1(2) q 16π 2 Q2 q=t. (2.2  # m2q × ln 2 − 1 .25].16) ¯ 2q /∂φ1(2)i. where xt = 14 (gw b 3 4 w 3 After having set the stage. (2.3).25] allows us to perform an orthogonal rotation in the space spanned by the ‘CP-odd’ scalars a1 and a2 . As usual. h∂LV / ∂G0 i = 0. we now derive the minimization conditions governing the ground state of the MSSM one-loop effective potential and then determine the Higgsboson mass matrices. .b i=1. and the G0 field becomes the true would-be Goldstone boson eaten by the longitudinal component of the Z boson.2 .  0    cos β − sin β G a1 .. This last fact [24.

such as scalar-pseudoscalar transitions.25]. After performing the above substitutions. For example. However.2    m2 e2qk i(1 − δij ) ∂ m q˜ m2q˜k ln 2k − 1 vi ∂aj Q #  2    2 2  ¯q ¯q m2q ∂ m δij ∂ m 2 − mq ln 2 − 1 . as well as the tadpole terms h∂ m . we can express the charged-Higgs-boson mass matrix as follows:    1 2 v1 v2 v1 v2 Re m212 + gw M2± ij = (−1)i+j vi vj 4 "    2  e2qk ∂ 2m eqk δij ∂ m 3 X X − + + − 2 vi ∂φj 16π ∂φi ∂φj q=t.b k=1.16) that a relative phase ξ between the two Higgs vacuum expectation values is induced radiatively in the MS scheme [24. such as Higgs-boson masses and tan β. As we detail below. Carena et al. in which ξ is set to zero order-by-order in perturbation theory [24]. the square of the charged Higgs-boson mass MH + 2 determined by the matrix element (M± )12 : (  v2 1 2 2 v1 v2 Re m212 + gw MH + = v1 v2 4 "   2 2   2   m2 eqk ∂ m eqk i ∂m 3 X X q˜ k 2 − mq˜k ln 2 − 1 − 16π 2 v1 ∂a2 Q ∂φ1+ ∂φ2− q=t. are ∂φ1 ∂φ2 i. we should stress that the phase ξ is renormalization-scheme dependent. Employing the tadpole conditions Tφ1 = Tφ2 = 0 and Ta1 = Ta2 = 0 allows one to substitute the mass parameters µ21 and µ22 . but to receive an imaginary counter-term (CT).18) 2 may be Since det M2± = 0 in (2. one may adopt a renormalization scheme.19) mq ln 2 − 1 −2 Q ∂φ1+ ∂φ2− ¯ 2q /∂φ1+ ∂φ2− i and h∂ 2 m e2qk / The self-energy derivatives appearing in (2. as irrelevant ξ -dependent phases in the effective chargino and neutralino mass matrices can thereby be completely avoided. which is determined by the vanishing of the CP-odd tadpole parameters Ta1 and Ta2 for ξ = 0. h∂ m eqk /∂φj i and h∂ m eqk /∂aj i. In fact. and Im m212 into the effective potential (2. are independent of the renormalization subtraction point Q2 and the choice of phase ξ .100 M.b k=1. −2 vi ∂φj Q ∂φi+ ∂φj−  − (2. (2.18) and (2. we evaluate the one-loop effective Higgs-boson mass matrices. Im m212 . This can be achieved by requiring the bilinear Higgs-mixing mass m212 to be real at the tree level. In the following.19): h∂ 2 m + − 2 2 2 ¯ qk /∂φj i. slightly different from the MS one.6). at higher orders. / Nuclear Physics B 586 (2000) 92–140 We observe from (2.18).2 #)  2 2   ¯q m2q ∂ m 2 . it has been explicitly demonstrated in [24] that physical CP-violating transition amplitudes. we adopt the ξ = 0 scheme of renormalization. In the remaining part of this section. the scheme of renormalization of Im(m212 eiξ ) does not directly affect the renormalization scheme of other physical kinematic parameters of the theory to one loop.

(2.2  2 eqk ∂m e2qk ∂ m  ∂φj  2 ¯ 2q ∂ m ∂φi ∂φj  2  2  ¯q ∂m ¯ q ∂m ∂φi ∂φj ln − ln m2q˜k  Q2  2  ¯q δij ∂ m vi  m2q Q2 " ∂φj ) m2q   m2q ln 2 − 1 Q . a2 . φ1 . Carena et al. M2P and M2SP denote the two-by-two matrices of the scalar. pseudoscalar and scalar–pseudoscalar squared mass terms of the neutral Higgs bosons. By analogy. φ2 . . × mq˜k ln 2 − 1 + Q ∂φi ∂aj Q  v1 v2 Re m212 M2P ij = (−1)i+j vi vj (   2  X  ∂ 2 m e2qk eqk δij ∂ m 3 X + − 2 16π ∂ai ∂aj vi ∂φj  M2SP ij 3 X X = 16π 2  × m2q˜k q=t. / Nuclear Physics B 586 (2000) 92–140 101 exhibited in Appendix A.b ∂φi  −2 + k=1. The individual matrix elements of M20 are given by M2S  ij  1 2 v1 v2 Re m212 + gw + g 02 vi vj vi vj 4 (     2   m2 X e2qk ∂ 2m eqk δij ∂ m 3 X q˜ k 2 + − mq˜k ln 2 − 1 ∂φi ∂φj vi ∂φj 16π 2 Q = (−1)i+j  + q=t.2  −1 +   − ∂m e2qk  ∂m e2qk  (2. as the neutral-Higgs-boson mass matrix M2 in that 0 work is expressed in the weak basis {a1 .22) m2q˜k  ln 2 ∂ai ∂aj Q   2    2 2  2 ¯q ¯q mq ∂ m δij ∂ m − m2q ln 2 − 1 . (2. a1 . φ2 }. −2 ∂ai ∂aj vi ∂φj Q (2.b k=1.M.b ln m2q˜k Q2 ∂ 2m e2qk k=1. Therefore the CP-odd Higgs-boson mass MA cannot be identified with any of the neutral Higgs-boson masses. in the weak basis {φ1 . a2 }.23) 4 Notice that our convention differs from that given in [25]. the neutral-Higgs-boson mass matrix takes on the form 4   M2S M2SP  M20 = .2   2  2  m2 #  m2 ∂m eqk ∂ m eqk q˜ k q˜ 2 ln 2k . Observe that the presence of CP-violating self-energy terms leads to mass eigenstates with no well-defined CP quantum numbers.20) T M2P M2SP where M2S .21)  2  eqk (1 − δij ) ∂ m ∂φi ∂aj vi ∂aj q=t.

In the weak basis {φ1 . the (4 × 4) matrix M20 reduces to a (3 × 3) matrix. which we assume to be higher than the electroweak scale. in which the improvement of the Higgs-boson mass-matrix elements was performed by carefully applying the process of decoupling of the third-generation squarks. In the language of the RG approach.24).49]. describes the threshold effect of the decoupling of the heavier stop and sbottom squarks and their respective mixing with the lighter states. the reduced neutral mass-squared matrix M2N may be expressed by      1 M2 M2 M2  M2N =   S 11 S 12  2 MS 1 cos β 21 M2SP cos β  2  12 MS 22 − sin1 β M2SP  21 SP 12 − sin1 β M2SP  21  2 − sin β1cos β MP  .24). Using the expressions (2. and likewise for M2P .24) 12 in (2. Let us denote by Qt b the scale of the heaviest third-generation squark. the dominant contributions to the Higgsboson mass matrix M2 may be written conceptually as a sum of two terms: M2 (mt ) = M 2 (mt ) + M2. which we denote by M2N . M2. the upper bound on the lightest CP-even Higgs mass was found to be strongly affected by the two-loop logarithmic corrections [45–49]. These contributions would be present. we perform a one-loop RG improvement of the squared charged Higgs2 2 boson mass MH + and of the squared neutral Higgs-boson mass matrix MN . we have used the properties of the matrix elements of M2SP : In writing 2 (MSP )11 = − tan β(M2SP )12 and (M2SP )22 = − cot β(M2SP )21 . M2N 3. M 2 (mt ). Carena et al. The second term.1). In (3. the analytic expressions for the self-energy and tadpole derivatives with respect to the background Higgs fields are given in Appendix A. even if the left-right mixing of the stop and sbottom states were absent. Within the framework of the RG approach. we are interested in evaluating the effective potential at mt . Since G0 does not mix with the other neutral fields.1) The first term. As we explain below. / Nuclear Physics B 586 (2000) 92–140 Again. The RG improvement incorporates all leading two-loop logarithmic corrections to the Higgs-boson mass-matrix elements. we have . In carrying out the RG improvement. since it has been shown [46] that this is the scale at which two-loop corrections are minimized. we follow the procedure outlined in [48. (3. RG-improved Higgs-boson mass matrices In this section. In particular.102 M.  (2. At the oneloop level. the second term is manifestly scale independent. contains the genuine logarithmic contributions which determine the whole scale dependence of the one-loop effective potential. the renormalization of the above two contributions must proceed in different ways. which were already found in the CP-conserving case to give rise to significant contributions to the Higgs-boson masses and couplings. φ2 .th(mt ). a}.th (mt ).19) and (2. we determine the analytic forms of the RGimproved charged and neutral Higgs-boson masses in the next section.

/ Nuclear Physics B 586 (2000) 92–140 103 first to consider that the aforementioned threshold contribution is ‘frozen’ at the scale Qt b . First. the one-loop β functions of λi can generally have appreciable values. the anomalous-dimension factors in combination with the one-loop relation between the quark masses at scales Qt b and mt yield sizeable two-loop corrections to the mass-matrix elements originating from the one-loop threshold effects. M2. the contribution M 2 (mt ) of the third-generation squarks to the effective potential describes the genuine one.2) where ξi (mt ) is the anomalous dimension factor of the Hi state to be determined below. we first need to compute the one-loop values of the quartic couplings λi . 6. 4. Carena et al. 4.M. with i = 1. Thus.th (mt ) = 1M2ij (Qt b )ξi−1 (mt )ξj−1 (mt ). where the decoupling of the stop and sbottom contributions at their appropriate thresholds is properly taken into account. On field-theoretic grounds. there are two important technical details that should be mentioned. these quartic couplings must have vanishing one-loop β functions. (3. given by (3.and bottom-quark Yukawa couplings. As a result. Following the above discussion. which have to be conveniently re-expressed as functions of the corresponding running masses at mt .3) ln + |hb | − + 9 Q2 6 Q2 . with i = 5. the low-energy values of λi differ significantly from their tree-level ones.6) in the limit where the squark mixing parameters vanish. In this context. the tree-level Higgs quartic 2 couplings λi . with all the involved kinematic parameters defined at this particular scale. are all proportional to the squared gauge couplings gw and g 02 (cf. 2.44]). as they are proportional to the fourth power of the top. 3. which are absent in the Born approximation to the MSSM Higgs potential. To this end. However. The pertinent one-loop running quartic couplings. 2. Then. As was already mentioned. with i = 1. we have to rescale the threshold contribution with the anomalous dimension factors of the relevant Higgs fields: 2. The second technical remark pertains to the RG evolution of the Higgs quartic couplings λi . i. we now proceed with the RG improvement of the Higgs-boson mass-matrix elements.2). denoted by λi . 7 (for the notation.and two-loop leading-logarithmic running of the Higgs quartic couplings. The RG-improved approach followed here is crucial for implementing properly the potentially large logarithmic corrections to the Higgs quartic couplings.. 3. The best way to calculate the latter effects is to consider the logarithmic part of the effective potential (2. may then be obtained by "   e2  2 MQ + m2t g 02 2 3 gw (1) − ln λ1 = − 4 12 32π 2 Q2   e2   MQ + m2b g 02 2 g2 ln + |hb |2 − w − 4 12 Q2  2   e2   # et + m2t Mb + m2b M g 02 2 g 04 2 ln . The one-loop matrix elements 1M2ij (Qt b ) depend on the running quark masses at the scale Qt b . these quartic couplings are radiatively induced by threshold effects. (2.8)). and cannot be generated by RG running. However. see [25. (1) µ = At = Ab = 0.e. Mij (3. and have already been taken into account in 1M2 (Qt b ). in the MSSM. we notice that.th(Qt b ) ≡ 1M2 (Qt b ).

7) + |hb |2 Re(µAb ) ln Q2 The analytic form of 1M2 (Qt b ) in (3. we need to know the one-loop running of the soft supersymmetry-breaking parameter Re m212 . 1M (Qt b ) is given by   (1) (3.2) can now be obtained by subtracting the oneloop Born-improved mass matrix M2(0) from its total one-loop contribution M2(1) .8) 1M2 (Qt b ) = M2(1)(Qt b ) − M2(0) Re m2(1) 12 (Qt b ). M MQ + m2t max(M 3 (1) b b 2 2 | |h | |h ln + ln λ3 = − t b 16π 2 Q2 Q2   2   e2   2 2  MQ + m2t g 02 gw g 02 gw gw g 02 2 + |ht | − − − ln − 4 12 4 12 4 12 Q2       2 e 2 + m2  2 2 M g 02 gw g 02 gw gw g 02 Q b 2 − + |hb | − + ln − 4 12 4 12 4 12 Q2    2  et + m2t g 02 M g 02 2 |ht | − ln + 3 3 Q2 )    02 e 2 + m2  M g g 02 b b |hb |2 − ln . M MQ + m2t max(M 3 (1) b b 2 2 | |h | |h ln + ln λ4 = t b 16π 2 Q2 Q2 ) e2 + m2t  g 2  e 2 + m2  2  2  M 2  M g g gw Q Q b |ht |2 − w ln − w |hb |2 − w ln .5) + 6 6 Q2 (   e2   e 2 + m2 )  et2 + m2t . (3. Gathering the relevant logarithmic terms present in the effective potential (2. More explicitly. M et2 + m2t )  max(M 3 Q 2(1) 2 |ht | Re(µAt ) ln Re m12 = 16π 2 Q2  2 2 e2 + m2 )  e +m . (3. (3.6).104 M. expressed in terms of λ(1) i and Re m2(1) . − 2 4 2 4 Q2 Q2 (1) λ2 (3. 2(1) 2(1) we have in mind the charged and neutral Higgs-boson mass matrices M± and MN 2 calculated in Section 2.4) + |ht | − 3 36 Q2 Q2 (   e2   e 2 + m2 )  et2 + m2t . Here. we find   e 2 + m2t . M max(M Q b b b . Furthermore. λi (Qt b ) . it is essential to stress again that the kinematic parameters 12 . Carena et al.6) Moreover. / Nuclear Physics B 586 (2000) 92–140 "   e2  2 MQ + m2t g 02 2 3 gw 2 + =− ln |ht | − 32π 2 4 12 Q2   e2   2 MQ + m2b g 02 2 g ln + w+ 4 12 Q2   e2  e2  #  Mb + m2b g 02 2 Mt + m2t g 04 2 ln ln + . where M2(0) represents the tree-level functional form of M2 .

10) × ln Q2 Q2  3|ht |2 |hb |2 = − |ht |2 + |hb |2 − 8gs2 λ(2) 3 2 2 (16π )     e2 2 e2 2  e2 MQ + m2t 2 2 max(Mt + mt . by appropriately considering the stop and sbottom thresholds. (3. are evaluated at the scale Qt b . As has been done in (3.15). As has also been stated explicitly in (3. we can express the one. (3. 4.13) with i = 1. can be determined by solving iteratively the RG equations [48.13) and (3. (3.9) × ln Q2 Q2   6|ht |4 3 1 |ht |2 + |hb |2 − 8gs2 = − λ(2) 2 2 2 2 (32π ) 2   e2   e2 2 2  MQ + mt 2 2 Mt + mt + ln . Mb + mb ) + ln . (3.3)–(3. 3. Another important ingredient in the RG improvement of the Higgs-boson mass matrices is the analytic two-loop result for the Higgs quartic couplings λ1 . The last ingredient for completing the programme of RG improvement of the Higgsboson mass matrices is knowledge of the analytic expressions for the anomalous dimension .11) × ln Q2 Q2 (2) (2) λ4 = −λ3 .12) For later convenience. In this way.14). Carena et al.and two-loop leading logarithmic contributions M 2 (mt ) to M2 (mt ) by means of the two-loop Born-improved mass matrix:   M 2 (mt ) = M2(0) Re m (3. we define collectively the sum of the tree. which we denote by λi .8).49]. .15) ¯ 212 (mt ). λ¯ i (mt ) . Similarly. one-loop and two-loop quartic couplings as follows: (2) λ¯ i = λi + λ(1) i + λi . (3. . 2.M. . . we have to include two-loop leading logarithms. we obtain   6|hb |4 3 1 (2) 2 2 2 |hb | + |ht | − 8gs λ1 = − (32π 2 )2 2 2  e2   e2 2 2  + m M Q b 2 2 Mb + mb + ln . the sum of the tree. one-loop and two-loop contributions to the soft-bilinear Higgs mixing may be defined as 2(2) Re m ¯ 212 = Re m212 + Re m2(1) 12 + Re m12 . These two-loop leading (2) logarithmic contributions to the Higgs quartic couplings. M 2 (mt ) is expressed in terms of mass and coupling parameters evaluated at the top-quark-mass scale. such as masses and couplings. / Nuclear Physics B 586 (2000) 92–140 105 involved in (3.6). Given the above definitions of the quartic couplings and the soft Higgs-mixing parameter in (3. (3. λ4 . knowledge of the two-loop contribution Re m2(2) 12 is not required in the one-loop RG improvement of the MSSM Higgs potential.14) As we see below. Note that M 2 (mt ) also includes the tree-level terms.

2(1) t b 2 (see (3. the following relations among the matrix elements of the RG-frozen part 1M2± of the charged-Higgs-boson mass matrix are obtained:   1M2± 11 (Qt b ) = − tan β(Qt b ) 1M2± 12 (Qt b ). Finally. + MH + (3. After including the RG running due to the Higgs-boson anomalous dimensions. Adopting the framework outlined above [48. 32π 2 mt (3. As a consequence. 32π 2 mt ξ2− (mt ) = 1 + 3|ht |2 Q2t b ln 2 .16)). MH2 + (mt ) is the squared two-loop Born-improved charged Higgs-boson mass given by MH2 + (mt ) = Re m ¯ 212 (mt ) 1 + λ¯ 4 (mt )v 2 (mt ) sin β(mt ) cos β(mt ) 2 (3. we find . where (M± )t12 2 denotes the one-loop contribution of the third-generation squarks to (M± )12 .19).20) 1M2± 22 (Qt b ) = − cot β(Qt b ) 1M2± 21 (Qt b ). describes the one-loop quark contribution to MH (MH + ) + We remark that the charged-Higgs-boson mass matrix receives the common anomalous dimension factor ξ1+ (mt )ξ2− (mt ). 2 where the scale at which the kinematic parameters are to be evaluated has been indicated explicitly.16) e 2 + m2t .2).19) + λ4 (Qt b )v1 (Qt b )v2 (Qt b ) .   (3.106 M. M e 2 + m2 ). as one of its mass eigenstates must correspond to the massless would-be Goldstone boson G+ that forms the longitudinal component of the W + boson. it is not difficult to compute the RG2 (m ) at the top-mass scale through the relation: improved charged Higgs-boson mass MH t +  + −1 t˜b˜ − 2 2 2 (Qt b ) 1MH MH + (mt ) = M H + (mt ) + ξ1 (mt )ξ2 (mt ) +  2(1) t b (mt ). respectively: ξ1+ (mt ) = 1 + 3|hb |2 Q2t b ln 2 . and ξ + (mt ) and ξ − (mt ) are the where Q2t b = max(M Q b b 1 2 anomalous dimension factors of the charged Higgs fields φ1+ and φ2− . In (3.49]. This is because M2± must possess a vanishing determinant at any RG scale Q2 . even though different matrix elements of M2± are involved. These analytic expressions are given below for the charged and neutral Higgs-boson cases separately. Carena et al. the term between brackets is the threshold contribution to the off2(1) ˜b˜ diagonal matrix element of the charged-Higgs-boson mass matrix.17) Further.18) ˜ 2 )t˜b is the one-loop scale-invariant part that contains the stop and sbottom and (1MH + contributions:  t˜b˜ t˜b˜ 1 2 M2(1) (Qt b ) + Re m2(1) 1MH + (Qt b ) = − ± 12 (Qt b ) 12 sin β(mt ) cos β(mt )  1 (1) (3. / Nuclear Physics B 586 (2000) 92–140 factors that occur in (3. M et2 + m2t .

it is evident that the RG running of the different matrix elements of 1M2± may be expressed in terms of the running of (1M2± )12 and the value of tan β at the scale mt . The parameters ξij (with q = t. (1M2± )12 (Qt b ) . Q2t = max(M Q b Q b b b Notice that. with ξ1t˜(mt ) = 1 + ˜ ξ1b (mt ) = 1 + 3|hb |2 Q2t ln 2 .M. M et2 + m2t ) and Q2 = max(M e 2 + m2 . 2. 32π 2 mt ˜ 3|ht |2 Q2b ln . 32π 2 m2t (3. b) are the anomalous-dimension factors related to the neutral Higgs-boson fields ( q˜ q˜ ξ (mt )ξj (mt ). e2 + m2t . 32π 2 mt ξ2b (mt ) = 1 + 3|ht |2 Q2t ln 2 . ξ1+ (mt )ξ2− (mt ) ξ1+ (mt )ξ2− (mt ) 107 (3. the RG-improved neutral-Higgs-boson mass matrix M2N may be computed by  −1   t˜ 1M2N ij (Qt ) M2N ij (mt ) = MN2 ij (mt ) + ξijt˜ (mt ) −1 b˜  ˜ 1M2N ij (Qb ) + ξijb (mt ) 2(1) t b (3. 2.26) . j = 1. for the very same reasons as in the charged-Higgs-boson case. j = 1. the vanishing of the determinants of M2P and M2SP at any Q2 scale leads to the common anomalous dimension factor ξi3 (mt ) = ξ3i (mt ) = ξ1 (mt )ξ2 (mt ) in the calculation of M2N (mt ) in (3.24) ξij (mt ) = ξj i (mt ) = iq˜ q˜ ξ1 (mt )ξ2 (mt ). M e2 + m2 ). 32π 2 mt ξ2t˜(mt ) = 1 + 3|hb |2 Q2b ln 2 . / Nuclear Physics B 586 (2000) 92–140 (1M2± )11 (Qt b ) [ξ1+ (mt )]2 (1M2± )22 (Qt b ) [ξ2− (mt )]2 = − tan β(mt ) = − cot β(mt ) (1M2± )12 (Qt b ) . Carena et al. for i = 1. Correspondingly. cos β(mt ) (Qq ) = − 33 1M2N q˜ (Qq ) = − 23 q˜ (M2SP )21 (Qq ) .22) As a consequence of this last relation. the corresponding matrix elements of (1M2N )i3 in (3. sin β(mt ) q˜ (1M2P )12 (Qq ) . for i. one has to introduce here two decoupling scales Qt and Qb .23). sin β(mt ) cos β(mt ) (3. Since this last fact involves the matrix elements (M2P )12 . 2. with i. 3 and j = 3. as the stop/top and sbottom/bottom loop effects occur separately in the q˜ threshold contributions. unlike the charged-Higgs-boson case. (M2SP )12 q˜ and (M2SP )21 . q˜ q˜ (3. (3. 3.25) We should observe that.21) where we have used the RG relation: tan β(Qt b ) = ξ1+ (mt ) ξ2− (mt ) tan β(mt ).23) are given by 1M2N 1M2N q˜ (Qq ) = 13 q˜ q˜ (M2SP )12 (Qq ) .23) + MN ij (mt ).

0 0 0 q˜ with λ34 = λ3 + λ4 .30) differs from that chosen in [25]. and can therefore be diagonalized by an orthogonal transformation as follows:  2  2 2 (mt ). Analytic expressions for MHi (mt ) and O are presented in Appendix B.29) − v 2  λ34 sin β cos β 2λ2 sin2 β 0. we write (M2(1) N ) to denote 2 the one-loop part of MN containing the contributions of the third-generation squarks. λ¯ 34 (mt ) . λ2 (Qq ). the soft Higgs-mixing parameter Re m ¯ 212 (mt ) may be substituted by the squared 2 RG-improved mass MH + (mt ) of the charged Higgs boson (cf. MH (mt ) . Before closing this section. we observe that the free kinematic parameters of the MSSM Higgs sector are MH + (mt ).23). M Ab (Qt b ). M eb2 (Qt b ). The resulting RG-improved Higgs-boson mass matrix M2N (mt ) is a symmetric. as we assign the Higgs fields in the reversed order. MH (mt ).31) Notice that our convention in (3. we reiterate the fact that Im m212 can be renormalized independently.30) O T M2N (mt )O = diag MH 1 2 3 where we have defined the Higgs fields such that their RG-improved masses satisfy the inequality: MH1 (mt ) 6 MH2 (mt ) 6 MH3 (mt ).108 M. the (3 × 3) matrices MN2 and (1M2N )q˜ (with q = t. In (3.  2(1) q˜ 2 q˜ (Qq ) 1MN (Qq ) = MN   2(1). b) describe. positive-definite (3 × 3) matrix. / Nuclear Physics B 586 (2000) 92–140 where the RG-scale dependence of the involved quantities has been displayed explicitly. M tan β(mt ). First.16) and (3. two important remarks are in order.q˜ (1).27) ¯ 212 (mt ).18)) in the neutral Higgs-boson mass matrix M2N (mt ) in (3. λ¯ 1 (mt ). the two-loop Born-improved effects and the one-loop threshold contributions associated with the decoupling of the heavy squark states:  2(0)  MN2 (mt ) = MN Re m (3. et2 (Qt b ). As was . (3.28) − M2(0) N (1) (1) (1) 2(0) where λ34 = λ3 + λ4 (likewise λ¯ 34 = λ¯ 3 + λ¯ 4 ) and MN is the tree-level functional 2 form of MN :   sin2 β − sin β cos β 0 2 Re m 2(0) 12  0 − sin β cos β cos2 β MN = sin β cos β 0 0 1   2 λ34 sin β cos β 0 2λ1 cos β (3.23). and tb (M2(1) N ) to denote its fermionic one-loop counterpart. λ1 (Qq ). (3. Carena et al. At (Qt b ). λ34 (Qq ) . µ(Qt b ). (3. λ¯ 2 (mt ). respectively. (3. (3. As in the charged-Higgs-boson case. 2 eQ (Qt b ).q˜ (1). without affecting the renormalization of the physical parameters of the theory [24].q˜ (1). Secondly.q˜ Re m12 (Qq ).32) In fact.

we assume that only the third 12 generation of squarks contributes to γRe m2 .90–93]. the mass renormalization of H + may be entirely reabsorbed by a corresponding renormalization of Re m212 and λ4 .. so that the actual size of the radiative corrections to the Higgs-boson masses and couplings can properly be extracted [50.35) as it should be. Employing (2. As can also be seen from an important role in this proof is played (3. the corresponding enhanced tan β terms are much smaller.and bottom-quark Yukawa couplings.35). / Nuclear Physics B 586 (2000) 92–140 109 stressed in Section 3. because they are proportional to the weak gauge couplings [50. the one-loop RG relation between the bottom mass and the bottom-quark Yukawa coupling at the scale Qt b receives quantum corrections that also include terms proportional to tan β [89–98]. the effective top. apart from the usual RG running. For simplicity. The analytic forms of γRe m2 and βλ4 are given by 12  3  2 |ht | Re(µAt ) + |hb |2 Re(µAb ) .99]. f4 /∂a2 [24]. For the top-quark case. since fermions do not contribute to the RG 12 running of Im m212 . . which are induced by the decoupling of the heavy SUSY states at a high scale. the RG running of Re m212 due to t˜ and b˜ is only relevant for non-zero values of µAt and µAb .33) (3. Since these last terms can be significant for large values of tan β. Carena et al. Specifically. γRe m2 = − 2 12 16π  (1) 2 4   gw dλ4 3 gw 2 2 2 2 |ht | + |hb | + = − . 32π 2 ∂φ1+ ∂φ2− v1 ∂a2 (3.19) and examining only the ln Q2 -dependent part. the respective corrections can still give rise to 5 For the τ -lepton Yukawa coupling. the ξ = 0 scheme of renormalization gives rise to a considerable simplification. by the necessary CP-odd tadpole term ∂ Tr M 4. after including the RG running of Re m212 and λ4 . 5 we must resum them within the RG approach. Qt b . 2|ht | |hb | − βλ 4 ≡ d ln Q2 16π 2 2 4 (3.M. within the above ξ = 0 scheme. Effective top and bottom Yukawa couplings In addition to the RG improvement of the Higgs-boson mass matrices discussed in the previous section.and bottom-quark Yukawa couplings obtain additional non-logarithmic threshold contributions. In other words. instead.34) Obviously. it can be shown that MH + is Q2 -independent. we consider here a further improvement related to the non-logarithmic threshold corrections to the top. although one-loop suppressed. denoted as γRe m2 and βλ4 . For example.g. e. since we can get rid of the radiatively-induced phase ξ between the two Higgs vacuum expectation values in the analytic expressions of the Higgs-boson masses and mixing angles. one can verify that 2 dMH + d ln Q2 1 ∝ −γRe m2 − βλ4 v1 v2 12 2  2 f4   f4  3 ∂ Tr M i ∂ Tr M + − = 0. For the bottom-quark Yukawa case.

In particular.and topquark Yukawa couplings. In the following.4) I (a.5) hb = √ 2 MW cos β[1 + δhb / hb + (1hb / hb ) tan β] where δhb / hb and 1hb / hb are given in (4. (4. |mg˜ |2 + = A∗t µ∗ I m2t˜ . There may also be important CP-violating one-loop corrections to the bottom. We start our discussion by considering the effective Lagrangian of the b-quark Yukawa coupling [25.3). However. the reader is referred to the original literature [90– 98].2) and (4. Carena et al. m2t˜ . and I (a. In the leptonic sector. which were considered in Sections 2 and 3 in detail. c) = (a − b)(b − c)(a − c) The b-quark Yukawa coupling hb (Qt b ) is then related to the running b-quark mass mb (Qt b ) by gw mb . In (4. if the respective Higgs-mass eigenstate couples predominantly to the Higgs doublet Φ2 [90–98].10) and (2. b. Under such a field redefinition. m2b˜ . since only the moduli of the Yukawa couplings hb and ht enter the field-dependent quark and squark masses in (2. / Nuclear Physics B 586 (2000) 92–140 an enhancement of up to 4 GeV in the prediction for the lightest Higgs-boson mass [100.c. (4. these CP-violating vertex corrections are generally small [102]. and pay special attention to the CP-violating vertex effects. m2b˜ . m2t˜ . we have redefined the right-handed b-quark superfield. 2 1 2 1 2 hb 3π 16π 2   |h 1hb 2αs ∗ ∗ t| mg˜ µ I m2b˜ . the CP-violating radiative corrections to the couplings of the Higgs bosons to b quarks are significant [25]. in addition to the CP-violating effects induced by the radiative mixing of the Higgs states. the radiative effects of the Higgs-boson couplings to the bottom quarks can be further enhanced. 2 1 2 1 2 hb 3π 16π (4. while the phases of δhb / hb and 1hb / hb as well as those of the supersymmetry-breaking parameters do not change. b. The running b-quark mass mb (Qt b ) is obtained by means of the RG running of the b-quark mass from the scale mb . Moreover. For a general discussion of the form and the origin of these finite Yukawa corrections to the third-generation quark masses. only the Yukawa coupling hb becomes complex. because of the large Yukawa and colour-enhanced QCD interactions [89]. we give a brief discussion of the non-logarithmic corrections to the top and bottom Yukawa couplings. and therefore should be included in the computation.3) where αs = gs2 /(4π) is the SU(3)c coupling strength. respectively. |µ|2 .50]: 0∗ 0∗ −Lφ 0 bb ¯ = (hb + δhb )φ1 b¯ R bL + 1hb φ2 b¯ R bL + h.5). as the tree-level b-quark Yukawa coupling is suppressed in this case. (4.2) (4. |µ|2 . 101]. |mg˜ |2 − =− |µ|2 I m2t˜ . c) is the one-loop function ab ln(a/b) + bc ln(b/c) + ac ln(c/a) .. the neutral Higgs-boson mass matrices remain unaffected by the above .1) with  |ht |2  2αs ∗ δhb mg˜ Ab I m2b˜ .13).110 M. so that the physical b-quark mass is positive.

|mg˜ |2 − =− |µ| I m2b˜ . Again. The QCD correction is proportional to the hermitean conjugate of the sbottom-mixing parameter Xb = Ab − µ∗ tan β. |mg˜ |2 + = Ab µ I m2b˜ . the decoupling procedure for the heavy squark states requires that the effective b. Taking into account both CP-violating self-energy and vertex effects. m2t˜ . we have to make a judicious phase rotation of the right-handed t-quark superfield.and down-type quarks.8) 2 1 2 1 2 ht 3π 16π  |hb |2 2  2αs ∗ δht mg˜ At I m2t˜ . within the RG-resummation approach described in Section 3. one originating from QCD effects and another from a charginomediated graph. m2b˜ .c. it is interesting to observe that the sum δhb + 1hb tan β in (4. / Nuclear Physics B 586 (2000) 92–140 111 field redefinition.5) and (4.6) The corresponding relation for ht as a function of mt may easily be determined analogously by the effective Lagrangian (4. Therefore. one can show that such a field redefinition does not change the analytic results of the RG analysis.7) are evaluated at the scale Qt b . (4. Also.10) with S gH i dd  1 O1i O2i Re(hd + δhd ) + Re(1hd ) = hd + δhd + 1hd tan β cos β cos β    − Im(hd + δhd ) tan β − Im(1hd ) O3i . LH f¯f = − 2MW 2MW i=1 (4.5) receives two sorts of quantum corrections. As we discuss in Section 5.M. it is important to remark that. m2t˜ . such that the physical top-quark mass becomes positive. generically denoted as 1M2 (Qt b ).7) ht = √ 2 MW sin β[1 + δht / ht + (1ht / ht ) cot β] with  |hb |2 ∗ ∗  2αs ∗ ∗ 1ht mg˜ µ I m2t˜ . Carena et al. whilst the chargino-induced diagram [90– 93] depends linearly on the stop-mixing parameter Xt = At − µ∗ cot β. the non-logarithmic corrections must be treated as threshold effects and hence they should only contribute to the RG-frozen part of the Higgs-boson mass matrices. |µ|2 . we find   3 X   gw mu gw md ¯ S P S P d gHi dd + igHi dd γ5 d + Hi u¯ gHi uu + igHi uu γ5 u . At this point. we have checked that the very same property of invariance under rephasings of ht and hb persists for the charged Higgs-boson mass matrix as well. m2b˜ .11) . (4. (4. and reads gw mt . collectively denoted as u and d. It is now straightforward to obtain the interaction Lagrangians of the Higgs-boson mass eigenstates Hi to the up. |µ|2 . (4. (4.9) 2 1 2 1 2 ht 3π 16π As in the case of the b-quark Yukawa coupling. these additional Yukawa corrections can lead to observable effects in Higgs-boson searches. The effective Lagrangian describing the t-quark Yukawa coupling is given by −Lφ 0 t¯t = 1ht φ10 t¯R tL + (ht + δht )φ20 t¯R tL + h.6). At this stage.and t-quark Yukawa couplings given by (4.

− Im(hd + δhd ) cos β cos β  1 O2i O1i S = Re(hu + δhu ) + Re(1hu ) gH i uu hu + δhu + 1hu cot β sin β sin β    − Im(hu + δhu ) cot β − Im(1hu ) O3i . On the other hand. we demonstrate explicitly the aforementioned (non-)decoupling features of CP violation by analyzing specific phenomenological examples. Thus. For similar reasons.12) (4.112 M. neutral Higgs bosons are predominantly produced via the Higgs-strahlung processes in e+ e− and q q¯ collisions. such as e+ e− → Z ∗ → . 2  1 (4.15) mu = √ hu + δhu + 1hu cot β v2 2 have already been defined to be positive in (4.u = 0. In the next section.    1 P Re(1hu ) − Re(hu + δhu ) cot β O3i gHi uu = hu + δhu + 1hu cot β  O2i O1i − Im(1hu ) − Im(hu + δhu ) . 5.13) (4. and also comment on the enhanced search capabilities offered by the LHC. Carena et al. We focus our attention on the physics potential for discovering Higgs bosons with mixed CP parities at LEP 2 and the upgraded Tevatron collider. sin β sin β P gH = i dd (4. We can therefore conclude that CP-violating self-energy and vertex effects do not decouple in the heavy neutral Higgs sector. respectively. one has O31 → 0.7). for values of the charged Higgs mass MH +  MZ . In this limit. / Nuclear Physics B 586 (2000) 92–140    1 Re(1hd ) − Re(hd + δhd ) tan β O3i hd + δhd + 1hd tan β  O1i O2i − Im(1hd ) . it is interesting to investigate the behaviour of self-energy. At the LEP 2 and Tevatron colliders. while O11 → cos β and O21 → sin β.14) where the Higgs scalar and pseudoscalar couplings are normalized with respect to their SM values. given by gw md /(2MW ) and gw mu /(2MW ). in the same large MH + limit. Phenomenological discussion In this section we discuss the phenomenological implications of radiative Higgs-sector CP violation in the MSSM for Higgs-boson searches at high-energy colliders.5) and (4. the scalar components of the H1 dd and H1 uu couplings acquire the known SM form. the scalar and pseudoscalar couplings of both the two heaviest Higgs bosons H2 and H3 to the up and down fermions do not vanish. where  1 md = √ hd + δhd + 1hd tan β v1 . as they are proportional to O31 and Im md. the pseudoscalar parts of the H1 dd and H1 uu couplings vanish.and vertex-type CP violation in the decoupling limit of a heavy charged Higgs boson in the MSSM. Finally.

with i = 1. Note that the couplings Hi ZZ and Hi W + W − are related to the Hi Hj Z couplings through gHk V V = εij k gHi Hj Z . 2 cos θw (5. (5. if two Higgs-boson couplings to gauge bosons are known. the soft supersymmetry-breaking and µ parameters are set to the values . gHi Hj Z = O3i (cos βO2j − sin βO1j ) − O3j (cos βO2i − sin βO1i ). For definiteness.6) For completeness.4) (5. proper vertex corrections to the Hi ZZ and Hi Hj Z couplings do not contain strong-coupling.2) the interactions of the charged Higgs bosons H ± with the neutral Higgs and W ∓ bosons. (5. ∂ µ ≡ ∂ µ ← − ∂ µ .5) (5. / Nuclear Physics B 586 (2000) 92–140 113 ∗ ZHi [103–105].1) LH V V = gw MW i µ 2 cos2 θw i=1 LH H ∓ W ± = LH H Z = gw 2 3 X ↔  gHi H − W + Hi i ∂ µ H − W +.or tan β-enhanced diagrams. we neglect proper vertex corrections to the Hi ZZ and Hi Hj Z couplings. gHi H − W + = cos βO2i − sin βO1i + iO3i . Therefore. naive dimensional analysis suggests that these corrections are suppressed relative to their tree-level values by loop factors of the kind: (3αw /4π)(|µAt |/m2t˜ ). we have assumed that the dominant contributions arise from Higgs-mixing effects. In addition to the Higgs-boson masses. unless stated otherwise. (3αw /4π)(|µAt |2 v 2 /m6t˜ ) . iV V (5.3) j >i=1 where ↔ → cos θw = MW /MZ . Carena et al. For our phenomenological discussion of CP violation. (5. As opposed to the b-quark Yukawa case discussed in Section 4. In the above calculation of the effective Higgs-gauge-boson couplings.µ + H Z Z . 2. where t˜1 is the heaviest stop squark (see (2. 1 1 10−2 ..108]. this is sufficient to determine the complete set of the couplings gHi V V and gHi Hj Z [107.µ + h.M.8) i=1 Evidently. we consider the following two representative values for tan β: (i) tan β = 4 and (ii) tan β = 20. q q¯ → Z ∗ → ZHi and q q¯ → W ± → W ± Hi [106]. and gHi V V = cos βO1i + sin βO2i . In the following. we have included in (5. Higgs bosons can also be produced copiously in pairs through the reactions: e+ e− → Z ∗ → Hi Hj and q q¯ → Z ∗ → Hi Hj . If the next-to-lightest Higgs boson is not too heavy.7) Moreover. (5.2) i=1 3 X ↔  gw gHi Hj Z Hi ∂ µ Hj Z µ .c. unitarity provides the constraint 3 X 2 gH = 1. The effective Lagrangians governing the interactions of the Higgs bosons with the W ± and Z bosons are given by [25]   3 X 1 µ gHi V V Hi Wµ+ W −. the Higgs-boson couplings to the gauge fields play an essential role in our forthcoming discussion.13)). 3.

. one might then need to arrange for cancellations among the different one.49]. the two-loop EDM contributions are of the order of the experimental EDM upper bounds mentioned above. Ab and µ.45. / Nuclear Physics B 586 (2000) 92–140 eQ = M et = M eb = 0. for all the parameter space for which the H1 -boson mass acquires. For our illustrations. as one way to avoid the one-loop EDM constraints [72–74].and right-handed squark mass MSUSY . in such a scheme. so as to violate present experimental constraints. large CP-odd phases may lead to rather large EDM contributions. Again. Then. we shall take the µ parameter to be real and assume that the only CPodd phases in the theory are arg(At ) = arg(Ab ) and arg(mg˜ ). apart from the bounds derived by EDM constraints. driving the latter to very small. thereby violating the known upper bounds on the electron and neutron EDMs de and dn : de /e < 0.5 TeV. as |Xt | further increases. mt˜2 is always larger than about 300 GeV.5 TeV. the requirement that the lightest Higgs-boson mass is positive can be used naively to set an upper bound on the phase of At . MSUSY = M µ = 2 TeV.44. For similar reasons. |mg˜ | = 1 TeV. mt˜2 & 100 GeV. experimentally excluded values. In such a scenario. which leads to enhanced CP-violating effects of the CP-odd phases arg(µAt. but holding fixed |At | and |µ|. Specifically.9). Furthermore. we have chosen relatively large values for the stop and sbottom mixing parameters At . one has to worry about the fact that Higgs-boson two-loop contributions to the EDMs [80–83] might still become sizeable..e. One phenomenologically interesting possibility for avoiding the possible CP crisis is to make the first two generations of squarks rather heavy with masses much above the TeV scale [66–71]. reaching a maximum when |Xt |/MSUSY ≈ 2. As was mentioned in the introduction. 300 GeV and large |Xt | values. For the scenarios under discussion.and twoloop EDM terms [72–74] at the level of 10% for the scenario with tan β = 20. For the low-tan β scenario in (5. with mt˜2 . the radiative corrections to the lightest Higgs boson H1 depend crucially on the stop mixing parameter |Xt | = |At − µ∗ cot β|. high values of |Xt | induced by large values of arg(At ) can make the mass of the lightest stop squark t˜2 very low. |At | = |Ab | = 1 TeV. Therefore. the radiatively-corrected H1 -boson mass increases as |Xt | increases. can give rise to observably large contributions to the electroweak precision parameter 1ρ [109. we have taken a gluino mass of 1 TeV in (5. Since these two-loop EDM effects depend almost linearly on tan β. light stop quarks.9).5 × 10−26 cm [55] and dn /e < 0.6 × 10−25 cm [56] at the 2-σ level. |mBe| = |mW e | = 0.48.b) on the Higgs sector.3 TeV. Carena et al.114 M. as well as a common left. i. However.9) As can be seen in (5. the radiative corrections to H1 -boson mass decrease and may even become negative. (5. We believe that this can be achieved without excessive fine-tuning of the CP-violating parameters of the theory.110]. keeping the third generation relatively light with masses of order 0. CP violation may only reside in the third generation. A distinctive feature of the CP-violating SUSY scenario compared to the CP-conserving one is that |Xt | can be increased by varying only the phase arg(At ) from zero to higher values. As was already noticed in [25.9).

1. Numerical estimates of (a) MH1 and MH2 and (b) gH as functions of arg(At ). 1 we give numerical predictions for the two lightest Higgs-boson masses MH1 and MH2 . / Nuclear Physics B 586 (2000) 92–140 115 2 Fig. for two different values of arg(mg˜ ): arg(mg˜ ) = 0 (solid lines) and arg(mg˜ ) = π/2 (dashed lines). for the i ZZ indicated choices of MSSM parameters. Solid lines correspond to arg(mg˜ ) = 0. We first discuss the scenario with tan β = 4. for which the values of the remaining . dashed lines to arg(mg˜ ) = 90◦ . and for the three relevant Hi ZZ couplings squared as a function of arg(At ). In Fig.M. Carena et al.

which gH 1 ZZ the H1 ZZ coupling gets sizeable again. the H2 boson has 2 ≈ 0. arg(At ) = 80◦ –95◦ . Since our interest is to analyze dominant CP-violating effects for a light Higgs sector. the H2 and H3 bosons have significant couplings to the Z bosons. the H2 boson becomes too heavy to allow for 2 . In the CP-conserving limit of the theory (arg(At ) = 0).2 almost channel e+ e− → Z ∗ → H1 H2 . 1(b). especially for the region in 2 is very small.e.9). In Fig. As the phase of At increases. for arg(At ) ≈ 80 degrees. the mass MH1 of the lightest neutral Higgs boson 2 . Carena et al. In fact. discovery via e+ e− → Z ∗ → ZH2 . while gH 1 ZZ a slightly different quantitative dependence on the phase of At . giving rise to larger H1 -boson masses. the stop mixing parameter |Xt | becomes larger. one has |µ| cot β  |At |. Since the vertex corrections are generally small for low or moderate values of tan β. 1 we also present predictions for a gluino phase of 90 degrees (dashed lines).7.6. In this large-tan β scenario. which may be probed at LEP 2 in this year’s a mass of 105–110 GeV and gH 2 ZZ run. First. two important effects take place. even though the coupling gH 1 ZZ 2 ZZ (larger) for larger values of the phase of At . a careful experimental analysis of the parameter region of interest will be necessary to determine whether the two lightest Higgs bosons can be observed for such large mass differences (MH1 − MH2 > 40 GeV) and such small gH1 H2 Z couplings. implying that LEP 2 cannot detect the Higgs boson H1 via gH 1 ZZ + e e− → Z ∗ → ZH1 . Second. / Nuclear Physics B 586 (2000) 92–140 soft supersymmetry-breaking parameters and µ are given in (5. i. 1. the discovery of a Higgs boson at LEP 2 is more challenging. Fig. this situation changes crucially once CP-violating phases become relevant. The lightest Higgs boson H1 acquires a mass below 90 GeV. 2 shows the changes in the predictions for the same choice of parameters as in Fig. as was mentioned above. In addition. with gH 2 ZZ As was discussed in [25]. 95◦ < arg(At ) < 110◦ ..116 M. µ and |At | = |Ab | rescaled by a factor of 2. although MH1 becomes small for arg(At ) > 115 degrees. is approximately is close to 85 GeV. thereby effectively leading to large modifications in the couplings of the Higgs bosons to gauge bosons. Since the squared coupling gH 1 H2 Z 3 ZZ independently of arg(At ). the mass-matrix terms describing the scalar– pseudoscalar mixing are enhanced. but with MSUSY . and the .8. We adopt the scenario given in (5. 2 gets very suppressed. they are expected to induce only small corrections to the Higgs-boson masses and mixings. On the other hand. Thus. with tan β = 20 and MH + = 150 GeV. This 2 2 (gH ) gets slightly smaller last fact is reflected in Fig. Although the H3 boson is too heavy to be detected at LEP 2 in this case.8–0. gH 1 ZZ equal to 0.9). the H1 and H2 bosons may also be searched for in the 2 2 = gH ≈ 0. while the square of the H1 ZZ coupling. we present predictions for a relatively small charged-Higgs-boson mass. We now investigate more quantitatively the predictions of a large-tan β scenario for the Higgs-boson masses and couplings. as can be seen from Fig. for the same range of values of arg(At ). This second effect can be attributed entirely to CP violation. These values of masses and couplings are now excluded by Higgs-boson searches at LEP 2 [54]. 1. well within the capabilities of LEP 2 to test. MH + = 150 GeV. but the H1 ZZ coupling is too small to allow experimental detection through the reaction e+ e− → Z ∗ → ZH1 . 0. This rescaling leads 2 exhibits to a slight increase (decrease) of the Higgs-boson mass MH1 (MH2 ). For larger values of the phase of At . However.

the next-to-lightest Higgs boson H2 is heavy enough to render its search . As Fig. 3(b) 2 on arg(At ). as is seen in Fig. but with µ = 4 TeV. effective stop mixing parameter |Xt | ≈ |At | is almost independent of arg(At ).M. Carena et al. Fig. In contrast to the Higgs-boson masses. Therefore. / Nuclear Physics B 586 (2000) 92–140 117 Fig. |At | = |Ab | = 2 TeV and MSUSY = 1 TeV. 3(a). 1. shows that there is a non-trivial dependence of the squared couplings gH i ZZ Furthermore. the Higgs-boson masses MH1 and MH2 do not exhibit any significant variation as a function of arg(At ). 2.

For similar reasons. 1. As in Fig. 3. Carena et al. through the e+ e− → Z ∗ → ZH2 reaction kinematically inaccessible at LEP 2. for all values of arg(At ). Higgs-boson searches at LEP 2 tend to be more efficient for small values of the At phases. / Nuclear Physics B 586 (2000) 92–140 Fig. As a result. but with tan β = 20.118 M. we find that. the H1 and H2 bosons are rather too heavy to be produced via the e+ e− → Z ∗ → H1 H2 channel at LEP 2. for which 2 the lightest neutral Higgs-boson mass is close to 100 GeV and gH is non-negligible 1 ZZ .

However. 500 GeV. For arg(At ) = 180◦ . but with MH + = 135 GeV. Hence. for the same choice of parameters as in Figs. these two Higgs bosons cannot be produced via e+ e− → Z ∗ → ZH2 or e+ e− → Z ∗ → ZH3 at LEP 2.25]. As was done in [25]. even though the H1 H3 Z coupling is close to 2 2 = gH ) for arg(At ) > 100◦ . the H2 and H3 bosons may still be accessed via e+ e− → Z ∗ → H1 H2 or H1 H3 . However. the H1 V V coupling (with 1 ZZ V = Z. and the H1 ZZ coupling rapidly decreases as the phase of At increases. the two heaviest neutral Higgs bosons become almost degenerate in mass. µ and |At |. it is interesting to mention that large CP-violating scalar-pseudoscalar mixings can lead to observable phenomena of resonant CP violation at high-energy colliders [112–125]. the H1 -boson mass varies approximately between 80 and 65 GeV. Further studies will be necessary to investigate the potential of this production mechanism at the Tevatron. and would have a reach up to MH2 ≈ 120 GeV [111]. The two heaviest neutral Higgs bosons H2 and H3 have masses in the range between 120 and 130 GeV. As was already observed in [24. However. for large At phases. such as H2 . / Nuclear Physics B 586 (2000) 92–140 119 2 (gH ≈ 0. although it becomes lighter with a mass in the range 90–95 GeV. 1 and 2. 4. depending on the phase of the gluino mass. 5(a) with (b) reveals that this last result is almost independent of the common scale factor of MSUSY . since MH3 ≈ MH + .M. The crossing point of the two squared couplings is when arg(At ) ≈ 90◦ . one of the squared couplings goes to 0 and the other to 1. Also. In Fig. MH + = 150 and 300 GeV. 400. Interestingly. 5(a) and (b) show the degree of mass splitting between the two heaviest neutral Higgs bosons H2 and H3 . and may also escape detection via the corresponding channel at the upgraded Tevatron. the degree of mass splitting is not much affected by the value of the gluino phase arg(mg˜ ): the results for arg(mg˜ ) = 90◦ are slightly higher than those of arg(mg˜ ) = 0. so that the H1 Higgs boson. the squared 2 2 2 2 = gH and gH = gH exhibit a cross-over as a function of couplings gH 1 H2 Z 3 ZZ 1 H3 Z 2 ZZ arg(At ). we plot numerical estimates for the same kinematic parameters as in Fig. For the whole range of values of arg(At ). In this case. but for MH ± = 200. It is interesting to present predictions for the neutral Higgs-boson masses and their couplings to gauge bosons for lower values of the charged Higgs-boson mass MH + in the above large-tan β scenario. In addition. the Higgs-pair production of H1 and H3 is unity (gH 1 H3 Z 2 ZZ not kinematically allowed at LEP 2. W ) is rather suppressed. either the H2 or H3 Higgs boson can be tested at the upgraded Tevatron collider provided a total integrated luminosity of 10 fb−1 per detector is available [111]. In Figs. we find that the . even though the H2 and H3 bosons are almost degenerate in the CP-conserving limit of the theory. for two different charged Higgs-boson masses. arg(At ) & 80◦ . the next-tolightest Higgs boson H2 has couplings of order unity to the Z and W bosons. 3. In this vein. they can have a degree of splitting up to 30% for a maximal CP-violating phase arg(At ) ≈ 90◦ . Figs. In this case. 6 and 7. Finally. Carena et al. The comparison of the Fig. we examine the behaviours of the scalar and pseudoscalar parts of the H1 bb coupling as functions of the CP-odd phase arg(At ). will be elusive at LEP 2.3). discovery of such a Higgs boson at the 5-σ level would demand a total integrated luminosity of 30 fb−1 . 300. Present simulations show that a neutral Higgs boson. with MH2 ≈ 180 GeV and a SM-like coupling strength to vector gauge bosons can be tested at the Tevatron collider with a total integrated luminosity of 10 fb−1 .

If we compare the . 3. while the second one will only S S P g P /[(gH )2 + (gH )2 ] occur in CP-violating observables. 4. Higgs-boson [(gH 1 bb 1 bb 1 bb H1 bb 1 bb 1 bb branching ratios are proportional to the first quantity.120 M. / Nuclear Physics B 586 (2000) 92–140 Fig. Carena et al. As in Fig. best way of analyzing such a behaviour is in terms of the CP-even and CP-odd quantities: S S S P P )2 + (gH )2 ] and 2gH g P /[(gH )2 + (gH )2 ]. In other words. but with MH + = 135 GeV. For example. 2gH 1 bb H1 bb 1 bb 1 bb gives a measure of the CP-violating component in the H1 bb coupling.

M. 5. which we already made at the end . for large values of the charged Higgs-boson mass. predictions for MH + = 150 GeV with those for MH + = 300 GeV in Figs. Such a decoupling behaviour of the CP-violating H1 bb component is in agreement with our observation. the higher ones to arg(mg˜ ) = 90◦ . Lower values of the same line type correspond to arg(mg˜ ) = 0. we find that the CP-violating component of the H1 bb coupling reduces in magnitude. Numerical estimates of (a) MH1 and (b) 2|MH2 − MH3 |/(MH2 + MH3 ) as functions of the CP-violating phase arg(At ). 6 and 7. Carena et al. / Nuclear Physics B 586 (2000) 92–140 121 Fig.

which crucially depends on the gluino phase and tan β. 1 of Section 4. / Nuclear Physics B 586 (2000) 92–140 S S S P P Fig. 6. Solid lines correspond to arg(mg˜ ) = 0. From Figs. Numerical estimates of (a) (gH )2 + (gH )2 and (b) 2|(gH )(gH )|/[(gH )2 + 1 bb 1 bb 1 bb 1 bb 1 bb P 2 (gH bb ) ] versus arg(At ). .122 M. 6(b) and 7(b). we see that the impact of the gluino phase on the CP-violating component of the H1 bb coupling is more important for the large-tan β scenario. has a dominant contribution to the H1 bb coupling. Carena et al. This may be attributed to the fact that the radiatively-induced term 1hb tan β. dashed ones to arg(mg˜ ) = 90◦ .

Ab µ and mg˜ µ. This cancellation usually takes place for moderate values of the charged Higgs mass and large values of tan β. There can be a cancellation or a strong suppression of the coupling of the lightest Higgs boson H1 to the bottom quarks. 7. / Nuclear Physics B 586 (2000) 92–140 123 Fig. Such an effect is also present in the CP-conserving case. depending on the magnitude of the CP-violating phases and of the products At µ.M. As in Fig. but with tan β = 20. 6. Carena et al. for specific signs and magnitudes of the above products involving .

124 M. 8(b) illustrates such a cancellation for the CP-violating SUSY model under discussion. Moreover. the H1 bb coupling can be strongly suppressed for arg(At ) = arg(Ab ) ≈ 15◦ . the trilinear terms At. 8.b and the gluino mass. we observe that for the set of SUSY parameters considered in Fig. / Nuclear Physics B 586 (2000) 92–140 S P Fig. Carena et al. Fig. Solid lines are for arg(mg˜ ) = −90◦ and dashed lines for arg(mg˜ ) = 0◦ . and has been discussed in detail in [126. Numerical estimates of (a) MH1 and MH2 and (b) [(gH )2 + (gH )2 ] versus arg(At ).127]. 8. For example.6 TeV and Mt = 0. we . 1 bb 1 bb e e e with soft squark masses: MQ = Mb = 0.

Finally. Moreover. since the lightest Higgs mass is in the mass range that may be within the reach of LEP 2. Fig.155]. Higgs bosons may be copiously produced via a wide variety of processes which depend in many different ways on the couplings of the neutral and charged Higgs bosons both to gauge bosons and fermions [145–153]. m2t˜ − m2t˜ . and hence has more chances to unravel the complete Higgs-boson spectrum of the MSSM with explicit CP violation. the deviations of the results presented in [25] with our results are expected . the two-loop effect originating squared stop masses. However. or into other hadronic modes. hence. in such a scenario. and its production cross section will be SM-like. of the type necessary to handed stop squark and moderate mixing parameter. The LHC has a considerably higher reach than LEP 2 and the upgraded Tevatron collider in the search for heavier Higgs bosons. Carena et al. Most intriguingly. In the latter work. the effective RG-improved potential was expanded up to renormalizable operators of dimension 4. MSUSY 1 2 from the one-loop radiative corrections to the Yukawa couplings of the top and bottom quarks was not taken into account in the computation of the effective potential in [25]. 8(a) shows that the mass of the lightest Higgs boson is of order 105 GeV. will be capable of providing a thorough test of the MSSM Higgs sector and shed light on the possibility of explicit radiative breaking of CP invariance in supersymmetry. The LHC. we have shown that mixing between states with different CP parities can dramatically modify those couplings and. In addition. even in the final run of LEP 2. together with the information gathered from experiments at LEP 2 and the upgraded Tevatron collider. In the case of the CP-violating version of the MSSM under study. |Xt |/M allow the possibility of electroweak baryogenesis [128–144]. the detection of the H1 boson may in principle be impossible. if such a Higgs boson cannot be discovered at LEP 2 via other decay channels. the set of parameters considered in this example also allow for a light righteQ . The expansion was performed in powers of the stop-mass splitting. H1 → b b. relative to the arithmetic average of the 1 2 2 = 12 (m2t˜ + m2t˜ ). the final phase of LEP 2 will leave an open window for electroweak baryogenesis. and arg(mg˜ ) lie in a specific range. Our work provide the basic tool to improve further those studies.M. one should study in detail the capability of LEP 2 to detect such a light H1 boson via its decays into τ pairs. Studies including CP-violating effects on the gluon-fusion production of Higgs bosons at the LHC have already been considered in the literature [154. for moderate values of tan β and for all soft supersymmetry-breaking masses equal to MSUSY . and to perform a complete analysis of the CP-violating effects on the many other Higgsboson search mechanisms available at LHC. practically independent of the CP-violating phase. the main ¯ can be strongly suppressed if the CP-violating phases arg(At ) decay channel. it is interesting to make a comparative analysis between our results and those obtained previously in [25]. / Nuclear Physics B 586 (2000) 92–140 125 have checked that for this same set of parameters the H1 ZZ coupling is almost SM-like. At the LHC. Therefore. Hence. A careful study of the CP-violating phases required for electroweak baryogenesis and the detection capabilities of LEP 2 for alternative decay modes becomes essential for testing this exciting scenario. importantly affect the associated production and decay mechanisms. Nevertheless. To make a conclusive statement on this possibility. This is therefore an extremely interesting example. We shall briefly comment on the enhanced LHC capabilities for Higgs-boson searches [145–153].

9. the deviations will only grow for increasing values of Xt . Carena et al. to be small. the impact of the bottom-mass quantum corrections. which were omitted in [25]. in which case the bottom-mass corrections are small. For larger values of tan β. In fact. Numerical predictions for MH1 and gH as obtained by the present complete RG approach 1 ZZ (solid lines) and the operator-expansion method of [25] (dashed lines). . only in the limit of small values of |µ|. instead. for small values of the stop-mixing parameter Xt = |At − µ∗ cot β|. / Nuclear Physics B 586 (2000) 92–140 2 Fig. are both approaches guaranteed to give comparable numerical estimates.126 M. is significant.

M. In particular. may be found in [156]. for the Higgs-boson masses and their couplings to fermions and to the W ± and Z bosons. we observe larger quantitative differences in the results obtained by the two approaches. we find that the predictions of the two works are in excellent agreement with one another for small values of Xt . even though the qualitative behaviour of the two predictions exhibits a quite analogous functional dependence. Our one-loop RG-improved approach overcomes the limitations present in earlier analyses. in which CP violation is induced radiatively by soft CP-violating trilinear interactions that involve the Higgs fields and the stop and sbottom squarks. the important effects of the one-loop corrections to the quark Yukawa couplings are incorporated in the computation of the Higgs-boson masses and in their respective Higgs-boson couplings to gauge and fermion fields. Carena et al. including all the CP-violating effects as presented in this work. The present work goes well beyond those studies. 9. we show the predictions for the mass of the lightest neutral Higgs boson H1 and its coupling to Z bosons.100. As was discussed above. More specifically. It offers very accurate predictions. for the Higgs-boson masses and for the whole range of the MSSM parameter space in the presence of non-trivial CP-violating phases. for vanishing gluino phase and three different values of the charged Higgs-boson mass: MH + = 150. Also. the present study also includes two-loop leading logarithms associated with QCD effects and t. assuming a moderate mass splitting among the stop squarks. at the same level as the most accurate calculations in the CP-conserving case (implying an uncertainty of order 3 GeV) [51. For the sake of comparison. A Fortran code that computes the Higgs-boson masses and couplings. 200 and 500 GeV.and b-quark Yukawa couplings due to the decoupling of the third-generation squarks . Conclusions We have performed a complete one-loop RG improvement of the effective Higgs potential in the MSSM. within our RG approach. which are induced by one-loop threshold effects on the t. 101].and right-handed stop masses. the qualitative phenomenological features found in [25]. / Nuclear Physics B 586 (2000) 92–140 127 In Fig. For large values of Xt . Earlier studies [24–26] of the neutral Higgs-boson mass spectrum were based on a number of particular assumptions and/or kinematic approximations. 6. and extends the most detailed analysis [25]. even in cases of large stop mixings and/or large hierarchies between the left. as obtained obtained by our RG approach (solid lines) and the operator-expansion method of [25] (dashed lines). we consider the same input parameters as those chosen in Fig. in which the oneloop RG-improved effective potential was expanded up to renormalizable operators of dimension 4. instead. It also contains all dominant two-loop non-logarithmic contributions to the one-loop effective potential. This assumption seemed to impose a serious limitation. The results obtained using the present RG approach confirm.and b-quark Yukawa couplings. however. 1. given that CP-violating effects exhibit an enhanced behaviour for large values of the stop-mixing parameter Xt . our RG approach allows for a rather precise determination of the radiative effects on a generic Higgs-boson mass spectrum.

P. which may remain at the edge of accessibility even during the final LEP 2 run. under Contract W-31-109-Eng-38. CP violation in the lightest Higgs sector becomes relevant for a relatively light charged Higgs boson. even a neutral Higgs boson as light as 60 GeV may escape detection at LEP 2. These CP-violating effects of level crossing in the Higgs sector modify drastically the Higgs-boson couplings to the up. we have also found that the product of the scalar times the pseudoscalar coupling of the lightest Higgs boson H1 to the bottom quarks has a very strong dependence on the gluino phase. . where the heavy MSSM Higgs bosons can be resonantly produced. This product of couplings gives a measure of CP violation in the H1 bb coupling. which can even be of order unity for relatively small charged Higgs-boson masses. strongly depend on the phase of the gluino mass and so introduce new CP violation into the MSSM effective potential at the two-loop level. with MH + . Finally. Carena et al. However. We have explicitly demonstrated that the radiatively-induced CP violation in the MSSM Higgs potential can lead to important effects of mass and coupling level crossing among the three neutral Higgs particles. In this context. We have also studied the effects induced by a non-trivial CP-odd gluino phase. A. In particular. CP violation does not decouple from the heaviest Higgs sector in the MSSM. Department of Energy. Work supported in part by the U. the predictions for the lightest Higgs-boson mass and its couplings to gauge bosons in the above decoupling regime of the theory will be practically identical to the corresponding predictions in the CP-conserving case. the upgraded Tevatron may have the physics potential to explore such CP-violating scenarios at low tan β values. opening up new possibilities for studying enhanced effects in CP-violating Higgs scalar-pseudoscalar transitions at the LHC or future muon colliders [112–125]. δhu. / Nuclear Physics B 586 (2000) 92–140 (as considered in [100.128 M. for MH + = 150 GeV and tan β = 4. In conclusion: the present analysis has shown that the MSSM with explicit radiative breaking of CP invariance constitutes a very rich theoretical framework. The presence of a gluino phase gives rise to small but non-negligible changes in the Higgs-boson mass spectrum and in the couplings of the Higgs fields to the W ± and Z bosons. This decoupling property. unlike the lightest Higgs sector. it is worth stressing that CP violation decouples from the lightest Higgs sector in the large-mass limit of a heavy charged Higgs boson. As a result. Acknowledgements We thank Manuel Drees for useful discussions. These one-loop threshold terms.S.and down-type quarks. For instance. the Tevatron and the LHC. Large radiative effects of CP violation in the Higgs sector of the MSSM can have important phenomenological consequences on Higgs-boson searches at LEP 2.d .25].101] in the CP-conserving limit). which was known to hold for the CP-violating self-energy effect [24. introducing new challenges in the search for fundamental Higgs scalars at LEP 2. the Tevatron and the LHC.d and 1hu. and to the W ± and Z bosons. High Energy Physics Division. which enters the effective potential at the two-loop level. thanks the theory group of Fermilab for the kind hospitality extended to him while part of this work was done. has now been shown to be valid for the CP-violating vertex effects as well. However. 160 GeV.

respectively. 2 2 2 ∂φ2 ∂a2 ∂φ1 ∂a12  2 2   2 2  ∂ m ¯b |hb |2 m2b ¯t |hb |2 m2t ∂ m . Appendix A. and (iii) include the dominant two-loop non-logarithmic corrections to the effective potential. . = = ∂φ2+ ∂φ2− ∂φ2+ ∂φ2− m2t − m2b m2b − m2t  2 2   2 2   2 2  ¯b ∂ m ¯t ¯t ∂ m ∂ m = =− ∂φ1+ ∂φ2− ∂φ2+ ∂φ1− ∂φ1+ ∂φ2−  2 2  ¯b ∂ m |ht hb |mt mb . we list the derivatives of quark masses with respect to Higgs fields. we list analytic expressions pertaining to derivatives of quark and squark masses with respect to their background Higgs fields. We have divided the appendix into three subsections.2) = =− + ∂φ2 ∂φ1− m2b − m2t . Derivatives of background-field-dependent masses In this appendix. = = |hb |2 . Carena et al. = |hb |2 . we give the derivatives related to non-vanishing tadpole contributions:  2  2 ¯b 1 ∂m ¯t 1 ∂m 2 (A. the self-energy-type derivatives involving neutral and charged Higgs bosons may be listed as follows:  2 2  2 2  2 2  2 2 ∂ m ¯b ¯b ∂ m ¯t ¯t ∂ m ∂ m 2 = = |ht | . A. = = ∂φ1+ ∂φ1− ∂φ1+ ∂φ1− m2t − m2b m2b − m2t  2 2   2 2  ∂ m |ht |2 m2b ¯b ¯t |ht |2 m2t ∂ m .1) = |ht | . Here we further (i) perform a more complete RG improvement of the effective potential.1. Quark derivatives First. which may also have an important impact on the H1 bb coupling. . Then. while the next two subsections contain the corresponding expressions for derivatives of squark masses with respect to neutral and charged Higgs fields. v2 ∂φ2 v1 ∂φ1 where the operation · · · denotes that the above expressions should be evaluated in the ground state of the Higgs potential. / Nuclear Physics B 586 (2000) 92–140 129 Note added After completion of the work described here. taking into account the crucial one-loop relation between the charged Higgs-boson and neutral Higgsboson mass matrices. (A. In the first subsection. we saw [157]. These derivative expressions are very useful. which also computes the one-loop RG-improved effective potential of the MSSM with explicit CP violation.M. In the limits where a comparison between the results was possible. we find reasonable agreement with [157]. (ii) develop a self-consistent treatment of the whole MSSM Higgs sector. as they constitute the building blocks of the general one-loop effective potential presented in Section 2.

/ Nuclear Physics B 586 (2000) 92–140 Note that we have not listed derivatives that vanish.  2  et1 (t2 ) g 2 + g 02 1 ∂m = |ht |2 − w v2 ∂φ2 8   1 2 et2 + 1 xt v 2 cos 2β eQ  xt M −M − (+) 2 2 2 2 mt˜ − mt˜ 1 2  + 2|ht |2 Re(µAt ) cot β − |At |2 .2 .2. Carena et al.2 This section contains the derivatives of the field-dependent squark masses m e2t1.  2  eb1 (b2 ) g 2 + g 02 1 ∂m = w v2 ∂φ2 8   1 2 eb2 − 1 xb v 2 cos 2β eQ  xb M −M + (−) 2 2 2 2 m˜ − m˜ b1 b2  − 2|hb |2 Re(µAb ) cot β − |µ|2 . Then.  2  eb1 (b2 ) g 2 + g 02 1 ∂m = |hb |2 − w v1 ∂φ1 8   1 2 eb2 − 1 xb v 2 cos 2β eQ  xb M −M − (+) 2 2 2 2 m˜ − m˜ b1 b2  + 2|hb |2 Re(µAb ) tan β − |Ab |2 .  2   2  et1 (t2 ) et1 (t2 ) 1 ∂m |ht |2 Im(µAt ) 1 ∂m . j = 1. (A. A. eQ  xt M −M + (−) 2 2 2 2 mt˜ − mt˜ 1 2 .2 and m with respect to neutral Higgs fields φ1.3) =− = −(+) v2 ∂a1 v1 ∂a2 m2˜ − m2˜ b1 b2 where the coupling parameters xt and xb are defined after (2.2 and a1. Derivatives of squark masses with respect to neutral Higgs fields e2b1.13).130 M. b1 . are found to be type terms h∂ 2 m  2 2  2 + g 02 et1 (t2 ) ∂ m gw = 8 ∂a12    1 2 et2 + 1 xt v 2 cos 2β + 2|ht |2 |µ|2 . the self-energye2qk /∂ai ∂aj i. We first give the tadpole terms:  2  et1 (t2 ) g 2 + g 02 1 ∂m = w v1 ∂φ1 8   1 2 et2 + 1 xt v 2 cos 2β eQ  xt M −M + (−) 2 2 2 2 mt˜ − mt˜ 1 2  − 2|ht |2 Re(µAt ) tan β − |µ|2 . with qk = t1 . 2. =− = −(+) v2 ∂a1 v1 ∂a2 m2t˜ − m2t˜ 1 2  2   2  eb1 (b2 ) eb1 (b2 ) 1 ∂m |hb |2 Im(µAb ) 1 ∂m . b2 and i. t2 .

3 2 2 mt˜ − mt˜ m2 − m2 1  ∂a1 ∂a2 b2 + g 02 g 2 + g 02 = w 8  ∂a1 ∂a2 2 gw 8    1 2 et2 + 1 xt v 2 cos 2β − 2|ht |2 |At |2 . eQ  xt M −M − (+) 2 2 2 2 mt˜ − mt˜ + (−)  131 |hb t˜1 2 |2 Re(µA b) − m2˜ b2 m2˜ b1 + (−) 2v1 v2 |hb t˜2 |4 Im2 (µA 3 − m2˜ b2 m2˜ b1 b) .M. the non-vanishing CP-violating self-energy terms h∂ 2 m by  e2t1 (t2 ) ∂ 2m  = +(−) ∂φ1 ∂a2 |ht |2 Im(µAt ) m2t˜ − m2t˜  × 1− 1 2 v12   2 et2 + 1 xt v 2 cos 2β eQ − M x M t  2 2 2 m2t˜ − mt˜ 1 2 − 2|ht | Re(µAt ) tan β − |µ| 2  e2b1 (b2 ) ∂ 2m  = +(−) ∂φ1 ∂a2 2  |hb |2 Im(µAb ) m2˜ − m2˜  × 1+ b1 b2 v12 m2˜ − m2˜ b1   2 eb2 − 1 xb v 2 cos 2β eQ −M 2 xb M 2 b2 + 2|hb |2 Re(µAb ) tan β − |Ab |2  e2t1 (t2 ) ∂ 2m ∂φ2 ∂a1  = −(+)  . . eQ  xb M −M − (+) 2 2 2 2 m˜ − m˜ = |hb |2 − ∂a12   e2t1 (t2 ) ∂ 2m = |ht |2 ∂a22 e2b1 (b2 ) ∂ 2m  ∂a22 b1 − e2t1 (t2 ) ∂ 2m  e2b1 (b2 ) ∂ 2m 1 = −(+) 2 1 − m2˜ b2 m2˜ b1 2 = −(+)    2 eb2 − 1 xb v 2 cos 2β + 2|hb |2 |µ|2 . / Nuclear Physics B 586 (2000) 92–140  e2b1 (b2 ) ∂ 2m   2 + g 02 gw 8    1 2 eb2 − 1 xb v 2 cos 2β − 2|hb |2 |Ab |2 . Carena et al. (A. |ht |2 Im(µAt ) m2t˜ − m2t˜ 1 2   .4) e2qk /∂φi ∂aj i are given In addition. eQ  xb M −M 2 2v1 v2 |ht |4 Im2 (µAt ) |ht |2 Re(µAt ) + (−) .

+ g 02   2 eb2 − 1 xb v 2 (1 + 2 cos 2β) eQ  xb M −M 2 − 2|hb |2 |Ab |2 − (+) 2    2 eb2 − 1 xb v 2 cos 2β . the CP-conserving self-energy-type derivatives h∂ 2 m lated to be  e2t1 (t2 ) ∂ 2m ∂φ12  = 2 + g 02 gw 8 + (−) 1 2 − m2t˜ 2 m2t˜ 1   2 et2 + 1 xt v 2 (1 + 2 cos 2β) eQ  xt M −M 2 + 2|ht |2 |µ|2 − (+) v12  ∂ 2m e2b1 (b2 ) = |hb |2 ∂φ12   2 et2 + 1 xt v 2 cos 2β eQ −M xt M  2 3 2 2 m2t˜ − mt˜ 1   2 − 2|ht |2 Re(µAt ) tan β − |µ|2 − − (+) 2 2 gw 8 1 m2˜ b1 − m2˜ b2 v12 2 m2˜ − m2˜ b1   e2t1 (t2 ) ∂ 2m = |ht |2 ∂φ22 . Carena et al. eQ −M 3 xb M 2 b2 + 2|hb |2 Re(µAb ) tan β − |Ab |2 − 2 gw 2 + g 02 8   1 2 et2 + 1 xt v 2 (2 cos 2β − 1) eQ  xt M −M − (+) 2 2 2 2 mt˜ − mt˜ 1 2  − 2|ht |2 |At |2 . (A.5) e2qk /∂φi ∂φj i have been calcuFinally. |hb |2 Im(µAb ) m2˜ − m2˜ b1  × 1− b2   2 eb2 − 1 xb v 2 cos 2β eQ −M 2 xb M 2 v22 m2˜ − m2˜ b1 b2 − 2|hb |2 Re(µAb ) cot β − |µ|2   . . / Nuclear Physics B 586 (2000) 92–140  × 1+   2 et2 + 1 xt v 2 cos 2β eQ − M x M t  2 2 2 v22 m2t˜ − mt˜ 1 2 + 2|ht | Re(µAt ) cot β − |At | 2  e2b1 (b2 ) ∂ 2m  = −(+) ∂φ2 ∂a1 2   .132 M.

one may make use of the .11)). and their analytic form is very complicated.6) A. To calculate the expressions h∂ 2 m be a formidable task. m2t˜ 1 1  . First.  (A. using a mathematical trick which was first applied in [88]. / Nuclear Physics B 586 (2000) 92–140 − (+) v22  e2b1 (b2 ) ∂ 2m = ∂φ22 v22 2 m2˜ − m2˜ ∂ 2m e2b1 (b2 ) ∂φ1 ∂φ2    2 eb2 − 1 xb v 2 cos 2β eQ −M 3 xb M 2 b2 − 2|hb |2 Re(µAb ) cot β − |µ|2   1  12 xt2 v 2 sin 2β + 2|ht |2 Re(µAt ) 2 2 − mt˜ 2   v1 v2 2 et2 + 1 xt v 2 cos 2β eQ −M + (−) 3 xt M 2 2 2 2 mt˜ − mt˜ = −(+) 2 . we notice that h∂ m e2qk /∂φi± i = 0. squark mass matrix M we proceed differently. (2. as a consequence of the fact that the true ground state of the effective potential should conserve charge.   1 2 eb2 − 1 xb v 2 (2 cos 2β − 1) eQ  xb M −M 2 2 m2˜ − m2˜ b1 b2  + 2|hb |2 |µ|2 b1 ∂φ1 ∂φ2 2 2 + g 02 gw 8 − (+)  e2t1 (t2 ) ∂ 2m 2 + 2|ht |2 Re(µAt ) cot β − |At |2 + (−)    2 et2 + 1 xt v 2 cos 2β eQ − M x M t  2 2 3 2 m2t˜ − mt˜ 1  133 2 − 2|ht |2 Re(µAt ) tan β − |µ|2   2 et2 + 1 xt v 2 cos 2β eQ −M × xt M 2  + 2|ht |2 Re(µAt ) cot β − |At |2 . Carena et al. Then. Therefore.    1  1 xb2 v 2 sin 2β + 2|hb |2 Re(µAb ) 2 m2˜ − m2˜ 2 b1 b2   v1 v2 2 eb2 − 1 xb v 2 cos 2β eQ −M + (−) 3 xb M 2 2 2 2 m˜ − m˜ = −(+) b1 b2 + 2|hb |2 Re(µAb ) tan β − |Ab |2   2 eb2 − 1 xb v 2 cos 2β eQ −M × xb M 2  − 2|hb |2 Re(µAb ) cot β − |µ|2 .M. Derivatives of squark masses with respect to charged Higgs fields Here we evaluate the derivatives of the field-dependent squark masses with respect to e2qk /∂φi+ ∂φj− i directly turns out to charged Higgs fields. The reason is that m e2qk are the eigenvalues of a non-trivial (4 × 4) 2 f (cf.3.

134 M.and b-(2 f2t and M ˜ × 2)-mass matrices. e8qk + Ae det M (A. In particular. B12 . C12 and D12 may be expressed in a compact form as follows: hA12 i = 0. t2 . b1 . f2 − Tr M f2 − Tr M C = Tr3 M 3 2  1 2 8 f f6 + BTr M f4 + CTr M f2 . b2 . For our purposes. f+ M hB12 i = −Tr M    f− − Tr M f+ M f− f2t + Tr M f2b Tr M f+ M f2t M hC12 i = Tr M  f− M f+ .8) (A. B = Tr2 M 2   1 f2 Tr2 M f6 − 1 Tr M f4 . A = −Tr M  1 f2 − Tr M f4 . it proves convenient to use a representation in which the columns and rows ‘2’ and ‘3’ of f2 have been interchanged. respectively.7) with f2 . it is sufficient to calculate the derivatives with respect to φ1+ and φ2− . . and Aij = ∂ 2 A/∂φi+ ∂φj− . f− = −  2 (A. qk = t1 . = − Q e2ql m e2qk − m ∂φi+ ∂φj− (A. Bij = ∂ 2 B/∂φi+ ∂φj− . etc.11) M   1 hb µ∗ √ h∗t hb v1 2 Then.9) ql 6=qk with ql .10) → f M− 0 ∂φ2− f2 are the usual t˜. the relevant coefficients A12 . / Nuclear Physics B 586 (2000) 92–140 eigenvalue equation:  f2 − m e2qk 14 = m m6qk + B m e4qk + C m e2qk + D = 0. we find the (4 × 4) matrix M   2   f2   f f+ 2 ∂M 0M f → Mt 0 2 .  f− . f+ =  2 M   1 ht µ∗ √ ht h∗b v2 2     1 2 1 2 ∗ ∗ g |h √ | − h A v t 2 t t   2 w . f2 M − Tr M b . Carena et al. and where M b     1 1 2 2 ∗ ∗ g |h | − h A v √ b 1 b b   2 w . → M fb 0 0 0 M ∂φ1+    f2  ∂M 0 0 . With such a reordering. (A. f D = det M = − Tr M + ATr M 4 to obtain    2 2  eqk e6qk + Bij m e4qk + Cij m e2qk + Dij Aij m ∂ m  .

12) e2qk /∂φ1+ ∂φ2− . we do not display explicitly the functional dependence of M2N on mt . Carena et al. With the help of (A.13) 1 Appendix B. (B.M. / Nuclear Physics B 586 (2000) 92–140 135    f+ M f2 M f− − Tr M f+ M f− − Tr M f− M f+ f2t M f4t M f4 M hD12 i = −Tr M b b    f2 Tr M f+ M f− + Tr M f− M f+ f2t M f2 M f2t + Tr M − Tr M b b    1  f4 f− . 3) and the corresponding (3 × 3) orthogonal matrix O.1)  r = −Tr M2N . f4 f2 f2 2 Tr M f+ M + Tr M t + Tr Mb − Tr Mt + Tr Mb 2 (A. we have  2 2  hB12 im4t˜ + hC12 im2t˜ + hD12 i et1 ∂ m 1 1 . D= 27 4 p= (B. after diagonalizing the RG-improved Higgs-boson mass matrix (M2N )(mt ). m2˜ − m2t˜ m2˜ − m2t˜ m2˜ − m2˜ b1 b1 2 4 hB12 im ˜ + hC12 im2˜ b2 b2   m2˜ − m2˜ m2˜ − m2t˜ b2 b1 b2 2 1 b1 b2 + hD12 i m2˜ − m2t˜ b2 .   1 s = Tr2 M2N − Tr M4N .2) To this end. Higgs-boson masses and mixing angles Here we present analytic expressions for the Higgs-boson masses MHi (mt ) (i = 1. it is straightforward to obtain the derivatives ∂ 2 m More explicitly. =− 2 mt˜ − m2˜ m2t˜ − m2t˜ m2t˜ − m2˜   e2b1 ∂ 2m =− ∂φ1+ ∂φ2−   e2b2 ∂ 2m =− ∂φ1+ ∂φ2− 2 b1 2 1 2 b2 hB12 im4˜ + hC12 im2˜ + hD12 i b1 b1   .12). The mass eigenvalues of the (3 × 3) matrix M2N are then obtained by solving the characteristic equation of cubic order: with x 3 + rx 2 + sx + t = 0. (B. 2  t = − det M2N . (A. = − ∂φ1+ ∂φ2− m2t˜ − m2˜ m2t˜ − m2t˜ m2t˜ − m2˜  e2t2 ∂ 2m ∂φ1+ ∂φ2−  1 b1 1 2 1 b2 hB12 im4t˜ + hC12 im2t˜ + hD12 i 2 2 . q= 27 3 p3 q 2 + . For notational simplicity.3) . 3 2r 3 rs − + t. it proves useful to define the following auxiliary parameters: 3s − r 2 . 2.

 2 2 2 2 O1i + M32 O2i + M33 − MH O3i = 0. If Mij2 .5) satisfy the desired mass hierarchy in accordance with the inequality of (3.5) = − r + 2 −p/3 cos MH 3 3 3 with   q p and 0 6 ϕ 6 π. |z3 |/13 (B. denote the matrix elements of M2N . r < 0. O1i + M12 M11 i  2 2 2 2 M21 O1i + M22 − MHi O2i + M23 O3i = 0.30).31). Carena et al. it is necessary and sufficient to require that D < 0. M31 i More explicitly. s > 0. we may express the three mass eigenvalues of M2N as   p 1 ϕ 2π 2 + . (B. we can diagonalize it by means of an orthogonal rotation O as stated in (3. / Nuclear Physics B 586 (2000) 92–140 To ensure that the three eigenvalues are positive. we have  |x1 |/11 x2 /12 O =  y1 /11 |y2 |/12 z1 /11 z2 /12 where 1i = and (B.4) Imposing these inequalities on the kinematic parameters of the theory. 2. t < 0. MH1 = − r + 2 −p/3 cos 3 3 3   p 1 ϕ 2π 2 − = − r + 2 −p/3 cos . with i. (B. ϕ = arccos − 2 −p3 /27 (B.6) Since the Higgs-boson mass matrix M2N is symmetric. MH 2 3 3 3   p 1 ϕ 2 . P 2 2 k Mik Okj = MHj Oij :  2 2 2 2 − MH O2i + M13 O3i = 0.8) q xi2 + yi2 + zi2 . 3. the elements Oij can then be obtained by appropriately solving the underdetermined coupled system of equations.136 M.7)  x3 /13 y3 /13  . j = 1. Furthermore. one can show [158] that the Higgs-boson mass eigenvalues in (B.

.

.

.

M 2 − M 2 M2 .

.

|x1 | = .

.

22 2 H1 2 23 2 .

.

M32 M33 − MH1 .

2 .

2 2 .

.

M M22 − MH 1 .

. z1 = sx1 .

.

21 2 2 .

9) . M31 M32 (B.

.

.

.

.

.

.

.

. .

.

.

.

y1 = sx1 .

.

.

2 2 .

M23 M21 .

2 2 2 .

M33 − MH M31 1 ..

/ Nuclear Physics B 586 (2000) 92–140 .M. Carena et al.

.

2 2 .

.

M13 M12 .

.

x2 = sy2 .

2 2 2 .

. M33 − MH M32 2 .

2 .

2 2 .

.

M12 M11 − MH .

.

. 2 z2 = sy2 .

2 2 .

M32 M31 .

.

2 .

.

M2 M13 .

x3 = sz3 .

.

2 12 2 2 .

M22 − MH3 M23 ..

.

.

.

M 2 − M 2 M2 .

.

|z3 | = .

.

11 2 H3 2 12 2 .

.

M21 M22 − MH3 137 .

.

.

.

M 2 − M 2 M2 .

.

|y2 | = .

.

11 2 H2 2 13 2 .

.

M31 M33 − MH2 .

2 2 2 .

M M11 − MH 3 y3 = sz3 .

.

13 2 2 M23 M21 .

.

.

.

.

.

.

.

. .

.

.

.

.

.

.

.

.

.

. .

.

.

.

.

..

Christenson. V. D 8 (1973) 1226. S. M. A. Fitch. Winstein.W.. Phys. Rev. B 160 (1985) 117. hep-ph/9806471. A. Fortschr. M. G. Report on the Heavy Flavour Working Group (LEP/SLD/CDF) to the Open Session of the LEP Experiments Committee on March 7th 2000 available at http://delphiwww. Lett. Lett. Lee. Rebelo. Kabir. Phys.S. Phys. B.-L. J. Rev. A.L. Phys. D 10 (1974) 1246. B 419 (1998) 217. Pais. A. Phys. Lett. Lett. 1968. Kong. Phys. Phys. Phys. 5 (1967) 24.J. Neubert. Lett. Mod. 85 (1991) 1. Maekawa. Zwirner. Phys.N. Phys. Branco. J. D 41 (1990) 225. Phys. P. 65 (1993) 1113. Prog. 37 (1976) 657. C. N. Pilaftsis. B 317 (1993) 233.html. A 11 (1996) 4173. A. Phys. Eksp. B 282 (1992) 387.C. Demir. Pis’ma Zh. Lett. Lett. Cronin. A. Phys. E. G. Phys. Phys. Phys. G.E. Phys. B 435 (1998) 88. D 58 (1998) 096010. Alavi-Harati et al. J.10) In (B. Phys. R. L. B 257 (1991) 83. Phys.D. Decker. Y. L. Rev.ch/˜offline/physics_links/lepc. T. B 289 (1987) 1. Rev.M. Fiz. B 553 (1999) 3. Yamaguchi. London. Georgi. W. Grimus. M. JETP Lett. Rev.A. Wagner.K.C. N. R. Teor. T. cern. Lett. Weinberg. 13 (1964) 138. 66 (1991) 1815. KTeV Collaboration. Ellis.D. Liu. Phys. H. B 262 (1991) 54. 37 (1989) 657. Sakharov. Mod. D 60 (1999) 055006. Branco. Hempfling. Lett. Phys. Ridolfi. Okada. 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] J. E. Theor. G. J. F. Türke. O. Yanagida. Pilaftsis. Berberio. Rev. Int.. Rev. Lett. Buras. U. Barr et al. Phys. Turlay. Phys.H.E. Lett. The CP Puzzle.C. Phys. 36 (1988) 201. 83 (1999) 22. Rev. 178 (1989) 147. R. 5 (1967) 32. Nucl. Berger. NA48 Collaboration.10).W. Phys. Rep. Paschos. the abbreviation sx ≡ sign(x) is an operation that simply gives the sign of a real expression x. Lett. Haba. A. F. (B. B 465 (1999) 335.. V. . Lett. B 398 (1997) 305. M.D. 44 (1980) 504. Phys. Rev. Phys. Fanti et al. Fortschr. Lett. Rev. Lett. Wolfenstein. D. Pomarol. H. Rev. NA31 Collaboration. B 287 (1992) 331. Haber. Phys. Academic Press.A. Nucl. Wolfenstein. Lin.

Dimopoulos. Pierce. Wells. J.E. M. Espinosa. G.M. Phys. Lett.E.E. Phys. C. Phys. Phys. D 49 (1994) 4908. B 418 (1998) 98. Phys. J. S. Hoang. Carena. Bock et al. J. Y. Grinstein. Savoy. Lett. Rev. B 423 (1994) 437. S. Ibrahim. D 50 (1994) 2199.B. Lett. Sasaki. Dvali. M. Polchinski. R. C. Barger. Pokorski.html.D.B. Phys. A. Hoang.R. Olive. Johnson.S. K. S. P.H. Mariotti (DELPHI Collaboration). D. Kizukuri. B 389 (1996) 503. E. Phys.R.H.G. Lett. Phys.E. D 45 (1992) 2482. E. N. F. Frigeni. Rev. Grifols. Wise. Lett. W. B 331 (1994) 99. Brignole. D. P. A. Wagner. Ohmann. A 50 (1994) 2960. Phys. F. C.A. Rev. Nath. Nucl. . Roszkowski. D. Lett. P. hep-ph/9910553. del Aguila. M. Lett. Phys. 82 (1999) 904. A. Hempfling. D 58 (1998) 091701. C 75 (1997) 539. Haber. Barbieri. Sasaki.cern. Carena. M. Lett. P. Wagner. Phys. Carena. B 439 (1995) 466 (Erratum). Hempfling.-J. Regan. J. Berger. B 447 (1999) 89. D 60 (1999) 075010. Phys. Lett. Chankowski. Phys.ch/LEPHIGGS/papers/index. P. B 128 (1983) 61. Lett. R. R. Phys.H. Phys.R.M. Nucl. Lett. B 354 (1995) 99. Gavela. B 440 (1998) 96. Rev. Nucl. Rev. Nanopoulos. D 46 (1992) 3025. Lett. M. Mrenna. Wyler. J. Phys. S. Riotto. Rev. Lett.M.cern.. Phys.E. C. S. Binétruy. S.-J. Cohen. Pokorski. Langacker. B 266 (1991) 397. Rev. A. Rev.D. Phys. B 266 (1991) 389. J. Carena. Phys.L. Ellis. Kane. Lett. A. Zhang. M.E. Dudas. 77 (1996) 3728. Rev. D. B 281 (1992) 100. D. Zhang. hep-ph/9906206. M. C. Phys. Phys. Rev. Reports to the Open Session of the LEP Experiments Committee on March 7th 2000.L. Weiglein. Drees.E. J. Z. Nanopoulos. S. N. T. Phys.H. K. M. Phys. R. R. Rev. D 57 (1998) 478. M. M. L. hep-ph/9912236. Rev. Polonsky. B 281 (1992) 284. A. 68 (1992) 3678. A. M. Lett. R. 66 (1991) 2565. Kolda. Wagner. Phys. D 48 (1993) 4280.ch/˜offline/ physics_links/lepc. Wagner. H. Quirós. M. Falk. C. V. B 121 (1983) 321. S. B. H. hep-ph/9907446. B 125 (1983) 393. J.V. B 126 (1983) 71.M.138 [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] [49] [50] [51] [52] [53] [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] M. R. Lett. Srednicki. J. B 258 (1991) 167. C. D.B. Lett. Commins. D. Phys. Chankowski. B. Phys. Grünewald (L3 Collaboration. Phys. Espinosa. Lett. G. D 58 (1998) 111301. D 50 (1994) 7035. Buchmüller. Lett. available from http://www. Phys. Rev.F. Lett.A. Rev. Phys. Rosiek. Quirós. Hempfling.V. Phys. A. Lett. Pomarol. Lett. Phys. Rosiek. L. Dugan. J. Nucl. Nojiri. D 49 (1994) 6173. Phys. Phys. Espinosa. B 436 (1995) 3. Nath. W. A. B 274 (1992) 191. K. McPherson (OPAL Collaboration). D 58 (1998) 019901 (E). Dutta. Phys. Heinemeyer. Phys. B 388 (1996) 588. Mendez. Searches for Higgs bosons: Preliminary combined results using LEP data collected at energies up to 202 GeV.C. E.. P. Caravaglios. Rev. Accomando. available from http://delphiwww. Phys. J. Rev. Papadopoulos. Nucl. B 357 (1995) 573. Lett. B 114 (1982) 231. Pokorski.J. P. Hall. M. Phys.V. Oshimo. Kaplan. Phys. B. LEP working group for Higgs boson searches. Dornan (ALEPH Collaboration). Carena et al. Ferrara. Nucl.G. Rev. Lett. P. G. Phys. B 461 (1996) 407. P.M. Yasui. Phys. M. Srednicki. Hollik. Rosiek. hep-ph/9907422. M. S. Lopez. Haber. B 355 (1995) 209. Casas. Ross. Quiros. M. Phys.html. B 255 (1985) 413. Y.R. Espinosa. M. P. Nanopoulos. DeMille. J. M. Phys. Giudice. / Nuclear Physics B 586 (2000) 92–140 R. Phys. J. B 381 (1992) 66. J. T. Harris et al.M. Phys. Nelson. Arnowitt. G. Kodaira.A. Quirós.

A.). M. Part. 1997.M. Olechowski. Pierce. Rev. Wells.A. B 106 (1976) 292.J. Rev. S. Drees. F. S. D 59 (1999) 115004. Mod. Barr. Phys. J. D 60 (1999) 073003. B 389 (1996) 312. F. Masiero. Phys.F. Pierce. pp. H. Garcia. hep-ph/9711470. W.M. Lett. Paban. J. Results available at http://fnth37. Pokorski. Stockinger. B 271 (1991) 123. K. A. B 242 (1990) 547 (E). A.-Y. L. B 377 (1996) 83. A. A. to appear in Phys.E. Lett. Nucl. Brhlik. B. Jimenez. Porod. S.J. S.W. Rev. J. Matchev. W. A.L.L. Phys. Ellis. Gabbiani. Phys. M. J. Gaillard. Phys. Nanopoulos. Phys.-Y. and references therein. D 39 (1989) 1922. M. J. Carena. M. J. Phys. T. Dykstra. Wagner. D. Pomarol. S. Int. B 544 (1999) 39. Lett. G.S. C. Kalinowski. Ellis.L. Phys. Heinemeyer. Yildiz. D 18 (1978) 1724. Report of the Physics at Run II Supersymmetry/Higgs Workshop. Silvestrini. J. Rev. D. Kane. S. H. Carena. Lett. Lett. hep-ph/9912253. D 59 (1999) 016004.html. A. Phys. Lykken (Eds. J. . 65 (1990) 21.-M. J. J. Lett. G. Zhang.F. M. Jimenez.-F. Rev. P.A. Olechowski.V. Gunion.E. Buras. Weiglein. D. V. Grzadkowski. Pilaftsis. 79 (1997) 982.B. Abel. A. Proceedings of the International Workshop on Quantum Effects in the Minimal Supersymmetric Standard Model. 82 (1999) 900. Rev. Quigg. Flores. Phys. Zee. Haber. H. Hollik. Phys. Bagger. Weinberg. G. Dicus. L. Gunion. Phys. Babu. B. 77 (1996) 4996. G. / Nuclear Physics B 586 (2000) 92–140 [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] [107] [108] [109] [110] [111] [112] [113] [114] [115] [116] 139 M. A. Stremnitzer. C. Barcelona. A. Lett. M. Phys. K. Everett. Choi. S. E. Rev. to appear in Heavy Flavors II. Kalinowski. Rev. J. U. Lett. Phys. R. Phys. Rev. Pokorski. Phys. Sola. D. 81 (1998) 5509. J. L. Rev. 83 (1999) 3972 (E). Carena. in preparation. D 16 (1977) 1519.A. H. W. Grzadkowski. Gabrielli. Lykken. Pilaftsis. D 50 (1994) 7048.S. Chankowski. R. C. Phys.H. J. Chang.L. Rev. Brhlik. Rev. Sep. Cvetiˇc. Bartl. Leigh. Rattazzi. E.M. Choudhury. S. Rev. Phys. J. Rev. Phys. Wagner. Dai. A. Ellis. in press. Sola.E. S. W.M. A. Phys. D 55 (1997) 1611. Phys. Spain. R. B 477 (1996) 321. K. D. Phys. Rosiek. C. Rev. Garisto. M. B 471 (1999) 174. G.T.A. Carena. Carena. Lett. Pokorski. B 504 (1997) 61. Phys. K. Rev.E. Pilaftsis. Nucl. Nierste. March-Russell. Chang. M. B 445 (1999) 331. Phys. Phys. S. C. Keung. M.A. D 60 (1999) 075011. Nucl. Rev. Nucl. Nucl. Wagner. Haber. J. B. Rev. U. Ma. hep-ph/9912223.M. G. Lett. Hall. Glashow. M. B 272 (1991) 313. D. C. Kolda. hep-ph/970344. Rev. J.M.E. D 49 (1994) 6168. A. Ridolfi. hep-ph/0001093. Pilaftsis.E. C. Phys. Phys. Lett. Wagner. A 9 (1994) 1097. Lett. Wilczek. R. Phys. Fermilab 1999.V. Nucl. 83 (1999) 2124. J. Lett. R. J.D. P.fnal. R. hep-ph/9811308. B 491 (1997) 3.M. F. hepph/0001002. Wagner.K.G. hep-ph/9910465. Hempfling. M. Chang. Lindner (Eds). Chankowski. Coarasa. B 237 (1990) 216. D. hep-ph/9908326. Matchev. Pilaftsis. Lee. Lett. R. M. Phys. B. Lett. J. 9 (1978) 50. J. Ellis. D 48 (1993) 5280. S. P. D 55 (1997) 1623. G. M. B 414 (1997) 92.-Y. Phys. 87–102. Rev. Lett. Ioffe. S. Nowakowski. Khoze. Méndez. Phys. Babu. M. Kane. Nanopoulos. Thacker. D. Heinemeyer. D. Nucl. Lett.M. B 426 (1994) 269. Frère. Raychaudhuri. Gajdosik. Rev. Phys. J.A. D. Phys. 63 (1989) 2333. Carena et al. Brignole. Phys. H. Phys. Keung. Nucl. Misiak.A. R. B 389 (1996) 53. hep-ph/9912516. Sarid. Lett.gov/higgs. Sov. C. Carena. Phys. Phys. Zwirner. Good. W. Phys. B.E. Kolda. Weiglein.

Phys. Prog. D. Rev. Phys. Nucl. G.M. Grzadkowski. Osland. Gonzalez-Felipe. B 445 (1998) 178.C. B 537 (1999) 3. M. Rummukainen. J. Nucl. D 60 (1999) 115018.F. Skjold. Joyce. De Carlos. Zeppenfeld. Lin.S. R. CERN-LHCC-9710. C. M. Grzadkowski. Lahanas. Lee. Rev. Rev. O. D. B 545 (1999) 141. Phys. K.D. Osland. Phys. hep-ph/9909315. Nucl. C. Espinosa. D 59 (1999) 116014.H.R. J. Int. G.. Gunion. Lee. Choi. J. A. 101 (1999) 415. A. A. Losada. Phys. B 453 (1995) 3. B 535 (1998) 423. Carena. I. See also Ref. Delepine. M. Baer. S. Grant. Phys. M. Han.E. Theor. Phys. J. Plehn. Cline. Loinaz. Gérard. D 61 (2000) 015003. Phys. S. Lett. M. Phys. Phys. Lett. K.ch/p/pilaftsi/www/.-Z. M. M. Phys. M.f is available at http://home. A. Rev. Lassila-Perini. Rev. A. Phys. S. Funakubo. Wells. Rev. D. H. F. C. Hindmarsh.M. Laine. hep-ph/9909418. E. Moore. B 338 (1994) 71. J. Zeppenfeld.S. Weyers. J. Resconi.Y. Pliszka. Wagner. Quirós. B 350 (1995) 218. D. Huang. J. J. Zarikas. B 503 (1997) 387. C 9 (1999) 413. Phys. Mod. 102 (1999) 389. Schmidt. Rev. J. Kinnunen. Rainwater. Phys. Zhang. John. Z. B 472 (2000) 119. Moretti. Choi. T. Vilja. D.F. / Nuclear Physics B 586 (2000) 92–140 B. Lett. Cline. CMS Thesis-1998-147. Cavalli. Bernreuther. D. B 386 (1996) 183. Panagiotakopoulos. B. P. M. B 503 (1997) 24. Phys.M. Nucl. Dedes. P. Nucl. hep-ph/9812387. M.M. D 56 (1997) 90. Phys. B 454 (1999) 297. M. Rainwater. Phys. Laine.Y. Zhang. Rummukainen.140 [117] [118] [119] [120] [121] [122] [123] [124] [125] [126] [127] [128] [129] [130] [131] [132] [133] [134] [135] [136] [137] [138] [139] [140] [141] [142] [143] [144] [145] [146] [147] [148] [149] [150] [151] [152] [153] [154] [155] [156] [157] [158] M. hep-ph/0003091. Boudjema. S. Phys. B 417 (1998) 79. Rev. J. Poggioli. Eur. Richter-Was. Belanger. D 61 (2000) 099901 (E). Nucl. S. hep-ph/0002287. hep-ph/9907422. Wagner. D. Phys. B 524 (1998) 3. Flesch. CMS Collaboration. L. J. V. Nucl. Theor. Boe. Lett. Acquistapace et al. Sridhar. Brandenburg. J. Wang. J. Phys. . Pliszka. Riotto. J. [50].X. Carena. Ogreid. Bödeker. Phys. Denegri. Quirós. hep-ph/0003122. J. Phys. B. M. Davidson. Nucl.R. Carena et al. A.M. Prog. Losada. S. Rev.G. hep-ph/9904491. T. Lett. S. Phys. Rev. CMS-NOTE 1997-057. B. M. B 475 (1996) 273. R. Phys. Carena. Grzadkowski. M. Espinosa. J. Lett.B. Huber. J. Schmidt. Phys. hep-ph/9908516. Lett. D 60 (1999) 113004. Gunion. K. M. J. B 497 (1997) 387.J. Laine. hep-ph/9910557. Grzadkowski. Nucl. G. D 57 (1998) 4446.A. Phys. W. B 380 (1996) 81. Choi. D 53 (1996) 5834. 13 (1998) 1371. V.G. Riotto. Drees. D 61 (2000) 015006. W. K.E. S. S. Lett.M. Phys. hep-ph/9904348. Lett. Mrenna. P. 80 (1998) 5259. Phys. Gianotti. Dedes. Rev. M. M. Phys. B. Kainulainen. Moretti. hep-ph/9911385. D. K. Riotto. The Fortran code cph. 81 (1998) 3317. J.cern. Froidevaux. Phys. Lee. S. Phys.Y. Wagner.S. hep-ph/0001301. T. C. Lett. A. Spanos. Wells. C. private communication. D. A. X. Phys. J.

gherghetta@cern.nl/locate/npe Bulk fields and supersymmetry in a slice of AdS Tony Gherghetta a. MP ∼ 1018 GeV.ch 1 On leave of absence from IFAE. a model is proposed where the warp factor generates a small (TeV) supersymmetry-breaking scale. Introduction One of the major puzzles in particle physics is the large hierarchy of scales that appear in the standard model (SM). The induced metric on these boundaries differs by an exponential (“warp”) factor. in this theory there are two fourdimensional boundaries. gauge bosons and scalars in this geometry and their implications on electroweak physics.Nuclear Physics B 586 (2000) 141–162 www.8 . revised 23 May 2000. kR & 1. on one of the boundaries. Alex Pomarol a.V.V. Finally. An explicit example of this is the Randall–Sundrum model [2]. All rights reserved. we study the behaviour of fermions. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. We also consider the case of supersymmetry in the bulk. PII: S 0 5 5 0 . E-mail: tony.1 a Theory Division. Thus. and the bulk geometry is a slice of anti-de Sitter (AdS5 ) space of length πR. Universitat Autònoma de Barcelona. is much larger than the electroweak scale (∼ 100 GeV). with the gauge interactions mediating the breaking to the scalar sector. Specifically. CH-1211 Geneva 23. The Planck scale. one can obtain an exponentially small scale. The electroweak scale is generated by a “warp” factor of the induced metric on the brane where the standard model fields live. University of Lausanne. It has recently been realized that the origin of these hierarchies in scale can be related to the presence of extra dimensions with nontrivial spacetime geometries [1.b. Hence. and analyze the conditions on the mass spectrum. accepted 20 June 2000 Abstract Five-dimensional models where the bulk is a slice of AdS have the virtue of solving the hierarchy problem.elsevier.  2000 Elsevier Science B. Barcelona. Switzerland Received 21 March 2000.3 2 1 3 ( 0 0 ) 0 0 3 9 2 . it is not necessary to confine the standard model fields on the brane and we analyze the possibility of having the fields actually living in the slice of AdS. Switzerland b IPT. If the SM ∗ Corresponding author. MP e−πkR . These scenarios may provide an explanation of the fermion mass hierarchy by warp factors. However. for a moderately large extra dimension.∗ . MP and MP e−πkR (where k is the AdS curvature scale of order MP ). This consists of a five-dimensional theory where the extra dimension is compactified on an orbifold. E-08193 Bellaterra.2]. and even larger than the electron or neutrino mass. CERN. and generates two effective scales. 1. All rights reserved. CH-1015 Lausanne.

we will then study the phenomenological consequences of having the SM fields in the 5D bulk. The role of scalar fields in the bulk was first considered in Ref. this model at low energy resembles the ordinary MSSM. [8. such as having all the SM fields in the 5D bulk. then proton decay can be sufficiently suppressed. It is interesting to study other possibilities. For example. Gauge superfields living in the bulk will be responsible for mediating TeV-brane physics to fields that are far away from it. Pomarol / Nuclear Physics B 586 (2000) 141–162 fields live on this boundary. guaranteeing a universal scalar mass spectrum. together with the fermions.8]. [15–18]. the boundary TeV-brane can now be the source of supersymmetry breaking and a window to “Planckian” physics. The complete SM in the bulk was considered in Refs. a small (Planckian) extra dimension was advocated. [4]. For instance. We will show that SM fermions and gauge bosons can propagate in the AdS5 bulk without conflicting with any experimental data. then the electroweak scale can be associated with the effective scale on this boundary. since they can induce large proton decay rates or large flavour violating interactions. and fermions [7] in the bulk. In the model of Ref. but this is not the case for proton decay.6]. These operators can be suppressed by allowing the fermions to be off the TeV-brane. a large (TeV) extra dimension was required that implied a cutoff (string) scale of O(TeV). while at the same time allowing for the fermion masses to be generated. Gherghetta. Nevertheless. We will show that having the SM fermions in the bulk can explain the fermion mass hierarchy by means of the metric warp factor. [2] only gravity propagates in the 5D bulk. This was done either by breaking supersymmetry in the bulk [11–14] or on a distant brane [15–18]. including for example. Similar scenarios were considered in the case of a flat extra dimension. Part of this analysis has been carried out previously in the literature. We will first analyze the non-supersymmetric case. A. where we will study the behaviour of various spin fields in a slice of AdS5 . This is the subject of this article. [11–14]. The inclusion of supersymmetry allows for new alternatives. However.9]. but in this case the resultant phenomenology was similar to the gravity- . In Refs. These operators are known to be a problem in theories with a low (TeV) cutoff scale (including in the original Randall– Sundrum setup [10]). Armed with the above.142 T. We will present here a more compact and general analysis. We will also study the magnitude of higher-dimensional operators. The Higgs field can now be delocalized from the TeV-brane. This delocalization is possible thanks to supersymmetry that protects the Higgs massless mode far from the TeV-brane. This leads to partial success because flavour violation can be suppressed to the desirable level. the case when scalar fields have boundary mass terms. the Higgs field must arise from a Kaluza–Klein excitation that is localized by the AdS metric on the TeV-brane [4. In Refs. In fact. this will be O(TeV) and the brane is referred to as the TeV-brane [3]. If fermions are also distant from the TeV-brane. In particular for kR ∼ 12. while the SM fields are on the TeV-brane. it does have important differences. Subsequent studies considered the SM gauge bosons [5. the gauge interactions can be the messengers of supersymmetry breaking to the quark and lepton sector. As an important new ingredient. We will derive the conditions necessary for supersymmetry and the implications for the particle spectrum. we will consider supersymmetry in the AdS5 slice.

Gherghetta. to obtain the AdS slice. the back-reaction from the bulk fields is neglected. Nevertheless. based on a non-factorizable geometry with one extra dimension. there have been recent unsuccessful attempts at deriving the scenario of Ref. y ∗ = 0 and y ∗ = πR are also the locations of two 3-branes. Nevertheless. the possibility here of mediating supersymmetry-breaking by five-dimensional gauge bosons leads to a very different scenario with respect to the previous examples. the fifth dimension y is compactified on an orbifold. [2] requires a negative tension. which form the boundary of the five- . we introduce the compactification scenario of Ref. S 1 /Z2 of radius R. with −πR 6 y 6 πR. Finally. In Section 3. Our conclusions and final comments appear in Section 5. Finally. especially given the fact that the fundamental theory is most likely to be supersymmetric. Further studies must still be carried out to settle this issue. three important remarks are in order. This will violate the weak-energy condition [27]. Theories that solve the hierarchy problem with other compactifications [1. This will lead us to consider supersymmetric models. we derive the mass spectrum and wavefunctions of the gauge supermultiplet and hypermultiplet. it is conceivable that a setup. equivalently. where particular attention is paid to constraints from electroweak data. [2]. a large volume in the extra dimensions). The possibility of having the SM fermions and gauge bosons in the 5D bulk is studied in Section 4. we study the conditions necessary for supersymmetry in a slice of AdS5 . The article is organized in the following way. In this scenario. if SM fields propagate in these extra dimensions. In Section 2. the TeV-brane in Ref. the couplings become too weak. A. together with their resulting phenomenological predictions. [2] from a five-dimensional supergravity Lagrangian [22–26]. we will not have anything more to add here. As we will see. small neutrino masses). In other words. quantum gravity effects at the TeV). We will assume that the solution of Ref. The orbifold fixed points at. and derive the Kaluza–Klein decomposition of fields with different spins. [2] exists and study its compatibility with supersymmetry. Second. we will assume that the AdS metric is not modified by the presence of these bulk fields. Pomarol / Nuclear Physics B 586 (2000) 141–162 143 mediated models of supersymmetry-breaking proposed earlier with no new low-energy implications. 2. Even though we will be considering fields other than the graviton living in the 5D bulk. While there is no general proof that a description from a more fundamental theory does not exist. Therefore.T. with the new phenomenology of having an extra dimension (Kaluza–Klein states at collider physics. like we are considering. we can combine the good implications of having a large (∼ MP ) cutoff scale (such as small proton decay. is possible without violating the weakenergy condition [27]. Bulk fields in a slice of AdS5 We will consider the scenario of Ref. In particular.19–21] usually require a large compactification radius (or. It is important to point out why the slice of AdS5 is interesting and different from other compactifications [19–21]. [2].

VM . the space–time between the two branes located at y ∗ = 0 and y ∗ = πR is simply a slice of AdS5 geometry. Λ(0) = −Λ(πR) = − Λ . Consequently. 4. 1) with µ = 1. k (4) The four-dimensional reduced Planck scale MP is related to M5 in the following way: MP2 =  M53 1 − e−2πkR . living in a slice of AdS5 given by the metric Eq.144 T. constructed from the five-dimensional metric gMN . The metric solution (2) is valid provided that the bulk and boundary cosmological constants are related by Λ = −6M53 k 2 . . M = (µ. The effective mass scale on the brane at y ∗ = 0 is the Planck scale MP . 5). 1. φ. Ψ . . A solution to the five-dimensional Einstein’s equations. A. The five-dimensional (5D) bulk action. (2). with the 5D coordinates labelled by capital Latin letters. The four-dimensional metric is ηµν = diag(−1. Similarly. Pomarol / Nuclear Physics B 586 (2000) 141–162 dimensional space–time. and we will refer to this 3-brane as the Planck-brane. 2 (1) where M5 is the five-dimensional Planck mass scale and R is the five-dimensional Ricci scalar. the effective mass scale is MP e−πkR . a complex scalar. and Λ(y ∗ ) . respectively. 1. which will be associated with the TeV scale provided kR ' 12. S5 . is given by [2] ds 2 = e−2σ ηµν dx µ dx ν + dy 2 . the classical action for this configuration is given by ZπR Z S=− 4 d x dy √ −g −πR     1 × Λ + M53 R + δ(y) Λ(0) + L(0) + δ(y − πR) Λ(πR) + L(πR) . . and a Dirac fermion. It is also interesting to note that the relations (4) arise in the effective fivedimensional Horava–Witten theory [28] when the Calabi–Yau moduli are fixed. which respects four-dimensional Poincare invariance in the x µ directions. We will now study a U(1) gauge field. at y ∗ = πR. . So far we have only assumed gravity to be present in the five-dimensional bulk. (3) and 1/k is the AdS curvature radius. This 3-brane will be referred to as the TeV-brane. (2) where σ = k|y| . Gherghetta. The cosmological constants in the bulk and boundary are Λ. k (5) From the form of the metric solution (2). has kinetic energy and mass terms given by .

the vacuum configuration of the scalar field can be thought of as an infinitely thin domain wall. for the metric defined by Eq. Therefore.T. We then have that Ψ¯ Ψ is odd under the Z2 symmetry. The scalar mass term is even under the Z2 symmetry and can be either a bulk or boundary term. where the arbitrariness of the sign can only be determined by the fermion interactions. b and c are arbitrary dimensionless parameters. On the other hand. This is very similar to the background 3-form field in the dimensional reduction of Horava– Witten theory [28]. (2). The equations of motion for the gauge. we have 1 dσ and Γ5 = 0 . dy   d2 σ (9) σ 00 = 2 = 2k δ(y) − δ(y − πR) . since Ψ¯ Ψ vanishes there. and δ(y) is the Dirac delta-function. Gherghetta. This nontrivial transformation of the fermion mass parameter implies that mΨ must arise from an underlying scalar field that receives a vacuum expectation value (VEV) with an odd “kink” profile. In particular. dy The step function (y) is defined as being 1 (−1) for positive (negative) y. the mass parameters of the scalar and fermion fields can be parametrized as 2 m2φ = ak 2 + bσ 00 . the Z2 transformation is given by Ψ (−y) = ±γ5 Ψ (y). . For the fermion Ψ . The generalization to different masses for different boundaries can be easily obtained from the analysis here. Pomarol / Nuclear Physics B 586 (2000) 141–162 Z S5 = − d4 x  × Z dy 145 √ −g  1 2 2 M 2 2 ¯ ¯ Ψ Ψ . In other words. a Majorana mass term for the fermion can be added. and consequently the Dirac mass parameter mΨ must also be odd. and the derivatives are defined as dσ = k(y). either in the bulk or on the boundaries. where e α The gamma matrices. where ΓM is the spin connection. this is required by supersymmetry. Under the Z2 parity the boson fields can be defined either odd or even (depending on their interactions). This term will necessarily break supersymmetry. respectively. given by σ0 = 2 We are assuming that the magnitude of the boundary mass for the scalar is the same on the two boundaries. F + |∂ φ| + i Ψ γ D Ψ + m |φ| + im M M Ψ MN φ 4g52 (6) where g = det(gMN ). scalar and fermion fields are. note that a Dirac mass term cannot be induced on the boundaries. γ5 ) are defined in curved space as γM = eM α M is the funfbein and γα are the Dirac matrices in flat space. (8) where a. Furthermore. γM = (γµ . mΨ = cσ 0 . but with opposite sign. A. As we will see. and we will comment on this in Section 4. (7) Γ µ = γ5 γµ 2 dy α γ . The U(1) gauge field strength is FMN = ∂M VN − ∂N VM and in curved space the covariant derivative is DM = ∂M + ΓM .

ak 2. in the gauge where V = 0 together with the constraint 5 ∂µ V µ = 0.  √ −g g MN ∂N φ − m2φ φ = 0. (10) Using the metric of Eq. (14). Φ(x µ . The corresponding eigenfunctions fn . (13) −πR Thus from Eq. s = {2. we have introduced the exponential factor. y) = 0.  g MN γM DN + mΨ Ψ = 0. 2. For the fermion. one can write a general second-order differential equation 3  2σ µν  (11) e η ∂µ ∂ν + esσ ∂5 (e−sσ ∂5 ) − MΦ2 Φ(x µ . 4. M Φ 00 without the boundary mass terms proportional to σ . (13). an expression for the normalization constant Nn is      ZπR 1 mn σ mn σ 2 2 e + bα (mn )Yα e dy e2σ Jα . These terms will be considered later when we impose the boundary conditions on the solutions. α = (s/2)2 + M Φ Using the orthonormal condition Eq. which takes into account the spin connection. The Bessel functions Jα and Yα are of order q b 2 /k 2 . ak 2 + bσ 00 . A. y) = √ 2πR n=0 (12) where the Kaluza–Klein modes fn (y) obey the orthonormal condition 1 2πR ZπR dy e(2−s)σ fn (y)fm (y) = δnm . where Φ = {Vµ . (15) fn (y) = Nn k k where mn is the mass of the Kaluza–Klein excitation Φ (n) and the normalization factor Nn and coefficients bα (mn ) are constants.R = ±γ5 ΨL. 1} and MΦ2 = {0. c(c ± 1)k 2 }. (14) −e ∂5 (e−sσ ∂5 ) + M b2 = {0.R ). are      esσ/2 mn σ mn σ e + bα (mn ) Yα e Jα . (11). Pomarol / Nuclear Physics B 586 (2000) 141–162 1 ∂M √ −g  √ ∂M −g g MN g RS FNS = 0. and we have explicitly separated the left and right components (defined by ΨL. In Eq. (14). e−2σ . the Kaluza–Klein eigenmodes fn (y) satisfy the differential equation   sσ bΦ2 fn = e2σ m2n fn . we have defined the mass-squared parameter. . obtained by solving Eq. φ. (2). e−2σ ΨL. c(c ± 1)k 2 ∓ cσ 00 }. Kaluza–Klein decomposition Let us decompose the 5D fields as ∞ 1 X (n) µ Φ (x )fn (y). as in the graviton case [2]. (16) Nn = πR k k 0 3 For the gauge field we work. Gherghetta.1.146 T.R }.

The Kaluza–Klein masses mn and the coefficients bα (mn ) are determined by imposing the boundary conditions on the solution (15).T. then we have  fn (y) → fn |y| . the parameter s. Gherghetta. A.1. does not depend on the particular type of field. Pomarol / Nuclear Physics B 586 (2000) 141–162 147 In the limit mn  k and kR  1 an approximate expression for the normalization constant is   eπkR mn πkR eπkR/2 Jα e .1. (17) 'p Nn ' √ k 2πkR π 2 Rmn Note that in this limit the normalization constant. Thus we must impose  . σ 00 . has delta function terms. 2. The appropriate boundary condition depends on whether the field is odd or even under the orbifold Z2 symmetry. (11). MΦ2 in Eq. Even fields If the 5D field is even under the Z2 symmetry.. Nn . (18) The derivative of fn must be either discontinuous (proportional to (y)) or continuous on the boundaries depending on whether or not the mass-squared parameter. i.e.

.

dfn − rσ 0 fn .

.

. k The continuity of fn at the boundaries implies that . Imposing Eq. the larger the value of n. Odd fields If the 5D field is odd under the Z2 symmetry. . and α > 0 are   α 3 πke−πkR .R }. e−2σ ΨL. then we have fn (y) →  σ0 fn |y| . b.1. = 0. . k Yα ( k ) bα (mn ) = bα (mn eπkR ). ∓c} for Φ = {Vµ . (20) (21) These two conditions determine the values of bα and mn . 2.πR where the parameter r has the values r = {0. 2. (22) mn ' n + − 2 4 The approximate mass formula (22) becomes more exact. In the limit that mn  k and kR  1 the Kaluza–Klein mass solutions for n = 1. (19) dy 0.2. φ. respectively. . (19) gives rise to the two equations bα (mn ) = − (−r + s/2)Jα ( mkn ) + (−r + s/2)Yα ( mkn ) + mn 0 mn k Jα ( k ) mn 0 mn .

fn .

and consequently (23) (24) .πR = 0.0.

an approximate solution for the Kaluza–Klein tower in the limit that mn  k and kR  1 is   α 1 πke−πkR . Supersymmetry in a slice of AdS5 An AdS space is compatible with supersymmetry [29. . 1)). Yα ( mkn ) bα (mn ) = bα (mn eπkR ). AdS supersymmetry requires that different fields belonging to the same supersymmetric multiplet have different masses. . . the graviphoton B and The on-shell supergravity multiplet consists of the funfbein eM M i (i = 1. 2). The supermultiplet mass-spectrum has been derived in Ref. . 1. P .1. 3. Note that the difference in the approximate mass formulae between the even and odd mass solutions is π/2. [31] for a five-dimensional AdS space. Thus. in contrast to the flat space case. (25) (26) In this case one can check that the derivative of fn is continuous on the boundaries and does not lead to further conditions. and α > 0. In a slice of AdS5 . Gherghetta. In the next section. satisfy ψ¯ i γ M · · · γ P χ j = − ik  j l χ¯ l γ P · · · γ M ψ k . where Ψ 3. Here we will extend the analysis to the case where the fifth dimension is compactified on an orbifold S 1 /Z2 . the supergravity Lagrangian has extra terms proportional to the cosmological constants: 4 We will follow the conventions of Ref.148 T. 1. . ψ i and χ i . [32] (except for our metric convention η µν = (−1. (27) mn ' n + − 2 4 where n = 1. The index i labels the fundamental two symplectic-Majorana 4 gravitinos ΨM representation of the SU(2) automorphism group of the N = 1 supersymmetry algebra in five dimensions. As in the even case. we will see that the supermultiplet mass-spectrum in a slice of AdS5 will have interesting phenomenological implications. This is because the momentum operator. where the two symplectic-Majorana spinors. in AdS space does not commute with the supersymmetric charges and P 2 is not a Casimir operator. Pomarol / Nuclear Physics B 586 (2000) 141–162 bα (mn ) = − Jα ( mkn ) . 2. However. γ µ ∂µ Ψ (28) bL − ∂5 Ψ bR + mΨ Ψ bR = 0. the even and odd fermion fields must also satisfy these equations. γ ∂µ Ψ (29) µ b = e−2σ Ψ .30]. A. Supergravity multiplet α . In the case of fermion fields one should note that the even and odd functions are related by coupled first-order differential equations bR + ∂5 Ψ bL + mΨ Ψ bL = 0.

(32) i = 0.2. we have defined the Z2 transformation of the symplectic-Majorana spinor as i = DM ηi + δΨM ηi (−y) = (σ3 )ij γ5 ηj (y) . (31) this leads to the Killing spinor equation σ0 γM (σ3 )ij ηj . (35) to be invariant under the supersymmetric transformations . in the following way σ0 γM (σ3 )ij ηj . Therefore. (35) × M M λ 3 MN Σ 2g52 In flat space. By requiring Eq. where fields in the same supermultiplet must have different masses [31]. The condition that the AdS5 background does not break supersymmetry is δΨM and using Eq. 3.T. λi . (34) This condition implies that only half of the 5D supersymmetric charges are preserved. Σ) where VM is the gauge field. Without loss of generality. For simplicity we will consider a U(1) gauge group. In order to respect supersymmetry in (30). the boundary terms require an extra condition to be satisfied. with respect to the supergravity case with no cosmological constant. However. (33) 2 In a non-compact five-dimensional AdS space this condition is always fulfilled. supersymmetry requires that mΣ = mλ = 0. A. (30) we do not show the dependence on BM . Pomarol / Nuclear Physics B 586 (2000) 141–162 Z 149 Z √ 1 d4 x dy −g 2     3 0 ¯ i MN Λ 00 3 i MNR i ij j ¯ DN ΨR − i σ ΨM σ (σ3 ) ΨN + 2Λ − 2 σ . after compactification. γ N ]/2. This is to be contrasted with an AdS5 background. the supersymmetric transformation of the gravitino must be changed. × M5 R + iΨM γ 2 k S5 = − (30) P where γ MNR ≡ perm (−1)p γ M γ N γ R /3! and σ MN = [γ M . namely DM ηi = − γ5 ηi = (σ3 )ij ηj . −1) and the symplectic-Majorana spinor ηi is the supersymmetric parameter. and Σ is a real scalar in the adjoint representation. In Eq. in the orbifold compactification. The action has the form Z Z √ 1 4 d x dy −g S5 = − 2   1 2 2 i M i 2 2 i ij j ¯ ¯ λ F + (∂ Σ) + i λ γ D λ + m Σ + im (σ ) λ . Gherghetta. one has in 4D a N = 1 supersymmetric theory instead of N = 2. Vector supermultiplet The on-shell field content of the vector supermultiplet is V = (VM . since in the AdS5 background we have BM = 0. λi is a symplectic-Majorana spinor. (31) 2 where σ3 = diag(1.

Y0 ( mkn ) Y0 ( mkn eπkR ) (39) Thus. (38). This implies that the action (42) in the background metric (2) is independent of the y coordinate. y) = √ 2πR e3σ/2 1 (0) (41) λL (x) + · · · . . . A. 2. 2 The Kaluza–Klein decomposition can be obtained easily from the analysis of the previous sections. δΣ = η¯ i λi . Pomarol / Nuclear Physics B 586 (2000) 141–162 δVM = −iη¯ i γM λi . (40) mn ' n − 4 and is consistent with the mass formulae (22) and (27). the approximate solution for the mass of the Kaluza–Klein modes with n = 1. (36) where ηi satisfies the Killing equation (33) and the condition (34). From Eq. even though the values of α are different. λ1L (x. while Σ and λ2L are odd. we find as expected that all fields of the supermultiplet have identical Kaluza–Klein masses.150 T. the five-dimensional masses of the scalar and spinor fields in the vector supermultiplet must be m2Σ = −4k 2 + 2σ 00 . Consequently. one finds that in a slice of AdS5 . Vµ (x. 1 (37) mλ = σ 0 . . b = 2. and c = . If we assume that Vµ and λ1L are even. while α = 0 for Σ and λ2L . Gherghetta. y) = √ 2πR Notice that the zero modes have the proper y dependence to make the 4D kinetic terms of these modes in the action ZπR √ 1 dy −g − d x 2πR −πR   1 ν µ (0) 1 (0) (0) 3σ ¯ 1 (0) µ λ ∂ V (x)∂ V (x) + ie (x)γ ∂ λ (x) . The even fields Vµ and λ1L will have a massless mode with the following y dependence: 1 Vµ(0) (x) + · · · . . (37) and Eq. In fact. then the Kaluza–Klein masses are determined by the equation J0 ( mkn ) J0 ( mkn eπkR ) = . 2 (38) Using Eq. (8) we have 1 a = −4.  δλi = −σ MN FMN + iγ M ∂M Σ ηi − 2iσ 0 Σ(σ3 )ij ηj . is given by [6]   1 πke−πkR . the . × ν µ µ L L 2g52 Z 4 (42) invariant under the conformal transformation gµν → e2σ gµν . we find that α = 1 for Vµ and λ1L .

the massless sector from Vµ and λ1L forms an N = 1 supersymmetric vector multiplet. Pomarol / Nuclear Physics B 586 (2000) 141–162 151 zero modes are not localized by the AdS space–time and they behave like zero modes in flat space. A. Furthermore. their couplings to the two boundaries are of equal strength. 3. Therefore. Hypermultiplet The hypermultiplet consists of H = (H i . The action has the form   Z Z . The odd fields Σ and λ2L do not have massless modes because this is not consistent with the orbifold condition.T. Ψ ) where H i are two complex scalars and Ψ is a Dirac fermion. Gherghetta.3.

.

2 √ S5 = − d4 x dy −g .

∂M H i .

then the Kaluza–Klein masses (identical for both the even and odd modes) are determined by the equation J|c+1/2| ( mkn ) J|c+1/2| ( mkn eπkR ) = . . H 1 (x. (22) for the even modes. 2. Y|c+1/2| ( mkn ) Y|c+1/2| ( mkn eπkR ) (47) In fact using Eq. (46) a = c2 ± c − 4 2 Thus. k + mH 1.2 = c ± c − 4 2 (45) mΨ = cσ 0 . +iΨ¯ γ M DM Ψ + m2H i |H i |2 + imΨ Ψ¯ Ψ . and Eq. we find that α = |c + 1/2| for H 1 and ΨL . we can again see that the odd and even Kaluza–Klein modes. n = 1. while H 2 and ΨR are odd. (43) Invariance under the supersymmetric transformations √ δH i = 2 i ij η¯ j Ψ . (27) for the odd modes. and α = |c − 1/2| for H 2 and ΨR . 2 (44) where ηi satisfies the Killing equation (33) and the condition (34). where c remains an arbitrary dimensionless parameter. y) = √ 2πR N0 e(2−c)σ (0) ΨL (x) + · · · . mn ' n + − 2 2 The massless modes of H 1 and ΨL are given by e(3/2−c)σ 1 (0) H (x) + · · · . . Assuming that H 1 and ΨL are even. requires that the fivedimensional masses of the scalars and fermion satisfy     15 2 3 2 2 ∓ c σ 00 .   √ 3 δΨ = 2 γ M ∂M H i  ij − σ 0 H i (σ3 )ij − mΨ H i  ij ηj . ΨL (x. y) = √ 2πR N0 (49) . (8) we can identify 3 15 and b = ∓ c . Using Eq. . have identical masses which are approximately given by   c 1 (48) πke−kπR .

A non-supersymmetric model can be constructed. the massless modes are not localized by the AdS space and will couple to the two boundaries with equal strength. It is given by . it would appear that the hierarchy problem is again reintroduced. the Kaluza–Klein excitations of the gauge bosons can induce four-fermion interactions that are severely constrained by the experimental data [33–42]. even though the scalar zero mode becomes heavy. However. then the gauge bosons must necessarily reside in the bulk. This dire conclusion is not in fact realized. In this case the localization becomes more effective the larger the value of c. This is effectively like having the Higgs field on the TeV-brane. due to the fact that chiral symmetry protects the massless mode. Gherghetta. where the kinetic terms are again independent of the y coordinate. As c becomes more negative. the mass of the lightest nonzero Kaluza–Klein mode is still of order the TeV scale. 2πkR(1/2 − c) (50) When c = 1/2. the massless sector of the hypermultiplet (from H 1 and ΨL ). the SM fields were assumed to be confined on the TeV-brane. and therefore the Higgs can be associated with this lightest Kaluza–Klein state. but eventually we will see that many more interesting possibilities exist for supersymmetric models. we have N0 = 1 which signals the conformal limit. the AdS space will localize the massless modes towards the boundary y ∗ = πR. since this is not consistent with the orbifold condition. Thus. depending on the value of the dimensionless mass parameter c. the localization at the y ∗ = πR boundary becomes increasingly more effective. Therefore. The odd fields do not have a massless mode. Thus. 4. forms an N = 1 supersymmetric chiral multiplet. the massless modes will be localized towards the y ∗ = 0 boundary. the massless mode of the bulk scalar field will receive radiative corrections of order the Planck scale. Thus. We will first consider models with bulk fields but without supersymmetry. The question naturally arises then of whether the SM fields can actually live in the bulk. If the fermion fields are located in the bulk.152 T. 4. as in the case of the vector multiplet.1. let us consider the five-dimensional gauge coupling between the bulk gauge boson and fermion. even without supersymmetry. [2]. A. The standard model in the bulk We have seen that the massless modes of a 5D scalar or fermion field in the AdS slice can be localized at either of the two boundaries. Pomarol / Nuclear Physics B 586 (2000) 141–162 where N02 = e2πkR(1/2−c) − 1 . On the other hand when c > 1/2. as in flat space. without supersymmetry. This continues to be true for massless fermions. For values of c < 1/2. In this case. because the nonzero Kaluza–Klein scalar modes are localized near the TeV-brane and consequently have TeV scale masses. Phenomenological implications In the original proposal in Ref. To study these constraints in our case. as explained in the vector multiplet case.

and c is defined in Eq. When c is large and negative. A. 153 (51) where g5 is the five-dimensional gauge coupling. the fermion is localized near the TeV-brane and the ratio g (1) /g approaches the √ (1) asymptotic limit g /g ' 2πkR ' 8. where the wavefunction of the Kaluza–Klein gauge bosons is constant. . Gherghetta. the coupling quickly becomes universal for all fermion masses. g (n) /g. y)Ψ (x. 1. At c = 0. At the conformal limit c = 1/2. y). Pomarol / Nuclear Physics B 586 (2000) 141–162 Z Z d4 x dy √ −g g5 Ψ¯ (x. (52) k k √ where g = g5 / 2πR is the four-dimensional gauge coupling. agrees with the case of a Planck-brane fermion [6]. leading to a restrictive lower bound on the first excited Kaluza– Klein mass scale m1 & 20 TeV [5]. Using the expression for the zero-mode fermion (49). the bound from electroweak precision experiments on the first excited Kaluza–Klein mass is (n) 1 − 2c Fig. we show the ratio g (n) /g. In Fig. the coupling completely vanishes due to the fact that the fivemomentum is conserved in this limit. for n = 1 (dotted line). In this region the asymptotic limit g (1) /g ' −0. for n = 1. as a function of c.4. 3. As the value c of the bulk fermion mass increases. the gauge coupling of a gauge boson Kaluza–Klein mode n to the zero-mode fermions is   ZπR k dy e(1−2c)σ g = g (1−2c)πkR e − 1 Nn 0      mn σ mn σ e + b1 (mn )Y1 e × J1 . y)iγ µ Aµ (x. For c > 1/2.T. n = 2 (solid line) and n = 3 (dashed-dotted line). (8). which corresponds to a TeV-brane fermion [5. as a function of the dimensionless fermion mass parameter c. because the fermion zero mode couples less strongly.6].2. 1. we find that the bound considerably weakens. This is because the fermions are now localized near the Planckbrane. If all fermions have equal five-dimensional masses. The ratio of the gauge couplings. [8]. we recover the massless fermion case considered in Ref. 2. In this region the first excited Kaluza–Klein mode of the gauge boson couples strongly to the fermion zero-mode.

ΨiL and Ψj R . Note that in deriving (55). and thus gives negligible contributions to any four-fermion operator (unless the fermion is on the TeV-brane). Here we will analyze it for the quarks and leptons. A. Pomarol / Nuclear Physics B 586 (2000) 141–162  m1 & 2. 4. then the fermion zero modes will develop an exponential profile which gives rise to the following four-dimensional Yukawa couplings (5) λij = λij k NiL Nj R e(1−ciL −cjR )πkR . the lower bound is fairly innocuous. will depend on the values of ciL and cj R . Thus we see that in the region where c & 1/2. which in some sense represents . λij . (55) where 1/2 − ciL 1 . The fermion mass hierarchy problem was not addressed there because the fermions were confined to the brane and the Yukawa interactions were conformally invariant. [7]. Our setup is as follows. was used to solve the gauge hierarchy problem by generating the TeV scale from the Planck scale. The precise flavour structure of the Yukawa coupling matrix. a similar use of the warp factor can be used to explain the fermion mass hierarchies. (53) where m1 ' 2. the warp factor e−σ . g (2) /g. while for simplicity the Higgs H (x) will be localized on the TeV-brane. The coupling of the higher-order Kaluza–Klein modes (n > 1) is always smaller than the coupling of the first excited state. and ΨiR (x. Gherghetta. (55). continue to remain c-independent for c & 1/2. y). We can consider two simple limits. We see then from Eq. As can be seen in Fig. For neutrino masses this idea has been considered in Ref.c. y). δ(y − πR) Z  (0) (0) (54) ≡ d4 x λij H (x) Ψ¯ iL (x)Ψj R (x) + h. the Higgs must be rescaled by an amount H (x) → eπkR H (x). If each fermion field has a bulk mass term parametrized by ciL and cj R . that exponentially-small Yukawa couplings can be generated for values of ciL and cj R slightly larger than 1/2.2. Fermion mass hierarchy In the original scenario [2]. if we allow the fermions to live in the bulk. in order to obtain a canonically normalized kinetic term.c. (5) where λij are the five-dimensional Yukawa couplings and λij define the effective four(0) (0) dimensional Yukawa couplings of the zero modes.5 ke−πkR . 1.1 g (1) g  TeV.154 T. the ratio of the couplings. ≡ (1−2c )πkR 2 iL e −1 NiL (56) and similarly for Nj R . y) + h. + · · · . the five-dimensional action will have the Yukawa coupling term Z Z  √ (5) d4 x dy −g λij H (x) Ψ¯ iL (x. and g (3) /g. However. we have two five-dimensional Dirac fermions ΨiL (x. For each fermion flavour i. Thus. y)Ψj R (x.

In this way we see that the mass of the lightest fermion states of the SM can be naturally generated for values of ciL > 1/2. in the limit that kR  1 and ciL > 1/2 become λ k p ciL − 1/2 e−(ciL −1/2)πkR . In this case the values of the mass parameters must be close to the conformal limit (ciL ' 1/2). (55) simplifies to (for kR  1) −2(ciL−1/2)πkR . It is easy to understand the behaviour of the curves in Fig. (55). for simplicity. Suppose first that we have a left–right symmetric theory such that ciL = ciR . where the diagonal elements of Eq.46). This can clearly be seen in Fig. (55). We have assumed.64 (assuming kR ∼ 12. .T. 2. we can suppose that the right-handed fermions all have cj R = 1/2. (58) λii ' √ ii 2πkR In this case the exponential factor is more dominant compared to the left–right symmetric assumption. Gherghetta. 2 the exponential damping of Eq. Pomarol / Nuclear Physics B 586 (2000) 141–162 155 the two possible extremes for the flavour structure. that λ(5) ij k ∼ 1. a top Yukawa coupling λt ∼ 1 can be obtained for ct L = −1/2. Eq. The heavier fermions (c. Alternatively. In Fig. the effective Yukawa couplings. 2. in this left–right symmetric limit. and must be localized on the TeV-brane (ct L . The top quark having a comparable mass to the VEV of the Higgs. Again we must stress that this scenario does not predict the fermion mass spectrum (which would correspond to predicting the values of ciL and (5) Fig. For ciL > 1/2. −1/2). we clearly see in Fig. for the case ciL = ciR (solid line) and ciR = 1/2 (dotted line) . λii ' λ(5) ii k (ciL − 1/2) e (57) An electron Yukawa coupling λe ∼ 10−6 can be generated for ceL ' 0. which is located on the TeV-brane. 2 we have plotted the diagonal element of the Yukawa matrix. since the exponential factor in (55). τ. the Yukawa couplings quickly become exponentially-small since fermions are localized near the Planck-brane and have an exponentially small overlap with the Higgs. The effective four-dimensional Yukawa coupling λii as a function of ciL . 2. where fermions are not localized on either of the two branes. Now. A. needs the largest overlap of all. For ciL > 1/2. On the other hand. b) require a larger overlap with the Higgs and cannot be confined near the Planck-brane. disappears for ciL < 1/2.

of course. Gherghetta. (d s¯)2 /M42 needs to be suppressed by at least an effective four-dimensional mass scale M4 ∼ 1000 TeV.2−2. for ci & 1 fermion masses are much too small (see Eq. and depending on the values of ci can lead to an exponential suppression of mass scales. It is important to realize. However. this of course leads to proton decay problems. M4 ranges from the TeV scale (for ci ' 1/2) to MP (for ci ' 1).5 TeV in the case of Eq. If the fermion fields are instead in the bulk. then we can ask whether effective higherdimensional operators are suppressed. [44–46]. each fermion zero mode has a wavefunction proportional to e(2−ci )σ . however. For the values 1/2 . = M42 M53 Ni Nj Nk Nl (4 − ci − cj − ck − cl ) (60) and Ni are defined in Eq. In the absence of any discrete symmetries. Higher-dimensional operators When fermions are confined on the TeV-brane. higherdimensional operators are suppressed by the TeV scale and not the Planck-scale [10]. the Kaluza– Klein gauge-bosons couple almost universally to any fermion with ciL & 1/2. but in this case the FCNC constraints are much weaker [42]. For example. Unfortunately. This idea was first discussed in the context of string theory by Ref. The third-family fermions do not have a universal coupling.156 T. we can only obtain the necessary suppression for proton decay if the mass parameters satisfy ci & 1. This would also seem to be a problem for our scenario. This is the case for the first and second family. where the exponential overlap between the Higgs and fermion wavefunctions also generated fermion mass hierarchies. 1. then they will couple differently to the Kaluza–Klein gauge bosons. Therefore. in our scenario the FCNC constraints lead to a much smaller lower bound on the first excited Kaluza–Klein mass m1 . Let us consider first the following generic four-fermion operators which are relevant for proton decay and K−K¯ mixing: Z Z Z √ 1 1 (0) (0) (0) (0) (59) d4 x dy −g 3 Ψ¯ i Ψj Ψ¯ k Ψl ≡ d4 x 2 Ψ¯ i Ψj Ψ¯ k Ψl . in the model of Ref. a lower bound on the compactification scale of 25−300 TeV is obtained [42] (the range in the bound depends on whether or not the induced FCNC processes violate the CP symmetry). This. 4. 1. [43]. This can give rise to large flavour-changing neutral current (FCNC) processes [42]. [44]. since ciL < 1/2. ci . Pomarol / Nuclear Physics B 586 (2000) 141–162 cj R ). The lower bounds are much reduced because. as in the original setup [2]. This is because (in non-supersymmetric models). and is also similar to Ref. A. M4 M5 where the effective four-dimensional mass scale M4 is 2k 1 e(4−ci −cj −ck −cl )πkR − 1 1 . (56). that if the fermions are localized at different points (by assuming that they have different 5D masses ci ). (57) we have m1 & 2−30 TeV. There is also the problem of generating large neutrino masses. As we saw in Section 3.3. means that the light fermions are localized on the Planck-brane. the Higgs field must live on . the four-dimensional suppression scale. For the case represented by Eq. (55)). However. as shown in Fig. constraints on K–K¯ mixing requires that the dimension-six operator. it does offer a possible explanation for the mass hierarchy between fermion families. while m1 & 0. (58). In addition.

This constraint is easily avoided if the neutrinos decay. We will consider two new possibilities. but this now requires a more detailed analysis. A.. as discussed in the previous section. ντ ). (a) and (b).e. M4 is given by M4 ' M52 −2(1−ciL )πkR e . Gherghetta. The Higgs can be either (a) not localized at all. (63) and leads to neutrino mass values of (1. The Higgs fields can now live anywhere in the bulk since their masslessness will be stable from radiative corrections. we find that the constraints on K−K¯ mixing. if supersymmetry is realized. This occurs if either the SM fermions are actually . On the other hand. it is impossible to suppress proton decay operators and have fermion masses. although FCNC constraints can be safely avoided if fermions live in the bulk.4. allow the fermions (and their superpartners) to be placed on the Planck-brane. These two possibilities. with a small overlapping on the Planck-brane. or (b) localized on the Planck-brane. then the masses do not satisfy the bound from overclosing the universe. (49). the requirement that the Higgs field is localized on the TeV-brane. 103. While these neutrino masses are not ruled out by collider experiments. let us consider the dimension-five operator responsible for generating Majorana neutrino masses. (61) ≡ d4 x M4 iL where C5 is the charge-conjugation operator. we can relate the neutrino masses to the fermion masses: mνi ' 102λ2ii GeV. if the neutrinos are stable. Supersymmetric models In the case of non-supersymmetric theories we were forced to have the Higgs field localized on the TeV-brane. k (62) Using the relation (58). 107 ) eV for (νe . Thus. (49) (i. 4. that require M4 & 1000 TeV. In particular.T. is in direct conflict with proton decay. In conclusion. However. could induce large proton decay. in order to solve the gauge-hierarchy problem. conformally flat limit). Pomarol / Nuclear Physics B 586 (2000) 141–162 157 the TeV-brane and will have a very small overlap with the fermions. c = 1/2 in Eq. To delocalize the Higgs from the TeV-brane and allow new scenarios we need supersymmetry. c  1/2 in Eq. they can be problematic for the recent SuperKamiokande results and cosmology. can be satisfied for values of ci compatible with generating fermion masses. For ciL > 1/2. Finally. any higher-dimensional operator will be suppressed by MP [6] (or the slightly smaller GUT scale of 1016 GeV). we can relax the above assumption. νµ . Again assuming that the left-handed fields are in the bulk we obtain Z Z √ 1 T C5 ΨiL H (x)H (x)δ(y − πR) d4 x dy −g 2 ΨiL M5 Z 1 (0) T (0) Ψ C5 ΨiL H H. The main motivation for having the SM fermion living on the Planck-brane is to avoid dangerous higher-dimensional operators that. By placing the fermions on the Planck-brane.

4. are rescaled to MP e−πkR . solving the supersymmetric flavour problem. Finally. (37). There will no longer be a massless gaugino mode in the spectrum. In particular we expect a mass gap between the scalar masses and the gaugino masses. In the latter case. Planckian physics is rescaled down to TeV energies. are localized by the AdS metric on the Planck-brane. In particular.158 T. the fermions form part of the five-dimensional hypermultiplets that. Supersymmetry breaking on the TeV-brane Let us assume that supersymmetry is broken on the brane at y ∗ = πR. Therefore we expect that the sparticle spectrum will be very similar to that in theories of supersymmetrybreaking by TeV compactifications [11–14]. For example. the scalar masses will be flavour diagonal. Since gauge interactions are flavourindependent. Therefore a gauge boson can be produced at the Planckbrane with a TeV energy and then propagate to the TeV-brane. communicating any physics on the TeV-brane to the Planck-brane. Thus. This will modify the Kaluza–Klein decomposition of the gaugino fields given in Section 3. Unfortunately with the Higgs off the TeV-brane. as shown in Section 3. There is also another motivation here. the full Kaluza–Klein mass spectrum will be shifted up with respect to the boson sector. A. which now have TeV masses. they will communicate it to the Planck-brane. [11–14].2. or they are bulk fields with fivedimensional masses c  1 (see discussion below Eq. Squarks and sleptons on the Planck-brane will receive masses at the quantum level. they will behave as messenger fields between the two branes. gaugino mass terms can be generated from the TeV-brane spoiling the supersymmetric condition Eq. the Yukawa couplings are no longer exponentially suppressed. if supersymmetry is assumed to be broken on the TeVbrane. This is a necessary consequence if some of the SM fermions (or the Higgs) live in the bulk. and the fermion mass hierarchy cannot be explained by the mechanism outlined in Section 4. This presents a new alternative to models with gaugino condensation where the small scale of supersymmetry breaking is generated by dimensional transmutation. we expect that in this scenario we will be able to test GUT or theories of quantum gravity at the TeV. Note that at the TeV-brane. they couple to the TeV-brane and can feel the breaking of supersymmetry at tree-level. But also Planckian (stringy) physics can be communicated to the Planck-brane. we can infer here some of the consequences. This is because the scalars will receive masses at the one-loop level . Although the precise value of the Kaluza–Klein masses and eigenfunctions will be left for future work. (60)). Pomarol / Nuclear Physics B 586 (2000) 141–162 confined to the Planck-brane as four-dimensional fields. Since we assume that the SM gauge superfields live in the 5D bulk.1. The breaking of supersymmetry on the y ∗ = πR brane will be communicated to the rest of SM particle superfields by the gauge superfields. one can generate small supersymmetry-breaking soft masses. If the SM gauge-bosons live in the bulk. At the TeV-brane. Gherghetta. we will consider the SM gauge-boson as a five-dimensional bulk field. Since all mass scales at y ∗ = πR. In particular. 4. it will interact with Planck scale excitations of the gauge boson. Notice that not only the massless mode but all the Kaluza–Klein tower will mediate the breaking of supersymmetry similarly as in Ref.

but in this case the squark and slepton masses. even if the Higgs soft masses are zero at tree-level. This can be problematic since the electroweak scale will be of order TeV. Supersymmetry breaking arising from the brane at y ∗ = πR. What about the Higgs? In model (a) the Higgs will also couple to the TeV-brane and Higgsino masses can be induced at tree-level. where only gravity lives in the five-dimensional bulk. 4π 159 (64) where αi is the gauge coupling of the gauge group under which the scalar i transforms. This gives rise to the usual µ-problem. and implies that the radius of the extra dimension satisfies kR ' 6. they will be induced at the loop level. The scalar masses will be an order of magnitude smaller than the gaugino masses and fulfill the relation m2i m2j = αi . the Higgsino mass is zero since the Higgs superfields do not couple to the TeV-brane. Of course. . Nevertheless. its mass will be of order TeV2 /MP . This represents a very light gravitino with mass m3/2 ∼ 10−4 eV and satisfies the usual constraints from cosmology and collider experiments [47. soft masses will be of order (MP e−πkR )2 /MP . however. This singlet can get a VEV at the electroweak scale and induce Higgsino masses. However. while the complete SM. that arise at the loop level. the scalar Higgs will also get masses at tree-level. and having the Higgs slightly localized towards the Planck-brane. lives on the Planck-brane. we must require that only the Higgsino mass is induced at the tree-level.48]. In general. for the Higgs coupled to the top. the lightest supersymmetric particle (LSP) will be the gravitino. 5 Therefore. Finally. In this case. c & 1/2. will be too small.T. αj (65) The right-handed slepton will be the lightest scalar. 5 We could lower the supersymmetry breaking scale from a TeV to 100 GeV. For model (b). Gherghetta. Pomarol / Nuclear Physics B 586 (2000) 141–162 m2i ∼ αi (TeV)2 . including the gauge fields. The origin of such a pattern of supersymmetry-breaking masses must be addressed by the fundamental (string) theory. This model is very similar to gravity-mediated models of supersymmetry-breaking and we do not expect any new interesting phenomenology. A. we expect that the soft Higgs masses will be positive at the one-loop level. we want to comment on a different alternative to the models considered here. there are also negative two-loop effects coming from the top/stop that can dominate the one-loop contribution and make the Higgs mass negative [11–14]. The simple solution that we envisage is to introduce a singlet field that couples to the Higgsinos. Another alternative would be to reduce the couplings of the Higgs to the TeV-brane by increasing its 5D mass. Since it couples to the TeV-brane by 1/MP -suppressed interactions. As in the case of the squark and slepton masses. This can lead to electroweak breaking at a scale an order of magnitude smaller than the supersymmetry-breaking scale. would then be communicated by gravity to the SM fields.

For these values of mΨ . It is appealing that the hierarchies in the fundamental scales of physics can be directly related to the geometry of space–time. there is no significant bound if the 5D fermion masses satisfy mΨ > σ 0 /2. the warp factor needed to explain the hierarchy becomes impotent. since the Higgs and fermions are located near the Planck-brane. and the heavy fermions to the TeV-brane. the supersymmetric flavour problem is naturally solved. Furthermore.160 T. We have shown that this could explain the fermion mass hierarchy by confining the light fermions to the Planck-brane. Thus. Thus. In non-supersymmetric theories. Remarkably. In this case the Higgsino mass could also be induced at tree-level. and strongly depends on the 5D fermion masses as shown in Eq. we reiterate that the whole scenario. If the gauge fields live in the bulk. the four-dimensional massless fermions are slightly localized near the Planck-brane and have an exponentially small coupling to the Higgs. Finally. these constraints forbid the generation of fermion masses. This fact alone. must eventually be embedded into some underlying theory (such as string theory). depending on their spin. We have studied the Kaluza–Klein mass spectrum and wavefunctions of scalars. including the details of the source branes. This model leads to a supersymmetric mass spectrum where the right-handed slepton is the lightest scalar and the LSP is a very light gravitino. then gaugino mass terms will be generated at tree-level. A. This provides a new alternative to gaugino condensation models. gauge bosons and fermions and analyzed the conditions for supersymmetry. We have also studied the phenomenological implications of having the SM fields propagating in a slice of AdS. Conclusion Bulk fields living in a slice of AdS have different behaviour. the gauge bosons couple to each brane with equal strength. we have shown that constraints from proton decay forces the fermions to be strongly confined towards the Planck-brane. In the massless sector we find that the scalar and fermion fields can be localized by the AdS geometry on either brane (depending on their 5D masses). Assuming that the source of supersymmetry-breaking is on the TeV-brane. is when the Higgs field is conformally flat (c = 1/2) and nonlocalized. such a fundamental theory must provide a tree-level supersymmetry-breaking mechanism on the TeV-brane. 1. However. Pomarol / Nuclear Physics B 586 (2000) 141–162 5. Similarly. higher-dimensional proton decay operators can be suppressed. while at the same time having fermion masses. The gauge fields will communicate the supersymmetry breaking to the Planck-brane where the squarks and sleptons live. warrants further investigation of these scenarios. because now there is very little wavefunction overlap with the Higgs. the massless Higgs mode is protected from radiative corrections and the Higgs can now live off the TeV-brane. The precise scale of the TeV-brane is constrained by electroweak precision data. by studying higher-dimensional operators. the scale of supersymmetry-breaking will be of O(TeV). while the gauge bosons are always nonlocalized. (53) and Fig. If supersymmetry is present. and therefore provide a window for physics of the distant brane. Gherghetta. only the Higgs is required to be localized on the TeV-brane. One particularly favourable model. In particular. Unfortunately. .

Lett.D. Lu. One of us (AP) acknowledges the ITP in Santa Barbara for hospitality during the initial stages of this work. B 466 (1999) 107. Rev. Cvetic. hep-ph/0002297. M. M. L. Deser. Ponton. [38] A. [22] H. Antoniadis.N. [2] L. Quiros. D 58 (1998) 065002. JHEP 0001 (2000) 003. Rev. contract no. hep-ph/9911293.J. M. hep-th/9911218. The work of TG is supported by the FNRS. Phys. J. D 60 (1999) 093006. [31] E. A. A. G. Phys. T. M. Schmaltz. A.G. Benakli. [36] T. Cvetic. Mirabelli. Neubert. C.98. Rev. Lett. Pomarol. Phys. hep-th/0001159. Phys. Verlinde. J. 83 (1999) 4922. M. . 38 (1977) 1433. Dudas. M. Cline. Lett. Yamaguchi. Oriol Pujolàs. J. R. hep-th/9911134. [19] A. A. Zumino. [4] W. A. Grojean.D. Townsend. Schmaltz. Witten. [8] S. Phys. B. Grossman. Antoniadis. B 469 (1999) 123. M. Rev. [29] P. Nucl. [37] W. hep-th/9908076. H. E.E. Gherghetta. H. N. A. D 60 (1999) 116004. Goldberger. M. D 60 (1999) 095008. Chacko. Munoz. hep-ph/0001172. [5] H. Quiros. Phys. [28] A.E. Behrndt. Rizzo. Lett. Davoudiasl. Gabadadze. Antoniadis. K. [35] M. Stelle. [12] A.B. [24] D. Pomarol. Rev. D 59 (1999) 086001. Phys. [3] J.L. Phys. Phys. Pomarol. Phys. Phys. [30] S. B 397 (1993) 515. S. D 61 (2000) 016007. Phys. B 438 (1998) 255.K. Lett. hep-th/0001166. [17] K. hep-ph/9911517. Gherghetta. Rizzo. Rev. 2155560. Dienes.R. D 60 (1999) 107505. M. D 59 (1999) 086004. Nucl. and Daniel Waldram for helpful discussions. Chang. 83 (1999) 3370. [18] M. Pomarol. Bajc. M. Peskin. Dimopoulos. Dvali. Lett. Quiros. [25] K. W. [9] B. [14] A. B 544 (1999) 503. M. Pomarol / Nuclear Physics B 586 (2000) 141–162 161 Acknowledgements We wish to thank Emilian Dudas. Phys. Strumia. Rev. Rev. G. G. while that of AP is partially supported by the CICYT Research Project AEN990766. B 326 (1994) 69. Lukas. Nucl. hep-ph/9912408. Skiba. hep-th/9912232. Dimopoulos.S. Linde. [15] D. Brandhuber. [16] Z. Phys. Kehagias. hep-th/9906182. hep-ph/9911262. M.T. E. Nath.A. Phys. [20] A. Arkani-Hamed. Phys. [13] I. Phys. Delgado. Nelson. Hisano. Yamaguchi. Randall. Luty. [23] R. Lykken. D. Wells. S. Wise. Ovrut. Marciano. [21] C.A. JHEP 0002 (2000) 005. hep-th/0002054.A. B. C.G. Shuster. [10] K. Nakano. K.D. [6] A. D 15 (1977) 2802. Rev. Benakli. K. Youm. Rev. Hewett. [33] I. [7] Y. Kribs. Pomarol. [26] M. [32] E. Masip. Kaplan. Waldram. [34] P. hep-ph/9911294. D 60 (1999) 096005. Quiros. hep-ph/9908530. Pope. Okada. Sfetsos. Lett. hep-ph/9912498. Kallosh. Phys. Rev. Rev. [11] I. [27] E. Servant.E. G. Phys. Sundrum. Phys. Randall. JHEP 9910 (1999) 013. B 429 (1998) 263. Phys. Lett. Rev. J. T. References [1] N. hep-th/9910081. B 554 (1999) 198.

Phys.D. B 462 (1999) 48. R. A. Pomarol / Nuclear Physics B 586 (2000) 141–162 R. Lett. Lett. F. Rev. F. Phys. D 61 (2000) 015008. Delgado. T. hep-ph/9912265. Rev. Rico. Feruglio. Mirabelli. S. Gherghetta. Lett. F. M. Gatto. B 423 (1998) 311. D 61 (2000) 033005. Ibanez. Brignole. Phys. G. . M. B 516 (1998) 13. Rev. M. N. D 61 (2000) 037701. B 475 (2000) 295. Phys. Phys. J. Casalbuoni. A. Quiros. De Curtis. Phys. Cornet. Schmaltz. Schmaltz. Gherghetta. M. Carone. Arkani-Hamed. M. Lett. Nucl. Shifman. C. L. Phys. JHEP 0001 (2000) 030.162 [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] T. Phys. Zwirner.E. Relano. Dvali. E. D. B 181 (1986) 269. Pomarol.A. A. A. Dominici.

R. 12. In the conformal case the absence of confinement leads to a different solution. In view of the recent developments of the AdS/CFT correspondence [1–4] it is thus natural to address this problem in the new setting proposed in this way.saclay. An exact correspondence for QCD is not yet known.  2000 Elsevier Science B. however useful information can be obtained from known realizations for confining theories.3 2 1 3 ( 0 0 ) 0 0 4 0 5 . in a confining theory.cea. revised 14 June 2000. PII: S 0 5 5 0 . which sets the scale for the Regge slope.∗ . 30-059 Cracow. AdS/CFT correspondence. Introduction The theoretical calculation from “first principles” of high energy scattering amplitudes in the so-called “soft” regime of QCD is among the oldest and yet unsolved problem of strong interaction physics.b.38.V.A. PACS: 11. France b M.Nuclear Physics B 586 (2000) 163–182 www. E-mail: janik@spht. Infra-red divergences appearing for Wilson lines can be factorized out or can be cured by considering the IR finite case of correlation functions of two Wilson loops. 11. Janik a. All rights reserved.fr 0550-3213/00/$ – see front matter  2000 Elsevier Science B. ∗ Corresponding author.nl/locate/npe Minimal surfaces and reggeization in the AdS/CFT correspondence R. when varying impact parameter. High energy amplitudes 1. Peschanski a a Service de Physique Theorique CEA-Saclay F-91191 Gif-sur-Yvette Cedex. In non-conformal cases related to confining theories. reggeized amplitudes with linear trajectories and unit intercept are obtained and shown to come from the approximately flat metrics near the horizon. Jagellonian University Reymonta 4. All rights reserved. the amplitudes are shown to be related to minimal surfaces generalizing the helicoid in various AdS5 backgrounds. Using the AdS/CFT correspondence in the classical approximation.elsevier. The main reason is that it requires a good understanding of 4dimensional gauge field theories at strong coupling which we do not possess till now. accepted 21 June 2000 Abstract We address the problem of computing scattering amplitudes related to the correlation function of two Wilson lines and/or loops elongated along light-cone directions in strongly coupled gauge theories.25.-t Keywords: Wilson loops. Poland Received 10 March 2000. Smoluchowski Institute of Physics.15-q.V. A transition between both regimes appears.-w.3 .

e. when α 0 → 0. where inelastic channels are expected to play an important rôle. In [7.e. we will analyze the correlation function of Wilson lines leading to an evaluation of q q¯ and qq scattering amplitudes at high energy. t) = s α(t ) × (prefactors). In particular. [7. In the AdS/CFT correspondence. We will concentrate on a situation where the difficulty with supergravity field exchanges does not arise. In this approach we will start by considering the correlation function of two Wilson lines elongated along the two light-cone directions. The case considered in our previous paper Ref. where a is the transverse extension of the Wilson loop. these correlation functions are related to minimal surfaces in the AdS5 geometry which have the Wilson loops as boundaries. the produced gravitational field in the dual AdS theory becomes strong.. since the scale of the bound is fixed by the particle of smallest mass (e. The gauge theory scattering amplitude is linked with a correlation function of tilted Wilson loops elongated along the light-cone directions [9– 12]. Janik. The case of IR finite correlators of Wilson loops will be dealt with in a second stage. namely L/a  s 2/7 . [7. i. the pion). by the massive W bosons arising from breaking U (N + 1) → U (N) × U (1). purely elastic scattering..A. Unitarity. The rôle of the quarks in the AdS/CFT correspondence will be played. where s. This will allow us to extend our study to small impact parameters. as in [13. The dominant contributions were identified and all correspond to real phase shifts. since there exists a single connected minimal surface which gives the dominant contribution to the scattering amplitude in the strong coupling regime. This will be done in the context of the black hole geometry in AdS space [15] (static Wilson loops were first studied in this background in [16. Reggeization of scattering amplitudes is expected to occur. The plan of our paper is as follows: in Section 2. Peschanski / Nuclear Physics B 586 (2000) 163–182 We would like to discuss relevant physical properties of scattering amplitudes at high energy expected from the S-matrix theory of strong interactions [5]. R.g.17]). the contribution of the bulk graviton gives an unexpected “gravity-like” s 1 behaviour of the gauge theory phase shift in a specific range of energies and (very) large impact parameters. In particular the Froissart bound [6] implies that α(t = 0) 6 1 and the prefactors of the amplitude are at most like log2 s.164 R.8] was that the weak field approximation in supergravity was shown to be broken unless the impact parameter L was sufficiently large. i. preventing perturbative calculations to be done in this background.. a configuration which can be used for the description of high-energy quark–quark or quark–antiquark amplitudes in gauge theories [9–11].14].8]. we considered large impact parameter and high energy scattering of colourless states for SU(N) supersymmetric gauge theories in the strong coupling. α(t) is the Regge trajectory corresponding to singularities of partial waves at j = α(t) in the t-channel.e. analyticity and crossing relations implied by the S-matrix theory impose constraints on α(t). In particular. Note that the Froissart bound assumes an underlying confining field theory. high-energy two-body amplitudes behaving as A(s. i. If the above condition is not met. or at least a mass gap. t are the well-known Mandelstam variables. large N limit using the AdS/CFT correspondence.. where one can use a flat .8] involved disjoint minimal surfaces and thus the necessity of including supergravity field exchanges between the two corresponding string worldsheets. The main but stringent difficulty which limited the scope of Refs.

The problem of the cancelation of the infra-red divergences is analyzed by considering Wilson loop correlators in Section 4. Geometry of the Wilson lines in Euclidean space. The AdS/CFT correspondence gives a recipe [13. Peschanski / Nuclear Physics B 586 (2000) 163–182 165 metric as a good approximation scheme near the horizon. It is convenient to pass from transverse momentum t = −q 2 to impact parameter space Z i 1 ˜ l). R. while the conclusions and open problems are pointed out in the final section. s 2π where l is the 2-dimensional impact parameter (in the following we will denote its modulus by L). Fig. t). Janik. see Fig. In the eikonal approximation the impact parameter space amplitude for q q¯ scattering is given by a correlation function of two Wilson lines [9–11] which follow the classical µ µ µ µ µ straight line quark trajectories W1 → x1 = p1 τ and W2 → x2 = x⊥ + p2 τ . 1.18] for calculating this correlation function through 1 ˜ l) = e− 2π α0 Aminimal . We analyze the factorizable structure of the IR divergences and isolate a cutoff independent inelastic amplitude leading to reggeization. t) = (1) d2 l eiq·l A(s. (2) . hW1 W2 i ≡ A(s. The IR cutoff will correspond to a fixed temporal extent of the lines −T < τ < +T . we consider the so-called “conformal” case of the AdS/CFT correspondence for N = 4 supersymmetric SU (N) gauge theory. In Section 3. with |x⊥ | = L. A(s. and A˜ is the amplitude in the impact parameter space. leading to the (approximate) derivation of scattering amplitudes between colourless states. where the AdS5 metric gives rise to a different minimal surface solution.R.A.14. 1. Wilson lines and minimal surfaces in “quasi-flat” geometry Let us start by defining an appropriate gauge theory observable for q q¯ scattering amplitudes A(s. 2.

m T → iT .g. an AdS black hole [15–17] among other geometries [19] for confining theories) bounded by the Wilson line segments limited by the cutoff T .A. this procedure defines the appropriate colour decomposition of the associa = −(1/2N)δ δ + ated amplitude. which is thus implicitly 1 2 normalized by 1/ hW1 i hW2 i. l) + AN 2 −1 (s. Using the well known colour decomposition tija tkl ij kl (1/2)δil δj k . where hW1 W2 i is the Wilson line correlator 1. A different approach to discuss the minimal surface problem in the conformal AdS5 was considered in [20]. In the case considered in this paper the analyticity structure contains branch cuts in the complex plane which have to be taken into account. the procedure for finding a minimal surface is ambiguous.8] since there.1. We will adopt a prescription for finding the minimal surface for infinitely long lines and then truncating it to a finite temporal extent parameterized by the IR cutoff T . Peschanski / Nuclear Physics B 586 (2000) 163–182 q 2 N in units of the AdS radius. AdS5 × S 5 for the conformal N = 4 SYM. Using the same strategy as in our first paper [7. which concentrated on the elastic part of the amplitude. Since the disjoint contour formed by the two Wilson line segments is not closed. 2 where A˜ 0 (resp. 9 f (z) z2 z2 (5) 1 The free propagation of the q and q¯ states is not included in the correlator hW W i. the hW W i correlation function had only simple poles in the complex θ plane. Janik. . In turn. we will perform the calculation with Euclidean signature for Wilson lines in the boundary R4 forming a relative angle θ in the longitudinal plane and then we will make an analytical continuation into Minkowski space by rotating the euclidean time coordinate clockwise and the angle anticlockwise (see [21] in this context): s θ → −iχ ∼ −i log 2 . (4) Note that a priori there is an ambiguity in making the analytical continuation depending on the precise choice of the path.. This implicitly consists of forming a “big” Wilson loop closed at large temporal distance by curves drawn on the infinite minimal surface. This phenomenon did not appear in the context of large impact parameter near forward scattering discussed in [7. adjoint) representations. l) . we have   1 ˜ ˜ ˜ (3) A(s. A˜ N 2 −1 ) are the amplitudes in the singlet (resp. α 0 = 1/ 2gYM and Aminimal is the area of the minimal surface in the appropriate background geometry (e. 2.8].166 R. R. l) ≡ N A0 (s. AdS black hole solution and its flat space approximation In [15] a proposal was made that a confining gauge theory is dual to string theory in an AdS black hole (BH) background the relevant part of which can be written as 2 = dsBH 16 1 dz2 ηµν dx µ dx ν + + ···.

so the resulting area which enters the formula for the amplitude (2) may be approximated by the area of the “flat space” piece. Property (ii) effectively induces this twisting to occur near the horizon as we shall show below. we will use this background to study the interplay between the confining nature of gauge theory and its reggeization properties.R.A. . Therefore we put the two tilted lines depicted in Fig. Peschanski / Nuclear Physics B 586 (2000) 163–182 167 where f (z) = z2/3 (1 − (z/R0 )4 ) and R0 is the position of the horizon. In all cases there is a scale. A consequence of (i) is that it is most efficient for a minimal surface to perform the “twisting” between the two Wilson lines as far away from the boundary as possible. Now near the horizon. Janik. 1 on the boundary at z = 0. (6) ds∼horizon R0 and the motion in the z direction is “frozen out”. Two salient features of the metric (5) are (i) the standard AdS prefactor 1/z2 close to the boundary (z = 0). 2 Although it was later found that the S 1 KK states do not strictly decouple in the interesting limits [22]. R. They have the property that for small z.e. At this stage we thus have to find a minimal surface between the lines at an angle θ in the flat space metric (6).14]. we have to evaluate the correlation function (2).25]. As is well known for the Plateau problem of minimal surfaces [26] the boundary conditions determine the solutions. The area of the “vertical” pieces is removed by the standard subtractions [13. similar to R0 above. following property (ii). the geometry looks like AdS5 × S 5 (in [24] up to logarithmic corrections related to asymptotic freedom) giving a coulombic q q¯ potential. 2). the minimal surface between well separated lines rises “vertically” in the z direction up to the horizon without sizable motion in the other R4 coordinates (see a schematic representation in Fig. Actually the qualitative arguments and approximations should be generic for most confining backgrounds 3. which marks a transition between the small z and large z regimes. Let us determine under what conditions the 2 Compared to standard coordinates [17] we used U = z−4/3 and U = R −4/3 . Due to property (i). In order to calculate the scattering amplitude. The relative angle (tilt) in the t–y plane and the separation in the transverse direction x (impact parameter) therefore define the boundary conditions for the geodesic equations for the string. Next we have to find the minimal surface in the appropriate geometry which has the two lines as its boundaries. i. the minimal surface performs the “twisting” (not displayed in Fig. The metric at the horizon is effectively flat 1 2 ∼ 2 ηµν dx µ dx ν . close to the boundary. Although an exact solution for the minimal surface spanned by the tilted Wilson lines is unknown for the metric (5). [19. (ii) the existence of a horizon which limits from above the values of z. The appropriate minimal surface in the BH geometry will look as follows. Finally the surface falls off again vertically towards the boundary. the properties of the black hole (BH) geometry allow for quite a good approximation scheme. 2) corresponding to the tilt angle θ between the initial Wilson lines.. We will now substantiate this intuitive picture with a more quantitative study of the geodesic equation for the string.24]. T 0 3 Two other geometries for (supersymmetric) confining theories have been discussed recently [23. as already discussed in Refs. For large z the geometry is effectively flat.

In all cases the ratio d/ l(τ ) → 0 when l(τ ) → ∞. R. In Ref. Gij is the background metric. 2.168 R. near the boundary (z → 0). The minimal surface equations follow from the Nambu–Goto action: 1 S= 2πα 0 ZT l(τ Z )/2 dτ −T p det hab . τ ) . x µ ) in (5). As a first remark we note that using the background metric (5) the terms in the induced metric hσ σ corresponding to the twisting are of the form  2  2  ∂t ∂y 1 + . v 1 ≡ Xi stands for general √ τ . (7) −l(τ )/2 where the induced metric on the worldsheet is hab ≡ Gij ∂Xi (σ. Thus the boundary conditions are “frozen” and transported to the vicinity of the horizon. which measures the transverse distance (on the boundary) over which the string worldsheet significantly deviates from being flat. v 0 ≡ σ . The Wilson lines are drawn here with vanishing angle of tilt θ = 0. ∂v a ∂v b (8) coordinates (z. For further discussion we shall make an approximation (similar to [20]) of neglecting explicit τ dependence in the Euler–Lagrange equations following from (7) and leaving it only in the implicit dependence on the boundary conditions through l(τ ). made for the case of the static q q¯ potential may be directly applied to our problem. Within this approximation the estimate of [19]. the minimization will not change noticeably the twist angle. for all metrics giving confinement. [19] a distance d is defined. minimal surface spanned by the two tilted Wilson lines is indeed predominantly flat and concentrated near the horizon following the general line of discussion of [19]. Janik. (9) ∂σ ∂σ z2 Hence. Depending on . The minimal surface in the black hole geometry. τ ) ∂Xj (σ. and l(τ ) = L2 + θ 2 τ 2 is the Euclidean distance between points on the two Wilson lines with the same value of the time coordinate τ . Peschanski / Nuclear Physics B 586 (2000) 163–182 Fig.A.

In this paper we will consider the generic case (from the 4D (S)YM point of view) when this limit resembles the original AdS5 × S 5 geometry [1–3]. We will not consider this problem in the present paper.A. The truncated helicoid solution may be parametrized by 4 For finite cutoff T in flat space. The precise dependence of d(l(τ )) on the horizon scale R0 depends on the metric considered. 0. We are thus led first to calculate the area of the minimal surface bounded by the tilted lines in the flat geometry (6) at the horizon. which is equal to L. L2 : τ → (τ cos θ. by keeping the impact parameter fixed and putting the scale R0 → ∞. In this case (see Fig. We will consider this conformal (non-confining) regime in detail in a further section. R. To summarize the discussion. the string is then to a large degree concentrated in the region near the horizon (6) with the boundary conditions essentially transported from z = 0. d behaves as a logarithm or a power of l(τ ) smaller than one. 0). The precise behaviour at these shorter distances will depend on the type of gauge theory and. We note that the same behaviour can be equivalently obtained through rescaling. Note. We will also be interested by the “truncated” helicoid 4 where τ = −T . τ sin θ.24]) and will just probe the small z region of the geometry. . Janik. 0. Peschanski / Nuclear Physics B 586 (2000) 163–182 169 the confining metric considered. corresponding to the two Wilson lines separated by a distance L in the “transverse” direction x and with a relative angle θ in the “longitudinal” plane: L1 : τ → (τ. Therefore.2. as long as the impact parameter is large with respect to R0 the approximations considered in this section should be valid. L). Let us recall the minimal surface solution in flat space. it may of course happen that the impact parameter distance between the two Wilson lines becomes much smaller than R0 . . However. Let us concentrate in the following on the case when the impact parameter is larger than the scale R0 . that with these boundaries the helicoid may be an unstable [27] minimum for a too large value of the cutoff. 0. (10) It is well-known that in the flat R4 geometry the minimal surface with infinite boudaries τ = −∞ · · · + ∞ is a helicoid. It is interesting to note that the condition d/ l(τ )  1 leads to a lower bound on the impact parameter L since the above condition is most restrictive for the smallest value of l(τ ). in particular. We will first perform the calculation in euclidean signature and then perform the analytical continuation (4).R. 2. on the small z limit of the appropriate metric. T as new boundaries. . 2) the minimal surface problem becomes less affected by the black hole geometry (or the large z behaviour of the different metrics [23. The helicoid is the only regulated (spanned by straight lines) minimal surface. For instance for the metric (5) rescaling arguments lead to a dependence 2/3 d(l(τ )) ∼ R0 log l(τ ). Helicoid geometry The basic building block of our construction is a minimal surface spanned by two straight line segments of length 2T . however. the truncated helicoid obviously remains a solution if one adds the boundary helices at τ = −T . T .

L and θ is the total twisting angle. T and σ = 0 . we will relate the analogue of the label n which appears in the calculation of Wilson loop correlators with multivalued saddlepoint minima of a minimization equation and thus to different classical solutions.A. Within the classical approximation which we have been using it is not possible to determine the value of n. x = σ. . The determination of the relative weights of the various contributions goes beyond the classical approximation used throughout this paper. Peschanski / Nuclear Physics B 586 (2000) 163–182 θσ θσ . log q i LT 1 + + (13) exp  2πR02 2χ L2 T 2χ 2 T  1 + L2 − χ L q 2 N/(2πR 2 ) coming from the where 1/(2πα 0 ) in (2) has been replaced by the factor 2gYM 0 flat metric (6). Its area is given by the formula s ZL ZT τ 2θ 2 dτ 1 + 2 Area ≡ S(T ) = dσ L 0 −T q s 2 2 2 2 2 1 + TLθ2 + θ TL T θ L q log = LT 1 + + . 2 2 L2 2θ 1+ T θ −θT t = τ cos L2 (11) (12) L Let us now perform the analytical continuation (4). (14) exp −n  χ 2R02  the form of which is uniquely fixed by the euclidean expression (12) up to a choice of the integer n. .. A naive continuation of the area formula (12) leads to a pure phase factor in (2): q q  s  2 2 2 N  2gYM 2 2χ 2 1 + T Lχ2 + χ TL  T L  . . L L where τ = −T . In fact when performing the analytical continuation we have to specify the Riemann sheet of the logarithm (i. Janik. However. L and χ dependence in (14) with an analogous factor exp(−nL2 /(π χ )) in the imaginary part of the D-brane scattering amplitude [28] where n labels the poles of the appropriate string partition function between the branes. the analytic structure of the euclidean area (12) involves cuts in the complex T . This method can be generalized to more complicated background 5 We also note the close similarity of the n.170 R. On a more physical ground.e. . Another useful way of deriving the above factor (14) can be directly obtained from the integral leading to (12). log → log +2πin). R. which links euclidean correlation functions in gauge theories with minkowskian ones directly related to scattering amplitudes. in Section 4. . y = τ sin . θ planes and thus leads to an ambiguity coming from the branch cut of the logarithm. 5 As can be seen the contribution (14) is cutoff-independent. This leads to an additional real multiplicative factor in (2):   q 2 N  2gYM L2  .

L2 Then we get L dσ 0 = niπ L2 χ (16) 0 which is effectively twice (×in) the area of a minimal surface bounded by a “half-elipse” of radii L and L/χ. We interpret this divergence as reflecting the expected IR divergence of the q–q¯ scattering amplitude [10. the superposition of long range coulombic and short range interactions leads to a factorization . which we will consider later.R. Janik. It is known since a very long time that the superposition of long range and short range potentials in the Schroedinger equation leads to a factorization formula for the relevant S matrix elements for each partial wave [29]. Peschanski / Nuclear Physics B 586 (2000) 163–182 171 geometries. for instance to the conformal case. and leads directly to the factor (14).11]. This T -independent imaginary part is unaffected by the second part of the analytical continuation (4).A. for which we lack an exact expression of the form (12). A consistent way to eliminate this cutoff dependence is to consider an IR finite physical quantity like scattering of two q q¯ pairs (see Section 4). the specific factorized form of (17) allows for a determination of an IR finite contribution. For instance in nuclear physics. (15) dτ dσ 1 − 2 = L 2χ −L/χ 0 We see that the imaginary part may be obtained by integrating (n times) around the branch cut of the square root. R. In the present case of Wilson lines.3. A convenient reinterpretation of the above formula follows from q performing the change of variables σ → σ 0 = σ 1 − r +L/χ Z 2in −L/χ 1− τ Z dτ 2χ 2 L2 τ 2χ 2 . This procedure yields the expression: s L/χ Z ZL τ 2 χ 2 πL2 . 2. (17) × exp −n  χ 2R02  There is a divergent phase in the above amplitude when the temporal length of the lines T goes to infinity. Reggeization in quark–(anti)quark scattering Our result for the Wilson line correlation function for the AdS BH geometry gives rise to the following contributions q q  s  T 2χ 2 2 N T   2gYM 2 2 2 1 + + χ T χ L L L2 log q i LT 1 + + A˜ n = exp  2πR02 L2 2χ T 2χ 2 T  1 + L2 − χ L   q 2 N  2gYM L2  . Let us perform only the first part of the analytical continuation (4) θ → −iχ but otherwise remain with the time variable T in Euclidean space. which can be interpreted as an effect of inelastic channels on the Wilson line correlator.

172 R. Within the above framework. In this context the amplitude (18) can be related to the inelasticity (overlap matrix) in the scattering namely √ 1 − 1 − 2f (l. The contribution of the real phase shifts behaves in the large T limit like  q  √  2 N    2gYM 2  2 eχT L 2 2 log + O 1/T i T χ + . the real phase shift exhibits a divergence which can be written as  2   2 L e coth χ log . s) are defined from the 2-body S matrix contribution to unitarity |S(l. in the QED result for electron scattering [30. The amplitude reads A(l. s) is purely real and thus contains information only on the inelasticity. We are led to interpret our resulting amplitude (17) in the same way. The factor (13) can be treated as redefining the initial and final q q¯ states due to long range interactions. s). The elastic S matrix may be treated as a redefinition of the asymptotic initial and final states. Note that this physical interpretation requires the integer n to be positive. s) .s) ∝ exp i 4π 4T 2 where 1/T has been substituted for a fictitious photon mass (IR regulator). s) = e2iδ(l. The effect of the confining potential is expected to generate inelastic channels through the phenomenon of string breaking and/or closed string emission.A. Peschanski / Nuclear Physics B 586 (2000) 163–182 into the elastic coulombic S matrix element and a short range amplitude modified by the long-range background. s) = 2 where the overlap matrix elements f (l.31]. For instance. arises from the contribution of many inelastic channels to the 2-body S matrix [33]. this phase is IR-divergent. . 6 It is clear that there remains a freedom in attributing a finite real phase shift either to the redefinition of the states or to the interaction. (18) where δ is the real phase shift due to the elastic long range interactions and T is the short range part of the amplitude. this contribution is expected to appear as an inelastic real factor in the amplitude. In hadronic interaction physics [32].s) · T (l. where we select initial and final q q¯ states. s) is obtained to be the cutoff independent factors (14). The analogous inelastic contribution Tn (l. (21) exp   2πR02 χ L The appearance of the IR divergent T 2 and L2 log T terms in the phase shift can be linked with the linear confining potential of the theory. Let us discuss both factors of the amplitude (17). Here we adopt the convention that T (l. while the phase factor diverges with T → ∞. Janik. We will return to the discussion of the n dependence in a further section. s) which. (20) T (l. R. (19) e2iδ(l. s). 6 Naturally. by unitarity of the S matrix. s)|2 ≡ 1 − 2f (l. a similar factorization appears for the S matrix elements for 2-body channels in terms of an elastic contribution and an amplitude T (l.

A. The (renormalon improved) 1-loop QCD result [11. (24) Upon analytical continuation this means that χ ∼ log s changes to χ − iπ ∼ log se−iπ . Finally let us compare our result with the general structure of Wilson line correlators at weak gauge coupling. it would be a candidate for the intercept one trajectories (the so-called pomeron and odderon) which are expected to emerge from a confining strongly interacting gauge theory. . Our nonperturbative result gives a hint on the scale and coupling dependence. For instance in the case of QED the whole dynamics is contained in the infinite phase factor (19) and the divergence is logarithmic. Tn (s. the twisting angle of the helicoid changes as θ → θ − π. the large T dependence of the q q¯ amplitudes we discuss reflects IR divergences which appear already in perturbative (weak coupling) calculations of the same quantities. t) (see formula (23)) is not spoiled by the different weights corresponding to fluctuations of the worldsheet around classical solutions. It is worthwhile to consider what changes in the preceeding discussion if we go from q–q¯ scattering to q–q scattering. In geometric terms. Janik. s) = exp −n  χ 2R02  173 (22) where the initial and final states are both q q.34] for q q¯ scattering is     T L2 ρ 1 αs log − Λ2 . ¯ It can be easily fourier transformed into transverse momentum space giving R2 0 iR 2 ln s 1+ 2n√2g2 N t YM s . R. Keeping the next to leading correction corresponding to log s → log se−iπ preserves the crossing relations between those channels.R. which are seen to have here a very simple geometric interpretation. Tn (l. (25) exp − χ π L π χ where ρ is an undetermined nonperturbative parameter. and since the string worldsheet spanned on the Wilson lines is oriented. Peschanski / Nuclear Physics B 586 (2000) 163–182 The inelastic q q¯ interaction amplitude at level n is  q  2 N  2gYM L2  . We note the compatibility between the nonperturbative cutoff independent piece in (25) and an analogous term in our result (22). Indeed. We note that in the asymptotically high energy limit log s  1. If the intercept one common to all contributions Tn (s. t) = q 0 2 N n 2gYM (23) This contribution is thus reggeized with a linear Regge trajectory with unit intercept and the slope given by the string tension related to the horizon distance R02 . as required by crossing properties. one obtains the same factors (17) for both qq and q q¯ channels. this corresponds to changing the orientation of one of the lines.

If we go to smaller impact parameters L < R0 (and also for T < R0 ).24] interpolating between a confining geometry at large z and approximately AdS5 × S 5 near the boundary z = 0). Janik.1. L θσ . As in the previous case. This different geometry leads to a qualitatively new behaviour which we now analyze. The analytic continuation θ → −iχ changes the . dτ dσ 2 1+ 2 z L −T (30) 0 p We perform a change of variables σ → σ 0 = σ 1 + τ 2 θ 2 /L2 which yields r 1 2πα 0 L 1+ τ ZT Z dτ −T 2θ 2 L2 dσ 0 1 z2 q 1 + zτ2 + zσ2 0 .A. We use the method outlined in Section 2 leading to formulae (15)–(16). y = τ sin L x = σ. the metric becomes closer and closer to the conformal AdS5 case. t = τ cos (26) (27) (28) z = z(σ. (29) Evaluation of the induced metric gives rise to the following area functional: s  ZT ZL  1 τ 2θ 2 1 + zτ2 + zσ2 . and some approximation scheme is needed. In the scenarios which behave better at short distances than the original BH proposal. Peschanski / Nuclear Physics B 586 (2000) 163–182 3. R. In this regime the dominant contribution to the amplitudes came from the part of the string worldsheet stretched near the horizon. Within a variational approximation approach. The conformal AdS5 case In the case of N = 4 SYM corresponding to the AdS5 × S 5 background we do not know yet the exact generalization of the helicoid. We note that the AdS5 × S 5 setting is directly related to scattering in the N = 4 SYM. τ ). 3. (31) 0 As in the previous section the cutoff independent part is obtained from the branch cut structure of the area functional (31). 2. we will look for a minimal solution in a restricted set of surfaces (“generalized helicoids”) parameterized by θσ . the minimal surface would only penetrate into a limited region near the boundary z = 0. we will concentrate on extracting the inelastic contribution which appears also here to be independent of the IR temporal cutoff T . Conformal case The flat metric approximation which we have used to derive the resulting area (12) assumed that the impact parameter L is sufficiently large with respect to the scale set by the horizon radius R0 (or a similar scale in the backgrounds [23.174 R. see Fig.

s) are not specified. 2π χ (33) where F (Ω) is a complicated function calculated in [35.36]. It is possible to obtain an approximate result in the high energy χ → ∞ limit from known properties of Wilson loop expectation values [13. in the window of convergence (the exponent of L in (35) between −2 and −3/2). We note that for n positive a similar energy dependence (i. change gYM YM Considering the impact parameter dependence and its Fourier transform to momentum space. (32) where  is the “5th” AdS coordinate where we put the D3 brane probe.  translates directly into the mass of the W bosons which play here the role of quarks.14]) which appears as the q 2 N → g 2 N in the exponent of s. Janik. s) ∼  log s Here. as in the case of the confining theory. Due to conformal invariance we know that the minimal area has the following form: Aminimal = f (L/..R. An approximate evaluation is then given by integrating the coulombic potential [13. The half-elipse has two cusps each with an angle π/2. This leads to the following logarithmic terms: −2 L 1 F (π/2) log . In the conformal case. we get . there remains a nonperturbative screening effect (already present for the static q q¯ potential [13.35. R. roughly 2L/χ apart. which are in the standard way subtracted out [13. In all cases we obtain a factorized energy behaviour with no moving Regge trajectories. Let us comment on the behaviour of the various components.14. (35) Tn (l. 14]: ZL −c 0 π dσ 0 = −c χ.36].36] (F (π/2) ∼ 0. whose contribution to Aminimal can be obtained from the results of [35.36]. Peschanski / Nuclear Physics B 586 (2000) 163–182 175 boundary conditions for the minimal surface to be a half-elipse of width L/χ and height p L (the upper integration limit in (31) then becomes L 1 − τ 2 χ 2 /L2 ). so we have at most a logarithmic behaviour in L/. We do not expect higher poles in  than first order. √ 2 2 02 4 χ L −σ (34) where c = 8π 3 /Γ 4 (1/4) is the coefficient in front of the (screened) coulombic potential.e. So we get √ 2 √ 2 N  n F (π/2) 2gYM 4 2gYM N π 2π L n 2π 4 2π s Γ (1/4) . with intercept greater than 1 and a (nearly) flat Regge trajectory) is obtained by resumming the leading log s terms in the perturbative expansion at weak coupling ([37–41].14.A. for a discussion in impact parameter space see [41]) for the singlet exchange amplitude. The  independent term g(χ) in (32) can be approximated by noting that at high energies the half-elipse is very much elongated and looks like parallel lines of length L. the values of n and the weights of the different components Tn (l.35.3π ). χ) + g(χ).

we expect a transition from the set of components (22) to the results (35). Peschanski / Nuclear Physics B 586 (2000) 163–182 Tn (s. Conformal/non-conformal transition As already mentioned. when the impact parameter is decreased and gets smaller than R0 . the result (36) obtained for the pure AdS5 × S 5 case should give the dominant behaviour also for the confining theory for impact parameters small with respect to the horizon scale 7 R0 (or more generally an analogous transition scale in the geometries [23. (36) Otherwise one observes either an UV divergence (for exponent values less than −2) or an IR one (for values larger than −3/2). while the conformal case (corresponding here to L  R0 ) used an approximation using coulombic potential. 4. In order to remedy this. For positive values of n the IR divergence requires a careful treatment which is beyond the scope of this paper. Thus even in the confining theory. However. Wilson loop correlators and scattering amplitudes We saw that an inherent feature of the q–q¯ scattering amplitude is its IR divergence. . This process can be observed by noting that both (22) and (35) were derived from a minimal surface spanned on a semielipse.42]. and also to show a context where the finite behaviour of the inelastic amplitudes calculated in the previous section appears directly without the infinite phases. the cutoff dependence on T is expected to be removed together with the related IR divergence which was present for the case of Wilson lines. and impact parameter distance L. R. Janik.A. When 7 We need also a sufficiently small T parameter. as long as the relevant geometry for small z is similar to AdS5 × S 5 . Note that. Indeed this R0 provides a natural value of the impact parameter cutoff L0 . we are led to consider the scattering of two q q¯ pairs of transverse size a. 3. as is the case already in the perturbative limit [11.2. where the loops are choosen to be elongated along the “time” direction and have a large but arbitrary temporal length T (the exact analogue for Wilson loops of T considered in the previous section). it can lead to infra-red divergent pieces also in the inelastic amplitude. t) ∼ is 4 1+n 2π 4 · Γ (1/4) √ 2 N 2gYM 2π  1+n F (π/2) 2π 1 t √ 2 N 2gYM 2π . For this setup we have to calculate the correlation function of two Wilson loops [12]. in the case of N = 4 SYM. since it gives some information on the scattering amplitudes between colourless states in gauge theories at strong coupling. The solution of the appropriate minimal surface problem in the full geometry would lead to an interpolation between the two extreme cases.24]).176 R. This process is interesting to study in itself. The result for impact parameters L > O(R0 ) was obtained by using the area law for Wilson loops. For large positive and negative times the minimal surface will be well approximated by two seperate copies of the standard minimal surfaces for each loop separately.

L > R ). We thus find a set of solutions very similar to the inelastic factor obtained in Section 2. Since we want to calculate the normalized correlator hW1 W2 i / hW1 i hW2 i. Janik. one can lower the area by forming a “tube” joining the two worldsheets. The modification due to . It is at this stage that we see that the result does not depend on the maximal length of the Wilson loops T . aL 1 + Ttube 0 The total area corresponding to the two Wilson loops is then given by s s T Ztube T 2 θ2 τ 2θ 2 dτ 1 + 2 + 2aL 1 + tube2 − 4a · Ttube. Inserting this solution into the area (38) we find Area(φ) = − 2L2 nπ + 2aL(−1)n . The whole contribution to the amplitude will just come from the area of the tube. see (12) for the euclidean case and (13) for theqminkowskian one). Peschanski / Nuclear Physics B 586 (2000) 163–182 177 we come to the interaction region. Analytically continuing the area formula (37) to the Minkowskian case and using a convenient change of variables. Ttube ). and subtract from it the area of the two independent worldsheets. and hence is IR finite. which has the interpretation of an “effective” time of interaction. there exists a solution with φ ∼ ±nπ . Then we will make a saddle point minimization of the area as a function of this parameter. ∂φ 2 (39) It is easy to realize that for large enough energy. the contributions of the regions outside the tube will cancel out (in a first approximation neglecting deformations near the tube). parameterized by Ttube . we will consider a family of surfaces forming the tube. Since we cannot obtain an exact minimal surface for these boundary conditions. Therefore we have just to find the area of the tube. (38) φ+ Area(Ttube ) = χ 2 the new variational parameter. Suppose that the tube linking the two Wilson lines is formed in the region of the time parameter t ∈ (−Ttube. Namely. In our approximation its two “sides” are formed by sheets of the helicoid solution (of area S(Ttube ).R. let us perform a variational approximation. R. where ρ ≡ a/L and sin φ = iχTtube /L isq 2 N → 0) the parameter φ is dynamically In the strong coupling limit (α 0 = 1/ 2gYM determined from the saddle point equation: 0= ρχ ∂Area(φ) = cos φ(cos φ − ρ) − sin φ.A. χ (40) where we retain the physical solutions with n positive integer. Area(Ttube ) = 2L L L (37) −Ttube where −2aTtube is the contribution of each individual Wilson loop to the normalization 1/(hW1 i hW2 i) of the Wilson loop correlation function. and for L sufficiently small. The front and back will be each approximated by strips of area 2 θ 2 /L2 (we assume a. the Minkowskian area can be put in the following simple form   2L2 sin 2φ + ρχ cos φ − 2ρ sin φ .

using R s j dj the usual Mellin transform s α ≡ 2iπ(j −α) . As a word of caution (and incentive for further study) we note that the saddle point in terms of Ttube is mainly driven to complex values. it is convenient to analyze the components of definite signature [5] with the even and odd contributions given by √ 2 √ 2 T˜n± (l. (41) Note the relative factor of 2 in the exponent in comparison with (22) due to the two sheet structure of the minimal surface. For lower energies. Peschanski / Nuclear Physics B 586 (2000) 163–182 the front–back contribution 2aL is negligible in the Fourier transformed amplitude for √ momentum transfer −t  a/R02 . For too large impact parameters we may enter the purely elastic regime found in [7.8] which does not correspond to connected minimal surfaces (the Gross–Ooguri transition [18]). 2 2 2n 2gYM N 2n 2gYM N (42) where αn (t) = 1 + R2 q 0 t. 2 N 4n 2gYM (43) Let us consider the contribution with n = 1 which is dominant at large L. It is easy to realize that the amplitude (42) corresponds to specific Regge singularities in the S-matrix framework.178 R. R. i. Recalling that charge conjugation acting on one of the q q¯ pairs is equivalent to considering the transformation χ → χ − iπ . Let us analyze the properties of the resulting amplitude. Indeed. For completeness. s) = e −n 2gYM N L2 χ R02 ±e −n 2gYM N L2 χ−iπ R02 . Using the Fourier transform (1) we finally get Tn± (s. namely double Regge poles whose trajectory is given by α1 (t).e. let us briefly discuss the general saddle point solution.. Janik. The study of these solutions is beyond the scope of the present paper. The determination of the weights of each component to the total scattering amplitude is beyond the reach of the classical approximation. This indicates that a complete treatment and an investigation of the Gross–Ooguri transition requires a more refined study of the tube minimal surface. For χ small there exist solutions with φ imaginary and thus leading to elastic parts of the amplitude. a classical solution. It is interesting to note that the minimization (39) gives rise in a natural way to a similar set of solutions parameterized by integers as found from the branch cut arguments in Section 2. there are families of solutions also leading to reggeized behaviour but with distorted trajectories.A. t) = iR02 ln s iR02 ln(−s) q s αn (t ) ∓ q (−s)αn (t ) . Also this term is probably more dependent on the treatment of the front–back parts of the tube in our approximation. it can be written in the following equivalent forms: . Each value of n corresponds to a saddle point.

33] leads to a similar decomposition where the Tn correspond to Regge pole/cut singularities. The transition between both situations and thus the relation between both sets of states remains an open problem. the latter states have masses proportional just to 1/R02 and spin limited by J 6 2. s) ∼ f0 exp − χ 2 2R0 this framework [33] unitarity is fulfilled whenever 0 < f0 < 1/2. R. . Janik. (45) where J is the spin and n labels the different trajectories.R.A. We should note that our approximations for calculating the Wilson loop correlator (which is the channel relevant for glueballs) are rather crude and become problematic at small t (consider the discussion after (40)). if   2 N 2gYM L2 for the inelasticity. Therefore the extrapolation of the linear trajectory into the glueball regime can easily break down. Because of the appearance of coupling constant dependence it is easy to see that these states correspond to massive string states and not to supergravity fields associated with the glueballs found in [43–46]. In particular. Unfortunately the complexity of the minimal surface problem with the Wilson loop boundary conditions does not allow us to make more quantitative estimates. In the absence of a direct determination of their relative weights. However. In we assume a gaussian distribution f (l. Finally let us comment on the relation of our results on the trajectory α(t) with the glueball spectrum calculations [43–46]. see (20). Let us discuss the contributions Tn to the amplitude with n > 1. leads to a specific model forqthe relative weights of the Tn ’s. π (j − α1 (t))2 cos πj/2 2 N 2 2gYM C 179 (44) where the complex contour C can be taken around the Regge (di)pole trajectory α1 (t) and the signature factors are either sin(πj/2) or −i cos(πj/2) depending on the positive or negative signature. the overlap matrix formalism [33]. Peschanski / Nuclear Physics B 586 (2000) 163–182 iR 2 ∂  α1 (t ) s T1± (s. The appearence of massive string states is not surprising in our case as we consider an extended string worldsheet between the two Wilson loops instead of a supergravity field exchange. it is interesting to note that unitarization of Regge amplitudes in the S matrix framework [32. t) = q 0 ∓ (−s)α1 (t ) ∂α 2 N 2 2gYM   Z R02 dj e−iπj/2 s j i sin πj/2 = q . Indeed. An extrapolation of the trajectory (43) to positive t leads to masses of the form q M 2 = 4n(J − 1) 2 N 2gYM R02 . the derivation of the Wilson line/loop correlation function does not allow us to give model-independent predictions for these weights in the total amplitude.

A. However the geometrical configurations in [47.. These results call for some comments. the minimal surface spanned between infinite lines forming an angle θ at the boundary. Conclusions and outlook Let us give our main conclusions. As expected. The slopes of the trajectories are the same as in our case (43). minimal surfaces with straight line boundary conditions corresponding to classical trajectories in Minkowski space. we want to emphasize the interest of solving exactly the well defined mathematical problem of finding the generalization of the helicoid for various AdS metrics. from behaviour of type (i) to (ii) when the impact parameter decreases below the interpolation scale set by the horizon radius. (ii) an approximate evaluation for the conformal AdS5 × S 5 geometry leading to flat Regge trajectories 8 and (iii) evidence for a transition. Janik.48]. along the lines of [28]. Peschanski / Nuclear Physics B 586 (2000) 163–182 5. This configuration corresponds to a high energy scattering amplitude between colourless q q¯ states. Even in the flat space approximation it would be useful to have a direct string calculation of the tube configuration. .48] is quite different from the one we considered in Section 4.e. in a confining theory. i. R.48].e. we isolate in certain cases IR finite inelastic amplitudes coming from the branch cut structure of the analytical continuation of helicoid-like surfaces. when the “effective” string length α 0 is taken to be set by the horizon radius in our case. Beyond the flat space approximation.180 R. By computing Wilson line and Wilson loop correlation functions in the framework of the AdS/CFT correspondence we show a relation between minimal surface problems in AdS5 metrics and reggeization in gauge field theory at strong coupling. For Wilson line correlators. confinement provides a natural IR cutoff scale. [47. In this case. In a second stage we considered the correlation function of two Wilson loops elongated along the light cone directions for the confining geometry. For high energies the amplitude is imaginary and thus mainly reflects the inelasticity of the process. Another goal is to go beyond the classical approximation in order to derive the n-dependent weights to the scattering amplitudes.. We note that the structure of our resulting amplitudes for the confining case (in particular the n. i. while the intercepts are different. 8 A remaining IR divergence in the inelastic amplitude is still present in the absence of confinement. It is interesting to note that the imaginary part in those calculations is generated from the singularities of the string amplitudes which are an infinite set of poles. We considered three cases: (i) flat metric approximation of an AdS black hole metric giving rise to Regge amplitudes with linear trajectories. χ and L dependence) matches the calculations of the imaginary part of flat space D-brane scattering amplitudes [28]√and some specific Wilson loop correlators 9 [47. 9 We have extracted the imaginary part from the formulae in Ref. We use a variational approximation where the minimal surface is constructed from two helicoidal sheets. the obtained amplitude is free from IR divergences and gives rise to reggeization with a linear trajectory with unit intercept.

B 565 (2000) 193.S. Maldacena. Fomenko. Lett. Yee. Peschanski. Phys. Giraud for useful remarks. see also Section 11 in [25]. B 374 (1996) 37. J.R. Sin.A. Aharony. Kinar. Sonnenschein. Phys. S. Verlinde. hepth/9302104. O. B 428 (1998) 105. J. Lett. J. A. Macroscopic strings as heavy quarks in large N gauge theory and anti-de Sitter supergravity. Sfetsos. H. D 53 (1996) 3835. G. Schreiber. H. QCD at high energies and two-dimensional field theory. I. Lett. . Patrìcio. Garel and B. Lett. S. Phys. J. S. P. Lett. Gubser. Rey.S. E.P. Phys. J. Yee. A. Acknowledgements R. Nachtmann. Maldacena. Witten. Oz.g.J. Y. Eq. H. 83 (1999) 3836. Quantum fluctuations of Wilson loops from string models. Phys. 1963. Nucl. North-Holland. R. N. hep-th/0003050. Adv. E. Robins. Gubser. S. We thank T. Lett. Yankielowicz. hep-th/9911123. J. Janik. Phys. Math. D 58 (1998) 106002. Frontiers in Physics. Peschanski / Nuclear Physics B 586 (2000) 163–182 181 Indeed the generalization of the helicoidal geometry in AdS space seems to be a building block for high energy scattering amplitudes in gauge theories at strong coupling. A. Rev.A. Boudaoud. Theor.-J. B 434 (1998) 36. E. Quantum Mechanics. B 456 (1999) 22. Zahed. Kehigas. A. Messiah.-J. Itzhaki. A. Phys. Adv. B 437 (1998) 77. Nucl. Benjamin. R. hep-ph/9609365 (see.87) for colourless state scattering). Lett. 2P03B08614. Klebanov. The Plateau Problem. Sonnenschein. Phys. M. Schreiber. E.. Froissart. O. J. Theisen. Armoni. RG flow. Polyakov. S. Angelantonj. A. Verlinde. Ooguri. Phys. 1989. C 76 (1997) 523. Theor. M. E. Rev. Kinar. Ooguri. Y. S. Ooguri. K. I. B 527 (1998) 171. Ann. Nachtmann. Large N field theories. R. Math. D. Y. e. Wilson loops and the dilaton tadpole. N. Witten.-T. Lett. Maldacena. hep-th/9803001. Phys. 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] J. Phys. H. 80 (1998) 4859. 123 (1961) 1053. Omnes.-J. Ben Amar.A. Rey. Gross. O. 2 (1998) 253. hep-th/9905111. E. C. Weiss. 1961. Janik. Phys. string theory and gravity. 2 (1998) 505. Mandelstam Theory and Regge Poles. Z. C.R. J. Rev. Janik. R. Brandhuber. B 325 (1994) 459. Lett. Phys. C 4 (1998) 101.T. Math. A. Rev. High energy collisions and nonperturbative QCD. B 566 (2000) 103. Phys. Korchemsky. 209 (1991) 436. hep-th/9909124.J. M. Phys. Meggiolaro.M. Tannenhauser. Froissart. 2 (1998) 231. Adv. R. Theor. Bachas. Gauge theory scattering from the AdS/CFT correspondence. J. H. Phys. Nucl. was partially supported by KBN grants 2P03B00814. Phys. Rev. Phys. M. (3. Phys. B 466 (1999) 199. Sonnenschein. S. Gordon & Breach. Phys. Eur. Phys. Cargese Summer School 1999. Rho.

J. L. J. Balitsky. E. Ooguri. Terning. S. Białas. Novak. Z. H.A.P. H. PhD thesis. 1996. J. Chen. C. N.N.I. Nunes. talk given at DIS 97. Sov. Ooguri. A. R.A. Phys. B 490 (1997) 306. Lipatov. Rev.182 [30] [31] [32] [33] [34] [35] [36] [37] [38] [39] [40] [41] [42] [43] [44] [45] [46] [47] [48] R. Prog. Mathur. Korchemskaya. Rev. 22 (1969) 666. Phys. L. Rozowsky. Lett. Nucl. hep-th/0003115. Phys. Glueball spectrum for QCD from AdS supergravity duality.S. van Hove. H.N. C. Theor. Peschanski / Nuclear Physics B 586 (2000) 163–182 H. . N. V. Two loop N = 4 supersymmetric amplitudes and QCD. Fadin. Ooguri. L. S. Kuraev. R. talk given at DIS 97. S. Sov. Sov. Myers. A. B 60 (1975) 50. Wu. Rev. Rev. Fadin. Itzykson.A. Y. D 60 (1999) 125006. E.P. Phys. D 58 (1998) 105009. Bern. Nuovo Cimento 38 (1965) 1385.N. V. JETP 44 (1976) 45.N. C. L. T. 36 (1964) 655.D. Phys.T. Lett. 23 (1976) 642. hepph/9706392. Y. Phys. G. Chaudhuri. Supersymmetric pair correlation function of Wilson loops. van Hove. de Mello Koch.C. I. JHEP 9910 (1999) 037. Phys. H. Phys. Drukker. Suppl.J. Jevicky. Phys. Chaudhuri. Quarkonium Scattering at High Energies. Lett. L. Nucl. J. Janik. 28 (1978) 822. Mihailescu. Phys. 23 (1969) 53. D. Csaki. Brower. hepth/0002046. R. Lipatov. Mod. University of Cambridge. B.A. Yan. Oz. Tan. Lipatov. E. JETP 45 (1977) 199. Sov.C. Rev. Salam. JHEP 9901 (1999) 017. Phys.R. Kuraev. I.S. Lipatov. Pair correlation function of Wilson loops. E. M. Abarbanel. J. Novak. Cheng. hep-th/9910183. Gross. Constable.-I. R. L. Nucl. Phys. 134 (1999) 153.

11.-q Keywords: Massless fields.V.N. AdS massless mixed symmetry fields reduce to a number of irreps of o(d − 2) algebra. Metsaev b. R.8 . D-14476 Golm.Nuclear Physics B 586 (2000) 183–205 www. in general. Germany Received 31 May 2000.R. The example of the three-cell “hook” Young diagram is discussed in detail. Leninsky prospect 53.A. Vasiliev d. Gauge invariance.15.elsevier. It is demonstrated that in contrast to massless fields in Minkowski space whose physical degrees of freedom transform in irreps of o(d − 2) algebra. The pattern of necessary flat-space ∗ Corresponding author. In particular. 117924 Moscow. From the field theory perspective this means that not every massless field in flat space admits a deformation to AdSd with the same number of degrees of freedom. Columbus. An equivalent statement is that. M. PII: S 0 5 5 0 . OH 43210-1106. All rights reserved. P.1 a Department of Theoretical Physics.c.ru 1 On leave of absence from Lebedev Physical Institute. because it is impossible to keep all of the flat space gauge symmetries unbroken in the AdS space. PACS: 11.ac.V.Kk.∗ . A conjecture on the general pattern of the flat space limit of a general AdSd massless field is made. an irreducible AdSd massless field does not classify according to irreducible representations of the flat space massless little algebra o(d − 2). USA d Albert-Einstein-Institut. it is shown that only a combination of the three-cell flat-space field with a graviton-like field admits a smooth deformation to AdSd . Lebedev Physical Institute. Max-Planck-Institut für Gravitationsphysik. All rights reserved. accepted 16 June 2000 Abstract Massless fields of generic Young symmetry type in AdSd space are analyzed. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. generic irreducible AdS massless fields reduce to certain reducible sets of massless fields in the flat limit. Chalmers University of Technology.nl/locate/npe How massless are massless fields in AdSd L. The Ohio State University. Moscow. E-mail: metsaev@td. Am Mühlenberg 1. AdS space–time 1.  2000 Elsevier Science B. Brink a .lpi. Introduction The aim of this paper is to demonstrate some peculiarities of generic free massless fields in anti-de Sitter (AdS) space of an arbitrary space–time dimension d. S-412 96 Göteborg. The main conclusion will be that. but reduces to a certain set of irreducible flat space massless fields. Sweden b Department of Theoretical Physics.3 2 1 3 ( 0 0 ) 0 0 4 0 2 .10. Russia c Department of Physics.

secondly.).6] higher spin gauge theories require infinite collections of higher spin gauge fields with infinitely increasing spins. For the first four Dynkin indices the bosonic and the fermionic numbers match up. Another important feature discovered in [7] is that gauge invariant higher spin interactions require the cosmological constant λ2 of the background AdS space to be non-zero to compensate the extra length dimensions carried by the higher derivative interactions required by the higher spin gauge symmetries.) The fact that higher spin theories require an AdS background was regarded as rather surprising until it was realized that it plays a distinguished rôle in the superstring theory as well [8–10]. 11. e.4]): firstly to overcome the well-known barrier of N 6 8 in d = 4 supergravity models and. A conjecture on the possible form of the higher spin symmetries and equations of motion for higher spin spin gauge fields was made in [11.. the dimension of M-theory.184 L. A generalization like this to higher dimensions is not straightforward because of the use of certain auxiliary twistor type variables.12] as a certain generalization of the d = 4 results [13. . As a starting point. to investigate if there is a most symmetric phase of superstring theory that leads to the usual string theory as a result of a certain spontaneous breakdown of higher spin gauge symmetries. where it was shown that there exists an infinite collection of triplets of higher spin fields having equal numbers of bosonic and fermionic degrees of freedom..g. Another motivation comes from the flat space analysis of certain massless (nonsupersymmetric) triplets in d = 11 in [15. In superstring theory all types of representations appear at the higher massive levels. d > 4 (e.16]. / Nuclear Physics B 586 (2000) 183–205 massless fields will be given. bringing in exceptional groups into the . This phenomenon does not take place. . [3. To investigate a possible relationship between the superstring theory and higher spin theories one has to build the higher spin gauge theory in higher dimensions. Brink et al. because it holds only for the representations of the space–time symmetries described by non-rectangular Young diagrams. The effect discussed in this paper takes place for d > 6. (In this perspective λ plays the rôle analogous to α 0 in superstring theory. d = 10. it is therefore important to analyze more carefully the notion of a general massless field in AdSd . This phenomenon is rather special for 11 dimensions and follows from the fact that the little group SO(9) is an equal-rank subgroup of F4 . These two motivations lead in fact into the same direction because. . for all those types of massless fields that appear in the usual low-energy massless sectors of the superstring models and supergravities.14] which were proved to describe interactions of all d = 4 massless fields. These triplets show some remarkable properties. though. This is the main goal of this paper.g. The study of higher spin gauge theory has two main motivations (see. For the same reason it cannot be observed in AdS4 higher spin gauge theories [1–3] (all massless fields in AdS4 are described by one-row Young diagrams). as shown for the d = 4 case [5. since it is impossible to keep all the flat space gauge symmetries unbroken in the AdSd space. Another (rather unexpected) manifestation of this fact is that not every massless field in flat space admits a deformation to AdSd with the same number of degrees of freedom.

Unitarity implies compatibility with quantum mechanics. ˆ k. b = 1 ÷ d − 1. The generators Mmˆ nˆ are Hermitian. t± (2) (3) (4)  1  a b . Lab = −iM ab . Massless unitary representations in AdSd AdSd is a d-dimensional space–time with signature (d − 1. ˆ n. and span the Lie algebra o(2) ⊕ o(d − 1) of the maximal compact subgroup a are combinations of AdS of the AdS group SO(2. t− 2 (5) [Lab . where ηnˆ mˆ = (−. · · · +. m. (1) ˆ lˆ = 0 ÷ d. Physically meaningful relativistic fields in AdSd are classified according to the lowest weight unitary representations of o(d − 1. respectively. 2. The commutation relations are explicitly Z-graded with . The commutation relations (1) take the form   a a . t+ = Eδ ab − Lab . If these triplets have anything to do with higher spin gauge field theories and/or M-theory it is an interesting question whether it is possible to extend the analysis of [16] to the AdS case. 2) space. (8) The generators E and Lab can be identified with the energy and angular momenta. +. / Nuclear Physics B 586 (2000) 183–205 185 picture.L. 2 E = M0d . 2). It is most useful to identify AdSd with the universal covering space of the appropriate hyperboloid embedded into Minkowski space–time with signature (d − 1. The non-compact generators t± translations and Lorentz boosts. −) is the flat metric in the (d − 1. where a. Mkˆ lˆ] = i(ηnˆ kˆ Mmˆ lˆ − ηmˆ kˆ Mnˆ lˆ − ηnˆ lˆMmˆ kˆ + ηmˆ lˆMnˆ kˆ ). The hermiticity conditions are  † a † a t± = t∓ . 2). t±c ] = δbc t±a − δac t±b . (6) [Lab . E † = E. while lowest weight of a unitary representation guarantees that the energy is bounded from below. General facts The commutation relations of o(d − 1. 2. = ±t± E. 1) and the group of motions SO(d − 1. That question in fact triggered this investigation. d − 1). Let us choose the following basis in the algebra: a = t±  1 M0 a ± iMd a . Brink et al. 2) are [Mmˆ nˆ . Lab = −Lab . Lce ] = δbc Lae − δac Lbe − δbe Lac + δae Lbc (7)  with all other commutators vanishing. 2).1.

[17] a and [18]) starting with the vacuum space |E0 . 2). 2) are then realized as bilinears in the quantum fields.g. We see that massive states are classified by the parameter E0 (which is the analog of mass) and a representation of o(d − 1). We will here not discuss the self-dual representations that can be singled out with the aid of the Levi-Civita symbol. . t+ 2 (12) 2 The self-dual and antiself-dual representations which appear for odd d are usually distinguished by a sign of the s(d−1)/2 . As a result. . Let us emphasize that the elements of the module (11) can be identified with the modes of a one-particle state in the corresponding free quantum field theory. Brink et al. The elements of the AdSd algebra o(d − 1. s) is irreducible and unitary. si = E0 |E0 . If E0 gets sufficiently small. the corresponding tensors are traceless.. The full representation of the Lie algebra o(d − 1. s is a generalized spin characterizing the representation. si t+ (11) for all k. = t−a . . The lowest weight unitary representations t± are now constructed in the standard fashion (for a review of the AdS4 case see. In terms of Young tableaux si is the number of cells in the ith row of the Young tableaux. This picture is in agreement with the standard description of massive relativistic fields in flat space–time in terms of the Wigner little group SO(d − 1).186 L. si. Since we are talking about representations of orthogonal algebras. Such representations are identified with massive representations of AdSd . is spanned by the vectors of the form a1 ak · · · t+ |E0 . . the analysis of unitarity can be reformulated as the check of the positivity of the norms of the finite-dimensional subspaces at every level. Field-theoretically this corresponds to a finitecomponent field carrying a finite spin. si = 0 (9) and forms a unitary representation of the compact subalgebra o(2) ⊕ o(d − 1) that means in particular that E|E0 . For sufficiently large generic E0 it is intuitively clear that the representation D(E0 . 2 Note that s describes a finitedimensional representation of o(d − 1). Note that states with pairwise different k are orthogonal as a consequence of (4). / Nuclear Physics B 586 (2000) 183–205 a having grade ±1.. Note that at the level of equations of motion the light-cone field-theoretical realization of generic AdSd massive representations for arbitrary E0 and s has been developed recently in [20].e. E is the grading operator. The states with fixed k are called level-k states. . i. e. sν ) with ν = [(d − 1)/2]. the norm cannot stay positive as is most obvious from the following consequence of (5)   d −1 a E. all norms are strictly positive. (10) Here s denotes the type of representation of o(d − 1) carried by the vacuum space: s = (s1 . s). si that is annihilated by t− a t− |E0 . denoted in [19] D(E0 .

24] identified with the conformal fields at the boundary of the AdS space [22]. (Otherwise. The question we address here is whether or not the shortening in AdSd can be interpreted in terms of irreducible representations of SO(d − 2). Obviously. E0 = E0 (s). .) Therefore. This will be demonstrated both at the algebraic level using 3 Singleton-type fields live at the boundary of AdS and cannot be interpreted as bulk massless fields. In particular. The fact that a singular submodule can be factored out admits an interpretation as some sort of a gauge symmetry (true gauge symmetry for the case of massless fields. massless fields are classified according to the representation of the massless little group SO(d − 2). Brink et al. one can build a negative norm state that is in contradiction with the assumption that we are at the boundary of the unitarity region. It can be shown that the same result is true for odd d (provided that sν is replaced by |sν |).. Note that this bound for d = 4 was originally found in [21]. or independence of bulk degrees of freedom for singletons and doubletons).e. By factoring out this subspace one is left with a unitary representation which is “shorter” than the generic massive representation.L. (13) (14) where t1 is the number of rows of the maximal length s1 . / Nuclear Physics B 586 (2000) 183–205 187 which implies that for negative E0 some level 1 states cannot have positive norm for a positive-definite vacuum subspace. We will show that the answer is no for a generic representation. Some zero-norm vectors then appear for E0 = E0 (s). the zero-norm states form an invariant subspace called a singular submodule. The resulting “shortened” unitary representations correspond either to massless fields [21] or to singletons [22] and doubletons [23. Massless representations are “shorter” than massive ones classified according to the parameter E0 (equivalent of mass) and a representation of o(d − 1). There is therefore a boundary of the unitarity region E0 = E0 (s) > 0 such that some states acquire negative norm for E0 < E0 (s). Starting from the inside of the unitarity region decreasing E0 for some fixed s one approaches the boundary of the unitarity region. this is true for the second rank antisymmetric tensor representation in the AdS5 case that can be identified with the field strength of the Yang–Mills fields of the N = 4 SYM at the boundary of AdS5 . The final result is E0 > E0 (s). i. d Presumably. 3 The analysis of positive definiteness of the scalar product of level-1 states was done in [25] for an arbitrary even d and an arbitrary type of representation of s carried by the vacuum state. these states should have zero norm for E0 = E0 (s). In flat space. For this reason we choose this definition of masslessness for all fields in AdSd except for the scalar and spinor massless matter fields which are not associated with any gauge symmetry principle and singletons. s1 = · · · = st1 −1 = st1 > st1 +1 > st1 +2 > · · · > sν . E0 (s) = s1 + d − t1 − 2. For the case d = 5 see [26] and references therein. all singletons except for scalar and spinor correspond to maximally antisymmetrized representations of the AdSd algebra equivalent to their duals. These necessarily should have vanishing scalar product to any other state.

a |E00 . s0 i = 0 t− (15) E|E00 .2. it forms some irreducible representation s0 of o(d − 1). namely the Cartan subalgebra. s0 i that has the properties analogous to (9) and (10). . The most important field-theoretical conclusion is that a generic irreducible massless field in AdSd decomposes into a collection of massless fields in the flat limit. (17) if |E00 . singular vacuum spaces form representations of the algebra o(2) ⊕ o(d − 1) and therefore decompose into a direct sum of irreducible representations of o(d − 1) on different levels. . An important consequence of this fact is that not every massless field in flat space can be deformed into AdS geometry. 0). . the singular submodule spanned by zero-norm states is itself a lowest weight representation. 2. 0. 4 Clearly. Obviously. . E00 = E0 + k. . . In that sense. s0 i. s1 s2 (19) with 0 6 s2 6 s1 . 0) corresponding to a two-row Young diagram.. . The standard situation is with a single irreducible singular vacuum space. Let us now consider the example of a vacuum state |E0 . a massless field in the AdSd is generically “less massless” than elementary massless field in flat space. 4 This terminology is very closely although not exactly coinciding with that used for the Verma module construction when irreducible vacuum subspaces are one-dimensional because the grade zero subalgebra. t+ (18) The factorization to a unitary irreducible representation is equivalent to identifying all these vectors to zero. Singular vectors It is useful to reformulate the problem in terms of singular vectors.e. . s0 i belongs to the level-k subspace. s0 i = E00 |E00 . . si with s = (s1 . Since the energy E is bounded from below for the whole representation. 1. s2 .188 L. (16) and i. / Nuclear Physics B 586 (2000) 183–205 the language of singular vectors and at the field-theoretical level focusing on the simplest nontrivial massless field with s = (2. s0 i we will call singular vacuum spaces while any of their elements will be called a singular vector. 0. Such spaces |E00 . Brink et al. Therefore it contains at least one nontrivial subspace |E00 . is Abelian. s0 i. The singular module as a whole then has the structure a1 ak · · · t+ |E00 .

bs1 } (E0 ) = 0.e...b1 .bs1 (E0 ) which is symmetric both in the indices a and in the indices b.bs1 (E0 ).. Let us consider the two representations with cells cut.bs1 (E0 ) ≡ 0 (23) t− for some E0 (Πα is a projector to one or another irreducible component in (22)).. satisfying the antisymmetry property va2 .L.. 2 As a result.bs1 (E0 ) = 0...b a (E0 ) s1 − s2 + 1 1 s2 −1 1 s1 −1 s2  (24) (symmetrizations within each of the groups of indices a and b are assumed) and va21 .a c.bs −1 2 1   1 = E0 − (d + s1 − 3) va1 .. Taking (20) into account it is enough to require ηb1 b2 va1 .as .as2 .b1 ..bs1 −1 c (E0 ) 2  s2 va . second or an (additional) third row.as2 {bs1 +1 .b1 .b1 .....b1 ...b a (E0 ) + s1 − s2 + 1 1 s2 −1 1 s1 −1 s2 (26) and  1 E0 − (d + s2 − 4) va1 ..bs1 −1 c (E0 ) + 2 s2 vca ... s2 and s1 it contains five irreducible components: two Young diagrams with one cell less (index c is contracted to either one of the indices a or one of the indices b) and three diagrams with one cell more: adding one cell to the first.. which means that contraction of any two indices with the o(d − 1) invariant flat metric ηab gives zero...a .bs1 −1  c = t+ va1 .b ..as2 −1 ..as2 .bs1 (E0 ). Brink et al. For generic d..bs1 (E0 ).as2 ....b1 .as 2 −1 .b1 ....as2 .b . singular vectors appear at t−c va21 .. Our problem therefore is to check whether any of these irreducible representations can be a singular vacuum space.b1 ..as2 −1 c..as2 .......b1 . (20) which implies that symmetrization over any s1 + 1 indices a and/or b gives zero.. / Nuclear Physics B 586 (2000) 183–205 189 This means that |E0 .as 2 −1 ... t+ (22) These states form a reducible representation of o(d − 1) (the tensor product of the vector representation with the representation (19)).b1 . si can be realized as a tensor va1 ..bs1 c = t+ vca1 ..........b1 .b1 .bs1 E01 = d + s1 − 3 and = (27) (28) . i.....as . (25) Elementary computations give that t−c va11 ... The appropriate projections are given by the following formulae describing irreducible o(d − 1) tensors va11 . The tensor is traceless..as2 .. (21) The level one states are c va1 ..b1 .  e c Πα t+ va1 .

b1 ..bs2 . b. where it was also shown that only the representation resulting from the factorization of the singular submodule with the highest energy of a singular vector is unitary (this fact is natural from the singular vector description: unitarity can be preserved only when the boundary of the unitarity region is approached.b1 . One can analogously investigate singular spaces with the boxes added to make sure that they have negative energies and therefore do not play a rôle in our analysis.. (31) and The analysis in terms of singular vectors is simple enough but can be simplified further with the aid of the technique proposed in [27] with the tensors corresponding to various Young diagrams realized as certain subspaces of an appropriate Fock space. We therefore conclude that unitary massless particles appear for E0 = d + s1 − 3. for s1 > s2 (30) E0 = d + s1 − 4.. = 0..as1 . . one can decompose the o(d − 1) tensor Aa1 . 3. . = 1 ÷ d − 1) of a specific symmetry type. (32) then it describes a representation of o(d − 2) of the same symmetry pattern.. the values of the “singular” energies (28) and (29) are in agreement with the general analysis of [25]. this implies the highest E0 ). We here use the tensor language to make most clear the interpretation in terms of the representations of the massless little algebra o(d − 2).190 L.as1 . If the tensor Aa1 ..as1 . = 0 .. into irreducible representations of o(d − 2) with the aid of the projection operators constructed from na or. As expected. We expect that the analysis of higher levels does not affect our conclusions.. (29) A few comments are now in order.b1 .1 since we will use it in the field-theoretical part of the paper. for a vector. performing dimensional reduction..... . . One reason is that the appearance of singular vectors at higher levels within the unitarity region would imply existence of higher spin gauge fields with gauge transformations having more than one derivative acting on a gauge parameter..... The cases s1 = s2 and s1 > s2 are different because v 1 ≡ 0 for s1 = s2 as a consequence of the antisymmetry property (20). nb1 Aa1 ..bs2 .. . Flat space pattern of AdS massless fields Let Aa1 .b1 ... be an irreducible tensor of o(d − 1) (a.. is orthogonal to na with respect to all possible contractions of indices na1 Aa1 . .bs2 . The general analysis of massless fields in flat space–time of an arbitrary dimension [28] shows that this does not take place.. If some contractions with na are nonzero... in other words...as1 .as1 ..bs2 .bs2 . o(d − 2) can then be identified with the stability subalgebra of o(d − 1) that leaves na invariant. For example... for s1 = s2 .. / Nuclear Physics B 586 (2000) 183–205 E02 = d + s2 − 4... For s1 > s2 both v 1 and v 2 are nontrivial. Brink et al.. This technique is explained in Section 4..b1 . Let na be a nonzero vector of o(d − 1)..

Brink et al. from s columns of equal height t). A = A .2. A⊥a = Aa − s t (34) In other words. At the field-theoretical level this means that it will not be possible to preserve all necessary gauge symmetries for λ 6= 0. containing a number of irreducible representations of o(d − 2). Let us consider an arbitrary Young diagram with row lengths s1 > s2 > s3 > · · · . This phenomenon is demonstrated in the field theoretical example in Section 4. heights) . The deformation will be possible however. (33) nc nc nc nc The analysis of singular vectors in Section 2. when a singular vector is present.2 admits a similar interpretation.L. Reintroducing λ and taking the flat limit λ → 0. This means that in the flat limit different singular vectors may decouple simultaneously and therefore a natural possibility a orthogonal) representation of consists of the flat space reduction to one (totally t+ o(d − 2). a as a vector na analogous to the momentum operator in flat-space field let us interpret t+ a commute with themselves). Let us note that the energy in AdSd is measured in units of the inverse AdS radius λ that was set equal to unity in our analysis. However such massless representations of the Poincaré algebra may not admit a deformation to a representation of the AdS algebra with λ 6= 0. Indeed. Therefore. the “reduced representation” is effectively smaller than an irreducible representation of o(d − 1) but may be larger than the corresponding irreducible representation of o(d − 2). if one starts with an appropriate collection of massless fields in flat space dictated by the “incomplete” dimensional reduction via decoupling of singular vectors. For our analysis it is more convenient to build Young diagrams not from rows as elementary entities but from rectangular blocks of an arbitrary height t and length s Aa = Aka + A⊥a . The fact that a singular theory (note that the operators t+ a with the vacuum vector decouple vector appears means that some contractions of t+ a from the spectrum and therefore are equivalent to zero. / Nuclear Physics B 586 (2000) 183–205 191 na nb b na nb b ka A . If all possible contractions of t+ would decouple this would mean that the o(d − 1) vacuum representation would reduce to the o(d − 2) of the same symmetry. The main aim of this section is to formulate a conjecture on the pattern of massless fields in flat space compatible with the deformation to AdSd . all energies of singular vectors tend to zero. in agreement with the standard analysis [29] of massless representations of the Poincaré algebra. A general Young diagram is a combination of blocks with decreasing lengths (equivalently. a block is a Young diagram composed of t rows of equal length s (equivalently. Since the energies (28) and (29) are different this cannot be true simultaneously.

while t1 is the number of rows of the length s1 . Brink et al. It can either be aligned along na or along the d − 2 perpendicular directions. s2 is the maximal row length of the remaining rows and t2 is the number of rows of length s2 . 2 (36) In other words. t)] is described in these terms by a single pair of integers (s. Let us now address the question what is the result of a dimensional reduction to one dimension less of a general diagram Y [(si . / Nuclear Physics B 586 (2000) 183–205 s1 t1 s2 t2 q q q q q q q q q q q q q q q sp−1 tp−1 sp tp (35) In these terms. Note that an elementary block Y [(s. ti )]. a Young diagram Y [(si . s1 is the maximal row length in the Young diagram. Every cell can be identified with some vector index.192 L. Therefore any number of cells from 0 . ti )] is described by a set of pairs of positive integers (si . But any number of indices from 0 to s1 can be chosen to take the extra value d − 1. There cannot be more than s1 indices along na because symmetrization with respect to more than s indices gives identically zero by the definition of a Young diagram. In the first case we cancel a cell. ti ) with s1 > s2 > s3 > · · · > sp > 0 and arbitrary ti such that p X i=1 1 ti 6 (d − 1). t). while in the second case we keep it.

. (s1 . (If not.. First consider a representation (e. t − 1). we can use the symmetry properties of the Young diagrams to reduce any contraction with N a1 a2 .g.. t)] (all indices are orthogonal to na ). the dimensional reduction of the block Y [(s. nak is totally symmetric. t2 )] composed from two blocks.ak of the original tensor to a contraction of N a1 a2 . consider a representation T described by a Young diagram Y [(s1 . Y [(s.L.) Let us consider some examples. Now. Of course only such cancelings are allowed that result in a Young diagram. (s2 . .ak = na1 na2 . s1 t1 s2 t2 (38) . 1)] which consist of two blocks with the bottom block having an arbitrary length 0 < s1 < s and height 1. t1 ). Since components of tensors along na are automatically symmetrized because the tensor N a1 a2 . dimensional reduction will lead to a number of tensors resulting from cutting an arbitrary number of cells in the bottom row of the block. tensor T ) described by the Young diagram Y [(s.ak with the bottom row of the block. t)] gives rise to the following representations of o(d −2): Y [(s. As a result. Brink et al... any resulting tensor is identically zero. / Nuclear Physics B 586 (2000) 183–205 193 to s1 can be canceled. every tensor appears once (note that the condition that the tensor is traceless does not affect this analysis since the reduced o(d − 2) tensors are also assumed to be traceless). t −1)] (a maximal possible number s of indices is contracted) and all diagrams Y [(s.. . t)] which is itself an elementary block.. s1 t1 − 1 n1 (37) In other words.

take away an arbitrary number of cells from the bottom line of the second box. t2 − 1) Y (s1 . 1). and two-block diagrams     Y (s1 . (s20 . t2 ) . (s2 . 1) . 1) . t2 ) . t2 ).) One can.     Y (s1 . (s10 . t1 − 1).g. (s2 . 1). t1 ). t1 − 1). (s2 . / Nuclear Physics B 586 (2000) 183–205 Again. t1 ). (s2 . t2 − 1). t2 − 1).. (s20 . t1 ). The rule therefore is: take away an arbitrary number n1 such that s1 − s2 > n1 > 0 from the bottom line of the top block and take away an arbitrary number n2 such that s2 > n2 > 0 of cells from the bottom line of the bottom block s1 t1 − 1 n1 s2 t2 − 1 n2 (39) Note that with this prescription we have n1 + n2 6 s1 in accordance with the general argument that one cannot cut a number of cells exceeding the maximal row length in the Young diagram. one cannot cut an arbitrary number of boxes in the top block because the cut line cannot be shorter than the length of the second block s2 . t1 − 1).     Y (s1 . e. Y (s1 . 1)] with arbitrary integers s10 and s20 such that s1 > s10 > s2 > s20 > 0 (40) and their degenerate versions described by the three-block diagrams     Y (s1 . (s2 . (s20 . 1). dimensional reduction means that one can cut some cells from the bottom rows of the upper and lower blocks (all cuts inside a block are equivalent by the properties of the Young diagrams to cutting its bottom line). t2 + 1) . t1 − 1).194 L. Analogously one proceeds for Young diagrams built from a larger number of blocks. [29] and references therein): . (s2 . Y (s1 . t1 − 1). The final result is that the dimensional reduction of a general Young diagram to one dimension less consists of the Young diagrams of the form (every diagram appears once. But now. (s10 . The pattern of the dimensionally reduced representation therefore consists of four-block diagrams Y [(s1 . t2 ) . however. (s2 . Brink et al. Y (s1 . see. (s2 . t2 − 1) . (s2. (s10 . t1 − 1). (Such tensors vanish identically.

e. require n1 = 0. A few comments are now in order. Note that massless fields corresponding to arbitrary irreducible representations of flat space massless little algebra o(d − 2) were considered in [28].L. i. . Take away any number of cells ni satisfying the conditions (42) of the bottom lines of all blocks except for the top one. Any diagram that appears as a result describes some irreducible representation of the massless little algebra o(d − 2) corresponding to some flat space massless field that should be present in the full set compatible with the AdSd geometry and unitarity. Let us now formulate the final result concerning a pattern of flat space massless fields that admit a unitary deformation to AdSd . Consider a AdSd massless field characterized by a diagram (35). Brink et al.. Conjecture. / Nuclear Physics B 586 (2000) 183–205 195 s1 t1 − 1 n1 s2 t2 − 1 n2 q q q q q q q q q q q q q q q sp−1 tp−1 − 1 sp np−1 tp − 1 np (41) One is allowed to take away any numbers ni of cells from the ith block provided that 0 6 ni 6 si − si+1 (42) (with the convention that sj corresponding to the “next to last” block equals zero).

An important consequence of the analysis of this section is that totally antisymmetric gauge tensors (i. 1)] decomposes into the following irreps of o(d − 2) algebra s2    X    ⊕Y (s1 . 1). (s2. it has been observed already in [25] at the level of equations of motion in the Lorentz gauge that some . ghost containing) sets of fields can be obtained by a similar procedure with one of the lower blocks remaining untouched instead of the top one as in the unitary case.. Some standard massless (gauge) fields may be needed as ingredients of AdS massless fields with nontrivial diagrams.e. 1).e. either by single rows. In fact all examples of massless fields that appear in supergravity and low energy string theory are described by elementary blocks (specifically.. Field theoretical example Now let us explain what happens from the field-theoretic perspective. Brink et al. 1).e. 1) . (s2 . The space of differential forms (including the spin one gauge fields) therefore is closed with respect to the deformation to AdSd . Note also that for the well studied case of lower dimensions d 6 4. or by single columns). For Young diagrams being themselves elementary blocks (i. This means that elementary block massless fields in AdSd classify according to irreducible representations of o(d − 2) as in the flat case (i.e. This is to say that AdSd massless field corresponding to Y [(s1 . differential forms) corresponding to the diagrams Y [(1. 0 6 s2 6 s1 . It is tempting to speculate that this may correspond to a nontrivial deformation of gravity to the AdS geometry in the presence of other fields. (s2 . The spectrum of flat-space massless fields to which an elementary AdS massless field decomposes is non-degenerate. That is why the phenomenon discussed in this paper was not observed before. (s. no additional massless fields should be added to deform to AdSd ). Y (s1 . Analogously one can find a totally symmetric spin s1 > 2 field corresponding to the diagram Y [(s1 .2). 1).. 1)] a graviton-type flat space massless field Y [(2. 1)]. 4.e. t)] can never appear as a result of a decomposition of a certain irreducible (unitary) AdSd massless field in the flat limit. In fact. for the representation Y [(2.. Non-unitary (i. i. ti = 0 for i > 0) only one o(d − 2) representation appears. singular vector) components that decouple. For example. 1)] among the fields resulting from the decomposition of the AdSd field Y [(s1 . only block-type massless representations are nontrivial (propagating) and therefore this phenomenon does not occur either. 1) ⊕ (43) s=1 where each term under summation appears just once. 1)] will be present (see example in Section 4. Therefore canceling out a box from the upper block corresponds to pure gauge (i. 1)... all the representations of o(d − 2) are pairwise different.e. 1) AdS → Y (s1 . described by the same block. (1.196 L. / Nuclear Physics B 586 (2000) 183–205 The role of the upper block is singled out by the unitarity condition: only singular a to the upper block have maximal energies and vectors corresponding to contractions of t+ describe unitary representations [25].

n = nm Λmn as (x) = −Λas (x). 4 The corresponding action is invariant under the gauge transformations (46)  1 m1 m2 n m1 n ∂ Λas + ∂ m2 Λas .) The Lagrangian can be chosen to be of the form 1 L = Φm1 m2 .m3 } = 0.n 2 1 3 − Φm1 m2 .n + Φ m. 1). satisfying the condition that full symmetrization with respect to all three indices gives zero Φ{m1 m2 . 1)] (44) In flat space the massless field of this symmetry type was described in [30].1.n2 . 1)] of the massless little algebra o(d − 2). Brink et al. 1).n m1 m2 . Here we analyze the problem at the Lagrangian level focusing on the explicit comparison with the flat-space limit. δas Φ m1 m2 . (47) 2  1 m2 n m1 n 1 m2 + ∂ m2 Λsym (48) − ∂ n Λm δsym Φ m1 m2 .n being a symmetric tensor in m1 and m2 . (1.n 2Φ m1 m2 . nm Λmn sym (x) = Λsym (x).n2 − Φ m1 m1 . / Nuclear Physics B 586 (2000) 183–205 197 of the redundant gauge symmetries expected in the flat-space description are absent in the AdS case.n1 ∂ n1 ∂n2 Φ m1 m2 . (45) (The authors of [30] used an equivalent representation with explicit antisymmetry in two indices.n1 ∂ n1 ∂n2 Φm2 m2 . Flat space Let us consider the simplest nontrivial example of a massless field having the symmetry properties of a non-block diagram with three cells Y [(2. 4.n 2 4 3 m 3 m1 m2 .n ∂ m1 ∂m2 Φ mm2 . Because this Lagrangian is fixed by the gauge transformations one can say that it is the gauge invariance with respect to the both gauge transformations (47) and (48) that ensures irreducibility of the massless field.n 4 4 3 + Φ m1 m1 . although equivalent way.n ∂m1 ∂m2 Φ + Φ ∂m1 ∂m2 Φ m m. We take the representation with the field Φm1 m2 . (1.n = ∂ m1 Λsym sym 2 mn with antisymmetric gauge parameter Λmn as (x) and symmetric gauge parameter Λsym (x). (49) In [30] it was proved that the Lagrangian (46) describes a physical massless field in flat space corresponding to the irreducible representation Y [(2.n − Φmm1 .L. .n 2Φm2 m2 . To begin with. let us reformulate the results of these authors in a somewhat different.

α¯ A . The fact that we deal with the Young diagram (44) is equivalent to imposing the following constraints on the generating function |Φi N11 |Φi = 2|Φi. d − 1. . The constraints (54) and (55) tell us that the oscillators respectively. D Dm ≡ em m Dm . + 4 P11 D 5 Tangent space (fiber) indices m. α1m (57) α2m and occur twice and once. α¯ A .2 mn where ωm is the Lorentz connection of space. The constraint (56) is equivalent to the condition (45).. The Lorentz covariant derivative for the representation |Φi takes the form X  1 m n n m M mn = α¯ A − αA α¯ A . αA (58) Dm ≡ ∂m + ωmmn M mn . n and target space (base) indices m. E = 1. SA ≡ α¯ A m Dm . (52) and |0i is a Fock vacuum m |0i = 0. (60) . α¯ B = 0. In the sequel we will often use the notation DA ≡ αA m Dm .g. NAB ≡ αA m PAB ≡ αA αBm .198 L. . In curved space–time it is convenient to use fiberwise fields Φ m1 m2 m3 ≡ em1 m1 em2 m2 em3 m3 Φ m1 m2 m3 . Brink et al.n α1m1 α1m2 α2n |0i. . (54) N22 |Φi = |Φi. on the right-hand side of Eq. 1. (51). m PSAB ≡ α¯ A α¯ Bm . n take the values 0. while M mn forms a representation of the Lorentz algebra so(d − 1. αB = 0. α¯ A (53) The indices A. (59) The flat-space Lagrangian (46) now takes the form 1 S1 − D2 D S2 − 3 P11 2PS11 L = hΦ|2 − D1 D 4 2  2 2 3 S1 + D1 PS11 + P11 D2 D S2 PS11 |Φi. / Nuclear Physics B 586 (2000) 183–205 Let us now introduce a notation that simplifies computation. 2 label two sets of oscillators. (56) where we use the notation m α¯ Bm . . (55) N12 |Φi = 0. (50) where em m is the vielbein of an appropriate space–time (e. C. 2 A=1. 1). B. 2 (51) m and α¯ Bn are auxiliary creation and annihilation operators where αA  m n  m n  m n αA . Minkowski or AdS). αB = ηmn δAB . 5 It is most convenient to formulate the action in terms of the following Fock-type generating function [27] 1 |Φi = √ Φm1 m2 .

With these SB satisfy a number of useful relationships conventions the covariant derivatives DA and D summarized in the appendix. Dn ] = −λ2 Mmn . D and Lorentz connection n . D2 ≡ Dm (67) where the second term accounts for Dm being rotated as a tangent vector.2. em n = δ m ωm kn = 0. It is sometimes convenient to combine them into a gauge parameter Λm. As in the flat case we will analyze gauge symmetries with totally symmetric and totally mn antisymmetric gauge parameters Λmn sym and Λas . The symmetric and antisymmetric parts can be singled out as |Λsym i ≡ |Si = N12 |Λi. α¯ nA = en n α¯ nA . (65) The condition (64) is just the standard zero torsion condition Dm en a − Dn em a = 0 (66) while (63) is the equation of the AdS space. The Lorentz covariant derivatives (58) are no longer commuting but satisfy the commutation relationships [Dm . Brink et al. (61) (62) 4. δsym |Φi = 1 2 D1 N21  − D2 |Λsym i. Dm αnA − Dn αmA = Dm α¯ nA − Dn α¯ mA = 0 (63) (64) with the convention αnA = en n αnA .n α1m α2n |0i. |Λas i ≡ 1 − 1 2 N12 N21  (68) |Λi. 2 1 |Λsym i = √ Λsym mn α1m α1n |0i. The covariant D’Alembertian is 2 + ωm mn Dn . Gauge symmetries in AdSd Let us now analyze the situation in the AdS case.L. / Nuclear Physics B 586 (2000) 183–205 199 SB and 2 = D m Dm are defined via (58) and (59) with the flat space vielbein where DA .n having no definite symmetry properties |Λi = Λm. (69) . (48) take the form δas |Φi = D1 |Λas i. 2 Now the gauge transformations (47). The Lagrangian (60) is invariant under two gauge symmetries generated by the antisymmn metric gauge parameter Λmn as and the symmetric gauge parameter Λsym which can be conveniently described as Fock vectors 1 |Λas i = √ Λas mn α1m α2n |0i.

A straightforward but rather tedious computation with the use of the identities collected in the appendix leads to the following result Z  S1 |Λi e hΦ|(f + 3)D1 + 32 (d − 5 − g)P11 D δS ΦΦ = −λ2 AdSd S1 + hΦ|(6 − 3d − f )D2 + 3(d − 3)P21 D  + 3 (1 − d − g)P21 D1 PS11 |Si . The best one can do is to find a Lagrangian invariant either with respect to the gauge symmetry with the parameter |Λas i or the one with |Λsym i. This is in . the case with gauge invariance with respect to |Λsym i does not lead to unitary dynamics. / Nuclear Physics B 586 (2000) 183–205 The gauge transformation δ|Φi which respects the constraints (54)–(56) is δΛ |Φi = D1 |Λi − D2 |Λsym i. we set  1 S1 − D2 D S2 + 3 P11 −D2 + λ2 (d − 5) PS11 LΦΦ = hΦ|D2 + 3λ2 − D1 D 4 2  3 S12 + D12 PS11 + P11 D2 D S2 PS11 |Φi. The most general deformation of (60) to the AdS case without higher derivatives is of the form  1 S1 − D2 D S2 + 3 P11 −D2 + gλ2 PS11 LΦΦ = hΦ|D2 − f λ2 − D1 D 4 2  S12 + D12 PS11 + P11 D2 D S2 PS11 |Φi. g = d − 5. 2 (73) From this expression it is clear that the freedom in the parameters f and g is not enough to warrant an action invariant under both types of symmetries.200 L. Since. (74) since the antisymmetric part of the gauge parameter enters only via the first term. (60)) to AdSd . Brink et al. (70) It can equivalently be written as a combination of the gauge transformations with symmetric and antisymmetric gauge parameters δΛ |Φi = δas |Φi + δsym |Φi (71) with the gauge transformations of the form (61) and (62) but now with the derivatives D1 and D2 as in AdSd . we focus on the Lagrangian possessing the |Λas i invariance. + 34 P11 D (72) where f and g are arbitrary parameters. Thus. according to the general analysis of unitary representations of AdSd in [25]. the Lagrangian (75) describes more degrees of freedom than the original flat-space Lagrangian we started with. Next we analyze whether there exists a Lagrangian that generalizes (46) (equivalently. + P11 D (75) 4 Because one of the gauge symmetries is lost. From (73) it is obvious that this is achieved by setting f = −3. Note that at the level of equations of motion in the Lorentz gauge an analogous phenomenon was observed for redundant gauge symmetries in [25].

o. Since this field should describe a massless spin 2 field in the flat limit. The gauge fixing |χi = 0 is impossible however for λ = 0. as expected. It turns out that this is indeed possible. The final result is that the action Z   e LΦΦ + LΦχ + Lχχ . Therefore it describes properly the irreducible AdS representation. Brink et al. one has to add a massless spin two field analogous to a graviton field. This phenomenon can be interpreted as some sort of nonanalyticity of the flat limit exhibited already at the free field level. Therefore. of massless AdS fields may not be described by an irreducible representation of o(d − 2). the cross term LΦχ becomes nontrivial so that the system does not decompose into a sum of elementary subsystems.f. (77) S= AdSd where 3 S1 + P11 D2 PS11 |χi. This is why the naive flat limit of the action (75) describes not two fields but only one in agreement with [30]. (81) Note that the Lagrangian LΦχ is proportional to λ and tends to zero in the flat limit. (80) δ|χi = D1 |ξ i + λ|Si. the action reduces in the flat limit to the sum of two actions for the irreducible fields.L. In the AdS case. n. LΦχ = (d − 3)λhΦ| − 2D2 + 2P12 D 2  3 S1 − 1 P11 D2 + λ2 PS11 Lχχ = (d − 3)hχ|D2 + dλ2 − D1 D 2 2  S12 + D12 PS11 |χi + 12 P11 D (78) (79) is invariant under the gauge transformations of the form δ|Φi = D1 |Λi − D2 |Si + λ(P12 − P11 N21 )|ξ i. it has its own gauge symmetry with the gauge parameter |ξ i ≡ ξm α1m |0i. / Nuclear Physics B 586 (2000) 183–205 201 agreement with the general conclusion of Section 3 that physical d. . The resulting gauge fixed action is nothing but the action (75) invariant under the gauge symmetry with antisymmetric gauge parameter. (76) The idea is that starting from the sum of the free Lagrangians LΦΦ + Lχχ one should add cross terms LΦχ which (i) reestablish all (appropriately deformed by λ-dependent terms) gauge symmetries with the parameters |Λi and |ξ i and (ii) tend to zero in the flat limit. described by the Fock vector |χi ≡ χmn α1m α1n |0i. in order to make the AdS deformation consistent for the case under consideration. The conjecture of Section 3 suggests that the flat space and the AdS dynamics can match only once one starts with specific (reducible) collections of fields in flat space. Let us therefore introduce the field χ mn symmetric in indices m. From the general analysis of Section 3 it follows that. Another comment is that according to (81) the field |χi becomes a Stueckelberg field that can be gauged away for λ 6= 0. however.

/ Nuclear Physics B 586 (2000) 183–205 We expect that one can analogously find a deformation of the flat space Lagrangian (60) to AdSd by adding an antisymmetric second rank gauge tensor. A similar phenomenon is expected to be true for more complicated Young diagrams: only the sets of fields predicted in Section 3 will have all signs of kinetic terms of the flat-space Lagrangians correct. we are left with  (84) D2 + 3λ2 |Φi = 0. (82)  S2 − D1 PS12 )|Φi = 0. Brink et al. . incompatible with unitarity).e. Taking into account that one can identify the Stueckelberg field |χi with |Si one Si |χi = 0. There is a leftover symmetry with the parameter |Si satisfying certain differential conditions. because the corresponding representation of the AdS algebra is not unitary. the tracelessness condition PSij |Φi = 0 and the By imposing the Lorentz gauge D condition |χi = 0. This would correspond to keeping the symmetry with symmetric parameters. Plugging these values into (86). 1. . .2 found in [25]. the resulting gauge invariant Lagrangian is expected to have a wrong relative sign of the kinetic terms of the elementary flat-space Lagrangian (i. 0). 0.202 L. (87) we indeed arrive at the equations (84) and (85). (87) . . s = (2.2 and the conditions on the leftover gauge parameter |Si D2 − λ2 (s2 − 2)(s2 − 3 + d) + λ2 X ! sA − λ2 |Si = 0 A=1. However. s) ! X 2 2 sA |Φi = 0. Let us compare these results with the equations for the gauge field corresponding to the AdSd massless representation D(E0 . (86) D − E0 (E0 + λ − dλ) + λ A=1. For the case under consideration the E0 and s are E0 = λ(d − 1). The equations of motion for the fields |Φi and |χi that follow from the Lagrangian (77) can be reduced to the form S1 − D2 D S2 + 1 D12 PS11 + D2 D1 PS12 D2 − D1 D 2  − λ2 P11 PS11 − 2λ2 P12 PS12 + 3λ2 |Φi S1 + P11 N21 D S1 + λ (d − 3)(D1 N21 − 2D2 ) − P12 D  + (P12 D1 − P11 D2 )PS11 |χi = 0.. can derive these conditions from the equations of motion for |χi in the gauge D S Pij |χi = 0  (85) D2 + dλ2 |Si = 0. S1 + 1 D12 PS11 − 1 P11 PS11 + dλ2 |χi + λ(D D2 − D1 D 2 2 (83) Si |Φi = 0.

It is tempting to speculate that there may exist some new version of gravity associated with this type of field. On the other hand we have argued that totally antisymmetric tensors can never result from the flat limit decomposition of other types of AdSd unitary representations. Metsaev and M. Grant No. From this conjecture it follows that there is a unique nontrivial situation when a flat space spin two massless field appears as a result of a nontrivial reduction of AdSd massless field with mixed symmetry properties. This example has been considered in detail. Gelfond for the help in LATEX drawing of Young diagrams. Conclusions We have shown that generic irreducible massless (gauge) fields in AdSd in the flat limit decompose into nontrivial sets of irreducible flat space massless fields. Albert Einstein Institute where some part of this work was done. Algebra of commutators In this appendix we collect some formulas that are used in the computations of the Section 4. 960538 and Grant No. ! X   2 2 S NCC + λ2 (1 − d)NBA DA . We made a conjecture on the pattern of the flat-space reduction of a generic AdSd massless field. Vasiliev would also like to express his gratitude to Prof. 99-0590 and by the RFBR Grant No. the space of differential forms is closed with respect to the flat space limit decomposition.2. In other words. DB = δAB D + λ +λ 2 X C C  PBC PSAC − NCA NBC . in particular. Appendix A. The authors are grateful to O. Vasiliev would like to thank for hospitality at Chalmers University. Brink et al. M. Brink is supported by NFR F 650-19981268/2000. / Nuclear Physics B 586 (2000) 183–205 203 5. M. In that sense AdSd massless fields are “less massless" than flat space massless fields. An interesting problem for the future is to generalize these results to the supersymmetric cases to analyze generic AdS supermultiplets in higher dimensions and. however.L. in AdS11 . 99-02-16207. smaller than the result of a dimensional reduction to one less dimension of a corresponding massive field. R. Metsaev is also supported by DOE/ER/01545-787. Vasiliev was also supported by NFR F-FU 08115347. These sets are. The work of L. Vasiliev is supported in part by INTAS. Acknowledgements The work of R. Metsaev and M. Nicolai for the hospitality at the MPI für Gravitationsphysik. H. R. (88) . Another problem is to consider the multiplets occurring in [15] to see how they group themselves in the case of AdS.

PCE = δBC NEA + δBE NCA + δAC NEB + δAE NCB + d(δBC δAE + δBE δAC ) is elementary. Vasiliev. Fortschr. Polyakov. [13] M. PAC D C  SC − DC PSAC . 2 (1998) 253. Dubna.D 2 2 X  SC − DC NAC . b D . D   SA + δAC D SB . Klebanov. Ann. [5] E. hep-th/0002008. DB ] = λ2  C  2  2  2 SB = λ SA .). World Scientific. Rev. Brink et al. Phys. Phys. D 18 (1978) 3630. D D   X (PBC NAC − PAC NBC ). Fronsdal. [10] E. hep-th/9802150. in: M. [14] M. 1999. Fradkin. The derivation of the following relationships: [NAB . Ann.A. Fortschr. [11] M. Shifman (Ed. [6] M. Nucl. Lett. M. B 285 (1992) 225. Phys. 2 (1998) 231. 35 (1987) 741.A.204 L. P. Lett. Lett. B 291 (1987) 141. PBC = δAB DC + δAC DB . [9] S.A. Phys. World Scientific. A. Fronsdal.A. Vasiliev. Russia. Grav.A.R. X  l NCB PSAC − NCA PSBC . B 257 (1991) 111.   SC . Brink. NCE ] = δBC NAE − δAD NCE . Vasiliev. Phys. Phys. PSAB . M. Contributed article to Golfand’s Memorial Volume. Rep. Theor. hep-th/9910096. light cone supersymmetry. July 27–31. D   SA . Phys. Quant. . P. hep-th/9908208. Phys. and superconformal algebras. Phys. [12] M. Phys. Dirac equations. [3] M. Vasiliev. Adv. Adv. hep-th/9711200. Shifman (Ed. B 301 (1988) 26. Phys. 315 (1999) 137. Lett. Rev.S.R. Math. Phys. Phys. 177 (1987) 63. 36 (1988) 33. [8] J.A. NBC = δAB D [NAB . 8 (1991) 1387.M. Fang. Class.S. Pengpan. B 428 (1998) 105. Theor. DC = δBC D   PSAB . Based on talk given at International Workshop Supersymmetries and Quantum Symmetries. Contributed article to Golfand’s Memorial Volume. [2] J. Math. D 22 (1980) 1361. hep-th/9802109. / Nuclear Physics B 586 (2000) 183–205 [DA . Vasiliev. Rev. DA = λ (1 − d)DA + 2λ2 2 SA = λ (d − 1)D SA + 2λ D . Phys. D 20 (1979) 848. [7] E. These formulas can be where derived by straightforward but sometimes lengthy calculation. Vasiliev. Vasiliev. B 243 (1990) 378. Fradkin. Lett.A. NCA D X C D2 in (88) is a covariant D’Alembertian operator (67). Metsaev. Maldacena. Gubser. I. DC ] = δBC DA . M. Phys.). Vasiliev. Phys. Ramond. References [1] C.S. [4] R. SA . Higher spin gauge theories: star-product and AdS space. B 189 (1987) 89. IIB supergravity and various aspects of light-cone formalism in AdS space–time. Nucl. [16] L. Phys. 190 (1989) 59. [15] T. hep-th/9808190. Witten. D 18 (1978) 3624. Ramond. Rev.A. C. Phys.

Evans. Nucl. [22] P. [25] R. in: B. Poland. Ogievetsky. Gunaydin. Vasiliev. 1984. Metsaev.O. B 322 (1989) 185. Phys. M. 4 (1963) 901.F. [29] A. Polanica. Nicolai.A. S.R. 55 (1977) 1. Lett. hep-th/9906217. Lett. [27] V. [28] J. 1984. hep-th/9810231. Lectures given at 35th Winter School of Theoretical Physics: from Cosmology to Quantum Gravity. Singapore. Warszawa. J. 8 (1967) 170. 22–26 July. Lopatin. hep-th/9802047. I. Nucl. [24] M. Koh. hep-th/9908005.G.A. 2–12 February 1999. on Supermembranes and Physics in (2+1) Dimensions. 1977. Invited talk given at Trieste Conf. [23] M. Singleton and doubleton supermultiplets of space–time supergroups and infinite spin superalgebras. Supersymmetry and supergravity.E. Talk given at Dubna International Seminar ‘Supersymmetries and Quantum Symmetries’ dedicated to the memory of Victor I. PWN. de Wit. Dirac. Mack. B 523 (1998) 145. D 12 (1975) 3819.). Arbitrary spin massless bosonic fields in d-dimensional anti-de Sitter space.M. [19] C. [21] N. 1989. de Wit. Brink et al. Raczka. Dubna. Phys. Representations of supersymmetry in anti-de Sitter space.R. D. Phys. Anti-de Sitter Supersymmetry. R. Labastida. Math.L. 1997. Lett. J. Phys. B 354 (1995) 78.M. Italy. Aulakh. World Scientific. [18] B. July 17–21. . Metsaev. B 563 (1999) 295.T. Math. Phys. Herger. Phys. I. van Nieuwenhuizen (Eds. Ouvry. Barut. Math. Nucl. A 3 (1988) 257. Mod. Phys. Phys. Phys.S. Fronsdal. Theory of Group Representations and Applications. Fayet. Comm. Minic. Gunaydin. / Nuclear Physics B 586 (2000) 183–205 205 [17] H. [20] R. Rep. Phys. P. P. Trieste. B 173 (1989) 284. [26] G. [30] C.

Università di Torino and INFN.25. E-mail addresses: gallot@to. They form short (or ultra short) supersymmetric multiplets and. are stable and protected from quantum radiative corrections.infn. we find perfect agreement. Introduction The spectrum of supersymmetric string theories usually contains a special class of states known as BPS states.3 2 1 3 ( 0 0 ) 0 0 4 0 6 . Gallot). Gallot a . Lerda). Lerda a. The gravitational interactions are obtained using the boundary state formalism while the SO(32) gauge interactions are determined by evaluating disk diagrams with suitable insertions of boundary changing (or twist) operators. I-15100 Alessandria. A. After taking into account the known duality and renormalization effects. D-branes. PACS: 11. Italy b Dipartimento di Scienze e Tecnologie Avanzate. Boundary states 1.∗ . P.5 . A well-known example of such BPS states is provided by ✩ Work partially supported by the European Commission TMR programme ERBFMRX-CT96-0045. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.Hf Keywords: String theory. I-10125 Torino.infn.  2000 Elsevier Science B. Strigazzi a a Dipartimento di Fisica Teorica. Italy Received 1 May 2000. by MURST and by the Programme Emergence de la région Rhône-Alpes (France). In particular the gauge coupling of a D-particle is obtained from a disk with two boundary components produced by the insertion of two twist operators. because of this fact.elsevier. Via P. lerda@to. strigazzi@to.infn. All rights reserved.it (A. thus confirming at a non-BPS level the expectations based on the heterotic/type I duality.b.it (L. PII: S 0 5 5 0 . ∗ Corresponding author. which are characterized by the property that their mass is completely determined by their charge under some gauge field.Nuclear Physics B 586 (2000) 206–230 www. All rights reserved.V. Giuria 1.it (P. accepted 21 June 2000 Abstract We study the gauge and gravitational interactions of the stable non-BPS D-particles of the type I string theory. Università del Piemonte Orientale. Sezione di Torino.V.nl/locate/npe Gauge and gravitational interactions of non-BPS D-particles ✩ L. Strigazzi). We also compare our results with the amplitudes among the non-BPS states of the heterotic string which are dual to the D-particles.

then. they are stable since they cannot decay into anything else. 1 The mass of these D-particles turns out to be 1 For the description of other non-BPS D-branes using the boundary state formalism see Refs. Such states do indeed exist and were identified by Sen as the non-BPS D-particles of type I [4. among them we recall the fact the non-BPS D-branes might be useful for analyzing the non-perturbative properties of the non-supersymmetric field theories that live on their world-volumes. has attracted a lot of interest during the last year (for reviews see Refs. The study of the stable non-BPS D-branes in string theory. One of the most notable examples of stable non-BPS configurations is provided by the perturbative states at the first excited level of the SO(32) heterotic string [10] which carry the spinor representation of the gauge group and whose mass is given by 2 gYM . an explicit boundary state description for them was provided in Ref. these states are non-BPS. [11] which are also reviewed in Appendix B. pioneered by Sen in a remarkable series of papers [2–6]. also in the type I theory there should exist stable non-BPS configurations that are spinors of SO(32).L. However. Mh = √ = 0 κ10 α (1. or equivalently they should appear as non-perturbative (D-brane type) configurations in the weakly coupled dual theory. In this process however. but being the lightest ones carrying the spinor representation of SO(32). or the fact that they may lead to novel types of relations among string theories [7]. [13]. Usually. [14–18]. Thus we can write Mh = gYM f. it is not difficult to find such non-BPS states with the standard string perturbative methods and analyze their properties at weak coupling. Gallot et al. Being at the first massive level. but. To verify the existence of these non-BPS states is therefore a very strong test on the duality relations between two string theories which does not rely on supersymmetry arguments. . they are stable and should be present also when one increases the heterotic string coupling constant gh . For them there is no particular relation between their mass and their charge.2) where the renormalization function f can in principle be computed perturbatively in the heterotic string and is such that f ∼ 1 for gh → 0. the mass Mh gets renormalized since there are no constraints on it coming from supersymmetry. since they cannot decay. being the lightest states with a given set of conserved quantum numbers. κ10 (1. supersymmetric string theories quite often contain states that are stable without being BPS.1) In the last equality we have introduced the low-energy gauge and gravitational couplings gYM and κ10 of the heterotic string following the conventions of Ref. they should be present also in the strong coupling regime. However.5]. If the heterotic/type I duality [12] is correct. [7–9]) also for several other reasons. These are in general the lightest states which carry some conserved quantum numbers. they form long multiplets of the supersymmetry algebra and receive quantum radiative corrections. / Nuclear Physics B 586 (2000) 206–230 207 the supersymmetric Dp-branes of the type II theories (with p even in type IIA and p odd in type IIB) [1].

the interactions of the non-BPS D-particles must be obtained using less standard methods and have not been fully investigated so far.2) and (1. we cannot use any more the boundary state formalism since this accounts only for the couplings of the D-particles with the closed strings that live in the bulk. that. gYM and κ10 are. we find that the gravitational potential energy of two D-particles exactly agrees with the one of their heterotic duals. On the other hand. by taking the large distance limit to which only graviton and dilaton exchanges contribute. we can deduce that the renormalization function f must be such that f ∼ 2−3/4 gI −1/2 for gI → 0 in order for the masses to agree on both sides.3) where gI .208 L. Clearly. only the general rules for computing string amplitudes with these D-particles have been given in the literature [5. we will concentrate on processes involving massless string modes that are responsible for the long range interactions among D-particles. κ10 (1. by the Newton’s law and the Coulomb’s law for massive and charged point-like objects in ten dimensions. and require the use 2 The T-dual heterotic/type I’ correspondence has been analyzed at the non-BPS level in Ref. On the type I side. . For the gauge interactions. the string. [11] (see also Appendix B). to study the gauge interactions of our D-particles we should consider scattering amplitudes involving open strings with mixed boundary conditions in an odd number of dimensions. instead. In this way one can show. and remembering that under the duality map the heterotic gauge and gravitational couplings turn into the corresponding ones of type I. respectively.20]. this result cannot be obtained using perturbative methods. respectively. / Nuclear Physics B 586 (2000) 206–230 Me0 = √ 1 α 0 gI = gYM −3/4 −1/2 2 gI . Then. these can be easily obtained using standard perturbative techniques from correlation functions of vertex operators. Gallot et al. On the heterotic side.26] (for recent reviews on the boundary state formalism and its applications see Ref. [19]. Comparing Eqs. Calculations of open string amplitudes with mixed boundary conditions have already appeared in the analyses of systems of several D-branes with different dimensionality (see for instance Ref. the situation is a bit more involved. 2 In this paper we elaborate further on these stable non-BPS particles and study in detail their gravitational and gauge interactions.3). Therefore. the open strings attached to the D-particles have Neumann boundary conditions only along the time direction. but is completely blind to the other bulk sector of the type I theory consisting of open strings with Neumann boundary conditions in all directions to which the SO(32) gauge fields belong. (1. In fact. [27]). It is the purpose of this paper to fill this gap. but is a prediction of the heterotic/type I duality. In particular. instead. indeed. [28]). the gravitational and gauge potential energies of two such particles at large distance are given. provided that the duality relations and the mass renormalization previously discussed are taken into account. at the lowest order in the heterotic string coupling constant. the gauge and the gravitational coupling constants of the type I theory in the conventions of Ref. for example. To study the gravitational interactions we adopt the boundary state formalism [21–24] and obtain the energy due to the exchange of closed string states between two D-particles by simply computing the diffusion amplitude between the two corresponding boundary states in relative motion [25.

Finally. 32) labeling the fundamental representation of the SO(32) gauge group. and then determine the gauge potential energy between two D-particles. weighted by Chan–Paton factors 0 and 01 . an important check of the heterotic/type I duality has been the discovery by Sen [3–5] that the stable non-BPS heterotic states carrying the spinor representation of SO(32) at the first massive level are dual to the non-BPS D-particles of type I. Appendices B and C contain our conventions and a list of more technical formulas. This paper is organized as follows: in Section 2 we compute the gauge coupling of a non-BPS D-particle of type I by evaluating a disk diagram with two twist insertions. Finally. and have been recently reconsidered from an abstract conformal field theory point of view [31].20]. / Nuclear Physics B 586 (2000) 206–230 209 of twist operators to produce mixed boundary conditions in certain directions. due to the existence of an odd number of fermionic zero modes. . Special care is required in these calculations because the twist operators that we use change the boundary conditions in an odd number of directions. Gallot et al. In Appendix A we show how to compute the gauge interactions between two BPS D-strings of type I by extending the method of Section 2 and verify the no-force condition. In Section 3 we use the boundary state formalism to compute the gravitational contribution to the potential energy of two (moving) D-particles. and discuss their relations. [5. Besides these factors. In particular. Specific rules for computing amplitudes involving such D-particles have been given by Sen [5] and Witten [20] in two different ways which we briefly recall here. Sen’s approach heavily relies of the use of Chan–Paton factors to distinguish the various kinds of open strings.30]. namely the non-BPS D-particles couple minimally to the gauge field. the latter a Chan–Paton factor σ1 . the 9–9 strings are the . The result of this calculation is extremely simple. we will describe how to compute scattering amplitudes involving non-BPS D-particles and bulk open strings of type I. Using such twist operators and applying the rules of Refs. this exactly agrees with the corresponding energy computed in the heterotic theory. . . Exploiting this fact. we will explicitly determine the gauge coupling of the D-particles by evaluating a correlation function on a disk with two boundary components produced by the insertion of two twist operators. respectively. the former carry a Chan–Paton factor 1. we then determine the gauge potential energy of a pair of D-particles at large distance and see that after taking into account the duality map. contain both states that are even and states that are odd under (−1)F . their vertex operators comprise the standard GSO-even  as well as the corresponding   part 1 GSO-odd part. Type I D-particle interactions: the gauge amplitude As we mentioned in the introduction. . These twist operators were used in the past to study strings on orbifolds [29.L. The 0–0 strings. 2. The 9–0 strings stretching between one of the 32 D9-branes of the type I background and a D-particle contain only (−1)F even states but. the 9–0 strings also carry a Chan–Paton factor λA (A = 1. In Section 5 we compare the results for the non-BPS D-particles and for their dual heterotic states. In Section 4 we study the gauge and gravitational interactions of the non-BPS heterotic states that are dual to the D-particles. whose end-points lie on the non-BPS D-particle.

which are present also on the BPS D0 brane of the type IIA theory. in the 9–0 sector one recovers an even number of fermionic zero modes and can perform the usual GSO projection. 1) generated by ψ00 and η. / Nuclear Physics B 586 (2000) 206–230 usual open strings of the type I theory which are GSO projected and carry only the standard Chan–Paton factors of the gauge group. In this way. These modes. In this case. We now turn to the 9–0 strings which are more relevant for our purposes. Let us now consider the R sector of the 0–0 strings. we just consider the R sector. . but since the extra fermion η is odd under this GSO parity. these vertex operators are simply = ψ i e−φ . [20] Witten has proposed to introduce an extra one-dimensional fermion η on each boundary of the string world-sheet lying on a D-particle. Upon quantization. 10). (2.2) 2 where S α is the spin field of conformal dimension 10/16 associated to the ten world-sheet fermions. We start with the NS sector of the 0–0 strings where at the massless level there are nine scalars x i (i = 1. similarly to what found by Sen. of conformal dimension 9/16. Vx(−1) i (2. Thus. 10 or 01 shows that the states of the 0–0 and 9–0 sectors have a non-trivial structure which is really due to the presence of an odd number of fermionic zero modes.1) Notice that there is no factor of eik·X in (2. Since the NS sector does not contain massless states. 1) (1 + ψ00 η)/2. • a GSO (or chirality) projector for the Clifford algebra of SO(1. Hence. correspond to vertex operators Vx i that do not depend on the boundary fermion η.1) because massless states of 0–p strings have no momentum. one obtains two types of 0–0 states. . Also in the 0–0 sector one performs a (generalized) GSO projection to obtain physical states. 9) corresponding to the freedom of moving the D-particle in its nine tranverse directions.2) account for the 216/2 = 256 degeneracy of the non-BPS D-particle. Let us now give some details on how to construct the massless states in the various open string sectors using Witten’s rules. of conformal dimension zero. In order to remedy to this oddity. Here both the ten worldsheet fermions ψ µ and the boundary fermion η possess zero modes so that the massless R states form a GSO-even spinor ξ α of SO(1. of conformal dimension 1/16. Gallot et al. the vertex operator describing the massless modes of the 9–0 sector should contain • a spin field S associated to the fermion ψ 0 . 10) by adding η. 9) to SO(1. the only world-sheet fermion to have a zero mode is ψ 0 so that the ground state is a GSO-even (chiral) spinor of the algebra SO(1. in Ref. (−1/2) Vξ α = . in the (−1/2) superghost picture the vertex operator for the massless R states reads 1 + η α −φ/2 S e .210 L. In the (−1) superghost picture.     The presence of the unusual Chan–Paton factors 1. . Note that in this case the GSO projection is simply the ten-dimensional chirality projection which is natural when one extends SO(1. the 16 massless fermionic modes described by (2. • a boundary changing operator ∆ for the nine space directions transverse to the D-particle. σ1 . .

30] which allow emissions with non-zero momentum in the transverse directions. on the other hand. in particular in the NS sector at the massless level we find the SO(32) gauge bosons which are described by the following vertex operators in the (−1) superghost picture (−1/2) V09 (−1/2) = ΩV90 = λt A (−1) = ΛAB Aµ ψ µ eik·X e−φ . The obvious choice is then to make insertions of the vertices V90 and V09 given in (2.4). Since in type I the strings are unoriented. as we mentioned above. Notice that the GSO projection in (2.3) keeps only one fermionic degree of freedom for each value of the index A of the fundamental representation of SO(32). the twist operators ∆ are reservoirs of transverse momentum [29. Note also that. This is merely related to the 9–0 sector through the action of the world-sheet parity Ω.5) where ΛAB are the generators of SO(32) in the fundamental representation (see Appendix B for our conventions) and Aµ is the polarization vector. Recalling that Ω simply acts by transposition on the Chan–Paton factors without changing the physical content of the vertex operators. we should consider also the 0–9 sector. (2. of conformal dimension 3/8.4) 2 Notice in particular that the SO(1. / Nuclear Physics B 586 (2000) 206–230 211 • a superghost contribution in the −1/2 picture e−φ/2 .L.3) 2 It is easy to check that the operator (2. this is to be expected because the presence of a D-brane breaks the translational invariance in transverse space. and hence we can conclude that the marginal operator (2. We now face the problem of finding the coupling between the non-BPS D-particle and the gauge field. Vgauge (2. (2.3) has indeed conformal dimension 1 as it should be for a physical vertex operator. are the usual open strings of the type I theory. We thus have to insert two vertices containing the boundary changing operator ∆ which turns a boundary of type 0 into one of type 9 (or viceversa). Since the latter belongs to the 9–9 massless sector. due to the insertion of the two vertices V90 and V09 . • a gauge Chan–Paton factor λA to specify which of the 32 D9-branes one is considering.4) which correspond to the SO(32) degeneracy of the D-particle. 1. although no momentum is carried by these vertices. there are the 9–9 strings which. the emitted gauge boson may have non-zero space momentum. This is represented in the Fig.3) and (2.3) and (2. Finally. Thus. Indeed. Upon quantization. we have 1 + ψ00 η S∆ e−φ/2 . Gallot et al. the diagram we have to compute corresponds to a disk with a part of its boundary on the D-particle and a part on the D9-branes from which the gauge boson is emitted. we have (−1/2) V90 = λA 1 + ψ00 η S∆ e−φ/2 . Note that. the boundary component associated with the D-particle carries indices in the bi-fundamental .3) accounts for the SO(32) degeneracy of the non-BPS D-particle. 1) GSO projection is the same in both vertices (2. the states described by V90 form a spinorial representation of SO(32).

(2. both massless and massive. . Gallot et al. as it should be for a minimally coupled particle at rest. / Nuclear Physics B 586 (2000) 206–230 Fig.32] for their derivation. perturbative open string states which group in the adjoint or in the symmetric representation of SO(32). as we will explicitly see in the following.9) ∆(z1 )ψ i (z2 ) eik·X(z2) ∆(z3 ) = 0 for i = 1. F gauge = hΩ09 |Vgauge (2. . The correlation function in (2. This is consistent with the fact that a D-particle should emit all.3) and (2. 1. The gauge coupling of a (static) D-particle is then given by the expectation value of the gauge boson emission vertex (2. and NNS for the NS vertex (2. 1 corresponds to (−1) |Ω90 i. . the normalization factors NR for the R vertices (2.212 L. . [11. there is no emission of gauge bosons with polarization Ai along the transverse directions. the expectation value of the gauge emission vertex is not vanishing. 9. [20]. Thus.6) where (−1/2) |Ω90 i = lim V09 z→0 (−1/2) (z) |0i and hΩ09 | = lim h0| V09 z→∞ (z).4). The disk diagram describing the gauge coupling of a type I D-particle. The notation h iη means that the correlator must be evaluated by including the action for the boundary fermion η as explained in Ref.8) F where we have also added a ghost c in each vertex operator. 3 F gauge may be reexpressed as a 3-point function on the world-sheet and reads Z (−1/2) (−1/2) gauge 2 (−1) = Cdisk NR NNS dµ(zi ) c V09 (z1 ) c Vgauge (z2 ) c V90 (z3 ) η .7) Note that due to the presence of the twist operators in V09 and V90 . .5) in the “vacuum” representing the D-particle. (2.5). After including the normalization factor Cdisk appropriate of any disk amplitude. the diagram of Fig. Thus. We then consider the longitudinal part for which the basic correlators are 3 We refer to Appendix B for the explicit expression of these normalization factors and to Refs. The latter vanishes because (2.8) may be decomposed into a longitudinal and a transverse piece. representation of SO(32).

13) 2 2 η This splits into four pieces.17) represents the amplitude for the emission of a gauge boson with longitudinal polarization ξ0 and color index (BC) from a 0-boundary in the bi-fundamental of SO(32). −1/2 −1/4 −1/2 −φ(z1 )/2 −φ(z2 ) −φ(z3 )/2 e e = z12 z13 z23 . (2. Inserting Eqs.L. we easily get   0 2 Z  Cdisk NR2 NNS z13 α k gauge t A BC D Tr λ Λ λ A0 dµ(zi ) =− F √ z12 z23 2  Cdisk NR2 NNS A Tr λt ΛBC λD A0 . the magnetization) while the other spin field ψ00 S plays the role of the disorder parameter µ [33]. The appearance of this representation is a direct consequence of our construction in which the D-particle is represented by the 0-component of a disk boundary produced by the insertion of the vertex operators V90 and V09 . The remaining correlator to be considered is     1 + ψ00 η 1 + ψ00 η S(z1 ) ψ 0 (z2 ) S(z3 ) . =− (2. (2.10)–(2. Eq. we can rewrite F gauge as follows  gYM (2. using the explicit expressions of the normalization coefficients and Chan–Paton factors given in Appendix B.11) (2. the SO(32) spinor degeneracy of the non-BPS D-particle of type I arises from the (second) quantization of the 32 massless fermionic zero-modes of the 0–9 open strings. the spin field S may be identified with the order parameter σ (i.8) and exploiting the projective invariance to fix the position of the three punctures at arbitrary values.e.17) F gauge = −i √ δ AB δ CD − δ AC δ BD A0 . a boundary state is a single state that correctly represents . [20] the only non-vanishing correlation functions are those containing one factor of η. This fact should not be surprising because a completely analogous situation occurs in the familiar description of Dp-branes using boundary states. we have 0 −1/2 3/8 −1/2 ψ0 S(z1 ) ψ 0 (z2 ) S(z3 ) = S(z1 ) ψ 0 (z2 ) ψ00 S(z3 ) = z12 z13 z23 .10) (2.12) the transverse momentum k i of the emitted gauge boson is not subject to any constraint. Indeed.15) into (2. (2. Indeed.12) Notice that in (2.16) √ 2 Then. In particular one has √ h1iη = 0. (2. Indeed. Gallot et al.15) Notice that a correlation function similar to (2. 2 where gYM is the gauge coupling constant of the type I theory. and thus it is clear that such a degeneracy cannot be seen in our operator formalism.14) hηiη = 2. two of which vanish.. e −α 0 k 2 α 0 k 2 −9/8 −α 0 k 2 z13 z23 . (2. as we have anticipated. On the other hand. / Nuclear Physics B 586 (2000) 206–230 hc(z1 ) c(z2 ) c(z3 )i = z12 z13 z23 .15) appears in the 2D Ising model. Finally. ∆(z1 ) eik·X(z2) ∆(z3 ) = z12 213 (2. according to Ref.

as required by the heterotic/type I duality.214 L. As indicated in Fig.17) we can now easily compute the gauge potential energy VI due to the exchange of the SO(32) gauge bosons between two D-particles. gauge Using the result (2. 2. δ AE δ F D − δ AF δ DE 2 7Ω8 r 7 (2. In fact. after taking into account the known duality relations. The diagram describing the gauge amplitude between two D-particles. even if does not account for its degeneracy under the gauge group. (2. 2. 2 q (2. using this construction we are able to obtain nontrivial information about the gauge interactions between two D-particles at large distance. Gallot et al.18) yielding gauge VI =− 2 1 gYM δ AE δ F D − δ AF δ DE 2 . thus confirming the validity of our construction. Fig.20) where Ωq = 2π (q+1)/2/Γ ((q + 1)/2) is the area of a unit q-dimensional sphere. δ δ 2 q (2. even if it does not account for the degeneracy of the Dp-brane under the supersymmetry algebra. in our case the 0-component of the disk boundary produced by the insertion of V90 and V09 correctly describes a D-particle and allows to obtain its coupling with the bulk 9–9 open strings. . we get the following (static) gauge potential in configuration space gauge VI (r) = − 2  1 gYM . as we will see later and in the following sections.20) can be obtained also using the rules given by Sen in Ref. Eq. [5] for computing amplitudes with nonBPS D-particles.17) and (2. / Nuclear Physics B 586 (2000) 206–230 a Dp-brane and its couplings to the bulk closed strings.20) clearly represents a “Coulomb-like” potential energy for point particles at a distance r in ten dimensions. We conclude this section by mentioning that the same results (2. we will show that the results obtained in this way exactly agree with those in the heterotic theory. Similarly. this can be obtained simply by sewing two emission amplitudes F gauge with the gauge boson propagator P= ηαβ BB 0 CC 0 0 0 − δ BC δ CB .19) Performing a Fourier transform. Moreover.

we refer the reader to Refs. at the leading order in the string coupling constant.4) where a (e a ) is the left (right) intercept (aNS = 1/2. (3.24. In Eq. the boundary state for a BPS D-particle of type IIA may formally be written as 4 |D0iIIA = |D0iNS + |D0iR . For details and conventions on boundary states. namely T0 = 8π 7/2 α 0 3/2 . enforces on it the appropriate boundary conditions and represents the source for the closed strings emitted by the brane. we can write √ (3. P denotes the closed string propagator P= 1 α0 . if we consider the type I theory [34–37]. This is a closed string state which inserts a boundary on the world-sheet. In the closed string operator formalism.13] has √ instead only a component along the NS–NS sector and a tension Te0 greater by a factor of 2 than T0 . The partition function of such open strings may be obtained by evaluating the cylinder/annulus amplitude in the closed string channel which is given by IIA hD0|P |D0iIIA . (3.22].L. the Ω projection is implemented by adding the so-called crosscap state 4 In order to avoid clutter. R–R) component of a boundary state with the simplified subscript NS (respectively.2) The presence of both the NS–NS and the R–R components implies that the spectrum of the open strings living on the D-particle is GSO-projected. the tachyon is removed by the projection onto states invariant under the world-sheet parity Ω [3. However. R) .5) |De 0iIIB = 2 |D0iNS . aR = 0).1) where the NS–NS and the R–R components are both proportional to T0 which is the tension of the D-particle in units of the gravitational coupling constant. The boundary state for the non-BPS D-particle of the IIB theory [3. [13. / Nuclear Physics B 586 (2000) 206–230 215 3. Type I D-particle interactions: the gravitational amplitude The gravitational contribution to the scattering of two non-BPS D-particles of type I can be calculated.3).23. one describes a Dp-brane by means of a boundary state |Dpi [21. (3. As an example.3) and then performing a modular transformation. we shall denote the NS–NS (respectively. [5. The boundary state description of the non-BPS D-particles has already been given in Refs. Gallot et al. (3. e0 − a − e 2 L0 + L a (3. from the diffusion amplitude between two corresponding boundary states. As a consequence.13] from which we recall the results that are relevant in the forthcoming analysis.26]. In the boundary state formalism. Thus. there is no GSO-projection in the spectrum of the open strings lying on the non-BPS D-particle and the presence of a tachyon in the NS sector renders it unstable.13].

9) |De 0iI = √ 2 √ where we have added the same factor of 1/ 2 for consistency with (3. 2 The spectrum of open strings stretching between two different (distant) D-particles at rest. must be canceled by the introduction of 32 D9-branes. (3.8) 2 added to the torus contribution gives the partition function for unoriented closed strings. given by the sum of the annulus and the Möbius strip contributions.13) 2 where the factor of one-half indicates that. a tachyon develops in . Gallot et al. obtained by summing the contributions from the annulus and the Möbius strip. The boundary state of the non-BPS D-particle of type I reads  1 √ 2 |D0iNS = |D0iNS .6).6) √ |Ci + 32|D9i .11) NS hD0|P |D0iNS + 2 In this theory. the partition function for unoriented open 9–9 strings. Hence. at sufficiently small distance. only the Ω symmetric combinations are retained. The mass of the D-particle is then given by 1 Te0 T0 1 = =√ . 2 √ where the factor of 1/ 2 has been introduced to obtain the right normalization of the various spectra. is √  1 √ 2 hC|P |D0iNS + 2 NS hD0|P |Ci . one labeled with a prime. is  1 10 (3. has a partition function given by √  1 √ 2 ( 2 ) NS hD0|P |D00 iNS + ( 2 )2 NS hD00 |P |D0iNS . the type I theory possesses a background “boundary state” given by [22]  1 (3. / Nuclear Physics B 586 (2000) 206–230 |Ci [22]. Then. (3. compared to the IIB case. (3. Notice that.10) Me0 = √ κ κ 2 10 α 0 gI 10 where κ10 is the ten-dimensional gravitational coupling constant of the type I theory (see Appendix B). 2 while the contribution of the Klein bottle 1 hC|P |Ci (3. which corresponds to inserting on the closed string world-sheet a boundary with opposite points identified. The world-sheet parity Ω exchanges the two sectors 0–9 and 9–0 so that we only retain symmetric combinations corresponding to the partition function √  32 2 (3.7) 2 hD9|P |D9i + 25 hD9|P |Ci + 25 hC|P |D9i . The partition function for open 0–0 strings living on the D-particle. (3.216 L. there are also 0–9 and 9–0 open strings with one end on the D-particle and the other on one of the 32 D9-branes of the type I background. The negative (−32) charge for the non-propagating R–R 10form that the crosscap generates in the background.12) NS hD0|P |D9i + hD9|P |D0iNS .

Gallot et al. Our aim is to study the diffusion of a moving D-particle with a velocity v along one space direction. the interaction amplitude just reads AI (v) = NS hD00 |P Λ|D0iNS + NS hD0|Λ† P |D00 iNS . 6 See for instance Ref.14) reduces to the static one (3. This can be done generalizing the treatment for the BPS D-branes presented in Ref.15) is the acting on the boundary state of a particle at rest.26] and explicitly reads 6  1 2 0 −2 Z∞ AI (v) = 8π α Z∞ dτ −∞  × 2 sinh(πν) ds s −9/2 e− b2 +τ 2 v 2 γ 2 2π α 0 s 0  6 f3 (q) θ3(−iν|is) − f46 (q) θ4(−iν|is) f16 (q) θ1(−iν|is) (3. To do so we have to perform VI the limit s → ∞ in the integrand of Eq. Indeed. (3.18). Here we have v = th(πν) and generator of the Lorentz transformations. then integrate on the variable s and finally identify the potential energy according to 5 The signs ± correspond to the two possible implementations of boundary conditions for world-sheet fermions. (3. V01 = sinh(2πν) cosh(2πν) (3. only a suitable linear combination of them is retained. / Nuclear Physics B 586 (2000) 206–230 217 this open string spectrum signaling the instability of the configuration which decays into the vacuum [5].14) where Λ is the boost operator Λ = eiπνJ 01 (3. k k . We are now in a position to extract the long range interaction potential energy grav due to gravitational exchange between the two particles. (3. In a physical (GSO projected) boundary state. [39] for definitions and conventions about the modular functions f and θ . −18 ) with   cosh(2πν) sinh(2πν) . The interaction amplitude AI (v) can be evaluated using standard techniques [25.L. Such an interaction may be evaluated analyzing the spectrum of the open strings stretching between the two objects with modified boundary conditions in the 0 and 1 directions. say X1 . Notice that the amplitude (3.16) × exp ±i J µν r=1/2 µ=0 where the boundary conditions are encoded in the matrix S = (−V01 . but we find it simpler to use the method of the boosted boundary state [25.26]. The boosted boundary state 5 reads " ∞ # X T0 1 (8) 0 1 † † δ (x) δ(x v + x ) exp − an · S · e an Λ|D0.18) in which q = e−πs . [38].17) Note that cosh(πν) ≡ γ is the Lorentz factor. on another D-particle at rest at the origin.13) in the limit of vanishing velocity. ±iNS = 2 γ n=1 # 9 " ∞ X Y er† ψr† · S · ψ |k µ = 0i. τ is the proper time of the moving particle and b is the impact parameter.

Thus. The states of Eq. we will adopt the bosonized formulation of the heterotic string in which the gauge degrees of freedom are described by sixteen chiral eI (I = 1. . [µ ν ψ−1/2 ψ−1/2 ψ−1/2 |ki ⊗ |K I i.1) and (4. . . .2) transform in the 84 7 . . (4. ψ−3/2 |ki ⊗ |K I i. In doing so. one naturally expects that these heterotic states have the same kind of interactions as the D-particles of type I. whereas those of Eq. . Interactions of the heterotic non-BPS states The non-BPS D-particles described in the previous sections account for the presence in the spectrum of the type I theory of long super-multiplets of states carrying the spinorial representation of SO(32). ν] |ki ⊗ |K I i. ± 12 .1) describe massive degrees of freedom which coordinates X transform in the 44 representation of the Lorentz group. We now study the interactions of these states with the massless gauge bosons of SO(32). 4. / Nuclear Physics B 586 (2000) 206–230 Z∞ AI (v) = grav dτ VI . The level matching condition requires that K 2 = 4. we obtain the Newton’s law with its first correction   Me02 1 2 grav v 1 + + o(v 2 ). 9. . (4. Carrying the same quantum numbers. In these formulas k denotes the space–time momentum (k 2 = −Mh2 ) while K I is the adimensional momentum associated to the sixteen internal eI .19) −∞ In the non-relativistic limit. . In the bosonized formulation of the heterotic string we must distinguish between the states associated to the 16 Cartan generators that are given by 7 The fermionic states that complete this long multiplet transform in the 128 representation of the Lorentz group. = 0.218 L. bosons X The long super-multiplet of the stable heterotic states appears at the first massive level (Mh2 = 4/α 0 ). .1) (4. in the nonrelativistic limit the boundary state calculation reproduces correctly the gravitational potential energy that we expect for a pair of D-particles in relative motion. Gallot et al. (3. This may be realized for example by taking K I to be of the form (± 12 .2) will be denoted by V and can be found in Appendix C both in the (−1) and in the (0) superghost pictures. . v → 0. . α−1 ψ−1/2 (µ µ ν) ρ (4. In this section we will check this idea and investigate the gauge and gravitational interactions of the non-BPS heterotic states using standard tools of perturbative string theory. These non-perturbative states are dual to those appearing at the first massive level in the heterotic theory. . .20) where we have introduced the radial coordinate r 2 = b2 + v 2 γ 2 τ 2 . ± 12 ) with an even number of + signs. . 16) appropriately compactified [10].2) with µ. and contains the following bosonic states µ α−1 ψ−1/2 |ki ⊗ |K I i. thus obtaining the spinorial representation of SO(32) with positive chirality. . ν. VI (r) = (2κ10 )2 7Ω8 r 7 2 (3. . The vertex operators for the states (4.

.5) × ccV ¯ 1(−1) (z1 .L. . with q 2 = 0 and I = 1. Including the normalization factor C0 appropriate of any tree-level b for each vertex operator. 0. and by √ b 3 2α 0 C0 N gauge Q µ µ (ζ1 · ζ2 ) Aµ k1 − k2 CQ (Kα1 ) δKα1 +Kα2 +Q. 8 The polarization ζ can be either a vector.. ±1. . .2) is obtained by simply computing the 3-point function on the sphere among two vertex operators V and one vertex operator Vgauge (see Fig. we have closed string amplitude and a normalization factor N Z gauge  b 3 d 2 µ(zi .4). 8 This is because we want to compare our results with those of the non-BPS D-particles of type I obtained in the previous sections in which the Lorentz group structure was not manifest. 0. .3) and those associated to the remaining 480 generators which are instead given by µ Aµ ψ−1/2 |qi ⊗ |Qi.4) and the internal momentum Q of the form (0. . z¯ 1 ) ccV where α1 and α2 label the spinor representation of SO(32) carried by the non-BPS states and a ghost factor cc¯ has been added in each puncture. Using the explicit expression of the vertex operators reported in Appendix C. can be found in Appendix C in the (−1) and (0) superghost pictures. z¯ 2 ) ccV ¯ gauge (z3 .0 . (4. .1) and (4. z¯ i ) = C0 N F3 α1 α2 (0) ¯ 2(−1)(z2 . (4. with q 2 = 0 (4. . 3. Actually. namely in the terms where the polarizations ζ1 and ζ2 of the two spinor states are contracted between themselves. . we are not interested in the complete expression of this correlation function but only in the scalar part of it.7) = F3 α1 α2 2 Fig. . Gallot et al. / Nuclear Physics B 586 (2000) 206–230 µ I Aµ ψ−1/2 |qi ⊗ e α−1 |Q = 0i. . ±1. The 3-point function on the sphere.6) = F3 α1 α2 4 when Vgauge corresponds to the gauge bosons associated to the Cartan generators. Also the vertex operators for the states (4.3) and (4. . z¯ 3 ) . 16.0 . (4. .1) and (4. 3). which we denote collectively by Vgauge . The gauge coupling of the states (4. we find that the terms of (4.5) proportional to ζ1 · ζ2 are given by √ b 3 2α 0 C0 N gauge I µ µ (ζ1 · ζ2 ) Aµ k1 − k2 (Kα1 − Kα2 )I δKα1 +Kα2 . 0). a symmetric or antisymmetric two-index tensor or an antisymmetric i three-index tensor depending on which particular states (4. . . 0. 219 (4.2) are considered. . 0.

Here ΓαAB 1 α2 matrix elements of the antisymmetrized product of two Γ -matrices of SO(32) which represent the fusion coefficients among the adjoint and two spinor representations of SO(32).6) and (4. see. α2 α3 Vh 2 gYM 2 1 2 ΓαAB Γ AB 1 α4 α2 α3  1 . we finally obtain the gauge potential energy gauge  α1 α4 .8) F3 α1 α2 1 α2 2 where we have used the definitions of the normalization coefficients to express the prefactor are the in terms of the Yang–Mills coupling constant of the heterotic theory. Indeed. In fact. The dependence on the internal momenta looks different in the two expressions (4. (4.9) becomes gauge  α1 α4 . (4.12) Notice that this potential does not depend on the relative velocity of the particles involved in the interaction. moreover. 1 α4 α2 α3 (4. in this way we obtain gauge  α1 α4 . Refs. it is a “Coulomb-like” potential for point particles in ten dimensions carrying the spinor representation of the gauge group. [10. / Nuclear Physics B 586 (2000) 206–230 when Vgauge corresponds to the gauge bosons associated to the remaining generators (here Kα denotes the lattice vector corresponding to the spinorial state α and CQ (Kα1 ) is a cocycle factor. e. Then.18). Gallot et al. but a moment thought reveals that it is actually of the same form as required by gauge invariance. [41] for the expression of these Γ -matrices in terms of the internal momenta and cocycle factors.40] for details). α2 α3 Vh =− 2 gYM 1 2 t 1 2  ΓαAB Γ AB . and removing for simplicity the polarization factors. α2 α3 F4 = 2 gYM u−s (ζ1 · ζ4 ) (ζ2 · ζ3 ) 2 t 1 2  ΓαAB Γ AB . α2 α3 F4 2 = −gYM (ζ1 · ζ4 ) (ζ2 · ζ3 ) Mh2 (2 + σ ) t 1 2  ΓαAB Γ AB . for t → 0 Eq. we can rewrite both equations in the following form gYM gauge AB µ µ = √ (ζ1 · ζ2 ) Aµ k1 − k2 ΓαAB ..g. which in the limit t → 0 is related to the Lorentz parameter γ according to σ = 2(γ − 1).9) where s.11) which in configuration space becomes gauge  (r) = α1 α4 . t and u are the usual Mandelstam variables which satisfy s + t + u = 16/α 0 . 1 α4 α2 α3 (4. . we introduce the adimensional variable σ = −uα 0 /4.220 L. 7Ω8 r 7 (4. 1 α4 α2 α3 (4. For later convenience.10) Reverting to the standard field theory normalization by multiplying each external leg by √ 1/ 2E. as expected. this can gauge with the massless propagator be simply obtained by sewing two 3-point functions F3 (2. 9 See for instance Ref.7). 9 We are now in the position of evaluating the contribution to the diffusion amplitude among four particles due to the exchange of gauge bosons at tree level.

8 q · ` = 1.13) where the polarization is µν = hµν = hνµ . (4. 1 2  grav (B) F3 α α = 0. (4. it is not difficult to find that grav  b 3 α 0 (ζ1 · ζ2 )µν k µ k ν . z¯ 2 ) ccV ¯ grav (z3 . z¯ 1 ) ccV As before.19) with the appropriate . (4. also now we are interested only in the scalar part of this expression which is proportional to ζ1 · ζ2 . z¯ 3 ) . 1 2 Note that the vanishing of the heterotic non-BPS states coupling with the antisymmetric Kalb–Ramond field is consistent with the fact that the type I D-particle does not couple to the R–R 2-form. since we want to compare it with the boundary state calculation of Section 3. `2 = 0 (4. z¯ i ) F 3 α α = C0 N 1 2 (−1) (−1) (0) ¯ 2 (z2 . 1 2 √ grav (φ) (4. Gallot et al. and will be denoted generically by Vgrav . and 1 1 µν = √ Bµν = − √ Bνµ . The vertex operators corresponding to these states are written in Appendix C in the (−1) and (0) superghost pictures. φ µν = √ (ηµν − qµ `ν − qν `µ ).1) and (4. q µ hµν = 0 (4. the dilaton and the antisymmetric tensor field are grav (h) µ F3 α α = 4κ10 (ζ1 · ζ2 ) hµν k2 k2ν .2) following the same steps we have described for the gauge interactions. µν ψ−1/2 e with q 2 = 0. The gravitational coupling of the non-BPS particles can be determined by evaluating the correlation function among two vertex operators V and one vertex operator Vgrav .19) F3 α α = − 2 κ10 Mh2 (ζ1 · ζ2 ) φ. 2 2 q µ Bµν = 0 (4.16) for the antisymmetric Kalb–Ramond field. Now we can evaluate the diffusion amplitude of the non-BPS particles due to gravitational exchanges by gluing two 3-point functions (4.17) × ccV ¯ 1 (z1 .14) for the graviton. / Nuclear Physics B 586 (2000) 206–230 221 We now turn to the gravitational interactions of the non-BPS heterotic particles (4. namely Z grav  b 3 d 2 µ(zi .L.15) for the dilaton.18) F 3 α α = C0 N 2 2 1 2 from which we read that the couplings of the non-BPS states with the graviton. Using the expression of the vertex operators reported in Appendix C. Let us recall that the massless bosonic states of the graviton multiplet of the heterotic theory are µ ν α−1 |qi ⊗ |Q = 0i.

therefore they can be suppressed. T . Summing over graviton and dilaton exchanges and introducing the same notation adopted for the gauge interactions.22) In Eq. and (K) ≡ cK3 (K1 + K2 ) cK2 (K1 )(−)U/2 is a cocycle factor whose values are ±1 (see for instance Ref. one can show that this part of the 4-point amplitude is A4 = − 4 b 4 (ζ1 · ζ4 )(ζ2 · ζ3 )A(s. while those for T = −2 correspond to exchanges of the remaining 480 gauge bosons. In fact. using standard techniques.21) we recognize Newton’s law for point particles of mass Mh separated by a distance r in ten dimensions with the appropriate relativistic correction. we have . α2 α3 F4 = −(2κ10)2 (ζ1 · ζ4 )(ζ2 · ζ3 ) Mh4 (2 + σ )2 . as expected. Gallot et al. we must look for the poles of A4 with respect to t.21) which in the small velocity limit becomes grav Vh (r) = (2κ10 )2   Mh2 1 2 2 v 1 + + o v . Inspection of Eq. [40]). U ). (4.20) can be directly obtained by evaluating a 4-point function of non BPS states on the sphere. We conclude this section by mentioning that the same results (4.10) and (4. u. t. t. (4. T and U are the Mandelstam variables for the internal momenta which obey S + T + U = −16. × Γ −1 − 4 2 4 2 4 2 (4.20) depend on the indices αi that span the SO(32) spinor representations carried by the non-BPS √ particles. T . we obtain grav  α1 α4 . In the limit t → 0. (4. 7Ω8 r 7 (4. Since we are interested only in the contributions due to exchanges of massless states in the t channel. The poles for T = 0 correspond to exchanges of gravitons. S. 2 7Ω8 r 7 (4. t (4.20) the following gravitational potential energy grav Vh (r) = (2κ10 )2 Mh2 γ .24) shows that these occur only for T = 0 and T = −2.19) nor the 4-point function (4. u. / Nuclear Physics B 586 (2000) 206–230 massless propagators.222 L. S.20) Notice that.23) where A(s. neither the 3-point functions (4. U )   0    0    0  tα uα sα sin π sin π = (K) sin π 4 4 4       tα 0 uα 0 sα 0 Γ − Γ 3− ×Γ 3 − 4 4 4       tα 0 T uα 0 U sα 0 S − Γ −1 − − Γ −1 − − . dilatons and gauge bosons associated to the Cartan generators of SO(32). we can obtain from Eq.24) In this expression S. or more precisely its “universal” part in the t-channel which is proportional to (ζ1 · ζ4 )(ζ2 · ζ3 ). Normalizing each external leg by a factor of 1/ 2E and removing for simplicity the polarization terms. C0 N π2 (4.

In the heterotic theory one is able to describe the non-BPS particles in a complete way because they are perturbative configurations of the heterotic string.26) (2 + σ ) .L.26) for T = −2.20). Inspection of the coefficients and a little algebra show that these expressions indeed match with Eqs. This fact should not be surprising.12)).1) 1 + 12 v 2 + o(v 2 ) 7Ω8 r 7 in the non-relativistic limit (see Eqs.10)). A4 | T =−2 ' − 0 α t We can disentangle the gauge and gravity pieces of Eq. / Nuclear Physics B 586 (2000) 206–230 223 b4  1 16πC0N (4. (4. The reason for this is quite simple. the detailed gauge group structure is not the same in the two cases. (2. However. Gallot et al. (3. (4.25) (4 + S/2)(2 + σ ) + (2 + σ )2 .20) and (4. For the gauge interactions the situation is a bit different. Hence. thus providing a strong check on our previous calculations and on their interpretation. Conclusions We now compare the results obtained in the previous sections and discuss their relation in the light of the heterotic/type I duality. because of gauge invariance. A4 | T =0 ' 5. the gauge part at T = 0 should have the same dynamical dependence as the amplitude (4. For the gravitational interactions. because also in the case of the supersymmetric BPS D-branes one is not able to account for their degeneracy (with respect to both the Lorentz group and the gauge group) using open strings with Dirichlet boundary conditions or equivalently boundary states. This is why the gauge amplitudes involving these particles explicitly depend on the indices of the spinorial representation of SO(32) (see Eqs.20) and (4. In Section 2 we have introduced a method to describe the emission of a colored gauge boson from a non-BPS D-particle viewed as a source carrying not the spinorial indices of V grav (r) = (2κ10 )2 . Indeed with these methods one can compute only the “universal” parts of the interactions involving D-branes. in the type I theory the non-BPS particles are non-perturbative configurations of the type I string. (4. the comparison is quite simple since in both theories we have found a potential energy of the form  M2 (5. The only thing that one has to do to have complete agreement is to change the values of the gravitational coupling constant κ10 and of the mass M according to the duality map as we discussed in the introduction. we can conclude that the term of Eq. 0 α t b4 16πC0N 1 (4. On the other hand. while the term quadratic in (2 + σ ) comes from gravity.8) and (4. and in particular one can fully specify the polarizations of these states also with respect to the gauge group. What is nice to observe is that these changes make the two gravitational potential energies agree not only at the static level but also at the first non-trivial order in the velocity v. (4.10) and (4. and thus the description one is able to provide for them using perturbative methods is necessarily incomplete.25) by observing that.25) linear in (2 + σ ) is due to gauge interactions.22)). Both in the type I theory and in the heterotic string we have found that the gauge potential energy of the stable non-BPS states is in the form of Coulomb’s law (see Eqs.

but rather as a 1-point function in some definite background. In conclusion. We remark that it is not meaningful to perform this reduction directly on the heterotic 3-point function (4.20). α2 α3 (r) (5. (5.4) and (2.3)). all its couplings with the closed string states of the bulk. (2. In this framework. and thus the space–time structure of its amplitudes cannot match with that of the heterotic scattering amplitudes. (4.17) describing the emission of a gauge boson from a D-particle. i. In order to do such a comparison.3) This expression exactly agrees with the corresponding one for the type I theory given in Eq.8) and then compare it with the type I amplitude (2.12) into those of the bi-fundamental representation. . in the perturbative type I theory the D-particle is an infinitely massive object which acts as a reservoir of momentum. this amounts to transform the spinorial indices of gauge )α1 α4 . Our results thus provide a dynamical test of the heterotic/type I duality at the non-BPS level. A similar situation occurs also in the gravitational sector where the boundary state representing the D-particle generates all its 1-point functions. namely the open strings of type I. we obtain the following reduced gauge potential energy for the heterotic non-BPS particles gauge Vh (r) ≡ Tr(1)−2 Γ A Γ D =− α4 α1 ΓEΓF α3 α2 gauge  Vh 2  1 gYM δ AE δ F D − δ AF δ DE . It would be nice to extend these results to all states of this open string sector. what we have done in Section 2 is to find the 1-point function of the non-BPS D-particle with the massless states of the other sector of the bulk. In this sense. in this paper we have described how to compute the gauge and gravitational potential energies of the non-BPS D-particles of type I and shown that these agree with the corresponding ones computed for the dual heterotic states provided that one uses the known duality and renormalization effects. Gallot et al. 1 describing the gauge emission from a D-particle of type I must not be considered as a vertex or a 3-point function in the field theory sense..e.2) Then. the diagram represented in Fig. but then the comparison with the heterotic theory is not immediate.12). (2.224 L. In other words. we must “reduce” the heterotic gauge potential energy by taking into account the contribution of all pairs of states compatible with the emission of a gauge boson of definite color. In fact. using this identity and Eq. (4. / Nuclear Physics B 586 (2000) 206–230 SO(32). 2 7Ω8 r 7 α1 α4 . α2 α3 given in Eq. using the various kinds of open strings of type I we have been able to account for the gauge interactions of the non-BPS D-particles. From the group theory point of view. but rather those of the bi-fundamental representation formed with the Chan–Paton factors of the boundary changing vertex operators V09 and V90 (see Eqs. This (Vh can be easily done by noting that ΓAΓD αβ Γ BC  βα   = Tr Γ A Γ D Γ BC = Tr(1) δ BD δ AC − δ CD δ AB .

In the limit of large distance between the branes. P. thus we do not consider it. M. 1). / Nuclear Physics B 586 (2000) 206–230 225 Acknowledgements We would like to thank C. there are massless states. For the no force condition to be true. In the R sector. the massless R ground state is a GSO even (chiral) spinor of SO(1. As a consequence of the BPS condition. when only massless closed string states are exchanged. being an extended object. Schweigert and R. B. the interaction between two branes is mediated only by the exchange of closed strings and the vanishing of the force is easily seen using boundary states. Bachas. one must recall that. the D-string cannot be minimally coupled to the gauge potential and thus the naive conclusion does not apply. In the case of two parallel D-strings the interaction (A.1) at the leading order in the string coupling constant. following the scheme we proposed in Section 2. Appendix A In this appendix we describe the gauge emission from a D-string of type I. and ∆0 is a boundary changing operator for the eight space directions transverse to the D-string of . At first sight. C. this means that the attraction due to gravitons and dilatons is compensated by the repulsion due to the p + 1 R–R form under which the Dp-branes are charged. also here the NS sector is massive and does not represent any degeneracy of the D-string. Pioline. We first briefly discuss the spectrum of the 1–9 open strings stretching between a D-string and a D9-brane. Gallot et al. The aim of this appendix is to evaluate the gauge emission from a D-string of type I using the same methods applied in the case of the D-particle (but without the technicalities due to the boundary fermion η). Since the world-sheet fermions ψ 0 . M. and then to compute its contribution to the diffusion process between two D-strings. two parallel BPS Dp-branes of type I or type II do not exert any force on each other. We especially thank B. ψ 1 have zero modes.L. However. this disk amplitude has thus to vanish identically. (A. instead. Russo for several useful discussions.1) is globally invariant under the world-sheet parity Ω so that it vanishes also in type I theory. As for the 0–9 strings. Di Vecchia. In the type II theories. Indeed one finds that hDp|P |Dpi = 0 (A.2) where S + is a positive chirality spin field of conformal dimension 1/8. in this theory one has to consider also the exchange of open 9–9 strings which are present in the bulk and whose first contribution — associated to a disk with two boundary components on the D-strings and two on the D9-branes — appears at the next-to-leading order in the string coupling constant. Frau. The corresponding vertex operator reads (−1/2) V91 = λA S + ∆0 e−φ/2 . Billò. However. Pioline for valuable discussions and remarks. this seems striking since the D-string is charged under the gauge potential (in fact it carries the SO(32) spinorial representation).

6) For the same reasons discussed in the case of the D-particle.6) we need the following basic correlators [41] 0 −α 0 k 2 α 0 k 2 −1 −α 0 k 2 z13 z23 .226 L. / Nuclear Physics B 586 (2000) 206–230 conformal dimension 1/2.7) S (z1 ) ψ µ (z2 ) S + (z3 ) = Γ µ C −1 for µ = 0. F gauge = hΩ19 |Vgauge (A.3) We now evaluate the coupling of the D-string with the gauge bosons of type I. . Gallot et al.e. Inserting the explicit values. (A. i. A (A. 1. As anticipated. this is determined by the amplitude on a disk with one boundary component on the D-string and one component on one of the 32 D9-branes from which a gauge boson is emitted. q2 (A. F gauge = Cdisk NR2 NNS dµ(zi ) c V19 (A.6). we gauge gauge between two D-strings given by VI ≈ can obtain the gauge potential energy VI gauge gauge PF . ∆ (z1 ) eik·X(z2) ∆0 (z3 ) = z12  + ++ −1/2 1/4 −1/2 z12 z13 z23 (A.  Z dµ(zi ) z13 z12 z23 α 0 k 2 (A. thus we have emissions only in the two longitudinal directions µ = 0. (−1) (1)|Ω91i. Inserting them into Eq. and is represented by the 1-point function of the gauge boson in the vacuum representing the D-string.9) The vanishing of this contribution confirms our previous statements about the no force condition. thus (−1/2) V19 (−1/2) = Ω V91 = λt S + ∆0 e−φ/2 . we can express F gauge as Z (−1/2) (−1/2) (−1) (z1 ) c Vgauge (z2 ) c V91 (z3 ) .18) for the massless gauge bosons. we obtain  ++ A F gauge = Cdisk NR2 NNS Tr λt ΛBC λD Aµ Γ µ C −1  = i gYM δ AB δ CD − δ AC δ BD (A0 − A1 ). using this coupling and the propagator (2. this is not a minimal gauge coupling because the D-string is an extended object. 1. also here there is no emission of gauge fields with polarizations along the directions transverse to the D-string.8) where we have used that (Γ 0 C −1 )++ = −(Γ 1 C −1 )++ = 1. As in the case of the D-particle. Now. To evaluate (A.. we find F gauge VI 2 = gYM (δAE δDF − δAF δDE ) (−1 + 1) = 0. The vertex operator for the massless R states of the 1–9 sector is obtained by acting with Ω on V91 .5) After introducing the appropriate normalization factors. (A. but these can occur with arbitrary transverse momentum.4) where (−1/2) |Ω19 i = lim V19 z→0 (−1/2) (z)|0i and hΩ19 | = lim h0| V19 z→∞ (z).

The normalization of the disk diagrams is [11. Gallot et al. (B. In the heterotic string.10) .7) NNS = gYM 2α 0 . It reads  (B. 3 and 4. N (B.2) (B. (B. The gravitational coupling constant κ10 . / Nuclear Physics B 586 (2000) 206–230 227 Appendix B In this appendix we present the definitions of the various normalization factors that are needed for the calculations presented in Sections 2.4) (B.1) . With these expressions it is easy to see that   Tr ΛAB ΛCD = 2 δ AC δ BD − δ AD δ BC .9) λA B = δBA . (B.L. −4 0 −5 −2 C0 = (2π) α g b = 8π 5/2 α 0 2 g. (B. and are given by κ10 = 8π 7/2 α 0 2 g.32] −2 (2α 0 )−2 .5) In type I string theory one must consider also diagrams involving open strings. We now list the expression of the various Chan–Paton factors that were used in Section 2. The factor ΛAB carried by the gauge boson has indices in the adjoint representation of SO(32). the gauge coupling constant gYM is related to κ10 as follows [11] 4 2 κ = 28 π 7 α 0 3 gh2 .   A Tr λt ΛBC λD = i δ AB δ CD − δ AC δ BD . α 0 10 while in the type I theory the relation between gYM and κ10 is [11] 2 = gYM 2 = gYM 23/2 2 κ = 215/2π 7 α 0 3 gI .3) where g is the string coupling constant (gh for the heterotic string and gI for the type I theory). Cdisk = gYM (B.8) ΛAB CD = i δCA δD The Chan–Paton factor λA associated to 0–9 strings is a column vector with an index in the fundamental representation of SO(32). α 0 gI 10 (B. the normalization C0 of the closed-string treeb of the closed string vertex level diagrams (sphere diagrams) and the normalization factor N operators have the same expressions for both the SO(32) heterotic string and the type I theory. Its matrix elements explicitly read   B A B − δD δC .6) while the normalization factors of the open string vertex operators in the NS and R sectors are √ NR = gYM (2α 0 )3/4 .

¯z) CK ei(K/ 2α )·X(¯z) . (C. z¯ ) = √ Vgauge .2). Aµ ψ µ (z) e−φ(z) eiq·X(z. which depend only on the internal momenta and satisfy [40] CK (P ) CK 0 (P ) = CK+K 0 (P ). z¯ ) = − √ 2α 0 √ 1 0 e VC(−1) (z. and of the bosonic states (4. 3! where (C.4) In Eq. the vertices associated to the non-BPS states (4.5) Applying the picture-changing operator to V (−1) we can obtain the vertices V (0) in the (0) superghost picture. They are given by the same expressions (C.¯z) CK ei(K/ 2α )·X(¯z) . VB (z.6) The vertex operators for the gauge bosons (4.2) are √ 0 e (−1) (z.228 L.¯z) ∂¯ X (C.3) associated to the 16 Cartan generators of SO(32) in the (−1) superghost picture are i eI (¯z). (4. (C.2) (C. C µνρ (z) = ψ µ (z) ψ ν (z) ψ ρ (z) e−φ(z) . z¯ ) = ζµνρ C µνρ (z) eik·X(z.3) and (4. B µν (z) = ψ µ (z) ∂x ν (z) e−φ(z). Moreover.4). (C. respectively. 2α  i −iα 0 (k · ψ)ψ µ (z)ψ ν (z)ψ ρ (z) + ∂Xµ (z)ψ ν (z)ψ ρ (z) Cbµνρ (z) = √ 2α 0  − ψ µ (z)∂Xν (z)ψ ρ (z) + ψ µ (z)ψ ν (z)∂Xµ (z) .¯z) CK ei(K/ 2α )·X(¯z) . / Nuclear Physics B 586 (2000) 206–230 Appendix C In this appendix we write the vertex operators of the non-BPS heterotic states (4. (C.2) the polarization tensor ζ µν is symmetric or antisymmetric depending on whether one considers a state in the 44 or in the 84 representation of the Lorentz group.3) Aµ (z) = ∂ψ µ (z) e−φ(z). in all vertex operators we have introduced suitable cocycle factors CK .1) and (4. by   bµ (z) = √ i ∂ 2 Xµ (z) − iα 0 (k · ψ)∂ψ µ (z) .1) and (4.3) with Aµ .4) associated to the 480 remaining generators are (−1) (z.7) 2α 0 while those for the gauge bosons (4. z¯ ) = ζµ Aµ (z) eik·X(z.1) (C. Gallot et al.13) of the graviton multiplet. In the (−1) superghost picture. VA √ 1 0 e (−1) ζµν B µν (z) eik·X(z. A 2α 0  √ bµν (z) = i 2α 0 ψ µ (z)∂ψ ν (z) − (i/2)(k · ψ)ψ µ (z)∂Xν (z) B  1 + 0 ∂Xµ (z)∂Xν (z) .1)–(C. of the gauge bosons of SO(32) given in Eqs. B µν and C µνρ replaced.

B 267 (1986) 75. Gallot et al. hep-th/9812031. Nucl. Lerda.13) of the graviton multiplet are i (−1) eν (¯z)eiq·X(z. . Nucl. B 460 (1996) 525. hep-th/ 9806155. Stable non-BPS states in string theory: a pedagogical review. [17] M.¯z) CQ ei(Q/ √ e z) 2α 0 )·X(¯ . hep-th/9810188. Russo.H. Dasgupta.R.L. hep-th/9908144. B 572 (2000) 95. Stable non-BPS D-branes of type I. Gaberdiel. JHEP 9806 (1998) 007. E. hep-th/9903123. D-branes and K-theory. z¯ ) = Aµ ψ µ (z) e−φ(z) eiq·X(z. (0) (z. Gallot. [7] A. B 293 (1987) 83. P. [22] J. TASI lectures on non-BPS brane systems. hep-th/9910217. [14] O. Gallot.R. hep-th/9510169. Lett. Phys. hep-th/9904201. respectively.11) (C. Vol. [20] E. Lett. Phys. Yost. [4] A. Lerda. Gross. Nucl. JHEP 9903 (1999) 013. Phys. Sen. [11] J. JHEP 9812 (1998) 021. C. L.K. JHEP 9911 (1999) 008. Lerda. [5] A. Y.J. M. hep-th/9808141. Phys.G.¯z) CQ ei(Q/ √ e z) 2α 0 )·X(¯ . Frau. z¯ ) = √ Vgauge i 2α 0 Aµ [∂Xµ (z) − i α 0 (q · ψ) ψ µ (z)] eiq·X(z. hepth/9905006. 229 (C. z¯ ) = √ µν ψ µ (z) e−φ(z) ∂¯ X Vgrav 2α 0 in the (−1) superghost picture. [13] M.A. R. Aµ [∂Xµ (z) − i α 0 (q · ψ) ψ µ (z)] eiq·X(z.12) References [1] J. Strigazzi. Witten.8) In the (0) superghost picture these vertices become. Gaberdiel. Gaberdiel. JHEP 9810 (1998) 021. JHEP 9809 (1998) 023. the vertices for the bosonic states (4. Nucl.A. S. 1998. hep-th/9803194. [15] O. Callan. Stefanski.R. Stefanski. Bergman.R. II. B 564 (2000) 60. String Theory. [8] A. [21] C. Lovelace. Sen. [16] M. (C. (0) (z. Nucl. E. Frau. Polchinski. P. [18] M. [6] A. Nucl. Polchinski.¯z) µν [∂Xµ (z) − i α 0 (q · ψ) ψ µ (z)] ∂¯ X 2α 0 in the (0) superghost picture. A. [3] A. Rev. R. hep-th/9809111. B 441 (1998) 133. and 1 eν (¯z) eiq·X(z. hep-th/9805019. Press. Nucl. [10] D. [19] T. J. [12] J. Sen. Phys. A.R. A. Harvey. Sen. Nucl. TASI lectures on D-branes. M. Nappi. B. B 308 (1988) 221. JHEP 9808 (1998) 010. Bergman. Phys. Phys. 75 (1995) 184. Witten. hep-th/0003022. hep-th/ 9910109. Cai. Phys. hep-th/9908060.10) Finally. Martinec. Strigazzi. Sen. B 256 (1985) 253. hepth/9611050. B. z¯ ) = − Vgauge 1 eI (¯z). hep-th/9901014. Gaberdiel. C. Dirichlet branes on orbifolds. Polchinski. L. Cambridge Univ.¯z) (z. Schwarz. B 286 (1988) 91. z¯ ) = − Vgrav (C. Sen.9) and (0) (z. [2] A. [9] J. Polchinski. Rohm. Non-BPS states and branes in string theory. Sen. Phys. / Nuclear Physics B 586 (2000) 206–230 (−1) Vgauge (z.¯z) ∂¯ X 2α 0 (C. Phys.

(Eds. Lett. B 400 (1997) 63. W. Russo. Belavin. Introduction to superstring theory. I. [40] M. A. B 288 (1987) 173–232. V. Lett. [36] G. S. Phys. Lerda. Recknagel. P. P. [33] A. 1988. Phys. Lerche.B. Watamura. M. M. D. / Nuclear Physics B 586 (2000) 206–230 [23] P. Cangemi.A.M. Russo.A. hep-th/9601143.230 L. Nucl. A. 1987. hep-th/9709062. T. Nucl. Billó. E. Schomerus. L. D branes in string theories. B 216 (1989) 59. Ishibashi. Nucl. 521. [38] C. Schwarz. B 279 (1987) 465. Nucl. D branes in string theories. [31] J. Phys. Lerda. [28] A. hep-th/9608127. [30] L. Martinec. Russo. P. A. A 13 (1998) 2977. C. I. Di Vecchia. Cambridge Univ. p. Nucl. Lett.). Billó. R. [24] M. Di Vecchia. Mack et al. Kostelecky. Friedan. B 507 (1997) 259. Non-Perturbative Quantum Field Theory. Russo. Lett. Frau. hep-th/9912161. Lerda. [37] P. S. [25] M. [29] S. Di Vecchia. Phys. hep-th/9802088. hep-th/9912275. Sagnotti.B. Horava. I. Sagnotti. Nucl. S. Bachas. A. Sciuto. D. Hashimoto. Nucl. Dixon. Kiritsis. I. Vafa. Pesando. Di Vecchia. Pergamon Press. R. S. Pradisi. A. Witten. [35] N. Phys. Phys.H. [41] V. Press. B 469 (1996) 235. II. Lerda. J. Mod. Phys. Nucl. [26] M. E. Phys. Marotta. Sciuto. [27] P. Phys. Liccardo. Phys. Hamidi. A. Di Vecchia. B 310 (1989) 239. Billó. Polyakov. R. B 526 (1998) 199. B 327 (1989) 461. Phys. hep-th/9805091. M. hep-th/9511043. hep-th/9701190. S. Fundamental strings in Dp–Dq brane systems. Phys. Phys. Zamolodchikov. [39] E. S. B 496 (1997) 243. A. R. Grandjean. Lechtenfeld. Superstring theory. in: G. Magnea. B 374 (1996) 37. R. hep-th/9707068. Gallot et al. Samuel. [32] P. [34] A. B 282 (1987) 13. B 241 (1984) 333. Nucl. Shenker. Di Vecchia. Green. A. A. Fröhlich. . Sciuto. Frau. O. Onogi. hep-th/9912079. Nucl. Frau. O. Pesando. Vol. Phys.

Japan Received 28 April 2000.ac.Kc. 11. 0) supersymmetry.titech. Oh-okayama. naganuma@th. Sakai).ac. It is argued that this feature is a generic phenomenon of Nambu–Goldstone modes on domain wall junctions in the bulk flat space in any dimensions.jp (H. Subsequently a great deal of research activity has been performed to study and extend the proposal [8–11]. 11. Another fascinating possibility has also been proposed to consider walls in the bulk spacetime which has negative cosmological constant [3.titech. Tokyo 152-8551. 11.jp (M. All rights reserved. Hodaka Oda.Nuclear Physics B 586 (2000) 231–260 www.27.15.elsevier. Naganuma).4]. Phenomenological implications of the idea have been extensively studied from many aspects.10. PII: S 0 5 5 0 .phys. hoda@th. We formulate mode equations and show that fermion and boson with the same mass come in pairs except massless modes which can appear singly.phys.jp (N. ∗ Corresponding author.ac.V. Since walls typically have co-dimension one. Introduction In recent years an interesting idea has been advocated to regard our world as a domain wall embedded in higher dimensional spacetime [1.phys. it is desirable to consider intersections and/or junctions of walls in order to obtain our four dimensional world from a spacetime with much higher dimensions.2].nl/locate/npe Nonnormalizable zero modes on BPS junctions Kenji Ito.+d.phys. Most of the particles in the standard model should be realized as modes localized on the wall. Oda). All rights reserved. The model with the bulk cosmological constant has been extended to produce an intersection of walls [5–7].titech.jp (K. Masashi Naganuma ∗ .3 . Their wave functions extend along the wall to infinity (not localized) and are not normalizable. Meguro. Nambu–Goldstone modes on the BPS domain wall junction are worked out for N = 1 supersymmetric theories in four dimensions. accepted 5 July 2000 Abstract Using an exact solution as a concrete example. nsakai@th. The model can give large mass hierarchy or can give massless graviton localized on the wall.3 2 1 3 ( 0 0 ) 0 0 4 3 6 .30. Ito). Tokyo Institute of Technology.  2000 Elsevier Science B.V.ac. E-mail addresses: kito@th.P.titech. Norisuke Sakai Department of Physics. PACS: 11. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.-z 1. in accordance with unitary representations of (1.

Moreover. these domain walls meet at a one-dimensional junction. the domain wall junction preserves only one supercharge. Recently domain wall junctions have attracted much attention as another interesting possibility for BPS states [25–27]. Our exact solution showed that the central charge Y contributes negatively to the mass of the domain wall junction configuration. Therefore we should not consider the central charge Y alone as the mass of the junction. One such point is the sign and meaning of the new central charge Y which arises when walls form a junction. If three or more different discrete vacua occur in separate region of space. If the two spatial dimensions of all of these domain walls have one dimension in common. Massless modes can appear singly without accompanying fields with opposite statistics. Such a domain wall preserves half of the original supersymmetry and is called a 1/2 BPS state [14]. Consequently several misconceptions can be pointed out and rectified.29]. 0) supersymmetry are classified into doublets for massive modes and singlets for massless modes. Various other aspects of the domain wall junctions are also studied recently [34–47]. We will use our exact solution as a concrete example and will extract the generic properties of the BPS domain wall junctions. We work out . For instance a new topological charge Y is found to appear for such a 1/4 BPS state [25–27]. Various properties of domain walls in N = 1 supersymmetric field theories in four dimensions have been extensively studied [18–22]. In spite of all these efforts. Ito et al. / Nuclear Physics B 586 (2000) 231–260 Supersymmetry has been useful to achieve stability of solitonic solutions such as domain walls. segments of domain walls separate each pair of the neighboring vacua. Domain walls occur in interpolating two discrete degenerate vacua in separate region of space. especially the Nambu–Goldstone modes. there have been a number of numerical simulations which indicate the existence of the domain wall junction solutions [28. The exact solution allows a thorough study of the properties of the BPS domain wall junction. The solitonic configuration for the junction can preserve a quarter of supersymmetry and is called a 1/4 BPS state. We also show that unitary representations of the surviving (1. In particular the modes on the domain wall background have been worked out and are found to contain fermions and/or bosons localized on the wall in many cases [23. 0) supersymmetric theory in 1 + 1 dimensions [25] which offers an intriguing possibility of chiral fermions. Thanks to the topological charge. There has been progress to study general properties of such domain wall junctions. Consequently the resulting theory was expected to be a (1. We study the modes on the background of the domain wall junction. it has been difficult to obtain an explicit solution and to prove the existence of a BPS domain wall junction. If we start from N = 1 four dimensional supersymmetric field theories. 13]. It has also been noted that these BPS states possess a topological charge which becomes a central charge Z of the supersymmetry algebra [15–17]. We define mode equations and demonstrate explicitly that fermion and boson with the same mass have to come in pairs except massless modes. Domain walls in supersymmetric theories can saturate the Bogomol’nyi bound [12.24]. Recently we have succeeded to work out an exact solution for the BPS domain wall junction for the first time [30].232 K. these BPS states are guaranteed to be stable under arbitrary local fluctuations. The purpose of the present paper is to give a more detailed study of the properties of the BPS domain wall junction in N = 1 supersymmetric field theories.

Therefore the anti-de Sitter geometry along the wall plays an essential role to achieve the localization of the wave function on the intersection in models with cosmological constant. we introduce BPS equations and the exact solution for the domain wall junction and discuss representations of the surviving (1. Since their wave functions are extended along these walls without damping. However. they are not localized states on the junction. One should note that our conclusion need not apply to the case with negative cosmological constant in the bulk. Therefore the fact that the Nambu–Goldstone modes on the BPS domain wall junction are not normalizable is a generic feature of supersymmetric field theories in the bulk flat space. We observe that the origin of this property can be traced back to the fact that the supersymmetry is broken by the coexistence of nonparallel walls. In Section 5. We work out the Nambu–Goldstone mode explicitly and show that they are not normalizable.K. Therefore they are not normalizable. five-dimensional walls can intersect in anti-de Sitter space. In Section 2. / Nuclear Physics B 586 (2000) 231–260 233 explicitly massless Nambu–Goldstone modes associated with the broken supersymmetry and translational invariance. For instance graviton zero mode is exponentially suppressed away from the intersection along the wall direction to produce a normalizable wave function. We find that the Nambu–Goldstone fermions exhibit an interesting chiral structure in accordance with the surviving (1. Since anti-de Sitter space does not have translational invariance. This indicates that the resulting theory cannot be regarded as a genuine (1 + 1)-dimensional field theory with discrete particle spectrum even at zero energy. If one demands a flat space at the four-dimensional intersection. the wave function of the zero mode does not become constant along the wall asymptotically. we present mode equations which define the fluctuations on the background of domain wall junction. If one approaches the intersection along the wall. we have to keep in mind that the domain wall junction configuration is actually living in one more dimensions similarly to the domain wall itself. contrary to our situation. one has an anti-de Sitter space not only in the bulk but also even on the walls [5–7]. 0) supersymmetry algebra. The non-normalizability of Nambu–Goldstone modes on the junction configuration is not an accident in this particular model. Physical origin of the nonnormalizability is clarified and the general validity of this phenomenon is argued. one meets precisely the same situation as the wall in the five-dimensional anti-de Sitter space. In Section 3. the relation between the choice of BPS equations and the boundary condition is discussed for a general Wess–Zumino model. Ito et al. Zero modes on the junction turn out to have properties quite similar to those on the domain wall. . contrary to a previous expectation [25]. In the presence of a bulk negative cosmological constant in six dimensions. 0) which is characteristic to 1 + 1 dimensions. In Section 4. Although the remaining supersymmetry is just (1. we also find that any linear combinations of the Nambu–Goldstone modes associated with the junctions become a linear combination of zero modes on at least one of the domain walls asymptotically along these walls. the central charge density and the energy density are examined and an interesting behavior is observed. The fermionic contributions to the central charges and mode equations in a convenient gamma matrix representation are given in the appendices. 0) supersymmetry algebra.

1. . To find the BPS equations satisfied by these BPS states. k = 1. Φ ∗j + and compute the anticommutators (2. 2.5) Yk = i knm d 3 x Kij ∗ ∂n A∗j ∂m Ai . µ = 0. ν = 0. BPS equations and the (1. 2.234 K.3) L = d 2 θ d 2 θ¯ K Φ i . where Ai is the scalar component of the ith chiral superfield Φ i and Kij ∗ = ∂ 2 K(A∗ . come from the total divergence. Qβ } = 2i σ k σ¯ 0 α γ γ β Zk . 2. the N = 1 superalgebra in general receives contributions from central charges [16–32].1).4) Z  (2. [33] except that the four-dimensional indices are denoted by Greek letters µ. Two 1/4 BPS states and two BPS equations It is known that if the translational invariance is broken as is the case for domain walls and/or junctions. 3  (2. Therefore these charges are topological in the sense that they are determined completely by the boundary conditions at infinity. 3 instead of roman letters m. we consider a S with an arbitrary complex two-vector hermitian linear combination of operators Q and Q α α ˙ α ∗ β and its complex conjugate β¯ = (β ) as coefficients Sα˙ .  123 = 1.17] and BPS domain wall junction is a 1/4 BPS state [25.1) {Qα .2) to find the central charges. an arbitrary superpotential W and an arbitrary Kähler potential K(Φ i . 3 of the system  Sα˙ } = 2 σ µ Pµ + σαkα˙ Yk . 0) supersymmetry algebra 2. 3 besides the energy–momentum four-vector P µ . (2. . The anti-commutator between left. The anti-commutator between two left-handed supercharges has central charges Zk . (2. and they are non-vanishing when there are nontrivial differences in asymptotic behavior of fields in different region of spatial infinity as is the case of domain walls and junctions [26]. . .c. 0) and (1/2. (2. A)/∂Ai ∂A∗j is the Kähler metric. k = 1. . Here and the following we use two-component spinors following the convention of Ref. . For instance. we can take a general Wess–Zumino model with an arbitrary number of chiral superfields Φ i . 1/2) central charges in accordance with the transformation properties under the Lorentz group.2) {Qα . Ito et al. Φ ∗j ) Z  Z   d 2 θ W Φ i + h. K = β α Qα + β¯ α˙ Q (2. / Nuclear Physics B 586 (2000) 231–260 2. Zk and Yk . 1.6) 1 The central charge Y also receives contributions from fermionic components of chiral superfields which is k given in Appendix A.and righthanded supercharges receives a contribution from central charges Yk . Central charges. (2. The contributions to these central charges from bosonic components of chiral superfields are given by [26] 1 Z Zk = 2 d 3 x ∂k W ∗ (A∗ ).26]. n. Q α α˙ One may call Zk and Yk as (1. BPS domain wall is a 1/2 BPS state [16.

Ito et al. If we assume.K. for simplicity. The field configuration of static junction must be at least two-dimensional. Since K is hermitian. and the inequality implies in this case n . the expectation value of the square of K over any state is non-negative definite. / Nuclear Physics B 586 (2000) 231–260 235 We treat β α as c-numbers rather than the Grassmann numbers.5). hS|K 2 |Si > 0. x 2 then we obtain hZ3 i = hY1 i = hY2 i = 0 from Eqs. that it depends on x 1 .4) and (2. (2.

.

.

.

    −1 .

β 1 .

2 − .

β 2 .

ψ i . For the general Wess–Zumino model in Eq. when β 1 = β¯ 1 = 0.3). F i ) gives after eliminating the auxiliary field F i 2 ∗ ∂Ai ∗ ∂W = −Ω+ F i = Ω+ K −1ij . Q1 + |hiZ1 + Z2 i| If HII > HI . then supersymmetry can only be preserved at hH i = HII and the only one combination of supercharges is conserved   h−iZ1 + Z2 i S Q ˙ |hH i = HII i = 0. this is a 1/2 BPS state. is preserved by the state |Si K |Si = 0. for any β α and for any state. and the inverse of the Kähler ∗ metric K −1ij are introduced. then supersymmetry can only be preserved at hH i = HI and the only one combination of supercharges is conserved   hiZ1 + Z2 i S (2. (2. (2.10) In the case of HI 6= HII .13) where complex coordinates z = x 1 + ix 2 . For simplicity we . hiZ1 + Z2 i ˙ ˙ . |h−iZ1∗ + Z2∗ i| (2. HII }. z¯ = x 1 − ix 2 . β 2 = β¯ 2 = 0.11) and (2. K. We can also consider gauge interactions. (2. The equality holds if and only if the linear combination of supercharges. two candidates of BPS bounds coincide and BPS state conserves both of two supercharges.12) Q2 + |h−iZ1 + Z2 i| 2 In the case of HI = HII .12). the BPS bound becomes hH i > max{HI .8) In this case.7) + Re β 2 h−Z2 + iZ1 i .9) (2.11) for H = HI applied to chiral superfield Φ i = (Ai . the condition of supercharge conservation (2. If HI > HII . h−iZ1 + Z2 i ˙ ˙ . We find that there are two candidates for the saturation of the energy bound [30]: H = HI ≡ |h−iZ1 − Z2 i| − hY3 i.2 hY3 i + Re β 1 2 h−Z2 − iZ1 i hH i > 1 2 2 2 |β | + |β | o  2 (2. (2. the state |Si saturates the energy bound and is called a BPS state.11) Q1˙ |hH i = HI i = 0. ∂ z¯ ∂A∗j Ω+ ≡ i h−iZ1∗ + Z2∗ i . β¯ 2 = β 2 |h−iZ1 + Z2 i| (2. when β¯ 1 = β 1 |hiZ1 + Z2 i| H = HII ≡ |hiZ1 − Z2 i| + hY3 i.

we shall present a simple way to find the correspondence between the choice of boundary conditions and the choice of BPS equations (2. Then the derivative ∂Ai /∂ z¯ in the above Eq. ∗j ∂A  2  ∂ ei i i + i vµ A .13) and (2. ∗j ∂z ∂A Ω− ≡ i h−iZ1∗ − Z2∗ i . |h−iZ1∗ − Z2∗ i| (2. The exact solution of BPS domain wall junction In a previous article [30]. (2. the derivative ∂Ai /∂z should be replaced by the covariant derivative Dz Ai = 12 (D1 − iD2 )Ai . (2.12) for H = HII applied to chiral superfield in the Wess–Zumino model gives after eliminating the auxiliary field 2 ∗ ∂Ai ∗ ∂W = −Ω− F i = Ω− K −1ij .17) v12 = D = − 2 j In Section 4. λ. In this model there are three discrete vacua.18) U (1) 0 0 1 −1 1 −1 0 0 1 −1 1 −1 0 0 0 U (1) interacting with a superpotential f + (T + Λ)DD e + (T − m)QQ e − h2 T . we have found an exact solution of BPS domain wall junction in a model motivated by the N = 2 supersymmetric SU(2) gauge theory with one flavor broken to N = 1 by the mass of the adjoint chiral superfield. Dz¯ = (D1 + iD2 ). (2.16) If U (1) gauge interaction is present. W = (T − Λ)MM (2. / Nuclear Physics B 586 (2000) 231–260 take only the U (1) gauge interaction. This model has the following chiral superfields with the charge assignment for the U (1) × U (1)0 gauge group f D D e Q Q e T M M (2.16) and (2. In this case the BPS condition applied to U (1) vector superfield in the Wess–Zumino gauge becomes in the case of minimal kinetic terms 1 X ∗j A ej A j . v23 = v02 . (2. v01 = −v31 .2. Ito et al. v23 = −v02 . v01 = v31 .14) Dµ A = ∂x µ 2 Moreover the same BPS condition (2.236 K. 2. v03 = 0. D) gives after eliminating the auxiliary field D 1 X ∗j A ej A j . Similarly the condition of supercharge conservation (2.13) should be replaced by the gauge covariant derivative Dz¯ Ai ∗ 1 ∗ ∂W 2Dz¯ Ai = Ω+ K −1ij .17). .11) applied to vector superfield in the Wess–Zumino gauge V = (vµ . v03 = 0.15) v12 = −D = 2 j where vµν ≡ ∂µ vν − ∂ν vµ and ej is the charge of the field Aj .15) or (2. Here we assume for simplicity the minimal kinetic term both for the chiral superfield Kij ∗ = δij ∗ and for the vector superfield.19) where parameters Λ and h can be made real positive and a parameter m is complex [30].

Correspondingly we obtain the BPS equations (2.17) for H = HII with Ω− = −1. besides Q(1) = √ e−i 4 Q2 + ei 4 Q 2 2  5π 5π 1 S˙ .12) as  π π 1 S˙ . / Nuclear Physics B 586 (2000) 231–260 237 Fig.3 : T = −Λ. D = D √ and when m = i 3Λ. In fact we find that the domain wall junction configuration Q(1) = (e−i 4 Q2 + ei 4 Q 2 conserves precisely this single combination of supercharges. W2 = −h2 m. Ito et al.K.20) Vac. Q=Q W1 = −h2 Λ. Thus three half walls are expected to connect at the junction with relative angles of 2π/3.16) and (2. Q(2) = √ ei 4 Q1 + e−i 4 Q 1 2 (2. 1. we specify the boundary condition where the wall 1 extends along the negative x 2 axis separating the vacuum 1 (x 1 > 0) and 3 (x 1 < 0) as shown in Fig. The BPS equations (2. at wall 3 Q(4) = √ e−i 12 Q1 + ei 12 Q 1 2  π π 1 S˙ . M = M f= D = D e = 0. besides Q(1) = √ e−i 4 Q2 + ei 4 Q 2 2 (2.2 : T = m. e = h. M=M e = h. Vac. Q = Q e= D = D e = 0. W3 = h2 Λ.21) The other two walls have also two conserved supercharges  π π 1 S˙ . this model becomes Z3 symmetric. 1. at wall 2 Q(3) = √ e−i 12 Q1 + ei 12 Q 1 2  π π 1 S˙ . Q = Q e= M = M f = 0. Boundary condition of the model in [30]. even though it has also another central charge Yk contributing. f = h. (2. Q(1) = √ e−i 4 Q2 + ei 4 Q 2 2  π π 1 S˙ .1 : T = Λ. If we have only the wall 1. For definiteness.17) for the vector . (2.22) When these three half walls coexist. Vac.11) and (2. we obtain the central charge Zk (vanishing Yk ) and find the two conserved supercharges from Eqs. we can have only one common conserved supercharge √ π π S˙ )/ 2.

The BPS equations (2. A similar observation has also been made in Ref. Correspondingly we should take the superpotential as √  1 1 h2 1 W = (T − Λ)M2 + (T + Λ)D2 + T − i 3Λ Q2 − T . Unitary representations of (1. z¯ ) = Q(z.17) for vector f = |M|. z¯ ) = s+t +u √ 2 Λt e z¯ ) = .3. M(z. The solution for these BPS equations is given by [30]. we can simplify the model without spoiling the solvability to obtain a Wess–Zumino model consisting of purely chiral superfields by the following procedure. 2 2 2 2 (2. 3 5 1 (2. and one Lorentz generator J 03 .23) assuming the minimal kinetic term. The vector superfields actually serve to constrain chiral superfields to have the identical magnitude pairwise through D = 0 to satisfy the BPS equation (2. 2.26) This Wess–Zumino model has the same solution as ours by changing h2 → h2 /2. Therefore we can eliminate the vector superfields: |M| f = M. / Nuclear Physics B 586 (2000) 231–260 superfield can be trivially satisfied by vµ = 0 and D = 0. superfields and reduce the number of chiral superfields by identifying pairwise M e e D = D. |D| e = |D|. we define the hamiltonian H 0 = H − hH i measured from the energy hH i of the background configuration. z¯ ) = D(z. |Q| e = |Q|. [36]. P 3 . 0) supersymmetry algebra Let us examine states on the background of a domain wall junction from the point of view of surviving symmetry. D(z. T (z. ∂z ∂A (2.238 K.16) corresponding to H = HII . Q = Q. Ito et al. s +t +u √ 2 Λu e z¯ ) = . z¯ ) = √ s+t +u 3 3 1    1 2Λ s = exp √ Re ei 6 π z . Λ → √ 3Λ/2. √ 2 Λs f . s+t +u i 2Λ e−i 6 π s + e−i 6 π t + ei 2 π u + √ Λ. 3 (2.25) This model is motivated by the softly broken N = 2 SU(2) gauge theory with one flavor. By projecting .24)    5 2Λ t = exp √ Re ei 6 π z . In the case of the BPS states satisfying the BPS equation (2. Q(z. z¯ ) = M(z. 3    1 2Λ u = exp √ Re e−i 2 π z . However. Since we are interested in excitation modes on the background of the domain wall junction. two translation generators H .16) for chiral superfields become in this case 2 ∂W ∗ ∂Ai = − ∗i . we have only one surviving supersymmetry charge Q(1) . out of the N = 1 four-dimensional super Poincaré generators.

0) supersymmetry algebra on the domain wall junction as anticipated [25]. To obtain unitary representations.28) Other commutation relations are trivial       0 H − P 3 .27) Q(1) = H 0 − P 3 . (2. H 0 + P 3 = H 0 − P 3 .K. Ito et al. 2    03 0 J . We also obtain the Poincaré algebra in 1 + 1 dimensions  03 0  i   J .H − P3 = i H0 − P3 . Q(1) = H 0 + P 3 . we can diagonalize H 0 and P 3 .29)  This is precisely the (1. we immediately find 2 (2.2) with central charges in four dimensions. J 03 . (2. / Nuclear Physics B 586 (2000) 231–260 239 from the supersymmetry algebra (2.1). Q(1) = Q(1) . H + P 3 = −i H 0 + P 3 . (2. Q(1) = 0.

.

.

.

H 0 .

E. p3 = E .

E. p3 . (2.30) P 3 .

E. p3 = p3 .

p3 . E > |p3 |.E. |F i). (2. operating by Q(1) on the state gives an unphysical zero norm state . If E − p3 > 0. |F i = p 1 E − p3 Q(1) |Bi. If E − p3 = 0. we can construct bosonic state from fermionic state and vice versa by operating Q(1) on the state: |Bi = p 1 E − p3 Q(1) |F i. and combine them by means of Q(1) .31) Therefore we obtain a doublet representation (|Bi.

(1) .

.

.

.

.

 .

.

Q .

p3 .E.

2 = E. p3 .

Q(1) 2 .

E. p3 . p3 = E.

H 0 − P 3 .

1) supersymmetry and the left-moving massless states can appear as singlets. We also find that massive modes should appear in pairs of boson and fermion. Moreover we can replace our model. if . (2. p3 = E − p3 = 0.32) Then the massless right-moving state |E. 0) supersymmetry algebra. doublet and singlet. If another BPS equation (2. (2.1. 3. Mode equation on the junction Since the vector superfields have no nontrivial field configurations.13) corresponding to H = HI is satisfied instead of Eq. we have (0. Nambu–Goldstone modes have no component of vector superfield. Nambu–Goldstone and other modes on the junction 3. This singlet state can either be boson or fermion. Thus we find that there are only two types of representations of the (1.E. p3 = Ei is a singlet representation.16). whereas the massless rightmoving mode can appear singly without accompanying a state with opposite statistics. This provides an interesting possibility of a chiral structure for fermions.

Let us consider quantum fluctuations A0i .3) in the following. x 3 = 0. x 2 space     ∂2W∗ ∂3W ∂W ∗ ∂2W OB i j  ≡  − ∂12 + ∂22 δ i j + ∂Aicl ∂Akcl ∂A∗k ∂A∗j cl cl ∂3W∗ ∂W ∗j ∂Ak ∗k ∂A∗i cl cl ∂Acl ∂Acl  OB i j 0∗j An 0j An   = Mn2 A0∗i n A0in j ∗k ∂Aicl ∂Akcl ∂Acl ∂Acl   ∂2W ∂2W∗ − ∂12 + ∂22 δ i j + ∗i ∗k ∂Acl ∂Acl ∂Ak ∂Aj cl  . The quantum fluctuation for scalar can be expanded in terms of these mode functions to obtain a real scalar field equation with the mass Mn for the coefficient bosonic field an (x 0 . (3. (3. ψ i around a classical solution Aicl which satisfies the BPS equations (2. We shall describe one such representation in Appendix B. x 2 .39) (3. For simplicity we assume the minimal kinetic term here Kij ∗ = δij ∗ . x 3 )  X   an x 0 .15) for H = HI or (2. ψ j = 0. ∗j j ∗k ∂Aicl ∂Akcl ∂A∗k ∂Aicl ∂Akcl ∂Acl ∂Acl cl ∂Acl (3. (2. x 3 A0in x 1 . x 2 ) space is manifest. (3. x 3 and x 1 .38) Similarly the linearized equation for fermions is given by −i σ¯ µ ∂µ ψ i − −iσ µ ∂µ ψ¯ i − ∂ 2W ∗ ∗j ∂A∗i cl ∂Acl ∂ 2W j ∂Aicl ∂Acl ψ¯ j = 0. x 1 . we can define . The bosonic modes A0in (x 1 . x 1 . Transforming from such a representation to the Weyl representation which we are using. (3.  cl (3. x 2 we have to define mode equations on the background which has a nontrivial dependence in two dimensions. (3.240 K. x 2 ) can easily be defined in terms of a differential operator OB in x 1 . Ito et al.40) To separate variables for fermion equations.34) In order to separate variables in x 0 .35)  .37) A0i x 0 .36) where the eigenvalue Mn2 has to be real from Majorana condition. it is more convenient to use a gamma matrix representation where direct product structure of 2 × 2 matrices for (x 0 . x 3 = n ∂02 − ∂32 + Mn2   an x 0 .16) and (2. x 2 . Consequently we shall neglect vector superfields and consider the general Wess–Zumino model in Eq.13) and (2. x 2 . Ai = Aicl + A0i . / Nuclear Physics B 586 (2000) 231–260 we wish.17) for H = HII .33) We retain the part of the Lagrangian quadratic in fluctuations and eliminate the auxiliary fields F i to obtain the linearized equation for the scalar fluctuations −∂µ ∂ µ A0∗i + ∂ 2W ∂ 3W ∂ 2W ∗ ∂W ∗ 0j 0∗j A + A = 0. x 3 ) and (x 1 . by another model with purely chiral superfields without spoiling the essential features including the solvability.

1/2 (3. In this ordering. cn are real.16) corresponding to H = HII to find the differential operator for bosons OB  iπ/4 1/2   −iπ/4 −1/2  e Ω− 0 Ω− 0 e i O2i k O1k j = O j −1/2 1/2 . mn     1 + ρ3 cn (x 0 . the coefficient fermionic fields bn .45) ψα x . Ito et al. cn (x 0 . we can use the BPS equation (2. x 3 )ψn1 i 0 1 2 3 (3.K. / Nuclear Physics B 586 (2000) 231–260 241 i .43) O1 i j  n ˙ . we have two distinct real mass parameters mn .16) corresponding to H = HII guarantees that the existence ˙ i of fermionic mode equations implies the existence of a solution of of a solution ψ¯ ni 1 . x 3 ) 2 2 where we use Pauli matrices ρa .40) for the fermion gives a Dirac equation in 1 + 1 dimensions for the coefficient fermionic fields (cn . x 3 ) is a Majorana spinor. x . x 2 . Please note a peculiar combination of leftand right-handed spinor components to define eigenfunctions. x 2 )  X bn (x 0 . ψn2 (1) (2) bosonic mode equations with the mass squared Mn2 = mn mn −1/2 −iπ/4 A0∗i Ω− n =e ˙ ψ¯ ni 1 . To relate the mass eigenvalues of fermions and bosons. x 2 ) .44) O2 j n j 2˙ i ψn2 ψ¯ n (2) where the mass eigenvalues m(1) n . a = 1. 2.42) O2 j ≡  . (3.41) O1 i j ≡  . x = i (x 1 . We can expand ψ i in terms of these mode functions   i (x 1 . let us multiply two differential operators for fermions O2 to O1 .46) −i ρ1 ∂0 + iρ2 ∂3 − m(1) n n ibn (x 0 . x 1 . x 3 ) (2) 1 − ρ3 − m = 0. ∂ 2W   i −i(∂1 + i∂2 )δj − j ∂Aicl∂Acl   ∂ 2W i −i(∂1 − i∂2 )δj   − i j ∂Acl ∂Acl   i (3. 3 to construct the 2 × 2 gamma matrices ρ1 . i A0in = eiπ/4 Ω− ψn2 . 2W ∗ ∂   −i(−∂1 − i∂2 )δji − ∗j ∂A∗i cl ∂Acl   " i # ¯ nj 1˙ ψn1 ψ  = −im(1) (3. Since we have a Majorana spinor in 1 + 1 dimensions which does not (1) (2) allow chiral rotations. x . ibn ) with two mass parameters (1) (2) m n . mn are real. iρ2 in 1 + 1 dimensions. x 3 )ψn2 n Since ψ(x 0 .47) B 0 e−iπ/4 Ω− 0 eiπ/4 Ω− Therefore the BPS equation (2.48) . (3. The linearized equations (3.39) (3. (3. j ψ¯ ni 2 ψn2   " ˙# j ¯ i1 ψn1 i  (2) ψn  = im . mn .ψ ¯ ni β˙ combining components of left-handed and right-handed the fermionic modes ψnα spinors by means of the following operators   ∂ 2W ∗ i − −i(−∂ + i∂ )δ 1 2 j   ∗j ∂A∗i   cl ∂Acl (3.

O1 k O2 j = 1/2 −e−iπ/4 Ω+ 0 e Ω+ 0 (3.13) corresponding to H = HI guarantees that the existence ˙ i of fermionic mode equations implies the existence of a solution of of a solution ψ¯ ni 2 . we find it convenient to choose the three broken supercharges as the following real supercharges  1 S˙ .242 K. (3. 3.13) corresponding to H = HI is valid. 1/2 (3. i A0in = e−iπ/4 Ω+ ψn1 . operator multiplication with different ordering gives the same bosonic operator whose rows and columns are interchanged   −1/2  −1/2  0 eiπ/4 Ω+ 0 −eiπ/4 Ω+ i k i OB j −iπ/4 1/2 . Q1 + e−iπ/4 Q QIII = √ e 1 2  1 S˙ . QI = √ eiπ/4 Q2 + e−iπ/4 Q 2 2  1 iπ/4 S˙ .2. we obtain √     ∗ ∗ ξ ∂1 Aicl + iσ 2 ξ¯ + iΩ− ξ ∂2 Aicl .50) Therefore we find that all massive states come in pairs of boson and fermion with (1) (2) the same mass squared Mn2 = mn mn in accordance with the result of the unitary representation of the (1. Nambu–Goldstone modes Since we are usually most interested in a low energy effective field theory. If global continuous symmetries are broken spontaneously.51) δξ ψ i = i 2 σ µ ξ¯ ∂µ Aicl + 2 ξ Fcli .52) We see that there is one conserved direction in the Grassmann parameter: ∗ ξ iσ 1 ξ¯ = Ω− ∗ and σ 2 ξ¯ = −Ω− ξ. δξ ψ i = 2 iσ 1 ξ¯ − Ω− (3. ψn1 (2) bosonic mode equations with the mass squared Mn2 = m(1) n mn −1/2 iπ/4 A0∗i Ω+ n = −e ˙ ψ¯ ni 2 . we wish to study massless modes here.54) . x 2 ) to the transformation of fermions by a Grassmann parameter ξ . for instance. / Nuclear Physics B 586 (2000) 231–260 If another BPS equation (2. x 2 ) and Fcli (x 1 . For our exact solution. we perform the associated global transformations and evaluate the transformed configuration by substituting the classical field. If the BPS equation (2. 0) supersymmetry algebra.49) Therefore the BPS equation (2.53) The other three real Grassmann parameters ξ correspond to broken supercharges. there occur associated massless modes which are called the Nambu–Goldstone modes. since classical field configuration of fermion vanishes ψcli = 0 √ √ (3. For supersymmetry we obtain nontrivial wave function by substituting the classical field Aicl (x 1 . QII = √ e−iπ/4 Q1 + eiπ/4 Q 1 2 Then the corresponding massless mode functions are given by (3. To find the wave functions of the Nambu–Goldstone modes.16) for the junction background is valid. Ito et al.

On the other hand. x 2 eiπ/4    0 1 2  . no linear combination of these Nambu–Goldstone fermions can be formed which does not have support extended along any of the wall. However. We plot the absolute values of (a)i=T | of the i = T component of the wave function of the Nambu–Goldstone fermions |ψ0 a = I. x )ψn1 (x 1 . and (II) (III) c0 (x 0 + x 3 ). They become identical to fermion zero modes on at least two of the walls asymptotically and hence they are not localized around the center of the junction. and c0 (x 0 + x 3 ) are left-moving modes. This fact means that the low energy dynamics of BPS junction cannot be described by a (1 + 1)dimensional effective field theory with a discrete particle spectrum. (3. c0 (x 0 . x 2 eiπ/4  x 1.x + c0 x 0 . x 3 ). to obtain the Nambu–Goldstone component of the mode expansion      ψ i x 0 . x 2 e−iπ/4 .57) Since the transformation parameter should correspond to the Nambu–Goldstone field with zero momentum and energy. These three states are all singlets of the (1. are given by   (a) (3. Therefore these wave functions are not localized and are not normalizable.K. We can construct a linear combination of the Nambu–Goldstone fermions to have no support along one out of the three walls. a = 1. a = 1. 2∂1 Aicl x 1 . x 3 ).x = 2∂2 Aicl x 1 . x 3 ψ0(II)i x 1 . x ) We have explicitly displayed three massless Nambu–Goldstone fermion components distinguishing from the massive ones (n > 0). x 3 ψ0(I)i x 1 . x 2 = ∂a Aicl x 1 . III in Fig.46) shows that b0(I) (x 0 − x 3 ) is a right-moving massless mode. x 3 space. 0) supersymmetry algebra in accordance with our .58) + 0 3 i 1 2 n>0 cn (x . x 1 . x 2  (III)i 1 2  (III) x . Ito et al. x 3 = b0(I) x 0 .59) A0 x 1 . the three transformation parameters ξ should be promoted to (I) (II) (III) three real fermionic fields in x 0 . x 2 + c0(II) x 0 . 2. The right-moving modes are asymmetric in bosons and fermions: two Nambu–Goldstone bosons and a single Nambu–Goldstone fermion. x 2 ) . x 2 . x 2 = . b0 (x 0 . x 2 = 243  (3. we have seen already that there are two left-moving massless Nambu–Goldstone fermions and one right-moving massless Nambu–Goldstone fermion. Similarly the Nambu–Goldstone bosons corresponding to the broken translation P a . These two bosonic massless modes consist of two left-moving modes and two right-moving modes. 2. x )ψn2 (x . The Dirac equation for the coefficient fermionic fields (3. 0) supersymmetry algebra. We can see that Nambu–Goldstone fermions have wave functions which extend to infinity along three walls. 2. x 3 ).56) (3. x 2 . x 3 ψ0 ! 0 3 i X bn (x . and c0 (x 0 . / Nuclear Physics B 586 (2000) 231–260 (I)i ψ0 (II)i ψ0 ψ0(III)i   4∂z Aicl x 1 . x .55) (3. These two left-moving Nambu–Goldstone bosons and fermions form two doublets of the (1. 0    0  x 1. II.

Ito et al. / Nuclear Physics B 586 (2000) 231–260 .244 K.

(I)T

The wave function

ψ0

(II)T

The wave function

ψ0

(III)T .

The wave function .

ψ0 .

3. 3. . Nonparallel wall also breaks half of supercharges. A single domain wall breaks only a half of supercharges. linearly independent ones among these two sets of broken supercharges of nonparallel walls become 3/4 of the original supercharges.3. Non-normalizability of the Nambu–Goldstone fermions We would like to argue that our observation is a generic feature of the Nambu–Goldstone fermions on the domain wall junction in a flat space in the bulk: Nambu–Goldstone fermions are not localized at the junction and hence are not normalizable. If the junction configuration is a 1/4 BPS state. x 2 ) space. analysis in Section 2. some of which may be linear combinations of the supercharges already broken by the first wall. The bird’s eye view of the absolute value of the i = T component of the wave functions of the Nambu–Goldstone fermions on the junction in the (x 1 . Therefore we obtained a chiral structure of Nambu–Goldstone fermions on the junction background configuration. if they are associated with the supersymmetry breaking due to the coexistence of nonparallel domain walls. Fig. 2. The following observation is behind this assertion.

. Q(N) ) is broken. . . Q0(N/4) have a wave function which extends to infinity and approaches a constant profile along both N N the walls 1 and 2. QII = √ e−iπ/4 Q1 + eiπ/4 Q (3. (3. Therefore the Nambu– Goldstone fermions associated with the coexistence of nonparallel domain walls are not localized at the junction and are not normalizable. . Q0(N/2) is broken. Q0(N/4) . Altogether a quarter of the original supercharges remain unbroken. x 2 ) e−iπ/4 (x 1 ) e−iπ/4 2∂1 Aiwall (I)i 1 2  cl → . . Consequently we have Nambu–Goldstone fermions localized around the core of the wall and is constant along the wall. x = 0 0      0 0 → (3. There are only two independent supercharges among QII . let us first note that the junction configuration reduces asymptotically to a wall if one goes along the wall. Q0( 2 ) are unbroken on the wall 1. . say the wall 1. Thus we find that any linear combinations of the Nambu–Goldstone fermions have to be infinitely extended along at least one of the walls which form the junction configuration. However. say wall 2. The other broken supercharge is Q0II on the wall 2 and Q00II on the wall 3. 2∂1 Aiwall 2∂1 Aicl (x 1 . . . The other half. . . Those modes corresponding to Q0( 4 +1) . . . QI . Q0( 2 ) have support only along the wall 2. Q0II . . Q0(N/4) have support only along the wall 1. . Ito et al.54) are broken   1 1 S˙ .60) QI = √ eiπ/4 Q2 + e−iπ/4 Q 2 1 2 2 and there are two corresponding Nambu–Goldstone fermions which become domain wall zero modes asymptotically     4∂z Aicl (x 1 . . ψ0 x . we have other walls which are not parallel to the first wall. Q(N/2) . Asymptotically far away along one of such walls. . . On the wall. another half of the supersymmetry Q0(1) . . . . . . and those corresponding to the linear combinations of Q(1) . In the junction configuration. Therefore the associated Nambu–Goldstone fermions have support which is infinitely extended at least along two of the walls. Consequently the Nambu–Goldstone fermions corresponding to Q0(1) . . If the junction is a 1/4 BPS state. . In our exact solution. Along the other walls we find two broken supercharges one of which is identical to one of the broken supercharges. . . . we call these broken supercharges as Q(1) . say Q0(1) . x 2 ) eiπ/4 cl (x ) e These wave functions are localized on the core of the wall 1 in the x 1 direction and are constant along the wall. domain wall junction configuration reduces asymptotically to the wall 1 at x 2 → −∞ with fixed x 1 .K. . Q0( 4 +1) . We can construct a linear combination of the Nambu–Goldstone fermions to have no support along one out of the three walls. . . . . and Q00II . S˙ . . Q(N/2) broken already on the wall 1. Q(N/2) orthogonal to Q0(1) .61) ψ0(II)i x 1 . . . Denoting the number of original supercharges to be N . a half of these. only two supercharges in Eq. . any linear combination has nonvanishing wave function which becomes fermion zero mode on at least one of the wall asymptotically. Together with QI we obtain three independent broken supercharges. . x 2 = 1 iπ/4 . . / Nuclear Physics B 586 (2000) 231–260 245 To see in more detail. . . is a linear combination of N N Q(1) . On this first wall. . a half of the original supersymmetry (Q(1) .

Eqs. as illustrated in Fig. Let us take the origin in x 1 . The Nambu–Goldstone mode on the domain wall junction is the zero wave number limit of such a spin wave mode. These two supercharges are given. Z2 )I J = 2 WJ∗ − WI∗ · ω (4. we can consider excitation modes which reduce to the spin wave modes along each wall. we can explicitly construct a plane wave solution propagating along the wall.64) . (2. massless graviton is localized on the background of intersection of walls [5–7]. BPS domain wall junction is formed when nonparallel BPS walls meet at a junction. 3(b).246 K. Ito et al. . there are two sets of BPS equations. We set the origin of the W space at an arbitrary point inside this BPS polygon and denote the value of the superpotential at the I th vacuum as WI . BPS domain wall is a 1/2 BPS state and conserves two supercharges.17). If we denote θI J the angle of the half wall separating two vacua. In order to have a balance of force. This physical consideration suggests that the massless Nambu–Goldstone fermion is precisely the vanishing wave number (along the wall) limit of the massive spin wave mode. (2. Z  (4. Boundary conditions and central charges For a 1/4 BPS state.63) ω E I J ≡ cos(θI J + π/2).12). 3(a). N . Let us note that our argument does not apply to models with the bulk cosmological constant.3) and examine if a domain wall junction can be formed where N different vacua appear in asymptotic regions. 4. In this section we make explicit the relation between the boundary conditions and the choice of these BPS equations.13)–(2. the configuration approaches to isolated walls asymptotically.4) as   E I J ≡ (Z1 . (2. sin(θI J + π/2) . S˙ . although the massless mode is buried at the tip of the continuum of massive modes. Even if there are several walls forming a junction configuration. . The normalizability of the massless graviton is guaranteed by the anti-de Sitter geometry away from the junction or intersection including the direction along the wall.28. Q1 + ei(−αI J −θI J ) Q 1 where αI J = arg(WJ − WI ). These N vacua correspond to N points in the complex plane of superpotential W. x 2 space as the junction point of these BPS walls.62) E I J · (Area). Let us take a general Wess–Zumino model in Eq. as illustrated in Fig. The field configuration of the junction at infinity is mapped to a straight line connecting these N vertices [26. . this polygon has to be convex [26]. for I = 1. (2. . the central charges Z1 and Z2 of this wall are given by Eq. In such models. Q2 + ei(−αI J +θI J ) Q 2 (4.11) and (2. from Eqs. In that case.15) and (2. Thus two supercharges conserved at this wall are S˙ . which may be called a spin wave and is among massive modes on the wall background. I and J . / Nuclear Physics B 586 (2000) 231–260 If a single wall is present. corresponding to the two kinds of BPS domain wall junctions.16)– (2. in terms of central charges Z1 and Z2 for the wall.29]. massless mode is a distinct mode different from the massless limit of the massive continuum. They should be a massive mode on the domain wall junction background. In regions far away from the junction.

68) Y3 = dx 3 i d 2 x ∂1 Ki ∂2 Ai − ∂2 Ki ∂1 Ai = dx 3 i Ki dAi . BPS domain wall junction is a 1/4 BPS state and conserves only one supercharge. This contour integral in the field space should be done as a map from a counterclockwise contour in the infinity of . θJ K − θI J = αJ K − αI J . where Ki ≡ ∂K/∂Ai is a derivative of the Kähler potential K. 3.67) and that the field configuration at infinity should move clockwise in W space.15) for H = HI should be used. the other BPS equations (2. Therefore we find that the BPS equations (2.16)–(2. as we go around the origin counterclockwise in x 1 . · · · = Q2 + ei(−αI J +θI J ) Q 2 2 (4. / Nuclear Physics B 586 (2000) 231–260 (a) 247 (b) Fig. We can use the Stokes theorem to obtain an expression for the central charge Y3 as a contour integral [26. x 2 space. x 2 space and BPS polygon in W space. .17) for the case H = HII should be used if the phase of the superpotential W increases as we go around the origin counterclockwise in x 1 . . x 2 space. θJ K − θI J = −(αJ K − αI J ). . S ˙ is the common conserved supercharge in Similarly. . This must be the common conserved supercharge for all the walls S˙ = Q2 + ei(−αJ K +θJ K ) Q S˙ = · · · . Walls in x 1 . (a) Walls separating vacua I.66) Moreover the field configuration at infinity should move counterclockwise in W space.65) Then the relative angle of the two neighboring walls must be equal to the difference of two phases of the differences 1W of the superpotentials for the two walls . Let S ˙ is conserved as us consider the case of H = HII where a linear combination of Q2 and Q 2 shown in Eq. . .12).30] Z Z I Z    (4. a linear combination of Q1 and Q 1 the case of H = HI . We obtain in this case . . . (b) BPS polygon in W space. x 2 space. If the phase of the superpotential W decreases as we go around the origin counterclockwise in x 1 . . x 2 space. Ito et al. (4. . .13)– (2. J. . as we go around the origin counterclockwise in x 1 . x 2 space. . K in x 1 . (4.K. Next we discuss the sign of the contribution of the central charge Y3 to the mass of the junction configuration. (2.

These models. Let us assume for simplicity that there is only one field which can contribute to Y3 as in our exact solution. Only complex fields can contribute to Y3 . should correspond to the case of opposite sign of rotations in W space and field space. the central charge contributes to the mass of the junction configuration negatively.24) and then the central charge is negative in this solution. but the contribution to the mass H = HI becomes again negative. We shall study their densities and integrated quantities in finite regions in this section. if they exist. and positive if the sign of rotations is opposite. the sign of the contribution of the central charge Y3 to the mass of the junction configuration is determined by the formula H = HII = |hiZ1 − Z2 i| + hY3 i. x 2 is mapped into a counterclockwise (clockwise) contour in field space. The solution is just a configuration obtained by a reflection x 1 → −x 1 . Since the exact solution satisfies the BPS equation for the case H = HII .68) shows that the central charge Y3 becomes negative (positive). x 2 space.1. The choice of these mass formulas are in turn determined by the map of the asymptotic counterclockwise contour in x 1 . / Nuclear Physics B 586 (2000) 231–260 z = x 1 + ix 2 plane. Energy density and central charges 5. In either solution. Therefore we should not consider the central charge Y3 alone as the physical mass of the junction at the center. More recently this feature of negative contribution of Y3 to the junction mass is studied from a different viewpoint and it is argued that this feature is valid in most situations except possibly in contrived models [36]. Therefore central charge Y3 contributes negatively to the mass of the junction. we conclude that the contribution of the central charge Y3 to the mass of the junction configuration is negative if the sign of rotations is the same in field space Ai and in superpotential space W.1.1. Y charge density The Y3 charge density Y3 = i 3nm ∂n (T ∗ ∂m T ) is given in our exact solution by Y3 = −24Λ  4 √ e e 3 Λx 2 √ 3 Λx 2 +2 cosh(Λx 1 ) 3 . We also can find a solution for the other case of H = HI in our model. the rotation in field T space has the same sign as the rotation in superpotential W space. The field configuration moves counterclockwise in field space in our exact solution in (2. Then the central charge is positive.248 K. x 2 space to a counterclockwise or clockwise contour in the superpotential space W. On the other hand. (4. Charge densities Our exact solution is useful to examine how the topological charges Zk . Eq. Yk and energy of the domain wall junction are distributed in x 1 . Ito et al. if the asymptotic counterclockwise contour in x 1 . In the junction configuration. the junction at the center cannot be separated from the walls. Combining these two observations. 5. or H = HI = |h−iZ1 − Z2 i| − hY3 i. irrespective of the choice of H = HI or H = HII . 5.

Im Z1 = −Re Z2 = −48 3 Λ 4  √ e 3 Λx 2 +2 cosh(Λx 1 ) √ √   3 Λx 2 cosh(Λx 1 ) + e 3 Λx 2 2 + 3 cosh(2Λx 1 ) 4e . Re W = −8Λ 3  √ e 3 Λx 2 +2 cosh(Λx 1 ) √ √ √   3Λx 2 2 + e2 3Λx 2 +6 e 3 Λx 2 cosh(Λx 1 ) √ e ∗ 3 . A bird’s eye view of Y3 .70) Im W = 2 3 Λ 3  √ e 3 Λx 2 +2 cosh(Λx 1 ) The Z charge densities are given by Zk = 2∂k W ∗ (k = 1.2. (5. Ito et al. √  4 e 3 Λx 2 +2 cosh(Λx 1 ) √ √  2 2 √ 4 e 3 Λx sinh(Λx 1 ) 1 + 2 e 3 Λx cosh(Λx 1 ) . by inserting the solution (2.69) where the cylindrical coordinates r and θ is used to make Z3 symmetry explicit.1. 4. 2) and are found to be 2 + e2 Re Z1 = −48Λ4 √ 3 Λx 2 √ 3 Λx 2 cosh(2Λx 1 ) + 3 e cosh(Λx 1 ) . The density is localized near the origin and the Z3 symmetry is manifest. (5.K. Here and the following. Corresponding effective charge density is given by . A bird’s eye view of the Y3 is given in Fig. 4. (5.24) √  3 Λx 2 cosh(Λx 1 ) + cosh(2Λx 1 ) sinh(Λx 1 ) ∗ 3 2 + 3e . / Nuclear Physics B 586 (2000) 231–260 249 Fig. we shall take the unit of Λ ≡ 1 in drawing figures. Z charge density We obtain the superpotential as the function of x 1 and x 2 . = −24Λ4  1 e 2Λr √ 3 sin θ +e 2Λr √ 3 sin(θ+ 2π 3 ) +e 2Λr √ 3  sin(θ− 2π 3 ) 3 .71) Im Z2 = 48Λ 4  √ e 3 Λx 2 +2 cosh(Λx 1 )  We can define the effective value of the Z charge which contributes to the energy of the junction as Zeff = −Re Z1 + Im Z2 . 5.

/ Nuclear Physics B 586 (2000) 231–260 √ 3 Λx 2 41 + 2e Zeff = 96Λ √  2 cosh(Λx 1 ) + e2 3 Λx 1 + 2 cosh(2Λx 1 ) . 7. .g.73) A bird’s eye view of H is shown in Fig. H. H = 24Λ 4  √ e 3 Λx 2 +2 cosh(Λx 1 ) (5. (5. Ito et al. a small dip is found around RΛ ∼ 1. The energy density is Z3 symmetric as expected.3. Far from the origin there is practically no difference between Zeff and H because Y3 is localized near the origin.72) Let us note that the effective charge density is Z3 symmetric..2. whereas individual charges Z1 . 5. 6. Since the decrease of Zeff is faster than the increase of Y3 . 5. Y3 charge Integrating Y3 over a triangle whose inscribed circle has a radius R as shown in Fig. 5. The Zeff charge contributes to the energy positively while Y3 does negatively.74) Y3 4 Fig. √ √  3 Λx 2 cosh(Λx 1 ) + e2 3 Λx 2 3 + 8 cosh(2Λx 1 ) 44 + 6e . 4  √ e 3 Λx 2 +2 cosh(Λx 1 ) (5.250 K. Zeff and Y3 along one of the walls (e. A cross section of the densities.1. 5. Z2 are not. Energy density Adding Zeff and Y3 together. 5. 7.1. Charge densities integrated over a region of finite radius In this subsection we shall evaluate the central charge densities integrated over a triangular or circular region depicted in Fig. negative x 2 direction) is shown in Fig. the energy density of the junction is obtained. we obtain for large R (RΛ  1)   √ √ 2  3π −√3R Λ triangle −2 3R Λ e (R) = −2 3 Λ L3 1 − +O e . A bird’s eye view of the energy density of the junction.2.

7. The leading term agrees with our previous evaluation by the step function approximation [30] and the subleading terms vanish exponentially as R → ∞.75) Y3 45 3 . A cross section of the densities H (solid line).K. where L3 denotes the length along the x 3 direction. Ito et al. The bold lines denote the domain walls forming a junction. for small R (RΛ  1). Fig. / Nuclear Physics B 586 (2000) 231–260 251 Fig. we obtain the Y3 charge √    8 2 4 3 5 5 triangle 2 2 4 4 6 6 R Λ +O R Λ . The domains of integration: triangle (solid line). On the other hand. inscribed circle (short dashed line) and circumscribed circle (long dashed line). (R) = − √ Λ L3 R Λ − R Λ − (5. Zeff (short dashed line) and Y3 (long dashed line) along the negative x 2 direction. 6.

In the limit of R → ∞. 2) denote the upper and lower bound of the domain of integration. Z charges Since the Z1 charge is given by a total derivative in x 1 . and x 3 ). L Z3 /2 Y3circle (R) = Z2π dx −L3 /2 = 8RΛ3 L3 √ 3 × 3   0 ∂ 0 dθ iT ∗ T ∂θ 0 Z2π dθ 0 sin θ − 2π 3    2RΛ cos 2π e2RΛ cos θ + sin θ e2RΛ cos θ+ 3 + sin θ + 2π e 3 2π 2RΛ 2π 2RΛ  2RΛ  √ sin(θ+ √ sin(θ− √ sin θ 3 3 ) 3 ) e 3 +e 3 +e 3 θ− 2π 3  . =2 (5. we obtain . x 2 ) is odd in x 1 . Although there are no terms of the first or third degree in R.2. Ito et al.77) Y3circle (R) = − Λ2 L3 R 2 Λ2 − R 4 Λ4 + O R 6 Λ6 .2. 9 2 The leading term is again the density at the origin multiplied by the area of the circle. We can also integrate the Y3 over a circle of radius R. (5. x 2 − W ∗ x 1− x 2 . 5. (5.76) where the Z3 symmetry is manifest. we obtain    8π 1 (5. Y3 (R) for all the regions converge to −2 3 Λ2 L3 as expected. 8. and the following surface integral formula obtained from the Stokes theorem. inscribed and circumscribed √ in Fig. there is no term of odd degree in R. x 2 .70) shows that Im W ∗ (x 1 . we obtain Im Z1 (R) = 0 and Re Z2 (R) = 0 for an integration region symmetric in x 1 which we shall use. we can rewrite the Z1 charge as Z Z Z1 (R) = 2 dx 2 dx 1 ∂1 W ∗ (A∗ ) Zx 2+ h    i dx 2 W ∗ x 1+ x 2 . Since Eq. We compare the Y3 (R) evaluated for triangle.252 K. there is a fifth degree term. x 2 ) is even and Re W ∗ (x 1 . / Nuclear Physics B 586 (2000) 231–260 Notice that the leading term comes from the density at the origin multiplied by the area of the circle. For large R (RΛ  1).78) x 2− where x i± (i = 1. Let us note that Re Z1 (Im Z2 ) is negative (positive) definite. Expanding the integrand for small R (RΛ  1). In this case it is more convenient to use cylindrical coordinates (r. we have to perform numerical integration to evaluate the circle Y3circle (R). For large R (RΛ  1). θ . In contrast to the triangle case. Firstly we choose as a domain of integration the triangle region whose inscribed circle has a radius R.

√   √  3 π −√3 RΛ e +O e−2 3 RΛ . Im Z2 3 8 3 triangle Re Z1 (R) = −12Λ2 L3 (5. (5.79) The leading linear term represents the contribution of charge density per unit length of the wall. The effective value of the Z charge becomes triangle Zeff triangle (R) = −Re Z1 triangle (R) + Im Z2 (R)   √ √ √ 7π 16  − Λ2 L3 e− 3 RΛ +O e−2 3 RΛ . 3 (5. The effective value of Z charge is triangle Zeff  64 (R) = √ R 2 Λ4 L3 + O R 6 Λ6 .82) We can also choose a circle of radius R as a domain of integration.K. Z charges become √    32 2 1 4 4 2 3 5 5 triangle 2 2 R Λ + O R 6 Λ6 . = 24Λ3 L3 R + 3 2 3 (5. Y3 (R) evaluated for the triangle (solid line). For small R (RΛ  1) we obtain . It is interesting to observe that there are no constant terms. (R) = − √ Λ L3 R Λ + R Λ − Re Z1 2 9 3 √    32 2 1 4 4 2 3 5 5 triangle 2 2 6 6 R Λ + O R Λ . 8. The exponentially suppressed terms represent the way the domain wall junction configuration converges to isolated walls as R → ∞. Ito et al.80) For small R (RΛ  1). RΛ + 4 √     √  4 −√3 RΛ 3 π triangle e − (R) = 12Λ2 L3 RΛ + +O e−2 3 RΛ . and that there is no term of the first or third degree in R. inscribed circle of radius R (short dashed line) and circumscribed circle (long dashed line). / Nuclear Physics B 586 (2000) 231–260 253 Fig.81) (R) = √ Λ L3 R Λ − R Λ + Im Z2 2 9 3 Notice that the leading term represents the density at the origin multiplied by the area of the triangle.

circle (R) for inscribed circle converges to the same value as the asymptotic slope of Zeff that for the triangle. Im Z2circle (R) = 9 4 Re Z1circle (R) = − (5. In the case of the circumscribed circle.254 K. inscribed circle and circumscribed circle in Fig.2. and circumscribed circle (long dashed line). The effective value of the Z charge is    64π 2 1 4 4 circle 2 2 6 6 Λ L3 R Λ − R Λ + O R Λ . 9. the energy is   √ √ 2 √ 16 3 Λ2 L3 e− 3 RΛ Htriangle(R) = 24Λ L3 R − 2 3 Λ L3 + 3 5π − 3 √  −2 3 RΛ +O e .3. (5. Firstly we choose the triangle region whose inscribed circle has radius R.83) The leading term is again given by the densities at the origin multiplied by the area of the circle. 9 4    32π 1 Λ2 L3 R 2 Λ2 − R 4 Λ4 + O R 6 Λ6 .    32π 2 1 Λ L3 R 2 Λ2 − R 4 Λ4 + O R 6 Λ6 . Energy of the junction Since our exact solution satisfies the BPS equation corresponding to H = HII . ins. / Nuclear Physics B 586 (2000) 231–260 Fig. We compare the effective Z value evaluated for triangle.84) Zeff (R) = 9 4 A numerical evaluation is needed for large R. As R → ∞. For large R (RΛ  1). the energy of the junction is obtained by adding Y3 and Zeff together H = Zeff + Y3 = −Re Z1 + Im Z2 + Y3 . Ito et al. since the total length of the walls is twice as long as those of the other cases.85) . 9. (5. Zeff (R) evaluated for the triangle (solid line). 5. the inscribed circle of the radius R (short dashed line). the Zeff becomes twice as large as those of the other cases for large R.

(5. Finally we plot the θ -dependence of the energy and charges that are obtained by integrating the densities from r = 0 to r = R with θ fixed (see Fig. 11(a).O. In Fig. 11(b) and (c)). For large R (RΛ  1). reflecting the fact that the junction is a string-like object and symmetric around the x 3 axis. For small R (RΛ  1). 10. 11). / Nuclear Physics B 586 (2000) 231–260 255 Fig. all the quantities are almost uniform in θ near the junction at the center. . the energy H reduces to Zeff . This work is supported in part by Grant-in-Aid for Scientific Research from the Japan Ministry of Education.86) Htriangle(R) = √ Λ L3 R Λ + R Λ + 7 315 3 In the case of the circle of radius R. The energy of the junction configuration H (R) evaluated for the triangle (solid). As we move away from the origin. Y3 disappears and the energy H approaches Zeff (see Fig.87) Hcircle (R) = 9 14 The energy of the triangle region is compared to those of inscribed and circumscribed circles in Fig. As R grows. main contribution comes from the direction of the walls (in our case −π/2.) gratefully acknowledges support from the Iwanami Fujukai Foundation. the energy is √    56 2 1 4 4 4 3 5 5 2 2 6 6 R Λ +O R Λ . inscribed circle of radius R (short dashed line) and circumscribed circle (long dashed line). 11(d)). 10. Ito et al. The first linear term can be regarded as the contribution from the walls and the second constant term can be regarded as the contribution from the junction at the center. and 5π/6) (see Fig. (5.K. π/6. In each figure the energy H (solid line) is the sum of the Zeff (short dashed line) and Y3 (long dashed line) and Z3 symmetry is manifest in their θ -dependence. Science and Culture for the Priority Area 291 and 707. Acknowledgements One of the authors (H. the energy is given for small R (RΛ  1) as    56π 2 3 4 4 2 2 6 6 Λ L3 R Λ − R Λ + O R Λ .

The θ -dependence of the energy and charges that are obtained by integrating the densities over the radial direction up to R with θ fixed. (2.3) ∂Ai The canonical energy momentum tensor is given by . In each figure horizontal axis denotes θ . Kähler metric is assumed to be minimal Kij ∗ = δij ∗ : i i L = −∂µ A∗j ∂ µ Aj + F ∗j F j + ∂µ ψ¯ j σ¯ µ ψ j − ψ¯ j σ¯ µ ∂µ ψ j 2 2 ∗ 1 i j ∂W 1 i j ∂W ∗ ∂W ∗j ∂W ψ +Fj − ψ + F − . (A. and solid. This is the starting Lagrangian to derive central charges and we will not neglect any total divergences from now on. (a) RΛ = 0. 11.5 case. (A. The canonical supercurrent is found to be √ √   Jαµ = 2 σ ν σ¯ µ ψ i α ∂ν Ai∗ + i 2 σ µ ψ¯ i α F i √ √   ∂W ∗ = 2 σ ν σ¯ µ ψ i α ∂ν Ai∗ − i 2 σ µ ψ¯ i α . Fermionic contributions to central charges We shall derive the central charges including fermionic contributions in the case of a general Wess–Zumino model with an arbitrary superpotential W.2) ∂A∗i √ √ α˙ α˙ J¯µα˙ = 2 σ¯ ν σ µ ψ¯ i ∂ν Ai + i 2 σ¯ µ ψ i F i∗ √ √ α˙ α˙ ∂W = 2 σ¯ ν σ µ ψ¯ i ∂ν Ai − i 2 σ¯ µ ψ i . (c) RΛ = 2 case. For simplicity. short dashed and long dashed lines correspond to H . (b) RΛ = 1 case. Zeff and Y3 . / Nuclear Physics B 586 (2000) 231–260 (a) (b) (c) (d) Fig.1) ψ¯ ψ¯ ∂Aj 2 ∂Ai ∂Aj ∂A∗j 2 ∂A∗i ∂A∗j We have added a surface term to Eq. (A. (d) RΛ = 3 case. Ito et al. Appendix A. respectively.256 K.3) to make the variational principle meaningful.

Ito et al. / Nuclear Physics B 586 (2000) 231–260 i i T µν = ∂ µ Aj ∂ ν A∗j + ∂ µ A∗j ∂ ν Aj + ψ¯ j σ¯ µ ∂ ν ψ j + ψ j σ µ ∂ ν ψ¯ j 2 2 .K.

.

2  .

.

∂W i i + g µν −∂λ A∗j ∂ λ Aj − .

.

j

+ ∂λ ψ¯ j σ¯ λ ψ j − ψ¯ j σ¯ λ ∂λ ψ j
∂A
2
2 

1 i j ∂W
1 i j ∂W ∗
− ψ ψ
− ψ¯ ψ¯
.
2
∂Ai ∂Aj
2
∂A∗i ∂A∗j
Canonical quantization gives (anti-) commutation relations  

i
A (x), ∂0 A∗j (y) x 0 =y 0 = iδ 3 (x − y)δ ij ,
o
n
j
ψαi (x), ψ¯ β˙ (y) 0 0 = −δ 3 (x − y)δ ij σα0β˙ .
x =y

257

(A.4)

(A.5)

The anticommutator between supercharges of the same chirality gives the supersymmetry
algebra (2.1) with the central charge Zk in Eq. (2.4)
Z
(A.6)
Zk = 2 d 3 x ∂k W ∗ (A∗ )
which turns out to have only bosonic contributions. The anticommutator between
supercharges of the opposite chirality gives the supersymmetry algebra (2.2) with the
central charge Yk  

Z
1 ¯j
knm
3
∗j
j
j
Yk = i
(A.7)
d x ∂n A ∂m A − ψ σ¯ m ψ ,  123 = 1,
2
which has both bosonic and fermionic contributions.

Appendix B. Gamma matrices and fermion mode equations
In order to separate variables (x 1 , x 2 ) and (x 0 , x 3 ) for spinors, it is most convenient to
use a gamma matrix representation where the direct product structure of 2 × 2 matrices in
(x 1 , x 2 ) and (x 0 , x 3 ) becomes manifest. One such representation is
γ 0 = ρ1,

γ 3 = iρ 2 ,

γ 1 = iσ 1 ρ 3 ,

γ 2 = iσ 2 ρ 3 ,

(B.1)

are Pauli matrices acting on 2 × 2 matrices and
acting on indices of blocks
where
of these 2 × 2 matrices. The four component spinor can be decomposed into a pair of two
component spinors ξ and χ in 0 + 2 dimensions 

ξ
ψ=
.
(B.2)

σa

ρa

The B matrix for 1 + 3 dimensions can be defined as a product of B matrices B (1) for 1 + 1
dimensions and B (2) for 0 + 2 dimensions
B = B (1) B (2) ,

B (1) = ρ 3 ,

B (2) = −iσ 1 .

(B.3)

The Majorana condition for the (1 + 3)-dimensional spinor and the pseudo-Majorana
condition for the (0 + 2)-dimensional spinor are given by
ψ = Bψ ∗ ,

ξ = B (2) ξ ∗ ,

χ = B (2) χ ∗ ,

(B.4)

258

K. Ito et al. / Nuclear Physics B 586 (2000) 231–260

which implies ξ1 = −iξ2∗ for components of the two component spinor ξ = (ξ1 , ξ2 )T , and
similarly for χ .
The Dirac equation for four component fermions in the general Wess–Zumino model
reads 
2 

∂ 2 W ∗ 1 − iγ 5
∂ W 1 + iγ 5
µ
i
+
(B.5)
ψ j = 0.
−iγ ∂µ ψ −
∂Ai ∂Aj
2
∂A∗i ∂A∗j
2
j

The mode equation in x 1 , x 2 space is defined in terms of two component spinors ξn and
j
χn   

∂ 2W ∗ 1 + σ 3 j
∂ 2W 1 − σ 3
i
ξn = −m(1)

δ ij σ 1 ∂1 + σ 2 ∂2 −
n χn ,
∂Ai ∂Aj 2
∂A∗i ∂A∗j 2   

∂ 2W ∗ 1 − σ 3 j
∂ 2W 1 + σ 3
i

χn = −m(2)
−δ ij σ 1 ∂1 + σ 2 ∂2 −
n ξn .
∂Ai ∂Aj 2
∂A∗i ∂A∗j
2
(B.6)
The two component spinors ξni , χni satisfy the pseudo-Majorana condition (B.4). Then we
find that mass eigenvalues are real  

(1) ∗
(2) ∗
,
m(2)
.
(B.7)
m(1)
n = mn
n = mn
We can make a separation of variables for the Dirac equation (B.5) by means of real
fermionic fields cn and bn
X  cn (x 0 , x 3 )ξn (x 1 , x 2 ) 
.
(B.8)
ψi =
ibn (x 0 , x 3 )χn (x 1 , x 2 )
n
Using these mode functions we find that the fermion fields (cn , ibn )T satisfy the Dirac
equation in 1 + 1 dimensions in Eq. (3.46).
This representation can be related to the usual Weyl representation in Ref. [33] by the
following unitary matrix U
1 − σ3 3 1 + σ3 2
ρ +
ρ ,
2
2
µ
γWeyl = U † γ µ U,

U=

BWeyl = U † BU ∗ = σ 2 ρ 2 .

(B.9)

The two component spinor in the Weyl representation is related to the components of the
four component spinor (B.2) in the representation (B.1) in this appendix as


χ1  

 ξ2 
ψα
.
(B.10)
=
α˙
¯
ψ Weyl  iξ1 
−iχ2
Thus we obtain the mode equation (3.41)–(3.44) in the Weyl representation.

K. Ito et al. / Nuclear Physics B 586 (2000) 231–260

259

References
[1] N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys. Lett. B 429 (1998) 263, hep-ph/9803315.
[2] I. Antoniadis, N. Arkani-Hamed, S. Dimopoulos, G. Dvali, Phys. Lett. B 436 (1998) 257,
hep-ph/9804398.
[3] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 3370, hep-ph/9905221.
[4] L. Randall, R. Sundrum, Phys. Rev. Lett. 83 (1999) 4690, hep-th/9906064.
[5] N. Arkani-Hamed, S. Dimopoulos, G. Dvali, N. Kaloper, Phys. Rev. Lett. 84 (2000) 586, hepth/9907209.
[6] C. Csaki, Y. Shirman, Phys. Rev. D 61 (2000) 024008, hep-th/9908186.
[7] A.E. Nelson, A new angle on intersecting branes in infinite extra dimensions, hep-th/9909001.
[8] K. Behrndt, M. Cvetiˇc, Supersymmetric domain wall world from D = 5 simple gauged
supergravity, hep-th/9909058.
[9] K. Skenderis, P. Townsend, Phys. Lett. B 468 (1999) 46, hep-th/9909070.
[10] A. Chamblin, G.W. Gibbons, Supergravity on the Brane , hep-th/9909130.
[11] O. DeWolfe, D.Z. Freedman, S.S. Gubser, A. Karch, Modeling the fifth dimension with scalars
and gravity, hep-th/9909134.
[12] E. Bogomol’nyi, Sov. J. Nucl. Phys. 24 (1976) 449.
[13] M.K. Prasad, C.H. Sommerfield, Phys. Rev. Lett. 35 (1975) 760.
[14] E. Witten, D. Olive, Phys. Lett. B 78 (1978) 97.
[15] E.R.C. Abraham, P.K. Townsend, Nucl. Phys. B 351 (1991) 313.
[16] G. Dvali, M. Shifman, Phys. Lett. B 396 (1997) 64, hep-th/9612128.
[17] G. Dvali, M. Shifman, Nucl. Phys. B 504 (1997) 127, hep-th/9611213.
[18] A. Kovner, M. Shifman, A. Smilga, Phys. Rev. D 56 (1997) 7978, hep-th/9706089.
[19] A. Smilga, A. Veselov, Phys. Rev. Lett. 79 (1997) 4529, hep-th/9706217.
[20] B. de Carlos, J.M. Moreno, Phys. Rev. Lett. 83 (1999) 2120, hep-th/9905165.
[21] G. Dvali, G. Gabadadze, Z. Kakushadze, Nucl. Phys. B 562 (1999) 158, hep-th/990103.
[22] V. Kaplunovsky, J. Sonnenschein, S. Yankielowicz, Nucl. Phys. B 552 (1999) 209, hepth/9811195.
[23] V.A. Rubakov, M.E. Shaposhnikov, Phys. Lett. B 125 (1983) 136.
[24] B. Chibisov, M. Shifman, Phys. Rev. D 56 (1997) 7990, hep-th/9706141.
[25] G. Gibbons, P.K. Townsend, Phys. Rev. Lett. 83 (1999) 1727, hep-th/9905196.
[26] S.M. Carroll, S. Hellerman, M. Trodden, Phys. Rev. D 61 (2000) 065001, hep-th/9905217.
[27] A. Gorsky, M. Shifman, Phys. Rev. D 61 (2000) 085001, hep-th/9909015.
[28] P.M. Saffin, Phys. Rev. Lett. 83 (1999) 4249, hep-th/9907066.
[29] D. Bazeia, F.A. Brito, Phys. Rev. Lett. 84 (2000) 1094, hep-th/9908090.
[30] H. Oda, K. Ito, M. Naganuma, N. Sakai, Phys. Lett. B 471 (1999) 148, hep-th/9910095.
[31] P. Townsend, in: PASCOS/Hopkins, 1995, p. 0271, hep-th/9507048.
[32] P. Townsend, Class. Quant. Grav. 17 (2000) 1267, hep-th/9911154.
[33] J. Wess, J. Bagger, Supersymmetry and Supergravity, Princeton University Press, 1991.
[34] D. Binosi, T. tel Veldhuis, Phys. Lett. B 476 (2000) 124, hep-th/9912081.
[35] D. Bazeia, F.A. Brito, Bags, junctions, and networks of BPS and non-BPS defects, hepth/9912015.
[36] M. Shifman, T. tel Veldhuis, Calculating the tension of domain wall junctions and vortices in
generalized Wess–Zumino models, hep-th/9912162.
[37] S.M. Carroll, S. Hellerman, M. Trodden, BPS domain wall junctions in large extra dimensions,
hep-th/9911083.
[38] N. Arkani-Hamed, S. Dimopoulos, G. Dvali, N. Kaloper, Manyfold universe, hep-ph/9911386.
[39] S. Nam, JHEP 0003 (2000) 005, hep-th/9911104.
[40] I. Bakas, A. Brandhuber, K. Sfetsos, Domain walls of gauged supergravity, M-branes and
algebraic curves, hep-th/9912132.

260

K. Ito et al. / Nuclear Physics B 586 (2000) 231–260

[41] M. Cvetiˇc, H. Lü, C.N. Pope, Domain walls and massive gauged supergravity potentials, hepth/0001002.
[42] C. Csaki, J. Erlich, T.J. Hollowood, Y. Shirman, Universal aspects of gravity localized on thick
branes, hep-th/0001033.
[43] R. Kallosh, A. Linde, JHEP 0002 (2000) 005, hep-th/0001071.
[44] S. Nam, K. Olsen, Domain wall junctions in supersymmetric field theories in D = 4, hepth/0002176.
[45] R. Altendorfer, J. Bagger, D. Nemeschansky, Supersymmetric Randall–Sundrum scenario,
hep-th/0003117.
[46] N. Alonso-Alberca, P. Meessen, T. Ortin, Supersymmetric brane-worlds, hep-th/0003248.
[47] J.P. Gauntlett, G.W. Gibbons, P.K. Townsend, Intersecting domain walls in MQCD, hepth/0004136.

Nuclear Physics B 586 (2000) 261–274
www.elsevier.nl/locate/npe

Models of dynamical supersymmetry breaking with
gauged U (1)R symmetry
Noriaki Kitazawa a,1 , Nobuhito Maru b,∗ , Nobuchika Okada c,2
a Department of Physics, Tokyo Metropolitan University, Hachioji, Tokyo 192-0397, Japan
b Department of Physics, Tokyo Institute of Technology, Oh-Okayama, Meguro, Tokyo 152-8551, Japan
c Theory Group, KEK, Tsukuba, Ibaraki 305-0801, Japan

Received 5 April 2000; accepted 26 June 2000

Abstract
We present simple models of dynamical supersymmetry breaking with gauged U (1)R symmetry.
The minimal supersymmetric standard model and supersymmetric SU(5) GUT are considered as
the visible sector. The anomaly cancellation conditions for U (1)R are investigated in detail and
simple solutions of the R-charge assignments are found. We show that this scenario of dynamical
supersymmetry breaking is phenomenologically viable with the gravitino mass of order 1 TeV or 10
TeV.  2000 Elsevier Science B.V. All rights reserved.
PACS: 12.60.Jv; 11.30.Na; 12.60.Cn
Keywords: Supergravity; Gauged U (1)R ; Dynamical supersymmetry breaking

1. Introduction
Supersymmety is motivated to solve the gauge hierarchy problem. However, since
the superparticles have not been observed yet, supersymmetry should be broken at low
energies. Spontaneous supersymmetry breaking at the tree-level [1,2] does not explain
why the discrepancy between the supersymmetry breaking scale and the Planck scale
is so large. On the other hand, in the models of dynamical supersymmetry breaking
the supersymmetry breaking scale is related to the Planck scale via the dimensional
transmutation [3]. In the light of this fact, many people have constructed the models of
the dynamical supersymmetry breaking [4,5] and discussed the phenomenology so far [6,
7]. As for the mediation mechanism of supersymmetry breaking to the visible sector, there
∗ Corresponding author. E-mail: maru@th.phys.titech.ac.jp; JSPS Research Fellow; Department of Physics,

University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033, Japan.
1 kitazawa@phys.metro-u.ac.jp
2 okadan@camry.kek.jp; JSPS Research Fellow.
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 4 1 5 - 6

262

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

are mainly two ways. One is the gravity mediation [6] (including the anomaly mediation [8,
9]), the other is the gauge mediation [7] (including the anomalous U (1) mediation [10,11]).
Many authors have extensively studied both scenarios and proposed interesting models.
In our previous letter [12], we proposed a simple mechanism of the dynamical
supersymmetry breaking with gauged U (1)R symmetry. Although the attempts to make
use of the gauged U (1)R symmetry in supergravity theories can be found in Refs. [13,
14], the mechanism of supersymmetry breaking was not the main topics in these works.
Supersymmetry can be dynamically broken by the interplay between the Fayet–Iliopoulos
term of U (1)R and the dynamically generated superpotential due to the non-perturbative
dynamics of the gauge theory with vanishing cosmological constant. However, we did not
discuss in detail the anomaly cancellation for U (1)R taking into account the full particle
contents, since we focused on the dynamics of supersymmetry breaking in the hidden
sector. The main purpose of this paper is to pursuit this point. We discuss two models as
the visible sector: one is the minimal supersymmetric standard model (MSSM), the other
is the supersymmetric SU(5) grand unified theory (GUT). In the model of MSSM, we use
the Green–Schwarz mechanism for anomaly cancellation. We find quite simple solutions
of R-charge assignments in both cases. The spectrum of the supersymmetry breaking mass
is also discussed. The scalar fields receive the soft supersymmetry breaking masses, which
is the same order of the gravitino mass from the tree-level interactions of supergravity.
Moreover, the scalar fields with non-zero R-charges obtain additional soft supersymmetry
breaking masses through U (1)R D-term, which is also the same order of the gravitino mass.
The masses of gauginos in the MSSM depend on the form of gauge kinetic functions. If
the gauge kinetic function includes the higher-dimensional term, the gaugino masses are
generated in the same order of the gravitino mass or less. If the gauge kinetic function is
trivial, the gaugino masses are generated through the anomaly mediation in a few orders
smaller than the gravitino mass. We see that our scenario is phenomenologically viable
with the gravitino mass of order 1 TeV or 10 TeV. The comments on the generation of the
Higgs potential from higher-dimensional interactions are given.
This paper is organized as follows. In Section 2, we review our mechanism of dynamical
supersymmetry breaking with gauged U (1)R symmetry. Then this mechanism is applied to
MSSM of the visible sector in Section 3 and applied to the supersymmetric SU(5) GUT of
the visible sector in Section 4. The last section is devoted to the summary and discussion.

2. Dynamical supersymmetry breaking with gauged U (1)R symmetry
In this section, we review our previous letter, in which we proposed a simple mechanism
of dynamical supersymmetry breaking with gauged U (1)R symmetry in the context of the
minimal supergravity. The model is based on the gauge group SU (2)H × U (1)R with the
following matter contents 3 :
3 In our previous letter [12], we did not discuss the cancellation of the gauge anomaly of [U (1) ]3 and the
R
mixed gravitational anomaly of U (1)R . We simply assumed there that these anomalies are cancelled out, if all

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

Q1
Q2
S

SU(2)H
2
2
1

263

U (1)R
−1
−1
+4

The general renormalizable superpotential at the tree-level is
W = λS[Q1 Q2 ],

(1)

where the square brackets denote the contraction of SU(2) indices by the -tensor, λ is a
dimensionless coupling constant. We assume that λ is real and positive.
It is known that the superpotential is generated dynamically by non-perturbative
(instanton) effect of the SU(2) gauge dynamics [15]. The total effective superpotential
is found to be
Weff = λS [Q1 Q2 ] +

Λ5
,
[Q1 Q2 ]

(2)

where the second term is the dynamically generated superpotential and Λ is the dynamical
scale of the SU(2)H gauge interaction. Note that the supersymmetric vacuum lies at hSi →
∞ and hQ1 i, hQ2 i → 0, if only the F-term potential is considered.
Next, let us consider the D-term potential. The gauged U (1)R symmetry is impossible
in the globally supersymmetric theory, since the generators of the U (1)R symmetry and
supersymmetry do not commute with each other. On the other hand, in the supergravity
theory the U (1)R symmetry can be gauged as if it were a usual global symmetry [13,14,
16,17]. However, it should be noticed that the Fayet–Iliopoulos term of the gauged U (1)R
symmetry appears due to the symmetry of supergravity. This fact is easily understood by
the standard formula for supergravity theories [18]. Using the generalized Kähler potential
P
G = K + ln |W |2 , we have D = i qi (∂G/∂zi )zi , where qi is the U (1)R charge of the
field zi . Note that the contribution from the superpotential leads to the constant term, since
the superpotential is holomorphic and has U (1)R charge 2.
With the above particle contents, the D-term potential is found to be 
2
gR2
4S † S − Q†1 Q1 − Q†2 Q2 + 2MP2 ,
(3)
2

where MP = Mpl / 8π is the reduced Planck mass, gR is the U (1)R gauge coupling, and
the minimal Kähler potential, K = S † S + Q†1 Q1 + Q†2 Q2 , is assumed. 4 Note that the
supersymmetric vacuum conditions required by the D-term potential and by the effective
superpotential of Eq. (2) are incompatible. Therefore, supersymmetry is broken. This
consequence remains correct, if there is no other superfields which have negative U (1)R
charges, or if the other negatively charged superfields (if they exist) have no vacuum
expectation value.
VD =

particle contents are considered with an appropriate U (1)R charge assignment. In this paper, we will discuss this
issue in detail.
4 This assumption is justified by our result with Λ  M which means that the SU(2) gauge interaction is
P
H
weak at the Planck scale.

264

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

Let us analyze the total potential in our model. Here, note that the cosmological constant
should vanish. This requirement comes not only from the observations of the present
universe but also from the consistency of our discussion. Since it is not clear whether
the superpotential discussed above can be dynamically generated even in the curved space,
the space–time should be flat for our discussion to be correct. Note that we cannot take the
usual strategy, namely, adding a constant term to the superpotential, since such a term is
forbidden by the U (1)R gauge symmetry. Therefore, it is a non-trivial problem whether we
can obtain the vanishing cosmological constant in our model.
Assuming that the potential minimum lies on the D-flat direction of the SU(2)H gauge
interaction, we take the vacuum expectation values such that hSi = s and hQαi i = vδiα ,
where i and α denote the flavor and SU(2)H indices, respectively. We can always make s
and v real and positive by symmetry transformations. The total potential is given by 
2   

Λ5
K
2
2
2
2
V (v, s) = e (λv + sW ) + 2v λs − 4 + W − 3W
v
gR2
(4s 2 − 2v 2 + 2)2 ,
(4)
2
where K and W are the Kähler potential and superpotential, respectively, which are given
by
+

(5)
K = s 2 + 2v 2 ,
5
Λ
(6)
W = λsv 2 + 2 .
v
Here, all dimensionful parameters are taken to be dimensionless with the normalization
MP = 1. The first line in Eq. (4) comes from the F-term (except for W 2 term) and the
remainder is the D-term potential.
Since the potential is very complicated, it is convenient to make some assumptions
for the values of parameters. First, assume that gR  λ, Λ. Since the D-term potential
is proportional to gR2 and positive definite, the potential minimum is expected for VD to be
small as possible. If we assume s  1 and v ∼ 1, the potential can be rewritten as 

(7)
V ∼ e2 λ2 − 3Λ10 .
√ 5
It is found that λ ∼ 3Λ is required in order to obtain the vanishing cosmological
constant.
Let us consider the stationary conditions of the potential. Using the assumptions s  1
and v = 1 + y (|y|  1), the stationary conditions can be expanded with respect to s and
y. Considering the relations gR  λ ∼ Λ5 , we can expand the condition ∂V /∂y = 0 and
obtain
y ∼ s2 −

e2 λ2
.
2gR2

(8)

Using this result, we can also expand the expansion of the condition ∂V /∂s = 0 and obtain
s∼

λΛ5
.
− Λ10

8λ2

(9)

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

265

By the numerical analysis, the above rough estimation is found to be a good approximation.
The result of numerical calculations is the following:
y ∼ 4.7 × 10−3 ,
s ∼ 6.8 × 10

−2

(10)

.

(11)

Here, we used the values of Λ = 10−3 , λ ∼ 1.8Λ5 and gR = 10−11 . For these values
of the parameters, we can obtain the vanishing cosmological constant. Note that the
numerical values of Eqs. (10) and (11) are almost independent of the actual value of Λ,
if the condition gR  Λ5 is satisfied and the ratio λ/Λ5 is fixed. This can be seen in the
approximate formulae of Eqs. (8) and (9). We can choose the value of Λ in order to obtain
a phenomenologically acceptable mass spectrum.
As a result of the above analysis, we can estimate the gravitino mass as

Λ5
m3/2 = eK/2 W ∼ 3.0 × 4 .
MP

(12)

This non-zero gravitino mass means that supersymmetry is really broken in the framework
of the supergravity. The gravitino mass contributes to the masses of scalar partners via
the tree-level interactions of supergravity. Note that there is another contribution, if scalar
partners have non-zero U (1)R charges. In this case, they also acquire the mass from the
vacuum expectation value of the D-term, and it is estimated as
m2D-term

= gR2 hDiq  

Λ5 2
∼ 7.3 × 4 q,
MP

(13)

where q is the U (1)R charge. This mass squared is always positive for the scalar fields
with positive U (1)R charges. The mass is the same order of the magnitude of the gravitino
mass. This is because gR is canceled out in the above estimation (see Eq. (8)). Note that
m3/2 and m2D-term is controlled by the strong coupling scale of SU(2)H gauge theory Λ.
We give some comments. Our model has the same structure of the supersymmetry
breaking model with the anomalous U (1) gauge symmetry [10]. In the model, the Fayet–
Iliopoulos term is originated from the anomaly of the U (1) gauge symmetry [19–21]. On
the other hand, in our model the origin of the term is the symmetry of supergravity with
the gauged U (1)R symmetry. The Fayet–Iliopoulos term appears even if the U (1)R gauge
interaction is anomaly free.
The mediation of supersymmetry breaking to the visible sector is discussed in
succeeding sections. We address the highly non-trivial problem whether the anomaly
cancellation of U (1)R can be done when we include the visible sector superfields with
positive semi-definite U (1)R charges [13,14]. We discuss two explicit models. One is the
model that the U (1)R anomalies are cancelled out by the Green–Schwarz mechanism. We
refer this model to the anomalous U (1)R model. The other is the model that the U (1)R
symmetry is anomaly free. We refer this model to the anomaly free U (1)R model.

266

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

3. Anomalous U (1)R model
Before discussing the model in detail, it is instructive to review the Green–Schwarz
mechanism in four dimensions [22]. Suppose that we have a gauge symmetry U (1)X
with mixed gauge anomalies of U (1)X G2i , where Gi denote other gauge groups. The
Lagrangian is not invariant under the U (1)X gauge transformation Aµ → Aµ + ∂µ α(x):
δLgauge = α(x) 

Ci
ei ,
tr Fi F
2

(14)

ei ) are the field strength tensors (its dual) of Gi .
where Ci are anomaly coefficients and Fi (F
On the other hand, there could be another anomaly at the Planck scale through the dilaton
superfield S with the U (1)X transformation S → S + 2i δGS α(x): 

δGS 1
ei ,
ki tr Fi F
(15)
2
2
is the Green–Schwarz coefficient which is calculated in the string theory [19]

δLGS = −α(x)
where δGS
as

1
tr qi .
(16)
192π 2
Here, qi are R-charges of the fermionic components of the superfields. ki are the Kac–
Moody levels of Gi . Note that the Kac–Moody levels are integers for non-Abelian gauge
groups but are not necessary integers for Abelian groups. Therefore, we can assume the
value of the Kac–Moody levels for Abelian groups appropriately. All the Kac–Moody
levels must have the same sign, since all the gauge couplings are generated by
δGS =

1
= ki hSi,
gi2

(17)

where hSi is the vacuum expectation value of the dilaton superfield. From Eqs. (14) and
(15), the anomaly cancellation conditions lead to the Green–Schwarz relations:
Ci
= 2π 2 δGS .
ki

(18)

This relation should be satisfied for any i. It follows from this relation that all of Ci has the
same sign, since all of ki has the same sign.
Now, we discuss the model in detail. We consider the MSSM with all the Yukawa
couplings but the µ-term. The relevant anomaly coefficients are listed below:
1
CH = (q1 + q2 ) + 2,
2
3
C3 = (2q + u + d) + 3,
2 

1
3
C2 = (3q + l) + h + h¯ + 2,
2
2  

4
1
1
1
1
CY = 3 q + u + d + l + e + h + h¯ ,
6
3
3
2
2

(19)
(20)
(21)
(22)

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274 

CY Y = 3 q 2 − 2u2 + d 2 − l 2 + e2 + h2 − h¯ 2 = 0, 

Cg = 2(q1 + q2 ) + s + 3(6q + 3u + 3d + 2l + n + e) + 2 h + h¯ − 6, 

CR = 2(q13 + q23 ) + s 3 + 3 6q 3 + 3u3 + 3d 3 + 2l 3 + n3 + e3 

+ 2 h3 + h¯ 3 + 18,

267

(23)
(24)
(25)

where q1 , q2 and s are R-charges of the fermionic components for Q1 , Q2 and S in the
hidden sector, and q, u, d, l, n, e, h, and h¯ are those for the chiral superfields Q(3, 2, 16 ),
SS
S(1, 2, − 1 ),
S(S
3, 1, 13 ), L(1, 2, − 12 ), N(1, 1, 0) E(1, 1, 1), H (1, 2, 12 ), and H
U
3, 1, − 23 ), D(
2
respectively. The representations and charges in the parenthesis are those under SU(3)C ×
SU(2)L ×U (1)Y . The coefficients CH , C3 , C2 , CY , CY Y , Cg and CR represent the anomaly
coefficients for U (1)R (SU(2)H )2 , U (1)R (SU(3)C )2 , U (1)R (SU(2)L )2 , U (1)R (U (1)Y )2 ,
(U (1)R )2 U (1)Y , U (1)R and (U (1)R )3 , respectively. Here we simply assume a family
independent charge assignment. Note that the gravitino contribution for the mixed
gravitational anomaly is −21 times that of a gaugino, while the gravitino contribution
for the U (1)R gauge anomaly is three times that of a gaugino [23–26], and the dilatino
contributes (−1) to both Cg and CR . Note also that the anomaly coefficient CY Y has to
vanish identically, since it cannot be cancelled by the Green–Schwarz mechanism.
As mentioned earlier, all the Yukawa coupling are included:
SH + yd QD
SH
S + ye LE H
S + yn LNH.
W = yu QU

(26)

The resulting conditions for R-charges are
q + u + h = −1,
q + d + h¯ = −1,

(27)

l + e + h¯ = −1,

(29)

l + n + h = −1.

(30)

(28)

Furthermore, we need the Yukawa coupling in the hidden sector
W = λS[Q1 Q2 ],

(31)

which leads to the condition
s + q1 + q2 = −1.

(32)

One can immediately see that C3 = C2 = CY = 0 and Eqs. (27), (28) and (29) are
incompatible, since CY + C2 − 2C3 = −6. Therefore we have to use the Green–Schwarz
mechanism to cancel the anomalies.
Taking into account the simple assumption k2 = k3 , we have only two solutions. One
solution consists of all integer charges:
S = 6,
Q1 = Q2 = −2,
S
S
Q = U = D = L = N = E = 0,

(33)
S = 2,
H =H

(34)

where these are the charges of superfields. The corresponding anomaly coefficients are

268

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

C3 = C2 = −3,
CH = −1,

CY = −9,
CR = −9,

CY Y = 0,
Cg = −57.

(35)
(36)

Note that all the non-trivial anomaly coefficients have negative signs. Therefore, we assume
that all the Kac–Moody levels are positive with negative vacuum expectation value of the
dilaton superfield. We can freely choose the positive values of kR and kg to satisfy the
Green–Schwarz relation [22]. On the other Kac–Moody levels the following discussions
are required.
The Green–Schwarz relation
2π 2 Cg
Cg
C2 C3 CH
CY
=
=
=
= 2π 2 δGS =
=
kY
k2
k3
kH
96
192π 2

(37)

is satisfied, if we introduce 39 gauge singlets of vanishing R-charge with kY = 9, k2 = k3 =
3 and kH = 1. These values of Kac–Moody levels tell us the gauge coupling relations at
the Planck scale, namely,
α3 = α2 = 3αY ,

αH = 3α3 .

(38)

This coupling unification can be easily accomplished by introducing extra massive particles
which change the running of the gauge coupling constants and does not affect the anomaly
cancellation conditions. In fact, the gauge coupling unification is realized, if we introduce
extra massive particles so that the SU(3)C gauge coupling almost does not run and the
SU(2)L gauge coupling is asymptotically non-free.
Let us perform the potential analysis. We assume the SU(2)H D-flat condition. The
scalar potential in the present case is the same one in our previous letter except for the
D-term part 5 
2   

Λ5
K
2
2
2
2
V (v, s) = e (λv + sW ) + 2v λs − 4 + W − 3W
v
gR2
(6s 2 − 4v 2 + 2)2 ,
(39)
2
where K and W are the Kähler potential and the superpotential, respectively, which are
given by Eqs. (5) and (6). We assume λ, Λ  gR , v = √1 + y with y  1 and s  1, and
2
expand the potential with respect to y and s. The approximate formulae of the stationary
conditions ∂V /∂y = 0 and ∂V /∂s = 0 are
+

e(3λ2 − 16Λ10)
3
,
y ∼ √ s2 −

2 2
32 2gR2

(40)

10λΛ5
.
(41)
9λ2 − 32Λ10
These formulae are correct within the order of magnitude, since the convergence of the
expansion is not so good. The numerical calculation gives
s∼

5 Although the Fayet–Iliopoulos term due to the anomaly of U (1) should be included, but its magnitude is
R
suppressed compared to that of the gauged U (1)R symmetry. Hence, we simply neglect it.

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

1
v ∼ √ + 0.019,
2
s ∼ 0.13,

269

(42)
(43)

where we used the parameters, Λ = 10−3 , λ = 7.3Λ5 and gR2 = 10−11 . We have checked
that the cosmological constant can be fine-tuned to zero by adjusting λ for fixed Λ.
Next, we estimate the spectrum of supersymmetry breaking masses. The gravitino mass
is given by

Λ5
m3/2 = eK/2 W ∼ 4.1 × 4 .
MP

(44)

This gravitino mass contributes to the masses of scalar partners via the tree-level interaction
S, D,
S L, N, E vanish,
of supergravity. Since R-charges for Q, U 

S, D,
S L, N, E .
(45)
m2i ∼ m23/2 i = Q, U
S have R-charge 2,
On the other hand, since H and H
m2H,HS ∼ m23/2 + 2gR2 hDi,  

Λ5 2
2
∼ O(m23/2 ).
∼ m3/2 + 2 8.2 × 4
MP

(46)

The contribution from hDi 6= 0 is the same order of the gravitino mass.
The gaugino mass can be generated in two different ways. In the case that the gauge
kinetic function is trivial, the anomaly mediation contribution dominates as
mλ i =

β(gi2 )
2gi2

m3/2 ,

(47)

where β(gi2 ) is the beta function for the standard model gauge groups. 6 In the case that the
gauge kinetic function is non-trivial and includes the term S 2 [Q1 Q2 ]3 /MP8 , for example,
the gaugino masses are
1
m3/2.
(48)
23
In both cases, there could be further contribution from the vacuum expectation value of Fcomponent of the dilaton [28]. The experimental bound on gaugino masses in the MSSM
[29] determines the order of the gravitino mass as 10 TeV in both cases.
Here, we comment on the µ-term. We can have the higher-dimensional interaction
mλ i ∼

W =κ

S[Q1 Q2 ]2 S
HH
MP4

(49)

which respects all symmetries, where κ is a dimensionless constant. Plugging the vacuum
expectation value of S, Q1 and Q2 into Eq. (49), we obtain µ ∼ κhSi. By adjusting κ  1,
the electroweak scale is generated.
6 More precisely, there is a correction of the same order of magnitude, see Ref. [27].

270

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

4. Anomaly free U (1)R model
In the last section, we discussed the scenario that supersymmetry breaking effect is
mediated to the MSSM as the visible sector. The Green–Schwarz mechanism is used to
cancel gauge anomalies of U (1)R and the simple solution was found. However, the Green–
Schwarz mechanism requires the dilaton field, and this leads to new difficult problems
such as the dilaton stabilization and so on [28,30]. Therefore, it is desirable to consider the
case that U (1)R is anomaly free. As mentioned earlier, it is impossible to cancel all the
gauge anomaly for U (1)R in the MSSM. In this section, we consider the supersymmetric
SU(5) GUT instead of the MSSM as the visible sector. We have found that all the gauge
anomalies can be cancelled by introducing SU(5) × SU(2)H gauge singlets with non-trivial
R-charges.
Suppose that we have N of SU(5) × SU(2)H gauge singlets with R-charge 2. The
anomaly cancellation conditions are   

3
1
1
(50)
U (1)R [SU(5)]2 : 3 f¯ + a + h + h¯ + 5σ + 5 = 0,
2
2
2 

[U (1)R ]: 2(q1 + q2 ) + s + 3 5f¯ + 10a + n 

+ 5 h + h¯ + 24σ + N + 7 = 0,
(51)   

3
3
3
3
3
3
3
3
3
¯
¯
[U (1)R ] : 2 q1 + q2 + s + 3 5f + 10a + n + 5 h + h
+ 24σ 3 + N + 31 = 0,

(52)

where f¯, a, n, σ, h and h¯ are the R-charges of the fermionic component for the superfields
S(S
S(S
F
5), A(10), N(1), Σ(24), H (5) and H
5), respectively. The representations in the
parenthesis are those of SU(5). The superpotential of Eq. (31), and Yukawa couplings
SH
S + NF
SH
W ∼ AAH + AF

(53)

give the conditions of Eq. (32) and
2a + h = −1,
¯
f + a + h¯ = −1
n + f¯ + h = −1.

(54)
(55)
(56)

We have a simple solution, in the case of N = 36 as
S = A = N = 0,
F
Q1 = 0,

S = 2,
H =H

Q2 = −2,

S = 4,

Σ = 1,
(57)

where these are the charges of superfields. Note that Q1 and Q2 have different charges now.
This modification makes possible to have the special charge assignment with all positive
charges except for the fields which couple to SU(2)H gauge bosons.
The potential analysis is the same in Ref. [12] as long as we assume that the potential is
along the SU(2)H D-flat direction. The gravitino mass is estimated as

Λ5
m3/2 = eK/2 W ∼ 4 .
MP

(58)

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

271

The gravitino mass contributes to the masses of scalar partners via the tree-level
interactions of supergravity. Note that there is another contribution, if scalar partners
have non-zero U (1)R charges. In this case, they also acquire the mass from the vacuum
expectation value of the D-term, and it is estimated as 
5 2
Λ
q,
(59)
m2D-term = qgR2 hDi ∼
MP4
where q is the U (1)R charge. This mass squared is always positive for the scalar partners
with positive U (1)R charges. The mass is the same order of the magnitude of the gravitino
mass.
The gaugino mass can be generated in two different ways. In the case that the gauge
kinetic function is trivial, the anomaly mediation contribution dominates as
mλ i =

β(gi2 )
2gi2

m3/2 .

(60)

On the other hand, in the case that the gauge kinetic function is non-trivial and includes the
higher-dimensional term 7 S([Q1 Q2 ])2 /MP5 , for example, the gaugino masses are
mλi ∼ m3/2 .

(61)

Considering the experimental bound on gaugino masses in the MSSM [29], the gravitino
mass is taken to be of the order of 10 TeV or 1 TeV in the former case or the latter case,
respectively. This phenomenological constraint requires the dynamical scale of the SU(2)H
gauge interaction to be of the order of 1015 GeV for both cases. This also means that λ is
extremely small, λ ∼ 10−15 . Note that the requirement of this fine-tuned value of λ is the
result from the fine-tuning for vanishing cosmological constant. Furthermore, this small
Yukawa coupling is consistent with the above discussion in the following sense. Since
S has the vacuum expectation value, the mass for Qi is generated through the Yukawa
coupling in Eq. (2). The relation λhSi  Λ is needed not to change our result from the
SU(2)H gauge dynamics.
Here, we comment on the Higgs potential. We would like to discuss whether it is possible
to construct the higher-dimensional operators that realize the gauge symmetry breaking as
SU(5) → SU(3)C × SU(2)L × U (1)Y → SU(3)C × U (1)em and also realize the doublet–
triplet splitting. We can explicitly write down the following interactions respecting all
symmetries. 8
W∼

λ1
SH + λ2 [Q1 Q2 ]2 H
SΣ 2 H
[Q1 Q2 ]H
MP
MP5
λ4
λ3
+ 4 S[Q1 Q2 ]2 tr(Σ 2 ) + 3 [Q1 Q2 ] tr(Σ 4 ),
MP
MP

(62)

7 The higher-dimensional term S([Q Q ])2 /M 5 in the gauge kinetic function can be forbidden to all orders
1 2
P
by the discrete symmetry.
8 More precisely, there are other terms of contracting the gauge indices in different ways, but we simply write
the representative.

272

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

where λi (i = 1 ∼ 4) are constants. By choosing λi appropriately, the first two terms can
realize the doublet–triplet splitting, while the last two terms can realize the gauge symmetry
breaking.
Unfortunately, there exists unwanted operators, for examples,
1
MP3

S
[Q1 Q2 ] H F
1
MP5
1

MP10 

2

→ 

1
S 2,
HF
MP

[Q1 Q2 ]3 S 2 → MP S 2 ,
[Q1 Q2 ]5 S 3 → S 3 .

(63)

Although the first term seems to be harmless, the other terms are dangerous since
supersymmetry is restored if these operators are present. However, this problem is not
specific to our model, but inevitable and generic in supergravity models.

5. Summary and discussion
In this paper, we have shown that it is possible to construct simple and phenomenologically viable models of dynamical supersymmetry breaking with gauged U (1)R symmetry. Supersymmetry breaking occurs by the interplay between the dynamically generated
superpotential and the Fayet–Iliopoulos term which appears due to the symmetry of supergravity. The cosmological constant can be fine-tuned to vanish at the minimum of the
potential. What is the most non-trivial in this class of models is the anomaly cancellations for U (1)R . We have presented two explicit models with anomaly cancellations as the
visible sector.
One is the MSSM with all the Yukawa couplings and without the µ-term. Anomalies
are cancelled by the Green–Schwarz mechanism. We have found a quite simple solution
of R-charge assignments for matter superfields in our model compared to the models
in Refs. [13,14]. Our solution consists of all integer charges. We have discussed the
gauge coupling unification which follows from the Green–Schwarz relations. The gauge
coupling unification is easily accomplished by introducing extra massive particles. The
spectrum of supersymmetry breaking masses has been estimated and turned out to be
phenomenologically viable as follows. The gravitino mass is of order 10 TeV. The scalar
masses are the same order of the gravitino mass. The gaugino masses are a few orders
smaller than the gravitino mass. We have also shown that it is possible to have the µterm. Unfortunately, we have to introduce the dilaton superfield to cancel anomalies in this
model. This leads to new difficult problems such as the dilaton stabilization and so on [28,
30].
The other is the supersymmetric SU(5) GUT. In this model, U (1)R anomalies are
cancelled by introducing the SU(5) gauge singlet superfields. We have also found a quite
simple solution of R-charge assignment. Unlike the anomalous U (1)R model in this paper
and the anomalous U (1) model in Refs. [10,11], the dilaton is not necessary since the

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

273

gauged U (1)R is anomaly free. Therefore, there is no dilaton stabilization problem. We
have estimated the mass spectrum of supersymmetry breaking and found that these models
are phenomenologically viable. The gravitino mass is of the order 1 TeV or 10 TeV
depending on the form of the gauge kinetic function. The scalar masses are the same order
of the gravitino mass. The gaugino masses are a few orders smaller than the gravitino mass
in the case that the gauge kinetic function is trivial, while the same order of it in the case
that the gauge kinetic function have higher-dimensional terms. We have shown that it is
possible to have the Higgs potential for the doublet–triplet splitting and the GUT gauge
symmetry breaking.
It is interesting to apply our mechanism to the phenomenology of the anomaly mediation
scenarios. In this scenarios, if the visible sector consists of the MSSM, the slepton mass
squared becomes negative because the beta function coefficients of SU(2)L × U (1)Y are
negative. This problem is easily avoided 9 by using our mechanism since the contribution
to the mass squared from the vacuum expectation value of U (1)R D-term is always positive
for the positively R-charged field. Unfortunately, this situation is not realized in our models
since R-charges of the matter fields are zero. However, it would be possible to construct
the models in which lepton superfields have positive charges.
Finally, note that our model is effectively reduced to the well known Polonyi model
which is the simplest supersymmetry breaking model at the tree-level. One can understand
this fact by freezing the Qi with its vacuum expectation values in the superpotential of Eq.
(2). Here, we would like to emphasize that the Polonyi model is derived as the effective
theory of our model. The Polonyi model does not specify the dynamics of supersymmetry
breaking. Also, unlike the Polonyi model, there is no dimensionful parameters other than
the Planck scale in our model. They are induced dynamically from the Planck scale.

Acknowledgements
This work was supported in part by the Grant-in-aid for Science and Culture Research
form the Ministry of Education, Science and Culture of Japan (#11740156, #3400, #2997).
N.M. and N.O. are supported by the Japan Society for the Promotion of Science for Young
Scientists.
References
[1]
[2]
[3]
[4]
[5]
[6]
[7]

P. Fayet, J. Iliopoulos, Phys. Lett. B 51 (1974) 461.
L. O’Raifeartaig, Nucl. Phys. B 96 (1975) 331.
E. Witten, Nucl. Phys. B 188 (1981) 513.
E. Poppitz, S. Trivedi, Ann. Rev. Nucl. Part. Sci. 48 (1998) 307.
Y. Shadmi, Y. Shirman, Rev. Mod. Phys. 72 (2000) 25, and references therein.
H.P. Nilles, Phys. Rep. 110 (1984) 1.
G.F. Giudice, R. Rattazzi, Phys. Rep. 322 (1999) 419, and references therein.

9 For related works, see [31–33].

274

[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]

N. Kitazawa et al. / Nuclear Physics B 586 (2000) 261–274

L. Randall, S. Sundrum, Nucl. Phys. B 557 (1999) 79.
G.F. Giudice, M. Luty, H. Murayama, R. Rattazzi, JHEP 9812 (1998) 027.
G. Dvali, A. Pomarol, Phys. Rev. Lett. 77 (1996) 3728.
P. Binétruy, E. Dudas, Phys. Lett. B 389 (1996) 503.
N. Kitazawa, N. Maru, N. Okada, hep-ph/9911251.
A.H. Chamseddine, H. Dreiner, Nucl. Phys. B 458 (1996) 65.
D.J. Castano, D.Z. Freedman, C. Manuel, Nucl. Phys. B 461 (1996) 50.
I. Affleck, M. Dine, N. Seiberg, Nucl. Phys. B 241 (1984) 493; Phys. Lett. B 137 (1984) 187.
S. Ferrara, L. Girardello, T. Kugo, A. van Proeyen, Nucl. Phys. B 223 (1983) 191.
T. Kugo, S. Uehara, Nucl. Phys. B 226 (1983) 49.
E. Cremmer, S. Ferrara, L. Girardello, A. van Proeyen, Nucl. Phys. B 212 (1983) 413.
M. Dine, N. Seiberg, E. Witten, Nucl. Phys. B 289 (1987) 589.
J. Atick, L. Dixon, A. Sen, Nucl. Phys. B 292 (1987) 109.
M. Dine, I. Ichinose, N. Seiberg, Nucl. Phys. B 293 (1988) 253.
M. Green, J. Schwarz, Phys. Lett. B 149 (1984) 117.
N.K. Nielsen, M.T. Grisaru, H. Römer, P. van Nieuwenheuizen, Nucl. Phys. B 140 (1978) 477.
N.K. Nielsen, H. Römer, Phys. Lett. B 154 (1985) 141.
L. Alvarez-Gaume, E. Witten, Nucl. Phys. B 234 (1983) 269.
J.P. Derendinger, Phys. Lett. B 151 (1985) 203.
J. Bagger, T. Moroi, E. Poppitz, hep-th/9911029.
N. Arkani-Hamed, M. Dine, S.P. Martin, Phys. Lett. B 431 (1998) 3219.
Particle Data Group, Eur. Phys. J. C 3 (1998) 1.
T. Barreiro, B. de Carlos, J.A. Casas, J.M. Moreno, Phys. Lett. B 445 (1998) 82.
A. Pomarol, R. Rattazzi, JHEP 9905 (1999) 013.
Z. Chacko, M.A. Luty, I. Maksymyk, E. Ponton, hep-ph/9905390.
E. Katz, Y. Shadmi, Y. Shirman, JHEP 9908 (1999) 015.

Nuclear Physics B 586 (2000) 275–286
www.elsevier.nl/locate/npe

Consistent SO(6) reduction of type IIB supergravity
on S 5
M. Cvetiˇc a,1 , H. Lü a,1,∗ , C.N. Pope b,2 , A. Sadrzadeh b , T.A. Tran b
a Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, PA 19104, USA
b Center for Theoretical Physics, Texas A&M University, College Station, TX 77843, USA

Received 2 May 2000; accepted 14 June 2000

Abstract
Type IIB supergravity can be consistently truncated to the metric and the self-dual 5-form. We
obtain the complete non-linear Kaluza–Klein S 5 reduction ansatz for this theory, giving rise to
gravity coupled to the fifteen Yang–Mills gauge fields of SO(6) and the twenty scalars of the coset
SL(6, R)/SO(6). This provides a consistent embedding of this subsector of N = 8, D = 5 gauged
supergravity in type IIB in D = 10. We demonstrate that the self-duality of the 5-form plays a
crucial role in the consistency of the reduction. We also discuss certain necessary conditions for
a theory of gravity and an antisymmetric tensor in an arbitrary dimension D to admit a consistent
sphere reduction, keeping all the massless fields. We find that it is only possible for D = 11, with
a 4-form field, and D = 10, with a 5-form. Furthermore, in D = 11 the full bosonic structure of
eleven-dimensional supergravity is required, while in D = 10 the 5-form must be self-dual. It is
remarkable that just from the consistency requirement alone one would discover D = 11 and type
IIB supergravities, and that D = 11 is an upper bound on the dimension.  2000 Elsevier Science
B.V. All rights reserved.

1. Introduction
Non-trivial Kaluza–Klein sphere reductions of supergravity theories have been studied
in a number of contexts. Long ago it was demonstrated [1] that the linearised analysis
of the zero-mode fluctuations of the S 7 reduction of D = 11 supergravity [2,3] can be
extended to a fully non-linear and consistent embedding of maximal D = 4 gauged SO(8)
supergravity in D = 11. The fact that the truncation to the zero-mode sector is consistent,
despite the non-linearities of the theory, is rather non-trivial since there appears to be no
group-theoretic understanding of why it should work.
∗ Corresponding author. E-mail: honglu@dept.physics.upenn.edu
1 Research supported in part by DOE grant DOE-FG02-95ER40893.
2 Research supported in part by DOE grant DOE-FG03-95ER40917.

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

276

M. Cvetiˇc et al. / Nuclear Physics B 586 (2000) 275–286

More recently, a similar, although more explicit, demonstration of the consistency of the
S 4 reduction from D = 11 to the maximal D = 7 gauged SO(5) supergravity was given
[4,5]. The consistent reduction of massive type IIA supergravity on a locally S 4 space,
to give the N = 2 gauged SU(2) supergravity in D = 6, has also been obtained [6]. In
addition, certain reductions to truncations of the maximal gauged supergravities in various
dimensions have also been constructed. These have the advantage of being considerably
simpler than the maximal theories, allowing the reduction ansatz to be presented in a
more explicit form. Cases that have been constructed include the N = 2 gauged SU(2)
supergravity in D = 7 [7]; the N = 4 gauged SU(2) × U (1) supergravity in D = 5 [8]; the
N = 4 gauged SO(4) supergravity in D = 4 [9]; and maximal abelian truncations in D =
4, 5 and 7 [10]. One can also consider non-supersymmetric truncations of the maximal
gauged supergravities. In [11,12] the consistent truncations of the D = 7, D = 5 and
D = 4 supergravities to the graviton plus scalar subsectors comprising only the diagonal
scalars in the SL(N, R)/SO(N) scalar submanifolds were considered (with N = 5, 6, and
8, respectively), and the consistent embeddings in D = 11 and D = 10 were constructed.
Although the full consistency of the S 7 and S 4 reductions of D = 11 supergravity
has essentially been demonstrated, no similar complete result exists for the S 5 reduction
of type IIB supergravity. The field content of the full N = 8 gauged supergravity [13]
consists of gravity; fifteen SO(6) gauge fields; twelve 2-form gauge potentials in the 6
and 6 representations of SO(6); 42 scalars in the 1 + 1 + 200 + 10 + 10 representations
of SO(6), and the fermionic superpartners. It is believed that this can arise from an S 5
reduction of type IIB supergravity; at the linearised level, the reduction ansatz was given in
[14]. However, at the full non-linear level, the only complete demonstrations so far are for
the consistent embedding of the maximal abelian U (1)3 truncation [10], the N = 4 gauged
SU(2) × U (1) truncation [8], and the scalar truncation in [11,12]. The full metric ansatz
was conjectured in [15].
In this paper, we obtain the consistent reduction ansatz for a different truncation of the
maximal gauged D = 5 supergravity. One can consistently set the 1 + 1 + 10 + 10 of scalars
to zero, at the same time as setting the 6 + 6 of 2-form potentials to zero. The bosons that
remain, namely gravity, the 15 Yang–Mills fields, and the 200 of scalars, come just from
the metric plus the self-dual 5-form sector of the original type IIB theory. We shall obtain
complete results for the consistent embedding of this subsector of the gauged N = 8 theory,
ij
with all fifteen of the SO(6) gauge fields A(1) , and the twenty scalars that parameterise the
full SL(6, R)/SO(6) submanifold of the complete scalar coset. These can be parameterised
by a unimodular symmetric tensor Tij .
Another way of expressing the truncation that we shall consider in this paper is as
follows. The type IIB theory itself can be consistently truncated in D = 10 so that just
gravity and the self-dual 5-form remain. The fifteen Yang–Mills gauge fields and twenty
scalars that we retain in our ansatz are the full set of massless fields associated with Kaluza–
Klein reduction from this ten-dimensional starting point. (The counting of massless fields
is the same as the one that arises from a toroidal reduction from the same ten-dimensional
starting point.) We shall see below that the self-duality condition on the 5-form plays an
essential rôle in the consistency of the S 5 reduction.

the analogous truncations cannot be performed in the maximal gauged supergravities in D = 7 and D = 4. with the FFA term. R)/SO(n+1). / Nuclear Physics B 586 (2000) 275–286 277 This subsector of type IIB supergravity is particularly relevant for the AdS/CFT correspondence [16–18]. Furthermore. The fifteen SO(6) Yang–Mills gauge fields will be represented ij by the 1-form potentials A(1) . R)/SO(n + 1) for generic (D. n).M. while for D = 11 reduced on S 7 one must instead additionally retain 35 pseudoscalars as well as the 35 scalars Tij . 2  b(5) = 1 i1 ···i6 g −4 U ∆−2 Dµi1 ∧ · · · ∧ Dµi5 µi6 ∗ˆ G 5! − 5g −4 ∆−2 Dµi1 ∧ · · · ∧ Dµi4 ∧ DTi5 j Ti6 k µj µk  i1 i2 − 10g −3∆−1 F(2) ∧ Dµi3 ∧ Dµi4 ∧ Dµi5 Ti6 j µj . n). are represented by the symmetric unimodular tensor Tij . while in D = 10 the 5-form must be self-dual. In terms of these quantities. The twenty scalars. d sˆ10 b(5) + ∗ˆ G b(5) . By contrast. We show that in a D-dimensional theory of gravity coupled to an n-form field strength. (1) (2) (3) (4) 3 This is related to the fact that one can consistently truncate five-dimensional maximal gauged supergravity to the SO(6) gauge fields plus scalars of the coset SL(6. Cvetiˇc et al.) Of the three cases where such a consistent sphere reduction is possible. (Such a coset structure is absent for any other values of (D. b(5) = G H  b(5) = −gU (5) + g −1 T −1 ∗DTj k ∧ (µk Dµi ) G ij 1 ij k ` − g −2 Tik−1 Tj−1 ` ∗F(2) ∧ Dµ ∧ Dµ . or in D = 10 with n = 5. 3 2. . We also address the more general question of the circumstances under which a theory admits a consistent sphere reduction. the ten-dimensional one is singled out as the only case where the consistent reduction includes only gravity plus ij the SO(n + 1) gauge fields A(1) and the scalars Tij . The inverse of the scalar matrix Tij is denoted by Tij−1 . The SO(6) reduction ansatz on S 5 We parameterise the fields for this truncated theory as follows. which are in the 200 representation of SO(6). and so it would not be appropriate to look for a consistent reduction ansatz with scalars Tij of SL(n + 1. R)/SO(6). antisymmetric in i and j . we find that the Kaluza–Klein reduction ansatz is given by 2 = ∆1/2 ds52 + g −2 ∆−1/2 Tij−1 Dµi Dµj . in order to achieve consistent reductions. a consistent S n reduction that retains the full set of SO(n + 1) Yang–Mills fields together with coupled massless scalars is possible only when D = 11 and n = 4 or 7. in D = 11 the theory has to be the bosonic sector of eleven-dimensional supergravity. for D = 11 reduced on S 4 one must additionally retain a set of five 2-form potentials. where i is a 6 of SO(6). because it is the metric and the self-dual 5-form that couple to the D3-brane. By contrast. In all three cases the full set of massless scalars includes a subset Tij described by the coset SL(n+1.

The terms involving structures of the form X(4) ∧ Dµi ∧ Dµj .  mj −1 `k ∗F(2) ∧ F(2) D Tik−1 ∗DTkj = −2g 2 (2Tik Tj k − Tij Tkk )(5) + Tik−1 T`m  1  mp  −1 −1 `k − δij − 2g 2 2Tk` Tk` − (Tkk )2 (5) + Tpk T`m ∗F(2) ∧ F(2) . b(5) = 0. 5 (8) where the potential V is given by  1 (9) V = g 2 2Tij Tij − (Tii )2 . µi µi = 1. / Nuclear Physics B 586 (2000) 275–286 where U ≡ 2Tij Tj k µi µk − ∆Tii . subject to the constraint µi µi = 1. (5) b(5) is derivable and (5) is the volume form on the five-dimensional spacetime. It is consistent to truncate the fields of type IIB supergravity to the metric and self-dual b(5) .278 M. jk F(2) = dA(1) + gAik (1) ∧ A(1) . dH (6) The ansatz presented above satisfies these equations of motion if and only if the fivedimensional fields satisfy the equations  1 k1 k2 k3 k4 −1 k` D Tik−1 Tj−1 ` ∗ F(2) = −2gTk[i ∗DTj ]k − 8 ij k1 ···k4 F(2) ∧ F(2) . Note that ∗ˆ G from the given expressions (1) and (3). ij ij kj ∆ ≡ Tij µi µj . The Lagrangian is in agreement with the one for five-dimensional gauged SO(6) supergravity in [13]. where . bMN = 1 H R 96 b(5) = 0. These equations of motion can all be derived from the five-dimensional Lagrangian 1 1 ij −1 k` DT`i − Tik−1 Tj−1 L5 = R∗1 − Tij−1 ∗DTj k ∧ Tk` ` ∗F(2) ∧ F(2) − V ∗1 4  4 1 i1 i2 i3 i4 i5 i6 i1 i2 i3 i4 i5 j j i6 F(2) A(1) − gF(2) A(1) A(1) A(1) − i1 ···i6 F(2) 48  2 2 i1 i2 i3 j j i4 i5 k ki6 + g A(1) A(1) A(1) A(1) A(1) . with the ansatz To see this. (7) 6 together with the five-dimensional Einstein equation. The ten-dimensional equations for motion for these fields are then given by 5-form H bMP QRS H bN P QRS . Cvetiˇc et al. we look first at the ten-dimensional equation d H ij given above. DTij ≡ dTij + gAik (1) Tkj + gA(1) Tik . including the S 4 reduction ansatz from D = 11 that was constructed in [4. In obtaining the above ansatz we have been guided by previous results in the literature. ij Dµi ≡ dµi + gA(1) µj . 2 In (8) we have omitted the wedge symbols in the final topological term. parameterise points in the internal 5-sphere. The coordinates µi . to economise on space.5]. we have presented it here because it is quite an involved computation.

without the use of any five-dimensional from calculating d H equations of motion. . 5 The Kaluza–Klein S 5 reduction that we have obtained here retains the full set of massless fields that can result from the reduction of gravity plus a self-dual 5-form in D = 10. Cvetiˇc et al. the initial truncation of the type IIB theory to its gravity plus self-dual 5-form sector is itself a nonsupersymmetric one. Since µi Dµi = µi dµi = 12 d(µi µi ) = 0. It should also be emphasised that it would be inconsistent to omit the fifteen SO(6) gauge fields when considering the embedding of the twenty scalars Tij . Thus the additional requirement of self-duality seems to be essential for consistency. Thus once again we see that supersymmetry and Kaluza–Klein consistency for sphere reductions seem to go hand in hand. and we read off X(5) − 16 δij X(5) b(5) . all other structures arising Contributions of this structure come only from d G b5 = 0 vanish identically. It is interesting. These would comprise 10 = 4 + 6 vector potentials and 5 = 1 + 4 scalars. where X(5) represents d ∗ˆ G a 5-form in the five-dimensional spacetime. which is in the 15 of SO(6). that self-duality of the 5-form is apparently forced on us by the requirements of Kaluza–Klein consistency. (Of course. therefore.M. However the inclusion of these fields would still not achieve a consistent reduction ansatz. It is worth remarking that it is essential for the consistency of the reduction ansatz that b(5) should be self-dual. This is a quite different situation from a toroidal reduction. In other words. which have a source term gTk[i on the right-hand side. Finally. with (10) now combining with terms from d G 4 form part of the five-dimensional Yang–Mills equations given in (7). the field equation d ∗ˆ G produce the (unacceptable) constraint k1 k2 k3 k4 ∧ F(2) = 0. equation. where it is always consistent to truncate to the scalar sector. The new feature here in the sphere reduction is that the scalar fields are charged under the gauge 4 If one were to consider the reduction of a ten-dimensional theory with a non-self-dual 5-form there would be additional fields present in a complete massless truncation. could still not acquire an interpretation as a source term for the additional fields. there is a trace subtraction in the scalar ij kk = 0 as the five-dimensional equation of motion. [i1 ···i6 Vi7 ] = 0. no further truncation of massless fields has been performed. / Nuclear Physics B 586 (2000) 275–286 279 ij X(4) represents a 4-form in the five-dimensional spacetime. ij k1 ···k4 F(2) (10) b(5) together into the self-dual field H b(5) that a consistent b(5) and ∗ˆ G It is only by combining G b(5) to five-dimensional result is obtained. In deriving these results one needs to make extensive use of the Schoutens over-antisymmetrisation identity. whose ansatz b(5) = 0 would give perfectly b(5) in (3). give rise to the Yang–Mills b(5) and also from the final term in equations above. anti-self-duality would be equally good. after the initial truncation of the type IIB theory in D = 10. give rise to the scalar equations of motion in (7). This can be seen −1 ∗DTj ]k appearing from the Yang–Mills equations in (7). setting the gauge fields to zero.) 5 Of course. since the current (10). One cannot simply consider a reduction ansatz for a the 5-form H ten-dimensional theory consisting of gravity plus a non-self-dual 5-form. Although the Bianchi identity d G is given by G ij ˆ (5) = 0 would acceptable equations of motion for F(2) and Tij . The terms involving structures of the form X(5) ∧ (µi Dµj ). Contributions of this kind come from d G ij ij b(5) . but the crucial point is that the consistency of the Kaluza–Klein S 5 reduction is singling out a starting point that is itself a subsector of a supersymmetric theory. if we try to “invent” a ten-dimensional theory that can be reduced on S 5 .

12]. with appropriate additional truncations. and will be addressed in future work. Note in particular that φ and χ do not appear in the five-dimensional scalar potential V . to  e1 . since then Tk[i Our testing of the consistency of the reduction ansatz (1)–(4) has so far been restricted b(5) . including the ten-dimensional Einstein equations. The second possible truncation involves reducing the scalar sector still further. the consistency seems to be assured once it has been achieved for the equations of motion for the antisymmetric tensor fields. Cvetiˇc et al. 6 In the next section we shall show that the reduction ansatz that we have obtained in this paper reduces. to results that have been obtained previously. and thus in all cases it is inconsistent to include the full set of scalar fields without including the gauge fields as well (see also [4. 3. as in [11. This reduces to the embedding that was obtained in [11]. in order to make contact with previous results in the literature. this provides further supporting evidence for the complete consistency of the ansatz that we have constructed here. χˆ = χ. X e2 . X2 . The first of the three involves truncating the twenty scalars Tij of the SL(6.χ) = − ∗dφ ∧ dφ − e2φ ∗dχ ∧ dχ. setting all the gauge fields to zero.) scalars in Tij .19]). In this reduction they do not appear in the previous ansatz for the metric and self-dual 5-form. X5 . for which the complete proof of consistency was constructed in [12]. A full testing of the consistency of the tento checking the equations of motion for H dimensional Einstein equations would be quite involved. Since the complete consistency was proven in these earlier results. Q ij where i Xi = 1. R)-invariant Lagrangian 1 1 (12) L(φ. 2 2 which is added to (8). (14) Tij = diag X 6 The experience in all previous work on consistent reductions is that although the actual checking of the higherdimensional Einstein equations is the most difficult part from a computational point of view. to D = 4. Truncations to previous results We can consider three different truncations of the reduction scheme of the previous section. Finally in this section.280 M. X4 . R)/SO(6) coset to the diagonal subset (13) Tij = diag(X1 . X6 ). .5. which itself is an extremely stringent requirement. X e3 . This is a general feature of all the sphere reductions. we remark that the S 5 reduction that we have constructed here is also consistent if we include the dilaton φˆ and axion χˆ of the type IIB theory. / Nuclear Physics B 586 (2000) 275–286 group. (One can consistently truncate to the diagonal −1 ∗DTj ]k = 0. and setting all the fifteen gauge fields F(2) to zero. X e3 . (11) where the unhatted quantities denote the fields in five dimensions. These simply reduce according to the ansatz φˆ = φ. X3 . X e2 . and in D = 5 they just give rise to the additional SL(2. X e1 . D = 5 and D = 7.

Thus b(5) gives the same expression as the one obtained for ∗ˆ G ansatz (4) for ∗ˆ G we have verified that the gauged SO(6) embedding that we have obtained in this paper can be truncated to the SU(2) × U (1) embedding of the N = 4 theory whose consistency was . X−2 . (16) and (17) into the b(5) in [8]. b(5) given in (3). 2 1 3 03 12 A(1) = A(1) = √ A(1) . X−2 . (16) We also parameterise the coordinates µi of the internal 5-sphere as follows: µ0 + iµ3 = cos ξ cos 12 θ ei(ψ+φ)/2. We take all the gauge potentials A(1) to be zero except for the following: 1 1 23 A01 (1) = A(1) = √ A(1) . and describes five-dimensional U (1)-gauged simple supergravity coupled to two U (1) vector multiplets. As a consistency check. (16) and (17) into the metric ansatz (1). we obtain 2 2 = ∆1/2 ds52 + g −2 X∆1/2 dξ 2 + g −2 ∆−1/2 X2 s 2 dτ − gB(1) d sˆ10 X √ 2 1 + g −2 ∆−1/2X−1 c2 (18) σ i − 2 gAi(1) . F 34 and F 56 ea = 1. leads to Substituting (15). we can also verify that substituting (15). with ξ parameterising the foliating surfaces. This is the metric ansatz obtained in [8] for the embedding of fivedimensional N = 4 gauged SU(2) × U (1) supergravity in D = 10. but now retaining the three U (1) gauge fields F(2) where a X (2) (2) 3 of the maximal abelian U (1) subgroup of SO(6). (17) Substituting (15). (16) and (17) into the ansatz for G 2 i b(5) = −gU ε5 − 3sc X−1 ∗dX ∧ dξ + √c X−2 ∗F(2) ∧ hj ∧ hk εij k G g 8 2 g2  sc sc i − √ X−2 ∗F(2) ∧ hi ∧ dξ − 2 X4 ∗G(2) ∧ dξ ∧ dτ − gB(1) . 2 A45 (1) = B(1) . The internal 5-sphere now has a geometrical interpretation as a foliation by S 3 × S 1 .M. 2 g 2 2g (19) where U = −2(X2 c2 + X−1 s 2 + X−1 ). and the σi denote the three left-invariant 1-forms of SU(2). again in agreement with the ansatz obtained in [8]. The third possible truncation involves retaining just a single scalar field X. X. at the same time retaining only the gauge fields of SU(2) × U (1). The consistent embedding of this theory in type IIB supergravity was obtained in [10]. 4 i √ where c ≡ cos ξ . by looking at the five-dimensional equations of motion in (7). It is convenient now to ij take the SO(6) indices i to range over 0 6 i 6 5. 2 1 2 31 A02 (1) = A(1) = − √ A(1) . / Nuclear Physics B 586 (2000) 275–286 281 Q 12 . s ≡ sin ξ . ∆ = X−2 s 2 + Xc2 . µ4 + iµ5 = sin ξ eiτ . hi ≡ σ i − 2 gAi(1) . that this is a consistent truncation.) The truncated theory is supersymmetric. (It is easy to see. σ3 = dψ + cos θ dφ. X. Cvetiˇc et al. given by σ1 + iσ2 = e−iψ (dθ + i sin θ dφ). X. µ1 + iµ2 = cos ξ sin 12 θ ei(ψ−φ)/2. by taking  (15) Tij = diag X.

rather than an unconstrained one. therefore. 21] to SL(n + 1. in which all the massless fields (including the SO(n + 1) Yang–Mills fields) are retained. if φE denotes the set of n canonically-normalised dilatons that result from the T n reduction then all the dilaton/axion couplings should be of the form E ai )2 = 4 for each i. . Consistency conditions for sphere reductions One of the interesting outcomes from our analysis is that it is essential for the consistency of the 5-form reduction ansatz that it be a self-dual 5-form. and so this would be insufficient for allowing an SO(n + 1) gauging. as a consistent reduction on S n . R) global symmetry. it would be necessary that the ungauged (D − n)-dimensional theory obtained by reducing instead on the torus T n should have a global symmetry group G that contains SO(n + 1) as a compact subgroup. with the Lagrangian e−1 LD = R − 1 2 F . R)/SO(n + 1) coset structure. related to the dilaton vector aE by 7 It is always the case that the counting of dilatons and axions would be consistent with the numerology required for an SL(n + 1. the following argument seems to suggest that the supersymmetric cases may be the only ones that can allow consistent S n sphere reductions. This again provides further supporting evidence for the consistency of the our gauged SO(6) reduction ansatz. Consider a D-dimensional theory of gravity plus an n-form field strength. (In fact the full set eaEi ·φ (∂χi )2 . R). the strengths of the dilaton couplings to axions from these two sources must be the same. R) [20.) It was shown in [22] that the strengths of dilaton couplings can be conveniently characterised in terms of the quantity ∆. since an SO(n + 1) factor in the denominator group would be gauged in the spherical reduction. It seems.21].282 M. R) global symmetry can be enhanced [20. All the known examples are associated with supersymmetric higher-dimensional theories. that the requirement of consistency of the sphere reduction has singled out a ten-dimensional starting point that is itself embeddable in a supersymmetric theory. / Nuclear Physics B 586 (2000) 275–286 proven in [8]. 2n! n (20) √ where e = −g. which has SO(n) as its maximal compact subgroup. For this conspiracy to occur. A reduction on T n always produces a theory with a GL(n. This raises the more general question of what possible higher-dimensional theories might allow consistent Kaluza–Klein sphere reductions. However. In special cases the GL(n. but this happens only if there is a “conspiracy” between axionic scalars coming from the metric and axions coming from the form-field Fn . 4. but this does not in general imply that the enhanced coset actually occurs. but one might wonder whether this was just a reflection of the fact that these are the cases that have received the most attention in the literature. Cvetiˇc et al. If this were to give a (D − n)-dimensional theory with an SO(n + 1) gauge group. where the constant vectors aEi satisfy (E of aEi vectors would constitute the positive-root vectors of SL(n + 1. 7 Specifically.

M. 9 As we have seen in Section 2 of this paper. scalars Tij parameterising the coset SL(n + 1. including not only the ten Yang–Mills fields of SO(5) but also the five 3-form field strengths coming from the antisymmetric tensor. n) = (11. it is necessary to include an F(4) ∧ F(4) ∧ A(3) term in the original Lagrangian (20). A similar situation arises with the S 4 reduction from D = 11. n) values for the Lagrangian (20) have ij been suggested in [19]. 4) and (11. In particular. In order for an S 7 reduction that retains all these fields to be consistent. it was shown that the Kaluza–Klein vectors and axions coming from a toroidal reduction of the metric always have ∆ = 4. then the total set of 70 spin-0 fields decompose as a 35v of scalars and a 35c of pseudoscalars. For example. Furthermore. 5). This point also emphasises that in the S 7 reduction one cannot consider just the 35 scalars Tij in isolation. the full set of bosonic fields of N = 8 gauged supergravity (and not merely the 28 gauge fields) must be included if the full set of Tij scalars are present. Cvetiˇc et al. . The 35 scalars in Tij correspond to the 35v . n) = (10. namely the self-duality of the 5-form in the case of (D. (10. Including the full set of 14 scalars Tij in a consistent reduction will force the complete set of massless fields to be nonvanishing. since the quantity ∆ is preserved under toroidal Kaluza–Klein reduction. D−2 283 (21) for an n-form field strength in D dimensions. and it is this value that is preserved under toroidal reduction. if we consider the S 7 reduction from D = 11. Another possibility for obtaining further examples of consistent sphere reductions is to include a dilaton in the higher-dimensional Lagrangian. / Nuclear Physics B 586 (2000) 275–286 aE 2 = ∆ − 2(n − 1)(D − n − 1) . 5). 7). An enumeration of all possible cases shows that for Lagrangians of the form (20) one gets ∆ = 4 only for (D. and in turn these excite the remaining 35c of pseudoscalars. 9 Consistent sphere reductions with scalars T for more general (D. We expect that a more complete analysis of the consistency of the reduction would exclude such possibilities. whose coupling to the n-form field strength is arranged to have ∆ = 4: 8 If there is no dilaton in the original higher-dimensional theory then the “dilaton coupling” ∆ is given by setting aE 2 = 0 in (21). The consistency of the reduction is then only possible if the FFA term of D = 11 supergravity is included. (11. (22) The above considerations seem to single out the three cases in (22) as the only ones where an S n reduction of a Lagrangian of the form (20) could consistently yield the gauge fields of SO(n + 1) and the scalars Tij of the coset SL(n + 1. R) to take place. Turning on these forces all the 28 gauge fields of SO(8) to be excited. 4). R)/SO(n + 1) only occur in the three cases listed in (22). n) = (11. By the same token one can expect that a consistent reduction would only be possible in the cases (D. R)/SO(n + 1). It therefore follows that for the symmetry enhancement to SL(n + 1. 7) if additional structure is also present in the eleven-dimensional Lagrangian. with precisely the coefficient dictated by D = 11 supersymmetry. it can turn out that additional structure is also required in order to achieve consistency. the dilatoncoupling ∆ for the original n-form field 8 in (20) must also take the value ∆ = 4.

then we must have D 6 11. First of all. Before the gauging. R)/SO(6) submanifold of the full scalar manifold. where we retain the twenty scalars Tij . . D − 2)/(SO(D − 2) × SO(D − 2)). we cannot ignore the fifteen SO(6) gauge fields. for n = 3 one can achieve ∆ = 4 in an arbitrary dimension D. since the scalars act as sources for them. the Lagrangian (23) then corresponds to the low-energy effective theory of the D-dimensional bosonic string. (It is only necessary to consider forms with n 6 D/2 since Hodge duality maps those with n > D/2 into this range. and we obtain the scalar coset SL(D − 2. which itself is a consistent truncation of type IIB supergravity. 2 2n! with a 2 = 4 − 2(n − 1)(D − n − 1)/(D − 2). unlike the situation in ungauged supergravity. and the twenty scalars of the SL(6. comprising gravity. Similarly. R)/SO(D − 2) together with the additional gauge fields of SO(D − 2). for an n-form with 4 6 n 6 D − 4 it can be seen that we must have D 6 11. the scalar coset in D = 3 is SO(D − 2. For example.284 M. Conclusions and further comments In this paper we have constructed a consistent Kaluza–Klein reduction ansatz for embedding the subset of the fields of five-dimensional N = 8 gauged supergravity. we can expect that it should be consistent to reduce the theory with n = 3 on a (D − 3)-sphere. A consistent sphere reduction. and a singlet scalar (which is the original dilaton of the D-dimensional theory). (Lagrangians of this type without the restriction on the value of the constant a have also been discussed in [19] in the context of sphere reductions. The possibilities for achieving ∆ = 4 couplings are in fact rather limited. into type IIB supergravity in D = 10.) Interestingly. In the context of the truncation that we consider in this paper. but rather a warped product of a domain wall and S n . Cvetiˇc et al. has in fact been obtained in an example where there is a dilaton in the higher-dimensional theory that couples to the form-field.) In this case there will no longer be an AdSD−n × S n vacuum solution. A crucial point in the analysis is that in the gauged five-dimensional supergravity one cannot consistently set the Yang–Mills fields to zero. / Nuclear Physics B 586 (2000) 275–286 1 1 aφ 2 e−1 LD = R − (∂φ)2 − (23) e Fn . This embedding can equivalently be viewed as a complete reduction ansatz (with no truncation of massless fields) for the ten-dimensional theory comprising just gravity plus a self-dual 5-form. if n = 4. while retaining the full set of scalar fields. One of the SO(D − 2) denominator group factors can be gauged. given that a should be a real number. We can expect that the full 3-sphere reduction of this Lagrangian (keeping the complete SO(4) gauge fields of its isometry group. albeit of a slightly different kind. the fifteen SO(6) gauge fields. 5. not merely the SU(2) subset of leftinvariant fields) will be consistent. In [6] it was shown that a reduction of the massive type IIA supergravity on an internal 4-dimensional space that is locally S 4 gives rise to the N = 2 SU(2) gauged supergravity in D = 6. 10 We saw that the self-duality 10 This also implies that solutions built using these scalars will in general require non-vanishing Yang–Mills fields. while for n = 5 we must have D 6 10.

We also discussed the more general possibilities that might arise if one includes a dilaton in the higher-dimensional theory. The consistency of the S 7 truncation in D = 11 supergravity. This special feature of the D = 5 gauged supergravity may have implications in the dual four-dimensional N = 4 super Yang–Mills theory. Nilsson.E. R)/SO(n + 1) will also have to include at least the gauge fields of SO(n + 1). Singapore. C. and furthermore will only be possible for the S 4 and S 7 reductions of D = 11. / Nuclear Physics B 586 (2000) 275–286 285 of the 5-form field of the type IIB theory plays an essential rôle in the consistency of the reduction. In particular. and has not yet been pushed to completion. de Wit.W. [2] M. the construction of the ansatz for the antisymmetric tensor fields is rather involved. Pope. B 281 (1987) 211.N.). and the S 5 reduction of D = 10.P. while in D = 4 they must be supplemented by the 35 pseudo-scalars. in: S. and in [24]. in order to achieve consistency. the 10 and 10 of pseudo-scalars lead to a considerably more complicated metric reduction ansatz. for example. C. is grateful to the University of Pennsylvania for hospitality during the course of this work. then a consistent reduction on S n that retains the scalars Tij of the coset SL(n + 1. References [1] B. but in D = 5 it acts as a “source” for the Yang–Mills fields themselves. J.N. World Scientific. In a full S 5 reduction of type IIB supergravity there will be additional fields coming from the reduction of the NS–NS and R–R 2-form potentials.N. B. In fact the consistency will in addition require that the theories in D = 11 and D = 10 have the additional structures associated with D = 11 and type IIB supergravity. Cvetiˇc et al. namely the FFA term in D = 11. [3] M. and from the dilaton and axion. Vol. On the other hand in D = 5 no additional fields beyond the ij scalars Tij and the gauge fields A(1) are required for consistency. The possibilities for further examples of consistent sphere reductions seem to be rather limited. Supersymmetry and Supergravity. Kaluza–Klein supergravity. In fact in all three cases this is related to the presence of the terms of the form (10). 130 (1986) 1. Phys.G. Nucl. In D = 7 the scalars and Yang–Mills fields must be supplemented by the five 2-form potentials. A complete analysis of the S 5 reduction can therefore be expected to be extremely complicated. Pope.J. The ansatz for a subset of the fields that included one scalar and one pseudoscalar was derived in [23]. Nicolai. In D = 7 this term acts as a source for the five 3-form fields. Duff. and the self-duality of the 5-form in D = 10. 1983. as discussed in Section 4. Kaluza–Klein supergravity and the seven sphere. More generally. Rep. bilinear in Yang–Mills fields. van Nieuwenhuizen (Eds.J. . Duff. Taylor. we showed that if one starts from a theory comprising gravity and an nform field strength. 82. However even in that case. H. in D = 4 it acts as a source for the 35 pseudo-scalars. Phys. P. Acknowledgement C.M. Ferrara.

hep-th/9903214. van Nieuwenhuizen. Phys.S. H.J. Consistency of Kaluza–Klein sphere reductions of symmetric potentials. Phys. B 272 (1986) 598. C. [6] M. [19] H. [23] M. Pope. hep-th/9905075. B. Liu. Symmetric potentials of gauged supergravities in diverse dimensions and Coulomb branch of gauge theories. S. hep-th/9710119. C. C. C. hep-th/9909121. L. Consistency of the AdS7 × S4 reduction and the origin of self-duality in odd dimensions. Lü. Gauged six-dimensional supergravity from massive type IIA.N. B 475 (2000) 261. Sezgin. 2 (1998) 231. Pope. Geometry of the embedding of supergravity scalar manifolds in D = 11 and D = 10. Math. Dualisation of dualities. Gubser. H. Romans. Phys. Lu. M.286 M. Nucl.P. B. hep-th/9906221. A new supersymmetric compactification of chiral type IIB supergravity. Lü. R. D 32 (1985) 389. [16] J. Phys. P. Phys. 83 (1999) 5226. Lett. Stainless super p-branes. hepth/9802150.M.N. hep-th/9910252. [13] M. Exact embedding of N = 1. Theor. [10] M. Kim. / Nuclear Physics B 586 (2000) 275–286 [4] H. H. hep-th/9802109. [8] H.X. [24] K. Pope. Cvetiˇc et al. Lett. Lü. hep-th/0002192.R. C. Tran. Rev. Cremmer. Cvetiˇc. Cvetiˇc. Phys.T. Lü. Embedding AdS black holes in ten and eleven dimensions. I. Phys. H. [15] A. Theor.P. Math. Pope. P. D = 7 gauged supergravity in D = 11. Phys. van Nieuwenhuizen. Tran. H. Anti-de Sitter space and holography. Lü. Pope. Gauge theory correlators from non-critical string theory. [9] M. P. [5] H. Nastase. SO(4) gauged supergravity from D = 11. C. Pilch.A. Nucl. Pope. hep-th/9906168. Phys. [12] M.P. B 456 (1995) 669. van Nieuwenhuizen. hep-th/9909203. hep-th/9911238.N.N. On the non-linear KK reductions on spheres of supergravity theories. J. Cvetiˇc. [21] E. Nastase. Vaman. hep-th/0002028. H. D. Stelle. Polyakov. Cvetiˇc. Pope. Vaman. hep-th/9909099. Phys. Warner. [11] M. Rev. Lü. Pope. C.N. Romans. Nucl. N. Lett. 2 (1998) 253. New vacua of gauged N = 8 supergravity in five dimensions. H. hep-th/9812035. Consistent nonlinear KK reduction of 11d supergravity on AdS7 × S4 and self-duality in odd dimensions. Warner.N. J. Sati. Martinez-Acosta. Adv. B 558 (1999) 96. N.J. E. Four-dimensional N = 4. H. Cvetiˇc. Lü. D. A. Five-dimensional N = 4 SU(2) × U (1) gauged supergravity from type IIB. Higher-dimensional origin of D = 3 coset symmetries. N. C. hep-th/0002099. Phys. Witten. Lü. Pope. Maldacena. B 523 (1998) 73. B. to appear in Phys. [14] H. C. Julia. Phys. Duff. Nucl. K. B 467 (1999) 67.N. The large N limit of superconformal field theories and supergravity. [18] E.A. K. Cvetiˇc.S. D. Lü. D. [17] S. Lü. P. Klebanov.J.J. Compact and non-compact gauged supergravity theories in five dimensions.N. Nastase. Günaydin. [7] H. L. C. C. Vaman. T. [20] E. Julia. .N. Adv. A.S. B 428 (1998) 105. Sadrzadeh. to appear in Nucl. Khavaev. Lett. Lü. Pope. Cremmer. Warner. hep-th/0002056. Gubser. T. hep-th/9508042. Pilch. [22] H.N. Pope.N. Mass spectrum of ten-dimensional N = 2 supergravity on S 5 . hep-th/9711200. Rev. Hoxha. H.

All rights reserved. these models possess a rich structure of possible deformations which may ∗ Corresponding author. Introduction Four-dimensional N = 1 supersymmetric type IIB orientifolds (see [1–13] and references therein) provide a domain of perturbative string vacua with novel properties (as opposed to the perturbative heterotic solutions) with potentially interesting phenomenological implications. University of Pennsylvania. Flat directions 1.hep. even within the current fairly limited class based on symmetric orbifolds.and F-flat deformations of the T-dual model. All rights reserved.elsevier.upenn. When continuous Wilson lines become discrete the gauge symmetry is enhanced and the T-dual orientifold corresponds to branes sitting at the orbifold fixed points. is the development of techniques that would yield a larger class of solutions than those of based on symmetric orientifold constructions. and their field theory realization Mirjam Cvetiˇc.nl/locate/npe D = 4 N = 1 type IIB orientifolds with continuous Wilson lines. Philadelphia PA 19104-6396. where the branes sit at the origin (original model without Wilson lines) as well as a deformation of the T-dual model where sets of branes sit at the fixed points (the model with discrete Wilson lines). 0550-3213/00/$ – see front matter  2000 Elsevier Science B. USA Received 13 June 2000.Mj. One goal. Wilson lines. There is a field theoretic analog describing these phenomena as D. moving branes. 11. that is far from being achieved. and their T-dual realizations as orientifolds with moving branes. Paul Langacker ∗ Department of Physics and Astronomy. However.25.V.Ex Keywords: Orientifolds.4 . Langacker). We demonstrate these phenomena for the prototype Z3 orientifold: we present an explicit construction of the general set of continuous Wilson lines as well as their explicit field theoretic realization.  2000 Elsevier Science B. PII: S 0 5 5 0 .3 2 1 3 ( 0 0 ) 0 0 4 1 4 .Nuclear Physics B 586 (2000) 287–302 www.edu (P. accepted 28 June 2000 Abstract We investigate four-dimensional N = 1 type IIB orientifolds with continuous Wilson lines.15. E-mail address: plg@langacker.V. PACS: 11.

e. by identifying the moduli space of D. Z3 orientifold model [1]. i. both continuous and discrete. parallel with the field theory treatment.e.and F-flat directions of the effective theory. This phenomenon is notoriously difficult to describe within the full string theory context.18]. one interesting phenomenon to explore is the blowing-up of the orientifold singularities [14– 16].. also the full string theory construction. the explicit unitary representation has not been given.e..21] and references therein). was exhibited for a number of examples [8].12] and references therein).. and sets the stage for further investigations of more involved orientifold models with continuous Wilson lines [19]. However. one identifies specific D. On the other hand the explicit field-theoretic realization in terms of (non-Abelian) flat directions allows the determination of the surviving gauge groups and massless spectrum. where branes are located at the orbifold fixed points.e. P. and to find their T-dual interpretation. The purpose of this paper is to set the stage for constructions of four-dimensional N = 1 type IIB orientifolds with continuous Wilson lines. the construction of explicit continuous Wilson lines allows for an explicit string theory realization where sets of branes are located at an arbitrary distance away from the fixed points. The explicit realization of these complementary pictures provide a beautiful correspondence between the two approaches. where branes are located away from the orbifold fixed points. . and here one expects to have. i. and their T-dual realizations as orientifolds with moving branes. these deformations correspond on the string side (in the T-dual picture) to the “motion” of a set of branes away from the fixed points. Another set of deformations corresponds to the introduction of Wilson lines. described as a power-series in the magnitude of the vacuum expectation values of the fields responsible for the deformations. i. On the other hand the field theoretical approach is only perturbative in the vacuum expectation values (VEV’s). while anticipated in general ([8. The deformations in the space of supersymmetric four-dimensional solutions always have a field-theoretic realization. The purpose of this paper is to address the study of such continuous Wilson lines of four-dimensional N = 1 orientifolds. however. Cvetiˇc.. In particular. and thus in the string picture corresponds only to a deformation infinitesimally away from the undeformed model. Explicit examples of discrete Wilson lines have been constructed for a number of a different orientifold models (see [8. It is also believed that in general there should be a field theoretical realization of the same phenomena. Langacker / Nuclear Physics B 586 (2000) 287–302 provide a fruitful ground for further investigation of their phenomenological implications (with the ultimate goal to identify classes of models with quasi-realistic features). since the metric of the blown-up space is not explicitly known. Continuous Wilson lines were first addressed in [8]. The connection of models with continuous Wilson lines to the T-dual models. we concentrate on explicit constructions of the continuous Wilson line solutions and the corresponding field theory realizations within the prototype.288 M. The new (deformed) supersymmetric ground states correspond to the exact string solutions. i. by constructing explicitly these Wilson lines. as well as from the field theory side.and F-flat directions of the original (undeformed) model.11. In particular. from both the full string theory description. which are different in nature from that of perturbative heterotic orbifolds [17.20.

v2 . the orientifold group is G = G1 + ΩG2 = Z3 + ΩZ3 . (The case of additional discrete symmetries in the orientifold group may require the presence of multiple sets of D5 branes as well.1. g 0 ∈ G2 = Z3 .) The compactified tori are described by complex coordinates Xi .1). We show that these solutions allow a continuous interpolation between the original model without Wilson lines and the models with discrete Wilson lines. D = 4. P. In Section 2. We then (Section 3. i = 1. v3 ) (subject to constraint 3i=1 vi = 1): g : Xi → e2iπvi Xi .1 we summarize the salient features of the Z3 orientifold construction.and F-flat directions and demonstrate their one-to-one correspondence with the continuous Wilson line string constructions.2) provide explicit constructions of such D. requires the inclusion of an even number of D9 branes. Z3 orientifold with continuous Wilson lines 2. (1) For the Z3 orientifold v1 = v1 = v3 = The tadpole cancellation. we provide a systematic analysis of their realization on the field theory side. associated with the open-string modes.) An open string state is denoted as |Ψ. and construct for this model the most general set of continuous Wilson lines that in the T-dual picture correspond to the moving branes. In Section 3 we turn to the field theoretical analysis. In Section 2. 3.and F-flat directions (Section 3. first discrete and then continuous Wilson lines. (Closure requires ΩgΩg 0 ∈ G1 = Z3 for g. 2. One starts with type IIB string theory compactified on a T 6 /Z3 (T 6 -sixtorus. In addition. i. 2..M. Langacker / Nuclear Physics B 586 (2000) 287–302 289 This paper advances these topics in several ways. The action of an orbifold group ZN on the compactified dimensions can be summarized via a twist P vector v = (v1 . j the Chan–Paton indices associated with the end points on a D9 brane. In particular. When continuous Wilson lines become discrete the gauge symmetry is enhanced and the T-dual orientifold corresponds to branes sitting at the orbifold fixed points. which is chosen to be accompanied by the same discrete symmetry G2 ≡ Z3 .e. specifically constructed for the Z3 orientifold. G1 ≡ Z3 — the discrete orbifold group) and mod out by the world-sheet parity operation Ω. Cvetiˇc. The elements g ∈ G1 = Z3 act on open string states as follows: . by first recapitulating the techniques for a classification of D. The paper is organized in the following way. we provide the first explicit unitary representation of the continuous Wilson line.2 we proceed with the construction of. ij i where Ψ denotes the world-sheet state and i. N = 1 Z3 orientifold We briefly summarize the construction [1] of four-dimensional Z3 type IIB orientifold models.

ij i → (γg )ii 0 .  (2) g : |Ψ.

the elements of ΩG1 = ΩZ3 act as . 1 3. i 0 j 0 γg−1 j 0 j .g · Ψ. Similarly.

Ωg : |Ψ. −1  . ij i → (γΩg )ii 0 .

j 0 i 0 γΩg j 0j (3) .Ωg · Ψ.

3).9 −1 and λ(i) = e2iπvi γg. up to a phase. 13 . Gauge bosons in the D9 brane sector arise from open strings beginning and ending on D9 branes. However.9 will further reduce the group.9 = 116 0 where the subscript 9 denotes the D9 brane sector in which these matrices are acting. (5) the first constraint implies that the λ(0) are SO(32) generators. 3. γgk = ±γΩ∗ γgk γΩ . (10) For Z3 orientifold this yields a matter content of three copies of ψ α = (12. since the string vertices for the chiral matter superfields involve the oscillator modes of the target space toroidal coordinates Xi . give rise to space–time gauge bosons (i = 0) and matter states (i = 1. finiteness of string loop diagrams yields tadpole cancellation conditions which constrain the traces of γg matrices [1] Tr(γZ3 ) = −4. 2.9 −1 and λ(0) = γg. (6) The Z3 twist action on the tori is given by the twist vector v = ( 13 . (4) γg3 = ±1. in the following.9 λ(0) γΩ. χ α = (66.9 λ(i) γg. in accordance with the usual rules for multiplication of group elements. consistency with group multiplication implies some conditions on the γg . 1)+2 .9λ(0) γg. where ω = e2πi/3 . The result is the gauge group: U (12) × SO(8). i = 0. (11) . The explicit representation for the D9 brane sector γΩ is symmetric and can be chosen real:   0 116 . Since the γg form a projective representation of the orientifold group. 2. while the constraints from the γg. (5) γΩ. 3. 14 . 14 . Ω 2 = 1 and ΩgΩg 0 ∈ Z3 for g. P. (8) With Eq. . Cvetiˇc. Invariance of these states under the action of the orientifold group requires T −1 λ(0) = −γΩ. if we choose the plus sign for D9 branes (and the minus sign for D5 branes) [8]. 8)−1 . . (4) and (6). .9 . Langacker / Nuclear Physics B 586 (2000) 287–302 where we have defined γΩg = γg γΩ . Further. Note. It turns out that the tadpoles cancel. This choice satisfies Eqs. ω2 112 . the Chan–Paton matrices now transform under the orbifold action (in order to render the physical states invariant under the orbifold action). α = 1. thus implying: T −1 λ(i) = −γΩ. Consider the G1 = G2 = Z3 case. (9) The Chan–Paton matrices of the matter states have to be invariant under the action of the orientifold group as well. that Ω exchanges the Chan–Paton indices.290 M.9 λ(i) γΩ. 13 ) and its action on Chan–Paton matrices is generated by:  (7) γZ3 = diag ω112 . whose Chan–Paton matrices will be denoted by λ(i) . Open string states. g 0 ∈ Z3 respectively imply: ∗ γΩ = ±γΩT . g 3 = 1. .9 .

The above consistency conditions reduce to the unique solution of the discrete Wilson line:  (17) γW = diag 116−3n .. S = (1. b}-U (12) indices. 1. . tadpole cancellations require  Tr(γZ3 ) = Tr(γZ3 γW ) = Tr γZ3 γW2 = −4. Z∗ ⊗ 1n . n) (23) α α . 8 − 2n. 14−n . 1. n.−1 . let us rearrange the entries in the γZ3 matrix (7) as follows:  γZ3 = diag ω2 16−n . ω2 16−n . β. ω16−n . respectively. It satisfies the following algebraic consistency conditions: (γZ3 γW )3 = +1. Cvetiˇc. i. (20) The matter comes in three copies and has the following representation (the subscripts correspond to the self-evident U (1) charges): χ α = ((6 − n)(11 − 2n). (18) which further breaks the gauge group down to: U (12 − 2n) × SO(8 − 2n) × U (n)3 . 116−3n. . (12) where α. whose real and imaginary components arise from the NS–NS and R–R sector. γ are family indices. γW ] = 0. [γZ3 . P. ω2 13n . The surviving gauge symmetry is determined by a projection: λ(0) = γW λ(0) γW−1 . (22) ¯ 1. 1. which is represented by a matrix γW . is such that it commutes with γZ3 . Z ⊗ 1n . ω. 1. ω13n . ω = e2πi/3 . (21) ψ = (12 − 2n. 4}. 1 . . it describes a discrete Wilson line.e. ω16−n . The closed string sector yields the gravity supermultiplet and the 36 (chiral) supermultiplets corresponding to the 9 untwisted (“toroidal”) and 27 twisted (blowing-up) sector moduli. Let us focus on a Wilson line. 14−n . . Langacker / Nuclear Physics B 586 (2000) 287–302 291 where the subscript refers to the U (1) charge of U (12). and i-SO(8) indices. (19) The matter representation is determined by the condition: λ(i) = γW λ(i) γW−1 . Wilson lines Discrete Wilson line When the action of the Wilson line on the Chan–Paton matrices. The renormalizable superpotential is of the form βb γ W ∼ αβγ ψiαa ψi χ[a. (14) To simplify the notation. acting along the two-torus coordinate Xi . 1)1. say i = 3. 1. n = {0.b] .2. 1. it depends only on discrete values of parameters. {a. (13) Further.M. 1)−2. The moduli are gauge singlets. (15)  (16) Z = diag ω2 . 2.

3. Wn∗ . where Un is a general (n × n)dimensional unitary matrix and U30 is a (3 × 3)-dimensional diagonal unitary matrix. (27) The ansatz for the Wilson line is taken to be of the form: ∗  . 1n ⊗ Z. b}-U (12 − 2n) indices. . β. . 2. Q = (1. . to complex conjugate states of branes located at the other fixed point away from the origin. 116−3n. 14−n .b] + Si1 3 Pi2 1 Qi3 2 with α 6= β 6= γ . Namely. by orientifold projection. located away from the origin. 1. the third complex direction X3 . Starting with a general ansatz: . ω16−n .1 .292 M. we again rearrange the entries in the γZ3 matrix. . P.1 . except that now it need not commute with γZ3 . The action of discrete Wilson lines implies that one can take n sets (n = 1. away from the origin (in a particular complex plane). There is a T-dual interpretation of this solution. n.  γZ3 = diag ω2 16−n . γ are again family indices. ω2 16−n . . (This motion can be accomplished in sets of six D7 branes. Such a Wilson line can be uniquely determined up to unitary transformations that commute with γZ3 .) Note that the string states associated with branes located at one Z3 orbifold fixed point. {a. The part of such unitary transformations that affects γW0 is of the form: Un ⊗ U30 . W1 . say.3}-respective indices for the three U (n) factors. . γW0 (28) where γW0 is a (3n × 3n)-dimensional unitary matrix. T-dualizing the original model along. . 1n ⊗ Z∗ . γW = diag 116−3n . W1∗ . We are still after a unitary representation γW that satisfies conditions (13)–(14). . and {i1. (25) The superpotential is of the form: βb γ αi βi γi W ∼ αβγ ψiαa ψi χ[a. Wn . ¯ n)−1. where Wi are (3 × 3)-dimensional unitary matrices subject to the following consistency conditions:  (30) (ZWi )3 = 13 . Employing such transformations in turn enables one to cast γW in the following most general form:  (29) γW = diag 116−3n . 1. . (26) where α. For convenience.2. Continuous Wilson lines As the next step we generalize the construction to the case of continuous Wilson lines. . We have reduced the problem to finding an explicit representation of the matrices Wi . i-SO(8 − 2n) indices. γW0 . 116−3n. ω16−n . . of six since each such set is moded out by six elements of the combined Z3 and Ω group action. n. are related. Tr(ZWi ) = Tr ZWi2 = 0. α (24) a = 1. ¯ n. Langacker / Nuclear Physics B 586 (2000) 287–302 P α = (1. Cvetiˇc. 1. . 4) of branes (each set containing six D7 branes) to be at one of the two Z3 orbifold fixed points. corresponds to the model with 32 D7 branes sitting at the origin of the third complex plane. 14−n . 1)−1.

However. To further determine the phases φa . this is due to the fact that the co-factors of the diagonal entries in the above matrix are all equal to w(w + x) − aa 0 . (37) + (38) + (39) is precisely the condition det(W ) = 1. w. 0 −aa we 0 −aa we iφ iφ +a e 3 iφB +w e 3 3iφw (37) =a . we √ shall change the notation (|a|. we introduce φA = 2φa 0 + φa + φb − φc . P. |b| = |a 0 |. w) > 0. |a |. φb . (33) |c| = |a|. 0 At this point. 2aa 0w (43) . W0 =    aa 0 + ωwx aa 0 − ω2 x(x + w) 2 w−ω x b c (32) where a. (36) and reduce the remaining unitarity conditions to the following set of equations: −aa 0weiφ + a 03 eiφA = a 02 . cos(φB ) = 2a 3 (w4 + w2 − a 2 a 02 ) . Cvetiˇc. The unitarity conditions (equating W † entries with the corresponding co-factors of W ) allow one to further constrain the parameters: q |w| = 1 − |a|2 − |a 0 |2 . (35) φ = φ w + φa + φa 0 . x are complex numbers subject to the constraint det(W ) = 1.) The solution of (37)–(39) gives: (a 4 + a 2 a 02 + a 04 + a 02 − a 2 ) . |w|) → 0 2 02 (a. c. provided (37)–(39) hold. b 0 c0 w3 (31) the conditions (30) yield the following general form:   w a b   a0 w+x c . φw . 2a 03 (a 4 + a 2 a 02 + a 04 − a 02 + a 2 ) .M. subject to the constraint w = 1 − a − a . and due to unitarity they should be proportional to the complex conjugate diagonal entries in (32). unitarity imposes an immediate constraint x = 0. for the sake of simplicity. namely. b. (38) =w . cos(3φw ) = 2w3 cos(φA ) = (40) (41) (42) as well as cos(φ) = (a 4 + a 2 a 02 + a 04 − a 02 − a 2 ) . (39) 2 2 ((ZW )3 = 13 is automatically satisfied. a . Langacker / Nuclear Physics B 586 (2000) 287–302 293   w1 a b W =  a 0 w2 c  . (34) φB = 2φa + φa 0 − φb + φc . φa 0 . φc .

(40)–(42). the relationships: a 03 sin(φA ) = a 3 sin(φB ) = w3 sin(3φw ) = aa 0 w sin(φ) that follow from (37)–(39) are useful. a 0 . (44) One can show that the expressions (40)–(43) indeed ensure cos(3φ) = cos(3φw + φA + φB ). a 0 (or equivalently the Euler angles ϕ and ψ) in which 3φ0 varies from 0 to ±2π . eiφ0 ≡ A + i 1 − A2 2w3 cos(φw )1 = (46) (47) (48) (49) For the limit A = 1. 2 where  1/3 p (w4 + w2 − a 2 a 02 ) .. 2  1 cos(φw )2 = ωeiφ0 + ω2 e−iφ0 . P. a. (37)–(39) turn out to be automatically satisfied. 2  1 cos(φw )3 = ω2 eiφ0 + ωe−iφ0 . a 0 = sin ψ sin ϕ. which are respectively the case of no Wilson line. and the only remaining phases are the two “gauge invariant” phase parameters φA. The latter are further constrained by (37)–(39) and (44). (Equivalently the Wilson line can be specified in terms of the two Euler angles ϕ and ψ of the three-sphere. it is convenient to solve (42) for cos φw . 3φ0 = 0 and 3φ0 = π correspond to A = +1 and −1. which is the most general unitary matrix that commutes with Z. a = sin ψ cos ϕ. all specified.294 M.e. 3φ = 3φw + φA + φB . In particular.) To study the monodromy properties of this Wilson line. which occur on the boundaries of the allowed region at the points a = a 0 = 0 and a = a 0 = 23 .) Thus we have arrived at the following form of the matrix W :   aeiφa a 0 eiφb weiφw a 0 eiφa0 weiφw aeiφc  . φB . Thus. w = 1. ω. positive parameters w. which has three roots:  1 iφ0 e + e−iφ0 . corresponding (in the T-dual picture) to the set of D7 branes sitting at the origin or at one of the two fixed points.B and the phase of the diagonal element φw . i. eiϕ2 . and thus provides an equivalency class for the Wilson line representations. by Eqs. To summarize. introduced as: w = cos ψ.. up to signs and multiples of 2π . i. subject to the constraint w2 + a 2 + a 02 = 1.e. For that purpose one has to find a path in the space of w. two phase parameters in W can be “gauged away”. in terms of two real parameters. A≡ . The continuous Wilson line interpolates continuously between these limits. Langacker / Nuclear Physics B 586 (2000) 287–302 Two additional constraints of Eqs. ω2 . φw and φ are not independent. or discrete Wilson lines. (In proving this identity. There is . these solutions reduce to φw = 0. respectively. the final form of (45) is specified in terms of three real. In addition. a. It is straightforward to find such paths. Cvetiˇc. the phases φA . eiϕ3 ). a = a 0 = 0. (45) W = i(φa +φa 0 −φb ) 0 i(φa +φa 0 −φc ) ae ae weiφw It is determined up to diagonal unitary transformations U30 = diag(eiϕ1 .

P. In particular.g. allowing a continuous interpolation. The continuous Wilson line considered there corresponds to the case with W1 = W2 = W3 = W4 . It is convenient to use the freedom 4π in U30 to require φb = φc .M. n). Langacker / Nuclear Physics B 586 (2000) 287–302 295 a continuous path between these.a 0 by the same amount.g. This occurs at a = a 0 = 0. ω2 ) and the remaining (n − k) W -matrices remain off-diagonal (continuous). As 3φw passes through 2π . However for special values of the Wi parameters additional gauge enhancement can take place.. The Wilson line (29) has a T-dual interpretation in terms of n sets of six D7 branes each moving in. and thus for the full one-to-one correspondence with the continuous Wilson line picture the Wilson line should depend on two real parameters (a and a 0 ) as well. . the T-dual picture corresponds to k-sets of six D7 branes sitting at one (of the two) orbifold fixed point away from the origin. can actually be related by a continuous interpolation as 3φw varies from 0 to 2π to 4π as a function of a and a 0 . 3φw can start from 0 and increase first to 2π and then to 4π as one moves back and forth between A = +1 and A = −1. e. From (30) it is clear that if W is a continuous Wilson line solution. Cvetiˇc. An examination of (37)–(44) reveals that φA and φB each decrease by π as 3φw increases by 2π . The above construction of the Wilson lines for the Z3 orientifold provides a generalization of the discrete Wilson lines (with n = 4) discussed in [8]. This general continuous Wilson line reduces to a hybrid (continuous/discrete) Wilson line if k W matrices become diagonal with elements ω (discrete Wilson lines) (or equivalently. Here. We thus see that these solutions. W1 = W2 · · · = Wk (k 6 n) yields the gauge group enhancement of U (1)n to U (k) × U (1)n−k . The surviving gauge group is generically: U (12 − 2n) × SO(8 − 2n) × U (1)n . 3 From A = ±1 one can always move along either branch of the square root. say. a = a 0 = 1 − √1 . with an obvious generalization to two sets W1 = W2 and W3 = W4 . there is a relation between the magnitude and phase of the complex parameters. (When one imposes the unitarity constraint on the symmetric matrix in [8]. related by the discrete Z3 symmetry for fixed a and a 0 .. so the changes in the Wilson line parameters are continuous. then so are ωW and ω2 W . . (50) as long as W1 6= W2 6= · · · Wn and (ai . . In particular. in order to preserve the signs of the angles. The gauge group U (1)n now becomes enhanced to U (k)3 × U (1)n−k . where the phases are indeterminant.B and φ must increase discontinuously by 3π and 2π . the third complex plane away from the orbifold fixed at the origin. and the matrix becomes a special one parameter case (with a = a 0 ) of the continuous Wilson described above. respectively. in which case an increase of φw by 2π 3 (or 3 ) is accompanied by an increase in φa. we have generalized and given an explicit unitary representation of Wi in terms of two parameters. For example. . ai0 ) 6= 0 (i = 1. when a subset of k Wilson line elements become discrete. e. from 3φ0 = 0 to 3φ0 = π and then to 3φ0 = 2π and further. which passes 3φ0 = π2 (A = 0) at. .) In the T-dual picture the position of the branes (in the third complex plane) should be parameterized by two real parameters. φA.

We checked that this is indeed the case. Since each W and WII are fully specified by two real parameters. The T-dual model. Cvetiˇc. γWJ ] = 0. except for two phases. On general grounds such Wilson lines are of the form (29) with the Wi ’s being of the form (45). For the additional Wilson lines γWJ one cannot gauge away the two phases.3 . Again.e. i. (52) Then. corresponds to the C 3 /Z3 model of 32 D3 branes sitting at the origin. We chose the first Wilson line γW by fixing the gauge. . However. with the phases subject to the analogous constraint: φcIII = φaIII + φc − φa . say. 0 = φb . respectively.296 M. when any Wilson line element becomes diagonal (discrete) the solution corresponds to a particular set of branes reaching the orbifold fixed point in the particular plane. φaII0 = φbII . respectively). (Of course. parameterized by γWII and γW .2. for a single Wilson line these two phases can be removed (“gauged away”) by a diagonal unitary transformation U30 that commutes with Z.. The elements in (45) are uniquely specified in terms of two real parameters. (The introduction of additional Wilson lines does not break the generic gauge group (50). complex directions corresponds to the model with 32 D5 branes sitting at the origin of the second and third complex plane. As discussed above. One can introduce the third Wilson line γWIII . The introduction of the third Wilson line γWIII . (53) the two Wilson lines commute. one still has the freedom to make an overall unitary transformation U30 on all three Wilson lines. the second X2 and third X3 . parameterizes the independent motion of n sets of (six) D3 branes along the first plane (along with the independent motions in the second and third planes. Langacker / Nuclear Physics B 586 (2000) 287–302 Multiple continuous Wilson lines As the last step we proceed with the construction of multiple Wilson lines. P.) T-dualizing the original model along two. in which one has dualized all three complex directions X1. if the phases for the corresponding matrices WII in the second Wilson line γWII are chosen as φcII = φaII + φc − φa . which is also uniquely specified by two real parameters for each WIII matrix. The action of continuous Wilson lines γWII and γW then corresponds to the independent motion of n sets of six D5 branes in the second and the third complex planes. these parameters are in one-to-one correspondence with the motion of the (n) sets of branes in the two complex planes. so that the phases for the matrices W are fixed as: φa = φa 0 = φb . choosing a specific U30 . the Wilson lines have to commute: [γW . φaIII III (54) Such a Wilson line turns out to commute both with γW and γWII .) The additional Wilson line γWII has again a T-dual interpretation. (51) This condition should in principle fix the undetermined phases.

Here. are gauge rotations of the original monomial. We also show how to deform from the discrete Wilson line solutions corresponding to n sets of branes located at the orbifold fixed points away from the origin. Cvetiˇc. involving products of HIP’s which have common multiplets. Classification of F. Each field in the monomial will typically have the same vacuum expectation value (VEV). The flat directions correspond to products of monomials from each of the HIP’s. (The detailed analysis of the blown-up Z3 orientifold was given in [14].j 1 These techniques were previously developed (see [24]) to construct the moduli space of the flat directions for models based on perturbative heterotic string models.and D-flat directions In this section we classify the F. . P.a i. those with different phases for the VEV’s of the fields in the monomial. 1 ) We first construct a gauge invariant polynomial from the non-Abelian fields.1.and D-flat directions of the new supersymmetric ground states that correspond to the deformation of the original model (and whose string theoretical construction we provided in the previous section). In the present model. and thus the supersymmetric ground state solutions were achieved by HIP’s which have non-zero U (1) charges. Other flat directions. The D-flat constraints for both diagonal and off diagonal generators of the non-Abelian gauge group are automatically satisfied. in which case the flat directions correspond to gauge invariant holomorphic monomials. e. (See [14] for more detailed discussion of such issues as overlapping HIP’s. Field theory realization 3.) 3. which is a sum of monomials involving the components of the fields. The constraints of F-flatness further require that h∂W/∂Φp i = 0 and hW i = 0 for all of the massless superfields Φp in the model.g. formed as products of other HIP’s. (55) DI = α. Langacker / Nuclear Physics B 586 (2000) 287–302 297 3. One can also consider D-flat directions with more than one independent VEV. each with its own VEV. For that purpose we utilize the one-to-one correspondence [23] of D-flat directions with holomorphic gauge-invariant polynomials (HIP’s) built out of the chiral fields in the model.. Flat directions corresponding to Wilson line solutions Unlike the blowing-up procedure [14–16] in which the blow-up introduced Fayet– Iliopoulos terms [22] for the anomalous U (1) terms. the D-flat directions necessarily involve non-Abelian fields due to the matter content. For simplicity. it suffices to check D-flatness for each case.2.M. Then one monomial term defines a D-flat direction. the deformations corresponding to the continuous Wilson lines correspond to the polynomials that are gauge invariant under the “anomalous” U (1) as well. using this technique. The D-flatness condition for SO(8) is X X αa†  ψi AIij ψjαa = 0. the flat direction analysis in [24] considered only the non-Abelian singlet fields in the model.

and only the product is SO(8).d] . X X αa† XX † J α J αb α ψi Tbab ψi + Tab χ[b. n = {1. It is necessary to consider a 6th order polynomial because the cubic (ψψχ) vanishes for a single family index due to the symmetry (antisymmetry) in SO(8) [U (12)] indices.. Several generalizations of (59) are possible.2] The direction (59) appears to break SO(8) × U (12) to SO(6) × U (10).. with matrix elements (Akl )ij = δj k δil − δj l δik .e.b] F-flatness is ensured by taking all fields from the same family.2] 1 2 [3. consider a concrete representation of non-Hermitian SO(8) and U (12) generators labeled by I = (kl) and J = (cd).g. .c] DJ = α a.  αc αd α  α ψi ψj χ[c. one can construct D. VEV v.2] . i. 144.b] = 0. . βb γ αβγ ψi χ[a. e. and will henceforth be denoted by (ψ131 ψ232 χ[1. 4}. ..g. For example. in which each factor is separately U (12) invariant. Langacker / Nuclear Physics B 586 (2000) 287–302 where AI are generators of the vector representation of SO(8) and I = 1. α = 3 for definiteness. . . (58) ψiαa ψjαb χ[a. .and F-flat directions from HIP’s of the form (ψψχ)(ψψχ).4] not D-flat under U (12) because. (60) (61) respectively. P. there is an additional surviving U (1) generated by a combination of the broken SO(8) and U (12) generators. A flat direction corresponds to a specific monomial in (58). .b.298 M. .2] 3 )2 . For U (12). . and Tdc  ab = δad δbc .e. (57) For the case of the original Z3 orientifold. with each of the three fields having the common This is just the square 2 of ψ131 ψ232 χ[1.. (ψ 31 ψ 32 χ 3 )(ψ 33 ψ 34 χ 3 ) is 1 2 [1. defined by  j (62) Fij = i Tji − Ti + Aij . In fact. 28. it involves overlapping U (12) polynomials. invariant. It is then straightforward to see that the Hermitian generator F12 .c] . b] = [c.2] 3 . One can consider a product of flat directions of the following schematic form: n Y (ψψχ)2i . (63) i=1 2 It is easy to check that (58) is only D-flat for [a. e. (In the blown-up case one must add a constant Fayet–Iliopoulos term ξFI to the D-term for the anomalous U (1) [14]. Cvetiˇc. .) The F-flatness conditions of the original Z3 orientifold model are βb αβγ ψiαa ψi = 0.i a.b where T J bJ (T ≡ −T J T ) are the generator matrices for the fundamental (anti-fundamental) representation of U (12) and J = 1. .c α. in the terminology of [14]. (56) χ[a. is conserved. Akl and Tdc . . to  31 32 3  3 ψ1 ψ2 χ[1. (59) ψ131 ψ232 χ[1. i. d]. To see this. .

i. These solutions provide a field theoretical realization of the motion of branes in multiple complex planes. which in the T-dual picture realizes the the motion of n sets of (six) D7 branes with groups of pi of them at the same position. respectively. F34 . (These directions are always F-flat. where the two sets of fields have VEV’s v1 and v2 . v1 = ω or ω2 . Cvetiˇc. . For example. the third complex plane. where F e13 ≡ For example. e13 + F e24 . P. the direction in which sets of pi of the n VEV’s P are equal. F v v v e24 . away from the fixed point at the origin. whose T-dual string theory construction in terms of multiple continuous Wilson line solutions was given at the end of the previous section. F34 . leads to Y U (pi ).M.4] 2 (64) for the case n = 2.e. the starting gauge group there is (19) with the massless particle content 3 One can show that the enhanced symmetry also holds if each pair of VEV’s differs by a factor of ω or ω2 . These directions are still F-flat provided one does not mix families within the same factor. say.. The T-dual string theory construction in terms of a generic continuous Wilson line. was derived in the previous section. One can generalize the construction of flat directions to include fields with all three family indices. The direction in (63) has the generic surviving gauge group U (12 − 2n) × SO(8 − 2n) × U (1)n . Langacker / Nuclear Physics B 586 (2000) 287–302 299 where each factor is a sixth order polynomial analogous to (58). if v1 = v2 in (64). and each factor has a non-zero VEV vi for each field in 2i−1 2i ψ2i−1 ψ2i χ2i−1. the discrete Wilson lines (17) provided a T-dual realization in terms of n sets of (six) D7 branes sitting at a fixed point away from the origin.) Most generally. the third complex plane. the result is valid only in the power series expansion in terms of the VEV’s of the fields. On the other hand. and F e14 − F e23 . and form the surviving group U (2). in.2i . (The U (1) generator is F12 + F34 . 4}. For example. one for each family with its own VEV viα but the same gauge group structure. thus providing for an enhancement of gauge symmetries. When p of the VEV’s in (63) are equal 3 . . . the gauge factor U (1)p is enhanced to U (p). in (64) the conserved U (2) generators become F12 . each factor in (63) can be generalized to a product of three factors. We have not found an analog of this freedom in i( v1 T31 − v2 T13 + A13 ). note again that the field theory allows for a realization of such a motion of branes only in the neighborhood of the original fixed point. F13 + F24 .2] 2 3 ψ333 ψ434 χ[3. . one can consider 3 ψ131 ψ232 χ[1. we have suppressed the third family index. However. and with each set at a different location. it is straightforward to show that the generators F12 . and F14 − F23 are conserved. (66) U (12 − 2n) × SO(8 − 2n) × i in one to one correspondence with the continuous Wilson line solutions (29) with pi (3 × 3)-dimensional matrices Wi equal.) For example. (65) This flat direction provides a field theory realization of a motion of n sets of (six) D7 branes in. say. with i pi = n = {1. and similarly for F 2 1 2 the continuous Wilson line construction. . given by (29).

Cvetiˇc. Similarly. where pi 6 4 − n. (69) S131 P131 Q31 1 generically breaks U (n)3 to U (1)q × U (n − q)3 . Conclusions In this paper we focused on the study of continuous Wilson lines within fourdimensional N = 1 type IIB orientifold models. . t˜12 ≡ t˜21† . For example. However. (70) i j i j The surviving gauge group is   Y X  Y X  pi × SO 8 − 2n − 2 pi × U (pi ) × U (qj ) U 12 − 2n − 2 i  X 3 qj . This corresponds to the multiple continuous Wilson lines and in the T-dual picture to the motion of sets of branes in three complex planes. then t˜21 ≡ v1 t2 + t2 + v2 t2 . Again. 4 This breaks the U (n)3 symmetry down to U (1) × U (n − 1)3 . where the U (1) generator is the diagonal sum t11 of three broken U (1) generators. including the choice of the specific branches for the interpretation of the phases φw . one with its own VEV for each family. and t22 form a conserved U (2). 4. provided the family indices are not mixed within a factor. 5 There are also enhanced symmetries for the products (SP Q)q in the case in which the VEV’s of each factor have relative phases of ω or ω2 .. 3.g. qj 6 n. We enforced unitarity and constructed the most general set of continuous Wilson lines within the original Z3 orientifold [1] and demonstrated that these models are in one-to-one correspondence with the T-dual models. ×U n − i i j (71) j This field theory picture of course has an analogous Wilson line realization encoded in a special choices of Wi matrices. P.300 M. . . 4 Each factor SP Q can again be replaced by a product of three factors. . (68) In (68) t (l) . if the two factors in (69) have v1 = ω 2 v (1)1 (2)1 v (3)1 or ω2 . The generators of the enhanced symmetry are generalizations of (68) in which v there are corresponding phases in the coefficients of the t (l) . e. t11 . where tji ≡ tj(1)i + tj(2)i + tj(3)i . ensures F-flatness. is the generator of the lth U (n) factor. The flat direction corresponding to the moving of one set of branes away from the fixed point (located away from the origin) is associated with the HIP αi αi αi Si1 3 Pi2 1 Qi3 2 → S131 P131 Q31 1 . we have not 2 1 found an analog on the string theory side. for special points with equal VEV’s for each of the factors 5 there is an enhanced symmetry U (q) × U (n − q)3 . Langacker / Nuclear Physics B 586 (2000) 287–302 (21)–(25). (67) where the restriction to a single family. l = 1. . say. a product of (SP Q)q with distinct gauge indices. such as. One can choose a “hybrid” flat direction composed of: X X Y Y (ψψχ)2pi (SP Q)qj . α = 3.  32 32 32  S2 P2 Q2 .

the sub-block that commutes with the corresponding sub-block of the γZ3 element and depends on a discrete parameter. i. Similarly. The number of parameters of such a continuous Wilson line is in one-to-one correspondence with the parameters that specify the location in the complex plane of each set of branes. We also analysed the field theoretic analog. in order to establish the general (and precise) correspondence between the models with continuous Wilson lines. Another more general direction involves a study of a general class of four-dimensional N = 1 type IIB orientifold models.. The Wilson line solution continuously interpolates between the limit of no Wilson line and the discrete solutions. In particular. The field theory describes these string solutions only in the proximity of the original models. when the full Wilson line becomes discrete the gauge group is enhanced to U (12 − 2n) × SO(8 − 2n) × U (n)3 . In general these models contain not only D9 branes but also. e. . specified by the holomorphic gauge invariant polynomials that parameterize the moduli space of the supersymmetric deformations of the original models. thus allowing for even more involved models. When a sub-block of the continuous Wilson line becomes discrete. describing the above string constructions as D. 4) of (six) branes moving in one (of the three) complex planes. where they are point-like.e. their T-dual interpretation. i.. Cvetiˇc. but also the massive spectrum of the string models. . . when k sub-blocks are equal then U (1)k is promoted to U (k).e. thus quantifying the correspondence between the two complementary approaches. Langacker / Nuclear Physics B 586 (2000) 287–302 301 where each Wilson line has an interpretation of n sets (n = 1. e. .M. The latter can be located at different points on a particular two-torus T 2 . . The dependence of the mass spectra and the couplings on the continuous Wilson line parameters.. that are in one-to-one correspondence with the parameters of the string constructions with the continuous Wilson lines (and their T-dual interpretation). as well as their field theory realization [19]. A gauge enhancement takes place for special values of the Wilson line parameters. D5 branes..g. P. it allows only for a powerseries expansion in the vacuum expectation values of the chiral superfields. We find the explicit form of the holomorphic polynomials. and may in turn lead to a discovery of potentially realistic open-string solutions. deserves further study. the explicit construction of the continuous Wilson lines would allow one to construct not only the massless. Investigation of these general classes of string solutions (by determining the gauge group. the mass spectra and the couplings) would shed light on the properties of a broad class of open string models within symmetric type IIB orientifold constructions. both from the explicit string construction as well as from the (perturbative) field theory perspective.and F-flat deformations of the effective field theory of the original model as well as deformations of the models with discrete Wilson lines. The generic Wilson lines break the original gauge group U (12) × SO(8) down to U (12 − 2n) × SO(8 − 2n) × U (1)n . The work sets the stage for further investigations of models with continuous Wilson lines. implementing simultaneously moving branes and the actions of continuous Wilson lines. this corresponds in the T-dual orientifold to branes sitting at the Z3 orbifold fixed points.g.

[4] Z.H. D 58 (1998) 101901. hep-th/9804110. Phys. R. Uranga. Ibáñez. Romelsberger. Phys. L. Cvetiˇc. J. hep-th/9705163.C. hep-th/9805157. Phys. JHEP 0004 (2000) 004. hep-th/9708040. Lett. Cosmology and Composite Structures.H. L. F. Langacker / Nuclear Physics B 586 (2000) 287–302 Acknowledgements We would like to thank Angel Uranga for many communications and suggestions regarding the work presented in the paper as well as for collaboration on related topics. J. B 545 (1999) 47.R. Ya. B 385 (1996) 96. Plümacher. O’Driscoll. JHEP 9807 (1998) 002. Rev. JHEP 9904 (1999) 020. Rev. Cvetiˇc. Uranga. Wang. S. Phys.H. Lazaroiu. College Park. Cvetiˇc. Singapore. Cvetiˇc.). Leigh. hep-th/9606169. D 56 (1997) 3686. Plümacher. Lykken. Moore. Nucl. Plesser. Ibáñez. M. hep-th/9605049. Nucl. G. hep-th/9506098.). We also benefitted from discussions with L. in preparation. M. Langacker. [10] G. Proceedings of Superstrings. B 543 (1999) 105. hep-th/9603167. B 512 (1998) 221. [14] M. Beasley. G. Nucl.H. Bianchi. Aldazabal. hep-th/9806143. hep-ph/9711178. [8] G. B 520 (1998) 75hep-th/9706051. [3] G. Quevedo. hep-th/9704059. P. D 53 (1996) 3399. and J. Taylor. Angelantonj. B. Fiol. Kakushadze. . hep-th/9801114. Phys. [13] G. March 1987. Dixon. B 536 (1999) 29. Kakushadze. Tye. [7] D. J. Luty. Stanev. B 526 (1998) 378.R. hep-th/9806044. The work was supported in part by U. [21] L. 1987. Mohapatra (Eds. Phys. Phys.R. L. D 58 (1998) 106007. [23] A. Zwart.M. [24] G. World Scientific. Poppitz. S.R. Phys. hep-th/0003263. S. A. hep-th/9806008. Aldazabal. [5] Z. Everett. in: S. [11] Z. hep-th/0005067. Cvetiˇc. Department of Energy Grant No. Phys. Wang. Ibanez. Douglas. Nucl. Phys. Phys. B 535 (1998) 311. A. [19] M. 59 (1987) 2829. [22] M. L. hep-th/9802103. W. Phys. Ibáñez. hep-th/9806080. [20] J. Faccioli. [17] M. R. [9] Z. Lett. hep-th/9911021.P. J. L. Pradisi.I. G. Trivedi. B 525 (1998) 3. Berkooz. C. B. Shiu. References [1] C. Langacker. Phys. Phys. Lett.E.I. M. hep-ph/9711178. Nucl. Gates. Wang. Douglas.E. B 483 (1997) 187. Sagnotti. B 566 (2000) 599. Violero.S. G. Rev. P. Lett. JHEP 9807 (1998) 002. C. [6] L. Cvetiˇc. Nucl. Phys.J. [16] M. Rev. Rev.S. Cvetiˇc. Everett. Rev. Phys. hep-th/9804026. A. M. Phys. Nucl. Phys.R. Langacker. Wang. D 58 (1998) 126001. Katz. A. Plümacher. Nucl. Phys. [15] C. hep-th/9903051.L. 59 (1987) 1795.) and in part by the University of Pennsylvania Research Foundation award (M. M. [12] M. Kakushadze. DOE-EY-76-023071 (M. M. Cleaver. Espinosa.E. Nucl. Font. Nucl. hepth/9907186. P. S. E. B 434 (1998) 269.C. unpublished. P.G. and P. Kakushadze. hep-th/9806080. [2] M. [18] M. Shiu. Maryland. Rev. Tye.302 M. Greene.

Introduction Certain noncommutative field theories [1–3] can be systematically derived from open string theories in the presence of constant background NS–NS two-form field (B field) [5–10]. The analysis is given for both bosonic and supersymmetric open strings. For example. linear and logarithmic IR singularities when the D-branes embedded in space–time have the codimension zero. the simplest one-loop annulus diagram reveals a prominent stringy character. All rights reserved. Further considerations of loop effects in noncommutative field theories [11–13] add an intriguing new element in the analogy between open string theories and noncommutative field theories. PII: S 0 5 5 0 . Korea Institute for Advanced Study.re.V.  2000 Elsevier Science B. unlike the generic commutative field theories arising as decoupling limits of string theories.elsevier. The UV regime of the open string annulus amplitudes can naturally be interpreted as the IR closed string degrees of freedom. Kiem). E-mail addresses: ykiem@kias.kr (Y. the investigation of the UV regime of the open string amplitudes shows that certain IR closed string degrees of freedom survive the decoupling limit as previously predicted from the noncommutative field theory analysis. sangmin@kias. the UV/IR mixing.kr (S. In the decoupling limit of Seiberg and Witten. accepted 27 June 2000 Abstract We explicitly evaluate one-loop (annulus) planar and nonplanar open string amplitudes in the presence of the background NS–NS two-form field. From the open string perspective. respectively. 1. This behavior closely parallels the UV/IR mixing of Refs.3 2 1 3 ( 0 0 ) 0 0 4 3 0 . noncommutative field theories are T-duality invariant signaling its stringy nature [4]. The upshot of these developments is that noncommutative field theories are more stringy than what one might naively expect. one and two. Sangmin Lee ∗ School of Physics. 0550-3213/00/$ – see front matter  2000 Elsevier Science B.Nuclear Physics B 586 (2000) 303–314 www. South Korea Received 13 April 2000. All rights reserved. In particular. These degrees of freedom are responsible for the quadratic. the open/closed string channel duality.re. Seoul 130-012.V.2 . [12. namely. we find that the nonplanar string amplitudes reproduce the UV/IR mixing of noncommutative field theories. Lee).nl/locate/npe UV/IR mixing in noncommutative field theory via open string loops Youngjai Kiem.13] ∗ Corresponding author.

we extend our analysis to open superstring amplitudes.1) where 1 We note that our calculations have overlaps with the earlier literature on open string amplitudes. The world-sheet propagator is α0 Xµ (z)Xν (w) = ηµν G(z − w). except including the contributions from the massive string excitations. 2 (2. We discuss further directions and implications suggested by our analysis in Section 5. Their Wilsonian effective action decoded from the annulus amplitudes also turns out to be the same as the one proposed in Ref. such as Refs. First. open string theory calculations reproduce the quadratic divergences [12]. In Section 2. where one can interpret certain UV divergences coming from nonplanar loops of high energy virtual particles as IR divergences. Lee / Nuclear Physics B 586 (2000) 303–314 in noncommutative field theories. Kiem. we embark upon the detailed study of one-loop annulus open string amplitudes in the presence of constant background B field 1 and recover many features found in Refs. respectively. Since our set-up is the string theory framework. where D is the dimension of space–time. we evaluate the planar and nonplanar annulus diagrams in the bosonic open string theory. 2. It was first obtained in [15. massive excitations decouple. couple linearly to the D-brane world-volume open string degrees of freedom and live in the bulk space–time. [12]. In this paper. in the absence of the B field. In Section 3. Upon taking the decoupling limit of [10]. [14] and [15. we can unambiguously identify these extra degrees of freedom as IR closed string contributions. we compute the world-volume propagators on an annulus in the presence of the constant background B field. S. noncommutative quantum field theories arising as limits of open string theories include closed string degrees of freedom. [12] for the extra degrees of freedom responsible for the IR singularities of the one-loop noncommutative field theory amplitudes. consider a rectangular torus whose modulus parameter τ = iT is purely imaginary. This paper is organized as follows. . Via open/closed string duality. which survive the decoupling limit. Here we present an equivalent but more concise form of the propagator following the notations of [17]. The one-loop open string amplitudes turn out to be of the same form as Ref. linear divergences and logarithmic divergences [13] caused by the extra degrees of freedom.16]. In Section 4. In short. which survive the decoupling limit. for (D − 1) D-branes (original critical open string theory). In particular.304 Y.13]. we identify the IR closed string degrees of freedom. while some UV degrees of freedom do not in nonplanar diagrams. World-sheet propagator on an annulus One important ingredient in computing the one-loop string amplitude is the world-sheet propagator on an annulus. and study their properties. [12. via the standard open/closed string duality.16] using a world sheet coordinate in which the two boundaries of the annulus are concentric circles. (D −2)-branes and (D −3)-branes.

Lee / Nuclear Physics B 586 (2000) 303–314 305 Fig. .Y. Kiem. S. 1. The world sheet coordinate for annulus.

.

.

θ1 (ν|iT ) .

2 2π .

+ .

[Im(ν)]2 . G(ν) = − log.

0 θ1 (0|iT ) .

Our next task is to find the explicit form of analogous expressions that are valid when B 6= 0. Suppose for now that we turn on the B12 = B along . Xµ (z)Xν (w) = ηµν G(z − w) + G(z + w) 2 (2. As noted in [18]. z → z + iT . one can bring the B field into a block-diagonal form and consider each 2 × 2 block separately. and they satisfy the flux conversation. m=−∞ q = exp(2πiτ ).2) Here. ω = exp(2πiν). we will not need them for we will consider only open string vertex insertions. However. thereby turning the original torus to an annulus. θ1 is the theta function defined as θ1 (ν|τ ) = −i ∞ X 1 1 2 1 (−1)m q 2 (m+ 2 ) ωm+ 2 . we place a mirror charge at −w¯ (and at all their lattice translation points) for a source charge at w. The propagators are periodic under the two lattice transformations z → z + 1. To turn torus propagators to annulus propagators. This operation imposes Neumann boundary conditions along the two boundaries at Re(z) = 0 and Re(z) = 1/2.3) The propagators with Dirichlet boundary condition can also be straightforwardly written down. the propagators look like α0  ¯ . while maintaining the periodicity in z → z + iT . T (2. the integral of ∂ 2 G(z) over the torus vanishes. Written explicitly.

S. The boundary conditions at Re(z) = 0 and Re(z) = 1/2 should be modified into 2 . Lee / Nuclear Physics B 586 (2000) 303–314 X1 = X and X2 = Y directions parallel to the D-branes under consideration.306 Y. Kiem.

∂n X + iB∂t Y = 0.

6) When |z|. (2. log Y (z)X(w) = α0 θ1 (¯z + w) T 1 + B2 (2.7) 1 + B2 B (y).4) intact. .5) with modified boundary conditions involving a constant term on the right hand sides of (2.10. X(0)Y (1/2 + iy) = i(2πα 0 ) 1 + B2 T The quadratic term in (2.11) distinguishes between the planar and nonplanar vertex insertions. We note that hXY i itself is not periodic. 2 1+B 2T 1 + B 2 B y . (2.9) (2. we have πα 0 B 2 1 G(1/2 + iy) + . The coefficients of the θ -function terms are uniquely determined by comparison with the propagator on the disk. in Refs. (2. For nonplanar insertions. T .11) 1 + B2 T deserves further comments. For later convenience.4) The answer is: B 2 4π 1 − B2 2 [Re(z + w)] ¯ 2. we insert open string vertex operators along the boundaries at Re(z) = 0 and Re(z) = 1/2. Our choice in this paper is to keep the boundary conditions (2. (2.5) (2. the quadratic terms are negligible and θ1 (z|iT ) reduces to z. we note that for the planar insertions. In particular.10) 2B 2 2π [Re(z + w)] ¯ 2.8) X(0)Y (iy) = i(πα 0 ) 1 + B2 where we introduced the Heaviside step function (y) = y/|y|. 3 The periodicity in this context means the periodicity of the physical objects.Re(z)=0. which was obtained. We note that (2. [8. it vanishes for planar insertions along Re(z) = 0 but gives a non-vanishing contribution for nonplanar insertions where one vertex is separated from another. 2 As in [15. such as h∂XY i. |w|  1. [10] thus surviving in the effective noncommutative field theory. We will find that it will give finite contributions to the amplitudes under the decoupling limit of Ref.1/2.16].18]. so that we recover the propagators on a disk. for example.4). X(0)X(1/2 + iy) = α 0 (2. it is possible to trade the quadratic terms of (2. but this does not give an ambiguity when computing physical amplitudes. 3 For the computation of the open string amplitudes. The quadratic terms are required by the periodicity in z → z + iT and flux conservation. the propagators become X(0)X(iy) = α 0 1 G(iy).5). G(z + w) ¯ + X(z)X(w) = G(z − w) + α0 1 + B2 1 + B2 T   4πi ¯ θ1 (z + w) 2 2B + Re(z + w)Im(z ¯ + w) ¯ .

2) 0 The “partition function” part is computed in the same way as in [19]. Lee / Nuclear Physics B 586 (2000) 303–314 307 3. (3. Kiem. S. VT = g α 0 eikX . Schematically. In the following. y1 )V2 (−p. e Z(T ) = (2π)n (3. Z  (3. 3. y. One-loop bosonic open string amplitudes Utilizing the world-sheet propagators of Section 2. T ). (3. but the dimensions and dependence on coupling constant will be unambiguous.1) 0 dT (2πα 0 T )−n/2 f1 (q)−24T T 0 ZT dy I (p. y2 ) T (3. A simple dimensional analysis shows that g is related to string coupling by g 2 = (α 0 )n/2−2 gst . Planar and nonplanar bosonic open string amplitudes on an annulus The one-loop amplitude in the presence of an (n − 1)-brane is given by Z∞ A= dT Z(T ) T 0 ZT ZT dy1 0 2 0 Z∞ =g α dy2 V1 (p.4) m=1 The vertex operators for a tachyon and a gauge boson are. we explicitly evaluate the annulus amplitudes.Y.5) where the coupling constant g is the one that appears in the low energy effective action of open strings. respectively. q = e−2πT . we will ignore all numerical constants in the overall normalization of the amplitudes. for the tachyon– tachyon insertion.1.3) I f1 (q) = q 1/24 ∞ Y (1 − q m ). inserting two open string vertex operators along the world-sheet boundaries in the presence of D-branes. When B = 0. we have . Z dn k X −2πα 0 T (k 2 +M 2 ) I . √ VA = gε · ∂XeikX . We first give the answers for the planar amplitudes. (3.6) S = − dn x (∂φ)2 + m2 φ 2 + g 2 φ 4 + · · · .

.

  .

θ1 (iy|iT ) .

−2 .

exp 2π y 2 . .

(3.7) I =.

0 θ1 (0|iT ) .

8) . we have 0 2  I = 1 · 2 J 00 + α 0 (ε1 · p)(ε2 · p)(J 0 )2 eα p J  0 2 = −α 0 (ε1 · ε2 )p2 − (ε1 · p)(ε2 · p) (J 0 )2 eα p J + (total derivative in y). T while for gauge boson–gauge boson insertion. (3.

Kiem. the answer is exactly the same as the one for B = 0. We now consider the nonplanar insertions where B 6= 0 effect is conspicuous.9) Gµν ≡ ηµν − Bη−1 B µν . For the tachyon–tachyon insertions. we get . [10] as  (3. except that the external momentum squared is evaluated with respect to the open string metric Gµν defined in Ref. When we turn on the B field (B 6= 0). Lee / Nuclear Physics B 586 (2000) 303–314 where we define J (y) = G(iy).308 Y. S.

.

    .

θ2 (iy|iT ) .

−2 .

exp 2π y 2 exp − p ◦ p . (3.10) I = .

.

0 θ1 (0|iT ) .

2. [12] as 1  × p = µ Θ µν pν . The B dependence in the amplitudes come from two combinations p ◦ p and  × p. p ◦ p ≡ − pµ (ΘGΘ)µν pν . there are polarization dependences. For the higher spin world-volume fields. as exemplified in  × p for the gauge boson amplitudes. The relation between the nonplanar tachyon–tachyon amplitude Z∞ A= 0 dT (2πα 0 T )−n/2 f1 (q)−24 T T ZT 0 . (3. 3. we only consider the tachyon amplitudes in detail for simplicity. × exp α 0 p2 K − 2πα 0 T (3.11) where we define K(y) = G(1/2 + iy). the effect of noncommutativity becomes stronger as we approach the UV corner of the moduli integral. 4 where Θ is the noncommutativity parameter defined in [10]. We introduced ◦-product and ×-product following Ref. We first review the well-known world-sheet duality for B = 0 to contrast it with the B 6= 0 situation.  µν Θ µν ≡ −2πα 0 (η + B)−1 B(η − B)−1 .13) The sign in the definition of p ◦ p is introduced to make it nonnegative.12) (3. Open/closed string duality and the decoupling limit For the rest of this section. T 2πα 0 T and the gauge boson–gauge boson insertions yield   0 2 (ε1 × p)(ε2 × p)  0 2 I = − α (ε1 · ε2 )p − (ε1 · p)(ε2 · p) (K ) + α0 T 2   p◦p . Due to the prefactor 1/T in front of p ◦ p.

.

  .

θ2 (iy|iT ) .

−2 2π 2 .

.

y dy .

0 exp θ1 (0|iT ) .

when y  T . This intuitive picture is confirmed by an explicit calculation which shows that . the open string diagram looks very much like the field theory diagram in Fig. For example. 2(a).14) in e−2πT .14) and the one-loop amplitudes in the field theory of the type (3. T (3. where we can expand (3.6) is manifest in the region T  1.

Y. Lee / Nuclear Physics B 586 (2000) 303–314 A = g2 α0 X I Z aI X dT 0 2 (2πα 0 T )−n/2 T e−2πα T MI = T Z I dn k 309 aI g 2 . k 2 + MI2 (3. In particular.15) where aI are some numerical coefficients. the usual channel duality allows us to rewrite (3. In the opposite end T  1.14) from the point of view of the closed strings. using the modular transformation of the theta functions. Kiem. we find Z∞ A= 0 dS n/2−12 S f1 (q) ˜ −18 S Z1 . S.

.

−2 dx .

θ4 (x|iS).

16) can be expanded to give X bJ κ 2 X Z 0 2 2 bJ dS e−2πSα (p +MJ )/4 ∼ (3. (3. 2. The picture now is a closed string connecting the open string states as in Fig. where χ is a closed string field. For the case of the space–time filling 25-brane. the mass spectrum is known to be α 0 MI2 = NI − 1 (open) and α 0 MJ2 = 4(NJ − 1) (closed) for nonnegative Fig.16) 0 where S = 1/T and q˜ = exp(−2πS).18) S = dn x κχφ. . the external momenta are always put on-shell. We note that in the bosonic string theory.17) A= p2 + MJ2 J J for some numerical constants bJ . . 2(b). 4 When computing perturbative string amplitudes. Open/closed string channel duality. Here we are assuming that the final expression holds for off-shell amplitudes as well. (3. The noncommutative field theory arises in the decoupling limit α 0 → 0 while keeping µν G and Θ µν fixed [10]. 4 The coupling constant κ appears in the low energy effective action of the form Z (3.

When 2πα 0 T goes below the UV cutoff. S.17). the massless intermediate degrees of freedom give a trivial IR divergence that should be cancelled with other divergences via Fischler–Susskind type mechanism. Kiem. 3. . as can be seen from (3. (3. When B = 0. As shown in Fig.310 Y. therefore. Domain of moduli integral. we need to take the Wilsonian point of view regarding the cutoff and the effective degrees of freedom. For this purpose. When B 6= 0. (3.3. the regulated open string contribution comes from the region of the moduli space where 2πα 0 T > 1/Λ2 . we explicitly introduce a short distance UV regulator 1/Λ2 in the open string description. Eq. 5 In [23]. 3.19) e S 0 where the IR regulated closed string amplitude A(Λ) is defined as Fig. 3. In this limit. The contribution to the nonplanar amplitude from these IR closed string degrees of freedom can be computed as Z∞ AIR = A(∞) − A(Λ) ' dS n/2−12 −p◦pS/(2πα 0) 2 0  S − e−(p◦p+1/Λ )S/(2πα ) . What survives the decoupling limit The key issue is to identify the contributions from the T  1 UV regime to the open string moduli integral when B 6= 0. In this channel. the same amplitude was computed using the open–closed string field theory for B = 0. in the spirit of string field theory [20–22]. all but the contribution from massless intermediate states disappear. we have a factorization channel where the original nonplanar annulus diagram becomes two string states connected by a long closed string tube. 5 Via open/closed string channel duality. the open string UV cutoff 1/Λ2 transforms to the closed string IR cutoff Λ2 . 3. the open string UV regime gets mapped to the closed string IR regime S/(2πα 0 ) > Λ2 . Then. if we ignore the tachyons. as depicted in Fig. Lee / Nuclear Physics B 586 (2000) 303–314 integers NI and NJ . the possible extra UV degrees of freedom originating from the extreme UV open string loops should come from the corner of moduli space where 0 < 2πα 0 T < 1/Λ2 . their contribution is non-trivial as we will see shortly. it is natural to investigate 0 < 2πα 0 T < 1/Λ2 corner of the open string moduli space in terms of the closed string picture.16). Division of the moduli space into two connected parts is inherent in their formalism. We thus resort to the nonplanar amplitude expression in the closed string channel.

From our derivation. In (3. (3.19).21) is nothing but the propagator of the extra degree of freedom multiplied by the coupling constant.19). we wrote down only the part of the amplitude that survives the decoupling limit and we deleted the tachyonic intermediate contribution. respectively. (3. Reinstating the tachyonic contribution would produce the negative eigenvalue for the quadratic effective 2 in (3. it is clear that χ field with the effective action (3.21) = AIR = κ 2 (α 0 )2 2 p ◦ p p ◦ p + 1/Λ p ◦ p + Λ2 (p ◦ p)2 from (3.22) gives the effective description of the “long tube” IR closed strings at low energies. For the codimension one 24-brane. (3. we repeat the calculations of the previous section for the case of open superstrings. The main finding that the space–time supersymmetry makes the two point . The extra dimensions found in Ref.Y. From the “long tube” IR closed string picture of Fig. 4. it is clear that they also represent the IR closed string degrees of freedom. Lee / Nuclear Physics B 586 (2000) 303–314 Z∞ A(Λ) ' dS n/2−12 −(p◦p+1/Λ2)S/(2πα 0) S e . we get   AIR = κ 2 log(p ◦ p) − log(p ◦ p + 1/Λ2 ) . where φ is a generic world-volume open string scalar field.24) are the same as the 1PI amplitudes found in Ref. at least when the noncommutative field theory under consideration derives from the decoupling limit of open string theory. [13] for the extra low energy degrees of freedom.19).20) is the IR regulated amplitude where we restrict the closed string moduli integral to the distance up to the IR cutoff scale Λ2 . From the low energy effective description point of view. 2. [13] are indeed space–time dimensions transversal to the brane. For the space–time filling 25-brane (n = 26).23) and (3.23) AIR = √ p◦p p ◦ p + 1/Λ2 while for the codimension two 23-brane. we find that (3. The A(Λ) defined in (3.22) dx ∂χ ◦ ∂χ + Λ (∂ ◦ ∂χ) + dn x κχφ. Kiem. indicating action for the small value of p ◦p by shifting p ◦p into p ◦p +Mtachyon the tachyonic instability [12].20) 0 explicitly introducing the cut off Λ2 . The effective action (3. they live in the bulk space–time while the open string degrees of freedom are confined on a codimension one and two D-brane. S.24) We note that (3. From our derivation. S 311 (3.22) is identical to the one found in the noncommutative field theory one-loop analysis [12]. we have   1 1 κ 2 (α 0 )2 − (3. the extra degree of freedom (denoted as χ field) then has the effective Lagrangian of the form Z Z   26 2 2 (3. One-loop open superstring amplitudes In this section.19) yields  2 0  κ 2 α0 κ α −p .

11). its contribution to amplitudes is from the corner of the open string moduli space where the dual closed string interpretation is appropriate. This behavior. Lee / Nuclear Physics B 586 (2000) 303–314 amplitudes vanish regardless of the value of B is consistent with Ref. Therefore. In our present context with parallel D-branes. The answer for the amplitude is given by 8 ZT  Z∞ 4 X dT a 0 −n/2 fa (q) (2πα T ) (−1) T dy V1 (p. [19]. might seem rather puzzling from the open string theory point of view. y) = εµ (∂Xµ + ip · ψψ µ )eipX . 0) . Naturally. a familiar AdS/CFT correspondence [27–29]. the amplitude AIR becomes a simple divergence involving the cutoff Λ2 without the momentum dependence. it implies that there is a term in the mode expansion of X. [26] where the thermodynamic evidence for the “winding states” in noncommutative field theories is given. in this context. [26] do not find any wrapping states. while we are apparently considering open strings. via the Fischler–Susskind mechanism. we have the contraction of four world-sheet fermions.11) should be present in the world-sheet propagators as a consequence of the boundary conditions (2. Kiem. which is linearly proportional to Re(z). A= T f1 (q) a=2 0 (4. the two point function completely vanishes due to the Jacobi’s fundamental formulae [25]. For closed strings. see Ref. however. the dual supergravity background geometries of commutative field theories have the asymptotic isometry group isomorphic to the conformal group. which would correspond to higher extended objects than strings.312 Y. In Ref. Its contribution to nonplanar amplitudes is the crucial exponential factor exp(−(p ◦ p)/(2πα 0 T )) necessary for the emergence of IR closed string degrees of freedom. (3. In fact. The vertex operators for massless gauge bosons are V (p.8) and (3. In realistic models. [24]. y)V2 (−p. one attributes this property to the space–time supersymmetry. ultimately resulting the vanishing beta function. 5. In fact.2) In addition to the terms in Eqs. however. Discussions The term (2. only the terms with eight or more world-sheet fermions give nonzero contribution. We note that the authors of Ref. For the commutative field theories. we clearly have sixteen supercharges. the “winding states” of Ref. Its existence. As was shown from the calculations in Section 3. S.4). (4. [24].1) 0 where the index a labels spin structures. [26] represent nothing but closed string states. the direction Re(z) is the spatial direction along which the open string lies. the noncommutative IR singularities do not show up. however. we consider the perturbations around a conformal . After the summation over spin structures. it was argued that for the supersymmetric gauge theories with sixteen supercharges. is consistent with Ref. From this point of view. For the definitions of fa (q). the cutoff dependence in open string channel gets cancelled by closed string sigma model divergence. this would usually signal the presence of a non-trivial winding state.

hep-th/9711165. Iowa. 1994. Note added After the completion of the first version of this paper. Reinstalling the part that vanishes in on-shell. the conjectured dual supergravity backgrounds of the noncommutative field theories do not have the asymptotic isometry group isomorphic to the conformal group [30. would like to thank Chang-Yeong Lee for useful correspondence.31].42)–(2. especially for bringing Ref. which has some overlap with the material presented in Section 2 and Section 3. hep-th/9711162. (2. High Energy Phys. Contemp. as shown from the analysis in this paper.K. References [1] A. Naturally. By turning the viewpoint around. [32] should reveal interesting physics. Oper. the disentangling of higher loop diagrams in noncommutative field theories via open string perturbation theory should be an exciting venue. C. Eqs. we need variable changes 2πy = log |ρ/ρ 0 | and 2πT = − log k. After reading [34]. Acknowledgements We are grateful to Seungjoon Hyun. Academic Press. J. Math. J. we found some calculational errors in Section 3. the detailed investigation of the Fischler–Susskind mechanism in the noncommutative context along the line of Ref.20]. The emergence of the non-trivial IR closed strings in noncommutative field theories appears to be related to the nonconformality of the worldvolume theory.R. (3. To explicitly see this. Rieffel. 62 (1987) 237.11). 9802 (1998) 008. Kiem. especially in relation to string field theory [10. in Eqs.R. We emphasize that our boundary propagator expressions. Douglas.45) in [34] obtained by using the boundary state formalism. [34] appeared. Iowa City.10). 9802 (1998) 003. [34]. Kimyeong Lee. The noncommutative field theory calculations mimic the open string calculations to a remarkable degree. [4] M. [2] A. [26] to our attention. While this paper was being written. Ref. As noted in [34]. in: Operator Algebras and Mathematical Physics. Ref. Connes. S. are identical to Eqs. (3. Math. Noncommutative Geometry. M. Alg. we find that our gauge boson amplitudes are identical to the ones given in Ref. The corrected calculation revealed the (1 × p)(2 × p) part in Eq. Schwarz. some on-shell string calculations can be extended to offshell in field theory limits. Phys. (2. In this spirit. Connes. the [(1 · 2 )p2 − (1 · p)(2 · p)] part. . Hull. Connes.1.7)–(2.11). A. Hyeonjoon Shin and Piljin Yi for useful discussions. On the other hand. [33] came up.Y. one might consider using the noncommutative field theory as a useful guide that provides us with a systematic organization tool for the open/closed string loop diagrams.8) and (3. Y. Douglas. [3] A. Lee / Nuclear Physics B 586 (2000) 303–314 313 fixed point. 1985. M.1 for the gauge boson amplitudes in the original version of our paper. High Energy Phys.

Nucl. S.G. Ho. R. S. [18] S. Lett.M. Rev. hep-th/9802150. Chu. S. Kugo. [27] J. Witten. Takahashi. A.M. Phys.R. H. hep-th/9611050. hep-th/0003180. hep-th/9905043. B 280 (1987) 599. Verlinde. hep-th/0002186. hep-th/9909059. Van Raamsdonk. 1927. [32] D. J. . Phys. Polchinski. hep-th/9711200. Matusis. B 508 (1997) 107. Chu. Phys. A Course of Modern Analysis. [21] B. Russo. 467.-Y. 9909 (1999) 032. B 450 (1999) 119. A. Theor.R. J. Asakawa. hep-th/9712199. Cheung. B 550 (1999) 151. P. hep-th/9908142. H. 02 (1999) 016. 102 (1999) 427. Leigh. Callan. T. Gorbatov. hep-th/9810072. hep-th/0003113. Adv. Phys.A. T. [34] A. C. [9] F. Nucl. Van Raamsdonk. Lee. Phys. 2 (1998) 253. Nucl. Witten. [14] E. [22] J. N. [13] M. 9906 (1999) 030. hep-th/9810179. Math. hep-th/9906192. Phys. Gubser. M. Seiberg.-J. [29] E. [8] V.G. [16] A.N. Berenstein.S. Kashani-Poor. Susskind. B 428 (1998) 105. G. Whittaker. Nappi. Adv.-S. hep-th/9907166. N. N. Dorn. I. Polyakov. Phys. [23] T. High Energy Phys. 9909 (1999) 025. Andreev. B 528 (1998) 185. Phys. Rey. Nucl. Theor. High Energy Phys. Hyun. [33] O. M. B 425 (1998) 48.S. Fischler. D 60 (1999) 106002. Gomis. J. Lett. Phys. Tseytlin. Zwiebach. Bilal. hep-th/9803031. Lovelace. Nappi. [19] J. Klebanov. M. Maldacena. P. E. B 163 (1985) 123. R. [11] T.A. Math. B 268 (1986) 253. Phys. hep-th/9803067.R. J. Lee. hep-th/9812219. Cambridge Univ. C. hep-th/9706115. Lett. Pouliot. B 390 (1993) 33. hep-th/9912072. [30] A. Lett. H. Ardalan. Lett. 2 (1998) 231. Minwalla. Yost. Kiem. [15] C. Watson. B 288 (1987) 525. Y. Schomerus. Lee. [20] E. High Energy Phys. A. Witten. Lett. Arfaei. E.G.E. Maldacena. [26] W.-K. Sheikh-Jabbari. p. Press. [25] E. Phys. Itzhaki. J. [10] N. Hashimoto.M. Nucl. Phys. B 569 (2000) 262. hep-th/9908134. hep-th/0001056. [24] A. S. Theor. Krogh.A. Paban. L. hep-th/0002067. [7] C. B 465 (1999) 142. hep-th/9904104. Filk. Sheikh-Jabbari. S. Nucl. hep-th/9802109. Nucl. C. Lee / Nuclear Physics B 586 (2000) 303–314 [5] M. Toumbas. Prog. Seiberg. Phys. Phys.M.-M. Fradkin. Khoury. B 376 (1996) 53. hep-th/9903205. [6] Y. C. S. C. hep-th/9906161. Nucl.-S. [17] S.T. Seiberg. [28] S. Callan. [12] S.314 Y. Abouelsaood. [31] J. Kiem. Phys.G. Yost. Phys. Phys. High Energy Phys. J. Russo. hep-th/0002075. C. N. Phys.

Alexei Nurmagambetov a. can be considered as a specific compactification of a single D = 11 M5-brane down to lower dimensions (with or without its subsequent T-dualization). Introduction The M5-brane plays an important role in studying properties of M-theory [1.ua 2 dmitri.elsevier.+y Keywords: Superbranes. Marzolo. Self-dual gauge fields. NSC Kharkov Institute of Physics and Technology. PACS: 11. realized to be relevant to a brane description of non-Abelian gauge theories [3. A direct dimensional reduction of D = 11 space–time with an M5-brane down to tendimensional space–time produces a so-called NS5-brane of type IIA supergravity which has been intensively studied in relation to six-dimensional gauge theories [6] and “little string theories” [7]. Ukraine b INFN. Kharkov.it.9 . UA-61108.V. It has also been shown [11.infn.-q. 11. All rights reserved. NSC Kharkov Institute of Physics and Technology. Kharkov.kharkov. Dmitri Sorokin b. Via F. 61108. revised 6 June 2000. Dimensional reduction 1. Italia Received 6 April 2000. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. 8. 3 The double dimensional reduction of the M5-brane action [8–10] is well known to result in a type IIA D = 10 Dirichlet 4-brane. many physically important multibrane configurations.ua 1 bandos@kipt. 35131 Padova. the theory of strings and associated field theories. For instance. On leave from Institute for Theoretical Physics. accepted 19 June 2000 Abstract The kappa-invariant worldvolume action for the NS5-brane in a D = 10 type IIA supergravity background is obtained by carrying out the dimensional reduction of the M5-brane action. Sezione di Padova.17. PII: S 0 5 5 0 .15. 3 ∗ Corresponding author.2 a Institute for Theoretical Physics.12] how by reducing the M5-brane action one may arrive at a duality-symmetric D3-brane action.2]. Akademicheskaya 1.∗ .4] and a brane-world scenario [5]. E-mail: ajn@kipt.Nuclear Physics B 586 (2000) 315–330 www.1 .  2000 Elsevier Science B.3 2 1 3 ( 0 0 ) 0 0 3 9 8 .V.nl/locate/npe The type IIA NS5-brane Igor Bandos a.sorokin@pd. All rights reserved. Ukraine.kharkov.

finding a mechanism of anomaly cancellation in the presence of M5-branes.and NS5-branes and in Section 6 we describe the full target-superspace covariant and κ-invariant NS5-brane action. / Nuclear Physics B 586 (2000) 315–330 The verification of the quantum consistency of M-theory requires. 4 In contrast to [15. The fact that the NS5-brane can be regarded as an M5-brane propagating in a dimensionally reduced D = 11 supergravity background substantially simplifies the analysis of the NS5-brane model. in particular. Mechanisms for the M5-brane anomaly cancellation proposed recently in [15. therefore.16] include as an important feature the reduction of the structure group SO(5) of the normal bundle of the M5-brane down to its SO(4) subgroup. the dimensional reduction is a particular case of such an “M5-brane framing” [15. The bosonic sector consists of three degrees of freedom corresponding to the two-rank self-dual worldvolume field and five worldvolume scalars. . In the case of the M5brane the five scalar fields describe its oscillations in a D = 11 background in the directions transversal to the M5-brane worldvolume.316 I. It has been shown that the anomaly problem has a natural solution in the case of D = 10 NS5-branes [13. In Sections 2–5 we consider bosonic M5. and thus filling in a gap in the list of worldvolume actions for supersymmetric extended objects found in string theory. and the fifth scalar field (corresponding to the compactified dimension of the D = 11 space) “decouples” and becomes a “purely” worldvolume field. For both five-branes eight fermionic fields can be associated with brane “oscillations” in Grassmann directions of corresponding target superspaces. We are thankful to Jeff Harvey for clarifying this difference in the approaches.20].14]. in particular. The aim of this paper is to get a full. while in the case of the NS5-brane four scalar fields correspond to transversal oscillations in a D = 10 background. nonlinear and κ-symmetric. 4 These facts provide us with a motivation to study in more detail the dynamical and symmetry properties of the NSIIA five-brane by constructing a full worldvolume action describing its dynamics in a type IIA D = 10 supergravity background. Such a reduction implies an existence of a covariantly constant vector field and. To get the action describing the dynamics of the physical modes of the NS5-brane as a dimensionally reduced M5-brane action we first briefly remind the structure and properties of the latter.16]). Bandos et al. while in the case of the D = 11 M5-brane the situation is much more subtle and requires additional study [13–18]. This results in the abovementioned reduction of the M5-brane normal bundle structure group SO(5) down to SO(4). [17] is based on the assumption that a full understanding of anomaly cancellation requires keeping the full SO(5). For instance. the physical field content of the IIA D = 10 NS5-brane is the same as of the M5-brane. NS5-brane action in a curved IIA D = 10 target superspace by carrying out the direct dimensional reduction of the D = 11 M5-brane action [8–10]. allowing one to derive its symmetries and dynamical properties directly from those of the M5-brane.16] the analysis of Ref. By now the action for the NS5-brane has been constructed up to a second order in the field strength of a two-rank self-dual worldvolume gauge field of the five-brane and only in a background of the bosonic sector of IIA D = 10 supergravity [19. looks very much as a dimensional reduction (to be more precise.

n. In addition to the usual gauge symmetry of the b2 field δa(ξ ) = 0. (5) the action (1) is invariant under the following transformations [8–10] δa(ξ ) = 0. The M5-brane action In the absence of interactions with antisymmetric tensor fields of D = 11 supergravity the action for the bosonic sector of the M5-brane has the following form [8] "q # p Z  −gˆ 6 ∗mn r ∗ ˆ ˆ H − det gˆmn + iHmn + q Hmnr ∂ a .I. are vector indices of D = 11 target space coordinates Xˆ mˆ . . . . . − det gˆmn + iHˆ mn ∗ . 3! −gˆ (3) a(ξ ) is an auxiliary scalar field ensuring the covariance of the model. δa = ϕ(ξ ). H −gˆ δ Hˆ ∗mn LDBI ≡ q  ∗ . (1) S =− d ξ [ 4 −∂a∂a where m. = 0. b∗mn defined in (8) reduces to Hˆ mn Note that at the linearized level. δbmn = 2φ[m (ξ )∂n] a(ξ ). Bandos et al. . . ˆ n. In what follows the “hat” over quantities indicates that they correspond to or induced by the elevendimensional theory. . 10 m. / Nuclear Physics B 586 (2000) 315–330 317 2. . . . 5 are vector indices of d = 6 worldvolume coordinates ξ m . δbmn = 2∂[m ϕn] (ξ ). H (6) (7) (8) . . δbmn = q [ [ −∂a∂a −∂a∂a where b∗mn = − p2 δLDBI . Hˆ mn [ −∂a∂a Hˆ ∗mnl = 1 p  mnlrsq Hrsq . . = 0. ˆ . Hmnl (ξ ) = 3∂[m bnl] is the field strength of the worldvolume antisymmetric tensor field bmn (ξ ). " # δa ∂s a ∗ ps b Hmn − Hmnp gˆ q . 1 ∗ ∗ Hˆ mnr ≡q ∂ r a. (11) gˆmn = ∂m Xˆ mˆ gˆmˆ nˆ ∂n Xˆ nˆ (2) is the worldvolume metric induced by embedding the five-brane into a D = 11 gravity (11) background with a metric gˆmˆ nˆ (X) (we use the “almost minus” Minkowski signature (+ − · · · −)). . and [ ≡ ∂m a gˆ mn ∂n a ∂a∂a (4) denotes the scalar product of the d = 6 vector ∂m a with respect to the metric (2).

ˆ ∗ivˆ ∗ H3 = H3 ∧ v. Bandos et al. vˆk ≡ q ∂k a . (11) 3! 2 The variation of the first term in (1) with respect to the gauge field and the scalar a(ξ ) can be written in terms of differential forms as Z Z Z q  6 6 ∗ b∗ ∧ ∗δ Hˆ ∗ . ˆ H (12) dξ δLDBI ≡ dξ δ − det gˆmn + iHmn = 2 2 H3 ≡ M6 b∗ and Hˆ ∗ are constructed respectively from the tensors (8) and (3) where 2-forms H 2 2 b∗nm . ˆ (14) vˆ ∧ H3 ∧ ivˆ δH3 = vˆ ∧ ivˆ H3 ∧ δH3 + H3 ∧ δH3 . To this end note that the second term in the action (1) can be written in terms of differential forms Z Z p 1 Hˆ ∗mn Hmnr ∂ r a d6 ξ L1 ≡ − d6 ξ −gˆ q [ 4 −∂a∂a Z 1 vˆ ∧ H3 ∧ ivˆ H3 . / Nuclear Physics B 586 (2000) 315–330 The symmetries (6) and (7) are characteristic of the covariant approach [21] to the Lagrangian description of duality-symmetric fields. 2!4! ∗Ω4 = + √ 1 1 1 dξ m2 ∧ dξ m1 −g m1 m2 n1 ···n4 Ω n1 ···n4 ≡ dξm2 ∧ dξm1 √  m1 m2 n1 ···n4 Ωn1 ···n4 . Ω6 ≡ −ivˆ Ω6 ∧ v. ivˆ H3 ≡ dξ m ∧ dξ n vˆk gˆ kl Hlnm . 6 To have ∗∗ = I we define ∗Ω2 = − √ 1 dξ m4 ∧ · · · ∧ dξ m1 −gm1 ···m4 n1 n2 Ω n1 n2 . 2!4! 2!4! −g . They ensure the b2 field equation of motion to reduce to a self-duality condition. as well as the connection with non-covariant formulations [22–31] Let us briefly describe how one derives the symmetries (6) and (7) and gets the selfduality condition [8. 6 Using the identities ivˆ δ vˆ = 0.318 I. b∗ ≡ 1 dξ m ∧ dξ n H H 2 2 1 ∗ Hˆ 2∗ = ivˆ ∗H3 ≡ dξ m ∧ dξ n Hˆ nm 2 (13) and ∗ is the Hodge operation in d = 6 dimensions. (15) vˆ ∧ H3 ∧ iδ vˆ H3 = δ vˆ ∧ ∗H3 ∧ ivˆ ∗ H3 = −δ vˆ ∧ vˆ ∧ ivˆ ∗H3 ∧ ivˆ ∗H3 (16) 5 In our notation dξ m1 ∧ · · · ∧ dξ m6 = d6 ξ  m1 ···m6 . (10) [ −∂a∂a 1 m 1 dξ ∧ dξ n ∧ dξ l Hlnm .21]. (9) =− 2 M6 where 5 vˆ = dξ m vˆm .

H [ −∂a∂a (24) To prove this note that Hˆ 2∗ is invariant under the transformations (6) q [ vˆ ∧ φ1 δb2 = da ∧ φ1 ≡ −∂a∂a q   [ ivˆ ∗(vˆ ∧ dφ1 ) ≡ 0. the second-order equation (22) reduces to the first-order selfduality condition  1 b∗ − ivˆ H3 = 0. (20) [ −∂a∂a Since H3 = db2 . ⇒ δ Hˆ 2∗ ≡ δivˆ ∗Hˆ 3 = −∂a∂a 7 We use conventions where external derivative acts from the right: dΩq = 1 mq ∧ · · · ∧ dξ m1 ∧ dξ n ∂n Ωm1 ···mq . .I. From (21) it also follows that the equation of motion of b2 field is d(da ∧ F2 ) = 0. / Nuclear Physics B 586 (2000) 315–330 319 and b∗ = vˆ ∧ Hˆ ∗ ∧ Hˆ ∗ b∗ ∧ H vˆ ∧ H 2 2 2 2 ⇐⇒ abcdef ˆ ∗ ˆ ∗ b∗ H b∗  abcdef H Hbc Hde vˆf . the variation (18) can be written (up to a total derivative) in the following form 7 Z Z d(da ∧ F2 ) ∧ (δb2 − δaF2 ). H 2 [ −∂a∂a 1 or Fmn ≡ q 1 [ −∂a∂a b∗mn − Hmnl gˆ lk q H (19) ! ∂k a . (22) and the equation of motion of a(x) is a consequence of Eq. (18) dξ δL ≡ − 2 2 M6 where F2 ≡ q  b∗ − ivˆ H3 = 1 dξ m ∧ dξ n Fnm . using the symmetry (6). Bandos et al. bc de vˆ f =  (17) one can rewrite the variation of the Lagrangian (1) in the form  Z  Z 1 1 6 H3 ∧ δH3 − da ∧ F2 ∧ δH3 − dδa ∧ da ∧ F2 ∧ F2 . It can be shown [21] that. (22). (21) dξ 6 δL ≡ − M6 from which the invariance of the action under (6) and (7) becomes evident. H (23) F2 ≡ q [ −∂a∂a or in components b∗mn = Hmnl gˆ lk q ∂k a . dξ q! d(Ωp ∧ Ωq ) = Ωp ∧ dΩq + (−)q dΩp ∧ Ωq .

δF2 = − q [ −∂a∂a Eq. Φ(X) is the dilaton field and Am (X) is the Abelian vector gauge field of D = 10 IIA supergravity. (25) is simplified when one takes into account that ivˆ da = Then (25) q q [ ivˆ vˆ = − −∂a∂a. The U (1)-gauge transformations of Am (X) and y are δAm (X) = ∂m ϕ (0)(X). the worldvolume of the p-brane being not compactified. E aˆ = E a . E 10 ≡ dXˆ mˆ Emˆ aˆ Xˆ . Eq. the transformations of the two-form (19) reduce to 1 δivˆ H3 = −ivˆ (vˆ ∧ dφ1 ). Bandos et al. ∗ = H ∗ gˆ lk √ ∂k a b∗mn is defined in (8) and reduces to Hˆ mn at the Remember that H mnl [ −∂a∂a ∗ =H linearized level. Xˆ mˆ = Xm . A standard (string frame) ansatz for the target-space vielbein under the Kaluza–Klein reduction of one spatial dimension has the following form    y = Xˆ 10 . and the reduction means that the background fields. (26) is similar to the general solution of Eq. 32–39]. (27) emˆ = 2 0 e3Φ where y is the coordinate compactified into a torus. Am (X) and Φ(X) of the D = 10 metric gmn !  2 4 (10) e− 3 Φ gmn − e2Φ Am An e 3 Φ Am (11) gˆmˆ nˆ = (29) 4 4 e 3 Φ An −e 3 Φ . (28) This ansatz leads to the following expression for the D = 11 target space metric in terms (10) (X) = em a ema . (22) for da ∧ F2 . and δ(da ∧ F2 ) = −da ∧ dφ1 . (24) becoming the conventional self-duality condition Hˆ lmn lmn . This means that the general solution of Eq. ! 1 2 e− 3 Φ em a −e 3 Φ Am aˆ . such as components of (27). Further details on the classical dynamics of the M5-brane the reader may find in [8–10. (23) appears just as a result of such gauge fixing. (22) can be gauged to zero with the use of the symmetry (6). / Nuclear Physics B 586 (2000) 315–330 Hence. E a = e− 3 Φ dXm em (X). y . δy = ϕ (0) (X).  1 2 2 a E 10 = e 3 Φ dy − dXm Am ≡ e 3 Φ F . do not depend on y which is now considered as an intrinsic scalar field in the 5-brane worldvolume. [ −∂a∂a ˆ δF2 = −dφ1 + ivˆ dφ1 ∧ v. (26) We now observe that Eq. 3. and Eq. Dimensional reduction of D = 11 gravity and the NS5-brane action The procedure of the direct dimensional reduction assumes a compactification of some of target-space spatial dimensions (one in our case).320 I.

gmn = ∂m Xm gmn m = 0. . . to the following form of the six-dimensional induced metric (2)  2 gˆmn = e− 3 Φ gmn − e2Φ Fm Fn . consequently. H ∗mnl = √ (34) Hˆ ∗mnp = gˆ 3! −g and the M5-brane field strength Hˆ ∗mn (3) is related to its NS5-brane counterpart H ∗mn as r s 1 g (∂a)2 ∗mn ∂k a  kmnpqr Hpqr p . and to “rescale” worldvolume fields and their scalar products with respect to the worldvolume induced metric (31). To present the explicit form of the NS5-brane action we should rewrite all its constituents in terms of D = 10 fields. (33) gˆ mn = e 3 Φ g mn + 1 − e2Φ F 2 The NS5-brane action follows from the M5-action (1) with the background metric having a particular form (29) and the coordinate Xˆ 10 = y(ξ ) being considered as an intrinsic worldvolume scalar field. (∂a)2 (36) (37) In view of Eqs.I. . . H ∗mn = √ . the Hodge duality (3) is now redefined with respect to the metric (31) r 1 g ∗mnp H  mnlrsd Hrsd . . (30). . / Nuclear Physics B 586 (2000) 315–330 and. (36) and (37) the antisymmetric tensor entering the DBI-like part of the M5-brane action is reexpressed in terms of H ∗lm as follows p 1 ∗ = gˆml gˆ nk Hˆ ∗lk = gˆml gˆ nk g/gˆ e− 3 Φ N −1 H ∗lk . (32) where Am = ∂m Xm Am (X) is the worldvolume pullback of Am (X) and Fm is the pullback of the one-form F introduced in (27). It is invariant under the U (1) gauge transformations (28). with N standing for s N≡ 1 e2Φ (F ∂a)2 = e− 3 Φ 1+ (∂a)2 (1 − e2Φ F 2 ) s [ ∂a∂a . 9 (31) is the six-dimensional metric induced by embedding the 5-brane worldvolume into the ten-dimensional curved space–time and Fm = ∂m y − Am . For instance. (35). Hˆ mn  2 (38) gˆml = e− 3 Φ gml − e2Φ Fm Fl . In what follows we will also use an expression for the inverse worldvolume metric   2 e2Φ F m F n . Bandos et al. ∂a∂a 2 (∂a)2 ≡ ∂l ag lm ∂m a. (35) H Hˆ ∗mn = gˆ ∂a∂a 3! −g [ −(∂a)2 where the scalar product (4) has also been correspondingly redefined as [ ≡ ∂k a gˆ ks ∂s a = e 3 Φ N 2 (∂a)2. 321 (30) In (30) (10) (X)∂n Xn . Fm defined in (32) can be considered as a field strength of the worldvolume scalar field y(ξ ).

Thus. / Nuclear Physics B 586 (2000) 315–330 As a result. Bandos et al. (44) N 1 − e2Φ F 2 −(∂a)2 ∗ defined in (37) and (40).322 I. To rewrite the transformations and the self-duality condition (24) in the form adapted to the NS5-brane propagating in the D = 10 background.NS5 ∗ = −√ . let us b∗mn (8) introduce the NS5 counterpart of the tensor H 2 δLkin. (33). (41) [ ∂a∂a Taking into account Eqs. the dilaton and the gauge vector field of type IIA D = 10 supergravity Z q S = − d6 ξ e−2Φ − det(gmn − e2Φ Fm Fn ) s   eΦ (gmp − e2Φ Fm Fp ) H ∗np × det δm n + i p N det(δm n − e2Φ Fm F n )   Z √ ∂p a 1 e2Φ F k F p 1 p .NS5 denotes the first (DBI-like) term in the action (39). we get the action for a bosonic 5brane coupled to the metric. the action (39) (by construction) has gauge symmetries (5) and (28). its variation with respect to the gauge field b2 (ξ ) and the auxiliary scalar a(ξ ) has the form of Eq. as we have seen. we have obtained the action describing the worldvolume dynamics of the bosonic 5-brane propagating in the “Kaluza–Klein” part (29) of the IIA D = 10 supergravity background. produce the self-duality condition (24). (40) Hmn −g δH ∗mn where Lkin. substituting (30)–(38) into the action (1). (39) d6 ξ −g 2 H ∗mn Hmnk g kp + − 2Φ 2 4 N 1−e F −(∂a)2 Since the action (39) is nothing but the M5-brane action for a special choice of the D = 11 metric (29). " #  2Φ F p F s  δa 1 e a ∂ s ∗ p − 2 Hmnp g ps + Hmn δbmn = p N 1 − e2Φ F 2 −(∂a)2 −(∂a)2 (42) (43) and the self-duality equation for the NS5-brane gauge field b2   1 e2Φ F p F s ∂s a ∗ ps p Hmn = 2 Hmnp g + . and hence (39) is also invariant under the symmetries (6) and (7) which. with N and Hmn In addition to the worldvolume diffeomorphisms and the symmetries (42) and (43). δbmn = −2∂[m a φn] (ξ ). which is just the DBIlike term of the M5-action (1) written in the D = 10 adapted worldvolume frame. . δa = ϕ(ξ ). (36) and (41) we obtain the following form of the local worldvolume symmetries δa = 0. In the next section we extend this action to describe coupling of the NS5brane to antisymmetric gauge fields of IIA D = 10 supergravity. it is easy to find the relation between H s (∂a)2 ∗ ∗ bmn Hmn =H . Using ∗ b∗mn and Hmn (35). (24). (21).

In (47) Cˆ (6) is the pullback of a six-form gauge potential whose field strength is D = 11 Hodge-dual to the field strength of Cˆ (3) 1 dCˆ (6) + Cˆ (3) ∧ dCˆ (3) = ∗ dCˆ (3). (21)  Z  Z 1 d(da ∧ F2 ) ∧ (δb2 − δaF2 ) + dCˆ 3 ∧ δb2 . the M5brane action (1) extended by the Wess–Zumino term (47) is invariant under the following transformations of the antisymmetric gauge fields 1 δ Cˆ (6) = dϕˆ (5) − δ Cˆ (3) ∧ Cˆ (3) . and the dimensional reduction of Cˆ (6) produces a ten-dimensional five-form C (5) and a six-form B (6) which are dual to C (3) and B (2) . 2 (48) In addition to the symmetries (6) and (7) with H (3) generalized as in (45). δ Cˆ (3) = dϕˆ (2). / Nuclear Physics B 586 (2000) 315–330 323 4. (52) . Bandos et al. the variation (18) of the action (1) with respect to bmn (ξ ) and a(ξ ) acquires an additional term in comparison with Eq. (46) dξ 6 δL = − 2 The symmetries (6) and (7) spoiled by the last term of (46) are restored if to the action (1) one adds the Wess–Zumino term [40]  Z  1 (2) ˆ (3) (6) ˆ . the duality relations can be easily derived by the dimensional reduction of Eq. 2  (2) (2) ˆ δb = ϕˆ X(ξ ) . (45) As a result. (48). (49) (50) To get the form of the coupling of the NS5-brane to the antisymmetric gauge fields of type IIA D = 10 supergravity we should dimensionally reduce Cˆ (3) . up to a total derivative. respectively. Cˆ (6) and the Wess– Zumino term (47) of the M5-brane. (47) C + db ∧ C SWZ = 2 M6 As it was shown in [8]. the symmetries (6) and (7) uniquely fix the relative factor between SWZ and the action (1). Cˆ (6) = B (6) + C (5) ∧ F . Coupling to the background gauge fields When the M5-brane couples to the 3-form background field Cˆ (3) of D = 11 supergravity the field strength H3 gets extended by the worldvolume pullback of Cˆ (3) H (3) → Hˆ (3) = db(2) − Cˆ (3) . The dimensional reduction of Cˆ (3) produces a tendimensional R–R three-form C (3) and an NS–NS two-form B (2)  1 ˆ Cˆ (3) = dXˆ l ∧ dXˆ nˆ ∧ dXˆ mˆ Cˆ mˆ nˆ ˆl Xˆ 3!  1 1 = dXl ∧ dXn ∧ dXm Cmnl (X) + dXn ∧ dXm Bmn (X) ∧ dy − dXl Al 3! 2 (51) ≡ C (3) + B (2) ∧ F .I.

2 2 2 Before proceeding with the consideration of the full super-NS5-brane action let us demonstrate how the action of Refs. We have now obtained the action for the NS5-brane propagating in a background of the bosonic sector of type IIA D = 10 supergravity Z q S = − d6 ξ e−2Φ − det(gmn − e2Φ Fm Fn ) s   eΦ (gmp − e2Φ Fm Fp ) n ∗np H × det δm + i p N det(δm n − e2Φ Fm F n )   Z ∂p a 1 e2Φ F k F p 1 6 √ ∗mn kp Hmnk g + p d ξ −g 2 H − 4 N 1 − e2Φ F 2 −(∂a)2  Z  1 1 + (55) B (6) + C (5) ∧ F + db(2) ∧ C (3) + db(2) ∧ B (2) ∧ F . δy = ϕ (0). [19.324 I. Bandos et al. To get the second-order action we should expand (55) in series of H (3) and truncate it down to the second order in H (3) assuming the worldvolume gauge field to be . (53) and the NS5-brane action (39) is enlarged with the following Wess–Zumino term  Z  1 1 B (6) + C (5) ∧ F + db(2) ∧ C (3) + db(2) ∧ B (2) ∧ F . SWZ = 2 2 (54) M6 where F = dξ m (∂m y − Am ) is now the worldvolume pullback of the two-form (51).20] is a second-order approximation in powers of Hmnk of the NS5brane action. and 1 (2) 1 dϕ ∧ C (3) − dϕ (1) ∧ A ∧ C (3) . δB (2) = dϕ (1) . δA = dϕ (0). with H (3) now having the form (53). (55). with the self-duality condition being regarded as an extra (actually on-shell) constraint. the field strength of the self-dual gauge field of the NS5-brane coupled to the D = 10 background gauge fields is extended as follows H (3) = db(2) − C (3) − B (2) ∧ F . (42). / Nuclear Physics B 586 (2000) 315–330 Thus. and under target-space gauge transformations δC (3) = dϕ (2) + dϕ (1) ∧ A. 2 2 1 1 1 (57) δC (5) = dϕ (4) − dϕ (2) ∧ B (2) + dϕ (1) ∧ C (3) − d ϕ (1) ∧ A ∧ B (2) . (56) under which δH (3) = 0.20] is obtained from Eq. NS5-brane action in the second order approximation The action of [19. δb (2) = ϕ (2) − ϕ (1) ∧ dy. δB (6) = dϕ (5) + dϕ (4) ∧ A − 5. (43) and (28). 2 2 M6 This action is invariant under the worldvolume gauge transformations (5).

C (3) − B (2) ∧ A → C (3) . B (6) − C (5) ∧ A → −B˜ (6) . B (2) → B (2) . Hmnl 1 − e2Φ F 2 8 We should note that our choice of the dimensionally reduced C and C differs from that in [19. / Nuclear Physics B 586 (2000) 315–330 325 weak.I. we shall write down only the “kinetic” part of the action. Hmnl − Hˆ mnl (59) Taking into account the expressions 1 ∗ = p mnkqrs H qrs Hˆ mnk 3! −gˆ and mnlpqr  mnlst v = −(3!)2 δp[s δqt δrv] . 8 To carry out such a truncation the simplest way is to first truncate the M5-brane action (1) and then perform its dimensional reduction. 2 The WZ term of [19. Since the Wess–Zumino term is already linear and quadratic in H . and discarding terms containing the auxiliary field and the linearized NS5-brane selfduality condition (44)   e2Φ F p Fl p ∗ − Hmnp δl + = 0. after some algebra one can rewrite (58) in the following form [21] Z p  1 S = − d6 ξ −gˆ 1 − Hmnl H mnl 24    p 1 ∗ ∂m a H mnl − Hˆ ∗mnl Hnlp − Hˆ nlp − ∂ a + · · · .20].20] by the following field redefinition: y → c(0) . [19. and carrying out the direct dimensional reduction of (60) we recover the NS5-brane action of [19. so the 3 5 Wess–Zumino term in Eq. (55) is related to the one of Refs. Bandos et al. 1 C (5) → C (5) − C (3) ∧ B (2) .20] also contains the curl of an auxiliary worldvolume 5-form field which ensures the exact gauge invariance of the WZ term.20] Z q S = − d6 ξ e−2Φ − det(gmn − e2Φ Fm Fn )     e4Φ 1 2Φ mnk mnk p +3 Fm H Hnkp F + · · · . Up to the second order in H (3) the M5brane action has the form  Z p  1 ˆ ∗ ˆ ∗mn 1 ˆ ∗mn 6 p + H Hmnp vˆ + · · · (58) S = − d ξ −gˆ 1 − Hmn H 4 4 with the self-duality condition (24) reducing to ∗ = 0. e Hmnk H × 1− (61) 24 1 − e2Φ F 2 Alternatively. b(2) → a (2) . . the action (61) can be obtained directly by truncating the NS5-brane action (55). (60) [ 8∂a∂a Discarding in (60) the term containing the auxiliary field a(ξ ) and the anti-selfdual tensor H − Hˆ ∗ (which is zero on the mass shell (59)).

y . δκ b2 = iκ Cˆ 3 ≡ dZˆ M 3 ∧ dZˆ M 2 δκ Zˆ M 1 Cˆ Mˆ Mˆ Mˆ Zˆ . A MA (65) and by a three-superform generalization of the bosonic gauge field (51) 1 ˆ ˆ ˆ ˆ (66) Cˆ (3) = dZˆ N ∧ dZˆ M ∧ dZˆ L CLˆ Mˆ Nˆ (Z).45]. In addition to all symmetries discussed above and target-space superdiffeomorphisms the super-M5-brane action is invariant under the following fermionic κ-symmetry transformations [9. The κ -symmetric super-NS5-brane action To generalize the results of previous sections to describe the propagation of an NS5brane in a curved IIA D = 10 target superspace parametrized by ten bosonic coordinates Xm and 32-component Majorana-spinor fermionic coordinates Θ α forming a IIA. where we have separated the eleventh coordinate y = X10 keeping in mind the dimensional reduction of D = 11 superspace down to type IIA D = 10 superspace. we again start with an M5-brane propagating in a generic D = 11 supergravity background parametrized by eleven bosonic coordinates Xˆ mˆ and 32-component Majorana-spinor fermionic coordinates Θ α forming a D = 11 superspace coordinate system   ˆ (63) Zˆ M = Xˆ mˆ .10]   αˆ βˆ ˆ ˆ aˆ αˆ κβˆ . Eˆ α . where the worldvolume induced metric is now   ˆ ˆ aˆ (67) gˆmn = ∂m Zˆ M ∂n Zˆ N Eˆ ˆ Zˆ Eˆ Mˆ aˆ Zˆ . 3! The supervielbein. D = 11 superspace geometry is described by a supervielbein   ˆ ˆ ˆ A (64) Eˆ A = dZˆ M Eˆ ˆ Zˆ = Eˆ aˆ . M ˆ = (a. by a superconnection where A  ˆ Bˆ Bˆ wˆ ˆ = dZˆ M wˆ ˆ ˆ Zˆ . 1 2 3 2 where the spinor matrix Γ¯ has the following expansion in products of D = 11 Dirac matrices . Θ α .326 I. / Nuclear Physics B 586 (2000) 315–330 6. Bandos et al.37–39. the superconnection and the gauge superfield are subject to supergravity constraints which put the superfield formulation of eleven-dimensional supergravity on the mass shell. An explicit form of the D = 11 supergravity constraints relevant to the description of M5-brane dynamics the reader may find in [9. N Cˆ (3) Cˆ (6) and are worldvolume pullbacks of the three-superform (66) and its sixand superform dual [45. The super-M5-brane action has the similar form as the bosonic action (1) enlarged with the WZ term (47).46]. M M  1 ˆ ˆ ˆ (68) δκ a = 0. iκ Eˆ αˆ ≡ δκ Zˆ M Eˆ ˆ Zˆ = I − ΓS iκ Eˆ aˆ ≡ δκ Zˆ M Eˆ ˆ Zˆ = 0. Θ α = Z M . ˆ α) are locally flat tangent superspace indices. D = 10 superspace coordinate system  (62) Z M = Xm .

I. (71) Eˆ a = e− 3 Φ(Z) E a . (74) n! Note that the spinor superfield χ α does not appear in the action (55). This is the following superfield generalization of Eq. from (68) we get  ˆ δκ Hˆ 3 = −iκ Fˆ4 = Eˆ aˆ ∧ Eˆ b ∧ Eˆ α Γaˆ bˆ I − ΓS αβ κ β . / Nuclear Physics B 586 (2000) 315–330 p ΓS = q  −gˆ − det(gˆ + iHˆ ∗ ) 327    i ˆ∗ (6) mnl mn ˆ ˆ ˆ ˆ Γ + Hmn vˆl Γ + tm vˆn Γ . (73) and all the bosonic background fields are replaced with corresponding superfields. Eˆ α = e− 6 Φ(Z) E α (Z) + F χ α (Z). (68) by substituting into the latter the ansatz (71) and (72) αβ α κβ . (55) where now the worldvolume induced metric is a gmn = ∂m Z M ∂n Z N EN (Z)EMa (Z). A consistent ansatz for the dimensionally reduced supervielbein (64) was proposed in [48]. C (5) . (72) and solving for Bianchi identities. We do not write the super-NS5-brane action explicitly since it has exactly the same form as Eq. C (3) and B (2) are the worldvolume pullbacks of the type IIA D = 10 superforms 1 An E ∧ · · · ∧ E A1 CA1 ···An (Z). which is actually the Grassmann derivative of the dilaton superfield Φ(Z). (27)  1 2 2 Eˆ 10 = e 3 Φ(Z) dy − dZ M AM (Z) ≡ e 3 Φ(Z) F . Γˆ (6) = δκ gˆmn = −4iEˆ α (m Γn) I − ΓS  αβ κβ. (70) We now turn to the consideration of the super-NS5-brane action.45]. B (6) . Γˆm1 · · · Γˆm6 .39. AM (Z) are components of the one-form gauge superfield A = dZ M AM (Z). (75) iκ E αˆ ≡ δκ Z M EM (Z) = I − Γ¯ C (n) (Z) = . It can be obtained from the super-M5-brane action by the direct dimensional reduction of the D = 11 supergravity superfields. 2  1 m1 ···m6 ˆ aˆ ε Γˆm ≡ ∂m Zˆ M Eˆ ˆ Zˆ Γˆa . Φ(Z) is the dilaton superfield. Bandos et al. in particular. for the M5-brane action to be κ-symmetric the superbackground must satisfy the supergravity constraints [9. M 6! 1 1 ∗ ˆ∗ b∗ H b∗ vˆq . When they are taken into account. and χ α (Z) is a Grassmann-odd Majorana spinor superfield. 1 (72) A where E A (Z) = dZ M EM = (E a . Different forms of these constraints have been considered in [48–51]. The super-NS5-brane action is invariant under κ-symmetry transformations obtained from Eqs. (69) Hpl vˆq ≡ εmnkplq H tˆ m = εmnkplq Hˆ nk nk pl 8 8 As is characteristic of all superbranes. E α ) are supervielbeins of type IIA D = 10 supergravity. The superfields which describe IIA D = 10 supergravity are subject to the constraints which are obtained from the D = 11 supergravity constraints using the ansatz (71).

Acknowledgements The authors are grateful to Kurt Lechner. (76) where Aα (Z) is a fermionic component of the Kaluza–Klein connection form A ≡ dZ M AM = E α Aα + E a Aa . An interesting problem for future study is to construct the Lagrangian description of the consistent coupling of a type IIA supergravity action to an NS5-brane source. also acknowledge kind hospitality extended to them at the Abdus Salam International Centre for Theoretical Physics where . δκ a = 0.B.53].56]. the type IIB D = 10 KK monopole is expected to be a self-dual object under the S-duality symmetry of type IIB supergravity. and A. The latter requires the construction of a duality-symmetric version of type IIA supergravity by the dimensional reduction of the duality-symmetric D = 11 supergravity [46]. Jeffrey Harvey. 7. κ-symmetry and background supergravity symmetries the super-NS5-brane action possesses special local symmetries ensuring the covariance of actions with self-dual gauge fields and serving for deriving the self-duality condition directly from the action as a consequence of the equation of motion of the gauge field. Conclusion and discussion To summarize. which should naturally couple to a heterotic five-brane [52. In addition to worldvolume diffeomorphisms. gauge symmetry. As it was noted in [47] and proved in the second order approximation in [20]. One of possible ways of deriving appropriate T-duality transformation rules for the antisymmetric gauge fields is to T-dualize the duality-symmetric version of type IIA supergravity to the duality-symmetric version of type IIB supergravity [55. we have obtained the covariant κ-symmetric action for the super-NS5brane in a IIA D = 10 supergravity background by the direct dimensional reduction of the M-theory super-five-brane action. I. Bandos et al. ⇒ iκ F = 0 δκ y = iκ E α Aα (Z). D = 10 supergravity.N. Bernard Julia and Kellog Stelle for useful comments. δκ b2 = iκ C3 + F ∧ iκ B2 . Paolo Pasti and Mario Tonin for interest to this work and valuable discussions and to Christopher Hull. The construction of the complete action for the type IIB KK monopole should allow one to explicitly verify this statement. A quadratic approximation for the bosonic part of this action has been constructed in [20]. The truncation of such a IIA supergravity action shall produce the duality-symmetric version of the N = 1. / Nuclear Physics B 586 (2000) 315–330 iκ E a = 0. Note that recent investigations of interacting brane actions [54] may provide one with a possibility of making this coupling supersymmetric.328 I. Another problem for further studying is to perform the T-duality transformation of the complete NS5-brane action and to arrive at a non-linear and supersymmetric action for a type IIB D = 10 Kaluza–Klein (KK) monopole.

I. Bandos et al. / Nuclear Physics B 586 (2000) 315–330

329

part of this work was done. This work has been partially supported by the Ukrainian GKNT
Grant 2.5.1/52 and INTAS Grant No 96-0308.

References
[1] P.K. Townsend, Nucl. Phys. (Proc. Suppl.) 58 (1997) 163, hep-th/9609217.
[2] P.K. Townsend, Four lectures on M-theory, in: Proc. of the ICTP Summer School on HEP and
Cosmology, Trieste, 1997; hep-th/9612121.
[3] A. Hanany, E. Witten, Nucl. Phys. B 492 (1997) 152.
[4] E. Witten, Adv. Theor. Math. Phys. 2 (1998) 61, hep-th/9710065.
[5] K. Bachas, “Desert” energy or transverse space, hep-th/9907023; Class. Quant. Grav. 17 (2000)
1, hep-th/0001093, and references therein.
[6] E. Witten, JHEP 01 (1998) 001.
[7] O. Aharony, A brief review of “little string theories”, Class. Quant. Grav. 17 (2000) 929.
[8] P. Pasti, D. Sorokin, M. Tonin, Phys. Lett. B 398 (1997) 41, hep-th/9701037.
[9] I. Bandos, K. Lechner, A. Nurmagambetov, P. Pasti, D. Sorokin, M. Tonin, Phys. Rev. Lett. 78
(1997) 4332, hep-th/9701149.
[10] M. Aganagic, J. Park, C. Popescu, J.H. Schwarz, Nucl. Phys. B 496 (1997) 191, hep-th/9701166.
[11] D. Berman, Phys. Lett. B 409 (1997) 153–159, hep-th/9706208; Nucl. Phys. B 533 (1998) 317–
332, hep-th/9804115.
[12] A. Nurmagambetov, Phys. Lett. B 436 (1998) 289.
[13] E. Witten, J. Geom. Phys. 22 (1997) 103, hep-th/9610234.
[14] K. Becker, M. Becker, Fivebrane gravitational anomalies, hep-th/9911138.
[15] L. Bonora, C.S. Chu, M. Rinaldi, JHEP 12 (1997) 007, hep-th/9710063.
[16] L. Bonora, M. Rinaldi, Normal bundles, pfaffians and anomalies, hep-th/9912214.
[17] D. Freed, J.A. Harvey, R. Minasian, G. Moore, Adv. Theor. Math. Phys. 2 (1998) 601, hepth/9803205.
[18] E. Witten, Duality relations among topological effects in string theory, hep-th/9912086.
[19] E. Bergshoeff, Y. Lozano, T. Ortin, Nucl. Phys. B 518 (1998) 363.
[20] E. Eyras, B. Janssen, Y. Lozano, Nucl. Phys. B 531 (1998) 275, hep-th/9806169.
[21] P. Pasti, D. Sorokin, M. Tonin, Phys. Lett. B 352 (1995) 59; Phys. Rev. D 52 (1995) R4277;
Phys. Rev. D 55 (1997) 6292.
[22] D. Zwanziger, Phys. Rev. D 3 (1971) 880.
[23] S. Deser, C. Teitelboim, Phys. Rev. D 13 (1976) 1592.
[24] R. Floreanini, R. Jackiw, Phys. Rev. Lett. 59 (1987) 1873.
[25] M. Henneaux, C. Teitelboim, in: Proc. Quantum Mechanics of Fundamental Systems 2,
Santiago, 1987, p. 79; Phys. Lett. B 20 (1988) 650.
[26] J.H. Schwarz, A. Sen, Nucl. Phys. B 411 (1994) 3.
[27] M. Perry, J.H. Schwarz, Nucl. Phys. B 489 (1997) 47.
[28] J.H. Schwarz, Phys. Lett. B 395 (1997) 191.
[29] A. Maznytsia, C.R. Preitschopf, D. Sorokin, Nucl. Phys. B 539 (1999) 438.
[30] Y.-G. Miao, H.J.W. Mueller-Kirsten, Self-duality of various chiral boson actions, hepth/9912066.
[31] Y.-G. Miao, R. Manvelyan, H.J.W. Mueller-Kirsten, Self-duality beyond chiral p-form actions,
hep-th/0002060.
[32] P.S. Howe, E. Sezgin, Phys. Lett. B 394 (1997) 62.
[33] P.S. Howe, E. Sezgin, P.C. West, Phys. Lett. B 399 (1997) 49.
[34] I. Bandos, K. Lechner, A. Nurmagambetov, P. Pasti, D. Sorokin, M. Tonin, Phys. Lett. B 408
(1997) 135.

330

[35]
[36]
[37]
[38]
[39]
[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]
[52]
[53]
[54]
[55]
[56]

I. Bandos et al. / Nuclear Physics B 586 (2000) 315–330

D. Sorokin, P.K. Townsend, Phys. Lett. B 412 (1997) 265.
E. Bergshoeff, D. Sorokin, P.K. Townsend, Nucl. Phys. B 533 (1998) 303.
P. Claus, R. Kallosh, A. Van Proeyen, Nucl. Phys. B 518 (1998) 117.
P. Claus, Phys. Rev. D 59 (1999) 066003.
D. Sorokin, Superbranes and superembeddings, hep-th/9906142; Physics Reports (in press).
O. Aharony, M theory and string dualities, in: Gauge Theories, Applied Supersymmetry and
Quantum Gravity, Imperial Colledge Press, 1997.
E. Bergshoeff, M. de Roo, T. Ortin, Phys. Lett. B 386 (1996) 85, hep-th/9606118.
M.J. Duff, J.X. Lu, Nucl. Phys. B 354 (1991) 129.
M.J. Duff, R.R. Khuri, J.X. Lu, Phys. Rep. 259 (1995) 213.
E. Cremmer, B. Julia, H. Lü, C.N. Pope, Nucl. Phys. B 535 (1998) 242, hep-th/9806106.
A. Candielo, K. Lechner, Nucl. Phys. B 412 (1994) 479.
I. Bandos, N. Berkovits, D. Sorokin, Nucl. Phys. B 522 (1998) 214.
C. Hull, Nucl. Phys. B 509 (1998) 216, hep-th/9705162.
M.J. Duff, P.S. Howe, T. Inami, K.S. Stelle, Phys. Lett. B 191 (1987) 70.
J.L. Carr, S.J. Gates Jr., R.N. Oerter, Phys. Lett. B 189 (1987) 68.
M. Cederwall, A. von Gussich, B.E.W. Nilsson, P. Sundell, A. Westerberg, Nucl. Phys. B 490
(1997) 179, hep-th/9610148.
E. Bergshoeff, E. Cowdall, P.K. Townsend, Phys. Lett. B 410 (1997) 13, hep-th/9706094.
E. Witten, Nucl. Phys. B 460 (1996) 541, hep-th/9511030.
J. Mourad, Nucl. Phys. B 512 (1998) 199, hep-th/9709012.
I. Bandos, W. Kummer, Phys. Lett. B 462 (1999) 254, hep-th/9905144; Nucl. Phys. B 565
(2000) 291, hep-th/9906041.
G. Dall’Agata, K. Lechner, D. Sorokin, Class. Quant. Grav. 14 (1997) L195.
G. Dall’Agata, K. Lechner, M. Tonin, JHEP 9807 (1998) 017.

Nuclear Physics B 586 (2000) 331–345
www.elsevier.nl/locate/npe

Five-dimensional gauge theories and local
mirror symmetry
Tohru Eguchi a , Hiroaki Kanno b,∗
a Department of Physics, Faculty of Science, University of Tokyo, Tokyo 113, Japan
b Department of Mathematics, Graduate School of Science, Hiroshima University,

Higashi-Hiroshima 739, Japan
Received 2 May 2000; accepted 13 June 2000

Abstract
We study the dynamics of 5-dimensional gauge theory on M4 × S 1 by compactifying type II/M
theory on degenerate Calabi–Yau manifolds. We use the local mirror symmetry and shall show that
the prepotential of the 5-dimensional SU (2) gauge theory without matter is given exactly by that of
the type II string theory compactified on the local F2 , i.e., Hirzebruch surface F2 lying inside a noncompact Calabi–Yau manifold. It is shown that our result reproduces the Seiberg–Witten theory at
the 4-dimensional limit R → 0 (R denotes the radius of S 1 ) and also the result of the uncompactified
5-dimensional theory at R → ∞.
We also discuss SU (2) gauge theory with 1 6 Nf 6 4 matter in vector representations and show
that they are described by the geometry of the local F2 blown up at Nf points.  2000 Elsevier
Science B.V. All rights reserved.
PACS: 11.15.-g; 11.30.Pb; 02.10.Rn
Keywords: Mirror symmetry; Seiberg–Witten theory; Instanton expansion; Supersymmetric gauge theory

1. Introduction
Recent developments in non-perturbative string theory and M theory have led to new
insights into the relation between low energy field theory and string theory: it has been
argued in particular that non-perturbative dynamics takes place in low energy field theory
and higher gauge symmetries emerge when compactifying Calabi–Yau and K3 manifolds
degenerate and some of their homology cycles vanish. For instance, when K3 surface
develops an A-D-E singularity, there appears an enhanced A-D-E gauge symmetry in
6-dimensions. Similarly, when a family of P1 ’s shrinks to zero size along a rational curve
∗ Corresponding author. E-mail: kanno@math.sci.hiroshima-u.ac.jp

0550-3213/00/$ – see front matter  2000 Elsevier Science B.V. All rights reserved.
PII: S 0 5 5 0 - 3 2 1 3 ( 0 0 ) 0 0 3 7 5 - 8

332

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

in Calabi–Yau threefold, we obtain a 4-dimensional SU (2) gauge theory with N = 2
supersymmetry.
In this article we would like to study the dynamics of SUSY gauge theories in 5dimensions by compactifying the M theory on degenerate Calabi–Yau manifolds. We
construct an effective action of an SU (2) gauge theory on a 5-dimensional space M4 × S 1
where M4 is the Minkowski space and S 1 is a circle of radius R. In the limit of R → 0 this
theory reproduces the standard N = 2 SUSY gauge theory in 4-dimensions, i.e., Seiberg–
Witten theory [1,2]. Thus our 5-dimensional model gives an M-theoretic generalization of
Seiberg–Witten theory by incorporating Kaluza–Klein excitations. In the opposite limit
R → ∞ our model reduces to the gauge theory in uncompactified 5-dimensions M5 .
Characteristic features of this theory have been studied using the brane-probe picture [3]
and also from the point of view of classical geometry of collapsing del Pezzo surfaces
[4–6] and the behavior of the low-energy effective gauge coupling has been determined
exactly.
In this paper we would like to propose an exact solution of the 5-dimensional theory on
M4 × S 1 which reproduces the known results at both limits R → 0 and R → ∞. As it turns
out, our prepotential follows directly from that of the type II string theory compactified on
singular Calabi–Yau manifolds using the method of local mirror symmetry [7–10]. In the
case of pure SU (2) gauge theory without matter our result is obtained from the type II
theory compactified on the local F2 , i.e., Hirzebruch surface F2 lying inside a Calabi–Yau
threefold which is the canonical bundle over F2 . Similarly, SU (2) gauge theories with Nf
matter in vector representations are also obtained from the type II theory compactified on
the local F2 blown up at Nf points (0 6 Nf 6 4). We will find that our model at R = ∞
has an infinite bare coupling constant and yields a non-trivial interacting field theory in the
infra-red limit.
The local mirror symmetry is a method of mirror symmetry adapted in the case of noncompact Calabi–Yau manifolds. Suppose, for instance, we are given a compact Calabi–Yau
threefold which is an elliptic fibration over Fn . One considers the limit of the size of the
fiber tE going to ∞. Then the resulting non-compact manifold is modeled by the local Fn ,
i.e., Fn inside a Calabi–Yau with the normal bundle being given by the canonical bundle of
Fn . The limit of tE → ∞ may also be considered as the limit of shrinking Fn with the size
of the fiber kept fixed. Fn (n = 0, 1, 2) and its various blow ups are the del Pezzo surfaces
and these are in fact the type of manifolds which featured in the geometrical interpretation
of 5-dimensional gauge theory [4–6].

2. SU (2) gauge theory without matter
Let us start from the case of 5-dimensional gauge theory without matter. We consider
the local F2 model which is described by the toric data given in the Appendix. Following
the standard procedure [7–10] one obtains a curve in the B-model given by
P = a0 x + a1 x 2 + a2 ζ + a3 + a4

1
= 0.
ζ

(2.1)

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

333

Introducing a new variable y = a2 ζ − a4 /ζ the curve is rewritten as
y 2 = (a1 x 2 + a0 x + a3 )2 − 4a2a4 .

(2.2)

Complex moduli of the B-model are defined by
zF =

a1 a3
,
a02

zB =

a2 a4
.
a32

(2.3)

If we choose a1 = a3 = 1, a0 = s, 4a2 a4 = K 4 , we find
y 2 = (x 2 + sx + 1)2 − K 4 ,

zF =

1
,
s2

zB =

K4
.
4

(2.4)

(2.4) is in fact the curve proposed by Nekrasov [11] for the description of 5-dimensional
gauge theory and its properties have been studied in [12] in detail.
If we introduce a variable U which is the analogue of u of the Seiberg–Witten solution,
the parameter s is written as
s = 2R 2 U,
where R is the radius of S 1 . In terms of U and R the curve reads as   

2
1
− K 4.
y 2 = x 2 − R4 U 2 − 4
R

(2.5)

(2.6)

By comparing (2.6) with the Seiberg–Witten curve y 2 = (x 2 − u)2 − Λ4 , we find the
correspondence between the parameters U and u  

1
2
2
(2.7)
u ⇐⇒ R U − 4 .
R
As we see from (2.7), U variable describes two copies of the u-plane. Strong coupling
region u ≈ 0 maps to U ≈ ±1/R 2 and thus the two copies are separated by a distance of
order 1/R 2 .
In the brane-probe interpretation of Seiberg–Witten solution [13,14], u-plane is
identified as a local region around one of the four fixed points (O-7 planes) in type I0
theory compactification to 8-dimensions on T 2 (u = 0 is identified as the location of
the O-7 plane). Then the curve (2.6) describes a theory which contains two of these O7 planes. Since the fixed plane acts like a reflecting mirror, D3-brane probe will possess an
infinite number of mirror images when inserted into a background of two orientifold planes.
These mirror images are separated by distances n/R 2 , n ∈ Z and open strings connecting
them generate Kaluza–Klein modes of supersymmetric gauge fields. Thus the curve (2.6)
effectively describes a theory on a 5-dimensional manifold M4 × S 1 with R being the
radius of S 1 . In the limit of R → 0 one of the u-planes moves off to ∞ and the model (2.6)
is expected to reduce to the Seiberg–Witten theory.
Periods of the B-model of local F2 (2.4) is determined by solving differential
equations (Gelfand–Kapranov–Zelevinskij (GKZ) system) associated with the toric data
(see Appendix). Differential operators are given by

334

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345  

L1 = zF 4θz2F + 2θzF + θzF 2θzB − θzF ,  

L2 = zB 2θzB − θzF + 1 2θzB − θzF − θz2B ,

(2.8)
(2.9)

where
θzF ≡ zF


,
∂zF

θzB ≡ zB


.
∂zB

(2.10)

These operators have a regular singular point at zF = zB = 0: they possess “singlelog” solutions ωF , ωB behaving as ωF = log zF + · · · and ωB = log zB + · · · at
zF , zB ≈ 0. There also exists a “ double-log” solution Ω behaving as Ω = (log zF )2 +
(log zF )(log zB ) + · · · . We identify the two single-log solutions as the Kähler parameters
tF , tB of the A-model: tF represents the size of the P1 fiber of F2 and tB the size of its base
P1 . tF is given by
"
#
X (2m − 1)!
X
2(2n − 1)!
n m
m
z z −
zB ,
−tF ≡ ωF = log zF +
(n − 2m)!n!m!2 F B
m!2
n>1,m>0
m>1
s
!
1
1
+
−1
= −2 log √
4zF
4zF
#
"
X (2m − 1)!
X
2(2n − 1)!
n m
m
z z −
zB .
(2.11)
+
(n − 2m)!n!m!2 F B
m!2
n>1,m>1

Similarly, tB is given by

m>1

s

1
+
−tB ≡ ωB = −2 log √
4zB

!
1
−1 .
4zB

(2.12)

In our interpretation as the 5-dimensional gauge theory, the size of the fiber tF is
identified as the vacuum expectation value A of the scalar field in the vector-multiplet
tF = 4RA.

(2.13)

On the other hand, the size of the base tB is related to the dynamical mass parameter Λ as
e−tB = 4R 4 Λ4 .

(2.14)

Then the mirror transformation (2.12) becomes
K=√

2RΛ
1 + 4R 4 Λ4

.

(2.15)

Note that (2.13) and (2.14) are in fact the identification of variables suggested in [7,15].
One can invert the relations (2.11), (2.12) perturbatively and express the B-model
parameters zF ,zB in terms of tF , tB . In the case of local mirror symmetry the holomorphic
solution of GKZ system is a constant and the mirror transformation is simpler than in the
compact Calabi–Yau case. One may then represent the double-log solution in terms of the
Kähler parameters

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345 

X 

X 1
n
2 n
q
+
4q
n
q
B
F
n2 F
n=1
n=1  

qF2 + 36qF3 + 260qF4 + 1100qF5 + · · · + O qB3 ,

335

Ω = tF2 + tF tB + 4
+ qB2

(2.16)

where
qF = e−tF ,

qB = e−tB .

(2.17)

We identify the double-log solution as AD , the dual of the variable A
Ω = −8πiRAD .

(2.18)

First two terms of (2.16) represent the classical intersection numbers of the Calabi–Yau
manifold and the remaining terms represent the contribution of world-sheet instantons.
According to Ref. [10] the double-log solution has a generic form 

2
X X X
X
q kn q km
i j
i
ti tj hJ J i +
x ∂ti dn,m 1 32 ,
(2.19)
Ω=
k
i,j =1

k=1 n,m>0

i

for a local model of a surface S (with two Kähler parameters). Ji denotes the Kähler classes
of S and hJi Jj i their intersection numbers. Numerical coefficients x i are defined by
X
x i Ji ,
(2.20)
c1 (S) =
i

where c1 (S) is the first Chern class of S. dn,m gives the number of rational holomorphic
curves in the homology class nJ1 + mJ2 . Sum over k in (2.19) represents the multiplecover factor.
By comparing (2.16) and (2.19) (set q1 = qF , q2 = qB ) we find [7,10]
d1,0 = −2,
d1,2 = d2,2 = 0,

dn,0 = 0,

n > 1,

d3,2 = −6,

dn,1 = −2n,
d4,2 = −32,

...

By integrating AD over A we have the prepotential F for the local F2 model
" 
X 

X 1
tF2 tB
tF3
1
n
n
+4
q + 4qB
nqF
F=
− −
3
2
32πiR 2
n3 F
n=1
n=1
#   

2 1 2
3
4
5
3
q + 12qF + 65qF + 220qF + · · · + O qB .
+ qB
2 F

(2.21)

(2.22)

3. Small and large radius limits
Let us next examine the small and large radius limits R → 0, ∞ of (2.22). We first
consider the 4-dimensional limit R → 0. Due to the relations (2.13), (2.14) small R
corresponds to the base of F2 becoming large (tB → ∞) and the fiber becoming small
(tF → 0) while the ratio e−tB /tF4 is kept fixed. At small R, we have K ≈ 2RΛ and U ≈
cosh(2RA). As explained by Katz, Klemm and Vafa [7], quantum parts of F are suppressed

336

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

because of the powers of qB ≈ R 4 and the surviving contributions come from the divergent
P
parts of the series dn,m qFn over qF as qF = e−4RA → 1. At small R the gauge coupling
τ = ∂ 2 F /∂A2 behaves as
X
i
8i
2i X qFn
4i
log(4R 4 Λ4 ) −
− R 4 Λ4
n3 qFn + · · ·
τ = RA −
π

π
n
π
n=1
n=1  

4

A
3i Λ
2i
+ ···.
(3.1)

≈ log 2 2
π
Λ
16π A4
(3.1) reproduces the one-loop beta function and the one-instanton contribution to the
Seiberg–Witten solution [1]. We can check the agreement with Seiberg–Witten theory for
higher instanton terms.
In (3.1) the world-sheet instanton expansion of type II theory is converted into the
space-time instanton expansion of gauge theory in the R → 0 limit. Coefficients of the
m-instanton amplitudes of gauge theory are determined by the asymptotic behavior of the
number of holomorphic curves dn,m as n → ∞ with fixed m.
Let us next examine the R → ∞ limit of the uncompactified 5-dimensional gauge
theory. It is known [16,17] that the gauge theory on M5 has no quantum corrections and its
gauge coupling is simply expressed in terms of classical intersection numbers of Calabi–
Yau manifold. In fact by taking R → ∞ world-sheet instanton terms disappear and we
have
2
τ
≡ τ5 = 2 A
(3.2)
lim
R→∞ 2πiR
π
(we have rescaled τ so that τ5 corresponds to the gauge coupling of 5-dimensional theory,
τ5 = 1/g52 ). In the next section we will discuss local F2 model blown up at Nf points. We
then find that the above formula is generalized as
!
Nf
Nf
X
X
2
1
1
|A − Mi | −
|A + Mi | .
(3.3)
τ5 = 2 A −
16
16
π
i=1

i=1

(3.3) is exactly the behavior of gauge coupling of SU (2) theory with Nf matter in vector
representations (at infinite bare coupling) [3–5]. Thus we reproduce correct results also in
the uncompactified limit R → ∞.
We should note that the local models of F0 and F1 also reproduce Seiberg–Witten
solution in the R → 0 limit [7] since the asymptotic behavior of the number of holomorphic
curves dn,m are the same for all models Fi , i = 0, 1, 2. F0 , F1 , however, have a different
classical topology from F2 and do not reproduce (3.2) at R = ∞. Thus the F2 model is
singled out as the unique candidate for the description of 5-dimensional gauge theory on
M4 × S 1 .
4. SU (2) theory with matter
Let us next consider the case of SU (2) gauge theory coupled to Nf matter (1 6 Nf 6 4)
in vector representations. We first discuss the Nf = 1 case. Relevant geometry is given by

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

337

the local F2 with one-point blown up. Corresponding curve is given by (see Appendix)  

1
2
2
2
3
.
(4.1)
y = (x + sx + 1) − K xw +
w
Comparison with the Nf = 1 Seiberg–Witten curve suggests the identification
w = eRM ,

(4.2)

where M denotes the bare mass of the matter multiplet. Complex moduli of the B-model
are given by (we denote zF −E as zF and zF +E as zF 0 for notational simplicity)
w2
1
K3
,
zF 0 =
.
(4.3)
,
zB =
2
s
sw
4w
In this case the GKZ system is given by five partial differential equations in three variables
(see Appendix) and is of considerable complexity. Here we content ourselves with the
analysis of prepotential at the tree and one-loop level ignoring the space–time instantons.
This suffices for our purpose of extracting the R → ∞ behavior of the theory. The small
R behavior has already been studied in [8] and argued to reproduce Seiberg–Witten theory
(we have also verified that the 1-instanton term is correctly reproduced).
When we ignore instantons, single log solutions are given by
X (2n − 1)!
(zF zF 0 )n ,
(4.4)
−tF ≡ ωF = log(zF ) +
n!2
n=1
X (2n − 1)!
(zF zF 0 )n ,
(4.5)
−tF 0 ≡ ωF 0 = log(zF 0 ) +
n!2
zF =

n=1

−tB ≡ ωB = log(zB ).

(4.6)

Identification with the variables of the gauge theory is given by
tF = 2R(A − M),

tF 0 = 2R(A + M),

e−tB =

2R 3 Λ3
.
w

(4.7)

Then by inverting relations (4.4), (4.5) we find
e−2R(A−M)
,
1 + e−4RA
The double log solution
zF =

zF 0 =

e−2R(A+M)
.
1 + e−4RA

(4.8)

n
X 4(2n − 1)! X
1
1
(zF zF 0 )n
Ω = tF2 + tF2 0 + 2tF tF 0 + tB (tF + tF 0 ) +
2
2
n!
j
n=1
j =1
!
X X (n + ` − 1)! 1
zn z` 0 ,
+ −
+
n!`!
n−` F F
n>`

(4.9)

n<`

can then be re-expressed as
1
Ω = tF2 + tF2 0 + 2tF tF 0 + tB (tF + tF 0 )
2
X 1
X 1
X 1
−4nRA
−2nR(A+M)
e

e

e−2nR(A−M) ,
+4
n2
n2
n2
n=1

n=1

n=1

(4.10)

338

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

using (4.8).
When we take the derivative in A, (4.10) gives    

1
i
2 log sinh(2RA) − log sinh R(A + M) sinh R(A − M) .
τ=
π
4
In the 4-dimensional limit (4.11) becomes  

1
i
2−
log A
τ≈
π
2

(4.11)

(4.12)

which gives the 1-loop beta function of Seiberg–Witten theory. On the other hand in the
5-dimensional limit we find  

2
1
τ
1
= 2 A − |A − M| − |A + M|
(4.13)
τ5 = lim
R→∞ 2πiR
π
16
16
as we have claimed before. We note that in formulas (3.2) and (4.13) τ5 does not contain an
2 . Thus the
additive constant which is identified as the bare coupling constant τ5,B = 1/g5,B
2 = ∞ which yields non-trivial
theory sits at the infinite bare coupling constant limit g5,B
5-dimensional theory in the infra-red regime [3].
We can similarly study the system with more matter up to Nf = 4 using the local mirror
symmetry. Curves of the B-model are given by (see Appendix)
y = (x + sx + 1) − K
2

2

2

4−Nf

Nf  

Y
1
wi x +
,
wi

(4.14)

i=1

where parameters wi correspond to the bare masses of the matter multiplets
wi = eRMi ,

i = 1, . . . , 4.

(4.15)

(4.14) agrees with the curve suggested by [11] and [18] (at Nf = 4 the factor K 4−Nf
should be replaced by a dimensionless parameter q).
In the cases Nf > 2 the analysis of GKZ system becomes further involved: we will
instead use a simpler method based on the Picard–Fuchs (PF) equation derived for the
elliptic curves (4.14). It turns out that the PF equation of the elliptic curve is an ordinary
differential equation of third order in the variable A and can be studied relatively easily.
This is the method used in Ref. [12]. This system, however, is not complete unlike that of
GKZ case. The periods are determined only up to integration constants and one can not
precisely fix the mirror transformation. This ambiguity, however, affects only the quantum
part of the computation and one still obtains precise results for the prepotential at the tree
and one-loop level.
First we note that the quadratic curve y 2 = ax 4 + 4bx 3 + 6cx 2 + 4dx + e is transformed
into the Weierstrass form
y 2 = 4x 3 − g2 x − g3 ,

(4.16)

by the relation
g2 = ae − 4bd + 3c2,

g3 = ace + 2bcd − ad 2 − b2 e − c3 .

(4.17)

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

339

We regard g2 , g3 as functions of the parameter s. Periods ω of the elliptic curve (4.16) obey
the PF equation [19,20]

d2 ω
+ c0 (s)ω = 0,
+ c1 (s)
2
ds
ds

(4.18)   

3
dg2
d
dg3
2g2
−3
g3 ,
c1 = − log
ds
2∆
ds
ds

(4.19)

c0 =

(4.20)

where  

  


1 d2
1
1
dg3 2
d
dg2 2
c1 log ∆ +
log



12
g
2
12 ds
12 ds 2
16∆
ds
ds

and ∆ = (g2 )3 − 27(g3 )2 denotes the discriminant of the curve. Since dA/ds is one of the
periods of the curve, it satisfies the PF equation. Regarding s as a function of A, we obtain 
2 2 
2 2 
4
d s
d s
ds
ds
ds d3 s
−3
+ c1
− c0
= 0.
(4.21)
3
2
2
dA dA
dA dA
dA
dA
This determines s in terms of A.
Similarly dAD /ds = dA/ds · d2 F /dA2 satisfies the PF equation and we obtain 
−2 2 3 
−1 4
ds
d F
d s d F
d3 F
ds

3
+
c
= 0.
1
dA
dA
dA4
dA2 dA3
dA3
This equation can be integrated once and we find  

3
dg2
const
dg3
ds
d3 F
2g2
−3
g3
=
.
3

ds
ds
dA
dA

(4.22)

(4.23)

Solving (4.23) determines the prepotential (const in the right-hand-side is fixed by a
suitable normalization of F ).
In the case Nf = 2, we find the gauge coupling at the classical and one-loop level as  

2  

1Y
i
log sinh R(A + Mi ) sinh R(A − Mi ) . (4.24)
2 log sinh(2RA) −
τ=
π
4
i=1

In the 4-dimensional limit (4.24) reads as  

1
2i
2 − · 2 log A,
τ≈
π
2

(4.25)

which gives the beta function of Nf = 2 Seiberg–Witten theory. On the other hand, in the
five-dimensional limit we have
!
2
2
X
X
2
1
1
|A − Mi | −
|A + Mi |
(4.26)
τ5 = 2 A −
π
16
16
i=1

i=1

in agreement with (3.3).
We have checked that the local model of F2 blown up at 3 and 4 points (see Appendix)
also reproduce (3.3). Thus we have obtained a model which appears to describe correctly
the physics of 5-dimensional theory up to the number of flavors Nf = 4 making use of the
mirror symmetry.

340

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

5. Discussions
Since F2 is obtained from F1 which is a one-point blow up of P2 , we effectively have up
to 5-point blow ups of P2 describing the gauge theory. Unfortunately, beyond the 5-point
blow up mirror symmetry of del Pezzo surfaces can not be described by toric geometry
and we can not apply our analysis for these cases. In fact in the range Nf > 5 something
drastic must happen, since in this asymptotically non-free region Seiberg–Witten solution
does not exist and we should not have a smooth 4-dimensional limit. On the other hand, in
this range the 5-dimensional gauge theories are expected to possess En , n = 6, 7, 8 global
symmetries and are of particular interests. It is a challenging problem to clarify the physics
of gauge theory with En symmetry. It may shed some light on the nature of asymptotically
non-free field theories in 4-dimensions.
Our result shows that from the point of view of the type II/M theory compactified
on Calabi–Yau manifold, low energy 5-dimensional theory on M4 × S 1 emerges most
naturally with its prepotential being exactly the same as that of the string theory. On the
other hand, 4-dimensional Seiberg–Witten theory appears only in the fine-tuned limit of
the Kähler parameters. It seems possible that our 5-dimensional model is an example of
an “M-theoretic” lift of various 4-dimensional quantum field theories. We may imagine
most of the 4-dimensional SUSY field theories in fact have a lift to 5-dimensions where
the quantum effects of the loops and instantons are replaced by purely geometrical effects
of world-sheet instantons. It will be also quite interesting to see if there is a further lift of
quantum field theory to 6-dimensions as suggested by the duality between F- and M-theory.

Acknowledgements
We would like to thank M. Jinzenji, M. Naka and Y. Ohta for discussions. Researches
of T.E. and H.K. are supported by the fund for the Special Priority Area No.707, Japan
Ministry of Education. We also acknowledge the stimulating atmosphere at Summer
Institute ’99 where this research was started.

Appendix A
A.1. Nf = 0
For pure Yang–Mills case we take the Hirzebruch surface F2 lying inside a Calabi–Yau
manifold. Its toric diagram has five vertices;
ν0 = (0, 0),
ν3 = (−1, 0),

ν1 = (1, 0),

ν2 = (0, 1),

ν4 = (−2, −1).

(A.1)

This is a two-dimensional reflexive polyhedra No. 4 in Fig. 1 of [10]. The charge vectors
satisfying the linear relation

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

X

`(k)
i νi = 0

341

(A.2)

i

are given by
`(F ) = (−2; 1, 0, 1, 0),

`(B) = (0; 0, 1, −2, 1).

The constraint among B-model variables Yi , i = 0, 1, . . . , 4
Y −`(k)
Y `(k)
Yi i =
Yi i
(k)
`i >0

(A.3)

(A.4)

(k)
`i <0

gives
Y02 = Y1 Y3 ,

Y2 Y4 = Y32 ,

(A.5)

and we have a solution  

s4
.
Y = sx, x 2 , ζ, s 2 ,
ζ

(A.6)

Set s = 1. Then we obtain the following curve in the B-model side
X
ai Yi ,
PNf =0 =
= a0 x + a1 x 2 + a2 ζ + a3 + a4

1
= 0.
ζ

(A.7)

Introducing a new variable y = a2 ζ − (a4 /ζ ), we can rewrite the curve as
y 2 = (a1 x 2 + a0 x + a3 )2 − 4a2a4 .

(A.8)

For each charge vector `(k) , a corresponding complex structure modulus is given by
Y `(k)
ai i .
(A.9)
zk =
i

In the present case we have
a1 a3
a2 a4
zB = 2 .
zF = 2 ,
a0
a3

(A.10)

By setting
a1 = a3 = 1,

a0 = s,

4a2 a4 = K 4 ,

(A.11)

we have
1
K4
.
,
z
=
B
4
s2
GKZ system is defined by a system of differential operators
(k)
(k)
Y  ∂ −`i
Y  ∂ `i
=
.
∂ai
∂ai
(k)
(k)
zF =

`i >0

`i <0

In the F2 case it is given by

(A.12)

(A.13)

342

T. Equchi, H. Kanno / Nuclear Physics B 586 (2000) 331–345

L1 = zF (4θz2F + 2θzF ) + θzF (2θzB − θzF ),

(A.14)

L2 = zB (2θzB − θzF + 1)(2θzB − θzF ) − θz2B ,

(A.15)

θzi ≡ zi


,
∂zi

i = F, B.

A.2. Nf = 1
It is known that matters in the vector representation are generated by blowing up the
manifold (see, for instance, [21]). We expect that a blow up of the local F2 provides the
description of gauge theory with Nf = 1 matter. The corresponding reflexive polyhedra is
No. 6 of [10] and given by the vertices
ν0 = (0, 0),

ν1 = (1, 0),

ν3 = (−1, 1),

ν2 = (0, 1),

ν4 = (−1, 0),

ν5 = (−1, −1).

(A.16)

We see that the charge vectors are
`(B) = (0; 0, 0, 1, −2, 1),
`(F −E) = (−1; 0, 1, −1, 1, 0),
`(E) = (−1; 1, −1, 1, 0, 0).

(A.17)

The constraint
Y3 Y5 = Y42 ,

Y0 Y3 = Y2 Y4 ,

Y0 Y2 = Y1 Y3 ,

(A.18)

is solved by  

tx s
Y = sx, sx 2 , , , s, ζ s .
ζ ζ
Setting s = 1 gives the curve
x
1
PNf =1 = a0 x + a1 x 2 + a2 + a3 + a4 + a5 ζ,
ζ
ζ
or

(A.19)

(A.20)

y 2 = (a1 x 2 + a0 x + a4 )2 − 4a5(a2 x + a3 ).

(A.21)

If we set
a1 = a4 = 1,
a2 = w,

a0 = s,

4a5 = K 3 ,

a3 = w−1 ,

(A.22)

we obtain the curve [11]  

1
.
y = (x + sx + 1) − K wx +
w
Complex moduli are given by
2

zB =

2

a3 a5 K 3
,
=
4w
a42

2

3

zF −E =

a2 a4 w 2
,
=
a0 a3
s

Complete set of differential equations is given by [10]

(A.23)

zF +E =

a1 a3
1
=
.
a0 a2 sw2

(A.24)

(A. A. The curve is given by PNf =2 = a0 x + a1 x 2 + a2 x 1 x2 + a3 + a4 + a5 ζ + a6 = 0.34) A.30) This is a reflexive polyhedra No. (A.25) L2 = (θF −E − θF +E )(θF −E − 2θB ) − zF −E (θF −E + θF +E )(θF −E − θB − θF +E ).26) L3 = θF +E (θF +E + θB − θF −E ) − zF +E (θF +E + θF −E )(θF +E − θF −E ). 8 of [10]. (A.3.30) (No. ν6 = (1. (A.36) . 1). ζ ζ ζ (A. a3 = (w1 w2 )−1 . ν4 = (−1. 0). ν5 = (−1.27) L4 = (θF −E − θF +E )θB − zB zF −E (θF −E + θF +E )(2θB − θF −E ). Nf = 2 We choose the following vertices ν0 = (0. 0). −1).35) or y 2 = (a1 x 2 + a0 x + a4 )2 − 4(a6x 2 + a2 x + a3 )(a7 x + a5 ). 4a5 = K 2 . ν1 = (1. 0).T.32) Substituting the relation a1 = a4 = 1. w1 w2 (A.   w2 w1 a2 = + . = a0 a2 s(w12 + w22 ) zF1 = a2 a4 w12 + w22 . a0 a6 sw12 w22 (A. (A. a6 = w1 w2 . H. (A.33) we obtain the complex structure moduli zB = a3 a5 K2 = . −1) to (A. = a0 a3 s zF3 = a1 a2 w12 + w22 = . ν3 = (−1.29) We note that the last two operators are necessary to obtain a unique double log solution Ω. 1). Kanno / Nuclear Physics B 586 (2000) 331–345 L1 = θB (θB − θF −E + θF +E ) − zB (2θB − θF −E )(2θB − θF −E + 1). ν2 = (0. Nf = 3 Polyhedron for the 3-point blow up is obtained by adding a vertex ν7 = (0. 343 (A. 4w1 w2 a42 zF2 = a1 a3 1 .4. Equchi. a0 = s. (A. 12 of [10]). Elliptic curve is given by PNf =3 = a0 x + a1 x 2 + a2 x 1 x2 + a3 + a4 + a5 ζ + a6 + a7 xζ = 0 ζ ζ ζ (A.28) L5 = θF +E (θF −E − 2θB ) − zF −E zF +E (θF −E + θF +E )(θF −E + θF +E + 1).31) or y 2 = (a1 x 2 + a0 x + a4 )2 − 4a5(a6 x 2 + a2 x + a3 ). 1).

a0 a8 s w32 w42 zF1 = (A.38) A. H.5. z F 4 a0 a6 s w12 w22 a0 a5 s zB = zF2 zF3 (A. w4 w3 (A.344 T. Equchi. a6 = w1 w2 q.39) By eliminating ζ it becomes y 2 = (a1 x 2 + a0 x + a4 )2 − 4(a6x 2 + a2 x + a3 )(a8 x 2 + a7 x + a5 ). zF1 = = w1 + w22 . a5 =   K w1 w2 K + . Elliptic curve is given by PNf =4 = a0 x + a1 x 2 + a2 + a5 ζ + a6 x 1 + a3 + a4 ζ ζ x2 + a7 xζ + a8 x 2 ζ = 0. a5 = . = = + = = . a0 a6 s w12 w22   a1 a7 1 1 1 zF5 = = + . By choosing the variables as (A. ζ (A. a0 = s. a2 = + q.42) . a3 = 4 w2 w1 4w1 w2 Kw1 w2 . 2 4w1 w2 w3 a0 a3 s a4 1 a1 a3 1 = = . a3 = w1 w2 w3 w4   w3 w4 a7 = + . Nf = 4 Polyhedron for the 4-point blow up is obtained by further adding a vertex ν8 = (1. 2 w1 w2 w3 w4 a4 1 a1 a3 1 = = .37) we find the complex moduli  a3 a5 K a2 a4 1 2 = . = a0 a5 s a2 a4 1 2 = (w + w22 ). a6 = 4 a2 = 1 .41) we find the complex moduli a3 a5 q = . −1) to the polyhedron of Nf = 3 (No. . a8 = w3 w4 . a1 = 1. a0 a3 s 1   a1 a2 1 1 1 zF3 = = + . 15 of [10]). a7 = w3 . a4 = 1. w2 w1 1 1 q. w3 (A.40)   w1 w2 a1 = 1. a4 = 1. a0 a2 s w12 + w22 zB = zF2 zF4 = a4 a7 w32 + w42 . a0 a2 s w12 + w22   a1 a2 1 1 1 a4 a7 w32 . Kanno / Nuclear Physics B 586 (2000) 331–345 By choosing the variables as a0 = s.

Seiberg. J. . Phys.C. [8] A. non-trivial fixed points and string dynamics. S. Phys. C. B 387 (1996) 278. Y. Probing F-theory with branes. Yankielowizc. Klemm. [5] M. Phys. P. hep-th/9609071. Five-dimensional supersymmetric gauge theories and degenerations of Calabi–Yau spaces. N. H. Vafa. D. Instanton sums and five-dimensional gauge theories. Ferrara. Kanno / Nuclear Physics B 586 (2000) 331–345 345 References [1] N. Phys.T. Local mirror symmetry: calculations and interpretation. B 388 (1996) 753. Klemm. Seiberg. S. Sonnenschein. Mirror symmetry and exact solution of 4D N = 2 gauge theories I. Chiang. hep-th/9709010. Adv. Lett. Theisen. S. B 531 (1998) 323. Klemm. R. hep-th/9606086. N. Matter from geometry. Seiberg. Nucl. E. Theisen. Five-dimensional gauge theories and relativistic integrable systems.G. hep-th/9408099. Small instantons. Monopoles. Theor. [7] S. Math. Nucl. E. Seiberg. Yau. [14] T. B 475 (1996) 526. S. [2] N. Lawrence. Katz. [4] D. Klemm. Intriligator. N.-M. M.R. B 497 (1997) 146.R. [18] A. Seiberg. Phys. Zaslow. A. Electro-magnetic duality monopole condensation and confinement in N = 2 supersymmetric Yang–Mills theory. Nucl. Cadavid. M. Lett. Katz. Nucl. Witten. S. hepth/9407192. Five-dimensional SUSY field theories. Ceresole. hepth/9603150. On the M-Theory approach to (compactified) 5D field theories. Extremal transitions and five-dimensional supersymmetric theories. F-theory and orientifolds. Geometric engineering of quantum field theory. hep-th/9506144. [21] S. Phys. [17] E. hep-th/9609239. hep-th/9609070. S. B. Nucl. B 431 (1994) 484. Katz. B 426 (1994) 19. E. hep-th/9605150. hep-th/9407087. N. Picard–Fuchs equation and prepotential of five-dimensional SUSY gauge theory compactified on a circle. Phys. Douglas.S. A. Ohta. Vafa. Phys. [10] T. On the geometry behind N = 2 supersymmetric effective actions in four dimensions. [20] A. hep-th/9706110. Lian.H. B 530 (1998) 73.T. [13] A. Brandhuber. Phys.R. hep-th/9608111. hep-th/9903053. D’Auria. [16] A. Nucl. Int. Nucl. hep-th/9605199. Douglas. Morrison. C. Phys. [11] N. hep-th/9801036. B 497 (1997) 56. A. N. Equchi. B 513 (1998) 239. Nucl. S. Phys. Mayr. 1 (1998) 53. B 357 (1995) 76. B 483 (1997) 229. hep-th/9705131. Nucl. C. Nucl. Banks. del Pezzo surfaces and type I0 theory. B 415 (1997) 127. Nucl. Katz. hep-th/9706025. B 497 (1997) 155. A 7 (1992) 6215. Vafa. B 471 (1996) 195. Witten. 11-dimensional supergravity compactified on Calabi–Yau threefolds. Phys. Correlation functions for topological Landau–Ginzburg models with c 6 3. A note on ODEs from mirror symmetry. [9] S. Vafa. Sen. J. Schmidt.R. Phys. Phys. Lett. Morrison. Kanno. Itzhaki. Nucl. Phase transitions in M-Theory and F-Theory. Klemm. [3] N. Phys. hep-th/9609219. Lett. Mod. [6] K. Yau. duality and chiral symmetry breaking in N = 2 supersymmetric QCD. Witten. [12] H. B 497 (1997) 173.-T. Nekrasov. Phys. C. Seiberg. [15] A. [19] A. Phys. hep-th/9702198. S. Roan. Nekrasov. Phys.

PII: S 0 5 5 0 . Here q denotes the momentum of the virtual photon exchanged in the deep inelastic scattering process or the annihilation process.and timelike splitting functions up to next-to-leading order (NLO) and the coefficient functions up to NNLO both for unpolarized and polarized scattering.to the s-channel.3 . The hadronic ∗ Corresponding author. All rights reserved. V. Q2 ). Blümlein a. to all orders in QCD leads to a connection between the two evolution kernels determining the q 2 -dependence of the DIS structure functions and the fragmentation functions.The Netherlands Received 18 April 2000.O. which originally was only derived in the scaling parton model.  2000 Elsevier Science B. and Yan mentioned the possibility [1. van Neerven a. Allahabad.L. Platanenallee 6. 211019.3 2 1 3 ( 0 0 ) 0 0 4 2 2 .Nuclear Physics B 586 (2000) 349–381 www.nl/locate/npe On the Drell–Levy–Yan relation to O(αs2) J. respectively. Box 9506.2] that the deep inelastic scattering structure functions at the one side and the nucleon fragmentation functions in e+ e− pair-annihilation on the other side may be related by an analytic continuation from the t. D-15735 Zeuthen.∗ .c a DESY Zeuthen. Levy and Yan (DLY) connecting the deep inelastic structure (DIS) functions and the single-particle fragmentation functions in e+ e− annihilation which are defined in the space-like (q 2 < 0) and time-like (q 2 > 0) regions respectively. Already before the advent of Quantum Chromodynamics (QCD) Drell.elsevier. Q2 ) where Q2 = |q 2 | or F2 (x.V.b . E-mail: blumlein@ifh. Universiteit Leiden. Levy. Q2 )/∂ ln(Q2 ) and the corresponding fragmentation functions satisfy the DLY relation up to next-to-leading order. Introduction Neutral current deep inelastic lepton–nucleon scattering and single nucleon inclusive production in e+ e− pair-annihilation are formally related by crossing the kinematic channels.V. Germany b Mehta Research Institute of Mathematics and Mathematical Physics (MRI). Chhatnag Road. It is shown that the evolution kernels describing the combined singlet evolution for the structure functions F2 (x. We also comment on a relation proposed by Gribov and Lipatov. 2300 HA Leiden. Q2 ). Ravindran a.de 0550-3213/00/$ – see front matter  2000 Elsevier Science B. ∂F2 (x. India c Instituut-Lorentz. All rights reserved. P. FL (x. An extension of the DLY-relation. Jhusi. W. accepted 29 June 2000 Abstract We study the validity of a relation by Drell. In relation to this we derive the transformation relations between the space. 1.

i. To establish this crossing relation 1 Later on massive vector meson ladder-models were studied in [12. Similar relations hold in the polarized case. One therefore may calculate the respective one-particle evolution kernels and study their behavior under the crossing from the t. for the contributions of lowest twist.to the s-channel.e. it was already shown [1] that for e+ e− -annihilation also diagrams of distinct connectedness appear (cf. also [14–16]) which are absent in DIS so that a proof of Eq. (1) at the non-perturbative level becomes very difficult. However. One may think off a general representation of the structure and fragmentation functions in terms of current– current expectation values. exhibits a similarity with the approach by Drell.28] the kernels become distribution-valued for x = 1 [29]. Levy.7] equations and their generalization [8–11] 1 aiming at a perturbative description of the structure and fragmentation functions.14–21]. In this order these kernels are nothing but the lowest order splitting functions Pkl(0) (x). 2 For a review on the early developments see [22]. (1) was originally postulated assuming that those terms are absent [1]. At that time the physical nucleons were considered as bound states built up of bare nucleons and pions in the context of the Yukawa theory. 2 Within QCD the picture changes. In these theories neither infrared nor collinear singularities are occurring. The interactions were described by Bethe–Salpeter [3–5] or Faddeev-type [6. A further complication in the case of a vector-theory as QCD is the emergence of infrared and also collinear singularities which have an essential impact on the crossing because of the behavior of the kernels at x = 1. p) = −Wµν (q. However.350 J. . and Yan. It is the aim of the present paper to investigate the validity of the Drell–Levy–Yan (DLY) relation. if applied to perturbatively calculable partonic structure functions and quantities related to them up to the level of two-loop order. Wµν (1) Here p denotes the nucleon momentum and q is the 4-momentum transfer to the hadronic system. moreover. 3 Eq. which are obtained from the N(0) inverse Mellin transforms of the anomalous dimensions γkl [31–33]. In leading order for unpolarized scattering the crossing relations mentioned above were given in [30].. / Nuclear Physics B 586 (2000) 349–381 tensors for the space-like process of deep inelastic scattering (DIS) and the time-like single nucleon inclusive reaction would therefore be related by (S) (T ) (q. the non-perturbative contributions factorize. See also the subsequent discussion in [23] pointing out that the exponent of the structure functions ∝ (1 − x)p near x = 1 [24–26] needs not to be integer. Blümlein et al. −p). Here it turns out that a thorough perturbative description of the structure functions and fragmentation functions is not possible. since in the range where single parton states are dominating. with q 2 < 0 for deep inelastic scattering and q 2 > 0 for e+ e− -annihilation.13]. The QCD-improved parton model. where the crossing relation Eq. perturbation theory may be used to describe the scaling violations of these functions at large values of |q 2 | where the running coupling constant is small. 3 Here x denotes the Bjorken scaling variable which will be defined differently for the time-like and space-like region except for x = 1 where both definitions lead to the same value for x. (1) was verified for the respective kernels. The relation could be established for the aforementioned ladder-models [1. As a consequence of the Bloch–Nordsiek theorem [27.

The paper is organized as follows. At asymptotic values of the kinematic variables structure functions only depend on Q2 and the Bjorken scaling variable xB = Q2 . Blümlein et al.and time-like processes both for unpolarized and polarized deep inelastic reactions where we consider two principal examples. We also comment on a relation by Gribov and Lipatov [36] which emerged in the same context. In the case of forward scattering the scattering matrix element can be written in terms of the leptonic tensor Lµν and the hadronic tensor Wµν by |M|2 = Lµν Wµν .J. g) 4 4 Similar relations hold for the polarized structure functions g (x. In Section 3 scheme-invariant combinations of coefficient and splitting functions are constructed for the space.t. 2p.and time-like coefficient functions at O(αs2 ) as well as the convolution relations which are needed for the investigation of the DLY-relation. Other relations between the splitting functions such as supersymmetric relations and relations due to conformal symmetry were discussed elsewhere.r.g.. (2) where ‘X0 represents an inclusive final state. Structure functions and fragmentation functions Deep inelastic scattering (DIS) of a lepton (l) off a hadron target (P ) is described by the process l(k1 ) + P (p) → l(k2 ) + ‘X0 . If the process is mediated by photon only we have i = 1. The Drell–Levy–Yan relation is studied in detail in Section 4. In the context of the parton model the structure functions can be expressed in terms of quark and gluon densities and the corresponding space-like (S) coefficient functions Ci. Basic relations for the deep inelastic structure and fragmentation functions are summarized in Section 2. (3) The hadronic tensor [37–40] contains the unpolarized and polarized deep inelastic structure functions Fi and gi . q 2 . q = k1 − k2 .q 0 6 xB 6 1. Q2 ) and g (x. Section 5 contains the conclusions. In the appendix we present the differences between to the space. 2. conditions are derived for the transformation of the splitting and coefficient functions from the space. Furthermore. 1 2 . q 2 = −Q2 < 0.35]. e. Notice that instead of F1 we can also take the longitudinal structure function FL . When a single gauge boson is exchanged between the incoming lepton and the hadron the above process factorizes into the leptonic part and the remaining hadronic part.k (k = q. 2 in both the polarized and unpolarized case. cf.and time-like processes one has to study scheme-invariant quantities which are the physical evolution kernels for specific choices of observables as the unpolarized and polarized structure and fragmentation functions or derivatives of them w. [34. / Nuclear Physics B 586 (2000) 349–381 351 between space. For the coefficient functions we extend the discussion to the NNLO level..to the time-like case. Q2 ) on the level of twist 2. (4) In QCD the Q2 dependence of the structure functions is only logarithmic and it accounts for the violation of scaling.

in the QCD improved parton model these functions can be expressed in a similar way in terms of parton fragmentation densities Dk (k = q. For convenience this scale is put equal to the factorization scale in the following. z. In Eq. (2). l(k1 ) + l(k q = k1 + k2 . Q2 ) = xE nf X j =1 Z1 ej2 xE "     dz 1 Q2 (T )S S xE 2 . µ2f . µf Ci. (5) where ej denotes the charge of the j th quark flavor and Nf represents the number of light flavors. satisfy renormalization group equations which will be shown below. g) and 1Ci. µ2f + fq¯i z. 2). This information is contained in the fragmentation functions observed in the reaction [1] ¯ 2 ) → P¯ (p) + ‘X0 . Therefore. Whereas in deep inelastic scattering the constituent structure of the nucleons is studied. L. However the parton densities and the coefficient functions. µ2f ) (k = q. 2 f z Nf q z µf       # xB 2 Q2 Q2 (S) (S)NS NS xB 2 . µf Ci. In this case the polarized parton densities and polarized coefficient functions are denoted by 1fk (z. in Eq. Corresponding formulae hold for polarized scattering. Blümlein et al. µ2f . 2 D z Nf q z µf . (5) the index (S) in the structure functions indicates the space-like nature of the process (q 2 < 0).q z. hadroproduction at e+ e− colliders provides us with information about the fragmentation process of these constituents into the hadrons. The scale µf . z z µf µf i = 2. / Nuclear Physics B 586 (2000) 349–381 (S) Fi (xB . 2 + fqj .352 J. 2 . Q2 ) = xB Nf X j =1  + fg Z1 ej2 xB "     dz 1 S xB 2 Q2 (S)S . In addition one encounters a dependence on the renormalization scale µr which arises in the renormalization procedure. (5) appear the singlet ‘S’ and non-singlet ‘NS’ combinations of parton densities which are defined by fqS z. µ2f − fqNS i Nf (7) respectively. Q2 /µ2f ) (i = 1. which do depend on these scales. (6) i=1 and     1 S fq z. (8) where the symbols have he same meaning as in Eq. i.k (z.. Fi(T ) (xE . q 2 ≡ Q2 > 0. µ2f + fq¯i z. appearing in the above equation. µ2f  = Nf X    fqi z.g z. denotes the factorization scale which is introduced while removing the collinear singularities from the partonic structure functions. µf Ci.q z. Furthermore. Notice that the structure functions Fi and gi do not depend on these scales. µf Ci.q z. g) multiplied by time-like coefficient functions.e. These fragmentation functions are the analogues of the DIS structure functions. µ2f = fqi z.

Q2 − FL(R) x.42] all the final state mass singularities are canceled too since the DIS structure function is an inclusive quantity. After coupling constant renormalization the hadronic structure function can be written as follows .g z. This entails the definition of the polarized fragmentation functions (T ) (T ) denoted by g1 and g2 for which one can present a similar formula as in Eq. (9). L. Scheme-invariant combinations In this section we give a short outline of the origin of the factorization scheme dependence of the anomalous dimensions (splitting functions) and the coefficient functions. µf Ci. First these divergences have to be regularized for which the most convenient way is to choose the method of n-dimensional regularization. (8) where the hadron P is polarized. the definitions for the singlet and non-singlet parton fragmentation functions are the same as those for the parton densities given in Eqs. The partonic structure functions bi. (5) where like in DIS we set the renormalization scale equal to the factorization scale. (11) 3. R = T (x = xE ). Q2 0 6 xE 6 1. The first one originates from the ultraviolet region. k = q.t. contain denoted by F various divergences.q z. (7). Similarly as in DIS one can also study the annihilation processes in Eq. g). An example is the coupling constant which becomes equal to αs (µ2r ) where µr is the renormalization scale. L. The discussion below deals with the DIS structure functions but the conclusions also hold for the fragmentation functions. Very often one also encounters the transverse structure function which in the time-like and space-like case is given by   i 1 h (R) F2 x. (10) The symbol T appearing within parentheses in Eq.q .r.28]. / Nuclear Physics B 586 (2000) 349–381  + Dg       # xE 2 Q2 Q2 (T ) (T )NS NS xE 2 . with n = 4 + . (9) is defined by xE = 2p. Using this method the singularities reveal themselves in the form of pole terms of the type 1/ j . The infrared divergences cancel between virtual and bremsstrahlung in the quantity F contributions by virtue of the Bloch–Nordsieck theorem [27. Blümlein et al. F1(R) x. (6). Due to the Kinoshita– Lee–Nauenberg theorem [41. bi. 2. 2 + Dqj . Q2 . z z µf µf 353 i = 2. The scales µf and µr are defined in the same way as in Eq.k (i = 1. (9) denotes that the fragmentation functions are measured in time-like processes. This type of singularities is removed via a redefinition of the parameters appearing in the QCD Lagrangian. µf Ci.k . (9) where the corresponding scaling variable for the process in Eq. representing the QCD radiative corrections. Furthermore. We also show how this dependence disappears in the evolution of the structure functions w.J. 2 . Q2 = 2x with R = S (x = xB ). the kinematic variable Q2 . Then one is left with only two types of singularities.

2 . The second type of singularity originates from the collinear region which can be attributed to the vanishing mass of the initial state parton represented by either the (anti-)quark or the gluon.354 J. Q . / Nuclear Physics B 586 (2000) 349–381 X Fi (x. Hence physical quantities are invariant under scheme transformation.  F µ2 µ2r !    X  2 µ2f  µ2f µ2f 2 Q 2 (14) Ci. 2 (x). The renormalization group equation of the parton densities . (12). 2 .  ⊗ fk (x). 2 . the coupling constant. 2 . 2 . mass factorization leads to the introduction of a scale µf called mass factorization scale which is related to the factorization scheme dependence. Blümlein et al. The change of the parton densities and the coefficient functions with respect to a variation in the scales µr and µf is determined by the renormalization group equation (RGE) [43–45].g This procedure provides us with the coefficient function denoted by Ci. 2 .  ⊗ fˆk (z). (15) Fi (x. 2 .  (z).r. Q ) = 2 k=q. 2 . 2 .l αs µr .g where the renormalized parton density is defined as !     X  µ2f µ2f  µ2f µ2f 2 2 Γlk αs µr .t. Substitution of Eq. 2 = µ µr µ µr (16) k=q. 2 . µ µr (12) where the symbol ⊗ denotes the Mellin-convolution defined by Z1 (f ⊗ g)(z) = Z1 dz2 f (z1 )g(z2 )δ(z − z1 z2 ). Hence the  in Eq. 2 ⊗ fl αs µr . Q ) = µ µr µf µr l=q.g !    2 2 2 Q µ ˆ Fik αs µr . = µ µr µf µr l=q. Like in the case of renormalization. µ .l αs µr . 2 ⊗ Γlk αs µr . (14) into Eq. The former only shows up in the parton densities and the coefficient functions and it only disappears in specific combinations representing physical quantities.g Since the mass factorization can be carried out in various ways one is left with an additional scheme dependence which comes on top of the renormalization scheme dependence entering the coupling constant in Eq. The parameter µ originates from n-dimensional regularization because in this method the coupling constant gets a dimension. (12) represents the collinear singularities which are removed from the partonic structure function via mass factorization and transferred to a transition function Γlk as follows    2 2 bik z. Like in the latter case µf drops out in physical quantities as the DIS structure functions or fragmentation functions. αs µ2r . fˆk is defined as the bare parton density which is scale independent and is an unphysical object because of the singular behavior of Fik . fl z. αs µr . (12) leads to the result    ! X  2 µ2f  µ2f µ2f 2 2 Q 2 Ci.l . dz1 0 (13) 0 Furthermore. Notice that the latter depends on the scale µr and therefore on the renormalization scheme w.

(24) + C¯ i.q ⊗ Zqq + C¯ i. Ci. 2 . 2 . (23)     (1) (2) (1) (2) (1) (1) (1) + Zqg + Zqg + Zqg ⊗ Zgg + Zqq + as2 C¯ i.g {m.q = δ(1 − z) + as C¯ i. µ  2  αs µf . 1 = δ(1 − z).g will not alter the physical observable like. The latter takes the following form       1 ∂ ∂ 2 2 + β a  1δ − P µ a µ . the beta-function is defined by  2    2 2 d a s µr = −β0 as2 µ2r − β1 as3 µ2r · · · (18) β a s µr ≡ µr 2 dµr The same equation as in Eq. / Nuclear Physics B 586 (2000) 349–381 355 follow from the one presented for the transition functions Γlk . The relation between the splitting and coefficient functions computed in two different schemes is found to be X X   d Zlm ⊗ P¯mn ⊗ Z −1 nk − 2β(as ) Zlm ⊗ (21) Z −1 mk .q ⊗ Zqg .g = as C¯ i.  (z) = 0.q + Zqq  (1) (1) (1) (1) + C¯ i.m ⊗ Zml Γlk → m=q.q + Zqq Ci. (17) also applies to the parton density because of the definition in Eq.m ⊗ Z −1 ml .l → .. z) appearing in the above equation are the splitting functions.l = m=q.g (20) m=q. Notice that we have chosen here µ2f = Q2 in order to get rid off the logarithms ln(Q2 /µ2f ) which usually appear.g X  C¯ i. Up to O(as2) one obtains     (1) (2) (1) (2) (1) 2 (1) (1) + Zqq + Zqg ⊗ Zgq + as2 C¯ i. 1.m as µ2f . The scale dependence of the coefficient function in Eq. C¯ i.g ⊗ Zgg + ···. The functions Pij (as .g Below we present the relation between the coefficient functions computed in two different schemes up to order as2 (Q2 ). µf As has been mentioned above scheme transformations such as X X −1 Zlm ⊗ Γ¯mk .  + β a s µf µf 2 ∂µ2f ∂as µ2f !    Q2 (19) ⊗ Ci.g. (16).n}=q.g ⊗ Zgq + ···. the structure functions and fragmentation functions. Blümlein et al. 1 (z) = 0.g  (1) (1) (1) (1) + C¯ i. (15) is given by       1 ∂ 2 ∂ 2 2  1δlm + Plm as µf .g Ci. e. . (22) Ci.  µ2f s lm lm s f f 2 ∂µ2f ∂as µ2f !    µ2f 2 ⊗ Γmk as µf .J. Plk = da s m=q. Furthermore. (17) as µ2f ≡ 4π where we have set µr = µf for simplicity.

(13) can be written as Z1 (f ⊗ g) = dz zN−1 (f ⊗ g)(z) = f N · g N . (19) and (15). 2 CI. Their influence is expected to become smaller when higher order terms in the perturbation series are taken into account [50]. B. the independence of the structure function on the scales µf and µr is not manifest when multiplying parton densities with coefficient functions. N (26) 0 Since the structure functions are scheme independent they become renormalization group invariants. Q2 ) = 0. (28) + fgN as µ2f . B. µ2f s f ∂µ2f ∂as µ2f The equation above follows from combining Eqs. For such an equation one needs two different structure functions called FA (x. Q2 ).l above has the form  N  N N  N  CAq CAg fq FA = . To avoid the problem of the factorization scheme dependence of the structure function when the perturbation series is truncated up to fixed order it is better to study evolution equations for the structure functions with respect to a physical scale which is represented by a kinematic variable like Q2 . / Nuclear Physics B 586 (2000) 349–381 In the subsequent part of the paper it is much more convenient to derive the expressions in the Mellin transform representation so that one can avoid the convolution symbol ⊗. 2 .356 J. In these type of evolution equations the kernels are factorization scheme independent order by order in perturbation theory. The Mellin transform of a function f (z) is given by Z1 f (N) = dz zN−1 f (z). the evolution equation for the non-singlet structure functions is even more simple. e. g) as matrix elements so that the equation Here one can view the CI. 2 Q0 µf     2  µf  Q2 N as µ2f . Limiting ourselves to the singlet case. However. l = q.q as µf .. Examples are A = 2 and B = L or FA and Q2 dFA /dQ2 . one can write      µ2f  Q2 N 2 N 2 N 2 FI (Q ) = fq as µf . However. Blümlein et al. [46–49]).g Q0 µf N (I = A. 2 CI. (25) 0 In this way Eq. I = A. the dependence on the choice of renormalization scheme and therefore the dependence on µr remains so that one is able to obtain a better estimate of the theoretical error on αs . (17). In particular when the perturbation series of a physical quantity is computed up to finite order there is a residual dependence on these unphysical scales (see. Hence they satisfy the RG equation    ∂ ∂ 2  + β a (27) µ FiN (x. Q2 ) and FB (x.g. (29) N N N FB fgN CBq CBg .

l . 2 = CI. 1 γmn as Q2 C N n. (32) also applies to FIN in Eq. 1 Tas exp − 2β(x) µf  ml as µ2f where γ N is the anomalous dimension matrix defined by Z1 γlkN =− dz zN−1 Plk (z).J.1 C = β(as Q s m.J ∂t β(as Q2 ) ∂t (36) (37) The kernels KINJ depend both on the anomalous dimensions γlkN and the coefficient N (Q2 ) but the latter two quantities are combined in a factorization scheme functions CI.J (t) − KI J = −4 mn n. CI.m    2 N −1 2  a Q . C m. (30) dx CI.m 2 µf One can show that the expression above is invariant under scheme transformations. N N FBN KBA KBB FBN where the physical (scheme invariant) kernel is given by    N (t) ∂CI.g {m.g Q2 -dependence only resides in the coefficient function the same evolution Since the equation as in Eq.J as Q2 .n}=q. C m.I .k as µ2f . (31) 0 We will now differentiate the coefficient functions w.l m=q. γlkN = ∂a s m=q.r. (28).m   β0 as Q2 N N −1 N N N −1  CI. 1 CJ. Q2     N N ∂CI.t. For a short-hand notation we introduce the evolution variable t   as Q2 2 .m (t)γ (t) C (t) . Blümlein et al.g X X −1 −1  N N N ¯ N −1 C¯ I. (35) t = − ln β0 as (Q20 ) so that we obtain   ∂ FAN 1 =− 4  N KAA N KAB  FAN  . (19) and the solution is given by the T-ordered exponential [51]  )#! " ( asZ Q2    Q2   γ N (x) N 2 N 2 . 2 .l as µf . (32) − CI.m = C N l.J ∂as Q2      N  −1   N  Q2 1 N as Q2 .g m=q. Q2 /µ2f  ∂CI. The latter are given by X X −1 −1 N N N ∂ Zlm γ¯mn Z N nk + 2β(as ) Zlm (33) Z N mk .k 2 2  Q = β as Q ∂Q2 ∂as Q2    N as Q2 .m as Q . Q2 /µ2f as µ2f . 1  ∂CI.k as µ2f .I = Zlm (34) Z N ml . / Nuclear Physics B 586 (2000) 349–381 357 The coefficient functions satisfy the RG-equation in Eq.

Blümlein et al. Q2 = n=0 n=0 l. Therefore this normalization accounts for keeping the same order in the coupling constant for the two quantities     FLN Q2 N(S) (40) Q2 . (38) one can compute order by order the coefficients in the perturbation series of the kernel KINJ = ∞ X  (n) asn Q2 K N I J . N(1) qg CL. Q2 ) and FL (x. F2 (x. (39) is resummed in all orders. Q2 ) Let us consider now two specific examples. e.l asn Q2 C N I. N(0) N(0) = γqq − K22 N(0) KL2 N(0) = γgq − N(0) N(0) = γgg + KLL N(1) CL. This factorization scheme independence of KINJ holds order by order in perturbation theory. + N(1) CL.g N(0) N(0) K2L = γqg . I = A. (39) n=0 (n) Notice that the coefficients (K N )I J are not invariant with respect to a finite renormalization of the coupling constant. ∞  X  (n) N CI. Eq. Q2 )/∂t as the observables FA. Q2 ).g (1) (1) Since both the coefficient functions CLq and CLg are scheme invariants such a normalization is possible. Q2 ) or the structure function F2 (x. (11). cf. g. / Nuclear Physics B 586 (2000) 349–381 independent way. [59].. This dependence is removed when the perturbation series in Eq. Using the series expansions for the anomalous dimensions and coefficient functions in terms of the strong coupling constant γlkN = ∞ X  (n) asn+1 Q2 γ N lk ..q N(1) CL. (41) .g To next-to-leading order in as (Q2 ).l . The lowest order contribution is well-known.1. choosing the structure functions F2 (x.g N(1) CL. We expand now the kernels KINJ for this choice of observables into a series in as . This is because FL (x. Q2 ) to its gluonic contribution in lowest order.q γ N(0) . 3. N(1) qg CL.g N(1) !2 CL. contrary to the structure function F2 (x.q N(0) γqg N(1) CL. k = q. Q2 ) and FL (x. Q2 ). In this combination of observables it is convenient to normalize the structure function FL (x. cf.B (x.q N(0) γ . Q2 ) vanishes in zeroth order of αs due to the Callan–Gross relation. FBN Q2 = FAN Q2 = F2  N(1) . one finds  N(0) N(0) − γgg γqq .358 J. B. Q2 ) and its slope ∂F2 (x.g. as Q2 CL.

g " N(2) # N(1) !2 N(1) CL.g CL.g N(1) N(1) N(0) N(0) − + C − + C2.g CL.q CL.q CL.g ! N(1) N(1) CL.q N(1) N(1) N(0) − + C2.g CL.g " N(2) # N(1) N(1) N(2) N(1) !2 CL.g γqq + N(1) N(1) CL.g γqq β0 β 0 CL.g CL.g γgg + 2β0 C2.g CL.J.g CL.q + C2.g CL.g CL.g N(1) N(2) CL.g   β1 N(0) N(1) N(1) N(1) N(1) N(0) N(0) − γqg − C2.g CL. CL. / Nuclear Physics B 586 (2000) 349–381  N(1)  N(1) CL.q β1 N(0) β1 N(0) N(1) N(1) − γ − γ γ − γ − qg gg N(1) N(1) β0 qg β0 gg CL.g N(2) N(1) N(2) ! CL.g − N(1) γqg .q N(1) N(0) N(1) N(0) − N(1) N(1) + C2.q + N(1) γgq N(1) N(1) 2.g CL.g " N(1) !3 N(1) N(2) N(1) !2 N(2) CL.q CL.g CL.g CL.q CL.q N(1) CL.g .g CL. (45) CL.g CL.g  N(1)  β1 N(0) CL.g N(1) N(1) N(0) + C2.q CL.g N(1) N(1) N(1) N(1) CL.q γqg + C − C2.q CL.q γgg N(1) N(1) CL.g CL.g K2L = γqg β0 N(1) N(2) ! CL.g CL.q N(1) N(0) N(1) N(1) − N(1) C2.q N(1) N(0) − N(1) C2.q CL.g γgq γqg 2.q − N(1) C2.q CL.g CL.g CL.q L.g CL.g CL.g # " N(2) N(1) N(1) !2 CL.q β1 N(0) CL.g CL.q β1 N(0) N(1) N(0) N(1) γqq − N(1) γqg − γqg + N(1) C2.q C C L.g N(1) N(1) N(1) N(0) N(0) − C2.q CL.q CL.q β1 N(0) N(1) N(1) N(1) = γgq − γgq + N(1) γqq − γqq KL2 β0 β0 C N(1) K22 359 N(1) = γqq − (42) (43) L.g γgq + N(1) C2.q CL.g CL.g # N(1) !2 N(1) N(2) ! CL.g CL.g + 2 N(1) N(1) − N(1) N(1) N(1) CL.q CL.g γqq + N(1) N(1) CL.q CL.q CL.g  N(1)  β1 N(0) CL.q CL.g  .g γqg − N(1) C2.g N(1) − C2.g γqq − γgg + 2β0 C2.g N(1) N(0) N(1) N(0) − C2. CL.g − N(1) C2.g γgg + 2β0 N(1) .q CL.q + 2β0 − N(1) N(1) .g " N(2) N(1) !2 N(1) N(2) # CL.g   N(1) 2  N(1)  CL.g CL.g CL.q CL.g CL.g CL.q + N(1) C2.q β1 N(0) N(1) N(1) N(1) KLL = γgg − γgg + N(1) γqg − γqg β0 β0 CL. Blümlein et al.q CL. (44) N(1) CL.

N(0) Kdd N(0) N(0) = γqq + γgg . (50) − N(0) 2 γqg  β1  N(0) N(1) N(1) N(1) N(0) + γgg − γqq + γgg Kdd = γqq β0  i 2β0 h N(1)  N(0) N(1) N(0) N(1) − 2β0 − γqg − 2β1 . 3.2. The observables FA. / Nuclear Physics B 586 (2000) 349–381 It is evident that the representation in terms of Mellin-moments is of advantage when compared to the corresponding x-space expressions. Q2 ) are here   FAN Q2 = F2(S)N Q2 . Although enforced by Eq.q γqq − 2β0 2 N(1) h i  β0 C2. Kd2 = γqq 4 The next-to-leading order kernels read: K22 N(1) K22 (46) = 0. (47) (48) N(1) K2d = 0. (47)–(51) are scheme-independent by an explicit calculation.   1 N(0) N(0) N(0) N(0) N(0) γgg − γqg γgq .g γqq − γgg γqg For this combination in next-to-leading order the evolution depends on two evolution kernels only. (51) + 4β0 C2. Blümlein et al.360 J.B (x. In leading order the scheme-invariance .k kl the expressions above which in general is not possible.g N(0) 2 N(0) N(0) N(0) N(0) N(0) − γ γ + 2γ γ − 2β γ γ − 0 qq qq gg qg gq qq N(0) 2 γqg ! N(0) N(1) γqq γqg β0 N(1) − γqq . Both quantities are well measurable in the present-day deep inelastic scattering experiments. In leading order one obtains N(0) N(0) = 0. In the case of polarized deep inelastic scattering similar relations apply considering the structure function g1 (x.q − N(0) C2. ∂t This example has been considered before in [54]. In the latter case one has to find the N and γ N appear in the denominators of inverse Mellin transforms of quantities where Ci. Q2 ) and its slope. Q2 ) and its slope. which we have done using Eqs.   ∂ FBN Q2 = F2(S)N Q2 . Q2 ) and ∂F2 (x. (34) for the next-to-leading order contributions. The anomalous dimensions and coefficient functions of the unpolarized case have to be substituted by those for polarized scattering. (36) one has still to show that the kernels (41)–(45). K2d = −4. Q2 )/∂t A second example concerns the structure function F2 (x. (33). (49) h i 1 N(1) N(0) N(1) N(1) N(0) N(1) N(0) N(0) N(1) γqq + γgg γqq − γqg γgq − γqg γgq Kd2 = γgg 4     β0 N(1) N(0) β1 N(0) N(0) N(0) N(0) N(0) γgg − γgq γqg + γgg − 2β0 γqq + C2. F2 (x.

Levy and Yan [1]. The crucial difference. (36). the scaling structure and fragmentation functions satisfy the following relation 5 :   1 (S) (T ) . In the original work of DLY [1] the Yukawa theory was discussed which does not contain gauge bosons. Wµν (52) where the momenta within the respective parentheses of the above quantities are the same as those defined in the beginning of the paper. . as in Eq. since the leading order anomalous dimensions and the lowest order coefficient functions for FL (x. the deep inelastic e+ e− annihilation can be related to matrix elements of hadronic electromagnetic current operators similar to that of deep inelastic lepton–hadron scattering. With the help of the evolution Eq.and time-like quantities in perturbative QCD.q → ∞ and q 2 = Q2 . Nevertheless. From the structure of the hadronic S (q. Relations of this kind are henceforth called Drell–Levy–Yan (DLY) relations.J. 2. Blümlein et al. apart from the ones which originate from the kinematics. 4. p. p) (space-like) and W T (q. p. Even if a crossing relation for these quantities does not exist. One condition to ask such a question at all is that the behavior of all contributing parts under crossing from space. although the original reasoning of these and other authors was quite different. L. Here we first briefly illustrate the idea behind the work of DLY for completeness. p) (time-like) and using the standard reduction tensors Wµν µν formalism one can infer that T S (q. cf. [60]. one still may investigate whether it exists for the perturbative evolution kernels.to time-like momentum transfer are controlled. A further condition for this investigation is that the latter quantities are scheme-invariant. This makes it possible to study the respective evolutions kernels without reference to the nonperturbative input densities. i = 1. This motivated DLY to study the process in detail and then relate it to the deep inelastic scattering process. is that the annihilation process is not related to the forward Compton amplitude contrary to deep inelastic scattering because in the former process the hadron is observed in the final state. (36) we are now prepared to ask for the validity of crossing relations between different space. Drell–Levy–Yan relations In the following we study in detail an interesting relation between deep inelastic lepton– hadron scattering and e+ e− annihilation into a hadron and anything else. p) = −Wµν (q. both processes are related by crossing symmetry which any field theory enjoys. proposed by Drell.q → ∞. Q2 ) are scheme invariants. / Nuclear Physics B 586 (2000) 349–381 361 is visible explicitly. respectively. (53) Fi (xB ) = −(−1)2(s1+s2 ) xE Fi xE 5 Here we indicate the overall signs in case of the scattering of particles of different spin. In field theory. At large momentum transfers |q 2 | the single parton picture applies and the non-perturbative parton densities factorize. −p). In the Bjorken limit for both deep inelastic scattering and deep inelastic annihilation for q 2 = −Q2 .

/ Nuclear Physics B 586 (2000) 349–381 Here it has been assumed that non-perturbative input parton densities can be decoupled (T ) trivially and are the same. Then we show how these relations are preserved for the physical quantities by looking at the kernels discussed in the previous section. The Drell–Levy–Yan relations at leading order In the case of the scheme-independent evolution kernels describing the evolution of F2 (x. Blümlein et al. Here the distribution (lni (1 − z)/(1 − z))+ is represented by   i lni (1 − z) lni+1 δ ln (1 − z) + θ (1 − δ − z) . only the transformation of two combinations of the leading order anomalous dimensions has to be considered. It is particularly interesting to study the above transformation at the level of the splitting functions and coefficient functions which constitute the physical quantities such as the structure and fragmentation functions. Q2 )/∂t. i = 0. Pqq (1 − z)+ 2   (0) (z) = 8TR Nf z2 + (1 − z)2 . Here the relation is already violated up to oneloop order for the coefficient functions. 1−z + which destroy the continuation through z = 1. Q2 ) and ∂F2 (x. (47). In both quantities the color factors of the off-diagonal elements enter only as a product. Apart from scaling violation one also encounters distributions of the type   i ln (1 − z) .362 J. It turns out that the DLY-relation is violated for the coefficient functions and splitting functions separately because both are scheme dependent. 4. respectively. This is true only when the continuation is smooth. etc. . cf. In this section. if there are no singularities for example at x = 1. This relation is called DLY-relation in the literature. (41). . the functions Fi (xE ) are the analytic (S) continuations of the corresponding functions Fi (xB ) from 0 < xB 6 1 to 1 6 xE < ∞. (54) δ(1 − z). . i. (57) Pqg  2  (0) 2 (58) 1Pqg (z) = 8TR Nf z − (1 − z) . In other words. we study this property in more detail extending earlier work [34]. Q2 ). . (0) (z) = 4CF Pgq 1 + (1 − z)2 . or its polarized counterpart for the structure function g1 (x. This in particular happens when we adopt the MS-scheme. These are the determinant and the trace of the singlet anomalous dimension matrix at leading order. z (59) . Although one can choose other schemes in which Eq.e. (55) = δ(1 − z) 1−z (i + 1) (1 − z) + where δ  1.. Let us start with the simplest examples and consider the scheme-invariant evolution kernels Eqs. (53) is preserved (see [34]) up to one-loop order we do not know whether this will hold up to any arbitrary order in perturbation theory. 2.1. The unpolarized and polarized leading order splitting functions read   3 1 + z2 (0) (0) (56) (z) = 1Pqq (z) = 4CF + δ(1 − z) . (47). 1.

– The off-diagonal elements of the singlet splitting functions matrix have to be multiplied by CF /(2Nf Tf ) for Pqg and 2Nf Tf /CF for Pgq . care should be taken when dealing with quark and gluon states which was not the case in the work by DLY. Blümlein et al. NLO splitting function As we know. which is indeed the case in most of the schemes.e.g. where a color and flavor neutral field theory was discussed. it simply amounts to finding finite renormalization factors. CF z z 363 (60) (61) (62) (63) where one demands δ(1 − z) → −δ(1 − z). Pgg have to be multiplied by (−1).65] that by an appropriate modification of the continuation rule in the MS scheme one can show that the time-like splitting functions are related to their space-like counter parts.. For the second set of physical evolution kernels the DLY-relation follows at leading order referring to the transformation relations for the leading order longitudinal coefficient functions. / Nuclear Physics B 586 (2000) 349–381 1 − (1 − z)2 . (79) in an analogous way. 1Pgg (z) = 8CA (1 − z)+ (0) 1Pgq (z) = 4CF The crossing relations of the leading order splitting functions are     CF (0) (0) 1 (0) (0) 1 . It was demonstrated by Curci.2. in the MS scheme characteristic of n-dimensional regularization. the validity of the DLY-relation for this case. (64) Eq. In addition to this.. z   1−z z (0) + z(1 − z) + 2β0 δ(1 − z). . accounting for the interchange of the initial and final state particles under crossing. + Pgg (z) = 8CA (1 − z)+ z   1 (0) + 1 − 2z + 2β0 δ(1 − z). It was shown that the finite renormalization factors can be constructed from the -dependent part of the splitting function when computed in dimensional regularization [66]. (78). Since the modification of the continuation rule has to do with the scheme one adopts. the naive continuation rule for these quantities may be violated. Hence.J. 4. Pgg = −zPgg . The transformation rules are: – The diagonal elements of the space-like flavor singlet splitting functions Pqq . = −zPqq = zPqg P¯qq z 2Nf Tf z     2Nf Tf (0) (0) 1 (0) (0) 1 ¯ ¯ Pgq = zPgq . P¯qg . Eqs. (63) is easily verified and implies the validity of the crossing relation from spaceto time-like evolution kernels Eq. Furmanski and Petronzio [64. (47). the splitting functions and coefficient functions are not physical quantities due to their factorization scheme dependence. respectively. i. e.

66] as well. (12) from regularization. P¯qq  (1)(S) (1)(T ) (T )(1) (T )(1) (0) (0) (0) − Pgq = −2β0 Zqg + Zqg ⊗ P¯gg − P¯qq + P¯qg P¯qg   (T )(1) (T )(1) ⊗ Zqq − Zgg . 2 whereas in the polarized case aqq = agg = 0. The z-independent constant aij . P¯gg (67) (68) (69) (70) where the quantities with a bar denote that they are continued from z → 1/z with the appropriate factors in front. (1)(S) (1)(T ) (T )(1) (T )(1) (0) (T )(1) (0) − Pgg = −2β0 Zgg + Zgq ⊗ P¯qg − Zqg ⊗ P¯gq . Pgg (z) = −zPgg . 63.   (1)(S) (1)(T ) (T )(1) (T )(1) (0) (0) (0) − Pqg = −2β0 Zgq + Zgq ⊗ P¯qq − P¯gg + P¯gq P¯gq   (T )(1) (T )(1) ⊗ Zgg − Zqq . In dimensional bi. (66) one finds that (1)(S) (1)(T ) (T )(1) (T )(1) (0) (T )(1) (0) − Pqq = −2β0 Zqq + Zqg ⊗ P¯gq − Zgq ⊗ P¯qg . zP Pqq (z) = −zPqq z 2Nf Tf qg z (71)     2Nf Tf (n) (n) 1 (n) (n) 1 ¯ ¯ zPgq . aij = 0. (65) ln(δ) → ln(δ) + iπ. one has the function Pij(0) along with ln(z). These quantities read in explicit form:     CF (n) (n) 1 (n) (n) 1 ¯ ¯ . Such an average is not needed for the annihilation process since here . For unpolarized scattering they read 1 aqg = − . The renormalization factors appearing in the Eqs. The constants aij are different in the unpolarized and polarized case.364 J. results from the polarization average.65. Since the pole is always associated with the splitting functions. Pqg (z) = . Blümlein et al. which is also multiplied by the splitting function.k in Eq. when one continues the partonic structure function F the space-like to the time-like region one obtains an additional factor z which when multiplied with the pole in  yields ln(z). (67)–(71) remain true for the polarized splitting functions [62.64. 2 (73) (74) The logarithms in the renormalization factors originate from the kinematics.67] and the continuation rules ln(1 − z) → ln(1 − z) − ln(z) + iπ. For deep inelastic scattering one averages the processes with one gluon in the initial state by a factor 1/( + 2). 1 agq = . (67)–(70) are given by  (72) ZijT (1) = Pj(0) i ln(z) + aj i . / Nuclear Physics B 586 (2000) 349–381 Note that these transformations are automatically accounted for in the case of the leading order physical evolution kernels discussed in the previous paragraph. Keeping this in mind and using the known splitting functions [61. Pgq (z) = CF z z The relations given in Eqs.

At next-to-leading order. Hence.J. (77) are essential to get the constant ζ (2) right when one goes from the space-like to the time-like region. This implies that there are no pole terms in bL. NLO coefficient functions Now. (65) one would obtain an additional term 12ζ (2) on the right-hand side of Eq.q (z) + z C1.q z     CF 1 (T )(1) (S)(1) 1 (T )(1) (z) − . The second term originates from the polarization average which is again absent in the time-like case. / Nuclear Physics B 586 (2000) 349–381 365 the gluon appears in the final state. This leads to the following continuation rule  2  2 Q Q → ln 2 − iπ.65] where also the relations for the NLO non-singlet coefficient functions were presented. The zeroth order longitudinal coefficient functions are identically zero so that the first order contributions are scheme independent. one has to replace the space-like q 2 by the time-like q 2 in addition to Eqs. The same also holds when other regularization methods for the collinear divergences are chosen. The transformation behavior of the non-singlet splitting functions in NLO have been worked out in [64. It is worth noticing that if one would replace ln(1 − z) → ln(1 − z) − ln(z) contrary to the prescription in Eq. (65)–(64). The pole term originates from the collinear divergence in n-dimensional regularization. (76) C = Zqg 2zC1.or n-dimensional polarization average. L. Notice that the average over the polarization sum does not show up in the polarized structure functions. The continuation rules given in Eqs. in the MS scheme [68–70].k . (65)–(66).g 2Nf Tf z Since the coefficient functions depend on the hard scale of the process. The leading order transverse coefficient functions are identical. (72) as follows:    (T )(1) (S)(1) 1 (T )(1) = Zqq . The time-like phase space acquires an extra factor z which gives a finite contribution when being multiplied with the pole terms 1/. 4. k = q. g). The difference which is 12ζ (2) can be understood to originate from the one-loop vertex correction when one continues from the space-like to time-like region in Q2 . let us study how space-like and time-like coefficient functions are related. Therefore in this case the constants aij are zero. (77) ln 2 µf space-like µf time-like The Z-factors get contributions from two sources. (75) C1. Notice that the space-like coefficient function contains −4ζ (2)δ(1 − z) and the time-like one contains 8ζ (2)δ(1 − z). (75). We find .g 2 1. The coefficient functions are expected to violate the DLY-relation due to their scheme dependence. the coefficient functions are related by the Z-factors in Eq. The first one is z-dependent and comes from the phase space integrals.k (z) (i = 1. Here we first present the relations between the space-like and time-like coefficient functions Ci. Blümlein et al.3. there is no left-over finite piece the corresponding partonic structure function F  which could arise from the z .

1. (85) Li3 1 − z 6   1 1 (86) = Li3 (−z) + ln3 (z) + ζ (2) ln(z). (80) = Zqq − CL. We have found this pattern by comparing the scheme transformation which we derived in the last section.q 2 z      z (1)(S) 1 CF z (S)(1) 1 (T )(1) (T )(1) ⊗ CLq ⊗ + Zgq . NNLO coefficient functions 4. 2 z     CF 1 (T )(1) (S)(1) 1 CL.80] for the time-like case.2 (1 − z) − Li3 (1 − z) + ln(z)Li2 (1 − z).4.366 J. The reason for this structure relies on the fact that CL.g 2 2Nf Tf z      z CF 1 z (1)S 1 (S)(1) (T )(1) (T )(1) + Zgg .2 1 − z 6   1 1 = ln3 (z) + S1.2 − z 6 If we do not continue ln(1 − z) and replace ln(1 − z) → ln(1 − z) − ln(z).q (z) − CL.q ⊗ = Zqg − CL.4.g 2 z 2Nf Tf 2 z The right-hand side of the above equation contains the convolutions of Z-factors with the continued NLO longitudinal space-like coefficient functions. then terms proportional to ζ (2) are not compensated between space-like and time-like coefficient functions and hence the relations given in Eqs.g (z) + = 0. / Nuclear Physics B 586 (2000) 349–381   z (S)(1) 1 (T )(1) CL.g 2 2Nf Tf z (78) (79) 4.73–77] for the space-like and [79. we used the following identities to simplify the expressions:   1 1 (82) = −Li2 (−z) − ln2 (z) − ζ (2). We follow the results given in [71.i . (81) ⊗ CL.2 (−z) + Li3 (−z) − ln(z)Li2 (−z) − ln3 (z) + ζ (3).2 (1 − z).g (z) + zCL. z CL. It turns out that the coefficient functions are related by the Z-factors through the matrix-valued convolutions    z (S)(2) 1 (T )(2) CL. S1. Li2 − z 2   1 1 = − ln2 (z) − Li2 (1 − z). (83) Li2 1 − z 2   1 1 (84) = − ln3 (z) + S1. Although . S1. Blümlein et al.i and C1. (80).g 2 z 2Nf Tf 2 z     CF 1 (T )(2) (S)(2) 1 CL. Longitudinal coefficient functions We consider the NNLO correction to the longitudinal coefficient function. (81) are no longer true. Li3 − z 6   1 1 (87) = −S1.i is obtained as the difference between C2.q (z) + − CL. Since the NLO coefficient functions involve various Nielsen integrals.q = 0.

g z   2 2 Q 1 (0) (0) ⊗ P¯gq + 12CF2 ζ (2) 2 ln 2 − 3 δ(1 − z). In Section 4.2. (80).g 2Nf Tf z "   CF 1 (S)(1) 1 (T )(1) (T )(1) (0) + 2β0 Zqg 2zPqg + Zqg ⊗ P¯qq = 4 2Nf Tf z #  1 (T )(1) (0) (T )(1) (0) (T )(1) (0) − Zqq ⊗ P¯qg + Zgg ⊗ P¯qg − Zqg ⊗ P¯gg ln(z) + 2    1 (T )(1) (S)(1) 1 (T )(1) (T )(1) ⊗ Zqq + Zqg ⊗ −zC1.and time-like coefficient functions using the transformation relations (64).1):    (T )(2) (S)(2) 1 (z) + z C1.4. the coefficient functions CLq(G) (z) may be combined to physical evolution kernels together with the NLO splitting functions as shown in Section 3.q + Zqq 2 z    C 1 1 (T )(1) F (S)(1) (T )(1) (T )(1) ⊗ Zqg + Zgq ⊗ zC + Zgq 2 2Nf Tf 1.q C1.q z " #   1 (1)S 1 (T )(1) (T )(1) (0) (T )(1) (0) ¯ ¯ 2(−z)Pqq + 2β0 Zqq + Zgq ⊗ Pqg − Zqg ⊗ Pgq ln(z) = 4 z    1 (T )(1) (S)(1) 1 (T )(1) (T )(1) ⊗ Zqq + Zqq ⊗ −zC1.q + Zqg 2 z .J. we find that the form of Eqs. (88) is the same but the term 1 ¯ (0) (0) P ⊗ P¯qg 8 gq does not occur. Similarly for the gluonic coefficient functions we find "   # CF 1 (T )(2) (S)(2) 1 C (z) − 2z C1. (88) + P¯qg 8 µf For the polarized NNLO coefficient functions which were derived in [73–77] and [79. The transverse coefficient functions are related by (see Appendix A. Transverse coefficient functions In NNLO physical evolution kernels for the transverse structure and fragmentation function can only be constructed when the space-like and time-like three-loop splitting functions are known. (66) and (77) for unpolarized and polarized scattering.T formally of NNLO. (47)–(51) up to second order. (65).5 we will show that because of the transformation in Eqs.g 2 1.80].1. (81) the physical evolution kernels in Sections 3.1 and 3. If they become available one can extend Eqs. 4.80]. Blümlein et al. The space-like coefficient functions for unpolarized scattering are computed in [73–77] whereas the time-like ones can be found in [79. Here we consider the relation between the space.2 remain DLY-invariant. / Nuclear Physics B 586 (2000) 349–381 367 (2)S.

(89) are absent.g + 2β0 δC2.g N(1) N(1) C¯ L.q − N(1) δC2. (72).g C¯ L.q δCL. Application of the DLY-transformations provides us with the following results N(1) N(1) = δγqq − δK22 " − N(1) C¯ L.q N(1) C¯ L. (67). δK22 For the remaining evolution kernels one obtains (91) using Eqs. / Nuclear Physics B 586 (2000) 349–381    CF 1 (T )(1) (S)(1) 1 (T )(1) (T )(1) + Zqg ⊗ Zgg + Zgg ⊗ zC1.g N(1) N(0) + γ¯gq δC2.5. (74).q C¯ L.g N(1) C¯ L. N(1) Substituting the expressions for δγqq N(1) and δγgq N(1) = 0.g N(2) δCL. Since we do not have to average over the initial state gluon polarization in the case of polarized scattering the term ln(z) + 1/2 multiplying the first bracket in Eq. NLO physical evolution kernels After having found the relations between space-like and time-like splitting and coefficient functions. the terms  1 (0) 1 (0) (0) (0) + P¯qg ⊗ P¯gg − P¯qq − β0 P¯qg 8 16 in Eq. 4. (89) is replaced by ln(z).g N(1) N(0) δC2.q  (T )N(1) N(0) (T )N(1) N(0) − Zgg γ¯qg + Zqg γ¯gg . we define the difference between the time-like quantities KITJ and the continued space-like quantities K¯ ijS by δKI J = KITJ − K¯ ISJ .368 J. also Eqs.2.q  N(1) (T )N(1) (T )N(1) N(0) (T )N(1) N(0) + N(1) −δγgq + 2β0 Zqg − Zqg γ¯qq + Zqq γ¯qg C¯ L.g # N(0) γ¯qg ! N(1) C¯ L.g C¯ L. In order to do this.g C¯ L. we investigate the DLY-transformation for the physical evolution kernels presented in Sections 3. (64). − β0 Pqg + Pqg ⊗ P¯gg 8 16 (89) For the polarized case. (66).g γ¯qq N(1) δC2. Blümlein et al.g 2 2Nf Tf z  1 ¯ (0) 1 ¯ (0) (0) (0) − P¯qq .q N(1) C¯ L.g N(1) C¯ L. (77). cf.g N(1) (T )N(1) N(0) (T )N(1) N(0) (T )N(1) − 2β0 Zqq − γ¯gq Zqg + γ¯qg Zgq = δγqq N(1) C¯ L. (69).g − N(1) N(2) C¯ L. we get (92) .q N(1) C¯ L.g N(1) δγgq + N(1) +  ¯ N(1) 2 CL.q N(1) N(1) N(1) N(0) − N(1) γ¯gg δC2. (90) where K¯ ISJ is obtained by transforming KISJ to the time-like region using the continuation rules (65).1 and 3.

g C¯ L. ∂F2 /∂t as a basis. (67)–(70).J for the evolution of the structure functions F2 and FL are thus DLY-invariant to next-to-leading order if continued from the space-like to the time-like region. (98) The physical evolution kernels KI.g C¯ L. leads to δKL2 N(1) = 0.g i i i (T )N(1) (T )N(1) N(0) (T )N(1) N(0) − Zqg γ¯gq + Zgq γ¯qg .q N(1) C¯ L.g (94) (T )N(1) (T )N(1) N(0) (T )N(1) N(0) N(0) − 2β0 Zgq − Zgq γ¯qq + Zqq γ¯gq γ¯gq (T )N(1) (T )N(1) N(0) − Zgg + Zgq γ¯gg + + N(1) h C¯ L. which are necessary in the explicit calculation. (96) N(1) δK2L N(1) δKLL = 0. − γ¯qg (93) N(1) N(1) N(0) (T )N(1) N(0) (T )N(1) N(0) = δγgg − 2β0 γ¯gg + Zqg γ¯gq − Zgq γ¯qg δKLL + N(1) C¯ L.q N(1) C¯ L.g (T )N(1) (T )N(1) N(0) (T )N(1) N(0) − Zqg γ¯gq + Zgq γ¯qg −2β0 Zqq N(1) C¯ L. which change under the DLY-transformation as follows: δKd2 =   β0  N(1) (T )N(1) N(0) N(0) δC2q − Zqq + γ¯gg − 2β0 γ¯qq 2 .q N(1) N(1) N(1) N(1) N(1) δγgq − N(1) δγgg δKL2 = δγqg + N(1) δγqq − N(1) C¯ L.g N(1) (T )N(1) (T )N(1) N(0) (T )N(1) N(0) − 2β0 Zqg + Zqg γ¯qq − Zqg γ¯gg δγgq i (T )N(1) N(0) (T )N(1) N(0) − Zqq γ¯qg + Zgg γ¯qg .1 as well as a series of involved Mellin convolutions leading to Nielsen integrals (see Appendix A. Using Eqs. Blümlein et al. Here only two evolution kernels are contributing.q N(1) C¯ L.2). We turn now to the physical evolution kernels in next-to-leading order where we choose the physical quantities F2 . (97) = 0.g !2 h (T )N(1) (T )N(1) N(0) (T )N(1) N(0) − Zqg γ¯qq − Zgg γ¯qg 2β0 Zqg (T )N(1) N(0) (T )N(1) N(0) + Zqq γ¯qg + Zqg γ¯gg + N(1) h C¯ L.q L.q L.J.q h N(1) C¯ L. ! N(1) ¯ N(1) ¯ N(1) 2 C C C¯ L. 2β0 Zgg (95) The explicit expressions for the differences in the coefficient function are given in Appendix A. / Nuclear Physics B 586 (2000) 349–381 N(1) δK2L 369   N(1) (T )N(1) (T )N(1) N(0) N(0) = δγgq − 2β0 Zqg + Zqg − γ¯gg γ¯qq   N(0) (T )N(1) (T )N(1) − Zgg Zqq .

/ Nuclear Physics B 586 (2000) 349–381 − δKdd    N(1) (T )N(1) N(0) 2 N(0) N(0) − Z − γ¯qq γ¯gg δC γ¯qq qg 2g N(0) β0 2γ¯gq  N(0) N(0) N(0) + 2γ¯qg γ¯gq − 2β0 γ¯qq . without changing the δ-function. In the future one can extend the investigation performed in this section to physical evolution kernels at the NNLO-level. We finally would like to comment on a relation derived by Gribov and Lipatov in [36] for the leading order kernels for a pseudoscalar and a vector field theory. Since in both reactions non-perturbative quantities such as the structure and fragmentation functions contribute the above question cannot be answered by means of perturbation theory for 6 See also [81. + 4β0 δC1q (99) (100) From Eqs. provided the three-loop anomalous dimensions are calculated. (75). For the physical nonsinglet evolution kernels this was shown in [64. FL ) one also needs the three-loop coefficient functions. (42)–(45) . Q2 ). (76). (101) δKdd = 0. (65).and space-like evolution kernels. The Z T -factors which are needed for the transformation of the splitting and coefficient functions cancel in the expression above. Starting with next-to-leading order. 5.    β0 N(1) (T )N(1) N(0) N(0) = −2 N(0) δC2g − Zqg − γ¯gg − 2β0 γ¯qq γ¯gq   N(1) (T )N(1) − Zqq . this relation is violated as well. that also for the physical singlet combinations. Eqs. One verifies. Eqs.67] and for some singlet combinations in [67]. that this relation holds in leading order for the space. (64). Eq. respectively.to s-channel was newly discussed. (64). (102) From these results it is clear that the time-like physical evolution kernels KijT can be directly derived from the space-like physical evolution kernels using the continuations in Eqs. (56)–(62). and xB = 1/xE . (77) where one has to account for the corresponding changes in the overall color factors.and time-like splitting functions of QCD. Q2 ) = K(xB . (79) we can derive that δKd2 = 0. whether the scattering cross sections of deep inelastic scattering e− + P → e− + ‘X0 are related to the annihilation cross section e+ + e− → P¯ + ‘X0 by a crossing relation changing from t. (103) where K and K denote the time.370 J. (78). this relation is not preserved. 6 One may write it in the form K(xE . (66). (48)–(51). Blümlein et al. . We find.65. For the choice of observables (F2 . Conclusions The old question.82] for related work.

For that purpose transformation relations have been derived for the splitting functions up to NLO and the coefficient functions up to NNLO.and time-like) as well the three-loop longitudinal coefficient functions in the first example above. Coefficient functions In this appendix we list the difference of the space. Two distinct renormalization group equations are implied. which are used in Section 4 to study the validity of the DLY-relation. However one can construct factorization-scale independent evolution kernels which describe the scheme-invariant evolution of these physical quantities in terms of a kinematic variable given by Q2 . which is valid in leading order. respectively. The first example of the application of the physical evolution kernels is the coupled structure functions F2 (x. Contrary to the splitting functions (anomalous dimensions) and coefficient functions the evolution kernels of the examples above are factorization scheme independent. Blümlein et al. Here the expressions also contain the logarithms . A second example is given by F2 (x. Q2 ) associated with the corresponding fragmentation functions in e+ e− -annihilation. Kurth in an early phase of this work are acknowledged. the collinear singularity and the ultraviolet singularity. Appendix A A. This work was supported in part by EU contract FMRX-CT98-0194 (DG 12-MIHT). In the calculation of both inclusive processes only two types of singularities occur. Q2 ) and ∂F2 (x. They quantify the impact of the factorization and the renormalization scale on the DIS structure functions and fragmentation functions when the perturbation series is truncated up to a given order.1. which presumes the knowledge of the yet unknown three-loop splitting functions (space. It remains to be seen how the physical evolution kernels behave under the DLY crossing relation at NNLO. which are computable within perturbation theory. This scheme invariant evolution is guaranteed up to any finite order in perturbation theory. / Nuclear Physics B 586 (2000) 349–381 371 the process as a whole. Q2 )/∂ ln(Q2 ). Acknowledgement We would like to thank P. These divergences are absorbed into the bare parton densities and the coupling constant. is already violated at next-to-leading order. Discussions with S. Q2 ) and FL (x.and time-like coefficient functions in the MS scheme. We have also shown that these kernels are invariant under the Drell–Levy–Yan-transformation up to next-to-leading order. However. since both the parton densities involved in the space. Menotti for providing us a reprint of [22]. Notice that in finite order this method does not remove the dependence of the physical quantities on the renormalization scheme of the strong coupling constant or its scale µr .J. On the other hand the Gribov–Lipatov relation.and time-like process factorize if the virtuality Q2 = |q 2 | of the four-momentum transfer is large a related question can be asked for the crossing behavior of the respective evolution kernels.

4).4) + 32 1 − z 9z 9 Notice that for the computation of the coefficients functions above and the ones following hereafter one also needs the virtual contributions to the zeroth and first order partonic processes. Li2 (1 − z) − 8 + (A. (A. are given by    (2)NS (A.3) − 9z 9 The same is done for the longitudinal gluonic coefficient functions corresponding to the ¯ respectively.g = CF2 8 1 + − z ln(z) + 8 ln2 (z) − 28 + z z      1 2 2 1 ln2 (z) + CA CF 16 4 − + z − z ln(z) − 16 1 + z 3 z    40 248 1 − 24z + z2 . respectively. The difference between the longitudinal purely singlet coefficient functions ¯ respectively. The difference between the longitudinal non-singlet coefficient functions corresponding to the processes γ ∗ + q → q + g + g and γ ∗ → ‘q¯ 0 + q + g + g.2). where ‘q 0 denotes the quark in the final state which undergoes fragmentation into a hadron P (see Eq. (A. In the non-singlet case we have (2)NS δC1.372 J.2) δCL.q    1 ln(z) 2 = CF − CA CF 8 −2ζ (2) − 4 ln(z) ln(1 + z) + ln2 (z) 2 1+z + 4 2(1 − z)ζ(2) + 4(1 − z) + 4(1 − z) ln(z) ln(1 + z)    2 2 −1+z + 2(1 + z) ln(z) − (1 − z) ln (z) ln(z) − 16 Li2 (−z) ln(z) 1+z     10 8 − 1 + 11z ln(z) − + Nf Tf CF 9 1−z    ln(z) 67 2 + ln (z) − 2ζ (2) + CA CF 4 1−z 9    53 187 2 − z + 2(1 + z) ln(z) − (1 + z) ln (z) ln(z) + 4ζ (2)(1 + z) ln(z) +2 9 9 .q = Nf Tf CF −8 6 + 4z − z ln(z) + 16 ln2 (z) − 16 3  128 2 304 + 64z − z . The difference in processes γ ∗ + g → g + q + q¯ and γ ∗ → ‘g 0 + g + q + q.q = CF2 4 2z − ln(z) ln(z) − 16 Li2 (1 − z) + 8 − 8z .3). Blümlein et al. the coefficient function is given by     24 2 (2) + 4z δCL. are given by    4 2 (2)PS δCL. / Nuclear Physics B 586 (2000) 349–381 Lµf = ln Q2 /µ2f  (A. (A. The differences between the transverse coefficient functions emerge from the same processes as mentioned above Eqs. (8)).1) which arise when the factorization scale µ2f is chosen to be different from Q2 . corresponding to the processes γ ∗ + q → q + q + q¯ and γ ∗ → ‘q¯ 0 + q + q + q. (A.

6) . Blümlein et al. / Nuclear Physics B 586 (2000) 349–381 " + CF2 4 ln(z) 1−z 373  8Lµf − 6 + 4 ln(1 − z) ln(1 − z) + 6Lµf − 18 !   16 ln(z) ln(z) + −4Lµf + 6 − 3  + 4 −4(1 + z)Lµf + 1 + 5z − 2(1 + z) ln(1 − z) ln(z) ln(1 − z) + 2 −2(5 + z)Lµf + 14 + 40z + 2(1 + z) ln(z) ln(1 − z)  + 6(1 + z)Lµf ln(z) − 8(2 + z) ln(z) + 7(1 + z) ln2 (z) ln(z) !   12 + 7 + 7z ln(z) − 10 + 22z + 2Li2 (1 − z) 4 − 1−z     2 24 + 13 + 13z S1.q "    8 −8 4 + 6z + z2 Lµf + ln(1 − z) 3   160 368 2 2 + 2z ln(z) − 160 − − 112z − z − 16 1 − 3z 9z 9 !   10 ln(z) ln(z) ln(z) + 4(1 + z) 4 ln(1 − z) + 4Lµf + 3    38 2 38 −z− z Lµf + ln(1 − z) −8 1+ 9z 9   4 2 + 16 2(1 + z) ln(z) − 2 − 3z − z Li2 (1 − z) 3 = Nf Tf CF # 1808 2 1168 224 640 − + z+ z .2 (1 − z) + 36(1 − z) + 32ζ (2) − 1 − z ln(z) +8 − 1−z 1−z #  (A.5) + 12ζ (2) 9 − 12Lµf + 4L2µf δ(1 − z) .J.2 (1 − z) − 9 27z 9 27 The difference between the gluonic coefficient functions equals (2) δC1. + 32(1 + z)S1. For the purely singlet difference we obtain (2)PS δC1.g "   184 2 32 24 704 + 66z + z + 48 − + 16z + z2 Lµf 9z 9 z 3     16 32 100 + 10z ln(z) + 40 − + 12z + z2 ln(1 − z) + −4 − 3z z 3     1 2 + 4 2 + + z ln(1 + z) − 16 1 + + z ln(z)Lµf z z = 2CA CF 188 + (A.

/ Nuclear Physics B 586 (2000) 349–381     1 2 + 8 2 + + z ln(z) ln(1 + z) − 16 1 + + z ln(z) ln(1 − z) z z !     40 64 52 2 2 2 + + z ln (z) + 4 2 − − z ln (1 − z) ln(z) − 3 3z 3 z   104 2 356 + 4z − z ln(1 − z) + −32 + 9z 9     104 2 284 2 + 4z − z Lµf + 2 2 − − z ln2 (1 − z) + −24 + z 9z 9   3 + 32 1 − − 2z ln(z) z !    32 2 8 2 + 16 2 − − z ln(1 − z) + Lµf + 16 + + 16z + z Li2 (1 − z) z z 3      2 2 + 4 2 + + z 1 + 2 ln(z) Li2 (−z) + 16 −2 + + z Li3 (1 − z) z z   856 2 418 4168 668 16 − 10z S1.2 (1 − z) − 169 + z z z (A.2 (1 − z) − + − z− z +8 8− z 9 27z 9 27 #    4 + 8ζ (2) −3 + + 2z 1 + 2 ln(z) z "   3 8 2 + 2CF 42 + − 29z − (14 − 11z) ln(z) + 16 3 − − z ln(1 − z) z z   4 + 4(2 − z) ln(z)Lµf − 4 2 − − z ln(z) ln(1 − z) z   10 2 − 16 2 − − z ln(1 − z)Lµf + (2 − z) ln2 (z) z 3 !   2 − 12 2 − − z ln2 (1 − z) ln(z) z !     2 2 52 − 18z − 8 2 − − z Lµf − 6 2 − − z ln(1 − z) ln(1 − z) + 64 − z z z   10 − 3z Lµf + 2 16 − z !    2 6 + 8 (2 − z) ln(z) − 2 2 − − z ln(1 − z) + Lµf + 6 − − 3z Li2 (1 − z) z z #     8 106 2 + 50z .7) . + 16 2 − − z Li3 (1 − z) + 8 10 − − 5z S1. Blümlein et al.374 J.

Blümlein et al.2 (1 − z) + 36(1 − z) + 32ζ (2) − 1 − z ln(z) +8 − 1−z 1−z #  (A. (A.J.8) + 12ζ (2) 9 − 12Lµf + 4L2µf δ(1 − z) . The analogues of Eqs.q = Nf Tf CF "  16(1 − 4z) Lµf + ln(1 − z) − 184 + 24z − 16(2 + 3z) !   10 ln(z) ln(z) ln(z) × ln(z) + 4(1 + z) 4Lµf + 4 ln(1 − z) + 3  − 48(1 − z) Lµf + ln(1 − z) . (A.6). (2)PS δ1C1.5).7) are given by (2)NS δ1C1. Q2 ) which describes polarized scattering.q  "  1 ln(z) 2 −2ζ (2) − 4 ln(z) ln(1 + z) + ln2 (z) = CF − CA CF 8 2 1+z + 4 2(1 − z)ζ(2) + 4(1 − z) + 4(1 − z) ln(z) ln(1 + z) # 2 −1+z + 2(1 + z) ln(z) − (1 − z) ln (z) ln(z) − 16Li2 (−z) ln(z) 1+z # "   10 8 − − 1 + 11z ln(z) + Nf Tf CF 9 1−z "   ln(z) 67 + ln2 (z) − 2ζ (2) + CA CF 4 1−z 9 #   53 187 − z + 2(1 + z) ln(z) − (1 + z) ln2 (z) ln(z) + 4ζ (2)(1 + z) ln(z) +2 9 9 "   ln(z) 2 8Lµf − 6 + 4 ln(1 − z) ln(1 − z) + 6Lµf − 18 + CF 4 1−z    16 lg(z) ln(z) + −4Lµf + 6 − 3  + 4 −4(1 + z)Lµf + 1 + 5z − 2(1 + z) ln(1 − z) ln(z) ln(1 − z) 2   + 2 −2(5 + z)Lµf + 18 + 36z + 2(1 + z) ln(z) ln(1 − z)  + 6(1 + z)Lµf ln(z) − 2(7 + 5z) ln(z) + 7(1 + z) ln2 (z) ln(z)     12 + 7 + 7z ln(z) − 2 + 14z + 2Li2 (1 − z) 4 − 1−z     24 2 + 13 + 13z S1. (A. / Nuclear Physics B 586 (2000) 349–381 375 We have computed the same differences between the coefficient functions corresponding to the structure function g1 (x.

/ Nuclear Physics B 586 (2000) 349–381 #  + 16 2(1 + z) ln(z) + 1 − 4z Li2 (1 − z) + 32(1 + z)S1. Convolutions Here we list the convolutions of a series of functions.376 J.11)  ln(z) 1 ln(z) ⊗ = − 2 S1.g = 2CA CF  212 + 48z − 32(1 − 2z)Lµf + 4(8 + 5z) ln(z) − 16(1 − 3z) ln(1 − z) − 8(4 + z) ln(z)Lµf + 8(2 + z) ln(z) ln(1 + z) − 8(4 + z) ln(z) ln(1 − z)  4 − (26 + 5z) ln2 (z) − 4(2 − z) ln2 (1 − z) ln(z) 3  + 32(1 − z) ln(1 − z) + Lµf   + 16 2 + z − (2 − z) Lµf + ln(1 − z) − (8 − z) ln(z) Li2 (1 − z) + 8(2 + z) ln(z)Li2 (−z) + 16(2 − z)Li3 (1 − z) − 32(5 − z)S1.2 (1 − z) + 96 − 96z .10) A.9) (2) δ1C1. (A.2(1 − z) + ln(z)Li2 (1 − z) . which are needed for the investigation of the DLY-relation in Section 4. (A.13) . (A. (1 − z) z z (A.12) 1 ln(z) ⊗ z2 ln(z) = − 3 − 24z + 27z2 − 24S1.2. (13) we obtain   ln(z) 1 1 3 ln(z) ⊗ = − 4S1.2(1 − z)z2 (1 − z) 12  − 3(1 + 4z + 5z2 ) ln(z) − 2z2 ln3 (z) − 12z2 ln(z)Li2 (1 − z) .2(1 − z) − 2 ln(z)Li2 (1 − z) − ln (z) . Blümlein et al.2 (1 − z) − 112 + 112z . (1 − z) (1 − z) (1 − z) 6 (A.2 (1 − z)  + 224 − 224z + 8ζ (2)(8 − 3z) ln(z)  2 + 2CF 22 + 40z + 4(2 − z)Lµf − 4(14 − 9z) ln(1 − z) + 4(8 − 5z) ln(z) + 16(2 − z) ln(1 − z)Lµf − 4(2 − z) ln(z)Lµf + 4(2 − z) ln(z) ln(1 − z)  10 − (2 − z) ln2 (z) + 12(2 − z) ln2 (1 − z) ln(z) 3  − 8(1 − z) ln(1 − z) + Lµf   + 4 − 10 + 5z + 4(2 − z) ln(1 − z) + Lµf − 2(2 − z) ln(z)  × Li2 (1 − z) − 16(2 − z)Li3 (1 − z) − 40(2 − z)S1. Using the definition in Eq.

2 (1 − z) − ln(1 − z)Li2 (1 − z) + Li3 (1 − z) . 2   1 1 1 ln(z) ⊗ ln(z) ln(1 − z) − ln2 (z) .2(1 − z) − ln(1 − z) ln2 (z) − 2 ln(z)Li2 (1 − z) .(A.2 (1 − z) = (1 − z) 1−z (1 − z) 2 +  1 2 (A.2 (1 − z) + ln(1 − z) − z ln(1 − z) (1 − z) z − ln(z) + ln(1 − z) ln(z) − ln(1 − z) ln2 (z) 2   − − 1 + z ln(1 − z) + z ln(z) Li2 (1 − z) + zLi3 (1 − z).21) (1 − z) (1 − z)+ (1 − z) 2  1 ln(z) ln(1 − z) ⊗ = − 2 S1.2(1 − z) − ln(1 − z) ln2 (z) (1 − z) 2   − ln(1 − z) + ln(z) Li2 (1 − z) + Li3 (1 − z). 6 1 ln(z) ⊗ ln(z) = −2S1. Blümlein et al. (A. / Nuclear Physics B 586 (2000) 349–381 377 ln(z) ⊗ z ln(z) = −2 + 2z − 2z S1. = (A.16) (1 − z) z z 1n ln(z) ⊗ z2 ln(1 − z) = −4 − 5z + 9z2 − 8 S1. (A.15) (1 − z) 6  ln(1 − z) 1  ln(z) ⊗ = −S1.22)    ln(1 − z) 1 1 2 ln(z) ⊗ ln (1 − z) ln(z) = z 1−z z 2 +  (A.2 (1 − z) − ln(z) − z ln(z) (1 − z) z (A. (A.23) + ln(1 − z)Li2 (1 − z) − Li3 (1 − z) .19)    ln(1 − z) 1 1 ln(z) ⊗ ln(z) ln2 (1 − z) − 2S1.20) − ln(z)Li2 (1 − z) − ln (z) ln(1 − z) .17) ln(z) ⊗ z ln(1 − z) = −2 + 2z − 2z S1.2(1 − z) − ln3 (z) − ln(z)Li2 (1 − z). .2 (1 − z)z2 + ln(1 − z) (1 − z) 4 + 4z ln(1 − z) − 5z2 ln(1 − z) − 3 ln(z) − 4z ln(z) + 2 ln(1 − z) ln(z) + 4z ln(1 − z) ln(z)  − 2z2 ln(1 − z) ln2 (z) + 2 + 4z − 4z2 ln(1 − z) o  − 4z2 ln(z) Li2 (1 − z) + 4z2 Li3 (1 − z) .18) 1 ln(z) ⊗ ln(1 − z) = −2 S1.J.14) − ln3 (z) − z ln(z)Li2 (1 − z). z z 2z (A. (A.

2 (1 − z) + 4 4z 4 (A. z (1 − z)+ z   ln(1 − z) 2 z ln(z) ⊗ 1−z +  5 3 3 + ln(1 − z) = z2 − − S1. z2 ln(z) ⊗ z2 ln(1 − z) = − z2 ln(z) ⊗ (A.28) + z ln(z) ⊗ z 1 = −z ln(z) − 1 + z − ln2 (z) + zLi2 (1 − z) + z ln(z) ln(1 − z).30)  + = −S1. / Nuclear Physics B 586 (2000) 349–381  1 1 ln(z) ⊗ = −Li2 (z) + ζ (2) . (A.2 (1 − z) + ln(1 − z) − ln(1 − z) − ln(z) z ln(z) ⊗ 1−z z + − 1 2 1 ln (1 − z) ln(z) − ln(z) ln(1 − z) + ln2 (z) 2 2 1 − ln(1 − z) ln2 (z) − Li2 (1 − z) + ln(1 − z) 2  × Li2 (1 − z) − ln(z)Li2 (1 − z) − Li3 (1 − z) . 2 (A.2(1 − z) + ln(1 − z) ln2 (z) + 2 ln(z)Li2 (1 − z) .26) 5 1 1 3 = − − z + z2 − z2 ln(z) (1 − z)+ 4 4 2 z2 2 ln (z) + z2 ln(z) ln(1 − z) + z2 Li2 (1 − z).29) z ln(z) ⊗ z ln(1 − z) = −z S1.25)  z2 2 S1.2 (1 − z) −  ln(1 − z) ln(z) ⊗ 1−z z 2 ln (z) ln(1 − z) − z ln(z)Li2 (1 − z). 2 (A. (1 − z)+ 2 (A. Blümlein et al. (A.24) 3 1 1 ln(1 − z) − ln(1 − z) − ln(z) 4z2 z 4 1 3 − ln(1 − z) ln(z) + ln2 (1 − z) ln(z) 2 2 1 3 3 + ln2 (z) − ln(1 − z) ln2 (z) − Li2 (1 − z) 4 2 2 −  + ln(1 − z)Li2 (1 − z) − ln(z)Li2 (1 − z) − Li3 (1 − z) .27) 2    1 ln(1 − z) = z − S1.2 (1 − z) + 1 2 ln (1 − z) ln(z) 2 .378 J.

Phys. Schwinger. z (1 − z)+ z   (A. (1 − z)+ 2 ln(z) 1 ln(1 − z) 1 − − δ(1 − z)ζ (2).2 (1 − z) − ln(z) ⊗ 1 2 ln (z) ln(1 − z) − ln(z)Li2 (1 − z). Phys. Rev. Yan.D. 2 1 1 = − ln2 (z) + Li2 (1 − z) + ln(z) ln(1 − z). Phys. Adv. 2 1 = 1 − z + z ln(1 − z) − z ln(z). Nat. D 1 (1970) 1617. Sci. 37 (1951) 455. Sci. Chen. (A. T. A. (1 − z)+ (A. Molecul.39) (A. / Nuclear Physics B 586 (2000) 349–381 1 ln(1 − z) ln2 (z) + ln(1 − z)Li2 (1 − z) 2 − ln(z)Li2 (1 − z) − Li3 (1 − z). Phys. D. Pestieau. B 30 (1969) 483. Weinberg.41) (A.35) (A. J. Lett.A. Bethe. S.M. .37) 2 3 1 1 = + z − z2 − z2 ln(z) + z2 ln(1 − z). Levy. Rev. Drell. 379 − ln(z) ⊗ ln(1 − z) = −S1. 1⊗ 1−z 2 + z⊗ 1⊗ 1 = ln(1 − z) − ln(z).38) z2 ⊗ (1 − z)+ 2 2 z2 ⊗  ln(1 − z) 1−z ln(1 − z) z⊗ 1−z =  + = (1 − z) ln(1 − z) + z ln(z) − z ln(z) ln(1 − z) 1 − zLi2 (1 − z) + z ln2 (1 − z). Phys. Blümlein et al. H.D. 37 (1951) 452. 133 (1964) B232. P. Roy. Nat. J.33) (A.E. JETP 12 (1961) 1014.42) References [1] [2] [3] [4] [5] [6] [7] S. J. Chen.40) (A. Sov. Phys.C. Rev. Salpeter.C.36)   3 3 1 1 + z − z2 ln(1 − z) − z(1 − z) + z2 ln2 (z) 2 2 2 2 + 1 2 2 + z ln (1 − z) − z2 ln(z) ln(1 − z) − z2 Li2 (1 − z). Phys. = z 1−z 2z + 1 1 1 ⊗ = ln(1 − z).34) (A. ⊗ =2 (1 − z)+ (1 − z)+ (1 − z) (1 − z)   ln(1 − z) 1 2 1 ⊗ ln (1 − z). 84 (1951) 1232. Rev. Proc.J. E.J. Faddeev. 8 (1972) 71. (1 − z)+   1 ln(1 − z) = −Li2 (1 − z) − ln(z) ln(1 − z) + ln2 (1 − z).31) (A. 187 (1969) 2159. L. Atom. Acad. Acad. Proc.32) (A. (A.Y.

B 39 (1971) 49. Polkinghorne. Mod. Martin. [53] P. R. D 1 (1970) 1035. [11] S. Kinoshita. I. Nucl. Schäfer. 52 (1980) 199. Phys. 52 (1977) 11. Blümlein. D 6 (1972) 3708.D. 24 (1970) 847. Phys. Nuovo Cimento A 2 (1971) 881. Shilov. Ann. West. Phys. Phys. Menotti. Frautschi. N. Parisi. Derman. W. T.N. Steiner. Phys.M. Lett. [34] J.D. [25] G. Walecka. Yan. Suura. and references therein.D. Rev. Academic Press.J. Fishbane. [43] C.C.B. Drell. Ravindran. L. Lee. B 450 (1999) 126. 20 (1975) 181. Belitsky. [31] D. Rev. T. Phys. B 32 (1970) 57.. F. Rev. Generalized Functions. [27] F. Phys. Math. [32] H. M. B 266 (1991) 173. B 29 (1998) 2581. Rev. Drell. Nucl. [16] H. Rev. Phys. Lett. Lett. D 5 (1972) 1738. B (Proc. A. Polkinghorne. Dahmen. [46] A.J. Lett. Comm. L. Tkabladze.M. Math. W. Nucl. [29] I. Wilczek. Phys. Comm. A.N. Parisi. [52] G.D. A. G. Nucl. Phys. Rev. Milsztain. Phys.M. Phys. Phys. 52 (1937) 54. [35] A. Phys. Phys.G. . Stirling. Rev. [9] S. Nuovo Cimento A 7 (1972) 118. Drell. [30] L. Phys. Gribov. Phys. Levy. Phys. [22] P. Lett. Phys. Nucl. B 28 (1971) 225. T.D. Phys. [28] D. [13] P. A. [10] S.J. Vogt. 1972. [20] P. Vol.V. A. R. [24] S. Gilman. Bloom. [38] E. Vogt. D 2 (1970) 1084.D. J. Gel’fand. Phys. Phys. Phys. B 126 (1977) 298. D. Nuovo Cimento 4 (1972) 79. D. [23] P. Ann. Rev. P. 133 (1964) B 1549. D 9 (1974) 980. Phys.J. Phys. Blümlein. Lett. (New York) 28 (1964) 18. Lipatov. Nordsieck.L. Lee. 24 (1970) 181. Sullivan. van Neerven. Gribov. Rev. / Nuclear Physics B 586 (2000) 349–381 [8] S. J. Phys. B 41 (1972) 480.V. Comm. 24 (1970) 1206. Phys. P. [48] J. (New York) 13 (1961) 379. B 44 (1972) 189. Riemersma. Maiani. Maison. Müller. Politzer. Yan. B 37 (1971) 78. Lett. Lett. [17] R. D 2 (1970) 1173. T. Rev. G.M. [49] J. [21] G. Sov.N. 1. 15 (1972) 675.M. Phys. Lett. Nucl. D. L. B 568 (2000) 263. 1964.C. J. Roberts.J. Yan. Virchaux. B 47 (1972) 313.G. Phys. Levy. Lett. [50] W. Phys. Veltman.) 51C (1996) 97. Altarelli. Yennie. Callan Jr. Phys. J. Blümlein. Gatto. W. Math. B 274 (1992) 221. [12] V. D 19 (1979) 133.D.R. Gatto.L. Phys. [33] G. S. B 37 (1971) 68. Georgi.D. [45] G. B 498 (1997) 285. Altarelli. Rev. 25 (1970) 1140. 22 (1969) 744. Rev. [36] V. DESY 99-072. Menotti. Suri. Preparata. F. S. Drell.L. Phys. (New York) 79 (1973) 1. V.E. Landshoff. Blümlein. H. [44] K. J. A. Blümlein. 18 (1970) 227. Chang. van Neerven. [14] A. G. Breitenlohner. Phys. Kochelev. Nauenberg.D. Frascati.380 J. Phys. [51] A. M. Nucl. Phys. B 53 (1973) 473. Lett. Gross. D 2 (1970) 1541. J. Rev. Symanzik. Rev. Phys. B 39 (1972) 643.D. May 2–3. P. Menotti. Phys. Suppl. Lett. Rev. Pisa Preprint SNS 3/72.J. B 51 (1973) 509. [42] T. [26] E.V. D. D 4 (1971) 510. Rev. Landshoff. Buras. in: Proceedings of the Informal Meeting on Electromagnetic Interactions. Lett. Sov. Lett. Polon. New York. Phys. Blümlein et al. Bloch. Rev. B 43 (1973) 217. Phys.C. Rev. Phys. D. Nucl. [39] J.V. Rev. D 8 (1973) 3633. Nucl. Acta Phys. [47] M. Lipatov. Math.D.N. t’Hooft.M. Gatto. [37] S. Ann.D. Phys. [19] P. Fishbane. Short. Lett. [40] J. [18] R. Phys. Drell. Phys.N. A. Lipatov. Nucl. [15] R. B 553 (1999) 427. Nucl. van Neerven. Nucl. Phys. J. Phys. [41] T. Vogt. D 9 (1974) 416. F. 3 (1962) 650. Vendramin. Phys. Landshoff.D.

Phys. P. S. van Neerven. Ph. B 475 (1996) 47. 1993. W. D 45 (1992) 3986. W. N. Nucl. M. Kounnas. van Neerven. Vermaseren. Phys. R. B 98 (1981) 285. V. Phys. Phys. Nucl. van Neerven. W. Nucl.J.J. Rev. Nucl. / Nuclear Physics B 586 (2000) 349–381 [54] [55] [56] [57] [58] [59] [60] [61] [62] [63] [64] [65] [66] [67] [68] [69] [70] [71] [72] [73] [74] [75] [76] [77] [78] [79] [80] [81] [82] 381 W.L. Ferrara. Phys. B 417 (1994) 61. Lipatov. Lett. Phys. R. B 273 (1991) 476. J. Vogelsang.P. Petronzio. Blümlein. B 97 (1980) 437. Phys. Humpert. Rijken. G. Phys. R. Phys. Stratmann. Hasslacher. D 25 (1982) 1211. Christ. Rev. Lacaze. T. Sov.L. J. C. M. E. N. B 175 (1980) 27. Rev. A. B 272 (1991) 127. Nucl. Ellis. . Phys. Moch. S. B 523 (1998) 245. D 60 (1999) 014018. Blümlein. Mueller. Nucl. Zijlstra. Phys. Lett. W. Van Neerven. C 11 (1982) 293. Lett. Lacaze.D. R.J.J. R. Sanchez Guille’n et al. B 157 (1979) 461. van Neerven. Phys.A. Floratos. B 487 (1997) 233. Lett. Nucl. B. Phys. R. D 54 (1996) 2023. C 70 (1996) 637. Catani. B 143 (1978) 521. E. W. B 383 (1992) 525. C. Reya. Curci. B 297 (1992) 377. Blümlein et al. Lett. Z. Petronzio. B 392 (1997) 207. Bukhvostov. Zijlstra. Phys.. Phys.B. Zijlstra. Popov. Phys. W. G 25 (1999) 1551. Floratos. B 353 (1991) 337. Phys. Nucl. J. Phys. B 192 (1981) 417. S. DESY-97-160. Lett.P. B. van Neerven. Rev. Phys. Phys. Phys. Furmanski. E. Leiden University. Kounnas. Phys. R. Muta. E.B. Parisi. D. W. Gatto. Ph.N. W. Phys.L. Duke. Erratum. D 6 (1972) 3543. Mertig.B. Phys. Thesis. P.H. Thesis. Glück. G. Phys. Nucl. van Neerven. J. G.K. Rev. Bardeen. Phys. Nucl. Rev. 1997.L. Furmanski. Rijken. S. Martinelli. Phys.G. Phys. W. Leiden University. Grunberg. B 98 (1981) 89. Rijken. E.L. B 496 (1997) 41. W. E.J. J. R. Nucl. Phys. Nucl. D 18 (1978) 3998. Lett. Glück. Phys. Ravindran.W. B 426 (1994) 245. J. Reya. hep-ph/0003100. Rev. C 75 (1997) 665. Petronzio. D 29 (1984) 2315.L. Nucl. Kurth. Lett. M.L. Vogt. B 386 (1996) 422. Phys. L. P. Z. B 146 (1978) 544 (Erratum). Phys. A. E.D. Lett. Nucl. Phys. G. J. A. Phys. G. Rev. hep-ph/9912355.G. Furmanski. 20 (1975) 287. Vogelsang. Buras. D 48 (1993) 116. Phys. A. hep-ph/9708388. B 184 (1981) 225. Phys. Blümlein. Z. Nucl. B 44 (1973) 381. W. Altarelli. W.

Capella b.elsevier.fr 2 kaidalov@vxitep. ΛQCD and in the gluon distribution function. PACS: 12. Our calculations show that nonperturbative corrections to a B-meson azimuthal asymmetry are negligible. We estimate also nonperturbative contributions to azimuthal distributions due to the gluon transverse motion in the target and the final quark fragmentation.2.88. Heavy flavor photoproduction. All rights reserved. A.itep. F-91405 Orsay Cedex. these NLO results have been widely used for a phenomenological description of available data (for a review see [9]). µF .0 .  2000 Elsevier Science B.V. A.Nuclear Physics B 586 (2000) 382–396 www. At leading order.am 1 capella@qcd.85. During the last ten years. accepted 20 June 2000 Abstract We analyze in the framework of pQCD the properties of the single spin asymmetry in heavy flavor production by linearly polarized photons. Russia Received 22 November 1999.38.yerphi. µR . All rights reserved.Cheremushkinskaya 25. Both leading and next-to-leading order predictions for the asymmetry are insensitive to within few percent to theoretical uncertainties in the QCD input parameters: mQ . revised 29 May 2000. 117259 Moscow. Our analysis shows that the next-to-leading order corrections practically do not affect the Born predictions for the azimuthal asymmetry at energies of the fixed target experiments.Bx. Single spin asymmetry 1.B. We conclude that measurements of the single spin asymmetry would provide a good test of pQCD applicability to heavy flavor production at fixed target energies. the key ∗ Corresponding author. At the same time.+e. France c Institute of Theoretical and Experimental Physics.nl/locate/npe Single spin asymmetry in heavy flavor photoproduction as a test of pQCD N.2 a Yerevan Physics Institute.th. Introduction In the framework of perturbative QCD. the basic spin-averaged characteristics of heavy flavor hadro-. Universite de Paris XI. Armenia b Laboratoire de Physique Theorique.Ni Keywords: Perturbative QCD.V. PII: S 0 5 5 0 . 13. 13. the parallel–perpendicular asymmetry in azimuthal distributions of both charm and bottom quark is predicted to be about 20% in a wide region of initial energy. Batiment 210. E-mail: nikiv@uniphi. Kaidalov c.u-psud.3 2 1 3 ( 0 0 ) 0 0 3 9 9 . photo.and electroproduction are calculated up to the next-to-leading order (NLO) [1–8].1 .ru 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Ivanov a. Alikhanian Br.∗ .Ya. B. 375036 Yerevan.

The main results of our analysis can be formulated as follows: − At fixed target energies. ϕ).2) Here dσ (s. Since the spin-averaged characteristics of heavy flavor production are not well defined quantitatively in pQCD it is of special interest to study those spin-dependent observables which are stable under variations of input parameters of the theory. µF and µR . ϕ = 0) + dσ (s. In particular.1) γ ↑ +N → Q Q + X. ϕ = π/2) (1. . we calculate the single spin asymmetry parameter. without additional assumptions. the pQCD predictions for spin-averaged cross sections with experimental data because of a high sensitivity of the theoretical calculations to standard uncertainties in the input QCD parameters.10]. Pγ is the degree of linear polarization of the incident photon √ beam. the reactions  (1. however they are also significant. especially for the c-quark case. The total uncertainties associated with the unknown values of the heavy quark mass.N. / Nuclear Physics B 586 (2000) 382–396 383 question still remains open: How to test the applicability of QCD at fixed order to the heavy quark production? On the one hand. For this reason. which measures the parallel–perpendicular asymmetry in the quark azimuthal distribution: A(s) = 1 dσ (s. mQ . namely. the LO predictions for azimuthal asymmetry (1. ϕ = π/2) . In particular. In this paper we analyze the charm and bottom production by linearly polarized photons. at the energies of the fixed target experiments the theoretical calculations are especially sensitive to the value of the heavy quark mass which plays the role of a cutoff parameter for infrared singularities of the theory.Ya. For instance. they increase the leading order (LO) predictions for both charm and bottom production cross sections approximately by a factor of 2.8]. Ivanov et al. ϕ) ≡ dσ dϕ (s. it is very difficult to compare directly. the NLO corrections are large. s is the centre of mass energy of the process (1. ΛQCD and the parton distribution functions are so large that one can only estimate the order of magnitude of the pQCD predictions for total cross sections [7.2) are not small and can be tested experimentally. ϕ = 0) − dσ (s.1) and ϕ is the angle between the beam polarization direction and the observed quark transverse momentum. In fact these uncertainties are of the same order as the contributions of nonperturbative effects (such as the primordial transverse motion of incoming partons and the hadronization phenomena) which are usually used for a phenomenological description of the charm and bottom differential spectra [9. A(s). For the shapes of the one. On the other hand. one could expect that the higher order corrections as well as the nonperturbative contributions can be essential in these processes.and two-particle differential distributions the above uncertainties are moderate in comparison with the ones for total cross sections. the factorization and renormalization scales. Pγ dσ (s.

LO .

3) A s = 400 GeV2 .LO (1.

Charm ≈ A s = 400 GeV2 .

both real and virtual. practically do not affect the Born predictions for A(s) at fixed target energies. − The NLO corrections. − At energies sufficiently above the production threshold. both leading and nextto-leading order predictions for A(s) are insensitive (to within few percent) to .18.Bottom≈ 0.

In Section 3 we discuss the nonperturbative contributions caused by both kT -kick and Peterson fragmentation mechanisms. mc . Leading order predictions At leading order. 2. pT . the kT -kick corrections are essential at low values of pT . Because of the smallness of the c-quark mass.Ya.and B-meson photoproduction. A comparison of the QCD predictions with the Regge model ones is also given. In Section 2 we analyze the properties of heavy quark azimuthal distributions at leading order. and rapidly vanish at pT > mc . µF .1) . For this purpose we use a parametrization of the intrinsic transverse momentum distribution proposed in Ref. For the D-meson. (2. / Nuclear Physics B 586 (2000) 382–396 uncertainties in the QCD parameters: mQ . Another important question is how the fixed order predictions for asymmetry parameters are affected by nonperturbative contributions usually used for the description of unpolarized spectra.384 N. respectively) cancel each other with a good accuracy. Ivanov et al. ΛQCD and in the gluon distribution function. Modeling nonperturbative effects with the help of the kT -kick and the Peterson fragmentation function we analyze also the pT . Measurements of asymmetry parameters would provide a good test of the fixed order QCD applicability to charm and bottom production. Our analysis shows that the pT . they will be reported separately in a forthcoming publication [11].2). they are of order of 20%. It is shown that for the B-meson both the kT -kick and the fragmentation contributions are small almost in the whole region of kinematical variables pT and xF . µR . This implies that theoretical uncertainties in the spin-dependent and spin-averaged cross sections (the numerator and denominator of the fraction (1. the only partonic subprocess which is responsible for heavy quark photoproduction is the two-body photon–gluon fusion: γ (kγ ) + g(kg ) → Q(pQ ) + Q(pQ¯ ). We estimate the nonperturbative corrections to A(s) due to the transverse motion of partons in the target. To be specific: Theoretical uncertainties in the pQCD predictions for the azimuthal asymmetry differential distribution are always significantly smaller than the ones in calculations of the shape of the corresponding spectrum of unpolarized cross section.and xF -spectra of the single spin asymmetry also depend weekly on the variations of QCD parameters. The details of our calculations of radiative corrections are too long to be presented in this paper. O(αem αS ). [10] (so-called kT -kick). Single spin asymmetry in pQCD 2.and xF -behavior of azimuthal asymmetry in D. We conclude that the single spin asymmetry is an observable quantitatively well defined and rapidly convergent in pQCD. It is shown that the kT -kick corrections to the b-quark azimuthal asymmetry A(s) are negligible. We also give the physical explanation of the fact that the pQCD predictions for A(s) are approximately the same at LO and at NLO. the kT -kick corrections to A(s) in the charm case are larger. The paper is organized as follows.1.

T =  sˆ 2 β − xˆ 2 . pEQ. t = (kγ − pQ )2 = tˆ.5) dtˆ dϕ Here g(z. For this reason we do not take into account the contribution of so-called hadronic or resolved component of the photon. and the beam polarization direction. tˆ.6) In this paper we will discuss photoproduction processes only at fixed target energies. and eQ is the quark charge in units of electromagnetic coupling constant. xˆ = 1 + 2 sˆ sˆ = (kγ + kg )2 . can be written in the form EQ d 3 σ (s. We use the following definition of partonic kinematical variables: tˆ = (kγ − pQ )2 . In (2. tˆ. Photon–gluon fusion diagrams at leading order. s 4m2Q .2) C is the color factor. β = 1− sˆ 2 pEQ.N. uˆ = (kg − pQ )2 . u. ϕ) = d3 pQ Z (2.2) where Pγ is the degree of the photon beam polarization. γ (kγ ) + N(kN ) → Q(pQ ) + X(pX ). µ2F ) describes gluon density in a nucleon N evaluated at a factorization scale µF . ϕ is the angle between the observed quark transverse momentum. (2. 1. u = (kN − pQ )2 = u/z ˆ − m2Q (1/z − 1). t. 4 (2. the invariant cross section of the single inclusive hadronic process.Ya. ϕ. The hadronic variables are related to the partonic ones as follows: s = (kγ + kN )2 = sˆ /z. / Nuclear Physics B 586 (2000) 382–396 385 Fig. Ivanov et al. C = TF = Tr(T a T a )/(Nc2 − 1) = 1/2.T . The cross section corresponding to the Born diagrams (see Fig. ϕ. × 1 − xˆ 2 (1 − xˆ 2 )2 (2. tˆ − m2Q . µ2R = C dxˆ dϕ sˆ   1 + xˆ 2 2(1 − β 2 )(β 2 − xˆ 2 ) + (1 + Pγ cos 2ϕ) . µ2R .3) In the Born approximation.4)   2ˆs d2 σˆ  dz δ sˆ + tˆ + uˆ − 2m2Q g z. µ2F sˆ . 1) is: 2 α α (µ2 ) eQ  em S d2 σˆ R sˆ . (2. This contribution is small at the energies under consideration [7]. .

2 2 2 respectively. pT2 .386 N.5) integrated over xF and over pT2 . and on the Feynman variable. 2. The quantities d σ (s. A(Eγ ). pT2 .5) integrated over both xF and pT2 . xF . Ivanov et al. in the c-quark production as a function of beam energy Eγ = (s − m2N )/2mN .max . . Fig. ϕ = π/2) . ϕ = π/2) Ax (xF ) = 1 d2 σ (s. Apart from the quantity A(s) defined by (1. 3. ϕ = 0) − d2 σ (s.  1 d2 σ (s. pT2 . ϕ = π/2) (2.2). ϕ) and d σ (s. of observed particle. xF . x = p /p pT2 ≡ pEQ. ϕ = 0) − d2 σ (s. ϕ = 0) + d2 σ (s. The same as in Fig. Fig. xF . xF . Ap pT2 = Pγ d2 σ (s. we will consider also the parameters Ap (pT2 ) and Ax (xF ). 2 .Ya.T F l l. pT2 . Single spin asymmetry. ϕ) are the cross section (2. the QCD LO predictions with and without the inclusion of kT -kick effect. / Nuclear Physics B 586 (2000) 382–396 To analyze the azimuthal distributions in photoproduction. it is convenient to use the parallel–perpendicular asymmetry parameters. The quantity dσ (s.7) (2. but for the b-quark case. ϕ = π/2) . xF . ϕ = 0) + d2 σ (s. Pγ d2 σ (s.2) corresponds to the cross section (2.8) which describe the dependence of azimuthal asymmetry on the transverse momentum. pT . 2. respectively. ϕ) in (1.

5 GeV and mb = 4. The most interesting feature of LO predictions for A(Eγ ) is that they are practically insensitive to uncertainties in QCD parameters. we use µF |Charm = q q p. Scaling behavior of the asymmetry parameter A(η) as a function of η = 4m2Q /s at QCD LO. changes of the charm quark mass in the interval 1.2). we use µR = µF . We have verified also that all the CTEQ3 versions of the gluon distribution function [12] as well as the CMKT parametrization [13] lead to asymmetry predictions which coincide with each other with accuracy better than 1.x 2 m2c + pT2 for charm and µF |Bottom = m2b + pT2 for bottom. Remember that analogous changes of mc lead to variations of total cross sections from a factor of 10 at Eγ = 40 GeV to a factor of 3 at Eγ = 1 TeV.8 GeV affect the quantity A(Eγ ) by less than 6% at energies 40 < Eγ < 1000 GeV. Calculating p.7 at Eγ = 1 TeV.5 and mb = 5 GeV lead to 3% variations of the parameter A(Eγ ) in the case of bottom production at energies 250 < Eγ < 1000 GeV.5 at Eγ = 1 TeV. One can see that at energies sufficiently above the production threshold the single spin asymmetry A(Eγ ) depends weekly on Eγ . the CTEQ3M [12] parametrization of the gluon distribution function is used. For the renormalization scale. 4. The total cross sections in this case vary from a factor of 3 at Eγ = 250 GeV to a factor of 1.5%. In particular. (2. The changes of A(Eγ ) are less than 3% for choices of µF in the range 1 2 mb < µF < 2mb .x the pT . / Nuclear Physics B 586 (2000) 382–396 387 Unless otherwise stated.Ya. For the total cross sections.and b-quark are presented by solid lines in Fig.2 < mc < 1. Ivanov et al.N.7) and (2.75 GeV. The extreme choices mb = 4.10]. The default values of the factorization scale µF chosen for the A(s) asymmetry calculation are µF |Charm = 2mc for the case of charm production and µF |Bottom = mb for the bottom case [9.and xF -dependent asymmetries Ap (pT2 ) and Ax (xF ). The default values of the charm and bottom mass are mc = 1. however. Let us discuss the pQCD predictions for the asymmetry parameters defined by (1. such changes of µF lead to a factor of 2. that the fixed order predictions for azimuthal asymmetry are insensitive to uncertainties in . 2 and Fig. Λ4 = 300 MeV and Λ5 = 200 MeV. Our calculations of A(s) at LO for the c. 3. Eγ = (s − m2N )/2mN . µR .7 at Eγ = 250 GeV and of 1. 3 Fig. respectively. 3 Note.8).

Ivanov et al. so that .Ya.388 N. η = 4m2Q /s. / Nuclear Physics B 586 (2000) 382–396 In the Born approximation scaling holds: with a good accuracy the quantity A(s) is a function of the variable η.

LO .

≈A s .

.

Bottom 2 m .

LO A(s).

e. Fig.9). (2.25) ≈ A(η = 0. 4 Eq. 4. 5. The same as in Fig. its independence of ΛQCD /mQ ) is demonstrated in Fig.9) Charm The scaling behavior of A(η) (i. (1. 5. For instance. the asymmetry parameter A(η) is not a monotonic function of variable η.  m2c b . in a D-meson production. (1.3) reflects only the fact that A(η = 0. (2. but for the pT -distribution of azimuthal asymmetry.. Ap (pT2 ). 6.3) does not contradict to Eq. QCD parameters only at energies sufficiently above the production threshold.2 < mc < 1. changes of the charm quark mass in the interval 1.025). pT2 -distribution of azimuthal asymmetry. Eq. . As one can see from Fig. 4 Fig.8 GeV lead to 100% variations of the QCD predictions for A(s). Ap (pT ). 4. So. in a B-meson production. at Eγ < 10 GeV. the QCD LO predictions with and without the inclusion of the kT -kick and Peterson fragmentation effects.

mb . We have also verified that a change in the other indpT put parameters has practically no effect on the computed values of Ap (pT2 ) for both charm and bottom. the form of the pT2 -spectrum is sensitive to the value of charm quark mass. . Ap (pT ). at QCD LO. contrary to A(s). Dependence of the pT2 -spectrum. In the Born approximation the pT -distribution of azimuthal asymmetry is practically a function of two variables: η = 4m2Q /s and ξ = pT2 /m2Q . pT -distributions grow rapidly with pT up to pT ≈ mQ and fall slowly as pT becomes larger than the heavy quark mass. Dependence of the pT -spectrum. The dependence of pT -spectra on variations of mQ is shown in Fig. mc . 9. pT2 ) [9]. The function Ap (ξ ) at different values of η is shown in Fig. 8.N. One can see that.Ya. 8. Ivanov et al. / Nuclear Physics B 586 (2000) 382–396 389 In Fig. Note however that variations of the Ap (pT2 ) shape due to changes of mc are significantly smaller than the corresponding ones for the unpolarized distribution dσ2 (s. on the b-quark mass. Ap (pT2 ). 5/Fig. 7. on the c-quark mass. 6 we show by a solid line the LO predictions for the pT2 -/pT -dependent parameter Ap (pT2 )/Ap (pT ) describing the azimuthal asymmetry in charm/bottom production. Fig. Fig. at QCD LO. 7 and Fig.

10. Fig. ξ = pT2 /m2Q . 10. / Nuclear Physics B 586 (2000) 382–396 Fig. xF -distribution of the single spin asymmetry. 11. but for the case of a B-meson production. Scaling behavior of the asymmetry parameter Ap (ξ ) as a function of ξ .Ya. Ivanov et al.390 N. The same as in Fig. Fig. in a D-meson production. the QCD LO predictions with and without the inclusion of the kT -kick and Peterson fragmentation effects. Ax (xF ). at different values of η at QCD LO. 9. .

4].8) in the case of c. the threshold enhancement of the ISGB mechanism is due to the gluon propagator pole in the diagrams in Fig. Eγ . originates from the so-called initial-state gluon bremsstrahlung (ISGB) mechanism which is due to the Feynman diagrams with the t-channel gluon exchange (see Fig. It is well known that the shapes of differential cross sections of heavy quark production in photon–hadron reactions are not sensitive to radiative corrections [3. .5 and fall to zero as xF → 1.and b-quark production are presented by solid lines in Fig. For instance. π dtˆ dtˆ sˆ ∼ 4m2Q . the order-αS2 photon– hadron cross section is determined at Eγ . xF -distributions take their maximal values approximately at xF ≈ 0. the main heavy quark photoproduction mechanism is the real gluon emission in the photon–gluon fusion: γ (kγ ) + g(kg ) → Q(pQ ) + Q(pQ¯ ) + g(pg ). Since the azimuthal angle ϕ is the same for both γ g and QQ centre of mass systems in the collinear and soft limits. t -channel gluon exchange diagrams in the photon–gluon fusion.T → 0. 12.11) by the ϕdependent ones. Indeed. limits [3. pEg. Radiative corrections Let us briefly discuss the NLO corrections to the single spin asymmetry A(s). tˆ . / Nuclear Physics B 586 (2000) 382–396 391 We have calculated also the dependence of single spin asymmetry on xF .11) can be generalized to the spin-dependent case substituting the spin-averaged cross sections in (2. sˆ ∼ 4m2Q . (2.Ya. Ivanov et al. 1 TeV. 12).4]. 2. Near the threshold.N. at order O(αem αS2 ). pEg → 0. (2. a large logarithmic enhancement of the t-channel gluon exchange diagrams contribution takes place in the collinear. Effectively. tˆ ≈ sˆ. g(z) ∝ 1/z.10) According to the analysis [3. it seems natural to expect that the Eq. and soft.4]. It is well known that. the ISGB contribution is proportional to the Born γ g differential cross section:  αS  dσˆ LO dσˆ ISGB K(ˆs ) sˆ. The LO predictions for Ax (xF ) defined by (2. Since the gluon distribution function increases very steeply at small z. behavior of the γ g cross section. 12 which is a common factor for both spindependent and spin-independent amplitudes.11) where K(ˆs ) is an enhancement factor. 1 TeV by the threshold. practically whole contribution to the heavy quark production in photon–hadron reactions at fixed target energies. respectively. (2. 11.2. 10 and Fig. the normalized quantity Fig.

12) 2 .T . has a random Gaussian distribution:   kT2 1 1 d2 N exp − 2 . µF and ΛQCD ) at NLO too. Moreover.392 N. pT2 = where σint (s) = R  dσ 1 s. pT2 ) ≈ dz g(z) dσ ISGB (zs.2) N d2 kT πhkT2 i hkT i 2 . kEg. the azimuthal asymmetry is independent (to within few percent) of the theoretical uncertainties in the QCD input parameters (mQ . pT2 ) and pT2 ≡ pEQ. For instance. is practically the same at LO and at (s.10]. kEg = zkEN + kEg. 12.12 ) with a good accuracy.Ya. pT2 .T NLO. also arising from the diagrams with the t-channel gluon exchange in Fig. This fact means that non-small contributions of the enhancement factor K(ˆs ) (NLO corrections) to theR pT -dependent and to the pT -integrated photon–hadron cross sections. In our analysis. However the contribution of the FE mechanism may be essential at superhigh energies only [4]. 1 TeV.2) (so-called kT -kick) together with the Peterson fragmentation function [14]. these contributions are especially important in the case of charmed particle production. (3. both real and virtual. the parametrization (3.1) According to [10]. µR . the primordial transverse momentum. / Nuclear Physics B 586 (2000) 382–396  f s. to the photon–gluon fusion mechanism practically do not affect the Born predictions for the single spin asymmetry in heavy flavor photoproduction at fixed target energies. 2 σint (s) dpT dpT2 dσ dpT2 (2. As to the photon–quark fusion. Because of the c-quark low mass. The details of our NLO analysis are too long to be presented here and will be given in a separate publication [11].T . Radiative corrections. the dominant production mechanism in this reaction is the so-called flavor excitation (FE). Nonperturbative contributions Let us discuss how the pQCD predictions for single spin asymmetry are affected by nonperturbative contributions due to the intrinsic transverse motion of the gluon and the fragmentation of produced heavy quark. dσ NLO(s. It is evident that an account of the photon–quark reactions cannot affect significantly the predictions of the photon–gluon fusion mechanism at fixed target energies in the polarized case too. = (3. . Our calculations show that it is really the case. In [9. pT2 ). γ q → QQq. cancel each other in the ratio (2. Ivanov et al.T azimuthal asymmetry. the contribution of the γ q reactions to the unpolarized bottom photoproduction makes only 5–10% from the contribution of the γ g processes at Eγ . It is evident that the inclusion of this effect results in a dilution of where kT2 ≡ kEg. we use the MNR model [10] parametrization of the gluon transverse momentum distribution. 3. One can assume that the same situation takes place also for the single spin asymmetry which is a ratio of the ϕ-dependent cross section to the ϕ-averaged one.

3) leads to a reduction of the pT -spectra. Nonperturbative contributions to the pT -dependent asymmetry parameters Ap (pT2 ) and p A (pT ) are shown in Fig.Ya. they are negligibly small in the whole region of pT . 2 and Fig. 2 ≈ 0).N. It is seen that the kT -kick corrections decrease rapidly with the increase of the heavy quark transverse mass. In the The latter can be nonperturbative. It was shown that the available data on charm photoproduction allow to choose for the averaged intrinsic transverse momemtum. / Nuclear Physics B 586 (2000) 382–396 393 aε (3. 6 that for pT > mQ the account of the fragmentation function (3.5 GeV2 and by 20% at hkT2 i = 1 GeV2 .7) due to the kT -kick and Peterson fragmentation effects practically do not affect predictions of the underlying perturbative mechanism: photon–gluon fusion. In the case of a D-meson production. 6.006 for a B-meson [15]. reggeons at pO Born approximation. color or white objects (say. we can conclude that nonperturbative corrections to the b-quark asymmetry parameters (1.2) and (2. have been used to describe the single inclusive spectra and the QQ correlations in photo. Fragmentation corrections to Ap (pT2 ) are of order of 15% in the case of a charmed meson production. Our calculations of the parameter A(s) at LO with the kT -kick contributions are presented in Fig. It is seen from Fig. hkT2 i. Ivanov et al. 5 and Fig.4) have the following form:   2α α eQ em QSQ 4(1 − β 2 )(β 2 − xˆ 2 ) 2 d2 σˆ O=S =C − (1 + Pγ cos 2ϕ) . any value between 0. in Fig.5 GeV2 and hkT2 i = 1 GeV2 . The kT -kick corrections to the bottom asymmetry are systematically smaller.5 and 2 GeV2 . 11 we presented nonperturbative contributions to the xF -distributions of single spin asymmetry. 10 and Fig. Calculating the hadronization effect contributions we use for the parameter ε that characterizes the Peterson fragmentation function the values εD = 0.06 for a D-meson and εB = 0. For completeness. mc . (3. the kT kick corrections are essential only in the region of low pT . 5 and Fig. the application of the fragmentation function formalism can only be justified in the region of high pT . the xF -distribution of the c-quark azimuthal asymmetry is more sensitive to the kT -kick corrections. Strictly speaking. One can see that the kT -kick corrections to Ax (xF ) in the bottom case are small in the whole region of xF . the partonic cross sections corresponding to the reactions D(y) = γ +O→Q+Q (3. To illustrate how strongly the azimuthal distributions depend on the basic subprocess dynamics let us consider the mechanisms of the photon fusion with nonvector primordials. according to the factorization theorems.5) dxˆ dϕ sˆ 1 − xˆ 2 (1 − xˆ 2 )2 . So.3) y[1 − 1/y − ε/(1 − y)]2 R1 (where aε is fixed by the condition 0 dyD(y) = 1). One can see that in the c-quark case the kT -kick reduces the value of A(s) approximately by 15% at hkT2 i = 0. 3 by dashed (hkT2 i = 0.5 GeV2 ) and dotted (hkT2 i = 1 GeV2 ) curves. For the B-meson. they do not exceed 5% for both hkT2 i = 0. they are less than 5% for a B-meson case.and hadroproduction at NLO. As expected.

The first. the spin dependent part of a D-meson photoproduction cross section can be of two types. the Regge pole model is only applicable in the narrow region xF ≈ 1 and pT2  s (so-called triple reggeon exchange limit). the cross section of the inclusive reaction is determined by the contribution of the leading (rightmost in the j -plane) Regge trajectory. describes the contribution of the t-channel singularities of unnatural spin-parity. is of natural spinparity.Ya.394 N. In the above expressions. which describes the pseudoscalar meson photoproduction in the triple reggeon exchange limit. (3.2) and (3. Note that the expressions (3. Since the relevant leading Regge trajectory. they are presented only as illustrative exercises demonstrating how the azimuthal distribution reflects the nature of the underlying subprocess. respectively.5). In principle. at arbitrary values of variables xF and pT . In this region. |(EepET )|2 = pET2 cos2 ϕ. π 0 -meson) leads to an ϕ-independent cross section. the asymmetry parameters are predicted to be negative.6) where the cross sections σˆ O=S and σˆ O=P describe the photon fusion with a scalar (O = S) and a pseudoscalar (O = P ) object. nE 0 ]pET |2 = pET2 sin2 ϕ. Really. (3. is written as: . dxˆ dϕ sˆ 1 − xˆ 2 2 (3.6) do not correspond to any concrete models. nE0 is a unit vector in the direction of the photon momentum and ϕ is the azimuthal angle of the D-meson momentum about nE0 : tan ϕ = ([Ee. / Nuclear Physics B 586 (2000) 382–396 eQ αem αQP Q 2 d2 σˆ O=P =C . the spin structure of the corresponding reggeon-particle vertex. is sketched in Fig. eE is the polarization vector of the photon.6) that only the gluon contribution leads to a positive azimuthal asymmetry. Completing the section let us compare the QCD predictions for a D-meson single spin asymmetry with the Regge pole model ones. In the case of a scalar primordial. 13. Ivanov et al.5). One can see from (2. nE 0 ]pET )/(EepET ). |[Ee. The diagram. The photon fusion with a pseudoscalar object (say. αD ∗ . corresponds to the contribution of the t-channel exchanges with natural spin-parity while the second. γ αD ∗ D.

.

β γ α D∗ D .

2 .

.

2 1 (pT ).

∝ .

[Ee. nE0 ]pET .

7) that. = pET2 (1 − cos 2ϕ).7) It is seen from (3. Triple reggeon diagram for the process γ N → DX. the Regge model predicts in the region xF ≈ 1 a large negative azimuthal asymmetry such that Ax (xF )|Regge → −1 as xF → 1. Fig. . 13. in contrast to QCD. 2 (3.

J.1) with a linearly polarized photon beam are of special interest. Phys. the reactions (1.LG971390. [16. As it is shown in the present paper. A(s) is independent to within few percent of the theoretical uncertainties in mQ . Phys. B 327 (1989) 49. Ellis. As it is shown in our paper. . R. Nucl. Measurements of the azimuthal asymmetry in bottom production would be a good test of the conventional parton model based on pQCD. Phys. R. Dawson. Unlike the unpolarized cross section. So. Smith. Rev. Our analysis shows that the NLO corrections practically do not affect the Born predictions for A(s) at fixed target energies. van Neerven. Beenakker. Our calculations show that the pT .K. In particular. the linearly polarized high energy photon beams can be generated using the Compton back-scattering of the laser light off the lepton beams (see. H. J. µR . Dawson. Ellis.L. S. B 374 (1992) 36. nonperturbative contributions to the charm production can be essential. Nason. Due to the c-quark low mass. According to the above references. rapidly convergent and insensitive to nonperturbative contributions. Acknowledgements This work was supported in part by the grant NATO OUTR.L. Ivanov et al. Phys. B 335 (1990) 260 (Erratum).K.Ya. Ellis. Phys. van Neerven. the QCD LO predictions for the single spin asymmetry in both charm and bottom production are non-small and can be tested experimentally. the single spin asymmetry is an observable quantitatively well defined in QCD. for instance.17] and references therein).K. D 40 (1989) 54. Nason.N. P. R. References [1] [2] [3] [4] [5] P. B 312 (1989) 551. / Nuclear Physics B 586 (2000) 382–396 395 4. Conclusion In the investigation of mechanisms which are responsible for heavy flavor production at fixed target energies. in a wide region of initial energy the parameter A(s) defined by (1. P. Nucl. this method promises to provide the beams of real photons with a definite polarization and high monochromaticity. Smith. W. µF . B 303 (1988) 607. Kuijf. Data on the pT . S.2) is of order of 20%. the single spin asymmetry in heavy flavor production by linearly polarized photons is an observable quantitatively well defined. at both LO and NLO. W.and xF -distributions of the azimuthal asymmetry in D-meson production could make it possible to discriminate between these mechanisms.and xF -distributions of the azimuthal asymmetry in bottom production are practically insensitive to nonperturbative contributions due to the primordial transverse motion and Peterson fragmentation effects. Nucl. Nucl. W. ΛQCD and in the gluon distribution function. Nason. the pQCD and Regge approaches lead to strongly different predictions for the single spin asymmetry in the region of low pT and xF ≈ 1. Phys. Concerning the experimental aspects. Nucl.

L. / Nuclear Physics B 586 (2000) 382–396 [6] W. G.. Nucl. [12] H. Phys. Ridolfi. Ivanov. Nucl. Ridolfi. Serbo. [17] V. P.L. Mangano. . Instr. M. 219 (1984) 5. Kaidalov. B 373 (1992) 295. Capella. [8] S. B 337 (1994) 358. J. Smith. Ginzburg. in: P. G.Ya. A. Frixione. Nason. M. [7] M.I. Phys. Quantum aspects of Beam Physics: Proceedings. B 351 (1991) 507. Kaidalov. Peterson. G.B. World Scientific. G. M. B 412 (1994) 225. Chrin. A. Ivanov et al.L. hep-ph/9805002. 1998. Nucl. Nucl. Nason. van Neerven. D. Nason. [13] A. Phys. Mangano. S. Rev. Ridolfi. Phys. Nucl. Zerwas. J. Merino. A. Nason.Ya. Schmitt. G. Singapore.L. Schuler. C. C 36 (1987) 163.396 N. Lai. [11] N.J. Phys. p.).B. Heavy Flavours II. Phys. Schlatter. Mangano.L. World Scientific. Botts. [10] M. Meth. Chen (Ed. [16] I.L. Phys. Buras. Mangano. Panfil. B 405 (1993) 507. D 27 (1983) 105. Capella. D 51 (1995) 4763. Singapore. R. V. I.). Ridolfi.A. Tran-Thanh Van.L. [15] J. G.L.F. Huston et al. J. 173. hep-ph/9702287. V. P. Frixione. in: A. Telnov. Kotkin. Phys. Meng. J. Advanced Series on Directions in High Energy Physics. P. Beenakker. in preparation. Lindner (Eds. P. Z. Lett. Rev. [9] S. Telnov.G. [14] C. P. 1998. W.

ch (M. we calculate the ADM of 1F = 2 operators again. In the case of 1F = 2 operators.nl/locate/npe Two-loop QCD anomalous dimensions of flavour-changing four-quark operators within and beyond the Standard Model Andrzej J.  2000 Elsevier Science B. Germany b Theory Division. An essential ingredient in any renormalization group analysis is the anomalous dimension matrix (ADM).elsevier. Introduction Renormalization group short-distance QCD effects play an important role in the phenomenology of non-leptonic weak transitions of K-. Jörg Urban a a Physik Department. CH-1211 Geneva 23. CERN. E-mail address: mikolaj. Switzerland c Institute of Theoretical Physics. which describes the mixing of the relevant local four-quark operators under renormalization [1.2].and B-mesons. D. 0550-3213/00/$ – see front matter  2000 Elsevier Science B. Mikołaj Misiak b. accepted 6 July 2000 Abstract We calculate the two-loop QCD anomalous dimension matrix (ADM) (γˆ (1) )NDR in the NDR–MS scheme for all the flavour-changing four-quark dimension-six operators that are relevant in both the Standard Model and its extensions. β β Ψ¯ 1α ΓAk Ψ2 Ψ¯ 3 ΓBk Ψ4α . In order to explain the difference.5 (1. The operators considered in the present paper have the form β β Ψ¯ 1α ΓAk Ψ2α Ψ¯ 3 ΓBk Ψ4 . All rights reserved. Poland Received 25 May 2000. PII: S 0 5 5 0 .3 2 1 3 ( 0 0 ) 0 0 4 3 7 .Nuclear Physics B 586 (2000) 397–426 www. Misiak). Both current–current and penguin diagrams are included. extracting it from the ADM of 1F = 1 operators.1) . 1. in the regularization-independent renormalization scheme. Ho˙za 69. D-85748 Garching. our results are compatible with the ones obtained by Ciuchini et al. PL-00-681 Warsaw. Some of our NDR–MS results for 1F = 1 operators overlap with the previous calculations.misiak@cern.V.c.V. All rights reserved. Warsaw University.∗ . ∗ Corresponding author. Technische Universität München. but differ from their NDR–MS results. Buras a . but several others have never been published before.

in the Supersymmetric Standard Model (SSM) (see. Ref. such as neutral meson mixing. one has Ψ 1 = Ψ3 . / Nuclear Physics B 586 (2000) 397–426 k where α. Our paper is organized as follows. For instance.3) For these operators.1)): ΓAk ⊗ ΓBk = (1 ± γ5 ) ⊗ (1 ± γ5 ). [3]). The subscripts i in Ψi are flavour indices. Buras et al. and give the corresponding anomalous dimensions.g. e. Section 6 is devoted to performing a consistency check of our 1F = 1 and 1F = 2 results.g. e. Other operators need to be considered in many extensions of the SM. respectively. [4]. no new calculation is actually necessary — all the two-loop results can be extracted from the existing Standard Model ones. Four-quark operators that occur in the SM calculations of flavour-changing processes do not form a complete set of all the dimension-six four-quark operators.. The current–current ADM of 1F = 1 operators is transformed there to such an operator basis.. (1. [4]. we compute the NDR–MS ADM for such 1F = 1 operators.B are generic Dirac matrices given explicitly below. The subject of Section 4 are 1F = 1 operators containing one quark–antiquark pair of the same flavour.g.398 A. but are compatible with their RI-scheme ADM. our two-loop results differ from the NDR–MS ones of Ciuchini et al. to which only the current–current diagrams are relevant. e. Comparison with this article and other existing literature is the subject of Section 7. The main purpose of the present paper is a calculation of the two-loop ADM for all the dimension-six flavour-changing four-quark operators in the NDR–MS scheme ( MS scheme with fully anticommuting γ5 ). . Refs.2) ¯ V −A (bd) ¯ V −A relevant in the Known examples are the operators (¯s d)V −A (¯s d)V −A and (bd) 0 0 0 0 Standard Model (SM) to K –K¯ and Bd –B¯ d mixing. β are colour indices and ΓA. In Section 5. For all the other operators.J. In Section 3. [5–8]). Ψ2 = Ψ4 . We conclude in Section 8. We identify the operators to which the so-called penguin diagrams are relevant. In the case of FCNC transitions with 1F = 2.. since all the methods are already known from similar SM calculations (see. (1. (1. The only novelty here is the introduction of evanescent operators that vanish by the Fierz identities. In Section 2. in which the 1F = 2 results can be easily read off. The ones that are not new agree with the old SM calculations. Eq. Our main findings are the NDR–MS anomalous dimensions of the operators with Dirac structures (cf. This is a relatively straightforward computation. Some of the 1F = 1 results have never been published before. we derive the matrix 1ˆr necessary for transforming the Wilson coefficients from the NDR–MS to the RI scheme (originally called the MOM scheme) that is more useful for non-perturbative calculations of hadronic matrix elements [9].     ΓAk ⊗ ΓBk = σµν (1 ± γ5 ) ⊗ σ µν (1 ± γ5 ) . This calculation serves also as a preparation for the comparison with Ciuchini et al. we perform a direct calculation of the NDR–MS-scheme ADM of 1F = 2 operators. the SSM and SM predictions for K 0 –K¯ 0 and Bd0 –B¯ d0 mixing can have similar precision only if the two-loop ADM for all the 1F = 2 operators is known.

and QSLL As an example. (2. the four-quark operators from Eq. s¯ γ PR d β .1) mix at one loop into the evanescent operators listed in Appendix A.J. LR and SLL sectors. Moreover. Appendix C contains additional evanescent operators that become important only when one wants to derive the 1F = 2 results from the 1F = 1 ones. QVLL 1  β µ    α α = s¯ α PL d α s¯β PL d β .11]. Buras et al. Finally.. they have to be treated in dimensional regularization in the same manner as the identity γµ γν γρ = gµν γρ + gνρ γµ − gµρ γν + iαµνρ γ α γ5 .4) Q2 = s¯ σµν PL d s¯ σ PL d and QSLL are generated.and two-loop ADMs for 1F = 1 operators with Dirac structures (1. In dimensional regularization.e.3). an analogous list for the 1F = 1 case is presented. QSLL QLR 1 = s¯ γµ PL d 1  β    α α = s¯ α σµν PL d α s¯ β σ µν PL d β . They can be split into 5 separate sectors. The operators belonging to the first three sectors (VLL.8. respectively. In Appendix B. the operators   = s¯α PL d β s¯β PL d α . The operators belonging to the two and QSLL by interchanging remaining sectors (VRR and SRR) are obtained from QVLL i 1 PL and PR .1) s¯ PR d β .10. The Fierz identities cannot be analytically continued to D dimensions. Appendix E contains a list of separate contributions from different diagrams to the one. we list the evanescent operators relevant to the 1F = 2 calculation. Appendix D is devoted to recalling and generalizing the notion of “Greek projections”. (2. (2. QSLL QLR 2 = s¯ PL d 2 where σµν = 12 [γµ . we shall consider only the VLL. the ADMs in the VRR and SRR sectors are the same as in the VLL and SLL sectors.2) i. In 4 dimensions these operators can be expressed through QSLL 1 2 by using the Fierz identities . Direct calculation of the ADM in the 1F = 2 case For definiteness. Since QCD preserves chirality. we shall consider here operators responsible for the K 0 –K¯ 0 mixing. appropriate evanescent operators have to be introduced. 2. In the following. (2.and two-loop diagrams. consider the operators QSLL 1 2 . there is no mixing between different sectors. according to the chirality of the quark fields they contain. There are 8 such operators of dimension 6. as in Section 6. Specifying these evanescent operators is necessary to make precise the definition of the NDR–MS scheme in the effective theory [5. When these operators are inserted into one. γν ] and PL. An important novelty in the present case (when compared to 1F = 1 calculations) is the necessity of introducing evanescent operators that vanish in 4 dimensions by the Fierz identities.3) Q˜ SLL 1  β µν  SLL α β α ˜ (2.R = 12 (1 ∓ γ5 ).A. LR and SLL) read   = s¯α γµ PL d α s¯ β γ µ PL d β . / Nuclear Physics B 586 (2000) 397–426 399 In Appendix A. Therefore.

2 8 1 µν (σµν PL )ij (σ PL )kl = −6(PL )il (PL )kj − (σµν PL )il (σ µν PL )kj . 2 4π (4π) (2.11) governed by the ADM γˆ (µ) that has the following perturbative expansion: γˆ (µ) =  αs (µ) (0) αs2 (µ) (1) γˆ + γˆ + O αs3 . these relations have to be generalized to 1 1 = − QSLL + QSLL + E1SLL . / Nuclear Physics B 586 (2000) 397–426 1 1 (PL )ij (PL )kl = (PL )il (PL )kj − (σµν PL )il (σ µν PL )kj .and two-loop counterterms in the effective theory.26)–(4. where Dirac algebra has to be performed in D 6= 4 dimensions.13) − 2bˆ c.9) Q˜ SLL 2 1 2 2 Here.7) Q˜ SLL 2 1 D=4 2 2 These relations can be used in the calculation of one-loop ADM. according to the following relations (equivalent to Eqs. [5]): γˆ (0) = 2aˆ 11.15) . V t d t s 2 4π (2. 2 which give (2. (2. γˆ (1) = 4aˆ 12 (2. (2. In the case of two-loop calculations.37) of Ref. It takes the form Leff = − 2  2 2 X  G2F MW ∗ V Zq Ci (µ) Qi + (counterterms)i . in the NDR–MS scheme. (2.9) and are given in Appendix A. (2. ˆ (2. The coefficients Ci (µ) satisfy the Renormalization Group Equation (RGE) µ d E E C(µ) = γˆ (µ)T C(µ) dµ (2. aˆ 12 and bˆ in the above equations parametrize the MS-scheme counterterms in Eq. (2.400 A.8) and (2. (2.8) Q˜ SLL 1 2 1 8 2 1 = 6 QSLL + QSLL + E2SLL .12) The ADM in the MS or MS scheme is found from one. Buras et al.10) (for D = 4 − 2)     X αs X 11 αs2 X 1 22 1 12 aik Qk + bik Ek + a + aik Qk (counterterms)i = 4π (4π)2  2 ik  k k k  + (two-loop evanescent counterterms) + O αs3 . (4.10) i where Zq is the quark wave-function renormalization constant.J. The effective Lagrangian can be written separately for each sector.14) The matrices aˆ 11 . E1SLL and E2SLL are the evanescent operators that vanish in 4 dimensions by Fierz identities.6) Q˜ SLL 1 1 D=4 2 8 2 1 = 6 QSLL + QSLL .5) 1 1 = − QSLL + QSLL . They are simply defined by (2.

The finite coefficients of these matrix elements define the matrix cˆ as follows:   X X 1 X dkj hEj itree + ekj hFj itree − cki hQi itree + O(). (2. Let us denote by hEk i1loop the one-loop K 0 –K¯ 0 amplitude with an insertion of some evanescent operator Ek .16) hEk i1loop = −  j j i Here. as well as the evanescent operators Ek .18) γˆ −24 − 48 2N + 6 − N2 N . but γˆ (1) does not. In the second method. i. In both of them. We calculate only the “annihilation-type” diagrams. All the one.and two-loop diagrams considered in the present article are calculated using two different methods. We begin with presenting the ADM in the SLL sector. One-loop diagrams. at one loop and (later) at two loops does not alter the final results for the renormalization constants. the external momenta are set to zero. (2.e. we drop all the diagrams where fermion lines connect the incoming and outgoing particles.A.. The remaining part in the limit D → 4 can be expressed by tree-level matrix elements of the physical operators Qi . In the first method. a covariant gauge-fixing term Lgf = − 1 µ a  ν a ∂ Gµ ∂ Gν 2λ (2. / Nuclear Physics B 586 (2000) 397–426 401 Fig. The ADMs calculated from these renormalization constants with the help of Eqs. 1 with insertions of the physical operators Qi . [4]. because in this very sector our results are going to differ (at two loops) from those of Ref.17) is used. The matrix cˆ is recovered from one-loop matrix elements of the evanescent operators. Dropping such diagrams consistently at the tree level. and the physical masses are set to zero. (2. Fj stand for such evanescent operators that are not necessary as counterterms for the one-loop Green functions with insertions of the physical operators Qi . but a common mass parameter is introduced in all the propagator denominators as IR regulator [12]. The matrices γˆ (0) = 2aˆ 11 . bˆ and cˆ in each sector are found from the one-loop d s¯ → s d¯ diagrams presented in Fig. Buras et al.J. the external quarks are assumed to have momentum ±p. The matrices cˆ and aˆ 12 depend on the structure of Fj . 1.14) are independent of the gauge-fixing parameter λ. The pole part of such an amplitude is proportional to some linear combination of tree-level matrix elements of evanescent operators. The two methods give the same results for the MS renormalization constants.13) and (2. The matrices γˆ (0)SLL and bˆ SLL are found to be the following: ! 1 1 −6N + 6 + N6 2 − N (0)SLL = .

Buras et al. Its entries are found to be the following: 136 12 107 2 10 203 2 107 10 N + N+ − − f. N 22 39 57 2 19 2 + − f. our results for the LR sector read ! 6 12 N (0)LR = . / Nuclear Physics B 586 (2000) 397–426 0 bˆ SLL = 1 2 0 1 −8 −8 − 2N 0 ! 1 2 . γˆ 0 −6N + N6 γˆ (1)LR 137 15 22 6 + 2N 2 − 3N f 71 9 4 N + N − 2f = bˆ LR =  200 6 44 3 N− N − 3 f 479 15 10 2 − 203 6 N + 6 + 2N 2 + 3 Nf 1 1 0 −5 − 2N 0 0 2 1 1 0 0 0 0 − 2N 2 (2. Proceeding analogously in the VLL sector. In order to find the matrix aˆ 12 . we need to calculate two-loop diagrams obtained from the ones in Fig. =− N − γ12 36 9 N N 18 9N 704 208 320 136 176 364 (1)SLL N− − − 2 + f+ f.23) ! − 22 3N f .24)  .3) in the four-quark vertices.21) The matrix bˆ in the VLL sector reads  1 1 bˆ VLL = −5 − 2N 2 .J. (2.14). All the two-loop diagrams and the corresponding colour factors are the same as in Fig. (2. we reproduce the well-known results for the [13]: one.19) where N stands for the number of colours. + Nf − f − 2 6 3 3 N 3 3 3N 2N 31 9 4 1 1 1 (1)SLL + − 2− f+ f.22) Finally. 2 and Table 2 of Ref.25) . Of course. =− γ21 3 3 N N 3 3N 21 2 343 2 188 44 26 (1)SLL N + 21N − + + f. (2. (2. we also consider additional Dirac structures (1. Inserting the calculated matrix aˆ 12 into Eq. we obtain the two-loop ADM. in the present article. too.402 A. (2. + f− γ (1)VLL = − N − 6 3 N 3N 2N 2 3 γ (0)VLL = 6 − (2. [5]. The above equation is one of the main results of the present paper. (2.and two-loop anomalous dimensions of the operator QVLL 1 (1)SLL γ11 =− 6 .20) γ22 = − Nf − 6f + 18 9 N 9 9N 2N 2 where f stands for the number of active flavours. 1 by including one-loop corrections on the gluon lines or adding another gluon that couples to the open quark lines. However. one-loop diagrams with counterterm insertions need to be included.

The evanescent operators for the VLL. the only possible diagrams are the current–current ones. QSLL 3   = s¯α σµν PL d α u¯ β σ µν PL cβ = QTL TL .A. SRL and SRR sectors will have exactly the same properties under QCD renormalization.and two-loop diagrams in the cases of VLL. VRL. In such a case. QSLR 2   = s¯α PL d β u¯ β PL cα = Q˜ LL . QVLR 1   = s¯α γµ PL d α u¯ β γ µ PR cβ = QVL VR .h. VLR. (3. it is sufficient to calculate the ADMs only for the VLL. QSLL 1   = s¯α PL d α u¯ β PL cβ = QLL . Calculation of the renormalization constants and the ADMs proceeds along the same lines as in the previous section. SLR and SLL sectors. QSLL 2   = s¯α σµν PL d β u¯ β σ µν PL cα = Q˜ TL TL . u.J. we have shown the notation of Ref. QVLR 2   = s¯ α PL d β u¯ β PR cα = Q˜ LR . The “mirror” operators in the VRR. For this purpose. d. Our final results for the 1F = 1 ADMs are as follows: ! 6 − N6 (0)VLL = . Obviously. [5] and [6].1) QSLL 4 where on the r. / Nuclear Physics B 586 (2000) 397–426 403 As mentioned in the introduction. For completeness we give in Appendix E the corresponding results for the SLL sector. all the comparisons with existing literature are relegated to Section 7. Current–current contributions to the ADM of 1F = 1 operators In the present section. VLR. The operators belonging to the four remaining sectors (VRR. QSLR 1   = s¯ α PL d α u¯ β PR cβ = QLR . These have not been published so far in the NDR–MS scheme. VLR and SLR sectors are given in Refs. SLR and SLL sectors are listed in Appendix B. VRL. between which there is no mixing. QVLL 2   = s¯α γµ PL d β u¯ β γ µ PR cα = Q˜ VL VR . VLR. Twenty linearly independent operators can be built out of four different quark fields.2) γˆ 6 − N6 . SRL and SRR) are obtained from the above by interchanging PL and PR . Buras et al. ¯ c.s. The relevant divergences in one. (3. SLR and SLL) read   = s¯α γµ PL d β u¯ β γ µ PL cα = Q˜ VL VL . The operators belonging to the first four sectors (VLL. we evaluate contributions from the current–current diagrams to the ADM of 1F = 1 operators. 3. we choose the operators in such a manner that all the four flavours they contain are different: s¯ . [4]. They can be split into 8 separate sectors. QVLL 1   = s¯α γµ PL d α u¯ β γ µ PL cβ = QVL VL .

137 15 22 6 + 2N 2 − 3N f 71 18 − 2 N − N + 4f γˆ (1)SLR = − 22 3 − ! −6 6 N − 19 6 N+ . (3. =− − 2 + γ24 18 N 9N 176 676 2 1880 320 88 (1)SLL N − − 2 − Nf + f.J. Buras et al. = γ31 3 3 N 3 3N 448 88 820 (1)SLL N+ − f. − 2Nf − 2 2 3 2N 3N 64 16 178 (1)SLL N+ + f. =− + Nf + γ33 18 9 9 9N 2N 2 (1)SLL =− γ11 ! (3. (3.4) −6  0 −6N +  γˆ (0)SLL =  48  − N + 24N 24 48 39 2 N + 3f 57 2 − 3N f 2N 2 .404 A.8) . = γ32 3 N 3 21 2 257 2 116 22 (1)SLL N − + f. / Nuclear Physics B 586 (2000) 397–426 γˆ (1)VLL γˆ (0)VLR − 22 3 − = − 19 6 N −6N + = γˆ 6 N −6 0 ! 479 15 6 + 2N 2 3 − 100 3 N+ N = 0 −6N + 6 N  6 N + + 10 3 Nf 22 3 f − .6) − 100 3 N+ 2 − 203 6 N + − 48 N N 2 6 N − 479 6 1 N + 3 N 15 2N 2 1 − N1 − N2 − 4N 6 0 2N − + + 22 3 f 10 3 Nf ! − 22 3N f . 6 N 2 − 203 6 N + γˆ (1)VLR = (0)SLR 57 2 − 3N f 2N 2 39 2 + N + 3f   . =− γ12 3 N 3 4 f 107 2 71 1 (1)SLL N − − . γ23 18 N 9 4 53 1 (1)SLL f. (3.5) 22 3N f (3. = − Nf + γ13 36 18 N 2 18 9N 8 f 109 (1)SLL N+ − .3) − 71 2 N− 22 3N f 137 6 + 15 2N 2 18 N + 4f − ! . =− + Nf − γ22 6 3 3 3N 2N 2 2 1 89 (1)SLL = N + − f. = −26N + γ21 N 107 10 203 2 28 10 (1)SLL N + − f.  2 N (3. =− γ14 36 N 18 104 (1)SLL .7)  1 2 107 N 2 148 10 + − f.

VLR and SLR sectors are preserved at two loops in the NDR–MS scheme only by coincidence.  2  0 6 0 12  6 0 1 2 (3. the one-loop matrix γˆ (0)SLL must satisfy the following identity (cf.9) = − Nf + γ44 2 18 9 2N 9 9N Equation (3. [4]): b bγˆ (0)SLLF γˆ (0)SLL = F with  0 − 12 0 1 8   −1 0 1 0   8 b=  F . Eqs. the relations (3. γ11 VLL VLL γ12 = γ21 . (3.14) 0 No similar relation holds for γˆ (1)SLL in the NDR–MS scheme.12) as well as the flavour.and chirality-blind character of QCD interactions. A detailed discussion of this fact can be found in Ref. this question is related to the properties of one-loop matrix elements of certain evanescent operators. 1 On the contrary.10) could be potentially broken at two loops in the NDR–MS scheme. this is not surprising. = γ41 3 N 3 776 320 176 (1)SLL − 2 + f.10) for the VLL. Since the Fierz identities are satisfied in four spacetime dimensions only. It is unclear to us whether the symmetries (3. The careful reader has already noticed that the following equalities hold up to O(αs2 ): (1)SLL γ34 = VLL VLL = γ22 .5). As it has already been said. these equalities are a consequence of the Fierz identities   (γµ PL )ij γ µ PL kl = −(γµ PL )il γ µ PL kj .9) is one of the main results of this work. where the penguin diagrams are considered.13) (3.A. VLR SLR = γ11 . or if there is some reason beyond this. analogous relations are broken at two loops in the SLL sector. =− γ42 3 N 3N 40 8 22 (1)SLL + f. 1 In Section 4.J. Surprisingly.10) (3. Because of the Fierz relations (2. 3 3 416 176 488 (1)SLL N+ − f. VLR SLR γ21 = γ12 .  (γµ PL )ij γ µ PR kl = 2(PR )il (PL )kj . because the Fierz identities are not true in D 6= 4 dimensions. As we shall see in Section 6. γ22 VLR SLR γ12 = γ21 . [6]. (3. they are not. VLR SLR γ11 = γ22 . / Nuclear Physics B 586 (2000) 397–426 405 50 8 N − f. = N− γ43 3 N 3 21 2 343 28 26 (1)SLL N2 + + f. Buras et al. . At one loop.11) (3. no invariance under Fierz rearrangement is observed at two loops for the operators with VLL Dirac structure. (9) and (10) of Ref.

it does not affect their ADM. we shall describe additional contributions to the ADM of 1F = 1 operators that are due to penguin diagrams. They were also obtained in Ref. s¯ γµ PL d β 2 q (4. .    Q12 = s¯α γµ PL d β d¯ β γ µ PR d α + s¯ β γ µ PR s α . [6]. Beyond SM. The factor of 4 can be absorbed into the global normalization factor of the effective Lagrangian.. The next two have the structure of Q5 and Q6 . [6]. for the operators obtained from the ones in Eq. including the Fierz relations (2. Of course. Q3 = s¯α γµ PL d α q Q5 = s¯ γµ PL d α α X q¯ γ PR q β µ β  q .and two-loop ADMs.12). q X  3 eq q¯ β γ µ PR q β .e. In the SM analysis of Ref.1) Their one.10). . It turns out that only 3 additional operators (and their mirror copies) undergo mixing via penguin diagrams into other fourquark operators in Eq. X β µ  X β µ  q¯ γ PL q β . For definiteness. Such contributions may arise only when the operators contain one quark–antiquark pair of the same flavour. These are    Q11 = s¯α γµ PL d α d¯β γ µ PL d β + s¯ β γ µ PL s β . new linearly independent operators appear.1). [6] by a global normalization factor of 4. [6]. The same results hold for the mirror copies of the SM operators. q Q6 = s¯ γµ PL d α β X  q¯ β γ µ PR q α .    (4. of Eq. Q8 = s¯α γµ PL d β 2 q Q9 = Q10 = X  3 α eq q¯ β γ µ PL q β . let us consider 1S = 1 operators.1) by PL ↔ PR interchange. but the sum over flavour-conserving currents 2 Our operators here differ from the ones in Ref. 10 such operators were considered 2     Q1 = s¯α γµ PL uβ u¯ β γ µ PL d α . (4. Q7 = s¯ α γµ PL d α 2 q X  3 eq q¯ β γ µ PR q α .1). (3. . The first three of the remaining operators have the structure of Q11 . Q4 = s¯ α γµ PL d β q¯ γ PL q α .s. The remaining elements of the operator basis can be chosen in such a manner that massless penguin diagrams with their insertions vanish. Penguin contributions to the ADM of 1F = 1 operators In the present section. Our aim is to find a minimal set of linearly independent new operators. (3. we shall use four-dimensional Dirac algebra. but with a relative minus sign between the two terms.2) Q13 = s¯α γµ PL d α d¯β γ µ PR d β + s¯ β γ µ PR s β .5). . In this case. . the Wilson coefficients of our operators are exactly the same as those in Ref. including current–current and penguin diagrams. as the first ratio on the r.406 A. can be found in Appendices A and B of Ref. Their Dirac structures are as in Eq. Q2 = s¯ α γµ PL uα u¯ β γ µ PL d β .J. [7].11) and (3.h. Buras et al. i. (2. / Nuclear Physics B 586 (2000) 397–426 4. (4. Q13 . In the process of identifying these operators. s¯ γµ PL d α 2 q X  3 α eq q¯ β γ µ PL q α .

mix only due to current–current diagrams.J. This means that all the 1S = 1 operators. this complication can be avoided by introducing appropriate evanescent operators that vanish in four dimensions by Fierz identities.L q β . This can be easily verified by using the results of Ref. as well. differs from the one of Q11 at the two-loop level.4) s¯ α PL.6) Q011 = s¯ α γµ PL d β d¯β γ µ PL d α + s¯β γ µ PL s α with the operators in Eq.R = 0. because the Fierz identity (3.1). (4. except for Q1 . their ADMs are identical to the ones we have already calculated in Sections 2 and 3. i. Q011 was not included in the basis (4. . / Nuclear Physics B 586 (2000) 397–426 407 is replaced by a difference between the analogous u-quark and c-quark currents.R d β q¯ β PR. By working with Q11 and the evanescent operator Q011 − Q11 .3).L q α . On the other hand. Massless penguin diagrams with insertions of the operators (1. It is straightforward to convince oneself that we have not missed any linearly independent 1S = 1 operator in the above considerations. Q13 and their mirror copies.A. . The above discussion implies that the only additional ADMs we need to find in the present section are: • The 3×3 matrix γˆcc describing the mixing of Q11 .R d α q¯ β PR. our use of these relations in the identification of linearly independent operators could be put in question.2). At the two-loop level. because Tr(Sodd PL.11) is broken at two loops in the NDR–MS scheme through penguin diagrams. as we have already discussed in Section 2 and will elaborate in Section 5. [6].2) remains unchanged.2). Consequently.3) s¯ α PL. Their mirror copies have to be included. a complication arises because generally the Fierz relations could be broken in D 6= 4 dimensions. . even at the two-loop level.11). The introduction of evanescent operators that vanish in four dimensions by Fierz identities turns out to have no effect on the two-loop ADM in the case of the operators with VLR and SLR structures.12) remains valid at two loops in the NDR–MS scheme. However.   (4. only massless penguin diagrams can cause mixing into other four-quark operators. even if the penguin insertions are considered [6]. . Buras et al. the Fierz identity (3.R ) = 0 and PL. as pointed out in Ref. Q13 among themselves. where q has flavour different from s and d. [6].4) vanish. through penguin diagrams. This allows us to restrict the basis of new physical operators (undergoing penguin mixing) to the one in Eq.5) where Sodd is a product of an odd number of Dirac γ -matrices. Further operators have the SLL and SRR Dirac structures as in Eq. For dimensional reasons. As Q011 = Q11 in D = 4 dimensions due to the Fierz identity (3.R Sodd PL. . so that the basis (4. the explicit appearance of Q011 can be avoided at any number of loops. or they have the form   (4. . As a result. .e.. . (4. (4.3). although it remains valid for current–current diagrams. (1. (4.3) and (4. the mixing of the operator    (4. .

8) γˆp(0) = 43 . However.1) γ R 4 ∂pρ p 2 =−µ2 where µ is the subtraction scale. It takes the form ! VLL γ1F 0 =2 (4. in which F is the up-type flavour. [6].7) γˆcc = VLR 0 γˆ1F =1 VLL and γˆ VLR taken from Eqs. S(p)R = −i/ 4π 2 µ . . . We find T  (4. Q6 via penguin diagrams. Below. all the contributions from penguin diagrams can be easily extracted from Ref. .9)  244 20 188 140 148 4  6N + 3N  −6N − 27 + 3N − 27N 2  27 + 27N 2 172 27 N − 2 3 + 260 27N 220 27 N − 508 27N − 22 3 The above discussion changes very little in the case of 1F = 1 operators. Transformation of the Wilson coefficients to the RI scheme The ADMs calculated in the present work are given in the NDR–MS scheme that is most convenient for perturbative calculations. 1. A simple one-loop calculation is necessary to verify that the renormalized inverse propagator in the RI scheme reads     −p2 αs 1 −1 CF λ − ln 2 p 1− (5.3 of Ref. . . . . (5. Buras et al. we begin with the definition of the RI scheme.408 A. with γ1F =2 1F =1 (0) αs (1) γˆp + · · · that originates from penguin diagrams can be The matrix γˆp = γˆp + 4π easily extracted from Sections 3. (3. 5. 1 . .2) + O αs2 .) The matrix γˆcc is given by current–current diagrams only. (Only Q3 . Similarly to the 1S = 1 case. For completeness.2) and (3. [4]. Q6 are generated by massless QCD penguin diagrams with four-quark operator insertions.   16 172 356 40 − 112 6N − 64 27 − 3N + 27N 2 27 − 27N 2 −6N + 3N   2 460 500 352   − 32 − 22 27 N − 3 − 27N 27 N + 27N 3 . .3).2 and 5. 0 × − N1 .21). / Nuclear Physics B 586 (2000) 397–426 • The 3×4 matrix γˆp describing the mixing of Q11 . For the massless quark propagator. One such scheme is the socalled Regularization-Independent (RI) scheme (originally called the MOM scheme) used in Ref. it might be necessary to transform them to another scheme that is more appropriate for nonperturbative calculations of hadronic matrix elements [9]. Q13 into Q3 .11) down to a low energy scale. − N1 . after the Wilson coefficients are evolved with the help of RGE (2.J. (2. 43 . the renormalization condition can be written as   i ρ ∂ −1 S(p) = 1. [6]. . . we shall give relations between the NDR–MS-renormalized and RIrenormalized Wilson coefficients of all the operators considered in Sections 2 and 3. γˆp(1)T =  (4. .

11) and (3.A. pµ pν σµρ PL ij σ νρ PR kl = 2 p 2 No Bij formfactors occur in the SLL sector thanks to the four-dimensional identity p / PL  ij p / PL  kl  il p / PL  kj   1 = p2 σµν PL ij σ µν PL kl .1) are imposed on renormalized matrix elements of the operators (2. (2. Q2 R = ASLL 21 (p ) Q1 22 (p ) Q2 tree tree (5. Q1 R = ASLL 11 (p ) Q1 12 (p ) Q2 tree tree SLL SLL SLL 2 2 + ASLL .3) ZqRI = 1 − 4π  2 provided the subtraction scale µ is identified with the standard MS renormalization scale. 22 (p ) Q2 tree + B22 (p ) p pν s¯ σ