Dynamical Supersymmetry Breaking in Four Dimensions and Its Phenomenological Implications

You might also like

You are on page 1of 43

Nuclear Physics B256 (1985) 557-599

~': North-Holland Publishing Company

D Y N A M I C A L S U P E R S Y M M E T R Y B R E A K I N G IN F O U R
DIMENSIONS AND ITS PHENOMENOLOGICAL IMPLICATIONS

lan AFFLECK*
DPh T Cen- Saclav, 911 ql -Gif-sur- Yt'ette Cedex, France

Michael DINE and Nathan SEIBER(i


The Institute for Advanced Study. Princeton, New Jersey 08540, USA

Received 28 November 1984


(Revised 25 February 1985)

It is now established that the non-renormalization theorems are violated by non-pcrturbative


effects in many four-dimensional supersymmetric theories, and that dynamical supersymmetry-
breaking (DSB) occurs in some chiral gauge theories. Here we present a set of general techniques
which allow one to determine whether or not supersymmetry is broken in almost any theory, and
permit a quite detailed analysis of the dynamics of many theories. These techniques are illustrated
by analysis of a wide variety of instructive models, including one in which DSB occurs and in
which we compute the spectrum and ground state energy. A careful discussion of the prospects for
realistic model building is presented, including: an argument that additional gauge interactions are
required to build realistic theories; dynamical supersymmetry-breaking in N = 1 supergravity
theories; a detailed discussion of the problems and prospects for dynamical supersymmetry-break-
ing in the multi-TeV energy range.

1. Introduction and conclusions

D y n a m i c a l s u p e r s y m m e t r y - b r e a k i n g (DSB) holds some hope for an elegant ex-


p l a n a t i o n of the gauge hierarchy [1,2]. Thus there has been considerable interest in
the n o n - p e r t u r b a t i v e properties of supersymmetric gauge theories [3-6]. In a series
of papers, we have explored several of these models. In refs. [7, 8], we gave an almost
c o m p l e t e d e s c r i p t i o n of the behavior of supersymmetric Q C D for small or zero
q u a r k mass, for arbitrary n u m b e r of flavors and colors. In refs. [9-11], we studied
three chiral gauge theories and argued that in each s u p e r s y m m e t r y is d y n a m i c a l l y
b r o k e n . A m a z i n g l y , one of these theories could, in fact, be solved in a systematic,
weak c o u p l i n g expansion, and its spectrum and other features exhibited [10]. Related
results have been o b t a i n e d by other authors [12-14].

* On leave from Joseph Henry Laboratories, Princeton University,Princeton, New Jersey 08540.

557
558 I. Affleck et aL / Dvnamwal supersvmmetrv breaking

The purpose of the present paper is to present a set of general techniques which
allow a rather complete analysis of many theories, to illustrate these techniques by
analyzing a variety of instructive models, and to discuss the prospects for building
realistic models based on dynamical supersymmetry-breaking. This section will give
a rather extensive overview of the whole paper; the following sections will go into
greater detail on techniques, analysis of models and realistic model building.
From our perspective, the most important feature of supersymmetric gauge
theories is a purely classical one; they have, in many cases, flat directions, i.e. lines
and surfaces in field space running to infinity along which the classical potential
energy is zero. In a "normal" field theory this degeneracy would be lifted by
perturbative corrections; in a supersymmetric theory this does not happen to any
finite order in perturbation theory [if the gauge group contains no U(1) factor]. A
general proof of this uses the non-renormalization theorem [15] to forbid perturba-
tire corrections to the superpotential and arguments by Witten [4] and Fischler et al.
[16] that D-terms cannot lift a zero-energy state if the gauge group has no U(1)
factor. Actually in most of the theories we have considered we do not need to invoke
the non-renormalization theorems to prove this statement; the symmetries respected
by perturbation theory (which include anomalous U(l)'s) do not allow the genera-
tion of new terms in the superpotential. Perhaps the most important (and most
unpleasant) thing we have learned is that, in most cases flat directions are not lifted
by non-perturbative effects either. Instead one of the following options is realized:
(i) the flat directions remain flat after the inclusion of all non-perturbative effects
and the theory has an exact degeneracy of inequivalent supersymmetric vacuum
states, or
(ii) a non-zero potential is generated which slopes to zero at infinity. We present
strong arguments that, in this case, the theory has n o s t a b l e v a c u u m s t a t e .
This result, which we present in sect. 2, restricts our search for theories with stable
vacuum states with dynamically broken supersymmetry to the rather small class of
chiral theories that do not have classical flat directions. The problem of finding or
proving the non-existence of flat directions is, in general, a rather complicated
algebraic exercise. We review some useful techniques in sect. 2.
In refs. [7, 8] we analyzed non-perturbative effects in a class of theories with flat
directions: supersymmetric QCD with arbitrary numbers of colors and massless
flavors. In the next sections we generalize the analysis to a variety of other theories.
Our basic approach is to study the effective lagrangian, Leff, at large field strength
along a flat direction. Our results are most rigorous when there exist flat directions
in which the scalar expectation values completely break the gauge group. Then, in
perturbation theory there is a continuum of inequivalent vacuum states in the
"Higgs phase" with the scale of gauge symmetry breaking ranging from zero to
infinity. (There are not really separate Higgs and confining phases in these theories;
a complementary description in terms of composite gauge invariant operators exists.)
Since the effective gauge coupling is small at large field strength we are able to
L Affleck et al. / Dynamical supers.vmrnetry breaking 559

control the non-perturbative effects and calculate the effective lagrangian to arbi-
trary precision. The only assumption we make is that supersymmetry is not anoma-
lous so that the effective lagrangian is supersymmetric. It then follows that from
Witten's arguments [1] that only the generation of an effective superpotential can lift
the flat directions. In some theories we find that no superpotential is allowed by the
(non-anomalous) symmetries of the theory; in other cases we find a unique super-
potential is allowed. In the latter case we can determine the precise coefficient of this
term in the effective lagrangian using instanton methods. In every case, we find the
term is generated at the one-instanton level. We extract the coefficient by calculating
fermion masses and the expectation values of F-operators. These arise from a single
instanton. It is more difficult to calculate the effective scalar potential directly
because it is an instanton-anti-instanton effect. However, the assumption that
supersymmetry is not anomalous determines the scalar potential uniquely. Lef~ can
also contain D-terms generated by loop diagrams, but, due to asymptotic freedom,
all such effects become negligible at large field strength. Thus we may calculate the
effective potential using the tree-level lagrangian plus the instanton generated
superpotential, to arbitrarily high precision for sufficiently large field strength, along
the classical flat direction.
We c a n n o t calculate Leff reliably at small field strength where strong interaction
dynamics become important. However we can often deduce relevant information by
using the Witten index, T r ( - 1 ) F [3].
In some theories, all flat directions leave unbroken, non-abelian gauge groups.
Then one has, at large field strength, some heavy gauge fields and some massless,
Goldstone-like objects weakly coupled to a strongly-interacting gauge theory. We
cannot proceed as rigorously in this case, but the decoupling of the strongly coupled
sector from the other fields allows us to use any known (or assumed) information
about this sector to obtain information about the full theory.
Theories of this type contain two scales. In addition to the scalar v.e.v.'s, ep, there
is now also the scale of the low-energy gauge group, Az.(cp). It is then natural to
work with an intermediate scale lagrangian valid at scales, M, AL(Cp)<< M << (p.
This lagrangian will be supersymmetric. It will contain in addition to renormalizable
terms, higher dimension operators suppressed by powers of 1/cp; these can play an
important role in determining the behavior of the full theory. Two cases must now
be distinguished, depending on whether or not the strong gauge sector breaks
supersymmetry. If it does, the low-energy effective lagrangian LLE, at scales
M << AL(Cp), need not be supersymmetric (i.e. supersymmetry will not be realized
linearly). LLE will describe the Goldstone-like objects associated with the flat
directions and a goldstino (and any massless objects connected with symmetries of
the strong sector). If the strong gauge sector does not break supersymmetry, L LF:
will be manifestly supersymmetric. It may or may not exhibit spontaneous super-
symmetry breaking. (This latter situation already arose in supersymmetric QCD for
N/< N,- 1 [8]).
560 I. Affleck et al. / l)vnamwal supersymmetrv breal~ing

This sort of reasoning leads to a rather complete picture of the behavior of most
theories with flat directions. We cannot, in general, proceed as rigorously in a theory
that does not have flat directions because the vacuum can be strongly interacting.
However, there is one class of theories without flat directions which we can analyze
rigorously in a particular limit. These are theories which possess flat directions when
some parameter (a term in the superpotential or a weak gauge coupling) goes to zero.
Call this parameter k. For small values of X, these theories almost have fiat
directions. In cases where instantons generate a superpotential along a flat direction
of the X = 0 theory, we typically find that for small, non-zero ~,, Vcrf has a minimum
near this fiat direction which goes to infinity as k -~ 0. In some cases this vacuum
state is supersymmetric, but in others it has spontaneously broken supersymmetry.
When the parameter is very small we can compute all properties of this theory
(vacuum energy, spectrum, couplings, etc.) in a systematic weak coupling expansion.
We analyze the simplest model of this type in which supersymmetry is broken in
sect. 4. The only uncertainty in this calculation is the shape of Vcrf at small field
strength, far from the minimum. We argue, based on the Witten index, that Vcrr
probably does have a zero near the origin. However, even if there were such a zero,
the vacuum we explore becomes more and more stable as ~ ~ 0 for three reasons: its
energy vanishes in this limit; its location moves to infinity; the barrier height, which
should be determined by the strong interaction dynamics, tends to a finite limit. We
regard such models as a proof by construction that dynamical supersymmetry-break-
ing occurs in four-dimensional gauge theories. If ~ is made larger, we expect
supersymmetry to remain broken, by continuity of the vacuum energy, although we
eventually lose our ability to solve the theory. (A phase transition to a supersymmet-
ric state is, of course, a logical possibility, but we see no indication of it.) Although
theories of this kind have asymptotically free gauge couplings, the examples we have
found have non-asymptotically free Yukawa couplings. Since we choose these to be
small at the scales relevant to the vacuum state (i.e. the scalar expectation values) we
expect that any effects due to some further unification of the theory at a much
higher energy scale (which would probably he necessary) will be negligible.
We also analyze theories that do not have even approximate flat directions. In
such cases we must resort to less direct arguments, and our conclusions, however
reasonable, cannot (currently) be proven. However, the following general result is
very useful [9]:
If any continuous global symmetry is spontaneously broken in a supersymmetric
theory (in four dimensions) then either
(a) there is an exact non-compact vacuum degeneracy (flat directions), or
(b) there is no vacuum state, or
(c) there is a vacuum state with spontaneously broken supersymmetry.
Since options (a) and (b) are not reasonable in a theory which possesses no fiat
directions classically, we conclude that in such theories if any chiral symmetry is
spontaneously broken then so is supersymmetry. We consider this result to be on a
I. Affleck et al. / Dvnamtcal super,~ymmet O. breaking 561

very firm footing; the weak point in using it to prove supersymmetry breaking is that
we can never prove conclusively (in strongly interacting theories) that chiral symme-
tries are spontaneously broken. However, we can make it plausible in some cases
based on the difficulty of satisfying 't Hooft anomaly conditions [17]; it may also
follow from the arguments of refs. [12-14, 18].
The third section of this paper applies our techniques to a variety of models. In
each case, we first analyze the classical potential for flat directions. If they exist, we
analyze the effective potential using the methods described above and determine
which of options (i) and (ii), above, is realized, if they do not exist we analyze the
model by going to a limit where there are almost flat directions, if possible, or by
using the above result on symmetry breaking, if all else fails. The upshot is that, in
most cases there are flat directions and options (i) or (ii) are realized. Among
theories without flat directions we find that all the theories with real fermion
representations have unbroken supersymmetry (consistent with Witten's arguments
[3]) whereas chiral theories without flat directions almost always break supersymme-
try if they possess global symmetries. We first consider minimal N = 2 supersymme-
try and show that the classical vacuum degeneracy is an exact property of the full
theory. We then consider a generalization of QCD containing a singlet field in which
some flat directions are lifted at infinity by non-perturbative effects. After this, we
consider an S U ( N ) gauge theory with an antisymmetric tensor and ( N - 4 ) ~ V ' s
(these theories have also been considered in ref. [13]). We conclude that supersym-
merry is probably broken for odd N (and we estimate the relevant scales), while
there is no ground state for even N. We then consider an O ( N ) gauge theory, with a
vector superfield and argue that the discrete Z~2x -4~ chiral symmetry of the massive
theory is probably spontaneously broken to Z 2, giving ( N - 2 ) supersymmetric
vacua. This is one case where our results differ from those expected from computa-
tion of the index in Born-Oppenheimer approximation [3]. To support our result, we
review a simple argument (due to Witten) that the discrete symmetry of these
theories is spontaneously broken at infinite volume [19]. Finally, we discuss SU( N )
with a symmetric tensor and ( N + 4) anti-fundamentals.
Sect. 4 contains the analysis mentioned earlier of the simplest theory with
calculable supersymmetry breaking.
In the final two sections of this paper we turn to realistic model building. We first
consider whether one could construct a realistic grand unified model with dynamical
supersymmetry-breaking without additional ("supercolor") gauge interactions. On
quite general grounds, we argue that this is not possible. In such theories, one can
integrate out heavy degrees of freedom to obtain a low-energy effective lagrangian,
Letr. This process of integrating out has various effects. From tree graphs containing
heavy fields, small, non-renormalizable terms in the superpotential are generated.
Also, very small instantons may violate non-renormalization theorems and induce
additional, small, renormalizable terms in the superpotential of Letr (e.g. Higgs
masses). Suppose, first, that the low-energy theory consists only of quarks, leptons
562 I. Affleck et al. ,/ Dynamical .super,wmmetrv breaking

and Higgs doublets. We show quite generally that in this case the effects of small
instantons are too small to be interesting. Then, if one ignores possible non-renor-
realizable terms, the low-energy theory possesses flat directions. Any potential which
is generated along these directions has scale, at most, A QCl), and falls to zero at
infinity. Even if the non-renormalizable operators raise these directions, the scale of
any supersymmetry breaking will be much less than a GeV. We illustrate these
considerations with two examples based on SU(5). First, in a theory with a 24 and a
5 and 5 of Higgs, we show that instanton effects of SU(2) L plus non-renormalizable
operators, do generate a hierarchy, but that supersymmetry is unbroken. Then we
point out that a minimal SU(5) theory with three generations of quarks and leptons
has flat directions. While instantons generate a potential in these directions, it falls
rapidly to zero at infinity.
We then consider the possibility that the low-energy theory contains additional
fields, such as gauge-singlet chiral fields or additional U(I) gauge interactions. We
argue that realistic supersymmetry breaking in such theories would require small
( - M ~ T ) instantons to induce either a U(I) D-term or a superpotential linear in a
chiral field. (It would, of course, be necessary that such terms lift any flat directions.)
We review an argument due to Witten that the former is impossible and prove that
the latter is impossible as well.
We next deviate briefly to consider non-perturbative effects in models with
tree-level O'Raifeartaigh breaking. These models often appear unstable against
dynamical supersymmetry restoration but we find this does not happen.
The supercolor theories which have been proposed to date [1,2] have all involved
matter fields in real representations and thus do not break supersymmetry. We are
thus led to consider the viability of chiral supercolor theories. We first note that
they appear to be promising candidates for the hidden sector of N = 1 supergravity
theories [20, 21]. The intermediate scale ( - 10 l° GeV) required in these models is
then generated dynamically rather than put in by hand. Moreover, as we will see,
even if one allows the most general non-renormalizable couplings at the Planck
mass, soft supersymmetry breaking in the low-energy lagrangian is described by a
single parameter. This is in contrast to the usual case [20, 21], where every soft-break-
ing mass and coupling is an independent parameter. (Even when approximate
symmetries are postulated, there are, in general, many relevant parameters.) This
simplification arises because the ground state occurs at scales much less than the
Planck mass, reducing much of the uncertainty due to quantum gravity effects, and
because gauge invariance significantly restricts the allowed couplings. In addition,
these theories have the feature that CP violation is comfortably small [22], and that
certain cosmological problems associated with very weakly-interacting hidden sec-
tors are avoided [231.
We finally consider the possibility that the supercolor sector could communicate
with the observable world at scales not much larger than M w. Due to the chiral
nature of our supercolor models, this communication can only occur via gauge
I. Affleck et al. / Dvnamwal super,*3'mmetrv hreaking 563

interactions; no renormalizable superpotentials coupling the two sectors exist. A


variety of problems then appear:
(i) 't Hooft anomaly conditions [17], in many cases, require massless, electrically
charged fermions in the supercolor sector. It is often difficult to break the symme-
tries which keep these fields massless.
(ii) Ignoring SU(3)c x SU(2)I- anomalies, none of our models explicitly breaks
R-invariance, so this must be done with additional gauge interactions (R-color).
(iii) In the simplest models, the supercolor sector does not couple to the ordinary
strong interactions, so the gluino is extremely light (of order 10 MeV). If some
supercolored fields do carry ordinary color, QCD becomes strong within two or
three decades of the supercolor scale. This requires early unification, presumably in
a theory with an exact baryon number symmetry. An approach in which B - L is
gauged is suggested which may avoid this problem.
(iv) If SU(2) L x U(1) is broken by the supercolor interactions, scalar quarks and
leptons are generally too light. Fortunately, the R-color interactions introduced to
break R-invariance can induce Higgs expectation values. It is. of course, necessary
that radiative corrections give positive mass-squared to scalar quarks and leptons.
These must be calculated on a case by case basis.
These difficulties can be traced to the fact that the supercolor theory is chiral and
the supercoior group is much larger than the global symmetries which can be weakly
gauged. It is possible that some of the supercolor models we discuss can be made
realistic; considerably more work is needed to determine if this is so. It is also
possible that more promising theories with DSB may be found to act as the
supercolor sector.
In conclusion, DSB breaking occurs in many chiral gauge theories. While these
models could make excellent hidden sectors in N = 1 supergravity, the prospects for
dynamical supersymmetry-breaking at accessible energies appear less favorable.

2. Techniques and general results


The classical scalar potential of a supersymmetric gauge theory is given by

V(qg) -- 1--~- ~ D"2 + }"~IF, 12 '


2g 2

D" = q0*/T/"~q0~,

aw( /
F,- 0~, (2.1t

where T f k are the gauge group generators (a labels the generator) in the representa-
tion (in general reducible) of the matter fields. W(~) is the superpotential.
564 I. Affleck et al. / Dynamical super~wmnetrv hrcaktng

Often, the classical potential vanishes at many points in field space and the
classical theory has several inequivalent vacua. Because of the non-renormalization
theorems, such an "accidental degeneracy" is not removed in perturbation theory.
Here, we will describe some techniques for finding these flat directions. Later, we
will discuss the circumstances in which this degeneracy is lifted by non-perturbative
effects.
A necessary condition for the potential to vanish is that D " = 0 for all a. ~ = 0 is
obviously a solution. Other solutions can be found by various methods [11.24, 25].
Here we describe the techniques of refs. [11,25]. For S U ( N ) groups, define the
N × N matrix

D,'=, (A,), %.
"~1 t /'
(2.2)

where (Atj)~ are the real generators of G L ( N ) (the pairs of indices i, j = 1 . . . . . N


label the generators). For q~ in the fundamental representation, ( A'~ )~ = 8j8, ~. For a
general representation ¢p<a
ab with upper and lower indices covariant and contra-
variant,

¢PtA'seP = ~ , .a
, . . . .ePsh
. 1 +mpD, <,1 + ' ' "
qz,,j

~,d ~f.,#> ,i i.'h


-<,h ,a --%h qJ,/ - .... (2.3)

It is easy to see that D ' = X",q); where X" are the generators of S U ( N ) in the
fundamental representation. The DU's thus all vanish if and only if D,' is propor-
tional to the unit matrix, D / = aS/.
As an example consider an SU(3} gauge theory with two massless flavors [6, g].
The scalar quarks are Q,s and ~,l. i = 1,2.3 is the color index and f = 1,2 is the
flavor index. We have to solve

D,' = Q*'tQ , t - Q'S Q] t = ~8,'. (2.4)

Q * , t Q , / i s a 3 × 3 hermitian positive semi-definite matrix of rank two. Diagonalizing


it by an SU(3) transformation we find at least one zero eigenvalue. Eq. (2.4) implies
that in this basis ,¢ ~ / t is also diagonal. Since ~,lt3t
~,/t3* ~ z d has at least one zero
eigenvalue and the other two eigenvalues are non-negative, a must vanish. Thus up
to an SU(3) transformation

Q'"O_,l = 0"0',, = (2.5)

Here, v ~ , v 2 are real. The most general solution of eq. (2.5), up to symmetry
I. Affleck et aL / Dvnamwal super.s3"mmetO" breaking 565

transformations, is easily shown to be [6, 8]

_ , _ [I)1.8)/, j= 1,2
(2.6)
Q'J=QJ/-~ 0, j=3

Thus the scalar potential vanishes along certain directions in field space (labeled by
v/); these directions are referred to as flat directions.
Several general tricks have proven to be useful in solving D~ = aS/. As in the
previous example, it is convenient to write D = ~,D,, where D, denotes the contribu-
tions of sets of fields, { ~o,} with identical transformation properties. One of the D,'s
(say, D~) can be diagonalized by a gauge transformation. If D = D~ + D 2, then both
D~ and D 2 must be diagonal in the same basis. The specific color representation of
the fields in D 1 may imply relations among its eigenvalues. In the example above,
one eigenvalue vanished. If D~ is generated by an antisymmetric tensor, its non-zero
eigenvalues are equal in pairs. Thus for S U ( N ) for odd N, D~ has at least one zero
eigenvalue. Methods such as this (as well as other tricks) will be used extensively in
sects. 3 and 4 when we analyze specific models. The reader is invited to develop
more efficient ways to find flat directions.
In one set of theories, it is particularly easy to find flat directions. These are
theories with matter in real (possibly reducible) representations. In particular, the
D ' s vanish for ¢p = ¢p*. The previous cases and S U ( N ) gauge theories with matter in
the adjoint representation are two simple cases of this rule.

2.1. L O W - E N E R G Y L A G R A N G I A N S A N D F L A T D I R E C T I O N S

A convenient way of looking for fiat directions in a theory with massive and
massless (or light) sectors is to study the low-energy theory. Clearly, if the light
theory has no flat directions, then the full theory has none connected to the vacuum
state under consideration. However, it may be necessary to keep higher dimension
operators in Lcfr to correctly analyze the low-energy theory. As a simple example
consider a Wess-Zumino model with

W = I ~M%2 + Xqv0~0~ . (2.7)

The low-energy theory consists only of ~ot and has no superpotential if only
renormalizable operators are retained in Lef t. Thus the theory appears to have flat
directions although the only solution of

OW
3% M % + Xcp~ O,

OW
3tp~ = 2MP°cPn = O, (2.8)
566 I. Affleck et al. / Dvnanu¢'al super,s3'mmetrr breaking

Fig. 1. Tree graphs which generate the higher-dimension operator of eq. (2.9). The cross in this
supergraph denotes a mass insertion.

is

cP0 = qPl = 0 .
The resolution of this paradox lies in the inclusion of a higher dimension term in the
low-energy superpotential
4
wolf (2.9)
This operator arises from integrating out cp~ at tree level. This can be done by
solving the equation aW/Oepo=0 to find e p 0 ( ~ ) = -Xo#~/M and substituting this
into W. To see that that is correct consider integrating out qo0 after eliminating the
auxiliary field F0 but before eliminating F~. Then the F~-independent part of
qv0(qo~, FI) is determined by OW/O~o= 0. The F~-dependent part leads to terms in
Lefr of quadratic or higher order in F l and so corresponds to f d ~ - t y p e terms. These
D-terms cannot remove flat directions. (We note in passing that a superpotential can
only be generated by tree graphs; loops of heavy fields can only generate D-terms by
the non-renormalization theorems.) The inclusion of this higher-dimension operator
lifts the flat direction of the low-energy theory, reflecting the fact that the full theory
had no flat directions. The procedure above is equivalent to evaluating the graph in
fig. 1.

2.2. D Y N A M I C S IN FLAT D I R E C T I O N S

It is possible to determine the behavior of almost any theory with flat directions,
for sufficiently large values of fields. The key is to study the low-energy effective
theory at a given point in the flat direction. We start by considering the previous
example, SU(3) with two flavors. We first analyze this theory at tree level; later we
will consider the quantum corrections. For non-zero u/, the expectation values of eq.
(2.6) completely break the gauge symmetry. Eight of the matter fields are "eaten":
the remaining four chiral fields are massless. Of the eight real scalar components of
these fields, six are Goldstone bosons of broken global symmetries, while two
correspond to the parameters u/ of the flat directions. Since the gauge symmetry is
broken by fields in the fundamental representation, the massless fields can be given
a simple gauge-invariant description [26]. The four light fields are 4)/, = O'/Q,r,.
I. Affleck et al. / Dynamical supersvmmetty breaking 567

Expanding about the expectation values of eq. (2.6), Q = (Q) + 6Q, we find

4~//, = Iv/128~, + v f S Q / f + of 6 0 / f + 0 ( 8 Q 2 ) .

Since • is gauge invariant, these superfields are orthogonal to the "eaten" fields.
Infinitesimally, they parameterize the solutions to eq. (2.4), and hence indeed
represent motions along the flat directions. The global symmetry breaking pattern is
easily found by examining the expectation value of 4; in particular, it is not
necessary to consider linear combinations of gauge and global symmetries in this
formulation. Note that the global symmetry is realized linearly in terms of ~, and
hence the low-energy lagrangian (in terms of 4 ) is a linear, (rather than a
non-linear) o-model.
This last remark is not generally true. It may happen that the global symmetry is
realized non-linearly. The light fields are still described by gauge-invariant monomi-
dis, but they are subject to constraints. An example is S U ( N ) gauge theory with
N/> N flavors. The Nfl components of ~//, = QfQ'/, are not all independent.
At the classical level, the low-energy effective lagrangian contains a D-term only:
the projection of (Q*'/Q,! + Q t,/Q_'/) D on the massless particles. This expression can
be written in terms of 4. Multiplying (2.4) by ~tg~jg' and using the definition of 4,
( ~ t ~ ) ~ ' = ~_,/~*g~_Js'~,~7 ' in the flat direction. One can obtain a similar relation for
Qt'/Q,,~. Thus, in the flat direction, we have

(Q~Q + 0"t0)flat dir. ~-" 2 Tr(~t¢~) 1/2. (2.11)

Perturbative corrections will generate additional D-terms. However, for large vf, the
theory is weakly coupled and these corrections are small. Moreover, they play no
role in lifting the vacuum degeneracy. Non-perturbative effects may generate a
superpotential, War. Such a superpotential, written in terms of 4, must be invariant
under all of the symmetries of the theory [6, 8], whether or not they are broken in a
given vacuum. This requirement determines the form of Weft uniquely [6, 8]:

bA 7 A7
Wee' = d e t ~ - b det (OQ)// . (2.12)
ff'

Here b is a numerical constant and A is the scale of the theory. A detailed instanton
computation [7, 8] shows that this term is indeed generated.
We now turn to a more general discussion of non-perturbative effects that may
violate the non-renormalization theorems and lift the vacuum degeneracy. We
consider, first, theories with no tree-level superpotential. Let the matter field ¢p,
transform according to the irreducible representation r, with the index /~,. Define G
to be the index of the adjoint representation and /z = ~,~t,. Consider the low-energy
568 I. Affleck et al. / Dynamical super.sTmmetry breaking

effective lagrangian along the flat directions. We must distinguish between two
different cases:
(i) The gauge symmetry is completely broken or is broken to an abelian subgroup.
(ii) There is a non-abelian, unbroken gauge interaction.
In the first case, the low-energy theory is (at least) approximately supersymmetric
and should be described by a supersymmetric lagrangian. Any lifting of the classical
flat directions will be described by a dynamically generated superpotential, as in the
previous example. Such a superpotential must be invariant under all the non-anoma-
lous symmetries. If no tree-level superpotential is present, the form of the dynami-
cally generated superpotential is quite restricted. In order to be invariant under all
ordinary (non-R) U(1) symmetries, it must be a function of

a = I-[~,~, . (2.13)
I

Here, the notation means that each irreducible representation appears/~, times. All
color and any flavor indices in this expression must be contracted so as to form a
singlet. It may happen that such a contraction does not exist. Then, no invariant
F-term exists and no superpotential can be generated. The R-invariance completely
determines the functional form (the R-charge of ,~ is G - ~):

W e f t = b A (3(;-u)/(~; ~')A - 2/1(; ~) (2.14)

Here the power of A, the dynamically generated scale, is determined by dimensional


analysis and b is a numerical coefficient. This superpotential may be generated by
instanton only if there are two more gaugino zero modes than matter field zero
modes [7, 8], i.e. G - ~ = 2. For that case

Weft = bA ~('- ':~'A * = b~ ;(; ':~'e s":/.~:(~')A i (2.15)

where g ( ~ ) is the gauge coupling at the scale ~ and the one-loop beta function has
been used. We see that the power of A that was determined by the R-invariance and
dimensional considerations agrees with the power obtained from an instanton
computation, i.e. the power in e s,,:/.g:~,). This coincidence was first observed in
supersymmetric Q C D [7, 8]. We see here that it is quite general.
When a tree-level superpotential is present, the symmetry of the model is smaller
and the form of any dynamically generated terms in the superpotential (W,~,.,) is not
so restricted. Moreover, the condition G - / z = 2 for an instanton-generated W~t)n
becomes G - ~ ~< 2. The gauge interactions " t i e " G - 2 of the gaugino zero modes
with G - 2 of the matter field zero modes as before. More matter field zero modes
may be " t i e d " by the tree-level superpotential, and hence Way n may be generated if
G - / z ~< 2. If instantons do generate a superpotential, it should behav,, like

Way" _ q~3e- s , : / . F ( , ) _ q~3 ~ , (2,16)


I. Affleck et al. / Dynamical super.~3'mmetO, breaking 569

(where ~ denotes a generic field and we have assumed that the tree-level superpoten-
tial contains only dimensionless couplings). Correspondingly, the additional terms in
the effective potential behave (in the flat directions) as

V- -- . (2.17)

When there is an unbroken non-abelian gauge interaction, non-perturbative


effects other than instantons may violate the non-renormalization theorems*. In this
case it is useful to consider first an effective theory at a scale M much less than q~
but much greater than A L, the scale of the unbroken group. At this scale all
couplings are weak, and the effective lagrangian may be obtained by integrating out
high-momentum modes in perturbation theory. It includes the kinetic terms for the
unbroken gauge interactions,

1 W"W. (2.18)
f d2° t2.( '

and kinetic terms for the various massless chiral fields. It may contain a renormaliz-
able superpotential for some of the light fields, if the original lagrangian contained a
superpotential. In addition the lagrangian may contain non-renormalizable terms. At
the scale, (~0), the coupling constant gL is related to the coupling constant of the
full theory by a group-theoretical integer factor a: g2 = ag21**. Let t/.L be the index
of the light matter fields and G L the index of the light gauginos. If no tree-level
superpotential is present, a ( G - F ) = G L - P ' L . Otherwise, more matter fields can
acquire masses due to ( ~ ) , and a F - Ft. >1 a G - G L. At the one-loop level, the scale
A L is related to A and (¢p) by

(2.19)

At the scale A L, the low-energy gauge interactions become strong. (By assump-
tion, this gauge theory has no flat directions, so it is truly a strongly interacting
theory.) Various cases must be distinguished. First, the low-energy theory may break
supersymmetry. For example, the low-energy theory may be one of the models
discussed in ref. [9] or [10] (examples appear in the next section). If this is the case.
the vacuum energy is necessarily of order A4t., and thus a potential is generated in

* This is only the case if this theory is asymptotically free; otherwise, there are presumably no
interesting non-perturbative effects as we evolve toward the infrared. We will assume that this is the
case from now on.
** Usually this constant is one. An example where it is not is an SU(3) gauge theory broken to its 0(3)
subgroup under which the triplet of SU(3) is the triplet of 0(3). In this case, a = 4.
570 I. Affleck et al. / Dynamicalsupersymmeto' breaking
the flat direction:

4(A)4°o°' ~')/3(;L-~'L
V(~) - At = ~0 -~ (2.20)

At energies below A L, supersymmetry is broken and the effective lagrangian need


not be supersymmetric (more precisely, supersymmetry need not be realized linearly).
Therefore we should not be surprised that a potential is generated without a
corresponding superpotential.
If, on the other hand, the low-energy theory does not, by itself break supersymme-
try, it may do so when higher-dimension operators are included. A dimension-five
operator is, in general, generated at the one-loop level. This can be seen by explicit
calculation [8, 27] or by the following indirect arguments. The effective coupling of
the low-energy theory is given by the low-energy fl-function

1 a 3GL -- #l.
+ In ~t (2.21)
g2.(~.) g2(M ) 16w2 M"

where g 2 ( M ) is the bare coupling of the full theory, defined at some cut-off scale,
M. The effective coupling of the full theory is

1
+ 36" -__~In .~-~. (2.22)
g'(lx) g2( M) 16rr2 M

But the couplings should be equal (modulo the factor a) at /.t = (~>, the symmetry
breaking scale. This requires a modification of L~fr:

L.fr=gl/d2~(l__647r
2g2[(3GL_tXi.)_a(3G_lx)]ln(~))W.W
." (2.23)
Here ~ is some field which obtains an expectation value along the flat direction. If
we now consider some particular vacuum state q~ = (~> + ~', this extra term corrects
the fl-function and also produces a dimension-five operator - i'd 20 (cp'/(~p >) W 2.
The necessity of this term also follows from symmetry arguments. ~ must first of all
be chosen to be invariant under all non-R symmetries. (This means, in general, that
it is really a fractional power of some invariant product of fields.) The non-anoma-
lous R-symmetry of the original theory must also be realized in Lcrf. But the
anomalies of the massless gauge and matter field fermions of the low-energy theory
do not, in general, cancel, leading to a shift in Lcff- F*F (the topological charge
density). Such a shift also arises from the extra term in L~fr (if ~ has a non-zero
R-charge). The coefficient of this term must be such that the two effects cancel.
Relationships between anomalies and fl-functions in supersymmetric theories ensure
I. Affleck et al. / Dynamical super.wmmett T breaking 571

that these two methods of determining the coefficient agree. We show this explicitly
in the case where the theory contains no tree-level superpotential. We must choose
the field ¢p to be A1/~' in this case. A has R-charge G - ~ t . Using the relationship
a ( G - / ~ ) = G L - P'L, the shift in L e tf under an R-transformation due to the extra
term is ( G / # - 1)[(G L - a G ) / 3 2 ~ r 2 ] F * F . On the other hand. the shift due to the
R-anomaly of the low-energy theory is (G L - # L G / # ) F * F / 3 2 ~ r 2 (recall that all
fermion matter fields have R-charge G/it). Using again a ( G - I~) = GL - t~L, we see
that these two quantities cancel.
The extra term in L¢ff also contains a component ( F ' / ( ~ p ) ) k ~ where F' is the
F-component of the shifted superfield ~', in a particular vacuum state. This term is
small and may be treated perturbatively. It is natural to expect that the discrete
chiral symmetries of the low-energy theory are broken by an expectation value for
the gluino bilinear k~,:

(~') A3 993[.~1"~3c'-~'1/~G/a~
- = -- (2.24)

Through the coupling of eq. (2.23), this induces a superpotential of precisely the
form of eq. (2.14). The apparent branches of the superpotential can be shown to be
related to the different values of ( ~ k ) , and correspond to the spontaneous breaking
of a discrete chiral symmetry [8]. Theories in which there are light matter fields
transforming under the unbroken gauge group can be analyzed along similar lines.
Having described the general procedures for analyzing theories with flat direc-
tions, we face an obvious, important question: does the dynamically generated
potential, Vcn(cp), ever rise at large field strength? It is worth mentioning, first, that
asymptotic freedom does not imply Vcrf ~ 0 as cp ~ oo, but simply that Vcfl-/~[9 4 ~ O.
TO answer this question, we should distinguish between the various cases we have
discussed above.
If the gauge symmetry is completely broken and instantons generate a superpoten-
t i a l , Veff ~ ~ 0 4 ( A / ~ ) 3G-~ (3G - / x is always even, if there is no Witten anomaly). For
3 G - g < 0, the theory is not asymptotically free. For 3 G - # = 0, the two-loop
fl-function determines whether the theory is asymptotically free or not. If it is, V~ff
rises at infinity. For 3 G - # = 2 or 4, the potential rises or becomes a constant at
large fields. For 3G - g > 4, VCff vanishes at infinity and the theory does not have a
vacuum (apart from the rather remote possibility of a supersymmetric, strongly
interacting ground state near the origin). When there is no tree-level superpotential,
G - / ~ = 2, and V~fr= ~ -2~c-x). Thus, in such theories the potential always vanishes
at infinity.
When the gauge symmetry is not completely broken, we have seen that typically
V¢n - A4(ep) or A6(ep)/rp 2. Depending on G, /.t, GI, gL and a, the potential may or
may not rise at infinity. However, when no classical superpotential is present, the
potential always falls at infinity, since a ( G - g ) = G L - ~ L , SO (eq. (2.19)) A L
572 1. Affleck et al. / D v n a m t c a l supers vn metry breaking

q~ 2~6 ~ ; ~ / o ~ , , - ~ . (Recall that G > G L, a >/1 and 3G L - ~ L > 0 in order that the
low-energy theory be asymptotically free.)
We thus see that in any theory with flat directions and no tree-level superpoten-
tial, any dynamically generated potential falls to zero at infinity. Applying the
criteria described above, this turns out also to be the case for most theories with
tree-level superpotentials. In fact, we have found no examples of theories in which
all flat directions are raised. An example in which some are raised is presented in
sect. 3.
In view of these results, we are led to consider theories without flat directions.
Several general arguments indicate that such theories may break supersymmetry,
under certain conditions [9]. Suppose that in some supersymmetric theory an
exact global symmetry is spontaneously broken. Then there must exist a massless
Goldstone boson, ~'. If supersymmetry is unbroken, this particle is accompanied by
superpartners, and in particular by a massless, spin-0 particle, o. Since ~ has no
potential, the same is true for o, and thus ( o ) is undetermined. If o itself is a
Goldstone boson of another broken symmetry, then corresponding vacua are related
by symmetry transformations. This is the case in the non-linear supersymmetric
G-models based on compact K~hler manifolds [28]. If o is not a Goldstone boson, its
range must be non-compact. The possibility of a compact range can be ruled out by
either of two arguments. First, if the underlying theory were coupled to gravity, then
N e w t o n ' s constant would be quantized [9,29]. Another argument is based on the
impossibility of writing a supersymmetric Wess-Zumino term on such manifolds
[30]. Therefore, the anomalies of the underlying theory cannot be realized in the
low-energy effective lagrangian if the manifold is compact. The vacuum degeneracy
is thus larger than that expected from symmetry considerations. (In particular, vacua
labeled by different values of o will have different physical properties.) This
phenomenon, first noticed by Ovrut and Wess [31], is referred to as "complexifica-
tion of the symmetry group." Such a degeneracy is perfectly reasonable in theories in
which the classical theory has flat directions, but it makes no sense in cases in which
the tree-level potential rises in all directions. We therefore conclude that if the
classical theory has no flat directions and a continuous global symmetry is sponta-
neously broken, then supersymmetry must be spontaneously broken.
Further support for this argument may be obtained from an anomaly recently
discovered by Konishi [32]. Konishi has shown that, if % and q~, are components of
a charged chiral field ~,, then

IT - ,,~ aW g2
~ v ~ { O ' ~ ' ~ k ' % } = - ~ , % + ( ~ 3 2 ,/./,2 ~ " " (2.25)

Here i is not summed, ~),~ is the supersymmetry charge and C is the Casimir of the
q~, representation. The first term on the right-hand side is the canonical one; the
I. Affleck et al. / Dynamical super,~vmmetrv breaking 573

second represents the anomaly. If the superpotential is independent of O,, then

i ,ri-/~ ,t,tar~ \ g2
~/-~ \ ~ a , ,r, , r , / = C (~,~) (2.26)
32w2

Thus, provided that the theory contains at least one charged field which does not
appear in the superpotential, (~,2~) is an order parameter for supersymmetry
breaking. Any such theory possesses a non-anomalous R-invariance, and ( ~ ) is a
natural order parameter for R-invariance breaking. Thus, in such theories there is a
close connection between R-invariance breaking and supersymmetry breaking.
As discussed in the introduction, there are two classes of theories without flat
directions. In some theories, there are "almost flat directions," i.e., flat directions
appear as some coupling (a term in the superpotentiai or a weak gauge coupling)
tends to zero. In these theories, one may compute the effective potential explicitly
and determine which symmetries are broken. An example of this type has been
analyzed in ref. [10], and a further example appears in sect. 5. In all cases, if a global
symmetry is spontaneously broken, supersymmetry is broken, in agreement with the
discussion above. Theories with no such parameter are genuine, strongly interacting
theories. The difficult problem here is to demonstrate that some global symmetry is
in fact broken. We have at our disposal only certain indirect methods. In particular
't Hooft anomaly constraints usually strongly suggest some pattern of symmetry
breaking. These may be supplemented by instanton computations [12, 14, 18].
Clearly, then, it is important to search for and analyze theories without flat
directions. Several such theories have been considered in refs. [9-11 ]. Others will be
discussed in the next sections.

3. Models

In this section we catalog a variety of interesting models, using the techniques


described in the last section. Some of the theories analyzed in this section exhibit
dynamical supersymmetry breaking; some do not. We do not claim this catalog to be
in any sense exhaustive. In particular, we expect that there exist other theories which
break supersymmetry besides those presented here. Rather, these are meant to
illustrate the applications of our methods, and to provide examples for the discus-
sion of realistic models in sects. 5 and 6. We will present the results in the form of a
list. For each model we first give the gauge symmetry and particle content, followed
by a brief description of the flat directions (if any), and then determine whether the
flat directions are lifted, and whether or not supersymmetry is broken.
574 I. Affleck et at / Dvnamical super,symmetrv hreakmg

31. S U ( N ) WITH MATTER H E L D IN THE ADJOINT REPRI-SENTATION q~',

3.1.1. No superpotential. This is a theory with N = 2 supersymmetry, It has a


large set of flat directions: q0 = % (up to gauge transformations), with cpo diagonal,
complex. In general, these vacua possess the unbroken gauge symmetry [U(1)] '~
These theories have a non-anomalous R-symmetry under which cp has R-charge
zero. This forbids any superpotential, as do the SU(2) symmetry which mixes (X, k )
and the N = 2 supersymmetry. (An explicit examination of zero modes in an
instanton background, along the lines of refs. [7, 8], shows that four zero modes are
not lifted, and thus no superpotentiai is generated). We thus conclude that the model
has a non-compact set of vacua.
3.1.2. I V : = ~mTrcp2+~Trq~ 3. This lifts the flat directions. There are no re-
maining chiral symmetries and so no reason to expect dynamical supersymmetry
breaking. Also, as m--* oo, the matter field decouples and supersymmetry is un-
broken [3].

3.2. QCD WITH N COLORS AND ,%,) FLAVORS, PLUS A SIN(iI.ET, S WITH COUPLIN(;
W~.I = X/SC)/QI
For simplicity, we consider the case where all the ?~j's are equal. The theory
without a singlet has flat directions [6,8]. Flat directions still exist with S = 0 (e.g.
Q.~l = ~)12 all others vanishing). Any potential which is generated in these direc-
tions falls to zero for large fields. We consider, instead, the flat direction with
Q = Q = 0, s 4= 0. The intermediate energy theory has only the singlet, S, and the
S U ( N ) gauge multiplet. There is a coupling between these sectors:

3~i(lnS)W2, (3.a)

by the arguments of sect. 2. The non-zero value of ( A ~ ) in pure supersymmetric


SU(N) Yang-Mills induces an effective superpotential in the low-energy theory:

Weft ~ s N / / : ~ A 3 - Nc/N (3.2)

V ~ S 2<u:/~ ])

(Note that this form for Wen is uniquely determined by a non-anomalous R-invari-
ance.) Thus, for N! >1N, this fiat direction is raised at infinity by non-perturbative
effects. As discussed earlier, we have no example in which all flat directions are
raised.
L Affleck et al. / Dynamical supersrmmeto' breaking 575
3.3. SU(N) WITH AN ANTISYMMETRIC TENSOR, A,j, AND ( N - 4)
ANTI-FUNDAMENTALS, F'
3.3.1. N even.
(i) N = 4.
This theory is equivalent to 0(6) with a vector, V,,. The flat directions have
AtA = Icl2a~.By a gauge transformation, we can take

0 1 0 0
A =c -1 0 0 0 (3.3)
0 0 0 1
0 0 -1 0

The unbroken symmetry is SP(4)(O(5)). The only light matter field is one singlet. In
the notation of sect. 2, G = 8, ~ = 2, GI. = 6, ~L = 0. The unique invariant super-
potential is thus

wo,- (vyo) ,,3= (,,j,~,,A,) ,J3 (3.4)

W~ff is generated by (X)~) of SP(4). (The condensation of ( ~ X ) is consistent with


the Witten index for SP(2N) groups [3].)
(ii) N >i 6.
The flat directions satisfy

O = 2A*A - E f t / i f / s = 1cl21. (3.5)


/

It is convenient to diagonalize A ' A , using our S U ( N ) freedom. Since the eigenvalues


are equal in pairs, we find that the only solutions are

ff/=c~{diag(a,a,b,b ..... d,d), i,f=l ..... N - 4 } , (3.6)


0, ,=N-3 ..... N

AtA = diag(lalZ, lalZ, lbl2, lbl2 . . . . . 0 , 0 , 0 , 0 ) + M .

The unbroken gauge symmetry is SP(4). Again, the low-energy theory has one
singlet. The theory has G = 2N, /~= 2 N - 6, G t = 6, #t. = 0. Following the rea-
soning of sect. 2, the dynamically generated superpotential must have the form

w~ff_ [A,,-2r,... ~ ,1-'- (3.7)

Using methods similar to those of ref. [10], one can in fact show that this
superpotential is unique, i.e. that there is a unique contraction of the gauge indices.
Again, Weft is generated by ()~X) of SP(4).
576 I. Affleck et aL / l)ynamical super.sTmmetrv breaking

At tree level, one can add a superpotential to this theory:

we, = (xT= - x ) . (3.8)

This does not lift all of the flat directions. The theory, again, probably has no
ground state.
3.3.2. N odd.
(i) N = 5 [3,9,131.
This theory has no flat directions. No invariant superpotential (renormalizable or
otherwise) exists for this theory [9]. The complexity of all solutions to 't Hooft
anomaly conditions [9], as well as the instanton argument of ref. [13], strongly
suggest that R-symmetry and hence supersymmetry are spontaneously broken. The
vacuum energy is necessarily of order A4.
(ii) N >/7.
With no tree-level superpotential, these theories have flat directions. Proceeding as
above,

ff/=v~{diag(a,a,b,b ..... d,d,O), i , f = l . . . . . N-4t, (3.9)


O, i=N-3 ..... N)

A*A = diag(la 12, lal 2, [bl 2, [bl 2. . . . . 0, 0,0, 0,0)

The unbroken gauge symmetry is SU(5); the low-energy theory contains a 10 and a
5, and ~ ( N - 4 ) ( N - 5) singlets. There is, again, no superpotential which can be
generated in the low-energy lagrangian. The theory has G = 2N. p, = 2N - 6, G 1 = 10,
~tL = 4. The scale of the SU(5) theory is A t , - ~- i.~x ~o~/~, and the scalar potential
satisfies
V(ep) - A ] . - q~-,',,2x-,o, (3.10)

These theories probably have no ground state.


These theories provide an example of another phenomenon discussed in sect. 2.
The low-energy theory (at scales E << A I_), even in the approximation that we ignore
the coupling between the singlets and the SU(5) sector, is not supersymmetric. We
expect it to consist of Goldstone bosons, a goldstino and a massless fermion required
by 't Hooft anomaly conditions in the SU(5) sector [9]. (It is also possible that there
is an additional Goldstone boson instead of the massless fermion.) In addition, there
are the components of the singlet supermultiplets. Thus there is no supersymmetric
low-energy lagrangian. In the even-N case, by contrast, the scale of supersymmetry
breaking is A3(2/ef/2 << A t., so the low-energy lagrangian is supersymmetric.
T o these theories we can add a superpotential [13]:

IV., =X/gA,jk}'E~ [Xr = - X ] . (3.11)


L Af/leck et a L / Dvnamtcalsuper~ymmetO' breaking 577
If A is real, by redefining the f f ' s we can choose

2` = ( 2`10'2
2`2o2 (3.12)

This superpotential raises all of the flat directions. (Proof: diagonalize A ' A , and note
a W / O f f l = 0 implies A ' A f t 2 = 0 which gives F 2' = 0 unless N - 4 <~ i ~ N, if a, b, c . . . .
are all non-zero. This cannot yield D / - 8/. Taking a, b, c, to be zero does not help
either.) For all 2,, << 1, in the previous flat directions there is a coupling of the
singlets, v - XEq04. The SU(5) sector gave E ~ ~-4/13(2,v 10). The coupling of the
singlets to the SU(5) sector gives smaller effects. Thus supersymmetry is broken, and
the m i n i m u m of the potential occurs for

q)--2` 13/2(2N+3)A, Evac~A (4N 2°)/(2N+3)A4. (3.13)

3.4. O(N) WITH A VECTOR, V~


3.4.1. N = 7. This theory has flat direction V, = c8,1. The unbroken gauge group
is 0 ( 6 ) ( = SU(4)). The low-energy theory has one singlet. This theory has G = 10,
= 2, GL = 8, P'L = 0. A superpotential is generated by (2`2`) of 0(6) (SU(4)) (we
expect (A,A.), 0, since 0(6) is isomorphic to SU(4), and it is believed that 2`2`
condenses for all SU(N)):

W~ff- (V. V ) - 1 / 4 (3.14)

The theory thus has no ground state. One can include in this theory a tree-level
superpotential:
W~l= m V . V. (3.15)

This lifts the flat direction. For small m, we can minimize the sum of Wef t and Wd.
There are 5 supersymmetric vacua at V, ~ 8ix(A7/5/mZ/5)e2*'"/5, n = 0 . . . . . 4. The
discrete Z10 chiral symmetry of the massive theory is spontaneously broken to Z 2.
We might expect, by continuity, that as m --, oo and we recover the pure 0(7) gauge
theory, the discrete symmetry remains broken. (Our calculation, of course, is only
reliable for rn << A.)
3.4.2. N >>.8. Again, these theories have flat direction V, = c6,~. The low-energy
theory has one singlet. G = 2 N - 4 , ~=2, G L=2N-6, /.t L = 0 . The allowed
invariant superpotential is

Werr=(V-V) 1/(,v-3) (3.16)

If we assume that the above continuity argument is correct, then we conclude, by


578 I. Affleck et at / Dvnamwal super.~3,mmetrv breaking

induction, that l,V~ff is generated by @ X ) of O ( N - 1 ) . If we add a classical


superpotential, Wc~=mV. V, there are N - 2 supersymmetric vacua at V -
m - ( N 3)/(2N- 4~ and the Z2(.~, 2~ symmetry is spontaneously broken.
F r o m Witten's computation of the index [3], it is not surprising that supersymme-
try is unbroken. However, the value of the index suggested by our calculation is
N - 2, whereas Witten obtained ~(N + 1) or ~(N + 2) for N even or odd, which is
not consistent with chiral symmetry breaking. For O ( N ) groups, the index calcula-
tion could not be performed with twisted boundary conditions and instead relied on
a Born-Oppenheimer approximation. It is possible that the index changes as the
volume is taken to infinity. As Witten observed [19], we would expect chiral
symmetry breaking for large N since then O ( N ) becomes equivalent to S U ( N ) and
the index suggests chiral symmetry breaking in the S U ( N ) case. There are also
theories (e.g. O(10) with a 16, a 16, and 2 10's) where one can do a reliable instanton
computation of the superpotential. In the presence of a small mass term, the discrete
symmetry will certainly be broken. Chiral symmetry breaking in pure O( N ) theories
should also follow from the methods of refs. [12, 18]. We note finally that in all the
other theories we have studied, whenever a superpotential was allowed by the
symmetries, it was generated.

3.5. S U ( N ) W I T H A S Y M M E T R I C TENSOR S,~ A N D ( N 4 4) A N T I - F U N I ) A M E N T A L S . ~-)'

This theory has flat directions with

ff/= { a,3}, i,f= l ..... N 1. (3.17)


O, otherwise )

StS= diag(lall 2, la212 . . . . ) + M .

The gauge symmetry is completely broken. In this theory, G = 2N, ~t = 2N + 6, so


no Wefr is allowed. The vacuum degeneracy is not lifted.
One can add a tree-level superpotential:

Wd = X/~S,,Ft'F ~ (Xx = X). (3.18)

For SU(3) two other terms are possible: det S and

X/s%,,k b)'F~,F~ . (3.19)

We only discuss the first term eq. (3.18) and only in the most symmetric case:

X/'~ = X3 f~. (3.20)

Again there are fiat directions. If S ~ O, diagonalize S+S. Then 8W/3F = 0 implies
I. Affleck et al. / Dynamical super.sTmmet O" brea~mg 579

S*Sff~=O. Together with D = S * S - 2 F . f f / f f / t - l , we deduce i f = 0 and S * S - 1 .


This implies S = c 1, up to a gauge transformation. If S = 0, we can use F / F / = 0,
~ ' f f J / - 3 } to show that there are no flat directions for N/> 5. (Proof: make
F/I - ( 1 , i , 0 , 0 . . . ) by an O ( N + 4) transformation and the condition /~lfft~ = 0.
Then ff/'ff/2= ff/tF~= 0 implies i f / z = 0 for f = 1,2. By a gauge transformation make
fir2 = (0, 0, 1, i, 0, 0 . . . ) and so on. Proceeding in this way, we need at least 2 N flavors
to obtain a flat direction.) We consider only the S 4:0 flat direction. This has
unbroken gauge symmetry O ( N ) . The low-energy theory has one singlet (from S);
all f f ' s get large masses from W~. A coupling between the O( N ) gauge multiplet and
the singlet, (1/32~r2)(trlnS)W 2 is required by the arguments of sect. 2. This
produces a potential energy

V I(AA)I2 A6 (3.21)
S2 S2 •

assuming that (AA) 4= 0 in the pure O ( N ) theory. Since G L = 2N - 4 ( N >/4),

A L -- S - ~,'/3~,v- 2)

W- (detS) t/~.~ .. 2),

V - S -"N-1)/~'' 2~ (3.22)

which agrees with the power obtained by R-invariance arguments.


N o w consider the cases N = 3 or N = 4, where there are additional flat directions
where the .K's are non-vanishing. In these directions, the gauge symmetry is
completely broken, and instanton methods are reliable. In both theories, however,
the index requires more than two zero modes, so no superpotential is generated and
the flat directions are not lifted.
Let us return to the flat direction with S 4= 0. For A = 0, the full effective potential
was flat in this direction. For non-zero A, it slopes to zero at large S. Suppose A is
very small. Our previous analysis was valid for AS >> A t . But there is a range of S
such that AS << A e but S >> A t . In this region, the F ' s are light, (m << AL) and the
low-energy theory has approximate flat directions, lifted by amounts of order ~.
Thus, starting from the origin, the potential of the full theory grows to a local
m a x i m u m for some non-zero S, and then falls. (Of course, there might be several
local minima and maxima in the strong coupling regime.) At the origin, the potential
m a y or may not vanish. For N = 3 or N = 4, it almost certainly does vanish there,
since we have seen that the potential vanishes in the S = 0, ff:~ 0 flat directions.
580 I. Affleck et al. / Dvnamwal super,~Tmmetrv breaking

4. SU(3) x SU(2) model

In this section we consider what is probably the simplest model with calculable
dynamical supersymmetry-breaking. The theory is QCD, with three colors and 2
flavors, Q,! and ~2'/. We gauge the SU(2) flavor symmetry acting on Q, and add a
weak doublet L/ to cancel the SU(2) anomaly. (This would be the Weinherg-Salam
model with one generation of quarks and leptons if we added the positron, e ' . and
gauged hypercharge. Doing this would not alter the following discussion signifi-
cantly.) With no superpotential, this theory has fiat directions:

Q,/= Q ' I = , L = (O,v/a z - b-2-). (4.1)

For simplicity, we will assume that the SU(2) interactions are much weaker than
the SU(3) interactions. Then SU(3) instantons generate the superpotential

2A~3
W~.= d e t ( Q Q ) " (4.2)

and we expect the theory to have no ground state. (Here we have picked our
subtraction scheme so that the constant b of eq. (2.12) is 2,)
T o this theory we can add a tree-level superpotential:

Wcl = X eJ~Q,/ff'L.~. (4.3)

(We refer to the flavors as u and d.) The resulting theory is easily shown to have no
flat directions. This coupling preserves hypercharge and a non-anomalous R-symme-
try. Combining Wd and Wef f, we obtain a potential with a non-zero minimum; this
follows from the fact that the potential blows up both at zero and infinity. At this
minimum, the R-symmetry is necessarily broken, so supersymmetry is broken (as
discussed in sect. 2). We will shortly verify this by explicitly minimizing the
potential. For now, note that V contains a term

oqW - ~

OL = ]XQ,fd ]-, (4.4)

and where this vanishes, det(~)Q) - 0, so V is infinite.


A simple scaling argument indicates that the minimum occurs at

A3
(,~) x '/7 , E- x2(,~} 4. (4.s)

Thus as X becomes small, the vacuum moves to large q)( >> A 3) where the SU(3)
coupling is small. Thus, in this limit, the vacuum state is weakly coupled.
1. Affleck et al. / Dvnamtcal supersvmmet O" breaking 581

For X --- 0 the gauge symmetry is completely broken along the flat direction so 11
chiral fields are eaten by the super-Higgs mechanism and get masses g(cp), leaving 3
light superfields. These can be conveniently represented by the 3 gauge-invariant
composite operators

X / - O/'Q,zLhe sh, Y = det(~gQ). (4.6)

The low-energy theory is a supersymmetric o-model. The classical K~ihler poten-


tial can be determined by solving for the operators that appear in the original
iagrangian (Q+Q, Q~ Q, L ~L) in terms of X and Y using the equations D 3 = D 2 = O,
as in sect. 2. We find

K(X/,y ) 6A+2B(A+VrA2+Ba)'/3+2B(A-C-~-B3) '/3


= (4.7)
[(A + 6T -8 )1J3 +(A_ CA - 2

where

A = IX;X:, 8= ~ ( r ~ r) '/2.

Loop corrections renormalize K but these effects are small at large % Finally,
instantons add a term 2ATX i to the superpotential. Thus, for small ~, the full
low-energy effective lagrangian is simply

L~,f= fd',~K(X:,Y)+ fd20(XX'2)+2A~3Y ~). (4.8)

All properties of the theory can be read off from La6 the location of the minimum,
vacuum energy, spectrum, etc. Hypercharge is clearly unbroken if ( Y ) = 0, which we
find to be the case. (More accurately a linear combination of hypercharge and SU(2)
and SU(3) transformations is unbroken.) Note that the smallness of X and the SU(2)
coupling and the asymptotic freedom of the SU(3) coupling permit a systematic
weak coupling expansion in this model. Our renormalization scheme, in which the
coefficient in W~ff is 2, can be determined straightforwardly by a one-loop computa-
tion [331.
From a computational standpoint we found it more convenient to write the
potential in terms of the original fields, due to the complexity of K(X, Y). We
introduced four real parameters (a and b plus two angles) for the flat directions of
the D-terms, computed V = Z, I OW/O~,l 2 as a function of these parameters and
minimized. We found the minimum had the form of eq. (4.1) with

A A
a = 1.286 . . . . ~1/7 ' b = 1.249 . . . . X1/7 ' E = 3.593 . . . Ax°/TAa. (4.9)
582 I. Affleck et al. / Dvnamwal .supers.vmmetry breaking

The light fermion spectrum consists of the Goidstone fermion, a massless fermion of
charge one required by the 't Hooft anomaly condition for hypercharge (the
"electron") and a neutral fermion of mass 1 1 . 2 9 . . - ~6/7A. We calculated this by
projecting the 14 x 14 matrix 82W/Ocp,aq~jonto the 3-dimensional subspace spanned
by OX//Sq~,, OY/Oq~,, and then finding the eigenvalues. Following a similar proce-
dure for the scalars we find a massless Goldstone boson (of the broken R-symmetry),
a charged scalar and three additional neutral scalars of mass - h6/~'~.

5. Non-perturbative effects in realistic models

5.1. M O D E L S W I T H U N B R O K E N SUPERSYMMF.TRY AT T R E E I.EVt-I.

The simplest realistic supersymmetric grand unified theory [1,34] is SU(5) with a
24 (Z) and a 5 and 5- (H, H) of Higgs fields as well as quarks and leptons in 5 and
10 representations with superpotential

W= ~MtrX2 + ~trX3 + A'H(Z + 3-~-~)H

+ matter-field couplings.

The H H mass has been fine-tuned to make the doublets massless. The addition of
soft supersymmetry-breaking terms (which might be generated by supergravity or an
O'Raifeartaigh sector) makes the model phenomenologically acceptable. Here we
wish to examine the non-perturbative behavior of the model in the absence of any
explicit supersymmetry breaking [1 ].
We first consider the model without the quark and lepton superfields. The
low-energy theory is then an S U ( 3 ) x S U ( 2 ) x U ( 1 ) gauge theory with a pair of
massless Higgs doublets, H v, H D. The SU(3) gauge muhiplet is decoupled; the
SU(2) sector is what we have called QCD with two colors and one flavor (not to be
confused with real QCD). There appear to be flat directions in this theory but, as
discussed in sect. 2, we must look for higher-dimension terms in W~ff, generated by
tree graphs.
T o lowest order, the shift in Z induced by a non-zero value of H-, H is given by
m

( M'SZ)', + X'H'H, = 0, (5.1)

where M' is the X-mass matrix. Shifts are induced only in the components of X not
eaten by the super-Higgs mechanism (since M is zero on the Higgs components). We
find
~k'2
W~}~ 20M ( H H )2. (5.2)

Thus there are no flat directions; the classical minimum lies at /4 = H = 0.


L A/fleck et al. / Dynamical supersvmmettT breaking 583

We now turn to non-perturbative effects. The low-energy, SU(2) instantons


generate an additional term in W~u:

Weu - x '2- 'tt I=t. / ) ,2 +


2-OM & ' (5.3)

where A 2 is the scale of the SU(2) theory:

(-'-~) 5 - e "8"2/~r2(M) (5.4)

Minimizing Veu we find a supersymmetric vacuum with SU(2)× U(1) broken to


U(1) .... at a scale

(H) 1 e_,,,~/3g:¢g)M. (5.5)


(X,) 1/3

Thus instantons generate an exponential gauge hierarchy; the required hierarchy of


order 10- x5 would result for a ( M ) = ~ (and ~' = 1). However, the resulting mass of
the Higgs multiplet is

M n - Me-S":/3gqa4~(~') a/3 --- 10 4eV! (5.6)

One might worry that this discussion has ignored possible non-perturbative effects
from very small instantons, i.e. instantons of size less than or of order M 1. Such
instantons do not respect the approximate, accidental symmetries of the low-energy
theory, nor need they respect the non-renormalization theorems. Thus they can
generate low dimension, renormalizable terms, such as mass terms for Higgs. and
could be of some importance. These are extremely small, however. Any such effects
would generate corrections to W~u of the form

8Wetf = M 3 e 8,,'-/g'-{MV( H H / M z), (5.7)

where f is some function of order one. (Indeed, a Higgs mass term appears to
be generated.) This correction, however, is insignificant at the minimum we ob-
tained above. It is suppressed, relative to the terms computed previously, by
e x p ( - }8~r2/g2).
Let us now include three generations of quarks and leptons. In this case there is a
much larger class of flat directions in the low-energy theory before inclusion of
higher-dimension operators. In fact, there are many directions which remain flat
even when higher dimension operators are included [35]. For instance, the up, down,
and strange anti-squarks can get expectation values of equal magnitude (with
584 1. Affleck et al. / Dvnamtcal super.$yrnmetry breaking

H = H = 0) completely breaking SU(3) color. There is no higher-dimensional mono-


miai that could lift this flat direction which preserves B - L and the low-energy
gauge symmetries. Instanton effects, at best, will produce a potential that slopes to
zero at infinity since, for q~---, :~

W ~ q~3e ~ 2 / C ~ ) <~ A 3 . (5.8)


One might again hope that small instantons (p - M - 1) could have important effects,
here, since they respect less symmetry. However, any instanton generated super-
potential takes the form

W - e-S": /g:~M)M 3f( ep/M ) + a~3e - 8,,:/.~i~o), (5.9)

where the first term represents the effects of the very small instantons. Since

(M) (3(;| /11)/2


e s,,:/g~(,~) = e - s~-"/.~'(.~t) (5.10)

the second term in W above has the same form as the first, so any minimum must
occur at ~ - M (or zero).
l e t us now ask, more generally, if a realistic model exists in which supersymmetry
is broken dynamically without the existence of additional low-energy gauge interac-
tions (supercolor). Let us demand that supersymmetry be broken at a scale M~ much
below M and let us assume (for simplicity) that no other mass scales are present
between M and M~. Lcrf, at scales _< M~ should be supersymmetric. It will contain a
renormalizable part L a as well as higher-dimension operators, suppressed by 1 / M .
The only gauge interactions in it are SU(3) color and G,~¢,,k (presumably S U ( 2 ) ×
U(1)). A realistic model can only result if supersymmetry is broken at tree level by
L~ft. For, if supersymmetry were broken by dynamical effects, the scale of this
breaking would be suppressed by powers of A/M~, where A is the SU(3) or G,~c~k
scale. (All gauge interactions are weak at scale M~.) We will assume that the
non-renormalizable interactions play no role in supersymmetry breaking. If they did.
the breaking scale would be less than M~ by a power of M~/M. It is hard to imagine
any effect in a weakly coupled G U T which would then generate a sufficiently large
value of M.~. (Once all matter field zero modes are accounted for, any instanton
generated scale is not likely to be too large.) With this restriction, we will now show
that L~fr cannot break supersymmetry at tree level. (One might ask if non-pcrturba-
rive effects associated with Yukawa couplings could break supersymmetry. All our
work has assumed Yukawa couplings to be small. This was necessary since they are,
in general, not asymptotically free. It is conceivable that dynamical effects associated
with the scale of an asymptotically free Yukawa coupling could break supersymme-
try, but we are not aware of such an effect.)
I. Affleck et al. / Dynamical super.~vmmetrv breaking 585

The tree-level breaking could, in principle, be of the O'Raifeartaigh or Fayet-


lliopoulos type. The supersymmetry breaking scale is determined by the coefficients
of low-dimension operators in Left; in the former case by a term #2fd4OS, where S
is a singlet; in the latter by ~fd%~D, where D is the auxiliary field of a U(1) factor
in Gwe~. If the supersymmetry breaking is to be dynamical in origin then these terms
must not be generated at tree level, but rather by non-perturbative effects. The only
relevant non-perturbative effects are those of instantons of scale size much smaller
than M~-1. These might, in principle, generate a value of /12 or ~ of order
M~,e- 8"2/g2<M~ (times matter field Yukawa couplings). With an appropriate value of
et(M) a TeV scale might be obtained. (Non-perturbative effects associated with large
instantons in the unbroken group are part of the dynamics of the low-energy theory.
As mentioned above, they have too small an effect.)
Witten has observed that no operator invariant under a simple unified gauge
group reduces to the U(1) D-term in Let t. This rules out the generation of this term
to any order in perturbation theory and presumably also rules out its non-perturba-
tive generation. Witten has informed us of the following, more general, argument
against the generation of a D-term [36]: if a theory with a D-term is coupled to
supergravity, a splitting of boson-fermion charges results [37]. But such a splitting
would not be consistent in a unified theory, which contains monopoles. Thus either
such a theory cannot be coupled to supergravity (even for arbitrarily weak values of
Newton's constant) or else no D-term is generated.
If the /t2fd20S term is absent at tree level, it will be absent to all orders in
perturbation theory due to non-renormalization theorems. Could it be generated by
instantons? Consider as an example an SU(5) model with 5 and 5 fields Q,,~9,
(i = 1,2) and a 24, Z with

W = mlO_lQ , + m2Q2Q 2 + MY 2 + ~,1~73 + QI( ~ 2.~--.,""t- h s S ) Q 2 + t~2S. (5.11)

Suppose/t 2 = 0 classically. The signal for the generation of a non-zero value of ~2 by


instantons would be a non-zero expectation value for F~:

( r , ) = (O~Q~). (5.12)

But such a scalar expectation value could never be generated by instantons because
there are always at least two fermionic collective coordinates (two supersymmetry
zero-modes). Quite generally, the generation of a linear term requires a scalar
expectation value and so cannot occur (non-linear terms result from fermion bilinear
expectation values and so can be generated). Thus we conclude that dynamical
supersymmetry breaking requires supercolor.
586 I. Affleck et al. / Dynamical super~)'mmetry breaking

5.2. N O N - P E R T U R B A T I V E EFFECTS IN O'RAIFF_ARTAIGH MODEI.S

A class of realistic models with supersymmetry broken by the O'Raifeartaigh


mechanism have superpotential [38]

W = MIO.,QI + M20_2Q2 + s(XOxQ2 -/~2). (5.13)

where Q, transform as some non-trivial representations of the gauge group, ~),,


transform as the conjugate representation and S is a singlet. At tree level, the
conditions

OW
= MxO_1 = O,
OQI

OW
OQ2 M2Q2=0'

3W
OS = X 0 t Q 2 - ~2 = 0, (5.14)

are incompatible, so supersymmetry is spontaneously broken. Such a theory is


potentially unstable against dynamical supersymmetry restoration since the non-
anomalous symmetries allow terms in W which would lead to a supersymmetric
minimum (e.g. MS2).
We will analyze non-perturbative effects in the weak-coupling limit X << 1. and
with ~t2 small. We then regard ~<S> as a term in the mass matrix of the Q,.~),
theory, so that the theory becomes supersymmetric QCD with two flavors (if Q is in
the fundamental representation). We have shown that additional terms are generated
in W which can make (Q,Qj) ~ 0. It might then appear that for some value of <S>,
X(~)IQ2) =/.t 2, and supersymmetry is restored. However. this does not happen. The
simplest way of seeing this is to use the Konishi anomaly [32] which is consistent
with our non-perturbative results on supersymmetric QCD [8] and presumably, for
these purposes, more general (i.e., valid for M > ,4 ). This gives (recall that supersym-
metry is unbroken in massive QCD):

OW 3W

= <~),U2> = O. (5.15)

Thus, even though ( Q , Q , ) and ( 0 2 Q 2 ) can be non-zero, (~)~Q2) remains zero


when non-perturbative effects are included, and thus does not disturb the
O'Raifeartaigh mechanism.
I. Affh'ck et al. / Dynamical supersymmet O" hreaAmg 587

6. Model building: problems and prospects


In the last several sections, we have seen many examples of theories which exhibit
dynamical supersymmetry-breaking. We have also seen that if we are to build a
realistic model with DSB, we must have additional gauge interactions beyond
S U ( 3 ) × S U ( 2 ) × U(1) in the low-energy theory, even if this theory is eventually
grand-unified. In the present section, we will consider the prospects for realistic
model building with DSB. We will explore two basic approaches. First, we will
consider the possibility of coupling one of these theories to N = 1 supergravity [39],
using it as a "hidden sector" [20,21]. We will see that this approach has some
advantages over existing models. In particular, it is more predictive. In considering
N = 1 supergravity theories, since they are non-renormalizable, a conservative ap-
proach is to write at the Planck scale (or slightly below) a lagrangian which is the
most general one consistent with its symmetries, with all dimensionful couplings
given by appropriate powers of M e . If one proceeds in this way in existing
supergravity models, the low-energy theory always has an enormous number of
parameters (every superpartner mass, for instance) [21]. To avoid this difficulty, one
must assume that the lagrangian, for unknown dynamical reasons, possesses certain
approximate symmetries. Even when this is done, the low-energy theory depends on
several unknown quantities, such as the form of the K~hler potential at large field
strength. As will be explained below, the situation is dramatically improved if for the
hidden sector we take a theory which dynamically breaks supersymmetry. In fact.
soft supersymmetry-breaking in the low-energy theory is then described by one
parameter. This occurs because none of the fields of the hidden sector acquire large
(O(Me)) expectation values, and because gauge invariance significantly restricts the
form of the allowed couplings at Mp. This extra predictive power comes at a price:
the quark whose Yukawa coupling drives electroweak symmetry-breaking [20,21]
must be quite heavy.
Second, we will consider the possibility that supersymmetry is broken by some
new interactions at a scale not too different than the weak scale. In such theories, the
" n e w s " that supersymmetry is broken is transmitted to ordinary fields by gauge
interactions, along the lines of the original "supersymmetric technicolor" [I,2]
suggestions. We will see that there are several serious difficulties which such an
approach must overcome, and we will illustrate these with some specific examples.
To show that these problems can, in principle, be solved, we will construct a model
with a promising phenomenology, The model is not beautiful, but it gives some
reason to hope that a compelling supersymmetric technicolor model can be built.

6.1. HIDDEN SECTOR SUPERGRAVITY MODELS

We have seen that globally supersymmetric theories may exhibit a variety of


behaviors, and that a crucial element in understanding these behaviors is the
enumeration of the classical flat directions of these theories. In theories with classical
588 I. Affleck et al. / Dvnami(al supersymmetry breaking

flat directions, one of two things almost always happens. Either non-perturbative
effects generate a potential in these directions which falls to zero at infinity (case A),
or the flat directions are not raised at all (case B). In theories without flat directions,
supersymmetry is broken if a global symmetry is spontaneously broken (case C);
otherwise it usually is unbroken (case D). All of these possibilities may be of interest
when coupled to supergravity. We will focus on cases A (falling potential generated
in a flat direction) and C (supersymmetry broken).
When coupled to supergravity, theories of type C (broken supersymmetry) are
perhaps the most interesting. As an illustration, we can consider the SU(3) x SU(2)
model discussed in sect. 4; most of our remarks are easily generalized to the other
supersymmetry-breaking theories. (We will comment about strongly coupled hidden
sectors shortly.) The gauge group of the theory is S U ( 3 ) x S U ( 2 ) x G, where G
includes the standard SU(3)c X SU(2)i. x U ( l ) r (e.g. G = SU(5)). The supersymme-
try-breaking SU(3) x SU(2) interactions will be referred to as "supercolor." None of
the ordinary fields (quarks, leptons, etc.) carry supercolor quantum numbers;
similarly, none of the supercolor fields (the fields Q, U, D,
and L described in sect.
4) transform under G.
In hidden sector models [20,21], the scale of supersymmetry breaking (the SU(3)
scale, up to powers of coupling constants) is taken to be of order
[ r
tt- vmgMp- ~/rnwMr,, (6.1)

where rng is the gravitino mass and governs the size of the soft supersymmetry-
breaking terms in the low-energy theory. Because the expectation values of the
hidden sector fields are of order ~t << Mr,, gravitational effects (e.g. corrections to the
scalar potential) are very tiny, and the minimum of the potential and the spectrum
are just as described in sect. 4, up to small (O(mg)) corrections.
T o be more precise, let us consider what such a theory looks like when coupled to
N = 1 supergravity [39]. Of course since the theory is non-renormalizable it can at
best be viewed as a cutoff theory, with cutoff of order Mr,. In that case, we should
first consider the effective lagrangian of the theory at a scale ~, slightly below Mr,.
Physics at lower scales should be obtained by integrating out higher momentum
modes from this theory. The lagrangian at x should be the most general one
consistent with its symmetries, with all dimensionful quantities scaled by appropriate
powers of Mr,. In particular, recall that for a general K~hler potential, d(q~, ~p*), the
scalar potential is given by [39]

V = exp(8~'Gd) [
~'M OOp'
x{ O~pMO--f-f+8rrG
8q)Modf)*_24~rGlf12 ] + (gauge terms) (6.2)
L Affleck et al. / Dynamicalsupers),mrneto, breaking 589

where f is the superpotential, and gXM= 02d/&puOq~4. Let us assume that for
[cp[ << Mr,, d may be expanded in a power series in ¢p/Mp (modulo logs of [~[2/Kz,
which, for the quadratic terms are connected with ordinary wave function renormal-
ization, and which, for simplicity, we will drop; their inclusion does not change the
discussion in any essential way). Then, since for all the fields of interest, 1~[ << M r,
we need only keep the quadratic terms; also, the exponential prefactor may be
dropped. Thus, after appropriate rescalings, the potential is

V=~, O--f-f+8~rG~p~f 2 - 2 4 ~ r G l f 1 2 + ( g a u g e t e r m s ) . (6.3)


N 0¢P~

The gauge terms are also quite simple. They are, in general, of the form

f d2OhA(w)W~W~A. (6.4)

Ordinarily, such terms contribute to gaugino masses, which are then simply parame-
ters of the low-energy theory. This is not the case here, however. Because all of the
chiral fields of the theory carry non-trivial gauge quantum numbers, h(tp) contains,
apart from dimensionless constants, only terms of dimension higher than one. Since,
e.g.

f d20 Mg
~_..2_2W 2 (6.5)

generates a gluino mass of order

M~ <F)<(p) << Mw ' (6.6)

(recall (q0) < it), these higher-dimension terms may be ignored. Thus gaugino masses
vanish at the scale K, and their values in the low-energy theory are calculable
numbers, rather than parameters. (Their precise values depend, for example, on the
details of the G U T sector [21].)
It is, of course, necessary to require vanishing of the cosmological constant. In the
present context, this can only be arranged by including in f a constant, ap-
propriately tuned so that ( V ) = 0. This procedure fixes ( f ) . Then m s = 8~rG(f).
The potential for the "observable" fields is then the usual one [20,21] calculated
with a fiat K~ihler metric, with the so-called A-parameter set equal to zero. In
particular, the scalar partners of ordinary fields all have mass [rng[2 at the scale x.
In order to obtain SU(2) × U(1) breaking in this context, it is necessary to invoke
radiative corrections below the scale K. The only mechanism we know of is the
presence of a heavy (third or possibly fourth generation) top quark [20,21]. The
59() I. Affleck et al. / Dynamical super.~Tmmetrv I,reakmg

calculation is the same as that performed by several authors, now with A = 0 and a
small gluino mass. The top-quark mass must be 160-190 GeV [20,21] for S U ( 2 ) x
U(1) breaking to occur. The size of the gluino mass and other R-symmetry breaking
terms in the low-energy theory depend on the details of the full grand-unified theory.
In particular, if the interactions of MC;UT mass particles violate R-invariance (even
for M v --, oe), then these terms can easily be large enough [20,21].
One major bonus of this type of theory concerns the question of CP violation.
Generically, N = 1 supergravity theories give a contribution to the neutron electric
dipole moment three orders of magnitude above present limits [40]. This can be
avoided only if the quantity A and the gluino mass are real at the scale ~, in which
case, the dipole moment is typically of order the current experimental bound [22]. In
general, this may represent a significant fine-tuning. In theories in which supersym-
merry is dynamically broken, both A and the gluino mass vanish at •, and there is
no CP problem.
Of course, all fine-tuning problems have not been solved in this approach: the
cosmological constant remains a serious embarrassment. One might hope to solve
this problem by including additional light fields in the theory, but we do not
currently have a reasonable proposal. Three other points in favor of these models
should be stressed. First, unlike all models studied to date, it is not necessary to
omit couplings by fiat. The lagrangian can be the most general one consistent with
its symmetries. Second, because these theories do not have approximate flat direc-
tions in the hidden sector, they avoid certain cosmological problems of existing
theories [23]. T h i r d - - a n d this was, after all, a principal motivation for studying
low-energy supersymmetry breaking [1,2], supersymmetry is dynamically broken.
The hierarchy problem is completely solved in these theories. (Of course, the
problem of the cosmological constant is not.)
So far we have considered a hidden sector which is weakly coupled, such as that of
the SU(3) x SU(2) theory. In such a case, an equation such as (6.3) has a straightfor-
ward interpretation in terms of the fields of the original theory. For truly strongly
coupled hidden sectors, such as the SU(5) model with a single 5 and 10, a detailed
description is more difficult. Presumably one should describe the situation with some
type of effective theory, in which supersymmetry is non-linearly realized, coupled to
supergravity. However, since there is no qualitative difference between these and the
weakly coupled models, we expect all of our comments above to remain correct for
these theories as well. In particular, the low-energy theory should be described by a
single parameter.
Another class of models deserves at least brief comment. These are models with
hidden sectors consisting of theories of type A, i.e. theories (like supersymmetric
QCD) in which a potential is generated along a classical flat direction which falls to
zero at infinity. The behavior of such a theory, when coupled to gravity, depends on
the form of the K~ihler potential and the superpotential at large fields. In particular,
we might expect higher-dimension, non-renormalizable operators to appear in the
L Affleck et al. / Dynamical super,~ymmetry breaking 591

superpotential, with coefficients scaled by appropriate powers of Mp. Such terms


could only be forbidden by an R-invariance, but this invariance would necessarily
be broken by the constant, ~ in the superpotential required to cancel the cosmologi-
cal constant. In such a case, the equations O(W/Oq~')=O and D ' ~ = 0 would
typically have solutions. By choosing ~ so that ( W ) = 0, we would obtain a
supersymmetric minimum with zero cosmological constant. For other choices of ~j, it
might be possible to obtain a local (and stable or highly metastable) minimum with
zero cosmological constant. We will not explore this possibility here. Such an
approach is likely to suffer from many of the difficulties of standard hidden sector
theories. In particular, because some fields are likely to acquire large expectation
values, such theories will have limited predictive power. A class of models with
similar features has been considered by Ferrara, Girardello, and Nilles (ref. [47]).
These authors considered the effects of gaugino condensates, when pure gauge
theories are coupled to supergravity, with a non-trivial h(~o) (eq. (6.4)). A potential
is generated for the q>fields by ( ~ . ) . When proper account is taken of the
dependence of (~k,) on the fields ~o, as in sect. 2:
()~X) - A ((p)3 (6.7)

the resulting potential also falls at infinity, unless other terms in the lagrangian
stabilize it (e.g. a superpotential).

6.2. SUPERSYMMETRIC TECHNICOLOR

We next consider a class of models in which supersymmetry is dynamically broken


at scales of order a few TeV (modulo factors of ten or so) [1,2]. In such theories,
supersymmetry is broken by a new set of gauge interactions ("supercolor"), with a
scale A ~. Some of the supercolored fields also carry ordinary (SU(3) x SU(2) x U(1))
quantum numbers. Ordinary fields (quarks, leptons, and Higgs) "learn" of super-
symmetry breakdown through gauge interactions; because the supercolor theories
are necessarily chiral, no renormalizable superpotential coupling supercolor to
ordinary fields exists. Reviewing the theories which have been shown (sects. 3, 4) to
exhibit DSB, we see that all of these theories possess global symmetries, some of
which may be gauged and identified with a subset of the standard SU(3) x SU(2) x
U(1), or possibly some additional gauge quantum number which may also be carried
by quarks and leptons (B - L). However, examining these theories carefully, we see
that there are a number of problems which such an approach must face from the
outset. All of these problems stem from the fact that the candidate supercolor
theories are chiral.
Before studying specific examples, which illustrate a n d / o r solve these difficulties,
it will be helpful to list the most serious of these problems here.
(i) Anomalies. Frequently the global symmetries of the supercoior sector which
one wishes to gauge possess Adler-Bell-Jackiw or Witten [41] anomalies. These can
592 I. Affleck et al. / l)vnamical ~'uper.~ymmetrv breaking

be dealt with in a variety of ways. In the case that the symmetry which one wishes to
gauge is a subset of S U ( 3 ) x S U ( 2 ) x U(1), there are various possibilities: add
supercolor-neutral fields with quantum numbers which cancel the anomalies; if the
symmetry is unbroken by the supercolor interactions, identify the corresponding
massless fermions of the supercolor spectrum with some ordinary fermions (if
possible) so that the anomalies automatically cancel; take for the supercolor sector
two copies of one of the supersymmetry-breaking theories, and assign opposite
q u a n t u m numbers to the fields in each sector. In the case where one gauges some
other symmetry (e.g. B - L, see below), it might be possible to cancel anomalies
from the outset by appropriate choice of quantum numbers. All of these possibilities
will be illustrated shortly.
(ii) R-invariance. A much more serious problem is that all the candidate super-
color theories we now know possess a spontaneously broken R-invariance with no
supercolor anomalies. This symmetry must be explicitly broken by a large amount if
one is to avoid a dangerous axion or Goldstone boson. Because of the chiral nature
of these theories, it is very difficult, even if additional fields are added, to write
couplings which break these extra symmetries, if we restrict ourselves to dimension-
less couplings. We see, at present, only two ways to avoid this problem. One is to
introduce dimensionful couplings. A more appealing approach is to introduce some
additional gauge interactions ("R-color") which, while not breaking supersymmetry
by themselves, do provide an explicit breaking of R-invariance. This approach also
requires the introduction of a gauge-singlet field into the theory. As we will see,
" R - c o l o r " has the additional virtue that it tends to induce S U ( 2 ) × U(1) breakdown.
(iii) Massless fermions. Most of the theories which we have considered have
massless fermions in their spectra, which are necessary in order to satisfy 't Hooft
anomaly conditions [17]. If one wants to gauge the corresponding symmetry, in such
a way that some of these fermions carry electric charge, one must find some
mechanism by which these fields can gain mass. Again, because of the chiral nature
of the underlying theory, it is difficult to break all of the unwanted chiral symme-
tries. Some examples where this breaking is achieved are described below.
(iv) Light gluinos versus asymptotic freedom. Many of the candidate supercolor
theories do not possess a global symmetry as large as SU(3). Thus one can gauge at
most a U(1) or SU(2) group. As a result, the gluino typically does not gain mass
until rather high order (typically three-loop order) and is simply too light (1-10
MeV). An exception will be described below. Alternatively, one can consider
supercolor theories which possess a global SU(3) symmetry. An example comes from
the class of S U ( N ) models with N - 4 fundamentals and an antisymmetric tensor. If
N = 15 (or larger) and the superpotential is chosen appropriately, one can gauge an
SU(3) x S U ( 2 ) x U(I) subgroup. However, above the SU(15) scale, the SU(3) /3-
function becomes violently non-asymptotically free, and the Q C D coupling becomes
large within a few (typically three) decades. As a result, some type of early
unification will be required if such theories are to work. Clearly this preunified
I. Affleck et al. / Dynarnical super.~vmmet O, breaking 593

theory must be baryon-number conserving. Finding such a model will be a quite


non-trivial group-theoretic exercise.
(v) Higgs fields and SU(2) × U(1) breaking. The theory must possess a mechanism
for S U ( 2 ) × U(1) breaking. The most elegant possibility is that the supercolor sector
itself breaks this symmetry. However, since scalar quark and lepton masses arise
through loop graphs, and are thus suppressed by powers of a relative to the
supercolor scale, they will tend to be too small if this scale is of order the weak scale.
Alternatively, as mentioned above, the additional interactions required to break
R-invariance may induce the breaking of SU(2)x U(I). This possibility will be
discussed in some detail below. Finally, it may be that in some theories the gauge
field contribution to the Higgs potential, has a minimum away from the origin. We
will not discuss this possibility in detail here.
(vi) Scalar quark and lepton masses. Obviously, these must be positive. There is
no simple general argument for their sign, and they must be calculated on a case by
case basis, as in ordinary theories [38]. In theories with strongly-interacting super-
color sectors, these masses cannot be calculated with current theoretical technology.
In theories such as that of sect. 4, they can be calculated in perturbation theory.
(vii) Peccei-Quinn symmetry. The singlet field a n d / o r additional gauge fields
mentioned in points (ii) and (v) are almost certainly required for another reason. For
without them the theory possesses a symmetry of precisely the type proposed by
Peccei and Quinn, with its resulting axion (or Goldstone boson) [43, 2].
In order to illustrate these problems and their possible solutions, let us first take
as an example the SU(3) × SU(2) theory of sect. 4. Neither the SU(3) nor the SU(2)
can be identified with ordinary gauge symmetries, since the resulting theory would
have flat directions. The theory does have a global U(1), which we might wish to
gauge and identify with hypercharge. In order to cancel anomalies, we can assign the
U(1) charges to the supercolor sector, and introduce spectators (additional fields
with U(1) charges chosen to cancel anomalies). The charges must be chosen so that
all of the fermions of the theory can gain mass. In particular, it must be possible for
the massless Weyl fermion of the supercolor sector to pair with one of the spectators,
and for the rest to pair among themselves. A much more elegant possibility is the
following. We can assign the U(1) v charges so that the massless fermion of the
supercolor sector has the charge of (say) the positron, and omit the fundamental
positron. Then the anomaly cancellation is automatic, and it seems in principle
possible for the fermion to gain a mass by pairing with the electron. But, it is in fact
impossible for this fermion to gain a mass, however we choose to cancel the
anomaly, for it is still protected by a chiral symmetry. This symmetry is just the one
which kept it massless before the weaking gauging. The gauging has not broken this
symmetry, and no renormalizable coupling which we can write down will break it.
This model also illustrates clearly the other problems listed above. The supercolor
sector of the theory contains a spontaneously broken, non-anomalous R-invariance.
The theory also possesses a "Peccei-Quinn" symmetry [42], an ordinary global U(1)
594 I. Affleck et al. / Dvnamu'al .;upersymmetry hreaking

under which the Higgs fields rotate with the same phase. This latter symmetry can be
broken by including a singlet field, S, with couplings X I S H H + X2S 3. However, the
theory, in this case, still possesses an R-symmetry with ordinary color SU(3)
anomalies. As a result, the theory produces an unacceptable axion [43,2]. This
problem could be solved by introducing a term ~ZS, but, as explained above, the
presence of such a small, dimensionfui coupling appears highly unnatural.
T o solve this problem, we can add to the theory another set of gauge interactions,
for which the R-symmetry is anomalous and with a scale much larger than A OCD (SO
that the axion receives a large mass). Consider, for example, adding to the theory an
additional gauge interaction, SU(3)R, and a set of fields, 0 " , transforming as an
octet under this symmetry. We now include in the superpotential

W R = ~ I S H H + ~,2 S3 + g ' S O 2 + gO 3. (6.8)

Consider, first, g ' = 0. The SU(3)R theory has no flat directions. It also has no
non-anomalous global symmetries. Following the reasoning of sect. 2, it is a strongly
coupled theory which we expect to leave supersymmetry unbroken. The spectrum
should consist only of massive excitations, with masses of order the scale of the
SU(3)R theory, A R, No qualitative change is expected to occur once g' is turned on,
in particular, no flat directions appear in the R-color sector.
The would-be axion of the supercolor can now gain mass. Also, the dynamics of
the R-color sector can now generate S U ( 2 ) x U ( 1 ) breaking. In particular, the
operator 0 z almost certainly obtains a v.e.v, of order A~. Thus, at scales below 3, R,
the effective superpotential obtained by integrating out R-colored fields is

Weff = ~ I S H H .4- ~)k233 + ~2S, (6.9)

where p 2 = g,(O2). This superpotential possesses two sets of minima, one with
S 2 = ~2/~. 2, H = H = 0, and one in which H H = tt2/~.l and SU(2) x U(1) is broken
to U(1). Of course, the effective lagrangian also contains terms (arising from
integrating out the supercolor interactions) which are supersymmetry violating.
These will lift the degeneracy between states. Moreover, once the theory is coupled
to supergravity, there are additional terms in the potential which lift the degeneracy.
As Weinberg has stressed [44], even if the desired vacuum is not the lowest energy
state, it may well be stable against tunneling. As in all supersymmetric models, we
will simply assume that the desired minimum is somehow selected. Thus in addition
to giving mass to the axiom the R-color interactions provide a mechanism for
electroweak symmetry-breaking.
Before considering the supersymmetry breaking in the Higgs sector arising from
supercolor interactions, let us estimate the mass of the axion. It may he calculated
using Dashen's formula. For simplicity, suppose that the Higgs v.e.v.'s are both
equal to ~t. Then, if we call the R-symmetry the symmetry which is non-anomalous
I. Affleck et aL / Dynamical super,wmmett T breaking 595

A/~ @ ,w,~
4' 4' + -4,---

4,
+ ___

4, ".$s." 4,
Fig. 2. Some diagrams contributing to scalar quark and lepton masses in supersymmetric technicolor.
Solid lines denote fermions, wavy lines gauge bosons, and dashed lines scalars. The subscript sc denotes
supercolor. Gauge bosons are generically called A~. M w, h are gauginos of SU(31. SU(2) and U(1)
respectively.

in the supercolor sector, and under which all other fields carry R-charge ~, the
corresponding current does not create the U(1)r Higgs boson from vacuum. Using
Wett, Dashen [45] then instructs us that the axion mass, m A, satisfies

m a2 f A = ~<p.2()~$2 + )~,HH)> + c.c., (6.10)

where fA is the axion decay constant, of order A ~,..


For reasonable values of the parameters, the axion mass is of order 3-100 MeV.
Such a particle would not have been detected in beam dump experiments, because
A~: is so large. It would not be produced in stars, because of its large mass. Thus the
R-color interactions can solve both the problem of SU(2) × U(1) breaking and the
axion problem of these models. Note, however, that these theories still have a strong
CP problem. This, presumably, requires an "invisible axion" [46].
In order that this picture for SU(2)× U(1) breaking not be upset by radiative
corrections coming from the supercoior sector, the supercolor scale cannot be too
large. The graphs of fig. 2 give masses to scalar quarks and leptons, and to Higgs
fields, proportional to the square of the hypercharge. It is, of course, necessary that
these contributions be positive; experience with explicit models indicates that this is
frequently the case [38]. (In the SU(3) × SU(2) case, it is a straightforward matter to
calculate these masses explicitly, but we will not attempt this here.) They must be
large enough that the lightest charged scalars have not yet been observed. On the
other hand, the contributions to Higgs masses should not be tOO large; otherwise
S U ( 2 ) × U(1) breaking could arise only through fine-tuning. Thus (a~/4~')A,~
should be of order the weak scale or smaller.
Apart from the problem of the massless charged fermion, this theory has another
very serious failing: the gluino is far too light. While we saw that scalar masses-squared
596 I. Affleck et aL / D v n a m t c a l super.wmmetrv hreaktng

Fig. 3. Diagram contributing to gluino mass in supcrsymmetric technicolor.

arise at two loops, the gluino mass would only arise at three loops (fig. 3), and would
be very tiny. These two problems insure that this theory cannot be realistic.
It is natural to ask to what extent these problems are generic, and what strategies
might be employed to overcome them. Let us first consider the problem of the
massless charged fermion. One possibility is to gauge a U(1) which is spontaneously
broken. For such a U(I), one does not expect a massless fermion, and the corre-
sponding Goldstone boson is "higgsed." An example of this type will be discussed
below, when we present a potentially realistic model.
A second possibility arises if the unbroken, gauged U(1) is a subgroup of a larger,
non-abelian group. Then, in some cases, there is no U(I) symmetry of the full theory
which protects the massless fermion of the supercolor sector. A rather baroque
example of this type is the following, in ref. [10], we showed that an SU(5) theory
with two generations and a superpotentiai breaks supersymmetry. We were able to
compute the effective potential for this theory, but we did not obtain its minimum. It
appeared likely that the global SU(2) x U(1) symmetry of that theory was sponta-
neously broken to U(1). (We could not, without explicitly minimizing the potential,
rule out the possibility that the U(1) was broken.) Let us assume that this is the case.
If this is so, the theory possesses a massless fermion, charged under the unbroken
U(1). Now consider a supercolor sector which consists of two copies of this SU(5)
theory. A non-anomalous S U ( 2 ) x U(1) symmetry can be weakly gauged, provided
we take the U(I) charges opposite in the two sectors. No symmetry now prevents the
massless charged fermions from pairing to acquire a Dirac mass; this mass need not
be exceptionally small.
Of course, it would be quite elegant if we could identify the SU(2) x U(1) here
with the SU(2) × U(1) of Weinberg-Salam. There is, however, a problem with this
identification. The masses of scalar quarks and leptons will almost certainly be too
small in such a theory. These will arise from one-loop graphs. The lightest masses
will be of order ( a / 4 c r ) l / 2 M w. The gluino mass would arise at two loops in such a
theory, and would be very small. The scalars could be heavier if we did not identify
the SU(2) of the supercolor sector with ordinary SU(2)L (while still identifying the
U(1) with hypercharge) but we would still face the problem of the light gluino.
L Affleck et al. / l)vnamicalsupersvmmetry breaking 597

Thus the problem of light gluinos emerges as one of the most severe in supersym-
metric technicolor. Clearly one way to avoid this problem is to take a supercolor
sector which has at least an SU(3) global symmetry which may be gauged. An
example of a theory with an SU(3)× SU(2)× U(1) global symmetry is an SU(15)
gauge theory with 11 ff's and an antisymmetric tensor. This model is in the class
studied in sect. 2. The ,K's are taken to transform as (3,1).,,: (3,1)_,.: (1,2),. 2.
(1,2)_v 2, (1,1)o, under the S U ( 3 ) × S U ( 2 ) × U(1), with Yl and )'2 chosen conve-
niently. The superpotential is taken as the most general one allowed by the gauge
symmetries. As discussed in sect. 3, this theory has no classical flat directions and
breaks supersymmetry. Before worrying about finer details of the resulting model
(patterns of symmetry breaking, etc.) we can see an immediate problem. Above the
supercolor scale, the B-function of QCD is large and positive: 12 g3/16~r2 for three
generations of quarks and leptons. If we take 10 TeV for the supercolor scale, we
find that a ~ - 1 for scales of order 10 4 TeV. Thus some sort of early unification
must be invoked. This unification must involve very large groups, and must keep the
proton stable. This may be possible, but we have not yet explored this idea
thoroughly. It will, of course, be necessary to study the patterns of symmetry
breaking in these theories, and to deal with possible light fermions and SU(2) × U(1)
breaking. In particular, it may be necessary to study even larger gauge groups to
obtain a realistic model.
Finally, we consider a somewhat different approach to model building, which
leads to a theory which is potentially realistic. We consider the possibility of weakly
gauging a U(1) symmetry which is spontaneously broken. Acting on ordinary quarks
and leptons, this symmetry might be B - L. Of course, the ordinary, g l o b a l . B - L
should not be spontaneously broken, but this is trivially arranged. A candidate
supercolor theory which may have a spontaneously broken U(1) is the SU(5) theory
with a single generation studied in ref. [9]. Because that theory is strongly coupled, it
was not possible to establish whether or not the ordinary (non-R) global U(I)
symmetry is spontaneously broken. (We referred to this as the A-symmetry.) In
particular, there is a simple solution to the 't Hooft anomaly conditions for
unbroken U(1)A, and simple order parameters which could describe the breaking of
U(1) A. Let us assume that U(1)A is spontaneously broken, and take this model for
our supercolor sector. We can identify A with B - L , and weakly gauge the
symmetry provided the A-charges are chosen to cancel the (B - L) 3 anomaly. Also,
we must add an R-color sector, as discussed above, to break SU(2)× U(1) and
explicitly break the R-symmetry.
Such a theory has a good chance of being realistic. Scalar quark and lepton masses
will arise at one loop. Contributions to Higgs masses, however, are highly sup-
pressed, since Higgs fields do not carry B - L. If one ignores Yukawa couplings,
they arise only at three-loop order; if one allows the suppression due to Yukawa
couplings, they arise at two-loop order. As a result, A~ may be taken quite large.
598 I. Affleck et al. / Dvnamical.~upersvmmetrv #real~mg

This is good, because gluino (and photino) masses arise only at two-loop order. It
thus appears that these masses can be large enough.
Thus, while it is not possible to compute in this theory, it appears that such a
model can be realistic. If one had a weakly interacting theory (such as the
SU(3) x SU(2) theory) with such a U(1), one could actually compute all of the
masses. Such a computation, while tedious, would be possible in the two-generation
SU(5) model if the global SU(2) × U(1) is completely broken in that theory.
In sum, it appears that chiral theories with dynamical supersymmetry-breaking are
excellent candidates for the hidden sectors of N = 1 supergravity theories. It is also
possible that supersymmetry might be dynamically broken at a "'nearby" scale.
However, considerably more work is required to provide an "existence p r o o f ' for
this possibility, and an elegant model is not yet in hand.

We thank T. Banks, M. Peskin, G. Veneziano and E. Witten for fruitful discus-


sions. We are grateful to E. Witten and M. Peskin for communicating to us several
unpublished results. The work of I.A. was supported in part by the National Science
Foundation through grant no. PHY-80-19754 and an A. P. Sloan Foundation
fellowship; that of N.S. in part by the US Department of Energy through grant no.
DE-AC02-76ER02220; that of M.D. in part by the National Science Foundation,
grant no. PHY-82-17352.

References
[1] E. Witten, Nucl. Phys. B188 (1981) 513
[2] M. Dine, W. Fischler and M. Srednicki, Nucl. Phys. B189 (1981) 575:
S. Dimopoulos and S. Raby, Nucl. Phys. B192 (1981) 353
13] E. Witten, Nucl. Phys. B202 (19821 253
[4] G. Veneziano and S. Yankielowicz, Phys. Left. 113B (1982) 321:
T.R. Taylor, G. Veneziano and S. Yankielowicz, Nucl. Phys. B218 (1983) 493
[5] M.E. Peskin, in Problems in unification and supergravity, e d (i. Farrar and F. Henyey (AIP, New
York, 1984)
[6] A.C. Davis, M. Dine and N. Seiberg, Phys. Lett. 125B (19831 487
[7] I. Affleck, M. Dine and N. Sciberg, Phys. Rev. Lett. 51 (1983) 1026
[8] I. Affleck, M. Dine and N. Seiberg, Nucl. Phys. B241 (1984) 493
[9] I. Affleck, M. Dine and N. Sciberg, Phys. Lett. 137B (1984) 187
[10] I. Affleck, M. Dine and N. Seiberg, Phys. Rev. l.ett. 52 (1984) 1677
[11] I. Affleck, M. Dine and N. Sciberg, Phys. Lett. 140B (1984) 59
[12] (i.C. Rossi and G. Veneziano, Phys. Lett. 138B (1984) 195"
D. Amati, G.C. Rossi and (i. Veneziano, Nucl. Phys. B249 (1985) 1
[13] Y. Meurice and (i. Veneziano, Phys. Lett. 141B (1984) 69
[141 V.A. Novikov, M.A. Shifman, A.I. Vainshtcin and V.I. Zakharov, Nucl. Phys. B223 (1983) 445:B249
(1985) 445
[15] M.J. Grisaru, W. Siegel and M. Ro6ck, Nucl. Phys. B159 (1979) 429
[16] W. Fischler, H.P. Nilles, J. Polchinski, S. Raby and L. Susskind, Phys. Rcv. Lett. 47 (19811 757
[17] G. "t Hooft, in Recent developments in gauge theories, ed. G. "t Hooft ct al. (Plenum, Nev, York,
1980);
Y. Frishman, A. Schwimmcr, T. Banks and S. Yankielowicz, Nucl. Phys. 1]177 (1981) 157:
S. Coleman and B. Grossman, Nucl. Phys. B203 (1982) 205
1. Affleck et al. / Dynamical super.wmmetry breaking 599

[18] V.A. Novikov, M.A. Shifman, A.I. Vainshtein, M.B. Voloshin and V.I. Zakharov. Nucl. Phys. B229
(1983) 394;
V.A. Novikov, M.A. Shifman, A.I. Vainshtein and V.I. Zakharov, Nucl. Phys. B229 (1983) 381,407
[19] E. Witten, private communication
[20] H.P. Nilles, Nucl. Phys. B217 (1983) 366:
R. Arnowitt, A. Chamsedine and P. Nath, Phys. Rev. Lett. 49 (1982) 070:
L. Ibahez, Phys. Rev. Lett. l18B (1982) 73;
R. Barbieri, S. Ferrara and C. Savoy, Phys. Lett. l19B (1982) 343
[21] L. Alvarez-Gaume, M. Wise and J. Polchinski, Nucl. Phys. B221 (1983) 495:
L. Hall, J. Lykken and S. Weinberg, Phys. Rev. D27 (1983) 2359
[22] I. Affleck and M. Dine, Nucl. Phys. B249 (1985) 361
[23] (i.D. Coughlan, W. Fischler, E. Kolb, S. Raby and G.G. Ross, Phys. Lett. 131B (1983) 59:
M. Dine, W. Fischler and D. Nemeschansky, Phys. Lett. 136B (1984) 169
[24] F. Buccclla, J.-P. Derendinger, C.A. Savoy and S. Ferrara, CERN preprint TIt 3212
[25] M. Peskin, private communication
[26] G. 't Hooft, in Recent developments in gauge theories, ed. (1. 't Hooft et al. (Plenum, New York,
1980):
T. Banks and E. Rabinovici, Nucl. Phys. BI60 (1979) 349;
E. Fradkin and S. Shenker, Phys. Rev. DI9 (1979) 3682:
S. Dimopoulos, S. Raby and L. Susskind, Nucl. Phys. B173 (1980) 208
[27] J. Polchinski, Phys. Rev. D26 (1982) 3674
[28] B. Zumino, Phys. Lett. 87B (1979) 203
[29] J. Bagger and E. Witten, Phys. Lett. l15B (1982) 202:
E. Witten, private communication
[30] D. Nemeschansky and R. Rhome, Nucl. Phys. B249 (1985) 157
[31] B. Ovrut and J. Wess, Phys. Re','. D25 (1982) 409
[32] K. Konishi, Phys. Lett. 135B (1984) 439
[33] G. 't Hooft, Phys. Rev. D14 (1976) 3432
[34] S. Dimopoulos and H. Georgi, Nucl. Phys. B193 (1981) 150
[35] I. Affleck and M. Dine, Institute for Advanced Study preprint (1984)
[361 E. Witten, unpublished
[37] R. Barbieri, S. Ferrara, D.V. Nanopoulos and K.S. Stelle, Phys. Lett. l13B (1982) 218:
J. Bagger, Nucl. Phys. B211 (1983) 302
[38] E. Witten, Phys. Left. 105B (1981) 267:
T. Banks and V. Kaplunovsky, Nucl. Phys. B206 (1982) 405:
M. Dine and W. Fischler, Phys. Lett. ll0B (1982) 227: Nucl. Phys. B204 (1982) 346:
L. Alvarez-Gaum6, M. Claudson and M. Wise, Nucl. Phys. B207 (1982) 96:
C. Nappi and B.A. Ovrut, Phys. Lett. l13B (1982) 175
[39] E. Cremmer, B. Julia, J. Scherk, S. Ferrara, L. Girardello and P. van Niewenhuizcn, Nucl. Phys. B147
(1979) 105:
R. Barbieri, S. Ferrara, D.V. Nanopoulos and K.S. Stelle, Phys. Lett. lI3B (1982) 2181
J. Bagger and E. Witten, Phys. Lett. 115B (1982) 202
[40] W. Buchmuller and D. Wyler, Phys. Lett. 121B (1983) 321:
F. del Aguila, M. Gavela, J. Grifols and A. Mendez, Phys. Lett. 126B (1983) 71:
J. Polchinski and M. Wise, Phys. Lett. 125B (1983) 393
[41] E. Witten, Phys. Lett. l17B (1982) 324
[42] R.D. Peccei and HR. Quinn. Phys. Rev. Lett. 38 (1977) 1,:140: Phys. Rev. I)16 (1977) 1791
[43] S. Weinberg, Phys. Rev. Lett. 40 (1978) 223:
F. Wilczek, Phys. Rev. Lett. 40 (1978) 279
[44] S. Weinberg, Phys. Rev. Lett. 48 (1982) 1776
[45] R. Dashen, Phys. Rev. 183 (1969) 1245
[46] M. Dine, W. Fischler and M. Srednicki, Phys. Rev. Lett. 104B (1981) 199;
J. Kim, Phys. Rev. Left 43 (1979) 103
[47] S. Ferrara, L. Girardello and H.P. Nilles, Phys. Lett. 125B (1983) 457

You might also like