You are on page 1of 5

Volume 96B, number 1,2 PHYSICS LETTERS 20 October 1980

THREE-LOOP FINITENESS OF THE N = 4 SUPERSYMMETRIC NON-LINEAR a MODEL

M. RO(~EK
DAMTP, University of Cambridge, UK
and
P.K. TOWNSEND
CERN, Geneva, Switzerland

Received 14 August 1980

We formulate theN = 4 supersymmetric non-linear a model (in two spacetime dimensions) in terms of N= 1 four-dimen-
sional superfields, and give the explicit form of all supersymmetry transformations. Using supergraph techniques we show
that all divergent contr~utions to the two-point function vanish up to the three-loop level. We also show finiteness to lead-
ing order in the 1/n expansion.

1. Introduction and results. Non-linear o models two-loop finite. It is important to investigate whether
are the field theorist's favorite two-dimensional play- finiteness persists in higher orders, and from elegant
ground for the investigation of properties conjectured general considerations it has been argued [4] that this
to hold in more physical, and more complicated, four- is the case. We start our investigation of this problem
dimensional theories. Supersymmetric theories are by formulating the model in terms o f N = 1 four-di-
known to have improved ultra-violet behaviour over mensional superfields, which by dimensional reduc-
their non-supersymmetric counterparts, in some cases tion are equivalent to N = 2 two-dimensional super-
rendering the theory finite, at least for the leading or- fields. (Unless otherwise stated, our subsequent discus-
ders of perturbation theory [1,2]. The supersymmetric sion is four dimensional. In particular, the four two-
non-linear o models [3] are therefore natural two-di- dimensional supersymmetries will be referred to as two
mensional candidates to study in this regard, and have four-dimensional supersymmetries.) Because our for-
the advantage of being considerably simpler than four- mulation is in superspace one supersymmetry is mani-
dimensional theories. Freedman and Alvarez-Gaum6 fest. The other one we give explicitly; it appears in a
have recently studied the ultra-violet behaviour of gen- supermultiplet of transformations that includes the
eral supersymmetric a models (where the Bose fields non-holomorphic U(1) isometry of ref. [6]. We show,
are the coordinates of an arbitrary riemannian mani- using supergraph techniques [7], the finiteness of the
fold) with striking results [4]. In particular, those mod- theory to leading order in 1 In, and independently, for
els defined on Ricci flat manifolds are finite to two n = 2, the three-loop finiteness of the two-point func-
loops, as a consequence of general arguments and as tion.
verified by explicit calculation. Perhaps the most in-
teresting supersymmetric o models are those with an 2. Superspace formulation. Consider the following
extended N = 4 supersymmetry (N = 2 in four dimen- lagrangian density:
sions); like the N = 4 super-Yang-Mills and the N = 8
supergravity such theories are self-conjugate. Remark- .~(x) = fd40 (e v~+q~+ + e- v~_~_)
ably, the only known example [5] of such an N = 4
supersymmetric o model is defined on a manifold with
vanishing Ricci tensor [6], and therefore is one- and
+ f d 2 0 [S(q~+4~_ + ib) + h.c.]. (2.1)

72
Volume 96B, number 1,2 PHYSICS LETTERS 20 October 1980

V is a real (N = 1, four-dimensional) superfield, q~+and To make contact with the work of ref. [5] we
¢_ are 2n independent chiral superfields satisfying make the following definitions:
Dk ~b+ = D& q~_ = 0, with Do, D& the usual two-com-
ponent spinor covariant derivatives, and ~+ and ~_ ~1 = ~+1o =o, ¢2 = ~-Io =o,
are the complex conjugate antichiral superfields. S is
a Lagrange multiplier chiral superfield satisfying D&S ~1 = ~+10=0, ~2 = ~b-10=0,
= Da~q= 0 and n is the number of components of qS+
and q~_ which can be chosen to be in a representation Vu=DouDVIo= o, M + iN=SIo=o,
of SU(N) such that (2.1) is SU(N) invariant. In addi-
tion, (2.1) is invariant under the following gauge trans- = (D~+ + D~b_)[ 0=0. (2.7)
formation with chiral parameter A(x, 0): Then, performing the 0 integrations, eliminating all
fields not defined in (2.7) (Lagrange multipliers and
8 V = i(A - ~.), 8S = 0,
auxiliary fields), we find the lagrangian of ref. [5 ] [eq.
(7)] with its constraints [eqs. (4) and (5)]. In partic-
6~b+ = - iA~+, 8~+ = iTkq~+,
ular, in the constraint ~i~i/qa/= b, b is found to have
the specific direction (0, 2b, 0), as chosen in ref. [6].
6~_ = iAq~_, 8~_ = - i A ~ _ . (2.2)
The form of the lagrangian (2.1) allows us to study
The invariance of the action under one supersym- easily the leading order in the 1/n expression. The ef-
merry is manifest, but there is also a second super- fective action F 0 + Pl to leading order in 1/n is given
symmetry which is contained in the following trans- by the tree graphs, F0, and the one-q~-loop graphs with
formations: arbitrary S and V external lines, Pl" The Feynman
6q~+ = iDZ(e- V ~ _ ) , 8 V = 4 i ( e g - ~S), rules are standard and we refer the reader to, e.g.,
ref. [7]. From the form of the lagrangian I"1 is finite
&b_ = -iD2 (eV~ q~+), ~S = 1 i~2(De DV), if it is S or S dependent. To see that F I ( V ) is finite
we observe that a general graph, as in fig.2.1a, is fi-
8q~+ = - i D 2 ( e - Veq~_), 6S---~-1 iD2(D~ DV), nite b~¢ power counting (in two dimensions) unless all
the D ~ and ~ 2 factors at the vertices are brought to-
8~_ = iD 2 (eVe~b+). (2.3) gether by partial integration to give the maximal num-
ber o f p 2 factors in the numerator, according to the
In order that these transformations leave the ac- formulae
tion invariant the complex superparameter e must
satisfy D2~ZD 2 = _ p Z D 2 ' ~2D2~2 = _p2~2. (2.8)

DZe + ~2g = 0, [)De = DUe = 0, (2.4a, b) This means that'effectively all but one propagator are
cancelled, so that a general graph with n propagators
whose general solution is is contracted to a graph with a single propagator times
e=e0+el(X), De0=De0=0, Del(x)=0. (2.5) a factor of ( - 1 ) n - 1 . Up to a common factor of fd2p/
p2 this is equivalent to taking all ~b~ propagators to be
That is, e is the sum of a constant chiral superfield e 0 - 1 ; thus the infinite contribution to the effective ac-
and an arbitrary antichiral superfield e 1 (x). The ac- tion is described by the following reduced generating
tion of (2.3) restricted to e = e I (x) reduces to a gauge function:
transformation when the field equations are satisfied.
This "off-shell central charge" [8] is not relevant to
our investigations and will not be discussed further.
However, it can be used to remove the 0-independent
part of e 0 which then takes the form
e0 = 0n X~ + i02f, (2.6) V2 V2 2 3

where, as a consequence of (2.4a), f i s real. Fig. 2.1. A typical contr~ution to r l ( v ) .

73
Volume 96B, number 1,2 PHYSICS LETTERS 20 October 1980

exp ( - F 1 (V)ldiv) 6u = ~2 [/3~0(0 _ u)], 6v = ~2 [~-1~0(o _ if)], (3.5)

=Nfd~b+ dq~_ de+ de_ exp [ - ( e v~+q~+ + e - vqs_ q~_)l where/3 and ~ are given by

= d e t - n (e v) det - n (e- V) = 1, (2.9) ~(l+u'ff)(l+u'~)~/2-1+u'u~ -1. (3.6)


/3=[_(1+o.0) ( l + f f - l+v'0
where N is an appropriately chosen normalization fac- The superparameter e 0 is given in eq. (2.6). The spinor
tor. It follows that the divergent part of I"1 (V) vanishes. X~ is the second supersymmetry parameter. To see the
significance of the remaining real scalar parameter f we
3. Unconstrained superfieMs. To proceed we now set X = 0 and then take the 0 = 0 part of (3.5) to get
eliminate S and V from (2.1). V is eliminated by its
field equation 6u[0= 0 = i f ~ ( u - u)10= 0 ,
1 1
V - ~ ln(q~_ ~b_) - ~- in (q3+$+), (3.1) 6ol0=0 = i f ~ - l ( v - u)[0=0 " (3.7)
while S is eliminated by resolving the constraint q~+qS_ This we recognize as the non-holomorphic U(1) isome-
+ ib = 0. This we do by introducing 2(n - 1 ) chiral try of ref. [6]. The origin of this previously rather mys-
superfields u and v, as follows: terious symmetry is now explained. The commutator
of the first (manifest) supersymmetry with this U(1)
q ~ + = ( l + u . o ) -1/2 (u,1), ¢ =(l +u.o) -1/2(o,1)
symmetry gives the second supersymmetry; that i s , f
(3.2) is the parameter for the rotation of one supersymmetry
(We have chosen b = 1 for convenience.) This not only generator into the other. From this observation we see
solves the constraint but also fixes the gauge for the in- that the additional U(1) symmetry is essential for the
variance of (2.2). In terms of u and v and the antichiral full N = 2 supersymmetry. All the other isometrics of
tT, ~, the lagrangian now takes the form ref. [6] can be directly carried over to superfields, be-
(1 + u.ff)(1 + o" fi)~]l/2 cause, being holomorphic, they respect the chirality of
~°=2f d40 O+u'v)(l~'~J u and o. In fact, the notion of a holomorphic function
in component language is directly related to the notion
of a chiral superfield in superspace.
- 2 f d 4 0 F(u, if, v, 0). (3.3)
Zumino has shown [9] that a lagrangian of this form 4. Superloops. From now on, for simplicity, we will
with any F will describe an N = 1 (N = 2 in two dimen- restrict the lagrangian (3.3) to depend only on u, v
sions) supersymmetric o model defined on a K/ihler with single components, i.e., n (of section 2) = 2. Ex-
manifold with K~ihler potential F. The particular poten- panded in powers of u, v the lagrangian is
tial of (3.3) was found previously [6] as the K~_hler po- 4u~o~ + 2u2ffo
tential for the N = 2 supersymmetric model, so the
~ = uff +vO - z1 (u2ff 2 + 02~2
above superspace form o f this model is not surprising. + 2fi2uO + 2020u + 202 oti)
But only for a special form o f F can we expect to have
N = 2 supersymmetry as this is not guaranteed by the + 1 (u 3t23 + 0303 + u2/~ 200 + v202uu
general form with arbitrary F. We can find the second
supersymmetry by expressing the transformations o f -- 5u202K0 _ 5t~202uv + 3u3ffo 2 + 3uff302 + 3o30u 2
eq. (2.3) in terms of u and o. But because eqs. (3.2) al-
so fix the gauge we must simultaneously perform a + 3003~ 2 + u 3 t ~ 2 +uo302 +0ff3u 2 +t703u 2)
compensating gauge transformation to assure that we
remain in the same gauge. In fact, the transformation + 1 (7U2ff2V202 _~U4tj 4 _ ~V4~4 + . . . ) + . . . .
6 = 6sup(e0) + 6gauge{A = - i D 2 [½(/3 + ~ - 1 ) g] } (3.4)
(4.1)
is compatible with (3.2) and this transformation
acting on u, v is

74
Volume 96B, number 1,2 PHYSICS LETTERS 20 October 1980

where all purely chiral or antichiral pieces have been On partial integration it may be possible that a D 2 fac-
dropped. We will now examine the infinite corrections tor acts on an external line in which case the propaga-
at one, two and three loops to (uff), the u~ two-point tor cancellation does not take place. But in this case
function. The v~ two-point function is trivially the the resulting diagram is finite (in two dimensions!).
same while all other two-point functions vanish. The Thus fig. 4.3a can be reduced to the form o f fig. 4.3b;
Feynman rules are given in fig. 4.1. fig. 4.3c vanishes b y virtue o f the previous one-loop
At one loop we have the diagram shown in fig. 4.2. result. Simple combinatorics is then all that is required
There is a direct cancellation between u and o and the to show that the sum of infinite contributions vanishes.
entire diagram vanishes. At two loops we have the dia- At three loops the contributions to the uti two-
grams shown in fig. 4.3. In the first o f these, (a), we point function are shown in figs. 2L5. Those o f fig.
can integrate b y parts a factor o f D 2 or ~2 to get a 4.5e vanish as a consequence of the previous one-loop
factor o f _ p 2 , as in eq. (2.8). This cancels one propa- result. Upon collapsing one propagator in fig. 4.5a it is
gator, leading to the collapse o f this diagram to one o f reduced either to the form o f fig. 4.5b or 4.5c. Fig.
a different form, as shown in the sequence of fig. 4.4. 4.5b can be similarly reduced to the form of fig. 4.5d
but in this case we must keep the contributions that
come from partial integration of spinor derivatives on-
u,v ~~ 84(0j- Oz) to external lines because o f the subdivergence at one
, 'p 2
vertex. These one-loop subdivergences cancel against
u -2 u" other one-loop subdivergences coming from collapsing
-I the second propagator in fig. 4.5a. This is a consequence
o f the one-loop finiteness of the four-point functions,
as shown in fig. 4.6. Fig. 4.5c can, b y partial integra-
Fig. 4.1. Feynman rules for the lagrangian (4.1). Only one ver- tion of spinor derivatives, be reduced either to the
tex is shown. The others follow in similar fashion with the
coupling constant as determined in the standard way from the
lagrangian (e.g., - 1 for the vertex shown). External lines have
no D2 or ~2 factors.

u ~ + u 0 =0
k_J
Fig. 4.2. Vanishing of one-loop contributions to <uE>. ©
u
a) b) c)
Fig. 4.3. Two-loop contr~utions to (uff).

+ O kj
( ~ +finite Fig. 4.5. Three-loop contr~utions to <uff>.
terms

+ finffe + ~<CO
terms

Fig. 4.4. Collapsing a propagator. Fig. 4.6. One-loop finiteness of four-point functions.

75
Volume 96B, number 1,2 PHYSICS LETTERS 20 October 1980

@x(k+p )z super-Yang-Mills theory [2]. In particular, the lagran-


gian in the form of (2.1) and the transformation laws
of (2.3) are strikingly similar. If indeed the supersym-
metric o model is completely finite, a hope that is en-
x(k+p )Z=2k.p x couraged b y our results, then a proof to all orders in
perturbation theory in this formulation may be gen-
eralizable to the N = 4 super-Yang-Mills theory. A
similar formulation o f the N = 8 supergravity theory
should also be possible.
= (-2./5+2) --~= 4/3
One of us (P.K.T.) would like to thank Professor
Fig. 4.7. (a) Integral coming from reduction of figs. 4.5a, c. S. Hawking and the members o f DAMTP, Cambridge
(b) Relation between (a) and the integral of fig. 4.5d. for their hospitality during the course o f this work.

form of fig. 4.5d or to that o f fig. 4.7a. But by manipu-


lation o f the integrand, as shown schematically in fig. References
4.7b, the latter integral can also be reduced to the
[1 ] E. Poggio and H. Pendleton, Phys. Lett. 72B (1977) 200;
form o f fig.4.5d. As all contributions to the infinite
D.R.T. Jones, Phys. Lett. 72B (1977) 199.
part of the two-point function have now been reduced [2] M. Grisaru, M. Ro~ek and W. Siegel, Zero 3-loop #-func-
to a common form it remains to calculate the combina- tion in N = 4 supersymmetric Yang-Mills theory, Brandeis
toric factors. We find preprint (July, 1980).
[3] P. DiVecchia and S. Ferrara, Nucl. Phys. B130 (1977) 93;
(a): 2 + 4~ . ~3= 4 , (b): 0, E. Witten, Phys. Rev. D16 (1977) 2991; Nucl. Phys. B147
(1979) 285;
4
(c):-1 +~'-~ =0,
3
(d):-4. A. D'Adda, P. Di Vecchia and M. Liischer, Nucl. Phys.
B152 (1979) 125.
Therefore, the three-loop contribution to the two- [4] L. Alvarez-Gaum6 and D.Z. Freedman, K/ihlcr geometry
and the renormalization of supersymmetric a-models,
point function is finite! Presumably the non-linear in-
Phys. Rev. D, to be published.
variance o f the theory, eq. (3.5), can be used to relate [5 ] T.L. Curtright and D.Z. Freedman, Phys. Lett. 90B (1980)
the two-point function to the other higher-point func- 71.
tions and thereby to deduce their finiteness to three [6] L. Alvareza~aum6 and D.Z. Freedman, Phys. Lett. 94B
loops. (1980) 171.
[7] M. Grisaru, W. Siegel and M. Ro~ek, Nucl. Phys. B159
(1979) 429.
5 Discussion. Our approach to the N = 4 two-di- [8] M. Sohnius, K. Stelle and P.C. West, Proc. Erice Conf. on
mensional non-linear supersymmetric o model is simi- Unification of the fundamental interactions (1980).
lar to a recent treatment o f the N = 4 four-dimensional [9] B. Zumino, Phys. Lett. 87B (1979) 203.

76

You might also like