sign in sign up
<<
Home , E7 (mathematics)

AEI-2010-146, MCTP-10-40, MIT-CTP-4176, MPP-2010-122, PUPT-2348

E7(7) constraints on counterterms in N = 8

N. Beiserta, H. Elvangbc, D. Freedmande, M. Kiermaierf , A. Moralesd, and S. Stiebergerg

aMax-Planck-Institut f¨urGravitationsphysik, Albert-Einstein-Institut, Am M¨uhlenberg 1, 14476 Potsdam, Germany bMichigan Center for Theoretical (MCTP), Department of Physics, University of Michigan, Ann Arbor, MI 48109, USA cInstitute for Advanced Study, Princeton, NJ 08540, USA dDepartment of and eCenter for Theoretical Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA f Department of Physics, Princeton University, Princeton, NJ 08544, USA , and gMax-Planck-Institut f¨urPhysik, Werner-Heisenberg-Institut, 80805 M¨unchen,Germany (Dated: September 10, 2010) We prove by explicit computation that the operators D4R4 and D6R4 in N = 8 supergravity have non-vanishing single-soft scalar limits at the 6-point level, and therefore they violate the continuous E7(7) symmetry. The soft limits precisely match automorphism constraints. Together with previous 4 results for R , this provides a direct proof that no E7(7)-invariant candidate counterterm exists below 7-loop order. At 7-loops, we characterize the infinite tower of independent supersymmetric 4 6 8 2 8 operators D R , R , ϕ R ,... with n > 4 fields and prove that they all violate E7(7) symmetry. This means that the 4- amplitude determines finiteness at 7-loop order. We show that the corresponding candidate counterterm D8R4 has a non-linear supersymmetrization such that its single- and double-soft scalar limits are compatible with E7(7) up to and including 6-points. At loop orders 7, 8, 9 we provide an exhaustive account of all independent candidate counterterms with up to 16, 14, 12 fields, respectively, together with their potential single-soft scalar limits.

INTRODUCTION each external scalar line [12–14]. The scalars of N = 8 su- pergravity are the ‘pions’ of this soft-pion theorem since N = 8 supergravity has maximal , and they are the 70 Goldstone bosons of the spontaneously the classical theory has global continuous E7(7) symme- broken generators of E7(7). It was recently proven [15] try which is spontaneously broken to SU(8). Explicit that the soft scalar property fails for 6-point matrix el- 4 calculations have demonstrated that the 4-graviton am- ements of the operator R (see also [16]). Thus E7(7) plitude in N = 8 supergravity is finite up to 4-loop order excludes R4 and explains the finite 3-loop result found in [1]. Together with - and -based obser- [1]. vations [2, 3], this spurred a wave of renewed interest in In the present paper we show first that the 5- and 6- the question of whether the loop computations based on 4 4 6 4 generalized unitarity [4] could yield a UV finite result to loop operators D R and D R are incompatible with all orders1 — or at which loop order the first divergence E7(7) symmetry because their 6-point matrix elements might occur. have non-vanishing single-soft scalar limits. Previous In gravity, logarithmic UV divergences in on-shell L- [17] and superspace [18] arguments sug- loop amplitudes are associated with local counterterm gested this E7(7)-violation. Our results mean that no operators of mass dimension δ = 2L + 2 composed of UV divergences occur in N = 8 supergravity below the fields from the classical theory. The counterterms must 7-loop level. respect the non-anomalous symmetries of the theory. It arXiv:1009.1643v1 [hep-th] 8 Sep 2010 We then survey the candidate counterterms for loop or- was shown in [7–9] that below 7-loop order, there are ders L = 7, 8, 9 using two new algorithmic methods: one only 3 independent operators consistent with linearized program counts monomials in the fields of N = 8 super- N = 8 supersymmetry and global SU(8) R-symmetry gravity in representations of the superalgebra SU(2, 2|8), [10]. These are the 3-, 5- and 6-loop supersymmetric 4 4 4 6 4 the other applies Gr¨obnerbasis methods to construct candidate counterterms R ,D R , and D R . their explicit local matrix elements. Our analysis shows The perturbative S-matrix of N = 8 supergravity that at each loop level 7, 8, 9, there is an infinite tower of should respect E7(7) symmetry [11], so one must subject 4 4 4 6 4 independent n-point supersymmetric counterterms with R ,D R , and D R to this test. A necessary condition n ≥ 4. At 7-loop order we find that none of the n-field for a counterterm to be E7(7)-compatible, is that its ma- µ operators with n > 4 are E7(7)-compatible. This leaves trix elements vanish in the ‘single-soft limit’ p → 0 for D8R4 as the only candidate counterterm at L = 7. We show that its matrix elements are E7(7)-compatible at least up to 6-points. We observe that it requires remark- 1 This question is well-defined whether or not N = 8 supergravity able cancellations for E7(7) to be satisfied to all orders is sensible as a full quantum theory [5, 6]. for any L ≥ 7 candidate counterterm. 2

4 4 6 4 4 4 E7(7)-VIOLATION OF D R AND D R 5-loop counterterm D R At order α05, the SU(8)-average (2) of the string the- To investigate E7(7) we study the soft scalar limit of the ory amplitudes directly gives the matrix elements of the 6-point NMHV matrix elements h++−−ϕϕ¯iD2kR4 . The unique SU(8)-invariant supersymmetrization of D4R4. external states are two pairs of opposite helicity gravi- The result is a complicated non-local expression, but its tons and two conjugate scalars. These matrix elements single-soft scalar limit is very simple and local, viz. contain local terms from nth order field monomials in the nonlinear SUSY completion of D2kR4 as well as non-local 6 4 4 X 2 lim h++−−ϕϕ¯iD4R4 = − 7 [12] h34i sij . (4) p6→0 pole diagrams in which one or more lines of the operator i

At tree level, the closed string effective action takes 33 4 4 X 3 -12φ 6 4 lim h++−−ϕϕ¯i(e D R )avg = − 35 [12] h34i sij . the form p6→0 i

03 -6φ 4 05 -10φ 4 4 (5) Seff = SSG − 2α ζ(3)e R − ζ(5)α e D R (1) It is important to realize that at order α06, the 6-point 2 06 2 -12φ 6 4 1 07 -14φ 8 4 + 3 α ζ(3) e D R − 2 α ζ(7)e D R + .... NMHV closed string amplitudes receive contributions from diagrams involving one vertex from e−12φD6R4 (to- All closed string amplitudes in this work are obtained via gether with vertices from the supergravity Lagrangian) KLT [19] from the open string amplitudes of [20]. The and from pole diagrams with two 4-point vertices of amplitudes confirm the structure and coefficients of (1). e−6φR4 (which coincides with R4 at 4-points). Since R4 Couplings of the φ break the SU(8)-symmetry is not present in N = 8 supergravity, its contributions of the supergravity theory to SU(4)×SU(4) when α0 > 0, must be removed to extract the matrix elements of the 6 4 and thus the supersymmetric operators of Seff are not the supergravity operator D R . The removal process must desired SU(8)-invariant operators. As explained in [15], be supersymmetric. an SU(8)-averaging procedure extracts the SU(8) singlet We first compute the R4 − R4 pole contributions to contribution from the string matrix elements. Specifi- the 6-graviton NMHV matrix element h−−−+++i as cally, the SU(8) average of the h++−−ϕ ϕ¯ie-(2k+6)φD2kR4 follows. This amplitude has dimension 14. Factorization matrix elements from string theory is at the pole determines the simple form

1 1234 5678 4 4 4 2 h++−−ϕϕ¯iavg = 35 h++−−ϕ ϕ i (2) h12i [45] h3|P126|6] /P126 + 8 permutations , (6) 16 123|5 4|678 18 12|56 34|78 − 35 h++−−ϕ ϕ i + 35 h++−−ϕ ϕ i . up to a local polynomial. The 9 terms correspond to the 9 distinct 3-particle pole diagrams. The result (6) is The 3 terms on the right side correspond to the 3 in- then checked by computation of the Feynman diagrams equivalent ways to construct scalars from particles of from the R4 vertex [21]. As the non-linear supersym- the N = 4 , namely from gluons, gluinos, metrization of R4 may contribute additional local terms, and N = 4 scalars. There are 35 distinct embeddings of we also consider adding the most general gauge-invariant SU(4) × SU(4) in SU(8). Averaging is sufficient to give and bose-symmetric polynomial of dimension 14 that can the matrix elements of the N = 8 field theory operator contribute to h−−−+++i, namely R4, as done in [15], and we extend it here to D4R4. For D6R4 a further correction is necessary and is discussed 2 2 h12ih23ih31i[45][56][64] P123 . (7) below. Before proceeding, we note that the operators in the To incorporate SUSY, we separately show that there action (1) are normalized such that their 4-point matrix is a basis for SU(8)-invariant 6-particle NMHV superam- elements are h++−−i = g(s, t, u)[12]4h34i4 with plitudes (an alternative to the basis in [22]) consisting of h−−−+++i and 8 distinct permutations of the states. 2 2 2 gR4 = 1 , gD4R4 = s + t + u , In this basis we write a superamplitude ansatz as the 3 3 3 2 2 2 2 (3) sum of the pole amplitude (6) plus a multiple of (7). We gD6R4 = s + t + u , gD8R4 = (s + t + u ) . then impose full S6 permutation symmetry on the ansatz. 3

This fixes the coefficient of the polynomial (7) to vanish Here, ∆ is the Laplacian on E7(7)/SU(8); in terms of the and determines the SUSY completion of the desired pole scalars ϕabcd of N = 8 supergravity, its leading terms are diagram uniquely! Finally we project out the scalar-graviton matrix el- h ∂ ∂ i ement from this superamplitude and take its single-soft ∆ = + 34 inequiv. perms + .... (12) scalar limit to find ∂ϕ1234 ∂ϕ5678

1 4 4 X 3 It is easy to see that the function (10) with the above lim h++−−ϕϕ¯i(R4)2 = − 70 [12] h34i sij . (8) p6→0 i

6 4 This non-vanishing result shows that the operator D R 2 λR4  2 is also incompatible with continuous E7(7) symmetry. (∆ + 60)fD6R4 (ϕ) = − fR4 (ϕ) . (13) λ 4 4 R4, D4R4 and D6R4 are the only local supersymmet- D R ric and SU(8)-symmetric operators for loop levels L ≤ 6 From (5) and (9), we can reconstruct the coefficient [7–9]. Hence N = 8 supergravity has no potential coun- a in (10) of the functions associated with the SU(8)- terterms that satisfy the continuous E symmetry for 7(7) averaged string-theory operator (e-12φD6R4) and with L ≤ 6. We stress that string theory is used as a tool to avg the supergravity operator D6R4. We find extract SU(8)-invariant matrix elements that must agree with the matrix elements of the N = 8 supergravity op- 66 60 a -12φ 6 4 = , a 6 4 = . (14) erators R4, D4R4 and D6R4 because each of these oper- (e D R )avg 35 D R 35 ators is unique. No remnant of string-specific dynamics -12φ 6 4 remains in the final results. For (e D R )avg, the couplings λ in (13) must take 03 their string theory values λR4 = −2α ζ(3) and λD4R4 = 2 α06ζ(3)2. The N = 8 operator D6R4, on the other Matching to automorphism analysis 3 hand, must satisfy (13) with λR4 = 0 because the opera- The non-vanishing single-soft scalar limits found above tor R4 does not appear in the action of N = 8 supergrav- have their origin in local 6-point interactions of the ity. Indeed, our results for f for both operators satisfy schematic form ϕ2D2kR4 which appear in the non-linear the Laplace equation with the expected choice of λ’s. completion of D2kR4. Let us encode this completion as f(ϕ)D2kR4, with CONSTRUCTION AND COUNTING OF  1234 5678  f(ϕ) = 1 − a ϕ ϕ + 34 inequiv. perms + .... COUNTERTERMS (10) The “. . . ” indicate higher order terms. The constant a 6 4 We now discuss the techniques used to classify and con- depends on the operator; for example a 4 = for R R 5 struct local supersymmetric operators, especially those [15]. We can determine a for D4R4 by taking a further needed for L ≥ 7. We are interested in SU(8)-invariant single-soft limit p →0 on (4) and comparing the resulting 5 operators, which are candidate counterterms, and in op- s, t, u-polynomial with the 4-point normalization of (3). 12 erators transforming in the 70 of SU(8). The latter The result is a 4 4 = . D R 7 are candidate operators for local single-soft scalar limits Ref. [17] used supersymmetry and duality consider- (SSL’s) of the matrix elements of singlet counterterm op- ations in d dimensions to constrain the moduli depen- erators. First we use of the super- dent functions f(ϕ) of the BPS operators R4, D4R4 and algebra SU(2, 2|8) to determine the spectrum and mul- D6R4. Specifically, for R4 and D4R4 in 4 dimensions, tiplicity of these operators. The spectrum is classified they found that f(ϕ) should satisfy the Laplace equation by the number n of external fields, the scale dimension, and the order k of the NkMHV type. Then we construct matrix elements of several operators explicitly using al- (∆ + 42)f 4 (ϕ) = 0 , (∆ + 60)f 4 4 (ϕ) = 0 . (11) R D R gorithms which incorporate Gr¨obnerbasis techniques. 4

Spectrum of local operators Note that locality requires that the exponents in (16) are non-negative numbers. In particular, 3p − q − n ≥ 0 Counterterms of N = 8 supergravity are supersymmet- implies 2L ≥ 6 + 2n − 4k using (17). This bound on the ric, SU(8)-invariant, Lorentz scalar local operators C in- existence of local non-BPS operators was conjectured in tegrated over spacetime. These local operators involve n- [8] and confirmed very recently in [9]. fold products of the fundamental fields and their deriva- As a simple illustration, consider (16) with n = 4 and tives. We restrict to diffeomorphism-covariant combina- p = q. We find C = D2p−4ϕ4, and after application of tions of the fields, such as the Riemann tensor R. In 0 D16 it becomes D2p+4R2R¯2; this is just the 4-point MHV enumerating all local operators of a given order n (up to local counterterm D2p+4R4. Locality of C requires p ≥ covariance), the equations of motion set the Ricci tensor 0 2, so the first available non-BPS operator is D8R4. equal to a combination of fields of quadratic order (and In practice, we use a C++ program to enumerate all lo- higher), which is automatically included at order > n in cal operators with 2L ≤ 30 − n amounting to ∼ 4.8 · 1022 the enumeration. The remaining 10 components of the terms.2 These are decomposed into irreps of SU(2, 2|8) on-shell Riemann tensor group into fields with Lorentz by iteratively removing the lowest weights and their cor- spin (2, 0) and (0, 2). The collection of all on-shell super- responding supermultiplets [23]. Special attention needs gravity fields span a representation of the N = 8 super- to be paid to BPS and short supermultiplets [24]. In to- Poincar´ealgebra as well as an ultrashort representation tal we obtained around 8.8 · 105 types of supermultiplets of N = 8 superconformal symmetry (see [9] for a recent along with their multiplicities.3 Finally we extract su- discussion). Using the SU(2, 2|8) permultiplets with scalar SU(8) singlets and 70’s as top –⊗– – – – – – – –⊗– , (15) supersymmetry components. The results at L ≤ 9 are SU(2)L | {z } SU(2)R SU(8) presented in Table I. 1 1 1 Our analysis shows that there are unique 2 , 4 , 8 the Dynkin labels of this lowest-weight representation BPS counterterms R4, D4R4 and D6R4, in agreement read [0,0,0001000,0,0], where the SU(8) labels [0001000] with earlier results [7–9]. They correspond to the lowest describe a 70 and the SU(2)L × SU(2)R Lorentz spins weights indicate a scalar. The graded symmetric tensor product of n copies of [0,0,0004000,0,0], [0,0,0200020,0,0], [0,0,2000002,0,0] . the above multiplet provide all local operators with n (18) fields. We are interested in supersymmetric operators: In the previous section, we showed that their 6-point ma- there is typically one such operator C in each irrep of trix elements have non-vanishing single-soft limits origi- the tensor product. For long supermultiplets it is the nating from the non-linear completion of the operators. 4 unique top component, obtained by acting with SUSY The limits correspond to local 70 BPS operators ϕR , 16 16 4 4 6 4 1 1 generators Q Q˜ on the lowest-weight component C0. ϕD R and ϕD R , which are descendants of the 2 , 4 , 1 5 (In superspace approaches, this is equivalent to the full 8 BPS superconformal primaries ϕ with SU(8) Dynkin superspace measure R d32θ.) For short or BPS supermul- labels [0005000], [0201020], [2001002]. The relationship tiplets fewer are needed to get from C0 between BPS operators are illustrated in Table I. to the top component(s). Hence it is sufficient to enu- merate the lowest superconformal weights C0. Its super- Explicit matrix elements and superamplitudes conformal transformation properties determine the spin, The matrix elements of potential counterterms such as k SU(8) representation as well as loop and N MHV level. D2kRn must be polynomials of degree δ = 2(k + n) in More concretely, Dynkin labels translate to scalar local angle and square brackets hi ji, [kl] which satisfy several operators as follows (assume q ≥ p) constraints. If ai and si denote the number of angle |ii [0,p,0000000,q,0] → D3p−q−nϕn+p−qRq−p (singlet) , and square |i] spinors for each particle i = 1, 2, . . . , n, then the total number of spinors is fixed by the dimen- [0,p,0001000,q,0] → D3p−q−n+1ϕn+p−qRq−p (70) . sion of the operator to be P (a + s ) = 4(k + n). For (16) i i i each particle i, of helicity h , there is a helicity weight Note that we display only prototypical terms, mixture i constraint a − s = −2h . We need polynomials which with other fields is implied: E.g. D4ϕ2 ' RR¯. (Here i i i are independent under the constraints of momentum con- we distinguish between the chiral and anti-chiral compo- servation and the Schouten identity, nents of the Riemann tensor.) To get from C0 to the su- 16 ˜16 16 X persymmetric C in long multiplets, apply Q Q ' D . hiji[jk] = 0 , hijihkli+hjkihili+hkiihjli = 0 , (19) k A lowest weight as in (16) then corresponds to a N MHV j counterterm at L loops with n 14 + p + q − n (singlet), 2k = n + p − q − 4 , 2L = 15 + p + q − n (70). 2 The computation took 3.5 hours on a desktop PC. (17) 3 The decomposition took 42 hours. 5 and a similar Schouten identity for square brackets. obtain These polynomials must satisfy bose and fermi symme- -14φ 8 4 tries when they contain identical particles. Finally, the lim h++−−ϕϕ¯i(e D R )avg p6→0 polynomials must satisfy SUSY Ward identities. This 2 4 4h 3 X 4 1  X 2  i (20) was ensured in [8] by packaging n-point matrix elements = −2[12] h34i 4 sij + 16 sij . into the manifestly SUSY- and SU(8)-invariant superam- i

3-loop 4-pt 5-pt 6-pt 5-loop 4-pt 5-pt 6-pt 6-loop 4-pt 5-pt 6-pt  2 ¨4  2 44  2 64 R4 ϕ2D2R3 ϕ ¨R D4R4 ϕ2D6R3 ϕ D R D6R4 ϕ2D8R3 ϕ D R singlet  ¨ singlet   singlet   1×MHV R4 non-linear 1×MHV D4R4 non-lin. 1×MHV D6R4 non-lin.

.soft .soft .soft 70 1×ϕR4 70 1×ϕD4R4 70 1×ϕD6R4

7-loop 4-pt 5-pt 6-pt 7-pt 8-pt 9-pt 10-pt 11-pt 12-pt 13-pt 14-pt 15-pt 16-pt 8 4 65 4 6 27 8 227 2 8 427 4 8 627 6 8 827 8 8 singlet D R D R D R D R R ϕ D R ϕ R ϕ D R ϕ R ϕ D R ϕ R ϕ D R ϕ R 1×MHV 2×NMHV 3×N2MHV 4×N3MHV 6×N4MHV 8×N5MHV 10×N6MHV

.soft .soft .soft .soft .soft .soft 8 4 4 6 8 3 8 5 8 7 8 70 ϕD R ϕD R ϕR ϕ R ϕ R ϕ R 2× 4× 6× 9× 14× 19×

8-loop 4-pt 5-pt 6-pt 7-pt 8-pt 9-pt 10-pt 11-pt 12-pt 13-pt 14-pt

10 4 8 5 6 6 4 7 2 8 9 2 2 8 2 9 4 2 8 4 9 6 2 8 singlet D R D R D R D R D R R ϕ D R ϕ R ϕ D R ϕ R ϕ D R 1×MHV 1×MHV 3×NMHV 3×NMHV 8×N2MHV 8×N2MHV 25×N3MHV 22×N3MHV 66×N4MHV 51×N4MHV 153×N5MHV ...... 10 4 8 5 6 6 4 7 2 8 9 3 2 8 3 9 5 2 8 70 ϕD R ϕD R ϕD R ϕD R ϕD R ϕR ϕ D R ϕ R ϕ D R 3× 4× 17× 16× 81× 63× 232× 211× 1033×

9-loop 4-pt 5-pt 6-pt 7-pt 8-pt 9-pt 10-pt 11-pt 12-pt D12R4 D10R5 D8R6 D6R7 D4R8 D2R9 R10 ϕ2D2R9 ϕ2R10 singlet 2×MHV 1×MHV 12×NMHV 14×NMHV 117×N2MHV 123×N2MHV 780×N3MHV 783×N3MHV 4349×N4MHV 2×MHV 7×NMHV 36×N2MHV 169×N3MHV ...... ϕD12R4 ϕD10R5 ϕD8R6 ϕD6R7 ϕD4R8 ϕD2R9 ϕR10 70 5×N0.5MHV 8×N0.5MHV 122×N1.5MHV 194×N1.5MHV 1814×N2.5MHV 2317×N2.5MHV 16485×N3.5MHV 5×N0.5MHV 52×N1.5MHV 469×N2.5MHV

TABLE I. Supersymmetric SU(8)-singlet L-loop counterterms and the SU(8) 70 operators which describe their potential single-soft scalar limits. When the singlet operator is in the NkMHV classification, the single-soft scalar limit operator belongs (k− 1 ) to the N 2 MHV sector. For L < 7, there are no independent singlet operators with n > 4, but the non-vanishing single-soft scalar limits arise from the non-linear completions of the 4-point operators D2kR4. one can choose the basis contractions given in (23). A similar construction can 2 ˜   be carried out for the three 8-point N MHV superampli- AD4R6 = δ(Q)δ(Q) (ϕ1, ϕ2)(ϕ3, ϕ4)(ϕ5, ϕ6) + perms , tudes of R8. Again, one can immediately propose three ˜   8 AD4R60 = δ(Q)δ(Q) (ϕ1, ϕ2, ϕ3, ϕ4)(ϕ5, ϕ6) + perms . superamplitudes that span the space of R counterterms, (23) for example by considering three order-8 contractions in- volving the ϕ-products (25). Here Q, Q˜ are the usual supercharges that act on the Grassmann η-variables of the superamplitude as differ- The infinite tower of 7-loop counterterms entiation and multiplication, respectively, and thus We now examine the multiplicity of potential 7-loop 8 8 n-point counterterms [8] Y X ∂2 Y X δ(Q) = [ij] , δ(Q˜) = hijiηiaηja . ∂η ∂η 8 4 4 6 8 2 8 4 8 a=1 i

a1a2a3a4b1b2b3c1 d1d2d3d4c2c3c4b4 them schematically as (ϕi, ϕj, ϕk, ϕl)≡  (25) Q4 16 ˜16 2q × t=1ηiat ηjbt ηkct ηldt . C ' Q Q ϕ . (27) Of course, the choice of contractions is not unique, The lowest weight ϕ2q must be in an SU(8)-singlet com- and only through the previously established multiplicity bination, and every such singlet gives rise to a long super- count do we know that it is sufficient to consider the two multiplet. Hence there is one 7-loop n-point counterterm 7 for each singlet in the decomposition of the symmetric expects the soft-limits of the 6-point matrix elements of tensor product of n = 2q 70’s. With increasing q there D4R4 and D6R4 to be proportional to this operator with is a (swiftly) increasing number of singlets, as illustrated some non-vanishing coefficient. This is precisely what by the explicit multiplicities up to n = 16 in Table I. happens. Consequently, there is an infinite ‘tower’ of independent At 7 loops with n > 4 points, the number nS of SSL op- 7-loop operators that are potential counterterms. Opera- erators D16ϕn−1 is always at least as large as the number 16 n tors corresponding to their SSL are also listed in Table I. nC of potential counterterms D ϕ . Generically, one Their construction is similar, and their multiplicities is therefore expects the soft limits of the potential coun- the number of 70’s in a product of (2q−1) 70’s. terterms to span an nC -dimensional subspace in the nS- dimensional space of SSL operators. It would follow that E7(7) violation of higher-point 7-loop operators all potential counterterms with n > 4 violate E7(7). In- deed, this is what we explicitly proved above for the 7- Consider the leading n-point matrix elements of a local loop case. counterterm C. If non-vanishing, the SSL produces a By the same logic, it is not at all surprising that there local (n − 1)-matrix element, which can be generated by is a non-linear supersymmetrization of D8R4 that pre- an (n − 1)-point local operator dC in the 70. Locality serves E at the 6-point level. The number of 5-point of C ensures that the SSL operation C → dC commutes 7(7) SSL operators precisely matches the number of D4R6 op- with the SUSY generators Q and Q˜. For a long-multiplet erators. Therefore, the 6-point soft limit of D8R4 can be (L ≥ 7) counterterm C = Q16Q˜16C we can therefore 0 made to vanish after adding an appropriate combination write of the two D4R6 operators, just as we found above. For 16 16 the 8-point soft limits of D8R4, however, there are 4 SSL dC = Q Q˜ dC0. (28) operators available; more than the 3 potential 8-point 8 8 4 E7(7) requires that the SSL vanishes. Now there are two counterterms R . If the 8-point soft limits of D R take ways to obtain dC = 0: either the single-soft limit of a generic value in the 4-dimensional space of SSL op- 16 16 8 C0 vanishes (dC0 = 0), or dC0 is annihilated by Q Q˜ . erators, no linear combination of R operators can be 2q At the seven-loop level, C0 = ϕ consists of only scalars chosen to give an E7(7)-preserving supersymmetrization 8 4 with no derivatives, and consequently dC0 6= 0. C0 is also of D R ; a remarkable cancellation is thus required for 16 ˜16 2q−1 8 4 not annihilated by Q Q because dC0 = ϕ in a 70 D R to be compatible with E7(7). satisfies a shortening condition only for n = 2q ≤ 4 [24]. As Table I illustrates, E7(7) becomes more and more Therefore, all 7-loop linearized counterterms with n > 4 constraining as we increase the number of points and have non-vanishing single-soft scalar limits and thus vio- loops. For example, the 14-point soft limits of D10R4 late E7(7). (This was also observed in [28].) These oper- have to lie in a specific 153-dimensional subspace of the ators may, however, play an important role as dependent 1033-dimensional space of SSL operators in order for 8 4 10 4 terms in the non-linear completion of the D R operator, D R to satisfy E7(7) after an appropriate addition of as we demonstrated above at the 6-point level. independent 14-point operators. It follows that E7(7) is a very constraining symmetry even for L = 7 and be- yond. Although there is an infinite tower of independent SSL STRUCTURE: 7-, 8- AND 9-LOOPS counterterms at each of loop L ≥ 7, we cannot expect any of these operators to preserve E7(7) ‘accidentally’. We now show that all our findings on E7(7)- There may, however, be a very good reason for the can- (in)compatibility of operators have a natural explanation cellations of terms that is needed for E7(7)-invariant op- in terms of the multiplicities of SSL operators in the 70 erators to exist for L ≥ 7; namely, when there is a con- 4 that is displayed in Table I. Let us first revisit the case struction of a manifestly E7(7)-invariant supersymmetric of D4R4 and D6R4. At the 5- and 6-loop level, there are operator [29, 30]. At the 7-loop level, for example, we can no potential 3-point or 4-point SSL operators available. only expect the 8-point single-soft limits of D8R4 to van- Therefore the matrix elements of D4R4 and D6R4 must ish after an appropriate addition of R8, if the manifestly have vanishing soft limits at 4- and 5-points, and this is E7(7)-invariant superspace integral that was proposed as indeed the case. There exists, however, one potential 5- a candidate counterterm in [29] is indeed non-vanishing. point SSL operator at L = 5 and L = 6. Generically, one One new feature that emerges at L = 8, 9 is the exis- tence of n > 4 operators that have vanishing soft-limits at the linearized level. This holds for the MHV operators D8R5, D10R5 and 2 × D8R6 as well as for at least 7 of 4 Throughout this section we are only concerned with the lowest- the 12×D8R6 NMHV operators. The latter follows from point non-vanishing SSL’s of an operator. If an operator has non-vanishing SSL’s at n-point, its higher-point matrix elements the multiplicity 5 of 5-point SSL operators at 9 loops. will generically have non-local SSL’s, which are not classified by E7(7)-invariance beyond the linearized level, however, is our analysis. a highly non-trivial constraint on all of these operators. 8

ACKNOWLEDGEMENTS K. S. Stelle, “Is N=8 supergravity a finite field theory?,” Fortsch. Phys. 57, 446 (2009). [4] Z. Bern, L. J. Dixon, D. C. Dunbar and D. A. Kosower, We thank M. Headrick for suggesting the use of the “One loop n point gauge theory amplitudes, unitarity Gr¨obnerbasis for the study of counterterms in super- and collinear limits,” Nucl. Phys. B 425, 217 (1994) gravity. We thank T. McLoughlin and P. Vanhove for [arXiv:hep-ph/9403226]; valuable discussions. Z. Bern, L. J. Dixon, D. C. Dunbar and D. A. Kosower, “Fusing gauge theory tree amplitudes into loop am- The research of DZF is supported by NSF grant PHY- plitudes,” Nucl. Phys. B 435, 59 (1995) [arXiv:hep- 0600465 and by the US Department of Energy through ph/9409265]; cooperative research agreement DE-FG-0205FR41360. Z. Bern, L. J. Dixon and D. A. Kosower, “Two-Loop g HE is supported by NSF CAREER Grant PHY-0953232, → gg Splitting Amplitudes in QCD,” JHEP 0408, 012 and in part by the US Department of Energy under DOE (2004) [arXiv:hep-ph/0404293]. grants DE-FG02-95ER40899 (Michigan) and DE-FG02- [5] , talk at Supergravity vs Superstring The- ory in the Ultraviolet Workshop at Penn State, Au- 90ER40542 (IAS). The research of MK is supported by gust 27-30, 2009. http://www.gravity.psu.edu/events/. . . the NSF grant PHY-0756966. . . . /supergravity/proceedings.shtml [6] M. B. Green, H. Ooguri and J. H. Schwarz, “Decoupling Supergravity from the Superstring,” Phys. Rev. Lett. 99, 041601 (2007) [arXiv:0704.0777 [hep-th]]. [7] J. M. Drummond, P. J. Heslop, P. S. Howe and S. F. Ker- [1] Z. Bern, J. J. Carrasco, L. J. Dixon, H. Johansson, stan, “Integral invariants in N = 4 SYM and the effective D. A. Kosower and R. Roiban, “Three-Loop Superfinite- action for coincident D-,” JHEP 0308, 016 (2003) ness of N=8 Supergravity,” Phys. Rev. Lett. 98, 161303 [arXiv:hep-th/0305202]. (2007) [arXiv:hep-th/0702112]. [8] H. Elvang, D. Z. Freedman and M. Kiermaier, “A sim- Z. Bern, J. J. M. Carrasco, L. J. Dixon, H. Johansson and ple approach to counterterms in N=8 supergravity,” R. Roiban, “Manifest Ultraviolet Behavior for the Three- arXiv:1003.5018 [hep-th]. Loop Four-Point Amplitude of N=8 Supergravity,” Phys. [9] J. M. Drummond, P. J. Heslop and P. S. Howe, “A note Rev. D 78, 105019 (2008) [arXiv:0808.4112 [hep-th]]. on N=8 counterterms,” arXiv:1008.4939 [hep-th]. Z. Bern, J. J. M. Carrasco and H. Johansson, [10] N. Marcus, “Composite Anomalies In Supergravity,” “Progress on Ultraviolet Finiteness of Supergravity,” Phys. Lett. B 157, 383 (1985). arXiv:0902.3765 [hep-th]. P. di Vecchia, S. Ferrara and L. Girardello, “Anomalies Z. Bern, J. J. Carrasco, L. J. Dixon, H. Johansson and Of Hidden Local Chiral Symmetries In Sigma Models R. Roiban, “The Ultraviolet Behavior of N=8 Supergrav- And Extended Supergravities,” Phys. Lett. B 151, 199 ity at Four Loops,” Phys. Rev. Lett. 103, 081301 (2009) (1985). [arXiv:0905.2326 [hep-th]]. [11] G. Bossard, C. Hillmann and H. Nicolai, “Perturba- L. J. Dixon, “Ultraviolet Behavior of N=8 Supergravity,” tive quantum E7(7) symmetry in N=8 supergravity,” arXiv:1005.2703 [hep-th]. arXiv:1007.5472 [hep-th]. [2] N. Berkovits, “New higher-derivative R**4 theo- [12] M. Bianchi, H. Elvang and D. Z. Freedman, “Generating rems,” Phys. Rev. Lett. 98, 211601 (2007) [arXiv:hep- Tree Amplitudes in N=4 SYM and N = 8 SG,” JHEP th/0609006]. 0809, 063 (2008) [arXiv:0805.0757 [hep-th]]. M. B. Green, J. G. Russo and P. Vanhove, “Non- [13] N. Arkani-Hamed, F. Cachazo and J. Kaplan, “What is renormalisation conditions in type II string theory the Simplest Quantum Field Theory?,” arXiv:0808.1446 and maximal supergravity,” JHEP 0702, 099 (2007) [hep-th]. [arXiv:hep-th/0610299]. [14] R. Kallosh and T. Kugo, “The footprint of E7 in am- M. B. Green, J. G. Russo and P. Vanhove, “Ultraviolet plitudes of N=8 supergravity,” JHEP 0901, 072 (2009) properties of maximal supergravity,” Phys. Rev. Lett. 98, [arXiv:0811.3414 [hep-th]]. 131602 (2007) [arXiv:hep-th/0611273]. [15] H. Elvang and M. Kiermaier, “Stringy KLT relations, M. B. Green, J. G. Russo and P. Vanhove, “Modular global symmetries, and E7(7) violation,” arXiv:1007.4813 properties of two-loop maximal supergravity and con- [hep-th]. nections with string theory,” JHEP 0807, 126 (2008) [16] J. Broedel and L. J. Dixon, “R4 counterterm and E7(7) [arXiv:0807.0389 [hep-th]]. symmetry in maximal supergravity,” arXiv:0911.5704 N. Berkovits, M. B. Green, J. G. Russo and P. Vanhove, [hep-th]. “Non-renormalization conditions for four-gluon scatter- [17] M. B. Green, J. G. Russo and P. Vanhove, “Automorphic ing in supersymmetric string and field theory,” JHEP properties of low energy string amplitudes in various di- 0911, 063 (2009) [arXiv:0908.1923 [hep-th]]. mensions,” arXiv:1001.2535 [hep-th]. [3] P. S. Howe and K. S. Stelle, “Supersymmetry coun- M. B. Green, J. G. Russo and P. Vanhove, “String theory terterms revisited,” Phys. Lett. B 554, 190 (2003) dualities and supergravity divergences,” JHEP 1006, 075 [arXiv:hep-th/0211279]. (2010) [arXiv:1002.3805 [hep-th]]. K. Stelle, “Finite after all?,” Nature Phys. 3, 448 (2007). M. B. Green, S. D. Miller, J. G. Russo and P. Vanhove, G. Bossard, P. S. Howe and K. S. Stelle, “The ultra- “Eisenstein series for higher-rank groups and string the- violet question in maximally supersymmetric field theo- ory amplitudes,” arXiv:1004.0163 [hep-th]. ries,” Gen. Rel. Grav. 41, 919 (2009) [arXiv:0901.4661 [18] G. Bossard, P. S. Howe and K. S. Stelle, “On duality [hep-th]]. symmetries of supergravity invariants,” arXiv:1009.0743 9

[hep-th]. mal Supersymmetry,” Lett. Math. Phys. 8, 159 (1984). [19] H. Kawai, D. C. Lewellen and S. H. H. Tye, “A Relation V. K. Dobrev and V. B. Petkova, “All Positive Energy Between Tree Amplitudes Of Closed And Open Strings,” Unitary Irreducible Representations Of Extended Con- Nucl. Phys. B 269, 1 (1986). formal Supersymmetry,” Phys. Lett. B 162, 127 (1985). [20] S. Stieberger, “On tree-level higher order gravitational B. Binegar, “Conformal Superalgebras, Massless Repre- couplings in ,” arXiv:0910.0180 [hep- sentations And Hidden Symmetries,” Phys. Rev. D 34, th]. 525 (1986). S. Stieberger and T. R. Taylor, “Complete Six-Gluon B. Morel, A. Sciarrino and P. Sorba, “Unitary Mass- Disk Amplitude in Superstring Theory,” Nucl. Phys. B less Representations Of Conformal Superalgebras,” Phys. 801, 128 (2008) [arXiv:0711.4354 [hep-th]]. Lett. B 166 (1986) 69. S. Stieberger and T. R. Taylor, “Supersymmetry Re- [25] David Cox, John Little, Donal O’Shea, Ideals, Varieties, lations and MHV Amplitudes in Superstring Theory,” and Algorithms, Springer, New York N.Y. (2000) Nucl. Phys. B 793, 83 (2008) [arXiv:0708.0574 [hep-th]]. [26] Bernd Sturmfels, Algorithms in Invariant Theory, [21] D. J. Gross and E. Witten, “Superstring Modifications Springer, New York N.Y. (2008) Of Einstein’s Equations,” Nucl. Phys. B 277 (1986) 1. [27] http://www.math.uiuc.edu/Macaulay2/ D. J. Gross and J. H. Sloan, “The Quartic Effective Ac- [28] R. Kallosh, “The Ultraviolet Finiteness of N=8 Super- tion for the Heterotic String,” Nucl. Phys. B 291, 41 gravity,” arXiv:1009.1135 [hep-th]. (1987). [29] G. Bossard, P. S. Howe and K. S. Stelle, “A note on the [22] H. Elvang, D. Z. Freedman and M. Kiermaier, “So- UV behaviour of maximally supersymmetric Yang-Mills lution to the Ward Identities for Superamplitudes,” theories,” Phys. Lett. B 682, 137 (2009) [arXiv:0908.3883 arXiv:0911.3169 [hep-th]. [hep-th]]. [23] M. Bianchi, J. F. Morales and H. Samtleben, “On stringy [30] P. S. Howe and U. Lindstrom, “Higher Order Invari- AdS(5) x S**5 and higher spin holography,” JHEP 0307, ants In Extended Supergravity,” Nucl. Phys. B 181, 487 062 (2003) [arXiv:hep-th/0305052]. (1981). N. Beisert, M. Bianchi, J. F. Morales and H. Samtleben, R. E. Kallosh, “Counterterms in extended supergravi- “On the spectrum of AdS/CFT beyond supergravity,” ties,” Phys. Lett. B 99, 122 (1981). JHEP 0402, 001 (2004) [arXiv:hep-th/0310292]. [24] M. Flato and C. Fronsdal, “Representations Of Confor-