Research Press ARTICLE Unconstrained canonical action for, and positive energy of, massive spin 2 S. Deser Abstract: Filling a much-needed gap. we exhibit the D = 4 Ferz-Pauli massive ¢= 2 action, and its — manifestly postive — ‘energy, in terms of ts 25 +1= 5 unconstrained helicity @ 1, 0) excitations, after reducing and dlagonatizing the troublesome heliity-zero sector, PACS Nos. 1L10-2, 14.803, Résumé : Nous complétons ii un domaine bien connu en présentant action massive: D=4de Fiere-Pauli, avecson énergie ‘manifestement positive, en fonction de ses 25 + 1= 5 excitations libres, heleite (#2, #1, 0). Nous réduisons et diagonalisons ‘Cabord le secteur le plus deliat, celui rhe! 1. Introduction ‘We begin with an apologia, because there is nothing still not, known about the Fierz-Pauli(F7) model, However, there seems 0 bbe no published deriation'—fom its covariant and highly constrained form—to the final unconstrained canonical action in terms fits2s+ = 5 halicity (22, #1, 0) components. The later form will display that teach mode propagates correctly and has manifestly postive energy. that is, asthe usual L= pj — H, H = (2Xp" + q(—V + hgh While Ferz [i] realized that postive energy was essential, it was displayed rather opaquely: a subsequent formulation [3] waslikewise less than transparently presented (and contains distracting typos). indeed, proper use of the constraints is not altogether trivial, making the correct process instructive as a (minor) exercise in fee field theory ‘Historically twas not andi FP’ 1929 work thatthe problem of representing massive spins >1 by tensor flelds (involving many ‘more than 2s + I components) was raised, let alone solved. Given the current interest in massive gravity with Einstein Kinetic terms plus nonderivative mass terms involving a fixed, say flat, backe ground. our summary may be useful. Indeed, this isa good place to note that, contrary o statements in the massive gravity ltera- ture, the mass terms destroy the whole (Armowitt—Deser-Misner) asymptotic energy formulation of general relativity as a to: dimensional surface integral a spatial infinity, just as the Coulomb asymptotic integral that counts total charge is lost in massive (Proca) vector theory: the Newtonian-Coulomb fields decay much too fast there for these integrals to conteibute at all. 2. The derivation ‘The action and fleld equations of the theory are the sum of linearized general relativity and the-uniquemass term that elim- nates the sath, ghost elicity zero, degree of freedom (DoF). We work throughout in firstorder 3+ 1, canonical form, which sim- plies the procedure and indeed starts directly in verms of the six ér0, [Traduit parle Réaction) ‘conjugate pairs (nh) rather than the 10 covariant is (see eg, ref 2) 1 fasfeiy — mem] 11+ (Be) +t + anna +e) 2 0 ‘where under the integral) 1 4 ltt tre — Vs) a y= (oP = Phy + and n ~ (12) is. Lagrange multiplier enforcing the linea Straint Ry = 0; while N, = ho, Decomes an auxiliary field t0 be eliminated by. completing squares, leaving only the six (>) pairs — indeed our whole process consists of juggling quadratic forms. Finally, we recall that the linearized threedimensional Einstein tensor, Gh), is both denticaly conserved and indepen- dent of the longitudinal, gauge, parts of hy, The second insted ent, essential tothe separation of the varius helicity Do in) is {the usual orthogonal decomposition of any symmetric tensor, = sit + 4a, — aps" + (Ost + ast) + 5, = SFT + 38, — 5p8" + (if + 381) @ ate jnaVe Completing squares in (1) removes the N, dependence of Hin favor of adding the verm 2m 92" to H. There remains the elim Received 27 July 201, Accepted 21 August 2014 S. Deser. Walker Burke Instity Univers, Waltham, MA 02454, USA Email for correspondence: deserobrandeiseds, for Theoretical Physies, California Institute of Technology, Pasadena, CA 91125, USA. Physics Department, Brandeis Dedicated to Chris Isham, the perpetrator of massivees gravity on his 70th bethday is was acwually done [for the more general ease ofthe system embedded in dS, rather th fat, space: however that devvation involved (an even ‘number of steps using inverse powers ofA, hence singularand not applicable here; is inal result ofcourse does init to ours. Ie was also shown in detall there that the hlicin-zero mode can be remaved by suitably tuning w/in dS ca J Phy. 98:395-357 20 ee do orgO-MIjpINL4020 “# Pubshed at wr eresearchpress comp on 10 September 208ination of the R, constraint, hence of one linear combination of the two helicity zero TL) modes. Itwil be equally essential to use R, = 0 to further eliminate one combination of their conjugate momenta (=, z!) using h~ x, constraints “strike twice” in our firstonder form, because they are valid forall times® “The task before us then is to decompose i) into a sum of three (noninteracting) orthogonal, two-DoF sectors: helicky + 2(TD, ele city # 1-717. and the (F, l)helictyzero. The latter's Hamitonian is the source of difticulty, being priori nonpositive before using the R(T. [)=0 constraint. To keep the discussion compact, we first dispose of the helicity > 0 sectors: that of TT is trivial to obtain, being unconstrained; we simply add up the TT terms in (1; drop- ping “FT", we have La stig Hm ape Rtas + mt) 20 “ [Note that H vanishes only for TT vacuum, #= 0 =f. The same is, true ofthe transverse vector (T) part though it sil requires some field redefinitions to achieve the same final form: here (1) also easily ylelds (omitting “1") He 2 + 2m a) + Smit 20 6 \while noc (yet) very prety, this Hs also positive and vanishes at (7) vacuum, a result unaffected by the further field redefinitions required to reach the final p ~ H form; we outline the process in Appendix A. (Recall, however, that correct energy functional form is only reached when the *(p, a) variables are redefined to ensure that the associated kinetic, “pq”, term is itself free of unwanted ‘numerical coefficients) ‘We now face the final, H(T 1), sector, where Ry must be used — twice. There, aay faa + a — When) Re 4¥{0) = HV? — fey — 2h 6) Roa = Soe = ae a ae me We now show that both potential and Kinetic parts of Hf are positive, using the Ry constraint and its time derivative, respec tively, liminating h, yields Wihetid = Sftigg? + mf] = 0 a again, Vonly vanishes at vacuum, ly = 0. Next we find the Ry = 0 Constraint between 7 and nt; The two field equations for 3 ~h obtained by varying (6) with respect to the rare ho eat amt = 0 tye at =0 6 ‘Taking their appropriate vanishing linear combination, we learn that ital = (-20 = mat 0 Can. Pigs. VoL 93,2015, hence finally (00) We have now established £0 forthe ful theory, but one task has yet to be completed: putting the helicity action into exact ‘pj — Hipa) form. Bven before this is done, one can already see that the second ordet field equations are uniformly (€—m®h= 0, but it isan amusingexercise —as well asa check-on the result —to do 0. Using (9 irs easy to translate the Fl sector's wh, + 1) into ‘rj form. AC this penultimate point, unt) thea + te om the obvious rescaling (a) > V3B(— i) achlees the desied final canonical frm ofthe beliiy-rero sector. a ee (e) Together with the vector mode in Appendix A, then, the total =D fit Hoth + air wad} ‘o) ‘after the (cosmetic) rescaling 7+ p'V2, h > q,V2. 3. Summary The physical correctness of the massive s = 2 FP model has been displayed: each of its 25 + 1= 5 helicity excitations obeys (O- mj = 0, Feo. Acknowledgements ‘This work was supported in partby grants NSF PHY-1266107 and. DOE # DE-SCoO11632. Collaboration with A. Waldron on ref. 2 facilitated the present effort, as did composing help from J. Prank- lin and CPU help from G, Conrad. References 1M Riere Hel Phys. Ata 1231929} a Geman doi s169ea10500, 2. 8, Deser and A. Waldron, Phys Vet 8,508,347 (201, do 101680370 2osrionnon3s ‘4S Dever J Trobatch. and. Trbatch Can J Phys 44, 17151966 do03139) pons. 4. Capone, Deser. and Yang. Aan Phy. 199.76 (987) doo 0168000 soias7eon0s2. Appendix A: The helicity + 1 sector ‘We consider here the remaining, helicityone, subspace involv ingonly the“ parts of (iin (1), Cleary, neither WG) nor the R, constraint involve hy ¢ only the mass term does: it contains (2}m i). Rs = sector involves (x3) as well asthe quadratic term 2N"9 7. Using the (1/2) from the mass term, we com plete the square to leave a net contribution m-%c!j!~ 2m-* (iv) there. Acthis pint, then, dropping the Tsndies, we find ‘That the orignal number sa, of kinetic terms decreases by one for every constaiat just Darbous's theorem on quadati forms; in massless theory there are four constraint, lav just the ewo (2, | pars. We wil ee this move explicitly Iter in this section, “& oblished by NRC Research PressUn) = ~~ Lint? + am %aft — Wh} 0) the necessary redefinition is obvious: adm eh = 05) and leads to the desired heliityone canonical Lagrangian, 4 Un) ah — Sle? + ent — (05) Another approach tothe problem is “from above": D=4 massive theories are descendants of D= 5 gauge theories, compactified on ‘circle of radius m [4], This fact immediately implies both positive ‘energy and correct DoF count in D = 4° the former, because the ‘energy of (abelian) gauge theories is manifestly positive, the latter because n has Y2D(D ~ 3) ~5 in D=5. One even gets the correct massdependence ofthe Hamiltonian, va) ~ +f ti v io V5 + tH Vi a where m,n, = 1.23, .and ;—im. However, isin the na step dsentangling the contains among the component ofthe foursimensonal TT" Rana the pis teediiensionl ones. i iy hg). to disgonalze {ino is helices avin (2) — that the sane, nae fynnastics catches up in this approach, 1 thank W. Sigel for Rggesnga al of ts altenat out “& oblished by NRC Research Press