Front. Phys. Frontiers in Physics Front. Phys. 2296-424X Frontiers Media S.A. 10.3389/fphy.2018.00129 Physics Original Research Spin-2 Twisted Duality in (A)dS Boulanger Nicolas 1 Campoleoni Andrea 1 2 Cortese Ignacio 3 Traina Lucas 1 * 1Service de Physique de l'Univers, Champs et Gravitation, Université de Mons, Mons, Belgium 2Institut für Theoretische Physik, ETH Zurich, Zurich, Switzerland 3Departamento de Física de Altas Energías, Instituto de Ciencias Nucleares, Universidad Nacional Autónoma de Mèxico, Ciudad de México, Mexico

Edited by: Ashkbiz Danehkar, Harvard-Smithsonian Center for Astrophysics, United States

Reviewed by: Yurii M. Zinoviev, Institute for High Energy Physics, Russia; Ioannis Papadimitriou, Korea Institute for Advanced Study, South Korea

 *Correspondence: Lucas Traina lucas.traina@umons.ac.be

This article was submitted to High-Energy and Astroparticle Physics, a section of the journal Frontiers in Physics

 27 11 2018 2018 6 129 12 06 2018 26 11 2018 Copyright © 2018 Boulanger, Campoleoni, Cortese and Traina. 2018 Boulanger, Campoleoni, Cortese and Traina

This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.

Starting from the dual Lagrangians recently obtained for (partially) massless spin-2 fields in the Stueckelberg formulation, we write the equations of motion for (partially) massless gravitons in (A)dS in the form of twisted-duality relations. In both cases, the latter admit a smooth flat limit. In the massless case, this limit reproduces the gravitational twisted-duality relations previously known for Minkowski spacetime. In the partially-massless case, our twisted-duality relations preserve the number of degrees of freedom in the flat limit, in the sense that they split into a decoupled pair of dualities for spin-1 and spin-2 fields. Our results apply to spacetimes of any dimension greater than three. In four dimensions, the twisted-duality relations for partially massless fields that appeared in the literature are recovered by gauging away the Stueckelberg field.

 electric-magnetic duality dual graviton Stueckelberg formalism constant curvature spacetimes partially massless fields 1. Introduction and conventions

Electric-magnetic duality, the symmetry of vacuum Maxwell equations under the exchange of electric and magnetic fields that interchanges dynamical equations with Bianchi identities, has counterparts in other physical systems, including supersymmetric field theories, linearised gravity and free higher-spin gauge theories. In supersymmetric Yang-Mills theories, electric-magnetic duality—see Olive and West [1] and references therein—acts as a strong/weak duality and, as such, has found applications in the study of non-perturbative phenomena like confinement (see e.g., [2]). In extended supergravity theories it is part of the U-duality symmetry and, since the pioneering work [3], it has been extensively studied.

If the dimension of spacetime is bigger than four, electric-magnetic duality actually links different descriptions of the same physical system. For linearised gravity on a flat background in n dimensions, for instance, it relates the Fierz-Pauli description in terms of the tensor hab with a description in terms of an irreducible mixed-symmetry tensor Ta1 … an−3|b, completely antisymmetric in its first n−3 indices. This link has however been established only at the linearised level: non-linear Einstein gravity cannot be reproduced in the dual mixed-symmetry picture by means of local interactions [4, 5]. The problems encountered in the attempts to lift gravitational dualities from the linearised formulation in flat background to the interacting level suggest that the study of electric-magnetic duality in curved backgrounds may be particularly promising. Positive results about interaction vertices for mixed-symmetry fields are indeed available in this context [6] and may indicate a way to extend the duality to the interacting theory.

Recently Boulanger et al. [7], built manifestly covariant action principles in the Stueckelberg formulation for dual massless, partially massless and massive spin-2 fields in maximally symmetric spacetimes of arbitrary dimensions n > 3, such that the degrees of freedom are preserved in the flat limit. The action principles for the dual fields were also related to the standard ones for such field theories by building on the previous works [8–19]. See also Mignaco [20] for references on earlier works.

In this Letter we focus on the massless and partially-massless cases and formulate the field equations derived from the actions of Boulanger et al. [7] as twisted-duality relations. In the massless case, our twisted-duality relation—see Equation 2.34—generalizes to (A)dS backgrounds the twisted-duality relation written in Hull [21, 22] for linearised Einstein gravity in flat spacetimes. Our duality relation actually smoothly reproduces the latter duality relation in the flat limit, thanks to the crucial role played by the Stueckelberg fields.

In the case of a partially-massless spin-2 field [23], the twisted-duality relation that we obtain—see Equation 3.35—has a smooth flat limit that reproduces a couple of twisted-duality relations in flat background, one for a massless spin-2 field and the other for a massless spin-1 field, thereby correctly accounting for the degrees of freedom of a partially-massless spin-2 field. Moreover, keeping the cosmological constant non-zero and setting the dimension of spacetime to n = 4, our twisted-duality relation reproduces the one given in Hinterbichler [24], upon eliminating the Stueckelberg field.

Twisted-duality relations are interesting for many reasons. In particular they relate, for a pair of dual theories, the Bianchi identities of one system to the field equations of the dual one, and vice versa. In the present work, we show that the field equations of two dual theories are formulated as a twisted-duality equation, although we note that the latter is not obtained from a variational principle that is manifestly spacetime covariant. Forgoing the latter requirement, for linearised Einstein theory around flat spacetime Bunster et al. [25] gave an action principle that yields the twisted self-duality conditions as equations of motion, keeping the graviton and its dual on equal footing. Finally, let us mention that, for the fully nonlinear Einstein-Hilbert theory, an action principle was given in Boulanger and Hohm [26] where both the graviton and its dual appear inside the action, albeit not on an equal footing and together with extra auxiliary fields. For recent interesting works where twisted (self) duality relations play a central role and for more references, see Henneaux et al. [27–29] and Lekeu [30].

As for our conventions, we work on constant-curvature spacetimes with either negative or positive cosmological constant Λ. We denote the number of spacetime dimensions by n and define the quantity λ2=-2σΛ(n-1)(n-2), σ = ±1, that is always positive. When the background is AdSn one has σ = 1, while σ = −1 for dSn. The commutator of covariant derivatives gives [∇a,∇b]Vc=-σλ2(gacVb-gbcVa), where gab is the background (A)dSn metric. The symbols ϵa1 ⋯ an and ϵa1⋯an denote the totally antisymmetric tensors obtained from the corresponding densities upon multiplication and division by -g.

 2. Massless spin-2 twisted duality 2.1. Fierz-pauli formulation

In the Fierz-Pauli formulation for a massless spin-2 field around a maximally-symmetric spacetime of dimension n, the Lagrangian (where we omit the factor -g for the sake of conciseness) is given by

LFP=−12∇​ahbc∇ahbc+∇​ahbc∇chba+12∇​ah∇ah              −∇​ah∇bhab−(n−1)σλ22(2habhab−h2).(2.1)

It is invariant, up to a total derivative, under the gauge transformations1

δhab=2∇​(aξb) .(2.2)

The primary gauge-invariant quantity for the Fierz-Pauli theory is given by

Kab|mn=−12(∇a∇[mhn]b−∇b∇[mhn]a+∇m∇[ahb]n                  −∇n∇[ahb]m)+σλ2(ga[mhn]b−gb[mhn]a).(2.3)

It possesses the same symmetries as the components of the Riemann tensor,

K[ab|c]d≡0 ,(2.4)

and obeys the differential Bianchi identity

∇[aKbc]|mn≡0 .(2.5)

The field equations derived from the Lagrangian LFP imply the tracelessness of the curvature:

Kmn:=gabKma|nb≈0 ,(2.6)

where weak equalities are used throughout this paper to indicate equalities that hold on the surface of the solutions to the equations of motion. More precisely, defining K=gabKab, the left-hand side of the field equations read

δLFPδhab ≡−2(Kab−12gabK).(2.7)

By virtue of the differential Bianchi identity for the curvature, one also finds that, on-shell, the curvature has vanishing divergence:

∇mKmn|ab≈0 .(2.8)

To summarize, the important equations in this section are 2.4, 2.5, and 2.6. The latter relation was derived from the Lagrangian LFP. For the purpose of deriving a twisted-duality relation, we can actually forget the origin of 2.6 and focus on the three Equations 2.4, 2.5, and 2.6.

 2.2. Dual formulation

We start from the dual formulation of the massless spin-2 theory as given by the Lagrangian L0(Ŷ,W) in Equation (20) of Boulanger et al. [7]:

L0(Y^,W)=1λ2[12∇​cWabc|d∇eWdbe|a+λY^ab|c∇eWcbe|a                           +σ2(n−2)∇​bY^ab|c∇dY^cd|a+λ22Y^ab|cY^ac|b] .(2.9)

This Lagrangian describes the propagation of the same degrees of freedom as the Fierz-Pauli one in Equation 2.1. It has been built in two steps: a Lagrangian depending only on the field Ŷab|c (that is a traceless combination of the components of the spin connection) is obtained by eliminating the vielbein from the first-order formulation of linearised gravity in (A)dS. The full Lagrangian 2.9 then results from the Stueckelberg shift

Y^bc|a→Y^bc|a+1λ∇​dWbcd|a .(2.10)

The Lagrangian 2.9 possesses as many differential gauge symmetries as the Lagrangian obtained in Boulanger et al. [13] that describes dual (linearised) gravity in Minkowski background.

We now define the following quantities

Rab|cd:=2∇​[a|​(∇​eWcde|b]+λY^cd|b]),(2.11) Kab|d:=2∇​[a|∇cY^cd|b]+2σ(n−2)λ(∇​cWcd[a|b]+λY^d[a|b]),(2.12)

together with their various non-vanishing traces

Ra|c=Rab|cb ,    Ka=Kab|b .(2.13)

Further introducing the traceless tensor Vab|cd encoding the traceless projection of Rab|cd,

Vab|cd=Rab|cd−4n−2δ[c[aRb]|d] ,(2.14)

we find that Vab|cd is invariant under the following gauge transformations:

δY^bc|a ​​=​​∇​dζbcd|a​​+​​∇​aΛbc​​+​​2n−1δa[b∇​dΛc]d+(n−3)σλχbca,(2.15) δWbcd|a=∇​eυbcde|a+∇​aχbcd−3n−2δa[b∇​eχcd]e−λζbcd|a .(2.16)

Finally, the traceless tensor

Xab|c:=Kab|c+2n−1δc[aKb](2.17)

is also found to be gauge invariant.

As in Boulanger et al. [7], one can also express the fields W and Ŷ in terms of their Hodge duals, that we denote by C and T:2

Wabc|d=−1(n−3)!ϵe[n−3]abcCe[n−3]|d ,      Y^ab|d=−1(n−2)!ϵe[n−2]abTe[n−2]|d .(2.18)

The corresponding curvatures are obtained from the previous gauge-invariant tensors V and X as follows:

Ka[n−2]|bcC=12!ϵa[n−2]deVbc|de ,Ka[n−1]|bcT=ϵa[n−1]dXbc|d .(2.19)

In components, the curvature tensors read

Ka[n−2]|Cbc= 2(n−2)(−1)n−1∇[b∇​aCa[n−3]|c]+2λ∇[bTa[n−2]|c]+… ,(2.20) Ka[n−1]|Tbc= 2(n−1)(−1)n∇[b∇​aTa[n−2]|c]                      −2σλ(n−1)(n−2)2δ[ba∇​aCa[n−3]|c]                     ​+2σλ2(n−1)(n−2)Ta[n−2]|[bδac]​+… ,(2.21)

where the ellipses denote terms that are necessary to ensure GL(n)-irreducibility of the curvatures KCa[n−2]|bc and KTa[n−1]|bc on the two-column Young tableaux of types [n−2, 2] and [n–1, 2], respectively. Pictorially, they are represented by

Indeed, tracelessness of Vbc|de and Xab|c implies that the Hodge dual tensors Ka[n-2]|bcC and Ka[n-1]|bcT obey the following algebraic Bianchi identities:

Ka[n−2]|acC≡0 ,  Ka[n−1]|acT≡0 .(2.22)

The two curvatures are linked via the following differential Bianchi identities:

∇​aKa[n−2]|Cbc≡λ(−1)n(n−3)(n−1)(n−2)Ka[n−1]|Tbc ,(2.23) ∇[bKCa[n−2]|cd]≡λ1n−2KTa[n−2][b|cd] .(2.24)

These are equivalent to the following two identities:

∇​dVab|cd ≡ −λn−3n−2 Xab|c ,     ∇​[bVcd]|ij ≡ λ2n−2δ[b[iXcd]|j] .(2.25)

The equations of motion for the dual gauge fields Ca[n−3]|b and Ta[n−2]|b derived from the Lagrangian L0(C,T)—obtained by substituting 2.18 in 2.9 and given in Equation (23) of Boulanger et al. [7]—can be written in terms of the traces of the gauge-invariant curvatures Ka[n-2]|b[2]C and Ka[n-1]|b[2]T. Explicitly, one has

δL0δCa[n−3]|b≡1λ2(n−3)!(KCa[n−3]c|cb+n−32KCa[n−4]cd|cdgab)≈0 ,(2.26) δL0δTa[n−2]|b≡σλ2(n−2)2(n−3)!(KTa[n−2]c|cb+n−22KTa[n−3]cd|cdgab)≈0.(2.27)

The field equations 2.26 and 2.27 can easily be obtained by starting from the field equations of the Lagrangian L0(W,Ŷ) and then expressing the fields Wabc|d and Ŷab|c in terms of their Hodge duals Ca[n−3]|b and Ca[n−3]|b in terms of their Hodge duals Ta[n−2]|b and Ta[n−2]|b, respectively. More in details, the left-hand sides of the field equations derived from L0(W,Ŷ) read

δL0(W,Y^)δWabc|d=12λ2V[ab|c]d ,    δL0(W,Y^)δY^ab|d=−σ2(n−2)λ2Xab|d ,(2.28)

and the gauge invariant tensors X and V can be expressed as

Xab|d=−1(n−1)!ϵc[n−1]dKTc[n−1]|ab ,  Vab|cd=−1(n−2)!ϵe[n−2]cdKCe[n−2]|ab .

The field equations 2.26 and 2.27 imply the tracelessness of the curvatures:

KCa[n−3]c|cb≈0 ,KTa[n−2]c|cb≈0 .(2.30)

In fact, from a result in representation theory of the orthogonal group—see the theorem on p. 394 of Hamermesh [31]—, the second equation above implies that

KTa[n−1]|bc≈0 .(2.31)

The curvature for the field T thus vanishes on shell, consistently with the observation that this field does not propagate any degrees of freedom in the flat limit [7, 19].

Upon using the first and second differential Bianchi identities 2.23 and 2.24, we also find the following two relations that are true on shell:

∇​aKa[n−2]|Cbc≈0 ,    ∇[bKa[n−2]|Ccd]≈0 .(2.32)

These equations, together with 2.30, imply that the divergences of the curvature KC vanish on shell:

∇aKab[n−3]|cdC≈0 ,     ∇bKa[n−2]|bcC≈0 .(2.33)

To summarise, the important equations of this section are the equations of motion 2.30 and the Bianchi identities 2.22, 2.23 and 2.24. In the following section we will relate them to the field equations and the Bianchi identities of the Fierz-Pauli formulation via a twisted-duality relation.

 2.3. Massless twisted duality

The twisted-duality relations for the massless spin-2 theory around (A)dS backgrounds are

Ka[n−2]|bcC≈12ϵa[n−2]ijKij|bc .(2.34)

As usual for twisted-duality relations, the Bianchi identities in a formulation of the theory are mapped to the field equations of the dual formulation, and vice versa, as we now explain in details.

First, the algebraic Bianchi identity 2.22 for the left-hand side of the twisted-duality relation 2.34 implies that the trace of Kab|cd vanishes on-shell, which is the field equation 2.6 in the metric formulation. The converse is true: If one takes the trace of the relation 2.34, the right-hand side vanishes by virtue of the algebraic Bianchi identity 2.4. This implies that the trace of the left-hand side of 2.34 vanishes, which enforces the field equation 2.30 in the dual formulation.

Second, starting again from the twisted-duality equation 2.34, the differential Bianchi identity 2.24 on the second column of KC combined with the Bianchi differential identity 2.5 imply the on-shell vanishing of KT, that is, 2.31. Using this result, the differential Bianchi identity 2.23 on the first column of KC gives the first equation of 2.32 that implies in its turn, via 2.34, the field equation 2.8 in the metric formulation of the massless spin-2 theory. The converse is also true: acting on the twisted-duality relation 2.34 with ∇a gives identically zero, from the right-hand side and as a consequence of the differential Bianchi identity 2.5 for the curvature in the metric formulation of linearised gravity around (A)dS. This implies the first field equation 2.33 for the dual graviton. Moreover, acting on 2.34 with ∇d and antisymmetrising over the three indices {b, c, d} gives identically zero from the right-hand side of 2.34, as a consequence of 2.5. That implies the field equation 2.31 (and therefore the second field equation 2.32) by virtue of the identity 2.24. Finally, the field equation 2.8 is mapped to the second field equation in 2.33.

Third, the twisted-duality relation 2.34 exactly reproduces, in the limit where the cosmological constant goes to zero, the twisted-duality relations given by Hull [22] for linearised gravity in flat spacetime, see also section 4 of Bekaert and Boulanger [32].

 3. Partially-massless spin-2 twisted duality 3.1. Standard stueckelberg formulation

We consider the Stueckelberg Lagrangian for a partially-massless, symmetric spin-2 field in which both signatures are allowed (making AdS manifestly non-unitary at the classical level):

LPM=−12∇​ahbc∇ahbc+∇​ahbc∇chba+12∇​ah∇ah−∇​ah∇bhab              −(n−1)σλ22(2habhab​​−​​h2)+σ∇​[aAb]∇[aAb]​​+​​(n−1)λ2AaAa              −2m˜Aa(∇ah−∇​bhab)+σm˜2(habhab−h2) ,(3.1)

where the partially massless theory really appears in the limit

m˜2 → (n−2)λ22  .(3.2)

The last two lines in the expression 3.1 are new terms in comparison with the Lagrangian for a strictly massless spin-2 field in (A)dS, see 2.1. In the limit 3.2, the Lagrangian LPM is invariant, up to total derivatives, under the gauge transformations

δhab=2∇​(aξ b)+2m˜n−2gabϵ ,  δAa=∇aϵ+2σ m˜ ξa .(3.3)

The quantity

Hab=hab−σm˜∇(aAb)(3.4)

is invariant under the gauge transformations with parameter ξa, but not under the gauge transformations with parameter ϵ. A fully gauge-invariant quantity is provided by the antisymmetrised curl of Hab. Indeed, defining

Kab|c:=2∇​c∇​[aAb]−4σλ2gc[aAb]−4σ m˜∇​[ahb]c≡−4σm˜∇​[aHb]c ,(3.5)

we have that Kab|c is fully gauge invariant in the partially massless limit 3.2, hence so is ∇[aHb]c. We further define the derived quantity Qab|mn as follows:

Qab|mn​=​ ​−​12(∇a∇[mHn]b​​−​​∇b∇[mHn]a​​+​​∇m∇[aHb]n​−​∇n∇[aHb]m)                         +(1−2m˜2(n−2)λ2)σλ2(ga[mHn]b−gb[mHn]a) .(3.6)

It possesses the symmetries of the components of the Riemann tensor, like Kab|cd in the massless case. The second line of the above expression is identically vanishing in the limit 3.2, so that Qab|mn is indeed a composite object purely built out of the gauge-invariant quantity ∇[aHb]c. The writing that we adopted in 3.6 facilitates the relation between Kab|cd and Qab|cd. The interest in defining 3.6 rests in the fact that the field equations for hab read

δLPMδhab≡−2Gab , where Gab:=(Qac|bc−12gabQcd|cd) .(3.7)

As a consequence, the field equations for hab imply that the curvature Qab|cd is traceless on-shell, as it was for Kab|cd in the strictly massless case.

The Noether identities associated with the gauge parameter ξa give the left-hand side of the field equations for the vector Aa:

δLPMδAa≡−2σm˜ ∇bGab .(3.8)

The non-vanishing of the covariant divergence of Gab is also related to the Bianchi identity

∇[aQbc]|mn≡−m˜n−2δ[a[mKbc]|n] ,(3.9)

where the gauge-invariant quantity Kab|c was defined above in 3.5 and satisfies the identity K[ab|c]≡0. In terms of Kab|c, the left-hand side of the field equations for Aa reads

δLPMδAa≡σKab|b ,(3.10)

so that the field equations for Aa imply that the curvature Kab|c is traceless on-shell.

 3.2. Dual formulation

We now consider the dual formulation of the partially-massless spin-2 theory that is described by the Lagrangian LPM(W,U) in Equation (39) of Boulanger et al. [7]:

LPM(W,U)=−12λ2∇​dWbcd|a∇eWabe|c+σm˜Uabc∇dWabd|c                             −σ2(n−2)m˜2∇​cUabc∇dUabd−λ22m˜2UabcUabc .(3.11)

A Lagrangian depending only on the field Wabc|d has first been obtained by solving the equations of motion given by the variation of the vielbein in a first-order formulation of the partially-massless theory. In analogy with the massless case, the additional field Uabc has then been introduced by a Stueckelberg shift.

Starting from 3.11 one can define the following quantities

Rab|cd:=2∇​[a|​(∇​eWcde|b]−σλ2m˜U|b]cd),(3.12) Kab|Uc:=2∇​[a∇eUb]ce+2(n−2)m˜(∇​eWec[a|b]−σλ2m˜Uabc),(3.13)

together with the successive traces

Ra|c=Rab|cb ,  R=Ra|a≡0 ,  KaU=Kab|Ub≡0 .(3.14)

In a similar manner to the massless case, we introduce the traceless tensor Vab|cd according to

Vab|cd=Rab|cd−4n−2δ[c[aRb]|d] ,(3.15)

and we find that the tensors Vab|cd and KabcU are invariant under the following gauge transformations:

δWbcd|a=​∇​eυbcde|a​+​∇​aχbcd−3n−2δa[b∇​eχcd]e−σλ2m˜ρbcda ,(3.16) δUabc=∇​dρabcd−(n−3)m˜χabc.(3.17)

Also in this case, we then express W and U in terms of their Hodge duals

Wabc|d​=​−1(n−3)!ϵe[n−3]abcCe[n−3]|d ,​​​​Uabc​=​−1(n−3)!ϵd[n−3]abcAd[n−3].(3.18)

The curvature tensor for C is defined, as in the massless case, by

Ka[n−2]|bcC=12!ϵa[n−2]deVbc|de .(3.19)

We also define the curvature K˜a[n−2]| via

KUab|c=(−1)n−12(n−2)!ϵd[n−2][acK˜d[n−2]|b] .         (3.20)

In order to invert this relation, we first compute

(−1)n2ϵd[n−2]abKUab|c=K˜d[n−2]|c−(n−2)δdcK˜ed[n−3]|e(3.21)

and take the trace of the above relation, which produces

K˜ab[n−3]|a=14ϵb[n−3]cdeKUcd|e.(3.22)

Inserting this relation back in 3.21 gives

K˜a[n−2]|b=(−1)n2ϵa[n−3]cde(δaeKUcd|b−n−22δabKUcd|e) .(3.23)

Explicitly, we have

K˜a[n−2]|b=(n−2)(∇b∇aAa[n−3]+(n−2)m˜∇aCa[n−3]|b                     −σ(n−2)λ2gabAa[n−3]) ,(3.24)

which is gauge invariant under [7]

δCa[n−3]|b=(−1)n−1(n−3)(∇​aυ˜a[n−4]|b−σλ2m˜gbaρ˜a[n−4])                          +n−3n−2(∇​bχ˜a[n−3]+(−1)n∇​aχ˜a[n−4]b) ,(3.25) δAa[n−3]=(n−3)((−1)n−1∇​aρ˜a[n−4]−m˜χ˜a[n−3]) .(3.26)

The curvatures obey the following algebraic Bianchi identities

KCa[n−2]|ab≡0 ,K˜a[n−2]|a≡0 ,(3.27)

which means that KCa[n−2]|bc and K~a[n-2]|b are projected on the following GL(n)-irreducible Young tableaux

The left-hand sides of the equations of motion derived from the Lagrangian 3.11 are given by

δℒPMδWabc|d=12λ2V[ab|c]d ,δℒPMδUabc=σ2(n−2)m˜2KU[ab|c] .(3.28)

Combining with what we obtained above, the field equations therefore imply

K˜a[n−3]b|b≈0 ,KCa[n−3]b|bc≈0 .(3.29)

The Bianchi identities read

∇dVab|cd≡−σλ2(n−3)(n−2)m˜KUab|c ,∇[aVbc]|de≡2σλ2m˜(n−2)δ[d[aKUbc]|e] .(3.30)

In terms of the curvatures KC and K~, they become

∇aKCa[n−2]|bc≡(−1)n2σλ2(n−3)m˜(n−2)δ[baK˜a[n−2]|c] ,(3.31) ∇[aKCd[n−2]|bc]≡−2σλ2m˜K˜[ad[n−3]|bδc]d.(3.32)

By taking a trace of the Bianchi identity and using the field equations, one therefore deduces that

∇bKCa[n−2]|bc≈(−1)n(n−3)σλ2(n−2)m˜K˜a[n−2]|c ,(3.33) ∇bKCa[n−3]b|cd≈−2σλ2(n−2)m˜K˜[ca[n−3]|d].(3.34) 3.3. Partially-massless twisted duality

The twisted duality that mixes the field equations and Bianchi identities of the two dual theories, the one for LPM(hab,Aa) on the one hand, and the one for LPM(Ca[n-3]|b,Aa[n-3]) on the other hand, is

Ka[n−2]|bcC≈12ϵa[n−2]ijQij|bc .(3.35)

This equation plays the same role as 2.34 in the strictly massless case.

What is new in the partially massless case compared to the massless case is that the flat limit of 3.35 is not enough to describe all degrees of freedom of a partially massless field. In fact, the twisted-duality relation 3.35 also induces a duality relation between the curvatures K~a[n-2]|b and Kab|c. This can be viewed by acting on 3.35 with ∇a and contracting the result with ϵa[n−1]d. One then uses 3.31 and the trace of 3.9, taking into account that, on shell, the traces of the four curvatures K~a[n-2]|b, Ka[n-2]|bcC, Qab|cd and Kab|c vanish. We obtain

K˜a[n−2]|b≈(−1)n−1σ m˜24λ2ϵa[n−2]cdKcd|b ,(3.36)

where we stress that 3.35 and 3.36 are equivalent for non-zero cosmological constant.

Now, taking the flat limit of both 3.35 and 3.36, we obtain two decoupled twisted-duality relations for the two decoupled pairs of fields (Ca[n−3]|b, hab) and (Aa[n−3], Aa). Both together, they propagate the correct degrees of freedom for a partially massless spin-2 field in the flat limit, as was found and discussed in section 4.3 of Boulanger et al. [7]. The flat limit of 3.36 gives

∂bF˜a[n−2]≈(−1)n(n−2)σ8ϵa[n−2]cd∂bFcd ,(3.37)

where F~a[n-2]=(n-2)∂aAa[n-3] and Fab = 2∂[aAb] are the field strengths for Ab[n−3] and Ab, respectively. In the flat limit, these latter quantities are gauge invariant, therefore the gradient ∂b on both sides of the above relation 3.37 can be stripped off to give, up to an unessential coefficient that can be absorbed into a redefinition of Aa[n−3], the usual electric-magnetic duality between a 1-form and its dual (n−3)-form in dimension n.

As a consistency check for the second duality relation 3.36, one can start from the twisted-duality relation 3.35 and this time take the curl of KC on its second column of indices, which yields

∇[bKCa[n−2]|cd]≈12ϵa[n−2]ij∇[bQcd]|ij .(3.38)

We then use the Bianchi identities 3.32 and 3.9 and take a trace, taking into account the field equation 3.29, which allows us to obtain the relation

K˜a[n−3][b|c]≈(−1)n−1σ m˜24λ2ϵa[n−3]ij[bKij|c] ,(3.39)

which is fully consistent with 3.36.

Finally, we come back to the twisted-duality relation 3.36 and gauge fix to zero both Aa and Aa[n−3] since they are Stueckelberg fields as long as λ is different from zero. In these gauges for the dual formulations, our second twisted-duality relation 3.36 becomes

(n−2)∇aCa[n−3]|b≈(−1)n2ϵa[n−2]cd∇chdb ,(3.40)

while the first twisted-duality relation 3.35 is just its curl, as one can readily check. This duality relation makes immediate contact with the one proposed for the specific case n = 4 in Equation (2.3) of Hinterbichler [24]. Relation 3.40 identifies the dual curvature F~ab|c in Hinterbichler [24] with 4∇[aCb]|c, the curl of the dual potential Cb|c = Cc|b. Note that, once the Stueckelberg fields Aa and Aa[n−3] have been set to zero, one cannot take a smooth flat limit any longer in the sense that physical degrees of freedom are lost in the flat limit.

The advantage of our Stueckelberg formulation for the twisted-duality relation is that the identification of the helicity degrees of freedom is manifest and does not require any specific system of coordinates to be seen. In the original Stueckelberg formulation, hab and Aa carry the helicity two and one degrees of freedom, and the twisted-duality relations 3.35 and 3.36 identify these degrees of freedom with the dual fields Ca[n−2]|b and Aa[n−3], respectively, in a manifestly covariant way.

 Author contributions

All authors listed have made a substantial, direct and intellectual contribution to the work, and approved it for publication.

Conflict of interest statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

We performed or checked several computations with the package xTras [33] of the suite of Mathematica packages xAct. The work of NB has been supported in part by a FNRS PDR grant (number T.1025.14), while the work of AC has been supported in part by the NCCR SwissMAP, funded by the Swiss National Science Foundation. AC and LT are, respectively, a Research Associate and a Research Fellow of the Fund for Scientific Research-FNRS, Belgium.

 References Olive DI West PC. Duality and supersymmetric theories. Cambridge: Cambridge University Press (1999). Strassler MJ. On confinement and duality. In: Proceedings of the 2001 Spring School on Superstrings and Related Matters. Trieste (2001). Available online at: http://users.ictp.it/ ~pub_off/lectures/lns007/Strassler/Strassler.pdf Cremmer E Julia B. The SO(8) supergravity. Nucl Phys. (1979) B159:141. 10.1016/0550-3213(79)90331-6 Bekaert X Boulanger N Henneaux M. Consistent deformations of dual formulations of linearized gravity: a no go result. Phys Rev D (2003) 67:044010. 10.1103/PhysRevD.67.044010 Bekaert X Boulanger N Cnockaert S. No self-interaction for two-column massless fields. J Math Phys. (2005) 46:012303. 10.1063/1.1823032 Boulanger N Skvortsov ED. Higher-spin algebras and cubic interactions for simple mixed-symmetry fields in AdS spacetime. J High Energy Phys. (2011) 9:63. 10.1007/JHEP09(2011)063 Boulanger N Campoleoni A Cortese I. Dual actions for massless, partially-massless and massive gravitons in (A)dS. Phys Lett. (2018) B782:285–90. 10.1016/j.physletb.2018.05.046 Curtright TL Freund PG. Massive dual fields. Nucl Phys. (1980) B172:413–24. 10.1016/0550-3213(80)90174-1 Fradkin E Tseytlin AA. Quantum equivalence of dual field theories. Ann Phys. (1985) 162:31. 10.1016/0003-4916(85)90225-8 West PC. E(11) and M theory. Class Quant Grav. (2001) 18:4443–60. 10.1088/0264-9381/18/21/305 Zinoviev Y. On massive high spin particles in AdS. arXiv:hep-th/0108192. West PC. Very extended E(8) and A(8) at low levels, gravity and supergravity. Class Quant Grav. (2003) 20:2393–406. 10.1088/0264-9381/20/11/328 Boulanger N Cnockaert S Henneaux M. A note on spin s duality. J High Energy Phys. (2003) 6:60. 10.1088/1126-6708/2003/06/060 Matveev A Vasiliev M. Dual formulation for higher spin gauge fields in (A)dS(d). Phys Lett. (2005) B609:157–66. 10.1016/j.physletb.2005.01.032 Zinoviev Y. On dual formulations of massive tensor fields. J High Energy Phys. (2005) 510:75. 10.1088/1126-6708/2005/10/075 Zinoviev Y. On dual formulation of gravity. J High Energy Phys. (2006) 10:9. 10.1088/1126-6708/2006/10/009 Gonzalez B Khoudeir A Montemayor R Urrutia LF. Duality for massive spin two theories in arbitrary dimensions. J High Energy Phys. (2008) 9:58. 10.1088/1126-6708/2008/09/058 Khoudeir A Montemayor R Urrutia LF. Dimensional reduction as a method to obtain dual theories for massive spin two in arbitray dimensions. Phys Rev. (2008) D78:65041. 10.1103/PhysRevD.78.065041 Basile T Bekaert X Boulanger N. Note about a pure spin-connection formulation of general relativity and spin-2 duality in (A)dS. Phys Rev. (2016) D93:124047. 10.1103/PhysRevD.93.124047 Mignaco JA. Electromagnetic duality, charges, monopoles, topology, .… Braz J Phys. (2001) 31:235–46. 10.1590/S0103-97332001000200014 Hull C. Strongly coupled gravity and duality. Nucl Phys. (2000) B583:237–59. 10.1016/S0550-3213(00)00323-0 Hull C. Duality in gravity and higher spin gauge fields. J High Energy Phys. (2001) 109:27. 10.1088/1126-6708/2001/09/027 Deser S Nepomechie RI. Gauge invariance versus masslessness in de Sitter space. Ann Phys. (1984) 154:396. 10.1016/0003-4916(84)90156-8 Hinterbichler K. Manifest duality invariance for the partially massless graviton. Phys Rev. (2015) D91:026008. 10.1103/PhysRevD.91.026008 Bunster C Henneaux M Hörtner S. Twisted self-duality for linearized gravity in D dimensions. Phys Rev. (2013) D88:064032. 10.1103/PhysRevD.88.064032 Boulanger N Hohm O. Non-linear parent action and dual gravity. Phys Rev. (2008) D78:064027. 10.1103/PhysRevD.78.064027 Henneaux M Lekeu V Leonard A. Chiral tensors of mixed young symmetry. Phys Rev. (2017) D95:084040. 10.1103/PhysRevD.95.084040 Henneaux M Lekeu V Leonard A. The action of the (free) (4, 0)-theory. J High Energy Phys. (2018) 1:114. 10.1007/JHEP01(2018)114 Henneaux M Lekeu V Matulich J Prohazka S. The action of the (free) N = (3,1) theory in six spacetime dimensions. J High Energy Phys. (2018) 6:57. 10.1007/JHEP06(2018)057 Lekeu V. Aspects of Electric-Magnetic Dualities in Maximal Supergravity. PhD thesis, Université Libre de Bruxelles (2018). Hamermesh M. Group Theory and Its Application to Physical Problems (Dover Books on Physics). Dover Publications. Bekaert X Boulanger N. Tensor gauge fields in arbitrary representations of GL(D,R): duality and Poincare lemma. Commun Math Phys. (2004) 245:27–67. 10.1007/s00220-003-0995-1 Nutma T. xTras: a field-theory inspired xAct package for mathematica. Comput Phys Commun. (2014) 185:1719–38. 10.1016/j.cpc.2014.02.006

1Indices enclosed between (square) round brackets are (anti)symmetrised, and dividing by the number of terms involved is understood (strength-one convention). Moreover, we will use a vertical bar to separate groups of antisymmetrised indices, see e.g., Equation 2.3.

2We substitute groups of antisymmetrised indices with a label denoting the total number of indices, e.g., ϵa1⋯an≡ϵa[n]. Moreover, repeated indices denote an antisymmetrisation, e.g., AaBa≡A[a1Ba2].