Arxiv:1608.03021V1 [Gr-Qc]

Home , Ernst equation

arXiv:1608.03021v1 [gr-qc] 10 Aug 2016 rvtto.I hspprw hwta h formulation the that show we to of theories paper extends Wick general this or a other In whether performing to ask gravitation. extends by to formulation logical form this therefore not Ernst is It an dependent [16–18]. into time rotation cast but can be symmetric as can cylindrically [13], metrics form are GR Similarly, ‘Ernst’ which 15]. similar [14, a theories Brans-Dicke into some dimensions moulded also higher in be for has (GR) can one work relativity and general Recent that metric shown the 4-potential. complex- two for electromagnetic for one the equations functions, two equa- scalar to field valued reduced the be where can [12], tions theory Einstein-Maxwell [11]. the Ernst system scalar Einstein tensorial the the solving than to rather suited numerical equation many better advantages, are analytic techniques offering [8–10]. as solu- transformations well old other from As and generated Kinnersley variable be using Ernst can the tions solutions off of new directly moments and read [5–7], several multipole be can the offers spacetime example formulation the For Ernst [4]. The the equations advantages encode Einstein equation construction. the Ernst satisfy by the which to coefficients, solution com- metric imaginary a and of real of ponents The function scalar coordinates. differential complex-valued, spacetime partial the single, non-linear a a for Einstein to equation the vacuum reduce to in possible representation is useful equations it that especially showed an Ernst be [3]. to proved formula- has Ernst tion sta- the For spacetimes, 2]. axisymmetric [1, and ways tionary different many in rewritten been have o eiwof review a for olnaiisapa nteEnteuto eae othe to related equation Ernst the in appear nonlinearities rs omlto faiymti ed in fields axisymmetric of formulation Ernst ∗ † [email protected] [email protected] rs eto oso htti omlto xed to extends formulation this that show to on went Ernst equations Einstein the presentation, initial their Since f ASnmes 42.b 43.k 45.d 04.70.Bw 04.50.Kd, 04.30.Nk, 04.20.Jb, numbers: shock-w PACS display that solutions soliton-like of context the eeaie rs qain odrv a derive function to scalar equations complex-valued Ernst a generalised cylindricall for or equation stationary either differential is that spacetime vacuum fgaiy ti hw ht si eea eaiiy h ax the relativity, general in as that, shown is It gravity. of h hoyt netgt h hs pe flarge-amplitud of e an speed of phase outside the field investigate gravitational to the theory describe the to used be may ( R f h rs omlto fteEnti qain sgenerali is equations Einstein the of formulation Ernst The rvt nantrlwy e e.[19] Ref. see way; natural a in gravity ) ( .INTRODUCTION I. R hois ttrsotta additional that out turns It theories. ) colo hsc,Uiest fMlore akil I 3 VIC Parkville Melbourne, of University Physics, of School rhrGog Suvorov George Arthur n rvttoa waves gravitational and Dtd coe 0 2018) 30, October (Dated: f ( R eeaiaino h io-oresmti,which metric, Zipoy-Voorhees the of generalisation ) ∗ f 1 n nrwMelatos Andrew and pe a ederived. be can speed terms function forcing the and to dissipative related with equations wave linear qain fmto o h erc nScinI are II Section in metrics Ernst-like the the for III, motion Section Wick-rotated In of its II. equations and Section metric in Papapetrou counterpart the for revisit this into the falls metric does Kerr as the class, particular, In com- rotating object. a pact surrounding geometry example, the for represent contexts; they important physically arise numerous spacetimes in Stationary symmetric. cylindrically and adn h ul o-iercs.I hspprw use we paper this re- In known are case. cylindrical results non-linear the few fully However, the garding 25]. phase dependent [24, frequency velocities with equa- propagate Klein-Gordon thus and the tion satisfy waves gravitational that “point-like” cylindrical non-linear vacuum exact, in that waves show gravitational We further. lem function rls eet the to here eralise R nG.I a lobe hw nlinearised in shown time been light long also of a has speed for It the known GR. at in been propagate has waves solu- gravitational it wave that particular, gravitational In Cylindri- and tions. example, cosmological [23]. GR for valuable; include regime if also they field testing are as strong solutions symmetric such the cally non-GR reasons, in in of down variety behave breaks important a bodies is for compact It theories how 22]. understand [21, to solutions deformed-Kerr other ii ta.[6 7 n eest rcdr hrb n ass one t whereby by procedure spanned space a configuration to of the refers components with and independent Lagrangian 27] a [26, ciates introduced al. was et terminology this Ritis knowledge, authors’ the To aaersdb oriae.I hswyoeotisasyste [28]. freedom a of obtains degrees of one number way finite this a only In on depending coordinates. by parametrised ( R si R the GR, in As hr r w aor fEnteuto htw gen- we that equation Ernst of flavours two are There n t derivatives. its and ) v eaiu costecua boundary. causal the across behaviour ave ymti euet ige non-linear single, a to reduce symmetric y rvt:apiain onurnstars neutron to applications gravity: sawre xml,w pl the apply we example, worked a As . rvttoa ae in waves gravitational e isymmetric f lpodlnurnsa.W loapply also We star. neutron llipsoidal e oaccommodate to sed n t eiaie,a ela h ic scalar Ricci the as well as derivatives, its and 1 arnintcnqe 2–0,wihwe which [26–30], techniques Lagrangian f ( f R ( rs qaint nls hsprob- this analyse to equation Ernst ) f R f f 1,Australia 010, ( ( rs qain r eie using derived are equations Ernst ) R ( † R R -erNwa erc[0,and [20], metric )-Kerr-Newman edeutosfra for equations field ) f hoyo rvt:stationary, gravity: of theory ) g rmteeeutosaphase a equations these From . µν nta ftepyia spacetime physical the of instead f ( f R f ( R rvt in gravity ) ( R theories ) rvt bynon- obey gravity ) f ( R theory ) yde by he o- m 2

derived and are shown to reduce to their GR counter- ponents gµν leads to the vacuum f(R) field equations2 parts when f(R) = R. Equipped with the generalised (e.g. [38]) Ernst equations, we work through a simple, formal example in Section IV to demonstrate how one may use f(R) the Ernst formulation to derive new exact metrics. This 0= f ′(R)Rµν gµν + [gµν µ ν ] f ′(R), (3) idealised example is potentially useful for studying ellip- − 2 − ∇ ∇ soidal compact objects like neutron stars, although its utility is mainly formal at the time of writing. In Section where R = Rα is the Ricci tensor, and = µ V we use the time-dependent Ernst equation to inves- µν µαν µ symbolises the d’Alembert operator. ∇ ∇ tigate some properties of large-amplitude gravitational waves in f(R) theories, in particular their speed of prop- Instead of working with the physical spacetime, one agation. The results are discussed in Section VI. can express the action (2) directly in terms of the configuration variables U,ω,B, and γ, and their first derivatives with respect to the spacetime coordinates. For static, II. EQUATIONS OF MOTION IN f(R) GRAVITY spherically symmetric metrics in f(R) gravity, a set of field equations equivalent to (3) has been derived by con- We derive the point-like Lagrangian associated with figuration space techniques [29, 30, 39]. In our case, we a stationary spacetime in Section II A. The formalism are considering the metric (1), and so our configuration for the cylindrically symmetric case, which is completely variables are U,ω,B,γ, and their first derivatives with analogous, is covered in Section II B. respect to ρ and z. The Ricci scalar is uniquely determined by the metric coefficients. This information can be self-consistently A. Stationary spacetime absorbed into the action (2) by imposing a constraint equation. To this end, we introduce a Lagrange multi- Following Ernst [3, 12], we consider a stationary, ax- plier λ [29–31], isymmetric spacetime endowed with the Weyl-Lewis- Papapetrou line element in Weyl coordinates t,ρ,φ,z { } A = d4x√ g f(R) λ R R¯ , (4) [32–34], − − − Z ds2 = U 1 e2γ dz2 + dρ2 + B2dφ2 U (dt ωdφ)2 , − ¯ − − (1) where R is the Ricci scalar expressed explicitly in terms where U, ω, B, and γ are functions of ρ and z only. of the configuration variables (as opposed to R which is In GR, it was shown by Papapetrou that the vacuum to be thought of as a function of the spacetime coordinates). Variation of the action (4) with respect to the Einstein equations imply B,ρρ + B,zz = 0 (see also below) [32]. Hence one can always adopt a set of harmonic configuration variables then leads to the equations of mo- ¯ coordinates t, ρ,¯ φ, z¯ with the properties dz2 dz¯2, tion subject to the constraint R = R. For our case, with dρ2 dρ¯2,{ and B }=ρ ¯ [35]. Therefore, in GR,7→ the respect to (1), we find function7→ B is redundant and the number of free functions reduces to three without loss of generality. In f(R) 1 U 2 3B √ gR¯ = ∇B ∇U + ∇ω ∇ω ∇U ∇U gravity this transformation is not always possible because − U · 2B · − 2U 2 · the equations governing the variable B are more compli- B cated, so we must use the more general line element (1) 2∇2B 2B∇2γ + ∇2U, − − U [13, 36, 37]. It should be noted that the particular form (5) of the metric (1) holds in vacuum, and a more general where the operator ∇ forms a 2-gradient with respect form may be required when considering arbitrary matter to the embedded 2-dimensional metric dσ2 = dz2 + dρ2, 2 sources. i.e. we have ∇α = (α,z, α,ρ) and ∇ α = α,zz + α,ρρ for The f(R) theory of gravity is a natural generalisa- any scalar function α(z,ρ). In equation (5), R¯ is a function of GR, where the Ricci scalar, R, appearing in the tion of the configuration variables U,ω,B,γ, and their Einstein-Hilbert action, is replaced by an arbitrary func- derivatives. We obtain λ = f ′(R) by varying the action tion of this quantity, f(R). The f(R) action reads (4) with respect to R. Any second order terms (e.g. U,ρρ) can be removed from the action (4) through integration A = d4x√ gf(R). (2) by parts, and total divergence terms may be removed − Z by invoking Gauss’s theorem (see [38] and Appendix A for details). The Lagrangian, being the integrand of the Variation with respect to the contravariant metric com- action (4), reads

2 Throughout, Greek symbols range over spacetime indices 0, 1, 2, 3, while Latin indices are reserved for spatial indices 1, 2, 3. 3

2γ L e B f ′(R) 2∇ ∇ 4∇ ∇ 2∇ ∇ = [f(R) Rf ′(R)] + 2 4BU B γ + U ω ω B U U U − 2BU · · − · (6) f ′′(R) + 2U [∇R ∇B + B∇R ∇γ] B∇U ∇B , U · · − · n o where we have made use of relations (A2)–(A4) de- are not written down explicitly here, because they are rived in Appendix A. Equation (6) reduces to the La- presented in a simpler form in Section III. Variation of grangian used by Ernst upto ignorable divergence terms L with respect to γ, (see above equation (4) in Ref. [3]) in the special case f(R)= R,B = ρ. The point-like ﬁeld equations may now be written down in their entirety by varying the Lagrangian (6) with ∂L ∂ ∂L 0= i , (9) respect to U,γ,ω, and B, as well as their derivatives. The ∂γ − ∂x ∂γ,i equations of motion for U and ω [28],

∂L ∂ ∂L 0= i , (7) ∂U − ∂x ∂U,i yields an integrability condition for the f(R) theory and not a diﬀerential equation for γ, because the Lagrangian and (6) depends only linearly on derivatives of γ. If we have 3 L L Rf ′(R)= f(R), as in GR, γ becomes a cyclic coordinate ∂ ∂ ∂ L 0= i , (8) for the Lagrangian . Evaluating (9) explicitly we ﬁnd ∂ω − ∂x ∂ω,i

2γ e B 2 2 0= [Rf ′(R) f(R)] + f ′(R)∇ B + f ′′(R) 2∇B ∇R + B∇ R + Bf ′′′(R)∇R ∇R. (10) U − · ·

Equation (10) demonstrates a significant difference between theories with f(R) = R and GR. When one has 6 2 Rf ′(R) = f(R), γ can be deduced from the variables U In the GR limit, equation (12) also reduces to ∇ B = 0. and B by6 inverting equation (10). As a result, in some For GR with nonzero cosmological constant, where we ways, the f(R) field equations admit a simpler structure have f(R)= R 2Λ, equation (12) reads than GR for the metric (1). In GR, equation (10) reads − 2 2γ 2 ∇ B = 0, and does not constrain γ. However, since the e− U∇ B + 2ΛB =0, (13) Ricci scalar must be fixed as zero in GR, equation (5) fills the role of a differential equation for γ given U, ω, and B which is a Helmholtz equation for B [41, 42]. In this case, (solved for through equations (7), (8), and (10), respec- equation (10) is again identical to (13), and the degrees tively) subject to appropriate boundary conditions. In of freedom in the system are reduced self-consistently; R either case, we have four equations in four variables; see is still fixed (with value R = 4Λ), so again (5) is an equa- also [40] and equations (13.8) in Ref. [1]. tion for γ rather than for R, and equation (13) becomes redundant. After some manipulations, the equation of motion for Equation (12) demonstrates the importance of keeping B, the function B in the line element (1) in general for f(R) gravity. If we were to take B = ρ, equations (10) and ∂L ∂ ∂L (12) immediately tell us that there are no f(R) solutions 0= , (11) i parametrisable by the Papapetrou metric (1) that admit ∂B − ∂x ∂B,i R = R = constant = 0 unless f(R )= R f ′(R )=0. It 0 6 0 0 0 is well known that f(R) gravity with R = R0 = constant reads, is equivalent to the Einstein equations with effective cos- f(R0) ′ mological constant Λeff = 2f (R0) [19], provided that 2γ ∇2 ∇ ∇ f ′(R0) = 0. Therefore, there are no GR solutions with 0=e− B f ′(R) B f ′′(R) B R 6 − · Λeff = 0 for B = ρ. h 2 6 2 B f(R) In an f(R) theory where R is not constant, equation 2f ′′(R)B∇ R 2Bf ′′′(R)∇R ∇R + . − − · U (10) can be used to eliminate γ from equation (12) re- (12) i sulting in an equation relating B and f that reads 4

∇2 2f(R) B 2 2 0=Rf ′(R) f ′(R)∇ B + f ′′(R)∇B ∇R +2Bf ′′(R)∇ R +2Bf ′′′(R)∇R ∇R R − · · (14) ∇ ∇ 2 B R Bf(R) f ′′(R)∇ R f ′′(R) · + Bf ′′′(R)∇R ∇R . − − B ·

Expressions (7), (8), (10), and (14) are not much sim- B. Cylindrically symmetric spacetime pler than equation (3). In Section III we show how one can reduce the expressions obtained above down to a sim- Consider the Jordan-Ehlers-Kompaneets line element pler Ernst form. [43],

2 1 2γ 2 2 2 2 2 ds = U − e dt + dρ + B dφ +U (dz ωdφ) , − − (15) where nowf, γ, B, and ω are functions of t and ρ. Though we keep the same set of configuration variables, the line element (15) is of a fundamentally different structure to the Papapetrou metric (1), and describes different physical scenarios (see Sections IV and V). Following the procedure in the previous section, we find that the integrand of the action (4) for the metric (15) reads

e2γ B f (R) ′ 2 / / 4 / / 2 / / = [f(R) Rf ′(R)] + 2 4BU B γ U ω ω B U U L U − 2BU ∇ · ∇ − ∇ · ∇ − ∇ · ∇ (16) f ′′(R) + 2U / R / B + B / R / γ B / U / B , U ∇ · ∇ ∇ · ∇ − ∇ · ∇

where the complex 2-operator / acts on scalar functions III. ERNST EQUATION ∇2 α(t,ρ) as / α = (iα,t, α,ρ) and / α = α,ρρ α,tt. In particular, the∇ operator / is formally∇ related− to ∇ through A. Stationary spacetime the Wick rotation z ∇ it (see Ref. [44] for a discussion 7→ of Wick rotations in curved spacetime). Though the line The Ernst equation (equation (2) in Ref. [3]) is re- elements (15) and (1) are different, and are introduced in markably simple, because the GR Lagrangian (5) does unconnected contexts, we see that the Lagrangians (16) not depend on the generalised position ω explicitly, and (6) are equivalent under the Wick rotations z it meaning that ω is a cyclic coordinate, which implies that and t iz and the identification ω iω. As a7→ re- 7→ − → the associated momentum is conserved [28]. This is also sult, the equations of motion, namely equations (7), (8), true for the f(R) case, as can be seen from expression (10), and (14), are also identical to the equations of mo- (6). In particular, the field equation (8) reads tion for the metric (15), provided one replaces ∇ with / in each of the expressions and writes iω in place of ω U 2 ∇ / 0= ∇ f ′(R)∇ω . (17) [16, 17, 28]. Furthermore, since the operator only ap- · B pears quadratically in the Lagrangian (16), we∇ have that is strictly real. Although the operators ∇ and / are In GR, when B is fixed as B = ρ, equation (17) con- L ∇ 1 formally related by a complex Wick rotation, the func- tains the coordinate factor ρ− . Ernst showed that one tions U,ω,γ, and B appearing in (15) are real functions of may introduce a potential function ϕ related to ω which real coordinates. Similar Wick rotation techniques have removes this coordinate dependency [3, 12]. Such a con- been applied in the GR case to transform stationary and struction is possible in f(R) theories when B = ρ and is axisymmetric solutions into cylindrically symmetric and discussed in Appendix B. However, in general, for B = ρ, time dependent ones [45–47]. equation (17) is already coordinate independent, because6 ∇ is defined as the covariant derivative with respect to the 2-metric dσ2 = dz2 +dρ2 (and not with respect to the cylindrical 3-metric dΣ2 = dz2+dρ2+B2dφ2, which is not flat when B = ρ, as it is in Ernst’s work [3, 12]). Any co- 3 Something similar happens in spherical symmetry; see §3 of [29]. ordinate transformations6 involving z or ρ self-consistently 5

modify the ∇ operator through the Christoﬀel symbols. The Ernst equation in GR is obtained by constructing As a result, in f(R) gravity, we do not need to, in general, a complex equation, where the vanishing of the real com- introduce the variable ϕ. The reader who is more famil- ponent implies (7) and the vanishing of the imaginary iar with the usual GR construction of the Ernst equation component implies (17) [3]. involving ϕ can make use of the equations presented in Appendix B to express equation (17) and others in terms We can obtain an Ernst-type equation for f(R) gravity of ϕ rather than ω [see equations (B4) and (B5)]. We by introducing a complex-valued function4 E˜ = U + iω, elect instead to express our results in terms of ω to avoid making use of equation (10), and recasting both equa- coordinate terms appearing in the general case B = ρ. tions (17) and (7) into a single equation for E˜, 6

E˜ 2 E˜ 2 Re( ) Re( ) 2 0=f ′′(R)Re(E˜) BRe(∇E˜) Re(E˜)∇B ∇R + i ∇ f ′(R) Im(∇E˜)+ f ′(R)Im(∇ E˜) − · ( " B # · B ) h i (18) E˜ 4 Re( ) 2 2 + f ′(R) Im(∇E˜) Im(∇E˜) + Re(E˜)∇ BRe(∇E˜) Re(E˜) ∇ B BRe(∇E˜) Re(∇E˜) , ( B · · − − · ) h i

which is to be solved for E˜ given B and f. Equation (18) equivalence between certain f(R) and scalar-tensor the- generalises the Ernst equation to f(R) gravity. ories of gravity [19, 48] (see [49] for a dissenting view how- As a consistency check, if we set f(R) = R, B = ρ, ever). The Ernst equation (18) reduces to known scalar- and introduce the potential ϕ through equation (B4) (see tensor forms under an appropriate conformal transforma- Appendix B), then equation (18) reduces correctly to the tion [14, 15]. In particular, we recover equations (3.3a)– GR Ernst equation (in our notation) (3.3c) of reference [14] and equations (16a) of reference [15] (with the exception of the Maxwell fields; see the discussion in Sec. VI) as a subcase of equation (18), where ∂ 0 = Re(E ) ∇2 + ρ E ∇E ∇E , (19) the f(R) theory is identified with a scalar-tensor theory ρ − · with a massless scalar field in the Jordan frame (see also Sec. 10.1 of [19]). We recover the scalar-tensor quadra- with E = U + iϕ. ture relations for γ, equations (3.3d)–(3.3e) of [14], from To solve the f(R) field equations in practice we may equation (5) together with (18). proceed as follows. First, choose an ansatz for the function f and scalar curvature R to investigate the properties of a particular theory of gravity. In principle, equa- B. Cylindrically symmetric spacetime tion (5) can be applied to eliminate R in the Ernst equation (18) and all other equations appearing in Section The Ernst formulation derived in the previous section II. However, if one wishes to look for solutions that are can also be applied to cylindrically symmetric, time- asymptotically flat, specifying a suitably decaying R a dependent spacetimes. The field equation for ω under priori results in a simpler, decoupled system. The linear the line element (15), which is equivalent to (8) under equation (14) can be integrated (in principle) to uniquely the maps z it and t iz, reads 7→ 7→ − determine B given any choices of R and f. In turn, if B is known, the Ernst equation (18) can be solved for U and ω. Finally, the remaining metric coefficient γ can be U 2 immediately determined using equation (10). The metric 0= / f ′(R) / ω . (20) ∇ · B ∇ is now completely constructed, and one need only check that the constraint equation (5) holds. If equation (5) does not hold, the implication is that no f(R) spacetime, Furthermore, the field equation for U reads the same as parametrisable by the Papapetrou metric (1), exists for the real part of (18) but with / in place of ∇ and a sign the initial ansatz. flip in ω terms. The Ernst equation∇ for a cylindrically It is worth emphasising that there is a well-studied symmetric spacetime in f(R) gravity is then

4 E˜ Note that the function will not be complex diﬀerentiable in gen- of a source in any physically reasonable spacetime, but one must be eral since it does not satisfy the Cauchy-Riemann equations (e.g. cautious when seeking to apply complex analysis techniques (e.g. for static solutions one ﬁnds ω = 0 but U =6 constant except for the residue theorem) to E˜. Minkowski spacetime). Both U and ω are real and smooth outside 6

2 2 Re(E˜) Re(E˜) 2 0=f ′′(R)Re(E˜) BRe( / E˜) Re(E˜) / B / R + i / f ′(R) Im( / E˜)+ f ′(R)Im( / E˜) ∇ − ∇ · ∇ (∇ " B # · ∇ B ∇ ) h i (21) E˜ 4 Re( ) 2 2 + f ′(R) − Im( / E˜) Im( / E˜) + Re(E˜) / BRe( / E˜) Re(E˜) / B BRe( / E˜) Re( / E˜) , ( B ∇ · ∇ ∇ · ∇ − ∇ − ∇ · ∇ ) h i

E˜ 2 2 2 2 with = U + iω. If we let f(R) = R, B = ρ, and (1+ ǫ) (R+ + R ) 4(1+ ǫ) /M introduce the Wick-rotated potentialϕ ˆ through (B7), we γZV = log − − , 2 " 4R+R # obtain the equation, − (25)

2 ∂ 0 = Re(E ) / + ρ E / E / E , (22) ω =0, (26) ∇ ρ − ∇ · ∇ which is a known cylindrical variant of the Ernst equation B = ρ, (27) (see equation (22.5) in Ref. [1]). There is an important distinction between equations with (18) and (21). Since t is a time-like coordinate, the latter equation is hyperbolic, while the former is elliptic. This R = ρ2 + [z (1 + ǫ) /M]2, (28) ± ± may have some implications regarding the stability of q numerical codes designed to solve such equations (e.g. where M is the mass of the object, and ǫ is the (formally [50]). Nevertheless as in Section III A, the real functions arbitrary) ellipticity parameter. In particular, ǫ> 0 cor- U and ω defining the metric (15) may be determined from responds to an object more oblate than a Schwarzschild the real and imaginary components of E˜, respectively. black hole, ǫ < 0 corresponds to a more prolate object, ǫ = 0 reduces the metric functions to the Schwarzschild ones, and ǫ = 1 reduces the metric functions to the − IV. WORKED EXAMPLE: ELLIPSOIDAL Minkowski ones [51, 53]. For the Zipoy-Voorhees metric NEUTRON STARS we have Rµν = 0. One possible way to search for a suitable generalisation We consider here a simple example of an f(R) the- of any GR metric is to fix one of the metric functions ory to demonstrate the method presented. Specifically, to be the same as their GR counterpart and see if the we search for a solution which generalises the Zipoy- structure of the f(R) theory allows for variation in the Voorhees metric of GR [51]. The latter metric represents other metric components. As a simple example, we make the spacetime exterior to a static compact object which is the simplifying assumption that γ is unchanged from its not spherically symmetric. It tends to the Schwarzschild GR counterpart in (25), i.e. γ = γZV. Searching for solution, when the ‘oblateness’ parameter tends to zero. solutions where the function f has power-law form [55] For example, the metric could describe the gravitational f(R)= f Rα, (29) field outside a neutron star that, through magnetic or 0 other internal stresses, has become deformed [52–54]. for some constant α, we find that the only possible solu- We begin by assuming that the Ricci scalar takes the tions compatible with equations (10) and (12) are ones simple form with R = 0, ∇2B = 0 and α 1. This result is one of 0 ≥ non-existence; for R0 = 0, there does not exist a Γ which 2 2 Γ 6 R = R0 ρ + z , (23) allows a power-law f(R) solution with γ = γZV (though there are non-trivial solutions with R = 0 which we de- 0 where R0 is a constant, and we demand either Γ 1 or rive below). Since ∇2B = 0 we may take B = ρ, as in ≤−5 R0 = 0 to obtain an asymptotically flat spacetime . The GR, without loss of generality. static Zipoy-Voorhees line element takes the form of (1) Suppose we introduce the ansatz with the definitions 2Q 1+ǫ U = e UZV, (30) R+ + R 2(1+ ǫ) /M UZV = − − , (24) R+ + R +2(1+ ǫ) /M for some function Q which tends to zero at infinity (so − that gtt tends to unity). The Ernst equation (18) may be written down in full, though the expressions are lengthy, so we avoid them here. However, if we further assume 5See the discussion surrounding equation (19) in [39] for a general α> 1, so that we work within the realm of strictly non- discussion on sufficient decay conditions required on R for asymp- GR theories, then the Ernst equation (18) is satisfied for totic flatness. any choices of ω and Q. As such, we have that the Ernst 7 equation (18), equation (14) for B, and equation (10) for V. PHASE SPEED OF NONLINEAR γ are all satisfied for the above choices. The remaining GRAVITATIONAL WAVES equation is the consistency relation for the Ricci scalar, equation (5), which forms an eikonal equation for ω, In this section we demonstrate a physical application of the Ernst equation (21) to gravitational waves. In par- 4e 4Qρ2 ∇ ∇ − ∇ ∇ ticular, we examine the behaviour of freely propagating, ω ω = 3 UZV Q · UZV " · nonlinear gravitational waves (solitons) in a vacuum f(R) (31) theory. Gravitational waves are often studied within the Q + U ∇Q ∇Q ∇2Q ,ρ . framework of perturbation theory, whereby the linearised ZV · − − ρ # theory, valid far away from the source, provides both an Equation (31) is subject to Dirichlet boundary condi- equation for the wave amplitude and a dispersion relations, i.e. ω must vanish at infinity. Clearly ω = Q = 0 tionship which allows for the definition of a phase speed is a solution to (31), which simply reproduces the Zipoy- (e.g. [60]). However, such an analysis does not neces- Voorhees solution. Equation (31) suggests that there is a sarily extend to the nonlinear theory, as nonlinearities great deal of freedom in obtaining rotating (or static) can introduce modified dispersion relations or dissipa- generalisations of the Zipoy-Voorhees metric in f(R) tion mechanisms (compare the Korteweg-de Vries equa- gravity. It is well known that the Dirichlet eikonal equation [61], for example). The analyses of Einstein, Rosen, tion (31) admits unique solutions for ω for any well- and others demonstrated that the nonlinearities of the behaved choice of Q (e.g. [56]). As such, there are in- field equations of GR do not allow for phase speeds dif- finitely many generalisations of the Zipoy-Voorhees met- ferent from the speed of light [62, 63]. Perturbation the- ric, each uniquely corresponding to a particular choice ory in f(R) gravity, however, demonstrates that grav- of the function Q (in contrast to GR, where the Ernst itational waves have frequency-dependent phase speeds equation (19) further restricts the choices of Q). As an in general [64] (this is true even in GR with nonzero example, if we take cosmological constant [65–67]). A nonlinear analysis is lacking for the general f(R) theory mainly because of Q = ln (1 σUZV) , (32) the absence of exact solutions describing gravitational − − where σ is an arbitrary constant, we obtain another static waves [68]. By using the Ernst formalism presented in solution with ω = 0 since the right hand side of (31) Section III for cylindrically symmetric, time-dependent vanishes. As can be verified directly by substitution, the metrics, we can construct gravitational wave solutions to metric given by (24)–(32) does solve the f(R) field equa- the nonlinear theory. Specifically, we construct a solution tions (3) with R = 0 for any constant σ, but has non- which has an arbitrary phase speed for a particular choice vanishing Ricci tensor unless σ = 0. In the zero ellipticity of f. While this does not represent a full treatment of limit, ǫ 0, we obtain the Reissner-Nordström metric the large-amplitude problem, it does suggest that phase [57]. A physical→ interpretation of σ is not readily avail- speeds other than the speed of light are possible in f(R) able without performing some additional analysis, i.e. by gravity, as the linear perturbation theory in f(R) gravity constructing the multipole moments and matching them implies. with a suitable Newtonian solution [7, 20]. Such an anal- Some immediate observations can be made by swap- ysis will be performed elsewhere. It is easy to see that ping the variable U for ψ defined through the relation ψ = 1 ln U and letting B = ρb for some function b(t,ρ). the function UZV from (24) is bounded for any ǫ 1, − 2 and so we may take σ small if necessary to ensure≥ that − The real and imaginary parts of equation (21) read, re- 1 spectively [f ′(R) = 0], UZV < σ− everywhere, so that the presence of Q does 6 not introduce singularities into the spacetime. 4ψ 2 2 ∂ e / b / b / ψ It is likely that more general metrics that include the 0= / + ρ ψ / ω / ω ∇ + ∇ · ∇ ∇ ρ − 2ρ2b2 ∇ · ∇ − 2b b Zipoy-Voorhees metric as limiting cases exist, where the form of the Ricci scalar differs from (23). In particular, b,ρ f ′′(R) / b / R R,ρ the choice made in (23) resulted in the somewhat trivial + / R / ψ ∇ · ∇ , − ρb f ′(R) ∇ · ∇ − 2b − 2ρ property R0 = 0. Several other choices, such as taking 2 2 (33) (ρ +z ) the simple exponential R e− , appear to lead to and the same non-existence result.∝ In any event, the metric 2 ∂ 2ω given by (24)–(32) can be used to describe the metric 0= / + ρ ω ,ρ +4 / ω / ψ ∇ ρ − ρ ∇ · ∇ exterior to deformed neutron stars in f(R) gravity. The (34) presence of σ (and Q) indicates that neutron stars are / b / ω f (R) ′′ / / arbitrarily ‘hairy’ in f(R) gravity; parameters other than ∇ · ∇ + R ω, − b f ′(R) ∇ · ∇ their mass and angular velocity influence their properties as seen by observers at infinity [20, 58]. Neutron stars are which form a coupled set of non-linear hyperbolic wave 2 1 also known to be hairy in scalar-tensor theories of gravity, equations. The second-order piece, / + ρ− ∂ρ = ∂tt + 1 ∇ − so the equivalence between certain Brans-Dicke and f(R) ∂ρρ + ρ− ∂ρ, corresponds to the flat-space wave oper- theories supports the conclusions outlined above [59]. ator in cylindrical coordinates. The fact that the met- 8

ric functions obey wave equations demonstrates explicitly for some β 0, equations (36)–(38) yield the solution that f(R) theories predict the existence of gravitational ≥ 2 2β 2 2 1/3 2√1 12β/3 waves [25, 69]. In particular, for the GR case f(R)= R, ds = ( u)− dt + dρ + A0 ( u)− − − restoring dimensional factors of c shows that the waves − − − 2 √1 12β 2 2 propagate at the speed of light [62]. ( u)dz + ( u) − ρ dφ , × − − Let us now introduce the retarded time u = t κρ (40) − h i for some κ > 0, and assume that all metric functions where A0 > 0 is an arbitrary amplitude, which could be ψ,γ,b, and ω are functions of u only, as for a traditional fixed by specifying a wave amplitude at some point in ‘soliton’ solution. The constant κ is effectively the phase space at t = 0. We have u < 0 inside the causal cone, speed of the gravitational wave; it describes the rate at and so the metric (40) is real with Lorentzian signature which disturbances propagate in the spacetime. We con- for t < κρ provided 0 β 1/12. The metric (40) fine the metric to the interior of the causal cone C given is singular across the causal≤ boundary≤ ∂C (i.e. u = 0), by C = (t,ρ,φ,z): t κρ , as is typical of gravi- as can be seen from the divergence of the tt-component { ≤ } tational wave solutions in GR [70–73]. Outside of the of the metric, but it is smooth for all t < κρ. The so- causal cone, i.e. for t > κρ, we set6 γ = ω = ψ = 0 lution (40) is similar to the simplest Belinski-Zakharov and b = 1 (Minkowski space). In this way we construct one-soliton solution of GR [70, 71], which represents the a spacetime that has a discontinuous wave front repre- late time behaviour of a particular Einstein-Rosen pulse senting a propagating gravitational wave in an otherwise profile [62, 73]. empty universe. The metric functions may suffer discon- While only a toy model which is unlikely to describe tinuities in their derivatives on the boundary of the causal a real gravitational wave, the metric (40) demonstrates cone like gravitational shock waves (see below). Setting that the phase speed, κ, of gravitational waves, may take κ to unity results in the causal cone coinciding with the arbitrary values in particular f(R) theories. More com- light cone. It has been proved that one must have κ =1 plicated solutions can be built by considering different in GR (e.g. [76]). However non-GR theories may permit functional forms for f using the machinery developed in κ to be either greater than unity (superluminal) or less Sec II and Sec III. than unity (subluminal). Simple solutions of the above form can be constructed by taking VI. DISCUSSION α f(R)= f0R . (35) In this paper we derive two generalised Ernst equa- For α > 1, we find that the Ernst equation (21), the tions for the f(R) theory of gravity in the special cases Wick-rotated equations (10) and (12) for γ and B, and of stationary and cylindrically symmetric spacetimes. We the constraint equation (5) are satisfied for explicitly derive a class of simple solutions for each case individually and verify that the associated metrics do ω(u)=0, (36) indeed solve the f(R) field equations. As a physical and application, we show that is possible to generalise the Zipoy-Voorhees metric of GR to f(R) theories [51]. The b(u) = exp [ψ(u)/2] , (37) Zipoy-Voorhees metric describes the gravitational field around an ellipsoidal compact body. The generalisation provided ψ satisfies the Riccati equation describes a similar object but with some added ‘hair’, i.e. some additional parameters other than mass and angular 0=2ψ¨(u) 3ψ˙(u)2 4¨γ(u), (38) − − momentum which appear in the metric coefficients. Ad- where an overhead dot refers to differentiation with re- ditionally, we construct a simple time-dependent metric spect to the retarded time u. For example, the solution which seeks to approximate a large-amplitude gravita- ψ = γ = 0 yields the Minkowski metric everywhere. It tional wave with arbitrary propagation speed. In GR, can be easily verified by direct substitution that metrics it is well known that gravitational waves must travel at (15) satisfying the equations (36)–(38) solve the f(R) the speed of light. However, in an f(R) theory, small- field equations (3) for any α> 1. amplitude wave solutions exist which have either sub- or If we set super-luminal propagating wave fronts [64, 65, 77]. The small-amplitude result is generalised to arbitrary ampli- 2γ 2ψ 2β e − = ( u)− , (39) tude here for a particular, time-dependent, cylindrically − symmetric metric. Although the result is restricted to this particular metric, it may open a path to more general results in future work. 6In general, matching conditions at the boundary of the causal cone impose boundary conditions on the metric functions [74]. We do The f(R) Ernst equations (18) and (21) offer a few not consider the details of the matching procedure here, as they advantages over the usual tensor system (3). The Ernst are not germane to the question of the phase speed (however see equations, while still nonlinear, are more decoupled than [75]). (3). The decoupling arises naturally because of the con- 9

figuration variable approach, which isolates the equations fields, e.g. in the vicinity of highly magnetised compact of motion for each metric coefficient. Furthermore, be- objects [78–81]. cause of the decoupling, there is a sequential recipe for Finally, it is worth noting that the Ernst formulation solving these equations, namely for the variable B, fol- outlined here is not unique to the f(R) theory of grav- lowed by U and ω, and finally for γ. Aside from the ity. Indeed, it applies to any metric theory of gravity practical value in obtaining exact solutions, the Ernst for- that generalises GR and admits a point-like description, mulation reveals something about the underlying struc- for which the procedures outlined in Sections II and ture of the f(R) field equations. For example, there III can be replicated. In particular, it can be verified exists a complex Wick rotation that transforms neatly by direct calculation that theories of gravity whose La- µν between solutions for compact bodies and gravitational grangian is a function of the curvature invariants Rµν R µναβ waves. The Lagrangians associated with the Papapetrou or Rµναβ R , have point-like counterparts indepen- (1) and the Jordan-Ehlers-Kompaneets line elements (15) dent of ω, i.e. ∂L /∂ω = 0 for either of the parametri- are also related by a Wick rotation [45], despite having sations (1) or (15). Such theories include generalised been introduced in different contexts. Gauss-Bonnet gravity or the one-loop quantum corrected We speculate without proof that the formulation pre- version of GR [82, 83]. sented here extends to the f(R)-Maxwell theory, along the lines of Ernst’s work on the Einstein-Maxwell theory [12]. If such an extension can be found, it will be in- ACKNOWLEDGEMENTS teresting to see how the additional nonlinearities in the f(R) field equations interact with the electromagnetic We thank Peter Farrell for discussions. We thank the field. Following the outline presented in Section V, it anonymous referee for their helpful suggestions which im- may also be interesting to investigate the properties of proved the clarity of this manuscript. This work was sup- gravitational waves in the presence of electromagnetic ported in part by an Australian Postgraduate Award.

[1] J B. Griffiths and J. Podolsky, Exact spacetimes in Ein- [19] A. de Felice and S. Tsujikawa, Living Rev. Relativity 13, stein’s General Relativity, (Cambridge University Press, 1002 (2010). Cambridge, 2009). [20] A. G. Suvorov and A. Melatos, Phys. Rev. D 93, 024004 [2] R. A. Matzner and C. W. Misner, Phys. Rev. 154, 1229 (2016). (1967). [21] T. Johannsen and D. Psaltis, Phys. Rev. D 83, 124015 [3] F. J. Ernst, Phys. Rev. 167, 1175 (1968). (2011). [4] B. K. Harrison, Phys. Rev. Lett. 41, 1197 (1978). [22] T. Johannsen, Phys. Rev. D 88, 044002 (2013). [5] G. Fodor, C. Hoenselaers, and Z. Perjés, J. Math. Phys. [23] K. Glampedakis and S. Babak, Classical Quantum Grav- 30, 2252 (1989). ity 23, 4167 (2006). [6] T. P. Sotiriou and T. A. Apostolatos, Classical Quantum [24] S. Capozziello, C. Corda, and M. F. De Laurentis, Phys. Gravity 21, 5727 (2004). Lett. B 669, 255 (2008). [7] G. Pappas and T. P. Sotiriou, Phys. Rev. D 91, 044011 [25] C. P. L. Berry and J. R. Gair, Phys. Rev. D 83, 104022 (2015). (2011). [8] W. Kinnersley, J. Math. Phys. 14, 651 (1973). [26] R. de Ritis, G. Marmo, G. Platania, C. Rubano, P. Scud- [9] N. R. Sibgatullin, Oscilations and Waves in Strong Grav- ellaro, and C. Stornaiolo, Phys. Rev. D 42, 1091 (1990). itational and Electromagnetic Fields, (Nauka, Moscow, [27] S. Capozziello and R. de Ritis, Classical Quantum Grav- 1984; English translation: Springer-Verlag, Berlin, 1991) ity 11, 107 (1994). [10] T. P. Sotiriou and G. Pappas, Journal of Physics: Con- [28] V. I. Arnold, Mathematical Methods of Classical Mechan- ference Series 8, 23 (2005). ics, (Springer-Verlag, New York, New York, 1998). [11] C. Klein and O. Richter, Ernst Equation and Riemann [29] S. Capozziello, A. Stabile, and A. Troisi, Classical Quan- Surfaces: Analytical and Numerical Methods, (Springer- tum Gravity 24, 2153 (2007). Verlag, Berlin, Heidelberg, 2005). [30] L. Sebastiani and S. Zerbini, Eur. Phys. J. C 71, 1 (2011). [12] F. J. Ernst, Phys. Rev. 168, 1415 (1968). [31] P. Teyssandier and P. Tourrenc, J. Math. Phys. 24, 2793 [13] C. Charmousis, D. Langlois, D. Steer, and R. Zegers, J. (1983). High Energy Phys. 2, 64 (2007). [32] A. Papapetrou, Annalen der Physik 447, 309 (1953). [14] T. Tsuchida and K. Watanabe, Prog. Theor. Phys. 101, [33] S. Chandrasekhar, The Mathematical Theory of Black 73 (1999). Holes, (Oxford University Press, Oxford, 1998). [15] P. Kirezli and O.¨ Delice, Phys. Rev. D 92, 104045 (2015). [34] R. Konoplya, L. Rezzolla, and A. Zhidenko, Phys. Rev. [16] G. Bossard, H. Nicolai, and K. S. Stelle, Gen. Rel. Grav. D 93, 064015 (2016). 41, 1367 (2009). [35] H. Weyl, Ann. Physik 54, 117 (1917). [17] D. Korotkin and H. Nicolai, Nucl. Phys. B 429, 229 [36] A. Azadi, D. Momeni, and M. Nouri-Zonoz, Phys. Lett. (1994). B 670, 210 (2008). [18] F. J. Hernández, F. Nettel, and H. Quevedo, Gravitation [37] D. Momeni and H. Gholizade, Int. J. Modern Phys. D and Cosmology 15, 109 (2009). 18, 1719 (2009). 10

[38] S. Capozziello and V. Faraoni, Beyond Einstein Gravity: [64] A. A. Starobinsky, JETP Letters 86, 157 (2007). A Survey of Gravitational Theories for Cosmology and [65] I. Ozsváth, I. Robinson, and K. Rózga, J. Math. Phys. Astrophysics, (Springer Publishing, New York, 2010). 26, 1755 (1985). [39] S. Capozziello, A. Stabile, and A. Troisi, Classical Quan- [66] J. Näf, P. Jetzer, and M. Sereno, Phys. Rev. D 79, 024014 tum Gravity 25, 085004 (2008). (2009). [40] F. D. Ryan, Phys. Rev. D 52, 5707 (1995). [67] J. Bernabeu, D. Espriu, and D. Puigdoménech, Phys. [41] N. O. Santos, Classical Quantum Gravity 10, 2401 Rev. D 84, 063523 (2011). (1993). [68] M. Sharif and Z. Yousaf, Astrophysics and Space Science [42] M. Astorino, J. High Energy Phys. 06 (2012) 086. 357, 1 (2015). [43] P. Jordan, J. Ehlers, and W. Kundt, Akad. Wiss. Mainz. [69] D. Christodoulou, Phys. Rev. Lett. 67, 1486 (1991). Math. Naturwiss. Kl. 2, 21 (1960). [70] V. A. Belinski and V. E. Zakharov, Sov. Phys. JETP 48, [44] M. Visser, Gravity Research Foundation (unpublished), 985 (1978). (1991). [71] A. Tomimatsu, General Relativity and Gravitation 21, [45] G. Beck, Z. Physik 33, 713 (1925). 613 (1989). [46] N. Sanchez, Phys. Rev. D 26, 2589 (1982). [72] T. Piran, P. N. Safier, and R. F. Stark, Phys. Rev. D 32, [47] B. J. Carr and E. Verdaguer, Phys. Rev. D 28, 2995 3101 (1985). (1983). [73] A. Ashtekar, J. Biˇcák, and B. G. Schmidt, Phys. Rev. D [48] T. P. Sotiriou, Classical Quantum Gravity 23, 5117 55, 687 (1997). (2006). [74] W. Israel, Nuovo Cimento 44B, 1 (1966). [49] S. Capozziello, P. Martin-Moruno, and C. Rubano, Phys. [75] R. H. Gowdy, Phys. Rev. Lett. 27, 826 (1971). Lett. B 689, 117 (2010). [76] S. W. Hawking and G.F.R Ellis, The Large Scale Struc- [50] J. Celestino, H. P. de Oliveira, and E. L. Rodrigues, Phys. ture of spacetime, (Cambridge University Press, Cam- Rev. D 93, 104018 (2016). bridge, 1975). [51] D. M. Zipoy, J. Math. Phys. 7, 1137 (1966). [77] M. E. S. Alves, O. D. Miranda, and J. C. N. de Araujo, [52] K. Boshkayev, H. Quevedo, and R. Ruffini, Phys. Rev. D Phys. Lett. B 679, 401 (2009). 86, 064043 (2012). [78] F. C. Michel, Rev. Mod. Phys. 54, 1 (1982). [53] L. Herrera, F. M. Paiva, and N. O. Santos, J. Math. Phys. [79] A. G. Suvorov, A. Mastrano, and U. Geppert, Mon. Not. 40, 4067 (1999). R. Astron. Soc. 459, 3407 (2016). [54] H. Quevedo, S. Toktarbay, and Y. Aimuratov, Int. J. [80] A. G. Suvorov, A. Mastrano, and A. Melatos, Mon. Not. Math. Phys. 3, 133 (2012). R. Astron. Soc. 456, 731 (2016). [55] T. Clifton and J. D. Barrow, Physical Review D 72, [81] I. T. Drummond and S. J. Hathrell, Phys. Rev. D 22, 103005 (2005). 343 (1980). [56] H. Ishii, Proc. American. Math. Society 100, 247 (1987). [82] B. Li, J. D. Barrow, and D. F. Mota, Phys. Rev. D 76, [57] L. Richterek, J. Novotný, and J. Horský, Czechoslovak 044027 (2007). Journal of Physics 52, 1021 (2002). [83] G. t‘Hooft and M. Veltman, Annales de l’IHP Physique [58] D. D. Doneva, S. S. Yazadjiev, and K. D. Kokkotas, Phys. théorique 20, 69 (1974). Rev. D 92, 064015 (2015). [59] G. Pappas and T. P. Sotiriou, Mon. Not. R. Astron. Soc. 453, 2862 (2015). Appendix A: Calculation of the Lagrangian L [60] K. S. Thorne, Rev. Mod. Phys. 52, 299 (1980). [61] P. D. Lax, Communications on pure and applied mathe- matics 21, 467 (1968). The derivation of the Lagrangian (6), for the Papa- [62] A. Einstein and N. Rosen, J. Franklin Inst. 223, 43 petrou metric (1), comes through several applications (1937). of integrations by parts, after which total divergence 1 2γ [63] J. J. Stachel, J. Math. Phys. 7, 1321 (1966). terms are discarded. Noting that √ g = U − e B and − λ = f ′(R), the definitions (4) and (5) give us

2γ 2γ 4 e B e B A = d x f(R) Rf ′(R) ( U − U Z (A1) 2 1 U 3B B 2 2 2 f ′(R) ∇B ∇U + ∇ω ∇ω ∇U ∇U + ∇ U 2B∇ γ 2∇ B . − U · 2B · − 2U 2 · U − − )

The action (A1) contains second order derivative terms, bility [19]. In general, we have the elementary formula which must be removed to avoid the Ostrogradsky insta- for well behaved X and Y , 11

2 dzdρY ∇ X = dzdρY X,zz + dzdρY X,ρρ (A2) Z Z Z = dρY X dzdρY X + dzY X dzdρY X (A3) ,z − ,z ,z ,ρ − ,ρ ,ρ Z Z Z Z = dzdρ (YX ) + (YX ) Y X Y X . (A4) ,z ,z ,ρ ,ρ − ,z ,z − ,ρ ,ρ Z h i

The ﬁrst two terms in the integrand in equation (A4) are equivalent to those for the Lagrangian are total divergence terms. Hence, for any X and Y , these terms can be removed from the action (A1) without L˜ = ∇Y ∇X. (A6) modifying the equations of motion [29], i.e. the equations − · of motion for the Lagrangian Making use of relation (A4) and expanding the integrand L = Y ∇2X, (A5) in (A1) we have

e2γB e2γ B ∇B ∇U U 2 3B L = f(R) Rf ′(R) f ′(R) · + ∇ω ∇ω ∇U ∇U (A7) U − U − U 2B · − 2U 2 · f ′(R)B f ′(R)B + ∂ U + ∂ U 2∂ [f ′(R)B] γ 2∂ [f ′(R)B] γ 2∂ [f ′(R)] B 2∂ [f ′(R)] B z U ,z ρ U ,ρ − z ,z − ρ ,ρ − ρ ,ρ − z ,z (A8) 2γ e B f ′(R) 2 4 2 = [f(R) Rf ′(R)] + 4BU ∇B ∇γ + U ∇ω ∇ω B ∇U ∇U (A9) U − 2BU 2 · · − · f ′′(R) + 2U [∇R ∇B + B∇R ∇γ] B∇U ∇B . (A10) U · · − · n o

Equation (A10) is precisely the form of the Lagrangian coordinates ρ,φ,z as (6). Following the procedure presented here, a similar { } 2 Lagrangian could be derived for the case when the metric U 0= ∇3 f ′(R)∇3ω , (B1) variables depend on an arbitrary number of coordinates. · ρB where ∇3 is the usual cylindrical 3-gradient (e.g. [1, 33]). The well-known identity for any diﬀerentiable function ϕ independent of φ (e.g. [3])

Appendix B: Coordinate independence of the Ernst 1 0= ∇ ρ− φˆ ∇ ϕ , (B2) equations 3 · × 3 implies that there exists a ‘potential’ ϕ such that ∇ Our notation in this article for the operator diﬀers U 2 from Ernst’s original presentation [3] because we allow f ′(R)∇3ω = φˆ ∇3ϕ, (B3) for a slightly more general line element in (1) (i.e. we B × do not demand B = ρ). When B = ρ, the GR Ernst where φˆ is the unit vector in the azimuthal direction. In equation (19) appears to have a coordinate dependency particular, the relation (B3) is equivalent to 1 due to the ρ− ∂ρ term. As Ernst showed, such terms may ∇ B be removed by introducing the cylindrical 3-gradient 3 ∇ ϕ = φˆ ∇ ω, (B4) ∇ 2 3 3 (as opposed to the 2-gradient ) and a new variable ϕ in f ′(R)U − × place of ω such that terms may be removed to write an equation which respects covariance. For completeness, which implies that equation (17) may be written as we show that the same is true for f(R) gravity. ∂ B ∂ B 0= ϕ + ϕ . (B5) One can re-write equation (17) in terms of cylindrical ∂ρ ρf (R)U 2 ,ρ ∂z ρf (R)U 2 ,z ′ ′ 12

The variable ϕ generalises the quantity introduced by is equivalent to Ernst (see equation (6) of [3]) to f(R) gravity. If B = ρ, as it must be in GR, equation (B5) reads

∇ϕ 0= ∇ , (B6) · f (R)U 2 ′ ∂ B ∂ B 0= ϕˆ ϕˆ (B9) which does not contain any coordinate dependent terms. ∂ρ ρf (R)U 2 ,ρ − ∂t ρf (R)U 2 ,t ′ ′ However, in general, B is a function of ρ and z which is / ϕˆ unknown a priori, so introducing ϕ is unnecessary. = / ∇ . (B10) ∇ · f (R)U 2 In cylindrical symmetry, the above analysis carries ′ over. We deﬁne a potentialϕ ˆ obeying B / φˆ / 2 3ϕˆ = 3ω. (B7) f ′(R)U ∇ − × ∇ If one is interested in f(R) solutions such that B = ρ Hence equation (20), viz. is ﬁxed, substituting the variable ϕ through (B4) or its U 2 Wick-rotated counterpartϕ ˆ through (B7) ensures that 0= / f ′(R) / ω , (B8) the resulting equations are coordinate insensitive. ∇3 · ρB ∇3