Constraints on Lorentz invariance violation from gamma-ray burst GRB090510 Zhi Xiao and Bo-Qiang Ma∗ School of Physics and State Key Laboratory of Nuclear Physics and Technology, Peking University, Beijing 100871, China We obtain modified dispersion relations by requiring the vanishing of determinant of inverse of modified photon propagators in Lorentz invariance violation (LIV) theory. Inspired by these dispersionrelations,wegiveamoregeneraldispersionrelationwithlessassumptionandapplyitto therecentobservedgamma-rayburstGRB090510toextractvariousconstraintsonLIVparameters. Wefindthattheconstraint on quantumgravity mass isslightly larger than thePlanck mass butis consistent with other recent observations, so the corresponding LIV coefficient ξ1 has reached the naturalorder(O(1))asoneexpects. Fromouranalysis, thelinearLIVcorrectionstophotongroup velocity might be not excluded yet. 0 PACSnumbers: 11.30.Cp,11.30.Er,11.55.Fv,98.70.Rz 1 0 2 I. INTRODUCTION n a J Lorentz invariance violation (LIV or LV) has been intensively investigated both theoretically and experimentally 8 in recent years. The revival passion of relativity violation in theoretical construction originates from the attempt to 1 compromise general relativity with quantum mechanics. On the other hand, the experimental searches may provide us with concrete evidence to sift a most hopeful candidate of quantum gravity from a vast number of theories. ] From theoretical aspect, some theories expect LIV to happen at high energies. For example, spontaneous Lorentz h symmetry breaking may happen in string theory as the perturbative string vacua is unstable, thus some tensor fields p - generatenonzerovacuumexpectationvalues[1]. ThebreakingofLorentzsymmetryalsohappensinotherframeworks, p such as loop gravity [2], foamy structure of spacetime [3], torsion in general gravity [4], etc.. More recently, Ho˘rava e proposedapowercountingrenormalizabletheoryofgravity[5]witha“dynamicalcriticalexponent”z tocharacterize h [ the anisotropic scaling properties between space and time. While Lorentz symmetry is breaking at high energies, it restores when this dynamical critical exponent flows to z = 1 at low energies. There are also some other proposals, 6 such as the so called double special relativity [6], which preserves relativity principle with a nonlinear realization v of Lorentz group, thus conventional Lorentz symmetry is also broken. One striking consequence of LIV is that the 7 photon propagation speed is no longer a unique constant, generally, it depends on energy and propagation direction. 2 9 These theoretical investigations have promoted various experiments to search for the deviation from conventional 4 linear dispersionrelationfor photons [7]. However,as the possible violationeffects for photons must be very tiny, the . detection of these effects present a significant challenge to experimentalists. In addition to improve the precision of 9 0 measurements to find any possible evidence of LIV, we should also take efforts on searching for certain accumulating 9 processestoamplifythesetinyeffects. Suchideahasalreadybeenproposedontheobservationofcertainastronomical 0 objects such as gamma-ray bursts (GRB) [3, 8], pulsars [9] and active galactic nuclei (AGN) [10], etc., and the tiny : LIV effect could manifest itself through the observation of rotation of linear polarization (birefringence) [11] or time v i of flight lag [8] for photons with different energies. X Thepaperisorganizedasfollows. InSection2,wereviewcertainmodifiedphotondispersionrelationsderivedfrom r several LIV models, including standard model extension (SME) with power counting renormalizable operators [11], a effective field theory with dimension 5 operators [12] and Ho˘rava’s anisotropic U(1) theory. In Section 3, we focus on timeofflightanalysisofGRBandtrytoextractsomeLIVparametersfromtherecentobservationofGRB090510[13]. We briefly discuss the time of flight analysis of photons from cosmological distant objects, then we give a general dispersion relation used conventionally in the astrophysical analysis of LIV [10]. This general dispersion relation contains those terms derivedfrom the models in Section 2 as special cases. We then extractconstraintsto linear LIV parametersfromGRB090510to O(0.1), improvedby 1 or2orderofmagnitude thanthose in[14]and [10,15]. From the analysis of the time-lag formula we find that it is hard to significantly improve the constraints from this simple and rough analysis, unless other time-lag effects (like source effect [16], which is a major uncertainty in the time of flight analysis) can be clarified or other methods will be used. ∗ Correspondingauthor. Emailaddress: [email protected] 2 II. PHOTON DISPERSION RELATIONS A. Background tensor field induced LIV A systematical treatment of LIV to incorporate particle standard model with power counting renormalizable La- grangian, called standard model extension (SME), was proposed by Kostelecky´ and Colladay in Ref. [11], where the photon sector reads 1 1 1 L =− F Fµν − (k ) FκλFµν + (k ) ǫκλµνA F . (1) photon µν F κλµν AF κ λ µν 4 4 2 From (1), we deduce the equation of motion below ∂αF +(k ) ∂αFβγ +(k )αǫ Fβγ =0. (2) µα F µαβγ AF µαβγ By expressing (2) in terms of 4-vector potential A and assuming that the Fourier decomposition µ A (x)≡ a (p)exp(−i p·x) (3) µ µ is still reliable, we express (2) in the momentum space as M (p)aν(p)=0, (4) µν where M (p)=η p2−p p −2(k ) pκ pλ−2i(k )κǫ pλ. (5) µν µν µ ν F µκλν AF µκλν We impose gauge fixing condition and require the determinant of the reduced matrix to vanish, then we can obtain an implicit function p0(p~), which is an eighth order-polynomialin p0. Otherwise one can verify that the determinant of M vanishes as a consequence of gauge invariance of equation (2). For example, we use Lorentz gauge µν ∂ Aα =0, (6) α in momentum space, i.e., p aα(p)=0. (7) α So we have the gauge fixed reduced matrix Mgf(p)=η p2−2(k ) pκ pλ−2i(k )κǫ pλ. (8) µν µν F µκλν AF µκλν For our purpose, we just try to extracta simplified result by assuming that (particle) rotationalinvariance still holds regardless of the explicit violation of Lorentz symmetry, i.e., only (k )0 and α (a combination of (k ) , for AF F κλµν details, see Appendix or [17]) are nonzero. With this assumption, we have the following matrix: Mred(p)= p2−αp~2 αp0p1 αp0p2 αp0p3 αp0p1 −(p2+α((p0)2+~p2−(p1)2)) αp1p2+2ik0 p3 αp1p3−2ik0 p2  AF AF .(9) αp0p2 αp1p2−2ik0 p3 −(p2+α((p0)2+~p2−(p2)2)) αp2p3+2ik0 p1 AF AF  αp0p3 αp1p3+2ik0 p2 αp2p3−2ik0 p1 −(p2+α(p0)2+p~2−(p3)2))  AF AF  Its determinant reads det(Mred(p))= 4(k0 )2p~2− (1+α)(p0)2−(1−α)p~2 2 (1+α)(p2)2. (10) AF n (cid:0) (cid:1) o By requiring det(Mred(p)) = 0 (otherwise there would be no solution for a photon field), we have two dispersion relations, one is the conventional p2 =0 and the other is 1 (p0)2 = (1−α)p~2±2k |p~| . (11) AF (1+α) (cid:0) (cid:1) 3 We can also use another equivalent method to obtain these two dispersion relations. First, we rewrite (1) in an explicitly quadratic form in the photon field, i.e., 1 1 1 1 L =− F Fµν − (k ) FκλFµν + (k ) ǫκλµνA F − (∂·A)2 photon µν F κλµν AF κ λ µν 4 4 2 2ξ 1 ηµν =− ∂ A Fµν +2(k )κλµν∂ A + ∂·A−2ǫκλµν(k ) A µ ν F κ λ AF κ λ 2 (cid:18) ξ (cid:19) 1 =total derivative+ A (D−1)νλA , (12) 2 ν F λ where we have added gauge fixing terms − 1 (∂·A)2 in (12) and 2ξ 1 (D−1)νλ ≡ (cid:3)ηνλ−∂ν∂λ(1− )−2(k )νµκλ∂µ∂κ−2ǫνµκλ(k ) ∂κ . (13) F (cid:18) ξ F AF µ (cid:19) Using the same Ansatz (3), we define a matrix Σ in momentum space 1 Σ(p) =− p2η −(1− )p p +2(k ) pµpκ−2iǫ pµ(k )κ. (14) νρ νρ ν ρ F νµκρ νµκρ AF (cid:18) ξ (cid:19) From the conventionalfree field theory, the differential operator inside the two fields in the quadratic form of certain Lagrangian(e.g., (12)) is just the inverse of free field propagatorin position space (see [18] or [19] ), thus (14) is just the inverse of photon propagator expressed in momentum space. We know that generally the inverse of propagator is just the dispersion relation, from which one can find the pole of the corresponding particle, so we expect that the determinant of (14) in case of ξ → ∞ (i.e., without gauge fixing) is zero. Similarly, we can find the explicit dispersion relation in a special gauge by choosing the corresponding specific value of ξ. For example, we find that Σ(p) | = −Mgf(p) (ξ = 1 is just the Lorentz gauge (6) used to obtain (9), and this choice can avoid the νρ ξ=1 νρ inequivalentgaugechoicecomparison. As pointedoutin [11],different gaugechoicesareinequivalentwith eachother intheLIVelectrodynamics). Sointherotationalinvariantcase,thismatrixcanalsoleadto(11)andtheconventional dispersion relation. We mention here that similar method to obtain photon propagator in the SME framework has also been obtained recently in [20], with a more systematic and complete treatment. TheleadingordernonrenormalizableLIVoperators(dimension5)weresystematicallystudiedin [12],whereMyers and Pospelov also introduced explicitly a timelike four-vector na to take LIV into account, thus this theory can be regardedas a leading nonrenomalizable part of SME. Since we are only interested in the study of the consequence of LIV to the propagationof GRB, we focus our attention only on photon field there. The corresponding Lagrangianis ξ δL = ǫµνκρnαF n·∂(n F ). (15) photon αρ κ µν 2M Pl We write it in another equivalent form, i.e. 1 2ξ L = A (cid:3)ηνρ− n·∂(n·∂ n ∂ ǫνµκρ+n ∂ ∂ ǫνµκαnρ) A +total derivative, (16) photon ν κ µ κ µ α ρ 2 (cid:18) M (cid:19) Pl where we have added the Lorentz gauge fixing term. Then by performing the same procedure as before, we have the reduced inverse of propagator 2iξ Π(p)νρ =−p2ηνρ− (ǫνµ0ρp2p +ǫνµ0αp p p δρ) (17) M 0 µ 0 µ α 0 Pl when expressing explicitly the time-like four-vector n in a preferred frames as nρ =(1,0,0,0). Then by imposing −p2 0 0 0  0 p2 −i 2ξ (p0)2p3 i 2ξ (p0)2p2  2ξ detΠ(p)=det MPl MPl ) =p4 ( ~p)2(p0)4−p4 =0, (18)  00 −iiM2P2ξξl(p(p00)2)2pp32 i 2ξ (pp20)2p1 −iM2Pξlp(2p0)2p1  (cid:18) MPl (cid:19)  MPl MPl  we obtain the dispersion relation 2ξ (p0)2 =p~2± (p0)2|p~|, (19) M Pl which was obtained in [12] plus the conventional one p2 =0. 4 B. Anisotropic scaling induced LIV Now we turn to another framework of LIV proposedrecently by Ho˘rava [5]. His originalproposalwas to provide a UVcompletionofquantumtheoryofgravity. Lorentzsymmetryappearsnaturallyinthistheorywhenthedynamical critical exponent flows to z =1 at low energies. While at high energies, space and time present anisotropic scaling t→λzt, ~r →λ~r, (20) thusLorentzsymmetrybreaksdown. However,thisformalismdoesnotbreakspatialisotropy,thusthereisnoneedto assumeaspecialbackgroundfieldconfigurationtorealizerotationalinvariance,unlikethebackgroundtensorformalism discussed above. Aside from gravity, Ho˘rava also constructed an anisotropic Yang-Mills theory with critical spatial dimension D =4 [5]. As Chen and Huang recently gave a generalconstruction of bosonic field theory demonstrating this anisotropic scaling behavior [21], we follow this new approach instead of [5]. In the new formalism, the photon action reads 1 1 1 nJ λ S = dtdDx E~2− (−1)n J,n ∂2n⋆FJ. (21) 2Z gE2 JX≥2gEJ−2 nX=0 M2n+12(D+1)(J−2)   For simplicity, we consider the case z = 2 and D = 3. Then one immediately reads from the action that the scaling dimensions of the couplings are 1 1 [g ] = (z−D)+1, [λ ] =z+D+ (z−D−2)J −2n. (22) E s J,n s 2 2 Thus in this case the renormalizablecondition ([g ] ≥0:z ≥ D−2) for E~ is automatically satisfied. Actually, it is E s superrenormalizable. Ifthecriticaldimension(i.e.,[g ] =0)isD =3,thenz mustequalto1,whichjustcorresponds E s to the conventionalLorentz invariantgauge theory. Renormalizability also imposes the condition [λ ] ≥0, and for J,n s a free field theory, J =2, n≤ z−1=1. For simplicity, we set λ = 1, then the free Lagrangian(with gaugefixing 2,0 2 term) is 1 1 λ 1 L = E~2− F Fij − 2,1(∂ F ·∂ F +∂ F ·∂ F ) − (∂·A)2 free g2 (cid:18) 2 ij M2 i ik j jk i jk i jk (cid:19) ξ E 1 1 3λ = A (cid:3)ηνρ−∂ν∂ρ(1− ) − 2,1∆(∆δ −∂ ∂ )δνδρ A . (23) 2g2 ν(cid:26)(cid:18) ξ (cid:19) M2 kj k j k j(cid:27) ρ E By performing the same trick, we can obtain 3λ Γ(p)νρ =−p2ηνρ+ 2,1δνδρ(p p −p~2δ ), (24) M2 k j k j jk and the corresponding dispersion relations read 3λ p2 =0, (p0)2 =p~2(1+ 2,1~p2). (25) M2 III. TIME OF FLIGHT ANALYSIS OF GRB IN LIV THEORY In this section, we discuss the LIV effect on the observed GRBs, focusing especially on the time of flight of γ rays. Gamma-ray bursts (GRBs) are sudden, intense flashes originating from distant galaxies with cosmological distances, and they are the most luminous electromagnetic events we ever known. As already shown in the above formulas, LIV can modify conventional Maxwell equations and hence leads to modified dispersion relation in addition to the conventional one. However, due to the large mass scale suppression, these LIV effects must be very tiny to account forthe conventionalstringentterrestrialtest. Fortunately,asfirstpointedoutin[3],the cosmologicaloriginplushigh energyandthemillisecondtimestructureofGRBmakeGRBanidealobjecttoobservethepossibleminusculeeffectsof LIV.Actually,manyknownstringentconstraintstoLIVparametersweredrawnfromastronomicalobservations,such as AGN [10], ultrahigh-energy cosmic rays (UHECR) [22], CMB [23, 24], etc.. The LIV induced modified dispersion relation can lead to many interesting phenomena. The most apparent consequence is the frequency dependence of photongroupvelocity,thoughthisisnotalwaysthecase. Forexample,ifonlyk 6=0intheSMEframework,photons F 5 propagate independently with their energies. So if photons with different energies are emitted simultaneously, this frequency dispersion of group velocity then leads to the so called time-lag phenomenon. In addition to time lag, certain models, e.g. SME, indicate that photons with independent polarizations obey distinct dispersion relations. This was demonstrated in the above two models, see (11) and (19). All these models involve a conventional mode withanextraordinaryhelicitydependentone,thuscanleadtothesocalledvacuumbirefringenceeffects[11,26]. The tiny changes in polarization grow linearly with propagation distance and hence can be accumulated to be observable for cosmological sources. This can provide a sensitive probe to LIV [11, 23–26]. Aside from purely kinematic effects, the tiny LIV correction to dispersion relation can also dramatically change the thresholds of high-energy particle reactions, hence leads to distinct particle spectrum of UHECR with respect to that of Lorentz invariance cases. The observationofthis spectrumcanprovideaunique signatureofLIV[27]. Onthecontrary,thenonobservationofthese effects can put very stringent constraints to LIV parameters [22]. Below we will primarily discuss the GRB time-lag caused by LIV. First we make a brief review of the formula used in the description of GRB photon time-lag with respect to the source redshift. For an isotropic and homogeneous universe, one can derive a differential relation dz dt=− (26) H (1+z) Ω +Ω (1+z)2+Ω (1+z)3+Ω (1+z)4 0 Λ K M R p from the Friedman equation a˙ K 8πG ρ ( )2+ = N , (27) a a2 3 see[28]for details,wherethe present-dayHubble constantH ≃71km/s/MpcandΩ =0fora nearlyflatuniverse. 0 K The matter density Ω ≃0.27,radiationdensity Ω ≃0 andvacuum energy density Ω ≃0.73 arethe cosmological M R Λ parametersevaluatedtoday. Wenotethat(26)isthe standardresultderivedfromgeneralrelativitywhichisalocally Lorentz invariant (LI) theory. So in dealing with photon time-lag below, we implicitly assume that the gravity side is untouched. Though a unified treatment should also include the change of gravitydue to possible LIV effects hence may also change (26) and the time-lag formula used below. Then we make a general assumption of photon dispersion relation E2 =f(p;M,{ξ }), (28) i where f(p;M, {ξ }) is a general function of p (p = |p~|), some unknown large scale M relevant to LIV and a set of i parameters {ξ }. Inspired by the dispersion relations (19) and (25) derived from particular models discussed above i andthe fact thatLIV correctionsmustbe verytiny atlowenergies,weassume thatthe expansionof(28) aroundthe conventional dispersion relation E2 =p2 is N p E2 =p2(1+ ξ ( )i), (29) i M Xi=1 where N is a large number marking the precision of our expansion. Thus (19) and (25) can be regarded as just two special cases of (29), i.e., only ξ 6= 0 and ξ 6= 0 respectively (where (19) just adds helicity dependence assumption 1 2 of LIV). In addition to its generality, the reason for beginning with (29) instead of those particular models is that various experiments have already ruled out (19) to a convincing level (see [29] and a recent review [30]). Indeed, dimension 5 LIV operatorshave suffered very stringentconstraints both from frequency-dependent birefringence test with GRB [31, 32], Crab Nebula [33], and the UHECR spectrum analysis under special assumptions [22] (i.e. LIV corrections to electron dispersion relations are smaller than those of photon ones, e.g., in the Liouville string models of foamy structure of space-time [34], where only neutral gauge bosons receive quantum-gravity corrections). By taking into account of the expansion of universe [35] and the assumption that gravity side remains intact,the time-lag led by modified dispersion relation (29) with leading order LIV correction of order n is 1+n δEn z (1+z′)n δt= ξ 0 dz′, (30) 2 nMn Z h(z′) 0 where h(z)=H Ω +Ω (1+z)2+Ω (1+z)3+Ω (1+z)4, (31) 0 Λ K M R p 6 E is the redshifted photon energy observed on earth. δEn = En−En, where E and E denote lower and higher 0 0 l h l h energiesofobservedphotonsrespectivelyinthe time delay. The n-thordercorrectioncorrespondstodimensionn+4 LIV operators. This can be seen from another formula z (1+z′)d−4 δt=δwd−4 dz′ Y (nˆ)k(d) , (32) Z h(z′) 0 jm (I)jm 0 Xjm which is suitable to the analysis of time-lag in the SME framework given recently by Kostelecky´ and Mewes [36]. Before we discuss the linear (n = 1) and the quadratic (n = 2) corrections to the photon dispersion relation, we first utilize the observed time delay in GRB090510 located at redshift z = 0.903±0.003 to give a rough estimate to photon mass. As is wellknown, photon mass is representedby dimension 2 operator (quadratic in photon fields) and may spoil gauge invariance (this is not the case in Chern-Simons theory [19] in space-time dimension 3). However, the presence of photon mass does not necessarily implies LIV. In a LI theory (Proca’s theory), the existence of a unique speed c could be regarded as the limiting speed of light for arbitrary-high energy photons. So photon mass is absent in a LIV theory (e.g. SME [31]) if gauge invariance is still valid. But we can still give a rough estimate of its magnitude by using (32) and the fact that the bulk of the photons above 30 MeV arrived 258±34 ms later than those below 1 MeV [13] in the observation of GRB090510 : Y (nˆ)k(2) ≤1.4801∗10−24 GeV2. (33) 0 jm (I)jm Xjm This could be translated to the photon mass bound as m ≤ 1.217∗10−3 eV, much larger than the mass upper γ bound given in [37], m ≤ 1 ∗ 10−18 eV. This confirms the remarks given in [38]: “(departures of electrostatic γ and magnetostatic fields from the gauge invariant one) give more sensitive ways to detect a photon mass than the observation of velocity dispersion.” Of course, one can obtain an effective mass bound comparable to this as m ≤1.217∗10−19 eV [24], but the origin is different. From the bound derived we see that one would need a much γ largerphotonmasstoexplainthetimeofflightdataifonedoesnotintroducetheLIVeffect(orothereffect,e.g. source effect). Morethanthat, as the presenceofphotonmassindicates that highenergyphotons propagatefaster thanlow energy ones, this time advance of high energy photons may cancel possible time-lag induced by certain LIV models (ξ < 0 in (29), see also (35)). Thus mass effects may conspire with LIV effects to produce a nearly nonobservation i of time-lag in certain time of flight analysis [39]. On the contrary, in some scenario with ξ > 0, the time-lag might i be caused by the combined effects of mass and LIV, thus the situation is still complicated. Fortunately, due to the high precision laboratory experiment [40] (constrain m to 10−17 eV level), those scenarios mentioned above do not γ happen and we can safely ignore the mass effects in our discussion about LIV constraints drawn from the time of flight analysis of GRB090510. Without the trouble of possible mass effects, we can then securely discuss LIV effects in the time-lag phenomena below. As a byproduct of (32), we give a rough estimate to mass dimension 3 LIV operators Y (nˆ)k(3) ≤1.1558∗10−21 GeV. (34) 0 jm (I)jm Xjm We see that this bound is comparable to that obtained from the LIV effects on Schumann resonances in a natural earth-ionosphere cavity [41], though it is much weaker than the other astronomical constraints [23, 24] (which are constrained to less than 10−43 GeV). ThenweturntononrenormalizableLIVoperatorsbutusingformula(30)oftime-laginstead,asitismoresuitableto our simple analysis basedon generaldispersionrelation(29). As usual, we only discuss energy dependent corrections to photon group velocity to the quadratic level. Before looking into details, from (29) we derive modified group velocity ∂E p 1 p v ≡ = 1+ (i+2) ξ ( )i g i ∂p E (cid:18) 2 M (cid:19) 1+ 1(i+2) ξ ( p )i = 2 i M (cid:0) 1+ξ ( p )i (cid:1) i M 1p p ≃1+ (i+1) ξ ( )i, (35) i 2 M where Einsteinsumoverindex i from1 to N is indicated. Then by the same procedurein [35], we cangivea time-lag formula which is accurate to 2nd order of E and first order in ∆z as M E −E z (1+z′) 3 E2−E2 z (1+z′)2 δt=ξ l h dz′+ (4ξ −ξ2) l h dz′. (36) 1 M Z h(z′) 8 2 1 M2 Z h(z′) 0 0 7 It can be seen that (36) is consistent with (30) when ξ = 0 and ξ = 0 respectively for linear (though the linear 2 1 correctioniscalculatedto secondorderinthelargemasssuppression,itwillbe checkedthatthis cannotimprovethe linear constraintsanymore,thus in practicalcalculation,(30) is enough)and quadraticcorrectionsto groupvelocity. Forlinearenergydependentcorrectionto1storder,wederivefromthemostconservativeclaimthat,thebulkofthe photonsabove30MeVarrived258±34mslaterthanthosebelow1MeV[13],theLIVscale M ∼−5.02689∗1016GeV, ξ1 which is 3-order less than the Planck scale if |ξ | is of order 1. However, if utilizing the more stringent claim that, a 1 singlehighestdetectedphotonfromGRB090510with31GeVarrives0.179slaterthanthemainLATemissionabove 100 MeV, we can deduce a significant higher quantum gravity mass scale M ∼−7.72017∗1019 GeV, (37) ξ 1 wherethe minussignindicatesthe factthatphotonswithhigherenergiespropagateslowerthanloweronesasalready mentioned in the discussion of photon mass (ξ < 0). By direct calculation of solving 2nd order equation of M (i.e. 1 ξ1 setting ξ =0 in (36)), we find that this can not improvethe resultany more as mentioned. This illustrates that in a 2 roughestimateoflinearcorrectiontogroupvelocity,thereisnoneedtotakeintoaccount2ndordercorrectionof(M)2 ξ1 as(36). Theresult(37)meansthatlinearcorrectiongivesaLIVmassscalenearly6.32M ifξ ∼ O(1),whichisvery Pl 1 closeto thatof[13]. Ofcourse,if onechoosesotherdata fromthe Table 2in[13],one canobtainthe sameconclusion thatquantum-gravitymassscaleissignificantlyabovethePlanckmass(atmostoforder102M [13])fromthissimple Pl analysis. This conclusion is nothing more than a translation of the claim that the constraint on the linear energy dependent LIV parameter |ξ | can be placed in the range 10−1 ∼ 10−2, if we regard M ∼ M . If this constraint 1 Pl is confirmed by other astrophysical observations, then it puts the constraints at least 2 order of magnitude stronger than[10]and[15],whichgives|ξ |<17and|ξ |<58respectively. However,theseconstraintsarenotstrongerenough 1 1 asthoseobtainedin[22]extractedfromtheUHECRspectrumandthosein[31]fromthefrequency-dependenthelicity observations of GRB930131 and GRB960924. However, we note that those most stringent constraints (ξ ≤ 10−14) 1 uptonowrelyeitheronparticularassumptions(see[22])orhelicitydependentmodels,e.g. (19). Thusitisnecessary toputthe constraintsobtainedfrom(29)onthe linearLIVparameterξ tothe samelevel(stillahardtaskasaspan 1 of12ordersto be conquered)fromfuture observations. If so, wecanfinally makea morefirmclaimthat dimension5 LIV operators can be excluded firmly [30]. Then we may reach the conclusion in the near future that either Lorentz symmetry is exact, or at high energies there are some other symmetries such as SUSY plus CPT to protect our low energy theory from receiving CPT odd corrections [29]. For quadratic energy dependent correction, i.e. ξ = 0, we obtain from the most conservative claim mentioned 1 above the constraints on quantum-gravity mass scale M ∼ 5.84718∗107 GeV if ξ ∼ O(1). While for the single 2 31 GeV event, we obtain the constraintas M2 ∼−5.26767∗1021 GeV2, thus M ∼ 7.25787∗1010 GeV. It is obvious ξ2 that the constraints obtained from quadratic correction are much weaker than those from linear one as quadratic correction being suppressed more than one power of M. Thus in the future we should take more efforts to the search of more stringent constraints on quadratic LIV correction to photon group velocity. As mentioned above, the quadratic correction is produced by dimension 6 operators, which are the leading order nonrenormalizable CPT even LIV operators. As we known, various current constraints to dimension 6 operators are also much weaker than dimension 5 ones [31]. Before we close this section, we observe that our results are similar to that obtained recently in [10] and [14, 15]. We give these results in the table below: Source Mkn501 [15] PKS 2155 - 304 [10] GRB080916C [14] GRB090510 [13] redshift 0.034 0.116 4.35 0.900 δt(s) 240 27 16.54 0.179 Eh to El(GeV) 104 to 250 600 to 210 13.22 to 10−3 31 to 0.1 M(GeV) 6.06∗1017 7.51∗1017 1.55∗1018 7.72∗1019 ttotal 6.01∗1013 1.72∗1015 2.34∗1016 1.29∗1018 δt TABLEI:Wherethefirstthreerowsbelow thesourcerowarethedatagivenfrom [10,13–15]respectively,thelast 2rows are thelinear quantum-gravitymasses and total time to time-lag ratios calculated from (30). We findthatthe roughestimatesaboutthe linearmassscaleareconsistentwiththosegivenbythe referenceabove and the order of magnitude of the linear mass scale ranges from 1017 to 1019. It is easily seen from Table I that one can approach this large magnitude with the advantages both from the quotient of total time to time-lag (which 8 originates from the cosmological distance and short pulse nature of GRB) and the large absolute energy difference (range from GeV to TeV). To further constrain the linear order quantum-gravity mass hence the coefficient ξ in 1 the future, we need to amplify the large ratio of total time to time lag, as photons with energies much higher than alreadyobserved(TeV)cannotreachus fromcosmologicaldistancedue tothe paircreationinteractionwithinfrared backgroundphotons. Onewaytoamplifythelargetimeratioistoexcludeothernon-LIVinducedtime-lagfactors,like different response time of detectors [42] (which slightly increase linear quantum-gravity mass scale obtained in [13]), or one turns attention to other methods like spectrum analysis [22], otherwise the constraints can not be improved significantly. Furthermore,astatisticanalysisbytakingintoaccountofstatisticerror[15]andamulti-sourceanalysis to calculate the correlation between distance and time-lag [8] will make the results more concrete. IV. CONCLUSION Inthispaper,wereviewedseveralmodifieddispersionrelationsfromstandardmodelextensionandHo˘ravatheoryin the photon sector. Dispersion relations are derived consistently from the inverse of photon free propagators,without taking quantum corrections into account. Inspired by these dispersion relations we give a more general one (29) to avoid some particular assumptions (e.g. helicity dependence). Then we apply this relation to the time of flight analysis of recently reported GRB090510. We obtain constraints on the linear LIV energy dependent coefficients to the level of ξ ∼ O(.1), which is equivalent to the statement that the relevant linear quantum-gravity mass scale 1 is M ∼ 7.72∗1019 GeV. Using the same method we also get the quadratic mass scale M ∼ 7.26∗1010 GeV, q denoting a much loose constraints to dimension 6 operators. These results are consistent with those in [10, 13–15]. As a byproduct of the time-lag formula (32), we point out that one can safely ignore the photon mass effect in the discussionofLIVeffects inthe time-lag analysisofGRBs due to the stringentterrestrialconstraintsonphotonmass, and we obtain a constraints Y (nˆ)k(3) ≤1.1558∗10−21 GeV to the dimension 3 operator. jm0 jm (I)jm From our analysis, we findPthat though the constraints obtained are far from reaching those from the spectrum analysis[22]andthose fromthe helicity dependentanalysis[31], ouranalysisrelieslittle onextraassumptionsexcept theexpansion(29)andtheformula(30). Thustoexcludethelinearorderquantum-gravitycorrectiontophotongroup velocityis stilltooearlyaslongasthe constraintsto the generallinearordercorrection(29)havenotapproachedthe samelevelasin[30]. We findthatthe resultshavealreadyreachedthe precisionofprobingPlanckmassscaleoreven higher, slightly better than [10, 15]. If one can largely clarify other time lag uncertainties like [42], the constraints could be improved more. Acknowledgments We thankHongboHuandBinChenforremindingusaboutreferences[13,35]andthe helpfuldiscussionswithBin Chen, Alexandre Sakharov,Lijing Shao, Zhi-bo Xu, and Shouhua Zhu. This work is partially supported by National NaturalScienceFoundationofChina(No.10721063andNo.10975003),bytheKeyGrantProjectofChineseMinistry of Education (No. 305001),and by the Research Fund for the Doctoral Programof Higher Education (China). Appendix The previous referred parameter α, is just one of the parameters defined from various combinations of (k ) . F κλµν These definitions arise for convenience from the consideration of the symmetry of this tensor. From the Lagrangian 1 δL=− (k ) FκλFµν, (38) F κλµν 4 wefindthat(k ) isantisymmetrictothetwoindicesκλandµν respectively,andissymmetrictotheinterchange F κλµν of these two pairsof indices. As we do not wantto include a conceivable θ-type term proportionalto 1ǫ FκλFµν, 2 κλµν we require that ǫ (k )κλµν = 0. By requiring that (k ) is doubletraceless as any trace term would serve κλµν F F κλµν merely as a redefinition of kinematic terms and hence a field redefinition, (k ) has the symmetry of Riemann F κλµν tensor. Then we can define the decomposition of (k ) in terms of its spatial and time indices, i.e., F κλµν 1 (k )jk ≡−2(k )0j0k, (k )il ≡ (k )jkmnǫijkǫlmn, DE F HB F 2 1 (k )jk ≡−(k )kj ≡ (k )0jmnǫkmn, (39) DB HE F 2 9 with Latin indices run from 1 to 3. Moreover,we define 1 1 α = tr(k ), α = tr(k ), (40) E DE B HB 3 3 and double tracelessness gives tr(k +k )=0, i.e., α≡α =−α . So we can extract the trace term to define HB DE E B (β )jk =(k )jk −αδjk, −(β )jk =(k )jk+αδjk. (41) E DE B HB ByusingBianchiidentity(k ) =0,wehavetr(k )=0. Withthesedefinitions,wecanrewritetheLagrangian F κ[λµν] DB (1) as 1 1 1 L = (E~2−B~2)+ α(E~2+B~2)+ (β )jkEjEk+(β )jkBjBk+(k )jkEjBk photon E B DB 2 2 2 (cid:0) (cid:1) +k0 A~·B~ −φ~k ·B~ +~k ·(A~×E~). (42) AF AF AF [1] V.A.Kostelecky´ and S.Samuel, Phys.Rev. Lett. 63, 224 (1989); ibid 66, 1811 (1991); Phys. Rev.D 39, 683 (1989); ibid 40, 1886 (1989). [2] R. Gambini and J. Pullin, Phys. Rev. D 59, 124021 (1999); J. Alfaro, H.A. Morales-Tecotl, and L.F. Urrutia, Phys. Rev.Lett. 84, 2318 (2000). [3] G. Amelino-Camelia, J.R. Ellis, N.E. Mavromatos, D.V. Nanopoulos, and S. Sarkar,Nature 393, 763 (1998). [4] M.L. Yan,Commun.Theor.Phys. 2, 1281 (1983); V.A.Kostelecky´, Phys. Rev.D 69, 105009 (2004). [5] P. Horava, Phys. Rev. D 79, 084008 (2009); JHEP 020, 0903 (2009); Phys. Rev. Lett. 102, 161301 (2009); arXiv:hep- th/0811.2217. [6] G.Amelino-Camelia, Int.J. Mod. Phys. 11, 35 (2002); ibid. 11, 1643 (2002); J. Magueijo and L. Smolin, Phys.Rev.Lett. 88, 190403 (2002); Phys. Rev.D 67, 044017 (2003). [7] H. Mu¨ller et al., Phys. Rev. Lett. 91, 020401 (2003); P. Wolf et al., Phys. Rev. Lett. 90, 060402 (2003); H. Mu¨ller et al., Phys.Rev.Lett. 99, 050401 (2007). [8] J.R. Ellis, K. Farakos, N.E. Mavromatos, V.A. Mitsou, D.V. Nanopoulos, Astrophys.J. 535, 139 (2000); J.R. Ellis, N.E. Mavromatos, D.V. Nanopoulos, and A.S. Sakharov,Astron.Astrophys.402, 409 (2003); J.R. Ellis, N.E. Mavro- matos, D.V. Nanopoulos, A.S. Sakharov and E.K.G. Sarkisyan, Astropart. Phys. 25, 402 (2006); [9] P. Kaaret, Astron.Astrophys.345, L32 (1999). [10] F. Aharonian1 et al., Phys. Rev.Lett. 101, 170402 (2008). [11] D.Colladay and V.A.Kostelecky´, Phys.Rev. D 55, 6760 (1997); ibid 58, 116002 (1998). [12] R.C. Myersand M. Pospelov, Phys. Rev.Lett. 90, 211601 (2003). [13] Fermi GBM/LAT Collaborations, A.A. Abdo et al., Nature (2009) in press (doi:10.1038/nature08574), arXiv:astro- ph/0908.1832. [14] The FermiLAT and Fermi GBM Collaborations, A.A. Abdo et al., Science 323, 1688 (2009). [15] MAGIC Collaboration and J.Ellis et al., Phys.Lett. B 668, 253 (2008). [16] L. Shao, Z. Xiao, B.-Q. Ma, arXiv:hep-ph/0911.2276. [17] V.A. Kostelecky and M. Mewes, Phys.Rev.D 66, 056005 (2002). [18] L.H. Ryder,Quantumfield theory (Cambridge UniversityPress, 1996), see chapter7, P242-P246. [19] G.V. Dunne,arXiv:hep-th/9902115, see section 2.2 and exersises there. [20] R. Casanaa, M.M.F. Jra, A.R. Gomesb, and P.R.D. Pinheiroa, arXiv:hep-th/0909.0544. [21] B. Chen and Q.G. Huang, arXiv:hep-th/0904.4565 (2009). [22] M. GalaverniandG. Sigl,Phys.Rev.Lett.100,021102(2008); L. MaccioneandS. Liberati,arXiv:astro-ph/0805.2548; L. Maccione, A.M. Taylor, D.M. Mattingly and S. Liberati, arXiv:astro-ph/0902.1756 (2009). [23] V.A. Kostelecky and M. Mewes, Phys.Rev.Lett. 99, 011601 (2007). [24] T. Kahniashvili et. al., Phys.Rev.D 78, 123006 (2008). [25] V.A. Kostelecky and M. Mewes, Phys.Rev.Lett. 97, 140401 (2006). [26] S.M. Carroll, G.B. Field and R. Jackiw, Phys.Rev.D 41, 1231 (1990) [27] U. Jacob and T. Piran, Phys. Rev.D 78, 124010 (2008). [28] StevenWeinberg, Cosmology (Oxford University Press, 2008). See section 1.5, Dynamics of expansion. [29] S.G. Nibbelink and M. Pospelov, Phys. Rev. Lett. 94, 081601 (2005); P.A. Bolokhov, S.G. Nibbelink and M. Pospelov, Phys.Rev.D 72, 015013 (2005) [30] S. Liberati and L. Maccione, arXiv:astro-ph/0906.0681 [31] V.A.Kostelecky´ and N. Russell, arXiv: hep-ph/0905.0031, Table XIV. [32] I.G. Mitrofanov, Nature426, 139 (2003). [33] T. Jacobson, S. Liberati, D. Mattingly and F.W. Stecker,Phys. Rev.Lett. 93, 021101 (2004). 10 [34] J.R. Ellis, N.E. Mavromatos, D.V. Nanopoulos, and A.S. Sakharov,Int.J. Mod. Phys.A 19, 4413 (2004) [35] U. Jacob and T. Piran, JCAP 0801, 031 (2008). [36] V.A.Kostelecky and M. Mewes, Astrophys.J.689, L1 (2008). [37] Particle Data Group, C.Amsler et al., Phys.Lett. B 1, 667 (2008). [38] A.S. Goldhaber and M.M. Nieto, arXiv:hep-ph/0809.1003. [39] T. Kahniashvili,G. Gogoberidze,andB. Ratra1,Phys.Lett.B643,81(2006);M.R. Martinez,T. Piran,andY. Oren, JCAP 017, 0605 (2006). [40] R. Lakes,Phys.Rev.Lett.80,1826(1998); JunLuo,Liang-ChengTu,Zhong-KunHuandEn-JieLuan,Phys.Rev.Lett. 90, 081801 (2003). [41] M. Mewes, Phys. Rev.D 78, 096008 (2008). [42] G. Ghirlanda, G. Ghisellini and L.Nava, arXiv:astro-ph/0909.0016.