Effective field theory for photons in media Finn Ravndal, Institute of Physics, University of Oslo, Norway. • Physical optics • Maxwell equations and symmetries • Minkowski (1908) and Abraham (1909) theories • Covariant theory and photon mass • Effective field theory ArXiv: 0804.4013, 0805.2606 • Conclusion Refraction: sin θ0 = n sin θ • In vacuum: •ν λ =c ≡1 0 0 0 • • In medium: !0 "0 n=1 n>1 " νλ = c = ν0 · λ0 /n = 1/n ! Light wave going through a material medium: ! A(x, t) = 1 " Ak (t)eik·x V k !0 (1) ! P Figure 1: Photon with wavevector k being emitted in the xz-plane. Initial atomic spin is along the z-axis. for eneach 3-dimensjonal harmonisk oscillator er in the forward ScatteredHamilton-operatoren waves from atom interfere 1 1 # +with p mω #r shorter wavelength. = direction to give a newH#wave 2m 2 2 2 2 Maxwell equations: ∂B ∇×E+ = 0, ∂t ∂D ∇×H− = 0, ∂t ∇·B=0 ∇·D=0 Constitutive relations: Gives wave equation with n= √ εµ B = µH D = εE ! 2 ∂ 2 n − ∇ E(r, t) = 0 2 ∂t 2 " efields atoms the while H describes the similar modification E in and B material, be the components of the standard antisymmetric tensor + ∂ν Fλµ + ∂µ Fνλ = 0 λ Fµν field B due to magnetization of∂the atoms. " ! " ! 0 −E −E 0 varying eral local fields on the microscopic scale between µν µν rapidly here the covariant gradient operator has the components ∂µ = (∂ F = (3) F = E −B is scale we expect ordinary ijLorentz invariance to be valid. Indeed, E −B ij mbining D and H into the corresponding antisymmetric tensor erified by writing these four equations on covariant form. For that Bij = εijk Bk . Then we can write the first two Maxwell equations (1) " the first two Ma ! we antisymmetric when write Bij = in εijkof Bthe can write fields EMaxwell and B we be the components standard k . Then equations rest frame: Covariant formulation as ∂λ Fµν + ∂ν ! Fλµ + ∂µ Fνλ H "0 = µν = 0 −D D −Hij (4) 0 −E F = (3)= 0 ∂ F + ∂ F + ∂ F λ µν ν λµ µ νλ iant gradient operator has the components ∂ = (∂ , ∇). Similarly, E −Bij µ t µν th two last equations (2) become simplified to ij = ijk Hk , the antisymmetric d HH into theεcorresponding tensor e= taken to. be constants. These constitutive relations now represent an effect where the covariant gradient operator has the components ∂µ = " ! ε B Then we can write the first two Maxwell equations (1) j ijk k −D medium 0 and iption ofcombining the fields in the where microscopic variations are averag µν µν D H into the corresponding antisymmetric tens (5) H = ∂ H = 0 ν D −Hij n some sense. " ! ∂ F +∂ F +∂ F =0 (4) to theseshown field e relations can also be formulated in an arbitrary frame as already √ V = Boost velocity: D Similarly, −Hij 2 (∂ , ∇). µν ant gradient operator has the components ∂ = 1 − v µ t peared on the right hand side of this equation. owski[2]. Let the∂νmedium with correspon H = 0 move with velocity v and therefore (6) H into the corresponding our-velocity V µ = γ(1, v).antisymmetric Here γ = (1tensor − v 2 )−1/2 as follows from V µ V = λ µν ν λµ νλ −D 0coupled µ(1, µν Hk , the last equations (2) µbecome simplified to v) ad we two included a four-current J = (ρ, J) µ H = µ = εijk Hk , the two last equations (2) become with H ij d a four-current J = (ρ, J) coupled to these fields, it would have " ! µ Covariant µν form of constitutive eqations: the rest frame of the medium are related by ing F hand = FsideVofν ,this oneequation. can−D then write the fields 0 right µν µ simplified D = εE (5) the magneti H = Both the electric permittivity µν tropic matter. ε and ∂ H = 0 D −Hij ν 2 µHfields Fµν + (nby −D1)(F Fν=VµH µν =are µ Vand ν −B µ ) for e of the medium the related = εE . Both the electric permittivity ε and the magnetic µpermeability µ the two 2 last equations (2) become simplified to k, µHµν = Fµν + (n − 1)(Fµ Vν − Fν Vµ ) n2 = εµ. Several years later Gordon[13] rewrote this on the more su 2 ere now n = εµ. Several years later Gordon[13] rewrote this on the µ µν m F = F Vν with tive form ρ σ µH = X X ( µν µ ν Fρσ µ ρ σ we write A = (Φ, A). µH = X X Can then write µν µ ν Fρσ ducing the tensor an derive thethe wave equation from a variational principle. The Lagrangian er also introducing tensor from with the requirement theory Xµν that = ηµνthe+vacuum (n2 − 1)V µ Vν should result in the limit n → (1 2 Hµν → Fµν . It then becomes the simplest Xµνclear = ηthat + (n − 1)Vµ Vchoice µν ν 2 2 erm ∝ (n − 1) in (7) will not contribute because of the antisymmetry 1 2 2 µν contribute because of the antisy eηµν lastisLagrangian: term ∝ (n − 1) in (7) will not the Minkowski metric that Xµν plays the role of a generaliz L = −so Fµν H (13 4metric so that X plays the role of a . Here η is the Minkowski σ µν µν etric[14]. desired properties. In and magnetic vector fields i 1 termsµνof ρthe electric σ es L =or (E DµL −B H)/2 when inserting the tensors (3) and (5). Using X F X =· − o find an · equation ofFmotion of the one must combine the th tw ρσ ν field, µ 4 frame of the medium, it is utive relations in the rest order to equations find an equation of motion of the field, one combin Maxwell (4) and (6). The first is satisfied as must an identity variant Maxwellpotential equations and (6). an t g a four-vector Aµ1(4) so2that Fµν ∂µ Afirst then ν − ∂is ν Asatisfied µ . Using as 2 2 =The µL = (n E are − ) symmetry (14 µ Bnot But Lorentz boosts e relation a(7) in the second equation the corresponding roducing four-vector potential A (6), so that Fµν = ∂µ Aν −wave ∂ν Aµequati . Usin 2 . It takes the form when chooses gauge equivalent to t nstitutive relation (7) instead intransformations! the second equation (6),arelations the corresponding wav eneral frame onesimplest must use theone constitutive on the covarian uge in vacuum, i.e.the ows[15]. It takes simplest form of when one chooses a gauge equiva 7). The Lagrangian is then a function the 4-vector potential Aµ , but wil epend explicitly on the µfour-velocity Vµµ of the νmedium. It is therefore no renz gauge i.e. 2 µν in vacuum, Symmetries Lorentz transformations • Invariance underspeed of light c=1/n with reduced • Gauge invariance: B = ∇ × A and E = −∂A/∂t − ∇Φ invariant under A → A + ∇χ Φ → Φ − ∂χ/∂t Massless photons ? Energy density and conservation: where N = E × H Momentum density and conservation where " 1! E = E·D+B·H 2 ∂E +∇·N=0 ∂t is the Poynting vector G=D×B ∂G +∇·T=0 ∂t 1 Tij = −(Ei Dj + Bi Hj ) + δij (E · D + B · H) 2 is the Maxwell stress tensor 1 µν µνµ αβ µ αν η F H T = F H + x = (t,αx) can combineαβ With M to 4-dimensional 4 µν conservation ∂ν T can be displayedlawin the matrix M =0 µν TM T 00 = ! E N G Tij where " isis the energy density the Minkowski energy-momentum tensor 1 E = (E · D + B · H) Tensor only symmetric in vacuum where G = N 2 the components of the Poynting vector N = E × on-symmetric with the components T k0 forming th nally, the space-space components make up the ten split up Minkowski the Minkowski tensor (16) twoterms, terms, up the tensor (16) inintwo up the Minkowski tensor (16) in two ! terms, " e first part Abraham (1909): ! 0 µν µν 2 " 0 ! 0 0" T = T + (n − 1) µν M µν A 2 TMµν = TAµν + (n2 − 1) 0 N 0 0 TM = TA + (n − 1) N 0 N 0 tt part tensor " ! partwhere symmetric energy-momentum E N" ! ! = E N " NNTij E = = TAµν µν µν T A TA(22) N T Tij N ij braham tensor[3]. It is symmetric by construction. From the µν lawtensor[3]. for the the Minkowski tensor onconstruction. the second index, it follo m It is symmetric by From the = 0 ∂ T Conservation gives ν n. From the above am tensor[3]. It isconsymmetric by construction. From the M µν tensor is not generally tensor conserved. Instead, it index, satisfiesit∂νfollo TA for the the Minkowski on the second it follows that the tensor on the second index, for the the Minkowski it folloµ µndex, µ is is a force density.µ Inconserved. the rest frame we see that K = (0, K) µν µν sor not generally Instead, it satisfies ∂ T atisfies ∂ T + K = 0 where Abraham force ν ν sor is notAgenerally conserved. Instead, it satisfies ∂ν TAA µ µµ = (0, K) K = (0, K) where force density. In the rest frame we see that K a force density. In the rest ∂ we see that K = (0, K) 2 frame K = (n − 1) (E × H) ∂t ∂∂ (23)2 Quantization Photon energy: E = h̄ω Photon momentum: p = h̄k = nE from Minkowski momentum density G = D × B and p = h̄k/n2 from Abraham momentum density N = E × H = G/n 2 Photon mass: m2γ Minkowski 1 − 1/n > 0 Abraham 2 <0 1 − n =E −p =E · 2 2 2 2 { Effective Theory Light cone in medium rest frame: x = ±t/n Metric: 2 2 2 2 ds = dt /n − dx = ηµν dx dx µ = (t/n, x) or ∂µ = (n∂t , ∇) x with µ ν so that covariant 4-momentum p = (nE, p) µ Massless excitation p pµ = 0 µ as for Minkowski photon. or E = p/n not invariant under the corresponding transformations. munder itself the is not invariant under the corresponding transformations corresponding transformations. mmagnetism is in thisisrespect different since the vacuum is invariant in this respect different since the vacuum is invarian µ pect different the vacuum is invariant eansformations. write A =since (Φ, A). Lorentz transformations. Covariant also the wave equation from athe variational princ electromagnetism ke this derive material symmetry manifest, introduce the correspondin aterial symmetry manifest, introduce corresponding try manifest, introduce the corresponding µ µ nate four-vector x so vacuum that = th rom the requirement that the theory should res t , ∇). -vector x = (t/n,=x)(t/n, so x) that ∂µ =∂µ(n∂ , ∇). ForForthe t(n∂ (t/n,the x) so that ∂ (n∂ the µ = potential t ,µ∇).Aµ For form new four-vector = (nΦ, A). The electric an 4-vector potential → F . It then becomes clear that the simplest cho new four-vector potential A = (nΦ, A). The electric and µν µν µ r potential = (nΦ, A). The electrictensor and Fµν = ∂µ Aν −∂ν A ectors are stillAgiven by the antisymmetric still given by the antisymmetric tensor Fµν = ∂µ Aν −∂ν Aµ he tensor Fµν = ∂µ A1ν −∂ν Aµµν he antisymmetric components and field tensor L = − F H µν ! " 4 0 "−nE ! µν or" F 0 = −nE (24 desired terms of the electric and (24) mag nE In−B ij −nE F µν = properties. (24) the tensors (3 nE −B ij s L = (E · D − B · H)/2 when inserting E −Bij 1 µν takes the standar n termsLagrangian: of this, the rest-frame Lagrangian (14) F F µL = − µν tive relations in the rest frame of the medium, it istrans 2 4 /4)F . It is now explicitly invariant under material Lorentz µν the rest-frame Lagrangian (14) takes the standard f-frame this, Lagrangian (14) takes the standard first setexplicitly of field equations (4) under obviously remainsLorentz unchanged whil 1 is now invariant material trans2 trans2 2 tly invariant under material Lorentz µν − Bwe) make use of th well equation (6) is replacedµL by = ∂µ F (n=E 0 when of field equationsremains (4) obviously remains unchanged while 2 ons (4) obviously unchanged 2 while ν ν ations. Writing this out, it becomes ∂ A = ∂ (∂ · A) where now nents µν now also find a satisfactory energy-momentum tensor fro n (21)Standard appropriate to this Itαβgives 1group. µν µν new µLorentz αν energy-momentum tensor µT = F α F + η Fαβ F 4 1 µν αβ µν µ αν η F F µT = F F + αβ α ith components 4 onents with components T ! µν = ! E nN nN" Tij " E nN T = is now symmetric and conserved both indices is obviously symmetric, traceless conserved on both in nN and Tij on T νµ = 0. ∂InT the direction this gives energy conservati µν time µν = ∂ T = 0 ν µ hile the space directions gives momentum conservation uslyinsymmetric, traceless it and conserved on both indices, ∂E entum density of the D × energy B is therefore the same + ∇field · N this =G0 =gives . In the time direction conservation on giving ∂t heeory. space directions it gives momentum conservation as in ∂G Minkowskithe momentum ensity and of the field + G∇=· TD=×0 B i.e. is therefore same as in he rest frames of∂tthese systems are preferred frames. It is also density! For the electromagn ost natural to quantize the excitations. µν Physical consequences: ! 1! h̄|k| h̄ωk = n 1) Casimir effect: E0 = k k 1 vac i.e. E0 = E0 n I. Brevik and K. Milton, arXiv 0802.2542 ! 3 h̄ωk d k 2) Thermal radiation: U = (2π)3 eβh̄ωk − 1 i.e. U = n3 Uvac ∝ n3 T 4 Landau & Lifschitz, 1960 ItItisisequivalent equivalent to to EE· ·∇ ∇ EE by by aa partial partial integration integration in in the the action action integral whe appears. appears. The Thesimilar similar term term ∂∂ttEE··∂∂ttEE involving involving two two time time derivatives derivatives is for the s reason reasonequivalent equivalent to to EE··∇ ∇22EE when when we we use use the the equation equation of of motion. motion. An interac likeEE· ·∇∇22BBisisruled ruledout out by by parity parity invariance. invariance. like Interactions dielectricmaterials materialsmagnetic magnetic effects effects are are negligible, negligible, and and it it is is therefore therefore reason InIndielectric assumethat thatall allthe theterms termsinvolving involving the the magnetic magnetic field, field, can can be be dropped. dropped. Inclu totoassume µ = 1 : Dielectric has interactionterms termsup upto todimension-8 dimension-8operators, operators, we we then then have have the the effective effective Lagran interaction "" εε!! 212 dd22 22 22 aa dd11 2 2 22 22 2 µν 2 L = E − B /n + E · ∇ E + (∇ E) + (E L = E − B /n + E · ∇ E + (∇ E) + (E · E) Lef f = M22 M44 M44 22 − 4 Fµν F M M M Theparameter parameterM M isisan anenergy energyscale scale which which characterizes characterizes the the microscopic microscopic physic The dispersion havedescribes integratedout. out. In aa dielectric dielectric this this isis set set by by atomic atomic physics physics and and should ha have integrated In valuearound around 55 -- 10 10 eV. eV. ItIt isis possible possible to calculate calculate or at at least least estimate estimate the dim B2 B1to value or B1,2physics ∝ d1,2 +from n(λ) = sionless coupling coupling constants constants , dd2+ and a from the atomic of the mate 2 4 sionless dd11,A a the atomic physics 2 and λ λ Withaarealistic realistic value value for for the the energy energy scale scale M M,, the the Lagrangian Lagrangian contains contains only t With independent parameters. For each each material material they they can can therefore therefore be be determined and theparameters. Kerr effect independent For threedifferent differentmeasurements. measurements. ItItwould would then then be be able able to to predict predict the the outcome outcome of m three 2 ∝ a dominated n(E) =n+ otherexperiments. experiments. In In practice, practice, weKE expect the the term term ∝ ∝K to be be dominated by other we expect dd22 to -term. dd1 1-term. M = (5term − 10)eV LetCharacteristic usfirst firstconsider considerthe themass: effectof ofthe thedispersive dispersive term ∝ dd11 in in the the effective effective Lagrang Lagrang Let us effect ∝ It gives a contribution to the classical equation of motion for the field that is lin Conclusion • Electromagnetic field in medium is like any other excitation in condensed matter physics, theory defined in rest frame only • Theory similar to Minkowski formulations in most physical predictions • Effective theory describes dispersion in a natural way • Quantum corrections can be calculated