‘MSci 4261 ELECTROMAGNETISM LECTURE NOTES VI In these notes, we will tum to scattering, discussing how light interacts with matter. As an application, ‘we will look att the sky. 6.1 Scattering Ina typical scattering problem, the sourves (J, p) will be localised in some region of space, At some carly time we are given some (ive) fields Aig, dj Which propagate forward in time, and then interact with (ies, scatter from) the sources. Considering A. (a similar discussion applies to ) we have (= V? = deg) OAin =0, since the incident fields are free, and OA = pod, in which the feds A are a superposition A= Ant Ac. of the incident fidds and the scattered fields. The incident fields are in effect the complanentary function for the inhomogeneous equation which is the field equation (ie., a solution to the caresponding homogeneous ecuation), whilst the scattered fields are a particular integral, which physical considerations dictate should be the one with only oufgoing waves, and so are determined by the retarded potentials = 9 | etaee-r/03, giving 1 noi =rjo)h A Aint 2 [ x 30%,¢ ride. ‘Ther 3s a similar equation for the scalar potential ®. 6.2 Scattering from a Small Scatterer ‘Let us consider the case when the incident, field is a plane wave, with angular frequency w the scattered wave will then have the same frequency (at least for the simple cases we will consider), and is generated by the charge and current distributions induced in the scatterer by the incident radiation — which will oscillate With the same frequency w in this analysis. If the scatterer is small, we may use a multipole expansion, and, then in the far zone, keeping only the dipole contributions, ste Exc BF tmxp) xn-nxmid Bec = An x Esc whee p an m are the induced electric and magnetic dipole moments of the scatterer. If the incident plano-wave is in the direction of the unit. vectar mp, we have By = Ec™™,‘The wave vector is ko = kno, and the time-dependence is in the amplitude Eo, which also specifies the polarisation of the incident radiation, The incident flux, the power per unit area normal to the incident ‘wave, after taking the time average, is given by Similarly the flux of energy in the scattered wave is -— Lp. (Sec) = zl. ‘This flux is radially outward, ic, along the direction n, and falls off as -b. The polarisation of the scattered radiation is determined by the direction of Esc. Jn real scattering experiments, one is interested in the differential seattering cross-section. This is defined to be the power radiated in a direction n, per unit of incident flux (with direction no). We can also take account of the polarisation state of the incident and scattered waves, by introducing unit vectors (which may be complex if for example the polarisation is circular) e and € respectively. Then. do, le Excl? PF (mesnoseo) = ie al = ak (ax) xn—nxm/ah zy Ble pe taxany/ (We have used n-€ = 0 as the dipole cloctsc field is perpendicular to n & the far zone.) Note that unless the dipole moments vanish the radiation scattered has a fourth power frequency dependence, as contained in the factor M4 (= 2rv/c). This is Rayleigh’s law. 63 Many Scatterers Jh the previous section, we had supposed that the scatterer was at the origin, Had it been at, the point xj, the incident wave would have been modified by an additional factor exp[ikty +%j] in Eo, andl the phase at the (dissant) point at which the scattered wave is observed! would likewise be modified by a phase exp[-iln+x;)]. (ncluling the time dependence, the waves eit are incoming to the origin for the plus sign (there i8 an ‘— vf” dependence), and outgoing frem the origin for the mim sign.) ‘The resultant phase change overall between the scattered and the incident wave, expliq + xj], where = k(n), makes no diffaence to the resultant differential cross-section. Howevar, if thaw are a mumber of scattering centres, located at different points, the phase differences between them will kad to interference effects. For a collection of dipole scatterers, we will now find that. we first need to sum over the fiekls for the mov hs ph en eens nine & Pale les ~doxm)e ‘This dearly depends sersitively on the positions x; of the scatterers, If all the scatterers are identical, the effect is to multiply the cross-section for a single scatterer by Fa =|DeP, 7 which is called the structure factor. In the forward direction, where q =0, we have F umber of scatterers). In general, FQ=Ner), o IN? (where NV is the 2When the distribution of scatterers is random, the phases of the off-diagonal terms in this sum canoel except very close to the forward direction, and the result for this case, called incoherent scattering, is that F(q) = N. (This assures that Jal >> a7, a being a typical distance between the scatterers. This implies that |dl|x; —x;| >> 1 so that e#f°—©) oscillates rapidly as one considers different, pairs of sites.) On the other hand, if the scatterens are arranged in a regular array; as for example in a crystal, the ‘coherence of the contributions to the scattering leads to a characteristic pattem. For a simple cubic structure for example, with lattice spacing a, and 1Vi,N2, Ng lattice sites along the three axes of the exystal, and with = (ais 2545) when referred to these axes, ane finds the structure factor vo SE SSH (One may more quidsly understand this result: by considering a one-dimensional array of scatterers, at. sites xj = (@— Ya,0,0), for i = 1,2... ‘The sum over paits of sites in J then gives rise to the factor (Ls et + cat $+ ODN emi + HM +. + -IAN=D9) which leads to cnx of the ratios of sines in the above.) ‘Each factor in the above expression for F peaks sharply around ga = 0,27,475.24 te, whenever the [Bragg condition is satisfied, and then, just as in the forward direction, F = N2, The number of peaks is limited by the maximum value which ga can attain, ga < 2ka, s0 that at long wavelengths, only the forward peak occurs, This has a width determined by q: <2n/Nsa, comespanding to scattering, angles less than or of ander 1/L, where L & the linear sim of the crystal. The siurly and analysis of scattering from matter has had profound effects on our understanding of physics, chanistry and biology. Far example, X-ray scattering has been extensively used to determine crystal and molecular structure. Probably the most far-reaching result was the understanding of the molecular basis to genetics X-ray scattering experiments, notably those of Rosalind Franklin and Maurice Wilkins at: Kings College London, were crucial to the disoovery of the double helix structure of DN. 644 Scattering from the Sky ‘The light scattered from a cloudless sky is blue, The explanation of this may be given at various levels of sophistication. It is already apparent from the Rayleigh formula that the k* dependence of dipole scattering ‘vill lead to a dominance of the shorter wavelengths in the scattered radiation frem an originally white source. A mote detailed descziption of the origin of this dipole scattering will now be given. We will suppose that thee are no free charges or cumrents in the sky. This means previous approach fails to give any scattering at all. ‘The reason is that we have bee considering sours in ‘vacuo, but air is not a vacuum, so that: € = ever # @ and 1 = Hof4r # Ho. The relative permittivity er and dative permeability ji are close to unity for air. Their deviation from unity is not in itself enough to give scattering: what is important is that there are local fluctuations in the values of these quantities. From Maxwell's equations it follows that . Vx (D-aE)=VxD+aB, ‘so that, a Vx (Vx (D-@B)) =V x (xD) +@a(V xB). ‘We may recognise that the expression D — @E = P is the polarisation, so that Vx(VXP)=VX (VXD) Hod(VXB) 3Likewise, using B — oH = 0M, where M is the magnetisation, we have aD ot a a pV x M = FV x B=! ‘Subtracting the last two equations gives a & apo gV XM-Vx (VX P) = ayn 5gD - Vx (V xD) 1e pep For each frecquency w = he we have time dependence et, leading to (7 +)DeX) =8), =VD- whae i VV xM-Vx (VxP). ‘This is small, as we are considering small fluctuations away from the vacuum, We have solved this type of ‘cxqation earlier using the method of Green’s functions. In this case we then derive peel Dex) = Doe) — I ens ‘where Da is any solution to the homogencous equation (7+ P)Do (Note that we have as usual imposed outgoing boundary conditians.) So if Dp represents the incident (plano ‘wave) radiation on same region where ¢ # co, jt # Jo, the integral represents the resultant scattered radiation from that region, Tn the far zone, where [x =2¢| — r we have oO pom+ 24, or, writing x = rn, qf eters). In fact the only dependence an x is through its direction n, We then obtain for the differential cross-section for scattering of radiation with polarisation € from this region do _\e-AB a De ‘To proceed further we need to be able to say something more about s(x). As remark ‘it is small, and is expressad in terms of the polarisation P = D—@E and the magnetization M = 4a above. On integrating by parts [ec tnx (ax P) +nxM'/q. We now need to be able to say something about the (indhioad) polarisation and magnetésaion P and M. 4655 The Born Approximation ‘The induced polarisation and magnetisation will depend in detail on the medium—in this case air—and on the fields imposed. We don’t know the fields, so need to know how to make some suitable approximation. We do lnaw that P=D-@8 = (6 — DE= aserB, wher de, =e, — 1; and similarly M=(,-H= 4B, where djtr = 4p =. Because the scattered wave has an amplitude which is small compared with the incident, wave, wemay to fst order in the small quantities der and dir replace the fields F and B by the contxibutions to them from the incident radiation, ignoring the small contribution from the scattered radiation itself. So to first order we may write P® derDo MR dpipcng x Dy” where we have ako used Bo = no x Ep/c since the incident wave is taken to be a plane wave in the direction ‘Ng. This replacement of the full field by the contribution of the incident fidd alone, dropping the contribution: of the scattered companent, is the Born approximation. Using it, we have for the Bam approximation to the amplitude R Atom =F [Et eH x a x DEG, + 1% (ty DY But if the incident wave is plano-polarised with polarisation vector éy and is propagating in the direction 19, we have Dh =Dotx) = Decne, and so obtain: Ape = eps [ dete [nx (nx eit tmx (mo Xeni, from which it follows that: (2) n= (EY [ tecrre -taxeaxedien tux (oo xaonf =) | | erete ener + axe) xevindf A usefill check with previous results is to apply this Born approximation to the scattering from a small dielectric sphere of radius a This gives (um = (2) Ue terre 9h = (Efe rereeat| [tar [ aeao) [* asemonsl Ae se (6ey)2( Saae—aaersaa ? 7 Hoe @P (SE) wqas0. ‘The result for the long wave-length scattering from a small dielectric scatterer is Ry Ga) Fle -vF- 5But we also have for a sphere with relative pemnittivity ¢, that the induced dipole moment p is given by 5) Eo. ect ep: P= Aro? ( ‘Ths the two calculations agree when de, is small. 6.6 Rayleigh’s Explanation for the Blue Sky ‘The quantity der Which appears in the considerations above s the electric susceptibility Xe = er—Lof a medium, Since at optical frequencies dyir & extremely small, we may relate Xe to the réractive index n by 1? =Gllr ee =Ne tle ‘This & not yet enough to allow ws to use the previous discussion to give an explanation for the scattering of simlight in the atmosphere if the air & considered as a homogeneoss medi. As already indicated, iti best explained through density fluctuations, but we give first another, somewhat simpler explanation due to Rayleigh. The bass of the discussion is to recognise that, the air is not homcgencous, because it is constituted from molecules, and what we should do to consider the scattering from the random distribution of molecu, each of which may be thought of as an electaically polarisable dipole. ‘This mock gives for the dectric susceptibility 1 Dro) e where the molecular polarisability 7,0) i8 defined so that each molecule has an induced dipole moment xd Pi = inolEGs) determined by the fidd EX) it experiences, This field is of couse affected by the presence of other polarised molecuks, and we will return to its determination below. But what we can already do is to obtain. do Ry2 rege ¢ 2 e (B=) f teem oq Dimes) Ry = (Fe) "eer ale? 7 By = (Faq) el thas from previous results, using the Bam approximation, since the moleaules are randomly distributed, and the ‘resultant scattering is incoherent. N is the number of moleaules, and for simplicity they have all been taken to have the same polarisability. Thus the differential scattering cross-section per molecule is : ie ei Qua) 6 ‘The next step is to relate this molecular polarisability to the bulk polarisablity, and hence to the susceptibility and the refractive index, Here we turn to the Clausius-Mosotti relation we discussed in an earlier lecture. This is —s0(@-0 Pn (+2) where here ve introduce the volume V explicitly and define pyy between €, and the refractive index, one obtains Jmol — 32-1 reg Tipp re F2 6 N/V. Using the previously given relationknown as the Lorentz-Lorerz: equation (1880), which can then be used to express the differential scattering of light from the air in the sky as (ree nei HF Gees) whtle-e2(—_Yin- eKle-eP(S 5) OP where at the last step we have wed (n= 1) << 1. So far in this Jong calculation, we have considered the scattering of incident sunlight which has been taken to be polarised, with polarisation vector é. ‘Since sunlight is umpolarisedl, we should really take an ayeragp over the to independent directions normal to no. This then gives for the differential cross-section for scattering leading to a polarisation parallel to the scattering plane defined by the vectors mand ny (= M( =) (n= 1P ao, and for scattering leading to a polarisation perpendicular to the scattering plane Gi). - 3" Gm) OO ‘The scattered light is thus polarised on average, with a polarisation given by the difference between the perpendicular and parallel cross-sections divided by their sum, ic., cos? sin® Tras6 ~ Tra As usual is the angle of scattering, i.e., the angle between the incident direction ny and the scattered direction n. ‘The total differential cross-section is the sum of the two possible final polarisations, and so is, B= PGR) ovata, On integrating this over angles, one obtains for the total cras+section per molecule 11@)= =F e0-0 ‘This is the power scattered out of the incident sunlight. per mokeaile per unit incident flux. Then if 7 is the incident intensity, in traversing a thickness dr of the atmosphere, the intensity is tedueed by dl = =Ipyo dx so that after traversing a thickness 2 the intensity & given by ‘with a corresponding expression for its inverse A, the attenuation length. This expression for the attemution length i due to Raykeigh. Note espedally that the attemation length is proportional to pyr, the number of molecules per unit volume. If there were no atomicity, there would be no attermation, and the sky would not be blue! Note ao the i factor, which leads to very substantial differences in A across the spectrum of visible wavelengths. There is also a wavelength dependence in the refractive index n. 76.7 Critical Opalescence ‘The discussion above is not appropriate to dense media, where the scattering is primarily caused by fluctuations in density. To deal with this situation, consider a division of the volume V from which the scattering occurs into small cells of volume v. The cells, although small, with linear dimension mich Tess than a wavelength, are to be large enough each to contain many molecules. The number of molecules in cach cell will on average be pxyv, where py is, as before, the (average) number of molecules per unit. volume, ‘The fluctuations in density will result in the number of molecules in the jth cell being different. from the average number pvt let this difference be Aj, ie. there are pv +A, molecules in the jth cell, The small Aifference in dersity of the jth call produces a corresponding difference =o Se= pxo in the rdative permittivity of the jth coll. But by differentiating the logarithm of the ClausiusMossotti ‘exquation, eral _ dnp mah G@t2~ 3 treo cone obtains Dep (=e $2) dow 3px Tas we have De tDa, CEVICHE Serj ge Recall that (setting, djir = 0, as 8 appropriate at optical frequencies) for a single scatterer Ry a - EY | / tree code ‘Treating each od] as a single scatterer, and induding the contributiars from all V/v cells in the volume V, which contains a total of pyV molecules, we have Pa(EN|D[ Prcereosech, rhe = (GB) Dee ees? =(=y1 ie. eee oe r= Wer +2) a, Spne =(2) Feo [eMetay Ina. In going fram the first to the second line, we have assumed that the corrdation between fluctuations fram one cel] to the next extends nly over distances small compared with the wavelength, s0 that the exponential can be set to unity. Summing over the initial polarisations 9, averaging over the final polarisations €, and. integrating over dQ, we obtain for the total aus-section for scattering from the pyV molecu in the volume Vs "(poems nar. ‘The aoss-secticn per moleaule is then obtained by dividing by pV, an the attention cocci (0, which 5s pw times the cross-section per mokeaule is G@-DE+ 1x AuP 3 pn vo 8Now the quantity |D, AP? = A% is the square of the sum of the fluctuation in the number of molecules in each cell taken over all the cells in the volume V, ie, it is the square of the fluctuation in the mumber of molecules in the volume V’ from the average musnber xyV of molecules in that volume. That is in turn given by statistical mechanics arguments as a = = py kT! par = Pw kD Br where kp is the Boltzmann constant, and Sp is the isothermal compressbility, Sr % flows hae 8) = De +22 (= lee 2 ENEMA atm = De +2) “Taal oo Flow meron. ‘This is the FinsteinSmoluchowsld equation (Smoluchowski obtained an incomplete version in 1904, which swas completed by Einstein in 1910).. The factor [e, ~ 1)(¢r + 2)/6] i well approximated by n = 1 when fe 1 << 1, and for an idea gas, pk T Br — 1, a that we recover the earlier result a= K's Rayleigh. But we also lear that asthe temperattre apprenches a critical temperature at wl infinite, the attemsation length diverges, and this kad to the phencmenan of critical opalescence, ‘This ist't «nite the end cf the stary, sine the approximaticns we have mde break down near the critical point. For 2% the aitical point approached, the correlation length for density fluctuations diverges, and s0 exceeds the wavelengths and we assumed no carelation from one cll to the other. The mare detailed consideration needed to complete the argument wes given in 1914 by Omstein and Zernicke; bt that is outside the scope of this course, 6 The Optical Theorem A general result which was first encountered in the context of optics, which is a consequence of causality (like the Kramers-Krtnig relations) is the optical theorem. ‘This relates the total cross-section to the swattering amplitude in the forward direction. ‘We saw above that, when an incident dectromagnetic wave of frequency w % scattered from a Tocalised scatterer, at large distances from the scatterer we may write +(3b)_- From this pop+ 4, and the differential cross-section for scattering is then given by do _ le AP aa (DP It can be shown (Kinchoff) that Frcseskose) 1 Ap IDol which is the scattering amplitule for scattering from an initial wave vector and polarisation ko,€ to a final wave vector and polarisation ke (the square modulus of which gives the corresponding differential ‘aoss-scction), may also be expressed a F Li Bere [oe xB) +e (ex (nx Ba)]as wher Bs, Es describe the scattered wave, and the integration is over any dosed surface $ which encloses the ‘scatterer, the outward normal from which is written n. The derivation of this formula makes use of Green’s ‘theorem, and although not difficult will be omitted.‘The total field in the presence of the scatterer is described by EB=K)+E, B=B)+B,. ‘The scatterer will absorb radiation, as well as scatter it, and the time-averaged absorption power, Pajys is given fram the inward flow of energy as at a Pate =~ yy [UBB nha, (the minus sign because we want the invard flux of energy, and the factor of a half because of the time averaging). On the other hand the time averaged scattered power is Peattd = Hp | HOE BH) m5 so that the total power removed frm the incident wave, either by absorption or by scattering is, P Hay [MES +B} xB.) nas, Setting Bo = Faeye¥©™, By = fo x Bo gives P. ByMs [ome “In xB, +ko x (nx E,)/oK] dS}. ‘We have previously determined the differential cross-section for scattering, and by integrating, this over solid angles have obtained the scattering cross-section 7. What we are now interested in is the total cross-section: tots Which also takes into account the absorption as well as the scattering of radiation. This is defined as the power removed from the incident radiation divided by the incident power crossing a unit: area, Le., divided by the incident, flux. So ng =P tot elEoPe?2 = REEgtc [cP 1B, + ho x (XE) /oK] a5}. ‘Now compare this with the scattering amplitude in the forward direction, for unaltered polarisation (dastic scattering) F(kos€0sk0360) Belews [en xB, + lo x (nx B,)]d5, and condude that A tot = FSF eorand chstic). ‘This is the optical theoran. A review and a more complete derivation can be found in the Supplementary Notes. 10