l Introduction Physics would be dull and life most unfulfilling if all physical phenomena around us were linear. Fortunately, we are living in 2 nonlinear world. While linearization beautifies physics, nonlinearity provides excitement in physics. This book is devoted to the study of nonlinear electromagnetic phenomena in the optical region which normally occur with high-intensity laser beams. Nonlinear effects in electricity and magnetism have been known since Maxwell's time. Saturation of magnetization in a ferromagnet, electrical gas discharge, Tectification of radio waves, and electrical characteristics of p—n junctions are just a few of the familiar examples. In the optical region, however, nonlinear optics became a subject of great common interest only after the laser was invented. It has since contributed a great deal to the rejuvenation of the old science of optics. 1.1 HISTORICAL BACKGROUND The second harmonic generation experiment of Franken et al.! marked the birth of the field of nonlinear optics. They propagated a ruby laser beam at 6942 A through a quartz crystal and observed ultraviolet radiation from the erystal at 347t A. Franken’s idea was simple. Harmonic generation of electro- magnetic waves al low frequencies had been known for a long time. Harmonic generation of optical waves follows the same principle and should also be observable. Yet an ordinary light source is much too weak for such an experiment. It generally takes a field of about 1 kV/cm to induce a nonlinear response in a medium. This corresponds to a beam intensity of about 2.5 kW/cm. A laser beam is therefore needed in the observation of optical harmonic generation. Second harmonic generation is the first nonlinear optical effect ever ob- served in which a coherent input generates a coherent output. But nonlinear optics covers a much broader scope. It deals in general with nontinear 1Zz Introduction interaction of light with matter and includes such problems as light-induced changes of the optical properties of a medium. Second harmonic generation is then not the first nonlinear optical effest ever observed. Optical pumping is certainly a nonlinear optical phenomenon well known before the advent of lasers. The resonant excitation of optical pumping induces a redistribution of populations and changes the properties of the medium. Because of resonant enhancement, even a weak light is sufficient to perturb the material system strongly to make the effect easily detectable. Low-power CW atomic lamps were used in the earlier optical pumping experiments on atomic systems. Optical pumping is also one of the effective schemes for creating an inverted population in a laser system. In general, however, observation of nonlinear optical effects requires the application of lasers. Numerous nonlinear optical phenomena have been discovered since 1961. They have not only greatly enhanced our knowledge about interaction of light with matter, but also created a revolutionary change in optics technology. Each nonlinear optical process may consist of two parts. The intense light first induces a nonlinear response in a medium, and then the medium in reacting modifies the optical fields in a nonlinear way. The former is governed by the constitutive equations, and the latter by the Maxwell’s equations. At this point, one may raise a question: Are all media basically nonlinear? The answer is yes. Even in the case of a vacuum, photons can interact through vacuum polarization. The nonlinearity is, however, so small that with currently available light sources, photon-photon scattering and other nonlinear effects in vacuum are still difficult to observe. So, in a practical sense, a vacuum can be regarded as linear. In the presence of a medium, the nonlinearity is greatly enhanced through interaction of light with matter. Photons can now interact much more effectively through polarization of the medium. 1.2 MAXWELL'S EQUATIONS IN NONLINEAR MEDIA All electromagnetic phenomena are governed by the Maxwell’s equations for the electric and magnetic fields E(r, ¢) and B(r, £): 1 0B VXE= om 1ldE 4a VXBa Cato qa) Vv -E=42p, vy B=0 where K(r, ) and p(r, 1) are the current and charge densities, respectively. TheyMaxweli’s Equations in Nonlinear Media 3 are related by the charge conservation law gy 4 he v I+ 3 0 (1.2) We can often expand J and p into series of multipoles: =e Sieg. ves J=3,+ aT cev x Mt aly Qyt--, (13) P= Mm —F-P- VV Qt. Here, P, M, Q,..., are respectively the electric polarization, the magnetization, the electric quadrupole polarization, and so on. However, as pointed out by Landau and Lifshitz,* it is not really meaningful in the optical region to express J and p in terms of multipoles because the usual definitions of multipoles are unphysical. In many cases, for example in metals and semicon- ductors, it is more convenient to use J and p directly as the source terms in the Maxwell's equations, or to use a generalized electric polarization P defined by ap J dy + op (1.4) where Jj, is the de current density, In other cases, the magnetic dipole and higher-order multipoles can be neglected. Then, the generalized P reduces to the electric-dipole polarization P. The difference between P and P is that P is a nonlocal function of the field and P is local. In this book, we assume electric dipole approximation, P = P, unless specified. one With (1.2) and (1.4), the Maxwell’s equations appear in the form (1.5) where P is now the only time-varying source term. In general, P is a function of E that describes fully the response of the medium to the field, and it is often known as the constitutive equation. If we could just write the constitutive equation and find the solution for the resulting set of Maxwell's equations with appropriate boundary conditions, then all optical phenomena would be pre- dictable and easily understood, Unfortunately, this seldom is possible. Physi- cally reasonable approximations must be resorted to in order to make the mathematical solution of the equations feasible. This is where physics comes into play.4 Totreduction The polarization P is usually a complicated nonlinear function of E. In the linear case, however, P takes a simple linearized form PO) = [0 xO 01 e EE 1) ara? (1.6) a where x") is the linear susceptibility. If E is a monochromatic plane wave with Ear, t) = E(k, wo) = &(k, w exp(ik-r — iwe), then Fourier transformation of (1.6) yields the familiar relation P(r, 1) = P(k, w) (7) = x(k, @) E(k, o) with ef x(k, wo) =f xr, Dexpl—iker + iwt)dr dt. (1.8) The linear dielectric constant e(k, w) is related to xM(k, } by e{k, we) = 1+ 4a (k, 2), (1.9) In the electric dipole approximation, x(r, 1) is independent of r, and hence both x(k, w) and e(k, w) are independent of k. In the nonlinear case, when E is sufficiently weak, the polarization P as a function of E can be expanded into a power series of E: P(r, ¢} =f" xO — 0 tt) E(Y, td’ de -o +f x(t git a 2) Ele 4) 00 XE(r,, t,)dr, dt, dr, de® . (1.10) +f XO nb TE by 00 Tf fy): B41) E(t, t, ECs, ty), dt, dr, dt, diy dt; + - + where x‘") is the 1th-order nonlinear susceptibility. If E can be expressed as a group of monochromatic plane waves E(r, 4) = DE(k,, «), (layAnharmonic Oscillator Model 5 then, as in the linear case, Fourier transform of (1.10) gives P(k, @) = Pk, w) + PA(Ko) + PP(kw) ts (1.12) with POkK, o) = x(k, w) E(k, @), P&(k,o) =YO(k =k; + kj, @ = 0; + 0): Elk;, @JE(k,, @)), 1.13 P@(k,o) = x(k =k, +k, +h, o = 0, + a + a) (1.43) ZE(k,, a E(k, « JE(K;, @), and xO(K Hk, +k, t + +k, =a, + +--+ w,) 02 = fp eT _ [xP tit te to) a4) e771) ~ ert 4) + ~ Herta) walt Old ey dt, --+ dr, dt, Again, in the electric dipole approximation, x'(r, t) is independent of r, or x(k, w) independent of k. The linear and nonlinear susceptibilities characterize the optical properties of a medium. If x” is known for a given medium, then at Jeast in principle, the sth-order nonlinear optical effects in the medium can be predicted from the Maxwel!’s equations in (1.5). Physically, x") is related to the microscopic structure of the medium and can be properly evaluated only with a full quantum-mechanical calculation. Simple models are, however, often used to illustrate the origin of optical nonlinearity and some characteristic features of x0, We consider here the anharmonic oscillator model and the free electron gas model, 1.3 ANHARMONIC OSCILLATOR MODEL In this model, a medium is composed of a set of W classical anharmonic oscillators per unit volume. The oscillator describes physically an electron bound to a core or an infrared-active molecular vibration. Its equation of motion in the presence of a driving force is 2 3 + ré + bx + ax? = F, (4.15)6 Introduction We consider here the response of the oscillator to an applied ficld with Fourier components at frequencies +, and +04; Fm TLE (ei + elt) + Byler + oyh, (2.16) The anharmonic term ax? in (1.15) is assumed to be small so that it can be treated as a perturbation in the successive approximation of finding a solution: wax xOq xO oe, (1.17) The induced electric polarization is simply P = Ngx. (1.18) The first-order solution is obtained from the linearized equation of (1.15): x = xf) + xO la) + ec. (a/m)E, (1.19) x) (qg,) = pn gto («) (a - a? —ia,F) where c.c. is a complex conjugate. Then, the second-order solution is obtained from (1.15) by approximating ax? by ax‘: x = x(a, + gay) + xO(co, ~ oy) + xO(2Wy) + xP (2w,) + x0) + ccc =2a(q/my' E,E, (43 ~ ot ~ io P)(o} - 04 F ie, D) 1 [of Cw, +) i x(a, + o,) = ity toe at w,)T} (1.20) _ 2p x®(20,) = amy Pas (of — oF - to )"(08 - 40? - i200) x20) = -a( 4) 2,(-3— + __—. mM! w\ eg — of ~ fol grat) By successive iteration, higher-order solutions can also be obtained. As seen in the second-order solution, new frequency components of the polarization at @, + W, 2, 22, and 0 have appeared through quadratic interaction of the field with the oscillator via the anharmonic term. The oscillating polarization components will radiate and generate new em waves at @, + @, 2a, and 2a.Anharmonic Oscillator Model 7 Thus, sum- and difference-frequency generation and second-harmonic genera- tion are readily explained. The appearance of the zero-frequency polarization component is known as optical rectification. More generally, frequency compo- nents at w = 7,00, + 1,0, with », and #, being integers, are expected in the higher-order solutions. In this model, the anharmonicity @ determines the strength of the nonlinear interaction. Treating ax? as a smail perturbation in the foregoing calculation is equiva- Ient to the assumption that £ is small and P can be expanded into power series of £. We can give a rough estimate on how the nonlinear polarization should diminish with increasing order. Assuming the nonresonant case with w, > #, and w,, we find from (1.19) and (1.20), pe PO |_| gee Po . (21) may For an electron bound to a core, if x is so large that the harmonic force mw5x and the anharmonic force max? are of the same order of magnitude, then both will be of the same order of magnitudes as the total binding force on the electron |g£,,|: |qE ail ~ mwgx ~ max? or mt 4 aE al ~ “8 2. (1.22) Equation (1.21) then becomes P® E [5 Fal (1.23) In fact, one can show in general port E eae 020 such that |Z/£,,| acts as an expansion parameter in the perturbation calcula- tion, Typically, E,, ~ 3 x 10° V/om. The E field for a 2.5-W/crm laser beam is only 30 V/em with |£/E,,| ~ 10-’. The nonlinear polarization is much weaker than the linear polarization. This suggests that the observation of nonlinear optical effects requires high-intensity laser beams, Relation (1,24), however, is true only for optical frequencies away from resonance. Near resonance, the resonant denominators may drastically en-8 Introduction hance the ratio [P("+)/P™|. Consequently, the nonlinear effects can be detected with much weaker light intensity. Optical pumping is an example. With resonant enhancement, it may even happen that |PO*)/P(| > 1. When ihis is the case, the perturbation expansion is no longer valid, aid the full nonlinear expression of P as a function of E must be included in the calculation. The problem then falls into the domain of strong interaction of light with matter. 14 FREE ELECTRON GAS A simple but realistic model to illustrate optical nonlinearity in a medium is the free electron gas model. It properly describes the optical properties of an electron plasma. The simplified version of the model starts with the equation of motion for an electron ar dt = -E(E+ 2x8}, (1.25) m € Damping is neglected here for simplicity. Clearly, the only nonlinear term in this equation is the Lorentz force term. Since v « ¢ in a plasma, the Lorentz force is much weaker than the Coulomb force, and then (e/mc)y X B in (1.25) can be treated as a perturbation in the successive approximation of the solution. For E = dye"r'o "+ Geter! + 6.0, we obtain e jes, TOC) = Beltre + oe, mo; ie? mm'w0,(w, + w)) (1.26) x[@, x(k, x &) + & x(k, x &)] 1( a, + ay) = x ett the) KO —Kertorde 4 6, and so on. For a uniform plasma with an electronic charge density p, the current density is given by JaIVE IO $ ce = por ep O4 (1.27) °F (PD +724 ---) with, for example, IM, + 0) = pp e(w, + ay)Free Electron Gas 9 and so on. This shows explicitly how an electron gas can respond nonlinearly to the incoming light through the Lorentz term. In a more rigorous treatment of an electron gas, we must also take into account the spatial variations of the electron density p and velocity ¥. Two equations, the equation of motion and the continuity equation,® are now necessary to describe the electron plasma: yy. ~~ hE L ) ral v= mp met oy xB and {1.28} ap atv (oy=0 where p is the pressure and om is the electron mass. The pressure gradient term in the equation of motion is responsible for the dispersion of plasma reso- nance, but in the following calculation we assume Vp = 0 for simplicity. Then, coupled with (1.28), is the set of Maxwell’s equations 1 0B VX BO gy xp Loe _ and _ Aepe (1.29) c at c c VE = 4n(p — p), and V'B=0 We assume here that there is a fixed positive charge background in the plasma to assure charge neutrality in the absence of external perturbation. Successive approximation can be used to find J as a function of E from (1.28) and (1,29). Let® p= pM + pO + PMH oo, ya rO4yOH ---, and jD+je+ (1.30) with JO = py and (1.31) ji? = phy + poyth,10 Introduction We shall find the expression for j(2e) as an example assuming E = & x exp(/k+r — dwt). Substitution of (1.30) into (1.28) and (1.29) yields dv(w)} € a siev) = ert ia) aoe) = —iwp™ = — Vy -(p%™) (1.32) VE = dap), Bv2)(200) € a =o (yO. a) £. yay 3 Rew (¥®. vy mov XB. The second-order current density is then given by J@(20) = I 2a (1.33) + Tama - BNE. The last term in (1.33) has the following equalities: @ é i47me (v-ERE= - na l¥ “(vie 2 we =a LY {PEE (1.34) __te? | _ve-E le me? | 1 — 02 fo? where &, = (4mpe/m)'” is the plasma resonance frequency. With (1.34), = (cfiwy E, and the vector relation EX (y x E) +(E-V)E= 5 Tou. E), the current density in (1.33) can be written as ipMe 39 Qo) ~ PE o(E-F) (e? ie? | yp-E - ££ ver tala (0.35) Ames mney w/o Equation (1.33) shows explicitly that aside from the Lerentz term, there are also terms related to the spatial variation of E. They actually arise from the nonuniformity of the plasma. In a uniform plasma, yp =0 and hence Vv +E =0 from (1.34). This means that k is perpendicular to E and therefore (E+ VE also vanishes. The Lorentz term is then the only term in J@(2w),References ity The induced current density J@(2w) should now act as a source for second harmonic generation in the plasma. In a uniform plasma with a single pump beam, J°(2w) « EX B is only along the direction of beam propagation. Since an oscillating current cannot radiate longitudinally, no coherent second harmonic generation along the axis of beam propagation is expected from the bulk of a uniform plasma. In the bulk of a nonuniform plasma or at the boundary surface of a uniform plasma, however, it is possible to find second harmonic generation through the nonvanishing ve. Equation (1.35) shows that when vp ¥ 0, the nonlinear response of the medium J?(2w) is greatly enhanced if w is near the plasma resonance. From the general principle, the nonlinear response of a medium is resonantly enhanced when the incoming field hits the resonance of the medium. One should, of course, also expect a resonant enhancement in the second harmonic generation when 2 is at resonance. This actually does come in through the response of the second harmonic field to J(2w), As 2 — w,, the current density will excite the longitudinal field at 2w resonantly. As shown in (1.33) or (1.35), the current density J@(2w) depends exclu- sively on the spatial variation of E. In fact, using vector identities, the expression of J@(2) in (1.33) or (1.35) can be put into the form’ J@(2w) = eV X( )—i2wv -( ). Comparing it with (1.3), we recognize that the two terms in J@(2q) represent the magnetic dipole and electric quadrupole contributions, respectively. No induced electric dipole polarization exists in a plasma. Also, the induced electric quadrupole polarization depends on the gradient of the electric field, and therefore cannot show up in the bulk of a uniform plasma. The free electron model here is applicable to a number of real problems. First, it can be used to describe the optical nonlinearities due to plasmas in metals and semiconductors. Second harmonic generation from metal surfaces is readily observable.® Then, with some modification to take into account the net charge distribution, the nonvanishing vp, and so on, it can also be used to describe the optical nonlinearities of a gas plasma. Various nonlinear optical effects in gas plasmas have been observed. They will be discussed in some detail in Chapter 28. The model has also been used to describe the observation of nonlinear effect in a crystal in the X-ray region.’ The electron binding energy is much weaker than the X-ray photon energy, and therefore the electrons in the crystal will respond to the X-ray as if they were free. REFERENCES 1 P.A. Franken, A. E. Hill, C. W. Peters, and G. Weinreich, Pays, Reo. Leif. 7, 118 (1961). 2. See, for example, G. Rosen and F. C, Whitmore, Phys. Rev. 1378, 1357 (1965) 3 J.D. Jackson, Classical Electrodynamics (McGraw-Hill, New York 1975), 2nd ed. p. 739; W. K. H Panofsky and M. Phillips, Classical Etectricity and Magnetism (Addison-Wesley, Reading, Mass., 1962), p. 131.12 Introduction 4 LD. Landau and & M, Lifshitz, Electrodynamics in Continuous Media (Pergamon Press, New York, 1960), p. 252. 5. 1. D. Jackson, Classical Electrodynamics (McGraw-Hill, New York, 1975), 2nd ed, p. 469. 6 N. Bloembergen, R.K. Chang, S. §. Jha, and C, H. Lee, Phys. Rev 174, 813 (1968). 7 PS, Pershan, Phys. Rev. 130, 919 (1963). 8 P.E. Eisenberger and §. L. McCall, Phys, Rev. Lett. 26, 684 (1971). BIBLIOGRAPHY Blocmnbergen, N., Nonlinear Optics (Benjamin, New York, 1965). Bloembergen, N., Rev. Mod, Phys, 54, 685 (1982).2 Nonlinear Optical Susceptibilities For tower-order nonlinear optical effects, nonlinear polarizations and nonlin- ear susceptibilities characterize the steady-state nonlinear optical response of a medium and govern the nonlinear wave propagation in the medium. Chapter 1 showed how the nonlinear optical response can be calculated for two model systems. Chapter 2 gives a more general discussion of nonlinear susceptibilities starting from the microscopic theory. 2.1 DENSITY MATRIX FORMALISM Nonlinear optical susceptibilities are characteristic properties of a medium and depend on the detailed electronic and molecular structure of the medium. Quantum mechanical calculation is needed to find the microscopic expressions for nonlinear susceptibilities: Density matrix formalism is probably most convenient for such calculation and is certainly moze correct when relaxations of excitations have to be dealt with.” Let } be the wave function of the material system under the influence of the electromagnetic field. Then the density matrix operator is defined as the ensemble average over the product of the ket and bra state vectors p=) (2.1) and the ensemble average of a physical quantity P is given by P) = (P) = (IPR) (22) = Tr(pP} Tn our calculation here, P corresponds to the electric polarization. From the pK}“4 Nonlinear Optical Susceptibilities definition of p in (2.1) and from the Schrédinger equation for le), we can readily obtain the equation for motion for p, a 1 FF = ql, el), (2.3) known as the Liouville equation. The Hamiltonian is composed of three parts, Ho y+ Hing + aston (2.4) In the semiclassicat approach, % is the Hamiltonian of the Waperturbed material system with eigenstates jn) and eigenenergies E,, so that Hin) = E,.|n), 6. is the interaction Hamiltonian describing the interaction of light with matter, and random i8 8 Hamiltonian describing the random perturbation on the system by the thermal Teservoir around the system. The interaction Hamiltonian in the electric dipole approximation is given by Hig = FE. (2.5) We consider here only the clectronic contribution to the susceptibilities. For the tonic contribution, we would have to replace er - E by — £,4;R, + E with g, and R, being the charge and position of the ith ion, respectively. The Hamiltonian 4, om i8 Fesponsible for the relaxations of material excitations, or in other words, the relaxation of the perturbed p back to thermal equi- librium. We can then express (2.3) as?:* 4 1 4, om gl % + Ha 01 +(P) (2.6) with a, 1 (See Al Manson OL 2.7) If the eigenstates |n) are now used as base vectors in the calculation, and |¥) is written as a linear combination of Ja), that is, |p) = L,@,l2), then the physical meaning of the matrix elements of p is clear. The diagonal matrix element p,, = ¢ Hlplat) = ja, Tepresents the population of the system in state la}, while the off-diagonal matrix element 9, = {n|p|n’) = @,a*. indicates that the state of the system has a coherent admixture of Jn) and Ja’), In the latter case, if the relative phase of a, and a,, is random (or incoherent), then Pan = 0 through the ensemble average. Thus at thermal equilibrium a} is given by the thermal Population distribution, for example, the Boltzmann distribution in the case of atoms or molecules, and pf = 0 for n # n',Density Matrix Formalism 15 We can use a simple physical argument to find a more explicit expression for (09/08) ,.1... The population relaxation is a result of transitions between states induced by interaction with the thermal reservoir. Let Way de the thermally induced transition rate from (7) to \n’). Then the relaxation rate of an excess population in |7) should be (%) = TM ale ~ Worn danl (23) At thermal equilibrium we have an 0) 0) Br Ld Mar gB8t — Wy] = 0. (2.9) ‘s Therefore, (2.8) can also be written as a Felon aan — ALC] = E [Hy ale — 08%) — Wy ear Pan ~ ®P)]- (2.10) The relaxation of the off-diagonal elements is more complicated.” In simple cases, however, we expect the phase coherence to decay exponentially to zero. Then, we have, for n + 7’, ODay cd @.n) with [7,7 = Dp} = (Ten being a characteristic relaxation time between the states |) and |’). In magnetic resonance, the Population relaxation is known. as the longitudinal relaxation, and the relaxation of the off-diagonal matrix elements is known as the transverse relaxation. In some cases, the longitudinal relaxation of a state can be approximated by a - i Pen ~ A ete = (TDS (One — 02). — @42) Then 7, is called the longitudinal relaxation time, Correspondingly, 7; is called the transverse relaxation time. Thus, at least in principle, if 3%, Hi, and (40/9)... ate known, the Liouville equations in (2.6) together with (2.2) fully describe the response of the medium to the incoming field. It is, however, not possible in general to combine (2.6) and (2.2) into a single equation of motion for (P). Only in special cases can this be done. In this chapter we consider only the case of steady-state response with (P) expandable into power series of E. The tran- sient response is discussed in Chapter 21.16 Nonlinear Optical Susceptibilities To find nonlinear polarizations and nonlinear susceptibilities of various orders, we use perturbation expansion in the calculation. Let P= p+ pO 4 py... and (P) = (PM) 4 (PA 4... (2.13) with (PO) = Te(pP) (2.14) where p is the density matrix operator for the system at thermal equilibrium, and we assume no permanent polarization in the medium so that and so on, We are interested here in the Tesponse to a field that can be decomposed into Fourier components, E = L,Sexpik, «+ — i,t). Then, since Hig = LH ga (,) and H,(0,) © é,exp(—iw,t), the operator p™ can also be expanded into a Fourier series 1 = F(a). i With do™(a,)/at = ~ia,p"%(a,), (2.15) can now be solved explicitly for e(«o,) in successive orders. The first- and second-order solutions are [inn 2D) a Ala, — w,, + iT, — LM) Cod ag + alee), 00,)],. rela + ex) = * h(a, + wy — Wy + iT, _ 1 © Ala, +o, — 0, ily) (2.16) XE (ad) alee) ~ Ph (4) ine (0) , 4. + [ig ()] anrP 2h-(e5,) — PAPC) on (oe) yg We use here the notation Any = (n|Aln’). Higher-order solutions can be Pra) = ) (of) - po) 7. ~E TTERarreeeerMicroscopic Expressions for Nonlinear Susceptibilities 17 obtained readily, although the derivation is long and tedious. Whenever diagonal elements p{)(0) appear in the derivation, further approximation on (DP ym/ Ot) eetax im (2.8) is often necessary in order to find a closed-form solution. We also note that the expression for pG)(w, + 4) in (2.16) is valid even for n = n' as long as w, + w, + 0 since the term (90/7/42) ia, Cam then be neglected in the calculation. 2.2 MICROSCOPIC EXPRESSIONS FOR NONLINEAR SUSCEPTIBILITIES ‘The full microscopic expressions for the nonlinear polarizations (P‘) and the nonlinear susceptibilities (x'")) follow immediately from the expressions of oe), With #,, = ereE and P = —Ner in (2.14) and (2.16), the first- and second-order susceptibilities due to electronic contribution are readily ob- tained. They are given here in explicit Cartesian tensor notation: PMYw) E,(#) -»55 (te) ne) en ag ben ro - = F et eto, +iT,, wo, + iT, xe) = xXF(o = w, + @2) _ P2(@) E, edixles) Dan Daw Dag (o- Wqg + Tyg )(@, — Oye + Tyg) (alate, (o — , + Ty, )(0, — Og + Tyg) / 1 (deel) en Das 2.17) (w+ o,, + Tyo: + oy, + Ty.) : “yeltienl Dn (o + wy, + FT, Ma + 0, + Ty.) _ dag arate) gn 1 1 (o- 6, +71,,) (= Hyg + ily, |) — Oy + | aX, +. +§ —____ (0 = Wy + Tyg) 4 Oz — Og + Tyg | Oy + yy + Tyg _ (ota 1 1 ij18 Nontinear Optical Susceptibilities There are two terms in x{P and cight terms in x. The calculation can be extended to third order to find x9; (@ = @ + « i ws), which will have 48 terms. The complete expression for xP it Is given | in the literature? and is not reproduced here, The resonant structure of x§ke however, is discussed in Chapter 14. In nonresonant cases, the damping constants in the denominators in (2.17) can be neglected. The second-order susceptibility can then be reduced to | 2 form with six terms, noting that the last two terms in the expression for in (2.17) become CD planes (Dal Dancl Dag (1 — Wag h(ory + yg) (1 + ayy ey = Gg)” With W denoting the number of atoms or molecules per unit volume, the expressions in (2.17) are actually more appropriate for gases or molecular liquids or solids, and pf is given by the Boltzmann distribution. For solids whose electronic properties are described by band structure, the eigenstates are the Bloch states, and oo corresponds to the Fermi distribution. The expression for xf) and x@} should then be properly modified. Since the band states form essentially a continuum, the damping constants in the resonant denominators can be ignored. In the electric dipole approximation with the photon wavevec- lor dependence neglected, x7}. for such solids has the form? xii (@ = @ + oy) - -£ fa (See O Cate fo [e ~ ea) for — evea)] (oaleles acer aleade’, a) <<" algl2.q) [we - o,,(] [4 - oa] Cr alyle, DXe. aele” Ke". ake.) [w+ a,,(q)] [er + @,,(a)] (2.18) (o.alglesay Cc alrate ¢eglele.q) [e+ e-.(@} [oy + ea) (e.dbele, Xe, dele’ )< 6" alergl2 [a ~ #.,(@)] [4 + oa) (Balle, a " 1 a vp on | I 3 2 3 % ta> 1 oy . - Fig. 2.1 A representative double-Feynman \g> {g| to |n’) olf! < gl ig> pg og ol eg oig'] uy wy wf wy lg>agl pig ] 1 3 Ig> aaa, 1 3 Wz wr oy \g> <3t ig> Eq -“2P. (2.21) a The polarization can be expressed in terms of either microscopic polarizabili- ties and local fields or macroscopic susceptibilities and applied fields: Po) = N{ a [Bpe(oy]} + atl Biel] LEnc(@adde + = (2.22) = x (0) + ARE (oder) +: With (2.20) and (2.21), the frst expression in (2.22) becomes p(w) = N[1- 42 na(a)| (2.23) x {aE (ao) + af Eiger) sf Ervel 2) + °- }. If the contribution of P'” to E,, with 1 > 1 is neglected [which is usually an excellent approximation since |P‘,,, « |p], then the local field can be written as Ex (w,) = [a - Sel(o)| Ea). (2.24) Then, from (2.22) and (2.23), we find Naw) X(#) = Tag A) Nal a) xG,(@= @ + 2) (2.25) _ Naf) [1 — (42/3) Ne o)][1 — (4/3) Na" (o,)] [1 — (40/3) Na(; Permutation Symmetry of Nonlinear Susceptibilities 5 and more generally x! (@ = @, + a, + --- + 4,) 2.26, Na (a = 0, + a, + --- + 0,) (2.26) * F—(44/3) Nal (o)] [1 — (49/3) Na (w))] --- [1 — (49/3) Na™(,)] Since the linear dielectric constant 2) is related to x" by al + (89/3) Na? 51 + dq = LESH) Na . OK T= (40 7NO we can write aq “1 M42 — 7 Neh} = 5 = [a 3 No! | 3 and (2.26) becomes? x (@ = @, te to te,) = NLM aM (w = 2 +o + + my) (2.27) with a ad +2 e(a,) +2 co) = [lop +21 Mod +2) | la) +2 i | 3 3 3 228) being the local field correction factor for the #th-order nonlinear susceptibili- ties. In media with other symmetry, the expression (2.27) is still valid, but L°”? will be a complicated tensorial function of e@(w), e(0,),-.., and e(a,).® 2.5 PERMUTATION SYMMETRY OF NONLINEAR SUSCEPTIBILITIES There is inherent symmetry in the microscopic expressions of susceptibilities. As can be readily seen from (2.17), the linear susceptibility x{} has the symmetry xP(e) = xP(-o), (2.29) which is actually a special case of the Onsager relation. Similarly, the nonlinear susceptibility x93}.(w = ws + wy) in (2.17) or a similar expression for x Si(2w = w+ @) has the following permutation symmetry when the damping con-6 Nonlinear Optical Susceptibilities stants in the frequency denominators can be neglected (ic., the nonresonant casesy:}? XB = 0, + 0) = xBilor = — er + #) = xf} (a, = 0 ~ @), (2.30) xB Ge =e +e)= 1xB(w = le — 0) = fx@lla = -w + 20). In the permutation operation, the Cartesian indices are permutated together with the frequencies with their signs properly chosen, More generally, one can show that the ath-order nonlinear susceptibility also has the permutation symmetry® XA nah oo = coy Fy tO) = XR gale = aa + Oe + «) =e (2.31) =P, (Un =O Oy Ode If the dispersion of x‘ can also be neglected, then the permutation symmetry in (2.31) becomes independent of the frequencies. Consequently, a symmetry relation now exists between different elements of the same x“ tensor, that is, x§")__), remains unchanged when the Cartesian indices are permuted. This is known’ as Kieinman’s conjecture,” with which the number of independent elements of x" can be greatly reduced. For example, it reduces 27 elements of x to only 10 independent elements. We should, however, note that since all media are dispersive, Kleinman’s conjecture is good approximation only when ail frequencies involved are far from resonances such that dispersion of x) is relatively unimportant. 2.6 STRUCTURAL SYMMETRY OF NONLINEAR SUSCEPTIBILITIES As optical properties of a medium, the nonlinear susceptibility tensors should have certain forms of symmetry that reflect the structural symmetry of the medium. Accordingly, some tensor elements are zero and others are related to each other, greatly reducing the total number of independent elements. As an illustration, we consider here the second-order nonlinear susceptibility tensor (2 Each medium has a certain point symmetry with a group of symmetry operations {5}, under which the medium is invariant, and therefore x‘” remains unchanged. In real manipulation, S is a second-rank three-dimen- sional tensor S,,,- Then, invariance of © under a symmetry operation isTable 2.1 Independent Nonvanishing Elements of x{w = @; + 2) for Classes Ceystals of Certain Symmetry Symmetry Class Independent Nonvanishing Elements Triclinic. 1 All elements are independent and nonzero Monoclinic 2 AYE, KEY, AXP, LV, VAX, VY, VEE, VIN, PAZ, ZZ, 12y, 1xp, zyx (two fold axis parallel to §) m HUM, YY, NOE, XEN, XXZ, YYZ, VEY, WAV IPN ENN, 2), 222, 27%, zxz (maitror plane perpendicular to $) Orthorhombic 222 IE, KEY, VEX, PRE, 2MY, YX mm2 XIX, AXZ, VIZ, VZV, ZMK, IVY, ZIT Tetragona} 4 Aye —yxz, MZ = — VIX, XOX = yLy, XkZ = pyz, IXX = Zyy, 12z, 2xy = — zyx 4 Ay! = Nz, XIy — yox, XEX = — yey, Xxz = —yyz, 2Xx = — Zyy, xy = 2px 42 wyz = yur, xy = ~y2x, Dy = —2yx 4mm XZK = VZV, XAZ = YyZ, XX = Zyy, 22 Rm AYE = YUE, KEY * Yer, Ixy = ZX Cubic 432 aye = —xzy = yex = —yxz = 2xy = — 2px Bn xyz = xzy = yzx 23 = yaz = xy = x Trigonal 3 XXX = yy = —yyt yxy, AYE = —yxT, ZY = ~ yex, KIX = Y2V, HAZ YV2, YY > YK = — AY = AYA, Zxx = Iyy, £22, 2xy = 2px 32 XXX py pyd = — yxy, YI = —yxz, xy = —yzX, Ixy = — zyx 3m KIX = YOY, HXZ = YYZ, XX = ZV, 2275 YY = — YAK = — xaxy = —xyx (mirror plane perpendicular to ) Hexagonal 6 XyZ = — XZ, Xzy = —yZX, XZX = PX, XXZ = pyz, IXX = IVY, 222, 7x = — ZX 6 2o = Spy = —yxy = yy, Wy = — yet may axa 622 xyz = —ynz, zy = — pre, tay —2yx 6mm XZ = pry, HAZ = pyz, 2x = ZY, 122 Gm2 Wy sm ~yxx = —xxy = —xyx 27B Nonlinear Optical Susceptibilities explicitly described by (4 St) 9:8 JS) = xih- (2.32) For a medium with a symmetry group that consists of » symmetry operations, n such equations should exist. They yield many relations between various elements of x, although often only a few are independent. These relations can then be used to reduce the 27 elements of x to a small number of independent ones. ‘An immediate consequence of (2.32) is that x = 0 in the electric dipole approximation for a medium with inversion symmetry: with S being the inversion operation, 8 + @ = —é, (2.32) yields xf= x9}, = 0. This explains why x for a free electron gas does not have an electric dipole contribution as shown in Chapter 1, Among crystals without inversion symmetry, those with the zincblende structure such as the III-V semiconductors have the simplest form of x. They belong to the class of T,(43m) cubic point symmetry. Table 2.2 Independeut Nonvanishing Elements of x( = ©, + @ + 3) for Crystals of Certain Symmetry Classes ‘Symmetry Class Independent Nonvanishing Elements Triclinic All 81 elements are independent and nonzero Tetragonal HAIN = YPVYy TELZ, 422, 4mm, Y PEL = 22yY, IKK = XNZZ, AXP = YAK, YLYZ = ZVZV, 4/mmm, 2m LNZN = XOXZ, APXY = YAYX, YEAY = LVYZ, INNZ = XIIX, xyyx = aay Cubic XXKX = yyy = L2EZ, VYIZ = ZIXX ™ XXVY, 23, m3 LIYy = YYRX = ALL, ZVIy = EXE = YAVX, LYE = IKE = AYA, ZYVE = RTZK = YEAY, (PELy = LAME = KYYX 432, 43m, m3m XXKK = yyy 7222 pyte = ZZpy = 22XN = AXLZ = AXYY = VAX PLYZ = LyZy H IXTX = XENI = YXYX = APL PEI = Ly yl = IXXI = ALIX = YY = YRXY Hexagonal ELIZ, MXXX = Yyyy = Kxyy t ayyx + wpry 622, 6mm, AXYY — HYAA, APYX = YXXP, AYAY = YHVRs 6/mmm, 6m2 YE * KU, LEVY ~ ZUXX, Zyys = INKL, p2zp = XIIK, yZyz —* XEXL, LYZY = INN Isotropic XXX = yyy = 2z22, yA = LEVY LEXK ANCL = EXPY = YPNA, Yzy2 = Cydy = TREX = AZXZ = XYKY = YXYX OLY — ZYYE = EAKZ — AZZX = XYZ = YRKY, xxxx = xaxpy + xpxy + xyyxPractical Calculations of Nonlinear Susceptibilities a Although many symmetry operations are associated with 7,(43m), only the 180° rotations about the three four-fold axes and the mirror reflections about the diagonal planes are needed to reduce x. The 180° rotations make xf} = — xP = 0, xP = —x@ = 0, and x@ = —x@ = 0, where 7, j, and k refer to the three principal axes of the crystal. The mirror reflections lead to the invariance of x0.(i#j*£) under permutation of the Cartesian indices. Consequently, x5} (¢ + j # k) is the only independent element in x for the zincblende crystals. For other classes of crystals, the forms of x can be similarly derived through the corresponding symmetry operations. The symmetry consideration here is the same as the one used to derive the electrooptical tensor [which is actually a special case of x(a = w, + ,) with w, = 0] and the piezoelectric tensor.’ The forms of x for second-harmonic generation are in fact identical to the latter.'? We reproduce a part of x@(c = , + w,) for various classes of crystals in Table 2.1. The above symmetry consideration for x7 can of course be extended to higher-order nonlinear susceptibilities. In particular, symmetry forms for x? are most important in view of the many interesting third-order nonlinear optical effects that can be observed readily in almost all media, Table 2.2 lists the x tensors for the more commonly encountered classes of media.! 2.7 PRACTICAL CALCULATIONS OF NONLINEAR SUSCEPTIBILITIES Symmetry operations drastically reduce the number of independent elements in a nonlinear susceptibility tensor, but then for a given medium, we would also like to know the values of these independent elements, While they can often be measured (see, for example, Section 7.5), it is also important that they can be calculated from theory. A successful theoretical calculation can help in predicting x‘) for media not easily subject to measurements or for the design of new nonlinear crystals. In principle, the microscopic expressions, such as the one for xf in (2.17), with appropriate local-field correction, can be used for such calculations. However, in most practical cases these expressions are useless because neither the transition frequencies nor the wavefunctions for the material are sufficiently well known. This is especially true for large molecules or solids. Simplifying models or approximations often are needed. If ali frequencies involved are far from resonances, one simplifying assumption often used is to replace each frequency denominator in the microscopic expression of x‘) by an average one and bring all frequency denominators out of the summation (see, for example, x in (2.17). Then the summation over matrix elements can be greatly simplified through the closure property of the eigen- states and can be expressed in terms of moments of the ground-state charge distribution, The problem reduces to finding the ground-state wavefunction of the system.!?» Nonlinear Optical Susceptibilities The foregoing approximation, however, is too drastic to yield good results. A more successful calculation of x"" can be done by the bond model. Such a model was used in the early 1930s to calculate the linear polarizability of a molecule or the linear dielectric constant of a erystal.“* The bond additivity rule was assumed: the induced polarizations on a molecule (or a crystal) is the vector sum of the polarizations induced on all bonds between atoms. In other words, the bond-bond interaction is neglected. The same ruie can be used in the calculations of x"), We can. write x= Pay (2.33) K where of” is the nth-order nontinear polarizability of the Kth bond in the crystal (or medium), and the summation is over all the bonds in a unit volume. Thus, with known crystal structure, the calculation of x‘? reduces to the calculation of af for different types of bonds. We discuss here only the calculations of x”, using the zincblende crystals as an example. The general procedure is as follows. The linear bond polariza- bility a is first calculated as a function of the applied field using the recently well developed bond theory.!* The second-order nonlinear bond polarizability a® is then obtained from the first derivative of a{? with respect to the applied field. Finally, the summation of (2.33) over the bonds is performed to find x. We assume here that a simple crystal can be constructed entirely out of the same type of bonds, and the bonds are cylindrically symmetric. The linear susceptibility x of the crystal can then be written as xP = (Ea?) = Gal) + Bae (2.34) = (Gp + 0G) it where af" and a) are the polarizabilities parallel and perpendicular to the bond, g = af) /aj?, and Gj? and GQ are the respective geometric factors arising from the vectorial summation over the bonds. Both G{? and G?) are proportional to the number of unit cells per unit volume, For the zincblende structure, GP = £GQ = 4/3, and (2.34) becomes 4n xP = a The next step is to find an approximate expression for aj through x. The microscopic expression of x‘) in (2.17) away from resonance has the form (1 + 2) af”. (2.35) Ne? Vila xP= FY aye, of. (2.36) 5.0 OnEPractical Calculations of Nonlinear Susceptibilities RE In the low-temperature limit, 0g? = 0 for all states except the ground state. Then, following the approximation of replacing ,, in the denominator by an average ©, ,, and the sum rule* A rhe, Ogltlig = Fy? (2.37) (2.36) reduces to 2? w= —___? __ 2.38 xO Tae — a) (2.38) with 0% = 4%Ne?/m being the electron plasma frequency. This simplified expression for x{) has actually been shown more rigorously by Penn for solids in the limit of zero frequency.!? From (2.34), we now have @ CMa + EVa = (GO + pGP)aY = ——2-_ (2.39 ia at ( A ?) Hl 42(a, - «°) ) We are, however, interested in af’? as a function of applied field. The polarizability should depend on the field through the field perturbation on the transition frequencies and matrix elements. However, in the approximate form of (2. 39), af can depend on the field only through &?, jr: TO find an expression for a, is Where the bond theory comes in. Physically, aa, = =E, can be regarded as an average energy gap between the filled and unfilled states, It can be written as‘ E,=(ER+ 7]? (2.40) where £, and C are known as the homopolar and heteropolar gaps, respec- tively, and, in the bond theory, have the expressions and (2.41) In these expressions, a, b, and s are constant coefficients, Z, and Zp, are the valences, and 7, and r, are the covalent radii of the A and B atoms forming the bond, d = 1, + ry is the bond length, and exp(—k,d/2) is the Thomas—Fermi screening factor. If A and B are identical atoms, then C = 0. Equation (2.40) can be derived easily from molecular orbital theory.'* The bond electrons have two eigenstates, a bonding state and an antibonding state. The energy dif- po32 Nonlinear Optical Susceptibilities ference between. the two states is E,. For a homopolar bond (4 = 8), the bond electrons see a symmetric potential with respect to the bond center and E, = E,. For a heteropolar pond (4 # B), the bond clectrons see an anti- symmetric potential, and E? = Ef + C? with C proportional to the asymmet- ric part of the potential. The wavefunctions of the bonding and antibonding states along the bond are shown in Fig. 2.4. It is seen that in the heteropolar case, there is a charge transfer from the side of the less electronegative atom to the side of the more electronegative atom. According to the molecular orbital theory, the amount of transferred charge Q is related to the heteropolar gap C by Q=->. (2.42) Figure 2.4 also shows that there is a bond charge cloud between the two atoms. The magnitude of the bond charge derived from the bond theory is 2eE? 42 - ST E2+ way (2.43) Fig. 24 Sketches of electronic wavefunctions of (a) the bonding state and (6) the antibonding state along the bond connecting the atoms A and B. The solid curves are for the homopolar case and the dashed curves for the heteropolar case.Practical Calculations of Nonlinear Susceptibilities 33 Levine!? suggests that the bond charge may be considered as a point charge sitting at distances 7, and r,, respectively, from atoms A and 8. ‘We can now discuss how the bond polarizability changes when the bond is subject to an external field. The change occurs through the field perturbation on the charge distribution. In our description here, af? depends on the applied field E through the dependence of E, on E BE A (E PEn yp OC = E(B ad + SE (2.44) while E, and C depend on E through field-induced changes in the charge transfer and bond charge. However, since the applied field is not expected to change the bond length, we have 0#,/@E, = 0 from (2.41), The second-order nonlinear bond polarizability a} is obtained from dafP/dE,. If € and 4 denote the two directions parallel and perpendicular to the bond, respectively, then from the symmetry argument, only af, and a®), are nonvanishing. We atso neglect a®, by assuming that a field transverse to the bond will not significantly perturb the charge distribution. Thus af? is the only nonvanishing element of a, Using (2.39) and (2.44), we find dat aft = BE; 2 a2C ac (248) 4a(Gl0 + pG?)(E2 - Arar)? OFe Now, either (2.41) or (2.42) can be used to calculate 8C/d£,. The two, however, correspond to two different physical pictures. In (2.41), the applied field changes r, and rz, but keeps r, + rg = d. In terms of the simple model where the bond charge can be treated as a point charge sitting at distances r, and r, away from the atoms A and B, the field then simply shifts the position of the bond charge along the bond. This is known as the bond-charge model.!” In (2.42), on the other hand, it is the field perturbation on the charge transfer Q that relates C to the field. This is the charge-transfer model.” The bond-charge model involves, with Ar = Ar, = — Ary, ac -( ac ar (2.46) 9B, ~ \ Gr, ~ 3rq) 3B; and since gAr = afP(w’}AE,(w’) for w’ > 0, we find from (2.41), (2.45),cal Nonlinear Optical Susceptibilities (2.46), (2.34), and (2.38) 807aM(0")C Z Z, (0)nc= ar |e + toe ver Qag (GM + aGP (EZ - Wet) a fe (247) 2 , " = 80 [x0(o) xP Co") [E . Zales (cP +ucp)grat la The charge-transfer model following (2.42) gives Ee ac _ _ 16; ag (248) aE, ee €} OE, It is assumed in this model that the field-induced charge transfer is from atom B to atom A, treating the atoms as points. Since a{)(w)AE,(w'} = AQd, we have from (2.45) and (2.48) 202 7E3 Oi ay” (:@ex = —— ee) 20( Gl + nGP)}(E2 — A2a*) ede? __ BCE xP Co) xo} (GP + xe PY ede Ra? (2.49) We should, however, keep in mind that the description of how an applied field modifies the charge distribution in both models is still fairly crude. In reality, the electronic charges are broadly distributed in the region between the two atoms. The peak of the distribution is near the center of the bend. As an example, a contour map of the. valence electron distribution around a Ga—As bond obtained by empirical pseudopotential calculation is shown in Fig. 2.5.” In the presence of a de external field along the bond, the charge distribution becomes only slightly more asymmetric with its peak essentially unshifted. This is shown in Fig. 2.6 for the charge distributions along the Si-Si and Ga—As bonds,” The field-induced shift of the bond charge in the bond-charge model actually refers to the shift of the center of gravity of the valence electron distribution, while the field-induced charge transfer in the charge-transfer model refers to the redistribution of the valence charges around the bend from one side of the bond center to the other. Finally, we can obtain x of a given medium from af), for various bonds, where i,j, and & denote the three orthogonal symmetry axes in the crystal: x= L (OP) K ~ EGP ule), “Practical Calculations of Nonlinear Susceptibilities 35 ‘SUM OF VALENCE BANDS Fig. 25 Contour map of valence electron density distribution (in units of e per primitive cell) for Gas in the (1, - 1,0) plane. (From Ref. 21.) where (Gi), is a geometric factor for the A-type bonds reflecting the structure of the medium. We note that with (a{?,), expressed in terms of x rather than af! in (2.47) and (2.49), even the total field correction has been somehow taken into account in the above derivation. We now use InSb as an example to illustrate the calculation of x}. The crystal has a zincblende structure; therefore, the only nonvanishing elements of x are xf} with i 4 j + k, There is only one type of bond in the crystal: those connecting In and Sb. The geometric factor G2), is then given by 4N/3V3 and the density of unit cells N is related to the bond length d by N = 3v3 /16d?. We also have G) = 469) = 4N/3. From (2.47), (2.49), and (2.50), the bond- $i $i (a) ————_—_1-_| Ga As {w11) (b) Fig, 2.6 Sketches of the charge distribution along a bond in (a) Si and (b) GaAs. Solid and dashed curves refer to cases with and without an external field along the bond, respectively. (Courtesy of S. Louie.)36 Nonlinear Optical Suxceptibilities charge model gives y , (xB ye APLC w | Za Ze ]oe-nere (2.51) © 31 + 2p)*qh?@? Goon and the charge-transfer model gives sdud?CEs[x(o)]Tx™(o')] (x8. Ye tT. 4 (2.52) 3(1 + 2p)?ek} WO? We calculate here x9), in the low-frequency limit w ~ w’ ~ 0. For InSb, d= 25 AE, = 3.7 eV,E, = 3.1 eV, C=21 eV, x = 1.17 esu, AG, = 13 eV, Z,= 3, Zy=5, ty = ry = d/2, bexp(—Kk,d/2) = 0.12 e?, w= 4, and ¢ = 0.6 ¢,” we obtain (x2, )yc.= L6 X 1076 esu and (x@ Joy. = 2.3 x 10-8 esu. The results of both models are in fair agreement with the experimental value of x2, = (3.3 + 0.7) x 10~° esu. This should be considered satisfactory in view of the crude approximations in the models. The calculations can also be extended to higher-order nonlinear susceptibili- ties. However, because of the crude approximations involved, they become much less reliable. Also, since we use the covalent bonding picture in the models, the calculations are less suitable for ionic crystals. In nonlinear optics, we are often interested in materials with high nonlinearity. This discussion suggests that the materials should have high nonlinearity in bond polarizabili- ties, For large x@, the crystal structure should also be as asymmetric as possible so that there is a minimum of vectorial cancellation im summing over a of all bonds, The calculations here are good only in the low-frequency limit. The ap- proximations in the models break down when the optical frequencies are close to the absorption bands. Because of resonant enhancement, the transitions with transition frequencies closer to the optical frequencies contribute much more to the susceptibilities. In order to calculate x and its dispersion in these cases, we must use the full microscopic expression of x‘) such as those derived in Section 2.2. Then detailed information about the transition matrix elements and frequencies of the material is necessary. Such calculations have been carried out by several authors on x (2w) of zincblende semiconductors with various degrees of approximation. In most cases, constant matrix ele- ments are assumed. The more accurate calculations, however, are those with wavefunctions and energies of the band states derived from the empirical pseudopotential method,” which has been extremely successful in reproducing x(a) for zincblende semiconductors; it should therefore also yield accurate results for x2), An example is shown in Fig. 2.7 for InSb. The peaks and shoulders in the spectrum generally correspond to resonances of w or 2 with the critical point transitions. The results also show that it is important toMiller's Coefficient 37 & IXeya{ 2a)! x 1077 esu a lays (2ta} / 1x (2eo} quartz hooleV} Fig. 27 Dispersion of x{},(2«) of InSb calculated using the empirical pseudopoten- tial method. The peaks arise from interband transitions in the regions indicated. (From Ref. 24) include the dispersive effects of both the matrix elements and the density of states for transitions in the calculations. Full quantum mechanical caiculations of x of (2.17) for molecular crystals have also been carried out using semiempirical Hartree-Fock LCAO (linear combination of atomic orbitals) methods by many researchers.> They were able to predict quite satisfactorily the measured values of x. Highly asym- metric molecules with strong charge-transfer bands appear to yield large |x @ if the crystal structure is also highly asymmetric. 2.8 MILLER’S COEFFICIENT Miller defined a coefficient?® HH @5 = 0 + @,) xP Coes) xPKeo, J EPC oon) (2:53) Aig = and found empirically that 4, has only weak dispersion and is almost a constant for a wide range of crystals, This is known as Milter’s rule, It suggests that high refractory materials should have large norilinear susceptibilities. The weak dispersion of A,,, can be seen from either the bond-charge or the38 Nonlinear Optical Susceptibilities charge-transfer model. Equations (2.51) and (2.52} show that for w > 0, 4, = constant independent of frequencies. The constant is, however, proportional to the heteropolar gap C, and does change, although only mildly, from crystal to crystal. That the measured A, ,, is indeed proportional to C for a large number of semiconductors has been demonstrated by Levine.’ For a crystal with several different types of bonds, a weighted average C must be used. The values of 4,,, for most nonlinear crystals are around few times 10~* esu. 2.9 CONVENTIONS ON NONLINEAR SUSCEPTIBILITIES The definitions of nonlinear susceptibilities in the literature vary and have caused some confusion. This section clarifies the conventions used in this book. The definition of nonlinear susceptibilities Is governed by the following rela- tion between the nonlinear polarization P' and the electric fields E,: PO (a) = x (w = a + wy + +++ + @,) BE (o; JE2(#2}- ++ E, (oy) (2.54) with E, and P‘ expressed as complex quantities: E, = &exp(ik, +r — at) . . (2.55) PO(o) = PMexp(ik > — iw). assuming w, and w are both nonzero. Many authors have written the ampli- tudes of E, and P“) in somewhat different forms with E, = }Sfexp(ik, + — ia,t) 2.56 PO( wo} = 4 P"Mexp(ik «1 — dwt) (2.56) and defined a nonlinear coefficient 4") to connect the amplitudes Pe) md: Bg (2.57) or Ptr) = (2)" Ta"): EE, «+ E,. (2.58) Comparison of (2.54) and (2.58) gives tm) = (2p tty 09 (2.59)Conyentions on Nonlinear Susceptibilities 39 and in particular @ = 3. xB Equation (2.59), however, needs modification when there are de fields present. For w, = 0, the corresponding de field E, should be related to &, and &/ by E; = 2¢, = €/. Then, if 5 of the a fields, namely, E,,...,E,, are de, we have, following (2.54) and (2.57) as definitions for x and d, PO = OBL EA) Ban Seaplilkaay to, ~ (aay + oo + 0, 2.6 a MOE, Bey Sexplilkya tot k)er 280) ~i(oy + + 0,)¢], and hence GD = QE Ym, (2.61) More explicitly, (2.54) takes the form PP)(@ = @, + + +1 + ,) = Laden (@ = wy + wy + + + 0) B,C) E, (402) +> Bi, (,)- Fay by eee (2.62) Our convention is that the term xf"), )(w = a, + a + ---+ w,) E,,(@)E),(@2) +++ £,,{w,) can be written with the fields arranged in any order as long as the subindices of x‘" are arranged in the same order, but no additional contribution to P{”) should arise from permutation of the fields in (2.62). The conventional notation demands that the field arrangement should always follow the ordering of the frequencies in the argument of x"). This leads to the question of what happens if two or more fields involved have the same frequency. In our convention, permutation of the fields with the same frequency should yield no additional contribution to P<". For example, we have for second-harmonic generation, PPRe) = x9), (20 = w+ w)E,(w) E(w) # x, E(w) E,(w) + x2, E,(w) E(w). may Ez (2.63) In the convention using the d¢ coefficients, however, all terms derived from permutation of the fields with the same frequency must be included in the expression of the nonlinear polarization. For example, LPP (de) = dB Qu) E(w) E(a) + d®, sey = 242), (20) B,(w) E(w). (20 )£(w)E,(w) (2.64)40 Nonlinear Optical Susceptibilities Tn comparison with $PP (ws = a + 2) = ECs = oy + ey JE (JE (a2) (2.65) we notice that since the nonlinear response of the medium is not expected to have a sudden change as w, approaches u,, the coefficient dQ,(a, = w, + &,) should change smoothly to 2d@).(w, = 2c). The result that [d(2(w, = + lew, = 24 (20) with j + & has caused a great deal of confusion. A similar situation occurs when one or more fields have their frequencies ap- proach zero, as we discussed earlier. Our convention here avoids such diffi- culty: x9A(@, = @, + @,) changes continuously to x{7j.(w; = 20,) as a, ap- proaches «., or to xFCa, = 0 + w,) a3 w, approaches zero. The continuous variation of xf, with frequencies can be explicitly seen in the microscopic expression of x43 in (2.17). Another convention proposed by Maker and Terhune’ and often used for third-order nonlinearity is to indicate explicitly the number of terms one can obtain by permutation of different field components in the expression of the nonlinear polarization. For example, we write PA(o = a, + @, + 05) = EL Du Clow = wr + wy + oy) EC JE, (w E() Aad (2.66) where D,,, is the degeneracy factor for the particular terms. If £,(@,) # E, (#2) + E,(w,), then D,, = 6, indicating that six terms can be obtained by permut- ing the three fields. For E,(w,) = E, (2) # E,(w;), we have Dy, = 3, and for E(w,) = E,(0,) = E,(w,), we have D,,= 1, This convention also has the difficulty that the nonlinear coefficients C{),(w = a, + w, + 5) vary discon- tinuously as the frequencies become degenerate. : Further discussions of nonlinear optical susceptibilities appear in later chapters in connection with the specific nonlinear optical problems discussed. REFERENCES 1 J. A. Armstrong, N. Bloembergen, J. Ducuing, and P. §, Pershan, Pays. Rev. 127, 1918 (1962). 2 N. Bloembergen and Y. R. Shen, Piys. Rev. 133, A37 (1964). 3_N. Bloembergen, Nontinear Optics (Benjamin, New York, 1965} 4 CP. Slichter, Principles of Magnetic Resonance, 2ad ed. (Springer-Verlag, Berlin, 1978), Chapter 5. 5 N. Bloembergen, H. Lotem, and R. T. Lunch, Indian J. Pure Appt. Phys. 16, 151 (1978). 6 T.K Yee and T. K. Gustafson, Pays. Rev. Al8, 1597 (1978). 7 See, for example, C. Kittel, Introduction to Solid State Physics, Sth ed. (Wiley, New York, 1976), p. 406 8 D, Bedeaux and N. Bloembergen, Physica (Amsterdam) 69, 67 (1973).16 2 Be Bk BR Bibliography 41 YY. R. Shen, Phys. Rev. 167, 818 (1968). D. A. Kleinman, Phys. Rev. 126, 1977 (1962). See, for example, J. F. Nye, Physical Properties of Crystats (Oxford University Press. London, 1957), P. N. Butcher, Nonlinear Optical Phenomena (Ohio State University Press, Columbus, 1965), pp. 43-50. F.N. H. Robinson, Bell Syst. Tech. J. 46, 913 (1967); J. Phys. C1, 286 (1968); S. S. Tha and N. Bloembergen, Piys. Rev. 17, 891 (1968); C. Flywanis and J. Ducuing, Phys. Rev 178, 1248 (1965). K. G. Denbigh, Frans. Faraday Soc. 36, 936 (1940). See, for example, J. C. Phillips, Covalent Bonding in Crystals, Molecules, and Polymers (University of Chicago Press, Chicago, 1969); Bonds and Bands it Semiconductors (Academic Press, New York, 1973), The relation is known as the Thomas-Reiche-Kubn sum rule in solid state physics, See, for example, J. Ziman, Principles of the Theory of Solids (Cambridge University Press, Cambridge, 1965), p. 224. D. R. Penn, Phys. Rev. 128, 2093 (1962). See, for example, C. A. Coulson, Valences (Oxford University Press, London, 1961). B. F. Levine, Phys. Rev. Lert. 12, 787 (1969); Phys. Rev. BT, 2591 (2973), 2600 (1973). C.L. Tang and C. Flytanis, Phys. Reo. BA, 2520 (1971): C. L. Tang, IEEE J. Quant. Electron. QE-, 755 (1973); F. Scholl and C. L. Tang, Phys. Rev. BB, 4607 (1973) 1. P, Walter and M, L, Coben, Phys. Rev. Lett. 26,17 (1971). S. Louie and M. L. Cohen, persona] communication. Values of the various quantities are obtained from J. C. Phillips and J. A. Van Vechten, Phys Rev. 183, 709 (1969). C. Y. Fond and ¥. R. Shen, Phys. Rev. B12, 2325 (1975), See, for example, J. L. Oudar and J. Zyss, Phys. Reo. A26, 2106 (1982); J. Zyss and 7, L, Oudar, Phys. Rev. A26, 2028 (1982); C. C. Teng and A. F, Garito, Phys. Rev. Lett. $0, 350 (1983); and references therein. R.C. Miller, Appl. Phys, Lect. 5,17 (1964). P. D. Maker and R. W. Terbune, Phys. Rev. 137, A801 (1965), BIBLIOGRAPHY Bloembergen, N., Nonlinear Optics (Benjamin, New York, 1965). Butcher, P. N., Nontinear Optical Phenomenon (Ohio State University Press, Columbus, 1965). Ducuing, J.. and C. Flytzanis, in F. Abelés, ed. Optical Properties of Solids (North-Holland Publishing Co,, Amsterdam, 1972), p. 859. Flytzanis, C., in H. Rabin and C, L. Tang, eds., Quantum Electronics (Academic Press, New York, 1975). Shen, Y. R., in N. Bloembergen, ed., Nonlinear Spectroscopy, Proceedings of the International School of Physics, Envico Fermi, Course LXIV (North-Holland Publishing Co, Amsterdam, 1977), p. 170.3 General Description of Wave Propagation in Nonlinear Media Waves can interact through nonlinear polarization in a medium. Propagation of waves in the presence of wave interaction leads to various nonlinear optical phenomena. The quantitative description of a lower-order nonlinear optical effect usually starts with a set of coupled wave equations with the nonlinear susceptibilities acting as the coupling coefficients. This coupled-wave approach can also be generalized to include waves other than electromagnetic. This chapter is devoted to a general discussion of coupled electromagnetic waves in a medium and the solution of the coupled wave equations under certain approximations. Applications of the analysis to specific nonlinear optical phenomena appear in later chapters. 3.1 COUPLED WAVES IN A NONLINEAR MEDIUM The wave equation that governs optical wave propagation in a medium is 1 a 4a [» x(¥ X) 2S leo 2 pn ty, (3.1) which follows directly from the Maawell equations (15). Wave interaction gives rise to the nonlinear terms in P. We assume that both E(r, ¢) and P(r, ) 4a‘Coupled Waves in a Nonlinear Medium a can be decomposed into a set of infinite plane waves: E(t. #) = LE (k;, @) i = Letter od, t PO, 2) = PO(E, 1) + P(r, 1), POG, 2) = EPMA, 0) f = Ex Cor) Elk 2), 3.2) and P= YL PMs) a>2 = TPM (ky a) = TL PMeerr tant where é, is taken as essentially independent of time. With e(a,)= 1 + 4x ™(e,), (3-1) becomes wt a v x(v x) ~ ce-fE(k, 0) = PM kay 0). (3.3) Cc Pe Suppose PX¢(k,,, 0) = P’(k,,, #) is. a nonlinear polarization from the prod- uct of E,(k,,@,) --- £,(k,, @,). Then, for the » fields E,(k;, «,), there should be n corresponding wave equations similar to (3.3). Together with (3.3), they form a set of (2 + 1) coupled wave equations. Note that while ,, should be equal to w in P™(k,,,,) because of photon energy conservation in the steady-state case, k,, need not be exactly equal to k since wave momentum conservation is not strictly required in a finite medium. Equation (3.3) clearly indicates that the various waves E,(Kk,, o;) are nonlinearly coupled through the nonlinear polarization PN, and their propagations in the medium will conse- quently be very different from the linear case where Pp" = 0. Through nonlinear coupling, energy can now be transferred back and forth between waves, and the larger PN" is, the more pronounced the effect should be, The coupled wave approach was first used in the description of microwave para- metric amplification and later adopted by Armstrong et al.* for describing wave interaction in nonlinear optics. The simplest case of optical wave interaction deals with second-order nonlinear optical effects. We use it here as an example to illustrate the coupled“4 Genera! Description of Wave Propagation in Nonlinear Media wave formalism. Consider three waves E{k,, 1), E(k, 2), and E(k, w = 0, + «,) interacting in a medium by second-order nonlinear polarization. Then, the coupled wave equations from (3.3) are 4nud PPO%(o,) wt vx(v x)~ se E(k, 0)) = 4a? = Mt (ea, = -#, +0): EX(k,, «, JE (k, w), ¢ we Anat lv x(B X)- ees, @,) = a 2 P(s,) = EE (4, = 0 ~ 6) Blk, YET 1s 64 and 4ae* [v x(¥ x) Se fete, 0) PA(w) 2 = A xo = wy + w,) 2 E(k, JE(K2, «). As seen here, the nonlinear susceptibilities appear explicitly as the coupling coefficients. They determine the rate of energy transfer among the three waves. in the case of a dissipationless medium, the permutation relation x{ O(a =) + wy) = XP(@ = — wy + @) = XL}j(w, = @ — «,) exists (sce Section 2.5). This is actually a necessary condition | for the coupled wave equations to satisfy the requirement that the total energy in the three waves is a constant, as we shall see in Section 3.2. The photon energy and momentum conservations in the present case are w = 0, + #, and k=k, + k,, respectively. For most effective energy transfer among the waves, one naturally expects that both photon energy and momentum conservations should be satisfied in the wave interaction. Therefore, even though k = k, + k, is not required, as mentioned earlier, satisfaction of the relation is preferred for the maximization of the wave coupling. This photon momentum matching condition is known in nonlinear optics as the phase matching condition, and it is one of the most important considerations in many nonlinear optical processes. Detailed soku- tions of (3.4) appear in Jater chapters. 3.2. FIELD ENERGY IN A NONLINEAR MEDIUM. The Maxwell equations (1.5) lead to the following familiar energy relation for the fields: ev EXB)= eh let + BY - pi (3.5)Field Energy in a Nonlinear Medium 45 With E x B being the Poynting vector, it shows that the rate of electromag- netic energy flowing out of a unit volume is equal to the reduction rate of the stored electromagnetic energy density. If the dispersion of the medium can be neglected, then the polarization P can be written as PCr, 4) = x®-E(r, 1) + xO: EE+ --- (3.6) and (3.5) reduces to the form € a read -(E X B) = — x Ul) (3.7) with O(c, 1) = Ge (E? + BY) + FROME + EEO EE + (3.8) being the instantaneous electromagnetic energy density. This is clearly not valid in a medium with dispersion, since the susceptibilities are defined only in terms of the Fourier components of fields and polarizations. In reality, it is more meaningful to consider the time-averaged energy relation. Let us il- lustrate the problem by first assuming a linear medium. We begin by assuming a quasi-monochromatic field E(t, 1) = &(1) ef@r-e0 + Be (pe Meron) (3.9) where &(1) is a slowly varying amplitude. Expressing é(1) as a Fourier integral, we have E(t, 4) = fad (w+ hero + cc, (3.10) Then the linear polarization takes the form Pr, 2) = fanx™(o +a) €(w + ne™r-oO- it + Oe (3.11) This leads to Zpo, = fan(-)ot a) xo) + ee, + | (+n) xelerroint to, (3.12) a) = [-iex™(a)-€t0 Ao) ]ewrmn tee,6 General Description of Wave Propagation in Nonlinear Media With x{)(w) = x(— w)*, the time average of (3.5) yields? (a9 xB) = -Zu -@ (3.13) where wy Elen) 2-6) + Bene}, Q= Frer(d 0" S(t}, G4) and ere tie = 1+ aay. As may be inferred from the preceding equations, {U) is the average field energy density stored in the linear medium and Q is the average power density dissipated into heat through absorption because 2” # 0, Equation (3.13) is therefore am energy conservation relation one should expect physically. The above calculation can be extended to the nonlinear case. Additional terms in the energy relation are expected from wave coupling. Consider, for example, three waves with #, = @, + w, and k; =k, + k, interacting in a nonlinear medium via x”. One finds, with the help of the permutation symmetry relation of x, that the average field energy density has an additional term (UP) = EEA) Mo = ey + 5): ERNE) ay @ dx (ws, ax?) +63(1) [o.22 + od wy 2 Ae) al: (3.15) XSPEC) + ce. obtained from aus _f op oP at ze a f° Note that only when |a,, x /de,,| « |x| can we write (UY = IEF ¥P(w, = — 4, + 0): &Sy + cc. (3.16) The last equation, however, is frequently used in the literature® to describe the interaction free energy density for the wave coupling. One would write the free energy density as F® = —) using (3.16) and derive the nonlinear polari- zation from P®(w,) = — }0F°/a8s. From 6?P(a,)/ddy dé, =Slowly Varying Amplitude Approximation 47 PPO wy) /8EF A6, = APO *(w,)/dE¢ IEF, the permutation symmetry rela- tion of x immediately results, Although this is a practice ong cam indeed follow, we must realize that it is actually the permutation symmetry of x that leads to the expression of (U/) in (3.15), Conversely, it is the existence of (U) that physically justifies the permutation symmetry of x). Near reso- nances, when dissipation in the medium becomes important, the permutation symmetry relation of x breaks down, and accordingly, (3.15) is no longer valid. More generally, in a nonabsorbing medium, the time-averaged field energy density should be* «U) = >» wm) G17) aml where (U“")) arises from nonlinear coupling of (# + 1) waves via the nth-order nonlinearity in the medium, and is given by (U) = WEB XM ay beet ty) AEG nti @ 3.18 A Botte | ata 4 6s (3.18) i=l i The time-averaged energy conservation relation takes the form € a (ae? “EXB) = -ZU) (3.19) 33 SLOWLY VARYING AMPLITUDE APPROXIMATION {n actually solving the coupled wave equations, several simplifying approxima- tions are often made Among them are the slowly varying amplitude approximation, the infinite plane-wave approximation, and constant pump intensity approximation. We discuss here only the slowly varying amplitude approximation and leave the others to later chapters. As mentioned earlier, wave coupling in a nonlinear medium results in energy transfer among waves. Therefore, the wave amplitudes are expected to change in propagation. We assume for illustration a plane wave propagating along 2: E(w, 2) = &(z)e@-99, Since the energy transfer among waves is usually significant only after the waves travel over a distance much longer than their wavelengths, we expect #8 (z) az? a€ 2z (3.27) it ere) == 7 forz Se(z) = Ep(0) + 12EE fpr (aren tte -engys ke? 4g and 2a? Sol) = Ep(l) +1 (zeit +802", (3.32) z The corresponding differential equations for ¢p and &, are 86, 2ae? ik oer _ Ame -i{kr~wr) a7! ‘a PNU(w, Ze and 965 _ _ ,2au? ke +0 a 7 ~ Ee PM (wu, z)e an (3.32) Comparing (3.24) with (3.32), we recognize that (3.24) can be obtained by neglecting ¢, in & (ox neglecting , if &, is propagating with wavevector —k).50 ‘General Description of Wave Propagation in Nonlinear Media 3.4 BOUNDARY CONDITIONS The usual boundary conditions for electromagnetic waves should be valid here; for example, the tangential components of E and B at a boundary surface must be continuous for each Fourier component. In general, the solution of the wave equation (3.3) for E(k, #,) driven by PX(k,,, a, = «/) & exp(ik,, +r — ia;t) has the form E(k, .,,, 0) = (Eye + Sette) emir (3.33) where the &y and é'p terms correspond to the homogeneous and. particular solutions, respectively. At the boundary, an incoming wave E,(k,;,k,,,, 0) splits into a reflected wave Ex(kjg,k,g,%/) and a transmitted wave E;(kj7,k,,r,@;). Let z = 0 be the boundary plane, and &-Z be the plane of incidence. Then it is casily seen that the continuity of the tangential compo- nents of E and B leads to the following relations:? 61,2 + Spt2 t Onn.x + Opn,x = Sur.x t Spr. and (Kir X Egg) + Kant X Spr) x + Bip X Sr) x + Km X Spr): 3.34) = (Kir X Sur), + Kar % Sr), G34 with two similar equations for the » components, and Katie 7 Bante Kenge = Kai = Bere = Bt: (3.35) The last equation relating the various tangential wavevectors is most inter- esting, It prescribes the directions of propagation for all waves (homogencous and particular) in the media when one of them is given. This is therefore equivalent to Snell's law in linear optics. 3.5 TIME-DEPENDENT WAVE PROPAGATION Propagating waves with time-varying amplitudes should of course obey the time-dependent wave equation in (3.1). Here again, the slowly varying ampli- tude approximation is usually valid. We expect that both the second-order time derivative and the second-order spatial derivative of the field amplitude can be neglected in the wave equation. This is illustrated in the following by assuming a quasi-monochromatic plane wave propagating along a symmetry axis, 2, of the medium. The wave equation takes the form a ae E(z,£) - 12 1 ne e? ar? D(z, t) = 2 rn t) (3.36)Time-Dependent Ware Propagation 51 with D(z, 1) = E(z,r) + 40P%z, 0), and Elz, 0 = &(2, Nexp(ikz — jot). Then, as shown in Section 3.3, the slowly varying amplitude approximation gives 2 get We Jeter a, (3.37) ae . Fablads= (i28 If E(z, ¢) is expressed in terms of the Fourier integral H(z, 1) = [8(o + gle orm" dn, then we have D(z,r) ~ fel + n&(o + gel ot ay dy L@ ~(w +a) - za qa Dl t) - {hae + nE(w + nei Horne ay It 1 de -S flo) + 2ume(er) + ong [slo ay) (338) xeon gy ow 1@ af-2 io i(kez—et) elo) F(z, 4) ike Rp ols oe where v, = (dk /dw)~! is the group velocity. Insertion of (3.37) and (3.38) in (3.36) with the approximation of ?P™t/@r? = —w® PN" yields? g+id = — 280? prs, py_ite-on (4 +7 x) le. - aa een, (3.39) In fact, as we have shown in the time-independent case in Section 3.3, the field amplitude @ in (3.39) actually corresponds to &; for the forward Propagating ware. For the backward propagating wave, the corresponding equation is 2 = oa PM(z, 1) eitketu0), (3.40) c Equations (3.39) and (3.40) should be used for short pulse propagation in a nonlinear medium. The time-derivative term in the equations is negligible only. if the amplitude variation is insignificant during the time T = We /c it takes for light to traverse the medium. We use (3.39) and (3.40) later in the discussion of nonlinear optical effects with ultrashort pulses,32 General Description of Wave Propagation in Nonlinear Media REFERENCES 1 See, for example, W. H. Louisell, Coupled Mode and Parametric Electronics (Wiley, New York, 1960). 2 1. A. Anmstrong, N. Bioembergen, J, Ducuing, and P, S, Pershan, Phys. Reo. 127, 1918 (1962). 3 L. Landau and E. M. Lifshite, Efectrodynamics in Continuous Media (Addison-Wesley, Reading, Mass, 1959), p. 253. 4 Y.R. Shen, Phys. Rev. 167, 818 (1968). 5 PS, Pershan, Phys. Rev. 130, 919 (1963), and many books and review articles on nonlinear optics. 6 See, for example, N. Bloembexgen, Nonlinear Optics (Benjamin, New York, 1565). N. Bloembergen and P.S. Pershan, Phys. Rev, 128, 606 (1962). 8 S.A. Akhmanoy, A. S. Chirkin, K. N, Drabovich, A. L Kovrigin, R. V. Khokhlov, and A. P. Sukhorukov, FEEE J. Quant. Electron. QE-4, 598 (1968). ~ BIBLIOGRAPHY Akhmanov, $. A., and R. V. Khokhlov, Problems of Nonlinear Optics (Gordon and Breach, New York, 1972). Bloembergen, N., Nonéinear Optics (Benjamin, New York, 1965). Ducuing, in R. Glauber, ed., Quantum Optics, Proceedings of the Internation School of Physics Enrico Fermi Course XLH (Academic Press, New York, 1969), p. 421.4 Electrooptical and Magnetooptical Effects Optical properties of a material can be modified by an applied electric or magnetic field. The refractive index changes as functions of the applied electric and magnetic fields are responsible for many electrooptical and magnetooptical effects. Although these effects were well known long before the advent of lasers, they can be regarded as nonlinear optical mixing effects in the limit where one of the field components is of zero or nearly zero frequency. This chapter therefore offers a brief discussion on these effects. 4.1 ELECTROOPTICAL EFFECTS In the presence of an applied de or low-frequency field Eg(& ~ 0), the optical dielectric constant e(@,E,) of a medium is a function of Ep. For sufficiently small Ep, e(«, Fp) can be expanded into a power series of Eq: e(w, Ey) = ea) + eC + )-Ey + ew + 22): EgEy + «>. (4.1) Since E + 47P = e-E, and P= x™-E+ x®:EE+ ---, we recognize that 2a + Q) = 42x (w + Q) and e(@ + 28) = 4x (w + 22). (4.2) Then, in a medium with no inversion symmetry, the electrooptical effect is dominated by the e® term linear in Ey. This is known as Pockel’s effect. The symmetry forms of the nonvanishing e® or x‘? for the 20 classes of crystals 3354 Electrooptical and Magnetooptical Effects are already given in Table 2.1 with, in addition, x23(w = + 0)= xfi(o = a+ 0). In a medium with or without inversion symmetry, the quadratic ficld-dependent term (4.1) always exists and is known as the de Kerr effect. The symmetry forms of e® or x) for some classes of crystals are given in Table 2.2 with, in addition, x®,(o =o +0 +0) =x9(w = © + 0+ 0) and xi) = xh The ficld-induced refractive indices give rise to linear birefringence or double refraction, Traditionally, the electrooptical effect is defined through the idex ellipsoid! 2 Zz 2 2 2 Sree et ax 4 2s =1 (43) Nex My Ra, Hy, ex Fixy with nj? = (e-*)}/? being the refractive index tensor. The power series expan- sion is carried out for all the coefficients #;;7(Eq) of the index ellipsoid I 1 wre (=) + Dri Boe + Do pinsEounEo + + (4.4) aij(Ey) My ly ke fal The coefficient 5, often is called the linear electrooptical tensor, and 7; 4; is the quadratic electrooptical tensor. The values of 7,,, for many crystals are tabulated in the literature,” Physically, clectrooptical effects result from both ionic or molecular move- ment and distortion of electronic cloud induced by the applied electric field. Even if the induced refractive index change is only around 107° (typical values of rj, are around 107° to 10-8 cm/v), a medium 1 em long can already impose on a visible beam a phase retardation of more than 7/2. Therefore, electrooptical effects have been widely used as optical modulators. 42 MAGNETOOPTICAL EFFECTS The optical dielectric tensor e of a medium is also a function of an applied dc magnetic field, Hy. It has the symmetry telation > &;;(Ho) = &:(- Ho). (4.5) Here, even in the absence of dissipation, ¢,, is a complex quantity but it has the property of being Hermitian: «,,(Ho) = #};(Hy) + icf (Hg) = ef (Ho). (4.6)“Magnetooptical Effects 55 Then, in a nondissipative medium, we have ef (Hy) = #;(Ho) = e(;(-Ho) 47 eH) = ~ef (Ho) = -#4(-Ha). “ Thus the real part of the tensor is symmetric and is an even function of Hg, and the smaginary part is antisymmetric and odd in Ho. The dependence of e7; on Ho leads to circular birefringence or the Faraday effect, while the depen- dence of e, on Ho Jeads to linear birefringence or the Cotton—-Mouton effect. This can be illustrated with a medium of uniaxial symmetry, having Ho parallel to the axis. In this case, the only nonvanishing elements of & are £,, = ey and e;, even in Hg, and ex, = — ef, odd in Ho- Diagonalization of & in the coordinate system with orthogonal unit vectors ¢ 4 = (% + ip)/ ¥2 and? yields the three diagonal elements €, and e,,, where e,=e,, + ey, are the susceptibilities for right and left circulasly polarized waves, respectively. Since ef, © e,,, the wavevectors of the two circularly polarized waves can be written an ex? wile, — {Over 1s Jey, (48) c¢ € € . x and the circular birefringence in a medium of length / is (4.9) A linearly polarized beam propagating along 2 will have its polarization rotated by an angle go he 2 (4.10} which is known as the Faraday rotation. On the other hand, since 2), (Hg) — ef,,(0) is generally different from e,,(H o) — €7-(0), the linear birefringence in the %-2 plane is also altered by the presence of Ho, known as the Cotton—Mouton effect. For sufficiently weak H,, the power series expansion of e(Hy) yields 2'(a,Hg) = (wo) + 2 O(w@ + 22)*HoHy + °° and (aH) =e ( + B)-Hy t+ o'- {4.11)56 Electrooptical and Magnetooptical Effects Again, e®/4a, e®/47, and so on, can be regarded as nonlinear susceptibili- tics, although they now arise from the magnetic contribution. Analogous to the electrooptical effects, the magnetooptical effects can also be used for optical modulation. The Faraday effect or circular birefringence causes the linear polarization of a beam traversing the medium to rotate. In the low-ficld limit, the rotation is proportional te the applied magnetic field. Again, the induced change in the dielectric constant or refractive index is usually small (~ 10-°/gauss for glass doped with a few percent of rare earth ions}, but the rotation resulting form the relative phase shift between the two circular polarizations can be few tens of degrees ina 1-cm-long medium with a field of several thousand gausses. The Cotton—Mouton effect, however, is much weaker and has limited applications. REFERENCES 1 Sce, for example, A. Yariv, Quantum Electronics, 2nd ed. (Wiley, New York, 1975), p. 327. 2 RI. Pressley, ed, CRC Handbook of Lasers, {Chemical Rubber Co., Cleveland, Ohio, 1971), p. at 4 See, for example, J, van den Handel, Encyclopedia of Physics, vol. 15, 8. Fligge, ed. (Springer- Verlag, Berlin, 1956), p. 15. BIBLIOGRAPHY Landay, L. D., and E. M. Lifshitz, Electrodynamics in Continuous Media (Pergamon Press, Oxford, 1960). Nye, J. E., Physical Properties of Solids (Oxford University Press, London, 1964). Pressley, R. J., ed., Handbook of Lasers (Chemical Rubber Co, Cleveland, Ohio, 1971). Wemple, §. H,, in F. T. Arecchi and E. O. Schulz-Dubois, eds. Laser Handbook (North-Holland Publishing Co,, Amsterdam, 1972), p. 975.5 Optical Rectification and Optical Field-Induced Magnetization Modulation and demodulation are commonly known processes at radio wave and microwave frequencies. They form the basis of telecommunications. It is natural to believe that these processes should also exist in the optical region. Light modulation by electrooptical and magnetooptical effects has already been discussed in Chapter 4, In this chapter, optical rectification producing de electric polarization and magnetization is considered. 5.1 OPTICAL RECTIFICATION In the literature, optical rectification, which was among the first nonlinear optical effects discovered, usually refers to the generation of a de electric polarization by an intense optical beam in a nonlinear medium. The effect can be seen directly from the nonlinear polarization PO) = 4@(0 = w - w) : E(w)E*(w) (5.1) with EX) = @exp(#k+r — ior). The nonlinear susceptibility x(0 = w — @) here governs the magnitude of the effect. In a nonabsorbing medium, the permutation relation of x7 relates x°(0 = — w) to the electrooptical coefficients x9, 0 = 0 - @) = xB (w= o + 0) = xf (w= 0+) Oe = nk (5.2) 5758 Optical Rectificatiow and Optical Fiekd-Induced Magnetization in a principal-axis system. Thus, from the electrooptical coefficient rj» the + polarization generated in optical rectification can be predicted. In actual experiments, the induced de field or voltage, instead of P®), is measured, but the two are linearly related. Bass et al.’ and Ward? performed optical rectifica- tion experiments and measured x90 =o ~ a) fora number of crystals. The experimental arrangement can be simple. A slab of crystal is oriented with its # axis perpendicular to the two parallel faces of the slab. The faces are coated with silver to form a set of condenser plates. An intense light beam is then directed through the crystal in a direction perpendicular to / to generate PPX0) according to (5.1), and the induced de voltage across the condenser plates is measured, Let the de dielectric constant of the crystal along i be e, and assume that the beam intensity can be approximated as uniform over a rectangular cross section s X ¢ in the crystal, as shown in Fig. 5.1. Then, following the infinite plane approximation for condensers, the equations governing the de fields are Ey = E,(d—1) + By 5.3 eEg = En + 4aP? 63) and since there is no net charge on the plates -E,(w —s) = E,5. (5.4) The solution of these equations yields a de voltage across the plates as ve- ae) Ew (5.5) ~ -44( 5) Exe wero) In the experiments to find xO =a@—w), both the voltage V {in the See eee HE Sr od Fig. 5.1 Experimental geometry for measurement of optical rectification,Effective Free Energy Density 59 mv/MW range) and the laser intensity must be accurately measured. ‘The results of Bass et al. and Ward? on optical rectification show that the identity of (5.2) indeed holds within experimental error. 5.2. EFFECTIVE FREE ENERGY DENSITY The time-averaged field energy density (UV) in a nonlinear medium was derived in Section 3.2, where we saw that we can obtain the polarization P(w)in the medium from the derivative of the effective free energy density F- FO = —((U) - (U)) a(F- F) Pla) Eee) (5.6) if the dispersions of susceptibilities are neglected. Thus the effective free energy density corresponding to a not very intense quasi-monochromatic wave in 4 nonabsorbing medium subject to a dc electric field Ey is* F[B(e),¥0] = F(Es) ~ E [xje(Eo)£f()E(o) +--+]. (5.78) If x,(Eo) can also be expanded into power series of Eo, it becomes F= FOE) + FO + FO+ -- FO = — ExnOFro) EC), (5.70) PO= — FY xP Sy Eo)" Ee). hak The free energy density here governs both the electrooptical effect and the optical rectification. From (5.6), the polarization induced in optical rectifica- tion is given by Pe are 2) E* 0) = ~ Fem = Ext ero) (0) (58) fod as is expected. The same {9} in (5.7) and (5.8) is clearly responsible for the linear electrooptical effect. The above description can be extended to the magnetic case.** For a not too intense light beam propagating in a nonabsorbing medium in the presence of a magnetic field, the effective free energy density can be written as a power "Note that Ey = lime ,o}(@) + Ef()L.6 Optical Rectification and Optical Field-Induced Magnetization series of the optical field F= Fy(Hy) — Dxi(Hod Btw) F(a) + -+*- (5.9) ad In analogy to the electric case, [x, (Hp) — xi;(0)] here governs the magneto- optical effect. Then one also expects from M(0) = —dF/dH, that there should be a dc magnetization induced in the medium by the incoming light. Indeed, this has been observed and is known as the inverse magnetooptical effect. For illustration, we assume a medium of uniaxial symmetry with light propagating along the axis, say, the Z-axis. It is well known that if the dc magnetic field is also along 2, the two circular polarizations are the eigenmodes of propagation. The effective free energy density can therefore be written as F~ F(Ho) — x4 (Ho)IE(@)P? ~ x (Ho 1E_(o)? + (5-40) where X= Xxx t Xay ANd X_= Xow — (Xzy ATE respectively, the linear sus- ceptibilities for the right and left circularly polarized waves. Equation (5.10) can be rearranged inio F = FH) — 1x4 (He) ~ x-( Ho) (IEP — 1E-P) 5.1 —3[x. (Ho) + x-CHo) (IEP + 1EP} + em with [x,CHo) — x_(Ho)l and [x.(4o) + x-(Hp)] being odd and even func- tions of Hy, respectively. The Faraday effect is now proportional to OF OF - + =4x.(4y) ~ x4. art pp RAH =x) (5.12) _ £4(Hq) — ©-( Ho) 4a as discussed in Section 4.2. We can also show Xx Ho) = Xjy (th (Ho wy fo) (5.13) Xca( Ho) ~ Xeu(0} = Hoes (Ha) + X-(Ho) — x4 (0) ~ x-)]- This magnetic field-induced susceptibility change is connected to the Cotton-Mouton effect. On the other hand, the optical field—induced taagneti- zation can also be derived from (5.11). With the optical field on, the inducedInverse Faraday and Cotton-Mouton Effects 61 magnetization change along Z is given by a AM= 7a, FA) = AMp+ AMeu, (5.14) ait@y- Pig RP AM, = 5 99 > XIE? IE), and lia MMe = 5 TH x- MIE? +E). The term AM, even in H) comes from the term responsible for the Faraday effect in F, and AMoy, odd in Hy from the Cotton—Mouton term in F. The phenomena are therefore called the inverse Faraday effect and the inverse Cotton-Mouton effect, respectively. As seen from (5.14), even if H, = 0, the inverse Faraday effect is nonvanishing as long as |E,|*|£_|, and is a maximum for circular polarization. The light beam with |£,|? # |£_|? here plays the role of a dc magnetic field and breaks the time-reversal symmetry of the medium. The inverse Cotton-Mouton effect can, however, exist even with linearly polarized light but vanishes for Hy = 0 since 0(x,+ x_)/0H, is odd in Hp. The effective free energy density then allows us to predict the magnitudes of the inverse Faraday and Cotton—Mouton effects from the measured Faraday rotation and Cotton-Mouton effect in the medium. Physically, the Faraday and Cotton—Mouton effects originate from circular and linear dichroism induced by the de magnetic field, but how do we describe the inverse effects? This is the subject of the following section. 33 INVERSE FARADAY AND COTTON-MOUTON EFFECTS Microscopically, the light-induced de magnetization arises because the optical field shifts the different magnetic states of the ground manifold differently (known as optical Stark shifts), and mixes into these ground states different amounts of excited states. Let the interaction Hamiltonian be Hig = 0 +E #, = —e[r, E_(w) + r_E,(o)] (5-15) From the time-dependent perturbation calculation, we find the perturbed62 Optical Rectification and Optical Field-Induced Magnetization eigenstate as ln) = [to + Ale) aie neta 5.16 Any = Zwrlatecae - oe] ™ and the optical Stark shift for |1) as ae.= 5 [erst - eat | 6.17) nag l ACOH Wye) MO + y4) where hu,,, = E,, — E,. To demonstrate the inverse magnetooptical effect, we assume here a simple paramagnetic ion with only two pairs of states as shown in Fig. 5.2. The ground | + m) state is connected to the excited | ~ m2’) state only by the matrix element ( — m'|r_|m) and the | — ) state connected to the | +m’) state by (+ m'jr,|—m), with r= (x £ iy)/ v2. In an applied magnetic field along 2, the Zeeman splittings for the two pairs are respectively 2gBmH, and 2g‘Bm'Hp, where B is the Bohr magneton. The energy separation between the pairs of states is hw >» KT at Hy = 0. The de magnetization of a system of N ions per unit volume along 2 is given by —NghXJ,) —NgBlGnth mye, + (= mle] = 0 -m- M (5.18) Here J, is the angular momentum operator and p,,, afe the thermal popula- 2g'm'BHy teem | de. m\ Fig, 5.2. Energy level diagram of an ideal <-—m| paramagnetic system with only two paris of states connected by circularly polarized opti- 2 gmBHy cal fields.Inverse Faraday and Cotton-Monton Effects 6 tions in | + ) described by the Boltzmann distribution en E andkT Pam eT EW /AT 4 gE em /kT 5.19) with E,,, = tg8mH, + AE ,,,. Through its perturbation on E ,,, and |), the optical field induces a de magnetization 4M = M( 24) -— M(o#, = 0) =AM?P+AM?, (5.20) AM? = —NeBra[( Pq — Pm) —(08, - 0% m)]) and AM? = —NgB[(A(m)J,(Almy) o& + (AC — ml) J.(Al — m)) 0 af where £°, y= (Py mse, <0» and 4M? and AM? come from induced changes in the populations and matrix elements, respectively. From (5.15) to (5.20) with JAE , pl AT and |(n'[r,|— my]? =| — m’|r_|m)|?, we readily find _ 7 2Nghm(AE,, — AE _ 5) PnP m - kT 2NgBmp Pm, 4 = — SE ahem mer el — m)/? | AMP h(w— @) #7( 00 — tay)? —( gm + g'm') HG hw + ) ¥(w + wp)’ (gm + g’m') RHE (5.21) x(IE.P -[£_#) + (gm + g'm’) BH, f(a — wy)? — (gm + g’m’) RHE —_(on + geile —_le rs enh FC + ea) —(gm + g’m’) BHDo Optical Rectification and Optical Field-Induced Magnetization and AM? = -4NgBm'Km'ler.|— mI? +{( - . [Aw — @o} + (gen + g’m’) BHy| oy + [A(w + 9) —(Cgm + gm’) BH] =m [Ao ~ «2g) — (gre + gn") BH)” fin EP =|E_P "Tae + ey) + (gmt waa rl ‘sz (Nea [A(o — 14g) + (gm + gm’) BH)” pe [A(e + 9) — (gm + g'm’) BH] 0 ¢——— [a(a — 0) —(am + g'm")BHp]” om |r + e) a [Ale + oy) + (gen + g’m’)BHo]” Since AM? arises from the induced population change in the ground states due to optical Stark shifts, it vanishes in a diamagnetic system which has only a singlet ground state that is populated at ordinary temperatures. It is therefore designated as the paramagnetic part of AM. Because of the finite relaxation time for the population distribution to reach new equilibrium, AM P cannot respond instantancously to a short incoming light pulse. In fact, from the time variation of AM?, one should be able to deduce the 7, relaxation time of the ground states. The AM’ term arising from the wavefunction mixing by the optical field exists even in a diamagnetic system and is designated as the diamagnetic part of AM, It responds almost instantaneously to the incoming light, The paramagnetic part is proportional to 1/kT for |AE . _| kT, and the diamagnetic part is essentially independent of temperature. This is similar to the temperature behavior of ordinary paramagnetism and diamagnetism.’ Both 4M? in (5.21) and AM? in (5.22) have been explicitly decomposed into aInverse Faraday and Cotton-Mouton Effects 6s part proportional to (JE,{? ~|£_|2) and a part proportional to ({6 |? + |E_}?). The former corresponds to the inverse Faraday effect, and the latter to the inverse Cotton-Mouton effect. For light frequency far away from reso- nance, the inverse Cotton—Mouton effect is much smaller than the inverse Faraday effect. As seen from (5.21) and (5.22), the ratio of AMey, to AMr is about (z’m’ + gm)BH,/h(« — w,)| for the paramagnetic part, and is about Kg’m’ + gmt) BH, /Aw — @9)| or (69, — p° m)/(p%, + p°,,), whichever is larger, for the diamagnetic part. It may become comparable to 1 when A(@ — w) approaches the Zeeman splitting energy. However, close to resonance, the induced dc magnetization due to optical pumping often becomes dominant.’ In actual experiments, the inverse Cotton—-Mouton effect is distinguishable from the inverse Faraday effect by the fact that with a reversal of the magnetic field Hy, AMcy4 changes sign but A.M, does not. Finally, we realize that (5.21) and (5.22) can be derived from (5.14) if the microscopic expressions of x, and x_ for the system in Fig. 5.2, following (2.17) for x;,, are used. This is left as an exercise to the readers. The above calculation for AM can of course be generalized to a paramagnetic system with N ground states. In dense media, a local field correction factor should also be included. The experimental scheme for observing the light-induced de magnetization is seen in Fig, 5.3. The light pulse induces a pulsed AM(z) in the sample. The time-derivative d(AM)/di then induces a voltage across the terminals of a pick-up coil around the sample. As an example, consider the case of CaF, :3%Eu2*. The Faraday rotation at 4 = 7000 A is 2 x 10~* rad/cm-Oc at 4.2 K. From (4.9) and (4.10), we obtain [0(x4— x)/9Ho]y,-0 = 1.8 x 107" esu/Oe. Then (5.14) predicts that with a circularly polarized ruby laser beam of 10 MW/cnt in the sample, the induced de magnetization is AM = AM, = 7 x 10-* erg/Oc-cnr’. This is equivalent to that induced by a de field of about 0.01 Oe. If a Q-switched laser pulse with a 10-MW peak intensity and a Polarizer 4 Beam splitter Laser pulse Oscilloscope Fig. 5.3 Experimental arrangement for measurement of inverse magactooptical effects.66 Optical Rectification and Optical Field-Induced Magnetization a rise time of 2 X 10~® sec is used, and AM(z1) is assumed to follow instanta- neously the intensity variation of the laser pulse, then the voltage induced across the terminals of a 30-turn pick-up coil will be 1.3 my. This agrees with the experimental observation of van der Ziel et al.,5 who demonstrated that the inverse relationship between the Faraday effect and the inverse Faraday effect indeed holds for many paramagnetic and diamagnetic substances. 54 INDUCED MAGNETIZATION BY RESONANT EXCITATION The de magnetization can of course also be induced by light through direct optical pumping, and is generally much stronger than the inverse magnetoopti- cal effect discussed in Section 5.3. Optical pumping by circularly polarized light, for example, alters the population distribution in the magnetic sublevels of both the ground and the excited states, A net angular momentum and hence a magnetization result, Theoretically, rate equations can be used to calculate the population redistribution and hence the magnetization induced by the resonant optical excitation if transition probabilities and relaxation rates between levels are known. Optical pumping in gases and solids has long been a subject of extensive investigation. Polarized Auorescence is often a means for detection of the induced orientation of the angular momentum in the medium. With the setup in Fig. 5.3, however, it can also be studied by measuring the dc magnetization generated in the medium by the laser pulse.’ This may be useful in some eases for studying relaxation between magnetic sublevels in condensed matter. REFERENCES M. Bass, P. A. Franken, J. F, Ward, and G. Weinrich, Phys. Rev. Lett, 9, 446 (1962). 1, F, Ward, Phys. Rev, 143, 569 (1966). YR. Shen and N. Bloembergen, Bull. Aim. Phys. Sac. 9, 292 (1964). P.S, Pershan, J. P. van der Ziel, and L. D, Malmstrom, Phys. Rev. 143, 574 (1966). J.P, van der Ziel, P. S. Pershan, and L. D, Malmstrom, Phys. Rev. Lett. 15, 190 (1965) J. HL van Weck and M. H. Hebb, Pays. Rev. 46, 17 (1934). J. F, Hokrichter, R. M. MacFarlane, and A. L. Schawlow, Phys. Rev. Lett. 26, 652 (1972) MAW Rene6 Sum-Frequency Generation ‘Wave interaction in a nonlinear medium leads to wave mixing, The result is the generation of waves at sum and difference frequencies. Sum-frequency genera- tion is one of the first three nonlinear optical effects discovered in the early days.' With the recent advances in tunable lasers, it has become one of the most useful nonlinear optical effects in extending the tunable range to shorter wavelengths. This chapter deals mainly with the basic principle of sum- frequency generation. 6.1 PHYSICAL DESCRIPTION Bass ct al.’ in 1962 first observed optical sum-frequency generation in a crystal of triglycine sulfate. In their experiment, two ruby lasers, with their operating wavelengths 10 A apart, were used to provide the input beams. The output, analyzed by a spectrograph, exhibited three lines around 3470 A, two side lines arising from second harmonic generation and the middle one from sum- frequency generation by the two laser beams. The physical interpretation of sum-frequency generation is straightforward. The laser beams at w, and w, interact in a nonlinear crystal and generate a nonlinear polarization P°(w, = w, + «), The latter being a collection of oscillating dipoles acts as a source of radiation at #, = w, + @». In general, the radiation could appear in all directions; the radiation pattern depends on the phase-correlated spatial distribution of P®(.,), With appropriate arrange- ment, however, the radiation pattern can be strongly peaked in a certain direction. This can be determined by phase matching conditions. As discussed in Section 3.1, for effective energy transfer from the pump waves at @, and a, to the generated waves at «,, in the sum-frequency generation (Fig. 6.1), both energy and momentum conservation must be satisfied. The energy conserva- tion requires #, = w, + #, While the momentum conservation requires ky = k, + k,. The latter indicates that the sum-frequency radiation is most effec- er8 Sum-Frequency Generation ky) (esi (eos, ks) (09, kat Fig. 6.1 Schematic description of sum-frequency generation. tively generated in the so-called phase-matching direction defined by ky = k, + k,.” If the wave interaction length fis finite, momentum conservation needs to be satisfied only to within the uncertainty range of 1//. The radiation pattern should therefore be a finite phase-matching peak with an angular width corresponding to Ak ~ 1/1. Absorption in the medium, for instance, can limit the interaction length and broaden the phase-matching peak. In general, sum-frequency generation from the bulk, if allowed and phase-matched, dominates over that from the surface. In reflection, however, because of lack of phase matching, a surface layer ~ 4/2m thick could contribute significantly to the output. This description gives a qualitative picture of sum-frequency generation, which needs to be substantiated with a more formal treatment. 6.2 FORMULATION The coupled wave approach discussed im Section 3.1 finds a direct application here.? The three coupled waves are E(w,), E(w2), and the sum-frequency output E(w,). Each field can be decomposed into a longitudinal and a transverse part E{w;) = E,(@,) + E, (w,). They obey the wave equations wot Artes? VE Ce) + ela) Eo]. =~ FPP(w) and Vv -[E,(o,) + 4rPP(w,) + 42 PP(4,)] = 0 (6.1) where PC.) = Py(w,) + Py (o,), POC) = OMG, = oy + 3) Ew) E05), PO (0) = Xe, = wy — 1) Elo E*(021), and PO (cs) = xo = wa, + a): E(w,JE(@,), The general solution of (6.1) with boundary condi- tions is extremely complicated. Fortunately, in real situations, reasonable approximations often can be made to simplify the solution, To illustrate, consider a simple case with the following, assumptions: (1) all waves are infinite plane waves; (2) depletion of energy from the pump waves can be neglected; (3) the nonlinear medium is semi-infinite with a plane boundary surface; (4) the nonlinear medium is cubic, or the beams are propagating along a symmetry axis. These assumptions are of course not essential, and in the appropriate circumstances can be relaxed, as we shall discuss later.Formulation o These assumptions drastically simplify the solution. Negligible depletion of pump field energy means that the nonlinear polarization terms responsible for wave coupling and energy transfer in the equations for E(w} and E(w, ) can be neglected. Thus the pump waves propagate essentially linearly in the nonlinear medium with E(w,) and E(@,} governed by the linear wave equations. In the infinite plane wave approximation we have in the nonlinear medium E7(w,) = 8 pexpliflyyf — oyf)] and Ep(o,) = &rexpli(kyy - ¥ ~ 1)]. The only equations left to be solved are those for E(w,) in (6.1) with P@(w,) = Pexpli(ky, «1 — o;f)] and k,, =k,, + k,7. The solution for E{w,) in the medium is straightforward. It comprises a homogeneous solution (a free wave with wavevector k,,) and a particular solution (a driven wave with wavevector ha.) 7 4rw? 40 PP il . E,(a,) = Aebr + 3 2) giksee) pmiase _ 07(13,— K3,)** alos) (6.2) where the amplitude A of the homogencous solution is 4 coefficient to be determined from the boundary conditions, and we assume E,(o,) + Am PP) = &4(05 Er, (60). We now give a more complete description of the problem including the boundary conditions.* Let z = 0 be the boundary plane separating the semi- infinite nonlinear medium on the right and a linear medium on the left. All waves are propagating in the x-z plane with wavevectors described in Fig. 6.2. For each w,, there exists in the linear medium an incoming field E,(w;) from thejleft and a reflected field Eg(w,} to the left, and in the nonlinear medium a wansmitted field E;(w;) to the right. They are related to one another by the boundary conditions. An immediate consequence of matching of the field components at the boundary is that at each @,, all the wavevector components parallel to the boundary surface must be equal. This leads to Snell's law of reflection and refraction for the pump waves. For the sum-frequency wave, we have Kea = Kaan = Katyn = Fas,x+ (6.3) In terms of the propagation angles described in Fig. 6.2, this relation becomes kyp8in dy, = Kypsin Oy = Kypsin Oy = k,,sind,, (6.4) = kysin Oy, + &7sin 7 : = k,,sin@,, + k,sin ,,. Equation (6.4) can be regarded as the nonlinear Snell Jaw. When the wavevec- bo70 Sum-Frequency Generation x Fig. 6.2 Description af wavevectors of various waves involved in sum-frequency generation in a semi-infinite nonlinear medium with a boundary surface at z = 0. tors of the incoming pump waves are known, it determines the propagation directions of the nonlinearly generated output waves.* To complete the solu- tion, one must also find the amplitudes of the output waves. In 6.2), FPE(w,) = x: &767. For a given nonlinear medium, x is prescribed, and &,, and &,; are related to the incoming pump field amplitudes by the Fresnel coefficients. The only unknown in the expression of E;-(w,) in (6.2) is the coefficient A. Then we should also consider the incoming and reflected waves at w,, described by E,(w,) and E,(w,), respectively, in the linear medium E,{as) = &ye@orr os) and Eg (og) = ge}, (6.5) The incoming field amplitude &, is given, but the reflected field amplitude 2, is to be determined. Thus there are two unknown coefficients, A and #,, to be fixed by the boundary conditions. Clearly, the requirement that both electric and magnetic field components parallel to the surface must be continuous provides enough relations to solve for A and ¢,. We postpone the solution to a later section, considering first the case of sum-frequency generation in the bulk.Simple Solution of Bulk Sum-Frequency Generation n 63 SIMPLE SOLUTION OF BULK SUM-FREQUENCY GENERATION We are interested here only in sum-frequency generation in the bulk of a nonlinear medium, as described by the growth of E,(w,) along 2 in Fig. 6.2. Since the growth of the sum-frequency field is generally insignificant over a distance of a wavelength, the slowly varying amplitude approximation dis- cussed in Section 3.3 is applicable here. With E-(#;) = &,(z)exp[i(ky, 1 - @,f)], (6.1) then takes the form a 2m} Ain. (2) = ape ee a saat (68) a [ens Sor + dre! ane =0 where Ak = 2A = ky + kyy — kor (6.7) is the phase mismatch. Solution of (6.6) yields 2rwt : Si (2) = Fy + sO P(e! see) chy, Ak 41 PD 64) Ps, a Sypy(2) = E yp (0) — ale) ake Asa further approximation, we may neglect the effect of nonlinear polarization on reflection and refraction at the boundary. Then, (0) is directly related to the incoming field é,,(0) through the Fresnel coefficients. The intensity of the generated sum-frequency wave at z is given by (z) - 2) The corresponding output power is obtained from the integration of I, over the beam cross section. Here, the finite beam cross section seems to be in conflict with the infinite plane wave assumption, but as is well known, if the beam cross section is sufficiently large, then the ray approximation is valid, and each ray can be treated as an infinite plane wave, Thus with I, depending on the transverse coordinate p, the total output power at w, is IGr(2)P. (6.9) Plos 2) = fi(z. 0)dp ees} (6.10) fies, 0) I? dp.n Sum-Frequency Generation A case of practical interest occurs in the absence of an input at «w, that is, ,, = 0. In the present approximation, &,;(G) also vanishes. Then for |k,/AK} >» 1, we have |éy7,| « |&7, and the intensity /, following (6.8) and (6.9) . takes the form wee cfe(w,) cos’, As shown in Fig, 6.3, 7, versus Ak, z given here peaks strongly at phase-match- ing Ak, = 0. The peak value is F(z) = SIF PP) (6.11) sin(Ak z/2) |? Akz/2 2003 [5 Maas = > ——_— | |P? 222 6.32 cfe(w;) cost, al (69) and the half-width between the first zeroes is (Al) awe = (Seb kyp2 = 2m. (6.13) ty -3n/s —2nf2 ait ait Onis Snr/z Ak Fig. 63 Sum-frequency output as a function of the phase mismatch Ak near Ak ~ 0.Solution with Boundary Reflection B For z = 1 om, ky ~ 10° cm7? in a typical example, we find (AK)yy/k37 ~ 10~‘, which indicates that in terms of Sk, the phase-matching peak is often extremely narrow. The calculation of sum-frequency generation in anisotropic media requires slight modification. First, since P/? in the second equation of (6.1) is usually negligible, Ey (0,)/Ey., (w;) = tana, is a constant determined from linear wave propagation, where a, is the angle between E,(w,) and E, , (w,). With the infinte plane wave approximation, and the slowly varying amplitude approximation, (6.1) becomes a a ere gin isk: ge fur(z) = Top cteosta, 7 Pe! (6.14} where é3y is the unit vector along 637. The solution of (6.14) is 2 2705 &; = 657 (0) + ———_| Par(z) fer (| ) Kqz ,Ake*cos"a, er PEM 1) (6.15) Within the range of our approximation, (6.15) is consistent with (6.8) for the isotropic medium. 64 SOLUTION WITH BOUNDARY REFLECTION x In the more general solution of (6.2) and (6.5), E;(w,) can be rewritten in the form E,(@;) = &y7(z) eer 4nut t. /,.,, the output tends to saturate and may even decrease as / increases. For efficient sum-frequency generation, we must therefore have /,., sufficiently long, of the order of at least a few millimeters in practice. In actual experiments, to avoid reduction of the effective beam interaction length due to finite cross sections, collinear phase matching is required: AK = hyp + hyp — Ryp = 0. (6.20) In terms of the refractive indices 2(w,), (6.20) can be written as a, [a(@,) ~ m()] + @,[2 (05) - n(a,)] = 0. (621) Clearly, for isotropic or cubic materials with normal dispersion 2(«;)> {n(,), n(#)), this relation can never be satisfied. Therefore, collinear phase matching can be achieved only with (1) anomalous dispersion or (2) birefrin- gent crystals. In the latter case, the medium should be a negative uniaxial crystal with n,(w,) < 25(w,). By choosing the wave at w, to be extraordinary, it is possible to find [n,(w,)— a(,)] and [1,(w3) — n(,)] with opposite signs so that (6.21) can be satisfied. Two types of collinear phase matching are commonly used. In Type I, both n(@,) and n(w,) ave ordinary or extraor- dinary, while in Type IJ, either n(w,) or #(@,) is ordinary. 66 EFFECT OF ABSORPTION Absorption is detrimental to sum-frequency generation since it limits the effective interaction Jength to roughly the attenuation length, It also broadens the phase-matching peak and lowers the peak value. This can be seen by including absorption in the derivation in Section 6.4. With absorption, the wavevectors become complex: & = k’ + if, where B is the attenuation coeffi- cient. Equation (6.14) changes into 7 2 i2nw} a TO aye Pel MAE UBart d= (6,22) Php CO8"A, (4 + yr |ar(2)‘Sumt-Frequency Generation with High Conversion Efficiency 7 with Ey (a5) = Bp (2) e8P-. The resulting solution is mexeyr PO *kjy costa [ik — (Bur + Bor ~ Por] (6.3) x [eraser ey - enor]. Fyp(z) = Fyp(O)e Pe + If the absorption at either the pump frequencies #, and w, or the output frequency «, is appreciable so that fetes Birt Bard = en Bare? =1, then the output intensity can be approximated by . 2nahlés PPP cfe(«os) cos*Bypcos‘a,|(Ak’Y’ + B| where 6 = By + Bor with B,,~0 or B= Bsr with By, + B,y~ 0. The phase-matching curve /, versus Ak now takes on a Lorentzian shape with a half-width 8. Compared with the zero absorption case, the peak value here is independent of z and is reduced by a factor of 1/%z*. This shows that with absorption, the effective interaction length is reduced to 1/8, which is just the attenuation length. When both (f,\;+ Br) and f, are appreciable, the output intensity even decreases exponentially with z. 4,(2) (6.24) 6.7 SUM-FREQUENCY GENERATION WITH HIGH CONVERSION EFFICIENCY We saw in earlier sections that at perfect phase matching, the output power of sum-frequency generation in a nonabsorbing bulk medium is proportional to FP, the square of the length of the medium. Then as / > 00, the output power would increase without limit, in violation of energy conservation. This is the consequence of the assumption of negligible pump power depletion, which is not valid when the output becomes significant compared to the pump. The set of three coupled equations in (3.4) or (6.1) must now be solved together to find a complete solution. For sum-frequency generation with high conversion efficiency, the following conditions usually exist: (1} the coupled waves are collinearly phase matched: (2) the medium is nearly lossless, and (3) the slowly varying ampiitudeB Sum- Frequency Getterition approximation is valid. The coupled equations can therefore be written as [similar to (6.14}} a8, iw Fir 1g gu Bz kn costg, Mtr Sr ar 2008" (6.25) 06, fx} Ar ig gee 82 kyp costa, ? and aes, -iw} esr 83 lk ek az Key. 40080 3@1reor with 2a, - k= Bair O(a =~, + 05): bore yr, 2m, 33 K,= aba KOM = — 1); Brey, and From the permutation symmetry of x? in a lossless medium discussed in Section 2.5, we find K, = K, = K; = K. Equation (6.25) can be solved exactly? First, we can easily show from (6.25) that the total power flow W in the medium, 2, 2 2 2 2 2 w= o? | ky 08" irl + Kar, ,C0s’a,|8zrl + Ryr £08"03/E5r| 20 en @ os (6.26) is a constant independent of z. This is also known as the Manley-Rowe relation.’ Then the number of photons created at w, must be equal to the numbers of photons annihilated at w, and «,: G =I 87 (0)[* = Wor (2)? fey I (0) 2 ; ! ( ye = fe I rr ( I ¥ ! rar Mt (6.27) = gel = Gr (O)P hoySum-Frequency Generation with High Conversion Efficiency 9 In solving (6.25), we define i hy £08" “e we = ew Gi7{z), i kay 008% \'* were | Hk | far(2). die 2g \17 ue" = (: 37, :008"a3 E7(2). meh (6.28) 8(2) = o5(2) - dfz}- #:(2), t- «| 27 Welaiwr a . thy ghrp pk yp, :008"0C08'a,C08*0, am, = u}(0) + uf (0) = 43 + 05, am, = 03(0) + u2(0) = u2 + v2, and ms, = u2(0) — u3(0) = u? - 43. Equation (6.25) becomes | 3 = —uu,sing, | 2 = ~uynsind, (6.29) a = uju,sin 8, and = = 1 BF u8). The last equation in (6.29) can be integrated to yield Uyu,u,cos 8 = T where I’ is a constant independent of z. Then by eliminating sin @ in (6.29), we 12 a de = +2[upuju? - 17}, (6.30)30 ‘Sum-Frequency Generation The choice of sign “+” or “‘—” in (6.30) depends on the initial value of #. The general solution of (6.30) is given in Ref. 3. We consider here the frequently encountered case where the boundary condition is #,7(0) = 0, or w3(0)= 0, which leads to T= Q and @ = 2/2. Equation (6.30) becomes 8 = 2[u2(m, — u2)(my — 02 )]. (6.31) The solution takes the form of a Jacobi elliptical integral al gip)-= He = 2/2, 0) a 4; (0) 4uso/en@ [(1 — py?) — ¥2y?)]'? (6.32) assuming m, = u?(0) > m, = u3(0) and y = #,(0)/u,(0). From (6.32) and (6.28), we find the intensities of the three waves as 43(5) = u}(0)sn?[,(0)5, v], 6.33) 2A(8) = u}(0) - w2(0)s?[xy(0V5 6) and at ($) = uf(0) — 4} {0)sn* {ay (0)E, y} The elliptical function sn*[u,(0)f, y] is periodic in ¢ with a period & (6.34) 2 4 [a =») - ve)? Physically, this indicates that as the interaction length increases, energy is transferred back and forth between the wave at w, and the waves at w, and w, with a period L. While the process first pumps energy into the sum-frequency field, it reverses the energy flow after photons in one of the pump waves are depleted. A simple case of physical interest is the up-conversion process used, for example, to convert an infrared image to the visible. It often occurs with u4,(0) ® u,(0) and u,(0)=0. Since y « 1, the elliptical integral of (6.32) reduces to a simple form and we find 13 (8) = 3 (0)sin? [ x, (0)5], w(t) = a3 (0) — usin? u (0)8]. (635) and ui ( = 7 (0) — u3(0) sin? fu, (0) )e] = uf (0).Sum-Frequeticy Generation with High Conversion Efficiency 81 They are plotted in Fig. 6.5, which shows explicitly the periodic variation of energy flow back and forth between the waves at w, and «,. In this case, the depletion of the pump field at w, is negligible. Therefore, the solution in (6.35) can also be obtained from (6.25) by letting ,, be a constant. Another case of interest is when 4,(0) = u,(0} so that y = 1. The solution becomes u3(£) = uw? (O)tank?[ u,{0)5], (6.36) uf (5) = uB(f) = 0} (O)sect?[ w,(0) 1]. It shows that the period of interaction L is infinite. As. § > oo, we have w(6) > 4,(0) and 1,(f) = u,({}— 0. This applies to the case of second harmonic generation, which we discuss in mere detail in Chapter 7. The foregoing discussion is based on the assumption of infinite plane waves. In reality, the beam cross sections are finite with intensity variation over the transverse profile. Accordingly, the results here have to be modified, using, for example, the ray approximation. As a result, complete depletion of photons in any beam is impossible. Focused beams are often used in actual experiments to increase the laser intensity, and the theoretical treatment of the problem io? mm INITIAL DISTRIBUTION ay ng/n,=0.01 RELATIVE NUMBER OF PHOTONS o> 77 ™ z Fig. 6.5 Relative numbers of photons, as a function of z, in the three coupled waves with perfect phase matching (ww, ++ w», Ay = & + &2) in an up-conversion pro- cess, The initial distribution of the photons in the three waves is 2, = 100 , and ny = 0. (Alter Ref. 3)a2 Sum-Frequency Generation becomes more complicated. Boyd and.Kleinman’ have worked out the case with negligible depletion. Here we simply refer to their work and postpone our discussion on focused beams to the next chapter. 68 A PRACTICAL EXAMPLE In most applications, efficient sum-frequency generation is desired. A number of rules should therefore be followed : 1 A soslinear optical crystal with little absorption at 0, a, and wz is first chosen. It should have a sufficiently large nonlinear susceptibility x? and should allow collinear phase matching. 2 The phase matching directions in the crystal, generally in the form of a one, are determined from the known refractive index tensor of the crystal. 3. The particular phase-matching direction with the appropriate set of polari- zations for the three waves is selected to optimize the effective nonlinear susceptibility x = é5-x© : &)6). 4 The length of the crystal is finally chosen to give the desired conversion efficiency. We consider here a practical example of sum-frequency generation in a KDP crystal with the pump beams at A, = 5320 A and A, = 6200 A. The sum-frequency generated is at A, = 2863 A. The ordinary refractive indices of KDP at room temperature are 1o(@,) = 1.5283, 1 9{w,) = 1.5231, and ng(w;) = 1.5757. For a beam propagating in a direction at an angie away from the optical axis, the extraordinary refractive index is given by _ Fem 2) ( ;) («-@) [tan @) + alain] with n,,,(0,) = 1.4822, 2,,,(c2) = 1.4783, and n,,,() = 1.5231, For type I phase matching, we have from (6.21) neler, @) | = Polen) + Zemo(es) = 1.5258 from which we can find @ ~ sin] Mae), [B= MC) n.{o) gC) — Man s) = 76.6°.Limiting Factors for High Conversion Efficiency 3 Let the waves be propagating in a plane at an angle ® from the X-axis of the crystal. In the X—- Y—Z coordinates, the three polarization vectors are é, = @ = (sin ®, —cos ®, 0) é,= (-sos@) cos ®, -cos(#) sin , sin(#) ). The KDP crystal has a 4 2-m point group. Its nonvanishing x! elements are xPho = XV = xy = xPbx = 2.6 x 107% osu and and xPhy = XGhx = 2.82 X 107% esu, The effective nonlinear susceptibility* for type I phase matching is X= yx 2 = —yPhrsin GP) sin 2d = —2.74 X 107%sin 2® esu. To optimize }x|, we must choose # = 45°. Finally, in the limit of negligible depletion of pump power, the output power is, following (6.12), given by 85a} P,P, = Se ear } FYe(w,e(w,)e(o;) 4 =4x vo-‘z2( PiP2 MW, where A is the beam cross section in square centimeters, z in centimeters, and we have used the approximation P, = J, in megawatts. 69 LIMITING FACTORS FOR HIGH CONVERSION EFFICIENCY As a nonlinear effect, the output power of sum-frequency generation is expected to increase with the pump intensity if the pump power is kept the same. This seems to suggest that a tighter focusing of the pump beams should *The expressions of x} for type I and type I phase matching for the 13 uniaxial crystal classes can be found in Ref. 8.cy Sum-Frequency Generation be used to attain higher conversion efficiency, as Long as the jongitudinal focal dimension (the confocal parameter) is longer than the effective interaction length. There is, however, & limit to the focusing one can use. First, too high @ jaser intensity leads to optical damage in the crystal. Then, the reduced beam cross section due to focusing may decrease the effective interaction length even for collinearly propagating beams. This occurs in an anisotropic crystal. For and extraordinary wave, the directions of wave propagation and ray (energy) propagation are generally different. Therefore, although the waves are propa- gating collinearly, the rays are not, “Walk-off” of the rays effectively decreases the interaction length. The walk-off effect can of course be minimized if the beams are propagating in a direction along which the wavevector and ray vector are parallel. This may be achieved for sum-frequency generation in a uniaxial crystal in a plane perpendicular to the optical axis, and is known as 90° phase matching. Such phase matching has been found possible in many crystals over a certain frequency range by temperature tuning. The poor beam quality also reduces the conversion efficiency. A multimode laser beam can be considered crudely as 2 beam with hot spots. The small dimension of these hot spots increases the walk-off effect and decreases the interaction length. Therefore, for high conversion efficiency, beams with TEM mode should be used. Good crystal quality is also important for efficient sum-frequency genera- tion. Inhormogeneity prevents perfect phase matching throughout the crystal. Since |Akz| < 2/2 is needed for efficient energy conversion, the tolerable fluctuation of the refractive index due to inhomogeneity is An < A/4z = 25% 1073 for k= 1 pm and z = 1 om. This means that the requirement on the crystal quality is stringent. For the same reason, temperature uniformity throughout the crystal length is also important. For a typical case with dn/dt =5 x 1075, a temperature stability of AT < 0,5 K throughout the crystal is necessary to achieve An < 2.5 x 1075. This discussion generally applies to all mixing processes. REFERENCES 1M. Bass, P. A. Franken, AE. Hill, C. W. Peters, and G. Weinreich Phys. Rev, Lett. 8, 18 (1962). 2 PD. Maker, R. W. Terhune, M. Niscnboff, and C. M. Savage, Phys. Ret Lat. 8, 21 (1962); 1A. Giordmaine, Pays. Rev. Lets, 8, 19 (1962). 3. J. A Armstrong, N. Bloembergen, 3. Ducuing, and P. S. Pershan, Phys. Rev. 127, 1918 (1962). “LN, Bloembergen and B. 5. Persban, Phys. Rev, 128, 606 (1962). 5 J, Ducuing and N. Bloembergen, Phys. Rev. Lett, 10, 475 (1963); R. K. Chang and N. Bloembergen, Phys. Rev. 144, 775 (1966). 6 JM Manley and H, E. Rowe, Prec. IRE 47, 2115 (1959).Bibliography 85 7G. D. Boyd and D. A. Kleinman, J. Appl. Phys. 39, 3597 (1968). 8 F. Zermike and F, E, Midwinter, Applied Nonlinear Optics (Wiley, New York, 1973), pp. 64-65. BIBLIOGRAPHY Akhmanov, §. A., and R. V. Khokhlov, Problems of Nonlinear Optics (Gordon and Breach, New ‘York, 1972) Bloembergen, N., Nonlinear Optics (Benjamin, New York, 1965). Zernike, F., and I. E. Midwinter, Applied Nonlinear Optics (Wiley, New York, 1973).7 Harmonic Generation In the history of nonlinear optics, the discovery of optical harmonic generation marked the birth of the field.! The effect has since found wide application as a means to extend coherent light sources to shorter wavelengths. This chapter summarizes the important aspects of harmonic generation. As it is a special case of optical mixing, most of the discussion in Chapter 6 can be applied here without much modification. The application of harmonic generation to the measurements of nonlinear optical susceptibilities and to the characterization of ultrashort pulses is also discussed. 7.1. SECOND HARMONIC GENERATION The theory of second harmonic generation follows exactly that of sum-frequency generation discussed in Chapter 7. With #, ~@, = @ and w, = 2a, the derivation and results in Sections 6.2 to 6.6 can be applied directly to the present case. In particular, the plane wave approximation with negligible pump depletion yields a second harmonic output power 32n'a? liso x2s8,8 asin (Ak 2/2) P30) ce(w)ye(2u) (Akzyzp 4 P,,(2) = (7.1) For collinear phase matching in a crystal with normal dispersion, we must have, following (6.20), either nQQw)= mole) (type 1) (7.2) or n,(20) = 4[%o(w) +n-(@)] (type TI). (73) 86Second Harmonic Generation 87 The calculation in the limit of high conversion efficiency requires slight modification. Specifically, for type I collinear phase matching, the permutation symmeiry of x, following (2.30), gives 2 ; Ka Tay, xP Qu = + 0): 2,8, © (74) = ae Pw = -w + 2w): be, The coupled equations of (6.24) reduce to the form! a6 -— i axes shy, er ky, £0s"a,, | (7.5) Ey ~i20) ge G2 hay, 00870, The conservation of power flow and the conservation of the number of photons in (6.26) and (6.27), respectively, become w wn Lr | Ko 2087 HEI | Boe 2087216 ru” Qe 20 (0)? - 16.(2)P _ 9 !Sro(z)1? = 16.0)? (7) fas 2ho . Using the definition of u,, and u,, [u, and u, in (6.28) with w, = w, = @ and @, = 2], we obtain du, . a7 Zu uy, sind and Mae oats a uZsind (7.7) If we assume u,,,(0) = 0, then @ = 1/2, and the solution takes the form a(S) = Aopen 2 4,(0)¢], Wo,(5) = F u, {Oran 72 w,(0)¢] (78) The second harmonic output power is then given by P,.(2) = P.(O)tant?[C(P,(0)/4) "| (19)88 ‘Harmonic Generation where C = K(Qw/ ¥e\(2a0/ve)”?, assuming, for simplicity, hy, , = ke = fkyy = doe, and a, = Mao Following (7.9), Fig. 7.1 shows how P,,,(2) increases with z at the expense of P.,(z). As a practical example, we consider the use of KDP as a second-harmonic generator for a Nd: YAG laser beam at 1.06 um. Using the same calculation as in Section 6.8 with mow) = 1.4939 and 1,,,(2@) = 1.4706 at room tempera- ture, we find that for type 1 collinear phase matching, the beams should be propagating in a direction at an angle = 40.5° away from the Z-axis of the crystal. The pump field should be linearly polarized in a plane bisecting the X-Z and Y-Z planes in order to yield an optimum xB = x Poy (20 )sin! = 15% 107% esu. With the plane wave approximation, the efficiency of second harmonic generation, following (7.9), is mh mal" = iant?{4.7 x 10-2 P]'“} (P, in MW). (7.10) ‘As seen from (7.10), the efficiency q teaches 58% when [P,)/4}'z = 21¥MW , or P,,(0)/A = 1B MW/cn? for z = 5 cm. In an actual experiment, 7 is often less because of the finite dimension and transverse intensity variation of the pump beam. An efficiency as high as 40%. with giant pulses or 85% with ultrashort pulses has, however, been. achieved? The foregoing calculation docs indicate that in order to have an appreciable conversion efficiency (n > 10%) in a crystal like KDP, a pump intensity of the order of 10 MW /ent is needed —— FUNDAMENTAL += tad HARMON IC NORMALIZED AMPLITUDE an t at z Fig. 7.1 Decay and growth of the nonmalized fundamental and second harmonic amplitudes, respectively, for the perfect phase-matching case. (After Ref. 1)Second Harmonic Generation by Focused Gaussian Beams a with a crystal length of a few centimeters. (A much longer crystal is seldom practical.) In general, higher pump intensity leads to a larger n, except in the limit of very high conversion efficiency.? For a low-power pump beam, therefore, focusing is generally used to increase y and hence the second harmonic output. Focusing, however, increases the walk-off effect, but as mentioned in Section 6.9, it also increases the spread of the beam propagation. The spread hurts the conversion efficiency as a portion of the beam now deviates from the exact phase matching direction. For type I collinear phase matching at the angle Ga) , for example, it can be shown readily that a small deviation A@ of the beam propagation direction from @) leads to a phase mismatch? I 1 Ak = dkyng(o)| =o sin 2(H ) Ad. TAM inwito| ats ~ gee pe@O — on With Aki = m being the half-width of the phase-matching peak, the acceptance angle about one can allow for beam convergence is, from (7.11), ln 3 (2w) 1 A= Shs : (7.12) ‘ laa ean sin2(H For a KDP crystal with = 1 cm and (3) = 45° at A, = 1.06 pm, the a ance angle AG, is only 2.5 mrad. Equation (7.12) shows that 46 diverges as approaches 90°. This occurs because the higher order terms of 4@ were neglected in (7.11). The correct result for @) = 90° is, assuming no(2u) ~ n,(2w), yt (7:13) at, = | n,(20) kyl[no(2#) — n.(20)] For a 1-em KDP crystal at A, = 1.06 wm, the acceptance angle is 36 mrad, which is an order of magnitude larger than in the previous case. The large acceptance angle for 90° phase matching is clearly advantageous if beam focusing is used in second harmonic generation. 1.2 SECOND HARMONIC GENERATION BY FOCUSED GAUSSIAN BEAMS Single-mode laser beams usually are required for efficient second harmonic generation, The conversion efficiency may be greatly enhanced by focusing of the pump beam into the nonlinear crystal. At the focal waist a single-mode” Harmonic Generation beam has a Gaussian intensity profile described by exp(—p’/W(2) with WM being the beam radius. The longitudinal dimension: of the focal region is defined by the confocal parameter 5 = kW as the distance between two points on the focusing axis where the beam radii are /2 times larger than that at the waist. Within the focal region, the beam has approximately a plane wave front, so that the plane wave approximation can be used. We consider first the case of negligible double refraction or walk-off effect, eg. the 90° phase-matching case. Obviously, if the crystal length / is smaller than the confocal parameter }, the conversion efficiency for second harmonic generation can be described by the result of plane wave approximation in (7.10) with A = We, (0) |! n= ete [Ee th, (7.14) Here, as long as b> /, tighter focusing should increase P.,(0) awe and improve the efficiency. If 6 < /, however, the approximation breaks down, and tighter focusing tends to reduce the efficiency. Thus optimum focusing occurs when the confocal parameter is about equal to the crystal length, b ~ 4. Boyd and Kleinman‘ studied the focusing problem in detail with numerical calcula- tion, They introduced an efficiency reduction factor Ao(£) with € = 1/6 to take into account the effect of focusing on the efficiency*> They find a= wir © cp, AO) k lft (€) |? (715) The function 4g(£}is plotted in Fig, 7.2. For & = //6 < 0.4, we have #o(é) = & and 7 in (7.15) reduces to the form in (7.14) as we expected. In the tight focusing limit, > 80, the function Ag() takes the asymptotic form 4 {£) = 1.1879?/€, and the efficiency drops rapidly with increasing ¢ The maximum value of Ag(€) = 1.068 appears at $= 2.84, with Ag(¢)= 1 in the range 1 << 6, This shows that although optimum focusing occurs at b = [/2.84 for given /, the efficiency will not decrease appreciably even if 6 = /. With double refraction, the situation is more complicated. Boyd and Klein- man showed that in the limit of low conversion efficiency, (7.15) is still valid if hg(@) is replaced by A(B, £) [A(0, &) = Hg(£)} or _ £7P.,(0)k oth (B, $) T (7.16) where B = $9(k_J)'” is a double refraction parameter, and = tan- ni (w) 1 pre { 2 Ec 220) “Fo |Second Harmonic Generation by Focused Gaussian Beams 91 10 10-2107? 1 10 10? 19° Fig. 7.2 Efficiency reduction factor A(B,€) versus € = i/b at various values of the double refraction parameter §. (After Ref. 4.) is the walk-off angle between the Poynting vectors of the collineatly propagat- ing fundamental and second harmonic waves along a direction at an angle away from the optical axis. The function 4(B, ¢} for several values of B is seen in Fig, 7.2. Note that 4(B,¢) depends only weakly on € near its maximum hy-(B). The latter can be approximated to within 10% by the expression’ hy (0) 1+ (4B7/a)hy (0) 47) fy (B) = with A,,(0) = 1.068. This equation together with (7.16) indicate that the efficiency reduction due to double refraction becomes appreciable when (4B? /71)h yg (0) ~ 1. We can define an effective length l= ee = 7.18 keh (0) ye’ 18) Equation (7.17) becomes _ hu Aud 8) = TT (7.19) This shows explicitly that when /.¢ = /in the presence of double refraction, the efficiency for second harmonic generation with optimum focusing reduces by a factor of 2 as compared to the case without double refraction. When (.¢ « /,92 Harmonic Generation this efficiency becomes 2 x= Pala), (720) The effect of double refraction on 4,,. is insignificant only when / 2,G) in a pure alkali vapor. If a buffer gas (e.g., Xe) with normal dispersion 1,(w) Toye and hence P(2w) should decrease inversely with * as a result of decrease in the number of particles in the light path as shown in Fig, 7.9. With this technique, numerous materials have been surveyed. They can be divided into five groups:?° centrosymmetric, phase-matchable, non—phase-matchable, large nonlinear coefficient, and small nonlinear coefficient. 7.6 SECOND HARMONIC GENERATION WITH ULTRASHORT PULSES Second harmonic generation with ultrashort pulses requires some special consideration. With the pulse length smaller than the length of the medium, the nonlinear polarization varies drastically along the length at a given time. The only simple case occurs when the group velocities of the forward propagating fundamental and second harmonic waves are the same. Then the two pulses will propagate together and interact with each other as if the problem were stationary. This is the quasi-stationary case. The solution is identical to that of the stationary case if z — v,t replaces z where p, is the group velocity. If the group velocity dispersion is nonnegligible, then the solution becomes much more complicated. Physically, the velocity mismatch causes the fundamental pulse to displace against the second harmonic pulse as they propagate along. This reduces the effective interaction length and decreases the conversion efficiency.» A rigorous mathematical treatment of the problem has been worked out by Akhmanov et al.” Infinite plane waves propagating along z with slowly varying amplitudes are assumed. As shown in Section 3.5, the pulse propaga-104 ‘Hanmonic Generation tion of a wave, &(z, Hexp[ikz — iwt], in a nonlinear medium can be described by ae 188 _ idnat Hye ket PNU( 2, pe fee, (7.26) In the present case, the group velocities of the fundamental and second harmonic waves are 0,, and 02,, respectively. If we use, as independent variables, z and 4 = f — 2/0, instead of z and ¢, then the wave equations governing the fundamental and second harmonic wave amplitudes &,(z,7) and €,,,(z, ) under the phase-matching condition become a6 o6,b145 dz onle a6, ae. (7.27) 2 20 = gf? e +p an 08) where v = v;,| — oj, and @ = (220? kc? )bo,° XP: 2,2, assuming &,,(2 = 0) = 0 The solution of (7.27) is nontrivial. Akhmanov et al.” showed that the coupled nonlinear equations can be combined into a single second-order differential equation er. 1 Bla) roma (728) where F(q — 92) = 0?(62 + 62.) + ovdd,/dq is a function of (4 — »z) only, as can be shown by the vanishing Jacobian aF/an, aF/a2 a a =0. Aum) Glr-”) Equation (7.28) is now a linear equation with a varying coefficient F. It can be solved analytically for arbitrarily large conversion efficiency. Let the fundamental pulse at z = 0 take the form 6,(0) = ad + 5} (7.29) Here, 7 is the pulsewidth. In addition, we define a number of characteristic lengths for the problem: Z is the length of the medium; Ly, = 1/cA is the interaction length at which 75% of the fundamental power is converted intoSecond Harmonic Generation with Ultrashort Pulses 105 second harmonic power in the stationary case; L, = 7/ is the propagating distance over which the overlapping fundamental and second harmonic pulses of width 7 are clearly separated. With a new set of variables y= 1/7, £3 2/Ly, My = Lys f= (92/18 — WF and € = [7/1 — UY? x fan — tan71(% — )], the solution of (7.27) has the form 4 €,= (44 + wen ve yr + a rsiong) (7,30) gta | Zooshé +[f— A(a — £)/f]sinhé a-i) 4.7 48 cosh £ + (4/7 sinh & \/bewe YI. When L, > Ly, (7 > tq), the group velocity dispersion is clearly negligible so far as second harmonic generation is concerned. En this quasi-stationary case, the solution in (7.30) reduces to €, (2,0) = 5 sech} — olen) “alas (731) A z é = —4 tanh] ——2—_— ea(ze) 1+# | (+?) Ly They are exactly the same as the stationary solution for second harmonic generation given in (7.8) if we replace &,(z = 0) there by 4/(1 + 7), In the limit of negligible pump depletion, z « (1 + #) Ly, the second harmonic field from (7.31) is proportional to the square of the fundamental field (2,0) = 62 (W)z/AL ya. Then the second harmonic pulsewidth is approximately half of that of the fundamental. When L, < Ly,, the group velocity mismatch becomes important, and the general solution of (7.30) must be used. There is a relative displacement between the fundamental and the second harmonic pulses. Consequently, the conversion efficiency drops, and the second harmonic pulse broadens. This has been experimentally confirmed.’ On the other hand, one can also use the group velocity mismatch to sharpen an input second harmonic pulse through amplification. If the fandamental pulse is appreciably longer than the second harmonic pulse, and if ¥,9 > u,,, then the two pulses can be arranged in such a way that the leading edge of the second harmonic pulse always sees the undepleted part of the fundamental pulse and gets amplified more than the lagging edge, resulting in a sharper output pulse.106 Harmonic Generation The group velocity mismatch is generally more appreciable at higher fre- quencies because of anomalous dispersion duc to absorption bands in the uv region. For a 1-psec pulse propagation in KDP, for example, £, = 3.cm at = 1.06 pm and L, = 0.3 cm at A= 053 am. Thus the effect of group velocity mismatch is much more important for frequency doubling of picosec- ond pulses into the uv. REFERENCES 1 JA. Ammstong, N. Bloembergen, J. Ducuing, and PS, Pershan, Phys. Ret 127, 1938 (1962); N. Bloembergen, Nonlinear Optics (Benjamin, New York, 1965), p. 85. J, Reintjes and R. C. Eckardt, Appl. Phys. Let. 30, 91 (1977). R. L. Byer and R. L. Herbst, in Y. R. Shen, ed., Nonlinear Infrared Generation (Springer-Verlag, Berlin, 1977}, p. 81. G. D. Boyd and D. A. Kleinman, J. Appl. Phys. 39, 9597 (1968). D.R. White, EL. Dawes, and J, H. Marburger, EEE J. Quant. Electron. QE-6, 793 (1970). R. S. Adhav and R. W, Wallace, JEEE J. Quant. Electron. QE-9, 855 (1973). RW, Terhune, Solid State Design 4, 38 (1963). See, for example, the review article by R. Piston, Laser Focus 14(7), 66 (1978). RB. Miles and S. E, Harris, Appf. Phys. Lets. 19, 385 (1971); [EEE J. Quaret, Electron. QE-9, 470 (1973). 10 3. Young, G. C. Bjorklund, A. H. Kung, BR, B, Miles, and 8. E. Harris, Phys. Rev. Lett. 27, 1551 (1971). 11 D.M. Bloom, G. W. Bekkers, J. F. Young, and §. E, Hartis, Appl. Phys. Lett. 26, 687 (1975); D. M. Bloom, J. F. Young, and S. E. Harris, Appl. Phys. Lett. 27, 390 (1975) 12 AHL Kang, J. F. Young, G. C. Bjorklund, and §. E Harris, Phys, Rev. Lett. 29, 985 (1972); AH Kung, 1. F. Young, and 5. E. Harris, Appl. Phys. Lett. 22, 301 (1973) [Erratum: 28, 259 «1976)]. 13S. E Harris, Phys. Rev. Lett 31, 341 (1973). 14. 5. Reingjes, C. Y. She, R.C. Eckardt, N. E Karangelen, R. C. Etton, and R. A. Andrews, Phys. Rev. Lett. 37, 1540 (1976); Appl. Pays. Leit. 30, 480 (1977). 15 G.E. Francois, Pays. Rev. 143, 597 (1966); J. E. Bjorkholm and A.B. Siegman, Phys. Rev. 154, 851 (1967). 46. J. Jerphagnon and S, K. Kurt, Phys. Rev. BL, 1739 (1970). 17 PD. Maker, R. W. Terhune, M. Nisenoff, and C.M. Savage, Phys. Rev. Leit. 8, 21 (1962). 18 1. E. Wynne and N. Bloembergen, Pips. Rev. 188, 1211 (1969); R. C. Miller and W. A. Nordland, Phys. Rev. B2, 4896 (1970). 39 J, Ducuing and N. Bloembergen, Phys. Rev. Fett. 10, 474 (1963); R. K. Chang and N. Bloembergen, Pays. Rev. 144, 775 (1966). 20S. K. Kurtz and T. T. Perry, J. Appl. Pays. 39, 3798 1968); 8, K. Kurtz, EEE J. Quant. Electron. QE*d, 578 (1968). 21 SA. Akbmanoy, A. 5, Chitkin, K. N. Drabovich, A. I Kovrigia, R. V. Khokhlov, and A. P. Sukhorukov, FEEE J. Quant. Electron. QE-4, 598 (1968) J, Comly and EB, Garmire, Appl. Phys. Lett. 12, 7 (1968). $. Shapiro, Appl. Phys. Ler. 13, 19 (1968). wn Cea aAwe BRBBibliography 107 BIBLIOGRAPHY Akhmanov, 3. A, A. I. Kovrygin, and A. B, Sukhorukoy, in H. Rabin and C. L. Tang, eds.. Quantum Electronics (Academic Press, New York, 1972), Vol. 1, p. 476. Bloembergen, N., Nonlinear Optics (Benjamin, New York, 1965). Kleinman, D, A., in F.'T. Arecchi and E, 0. Schulz-Dubois, eds., Laser Handbook (North-Holland Publishing Co,, Amsterdam, 1972), p. 1229. Pressley, R. J., ed., Handbook of Lasers, (Chemical Rubber Co., Cleveland, Ohio, 1971), p. 489. Zernike, F. and J. E, Midwinter, “Applied Nonlinear Optics” (Wiley, New York, 1973)8 Difference-Frequency Generation Difference- frequency generation is theoretically not very different from sum- frequency generation, but the problem is of great technical importance in its own right, as it provides a means for generating intense coherent tunable radiation in the infrared. Traditionally, blackbody radiation has been the only practical infrared source. Yet, governed by the Planck distribution, it has weak radiative power in the medium- and far-infrared range. A l-cm?, 5000 K blackbody radiates 3500 W over the 4q solid range, but its far-infrared content in the spectral range of 50+ 1 cm! is only 3 X 107° W/cm? - sterad. Infrared lasers may seen to have all the desired properties as infrared sources but their output frequencies generally are discrete with essentially zero tunabil- ity. Tunable infrared radiation can, however, be generated by difference- frequency generation. Being coherent with high average or peak intensity, it may find many applications in the field of infrared sciences. This chapter deals mainly with infrared generation by difference-frequency mixing. The diffrac- tion effect is considered in the long wavelength limit. Far-infrared generation by ultrashort pulses is also discussed. $1 PLANE-WAVE SOLUTION In the infinite plane-wave approximation, the theory for difference-frequency generation follows almost exactly that of sum-frequency generation if the pump intensities can be approximated as constant. Then the output power at @, = @; — @, generated from the butk is Po, Co) seat, XP (doy = 8g — a) BE 2m (Akz/} 2 cl feCw, ela, ) Cos) ? 2 ON nk 2 x Plo IPles) (8.1) 108‘Plane-Ware Solution 109 With phase matching and in the presence of appreciable pump depletion, the solution of Section 6.7 must be used. However, the usual initial boundary condition is 4,(z = 0} = 0 [i.e., €(0) = 9; the notations here follow those in Section 6.7.] in the present case. The equation to be solved becomes (9 = — 2/2) 2 4.[3(m, — 23)(m, + 08))'7 (8.2) with m, = u}(0) and m, = u7(0). The solution takes the form u3(S) = — uf (O)su? [14 ,(0)5, v], u}(£) = u2 (0) — wf (O)sn*[ iu, (0)S, x], (8.3) u3(S) = 230) + wf (0)sn?[ iu, (0)S, 7] where ~1f 42 a fii) dy : sn (a) f [a -y7)(1 — yj]? ix, (0) 43 (0) ” In the particular simply case where u3($) « 4}(0) or |y?y7| <1, we have sn[ize,(0)$] = i sinh[w,(0)S] and hence 13(£) = up(0)sinh? [45(0)¢}, Hf (5) = up (O)cost? [u,(0)s], (8.4) 49 ($) = 03 (0) — uf (0)sinh? [w5(0)¢] = u3 (0). For |w;(0)§| «<1, this solution leads to (8.1) with Ak = 0. The above results can of course be obtained directly from the coupled wave equations of (6.25) by letting & be a constant. The more general solution with &, = constant, €,(0) * 0, & (0) ¥ 0, and Ak # 0 can also be obtained, but this is postponed to the next chapter in connection with a discussion of parametric amplification. The plane-wave approximation adopted here is good as long as the output wavelength is much smaller than the beam cross section. The foregoing results should describe fairly well near-IR and mid-IR generation by difference- frequency mixing. Experimental reports on the subject are numerous, and have been summarized in recent review articles»? An important fact to realize is that the efficiency of infrared generation is expected to be low because of its dependence on the square of the output frequency as seen in (8.1).10 Difference-Frequency Generation 82 FAR-INFRARED GENERATION BY DIFFERENCE-FREQUENCY MIXING The infinite plane-wave approximation ceases to be valid for far-infrared generation in the long-wavelength limit as diffraction becomes important when the pump beam diameter appears comparable to the far-IR wavelength. We must look for a better solution of the wave equation A103 Lo a PO(w) (8.5) |v x(x) ~ Eee) -|Elo) = with P&(w,) = xo) = 4 — ,) : E(e3)E*(w,). Since the conversion efficiency is expected to be small because of the small @,, the depletion of the pump fields can be neglected and the amplitude of P can be regarded as independent of propagation. If we neglect the boundary reflections by assuming that the nonlinear crystal is immersed in an infinite index-matched linear medium, the far-field solution of (8.5) has the familiar expression® P(r’, w,) e741 SOME Sere 6 7 (8.6) E(r,@,) = (2) fare - where V is the volume of interaction of the pump fields in the nonlinear medium. With P@(r’, o,) known, Er, t,) can be calculated. Consider, for example, the case where the pump fields F(«,) = E\explik yz — iwyt) and E(w;) = 8,(texplikyz — iw,t) can be approximated by &,(2) and £3(8) being, constant in a cylinder defined by (x2 + y?) ¢ a? and vanishing elsewhere. The nonlinear polarization is assumed to have the form (1 = FF) P(r, ta) = PMettae’-42) for G24) sa? =0 for(x?+y?) >a? (8.7) ka, = ha oh With this expression of P®, (he integral in (8.6) can be readily evaluated. Let r= droos} + Srsingd (see Fig. 8.1), 1’ = Xp’ cos @ + fp’ sind + iz’, and xp (ikelr - rl) = exp like = tky(e- sry] kor] rFar-Infrared Generation by Diflerence-Frequency Mixing i for the far field. Equation (8.6) then becomes** (kz reat) 2 @ G eo ih E(r, w)) = = poe f dale ike ~ 08 9+ Ak sk) r if xf “def 2H ate thr0' os 8 sing (8.8) 0 0 Bmw} yey eer 29 (sina) 742 fo" - ae 7 (S22 favor | | with koe a= eat + a - oso}, Bek,asng, Ak =k), — ky, 2 and J, being Bessel’s function. We now have wean? = [S]ncaear eee (Se)[MEP]. 9 P « Integration of cye(w,) {E(, @,)|*/27 over the detector surface (Fig. 8.1) yields the total far-infrared power P(,) collected by the detector as P(w,) = free. ©)|?2ar? sing dg. (8.10) This result is physically understandable. The [2/,( 8)/B}° term arises from diffraction from a circular aperture as usual, and the (sin a, ‘/ay? term describes ym Linear Media —, Oatector Nonlinear Crystal 4 ig. 8.1 Schematic for calculations of power output. The laser beams produce a nonlinear polarization in the crystal at the difference frequency; the polarization is then treated as a source for the difference-frequency generation.nz Differonce-Frequency Generation the phase-matching condition. In the limit of 4, >> 1 so that the diffraction effect is expected to be negligible, 2J( B)/B is appreciable only for ¢ s 1/k24, and then (sina/a)? reduces approximately to the usual phase-matching factor [ sin (AKI/2)/(AKE/DD. Also, for k, > 1, if the detector js large enough so that bua 2 1/k 2a, we have fe™™(2.4(8)/87 sin db = 2/k}a*. The output power P(w,) calculated from (8-10) can then be shown to have an expression exactly the same as that in (8.1) derived from the infinite plane-wave ap- proximation. The theory here properly takes into account the diffraction effect. Equations (8.9) and (8.10) can in fact be used for an order-of-magnitude estimate of the far-infrared output. In the long-wavelength limit, the output approaches the oy dependence on the frequency as one would expect from the dipole radiation theory. This suggests that the efficiency of difference-frequency generation should decrease drastically toward longer wavelengths in the far-IR region. Fyen so, with the commonly available lasers, the far-IR output from difference-frequency generation still can be much more intense than a blackbody radiation source. A number of simplifying approximations have been employed in the deriva- tion leading to (8.9). It is possible to use a more realistic expression. for P(r’, w,) in (8.6) and evaluate the integral numerically to yield a better result. However, the assumption that the nonlinear medium is immersed in an index-matched linear medium is fairly ideal and is usually a poor approxima- tion. In practice, a nonlinear crystal in air has a very different refractive index at far-IR wavelengths than that of the air. Consequently, reflections of far-IR waves at the boundary surfaces are very important. In treating waves at the boundaries of the nonlinear crystal, one cannot use the far-field approxima- tion. This makes the foregoing theoretical approach inappropriate for dealing with the boundary effects. In order to properly take into ‘account the boundary effects, one should decompose the spatially dependent far-IR field into spatial Fourier components and impose the boundary conditions on each component separately. The calculation naturally becomes tmuch more complicated, and numerical solution is often necessary for elucidation. We discuss here only some of the physical results and refer the readers to the literature for the details of the calculation.» Since the far infrared refractive index of a solid is usually large (~ 5). reflection at a solid-air boundary can be high. Even multiple reflections can be significant, and in a crystal slab they give rise to a Fabry-Perot factor to each Fourier component in the output. The long wavelength of the far-IR field makes the phase-matching angle less critical, so that phase matching can be satisfied approximately by far-IR output over a fairly broad cone. This cone is substantially broadened outside the crystal through refraction at the boundary. Part of the far-IR. radiation may not even be able to get out of the crystal because of total reflection. Focusing of the pump beams generally helps, but absorption hurts the far-IR output as expected. The output field in (8.6) incorporated with an average transmission coefficient can in fact be a veryFar-Infrared Generation by Ultrashort Pulses M3 good approximation if the realistic P(r’, w,) is used in the calculation. Equation (8.1) obtained from the infinite plane-wave approximation, however, gives a poor description of the far-IR generation. Experimentally, far-IR generation by difference-frequency mixing has been observed in numerous cases,** with output frequencies ranging from 1 to several hundred inverse centimeters. For example, in LINbO;, xf}y(«@ = ©, — @))= 4.5 X 107% esu for w, ~ #, around the ruby laser frequency. If the pump laser beams are of 1 M'W each over an area of 0.2 om’, a far-IR power of ~ 3 mW at 10 cm’ is expected from (8.10) to be generated from a LiNbO, crystal 0.05 cm thick under the phase-matching condition. In a real experi- meat, 1 mW at 8.1 cm~! was detected from a crystal of 0.047 cm.* Discretely tunable CW far-IR output of 10-? W has also been observed from mixing of two CO, lasers (25 W) in GaAs.” Tunable far-IR radiation can also be generated by stimulated polariton scattering and by spin-flip Raman transi- tions. We postpone their discussion to Chapter 10. 83 FAR-INFRARED GENERATION BY ULTRASHORT PULSES The discussion in previous sections on infrared generation by optical mixing applies to cases where the pump beams are quasi-monochromatic. The two pump pulses are assumed to be sufficiently long, and the spectral purity of the infrared output, generally correlated with the laser spectral widths, is limited by the pulsewidth. Here, however, we consider the case of far-IR generation by a single short laser pulse." If the laser pulsewidth is as short as 1 psec, the corresponding spectral width should be at least 15 cm~'. Then, in a nonlinear crystal, the various spectral components of the pulse can beat with one another and generate far-IR radiation up to the submillimeter range. One might consider this an optical rectification process in which a dc picosecond pulse is generated. However, unlike the case discussed in Section 5.1, we are here interested only in the radiative component of the rectified ficid. This generation of the radiative output is subject to the influences of phase matching, diffrac- tion, boundary reflection, radiation efficiency, and so on./* Fay-IR generation by ultrashort pulses is, as usual, governed by the wave equation 1 a 4a a? vy x(v XE) + Sek = - =P? 2). 3 (vx) e? ar? co? ae? (1) (1) Given Pfr, 4), (8.11) with appropriate boundary conditions can, at least in principle, be solved. The far-IR output and its power spectrum can then be calculated. The solution of far-IR generation by a single short pulse in a thin stab of nonlinear crystal has actually been obtained through the Fouriers! (Cg Joy sayV) “Kousys UoReTpes afodip 37) SOAS D BAIN “O = 7 7B BuTqOREUT aseqd 303 jooye Forgorent-aseyd oy saat q aainZ ~esind ynduy paynoas ay JO wonmjoads sanod ay} soart B aA “(27) UI wunsads yndyno peresyUI-rey 9qP 0} STONNAITTOS Jofeur sag} om Fupmoys soar (q} “qeys 2m) Jo Anouroed yorag-A1ge,j 3q) Wo.y sostre adojaaue paysep oy) Japan amo pros oil ‘sTxE-9 9q) Boe pazeyod st agmnd s2se] ou pue ‘quis am Jo saoeyms aueyd amp on ppqrezed smce-o ov WIM pormayio st yessK10 aqL “qeys “OGNTT UHU-T v Go WapwouT AleuTION agind sasey pn sasd-z ¥ Joy umuyoads indyno payesyurrey poops SL (>) ve Bu (a) (B} ua) o> Apvanbard BRIDE (-w2) fousebssg 31 s 9 (squn Goma! volinglaisip 1044998 § 4Far-Infrared Generation by Ultrashort Pulses 1s transform of Er, ¢) and P(r, ¢), neglecting the dispersion of e and x© in Pr, 1) = x2: Er, NE*(r, £).1° We present here a physical description of the solution, Figure 8.2a shows the calculated power spectrum of far-IR radiation generated by a 2-psec Nd laser pulse from a 1-mm LiNbO, slab. First, the Fabry-Perot geometry of the slab gives rise to the interference pattern under the dotted envelope. Then the dotted envelope of the spectrum is basically the product of three contributions seen in Fig, 8.25: curve @ represents the power spectrum of the rectified input pulse, curve b describes the w* dependence of the radiation efficiency with a much sharper low-frequency cutoff as w > 0 due - to diffraction; curve c is the phase-matching curve with its phase-matching peak at « = 0 for the particular crystal orientation with the é-axis along the face of the slab. Thus, the calculated power spectrum in Fig. 8.2a can be physically understood. Such theoretical calculation actually gives a very good description of the experimental observation. Figure 8.34 shows a comparison ‘between theory and experiment for far-IR generation from a 0.775-mm LiNbO, with the é-axis in the slab face by a train of normally incident mode-locked Nd/glass laser pulses.* The Fabry—Perot pattern is absent here because the spectrum has been averaged over the actual instrument resolution. By orienting the crystal to INTENSITY o 2 4 6 6 © @ 4 6 2 4 6 6 OM 2 4 6 cme! em! fa) te} Fig. 83 (q) Far-infrared spectrum generated by mode-locked pulses in LiNbO, phase matched at zero frequency. The experimental points were obtained from the Michelson interferogram and the solid curve from the theoretical calculation assuming Gaussian laser pulses with a 1.8 psec pulsewidth. (>) Far-infrared spectrum generated by mode-locked pulses in LINDO, oriented to have forward and backward phase matching at 13.5 and 6.7 cm}, respectively. The expeximental points were obtained from the Michelson interferogram. The solid and dashed curves were calculated by assuming Gaussian laser pulses with a pulse-width of 2.3 and 1.8 psec, respectively. (After Ref. 8)16 Difference-Frequency Generation achieve phase matching at finite w, one expects, from the above discussion, that a single phase-matching peak at w * 0 may dominate the output spec- trum. An example is seen in Fig. 8.35. Again, theory and experiment agree well. The two peaks in the theoretical curves correspond to phase-matched generation of the far-IR radiation in the forward and backward directions, respectively. This figure suggests that we can have tunable far-IR output by simply rotating the nonlinear crystal. As shown, the pulse stil] has a fairly broad linewidth, indicating that it is also a pulse of picosecond duration. Nevertheless, since the output linewidth is appreciably narrower than the laser linewidth, the output pulse must be appreciably longer than the input pulse, That the output is still significant after the input has more or less decayed away is an interesting fact considering that the medium response to the input pulse is essentially instantaneous in this case. With an input peak power of 0.2 GW over a cross section of 1 cm?, a far-IR output of 200 W peak power has been detected from a 0.78-mm LiNbO; crystal.* REFERENCES 1 RL. Byer, in ¥. R. Shen, ed. Nonéinear Infrared Generation (Springer-Verlag, Berlin, 1977). 2 F Zemike, in C. L, Tang, ed. Methods of Experimental Physics, Vol. XV: Quantum: Electronics, Part B (Academic Press, New York, 1979), p. 143. 4 See, for example, J. D. Jackson, Classical Electrodynamics, 2nd ed. (Wiley, New York, 1975), Chapter 9. D. W. Faries, K. A. Gebring, P. L. Richards, and ¥. R. Shen, Phys. Rev. 180, 363 (1969). Y.R Shen, Prog. Quant, Electron. 4, 207 (1976). J, R. Morris and Y. R. Shen, Pays. Rev. A15, 1143 (1977). B. Lax, R. L. Aggarwal, and G. Favrot, Appl. Phys. Lett, 23, 679 (1973); B. Lax and R. L. Aggarwal, J. Opt. Soc. Am. 64, 533 (1974). 8 K.H. Yang, P. L. Richards, and Y. R. Shen, Appl. Phys. Lett. 19, 285 (1974). 9 T, Yajima and N. Takeuchi, Zap. J. Appl. Phys. 10, 907 (1971). 10 J. R. Morris and ¥, R. Shen, Optics Comm. 3, 81 (1971). aA ek BIBLIOGRAPHY Shen, Y¥. R., Prog. Quant. Electron, 4, 207 (1976). Shen, Y. R.,ed., Noalinear Infrared Generation (Springer-Verlag, Berlin, 1977). Warmer, J., in H, Rabin and C. L. Tang, ed8. Quantum Electronics (Academic Press, New York, 1973), vol. 1, p. 703.9 Parametric Amplification and Oscillation The three-wave interaction discussed in previous chapters is manifested by energy flow from the two lower-frequency fields to the sum-frequency field or vice-versa. The latter happens in difference-frequency generation, which, in general, can be initiated with a single pump beam at the sum frequency. Difference-frequency generation can then be considered as the inverse process of sum-frequency generation, and is generally known as a parametric conver- sion process. Parametric amplification and oscillation in the radio frequency and microwave range were developed before the laser was invented.' The same process was expected in the optical region, and was actually demonstrated in 1965.7 It has since become an important effect because it allows the construc- tion of widely tunable coherent infrared sources through the controllable decomposition of the pump frequency. In this chapter we explore the theory of parametric fluorescence, amplification, and oscillation together with some practical considerations. 9.1 PARAMETRIC AMPLIFICATION As am inverse process of sum-frequency generation, the general theory of parametric amplification is the same as that for difference-frequency genera- tion. In fact, the only difference of the two processes is in the input conditions. Even there, the difference is not clear-cut, but we normally consider parametric amplification as a process initiated by a single pump beam while difference- frequency generation is initiated by two pump beams of more or less compara- ble intensities. The difference disappears after a significant fraction of the pump energy has been transferred to the two lower frequency fields. Thus the theoretical description of parametric amplification with infinite plane waves again starts from the set of three coupled wave equations (3.4). In the slowly 7Ae Parametric Amplification and Oscillation varying amplitude approximation with E(w,) = &{zexpli(k tr — of + ol and a plane boundary at 2 = 6, they become (see Sections 3.3 and 6.7) 8 ia} : P 2g = Ker! Ak 24190, ek, iw het = Ree Ak rill, (9.1) # a iw} wihke-d Ron Fy Reet Aka io where 27. a8 Ko ak (cay = wy + wy): 22, and Ak = ka, — ki, — Rar and & = $ — $1 — 2 js the initial phase difference of the fields at z = 0." We assume here & = —7/2. The solution of (9.1) with Ak = 0 has been discussed in previous chapters in connection with sum- and difference-frequency genera- tion. In parametric amplification, E(w) is known as the pump wave, E(w,) for £(o,)) the signal wave, and E(w,) [or E(w] the idler wave. We consider here first the case of negligible pump depletion with Ak + 0. The assumption of negligible pump depletion means that & can be tegarded as a constant. Equation (9.1) then reduces to a set of two linearly coupled equations between &; and &}. Writing , = Cye™ and 6 = Ce, we find immediately y, = —¥i + 44 and 2 aw Ms ek " =0. (9.2) -w Kes, (iy - iSk) 2s This leads to the solution nyo HAS ig), . [ 2 —(any* ae 4uied \ org? B= (80 1 80 ke: 3° a= [c..e0™* + C.F | eta sake ére [quiet + C_ett/Oe] eo AMBRE (93) i | Pa k : af2 Cs" ed [uaee 24 (0)-2(] 8} 2. Ar ~ 2g -stle ALA Gs =A lea enero v2 (Be) sof (0Parametric Amplification 19 This solution shows the following physical properties. If K' 6; is small so that ge < (Ak), then g is purely imaginary. If Ké; is sufficiently large so that gg > (Ak), then g is real and positive, and at large gz, both #, and & grow exponentially with 2. Thus g, = (Ak) is the threshold for parametric amplifica- tion, The parametric gain is clearly a maximum with g = go at phase matching, Ak = 0. Introduction of attenuation coefficients at «, and ow, in the above formalism is straightforward. As expected, they increase the threshold and decrease the gain. As an example, consider parametric amplification in LiNbO, with xG} = 2.7 % 107 esu at Ay ~ Az = 1.06 wm with m = 2 = 2.23. The maximum gain is found to be go = 0.9 x 10-78, em}. For a pump field of & = 100 esu corresponding to 2.5 MW. ‘/ont, the gain is 0.9 cm™ 1 Thus the overall exponential gain g/ even in a crystal of length / = 5 cm is not very large. To achieve an overall gain of gy! ~ 40, we must either use a pump beam of much higher intensity (which is attainable only with picosecond pulses if optical damage to the crystal is to be avoided) or use an optical cavity to increase the effective length. In the latter case, the system may become an oscillator, as will be discussed later. As noted in (9.3), the phase mismatch Ak suppresses the gain very effec- tively. Therefore, in the limit of high conversion efficiency, we need only consider the phase-matching case although the general solution of (9.1) with Ak + 0 has been worked out by Armstrong ¢t al.4 Following the notations and derivation in Section 6.7, we find, assuming @) = —a/2 in 9.3), oud = 2[ viens — WB)oms — oT (9.4) which has the solution [assuming uz(0) < 13)] 33(5) = (13 (0) — u?(0))sn” x [9 $uk(O) f+ sa? aaa , | , | oS (0 u3() = uh{(0) + u3(0) ~ w3(E) 95) ud(f) = WF(0) - 43(0) +2) 2 80) - dO) ub (0) + 43 (0) In the case of 2, (0) = 0, this result can be shown to reduce to (8.3) derived for difference-frequency generation.168 Parametric Amplification and Oscillation 9.2 DOUBLY RESONANT PARAMETRIC OSCILLATOR ‘As mentioned earlier, an optical cavity can be used to increase the overall gain in parametric amplification. Then, parametric oscillation can also occur. That tunable output is possible in the absence of input makes the parametric oscillator a practically more useful device than the parametric amplifier. As shown in Fig. 9.1, a parametric oscillator is composed of a nonlinear optical crystal sitting in an optical cavity. For simplicity, we assume that the cavity is formed by two plane parallel mirrors. Two types of cavity commonly are used. The doubly resonant cavity has the mirrors strongly reflecting at both , and w,, and the singly resonant cavity has the mirrors strongly reflecting at cither w, or o,. Usually, both mirrors are transparent to the pump wave. Tf the single-pass parametric gain is small, the pump intensity can be regarded as independent of the distance in the cavity. Let us consider the doubly resonant case first.>-* The fields in the cavity can be written as E(w,) = 24; (¢)sin kyze", E(w.) = 26,(¢)sin k,2e7'°™", (9.6) E( @) = syeitst iat with @ = yo, & 2 yp, and w, = w +: = Oro + wp + dw. The cavity condition requires _ Bip _ mart ® ges _ Wyo _ mat + OD, ~ ~} and kz (9.7) where m, and m, are positive integers, Tis the cavity length (we assume here, for simplicity, that the crystal length is equal to the cavity length), and 2®, and 2, are round-trip phase shifts at w, and @, due to boundary reflections and refractions, The coupled wave equations in this case become nd a # 4awt [a + a (na + =) E(w,) = + KEM (er EC), a(n, 0, & 4rw} [xz (ngee Galfer eae wronBled. 68 Noniinear crystal Pump beam wo — = ( —>" 03 a N o R R Fig, 9.1 Schematic diagram of a parametric oscillator.Doubly Resonant Parametric Oscillator 12 Here, I, and I, are the damping constants at and w,. They constitute the attenuation loss due to absorption and scattering in the cavity plus the mirror transmission loss, With an intensity attenuation coefficient of a,(a,) per unit length and two mirrors of equal reflectivity Ro,), FT; is obtained from the definition en thnl/e Rig tet or r= £[,- Fi Ri] met (9.9) = [a +70-%)| ifR,=1. We assume T, =. With the slowly varying amplitude approximation, |a°6,/81?| « ja, dE,/ at), (9.8) reduces to 2 [F + 5t| &,(t)sin kz = tou( =) Kés(1)8;sin kyze", (9.10a) a1 Nes . <¥ . hot £4 drag o)sin tgs = ~teny 4) Ké,(1) sin kyze" (9.100) 2 where 2a, 3a = Bay xa, = int 8) 28 and we treat &, as a constant. Multiplying (9.10a) and (9.10b) by sin Ayz and sin kz, respectively, and integrating the equations from z = O10 z =! we find jz + sao = eyo a } Kees) etki, % sin( OE 1/2) sei : 1) [f+ trate = -Henl 5) KAKO era The solution of this set of linearly coupled equations can be written in the form &(t) = (D,,e°*" + Dy_et emer, EF(1) = (De + D,_e?-el/2her, a l[— s,=4[-T + @], (9.12) = [ag - (au), G3 - pol" erga sin? (Ak 1/2) ning” (AkL/2y°12 Parametric Amplification and Oscillation where D,, and D, , are coefficients to be determined from (9.11) together with the initial values é,(0) and 4;+(0). In the discussion of parametric oscillation, however, it is not possible to find D, , and D, ,. We are more interested in the threshold condition for oscillation. Equation (9.12) shows that oscillation starts when G = G,, =I or the threshold pump intensity for parametric oscillation is nye T= 35 (a (Aw)? + 1? (013) 2162 joy? Pa] sin? (AK 1/2) ningns (Ak 1/2 Clearly, the threshold is a minimum for &u = 0 and Ak = 0. Thus if we use the same LiNbO, crystal described in the example of parametric amplification and insert it between two plane mirrors of R, = R, = 0.98 separated by the crystal length / = 5 om, we find a minimum pump threshold of (lain = 2-5 KW/emn? assuming that the attenuation loss a, is negligible. This is a fairly low intensity and is achievable even with a CW laser. The doubly resonant CW parametric oscillator was first demonstrated by Smith et al. in 1968.° The solution in (9.12) also shows that the output frequencies of the signal and idler waves are @, = ey + Faw and wa, = dy + Law (9.14) with @p and t9 being the cavity resonant frequencies given by (9.7) and Aw = w; — w19 — 49 18 automatically minimized to achieve the lowest pump threshold possible. Tuning of the output frequencies is discussed in Section 9.3. One serious disadvantage of the doubly resonant parametric oscillator is that it has jJow stability.’® Let us assume that originally the oscillator with a cavity length / operates at a, = Wg = mywe/Myl, w= Wy = myMC/Mal, and w, + W = w,. A small change of the cavity length A/ due to external perturbation shifts the output frequencies to w; = wig = (my + Am)re/n,(? + Af), w= thy = (nt, — Am) ae/nz(1 + AD, and w+ 0) = @;. We then find sae) /(\( Bom Am = «| r Fl ny The shift in the output frequencies due to A/ is therefore given by 8a = a] — 0, = &, ~ 0 ny Al a, 83 7 PLL a) -% (9.15) ny Al =——— 4,7, - ~ 3 mom OTDoubly Resonant Parametric Oscillator 13 Thus a change of Al/i = 1077, which shifts the frequency of a cavity mode by only 1077, will cause the output frequencies of the doubly resonant parametric oscillator to change by more than 107s, if jm, — "y|/z < 107%, This shows that the output of the oscillator will be very unstable, subject to external vibration and thermal fluctuations. For steady-state operation, an oscillator must have its gain clamped to the threshold value, since otherwise the output would continuously grow in time or decay to zero. This makes the calculation of conversion efficiency fairly simple.’ The pump field in the cavity is self-adjusted through pump depletion by parametric conversion to yield an oscillator gain clamped to the threshold. The signal and idler fields increase with the increasing pump energy, but being standing waves, their amplitudes are constant across the cavity. The part of the pump power coupled into the signal and idler fields should show up as power loss in the cavity and in the signal and idler output through the cavity mirrors. Let us assume the phase-matching case, Ak = 0, In the slowly varying ampli- tude approximation, the equation for é reduces to that in (9.1), With 6 = 1/2, the forward propagating pump field at z = is aye Ay #3 (p= 650) -[ 2] xe ey. (9.16) ‘Then the generated fields at w, and @, in the cavity can also react back and generate a backward propagating wave at o, with 8; (0)= Ke, &i. (9.17) Energy conservation requires a,{ 16 (0) — 16 (DF — 16s O)?} = mle? ~ eT) tagl@P7(b = eT) (9.18) = (aflaPT, + APT )2I/e That the number of photons generated at w, and , must be the same leads to 2 2 2, 2 mele? Ty _ nglél*T °, oy (9.19) Combining (9.16) to (9.19) gives 730/61? = nil 6? won (Ol? — oynsl€F (ODP 0 ef.1 1 (9.20) “oll yw ¥m4 Parametric Amplification and ‘Oscillation where w,0,c* v= [eeeeror|/nn Tf we take w, = @19, 2 = #202 and I = [;, and use the expression of the gain Gy in (9.12) for Ak = 0, we have N= G2/T? = G3/G§, Since Gi is propor- tional to the pump intensity, N has the physical meaning of how many times the pump intensity is above the threshold value, with N = 1 corresponding to the threshold. The total power output (signal and idler, including attenuation Joss in the cavity) is £) [egg Paty + age 2E! ale eee P+ P= (3 The overall conversion efficiency js then given by B+P,_ n= 2 WON -). (9.21) 3 which can teach 50% for N = 4. 93 SINGLY RESONANT PARAMETRIC OSCILLATOR Because of its intrinsic instability, doubly resonant parametric oscillators are seldom used in practice even though they have lower oscillation thresholds. The instability can be climinated by using a singly resonant cavity.) Let the mirrors be transpasent at o, and #5, and highly reflective only at #2. Then we can describe the three fields as E(w) = 4(z, pete iav, E(w,) = 26, (#)sin kaze *™', (9.22) E(ta5) = by{z er with wag = myte/ny! and w, + Wg = 43: Clearly, in this case, a small frac- tional change of A//I induces only a shift of 8@ = Gyg(Ai/1) in the output frequencies. To find the threshold of oscillation, we can again solve the coupled wave equations for E(w,) and E(w,). We shall, however, take a somewhat different approach here. We can start ‘with the solution of parametric amplification in (9.3) and impose the condition that for parametric oscillation, a single-pass gain must be equal to the round-trip loss. Assume the phase-matching caseSingly Resonant Parametric Oscillator 125 Ak =0. The round-trip attenuations of the two fields are eTunl/e and e-T2"!/*, yespectively. Then from (9.3) we have 1A €,(0) = eta 0}onh Bunt + {| &F(0)sinh} rull, i : (9.23) Es ad G3 (0) = e Temi“ [eriovcsntew - (2) 4,(Osahtsa!| 2 and hence n,\t2 coshd gy ~ efi/e, (2) sinhtgqé 4 = ne =0. (9.24) - {=| sinh}g,/, coshigy! + eh "“ This leads to the threshold condition for oscillation 1 Lt effin t Tinie cosh 3g yl = elimd/e g laml/e (9.25) (etre — 1)(e!7¢ = 1) =l+ eliml/e y glatale The result here is quite general and is applicable also to the doubly resonant case, where }2,,/, Tymi/c, and P,n//c are ail much smaller than 1. Equation (9.25) reduces to gi = 4n,2,T,0,/c’, which can be shown to be the same as the threshold condition (G,, = I’) derived in the previous section. Now, for the singly resonant oscillator, only }g,,/ and T,n,//c are much smaller than 1, but exp(T\n,//c) > 1 because of the large transmission loss of the mirrors. We then find the threshold condition to be og letin’ = I nplve 1 + eliniee BE pty! =e. gal? (9.26) In comparison with the doubly resonant case, the square gain threshold ratio, which is equal to the pump threshold ratio, is (23) soaly _ (Za, )sioaly (8%) boubiy Cin) deabiy 2 1-R," (9.27)126 Parametric Amplification and Oscillatioa For R, = 0.98, this tatio indicates that the pump threshold for the singly resonant oscillator is 100 times higher than that for the doubly resonant oscillator. Thus singly resonant parametric oscillators usually require pulsed lasers as pump sources, The conversion efficiency can again be calculated from energy or photon number consideration, knowing that the oscillator gain must be clamped to the threshold in the steady-state operation. In this case, é, is a constant, and EH) and &,(/) are obtained from the coupled wave equations in (9.1) AF, pos ie 7, KEE, and (9.28) a6, wx 37 ~ 7, KO: The backward waves at o, and w, are negligible. With &(0) = 0, we find recor = (SE 6 (Or sitar and (9.29) 18, (8)2 ~ 14s (0) Peostet where wee? te 2g 12 B= SOE RAE The photon number generated at #, ina single-pass gain must be equal to the photon number at w generated and then lost in a round trip around the cavity, so that Qni/al&? _ lee @ 4 (9.30) ~ (2 ies onrsine 3 The last equality leads to the relation sin'Bl 1 == 9.34 Be We (ot) where Ne= (oye? /ryma) K NE (0)P - & 8(T2,//c) SuFrequency Tuning of Parametric Oscillators ni again has the physical meaning of how many times the pump intensity is above the threshold value. The conversion efficiency is then given by _ mle? + (20 gmgl/e}itgl6o)? (9.32) 114,(0)7 = sin’ pf with sin’f/ determined from (9.31). For W* = 4, we find 7 = 90%. This is of course somewhat ideal since we have used the steady-state plane wave ap- proximation in the derivation. A more realistic calculation with Gaussian beam profiles has been worked out by Bjorkhobm." The output conversion efficien- cies at w, and «,, defined as output versus input, are ® Tout (@2) = (Ss. (ca) = dak 2] 633) Four \@2 Tyrgl/c \ ay n An overall output conversion efficiency, toy(@1) + Aou(@2), as high as 70% has been experimentally demonstrated.!? 94 FREQUENCY TUNING GF PARAMETRIC OSCILLATORS The output frequencies of a parametric oscillator are determined by energy and mornentum conservation =o, +0, and ky =k, +k. Together they yield the relation 05 [3(@) — (4 — @,)] = a [m,(o,) + alos — 0). (9.34) Equation (9.34) dictates the signal frequency «, if the dispersions of the refractive indices n,(«,) are known. As discussed earlier in sum-frequency generation, (9.34) can be satisfied only in anisotropic crystals. With negative uniaxial crystals in the normal. dispersion region, n;(w,) must be extraor- dinary, while m,(«,) and 2,(«,) can be either both ordinary (type I phase matching) or one ordinary and one extraordinary (type II phase matching).128 Parametric Amplification and Oscillation Equation (9.34) shows that the output frequencies of the oscillator (or the frequencies for maximum parametric gain) can be tuned if n(@) can be varied by external parameters. We consider here frequency tuning by angular rotation of the crystal and by temperature. We assume type I phase matching in the following discussion. In angle tuning, let the output frequencies be w, and w, when the crystal is oriented with an angle @ between its @ axis and the axis of the cavity. We have c2gM§ ung, 8) = wy? (wy) + pnd (2). (9.35) If the crystal is now rotated to + A®, the output frequencies should corre- spondingly change to w, + Sw and w, — Aw. Equation (9.35) becomes e oO wn] ntles.) + 2 aa] =(, + te)[rt(on) + Fate + ore] +(@, - se) ne) - Fae + o(au")| (9.36) or ans an? and Aw = oth ae/ {lage - wg| + n¥(0) — nia} (937) where O( Aw?) are terms of orders higher than or equal to (Aw). For uniaxial crystals, one finds # =- (x5) snaal(] -(3)| (9.38) Then, given the dispersions of n(@,), the angular tuning curve @, (or #;) versus @ can be calculated from (9.37) and (9.38). As @, approaches 0/2, however, the quadratic terms of Aw in (9.36) become nonnegligible. Keeping the (Aw)? terms in (9.36) near the degenerate operating point 6 = d and w, = Fu, we find, instead of (9.37), 1/2 = [oy 283 an, On la ao (os( 3), / 258 +22) aoe, 6%)Frequency Tuning of Parametric Oscillators 129 ANGLE TUNED LiNbOs PARAMETRIC OSCILLATOR Te25 °C WAVELENGTH [ut BANDWIOTH (em-!) Sem CRYSTAL LENGTH tt soe 49° 48° 8 647* 46° ase 44" CRYSTAL ANGLE Fig. 9.2 Tuning range and gain bandwidth for the angle-tuned, singly resonant, LiNbO, parametric oscillator. The pump wavelength is 1.06 wm. (After Ref. 14.} As examples, we show in Figs. 9.2 and 9.3 the angular tuning curves of LiNbO, pumped by a 1.06-ym laser beam and ADP pumped by a 0.347-um beam. The tuning is of the order of 1000 cm”! per degree of rotation away from the degenerate point. Temperature tuning follows a similar treatment. Assume at temperature T wayn§ (Way, T) = nC, P) + 209 (u, T). (9.40) At T + AT, the output frequencies change to w, + Aw and w, — Aw, and we have, away from the degenerate operating point, Ont oO a oO «| ni(0,,T) + ar = (w+ a1) ro, r)+ ohare gl an’ ang + (cy — be) rb. 7) + Sear- Fae] co( BS /8T) — w,(On0/9T) — &,(dn8/8T} no — 9 + w,( dnd /du) — w{ On$/de) (9.41) Aw =130 Parametric Amplification and Oscillation Wavelength in Angstroms 7900 10 15000 Rotation A @ in Degrees 20 1S 10 Photon Energy in ev & 4 2 o “2 “4 “8 Fractionol Frequency Shift A Fig. 93 Tuning curve for the angle-tuned ADP parametric oscillator. The pump wavelength is 0.347 pm. The angle A@ is measured with respect to the angle at which the signal and idler frequencies are degenerate. The dotted curve is a theoretical curve. {After D. Magde and H. Mahr, Pays. Rev. Lett. 18, 905 (1967). Around the degenerate operating point, T = 7 and #, = 5, /2, we find 1/2 _f famy 2) am, , 03 / an a so ~ fos] ar Or wf Fa + 7 Va? Hees cary”. (9.42) The temperature tuning curves of LiNbO, for a number of pump laser frequencies are shown in Fig, 9.4 as an example. With a pump wavelength at 0.53 ym, the degenerate point is at 7, = 49.3°C and tuning is 300 em™1/(°C). While Ak = k, — k, — &, = 0 determines the output frequencies, Ak? = 24 determines the gain bandwidth of a parametric oscillator. Since vf) -o( 23) ty — wy baa] ~ 22157, pee NN Ba, 7h deg AkE= u/c (9.43)Frequency Tuning of Parametric Ostillators BI 2000 5000 4000 Le 3000, (em) (am) WAVENUMBER WAVELENGTH 24735 18,000/- SIGNAL = 550 20,000/;- — 500 1 1 1 L 1 200 250 300 350 406 450 TEMPERATURE {°C} Fig. 9.4 Tuming curves for the temperature-tuned LiNbO, parametric oscillator at various pump frequencies. [After R. L. Byer, in H. Rabin and C. L. Tang, eds., Quantum Electronics (Academic Press, NY, 1975), vol. 1, p. 631.] the bandwidth is given by an. 80 = 2c f ny — na oy( 5a \- al 5: Around the degenerate operating point, a better approximation gives 1a = (2 fan, al an bo ={ i [|i >(Z)|_.) (9.45) The result here shows that near the degenerate point, the bandwidth can be fairty large, ~ 100 cm™! for / = 1 om. Away from the degenerate point, 8w is (9.44)—_- i we (6) Tet er 727 syg “}4dy Speussng °§ ‘S pure ‘OOH OOS Tf wa €y69'0 ‘gomng “gH pur uosumor Dad IT NOZ-99 9-01 ME fOQNTI Aga (pLev) 02s ‘Sz nay stag addy kg “1 (su ¢T-) Ww O0'L poe ‘Sunmayy ‘NW SQ9H “TU wnt yp PT 0b MW E-T0 fOQNtT DVA:PN wt 6690 ‘6LG 0 ‘ZES'O “Tir O SPBuapPAwA (LET) L6 ‘LI HOF (su 002-) STUOULTeY puodIIs 18 shyg dy SETA MA WIT SOTE-SEO asp AA OT-T0 faqnTI — 3upezado OWA: PN. (LL6D ae wit 997'0 sHOWIeR ‘gt nay sty ad y ‘Kassel (sa 2) ormoUTEY ypo 1 “yD pus Yano1oge A Wf urd €1'0-26'0 ESE ANT O01 aay = @apetsdo NVA! PN. L961 ‘BApjOCE ‘sondo e asapoyy “dinAg 32 paytasaly a ‘suysTeysid SW Pue UTBUAOD Tv pastry Hd ‘sOPOUN “AW ‘aeaped “AA saasungy) 'N ‘© ‘AoueURpAY "W'S (9961) (Guand ca! s¢0) Tee CMa ALL OITA It TT 96° ord e'0 PTE ZES'0 “ACU pur ‘scoped “AA PHEW 3SO-BP'O ve souOWey ‘seyeysig 'S "v AOSOTOY "W'A (Gund mrt Ze¢°0) (sw 92) pany pae pucoas ‘agmoy Tv sourmyY VW WT LTI-LS60 BE AML OOT dcx pure see = PN soUQa]Iy aguey sun Aousw UPL OSIM PUB RUSTE, Jase] dumg uoIsisauon) «= FaMOg INdiNg —-IWIUTTVON. SUOWBITLISO yowereg SabepussaMday T6eL{e146 Opp ‘te NAT “sty pddy “png ‘2 ‘A pure ‘soe seq] “q POUR “DC (ZL6D) oe “Oe WeT sty jddy “pS “7M poe UN NH ‘sae uN] g wUURH DC Gis GRE ‘pe naT sds id “Biaqp[0O *s “T PUE SSM “VE (Ze61) E15 ‘SET LAL ‘AOUIRUS "A “A Put 'Ad,PARS ‘CW ‘AOIOJORL “WW “TASASEN YY "y T ‘AopAaed “VW {cL61) RT AE 897 “syd dy ‘s9kg 71-8 PUR wa; “TU (ELOU) SL CL HAT “8h ddy ‘Greg ‘poe MEN ' (26D 6088 “uossapy uO “p 2gg 7 oydures “gy (Lb 9€ GL WaT Shug ‘iddy ‘Suey, “1D pre oppure ‘f-y COLET) 696 °LT “77 “skys jddy “8229p OD 'S “I (le6 E16 ‘CL WT dat OPIN Ad ursuaoy "FW Pur COL6L) LEE ‘TT ‘NTT ALL SPIN “A ‘d PUR “araDAOY “I "W ‘OMA[IZ] IV wn e-aet | STO wi ¢g-1g made ey or wit ops md ¢¢ £50 mt 16-01 wt E77 %0P MAT Z-SIhO BB ud gt-TT HOT wd p-11'0 %I- wit p'z-89°0 %OT (su sz) MOOT fgsyFdy {su o0¢-) M008 SPD (ou 08) ANTS ®$PO (sa O01) AAT- 2spo (us) AOL ‘Onl (a st) AAT OT ots (su 02) MAE ‘orl (a 0z) ‘ort AAA OT- ‘olH-» wr S9OT FOMPD = PN we 18Z 398] TH wit 967 sa ae wt egy DOWA:PN wi LytO STIOULIEY puoces 78 Saneisdo Jase] Aqny urd £769°0 daser Aqny wm er69'0 Josey Aquy md gs omMomrey prosas ye Buqeiado SSUID : PN 1331m Parametric Amplification and Oscillation GRATING CRYSTAL GVEN ETALON SAPPHIRE 2 po Fig, 9.5 Singly resonant cavity design for an angle-tuned parametric oscillator. (After Ref, 14.) typically around 5-10 cm7!, A more dispersive crystal yields a narrower bandwidth. The output linewidth can be narrowed appreciably by using frequency selection elements in the cavity. Figure 9.5 shows an angle-tuned singly resonant LINDO, parametric oscillator designed by Byer.?:!4 The beam splitter transmits 90% of the pump beam and reflects 99% of the signal beam. The angular tuning curve of this oscillator is shown in Fig, 9.2 together with the gain bandwidth. If a 600-line/mm grating blazed at 1.8 wm is used as the back mirror of the cavity, the signal output has a linewidth around 1 cm‘! for a beam spot size of 1.6 mm. With a prism beam expander in the cavity to increase the beam spot size on the grating, the Yinewidth can be narrowed by anoiher order of magnitude. Alternatively, a 1-mim tilted etalon can be inserted in the cavity to reduce the linewidth to less than 0.1 cm™!, Other frequency- selective elements, such as multiclement birefringent filters and etalon assem- bly, can also be used in the cavity for reduction of the output linewidth. In Table 9.1 we reproduce the list prepared by Byer and Herbst,* describ- ing the operating characteristics of a number of representative parametric oscillators. 9.5 PARAMETRIC FLUORESCENCE Parametric oscillation occurs through amplification of noise photons initiated by parametric scattering or fluorescence. In general, in the parametric process,Parametric Fluorescence 135 a photon at w, is scattered into a photon at w, and a photon at w, with #, + @, = w, and k, +k, = k;. Parametric scattering or fluorescence refers to the parametric scattering process where the initial numbers of photons at #, and «, are zero. This nonlinear optical emission can be properly described only by quantizing the fields.'* In the parametric process with negligible pump depletion, the intense pump field can be treated as a constant classical field. The Hamiltonian governing the problem can be written as XD hu,(aya, +) + pkGylapaze * + ayaye"] (9.46) i=12 where a7 and a, are creation and annihilation operators for photons at w, Tespectively, and G, is given in (9.12). The Heisenberg equation of motion for an operator X is a7 Sin, #) (9.47) from which we obtain dat rT = iva} + i4Gya,e*" ta (9.48) a = —iw,a, — ibGyafe™™*. The above set of equations can be solved to yield af (t) = [at ()cosb}Gye + ia, (0)sinh} Gye] ef (9.49) 23(1) = [a,(0)cosh} Gyt — ia, (0)sinh} Gos] e7**". Then, [assuming {4,(0)) = {2,(0)> = 0] the average numbers of photons at w, and w, are, respectively, &m (2) = Cat (A}ar(4)) = (1,0) poosh? EGyt + (1 + (1 (0)))sinb?} Gor (mz (4)) = (at (1)a,(z)) = (nz {O))cost?$Gyt + (1 + ¢n,(0)) sink? } Gy. (9.50) The result here shows clearly that the number of photons at w, and w, can grow out of zero in the parametric process. For (7,(0)) = 0 and (”,(0)) = 0, we have (a(t) = Ca, (2) = sinh’ 4Gor. (9.51)136 Parametric Aanplification and Oscillation While parametric fluorescence Jeads to parametric oscillation, it can also provide initial photons as input of a parametric amplifier. If the single-pass gain of the amplifier is large, the output can be appreciable. This single-pass parametric amplification of noise photons in a nonlinear crystal often is known as parametric superfluorescence."** To find the output power, we note in (9.50) that the output at o, (or @,) actually grows out of the initial noise photon at , (or w,). We cam assume one photon per mode is created at w, (or w,) by parametric scattering at the input end, z = 0, and use (9.3) to calculate the output of the parametric amplifier at «, (or #,). For one photon in each mode, the corresponding input intensity at ©, is ha,c/n,V, and the output at w, atz = lis th 1,(w) = [ 2p porte (9.32) gV The number of modes in the frequency range @, to &, + de and in a solid angle extended by k, from y) to Wy + dip, (see Fig. 9.6) is, for small #2, _ dakjsin yy dey dy Br /V vy (9.53) _ Vann > = Feit dw. The total output power in a beam of cross section 4 between a, and w, — dw collected by a faraway detector with a small collection angle @ is then given by P(w,) do = fileda dN : nin,helgtl?A po, sink?igh (9.54) = (do) ae d . (do) lene? f 41 (ast? wh Nonlinear erystal Detector Pump at w —_—_—_ _—_> ae ——_. k, gt vA ve ky Fig. 9.6 Geometry of parametric fluorescence collected by a photodetector.Parametric Fluorescence 137 Here g? = g3 — (Ak)}* is a function of 4, through Ak. In the forward direction, Hy = 0, the phase-matching condition Ak = 0 is satisfied at w, = and @, = «a? such that woln, + wn, = 03,73. From Fig. 9.6, we find for small ¥, and ee ke? —Uky + Ak — hy) = 2&2, (1 — 008 f,) Bk = (ky ~ ky ~ ky) +(e) a (9.55) 2 = -adw + byt ah) { dk ‘ da jo (do, } 9)’ Aw = 0, — 08 = af - w, kik 2K,” where Equation (9.54) becomes int?! g? —(— ado + b¥2)'|71/2 P(o,)= rinohiail f sin led (Cone oT} a a) daddy, me {fed -(-2dw + 544)'] ya} (9.56) This result shows that P(w,) is a maximum at w, = wf, that is, when the waves at @), ty, and w, are phase matched in the forward direction. The half-power points appear at w, = w$ + [Aw,| and w, = w? — |Aw_|, where jAw,| and |Aw_| are given approximately by sint{| xf ~ (067 # alte )]'°1/2} sink got?) {[ 53 (00? + aldo 41) 71/2)" ( g0t/2)" 2592 ay, = glee(b8? Faldo, 7) ‘fesel =4, which yields, for go >» 607, 1p JAo,| = 4] (28¢102) + vel 1I/ag. 9 (9.57) |Aw_| = 3 (?m2) - veel a i138 Parametric Amplification and Oscillation Therefore, the bandwidth of parametric superfluorescence observed by the detector is Aogy = [de + [Aol 2 12 = 2(#2in2] . (9.58) A crude estimate using (9.56) yields, for gof = 50, and w, in the near infrared, a P(«,)/A as large as 10" W/cm? per cm~* in a 10-mrad forward cone. Even in a 5-cm LiNbO, crystal, go! = 50 requires a pump beam intensity of about 150 MW/cmr. This intensity, however, is easily obtained with mode-locked laser pulses without damaging the crystal. Indeed, parametric snperfluorescence has become the most realistic scheme for producing tunable picosecond pulses in the near-infrared region.” The general time dependent description of parametric amplification or superflucrescence by ultrashort pump pulses has been given by ‘Akhmanov et al.,!® but the qualitative features are the same as those discussed in Section 7.6, In particular, if the group velocity mismatch can be neglected, then the quasi-stationary solution is still applicable in the moving frame of the propagating pulses. Parametric fluorescence or superfluorescence can also be used to determine experimentally the frequency tuning curves of parametric oscillators.'**° This is most useful when the refractive index data of the crystals are not readily available. 9.6 BACKWARD PARAMETRIC OSCILLATOR The case of parametric amplification with counterpropagating signal and idler waves shows interesting characteristics and deserves special consideration. The two waves grow in opposite directions and, through parametric interaction, impose a positive feedback on each other. Then, with sufficient gain, the system may become a mirrorless oscillator, that is, it may yield a finite output with a zero input.’® The principle is similar to the backward-traveling wave tube.”° The equations governing backward parametric amplification are a simple modification of (9.1). With negligible pump depletion and linear attenuation, they are 26 = iat gee et Ak ttiB0 1 Ry heres and (9.59) ia} ‘t op = KE Open i Ske” iB az 2 assuming £, = &exp[ik,z — fayt + ig), By = Gyexpl —tkyz — fant + ids), FyReferences 139 = constant, and Ak = k; — k, + Ky. The input boundary conditions are = &,(0) at z = Oandd}# = &F(!) atz = 1. The solution of (9. 59) for Ak = 0 is (% = 1/2 is assumed for simplicity) Ezy = nn j2i[a)” gs [ Boz / cost | + (B)” 89 (1))sin cos. (9.60) é(2)= S| 2) 4,0) sol 1) D / costs! ‘| 2 Boz Bol sets /awt] with gy given in (9.3), and hence the output is Bot k. + Sol é() = = ,(0) / cos 8° +i ste) ex()an 8, (9.61) oy f ky? t arto) ~ 12 (B) 6 (0tan $2 ds eH Df cos If gof/2 > 77/2, both &(/) and &(0} become infinite unless the input (0) and £3(/) are zero. This is therefore the oscillation threshold. The actual output will grow drastically as go/ > # and will be determined only by taking into account the pump depletion through the nonlinear coupling between the waves, Experimentally, backward parametric oscillation has not yet been observed. The reason is that the phase-matching condition Ak = 0 cannot be satisfied in the usual cases. It is possible to obtain Ak = 0 with #, (or @)) in the far infrared range,?! but then the gain coefficient go/ is too smail to reach the oscillation threshold. REFERENCES W. H. Louisell, Coupled Mode and Parametric Electronics (Wiley, New York. 1960). 1 A. Giordmaine and R. C. Miller, Phys. Rev. Lett. 14, 973 (1965). RG. Smith, J. Appl. Phys. 41, 4121 (1970). J. A. Armstrong, N. Bloembergen, J. Ducuing, and P. §. Pershan, Pays. Rev. 127, 1918 (1962). AE. Siegman, Appi. Opt. 1, 739 (1962). R. G. Smith, J. E. Geusie, H. J. Levinstein, S. Singh, and L. G. van Uitert, J. Appl. Phys. 39, 4030 (1968). Awnawene140 Parametric Amplification and Oscillation 7 1. A Giordmaine and R.C. Miller, in P. L. Kelley, B. Lax, and P. E, Tannenwald eds., Physics of Quantum Blectronics (McGraw-Hill, New York, 1966); Appl. Phys. Lett, 9, 298 (1966) 8 JE, Bjockhole, Appl. Phys. Lete. 13, 399 (1968). 9 J.B. Bjorkbola, Appl. Phys. Lett. 13, 53 (1968). 40 L.B, Kreuzer, Appl. Phys. Lett. 15, 263 (1969). LL J.B Bjorkbolm, [EEE J. Quant. Electron. QE-1, 109 (1971). 12 J. Falk and J. E. Murray, Appl. Phys. Lett. 14, 245 (1969). 13. §.J. Brosnan and R. L. Byer, ZEEE J. Quant. Electron. QE-AS, 415 (1979). [4 RL. Byer andR. L. Herbst,in ¥. R. Shen, ed., Nonlinear Infrared Generation (Springer-Verlag, Berlin, 1977), p. 81. 15 W. H. Louisell, A. Yariv, and A. E. Siegman, Phys. Rev. 124, 1646 (1961); T. G. Giallorenzi and C. L, Tang, Phys. Rev. 166, 225 (1968). 16 $, E, Harris, M. K, Oshman, and R. L. Byer, Phys. Rev. Lest. 18, 732 (1967); R. L. Byer and S. E. Harris, Phys. Rev. 168, 1064 (1968). 17. TA Rabson, H. J. Ruiz, P. L. Shah, and F. K. Tittel, Appi. Phys. Lett. 21,129 (1972); A. HL Kung, Appi. Phys. Lert, 25, 653 (1974), A. Scilmeier, K. Spanner, A, Lavbereau, and W. Kaiser, Opt. Comm. 24, 237 (1978). 18 S.A. Akhmanov, A. I. Kovrigin, A. P. Sukhorukov, R. V. Khokhlov, and A. §, Chirkin, JETP Lett, T, 182 (1968); S. A. Akhmanov, A. S. Chirkin, K.N. Drabovich, A. I. Kovrigin, R. ¥. Khokhiov, and A, P. Sukhorukov, [EEE J. Quant, Electron. QE-4, 598 (1968); S. A. Akhmanov, A. P. Sukhorukov, and A. S. Chirkin, Sov. Phys.-SETP 28, 748 (1969). 19 S.E. Harris, Appl. Phys. Lett. 9, 114 (1966). 20 LR. Pierce, Travelting Wave Tubes (Van Nostrand, Princeton, NJ, 1950). 2K. HE. Yang, P. L. Richards, and Y. R. Shen, Appl. Phys. Lett. 19, 320 (1971). BIBLIOGRAPHY Akhmanov, S. A, and R. ¥. Khokhlov, Uspekhi 9, 210 (1966). Brosnan, §. J., and R. L. Byer, [EEE J. Quant. Electron. QE-15, 415 (1979); R. A. Baumgartner and R. L. Byer, F. Quant. Efectron, QE-15, 432 (1979). Byer, R. L., in H, Rabin and C. L. Tang, eds., Treatise in Quantum Electronics, Vol. 1, Part B {Academic Press, New York, 1975), 587. Byer, R. L., and R. L, Herbst, in ¥, R. Shen, ed., Nonlinear Infrared Generation (Springer-Verlag, Berlin, 1977), p. 81 Giordmaine, J. A., in R. J. Glauber, ed., Quarture Optics (Academic Press, New York, 1969), p. 493, Harris, S. E. Proc. [EEE 57, 2096 (1969). Smith, R. G., in F. T, Arrechi and E. 0. Schulz-Dubois, eds, Laser Handbook (North-Holland Publishing Co, Amsterdam, 1972), p. 837. Tang, C. L,, in H. Rabin and C. L. Tang, eds., Quantum Etectronics, Vol. 1, Part A (Academic Press, New York, 1972), p. 419. Yariv, An Quancum Electronics (Wiley, New York, 1975}, 2nd ed., Chapter 17.OS™“ Ee 10 Stimulated Raman Scattering The wave coupling problems discussed in previous chapters are certainly not limited to electromagnetic waves. They can be generalized to involve both electromagnetic waves and other types of waves, allowing us to imagine a host of new nonlinear optical effects. In this chapter, we show that stimulated Raman scattering can be described classically as a parametric generation process with one of the electromagnetic waves replaced by a material excita- tional wave. More correctly, one would, of course, treat the material excitation quantum mechanically. Stimulated Raman scattering is then considered a two-photon stimulated process grown out of spontancous Raman emission. Stimulated Raman scattering is one of the few nonlinear optical effects discovered in the early 1960s. Over the years, many useful applications have been developed out of this process. Some of them are discussed in this chapter. 10.1 HISTORICAL REMARKS In 1962 Woodbury and Ng,’ while studying Q-switching of a ruby laser with a nitrobenzene Kerr cell, detected an infrared component in the laser output. Its frequency was 1345 cm~' downshifted from the laser frequency. This frequency shift coincided with the vibrational frequency of the strongest Raman mode of nitrobenzene. It was then recognized by Woodbury and Eckhardt? that the infrared output must result from stimulated Raman emission in nitrobenzene. This was soon verified in a large number of liquids by several research workers. Similar effects were also found in gases and solids. Table 10.1 shows a list of Raman modes of some materials in which stimulated Raman scattering was observed. An early theoretical description of stimulated Raman scattering was given by Hellwarth,> who treated it as a two-photon process with a full quantum i141142 Stimulated Raman Scattering Table 10,1 Frequency Shift, Linewidth, and Scattering Cross Section of Spontaneous Raman Scattering for a Number of Substances and the Corresponding Stimulated Raman Gain? Cross Section Raman Gain Raman Shift Linewidth do/dQ x 10° G, X 10° Substance (em™!) 20 (em73) (em™! = ster™?) (om /MW) Gas HE 4155 02 1.5 (300 K, 10 atm) Liquid 0, 1522 0.177 0.48 + 0.14 14.544 Liquid N, 2326.5 0.067 0.29 + 0.09 45 Benzene 992 245 3.06 2.8 cs, 655.6 0.50 7.55 24 Nitrobenzene 1345 66 64 21 LiNbO, 238 7 262 28.7 InSb 0 - 300 03 10 1,7 x 104 “After Y. R. Shen, in M. Cardona, ed., Light Scattering in Sotids (Springer-Verlag, Berlin, 1975), p. 275. oR E. Hagenlocker, R. W. Minck, W. G. Rado, Phys. Rev. 154, 226 (1967). °For a carrier concentration n, ~ 10" em7>, mechanical calculation. The simple theory, however, cannot account for the observed anti-Stokes radiation which often appears with intensity almost as high as the Stokes radiation, Garmire et al.“ and Bloembergen and Shen? Jater used the coupled wave approach to describe stimulated Raman scattering and were able to explain the anti-Stokes generation as well as the higher-order Stokes and anti-Stokes output. Yet the theories still could not explain many other important observations. These included an observed stimulated Raman gain much larger than the value predicted by the theories, extremely sharp threshold for stimulated Raman emission, asymmetry of forward-backward Raman intensity, and appreciable spectral broadening of the Raman radiation. It was later realized that these anomalies were actually induced by self-focusing of the laser beam in the medium, which is discussed in Chapter 19, Without self-focusing, the theories predicted experimental results satisfactorily. Early interest in stimulated Raman scattering arose because it could provide intense coherent radiation at new frequencies and because it was a possible loss mechanism in propagating high-power laser beams in a medium, for example, in the atmosphere or in a fusion plasma. More recently, stimulated Raman scattering was used to generate tunable coherent infrared radiation by either tuning the material excitation as in stimulated polariton scattering’ and stimutated spin-flip Raman scattering,’ or by tuning the exciting laser frequency with, for example, a tunable dye laser. Spectroscopic applications of stimulated Raman scattering have also been developed, with emphasis on high-resolution studies® Transient stimulated Raman scattering has been applied to the studyQuantum Theory of Stimulated Raman Scattering 143 of relaxation of material excitations in the picosecond region with mode-locked laser pulses.” 10.2. QUANTUM THEORY OF STIMULATED RAMAN SCATTERING Raman scattering is a two-photon process in which one photon at w)(k,) is absorbed and one photon at w,(k,) is emitted, while the material makes a transition from the initial state |/) to the final state |f, as shown in Fig, 10.1. Energy conservation requires 4(w, — 0.) = €,— E; = hw,,, which is the en- ergy difference between the final and initial states. Stokes and anti-Stokes scattering correspond to w,, > 0 and w,; < 0, respectively. To find the Raman transition probability, we use the standard second-order perturbation calculation.” The interaction Hamiltonian in the electric-dipole approximation is Hig = — er E + adjoint (10.1) i where E = deft) + & eat 92). The Raman transition probability per unit time per unit volume per unit energy interval is found to be dW, aW, Bn ?Neye, d(fw,) d(ha,) 88) M=) KAMAE K alas ayla,)/?g (Ade) . . . (10.2) ersé.| s)(s jer? er 8 i s)¢s ler é Filer, — 0,,) Ai(o, + wy} Here WN is the density of molecules or unit. cells in the medium, e¢ is the hun, Chy) hus (Ka} H> Fig. 10.1. Schematic drawing showing the Stokes (a, < «,} Raman transition from the initial state |#) to a final excited state |f), and the anti-Stokes (w; > #,) Raman transi- li> don from |f) to |7)-144 Stimulated Raman Scattering dielectric constant, ¢ denotes the field polarization, |s) is the intermediate state of the material system, ja) denotes the state of the radiation field, a* and a are the creation and annihilation operators of the field, and finally g(hde) describing the lineshape is the joint density of states of the Raman transition. For a Lorentzian lineshape, we have AL /a g( hha) = —————_— (ade) 1{de) + PT? (10.3) where AT is the halfwidth in energy units. The transition probability in (10.2) is proportional to Kayley ayla,)?. Te Co,| = Cy, ny) and (ay = (my — Ly + 1 with n, and 7m, being integers, then dW,,/d(ha,) © ny(mz + 1); spontancous and stimulated Raman scattering correspond to , = 0 and ny ¥ 0, respectively. The states of the radiation fields are often more complex in general. Then there exists 10 simple expres- sion of (a;|aj a;la;). However, if the average numbers of photons 7, and 7, at w, and @, are much larger than 1, the approximation Kaplatayta,>|? = Fh, is excellent In a certain sense, this is equivalent to saying that the field can be treated classically. Consider now the propagation of the #, and w, beams in the Raman medium. The Raman transition leads to the absorption of the w, beam and amplification of the w, beam. From a simple physical argument, the change of the average number of Raman photons in a single mode at w, per unit length of propagation is given by® dy _ (2M aM a de | da, *~ da, = (Ga — afin My 6 (0.4) where p, and p, are the populations in \) and |f)}, and a, is the attenuation coefficient at w,. Since W, = W,, from detail balancing, the quantity G, takes the form dW, ey? a —fi(y — p) Ga dw, (e~ %) iy {10.5} |M,|?F,g(hdwo)(p;,— py), forty, Hy > 1 _ 89° Now, h ale which is a constant if 7, can be treated as a constant. This is the case when depletion of the @, pump beam is negligible. The solution of (10.4) is then simply an exponentially growing function of z: ,(z) = Ay(O)e'Or" {10.6}Quantum Theory of Stimulated Rasian Scattering 445 with Gg playing the role of a stimulated gain coefficient proportional to the incoming laser intensity at w,. Since the gain is proportional to the Raman transition probability, one expects that it should also be proportional to the spontaneous Raman cross section, By definition, the differential Raman cross section #0 /d( hw) dQ is the scattering probability of an incoming photon at w, per unit area into a Raman photon of a certain polarization at w, per unit solid angle around Q and unit energy interval around frw,: ao aW,;, . , a(ha,) da * aha, 5 / NKayla asi.) c (10.7) = (28) on 2g(hAw)p, where see 0 Z2)/ (an)? is the density of radiation modes per unit solid angle at w,. From (10.5) and (10.7), we immediately find the relation 4n7c%s, =N. 1 E,\2 G wiakesp, (eo - lage aa ’l vas 4 . NATEE (6, - » 1 FH irene, @a}e)p; Thus, given the spontaneous Raman cross section do /d@ and the halfwadth T, the stimulated Raman gain in a medium can be estimated easily. Table 10.1 shows the cross sections, the halfwidths, and the estimated gains for the Raman lines of some materials in which stimulated Raman scattering has been observed. As an example, let us consider stimulated Raman scattering in benzene. From Table 10.1, we find that the Raman gain for the 992 em~! mode of benzene is 2.8 x 10-3 cm/MW. Thus, in order to generate e”” Raman photons from one noise photon (corresponding to an output of ~ 100 kKW/cm’) in a 10-cm benzene cell, a laser beam of 1 GW/cm¥ is needed. This shows that a high-power pulsed laser is necessary for the study of stimulated Raman scattering. In actual experiments, stimulated Raman scattering with a power gain of ~ e*° in benzene and in many other liquids has been observed with a laser beam of ~ 100 MW/cn® or less. The order-of-magnitude discrepancy between theory and experiment is the result of self-focusing of the laser beam in the medium (see Section 17.3).146 ‘Stimulated Raman Scattering Note that we use the approximation (a,\a}¢)|a,>|? = 1A) in describing stimulated Raman amplification. This approximation is certainly not valid when 7i, or 7, is small. Therefore, strictly speaking, this theory is rather crude for describing stimulated Raman amplification starting from noise or sponta- neous scattering. The complete quantum theory of stimulated Raman scatter- ing by spontancous scattering is a difficult but challenging problem in quantum optics and has not yet been fully developed. In some respects, it is a two-photon analog of the superfluorescence problem (Chapter 21). 10.3 COUPLED WAVE DESCRIPTION OF STIMULATED RAMAN SCATTERING Coupling of Pump and Stokes Waves From the wave interaction point of view, a two-photon transition is a third-order process, It can be seen from the relation that the net transition rate is equal to the rate of generation or loss of photons at @, OF w,: aW, 8 oy ri, — op) = are| 2" (e,) +B, ]|/Ae (10.9) =2 Rel FP (w,) -E, /fieo,. Since dW,,/du & \E\PIE I, we have PO(w,) © |B,)7B, and POH & |£,!?£,, both of which are therefore third-order nonlinear polarizations. Stimulated Raman scattering can then be described as a third-order wave coupling process between the pump and Stokes waves. The wave equations are 2 4 2 v X(0 X By) — bey Ey = PO (0) ¢ c and (10.10) 2 4 2 y x(W X E,) —Fe,k, = 2 PO(e,). ¢ c For simplicity, we consider the special case of an isotropic medium with E, and E, polarized in the same direction. The nonlinear polarizations take the form PO (a) = [xPIEP + xQlErP] A, and (10.11) PO) = [xBIP + xP LEAP] B- ‘As seen in (10.11), the xf and x§? terms in P only act to modify theCoupled Wave Description of Stimulated Raman Scattering 147 dielectric constants ¢, and e, in (10.10). They are responsible for the field- induced birefringence, self-focusing, etc., but have no direct effect on stimu- lated Raman scattering, We therefore neglect them in the following discussion. The xP terms in P®, on the other hand, effectively couple E, and E, in (10,10) and cause energy transfer between the two fields, They are then responsible for the stimulated Raman process, and they are known as the Raman susceptibilities. The microscopic expressions for x) can of course be obtained from the usual procedure outlined in Chapter 2. Each x@ should have a resonant term and a nonresonant term. Only the resonant term is connected to the Raman process. In fact, the microscopic expression of the resonant term [x@], can be obtained directly from the relation in (10.9) dW, 21m x GE, [| Esl fig oy xe |"|Eal Go, (8 (10.12) ~. Zim XP? LBs!” hi : From (10.5), (10.8), and the relation {E|"«/27 = fwit, we find 4mresd Ga-~ a, (lm x@ DIE, cK, Im x = —Im xq (10.13) c*e,(e, — ey) do Aa awh, aa7ehe) = —NIM;(P(o, ~ ,)ng( hao). If g(#Aw) is a Lorentzian, then from the Kramers-Kronig relation, the real part of [xl can be explicitly derived. We then have the microscopic expression x8h [x@le [xS]re + LxSbles _ NIMC — &) al(o, — & ~ %,) — iT] xl = xR. : (10.14)148 Stimulated Raman Scattering Knowing x@, we can now solve (10.10) with (10.11). For plane-wave propaga- tion along 2, and assuming slowly varying amplitude approximation, the wave equations reduce to ao. a _f 2awt (2+$)4- (252 aes 1 and (10.15) a & _; 203 \ 6 (3: + 2) 6 = ea xR EW a Anw} (a+ ay |e? = — Tay im xt PE ky and (10.16) a Ary (2 + an)iba? = ~ ps Pm xa These equations can new be identified with the equation for 7, in (10.4) and a similar equation for ii, if \E,|? and |E,|? are replaced by 2oho,f,/e, and Urhw, h/t, respectively, with Gz given in (10.13). Thus, when the depletion of the pump field JE, is negligible, we again have the exponentially growing solution of |E,(2)7 = |Ex(O)exp(Gaz — 2). If the attenuation constants a, and a, are vanishingly small, then even the exact solution of (10.16) can be obtained easily with the help of the conserva- tion of the total number of photons fa? fer + 87 (6,2 /e,] = K. The solution takes the form wae OP _gg| Ene AG) 4 K/a? 14 (0)? = wk Oe? |’ 10.17 A AC) a (ie) AP K/e 16:(0)P — aK er? 4,078? | Parametric Coupling of Optical and Material Excitational Waves Stimulated Raman scattering can also be considered as a parametric genera- tion process in which the optical pump wave generates a Stokes wave and a material excitation wave.** This can be seen as follows:Coupled Wave Description of Stimulated Raman Scattering 149 We recall that the expressions for the nonlinear polarizations in (10.10) can also be derived using the density matrix formalism of Chapter 2, Let us consider here only the Raman resonant term in P®. We can write, in the notation of Fig, 10.1, Pc) < NE [Csler = 2h eB on) + EMpETE:(0,— 9). Then, in the steady state, we have Mji(ei— 94) ooo SE Ala, — 0, — w+ Ty) (10.21) P(e) = [Sh] al Bu?E 2), P(e, - a)= with [x@}]x having exactly the same expression as in (10.14). One can find P2)(w,) similarly. The formalism here shows explicitly p/(c, — «,) as a material excitation resonantly driven by optical mixing Z7E,, Stimulated Raman scattering can therefore be considered a result of coupling of three waves E,(«,), £2(1,), and150 Stimulated Raman Scattering Py: (y — W2)}, as governed by the wave equations [° X(T x) +8 & “24 y0%(0,) ~ 72h, (xP yPEs + [aR all + NMP E,{ a py, (01 — ay)}, (1022) lv xv x) + atl, = {rai p0(u,) = os ba{ix wel EE, xP |EPE, + NM, E,() f(y ~ &)}> a i (F- tay+ Tafoner — ay) = EMA (6, e ETE: where E, is the pump wave, and E, and p,, are the generated waves. Aside from the x® terms, which are not essential for stimulated Raman scattering, (10.22) is very similar to (3.4), which was used in Chapter 9 to describe the parametric generation process, except that here the dynamic equation for pfp'(w, - @2) has replaced the idler wave equation. We assumed in the above discussion a localized Raman excitation between two energy levels. This is a good approximation for most Raman excitations including molecular vibration, optical phonon, electronic excitation, spin-flip transition, and optical magnon and plasmon, even though the dynamic equa- tions for different excitations are generally different. Since pf? has no disper- sion, phase matching for the wave coupling is automatically satisfied here. This makes the solution look different from that of the optical parametric process. The general formalism, however, is valid for any material excitation if the dynamic equation for of? is replaced by the appropriate dynamic equation for the excitation, with or without dispersion. We note that as long as the response of pf? can be considered as stationary, as given in (10-21), elimination of pf m (10.22) with (27/007) E(w) = —wE(w) immediately leads to (10.10), which was used earlier to describe stimulated Raman scattering. However, the set of equations in (10,22) is clearly more general as it also describes the transient case where of? does not respond instantaneously to the time variation of the driving force EE}. The popula- tion difference p, — p, in (10.22) can be approximated by its thermal equi- librium value under weak excitation, but in general, from physical argument, it should obey the relaxation equation dW, 34 + File = op) — (06? - of)] = Naas? -»,) (20.23)Coupled Wave Description of Stimulated Raman Scattering 181 where the right-hand side is the transition rate, which, from (10.9) and (10.19), is found to be aM ; toh — MREZE, 24 Way (POP) = Myo EbOh — MAE Pi]. (10.24) In the general description of stimulated Raman scattering, (10.23) should be solved together with (10.22). Strong excitation of the population leads to ~ saturation in the Raman gain. This description, however, applies only to the localized two-level excitation. For the boson-type excitation, (10.23) should be replaced by the equation Wn 10.25 Toy — p) (10.25) (§ + lle — iis) where n, and 7i, are the average numbers of bosons present with and without the driving fields, respectively. Stimulated Raman Scattering by Molecular Vibrations or ‘Optical Phonons The most common case of stimulated Raman scattering is by a molecular vibration or optical phonon, which is often described as oscillation of the normal coordinate @. In agreement with the conventional definition of Q, we can identify p(w) in (10.22) with (4/20)? and replace its dynamic equation by the driven harmonic oscillator equation for Q a a 2(w, — @) PO " (# +2raet foto, — &)= [Ae e) MyE, ES (0; - py); (10.26) which, with w, — w, > ay, reduces to a,. ; - (5; + jeg + re = i[2A(e, - «))) 7, BET(e, — 9). (10.26a) Then the equations for Z, and E, im (10.22) together with (10.26) properly describe stimulated Raman scattering by phonons. In particular, in the steady- state case, we again obtain the expression for the resonant Raman susceptibil- ity Exb]e in (10.14).152 Stimulated Raman Scattering 104 STOKES-ANTI-STOKES COUPLING We have thus far discussed only the stimulated Stokes emission. Experimen- tally, however, both Stokes and anti-Stokes waves are simultaneously generated with comparable intensity m the stimulated Raman process even at very low temperature. This is very different from the case in spontaneous Raman scattering and cannot be understood as a stimulated two-photon process described in Section 10.2 since the anti-Stokes wave then would be absorbed instead of generated. Yet it can be explained readily by the coupled wave description.“ A third-order nonlinear polarization PO(q,) at the anti-Stokes frequency w, = 20, — «, can be induced in the medium through mixing of the pump wave at w, and the Stokes wave at o, via the Raman resonance at @; — W, = &, — @;, as we see below. With the simultaneous presence of E,(«;), E,(a,), and £,(«,), the material excitation p,,, now driven by both £,E* and E,£}, obeys the equation a... i (Fie Ty ot — HLM ETE, + MA EEN (e,~ 07) 0927) where M,, and M, , are the Raman matrix elements for the transitions incurred by E,, E, and £,, E; respectively. Mixing of the material excitation with the em waves then gives rise to the resonant part of the nonlinear polarizations PQ (cop) = N(MBA. «EP + Mp, aE of)» PO (a,) = NMy, Eph, (10.28) PO(u,) = NMP, Epa If the response of p,, to the fields is stationary, then p,; can be solved from (10.27) and substituted into (10.28), The resultant P& combined with the nonzesonant part of P can be cast into the form PO (w= a + Wy + Oy) = XPE(09,) By wy) E,(w,) with x® containing both a resonant and a nonreso- nant term. The set of wave equations describing the steady-state anti-Stokes generation then is given by 2 Z oy 4Anuy . 7 [r x(V x) tafe = az [Xt 1E,)72; $ (x4 xP BEES + XQ EI, wt Ane? (10.29) le x(¥ x) fee, = Tae IAIE. + xPeres]. 2 2 @. 40, , Iv x(v x)- ade: = va eal xsa Er Et + xQlePes] cStokes—Anti-Stokes Coupling 353 where NIM, (9 — 0) x2 = (x® wr - xP = (x8 ben Alo, ~ «4; — Ty) NMy. Mp. «(Pr h(w, w; 7 Oy NIMy,, ol"( 0) — 9) h(a, wy, ~ iT) x= = (x2) INR > x2 = (x2 )na - and we have neglected terms of xE,|7E, in (10.29). The set of equations in (10.29) actually describes a four-wave parametric generation process with E, being the pump wave, £, and E, being the signal and idler waves, and x,, acting as the coupling constant between the signal and idler waves. The solution of (10.29) in the limit of ao pump depletion therefore follows closely that of parametric generation described in Chapter 9. We skip the derivation and present only the results here. Assuming an isotropic medium with a plane boundary at z = 0 and slowly varying amplitudes for £, and E,, we find? E, = [&,expAK,2) + &_exp(iAK_z)]explik, «r — a,2) and (10.31) Et = [&,exp(iAK,z) + 6% exp(iK_z)]exp[—ik, +1 — (14k + @,)2] where O= Bhs y — Kexy — Kaxsyr Ak = 2ky, — yp Kass Kp EE aK .= Se {(¢) -(anya]”, 2a? A= | ],018 2 “(eh als « we” ik tk, Vk Gp = —2Im(A). For the sake of simplicity, we neglect the dispersions of the absorption154 ‘Stimulated Raman Scattering Fig. 10.2 General relationship between the wavevectors of Stokes, anti-Stokes, and laser waves. coefficients a, and the coupling coefficients (27w?/c7k,)x®, The relationship between the various wavevectors is shown in Fig, 10,2, With ¢, =o and €, = E, at z = 0, we have &, AK,-A at t Eon A? 10.32 ¢ _ (eb K 5+ A) So + AGI { ) i AK ,- AK; . A number of important physical results come out of the above solution. First, through the Stokes-anti-Stokes coupling, two composite waves (Fea: Fa4) and (é,_, &,-) are formed with the eigenvectors AK, and the Stokes-anti-Stokes amplitude ratio (¢,/é,) ,. One of them may experience an exponential gain and the other a loss if Im(AK.) < 0. The coupling clearly depends on phase matching. If the phase mismatch is sufficiently large that |4k| > |Al, then the Stokes and anti-Stokes waves are effectively decoupled. This is explicitly seen by the fact that AK, and @, ,/¢,, reduce to AK_=A, AK,=Ak-A and en é,_ b A Ak -|Ale2 The first denotes a nearly pure Stokes wave with an exponential power gain of 21m(AK_)= Ga, while the second denotes a nearly pure anti-Stokes wave with an attenuation —G,. These are the results expected when the Stokes and anti-Stokes waves are decoupled from one another. The Stckes-anti-Stokes coupling is maximum when Sk = 0. The solution yields AK ,= 0 and‘Stokes—Anti-Stokes Coupling 155 |6%,/6, .| = 1. Then neither Stokes nor anti-Stokes waves should experience an exponential growth. This is because through coupling, the Stokes gain is completely canceled by the anti-Stokes attenuation, a result well known in the parametric amplifier theory where no gain can be expected at @, = w, — w; if the other sideband at w, = w,+ «, is not suppressed. As jAX| gradually deviates from zero, the gain [2 Im(A.X_)| increases rapidly from 0 toward G,, as shown in Fig, 10.3, while the anti-Stokes-Stokes ratio {\@%_/¢,_| decreases from 1 to 0. Appreciable anti-Stokes generation is therefore expected in the region |Ak| ~ {A| where both |Im(AX_)| and |¢* /¢,.] are sufficiently finite. The anti-Stokes radiation should appear in the form of a double cone around the phase-matching direction, as shown in Fig. 10.4. We should, however, remember that infinite plane waves are assumed in this theory. In reality, a pump beam of finite cross section has a spread of wavevectors. Consequently, the sharp dip in the gain curve in Fig. 10.3 may get smeared out, and instead of the double cone in Fig. 10.4, one may observe only a single cone of anti-Stokes radiation.? ‘STOKES GAIN COEFFICIENT —=- MOMENTUM KISKATCH —* Ak /Gp t L 1 4 1 -4 -2 © 2 4 6 Fig. 103 The Stokes power gain as a function of the normalized near momentum mismatch Ak/Gg in the z direction. The asymmetry is due to the nonresonant part xS% = Olin xP max (After Ref. 5.)156 Stimulated Raman Scattering POWER GAIN (db) [ttegl/Lo fons] L120 Gaz 530 100 " co Gar= 20 Lo ak/Gr Fig. 10.4 Anti Stokes intensity versus the linear phase mismatch Ak (normalized by the Stokes power gain Gq). The asymmetry is due to X yp ~ 0-1 xP lmax- (After Ref. 5) 10.5 HIGHER-ORDER RAMAN EFFECTS Intense higher-order Stokes and anti-Stokes radiation often shows in experi- ments accompanying first-order Stokes and anti-Stokes generation.’* Unlike the overtones in spontaneous scattering, they appear with frequencies at w@ £ oy, where n is an integer, This characteristic suggests that they are generated more or less successively. For example, when the first Stokes £,(«,) becomes sufficiently intense, it may also act as a pump wave to generate the second Stokes Ey, (@,3 = #, — @yi)- In general, however, we must find theHigher-Order Raman Effects 157 answer from the coupled wave approach, because an ath-order Stokes or anti-Stokes wave may couple to several Stokes and anti-Stokes waves of various orders through third-order nonlinear polarizations. Again, coupling of a particular set of waves is strongest if phase matching is satisfied. Unfor- tunately, with many complex waves nonlinearly coupled together, the problem becomes extremely complex. We consider here, as an example, a special case where only the Stokes generation in the +2 direction needs to be taken into account? This can be achieved in a real situation with short pulsed laser excitation such that the backward Stokes generation is suppressed (see Section 10.9), while the anti- Stokes generation can be neglected. The set of coupled wave equations is then given by (S a + ee, Ares 21 Ve, = -( FE) ete, 2 2, 40 2 a ale -( Janet, + x eF 6) (10.33) a + W282 E a2? ct 2 etc. The solution of (10.33) can be obtained by numerical calculation, as seen in Fig. 10.5. As the length of the medium or the pump power increases, the first Stokes power increases gradually at the beginning, and then suddenly builds up to a maximum while the pump power plunges to nearly zero by depletion. Then the first Stokes power remains roughly constant for a while and gets depleted into the sécond Stokes, and so on. This is in fact what one would Anat, ' -(3"|QeaPen + x9 Esl En], Psa / Petar o 100 200 300 400 Fig. 105 Generation of higher-order Stokes waves as a function of normalized cell length z = (165°? lm x $/k, Py > Py Stokes power (kW) G1 o 10 20 30 Call Fength (crm? Fig. £0.7 First-order forward and backward Stokes power versus the toluene cell length at three laser powers P, = 80, P, = 67, and P, = 53 MW/car. [After ¥. R. Shen and Y. J. Shaham, Phys. Rev. 163, 224 (1967),] 158Experimental Observations and Applications ts9 expect from the simple physical picture. The calculation here, however, as- sumes infinite plane waves. With a pump beam of finite cross section, the rise and fall of an nth-order Stokes wave should be more gradual, as demonstrated both theoretically and experimentally by von der Linde et al.? 10.6 EXPERIMENTAL OBSERVATIONS AND APPLICATIONS Stimulated Raman Scattering in Self-Focusing Media A typical setup for studying stimulated Raman scattering is seen in Fig. 10.6. As mentioned in Section 10.1, earlier results on the Stokes intensity measure- ments disagreed strongly with theory. Most of those experiments were on liquids with large Kerr constants and the results showed a sharp threshold for stimulated Raman scattering and an effective Raman gain an order of magni- tude larger than the predicted gain. An example is given in Fig. 10.7, which also shows a forward-backward asymmetry in the Stokes output that was not predicted by the theory, Other anomalies such as the Raman spectral broaden- ing and the anomalous anti-Stokes ring pattern were also observed. It was later realized that these anomalies were initiated by self-focusing, which readily occurred in Kerr media (see Chapter 17). Self-focusing has a threshold; it increases the laser intensity dramatically at the focal region, and breaks the forward—backward symmetry of the Raman amplification. This then explains qualitatively the results of Fig. 10.7. Raman anomalies constituted a subject of great confusion in the past. We do not go into any detailed discussion on the subject here. Interested readers should consult Section 17.3 and the relevant literature (see Bibliography). Stimulated Raman Scattering in Non-Self—Focusing Media Even without self-focusing, stimulated Raman scattering often shows a certain gain anomaly. An example is seen in Fig. 10.8, where the Stokes output is plotted against the laser input in liquid nitrogen. Self-focusing was not observed in this case. As the laser power increases, the Stokes output first increases linearly because it is generated from spontaneous scattering, and then grows quasi-exponentially when stimulated scattering sets in. At a certain input power, the output rises suddenly with an effective gain much larger than the theoretical prediction. It finally levels off as a result of depletion of the laser power. The sharp rise of the Stokes output is presumably due te feedback in the Stokes amplification from Rayleigh scattering or diffuse reflection from walls and cell windows. Experiments of the kind in Fig. 10.8 actually describe the build-up of Raman oscillation, that is, amplification from noise. As is well known, the output of an oscillator without saturation depends critically on small perturbation or feedback. This makes the quantitative interpretation of160 Stinmilated Raman Scattering wr? Saturation 10-* Strokes intensity Ix (arbitrary units} 10-8 0 O62 04 06 O8 10 1.2 14 16 18 Laser intensity fz, (arbitrary units) Fig. 10.8 Experimental result of the first-order Stokes output as @ function of the laser intensity in liquid N,. [After 3. B. Grun, A. K. McGuillan, and B. P. Stoichefl, Phys. Rev. 180, 61 (1969),] the oscillator output extremely difficult, especially if the perturbation or feedback cannot be well characterized. With careful elimination of feedback, Haidacher and Maier have shown that the sharp rise of the Stokes output can be greatly suppressed.’ Raman Gain Measurements ‘The theory developed earlier is clearly a theory of Raman amplification rather than oscillation. To check the theory, one should carry out experiments on Raman amplifiers.'> This can be done with the oscillator-amplifier setup in Fig. 10.9. The backward Stokes emission from the oscillator provides a Stokes input to the amplifier, and the amplification gain of the backward Raman scatteringStokes Raman Stokes Raman Hy Laser Absorber Amplifier Absorber Oscillator PM, PM, Fig, 10.9 Experimental set-up for measuring the backward Stokes gain. (After Ref. 15.) GAIN 1 1 1 L ! ‘ L L o 10 20 30 40 50, 60 70 80 p AMAGATS Fig. 10.10 Stokes Raman gain in 7, gas as a function of pressure. The data points ¥ for forward gain should be compared with the dashed theoretical curve (cell length: 80 em; peak input intensity: 20 MW/cnr’). The data points + for backward gain should be compared with the solid theoretical curve (cell length: 30 cm; peak input intensity: 60 MW cm at 0.6943 am.) (After Ref. 15.) 161162 Stimulated Raman Seattering in the amplifier is measured. Figure 10.10 shows the result on hydrogen gas exhibiting good agreement between theory and experiment. Such a Raman gain experiment was, however, not every successful in self-focusing media, where self-focusing in both the oscillator and the amplifier led to complica- tions, Anti-Stokes and Higher-Order Raman Radiation Anti-Stokes radiation of many order often can be observed in stimulated Raman scattering’? In condensed matter they appear in the form of bright multicolored rings on a plane perpendicular to the laser beam. Rings of different color correspond to different orders of anti-Stokes, Chiao and Stoicheff'? showed in calcite that the mth-order anti-Stokes radiation is emitted in the direction given by the phase-matching condition k,,, = Ke, n-1 + ky — k, ,. This is expected if we assume that the higher-order anti-Stokes is generated successively from the lower-order Stokes and anti-Stokes. The first- order anti-Stokes ring should be defined by k, = 2k, — k, according to the theory. In self-focusing liquids the situation is more complicated. The directions of the anti-Stokes rings now deviate from those defined by kg. = Ke n-1 + ky — k,,, a8 they are now affected by self-focusing. Ganmire’® had some success in interpreting these rings by assuming the existence of thin filaments of pump light resulting from self-focusing in the medium. The problem, however, remains since the assumption of filaments is not quite valid. 1.0, Power ratios —> S a 0 6 9 100 200 300 400 500 Laser intensity 2, (0, 0) [MW/em?] —>» Fig. 10.11 Normalized transmitted laser (R 1), first (Rss) and second Stokes (R52) power as a function of the incident laser intensity 1,(0,0). The experimental data of R,, Roy, and Rey are represented by circles, rectangles, and diamonds, respectively. The curves are calculated according to the theory in Section 10.5 with the finite beam cross section taken into account. (After Ref. 13.)Experimental Observations and Applications 183 Higher-order Stokes radiation appears mostly along the axis in the forward and backward directions. Quantitative studies generally are difficult. Using subnanosecond laser pulses, von der Linde et al.!? were able to carry out quantitative study in a special case. The short input pulse excited only the Stokes radiation in the forward direction, as already described in Section 10.5. Their results are shown in Fig. 10.11. Theoretical calculation following (10.33) but with a Gaussian beam profile shows good agreement with the experiment. Stimulated Raman Scattering by Broadband Pump Source Raman gain measurements in the absence of self-focusing may still show a forward—backward asymrietry. This results from the finite linewidth of the pump beam. If the pump linewidth is 2F,, and the Raman linewidth is 2I, theories predict that the maximum backward Raman gain is proportional to (r+ Tr)? and the maximum forward Raman gain is proportional to Th assuming no relative dispersion between the Stokes and pump frequencies.) This surprising result can be understood qualitatively from the following picture, In the forward direction, as a short Az section of the Stokes wave propagates in the medium, it always coherently interacts with the same Ar section of the pump wave. On the other hand, if a short section of the Stokes wave propagates in the backward direction, it constantly encounters a new wavefront of the pump wave. Consequently, the forward Raman gain follows the pump intensity variation and is proportional to T? as predicted by the stationary theory described earlier. The backward Raman gain is reduced because the amplification process averages over the amplitude and phase variation of the pump field. Explicit results can be obtained, for example, by considering the special case of stimulated Raman scattering by a short pump pulse (see Section 10.9) where the pump linewidth is given by the inverse of the pulsewidth. More generally, statistical theories should be used to describe the problem.” Competition between Raman Modes In stationary stimulated Raman scattering, only the mode with the maximum gain appears to participate in the process. It is usually the mode with both a large Raman cross section and a narrow Raman linewidth. Effective depletion of laser power into this Raman mode forbids the occurrence of stimulated scattering into other modes. Only in transient cases (see Section 10.10) will several competing modes show up simultaneously. Inverse Raman Effect A loss in pump radiation always accompanies the gain in the Stokes radiation in a stimulated Raman process. Thus with both the pump and the Stokes waves sent into a medium, one can observe simultaneously the gain of the164 Stinmulated Raman Scattering Stokes wave and the attenuation of the pump wave. The absorption of the pump radiation in a stimulated Raman process was first demonstrated by Jones and Stoicheff and is known as the inverse Raman effect.? Tunable Infrared Sources Obtained from Stimulated Raman Scattering Stimulated Raman scattering with its Raman-shifted output has long been considered a viable method for generating coherent radiation at new frequen- cies, Special interest is in tunable coherent sources. Since the Raman frequency of a medium is usually fixed, the tunability is often achieved by using a tunable pump laser. Two systems have received much attention, One is the atomic vapor system, mostly alkali and alkali-carth metal vapor. Raman transitions involved are often of the § + P > Sor S > P -> D type.!*™ The tunable pump source is near resonance with the S > P transition, so that the Raman cross section is greatly enhanced. As a result, even though the atomic vapor density is smail (-s 10” atoms/cm’ at a pressure of ~ 10 torr), the Raman gain is significant. Stimulated Raman scattering with emission m the near infrared can be readily observed. With a dye laser input of a few tens of kilowatts in an alkali vapor, the infrared output can have a continuous tuning range of several hundred inverse centimeters and a peak photon conversion efficiency as high as 50%.° The tuning range can be further extended by using different Raman transi- tions, For example, in Cs, with a tunable pump dye laser in the blue and uv, the observed Stokes output appears in the range of 2.5-4.75 pm, 5.67-8.65 pm, and 11.65-15 am sesulting from the Raman transitions 6S + 75,65 — 8S, and 6S > 9S respectively. It can be extended to ~ 20 ym using the 6S > 105 transition. If a broadband dye laser is used as the input, the infrared output will have the same broad bandwidth. With a pulsed dye laser, Bethune et al.?* generated the broadband infrared beam and used it to obtain single-shot absorption spectra of molecules. To increase the detection sensitivity, they up-converted the infrared signal transmitted through the sample into visible through nonlinear mixing in another alkali vapor cell. The technique allows the recording of infrared spectra of transient chemical species with nanosecond resolution using nanosecond laser pulses. At high pump power, the output power from stimulated Raman scattering in atomic vapor is often limited by the occurrence of multiphoton absorption and ionization, population saturation, field-induced spectral broadening, and other nonlinear optical processes.” However, output as high as 30 mJ at ~ 2.9 um with a photon conversion efficiency of 40% has been observed in Ba vapor. The Stokes owtput from atomic vapor tends to have a linewidth increasing with the pump intensity. This line broadening may result from the Stark effect caused by the photoionized atoms, or saturation in the Raman transition, or others; the dominant mechanism has not yet been identified. Another system of immense practical interest is the molecular gas system, such as H), Nj. These simple molecules have very strong Raman modes,Experimental Observations and Applications 165 Intense Raman output can be expected from a gas cell of a few tens of centimeters long at a few atmospheric pressure with several megawatts per square centimeter laser intensity in the visible.” Molecular hydrogen is proba- bly most interesting because of its large Raman shift (4155 cm~'). With a pump source at A > 7500 A, the third-order Stokes output has a wavelength larger than 10 pm. Figure 10.12 shows the accessible wavelength range the various orders of Stokes and anti-Stokes output from H, and D, can cover with a tunable pump source between 2000 and 8000 A. The output power can be significant, as indicated in Table 10.2 for a commercial unit. It can be Table 10.2 Raman Output from RS-1¢ Wavelength (nm)? Energy (mJ) Pressure (psi) With 560-nm Pump Beam’ 195 (ASs) 0031 125 213 (AS) 0091 125 234 (AS,) 024 110 259 (AS) 054 15 290 (AS,) 10 145 330 (Sy) 26 160 382 (AS,) B “190 454 (AS,) 21 200 730{S,) 7 90 1048 (S,) 62 300 1855 (55) 60 275 With 280-nm Pomp Beam’ 207 (AS;) 038 100 227 (AS,) ag 125 251 (AS,) SA 125 317(S,) 22 40 365 (54) 31 300 430 (3) 12 160 $24(S,) 0.24 110 669(S,) 0.060 100 “After Quanta-Ray, Inc., advertising brochure on RS-1 Raman Shifter. 's denotes ith Stokes wavelength; 45, denotes ith Anti- Stokes wavelength. *Gas: H, at 300°K Pump; 85 mJ at 560 nm from a Quanta-Ray PDL-1 dye laser pumped by a DCR-1A Nd ; YAG laser. ‘Pump: 17 mJ at 280 nm from a frequency doubled Quanta-Ray PDL-1 dye laser pumped by a DCR-1A Nd: YAG laser.Re NE (251 “Kery-pimend Aq Jaljius UEWEY T-Sy UO amMyoosg FuIsMUAApe ay daiyy) ‘sese3 durnd se *,_ar> 1962 Jo UMS UemeY @ Ds “cl (g) pu "|W slp Jo YRS eeuTY & MLA TH (0) Jo 3s0 WO a[gissaeoe suPBuazsaem a[qissod ATeops1O3T} 9) AOYs SSAIMO JO SOTNUTEY UL zl Bu 9) @) (WU) unbuajanem dung (wu) yaBuajasem ding O08 aoe oo3 005 oor 006 00z 008 O04 009 oog or ODE One fe TT TTT T ° 166 (ust) yndina soyyus veweY,Experimental Observations and Applications 167 improved by using the oscillator-amplifier scheme as in the laser-pumped dye laser case. Conversion efficiency as high as 80% in the Stokes output has been achieved. High-power infrared CO, lasers can also be used to obtain stimulated Raman scattering in molecular gases. It yields, for example, a Stokes output around 16 wm from CH,. A conversion efficiency of ~ 10% can be achieved.?4 The 16-1m radiation is most important for laser isotope separation of uranium through vibrational excitation of UF,. It can also be obtained by stimulated Raman scattering via rotational transition in para-hydrogen molecules using CO, lasers. An output energy in excess of 1 J and a peak power of ~ 20 MW with a photon conversion efficiency of 85% has been ebserved,® Tunable far-infrared output down to 257 wm has also been obtained from stimulated Raman scattering via rotational transitions Q(J7) in HF using the tunable infrared output from stimulated Raman scattering in H, by a flash-pumped dye laser as a pump source.” Tunable U¥ Source Obtained from Anti-Stokes Scattering Stimulated anti-Stokes Raman scattering is possible if a Raman transition has an inverted population [p, < p, in (10.5) leading to a positive exponential gain 30 24 * 40 a 45 2 49) INVERSE RAMAN SIGNAL /410) 1 MN 908.2 908.4 9086 908.8 9030 3092 RAMAN SHIFT (om!) Fig, 10.13 Inverse Raman spectrum in the vicinity of », fundamental of CF, at 4 Torr. Fundamental and hot band transitions are labeled by J and J’, respectively. Pump and probe laser powers of 2 MW and 100 mW were used, respectively. A 3-sec time constant was used to average the 10 pps signs. [After A. Owyoung, in W. ON. Guimaraes, C. T. Lin, and A. Mooradian, eds., Lasers and Applications (Springer-Verlag, Berlin, 1981), p. 67.)168 Stimulated Raman Scattering for the E, field]. This has been suggested as a means to obtain a potentially powerful uv source which is broadly tunable. To achieve the inverted popu- Jation, it is generally necessary to choose a metastable state as the upper state in the Raman transition. The population can be pumped into the metastable state by various methods. In a recent demonstration,?’ photodissociation was used to obtain an inverted population between a metastable state and the ground state of a dissociation fragment, and stimulated anti-Stokes Raman scattering from the metastable state was then observed, For example, through AtF laser pumping, TICt was dissociated into TI and Cl. The thallium product was in the 6p?P?,, metastable state. It could reach a concentration of 4 x 10% atoms/cm? out of the original TIC] concentration of 6.9 x 106 molecules /enr. If the second- or third-harmonic output from a Q-switched Nd: YAG laser was then propagated into this photodissociated system, stimulated anti-Stokes Raman emission from 6p*P2,, to the ground state 6pP{,, of Tl was readily seen. A 10% conversion efficiency was achieved in a cell 25 em long. Even spontaneous anti-Stokes Raman scattering can be useful as a radiation source.” Using a high-lying metastable state, the emission can be tunable over narrow tegions in the XUV. Such a source can have the unique properties of AGAIN (ARB) 1570 4990 1610 1630 AV tem} Fig. 10.14 The Raman gain spectrum of a monolayer of p-nitrobenzoic acid (PNBA) on a thin film of aluminum oxide supported by sodium fluoride. Three principal features are marked. [After J. P, Heritage and D. L. Allara, Chem. Phys. Lett. 74, 507 {1980).)Stitnulated Polariton Scattering 169 narrow linewidth, ultrashort pulsewidth, and relatively high peak spectral brightness. Stimulated Raman Scattering as a High-Resolution Spectroscopy Technique The theory of stimulated Raman scattering in Section 10.2 or 10.3 shows a Stokes amplification gain G,(w,— w,) and a corresponding pump attenuation both proportional to the Raman lineshape. Therefore, measurements of the Stokes gain versus w, — w, (known as stimulated Raman gain spectroscopy) or measurements of the pump attenuation versus w,— w, (known as inverse Raman spectroscopy) should yield a Raman spectrum identical to that ob- tained from spontaneous Raman scattering. The coherent spectroscopic tech- nique, however, has two important advantages. First, no spectrometer is needed, so that the spectral resolution is limited only by the laser linewidths which can be many orders of magnitude better than the resolution of a spectrometer. It can be used to obtain high-resolution Raman spectra of gases not realizable by spontaneous Raman scattering? An example is seen in Fig, 10.13. Second, with CW mode-locked laser pulse and a locked-in detection scheme, the coherent technique can be extremely sensitive and can be used to study Raman spectra of thin films and adsorbed molecules”? Monolayer detection is possible, as has been demonstrated by Heritage shown in Fig. 10.14, 10.7. STIMULATED POLARITON SCATTERING The material excitation p,, discussed in Section 10.3 can in general be both infrared and Raman active, that is, it can be excited by both the two-photon Raman process and the one-photon infrared absorption process. This is the case, for example, with phonons in polar crystals. The direct coupling between infrared and phonon waves actually forms a mixed excitational wave which is usually known as polariton.” A typical polariton dispersion curve of a polar crystal is seen in Fig, 10.15, Because of the strong dispersion in the k ~ we! /e region, Raman scattering by polaritons shows a Raman frequency shift that depends on the scattering angle, as dictated by the frequency and wavevector matching conditions a, = o, + , and k,; =k, + k;, where w, and k, are the frequency and wavevector of the polariton. Stimulated Raman scattering by polaritons (or stimulated polariton scatter- ing) occurs when the pump excitation is sufficiently strong. It can again be described by the coupled wave approach. Four interacting waves are now involved in the problem: the pump E,, the Stokes E,, the infrared E,, and the material excitation p,,. The process can be considered a combination of the pafametric generation process discussed in Chapter 9 and the stimulated170 Stimulated Raman Scattering Real K —-~+ Magnitude of K when K is pure imaginary wor = LX 10" rad s-! Phonontike, tap = 1X 10"*rad s-? bat w, in 104 st Phononlike 0 05 1.0 13 Kin 10‘ cm" Fig. 10.15 Coupled modes of photons and transverse optical phonons in an ionic crystal. The fine horizontal line represents oscillators of frequency w, in the absence of coupling to the electromagnetic field, and the fine line labeled w = cK/ Je(co) corresponds to electromagnetic waves in the erystal, but uncoupled to the lattice oscillators at w7. The heavy lines are the dispersion relations in the presence of coupling between the lattice oscillators and the electromagnctic wave. One effect of the coupling is to create the frequency gap between #, and ,; within this gap the wavevector is purcly imaginary of magnitude given by the broken line in the figure. [After C. Kittel, Introduction to Solid State Physics, Sth ed, (Wiley, New York, 1976), p. 304) Raman process discussed in previous sections, The wave equations for the four waves") are Anu; v x(v x)- =o xO" E,E, + NMZE,p,,], Anw? vw x(v x)- = 2, [xR EF + NM Evi]. mag (10.34) 2, [xEE} + NAtp, |, 2 and a... i . (= + fay, + Tye = ql Anes + My E,ET (9; — 0) where x? is the usual second-order nonlinear susceptibility and A); = (f| — er él) is the transition matrix element for the infrared excitation ofStimulated Polariton Scattering in the material system from |/) to {/). The third-order nonresonant terms xGR{£|7E, in (10.34) have been omitted for simplicity. Tf the response of Pp: is stationary, then by eliminating p,,, (10.34) reduces to a Anu? , . . [ext 9 - Shae, “5s [ree + PEPE, lr x(v x) ~ ae, = reo XIE EY +x@1ere,], (10.35) v x(v x) ~ bo ales = Ane a REE? where xn x9 — Molo Be) (ws — wy, + iT) we NIM,1?(9; ~ 9) (10.36) h( oy ~ wy, + Tp) NiApl?(0, — 0) 83 et = 83 — Soo var Aas — a, + D,) where k; = («4/c)eh/2 actually describes the polariton dispersion curve. The solution of (10.35) again resembles that of parametric generation. With a plane boundary at z = 0, and assuming no pump depletion, it takes the form E* = [8% exp(iAK,z) + &* exp(iAK_z)]exp(—ik,-1), (10.37 Ey = [ 4, exp(iAK,z) + &_exp(iAK_z)]exp(ik, +e + idkz), ) where ® k= oles, ay”, Aka ky ~ky— ky ky eke dy ak .= Hy, —») 44 [Oy + 9) - 4a], k (10.38) Ys = ag Ula, + 2ke)s to? Ke = 4 exRlE 2k,,¢7 272 Stimulated Raman Scattering ke y= Ak ile). an2wta - ay 2 SEH OP Ey? Ok, Ks, xy EW", 2 &, ai, wl=|4" M(AK,+ ¥%)I- 3 (es WA(aK + nh As one would expect, the solution here reduces to that of simple Raman-Stokes generation and parametric amplification, respectively, when the coupling of the infrared field E, to the other waves vanishes [4,, = 0 and x® = 0] and when the nonlinear coupling of p,, with E, and E, vanishes (M,, = 0). Numerical examples on GaP and LiNbO, can be found in Henry and Garrett and in Sussman. By adjusting properly the relative angle between k, and k, to tune along the polariton dispersion curve, the output of stimulated polariton scattering is tunable, This has been demonstrated in LiNbO.” In a resonator, up to 70% of the laser power can be converted into Stokes. The infrared output is tunable from 50 to 238 cm, Its peak power was found to be 5 W when a 1-MW pump beam from a Q-switched ruby laser with a beam diameter of 2 mm was focused into a 3.3-cm LiNbO, crystal by an f = 50 cm lens. This output is limited because LiNbO, has a low damage threshold. in practice, single-mode lasers should be used to avoid hot spots in the beam which increase the damage probability. Observation of far-infrared output from stimulated polariton scattering in quartz has also been reported. 10.8 STIMULATED SPIN-FLIF RAMAN SCATTERING An alternative way to obtain tunable output from stimulated Raman scattering is to use a fixed pump frequency and tune the resonant frequency of the material excitation. Stimulated Raman scattering between Zeeman levels is an example: the resonant frequency is tumed by the applied magnetic field. Unfortunately, the tuning range is usually small. The Zeeman splitting 24,33 = gB/21.4 em—', with B in kOersted, is only ~ 1 em™* for g = 2 and B = 10 kOersted. In some solids, however, the effective g factor can be much larger, for example, g = 50 in InSb. Then, when B is varied from 0 to 100 kOersted, the Zeeman splitting can be tuned over ~ 240 em~?. This is a reasonably broad tuning range. It happens that InSb is also a good Raman scatterer in the infrared. Zeeman levels of band electrons in semiconductors are usually known as Landau levels, and Raman scattering between spin-up and spin-down states is called spin-flip Raman scattering, described schematically in Fig. 10.16. Wolff and Yafet**Stimulated Spin-Flip Raman Scattering 13 s=-Ve s= V2 hay fk hwol ka) Fig. 10.16 Schematic of spin-ftip Raman process in n-InSb. showed that the spin-flip Raman cross section is given by do) _{ e \Vyo\f_ Behe (alee ~ (5] (2s — Pe} (10:38) for pump and Stokes polarizations perpendicular to each other, where m, = 2m/|g| is the spin electron mass, m is the natural electron mass, and E, is the energy gap. In InSb, we have g = 50 and £, = 1900 om™!, Then, with E,he,/(Ey — fw?) ~ 1, the spin-flip Raman cross section is already ~ 600 times larger than the Thomson scattering cross section for free electrons, which is (e2/me?¥? ~ 10°% cm?/ster. This is what was actually observed in InSb with a CO, laser (w, = 940 om"). As a comparison, the Raman cross section for the strongest mode of benzene at 992 cm™! is 3x 107% cor’/ster in the visible. For Aw, = E,, (do/d@)cp can be even much larger as a result of resonant enhancement, as seen in Fig, 10.17. With a CO laser at w, = 1800 m7, (da/dQ)gp was found to be ~ 105 times stronger than the Thomson scattering cross section. According to (10.8), the stimulated Raman gain Gp is proportional to N(do/dQ)T~', and the density N here for spin-flip scattering refers to the carrier density. In the case of n-type semiconductors (Fig. 10.16), N is the electron density in the conduction band. As a result, G, here is badly hurt by N, which is always much smaller than the atomic or molecular density in condensed matter, Even so, form > 10'/en’, it is still larger than those in other condensed media if the halfwidth I is assumed to be the same. Actually, T of a-InSb is quite narrow at low temperature and depends on the carrier concentration and k,+H. It can be as narrow as 0.15 em”! with n= 1x 10° /cor’,?5 Assuming T = 2 cm7}, N = 3 x 10" /em?, and p, — p= 3 in (10.8), we find in n-In$b a spin-flip Raman gain Gy = 1.7 x 10° °Y cm~’, where J is the174 Stimulated Raman Scattering Incident photon wavelength (am} 6.1 6.0 6.95.8 57 56 5.5 54 53 5.2 Relative cross section 200 205 210 215 220 225 290 235 240 incident photon energy (meV) Fig. 10.17 Resonance enhancement of spontaneous spin-flip Raman scattering as a function of input photon energy (n = 1 X 10'S cm~*, H = 40 kOe, and T ~ 30 K). {Alter §, R. J. Brucck and A. Mooradian, Phys, Rev. Lett. 28, 161 (1972).] CO, laser intensity in watts per square centimeter. This is the largest known Raman gain for all materials. The gain can be increased further by adjusting V properly to yield an optimum value for NI? and by moving fw, closer to £,. At the CO laser frequency, the gain becomes 6 x 10~“f cm7'. From these estimates, one expects that stimulated spin-flip Raman scattering should be observable in InSb of a few millimeters in length with a pump beam of ~ 10° W/crr at 10.6 um or ~ 104 W/crr at 5.3 pm. In practice, optical feedback at the air~sample interfaces can result in Raman oscillation. Patel and Shaw,” using a Q-switched CQ, laser of 1 kW at 10.6 4m focused to a spot of 10-? cm? in a 5-mm n-InSb sample with N = 10™*/cm’ at T = 18 K, observed a Stokes output of 10 W. The output was tunable from 10.9 to 13.0 pm with BStimulated Spin-Flip Raman Scattering ms varied from 15 to 100 kOrsted. The output linewidth was less than 6.03 cm™?. Using a single-mode CW CO laser at 5.3 pm focused to ~ 5 x 107° cm’ ina 48mm x-InSb sample with n= 10'*/cm> at T= 30 K, Brueck and Mooradian®*® obtained a Raman oscillation threshold at a pump power less than 50 mW, a power conversion efficiency of 50%, and a maximum Stokes output in excess of 1 W. The output linewidth can be as narrow as 1 kHz. With samples in a low magnetic field, a conversion efficiency of 80% has been achieved.” Anti-Stokes radiation and Stokes radiation up to the fourth order have also been observed. Detailed operation characteristics of InSb spin-flip Raman lasers are listed in Ref. 38. Stimulated spin-flip Raman scattering can also occur in other semiconduc- tors. Among those reported in the literatures are CdS pumped by a dye laser, InAs pumped by an HF laser, and Pb,Sn,_,Te, Hg,Cd,_,Te, and Hg,Mn,_,Te pumped by a CO, TEA laser. Raman Gain ( em’) [P (w.)/ P(wa)] x 10° [P(g / Play) ] x10? (Non collinear, Phase- Matched ) ( Collinear, Non- Phase- Matched ) -6 -4 =z ° 2 4 6 (w3- 9) / 7 Fig, 10.18 Theoretical curves of the Raman gain g, and the ratios of the far-infrared output P(.;) to the Raman output P(w2) for the collinear phase-mismatched case and for the noncollinear phase-matched case. (After Ref. 37.)176 Stimulated Raman Scattering Actually, the spin-flip uansition can be excited by both the Raman process and the direct one-photon absorption process, although the latter is weak because it is a magnetic-dipole transition. Therefore, strictly speaking, stimu- lated spin-flip Raman scattering is a special case of stimulated polariton scattering,” and the theory developed in the previous section should apply here. In addition to the Stokes output, a far-infrared output at the spin-flip transition frequency is expected. In the present case, the calculation is, how- ever, relatively simple, because the free carrier absorption at «, is quite strong and we always have (y, + ¥,)” > A in (10.38). As a result, the gain is almost exactly equal to the stimulated Raman gain with A = 0. We show in Fig. 10.18 a calculated example of both the gain and the ratio of the far-infrared output to the Stokes output. Collinear phase matching is not possible in this example, so that the far-infrared output in the forward direction is relatively low. It becomes stronger in the noncollinear phase-matching direction. In fact, be- cause the direct excitation of the spin-flip transition is weak, the far-infrared output can be calculated through iteration by first finding the Stokes output from stimulated Raman scattering and then the difference-frequency output from the pump and Stokes mixing. Far-infrared output from a spin-flip Raman oscillator has not yet been reported. Only the collinear phase-mismatched case has been tried. Far-infrared generation by optical mixing of pump and Stokes waves in InSb has, however, been observed with its maximum appearing at the spin-flip resonance. The results agree very well with the theory. This far-infrared output is, of course, tunable over the same range as the Stokes output, and constitutes a potential tunable coherent source in the far-infrared which can be both intense and narrow in linewidth. 10.9 TRANSIENT STIMULATED RAMAN SCATTERING Pulsed lasers often are used in stimulated Raman scattering experiments. We must therefore consider the time dependence of the oytput. If the pulsewidth is much longer than the relaxation time of the Raman excitation and the time required for light to traverse the medium, we can expect from physical argument that the output pulse will follow the temporal variation of the input pulse. This is the quasi-steady-state case. Otherwise, the output should exhibit a transient behavior. To describe the transient effect, we should, in general, use the coupled wave approach of Section 10.3. In this approach, the dynamic equation for the material excitation explicitly takes into account the possible transient response of the medium, Even in the case of strong Raman gain, the slowly varying amplitude approximation of the fields is usually still valid. For Stokes genera- tion in the forward direction along 2, the set of coupled equations in (10.22),‘Transient Stimulated Raman Scattering 7 following the derivation of Section 3.5, can be written as a { 221 * (e+23 a &,(z,1) %, NMRE,A éo1 220} + ($+ ty g)at, t)= f hy aon (10.40} ; . (5+ Ta) at eee) = paler aeee with E,(«,) = &(z, Dexplik;z — ijt) and p,,(o, — w,) = A(z, Hexpli(k, — k,)z — i(w, — @,)f] in (10.22). Here, v, and v, are the group velocities at w, and w,, respectively, and we have neglected for simplicity the nonresonant driving terms in the equations of & and 4. Consider first the simpler case where the amplitude variations of é, and are sufficiently slow so that |@4/dt| is negligible compared to |['A|, Then A(z, 1) = iM,,(0; — py)é,é7 /AT,,, and if we assume for simplicity 9, = »,, and use the transformation of variable 2’ = z and t' = i — z/v, (10.40) reduces to ex? any jer. e ar ‘ (10.41) siei dma} xPlIA PE dz’ ck, where x? = N[M,|’(p, — 9,)/iRT. These equations are identical to (10.15) with a = 0 except that @, and & are now functions of z and ¢ — z/v. In other words, , and & should obey the steady-state solution in the retarded time coordinate. Physically this result follows from the fact that a differential section of the laser pulse always interacts with one and the same differential section of the Stokes pulse throughout the medium. This is the quasi-steady-state solution. For backward scattering, we must replace #, in (10.40) by —»,, and (10.41) is no longer valid. The quasi-steady-state solution applies only when the amplitude variations of the input pulses are negligible in the time duration for light to traverse the entire length of the medium. However, the general solution for this case can still be found for |v,| = |2,| as p42 ,—-2\e IP (e+ 2/u,0) leaP(e+ 54 a) Ee te AO ee Bi f)= [ate Pe, (10.2) Fest B(r+2 )-f f hb, 9,0) dy,mg Stimulated Raman Seattering and 4703 (48) Tt reduces to the quasi-steady-state solution when both the laser and the Stokes input pulses are so long that the integrands in F, and F, can be approximated by constants. On the other hand, if both input pulses are much shorter compared to. the time for light to traverse the length of the medium, then the backward Raman amplification is greatly reduced in comparison with the forward Raman amplification due to the limited interaction range of the laser and the Stokes pulses. This is seen in (10.42) that y(t ~ 2/v) < gi&(f — z/v)|4, where fis the length of the medium. Another case of interest is the amplification of a backward Stokes pulse in a tong medium against a relatively long laser pulse. If the leading edge of the laser pulse is sufficiently steep, Stokes pulse sharpening may occur as the wavefront of the backward Stokes pulse continuously sees the fresh undepleted incoming laser beam and gets full amplification while the tagging part of the pulse does not." An example is seen in Fig, 10.19. This pulse-sharpening phenomenon can be observed in some liquids when the initial Stokes pulse is generated by self-focusing near the end of the cell, We now consider the more general case where |¢.4/@t| is no longer negligible compared with |I'4|. This happens when 4, 67 varies rapidly so that the material excitation cannot respond instantaneously, or more quantitatively, when the laser pulsewidth 7, is smaller than or comparable with the dephasing TIME tot giao 4, PULSE INTENSITY LASER INTENSITY @ 7 6 . 2 t (4 t.36cin Fig. 10.19 Calculated normalized Raman pulse intensity as a function of time for an initial condition |E, = |E,ol(f — fq) for t > fo. The curves show the pulse development at length intervals of A/ = 2.77/G. Gis the Raman gain and was determined to be 0.7 cm7! in CS,. Lower scale is in dimensionless units; upper scale describes the experi- mental conditions. (After Ref. 39.)‘Transient Stimutated Raman Scattering 19 time 7, = 1 /T, of the Raman excitation (or more correctly, as we shall see later,* when 7, < Gault, where Gp, is the steady-state Raman gain from (10.8) by assuming the laser intensity to be the peak intensity of the input pulse and / is the length of the medium). Although Q-switched pulses may be short enough for studying transient stimulated Raman scattering in gases, picosec- ond pulses are needed for liquids since 7, is usually of the order of picosec- onds. With a picosecond pump pulse, the backward Raman scattering is hardly detectable because of the very limited length of interaction between the backward Raman and the incoming pump pulse. We discuss here only the forward Raman scattering. Consider the case where both the depletion of pump power and the induced population change are negligible. Then (assuming v, = v,) (10.40) reduces to (Z + aL = méi{t - 2) a, zea ° (10.43) a . Zz (a+ rat = -inez(s-2)6, where Milo; 0) n= and #\(1 — z/v) is given by the initial condition, The solution of (10.43) describes the transient stimulated Raman scattering. Its derivation is somewhat long and tedious. We shall therefore only sketch the result here and refer the readers to Ref. 42 for details. With ¢ = ¢ — z/o and z' = z, (10.43) can be transformed into a second-order partial differential equation 2 [aoe -amietne]o~o (10.44) where U = Fexp(Iv’) and F stands for either 6, or A*. The equation can be further simplified to a? (sar - nim|U =0 (10.45) by defining + = f% )4,(2”)[? de”. Equation (10.45) is in the standard form of a hyperbolic equation which can be solved with arbitrary initial conditions. In180 Stimulated Raman Seattering the present case, the solution takes the form (24,0) = 800.0) Hmm) 8) xf Pear VE) — x42 arta e(1) - eee} )}} dt”, (10.46) wea) ein fh Pe aE 0,07) x Jg(2[maaCr(") - a(t’))2']'7) } at”, where the input conditions are A*(z’) = Oat! > — 0 and 6(2', 1°) = 60, 7) at z =z’ = 0, and J, is the ith order Bessel function of imaginary argument. The solution in (10.46) has the following characteristics: 1) Since fy(x) = 1 and /,(x) = x for x « Land I(x) = Qax)” Fexp(x) for x » 1, the Stokes amplitude & first increases linearly with z and then, in the limit of large amplification, increases exponentially in the form Sle PECL BIE LER) — AT “38 (10.47) xexp{ P=) + nme (r) ~ e277}. 2 If the pump pulse is sufficiently long, then &, takes on a quasi-steady-state exponential gain. This can be seen from (10.47) when (4 — t9) > Gp2T; for a rectangular pump pulse starting at fo. For this reason, 7, < GamlT, can be used as the condition for the observation of transient stimulated Raman scattering, as mentioned earlier. 3 HT, < Ff, the factor exp[—T(r’ — 7’)] can be approximated by 1 in the integrals of (10.46) and (10.47). The Stokes amplitude grows rapidly only toward the middle part of the pump pulse. It then drops off following the pump pulse at the tail, The Stokes pulse is therefore always narrower than the pump pulse. The material excitation A behaves in the similar manner but has an exponential decay tail exp( ~T1) after the pump pulse is switched off or dropped to nearly zero. 4 In the limit of large amplification, (10.47) gives (Ena ex0( SF) (10.48)Mensurements of Excitational Relaxation Times 181 12 z| , ae), =f aor where Gy is a transient gain given by Gra alain) (10.49) The transient gain here is independent of the laser pulseshape. For a pulsc of the form &,(1') = &,exp(—|1'/T|"), the peak of the Stokes pulse is delayed from the peak of the pump pulse by a time r, = T(4log Gam)!" Numerical calculations of transient stimulated Raman scattering have been carried out by Carmen et al. for various pump shapes. The results confirm the characteristic features presented above. Transient behavior of stimulated Raman scattering was first noticed in gases by Hagenlocker et al.” Later, with picosecond pulses, it was also observed in liquids. Quantitative measurements have shown goed agreement with theoreti- cal predictions.“ A better experiment yet to be carried out is 10 measure the temporal variation of Stokes amplification in an amplifier cell (see Section 10.6C). The transient gain G, is different from the steady-state gain G, in the fact that the former depends only on the Raman cross section (& 7,1;), while the latter is also inversely proportional to the halfwidth T. Therefore, it is possible to observe in transient stimulated Raman scattering some Raman modes which are suppressed in the steady-state case. More than one Raman mode can in fact simultaneously show up in transient Raman scattering.” 10.10 MEASUREMENTS OF EXCITATIONAL RELAXATION TEMES Relaxation of a material excitation can be measured directly by probing the decay of the excitation. In analogy to the magnetic resonance cases, two relaxation times are often used to characterize the relaxation process: the longitudinal relaxation time Jj, which governs the decay of the induced population change in the excited state, and the transverse relaxation time 7), which is the dephasing time of the excitational wave (see Section 2.1). In condensed matter, 7, and 7; are usually of the order of picoseconds, and therefore picosecond pulses are needed to excite and to probe the material excitation in the measurements of T, and T,. We consider here only Raman- allowed excitations, with both excitation and probing achieved through Raman transitions. The general principle, however, is applicable to other types. of transitions. We also note that only in the limit of homogeneous broadening is T, equal to the inverse halfwidth, but even then 7, can be very different from qT.182 Stimulated Raman Scattering, In the preceding section, it was seen that transient stimulated Raman scattering yields a material excitational wave 4 which decays exponentially as exp(—i/T;) even after the pump pulse is switched off. This material excita- tional wave at (@, — @)) = 4, can be probed by mixing 4 with a probe pulse Ey at k, and «w, to generate a coherent anti-Stokes wave E, = &exp(ik,*t — ja,t) with k, =k, — kz + ky and a, = @; — a2 + 43. The equation govern- ing the forward anti-Stokes pulse amplitude is 2 270, a.1¢@ =} (3+ 2 Fale = i ok, from which we find, with the help of transformation of variables 2° = 2 and 1 =1-—2/v,, the solution Ja aile4C.0 (10.50) 6, (1,1) « fez, vyA(z’, 1’) de’. o ‘The time-integrated coherent anti-Stokes signal is therefore given by Sra & f 1q( bP at 00 (10.51) « { [fac 1) A(z‘, t'} de" P de, This signal is, of course, a function of the time delay #, between the exciting and the probing pulses. If /p > 7, > pulsewidth 7,, then it is clear from (10.51) that So, © exp(—2tp/T)). Therefore, F, can be easily deduced from the result of Sig, versus fy. More rigorously, the effect of the finite pulsewidth T, should be taken into account in the time convolution in deducing 7. The longitudinal relaxation time T, describes the decay of the induced population change Ap as governed by (10.23) or, more explicitly, (s + a = Fl Mphere - Mpazé,4](o! - of) (10.52) with Ap = $(p,— pp of + OP) ® (p? — p?). The equation shows that after the pump pulse is over Ap should decay exponentially as exp(—1/T,). Incoher- ent (spontancous) anti-Stokes scattering is directly proportional to Ap and therefore can be used to probe the decay of Ap. With a probe pulse F, at #2, the time-integrated signal at w, = w, + u,, is given by Sigg & fleet. 1)PAp(z, t) dz dt, (10.53)Measurements of Excitational Relaxation Times 183 which is a function of the time delay #, between the exciting and the probing pulses. When fp > 7, 2 T,, this signal is proportional to exp( ~ fp/T,) so that T, can be deduced from Sj,, versus fp. The method was first used by DeMartini and Ducuing® to measure 7, of the 4155 cm™! vibrational excitation of gaseous H,. At 0.03 atmospheric pressure, T, = 30 psec, so that Q-switched laser pulses are short cnough to AMPUFIER switcn }—=——{M-1 LASER TO SCOPE TO SCOPE TO SCOPE tee TO SCOR Ps Kor) pap 3 vO I Wyo” | Fe . Be \) 0 SCOPE fl FO RS. ne Fig, 10.20 Schematic of the experimental system for photon lifetime measurement. The pump beam B1 at A ~ 1.06 4m and the probe beam B2 at A = 0.53 pm interact in the Raman sample RS. Glass rod for fixed optical detay, FD; glass prisms for variable delay, WD; filter, F; photodetector, P; two-photon fluorescence system, TPF. [After A. Laubereau, D. von der Linde, and W. Kaiser, Phys. Rev. Lett. 28, 1162 (1972).] 1 3 en a | Sinc(zp) °—~, 3 t \ Boh | ‘ g I Z \ Seah (tp) & | \ ! \ Boy I \ T, = 20+ 5 psec no | oy T/T, 780 I ‘ Ly 0 10 20 30 40 Delay time tp (psec) Cig. 10.21 Measured incoherent scattering S,,,(#p)/Sincgay (Closed circles) and coher- ent scattering §..4(fp)/S.on,,,.» (open circles) versus delay time t, for ethy] alcohol. The : solid and dashed curves are caleulated. [After A. Laubereau, D. von der Linde, ané W. Kaiser, Phys. Rev. Lest. 28, 1162 (1972}]184 Stimulated Raman Scattering carry out the measurements. In condensed matter, however, 7, and 7; are in the picosecond range, and picosecond mode-locked laser pulses must ‘be used, as pioneered by Alfano and Shapiro and by Kaiser and associates.” Figure 10.20 is a typical experimental arrangement. The mode-locked pulse from an Nd laser is used to excite the material excitation by stimulated Raman scattering, and the second harmonic of the mode-locked pulse, after an adjustable time delay, is used to probe the excitation. The results on ethyl alcohol in Figure 10.21 is an example. The exponential tails of the S,4,(7p) and Sinc(tp) curves in the figure yield T, and T, readily. The technique can be extended to the study of decay routes of an excitation, and the dephasing dynamics of an excitation in large molecules or condensed matter.” Note, however, that this discussion of dephasing time measurements applies only to a homogeneously broadened Raman transition. In the case of inhomogeneous broadening, the material excitation has a distribution of reso- nant frequencies w,,, and A(z, t) in (10.50) and (10.51) should be replaced by an integration of the excitational waves over the distribution of w,,. The decay of S,,, with time is no longer in the form of exp(— 2tp/T,). If the width of the inhomogeneous broadening is considerably larger than 1/7,, the decay time of Sion Will be dominated by 7,; then no information about 7, can be obtained. It is, however, possible that when the pump pulse is so intense as to cause coherent saturation in the Raman transition, the 7, value can still be deduced from the decay of S,,,,.°° More details on theory and experiments of vibrational relaxation studied by ultrashort pulses can be found in Refs. 47 and 48. REFERENCES E. J. Woodbury and W. K. Ng, Proc. [RE 50, 2347 (1962). EJ, Woodbury and G. M. Eckhardt, U.S, Patent No, 3,371,265 (27 February 1968). R. W. Hellwarth, Phys. Rev. 130, 1850 (1963); Appl. Ope. 2, 847 (1963). E. Garmire, B, Pandarese, and C. H. Townes, Phys. Reo. Lett. 11, 160 (1963). N, Bloembergen and Y. R. Shen, Phys. Rev. Lett. 12, 504 (1964); Y. R. Shen and N. Bloembergen, Phys. Rev. 137, A1786 (1965). 6 SX. Korte and J. A. Giordmaine, Phys. Rev. Lett. 22, 192 (1969); I. Gelbwachs, R. H. Pantell, H. E. Putboff, and J. M. Yarborough, Appi. Phys. Lett. 14, 258 (1969). 7 C.K.N. Patel and E. D. Shaw, Pays. Rev. Lett. 24, 451 (1970); Phys. Rev, B3, 1279 (1971). 8 A. Qwyoung, [EEE J. Quant. Electron. QE-14, 192 (1978); in H, Walther and K. W. Rothe, eds,, Laser Spectrascopy, vol. IV (Springer-Verlag, Berlin, 1979), p. 175; in W. O. N. Guimaraes, CT. Lin, and A. Mooradian, eds, Lasers and Applications (Springer-Verlag, Berlin, 1981), p. 67 9 D.vonder Linde, A. Laubereau, and W. Kaiser, Phys. Rev. Lett, 26, 954 (1971); R. R. Alfano and S, L, Shapiro, Pays. Rev. Lett. 26, 1247 (1971). 10 See, for example, W. Heitler, Quaattun Theory. of Radiation, 31d ed. (Cambridge University Press, Cambridge, 1954), p. 192. IR. Glauber, Phys. Rev. 130, 2529 (1963); 131, 2766 (1963). we eneReferences 185 R. W. Terhune, Solid State Design 4, 38 (1963); R. ¥. Chiao and B. P. Stoichelf, Phys. Rev. Lett, 12, 290 (1964); H. 5. Zeiger, P. E, Tanneawald, $. Kem, and R. Burendeen, Phys. Rev. Lett. U1, 419 (1963). D. von der Linde, M. Maier, and W. Kaiser, Phys. Rev. 178, 12 (1969), G. Haidacher and M. Maier, VII Int. Quant. Electron. Conf., San Francisco (1974), post- deadline paper Q.7. N. Bloembergen, G. Bret, P. Lallemand, A. Pine, and P. Simova, [EEE J. Quant. Electron QE-3, 197 (1967); P. Latlemand, P. Simova, and G. Bret, Phys. Rev. Leu. 17, 1239 (1966). E, Garmire, Phys. Lett. 47, 251 (1965). N. Bloembergen and Y.R. Shen, Phys. Rev. Leis. 13, 720 (1964); R. L. Carman, F. Shimizu, C.S. Wang, and N. Bloembergen, Phys. Rev. A2, 60 (1970); S. A. Akhmanov, Yu. E. D'yakov, and L. 1. Pavlov, Sov. Phys. JETP 39, 249 (1974); M. G. Raymer, J. Mostowski, and J. L. Carlsten, Phys. Rev, AMY, 2304 (1979); LA. Walmsley and M. G. Raymer, Phys. Rev. Lert. 50, 962 (1983). W. J Jones and B. P. Stoicheif, Phys. Rev. Lett. 13, 657 (1964). P. P. Sorokin, N. §. Shiren, J. R. Lankard, E, C. Hammond, and T. G, Kazyaka, Appl. Phys Lett, ¥2, 342 (1973); M. Rokni and §. Yatsiv, Phys. Lett, A, 277 (1967); J. J. Wynne and PP. Sorokin, in Y. R. Shen, ed. Nonfinear Infrared Generation (Springer-Verlag, Berlin, 1977), p. 159. D.C. Hanna, M. A. Yuratich, and D. Cotter, Nonlinear Optics of Free Atoms and Molecutes (Springer-Verlag, Berlin, 1979), Chapter 5, and the references therein D. S, Bethune, J. R. Lankard, and P. P. Sorokin, Opr. Lert. 4, 103 (1980). R. Frey and F. Pradere, Opt. Comm. 12, 98 (1974); V. Wilkie and W. Schmidt. Appi. Phys. 16, 151 (1978); T. R. Loree, R. C. Sze, D. L. Barker, and P. B. Scott, /EEE J. Quant. Eieciron. QE-15, 337 (1975). R. Frey and F. Pradere, Opt. Lett. 5, 374 (1980). J, J. Tice and C. Wittig, Appl. Phys. Lett. 30, 420 (1977). R.L. Byer, EEE J. Quant. Electron. QE-12, 732 (1976); R. L. Byer and W. R. Trunta, Ops. Lett, 3, 144 (1978); P. Rabinowitz, A. Stein, R. Brickman, and A. Kaldor, Appi, Pays. Let. 35, 739 (1979). R. Frey, F. Pradeze, J. Lukasik, and J. Ducuing, Opt. Comm. 22, 455 (1977); R. Frey, F. Pradere, and J. Ducuing, Ops. Comm. 23, 65 (1977); A. DeMartino, R. Frey, and F. Pradere, Opt. Camm, 27, 262 (1978). J.C. White and D. Henderson, Phys. Rev. A 25, 1226 (1982); J. C. White, in H. Weber and W. Luthy, eds., Laser Spectroscopy, V1 (Springer-Verlag, Berlin, 1983). S. E, Hartis, Appl. Phys. Lett. 31, 498 (L977); L. J. Zych, L. Lugasik, J. F. Young, and 8. E Harris, Phys. Rev. Lett, 40, 1493 (1978); S. E. Harris, R. W. Falcone, M. Gross, R Normandia, K. D. Pedrotti, J. E, Rothenberg, J.C, Wang, J. R. Willison, and J. F. Young, in AR. M. McKellar, T. Oka, and B. P. Stoicheff, eds., Laser Spectroscopy, V. (Springer-Verlag, Berlin, 1981); 8. E. Harris, R. G, Caro, R. W. Faloone, D. E. Holmgren, K. D. Pedrotti, J.C. ‘Wang, and J. F. Young, in H. Weber and W. Luthy, ods., Laser Speciroscopy, VI, (Springes- Verlag, Berlin, 1983). 3B. F. Levine, C. V. Shank, and J.P. Heritage, FEEE J. Quant. Electron. QE-15, 1418 (1979); J.P. Heritage and D. L. Allara, Chem. Phys. Lets. 74, 507 (1980); B. F. Levine, C. G. Bethea, A. R. Tretola, and M. Korgor, Appl. Phys. Lett. 37, 595 (1980), K. Huang, Proc. Roy. Soc. (London) A208, 353 (1951); M. Bom and K. Huang, Dynamic Theory of Crystal Lattices (Oxiord University Press, Oxford, 1954), Chapter II; I. I. Hopfield, Phys. Rev. 112, 1555 (1958). R. Loudon, Proc. Phys. Soc. BE, 393 (1963); Y. R. Shen, Phys. Rev. 138, A1741 (1965)186 Ky} Rod ai a ee a ad Stimulated Raman Scattering C. H. Henry and C. G. B, Garrett, Phyy. Rev, 171, 1058 (1968); §. S. Sussman, Microwave Lab Report No. 1851, Stanford University (1970), J. M. Yarborough, S. §. Sussman, H. E. Puthoff, R. H. Pantell, and B. C. Fohnson, Appl. Phys: Lett, V5, 102 (1969), PA Wollf, Phys, Rev. Lett, 16, 225 (1966); Y. Yafct, Phys, Rev. 152, 858 (1966). 5. R. J. Brueck and A. Mooradian, Phys. Rev. Lett, 28 1458 (1972). A. Mooradian, SR, J. Brueck, and F, A. Blum, Appl, Phys. Zett. 17, 481 (1970); S. RJ. Brueck and A. Mooradian, Appl. Phys. Lett. 18, 229 (1971). C.S, DeSilets and C.K.N. Patel, Appé. Phys, Lett, 22, 543 (1973); C.K. N. Patel, Phys. Rev. Lett, 28, 649 (1972). HL Pascher, G. Appold, R. Ebert, and H. G, Hafele, Appl, Phys. 15, 53 (3978); B. Walker, G. W. Chantry, D. G. Moss, and C. C. Bradley, J. Phys. D9, 1501 (1976); M. J. Colles and C. R. Pidgeon, Rep. Prog. Pays. 38, 329 (1975); S. D. Smith, R. B, Dennis, and R. G. Harrison, Prog. Quant, Efeciron, 5, 205 (1977). Y.R. Shen, Appl. Phys. Lett, 26, 516 (1973). Y. T. Nguyen and T, J. Bridges, Phys. Rev, Lett. 29, 359 (1972); Appl. Phys. Lett. 23, 107 1923} M. Maier, W. Kaiser, and J. A. Giordmaine, Phys. Rev. 177, 580 (1969). R. L, Carmen, F. Shimizu, C. S, Wang, and N. Bloembergen, Pays. Rev. A2, 60 (L970); C. S Wang, in H. Rabin and C. L. Tang, eds, Quantum Electronics, vol. 1 (Academic Press, New York, 1975), p. 447. E. E, Hagenlocker, R. W. Minck, and W. G. Rado, Phys. Rev. 184, 226 (1967). BR. L. Casman and M. E, Mack, Phys. Rev. AS, 341 (1972). R. L, Carman, M. E. Mack, F, Shimim, and N. Bloembergen, Phys. Rev. Lett, 23, 1327 (1969); M. C. Mack, R. L. Carman, R. Reinsjes, and N. Bloembergen, Appl. Phys. Lev, 16, 209 (1970). F. DeMastini and J. Duening, Phys, Rev, Lett, 17, 117 (1966). A. Laubereau and W. Kaiser, Rev. Mod. Phys. 50, 607 (1978). W, Zinth, Appl, Phys. B 26, 213 (L981). S. M. George and C. B, Hares, Phys. Rev. A28, 863 (1983). BIBLIOGRAPHY Bloembergen, N., Amt. J. Phys. 35, 989 (1967). Shea, Y. R., in M. Cardona, ed., Light Scattering in Solids (Springer-Verlag, Berlin, 1975), p. 275. Kaiser, W., and M. Maier, in F. T. Arecehi and E. O. Schulz-Dubois, eds., Laser Handbook (North-Holland Publishing Co., Amsterdam, 1972), p. 1077.ll Stimulated Light Scattering In Chapter 10 stiraulated Raman scattering was treated as a result of paramet- tic interaction between light and material excitation. Examples were given in which the material excitation was either electronic or vibrational. The Raman shift involved could in principle range from zero to a frequency as barge as the pump laser frequency. Some material excitations have very low frequencies, of the order of <1 cm7!. They are usually related to atomic or molecular motion. Light scattering by such materia! excitations frequently is called something other than Raman scattering. For example, Brillouin scattering involves acoustic wave excitation, Rayleigh scattering deals with entropy wave excitation, and Rayleigh-wing scattering relates to molecular orientational excitation.’ With sufficiently high pump laser intensity, all these spontaneous light scattering processes could become stimulated. Some of them are discussed briefly in this chapter. 11.1 STIMULATED BRILLOUIN SCATTERING Stimulated Brillouin scattering results from parametric coupling, between light and acoustic waves. The theory follows the general formalism given in Section 10.3 for stimulated Raman scattering, with the material excitational wave here referring to the acoustic wave. We consider only Brillouin scattering in liquid. The coupled wave equations, similar to (10.22), are 2, g Ane, [r x(¥ +8 She, ~ ZT PME (a) and (14.1) 2, Aned [- x(v xt 2 =e = ar PM) 187188 Stimulated Light Scattering together with the driven acoustic wave equation a é to? 3a eae VilAp= -U'E (11.2) We use the density variation Ap to describe the acoustic wave; o is the acoustic velocity, and [, is the acoustic damping coefficient or the halfwidth of the Brillouin line in spontaneous scattering. The driving force f and the nonlinear polarizations PN! arise from the nonlinear coupling of the three waves and can be obtained as follows: a [4E. 1 de PM (a) = 35 | ap = de SEE, b0: 1 de PM (02) = ga Gp ear: (11.3) 1 f=vp= o( pre: Et] where p is the electrostrictive pressure, y = py Ge/9p is the clectrostrictive coefficient, and py is the mass density of the liquid. The problem is just another example of parametric wave interaction, and the solution of the coupled wave equations (11.1) and (11.2) follows those described repeatedly in Chapters 9 and 10. We consider here only backward stimulated Brillouin scattering in the steady-state case, The forward Brillouin scattering does not occur as it has a zero frequency shift. Let E, = 8, exp(ik,z — ian,t), Ey = 2,8, exp(—ik,2 — a,f), and Ap = A exp(ik,z — iw,t), with a, = a, + o, and k, = w,/». Following the slowly varying ampli- tude approximation, (11.1) and (11.2) reduce to ies? (e+5)4 (ot 9: (at. 3), aeneh 2] ake Op 88) ge 9 BEC a \ande Be: (a2 3) 8 aa wll @,)8 tae *™, a4) @ Tay, Ke | f 88 \ iat. 5 \te gueink: (Z+ d= eo ip (at - 2)" oe where Ak = k, + k,—k,. This is analogous to the backward parametric amplification case described in Section 9.6. If the damping constants « and I’, are sufficiently small, the amplification can in principle go into oscillation. For the acoustic waves involved in backward Brillouin scattering in liquids, how- ‘ver, w,/27 is typically of the order of $ GHz and the corresponding T’,/27 is around 0.1 GHz (see Table 11.1). the attenuation coefficient [,/v is ~ 104Stimulated Brillouin Scattering 189 ‘Table 11.1 Frequency Shift vg, Linewidth T,, and Maximum Steady-State Gain Factor of Stimulated Briflouin Scattering, (28) nar © Bf. for a Number of Liquids* rr in Frequency Linewidth Gain Factor shift Tt Calculated Measured ss (spont.) gp(man) £5(max) Substance (Mz) (MHz) (cm/MW) (om/MW) CS, 5850 52.3 0.197 0.13 Acetone 4600 224 0.017 0.020 n-Hexane 222 0.027 0.026 Toluene 5910 579 0.013 0.013 cc, 4390 520 0.0084 0.006 Methanol 4250 250 0.013 0.013 Benzene 6470 289 0.024 0.018 H,@ 5690 317 0.0066 0.0048 Cyclohexane 5550 74 0.007 0.0068 *After L. L. Fabellinskii, Molecular Scattering of Light (Plenum, New York, 1968). om7! with » ~ 10° cm/sec. This is often much larger than the gain coefficient of stimulated Brillouin scattering estimated below. Consequently, the acoustic wave excitation here can be considered as highly damped, and (11.4) can be solved by first eliminating 4 using the approximation dA/dz = iAkA. We then have @ ia _ awit + (e +34 =e ONE 6 a i2mwt ars) o_& a 61 Og | (a2 > 3 )e ~ aPlarer where ge (de/ap (et - 22) P__ kao = 4q0 Bk — iT; /o° Equation (11.5) looks exactly the same as (10.15) in the stimulated Raman scattering case except that the Britlouin susceptibility x§ now replaces the Raman susceptibility x in (10.15). Since Imy? > 0, (11.5) shows that & would grow in the backward (—z) direction if (27w,/k,c7)Imx} > a/2, while &, would decay in the forward direction. In the case of negligible pump depletion, the solution of (11.5) is 1E,(2)/? = 6,2) Petre) (Ls)190 Stimulated Light Seattering where the Brillouin gain Gy is given by 4auh Gy = — lm xP 4," W.7 ara xP (11.7) Calculated values of Gp for a number of liquids are listed in Table 11.1.7 It is seen that with 100 MW/cm? pump intensity in a 10-cm cell, the exponential gain G,/ can be of the order of 10 (much larger in CS, because of the narrower Brillouin width), and therefore stimulated Brillouin scattering should be easily observable. The solution of (11.5) including the pump depletion, but with w = 0, can also be obtained readily.’ One finds the following algebraic relations between the inputs |£,(0)|?, |E,(/)|° and the outputs JECOP, [E,O?: ECO)? — COP = Ex)? — (OP and (11.8) ECP 1 = 1B (0)/E,(0)? TESCO) exp{ [1 — 12 (0)/8, OVP] GalEs(O)] ~ E00 FO)E 0 1 - 3S 2 = : 3 Q 1 Py {7 ™ 0 2 50 TIME [nsec] Fig, 11.1 Oscilloscope traces of the incident laser pulse P,, the backward Brillouin pulse Pp, and the transmitted laser pulse Py in ethyl ether. [After M. Maier, Phys, Rev. 166, 113 (1968).]Stimulated Brillouin Scattering it Significant pump depletion in stimulated Brillouin scattering is in fact a common phenomenon. A typical example is seen in Fig. 11.1, where depletion of the pump pulse into a backscattered Brillouin pulse is clearly dernonstrated.* Energy conversion as high as 90% has been reported. The lifetime of acoustic waves is 13 = 1/21, which is of the order of 1 nsec for w,/24 ~ 5 GHz. Transient stimulated Brillouin scattering is expected if the pump pulse has a width comparable to 1 nsec or less. The transient solution again resembles that of the Raman case. However, we do not dwell on it here, but refer the readers to Kroll.+ Experiments actually found strong dependence of the gain on the lifetime of the acoustic wave for pulsewidth less than ~ 100 7). This is especially so near the threshold for stimulated scatter- ing. Careful measurements have shown quantitative agreement between theory and experiment Stimulated Brillouin scattering was first observed by Chiao et al.’ in quartz and sapphire using a Q-switched ruby laser. They analyzed the reflected light from the medium with a Fabry—Perot interferometer and found the Brillouin- shifted component. Because of the high conversion efficiency, the backscattered Brillouin pulse frequently is so intense that it can be detected by eye. Without proper isolation between the sample and the laser system, the backscattered Brillouin pulse will propagate into the laser medium and be further amplified. The result is that a Brillouin-shifted component will now appear in the laser output. Such a process can repeat a number of times, and the laser output will then have a spectrum containing several orders of Brillouin-shifted compo- nents,® This is what one would avoid in experiments requiring a single-mode laser beam. gh (1073 cm/ MW) —400 -—200 0 200 «400 Frequency shift Aca/ 2m (MHz) Fig. 11.2. Experimental Brillouin gain factor, g§ « Bj, versus Irequency shift Au/2a for a nonabsorbing liquid (66% CS, and 34% CCl,). The theoretical fit is a Lorentzian curve. {After Ref. 15.)192 Stimulated Light Scattering ‘As in the Raman case, the best way to test the theory of stimulated Brillouin scattering is to conduct gain measurements with an oscillator—amplifier arrangement similar to that in Fig. 10.9. An example is shown in Fig. 11.2.° A theoretical fit to the data using (10.7) allows the deduction of T,, which compares well with [, obtained from spontaneous scattering. Comparison of the Brillouin gains in Table 11.1 with the Raman gains in Table 10.1 shows that stimulated Brillouin scattering should usually dominate over stimulated Raman scattering in most liquids in the steady-state case. With a strong pump laser beam, especially one that is focused, stimulated Brillouin scattering can actually have enough gain to occur in all directions. Experimen- tally, acoustic waves of various frequencies are generated at the focal point in different directions. An audible sound can be heard when it happens. It is likely that a shock wave is initiated at the focal point. Cell windows often are shattered by the strong pressure waves generated, but the detailed mechanism involved is not yet understood. 41.2 STIMULATED THERMAL BRILLOUIN AND RAYLEIGH SCATTERING We have assumed that the acoustic wave is described by the density variation Ap. This is, however, only an approximation first used by Einstein and Brillouin" to describe spontaneous light scattering by low-frequency thermo- dynamic fluctuations in a single-component medium. Actually, p is a function of pressure p and entropy S, and one can write Ap = (dp/dp),Ap + (89/05), AS, Then Sp(t) describes the acoustic wave, while AS(1) describes the entropy wave at zero frequency with a diffusion-type equation of motion. In spontaneous scattering, Ap is responsible for the Brillouin doublet of the spectrum, and AS for the central Rayleigh component.’ Therefore, for the stimulated Brillouin scattering discussed in the previous section, a more correct formalism will have to replace Ap by Ap and e/dp by (8c/dp)5. In some cases, however, it is more convenient to use the independent thermodynamic variables g and T instead of p and S. This is particularly true when the temperature ¥ varies in direct response to the external heating of the medium. In the equations of motion, we expect that Ap(t) and AT(1) are coupled since both, being linear combinations of 4 p(t) and A5(¢), are mixtures of acoustic and entropy waves, Light scattering under the effect of heating due to optical absorption is known as thermal light scattering.’ The equations of motion for Ap and AT are, respectively, the Navier-Stokes equation in conjunction with the equation of continuity,!? 2 2 ay + 5 (de) + Bb y(AT) -ave¥ = (EE) -2(#),@epver), (11.9) Q a ae + mov y=0Stimulated Thermal Brillouin and Rayleigh Seattering 193 and the energy transport equation -i1 (osc$, - Ay? ] aT — GS-) 94, a a a (1120) nea 1{ de a ~ 22 ¢p, es) + 2( Fe) oP EF where v is the acoustic wave velocity, & = C,/C, is the ratio of heat capacities at constant pressure and at constant volume, > ts the isothermal compressibil- ity, 9 characterizes the acoustic damping, y = #(0¢/dp},, Ay is the thermal conductivity, and a is the linear absorption coefficient. The two equations in (11.9) can be combined to yield a 2 a vB rp “ant E tay dp +? v7(AT) (12.11) 1/4 = Lv e,-es) ~ sl 5F),7 [E, EZ 9(A7)]. We notice that if the approximations § ~ 1, a = 0, and (#¢/0T), = 0 are used, (11.10) gives AT = 0 and (11.11) reduces to the acoustic wave equation in (11.2). Stimulated thermal Brillouin and Rayleigh scattering is now described by the coupling of (11.10) and (11.11) with the wave equations for E, and E, in (11.1) having y 1 f de PN (o.) = amp, P24? + als) eer and (11.12) L [ae PN (a,) = Fp Ee" + lo ,far" The solution of the coupled wave equations is similar to those of stimulated Raman and Brillouin scattering. We consider only the steady-state solution for backward scattering here, and assume that both excitations Ap and AT are highly damped. We can then replace 0/dt by —iw, = ~i(w@, — w;) and Vv = d/az by i(k, + k) in (11.10) and (11.11) and solve for Ap and AT. Substituting the expressions of Ap and AT into (11.12) and then PN. into (11.1), and using the slowly varying amplitude approximation for E, and Ef, we again find the amplitude equations for &, and &* in the form of (11.5), or a apt a}lP = -B1eP iG? (e+ 16 _ (11.13) a (3p - #)l&0 = -BIA Pz?194 Stimulated Light Scattering With the approximation (de/dp}r? > (ae/0T),T, we have 2(Aa/T, 20,/T) _200T) +(Be + pg, )—o 1 = st a. Po Pi - 1+ (a/Tex) +(an/T,y 1 +(A8/Ts) where ky tka)? = (ky 1p) (0, — 0), 22 any AQ Ba = Anc*pyel, ay? 1 Bretpely __ aint 4ncippl ee. (1.44) en 4ac*ppeT ex. o 20 Be Cyto ° ye (8 = VWerPar anew, ° p= wet ka 5 Po ° Y and = Are, + ka). x Pp In the limit of negligible pump depletion, |é|? grows exponentially in the backward direction with a gain G-a= Bla - @- The first two terms in (11.14) are responsible for stimulated Brillouin scattering, and the last term for stimulated Rayleigh scattering. The Bg term is for normal stimulated Brillouin scattering. It teads to the same Brillouin gain derived in the previous section. The gain spectrum is centered at AQ = 0 or wo, — @ =A k,)o/8?, which is the frequency of the acoustic wave excitation involved in the backward Brillouin scattering. The 83 term corre: sponds to thermal Brillouin scattering, since it vanishes if the absorptionStimulated Rayleigh- Wing Scattering 195 (10-30 / MW) -400 +200 G 200 «400 Frequency shift Ses /2m (MHz) Fig. 113 Experimental thermal Brillouin gain factor, g5 % 83, versus frequency shift Aw/27 for an absorbing liquid (66% CS, and 34% CC1, with a small amount of 1,) with an absorption coefficient « = 0.83 cm7!. The theoretical curve is the dispersive counterpart of a Lorentzian. (After Ref. 15.) coefficient a is zero, Its gain spectrum is antisymmetric about A& = 0. The BR, term also vanishes when a = 0, and corresponds to thermal Rayleigh scattering with a gain maximum at w, = w, ~ a, = Igy. Finally, the 8%, term corte- sponds to ordinary Rayleigh scattering with the same gain spectrum as B,. Experimentally, stimulated Rayleigh scattering is most difficult to observe because of the small B{,, but has actually been observed."? The maximum Rayleigh gain in liquids is estimated to be two orders of magnitude lower than the Brillouin gain.? With absorption, Rayleigh scattering can be greatly en- hanced through the §g, term. Indeed, stimulated thermal Rayleigh scattering can be readily observed in absorbing gases and liquids.’* Stimulated thermal Brillouin scattering in absorbing media is also easily observable’? Its occur: tence is evidenced by a small upshift of the Brillouin-shifted frequency, since the combined gain spectrum exhibited by the 8% and 85 terms in (11.14) has a maximum at AQ > 0. The best way to study the effect of thermal Brillouin scattering is to measure the Brillouin gain as described in Section 11.1. The measured gain spectrum can then be compared directly with the theoretical spectrum in (11.14). An example is presented in Fig, 11.3, which shows a good agreement between theory and experiment. 113 STIMULATED RAYLEIGH-WING SCATTERING Fluctuations of molecular orientation and distribution in a fluid medium result ; in fluctuations of the dielectric constant and lead to spontaneous light scatter-196 Stimulated Light Scattering ing. This is known as Rayleigh-wing scattering, which has a spectrum similar to Rayleigh scattering but much broader, in width. Stimulated Rayleigh-wing scattering can also be expected at high pump intensity. The physical picture is as follows: mixing of E, and E, reorient and redistributes the molecules; the molecular reorientation and redistribution, which vary in space and time, in turn beat with E, to enforce E,. Stimulated Rayleigh-wing scattering is then simply the result of coupling between E,, E,, and the induced variation in molecular reorientation and redistribution. For quantitative description, we must find the equation of motion for molecular reorientation and redistribu- tion. We consider here only the reorientation mechanism. We assume anisotropic molecules with cylindrical symmetry. The optical polarizabilities parallel and perpendicular to the molecular axis are denoted by ag and «, , respectively. Let the molecular axis tilt at an angle 6 from the X-axis (Fig. 11.4). Then a linearly polarized E field along £ induces a dipole p on the molecule with p, = a,,% and a,, = 0,cos?@ + a, sin’? 4 * (11.15) = & + Aa(cos*# — $) where a=4}(a,+2a,) and Aa =a - a. The applied field E now interacts with the induced dipole and tends to align the molecule along % against the thermal randomization. Let us assume a collection of noninteracting molecules with a density V and a random orienta- tional distribution in the absence of E. With the applied E, the orientational Fig. 1,4 Sketch showing a uniaxial molecule lying at an angle @ from the electric field E along 2.Stimulated Rayleigh-Wing Scattering wT distribution function at equilibrium becomes (11.16) f(0)= ex0(= 24,s1BV'/tsT) where Z= fos of apt sine ky is the Boltzmann constant, and the factor 2 instead of 1/2 appears in the exponential because of our convention on the amplitude of E (Section 2.9). Then the tensor component x,, of the optical susceptibility, following (11.15) and (11.16), is given by <2 xe Xt 3 Naa with = Na, = baat - 4) = [Hoos — 4 ecp[— 2AalE]2{cos’# — 4) /k,T]sin 6 48 0 (11.17) fexp[—2de| Z/*(cos"# - }) /ky7|sin ad 0 where ¢ ) denotes the orientational average. Similarly, we find — 1 yy 7 Xe =X — | NMA. (1.18) Physically, the quantity @, which describes the degree of molecular alignment, often is known as the orientational order parameter. For random distribution, Q = 0, and for perfect alignment Q = 1. The polarization induced by E takes the form P= ix, 8 11.19 = 3 xB + NSaQE. (11.18) Since @ is also a function of E£, the second term in (11.19) represents a nonlinear polarization PN™ pre = #2 Naage (11.20)198 Stimulated Light Scattering For refatively weak fields, we have Q & |Z[?, and then PN is a third-order nonlinear polarization. In stimulated Rayleigh-wing scattering, the molecular orientational distribu- tion changes in response to the beating of two optical fields E, and E,. The equation of motion for molecular reorientation by this combined action of E, and E, is provided by the Debye rotational diffusion equation for the distribu- tion function f(@):! af _ 1 A, af af , AdalE*sin 9 cos 2 Salome + wf? Equation (11.21) can be reduced to a simple equation of motion for Q by multiplying both sides by 4(cos’# — }) followed by an integration over 8, and neglecting the field-dependent terms of higher orders than [EP /k aT: (11.21) 90 _ _ 2, Ang ep Wat hel (11.22) where 1p = 7/5 kgT is the Debye relaxation time and v is a viscosity coefficient for an individual molecule. Now that E = E, + E, = %é,exp(ik, +1 — iat) + ZG,exp(ik,°r ~ ia,t) and we are interested in the orientational redistribu- tion excited by the beating of E, and E,, the |E/? term in (11.22) should be replaced by E,+ Ef. Equation (11.22) is then coupled to the wave equations (11.1) for E, and E, via (11.20). In the steady-state case, one finds _ toh, EF 3v(1 — iaty) PM (an) = xBWIEP Er (11.23) 3 8Nr,(day° KRW Gon oan) 9v{k + twry) with w = a, — @;. The Rayleigh-wing susceptibility x {ly has a negative imag- inary part. In analogy to the other stimulated light scattering cases, this. indicates that £, can experience an exponential gain exp(Gpw — #)2 with 2 Grw = “E22 | xit) ie? bn N(Aa) 7p 1641 Te en ASK gT(1 + waz} (11.24) which has its maximum at w = 1/tp.Other Stimulated Light Scattering 19 For a liquid with sp ~ 1071 sec and 16q7N(Aa)?/45k ,T ~ 107! crn’ /erg in a typical case, we have (Ggw )max ~ 10-3 cm/MW, comparable to Raman gains in many liquids? Stimulated Rayleigh-wing scattering is therefore ex- pected to be easily observable, as is indeed the case!” In fact, it is very much analogous to stimulated Raman scattcring by vibration: the material excitation 1Q| is independent of the wavevector so that the stimulated gain is isotropic. Consequently, Stokes-anti-Stokes coupling {see Section 10.4) that leads to the generation of anti-Stokes radiation in the near forward direction in stimulated Raman scattering also occurs in stimulated Rayleigh-wing scattering,” The results are somewhat different because of the difference in the resonant frequencies of the material excitations. Unlike the Raman case, the maximum gain with Stokes—anti-Stokes coupling in stimulated Rayleigh-wing scattering appears at w,— w, = w,,— = 0 with k, and k,, making an angle 6,,, with k,. In other words, the laser beam, generates through stimulated Rayleigh-wing scattering a cone of radiation of the same frequency at an angle @,, from the laser beam. In reality, a laser beam of finite cross section has a spread of &,. The effect of stimulated Rayleigh-wing scattering, is to broaden this spread of k; of, equivalently, to reduce the laser beam cross section. The laser beam therefore appears to self-focus as it propagates in the medium. This is an unconventional way of describing self-focusing of light. The conventional way will be discussed in Chapter 19. We note that it is amplification of the existing off-axis k, components that leads to self-focusing. Hf stimulated Raman scatter- ing also occurs in the medium, it is initiated by amplification of noise. Since the Raman gain and Rayleigh-wing gain are comparable in many liquids, we can expect that the occurrence of self-focusing often precedes that of stimu- lated Raman scattering. 11.4 OTHER STIMULATED LIGHT SCATTERING In a multicomponent system, local concentrations of the components can fluctuate, causing variation in the dielectric constant and scattering of light, known as concentration scattering.’ In thermodynamics, concentrations to- gether with p, T or p, 5 form a set of thermodynamic variables. Variation of the dielectric constant of a two-component system can be written for example, as, de be de ac= (3) 20+ [ae]. 07 +(Fe), 186 (11.25) where C is the relative concentration. In stimulated concentration scattering, AC is excited by the beating of E, and E,. It obeys a driven diffusion equation” a a 2 =~ (F DY Jac D(de/4C),,.777(E, ES) 11.26 Sang ( 34/80). 7 (11.26)200 Stimulated Light Scattering In (11.26) we neglected the coupling of AC to Ap and AT; D is the diffusion coefficient and p is the chemical potential. In a more rigorous treatment, the coupling between AC, Ap, and AT can be taken into account.” Analogous to the other stimulated light scattering cases, stimulated concentration scattering is then described by the solution of the coupled equations (11.26) and (11.1) with PN4(a,) = (4¢/8C), 7E,AC/4m and PN4(w,) = (d¢/8C),, pE,AC*/ 4q. The stimulated gain has a spectrum proportional to w7(1 + w*72) with ~' = Dk? and k =k, — k,. Details of theory and experiment of stimulated concentration scattering can be found in Ref. 20. There are, of course, many other types of light scattering: light scattering by molecular fibration (rotational oscillation), by sheer waves, by spin waves, by surface waves, etc.| In principle, with sufficient pump intensity, they can all become stimulated: knowing the dynamic equation for the material excitation, the theoretical treatment again follows the coupled-wave approach. However, the threshold for a certain stimulated scattering may be higher than the optical damage threshold; if so, such stimulated scattering will not be observable. A very different type of stimulated light scattering is stimulated Compton scattering, first proposed by Pantell et al.’ By backscattering microwaves from a relativistic electron beam, tunable far-infrared radiation could be generated. Tunability could be achieved by varying the electron energy. Sukhamte and Wolff”? showed that stimulated Compton scattering could be greatly enhanced in a magnetic field if the microwave frequency is equal to the cyclotron resonance frequency. Experiments on stimulated Compton scattering have not yet been reported. However, in a related problem, tunable microwave and far-infrared radiation have been generated by relativistic electrons performing cyclotron motion in a magnetic field.* Also, intense microwave emission with peak power > 500 MW and conversion efficiency ~ 17% has been observed in coherent Cherenkov radiation from a relativistic electron beam interacting with a slow wave structure. REFERENCES 1 See, for example, I. L. Fabellinskii, Molecular Scatsering of Light (Plenum, New York, 1968). W. Kaiser and M. Maier, in F. T. Arrecchi and E. 0. Schulz-DuBois, eds., Laser Handbook (North Holland Publishing Co., Amsterdam, 1972), p. 1077. C.L. Tang, J. Appl. Phys. 37, 2945 (1966), M. Maier, Phys. Rev. 166, 113 (1968). NM. Kroll, J. Appl. Phys, 36, 34 (1965). M. Maier and G. Renner, Phys. Leit. 34A, 299 (1971); Opt. Comm. 3, 301 (1972) RY. Chiao, C, H. Townes, and B. P. Stoichelf, Phys. Rev. Lett. 12, 592 (1964). 8 E. Garmire and C. H. Townes, Appl. Phys. Lett. 5, 84 (1964). 9D. Pohl and W. Kaiser, Phys. Rev. BL 31 (1970). 10 A. Einstein, Ann. Phys. 33, 1275 (1910). ULL. Brillouin, Ava. Pays. (Paris) 17, 88 (1922). w 4AM ae12 3 14 18 16 0 18 9 Sere Bibliography 201 R. M. Herman and M. A. Gray, Phys. Rev. Lest. 19, 824 (1967) G.L Zaitsev, Yu. L Kyzylasov, V, S, Starunov, aud I. L. Fabellinskii, JETP Lett. 6, 255 (1967); 1 L. Fabellinski, D. I. Mash, V. V. Morozov, and V. §. Starunoy, Pays, Lett, 27A, 253 (1968). D. H. Rank, C. W. Che, N. D. Foltz, and T. A. Wiggins, Pays. Rev. Lett. 19, 828 (1969). D. Pohl and W. Kaiser, Phys. Rev. BI, 31, (2970). P. Debye, Polar Molecules (Dover, New York, 1929). D. I. Mash, V. ¥. Morozov, ¥. §. Stansnov, and I. £. Fabellinskii, JETP Lert. 2, 25 (1965). M. Denariez and G. Bret, Compr. Rend. 265, 144 (1967), Phys. Rev. 171, 160 (1968). R. Y. Chiao, P. L. Kelley, and E. Garmire, Pays. Rev. Lett. 17, 1158 (1966); R. L. Carman, R. Y, Chiao, and P. L. Kelley, Phys. Rev. Lett. 17, 1281 (1966). W. H. Lowdermilk and N. Bloembergen, Phys. Rev. A5, 1423 (1972) R. H. Pantell, G. Soncini, and H. E. Putholt, [EEE J. Quani. Electron. QE-4, #)5 (1968). V.P. Sukhamte and P. A. Wolff, IEEE J. Quant. Eleceron. QE-10, 870 (1974) V.L. Granatstein, M. Herndon, R. K. Parker, and 8. P. Schlesinger, FEEE J. Quant. Electron QE-10, 651 (1974) 'Y. Carmel, J. Ivers, R. E. Kribel, and J. Nation, Phys. Rev. Lert. 33, 1278 (1974). BIBLIOGRAPHY Fabellinskii, I. L., Molecular Scattering of Light (Plenum, New York, 1968) Kaiser, W., and M. Maies, in F. T. Artecchi and EQ. Schulz-Dubois, eds., Laser Handbook (North Holland Publishing Co., Amsterdam, 1972), p. 107712 Two-Photon Absorption One-photon and two-photon transitions follow different selection rutes. They are therefore complementary to each other as spectroscopic tools. A well-known example is infrared absorption versus Raman scattering, In a two-photon absorption process, two photons are simultaneously absorbed to excite 4 material system. Being a higher-order process, its absorption cross section often is many orders of magnitude smaller than that of a one-photon absorp- tion. Even so, two-photon absorption is readily observable with lasers and has become a valuable spectroscopic technique complimentary to linear absorption spectroscopy. This chapter briefly describes the basic theory, measuring tech- niques, and various applications of two-photon absorption. 12.14. THEORY The transition probability of a two-photon process was first derived by Géppert-Mayer using second-order perturbation theory.! The derivation was given in Section 10.2 for the case of Raman scattering. For two-photon absorption, the transition probability per unit time per unit volume per unit energy interval closely resembles that of (10.2) and is given by AW, Wy, BPNiyay a(he) d( hey) 82 m=Z s K SIME) PKaylan ala, )[7e( A Aw), (12.1) er: é|s)¢sler* ey + er7é,|s)(sler7éy h(a, — &,,) BC, — wy (see Fig. 12.1). Notation here follows Section 10.2, with de = 0) + 0, — wy; In the semiclassical approximation, Keay |azay]e,)|? = nn, can be replaced by (£402) |EyI21E|2/(2m)?(ReayX Ay) = Fyda(e,07)'/7/e7(few, (ftw), where J, and /, are the beam intensities at w, and «, respectively. The two beams 2lf> Fig. 12.8. Two-photon excitation of a system from lip |i) to|f) via the virtual intermediate state |s). propagating along 2 in such a nonlinear absorbing medium have their attenua- tions governed by the equation dl, dl, = - whl. = = -wyhl, (12.2) __ [aw,/d(ho;)] (0, — 67) re AL orn 5 wear eth dw )€p; — py). As in the Raman case, the above equation for y can also be derived from the coupled wave approach. It is easily shown, following a derivation similar to that of Section 10.3, that the two-photon absorption coefficient y is linearly proportional to the imaginary part of the third-order nonlinear susceptibility x for the two-photon absorption process: 80? Y= lm x cree? (12.3) Imx® = NalMy|(h dw) 9; — py). The same result can of course be obtained by treating two-photon absorption as a wave-mixing process in which the two optical waves at w, and w, jointly excite the material excitational wave p{?(c, + «;). The derivation follows that in Section 10.3.204 Two-Photon Absorption The coupled equations in (12.2) can be solved analytically, noting that which is the consequence of having equal numbers of photons absorbed at , and w, in a two-photon absorption process. If fo and yy are the beam intensities at the entrance of the medium, we have honk _ beak (12.4) ey ey The solution of (12.2) can then be obtained by first eliminating either J, or 2). Assuming fig > 9, we find Le=I (Lo/e1) - Un/@2) 1 Eg/ er) — Can/ 2 Jexp(~ Kz)’ [Uio/e1) ~ Uan/e)lexp(= Kz) b= : 12.5 2 tao 7 jean) = Ulap/onenp(— Ke) (025) K = way 4 — 22). If Jy) > Iq, then the attenuation of Ff, is negligible, and the solution reduces to 4, = ho: (12.6) I, = Eyexp(~ Kz). A special case of interest is when w, = «,. The conventions of Section 2.9 should be used in dealing with the two-photon absorption coefficients in this case, Equation (12.2) becomes 2 -ey (2.7) and the solution takes the form = fio hay Typo yz" (12.8) In the case of weak absorption, it reduces to Fy =ip(k - How yz). (12.9)Experimental Techniques 25 The two-photon absorption coefficient y and the corresponding third-order susceptibility x are in general tensor quantities. Analogous to Raman scatter- ing, the selection rules can be derived from group theory. They have been obtained by Inoue and Toyozawa’ for the 32 crystal point groups and by McClain? for molecular fluids. Spin-orbit coupling has been included by Bader and Gold‘ in an extension of Inoue and Toyozawa's calculation. 122 EXPERIMENTAL TECHNIQUES Two-photon absorption can be measured directly from beam attenuation if the absorption is sufficiently strong. Let us assume, as an example, a typical two-photon absorption coefficient with Im x = 10~"? esu for a condensed medium. Then, from (12.3) and (12.5), the induced attenuation coefficient K is of the order of 10-* for I, ~ few MW/cm*. This corresponds to a ~ 1% attenuation of the a, beam in traversing through a medium 1 cm long, and it should be easily measurable. Direct attenuation measurement of two-photon absorption is therefore fairly straightforward with pulsed lasers unless x) is orders of magnitude smaller than 10-7 esu. For two-photon spectroscopic work, one of the two input beams must be tunable. In the carly days, only fixed-frequency lasers were available. The tunable beam was provided by an incandescent or arc lamp in conjunction with a monochromator. A two-photon absorption spectrum was obtained by measuring the laser-induced attenuation as a function of the frequency of the tunable beam in the medium. A typical experimental arrangement is shown in Fig. 12.2.5 Over the years, several research groups have constructed more sophisticated, automated versions of the setup.’ The incoherent lamp can now be replaced by a tunable laser with great improvement on the signal-to-noise ratio. Unfortunately, the tunability of a laser is siill limited. Replacement of the Jamp by a tunable laser is preferable only if a narrow spectral range is of interest. Weak beam attenuation is generally difficult to measure. One would like to find other methods of higher sensitivity for two-photon absorption measure- ment. In many media, luminescence may appear following excitation. This is nearly always the case in gases, and is also fairly common in condensed matter although the quantum yield could be small. Since luminescence is easily detectable, it provides a means to monitor two-photon absorption with a sensitivity many orders of magnitude higher than the beam attenuation mea- surement. The first two-photon absorption experiment was actually done with this technique.’ However, in two-photon absorption spectroscopy, one must be sure that the quantum yield of luminescence does not depend strongly on the excitation frequency; otherwise, the spectrum will appear distorted. Two-photon excitation near or above the ionization level of an atom or molecule may lead to ionization, and the resulting electrons and ions are easily detectable? Therefore, photoionization can also be a sensitive method for206 ‘Two-Photon Absorption s emple chamber Nitwogen temperature Grating Xenon = Shutter i monochromator orc lomp Nickel sulfate —~] filter Ultraviolet photomultiplier Laser manitor Chopper | AG and BC Signal to Hgnals to oscilloscope oscillescope Ruby laser Fig. 12.2 Schematic diagram of an experimental set-up for two-photon absorption spectroscopy. (After Ref. 5.) detecting two-photon absorption. Its application, however, is limited to the case where the final excited state is near or above the ionization level. The observed spectrum is the two-photon absorption spectrum weighted by the ionization rate, which generally depends on the energy of the final excited state. By the same token, photoconductivity can also be used to detect two-photon absorption in a solid. If heat released through relaxation after the two- photon excitation can be monitored, it can also be used to measure two-photon absorption. An example is photoacoustic spectroscopy, in which heat released appears as an acoustic signal detectable by either a microphone or a trans- ducer. Less conventional methods of detecting two-photon absorption include photoemission, photodissociation, photochemical reaction, and optogalvanic effect. 123 TWO-PHOTON ABSORPTION SPECTROSCOPY Solids The first spectroscopic measurements of two-photon absorption were carried out by Hopfield et al.* on alkali halides near the band edges, using the setup shown in Fig, 12.2. Since the crystals have inversion symmetry, the states near the band edges have more or less definite parities. Thus one-photon and‘Two-Phaton Absorption Spectroscopy 27 [ot ot ot 7 cm] 4 ++ 2-Photon Absorption \—— 1-Phofon Absorplion 80° ao 750 700 eV -+=2-Photon Absorption oh Phelon Absorption goth on 1-Photon Absorption (Optical Density in Arbilr Units) 2-Photen Absorption Coefficient Photon Energy Fig. 123 Qne- and two-photon absorption spectra of KBr and RbBr. [After D. Frohlich and B, Staginnus, Phys. Rev. Lert. 19, 496 (1967). two-photon absorption spectra are expected to be different, as seen in Fig, 12.3. In particular, no exciton peak is present in the two-photon absorption spectrum. The result was used by Hopfield et al. to test the validity of the various exciton models for the alkali halides, Two-photon absorption in semiconductors is the subject of numerous studies.” It was hoped that the technique would lead to new information about the band structures. Thus far, the results have been disappointing mainly because, first, the band structures of those semiconductors are already very well known; second, the two-photon absorption data are not very accurate due to laser fluctuations; and third, the spectral ranges covered by two-photon absorption are very limited. Two-photon absorption is, however, a useful tool to study excitions and exciton-polaritons in a semiconductor. With one-photon absorption, only the existence of the exciton-polaritons can be shown by the observation of the reststrahlung band. With two-photon absorption, the disper- sion curve of the exciton-polaritons can be measured.’° An example is given in Fig. 12.4. In this case the excitons can be excited by both one- and two-photon208 ‘Two-Photon Absorption g ze} Of 7 & Se seo" & ; qQ ‘K, Q Nw || z= é s Lv 20° a 1 LE Qo sof = = 328 3212 32 3210 3209 3200eV — FWO-PHOTON ENERGY fa) Cuct ate 2 ENERGY WAVE VECTOR K mn ca’ co Fig. 12.4 (a) Two-photon absorption peaks near the first excitonic excitation of CuCl for different angles @ between the two incoming bearns at 1.5 K. (6) Dispession curves of the transverse exciton polariton (TP) and the longitudinal exciton (LE) of CuCl. The lines are theoretical curves calculated from reflectivity data. The open squares and closed circles are measured results from two-photon absorption. (and the crosses are from second harmonic generation.) [Alter D. Frohlich, Festkarperprobleme XXT, 363 (1981).] transitions. The observed two-photon absorption is due partly to excitation of excitons and partly to sum-frequency (or second harmonic) generation.” The correct theoretical treatment of the probiem follows closely the derivation in Section 10.7.7 Two-photon absorption has also been used to probe the states of excitonic molecules!? that cannot be reached by one-photon excitation, In other applica- tions, two-photon absorption can be used to yield a uniform excitation of carriers in the bulk. This could be useful in both physics and device studies,References 9 Molecular Fluids and Gases Two-photon absorption can be used to probe excited states that cannot be reached by one-photon excitation. In molecules with centers of symmetry, the electronic states can be divided into gerade (g) and ungerade () states. One-photon transitions from g to g or from w to wu are forbidden, but two-photon transitions are allowed. Thus with two-photon absorption, it is now possible to study a new set of electronic, vibrational, and rotational states which cannot be reached by one-photon absorption. Numerous examples are. cited in Ref. 14. McClain’ has pointed out that even though the molecules are randomly oriented in a gas or liquid, two-photon absorption with @ # w, still shows polarization properties that allow us to determine the symmetry of the excited states of the molecules. Therefore, two-photon absorption has become an important tool in the field of molecular spectroscopy, as evidenced by the large number of references cited in Ref. 14. Atoms Two-photon absorption also can be used to study excited electronic states of an atom that cannot be probed by one-photon absorption. Examples are the rs and nd states of an alkali atom. Because of the large transition matrix elements between the atomic states, two-photon absorption in atomic gases is generally much stronger than in molecular gases. Yet it is still too weak to be observed by the measurement of beam attenuation, Fortunately, other methods, such as photoluminescence and photoionization, can be used, They are sensitive enough to detect two-photon absorption in a vapor of less than 1 torr Pressure. With counterpropagating beams of the same frequency, two-photon absorption in gases can yield Doppler-free spectral lines. This is described in Chapter 13. Applications of two-photon absorption to the atomic studies of high Rydberg. States, quantum defect theory, and autoionizations are discussed in Chapter 18. REFERENCES M. Goppert-Mayer, Anz, Physik 9, 273 (1931). M, Inoue and Y. Toyorawa, J. Phys. Soc. Japan 20, 363 (1965). W. M. McClain, J. Chern. Phys. 55, 2789 (1971). T. R. Bader and A. Gold, Phys, Rev. 171, 997 (1968), J. J, Hopfield, J. M. Wortock, and K. J. Park, Phys. Rev. Lett. U1, 414 (1963); J. J. Hopfield and J. M. Worlock, Phys. Rev. 137, A1455 (1965). 6 See, for example, B. Staginnus, D. Frohlich, and T. Caps, Rev. Sci. Inst. 39, 1129 (1968); M, W. Dowley and W. L. Peticolas, (BM Res. Dev, 12, 188 (L968); R. L. Swofford aad W. M. McClain, Rev. Sci. Inst, 44, 978 (1973). 7 W. Kaiser and C.G. B. Garrett, Phys. Rev, Lett, 7, 229 (1961) Amun210 10 W 13 14 ‘Two-Photon Absorption See Chapter 18 on multiphoton ionization. See, for example, ©. C, Lee and H. Y. Pan, Phys. Rev. BY, 3502 (1974) D. Froblich, £. Mahler, and P. Wiesner, Phys. Rev Lett. 26, 554 (1971). ‘D.C. Haucisen and H. Mahr, Phys. Rev. Lett. 16, 838 (1971). D. Boggett and R. Loudon, Phys. Rev. Lett. 28, 1051 (1972). G. M, Gale und A. Mysyrowicz, Phys. Rev. Lett, 32, 727 (1974), L. L. Chase, N. Peyghambarian, G. Grynberg, and A. Mysyrowicz, Phys. Rev. £ere. 42, 1231 (1979) M. W, McClain, Ann. Rev. Phys. Chem. 31, 559 (1980). BIBLIOGRAPHY Goid, A,, in R. Glauber, ed., Quantum Optics (Academic Press, New York, 1969). p. 397. McClain, W. M., Acc. Chem. Res. 7, 129 (1974)’ McClain, W. M., Ana, Rev. Phys. Chem, 31, 559 (1980). Mabr, H., in H. Rabin and C. L, Tang, eds., Treatise in Quantum Efectronics (Academic Press, New York, 1975), p. 472. Worlock, J. M., in F. T. Arrecchi and E. ©. Schulz-Dubois, eds., Laser Handbook (North-Holland Publishing Co., Amsterdam, 1972), p. 1323,13 High-Resolution Nonlinear Optical Spectroscopy ‘Lasers are known to have extremely narrow intrinsic linewidths. They are thus ideal tools for high-resolution spectroscopic studies. For a He-Ne laser at 3.39 ym, a linewidth as narrow as 3 Hz has been reported} This represents a resolving power of 2 x 105, which is only two orders of magnitude jess than the Méssbauer effect. In ordinary spectroscopy, however, the studies of spec- troscopic details often are limited by inhomogeneous broadening rather than by instrument resolution. The Doppler width of the sodium D lines at room temperature is ~ 1.3 GHz, while the hyperfine splittings of the lines are only several hundred megahertz. In solids, the inhomogeneous width of a tine can be even much higher. Thus for high-resolution spectroscopy it is of prime importance to find ways to reduce the effect of inhomogeneous broadening. This chapter introduces a number of nonlinear optical spectroscopic tech- niques which serve the purpose. These methods have, in recent years, revo- Jutionized the field of atomic and molecular spectroscopy and stimulated a great deal of interest in the area of high-resolution solid-state spectroscopy. 13.1. GENERAL DESCRIPTION Inhomogeneous broadening of a spectral transition arises because atoms, molecules, or ions in an ensemble do not all have the same local environment. Consider a transition between two states |x) and jn’) with a resonance frequency w,,,. In general, o,., is a function of a number of parameters, a, B, y,... describing the local environment. These local parameters are random variables and should obey a certain statistical distribution function, say, gia, B, 77>: ) with fgdadBdy--- = 1. A physical quantity x, which is a an212 High Resolution Nonlinear Optical Spectroscopy function of w,,,, should then have its average value given by X= fX [py (es Boys g(a 8.7 .Jdadpe--. (13.1) For example, a Lorentzian absorption line with inhomogeneous broadening has the expression SoT8(a, B, } S(o} = {[ —<—— 7? 13.2 “) SFoeraas 8 r 032) ‘The number of local parameters necessary to characterize the local environ- ment can be many. For impurity ions in a solid, each ion has its own local environment, and sees a local crystal field resulting from the Coulomb force of neighboring atoms or ions.’ The crystal field can be described by a set of local parameters, and the distribution of ions over the local sites therefore, can be characterized by the distribution function of these local parameters. The total number of such local parameters depends on the local symmetry of the ion site. Tt can be very large for a low-symmetry site, for example, larger than 10 for a C, symmetry. The inhomogeneous broadening of the impurity ion spectrum is in principle determined by the statistical variation of the many local parame- ters characterizing the ion site. In practice, however, one or a few parameters describing the high-symmetry components of the crystal field may dominate the rest. In gases, the situation is most fortunate, since the velocity of atoms or molecules is the only local parameter contributing to the inhomogeneous broadening, which is just the Doppler broadening. The thermal velocity obeys the Maxwellian distribution si) =e (13.3) where u2 = 2kT/m, and m is the mass of a single atom or molecule, The Doppler width is then given by the well-known expression (opin = 2a [(2KT/me? a2]? (13.4) = 7,163 X LO“ (T/A) erg where T is in degrees Kelvin and A is the atomic or molecular weight. For A = 100, we find Aw, ~ 0.02 cm7! (0.6 GHz) in the green, and Awp ~ 10-7 cm! (30 MHz) in the infrared around 10 »m. These inhomogeneous broaden- ings may seem to be narrow by the standard of ordinary spectroscopy, but they often are much broader than homogeneous linewidths. The natural lifetime broadening is 10° to 107 Hz for atomic transitions, and 10 to 10° Hz forQuantum Beats a5 molecular vibrational transitions. The pressure broadening due to atomic or molecular collisions is around 10* Hz at 1 mtorr, and broadening due to collisions with walls of a container of a few centimeters in dimension is 10? 10 10* Hz. The power broadening (or saturation broadening) of an atomic transition can be ~ 10 MHz/mW/cn’. If a spectroscopic technique has sufficient resolution, then elimination of the Doppler broadening allows us to measure the homogeneous linewidth and lineshape and to probe the various physical mechanisms for the homogeneous broadening. Furthermore, level shifts and splittings smaller than the Doppler width but larger than or comparable to the homogeneous width can also be studied. These include many interesting problems such as Zeeman and Stark effects, collisional effects, hyperfine splittings, isotope shifts, Lamb shifts, quantum defects of Rydberg States, measurements of rotational splittings, and the like. To eliminate Doppler broadening, the classical approach is to use a mono- energetic atomic or molecular beam. For a small beam divergence of 26, the residual Doppler width seen by a perpendicular probe beam is (Bwp}y, = Qu/c)a,,¢ where u is the forward beam velocity. With >= 10°? rad, (8@p)9, can be more than one order of magnitude smaller than (Aw)... High-resolution nonlinear optical spectroscopy, however, uses nonlinear optical methods to reduce the effect of inhomogeneous broadening. There exist a number of such techniques. They generally follow the basic idea of either using a resonant effect that is independent of inhomogeneous broadening or selectively studying only a group of molecules with the same resonant frequency. Most of the techniques discussed in the following sections are applicable, in principle, to both gases and condensed matter, although the spectral lines in condensed matter frequently are 100 broad to require high-resolution spectro- scopic measurements. 13.2, QUANTUM BEATS? Consider a system with two closely spaced excited states as seen in Fig. 13.1. If a laser pulse with an inverse pulsewidth larger than the spacing between the two excited states is used to resonantly excite the system, then after the pulse <2l <1) wad OO l? = lay(dylervaye mt! + ax Polervyye eM? (13.6) = Aye EY Age TH + Be *TIGOS[( yg — Wyo) + |, ¥5q= 82.9 MHz ¥43> 66.5 ¥32* 49.9 No hoa BEATS AT 116.4, 66.5, 49.9 MHz 149.4, 82.9 66.5 MHz Ag =9193 MHz 3 (a) Fig. 13.2. (a) Hyperfine structure of the 678, yp and PP, 72 levels of cesium. The sets of quantum beat frequencies expected for the a and } excitations are indicated. (6) Observed oscilloscope traces of the quantum beats in fluorescence resulting from the a and b excitations, respectively. The corresponding theoretical plots of the beats are shown under the experimental traces. (c) Time recording and frequency analysis of the I, — 1, signal in (b). (After Ref. 4.)Quantum Beats 215 EXCITATION FROM THE F=3 GROUND STATE () EXCITATION FROM THE F«3 GROUND STATE Seth) ' Vag Vay Van" Van t Haz 4 val a . — 0 100 tise) 0 SO ¥(MHz) fe} Fig. 3.2 (Continued ). which shows a damped oscillation with a beat frequency (w5 — 49). For an ensemble of such systems, each system may have slightly different @jq and tz) due to the Doppler effect or other inhomogencous effects, but the oscillation frequency (3) — #19) should be the same for alk systems. Conse- quently, the spontaneous radiation power from the ensemble after the pulse excitation is still given by (13.6). The oscillation is observable as long as Jeng — @y9! 2 (1) + Ty). The oscillation frequency yields directly the level SPACiNg Wy, = Wyqg — Wy216 High Resolution Nonlinear Optical Spectroscopy This quantum beat technique can be extended to systems with several closely spaced transitions. The time-varying signal becomes more complicated, but if the level spacings are much larger than the damping coefficients T, the spectrum can be obtained readily from a Fourier wansform of the time-varying signal, The technique is most useful for finding small level splittings, so that the signal can be accurately measured using the conventional transient detec- tion system. Both the time-varying signal and its Fourier spectrum can be directly displayed on the oscilloscope. Figure 13.2 is an example of the results from a quantum beat experiment.* 133 SATURATION SPECTROSCOPY The basic idea of saturation spectroscopy is as follows. A monochromatic light beam resonantly excites only a small group of atoms or molecules under the inhomogeneously broadened profile and induces in them a population change. This group of atoms or molecules, marked by the population change, can then be selectively studied by either absorption or luminescence. The effect of inhomogencous broadening is thus suppressed. Saturation in Excitation Consider first the induced population change due to resonant excitation in a two-level system shown in Fig, 13.3. The rate equation for the population change is @o1 (3 + 7 (ae ~ Ap") = -2W,, Ap (43.7) where Ap =?) — 0, is the population difference between the two levels, Ap® is the corresponding thermal equilibrium value, H4, = 2nQ7g(a) is the transi- tion rate, @ = (1/h)|(ljer*E|2)] is the Rabi frequency for the transition, and g(w) is the lineshape function of the transition. We assume that g(#) = <2 ww wy Fig. 133 A two-level system with a resonant excita- <1 tion.Saturation Spectroscopy aT T/a[(e - ey)? + T?4. The steady-state solution of (13.7) can be written as —(T/E) Ae” —___ 3.8 (@— wy) + 2 + 4/6) (13.8) Ap — Ap? = with I/L, = 42°T,/T. Here I= clEP/2an is the intensity and J, = cT}E[?/87@27, is defined as the saturation intensity. Physically, (Ap — Ap®) of the population is resonantly pumped from the lower state into the excited state. When I/I, approaches infinity, Ap approaches Apy. This is known as the pump saturation effect. The absorption of the pump beam is proportional to W, Ae and can be described by an absorption coefficient al? 22 —— (13.9) (@- wy) +170 + 1/5) where ag is the peak value at w = w,, and £/7, > 0. Equation (13.9) shows that for nonnegligible 1//,, the absorption line is broadened by a factor of (1 + J/LY. This is the well-known power broadening effect. In the weak saturation limit of 1/1, « 1, (13.8) and (13.9) reduce, respec- tively, to Ap — Ap? = EEL (13.10) and qe - “a (13.41) (@— ey) + I? (@ — ey, ¥ + T? The change in the absorption coefficient a then is proportional to f or |£ and can be regarded as a third-order nonlinear optical effect. With inhomogeneous broadening, a monochromatic laser beam can reso- nantly excite only a small fraction of the atoms or molecules. Consider a gas medium in which the Doppler effect dominates the innomogencous broaden- ing, For the group of atoms or molecules with a velocity component v, along the laser beam propagation, the resonant frequency in the laboratory frame is wy, — ko, where w,, is the resonant frequency in the rest frame. Then, following (13.8), the population excitation in this group of atoms or molecules is given by (Uz/1,)Ae*(v,) Ap(o,) — Ap*(x,) =>. )~ aetla) (o — wy + ko, + 714+ 7/1) (13.12)218 High Resolution Nondinear Optical Spectroscopy de "s 0 Fig. 13.4 Hole burning in the Maxwellian distribution resulting from resonant excita- tion. Here Ap%(v,) = (Ap™/ vrw)e""/"" and Ap™ is the thermal population difference between the two levels {1) and [2), Clearly, appreciable pumping of the population, Ap(v,) # 4p°(u,), can occur only in those atoms or molecules with v, = (w — w2,)/k, as shown by the dip in the population distrjbution in Fig, 13.4, This is the hole-buming effect.5 The pump excitation has modified 4p%(o,) in such a way that it creates a hole with a halfwidth (1 + £/1,)'7/k in the Maxwellian distribution. Absorption of a Weak Probe in the Presence of a Strong Pump Beam One would think that the hole-burning effect can be probed by a weak beam with a frequency #’ scanned across the hole, The absorption coefficient of the probe beam would be dominated by those atoms or molecules with v, = (w’ — ,,)/k for co-propagating pump and probe beams, and therefore as w’ = w the probe beam should feel the presence of the hole in Ap(v,). This simple description, however, neglects the coherent interaction between the pump and probe beams. As we shall see below, the coherent interaction can significantly modify the absorption spectrum, especially in the strong saturation limit when Li, 21. The calculation of probe absorption can begin with the density matrix element p,(«’), knowing that the absorption coefficient is given by Ari’ aot) = (22 )im x() with (13.13) ot) = NPubn(o’) x(a’) Fo’)Saturation Spectroscopy 29 and p,) = (ller[2>. In the presence of the strong pumping field E(w), we want to find p2,(w’) to all orders of E(w), but linear in the probe field E(«")5 This is possible for a two-level system. Directly from the Liouville equation (2.6) for the density matrix, we can obtain the following set of couple equations (see Section 2.1): Alo’ — wy, + T) pn (o’) = ~PnE(#’)[ Pn) - P2(0)] ~ Pn E(w)[eylo’ - ©) ~ pnw’ — #)], 1 , A[o -wt + Yeuls = «) - py{a’ - @)] 1 = —2pyE*(wo)on(o’) + 2p E(w’) ep(—o) +2pnE(e)epy(o’ - 20), (13.14} h(a! - 26 — oy + T py (o" — 2w) = - 72 E(w) om(o' - 4) — en (e’- #)], [00 raf] = Se 86/|+ ri. | (@ ~ yy +1? ~ Ap, B(- 0) do Pul-#) = @ oy iT We ignore py,(2) and p,.(2w) in (13.14) because they are smal] in compari- son, The solution of p,,(c") from (13-14) is 1 é pala} = qenk (wo yA -1 22? x | (0 — @, +) - (eo! - oy ) ] 39? 20+ w,, +1 2 2 fe F PhE(o) Eo!) on(-e) + . 13.13 ol 29? (13.15) wo wf +i | - ————— Ti} wf -Ww+oy+iF220 High Resolution Nonlinear Optical Spectroscopy which can be written as = = (1/A) pn E(o') Se we — oy, + iT ; =(1/h) py E(w’ (o’ — 20 + wy + AT )(w + w’ — Duy, )22? dp (@ — @y — To" = on + iT)D Pale’) D=(w'-a,+ ini{(w -ot Flo — 20 + ey, + iT) — 22? 1 =20?(w' — 20 + oy, + iT), = lpn E(o}P (13.16) # The first term in p,,(w) of (13.16) arises from the hole-burning effect, while the second term comes from the coherent interference of the pump and probe beams. Physically, this interference sets up an oscillation in the populations ,(o’ — @) and pz,(w” — w), which in turn scatters the pump beam to yield a coherent output at w’. The denominator of p(w’) now has three zeroes, corresponding to three distinct resonances when E(a) is sufficiently strong. This result of strong interaction of light with a resonant two-level system can also be understood from the picture of dynamic (or ac) Stark splitting, or the dressed atom picture, which will be described in Chapter 22. In the presence of Doppler broadening, there is a p,(#") for each velocity group of atoms or molecules. (We assume, for simplicity, that the Doppler broadening is much larger than the homogeneous broadening.) The expression of py,(w’, u,) can be obtained from (13.15) or (13.16) by replacing w and «’ by w — kev, and w’ — k+¥,, respectively. To find the absorption coefficient seen by the probe beam in this case, we must integrate p(w’, v.) over the Doppler profile. In other words, p(w’) in (13.16) should be replaced by {? 40, Py (w’, v,). For counterpropagating pump and probe beams, the hole- burning term in (13.16) yields an absorption coefficient AnNo'| py? ayp(o') = he xf 89°C) de ~e PU/T, | 1+ 2 — wy + kv.) + T? (w= 0, ee) + 1] nity —|y- —— 2 ~ ew : acim | P (13.17) x [(#’ = #) + (0 — wy] +Saturation Spectroscopy 2k where wie) £ r= rp «(a + ry") and The above result shows that over the entire Doppler profile, a_(w’) is approximately equal to ap(’) & Ap"[0, = (@" ~ w2,)/k'| except around w” = 2, — @, where it exhibits a dip with a halfwidth P. Both the amplitude and the halfwidth of the dip increase as JF, increases. This resonant dip seen by the probe beam at w’ = 2w; — @ originates from the hole at w’ ~ ku, = @y created by the pump beam with frequency @ = #,, — £y,. The weak saturation limit (7/I, « 1) is of particular interest. Equation (13.17) reduces to 2rU/i, [(o’ = 0) +(e — ay)] + aT? ] (2338) Oya (’) = ay(w’}(1 — which shows that the absorption dip now has a halfwidth equal to the natural width 2T of the individual atoms. For co-propagating pump and probe beams, the formulation and results are essentially the same except that the probe beam sees a hole w" = w as expected. ‘The term in p(w’) of (13.16) due to the coherent effect is more com- plicated, It can modify the absorption spectrum significantly. Only a qualita- tive discussion is presented here; Ref. 7 provides mathematical details, For counterpropagating pump and probe beams, the coherent part of py,(w', v,) has three poles, but in the contour integration over o, the integration path should be closed in the upper plane that contains only one pole. It gives a contribution superimposed on the hole caused by the saturation effect to form a broader but shaliower dip in the absorption spectrum. This is seen in Fig. 13.5. It is seen that when Z/I, is large, the coherent effect can drastically change the absorption spectrum seen by the probe beam. However, if [/t, <1, the coherent effect is not very significant, and the saturation effect alone gives a good description of the absorption spectrum. For co-propagating pump and probe beams, the contour integral of the coherent part over v, should be closed in the lower plane that contains two poles, The coherent effect again broadens the saturation hole, For 1/I, « 1, the hole is composed of two Lorentzian dips at the same resonant frequency w’ = w, one with a halfwidth 2T and a depth 41/I, and the other with a halfwidth t/7, and a depth 3//T,. [If the two levels have different lifetimes yy! and y;! in the more general case, then them2 High Resolution Nonlinear Optical Spectroscopy 0.5 -02 0 0.2 05 Rfku Fig. 13.5 Absorption spectra with a saturation dip seen by a weak probe beam in the presence of a strong counterpropagating pump wave of different strengths ///,. The solid curves include the coherent effect, but the dashed curves do not. (After Ref. 7.) absorption dip is a superposition of three Lorentzian dips, with halfwidths 2T, Ny Yy and depths 11/T,, 1T/F Y/Y + Ya) FU/GIU/N + Yade Pespec- tively.] By analyzing the shape of the overall dip it is possible to deduce T' and T, (or T, y, and 2). As mentioned earlier, the coherent effect is caused by coherent scattering of the pump wave at « by the population modulation at (@ — w’} resulting from the beating of the pump and probe waves at w and w’. The coherently scattered output at »’ can constructively or destructively interfere with the incoming probe wave causing a decrease or increase in the absorption of the probe beam. Then it is obvious that if pulsed lasers are used, the coherent effect can be eliminated by delaying the probe pulse from the pump pulse. The preceding discussion shows that one can deduce the homogencous linewidth of a transition from the observed saturation dip, but the main purpose of saturation spectroscopy is to resolve the closely spaced lines normally hidden under the inhomogencously broadened profile. This can not be achieved with co-propagating pump and probe beams since the saturation dip always appears at w= w. With counterpropagating pump and probe beams, however, the dip appears at w’ = 20, — @; different transitions with different resonant frequencies w., then show up as different dips and can be well resolved as long as the frequency separation is larger than the dip width. Absorptien of the Probe Beam in the Presence of a Counterpropagating Pump Beam of the Same Frequency The spectroscopic method described above requires two tunable lasers of high monochromaticity, which is seldom affordable. The saturation spectroscopy can, however, be carried out with counterpropagating pump and probe beamsSaturation Spectroscopy 2B of the same frequency. This can be seen from the fact that the counterpropa- gating pump and probe waves interact with the same velocity group of atoms under the Doppler profile when w — ky, = w + ku, = w,, or u, = 0, that is, when w is tuned to the center of the Doppler-broadened transition line. In the case of a strong pump and a weak probe, the calculation is similar to that in the previous section. In (13.14), for the sake of simplicity, the wavevec- tor dependence in the density matrix components is not explicitly shown. Actually, for counterpropagating waves we have p,,(w — w, k° + k). Then even when w = w' the coherent effect is still present because the pump and probe waves can interfere and yield a spatial modulation in the population difference, Ap{K’ + &). However, as pointed out earlier, the coherent effect is relatively unimportant when 1/7, *« 1. We therefore assume //Z, < 1 and neglect the coherent effect in the following discussion. The absorption coeffi- cient of the weak probe beam is then given by (13.18) with @ = «’, A typical experimental arrangement is seen in Fig. 13.6.* The error arising from the beams being not exactly antiparallel is not significant for most applications, As an example, Fig. 13.7 compares the saturation spectrum of the Balmer a-line of atomic deuterium with the emission line profile of a cooled deuterium gas discharge.* The Lamb shift is resolved in the saturation spec- trum. Note that the spectrum in Fig. 13.7 was obtained by subtraction of the saturation spectrum from the original Doppler-broadened spectrum. Each resonant peak here corresponds to an absorption dip in the saturation spec- trum. faser tock-in amplifier signat defector saturating beam probe. Fig. 13.6 Experimental arrangement of saturation spectroscopy with two counterprop- agating waves of the same drequency. (After Ref. 8.)Psy . =I + DaP ae oe { < Site ‘Pip a) Psp n=? Sp Pp SSK I aoe Lamb shift o L L 4 4 1 L —02 —4) 0 od a2 03 a4 frequency (cm) Fig. 13.7 Balmer a spectrum of atomic deuterium. (a) Fine structure of the n = 2 and n= 3 levels. (6) Emission line profile of a cooled deuterium gas discharge and theoretical fine-structure lines with relative transition probabilities (T = 50 K). (c) Observed saturation spectrum with the optical Lamb shift, (After Ref. 8.) 4Saturation Spectroscopy 25 More generally, the pump and probe beams can be of comparable intensi- ties. We consider the special case of two beams of equal intensities and neglect the coherent effect. Under the resonant excitation, the population difference between the two levels of an atom with a velocity component », is obtained from the usual saturation formula A = Ap® plo.) = Ap%o,) ar ; 7 (o— oy they, 4E?2 (wo — ko.) +1? (13.19) 1+ where Fis the intensity of each beam. The absorption coefficient for each bearn is given by = ofa ee} (13.20) “\a[(e — wy + ke) + T?] where ay = (4n0’/cN|P1217/F). In the weak saturation limit, //f, « 1, we find 1+ al (13.21) (@— ey) +0 a(w) = sts - 3(¢) with ay(w) = apdp'( -= — }. The result shows that when |w — w,,| >> I’, the absorption coefficient is a(w) = ay(ol ~ 1/27,), but when Jo — ey} - T, additional absorption appears in the form of a dip, with a depth (4//I,)a(w) and a halfwidth equal to the homogeneous halfwidth F. This is sketched in Fig. 13.8¢. In practice, the two counterpropagating waves of equal intensities can be provided by the field in a laser cavity. The medium in this case can be the laser medium with a negative Sp corresponding to an inverted population, A reduction in |Ap| decreases the gain and ence the laser output. Thus the hole-burning effect appears here as a dip in the gain spectrum or a dip in the laser output spectrum. This was first proposed by Lamb, and is known as the Lamb dip.’ It was the observation of the Lamb dip that opened the field of high-resolution laser spectroscopy. ‘One can, of course, also insert an absorbing medium into a laser cavity. In this case, the absorption generally reduces the laser gain and decreases the laser(Lit) ag (on 4, \ eo (oo) S =, a o (a (e) Fig. 13.8 (a) Lamb dip and (4) inverted Lamb dip in saturation spectroscopy with two counterpropagating waves of the same frequency and intensity. 4/20 oO 4 a 12 16 UE, Fig. 13.9 Dip width as a function of pumping strength 1/J,. Curve A is. exact calculation; B is an approximation ignoring the coherent effect. (After Ref. 7.) 126Saturation Spectroscopy 227 output. Then a dip (decrease) in the absorption arising from hole burning should lead to an inverted dip in the laser gain spectrum, as in Fig. 13.85. This is known as an inverted Lamb dip. The Larmb-dip experiment requires the coincidence of the laser frequency with the transition frequency. It is therefore more limited than the general absorption spectroscopic technique described earlier, We have neglected the coherent effect here. To see that it is indeed negligible for //F, « 1, Fig. 13.9 shows the dependence of the dip width as a function of Z/I, with and without the coherent contribution.’ The difference clearly is appreciable only when /7, >> 1. Multilevel Saturation Spectroscopy The preceding discussion can be extended to a three-level system with two transitions sharing a common level (Fig. 13.10).'° A strong monochromatic beam with frequency = |w,| is used to induce a population change in levels 0 and 1 for a selected group of atoms or molecules. A weak beam at w’ = |wg| is used to probe the induced population change. Since only a selected group of atoms or molecules is seen by the probe beam, the inhomogeneous broadening, of the (Oj > (2| transition is greatly reduced. For this case, the calculation is fairly straightforward. The induced population change has already been de- Tived, so that the absorption or gain spectrum obtained by the probe beam can be calculated readily. The coherent effect is negligible when w and w’ are very different. When w ~ w’, it is still negligible for counterpropagating pump and probe waves in the weak saturation limit, 7/Z, « 1, The technique is most useful for resolving transitions between two sets of closely spaced multilevels. If the difference of the two resonant frequencies |«,9| and fw29| is smaller than or comparable to the Doppler width im a gas medium, saturation spectroscopy can also be carried out with two counterpropagating waves of the same frequency (w = w’))° They interact with the same velocity group of atoms when w and o, satisfy the relations @ — kv, = |@y9| and @ + kv, = |ay9)- The saturation dip appears at w = }{|19| + |e. [). The experiment can be performed in either an absorption cavity or a laser cavity as in the Lamb-dip <2) <2t <0! <1 to wo faa? i(o, + kv ~ @,) h(w, + ky-v—o,,) J Lf no intermediate resonance is involved, the v dependence in [ty can be neglected. Then the only dependence of W), on ¥ is through the lineshape function g(h dw) = g[a(oy + qty + oy + ky*¥ — &)]- (13.23) It is readily seen that if w, = #) and k, 4 ~k,, then the first-order Doppler effect of g(A Aw) is completely eliminated, with g(h Aw) reduced to g[f Qo — @,,) In the case of a Lorentzian lineshape, we have Ufa . (13.24) (20 - wp) +7 g(h Aw) = With g(/ Aw) independent of ¥ to the first order cf v, the two-photon absorption spectrum of a gas medium js then the same as that of stationary atoms or molecules. Here, unlike saturation spectroscopy, all molecules con- tribute equally to two-photon absorption. The absorption coefficient is there- Absorption Atom ann Ow Beam 4 Beam 2 Beam 1 + Beam 2 Beam 1+ Beam 1 and Beam 2 + Beam 2 ao esse wo Fig. 13.13 A two-photon absorption line obtained by counterpropagating waves in a gas medium. The sharp peak is the Dopples-free Hine and the broad background is the Doppler-broadened spectrum.(a} photomultiplier to recorder UY filter AL lignt pipe opticad fsotator merror R=10cm to monitor taser frequency (GH2) (*) oF Fig. 13.14 (a) A typical experimental arrangement for Doppler-free two-photon spectroscopy with a CW dye laser. (6) The 3s-4d transitions of Na observed by two-photon Doppler-free spectroscopy with circularly polarized light. (c) Zeeman splitting of the 3s-4d transitions at H = 170 G observed by two-photon Deppler-free spectroscopy. (After Ref. 8.) Bia2 High Resolution Nonlinear Opticat Spectroscopy fore proportional to the density of molecules, but the observed spectral width corresponds to the homogeneous linewidth. However, with counterpropagating beams, the two participating photons in the two-photon absorption process may come in the manner of either one from each beam or beth from the same beam. While the former gives a Doppler-free line, the latter still leads to a Doppler-broadened peak. The spectrum should appear as a sharp line sitting on a broad background, as in Fig. 13.13. If the intensities of the two counterpropagating beams are the same, the integrated strength of the background should be equal to that of the Doppler-free line. Fortunately, the Doppler width is (ku/T) times larger than the homogeneous width. In the visible range, (ku/T) can be 10? to 10°. Therefore, the back- ground appears to be fairly weak and does not hurt the quality of the Doppler-free spectrum. In some cases the background can also be completely eliminated by using appropriate pump polarizations. Two-photon Doppler-free spectroscopy. is attractive because of its simplic- ity.!° An example is shown in Fig. 13.14, where the hyperfine structure and the Zeeman splittings of the 3S-4D transition in sodium are clearly resolved.’ The technique allows the use of a single laser to probe a transition at twice the laser frequency. The sensitivity of the technique is good, since all the atoms or molecules in the beam participate in the absorption. It can be further improved. with the Auorescence or ionization detection schemes mentioned in Section 12.2. In general, however, two-photon absorption spectroscopy is still less sensitive than saturation spectroscopy because of the lack of an intermediate resonance. Pulsed lasers with high peak power must then be used, and the resolution often is limited by the laser linewidth. If a close intermediate resonance exists, then a CW laser can be intense enough to be used to obtain a high-resolution two-photon absorption spectrum. Figure 13.14 is a good exam- ple. 13.5 HIGH-RESOLUTION POLARIZATION SPECTROSCOPY A polarized pump beam, resonantly exciting a transition in a medium, is expected to induce a dichroism and a corresponding birefringence in the medium through the induced population change. Both the dichroism and the birefringence should exhibit a resonant behavior at transitions involving levels with the induced population change, and can be measured by the polarization variation of a probe beam through the medium. This is the underlying principle of polarization spectroscopy.!® We consider it here as a modification of the saturation spectroscopy discussed in Section 13.3. Only a qualitative discussion is attempted; quantitative details can be found in Ref. 19. Consider a circularly polarized monochromatic laser beam which resonantly excites a transition of a selected velocity group of atoms to saturation.'* This group of atoms will then preferentially absorb the oppositely polarized compo- nent of a probe beam probing the saturation. In addition, there is an inducedHigh-Resolution Polarization Spectroscopy 233 circular birefringence since, as a result of the selective excitation, the refractive indices for the two circularly polarized components are no longer the same. Consequently, in traversing the medium, a linearly polarized probe beam, with frequency in the region of the saturation dip, becomes elliptically polarized along with a rotation of the major polarization axis. If the probe frequency is away from the saturation dip, then the polarization of the probe remains essentially unchanged. The transmitted probe beam can now be analyzed by a crossed analyzer. Only at the saturation dip in the Doppler-broadened peak will the probe light leak through the analyzer. Unlike the absorption saturation spectroscopy, the polarization spectroscopy here yields a Doppler-free spec- trum with no inhomogeneous background. This is certainly an advantage since the sensitivity can be greatly improved without the strong background. It allows the use of lower laser power and lower gas pressure. In general, an elliptically polarized pump beam can also be used as long as the analyzer is set correspondingly* The technique applies to both two- and three-level systems. In the fatter case, there should be a common level for the pump and probe transitions as described in Fig. 13.10. With a system of many closely spaced transitions, as in the case of a molecule, the polarization spectroscopy also has the advantage of being able to greatly simplify the spectrum.”* 7! The polarized pump beam modifies the population of the common level with a particular molecular orientation. The probe beam monitoring transitions from this common level to other levels should experience a dichroism and a birefringence. Then, with a crossed analyzer, these transitions can be casily distinguished from the others. The resulting spectrum seen by the probe beam through the analyzer is usually far simpler and better characterized than the ordinary absorption spectrum. In this technique, the common level is labeled through saturation pumping by the polarized pump beam. It is therefore known as polarization labeling spectros- copy.20.21 Figure 13.15 shows the polarization labeled spectra of Na, as an example.” The pump laser beam was circularly polarized and tuned to a X’2¢ —> BTL, transition of Na, near 4825 A. The first row of spectrum in Fig. ‘b. 15 was obtained with the pump frequency adjusted to the (vo = 0, J = 49) -» (4,50) transition. Transitions from (0, 49) to the various (v’, J’) states were monitored by the polarization change of the probe light induced by the population change in (0,49). For each pv > v’ transition, three spectral lines are expected accord- ing to the selection rules AJ = +1, 0. The signal intensity for AY = +1 is, to the first order, proportional to a}, and for AJ = 0, proportional to 1//?, where ag is the absorption coefficient in the absence of pumping. Therefore, for large J, the AJ = 0 transitions are much weaker and may not show up in the spectrum. We can then identify the successive doublets in the first row of the spectrum in Fig. 13.15 as (0,49) -— [¢0,48), (0, 50), [(1,48), (1,50)], [(2, 48), (2,50). ++, [(6,48), (6,50)], respectively, Similarly, the successive rows of spectra in Fig. 13.15 were obtained with pumping transitions (1,25) > (5,24), E (0,42) > (4,41), (1, 29) + (6,29) and (1,33) + (6, 34), respectively. The » = 1PROBE LASER Ads 300 PUMP LASER B) 60.0! POLARIZER Af4- PLATE GB aT] POLARIZER SODIUM ‘OVEN i) Fig. 13.15 (2) Experimental arrangement of polarization labeling spectroscopy. (8) Polarization-labeled spectra of Na, B'II, band. For the top five spectra, the pump laser was circularly polarized. For the bottom spectrum, the pump was linearly polarized. (After Ref. 21.) 24 ANALYZER SPECTROGRAPHOptical Ramsey Fringes 25 — ¢’ = 3 wansitions in the second and fourth rows are missing because of the small Franck-Condon factors. The upper-state vibrational quantum numbers in the observed transitions can be assigned here easily because each row of spectrum should end up at the low-frequency side with a »° = 0 doublet. The spectrum in Fig. 13.15 is much simpler than the ordinary absorption spectrum, where the transitions from many rotational states of the lower v = 0 anda = 1 vibrational states will create a nearly intangible forest of lines in the spectrum. The polarization labeling technique can also be extended to the effective four-level systems described in Fig. 13.12. The polarized pumping partially orients the molecules in the 0) level. Through molecular collision, the orienta- tion is transferred from |0) to (0). The probe beam probing the transition from 0’) to 2) should again experience an induced dichroism and birefringence.”° Polarization spectroscopy also can be used to study two-photon transitions. Instead of absorption, the polarization variation of the incoming beams is measured, We consider this in more detail in Chapter 15 under the rubric of Raman-induced Kerr effect, recognizing that Ramat transitions are just a special case of two-photon transitions. 13.6 OPTICAL RAMSEY FRINGES In radio atomic spectroscopy, an ingenious high-resolution technique was invented in the late 1940s by Ramsey using an atomic beam.” As seen in Fig. 13.16, an atomic beam interacts with the applied radio fields in two regions. Atoms passing through the first region are coherently excited. If the coherence persists when the excited atoms reach the second region, they may absorb er emit, depending on whether the atomic coherence is in phase or out of phase with the exciting radio field. The absorption in the second region therefore appears as interference fringes as the radio frequency scans through the atomic resonance. These are known as Ramsey fringes. The fringe pattern depends on the coherent lifetime (dephasing time) of the atomic excitation from which the high-resolution spectrum can be deduced, The technique can be extended to optical spectroscopy.”? Since the dephas- ing time of an optical transition is usually very short, a modification of the technique is necessary. Instead of the em fields exciting an atomic beam at two separate spatial points, one can use two laser pulses separated in time to excite the same group of atoms in a vapor cell.2*2> The effect is the same. The Radio fields gh er bes Fig. 13.16 Schematic deseribing the interaction of an atomic beam with two radio frequency fields leading to the observation of Ramsey fringesBE High Resolution Nonlinear Optical Spectroscopy absorption spectrum seen by the second pulse or the emission spectrum after the second pulse exhibits the Ramsey fringes. It is interesting to note that the spectroscopic resolution of the Ramsey-fringe technique is not limited by the pulsed laser linewidth, which, by the uncer tainty principle, is given by the inverse of the pulsewidth, This can be seen as follows, For two identical coherent pulses separated by time T, the field can be written as E(Q4 fdo[8(w)sin wt + 6(w)sin(we + oT] (13.25) = fav [2e(w)sin eos we +f) with w = w, + Aw. As seen in Fig. 13.17, it has a sawtooth spectrum whose envelope is the spectrum of a single pulse. Then, in tuning the laser over a resonance, it is the sawtooth spectral profile of the laser scanning over the resonance. Therefore, the spectral resolution is now limited by the width of the sawtooth instead of the width of the envelope. In scanning w, over a 5-function resonance, the linear transmission spectrum should reproduce the sawtooth spectrum of the laser excitation. If the resonance has a finite linewidth, then the sawteeth in the transmission spectrum are broadened with a decreased contrast of peaks to background. The Ramsey fringes depend on the phase correlation of the two pulse excitations. If the pulses are not phase-correlated, no fringes can be seen. Also, if the coherent excitation imposed on the atoms by the first pulse dephases before the second pulse arrives, the fringe pattern will disappear. The dephas- Fig. 13.17 Single and double coherent laser pulses with their corresponding frequency spectra.(Oprical Ramsey Fringes. BT ing effect on the observed spectrum is in broadening the sawteeth and. decreasing the contrast of the fringes. In an actual experiment, atomic motion also may cause dephasing. This, however, can be eliminated if two-photon excitation is used.” As discussed in Section 13.4, the former is independent of atomic velocity and has all atoms participate in the absorption. In the following discussion, we use this case as an example to give a simple mathematical derivation of the Ramsey fringes.” Consider a two-level system excited via two-photon transitions by two square pulses of duration 7; one is switched on at ¢ = 0 and the other at = FT. At? < 0 the system is in state (1), and at ¢ > 0 the system is in a coherent state We) = a(e}fl) + (1). (13.26) Assume that the perturbation theory is valid. Then a(7) = 1, and b(r) can be obtained from the Schrédinger equation or ina(s) = Ae ent 13.27 a where A is a coupling coefficient proportional to the square root of the two-photon transition probability. We find, for : = 7, Be) = A [eter — 1] (13.28) 2 and forr >= T+ 17, = A -HQw—wyXT+7) gee) T b(i= Jaw, 1 e 2T) + b(t} A (13.29) = -iQw-wy)t Hien wy} F Jeon [e wt 1]fe wT +], If fluorescence from ¢2{ is detected, the signal should be proportional to \b(1}?, which, after the second pulse, is given by 7). (13.30) al _ 2 = p(T + P= wee] SICe— eat ew cos? 2 5 ou (Qe — wy )}r } The spectrum of |b(T + 7)|? versus (2w — w,) is seen to be a fringe patiern . shown in Fig. 13.18, The periodic spacing of the fringes is 1/27. In the f presence of a dephasing rate I with x < ['-}, (13,29) should be modified by28 High Resolution Nonlinear Optical Speciroscopy Await i i i Fence Fon fogtair © Fig, 13.18 Ramsey fringes resulting from two-photon excitation of a transition by two square pulses of duration 7, one switched on at ¢ = 0 and the other at ¢ = 7. replacing b(7') by b(T )e'7. We then find (T+ 1)P = (ota) tee — Afertem enh ee PTYP. (13.31) The fringe contrast is given by (T+ bag = I(T + ban 2 2 crs la +1B(T + Yon (3.32) which is 1 for T = 0 and decreases rapidly as I’ increases. Figure 13.19 is an example of the optical Ramsey fringes obtained by two-pulse two-photon excitation in sodium together with the experimental arrangement.® On each spectral line, the interference fringes are clearly visible. The observed periodic spacing of the fringes is indeed given by 1/2T and decreases with increase of 7. An obvious improvement of the Ramsey fringes can be achieved by using a series of many equally spaced pulses.” It can be obtained from either a mode-locked laser or a pulse through a resonant cavity with partially transmit- ting mirrors. If the pulses are phase-correlated, the frequency spectrum will appear as a series of equally spaced spikes, as in Fig. 13.20. The spectral resolution is now limited by the width of cach spike. With ¥ pulses coherently exciting the atoms within the dephasing time, the width of the spikes isMirror Sodium ceil and oven fa) 325-4? TWO PHOTON TRANSITIONS IN No ! LASER FREQUENCY (GHz) (8) Fig. 13.19 (a) Arrangement of the Ramsey fringe experiment in Na. (5) Observed spectra of the 37S — 47D two-photon transitions in Na; the lower trace, showing the Ramsey fringes on each peak, was obtained with a delay of 25 nsec between the two excitation pulses. (After Ref. 27.)240 High Resolution Nonlinear Optical Spectroscopy Fig. 13.20 A series of equally spaced laser pulses and: the corresponding frequency spectrum. inversely proportional to N, while the signal intensity is proportional to N?. This technique has been demonstrated by Teets et al.” 13.7 OTHER HIGH-RESOLUTION SPECTROSCOPIC TECHNIQUES There are a number of other high-resolution spectroscopic techniques. Most of them are variations of the techniques we have already discussed, including multiphoton Doppler-free absorption and multiphoton saturation spectros- copy. Coherent transient spectroscopy and four-wave mixing spectroscopy, however, are unique and deserve a more detailed discussion. The former is discussed in Chapter 21 on coherent transient effects, and the latter in Chapter 15. REFERENCES 1 IL. Hall, in M.S. Feld, A. Javan, and N. Kumit, eds. Fundamentaf and Applied Laser Physics (Wiley, New York, 1973), p. 463; A. Brillet and I. L. Hall, Phys. Rev. Lert. 42, 349 (1979) 2B. Bleaney and K, W. H. Stevens, Rep. Prog. Pys. 16, 108 (1953). EB. Aleksandrov, Opt. Spectrosc. 17, 957 (1964); J. N, Dodd, R. D. Kaul, and D. M. ‘Warrington, Proc. Pays, Soc. (London) 84, 176 (1964). 5. Haroche, J. A. Paisner, and A. L. Schawlow, Phys. Rev. Lett. 30, 948 (1973). See, for example, N. Bloeaibergen and Y. R. Shen, Phys. Rev. 133, A37 (1964). W. R. Bennett, Phys, Rev. 126, 580 (1962). Y_ 5, Letokhov and V. P. Chebotayev, Nontinear Laser Spectroscopy (Springer-Verlag, Berli, 1977), Chap. IL we AAW ekBibliography ul 8 T. W. Hansch, in N. Bloembergen ed., Nonlinear Specirascepy {Norlh- Holland Publishing Co.. Amsterdam, 1977), p. 17. 9 W.E. Lamb, Phys. Rev. 134A, 1429 (1964); R.A. McFarlane, W. R. Bennett, and W. E. Lamb, Appl. Phys. Lest. Z, 189 (1963); A. SzOke and A. Javan, Phys. Rev. Lett. 10, 521 (1963) 10 H.R. Schlossberg and A. Javan, Phys, Rev, 150, 267 (1966); erratum: Pays. Reo. AS, 1974 972). It J. L, Hall and C, Bordé, Phys. Rev. Lett, 30, 1101 (1973), 12 R.G. Brewer, RL, Shoemaker, and §, Stenholm, Phys. Rev. Lett. 33, 63 (1974). 13. M.S. Feld and A. Javan, Phys. Rev. 177, 540 (1969). 14 A, Sabo, Phys. Rev. Lett. 25, 924 (1970). 15. L,S, Vasilenko, V. P. Chebotagev, and A. V. Shishacy, JETP Leet. 12, 113 (1970). 16 F. Biraben, B. Cagnac, and G. Grynberg, Phys. Rev. Lete. 32, 643 (1974); M. D. Levenson and N. Bloembergen, Phys. Rev. Leit. 32, 645 (1974); T. W. Hansch, K. C. Harvey, G. Meisel, and ALL. Schawlow, Opt. Comm 1, 50.1974). 17 F. Birabeo, B. Cagnac, and G. Grynberg, Phys. Lett. 48A, 469 (1974). 18 C. Weiman and T. Hansch, Phys. Rev. Lett. 36, 1170 (1976). 19 C. Weiman, Ph. D. dissertation Stanford University (1976). 2 R. Tels, R. Feinberg, T. W. Hansch, and AL L. Schawlaw, Phys. Rev. Lett, 37, 683 (1976) 21M. B. Kaminsky, R. T. Hawkins, FV. Kowalski, and A. L. Schawlow, Piys. Reu, Lett. 36, 671 (1976). 22 N, F, Ramsey, Phys. Rev. 76, 996 (1949), 23 J, C, Gergguist, S.A. Lee, and J. L. Hall, Phys. Rev. Leis. 38, 159 (1977) WA Ye, ¥, Baklanov, V, P, Chebotayev, and B. Ts, Dubetsky, App/. Phys. 11, 201 (1976). 35. M, Salour and C. Cohen-Tannoudji, Phys. Rev. Lett, 38, 757 (1977) 2% M, Salour, Rev. Mod. Phys. $0, 667 (1978). 27 R. Teets, J, Eckstein, and T. W. Hansch, Phys. Rev. Lert, 38, 760 (1977) BIBLIOGRAPHY Bloembergen, N., ed., Nonlinear Spectrascopy (North-Holland Publishing Co., Amsterdam, 1977) Desntrdder, W., Laser Spectroscopy (Springer-Verlag, Berlin, 1981). Hansch, T. W., Physics Today 30, No. 5, 34 (1977). Letokhov, ¥. 8. and V. P. Chabotayev, Nonlinear Laser Spectroscopy (Springer-Verlag, Berlin, 1977). Proceedings of International Laser Spectroscopy Conferences: 1. R. G. Brewer and A. Mooradian, eds. (Plenum Press, New York, 1973); II, §, Haroche, J. C. Pebay-Peyroula, T. W. Hansch, and 5. E. Harris, eds. (Springer-Verlag, Berlin, 1975); INI. J. L. Hall and J. L. Caristen, eds. (Springer-Verlag, Berlin, 1977); TV. H. Walther and K. W. Rothe, eds. (Springer-Verlag, Berlin, 1979}; V. A. W. McKellar, T. Oka, and B. P. Sloicheff, eds. (Springer-Verlag, Berlin, 1981); VL H. Weber, ed. (Springer-Verlag, Berlin, 1983). Schawlow, A. L., Rev. Mod. Phys. 54, 697 (1982). Shimoda, K., ed., High-Resolution Laser Spectroscopy (Springer-Verlag, Berlin, 1976) Shimoda, K., and T. Shimizu, Prog. Quant. Efectron. 2, 47 (1972). Walther, IL, ed., Laser Spectroscopy of Atoms and Mofecules (Springer-Verlag, Berlin, 1976).14 Four-Wave Mixing Four-wave mixing refers to the nonlinear process with four interacting electro- magnetic waves. In the weak interaction limit, it is a third-order process and is governed by the third-order nonlinear susceptibility. Unlike second-order processes, a third-order process is allowed in all media, with or without inversion symmetry. Yet it is generally much weaker than an allowed second- order process because of disparity in the sucsceptibilities, be] « [xO , With high-intensity, lasers, however, it is still easily observable, as first demonstrated by Maker and Terhune.’ This is particularly true if |x| shows resonant enhancement, When more than one tunable laser is used for pumping, even multiple resonances of x‘ can be excited. Being flexible and easily observable in all media, four-wave mixing has many interesting applications. It extends the frequency range of tunable coherent sources to the infrared and ultraviolet.? In the degenerate case (i.c., four waves having the same frequency) it is used for wavefront reconstruction in adaptive optics.’ With resonances, it can be adopted as a powerful spectro- scopic and analytical tool for material studies. We consider the fundamentals as well as some applications of four-wave mixing in this chapter, and leave the discussion of four-wave mixing spectroscopy to Chapter 15. 14.1 THIRD-ORDER NONLINEAR SUSCEPTIBILITIES In media with inversion symmetry, third-order nonlinearity is the lowest-order nonlinearity allowed under clectric-dipole approximation.’ The microscopic expression of a third-order nonlinear susceptibility x can be derived from 2 perturbation calculation using, for example, the diagrammatic technique out- lined in Section 2.3. In general, it consists of 48 terms, explicitly shown in Ref. 5. While x is governed by the overall symmetry of the bulk medium, each term of x© is governed by the selection rules on its matrix elements. waz‘Third-Order Nonlinear Susceptibitities 3 Near resonances, a few terms of x® are resonantly enhanced through the tesonant denominators. The resonant part of x® can be separated from the nonresonant part through resonant dispersion, The former is a complex quantity as the damping coefficients in the resonant denominators become nonnegligible. A few examples are discussed here. Singly Resonant Cases Assume three input pump frequencies «,, «,, and w;. Single resonance of x) occurs when any of the three frequencies or their algebraic sums approach a transition frequency of the medium. Consider as an example the case seen in Fig. 14.1 where w, — w, is at resonance. The third-order susceptibility can be written as the sum of a resonant part x@ and a nonresonant part xh xP = x 8h + xk. (14.1) The expression of x can be obtained either from the general expression of x© or from the derivation in Section 2.3 or 10.4. We find N( Mera )in( Miers) g(x — Pe) h(a, — w, — wy, — Ty} [xQ(w, = os — 62 + Oy) yer = — (14.2) where (MdyD| Aw, — - ee) | er,|# > aler; ‘Jer,[n {mi er, (42,),,=E Calenladcalenig) _ By | 2a, — 0, — Opyy 2, ~ 0) + ig ge 1 (14.3a) x Ce Vale Cade Oey (0, — ayy + Tyg) (Wy — @y — gig + Tyg) + terms with j and / interchanged [x@(o; = 0, — @ + 04)] sae _ Key églrbe’> BOTY day Wy ~ ayy + Tyg (14.3b) x Cals cnlraly Cael ocala) (alts dee @] — @y — Ong 20, — Omg @ ~ a, + iT, + terms with / and / interchanged [x(ey = 2, — 2 + 0y)f ine Ne* (alein’)< airs 678) Oe x ([(4) — og + Bee) (eh — 02 - Oy - iTy,) - 144 x0} — yy — ley) t+ (0h op tity) OO [lor ~ Pag Ty)" — (oj Wy, + Tye) )} [xQP( us = wy — 0h + oy) yer X (wy — wf + i/ Thy 4 = AS calaln) niga Xatrle’ do) x {(wp og + Tag) (et — #2 — Oy Tie) . - 14.4b ‘5 Tys) ‘| ( ) [cy ~ eg + Tag) — (eh ~ + (003 — cay + Tyg) (oy - 0 + iT) x (e, - & = Ty) —(w{ ~ + Tye) |}: These expressions are for isolated molecules or ions, As discussed in Section 13.1, the resonant frequencies of molecules or ions often depend on the localGeneral Theory of Four-Wave Mixing wT environment. The effective x) of an ensemble of molecules or ions should be a weighted average of x) over the distribution of the resonant frequencies. 14.2. GENERAL THEORY OF FOUR-WAVE MIXING The theory of four-wave mixing follows closely the general theory of optical mixing. We assume here, for simplicity, a cubic or isotropic medium. Three different cases are considered in this section. Three Pump Fields Let the pump fields be E,,(0,,) = &,exp(ik,, 6 — iu,,¢) with m = 1, 2, 3 (Fig. 14.4a). The output field, E,(o,) = &exp(/k, +r — ia, f) with w, = w) + & + 4, is governed by the wave equation ut 2+ Latasfe,~ Epa) (14s) where PO(,) = xe, = @, + w2 + &5): E(w, E(w, E(u). The solution of (14.5) follows that described in Chapter 6. With the usual slowly varying atoplitude approximation, negligible pump depletion, and the simplifying boundary condition, it yields 2aw? 2) 5- (ake Fa Sl — elk yore (14.6) 3 E, Ey E, = (}- Es (a) (a) 2) = — om ED E;(0) 50 <=, a EM EJ’ (od Fig. 14.4 Three different types of four-wave mixing discussed in Section 14.2.248 Four-Ware Mixing where Ak = Ak’ + (Ak” = (kj + ky + - ki) + Coq) + oy + Oe a), Ak’ is the wavevector mismatch, and the «,,; are the attenuation coefficients of the waves along 2. ‘As is common for optical mixing, phase matching (Ak = 0) is of prime importance here, since it greatly enhances the signal output. In four-wave mixing, phase matching can be achieved in an infinite number of ways by properly adjusting the directions of propagation of the three pump waves. Which arrangement is preferred often depends on practical considerations, such as optimum beam overlapping length and better spatial discrimination against scattering background. Output Field in the Same Mode as One of the Input Fields In this case, we take E,, = Es; (Fig. 14.46). The input field E,, should then experience gain or loss induced by the nonlinear wave interaction. Since o, = 0, and k, =k, in (14.5), we must have @, = — #2 and Ak = k(#,) + k,(@,). With negligible depletion of E, and Ep, the solution becomes 8, (2) = &,(Oexp[ gz) ~ a2], mw? ae (4.7) gdz)= Tax ye eto — else), The real part of g,(2) represents a gain. For the special case of £, (a) = Ed,(w) and Ak” = 0, we find 2a? Ref g)(2)] = ee tf 163172. (148) kc which can be compared with the result of Raman gain discussed in Section 10.3, Backward Parametric Amplification and Oscillation This is a special case of four-wave mixing in which two strong waves act as the pump fields and two counterpropagating weak waves g¢t amplified (Fig. 14.4c). It resembles the parametric amplification case of Section 9.6, except that two pump fields instead of one are used here. The two weak waves are the signal and idler waves, respectively. The solution then is essentially the same as that described in Section 9.6. ‘Assuming perfect phase matching, which can be achieved easily in this case, and negligible pump depletion, we find, followingDegenerate Four-Wave Mixing 19 the derivation in Section 9.6, for the signal and idler fields, E, and E,, propagating along +2, respectively, V/2 6,(2 = 0) = 8,(0) /c0s52t + (| gp(ojtan®e, 14.9) ta ( ky Bol wt | 62-9) ~i24 | ai )ianSt! + 6.4(0)/008- where Bat _ ( siet (3) (Fe, eel" (14.10) ln, og sae K= Sex’ Mw, = —@; + @ ety) 62182 Cc and £, and E, are the pump fields. As gof approaches 7, both (0) and &(/) diverge according to (14.9). This indicates the onset of oscillation, which yields an output even in the absence of any input, &,(?) = &(0) = 0. In the case of sufficiently weak pump fields, gol « 1, and (2) = &,(Q), the signal output should reduce to the expression for an ordinary four-wave mixing process with three pump fields. 443 DEGENERATE FOUR-WAVE MIXING e of four-wave mixing in which all the four The third-order nonlinear polarization govern- three components with different wavevectors: We now consider a special cas waves have the same frequency. ing the process has, in general, PO(a) = PO(K, + ky - k,,@) + PO(k, — kL + k,,0) gaat) + POY ky + ky +k) ‘ where PO (hy + kj — @) = Oo) Ey (Es (KEP) P(e — Wi + ki, &) = x(w) By (k EV (KE). PO(-k, + ki +h, 0) = Oe) TERK EK JE (k,)- E,(k,), Exkp, and E,(k,) are the three input fields, and all xO(w) in the above equations are the same under the electric dipole approximation. We note that in this case xu) has at least a singly resonant term arising, from the250 Four-Wave Mixing two-photon zero-frequency resonance, i.¢., a term with (w — w + i/7;) in the denominator. It may also have a two-photon resonant term if w + # is in resonance with a transition of the medium, Finally, x can be triply resonant if w is near a resonance. Because of the strong resonant enhancement, x for degenerate four-wave mixing can be very large in some media, As a result, such a third-order process is even observable with CW laser beams. The output of degenerate four-wave mixing can be calculated using the theory in Section 14,2, but it can also be easily understood from the folowing physical picture. Two of the three input waves interfere and form either a static grating or a moving grating with an oscillation frequency 2a; the third input wave is scattered by the grating to yield the output wave. In many cases where 2w is away from resonance, the contribution from the static gratings should dominate. With three input waves, three different static gratings are formed. The grating formed by the k, and k, waves scatters the k; wave to yield outputs at k, = ky + (k, — k;). The one formed by the kj and k, waves scatters the k, wave to yield the outputs at k, =k, + (kj — k,). The one formed by the k, and kj waves scatters the k, wave to yield the outputs at k, =k, + (k, — kj), They are illustrated in Fig. 14.5 for the special case of ki = —k,. Altogether, three output waves with different wavevectors, k, = k, +k, -—k, k,-kj+k,, and —kj + kj +k,, are expected. However, we realize that since k,|, in general, is not equal to wel/*/c, the generation of the three output waves may not be all phase matched. Consider, for example, the case with kj = ~k,. The output waves are expected to have k, = —k, and k, + 2k,, While the generation of the output at k, = —k, is always phase matched, that of the other two is not. Thus usually only the output at k, = —k, needs to be considered. It is interesting to see the connection between this case and holography. In both cases, the output waye (k,), arising from scattering of one of the pump waves (k, or kj) off the interferogram formed by the object wave (k,} and the other pump wave (kj or kj), retraces back the path of the object wave (k, = —k,). Now that an object can be represented by a group of k, waves, we see that an image of the object can be reconstructed by the corresponding output waves. In an isotropic medium, if we require the output of degenerate four-wave muxing not to be in the same mode as one of the pump waves, then for phase-matched output we must have kj = —k, and k, = —k,. The effective nonlinear polarization can be written, from symmetry consideration, as PA(k, = —ky, 0) = x(a) + Ey (ky E((—k EF (K,) ~ A(E, ESE; + B(E{ Es )E, + C(E,E{)Et (14.12) where A, B, and C all depend on the angle @ between E, and E,, and B(8) = A(a — @). The brackets (E,*E*) and (E,*E}) in the A and 8B terms in (14.12) describe the static gratings formed by the wave interference, whilePhase Conjugation by Four-Wave Mixing 231 kya ky ey ky =k; + ky ~ ky = kj + 2k, ky kg = hi (8) te) Fig. 14.5 Degenerate four-wave mixing resulting from scattering of an incident wave by the static grating formed by the other two incident waves: (a) grating formed by the k, and k, waves, (6) grating formed by the kj and k, waves, and (c) grating formed by the k, and kj waves. (E, °E{) in the € term is a moving grating with an oscillation frequency 2. By properly arranging the polarizations of the three incoming waves, it 1s possible to have only one particular term in (14.12) nonvanishing. The output is polarized along P°). Being in the backward direction with respect to the incoming k, wave, it can be described by the solution of (14.9).” We then notice that with &,(/) = 0, the output field &,(0) has a magnitude proportional to that of the input field ¢,(0), and a phase complex conjugate to that of €,(0). 14.4 PHASE CONJUGATION BY FOUR-WAVE MIXING Phase conjugation® is defined as the process in which the phase of the output wave is complex conjugate to the phase of an input wave. In other words, the process reverses the phase of the input. This happens, for example, in difference-frequency generation, parametric amplification, and four-wave mix- ing. If the phase-conjugated output propagates in the backward direction with respect to the corresponding input wave, then it can be used to correct abberation due to phase distortion experienced by the input wave. As il- lustrated in Fig. 14.6, the input beam in passing through a medium suffers aIncident Mirror Distorting medium Reflected wave A Phase —conjugate Incident mirror wave Distorting medium 4 + \ | ! vs I a | Reflected wave (a) (b} Fig. 14.6 Sketches showing how the wavefront of a beam changes in passing through a distorting medium back and fourth: {@) an ordinary mirror is used, and (#) a phase-conjugate mirror is used to reflect the beam. Fig. 14.7 Photographs showing correction of aberration; {@) an unperturbed laser beam, (5) the same laser beam after passing through an etched glass plate, and (c) the same beam after a phase-conjugate reflection and a second pass through the etched plate. (After Ref. 9.)Phase Conjugation by Four-Wave Mixing 253 wavefront distortion. Unlike an ordinary mirror, the phase-conjugate mirror reverses the wavefront distortion of the input beam upon reflection. Then, as the phase-conjugated wave reflects back through the medium again, the wave- front distortion is completely removed (as long as diffraction is negligible}. An example is seen in Fig, 14.7. In the previous section, we saw that the output of degenerate four-wave mixing is a reflected phase-conjugate wave with respect to one of the input waves [&(0) « &*(0)|. Thus a nonlinear medium for degenerate four-wave mixing can be used as a phase-conjugate mirror.’ According to (14.9) with &(1) = 0, the phase-conjugated output |¢,(0)|* can even be more intense than the input |&(0)|? if gol/2 > 7/4, and the amplifications approaches the oscillation limit as gof/2 — 2/2, fa} | Conjugator e tilumination Wluminetor Reflection and amplification ) ——_ (6) | Conjugator Ampiitier } : 7 \ Conjugation {c) | Conjugator | \\—>|_ Amplifier 4 ( {@)_ | Conjugator Amplifier > > Correction of amplified wave Fig. 14.8 Wavefront reconstruction applied to laser fusion. The sequence of events is (a) illumination of the target with a probe beam, (4) reflection and amplification resulting in a distorted wave, (c) phase conjugation, and {d) a second pass through the distorting amplifying medium, producing a correct focusing of the beam on the target. (After Ref. 9.)154 Four-Wave Mixing There are many interesting and potentially important applications of phase conjugation based on its ability to remove wavefront distortion.? One is for correction of distorted images. As an example, consider the amplification of a laser beam in an amplifier. The beam quality may be seriously deteriorated after the beam traverses the amplifying medium. If, however, a phase-con- jugate mirror is used to send the beam back through the amplifier once again, then the amplifier output can have its input beam quality restored. This allows the construction of high-power laser systems with beam quality comparable to that of a single-mode oscillator. One can also have a laser oscillator with one of its mirrors replaced by a phase-conjugate mirror, which tends to help in establishing a better beam quality, improve the mode stability, and possibly provide additional gain to the oscillator. In Jaser fusion work, beam focusing on the target may be impaired by wavefront distortion of the beam in its propagation from the laser to the target, This can be remedied by the scheme of Fig. 14.8. The target is first illuminated; the wave radiated from the target is then amplified by the amplifier, reflected by a phase-conjugate mirror, amplified again, and finally automatically focused onto the target with no net distortion of its wavefront. The scheme can actually be used on any target, not necessary in laser fusion work. In principle, it can also be applied to a moving target ata distance as the beam automatically tracks the target. This may have great potential in military applications. With sufficient gain in the amplifier, it is even possible to have laser oscillation with the target and the phase-conjugate mirror forming the cavity, so that no external illumination en the target is necessary. Phase conjugation can also be obtained from stimulated scattering of a highly multimode pump beam." In this case, the output is phase conjugated to the input pump field. This has been observed experimentally. Theoretically, however, it is shown that only approximate phase conjugation can be achieved, but the approximation becomes better when the number of pump modes increases. A better physical understanding of the effect is still needed. 145 TUNABLE INFRARED AND ULTRAVIOLET GENERATION? Four-wave mixing can be used to extend the frequency range of coherent radiation to infrared and ultraviolet. The process is again governed by the general theory of optical mixing, It is the third-order susceptibility x(w, = , + @) + w;) that determines how efficient the frequency conversion process can be with given pump intensities, As shown in Section 7.4, x for third harmonic generation can be greatly enhanced by resonances, In alkali vapor, for example, it increases from less than 107 esu/atom to ~ 1077! esu/atom near a single resonance, Now, in four-wave mixing, w,, w;, 3 are not necessarily equal; it is even possible to have x doubly or triply resonant. A few examples are schematically shown in Fig. 14.9 using potassium vapor asTunable Infrared and Ultraviolet Generation 255 G Bs Pp Ad 5p-—— w “5 wy Ss aa wo, % we © w aw * Ws — ‘3, 3 wg ony w ty 4s @ ce) (2 (4) Fig. 14.9 Energy level diagram of a potassium atom and schematics for a number of resonant third-order optical mixing processes: (a) third-harmonic generation, (6) sum frequency generation with w,, a, + w2, and w, + w, + w, near resonances, (¢) in- frared generation with w, and w, — w, near resonances, and (d) infrared generation with w, and w, + w near resonances. the nonlinear medium. As one would expect, x doubly or triply resonant can be orders of magnitude larger than that of the singly resonant case. We use here the process of Fig. 14.9c for illustration. Let us assume that in Fig. £4.9c, #, is near the 4s > Sp resonance, a, — «2 is exactly on the 4s > 5s resonance, and w, is not too far from the 4s > 4p resonance. The dominating resonant term of x" in this case is Ne 4 (oP ).er = (AE ] apts etp dtpirse) (Sein Br) Spins) SP Ag = [Ce — spas + ITs, (oy ~~ 554, + 9T5,) (14.13) X(a— Oya] where D, , sums over the fine structures in the s and p levels. With @, — Ws, - 45 = 30 om}, Ty, = 0.1 om™’, and w, — @4,-4, = 5000 cm™, we find x = 6 x 107” esu/atom.™ For an atomic density of 10'’ atoms/cm’, the value of x = 6 x 107 esu is already larger than the nonresonant x"? of a typical condensed matter, We notice that since «, is still far away from resonance, x? does not vary appreciably with w,. This, together with the large value of x, means that efficient generation of tunable output over a broad range of infrared frequency w, = w, — @, — w, is possible as long as collinear phase matching can be achieved. If w, is tuned toward the 4s > 4p resonance, then x°" is further enhanced. For #; — w,,_ 4, = 50m, we have x = 6 x 107% esu/atom.256 Four-Wave Mixing Collinear phase matching is essential for high conversion efficiency. In the above infrared generation, it may be achievable using anomalous dispersion of the alkali vapor. The collinear phase matching relation k, = ky + ks + k, for the process of Fig. 14.9¢ can be written in the form Oy, = W_N, t agNs + OT. (14.14) Since both w, and w, are very far from resonances involving the ground level, the normally dispersive refractive indices nr» and n, do not vary appreciably with w, and «,. For a prescribed output frequency «,, the frequency 1; is fixed from the relation a, 4, = #3 + ©, = @) — 2, and hence both », and #3 are fixed, Now that 7, has an anomalous dispersion while 7, is nearly independent of w), it is possible to satisfy (14.14) by adjusting , and @ properly. For smaller ,, we must have «, closer to wsp_4, to achieve phase matching. This unfortunately increases the absorption of the pump field at w, and decreases drastically the efficiency of infrared generation, The problem may be alleviated by mixing foreign gas into the vapor, for example, sodium in potassium, and utilizing the additional dispersion in the refractive index provided by the foreign gas.'* Tunable infrared generation using the process of Fig. 14.9¢ was actually demonstrated by Sorokin et al.'?"" In their experiment, only two pump beams, at w, and «,, were used, while the @, beam was automatically generated in the cell by stimulated Raman scattering of the w, beam. With peak powers of 1 kW at w, and 10 kW at w; and an active length of 30 cm in potassium vapor, they observed a tunable infrared output from 2 to 25 »m having peak powers of 100 mW at 2 pm and 0.1 mW at 25 am. Extension to a broader tuning range may be possible. The theoretical analysis of this experiment is not yet complete, Strictly speaking, it is a four-wave parametric amplification process with waves at @, and w, being the signal and idler waves. The calculation should be a straightforward extension of that given in Chapter 9, including a Raman resonant transition. As a variation of the process, the Raman transition w, — @, = Ws, 4, Cal be replaced by a two-photon transition @, + @) = @s,—4;- The process should be at least equally efficient if w, is also near the w,,_ 4, Tesonance, as in Fig. 14.9d, Then, for infrared generation, #, can be either larger or smaller than s,_ 4, in both Fig, 14.9¢ and Fig. 14.9d. A reverse four-wave mixing process is, of course, also possible, Thus, by using «,, 2, and w, as pump frequencies in Fig, 14.9¢, the w, wave is generated. The process can therefore be used to construct an infrared-to-visible converter.'4 With laser powers much higher than 10 kW, the conversion efficiency of infrared generation by four-wave mixing can, in principle, be high. It may, however, be limited by other simultaneously occurring nonlinear effects, such as sélf-focusing, saturation, and jonization. A thorough study of how the conversion efficiency can be improved has not yet been reported.Tunable Infrared and Ultraviolet Generation 257 Four-wave mixing can also be used to generate tunable ultraviolet radiation.”"* In comparison with third-harmonic generation, the four-wave mixing process has the advantage that x°)(w, = a, + w, + ,) can be greatly enhanced through multiple resonances. An example appears in Fig. 14.9b. Alkali vapor, however, is not very good for vacuum uv generation because its ionization energy is too low and there is no discrete state in the ionization continuum to resonantly enhance x® at the uv output frequency. Alkali earth vapor appears to be much better. The process of Fig. 14.10 is used in Sr vapor for an example. The pump frequencies w, and w, are adjusted so that w, is near the (5s)? > (Ss}(5p) resonance and w, + w, is on the (5s)? > (Sp)? resonance. The other pump frequency w, is near the (Spy > (6s)(6p) reso- nance, and can be tuned to yield tunable uv output at w, = @, + @ + 3. Note that (65)(6p) is a discrete autoionization state in the ionization con- tinuum. Because of the multiple resonances, x®(w,) can be greatly enhanced. (6s) (5p) i lonization limit —] Lis 40,000 F/- sp? (85) (75) (5s) (6p) a |_ (5s) (6s) a (85) (6s) - s a (85) (5p) 20,000 }— Ho OL I {Bs}? Fig. 14.10 Energy level diagram of Sr with arrows showing a resonant four-wave mixing process for tunable uv generation at 2w, + 5.258 Four-Wave Mixing To avoid strong attenuation of the pump beams, the pump frequencies should be sufficiently far away from resonant absorption. Then, with phase matching, the conversion efficiency for uv generation can be significant. Molecular gases, such as CO and NO, also can be used as effective media for tunable uv generation by four-wave mixing.** Their resonantly enhanced nonlinear susceptibilities are Jower in comparison with those of metal vapors because of the weaker resonances, but the gases are much casier to handle. The medium can be in the form of a molecular beam, which, for vacuum uv spectroscopy, has the advantage of requiring no window between the uy source and the sample in the vacuum.” Phase matching for uv generation can be achieved by using the anomalous dispersion of the refractive index as in the case of infrared generation, It can also be achieved by mixing of foreign gas into the vapor. In the actual experiment of Hodgson et al. on Sr, only two pump beams were used with @, = @,, and 2w, was tuned to the resonant transition from (5s)? to an even-parity excited state. By varying «, to have the resonant enhancement of various autoionization states successively coming into play, the uv output was observed to have a tuning range from 1578 to 1957 A. By using Mg, Hg, and Zn, the tunable uv generation can be extended to the 1060 A region.'® With optimal focusing of ~ 1 MW pump beams into a metal vapor of ~ 10 torr, conversion efficiency of ~ 1% is possible but is limited by the usual detrimen- tal effects—self-focusing, saturation, ionization, and so on. In principle, one can also use condensed matter as the nonlinear medium for efficient infrared and ultraviolet generation by four-wave mixing. In practice, however, a condensed matter is often strongly absorbing in the uv below 2000 A, and has rather broad absorption bands in the visible and infrared. Therefore, the output cannot be in the vacuum uv or in any of the absorption bands. If ali the four frequencies should stay out of the absorption region, then the resonant enhancement may not be strong enough to make x”) larger than the multiply resonant x of a metal vapor at ~ 10 torr. To conclude this section, consider an interesting application of four-wave mixing to time-resolved infrared spectroscopy.!® Figure 14.11 is the schematic of the experimental arrangement. A pulsed broadband infrared beam (oF) is first generated by a stimulated electronic Raman process in a metal vapor cell using a pulsed broadband dye laser. This infrared beam, in passing through a sample, carries the spectral information of the sample absorption. Tt then interacts with a narrow-band laser beam (w,) in a second metal vapor cell. The latter generates a narrow-band Stokes beam (w,) in the cell, and through four-wave mixing (®, = «) — 0, + o%,), up-converts wi, to a broadband visible output w®,, which can be recorded on a spectrograph. With this technique, the infrared absorption spectrum of the sample is displayed as a corresponding visible spectrum, and hence the detection sensitivity is greatly improved, Moreover, since nanosecond or picosecond laser pulses can be used, the technique has a time-resolving capability in the nanosecond or picosecond‘Transient Four-Wave Mixing 259 (oop, «5, off) tan Metal oh Spectrograph emf Sampie Metal vapor [<—— oy Fig. 14.1. Schematic of an experimental arrangement using four-wave mixing for time-resolved infrared spectroscopy. Tegime. It may therefore find important applications in studies of chemical Teactions and radical spectra. 146 TRANSIENT FOUR-WAVE MIXING We have thus far considered four-wave mixing only in the steady state. With pulsed resonant excitation, however, transient effect in four-wave mixing may become important. As in the steady-state case, transient four-wave mixing is also governed by the third-order polarization P®. The only difference is that P©) is now a time-varying function of excitation and relaxation of the medium. Here we discuss the derivation of P®), leaving the actual solution of the wave equation with P©) as the driving source to Chapter 21. Using the density matrix formalism of Section 2.1, we can write P? = Tr(- Nerp”). (14.15) Then, to find P®, it is necessary only to find the density matrix p®. We adopt here again the diagrammatic technique of Yee and Gustafson” for the deriva- tion of p®, The notation here follows Section 2.3. Consider the general case where three successive fields, &,(u,, fexp[ik, «1 — «8)], & (co, dexpli(ky-r — o,f), and &,(w,, Hexpli(k, +r — @3f)] interact with a material system at 1, 4, and ¢, with 4 < ¢, < 4,, In this given time order, p(w = w, + w) + w3,£) has eight terms derived from the eight diagrams in Fig. 14.12. In comparison, p(w) for the steady-state case has 48 terms from 48 diagrams as mentioned in Section 2.3. The rules for deriving, an expression from a diagram are the same as those given in Section 2.3 expect that the propagation from one vertex at ¢, to the next at /, is now represented by the phase factor KCablA(4, ~ aby) = eat tonne (14.16)wg Ws In wy ren [10> Dom pag <1 I> Ppa <8 | (a (2) fo } [] [ Pegg <7 | 20> Pum <7 | | 28> Beene < Ht | () (#) yp tp>) , wy Povo < Ht | b> perm <1 @ (h) Fig. 14.12 The eight Feymann diagrams representing the cight terms of pP(z). 260‘Transient Four-Wave Mixing m1 where |a) is the ket state on the left between #, and ¢, and (d| is the bra state on the right, and y,, is the damping constant. The final expression has integration over all possible time separations between vertices and summation over all possible initial, intermediate, and final states. Two terms of p°'(1) corresponding to the diagrams with Figs. 14.12¢ and 14.125 are used here as examples.2* From Fig. 14.12a, we have, with the interaction Hamiltonian given by #°’(w) = —P° E(w 9, (Pole - EL (Ps Mots Pad Ke tint Rps)t5Ciepr Ope Ae pat Spm (14.17} x¢pprei(t- a-ha ny) ebrr eat Nga) xdmp- &(t = — qetbr enter) x (rip: é,(t _ netrs-™)(Ip hale} With the substitution of variables § = ¢— 1 — 2 — £,=1t- 4 — 7, and £,= 1-5, this equation becomes [rhe = E feenrenl se) Undead mers Pes XC pip + 2a) Canip + @yir Crip = 235) (14.18) xf" dk ellie Hyd tieer Bro) F2lh (a Jel aE xf* Agent tr Alean tom) *erlo i (B, Jethe™ -« x f abyelierettndeitg (ge) -«262 Four-Wave Mixing Similarly, we find, from Fig, 14.125, [Mlw=- £ [een eur 5) (ippon mas XC pip: Galerie + eile) Cmlp - 23) xf" dg ge Kitss* Myst ipnt Pom) wsks (£5 Jert>"* -2 f° i emt Gen) — 1: ik, x Ag seU pm Pym) emt md Serta (£, petha'® ~% x f* dE,gliomrtaltalbg, (Je). (14.19) -@ The full expression of p°'(z) is the sum of all eight terms derived from the eight diagrams in Fig. 14.2, In an actual case of transient four-wave mixing, however, resonant or near-resonant excitations are involved, and terms in p? which are nonresonant can often be neglected. The effective number of terms of p then is greatly reduced. As in the steady-state case, transient four-wave mixing yields a coherent output, It belongs to the category of coherent transient optical effects, which will be discussed in more detail in Chapter 21. Here we consider only some general characteristics of transient four-wave mixing.” First, as is explicitly shown in (14.18) and (14.19), the nonlinear polarization PO « p® bas a wavevector k, =k, + k, + ky. For the mixing process to be efficient, the phase-matching condition should be satisfied, that is, k(w) = [a eC) /elk = k,, This is the same as in the steady-state case, as one would expect. Second, the transient behavior of p“(#) arises from resonant or near-resonant excita- tion of the medium and is governed by the time-dependent phase factors in p®. For illustration, we discuss the case of a molecular gas excited by three resonant pulses. Tn a gas, molecules with different velocities should interact with different fields at different times. Let s(#) be the position of a molecule at time # moving with a velocity v. The field scen by the molecule at an earlier time &, is E,CE,Jexplik, * t(E,) — i,£,] at e(&,). Since r(Z,) = 1(1) — Ct — &,)v, We have &,(E,Jexp [7k «r(£,)] = &(dexp[ik,-r(1) ~ eky-w(e - &)]. (14.20) As a result, p(s) for the molecules is a function of ¥. For example, from‘Transient Four-Wave Mixing ws (14.18) we find 3 [Aco Z fetewrn(h) Cnrbacodeh mrBs X (pips har) Carp *2ale (resis) xf’ gst eon Soh (E)) (14.21) x f* df elem 9" Ope epat hala (Ep) ~2 x f* db, elileon9)* Goat tI CE.) b a The overall p® for the gas system is then given by an average of pty, 2) over the velocity distribution n(v): (1) = f~ nla. dv. (14.22) Equation (14.21) shows explicitly that p has a wavevectork, = k, + k, + k3. The. transient behavior of p(y, 2) as governed by the time-dependent phase factor is now also a function of v, If we assume that the exciting pulses are resonant with @, = @,,,, @2 = Gyr and w, = @,,, and the pulses are short compared to the relaxation time |$, At and the velocity dephasing time Ik,*¥)77, then (14.21) reduces to 1 3 ie, a (06, Dlen= (2) Lovato x Ew iN Thal baad a(t foo) +t 6109) er Ppl! £0) — pel En Fam) — Pp mC $20 E10) (14,23) XC pip + eal) Crap air) (rip 2515) * és a 0 fi abe) J bata) [? 486 6) Or -« 0 ~e where £;, is the time when the center of the /th pulse arrives at r(7). It is seen that the only phase factor in pv, 1) depending on v is IRD a FETs F a) aE) th Eo | (14.24) If @(v) = 0 for all v, then molecules with different y will radiate in phase, and the transient four-wave mixing output will be a maximum. This happens at64 Four-Ware Mixing 1 = t, provided 1, > So, where &, = ky -[hidio + ka8o0 + Kako Akal”. (14.25) This result shows that the coherent output from transient four-wave mixing is expected to last only for a short duration during which 6(+) is small for all ¥. This is generally true for all coherent transient effects. The other factors in (14.23) determine the intensity of the four-wave mixing output, with EXPL bya? ~ E30) — Spel Ego — €30) — Sym ($20 ~ €10)] describing the decay of molecule excitation due to random perturbation, and hence the decay of the coherent output signal, We do not dwetl on the details of transient four-wave mixing here but postpone the discussion to Chapter 21 in conjunction with other coherent transient optical effects. REFERENCES 1 B.D. Maker and R. W. Terhune, Phys, Rev. A 137, 808 (1965). 2 See, for example, J. J. Wynne and P. P, Sorokin, in Y. R. Shen, ed., Nonlinear Infrared Generation (Springer-Verlag, Berlin, 1977), p. 159; D.C. Hanna, M.A. Yuratich, and D. Cotter, Nonlinear Optics of Free Atoms and Molecules (Springer-Verlag, Berlin, 1979). 3. BL. Stepanov, E. V. Ivakin, and A. S, Rubanov, Sov. Phys. Dokiady 16, 48 (1971); RW. Hellwarth, J. Opt. Soc. Am. 67, 1 4977). 4. RW. Hellwarth, Prog. Quant, Etectron, 8, 1 (1977). 5S N. Bloembergen, H. Lotem, and R. T. Lynch, fnd. J. Pure Appl. Phys. 16, 151 (1978). 6 JeL, Qudar and Y. R. Shen, Phys. Rev. A 22, 1141 (1981). 7 A. Yariv and D. M, Pepper, Opt. Lest. 1, 16 (1977). & A. Yariv, [EEE J. Quant. Electron. QH-14, 650 (1978). 9 CR Giuliano, Phys. Today 34, No. 4, 27 (1980), and references therein 10 B. Yn. Zel'dovich, V. L Popovicher, V. V. Ragul'skii, and F. 8. Faizullov, Sov. Phys. JETP 15, 109 (1972) 11 R.W. Hellwarth, J. Ops. Soc. Am. 68, 1050 (2978). 12 1.3, Wynne, P. P. Sorokin, and J. R, Lankard, in R. G. Brewer and A. Mooradian, eds., Laser Speeteascopy (Plenum, New York, 1974), p. 103. 13. P. P. Sorokin, J, J. Wynne, and J, R. Lankard, Appl. Phys. Lett. 22, 342 (1973). 14 DM. Bloom, J. R, Yardley, J. F. Young, and 8. E. Harris, Appl. Phys. Lett. 24, 427 (1974). 15. R.T, Hodgson, P. P. Sorokin, and J. J. Wynne, Phys. Rev. Lett. 32, 343 (1974). 16 K.K. Innes, B. P. Stoichetf, and 8. C. Wallace, Appl. Phys. Leit. 29, 715 (1976); F. Vallee, S.C. Wallace, and 3, Lugasik, Opt. Commun. 42, 148 (£982); F. Vallee and J. Lugasik, Opt. Commun, 43, 287 (1982). 17 AHL Kung, Opt. Lett. 8,24 (1983); Technical Digest, Conference on Lasers and Electro-Optics, Baltimore, (1983), paper Tu 13. 18 TJ. McKee, $. C. Wallace, and B. P. Stoicheff, Opt. Lett. 3, 207 (1978); F, S. Tomkins and R- Mahon, Ops. Lett 6, 179 (1981); J. Bokor, R.R. Freeman, R. L, Panock, and J. C. White, Opt. Latt, 6, 182 (1981), W. Jamroz, P. E, LaRoeque, and B. P. Stoichel, Opt, Lett. 7, 617 (1982). 19D, §. Bethune, J. R. Lankard, and P,P, Sorokin, Opt. Lett, 4, 103 (1979), Ph. Avouris, D. &Bibliography 265 Bethune, J. R, Lankard, J. Ors, and P. P. Sorokin, J. Ghent, Phys. 74, 2304 (1981). W® T.K. Yee and T. K. Gustafson, Pays. Rev. A 18, 1597 (1978)- 2 P. Ye and ¥. R. Shen, Phys. Rev. A 25, 2183 (1982). BIBLIOGRAPHY Fisher, R. A.,ed., Optical Phase Conjugation, (Academic Press, New York, 1983). Hanna, D.C, M.A. Yuratich, and D. Cotter, Nonlinear Optics of Free Atoms and Molecules (Springer-Verlag, Berlin, 1981). Hellwarth, R. W., Prog, Quant. Electron. 5, 1 (1977). Proceedings of the interpational Werkshop oa Optical Phase Conjugation and Instabiliies, C- Flytzanis et al,, eds. (Cargese, France, 1982).I5 Four-Wave Mixing Spectroscopy Four-wave mixing with resonant excitations can be a versatile spectroscopic technique. * It has already found important applications in many areas, such as analytical chemistry, combustion, and material studies. Its advantages over other techniques are in the capabilities for high resolution, for elimination of strong fluorescence background, and for time-resolving measurements of ultra- fast dynamic properties. In this chapter we discuss four-wave mixing spectros- copy in its various forms. Note that most of the high-resolution nonlinear spectroscopic techniques discussed in Chapter 13 can be regarded as four-wave mixing spectroscopy when the nonlinearity is limited to the third order. 15.1 GENERAL DESCRIPTION As shown in Chapter 14, the signal in a four-wave mixing process is directly related to the third-order nonlinear susceptibility ©, which exhibits reso- nances characteristic of the nonlinear medium. Therefore, from x as a function of pump frequencies, one can obtain spectroscopic information about the medium. The expressions of xO(w = a, + w, + w;) at the beginning of Section 14.1 show that it can have single, double, or triple resonances. Singly resonant four-wave mixing has the merit of a relatively simple experimental arrange- ment since only one pump frequency needs to be tuned. The corresponding theoretical analysis is also straightforward. The process can yield information one normally obtains from ordinary one-photon or two-photon transition measurements, Doubly and triply resonant four-wave mixing may require more than one tunable pump frequency and hence a more complicated experimental set-up. It can, however, give more selective spectroscopic information about the medium. Doppler-free spectra, for instance, can be obtained with double or 166‘Coherent Raman Scattering Spectroscopy 267 triple resonances. The high-resolution nontinear spectroscopic techniques dis- cussed in Chapter 13 are good examples. The two general types of four-wave mixing processes discussed in Section 14.2 can lead to two different kinds of four-wave mixing spectroscopic tech- niques. In the first case, the output is in a mode different from the input modes. The signal is proportional to |x|? and can be selectively detected through Gitering against the transmitted pump beams. The technique is most commonly used in singly resonant cases where the analysis is relatively simple. In the second case, the output is in the same mode as one of the input waves. The signal appears in the form of gain or loss of that particular input wave and is proportional to Im(x) (see Section 14.2, “Output Field in the Same Mode as One of the Input Fields”). This then allows a simple interpretation of the observed spectrum. In the following sections, we discuss the few well-known four-wave mixing spectroscopic methods in detail. While in most cases we consider #, — #, near resonance (the Raman case), keep in mind that the same description applies to cases with w, + @, near resonance. 15.2. COHERENT RAMAN SCATTERING SPECTROSCOPY Coherent Anti-Stokes Raman Scattering (CARS) As shown schematically in Fig. 15.1, CARS refers to the Raman scattering process in which the material excitational wave at w, — #, = w, is first coherently excited by the beating of two incoming waves at w, and w, and then mixed with the wave at , to yield a coherent output at the anti-Stokes frequency w, = 2w, — w,. Itis therefore a coherent version of the spontaneous Raman scattering process. Because of its numerous important applications, CARS is perhaps the best known four-wave mixing spectroscopic method. A large number of review articles already exist on the subject.* The theory of CARS is simple, as it follows the general theory of optical mixing discussed earlier in Chapters 6 and 14. With the input pump beams at @, and , the output field is governed by the wave equation (assuming an isotropic medium) 2 2 [v + Se(a)fe,- — 2% por(a,) (15.1) c c where P(a,) = xO, = @, + wy — W2): Ey()E;(w, JES(w), The solu- tion of (15.1) with negligible pump depletion and damping gives an output intensity /, from a slab medium of length /: 2a0 sim Ak? Fe XSPIE EP ‘ (244,!) = SiB,P = 1, = 5qlEol (15.2)268 Four-Wave Mixing Spectroscopy wr wy wa @y Fig. 15.1 Schematic diagrams describing the CARS process. where xD = 2 xO: bere Ak, = Ak-i, Ak =k, — (2k, — k,). We note that the anti-Stokes output is proportional to |x@\?. If the frequency scan is limited to a sufficiently narrow range, the variation of J, versus the pump frequencies is entirely dominated by the resonant dispersion of x@. As discussed in Section 14.1, x® can be decomposed into a resonant part x and a nonresonant part x). Here, x? is singly resonant at @, — @, = w, and can be written in the form xP= lola a) rE (15.3) @, — a, w,) +10 where, from (14.2), a = ~N( Me) (Meg 0, - py fh is essentially indepen- dent of w, and #, a8 @, — w, scans over the Raman resonance. Therefore, the anti-Stokes output has a spectrum given by ? ap +-—_4 r Zz . (a, — a, ~ w,) +T? (15.4) Figure 15.2 is an example. Because of the presence of the nonresonant part xh in x®, the spectrum appears asymmetric with respect to the resonant point w, ~ #, = ,, and has a peak and a dip at 1 2 1/2" (o,-a),-094{- 445) +ar| (15.5)Coherent Raman Scattering Spectroscopy 1 1 T T T T T T T 1 4 xo 4 ] > fab 4 z a # Zz 2 a yy Z iL 4 & S 2 25MM CALCITE - 4 1o7- 4 a) ° v 20 30 40 3G (W,-W,)- Wa ICM enc Fig. 15.2. A CARS spectral line around the 1088-cm~! resonance in calcite. (After M. D. Levenson, JEEE J. Quant. Electron. QE-10, 110 (1974}.] We find a (4 — 1). + (a) - @)_= 2o, Ta NR (15.6) z a¥ 2 [Coy - @)4-(@, - @)_]P= ~,} +4P’. xk Measurements of (w, — w,) , together with the output intensities at (w@, — 0),270 Four-Wave Mixing Spectroscopy can in principle determine all the four quantities w,, I, ¢, and xf. in practice, however, the accuracy of intensity measurements is usually poor but that of frequency measurements can be very good. Then, we note from (15.6) that if, or T is known from other measurements, 2/7 Pe can be obtained simply from the frequency measurement of (w, — #,),+(@, — @.)_. If a is also known from spontaneous Raman scattering, then x8 can be deduced very accurately? This contrasts with the inaccurate determination of x€h by the intensity measurements, Although (w, — ©,), alone cannot determine the resonant frequency «,, they do give an approximate value for «,, and therefore can be used to characterize the resonant medium, In many cases the Raman resonance is strong, while the nonresonant background x}, is weak, so that Ja/xQh| > I. This often occurs, for example, in pure gas media, where, one finds (, ~ #3) ,= @, and (@) — @)_= e, — a/xRh- The resonant frequency w, can be directly obtained from the peak position of the CARS spectrum. In other cases where |a/x&kI $I, both the peak and the dip of the spectrum appear around «, within a range of the order of I. However, if IxSk| is much larger than a/T, then the peak and dip will not be prominent in the spectrum. This is clearly a drawback of CARS as a spectroscopic technique because weak Raman reso- nances are not easily observable. The sensitivity of CARS can be greatly improved if the nonresonant background due to x}, can be suppressed. As we shall see later, this can be done by selectively detecting the proper polarization component of the output. Maker and Terhune’ first demonstrated the CARS process. Later, Wynne,’ Bloembergen and his associates,* and Akhmanov et al,’ used the method for spectroscopic studies in a large number of liquids and solids. A typical experimental arrangement is presented in Fig. 15.3. Dye lasers pumped by either a nitrogen laser or a frequency-doubled Nd : YAG laser are often used as tunable pump sources. The signal strength cam be estimated from (15.2). Assume |x) aa, = [a/T| > IxQal and perfect phase matching Ak = 0. Then, if input beams of 10 kW peak power at ~ 5000 A are focused to a spot of 1075 em? into a medium 1 mm long with [x@\nax = 107"? esu, we find that the output can have a peak power of 5 W. If the medium is 2 cm long with [x2)lmmax = 107" esu, the output can still have a peak power of 2 mW, which is readily detectable. The last example actually corresponds to the case of CARS in a gas medium at nearly the atmospheric pressure. In fact, while initial CARS work was on condensed matter, the more recent applications of CARS to material studies have been on gases, Even CARS from molecules in molecular beams has been observed.® Because the spectral resolution of CARS is basically limited by the laser linewidths, it can be used to obtain high-resolution Raman spectra of gases.” Conventional Raman scattering, in comparison, suffers not only poor spectral resolution limited by the spectrometer but also low signal intensity. The high sensitivity of CARS allows the method to be used for detecting trace molecules in gases, In particular, it has found applications in combustion studies as aCoherent Raman Scattering Spectroscopy 21 DUAL BOXCAR BANPUT A INPUT REFERENCE SAMPLE CHART RECORDER MONOCHROMATOR LT aw Ne LASER OCUBLE MONOCHROMATOR «POLARIZER TA FILTER _ ay | WEL COMPOSITE \0--¥ coe nON SLIT Fig. 15.3 A CARS spectroscopy system. Laser beam 1 (---) and laser beam 2 {--- ) are blocked after the samples, and the output duc to frequency mixing is collected into the double monochromators. The signal from the reference arm and the ratioing system are used for normalization in order to reduce the effect from laser fluctuations. {After M. D. Levenson, JEEE J. Quant. Electron. QE-10, 110 (1974).] means to monitor the temporal and spatial distributions of various species and to find the internal energy distribution of the species.'° Because the output is coherent and highly directional, temporal, spatial, and spectral filtering can be used to suppress the strong luminescence background from the combustion process. In fact, its ability to discriminate against luminescence takes coherent Raman spectroscopy in general a unique technique for combustion studies. Similarly, CARS can be used to probe the reaction products and their internal energy distribution from a chemical reaction, and to obtain Raman spectra of fluorescent materials. That CARS is capable of probing materials in an enclosed, hostile environment with good space and time resolution has prompted the suggestion of using CARS to study high-temperature plasmas! and to monitor target implosion in laser fusion.’* Although the foregoing discussion is focused on Raman resonances, the theoretical treatment applies equally well to two-photon resonances withm Four-Wave Mixing Spectroscopy Gy + @y = Wy, OF 201 = @o, OF 20, = wo. The coherent output is at either Mig £ Wy OT Wy $ 0. Polarization CARS It is easy to see from (15.4) that if Ix@lmex = l¢/Tl xGh- the resonant structure sitting on the nonresonant background in the CARS spectrum will be easily lost in the presence of background fluctuations. With an appropriate polarization arrangement in the detection scheme, however, the nonresonant background can be largely suppressed, and hence the sensitivity of CARS is greatly improved.">:1* The basic principle of the polarization CARS is described below. The solution of (15.1) shows that the output field E, is proportional to the nonlinear polarization P©(«,), which can be written as Po = PG + PR (15.7) with PQ = x, :E,E,Es and PQ = x: EE, £4. If PS) is in the direction 4,, then since PO and PG} are generally not in the same direction, the output field has the form [2,4 (8, « xBk: Ey ELES) + 2,A*(@, - xf E,EE3) B +2,A"(2,-xf): ELELES)| (15.8) where AN® and A® axe coefficients, and @, is orthonormal to é,. With an analyzer in front of the detector to block out the é, component, the output signal is then given by 1, © |A%, +x EE, Es}? (15,9) which is proportional to only |2,* x : 4,2,6,|". The nonresonant background can in principle be completely suppressed. In practice, because the analyzer is not perfect, a residual nonresonant background still shows up but is certainly greatly reduced from the original magnitude. For spectroscopic studies, one is interested in the dispersion of x and not in the absolute value of x@. Then, with weak resonances such that [é,-xQ0: @21é0l © 1e,.-xRk 212 122h it is most convenient to record the ratio |R[* of the é, and é, components of the output, which is given by AB(,-7: 2222) | IRP =| enip ay Qh ra ee ABR( 8 oy: 8,282) (15.10) Note that [R[’ is independent of the pump intensities, and therefore is littleCoherent Raman Scattering Spectroscopy 273 affected by laser intensity fluctuations. Since the dispersion of x is negligible in the resonant region, |R|* versus a, — «, essentially reproduces the spectrum of Ix1?. A possible experimental arrangement for this polarization CARS is sche- matically shown in Fig. 15.4. The é, and @, components, corresponding to the transmitted and rejected components from the analyzer, are simultaneously recorded by the two photodetectors, and their ratio is taken by the divider. The polarization CARS spectrum of benzene in CC], is shown in Fig. 15.5 as an example.!* The same spectral line would hardly be visible in ordinary CARS because of the enormous nonresonant background. In some cases, one would like to obtain the spectra of Re(x{?) and Im(x#) separately. This can be achieved with a simple modification of the polarization arrangement in Fig. 15.4. If we rotate the analyzer by a small angle @ from the null point, then & is replaced by R, = tané + R, and the ratio of the two output components by JR, |? = [tan @ + RY’. (15.11) Let us choose tan?@ > |Ri2, Then the output ratio can be approximated, with R= R’ + ik’, by IR, |? = tan?@ + 2R’ tan 6. (15.12) Since x@ is generally real, R’ is directly proportional to Re(é,-xQ): é)8,2)). The spectrum of [R,|? therefore yields the resonant spectrum of Refé,+ x@: é,4,8,) on top of a constant background. The relative strength of the resonant structure versus the background in this case is 2R’/tan 6, which is certainly much larger than that in ordinary CARS. In actual experiments, it ts more convenient to use a polarization rotator in front of the analyzer than to rotate the analyzer. This is to avoid problems that may arise because of sensitivity of the photodetection system to the variation of beam polarization."* Tf, in addition, a quarter wave plate is also inserted before the polarization rotator to change the relative phase of the é, component by +90°, then the /4 Plate rm Analyzer wy) we Mono- ro “| chramator po2 Sample Fig. 15.4 An experimental arrangement for polarization CARS. (After Ref. 15.)24 Four- Wave Mixing Spectroscopy tb) Imt Xp} Fig. 15.5 Polarization CARS spectra of 0.011 M benzene in carbon tetrachloride: (a) with suppression of the nonresonant background, (5) with Im x superimposed on a weak nonteso- nant background, and (c) with Re x{? superim- 3a5 990995. 1000 «posed on a weak nonresonant background. Stokes shift em) (After Ref. 14) ratio of the two output components becomes IR.) = |tan@ + IR}? (15.13) = tan’?@ F 2R”’tan 6, which yields the resonant spectrum of Im(é,- x? : ¢,2,é2) on top of a constant background. An example of |R,|? and |R,|? versus #, — a, is shown in Fig. 15.5 for 0.01 M benzene dissolved in CC1,.! The small value of [xQ/x(h| due to the small amount of benzene in this case would make the detection of the Raman resonance of benzene very difficult with ordinary CARS. The technique of measuring R’ and R” described here is essentially the same as the well-known heterodyne technique. The uncrossed coherent background signal here plays the role of the local oscillator. We realize that even in detecting |R|?, the resonant spectrum is not completely free of background because of the finite extinction coefficient of the analyzer. Fluctuations of the background still limit the sensitivity of the polarization CARS. Even though the background in the measurements of R’ and R” is higher than that in the measurement of |R|?, a signal-to-noise ratio analysis shows that the former can be orders of magnitude more sensitive.'* This is because heterodyning effectively increases the signal-to-background ratio. The greatly improved sensitivity of the polarization CARS makes itRaman-Induced Kerr Effect Spectroscopy (Rikes) 215 particularly useful for probing trace molecules, as in combustion studies."* Other Coherent Raman Spectroscopic Techniques Asa variation of CARS, one can detect, with the same input pump beams, the coherent output at the frequency 2@, — #1. This is known as coherent Stokes Raman scattering (CSRS). The theory of CSRS is the same as CARS, except that the values of x), and @ are somewhat different because of dispersion. The polarization arrangement can also be used to improve the sensitivity of CSRS. Raman gain and inverse Raman spectroscopic techniques have already been discussed in Section 10.6. They can be considered four-wave mixing processes with the output in the same mode as one of the inputs. The gain or loss, which is directly proportional to Imx{, shows a resonant spectrum without the nonresonant background. The techniques are particularly suitable for obtain- ing high-resolution Raman spectra of gases. With sufficiently strong imput laser intensities, the Raman transition can even be saturated. Saturation Raman spectroscopy, similar in principle to saturation spectroscopy discussed in Section 13,3, can be used to obtain sub-Doppler Raman spectra.’” Although the discussion in this section has been focused on Raman reso- nances, the theoretical treatment applies equally well to general two-photon resonances with w, + @ = Wo, OF 24) = Wy, OF 2, = Wp, where wy is the resonant frequency. The coherent output can be detected at cither |, + o| or jw, + @p|, Interference may show up if the output exhibits Raman and two- photon resonances al the same time."* 153 RAMAN-INDUCED KERR EFFECT SPECTROSCOPY (RIKES) A gain or Joss is always accompanied by a corresponding resonant birefrin- gence. Thus for an induced Raman gain or loss, there should be a correspond- ing induced birefringence seen by the probe beam. This is known as the Raman-induced Kerr effect,!° where the Kerr effect is used indiscriminantly to denote the field-induced birefringence phenomenon, (The optical Kerr effect is discussed in more detail in Chapter 16.) We consider here the same experimental geometry as in the Raman gain spectroscopy with a pump beam at w, and a probe beam at w,, both propagating along 2. The frequency w,— , is scanned over the Raman resonance at w,. For simplicity, let the probe beam be linearly polarized along 4. In propagating through the medium, the polarization of the probe beam becomes elliptical as a result of the Raman-induced birefringence. When: the birefringence is small, the polarization change can be treated as creation of a new polarization component at w, along j through the four-wave mixing process as follows. The incoming fields induce a third-order polarization with a216 Four-Wave Mixing Spectroscopy § component PEC) = H+ XO (oe, = cy — wy + Oy) OA RELEFE,. (15-14) In a medium of length /, this nonlinear polarization component generates a field component along > idm? Ze; E,(,) = Ok 2 P(e). (15.15) Measurement of the output f,(w,) © 1B, (a)? versus #, — # should there- fore yield the spectrum of |? - x: €,¢,4(7, which exhibits Raman resonance. In many media, if the collinearly propagating beams are along some symmetry direction, the subindices of x{}}, must appear in like pairs from symmetry consideration. We then have PPC wn) = XSyx BE Fax + XSpax Ey Et Ban (15.16) With the pump beam linearly polarized in the direction bisecting % and f, so that |£,,| = {£,,| we find (62) © Qn + XB. ya t xPecl Ere }Eal- (15.17) The phase-matching condition is automatically satisfied in this case. Since x) = yh + x8), the observed spectrum has the resonant structure superim- posed on the nonresonant background, similar to that obtained by CARS. If the pump beam is circularly polarized, then E,, = +i£,,, and we have Ty(r) & Rye — Xara? (15.18) Tn an isotropic medium, the nonresonant part has the symmetry retation ORD rar = SRI pyre but OR yy * (XB Iyer and we find Fy (00) © (XP) yaya — (XP) ypeal MEI*LEDP (15.19) The observed spectrum therefore shows no nonresonant background and is similar to that obtained in polarization CARS (Fig. 15.54). In experiments, 1,(10,)/F,(@,) should be recorded to minimize the effect of probe beam fluctuations. This can be done by simultaneously recording the transmitted and rejected components of £(«) from an analyzer in front of the detector. One can also use the heterodyne scheme to obiain Rex and Imx{? separately.” Again, by rotating the analyzer a small angle @ away from §, the ratio of the transmitted and rejected components becomes Ry? = fam + yf (xB) ere — OR Dye bP? (15.20)Multiply Resonant Four-Wave Mixing 2 where y is a real coefficient proportional to {E,|?. If tan’@ > WOR) ree > OR ),yx0}1?, then we have [Rift = tan?@ + 2y(tan @)Re{ (x) yoye — (XP Joveef> — 5.24) which yields a spectrum of Rex. On the other hand, with a quarter-wave plate inserted in the path, the ratio of the transmitted and rejected components becomes Ral? = tan + iy{( xP) are — OX) yee}? (15.22) = tan? + 2y(tandm{ (x2) yeye — (XP) pyar } which yields a spectrum of Imy{?. Again, while the discussion here is on RIKES, it can be applied equally well to other two-photon resonances with w, + w, = w), where wy is the resonant frequency. 15.4 MULTIPLY RESONANT FOUR-WAVE MIXING The resonant nonlinear susceptibility x can be further enhanced with double and triple resonances. It should give more selective spectroscopic information than in the singly resonant case. In general, multiply resonant four-wave mixing can provide more specific details about a particular one-photon reso- nant transition. A few interesting applications of such resonant processes are described here, High-Resolution Doppler-Free Spectroscopy Resonant four-wave mixing with appropriate input beams can yield Doppler- free spectra. A few examples were discussed in Chapter 13. The Doppler-free saturation and polarization spectroscopy with the lowest-order saturation is a triply resonant four-wave mixing process in which the pump field E() creates a population change 2 proportional to E(w)}E*(«) and then the probe field E(@") detects the change through the nonlinear polarization PO(«’) & E(w) E*(w) E(w’) with w and w either on the same or different resonant transitions. The two-photon Doppler-free absorption spectroscopy, on the other hand, is based on a singly resonant four-wave mixing process. In general, multiply resonant four-wave mixing can yield a Doppler-free spectrum if the damping coefficients of the resonant denominators of x? satisfy a certain condition. This is seen from the following mathematical derivation.”"28 Four-Wave Mining Spectroscopy Consider first a doubly resonant case, The resonant nonlinear susceptibility x@ has the general form (see Section 2.2) xQ = 7 (15.23) — 4 w, — @, + IT my — Os £ iT) where A is a coefficient with negligible frequency dependence in the present discussion, In a gas medium, x should be an average over the Doppler profile (see Section 13.1), while o,, and w,, should be replaced by wf,(1 — vk, /e) and w?,(1 — v-k,/c), respectively: xk) -{* de Ag(o) -o {w,— of {1 - v-k,/c) + Ty][e, — e841 - vk, /e) +i] (45.24) where g(v) = (1/ Yr ujexp(—v?/u?) and u = 2k7/m. Equation (15.24) can be written in the form oy =f? gy Rom) _ (x8) = [de EEG) (18.25) with» = 0/4, £, = [c/(8- kul — o2,/ 09, F iT,,/af;), and a similar expres- sion for &,. Then the integral can be evaluated in terms of the plasma dispersion function zener tier)= ref” ae” forg™> 0s z(§*) = -Z*(-€) and (x@) behaves differently near the double resonance £/ = £4, depending on the relative signs of £” and &. If £7 and &j are of the same sign, we have Z(G) — Zé oxy = tee) = 260, san 27 §b which shows no singularity as €; > £j, and as a function of £5 (or w,), has a resonant width more than twice the Doppler width. If, however, £% and £7 have opposite signs, we find, as > & OBR) & [f + ae (15.28)Multipty Resonant Four-Wave Mixing 219 where + and — refer to the cases of ” > 0 and £7 < 0, respectively. The last term in (15.28) has a Lorentzian resonant lineshape with a halfwidth 7 + ty = [e/(6- kyu, /af, + Ty/el,). In other words, as 0,/u!, > w,/w), at double resonance, (x?) versus w,/«), has a Lorentzian linewidth given by the sum of the normalized natural linewidths, (Tf, Joe, + Ty,/w},). Therefore, measurement of (x), or ((x@ 1? in this case, can yield a Doppler-free spectrum. For the various doubly resonance cases shown in Fig. 14.2, spectra obtained with 14.2d, g, and & are Doppler-free.'5 Note that Fig. 14.2¢ describes a CSRS process, while d and & are CARS and CSRS, respectively, probing a resonant transition between excited states. A similar result is obtained for the triply resonant case. We can write a exp”) GP {Ee eiecey 1579) At triple resonance, £; > 5 = $7, if one of the g” has a different sign from the E two others, then (15.29) gives a Doppler-free spectrum. In general, (x) versus «2, shows two Lorentzian peaks. They merge into one when £7 — £7. All the triply resonant processes in Fig. 14.3 should yield Doppler-free spectra. The discussion here can of course be extended to the reduction of inhomo- geneous broadening in solids.\> However, because of the many crystal-feld ; parameters governing the broadening (see Section 13.1), the elimination of crystal-field broadening is never perfect. Measurement of Longitudinal Relaxation Times ~ If the relaxation of an excess population in a state can be approximated by a Pan ~ Pn [Flew - | --! Fe ) (15.30) the resonant four-wave mixing spectroscopy can also be used to deduce the ; longitudinal relaxation time T,,.1° The physical idea is fairly simple. Two imput beams at @, and w/ near a resonance can beat in the medium and induce a time-varying population change oscillating at the Frequency @, — «{ with an amplitude inversely proportionat to the zero-frequency resonant denominator [(o, — #} + i/F,,]. The induced population change in the particular state is then probed by a resonant transition from that state. The output spectrum of &x®) versus w, — 0} has a halfwidth 1/7,,, although the halfwidths of the one-photon transitions involved are much larger than 1/7,230 Four-Wave Mixing Spectroscopy Consider the process in Fig. 14.3@ described by (14.42) as an example. For a gas medium, the average x{’ with 7, >» [~'is (x) @ —— fay 8) _ Oe ti/T fw ty Oy, + Tyg x L a HY = Beg Tyg ol HRY Wg + Tag h (15.38) Since T,,, >> Fp}, Uy, the integral in (15.31) is practically independent of w} in the range of |a, — wi) ~ ¥,,!, and therefore we have the spectrum of (x{? YETSUS Wy — Wy aS x) & (ee + /Tyy) (15.32) Yajima et al.”? proposed and demonstrated a method for measurements of T; and T, using four-wave mixing with two laser beams. It was also mentioned in Section 13.3 that T, and 7, can be deduced simultaneously from the saturation spectroscopy. Unlike the case described above, however, the lack of a third adjustable input frequency in the latter cases for selective resonant probing makes the interpretation of the results Jess straight-forward. Coherent Raman Spectroscopy of Excited States The doubly resonant case of Fig. 14.2 described by (14.3d) with an output at 2, — , shows a resonance at #, — @, = @,,,. Thus the four-wave mixing process can be used to probe a Raman transition between the two excited states |”) and |n’).” We, notice however, that the bracket in (14.3¢) is tt 8, Org t ils Wy Wy — iT yy yn + ya) + (Trin — Tre = Tag) (a, — Or, + Deg)(@2 — Ong - Ty) (15.33) which nearly vanishes at w, — &, = @,, if w, and «, are fay away from resonances so that I, and Fj, are negligible. Yet, even with the double Tesonance w, ~ w,, and #, = 4,,, no Raman resonance can be observed if Tn 7 Vive ~ Tag = 0 because the resonant denominator (a, — w. — @,, + Tyg) in x? of (14.3d) is canceled by the numerator in (15.33). This actually happens when the damping is dominated by spontancous emission, sinceForced Light-Seattering Spectroscopy 281 Tyg + Tag = Frvn-”? In the presence of collisions, this relation no longer holds, and we usually have [,, = 0)? + ¥,,p, where T}¥ is due to spontaneous emis- sion, p is the gas pressure, and ¥,, is the coefficient describing the collisional broadening. The relevant frequency factor in x§ of (14.3¢) then becomes 1 1 1 (0) = Wy = Wyg + Ty) ( Wy — yey + Tyg 2 — Ong — T| 1 (0 — yg + Tyg)(@2 ~ ne — Trg) xlo14 Cran ~ Ywe ~ Yaa) P a — @,~ wy, + AES, + tone) (13.34) which clearly shows a Raman resonance between |) and |n’) with an amplitude proportional to p. This is called the pressure-induced extra reso- nance in four-wave mixing (PIER-4) by Bloembergen et al.” They observed the PIER-4 signal between the 3P,,. and 3P, . states of sodium vapor. The signal strength is proportional to i( Yq — Yarg — Yng) P beQP? & NA =| 0 ~~ Oy + (LF, + tuP) (15.35) With buffer gas, y,;7 should be replaced by yp + ¥/;P ouner- The linewidth of the resonance should increase linearly with the pressures. The experimental results agree with these predictions. It shows that PIER-4 can be used as a spectroscopic method not only to probe the Raman resonance between excited states, but also to study molecular collisions. The PIER-4 process is, of course, not restricted to gases. In condensed matter, the relation T,,., = Oat [., usually does not hold, so that the PIER-4 signal should be observable. Note that the case of Fig. 14.2d can yield a Dopplez-free spectrum. Among the other processes in Fig. 14.2, case A with the output at 20, — , also probes the Raman resonance between excited states. The theoretical discussion is essentially the same as case a. 15.55 FORCED LIGHT-SCATTERING SPECTROSCOPY As described in Section 10.3, a transverse excitation, denoted by the density matrix p,,, with nm # n’, can be treated as a material excitational wave governed by a characteristic wave equation. Thus CARS can be considered a process in which the material excitation wave at w, — ~ w,, is resonantly excited by282 Four-Wave Mixing Spectroscopy optical mixing of two incoming waves at w, and w,, and then the material excitational wave coherently scatters the incoming wave at w, and yields the anti-Stokes output. Theoretically, therefore, the process can be described by coupling of light waves at w,, ,, and 2, — @, with the material excitational wave at w, — w,. In the steady-state case, elimination of the material excita- tional wave in the coupled equations leads to the same result given in Section 15.2. We realize that the material excitational wave is not restricted to vibrational or electronic excitation. It can be any excitation including acoustic wave, entropy wave, spin wave, and charge density wave. Each, however, generally has its own governing wave equation. A CARS process involving low-frequency material excitation is often known as a forced light-scattering process.” It differs from the spontaneous anti-Stokes scattering process in the sense that the material excitation is not thermally excited but is coherently driven by beating of two optical waves. (In general, it can of course also be directly excited by an electromagnetic wave with w ~ «,, or by other means.) In the discussion of stimulated light scattering in Chapter 11, the wave equations for a number of low-frequency excitations were given. As an example, consider here briefly the case of forced light scattering of concentration variation. The driven equation of motion for the concentration variation is, following (11.26), (Zz - DV? + Flac = —-AV?(E, Et) (15.36) where A = D(d2/AC), 7/Spo(0n/0C), 7, and we have generalized the equation by including a term AC//z to allow relaxation of AC. With monochro- matic incoming plane waves, E, = ¢,exp(ik, +r — fw, ¢) and E, = éexp(ik2° ¥ — i@f), the solution of (15.36) gives A(ky - k)(é°88) Ac = (@, — a) + ift/r + D(k, - k,)'] eff —kad-rter~o2), (15,37) This driven concentration wave then coherently scatters the E, wave to yield an anti-Stokes output, which is governed by (15.1) with ~417(w2/c*) x PO(w,) replaced by — (2/7? 8e/2C), 7-(AC)E,, From a medium of length /, the anti-Stokes output in the phase-matched direction k, = 2k, — k, has an intensity 2 ‘2 = (32), ACV E: + Sax QRETED (15.38) — a a Bae where the x@) term is the nonresonant contribution to the output. The spectrum of J, versus w, — w, in this case is dominated by the zero-frequencyReferences 283 resonance of the concentration wave, The width of the resonant line has contributions from both relaxation and diffusion of the local excess concentra- tion, 15.6 TRANSIENT FOUR-WAVE MIXING SPECTROSCOPY With resonant excitation in four-wave mixing, the relaxation of the excitation naturally leads to a time-dependent output signal when pulsed lasers are used for pumping and probing. Thus transient four-wave mixing spectroscopy can be used to study the relaxation behavior of a resonant excitation. In fact, transient CARS was already discussed in Section 10.10. It was shown that the transverse relaxation time of the Raman resonance can be obtained from the transient measurements. Transient forced light scattering is quite similar. Let us consider again the forced concentration scattering as an example. Assume that the pump pulses are much shorter than the characteristic time for concentration variation, and that at ¢~ 0, they induce a concentration grating in the £ direction, AC(x, ¢ = 0) = AC,4@1 + cos Kx), where K = (k, ~ k,)-% and AC, « #,6,, After the pump pulses are over, the induced con- centration grating gradually decays away, following the transient solution of (15.36) AC(x, = Ace" [1 + PK cog Ks| (15.39) A short probe pulse with intensity J, scattered by AC at time ¢ then yields a coherently scattered signal with a wavevector k, = k, + (k, — k;): S(t) 0 Rie rt ey, (15.40) From the measured exponential decay of the signal with time and its depen- dence on the grating spacing (27/K), we can then deduce separately the . relaxation time + and the diffusion constant D, More rigorousty, the finite widths of the pump and probe pulses should also be taken into account in the | calculation by proper convolution. Transient four-wave mixing spectroscopy can be more general, involving possible multiple resonances. This more general discussion is postponed to Chapter 21, where transient four-wave mixing is shown to be related to other coherent transient effects such as photon echoes and free-induction decays. REFERENCES 1_N. Bloembergen, in H. Walther and K. W. Rothe, eds., Laser Spectroscopy IV (Springer-Verlag, Berlin, 1979), p. 340; see also the general references. 2M, D. Levenson, Physics Today 30, No. 5, 44 (1977), A. B. Harvey, J. R. McDonald, and W. M. Tolles, in Progress in Analytica Chemistry (Plerum Press, New York, 1977), p. 211;anu hw Four- Wave Mixing Spectroscopy §, A. Akhmanov and Nf, Korotecy, Sop. Phys. Uspekhi 20, 899 (1977); 5. A, Akbmanov, in N, Blocmbergen, ed., Nonlinear Spectroscopy, Proceedings of International School of Physics “Enrico Ferri” (North-Holland, Publishing Co. Amsterdam, 1977), p. 217; W. M. Tolles, J. W. Nibler, J, R. McDonald, and G. V, Knighton, in A. Weber, ed., Topics in Current Physics (Springer-Verlag, Berlin, 1977; S. Druet and J. P. Taran, in C. B. Moore, e4., Chemical ond Biochemical Applications of Lasers, (Academic Press, New York, 1979), Vol. IV; M. D- Levenson, in A. Harvey, ed,, Chemical Applications of Nonlinear Optics (Academic Press, New York, 1980); M. D. Levenson and J. J. Song, in ¥. Letokhov and M. 5. Feld, eds., ‘Advances in Coherent and Nonlinear Optics (Springer-Verlag, Berlin, 1980}; G. [. Eesley, Cohereat Raman Spectroscopy (Pergamon, New York, 1981). M. D. Levenson and N, Blocmbergen, J, Chena. Phys. 60, 1323 (1974). P. D, Maker and R. W. Terhune, Phys, Rev. 137, A801 (1965). J.J. Wynme, Phys. Rev. Lett, 29, 650 (1972); Phys. Rev. B 6, 334 (1972) M. D. Levenson, C. Flytzanis, and N. Bloembergen, Phys, Rev. B 6, 3462 (1972) E Yablonovitch, C. Flytzanis, and N. Bloembergen, Phys. Rev. Lett. 29, 865 (1972), M. D. Levenson, JEEE J. Quant. Electron, QE-10, 110 (1974); M. D. Levenson and N. Bloembergen, Phys. Rev. B WO, 4447 (1974); J, Chem. Phys, 60, 1323 (1974). S, A. Akhmanov, V. G, Dimitriey, A. 1. Koverigin, N. I. Koroteev, V. G. Tunkin, and A. I Kholodnykh, JETP Lets. 15, 425 (1972). H. Huber-Walchli, D. M. Guthals, and J. W. Nibler, Chem. Phys. Lett. 67, 233 (1979); M. D. Duncan, P. Oesterlin, and R. L. Byer, Opt. Lett. 6, 90 (1981); M. D. Duncan, P. Ocsterlin, F. Konig, and R. L. Byer, Chem. Pays. Lett, 80, 253 (1981). F, DeMartini, F, Simoni, and E. Santamato, Opr. Comm. 9, 176 (1973), M. A. Henesian, L Kulevski, and &. L. Byer, J. Chem. Phys. 65, 5530 (1976). P.R. Regoier and J. P. Taran, Appl, Pays. Lett, 23, 240.(1973); J. W. Nibler, J. R. McDonald, and A. B. Harvey, Opi. Comm. 18, 371 (1976). H.C. Praddaude, D. W. Scudder, and B. Lax, Appl. Phys. Lest. 35, 766 (1979). R. L. Byer and E. Gustafson, private communications (to be published). S.A. Akbmanov, A. F. Bunkin, $. G. Ivanov, and N. L Koroteev, JETP Lett, 25, 46 (1977). JL. Oudar, R. W. Smith, and Y.R. Shen, Appi. Phys. Lett, 34, 758 (1979). FL. Oudar and Y. R. Shen, Phys. Rev. A 22, 1141 (1980) L.A Rahn, L. J. Zych, and P. L. Mattern, Opt. Comm. 30, 249 (1979). ‘A. Owyoung and P. Esherick, Opt. Lett. 5, 421 (1980). §. D. Kramer, F. G. Parsons, and N. Bloembergen, Phys. Rev. B 9, 1853 (1974). D. Heiman, R. W. Hellwarth, M.D, Levenson, and G, Martin, Phys. Rev. Lett, 36, 189 (1976). G.L. Eestey, M. D. Levenson, and W. M. Tolles, IEEE J. Quant. Biectron. QE-14, 45 (1978). 8. A.J, Duet, FP. E. Taran, and Ch. J. Bordé, J. Phys. (Paris) 40, 841 (1979). 22° T. Yajima and H. Souma, Phys. Rev. A 17, 309 (1978); T. Yajima, H. Souma, and Y. Ishiba, Phys. Rev. A 17, 324 (1978). ¥, Por, A. R. Bogdan, M. Dagenais, and N, Bioembergen, Phys. Ree. Lett. 46, 111 (1981); A. R. Bogdan, M, Downer, and N, Bloembergen, Phys. Rev. A 24, 623 (1981); N. Bloember- gen, A. R. Bogdan, and M. W. Downer, in AR. W. McKellar, T. Oka, and B. P. Stoicheff, eds., Laser Speciroscopy V (Springer-Verlag, Berlin, 1981), p. 157 P.L. Devola, J. R. Andrews, R. M. Hochstsasser, and H. P. Trommsdorif, J, Cem. Phys. Td, 4695 (1980). H. Eichler, G. Salje, and H. Stabl, J. Appi. Phys, 44, $383 (1973); D. Pobl, $. E, Schwarz, and Y. Imniger, Phys. Rev. Lett. 31, 32 (1973)Bibliography 285 16 D. W. Phillion, D. J. Kuizenga, and A. E. Siegman, Appl, Phys. Lets, 27, 85 (1975); J. Salcedo, A. E. Siegman, D. D. Diott, and M. D. Fayer, Phys. Rev. Lett, Al, 131 (1978), J. R. Salcedo and A. E. Siegman, FEEE J. Quant. Electron. QE-15, 250 (1979) BIBLIOGRAPHY Bloembergen, N., ed., Nonlinear Spectroscopy (North-Holland, Publishing Co. Amsterdam, 1977}. Laser Spectroscopy I-V (Springer-Verlag, Berlin, 1973, 1975, 1977, 1979, 1981). Letokhov, V., and M.S. Feld, eds., Aduances in Coherent and Nonlinear Optics (Springer-Verlag, Berlin, 1980). Levenson, M. D., facroduction to Nonlinear Laser Spectroscopy (Academic Press, New York, 1982)._16_ Optical- Field-Induced Birefringence An applied de electric or magnetic field can effectively modify the refractive index of a medium. Electrooptic and magnetooptic effects were discussed in Chapter 4. The same is possible with an optical field. A sufficiently intense laser beam can induce a significant change in the refractive indices of a medium, The refractive index change in turn affects the beam propagation and leads to a new class of nonlinear optical effects characteristically different from either optical mixing or nonlinear wave attenuation. This chapter describes the various physical mechanisms contributing to the optical-field-induced birefrin- gence, and the resulting effect on the beam polarization. Connection to four-wave mixing is made. Only media with inversion symmetry are consid- ered, to avoid complications caused by the presence of second-order processes. 16.1 GENERAL FORMS OF OPTICAL-FIELD-INDUCED REFRACTIVE INDICES The nonlinear polarization P"(«w) induced by an intense monochromatic field E(w} has a general form P*(a) = Ax[o, E(w) E*(w)] *E(a). (16.1) A similar expression exists for the nonlinear polarization PN'(w’) at the probe frequency w’ PN(w!) = Ax [o’, E(w) E*(w)] E(w’). (16.2) 286Physical Mechanisms 287 Here the probe field E(w’) is assumed to be sufficiently weak so that only the term linearly proportional to Ei’) in P™*(w’) needs to be taken into account. The induced susceptibility Ay is related to the induced refractive index An by the simple relationship A(nj,) = de, = 4udx;,. (16.3) We consider here only the third-order nonlinear polarization or Ax to the lowest order, Then, by the conventions given in Section 2.10, the third-order nonlinear polarization P® («’) can be written as P= E xthile’ =o" + 0 — wo) E(w) E(w) EP (w) Bk 6 E Ror = w+ a= we (wVEy(o) Era). OO i For w = w’, it becomes PM(w)=3 E CRw- wt aw E(wIE(w)EMe). (065) A general review of third-order susceptibilities can be found in Ref. 1. In an isotropic medium, the nonzero components of x are xf,, x82, xPho. and xf» with xffh1 = xfie + xBha + xfPu. 16.2, PHYSICAL MECHANISMS ; A number of physical mechanisms can contribute to the optical-field-induced tefractive indices. We discuss here only the few commonly encountered ones. Electronic Contribution | The applied optical field can distort the electronic charge distribution in a medium, which will lead to a change in the refractive indices. Microscopically, the electronic contribution to the third-order susceptibility can be derived from the third-order perturbation calculation, as outlined in Chapter 2. For a typical } transparent liquid or solid, x“) falls in the range between 107)3 and 107} esu. However, as the optical frequencies approach an absorption band, x can be P greatly enhanced. This is particularly true if the resonant absorption is sharp. A population redistribution induced by the resonant excitation often accounts for the major part of the enhancement. Consider, as an example, a monochromatic beam of frequency # propagat- ing in a gas medium with a transition frequency &,, Close to w. It is easy to F show, from the density matrix formalism, for example, that the beam sees,288 Optical-Field-Induced Birefringence aside from a nonresonant background, a resonant susceptibility Kalenlgb = NEA (16.6 xals) A(w — w,, + Tyg) ° ) where Ap = 9, — Pp, I8 the population difference between the states |g) and jn), and N is the density of molecules. Clearly, if the beam intensity is sufficiently strong, the resonant excitation will yield a significant population redistribution, with Op decreasing as the beam intensity increases. This is known as the saturation effects, as discussed in Section 13.3. Following (13.7) for an effective two-level system, we then have Spat Tag E(@)P/R? Ap = as 1- 2 2 2 2 pr (wo ~ oyg) + Tig + ApagP Tig E(@) PZB | (16.7) where p,, = Caler|g) and Ap® is the thermal population difference. In the weak saturation Limit, (16.7) reduces to (16.8) 4p? Ty TnglE(o) 12/4? tp = Apt} — see el ¢ MP ; (o-4,,) +12, The resonant susceptibility of (16.6) in the presence of an intense beam can therefore be written in the form xn(#) = x) + Axe (o, LE(@)?)- (16.9) Here x{P{o) is the resonant part of the linear susceptibility independent of the beam intensity, and 4pieT Tag (oP /P lolayethiattt apr 1) (o ~ egg) + Thy + Sere TT E(@)PZA Axe = — xk which, in the weak saturation limit, reduces to Axa = XPlE()P? with 4p TT neff Engl Vast (16.11) (a - Wn,)° +72 xP = —xPPhysical Mechanisms 289 It is interesting to see how large x can be for an atomic vapor system. Assume N = 10" /omt, p,,= 5X 107"? esu (for the s -> p transition in an alkali vapor, for example), TI, = 1, and |w — @,,| = 1 cm > T,,. Then we obtain xf? = 0.01 esu and x= 2.5 X 107? esu. This shows that even with |E€w)} = 1 esu only (corresponding to a beam intensity of 250 W/cm), the induced susceptibility is as latge as Ax, = 2.5 X 10~? esu, or, equivalently, the induced refractive index is Ang = 1.5 x 107? esu. The large value of x§° for atomic vapor has prompted researchers to use it as the nonlinear medium in degenerate four-wave mixing work such as phase conjugatior? and nonlinear optical diffraction? The saturation effect in semiconductors can aiso yield a large x. In InSb, for example, x can reach a value of ~ 1 esu when the optical frequency moves into the direct absorption band.* The mechanism responsible for the large x@ here is somewhat different from the atomic systems because of the band structure in semiconductors. The resonant optical field pump electrons into the conduction band and leaves holes in the valence band. Because of the fast relaxation rate of carriers within a band, the excited electrons and holes quickly relax to thermal population distributions in the conduction band and the valence band, respectively. The steady-state population distributions of electrons and holes are finally determined by the balance between the excita- tion and the electron-hole recombination across the band gap. This induced population redistribution (pump saturation) results in a change in the absorp- tion spectrum, which is related to the optical-field-induced refractive index through the Kramers-Kronig relation. The large x? resulting from the saturation effect is not limited to the cases with electronic resonance. In molecular systems, the same thing can occur near a vibrational transition, although x is usually not as large as that for the atomic systems because of the weaker oscillator strength of the vibrational transition. Raman or Two-Photon Contribution The third-order susceptibility x(w’ = @’ + @ — #) can also be resonantly enhanced if |’ — «| approaches a Raman transition. This fact forms the basis of the Raman-induced Kerr effect spectroscopy (RIKES) described in Section 15.3. It is seen there that the presence of the field E(w) can induce at the probe frequency w" a susceptibility change Ax(’) = xX"): E(w)E*(w), or a corresponding refractive index change An(w’). Typically, x ~ 3 x 107” esu in a liquid with |@— w'] near a strong Raman resonance. Therefore, with |E(w){ = 100 esu (corresponding to a beam intensity of 2.5 MW/cm?), one finds An ~ 107-8. Similarly, xO(w’ = 0” + w — w) can be resonantly enhanced if w + 0 approaches a two-photon resonance. The presence of E(w) again induces a susceptibility change Ax(o') = x%(w):E(w)E*({o) or a corresponding An(’) at wo’. For both Raman and two-photon cases, x(w’) is further290 Optical-Field- Induced Birefringence enhanced if w or w’ is also near a transition to an intermediate state. In sodium vapor, for example, with w, + w, on the 3s — 4d transition and w, being 10 cm! away from the 3s —* 4p transition, one finds x ~ 10-75N esu where N is the atomic density. For N ~ 10'S/cm® and |£|~ 100 esu, the induced refractive index change An is as large as 10-55 In molecular systems, x°(« = w@ + w@ — w) can often be greatly enhanced as 2 approaches the » = 0 > v = 2 vibrational transition because w, at the same time, is near the v= 0 > v=1 resonance. The large x®) seen by a 10,6-4m CO, laser in SF, gas is a typical example. Electrostriction Application of a de electric field to a local region of a medium causes an increase in the density of the medium in that region. This field-induced density redistribution occurs in order to minimize the free energy of the system in the presence of the field, and is known as clectrostriction. The same effect is expected with the optical field, since the de field energy and the opiical field energy are equivalent in this case. The induced density change then leads to a change in the susceptibility or refractive index. The essential mathematical derivation has already been given in Section 11.1 in connection with the discussion on stimulated Brillouin scattering. The induced density variation App should obey the driven acoustic equation of (11.2), which appears in the case of monochromatic pump beam as ee I Pri - 20, vy "|e. ~ -v3(Z16(4)?). (16.12) If the beam is CW, the time-independent solution is App = (16.13) The resulting nonlinear polarization seen by a probe beam at w* is PO(a) =f ee D ApoE’) (16.14) and hence ax(o} =e 22) ny Pp. (16.15) Here we assume that «” is very different from w so that App(w’ — w) induced by the beating of E(w) and E(w’) is negligible. Otherwise, an additional term in P@('), and hence Ax(«’), appears resulting from coherent scattering of 3Physical Mechanisins 291 Ei) by App(w’ ~ w). Typically, for liquid, Ax ~ 107? esu or An ~ 10-1! esu with |£| ~ 1 esu.’ In a transparent medium this electrostrictive contribu- tion often is appreciably larger than the electronic contribution. However, as we discuss later, the response of the density variation to the applied field is slow, and therefore, with short optical pulses, An can never reach its full steady-state value. Molecular Reorientation and Redistribution Assume a linearly polarized beam propagating in a liquid. If the molecules are anisotropic, the optical field tends to align the molecules through interaction of the field with the induced dipoles on the molecules. Furthermore, because of the presence of induced dipole interaction between molecules, the molecules will spatially redistribute themselves in order to minimize the free energy of the system. Both molecular reorientation and redistribution lead to a change in the refractive index of the medium. In Section 11.3, optical-field-induced molecular reorientation in an isotropic medium was described at length, We reviewed it here with a more general derivation. We can define an orientational order parameter Q by the macro- scopic relation® Xax =X + FAN O (36.16) assuming E along 4, where X = 4(x,. + Xy) + X..) is the average linear susceptibility of the medium, and Axp is the anisotropy of the medium when all molecules are aligned parallel (Q = 1). We also have Xyy = Xaz =X 7 FAX. (16.17) The order parameter Q defined here is a macroscopic quantity while the one in Section 11,3 is microscopic. The two become identical if the molecules are noninteracting, In the liquid phase, the induced ordering is expected to be small, so that Q <« 1. According to Landau’s theory, the free energy of the system can then . be expanded into power series of Q:° FoF t ha(T—T")Q? + 4BQ* + ++ — 2X + AXO)E(w)? (16.18) : where F, is independent of Q, T is the temperature, @ and B are constant coefficients, and 7'* is a fictitious second-order phase transition temperature at which the system would necessarily make a transition to the ordered phase if it had not already done so. In the limit of small Q, terms of orders higher than292 ‘Optical-Field- Induced Birefringence Q? in F can be neglected. Then minimization of ¥ immediately gives _ 4 Axl E(w)? Q= sa(T = T*) (16.19) The induced change in the susceptibility tensor seen by a probe beam at w, from (16.16) and (16.17), appears to be AX gp = T2AXyy = ~ 2X, = Phx _ BAxolo')bx0(#) py y)2 Sa =T*) Tr) |E{a)/?. (16.20) Again, we have assumed that |e - «| is much larger than the inverse of the relaxation time for molecular reorientation, so that Q(w’ — w) induced by the beating of E(w) and E(w’) is negligible. Otherwise, an additional term should appear in Q, and hence Ax. In the ideal limit of noninteresting molecules, one would, of course, have T* =0, and the foregoing results should reduce to those in Section 11.3, namely (11.23) with £, = E, = £ and w =o’. A direct comparison of the expressions of @ then yields a = 5Nk, for this ideal case. We have considered only molecular reorientation induced by a linearly polarized light. What happened if, instead, a circularly polarized light is used? Will there be a circular birefringence induced in an isotropic medium? Physi- cally, it is easy to see that the optical field should reorient the molecules toward the (2-) plane perpendicular to the direction of propagation (2), but in the 4-H plane, the molecules are randomly distributed since they cannot follow the field rotating at the optical frequency. Thus one expects a field-induced linear birefringence between 2 and %-J, and no birefringence in the £-) plane. The induced susceptibility changes should be ax, = Ay.= — 44x. (16.21) where the subindices “+” and “—” denote the right and left circular polari- zations, respectively. Since 4y,,, should be the same, irrespective of a linearly polarized aligning field along % or a circularly polarized field in 2-f, we have chy - 2oxolw}Axo(e) ax.= dx. Sy HEC) (26.22) which is four times less than Ax ,, induced by a linearly polarized field along 2. In ordinary liquids composed of anisotropic molecules, molecular reorienta- tion can yield a An of the order of 107" to 107" esu for |E| = 1 esu.’ The response time of molecular orientation is usually of the order of 10 psec.Physical Mechanisms 293 Therefore, with nanosecond and subnanosecond laser pulses, molecular re- orientation is often the dominant mechanism for the observed An. For liquid crystalline media in the isotropic phase, however, An can be much larger.” This is because in such media T* may be less than 1 K below the isotropic— mesomorphic transition, and according to (16.20), as T approaches the transi- tion temperature, Ax or Am increases inversely with (T — T*). This pretransi- tional behavior is commonly known as critical divergence. It has been found that An can seach a value of ~ 1079 esu for |E| = 1 esu even at a temperature 5 K above the transition. The response time of An, however, also increases with (7 — T*)7}, which is known as the critical slowing-down behavior. At 5 K above the transition, it is of the order of 100 nsec.” Molecules in the isotropic liquid phase are more or less uncorrelated. Their response to the external perturbation therefore is an unconcerted effort and cannot be very large. On the other hand, molecules in a liquid crystal phase are well correlated, at least in orientation. The orientational response of the molecules to the external perturbation is a group effort, and can be expected to be extremely large. Indeed, with |£] ~ 1 esu, the observed average An in a nematic liquid crystal film can be ~ 0.05.!° Because of the wall-aligning force on molecules, the induced molecular orientation is not uniform across the film. The induced Ax is usually a strong nonlinear function of |£|? even at |E} ~ 1 esu. Being a correlated response to the applied field, this induced molecular reorientation is very slow, having a response time of the order of 1 sec or larger. The molecular reorientation can contribute to An only if the molecules are anisotropic. However, it has been found that in transparent liquids composed of nearly spherical molecules or atoms, an appreciable An can still be induced by laser pulses. The observed An is much larger than what one would expect from electronic and electrostrictive contributions and must arise from optical- field-induced redistribution.’ The theory of molecular redistribution is unfor- tunately not yet well formulated. Both molecular reorientation and molecular redistribution can occur only if molecules are free to rotate and move in the medium. With a few exceptions, this is generally not the case in solids. Therefore, they should not contribute to the observed An in solids. Other Mechanisms There are many other possible mechanisms that may contribute to An. in an absorbing medium, the optical-field-induced temperature rise AT will certainly lead to a change in the refractive index. The induced concentration variation AC in a mixture is another possible mechanism. The expressions for AT and AC can be obtained following the derivation in Chapter 11. An extremely large An can also be obtained in photorefractive materials such as BaTiO,.'' The optical beam presumably excites and redistributes the charges trapped at various sites in the crystal. The charge redistribution sets up a strong internal electric field,294 Optical-Field-Lnduced Birefringence which in turn induces a large An via the electrooptical effect. Even with a 10-mW/cm? laser beam, the induced An can be larger than 1075, although the response is very slow (~ 1 sec) at such a low laser intensity. 16.3 OPTICAL KERR EFFECT AND ELLIPSE ROTATION The optical-field-induced birefringence can change the polarization of a beam propagating in the medium. Conversely, measurements of the change of the beam polarization should allow us to deduce values of the induced birefrin- gence. Consider first the optical Kerr effect, which usually refers to the phenome- non of linear birefringence induced by a linearly polarized optical field. We assume an isotropic medium here for simplicity. With a pump beam at w and a probe beam at o’, the third-order polarization at takes the form Pw) = L[xMa(o’ = of +o - oF (aE (w)E*() i $xQa lo! =o + w= WE (OE (e)E*(o) (16.23) $xBu(w = of + 2 = WE Mw E (0) E"(0)]- If the two beams are parallel and their polarizations are linear but at 45° with respect to each other, then with the pump field E(w) along % propagating along 42, one finds P(e") = (xP + xBbe + Pa E.(oE(W EC), Po) = xB, (o Eo) E*(a). (1624) The field-induced anisotropy in the susceptibility is therefore given by'* 8x(w') = Ox.. — Axyy (16.25) = (xB + xB IEW)? or the induced linear birefringence given by 2a an,(w’) = Zox(w/) (16.26) 2a = Oa + xPa)lZ()P. Then, in propagating through a medium of length f, the & and § components ofOptical Kerr Effect and Ellipse Rotation 295 the probe field experience a relative phase difference 8g = (w'/c) inl. (16.27) This phase difference changes the polarization state of the probe beam, and can be measured by an analyzer crossed with the polarization of the incoming probe beam in front of the detector. A typical experimental arrangement is seen in Fig, 16.1.!2 The signal should be proportional to sin’(89/2), from which $4, and hence &,, can be deduced. We note here that the same result with $< 1 can be obtained from the four-wave mixing approach, as de- scribed in Section 15.3. The four-wave mixing output corresponds to the generation of an orthogonal polarization component in the probe beam. As mentioned in the previous section, when w approaches w’, the beating of E(w) and E(w) provides an additional contribution to Ax. The degenerate case with w =o and k = k’ is particularly interesting since we now have a single monochromatic beam propagating in the medium. Will the polarization state of the beam vary during propagation as a resuit of the induced birefrin- gence? The answer is yes, in general, as one can expect from the four-wave mixing picture. We consider here the special case of an isotropic medium. The third-order polarization has the same general expression given in (16.23) with w’ = w. In terms of circular polarization components, we can write? PY(a) = (xflo + xf) Es PE. (#) 9 ; ; (16.28) + (xe + xBa + Axi NEx(@)PE.(o). Scope D-I 4 | 50% Beom Grating | Splitter | b-f--Ik-24-F +4 414-5 Ruby BS el P-I Somple P-3 Ye Loser P2 ai He-Ne Light. p Rocke! A eco xX Ruby Light D-2n,” BS pike Filter at 6328 He-Ne Scope Loser Fig. 16.1 Experimental arrangement for optical Kerr measurement. P-1, P-2, P-3 are polarizers, and D-1, D-2 are detectors. (After Ref. 9.)296 Optical-Field-Induced Birefringence Far a linearly polarized field, Fw) = £E(), (16.28) reduces to PO(a) = SxPulE(o)/7E(w) (16.29) and for a circularly polarized field, E = é , E(w), it becomes POW) = 8, OePe + Pa )IE(@)PE(o). (16.30) In both cases, the induced P®() has the same polarization state as the incoming field, and therefore no change in the beam polarization is expected as the beam propagates in the medium. More generally, however, for an ellipti- cally polarized field, E = @,£,+ @_E_, (16.28) shows that the induced refrac- tive indices for the two circularly polarized components ate, respectively, 2a An =~ n *[(xPe + xP EP + (xP + xBho + 2x EP h- (16.31) The difference between An, and An_ here indicates that the elliptically polarized beam can induce a circular birefringence in the medium Ang = Sn,~ An_= ~ (2m /n)2x8h(\E,I? — EP). (16.32) As discussed in Section 4.2, a circular birefringence renders a rotation of the beam polarization. In the present case, the rotation of the elliptical polarization across the medium of length / is given by”? = (w/2c)And. (16.33) Note that # depends on both the beam intensity and the ellipticity according to (16.32}. Measurement of this intensity-dependent ellipse rotation @ allows us to deduce dn, and hence x{,. Figure 16. 2 shows a typical experimental arrangement for ellipse rotation measurements.'? Because of the beam intensity variation in the transverse direction F, @ actually varies with r. To deduce x{}), from the measurement of 4, one should limit the detection to a small region in the beam profile where the intensity variation is negligible. Alternatively, one can use a focused beam with a Gaussian beam profile.!* The average ellipse rotation of the entire beam can be calculated in terms of An, and the beam characteristics, and compared with the measurements. From (16.26) and (16.32), it is seen that the combined results of the optical Kerr measurement (with w ~ w’) and the ellipse rotation measurement allow us to determine two of the three independent elemenis of x®, namely x95 and x{%,. The third, x{?,, generally can be obtained only by resorting to‘Transient Effects 297 Scope D-t [Ot PMD gpg ge HD rem Ruby ey v “yD bs ‘ Loser Fi = PLR Ul Sample LZ R2FZ Pe O02 Fig. 16.2 Experimental arrangement for ellipse rotation measurement. P-1, P-2 are polarizers, R-1, R-2, Fresnel Rhombs, and D-1, D-2, detectors. (After Ref. 9) another independent experiment, for example, a degenerate four-wave mixing experiment. The optical-field-induced variation of the beam polarization in an aniso- tropic medium is often complicated, because of the existing linear birefrin- gence even in the absence of an induced An. The problem is particularly interesting when An is so large that it changes drastically the beam propaga- tion characteristics in the medium. This may happen, for example, in liquid crystals.!4 16.4 TRANSIENT EFFECTS The foregoing discussion was limited to the steady-state case, The response of a medium, however, becomes transient if laser pulses with pulsewidths shorter than the response times are used. The response times are different for different physical mechanisms contributing to the induced refractive index. The elec- tronic contribution has a response time on the order of (w ~ w))7' ~ 107 sec, if the optical frequency « is far away from any absorption band, where w, is the position of the major absorption band, On the other hand, for An induced by the population redistribution, the response time is determined by the population relaxation. For the électrostrictive contribution, the transient response is governed by the equation of motion (16.12) for acoustic waves.’ With the laser pulse given, the time-dependent density variation App(r, £) can be solved from (16.12) and Ax() is directly proportional to App(r, 1). In this case, the response time is characterized by the inverse of the damping coefficient, T;', and the time it takes for the acoustic wave to travel across the beam radius, R/v. If the pulsewidth 7, is much longer than both Ty! and R/v, the response is quasi-steady-: State, Yet if 7, « R/v < Ty", the response becomes transient, with App ~ ApS {T7/T ura) near the end of the pulse, where Api, is the steady-state value. tt T, «Ty! « R/o, the response is also transient, with App ~ del AR/e) } For R = 1imm,v = 2X Wom/sec, Ty! ~ R/o = 5 X 10-7 sec, and T, = 10-* sec, we have App ~ 4X 10-* aps, The corre-28 ‘Optical-Field-Induced Birefringence sponding induced refractive index is also 4 X 10~* times smaller than the steady-state value. This example shows that the electrostrictive contribution to An is far less important with nanosecond or picosecond Laser pulses than with tonger pulses. For the molecular reorientational contribution, the order parameter @ obeys the driven Debye equation given in (11.22), or more generally, by the following dynamic equation for Q including dissipation: p22. _ oF a aQ (16.34) = -a(f- T*)Q + fAxglE(w)|? where » is a viscosity coefficient. The solution of the equation is simply 2 ot) =f SBE peenceay (16.35) =o with the relaxation time r defined as + =v/a(T — T*). (16.36) For ordinary liquids, r is of the order of 10 psec. Then the response of molecular reorientation to a nanosecond laser pulse is expected to be quasi- Steady-state, but to a picosecond pulse, it is still transient. For liquid crystals in the isotropic phase, + can be much larger, approaching 1 psec as T approaches the isotropic-mesomorphic transition temperature.” In the meso- morphic phase, the collective motion of the molecules slows down the response even more drastically. The response time can be longer than 1 sec.!° 16.5 APPLICATIONS A number of applications can be derived from the optical-field-induced refractive index change. In Section 14.3, we saw how degenerate four-wave mixing can be used for phase conjugation and image reconstruction. As described there, degenerate four-wave mixing can result from coherent scatter- ing of a light beam by a refractive index grating induced in the medium by the interference of two pump light waves. Thus a medium with a large optical- field-induced refractive index per unit field intensity is most useful for efficient degenerate four-wave mixing applications. The optical Kerr effect can be used in optical switching.” As shown in Fig. 16.3, the transmission of the weak signal beam passing through the nonlinear medium is normally blocked by the crossed analyzer. In the presence of an intense pump beam, however, the medium becomes birefringent, and the signalApplications 299 beam, experiencing a polarization change in traversing the medium, is no longer completely blocked by the analyzer. Therefore, if a picosecond pump pulse is used, a nonlinear medium with a picosecond response time can act as a fast optical gate for a signal beam with a picosecond on-off time. Such an optical switch is clearly very useful in many applications, especially in dynamic study of physical phenomena and mechanisms. If a Fabry-Perot interferometer is filled with a nonlinear medium, then because of the optical-field-induced refractive index, the transmission of a monochromatic beam through the interferometer depends on the beam inten- sity. For example, with the interferometer spacing tuned to a value for peak linear transmission, an intense light beam may find the interferometer badly mistuned because the field-induced refractive index has caused an additional phase change on the beam traversing the interferometer. In general, the transmission of the nonlinear Fabry—Perot interferometer is a highly nonlinear function of the beam intensity, depending on the initial phase setting of the interferometer. Three generat forms of transmission characteristics of the interferometer can be obtained, and can be used for intensity limiting, differen- tial amplification, and bistable operation.’ The last one has been the focus of very active research because of its great potential applications in optical data processing,?” Optical bistable operation of a nonlinear Fabry—Perot (FP) interferometer is illustrated in Fig, 16.4. The FP transmission curve T versus ¢ in Fig, 16.4a is described by the equation qh T= Ta Fantle/2) Fain®(6/2) (16.37) where Ty and F are constants, Because of the field-induced refractive index change in the medium inside the interferometer, the round-trip phase retarda- tion @ now depends on the field intensity. Assuming An = 1,/E {?, we can write $= bo + Kg? (16.38) With 4, and & being constants, (16.38) describes a straight line in Fig. 16.4@, Pump pulse 4 —_— Signal beam IN Y Polarizer Analyzer Kerr cell Fig. 16.3 Optical Kerr cell as a fast optical switch.300 Optical: Fiehi-Induced Birefringence UK ple (Unle Uns Tia {b) Fig. 164 (a) Graphical method of finding the operating points of a nonlinear Fabry-Perot interferometer. When /,, increases, the operating point moves along the path 1-2-4-3-3-4, and when J, decreases, the operating point moves along the path 4-3-B-22-1. (4) oy Versus Fi, in the form of a hysteresis loop corresponding to the operating path described in (2). the slope of which is proportional to the incoming laser intensity 1/J,,. Given Iq, the operating point of the interferometer is then determined by the solution of (16.37) and (16.38), corresponding to the crossing point of the straight line with the FP transmission curve. It is seen in Fig. 16.4a that if f,, is sufficiently large, more than one operating point can exist. Some of them (such as B and D in Fig. 16.4a) are unstable. Among the stable ones, the real operating point is selected by the operating condition that if the operating point is varied, the interferometer preferes to have it varied smoothly along the curve. Thus, asReferences 301 shown in Fig. 16.4@, when J,, increases, the operating point should move along 1-2-A-3-3'—4, but when 4, decreases, it should move along 4-3’-C-2'-2-1. As a result, the corresponding /,,, versus Ji,, sketched in Fig. 16.45, takes the form of a hysteresis loop. For J, between (J,,), and (J,,);, the output.can have either a high value or a low value depending on the operation path. This bistable behavior is the basis of binary switching elements. Therefore, the nonlinear FP interferometer can become an important optical device in optical data processing and all-optical logic and computing systems. Although the nonlinear FP interferometer is commonly used to study optical bistability, we should note that the samme phenomenon can arise in many other systems. Away from the operational point of view, a nonlinear FP interferometer is also interesting because its switching action resembles a phase transition," and it is a nonlinear system with positive feedback that may lead to bifurcations and chaos in the output.” Both the optical Kerr effect and the intensity-dependent ellipse rotation can be used for pulse shaping. The polarization state of an intense laser pulse in traversing a nonlinear medium is time-dependent, resulting from the intensity- dependent polarization rotation. Therefore, through the use of an analyzer, the transmitted pulse can be reshaped. The same effect can be used in a laser cavity to reshape the laser pulses.”° The optical-field-induced refractive index can also lead to an intensity- dependent distortion of the beam wavefront, resulting in self-focusing and other self-actions of light. This is the subject of discussion in Chapter 17. REFERENCES 1 R.W_ Hellwarth, Prog. Quant. Electron. §, 1 (1977). 2 D.M. Bloom, P. F. Liao, and N. P. Econorou, Opt. Lett. 2, $8 (1978). N, Tan-no, K. Ohkawara, and H. Inaba, Phys. Reo. Lett. 46, 1282 (1981), C. V. Heer, Opr. Lert, 6, 549 (1981). D. A. B. Miller, C. T. Seaton, M. E. Prise, and §. D. Smith, Phys. Rev. Lett. 47, 197 (1981). P. F. Liao and G. C. Bjorklund, Phys. Rev. Lett. 36, 584 (1976). D. G. Steel and L. F. Lam, Phys. Rev, Lett. 43, 1588 (1979) Y.R. Shen, Phys. Lett. 20, 378 (1966). P, G. De Gennes, Mol. Cryst. Lig. Cryst. 12, 193 (1971) G. K. L, Wong and Y. R. Shen, Phys. Reo. 4 10, 1277 (1974). In this reference, a tensorial order parasteter is used, and the coefficient a is three times less than the one defined here, ‘Also, the field energy is e|E|?/8x there, while bere it is defined as «| E|?/2. 10S. D. Durbin, 5, M. Arakelian, and Y. R. Shen, Pays. Rev. Lett, 47, 1411 (1981). 1 J. Feinberg, D. Heiman, A. R. Tanguay, and R. W. Hellwarth, J. Appl. Phys. 51, 1297 (1980). 12 See, for example, E. G. Hanson, Y. R. Shen, and G. K. L. Wong, Phys. Rev. 14, 1281 (1976); A. Owyoung, R. W. Hellwarth, and N. George, Phys. Rew. 4 4, 2342 (1971p. 13 P. D. Maker, R. W. Terhune, and C. M. Savage, Phys. Rev. Lett, 12, 507 (1964), rs CoH awe302 14 15 16 IF ‘Optical Field-Induced Birefringence S. D. Durbin, S. M. Arakelian, and Y. R. Shen, Ops. Lew. 6, 411 (1981); Pays. Rew. Lett, 47, 1411 (1981). M. A. Duguay and J. W. Hanson, Opt. Comer. 1, 254 (1969) HL. Seidal, US. Patent No. 3610731; A. Szoke, ¥, Danean, J. Goldhar, and N. A. Kurait, Appl. Phys. Lett, 1S, 376 (1969); H. M. Gibbs, 8. L. MeCall, and T. N. C. Venkatesan, Phys. Rev. Lett. 6, 1135 (1976). See C. M. Bowden, M. Ciftan, and H.R. Robl, eds., Opticaf Bistabitisy (Plenum, New York, 1981); C. M. Bowden, H. M. Gibbs, and 8. L. McCall, eds., Optical Biseabitity IZ, (Plenum, New York, 1983), See, for example, R. Bonafacio and L. A. Lugiato, in H. Haken, ed., Pacent Formation by Dynamic Systeras and Pattern Recognition (Springer-Verlag, Berlin, 1979). K. Ikeda, H. Daido, and 0. Akimoto, Phys. Rev, Lett, 45, 709 (1980); 48, 617 (1982). See, for example, J. Schwar, W. Weiler, and R, K, Chang, EEE J. Quant. Electron. QE-6, 442 (1970); L. Dahlstrom, Opi. Comm. 4, 289 (1971): K. Sala, M. C. Richardson, and N. R. Isenor, [EEE J. Quant. Electron, QE-13, 15 C97). BIBLIOGRAPHY Chang, T. Y., Opt. Eng. 20, 220 (1981). Hellwarth, R. W., Prog. Quant. Electron. 5, 1 (1977).Self-Focusing Self-focusing of light has fascinated many researchers in the past. It is typical of the type of nonlinear wave propagation that depends critically on the transverse profile of the beam. Theoretically, the wave equation governing the effect is the prototype of an important class of partial differential equations such as the Landau-Ginzburg equation for type-II superconductors and the Schrédinger equation for particles with self-interactions. Practically, the effect often is responsible for optical damage of transparent materials, is a limiting factor in the design of high-power laser systems, and sometimes plays an important role in the occurrence of other physical processes in a medium. Although a complete solution of self-focusing and related effects requires extensive numerical calculations, good physical understanding of the problem can still be obtained from solutions with approximations based on experimen- tal findings. This is the emphasis of our discussion in the present chapter. Aside from self-focusing, there also exist a number of other self-action phe- nomena. Only self-defocusing, self-phase-modulation, and self-steepening are briefly discussed here. 17.1 PHYSICAL DESCRIPTION We begin with a physical description of the self-focusing phenomenon. Briefly, self-focusing is an induced lens effect. It results from wavefront distortion inflicted on the beam by itself while taversing a nonlinear medium. Consider a single-mode laser beam with a Gaussian transverse profile propagating into a medium with a refractive index given by 7 = 2, + An(|E|*), where An(|E|?) is the optical-field-induced refractive index change (see Chapter 16). If Av is positive, the central part of the beam having a higher intensity should experi- ence a larger refractive index than the edge and therefore travel at a slower velocity than the edge. Consequently, as the beara travels in the medium, the original plane wavefront of the beam gets progressively more distorted, as seen 303304 Self-Focusing Fig. 17.1 Distortion of the wavefront of a laser beam leading to self-focusing in a nonlinear medium. in Fig. 17.1. The distortion is similar to that imposed on the beam by a positive lens. Since the optical ray propagation is in the direction perpendicular to the wavefront, the beam appears to focus by itself. However, a beam with a finite cross section should also diffract. Only when self-focusing is stronger than diffraction will the beam self-focus. Crudely speaking, the self-focusing action is proportional to An(|E£|?), while the diffraction action is inversely proportiona? to the square of the beam radius. Therefore, as the beam shrinks on self-focusing, both the self-focusing action and the diffraction action become stronger. If the latter increases faster than the former, then at some point diffraction overcomes self-focusing, and the self-focused beam, after reaching a minimum cross section (the focal point), should diffract. In many cases, however, the field-induced refractive index can be approximated by An = n,|E|?, where n, is a constant. Then, because |E| is inversely proportionat to the beam radius, the self-focusing action is always stronger than the diffraction action if it is initially so. The beam may keep on self-focusing until some other nonlinear optical effect sets in to terminate the process. In such a case, the cumulative action of the nonlinear iterative effect makes the beam shrink sharply and suddenly, as in Fig. 17.1. The focal point and the focal distance (z;) are then well defined. Which nonlinear effect actually sets in at the focal point to terminate the self-focusing process depends on the medium. It could be, for example, stimulated Raman scattering, stimulated Brillouin scattering, two-photon absorption, or optical breakdown. A special case of interest occurs when the self-focusing action and the diffraction action on the input beam just balance each other. One would then expect that the beam should propagate in the medium over a long distance without any change in its beam diameter, This is known as self-trapping. In practice, however, self-trapping in the above context is not a stable situation, Any smail loss of the laser power due to absorption or scattering can upset the balance between self-focusing and diffraction and cause the beam to diffract. As we shall see in Section 17.2, for self-focusing to be stronger than diffraction, we must have An > 1/k?a?, where & is the wavevector and a is the beam radius, Thus, fora ~ 1 mm, and k ~ 2 x 104 cm™}, self-focusing occurs only if An = 1077 esu. In most media, such a large An can be induced only by a laser intensity higher than several megawatts per square centimeter, normallyPhysical Description 305 Fig, 17.2. Image of small-scale filaments at the exit windows of a CS, cell created by self-focusing of a multimode laser beam. [After S.C. Abbi and H. Mahr, Pays. Ree. Lett, 26, 604 (1971}.] achievable only with pulsed lasers. Self-focusing should then have a time dependence resulting from the amplitude variation of the input laser pulse, Yet if the response of the medium to the field can be considered as instantaneous, the steady-state description of self-focusing is still applicable, except that the focal distance now varies with time in response to the laser intensity variation. This is known as quasi-steady-state self-focusing. If, however, the laser pulse- width is shorter than or comparable to the response time of An, then the time variation of An also becomes important in self-focusing since propagation of the leading part of the pulse can influence propagation of the lagging part. This is the regime of transient self-focusing. A more detailed discussion on quasi- steady-state and transient self-focusing witl be given later. Askar’yan! first suggested the possibility of self-focusing due to An(E|?). Hercher? found, in carly 1964, that by propagating a Q-switched laser beam of a few megawatts in a solid, one could obtain long threads of damage spots only a few microns in diameter, Chiao et al.* soon proposed the self trapping model to explain the observation, assuming that the damage tracks were induced byFig. 173 Images of a self-focused single-mode laser beam at the exit window of a toluene cell of different cell lengths: (a) short cell length, beam not yet self-focused (~ 700 zm); (6) cell length close to self-focusing threshold, the self-focused beam: at nearly one-tenth of its original size {~ 50 pm); (c}) cell length above the self-focusing, threshold, the setf-focused beam at its limiting size —the filament (10 wm). [After Y. R. Shen, Prog Quant, Electron. 4,3 (1975).] 306Theory 07 the self-trapped laser filaments. It was shown much later that the damage tracks were actually due to time-varying self-focusing with moving focal points.** In the meantime, stimulated Raman scattering was discovered, but it was found that in many solids and liquids, it had a very sharp threshold which could not be explained by the usual theory of stimulated Raman scattering.* It was then realized that stimulated Raman emission in such media was actually initiated by self-focusing at the focal point, and the sharp focusing of the self-focusing process was the cause of the sharp onset for stimulated Raman scattering,’ Self-focusing can also account for many other observed anomalies in stimulated Raman scattering. Earlier photographs of the self-focused beam in a Kerr liquid showed that the beam shrank upon self-focusing and then broke into many filaments with nearly constant diameters? A typical example is shown in Fig. 17.2. Each filament has a diameter of the order of 10 pm, which appears to be a characteristic of the medium. That many filaments could result from self-focus- ing of an apparently single-mode Gaussian beam was a surprising fact and attracted a great deal of attention. Later, however, it was found that the multiple filaments actually originated from the weak multimode structure in the beam. When a truly single-mode laser was used, self-focusing of the beam did lead to only one single filament, as shown in Fig. 17.3. Then the problem remained interesting because the formation and the characteristics of the filament were not understood. An important question was whether the ob- served filament was a manifestation of the predicted self-trapping phenome- non.? It turned out that the filament was simply the trajectory of the focal spot in the time-varying self-focusing process achieved with a pulsed input.” We discuss the filament problem in more detail in later sections, but first we consider a more quantitative theory of self-focusing. 17.2 THEORY The formal theory of self-focusing is fairly simple. It is described by the nonlinear wave equation VE — (82/c7817)|(ny + An)’E] = 0 a7) assuming that the medium is isotropic, the field is transverse, and the medium response is instantaneous so that An(|£|*) does not depend explicitly on ¢. In this case, the differentiation in the transverse direction is nonnegligible, but since the field amplitude is not expected to vary appreciably over a distance of a wavelength, we can still use the slowly varying amplitude approximation.HS Self-Focusing Then, for a quasi-monochromatic beam propagation along 2, (17.1) can be reduced to a first-order partial differential equation in z and 7. The equation can be further simplified by climinating @/@t with the substitution of the reduced time variable ¢ = ¢ — 2/0,, where v, is the group velocity. Thus by writing E = f(r, 2, £yexp(ikz — iw), (17.1) becomes kt 4 gt len —2n2{ A (ine gZ + vi}e~ e2| | (17.2) where the beam profile is assumed to be circularly symmetric with + being the radial coordinate. Both the absolute amplitude and the phase of the field are expected to be functions of r, z, and £ with & = 4 exp(id), (17.2) can be split into two coupled equations for the absolute amplitude A and the phase :"! a ka = -v, -(4’v 9) (17.3a) and aot 2_k( VIA An) _ a? + 3g (V9) 3 ea +2 =0. (17.3b) Equation (17.3a) is an energy relation, while (17.3b) describes the ray trajec~ tory. Since the phase function $(r, z, £) actually represents the wavefront of the beam, (17.3b) is a description of how the self-focusing action, represented by 2An/ng, and the diffraction action, represented by At, A/k?A, distort the wavefront. If at z = 29 there is an exact balance of self-focusing and diffraction such that An | ViA "A =0 (17.4) for all r, and if, in addition, the wavefront is flat at z) so that V_, ¢ = 0, then (17.3) yields #¢/dz = 0 and 94/8z = 0 for z > zp. This is the self-trapping case: the wave propagates in the medium with a plane wavefront and a constant transverse profile. The self-trapping solution of (17.4) for An = n|E|? tan be obtained analytically.’ However, it is rather unstable. A stnall deviation of A(r, z) from the specific form of the self--trapping solution will cause the beam to either self-focus or diffract, or partly self-focus and partly diffract. Equation (17.3b) has the same form as the Hamilton—Jacobi equation + 8S/ét = 9 in classical mechanics,!? where #’= p?/2m + V is the Hamil- tonian of a particle in a potential well V, and S(q, p, t) is the Hamiltonian’s principal function. In our case, #(r,z) plays the role of S, and z, 7, k‘Theory 309 correspond to f, q, m, respectively, with (V 1%)? equivalent to p? and 2 v= # +28) 2\ kA Ro Then, by knowing that the Hamilton-Jacobi equation should Jead to the usual equation of motion, md?g/dt? = aV/aq, for the particle, we obtain similarly, from (17.3b), the equation 2 2/214 yA), (1735) KA Ao which governs the optical ray trajectory r as a function of z. The solution of (17.5) can therefore be perceived from the motion of a Particle in the potential well V. However, in our case, V is known only if A(?, 2) is known, assuming the function An(|E]?) is specified, but A(r, z) can be obtained only by solving the coupled equations in (17.3). As an approxima- tion, one can assume a certain functional form for A(r, z). This approximation is found to be reasonable as long as ray bending during focusing is not significant. For example, we can assume that the central part of the beam Tetains its Gaussian profile as it propagates, but the beam radius varies with z: this means, for0 [—V(ay)/2k], then the beam can never self-focus but will diffract to infinity, although it may first focus if (da/dz)y < 0. In real experiments, however, saturation of An may not occur even in the focal region. Other nonlinear optical effects often set in to affect self-focusing long before the laser intensity reaches a level to saturate An. In fact, the paraxial approximation used in the above derivation also breaks down in the sharp focusing region in the common practical cases. Then, in reality, our calculation here applies only to the prefocusing region with An = n,(£/?. In such a case, for a Gaussian beam in the paraxial approximation r < a, the potential V takes the form _ 1 n Ad kg? to (17.11) = al - # ka? Py where 242 Moe £42 ngea*Ag p= f A 2ar de = “ES is the laser power, and Py = cd’/8e7n,, Using the particle analogy, we immediately see that self-focusing can occur only if P > Py. As long as the initial beam divergence (k/2)(da/dz)3 is less than — (ay), the self-focusing action is always stronger than the diffraction action, and the beam radius should eventually reduce to zero. In the special case of P = P,, we have V = 0, and if (da/dz), = 0, the beam should propagate without a change in the beam tadius. This corresponds to the self-trapping case. With V given in (17.11) the integral of (17.9) yields a P\ 22? da\ 2} Selt-zlaa tp + (SZ) 2 Taz a} Rhea [ dz a a ) which shows how the beam radius reduces as a function of the propagation distance z. The sharp focal point should appear at 2 = z, corresponding to a= 0. We find kaj V2 ” (P/Pe= 1? = (ka WE dade (748) 2312 Self-Focusing If (da/dz)g = 0, then it becomes kaz/¥2 yo lie (17.14) (P/%-1) The above solution obtained with the paraxial approximation is of course valid only for ray propagation close to the beam axis. The more rigorous solution of (17.3) can be obtained numerically on computers.) It is found that for An = 2,|EP and P > P,, = (.22A)'c/128n (P,, is known as the critical power for self-trapping>), the initiatly plane Gaussian beam can self-focus into a sharp focal spot at! 0.43ka 22a (17.15) {[(P7P..)'? 0.852]? - ow2is} The relation between z, and yP/P,, is plotted in Fig. 17.5. For P > 1.2F,,, 1505: 1ao) (ory? O55} L ° ay 27 ay a @ Fig. 175 Curve A describes the dependence of the self-focusing distance on input power; curve B is the asymptote of curve A at high powers; curve C describes o in (17.17) as a function of the input power. (After Ref. 14.)‘Quasi-Steady-State Self-Focusing 313 (17.15) can be approximated by the asymptotic form (17.16) K O” Yp — 0852/h, with K = 0.367ka2yP., . In a Kerr liquid with n, ~ 107" esu, we find P,, ~ 8 kW for a laser beam with X ~ 5000 A. The focal length z, is proportional to the square of the initial beam radius a3. For ag = 500 am and P/P., = 1.5, we have 2, 31 cm. The on-axis beam intensity as a function of z can also be calculated, and can be approximated by i(z) _ _ Z 2)-a/2 700) = f (2) (17.17) where a is a parameter depending on P as seen in Fig, 17.5. In the quasi-steady-state case, the field amplitude #, and hence J and P, are also functions of { = 1 — z/v,. Then (17.16) and (17.17) become _ K a(t) = WA ~ 080 ~ 08am (17.18) and (2,8) [, fz be? ep" (2) | . (17.19) An immediate consequence of this time-dependent solution is that the focal spot position, given by z,, should vary with time.” This moving focus picture describes the observed results of self-focusing of nanosecond laser pulses in liquids very well, as we shall see in the following section. 173° QUASI-STEADY-STATE SELF-FOCUSING For self-focusing of g-switched laser pulses in quids, having the pulsewidth (~ 10 nsec) much longer than the response time of the medium (~ 10 psec), the preceding discussion of quasi-steady-state self-focusing should apply. Being a strongly nonlinear effect, self-focusing depends critically on the input beam characteristics. Weak ripples on the otherwise smooth transverse profile may get strongly amplified in the self-focusing process and lead to the breaking of the beam into several independently self-focused sections. With n =n) + na|E|?, the critical size A below which a beam with intensity J becomes unstable against the transverse intensity variation can be estimated from the314 Self-Focusing expression of P,,.!5 1a 1/2 (2) = 1B (17.20) at (32an,1)? For 2, ~ 107" esu, J ~ 50 MW/em’, and A ~ 5000A, we find A ~ 10 am. The use of laser beams with relatively poor mode quality has led to the observation of beam break-up and multiple filaments. The results then become very difficult to interpret. To compare experiment with theory, therefore, single-mode lasers must be used. We consider here only self-focusing of single-mode laser pulses. Self-Focusing in the Prefocal Region Equation (17.19) describes self-focusing in the prefocal region in a medium with 1 = 9 + n,)£|. It has been confirmed experimentally by measuring the peak intensity on the beam axis as a function of z.'* The shrinkage of the beam tadius due to self-focusing has also been observed (Fig, 17.3). The polarization dependence of self-focusing is very interesting. It has been found that the output from the focal region is always linearly polarized irrespective of the input polarization. For a circularly polarized input beam, the direction of the output polarization is random. This can be understood as the result of nonlinear coupling between the two circularly polarized field components via the field-induced refractive indices in the medium. 1” As shown in (16.31), the field-induced refractive indices for the two circularly polarized fields can be written as An,= (#\lae.P + BIE.) (27.21) For ordinary liquids, A — B = —2xf}, <0. Therefore, if both circularly polarized fields are present, the weaker one sces a larger An and hence self-focuses more readily until its intensity equals that of the other component. The output then becomes linearly polarized. The above argument applies even to the case of a circularly polarized input beam, since, in practice, the beam can never be perfectly circularly polarized. The quantitative analysis of this self-focusing problem with mode coupling, however, has not yet been worked out. Filaments and Moying Foci The self-focusing threshold for a given medium of length / is determined by the condition z;(P,4x) = 4, where P,,,, is the peak power of the input pulse. Equation (17.18) describing z, as a function of P.,., has been experimentally confirmed by measuring the self-focusing threshold powers at different /.'8‘Quasi-Steady-Stute Self: Focusing 318 We now consider what happens when the input peak power is above the self-focusing threshold, In early experiments, it was found that after the beam self-fecused it broke inte a number of intense thin filaments.* > The multiple filaments were the result of the multimode structure in the input beam, as Mentioned earlier. It was shown later that a single-mode input laser actually sesulted in a single filament on the beam axis. For a given medium, the filament had a diameter constant to within + 20% and lasted over a distance of a few centimerers. The intensity of the filament could be as high as a few tens of gigawatts per square centimeter. From the picture of quasi-steady-state self- focusing, one can perhaps realize that the filament may correspond to the track of the moving focal spot as it appears on a time-integrated photograph. That this is indeed the case has been confirmed by motion pictures taken with streak camera.” The diameter of the filament then corresponds to the diameter of the focal spot, and the filament intensity to the intensity in the focal region. ‘We can obtain a better perspective of the moving-focus picture from Fig, 17.6. The upper U curve describes the position of the focal spot as a function of time. It is constructed from the input pulse P(r) using (17.18) and assuming K and P,, are known.” In practice, K and P, can be determined from the measurements of 2-(P,,,) versus P,,,,.'* Experimental determination of the U curve (at least partially) is also possible from some sort of time-of-flight measurements of the moving focus.”' It has been found that the measured curve agrees very well with the one calculated from (17.18), As seen from Fig. 17.6, the U curve has the following characteristics. If the length of the medium dis sufficiently long, then the focal spot first appears at z, inside the medium. It then splits into two: one moves backward and then forward after it reaches the minimum self-focusing, distance zy(P,,,..) corresponding to the peak of the input puise; the other moves forward with a velocity faster than light. Both branches of the U curve have their slopes approach light velocity as z > oo. Note that the faster-than-light feature does not violate the special theory of relativity because the focal spots appearing at different times actually come irom self-focusing of different parts of the input pulse, and therefore the “motion” of the focal spots does not transmit anything real. However, a strong polarization induced by the moving focal spot still appears in the medium and | cai have an apparent velocity faster than the light velocity. This is similar to the case of Cerenkov radiation, but the problem has not yet been worked out, : The unusual characteristics of the U curve for the moving focus lead to a number of interesting resuits.? First, the focal spot should spend a relatively long time at 2,(P,,,.), and hence optical damage is more likely at 2/( Pya.)- Indeed, in transparent liquids, laser-induced bubbles have been observed in this region. Second, when z,(P,,,,) is appreciably smaller than /, the light pulse diffracted fromm the filament within a few centimeters at the end of the cell has a very short pulsewidth, less than 100 psec for a nanosecond input pulse. t Third, the high laser intensity (~ 10 GW/cn?) in the focal region readily initiates other nonlinear optical processes. One is a strong phase modulation316 Seli-Focusing Distance in medium, Z {em} Input power, P(t) ° 7 2 3 4 5 Time, t {nsec} Fig. 17.6 Lower trace describes input power P(t) as @ function of time s, Peak power is 42.5 kW and the half-width at the 1/e point is 1 nsec. Upper trace calculated from (17.18) describes the position of the focal spot as a function of time. Values of (0.852), and K used are 8 kW and 11.6 cm (kW)'”, respectively, which corresponds roughly to an input beam of 400 ym in diameter propagating in €S,. The dotted lines, with the slope equal to the light velocity, indicate how light propogates in the medium along the z-axis at various times. (After Ref. 22.) and a resultant spectral broadening on the light diffracted from the filament region due to the large field-induced refractive index change An. We discuss the problem in more detail in a later section. Another is the initiation of stimulated Raman and Brillouin scattering, The stimulated scattering, in turn, drastically affects self-focusing. The interplay between the two, which is most intriguing and interesting, is discussed below. ‘The sharp stimulated Raman threshold in Kerr liquids (those in which Az is dominated by molecular reorientation) was a problem that attracted a great deat of attention in the carly development of nonlinear optics (See Section 40.6). We now understand that it results from self-focusing, The extremely high laser intensity in the focal region readily initiates stimulated Raman and Brillouin scattering. The sharp stimulated Raman and Brillouin thresholds should therefore nearly coincide with the self-focusing threshold. The buildupQuasi-Steady-State Self-Focusing M7 of the Raman and Brillouin intensities, however, depends on the characteristics of the two stimulated scattering processes in the particular medium.’*” Stimulated Brillouin scattering has a large steady-state gain but a slow tran- sient response (~ 10 nsec), whereas stimulated Raman scattering has a much smaller steady-state gain but an almost instantancous response (~ 5 pseo).23 As the laser intensity increases upon self-focusing, the Brillouin scattering may or may not appear earlier than the Raman scattering, depending on the medium. When the laser power is well above the self-focusing threshold, the stimu- lated Raman and Brillouin generation can be understood with the help of Fig. 17.7.% The moving focal spot initiates both forward Raman and backward Raman and Brillouin scattering along the U curve, The backward radiation initiated from the lower branch of the U curve intersects with the incoming laser light in the shaded region and gets effectively amplified. Since the Raman scattering has an instantaneous response, it appears first, and its strong amplification soon depletes the incoming laser power to a level below the self-focusing threshold. Termination of self-focusing then stops the Raman radiation, As a result, the backward Raman output appears in the form of an intense subnanosecond pulse, ™ as seen in Fig. 17.8. With the Raman emission fading out, the incoming laser power recovers from depletion and reaches the self-focusing threshold again. The backward Brillouin radiation then initiated can bave a larger transient gain than the Raman scattering. It builds up in intensity and depletes the incoming laser power. Through self-adjustment, the backward Brillouin scattering keeps the transmitted laser power just below the self-focusing threshold. If the transmitted laser power is too high or too low, the Brillouin scattering intensity would increase or decrease accordingly to deplete more or Jess laser power. This explains the observations in Fig. 17.8 that after the sharp dip afflicted by the Raman generation, the transmitted laser pulse shows a depleted flat top, while the sum of the transmitted laser power and the backward Brillouin power is equal to the incoming taser Fig, 17.7 The interaction between backward stimu- lated scattering and incoming laser radiation. Back- ward stimulated Raman and Brillouin radiation, ini- tiated along the upper branch of the U-curve, propa- gates along the dot-dashed tines and interacts with the non-self-focused incoming laser light in the shaded region. [After Y. R. Shen Prog. Quant. Electron. 4,12 t (1975)]ue Self-Focusing | ir oN rE POWER IMW] -o o 25 gs TIME tinsec} Fig. 178 Oscilloscope traces of the incident laser pulse (a), the total stimulated emission in the backward direction (8), the backward stimulated Raman emission alone (7), and the transmitted laser light (8). (After Ref. 24.) power."4 The depletion of the incoming laser power to a level below the self-focusing threshold also terminates the moving focus or filament, Conse- quently, the later portion of the lower branch of the U curve can never be observed. The forward stimulated Raman scattering can also be initiated in the moving focal region. Figure 17.7 shows that its amplification is through interaction with the diffracted laser light after the focal region, and therefore is expected to be much smaller than the backward Raman amplification. This is indeed the experimental observation. As the laser power or the medium length increases further, the focal region becomes longer, and so does the laser-Raman interaction length. As a result, the forward Raman output increases steadily. It can eventually deplete nearly all the laser power in the focal region. Then diffracted Raman, instead of laser light would show up in that focal region, and the laser filament resulting from the moving focus along the upper branch of the U curve would appear effectively terminated.”