660 CHAPTER 16 SemCONoUCTOR oPTICS Each pair of quantum numbers (1, ¢2) is associated with an energy subband that has a density of states o(K) — 1/1 per unit length of the wire and therefore 1/7rdh dp per unit volume. The corresponding quantum-wire density of states (per unit volume), asa function of energy. is (dda) (yme/ v2 xn) elE)- ) VE-E,-Ey— Ea 6, otherwise, 4192 152,8)..-, (161-40) E> Ect Ey + Ep ‘These are decreasing functions of energy, as illustrated in Fig. 16.1-29. Quantum Dots Ina quantum-dot structure, the electrons are narrowly confined in all three directions within a region that we take to be a box of volume di dz. The energy is therefore quantized to ed ee 16.14 fan faye 2me FE (aot / Qa)* FP (asx ds)’ ae iain 41142599 = 15,3). (16.142) > Fa Eqs ‘The allowed energy levels are discrete and well separated so that the density of states, presented hy a sequence of delta functions at the allowed energies, as illustrated in Fig. 16.1-29. Quantum dots are often called artificial atoms (see Sec. 13.1C). Even though they contain enormous numbers of suwongly interacting natural ators, the dis- crete energy levels of the quantum dot can, in principle, be chosen at will by proper design. 16.2 INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS We proceed « cousider suine uf dhe basic optical propestics uf semivouductors, with: an emphasis on the processes of absorption and emission important in the operation of photonic devices. This domain of study is known as semicanductor aptics. A. Photon Interactions in Bulk Semiconductors A number of ui semiconductors. scans cau Iead (u the absorption and emission of photons in bulk ‘The most important of these are: "= Band-t0-Band (Interband) Transitions. An absorbed photon can result in an elec- ton in the valence band making an upward transition to the conduction band, thereby creating an electron-hole pair (Fig. 16.2-1(a)], Electron-hole recombi- nation can result in the emission of a photon. Band-to-band wansitions may be assisted by one or more phonons. A phonon is a quantum of the lattice vibrations associated with molecular or acoustic vibrations of the atoms in a material162. INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS 661 Figure 16.2-1. Examples or absorption and emission of photons in bulk semiconductors. (a) Band- fortand transitions in GaAs can result an the ebsorption or emission of photons of wavelength Ye < Ay = heg/E, = O87 um, (4) The absorption of a photon of wavelength Xa = he./Ex = 14 jum ests ina valence-band to cceptor-level transition in Hg-doped Ge (Ge:He) (c) Free-arrier twansitons within the conduction band of Ge = Impurity-t0-Bund Transitions. Aw abyvibed phuwwu ean result iat & wasition De= tween a donor (or acceptor) level and a band in a doped semiconductor. In a prlype material, for example, a low-energy photon can lift an electron from the valence band to the acceptor lovel, where it becomes trapped by an acceptor atom [Fig. 16.2-1()]. A hiole is created in the valence band and the acceptor atom is ionized. Or a hole may be trapped by an ionized acceptor atom; the result is that the electron decays from its acceptor level to recombine with the hole. The energy ‘may be released radiatively (in the form of an emitted photon) or nonradiatively (Gn the form of phonons). The wansition may also be assisted by traps in defect states, as illustrated in Fig. 16.1-17. 1 Free-Carrier (Intraband) Transitions. An absorbed photon can impart its energy to an electron in a given band, causing it to move higher within that band. An electron in the conduction band, for example, can absorb a photon and move to a higher energy level within the conduction band [Fig. 16.2-I(c)]. This is followed by thermalization, a process whereby the electron relaxes down to the bottom of the conduction band while releasing its energy in the form of phonons. The strength of free-cartier absorption iy proportional to dhe carrier density; i decreases with photon energy as a power-law function, = Phonon Transitions. Long-wavelength photons can release their energy by di- rectly exciting lattice vibrations, i. by ereating phonons. Excltonic Transitions. The absorption of a photon can result in the formation of an exciton. This entity is much like a hydrogen atom in which a hole plays the role of the proton. The hole and electron are bound together by their mutual ‘Coulomb interaction. A photon may be emitted as a result ofthe clectron and hole recombining, thereby annihilating the exciton, ‘These transitions all contribute to the overall absorption coefficient, which is dis- played in Fig. 16.2-2 for Si and GaAs, and at greater magnification in Fig. 16.2-3 for ‘a nuimber of semiconductor materials” For photon energies greater than the bandgap energy Ey, the absorption is dominated by band-to-band transitions that form the basis ‘of many photonic devices. The spectral region where the material changes from being relatively transparent (hv < E,) to strongly absorbing (hiv > E,) is known as the ahcorption edge Ditect-handgap semiconductors havea more ahnpt absorption edge than indirect-bandgap materials, as is apparent from Figs. 16.2-2 and 16.2-3.062 CHAPTER 1® seneconoucTOR OPTICS ayvelength Ao (um) 00 10 19, ol! wrt Ries a & & Figure 16.2.2 Observed optical bsurption coefficient er versus pli- ton energy and wavelength for Stand GaAs in thermal equilibrium at 300° K. The bandsap energy E, is 112 eV for Siand 142 eV for Gaas. Silicon is relatively transparent in the ‘band A, - 1-1 to 12 pm, whereas in : tuuaie GaAs is vclavely tranaparet oor ot 10 10.0 nthe band Ay © 0.87 10 12 pm (see Absorption coeffient fn Photon cron te (9) Fg. 32-1), Wavelength Ao (um) Wse3 2 1s 190908 ooo nos os 3 5 Postion coficet oe) BS a ag tas Photon enery eV) Figure 16.2-3 Absorption coefficient versus photon energy end wavelength for Ge, Si, GaAs, GaN and selected other II-V binary semiconductors at T = S047 K, on an expanded scale, B, Band-to-Band Transitions in Bulk Semiconductors ‘We proceed 1 develop a simple theory of direct band-to-band photon absorption and emission in bulk semiconductors, ignoring the other types of transitions. Bandgap Wavelength Direct hand-to-hand absorption and emission can take place only at frequencies for which the photon energy hv > E,. The minimum frequency v necessary for this to uur is vy = Ey/h, su Ural the eonrexponing mains wavelength is Ag = Co/'¥q — ‘heo/ E,. It the bandgap energy is given in eV (rather than in J), the bandgap wavelength162 INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS 663 ica ey io jm tuens out to be we. (162-1) E; Bandgap Wavelength ‘hg (as) and Ey (@¥) “The quantity A, is known as the bandgap wavelength (or cutoft wavelength). ‘The bandgap wavelength ,, and its associated bandgap energy E,. are provided in ‘Table 16.1-2, and in Figs. 16.1-7 and 16.1-8, for a number of semiconductor materials ‘of importance in photonics. II-V ternary and quatemary semiconductors of different ‘compositions span a substantial range of bandgap wavelengths, from the mid-infrared to the mid-ultraviolet, as is evident in Fig. 16.2-4. stp ts tonne 03 oe oa as ay Atm) Figure 16.2-4 Bandgap wavelength ,, end corresponding bandgap energy Ey, for selected elemental and II-V binary, ternary, and quaternary semiconductor materials. Successive rows, starting atthe top, represent AllnGaN, AIGaN, InGaN, InGaAsP, AllnGaP, InGaP, GaAsP, AlGaAs, InGaAs. and GaAsSb, The shaded regions indicate compositions for which the materials ae drect- bend semiconductors Conditions for Absorption and Emission Electron excitation from the valence to the conduction band may be induced by the absorption of a photon of appropriate energy (J > Eq or \ < 4g). An electron hole pair is generated (Fig. 16.2-5(a)]. This adds to the concentration of mobile charge carriers and inereases the conductivity of the material. The material behaves as a photoconductor with a conductivity proportional to the photon flux. This effect is used to detect light, as discussed in Chapter 18. Electron deexcitation from the conduction to the valence band (electron-hole re- combination) may result in the spontaneous emission of a photon of energy hu > E, [Fig. 16.2-5(6)], or in the stimulated emission of a photon (Fig. 16.2-5(0)}, provided that a photon of energy hy > E, is initially present (see Sec. 13.3). Spontaneous emission is the underlying phenomenon on which the light-emitting diode is based, as will be seen in Sec. 17.1. Stimulated emission is responsible for the operation of semiconductor optical amplifiers and laser diodes, as wall be seen in Sees. 17.2, 17.3, and 17.4, “The conditions under which absorption and emission take place are summarized as follows664 CHAPTER 18. SEMICONDUCTOR OFS wo * o ° Figure 16.2-5 (a) The absorption of @ photon results inthe generation of an electron-hole par. ‘This process is used in the photodetection of light. (3) The ecomhination of an electron-hole pair results in the spontaneous emission of a photon. Light-emitting diodes (LEDs) operate on this basis. (6) Electton-hole recombination can be induced by a photon. The result isthe stimulated emission ‘of an ideutical photon, This isthe undeilying rovess responsible forthe epetation of semiconductor Taser diodes = Conservurion of Energy. The abyuiptivn ot eutission uf « plrton of cueigy tw requires that the energies of the tWo states involved in the interaction (say Ey and E, in the valence band and conduction band, respectively, as depicted in Fig. 16.2 5) be separated by hu. Thue, for photon emicsion to occur by electron hole recombination, for example, an electron occupying an energy level Ep 11ust {interact with a hole occupying an energy level E,, such that energy is conserved, E, E-hw. (16.2.2) = Conservation of Momentum, Momentum must also be conserved in the process of photon emission absorption, so that p» — py = hhvfe = h/d, or ka ~ ky = 2/2. ‘The photon-momentum magnitude h/ is, however, very small in comparison with the range of momentum values that clectrons and holes can assume, The semiconductor E-f diagram extends to values of of the order 2x/a, where the lattice constant « is much smaller than the wavelength A, so that 27/\ < 9 /a The momenta of the electran and the hole participating in the interaction rust therefore be approximately equal. This condition, ky * ky, is called the selection rule. Transitions dat obey this tule are represented in the E-* diagrat (Fig, 162-5) by vertical ins, indicating that the change in kis neizible onthe Seale of the diagram. Energies and Momenta of the Electron and Hole with Which a Photon Interacts As Is apparent from Fig. 16.2-5, conservation of energy and momentum requi that a photon of frequency 1’ interact with electrons and holes of specific energies ‘and momenta determined by the semiconductor E-K relation. Using (16.1-3) and (16.1-4) to approximate this relation for a direct-bandgap semiconductor by two parabolas, and writing E. ~ E, = Eg, (10.2-2) may be written in the form 242 242 66 = EE oe OE iy, (16.2.3) from which w= 2h (tw -€,), (162-4) ¥4162 INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS 665 where Lot,t, 162-5) Substituting (16.2-4) into (16.1-3) provides that the energy levels Ey and Ez with which the photon interacts are B= E+ re (lw - E,) (16.26) A (hy — E,) = Ea ~ hv. (162-7) my Inthe special ease when me = amy, we obtain Ez = Ec-+4(hu— Ey), as required by symmetry. Optical Joint Density of States. We now determine the density of states o(v) with which a photon of energy fu interacts under conditions of energy and momentum conservation in a direct-bandgap semiconductor. This quantity incorporates the density of states in both the conduction and valence bands and is called the ‘optical joint density of states. The one-to-one correspondence between E> and 1» embodied in (16.2-6) permits us to readily relate o(v) to the density of states (Ex) in the conduction band by use of the incremental relation vo(Ez) Ez — lv) dy, from which o(v) = (dE2/dv)g.(E»), so that (162-8) Using (16.1-7) and (16.2.6), we finally obtain the number of states per unit vol- ‘ume per unit frequency: oe) = On iw, he > By (16.2.9) a Optical Joint Density of States which is illustrated in Fig. 16.2-6, The one-to-one correspondence between E sand i (16.2-7), twxetber with yo(E1) finn (16.1-8), results in an eapression for o(v) identical to (16.2-9), eo) Figure 16.2.6 The density of states with whieh t t 4 photon of energy fu’ interacts increascs with . hy = in aceordance with a square-root law. Photon Fmission Is Unlikely in an Indirect-Randgap Semiconductor. Radiative electron hole recombination is unlikely in an indirect bandgap semiconductor. “This is because uausitivus frou near die bowunr of the euuduetion band w neat le top of the valence band (where electrons and holes, respectively, are most likely666 cHaPren 6. sewiconoucron oPnics to reside) requires an exchange of momentum that cannot be accommodated by the emitted photon, Momentum may be conserved, however, by the participation of phonons in the interaction. Phonons can carry relatively large momenta but typically have small energies (~ 0.01-0.1 eV: see Fig. 16.2-2).so their transitions. appear horizontal on the E-k diagram (see Fig. 16.2-7). The net result is that momentum i conserved, but the A-veleetion rule is violated. Because phonon- assisted emission involves the participation of three bodies (electron, photon, and phonon), the probability of its occurrence is quite low. Thus, Si, which is an indirect-bandgap semiconductor, has a substantially lower radiative recombina- tion coefficient than does GaAs, which is a direct-bandgap semiconductor (see ‘Table 16.1-4). Silicon is therefore not an efficient light emitter, whereas GaAs is. Figure 16.2-7 Photon emission in an indirect-handpap semiconducter. The recom ination of an electron near the bottom of the conduction bund with a hole near the top fof the valence bund tequites the exchange ff energy and momentum. The energy may be cared off by @ photon, but one oF ‘more phonons are also required to conserve Hiiianiiiotiiiieiis momentum. This type of multiparice inter. 4 getion is therefore unlikely. we = Photon Absorption Is Not Unlikely in an Indirect-Bandgap Semiconductor. Ak though photon absorption also requires energy and momentum conservation in an inditect-bandgap semiconductor, this is readily achieved by means of a two-step process (Fig. 16.2-8). The cleetronis frst excited to ahigh energy level within the Conduction band by a k-conserving vertical transition. It then quickly relaxes to the bottom of the conduction band by a process called thermalization in which its momentum is transferred to phonons. The generated hole behaves siilaly: Since the process occurs sequentially, it does not require the simultaneous presence Of tntee bodies and is thus not unlikely. Silicon is therefore an efficient photon detector, as is GaAs. Figure 162-0. Fintin sboxton in an Insrstcaragap secon a 4 vr tical (k-conserving) transition, The mn welis an ali ect in Oo oe {ction herd esving behind a hole inthe ‘Stem tand ie slecton and hole ten nor fast tontone to hemes end Wight pone exch nhc canton nd alee bands, respectively, releasing helt nergy inthe foto pens. ine the process is sequential tine C. Absorption, Emission, and Gain in Bulk Semiconductors ‘We now proceed 10 determine the probability densities of a photon of energy fu being emitted or absorbed by a bulk semiconductor material in a direct band-to-band182. INTERACTIONS OF PHOTONS WITHCHARGE CARRIERS — 67 transition. Conservation of energy and momentum, in the form of (16.2-4), (16.2 6), and (16.2-7), determines the energies E; and Es, and the momentum hk, of the electrons and holes with which the photon may interact. Three factors determine these probability densities, as discussed below: 1. Occupaney probabi 2. Transition probabilities 3. Optical joint density of states, Occupancy Probabi ‘The occupancy conditions for photon emission and absorption by means of transitions between the diserete energy levels E and E; are stated as follows: Emission condition: A conduction-band state of energy Ey is filled (with an electron) ‘and a valence-band state of energy Ey is empty (Le, filled with a hole). Absorption condition: A conduction-band state of energy E isempty and a valence band state of energy Er is filled. ‘The probabilities that these occupancy conditions are satisfied for various values of Ey and E, are determined from the appropriate Fermi fonctions f-(E) and fy(E) asco ciated wit the conduction and valence bands of a semiconductor in quasi-équilibium. Thus, the probability J.(v) that the emission condition 1s satistied for @ photon ot energy hh isthe provhict of the prohahilties that the upper state is lled and that the Tower state is empty (these are independent events), £0) = fe(Es) [1 = (Fi) (162-10) The energies Ey and Ep are related tov by (16.2-6) and (16.2-7). Similarly, the probability /,(v) that the absorption condition is satisfied is fol) = [1 f-(Es)| ful E). (162-11) el EXERCISE 16.2-1 Requirement for the Photon Emission Rate to Exceed the Absorption Rate. (@) Tor a bulk semiconductor in thermal equilibrium, show that f(r) is always smaller than Ja(v) so that the rate of exon emission cannex exceed the rate of phaxon absorption, (b) For a semiconductor in quasi-equilibrium (Ey, # Ey). with radiative transitions occuring between a conduction-hand state of energy Ez and a valence-band sate of energy E, with the same value of f, show that emission is more likely than absorption if the separation between the ‘quasi Fermi levels is larger than the photon energy, ie if Eje—E yy > hw. (102-12) ‘Condition tor Net Emission ‘What does this condition imply about the locations of E,- relative to E..and Ey. relative to E.? eee ae668 CHAPTER 16 SEMICONDUCTOR OPTICS Transition Probabilities Satisfying the emission /absorption occupancy condition does not assure that the emission /absorption actually takes place. These processes are governed by the proba- bilstic laws of interaction between photons and atomic systems examined at length in Secs, 13.3A-13.3€ (See also Exercise 13.3-1). AS they relate 10 semiconductors, these laws are generally expressed in terms of emission into (or absorption from) a narrow band of frequencies between v and » + de: ‘Summary A radiative transition between two discrete energy levels E, and Ey is charac terized by a transition cross section o(v) — (X2/Brtsp)g(v), where v isthe fre- ‘quency, fay is the spontaneous lifetime, and g(v) is the lineshape function {which has linewidth Av centered about the transition frequency vo = (Eo ~ Ei)/h ‘and has unity areal. In semiconductors, the radiative electron-hole recombination lifetime 7, which was discussed in Sec. 16.1D, plays the role of fap so that x OW) = Fee 00) (16.2-13) 1 TF the occupancy condition for emission is satisfied, the probahility density (per unit time) for the spontaneous emission of a photon into any of the availble radiation muodes in dhe narrow frequency band between v and +d is Fop(v) du = = alu) dv. (16.2-14) ‘= If the occupancy condition for emission is satisfied and a mean spectral photon-flux density dy (photons per unit time per unit area per unit fre- quency) al frequency 7 is present, the probability density (pet unit tine) for the stimulated emission of one photon into the narrow frequency band hetween v and + di is » Bam Wi() due ~ fy o(u) dv — bu — glu) do. (16.2-15) * If the occupancy condition for absorption is satisfied and a mean spectral photon-Mux density ¢, at frequency is present, ie probability density for the absorption of one photon from the narrow frequency band between v and also given by (16.2-15).. sition has a different central frequency vp, and since we are consid- ering 4 collection of such transitions, we explicitly label the central frequency of the transition by writing g(x) as g.o(). In semiconductors the homogeneously broadened lineshape function g,9(v) associated with a pair of energy levels generally has its origin in electron-phonon collision broadening. It therefore typically exhibits a Lorentzian Tineshape [see (13.3-34) and (13.3-37)] with width Av ~ 1/7 To, where the electron phonon collision time T: is of the order of picoseconds. If Tz ~ 1 ps, for example, then Av — 318 Gllz, corresponding to an energy width hAv ~ 1.3 meV. The radiative lifetime broadening of the levels is negligible in comparison with colttstonal broadening.162 INTERACTIONS OF PHOTONS WITH CHARGE CARRIERS — 669 Overall Emission and Absorption Transition Rates For a pair of energy levels separated by Ez — E, = hr, the rates of spontaneous ‘emission, stimulated emission, and absorption of photons of energy hv’ (in units of photons/s-Hz-cm® of the semiconductor material), at the frequency 1, are obtained as follows: The appropriate transition probability density Fa,(v) ot W(v) las provided in (16-14) o¢ (16 7-15)] is multiplied by the appropriate acenpation probability f.(%) ‘or fa(u0) [as given in (162-10) or (16.2-11)}, and by the density of states that can interact with the photon g(vp) [as set forth in (16.2-9)]. The overall transition rave for all allowed frequencies is then calculated by integrating over vn. ‘The rate of spontaneous emission at frequency v, for example, is given by fe) = f/f) eo) (62:16 ‘When the collision-broadened width Av is substantially less than the width of the product f-(vn)o(vn). which is the usual situation, q.o(v) may be approximated by {8(v — vp), whereupon the transition rate simplifies to Tap() = (1/7,)o(v) fe(1). The rates of stimulated emission and absorption are obtained in a similar fashion, which leads to the following formulas: rapt) = 2 eo) f0) (162.17) Fal) = oe ML) (162-18) ¥ Fal) —befee, EL) Eniscsnand Abeotton Rater “These equations, together with (16.2-9)-{16.2-11), permit the rates of spontaneous, emission, stimulated emission, and absorption arising from direct band-to-band transi- tions (photons/s-Hz-em®) to be calculated in the presence of a mean spectral photon- flux density 4, (photons/s-Hz-cm?). The products o(v)fe(v) and o(v) fo(v) are anal- ‘ogous to the products of the lineshape function and atomic number densities in the upper and lower levels, 9(1u)No and g(1-)Ns, respectively, used in Chapters 13-15 to study emission and absorption in atomie ayatems. ‘The determination of the occupancy probabilities fe(v) and fa(v) requires know!- edge of the quasi-Ferm levels E - and E ,,..Itis via the control of these two parameters, (by the application of an external bias to a p~n junction, for example) that the emission and absorption rates are modified to produce semiconductor photonic devices that carry ‘out different functions, Equation (16.2-17) is the basic result that describes the operi- tion of the light-emitting diode (LED). a semiconductor source based on spontaneous emission (see Sec. 17.1). Equation (16.2-18) is applicable to semiconductor optical amplifiers and laser diodes, which operate on the basis of stimulated emission (sce Sees. 17.2-17.4), Equation (16.2-19) is appropriate for semiconductor detectors that function by means of photon absorption (see Sec. 18.1B). ‘Spontaneous-Emission Spectral intensity in Thermal Equilibrium ‘A semiconductor in thermal equilibrium has only a single Fermi function so that (16.2- 10) becomes fe(v) = f(Ez)[1 — f(E1)}- If the Fermi level lics within the bandgap, away from the band edges by at least several times KT, use may be made of the670 CHAPTER te SeIACONDUCTOR OPTICS ‘exponential approximations to the Fermi functions, {(E») % exp|—(E» — E,)/kT| and 1 ~ f(Ex) = exp[-(E, — E1)/KT], whereupon f.(v) = exp|—-(E2 — E1)/kT], Sel) exw(-Zz)- (162.20) Substituting (16.2-9) for o(v) and (16:2-20) for J(u) into (16.2-17) therefore provides hw E, aa Tal) % Dy fiw Ee exo(- hy > E,, (162-21) where (162-22) is a parameter that increases with temperature at an exponential rate ‘The epontancows emission rate (16.2 21), which io plotted verous ha in Fig. 16.2 9, comprises two factors: a function associated with the density of states that increases as the square-root of iy ~ E,, and an exponentially decreasing function ot hi” — Ey g from the Fermi function. The spontaneous emission rate can be increased by augmenting fe(v). In accordance with (16.2-10), this ean be achieved by purposely ‘causing the material to depart from thermal equilibrium in such a way that (Ea) is made large and f,(E1) is made small. This assures an abundance of both electrons and holes, which is the desired condition for the operation of an LED, as discussed in See. 17.1 16 Figure 16.2-9 Spectral intensity of the direct band-to-band spontaneous emission rate Fy(7)_(photons/<-Hz- com) from a remiconductor in thermal ‘equilibrium, as a function of he The spect has & wrfieyociy eat at b= E,/h and extends over & range fy of frequencies of approximate width 2KT Gain Coefficient in Quasi-Equilibrium ‘The net gain coefficient 7o(1/) corresponding to the rates of stimulated emission and. absorption in (16.2-18) and (16.2-19) is determined by taking a cylinder of unit area and incremental length dz, and assuming that mean spectral photon-fiux density is directed along its axis (see Fig. 14.1-1). If $y(2) and gy(z) + dgo(2) are the mean spectral photon-flux densities entering and leaving the cylinder, respectively, d¢y(2) ‘must be the mean spectral photon-flux density emitted from within the cylinder. The162. INTERACTIONS OF PHOTONS WITHCHARGE CARRIERS 671 incremental number of photons, per unit time per unit frequency per unit area, is simply the number of photons gained, per unit time per unit frequency per unit volume fale) ~ role mpi by dhe eka of he eine ees (2) — alo) — Foy (v)] a. Substituting the rates set forth in (16.2-18) and (16.2-19) leads to oy olv) [falv) — fa(v)} bulz) = ral) bul 2), (16.2-23) (162-24) Gain Coefficient \where the Fermi inversion factor f,(v) Is given by $Salv) = Sulv) ~ falv) = Sel Ea) - f(Es), (16.2-25) 1as may be seen from (16.2 10) and (16.2 11), with E, and B, related to v by (16.2- 6) and (16.2-7). Comparing (16.2-24) with (14.1-4) reveals that g(v) fo(v) in the semiconductor system plays the role ot NVg{v) in the atomic system. Using (16.2-9), the gain coefficient may be east in the form ww) = Di VIW=E, fv), > By 192-264) anys with Di= eeuaie (162-260) ‘The sign and spectral form of the Fermi inversion factor f,(v) are governed by the quasi-Fermi levels Ey. and Ey., which in tur depend on the state of excitation of the carriers in the semiconductor. As shown in Exercise 16.2-1, this factor is positive (corresponding to a population inversion and net gain) only when Eye — Eyy > hv. ‘When the semiconductor is pumped toa sulfciently high level by means of an external source of power, this condition may be satisfied and net gain achieved, as we shall see in Sec. 17.2. This isthe physies underlying the operation of semiconductor optical amplifiers and laser diodes. Absorption Coefficient in Thermal Equilibrium au lay only a single Fer level Ey = Eye = ‘A seuiieguducior in dhesinal eyuil E jy $0 that SAE) = §.(E) = HE) = 1 ele Eye ‘The factor fy(v) ~ fe(E2) — folEs) ~ {(E2) — f(E:) < 0, and therefore the gain coefficient (0) is always negative (since E2 > Ey and (E) decreases monotonically ‘with E]. This is true whatever the location of the Fermi ievel Ey. Thus, a semiconduc- tor in thermal equilibrium, whether it be intrinsic or doped, always attenuates light ‘The attenuation (absorption) coefficient, a(v') — —o(v)sis therefore (162-27) a(x) — Dy fiw —E, [f(E,) — f(E2)] (16.2-28) Avsorption Coefficient672 CHAPTER 16. SEWICONOUCTOROPTICS where Ey and E; are given by (16.2-6) and (16.2-7), respectively, and Dy is (16.2-260). If; lies within the bandgap but away from the band edges by an energy of atleast several times kT, then {(E,) = Land f(E,) ~ 0 so that [f(E,) — f(E2)] = 1. In that ease, the diet band-to-band contribution to the absorption cocficient ix Vieni? em ae : ap v=, (162.29) Equation (16.2-29) is plotted in Fig, 16.2-10 for GaAs, using the following parameters: n= 3.6, me = 0.07 mo, my = 0.50™a, mo = 9.1 X 10 *" kg, a doping level such that 7. 4 ns (this differs from that given in Table 16.1-4 because of the difference in doping level), E, — 1.42 eV, and a temperature such that [f(E,) — f(E:)] ~ 1. As the temperature increases, /(E1) ~ f(Ez) decreases below unity and the absorption ‘coefficient set forth in (16.2-28) is reduced. given by av) = _Wevelenath em) 2 o4 otf boo i ] igure 16.2-10 cateuaes absorp esxict tion coeficient a(v) (em) resulting from direct band-to-band transitions. as funetion of the photon energy Ihe (eV) and wavelengths 3, (i, for GaAs. This curve should be compared 1 1 ‘wih the experiential resul show it "n= ei) Fig 102.3 whieh inctesal tory (0 (om) In accordance with (16,2-29), absorption near the hand edge in a direct-handgap semiconductor should follow the functional form liv — Ey. However, the sharp ‘onset of absorption at hv = E, is an idealization. AS is evident in Fig. 16.23, ditect- hhandgap comicandhctore generally evhibit an exponential absorption tail, known as the Urbach tail, with a characteristic width KT’ that extends slightly into the for- bidden band. This is associated wit thermal and static disorder in the exystal arising, from several factors, including phonon-assisted absorption, randomness in the doping tribution, and variations in material composition. Absorption near the band edge in indirect-bandgap semiconductors (¢,8., Ge, Si, and GaP in Fig. 16.2°3) follows the functional form (fv ~ E,)* tater than the square-root relation applicable for direct-bandgap semiconductors. ee EXERCISE 16.2.2 Wavelength of Maximum Band-to-Band Absorption. Use (16.2-29) to determine the (Iree- space) wavelength 2, at which the absorption coefficient of a semiconductor in thermal equilibrium is maximum, Calculate the value of \, for GaAs, Note that thi result applies only to absorption by direct hand. to-band transitions,