With available GaAs injection lasers it appears practicable to create electron-hole pairs in InP. The major difficulties in principle are that it is not clear that excitons exist at the densities required. The use of the laser source leads to the creation of energetic electron-hole pairs.
With available GaAs injection lasers it appears practicable to create electron-hole pairs in InP. The major difficulties in principle are that it is not clear that excitons exist at the densities required. The use of the laser source leads to the creation of energetic electron-hole pairs.
With available GaAs injection lasers it appears practicable to create electron-hole pairs in InP. The major difficulties in principle are that it is not clear that excitons exist at the densities required. The use of the laser source leads to the creation of energetic electron-hole pairs.
2 By focusing the light of the GaAs laser ("-' 1 W) onto a spot of diameter ",,0.1 mm, light intensities"" 1()4 W / cm 2 are attainable. To summarize, with available GaAs injection lasers it appears practicable to create electron-hole pairs in InP in sufficient density to observe collective effects associated with BE condensation of excitons. Line nar- rowing is expected. The major difficulties in principle are: (1) it is not clear that excitons exist at the densities required (1.3X 10 15 cm- 3 at 2K) and (2) assuming they exist, the effects of interactions between excitons on the condensation are not known. In addition, the use of the laser source leads to the creation of energetic elec- tron-hole pairs, a condition less favorable for the at- tainment of quasi-equilibrium then would be the case if the excitons were created directly (cf. Refs. 3, 6, 8). Sample purity and heating due to the intense SOurce are further practical difficulties. Nevertheless, an experi- mental study of dense electron-hole systems in rela- tively pure InP samples appears worthwhile. Note added in proof. N athan 21 has made preliminary measurements, irradiating an InP sample in liquid He at 1.9K with intensities up to 1 kW/cm 2 from a GaAs laser. He found a broadening of the emission line asso- ciated with "free" exciton decay at low excitation levels, perhaps indicating that the excitons dissociate as discussed above. 21 M. I. Nathan (private communication). JOURNAL OF APPLIED PHYSICS VOLUME 36, NUMBER 8 AUGUST 1965 Saturation Effects in High-Gain Lasers W. W. RIGROD Bell Telephone Laboratories, Incorporated, Murray Hill, New Jersey (Received 17 December 1964) Earlier calculations of the radiation intensity obtainable from lasers with homogeneous line broadening are generalized to include arbitrarily large loss fractions per pass. The conditions for maximum transmitted or internally dissipated power are derived, as well as the axial distribution of radiation within the active medium. The relevance of these calculations to high-gain gas lasers is discussed. INTRODUCTION T HE radiation intensity produced by steady-state lasers (optical masers) has been computed pre- viously,I.2 subject to the restriction of small fractional values of the net loss per pass. With the advent of high- gain lasers which can oscillate even in the presence of large dissipative or coupling losses, however, it is ex- pedient to extend these computations by removing the latter restriction. The analysis presented here treats only the case of homogeneous laser transitions, in which the line shape does not change during gain saturation. The resonator is assumed to support substantially plane electromagnetic waves, with its losses concen- trated near the mirrors. Despite these assumptions, the results should furnish a first-order description of the radiation distribution in most forms of high-gain lasers, when the losses exceed several percent. If we denote the normalized radiation intensities in the +z and -z directions of a laser oscillator by fh=w+/wo and fL=w_/wo, respectively, where Wo is the saturation parameter, the saturated gain coefficient g(z) is related to the unsaturated gain coefficient go by! (1) 1 E. O. Schulz du Bois, Bell System Tech. J. 43, 625 (1964). 2 W. W. Rigrod, J. Appl. Phys. 34, 2602 (1963). As the gain coefficient is isotropic, 1 d{3+ 1 d{3- g(z)=--=--- {3+ dz {3- dz (2) and we see that {3+t3- = const. = C. (3) As shown in Fig. 1, the coordinate z runs between the extremities of the uniform maser medium, which is bounded by the effective reflectances and (4) r2= 1-a2-t 2 at z=L, where a1,a2 are the dissipative losses and tlh the re- spective mirror transmittances. Oscillations will then stabilize at a level which satisfies the boundary conditions: ({32/ (33)r2= ({34/{31)r1 = 1. (5) From (3) we have (6) and therefore (7) 2488 W. W. RIGROD 1--------- 9oL----------+j I I I _ MASER MEDIUM _ - __ 1,13++ ;3-1 ------- ;3. o z- L the output intensity at each mirror is Wt, given by 2wt/wo= [(1-a-r)/ (1-r)][goL+ lnr J (17) = [t/(a+t)J[goL+ln(1-a-t)]. (18) Thus, when a1 = a2, we can always replace an asym- metrical resonator by an equivalent symmetrical reso- nator, given by (15), having the same relation between the total output power and the unsaturated gain goL. In addition, when one mirror is opaque and lossless, say t 1 =0 and r1= 1, the output intensity is given by Wt 12 -=/32t 2 =--[goL+ln(r2)!J Wo a2+t2 (19) corresponding to the output at each of the two identical mirrors of a symmetrical resonator with twice as much gain. Accordingly, we shall confine our attention hence- forth to symmetrical resonators. The optimum output coupling t opt , for which Wt in (18) is maximized, is defined by t/a= [(1-a-t)/ (a+t)J[goL+ln(l-a-t)J (20) FIG. 1. Schematic diagram showing normalized levels of flux glvmg intensity in both directions in an asymmetric laser oscillator; and 2 (Wt)max/WO= t 2 / a(l-a- t), (t= t opt ). (21) symbols employed as in the text. Integrating the gain expression (1) for power flow in the positive direction 1 d/3+ (8) we obtain The same procedure for gain in the negative direction yields goL= In(/34//3a) +/34- (3a-C(1/{34-1/{3a). (10) Adding these two equations, and making use of (5) and (6), we obtain (r1)l {32= [goL+ln(r1r 2)lJ (11) [Crt)l+ (r2);J[1- (r1r2)lJ and {34 is then given by (7). The total output radiation intensity, via both mirrors, is (W2+W4), given by (W2+W4)/WO= {34h+{32t2. When the losses at each mirror are equal, this reduces to W2+W4 l-a- (r1r2); --= [goL+ ln (r1r2) 1]. Wo 1- (r1r2)* For a symmetrical resonator, defined by r= l-a-t, (12) (13) (14) (15) (16) For small loss fractions, (a+t)l, these equations reduce to the approximate solutions previously derived 2 : topt/a= (goL/a)l-l 2 (Wt)max/WO= topt 2 / a. (22) (23)
8 a= 0.001 '0 "," '" 0.005 0.025 0.050 0.075 0.' 4 t OPT 0.2 a 0.3 0.4 0.5 --- EXACT ---- APPROX. 2 0.' 0=-------='20=----4:'::0--------::'6'=-0 40 UNSATURATED GAIN (De) FIG. 2. Optimum coupling coefficient (topt) for maximum oscil- lator output power, as a function of the loss fraction (a) and the unsaturated gain in dB, 4.343 goL. The solid lines represent the exact solution, and the dashed lines the approximate version. SATURATION EFFECTS IN HIGH-GAIN LASERS 2489 The exact and approximate solutions for t oPt are plotted in Fig. 2 for various values of the unsaturated gain in dB, 4.343 gaL, and the loss fraction a. The ap- proximate solution is adequate, for high-gain lasers, up to about a= 1 %. The error in the approximate solution increases rapidly, however, for greater losses, with both a and goL. We see that topt/a decreases rapidly with in- creasing a, such that i-a for a>0.5 at high gain values. The curves of Fig. 3 shows the normalized maxi- mum output intensity obtainable with optimum output coupling, as given by (21), for given values of a and 4.343 gaL. The enhancement of available output power by a decrease in a is greatest for small gain values. At high values of gain and loss (a>0.5), the normalized output (21) approaches (i-a) gaL. (Gain saturation due to amplified spontaneous emission sets an upper limit to the gain per pass in Figs. 2 and 3.) A similar procedure may be used to design a resonator for maximum power dissipation within it. In the sym- metrical resonator, the normalized radiation intensity reflected from each mirror is (24) which can be written in the form: (25) stating that the power generated per pass ({32- (31) equals the power dissipated in a plus that transmitted via t. The duality between the two forms of loss is shown by comparing the expression for the total (normalized) transmitted power 2{32t= [tl (a+t)J[gaL+ In(l-a- t) ] (26)
20 10 8 6 4 1 0.8 0.6 0.4 0.2 0.1 Wl. __ ....L. __ --" ____ ..l..... __ --L ____ .L-__ --'-__ -,-' o 20 40 60 80 100 120 140 UNSATURATED GAIN (DB) FIG. 3. Normalized values of the maximum output flux in- tensity, 2 (W')max/WO, of the homogeneous-line laser oscillator, ob- tainable with optimum output coupling for given values of a and the single-pass gain 4.343 goL(dB). 1.0 ,-----------------:::;::;=--_-:---, UJ 0.8 ..J n. I o 0.6
UJ >
til 0.2 (( 1 2r 2 l+r
o 0.2 0.4 0.6 0.8 1.0 REFLECTANCE, r FIG. 4. Ratio of the flux intensity at the midplane (z=L/2) of a homogeneous-line maser to that at either end (z=O,L), as a function of mirror reflectance. with the normalized form of the dissipated power: 2{32a= [al (a+t)J[gaL+ In(l-a-t)]. (27) For :fixed t, there is an optimum value of a for maximum power dissipation in a; and this power is greatest when t=O. It is curious to note that the variation of total (two- way) radiation intensity within the symmetrical laser cavity is independent of the unsaturated gain. When the gain expression (8) is integrated over half the laser length, we find that goL/2= In ({3+1 (31)+ ({3+ -C/{3+)+ ({32-{31)' (28) At the midplane z=LI2, the flux in both directions is the same, yielding (3+={3-= C/{3+ gaL/2= In({3+/{3-)+ ({32-{31)' Combining this result with (9), we obtain (29) (30) In ({3+/ (31) = t In ({32/ (31) = - t lnr. (31) The ratio of the two-way flux intensities at the midplane z=L/2, to that at the extremities of the laser, z=O and z=L, is then given by ({3++{3-h/2 2 ({3rf32)t 2(r)t ({3+ +{3-)O,L {31+{32 1 +r' (32) i.e., the ratio of the geometric mean of {31 and {32 to their arithmetic mean. This ratio is plotted in Fig. 4 as a function of r, the geometric mean of the mirror re- flectances. (The relatively small variation of total flux intensity with distance in a laser has been noted in- dependently by R. N. Zitter.) The expression is not valid, naturally, in the limit of r=O, as spontaneous emission noise input would then need to be considered. 2490 W. W. RIGROD The maximum variation in the saturated gain co- efficient g(z) along the laser axis is indicated by the ratio which does not differ appreciably from unity for r>t. Gas lasers saturate more or less homogeneously under several conditions: (1) when the half-width of the hole burned in the Gaussian gain profile by mono- chromatic radiation is comparable to that of the transi- tion itself3; (2) when the collision frequency is not small compared to the spontaneous decay rates of the excited states 4 ; and (3) in very long resonators, when the mode spacing is appreciably smaller than the hole widths, resulting in nearly uniform gain saturation over the oscillation bandwidth. One or more of these conditions is usually operative in very high-gain gas lasers, except for oscillations not far from threshold. In the last instance mentioned, the integrated radia- tion intensity of the laser over the transition can be computed, approximately, by assuming the gain co- efficient to saturate homogeneously in each frequency interval within the band. 5 This procedure, employed in Ref. 5 for small-loss lasers, (a+t)l, can be gen- eralized to arbitrarily large loss per pass in the same way as for a true homogeneously broadened laser. In each frequency interval dv under the Gaussian gain profile, the output radiation intensity is given by an expression similar to (18): 21)W,(v) = [tl (a+t)J[go(vo)L exp- [(v-vo)1 .1vcJ2 +In(l-a-t)J, (34) where Wo has been replaced by 1)-\ the saturation parameter per unit bandwidth, Vo is the center fre- quencyof the transition, (35) and .1vD is the Doppler width at half-maximum. The oscillation bandwidth, over which is given by (36) where X=go(vo)LI-ln(l-a-t). (37) Integration of (34) over the oscillation band yields the total multifrequency output intensity WI, as follows: 2W t/ 1)-1 (2.1v ma x) = [tl (a+ t) J X[(goL)+ln(l-a-t)J, (38) where (goL) is the average value of goL over the oscilla- 3 E. I. Gordon, A. D. White, and J. D. Rigden, Proceeding of the Symposium on Optical Masers (Polytechnic Press, Brooklyn, New York, 1963), Vol. 13, p. 309. 4 W. R. Bennett, Jr., Phys. Rev. 126, 580 (1962). 6 A. D. White, E. I. Gordon, and J. D. Rigden, Appl. Phys. Letters 2,91 (1963). tion bandwidth, given by (goL) = go(vo)L(1I"i/2)[erf (lnX)l/ (lnX)!]. (39) Comparison of (38) with (18) shows they are formally identical, except for the replacement of goL by its average value, and of Wo by an equivalent parameter Wo= 21)-1.1v max =1)-![2.1v c (lnX)lJ (40) for the uniformly saturated Gaussian line. The optimum output transmittance topt is found by setting dW t / dt= 0 in (38). After some algebra, we obtain an expression very similar to (20): t/a= [(1-a-t)1 (a+t)J[(goL)+ln(l-a-t)J, t=topt. (41) The maximum output intensity (Wt)max at each mirror is given by 2 (Wt)maxIWo=t2/a(l-a-t), t=topt (42) in the same form as (21). In practice, we are usually able to determine the peak gain go(vo)L and loss a per pass, but do not know the oscillation bandwidth or the corresponding value of (goL). However, as previously noted, the approximate solutions for topt/ a, derived for the condition (a+t)l, are adequate for a <0.01 and gains below about 20 dB. In this region, therefore, the solutions for t opt / a and (Wt)max given in Fig. 5 of Ref. 2 may be used. For higher values of loss and gain, on the other hand, the curves of Fig. 2 show that topt! a changes very slowly with goL, and for very high values of goL ap- proaches the value toptl a"" (1/ a)-1. (43) Thus for gains above about 20 dB and a> 0.01, the peak gain value go(vo)L can be used in place of (goL), with the curves of Fig. 2, to find topt/a to a quite good ap- proximation. If better accuracy is sought, the values so obtained can be used to compute X, and consequently (goL) from (39), and with them to redetermine topt/a. Having found X, we can compute .1v max from (35) and (36), evaluate Wo from (40), and thus find the maxi- mum output intensity via each mirror (WI)max from (42) or the curves of Fig. 3. Alternatively, for very high gain per pass, such that X1, erf (lnX) t "" 1 and (38) reduces to the expression 2W t t 1I"t ___ ,rv -go (vo)L 1)-1 (2.1vc) a+t 2 in which W t = (Wt)max when t=t opt . (44) (45) The writer is indebted to Mrs. C. A. Lambert for performing the numerical computations, and to E. 1. Gordon for valuable comments.