AlInGaAs 0.8- m High-Power Semiconductor Lasers ... - IEEE Xplore

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 41, NO. 9, SEPTEMBER 2005

1105

AlInGaAs 0.8-m High-Power Semiconductor Lasers With Delta-Doped Resonant Tunneling Quantum Wells Michael Yasin, Ilan Samid, Zeev Engelman, and Dan Fekete

Abstract—In this paper, we describe a new structure design for producing low-threshold high-efficiency and high-brightness 0.807- m lasers. In this structure, we incorporate a self-discriminating weak optical confinement asymmetrical waveguide, and an active region based on single or double AlInGaAs quantum well (QW) with Te n-type -doping. Optimized coupling between the -doping layer and the double QW, along with waveguide and doping profile optimization, yields th = 140 A/cm 2 per QW, a 2 close to one. far-field angle of 24 , and Index Terms—Aluminum alloys, indium alloys, n-type delta doping, quantum-well (QW) lasers, semiconductor lasers, strain.

I. INTRODUCTION

intensity distribution along with reduced optical power density and, hence, lower facet heating, which is critical for reducing COMD and spatial hole burning. Increasing the thickness of the waveguide layer can reduce the overlap between the fundamental optical mode and the highly doped cladding layers. This results in lower optical loss due to free-carrier optical absorption and a significant improvement of the external differential efficiency. A further advantage is the smaller vertical divergence, which is desired for fiber coupling and beam collimation. In order to obtain a high-brightness laser two main goals should be achieved: high power and narrow beam divergence. The brightness of a laser is defined as

H

IGH-POWER semiconductor lasers are of a great interest for many applications such as range finders, laser radars, sources for optical pumping of solid-state and fiber lasers, optical data storage, printer systems and more. Many of these applications also require narrow beam divergence and high brightness, which is not available in a conventional single quantum well device. In the design and optimization of high power diode lasers, particular care must be taken in order to avoid device overheating and facet degradation effects due to high optical power density, that are the primary factors limiting the performance and the lifetime of semiconductor lasers operating at high output power. The waveguide of the conventional separate-confinement heterostructure (SCH) laser is designed to maximize the overlap of the fundamental optical mode with the gain in the quantum wells (QWs), resulting in a low-threshold current density. While this design is suitable for lasers operating near threshold, other laser parameters such as external differential efficiency, power conversion efficiency, and the catastrophic optical mirror damage (COMD) limit the maximum output power achievable with a particular device [1]. The optical power density in the device can be reduced by the growth of a suitable epitaxial structure. The self-discriminating asymmetric waveguide [2] (SDAW) and the broad waveguide [3], [14] (BW) lasers represent a new approach in designing high-efficiency high power lasers. Using broad optical waveguides results in a wide optical near-field (NF)

Manuscript received January 9, 2005; revised May 11, 2005. The authors are with the Department of Physics, Solid State Institute, and Advanced Optoelectronics Research Center, Technion-Israel Institute of Technology, Haifa 32 000, Israel (e-mail: [email protected]; [email protected]; [email protected]; ssdan@ techunix.technion.ac.il). Digital Object Identifier 10.1109/JQE.2005.852791

(1) Where is the maximum continuous wave (CW) emitted optical output power, is the emitting area and is the solid angle into which the power is emitted, where , and are the far-field (FF) angles in the fast and slow axes, respectively. We will deal here only with which is determined by the waveguide structures and, hence, by the grown layers. The maximum continuous wave emitted power is given by [4] (2) where is the internal optical power density at the COMD limit that depends on the material of the active region. , the QW width divided by the confinement factor, is the effective vertical beam spot size. is the laser stripe width, and is the front facet reflectivity. When the laser operates in the fundamental Gaussian-like mode, directly scales where is the waist of the normalwith ized Gaussian beam and in the fast axis direction is given by . Since (3) and (4) As a result, for a given active region reducing the beam divergence increases the maximum power, so it is the main controllable parameter that determined the maximum power as well as the brightness.

0018-9197/$20.00 © 2005 IEEE

1106


In this paper, we describe a new structure design for making low-threshold high-efficiency high-brightness 0.807- m lasers. Using the mode selection mechanism of an SDAW waveguide, we designed a new waveguide structure capable of CW operation. In this waveguide, we incorporated an active region based on single or double AlInGaAs QWs. We used a Te n-type doping coupled to the QWs to reduce the threshold current. Beam divergence together with the threshold current, , and the quantum efficiency, are uniquely determined by the structure. Therefore, measuring these parameters under pulse conditions of unpackaged laser bars enabled us to optimize the structure, without the need of the more complex processing steps.

II. SELF-DISCRIMINATING ASYMMETRIC WAVEGUIDE One of the ways of achieving broader NF distribution is decreasing the refractive index difference between the waveguide and the cladding layers. This allows a weakly confined optical mode that penetrates deeper into the cladding thus broadening the NF distribution of the beam. Additional NF broadening can be attained by increasing the thickness of the waveguide region, but at the cost of allowing higher modes, which reduces the beam quality. A self-discriminating weak optical confinement asymmetrical waveguide (SDWAW) was designed in order to achieve narrow beam divergence as well as high discrimination of high-order modes. In symmetric waveguides, optical modes are divided into symmetric (even) and anti-symmetric (odd) modes. Even modes, including the fundamental one, have similar intensities at the center of the waveguide. In an asymmetric waveguide, the modes lose their symmetry properties, and the fundamental mode can achieve its maximum intensity at the position where all other modes have low intensities. Therefore, mode selection can be achieved by correct positioning of the active layer (the QW) within the structure [2]. This selectivity is enabled by much lower optical confinement of higher optical modes within the gain region than that of the fundamental one as well as theirs higher losses. Two SDWAW structures with single- (SQW) and double- (DQW) QWs were designed with a simple method of numerical analysis. This analysis considers only the real part of the index of refraction. Therefore, due to the small index difference between the waveguide and the cladding, a small index variation originated either by the gain, temperature or the growth may lead to large deviation between the calculated and the actual modes. Fig. 1 shows the calculated optical modes (all normalized to equal intensity) of the SDWAW and the Al-In profile for double QW lasers. The active region of the considered structures consists of one As QWs. Their waveguide or two quaternary Al In Ga thickness , that includes both waveguide layers and QWs, was slightly different; the widths of the QWs were 7.5 nm for SQW laser, and 7.5 and 8.5 nm for DQW lasers. The calculated fast axis optical mode effective beam spot size , NF and FF full-width at half-maximum (FWHM) for these lasers are shown in Table I. The structural details are given in Tables II and III.

Fig. 1. Al–In profile of DQW laser with SDWAW waveguide and the calculated amplitudes of the TE modes.

TABLE I CALCULATED WAVEGUIDE AND FUNDAMENTAL MODE PARAMETERS OF SDWAW STRUCTURES

The modal gain of different order modes is proportional to the confinement factor , which depends on the overlap of the optical mode pattern with the gain region of the laser. The relative intensity at the position of the QWs is lower (see Fig. 1) and, hence, the confinement factor is also smaller for high-order modes, than that of the fundamental mode. However, high material gain and larger reflectivity of higher order modes can lead to lasing of high-order even modes. The odd modes at properly designed asymmetric waveguides have negligible overlap with the QWs. Additional discrimination of high-order modes is possible by appropriate design of the doping profile of the cladding. This decreases the free carrier absorption of the fundamental mode and increases the absorption of the high-order modes. The nominal doping levels in the cladding were stepped down toward the active region to ensure lower losses for the fundamental optical mode. Both the n-side and the p-side of the cladding were divided into two regions of high and low doping. The borders of these regions are marked schematically by dashed lines in Fig. 1. and ) modes One can see that the high-order even ( spread more and have their global maxima inside the heavily doped regions of the cladding and, hence, their absorption is much higher than that of the fundamental mode. The overlap of the fundamental mode with the lower absorption n-type part of the waveguide is larger than that with the p-type part. As a result of these measures, the discrimination of high-order modes is very efficient.

YASIN et al.: AlInGaAS 0.8 m HIGH-POWER SEMICONDUCTOR LASERS

TABLE II STRUCTURE OF DQW SDWAW LASER

TABLE III STRUCTURE OF SQW SDWAW LASER

1107

to modified band structure and reduced heavy hole effective mass. Due to dislocation pinning in In-containing compounds, strained layer InGaAs–AlGaAs QW lasers have improved reliability over GaAs–AlGaAs lasers [1]. Therefore, the strained QW of GaAlInAs was utilized. The utilization of quaternary composition is rather involved, since a specific wavelength can be obtained by several combinations of variables such as QW width and the quaternary component. On the other hand the advantages provided by strained QWs can be obtained. In order to exploit this advantage, a higher indium and aluminum mole fractions are needed in order to work at the wavelength of 0.807 m, up to the point where the higher aluminum contents degrades the quality. The proper selection of parameters for achieving low threshold current density was obtained from a recent study of the metal–organic chemical vapor deposition (MOCVD) growth conditions of quaternary AlInGaAs QWs, in which the dependence of the threshold current density on the quaternary composition and QW size of quaternary AlInGaAs QW lasers was determined [6]. The modal gain decrease associated with a weak optical confinement can be circumvented by using multiple QW (MQW) structures in the active region. This increases the overlap of the optical mode with the gain region, and consequently the modal gain, almost without altering the vertical spot size. As was recently demonstrated, the increased transparency current density , associated with MQW lasers can be, at least in part, compensated by the use of n-type -doping [7]. The dependence of the material gain of QW lasers on the injection current density is [8], [15] (5) where is the transparency current density and is the gain coefficient. This semilogarithmic expression that accurately approximates gain-current density curves of QW lasers was derived both by fitting a curve to calculated values and an approximate analytical evaluation of overlap integrals that compromised the gain model for a single transition level in and are the nominal transparency current density [9]. and the gain coefficient from the small gain expansion that is linear with the excess current. is equal to the At the lasing threshold, the modal gain and the mirror loss . The sum of the intrinsic absorption depends on the cavity length and the back and mirror loss and . Thus, the threshold the front mirror reflectivities current density is given by

III. REDUCING THE THRESHOLD CURRENT DENSITY OF AlInGaAs–AlGaAs QW LASERS WITH Te n-TYPE -DOPING The high-brightness 0.807- m lasers are based on an AlInGaAs–AlGaAs material system. Strained QWs account for lower threshold current densities and higher efficiencies due

(6) This equation was derived in [9] for noninteracting MQWs. However, in the case of two coupled QWs with -doping, (6) versus [7]. Here also gives the correct dependence of is the average transparency current density per QW, is the average gain-current density coefficient per QW, is the average confinement factor per QW, and is the number of the and the differential gain QWs employed in the active layer. (where is the average QW width) are independent of device geometry and can be used to compare the quality of

1108


different laser quantum structures. Since we defined as the average transparency current density per QW, in the case of , although the transparency current density DQWs of two different QWs is different. It was recently demonstrated that by positioning Te n-type -doped layer near two coupled QWs, a low transparency current is obtained for a strained AlInGaAs–GaAs double QW laser designed to emit light at 0.807 m. The coupling between the three elements affects mainly while doesn’t depend on the number of QWs or the coupling [7]. The effect on is in part due to the higher initial electron population in the QWs and in part due to the enhanced coupling between the two QWs. This enhanced coupling results from the improved overlap of the confined fundamental electronic levels in the two QWs during current injection. In order to enable electron transfer by resonant tunneling from the first to the second QW, the electron level of the second should be deeper than that of the first, by an amount that depends on the injection current level at threshold. This can be obtained by a combination of increasing of In concentration, lowering Al concentration and/or increasing the width of the second QW. Since the band bending decreases, when the amount of injected carrier increases, a shallower second well is needed, when the threshold current is higher [7]. To demonstrate qualitatively the nature of the engagement of n-type -doping with two AlInGaAs QWs, the energy levels, and wave functions of the confined conduction band electrons were calculated [10] for the active region of the lasers described in this research. The structures of such typical regions consist of Al Ga As barriers and two coupled QWs. The n-type -doping is located in a 5.5-nm-thick Al Ga As layer, which is positioned at a distance of 8.5 nm from the first QW. The distance between the -doped layer and the first 7.5-nm-thick Al In Ga As QW is 10 nm and the two QWs are situated 5 nm from each other, a distance, which allows tunneling between the wells. The second In Al Ga As QW is 8.5 nm thick. This calculation shows that without current injection, the ground electron level of the second QW is higher than that of the first well, due to band bending [Fig. 2(a)]. Fig. 2(b) shows that carriers cm an injected carrier density of reduces the band bending and lowers the ground electron level in the second QW, so that the confined electronic levels in the two QWs and the -doped layer overlap. As seen from Fig. 2, the main effect of the injected carrier is on the coupling between the two QWs, while the injected carriers have a negligible affect on the coupling between the -doped layer and the first QW. When the energy levels of the three wells (formed by the -doped layer and the two InAlGaAs QWs) coincide, the wave function splits between the QWs and the -doped layer. Thus, the probability of the electrons to tunnel from the -doped layer straight into the ground energy levels of the two QWs is high and therefore consistently increases the initial carrier concentration in the QWs. This results in the decrease of [7]. IV. EXPERIMENTAL The structures were grown by low-pressure MOCVD on n-type (100) GaAs substrates. The high indirect bandgap AlGaAs waveguide and cladding layers provide optical confinement and electrical carrier injection into the intrinsic diode

Fig. 2. Electronic band structure, energy levels, and wave functions of n-type -doping located in the structure consisted from 1.25310 cm 5.5-nm Al Ga As layer 10 nm apart from two coupled QWs boarded by Al Ga As barriers. The distance between the wells is 5 nm; the first QW is a 7.5-nm-thick In Al Ga As and the second is a 8.5-nm-thick Al Ga As. In (b), the calculations were done with injected carrier In = 0:75 1 10 carriers=cm . density of n

junction region. The advantages of such layers with high content are in lower hole leakage current, higher Al thermal conductivity and smaller refractive indices. Linear graded refractive index layers (GRIN) and high doping in the cladding near the contact layer and buffer layer were used to improve carrier injection. The thickness and doping of the waveguide and cladding layers were designed to provide large absorption losses for high-order modes, while keeping loss low for the fundamental mode. The waveguide layers of SQW and DQW SDWAW lasers were undoped. The doping level was low [(2-4) 10 cm in the cladding regions near the optical confinement layers and as high as 1 10 cm near the contact and buffer layers. The growth conditions for growing the active region are described in [6]. The -doped layer was grown by exposing the bare surface to diethyl telluride and AsH with a prior and posterior purge of H and AsH [11]. In order to assess the sheet density in the -doped layer we used Van Der Pauw measurements (at room temperature and at 77 K) of the mobility and carrier concentration of a test structure with a -doped layer that was grown in the same manner as the lasers, but with an InGaAs QW replacing the AlInGaAs QW of the laser. Because of the limited achievable accuracy of the measurements of the -doping concentration due to unintentional doping in the surrounding layers and the lack of the knowledge about the accurate distribution of the donors in the -doping region, the application of these measurements in order to realize the calculated structures is extremely difficult. Therefore, we examined experimentally the effect of the -doping concentration and of the depth of the ground electron level of the second QW on the threshold current of the different structures. In each structure, a single n-type (Te) -doped layer was positioned at a distance of 100 Å from the QW. The nominal doping concentration was varied for


1109

different structures. The active and spacer regions of the lasers were undoped. The laser structures used in these experiments are described in Tables II and III. The actual doping of the layers may differ from the nominal doping displayed in the tables due to different growth tempertures. This is the case for the buffer and the contact layers, which were grown at a low temperature. The actual carrier concentration as measured by CV profiler exceeds 1 10 cm . Broad-area lasers with stripe widths of 100 m were realized by standard photolithography, using deep wet chemical etching of the epilayers down to the substrate in order to isolate the stripes optically and electrically from each other. V. STATIC ANALYSIS OF BROAD AREA LASERS Light power output versus current density curves of individual uncoated broad-area lasers probe-tested at the bar level with different cavity lengths were measured using 500 nsec current pulses at a 1-kHz repetition rate to obtain the laser parameter used in (7). For 600- m long lasers, the FF intensity distributions were analyzed at different output power levels and for the threshold current was the characteristic temperature determined over the temperature range of 20 C to 100 C. By plotting the inverse differential efficiency versus the internal efficiency and the intrinsic losses were determined according to the equation (7) was determined from the intersection of the linear fit with axis extrapolated for . is governed by varthe ious kinds of leakage currents and nonradiative recombination processes. Using the values for reflectivity of the cleaved facets , was calculated utilizing the slope of reflect the quality of the the fitted line. The extracted and structure. From the relation of and given in (6), and were extracted using the calculated value of , and the internal parameters, which were found using (7). We first studied the effect of the sheet carrier concentration in the -doped layer on DQW SDWAW structures with a 7.5-nm-thick Al In Ga As first QW and 8.5-nm-thick In Al Ga As second QW. The threshold current density versus inverse cavity length of the broad-area double QW in the lasers with sheet concentrations of 0–3.4 10 cm -doped layer is shown in Fig. 3. The lowest threshold current is obtained for a sheet carrier concentration of 1.4 10 cm . The left columns of Table IV summarize the basic parameters of these lasers. Next, using this value of sheet concentration, we changed the depth of the ground electron level of the second QW by changing its width. We examined 8.0, 8.5, and 9.5-nm-thick Al In Ga As QWs. Threshold current density versus inverse cavity length of the broad-area DQW SDWAW lasers with a sheet concentration of 1.4 10 cm in the -doped layer, is shown in Fig. 4. The right three columns of Table IV summarize the basic parameters of these lasers.

Fig. 3. Threshold current density versus inverse cavity length of the broad-area DQW SDWAW lasers with different doping sheet concentrations, the solid curves are the best fits of the data points to equation (6).

Due to the lower and thus lower modal gain, as compared to ordinary SCH lasers, the injected current at threshold of these new structures should be higher. The higher injected current density at threshold reduces the band bending in comparison to SCH lasers with the same carrier concentration at the -doped layer so a shallower second QW is needed for a DQW SDWAW in order to assure that the confined fundamental electronic levels is obtained for a in the two QWs coincide. Indeed the lowest 8.5-nm-thick second QWs as is seen in Fig. 4, while the lowest is obtained for a 9.5-nm-thick second QW in the case of a SCH laser with an identical active region [7]. Thus, the optimal coupling between the two QWs was obtained for structure M333 with QW widths of 7.5 and 8.5 nm, and highest (see Table IV). The resulting in the lowest 24 , average measured FF obtained for this structure was A/cm while the lowest threshold current density or A/cm per QW (Fig. 3) is only twice the value obtained for a double QW SCH laser [7], where all other parameters are comparable. The effect of the -doping on the different structures is demonstrated in Fig. 5, where the threshold current density versus cavity length for structures with and without -doped layer (identical growth sequence where only the exposure of the bare surface to diethyl telluride was omitted) is plotted. Table V summarizes the basic parameters of these lasers. As can be seen from Fig. 5, the effect of the -doping on SQW SDWAW lasers is rather small and under certain conditions it increases the threshold current of short cavity length lasers due to the reduction of . This can be explained from (6), where a reduced value of (in the case of a short cavity and SQW lasers that incorporate -doping) may increase the exponent term more than the reduction it causes when multiplying the exponent, thus resulting in higher threshold current. In contrast, the effect of the -doping on resonant tunneling DQW laser is extremely strong. Due to the tunneling and higher modal gain the DQW laser have lower threshold current density than SQW. The data shown in Tables IV and V indicates that both SQW and DQW structures have narrow FF distributions, due to the weak optical confinement. The measured vertical FF distributions of the laser beams are 22 and 25.5 for M216 SQW

1110


TABLE IV BASIC PARAMETERS OF DQW SDWAW LASERS WITH 7.5-nm-THICK FIRST Al In Ga As QW AND DIFFERENT SECOND QW THICKNESS AND SHEET CONCENTRATION IN THE -DOPED LAYER

Fig. 4. Threshold current density versus inverse cavity length of the broad-area DQW SDWAW lasers with different second QW width and doping with sheet concentration of 1.4310 cm , the solid curves are the best fits of the data points to equation (6).

SDWAW and M214 DQW SDWAW structures, respectively. These values agree quite well with the calculated values for the passive waveguides, considering that the actual waveguides may differ from the nominal waveguides due to nonuniformity of the growth (both composition and thicknesses) and growth reproducibility. The fact, that the differential gain and internal quantum efficiency in Tables IV and V have almost the same values for the structures with similar active regions, suggests that the difference in the performance of these structures indeed results from the waveguide structure and not from other causes associated with the growth. The values of the characteristic temperature, the gain coefficients and the transparency current densities obtained for the optimized DQW SDWAW structure are similar to that of the standard GRIN-SCH structure M186, but the threshold current is almost doubled due to the larger value of . On the other hand, the FF divergences of the new structures are reduced by 50%. As can be seen from Tables IV and V, the internal loss does not depend on the number of QWs and the -doping. This indicates that internal loss in the present structures mostly originates from the highly doped cladding, which is similar for all

Fig. 5. J versus cavity length of the different structures with and without -doping, the dashed curves are only to guide the eye.

structures rather than being due to the scattering at interfaces. Thus, stepping down the carrier concentration in the cladding toward the active region plays an important role in reducing considerably. Since for SCH structures with the same acvalues in the range 3–10 cm [6], [11] tive region we get we assume that the low doping levels and consequently low free carrier absorption of AlGaAs waveguide layers has a negligible contribution to the internal loss in the structures discussed herein. The losses and the confinement factor are the most important in QW lasers. The parameters determining the threshold gain confinement factor depends on the waveguide structure, as well as the number of QWs. Therefore, because of the smaller conof an SQW is lower than finement factor, the modal gain that of DQWs. In the short cavity uncoated lasers with relatively the mirror losses that depend low intrinsic absorption on the cavity length and facet reflectivity dominate. Hence, the threshold current density of short cavity SQW lasers with low confinement factors is high in comparison with a similar DQW lasers. Indeed as shown in Fig. 5, the increase of threshold current for short laser cavity lengths was observed to be stronger for SQW than for DQW lasers. For SQW lasers which incorporate -doping, the lower transparency current density leads to high


1111

TABLE V EFFECT OF -DOPING ON BASIC PARAMETERS OF SQW AND DQW SDWAW LASERS

values of the threshold current density for resonator lengths of 0.5 mm and shorter. In contrast, for DQW lasers which incorporate -doping, the lower transparency current density and higher lead to lower values of the threshold current density for any of the considered resonator lengths. for all structures is more than 18 smaller than that The FF of the standard GRIN-SCH structure (M186). The low internal losses and the small transparency current density indicate that both SQW and DQW SDWAW structures are suitable to produce high-power laser diodes (LDs). VI. BEAM QUALITY For laser applications, detailed information about the beam properties at high power level is required. The beam quality of the SDWAW laser diode was determined by analysis of the NF and FF beam distributions, and by consequent calculation of the vertical beam quality diffraction-limited factor [12]. All beam measurements were performed at identical pulsed drive current and temperature conditions as in the previous section. The measured NF intensity distributions of a 100- m stripe with unmounted laser diodes M214 at 20 C ambient temperature are shown in Fig. 6. The spatial resolution is limited to about 0.1 m. The corresponding calculated and measured intensity distributions in the FF of the laser are shown in the Fig. 7. The vertical FF was measured for diodes at injected currents of 1, 1.5, and 2.0 A. The FF along the fast axis for the SDWAW structures was 22 and 25.5 (FWHM), respectively. The behavior of the beam distribution parameters is similar for all the structures. The measured beam shapes are Gaussian-like (see Figs. 6 and 7). No signs of higher order modes were observed, even at high output power, and beam characteristics are very stable to variations in the output power. Both the aperture width and divergence angle depend mainly on the waveguide structure. There is the complicated influence of the interplay of the epitaxial layer structure, optical confinement, refractive index thermal changes, gain and losses upon the beam quality of laser diodes. In addition, the calculation considered only the real part of the index of refraction.

Fig. 6. Measured near field distributions of the DQW SDWAW lasers together with Gaussian fit values =2 for different powers.

Thus, good agreement between the measured lasing modes and the calculated waveguide modes is obtained with these assumptions. The fast axis beam quality of the diode lasers is proportional to the product of the vertical aperture width and vertical divergence angle of the emitted laser beam (8) where is the measured lasing wavelength, is the refractive index of air, and are the average values of the measured NF and FF. is the full width of the Gaussian curve fit at of its maximum. was evaluated in the vertical directions of the beam intensity profiles. As shown in Table VI, calculated values are higher for the DQWs that have higher gain. Nevertheless, with the measure, the beam quality factors in Table VI ment accuracy are very close to the desired value of 1. Hence, all considered structures lased only in the fundamental vertical mode (that is not always the case for high-power lasers see [13]). Therefore,

1112


Fig. 7. (a) Calculated and (b) measured far field distributions of the DQW SDWAW lasers together with Gaussian fit =2.

TABLE VI BEAM QUALITY PARAMETERS

the output of these SQW and DQW SDWAW lasers is diffraction-limited in the fast axis direction. From (3), the solid angle into which the power is emitted, entirely defines the brightness of the COMD limited laser output. Minimizing the laser fast axis FF divergence also minimizes . Consequently, the newly developed laser structures with extremely narrow vertical FF distributions generate maximum fast axis brightness. VII. HIGH-POWER QUASI-CW CHARACTERISTICS The optical output power is almost linearly proportional to the injected current above the threshold up to the level where the output power saturates and decreases.

To obtain a high quasi-CW (QCW) output power, laser bars with a filling density of ten 100- m–wide stripe lasers per 0.5-cm wafer width were fabricated. An insulator Si N was sputtered outside of the mesa stripe. Ti–Pt–Au metal contacts were evaporated on the p-side. The wafer was thinned to approximately 120 m to allow proper cleaving of the laser mirrors and Ni–AuGe–Ni–Au was used as backside n-metallization. After alloying of the contact at 380 C, the wafers were cleaved along crystallographic planes into bars. For soldering, a thick electroplated Au layer was deposited on the evaporated metallization. Because of high injected current, removing heat from the semiconductor lasers operating CW or QCW at high output power is necessary to prevent device overheating, reduction in power and facet damage. Laser bars containing 72 single emitters with a resonator length of 600 m were fabricated from all wafers, and they were mounted junction down on a cooled Cu heat spreader with Au submount with AuSn solder. The front and back side facets were antireflection (AR) and high-reflection (HR) coated, respectively. All the laser bars considered in this section, including reference bars, were fabricated together in the same run and mounted and tested in the same manner and conditions. The devices were tested under QCW conditions with a 200- s pulse duration at a 25-Hz repletion rate and at a temperature of 22 C. Standard commercial AlGaAs–GaAs GRIN-SCH wafers which contained 72 single emitters, were used as a reference. Such a structure is conventional for 0.8 m wavelength high brightness applications, but has a relatively low threshold for COMD. The tested standard devices lased at a wavelength of 800 nm and produced 83 W at 85 A corresponding to 1.2 W per emitter. The growth uniformity of the new structures was inferior to the standard commercial wafer. Therefore, only a portion that contained approximately 12 emitters was cleaved and mounted in order to reduce the possibility that one emitter is not functioning. The number of operating emitters was confirmed by the NF pattern. For structure M214 only four emitters were lasing. The QCW power-current-voltage characteristics of bars from wafers M214 and M216 are shown in Fig. 8. The maximum power level was 20 W from the DQW SDWAW structure (M216) at 60 A. The devices show a slope efficiency of 0.6 W/A for both M214 and M216and series resistance of 0.03 . Due to the lower and and higher and , it is expected that the slope efficiency of the laser bars based on SDWAW structures will be less than that of the standard GRIN-SCH structure laser. The expected reduction is improved by the enhanced coupling between the two QWs in structure M333. All SDWAW structures achieved greater maximum output power per single emitter than the GRIN-SCH. Wafers M214 and M216 yielded maximum power per emitter of 2.5 and 1.7 W, respectively. For M214, this is twice the maximum power per emitter than that obtained for the reference GRIN-SCH (and 1.4 times higher than for M216). Equation (4) relates the maximum to FF. The predicted maximum power increase power over the GRIN-SCH structure based on the experimental values of are 2.1 for M216 and 1.8 for M214. This discrepancy


1113

QW layers, the emission wavelength decreased while still maintaining desired improvements in device properties. Excellent broad-area lasers have been achieved due to: strained quaternary AlInGaAs QW; optimized coupling between the -doped layer and the QWs; waveguide optimization; and doping profile optimization. Structure (M333), that consists of a -doping with optimum coupling to double InGaAlAs QW as an active layer embedded in a SDWAW waveguide with weak optical confinement, yields FF of 24 , close to one, and A/cm per QW, that is twice obtained for standard SQW laser structure with FF of 46 . The brightness of the new structures is about double that of the GRIN-SCH structure used for comparison. ACKNOWLEDGMENT The authors would like to thank J. Ben-Joseph, D. Gur, and A. Philosoph for their dedicated technical contribution during the research. REFERENCES

Fig. 8. Output power characteristics from the front mirror and the current-voltage characteristics of high power diode laser bars: (a) M214: 0.61 W/A; = 812 nm; R = 0:03 four lasing emitters operating at maximum power of 2.5 W per emitter. (b) M216: 0.58 W/A, = 789 nm, R = 0:032 , 12 lasing emitters operating at maximum power of 1.7 W per emitter.

between the actual and the predicted values of the maximum obtainable power suggests that pulse measurements can only roughly assess the CW potential of diode lasers. There are two different mechanisms limiting maximum light output power. One is catastrophic optical mirror damage and the other is power saturation due to thermal effects (including leakage current). A large spot size combined with good slope efficiency permitted the device to achieve high output power before any sign of COMD. The DQW structure performed better than the SQW at high temperatures. Hence, improving the coupling between the two QWs, as was done in the structure M333, yields higher and . Optimization of the doping profile will further reduce and , so the potential maximum output power predicted from the relation in (4) may be achieved.

VIII. CONCLUSION AlInGaAs–GaAs strained QW broad-area laser bars with different waveguide and active regions emitting near 0.8 m were processed and characterized. The structures were grown using MOCVD. By incorporating aluminum into the strained InGaAs

[1] S. L. Yellen, R. G. Waters, A. H. Shepard, J. A. Baumann, and R. J. Dalby, “Reliability of InAlGaAs strained quantum well lasers operating at 0.81 m,” IEEE Photon. Technol. Lett., vol. 4, no. 8, pp. 829–831, Aug. 1992. [2] I. O. Lelong, M. Blumina, R. Sarfaty, D. Fekete, I. Samid, and M. Yust, “Pulsed high power quantum well laser using asymmetric waveguide,” Semicond. Sci. Technol., vol. 11, pp. 568–570, 1996. [3] I. B. Petrescau-Prahova, M. Buda, and T. G. Van de Roer, “Design of a 1 W single filament laser,” IEICE Trans. Electron., vol. E77-C, pp. 1472–1478, 1994. [4] D. Botez, “Design consideration and analytical approximations for high continuous-wave power broad-waveguide diode lasers,” Appl. Phys. Lett., vol. 74, pp. 3102–3104, 1999. [5] S. L. Yallen, R. G. Waters, Y. C. Chen, B. A. Soltz, S. E. Fischer, D. Fekete, and J. M. Ballantyne, “20 000 h InGaAs quantum well lasers,” Electron. Lett., vol. 26, pp. 2083–2084, 1990. [6] J. Gilor, I. Samid, and D. Fekete, “Threshold current density reduction of strained AlInGaAs quantum-well laser,” IEEE J. Quantum Electron., vol. 40, no. 10, pp. 1355–1364, Oct. 2004. [7] D. Fekete, “Reducing the threshold current density of AlInGaAs double Quantum-well laser with n-type delta doping,” Appl. Phys. Lett, vol. 86, 2005. [8] P. W. A. Mc Ilroy, A. Kurobe, and Y. Uematsu, “Analysis and application of theoretical gain curves to the design of multi-quantum-well lasers,” IEEE J. Quantum Electron., vol. 21, no. 12, pp. 1958–1963, Dec. 1985. [9] J. Z. Wilcox, G. L. Peterson, S. Ou, J. J. Yang, M. Jansen, and D. Schechter, “Gain- and threshold-current dependence for multiple-quantum-well lasers,” J. Appl. Phys., vol. 64, pp. 6564–6567, 1988. [10] O. Buchinsky, M. Blumin, M. Orenstein, G. Eisenstein, and D. Fekete, “Strained InGaAs–GaAs single quantum well lasers coupled to n-type -doping-improved static and dynamic performance,” IEEE J. Quantum Electron., vol. 34, no. 9, pp. 1690–1697, Sep. 1998. [11] O. Buchinsky, M. Blumin, R. Sarfaty, I. Samid, M. Yust, and D. Fekete, “n-type -doped quantum well lasers with extremely low transparency current density,” Appl. Phys. Lett., vol. 68, no. 15, April 8, 1996. [12] M. W. Sasnett, “Propagation of multimode laser beams: the M factor,” in The Physics and Technology of Laser Resonators. New York: Adam Hilger, 1989. [13] A. Al-Muhanna, L. J. Mawst, D. Boetz, D. Z. Garbuzov, R. U. Martinelli, and J. C. Connolly, “High-power (> 10 W) continuous-wave operation from 100-m-aperture 0.97-m-emitting Al-free diode lasers,” Appl. Phys. Lett., vol. 73, p. 1182, 1998. [14] R. K. Haung, J. P. Donnelly, L. J. Missaggia, C. T. Harris, J. Plant, D. E. Mull, and W. D. Goodhhue, “High-power nearly diffraction-limited AlGaAs-InGaAs semiconductor slab-coupled optical waveguide laser,” IEEE Photon. Technol. Lett., vol. 15, no. 7, pp. 900–902, Jul. 2003. [15] T. Makino, “Analytical formulas for the optical gain of quantum wells,” IEEE J. Quantum Electron., vol. 32, no. 3, pp. 493–501, Mar. 1996.

1114

Michael Yasin graduated from the Radiophysics Faculty of Nizhny Novgorod State University, Nizhny Novgorod, Russia, in 1992 and received the M.Sc. degree in physics from the Technion-Israel Institute of Technology (I.I.T.), Haifa, Israel, in 2002, where he is currently pursuing the Ph.D. degree in physics. From 2002 to 2004, he worked for the Research and Development Team of Lambda Crossing, Israel. He was involved in the design and development of a new class of integrated tunable optical components. His interests are in electrooptic semiconductor devices.

Ilan Samid was born in Israel in 1936. He received the B.Sc., M.Sc., and D.Sc. degrees in electrical engineering from the Technion-Israel Institute of Technology (I.I.T.), Haifa, Israel, in 1966, 1970, and 1978, respectively. He started his research on injection lasers in Israel on 1966. Since 1970, he has been working on hetero-epitaxy and research and development of high-power laser diodes and other optical and electronic devices, such as IR detectors, solar cells, GaN lasers, and GaAs microwave devices, both in Israel (I.I.T., Rafael, SCD-Semiconductor Devices, and PBC Lasers) and laboratories in the USA (Quantum Electronics Laboratory, California Institute of Technology, Pasadena, under Prof. Amnon Yariv from 1973 to 1975, and in Comsat Laboratories. on sabbatical leave from Rafael from 1982 to 1983). Since September 2003, he has also been the Chief Engineer at PBC Lasers, a nanophotonic start-up company in Israel.


Zeev Engelman received the B.Sc. degree in electrical engineering from the Technion-Israel Institute of Technology, Haifa, Israel. From 1978 to 1989, he worked as a Research Engineer of optical thin film at Rafael, Israel, where he developed vacuum deposition reactors. Since 1989, he has been at the Solid State Institute, Technion-Israel Institute of Technology, as a Research Engineer in the GaAs Semiconductor Device Laboratory where he is working on hetero-epitaxy and research and development of high-power lasers.

Dan Fekete received the B.Sc. degree in physics and mathematics and the Ph.D. degree in physics both from the Hebrew University of Jerusalem, Jerusalem, Israel. After graduating, he spent a three-year postdoctoral period at California Institute of Technology, Pasadena, and at Xerox PARC. Since 1980, he has been with the Physics Department, Solid-State Institute and the Advanced Optoelectronics Research Center, Technion-Israel Institute of Technology, Haifa, Israel, aside from three visiting years at Cornell University, Ithaca, NY, (1985–1986, 1996–1996) and at the Swiss Federal Institute of Technology (EPFL), Lausanne, Switzerland, from 2000 to 2001. His research interests include semiconductor diode lasers.