Quantum Electronics, IEEE Journal - IEEE Xplore

1 downloads 0 Views 115KB Size Report
From Arrhenius plots of the integrated luminescence intensity, it is found that carrier loss from the QW is dominated by a nonradiative loss mechanism with an ...
1338

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 8, AUGUST 1997

Photoluminescence Investigation of the Carrier Confining Properties of Multiquantum Barriers A. P. Morrison, Member, IEEE, L. Considine, S. Walsh, N. Cordero, J. D. Lambkin, G. M. O’Connor, E. M. Daly, T. J. Glynn, and C. J. van der Poel Abstract— A comparative luminescence study of two Ga0:52 In0:48 P–(Al0:5 Ga0:5 )0:52 In0:48 P single-quantum-well (SQW) samples with bulk and multiquantum barrier (MQB) barriers is presented. When excess carriers are only created in the quantum wells (QW’s) of the samples by resonant excitation using a dye laser, the luminescence efficiency of both samples as a function of temperature is found to be essentially identical. We find, therefore, no evidence for any enhancement in the confining potential of the MQB sample over the bulk barrier sample. From Arrhenius plots of the integrated luminescence intensity, it is found that carrier loss from the QW is dominated by a nonradiative loss mechanism with an activation energy considerably smaller than that expected from direct thermal loss of electrons and holes into the barriers. We suggest that the improved device characteristics reported for lasers containing MQB’s is due to effects other than the quantum interference of electrons. Index Terms— AlGaInP, multiquantum barriers, photoluminescence measurement, quantum-effect semiconductor devices, quantum interference, quantum wells, semiconductor lasers,

I. INTRODUCTION

S

HORT-WAVELENGTH (600–700 nm) visible laser diodes are of considerable technological interest. These lasers have applications in areas of optical storage, barcode readers, short-haul communication networks, and laser printers. Although the production of reliable lasers at wavelengths as short as 650 nm is now routine, it is found that as the emission wavelength is reduced, lasers increasingly suffer from high-threshold current densities, low powers, and low characteristic temperatures ( ). This behavior is caused by the small conduction and valence band-edge discontinuities between the Ga In P and (Al Ga ) In P alloys used in these visible lasers. The small band-edge discontinuities allow a significant portion of the injected electron current to leak over the cladding-waveguide heterobarrier, a problem which is then exacerbated as the operating wavelength is reduced [1], [2]. Manuscript received January 3, 1997; revised April 11, 1997. A. P. Morrison is with Technology Characterization and Modeling Group, National Microelectronics Research Centre, Lee Maltings, Ireland. L. Considine was with the National Microelectronics Research Centre, Optronics Ireland, Lee Maltings, Cork, Ireland. He is now with Swan & Co., Ltd., Harston, Cambridge, U.K. S. Walsh, N. Cordero, and J. D. Lambkin are with the National Microelectronics Research Centre, Optronics Ireland, Lee Maltings, Cork, Ireland. G. M. O’Connor, E. M. Daly, and T. J. Glynn are with the Physics Department, Optronics Ireland, University College, Galway, Ireland. C. J. van der Poel is with Philips Research Laboratories, 5656 AA Eindhoven, The Netherlands. Publisher Item Identifier S 0018-9197(97)05431-6.

Several techniques are used to minimize this problem such as increasing the level of p-doping in the cladding region [1], [2] and using highly disordered alloys to achieve the largest possible band-offsets [3]. Another imaginative technique to reduce electron leakage, first put forward by Iga et al. [4], is to include a multiple quantum barrier (MQB) in the pcladding region of the laser. A MQB consists of a thick wide ˚ followed by a thin superlattice bandgap layer (150 to 200 A) (approximately 10 periods) with well and barrier thicknesses ˚ of the order of tens of Angstroms. The initial thick layer is used to prevent tunnelling of low-energy electrons into the superlattice, while the superlattice itself sets up an interference of the electron’s wavefunction to achieve a high reflection coefficient. The MQB is therefore in many ways analogous to an optical distributed Bragg reflector. By judicial choice of layer thicknesses, electron reflection coefficients of greater than 99.9% can be maintained for energies approaching twice that of the conventional heterobarrier height. This makes MQB’s an attractive stratagem to improve carrier confinement in the active region. Several groups, including our own, have reported reduced threshold current densities and increased values as a result of using MQB’s in visible lasers. We have shown a 31% reduction in the room-temperature threshold current of a bulk double-heterostructure AlGaInP laser due to the use of a MQB [5]. Kishino et al. [6] reported a reduction in threshold current density from 1.2 kA/cm for a conventional laser design to 0.84 kA/cm for a laser containing a MQB, while Rennie et al. [7] measured a 20 C increase in the maximum operating temperature of a MQB laser over its conventional counterpart. Such reports have not been confined to AlGaInP lasers and several groups have also published, equally impressive improvements for AlGaAs- and InGaAsP-based lasers [8]–[10], despite these lasers already having reasonably large intrinsic heterojunction barriers. In virtually all these cases, the improvement in laser characteristics is attributed to the MQB providing a virtual increase in the heterobarrier. However, improved laser performance alone does not constitute unambiguous evidence for a virtual barrier. It is quite feasible, for example, that the improvement could be attributed to some other prosaic explanation associated with the MQB such as the superlattice helping to prevent the back diffusion of dopant into the active region. To substantiate the claim that MQB’s genuinely produce a barrier height increase, one needs to examine the MQB in isolation. The few authors that have attempted to do so usually compare bulk and MQB barriers in n-i-n tunnel diodes [11]. Unfortunately, most of this work has

0018–9197/97$10.00  1997 IEEE

MORRISON et al.: THE CARRIER CONFINING PROPERTIES OF MULTIQUANTUM BARRIERS

concentrated on AlGaAs-based devices and few quantitative measurements of barrier height improvement have as yet been reported with the notable exception of Rennie et al. [12]. Recently, however, Islam et al. [13] reported a comparative luminescence study of undoped single InGaP quantum wells (SQW’s) with: 1) bulk AlInP barriers; 2) InGaP–AlInP MQB barriers; and 3) a generic MQB barrier structure referred to as a strain-modulated aperiodic superlattice heterobarrier (SMASH). These SMASH barriers are essentially MQB’s in which the bandgaps of the superlattice well layers increase in successive layers away from the thick antitunnelling layer. The structure is designed to enhance the carrier capture in the SQW due to the “funnelling” potential of the superlattice. These authors found that at room temperature the SMASH and MQB barrier samples were, respectively, 20 and 4 times brighter than the SQW with bulk AlInP barriers. These authors conclude that the MQB structures are responsible for the improved luminescence efficiency by providing the expected virtual barrier. In a similar fashion, Takagi et al. [14] have also investigated the temperature dependence of the photoluminescence (PL) intensity from a GaAs–AlGaAs SQW with and without a MQB barrier. Again, due to the increased luminescence efficiency from the MQB sample, these authors also concluded that an enhanced virtual potential barrier is responsible. In this paper, we also present a comparative luminescence study of Ga In P–(Al Ga ) In P SQW’s with both bulk and MQB barriers and attempt to measure any increase in the effective total barrier height. Like Islam et al. [13] and Takagi et al. [14], we also observe an increase in luminescence efficiency at room temperature in the SQW sample with the MQB barriers when compared with the conventional SQW sample. However, we find no evidence for the presence of a virtual barrier. Instead we attribute the observed increase in the SQW luminescence efficiency to the simple thermalization of excess carriers trapped in the MQB superlattice back into the quantum well (QW).

II. EXPERIMENT The barrier height present in a GaInP–AlGaInP SQW can be defined as the total energy required to remove an electron from the ground state in the well to the conduction-band edge of the barrier and the energy required to remove a corresponding hole from its ground state to the barrier’s valence-band edge. This barrier height is experimentally estimated by measuring the difference between the AlGaInP barrier and the GaInP QW PL peak energies. We have previously established that for Ga In P–(Al Ga ) In P QW’s, the dominant loss mechanism for electrons and holes is indeed thermal activation over this barrier [15]. In an identical fashion to that found for In Ga As–GaAs QW’s [16], the integrated luminescence intensities from Ga In P–(Al Ga ) In P QW’s of various widths as a function of inverse temperature show an Arrhenius behavior which, when fitted, give thermal activation energies that agree exceedingly well with those expected for this particular mechanism. In principle, therefore, by measuring the luminescence as a function of temperature of a

1339

GaInP–AlGaInP SQW with both conventional bulk barriers and MQB barriers, the effect of the MQB on the barrier height can be directly measured and compared. Two Ga In P–(Al Ga ) In P SQW samples were measured, one with bulk (Al Ga ) In P barriers (SQW ) and the other with MQB barriers (SQW ). The SQW sample consisted of a 50˚ Ga In P well with 0.2- m (Al Ga ) In P A sample also consisted of a 50barriers. The SQW ˚ ˚ Ga In P well, either side of which were 150-A A (Al Ga ) In P antitunnelling layers followed by ˚ five pairs of (Al Ga ) In P–Ga In P, 35 A ˚ and 15 A thick, respectively, and a further 0.16 m of (Al Ga ) In P. The total thickness of the two SQW samples were identical. Both samples, nominally undoped, were grown by conventional atmospheric metal organic vapor phase epitaxy (MOVPE) at a growth temperature of 760 C on a Si-doped GaAs substrate misoriented by 3 toward the (111)A direction. This substrate misorientation, coupled with the high growth temperature, ensures that a high degree of group III sublattice disorder is achieved [17]. Cleaved-edge transmission electron microscopy confirmed that the measured ˚ of the nominal design values. layer thickness were within 2 A By using a simple effective mass model described elsewhere [18], and assuming Ga In P and (Al Ga ) In P electron effective masses of 0.11 m and 0.23 m , respectively, and a conduction-band offset of 213.5 meV [19], the MQB structure used in the SQW sample was designed to have a ˚ If a stringent cutoff criterion total thickness of less than 400 A. for the electron reflection of 99.9% is used, then the estimated virtual barrier height improvement is 77.3 meV larger than the intrinsic conduction-band offset. The CW PL spectra were collected from the SQW samples in two configurations. In the first instance, both samples were excited above the (Al Ga ) In P barrier energy using the 488-nm line of an Ar ion laser, while in the second the SQW’s were excited resonantly using a tuneable dye laser, both with an excitation density of 2.5 kW m . This latter mode of excitation only creates a population of electrons and holes in the SQW itself. Initial PL spectra were also taken at 4.2 K in a helium bath cryostat using 488-nm excitation. The temperature-dependent measurements from 10 K to room temperature were taken using a closed cycle cryostat. In all cases, the luminescence signal was dispersed through a SPEX 500 M spectrometer and detected with a silicon charge-coupled device.

III. RESULTS A comparison of the 4.2-K PL spectra for the SQW and SQW samples (see Fig. 1) shows that the QW and barrier emissions occur at the same energy for both samples. The QW emission occurs at 2.067 eV for the SQW and 2.065 eV for the SQW while the FWHM of these peaks are 15 and 12 meV, respectively. The barrier emissions occur at 2.328 and 2.329 eV, respectively, for the SQW and SQW samples, all of which indicate that the QW structures in the two samples are virtually identical. From

1340

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 8, AUGUST 1997

Fig. 1. PL spectra from the samples SQWBulk and SQWMQB .

the QW and barrier emission energies, the barrier height for this QW is determined to be 262 meV. Additional peaks located 32 and 51 meV below the bulk barrier peak in both samples are LO phonon replicas of the bulk emission. The origin of the background emission at lower energy, particularly sample, is unknown although it could be for the SQW acceptor related or related to ordering-induced emissions that can occur in these materials. Peaks located 47 meV below the emission peak of the QW in both structures are also attributed to a phonon replica as phonon energies for the InPlike and GaP-like LO phonons in InGaP have been measured to be 44.5 and 47 meV, respectively [20]. For the SQW sample, the MQB emission is located at 2.203 eV with a shoulder at 2.187 eV. From the comparison of the 4.2-K PL of these two samples, we can conclude that the QW’s are almost identical and hence a fair comparison can be made to reveal the influence of the MQB on carrier confinement. Fig. 2 shows Arrhenius plots of the integrated PL intensity as a function of temperature for both of the samples when excited above the barrier energy with the 488-nm line of the Ar ion laser. The difference in behavior between the two samples is quite significant. In the first instance, the decrease in intensity in going from 10 to 300 K is a factor of 1/443 sample compared to only a factor of 1/61 for the SQW (as shown in Table I), perhaps suggesting the for SQW presence of the Bragg confinement of carriers. However, a closer analysis of the data suggests otherwise. The SQW sample shows a slight increase in PL intensity at higher temperatures before dropping off sharply as room temperature is approached. Comparing this with the same plot for sample , the PL intensity begins to fall off gradually before SQW peaking again at 140 K and then falling off sharply as room temperature is reached. This resonant increase in PL intensity is indicative of the thermal emission of carriers trapped in the MQB superlattice and their subsequent recapture by the lower

TABLE I FACTORS

BY WHICH THE 10-K PL INTENSITIES ARE REDUCED IN GOING TO ROOM TEMPERATURE UNDER THE TWO MODES OF EXCITATION AS DESCRIBED IN THE TEXT

energy state of the QW. This process of thermalization and recapture has been previously seen in both AlGaInP–InGaP [15] and InGaAs–GaAs QW structures [21]. Fig. 3 shows a plot of the integrated PL intensity for both samples when the QW is resonantly excited using the dye laser. Unlike the previous excitation mode, the behavior of the two samples is now very similar. In both cases, the PL intensity gradually reduces with increasing temperature until a temperature is reached when the fall-off in PL intensity is then very sharp. Increases in intensity prior to quenching are no longer evident. The maximum difference in intensity between the two samples is only a factor of 2.8 and occurs at a temperature of 170 K; the total reduction in intensity in going from 10 K to room temperature for both samples is virtually identical (a factor of approximately 1/40) as shown in Table I. IV. ANALYSIS In an identical fashion to that previously described for PL data from Ga In P–(Al Ga ) In P QW’s [15], the Arrhenius behavior shown in Fig. 3 can be fitted very successfully across the entire temperature range using a simple model that assumes two thermally activated loss mechanisms. The low-temperature mechanism is characterized by an activation energy while the higher temperature mechanism is

MORRISON et al.: THE CARRIER CONFINING PROPERTIES OF MULTIQUANTUM BARRIERS

1341

Fig. 2. Integrated PL intensity using 488-nm excitation for samples SQWBulk and SQWMQB .

Fig. 3. Integrated PL intensity for SQWBulk and SQWMQB when the QW is resonantly excited using a dye laser. Solid lines are fits to the experimental data, using (1).

characterized by the activation energy . The luminescence intensity ( ) as a function of temperature ( ) is therefore simply given as (1) is the PL intensity at 4.2 K, is Boltzmann’s where constant, while and are the ratios of the 4.2-K radiative lifetime to the high-temperature nonradiative lif-time for mechanisms and , respectively. The prefactors and are assumed to remain constant as a function of temperature. The results of a least squares fit of (1) to the experimental data are shown in Fig. 3 (solid lines) while the fitted parameters are given in Table II. The quality of the fits to the experimental

data are excellent and allow a quantitative comparison of the two structures to be made. The nature of the low-temperature mechanism is not clear although it may be associated with the thermalization of carriers from band-edge fluctuations due to either alloy variations or well width fluctuations followed by nonradiative recombination, possibly at the QW interfaces. The higher temperature activation process is anticipated to be dominated by the thermal loss of carriers out of the QW, although this turns out not to be the case.

V. DISCUSSION There are several important points to note from the fitted parameters given in Table II. At first glance it may

1342

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 8, AUGUST 1997

FITTING PARAMETERS

OF

TABLE II (1) USED TO GIVE

appear promising that the activation energy for the SQW sample is almost a factor of two larger than sample, suggesting that the MQB has for the SQW indeed produced an enhanced barrier height. However, both activation energies are less than half the 262-meV barrier height deduced from the PL emission energies sample. Therefore, unlike the lower Al for the SQW composition Ga In P–(Al Ga ) In P QW’s mentioned previously [15], the major loss mechanism for carriers in both these higher Al composition SQW samples is not simply activation out of the QW as anticipated, but is instead complicated by the presence of a competing and less energetic loss mechanism. This behavior is also observed in InGaAs QW’s with AlGaAs barriers [16]. Unlike the major loss mechanism in InGaAs–GaAs QW’s, which is dominated by thermal activation of carriers out of the well, for InGaAs–AlGaAs QW’s the major loss mechanism is dominated by nonradiative recombination via traps in the barrier or interface which are assumed to be associated with Al–O complexes. It is possible, therefore, that the high Al composition in the AlGaInP SQW’s discussed here also introduces a similar loss mechanism. It is also possible that since the (Al Ga ) In P alloy is close to the cross-over, an additional loss mechanism might involve the nonradiative loss of electrons via states in the barrier. Such a mechanism can give rise to a significant proportion of the large excess leakage currents measured in tensile strained QW visible lasers [1]. It is worth stressing, therefore, that the assumption that the poor thermal characteristics of visible lasers is merely due to the small -minima band offsets is perhaps an oversimplification and that greater attention should be paid to other possible loss mechanisms. of both the SQW samDespite the activation energies ples being considerably lower than the experimentally determined conventional barrier height of 262 meV, it might be argued that the MQB has nevertheless resulted in an increase in the barrier height when compared to the bulk barrier. However, the mechanism discussed in the previous section, whereby carriers thermally ejected from the QW can be captured by the superlattice of the MQB and subsequently for the SQW sample recaptured by the QW, prevents from being determined uniquely. We do not, therefore, attach any great significance to the measured difference in between the samples. In addition, from a device perspective, the important parameter of merit for a QW is its roomtemperature luminescence efficiency. The combination of the and prefactors for both samples active energies ensures that their room-temperature luminescence efficiencies are virtually identical. From this perspective, therefore, the

THE

FITS

AS

SHOWN

IN

FIG. 3

two samples are indistinguishable and the MQB considered to have no additional effect over the bulk barrier at room temperature. This observation is in stark contrast to that made by Islam et al. [10] and Takagi et al. [14], who both reported significant enhancements of luminescence efficiency in MQB QW samples at room temperature. However, the PL measurements reported by these authors were made using an excitation wavelength of 488 nm and therefore, in an identical fashion to that reported here, the observed increased in PL intensity in their MQB samples is almost certainly due to the capture of carriers into the QW that have thermally escaped from the MQB’s superlattice. The data presented here prompt two questions. First, why is a barrier increase not observed? Secondly, why do lasers with MQB’s invariably show improved device characteristics? For a MQB to provide a virtual barrier, it is essential that the phase of the electron’s wavefunction remains constant throughout the structure, i.e., the electron is neither elastically nor inelastically scattered. It seems clear, therefore, that the total thickness of the MQB should be kept as thin as possible, at least as thin as the expected coherence length of an electron in the material. An estimate for the electron coherence length can be found by simply equating it to the electrons’ mean free path. For bulk intrinsic GaAs, for example, with a low field mobility of 8000 ˚ This cm /Vs, the mean free path is then approximately 350 A. estimate of electron coherence length is encouraging in that it does not rule out the possibility of the MQB generating a virtual barrier provided the MQB total thickness is kept below this value. Despite this, however, several factors, such as the relatively large number of interfaces typically found in a MQB and that many of the MQB’s reported in the literature ˚ almost certainly prevent are considerably thicker than 350 A, electron coherence from being maintained in these structures at room temperature. This is supported by photoreflectance measurements, made as a function of temperature, of Bragg confined transitions in a superlattice with MQB’s. Lipsanen et al. [22] observe Bragg confined states up to temperatures of 200 K but not at higher temperatures, presumably due to the breakdown of electron coherence. It is conceivable, therefore, that MQB’s provide enhanced barriers at low temperatures when they are not required, but that these barriers cease to exist at room temperature when they are required. For the case of a MQB laser, this situation is further exacerbated. The concept and modeling of the MQB for use in lasers has to date been based upon the highly simplistic model of a single electron traveling perpendicular to perfect flat-band heterojunction interfaces with absolutely no account taken of any scattering mechanisms (infinite coherence length). If there is going to be any possibility of observing the MQB

MORRISON et al.: THE CARRIER CONFINING PROPERTIES OF MULTIQUANTUM BARRIERS

effect at all, it will be in structures where flat bands are maintained, where heterojunction perfection is maximized, and where the phonon density is low, i.e., by performing optical measurements on undoped structures at low temperature as has been attempted here and in [22]. Lasers, on the other hand, are characterized by active regions with high carrier densities, nanosecond carrier lifetimes, efficient LO phonon scattering, and the presence of significant electric fields caused by both interfacial band-bending and Ohmic resistance. All these factors mitigate against any MQB virtual barrier by virtue of destroying electron coherence. The question of why MQB lasers show enhance characteristics still remains. We make the observation that all reported MQB lasers, as far as we are aware, show enhanced performance. This enhancement seems independent of the material system used; whether the MQB’s barrier material has a direct or indirect bandgap, which material parameters are chosen to model and derive the MQB’s optimum structure [23], and even whether the thick antitunnelling layer is present [15]. This strongly suggests that the effect of the MQB is other than producing a virtual barrier. We believe, therefore, an alternative explanation, of which there are several possibilities, is required. It is possible, for example, that electron transport through the superlattice is inhibited by scattering into the X minima of the MQB as suggested by Yen et al. [23], or that the superlattice simply inhibits the back diffusion of pdopants into the active region. We suggest, however, that if the MQB does prevent back diffusion of p-dopant or if the MQB acts as a hole trap, then a further benefit is gained. The prevention of dopant back diffusion will produce a doping spike at the cladding-waveguide interface since MQB’s are generally placed adjacent to the waveguide region on the pside of a laser. This local increase in doping will cause the quasi-Fermi level for holes in the cladding layer to move closer to the valence band edge and thereby increase the potential barrier for electron leakage. Although this argument is only supposition, we feel, however, that considerably greater experimental and theoretical attention should be paid to this type of mechanism. VI. CONCLUSION We have shown that measurements of PL intensity as a function of temperature from SQW’s with MQB barriers do not show any improved luminosity due to the Bragg confinement of carriers at room temperature. Previous reports of similar measurements which indicated otherwise failed to take account of the thermalization of carriers trapped in the superlattice of the MQB into the SQW. We suggest that although Bragg confinement of carriers may occur at low temperatures, at room temperature the breakdown in electron coherence prevents any generation of a virtual barrier.

1343

[2] D. P. Bour, D. W. Treat, R. L. Thornton, R. S. Geels, and D. F. Welch, “Drift leakage current in AlGaInP quantum-well lasers,” IEEE J. Quantum Electron., vol. 29, pp. 1337–1342, May 1993. [3] T. Suzuki, A. Gomyo, S. Iijima, K. Kobayashi, S. Kawata, I. Hino, and T. Yuasa, “Band-gap energy anomaly and sublattice ordering in GaInP and AlGaInP grown by metalorganic vapor phase epitaxy,” Jpn. J. Appl. Phys., vol. 27, pp. 2098–2102, 1988. [4] K. Iga, H. Uenohara, and F. Koyama, “Electron reflectance of multiquantum barrier (MQB),” Electron. Lett., vol. 22, pp. 1008–1010, Sept. 1986. [5] A. P. Morrison, J. D. Lambkin, C. J. van der Poel, and A. Valster, “Evaluation of multiquantum barriers in bulk double heterostructure visible laser diodes,” IEEE Photon. Technol. Lett., vol. 8, pp. 849–851, July 1996. [6] K. Kishino, A. Kikuchi, Y. Kaneko, and I. Nomura, “Enhanced carrier confinement effect by the multiquantum barrier in 660 nm GaInP/AlInP visible lasers,” Appl. Phys. Lett., vol. 58, pp. 1822–1824, Apr. 1991. [7] J. Rennie, M. Watanabe, M. Okajima, and G. Hatakoshi, “High temperature 90 C CW operation of 646 nm InGaAlP laser containing multiquantum barrier,” Electron. Lett., vol. 28, pp. 150–151, Jan. 1992. [8] T. Takagi and K. Iga, “Temperature dependence of GaAs/AlGaAs multiquantum barrier lasers,” IEEE Photon. Technol. Lett., vol. 4, pp. 1322–1324, 1992. [9] Y. Inaba, T. Uchida, N. Yokouchi, T. Miyamoto, K. Mori, F. Koyama, and K. Iga, “Growth of GaInAs(P)/InP multi-quantum barrier by chemical beam epitaxy,” J. Cryst. Growth, vol. 136, pp. 297–301, 1994. [10] T. Fukushima, H. Shimizu, K. Nishikata, Y. Hirayama, and M. Irikawa, “Carrier confinement by multiple quantum barriers in 1.55 m strained GaInAs/AlGaInAs quantum well lasers,” Appl. Phys. Lett., vol. 66, pp. 2025–2027, Apr. 1995. [11] T. Takagi, F. Koyama, and K. Iga, “Electron-wave reflection by multiquantum barrier in n-GaAs/i-AlGaAs/n-GaAs tunnelling diode,” Appl. Phys. Lett., vol. 59, pp. 2877–2879, Nov. 1991. [12] J. Rennie, M. Okajima, K. Itaya, and G. Hatakoshi, “Measurement of the barrier height of a multiple quantum barrier (MQB),” IEEE J. Quantum Electron., vol. 30, pp. 2781–2789, Dec. 1994. [13] M. R. Islam, R. V. Chelakara, and R. D. Dupuis, “InAlP/InGaP strainmodulated aperiodic superlattice heterobarrier for enhanced electron confinement in visible ( 650 nm) light emitting devices,” Appl. Phys. Lett., vol. 67, pp. 2057–2059, Oct. 1995. [14] T. Takagi, F. Koyama, and K. Iga, “Design and photoluminescence study on a multiquantum barrier,” IEEE J. Quantum Electron., vol. 27, pp. 1511–1516, 1991. [15] E. M. Daly, T. J. Glynn, J. D. Lambkin, L. Considine, and S. Walsh, “The behavior of quantum well luminescence as a function of temperature,” Phys. Rev. B, vol. 52, pp. 4696–4699, Aug. 1995. [16] J. D. Lambkin, D. J. Dunstan, K. P. Homewood, and L. K. Howard, “Thermal quenching of the photoluminescence of InGaAs/GaAs and InGaAs/AlGaAs strained-layer quantum wells,” Appl. Phys. Lett., vol. 57, pp. 1986–1988, 1990. [17] J. D. Lambkin, L. Considine, S. Walsh, G. M. O’Connor, C. J. McDonagh, and T. J. Glynn, “The temperature dependence of the photoluminescence intensity of ordered and disordered InGaP,” Appl. Phys. Lett., vol. 65, pp. 73–75, July 1994. [18] A. P. Morrison, J. D. Lambkin, E. O’Sullivan, and S. Fahy, “Simple effective mass model to include the effects of 0 X mixing in multiquantum barriers,” Opt. Eng., vol. 33, pp. 3926–3930, Dec. 1994. [19] K. F. Brennan and P. K. Chiang, “Calculated electron and hole steadystate drift velocities in lattice matched GaInP and AlGaInP,” J. Appl. Phys., vol. 71, pp. 1055–1057, Jan. 1992. [20] O. Madelung, Ed., Landolt-Bornstein New Series. Berlin, Germany: Springer-Verlag, 1992, vol. 22a. [21] M. Vening, D. J. Dunstan, and K. P. Homewood, Phys. Rev. B, vol. 48, pp. 2412–2417, 1993. [22] H. K. Lipsanen, K. Taskinen, and V. M. Airaksinen, “Optical spectroscopy of bragg confined transitions in a superlattice with multiquantum barriers,” Solid State Commun., vol. 93, pp. 525–528, 1995. [23] S. T. Yen, C. P. Lee, C. M. Tsai, and H. R. Chen, “Influence of X-valley superlattice on electron blocking by multiquantum barriers,” Appl. Phys. Lett., vol. 65, pp. 2720–2722, Nov. 1994.

=

0

REFERENCES [1] A. T. Meney, A. D. Prins, A. F. Phillips, J. L. Sly, E. P. O’Reilly, D. J. Dunstan, A. R. Adams, and A. Valster, “Determination of the band structure of disordered AlGaInP and its influence on visible-laser characteristics,” IEEE J. Select. Topics Quantum Electron., vol. 1, pp. 697–706, June 1995.

A. P. Morrison (S’92–M’96), photograph and biography not available at the time of publication.

1344

L. Considine, photograph and biography not available at the time of publication.

S. Walsh, photograph and biography not available at the time of publication.

IEEE JOURNAL OF QUANTUM ELECTRONICS, VOL. 33, NO. 8, AUGUST 1997

G. M. O’Connor, photograph and biography not available at the time of publication.

E. M. Daly, photograph and biography not available at the time of publication.

N. Cordero, photograph and biography not available at the time of publication.

T. J. Glynn, photograph and biography not available at the time of publication.

J. D. Lambkin, photograph and biography not available at the time of publication.

C. J. van der Poel, photograph and biography not available at the time of publication.