APPLIED PHYSICS LETTERS

VOLUME 73, NUMBER 26

28 DECEMBER 1998

Effect of detuning on the angular emission pattern of high-efficiency microcavity light-emitting diodes C. Dill,a) R. P. Stanley, U. Oesterle, D. Ochoa, and M. Ilegems Institute of Micro and Optoelectronics, De´partement de Physique, EPFL, Ch-1015 Lausanne, Switzerland

~Received 4 August 1998; accepted for publication 23 October 1998! Results on molecular beam epitaxy-grown microcavity light-emitting diodes with InGaAs/GaAs quantum wells and a hybrid top mirror are presented. An external quantum efficiency of 14.8% is achieved for a 400 mm diam light-emitting diode. The strong influence of the spectral overlap between the spontaneous emission spectrum and the cavity resonance mode on the radiation pattern is shown. The angular emission profile is compared with model predictions for different detunings, and a very good agreement is obtained when the asymmetric spectral broadening of the intrinsic spontaneous emission is taken into account. © 1998 American Institute of Physics. @S0003-6951~98!04052-2#

There is a great interest for vertical light emitters, as laser diodes and light-emitting diodes ~LEDs!, in the near infrared for many applications in optoelectronic systems. Light-emitting diodes are used in displays, as indicator lights, as emitters for remote free-space functions, and to a lesser extent, for short-distance communication and optical interconnects. The interest in LEDs for these applications lies in their simple design and low fabrication costs. In addition, LEDs are inherently more reliable than laser diodes, have a lower temperature sensitivity, operate without threshold, and are better in terms of eye safety. Despite these advantages, light-emitting diodes suffer from some major limitations such as the poor extraction efficiency of light ~typically, below 2% per facet for a simple planar design!, the wide spectral width, and the large divergence of the output beam. In the last 5–6 years, the concept of microcavity lightemitting diodes ~MCLED! has been introduced to overcome some of these limitations. In these structures, the optically active material is placed inside a microcavity whose fundamental optical mode is resonant with the natural emission spectrum.1 Typically, semiconductor quantum wells ~QWs! are placed inside a wavelength-size Fabry–Pe´rot ~FP! cavity at the antinode of the optical field. The cavity itself is bounded by two mirrors, which can be quarter-wavelength distributed Bragg reflectors, metallic mirrors, or hybrids of the two. A microcavity can modify the spontaneous emission characteristics of the source, such as the spontaneous emission lifetime,2 the spectral linewidth,3 and the directionality of the emission.4,5 Planar microcavity light-emitting structures exploiting these effects were first proposed and demonstrated in 1992 by Schubert et al.6 This structure was further elaborated and improved, particularly in terms of efficiency, by Blondelle et al.,7 using metal–organic chemical vapor deposition ~MOCVD! grown material. As was shown earlier,8 the performance of the MCLED is critically dependent on spectral overlap between the spontaneous emission a!

Electronic mail: [email protected]

spectrum from the active layer and the cavity resonance mode. Therefore, it is important to determine the optimum spectral overlap resulting in maximum light output and efficiency. In this letter we present a molecular beam epitaxy ~MBE! grown high-efficiency microcavity LED. We provide a detailed comparison of MCLED device performance with model predictions and we demonstrate the influence of the detuning between the wavelength of maximum spontaneous emission and the cavity resonance mode, i.e., Dl5l SE ~spontaneous emission!–l FP ~cavity mode!, on efficiency and radiation pattern. The LED structure that we used is a variant of the design used by Blondelle et al.,7 with a hybrid top mirror rather than a pure metal mirror. A schematic diagram of the MCLED is shown in Fig. 1. The bottom emitting structures are grown by MBE on n 1 -GaAs substrates. The active region, consisting of three 70 Å In0.17Ga0.83As strained QWs, is embedded in the center of an undoped l layer (Al0.1Ga0.9As). The bottom mirror consists of 7.5 pairs of an AlAs/GaAs l/4 stack; Si-(n)-doped 231018 cm23, while the top mirror consists of one Al0.6Ga0.4As/GaAs pair, Be-( p)-doped 531018 cm23, with a nonuniform doping profile and completed by a phasematch layer and a gold mirror. A high doping level used near the surface allows nonalloyed Ohmic contacting. In order to characterize the relationship between output efficiency and detuning, the thickness of the intrinsic l layer

FIG. 1. Schematic cross section of the MCLED.

0003-6951/98/73(26)/3812/3/$15.00 3812 © 1998 American Institute of Physics Downloaded 22 Feb 2006 to 128.178.175.84. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

Dill et al.

Appl. Phys. Lett., Vol. 73, No. 26, 28 December 1998

3813

TABLE I. Experimentally determined detunings from PL and reflectivity measurements compared to values obtained by fitting the angular distribution pattern. Corresponding external quantum efficiencies ~QE! are given.

Diode A Diode B Diode C

Dl~exp.!

Dl~calc.!

Ext. QE

0 nm 15 nm 30 nm

0 nm 12 nm 29 nm

11.7% 14.8% 12.4%

was varied by growing a wedge, i.e., a nonuniform growth profile, thus resulting in a shift of the cavity resonance to shorter wavelength towards the edge of the wafer. Prior to processing, all samples are characterized by x-ray diffraction, photoluminescence, and reflectivity measurements. The x ray provides information on the actual thickness of the periodic layers. The position of QW emission is determined through photoluminescence ~PL! prior to the deposition of the Au layer; the Fabry–Pe´rot mode position is measured by reflectivity afterwards. The diodes are prepared using standard lithography and wet-chemical etching technique. In the first process step a large mesa is defined with standard photoresist and etched down into the n 1 substrate. The metals for the bottom Ohmic n-contact ~Ni/Ge/Au! are deposited by e-beam evaporation and alloyed. The gold p contact acts both as a metallic mirror and as an Ohmic contact. In order to maintain its high reflectivity the p contact is nonalloyed. These nonalloyed contacts fabricated on the 131020 cm23 p-doped layer have a specific resistance of 531025 V cm2. Finally, the mesas are etched down around the p-contact dots ~400 mm diam! into the active region, to prevent leakage current across the p – i – n-junction perimeter. The substrate is thinned down to 60 mm to reduce absorption losses for substrate emission. Finally, the substrate side is antireflection coated. All optical and electrical measurements are performed at room temperature. First, the results of PL and reflectivity measurements are used to map the detuning across the wafers. Thus, the shift of the cavity mode is estimated according to the growth gradient across the wafer. The experimentally determined detunings ~Dl! of three diodes, labeled A, B, and C, are given in Table I. These diodes are processed from wafers which showed a PL maximum at about 940 nm in the center of the wafer. Maximum performance was obtained for diode B, which has an estimated detuning, Dl 5215 nm. The light output versus current (L – I) characteristic of this diode under pulsed condition ~20 ns, 1 kHz! is shown in Fig. 2. The L – I characteristic ~open circles! is linear except at the lowest current densities. The external quantum efficiency ~solid diamonds! remains nearly constant for currents between 20 and 100 mA. The slight decrease in efficiency at higher currents is attributed to the spectral broadening of the spontaneous emission at higher injection levels. The diode has an output power of 18.5 mW at a current of 100 mA. The maximum external quantum efficiency into air is 14.8%. This value should be compared to a theoretical value of 20.1%. The theoretical value is calculated assuming a 215 nm detuning, a Gaussian spontaneous emission with a full width at half maximum ~FWHM! of 35 nm, an internal quantum efficiency of 100%, and neglecting recycling effects. The wall-plug efficiency is 15.5%. This

FIG. 2. Light output and external quantum efficiency vs current at room temperature under pulsed conditions for a MCLED with an estimated detuning of 15 nm ~diode B!.

high wall-plug efficiency is due to a low turn-on voltage and a low series resistance. For the 400 mm diode the differential resistance is less than 1 V. All angular resolved measurements were performed at I F 550 mA, which corresponds to a current density of 40 A/cm2 for a device of 400 mm diameter. Figures 3~a!, 3~b!, and 3~c! show the emission profile—power versus angle—for the three diodes: A, B, and C. In Fig. 3~a! the spontaneous emission is resonant with the cavity mode at normal incidence and this results in a narrow bell-shape emission profile. The emission is highly directional and has

FIG. 3. Experimental angular emission profiles ~solid lines! of diodes A, B, and C at I F 550 mA ~40 A/cm2!. Corresponding model predictions ~dashed lines! using the detuning as a fitting parameter: ~a! 0 nm, ~b! 12 nm, and ~c! 29 nm. Downloaded 22 Feb 2006 to 128.178.175.84. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp

3814

Dill et al.

Appl. Phys. Lett., Vol. 73, No. 26, 28 December 1998

FIG. 4. Spontaneous emission spectrum of three 70 Å In0.17Ga0.83As strained QWs, at a current density of 40 A/cm2 in the absence of a microcavity.

its maximum light output in the normal direction. A double lobe @Fig. 3~b!# was found for the diode described above ~diode B!, where a detuning of 215 nm was estimated. The negative detuning between the spontaneous emission and the cavity resonance improves the overlap of the luminescence and the ‘‘escape’’ window, and increases the extraction efficiency. Figure 3~c! shows the emission profile of a cavity with a large detuning, Dl5230 nm, having an emission maximum at about 45°. Although the emission profile covers an even larger solid angle, it has a lower efficiency due to the increased reflection at the semiconductor–air interface and total reflection beyond the critical angle. The previous descriptions illustrate how the tuning of the cavity resonance can modify the emission pattern from being highly directional to one with a wider profile and two lobes that are used to achieve very high efficiencies. The external quantum efficiencies listed in Table I, assuming an internal quantum efficiency of about 75%, show the strong influence of the detuning on the optical performance. This change in efficiency with detuning is in good agreement with the findings in Refs. 7 and 8, showing also maximum efficiency at a detuning of about 15 nm. Besides estimating the detuning according to PL and reflectivity measurements, the detuning can also be determined comparing the angular emission pattern with model predictions. The theory used in the simulation tool is based on the plane-wave expansion of a dipole emitter inside the microcavity.9 The refractive indices are taken from standard tables. The spectrum of the source emitter used in the calculations was taken from an electroluminescence spectrum ~Fig. 4!, measured on a similar test sample, which also has three 70 Å In0.17Ga0.83As strained QWs, at the same current density of 40 A/cm2, but in the absence of a microcavity. This spectrum, having a FWHM of 35 nm, was used in the

program to calculate the angular emission spectra. A leastsquares procedure was used to compare the calculated and measured spectra, and only one free parameter, the detuning, was used to obtain the best fit. The experimental data of the three diodes A, B, and C are compared with simulation results. The dashed lines in Figs. 3~a!–3~c! represent the best fit corresponding to the experimental curves. The best fit found for diode A is a detuning of 0 nm. This result is in agreement with the estimation of the detuning according to the PL and the reflectivity measurements. A best fit @Fig. 3~b!# for diode B was with a calculated detuning of 212 nm. For diode C the detuning, estimated to be around 230 nm, was calculated to be 229 nm with a very accurate fit of the two lobes. These results show the excellent agreement between the measured radiation pattern and the theoretical predictions based on the simulation tool. Considering the small deviation between the experimental curves and the simulation results, the technique of fitting the radiation pattern is found to be a very sensitive method to determine the detuning. In summary, we reported the fabrication of a MBE grown high efficiency single facet, substrate emitting microcavity LED, having a maximum external quantum efficiency of 14.8% and a high wall-plug efficiency of 15.5%. The device performance is modeled based on a plane-wave expansion of a dipole emitter inside the cavity. We showed that the modeling technique allows us to accurately simulate the radiation pattern of the diode as a function of the detuning between the maximum of spontaneous emission and the cavity resonance wavelength. A good agreement is obtained between the experimental data on detuning and the best-fit values obtained from modeling, which shows that the fitting is a very sensitive technique to characterize the properties of the diodes. This work was supported by the European Commission within the framework of the ESPRIT-SMILES program. 1

F. de Martini, G. Innocenti, G. R. Jacobowitz, and P. Mataloni, Phys. Rev. Lett. 59, 2955 ~1987!. 2 A. M. Vredenberg, N. E. J. Hunt, E. F. Schubert, D. C. Jacobsen, J. M. Poate, and G. J. Zydzik, Phys. Rev. Lett. 71, 517 ~1993!. 3 N. E. J. Hunt, E. F. Schubert, R. A. Logan, and G. J. Zydzik, Appl. Phys. Lett. 61, 2287 ~1992!. 4 H. De Neve, J. Blondelle, R. Baets, P. Deemester, P. Van Daele, and G. Borghs, IEEE Photonics Technol. Lett. 7, 287 ~1995!. 5 D. G. Deppe and C. Lei, Appl. Phys. Lett. 60, 921 ~1992!. 6 E. F. Schubert, Y. H. Wang, A. Y. Cho, L. W. Tu, and G. J. Zydzik, Appl. Phys. Lett. 60, 921 ~1992!. 7 J. Blondelle, H. De Neve, P. Deemester, P. Van Daele, G. Borghs, and R. Baets, Electron. Lett. 31, 1286 ~1995!. 8 H. De Neve, J. Blondelle, P. Van Daele, P. Demeester, and R. Baet, Appl. Phys. Lett. 70, 799 ~1997!. 9 D. Ochoa, J. Appl. Phys. ~to be published!.

Downloaded 22 Feb 2006 to 128.178.175.84. Redistribution subject to AIP license or copyright, see http://apl.aip.org/apl/copyright.jsp