1342 IEEE PHOTONICS TECHNOLOGY LETTERS, VOL. 11, NO. 11, NOVEMBER 1999 Small-Signal Analysis of 1.3- m Microcavity Light-Emitting Diodes P. Landais, B. Roycroft, A. Shaw, B. Depreter, I. Moerman, Member, IEEE, and J. Hegarty, Member, IEEE Abstract—The modulation speed of 1.3- m microcavity light- emitting diodes (MCLED’s) has been measured using a small- signal modulation analysis. A speed of 260 MHz using a 25- m diameter sample at current density of 10 kA/cm has been achieved. The carrier confinement has been calculated for several carrier densities in order to investigate the origin of the speed lim- itation. By comparing the performance of the 1.3- m MCLED’s with that of the 990-nm devices we conclude that the limiting factor on the speed seems to be a lack of carrier confinement in the quantum wells and not a cavity effect. Index Terms— Communication systems, frequency response, light-emitting diodes, microcavity, quantum-well devices. I. INTRODUCTION P LANAR microcavity light-emitting diodes (MCLED’s) are formed by a vertical Fabry–Perot cavity containing an active layer. In the “weak coupling” regime, the fundamental mode of the confined dimension of the cavity corresponds to the emission peak wavelength of the active medium. This interference effect selects the emitted mode, increases its power at the expense of the other modes and modifies its spatial distribution. Consequently, MCLED’s feature an in- crease of the extraction efficiency, of the directionality of the emission and a narrowing of the emission spectrum [1]–[4]. These advantages over standard light emitting diodes make the MCLED’s very attractive for applications such as lighting [5], interconnects [6], gas sensing [7], and for fiber-optic communication at m [8] and from to m [9], [10]. Though 1.3- m MCLED’s have already been demonstrated [10]–[12] there has been no report to date of their amplitude modulation (AM) speed, a parameter of special importance in many applications at m. The purpose of this paper is to evaluate the modulation performance of the state-of-art 1.3- m MCLED’s and to investigate whether the use of a micro-cavity design introduces any limitations of the operating speed compared to conventional LED’s. Reduction of the speed could occur as a result of photon recycling, a mechanism which has been inferred in MCLED operating at nm [13]. We present the dependence of the small- signal bandwidth with bias current for a 1.3- m MCLED. Sample fabrication is described in the next paragraph, after Manuscript received April 14, 1999; revised August 2, 1999. This work was supported by European Contract ESPRIT SMILED 24997. P. Landais is with Optronics Ireland, Physics Department, Trinity College Dublin, Dublin 2, Ireland. B. Roycroft, A. Shaw, and J. Hegarty are with the Physics Department, Trinity College Dublin, Dublin 2, Ireland. B. Depreter and I. Moerman are with the University of Gent-IMEC, INTEC Department, B-9000 Gent, Belgium. Publisher Item Identifier S 1041-1135(99)08678-4. which the experimental setup is introduced. Finally the results achieved are presented and discussed. They show that the dominant effect in the limitation of the bandwidth (BW) is the lack of carrier confinement and not any cavity effect such as photon recycling. II. DEVICE STRUCTURE AND FABRICATION The components under test are 1.3 m substrate emitting MCLED’s. Their structure is a planar -scale Fabry–Perot cav- ity containing three In Ga As P /InP quantum wells (QW’s), grown by metal-organic chemical vapor deposition. The epitaxial growth was performed in two steps on an n- type InP substrate allowing precise positioning of the QW’s with respect to the cavity resonance. In the first growth step, the n-doped distributed Bragg reflector (DBR) consisting of 15-pairs of In Ga As P –InP layers was grown at low pressure. In the second step, the active layers and the p- type layers were grown. Growth of the p-type layers under atmospheric pressure permits a better incorporation of the Zn doping. The top mirror is a 200-nm evaporated layer of Au which also acts as a p-type contact. A circular mesa was etched to ensure current confinement. The n-type contact consists of two gold stripes on the bottom of the samples. An additional antireflection (AR) coating is added to minimize reflection at the air/substrate interface. At room temperature 85- m devices show a quantum efficiency of 5% at 1 mA and a maximum output power of 750 W at 100 mA. III. EXPERIMENTAL RESULTS AND DISCUSSION Fig. 1 shows the optical spectra of a 115- m diameter device for three values of bias current: 1, 10, and 25 mA. The emission consists of a main peak centered at 1281.6 nm with a linewidth of approximately 18 nm and a secondary peak at 1225 nm. The main line shape is largely independent of bias. The spectrum of the main line is shown on a linear scale in the inset. It is clear at 25 mA there is an added spectral modulation on the line with a separation between peaks of this submodulation of 3 nm, which corresponds to an optical thickness of approximately 100 m in InP. This implies that the substrate is acting as the spacer layer of a secondary Fabry–Perot. A small-signal analysis has been carried out on 25-, 40-, 70-, and 115- m diameter samples. No heat sink was used. A 5-dBm modulation power provided by the output of a 0.3–3000-MHz bandwidth network analyzer is superposed on a dc bias. This electrical signal is applied to the selected sample via a 20-GHz electrical probe. Emission is collected 1041–1135/99$10.00  1999 IEEE