Light emission profiles from AlGaInP LEDs are compared with those obtained via Monte-Carlo simulation and a lossy transmission line model. Results show good agreement for devices with thin current spreading layers. However, for thicker layers, geometric effects dominate the output profile, resulting in discrepancies between experimental data and the transmission line model. The simulation of emission profiles confirms this discrepancy, caused by light escaping from below the p-contact.
Introduction: The generation of light in the active region of an LED is a spontaneous and random process. The probability of a photon escaping the chip depends on the direction of photon emission with associated absorption and reflection processes. A Monte-Carlo ray tracing simulation has been developed to calculate the spatial photon density for various LED structures. Similar simulations have been reported [1–3]; however, this work specifically considers the AlGaInP material system and more importantly incorporates the current spreading transmission line model of Morgan et al. [4]. The noninvasive direct measurement of current spreading in the active region as a result of a current spreading layer is not possible. Therefore, the only practical experimental probe of current spreading is the measurement of light output. Previously, it has been assumed that the current injected at a given point in the active region is proportional to the light output from the device at that point. However, experimental data of light emission from low sheet resistance (Rsh) thick current spreading layers has been inconsistent with our current spreading transmission line model data but consistent with similar data produced by Fletcher et al. [5], who confirm the exponential form of current spreading for thin current spreading layers, but not for thick layers. The MonteCarlo ray tracing simulation of emission profiles confirms the discrepancy between the transmission line model and experimental data as a real effect, dominated by device geometry. Nonlinearity in the detection system was assumed to be the origin of the discrepancy. By incorporating the transmission line model into the Monte-Carlo simulation, good agreement between simulated and experimental profiles can be shown for a range of AlGaInP devices with different material properties and current spreading layer thicknesses. Simulation: The Monte-Carlo simulation is one-dimensional, therefore alterations to the two-dimensional model described previously [6] are made. A Monte-Carlo simulation is a useful method for modelling light emission from LEDs. Photons are represented as particles and originate in the active region at random positions weighted against the current distribution in the active region using the one-dimensional model for current spreading. The simulation then follows the random trajectory of a single photon generated in the active region through the device, taking into account most loss mechanisms such as absorption, scattering and interaction with semiconductor–semiconductor and semiconductor–air interfaces. This is repeated 108 times to give an overall view of device performance. Experiment: Light emission profiles were obtained using a 5 millionpixel digital camera mounted onto a microscope with 100 objective lens. Comparisons were made at a constant drive current of 15 mA for a 500 mm square device. The camera is remotely triggered; this ensured that each device was subjected to constant heating under DC operation. The devices used in this study were those described in [7] and consisted of an n-type GaAs substrate with an AlGaInP double-heterojunction active region. Two different current spreading layers were investigated; a thin, high sheet resistance GaAs p-layer (thickness t ¼ 30 nm, Rsh ’ 1000 O) and a thick, low sheet resistance GaP p-layer (t ¼ 5 mm, Rsh ’ 10 O).
1.2
GaP CSL-TLM GaP CSL-EXP GaP CSL-MC
contact
normalised intensity, a.u.
J. Kettle and R.M. Perks
expected, the light output falls exponentially from the p-contact, consistent with the transmission line model. The injected current is greatest below the contact; however, the current spreading layer is so thin, very little light escapes from below the contact. Light propagating towards the contact is assumed to be reflected back to the absorbing substrate or absorbed in the contact itself. Fig. 1 (right hand side) shows the emission profiles associated with the low sheet resistance current spreading layer device. Here, the fall-off rate of the experimental data and transmission line model is inconsistent. The current spreading layer in this device is much thicker. The effect of this is twofold: first, the current spreading is more uniform; secondly, the thicker current spreading layer allows for geometric effects to dominate the emission profile. Clearly the functional form of the experimental and Monte-Carlo profiles for this device is not exponential. It had been assumed that this shouldering effect in the emission profile was due to either localised saturation and heating [8] or surface recombination [9]. Surface recombination is not included in the Monte-Carlo simulation and thus is ruled out as having significant influence on the profile. Likewise, the general form of the profile is maintained for low drive currents, excluding local saturation effects. The fact that the current spreading layer is thicker in this case allows more light to emerge from the geometrical shadow; adding to the light emission around the contact. This may also be explained in terms of simple geometry, as illustrated in Fig. 2. As the thickness of the current spreading layer tends to zero, the diameter of the region of the active region that is completely shadowed by the contact (physical shadow) tends towards the diameter of the contact (2r0 where r0 is the contact radius). As the current spreading layer thickness increases, the diameter of this shadow zone decreases (see inset of Fig. 2). This effectively exposes more of the active region that can couple photons through the surface.
1.0 0.8
5 mm GaP CSL
no CSL-EXP no CSL-MC
0.6
no CSL-TLM
0.4
no effective CSL
0.2 0 -50
50 150 250 position at surface of die, mm
350
Fig. 1 Normalised light output profiles: experimental Monte-Carlo and transmission line model against current injected into active region for (left) 30 nm GaAs current spreading layer (CSL) using resistivity ¼ 0.02 Ocm and diode conductance ¼ 4.5 102 S and (right) 5 mm GaP current spreading layer using resistivity ¼ 0.02 Ocm and diode conductance ¼ 3.2 102 S 2r0 105
diameter of physical shadow D, mm
Monte-Carlo based simulation of surface light emission profiles from AlGaInP light emitting diodes
t1
a
100
95
D1
t1 > t2 D1 < D2
2r0 90 a
85
0
5 10 15 CSL thickness, mm
t2
20
D2 geometric shadow
Results and discussion: Fig. 1 (left hand side) shows the spatial emission profiles associated with the high sheet resistance current spreading layer device. Experimental data, Monte-Carlo simulation and the transmission line model show excellent agreement. As
ELECTRONICS LETTERS 27th April 2006
Fig. 2 Variation in diameter of physical shadow, D, against current spreading layer thickness a is critical angle of total internal reflection for semiconductor air interface (for GaP a ’ 19.47 )
Vol. 42 No. 9
Conclusions: Two conclusions can be drawn from this work. First, the effectiveness of the transmission line model of Morgan et al. [4] is highlighted. By considering the trajectories of the photons when they leave the active region and propagate out of the die surface, a profile consistent with experimental data can be obtained irrespective of current spreading layer thickness. Secondly, the shape of the emission profile for the thick GaP current spreading layer has been shown to be due to light generated from underneath the contact escaping through the surface. This has the effect of pushing the profile up around the contact, resulting in the inverted curve as opposed to the expected exponential form. Geometric effects dominate the emission profile over the current distribution at thicker current spreading layers.
3 4 5
6
7
# The Institution of Engineering and Technology 2006 15 February 2006 Electronics Letters online no: 20060463 doi: 10.1049/el:20060463
8
J. Kettle (Multidisciplinary Nanotechnology Centre, School of Engineering, Swansea University, Singleton Park, Swansea SA2 8PP, United Kingdom)
9
Lee, S.J., Badano, A., and Kanicki, J.: ‘Monte Carlo modeling of the light transport in polymer light-emitting devices on plastic substrates’, IEEE J. Sel. Top. Quantum Electron., 2004, 10, (1), pp. 37–44 Morgan, D.V., Al-Ofi, I.M., and Aliyu, Y.H.: ‘Indium tin oxide spreading layers for AlGaInP visible LEDs’, Semicond. Sci. Technol., 2000, 15, (62), pp. 67–72 Fletcher, R.M., Kuo, C.P., Osentowski, T.D., Huang, K.H., and Craford, M.G.: ‘The growth and properties of high performance A1InGap emitters using lattice mismatched GaP window layers’, J. Electron. Mater., 1991, 20, (12), p. 1125 Porch, A., and Morgan, D.V.: ‘Analysis of current spreading in transparent current spreading layers of finite size 5-7’ in Porch, A., Perks, R.M., Morgan, D.V. (Eds.): ‘Compound semiconductor devices and integrated circuits’ (Wocsdice, 2005, ISBN 0863415164), pp. 5–6 Aliyu, Y.H., Morgan, D.V., Thomas, H., and Bland, S.W.: ‘A1GaInP LEDs using reactive thermally evaporated transparent conducting indium tin oxide (ITO)’, Electron. Lett., 1995, 31, (25), pp. 2210–2212 Morgan, D.V., Porch, A., Perks, R.M., Jones, M.O., and Edwards, P.P.: ‘Transparent current spreading layers for optoelectronic devices’, J. Appl. Phys., 2004, 96, (8), pp. 4211–4218 Gessmann, T.H., and Schubert, E.F.: ‘High-efficiency AlGaInP lightemitting diodes for solid-state lighting applications’, J. Appl. Phys., 2004, 95, (5), pp. 2203–2216
R.M. Perks (Cardiff School of Engineering, Cardiff University, Queens Buildings, The Parade, Cardiff CF24 3AA, United Kingdom) E-mail:
[email protected] References 1 2
Lee, S.J.: ‘Analysis of light-emitting diodes by Monte Carlo photon simulation’, Appl. Opt., 2001, 40, (9), pp. 1427–1437 Hu, F., Qian, K.-Y., and Luo, Y.: ‘Far-field pattern simulation of flip-chip bonded power light-emitting diodes by a Monte Carlo photon-tracing method’, Appl. Opt., 2005, 44, (14), pp. 2768–2771
ELECTRONICS LETTERS 27th April 2006
Vol. 42 No. 9
