Vol. 5, No. 3, 1989
H E G Amot, M Watt, C M Sotomayor-Torres,
R Glew*, R Cusco**, J Bates and S P Beaumont.
Nanoelectronics Research Centre, Dept of Electrical and Electronic Engineering, Uni~crsity of Glasgow, Glasgow, G12 SQQ, U.K. *STC Technology Ltd, London Road, Harlow, Essex CM17 9NA. **Universitat Autonoma de Barcelona Dept Fisica, Bellaterra, Spain. Received
8th August 1988.
Abstract :Buried G&s-GaAlAs quantum dots have been grown by MOCVD for the first time . Free -standing quantum dots were formed in GaAs-GaAlAs single quantum well material by a combination of electron beam lithography and dry etching, then the dots were buried by regrowing GaAlAs. Low temperature (6OK) photoluminescence spectra shows an emission which is increased in intensity and shifted to higher energies from the dots after they are buried. The efficiency of the emission scales approximately with the volume of material remaining after patterning. Keywords : GaAs/AlGaAs Quantum Dots, Photoluminescence,
Introduction Recently it has been suggested that substantial improvements to semiconductor laser operation can be made by confining the carriers in three dimensions, as in a quantum box or dot lasert. To date no-one has constructed a quantum dot laser, although attempts have been made2. One way to achieve this is to define the quantum dots by electron beam lithography followed by reactive ion etching (RIE) and then to regrow over the dots. The main drawback to this method could possibly be the RIE damage. The first step towards fabrication is therefore to investigate the optical behaviour of the buried quantum dots (QDs). Here we report on the first all MOCVD buried quantum dots. Experiment MOCVD growth was performed in an atmospheric pressure horizontal reactor equipped with a vent run manifold. The starting material consists of a single lOOA quantum well in GQ.,AI,,,AS grown at 750’~, SOOA from the surface.
Patterning was performed by electron-beam lithography using a modified Philips PSEM-500 scanning electron microscope with an 8OA spot at 50kV. A negative resist (HRN) was used as this could be removed before regrowth by oxygen plasma etching in a Plasmafab 505 oxygen barrel etcher. The free standing dots were defined by reactive ion etching (RIE) with either silicon tetrachknide (SiCu using a Plasmatechnology RIESO machine opemted at 13.56MHz at a pressure of lOmT, a d.c. bias of 300V and a power density of 0.65W/cm2 or methane hydrogen (CI&/H2) using a Elecuotech SRS Plasmafab 340 machine with a gas ratio 1:5, pressure 15mT. d.c. bias of 1OOOVand a power density of 0.75W/cm2. The etch rate of GaAsGaAlAs in Sic14 is O.Zkm/min and in CHJH2 is 25OAlmin. Several dot sizes were investigated - nominally 700, 1100, and 3500A diameter dots in 100~m2 arrays with a separation between the dots of 3000, 5000 and 1OOOOA respectively (figure 1). The smaller dots were circular but the larger dots square. Two identical samples were fabricated but different gases - SiCl, and CH..+/H2were used in the reactive ion etching. The SiC14 etched dots
0 1989 Academic Press Limited
Vol. 5, No. 3. 1989
Figure 1 SEM micrographs of 3500A diameter quantum dots after RIE with a)SiCL b)CH&lz
Figure 2 SEM micrographs of 3500A diameter quantum dots after regrowth with 0.2pm of Gac6Alc,4As on a)SiCb and b)CH4/H2 etched dots
were three times deeper (3OOOA) than those etched in CH&-I> Beside each 1OOpm2array of dots was a 100pm2 unpatterned mesa to provide a control. Complete coverage of the dots was achieved by overgrowing by MOCVD with
0.2 t.trn of Gae,bA1e,,As and a 4OOA GaAs cap (figure 2) at 750°C. Low temperature (60K) photoluminescence (PL) spectra were obtained by exciting with the 633nm line of a
Vol. 5, No. 3, 7989
Figure 3 Quantum Well photoluminescence emission from SiCl, etched mesa (A) before and (B) after regrowth.
Helium Neon laser focused to a 50um spot. The overgrown layer was transparent to the laser wavelength. The typical spectral resolution was lOA. Results
Prior to regrowth, quantum well (QW) emission was obtained only from the mesas on each sample. It is expected that a significant contribution to the PL intensity will come from the photoexcited carriers excited in the Ga0,7Alc,sAs barriers which then recombine in the GaAs well. After regrowth over the SiCldetched QDs QW emission was recovered in the 3500A dots and remained in the mesas but in the CHdHa etched dots and mesas no quantum well emission was obtained. The photoluminescence spectra for the SiCl, etched sample are shown in figures 3 and 4. Spectrum A is the QW emission from the mesa before regrowth and spectrum
B after. The emission has shifted approximately 30A to higher energies and the intensity has decreased by a factor of about 7 after regrowth. This reduction in intensity may be due to nonradiative traps present in the regrown interface. The shift may be due to vertical diffusion of the aluminium into the QW on regrowth. Spectra C and D are the QW emissions from the 35OOA QDs before and after regrowth. Spectrum E corresponds to the adjacent mesa. Both the 35OOA QDs and the lOOurn* mesa show an emission peak at 796OA thus the QW emission from the dots has been recovered. The luminescence efficiency scales approximately with the volume of QW material remaining in the dots relative to that of the overgrown mesa. It is likely that no emission was seen from the dots after RIE due to significant sidewall damages. It is also suggested that MOCVD material may be more sensitive to the reactive ion etch damage as QW emission from dots fabricated on MBE
Vol. 5, No. 3, 1989
Figure 4 Quantum Well photoluminescence emission from the SiCl, etched patterns. Spectrum C - ZOOA dots before regrowth and Spectrum D after regrowth.
material can be readily obtain&. Emission is recovered after regrowth due to the annealing out of the sidewall damage at the regrowth temperature. It may only be seen inthelargerdotsas: a) RIR could completely destroy the QW in the smaller dots whereas the greater volume of material in the larger dots would be able to recover sufficiently on anneahng to give a measurable signal b) the effect could occur in the smaller dots but the signal is too weak compared to the background c) on regrowth the aluminium can diffuse completely through the smaller quantum dots destroying any structure in the material.
Spectrum E -emission from the mesa after regrowth. The rising signal in spectrum C is part of the high energy tail from the CiaAs Substrate emission.
The line broadening of the dot emission by a factor of 1.3 relative to the mesa is probably due to the RIE damage which is more significant for the 3500A dots than for the mesa due to the increased surface to volume ratio of the dots. On the microscopic scale there will be variations in the bandgap which result in the broadened peak. It is unclear why [email protected]
RIE QDs show no emission before or after regrowth. The most uniform regrowth seems to have been on the CHdH, etched sample (see figure 2). However in this case, this is most likely to be a depth effect because the CH&-I, etched dots are only one third as deep as those etched by SiC14. In a previous experiment where the
Vol. 5, No. 3, 1989
regrown layer on 35OOA SiCl, etched dots was thicker, the overgrown profile of the dots was almost identical to that of the shallower overgrown CH& etched dots reported heres. It was also found that an increase in growth time resulted in planarisation of the surface because the growth between the dots is faster than the growth on the dots. For fabrication of a quantum dot laser this is important for the regrowth of the remaining laser structure.
including CASE awards for H Arnot with GEC Hirst Research Centre and J Bates with RSRE Malvem and a SERC IT studentship for M Watt. R Cusco acknowledges financial support from the British Council and Glasgow University,
References Conclusion Regrowth of GaAlAs over SiCl, and C&./H, etched quantum dots has been successfully achieved. Photoluminescence emission is recovered consistently in the QDs etched by SiCL after regrowth, due to the annealing out of the RIE damage and the intensity scales with the volume of material remaining after patterning. However, MBE dots show PL after RIE unlike MOCVD dots, suggesting that the way to proceed is to fabricate dots on MBE material and then to overgrow by MOCVD. This is currently being investigated. Further work is needed to understand why RIE with different gases affects MBE and MOCVD material differently. Acknowledgements -This work was supported by the Science and Engineering Research Council (U.K.) grants
1 see for example M. Asada, Y Miyamoto & Y Suehatsu, IEEE Quantum Electronics - 22, 1915, (1986). 2 Y Miyamoto, M Cao, Y Shingai, K Furuya, Y Suematsu, K G Ravikumar & S Arai, Japanese Journal of Applied Physics - 26, L225, (1987). 3 S Thorns, S P Beaumont, C D Wilkinson, J Frost and C R Stanley in Microelectronic Engineering 5, 249 (1986) North Holland. Elsevier Science Publishers B.V. 4 H. Amot, S R Andrews and S P Beaumont published - Microelectronics 88, Vienna
5 R Glew, H Arnot, S P Beaumont, J Lambkin & D Dunstan, presented at the 4th Int Conf on MVPE, Hakone, Japan, May 1988, Poster P-81 and to be published.