Fabrication of 50–100 nm Patterned InGaNBlue Light Emitting Heterostructures

Lu Chen1) (a), Aijun Yin (a), J.-S. Im (a), A. V. Nurmikko (a), J.M. Xu (a), and

J. Han (b)

(a) Division of Engineering and Department of Physics, Brown University, Providence,RI 02912, USA

(b) Department of Electrical Engineering, Yale University, New Haven, CT, USA

(Received July 10, 2001; accepted August 17, 2001)

Subject classification: 78.55.Cr; 78.67.De; 81.40.Tv; 85.60.Jb; S7.14

We have developed fabrication approaches and studied the optical properties of arrays of submi-crometer sized InGaN/GaN MQW posts, created by patterning and etching of epitaxially grownwafer material. Two approaches have been employed: a) electron beam lithography, by which postsof individual diameter of 100 nm have been realized; b) the use of nanometer-scale self-assembledtemplates of porous alumina as masks, to fabricate posts on a 50 nm scale. Robust violet lightemission at room temperature has been observed in such nanoscale structures.

Introduction Laterally ordered, patterned subwavelength sized nitride structures areof interest for novel blue/ultraviolet light emitters as spontaneous and stimulated emis-sion may be enhanced in these artificial structures. The enhancement can be anticipatedeither in the “photonic bandgap” sense, or due to proximity interaction of coupledelectronic and electromagnetic excitations in structures which possess the lattice con-stants in the sub-50 nm regime. In this work we investigate approaches to fabricatingsuch small period structures by resorting to electron beam lithography but also exploit-ing recently developed techniques for synthesizing self-organized porous alumina [1]which is here used as a pattern transfer mask for the fabrication of high density nano-scale arrays of posts in InGaN/GaN QW heterostructures. The aim of the project wasto study the light emitting capability of such subwavelength scale structures where ver-tical side walls of the isolated MQW posts were exposed defined by etching.

Fabrication As a test heterostructure, we employed an undoped three-QW InGaN/GaN structure, with 40 �A wells and 100 �A barriers. The Indium concentration was ap-proximately 7%. The top GaN cladding layer was approximately 50 nm thick. First,electron beam patterning was used to define the template for the square lattice of postswhich were subsequently produced by reactive ion etching in a Cl2/BCl3 mixture, em-ploying a nickel-based etch mask structure. An etch depth on the order of 400 nmensured that the etching reached well beyond the MQW section in depth. A scanniningelectron beam microscope image is shown in Fig. 1 of a portion of the array, with100 nm diameter posts separated by approximately 200 nm center-to-center spacing.The fabrication process for the second type of arrays of QW posts began with elec-

trolytically assisted growth of porous alumina, where the self-organization of pores oc-

1) Corresponding author; Phone: +1 401 863 2333; Fax: +1 401 863 1387;e-mail: Lu_Chen@brown.edu

phys. stat. sol. (a) 188, No. 1, 135–138 (2001)

# WILEY-VCH Verlag Berlin GmbH, 13086 Berlin, 2001 0031-8965/01/18811-0135 $ 17.50þ.50/0

curs in a hexagonal lattice with typical pore di-mension ranging from 20 to 50 nm. The micro-meter thick alumina template, of up to 1 cm2 inarea, was subsequently thinned to approximately400 nm by a chemical procedure, prior to its at-tachment on the GaN heterostructures. We havefound that for adhesion of the template and for

subsequent processing, it is very important for the GaN films and heterostructures tohave a superior surface morphology, with mean roughness not larger than 4 nm. Anickel-based multilayer etch mask was deposited through the pores onto the nitride sur-face. After lifting off the alumina template, chlorine-based reactive ion etch was used tocreate the 500 nm deep posts, isolating the QWs, with a diameter of �50 nm and latticeconstant of �100 nm. The SEM images of Fig. 2 show a plan view of a portion of sucha nanoprocessed InGaN/GaN QW sample on two different scales of magnification.

Optical Properties The optical performance of e-beam lithographically 100 nm writtenstructures and those of the 50 nm scale arrays of such posts have been compared withunpatterned planar wafer material by resorting to standard cw and time-resolved photo-luminescence measurements at room temperature. As a source of excitation, we em-ployed a continuous-wave, modelocked Nd:YAG laser, whose third harmonic at355 nm corresponds to a photon energy just below the absorption edge of the GaNbarrier layers. The excitation pulses were approximately 50 ps in duration, at a 75 MHzrepetition rate, focused to a spot of less than 5 mm in diameter. Although visible lumi-nescence was easy to see at low levels of excitation, we employed an excitation level ofapproximately 20 mJ/cm2 per pulse to ensure that there was a high probability that thelocalized states in the InGaN QWs were filled so that the majority of the e–h pairswould be in the extended states [2] (see below for additional discussion). Figure 3shows the steady state PL spectra of the three types of structures in the form of rawlaboratory data where the amplitudes are normalized, with the multiplication factors

136 Lu Chen et al.: Fabrication of InGaN Blue Light Emitting Heterostructures

Fig. 1. Scanning electron microscope image of a squarearray of InGaN/GaN MQW 100 nm diameter posts

Fig. 2. Left panel: Global plan view of the patterned InGaN/QW nanopost structure on a micro-meter scale. Right panel: Detail top view of the 50 nm diameter posts

indicating the measured PL intensities. As a first approximation the lab data needs tobe corrected by considering the actual areal filling factors of 14.5% and 22.7% for thee-beam written and porous alumina template etched samples, respectively, in both opti-cal absorption and PL emission. We then find that that the true amplitude ratios for theroom temperature PL efficiency for the three samples are approximately 1:11.9:2.8 forthe unpatterned QW, the e-beam written and porous alumina etched samples, respec-tively. That is, in this admittedly oversimplified picture, both the e-beam written andnanoporus templated InGaN MQW post samples exhibit PL efficiency that exceedsthat from the unpatterned material, presumably due to higher photon escape probabil-ity due to the post-like texture of the light emitting medium.To study the recombination process further, we performed time-resolved photo-lumi-

nescence experiments on the three types of samples by using a spectrometer/streakcamera combination. Figure 4 shows the transient PL emission recorded from the peakspectral region of each sample (at room temperature). From the data, we extract corre-sponding lifetimes of 345, 125 and 50 ps, respectively, for the unpatterned QW, thee-beam written and porous alumina etched samples. The lifetime is obtained from the1/e point of the transient PL traces and while the value for the nanopatterned templatesample is rather shortened, it is nonetheless reasonably well time resolved in the experi-ment.

phys. stat. sol. (a) 188, No. 1 (2001) 137

Fig. 3. Room temperature PL spec-tra of the three types of samples

Fig. 4. Time-resolved luminescence forthe upatterned QW and two posts ar-rays, respectively

The most important conclusion from our measurements is that the PL efficiency ofthe patterned 100 nm and 50 nm scale InGaN QW posts is very robust and not im-paired by the very large surface-to-volume ratio. That is, in strong contrast to conven-tional III–V and II–VI semiconductors, surface recombination at the etch-exposed verti-cal walls of the QW is modest at best. Under low photoexcitation (such ascorrepsonding to an LED) such a result would not be entirely unexpected, since elec-tron–hole pair localization in the random potential wells induced by the pronouncedindium concentration fluctuations could be argued to trap the carriers away from theinterface. However, under our conditions of intense photoexcitation (with an excess of1018 e–h pairs per cm—3 in bulk equivalent terms) the carriers have been shown toreside in extended states [2], in the regime of concentration which approximates thatencountered in a violet diode laser. The nanopatterned structures correspond to thecase where the feature size is less than the diffusion length of the carriers.There are several issues in the data of Figs. 3 and 4 that are being investigated

further and to which we do not have a full explananation at this writing. First, theorigin of the finite spectral blueshifts in Fig. 3 for the patterned structures is not clearand could involve multiple factors such as strain relief and piezoelectric fields, for ex-ample. Secondly, the apparent enhancement of the PL efficiency in the patterned struc-tures, while very real, is likely to have a complex origin due to the different opticalstructures involved, from the standpoint of light extraction efficiency alone. At the ele-mentary level, the exposed vertical walls of the QWs reduced total internal reflectionlosses, relative to the unpatterned QW sample. Beyond that, however, the 100 nme-beam lithographically defined array has characteristic dimensions on the scale of atwo-dimensional photonic crystal and work is underway to investigate the impact ofsuch effects. Finally, the 50 nm feature size array represents a regime where near opti-cal field interactions begin to be important and require their own systematic study as afunction of the individual post size and array periodicity and symmetry.

Acknowledgement The work at Brown University was supported by the NSF, DAR-PA, and ONR.

References

[1] J. Li, C. Papadopoulos, and J.M. Xu, Appl. Phys. Lett. 75, 367 (1999);Nature 402, 253 (1999).

[2] A. Vertikov, I. Ozden, and A.V. Nurmikko, J. Appl. Phys. 86, 4697 (1999)

138 Lu Chen et al.: Fabrication of InGaN Blue Light Emitting Heterostructures