Journal of Crystal Growth 230 (2001) 481–486

Luminescence studies of defects and piezoelectric fields inInGaN/GaN single quantum wells

S.J. Henleya,*, A. Bewicka, D. Chernsa, F.A. Ponceb

aH.H. Wills Physics Laboratory, University of Bristol, Tyndall Avenue, Bristol BS8 1TL, UKbDepartment of Physics and Astronomy, Arizona State University, Tempe, AZ, 85287, USA

Abstract

Transmission electron microscopy (TEM), cathodoluminescence in the scanning electron microscope (SEM-CL) and

photoluminescence (PL) studies were performed on a 30 nm GaN/2 nm In0.28Ga 0.72N/2 mm GaN/(0 0 0 1) sapphiresingle quantum well (SQW) sample. SEM-CL was performed at low temperatures � 8K, and at an optimumaccelerating voltage, around 4–6 kV to maximise the quantum well (QW) luminescence. The CL in the vicinity of

characteristic ‘‘V-shaped’’ pits was investigated. The near band edge (BE) luminescence maps from the GaN showedbright rings inside the boundaries of the pits while the QW luminescence maps showed pits to be regions of lowintensity. These observations are consistent with TEM observations showing the absence of QW material in the pits.

Variations in both the BE and QWmaps in the regions between the pits are ascribed to threading edge dislocations. TheCL and PL QW luminescence was observed to blue-shift and broaden with increasing excitation intensity. This wasaccompanied by decreasing spatial resolution in the CL QW maps implying an increasing carrier diffusion length in theInGaN layer. The reasons for this behaviour are discussed. It is argued that screening of the piezoelectric field in the

material may account for these observations. # 2001 Elsevier Science B.V. All rights reserved.

Keywords: A1. Defects; A3. Quantum wells; B2. Piezoelectric materials; B2. Semiconducting III–V materials

1. Introduction

The luminescence properties of low-dimensionalheterostructures of the ternary alloys InGaN andAlGaN with GaN have recently become of greatcommercial and scientific interest [1–3]. The rapidadvances in the hetero-epitaxy of the group-IIInitrides has facilitated the production of lightemitting diodes and laser diodes that operateacross a wide spectral range [1,2]. The effect of

defects on the luminescence in these materials isstill poorly understood. The defects have beenshown to act as non-radiative recombinationcentres [4]. However a range of other factorsincluding piezoelectric fields, non-uniformity ofthe quantum well (QW) and excitation intensityaffect the luminescence from these materials [5–7].The aim of this paper is to clarify the role of thesefactors.

2. Experimental

In this paper we discuss a series of catho-doluminescence (CL) and photoluminescence

*Corresponding author. Tel.: +44-0117-9288750.

E-mail address: simon.henley@bristol.ac.uk (S.J. Henley).

0022-0248/01/$ - see front matter # 2001 Elsevier Science B.V. All rights reserved.

PII: S 0 0 2 2 - 0 2 4 8 ( 0 1 ) 0 1 2 4 5 - 3

(PL) measurements in a 30 nm GaN/2 nmIn0.28Ga0.72N/2 mmGaN/(0 0 0 1) sapphire singlequantum well (SQW) sample grown by MOCVDat temperatures of 10508C (GaN) and 8008C(InGaN).The CL measurements were carried out on a

modified Coates and Welter CWICSCAN SEM.This is a first generation cold field-emission gunscanning electron microscope. (FEG-SEM) produ-cing 1 nA probes of 3.5 nm at 15 kV and 20 nm at1 kV. The light collection optics are an OxfordInstruments aluminium parabolic mirror con-nected via fibre optic cable to a grating spectro-meter and a Peltier-cooled PMT operating inphoton counting mode. The temperature of thesample (�8K5T5300K) is controlled by anOxford Instruments liquid He cryostat. Thescanning and detector systems are computercontrolled allowing digital image acquisition.PL measurements were carried out on a

Renishaw Raman System 2000 using a nUVHe-Cd laser. The sample temperature was alsocontrolled with a Oxford Instruments liquid Hecryostat.The defect structure in the material was

examined by transmission electron microscopy(TEM) in a Philips EM430. Specimens wereprepared in plan-view and cross-section by backthinning using mechanical polishing and ion-thinning.

3. Results and discussion

3.1. TEM results

The microstructure of the material, was exam-ined by TEM, and is characterised by an array of‘‘V-shaped’’ pits [8,9] with two distinct size ranges,namely larger pits (approx. 2 mm in diameter) anda higher density of smaller pits (approx. 50.3 mmin diameter). The larger pits have been shown tooriginate from near the GaN/sapphire interfaceand cut through the QW. The smaller pits, mostlyassociated with threading dislocations and nano-pipes, originate from just below the QW. Fig. 1shows a cross-sectional TEM image in which thesefeatures can be seen. It is suggested that the

growth of these pits initiates after the growthinterrupt associated with the growing of the QWand subsequently spread out into pits [8,9]. InFig. 1 the strong contrast in the region containingthe InGaN layer is due to lattice displacementcaused by the larger interplanar spacing of InGaNalong [0 0 0 1] compared to that in GaN. From theabsence of this contrast from the region of the pitit can be concluded that no InGaN grows on theside walls of the pit. Moreover the uniformity ofthe contrast from the flat regions suggests that theQW is uniform on scales less than 50 (A [10]. Otherthreading dislocations were observed between thepits. A standard analysis showed these dislocationswere all of edge type with 1

3h1 1%2 0i Burgers

vectors.

3.2. CL results

Fig. 2 shows CL spectra taken from the sampleat 8K for a range of accelerating voltages. Thepeaks in the spectra are labelled. The near bandedge, D0X, luminescence from the GaN is labelledBE. The QW luminescence is labelled QW and theso-called ‘‘yellow luminescence’’ is labelled YL. Itcan be seen that an optimum voltage � 4–6 kVcan be chosen such that the QW luminescence is ata maximum due to the electron beam dumping themaximal amount of charge into the vicinity of theQW. The yellow luminescence was observed toincrease as the electron beam penetration, into thesample, increases. Fig. 3 shows CL spectra, taken

Fig. 1. Threading dislocations and a pit. A dislocation can be

seen terminating at the base of the pit.

S.J. Henley et al. / Journal of Crystal Growth 230 (2001) 481–486482

at 8K, at various electron beam currents. Thepeak of the QW luminescence varies with the beamcurrent used. As the beam current is increased theQW luminescence is observed to broaden andblue-shift. This blue shift was observed to remainafter the beam current was decreased. In PLspectra, taken at 8K at various laser intensities,

the QW peak was also observed to blue-shift in asimilar way with increasing excitation intensity.This effect was temporary however and the lowpower spectrum was regained when the intensitywas reduced.Fig. 4a shows a SEM secondary electron image

of an area containing one large and many smallpits. CL maps, at different wavelengths, of thesame area as Fig. 4a are shown in Figs. 4b–e. TheQW luminescence map at 460 nm and low beamcurrent (Fig. 4b) shows that there appears to be noQW luminescence from the pits as they appearblack in the map. CL linescans across large pitsshows that there is a rapid drop off of the QWluminescence towards the centre of the pit. Thisdrop off is attributed to the diffusion of carriers,generated in the GaN side walls of the pit, into theQW at the edge. From these linescans, and thegeometry of the pits, it was possible to estimate thecarrier diffusion length in GaN as 100–200 nm, inline with other measurements [13,14]. Simulationsof the diffusion of carriers from the GaN, from thepit wall, into the QW material at the edgeproduced very similar linescans to those observed.The BE luminescence map, from the GaN, at

358 nm (Fig. 4c) shows bright ‘‘rings’’ of lumines-cence, inside the edge of the pits, around the

Fig. 3. The blue shift with increasing beam current of the QW peak. CL spectra taken at 10K and 5kV.

Fig. 2. CL spectra taken from the sample at 8K for a range of

accelerating voltages.

S.J. Henley et al. / Journal of Crystal Growth 230 (2001) 481–486 483

centre. This shape is thought to be caused bydifferent factors. Near the edge of the pit therecombination in the GaN is reduced due topreferential recombination, in the QW, of excitonsthat have diffused as far as the QW. Nearer thecentre of the pit the luminescence is reduced by theside walls of the pit reducing the solid angleavailable for collection by the CL mirror. There isalso possible non-radiative recombination at the

centre of the pit due to the presence of adislocation at the bottom [4] (see Fig. 1).CL maps taken at the high energy shoulder of

the QW luminescence peak (445 nm) before andafter high beam current irradiation of the areastudied (Figs. 4d and e) show that the blue shift, ofthe QW peak, has occurred during the irradiation.Both maps were taken using the same beamcurrent. Some bright areas are visible in the map

Fig. 4. Images taken at 5 kV and 8K. (a) SEM secondary electron image; (b) CL map at 460 nm (QW) before high beam current

irradiation; (c) CL map at 358 nm (BE); (d) CL map at 445 nm (high-energy shoulder of QW peak) before irradiation; (e) CL map at

445 nm after irradiation.

S.J. Henley et al. / Journal of Crystal Growth 230 (2001) 481–486484

taken before the irradiation (Fig. 4d). It is alsoobserved that the spatial resolution in the QWmaps decreases after high beam current irradia-tion. This can be observed by comparing the CLmaps in Figs. 4b and d.A blue shift of the QW luminescence peak, with

increasing excitation, has been attributed toscreening, by newly generated carriers, of thestrain-induced piezoelectric field that is presentacross the well [6]. However, the screening of thepiezoelectric field alone cannot satisfactorily ex-plain the broadening of the QW peak.The QW peak is broad, compared to the BE

peak, even at low excitation intensity. This isevidence for indium or well width fluctuations.These fluctuations are thought to occur on scalesless than 50 (A [10]. The areas of high indiumcontent will have a smaller band-gap, and a higherpiezoelectric field, than the indium sparse regions.These variations would be expected to cause abroad luminescence peak. As the excitation densityis increased the QW peak was observed to broadendramatically. The indium fluctuation would beexpected to cause localisation of carriers at indiumrich regions [11,12]. As these potential minima arefilled carrier recombination will occur morefrequently in the indium sparse regions causing abroadening of the QW peak. However this is notsupported by spot mode CL spectra from differentareas in the studied region. The local spectrasimply show a varying shift and no significantbroadening of the QW peak. The broadeningobserved must be attributed to the blue shiftvarying from area to area. The variation of theblue shift from different areas on the film can beobserved by careful consideration of the CL mapsat different wavelengths. The spatial variation ofthe blue shift is on a scale larger than that of theindium fluctuations. This variation could beattributed to different amounts of screening ofthe piezoelectric field. The density of trap states forcharge carriers (e.g. impurity atoms) may varyacross the film. Localised charge will screen thepiezoelectric field. It is suggested that the differ-ence, in the duration of the peak shift, between theCL and PL blue shifts also supports this idea.SEM-CL introduces net charge into the sample,from the electron beam, whereas PL generates no

overall charging in the sample. This trappedcharge will have no carriers of the opposite chargeto recombine with so the peak shift would beexpected to remain until the net charge haddiffused away. It is suggested that the main effectof the indium fluctuation be in controlling thecarrier diffusion length in the QW. The effect oflocal potential minima would be to trap chargecarriers. As these states are filled the diffusionlength would be expected to increase. This isobserved as a decrease in the spatial resolution ofthe QW CL maps after high beam currentirradiation (see Fig. 4).

4. Conclusion

In summary it has been shown that the presenceof piezoelectric fields and fluctuations in indiumcontent in InGaN/GaN SQW’s have a large effecton the luminescence from these structures. Screen-ing of the piezoelectric field, due to localisedcharge carriers, can cause an excitation dependentblue shift of the QW peak. The blue shift wasshown to remain after CL excitation but not afterPL excitation agreeing with the hypothesis thattrapped charge carriers are causing the screeningof the piezoelectric field. It was suggested that thefluctuations in the indium composition can causevariation in the carrier diffusion length in the QW.The characteristic ‘‘V shaped’’ pits in these

samples reduce the QW luminescence as noquantum well material grows on the side walls ofthe pits. The growth of these pits has been shownto be associated with threading dislocations and itis argued that screw type threading dislocationsinitiate the growth of these pits.

Acknowledgements

The authors are grateful to Prof. F. A. Ponce(ASU) for provision of the sample and to Dr. JohnDay (IAC, University of Bristol) for softwaredevelopment. We are grateful for financial supportfrom the EPSRC (Grant GR/M03030), fromNATO (Grant No. 960690) and from Philips UK(for SJH).

S.J. Henley et al. / Journal of Crystal Growth 230 (2001) 481–486 485

References

[1] G. Fasol, S. Nakamura, The Blue Laser Diode, Springer,

Berlin, 1997.

[2] S. Nakamura, M. Senoh, S. Nagahama, N. Iwasa, T.

Yamada, T. Matsushita, H. Kiyoku, Y. Sugimoto, Appl.

Phys. Lett. 68 (1996) 3269.

[3] K.P. O’donnell, W. Van der Stricht, R.W. Martin, P.G.

Middleton, App. Phys. Lett. 74 (1998) 215.

[4] T. Sugahara, H. Sato, M. Hao, Y. Naoi, S. Kurai,

S. Tottori, K. Yamashita, K. Nishino, L. Romano,

S. Sakai, Jpn. J. Appl. Phys. 37 (1998) L398.

[5] L.-H. Peng, C.-W. Chuang, L.-H. Lou, App. Phys. Lett. 74

(1999) 795.

[6] T. Takeuchi, S. Sota, M. Katsuragawa, M. Komori,

H. Takeuchi, H. Amano, I. Akasaki, Jpn. J. Appl. Phys. 36

(1997) L382.

[7] C. Wetzel, S. Nitta, T. Takeuchi, S. Yamaguchi,

H. Amano, I. Akasaki, MRS Internet J. Nitride Semicond.

Res. 3 (1998) 31.

[8] I. Kim, H. Park, Y. Park, T. Kim, App. Phys. Lett. 73

(1998) 1634.

[9] X.H. Wu, C.R. Elsass, A. Abare, M. Mack, S. Keller, P.M.

Petroff, S.P. DenBaar, J.S. Speck, App. Phys. Lett. 72

(1998) 692.

[10] F.A. Ponce, D. Cherns, W. Goetz, S. Kern, Mater. Res.

Soc. Conf. Proc. 482 (1998) 453.

[11] S. Chichibu, T. Azuhata, T. Sota, S. Nakamura, Appl.

Phys. Lett. 70 (21) (1997) 2822.

[12] S. Chichibu, K. Wada, S. Nakamura, Appl. Phys. Lett. 71

(16) (1997) 2346.

[13] S. Rosner, E. Carr, M. Ludowise, G. Girolami, H.

Erickson, Appl. Phys. Lett. 70 (4) (1997) 420.

[14] S. Rosner, G. Girolami, H. Marchand, P. Fini,

J. Ibbetson, L. Zhao, S. Keller, U. Mishra, S. DenBaars,

J. Speck, Appl. Phys. Lett. 74 (14) (1999) 2035.

S.J. Henley et al. / Journal of Crystal Growth 230 (2001) 481–486486