InP quantum dots

ARTICLE IN PRESS

Journal of Crystal Growth 298 (2007) 586–590 www.elsevier.com/locate/jcrysgro

Tuning and understanding the emission characteristics of MOVPE-grown self-assembled InAs/InP quantum dots Bhavtosh Bansal, M.R. Gokhale, Arnab Bhattacharya, B.M. Arora Department of Condensed Matter Physics and Materials Science, Tata Institute of Fundamental Research, Homi Bhabha Road, Mumbai 400005, India Available online 8 December 2006

Abstract This paper examines the optical properties of self-assembled InAs/InP quantum dots grown by metalorganic vapour phase epitaxy (MOVPE). In a non-equilibrium regime of relatively low temperature (p450 1C) and higher growth rates (X1.5 monolayers/s), a rich variation in the peak emission wavelength, QD sizes, density and modality of the size distribution can be accomplished by changing the growth parameters. For example, the low-temperature peak emission wavelength of the ensemble can be tuned anywhere between 1.4 and 1.9 mm. Similarly, broadband emission with 250 meV bandwidth can be obtained from samples with bimodal dot distributions. The temperature dependent photoluminescence spectra show interesting dynamics associated with the thermally activated carrier transfer between dots of different ground state energies. r 2006 Elsevier B.V. All rights reserved. Keywords: A1. Nanostructures; A3. Metalorganic vapour phase epitaxy; A3. Quantum dot; B1: Semiconducting III–V materials

1. Introduction InP substrates [1–3] provide an interesting and technologically useful variant to the more canonical GaAs [4] substrates for growth of InAs quantum dots (QD) via the Stranski Krastanov route. Apart from a smaller band offset that allows for an emission in the 1.5 mm wavelength range, the anion exchange across the interface and a lower lattice mismatch makes the self-assembly less controlled and more susceptible to be affected by the growth kinetics parameters [5]. The self-assembled growth process is inherently statistical in nature, and the areal density, average size, size dispersion of the QD ensemble can be influenced by the interplay of energetics and kinetics of the growth process. The morphological characteristics of the ensemble in turn determine the electronic density of states and hence, the optical emission properties. Therefore, understanding and experimentally controlling the morphological characteristics of QDs have been a fundamental issue. In this paper, we discuss the influence of growth parameters on the optical emission from InAs/InP QDs Corresponding author.

E-mail address: [email protected] (B. Bansal). 0022-0248/$ - see front matter r 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2006.10.154

grown in a non-equilibrium regime of relatively low temperature (p450 1C) and higher growth rates (X1.5 monolayers (ML)/s). While many previous studies have accomplished a tuning of the emission wavelength by changing the barrier material [6], in this study, we shall demonstrate that the growth kinetics themselves provide a simple, viable and controllable means to drastically vary the dot sizes, modality of the dot-size distribution, and dot density. All these features have in turn allowed us to tune the peak emission wavelength of the ensemble over a wide range, and have also permitted us to modify the emission bandwidth. 2. Experimental procedure All the samples studied in this work were grown on epiready n+ doped (1 0 0) InP substrates using trimethylindium (TMIn) and arsine as group III and V sources in a horizontal reactor (CVD Inc.) at a pressure of 100 Torr with Pd–diffused hydrogen as the carrier gas. InAs layers were grown at a relatively low temperature of 430–450 1C. Before the InAs deposition, an InP-buffer layer was grown in three stages: first 500 A˚ at 625 1C and then with the temperature continuously ramped down to the InAs QD

ARTICLE IN PRESS B. Bansal et al. / Journal of Crystal Growth 298 (2007) 586–590

growth temperature and finally, another 500 A˚ at the stable temperature. To avoid switching transients, the TMIn flux for the buffer, the InAs, and the cap layers was kept the same. The QD growth was carried at the fixed V/III ratio of 50. The TMIn flow was varied between 20 and 55 Torrcm3/min across different growth runs. For a given set of growth conditions, a pair of samples was grown with identically deposited InAs layer in two growth runs. In the first case, the sample was immediately cooled and taken out of the reactor after InAs deposition itself to enable a study of surface morphology and, in the second case, an InP cap layer was grown for samples used for photoluminescence (PL) study. For these samples, about 50 A˚ InP cap was deposited at the InAs deposition temperature to minimize further ripening during the subsequent growth of the remainder of the cap at higher temperature. The morphology of the uncapped dots was characterized using a Nanoscope II atomic force microscope (AFM) in contact mode. PL spectra were recorded at 25 K with a 0.67 m McPherson grating monochromator and 325 nm helium–cadmium laser as the excitation source at a power density of 0.5 W/cm2. The measured spectra were corrected for the system response against a standard Oriel black body source heated to 1350 K. To understand the

587

optical properties of the QDs, we have studied: (1) dependence of the morphological characteristics on growth parameters by quantitative analysis of AFM measurements, (2) relationship between the morphological and electronic states of the ensemble through correlation between the size distributions in uncapped samples and the PL spectra in overgrown samples, and (3) the electronic states of the QD ensemble through temperature dependence of PL. 3. Results and discussion Fig. 1 shows 20 K PL spectra from different QD samples grown at fixed temperature (450oC) but varying coverage and growth rate (GR) as in (a) GR 5 ML/s, coverage 13 ML, (b) GR 2 ML/s, coverage 14 ML, and (c) GR 5 ML/s, coverage 5 ML. The emission peak has similar linewidth (100 meV) but very different peak energies— 1.87 mm for (a), 1.57 mm for (b), and 1.39 mm for (c)— showing that the emission wavelength can be tuned over an approximately 500 nm range just by changing the growth conditions as indicated. The emission efficiency for all three samples was similar. Since our interest was to understand the growth kinetics, no attempt was made to

Fig. 1. PL spectra from different QD samples with similar linewidths (100 meV) but very different peak energies changed by tuning the feature size which is strongly dependent on the growth rate (GR) and coverage. Temperature was 450oC. (a) GR 5 ML/s, coverage 13 ML, (b) GR 2 ML/s, coverage 14 ML, and (c) GR 5 ML/s, coverage 5 ML. The AFM images of 1  1 mm2 area from the corresponding uncapped samples are also shown. The peak at 1.1 eV is due to the wetting layer.

ARTICLE IN PRESS B. Bansal et al. / Journal of Crystal Growth 298 (2007) 586–590

588

minimize the linewidths. Nevertheless these values are reasonable for InAs/InP QDs grown without optimizing the capping sequences, and taking into account As/P exchange effects [7]. Fig. 2(i) shows the heights distribution obtained from 1 mm  1 mm AFM images for QD samples grown at different growth temperatures and growth rates. The growth rate in each case was assumed proportional to the TMIn flux, with V/III ratio being fixed at approximately 50 across different growth runs. The nominal GR, the growth duration (t) and the growth temperature (T) for samples is shown in the subfigures of Figs. 2(i) and (ii) are as follows: (a) GR ¼ 5 ML/s, t ¼ 2.5 s, T ¼ 450 1C, (b) GR ¼ 2.5 ML/s, t ¼ 5 s, T ¼ 450 1C, (c) GR ¼ 2.5 ML/s, t ¼ 5 s, T ¼ 430 1C, (d) GR ¼ 5 ML/s, t ¼ 3 s, T ¼ 450 1C, (e) GR ¼ 1.8 ML/s, t ¼ 8 s, T ¼ 450 1C, (f) GR ¼ 1.8 ML/s, t ¼ 8 s, T ¼ 430 1C. This figure, therefore, covers the space of three basic growth parameters—total coverage, growth rate, and temperature— and along with Fig. 1 is a demonstration of the rich variety trends that one can observe in the size distribution, and hence the emission wavelength of the QD ensemble. We also observe that the QD density can be tuned by over an order of magnitude, with further from equilibrium conditions yielding smaller and denser dots. The low-temperature PL spectra from the corresponding capped samples are shown in Fig. 2(ii). Coverage in (a)–(c) is nominally the same and these are representatives of the growth parameter dependence of the morphology and optical properties at an intermediate stage of growth. Samples in (d)–(f) also have similar coverage and these are at a late stage of growth when the QD density has saturated. From both the AFM morphology and the PL lineshape, we find that the effect of lowering the growth rate from (a) 5 ML/s to (b) 2.5 ML/s at Tg ¼ 450 1C can be qualitatively

25

10

200 0

0 25

d

20

1.0

e

120

f

0.4 0.2 0.0 0.6 1.0

15

90

20

10

60

0.4

5

30

0.2

0

0

0

0 5

10 15 20 25

0 5 10 15 20 25 Height (nm)

0.9

1.2

0.6

d

0.9

1.2

e

0.6

0.9

1.2

0.9

1.2

f

0.8

30

10

c

b

a

0.6

5

0

40

c

400

15

20

600

0.8

20

40 Density (x108 cm-2)

b

PL Intensity (normalized)

a 60

compensated by lowering the growth temperature to 430 1C at that lower growth rate (plot (c)). Similar behaviour is observed for the samples with higher coverage in subfigures (d)–(f). We have previously reported a detailed study of the influence of growth kinetics [8], and our observations are similar to those seen in recent growth simulations [9]. This indicates that under the regime of kinetically determined growth, the adatom deposition, surface diffusion and irreversible aggregation are the primary surface processes to determine the size and density of QDs, in direct analogy with very simple models of submonolayer homoepitaxy [10]. These facts suggest a relationship between not just the PL emission energies but also the inhomogeneous broadening on the growth-kinetics parameters. We often observe a bimodal distribution in the QD sizes at an intermediate stage of growth, for example, Fig. 2(b). While the origin of this bimodality was systematically studied through the complete QD evolution cycle [11], it is also found that a higher growth rate or lower temperature makes the QD distribution unimodal. A dependence on the modality of the size distribution has previously been also been observed in InP/GaInP quantum dots by Porsche et al. [12] and is theoretically not very well understood [13]. Temperature Dependence of PL: If the density of quantum dots in a sample is not too high, these dots form an inhomogeneously broadened ensemble of localized states with negligible coherent coupling (tunneling) between different dots. Yet, an incoherent coupling that involves a thermal excitation to the wetting layer and subsequent relaxation of charge carriers into a thermally preferred lower energy states is usually observed. These effects are most prominently seen as anomalies in the temperature dependent PL spectra and have direct bearing on the distribution function of the carriers and hence,

0.6

0.0 0 5 10 15 20 25

0.6

0.9

1.2 0.6

0.9

1.2

0.6

Energy (eV)

Fig. 2. (i) Heights distribution for samples grown at different growth temperatures and growth rates. Coverage in (a)–(c) is nominally the same and represents an intermediate stage of growth. Samples in (d)–(f) also have similar coverage and these are at a late stage of growth when the QD density has saturated. We find that the effect of lowering the growth rate (a)–(b) and (d)–(e) can be qualitatively compensated by lowering the growth temperature at that lower growth rate. (ii) PL spectra on the corresponding capped samples.

ARTICLE IN PRESS A

90 80

B

589

30 20 10 0

5 10 15 20 25 Height (nm)

60

C

50

20K

200

40

Large Dots Small Dots Wetting Layer

180 140K

30 200K

20 260K

10 294K

0

FWHM (meV)

PL Intensity (a.u.)

70

Count (x108 cm-2)

B. Bansal et al. / Journal of Crystal Growth 298 (2007) 586–590

160 140 120 100 80 60

0.6

0.8

0

1.0

50

Energy (eV)

100

150

200

250

300

Temperature (K)

Fig. 3. (a) The temperature dependence of the PL spectra for a sample with a bimodal distribution of dots. The histogram of QD heights is shown in the inset. Note that one obtains a nearly flat broadband emission at room temperature (linewidth 240 meV). (b) The temperature dependence of the FWHM of the different peaks in (a).

170 High Density QD Moderate density QD

FWHM (meV)

160 150 140 130 120 110 0

50

100

150

200

250

300

Temperature (K) Fig. 4. Temperature-dependent FWHMs for unimodal distributed QD ensembles with different densities and size distributions. (i) (open triangles) QD density 1011 cm-2 where PL linewidth first decreases with increase in temperature and is then, limited by intrinsic broadening effects. (ii) (solid circles) QD density 1012 cm2, where carrier transfer right up to room temperature results in linewidth decreasing with increase in temperature.

device performance. The phenomenon of carrier transfer from smaller (higher energy) dots to larger (lower energy) dots is most clearly seen in the non-monotonic behavior of the full-widths at half-maxima (FWHM). Fig. 3(a) shows PL spectra from a sample with a bimodal QD size distribution as a function of temperature. The heights histogram of a corresponding uncapped sample is shown in inset. At low temperature, three distinct peaks

from the (A) large dots, (B) small dots, and (C) wetting layer are seen. With increase in temperature, the ratio of peak intensities changes, with the low-energy peaks dropping faster due to a larger activation barrier for the larger dots. The dependence of the FWHM of the three peaks as a function of temperature is shown in Fig. 3(b). The opposite trends followed by the temperature dependence of FWHM corresponding to the peaks due to large and small dots indicate thermally activated carrier transfer. We further observe that at room temperature, the PL spectrum is broad (FWHM 250 meV) and flat. Such QD ensembles may be useful for fabrication of broadband light sources and detectors. For the case of unimodal dot distributions, we observe the characteristic anomalous minimum in the linewidth with increasing temperature plotted for example in Fig. 4 (triangles) for an ensemble with a QD density 1011 cm2. Here, the PL linewidth first decreases with increase in temperature. Beyond about 200 K, the FWHM starts to increase with temperature indicating that the thermal smearing of the distribution function dominates over the localization effects. The distribution function may be assumed to have partially thermalized above 200 K. However, for very high-density samples (1012 dots/cm2) we see a less frequently observed monotonic decrease of the linewidth right up to room temperature (Fig. 4, circles) suggesting an activated carrier transfer to larger dots right up to room temperature. 4. Conclusions We have grown InAs/InP QDs by MOVPE in a highly non-equilibrium regime, of relatively low temperature (p450 1C) and higher growth rates (X1.5 ML/s), which

ARTICLE IN PRESS 590

B. Bansal et al. / Journal of Crystal Growth 298 (2007) 586–590

permits the peak-emission wavelength, QD sizes, density and modality of the size distribution to be tuned using kinetic parameters. The low-temperature peak emission wavelength of the ensemble can be tuned over almost a 500 nm range from 1.4 to 1.9 mm. Similarly, broadband emission with 250 meV bandwidth is obtained from samples with bimodal-dot distributions. Our observations provide useful insights into light emission from InAs/InP QD ensembles. References [1] N. Carlsson, T. Junno, L. Montelius, M.-E. Pistol, L. Samuelson, W. Seifert, J. Crystal Growth 191 (1998) 347. [2] K. Kawaguchi, M. Ekawa, A. Kuramata, T. Akiyama, H. Ebe, M. Sugawara, Y. Arakawa, Appl. Phys. Lett. 85 (2004) 4331. [3] A. Michon, G. Saint-Girons, G. Beaudoin, I. Sagnes, L. Largeau, G. Patriarche, Appl. Phys. Lett. 87 (2005) 253114.

[4] D. Bimberg, M. Grundmann, N.N. Ledentsov, Quantum Dot Heterostructures, Wiley, Chichester, 1999. [5] C. Preister, M. Lannoo, Phys. Rev. Lett. 75 (1995) 93; Curr. Opinion Sol. Stat. Mat. Sci. 2 (1997) 716. [6] Q. Gong, R. No¨tzel, P.J. van Veldhoven, T.J. Eijkemans, J.H. Wolter, Appl. Phys. Lett. 85 (2004) 1404. [7] D. Fuster, M.U. Gonza´lez, L. Gonza´lez, Y. Gonza´lez, T. Ben, A. Ponce, S.I. Molina, J. Martı´ nez-Pastor, Appl. Phys. Lett. 85 (2004) 1424. [8] B. Bansal, M.R. Gokhale, A. Bhattacharya, B.M. Arora, Appl. Phys. Lett. 87 (2005) 203104. [9] M. Meixner, R. Kunert, E. Scho¨ll, Phys. Rev. B 67 (2003) 195301. [10] A.-L. Barabasi, H.E. Stanley, Fractal Concepts in Surface Growth, Cambridge University Press, Cambridge, 1995. [11] B. Bansal, M.R. Gokhale, A. Bhattacharya, B.M. Arora, arXiv:condmat/0505305. [12] J. Porsche, A. Ruf, M. Geiger, F. Scholz, J. Crystal Growth 195 (1998) 591. [13] R.E. Rudd, G.A.D. Briggs, A.P. Sutton, G. Medeiros-Ribeiro, R.S. Williams, Phys. Rev. Lett. 90 (2003) 146101.