Journal of the Korean Physical Society, Vol. 56, No. 6, June 2010, pp. 1828∼1832

Height-controlled InGaN Quantum Dots and Light-emitting Diode Applications Il-Kyu Park LED-IT Fusion Technology Research Center and Department of Electronic Engineering, Yeungnam University, Gyeongbuk 712-749

Seong-Ju Park∗ Department of Materials Science and Engineering, Gwangju Institute of Science and Technology, Gwangju 500-712 (Received 16 March 2010, in final form 27 April 2010) We report on the growth of InGaN height-controlled quantum dots (HCQDs) and the fabrication of light-emitting diodes (LEDs). InGaN HCQDs were grown by alternately depositing In0.43 Ga0.57 N QDs and In0.2 Ga0.8 N spacer layers on a seed In0.43 Ga0.57 N QD layer by using a metal-organic chemical vapor deposition system. The photoluminescence (PL) and the electroluminescence (EL) emission peaks of the InGaN HCQDs were red-shifted with increasing number of depositions. This indicates that the height of the InGaN HCQDs can be controlled by the number of deposition cycles of the In0.43 Ga0.57 N/In0.2 Ga0.8 N layers because the thin In0.2 Ga0.8 N spacer layer allowed electrical coupling between the vertically-stacked QDs. As the input current increases, the EL emission peak of the InGaN HCQD LED was blue-shifted, and the width of the EL peak increased, indicating a negligible piezoelectric-field-induced quantum-confined Stark effect in the InGaN HCQDs. PACS numbers: 68.55.ag, 73.21.La, 71.55.Eq, 78.55.Cr Keywords: InGaN, Quantum dots, Light-emitting diodes, MOCVD DOI: 10.3938/jkps.56.1828

I. INTRODUCTION

InGaN-based quantum dots (QDs) have attracted much attention for their potential nanophotonic applications, such as single-photon devices and light-emitting sources [1–4]. InGaN QDs are usually grown via a straininduced Stranski-Krastanov (S-K) growth mode. However, it has been known to be very difficult to control the size or shape of the QDs when using this approach because the strain relaxation-induced QD formation depends on the lattice-mismatched heteroepitaxy system [5,6]. Furthermore, the height of the InGaN QDs is usually much smaller than the diameter, resulting in a small aspect ratio well below 1 [5,6]. Williamson and Zunger showed that the electron wave function in QDs with a small aspect ratio can penetrate significantly into the quantum barrier, resulting in a reduced confinement of the electron wave function in the vertical direction and in an overlap integral between the electron and the hole wave functions [7]. This results in a reduced recombination efficiency of the electron and the hole in QDs with a lower aspect ratio. To circumvent the problems originat∗ E-mail:

ing from the difficulty in controlling the shape of QDs, a cycled deposition of a short-period superlattice on top of seed QD layers was proposed for InAs or InGaAs QDs [8– 10]. This method provides controllability of the height of the QDs by changing the number of deposition cycles and opens the way for artificial control of the electronic energy states along the vertical direction and the oscillator strength in the QDs. Even though the growth and the characteristics have been investigated much for InAs and InGaAs QDs deposited by using this method [8–10], little has been reported for III-nitride [11], which is an important material system for solid-state lighting devices and short-wavelength single-photon sources for quantum information applications [1–4]. In this paper, we report on the growth of InGaN height-controlled QDs (HCQDs) by alternating the deposition of In-rich In0.43 Ga0.57 N QDs and In-poor In0.2 Ga0.8 N spacer layers on a seed In0.43 Ga0.57 N QD layer grown via the S-K growth mode. Furthermore, by using the InGaN HCQDs to fabricate light-emitting diodes (LEDs), we were able to investigate the recombination process and the emission properties of InGaN HCQD LEDs.

[email protected]; Fax: +82-62-970-2309

-1828-

Height-controlled InGaN Quantum Dots and Light-emitting Diode Applications – Il-Kyu Park and Seong-Ju Park

-1829-

Fig. 1. (Color online) Schematic structure of a heightcontrolled InGaN quantum dot LED.

II. EXPERIMENTS Self-assembled InGaN HCQDs were grown on GaN surfaces on (0001) sapphire substrates at a pressure of 200 Torr in a metal-organic chemical vapor deposition (MOCVD) system. After the growth of a 2-µm-thick n-GaN layer at 1050 ◦ C In0.43 Ga0.57 N QDs, which act as seeds for the HCQDs, were grown at 650 ◦ C for 6 s via a strain-induced S-K growth mode using N2 as a carrier gas. To grow the HCQD structures, we grew 5, 10, and 15 periods of In-rich In0.43 Ga0.57 N as QDs and In-poor In0.2 Ga0.8 N layers as spacer layers on the seed In0.43 Ga0.57 N QD layer at the same growth temperature. From the growth rate of the thick In0.2 Ga0.8 N layer, the thickness of the In0.2 Ga0.8 N spacer layer was estimated to be less than 5 nm. The In compositions of the InGaN QD and spacer layers were 43% and 20%, respectively, from an X-ray diffraction analysis for a 20-nm-thick InGaN layer grown under the same condition. For photoluminescence (PL) measurements, after growing the HCQD layers, a 50-nm-thick GaN layer was grown at 750 ◦ C to avoid surface recombination of electrical carriers. The detailed procedure and a description of the mechanism of the InGaN HCQDs growth will be published elsewhere [11]. After growth of the HCQD layers with multiple deposition cycles, a 0.13-µm-thick pGaN layer was grown on the HCQD layer at 900 ◦ C The growth temperature for the p-GaN layer was 100 ◦ C lower than the conventional p-GaN growth temperature of 1000 ◦ C to protect the HCQD layers from thermal damage. To fabricate LEDs with a size of 300 × 300 µm2 , a p-GaN layer was etched by means of an inductivelycoupled Cl2 /CH4 /H2 /Ar plasma until the n-GaN layer was exposed, thereby achieving an n-type Ohmic contact. A detailed procedure for the fabrication of LEDs with a size of 300 × 300 µm2 was published elsewhere [2]. A schematic structure of the LED with InGaN HCQDs

Fig. 2. (Color online) (a) Room-temperature PL spectra of a single layer and of five multiple layers composed of InGaN QDs and 20-nm-thick GaN quantum barriers. (b) Room-temperature PL spectra of InGaN HCQDs with 2 and 5 deposition cycles.

is shown in Fig. 1. The electroluminescence (EL) of the LED was measured using a photodiode system equipped with an optical fiber for light collection.

III. RESULTS & DISCUSSION To examine the controllability of the height of InGaN QDs by depositing multiple InGaN QDs and thin InGaN spacer layers, we measured the PL spectra for two types of InGaN QDs with the same composition. One had InGaN QDs with a 20-nm-thick GaN barrier, and the other type had thin In0.2 Ga0.8 N spacer layers. In the case of the InGaN QDs with a 20-nm-thick GaN barrier layer, the PL spectra of the InGaN QDs shifted to the shorter wavelength side from 397 to 390 nm with increasing number of deposition cycles from 1 to 5, as shown in Fig. 2(a). This is assumed to be due to an increase in the bandgap energy of the InGaN QDs caused by the accumulation of compressive stress in the layers with increas-

-1830-

Fig. 3. (Color online) Normalized PL spectra of InGaN HCQDs with various numbers of deposition cycles measured at low temperature (10 K).

ing number of deposition cycles [12]. However, the PL spectra of the InGaN QDs with a thin In0.2 Ga0.8 N spacer layer showed a red-shift with increasing number of deposition cycles, which can be attributed to a piezoelectricfield-induced quantum-confined Stark effect (QCSE) or a quantum confinement effect. This will be discussed in detail later. The tunneling probability is well known to be inversely proportional to the potential barrier width and height, indicating that the tunneling probability of the electrical carriers through the 20-nm-thick GaN barrier layer is very low compared with that through the 5nm-thick In0.2 Ga0.8 N layer. Therefore, the electron and the hole wavefunctions between the InGaN QDs in the subsequent layer cannot be coupled to each other. On the other hand, In0.2 Ga0.8 N spacer layers less than 5 nm in the thickness are thin enough to allow an overlap of the individual wave functions between the subsequently deposited QD layers, finally resulting in an electrical coupling. The electrically-coupled QDs results in a merged energy structure, and the ground-state quantum confinement energy of the HCQDs may be tuned by using the number of deposition cycles of the QD and spacer layers to control the height. The structural properties investigated by using transmission electron microscopy (TEM) and atomic force microscopy (AFM) showed that the heights of the InGaN HCQDs were strongly correlated with the number of deposition cycles [11]. Figure 3 shows the normalized PL spectra of the In0.43 Ga0.57 N HCQDs with 5, 10, and 15 periods measured at 10 K. As the number of deposition cycles increased from 5 to 15, the PL peak from the HCQDs shifted from 456 to 509 nm. This result is due to an increase in the vertical height of the HCQDs with increasing number of deposition cycles, finally resulting in a decrease in the quantum confinement energy in the HCQDs. As previously explained, by repeating the

Journal of the Korean Physical Society, Vol. 56, No. 6, June 2010

Fig. 4. (Color online) Normalized EL spectra of InGaN HCQD LEDs with InGaN HCQDs of different numbers of deposition cycle at an input current of 20 mA.

growth of an In0.43 Ga0.57 N QD/In0.2 Ga0.8 N space layer, the In0.43 Ga0.57 N QDs were vertically coupled, which resulted in a merged energy structure. This indicates that the height of a HCQD can be systematically controlled by using the number of deposition cycles, and the emission wavelength can be tuned from the blue to the green spectral range. It should be noted that the InGaN HCQD with 5 periods showed two PL peaks, one around 420 nm and the other around 456 nm. However, the intensity of the PL peak around 420 nm decreased significantly at room temperature, as shown in Fig. 2(b). This peak is assumed to be due to localized excitons in the shallow potential wells originating from composition or thickness fluctuations in the InGaN spacer layers. With increasing temperature, carriers redistributed from shallow to deep potential wells, finally resulting in a single peak on the lower energy side, which was provided by InGaN HCQDs. The peak around 380 nm and its neighboring lines are attributed to donor-to-acceptor pair (DAP) transitions in GaN and their longitudinal optical (LO) phonon replicas, respectively [13]. The spacing between side peaks was 94 meV, which was very close to the GaN LO phonon energy of 91.5 meV [14]. Figure 4 shows the normalized EL spectra of the HCQD LEDs with three different periods at an input current of 20 mA, measured at room temperature. The EL emission peak was red-shifted from blue (455 nm) to green (508 nm) with increasing number of deposition cycles from 5 to 15. The EL results showed the same behavor as the PL results, which indicates that the InGaN HCQDs were not damaged thermally during the growth of the p-GaN layer at 900 ◦ C Figure 5(a) shows the EL spectra for 15 periods of an InGaN HCQD LED with increasing input current from 20 to 100 mA. A strong bluish-green EL emission peak from the InGaN HCQD LED can be seen at around 508

Height-controlled InGaN Quantum Dots and Light-emitting Diode Applications – Il-Kyu Park and Seong-Ju Park

-1831-

Therefore, the EL results for the LED show that the piezoelectric-field-induced QCSE in the InGaN HCQDs is negligible. The isotropic confinement and polarization in the QDs are known to weaken the effect of the piezoelectric field compared to the case of quantum wells [16,17]. The TEM investigation showed that the InGaN HCQDs were embedded in the In-poor InGaN spacer layer in three dimensions [11]. This induces isotropic confinement and polarization in the InGaN HCQDs and finally results in a negligible QCSE in the InGaN HCQDs. If the QCSE dominates the recombination process, the recombination efficiency of the electron and the hole would decrease with increasing number of deposition cycles because the polarization-induced electric field would enhance the spatial separation between the electron and the hole. However, as shown in Fig. 2(b), the PL intensity of the InGaN HCQDs did not decrease with increasing number of deposition cycles, indicating a negligible QCSE in the InGaN HCQDs. Therefore, the negligible piezoelectric-field-induced QCSE in InGaN HCQDs is beneficial for improving the luminescence efficiency of InGaN QD-based LEDs. From these results, InGaN HCQDs can be expected to serve as promising light-emitting sources for highly-efficient LEDs.

IV. CONCLUSION Fig. 5. (Color online) (a) EL spectra of the LED with InGaN HCQD with 15 deposition cycles (b) EL peak position and the FWHM of the EL peak for increasing input current for InGaN HCQDs with 15 deposition cycles. The solid lines serve as a guide.

nm. The EL intensity increased with increasing input current without showing saturation, as shown in Fig. 5(a). This result can be attributed to carrier recombination in the deep potential wells provided by the InGaN HCQDs. To better understand the EL emission properties of the InGaN HCQD LEDs, we plotted the change in EL peak energy and the full width at half maximum (FWHM) of the peak versus the input current, as shown in Fig. 5(b). The EL peak showed a blue-shift with increasing input current. It is known that there are two possible origins of the blue-shift of the EL peak with increasing input current in InGaN-based LEDs: 1) a screening effect of the QCSE due to increased free carriers, and 2) a band-filling effect of the localized energy states. Both can contribute to the recombination process of the electrical carriers in InGaN HCQDs. The FWHM increased with increasing input current, as shown in Fig. 5(b). If the QCSE dominated the recombination process of carriers in the QDs, the FWHM of the EL peak would decrease with increasing input current [3,6]. The increase in the FWHM of the EL peak is due to a band-filling effect in the InGaN HCQDs [3,6,15].

InGaN HCQDs were grown by alternately depositing In0.43 Ga0.57 N QDs and In0.2 Ga0.8 N spacer layers on a seed In0.43 Ga0.57 N QD layer by using a MOCVD system. Optical investigation using PL spectra showed that the emission peaks from the HCQDs shifted to the lower energy side from 465 to 509 nm with increasing number of deposition cycles from 5 to 15. The electrically-coupled QDs result in a merged energy structure. As a result, the ground-state quantum confinement energy of the HCQDs can be tuned by controlling the height with the deposition cycle. The EL spectra of LEDs fabricated with InGaN HCQDs showed that the emission peak was shifted to the long wavelength side from 455 to 508 nm with increasing number of deposition cycles from 5 to 15, as was the case in the PL results. The EL investigation of the LEDs with InGaN HCQDs showed that the piezoelectricfield-induced QCSE was negligible in the recombination process of electrical carriers, which was beneficial for improving the efficiency of InGaN QD-based LEDs. These results show that HCQDs may be a promising approach to creating long-wavelength or multi-colored InGaN QD light-emitting sources.

ACKNOWLEDGMENTS This work was partially supported by the World Class University (WCU) program through a grant provided

-1832-

Journal of the Korean Physical Society, Vol. 56, No. 6, June 2010

by the Ministry of Education, Science and Technology (MEST) of Korea (Project No. R31-2008-000-10026-0), the Ministry of Land, Transport, and Maritime Affairs (Grant No. 20090006), U.S. Air Force Office of Scientific Research (AFOSR)/Asian Office of Aerospace Research and Development (AOARD), Industrial Strategic Technology Development Program funded by the Ministry of Knowledge Economy (MKE) (Grant No. 10033630) and by Yeungnam University research grants in 2009.

REFERENCES [1] S. Kako, C. Santori, K. Hoshino, S. G¨ otzinger, Y. Yamamoto and Y. Arakawa, Nat. Mater. 5, 887 (2006). [2] L. S. Dang, G. Fishman, H. Mariette, C. Adelmann, E. Martinez, J. Simon, B. Daudin, E. Monroy, N. Pelekanos, J. L. Rouviere and Y. H. Cho, J. Korean Phys. Soc. 42, S657 (2003). [3] I. K. Park, M. K. Kwon, S. B. Seo, J. Y. Kim, J. H. Lim and S. J. Park, Appl. Phys. Lett. 90, 111116 (2007). [4] L. W. Ji, Y. K. Su, S. J. Chang, C. S. Chang, L. W. Wu, W. C. Lai, X. L. Du and H. Chen, J. Cryst. Growth 263, 114 (2004). [5] T. Xu, A. Y. Nikiforov, R. France, C. Thomidis, A. Williams and T. D. Moustakas, Phys. Status Solidi A 204, 2098 (2007).

[6] I. K. Park, M. K. Kwon, C. Y. Cho, J. Y. Kim, C. H. Cho and S. J. Park, Appl. Phys. Lett. 92, 253105 (2008). [7] A. J. Williamson and A. Zunger, Phys. Rev. B 59, 15819 (1999). [8] G. S. Solomon, J. A. Trezza, A. F. Marshall and J. S. Harris, Jr., Phys. Rev. Lett. 76, 952 (1996). [9] J. He, H. J. Krenner, C. Pryor, J. P. Zhang, Y. Wu, D. G. Allen, C. M. Morris, M. S. Sherwin and P. M. Petroff, Nano Lett. 7, 802 (2007). [10] J. S. Kim, P. W. Yu, J. I. Lee, J. S. Kim, S. G. Kim, J.Y. Leem and M. Jeon, Appl. Phys. Lett. 80, 4714 (2002). [11] I. K. Park, C. J. Choi and S. J. Park J. Cryst. Growth (to be published) [12] N. Khan and J. Li, Appl. Phys. Lett. 89, 151916 (2006). [13] M. A. Reshchikova and H. Morko¸c, J. Appl. Phys. 97, 061301 (2005). [14] J. A. Davidson, P. Dawson, Tao Wang, T. Sugahara, J. W. Orton and S. Sakai, Semicond. Sci. Technol. 15, 497 (2000). [15] Y. Narukawa, Y. Kawakami, M. Funato, Sz Fujita, S. Fujita and S. Nakamura, Appl. Phys. Lett. 70, 981 (1997). [16] Y. Narukawa, Y. Kawakami, S. Fujita and S. Nakamura, Phys. Rev. B 59, 10283 (1997). [17] I. K. Park, M. K. Kwon, J. O. Kim, S. B. Seo, J. Y. Kim, J. H. Lim, S. J. Park and Y. S. Kim, Appl. Phys. Lett. 91, 133105 (2007).