GaN heterostructures

GaN heterostructures

Superlattices and Microstructures 40 (2006) 507–512 www.elsevier.com/locate/superlattices Vertical electron transport study in GaN/AlN/GaN heterostru...

481KB Sizes 1 Downloads 107 Views

Superlattices and Microstructures 40 (2006) 507–512 www.elsevier.com/locate/superlattices

Vertical electron transport study in GaN/AlN/GaN heterostructures S. Leconte ∗ , E. Monroy, J.-M. G´erard Equipe mixte CEA-CNRS-UJF Nanophysique et Semiconducteurs, DRFMC/SP2M/PSC, CEA-Grenoble, 17 rue des Martyrs, 38054 Grenoble cedex 9, France Received 25 September 2006; accepted 6 October 2006 Available online 22 November 2006

Abstract In this work, we investigate the electronic structure and vertical electron transport through GaN/AlN/GaN single-barrier structures with different AlN thickness, grown by plasma-assisted molecular beam epitaxy. Conductive and capacitive characterization has been performed, and the experimental results are interpreted by comparison with 1D self-consistent simulations. Capacitive measurements reveal a complete depletion of the top GaN layer, and the formation of a two-dimensional electron gas at the bottom interface of the AlN barrier, even for barrier thicknesses of 0.5 nm (2 monolayers of AlN). Conductive atomic force microscopy reveals discrete leakage current locations with a density of ∼107 cm−2 , more than one order of magnitude lower than the dislocation density in these samples. These results are promising for the fabrication of resonant tunnelling diodes using the GaN/AlN material system. c 2006 Elsevier Ltd. All rights reserved.

Keywords: GaN; Resonant tunneling diode; Two dimensional electron gas; Conductive atomic force microscopy

1. Introduction III-nitride materials, with their large conduction-band offet – about 2 eV for the GaN/AlN system –, are promising candidates to develop intersubband (ISB) devices operating in the nearinfrared, and resonant tunneling diodes with high values of peak-to-valley ratio. However, we are still far from understanding unipolar vertical electric transport in nitride heterostructures. First reports of resonant tunnelling in Al(Ga)N/GaN double barriers [1–5] are controversial because ∗ Corresponding author.

E-mail address: [email protected] (S. Leconte). c 2006 Elsevier Ltd. All rights reserved. 0749-6036/$ - see front matter doi:10.1016/j.spmi.2006.10.008

508

S. Leconte et al. / Superlattices and Microstructures 40 (2006) 507–512

of the scarcity and irreproducibility of the published data. Current instabilities are admitted [2, 4,5], and their assignment to resonant tunnelling or trapping is under debate. Furthermore, the current densities and peak-to-valley ratios do not correspond to the values predicted by standard tunnelling models. Understanding the vertical transport and the electronic properties of AlN/GaN tunneling barriers is hence a priority in order to improve the design of multi-barrier quantum devices, such as resonant tunnelling diodes, quantum well infrared photodetectors or quantum cascade lasers. In this work, we investigate the charge distribution and vertical electron transport through GaN/AlN/GaN structures with different AlN barrier and GaN capping thicknesses. Capacitive measurements of samples with barrier thickness between 0.5 and 3 nm reveal the presence of a depletion region in the top GaN layer, and the formation of a two-dimensional electron gas (2DEG) at the bottom interface of the barrier, even for barrier thicknesses as small as 0.5 nm (2 monolayers of AlN). In addition, conductive atomic force microscopy has been used to identify the leakage current distribution. 2. Experimental Samples consisting of an AlN thin layer embedded in non-intentionally-doped (n.i.d.) GaN were grown by plasma-assisted molecular-beam epitaxy (PAMBE) in a MECA2000 chamber equipped with standard effusion cells for Al and Ga. Active nitrogen was supplied by a radiofrequency plasma cell. Substrates consisted of 10 µm-thick n.i.d. GaN-on-sapphire templates. A 600 nm-thick n.i.d. GaN buffer layer was deposited prior to the growth of the AlN barrier and the 50 nm-thick n.i.d. GaN cap layer. The barrier thickness varied between 0.5 and 3 nm. The substrate temperature was TS = 720 ◦ C, and growth proceeded under Ga excess during the deposition of both GaN and AlN, i.e. the Ga shutter remained permanently opened during growth, and AlN was obtained by additional opening of the Al cell. No growth interruptions were performed at the interfaces. We have previously reported that under these growth conditions we obtain abrupt GaN/AlN interfaces at the atomic-layer scale [6], due to the preferential incorporation of Al compared to Ga [6,7]. The electronic structure of the GaN/AlN/GaN single barrier structures was calculated from 1D self-consistent simulations using WinGreen and Nextnano3 software [8,9]. We assumed that the AlN barriers are strained on the GaN layer, and the GaN residual doping was 5 × 1017 cm−3 , as deduced from our capacitance measurements. Low-temperature (10 K) photoluminescence measurements were performed by exciting the samples with a power of 1 mW from a frequencydoubled Ar laser (244 nm). Capacitive characterization was carried out using an HP4274A LCR meter and a Keithley 4200 semiconductor characterization system. Conductive atomic force microscopy (C-AFM) experiments were performed in a Dimension 3100 microscope, and a Keithley 617 programable electrometer was used to detect the localized current between a Pt/Ir AFM tip and the sample when a dc bias was applied. 3. Results and discussion Fig. 1 illustrates the simulation of the band diagram of the 2 nm-thick AlN barrier. The difference in piezoelectric and spontaneous polarization at the GaN/AlN interfaces, induces an electric field of about 12 MV/cm in the barrier, strong band bending in the GaN cap layer and the formation of a two-dimensional electron gas (2DEG) at the AlN/GaN-buffer interface. The energy gap through the barrier (E GB in Fig. 1(a)), i.e. the energy difference between the bottom

S. Leconte et al. / Superlattices and Microstructures 40 (2006) 507–512

509

Fig. 1. (a) Band diagram of a GaN/AlN/GaN single-barrier structure with an AlN barrier thickness of 2 nm. The arrows indicate the energy bandgap through the barrier, E GB , i.e. the energy difference between the bottom of the GaN conduction band at the AlN/GaN buffer interface and the top of the GaN valence band at GaN cap/AlN interface. (b) Evolution of E GB as a function of the barrier thickness obtained by simulations at 7 and 300 K.

of the GaN conduction band at the AlN/GaN buffer interface and the top of the GaN valence band at GaN cap/AlN interface decreases linearly with increasing barrier thickness from 1 ML (0.25 nm) to 12 ML (3 nm) as illustrated in Fig. 1(b). For larger barriers (>3 nm), E GB becomes negative, allowing charge transfer from the conduction band to the valence band by tunnelling through the barrier. Moreover, the top of the valence band at the upper interface of the barrier lies above the Fermi level, resulting in the formation of a two-dimensional hole gas (2DHG) which pins the Fermi level at this interface and limits the potential drop in the AlN barrier (saturation of the curve in Fig. 1(b) for barrier thickness >3 nm). The photoluminescence spectra displayed in Fig. 2 present the typical features of GaN photoluminescence, with intense near-band-edge emission, and a broad yellow luminescence (YL, maximum at about 2.2 eV). We observe additional PL features at 3.10 eV and 2.54 eV for the 0.5 nm and 1 nm barriers respectively. These PL lines are attributed to recombination

510

S. Leconte et al. / Superlattices and Microstructures 40 (2006) 507–512

Fig. 2. Low temperature (10 K) photoluminescence spectra of GaN/AlN/GaN single barrier structures with different barrier thickness. The spectra are vertically shifted for clarity.

Fig. 3. 1/C 2 (V ) curves measured from 400 µm mesas, in samples with an AlN barrier thickness from 0.5 to 3 nm.

from the 2DEG at the AlN/GaN buffer interface to the top of the GaN valence band at the GaN cap/AlN interface, in good agreement with the values predicted by our simulations (see Fig. 1(b)). For thicker barriers, radiative recombination through the barrier is not observed since the tunnel probability decreases exponentially. In order to analyze the charge distribution in these structures, Ni/Au (30/100 nm) Schottky contacts were e-beam evaporated on top of mesas with a diameter of 200, 400 and 1000 µm, patterned with Cl2 -based reactive ion etching. An extended Ti/Al/Ti/Au (30/70/10/100 nm) ohmic contact was deposited around the mesas. Results of capacitive measurements are given in Fig. 3. The capacitance at zero-bias is approximately the same in all the samples, and corresponds to the capacitance of a 50 nm wide depleted GaN region (to be compared with the 130 nm wide

S. Leconte et al. / Superlattices and Microstructures 40 (2006) 507–512

511

Fig. 4. (a) Contact-mode AFM topographic image of the GaN(50 nm)/AlN(3 nm)/GaN structure. (b) Simultaneously taken C-AFM image under 4 V forward bias. White areas correspond to higher current levels.

depletion region in a GaN reference sample), which reveals that the GaN cap layer is totally depleted, as predicted by our simulations. At low bias the capacitance increases very slowly due to the presence of the 2DEG at the bottom interface of the barrier. The density of charge in the 2DEG was calculated from these data, and it increases with the barrier thickness from 2 × 1012 to 5 × 1012 cm−2 . The transport properties of the barriers in micrometer mesas are masked by current leakage induced by the high density of dislocations. Spatially resolved techniques are required to identify the hot current spots and separate intrinsic from defect induced properties. In this work, conductive atomic force microscopy (C-AFM) has been used to investigate the local conductivity of the barriers. The AFM topography and C-AFM images of the GaN(50 nm)/AlN(3 nm)/GaN structure are shown in Fig. 4(a)–(b). The surface morphology of all the samples satisfy the expectations for GaN layers grown under Ga-rich conditions: monolayer steps between hexagonal hillocks (see Fig. 4(a)). To study the current distribution, we scan the top of the mesas with Pt/Ir-metallized AFM tips. The tip-to-sample contact behaves as a microscopic Schottky contact, whereas the extended ohmic contact around the mesas is grounded. Fig. 4(b) displays a typical C-AFM image obtained when the tip is forward-biased. Current flow is concentrated within small areas which appear as white spots in the image, while the current density in the rest of the sample is below the noise level (∼100 pA) of our setup. The density of leakage current spots (∼107 cm−2 ) is approximately one third of the density of hexagonal hillocks, and much lower than the dislocation density of these samples (∼5 × 108 cm−2 ). This result agrees with previous C-AFM reports on thick GaN layers [10,11] which attribute leakage paths to pure screw dislocations. 4. Conclusion We have used photoluminescence and capacitive measurements to analyse the electronic structure and charge distribution in GaN/AlN/GaN single-barrier structures. We have verified a complete depletion of the top GaN layer, and the formation of a two-dimensional electron gas at the bottom interface of the AlN barrier, in good agreement with our simulation of the electronic structure, even for a barrier thickness as small as 0.5 nm (2 monolayers of AlN). Conductive atomic force microscopy images of these samples display a density of leakage current spots in

512

S. Leconte et al. / Superlattices and Microstructures 40 (2006) 507–512

the 107 cm−2 range, which is significantly lower than the dislocation density (∼5 × 108 cm−2 ). These results are promising for the fabrication of micrometer-size resonant tunnelling diodes using the GaN/AlN material system. Acknowledgement This work is partially supported by EC FP6 STREP project “NITWAVE” under IST contract #004170. References [1] A. Kikuchi, R. Banai, K. Kishino, C.-M. Lee, J.-I. Chyi, Appl. Phys. Lett. 81 (2002) 1729. [2] C.T. Foxon, S.V. Novikov, A.E. Belyaev, L.X. Zhao, O. Makarovsky, D.J. Walker, L. Eaves, R.I. Dykeman, S.V. Danylyuk, S.A. Vitusevich, M.J. Kappers, J.S. Barnard, C.J. Humphreys, Phys. Status Solidi (c) 0 (2003) 2389. [3] A.E. Belyaev, C.T. Foxon, S.V. Novikov, O. Makarovsky, L. Eaves, M.J. Kappers, C.J. Humphreys, Appl. Phys. Lett. 83 (2003) 3626; A. Kikuchi, R. Banai, K. Kishino, C.-M. Lee, J.-I. Chyi, Appl. Phys. Lett. 83 (2003) 3628. [4] M. Hermann, E. Monroy, A. Helman, B. Baur, M. Albrecht, B. Daudin, O. Ambacher, M. Stutzmann, M. Eickhoff, Phys. Status Solidi (c) 1 (2004) 2210. [5] S. Golka, C. Pfl¨ugl, W. Schrenk, G. Strasser, C. Skierbiszewski, M. Siekacz, I. Grzegori, P. Porowski, Appl. Phys. Lett. 88 (2006) 172106. [6] E. Sarigiannidou, E. Monroy, N. Gogneau, G. Radtke, P. Bayle-Guillemaud, E. Bellet-Amalric, B. Daudin, J.L. Rouvi`ere, Semicond. Sci. Technol. 21 (2006) 912. [7] E. Iliopoulos, T.D. Moustakas, Appl. Phys. Lett. 81 (2002) 295. [8] WinGreen. Available on-line: http://www.fz-juelich.de/isg/mbe/software.html. [9] Nextnano3 . Available on-line: http://www.nextnano.de. [10] J.W.P. Hsu, M.J. Manfra, R.J. Molnar, B. Heying, J.S. Speck, Appl. Phys. Lett. 81 (2002) 79. [11] B. Simpkins, E.T. Yu, P. Waltereit, J.S. Speck, J. Appl. Phys. 94 (2003) 1448.