- Email: [email protected]
Growth of high-quality N-polar GaN on bulk GaN by plasma-assisted molecular beam epitaxy
Solid State Communications 305 (2020) 113763
Contents lists available at ScienceDirect
Solid State Communications journal homepage: http://www.elsevier.com/locate/ssc
Growth of high-quality N-polar GaN on bulk GaN by plasma-assisted molecular beam epitaxy Christian Wurm a, *, Elaheh Ahmadi b, Feng Wu c, Nirupam Hatui a, Stacia Keller c, James Speck c, Umesh Mishra a a b c
Department of Electrical and Computer Engineering, University of California, Santa Barbara, CA, 93117, USA Department of Electrical Engineering and Computer Science, University of Michigan, MI, 48109, USA Materials Department, University of California, Santa Barbara, CA, 93117, USA
A R T I C L E I N F O
A B S T R A C T
Communicated by Bell Gavin
’There is an interest in growing N-polar GaN on bulk GaN for high electron mobility transistors (HEMTs). Current N-polar HEMT technology is dominated by devices grown on non-native substrates which possess a high density of dislocations due to lattice mismatch. N-polar GaN films grown directly on non-miscut GaN substrates have displayed a high density of pits and depressions on the surface. This work demonstrates that surface impurities play a major role in the formation of surface-pits in epitaxially grown film. By subjecting the GaN substrate to an ultra-violet (UV) O3 (ozone) clean prior to growth and initiating growth with a 2 nm coherently strained AlN layer, grown under metal-rich conditions, we have demonstrated N-polar GaN films with a nearly pit-free sur face. This AlN initiation layer (AIL) likely captures impurities on the substrate surface thus decoupling the substrate surface from the epitaxially grown film. It is demonstrated that utilizing thicker AILs, up to 8 nm, further improves film quality and surface morphology. The methods employed in this study to produce highquality N-polar GaN grown on bulk GaN will pave the way for future GaN devices with an order of magni tude or more lower threading dislocation density (TDD).
Keywords: A. Semiconductors B. Molecular beam epitaxy C. Nitrides D. HEMT
1. Introduction The need for high-quality ð0001Þ GaN, or N-polar GaN, films grown on bulk GaN comes from its applications in high power, high frequency electronics. N-polar high electron mobility transistors, or HEMTs, have shown superior properties with respect to their Ga-polar counterparts in terms of operating frequency, power output and scalability [1,2]. High quality N-polar GaN grown by metal-organic chemical vapor deposition (MOCVD) on vicinal sapphire and SiC has already been used extensively to make high performance HEMTs [3]. However, growing GaN hetero epitaxially on SiC, for example, where an in-plane 3.4% lattice mismatch [4] is present, results in vertically propagating threading dislocations (TDs). A threading dislocation density (TDD) on the order of 108 cm 2 has been reported for GaN grown on sapphire and SiC [5]. In high concentrations, threading dislocations can act as traps impeding mobility and lowering device performance [6]. Dislocations in GaN are also a major cause of vertical leakage when grown by metal-rich
plasma-assisted MBE (PAMBE); metal-rich growth results in metal-decorated threading dislocations thereby creating a vertical conductive path [7]. Through homoepitaxy on bulk GaN, HEMTs with an order of magnitude or lower TD densities can be produced. Though current HEMT devices grown on SiC exhibit exceptional device perfor mance it is worth investigating whether homoepitaxially grown N-polar HEMTs yield any improvement over current HEMT technology. Growth of N-face GaN is particularly challenging due to the nature of the N-face surface. Theoretical calculations made for Ga-polar surfaces predict a much higher diffusion barrier for N compared to Ga - 1.4 eV and 0.4 eV, respectively [8]. This leads to challenges in particular for epitaxy via MOCVD, using N-rich conditions, which were mitigated by growing at temperatures above 1000 � C on miscut substrates [3]. Growth by metal-rich MBE, however, occurs under a Ga-adlayer which modifies the surface energy such that step-flow growth can be achieved at lower temperatures on non-miscut substrates [9]. Thus, for Ga-rich growth, Ga acts as an auto-surfactant, lowering the adatom diffusion barrier promoting step-flow growth [10].
* Corresponding author. Present address: 792 Willow Walk apt G, Goleta, CA 93117, USA. E-mail addresses: [email protected] (C. Wurm), [email protected] (E. Ahmadi), [email protected] (F. Wu), [email protected] (N. Hatui), stacia@ece. ucsb.edu (S. Keller), [email protected] (J. Speck), [email protected] (U. Mishra). https://doi.org/10.1016/j.ssc.2019.113763 Received 6 June 2019; Received in revised form 13 August 2019; Accepted 9 October 2019 Available online 12 October 2019 0038-1098/© 2019 Published by Elsevier Ltd.
C. Wurm et al.
Solid State Communications 305 (2020) 113763
MBE yields films with highly abrupt interfaces and offers in-situ growth monitoring by refection high-energy diffraction, or RHEED, making it ideal for proof-of-concept devices. Unlike MOCVD, MBE Npolar GaN grown on bulk GaN and SiC has been achieved without the use of vicinal substrates [3,11,12]. Before high quality N-polar devices can be grown and processed on bulk GaN substrates growth conditions must be optimized to ensure smooth surface morphology and high quality films. Past N-polar homoepitaxially grown GaN resulted in films with pits on the surface [4, 13,14]. These pits have also been observed in Ga-polar films by Heying at al and Tarsa et al. in Refs. [15,16] respectively. Heying et al. attrib uted these pits, or surface depressions, to different types of surface terminated dislocations [15]. Pits which were observed in MBE grown N-polar films by Turski et al. and Cheze et al. on bulk GaN, in Refs. [13, 17] respectively, were attributed to slow growing parts of a step meander when two adjacent meanders connect [17]. It was shown, in MBE, that these surface pits could be suppressed by using miscut sub strates which reduced the terrace widths on the surface thereby making the steps more periodic; small periodic terraces reduce the effects of unusual step-meandering [13,17]. In this work we will present a 2-step approach to eliminating surface-pits. First, the substrates were subjected to an ex-situ UV-ozone clean. Next, the growth was initiated with a 2 nm thick AlN layer grown under Ga-rich conditions which we call the AlN initiation layer (AIL). By using these two steps N-polar homoepitaxial grown films almost completely free of pits were demonstrated. Growing thicker AILs (up to 8 nm) showed to improve film quality further while still maintaining good surface morphology. Interestingly, the surface pits could be recreated when the GaN films were highly doped with carbon (>5:3 � 1019 cm 3 ). The results suggest that these faceted pits tend to nucleate around impurities or clusters of impurities on the sur face such as Si and C. To the best of our knowledge, this is the first report on the growth of pit free N-polar GaN films by MBE on on-axis GaN substrates.
Fig. 1. (1.a - left) AFM micrograph for sample A1, (1.b - right) AFM micrograph for sample A2 which was subjected to a UV-ozone clean prior to growth. Surface RMS was found to be 12.2 nm and 1.41 nm for A1 and A2 respectively.
(RHEED). Growths were interrupted approximately every 10 min to desorb excess Ga on the surface preventing the formation of droplets. Carbon doping was carried out using a carbon tetra-bromide (CBr4) source. CBr4 was introduced to the system via an automated control valve to throttle the vapor pressure from the CBr4 source to a foreline and into the system [18]. Surface morphology for each sample was characterized by atomic force microscopy (AFM). Samples B3 and C1-C4 were characterized by x-ray diffraction (XRD) rocking-curve scans both on and off axis. Highresolution transmission electron microscopy (HRTEM) imaging was also carried out on sample B2 along with convergent-beam electron diffraction (CBED) to rule out any possibility of polarity inversion as a result of the AIL. Film thickness for all samples, except A1-2, were verified by measuring interference fringe spacing from XRD ω-2θ scans a technique which is explained in more detail in Ref. [19]. To verify that the thickest AIL was strained to the GaN a reciprocal space map (RSM) scan was generated by XRD on sample C4.
2. Experimental procedure
3. Effects of UV ozone clean (A-series)
Four sets of experiments were carried out to observe the effects of the UV-ozone clean (A1-A2), the effects of the 2 nm AlN initiation layer (B1B3), AIL thickness (C1-C4) and the effect of carbon doping on surface morphology (D1-D3). For every sample in this study semi-insulating (SI) N-face epi-ready bulk GaN substrates provided by NGK Insulators were used. With the exception of sample A1, every sample underwent a 15-min ultra-violet (UV) ozone clean followed by a 1-min dip in hydrofluoric (HF) acid. The UV-ozone treatment and the HF dip were repeated two more times before 500 nm of Ti was deposited on the back of the wafers using an electron-beam evaporator. Prior to loading into the growth system, the samples were cleaned in acetone, methanol and isopropyl alcohol. All samples were mounted to 300 Si substrates by In. The samples in this study were grown in a Varian Gen II MBE system with conventional Al, Si and Ga effusion cells. A Riber rf-plasma source was used to supply active nitrogen for growth. Ultrahigh-purity nitrogen (99.9995% purity) was used. Growth conditions consisted of flowing 3 sccm of N2 at a plasma power of 250 W. To maintain Ga-rich conditions the Ga-flux used for both the GaN and AlN growth was kept constant at a beam equivalent pressure (BEP) of 4:4 � 10 7 torr while Al-flux for AlN growth was 1:4 � 10 7 torr. All AlN layers were grown in the presence of a Ga-flux to maintain a Ga-adlayer during growth. Chamber pressure during growth was approximately 3 � 10 5 Torr. Under these conditions the growth rate was measured to be approximately 60 A/min. Substrate temperature was monitored during growth by an optical pyrometer calibrated to the melting point of Al. To ensure good surface morphology all samples were grown under Ga-rich conditions such that a Ga-adlayer was present on the surface throughout the entire growth. The Ga-adlayer was monitored using reflection high energy electron diffraction
A buildup of impurities at the epi/substrate interface, or regrowth interface, are a common occurrence and are believed to be caused by either impurities in the atmosphere accumulating on the surface, re sidual impurities left as a result of the chemical mechanical polishing of the substrate by the manufacture or from gettering inside the MBE chamber [18]. Results from secondary ion mass spectrometry (SIMS), not shown in this work, have confirmed this spike in impurity concen tration at the regrowth interface. Since the N-polar surface is particu larly sensitive to growth it would be natural to see if subjecting the substrate to some kind of cleaning process before growth would improve the surface morphology of the epitaxially grown GaN. In the past UV-ozone treatments had been used by Gupta et al. to remove Si particulates from a post RIE-etched surface of Si-doped GaN [20,30]. Exposing the GaN surface to O3 and UV light converts the GaN surface into a thin gallium oxide layer. Submerging the substrate in hydrofluoric acid (HF) strips the oxide from the surface along with any impurities absorbed by it resulting in a cleaner surface. The AFM mi crographs shown in Fig. 1 show a decrease in pit density as a result of the UV-ozone treatment with a corresponding surface RMS of 12.2 nm and 1.41 nm for A1 and A2 respectively. Although the UV-ozone treatment lead to smoother surface morphology and reduced the number of pits it did not eliminate them completely; more impurities could have been introduced after the UV-ozone clean either from the atmosphere or gettering in the chamber. This, however, was the first piece of evidence to suggest that impurities on the surface play some role in the generation of these pits.
2
C. Wurm et al.
Solid State Communications 305 (2020) 113763
Fig. 2. B-series AFM micrographs (top) with corresponding structures below.
Fig. 3. a. Left – HRTEM image taken for B1 at the AIL. Cross-section direction is (11-20). b. Right-TEM images and CBED patterns for B2 showing both the simulated and experimental patterns for above and below the AIL.
4. AIL (AlN initiation layer) B-series
initiated with AlN, the surface pits disappeared, while inserting AlN after only 3 nm of GaN growth does not lead to any improvements compared to the sample without AlN (B3). This, along with the results of the UV-ozone treatment shown in Fig. 1, suggest that the nucleation of pits occurs at the substrate interface and not later in the growth. SIMS results, not shown in this work, have confirmed a spike in impurity concentration in AlN interlayers used in GaN which is typically larger than what is seen at the regrowth interface. Given the higher bond en ergy Al has with other species with respect to Ga and In, it is thought that these AlN interlayers getter impurities from inside the growth chamber [25]. TEM images shown in Fig. 3 suggest two possible origins of pitgeneration when growing on the GaN substrate surface. The HRTEM image for B1 shown in Fig. 3 reveals that the AlN/substrate interface is not entirely flat possibly due to the chemomechanical polishing of the substrate by the manufacturer. Because of the higher bond energy of Al with respect to Ga and N [25], it is likely that the incoming Al-flux more readily bonds with impurities on the surface compared to Ga. As the AlN continues to grow the surface smoothens as is seen in sample B1 in Fig. 2. It is on this abrupt, impurity free, AlN surface that smooth GaN can grow in step-flow mode without generating pits. Provided the AlN is thin enough such that it is fully strained to the GaN, no new dislocations will be generated due to lattice mismatch between AlN and GaN. Results from the UV-ozone treatment shown in Fig. 1 also support this theory since we see a reduction in pits as a result of cleaning of the substrate
In the past thin layers of Al/group-V films have been used as nucleation layers and as diffusion barriers for impurities during epitaxial growth. A 20 nm AlAs layer, for example, had been used by Ibbetson et al. as a diffusion barrier between doped GaAs and low temperature grown undoped GaAs [29]. Similarly, for GaN grown on SiC, thicker layers, about 70 nm, of AlN nucleation layers grown under nitrogen-rich conditions have also been used to block impurities from the SiC sub strate from diffusing into the MBE-grown HEMT structures [6,21,22]. AlN has also been used as a buffer layer for GaN grown on Si(111) [23, 24]. Given the success AlN has shown as a nucleation layer for hetero epitaxially grown GaN and as a diffusion barrier for impurities it seemed interesting to see how it would affect growth on a native substrate. Three structures were grown to observe the effects of the AlN initi ation layer which are illustrated in the lower part of Fig. 2. Sample B1 was initiated with 2 nm of AlN followed by 250 nm of GaN. Sample B2 was initiated with 3 nm of GaN followed by 2 nm of AlN before the subsequent 250 nm of GaN was grown. Finally, for sample B3, 250 nm of GaN was grown directly on the GaN substrate. As stated above, growth conditions for the AlN layers did not differ from that of the GaN layers, that is, the AlN layers were also grown at 740� under the same Ga flux but with an Al BEP of 1:4 � 10 7 torr. AFM micrographs shown in Fig. 2 confirm that the 2 nm AIL leads to the elimination of pits on the surface of sample B1. Only when growth is 3
C. Wurm et al.
Solid State Communications 305 (2020) 113763
Fig. 4. a. (left) AFM micrographs for C-series samples where the surface RMS roughness for C1-4 is 0.50 nm, 0.51 nm, 0.21 nm and 0.56 nm respectively, b. (right) XRD w-scan FWHM values for all of the C-series samples as well as B3.
5. Employing thicker AILs to improve surface morphology and film quality Based off the dramatic results obtained by using the 2 nm AIL as seen in Fig. 2, further experiments were conducted to investigate if thicker AILs would further improve film quality and surface morphology. Samples C1-C4 were each initiated with a 2 nm, 4 nm, 6 nm and 8 nm AIL, respectively. The resulting AFM images and XRD rocking curve data are shown in Fig. 4 below. For all four samples the surface RMS roughness was less than 1 nm. Looking at Fig. 4.a we see that although the surfaces of C1 and C2 are free of pits, the steps appear more random compared to those in C3 and C4 where thicker AILs were used. Although the steps are not as straight in C4 compared to that of C3, which appears more characteristic of GaN grown on a miscut substrate, both C4 and C3 look smoother compared to C1 and C2 where thinner AILs were used. Another benefit of MBE is in its ability to grow thick AlN interlayers coherently strained to GaN [26]. As mentioned previously, if the AIL remains coherently strained to the GaN, no dislocations could be generated due to the GaN/AlN lattice mismatch. The RSM scan for sample C4 shown in Fig. 5 confirms that the 8 nm AIL is indeed fully strained to GaN. For future work it would be helpful to see what the critical AlN thickness is such that maximum film-quality and surface morphology is achieved without relaxation. From the results observed in Fig. 4 it can be concluded that using a thicker AIL will improve film quality and surface morphology. When the subsequent GaN starts to grow on the smoother and thicker AIL in stepflow mode there is less likelihood of generating pits or depressions. As discussed in section 3, the AIL helps to decouple the MBE grown GaN from impurities and defects on the substrate surface; from Fig. 4 it is demonstrated that this decoupling effect is intensified when going to a thicker AIL.
Fig. 5. XRD RSM scan done at the (105) plane for the sample C4 which was grown using the 8 nm AIL.
surface. Furthermore, the TEM image shown in Fig. 3b reveals a large number of defects in the substrate. Pits may also be generated where these defects intersect with the surface. Given that AlN buffers and nucleation layers have shown to produce smooth GaN when grown on non-native substrates [6,21,22], [24] it is likely that AlN is less sensitive to generating pits around defects compared to GaN. This dramatic improvement in surface morphology resulting from the AIL begs the question as to whether the initial AlN layer actually inverted the polarity of the GaN film. That is, did the film grown following the AIL become Ga-polar? It has been demonstrated that AlN buffer layers used to grow GaN by PAMBE on (111) Si under metal-rich conditions inverted the polarity of the AlN from N- to Al-polar by using highly Al-rich conditions [24]. However, CBED patterns shown in Fig. 3 for the substrate and the epi grown film for this study, compared with simulated results, reveal that no polarity inversion took place due to the AIL and both the substrate and the epi are indeed N-polar.
6. Carbon doping Carbon doping by CBr4 in MBE grown GaN has been used extensively to grow semi-insulating GaN buffer layers for HEMTs [27]. It has been demonstrated that when C-atoms occupy N-sites of the naturally N-type GaN, they act as deep level acceptors pulling the Fermi level away from the conduction band making the film more insulating [18,28]. This has enabled the fabrication of semi-insulating GaN for transistor applica tions. Furthermore, the AIL described above could induce a parasitic channel near the regrowth interface which could potentially be miti gated by doping with some deep acceptor such as C. Because of this it was necessary to see how C-doping effected the surface morphology for N-polar GaN grown on bulk GaN. 4
C. Wurm et al.
Solid State Communications 305 (2020) 113763
References [1] B. Romanczyk, S. Wienecke, M. Guidry, H. Li, E. Ahmadi, X. Zheng, S. Keller, U. K. Mishra, Demonstration of constant 8 W/mm power density at 10, 30, and 94 GHz in state-of-the-art millimeter-wave N-polar GaN MISHEMTs, IEEE Trans. Electron Devices 65 (2018) 45–50, https://doi.org/10.1109/TED.2017.2770087. [2] S. Wienecke, B. Romanczyk, M. Guidry, H. Li, E. Ahmadi, K. Hestroffer, X. Zheng, S. Keller, U.K. Mishra, N-polar GaN cap MISHEMT with record power density exceeding 6.5 W/mm at 94 GHz, IEEE Electron. Device Lett. 38 (2017) 359–362, https://doi.org/10.1109/LED.2017.2653192. [3] S. Keller, H. Li, M. Laurent, Y. Hu, N. Pfaff, J. Lu, D.F. Brown, N.A. Fichtenbaum, J. S. Speck, S.P. DenBaars, U.K. Mishra, Recent progress in metal-organic chemical vapor deposition of (000-1) N-polar group-III nitrides, Semicond. Sci. Technol. 29 (2014), 113001, https://doi.org/10.1088/0268-1242/29/11/113001. [4] M.H. Wong, S. Keller, N. Dasgupta Sansaptak, D.J. Denninghoff, S. Kolluri, D. F. Brown, J. Lu, N.A. Fichtenbaum, E. Ahmadi, U. Singisetti, A. Chini, S. Rajan, S. P. DenBaars, J.S. Speck, U.K. Mishra, N-polar GaN epitaxy and high electron mobility transistors, Semicond. Sci. Technol. 28 (2013), 074009, https://doi.org/ 10.1088/0268-1242/28/7/074009. [5] Power GaN Devices, Springer Berlin Heidelberg, New York, NY, 2016. [6] S.W. Kaun, M.H. Wong, U.K. Mishra, J.S. Speck, Correlation between threading dislocation density and sheet resistance of AlGaN/AlN/GaN heterostructures grown by plasma-assisted molecular beam epitaxy, Appl. Phys. Lett. 100 (2012), 262102, https://doi.org/10.1063/1.4730951. [7] Z. Yang, Y. Zhang, D.N. Nath, J.B. Khurgin, S. Rajan, Current gain in sub-10 nm base GaN tunneling hot electron transistors with AlN emitter barrier, Appl. Phys. Lett. 106 (2015), 032101, https://doi.org/10.1063/1.4906287. [8] T.K. Zywietz, J. Neugebauer, M. Scheffler, The adsorption of oxygen at GaN surfaces, Appl. Phys. Lett. 74 (1999) 1695–1697, https://doi.org/10.1063/ 1.123658. [9] T. Zywietz, J. Neugebauer, M. Scheffler, Adatom diffusion at GaN (0001) and (0001) surfaces, Appl. Phys. Lett. 73 (1998) 487–489, https://doi.org/10.1063/ 1.121909. [10] J. Neugebauer, T.K. Zywietz, M. Scheffler, J.E. Northrup, H. Chen, R.M. Feenstra, Adatom kinetics on and below the surface: the existence of a new diffusion channel, Phys. Rev. Lett. 90 (2003), https://doi.org/10.1103/ PhysRevLett.90.056101. [11] Z.P. Guan, A.L. Cai, J.S. Cabalu, H.L. Porter, S. Huang, Molecular beam epitaxy growth of GaN on C-terminated 6H–SiC (0001) surface, Appl. Phys. Lett. 77 (2000) 2491–2493, https://doi.org/10.1063/1.1318723. [12] E. Ahmadi, F. Wu, H. Li, S.W. Kaun, M. Tahhan, K. Hestroffer, S. Keller, J.S. Speck, U.K. Mishra, N-face GaN/AlN/GaN/InAlN and GaN/AlN/AlGaN/GaN/InAlN highelectron-mobility transistor structures grown by plasma-assisted molecular beam epitaxy on vicinal substrates, Semicond. Sci. Technol. 30 (2015), 055012, https:// doi.org/10.1088/0268-1242/30/5/055012. [13] C. Ch�eze, M. Sawicka, M. Siekacz, H. Turski, G. Cywi� nski, J. Smalc-Koziorowska, J. L. Weyher, M. Kry�sko, B. Łucznik, M. Bo�ckowski, C. Skierbiszewski, Step-flow growth mode instability of N-polar GaN under N-excess, Appl. Phys. Lett. 103 (2013), 071601, https://doi.org/10.1063/1.4818322. [14] S.S. Pasayat, E. Ahmadi, B. Romanczyk, O. Koksaldi, A. Agarwal, M. Guidry, C. Gupta, C. Wurm, S. Keller, U.K. Mishra, First demonstration of RF N-polar GaN MIS-HEMTs grown on bulk GaN using PAMBE, Semicond. Sci. Technol. 34 (2019), 045009, https://doi.org/10.1088/1361-6641/ab0761. [15] B. Heying, E.J. Tarsa, C.R. Elsass, P. Fini, S.P. DenBaars, J.S. Speck, Dislocation mediated surface morphology of GaN, J. Appl. Phys. 85 (1999) 6470–6476, https://doi.org/10.1063/1.370150. [16] E.J. Tarsa, B. Heying, X.H. Wu, P. Fini, S.P. DenBaars, J.S. Speck, Homoepitaxial growth of GaN under Ga-stable and N-stable conditions by plasma-assisted molecular beam epitaxy, J. Appl. Phys. 82 (1997) 5472–5479, https://doi.org/ 10.1063/1.365575. _ [17] H. Turski, F. Krzy_zewski, A. Feduniewicz-Zmuda, P. Wolny, M. Siekacz, G. Muziol, C. Cheze, K. Nowakowski-Szukudlarek, H.G. Xing, D. Jena, M. Zaluska-Kotur, C. Skierbiszewski, Unusual step meandering due to Ehrlich-Schwoebel barrier in GaN epitaxy on the N-polar surface, Appl. Surf. Sci. (2019), https://doi.org/ 10.1016/j.apsusc.2019.04.082. [18] D.S. Green, U.K. Mishra, J.S. Speck, Carbon doping of GaN with CBr4 in radiofrequency plasma-assisted molecular beam epitaxy, J. Appl. Phys. 95 (2004) 8456–8462, https://doi.org/10.1063/1.1755431. [19] M.A. Moram, M.E. Vickers, X-ray diffraction of III-nitrides, Rep. Prog. Phys. 72 (2009), 036502, https://doi.org/10.1088/0034-4885/72/3/036502. [20] C. Gupta, C. Lund, S.H. Chan, A. Agarwal, J. Liu, Y. Enatsu, S. Keller, U.K. Mishra, In situ oxide, GaN interlayer-based vertical trench MOSFET (OG-FET) on bulk GaN substrates, IEEE Electron. Device Lett. 38 (2017) 353–355, https://doi.org/ 10.1109/LED.2017.2649599. [21] S.W. Kaun, M.H. Wong, U.K. Mishra, J.S. Speck, Molecular beam epitaxy for highperformance Ga-face GaN electron devices, Semicond. Sci. Technol. 28 (2013), 074001, https://doi.org/10.1088/0268-1242/28/7/074001. [22] S. Rajan, M. Wong, Y. Fu, F. Wu, J.S. Speck, U.K. Mishra, Growth and electrical characterization of N-face AlGaN/GaN heterostructures, Jpn. J. Appl. Phys. 44 (2005) L1478–L1480, https://doi.org/10.1143/JJAP.44.L1478. [23] L. Ravikiran, M. Agrawal, N. Dharmarasu, K. Radhakrishnan, Effect of stress mitigating layers on the structural properties of GaN grown by ammonia molecular beam epitaxy on 100 mm Si(111), Jpn. J. Appl. Phys. 52 (2013), 08JE05, https:// doi.org/10.7567/JJAP.52.08JE05. [24] M. Agrawal, K. Radhakrishnan, N. Dharmarasu, S.S. Pramana, Effect of III/V ratio on the polarity of AlN and GaN layers grown in the metal rich growth regime on Si
Fig. 6. AFM micrographs for samples D1-D3 (left to right respectively) where C-doping is varied from 5:3 � 1019 ; 2:5 � 1020 and 8:3 � 1020 cm 3 respectively.
For samples D1-D3 growth was initiated with 2 nm of AlN followed by 200 nm of GaN:C with CBr4 foreline pressure of 20, 40 and 60 mTorr, corresponding to a C-doping concentration of approximately 5:3� 1019 ; 2:5 � 1020 and 8:3 � 1020 cm 3 respectively. C-concentration as a function of CBr4 foreline pressure was determined from a multi-layered C-doped SIMS stack carried out in a previous experiment. AFM micro graphs showing the surface morphology of D1-3 are depicted in Fig. 6 below. The AFM micrographs in Fig. 6 reveal that the surface pits return for high C-concentrations. Sample D1, where C-doping was 5:3� 1019 cm 3 , exhibited almost no pits however it can be seen that in certain regions, particularly where two adjacent step meanders meet, there exists de pressions which could potentially lead to the formation of pits had the film continued to grow. From Fig. 6 we see that these pits return for some C-doping concentration between 5:3 � 1019 2:5 � 1020 cm 3 . The results of the C-doping growth series provide further evidence suggesting impurities play a role in the generation of surface-pits. However, surface pits on N-polar GaN films can also be generated by other factors including surface defects, dislocations and slow-moving step meanders. To the best of our knowledge it is unknown as to whether this same effect is seen in Ga-polar films utilizing such high Cconcentrations. 7. Conclusion In this work we have demonstrated that smooth, pit-free N-polar GaN can be grown on non-vicinal GaN substrates by PAMBE using the AlN initiation layer or AIL. Considering the higher bond energy Al has with other species with respect to Ga, combined with AlN’s long history of being a reliable nucleation layer for heteroepitaxially grown GaN, it is likely the AIL grows over defects and impurities on the substrate surface thereby creating a smooth surface for which N-polar GaN can be grown on. Thus the AIL effectively decouples the epitaxially grown GaN from the substrate surface. It was shown that thicker AILs, up to 8 nm, further improve film quality and surface morphology while still remaining strained to the GaN. By varying C-doping for films grown with the AIL it was shown that pits can again be generated on the surface for C-doping concentration between 5:3 � 1019 2:5 � 1020 cm 3 . This further supported the hypothesis that impurities contribute to pit formation in N-polar GaN films grown by MBE. The findings in this study pave the way for the MBE growth of high performance N-polar HEMT structures on bulk GaN. Acknowledgements The authors gratefully acknowledge the funding from the Office of Naval Research (Dr. P. Maki) and the DARPA DREaM program. SIMS measurements were carried out by Dr. Tom Mates of the California NanoSystems Institute (CNSI). The authors would also like to thank Kelsey Jorgenson and Richard Cramer for their helpful discussions and advice regarding PAMBE of GaN.
5
C. Wurm et al.
Solid State Communications 305 (2020) 113763
(111) by plasma assisted molecular beam epitaxy, Jpn. J. Appl. Phys. 54 (2015), 065701, https://doi.org/10.7567/JJAP.54.065701. [25] M. Stutzmann, O. Ambacher, M. Eickhoff, U. Karrer, A. Lima Pimenta, R. Neuberger, J. Schalwig, R. Dimitrov, P.J. Schuck, R.D. Grober, Playing with polarity, Phys. Status Solidi B 228 (2001) 505–512, https://doi.org/10.1002/ 1521-3951(200111)228:2<505::AID-PSSB505>3.0.CO;2-U. [26] A. Adikimenakis, S.-L. Sahonta, G.P. Dimitrakopulos, J. Domagala, Th. Kehagias, Ph. Komninou, E. Iliopoulos, A. Georgakilas, Effect of AlN interlayers in the structure of GaN-on-Si grown by plasma-assisted MBE, J. Cryst. Growth 311 (2009) 2010–2015, https://doi.org/10.1016/j.jcrysgro.2008.10.085.
[27] C. Poblenz, P. Waltereit, S. Rajan, S. Heikman, U.K. Mishra, J.S. Speck, Effect of carbon doping on buffer leakage in AlGaN/GaN high electron mobility transistors, J. Vac. Sci. Technol. B Microelectron. Nanometer Struct. 22 (2004) 1145, https:// doi.org/10.1116/1.1752907. [28] J.L. Lyons, A. Janotti, C.G. Van de Walle, Carbon impurities and the yellow luminescence in GaN, Appl. Phys. Lett. 97 (2010), 152108, https://doi.org/ 10.1063/1.3492841. [29] J. Ibbetson, Electrical Characterization of Nonstoichiometric GaAs Grown at Low Temperature by Molecular Beam Epitaxy, Thesis, 1997. [30] S. Chowdhury, AlGaN/GaN CAVETs for High Power Switching Application, Thesis, 2010.
6
Contents lists available at ScienceDirect
Solid State Communications journal homepage: http://www.elsevier.com/locate/ssc
Growth of high-quality N-polar GaN on bulk GaN by plasma-assisted molecular beam epitaxy Christian Wurm a, *, Elaheh Ahmadi b, Feng Wu c, Nirupam Hatui a, Stacia Keller c, James Speck c, Umesh Mishra a a b c
Department of Electrical and Computer Engineering, University of California, Santa Barbara, CA, 93117, USA Department of Electrical Engineering and Computer Science, University of Michigan, MI, 48109, USA Materials Department, University of California, Santa Barbara, CA, 93117, USA
A R T I C L E I N F O
A B S T R A C T
Communicated by Bell Gavin
’There is an interest in growing N-polar GaN on bulk GaN for high electron mobility transistors (HEMTs). Current N-polar HEMT technology is dominated by devices grown on non-native substrates which possess a high density of dislocations due to lattice mismatch. N-polar GaN films grown directly on non-miscut GaN substrates have displayed a high density of pits and depressions on the surface. This work demonstrates that surface impurities play a major role in the formation of surface-pits in epitaxially grown film. By subjecting the GaN substrate to an ultra-violet (UV) O3 (ozone) clean prior to growth and initiating growth with a 2 nm coherently strained AlN layer, grown under metal-rich conditions, we have demonstrated N-polar GaN films with a nearly pit-free sur face. This AlN initiation layer (AIL) likely captures impurities on the substrate surface thus decoupling the substrate surface from the epitaxially grown film. It is demonstrated that utilizing thicker AILs, up to 8 nm, further improves film quality and surface morphology. The methods employed in this study to produce highquality N-polar GaN grown on bulk GaN will pave the way for future GaN devices with an order of magni tude or more lower threading dislocation density (TDD).
Keywords: A. Semiconductors B. Molecular beam epitaxy C. Nitrides D. HEMT
1. Introduction The need for high-quality ð0001Þ GaN, or N-polar GaN, films grown on bulk GaN comes from its applications in high power, high frequency electronics. N-polar high electron mobility transistors, or HEMTs, have shown superior properties with respect to their Ga-polar counterparts in terms of operating frequency, power output and scalability [1,2]. High quality N-polar GaN grown by metal-organic chemical vapor deposition (MOCVD) on vicinal sapphire and SiC has already been used extensively to make high performance HEMTs [3]. However, growing GaN hetero epitaxially on SiC, for example, where an in-plane 3.4% lattice mismatch [4] is present, results in vertically propagating threading dislocations (TDs). A threading dislocation density (TDD) on the order of 108 cm 2 has been reported for GaN grown on sapphire and SiC [5]. In high concentrations, threading dislocations can act as traps impeding mobility and lowering device performance [6]. Dislocations in GaN are also a major cause of vertical leakage when grown by metal-rich
plasma-assisted MBE (PAMBE); metal-rich growth results in metal-decorated threading dislocations thereby creating a vertical conductive path [7]. Through homoepitaxy on bulk GaN, HEMTs with an order of magnitude or lower TD densities can be produced. Though current HEMT devices grown on SiC exhibit exceptional device perfor mance it is worth investigating whether homoepitaxially grown N-polar HEMTs yield any improvement over current HEMT technology. Growth of N-face GaN is particularly challenging due to the nature of the N-face surface. Theoretical calculations made for Ga-polar surfaces predict a much higher diffusion barrier for N compared to Ga - 1.4 eV and 0.4 eV, respectively [8]. This leads to challenges in particular for epitaxy via MOCVD, using N-rich conditions, which were mitigated by growing at temperatures above 1000 � C on miscut substrates [3]. Growth by metal-rich MBE, however, occurs under a Ga-adlayer which modifies the surface energy such that step-flow growth can be achieved at lower temperatures on non-miscut substrates [9]. Thus, for Ga-rich growth, Ga acts as an auto-surfactant, lowering the adatom diffusion barrier promoting step-flow growth [10].
* Corresponding author. Present address: 792 Willow Walk apt G, Goleta, CA 93117, USA. E-mail addresses: [email protected] (C. Wurm), [email protected] (E. Ahmadi), [email protected] (F. Wu), [email protected] (N. Hatui), stacia@ece. ucsb.edu (S. Keller), [email protected] (J. Speck), [email protected] (U. Mishra). https://doi.org/10.1016/j.ssc.2019.113763 Received 6 June 2019; Received in revised form 13 August 2019; Accepted 9 October 2019 Available online 12 October 2019 0038-1098/© 2019 Published by Elsevier Ltd.
C. Wurm et al.
Solid State Communications 305 (2020) 113763
MBE yields films with highly abrupt interfaces and offers in-situ growth monitoring by refection high-energy diffraction, or RHEED, making it ideal for proof-of-concept devices. Unlike MOCVD, MBE Npolar GaN grown on bulk GaN and SiC has been achieved without the use of vicinal substrates [3,11,12]. Before high quality N-polar devices can be grown and processed on bulk GaN substrates growth conditions must be optimized to ensure smooth surface morphology and high quality films. Past N-polar homoepitaxially grown GaN resulted in films with pits on the surface [4, 13,14]. These pits have also been observed in Ga-polar films by Heying at al and Tarsa et al. in Refs. [15,16] respectively. Heying et al. attrib uted these pits, or surface depressions, to different types of surface terminated dislocations [15]. Pits which were observed in MBE grown N-polar films by Turski et al. and Cheze et al. on bulk GaN, in Refs. [13, 17] respectively, were attributed to slow growing parts of a step meander when two adjacent meanders connect [17]. It was shown, in MBE, that these surface pits could be suppressed by using miscut sub strates which reduced the terrace widths on the surface thereby making the steps more periodic; small periodic terraces reduce the effects of unusual step-meandering [13,17]. In this work we will present a 2-step approach to eliminating surface-pits. First, the substrates were subjected to an ex-situ UV-ozone clean. Next, the growth was initiated with a 2 nm thick AlN layer grown under Ga-rich conditions which we call the AlN initiation layer (AIL). By using these two steps N-polar homoepitaxial grown films almost completely free of pits were demonstrated. Growing thicker AILs (up to 8 nm) showed to improve film quality further while still maintaining good surface morphology. Interestingly, the surface pits could be recreated when the GaN films were highly doped with carbon (>5:3 � 1019 cm 3 ). The results suggest that these faceted pits tend to nucleate around impurities or clusters of impurities on the sur face such as Si and C. To the best of our knowledge, this is the first report on the growth of pit free N-polar GaN films by MBE on on-axis GaN substrates.
Fig. 1. (1.a - left) AFM micrograph for sample A1, (1.b - right) AFM micrograph for sample A2 which was subjected to a UV-ozone clean prior to growth. Surface RMS was found to be 12.2 nm and 1.41 nm for A1 and A2 respectively.
(RHEED). Growths were interrupted approximately every 10 min to desorb excess Ga on the surface preventing the formation of droplets. Carbon doping was carried out using a carbon tetra-bromide (CBr4) source. CBr4 was introduced to the system via an automated control valve to throttle the vapor pressure from the CBr4 source to a foreline and into the system [18]. Surface morphology for each sample was characterized by atomic force microscopy (AFM). Samples B3 and C1-C4 were characterized by x-ray diffraction (XRD) rocking-curve scans both on and off axis. Highresolution transmission electron microscopy (HRTEM) imaging was also carried out on sample B2 along with convergent-beam electron diffraction (CBED) to rule out any possibility of polarity inversion as a result of the AIL. Film thickness for all samples, except A1-2, were verified by measuring interference fringe spacing from XRD ω-2θ scans a technique which is explained in more detail in Ref. [19]. To verify that the thickest AIL was strained to the GaN a reciprocal space map (RSM) scan was generated by XRD on sample C4.
2. Experimental procedure
3. Effects of UV ozone clean (A-series)
Four sets of experiments were carried out to observe the effects of the UV-ozone clean (A1-A2), the effects of the 2 nm AlN initiation layer (B1B3), AIL thickness (C1-C4) and the effect of carbon doping on surface morphology (D1-D3). For every sample in this study semi-insulating (SI) N-face epi-ready bulk GaN substrates provided by NGK Insulators were used. With the exception of sample A1, every sample underwent a 15-min ultra-violet (UV) ozone clean followed by a 1-min dip in hydrofluoric (HF) acid. The UV-ozone treatment and the HF dip were repeated two more times before 500 nm of Ti was deposited on the back of the wafers using an electron-beam evaporator. Prior to loading into the growth system, the samples were cleaned in acetone, methanol and isopropyl alcohol. All samples were mounted to 300 Si substrates by In. The samples in this study were grown in a Varian Gen II MBE system with conventional Al, Si and Ga effusion cells. A Riber rf-plasma source was used to supply active nitrogen for growth. Ultrahigh-purity nitrogen (99.9995% purity) was used. Growth conditions consisted of flowing 3 sccm of N2 at a plasma power of 250 W. To maintain Ga-rich conditions the Ga-flux used for both the GaN and AlN growth was kept constant at a beam equivalent pressure (BEP) of 4:4 � 10 7 torr while Al-flux for AlN growth was 1:4 � 10 7 torr. All AlN layers were grown in the presence of a Ga-flux to maintain a Ga-adlayer during growth. Chamber pressure during growth was approximately 3 � 10 5 Torr. Under these conditions the growth rate was measured to be approximately 60 A/min. Substrate temperature was monitored during growth by an optical pyrometer calibrated to the melting point of Al. To ensure good surface morphology all samples were grown under Ga-rich conditions such that a Ga-adlayer was present on the surface throughout the entire growth. The Ga-adlayer was monitored using reflection high energy electron diffraction
A buildup of impurities at the epi/substrate interface, or regrowth interface, are a common occurrence and are believed to be caused by either impurities in the atmosphere accumulating on the surface, re sidual impurities left as a result of the chemical mechanical polishing of the substrate by the manufacture or from gettering inside the MBE chamber [18]. Results from secondary ion mass spectrometry (SIMS), not shown in this work, have confirmed this spike in impurity concen tration at the regrowth interface. Since the N-polar surface is particu larly sensitive to growth it would be natural to see if subjecting the substrate to some kind of cleaning process before growth would improve the surface morphology of the epitaxially grown GaN. In the past UV-ozone treatments had been used by Gupta et al. to remove Si particulates from a post RIE-etched surface of Si-doped GaN [20,30]. Exposing the GaN surface to O3 and UV light converts the GaN surface into a thin gallium oxide layer. Submerging the substrate in hydrofluoric acid (HF) strips the oxide from the surface along with any impurities absorbed by it resulting in a cleaner surface. The AFM mi crographs shown in Fig. 1 show a decrease in pit density as a result of the UV-ozone treatment with a corresponding surface RMS of 12.2 nm and 1.41 nm for A1 and A2 respectively. Although the UV-ozone treatment lead to smoother surface morphology and reduced the number of pits it did not eliminate them completely; more impurities could have been introduced after the UV-ozone clean either from the atmosphere or gettering in the chamber. This, however, was the first piece of evidence to suggest that impurities on the surface play some role in the generation of these pits.
2
C. Wurm et al.
Solid State Communications 305 (2020) 113763
Fig. 2. B-series AFM micrographs (top) with corresponding structures below.
Fig. 3. a. Left – HRTEM image taken for B1 at the AIL. Cross-section direction is (11-20). b. Right-TEM images and CBED patterns for B2 showing both the simulated and experimental patterns for above and below the AIL.
4. AIL (AlN initiation layer) B-series
initiated with AlN, the surface pits disappeared, while inserting AlN after only 3 nm of GaN growth does not lead to any improvements compared to the sample without AlN (B3). This, along with the results of the UV-ozone treatment shown in Fig. 1, suggest that the nucleation of pits occurs at the substrate interface and not later in the growth. SIMS results, not shown in this work, have confirmed a spike in impurity concentration in AlN interlayers used in GaN which is typically larger than what is seen at the regrowth interface. Given the higher bond en ergy Al has with other species with respect to Ga and In, it is thought that these AlN interlayers getter impurities from inside the growth chamber [25]. TEM images shown in Fig. 3 suggest two possible origins of pitgeneration when growing on the GaN substrate surface. The HRTEM image for B1 shown in Fig. 3 reveals that the AlN/substrate interface is not entirely flat possibly due to the chemomechanical polishing of the substrate by the manufacturer. Because of the higher bond energy of Al with respect to Ga and N [25], it is likely that the incoming Al-flux more readily bonds with impurities on the surface compared to Ga. As the AlN continues to grow the surface smoothens as is seen in sample B1 in Fig. 2. It is on this abrupt, impurity free, AlN surface that smooth GaN can grow in step-flow mode without generating pits. Provided the AlN is thin enough such that it is fully strained to the GaN, no new dislocations will be generated due to lattice mismatch between AlN and GaN. Results from the UV-ozone treatment shown in Fig. 1 also support this theory since we see a reduction in pits as a result of cleaning of the substrate
In the past thin layers of Al/group-V films have been used as nucleation layers and as diffusion barriers for impurities during epitaxial growth. A 20 nm AlAs layer, for example, had been used by Ibbetson et al. as a diffusion barrier between doped GaAs and low temperature grown undoped GaAs [29]. Similarly, for GaN grown on SiC, thicker layers, about 70 nm, of AlN nucleation layers grown under nitrogen-rich conditions have also been used to block impurities from the SiC sub strate from diffusing into the MBE-grown HEMT structures [6,21,22]. AlN has also been used as a buffer layer for GaN grown on Si(111) [23, 24]. Given the success AlN has shown as a nucleation layer for hetero epitaxially grown GaN and as a diffusion barrier for impurities it seemed interesting to see how it would affect growth on a native substrate. Three structures were grown to observe the effects of the AlN initi ation layer which are illustrated in the lower part of Fig. 2. Sample B1 was initiated with 2 nm of AlN followed by 250 nm of GaN. Sample B2 was initiated with 3 nm of GaN followed by 2 nm of AlN before the subsequent 250 nm of GaN was grown. Finally, for sample B3, 250 nm of GaN was grown directly on the GaN substrate. As stated above, growth conditions for the AlN layers did not differ from that of the GaN layers, that is, the AlN layers were also grown at 740� under the same Ga flux but with an Al BEP of 1:4 � 10 7 torr. AFM micrographs shown in Fig. 2 confirm that the 2 nm AIL leads to the elimination of pits on the surface of sample B1. Only when growth is 3
C. Wurm et al.
Solid State Communications 305 (2020) 113763
Fig. 4. a. (left) AFM micrographs for C-series samples where the surface RMS roughness for C1-4 is 0.50 nm, 0.51 nm, 0.21 nm and 0.56 nm respectively, b. (right) XRD w-scan FWHM values for all of the C-series samples as well as B3.
5. Employing thicker AILs to improve surface morphology and film quality Based off the dramatic results obtained by using the 2 nm AIL as seen in Fig. 2, further experiments were conducted to investigate if thicker AILs would further improve film quality and surface morphology. Samples C1-C4 were each initiated with a 2 nm, 4 nm, 6 nm and 8 nm AIL, respectively. The resulting AFM images and XRD rocking curve data are shown in Fig. 4 below. For all four samples the surface RMS roughness was less than 1 nm. Looking at Fig. 4.a we see that although the surfaces of C1 and C2 are free of pits, the steps appear more random compared to those in C3 and C4 where thicker AILs were used. Although the steps are not as straight in C4 compared to that of C3, which appears more characteristic of GaN grown on a miscut substrate, both C4 and C3 look smoother compared to C1 and C2 where thinner AILs were used. Another benefit of MBE is in its ability to grow thick AlN interlayers coherently strained to GaN [26]. As mentioned previously, if the AIL remains coherently strained to the GaN, no dislocations could be generated due to the GaN/AlN lattice mismatch. The RSM scan for sample C4 shown in Fig. 5 confirms that the 8 nm AIL is indeed fully strained to GaN. For future work it would be helpful to see what the critical AlN thickness is such that maximum film-quality and surface morphology is achieved without relaxation. From the results observed in Fig. 4 it can be concluded that using a thicker AIL will improve film quality and surface morphology. When the subsequent GaN starts to grow on the smoother and thicker AIL in stepflow mode there is less likelihood of generating pits or depressions. As discussed in section 3, the AIL helps to decouple the MBE grown GaN from impurities and defects on the substrate surface; from Fig. 4 it is demonstrated that this decoupling effect is intensified when going to a thicker AIL.
Fig. 5. XRD RSM scan done at the (105) plane for the sample C4 which was grown using the 8 nm AIL.
surface. Furthermore, the TEM image shown in Fig. 3b reveals a large number of defects in the substrate. Pits may also be generated where these defects intersect with the surface. Given that AlN buffers and nucleation layers have shown to produce smooth GaN when grown on non-native substrates [6,21,22], [24] it is likely that AlN is less sensitive to generating pits around defects compared to GaN. This dramatic improvement in surface morphology resulting from the AIL begs the question as to whether the initial AlN layer actually inverted the polarity of the GaN film. That is, did the film grown following the AIL become Ga-polar? It has been demonstrated that AlN buffer layers used to grow GaN by PAMBE on (111) Si under metal-rich conditions inverted the polarity of the AlN from N- to Al-polar by using highly Al-rich conditions [24]. However, CBED patterns shown in Fig. 3 for the substrate and the epi grown film for this study, compared with simulated results, reveal that no polarity inversion took place due to the AIL and both the substrate and the epi are indeed N-polar.
6. Carbon doping Carbon doping by CBr4 in MBE grown GaN has been used extensively to grow semi-insulating GaN buffer layers for HEMTs [27]. It has been demonstrated that when C-atoms occupy N-sites of the naturally N-type GaN, they act as deep level acceptors pulling the Fermi level away from the conduction band making the film more insulating [18,28]. This has enabled the fabrication of semi-insulating GaN for transistor applica tions. Furthermore, the AIL described above could induce a parasitic channel near the regrowth interface which could potentially be miti gated by doping with some deep acceptor such as C. Because of this it was necessary to see how C-doping effected the surface morphology for N-polar GaN grown on bulk GaN. 4
C. Wurm et al.
Solid State Communications 305 (2020) 113763
References [1] B. Romanczyk, S. Wienecke, M. Guidry, H. Li, E. Ahmadi, X. Zheng, S. Keller, U. K. Mishra, Demonstration of constant 8 W/mm power density at 10, 30, and 94 GHz in state-of-the-art millimeter-wave N-polar GaN MISHEMTs, IEEE Trans. Electron Devices 65 (2018) 45–50, https://doi.org/10.1109/TED.2017.2770087. [2] S. Wienecke, B. Romanczyk, M. Guidry, H. Li, E. Ahmadi, K. Hestroffer, X. Zheng, S. Keller, U.K. Mishra, N-polar GaN cap MISHEMT with record power density exceeding 6.5 W/mm at 94 GHz, IEEE Electron. Device Lett. 38 (2017) 359–362, https://doi.org/10.1109/LED.2017.2653192. [3] S. Keller, H. Li, M. Laurent, Y. Hu, N. Pfaff, J. Lu, D.F. Brown, N.A. Fichtenbaum, J. S. Speck, S.P. DenBaars, U.K. Mishra, Recent progress in metal-organic chemical vapor deposition of (000-1) N-polar group-III nitrides, Semicond. Sci. Technol. 29 (2014), 113001, https://doi.org/10.1088/0268-1242/29/11/113001. [4] M.H. Wong, S. Keller, N. Dasgupta Sansaptak, D.J. Denninghoff, S. Kolluri, D. F. Brown, J. Lu, N.A. Fichtenbaum, E. Ahmadi, U. Singisetti, A. Chini, S. Rajan, S. P. DenBaars, J.S. Speck, U.K. Mishra, N-polar GaN epitaxy and high electron mobility transistors, Semicond. Sci. Technol. 28 (2013), 074009, https://doi.org/ 10.1088/0268-1242/28/7/074009. [5] Power GaN Devices, Springer Berlin Heidelberg, New York, NY, 2016. [6] S.W. Kaun, M.H. Wong, U.K. Mishra, J.S. Speck, Correlation between threading dislocation density and sheet resistance of AlGaN/AlN/GaN heterostructures grown by plasma-assisted molecular beam epitaxy, Appl. Phys. Lett. 100 (2012), 262102, https://doi.org/10.1063/1.4730951. [7] Z. Yang, Y. Zhang, D.N. Nath, J.B. Khurgin, S. Rajan, Current gain in sub-10 nm base GaN tunneling hot electron transistors with AlN emitter barrier, Appl. Phys. Lett. 106 (2015), 032101, https://doi.org/10.1063/1.4906287. [8] T.K. Zywietz, J. Neugebauer, M. Scheffler, The adsorption of oxygen at GaN surfaces, Appl. Phys. Lett. 74 (1999) 1695–1697, https://doi.org/10.1063/ 1.123658. [9] T. Zywietz, J. Neugebauer, M. Scheffler, Adatom diffusion at GaN (0001) and (0001) surfaces, Appl. Phys. Lett. 73 (1998) 487–489, https://doi.org/10.1063/ 1.121909. [10] J. Neugebauer, T.K. Zywietz, M. Scheffler, J.E. Northrup, H. Chen, R.M. Feenstra, Adatom kinetics on and below the surface: the existence of a new diffusion channel, Phys. Rev. Lett. 90 (2003), https://doi.org/10.1103/ PhysRevLett.90.056101. [11] Z.P. Guan, A.L. Cai, J.S. Cabalu, H.L. Porter, S. Huang, Molecular beam epitaxy growth of GaN on C-terminated 6H–SiC (0001) surface, Appl. Phys. Lett. 77 (2000) 2491–2493, https://doi.org/10.1063/1.1318723. [12] E. Ahmadi, F. Wu, H. Li, S.W. Kaun, M. Tahhan, K. Hestroffer, S. Keller, J.S. Speck, U.K. Mishra, N-face GaN/AlN/GaN/InAlN and GaN/AlN/AlGaN/GaN/InAlN highelectron-mobility transistor structures grown by plasma-assisted molecular beam epitaxy on vicinal substrates, Semicond. Sci. Technol. 30 (2015), 055012, https:// doi.org/10.1088/0268-1242/30/5/055012. [13] C. Ch�eze, M. Sawicka, M. Siekacz, H. Turski, G. Cywi� nski, J. Smalc-Koziorowska, J. L. Weyher, M. Kry�sko, B. Łucznik, M. Bo�ckowski, C. Skierbiszewski, Step-flow growth mode instability of N-polar GaN under N-excess, Appl. Phys. Lett. 103 (2013), 071601, https://doi.org/10.1063/1.4818322. [14] S.S. Pasayat, E. Ahmadi, B. Romanczyk, O. Koksaldi, A. Agarwal, M. Guidry, C. Gupta, C. Wurm, S. Keller, U.K. Mishra, First demonstration of RF N-polar GaN MIS-HEMTs grown on bulk GaN using PAMBE, Semicond. Sci. Technol. 34 (2019), 045009, https://doi.org/10.1088/1361-6641/ab0761. [15] B. Heying, E.J. Tarsa, C.R. Elsass, P. Fini, S.P. DenBaars, J.S. Speck, Dislocation mediated surface morphology of GaN, J. Appl. Phys. 85 (1999) 6470–6476, https://doi.org/10.1063/1.370150. [16] E.J. Tarsa, B. Heying, X.H. Wu, P. Fini, S.P. DenBaars, J.S. Speck, Homoepitaxial growth of GaN under Ga-stable and N-stable conditions by plasma-assisted molecular beam epitaxy, J. Appl. Phys. 82 (1997) 5472–5479, https://doi.org/ 10.1063/1.365575. _ [17] H. Turski, F. Krzy_zewski, A. Feduniewicz-Zmuda, P. Wolny, M. Siekacz, G. Muziol, C. Cheze, K. Nowakowski-Szukudlarek, H.G. Xing, D. Jena, M. Zaluska-Kotur, C. Skierbiszewski, Unusual step meandering due to Ehrlich-Schwoebel barrier in GaN epitaxy on the N-polar surface, Appl. Surf. Sci. (2019), https://doi.org/ 10.1016/j.apsusc.2019.04.082. [18] D.S. Green, U.K. Mishra, J.S. Speck, Carbon doping of GaN with CBr4 in radiofrequency plasma-assisted molecular beam epitaxy, J. Appl. Phys. 95 (2004) 8456–8462, https://doi.org/10.1063/1.1755431. [19] M.A. Moram, M.E. Vickers, X-ray diffraction of III-nitrides, Rep. Prog. Phys. 72 (2009), 036502, https://doi.org/10.1088/0034-4885/72/3/036502. [20] C. Gupta, C. Lund, S.H. Chan, A. Agarwal, J. Liu, Y. Enatsu, S. Keller, U.K. Mishra, In situ oxide, GaN interlayer-based vertical trench MOSFET (OG-FET) on bulk GaN substrates, IEEE Electron. Device Lett. 38 (2017) 353–355, https://doi.org/ 10.1109/LED.2017.2649599. [21] S.W. Kaun, M.H. Wong, U.K. Mishra, J.S. Speck, Molecular beam epitaxy for highperformance Ga-face GaN electron devices, Semicond. Sci. Technol. 28 (2013), 074001, https://doi.org/10.1088/0268-1242/28/7/074001. [22] S. Rajan, M. Wong, Y. Fu, F. Wu, J.S. Speck, U.K. Mishra, Growth and electrical characterization of N-face AlGaN/GaN heterostructures, Jpn. J. Appl. Phys. 44 (2005) L1478–L1480, https://doi.org/10.1143/JJAP.44.L1478. [23] L. Ravikiran, M. Agrawal, N. Dharmarasu, K. Radhakrishnan, Effect of stress mitigating layers on the structural properties of GaN grown by ammonia molecular beam epitaxy on 100 mm Si(111), Jpn. J. Appl. Phys. 52 (2013), 08JE05, https:// doi.org/10.7567/JJAP.52.08JE05. [24] M. Agrawal, K. Radhakrishnan, N. Dharmarasu, S.S. Pramana, Effect of III/V ratio on the polarity of AlN and GaN layers grown in the metal rich growth regime on Si
Fig. 6. AFM micrographs for samples D1-D3 (left to right respectively) where C-doping is varied from 5:3 � 1019 ; 2:5 � 1020 and 8:3 � 1020 cm 3 respectively.
For samples D1-D3 growth was initiated with 2 nm of AlN followed by 200 nm of GaN:C with CBr4 foreline pressure of 20, 40 and 60 mTorr, corresponding to a C-doping concentration of approximately 5:3� 1019 ; 2:5 � 1020 and 8:3 � 1020 cm 3 respectively. C-concentration as a function of CBr4 foreline pressure was determined from a multi-layered C-doped SIMS stack carried out in a previous experiment. AFM micro graphs showing the surface morphology of D1-3 are depicted in Fig. 6 below. The AFM micrographs in Fig. 6 reveal that the surface pits return for high C-concentrations. Sample D1, where C-doping was 5:3� 1019 cm 3 , exhibited almost no pits however it can be seen that in certain regions, particularly where two adjacent step meanders meet, there exists de pressions which could potentially lead to the formation of pits had the film continued to grow. From Fig. 6 we see that these pits return for some C-doping concentration between 5:3 � 1019 2:5 � 1020 cm 3 . The results of the C-doping growth series provide further evidence suggesting impurities play a role in the generation of surface-pits. However, surface pits on N-polar GaN films can also be generated by other factors including surface defects, dislocations and slow-moving step meanders. To the best of our knowledge it is unknown as to whether this same effect is seen in Ga-polar films utilizing such high Cconcentrations. 7. Conclusion In this work we have demonstrated that smooth, pit-free N-polar GaN can be grown on non-vicinal GaN substrates by PAMBE using the AlN initiation layer or AIL. Considering the higher bond energy Al has with other species with respect to Ga, combined with AlN’s long history of being a reliable nucleation layer for heteroepitaxially grown GaN, it is likely the AIL grows over defects and impurities on the substrate surface thereby creating a smooth surface for which N-polar GaN can be grown on. Thus the AIL effectively decouples the epitaxially grown GaN from the substrate surface. It was shown that thicker AILs, up to 8 nm, further improve film quality and surface morphology while still remaining strained to the GaN. By varying C-doping for films grown with the AIL it was shown that pits can again be generated on the surface for C-doping concentration between 5:3 � 1019 2:5 � 1020 cm 3 . This further supported the hypothesis that impurities contribute to pit formation in N-polar GaN films grown by MBE. The findings in this study pave the way for the MBE growth of high performance N-polar HEMT structures on bulk GaN. Acknowledgements The authors gratefully acknowledge the funding from the Office of Naval Research (Dr. P. Maki) and the DARPA DREaM program. SIMS measurements were carried out by Dr. Tom Mates of the California NanoSystems Institute (CNSI). The authors would also like to thank Kelsey Jorgenson and Richard Cramer for their helpful discussions and advice regarding PAMBE of GaN.
5
C. Wurm et al.
Solid State Communications 305 (2020) 113763
(111) by plasma assisted molecular beam epitaxy, Jpn. J. Appl. Phys. 54 (2015), 065701, https://doi.org/10.7567/JJAP.54.065701. [25] M. Stutzmann, O. Ambacher, M. Eickhoff, U. Karrer, A. Lima Pimenta, R. Neuberger, J. Schalwig, R. Dimitrov, P.J. Schuck, R.D. Grober, Playing with polarity, Phys. Status Solidi B 228 (2001) 505–512, https://doi.org/10.1002/ 1521-3951(200111)228:2<505::AID-PSSB505>3.0.CO;2-U. [26] A. Adikimenakis, S.-L. Sahonta, G.P. Dimitrakopulos, J. Domagala, Th. Kehagias, Ph. Komninou, E. Iliopoulos, A. Georgakilas, Effect of AlN interlayers in the structure of GaN-on-Si grown by plasma-assisted MBE, J. Cryst. Growth 311 (2009) 2010–2015, https://doi.org/10.1016/j.jcrysgro.2008.10.085.
[27] C. Poblenz, P. Waltereit, S. Rajan, S. Heikman, U.K. Mishra, J.S. Speck, Effect of carbon doping on buffer leakage in AlGaN/GaN high electron mobility transistors, J. Vac. Sci. Technol. B Microelectron. Nanometer Struct. 22 (2004) 1145, https:// doi.org/10.1116/1.1752907. [28] J.L. Lyons, A. Janotti, C.G. Van de Walle, Carbon impurities and the yellow luminescence in GaN, Appl. Phys. Lett. 97 (2010), 152108, https://doi.org/ 10.1063/1.3492841. [29] J. Ibbetson, Electrical Characterization of Nonstoichiometric GaAs Grown at Low Temperature by Molecular Beam Epitaxy, Thesis, 1997. [30] S. Chowdhury, AlGaN/GaN CAVETs for High Power Switching Application, Thesis, 2010.
6