Reducing Thermal Resistance of AlGaN/GaN Electronic ... - IEEE Xplore

193 downloads 0 Views 235KB Size Report
Richard Lossy, Reza Pazirandeh, Frank Brunner, Joachim Würfl, and Martin Kuball ... K. P. Hilton, J. O. Maclean, D. J. Wallis, M. J. Uren, and T. Martin are with.
IEEE ELECTRON DEVICE LETTERS, VOL. 30, NO. 2, FEBRUARY 2009

103

Reducing Thermal Resistance of AlGaN/GaN Electronic Devices Using Novel Nucleation Layers Gernot J. Riedel, James W. Pomeroy, Keith P. Hilton, Jessica O. Maclean, David J. Wallis, Michael J. Uren, Trevor Martin, Urban Forsberg, Anders Lundskog, Anelia Kakanakova-Georgieva, Galia Pozina, Erik Janzén, Richard Lossy, Reza Pazirandeh, Frank Brunner, Joachim Würfl, and Martin Kuball

Abstract—Currently, up to 50% of the channel temperature in AlGaN/GaN electronic devices is due to the thermal-boundary resistance (TBR) associated with the nucleation layer (NL) needed between GaN and SiC substrates for high-quality heteroepitaxy. Using 3-D time-resolved Raman thermography, it is shown that modifying the NL used for GaN on SiC epitaxy from the metal– organic chemical vapor deposition (MOCVD)-grown standard AlN-NL to a hot-wall MOCVD-grown AlN-NL reduces NL TBR by 25%, resulting in ∼10% reduction of the operating temperature of AlGaN/GaN HEMTs. Considering the exponential relationship between device lifetime and temperature, lower TBR NLs open new opportunities for improving the reliability of AlGaN/ GaN devices. Index Terms—CVD, epitaxial layers, FETs, gallium compounds, MODFETs, resistance heating.

H

IGH ELECTRON mobility transistors (HEMTs) that are GaN-based offer excellent high frequency and power handling capabilities and will play a central role in future radar and communication applications [1], [2]. Outstanding device performance has been reported; however, long-term reliability still remains a major concern despite recent advances [3]–[5]. High operating temperatures associated with very localized Joule heating decrease the mean time to failure. This makes device thermal management a crucial aspect in tackling device reliability challenges. High-thermal-conductivity substrates, typically SiC, are used for AlGaN/GaN power devices to efficiently extract the generated heat and to minimize device temperature rise. However, to achieve the high-quality heteroepitaxial growth of GaN on SiC, interlayers must be introduced between the SiC and the GaN, normally an AlN nucleation Manuscript received October 24, 2008; revised November 13, 2008 and November 18, 2008. First published December 22, 2008; current version published January 28, 2009. The work in Bristol was supported by EPSRC, at QinetiQ Ltd. by the ES Domain of the U.K. Ministry of Defence and the KORRIGAN Program, at Ferdinand Braun Institute by ESA, and at Linköping University by the Swedish Foundation for Strategic Research and the KORRIGAN Program. The review of this letter was arranged by Editor J. A. del Alamo. G. J. Riedel, J. W. Pomeroy, and M. Kuball are with the H. H. Wills Physics Laboratory, University of Bristol, BS8 1TL Bristol, U.K. (e-mail: [email protected]; [email protected]). K. P. Hilton, J. O. Maclean, D. J. Wallis, M. J. Uren, and T. Martin are with QinetiQ Ltd., WR14 3PS Malvern, U.K. U. Forsberg, A. Lundskog, A. Kakanakova-Georgieva, G. Pozina, and E. Janzén are with the Department of Physics, Chemistry and Biology, Linköping University, 582 46 Linköping, Sweden. R. Lossy, R. Pazirandeh, F. Brunner, and J. Würfl are with the FerdinandBraun-Institut für Höchstfrequenztechnik, 12489 Berlin, Germany. Color versions of one or more of the figures in this letter are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/LED.2008.2010340

layer (NL), to compensate for lattice mismatch [6]. The thermal resistance of a device, i.e., temperature rise per input power density, is therefore not only determined by the intrinsic thermal resistance of the materials used in a device (AlGaN, GaN, SiC) but also by thermal-boundary resistance (TBR) of the NL that restricts heat transport from the GaN into the SiC substrate. The NL TBR has been found to be responsible for an additional 30%–50% increase in device channel temperature in today’s AlGaN/GaN on SiC devices [7]. Therefore, there is significant potential for improving AlGaN/GaN HFET thermal resistance by exploring novel NLs. In this letter, we demonstrate that, by changing the NL, from a metal–organic chemical vapor deposition (MOCVD)-grown standard AlN-NL to a hot-wall MOCVD-grown AlN-NL, a reduction in AlGaN/GaN device channel temperature can be achieved, opening new opportunities for improving the reliability of AlGaN/GaN electronic devices. Raman thermography was used to determine the NL TBR in AlGaN/GaN electronic devices containing standard (i) 30-nm-thick and (ii) 40-nm-thick AlN-NLs deposited at 1020 ◦ C–1050 ◦ C by MOCVD at Ferdinand Braun Institute and QinetiQ Ltd., and (iii) a 80-nm-thick AlN-NL grown at 1100 ◦ C using hot-wall MOCVD at Linköping, denoted in the following as 30 nmstd , 40 nmstd , and 80 nmhot−wall , respectively. The MOCVD-grown AlN-NL, which is the current standard NL for GaN on SiC epitaxy, typically contains very small crystallites and a very high density of extended defects [8]. In contrast, a novel hot-wall MOCVD results in an improved microstructure of the AlN-NL, being a more monocrystalline AlN layer with a dislocation density of 1010 cm−2 [9]. This is likely due to the reduced temperature gradients in the growth zone, different growth temperatures, efficient precursor cracking, and specific substrate surface preparation in the hot-wall MOCVD used. The relative contribution of these different factors, as well as the detailed NL microstructure, still needs further investigation. More details of this growth technique can be found in [9]–[11]. The subsequent device structures grown for 30 nmstd and 40 nmstd consisted of 25–30-nm undoped AlGaN on top of 2.35-μm undoped insulating GaN on 400-μm-thick c-plane 4H-SiC substrates. The 80 nmhot−wall structure was a 25-nm undoped AlGaN on a 1.9-μm undoped insulating GaN on a 390-μm-thick c-plane 6H-SiC substrate. Raman thermography is based on measuring changes in the temperature-dependent material phonon frequencies [12]–[14]. It was used to measure device and substrate temperature with submicrometer spatial and 10–30-ns time resolution [15], [16].

0741-3106/$25.00 © 2009 IEEE

104

Fig. 1. Temperature rise in the center of an ungated AlGaN/GaN device, grown on SiC with AlN-NL 80 nmhot−wall , as a function of depth from the device surface into the SiC substrate. The device was 125 μm wide with 20-μm contact separation and was dc operated at 5.3 W (50 V). Thermal simulation data with and without TBR are overlaid. Note that the discrepancies between simulation and experimental data near the interface are related to the finite depth resolution of confocal Raman microscopy [7].

Here, the GaN A1 (LO) and SiC E2 (PO) phonons were used. Ungated devices with standard ohmic contacts separated by 4–20 μm were operated under dc bias and with square electrical pulses of 300-ns duration during the measurements. The rise and fall time for current/voltage pulses at the devices was ∼10 ns. The Raman signal was excited by laser pulses of 30-ns duration generated by the acousto–optic modulation of a 532-nm diode-pumped frequency-doubled Nd-YAG laser. A 50 × 0.55-NA objective lens was used to focus the laser onto the device and to collect the Raman backscattered light, achieving a lateral spatial resolution of 0.5–0.7 μm. In the measurement, the device temperature is averaged through the GaN layer thickness, while the SiC substrate temperature is averaged over a depth of 4–5 μm below the GaN/SiC interface, given by the depth resolution of confocal microscopy. The temperature accuracy achieved was better than ±5 ◦ C for the device and ±2 ◦ C for the substrate. All wafers were attached to a copper heat sink, and the wafer backplate temperature was controlled by a Peltier element. Three-dimensional finite difference thermal simulations of the device structures were performed for comparison to the experimental data. The temperature-dependent thermal parameters used were similar to those of [7] and [15]. A uniform surface heat load between the contacts was used to simulate the ungated devices. A temperature–depth scan from the AlGaN/GaN device into the SiC substrate is shown in Fig. 1. Peak temperature is located near the AlGaN/GaN interface where heat is generated [13], and temperature decreases from the GaN, through the SiC, and toward the heat sink. The apparent temperature discontinuity between the AlGaN/GaN device and the SiC substrate in Fig. 1 highlights the presence of a TBR at the GaN/SiC interface restricting heat transfer from the GaN into the SiC. Fig. 2 shows the transient temperature of a similar device. Although the GaN device temperature rises rapidly by ∼100 ◦ C within the first 300 ns after turning on the device, there is very little temperature variation (6–8 ◦ C) in the SiC substrate close to the GaN/SiC interface. Additionally the SiC temperature rise

IEEE ELECTRON DEVICE LETTERS, VOL. 30, NO. 2, FEBRUARY 2009

Fig. 2. Temperature of an AlGaN/GaN device with AlN-NL 80 nmhot−wall and of its SiC substrate (inset) as function of time, recorded in the center of a 125-μm-wide ungated device with 5-μm contact separation. The device was operated at 4.4 W, with 35-V source–drain square electrical pulses of 300-ns length at a backplate temperature of 125 ◦ C. Temperature data obtained from finite difference simulation, averaged over the GaN layer thickness, are overlaid.

is not instantaneous, but is delayed by ∼30 ns with respect to the device temperature. Both, the small SiC temperature amplitude and the presence of a time delay between GaN and SiC temperatures are caused, in a sizable part, by the presence of a TBR at the GaN/SiC interface. By fitting thermal simulation data to the dc and time-resolved experimental thermal results (Figs. 1 and 2), the TBR can be quantified for the different AlN-NLs investigated. A number of simulation input parameters must be accurately known to determine the TBR. The SiC substrate thermal conductivity is the dominant parameter for the device temperature [17]. As the temperature of the SiC as a function of depth (Fig. 1) is almost exclusively dependent on the substrate thermal conductivity, this parameter was determined by fitting thermal simulation to the experimental data, as in [7]. Small variations in the GaN thermal parameters have little influence on the device temperature and its dynamics [12], e.g., a 10% increase in GaN thermal conductivity would result in less than 2% decreased device temperature. Therefore, GaN thermal parameters were assumed to be uniform across the layer and identical from wafer to wafer, taken from [7], as were GaN and SiC heat capacitance [16]. The AlN-NL TBR is the only remaining free parameter used in fitting the thermal simulation to the experimental transient thermal data in Fig. 2. The device and SiC substrate temperatures rise, and the delayed thermal responses of the SiC substrate are reproduced well in the simulation. Simulation results obtained with 30% lower and higher TBRs than the best fit value are also shown in Fig. 2, enabling us to estimate the accuracy of the TBR determined here to be better than ±15%. We note that TBR can also be obtained from the GaN/SiC temperature discontinuity at the interface (Fig. 1) as demonstrated in [7]; however, this requires assumptions about the microscope confocal depth resolution to be made, which are not needed using the approach taken here. Fig. 3 shows the TBR derived for the different NLs, 30 nmstd , 40 nmstd , and 80 nmhot−wall , as a function of

RIEDEL et al.: REDUCING THERMAL RESISTANCE OF AlGaN/GaN ELECTRONIC DEVICES USING NLs

Fig. 3. TBR of different AlN NLs for AlGaN/GaN on SiC epitaxy as a function of GaN/SiC interface temperature: MOCVD-grown standard 30-nmthick AlN-NL (30 nmstd ), 40-nm-thick AlN-NL (40 nmstd ), and hot-wall MOCVD-grown 80-nm AlN-NL (80 nmhot−wall ). The resulting rise in the AlGaN/GaN HEMT channel temperature during device operation obtained using thermal finite difference simulation, with respect to a device without any TBR, is also given. Interface temperature is defined as the average of GaN and SiC temperature at the interface. Insets show TEM micrographs of the NLs.

GaN/SiC interface temperature. The TBR of all AlN-NLs measured increases with rising temperature due to a decreasing thermal conductivity of the AlN caused by increasing phonon scattering at higher temperatures [18]. The TBR of 80 nmhot−wall is 25% lower over the whole temperature range than 30 nmstd and 40 nmstd , despite the NL layer being twice as thick as the standard NLs. In contrast to the commonly used MOCVD-grown AlN-NL, the 80 nmhot−wall NL is more monocrystalline (Fig. 3), resulting in less defect, grain and domain boundary phonon scattering, and consequently, in a higher thermal conductivity of the NL. Fig. 3 clearly shows that modifying the NL from an MOCVD-grown standard AlN-NL to a more monocrystalline AlN-NL can provide significant thermal benefits for AlGaN/GaN electronic devices. The 25% reduction in TBR for 80 nmhot−wall results in a 10% reduction in peak channel temperature rise in an AlGaN/GaN HEMT (Fig. 3), as well as in a reduction in the temperature of the whole device active area, providing significant benefits for device reliability. The similar TBR measured here for the standard MOCVD-grown AlN-NLs from separate laboratories suggests that these NLs are of similar microstructure and that the corresponding TBR determined can be considered as representative for the current standard MOCVD AlN-NLs. We note that TBR of 30 nmstd is possibly slightly smaller than that of 40 nmstd (Fig. 3), which one might expect as heat has to cross a thinner NL layer. The 80 nmhot−wall TBR is still responsible for 20%–40% of AlGaN/GaN device temperature rise with respect to the absence of TBR. Therefore, additional reductions in the AlGaN/GaN device temperature are possible, beyond what has been demonstrated here. Reducing the 80 nmhot−wall thickness, together with further improvements in the microstructure (dislocation, domain boundary density, and interface regions) of the AlN-NL, could open further opportunities for improving TBR

105

and therefore improve the device reliability. Corresponding improvements may also be possible using standard MOCVD or other growth techniques by modifying surface preparation and growth conditions or using different substrate types, including offcut angles. We note that for optimizing a NL not only a low TBR needs to be achieved but the NL also has to enable subsequent high-quality GaN epitaxy, as has been achieved here for the hot-wall MOCVD AlN-NL growth method investigated [10], [11]. In conclusion, the influence of AlN-NLs in GaN/SiC on AlGaN/GaN device temperature was studied using the timeresolved Raman thermography. As device failure rates increase exponentially with temperature, even a small reduction in the device temperature can have a large impact on device reliability. Improving the microstructure of the AlN-NL toward monocrystalline AlN by using novel growth approaches rather than using an MOCVD-grown standard AlN-NL was shown to provide a reduction of 25% in the TBR at the GaN/SiC interface, offering new opportunities for improving future AlGaN/GaN device reliability. ACKNOWLEDGMENT The authors would like to thank A. Sarua (Bristol) for fruitful discussions. R EFERENCES [1] S. Nuttinck, E. Gebara, J. Laskar, and M. Harris, “Development of GaN wide bandgap technology for microwave power applications,” IEEE Microw. Mag., vol. 3, no. 1, pp. 80–87, Mar. 2002. [2] M. Kuzuhara, H. Miyamoto, Y. Ando, T. Inoue, Y. Okamoto, and T. Nakayama, “High-voltage rf operation of AlGaN/GaN heterojunction FETs,” Phys. Stat. Sol. A, Appl. Res., vol. 200, no. 1, pp. 161–167, Nov. 2003. [3] Y. F. Wu, A. Saxler, M. Moore, R. P. Smith, S. Sheppard, P. M. Chavarkar, T. Wisleder, U. K. Mishra, and P. Parikh, “30-W/mm GaN HEMTs by field plate optimization,” IEEE Electron Device Lett., vol. 25, no. 3, pp. 117–119, Mar. 2004. [4] K. Joshin and T. Kikkawa, “High-power and high-efficiency GaN HEMT amplifiers,” in Proc. IEEE Radio Wireless Symp., 2008, pp. 65–68. [5] G. Meneghesso, G. Verzellesi, F. Danesin, F. Rampazzo, F. Zanon, A. Tazzoli, M. Meneghini, and E. Zanoni, “Reliability of GaN highelectron-mobility transistors: State of the art and perspectives,” IEEE Trans. Device Mater. Rel., vol. 8, no. 2, pp. 332–343, Jun. 2008. [6] N. D. Bassim, M. E. Twigg, C. R. Eddy, J. C. Culbertson, M. A. Mastro, R. L. Henry, R. T. Holm, P. G. Neudeck, A. J. Trunek, and J. A. Powell, “Lowered dislocation densities in uniform GaN layers grown on step-free (0001) 4H-SiC mesa surfaces,” Appl. Phys. Lett., vol. 86, no. 2, p. 021 902, Jan. 2005. [7] A. Sarua, H. Ji, K. P. Hilton, D. J. Wallis, M. J. Uren, T. Martin, and M. Kuball, “Thermal boundary resistance between GaN and substrate in AlGaN/GaN electronic devices,” IEEE Trans. Electron Devices, vol. 54, no. 12, pp. 3152–3158, Dec. 2007. [8] D. D. Koleske, R. L. Henry, M. E. Twigg, J. C. Culbertson, S. C. Binari, A. E. Wickenden, and M. Fatemi, “Influence of AlN nucleation layer temperature on GaN electronic properties grown on SiC,” Appl. Phys. Lett., vol. 80, no. 23, pp. 4372–4374, Jun. 2002. [9] A. Kakanakova-Georgieva, P. O. A. Persson, A. Kasic, L. Hultman, and E. Janzen, “Superior material properties of AlN on vicinal 4H-SiC,” J. Appl. Phys., vol. 100, no. 3, p. 036 105, Aug. 2006. [10] U. Forsberg, A. Lundskog, A. Kakanakova-Georgieva, R. Chiechonski, and E. Janzen, “Improved hot-wall MOCVD growth of highly uniform AlGaN/GaN HEMT structures,” J. Cryst. Growth, 2008, to be published. [11] A. Kakanakova-Georgieva, U. Forsberg, I. G. Ivanov, and E. Janzen, “Uniform hot-wall MOCVD epitaxial growth of 2 inch AlGaN/GaN HEMT structures,” J. Cryst. Growth, vol. 300, no. 1, pp. 100–103, Mar. 2007.

106

[12] M. Kuball, J. M. Hayes, M. J. Uren, T. Martin, J. C. H. Birbeck, R. S. Balmer, and B. T. Hughes, “Measurement of temperature in active high-power AlGaN/GaN HFETs using Raman spectroscopy,” IEEE Electron Device Lett., vol. 23, no. 1, pp. 7–9, Jan. 2002. [13] S. Rajasingam, J. W. Pomeroy, M. Kuball, M. J. Uren, T. Martin, D. C. Herbert, K. P. Hilton, and R. S. Balmer, “Micro-Raman temperature measurements for electric field assessment in active AlGaNGaN HFETs,” IEEE Electron Device Lett., vol. 25, no. 7, pp. 456–458, Jul. 2004. [14] A. Sarua, H. F. Ji, M. Kuball, M. J. Uren, T. Martin, K. P. Hilton, and R. S. Balmer, “Integrated micro-Raman/Infrared thermography probe for monitoring of self-heating in AlGaN/GaN transistor structures,” IEEE Trans. Electron Devices, vol. 53, no. 10, pp. 2438–2447, Oct. 2006. [15] G. J. Riedel, J. W. Pomeroy, K. P. Hilton, J. O. Maclean, D. J. Wallis, M. J. Uren, T. Martin, and M. Kuball, “Nanosecond timescale thermal

IEEE ELECTRON DEVICE LETTERS, VOL. 30, NO. 2, FEBRUARY 2009

dynamics of AlGaN/GaN electronic devices,” IEEE Electron Device Lett., vol. 29, no. 5, pp. 416–418, May 2008. [16] M. Kuball, G. J. Riedel, J. W. Pomeroy, A. Sarua, M. J. Uren, T. Martin, K. P. Hilton, J. O. Maclean, and D. J. Wallis, “Time-resolved temperature measurement of AlGaN/GaN electronic devices using micro-Raman spectroscopy,” IEEE Electron Device Lett., vol. 28, no. 2, pp. 86–89, Feb. 2007. [17] J. W. Pomeroy, M. Kuball, D. J. Wallis, A. M. Keir, K. P. Hilton, R. S. Balmer, M. J. Uren, T. Martin, and P. J. Heard, “Thermal mapping of defects in AlGaN/GaN heterostructure field-effect transistors using micro-Raman spectroscopy,” Appl. Phys. Lett., vol. 87, no. 10, p. 103 508, Sep. 2005. [18] B. C. Daly, H. J. Maris, A. V. Nurmikko, M. Kuball, and J. Han, “Optical pump-and-probe measurement of the thermal conductivity of nitride thin films,” J. Appl. Phys., vol. 92, no. 7, pp. 3820–3824, Oct. 2002.