0741-3106 (c) 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information. This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/LED.2017.2717539, IEEE Electron Device Letters Abstract— In this letter, traps related dispersion phenomenon of GaN/AlGaN/GaN HEMT grown on SiC substrate is investigated through low-frequency noise (LFN) measurements. Low-frequency drain noise measurements are performed over the frequency range of 20 Hz to 1 MHz and for varying chuck temperatures (Tchuck) ranging between 25ºC and 125ºC. This study demonstrates that two prominent deep-levels are present in the device-under-test (DUT) with apparent activation energies (Ea) of 0.51 and 0.57 eV, respectively. Moreover, signatures of deep level with Ea of 0.57 eV becomes visible only at higher operating temperatures, due to the fact that traps dispersion occurs at very low frequencies which is confirmed through TCAD physical simulations. Index Terms— Gallium-nitride (GaN), high electron mobility transistor (HEMT), deep levels, low-frequency noise measurement, TCAD physical simulations. I. INTRODUCTION he AlGaN/GaN HEMT devices have demonstrated their excellent capabilities to become promising candidates for RF high power and power switching applications [1]. Furthermore, the state-of-the-art for low noise figures have been already reported for these GaN HEMT devices [2]. In spite of their excellence performance, additional development related to reduction of charge-trapping dispersion effects is however essential for GaN based HEMTs in order to become a disruptive technology. The parasitic charges moving in and out of the deep-levels located on the surface or barrier or buffer layer of the device degrades its dynamic performance [3] and also undermines its long-term reliability [3]. The trapping time constants of these devices are in the range of several ns to a few seconds and hence limit the device performance significantly even at low frequencies. Moreover, the presence of defects or deep levels could contribute to low- This work was supported by DEFIS-RF project, France, under Contract ANR 13-CHN-0003. N. K. Subramani, J. Couvidat, A. A. Hajjar, J. -C. Nallatamby, M. Prigent and R. Quéré are with University of Limoges, CNRS, XLIM, UMR 7252, F- 19100 Brive, France (e-mail: nandhakumar2005@gmail.com; julien.couvidat@xlim.fr; ahmad.al-hajjar@xlim.fr; jean- christophe.nallatamby@unilim.fr; michel.prigent@xlim.fr; raymond.quere@xlim.fr). D. Floriot is with UMS, Villebon-sur-Yvette, F-91140, France (email:Didier.floriot@ums-gaas.com). frequency noise [4]. The origin of these noise sources are strongly related with the physical mechanisms governing the device operation. Among various noise sources, 1/f noise and Generation-Recombination (G-R) noise are the most important, however, these noise sources are reducible [5]. The 1/f noise could be originated due to material and device microscopic degrees of freedom interacting with quantum variables of nanodevices [6]. Moreover, it has been reported in [7], 1/f noise is generated on the surface or inside the volume, depending on the thickness and material of the grown sample. The GR noise arises due to the existence of deep-levels in the device that randomly capture and emit charge carriers, acting as generation-recombination centers [5]. A trapping- detrapping process results in fluctuations in the number of charge carriers, leads to change in device conductance [5]. Moreover, the G-R noise follows the Lorentzian frequency dependence and hence, it is characterized using a corner frequency and a characteristic time constant, related to the trapping process [8] [9]. Therefore, it is critical to understand the origin, their location and the physical mechanisms involved in these deep levels and also identifying the ways to minimize their influence on the device performance. There are several measurement techniques reported [3], [10]–[13] such as gate and drain lag transients, DLTS, frequency dependent transconductance dispersion, low frequency output admittance measurements and low-frequency noise characterization, in order to extract the trap characteristics, and understand their physical nature. Low-frequency noise characterization is undoubtedly an efficient tool for analyzing the deep levels in the device structure and to assess the device reliability [4]. LFN measurement has considerable advantages over conventional DLTS techniques, since the latter requires a large reverse bias voltage applied to device which might degrade the device performance. Moreover, the noise measurements at low temperatures can identify the shallow deep levels, which cannot be detected using DLTS [14]. Furthermore, the LFN measurement can be applied even to small area devices which may not be possible with capacitance DLTS measurement [3]. In this letter, the deep levels of a GaN/AlGaN/GaN HEMT transistor grown on SiC substrate have been characterized using low frequency noise measurement. The noise measurements are performed at different Tchuck ranging between 25ºC and 125ºC and this allow us to determine the Low Frequency Noise Characterization in GaN HEMTs: Investigation of deep levels and their physical properties Nandha Kumar Subramani, Julien Couvidat, Ahmad Al Hajjar, Jean-Christophe Nallatamby, Didier Floriot, Michel Prigent and Raymond Quéré, Fellow, IEEE T