Solid-State Electronics xxx (2016) xxx–xxx
Contents lists available at ScienceDirect
Solid-State Electronics journal homepage: www.elsevier.com/locate/sse
TMAH-based wet surface pre-treatment for reduction of leakage current in AlGaN/GaN MIS-HEMTs Young Jun Yoon, Jae Hwa Seo, Min Su Cho, Hee-Sung Kang, Chul-Ho Won, In Man Kang ⇑, Jung-Hee Lee ⇑ School of Electronics Engineering, Kyungpook National University, Daegu, Republic of Korea
a r t i c l e
i n f o
Article history: Received 21 April 2016 Received in revised form 21 June 2016 Accepted 24 June 2016 Available online xxxx The review of this paper was arranged by Prof. E. Calleja Keywords: GaN MIS-HEMT Surface treatment Leakage current Breakdown voltage TMAH
a b s t r a c t The pre-passivation surface treatment process with tetramethylammonium hydroxide (TMAH)-based wet solution was proposed for the minimization of the leakage current (Ileak) in AlGaN/GaN metalinsulator-semiconductor high electron mobility transistors (MIS-HEMTs). This process step contributes to the simultaneous decrease of the surface current (Isurf) in the active region of device and mesaisolated region by removing the surface states and traps related to nitrogen (N) vacancy, Ga-oxide, and dangling bonds. Using the surface treatment, the fabricated device achieves a lower off-state current (Ioff) of 10 12 A/mm, a higher on/off current ratio (Ion/Ioff) of 1011, a small subthreshold swing (SS) of 68.4 mV/dec. The reduced Ileak also improves breakdown voltage (BV). In addition, the interface trap density (Dit) between the SiN layer and the AlGaN surface was extracted to evaluate the quality of the SiN/GaN interface, which showed that the treatment decreases the Dit with reduction of the surface defects. Ó 2016 Published by Elsevier Ltd.
1. Introduction AlGaN/GaN-based metal-insulator-semiconductor highelectron mobility transistors (MIS-HEMTs) have been studied for high frequency and high power applications, owing to outstanding material properties of III-nitrides such as a high electron mobility, a high electron saturation velocity, and a high breakdown electric field [1–5]. However, the epitaxial AlGaN/GaN layers grown for the device fabrication suffer from the existence of many states and traps at surface as well as in buffer layer, related to dislocations, nitrogen (N) vacancy, or oxygen (O) incorporation. Especially, the trapping effects through the surface states leads to a high device leakage current (Ileak) which results in high off-state current (Ioff), the poor subthreshold swing (SS), low breakdown voltage (BV), high noise, and low power efficiency [6–8]. To suppress the Ileak, the process step for the reduction of surface current (Isurf) is required. Several groups have reported a surface passivation with SiN layer for reduction of the Isurf and surface states [9,10]. However, since this method occasionally suffers from difficulty in obtaining a high interface quality between SiN and GaN layer [11], an appropriate surface treatment prior to SiN passivation is important in improv⇑ Corresponding authors. E-mail addresses:
[email protected] (I.M. Kang),
[email protected] (J.-H. Lee).
ing the interface quality [12]. Further, the Isurf also flows through the leakage pass created by surface states and traps, which are related to plasma etching for the device isolation [13,14], and this current can be significantly increased if the plasma damaged surface is not properly recovered by an appropriate treatment. Reduction of the leakage current due to plasma etching required for the device isolation is therefore very important especially for a high voltage application. Consequently, to minimize the Ileak of the device, it is essential to eliminate the origins for the Isurf in mesaetched region as well as active region of the device. It is known that the wet etching based on tetramethylammonium hydroxide (TMAH) solution removes native Ga-oxide from the GaN surface because the Ga-oxide is dissolved in hydroxide solutions and the surface layer becomes N-terminated as hydroxide solutions such as KOH and NaOH do [15,16]. Since the OH ions in hydroxide solutions react with Ga atoms, the Ga-terminated surface with many N vacancy-related dangling bonds is converted into the Nterminated surface by removing the N vacancy [17]. It was also reported that the TMAH treatment on the plasma etched GaN surface effectively removes the plasma damage from the surface, but also smoothens the GaN surface [18–20]. In this work, the TMAH-based surface pre-treatment prior to SiN passivation was introduced, which simultaneously reduces the Isurf through both the active region and the mesa-isolated region. The effect of the treatment on the reduction of the Isurf
http://dx.doi.org/10.1016/j.sse.2016.06.009 0038-1101/Ó 2016 Published by Elsevier Ltd.
Please cite this article in press as: Yoon YJ et al. TMAH-based wet surface pre-treatment for reduction of leakage current in AlGaN/GaN MIS-HEMTs. Solid State Electron (2016), http://dx.doi.org/10.1016/j.sse.2016.06.009
2
Y.J. Yoon et al. / Solid-State Electronics xxx (2016) xxx–xxx
was verified by measuring the Isurf according to the treatment time. The interface quality between the TMAH-treated GaN surface and the SiN gate dielectric layer was also evaluated by extracting the interface trap density (Dit). 2. Device fabrication The AlGaN/GaN heterostructure was grown by metal–organic chemical vapor deposition (MOCVD) on c-plane sapphire substrates. The structure consisted of 2 lm-thick GaN buffer layer, 60 nm-thick undoped GaN channel layer, 22 nm-thick AlGaN layer, and 2 nm-thick GaN cap layer. Two-dimensional electron gas density (2DEG) of 9.7 1012 cm 2 and electron mobility of 1520 cm2/ Vs were estimated by Hall measurement at room temperature (RT), respectively. The schematic cross section of the fabricated MIS-HEMT is shown in Fig. 1(a). The gate length (LG) and the gate-to-source distance (LGS) are 2 and 5 lm, respectively. The gate-to-drain distance (LGD) is designed to be 5, 10, and 20 lm. The fabrication process started with mesa etching using Cl2/BCl3 plasma-based inductively coupled plasma (ICP) for device isolation. Etching depth for device mesa isolation was about 250 nm. The wet surface treatment was then carried out using TMAH solution (5% concentration) at 90 °C for 3 min prior to SiN deposition. It is noticed that the TMAH treatment was applied to the surface of the device active region and the mesa-etched region for the isolation, as shown in Fig. 1(b) and (c), respectively. In the active region, the treatment removes the native Ga-oxide formed at the GaN surface as well as the N vacancy-related dangling bond to change the surface to the N-terminated surface. Same treatment effect can be also achieved at the mesa-etched surface. Furthermore, the TMAH also smooths the rough GaN surface due to the plasma etching to remove the plasma damage from the etched surface as discussed above. After the treatment, 20 nm-thick SiN layer as a gate dielectric layer was deposited using plasma-enhanced chemical vapor
Fig. 2. (a) Output and (b) ID and IG characteristics at VDS of 7 V of the fabricated MISHEMTs with and without the TMAH-based surface treatment.
Fig. 1. (a) Schematic cross section of the AlGaN/GaN heterojunction-based MIS-HFET. (b) Schematic of the atomic arrangement at SiN/GaN interface and (c) SEM images of the etched surface without and with the TMAH-based surface treatment.
Please cite this article in press as: Yoon YJ et al. TMAH-based wet surface pre-treatment for reduction of leakage current in AlGaN/GaN MIS-HEMTs. Solid State Electron (2016), http://dx.doi.org/10.1016/j.sse.2016.06.009
Y.J. Yoon et al. / Solid-State Electronics xxx (2016) xxx–xxx
deposition (PECVD) at 370 °C. The SiN layer also plays a role of passivating the mesa-etched surface for isolation. To form ohmic contact in source and drain region, Au/Ni/Al/Ti/Si metal layer was deposited and annealed at 800 °C for 30 s in a N2 ambient. Contact resistances for samples without and with the TMAH treatment were measured to be 0.66 and 0.72 O cm using transfer length method (TLM), respectively. Finally, Ni/Al/Ni metal layer for the gate and pad was deposited by electron beam evaporator. For comparison, the reference device without the TMAH treatment was also fabricated with same process steps. 3. Results and discussion Fig. 2(a) shows the output characteristics of the fabricated MISHEMTs with and without the TMAH treatment. The device with the treatment exhibited maximum drain current (ID,max) of 425 mA/ mm at gate voltage (VGS) of 0 V and a specific on-resistance (Ron) of 6.13 mO cm2, and similar output characteristics were also obtained from the reference device without the treatment. As shown in drain current (ID) characteristics of Fig. 2(b), however, the device with the treatment exhibited much improved off-state performances; Ioff of 1.9 10 12 A/mm approximately four orders lower than that of the reference device, high Ion/Ioff ratio of 1.67 1011, and very low SS of 68.4 mV/dec. The gate current (IG) of the device with the treatment was also slightly decreased compared to that of the reference device. These clearly indicate that the treatment and subsequent deposition of the SiN layer well passivates the GaN surface with effectively reducing the surface states and traps. In order to further investigate the effect of the TMAH treatment on Isurf, we carried out the measurement for Isurf in active (Isurf,active) and isolation region (Isurf,isolation) with varying the treatment time, as shown in Fig. 3. Isurf components were evaluated by
Fig. 3. Measured (a) Isurf,active and (b) Isurf,isolation with different treatment time. Inset shows schematic structures for measurement of Isurf,active and Isurf,isolation.
3
using two different measurement structures simultaneously produced with device fabrication on the same wafer. The inset in Fig. 3(a) shows the double-gate structure used for measurement of Isurf,active. A distance between gate1 (G1) and gate2 (G2) is 5 lm. When a reverse bias is applied to the G2, the Isurf,active is occurred because electrons injected from G2 flow into the grounded-G1 though surface states and traps. The Isurf,active is unaffected by a bulk leakage current (Ibulk) flowed between G2 and source, since G1 is grounded. Therefore, the surface quality between G1 and G2 can be verified from the measured Isurf,active. The Isurf,active was greatly decreased by about 2 orders of magnitude with the treatment for 1 min and further decreased slightly as the treatment time increased to 3 min for the higher G2 voltage. This states that SiN passivation with sufficient pre-treatment time considerably reduces surface states and traps in the active region because the Isurf,active is due to the hopping conduction mechanism associated with states and traps [9,10]. The Isurf,isolation was evaluated by measuring the current between two ohmic contact electrodes connecting the mesa-etched region, as shown in inset of Fig. 3(b). A higher Isurf,isolation was observed for the structure without the treatment due to large amount of surface states and traps related to the dangling bonds on the etched surface [11]. In contrast, the structure with the treatment for 3 min exhibited much lower Isurf,isolation of 10 10 A/mm at voltage of 100 V, approximately 3 orders lower in magnitude than that of the structure without the treatment because the surface defects as well as plasma damage on the etched surface were effectively removed by the treatment [18–20] and the surface was well passivated with SiN deposition.
Fig. 4. (a) Off-state current characteristics of the fabricated MIS-HEMTs without and with the TMAH-based surface treatment at VGS of 9 V and (b) BVs in devices as a function of LGD.
Please cite this article in press as: Yoon YJ et al. TMAH-based wet surface pre-treatment for reduction of leakage current in AlGaN/GaN MIS-HEMTs. Solid State Electron (2016), http://dx.doi.org/10.1016/j.sse.2016.06.009
4
Y.J. Yoon et al. / Solid-State Electronics xxx (2016) xxx–xxx
CSiN is 323 nF/cm2. As shown in Fig. 5(b), the extracted Dit values for slow traps of the device with and without the treatment were 5.99 1011 and 1.53 1012 cm 2 eV 1 at a VGS,base in the up sweep of 1 V, respectively. Dit was increased for both devices as the VGS,base in the up sweep increases. This is believed to be due to the existence of the border and/or the oxide traps in the SiN gate dielectric layer. Further study is needed to understand the trapping behavior through these traps. 4. Conclusions The TMAH-based wet surface pre-treatment prior to SiN passivation has been employed to reduce the Ileak of AlGaN/GaN MISHEMTs. The pre-treatment significantly suppressed the Isurf not only in the active, but also in isolation regions by decreasing the density of surface state and trap. Using the pre-treatment for 3 min, the fabricated device exhibited very low SS of 68.4 mV/dec and maximum BV as high as 722 V. Furthermore, the reduction of Dit estimated from the pulsed-mode transfer characteristics for the device indicates that the TMAH treatment greatly improves the interface quality between SiN layer and the GaN surface. Acknowledgments This work was supported in part by the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (No. 2013-011522, 2011-0016222, 2016R1C1B2015979), and in part by Samsung Electronics Co. This work was supported by the Global Ph.D. Fellowship Program through the NRF funded by the MEST (2013H1A2A1034363). This work was supported by the BK21 Plus project funded by the Ministry of Education, Korea (21A20131600011). Fig. 5. (a) Pulse-mode transfer characteristics of the fabricated MIS-HEMTs without and with the TMAH-based surface treatment at VDS of 0.1 V and (b) extracted DVth and Dit in devices as a function of VGS,base.
Fig. 4 shows current characteristics at off-state and BV for both MIS-HEMTs fabricated with and without the TMAH treatment. BV was defined as drain voltage (VDS) value at ID of 1 lA/mm. ID of the device with the treatment is approximately one order lower in magnitude for the entire voltage ranges before breakdown compared to that of the reference device, which leads to much higher BV for the device with the treatment. BV of the reference device with LGD of 20 lm was 420 V, but BV increased to 722 V for the corresponding device with the treatment. BVs for devices with different LGD are shown in Fig. 4(b). It is noticed that BV of the device with the treatment seems to linearly increase with LGD due to the well passivated surface, while that of the reference device does not show the linear dependence because the hopping conduction through surface states and traps is significant. Fig. 5(a) shows pulse-mode transfer characteristics of the fabricated MIS-HEMTs with and without the TMAH treatment. The pulse width and period were 2 ms and 20 ms, respectively. Base VGS (VGS,base) in the up sweep is fixed at 10 V. Pulse-mode hysteresis measurements were employed for an accurate extraction of threshold voltage shift (DVth) related to slow traps with EC ET > 0.54 eV (sit > 2 ms), and low VDS of 0.1 V was applied to reduce the field-assisted detrapping [21,22]. In addition, DVth with VGS,base in the down sweep varying from 1 V to 3 V is also characterized using pulse-mode hysteresis measurements. The device with the treatment exhibits relatively smaller DVth of 0.3 V at a VGS,base in the up sweep of 1 V compared with the value of 0.8 V for the reference device. Dit is estimated using CSiNDVth/q, where
References [1] Mishra UK, Shen L, Kazior TE, Wu Y-F. Proc IEEE 2008;96:287. [2] Ikeda N, Niiyama Y, Kambayashi H, Sato Y, Nomura T, Kato S, et al. Proc IEEE 2010;98:1151. [3] Su M, Chen C, Rajan S. Semicond Sci Technol 2013;28:074012. [4] Ueda T, Ishida M, Tanaka T, Ueda D. Jpn J Appl Phys 2014;53:100214. [5] Kuzuhara M, Tokuda H. IEEE Trans Electron Dev 2015;62:405. [6] Deguchi T, Kikuchi T, Arai M, Yamasaki K, Egawa T. IEEE Electron Dev Lett 2012;33:1249. [7] Saito W, Kuraguchi M, Takada Y, Tsuda K, Omura I, Ogura T. IEEE Trans Electron Dev 2004;51:1913. [8] Saito W, Kuraguchi M, Takada Y, Takada Y, Tsuda K, Omura I, et al. IEEE Trans Electron Dev 2005;52:159. [9] Tan WS, Uren MJ, Houston PA, Green RT, Balmer RS, Martin T. IEEE Electron Dev Lett 2006;27:1. [10] Chen YH, Zhang K, Cao MY, Zhao SL, Zhang J, Ma XH, et al. Appl Phys Lett 2014;104:153509. [11] Liu ZH, Ng GI, Zhou H, Arulkumaran S, Maung YKT. Appl Phys Lett 2011;98:113506. [12] Arulkumaran S, Ng GI, Vicknesh S. IEEE Electron Dev Lett 2013;34:1364. [13] Lin Y-S, Lain Y-W, Hsu SSH. IEEE Electron Dev Lett 2010;31:102. [14] Xu Z, Wang J, Cai Y, Liu J, Jin C, Yang Z, et al. IEEE Electron Dev Lett 2014;35:1200. [15] Nowak G, Xia XH, Kelly JJ, Weyher JL, Porowski S. J Cryst Growth 2001;222:735. [16] Zhuang D, Edgar JH. Mater Sci Eng R 2005;48:1. [17] Li D, Sumiya M, Fuke S, Yang D, Que D, Suzuki Y, et al. J Appl Phys 2001;90:4219. [18] Kim K-W, Jung S-D, Kim D-S, Kang H-S, Im K-S, Oh J-J, et al. IEEE Electron Dev Lett 2011;32:1376. [19] Kim D-S, Im K-S, Kim K-W, Kang H-S, Kim D-K, Chang S-J, et al. Solid State Electron 2013;90:79. [20] Zhang Y, Sun M, Wong H-Y, Lin Y, Srivastava P, Hatem C, et al. IEEE Trans Electron Dev 2015;62:2155. [21] Yang S, Huang S, Schnee M, Zhao Q-T, Schubert J, Chen KJ. IEEE Trans Electron Dev 2013;60:3040. [22] Yang S, Liu S, Lu Y, Liu C, Chen KJ. IEEE Trans Electron Dev 2015;62:1870.
Please cite this article in press as: Yoon YJ et al. TMAH-based wet surface pre-treatment for reduction of leakage current in AlGaN/GaN MIS-HEMTs. Solid State Electron (2016), http://dx.doi.org/10.1016/j.sse.2016.06.009