Plasma Surface Pretreatment Effects on Silicon Nitride Passivation of AlGaN/GaN HEMTs

Several plasma and wet-chemical surface treatments have been tested to determine their effectiveness in improving SiN passivation of the AlGaN surface in HEMT devices. Mitigation of RF dispersion was evidenced by dramatic improvements in pulsed IV data after various plasma surface treatments and SiN passivation were applied to gate-level processed HEMTs. To examine surface chemistry as a result of these treatments, XPS was used to obtain atomic concentrations and bonding information. INTRODUCTION Although the development of high-power HEMT devices based on AlGaN/GaN heterostructures has progressed rapidly over the last decade, commercialization of circuits based on these devices has been delayed in part by concerns of reliability. In particular, these devices are prone to rapid and long-term degradation due to the presence of defectrelated charge traps. The relatively slow charging and discharging of these defect states, with time constants in the microseconds range, cause HEMTs to experience RF dispersion and current collapse.[1-4] Mitigation of RF dispersion has been demonstrated in the literature by passivating the AlGaN surface using a variety of dielectric films with the most commonly reported one being SiN deposited by PECVD. This work reports on our investigation of fundamental aspects of surface modification and passivation relating to manufacturability of reliable HEMT devices. We have found that the pretreatment of the AlGaN surface immediately prior to plasma deposition of SiN is a critically important step in achieving passivation efficacy. The effects of plasma and wet chemical treatments prior to passivation have been evaluated with regard to the resulting electrical characteristics of the HEMTs and the chemical properties of the AlGaN surface. EXPERIMENTAL The HEMT heterostructures used for this study were grown on sapphire substrates by MOCVD, and were provided by Emcore Corporation. The device epilayers consisted of a 1.7 μm thick undoped GaN layer followed by a 23 nm Al0.32GaN layer. Ohmic metallization was attained by e-beam evaporation of Ti/Al/Ni/Au (15/100/50/50 nm) followed by 60 seconds of rapid thermal annealing at 875 oC. The resulting ohmic contact resistance was 0.4±0.1 Ω mm. HEMT mesa isolation was achieved through Cl2 plasma in an ICP reactor and was followed by gate metallization of Ni/Au (50/100 nm) by e-beam evaporation. Gate contacts were defined by contact lithography and had dimensions of 1 μm × 100 μm. Unpassivated devices had average threshold voltages of Vth = -4.2±0.1 V and average maximum transconductance of gm,max = 132±60 mS/mm. After gate contact definition, several different prepassivation surface treatments were utilized followed by encapsulation in SiN. These treatments included various SF6, O2, NH3, N2 plasmas as well as wet treatments using either an SC1 clean (hot aqueous NH4OH/H2O2) or 10:1 BOE (aqueous NH4F/HF). Two different plasma systems were used to administer the prepassivation surface treatments. Plasmas involving SF6 and O2 were carried out in a magnetically enhanced ICP reactor that had backside cooling of the substrate. Chamber pressure during the treatment was held at 8 mTorr, while RF power was 400 W and 40 W to the inductive coil and substrate electrode, respectively. With these conditions, induced DC bias levels reached -100 V to -130 V. Plasmas that included NH3 and N2 gaseous precursors were applied to the sample directly before passivation, in the same PECVD chamber, utilizing a parallel plate, shower head geometry. The pressure was held at 2.7 Torr for these plasma treatments, as well as for the SiN deposition. In the case of the wet-chemically treated samples and the SF6/O2 plasma-treated samples, SiN encapsulation was performed within a few hours of pretreatment. RESULTS AND DISCUSSION Among the various surface pretreatments and SiN deposition parameters evaluated, we have found that the measured DC and pulsed IV electrical characteristics can vary greatly. Figure 1 shows the change in HEMT channel sheet resistance that resulted from 30 seconds (unless otherwise denoted) of plasma pretreatment followed by approximately 800 Å of SiN encapsulation. The refractive index of the SiN film used in these experiments was 305 CS MANTECH Conference, May 14-17, 2007, Austin, Texas, USA Figure 1 – Change in HEMT channel sheet resistance measured by CTLM after various plasma and wet chemical pretreatments and SiN (n = 1.99) encapsulation. -20.0% 0.0% 20.0% 40.0% 60.0% 80.0% 100.0% 120.0% 140.0%