Elsevier

Microelectronics Reliability

Volume 53, Issues 9–11, September–November 2013, Pages 1444-1449
Microelectronics Reliability

Statistics and localisation of vertical breakdown in AlGaN/GaN HEMTs on SiC and Si substrates for power applications

https://doi.org/10.1016/j.microrel.2013.07.117Get rights and content

Highlights

  • Probability distribution functions of GaN buffer breakdown voltage were determined.

  • A time dependent behaviour of vertical breakdown was observed.

  • Vertical breakdown paths were localised by backside infrared microscopy.

  • The strong localisation of the breakdown paths is due to current filamentation.

  • Roles of buffer defects and self-heating effects have been discussed.

Abstract

We analyse vertical breakdown signatures in normally-off and normally-on AlGaN/GaN HEMTs on Si and SiC substrate for power applications. The probability distribution function of the breakdown voltage VBD values shows mostly a bimodal distribution that is characteristic for substrate/epitaxy type and bias polarity. Different types of distribution functions are tested. The vertical breakdown is found to be a time dependent phenomenon and hypotheses for its initiation are discussed. Using backside infrared microscopy, we found that the breakdown occurs in localized spots, related to current filaments. Failure localisation under pulsed mode shows better spatial localisation compared to the DC conditions.

Introduction

AlGaN/GaN HEMTs are promising candidates for power switching applications [1], but there are challenges to improve their reliability [2], [3]. For low cost mass production, it is desirable to grow the III-N epitaxial layers on large area silicon substrates with reasonable quality [4]. For safe system operation under power-off condition, normally-off transistor operation is usually requested to reduce the requirements for gate driving circuitry. Normally-off AlGaN/GaN HEMTs with p-doped GaN gate, fabricated on SiC substrates, have demonstrated a breakdown voltage of 600 V [5]. It has been shown that the device breakdown voltage VBD of such power devices first linearly increases with the gate to drain distance LGD (i.e. lateral breakdown) and then becomes independent of LGD and saturates [6], [7]. In a floating substrate configuration, this saturation is caused by a vertical leakage current and vertical breakdown (VB) through the GaN buffer, between the Ohmic contact and the substrate [6], [7]. For sufficiently conductive substrates, the saturation value of VBD is thus a sum of breakdown voltages between the drain-and-substrate and substrate-and-source [6]. Furthermore, a large spreading in VBD values has been found, which indicates that the VB path can be due to current filaments related to extended defects [6], [8]. Based on area scaling of statistically distributed VB data, guidelines for optimal power device contact area have been determined [8]. The vertical leakage current mechanism across the GaN buffer was recently attributed to trap-mediated transport [9], [10] and the failure location was analysed from device top side [10]. In this paper, we report the statistical measures of VB data and localise the VB spots in normally-off and normally-on AlGaN/GaN HEMTs on n-type SiC and Si substrates. The cumulative distribution function (CDF) of VBD values is determined as a function of device area, substrate and heteroepitaxy. An analogy with statistics of time dependent dielectric breakdown (TDDB) in thin gate oxides [11] and off-state degradation of the AlGaN barrier [12], [13] is discussed. The position of VB paths is accurately obtained by backside infrared (IR) microscopy under DC and pulsed conditions. The mechanisms leading to the initiation of VB are discussed.

Section snippets

Experiments

The structures of the investigated devices are given in Fig. 1 and Table 1. Device A is a normally-on HEMT with threshold voltage Vth  −2.5 V (IDmax = 0.39 A/mm). Its structure is a Schottky-gate (Ir/Ti/Au) with a 25-nm-thick Al0.25Ga0.75N barrier, and a 2-μm-thick Al0.08Ga0.92N buffer grown on conductive Si substrate. The Ohmic contact is Ti/Al/Ni/Au/Ti/Pt. Device B is a normally-off device with Vth  0.5 V (IDmax = 0.35 A/mm). Its structure is a p-GaN gate with 14-nm-thick Al0.23Ga0.77N barrier

Statistical IV analysis

Typical examples of vertical leakage current IVs in voltage controlled mode between the top Ohmic contact and the substrate are given in Figs. 2a, b, 3a, b and 5a. In positive and negative biases, the leakage current has different voltage dependence and VB behaviour. Potential energy barriers at the 2DEG/GaN buffer interface (for negative drain), nucleation layer/substrate (for both bias polarities), but also bulk GaN traps can be involved in the transport. Resistive hopping-like transport,

Conclusions

Our results suggest that the statistical analysis of VB data can give specific information on the VB mechanism and could be used to characterise material quality and process technology. A time to breakdown behaviour has been found in VB. Together with the time to breakdown behaviour in the degradation of III-nitride barriers [12], [13], TDDB of GaN MIS HEMT dielectric stack [25] and TDDB of SiN passivation layer [26], this is yet another observation of time to breakdown behaviour in GaN based

Acknowledgments

This work was performed within EU Project HiPoSwitch (Grant agreement no. 287602) and supported by EU.

References (26)

  • D. Visalli et al.

    High breakdown voltage in AlGaN/GaN/AlGaN double heterostructures grown on 4 inch Si substrates

    Phys Stat Solidi C

    (2009)
  • R. Zhytnytska et al.

    Voltage power transistor design influence of metallic contact area on device breakdown

    WOCSDICE

    (2011)
  • C. Zhou et al.

    Vertical leakage/breakdown mechanisms in AlGaN/GaN-on-Si devices

    IEEE EDL

    (2012)
  • Cited by (0)

    View full text