Elsevier

Microelectronics Reliability

Volume 58, March 2016, Pages 103-112
Microelectronics Reliability

How to quantify and predict long term multiple stress operation: Application to Normally-Off Power GaN transistor technologies

https://doi.org/10.1016/j.microrel.2015.12.020Get rights and content

Highlights

  • The generalized BAZ model is refined and adapted to the GaN technology.

  • We have completed numerical Application on a Normally-off transistor GaN GS66508P-E03 650V enhancement mode manufactured by GaN Systems.

  • The concept of Maximum Rating limits and burnout conditions have been useful to derive reliability key parameters.

  • When multiple stresses are applied simultaneously, maximum rating limits values are imbricated to derive equivalent activation energy.

  • This helps to give reliability quantification rule for effective Ea and related condition of stress to assess RUL (Remaining Useful Life) condition.

Abstract

The present paper is implementing a numerical application of the Boltzmann–Arrhenius–Zhurkov (BAZ) model and relates to the statistic reliability model derived from the Transition State Theory paradigm. It shows how the quantified tool can be applied to determine the associated effective activation energy. The unified multiple stress reliability model for electronic devices is applied to Normally-Off Power GaN transistor technologies to quantify and predict the reliability figures of this electronic type of product when operating under multiple stresses in an embedded system operating under such harsh environment conditions as set for Aerospace, Space, Nuclear, Submarine, Transport or Ground application.

Introduction

The model of the Transition State Theory (TST) developed by E. Wigner [1] and M. Evans and M. Polanyi [2] in 1938 was considered to derive the unified reliability theory presented. Accordingly, the failure rates and reliability mathematics for Cumulative Distribution Function (CDF), Reliability function (R) and Probability Density Function (PDF) have been described when considering the TST concept. In former papers we have demonstrated how multiple stresses applied, may impact the effective activation energy suggested by the BAZ model [3], [4]. We will see here how to apply the pre-defined model to the case of a Gallium Nitride Normally-off high power transistor (Enhanced-mode) and a detailed numerical application. The quantified tool will be applied to determine the minimum value of stressor parameters named χ's and the equivalent single associated minimum effective activation energy to predict long term operation under multiple stresses in harsh environment. The completed numerical application on a Normally-off transistor GaN reference GS66508P-E03 650 V enhancement mode manufactured by GaN Systems is proposed to support the methodology. The concepts of maximum rating limits and burnout conditions are key factors which will give insight to derive related stressors as parameters χi's and γ's, both statistically represented by a normal distribution law.

The paper is organized as follows: after a recall of the principle of the generalized BAZ model, we will take the example of a Normally-Off GaN Power transistor detailing from the data sheet the maximum rating limits and will show other dynamic maximum rating limits to be considered in detail mainly related to the switching bias operation conditions (also related to design rules). Finally the BAZ model will be refined and adapted to this technology. A short discussion will recall the main highlights we have observed.

Section snippets

BAZ model and Transition State Theory

Shown in Fig. 1, is a free energy diagram of why things generally tend to degrade faster at higher temperatures as explained in chemistry and Reliability Physics books (see for example J.W. McPherson [5]). Stress-dependent activation energy observations seem to be general in nature (i.e., not confined just to a single failure mechanism or stress type), and works attempt to explore the conditions under which a stress dependent activation energy is theoretically expected. The generalized Eyring

A Normally-off transistor GaN: data sheet and maximum rating definitions

As an example, the data sheet given (at the date of this publication) in Table 1 defines the parameters extracted from a Power GaN Normally-Off transistor GS66508P-E03 650 V enhancement mode GaN transistor (reproduced from data sheet GaN Systems www.gansystems.com).

Generally, maximum rating limits are defined and are considered with a given margin compared to burnout failure limits (MI, MV or MP respectively for current, voltage and power dissipation).

Sudden catastrophic failures due to

A Normally-off transistor GaN: reliability impact of turn-on/turn-off switching voltage induced by commutation losses

Since Normally-off GaN transistors start to conduct significant current at VGS(th) = 1.6 V, care must be taken to ensure a low impedance path from gate to source when the device needs to be held off during dv/dt in a rectifier function. As the temperature coefficient of the eGaN FET is positive throughout its range of operation, this means that when the temperature of a localized region of the device increases, its current carrying capability is reduced causing the current to be dispersed to other

BAZ model applied to Normally-off GaN transistor

As a result the related parameters χi's are statistically represented by a similar distribution law and Fig. 7 represents a 3D-plot of such normalized stress factor χi vs the catastrophic burnout limit SiBO_failure of a Normally-off transistor GaN. When a parameter failure limit is affected by the temperature as for example observed on breakdown voltage decreasing when the temperature is decreasing [10], this effect can be easily considered within the reliability model thanks to the γV

A numerical application example for a Normally-off GaN transistor

In the following we are assessing the multi-stress induced effect the reliability figures on a GS66508P-E03 650 V enhancement mode Normally-off GaN transistor. The Arrhenius activation energy due to thermal stress only is assumed to be Ea = 2,1 eV.

In order to assess the reliability model and to define the limits of the Safe Operating Area (SOA) we need to define some specific condition of endurance testing with high enough stress to accelerate failure mechanisms.

Future reliability test programs are

The BATHTUB curve model

Considering N “good” devices of interest randomly selected from a homogeneous manufacturing lot. The term “good” means the devices are non-discernible, they are functional and their performance and electrical parameters are statistically normally distributed (Gauss or Normal statistics). Let's consider the most representative electrical parameters (biasing or leakage currents, biasing or breakdown voltages) to be a representative sensor of the healthiness of the devices. For the sake of

Discussion

The generalized BAZ model based on simultaneous multiple stress conditions has been presented and is fully depicted thanks to considering absolute maximum ratings and burnout limit normalization. The methodology implemented for the GaN transistor process in Section 5 can be easily generalized to any type of electronic device for any failure mechanism that is thermally determined as a rate function described by the Maxwell–Boltzmann distribution.

To complete our study, here are some views related

Conclusion

The principle of the generalized BAZ model exposed in Ref. [2] was recalled. An example of a Normally-Off GaN Power switch transistor detailing from the data sheet the maximum rating limits was considered. We have seen how other dynamic maximum rating limits must be pondered in detail and in particular for switching bias operating conditions (also to be related to specific Design Rules to define).

The generalized BAZ model was refined and adapted to the GaN technology. As an example we have

Acknowledgments

The study was conducted in the frame of Electronic Robustness contract Robustness Project IRT-008 managed by IRT Saint Exupery, Toulouse (France) and sponsored by the following funding partners: Agence Nationale de la Recherche, Airbus Operations SAS, Airbus Group Innovation, Continental Automotive France, Thales Alenia Space France, Thales Avionics, Laboratoire d'Analyse et d'Architecture des Systèmes — Centre National de la Recherche Scientifique (LAAS-CNRS), Safran Labinal Power Systems,

References (12)

  • J.A. Del Alamo et al.

    GaN HEMT reliability

    Microelectron. Reliability

    (2009)
  • J.B. Bernstein et al.

    Reliability Matrix Solution to Multiple Mechanism Prediction

    (December 2014)
  • E. Wigner

    The Transition State Method

    (1938)
  • M. Evans et al.

    Inertia and Driving Force of Chemical Reaction

    (1938)
  • E. Suhir

    Predicted Reliability of Aerospace Electronics: Application of Two Advanced Probabilistic Concepts

  • A. Bensoussan et al.

    Design-for-reliability (DfR) of Aerospace Electronics: Attributes and Challenges

There are more references available in the full text version of this article.

Cited by (0)

1

Permanent address: Thales Alenia Space France, Competence Center Electronics (CCEL), Toulouse, France.

View full text