On the temperature dependence of NBTI recovery

Abstract

Poly resistors around the device can be used to perform fast in situ heating on a single device on wafer level. This is a commonly used technique to apply time-saving NBTI stress in the production line [[1], [2] Ting-Kang, Chi-Shiun Wang, Kuan-Cheng Su. Self-heating p-channel metal-oxide-semiconductor field-effect-transistors for reliability monitoring of negative-bias temperature instability. Jpn J Appl Phys 2007;46(12):7639–42]. We demonstrate how such a structure can not only be used as a heating element but also as a fast tool for switching the temperature. The cool down process as well as the heating procedure are rigorously analyzed and found to be very fast (<1 s) and independent of the difference between actual and target temperature. Thus, we are able perform NBTI at a certain stress temperature, which generates a certain degradation level, while the recovery itself can be studied at arbitrary temperatures. By using this technique, our understanding of the recovery physics can be probed in an unprecedented manner. In order to guarantee that our measurements probe the ‘classic’ NBTI mechanisms, unpolluted by tunneling currents in thin oxides and the strongly process dependent impact of nitridation, we use PMOS transistors with 30 nm SiO₂ gate oxides.

Introduction

A remarkable variety of models concerning NBTI (negative bias temperature instability) degradation and recovery have been published [4], [5], [6], [7], [8], [9], [10], [11]. Unfortunately, those models are often inconsistent or even controversial. A reason for that might be the fact that people perform experiments on different device technologies with different gate oxide thicknesses, implantations and metallization stacks. If we consider diffusing hydrogen to play a crucial role in NBTI, as suggested by the RD (reaction–diffusion) model [4], [7], [9], the device geometry [6], the nature of the gate oxide [6], [8], [24] as well as the character of interface-near layers are likely to influence the situation significantly. A reliable comparison of phenomena observed is difficult, when such different devices are stressed under various bias and temperature conditions. Consequently, the dynamics of degradation and relaxation have been attributed to many effects like hole trapping/tunneling [12], hydrogen release at the interface [4], [7], [9], hydrogen diffusion/reaction/cracking within the gate oxide and at the interface [7], [9], [11], [13], [14], [15], [21], bonding and antibonding hydrogen configurations [16], [17], various oxide defects with wide spread energy trap levels [18] and some more.

To summarize the widely reported physical facts, temperature as well as electric field (respectively the interface free hole concentration) play an important role in NBTI degradation and recovery. The question if the total effect is a fusion of multiple components which can differ in their dependencies remains to be seen.

In order to detect fast recovering components a short stress-measurement delay is required. While fast measurements in the microsecond regime have been achieved for devices with thin gate oxides [19], we report for the first time sub-millisecond I_d–V_g measurements on thick gate oxides, which are more complicated to achieve due to the much larger gate voltages differences between stress and recovery. A large voltage switch involves a temporary high charging/discharging current which can exceed the measurement range and is therefore hard to handle. In order to achieve reasonable measurement speed and resolution we have used SMUs (source-measurement units) on highest novel industrial standard. By combining a fast sampling measurement with a conventional spot measurement, we are able to record fast recovering components (<1 ms) as well as long-term effects.

In this paper our aim is to investigate the temperature dependence of NBTI recovery. The motivation to do this is to clarify, if the observed effects are thermally activated or not. In the case of NBTI thermodynamic models are often linked to mobile hydrogen which can diffuse around and passivate or create dangling bonds at the substrate-oxide interface and inside the bulk of the gate-oxide [15], [16], [21], [24], [25].