Reliability Detection of Process-Induced Metallization Defects in GaAs Devices
- PDF / 1,126,819 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 122 Downloads / 246 Views
Reliability Detection of Process-Induced Metallization Defects in GaAs Devices Steve H. Kilgore Technology Reliability & Quality, Freescale Semiconductor, Inc., Tempe, AZ 85284, USA ABSTRACT Process-induced defects in electroplated Au interconnect metallization on GaAs devices were detected during the course of reliability testing. Abnormally high lognormal sigma values (σ > 0.7) indicated the existence of a bi-modal failure mechanism. A distinct early lifetime failure mode was observed along with the intrinsic electromigration metallization wear-out failure mode. Physical characterization of the electroplated Au film revealed as-deposited nanoscale voids. Elimination of these voids through process improvement as well as suggested mechanisms for the early failures are discussed. INTRODUCTION The growth of the gallium arsenide (GaAs) market has increased considerably as mobile and cellular devices comprise the largest consumer products. Annual revenue of GaAs devices has grown from under $3 billion to over $5 billion in the last ten years. This demand has spurred improvements in manufacturing yields and reduced production cost. Even so, GaAs technology maturity has not reached the level it is for silicon CMOS technology insofar as our understanding of reliability and defectivity. Devoted research studies are essential to ensure the quality and reliability of GaAs technologies. Of particular importance are relevant studies on the reliability of gold (Au) interconnects employed in compound semiconductors. The reliability of passivated electroplated Au interconnects and the impact of defectivity were examined in this investigation. Electromigration reliability assessments generate failure time data that routinely follow a lognormal statistical distribution. Deviation from a lognormal line fit is a strong indication that the data contains bi-modal or multi-modal failure mechanisms. In these cases, it is critical to have proper physical characterization and identification of the mechanism responsible for the failures. Experience with IC reliability over the last few decades has established that the failure characteristics of devices are described by the “bathtub” curve. At the initial segment of the curve are devices failing early due to manufacturing defects. These early failures are considered extrinsic failure mechanisms since the failures are not associated with the inherent design or materials used. Once the intended lifetime of a device is reached it begins to accumulate wearout failures as indicated by the increasing failure rate segment of the curve. These “end-of-life” failures constitute intrinsic failure mechanisms because the inherent materials are wearing out in well defined modes, for instance those due to electromigration wear-out. Some of the initial studies on thin Au films sparked controversy over the direction of electromigration [1, 2] which was resolved once it was discovered that sodium from soda-lime glass substrates diffused into the Au film causing voids near the anode end [3]. A fundamental aspec
Data Loading...