Automated STEM and TEM Sample Preparation in Semiconductor Process Support
The semiconductor industry is said to have unquestionably entered the realm of nanotechnology. Critical dimensions of many features are specified in terms of nanometers. Gate oxides are only a few nanometers thick. Barrier and seed layers for copper processes are not much more. Gate lengths are forecast at less than 20 nm by the end of the decade. More over, the drive to increase device density is leading to the adoption of FinFET and other new transistor designs that include complex three-dimensional structure. Even conventional planar CMOS designs now incorporate processes such as damascene interconnects that are inherently three dimensional. The need for higher spatial resolution combined with the need for cross-sectional imaging of complex structures has led to a significant increase in the demand for scanning transmission electron microscopy or STEM and transmission electron microscopy or TEM in operations of semiconductor manufacturing. Both TEM and STEM require very thin samples, typically less than 100 nm thick. This however can be hard, time consuming, and expensive to prepare. As they say, sample preparation can become a process bottleneck. Manual sample preparation is relatively very slow and it requires a skilled technician to do such job. Focused ion beam or FIB preparation provides an alternative that is fast but expensive. A method called automated pre-thinning systems, such as the EM2, perform the preliminary stages of the thinning quickly, reliably, and at a fraction of the cost of using FIB for the entire sample extraction and thinning process.
The primary limitation to SEM resolution in most applications is the beam spreading within the bulk of the microscopic specimen. Usually, the beam electrons would scatter as they enter the specimen, eventually giving up all of their energy through multiple scattering events. Typically, this happens unless one of those events directs the beam electron back out through the sample surface as abackscattered electron or referred to as the BSE. The region that encompasses these scattering events is known as the volume of interaction and the imaging signals used by SEMs can originate anywhere within this volume. It is typically many times larger than the beam diameter though it can be reduced by operating at low accelerating voltages. Both TEM and STEM form images from electrons transmitted through the sample and thus require sample thin enough to transmit most of the beam electrons. In a STEM, like the SEM, the beam is focused to a small spot that scans the sample surface. Since the sample is normally thin, most electrons scatter once or not at all as they pass through. The volume of interaction is greatly reduced and confined almost entirely to the region directly below the beam spot. In a sufficiently thin sample, STEM resolution is determined by the size of the beam.
A TEM uses a broad electron beam. It then focuses transmitted electrons into a real image that is projected onto a fluorescent screen or some other imaging device. The resolution of the TEM is determined primarily by the optical performance of its electron lenses. However, like STEM, sample thickness is a critical determinant of imaging performance and thinner is almost always considered better. The STEM and TEMs extensive sample preparation requirements impeded its acceptance in applications where high throughput and rapid results were important. Manual preparation techniques were difficult, time consuming, and believed to be often unreliable. Recently, the development of FIB based sample preparation techniques has been a significant factor in the growing acceptance of STEM and TEM in semiconductor manufacturing. The primary benefits of FIB based preparation are its ability to reliably create site specific sections of designated devices or defects, a precise control of the final thinning process. However, FIB systems are expensive and because of their versatility they are often a the most heavily used tools in the fab support laboratory. In many cases it is more economical to consider other methods for the stages of the preparation process, saving expensive FIB time for the final thinning process where its precise control adds the greatest and provides the highest return on investment.