Monitor Friction for CMP End Point Detection
By J. Scott Moore
According to engineers at Micron Technology, Inc. (Boise, ID), monitoring the change in friction might be a better way to determine chemical mechanical polishing end points (CMP) than more technically sophisticated methods like optical interferometry. A chief advantage of a friction monitoring approach is that it offers a number of options in just how the change in friction is measured. Engineers can select the most appropriate option for a particular process step.
As device dimensions shrink, the depth of focus of optical lithography systems is reduced (click here to read a related article), requiring nanometer-scale flatness of wafer surface topography. Processing steps like metallization and dielectric deposition create more severe topography variations as the number of layers increases. CMP, considered by many to be the only viable way to achieve such flatness, has fast become the preferred method for planarization.
Because of the complexity and relatively short history of CMP, semiconductor engineers must often rely on trial and error to determine processing parameters. One of the more troublesome issues is determining just when the CMP is finished. Stopping polishing too soon or too late can destroy equipment throughput and yields.
CMP end point usually occurs when surface topography is flattened, or when a layer is completely removed. The polishing removal rate is a function of at least eight parameters, some not understood and others that have a complex interrelationship. Machine-controlled variables include process duration, backpressure, conditioning technique and others. Process and material-dependent parameters include polishing step (metallization or dielectric, for example), composition of the polishing layer and underlying layers, pattern density and topography, wafer bow and warpage, slurry composition and flow, and many more.
The Micron Technology engineers based their friction approach on a simple premise: when a layer is polished away, the underlying layer emerges. With the new material, the friction between the wafer and pad will change, and thus the force required by the motor. The same thing happens when topography is eliminated. As a result of the change in friction, a variety of measurable parameters change. The easiest way to monitor the change in friction is by measuring changes in current drawn by the motor rotating the wafers and/or pads (click here to read a related article). Other monitoring schemes, depending on application, include measuring changes in heat produced, or measuring the motor's actual torque. However, in most cases the simplest and cheapest solution is current monitoring. The friction method is accurate within 100 angstroms, or a couple of seconds of processing.
Ultimately, the most accurate method for end point detection is a direct thickness measurement, generally accomplished by means of interferometry. Unfortunately, such optical thickness measurements must be performed through a moving slurry solution whose composition is changing. Monitoring changes in friction is a simple, accessible method of determining end points, with sufficient accuracy for most planarization steps.
More accurate control of CMP improves control of such device parameters as threshold voltage and contact resistance.
For more information: Trung Doan, Senior Engineer, Micron Technology. Tel: 208-368-4000.
Dr. J. Scott Moore is a consultant based in New York. Having worked many years as a scientist in the computer industry, Dr. Moore now writes technology based articles for several publications, and provides business and market intelligence services to a range of clients. His recent work includes conference preparation and a series of in-depth reports on emerging technologies, emphasizing semiconductor and electronics applications.