Bridging Nanotech's Metrology Void

As we grapple with true nanotechnology and seriously begin managing things at the molecular level, there is concern over whether we are prepared to measure what we are attempting. This is complicated by the move back toward "process and device design as competitive advantage." As we progress in that direction, we are losing the ability to perform crucial measurements.

For instance, strained silicon is important today. There are several localized strained processes, but we seem to lack the metrology capable of measuring the strain on the silicon — lattice changes. We can measure physical properties of films associated with the transistor and relate those to channel mobility through detailed modeling. There are many ways to stress PMOS, but each requires a different metrology. The best way to determine results appears to be to build the transistor and see if it works.

Before, manufacturers also produced exotic chips such as parametric devices, but there existed a metrology capable of providing the necessary diagnostics and control. This allowed a rational design of experiments without footing the heavy cost of building devices to determine how — or if — they worked. Fairly accurate measurements could be made to determine how results would change when process conditions were varied.

The problem is cost. The expense necessary to precisely measure these dimensions may be presently difficult to justify by the manufacturer — or metrology provider. Some years ago, a metrology company introduced an excellent electron microscope designed to measure CDs; however, few were bought because users could not justify its cost, therefore they continued measuring CDs optically. Optical measurement then became increasingly difficult, forcing the use of a different method. Herein lies the crux of the problem: When dimensions get to the 20 or 10 nm range, how will they be measured? Preliminary data shows that CDs can be measured down to ~20 nm (for isolated lines on a silicon substrate) using greatly improved CD-SEM or scatterometry. Will it be enough to tell us that self-assembly self-assembled?

Today, we are dealing with substrate damage or corrosion. When an etch or clean is done, there is contaminant buildup. In the case of trench etch, the point is to strip photoresist or clean it. Few techniques satisfactorily measure sidewall damage at a deep trench's bottom. With ultrashallow junctions and fine dimensions, not much material can be lost during the clean process. Cleaning efficiency and damage absence are quantified based on what is measured on a flat, uniform surface, because that is where metrology tools work best. Sometimes there is a weak correlation between what is measured on a blanket wafer and what one gets on a patterned wafer; often there is not. An etched device has geometry effects that come into play, synergistic effects with etch and the clean, and crystallographic effects in ion implant structures' case. Also, as we approach quantum levels, statistical control use becomes complicated. It now becomes necessary to look at multiple sites and generate statistical data rather than simply measure in one location.

This growing measurement void is not solely the metrology provider's problem. We must find an approach that is either less tolerant toward what we cannot measure, or novel ways to do the measurement and correlation. Metrology equipment is slowly but steadily progressing toward unaffordability, and some of it is too slow for practical inline inspection. Principally, metrology is being used for failure analysis and process development, but even here a point is being reached at which we will be unable to measure what we need to.

Sematech and others are doing valuable work in areas such as the measurement of interfacial properties through optical second-harmonic generation, localized measurement of stress through µ-Raman, and advanced material characterization through TEM. However, more resources are needed.

Perhaps an answer to this lies in a new paradigm of test structures that can be indirectly measured, a radically powerful way to process data, or some other resource. Whatever the answers are, we must find them fast, because nanotechnology's demands on metrology no longer exist in some futuristic science-fiction universe. They are already on the horizon.

And they are getting closer.