Advances in Micro-Contamination Studies

- S. Brennan
- P. Pianetta

It's surprising what a few atoms can do. Small clusters of metal atoms embedded in the gate oxide of a next-generation integrated circuit can cause it to fail. Thus, minimizing the contamination of silicon wafer surfaces during processing is, not too surprisingly, a constant battle for the IC industry. However, to quote my high-school shop teacher, "You can't make what you can't measure, because you don't know when you've got it made." As the industry produces cleaner and cleaner wafers, the challenge to measure the remaining levels of contamination becomes increasingly difficult.

Measuring the residual level of metal contamination on these ultra-clean wafers is the goal of a collaborative research effort involving scientists from SSRL, Hewlett-Packard and a group of semiconductor manufacturing companies including Advanced Micro Devices, Applied Materials, Balazs Labs, Digital Equipment, IBM, Intel, Motorola, Texas Instruments and Toshiba. The technique we are studying, known as Total X-ray Reflection Fluorescence, or TXRF, is widely used within the semiconductor industry with a rotating anode x-ray source. By employing the 54-pole wiggler on Beam Line 6 as the x-ray source, we have improved the sensitivity of this technique by a factor of 20. To understand the sensitivity of the SR-TXRF technique, you need to know that a silicon surface has ~1x10¹⁵ atoms/cm2. A conventional TXRF instrument is sensitive to elements such as iron, nickel, and copper at concentrations of 5x10 ¹⁰ atoms/cm². This is 0.00005 of a monolayer. Over the past several years the collaboration has improved the sensitivity of the technique to 3x10⁸ atoms/cm², which is equivalent to 0.0000003 of a monolayer. As another reference point, a commonly used technique for measuring surface cleanliness is Auger Electron Spectroscopy. This is typically sensitive to 0.01 monolayers.

The researchers involved in this project include Alice Fischer-Colbrie and Stephen Laderman of Hewlett-Packard; Sean Brennan, Srinka Ghosh, Piero Pianetta, and Nori Takaura at SSRL. This past year the efforts of the group have been in improving the detector, improving the cleanliness of the environment in the hutch, studying alternative experimental geometries and testing our techniques on samples of technological interest. The first of these was made possible by a CRADA, or Cooperative Research and Development Agreement. The goal of the CRADA was to develop a detector without "parasitic peaks," which is fluorescence in the spectrum caused by trace elements in the detector rather than on the sample. For instance, in one detector we have used, a beryllium-copper spring was used as an electrical contact to the silicon crystal used as the detector. Elastically scattered photons from the sample excited copper fluorescence from this spring that was then observed by the detector. The effect is to "see" copper on the sample even when none exists there. We have worked closely with the detector manufacturer to eliminate potential sources of parasitic scattering and now have a detector with the least parasitics of any we have seen. More importantly, the level of parasitic scattering is less than our theoretical detection limit, so if we see a copper signal that corresponds to 3x10⁸ atoms/cm² from a sample, we have confidence that it actually is coming from the sample.

On the cleanliness front, as many of you are aware, building 131 is not a class 10 clean room. That is the level of cleanliness required for us to measure wafers without adding contamination to them. Thus we have built a "mini-environment" clean room that can be installed inside the Beam Line 6-2 front hutch. It includes filters for cleaning the air inside the clean room and a vestibule where researchers don special clothes before entering the main room containing the chamber and samples. This portable clean room has turned out to be very successful in reducing the particle count and minimizing the contamination of wafers brought in for measurement.

The third area of effort has been in testing whether a vertical scattering geometry is worse than a horizontal one. We have always assumed that the best position for the detector is looking along the E-vector of the incident beam, as that is where the lowest level of scattered radiation exists. A single electron in a bend magnet produces radiation that is polarized in the horizontal plane. In that case, one would see no scattered radiation off of our samples with the detector placed along the E-vector. With bunches of electrons and a wiggler magnet, the degree of linear polarization in the horizontal plane is greater than 90%, but the rest is vertically polarized light which can scatter through 90 degrees and enter the detector. This scattered light is the dominant signal entering the detector in TXRF experiments and limits our ability to increase the angle of incidence on the sample because the signal saturates the detector. Unfortunately, the focused beam size from SPEAR is larger in the horizontal direction than in the vertical, so there would be an advantage to having the wafer horizontal. This places the detector above the sample, and is the geometry used in conventional TXRF instruments (where the incident light is unpolarized). Our measurements this past year have shown that the elastically scattered light into the detector in the vertical geometry (wafer horizontal) is roughly 10x higher than with the horizontal geometry. Given that the elastically scattered signal dominates in the horizontal scattering geometry, increasing that signal by a factor of 10 overwhelms the increase in fluorescence signal from the contaminants, and our detection limit is poorer in that case.

Lastly, despite all of the technique development that has occurred this year, we still had time to study a number of wafers that were technologically important to HP and other Sematech member companies. The wafers that have been studied include evaluations of incoming wafers from vendors to see what levels of contamination exist on these wafers prior to cleaning, evaluating new and existing processing tools to determine whether these tools add contaminants to the wafers during processing, and evaluating new and existing cleaning chemistries to see whether the contaminants added by steps one and two can be successfully removed or are simply replacing one contaminant with another.

This next year is likely to be just as exciting. We are implementing a new vacuum chamber for these measurements which enables us to test both 6" and 8" wafers, and to map the contaminants on these wafers by scanning them with respect to the detector. The chamber will include a robot for remote loading of wafers, which should reduce the contamination of these wafers during the measurement. This will allow the companies to perform trials sponsored by Sematech. Ultimately, if the trials are successful, we hope to implement a dedicated station on Beam Line 11 which can be used full time by the semiconductor industry.

[SSRL Welcome Page ] [Research Highlights]  [Beamlines] [Accel Physics]  [User Admin]   [Announcements] 

December 2, 1996