Photoluminescence mapping detects Cu contamination in Si wafers
02/01/2002
Steve Westrate, Victor Higgs, Accent Optical Technologies, San Jose, California
overview
Analysis with silicon photo-enhanced recombination has proven useful in detecting and monitoring copper contamination on product wafers in fabs. This new room-temperature method focuses a visible laser to create an excitation spot on a wafer with subsequent measurement and analysis of the resulting photoluminescence. Sensitivity of this new method is equal to, and correlates well with, total x-ray reflection fluorescence, which offers high sensitivity but is relatively difficult to use within a fab environment.
Several laboratory techniques can detect copper (Cu) contamination. The most common nondestructive chemical analysis method is total x-ray reflection fluorescence (TXRF). This method has excellent sensitivity and is well accepted in laboratories for determining quantitative elemental analysis of materials, but the technique is typically used off-line, samples only a small (~1 cm2) area of a whole wafer, and is very time-consuming. These sampling restrictions prevent TXRF from being incorporated directly on a fab line and limit the opportunity to check routinely for Cu contamination.
Recent development of silicon photo-enhanced recombination (SiPHER) to characterize the near surface region of Si wafers addresses the need for fast, nondestructive screening for Cu contamination in fabs [1]. Based on room-temperature photoluminescence (PL) technology, this stand-alone technique provides ease of use, speed of measurement, precision, accuracy, and usable detection limits, all nondestructively and on wafers up to 300mm dia. It is also capable of analyzing regions of interest using ~1µm high-resolution scans.
Briefly described, SiPHER focuses a 532nm visible laser to create a 1-2µm dia excitation spot. Beam modulation confines photo-generated carriers and greatly reduces carrier diffusion length; laser excitation effectively confines photo-generated carriers to a 1µm-deep, near-surface region. Subsequent luminescence is only detected from a small (~1µm2) area within the overall carrier diffusion envelope, which further enhances spatial resolution.
SiPHER can be used to map a 200mm wafer in 10 min, gathering 7000 data points that are presented as a color map. PL signal intensity and standard deviation across a wafer are stored in statistical process control files to determine objectively when contamination levels are exceeding acceptable limits.
Experimental results
To determine the effectiveness of SiPHER in detecting Cu contamination, we performed a number of scans to obtain PL maps of 200mm wafers. Simultaneously, we captured a reflected-laser microscopic surface map image of our sample wafers so surface structures could be identified and excluded from analysis. We used an optional in situ laser marking capability to identify defect areas for further analysis.
We used p-type 10Ω•cm wafers that were intentionally contaminated with a special metal-ion solution, then annealed at 400°C for 30 min. We limited the contamination to an approximate 1cm2 area in the center of the wafer. The maps in Fig. 1 clearly show the intentional contamination from the frontside scan (i.e., the blue spot in Fig. 1a) and also from the backside scan (Fig. 1b).
We used TXRF with an average detection limit of ~1 x 1010 atoms/cm2 for iron, chromium, nickel, and Cu to measure specified positions on our test wafers. The TXRF sample area was 10mm, and we made measurements as close as possible to the area of interest. TXRF showed that the surface concentration was ~6 x 1010 atoms/cm2 and no other impurities were detected compared to the control sample.
Overall, we determined that SiPHER shows that the contaminated area had a higher average photoluminescence signal compared to the background. In other experiments, PL measurements have shown that intentional metal contamination with iron and nickel reduces the average PL intensity, whereas Cu increases the average PL signal [2]. These results can be used to set up statistical process control limits that easily identify variations in a process quantitatively.
When we increased the level of copper contamination prior to annealing, the PL detected that the intensity of Cu contamination from the front- and backside of the wafer was greater. TXRF analysis revealed Cu at ~4 x 1012 atoms/cm2 on these wafers. This demonstrated our ability to monitor Cu contamination from the backside of the wafer. These experiments were designed to understand whether we could see low levels of copper contamination.
During wafer handling, backside contamination can occur that could lead to contamination on the wafer front surface in the device active region and the front surface contamination can go to the backs of wafers and contaminate other wafers used in the same process step.
 Figure 2. PL maps after backside Cu contamination showing a) no Cu contamination when mapped from the front side and b) clear Cu contamination when mapped from the back. |
Conversely, when we deliberately contaminated wafers from the backside, our PL map of both sides showed clear backside contamination, but the front-side maps revealed only circular dark lines and no evidence of Cu contamination (Fig. 2). Our TXRF comparison measurements at this location confirmed ~1012 atoms/cm2 of Cu contamination. We attributed the lack of Cu on the front surface of the wafer to intentional thermal processing to produce gettering sites that prevented Cu from diffusing to the wafer front surface.
 Figure 3. High-resolution PL maps showing a) no Cu contamination from the wafer center and b) Cu contamination from the wafer edge. |
The wider circular "marks" on the front surface of the wafer (depicted in Figure 2a) were caused by a contaminated wafer chuck. This contamination came from prior measurements of this wafer on a lifetime metrology tool that uses a heated, full-contact vacuum chuck. During this measurement, contamination on the chuck diffused to the front surface. The level of copper contamination produced by the wafer chuck was much higher and was not gettered effectively. (SiPHER uses a low-contact wafer chuck that contacts only the wafer edge.)
In one more test, we performed PL mapping of a set of wafers that had been unintentionally contaminated with Cu during wafer polishing. These wafers were used to fabricate test structures for subsequent studies of gate oxide integrity. Here, we used PL scanning to obtain higher-resolution images at different positions on the wafer.
The central regions of the wafer showed no unusual nonuniformities in the PL signal (Fig. 3a), but at the wafer edge we observed ~5-10 cm2 areas of nonuniform PL intensity (Fig. 3b) similar to those seen in our deliberate contamination tests. All wafers used in this test showed such edge effects.
In addition, when gate structures on wafers in this final test were examined to determine gate oxide integrity, oxide breakdown measurements showed a clear spatial dependence — there was a higher time-zero dielectric breakdown toward the edge of the wafer where we observed the nonuniform PL signal.
Conclusion
We have demonstrated the ability to detect and monitor Cu contamination on product wafers in fabs using SiPHER. Sensitivity is equal to — and correlates well with — TXRF. This method allows early detection of Cu contamination and prevents continuing contamination that occurs when only TXRF techniques are used sparingly. Other work has shown that SiPHER can detect impurities in epitaxial silicon and silicon on insulator, and can also detect degradation in gate oxide integrity [3].
Acknowledgments
This paper is based on original work published by V. Higgs, et al., at the 3rd Intl. Conference on Advanced Science and Technology of Si Materials, November 2000, Kona, Hawaii. SiPHER is a trademark of Accent Optical Technologies Inc.
References
- V. Higgs, et al., Semiconductor Si 1998, H.R. Huff, U. Gosele, H. Tsuya, Editors, The Electrochemical Society Proceedings Series, Pennington, NJ PV98-1, p. 1564, 1998.
- V. Higgs, Materials Research Symposium Proceedings, Vol. 588, 2000, p. 129.
- S. Koveshnikov, et al., "Degradation of Gate Oxide Integrity due to Ni and Cu Contamination and Impurity Gettering in Epitaxial Si Wafers," 197th ECS Meeting-Toronto, Ontario, Canada, May 2000.
Steve Westrate received his biology degree from Denison University and has attended the AEA-Executive Institute at Stanford University. He is VP of global sales, Silicon Division, Accent Optical Technologies, 2186 Paragon Dr., San Jose, CA; ph 408/543-3120, fax 408/451-0970, [email protected].
Victor Higgs received his PhD at Queen Mary College, University of London, and is currently product manager at Accent Optical Technologies in the UK.