Should We Analyze for Trace Metal Contamination at the
Edge, Bevel, and Edge Exclusion of Wafers?
Meredith Beebe, Chris Sparks, and Ron Carpio
InternationalSEMATECH, 2706 Montopolis, Austin TX78741.
Abstract. The edge, bevel, and edge exclusion area of a wafer has historically been difficult to monitor for trace metals.
Standard trace metal surface techniques such as total reflection x-ray fluorescence spectroscopy, time-of-flight
secondary ion mass spectrometry, and vapor phase decomposition inductively coupled plasma are currently not capable
or have difficulty measuring metals to the edge and bevel of the wafer. With shared metrology toolsets and new
materials being introduced into semiconductor fabs, it is important to measure possible contamination in these areas of
the wafer. Tools that have edge grip pins or centering and aligning pins, also are at risk to contaminate wafers at the
edge and bevel. A technique had been developed known as the beveled edge analysis tool that chemically extracts
contamination from the edge, bevel and edge exclusion of a wafer that is then quantified by inductively coupled plasma
mass spectrometry. In this study we will show correlation of this technique to standard trace element analysis methods.
We will also present data from characterizing processes and fab tools that will benefit from this measurement.
to its small spot size capabilities, but does not give a
value that is representative of the complete area of the
edge, bevel, and edge exclusion. It is also difficult for
TOF-SIMS to analyze on the angled bevel of a wafer.
INTRODUCTION
The edge, bevel, and edge exclusion of a wafer has
typically been an area difficult to monitor for trace
metal contamination. It is becoming an increasing
concern to monitor these areas of the wafer in the chip
fabrication process due to cross- contamination issues
with shared metrology toolsets between copper and
non- copper processing, and with new materials
introduced into production for high k gate dielectrics
[1-3]. Other possibilities for contamination at the
edge, bevel, and edge exclusion of a wafer are
centering pins in tools that align wafers, incomplete
backside etching of films, and contaminated cassette
boxes.
In order to analyze the edge, bevel, and edge
exclusion of a wafer a mechanical jig was constructed
(Universal Engineering, Lowell, MA) and is referred
to as the beveled edge analysis tool (BEAT). A wafer
is supported vertically on a vacuum chuck, and is
rotated through a solution that chemically extracts
contamination off the wafer using chemistries similar
to VPD. BEAT can be refigured to analyze 200 and
300mm wafers. More details on this device have been
previously discussed in other studies [4].
Traditional trace metal techniques have neglected
to fully characterize the edge, bevel, and edge
exclusion of a wafer due to various limitations. Total
reflection x-ray fluorescence spectroscopy (TXRF)
cannot operate close to the edge of a wafer due to
scattering of radiation and has an almost 10mm edge
exclusion. Vapor phase decomposition (VPD) can
quantify trace metals to the edge of the wafer,
however, VPD cannot analyze on the bevel of a wafer.
Time-of-flight secondary ion mass spectrometry
(TOF-SIMS) can measure to the edge of a wafer due
EXPERIMENTAL
The wafer is held in the mechanical jig (Figure 1)
by vacuum, which allows the edge, bevel, and edge
exclusion of the wafer to be suspended in an extraction
chemistry of dilute hydrogen peroxide and
CP683, Characterization and Metrology for VLSI Technology: 2003 International Conference,
edited by D. G. Seiler, A. C. Diebold, T. J. Shaffner, R. McDonald, S. Zollner, R. P. Khosla, and E. M. Secula
© 2003 American Institute of Physics 0-7354-0152-7/03/$20.00
309
Figure 2 were wafers
concentrations of copper.
S1 1.E-M3-
-H-
•| 1.E+12 H
§ 1.E+09
1.E+08 -
with
increasing
§
|
4
£
—r~
—1-
| 1.E-M1 fc 1.E-MO
spiked
)
1
2
3
wafer set
FIGURE 1. Wafer Resting on BEAT Alignment Pins with
Sample Boat.
FIGURE 2.
Techniques
hydrofluoric acid. The wafer is slowly rotated through
the SOOuL chemistry (~8 minutes/rotation) in order to
allow the extraction process to take place where the
hydrophilic surface is turned to hydrophobic. A
0.001-inch micrometer controls the depth of the wafer
in the extraction solution. The sample boats used are
made of molded high purity PFA set in a PTFE
housing (Savillex Corporation, Minnetonka, MN). To
avoid contamination from wafer to wafer, the
approximate one- third of the wafer between the wafer
pins is not scanned through the extraction solution.
Once the wafer has completed the rotation in the
sample boat, the extraction solution is poured into a
500uL PFA vial (Savillex Corporation, Minnetonka,
MN). The solution is ready for analysis by ICP-MS.
The majority of the analyses were done on a
quadrupole Agilent 4500 ICP-MS (Palo Alto,
California) operated under normal plasma conditions.
A lOOuL/ sec self- aspirating nebulizer (Elemental
Scientific Inc., Omaha, NE) was used in order to
conserve the small sample volume.
These wafers consistently had almost 6E10
atoms/cm2 levels of copper in comparison with the low
to mid E09 atoms/cm 2 detection limits of copper for
VPD-ICP-MS and TXRF. Whether this higher level
of copper was a result of actual copper present on the
wafer at the edge, bevel, and edge exclusion, or an
isobaric interference with another species was
examined. In order to prove we are measuring copper,
two new bare p-type 200 mm silicon wafers were run
on BEAT and contamination was collected at the edge,
bevel, and edge exclusion. The solution was then
analyzed by a Finnigan Element 2 Sector-Field high
resolution ICP-MS (Table 1).
RESULTS AND DISCUSSION
The first wafer analyzed showed the E l l level of
copper similar to what has been seen in previous
experiments. Inspection of the spectra (Figure 3) from
the HR-ICP-MS analysis determined that there is
copper on the wafer as well as an interference peak.
This copper is most likely present on the backside or
bevel of the wafer and would not have been detected
by any other methods such as VPD or TXRF, due to
the difficulties of measuring on these areas of the
wafer. Further examination of the spectra also shows
there is an interference with 63 Cu. The interference
with copper most likely is the 28Si19F16O species. The
silicon in this species is supplied by the wafer, the
fluorine from the hydrofluoric acid used in the
Copper Concentrations with Various
TABLE 1. Copper Results in atoms/cm2 from HRICP-MS
Cu63
Wafer 1 (Pass 1)
1.3E11
Wafer 1 (Pass 2)
1.9E09
Wafer 2 (Pass 1)
4.2E09
Wafer 2 (Pass 2)
1.9E09
This study developed into two parts, investigation
of relatively high contamination levels seen on virgin
wafers and characterization of an edge exclusion clean
process.
Technique Characterization
The first investigation developed after it was
noticed in previous studies that control wafers (wafer
set 1 of Figure 2) had higher copper measurements by
BEAT compared to VPD and TXRF. Sets 2-4 of
310
able to etch the tantalum barrier to silicon, even with
the increasing concentrations of hydrofluoric acid
used. Since the tantalum would not dissolve in the
extraction solution and was hydrophilic, the
hydrofluoric acid was pulled up into the copper film.
This left black oxidized copper residue along the edge
of the copper film in some spots.
extraction solution, and the oxygen from the water or
hydrogen peroxide.
FIGURE 3.
A second attempt to characterize the edge clean
was done using a wafer processed without the
tantalum barrier. This wafer had just the copper seed
layer, the electroplated layer, and the 2mm edge clean
processes. The BEAT measurement was taken with a
1mm edge exclusion analysis. Although the extraction
solution was not in contact with the copper film, there
was a slight halo formed on the edge of the
electroplated copper.
Perhaps vapor from the
extraction solution was reacting with the copper
causing a color change, which is what we were
referring to as a halo. The effect of the BEAT analysis
on the electroplated copper film without the tantalum
barrier will be a future study. Analysis of the solution
collected by BEAT on the ICP-MS found
approximately 10,000 ppb copper and such a large
quantity of tantalum that the detector shut off and was
unable to measure. These results were interesting
since this second wafer was processed without the
tantalum barrier, and yet tantalum was found in
extremely large quantities.
The tantalum could
possibly be contamination from the seed deposition
tool or residual from a reclaim process. These results
indicate that the clean is not removing all the copper
from the edge, bevel, or edge exclusion.
63
Cu Spectra from HR-ICP-MS
A second pass of the same wafer in BEAT and
analysis in the HR-ICP-MS shows the copper level
dropping to 1.9E09 atoms/ cm2. This reduction is
expected as the majority of the copper is extracted
from the first BEAT pass. The second wafer analyzed
had 4.2E09 atom/cm2 copper measured on the first
pass. After examining its spectra it was determined
that less copper was present on this wafer, but the
same level of interference from 28Si19F16O was still
present. However, in a quadrupole based ICP-MS this
interference would also be quantified with 63Cu. The
28 19 16
Si F O species would quantify to roughly 5E09
atoms/crn2 of copper if analyzed on a quadrupole ICPMS.
CONCLUSION
Process Tool Characterization
The second part of this study involved
characterizing an edge exclusion clean performed in a
Novellus Sabre XT at International SEMATECH.
This 300mm copper electroplater has a built in edge
exclusion clean of 2mm using a sulfuric acid and
hydrogen peroxide rinse. Setting the BEAT to a 1mm
edge exclusion analysis, the effectiveness of the clean
was evaluated. The first attempt to characterize this
clean was on a wafer processed in several steps. The
first step was a deposited tantalum barrier, then a
copper seed layer, then electroplated copper, and
finally the edge exclusion of copper seed was etched
off The tantalum barrier is not removed with the
copper during the tool's edge clean. BEAT was not
311
The relatively high level of copper measured on
virgin wafers shows copper contamination on the
edge, bevel, or edge exclusion, which highlights the
need for this analytical measurement. Also, an
interfering species of 28Si19F16O with 63Cu was
measured on a HR-ICP-MS that was not previously
resolved on the quadrupole ICP-MS.
After
characterizing an edge exclusion clean on an
electroplating tool it was determined that copper was
not being completely etched by the rinse of sulfuric
acid and hydrogen peroxide. Further study is needed
to investigate why the edge exclusion copper wafers
saw a halo effect in the copper when visibly the BEAT
extraction solution was not touching the copper film.
ACKNOWLEDGMENTS
The authors would like to thank Evelyn Ferrero of
AMD for collecting and analyzing data from the HRICP-MS. Kam Hettiaratchi for assistance in the
processing the tantalum barrier and copper seed
wafers.
REFERENCES
1. Gualhofer, E., Oyer, H., Tsui, B, "Wafer Backside Spin
Process Contamination Elimination for Advanced
Copper Device Applications," Semiconductor Fabtech,
11th Edition, January 2000. pp 289-293.
2. Simpson, R., Ritzdorf, T., Dundas, C., "Reducing Bevel
and Edge Contamination to Help Enhance Copper
Process Yields," Micro, October 2000. pp 41-53.
3. Geraghty, P. and Mclnerney J., "Using Exclusion Ring
Technology to Avoid CVD Tungsten Bevel
Contamination," Micro, July/ August 2000.
4. Sparks, C., Gondran, C., Lysaght, P., Donahue, J., "A
Novel Technique for Contamination Analysis Around the
Bevel and Edge Exclusion Areas of 200mm and 300mm
Silicon Wafers," Submitted to SPIE Advanced
Microelectronic Manufacturing proceedings in March
2003.
312