Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
DEVELOPMENT OF AN XRF METROLOGY METHOD FOR
COMPOSITION AND THICKNESS OF
BARIUM STRONTIUM TITANATE THIN FILMS
Thomas Remmel and Dennis Werho
Semiconductor Technology, Semiconductor Products Sector,
Motorola Inc., Mesa, Arizona, 85202
ABSTRACT
Thin films of barium strontium titanate (BST) are being investigated as the charge storage dielectric
in advanced memory devices, due to their promise for high dielectric constant.
Since the
capacitance of BST films is dependent upon both stoichiometry and thickness, implementation
into manufacturing will require precise metrology methods to monitor both of these properties.
This is a challenge, since the BST film thicknesses are 60 nm or less. A metrology method was
developed based on wavelength dispersive x-ray fluorescence and applied to the measurement of
both stoichiometry and thickness of BST thin films.
INTRODUCTION
The continually increasing demand for more memory and the trend towards higher operating
speeds of integrated circuits dictates the need for high density memory accessible at rapid rates.
The time-proven bit cell design of DRAMS (dynamic random access memories) relies upon the
storage of charge in an integrated capacitor. To date, the dielectric used in this capacitor has been
SiOZ; the challenge of maintaining sufficient charge storage density as the bit cell size decreases
has historically been met by decreasing the SiOZ thickness. However, this approach is reaching
its limit and although nitridation of Si02 affords some increase in dielectric constant, radically
higher dielectric constant films are required to replace SiOz. Barium strontium titanate (BST) is
particularly attractive because of its very high dielectric constant, low leakage, thermal stability,
low dissipation factor and promising fatigue behavior [l-4].
The dielectric constant of Bai$&Ti03
is dependent upon both the Ba/Sr ratio and the A/B
[(Ba+Sr)/Ti] site ratio [5-81. For example, a 10% deviation in A/B ratio fi-om 1.0 can result in a
30% decrease in the dielectric constant of the film. Since the capacitance per unit area is a
function of both the dielectric constant and the film thickness, implementation of BST into
manufacturing requires precise metrology methods to monitor both the stoichiometry and
thickness of the films. This is no small challenge, considering that BST film thicknesses are 60
run or less and that control of the stoichiometry to better than 1% relative will be required. In
addition, the metrology method needs to be rapid, non-destructive, non-invasive, wafer-fab
compatible and capable of handling wafers up to 300 mm in diameter. The ability to measure
across-wafer variations is also desirable.
99
99
This document was presented at the Denver X-ray
Conference (DXC) on Applications of X-ray Analysis.
Sponsored by the International Centre for Diffraction Data (ICDD).
This document is provided by ICDD in cooperation with
the authors and presenters of the DXC for the express
purpose of educating the scientific community.
All copyrights for the document are retained by ICDD.
Usage is restricted for the purposes of education and
scientific research.
DXC Website
– www.dxcicdd.com
ICDD Website
- www.icdd.com
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Several measurement methods were considered as potential metrology options, including
wavelength dispersive x-ray fluorescence (WDXRF), energy dispersive x-ray fluorescence
(EDXRF), Rutherford Backscattering Spectrometry (RBS), Fourier Transform Infrared
Spectroscopy (FTIR), Spectroscopic Ellipsometry (SE), Inductively Coupled Plasma-Atomic
Emission Spectroscopy (ICP-AES), Glow Discharge-Optical Emission Spectroscopy (GD-OES),
Glow Discharge Mass Spectroscopy (GD-MS), and small spot XRF. The inability of EDXRF
to resolve the peaks of interest from interfering peaks in BST precluded its use for this work.
The utility of SE for film thickness measurement is well accepted; however composition
measurement via SE is doubtful. FTIR shows some promise as an indirect “fingerprinting”
method for BST film process monitoring; however, the ability of FTIR to directly measure BST
film stoichiometry has not yet been proven. RBS was ruled out due to its high cost and lack of
wafer fab compatibility; ICP was not considered due to its destructive nature, lengthy analysis
time, poor spatial resolution and lack of fab compatibility.
GD-OES, GD-MS and small spot
XRF were also not considered viable options for various technical reasons. WDXRF was chosen
as the preferred method because it satisfied most of the criteria, is already an accepted, fabcompatible tool and offers a high degree of maturity in quantitative analysis of thin films. The
weakest attribute of WDXRF is a rather large analysis area, 35 mm on our instrument; however
this still affords the ability to measure film composition and thickness at several locations across
a wafer. Apertures can be used to reduce the analysis area to as small as 5 mm in diameter;
however, with this area reduction comes an associated reduction in count rate and a decrease in
statistical precision of the measurement method. To compensate, analysis time can be increased
substantially, but the time penalty is usually too high to be feasible except in unusual cases.
EXPERIMENTAL
METHOD
For x-ray analysis of BST on a Si wafer, several of the elements have characteristic x-ray lines
with wavelengths in close proximity to one another; thus the motivation to use WDXRF rather
than EDXRF.
On the XRF system utilized in this work, a fixed channel was used for
measurement of the Ti-Ka line and a scanning channel for both the Ba-LP and Sr-Ka lines,
whereas oxygen was not measured.
To quantify stoichiometry and thickness from x-ray intensity, calibration of the XRF is required
through the use of appropriate standards. Two series of Bai,Sr,Ti,03
standards were created:
(1) initially, a set of MOD (metal-organic-decomposition) films made from mixtures of strontium
titanate (STO) and barium titanate (BTO) precursors, spun onto 6” Si wafers and subsequently
baked to drive off the volatiles; and (2) a series of BST films sputtered from different
composition targets. The MOD films ranged from 50 to 200 nm in thickness, had an A/B ratio of
1.O, and included 7 different stoichiometries that covered the range from x = 0 to 1.O. Later on,
standards with sputtered films were created, which included nominal stoichiometries of x = 0.3,
0.5 and 0.75 and had A/B ratios of about 1.1. All films were deposited on oxidized <loo> Si
wafers. The thicknesses of the BST standard films were measured using SE and x-ray
reflectivity.
100
100
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
101
The stoichiometries of the MOD standard films were independently determined using three
methods: (1) weight ratios of the ST0 and BTO precursors; (2) PBS; and (3) ICP-AES; only
RBS was used on the sputtered films. The results from RBS and ICP analyses validated the
precursor weight ratio values of the MOD films, as shown in Fig. 1. Considering the
uncertainties associated with ICP and RBS analyses of these films, it was decided that the
stoichiometries as calculated from the weight ratio values would be accepted as the calibration
values for the MOD standard films.
All standard films were analyzed using XRF to obtain the x-ray intensities for Ba, Sr and Ti, and
then empirical relationships were derived between these intensities and BST stoichiometry.
Since the absoprtion of the Ba, Sr and Ti characteristic x-rays is negligible for thin BST films,
relationships of the form:
Sr/Ti (stoichiometry) = A * Sr/Ti (x-ray intensities) + B
were derived for both Sr/Ti and Ba/Ti. The empirical calibration curve of Sr/Ti for the MOD
standard films is shown in Fig. 2, and the linear fit is excellent; similar correlation was obtained
for Ba/Ti. In applying these empirical relationships to calculate the stoichiometry of unknown
BST films, we used the convention whereby the sum of (Ba + Sr) stoichiometry was normalized
to 1.0 to yield the Ti stoichiometry (y) in Bai-,Sr,Ti,03.
For thickness measurement, an empirical calibration of the form:
t = A * Ba (net kcps) + B * Sr (net kcps) + C
was also derived. The results are plotted in Fig. 3 against film thickness as measured by SE. The
agreement between the two methods is very good. Note that knowledge of the density of the
1.0
1.0
c 0.8
0.8
2
L8 0.6
.s
;i
cd 0.4
a
2
g 0.2
3
.s
0
.s
$
0.6
2
0.4
&i
0.2
0.0
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Ba stoichiomeQ calculated from
MOD solution weight ratios
FIGURE 1. Comparison of ICP and RBS versus
weight ratio stoichiometry of the MOD BST standard
films.
0.0
0.1
0.2
0.3
Sr/Ti X-ray Intensity Ratio
0.4
FIGURE 2. XRF calibration results for Sr in MOD
BST films.
101
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
102
BST films is required when determining film thickness via XRF. The density of the MOD films
was determined to be about 4.2 gm/cm3, whereas BST films deposited by other methods will
result in films with different densities. For example, densities of 5.1-5.6 gmkm3 have been
measured (using x-ray reflectivity) for sputtered BST films, depending upon process conditions.
JWPEATABILITY
The repeatability of the XRF method for measuring BST film stoichiometry and thickness was
determined by repetitive analysis of a nominal Bat. 25Sr 0,75Ti03 film over a 15 month time frame.
Shown in Figs. 4-6 are the results of this gauge capability study for Ba stoichiometry, A/B ratio
and film thickness, respectively.
0.30
8
y = 0.9989x
Rz = 0.9924
200
c
$j
0.28
E
0
Is
h
; 100
3
.3
0
2
a
E9
4
E
0.26
0.24
0.22
0
100
0
Apr
200
Jul
Ott
Jan
Apr
Thickness by SE (nm)
Film thickness calibration for MOD
BST films. Thickness derived Corn Ba and Sr x-ray
intensities; BST densiy 4.0 gmlcm3.
FIGURE
3.
FIGURE 4. XRF measurementsof Ba on the same
Ba,&ro.,5Ti03 film over a 15 month period. Mean
value is 0.258, sigma=O.O03
1.00
322095
*!b&+%w*“‘z-8
.r
go
$j 49
1
‘O 48
8
0.90
/
0.85
!
Apr
I
I
I
I
I
Jul
Ott
Jan
Apr
Jul
FIGURE 5. XRF measurements of A/B ratio
on the same Ba0,25Sr0,75Ti03 film over a 15
Mean value is 0.967;
month period.
sigma=0.008.
E
z 47
z
a
46
Apr
Jul
Ott
Jan
Apr
Jul
FIGURE 6. XRF measurements of BST film
thickness on the same Ba0.25Sr0,75Ti03 film
over a 15 month period. Mean value is 48.2
nm; sigma=0.44.
102
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
103
One challenging aspect of developing metrology methods is the transfer of the method to other
labs and the ability to achieve inter-lab agreement among results. This has been accomplished
with the BST XRF metrology method. Shown in Fig. 7 is a comparison of A/B ratios between
two different labs on a combination of similar and identical BST films. The agreement between
the two labs is within the + 2% precision of the method for A/B ratio measurements.
MEASUREMENT
OF ACROSS-WAFER
STOICHIOMETRY
As noted previously, the XRF method, with its 35 mm diameter analysis area, is capable of
measuring across-wafer variations in film stoichiometry and thickness. This has been used
extensively during development for process optimization, since uniformity of BST film
properties across the wafer is critical to yield functional and reliable devices. It can also readily
be used as a process monitoring tool, to ensure process control. Shown in Fig. 8 is an example of
the results from measurement of the A/B ratio across a 200 mm wafer.
In this case,
“overlapping” measurements were taken every 15 mm, from the wafer center to within a few mm
of the edge, along orthogonal x- and y- directions relative to the wafer flat. (The XRF tool has an
r-8 stage, so across-wafer measurements involve the wafer being rotated 90, 180 and 270 degrees
between radial measurements taken center-to-edge). The results in Fig. 8 show relatively good
across-wafer uniformity in A/B ratio for the BST film. The analysis time for measurements such
as those shown in Fig. 8 is on the order of 1.5 hours/wafer, of course, simultaneously yielding
across-wafer data for Ba/Sr ratio and BST film thickness variations.
.
SimilarWafers
A A A
.oeA*
.oeA*
’ l&5
1.00
0.9 1.0
1.1 1.2 1.3 1.4 1.5
Lab 1
FIGURE 7. Inter-laboratory comparison of A/B
ratio measurementsof BST films on similar and
identical wafers.
-100
-50
0
50
100
Distance from Wafer Center (mm)
FIGURE 8. Across-wafermeasurementsof the A/B
ratio of a BST film on a 200 mm wafer.
103
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
0.8
E!
‘g
2
3
i
g
104
1.6
y = 1.0077x - 0.0129
R’ = 0.9994
y = 0.994x - 0.0013
Rz = 0.9999
0.6
0.4
22
0.2
0.8
0.2
0.4
0.6
0.8
MOD Calibration
FIGURE 9. Comparison of Ba stoichiometry c6
several BST films derived using sputtered versus
MOD film calibrations.
COMPARISON
OF CALIBRATION
0.8
1.0
1.2
1.4
1.6
MOD Calibration
FIGURE 10. Comparison of A/B ratio for several
BST films derived using MOD versus sputteredfilm
calibrations.
USING MOD VS SPUTTERED
STANDARDS
A fundamental question that existed after the initial calibration using MOD standards was
whether those standards were valid for analyzing sputtered BST films; thus the second set of
calibration wafers were created composed of sputtered BST from different composition targets.
After calibrating with the sputtered standards using a manner identical to the MOD films, a third
set of BST films was analyzed using both MOD and sputtered calibrations. A comparison of
these results are shown in Figs. 9 and 10. The excellent agreement between these two calibration
methods verified the validity of using the MOD standards for sputtered film analysis; the
sputtered films also provided a second set of BST reference films over a narrower range of
thicknesses and stoichiometries.
COMPARISON
OF EMPIRICAL
VS. FUNDAMENTAL
PARAME TERS ANALYSIS
Although the empirical methodology for BST quantification was developed from x-ray
intensities, this approach requires time consuming off-line data processing to obtain
stoichiometries. It would be preferable to utilize the data processing capability of the XRF tool
itself, and thus the attempt was made to apply the vendor-provided fundamental parameters
(FP) [9, lo] treatment to the data to achieve on-line quantification. A comparison of the results
for Ba and Sr stoichiometries, using both empirical and fundamental parameters approaches
anchored on the MOD calibration wafer set, reveals that the agreement between the two
analytical approaches for Ba and Sr was excellent. However, when comparing the two methods
for A/B ratio measurement, the agreement was not as good. This is depicted in Table I, where the
A/B ratios of the entire matrix of MOD standard films are shown as calculated using the
empirical and fundamental parameters methods of quantification. The A/B ratios for these films
should be 1.O; however, the results indicate that the fundamental parameters method yields A/B
104
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
ratios that deviate from 1.O and show a trend of understating the A/B ratio for thin, Ba-rich BST
to overstating the A/B ratio for thick, Sr-rich BST. No such bias is evident in the empirical
results on the same data set. One can simulate this trend in A/B ratio as a function of film
thickness in the empirical approach by using incorrect values for the background count levels for
Ba and Sr. It is postulated that the fundamental parameters approach is not properly
compensating for the high background levels encountered in our analyses.
Ba:Sr Composition
loo:o
88:12
65135
44:56
25:75
8:92
0:lOO
60 nm
FP
Empirical
0.94
1.01
BST FILM THICKNESS
120 nm
FP
Empirical
0.97
1.00
0.95
0.97
0.98
1.00
1.00
1.00
0.98
1.00
1.00
1.00
1.00
0.99
0.99
0.99
1.02
1.00
FP
1.00
0.99
1.00
1.01
1.02
1.03
1.03
180 nm
Empirical
1.00
0.99
1.00
1.00
1.01
1.01
1.01
Table I. Comparison of A/B ratios for the MOD BST standards set calculated using empirical versus
fundamental parameters analysis methods for three different BST film thicknesses.
ERRORS DUE TO SUBSTRATE
DIFFRACTION
There are several sources of error in the analysis of films using XRF, including the accuracy of
the calculation algorithm (discussed previously), the accuracy of the standards, the quality of the
calibration process, counting statistics, instrument stability and knowledge of the background
count levels. Particularly problematic for WDXRF analysis of BST films is the issue of
substrate diffraction, which impacts the background levels [ 111. Substrate diffraction refers to
diffraction of the incident radiation by the single crystal substrate. This diffracted intensity,
which is dependent upon the geometrical arrangement of the radiation source, substrate and the xray detector, will vary as a function of crystallographic orientation of the substrate and its
angular position and manifests itself as additional background x-ray intensity.
In WDXRF, with fixed position detectors, it is not possible to scan the spectrometer wavelength
to measure backgrounds levels; additionally, for scanning channels, measurement of background
levels can increase the analysis time by 3X or more. Therefore background levels were measured
on BST film-free wafers (for Ti) or measured for Ba in ST0 films and for Sr in BTO films.
These “element free” background levels were then assumed to be applicable to the unknown
films. However, substrate diffraction can contribute to these background level measurements and
therefore must be well characterized from wafer to wafer. As an added complication, to achieve
full wafer coverage, many XRF tools use an “r-0” stage to reach all locations on a wafer rather
than an x-y stage. This introduces the added variable of angular geometry (e) for each
measurement location
105
105
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
106
4.1
0
90
180
270
3
ST?a
80
s &
3 8
4.0
3.9
3.8
3.7
;i
3.6
3.5
0
360
360
Angular Rotation - 8 (degrees)
FIGURE
11. DifEaction e&ct for Ti-Ka on bare Si
< 1OO> wafer. This shows background levels
measuredby Ti spectrometeras a function of angular
rotation of Si substrate.
FIGURE
12. Ditiaction effectfor Sr-Ka on bare Si
Cl OO>wafer.
Shown in Fig. 11 is a plot of the x-ray intensity collected by the Ti-Ka channel measured for a
bare Si x100> wafer (no BST), as a function of 8. Note the strong substrate diffraction effect;
although there is no Ti present, the background x-ray levels vary by as much as an order of
magnitude as a function of 8. The maximum background levels are as high as 10% of the
maximum Ti intensity measured in typical BST films. The repeating pattern with angular
rotation is due to the four-fold symmetry of the Si x100> wafer. Although diffraction effects are
significant for Ti, they can generally be minimized by performing analysis at an angular position
having a slowly varying and minimal substrate diffraction intensity, such as near 0, 90, 180 and
270” as shown in Fig. 11. However, the diffraction effect is much more complicated for Sr; for
which the diffraction intensity for the Sr channel from a bare Si <loo> wafer is shown in Fig. 12
as a function of angular rotation. For Sr, these background levels are 30-50% of the peak
intensities measured in typical BST films. As is evident in the graph, the selection of a slowly
varying, minimal intensity angular position for Sr is nearly impossible.
The impact of substrate diffraction on the quantification of a BST film is shown in Fig. 13. In
this example, across-the-wafer measurements were performed on a BST film, however, the
calculated A/B ratios for the same location at the center of the wafer varied considerably when
rotating the wafer 180 degrees. This is depicted in Fig. 13 as the apparent “discontinuity” in A/B
ratio values at the wafer center. The explanation of this effect is apparent when examining the
diffraction effect for this wafer, as shown in Fig. 14. Here the Ti x-ray peak intensity was
measured at the center of the BST/SiO#i <loo> structure as a function of angular rotation. The
diffraction effect on the Ti intensity is readily apparent; however, the repeating pattern, four-fold
symmetry seen in Fig. 11 is not apparent in this plot. After pole figure analysis using XRD, it
was determined that the <loo> test wafer substrate was cut 7 degrees off-axis, which explains
the loss in symmetry in the diffkaction scan shown in Fig. 14. The A/B ratio anomaly at the
wafer center is attributable to the difference in Ti background levels at 0 versus 180 degrees
angular rotation. Note that there is no difference in Ti intensity between the 90 and 270 degree
106
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
-45
0
45
90
135
107
180
225
270
315
Angular rotation
Distance from Wafer Center (mm)
FIGURE 13. Across-wafermeasurementsof A/B
ratio for a BST film exhibiting anomalousbehavior
due to substratediffraction.
FIGURE 14. Measurement of Ti-Ka intensity
versusangular rotation for wafer analyzedin Fig. 13,
showing loss of symmetry in diffraction eEct due to
Si <lOO> substratebeing cut off-axis.
rotation, which explains why the A/B ratio anomaly was only observed in one orthogonal
direction on this wafer.
DISCUSSION
AND CONCLUSIONS
WDXRP has been shown to be a viable method for measurement of the composition and
thickness of BST films. Accuracy and precision have been measured through comparison with
other measurement methods and repeatability studies. Current limitations to the WDXRP
method include substrate diffraction effects, counting statistics, accuracy of the independent
measurement techniques for validation of the standards and accuracy of the quantification
algorithm. The discrepancies between the empirical and fundamental parameters methods are not
fully understood. Accuracy improvements can be achieved through the acquisition of optimized
x-ray detectors for Ba and Sr, a more thorough understanding of the substrate-to-substrate
variability of the diffraction contribution and improved standard validation.
ACKNOWLEDGMENTS
The authors want to acknowledge the efforts of many process technologists within Motorola’s
Semiconductor Technology group who enabled the performance of this work. Special thanks for
the individual contributions from Peir Chu, Beth Baumert, Gary Mote, Steve Primrose and Dan
Sullivan. Motivation for this work came from the Materials Research and Strategic Technologies
group; many thanks to the support of Bradley Mehrick, Robert E. Jones, Jr., Peter Zurcher and
Sherry Gillespie.
107
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
Copyright(C)JCPDS-International Centre for Diffraction Data 2000, Advances in X-ray Analysis, Vol.42
[l]
[2]
C.-J Peng and S. B. Krupanidhi., J. Mater. Res., lo,708726 (1995).
C. S. Hwang, S. 0. Park, H.-J. Cho, C. S. Kang, H.-K. Kang, S. I. Lee and M. Y. Lee, Appl.
Phys. Lett. 67 (1995).
[3] W.-J. Lee, I.-K. Park, G.-E. Jang and H.-G. Kim, Jpn. J. Appl. Phys. 34, 196-199 (1995).
[4] H.J. Cho, C. S. Kang, C. S. Hwang, J.-W. Kim, H. Horii, B. T. Lee, S. I. Lee and M. Y. Lee,
Jpn. J. Appl. Phys. 36, L874-L876 (1997).
[5] Y. Miyaska and S. Matsubara, in Proceedings of 1990 IEEE Seventh Int. Symp. on
Application of Ferroelectrics, edited by S. B. Krupanidhi and S. K. Kurts, pp. 121-124.
[6] T. S. Kim, C. H. Kim, and M. H. Oh, J. Appl. Phys., 75,7998-8003 (1994).
[7] S. Yamamichi, H. Yabuta, T. Sakuma and Y. Miyasaka, Appl. Phys. Lett., 64, 1644-1646
(1994).
[8] R. Tsu, H-Y Liu, W.-Y. HSU, S. Summerfelt, K. Aoki and B. Gnade, Mat. Res. Sot. Symp.
Proc., 361,275-280 (1995).
[9] J. W. Criss and L. S. Birks, Analytical Chemistry, 40, 1080 (1968).
[lo] G. R. Lachance, Canadian Spectroscopy, 15,64-71 (1970).
[ 1 l] H. Kohno, H. Kobayashi, M. Kuraoka and R. Wilson, “Next Generation Semiconductor
Thin Films: Specific Analysis Problems and Solutions,” presented at the 45th Annual
Denver X-Ray Conference, Denver, Colorado, August 3-8, 1996.
108
108