1. 1
Preparation of Electron Transparent Samples for Transmission
Electron Imaging
Michael Berkson, Katayun Barmak, Xiwen Chen
Columbia University
The size effect in Cu has prompted interest in alternate metals for sub-30nm linewidths.
The criteria for alternate metals should include: (a) low resistivity, (b) minimal resistivity
size effect, (c) elimination of diffusion barrier, and (d) high melting temperature.
Electroless-deposited Ni, Co, Ru, and Pt have been identified as candidates for further
investigation. Exploratory characterization of electroless-deposited metal films using
transmission electron imaging and diffraction methods will be conducted. Comparing these
results with results for physically deposited films will help guide interconnect metal
process development. In order to image and analyzing materials using TEM, samples must
be thin enough to be electron transparent. In this project, a method for preparing electron-
transparent solid crystalline TEM samples is developed and practiced.
Introduction
The electron diffraction based metrology me-
thods developed by Profs. Barmak and Coffey
make use of crystal orientation maps obtained by
precession electron diffraction (PED) in the
transmission electron microscope (TEM) using
the ASTAR™ system. The orientation maps can
be used to obtain:
1. Average grain size and grain size
distribution.
2. Orientation distribution (OD), including
fiber fractions, i.e., fraction of grains
orientated with a give crystal axis
parallel to the film normal).
3. Misorientation distribution (MD), i.e.,
fraction of boundaries with a given
misorientation across the boundary.
4. Grain boundary character distribution
(GBCD), i.e., length fraction (in 2D) or
area fraction (in 3D) of boundaries with
a given misorientation and boundary
normal.
The size effect in Cu has prompted interest in
alternate metals for sub-30nm linewidths. The
criteria for alternate metals should include:
1. Low resistivity,
2. Minimal resistivity size effect,
3. Elimination of diffusion barrier, and
4. High melting temperature.
Electroless-deposited Ni, Co, Ru, and Pt have
been identified as candidates for further
investigation. Exploratory characterization of
electroless-deposited metal films using trans-
mission electron imaging and diffraction methods
will be conducted. Comparing these results with
results for physically deposited films will help
guide interconnect metal process development.
The anticipated results are:
 Microstructural characterization of
electroless deposited metal films
provided by Lam Corporation using
transmission electron imaging and
diffraction methods.
 Comparison of the results with results
for physically vapor deposited (PVD)
films
In order for us to use TEM to characterize these
films, which are deposited on Si wafers hundreds
of microns thick, we must thin our samples
enough to make them electron transparent. This
is a necessary step to obtain ASTAR™ images
from TEM for analysis with ASTAR™.
However, usually only a paragraph of such a
publication is devoted to the TEM sample
preparation procedure. In this report, we will
describe in more detail the process we used to
prepare the deposited thin-film samples for
TEM.

2. 2
Detailed TEM sample preparation procedures
have been published by Pakzad [1], Yao [2], and
others. Our approach will closely follow [1],
which consists of several mechanical thinning
methods and no chemical etching.
Experimental Methods
This TEM sample preparation procedure was
developed and practiced on 3-inch Si wafers with
Fe80Ni20 film deposited via sputtering.
Cutting
Si wafers with deposited thin films are cut face-
down with a diamond scribe using a glass slide
as a straightedge. The wafer is braced against a
ruler and cut into 1.5mm-square pieces.
Mounting
A square sample is mounted to a circular Cu grid
of 3mm in diameter and 1mm diameter central
circular opening using M-bond epoxy, with the
thin film facing the grid. This sample-grid
assembly is then fixed with wax to a Pyrex glass
stub whose size is compatible with the grinding
tools described below.
Grinding and Polishing
Mounted samples were ground using a South
Bay Technology Model 900 Grinder/Polisher lap
wheel (Figure 1) and Buehler SiC grinding
papers with particle sizes denoted 800 grit
(8.4µm) and 1200 grit (6.5µm). Samples were
ground to a target silicon thickness of 90µm.
Figure 1. South Bay Technology lap wheel, used for
grinding and polishing.
Samples were then polished with circular cloth
compatible with the lap wheel and 1µm diamond
paste manufactured by Buehler.
During the grinding and polishing process, stubs
containing the samples were held in place by a
tripod (Figure 2).
Figure 2. Tripod used with lap wheel.
Dimple Grinding
Samples were then thinned to a target silicon
thickness of 7µm using a Gatan Model 656
Dimple Grinder (Figure 3). This device grinds a
crater into the sample to thin it further while
preserving its structural integrity to prevent
fracture. The dimple grinder uses a bronze
grinding wheel with 1µm diamond paste as an
abrasive, followed by a felt wheel with colloidal
silica suspension.
Figure 3. Dimple grinder.
Ion Milling
Finally, samples are thinned to under 100nm in
the center using the Gatan PIPS II ion milling
machine, which mills the sample by bombarding
it with ionized argon gas. Since we have no way
of measuring sample thickness directly during
this step, our criterion for sufficient thinness is

3. 3
the appearance of an optical interference pattern,
shown in Figure 4.
Figure 4. Optical micrograph showing diffraction pattern
on ion-milled sample.
Results and Discussion
Originally, mounting the grid on the sample was
done between dimple grinding and ion milling.
Mounting at that point in the process proved
difficult because epoxy was too often smeared
onto the middle of the sample, interfering with
electron transparency, or mounting wax got in
the way and prevented the epoxy from bonding
the grid and the sample together.
By mounting the sample on the grid first, a
change borrowed from the doctoral thesis of G.
Lucadamo [3], the sample could maintain its
structural integrity throughout the process with
the additional support from the grid, and the
difficult mounting step later in the procedure was
eliminated.
During the time when the original procedure was
conducted, we attempted to prepare 8 plan-view
samples, none of which made it past the post-
dimple grinding grid-mounting step. With the
procedure described in this report, we prepared 5
plan-view samples in 6 attempts, and the one we
examined using TEM was sufficiently electron
transparent to produce an image (Figure 5).
The Moire patterns seem in the image of Fig. 5
suggest that the film has more than one layer of
grains through the thickness. We attribute this
grain structure to the sample itself and not to the
preparation procedure.
Figure 5. Transmission electron micrograph (top) and
electron diffraction pattern (bottom) for plan-view
permalloy thin film sample.
Conclusions
The sample preparation procedure described in
this report, which involves mounting the sample
on a Cu grid before grinding and polishing, rather
than after dimple grinding, is an effective method
for preparing thin film samples for TEM. The
two-layer grain structure in the sample examined
is more likely to be a product of the sputtering
process than of the TEM preparation process.

4. 4
References
1. Pakzad, A.; Granz, S.; Wise, A. Ready, set,
go! Ensuring an identical TEM sample
preparation route again and again. Gatan.
http://www.gatan.com/ready-set-go
-ensuring-identical-tem-specimen
-preparation-route-again-and-again
(accessed June 2, 2016).
2. Yao, B.; Petrova, R.; Vanfleet, R.; Coffey, K.
J. Electron Microsc. 2006, 55(4), 209–214.
3. Lucadamo, G. Ph.D. Dissertation, Lehigh
University, 2000.
Acknowledgments
Funds for this research were provided by the
Materials Research Science and Engineering
Center (MRSEC) and the NSF Engineering
Research Center grant DMR-1420634. Thank
you to Prof. Katayun Barmak for your guidance
throughout the project and for the language in the
introduction, and to Xiwen Chen for your help
with the sample preparation procedure as well as
several of the figures. And thank you to Jiaxing
Liu for the thin film samples.
Michael Berkson is a senior at
Columbia University studying
Materials Science. He is
continuing his studies next year
in the MS program in Materials
Science at Columbia, and he hopes to pursue a
multidisciplinary engineering career.