Microtensile Testing of Thin Films in the Optical and
Scanning Electron Microscopes
David T. Read, J. David McColskey, Roy Geiss, and Yi-Wen Cheng
Materials Reliability Division, National Institute of Standards and Technology, Boulder, CO 80305
Abstract. Because thin films are formed by processes different from those used to produce bulk materials, their
microstructures, and hence their mechanical properties, are quite different from those of bulk materials of the same chemical
composition. While the general principles of conventional mechanical testing are applicable to thin films, special test
equipment and techniques are required. These are briefly described here. Present specimen sizes are near the limit of what
can be tested in the optical microscope, so techniques useful in the scanning electron microscope are of interest. Test
techniques adapted for use in the SEM are presented. These test methods have been applied to pure aluminum films
deposited in our laboratory, aluminum films made in a commercial CMOS fab facility, electrodeposited copper, polyimide
films, and polysilicon films. The differences among the stress-strain curves for these very different materials were as
dramatic as would be expected. Now that some experience with these test techniques has been accumulated and the
reproducible results are becoming available, comparisons can be made to expectations based on well-established bulk
behavior. Current unresolved materials-science issues include the "deficit" of the quasi-static apparent Young's modulus
relative to bulk values of some metals, and the generally low elongation to failure found in tensile tests of free-standing
metal films.
INTRODUCTION
Microelectronic devices have taken on critical
functions in communications, aerospace, ground
transportation, and other aspects of the present-day
infrastructure; obviously, assuring the reliability of
microelectronic devices is necessary and important. The
reliability assessment approach has served the industry
well for many years. It is essentially a trial-and-error
approach where complete devices are tested under welldefined, high-stress conditions such as thermal,
humidity, and/or power and voltage cycling. The
realization is now widespread that tests where no
devices fail are not informative, so severe conditions
are generally used. Failure analysis provides feedback
to the next iteration of manufacturing process
refinement. However, this traditional approach has the
shortcoming that complete devices must be assembled
and stress-tested before a reliability assessment can be
made. Hence, the move toward 'predictive reliability' or
Virtual prototyping1 is well established in leading
commercial manufacturers, as well as in early adopters
like aerospace.
Predictive reliability of microelectronic devices has
evolved along the same general lines as predictive
reliability of high-performance mechanical structures
like aircraft and pressure vessels. The inputs are
knowledge of the mechanical design and detailed
configuration of the structures, the physical behavior of
the materials of construction, the service and storage
conditions, and the analytical tools to produce a
prediction of the lifetime of the device. The typical
analysis tool is finite element analysis, supplemented
by fracture mechanics as appropriate. The obvious
differences between large-scale mechanical structures
and microelectronic devices require different methods
for obtaining the input information, and a somewhat
different set of failure mechanisms and failure criteria.
Obtaining relevant characterization data for thin
film interconnect structures requires assurance that the
specimens are representative of the practical structures.
The clear experience to date is that the microstructures
seen in thin films are very different from those in bulk
specimens, and that there are special effects such as the
blocking of dislocation motion by the substrate. So,
microscale testing is required, with the ultimate goal of
testing specimens produced by semiconductor
fabrication equipment and similar in size to features on
integrated circuit chips. The tensile test provides an
unambiguous and useful measure of the mechanical
properties of practically any material. However, it
requires specialized apparatus and the fabrication of
specimens with a specific geometry. Because vapordeposited films are of the order of 1 um thick, the
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
353
FIGURE 1. Free-standing aluminum tensile specimen.
Note the scale bar. The gauge section width is 10 urn.
failure loads are of the order of gram-forces or less, and
the specimens cannot be handled directly. Thus, the
apparatus and the specimen handling techniques are a
significant challenge. An alternate approach,
indentation testing, has been developed and applied to
thin film specimens. Its big advantage is that only a
blanket film is required, rather than a special geometry
freed from its silicon substrate. However, indentation
tests, especially on thin films, require more complicated
interpretation than tensile tests.
FIGURE 2. CMOS aluminum (contact metal) fabricated
through the MOSIS service. The test section is 10 um wide by
190 prn long [1].
SPECIMENS AND TECHNIQUE
We have tested both specimens made in our local
fabrication facilities and specimens made at outside
fabsfl]. For aluminum, copper and polyimide, the
specimen's gauge or test section was approximately 10
\im wide by 190 |xm long. Figure 1 shows an early
homemade
specimen,
electron-beam-evaporated
aluminum. We have since found that a round hole for
the loading pin, as shown in Figure 2, works better.
Thicknesses from 0.3 to 10 \im thick have been tested.
For the thicker specimens, successful lithography
usually requires a wider gauge section. The film to be
tested is deposited on a bare silicon substrate and
patterned lithographically. Then, the substrate is placed
in a vacuum chamber and exposed to xenon difluoride
(XeF2) vapor [2]. The exposed silicon is removed in a
dry-etch process, leaving the specimen free-standing.
Considerable lore about how to carry out this etch has
been accumulated, mostly in conference papers [3]. We
warm the chip before etching, to drive off excess water
vapor, and etch in a pulsed fashion. To avoid
interference with the loading pin, the pit beneath the
specimen should be 60 \im deep or more. For our
etching system, this can be reached within two hours.
For the thinnest specimens, we have had problems with
the grip section tearing out; reinforcement of the grip
section by placing a photoresist overlay on the grip
FIGURE 3. Polysilicon ring-pull microtensile specimen
made at Sandia National Laboratories. The loading pin is
shown engaged in the pull-ring.
section seemed to solve the problem. For polysilicon
specimens made at Sandia National Laboratories,
Figure 3, the gauge sections varied from 50 to 1000 |4,m
long, and were narrower and thicker. The specimen
geometries shown seem to represent the approximate
lower limit for testing under the optical microscope.
Smaller specimens will have to be tested in the
scanning electron microscope (SEM).
We currently use the force-probe tensile test
technique, which is similar to that originally reported
by Greek et al. [4]. The apparatus includes a loading
system, shown schematically in Figure 4, operable
within the SEM or under an optical microscope. Force
is measured using the deflection of the flex strips
shown in Figure 4, and displacements are measured
using digital image correlation of a few hundred images
acquired during the tensile tests. Further development
of this technique to smaller specimens will require a
new force measurement scheme. Testing in the SEM
354
Slope, Elongation
Tensile
strength,
GPa
to failure,
%
MPa
22.5
EB Al [6]
151
28
1.4
74
CMOS Al [1]
35
24
5.5
Polyimide [7]
181
Polysilicon [8]
170
2900
2.1
234
255
79
EDCu
TABLE 1. Tensile properties of various thin film materials.
Material
FIGURE 4. Force-probe apparatus for testing of thin film
tensile specimens, shown in schematic. The apparatus is
mounted within the chamber of a scanning electron
microscope (SEM), or under an optical microscope.
will provide the needed increase in magnification for
imaging.
RESULTS
The particulars of specific thin film tensile specimens
and of experimental microtensile testing of thin films
are still far from routine. Every material, and
sometimes every test, seems to have some peculiarities.
Here only an overview, with a few examples, is
presented; the reader is referred to the original reports
for details. Figure 5 shows a stress-strain curve of the
electron-beam evaporated aluminum as shown in
Figure 1. While the values of the yield and ultimate
strengths for this material, tabulated below with the
other materials discussed, were not surprising, the
substantial elongation to failure was well above typical
results for thin metal films [5]. This film, tested at room
temperature, consistently reached an average strain of
22.5 % before failure. This unusual result is believed to
be related to the extremely small grain size, about 0.35
|im. The question of why thin films tested in tension
0.25
Yield
strengt
h,MPa
94
65
103
show such low elongations to failure, typically around
1 %, is an open issue in the present understanding of
the mechanical behavior of thin films. Besides
elongation to failure, the other outstanding
measurement issue with metal films is their elastic
behavior; specifically, the static Young's modulus has
been reported to be lower for thin films that for the bulk
form of the same material [9]. Measurement of the
static Young's modulus on tiny thin film specimens is
difficult, and errors have made it into the literature in
the past. But the evidence seems to be mounting for a
modulus deficit in these materials. The present results
support the existence of the modulus deficit.
A sufficient amount of data is becoming available on
the mechanical behavior of thin films to begin to
consider the causes of the observed behavior. A useful
tool for investigating failure modes in bulk specimens
is fractography, that is, microscopic observation of the
fracture surface. Few, if any, fractographs of thin film
tensile specimens have been reported, because of the
extreme difficulty of making such observations.
Figure 6 shows a SEM image of the fracture surface
of a CMOS contact metal tensile specimen. A few
pores and a grain boundary with a pore in it are visible
in the micrograph. This film, deposited as part of a
commercial fabrication process, has been subjected to a
typical CMOS heat treatment, likely around 450° C.
The average grain size is 2.75 |im. This fracture surface
0.30
FIGURE 5. Stress-strain curve with two unloading-reloading
cycles for Young's modulus measurements.
355
FIGURE 6. Fracture surface of thin film tensile specimen of
Al 0.5 % Cu deposited on bare silicon (contact metal), made
through the MOSIS service.
Read DT, Cheng Y-W, McColskey JD, Microtensile
Behavior of a Commercial Photodefinable Polyimide, in
Proceedings of the 2002 SEM Annual Conference,
Society for
Experimental
Mechanics, Bethel,
Connecticut:2002, pp 64-67.
LaVan DA, Tsuchiya T, Coles G, Knauss WG, Chasiotis
I, Read DT, Cross Comparison of Direct Strength
Testing Techniques on Polysilicon Films, in ASTM STP
1413: Mechanical Properties of Structural Films,
Muhlstein C, Brown SB, editors. American Society for
Testing
and
Materials,
West
Conshohoken,
Pennsylvania:2001, pp 16-27.
Huang HB, Spaepen F. Tensile testing of free-standing
Cu, Ag and Al thin films and Ag/Cu multilayers, Ada
Materialia 48 (12), pp. 3261-3269,2000.
shows some intriguing features that are very small and
have not yet been interpreted.
Pores, of course, would explain low values of the
Young's modulus, and possibly elongation as well, but
the number of pores visible in Figure 6 is not sufficient
to explain the Young's modulus values listed in Table 1.
DISCUSSION and SUMMARY
Mechanical characterization of thin films is becoming
a sufficiently robust measurement technique that the
materials-science implications of the results can be
considered seriously. Standardization will soon be
useful, and will become more important as external
fabrication facilities are used more and more. Though
low elongation to failure and low Young's moduli are
commonly reported for metal thin films, no general
explanation of these phenomena has been given.
ACKNOWLEDGEMENTS
This work was supported financially by the NIST
Office of Microelectronics Programs, as well as by the
continuing NIST appropriation for Standards, Testing,
and Research. This work is a contribution of the U. S.
government, and is not subject to copyright in the U. S.
A.
REFERENCES
1.
2.
3.
4
5
6
Read DT, Cheng Y-W, McColskey JD, Keller RR,
Mechanical Behavior of Contact Aluminum Alloy, in
Materials Research Society Symposium Proceedings,
Ozkan CS, Freund LB, Cammarata RC, Gao H, editors.
Materials
Research
Society,
Warrendale,
Pennsylvania:2002, pp 263-268.
Dagata JA, Squire DW, Dulcey CS, Hsu DSY, Lin MC.
Chemical Processes Involved in the Etching of Silicon
by Xenon Difluoride, Journal of Vacuum Science &
Technology B 5 (5), pp. 1495-1500, 1987.
Chang FI, Yeh R, Lin G, Chu PB, Hoffman E, Kruglick
EJJ et al., Gas-phase silicon micromachinmg with xenon
difluoride, in Proceedings of the SPIE, Bailey W,
Motamedi ME, Luo F-C, editors.
SPIE Press,
Bellingham,1995, pp 117-128.
Greek S, Ericson F, Johansson S, Schweitz JA. In situ
tensile strength measurement and Weibull analysis of
thick film and thin film micromachined polysilicon
structures, Thin Solid Films 292 (1-2), pp. 247-254,
1997.
Hardwick DA. The Mechanical-Properties of Thin-Films
- A Review, Thin Solid Films 154 (1-2), pp. 109-124,
1987.
Read DT, Cheng YW, Keller RR, McColskey JD.
Tensile properties of free-standing aluminum thin films,
Scripta Materialia 45 (5), pp. 583-589, 2001.
356