LETTERS

Mechanical annealing and source-limited

deformation in submicrometre-diameter
Ni crystals
Z. W. SHAN1,2 , RAJA K. MISHRA3 , S. A. SYED ASIF2 , ODEN L. WARREN2 AND ANDREW M. MINOR1 *
1

National Center for Electron Microscopy, Lawrence Berkeley National Laboratory, Berkeley, California 94720, USA
Hysitron Incorporated, 10025 Valley View Road, Minneapolis, Minnesota 55344, USA
3
General Motors Research and Development Center, Warren, Michigan 48090, USA
* e-mail: [email protected]
2

Published online: 23 December 2007; doi:10.1038/nmat2085

The fundamental processes that govern plasticity and determine

strength in crystalline materials at small length scales have been
studied for over fifty years13 . Recent studies of single-crystal
metallic pillars with diameters of a few tens of micrometres or
less have clearly demonstrated that the strengths of these pillars
increase as their diameters decrease47 , leading to attempts to
augment existing ideas about pronounced size effects8,9 with new
models and simulations1017 . Through in situ nanocompression
experiments inside a transmission electron microscope we can
directly observe the deformation of these pillar structures and
correlate the measured stress values with discrete plastic events.
Our experiments show that submicrometre nickel crystals
microfabricated into pillar structures contain a high density
of initial defects after processing but can be made dislocation
free by applying purely mechanical stress. This phenomenon,
termed mechanical annealing, leads to clear evidence of sourcelimited deformation where atypical hardening occurs through
the progressive activation and exhaustion of dislocation sources.
In 1952, Herring and Galt3 reported that the measured
tensile strength of Sn whiskers (filamentary crystals a few
micrometres in diameter and a few millimetres in length) may
approach the theoretical value for a perfect Sn lattice. Thereafter,
the relationship between the size of crystalline materials and
their mechanical properties has been the subject of numerous
experimental studies2,18 and scientific debates1,8,9 . The general
trend found throughout these studies is that the strength of the
whiskers increases with decreasing diameter1 . The recent advent of
compression tests on microfabricated pillar structures4,5 now makes
it possible to probe the size-dependent strength of materials in a
uniaxial fashion without requiring crystal growth in whisker form.
The results from these elegant mechanical tests are consistent with
the tenet that smaller is stronger, even for diameters larger than
those of the whiskers.
To account for the high strengths of microfabricated metallic
pillars, it has been proposed10,11 that mobile dislocations escape
from the crystal at the nearby free surfaces before multiplying
and interacting with other dislocations during the deformation
processthus leading to a state of dislocation starvation.
Presumably, high stresses would then be expected because new
dislocations would have to be nucleated for plasticity to proceed.
Although providing strong evidence for size effects in metals,

these ex situ tests2,4,7,18 do not permit a one-to-one relationship

between the mechanical data and the microstructure evolution.
In this study, we have carried out direct-observation, in situ
experiments on microfabricated Ni pillars by using a unique
instrument that enables quantitative nanoscale compression tests
inside a transmission electron microscope (TEM). In addition,
this technique enables us to examine potential factors such as the
surface oxide19 , pillar sink-in6 and the initial microstructure.
Single-crystal Ni pillars between 150 and 400 nm in diameter
were micromachined using an FEI 235 Dual Beam focused ion
beam (FIB; for more sample preparation details see Supplementary
Information). TEM observations showed that the pillars had a
high density of defects after FIB processing (Fig. 1a), and that
these defects could be classified into two types: small, loop-like
defects and long line defects often extending across the pillar.
Presumably, the former are dislocation loops resulting from the ion
beam irradiation and the latter are pre-existing dislocations in the
bulk single-crystal Ni. Counting both types of defect, the starting
dislocation density of the pillars was 1015 m2 .
In situ compression tests involving a diamond flat punch were
carried out on pillars of several different sizes. These tests were
executed either under loading rate control (micronewtons per
second) or displacement rate control (nanometres per second),
the details of which can be found in Supplementary Information.
Remarkably, the dislocation density always decreased markedly
during the deformation test and in some cases a dislocation-free
pillar was the end result. One typical example of this phenomenon,
which we denote as mechanical annealing, is shown in Fig. 1,
which illustrates two consecutive compression tests run on the
same pillar under loading rate control. The starting pillar geometry
is 160 nm in diameter at the free end, 850 nm in length and
had a sidewall taper angle of 4.5 . Figure 1a shows the starting
microstructure and Fig. 1d shows the force versus displacement
curve of the first test. After the initial test, the main section of the
pillar was left completely dislocation free (Fig. 1b). Examination
of the recorded video (see the Supplementary Information)
revealed that the pillar yielded immediately on contact with the
punch. During the initial portion of the test, the pre-existing
dislocations progressively left the pillar and were accompanied
by intermittent bursts of new dislocations, which correspondingly
caused the load to fluctuate (Fig. 1d). The sharp increase in load

nature materials VOL 7 FEBRUARY 2008 www.nature.com/naturematerials

2008 Nature Publishing Group

115

LETTERS
a

b
g = [111]
[111]

ZA = [110]

50

80

Force (N)

40
Force (N)

200 nm

200 nm

30
20

70

1.2

60

1.0

50
40
30

100
150
50
Displacement (nm)

200

0.8
0.6

0.2

10
0

Test 1
Test 2

0.4

20

10

1.4

Stress (GPa)

200 nm

0
0

50 100 150 200 250 300 350 400

Displacement (nm)

50 100 150 200 250 300 350 400

Displacement (nm)

Figure 1 Two consecutive in situ TEM compression tests on a FIB microfabricated 160-nm-top-diameter Ni pillar with h111i orientation. a, Dark-field TEM image of
the pillar before the tests; note the high initial dislocation density. b, Dark-field TEM image of the same pillar after the first test; the pillar is now free of dislocations.
c, Dark-field TEM image of the same pillar after the second test. d,e, Force versus displacement curve of the first (d) and second (e) test. f, Instantaneous stress versus
condition, zone axis
compressive displacement for the two tests; the apparent yield stress is similar for both tests. All dark-field images are shown in a g = [111]

( ZA) = [110].

starting at 170 nm (Fig. 1d) coincided with predominantly elastic

behaviour. Post-test, dark-field observation under multiple twobeam conditions confirmed that this test left behind a dislocationfree pillar (Fig. 1b).
The pronounced decrease in dislocation density during the
nanocompression test provides direct experimental support for the
dislocation starvation mechanism that has been hypothesized on
the basis of experiments10,11 and simulations17,20 . Neither the surface
oxide layer19 nor the FIB damage layer21 trapped dislocations inside
the pillar during deformation. Presumably, the driving force for the
escape of dislocations within the pillars is a combination of the
applied stress and the image forces from the surface22 . The former
will activate or nucleate dislocations and the latter will assist the
dislocations in moving towards the free surfaces of the crystal.
The extent to which both pre-existing and newly generated
dislocations ran out of the pillar was unexpected, and presented
an opportunity to examine the strength of a dislocation-free pillar.
Figure 1e shows the forcedisplacement curve of the second test
on the same pillar. The mechanical response was predominantly
elastic for the initial 20 nm. Discrete plasticity then occurred,
demonstrated by a series of discontinuities in the curve. After
the compressive displacement reached 140 nm, the load steadily
decreased over the next 185 nm. The video of this test clearly
showed that this steady decrease in force was due to buckling of the
pillar. The load rise following this buckling (starting at 325 nm)
occurred as the punch came into contact with material at the base
of the pillar (Fig. 1c).
One advantage of in situ testing is that the instantaneous
contact diameter can be measured from the still frames extracted
from the recorded movies (30 frames s1 ). This makes it possible
to determine an instantaneous contact stress imposed on the pillar

(force divided by the instantaneous contact area), which is more

accurate than a simple engineering stress (force divided by the
initial area of the top of the pillar), providing the assumption of
symmetrical deformation holds true. For the test shown in Fig. 1,
this assumption is reasonable, and the instantaneous stress versus
the compressive displacement for the two tests corresponding to
Fig. 1d and e are plotted together in Fig. 1f. Interestingly, despite
the approximately 15 orders of magnitude difference in starting
dislocation density, the apparent yield stress achieved in both tests
is quite similar. Therefore, the initial defects due to FIB processing
did not significantly affect the stress response of the pillar. This
suggests that the deformation of the pillar is controlled instead by
the nucleation of dislocations and their propagation through the
pillar. The low impact of initial microstructure on the stressstrain
curve is probably related to the fact that defects created by ion
beams typically do not extend much beyond the penetration depth
of the ions, which for 30 keV Ga+ in Ni is only 1020 nm even for
large incident angles23 .
However, owing to the tapered geometry of the microfabricated
pillars, the stress throughout the pillars is not homogeneous
and the simple analysis used for calculating the instantaneous
stress does not fully capture the complexity of the situation. A
recent computational study has reported that a taper angle to
the sidewalls of microfabricated pillar structures can result in an
overestimation of the elastic modulus and the apparent yield stress
during a microcompression test24 . Given that the microfabricated
pillar shown in Fig. 1 had a sidewall taper angle of 4.5 , it is
expected that the top of the pillar experienced a larger imposed
stress during compression, which might result in inhomogeneous
plastic deformation localized at the top of the pillar. This is in
fact what happened and Fig. 1b shows a remnant of this localized

116

nature materials VOL 7 FEBRUARY 2008 www.nature.com/naturematerials

2008 Nature Publishing Group

LETTERS
a

1.0

0.9

221 nm

Normalized stress

0.8

221 nm

0.7
0.6
0.5

Pillar free-end
diameter

0.4

150 nm (Fig. 2b)

160 nm (Fig. 1a)
290 nm (Fig. 3a)
2,000 nm

0.3
0.2

200 400 600 800 1,000

Distance from the free end (nm)

Figure 2 Aspects of taper leading to localized deformation. a, TEM image of a pillar before compression with a sidewall taper angle of 4.5 , a free-end diameter of
150 nm and a length of 800 nm. b, TEM image of the same pillar shown in a, after the compression test. Plastic deformation is found to concentrate on the part above
the red line, which is in the same location in both images. Note that the pillar is left almost dislocation free after the compression test, even below the red line. c, Normalized
initial stress (stress in the limit of infinitesimally small strain divided by the initial stress at the free end) versus axial distance from the free-end surface for pillars with
different diameters and a fixed taper angle of 4.5 .

deformation at the top of the pillar. In addition, the taper angle has
decreased almost to zero after the initial test, consistent with there
being greater deformation in regions of higher stress.
Clear evidence of inhomogeneous deformation in a second
pillar, with an initial geometry of 150 nm in free-end diameter,
800 nm in length and 4.5 in sidewall taper angle, is shown
in Fig. 2a,b. In this case, a change in diameter was not detected
below 360 nm from the original free end of the pillar (below
the red line in Fig. 2a,b) even after 130 nm of compressive
displacement. Yet, as Fig. 2b shows, the pre-existing dislocations
below this point were driven out of the pillar and the entire
length of the pillar was left almost dislocation free. Thus, a tapered
geometry in the investigated size regime makes it necessary to
determine local stresses and strains to fully quantify the observed
deformation behaviour.
To graphically illustrate how the presence of a taper can
lead to localized deformation, a purely geometric relationship
describing the normalized initial stress distribution (that is, the
stress distribution in the limit of infinitesimally small strain divided
by the initial stress at the free end) for pillars with different
diameters and a fixed taper angle of 4.5 is plotted in Fig. 2c in
terms of the axial distance from the free-end surface. For a set
of pillars having a fixed taper angle and a constant aspect ratio
(length divided by a consistent measure of diameter such as the
free-end diameter), the ratio of the initial stress at the free end
to that at the base would be independent of diameter on account
of geometric self-similarity. Although the fabrication method used
in this study yielded a set of pillars having a fixed taper angle, the
aspect ratio varied from 3:1 to 5:1 and this variation was dictated
primarily by differences in diameter rather than in length. As can
be seen in Fig. 2c, the decay in initial stress with increasing axial
distance from the free-end surface becomes more pronounced as
the diameter decreases. Therefore, a set of pillars having a fixed
taper angle and a constant length should exhibit increasingly more
localized deformation as the diameter decreases. We found that this
trend held true over the range of diameters in this study.
The extent of mechanical annealing is affected by the diameter
of the pillars, and is consistent with the tenet of smaller being
stronger. We observed that larger-diameter pillars were less likely
to be dislocation free at the end of the test. This observation is in

accordance with the image forces contributing to the annihilation

of dislocations at the surface, because this effect would decrease
with an increase in sample dimension22 . Figure 3 shows results
from a larger pillar. In this case, a pillar 290 nm in diameter
at its free end, 1 m in length and with a taper angle of
4.5 (Fig. 3a) was compressed under displacement rate control.
Comparison of the pillar before (Fig. 3a) and after (Fig. 3b) the
test revealed that most of the dislocations had been driven out
of the pillar, but not as completely as in the previous examples
concerning smaller-diameter pillars (Figs 1b and 2b). Figure 3c
shows a plot of force and displacement versus time of the test for
the pillar shown in Fig. 3a. Arrows superimposed on the plot mark
six succinctly chosen points of the data. The extremely irregular
appearance of the curve is due to the discrete nature of dislocation
nucleation and motion25 . The corresponding instantaneous stress
versus displacement curve is shown in Fig. 3d with the same points
transferred from Fig. 3c. Figure 3ej shows the microstructure at
each of the six points. The punch contacted the pillar at point 1
and then from point 2 to point 3 a serrated load/stress plateau was
measured. The video of this test revealed that the deformation of
the pillar during the serrated plateau was the result of the activation
of a single {111} slip system, which can be seen clearly in the video
as the crystal shearing against itself. The source indicated in Fig. 3f
became exhausted at point 3 (corresponding to Fig. 3g). Between
points 3 and 4, the stress built up to reach a high value of 2.6 GPa,
even though the pillar still possessed a high dislocation density
(Fig. 3h). This is consistent with our recent in situ nanoindentation
study on submicrometre Al grains, which demonstrated that high
stresses can be achieved even in metallic grains possessing a high
dislocation density26 . At point 4 in Fig. 3c, the pillar again yielded
but this time in a manner corresponding both to plasticity within
the pillar itself and to movement of the pillar as a whole into its
substrate (by more than 20 nm). Many trace lines not present before
this yield event (Fig. 3h) appeared along the [21 1] direction within
a time interval of no more than 1/30 s (Fig. 3i). The sudden sinkin of the pillar into the substrate can be thought of as a hard spike
punching into relatively soft ground, but in this case the spike
and the ground are of the same material and the difference in
hardness is purely a size effect. The sudden sink-in event caused
the punch to lose contact with the free end of the pillar for 1 s.

nature materials VOL 7 FEBRUARY 2008 www.nature.com/naturematerials

2008 Nature Publishing Group

117

LETTERS
a

Compression

200 nm

150

200 nm

3.0

200

100

50
1

50

10

2.0

4
2

1.5
1.0
1

0
20

15

Time (s)

150 nm

[111]

5(6)

0.5

Stress (GPa)

150

100

Displacement (nm)

Force (N)

2.5

50

3
100
150
Displacement (nm)

200

150 nm

150 nm

150 nm

150 nm

g = [111]

ZA = [110]

150 nm

[211]

Figure 3 Direct evidence of source-limited deformation. a, Bright-field TEM image of the pillar before deformation (290 nm free-end diameter). b, TEM image of the
same pillar after the displacement rate control test. Note the trapped dislocations. c, Force and displacement versus time of the test. d, Instantaneous stress versus
displacement of the test. ej, Frames extracted from the video that correspond to the microstructure of the pillar at the instances marked as 16 in c and d, respectively. All
condition, zone axis ( ZA) = [110].

images are shown in a g = [111]

The load rise after point 5 resulted from the punch re-contacting
the pillar, and the subsequent load reduction leading to point 6
was the consequence of withdrawing the punch at the end of the
test. Figure 3j corresponds to the microstructure immediately at the
end of the experiment. The observation that a high stress (2.6 GPa)
was achieved even in the presence of a high dislocation density is
consistent with the theory that dislocation source starvation is the
critical factor in determining strength at these small scales.
Traditionally, strain hardening is associated with strong
interactions between dislocations and an increase in dislocation
density throughout an experiment. In our case, however, the

increases in stress levels (ignoring the fluctuations) occur

more in a stepwise rather than a continuous fashion and
can be ascribed to the progressive exhaustion of dislocation
sources, as is seen in Fig. 3. The underlying physical mechanism
can be understood in terms of a competition between the
dislocation nucleation/activation rate and the (mobile) dislocation
annihilation rate. If there are enough mobile dislocations or
a productive enough dislocation source to accommodate the
imposed deformation, a stress/load drop will occur. However, if
there are not enough active sources or dislocations to accommodate
the imposed deformation, the stress/load will increase. Considering

118

nature materials VOL 7 FEBRUARY 2008 www.nature.com/naturematerials

2008 Nature Publishing Group

LETTERS
the large difference in stiffness between diamond (the punch
material) and nickel as well as the inevitable roughness on the
contact surfaces, the contact interface will serve to generate easy
dislocation sources at the initial deformation stage, and this was
observed in all of our experiments. However, eventually these
sources become exhausted and a hardening response (stress level
increase) is seen during the compression test (for example, Figs 1f
and 3d).
Our in situ, quantitative nanoscale compression tests have
demonstrated that high-resolution mechanical data can be
directly correlated with dynamic microstructural evolution in
submicrometre-diameter pillars. The extent to which the pillars
were found to anneal through mechanical deformation was
surprising. This led to the possibility of studying dislocation-free
volumes and also gave rise to atypical strain hardening where
deformation was controlled by the progressive activation and
subsequent exhaustion of dislocation sources. The observation that
the dislocation density can fall to zero and that deformation takes
place at high stresses is consistent with previous hypotheses that the
strength increases in small pillar structures are controlled by the
activation of new dislocation sources in a source-limited regime.
As a whole, this study is highly encouraging from the perspective
of bridging the gap between experimentation and computational
plasticity models.
Received 23 July 2007; accepted 19 November 2007; published 23 December 2007.

7. Greer, J. R. & Nix, W. D. Size dependence of mechanical properties of gold at the sub-micron scale.
Appl. Phys. A 80, 16251629 (2005).
8. Nix, W. D. Yielding and strain hardening of thin metal films on substrates. Scr. Mater. 39,
545554 (1998).
9. Nix, W. D. Mechanical properties of thin films. Metall. Trans. A 20A, 22172245 (1989).
10. Greer, J. R. & Nix, W. D. Nanoscale gold pillars strengthened through dislocation starvation. Phys.
Rev. B 73, 6 (2006).
11. Nix, W. D., Greer, J. R., Feng, G. & Lilleodden, E. T. Deformation at the nanometer and micrometer
length scales: Effects of strain gradients and dislocation starvation. Thin Solid Films 515,
31523157 (2007).
12. Parthasarathy, T. A., Rao, S. I., Dimiduk, D. M., Uchic, M. D. & Trinkle, D. R. Contribution to size
effect of yield strength from the stochastics of dislocation source lengths in finite samples. Scr. Mater.
56, 313316 (2007).
13. Sieradzki, K., Rinaldi, A., Friesen, C. & Peralta, P. Length scales in crystal plasticity. Acta Mater. 54,
45334538 (2006).
14. Guo, Y., Zhuang, Z., Li, X. Y. & Chen, Z. An investigation of the combined size and rate effects on the
mechanical responses of FCC metals. Int. J. Solids Struct. 44, 11801195 (2007).
15. Rabkin, E. & Srolovitz, D. J. Onset of plasticity in gold nanopillar compression. Nano Lett. 7,
101107 (2007).
16. Rabkin, E., Nam, H. S. & Srolovitz, D. J. Atomistic simulation of the deformation of gold nanopillars.
Acta Mater. 55, 20852099 (2007).
17. Tang, H., Schwarz, K. W. & Espinosa, H. D. Dislocation escape-related size effects in single-crystal
micropillars under uniaxial compression. Acta Mater. 55, 16071616 (2007).
18. Brenner, S. S. Plastic deformation of copper and silver whiskers. J. Appl. Phys. 28, 1023 (1957).
19. Uchic, M. D., Dimiduk, D. M., Florando, J. N. & Nix, W. D. Oxide surface films on metal
crystalsResponse. Science 306, 11341135 (2004).
20. Wei, Q. H. & Wu, X. L. Grain boundary dynamics under mechanical annealing in two-dimensional
colloids. Phys. Rev. E 70, 4 (2004).
21. Kiener, D., Motz, C., Rester, M., Jenko, M. & Dehm, G. FIB damage of Cu and possible consequences
for miniaturized mechanical tests. Mater. Sci. Eng. A 459, 262272 (2007).
22. Weertman, J. & Weertman, J. R. Elementary Dislocation Theory (Oxford Univ. Press, New York, 1992).
23. Ziegler, J. F., Biersach, J. P. & Littmark, U. The Stopping and Range of Ions in Solids, Stopping and
Range of Ions in Matter Vol. 1 (Pergamon, New York, 1985).
24. Zhang, H., Schuster, B. E., Wei, Q. & Ramesh, K. T. The design of accurate micro-compression
experiments. Scr. Mater. 54, 181186 (2006).
25. Koslowski, M. Scaling laws in plastic deformation. Phil. Mag. 87, 11751184 (2007).
26. Minor, A. M. et al. A new view of the onset of plasticity during the nanoindentation of aluminium.
Nature Mater. 5, 697702 (2006).

References

Acknowledgements

1. Brenner, S. S. Growth and properties of whiskers. Science 128, 568575 (1958).

2. Brenner, S. S. Tensile strength of whiskers. J. Appl. Phys. 27, 14841491 (1956).
3. Herring, C. & Galt, J. K. Elastic and plastic properties of very small metal specimens. Phys. Rev. 85,
10601061 (1952).
4. Uchic, M. D., Dimiduk, D. M., Florando, J. N. & Nix, W. D. Sample dimensions influence strength
and crystal plasticity. Science 305, 986989 (2004).
5. Dimiduk, D. M., Uchic, M. D. & Parthasarathy, T. A. Size-affected single-slip behavior of pure nickel
microcrystals. Acta Mater. 53, 40654077 (2005).
6. Volkert, C. A. & Lilleodden, E. T. Size effects in the deformation of sub-micron Au columns. Phil.
Mag. 86, 55675579 (2006).

Research performed at the National Center for Electron Microscopy, Lawrence Berkeley National
Laboratory, was supported by the Scientific User Facilities Division of the Office of Basic Energy
Sciences, US Department of Energy under Contract No. DE-AC02-05CH11231. This work was also
supported by an SBIR Phase II grant DE-FG02-04ER83979 awarded to Hysitron, which does not
constitute an endorsement by DOE of the views expressed in this article. Chris Gilde is thanked for his
assistance with video editing.
Correspondence and requests for materials should be addressed to A.M.M.
Supplementary Information accompanies this paper on www.nature.com/naturematerials.
Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

nature materials VOL 7 FEBRUARY 2008 www.nature.com/naturematerials

2008 Nature Publishing Group

119