Scripta Materialia 53 (2005) 177–182

www.actamat-journals.com

Bauschinger effect in thin metal films

Y. Xiang, J.J. Vlassak *

Division of Engineering and Applied Sciences, Harvard University, 29 Oxford Street, Cambridge, MA 02138-2901, USA

Received 28 February 2005; received in revised form 18 March 2005; accepted 25 March 2005
Available online 25 April 2005

Abstract

The Bauschinger effect in thin sputter-deposited Al and Cu films is studied by isothermally deforming the films alternately in
tension and compression. Passivated films exhibit an unusual Bauschinger effect with reverse flow already occurring on unloading,
while unpassivated films show little or no reverse flows when the film is fully unloaded.
Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Bauschinger effect; Thin films; Compression test; Plastic deformation; Dislocations

1. Introduction with superior mechanical behavior. Many experimental

It is well known that the mechanical response of a
metallic material depends not only on its current stress
state but also on its deformation history. One of the
most important examples is the observation that after
a metal is deformed plastically in one direction, the yield
stress in the reverse direction is often lower. Fig. 1 sche-
matically shows a typical stress–strain curve for metallic
materials. The stress rf is the forward flow stress, and rr
at the start of reverse plastic flow is the reverse flow
stress. If rr is equal to rf, the material hardens isotrop-
ically. For many metals, however, the reverse flow stress
is found to be lower than the forward flow stress. This
anisotropic flow behavior was first reported by Bausch-
inger [1] and is referred to as the Bauschinger effect. The
loss of strength due to the Bauschinger effect is of prac-
tical importance since the strength of a metal part may
be impaired if the working stress acts in the reverse
direction compared to the manufacturing stress. Fur-
thermore, a good understanding of the physical origin
of the Bauschinger effect may lead to more refined plas-
ticity theories and may ultimately result in materials
*
Corresponding author. Tel.: +1 617 496 0424; fax: +1 617 495 3897.
E-mail address: vlassak@esag.deas.harvard.edu (J.J. Vlassak).
and theoretical efforts have been devoted to studying
the Bauschinger effect in bulk metals since the phenom-
enon was first reported [2–15]. The physical origins are
generally ascribed to either long-range effects, such as
internal stresses due to dislocation interactions [6,7], dis-
location pile-ups at grain boundaries [8,9] or Orowan
loops around strong precipitates [10–14], or to short-
range effects, such as the directionality of mobile dislo-
cations in their resistance to motion or annihilation of
the dislocations during reverse straining [10]. Satisfac-
tory agreement has been achieved between models and
experimental results obtained in various bulk metals
and alloys [3–5,7–11].
Thin metal films are widely used in many advanced
devices across a wide range of industries. Reliability
problems encountered in these applications have moti-
vated a strong interest in the mechanical behavior of
thin films [16,17]. Besides this technological driving
force, thin films also provide a unique opportunity to
investigate fundamental problems in materials science.
For example, at least one dimension of a thin film is
comparable to the characteristic length scales associated
with material defects such as dislocations; thus, free sur-
faces and interfaces are expected to play an important
role in the mechanical behavior of thin films [18].

1359-6462/$ - see front matter Ó 2005 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.scriptamat.2005.03.048
178 Y. Xiang, J.J. Vlassak / Scripta Materialia 53 (2005) 177–182

inger effect in thin metal film and has been applied to

thin sputter-deposited Al and Cu films. The results pro-
vide for the first time unambiguous experimental evi-
dence of a strong Bauschinger effect in thin metal films.

2. Experiment and results

The new experimental method is based on the plane-

strain bulge test technique [26,27]. In this technique, the
film of interest is deposited on a Si wafer and long rect-
angular membranes are fabricated using standard
micromachining technology. Fig. 2(a) shows a perspec-
Fig. 1. A schematic of the typical stress–strain curve of a metallic tive view of a typical bulge test sample. The as-prepared
material that exhibits the Bauschinger effect.
membrane is initially flat and under tension. It is then
deflected by applying a uniform pressure to one side
Furthermore, as a result of the special manufacturing causing a state of plane strain in the film. The applied
techniques and the small materials dimensions, films pressure, p, and corresponding membrane deflection,
often have unique microstructures. For example, thin h, are measured and converted to a stress–strain curve
films often have a columnar grain structure with an using the following two simple formulae [26,27]:
average grain size that is much smaller than in bulk
pa2 2h2
materials; they are frequently highly textured [19]. As r¼ and e ¼ e0 þ ; ð1Þ
a result of these dimensional and microstructural con- 2ht 3a2
straints, many materials behave mechanically very dif- where t is the film thickness, 2a the membrane window
ferently in thin film form than in the bulk, especially width, as shown in Fig. 2(a), and e0 the residual strain
in the plastic regime [18]. Various theoretical models
and numerical simulations have been proposed recently
to describe thin-film plasticity including strain-gradient
plasticity theories [20,21], crystal plasticity theories
[22], and discrete dislocation simulations [23,24]. The
plasticity theories and the discrete dislocation models
explain the strengthening effects associated with the film
thickness and microstructure reasonably well; they pre-
dict, however, very different behavior in reverse loading.
In the discrete dislocation simulations [22,25] and some
crystal plasticity theories [25], passivated films show a
distinct Bauschinger effect upon unloading after plastic
pre-straining in tension. Reverse plastic flow starts early
even though the overall stress in the film is still in ten-
sion. This type of Bauschinger behavior is not predicted
in other models [20–22] and is also very different from
that typically found in bulk materials. Up to date, how-
ever, there has been no direct experimental evidence of
such a Bauschinger effect in thin metal films.
The most common experimental technique for reveal-
ing the Bauschinger effect in bulk material is cyclic or
unidirectional testing where pre-straining in tension is
followed by reverse loading in compression. This meth-
od cannot be directly applied to thin films because any
compressive stress in the plane of a freestanding thin
film causes it to buckle due to the large lateral dimen-
sion/thickness ratio in thin films. In this article, we re-
port on a new experimental technique to deform thin
metal films alternately in tension and compression and Fig. 2. A schematic of the compression test technique for thin metal
to measure the corresponding stress–strain curves. This films: (a) the plane-strain bulge test technique; (b) schematic of the
technique allows us to quantitatively study the Bausch- stress–strain curve for each layer of the composite film.
 Y. Xiang, J.J. Vlassak / Scripta Materialia 53 (2005) 177–182 179

in the film, which is equal to the residual stress divided sure required to deflect a freestanding Si3N4 membrane
by the biaxial modulus of the film. Fig. 2(b) illustrates to the deflection h. According to Eq. (5), p2(h) can be ob-
how this technique can be modified to test films in com- tained by subtracting the pressure–deflection curve of a
pression. A bilayer membrane is microfabricated con- freestanding Si3N4 film from that of the bilayer. The
sisting of the metal film of interest on a thin ceramic stress–strain curve of the metal film, f2, can be calculated
layer such as Si3N4 that has a much higher elastic mod- from p2(h) using Eq. (1). This method is generally appli-
ulus than the metal film. During the test, the metal film cable as long as the constitutive equation of layer 1, f1,
first flows in tension, while the Si3N4 layer only deforms does not change when layer 2 is removed.
elastically. Upon unloading, the tensile stress in the The detailed sample preparation method is discussed
Si3N4 layer drives the metal film into compression, while elsewhere [27]. Briefly, 1 lm thick Al films were sputter
the overall stress in the bilayer is kept in tension to pre- deposited onto a Si wafer, coated on both sides with
vent buckling of the membrane. The Si3N4 film thus 75 nm of Si3N4 by means of low-pressure chemical
serves a dual purpose: It provides the driving force to vapor deposition (LPCVD). Immediately prior to the
deform the metal film into compression and it passivates Al deposition, a thin TiN sticking layer was grown using
one of the surfaces of the metal film. By adjusting the reactive sputtering. Freestanding Al/TiN/Si3N4 compos-
thickness ratio of the Si3N4 layer and the metal film, dif- ite membranes were microfabricated by opening rectan-
ferent levels of compressive stress can be obtained in the gular windows in the Si substrate using standard silicon
metal film. micromachining techniques. Freestanding Cu/Si3N4
If the Si3N4 film and the metal film are denoted as bilayers were prepared by first microfabricating free-
layers 1 and 2 with thickness t1 and t2, respectively, standing Si3N4 membranes followed by sputter deposit-
the average stress in the bilayer, rb, is given by ing 600 nm Cu films directly on top of the Si3N4
t1 t2 membranes. Both sets of samples were vacuum annealed
rb ¼ r1 þ r2 ; ð2Þ at 300 °C to stabilize the microstructure. Freestanding
t1 þ t2 t1 þ t2
Al and Cu membranes were prepared by etching the
where r1 and r2 are the stresses in the respective layers. Si3N4 or Si3N4/TiN layer beneath the metal films using
These stresses are related to the corresponding strains in reactive ion etching (RIE). Transmission electron micro-
each layer through the following constitutive equations: scope (TEM) micrographs in Fig. 3 show that grains in
  both films are roughly equiaxed and that the average
2h2
r1 ¼ f1 e01 þ 2 and grain size is 2.1 lm for Al and 0.9 lm for Cu. Both films
3a
  ð3Þ have a columnar grain structure with grain boundaries
2h2 traversing the thickness of the film. Annealing twins
r2 ¼ f2 e02 þ 2 ;
3a are only found in Cu grains.
where e01 and e02 are the residual strains in layers 1 and Both composite and freestanding films were tested in
2, respectively. Equilibrium requires that the average multiple loading/unloading cycles with a bulge test appa-
stress in the membrane depends on the membrane ratus [27]. For the Cu/Si3N4 samples, the elastic contri-
deflection as follows: bution of the Si3N4 films was measured independently
by dissolving the Cu in dilute nitric acid and testing the
pa2 Si3N4 membranes separately. The pressure–deflection
rb ¼ ; ð4Þ
2hðt1 þ t2 Þ of the Cu/Si3N4 bilayer and the freestanding Si3N4 mem-
where p is the total pressure supported by the bilayer. brane are plotted in Fig. 4. The curve of the freestanding
From Eqs. (2)–(4), the pressure–deflection relation for Si3N4 membrane consists of both loading (solid line) and
the bilayer is found to be unloading (solid squares) sections. It can be seen that the
    deformation of the Si3N4 membrane is elastic. For a
2ht1 2h2 2ht2 2h2 given deflection, the difference between the pressures
pðhÞ ¼ 2 f1 e01 þ 2 þ 2 f2 e02 þ 2 . ð5Þ
a 3a a 3a supported by the bilayer and the Si3N4 membrane gives
We define the pressures p1 and p2 as follows: the contribution of the Cu film, as illustrated by the in-
  sert in Fig. 4. The stress–strain curve of the Cu film is
2ht1 2h2 then calculated using Eqs. (1) and (6). The same method
p1 ðhÞ ¼ 2 f1 e01 þ 2 and
a 3a could not be applied to the Al samples because the TiN
  had a large compressive stress that caused the freestand-
2ht2 2h2
p2 ðhÞ ¼ 2 f2 e02 þ 2 . ð6Þ ing Si3N4/TiN bilayer to buckle if the Al was dissolved.
a 3a
Instead, the properties of the Si3N4/TiN were obtained
Since the Si3N4 film (layer 1) deforms only elastically, from the elastic unloading curves of both Si3N4/TiN/Al
the constitutive function f1 is independent of whether and freestanding Al films [26]. The elastic contribution
the film is freestanding or part of a bilayer. Comparison of Si3N4/TiN bilayer can then be subtracted from the
of Eqs. (1) and (5) shows that p1(h) is equal to the pres- data of the Si3N4/TiN/Al composite films.
180 Y. Xiang, J.J. Vlassak / Scripta Materialia 53 (2005) 177–182
Fig. 3. TEM micrographs showing the grain structure of (a) a 1 lm Al
film and (b) a 0.6 lm Cu film.
Fig. 4. Pressure–deflection curves of the Cu/Si3N4 bilayer and the
freestanding Si3N4 film. For clarity, only every 20th data point of the
Si3N4 unloading curve is displayed (j).
The resulting stress–strain curves are presented in
Fig. 5(a) and (b) for passivated Cu and Al film, respec-
tively. Also plotted in Fig. 5 are typical stress–strain
curves for unpassivated Cu and Al films. For both mate-
rials, the passivated films show a very strong Bauschin-
ger effect: during unloading, reverse plastic flow starts
Fig. 5. Stress–strain curves of passivated and freestanding films for (a)
a 0.6 lm Cu film and (b) a 1 lm Al film. The stress–strain curves are
offset by the biaxial residual strains in the films, as represented by the
dashed lines starting from the origins in the stress–strain curves.
when the applied stress is still in tension and continues
when the films are loaded in compression. Each loading
cycle shows significant hysteresis, which increases with
increasing plastic strain. By contrast, the stress–strain
curves of unpassivated films have unloading cycles with
little or no hysteresis.
3. Discussion
A dislocation-based mechanism is proposed to ex-
plain the difference between passivated and unpassivated
films. Fig. 6 depicts a thin metal film with a columnar
grain structure. The lower surface of the film is passiv-
ated by a material that forms a strong interface, while
the upper surface is free. When dislocations reach an
obstacle such as a grain boundary or a strong interface
during forward plastic deformation, they form pileups;
when they reach a free surface, they simply exit the
material to form surface steps. In the presence of pile-
ups, dislocations in the film are subject to two types of
stress: the externally applied stress and the internal stress
imposed by the pileups; the former provides the driving
force for further plastic flow, while the latter acts as a
 Y. Xiang, J.J. Vlassak / Scripta Materialia 53 (2005) 177–182
Fig. 6. Dislocations interact with grain boundaries as well as free
surfaces and interfaces in thin films.
resistance. During tensile loading of a passivated film,
dislocation pile-ups form at the film/passivation inter-
face resulting in significant back stresses in the film.
Upon unloading, these back stresses cause dislocations
to glide in the opposite direction resulting in a nonlinear
unloading curve. The different behavior of passivated
and freestanding films arises as follows. In freestanding
films, many dislocations can exit the film due to the
proximity of the free surface and no significant Bausch-
inger effect is observed. If one of the film surfaces is pas-
sivated (Fig. 6) and the interface is strong, a boundary
layer with high dislocation density is formed near the
interface [23]. This boundary layer, which may consist
of simple pileups or may have a more complex structure,
has an associated back stress that is not present in free-
standing films. As a result of this high back stress, re-
verse plastic flow starts much earlier during the reverse
loading cycle. If the back stress is large enough, reverse
flow can start even when the externally applied stress is
still in tension, as observed in the stress–strain curves of
the passivated Cu and Al film in Fig. 5. It should be
noted that the Bauschinger effect will occur only when
the obstacle to dislocation motion is strong enough to
cause significant back stresses. The data in Fig. 5(b) sug-
gest that this is not the case for the thin native oxide that
forms on freestanding Al films.
In addition to the composite-film technique proposed
here, the substrate curvature technique can also be used
to load thin films deposited on a substrate alternating in
tension and compression. In this technique, a metal film
on a silicon substrate is loaded into compression by
heating the sample until the residual stress in the film be-
comes compressive due to the thermal mismatch with
the substrate. Upon cooling, the film goes into tension.
From the change in substrate curvature, one can deter-
mine the stress in the film and thus obtain an estimate
for the forward and reverse flow stress. Using this
technique, Baker et al. [28] observed ‘‘anomalous’’ ther-
momechanical behavior for Cu films encapsulated
between Si3N4 barrier and passivation layers, when a
181
small amount of oxygen was introduced in the deposi-
tion system during film growth: the films exhibited a
Bauschinger-like behavior that is large compared to that
observed for bulk Cu. The result is difficult to interpret,
however, because the applied strain cannot be decoupled
from the temperature change and because at elevated
temperatures other deformation mechanisms such as
diffusional creep may come into play [29]. Compared
to the substrate curvature technique, the composite film
technique proposed here has the unique advantage of
measuring the isothermal stress–strain behavior and is
thus ideal for studying Bauschinger effect in thin metal
films.
4. Summary
In conclusion, we have developed a new experimental
technique to deform thin metal films alternately in ten-
sion and compression and to measure the corresponding
isothermal stress–strain curve. This method was used to
investigate the Bauschinger effect in thin sputter depos-
ited Al and Cu films. A large Bauschinger effect is found
in both materials when one of the film surfaces is passiv-
ated with a thin layer of LPCVD Si3N4. We propose
that the strong interface between the film and passivat-
ing layer prevents dislocations from exiting the film
and allows the build-up of significant back stresses. On
unloading, these back stresses cause reverse plastic flow.
By contrast, stress–strain curves of unpassivated films
show little or no reverse flows when the film is fully un-
loaded because the free surface allows dislocations to
exit the film and no back stresses are generated. We
anticipate that these experimental results will be useful
for validating and refining existing models for thin-film
plasticity.
Acknowledgments
This research is funded by NSF (DMR-0133559 and
DMR-0215902). The authors gratefully acknowledge
John W. Hutchinson and Morton G. Gurtin for helpful
discussion.
References
[1]Bauschinger J. Civilingenieur 1881;27:289.
[2]Masing G. Wiss Veroff Siemens-Konzern 1923;3:231.
[3]Stoltz RE, Pelloux RM. Metall Trans A 1976;7:1295.
[4]Sowerby R, Uko DK, Tomita Y. Mater Sci Eng 1979;41:43.
[5]Bate PS, Wilson DV. Acta Metall 1986;34:1097.
[6]Hasegawa T, Yakou T, Kocks UF. Mater Sci Eng 1986;81:189.
[7]Pedersen OB, Brown LM, Stobbs WM. Acta Metall 1981;29:1843.
[8]Margolin H, Hazaveh F, Yaguchi H. Scripta Metall 1978;12:1141.
[9]Ono N, Tsuchikawa T, Nishimura S, et al. Mater Sci Eng
1983;59:223.
[10] Aran A, Demirkol M, Karabulut A. Mater Sci Eng 1987;89:L35.
182 Y. Xiang, J.J. Vlassak / Scripta Materialia 53 (2005) 177–182

[11] Atkinson JD, Brown LM, Stobbs WM. Philos Mag 1974;30:1247. [22] Bittencourt E, Needleman A, Gurtin ME, et al. J Mech Phys
[12] Brown LM, Stobbs WM. Philos Mag 1971;23:1185. Solids 2003;51:281.
[13] Brown LM, Stobbs WM. Philos Mag 1971;23:1201. [23] Nicola L, Van der Giessen E, Needleman A. J Appl Phys
[14] Brown LM. Scripta Metall 1977;11:127. 2003;93:5920.
[15] Hsu R, Arsenault RJ. Mater Sci Eng 1984;66:35. [24] Needleman A, Van der Giessen E. Mater Sci Eng A 2001;309:1.
[16] Nix WD. Metall Trans A 1989;20:2217. [25] Yefimov S, Groma I, Van der Giessen E. J Mech Phys Solids
[17] Vinci RP, Vlassak JJ. Ann Rev Mater Sci 1996;26:431. 2004;52:279.
[18] Arzt E. Acta Mater 1998;46:5611. [26] Vlassak JJ, Nix WD. J Mater Res 1992;7:3242.
[19] Thompson CV. Ann Rev Mater Sci 2000;30:159. [27] Xiang Y, Chen X, Vlassak JJ. J Mater Res; submitted for
[20] Triantafyllidis N, Aifantis EC. J Elast 1986;16:225. publication.
[21] Fleck NA, Muller GM, Ashby MF, et al. Acta Metall [28] Baker SP, Keller-Flaig RM, Shu JB. Acta Mater 2003;51:3019.
1994;42:475. [29] Kobrinsky MJ, Thompson CV. Appl Phys Lett 1998;73:2429.