Stress relaxation in tungsten films by ion irradiation

Article abstract of DOI:10.1063/1.119515

Tungsten films, grown by plasma-enhanced chemical vapor deposition for very large scale integration interconnect applications, suffer from a high internal tensile stress that leads to contact failure. We show with wafer curvature measurements that the stress can be relaxed via viscous flow at room temperature by irradiating the films with energetic ions after deposition. Transmission electron microscopy does not indicate significant structural changes in the W films during irradiation. We varied the irradiation conditions (from 140 keV B ions to 400 keV P ions) and find that the flow rate scales with the nuclear stopping power. Similarities and differences with beam-induced mixing and diffusion are discussed.

Full text of DOI:10.1063/1.119515

Stress relaxation in tungsten films by ion irradiation
E. Snoeks, K. S. Boutros, and J. Barone
Citation: Applied Physics Letters 71, 267 (1997); doi: 10.1063/1.119515
View online: http://dx.doi.org/10.1063/1.119515
View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/71/2?ver=pdfcov
Published by the AIP Publishing
Articles you may be interested in
Isothermal stress relaxation in electroplated Cu films. II. Kinetic modeling
J. Appl. Phys. 97, 103532 (2005); 10.1063/1.1904721
Isothermal stress relaxation in electroplated Cu films. I. Mass transport measurements
J. Appl. Phys. 97, 103531 (2005); 10.1063/1.1904720
Formation of Au 0.6 Ge 0.4 alloy induced by Au-ion irradiation of Au/Ge bilayer
J. Appl. Phys. 93, 903 (2003); 10.1063/1.1530356
Postdeposition relaxation of internal stress in sputter-grown thin films caused by ion bombardment
J. Appl. Phys. 85, 841 (1999); 10.1063/1.369249
Effects of ion irradiation on the residual stresses in Cr thin films
Appl. Phys. Lett. 73, 891 (1998); 10.1063/1.122029
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
132.174.255.116 On: Wed, 17 Dec 2014 17:51:34

Stress relaxation in tungsten films by ion irradiation
E. Snoeks,a) K. S. Boutros, and J. Barone
Philips Research, Briarcliff Manor, New York 10510
͑Received 31 March 1997; accepted for publication 14 May 1997͒
Tungsten films, grown by plasma-enhanced chemical vapor deposition for very large scale
integration interconnect applications, suffer from a high internal tensile stress that leads to contact
failure. We show with wafer curvature measurements that the stress can be relaxed via viscous flow
at room temperature by irradiating the films with energetic ions after deposition. Transmission
electron microscopy does not indicate significant structural changes in the W films during
irradiation. We varied the irradiation conditions ͑from 140 keV B ions to 400 keV P ions͒ and find
that the flow rate scales with the nuclear stopping power. Similarities and differences with
beam-induced mixing and diffusion are discussed. © 1997 American Institute of Physics.
͓S0003-6951͑97͒04428-8͔
Tungsten thin films are an alternative solution for a low
resistivity ohmic contact to silicon. Chemical vapor deposi-
tion ͑CVD͒ or plasma-enhanced CVD ͑PE-CVD͒ are among
the commonly used techniques to fabricate these W films.1,2
A drawback of these techniques is that the resulting W films
often experience internal mechanical stress that may result in
contact failure and reduction in yield. Developing a tech-
nique to reduce the internal stress in these films will result in
an improvement of reliability.
Ion beam assisted deposition, using ion energies of sev-
eral hundreds eV, has been used in an attempt to produce
such low-stress metal films.3,4 The success of this technique
has been limited, since an intrinsic stress scaling law seems
to exist caused by atomic peening, where an excess of inter-
stitial atoms result in compressive stress.5,6 In the present
letter, we report the use of ions at several hundreds keV
energy to relax the stress in W films after deposition is com-
pleted. This is an ex situ bulk treatment, rather than an in situ
surface treatment. Ion irradiations in the keV and MeV en-
ergy regime have been shown to induce plastic deformation
in amorphous covalent materials.7,8 Stress relaxation occured
via Newtonian viscous flow. The same apparent behavior
was found for stress relaxation of Al films on SiO2.7
The layers consist of textured columns; under a certain dif-
fraction condition only a few columns light up in this dark-
field image. The typical diameter of the columns in the W
film ranges from 50 to 150 nm, as determined by plan-view
TEM.
The wafer curvature was measured before and after the
W deposition, using a laser deflection apparatus ͑Flexus
F2418͒. The amount of wafer bending caused by the W
deposition ͑measured from the edge to the center of the wa-
fer͒ was typically about 25 ␮m. From this measurement, the
integrated in-plane stress in the W film can be derived using
the biaxial elastic modulus (YSiO ϭ1ϫ1011 Pa) and the
2
thickness and diameter of the substrate.9 The integrated ten-
sile stress in our W films at room temperature was about
88Ϯ9 N/m, which corresponds to an average local stress of
␴ϭ͑2.6Ϯ0.3͒ϫ108 Pa.
The tungsten layers were then irradiated with 400 keV P,
200 keV N, or 140 keV B ions. These energies were chosen
such that the range plus the ion straggle10 were about equal
for each ion and corresponded to 80% of the thickness of the
We now show that intrinsic tensile stress in polycrystal-
line W films decreases exponentially during ion irradiation at
room temperature. The relaxation rate is proportional to the
nuclear stopping, indicating that relaxation is activated by
ballistic interactions in the film. Structural changes upon
stress relief are assessed by transmission electron miscros-
copy ͑TEM͒. Comparison with other data on radiation-
induced flow suggests that this is a universal phenomenon.
The W films were deposited on 4-in.-diam quartz wafers
͑525 ␮m thick͒, on which a 5% TiW adhesion layer has first
been sputtered. During PE-CVD of W, via H2 reduction of
WF6, the substrates were held at 360 °C. Figure 1͑a͒ shows a
dark-field cross-sectional TEM micrograph, obtained with a
Philips EM430 microscope operated at 300 kV, of the layers
after deposition. The W and TiW layers can clearly be dis-
tinguished in the image. The thickness of the TiW layer is
0.13 ␮m; the average thickness of the W film is 0.34 ␮m.
FIG. 1. Dark-field cross-sectional TEM micrographs of W films ͑a͒ and ͑b͒
before irradiation and ͑c͒ after irradiation with 400 keV P to a fluence of
2.5ϫ1014/cm2.
a͒
Electronic mail: snoeks@amolf.nl
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
Appl. Phys. Lett. 71 (2), 14 July 1997 0003-6951/97/71(2)/267/3/$10.00 © 1997 American Institute of Physics 267
132.174.255.116 On: Wed, 17 Dec 2014 17:51:34

If we assume that ␨ϭ0 ͑i.e., no density changes͒, the
data from Fig. 2 can be used to estimate the radiation-
induced viscosity. The solution of Eq. ͑1͒ in that case is a
single exponential function
␴ ␾͒
͑
YW
ϭexp Ϫ
␾ ,
͑2͒
ͩ
ͪ
␴
6␩
rad
0
where ␴(␾)/␴ is the normalized stress after irradiation to
0
fluence ␾, as depicted in Fig. 2. The dashed curve is the best
fit of Eq. ͑2͒ to the data. Using YWϭ57ϫ1010 Pa,11 we find
that ␩
of W under P irradiation is (0.96Ϯ0.21)
rad
ϫ1025 Pa ions/cm2, (3.7Ϯ0.7)ϫ1025 Pa ions/cm2 under N
irradiation, and (7.7Ϯ0.7)ϫ1025 Pa ions/cm2 for B irradi-
ated W. Hence, the fluence needed for relaxation increases
with decreasing ion mass. This is summarized in the inset in
Fig. 2, where the fluidity ͑inverse of the experimentally de-
FIG. 2. Relative stress in the W film as a function of P fluence, measured ex
situ in one film after subsequent irradiations at 400 keV. The initial stress in
the film was about 2.6ϫ108 Pa tensile. The dashed line is the best exponen-
tial fit. The inset shows the radiation-induced fluidity (1/␩rad) of W vs
nuclear stopping power. Data are shown for P, N, and B irradiations, as
indicated. The line shows the best linear fit to the data.
termined ␩ ͒ is plotted against the maximum nuclear stop-
rad
ping power in the W film.10 As can be seen, 1/␩rad is close to
linearly proportional to the nuclear stopping power. For the
range of ions and energies used, the electronic stopping
power varied much less ͑from 1.3 keV/nm for 400 keV P to
0.66 keV/nm for 140 keV B͒ than the nuclear stopping
power. We therefore conclude that electronic stopping con-
tributes much less to the radiation-induced flow process than
nuclear stopping.
film. The ion beam current was held below 50 nA/cm2 to
keep the subtrates at room temperature. At specified ion flu-
ences, the wafer was taken out of the implanter in order to
measure the wafer curvature, after which it was reloaded in
the implanter for the irradiation to be resumed.
After the P irradiations were completed, the layer was
characterized by TEM. The film still consisted of the 50–
150-nm-diam columnar structure, as was observed in Fig.
1͑a͒ before irradiation. On smaller length scales, only minor
changes were observed. Figure 1͑b͒ shows a dark-field image
taken in cross section within a column before irradiation, and
Fig. 1͑c͒ after irradiation with the same magnification. In
Fig. 1͑b͒, some very small (ϳ3 nm) bright reflecting fea-
tures are seen, which could be defect loops or small crystal-
line inclusions or grains. These features are only observed
under specific orientations of the sample. These small bright
features could no longer be observed within the columns
after irradiation, as is illustrated in Fig. 1͑c͒. Extended de-
fects ͑such as dislocations, which may form as a result of
stress relaxation͒ were not observed either.
Since the structural changes are only very minor, we
conclude that the mechanism for stress relaxation is related
to the point defects within the ion cascade. From that point
of view, the atomic motion is a result of the defect density
and temperature which increase very briefly and strongly
within a few nm from where an energetic ion penetrates
through the material. The temperature in the collision cas-
cade may rise many thousands of degrees for a few picosec-
onds, sufficient to cause permanent rearrangement of the
atoms.12,13 This is known to lead to beam-induced mixing
and diffusion, of which the rate scales with nuclear energy
deposition if cascades or subcascades do not overlap.12,14–16
When a stress field is present in the film, the atomic rear-
rangement will be biased such that the elastically stored en-
ergy in the film is lowered. Radiation-induced flow thus
seems to be related to beam-induced diffusion.
Figure 2 shows the normalized stress in the W film mea-
sured on one single wafer as a function of P irradiation flu-
ence. The y axis is normalized to the value found for the
as-deposited wafer. The data show an exponential decrease
of the stress in the W with P fluence. After a fluence of 2.5
ϫ1014/cm2 ͑corresponding to about 0.6 displacements per
atom͒,10 the stress is relaxed to less than 10% of the original
value. Thermally activated creep processes at room tempera-
ture are many orders of magnitude slower than the creep
rates observed in this ion beam experiment. The experiment
was also performed with N and B irradiations. The stress was
found to decrease exponentially in both cases, similar to Fig.
2, with the difference that factors of 3.9 and 8.0 higher ion
fluences were needed, respectively, to reach the same amount
of stress relaxation as shown for P.
There are several possible contributions to in-plane
stress changes during irradiation, such as density changes, or
plastic shear deformation. The latter may be characterized in
terms of a radiation-induced viscosity,7,8 if the strain rate is
proportional to the stress ͑Newtonian shear flow͒ which is
indeed the case in Fig. 2. Since the observed stress change is
fully beam related, time is replaced by ion fluence, as was
explained in Refs. 7 or 8. Following these references the
general biaxial stress rate equation is given by
d␴
d␾
YW
ϭY ␨ ␾͒Ϫ
␴,
͑1͒
͑
W
6␩
rad
where ␾ is the ion fluence, YW is the biaxial elastic modulus
Based on conceptual similarities and differences be-
tween diffusion and plastic flow,8,17,18 it is perhaps possible
to modify well established models for beam-induced
diffusion14–16 in order to describe beam-induced flow. One
of poly crystalline tungsten, and ␩ the radiation-induced
rad
viscosity; ␨ denotes the sum of strain rates due to phenom-
enalike density changes, and may depend on fluence.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
268 Appl. Phys. Lett., Vol. 71, No. 2, 14 July 1997 Snoeks, Boutros, and Barone
132.174.255.116 On: Wed, 17 Dec 2014 17:51:34

beam-induced diffusivity scales linearly with nuclear stop-
ping power in the athermal regime only, i.e., at irradiation
temperatures below the transition temperature Tc . Brong-
ersma et al.20 have shown that the temperature dependence
of ␩ in SiO2, very similarly, shows a temperature depen-
rad
dent and independent regime with TcϷ300 K.20 Tc is a ma-
terials parameter which scales with the cohesive energy.15,16
Using a cohesive energy of 8.7 eV,17 we find that for tung-
sten TcϷ830 K. Because this is well above room tempera-
ture, it is expected that our measurements of ␩ in W films
rad
were performed in the temperature-independent regime. To
experimentally study the temperature dependence would be
an interesting next step towards understanding radiation-
induced flow of W.
In conclusion, the intrinsic stress in PE-CVD W films
can be relaxed at room temperature by ion irradiation. The
stress relaxes via creep or flow in an exponential fashion
with ion fluence. The relaxation rate scales linearly with the
portion of the stopping that is deposited into atomic colli-
sions. This is consistent with the idea that the relaxation
phenomenon is a ballistic effect, that will occur during irra-
diation with any ion. Based on conceptual similarities with
beam-induced diffusion, we speculate that the product of
beam-induced viscosity and diffusivity is a constant that de-
pends on materials parameters such as the elastic constant
and others.
The authors thank J. Burke for assistance with the ion
irradiations. The TiW adhesion layers were grown by J. Mur-
phy. Thanks are also due to R. Egloff, M. L. Brongersma,
and A. Polman ͑FOM-Institute, Amsterdam, The Nether-
lands͒ for valuable discussions.
FIG. 3. Double log plot of ␩rad /Y against nuclear stopping power for W
͑data from this work͒, Al ͑data from Ref. 7͒, SiO2 ͑Ref. 8͒, sodalime-
borosilicate glass ͑Ref. 8͒, and amorphous silicon ͑Ref. 7͒. The lines are best
fits, and both have identical slopes of Ϫ1.1.
aspect that has to be taken into account in the case of plastic
flow is how well a material can elastically accommodate an
imposed mechanical strain. To test this idea, we plotted
␩
rad devided by the bielastic constant of the film for different
materials in Fig. 3 as a function of maximum nuclear stop-
ping power.19 The data points for W ͑dots͒ are based on the
present work, while the data points for annealed Al ͑black
triangle͒ and pure amorphous silicon ͑open triangle͒ are
based on measurements by Volkert and Polman,7 and the
data for SiO2 ͑circles͒ and sodalime-borosilicate glass ͑ϫ͒
are obtained from Ref. 8. Although ␩ may differ widely
rad
from material to material, scaling by the biaxial elastic con-
stant Y, results in a quantity that is a clear function of nuclear
stopping for the polycrystalline metal films ͑filled data
points͒ as well as the amorphous covalent films ͑open data
points͒. On the log–log plot, the slopes of the best linear fits
are Ϫ1.1 for both classes of materials, which approximately
1 E. K. Broadbent, J. Vac. Sci. Technol. B 5, 1661 ͑1987͒.
2 A. J. Learn and D. W. Foster, J. Appl. Phys. 58, 2001 ͑1985͒.
3 R. A. Roy, R. Petkie, and A. Boulding, J. Mater. Res. 6, 80 ͑1991͒.
4 R. A. Hoffman, J. C. Lin, and J. P. Chambers, Thin Solid Films 206, 230
͑1991͒.
corresponds to an inverse relationship between ␩ and the
rad
nuclear stopping. This striking similarity in the behavior of
different materials may indicate that the ion beam induced
relaxation mechanism can be understood in a general theo-
retical framework, as has been the case for beam-induced
diffusion.14–16
5 H. Windischmann, J. Appl. Phys. 62, 1800 ͑1987͒.
6 C. A. Davis, Thin Solid Films 226, 30 ͑1993͒.
7 C. A. Volkert and A. Polman, Mater. Res. Soc. Symp. Proc. 235, 3 ͑1992͒.
8 E. Snoeks, T. Weber, A. Cacciato, and A. Polman, J. Appl. Phys. 78, 4723
͑1995͒.
However, more work is needed to understand which
other materials constants should be taken into account. For
instance, the relaxation rate in the amorphous covalent ma-
terials is faster than the metals for the same amount of
nuclear stopping. This might be a result of the recrystalliza-
tion process which happens during the cooling phase of the
collision cascade in the case of metals and not in the case of
amorphous materials. For metals this means that although the
stress may be fully relieved during the peak of the spike,
some stress is reintroduced by the crystallization. This effec-
tively results in a higher value for the radiation-induced vis-
cosity. In glasses this is not the case since the cooling down
does not cause a phase change inside the spike.20
9 G. G. Stoney, Proc. R. Soc. London Ser. A 82, 172 ͑1909͒.
10
TRIM-89: J. P. Biersack and L. J. Haggmark, Nucl. Instrum. Methods 174,
257 ͑1980͒.
11 S. W. H. Yih and C. T. Wang, Tungsten ͑Plenum, New York, 1979͒.
12 T. Diaz de la Rubia, R. S. Averback, Horngming Hsieh, and R. Benedek,
J. Mater. Res. 4, 579 ͑1989͒.
13 M. Ghaly and R. S. Averback, Phys. Rev. Lett. 72, 364 ͑1994͒.
14 W. Bolse, Mater. Sci. Eng. 12, 53 ͑1994͒.
15 Y.-T. Cheng, Mater. Sci. Rep. 5, 45 ͑1990͒.
16 M. Nastasi and J. W. Mayer, Mater. Sci. Eng. R. Rep. 12, 1 ͑1994͒.
17 P. A. Egelstaff, An Introduction to the Liquid State ͑Oxford University
Press, Oxford, 1992͒, Chap. 13.
18 P. A. Duine, J. Sietsma, and A. van den Beukel, Phys. Rev. B 48, 6957
͑1993͒.
It is outlined in, for example, Refs. 14–16 that beam-
induced diffusion is composed of a ͑beam-induced͒ athermal
contribution and a ͑beam-assisted͒ thermal migration. The
19 The average stopping power in the film is about 75% of the maximum
stopping power for all present irradiations.
20 M. L. Brongersma, E. Snoeks, and A. Polman ͑unpublished͒.
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
Appl. Phys. Lett., Vol. 71, No. 2, 14 July 1997 Snoeks, Boutros, and Barone 269
132.174.255.116 On: Wed, 17 Dec 2014 17:51:34