Surface & Coatings Technology 201 (2007) 8935 – 8938

www.elsevier.com/locate/surfcoat

Hot-wire assisted chemical vapor deposition of Cu by

direct-liquid-injection of CupraSelect®
G. Papadimitropoulos, D. Davazoglou ⁎
NCSR “Demokritos”, Institute of Microelectronics, POB 60228, 15310 Agia Paraskevi, Greece

Available online 1 May 2007

Abstract

A novel chemical vapor deposition reactor for Cu films is presented in which the chemical reactions were assisted by a heated tungsten
filament (hot-wire CVD, HWCVD). The vapors used were hexafluoroacetylacetonate Cu(I) trimethylvinylsilane (CupraSelect®), which is liquid at
room temperature and was injected into the reactor with the aid of a direct-liquid injection system using N2 as carrier gas. The deposition rates of
HWCVD Cu films obtained with this reactor were one and a half times higher than for typical thermally grown films at a filament temperature of
170 °C. Moreover, high quality Cu films were deposited on SiLK®, which cannot be done by thermal CVD. The resistivity of HWCVD Cu films
is two to five times higher than those for thermally grown films due to the possible presence of impurities into the Cu films. The reaction of the
CupraSelect® vapors with the W filament and the W-covered Si substrate, leads to the formation of Cu–W–O compounds in the Cu films. The
presence of such compounds does not degrade catastrophically the conductivity of Cu HWCVD films.
© 2007 Elsevier B.V. All rights reserved.

Keywords: Copper; Hot-wire CVD; Direct-liquid-injection; SiLK®; CupraSelect®

1. Introduction high resistivity, etc.) when deposited on insulators. This is an

In modern integrated circuit manufacturing, copper has replaced
the traditionally used aluminum for interconnects due to its lower
bulk resistivity (1.8 vs 2.7 μΩ cm) and higher electromigration
resistance [1]. The use of copper in interconnects enables cir-
cuits to run faster and to be more reliably than when aluminum
metalized [2]. Several methods have been developed to deposit
copper layers such as electroplating [3], which is the most
commonly used industrial method, and chemical vapor deposition
(CVD) [1]. CVD methods ensure the deposition of conformal
copper layers within small trenches and via holes without the need
of current conductive seed layer, which is necessary in electro-
plating [3]. One of the main drawbacks in using CVD methods for
copper deposition is that most of the copper precursors are solid
[1]. This renders these processes poorly repeatable, unreliable and
with low growth rate. Other problems of the CVD Cu films are
their poor adhesion and properties (e. g. large grain formation,
⁎ Corresponding author.
E-mail address: d.davazoglou@imel.demokritos.gr (D. Davazoglou).
0257-8972/$ - see front matter © 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.surfcoat.2007.04.085
important problem since such films are used in integrated circuits
metallization. In order to overcome the above drawbacks liquid
metal organic (MO) precursors were synthesized, in which the
metal appears with an oxidation number equal to +1, and direct-
liquid-injection (DLI) systems were developed [4]. The use of
liquid precursors and DLI has responded to the repeatability and
low growth rate problems. Moreover, copper (I) precursors offer
the advantage of selective CVD (SCVD) under specific deposi-
tion conditions [1,5–7]. However, the problems of poor adhesion
and non-conformal deposition on insulating surfaces have re-
mained unsolved.
In this work, in order to enhance further the growth rate of Cu
and to promote deposition on plastic insulating substrates we
combined the DLI of a liquid precursor in the deposition chamber
with the so called hot-wire CVD (HWCVD) technique. HWCVD
consists in the separate heating of the gas phase with the aid of a
tungsten wire (in the remaining of this paper Cu layers deposited
by DLI-MOCVD with a separate heating of the gas phase will
referred to as hot-wire CVD, HWCVD Cu layers). It was shown
that HWCVD Cu films were deposited on conductive surfaces
with a growth rate almost double compared to the conventional
8936 G. Papadimitropoulos, D. Davazoglou / Surface & Coatings Technology 201 (2007) 8935–8938
Fig. 1. Dependence of the growth rate of HWCVD Cu films on W and SiLK®-
covered Si substrates. The corresponding variation for thermally grown films is
also seen (the straight lines shown are only used as a guide for the eye).
MOCVD and, moreover, they are deposited on a low dielectric
constant (low-k) material used in IC interconnects (SiLK®).
2. Experimental details
The home-made hot-wire CVD system used [8] was com-
prised by a vertical type cold wall reactor of stainless steel and was
equipped with a tungsten hot-wire to enable the separate heating
of the gas phase. Copper films are deposited from CupraSelect®
vapors [9], which is the commercial name of the hexafluoroace-
Fig. 2. SEM micrographs of the surface of HWCVD (upper) and thermal Cu films (lower) deposited on W-covered Si Substrates at 180, 200 and 220 °C, from left to
right.
tylacetonate(hfac) Cu(I) trimethylvinylsilane(tmvs) and which is
liquid at room temperature. The precursor is injected with a rate of
6 g/h in the vaporization unit (Bronkhorst CEM 200) where is
heated at 45 °C and mixed with a stream of ultra-pure N2 (carrier
gas) set at mass flow rate 45 sccm. This DLI [4] system has a faster
response time, better stability and a lower working temperature
than conventional systems based on bubbling. Substrate and
plumbing temperatures were controlled separately with the aid of
Eurotherm temperature controllers.
All depositions were carried out at a pressure of 2 Torr and a
mass flow rate of CupraSelect® of 6 g/h. The tungsten filament
temperature was 170 °C and its presence did not affect the sub-
strate temperature. Cu depositions were made on 3 in. oxidized
(100) silicon substrates covered with LPCVD W, which was
shown to act as diffusion barrier efficiently at temperatures up to
550 °C [10], and SiLK® layers. Films were characterized with
X-ray diffraction (XRD) with a D500 Siemens diffractometer
using Cu Kα radiation and a secondary graphite monochroma-
tor. Scanning electron microscopy (SEM) measurements in top
view and cross section were also made with a FEI Inspect instru-
ment. The electrical resistance measurements were made with the
four-point probe method and combined with cross-section SEM
measurements to obtain thickness, yielded film resistivity.
3. Results and discussion
In Fig. 1 the growth rate of Cu films deposited on W and
SiLK®-covered Si substrates at temperatures of 180, 200 and
220 °C are shown. The depositions on W were both thermally
and hot-wire assisted while on SiLK®, since thermal deposition
 G. Papadimitropoulos, D. Davazoglou / Surface & Coatings Technology 201 (2007) 8935–8938 8937

Fig. 5. XRD patterns recorded on HWCVD Cu films deposited on SiLK®.

yielded no films at all at these temperatures, only HWCVD

results are reported. It can be observed that the separate heating
of the gas phase by the hot-wire induced a significant increase
(by a factor of approximately one and a half) of the deposition
rate. The kinetics of the reaction does not seem to be signifi-
cantly influenced by the hot-wire and the growth rate vs tem-
perature variation is simply parallel shifted towards higher
values. Anticipating results presented next it is noted that
SiLK® is more inert chemically vs the gas phase than W, so the
chemical reaction between substrate and gas phase is expected
to influence the sticking coefficients and surface diffusion
therefore, deposition kinetics.
In Fig. 2 SEM top views of thermally and hot-wire deposited
Cu films on W-covered substrates and at various temperature are
shown. It can be observed that the surface morphology changes
and the HWCVD films are more “grainy” and, therefore, rougher.
A safe comparison of grain size cannot be made since samples
Fig. 3. Surface micrographs of HWCVD Cu films deposited on SiLK® covered
Si substrates at 180 (upper), 200 (middle) and 220 °C (lower).
produced by thermal CVD have elongated grains while in those
by HWCVD the grains have a more spherical shape. However,
freely speaking, it can be reported that in Cu films grown ther-
mally at 180 °C grains have sizes of the order of 100 nm, which

Fig. 4. Variation of the Cu film resistivity with the temperature of deposition for
HWCVD films on W and SiLK® substrates. That for thermally grown films on
W is also reported. Fig. 6. XRD patterns recorded on thermally grown Cu films deposited on W.
8938 G. Papadimitropoulos, D. Davazoglou / Surface & Coatings Technology 201 (2007) 8935–8938
increase with temperature to approximately 150 nm at 220 °C.
HWCVD films have grains with size increasing rapidly with
substrate temperature from 100 to 400 nm for temperatures of 180
and 220 °C, respectively.
In Fig. 3 SEM top views of HWCVD Cu films deposited on
SiLK® covered substrates and at various temperature are shown.
It can be observed that, similarly to HWCVD Cu films deposited
on W, films have a granular morphology with grains almost
spherical. The grain size depends exponentially on the deposition
temperature and varies from 50 nm at substrate temperature of
180 °C to 200 nm for films at 220 °C. Comparing grain sizes of
HWCVD Cu films deposited on W (Fig. 2) and SiLK® (Fig. 3) at
the same temperature show that they are very similar. This means
that the grain size is mostly dependent on the deposition tempera-
ture and not gas phase reactions. This conclusion changes at
higher filament temperatures, above 650 °C, where the grain size
is of the order of 20–30 nm independent of the deposition
temperature.
In Fig. 4 the resistivity variation of Cu films thermally and hot-
wire deposited on W and SiLK® covered substrates are shown. It
must be noted that these values are probably overestimated
because the thickness measurements by SEM were not very
accurate (thickness was ranging between 200 and 500 nm).
Another cause of inaccuracy in the determination of resistivity of
Cu films was that they were not protected against oxidation after
deposition (in some cases resistance measurements were made a
few days after deposition). In any case, it is reasonable to believe
that the general trends of the resistivity variation with the various
parameters involved are preserved. It is observed that while the
thermal Cu films grown on W have a resistivity of the order of
2.5 μΩ cm independent on deposition temperature, the HWCVD
films exhibit higher resistivities, which drop exponentially with
temperature towards the values for thermally grown films. This
indicates that low film resistivity is obtained by decomposing the
CupraSelect® vapors on the surface [11], catalyzed by the pres-
ence of free electrons on it, and not in the gas phase. At higher
filament temperatures, above 650 °C, where the gas phase reac-
tions are even faster, the Cu films obtained were 2–3 orders of
magnitude more resistive. This means that the decomposition of
vapors is not complete; impurities remain into the film destroying
its conductivity. This conclusion is further corroborated by the
observation that the resistivity is slightly (if at all) affected by the
nature of the substrate (SiLK® or W).
Another point of concern was the reaction between the
CupraSelect® vapors, the tungsten filament and the substrate. In
Fig. 5 XRD patterns taken on HWCVD Cu films deposited on
SiLK® at various temperatures are shown. It is observed that
except for the peaks corresponding to Cu, another one corre-
sponding to WO3 is shown at deposition temperature of 180 °C.
At 200 °C this peak is attenuated and at 220 °C the WO3 peak is
almost completely absent. Obviously the origin of W is the heated
filament. The CupraSelect® contains oxygen which probably
reacts with W to produce volatile WOx species, which are incor-
porated into the growing film. At lower deposition temperatures
these species approach the substrate and condense on it easily
since they are produced in a way similar to thermal evaporation
while at higher temperatures this condensation process occurs less
intensively (if at all). It is to be noted that the presence of tung-
sten oxide in the Cu films is certainly related to its resistivity.
Moreover, the possible presence of other impurities such as C
and F originating from the incomplete decomposition of the
CupraSelect® vapors also affect film resistivity. In Fig. 6 XRD
patterns taken on thermally grown Cu films on W at various
temperatures are shown. It is noted that the W-related peaks with
constant strength, namely WO3 [122], W [110] W3O [211], W
[200] and W [211], are due to the substrate. It can be observed
now that at all deposition temperatures Cu2WO4 is formed, which
as pointed out before, shows the chemical reactivity of the
substrate vs the gas phase. For these depositions the substrate is at
temperatures comparable to those of the W filament for hot-wire
depositions on SiLK®. Therefore, the oxidation of the substrate,
at least during the initial stages of deposition is expected, but the
XRD peak due to the substrate probably masks the one due to the
oxidation of the substrate. In any case the formation of com-
pounds such as WOx and Cu2WO4 is most probably confined near
the interface Cu/W so it does not affect dramatically the resistivity
of the Cu film.
4. Conclusions
In summary, the HWCVD of Cu films on W and SiLK®
substrates by DLI CupraSelect® vapors was presented. The
separate heating of the gas phase by the hot-wire enhanced the
growth rate by a factor of one and a half and enabled the
deposition on SiLK®, which is not possible by thermal decom-
position of the CupraSelect® vapors. High quality smooth Cu
films were obtained on SiLK® with, however, higher resistivities
than thermally grown ones. The increased values of resistivity are
due to the incomplete decomposition of the CupraSelect® vapors,
and the incorporation of W impurities from the W filament into
the Cu film. Cu–W–O compounds were formed into Cu films
grown thermally on W, which does not seem to influence signifi-
cantly its resistivity.
References
[1] T. Kodas, M. Hampten-Smith, The Chemistry of Metal CVD, VCH, Weinheim,
1994.
[2] D. Edelstein, J. Heidenreich, R. Goldbatt, W. Cote, C. Uzoh, N. Lusting,
P. Roper, T. McDevitt, W. Motsiff, A. Simon, J. Ducovic, R. Wachnick,
H. Rathor, R. Schulz, L. Su, S. Luce, J. Slattery, Tech. Dig., IEEE Int.
Electron Devices Meet. (1998) 773.
[3] P. Andricacos, The Electrochem. Soc. Interface, Spring 1999, p. 32.
[4] H.J. Boer, Solid State Technol. (March 1996) 149.
[5] J. Norman, D. Roberts, A. Hochberg, Mat. Res. Soc. Symp. Proc. 282 (1993)
347.
[6] D. Davazoglou, S. Vidal, A. Gleizes, J. Vac. Sci. Technol. B19 (3) (2001)
759.
[7] D. Davazoglou, I. Raptis, A. Gleizes, M. Vasilopoulou, J. Vac. Sci. Technol.
B22 (2004) 859.
[8] G. Papadimitropoulos, D. Davazoglou, Microel. Engin. 84 (2007) 1148.
[9] J. Norman, B. Muratore, P. Dyer, D. Roberts, A. Hochberg, J. Physique,
Lett. IV (1) (1991) C2.
[10] D.N. Kouvatsos, V. Ioannou-Sougleridis, S. Tsevas, F. Christoforou,
D. Davazoglou, C. Boukouras, Microelectron. Eng. 70 (2003) 501.
[11] J.A.T. Norman, D.A. Roberts, A.K. Hochberg, P. Smith, G.A. Petersen,
J.E. Parmeter, C.A. Apblett, T.R. Omstead, Thin Solid Films 262 (1995) 46.