Sei sulla pagina 1di 10

Chemical vapour deposition of molybdenum carbides: aspects of

phase stability
J. Lu
a,
*
, H. Hugosson
b
, O. Eriksson
b
, L. Nordstrom
b
, U. Jansson
a
a
Department of Inorganic Chemistry, A

ngstrom Laboratory, University of Uppsala, Box 538, S-751 21 Uppsala, Sweden


b
Condensed Matter Theory Group, Department of Physics, University of Uppsala, Box 530, S-751 21, Uppsala, Sweden.
Received 16 February 1999; received in revised form 12 August 1999; accepted 14 January 2000
Abstract
Thin lms of different molybdenum carbides (d-MoC
12x
, g
0
-MoC
12x
and Mo
2
C) have been deposited from a gas mixture of MoCl
5
/H
2
/
C
2
H
4
at 8008C by CVD. The H
2
content in the vapour has a strong inuence on the phase composition and microstructure. Typically, high H
2
contents lead to the formation of nanocrystalline d-MoC
12x
lms while coarse-grained g
0
-MoC
12x
is formed with an H
2
-free gas mixture.
This phase has previously only been synthesized by carburization of Mo in a CO atmosphere and it has therefore been considered as an
oxycarbide phase stabilized by the presence of oxygen in the lattice. Our results, however, show that g
0
-MoC
12x
lms containing only trace
amounts of oxygen can be deposited by CVD. Stability calculations using a FP-LMTO method conrmed that the g
0
-MoC
12x
phase is
stabilized by oxygen but that the difference in energy between e.g. d-MoC
0.75
and oxygen-free g
0
-MoC
0.75
is small enough to allow the
synthesis of the latter phase in the absence of kinetic constraints. Annealing experiments of metastable d-MoC
12x
and g
0
-MoC
12x
lms
showed two different reaction products suggesting that kinetic effects play an important role in the decomposition of these phases. q 2000
Elsevier Science S.A. All rights reserved.
Keywords: Molybdenum carbide; Carbides; Chemical vapor deposition
1. Introduction
Molybdenum carbides have many interesting properties
such as high hardness and a high catalytic activity suggest-
ing a potential use of these compounds in many thin lm
applications. The synthesis of such lms, however, requires
a possibility to deposit carbides with a well-known struc-
ture. This is not a trivial task since the MoC system is very
complex and includes a number of stable and metastable
phases as shown in Fig. 1 and Table 1 [1,2] The relationship
between the various phases can be described by stacking
sequences of close-packed Mo planes with C in interstitial
sites. For example, the hexagonal g-MoC phase is isostruc-
tural with WC and has an AAAA packing sequence. Two
high temperature phases have also been reported: the hexa-
gonal h-MoC
12x
phase with a more complex ABCACB
stacking sequence and the cubic d-MoC
12x
phase with a
NaCl-type structure (ABCABC stacking sequence of the
metal layers). The two Mo
2
C phases exhibit slightly more
complicated structures with a hexagonal high-temperature
phase with an ABAB stacking sequence of metal planes and
a disordered distribution of carbon atoms in octahedral sites.
At lower temperatures this phase transforms into one, possi-
bly two, orthorhombic phases with the Mo atoms slightly
distorted from their positions in the close-packed planes.
The stacking sequence can still, however, be described as
ABAB packing of Mo but now with an ordered distribution
of carbon in interstitial sites. Velikanova et al. have also
reported a b
00
-Mo
2
C phase which is assumed to be a more
complicated variant of the b
0
-Mo
2
C phase [2]. The structure
of this phase, however, is unknown and it has therefore not
been included in Table 1. Finally, several authors have
reported the existence of a metastable phase, g
0
-MoC
12x
,
with a hexagonal AABB stacking sequence [3,4].
The relative stability and the true nature of all the phases
shown in Table 1 are yet not completely known and several
different phase diagrams of the MoC system have been
presented [1,2]. For example, earlier studies by e.g. Rudy
et al. showed that suggested that the equilibrium phases at
50 at.% C are Mo
2
C and C (i.e. the hexagonal g-MoC phase
(WC-type) is not an equilibrium phase) [1]. In contrast, Fig.
1 shows a more recent phase diagram from Velikanova and
co-workers, where g-MoC is indeed an equilibrium phase
and that it should be included in the phase diagram [2]. The
nature of the metastable g
0
-MoC
12x
phase is also not clear.
Thin Solid Films 370 (2000) 203212
0040-6090/00/$ - see front matter q 2000 Elsevier Science S.A. All rights reserved.
PII: S0040-6090(00)00750-1
www.elsevier.com/locate/tsf
* Corresponding author.
E-mail address: jun.lu@angstrom.uu.se (J. Lu).
In the literature it is generally assumed that this phase is
stabilized by oxygen and that it therefore more correctly
should be described as an oxycarbide [1]. This is supported
by the fact that the g
0
-MoC
12x
phase mainly has been
formed by carburization of metallic molybdenum in a CO
atmosphere [3,4].
Molybdenum carbide lms can be deposited by several
different techniques including sputtering and other PVD
methods (see e.g. Refs. [57]). Several authors have also
deposited molybdenum carbide lms by chemical vapour
deposition (CVD) using Mo(CO)
6
and different types of
activation (heat, laser, electron beam, synchrotron radiation)
[8,9]. A problem with Mo(CO)
6
as a metal precursor,
however, is that the lms are easily contaminated by
oxygen. In fact, it has been shown that the decomposition
of Mo(CO)
6
tends to give not a carbide but an oxycarbide
lm [9]. An alternative molybdenum source for carbide
CVD is MoCl
5
which has successfully been used for CVD
of metallic molybdenum at temperatures as low as 5008C
[10]. Norin et al. have also demonstrated that molybdenum
carbide lms can be deposited using MoCl
5
and C
60
as
precursors [11]. We have also recently used CH
2
I
2
to deposit
lms of nanocrystalline d-MoC
12x
at 6008C [12]. An inter-
esting observation is that most crystalline molybdenum
carbide lms prepared by CVD and PVD contain either
Mo
2
C and/or the cubic d-MoC
12x
phase. To the knowledge
of the authors, there are no reports on thin lm growth of h-
MoC
12x
, g-MoC or g
0
-MoC
12x
.
The objective with this work was to study the CVD of
molybdenum carbide lms from a reaction gas mixture of
MoCl
5
/C
2
H
4
/H
2
with a special emphasis on the phase forma-
tion and phase stability. An important part of the work has
been to determine conditions for the deposition of oxygen-
free carbide lms. Theoretical calculations of the total ener-
gies have been carried out to support experimental observa-
tions.
2. Experimental
A horizontal hot-wall quartz reactor (2.5 cm diameter
reactor tube) was used for the molybdenum carbide deposi-
tion. A schematic view of the deposition system can be seen
in Fig. 2. The system was pumped by a mechanical pump
yielding a base pressure of about 10
23
Torr. To minimize oil
back diffusion, a liquid nitrogen trap was placed between the
reactor and the pump. The total pressure in the system was
measured by a conventional capacitance manometer and
was kept constant to 1.7 Torr in most experiments. The
leak rate of the system was regularly checked with a leak
a detector and by pressure increase measurements and was
determined to be less than 10
25
Torr l/s. This leak corre-
sponds to a contamination level of less than 2 ppm.
The molybdenum precursor was introduced into the reac-
tor from a sublimator containing MoCl
5
powder .99%
purity) and by using Ar .99:9999% as a carrier gas.
The sublimator as well as all parts between the sublimator
and the furnace were heated to about 80858C by heating
tapes. The vapour pressure of MoCl
5
at 858C is 0.3 Torr. It is
important to note that MoCl
5
is a very air-sensitive
compound. The MoCl
5
powder was therefore loaded into
the sublimator under Ar atmosphere in a glove box. The
H
2
(99.9999%) and C
2
H
4
(99.95%) were introduced into
the reactor by a separate tube. The gas ows and total pres-
sure were controlled by mass ow controllers and a throttle
valve, respectively. The total gas ow rate was kept constant
to 360 sccm in all experiments while the ow rate of Ar
through the sublimator was kept constant to 180 sccm.
Variation in the ows of H
2
and C
2
H
4
were compensated
by additional Ar. Finally, the temperature was kept constant
at 8008C in most experiments.
Small pieces of single-crystal Al
2
O
3
with a (102)-orienta-
J. Lu et al. / Thin Solid Films 370 (2000) 203212 204
Fig. 1. The MoC phase diagram.
Table 1
Name, structure and stacking sequence of metallic planes for different
molybdenum carbide phases
Phase Structure Stacking sequence
b-Mo
2
C Hexagonal ABAB
b
0
-Mo
2
C Orthorhombic ABAB
h-MoC
12x
Hexagonal ABCACB
d-MoC
12x
Cubic ABCABC
g-MoC Hexagonal AAAA
g
0
-MoC
12x
Hexagonal AABB
Fig. 2. Schematic view of the hot-wall CVD system for MoC deposition.
tion were used as substrates. Preliminary experiments
showed that lms deposited on mirror-like substrate
surfaces yielded lms with a poor adhesion. For this reason,
the lms were deposited on non-polished Al
2
O
3
. Films
deposited on these substrates gave lm with a very good
adhesion. Prior to a deposition experiment, the substrates
were cleaned in acetone and placed into the reactor from the
end of the system with an Ar ush to prevent air contamina-
tion of the system. All the substrates were placed in the
system parallel to the gas ow.
The phase composition of the lms were determined by
X-ray diffraction (XRD) using a Siemens D5000 diffract-
ometer with Cu K
a
radiation using both u2u and glancing
angle geometry. For the cubic d-MoC
12x
phase the peak
positions and relative intensities were calculated by the
LAZY PULVERIX program using positions for Mo and C
in the NaCl-type structure (space group Fm3m) as input.
The cubic cell parameter a was set to 4.26 A

and the calcu-


lations were carried out with a 70% occupancy of the carbon
atom positions.
The chemical composition of the carbide lms was inves-
tigated with X-ray photoelectron spectroscopy (XPS) and a
scanning Auger microprobe (SAM) in a combined PHI 5500
Multitechnique instrument equipped with ion pumps.
Monochromated Al K
a
radiation was used in the XPS analy-
sis. The binding energy scale was calibrated by setting the
C1s peak for adsorbed hydrocarbons to 284.6 eV. The peaks
were deconvoluted using a Gaussian (80%)Lorentzian
function. The XPS data was used to estimate the composi-
tion of the lms using elementary sensitivity factors. These
factors were obtained from a standard MoC
0.8
sample with a
composition determined by Rutherford backscattering spec-
troscopy (RBS). Possible overlapping of the O1s signal
from features originating from Mo3s made it difcult to
analyse the low oxygen content in the lms with XPS.
Oxygen analyses were therefore carried out with SAM in
the same chamber using a beam current of 50 nA. Prior to
XPS or SAM analyses, surface oxides formed during trans-
port from the reactor to the analysis chamber were removed
in situ by argon ion sputtering using a 4 keV acceleration
voltage and a sputtering current of 6 mA. Finally, cross-
sections of the carbide lms were studied with transmission
electron microscopy (TEM) on a JEOL 2000 FXII instru-
ment with a 200 kV working voltage. The cross-sections
were prepared by gluing two pieces of the lms together
face-to-face. This sample was cut, polished, dimpled and
nally ion-milled. Film thickness was measured by TEM
and X-ray uorescence spectroscopy (XRFS). The latter
technique gives the total amount of deposited Mo atoms/
cm
2
. In a single-phase lm, this value give can be converted
into a true lm thickness providing that the lms are
uniform without porosity. In this paper, porosity has been
estimated by calculating an approximate thickness from
XRFS results (assuming a MoC density of 9.5 g/cm
3
) and
compare these values with actual thicknesses observed in
TEM.
3. Theoretical calculations
The stabilities of three carbide phases (g-MoC, d-
MoC
12x
and g
0
-MoC
12x
) were calculated within rst-prin-
ciples density functional theory using a full potential linear
mufn tin orbital method (FP-LMTO) [13]. This method
allows us to calculate the total energy as a function of
volume for the selected phases. The total energy vs. volume
curve was subsequently tted to a Murnaghan equation of
state to nd theoretical ground-state energies E
s
and equili-
brium volumes V
calc
for each structure. The calculations
were carried out for two compositions MoC
1.0
and MoC
0.75
to evaluate the inuence of vacancy concentration on the
stability. The energies of the stoichiometric and the substoi-
chiometric phases were calculated by taking the energy of
the stoichiometric phase, MoC
1.0
, and comparing it with the
energy of the substoichiometric phase, MoC
12x
, plus the
energy of x atomic fraction of carbon in the graphite struc-
ture. The energy of the carbon atom in graphite was taken to
be 1.029 keV per atom. A more detailed description of the
calculations will be described elsewhere [14].
The results from the calculations are shown in Fig. 3. As
can be seen the relative stability is dependent on the carbon
content. The most stable stoichiometric phase is the hexa-
gonal g-MoC (Fig. 3a). This phase is about 0.49 eV (about
47 kJ/mol Mo) more stable than g
0
-MoC
1.0
. The energy
separation between g
0
-MoC
1.0
and the least stable of the
three phases (d-MoC
1.0
) is about 0.08 eV (about 8 kJ/mol
Mo).
As can be seen in Fig. 3b, the three different phases have
similar energies for the substoichiometric composition
MoC
0.75
although the relative order of stability has been
changed. The least stable carbide is now the hexagonal g-
MoC. The energy of this phase has increased with about
J. Lu et al. / Thin Solid Films 370 (2000) 203212 205
Fig. 3. Calculated total energies vs. volume for (a) stoichiometric composi-
tion and (b) sub-stoichiometric MoC
0.75
. W, g-MoC
12x
, K, d-MoC
12x
, A,
g
0
-MoC
12x
.
0.62 eV (about 60 kJ/mol Mo) compared to the stoichio-
metric composition. This result is expected since the g-
MoC phase is known to exhibit a very narrow homogeneity
range (see Fig. 1). In contrast, the energy of d-MoC
0.75
decreases with about 0.1 eV (about 9 kJ/mol Mo) compared
to the stoichiometric composition, making this phase the
most stable of the three at this composition. The decrease
in energy is in agreement with the phase diagram in Fig. 1
which shows that this phase exhibits a fairly wide homo-
geneity range. Finally, we can see that substoichiometric g
0
-
MoC
0.75
is less stable than the stoichiometric composition of
this phase. The energy difference, however, is only found to
be 0.05 eV (about 5 kJ/mol Mo) making this phase slightly
less stable than the substoichiometric d-MoC
0.75
.
The results in Fig. 3 can be used to determine theoretical
volumes, V
calc
, and calculate bulk moduli, B
t
, for the differ-
ent phases. As can be seen in Table 2, the calculated
volumes are in good agreement with experimental data.
The bulk modulus, B
t
, which can be calculated from the
Murnaghan equation, shows values ranging from 330 GPa
to about 380 GPa depending on phase and carbon content.
These values are in agreement with values obtained for
other transition metal carbide phases.
It has previously been suggested that the g
0
-MoC phase
is stabilized by oxygen and therefore should not be
considered as a true carbide. We have therefore carried
out calculations for g
0
-MoC
0.75
O
0.25
and compared the
stability of this oxycarbide with g
0
-MoC. In this calcula-
tion, the stability of the oxycarbide system was related to
the pure carbide by comparing the energy of the system
g
0
-MoC
1:0
10:125 O
2
with the energy of g
0
-
MoC
0:75
O
0:25
10:25 C. The calculations show that the
addition of oxygen to g
0
-MoC actually stabilizes this
phase with about 0.88 eV (about 86 kJ/mol Mo). This
certainly indicates that the observations of a g
0
-MoC
phase which have been reported in the literature can be
due to the formation of an oxycarbide. However, the total
energies calculated in Fig. 3 show that the energy differ-
ence between g
0
-MoC
0.75
and d-MoC
0.75
is fairly small
(about 6.6 kJ/mol Mo). It is therefore not unlikely that
oxygen-free lms of the g
0
-MoC
12x
phase could possibly
be synthesized.
4. Experimental results
4.1. General observations
In a series of preliminary experiments, the reactants were
mixed prior to the furnace and allowed to enter a reactor
heated to 8008C. These experiments yielded unacceptable
deposition proles where rather thick lms of only metallic
Mo were deposited at the reactor entrance. The deposition
rate of the Mo quickly decreased and very thin carbide lms
could only be deposited about 20 cm into the heated parts of
the reactor. The explanation for this growth behaviour is that
MoCl
5
is a very reactive molecule which is reduced to
metallic Mo by H
2
already at about 5008C [10]. In contrast,
C
2
H
4
is a much less reactive molecule which has to be
thermally activated to higher temperatures prior to any
substantial decomposition. This difference in reactivity
leads to a fast, preferential reaction of MoCl
5
and an unac-
ceptable depletion of this reactant from the vapour. This
effect could be reduced by using a design shown in Fig. 2
where the C
2
H
4
/H
2
/Ar mixture is transported separately in
an inner tube and mixed with MoCl
5
about 30 cm into the
reactor. This set-up clearly reduced the MoCl
5
depletion and
made it possible to deposit carbon-rich carbide lms at a
distance of a few cm after the gas mixing point.
Further experiments with this set-up showed that the
deposition of molybdenum carbide is strongly dependent
on temperature and pressure. Carbide lms with a good
adhesion are easily formed at 8008C and above. Tempera-
tures below 8008C, however, quickly reduce the deposition
rate and gave lms with a low carbon content (Mo
2
C and
Mo). It was also found that a reduction in pressure yielded a
wider deposition zone and suppressed the codeposition of
free carbon. For these reasons, most experiments in this
study were carried out at 8008C and 1.7 Torr.
The experiments also showed that the hydrogen content
in the reaction gas mixture has a dramatic effect on the
deposition process. In fact, it was found that the partial
pressure of hydrogen could be used to control the phase
compositions of the carbide lms. For example, high H
2
content in the vapour .30% yielded a broad deposition
zone of d-MoC
12x
in the centre of the reactor at 8008C. A
reduction of the H
2
content to less than 10% changed the
phase composition of the lms in the centre of the reactor to
Mo
2
C. An interesting observation was that carbide lms
could also be deposited without the presence of H
2
into
the vapour. A mixture of only MoCl
5
and C
2
H
4
yielded
lms consisting of mainly the metastable g
0
-MoC
12x
phase. A general observation was that higher H
2
content
favoured a fast deposition of molybdenum at the gas inlet
region. This effect increased the deposition rate of molyb-
denum-rich phases (Mo and Mo
2
C) close to the gas-mixing
point in the reactor. A more even prole was obtained with
decreasing H
2
pressures. Finally, a rather surprising obser-
vation was that the phase composition was rather unaffected
by changes in the C
2
H
4
content. For example, C
2
H
4
concen-
J. Lu et al. / Thin Solid Films 370 (2000) 203212 206
Table 2
Experimental volume (V
exp
), calculated volume (V
calc
) and calculated bulk
moduli (B
t
) for different phases and compositions
Phase V
exp
(A

3
) V
calc
(A

3
) B
t
(GPa)
g-MoC
1.0
20.430
a
20.255 381
d-MoC
1.0
20.261 364
g
0
-MoC
1.0
20.391 369
g-MoC
0.75
19.838 333
d-MoC
0.75
19.614
a
19.396 337
g
0
-MoC
0.75
20.422
b
19.697 330
a
Ref. [1].
b
This study.
trations ranging from about 0.6% up to about 15% yielded
only lms consisting of d-MoC
12x
in the centre of the reac-
tor (the H
2
content was 25% in these experiments).
4.2. Deposition of nanocrystalline d-MoC
12x
As mentioned above, lms of cubic d-MoC
12x
can be
deposited in a wide deposition zone from gas mixtures
containing substantial amounts of hydrogen. The deposition
rate of this phase is favoured by a high H
2
content in the
vapour. Growth rates of about 1 mm/h can easily be attained
at 8008C using a gas mixture of 30% H
2
. No attempts,
however, have been carried out to optimize the growth
rate and it is likely that considerably higher growth rates
can be attained by a tuning of the deposition process.
XPS analyses showed that the d-MoC
12x
lms were
chlorine-free. Analysis with SAM on freshly prepared
lms showed that the oxygen content was less than about
0.5 at.% (close to the detection limit). It should be noted,
however, that the lms probably contain even less oxygen
since the samples are oxidized during transport to the analy-
sis chamber. This gives a small contribution in the oxygen
analysis also after argon ion sputtering due to e.g. shadow-
ing effects.
The XPS analysis also showed that the d-MoC
12x
lms
always contained free carbon. This can be seen in the C1s
XPS spectra showing two peaks at 284.3 and 283.2 eV. The
latter peak is due to carbidic carbon while the peak at 284.3
eV can be attributed to CC (or CH) bonds from free
carbon. The exact nature of this carbon cannot be deter-
mined from XPS spectra. It can be graphite, turbostratic
carbon, amorphous carbon or some other type of carbon
with more or less aliphatic and/or aromatic character. In
this following this carbon will be described as `free carbon'.
An interesting observation was that the surface of the lms
contained large amounts of free carbon, which could be
removed, but not completely eliminated, by prolonged in
situ argon ion sputtering. In general, a sputtering time
ranging from a few minutes to more than 10 min was
required to reduce the relative amount of free carbon to a
`steady-state' value which we have identied as the bulk
content. The bulk content of free carbon (given by the inten-
sity ratio between the peaks at 284.3 and 283.2 eV) was
estimated to vary from about 11 to 2 at.% for different
samples. The bulk content of free carbon as well as the
thickness of the carbon-rich layer were found to be depen-
dent on experimental conditions used during lm growth.
Thicker surface layers and higher bulk content of free
carbon were obtained by decreasing the C
2
H
4
partial pres-
sure or/and by decreasing the H
2
content in the vapour. The
formation of free carbon was also dependent on the substrate
position in the reactor. In general, the free carbon content
(both in the bulk and on the surface) increased further down-
stream in the reactor.
As mentioned above, the total C/Mo ratio in the bulk of
the lms was dependent on the experimental conditions. For
example, XPS analysis of a typical d-MoC
12x
lm deposited
in the centre of the reactor using a gas mixture of 33% H
2
and 17% C
2
H
4
showed a total C/Mo ratio of 0.79. The bulk
content of free carbon in this lm could be estimated to
about 5 at.% from the relative intensities of the peaks at
284.3 and 283.2 eV after sputtering of the surface. This
gives a total carbon content in the carbide of about 41
at.% corresponding to a composition of d-MoC
0.69
.
Analyses of a number of lms showed that no signicant
variation in the carbide stoichiometry is obtained at different
vapour compositions and substrate positions. The carbon
content in the carbide will affect the cell parameter of the
cubic cell and an estimation of the carbon content can be
therefore be carried out from XRD data providing that the
relationship between composition and cell axis is known.
The cell parameters for the compositions MoC
0.69
and
MoC
0.75
have been determined by Rudy et al. to 4.266 A

and 4.281 A

, respectively [1]. The d-MoC


12x
lms depos-
ited in this study typically exhibit a cell-axis of 4.260 A

.
This corresponds to a composition of about MoC
0.67
(40.1
at.% C) using a linear extrapolation of the data in Ref. [1].
No signicant variation in the cell parameter was found at
different vapour compositions. However, it was found that
the cell parameter of d-MoC
12x
will increase slightly at
higher total pressures.
Fig. 4a shows an X-ray diffractogram of a typical lm
where all peaks can be attributed to d-MoC
12x
[15]. The
diffractogram show that the lms are generally almost
texture-free although a slight (110)-texture could be
observed for lms deposited at low ethene concentrations.
A general observation, however, was that the diffraction
peaks are very broad. This can be due to several factors,
such as small grain sizes, stresses in the lms and composi-
tion variations. As will be shown below most of the line-
broadening can be attributed to a small grain size. The
average grain diameter can be estimated to about 6070 A

using the Scherrer formula [16]. No indication of graphite


was found in the X-ray diffractograms.
J. Lu et al. / Thin Solid Films 370 (2000) 203212 207
Fig. 4. X-ray diffractograms of (a) a d-MoC
0.67
lm deposited at 8008C for
30 min and (b) the Mo
2
C lm formed by annealing of the d-MoC
0.67
lm in
vacuum for 1 h at 10008C. d, d-MoC
0.67
; b, b-Mo
2
C.
The ne-grained microstructure of the d-MoC
12x
lms
was conrmed by cross-section TEM. Fig. 5 shows a typical
TEM micrograph of these lms with an average grain size of
50100 A

. The interface between the carbide lm and the


substrate was sharp and no indication of a lm/substrate
reaction could be observed at higher magnications. The
average grain size was constant in all parts of the lm.
This growth behaviour is in contrast to d-MoC
12x
lms
deposited from MoCl
5
/CH
2
I
2
/H
2
gas mixtures which exhibit
a zone of smaller grains close to the substrate material [12].
Finally, it should be noted that the grain size was affected by
the deposition conditions. In general, smaller grains were
obtained at high ethene concentrations. A comparison of the
lm thickness given by TEM and the amount of Mo atoms/
cm
2
given by XRFS shows that the lms deposited at with
less that 10% C
2
H
4
in the vapour are dense (porosity ,5%).
It was found, however, that a reduction of the C
2
H
4
content
below 10% yielded lms with a more porous microstructure.
Details on the inuence of process parameters on the micro-
structure of the d-MoC
12x
lms will be presented elsewhere
(J. Lu, U. Jansson, unpublished data).
4.3. Deposition of g-
0
-MoC
12x
As mentioned in Section 4.1, carbide lms can also be
deposited without any H
2
in the vapour. The absence of H
2
,
however, has a dramatic effect on the phase composition. As
can be seen in Fig. 6, XRD shows that the lms now mainly
consist of the hexagonal g
0
-MoC
12x
phase [17]. The growth
rate of this phase is about 30% lower than for the deposition
of d-MoC
12x
in an H
2
atmosphere.
XRD analyses of a series of lms showed cell parameters
of a 2:93 A

and c 10:98 A

. The total C/Mo ratio was


determined to about 0.78 by XPS. CC bonds could be
observed in the C1s spectra suggesting that about 2 at.%
free carbon is formed together with the g
0
-MoC
12x
phase.
This corresponds to a composition of about g
0
-MoC
0.74
. An
interesting observation was that the oxygen content in the
lms was very low. In fact, analysis with AES showed
oxygen levels close to the detection limit for this technique
(,0:5 at.% O). These oxygen levels are similar to those
observed in pure Mo or d-MoC
12x
lms prepared by this
process and may due to the fact that an surface oxide is
formed on the surface during transport in air from the
deposition reactor to the analysis chamber. For example,
shadowing effects, make it difcult to completely remove
all traces of oxygen also during prolonged Ar
1
sputtering.
Furthermore, no chlorine was observed in the lms.
The X-ray diffractograms exhibit fairly narrow diffraction
peaks suggesting that the g
0
-MoC
12x
phase has a larger
grain size than the d-MoC
12x
lms. This was also conrmed
by TEM observations which showed grain sizes of about
2000 A

. As can be seen in Fig. 7, these grains extend


through the entire lm from the surface to the substrate
region. An interesting observation in Fig. 7, however, is
that much smaller equiaxed grains are formed at the inter-
face. Electron diffraction shows that these grains consist of a
thin layer of the d-MoC
12x
phase with an average grain size
J. Lu et al. / Thin Solid Films 370 (2000) 203212 208
Fig. 5. TEM cross-section of a 3000-A

thick d-MoC
0.67
lm deposited at
8008C. Deposition time: 30 min.
Fig. 6. X-ray diffractograms of (a) a g
0
-MoC
0.74
lm deposited at 8008C, (b)
after annealing of the g
0
-MoC
0.74
lm in vacuum at 10008C for 1 h and (c)
annealing of the g
0
-MoC
0.74
lm in C
2
H
4
at 10008C for 3 h.
Fig. 7. Cross-section TEM micrograph of a 2500-A

thick g
0
-MoC
0.74
lm
deposited at 8008C. d and g
0
represent d-MoC
12x
and g
0
-MoC
12x
, respec-
tively.
of about 250300 A

. Furthermore, the TEM micrographs in


Fig. 8 show a lamellar-like contrast which can be observed
in all g
0
-MoC
12x
grains. The average distance between the
line-like structures varies not only from grain to grain but
also within individual grains. A typical distance in many
parts of the grains are about 11 A

. This corresponds to the


AABB stacking sequence of the unit cell in g
0
-MoC
12x
. The
variations in line distance in other parts of the grains suggest
that also other stacking sequences exist. This conclusion is
supported by electron diffractograms showing elongated
diffraction spots in the direction of the c*-axis which can
be attributed to stacking faults of the Mo layers. Finally, a
comparison of the lm thickness given by TEM and the
amount of Mo atoms/cm
2
given by XRFS shows that the
lms are dense (porosity ,5%).
4.4. Annealing experiments
As mentioned above the cubic d-MoC
12x
phase is only
thermodynamically stable at temperatures above 19568C,
while the g
0
-MoC
12x
phase should be metastable at all
temperatures. A series of annealing experiments were there-
fore carried out to determine the decomposition route for
these phases. Diffractogram b in Fig. 6 shows that annealing
of g
0
-MoC
0.74
at 10008C for 3 h in vacuum leads to a decom-
position of the g
0
-phase and subsequent formation g-MoC
and b-Mo
2
C [18,19]. The formation of the b
0
-phase could
be excluded since no splitting of diffraction lines was
observed in the X-ray diffractograms. However, XRD
shows that some g
0
-MoC
12x
still remain in the lm and
that equilibrium is not yet reached. The amount of g
0
-
MoC
12x
phase in the lm can be further reduced slightly
by a longer annealing for e.g. 20 h at 10008C but it cannot
be completely eliminated. The amount g-MoC formed can
be further increased by annealing in a carburizing atmo-
sphere. Diffractogram c in Fig. 6 shows that annealing in
2 Torr C
2
H
4
leads to an almost complete transformation of
the g
0
-MoC
12x
phase to mainly g-MoC and small amounts
of b-Mo
2
C.
Annealing of a d-MoC
0.67
lm (about 40.1 at.% C) at
identical conditions show a quite different behaviour. As
can be seen in the XRD of Fig. 4b, this lm decomposes
into b-Mo
2
C. Furthermore, the diffraction lines of the b-
Mo
2
C phase are sharp compared to those from the d-
MoC
0.67
lms. This suggests that a substantial grain growth
must have occurred during the phase transformation. This
was also conrmed by optical microscopy which showed
carbide grains with diameters ranging from about 0.5 to 1
mm. Mo
2
C contain less carbon than d-MoC
0.67
and a phase
transformation to the former carbide must therefore also
lead to the formation another carbon-containing phase. No
other carbide or graphite could, however, be seen in the
diffractograms. XPS analysis also showed a small decrease
in the bulk carbon content to about 36 at.%. It was found,
however, that the surface region of the annealed lms
contained considerably larger amounts of free carbon and
that a longer sputtering time was required to remove the
carbon-rich surface layer on the annealed lms compared
to as-deposited d-MoC
12x
lms. It can therefore be
concluded that the annealing lead to a segregation of carbon
to the lm surface and the formation of a Mo
2
C lm with a
surface layer of amorphous carbon and/or graphite. Anneal-
ing of d-MoC
12x
in C
2
H
4
give the same result as vacuum
annealing.
5. Discussion
The results show that three different carbides (d-MoC
12x
,
g
0
-MoC
12x
and b-Mo
2
C) can be deposited from a MoCl
5
/
C
2
H
4
/H
2
gas mixture in a conventional hot-wall CVD reac-
tor. Our results can be separated into two parts: (i) phase
stabilities in the MoC system and (ii) process characteris-
tics. In the following, the two parts will be treated sepa-
rately.
5.1. Phase stabilities in the MoC system
As mentioned earlier, there are still some unresolved
questions in the MoC system. For example, it has been
suggested that the g
0
-MoC
12x
phase is stabilized by oxygen
and rather should be considered as an oxycarbide than a true
binary carbide [1]. Earlier phase diagrams have also not
included g-MoC as stable carbide although a recent phase
diagram by Velikonova and coworkers claim that this
compound is thermodynamically stable [2]. To increase
our understanding of the MoC system, we have carried
out some theoretical calculations of phase stabilities using
an FP-LMTO method. The calculations only include a
comparison of stabilites for three phases at two different
compositions and can therefore not be used to construct a
complete, theoretical phase diagram. The results conrm
that the g
0
-MoC
12x
phase indeed is stabilized by oxygen.
It is therefore likely that the synthesis of this phase in the
presence of oxygen (e.g. carbonization of Mo in a CO atmo-
sphere) could lead to an oxycarbide phase containing
J. Lu et al. / Thin Solid Films 370 (2000) 203212 209
Fig. 8. TEM micrograph of lamellar-like structure in a g
0
-MoC
0.74
lm
deposited at 8008C.
substantial amounts of oxygen. Our results, however, show
that g
0
-MoC
0.75
is only slightly less stable than d-MoC
0.75
(about 5 kJ/mol Mo). The latter phase is well-known and has
been synthesized as a metastable phase in many thin lm
processes. It is therefore conceivable to assume that also the
g
0
-MoC
12x
should be possible to prepare in similar
processes. This was also conrmed by our experimental
results which showed that g
0
-MoC
0.74
could be deposited
when H
2
was completely eliminated from the reaction gas
mixture. AES analysis shows that these lms contain trace
amounts of oxygen (,0:5 at.%). The oxygen content is
close to the detection limit and on the same level as those
found in other metal and metals carbide samples where a
surface oxide has been removed by sputtering. The results
suggest that g
0
-MoC
12x
can be described as a true carbide
but it cannot, of course, be excluded that trace amounts of
oxygen is required to stabilize the growth of this metastable
phase. It should also be noted that the g
0
-MoC
12x
lms
prepared in our study contain many stacking faults.
The XRD analysis showed that the as-deposited carbon-
rich lms contained either the d-MoC
12x
or g
0
-MoC
12x
together with small amounts of free carbon. This means
that the equilibrium phases, as shown by any of the
previously published phase diagrams, are never deposited.
It is interesting, however, that d-MoC
12x
and g
0
-MoC
12x
follow different decomposition pathways during a
prolonged annealing at 10008C. The results show that nano-
crystalline d-MoC
12x
decompose to more coarse-grained b-
Mo
2
C and free carbon which seems to segregate to the lm
surface and the grain boundaries. These two phases are close
to the equilibrium phases in the earlier phase diagram
published by Rudy et al. (b
0
-Mo
2
C 1graphite) [1]. In
contrast, annealing of g
0
-MoC
12x
show a more complicated
behaviour. In vacuum, this phase decompose into b-Mo
2
C
and g-MoC, which are considered as equilibrium phases in
the most recent phase diagram by Velikanova et al. (Fig. 1)
[2]. The transformation of g
0
-MoC
12x
to g-MoC is under-
standable from a kinetic point of view. The difference
between the two phases is mainly a stacking fault between
the close-packed Mo layers. However, the lower carbon
content in the g
0
-MoC
12x
phase (g
0
-MoC
0.74
) compared to
g-MoC requires a signicant diffusion of carbon and/or
metal atoms as the b-Mo
2
C is formed. This can be avoided
by annealing in a carbon-containing atmosphere (e.g. C
2
H
4
).
This also leads to the formation of more g-MoC and less b-
Mo
2
C as shown in Fig. 6. Our results suggest that the two
different phase diagrams obtained by Rudy et al. and Veli-
kanova and coworkers can be due to the carbide phases
present in their samples prior to annealing.
5.2. Process characteristics
The results show that a temperature of 8008C was
required to deposit carbon-rich lms (d-MoC
12x
and g
0
-
MoC
12x
) from a MoCl
5
/C
2
H
4
/H
2
mixture. This is in contrast
to other carbon precursors such as Mo(CO)
6
and CH
2
I
2
which allow carbide deposition at 6008C and below
[8,12]. The results also show that MoCl
5
is more reactive
than C
2
H
4
and that the halide precursor quickly can be
reduced to metallic Mo and depleted from the vapour with-
out any substantial carbide formation if the reactants are
mixed prior to the heated zone of the reactor. This problem
could partly be solved by mixing the gases about 30 cm into
the reactor as shown in Fig. 2. However, to understand our
results we have to remember that a deposition prole of
different phases are formed also with this set-up providing
that H
2
is present in the vapour. Mo-rich phases (Mo and
Mo
2
C) are then formed immediately after the gases are
mixed and the more carbon-rich d-MoC
12x
phase is formed
only at some distance further into the reactor together with
small amounts of `free carbon'. The d-phase has a more or
less constant composition of d-MoC
0.67
but the amount of
free carbon is higher further into the reactor. This means that
the overall C/Mo ratio in the lms increase continually
downstream into reactor.
The experiments with H
2
in the vapour show that we have
a narrow transition zone in the reactor where the phase
composition changes from Mo-rich phases to d-MoC
12x
(and some free carbon). Increasing H
2
content pushes this
transition zone closer to the gas mixing point. This means
that we, in the centre of the reactor, will deposit d-MoC
12x
lms at high H
2
content, while Mo
2
C will deposited in the
centre of the reactor at lower H
2
content. This behaviour can
be understood from the differences in chemical reactivity
MoCl
5
and C
2
H
4
described in Section 4.1. The high reactiv-
ity of MoCl
5
in the presence of H
2
will lead to a fast initial
deposition of metallic Mo in the gas inlet region as soon as
the gases mix. In contrast, reaction with C
2
H
4
is much
slower and lms deposited in the gas-mixing area will
mainly contain Mo with small amounts of Mo
2
C. As a
consequence, MoCl
5
will be depleted from the vapour and
more carbon-rich lms containing d-MoC
12x
and free
carbon will be deposited in the centre of the reactor.
Lower H
2
content will reduce the fast deposition of Mo at
the reactor entrance and give higher MoCl
5
partial pressures
further downstream. The higher MoCl
5
/C
2
H
4
ratio will push
the Mo
2
C/d-MoC
12x
transition point further downstream
and yield lms of Mo
2
C in the centre of the reactor (and a
more even deposition prole). At all H
2
concentrations,
however, the total C/Mo content in the lms will increase
as a function of distance from the gas inlet. The prole can
be almost eliminated by the use of an H
2
-free reaction
mixture which reduces the MoCl
5
reduction rate. The prole
can also be reduced by a decrease in total pressure. This
effect has also been observed in the CVD of WC from WF
6
/
C
3
H
8
/H
2
and has been attributed to the fact that lower pres-
sures will reduce the decomposition rate of the metal halide
in the gas inlet region [20].
Our results also show that all d-MoC
12x
lms have a
constant carbide composition of about d-MoC
0.670.69
(the
amount of free carbon can vary in these lms). It is inter-
esting to note that the homogeneity range for d-MoC
12x
in
J. Lu et al. / Thin Solid Films 370 (2000) 203212 210
the phase diagram become more narrow and almost form a
line phase as the temperature is reduced. A rough extrapola-
tion of the boundary lines in Fig. 1 to the temperatures used
in this study give a composition very close d-MoC
0.67
. It is
possible that this composition is the most stable (or least
unfavourable) at these metastable deposition conditions.
The results also show that lms deposited from the
MoCl
5
/C
2
H
4
/H
2
mixture were dense with a porosity of less
than 5% at C
2
H
4
concentrations .10%. At lower C
2
H
4
content, however, a more porous microstructure was
evolved. This is in contrast to nanocrystalline d-MoC
12x
lms obtained with CH
2
I
2
as a carbon source [12]. These
lms were found to have a large amount of pores (up to 40%
of the total volume). The phase composition of the carbide
lms was also affected by the vapour composition. Surpris-
ingly, only the d-MoC
12x
phase was deposited together with
some free carbon in a wide C
2
H
4
partial pressure range (see
Section 4.1). The most important effect of higher C
2
H
4
content in the vapour were the formation of less free carbon
and a reduction in grain size of the nanocrystalline lms.
The inuence of C
2
H
4
on the deposition process in general
and the grain size in particular will be presented elsewhere.
It should be noted, however, that the somewhat surprising
observation that the free carbon content is higher at low
C
2
H
4
content was found to be due to a more porous micro-
structure yielding a larger surface area on which a carbon-
rich surface layer easily was formed (J. Lu, U. Jansson,
unpublished data).
An interesting observation was that the growth behaviour
and the nal phase composition in the lms were strongly
dependent on the hydrogen content in the vapour. For exam-
ple, deposition without any H
2
(i.e. a MoCl
5
/C
2
H
4
gas
mixture) gave coarse-grained lms consisting of mainly
the g
0
-MoC
12x
phase. In contrast, nanocrystalline d-
MoC
12x
lms were deposited at H
2
content higher than
10%. In both cases, the total carbon content in the lms
was almost identical although the deposition rate was
much higher with H
2
in the vapour. An important effect of
H
2
was therefore to switch growth mode from coarse-
grained g
0
-MoC
12x
to nanocrystalline d-MoC
12x
. The
reason for this behaviour is yet unclear but it is conceivable
to assume that the effect is due to some critical interactions
between hydrogen and either a molybdenum halide species
or a hydrocarbon species. Further studies involving in situ
analysis with e.g. mass spectrometry or surface analysis
techniques are required to give a detailed understanding of
this process.
6. Concluding remarks
Several different molybdenum carbide phases can be
deposited at 8008C from a MoCl
5
/C
2
H
4
/H
2
mixture. This
temperature is higher than for other carbon precursors
such as Mo(CO)
6
and CH
2
I
2
which can be a limitation for
certain applications. An advantage with C
2
H
4
, however, is
that the lms are much more dense (low porosity),
compared to lms produced by CH
2
I
2
. Another observation
is that the grain size of the nanocrystalline d-MoC
12x
lms
is affected by e.g. the C
2
H
4
partial pressure. This observa-
tion, together with the fact that free carbon (presumably
present in a non-crystalline form) is deposited together
with the carbide, suggests that it should be possible to
synthesize composite lms of nanocrystalline carbide grains
in an amorphous carbon (a-C) matrix. It has recently been
suggested that such lms may have unique mechanical and
tribological properties [21]. Such properties have also been
reported for nanocrystalline TiC/a-C lms [22]. It is possi-
ble that similar lms of a nanocrystalline d-MoC
12x
/a-C
composite structure can be deposited. Further studies,
however, are required to clarify the relationship between
experimental parameters and microstructure.
We have also for the rst time been able to deposit thin
lms of g
0
-MoC
12x
and g-MoC (after annealing in a
carbon-containing atmosphere). The results also show that
the g
0
-MoC
12x
phase can be synthesized as a binary carbide
with very low oxygen content. It can therefore be concluded
that this phase is not an oxycarbide containing signicant
amounts of oxygen as suggested in the literature. The theo-
retical calculations also suggest that oxygen stabilizes the
g
0
-MoC
12x
phase. The total energy difference to other
carbide phases, however, is rather small, which can explain
why this compound can be obtained as a metastable phase. It
is clear that more extensive theoretical calculations also
including other carbides such as the different Mo
2
C phases
can give more detailed knowledge about phase stability in
the MoC system. Finally, lms in the metastable phases d-
MoC
12x
and g
0
-MoC
12x
decompose into different products
during annealing. The products are in agreement with two
different phase diagrams reported in the literature and
suggest that kinetic factors may be important in the forma-
tion of equilibrium conditions in the MoC system.
Acknowledgements
The authors are grateful to the Swedish Natural Science
Research Council (NFR), the Swedish Research Council for
Engineering Sciences (TFR) and the A

ngstrom Consortium
for nancial support.
References
[1] E. Rudy, St. Windisch, A.J. Stosick, J.R. Hoffman, Trans. Metall. Soc.
AIME 239 (1967) 1247.
[2] T.Y. Velikanova, V.Z. Kulii, B.V. Khaenko, Sov. Powder Metall. 27
(1988) 891.
[3] K. Kuo, G. Hagg, Nature 170 (1952) 245.
[4] A. Chrysanthou, P. Grieveson, J. Mater. Sci. Lett. 10 (1991) 145.
[5] J. Wood, J.E. Chen, A.M. Kadin, R.W. Burkhardt, S.R. Ovshinsky,
IEEE Trans. Magn. 21 (1985) 842.
[6] E.L. Haase, J. Low Temp. Phys. 69 (1987) 245.
J. Lu et al. / Thin Solid Films 370 (2000) 203212 211
[7] F. Okuyama, Y. Fujimoto, S. Kato, H. Kondo, Appl. Phys. A 38
(1985) 275.
[8] A. Watanabe, Y. Imai, K. Osato, et al. Denki Kagaku 57 (1989) 1067.
[9] I.F. Ferguson, J.B. Ainscough, Nature 202 (1964) 1328.
[10] K. Yasuda, J. Murota, Jpn. J. Appl. Phys. 22 (1983) 615.
[11] L. Norin, U. Jansson, J.-O. Carsson, Thin Solid Films 293 (1997)
133.
[12] J. Lu, U. Jansson, in: M. Allendorf, C. Bernard (Eds.), Chemical
Vapour Deposition, Paris, France, September 49, 1997, The Electro-
chemical Society and EUROCVD-11 Proceeding 9725, The Electro-
chemical Society, 1997.
[13] H. Skriver, Solid State Sciences, Springer, Berlin, 1984.
[14] H. Hugosson, O. Eriksson, L. Nordstrom, U. Jansson, L. Fast, A.
Delin, B. Johansson, J.M. Wills, J. Appl. Phys. 86 (1999) 3758.
[15] K. Yvon, W. Jeitschko, E. Parthe, J. Appl. Crystallogr. 10 (1977) 73.
[16] B.D. Cullity, Elements of X-ray Diffraction, 2nd ed., Addison-Wiley,
London, 1978, p. 102.
[17] Joint Committee on Powder Diffraction Standards, Powder Diffrac-
tion File, Card 060546, ASTM, Philadelphia, PA, 1967.
[18] Joint Committee on Powder Diffraction Standards, Powder Diffrac-
tion File, Card 350787, ASTM, Philadelphia, PA, 1984.
[19] Joint Committee on Powder Diffraction Standards, Powder Diffrac-
tion File, Card 451015, ASTM, Philadelphia, PA, 1988.
[20] H. Hogberg, P. Tagtstrom, J. Lu, U. Jansson, Thin Solid Films 272
(1996) 116.
[21] S. Veprek, S. Reiprich, Thin Solid Films 268 (1995) 64.
[22] A.A. Voevodin, S.V. Prasad, J.S. Zabinski, J. Appl. Phys. 82 (1997)
855.
J. Lu et al. / Thin Solid Films 370 (2000) 203212 212

Potrebbero piacerti anche