Iron thin films from Fe(CO)5 and FeCp2/H2O under atmospheric pressure

Article abstract of DOI:10.1149/1.2352050

Iron layers were first obtained from iron pentacarbonyl in metallorganic chemical vapor deposition (MOCVD) process under atmospheric pressure, in the temperature range 473- 773 K, in a vertical cold wall reactor. Films of good purity were obtained with or without hydrogen as co-reactant, no chemical additives being used. The experiments showed that the velocity of the gas stream and, to a lower extent, the precursor molar fraction are the key parameters to be controlled, in order to monitor film growth rate and purity. In a second step, Fe thin layers were obtained by atmospheric pressure MOCVD starting from the reactive gas mixture FeCp2 and H2 O in the temperature range 973- 1073 K. A thermochemical simulation of the Fe-C-H-O system allowed optimum processing conditions to be approached. X-ray diffraction and microprobe analysis showed that the highest iron content in the layer was obtained for H2 O FeCp2 ratios between 4 and 6. Film growth occurs in two steps: the initial formation of a black, powdered, and porous layer that becomes densified as a result of the grain growth on increasing the deposition time in order to form compact gray metal films. This two-step mechanism was confirmed by kinetic and in situ IR pyrometric observations.The Electrochemical Society

Full text of DOI:10.1149/1.2352050

Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
G1025
0013-4651/2006/153͑12͒/G1025/7/$20.00 © The Electrochemical Society
Iron Thin Films from Fe„CO…5 and FeCp2/H2O under
Atmospheric Pressure
z
*
*
F. Senocq, F.-D. Duminica, F. Maury, T. Delsol, and C. Vahlas
Centre Interuniversitaire de Recherche et d’Ingénierie des Matériaux CNRS/INPT, ENSIACET, 31077
Toulouse Cedex 4, France
Iron layers were first obtained from iron pentacarbonyl in metallorganic chemical vapor deposition ͑MOCVD͒ process under
atmospheric pressure, in the temperature range 473–773 K, in a vertical cold wall reactor. Films of good purity were obtained with
or without hydrogen as co-reactant, no chemical additives being used. The experiments showed that the velocity of the gas stream
and, to a lower extent, the precursor molar fraction are the key parameters to be controlled, in order to monitor film growth rate
and purity. In a second step, Fe thin layers were obtained by atmospheric pressure MOCVD starting from the reactive gas mixture
FeCp2 and H2O in the temperature range 973–1073 K. A thermochemical simulation of the Fe-C-H-O system allowed optimum
processing conditions to be approached. X-ray diffraction and microprobe analysis showed that the highest iron content in the
layer was obtained for H2O/FeCp2 ratios between 4 and 6. Film growth occurs in two steps: the initial formation of a black,
powdered, and porous layer that becomes densified as a result of the grain growth on increasing the deposition time in order to
form compact gray metal films. This two-step mechanism was confirmed by kinetic and in situ IR pyrometric observations.
© 2006 The Electrochemical Society. ͓DOI: 10.1149/1.2352050͔ All rights reserved.
Manuscript submitted April 7, 2006; revised manuscript received July 18, 2006. Available electronically October 10, 2006.
Our aim was to study chemical vapor deposition of iron thin
films for metallurgical applications. Indeed, continuous deposition,
or strip coating, can be highly relevant as a form of surface treat-
ment in metallurgy. A suitable process needs at least to fulfill three
requirements: ͑i͒ deposition at low temperature to avoid dimen-
sional and structural changes of the steel substrate, ͑ii͒ a high
growth rate to compensate for the speed of the strip, and ͑iii͒ op-
eration at atmospheric pressure, which is highly preferable for tech-
nical and economical reasons. The use of metallorganic compounds
as molecular precursors satisfies the first requirement. For the other
two, new processes and optimization of metallorganic chemical va-
por deposition ͑MOCVD͒ processes already reported in the litera-
ture are necessary.
Compared to other metals, few open scientific publications refer
to thermal CVD of iron, even though many are devoted to the prepa-
ration of iron oxides, sulfides, or silicides. Little attention seems to
have been paid to the optimization of the growth rate, which is
nevertheless necessary for a continuous deposition process. More-
over, a limited number of molecular precursors have been studied,
namely, Fe͑CP͒2,1 Fe͑CO͒5,2-7 Fe2Cp2͑CO͒4,8 Fe͓N͑SiMe3͒2͔3,9
and Fe͑COT͒͑CO͒3.10 Most of the processes reported operate under
low pressure, and rare are those working under atmospheric pres-
sure. Highly textured ͓011͔ ␣ iron films ͑bcc͒ are reported to have
been prepared from ferrocene1 under 105 Pa of H2, in the tempera-
ture range 673–1173 K, but with selective deposition, and no
growth rate is given. In the case of iron pentacarbonyl, high growth
rates are reported, up to 2 ␮m/h,5at moderate temperature and under
atmospheric pressure of hydrogen. However, rising temperature
lowers the growth rate,5,6 but the use of additives like 1,2,3,4,5-
pentamethyl-cyclopentadiene in the vapor phase softens this trend,6
In addition to the open literature, patents refer mainly to two
where L is either a phosphine, a phosphite, or an amine group, or the
family ͓Fe͑CO͒ ͔ ͑C4R4͒, where each R can be H, a halide, OH, an
3 2
alkyl, or an aryl moiety. Each of the molecules proposed can be
prepared using Fe͑CO͒ as a starting material. For example,
2
Fe͑CO͒ ͑NMe3͒, Fe͑CO͒ ͓P͑OMe͒ ͔, and Fe2͑CO͒ ͑C4H4͒ were
4
4
3
6
used, without carrier gas, under dynamic vacuum ͑0.13–1.33 Pa͒ to
prepare smooth and shiny films with a metallic appearance, but con-
taining high amounts of hetero-atoms ͑as high as 34 atom % C, or
23 atom % P͒. This is evidence that a clean decomposition of these
metallorganic complexes is not observed. Moreover, the toxicity of
Fe͑CO͒ and its derivatives limits their use, at least in large-scale
5
processes. To bypass this shortcoming, ferrocene has been used, i.e.,
under oxidative atmosphere, to obtain thin oxide layers, as patented
by Kane and Schweizer in 1975.16 In other conditions, Mukaida et
al. patented the use of FeCp2 for the preparation of ␤-FeSi2.17 In
both cases, the toxicity of Fe͑CO͒ was put forward to account for
5
not using this molecule. The alkyl-cyclopentadienyl derivatives
͓general formula: ͑R1C5H4͒2-pM͑R2͒p, where R1 is alkyl, R2 is H, F,
or alkyl, and M is a metal, which can be Fe͔ have been proposed18
to obtain oxide layers under oxidative atmosphere. The precursors
belonging to this family are claimed to have vapor pressures over
13 Pa at 373 K, but no specific example of iron oxide deposition is
given. Besides these two families, iron acetylacetonate has been
proposed for magnetic coating of iron oxide spinels,19,20 with or
without O2, but it was established that pure iron films were out of
reach with this precursor. More recently,21 Choi et al. proposed an-
other kind of organometallic compounds ͓of the general formula Rm1
M͑PR23͒x, with R1 = H, D, H2, D2, alkyl; R2 = alkyl, aryl; and
M = VIIB and VIIIB group metal͔ to obtain high-purity metallic
films or powders.
A promising class of organometallic precursors for deposition of
transition metals is the metallocenes; among them, ferrocene
͑FeCp2͒ was previously used.22,23 FeCp2 is easy to synthesize and
purify, stable in air, and nontoxic. It has a relatively low vapor
pressure ͑ca. 1.3 Pa at 303 K͒, and the good thermal stability of this
compound allows heating at sufficiently high temperatures to sig-
nificantly increase its vapor pressure.24 Such thermal stability im-
poses relatively high pyrolysis temperatures ͑up to 773 K͒ at which
an autocatalytic decomposition reaction occurs in the gas phase and
leads to the formation of a black powder, which primarily consists of
iron contaminated by graphitic carbon and traces of Fe3C ͑cement-
ite͒. According to the literature, reasonably pure Fe films have been
deposited by low-pressure CVD to limit gas-phase nucleation, but it
is very difficult to obtain such films under atmospheric pressure
using inert carrier gas. Elihn et al. employed oxygen to decrease the
carbon contamination under atmospheric pressure with a gas mix-
families of precursors: ͑i͒ Fe͑CO͒ and its derivatives and ͑ii͒
5
Fe͑CP͒ and related compounds. As Fe͑CO͒ vapor pressure is too
2
2
high to easily control the low mole fractions required for micro-
electronic applications, Long11 chose Fe͑CO͒ ͑C4H6͒ and
3
Fe͑CO͒ ͑C8H8͒, for Fe doping of InP layers, between 923 and
3
973 K. He found that the butadiene derivative gave more promising
results than Fe͑CP͒2, and attributed this to its low decomposition
temperature ͓Յ523 K,12 to be compared to Ն723 K 13 for Fe͑Cp͒ ͔.
2
Owing
to
the
high
vapor
pressure
of
Fe͑CO͒5,
McCormick14,15suggested the use of the derivatives Fe͑CO͒4L,
*
Electrochemical Society Active Member.
z E-mail: Francois.Senocq@ensiacet.fr
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G1026
Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
Table I. Typical MOCVD conditions for the growth of Fe layers under atmospheric pressure starting from different H2O/FeCp2 molar ratios.
No. run
1
2
3
Growth temperature ͑K͒
Carrier gas flow rate ͑N2, sccm͒
Total gas flow rate ͑sccm͒
H2O/FeCp2 molar ratio
973
4950
5500
5.5
1023
4900–5100
5500
1073
4950
5500
5.5
1.9–6.3
Co-reagent H2O ͑sccm͒
FeCp2 sublimation temperature ͑K͒
FeCp2 mole fraction
82
373
31–82
82
373
368–375
͑2.3–3.5͒ ϫ 10−3
10–24
2.7 ϫ 10−3
15
2.7 ϫ 10−3
15
Deposition time ͑min͒
ture containing 99–99.5% argon and 0.5–1% of oxygen.25 They ob-
served a decrease of carbon in the nanoparticles thus formed, but
they also raised the question of the presence of Fe2O3, in addition to
iron.25,26
electron microscopy ͑SEM͒ was used to investigate surface mor-
phology, and film thicknesses were determined through observations
of cross sections and from the mass gain of the samples. The film
composition was analyzed by electron probe microanalysis ͑EPMA͒.
The selection of the molecular precursor is a key point for the
development of MOCVD processes. The choice strongly depends on
the application and the constraints of the deposition process.27 For
metallurgical coating applications, continuous deposition of pure Fe
films is advantageous as they could be used, for instance, in strip
coating. In this objective, high growth rates and atmospheric pres-
sures are significant requirements for the CVD process. In this pa-
per, we report an investigation of MOCVD of Fe films under atmo-
Results and Discussion
Fe͑CO͒
5.— Using Fe͑CO͒ without additive, incorporation of
5
carbon or oxygen from the carbonyl ligands in the layer could be
expected, so the first experiments were carried out under hydrogen.
With a carrier gas flow rate of 30 sccm of N2 and 200 sccm of H2,
and with precursor molar fractions of 2 ϫ 10−3 or 6.75 ϫ 10−3, no
deposition occurs for a substrate temperature of 443 K, this pyroly-
sis temperature likely being too low in the gas flow conditions cho-
sen. At 473 K, thick films were formed at growth rates as high as
50 ␮m/h, but with intense spallation, even during growth. The “self-
destruction” of the film was attributed to too high a growth rate,
which causes too much strain in the film and poor adhesion to the
substrate. From 523 to 673 K, adhesive shiny gray metallic films
were obtained.
spheric pressure using Fe͑CO͒ as an iron source under various
5
conditions, and from the reactive gas mixture FeCp2 and H2O in the
temperature range 973-1073 K. Preliminary results were recently
28
reported at an ECS meeting, using Fe͑CO͒
and FeCp2/H2O.29
5
Here, the experimental details are reported and the results discussed
in relation with data from the literature. Thermodynamic calcula-
tions were carried out to predict and optimize the growth conditions
from this gas mixture and the growth mechanism is discussed.
XRD analysis ͑Fig. 1͒ shows that these films are made of iron,
with Fe3C ͑cementite͒ as secondary phase ͑no crystallized graphitic
carbon was detected͒. A slight texturation along the ͑110͒ axis of
iron was observed ͑texturation coefficient Ϸ2͒. Application of the
Scherrer law gave a mean crystallite size of ca. 10 nm in the tem-
perature range 473–573 K, rising to ca. 25 nm at 623 K. As EPMA
analysis ͑see below͒ showed that the carbon content was steady ͑1–
2 atom %͒ in the temperature range 523–623 K, the difference in
the apparent Fe/Fe3C ratio in X-ray diagrams can be attributed to a
better crystallization of iron at higher temperatures.
The SEM observations show smooth and homogeneous surface
morphologies, while analysis of cross sections reveals an evolution
of the morphology with the temperature: porous with columnar
growth until 553 K, and lamellar beyond this temperature ͑Fig. 2͒.
The composition of the films was investigated with EPMA analy-
ses: at 473 K, iron was 90 atom % as carbon was up to 10%. The
large quantities of carbon may come from the deposition tempera-
ture being too low, which could lead to incomplete decomposition of
the precursor, and its subsequent incorporation into the film, owing
to the high growth rate. In the temperature range 523–623 K, the
iron rose to ca. 98 atom %, and carbon to ca. 1–2%. The oxygen
content was very low, below the detection limits. At 673 K, the iron
Experimental
The CVD experiments were conducted in a vertical tubular cold-
wall reactor, 46 mm i.d., under atmospheric pressure, the substrates
͓Si͑111͒ wafers or stainless steel coupons͔ being placed on a stain-
less steel sample holder ͑32 mm in diameter, 40 mm high͒ heated by
HF induction. Nitrogen was used as carrier gas, and either hydrogen
or nitrogen were used as dilution gas. The gas flows were monitored
with mass-flow meters.
Fe͑CO͒ and FeCp2 mole fractions were controlled by monitor-
2
ing the temperature of the thermostated saturators and were deter-
mined from precursor mass loss. The same procedure was followed
for the water vapor mole fraction in experiments with FeCp2. In
both cases, N2 was the carrier gas of the precursors, and, with
FeCp2, the total flow rate was 5500 standard cubic centimeters per
minute ͑sccm͒ to reach laminar flow above the surface of the sub-
strates. Typical MOCVD conditions for the growth of Fe layers
under atmospheric pressure starting from different H2O/FeCp2 mo-
lar ratios are presented in Table I, and those starting from Fe͑CO͒
5
in Table II.
The microstructure of the films obtained was studied by X-ray
diffraction ͑Cu K␣, diffracted beam monochromator͒. Scanning
Table II. Growth rates and composition (EPMA) of films deposited under atmospheric pressure starting from Fe„CO…5 for various experimental
conditions.
EPMA
Run
K
Flow rate ͑sccm͒
Mole fraction
V gas ͑cm/s͒
Growth rate ͑␮m/h͒
Fe
C
E6
D2
D12
D26
D14
D27
673
673
673
673
773
773
200H2ϩ30N2
1147 ͑N2͒
6.7 ϫ 10−3
1.0 ϫ 10−2
2.2 ϫ 10−3
7.0 ϫ 10−3
1.0 ϫ 10−3
7.0 ϫ 10−3
0.25
1.16
2.33
10.06
5.02
10.06
0.3
1.6
2.0
4.5
0.16
0.6
90
89
95
92
94
92
6
11
5
8
6
2320 ͑N2͒
10,000 ͑N2͒
5,000 ͑N2͒
10,000 ͑N2͒
8
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
G1027
Figure 3. Growth rates vs deposition temperature, for different molar frac-
tions, in the temperature range 473–673 K, 230 sccm total flow rate
͑30 sccm N2 + 200 sccm H2͒. The dotted curves are a guide for the eyes.
Figure 1. XRD diagrams ͑grazing incidence 2°͒ of iron films deposited from
Fe͑CO͒ and H2 on silicon wafer.
5
content dropped to 90%, as C and O rose to ca. 5% each. The lower
film purity can be correlated to the homogeneous decomposition of
͑Fig. 4͒ show the diffraction lines of iron. The other lines were
attributed to iron oxide Fe2O3, no graphitic carbon or iron carbide
being detected. Similar diagrams were obtained from samples pre-
pared at 773 K with a higher flow rate, although a slight increase of
iron oxide amount was observed. The SEM observations of the sur-
faces and of the cross sections of samples grown at 673 and 773 K
showed smooth and homogeneous layers, with regular thickness,
and equi-axial grain growth.
The growth rates increased with the total flow rate ͑Table II͒, and
then with the gas velocity. For similar mole fractions, the growth
rate was 0.3 ␮m/h at 673 K with 230 sccm flow rate ͑run E6͒, and
4.5 ␮m/h with 10,000 sccm ͑run D26͒. Nevertheless, as observed
before, the growth rate dropped when the temperature was increased
from 673 K ͑4.5 ␮m/h, run D26͒ to 773 K ͑0.6 ␮m/h, run D27͒,
keeping constant the flow rate and mole fraction. As homogeneous
decomposition is nearly completely avoided, the reverse reaction
that gives volatile iron species from iron film and releases CO
Fe͑CO͒ observed in the gas phase from ca. 623 K.
5
The growth rate was estimated by SEM observation of cross
sections of samples, or on scales for the thickest deposits where they
had become detached from the substrate ͑Fig. 2b͒. A decrease in
growth rate with increasing deposition temperature was observed
͑Fig. 3͒, the upper limit being apparently 673 K, for which the depo-
sition rate was very low ͑Յ0.3 ␮m/h for a molar fraction of 2
ϫ 10−3 or 6.75 ϫ 10−3͒. This behavior has already been reported for
Fe͑CO͒ 5,6 and for Ni͑CO͒4,5,30 and is imputed to ͑i͒ homogeneous
5
gas-phase reactions of the reactant and ͑ii͒ surface reactions of CO
released by deposition and decomposition reactions to re-form vola-
tile iron species.
An increase of growth rate with molar fraction was observed, but
it did not allow deposition beyond 673 K. Experiments done at
higher temperatures, and with low flow rates, show intense homo-
geneous decomposition of the precursor, leading to black soot in the
reactor and thin, black, nonadhesive crumpled films. As the thermal
limitation of the deposition reaction could be attributed to ͑i͒ homo-
geneous decomposition reactions and ͑ii͒ surface side reactions due
to poor byproduct elimination, and also to far too long residence
times of different chemical species in the reaction zone, it seemed
promising to shorten the residence time by increasing the total flow
rate. This was done by raising the overall flow rate ͑up to
10,000 sccm͒. As it is possible to get pure metal films from metal
seemed to dominate. Moreover, the higher the Fe͑CO͒ mole frac-
5
tion, the higher the incorporation of carbon in the film, which pre-
vents the growth of pure metal layers at high mole fractions.
FeCp
2.— When FeCp2 was used without any additive, a self-
catalyzed decomposition reaction occurred in the gas phase over
813 K leading to black soot and a black nonadhesive deposit mainly
composed of pulverulent graphitic carbon, untextured iron, and
some cementite Fe3C. Shortening the residence time by increasing
carbonyls such as Ni͑CO͒ without using a reductive atmosphere30
the total flow rate as previously done for Fe͑CO͒ did not improve
4
5
and in order to avoid the dangers of large hydrogen output, pure
nitrogen was used both as dilution and carrier gas.
the film quality. Below 813 K, the growth was very slow, due to a
low pyrolysis rate.
The first noticeable effect of the increase of flow rate was to drop
To allow the use of FeCp2, a well-chosen additive is therefore
necessary. As H2 is not recommended in large-scale applications for
security reasons, the effect of water vapor, which is cheap, nontoxic,
and risk-free, was investigated. Along with the experiments, a ther-
modynamic simulation of the FeCp2/H2O system was performed.
the homogeneous decomposition of Fe͑CO͒ previously observed
5
͑i.e., run E6͒. It was thus possible to prepare iron films in such
conditions at 673 K, with an overall gas flow rate of 1147 sccm. As
for the films prepared at low temperature under H2, XRD diagrams
Figure 2. SEM cross section of iron films
on silicon wafer obtained at 553 K ͑a, left͒
and 623 K ͑b, right͒ from Fe͑CO͒ and
5
H2. ͑b͒ shows a film flake detached from
substrate. The white bars below the pic-
tures represent 1 ␮m ͑left͒ and 2 ␮m
͑right͒.
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G1028
Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
Thermodynamic study.— For the thermodynamic calculations, a
list was made of all the possible gaseous and condensed phases that
can be formed from the combinations of the elements involved in
the system, namely, C, Fe, H, and O. The total Gibbs energy of this
system was minimized to obtain the number of molecules from each
compound that form at various temperatures and H2O/FeCp2 molar
ratios. The total Gibbs energy was minimized using the Gemini 2
software.31 A set of 172 gaseous species and 14 condensed stoichio-
metric compounds were considered in the calculations. Their ther-
modynamic description was obtained from the SGTE ͑Scientific
Group Thermodata Europe, Grenoble͒ data base. In composing the
data base, the C-Fe metastable diagram was used, as established by
Gustafson.32 The use of this diagram allowed us to predict the
amount of cementite that would be obtained in the film.
Figure 5 presents a deposition diagram calculated at 105 Pa as a
function of deposition temperature and of the H2O/FeCp2 molar
ratio in the input gas phase ͑95% Ar͒.
Figure 4. XRD diagram of iron thin film from Fe͑CO͒5, at 673 K, overall
flow rate 2320 sccm of N2, Fe͑CO͒ mole fraction: 2.2 ϫ 10−3 ͑run D12͒.
5
It reveals the possible existence of different domains composed
of one, two, or three condensed phases. H2O/FeCp2 ratios lower
than 10 yielded films with no contamination from carbon, but con-
taining cementite above 773 K. H2O/FeCp2 ratios higher than ap-
proximately 14.5 yielded ferrous oxide, FeO. Between these two
values, pure Fe is expected to be deposited. In these conditions, the
cubic phase Fe bcc is expected in the films above 850 K; it is re-
placed by Fe fcc above 1160 K. Figure 6 presents two yield dia-
grams of the calculated number of moles of gaseous species as a
function of the H2O/FeCp2 ratio in the input gas at 1000 K ͑left͒
and as a function of deposition temperature for an initial H2O/FeCp2
ratio of 12.5. These conditions correspond to the dotted lines shown
in Fig. 6. Dashed domains in the two diagrams of Fig. 6 correspond
roughly to conditions where pure Fe bcc is expected to be deposited.
In these conditions, the reaction of FeCp2 with water leads to a
reductive atmosphere mainly composed of carbon monoxide and
hydrogen. This overall reaction can be written as
4Fe͑C5H5͒2 + 45H2O → 4Fe ͑BCC͒ + 35CO + 65H2 + 5CO2
As is often the case with thermodynamic calculations, these re-
sults should be considered as trends rather than precise predictions.
However, they do indicate a window to obtain pure Fe with this
system. The next sections confirm the useful trends of this calcula-
tion.
Influence of the growth conditions.— CVD runs were carried
out at 973 K starting from a gas mixture of FeCp2 and H2O, but as
the growth rate was relatively slow, the thickness was extremely
weak after 20 min of reaction. Increasing the temperature to 1023 K
gave layers with reasonable thicknesses. In situ observations re-
Figure 5. Theoretical phase diagram for the Fe-C-H-O system under 1 atm
describing the film composition for various temperatures and H2O/FeCp2
molar ratios. Dotted lines correspond to the calculation conditions for the
yield diagrams of Fig. 6.
Figure 6. Calculated volumic fractions of gaseous species in the Fe-C-H-O system ͑Ar atmosphere͒. Species occurring at very low concentrations such as C2H4
and C2H2 are not plotted for clarity.
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
G1029
Figure 8. Atomic composition ͑EPMA͒ of iron layers deposited on Si͑111͒
during CVD runs starting from different H2O/FeCp2 molar ratios ͑deposition
time = 15 min͒.
Figure 7. XRD grazing patterns ͑incidence angle 2°͒ of Fe layers grown by
CVD on Si͑111͒ at 1023 K for different H2O/FeCp2 molar ratios ͑deposition
time = 15 min0: ͑a͒ H2O/FeCp2 = 0; ͑b͒ H2O/FeCp2 = 6; ͑c͒ H2O/FeCp2
= 8.
For example, on Si, after 10 min, a significant quantity of graphitic
carbon is incorporated into the layers and it decreases significantly
by increasing the deposition time.
vealed that during the first 10 min, a black powder layer was formed
on Si substrates. After 10 min, a gray layer, apparently more com-
pact and adherent, grew quickly over the whole sample, starting
from the edges. This compact gray layer did not grow on the top of
the black layer, but rather seemed to arise from the transformation of
the initial powder layer. The samples became entirely gray after
15 min. On steel substrates, the formation of the gray layer started
later ͑after 14–15 min͒ and the sample became entirely gray after
18–19 min. At 1073 K, an increase of the growth rate on Si͑100͒
was observed. The formation of the gray layer started after only
5 min, compared to 12 min at 1023 K.This indicates that the in-
crease in the temperature appreciably accelerated the formation of
the compact gray layer.
Figure 7 shows the XRD diagrams of layers carried out at
1023 K starting from FeCp2 and H2O on Si͑111͒ substrates. It con-
firms the tendencies anticipated by thermodynamic simulation, i.e.,
by increasing the water partial pressure, the amount of graphitic
carbon in the film is lowered to negligible values, while levels of
both iron and oxides rise. The layers are composed of C + Fe3C
+ Fe for H2O/FeCp2 molar ratio = 0, Fe3C + Fe + oxide traces for
H2O/FeCp2 = 6, and Fe + Fe3O4 for H2O/FeCp2 = 8.
Assuming there is no preferential orientation, the relative inten-
sity of the main diffraction lines suggests that the increase of the
deposition time involves a decrease of the graphitic carbon content,
which becomes practically undetectable after 20 min ͑Fig. 9͒. The
formation of cementite was optimal after 15 min ͑during the forma-
tion of the gray layer͒, and then decreased with increasing deposi-
tion time. The amount of Fe3O4 remained almost constant, while the
incorporation of FeO decreased after 12 min. On steel substrates, we
observed the same tendencies as on Si.
The morphology and the microstructure of the layers depended
significantly on the deposition time. Figure 10 shows SEM micro-
graphs of Fe layers obtained starting from FeCp2 and H2O at
1023 K on Si͑100͒ for different deposition times. A very porous
layer composed of nanometric grains was observed after 10 min of
deposition. It was black, powdery, and adhered poorly to the sub-
strate. Increasing the deposition time led to an increase in grain size.
After 12 min the grains started to grow, leading to the formation of
Electron-probe microanalysis ͑EPMA͒ confirmed the variation of
Fe, C, and O composition with the H2O/FeCp2 molar ratios ͑Fig. 8͒.
Thus, for a H2O/FeCp2 ratio = 0, the layers had a powdery appear-
ance, being primarily made up of carbon and carbide. The increase
of the H2O/FeCp2 ratio led to a fast reduction in the amount of
carbon, a slight increase in oxygen, and a significant increase in
iron. The highest Fe content was obtained for H2O/FeCp2 molar
ratios in the range 4–6, which is why a series of experiments was
carried out around these values. For H2O/FeCp2 molar ratios higher
than 6, an increase in oxygen to the detriment of Fe was observed to
the benefit of FeO and Fe3O4.
Analysis of the growth mode.— The XRD patterns show the
presence of Fe ͑majority͒, Fe3C, graphitic C, and oxides ͑FeO and
Fe3O4͒ whose ratio changes with the deposition time. The formation
of oxides is observed on Si͑100͒ for H2O/FeCp2 molar ratios higher
than 4. On steel, the formation of oxides starts for H2O/FeCp2 ratios
higher than 6, maybe because a small part of the oxygen is con-
sumed by oxidation of steel substrates. The deposition time seems to
modify the nature and the structure of the layers ͑on Si and steel͒.
Figure 9. XRD relative intensities of C͑002͒/Fe͑200͒, Fe3C͑210͒/Fe͑200͒,
FeO͑200͒/Fe͑200͒, and Fe3O4͑311͒/Fe͑200͒ ratios of Fe layers grown on
Si͑100͒ at 1023 K for different deposition times, starting from H2O/FeCp2
molar ratio = 4.
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

G1030
Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
Figure 12. Variation of the pyrometric signal vs the deposition time of an Fe
layer grown at 1023 K on steel substrate starting from H2O/FeCp2 molar
ratio = 6 ͑the continuous line represents the smoothed pyrometric signal and
the oscillations are due to the temperature control of the substrate͒.
Figure 10. SEM micrographs of Fe layers obtained starting from FeCp2 and
H2O at 1023 K on Si͑100͒ for different deposition times ͑H2O/FeCp2 molar
ratio = 4͒.
the gray layer mentioned previously but exhibiting better adhesion
to the substrate. After 20 min the layer becomes very compact and
uniform and seems to be composed of equiaxial micrometric grains.
The morphological diversity of the Fe layers formed after the
different durations of deposition made it difficult to determine the
growth rate. Kinetic data were obtained by measuring the mass of
deposited film on steel substrates for different deposition times ͑Fig.
11͒. For roughly 10 min, the mass of deposited film remained very
low. This can be explained by the high porosity of the layer and by
a significant deposition of graphitic C ͑light element͒ in the early
stages of the growth during the formation of the powdery layer ͑Fig.
11 and 12͒. A steady state is observed after ca. 10 min with a con-
stant growth rate of 0.156 mg min−1 cm−2 ͑ca. 12 ␮m/h͒, corre-
sponding to the formation of the compact gray layer.
This growth mode was confirmed by in situ IR pyrometry analy-
sis using a setup described elsewhere.33 Details on this diagnostic
tool for the control of CVD processes have recently been reported.34
The pyrometric signal recorded during the CVD of Fe at 1023 K on
steel starting from a gaseous FeCp2-H2O mixture is shown in Fig.
12.
responding to the nucleation step leading to the formation of the
powdery black layer. During the second period, the deposition of C
and oxides ͑FeO and Fe3O4͒, which both have higher emissivities
than the substrate, explains the higher values of the pyrometric sig-
nal. As a result, the period between 5 and 14 min corresponds to the
growth of the loose black layer. The third stage of growth appears
after 14 min with an abrupt reduction of the pyrometric signal ͑oc-
curring over only 1 min͒, before its stabilization to an approximately
constant value. In the intermediate period, a decrease of C and Fe
oxides in the layer was observed ͑Fig. 9͒. During this stage, the iron
grains rapidly grow to uniformly cover the surface of the substrate
forming the compact gray layer. Except for the nucleation step, the
transition between the two growth modes is only of a few tens of
seconds. Generally, the compact gray layer starts to be formed at one
edge of the sample and the growth rapidly extends across the sample
like a step-flow mechanism. Before the growth of the compact gray
layer, the carbon is entirely consumed and the level of FeO de-
creases significantly, as shown in Fig. 9. These observations indicate
reduction of the iron oxides by the C during the deposition of the
powdery black layer. The growth of Fe layers at 1023 K is in good
agreement with Ellingham’s diagram. Once the codeposited C is
consumed by the reduction of oxides, a metallic Fe film grows ͑bcc
structure of the metal͒ with only slight contamination by Fe3C.
Three main growth periods can be clearly distinguished. In the
first period ͑0–4 min͒, the pyrometric signal strongly increases,
likely due to a change of emissivity and of surface roughness cor-
Conclusion
Relatively pure thin layers of iron were obtained, with high
growth rates, by atmospheric pressure MOCVD, either from
Fe͑CO͒5 in the temperature range 473–773 K, or FeCp2 and H2O in
the temperature range 973–1023 K.
Fe͑CO͒ is a satisfactory precursor for iron thin-film deposition
5
under atmospheric pressure, even without additives in the gas phase.
With this precursor, the growth rate is closely related to gas velocity
as a short residence time in the deposition zone lowers the formation
of byproducts, which penalize the layer growth. Nevertheless, with-
out additive in the gas phase, an increase of the precursor mole
fraction favors the incorporation of carbon into the film. As a result,
the window for the growth parameters is relatively narrow so it will
be certainly hard to optimize the use of Fe͑CO͒ in a large-scale
5
reactor process, due to its sensitivity to the variation of the local
conditions.
The use of FeCp2 alone leads to a pulverulent and nonadhesive
deposit, but the addition of water vapor to the gas phase leads to the
formation of relatively pure Fe thin layers. The layers are essentially
composed of Fe, but also contain secondary phases like Fe3C, FeO,
Figure 11. Variation of the film masses deposited on steel substrates for
different deposition times at 1023 K ͑H2O/FeCp2 ratio = 6͒.
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

Journal of The Electrochemical Society, 153 ͑12͒ G1025-G1031 ͑2006͒
G1031
11. J. A. Long, V. G. Riggs, A. T. Macrander, and J. W. D. Johnston, J. Cryst. Growth,
77, 42 ͑1986͒.
12. D. L. S. Brown, J. A. Connor, M. L. Leung, M. I. Paz-Andrade, and H. A. Skinner,
Fe3O4, and C, the proportion of which depends on the CVD condi-
tions. The morphology and the structure of the layers change with
the deposition time. In addition to nucleation, growth occurs in two
steps: ͑i͒ the initial formation of a black porous layer that ͑ii͒ den-
sifies in a few minutes as a result of grain growth and coalescences
for increasing deposition times leading to compact gray metallic
films. This two-step mechanism was confirmed by kinetics studies
and by in situ IR pyrometric diagnostic. XRD and EPMA analyses
show that the highest purity of the Fe layers is obtained for
H2O/FeCp2 ratios between 4 and 6. The experimental results are in
fairly good agreement with the thermodynamic model made of the
Fe-C-H-O system under atmospheric pressure.
J. Organomet. Chem., 110, 79 ͑1976͒.
13. I. B. Johns, E. A. McElhil, and J. O. Smith, J. Chem. Eng. Data, 7, 277 ͑1962͒.
14. F. B. McCormick, W. L. Gladfelter, and Y. Senzaki, U.S. Pat. 5,314,727 ͑1994͒.
15. F. B. McCormick, W. L. Gladfelter, and Y. Senzaki, U.S. Pat. 5,372,849 ͑1994͒
16. J. Kane and H. Schweizer, U.S. Pat. 3,914,515 ͑1975͒.
17. M. Mukaida, I. Hiyama, T. Tsunoda, and Y. Imai, Thin Solid Films, 381, 214
͑2001͒.
18. A. Erbil, U.S. Pat. 4,927,670 ͑1990͒.
19. H. Torii, E. Fujii, M. Aoki, and K. Ochiai, U.S. Pat. 4,975,324 ͑1990͒.
20. E. Fujii, H. Torii, A. Tomozawa, R. Takayama, and T. Hirao, J. Cryst. Growth,
151, 134 ͑1995͒.
21. H. Choi, U.S. Pat. 6,777,565 ͑2004͒.
22. G. T. Stauf, D. C. Dristcoll, P. A. Dowben, S. Barfuss, and M. Grade, Thin Solid
Films, 153, 421 ͑1987͒.
National Center for Scientific Research assisted in meeting the publica-
tion costs of this article.
23. G. J. M. Dormans, J. Cryst. Growth, 108, 806 ͑1991͒.
24. F.-D. Duminica, F. Maury, and F. Senocq, Récents Progrès en Génie des Procédés,
Vol. 92, L2/1-8, Lavoisier, Paris ͑2005͒.
25. K. Elihn, F. Otten, M. Boman, F. E. Kruis, H. Fissan, and J.-O. Carlsson, Nano-
struct. Mater., 12, 79 ͑1999͒.
References
1. G. J. M. Dormans, J. Cryst. Growth, 180, 806 ͑1991͒.
2. G. T. Stauf, D. C. Dristcoll, P. A. Dowben, S. Barfuss, and M. Grade, Thin Solid
Films, 153, 421 ͑1987͒.
26. F. Otten, K. Elihn, F. E. Kruis, M. Boman, J.-O. Carlsson, and H. Fissan, J. Aerosol
Sci., 29, 125 ͑1998͒.
3. R. Kaplan and N. Bottka, Appl. Phys. Lett., 41, 972 ͑1982͒.
4. P. J. Walsh and N. Bottka, J. Electrochem. Soc., 131, 444 ͑1984͒.
5. P. A. Lane, P. J. Wright, P. E. Oliver, C. L. Reeves, A. D. Pitt, J. M. Keen, M. C.
L. Ward, M. E. G. Tisley, N. A. Smith, B. Cockayne, and I. Rex Harris, Chem. Vap.
Deposition, 3, 97 ͑1997͒.
6. P. A. Lane and P. J. Wright, J. Cryst. Growth, 204, 298 ͑1999͒.
7. H. J. Haugan, B. D. McCombe, and P. G. Mattocks, J. Magn. Magn. Mater., 247,
296 ͑2002͒.
27. F. Maury, Chem. Vap. Deposition, 2, 113 ͑1996͒.
28. T. Delsol, F. Maury, and F. Senocq, in Proceedings of the 15th European Confer-
ence on Chemical Vapor Deposition, R. Fischer, Editor, PV 2005–09, p. 638, The
Electrochemical Society Proceedings, Pennington, NJ ͑2005͒.
29. F. D. Duminica, F. Maury, C. Vahlas, F. Senocq, and T. Delsol, in Proceedings of
the 15th European Conference on Chemical Vapor Deposition, R. Fischer, Editor,
PV 2005-09, pp. 644, The Electrochemical Society Proceedings, Pennington, NJ
͑2005͒.
8. R. Feurer, M. Larhafi, R. Morancho, and R. Calsou, Thin Solid Films, 167, 195
͑1988͒.
30. F. Fau-Canillac and F. Maury, Surf. Coat. Technol., 64, 21 ͑1994͒.
31. http://thermodata.online.fr/thermafr/france.html
9. D. V. Baxter, M. H. Chisholm, G. J. Gama, A. L. Hector, and I. P. Parking, Chem.
Vap. Deposition, 1, 49 ͑1995͒.
10. W. Luithardt and C. Benndorf, Thin Solid Films, 290/291, 200 ͑1996͒.
32. P. Gustafson, Scand. J. Metall., 14, 259 ͑1985͒.
33. C. Gasqueres, F. Maury, and F. Ossola, Chem. Vap. Deposition, 9, 34 ͑2003͒.
34. C. Gasqueres, F. D. Duminica, and F. Maury, Chem. Eng. Proc., Submitted.
Downloaded on 2015-04-19 to IP 170.140.26.180 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).