Effects of processing parameters in the chemical vapor deposition of cobalt from cobalt tricarbonyl nitrosyl

Article abstract of DOI:10.1149/1.1391904

This paper reports the development of a thermal chemical vapor deposition process for pure cobalt from the source precursor cobalt tricarbonyl nitrosyl for incorporation in integrated circuit silicide applications. Studies were carried out to examine the underlying mechanisms that control Co nucleation and growth kinetics, including the effects of key process parameters on film purity, texture, morphology, and electrical properties. For this purpose, systematic variations were implemented for substrate temperature, precursor flow, hydrogen reactant flow, and deposition time (thickness). Resulting films were analyzed by Rutherford backscattering spectrometry, X-ray photoelectron spectroscopy, X-ray diffraction, four-point resistivity probe, scanning electron microscopy, and atomic force microscopy. These investigations identified an optimized process window for the growth of pure Co with resistivity of 9 ± 2 μΩ cm, smooth surface morphology, and root-mean-square surface roughness at or below 10% of film thickness.

Full text of DOI:10.1149/1.1391904

Journal of The Electrochemical Society, 146 (6) 2139-2145 (1999)
2139
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
The Effects of Processing Parameters in the Chemical Vapor Deposition
of Cobalt from Cobalt Tricarbonyl Nitrosyl
a
a
a
a
z
Ana R. Ivanova, Guillermo Nuesca, Xiaomeng Chen, Cindy Goldberg, Alain E. Kaloyeros,
b
c,
Barry Arkles, and John J. Sullivan *
aNew York State Center for Advanced Thin Film Technology and Department of Physics, The State University of New York at
Albany, Albany, New York 12222, USA
b
Gelest Incorporated, Tullytown, Pennsylvania 19007, USA
c
MKS Instruments, Andover, Massachusetts 01810, USA
This paper reports the development of a thermal chemical vapor deposition process for pure cobalt from the source precursor cobalt
tricarbonyl nitrosyl for incorporation in integrated circuit silicide applications. Studies were carried out to examine the underlying
mechanisms that control Co nucleation and growth kinetics, including the effects of key process parameters on film purity, texture,
morphology, and electrical properties. For this purpose, systematic variations were implemented for substrate temperature, precur-
sor flow, hydrogen reactant flow, and deposition time (thickness). Resulting films were analyzed by Rutherford backscattering spec-
trometry, X-ray photoelectron spectroscopy, X-ray diffraction, four-point resistivity probe, scanning electron microscopy, and atom-
ic force microscopy. These investigations identified an optimized process window for the growth of pure Co with resistivity of 9 Ϯ
2
␮⍀ cm, smooth surface morphology, and root-mean-square surface roughness at or below 10% of film thickness.
©
1999 The Electrochemical Society. S0013-4651(98)08-015-X. All rights reserved.
Manuscript submitted August 5, 1998; revised manuscript received December 22, 1998.
in situ doping with Co.19-21 There are also some recent reports on
Continued enhancement in integrated circuitry (IC) speed and
performance through scaling down of the minimum feature size be-
low 0.25 ␮m requires the implementation of innovative material and
2
2,23
CVD CoSi formation by codeposition.
In these applications,
2
metallorganic compounds are typically employed, since cobalt
halides are not suitable due to low volatility and high decomposition
1
process solutions. A case in point is cobalt silicide (CoSi ), which
2
2
4,25
is emerging as a viable silicide replacement for titanium silicide
temperatures.
2
(
TiSi ) in future self-aligned silicide (salicide) technology. The key
Within the scarce literature on cobalt CVD, considerable atten-
2
advantages of CoSi , in comparison with TiSi , are wider silicidation
tion has been given to the deposition of Co by thermal decomposi-
2
2
2
19,22,23,26
window, and easier scalability to finer line geometries. Additional-
tion of cobalt carbonyl Co (CO) .
The most serious draw-
2
8
ly, epitaxial CoSi exhibiting superior material integrity and thermal
back associated with its use is the existence of undesirable, but ther-
modynamically favorable, precursor reaction pathways which com-
pete with the deposition of pure Co. These pathways include poly-
merization reaction in the gas phase and reaction of the precursor
2
stability can be formed on Si(100) in a fashion which is highly com-
3
-5
patible with current microelectronics technology.
In a typical salicide process, the blanket metal is deposited first,
followed by annealing to form a silicide phase on the device regions
with H to form a highly volatile and extremely unstable hydro-
2
6
27
8
with exposed Si, and the unreacted metal is subsequently etched.
cobalt tetracarbonyl HCo(CO)4. Furthermore, Co (CO) decom-
poses during storage even under vacuum or inert atmosphere.
2
4
,7,8
9-11
21
Sputtering
or evaporation
techniques are commonly em-
ployed for Co deposition. Sputtering is typically preferred over
Other candidate precursors for Co CVD included cobaltocene
Co(C H ) , cyclopentadienylcobalt dicarbonyl C H Co(CO) , and
3
evaporation due to the higher purity of sputter-deposited films and
5
5 2
5
5
2
6
22,28,29
superior control of the sputter-deposition process. However, the
cobalt acetylacetonate Co(C H O ) .
All three sources
5
7 2 3
highly energetic ions in the sputtering process can potentially induce
substrate damage. This is particularly undesirable in salicide pro-
cessing, since the metal is deposited directly over the active source
and drain device regions which are formed in single-crystal Si.
Alternatively, chemical vapor deposition (CVD) can be em-
ployed to grow pure metal films without ion-induced substrate dam-
require H reduction and have been typically employed in atmos-
2
pheric pressure CVD (APCVD), yielding Co films with an unac-
ceptably rough surface morphology. At reduced reactor pressure,
2
8
deposition from cobaltocene does not occur on any substrate,
while use of C H Co(CO) results in Co films containing up to 50
5
5
2
2
2
atom carbon.
1
2
age. Other attractive features of CVD include the ability to uni-
formly coat aggressive substrate topographies and compatibility
It is thus apparent that the cobalt precursors discussed do not
exhibit the desired properties for applications in a CVD-based sali-
cide process. Accordingly, we have identified cobalt tricarbonyl
1
3
with emerging microelectronics fabrication flows. Furthermore,
the substrate surface could be actively involved in the CVD reaction
pathways, thereby allowing tight control of film texture and grain
size, including the formation of metastable phases and epitaxial
nitrosyl, Co(CO) NO, as a more suitable precursor for such applica-
3
tions. This assessment is based on its favorable chemical and physi-
cal properties, which include high volatility (vapor pressure 26 Torr
at 0ЊC and 100 Torr at 25ЊC), availability in liquid form, and rela-
1
4
growth. These features can possibly lead to silicide formation with
superior performance as compared to conventional physical deposi-
tion methods, especially given the strong dependence of the silicide
on the metal/Si interfacial conditions.
2
5,30
tively good stability (decomposition temperature >66ЊC).
Fur-
thermore, Co is in an oxidation state of zero in this compound, which
implies that metal-ligand bonds are easily cleaved to yield pure
3
1
A key component in the development of a viable Co CVD
process is the identification of a suitable precursor source with the
chemical characteristics required for growth of pure Co films at
acceptable delivery and process temperature windows. The majority
of CVD Co work has focused on fabrication of Co layers for mag-
metal films according to
Co(CO) NO(g) r Co(s) + 3CO(g) + NO(g)
[1]
3
In this respect, Co(CO) NO has not been studied to the same
3
extent as other metal carbonyls, and most of the research has focused
1
5,16
17,18
neto-optic applications,
magnetic Co-containing alloys,
or
15,16
on photolysis reactions.
Accordingly, we sought to develop a Co CVD-based route to the
*
Electrochemical Society Active Member.
E-mail: ak127@csc.albany.edu
formation of CoSi for applications in sub-0.25 ␮m microelectron-
ics technology. In this respect, the present paper focuses on the
2
z
Downloaded on 2015-01-31 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2
140
Journal of The Electrochemical Society, 146 (6) 2139-2145 (1999)
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
development and optimization of CVD processing conditions, lead-
ing to the deposition of pure Co films with viable properties and per-
formance. The incorporation of the CVD Co process in the forma-
Table I. Process regime investigated for CVD Co. For all experi-
ments, the precursor source temperature was kept at 0ЊC.
tion of device-quality CoSi is the subject of a subsequent report by
the present authors.
2
Substrate
temperature
Precursor
flow
(sccm)
H reactant
flow
(sccm)
Reactor Deposition
2
pressure
(Torr)
time
(min)
(ЊC)
Experimental
Process development was performed in a standard custom-de-
signed, stainless steel alpha-type CVD reactor.
The precursor was delivered to the reactor chamber using an
MKS type 1150C vapor source mass flow controller (MFC) to en-
sure reliable and reproducible delivery. The unit was calibrated to a
Substrate temperature series
2
2
2
3
330
360
390
4
4
4
10
40
70
00
0.51
0.51
0.51
0.51
0.51
0.51
0.51
0.51
0.51
0.51
1750
1750
1750
1750
1750
1750
1750
1750
1750
1750
1.51
1.51
1.51
1.51
1.51
1.51
1.51
1.51
1.51
1.51
15
15
15
15
15
15
15
15
15
15
maximum flow of 5 sccm of Co(CO) NO. During actual processing,
3
the 1150C MFC was maintained at 50ЊC, while the upstream and
downstream delivery lines were kept at 40 and 55ЊC, respectively, in
order to prevent precursor recondensation.
20
50
80
The precursor was maintained in a specialized glass reservoir,
which was immersed in an ice bath to ensure that the precursor tem-
perature was constantly maintained at 0ЊC. This source temperature
was mandated by the need to maintain a highly controllable precur-
sor vapor pressure, given that the performance of the MKS 1150
MFC is dependent on the inlet (upstream) and outlet (downstream)
pressures. From this standpoint, the vapor pressure of the source, as
driven primarily by its temperature in this case, plays a significant
role in the MFC’s delivery profile.
In the reactor geometry employed, the precursor delivery lines
and gas sidelines were connected to a cone-type showerhead, which
was mounted on the top of the reactor chamber and designed to pro-
vide uniform gas flow over the entire substrate surface. In this geom-
etry, reactants were flowed downward with pumping being carried
out from the bottom of the reactor. The CVD system was equipped
Deposition time series
3
3
3
3
3
3
90
90
90
90
90
90
0.51
0.51
0.51
0.51
0.51
0.51
1750
1750
1750
1750
1750
1750
1.51
1.51
1.51
1.51
1.51
1.51
13
15
10
15
20
30
Precursor flow series
3
3
3
3
90
90
90
90
0.25
0.51
111.
1.51
1750
1750
1750
1750
1.51
1.51
1.51
1.51
20
10
17
15
Hydrogen reactant flow series
3
90
390
90
0.51
0.51
0.51
0.51
0.51
1500
1650
1750
1850
1000
111.
1.28
1.51
1.67
211.
10
10
10
10
10
with two separate gas delivery lines for Ar and H , whose flow was
2
controlled using MKS type 1259C electronic MFCs. The process
chamber walls were maintained at 60 Ϯ 10ЊC.
3
390
390
In a typical CVD experiment, the samples were placed on a resis-
tive heater positioned directly under the showerhead. The reactor
Ϫ6
was pumped down to a base pressure of 10 Torr using a Seiko-
Film thickness was measured by RBS and cross-sectional SEM.
RBS analyses were carried out on a 4.5 MeV Dynamitron model
P.E.E. 3.0 linear accelerator. RBS spectra were taken using a 2 MeV
Seiki model STP-300 turbomolecular pump. The substrates were
then heated to the desired processing temperature and the reactants
introduced into the reactor chamber. A mechanical pump was used
for process pumping, and the pumping speed was controlled with a
manual throttle valve.
ϩ
He beam. SEM investigations employed a Zeiss DSM 940 micro-
scope using a 20 keV primary electron-beam.
Deposition rates were defined as T/t, where T is film thickness
and t deposition time. Deposition time was measured as the period of
time during which the precursor was delivered to the reactor. Four-
point probe measurements were performed to determine the sheet
resistance on the blanket samples. The film resistivity was then cal-
culated using the thickness values measured with RBS, since thick-
ness measurements by RBS and SEM were in excellent agreement.
Film composition and associated elemental chemical states were
investigated using XPS. The XPS measurements were carried out on
a Perkin-Elmer PHI 5500 multitechnique system with spherical
capacitor analyzer. The gold f7/2 line was used as a reference and the
analyzer calibrated accordingly. The primary X-ray beam was gen-
erated with a Mg K␣ X-ray source at operating power of 300 W and
15 keV applied to the anode. The results were standardized using a
sputtered Co sample.
XRD analysis were done on a Scintag XDS 2000 X-ray diffrac-
tometer equipped with a Cu K␣ X-ray source and a horizontal wide-
angle four-axis goniometer with stepping motors which allowed in-
dependent or coupled theta/2 theta axes motion. XRD spectra were
collected in the low-angle incidence geometry, with an angle of inci-
dence of 5Њ being used (angle as measured between the incident
beam and the sample surface). In this mode, data collection is per-
formed by fixing the incidence angle at 5Њ, and the XRD pattern is
collected for 2␪ ranging 25-65Њ. The advantages of this approach
include increased sensitivity due to the longer X-ray transmission
path through the thin film and elimination/minimization of the high-
intensity peak from the Si substrate. The collected XRD patterns
were compared to the Co reference patterns from the standard Joint
The silicon substrates used in this study were cleaned ex situ by
rinsing with acetone and methanol, and then dried using compressed
high-purity argon gas. The native oxide was subsequently removed
with a 1 min dip in a 10% hydrofluoric acid solution followed by a
deionized water rinse. Finally, the samples were again dried in a
high-purity argon atmosphere. Additional substrates used were blan-
ket silicon dioxide (SiO ) and silicon nitride (Si N ) on Si, and
2
3 4
Sematech produced patterned structures with nominal 0.25 ␮m, 4:1
aspect ratio (AR) features. These samples were subjected to acetone
and methanol rinse only.
As mentioned earlier, the effects of four key process parameters
on resulting film characteristics were investigated. These were sub-
strate temperature, deposition time, precursor flow, and H reactant
2
gas flow. The process regime investigated is given in Table I. It should
be noted that subsequent analyses showed that the resulting film char-
acteristics were not dependent on the type of substrate used (i.e., n-
or p-type Si, SiO , or Si N ). Thus, the results reported herein are pri-
2
3 4
marily from the analyses of films deposited on Si substrates.
Analytical Techniques Employed
The compositional, structural, and electrical properties of the
deposited Co films were thoroughly analyzed by Rutherford back-
scattering spectrometry (RBS), X-ray photoelectron spectroscopy
(
XPS), X-ray diffraction (XRD), four-point resistivity probe, scan-
ning electron microscopy (SEM), and atomic force microscopy
AFM). The resultant findings are presented in terms of the properties
of the Co films produced under the process regime given in Table I.
(
Downloaded on 2015-01-31 to IP 129.81.226.78 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, 146 (6) 2139-2145 (1999)
2141
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
Figure 3. Typical low angle incidence XRD pattern of Co film deposited at
80ЊC.
1
Results and Discussion
Study of reaction kinetics.—Figure 1 displays an Arrhenius plot
of the ln (growth rate) vs. the reciprocal absolute temperature for
thermal CVD Co from Co(CO) NO. As can be seen in Fig. 1, no film
Figure 1. Arrhenius plot for CVD Co from Co(CO) NO as a function of sub-
3
3
strate temperature.
deposition was observed below 150ЊC. Between 150 and 200ЊC, the
films were a cobalt oxide phase which was heavily contaminated
with carbon and nitrogen, as supported by the XPS and XRD data
shown in Fig. 2 and 3, respectively, and listed in Table II, and were
typically electrically insulating.
The observation of the CoO phase is supported by the work of
3
2
Roustan et al., which indicated the existence of Co(CO) NO reac-
3
tion pathways which are thermodynamically favorable to the growth
of CoO. Roustan et al. performed time-resolved Fourier transform
infrared (FTIR) and mass-spectrometry studies of the adsorption and
reaction of Co(CO) NO on a catalytic surface, namely, activated alu-
3
3
2
mina (Al O ). The latter was maintained at room temperature.
2
3
Under such experimental conditions, Roustan et al. observed the
adsorption of CoNOCO radicals to the alumina surface. Further-
more, they found that surface isocyanates (NCO) were readily
formed from NO and CO ligands, with the Co atom being the most
likely candidate to act as acceptor for the extra oxygen atom. The net
result was the formation of a CoO phase which was contaminated
with C and N.
Between 200 and 250ЊC, no carbon was detected in the CoO
phase, even though the films were still contaminated with N, as doc-
umented by the XPS and XRD data shown in Fig. 2 and 4a. This
finding is supported by the energetics of chemical bonding configu-
Figure 2. XPS compositional trends for CVD Co films as a function of sub-
strate temperature.
Committee for Powder Diffraction Standards (JCPDS) powder dif-
fraction file (PDF).
AFM was performed on a Digital Instruments Nanoscope III mul-
timode scanning probe microscope. Data was collected in tapping
mode AFM with Si cantilevers at resonance frequencies in the range
rations in Co(CO) NO. In particular, it is known that the Co–NO
3
bond in Co(CO) NO is a three-electron ␲-type bond, which is short-
3
3
1
er than the Co–CO bond. These characteristics indicate that the
Co–NO bond is thus stronger, and is thus expected to require even
higher energy to cleave, than the Co–CO bond. It is suggested that
in the temperature regime 200-250ЊC, the thermal budget available
is conducive to complete dissociation of the Co–CO but not the
Co–NO bond. This results in the presence of Co–NO radicals on the
200-300 kHz. Surface roughness was calculated as the standard devi-
ation of the mean height of surface structures. Multiple surface loca-
tions were probed for each sample using a variety of scan sizes rang-
ing from 500 ϫ 500 nm to 10 ϫ 10 ␮m. The average surface grain
size was determined using standard statistical averaging techniques.
Hydrogen incorporation in representative samples was analyzed
3
2
Si surface. As proposed by Roustan et al., these radicals subse-
quently react to produce a nitrogen-contaminated CoO phase.
Alternatively, in the process temperature range 250-350ЊC, the
films were mostly a Co phase, which included a small concentration
15
on the Dynamitron linear accelerator using a N beam. The nuclear
15
12
reaction N(p, ␣␥) C with resonance energy of 6.385 Ϯ 0.005 MeV
was employed.
Table II. Summary of film compositional characteristics as a function of substrate temperature.
Temperature range
150-200ЊC
210-250ЊC
250-350ЊC
350-480ЊC
Oxygen, atom % (XPS)
Nitrogen, atom % (XPS)
Carbon, atom % (XPS)
15-25
5-7
4
15-20
4-5
<1
2.5-10
<1
<1
<1 atom %
<1 atom %
<1 atom %
Resistivity, ␮⍀ cm (four-point probe)
Structure (XRD)
rms surface roughness, nm (AFM)
>1000
CoO
ϳ10
>1000
CoO
6-7
15-40
hcp Co
20-30
8.5-11 ␮⍀ cm
fcc Co
12-18 (up to 450ЊC)
1
10 (above 450ЊC)
Average surface grain size, nm (AFM)
ϳ70
63
135
75-112 (up to 450ЊC)
00-3000 (above 450ЊC)
5
Downloaded on 2015-01-31 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2
142
Journal of The Electrochemical Society, 146 (6) 2139-2145 (1999)
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
of the CoO oxide phase. Finally, in the temperature range above
50ЊC, the films were Co metal, as documented by the XPS and
3
XRD data shown in Fig. 2 and 4b. The observed trend in the depen-
dence of growth rate on substrate temperature indicates that the
CVD Co process is in the mass-transport-limited regime. Our results
seem to indicate that the deposition of pure cobalt occurs upon the
availability of sufficient thermal energy for complete precursor dis-
sociation and complete desorption of reaction by-products from the
substrate surface prior to the occurrence of the undesirable reactions
discussed.
Effects of substrate temperature on film properties.—The de-
pendence of film composition on substrate temperature has already
been discussed in the first section and summarized in Table II.
Therefore, only relevant data is presented herein for the sake of clar-
ity and completion.
Figure 5. Ttypical XPS depth profile of a pure, 140 nm thick, Co film
deposited at 390ЊC.
Compositional uniformity across the bulk of the films was deter-
mined by XPS depth profiling and found to be consistently uniform
throughout the entire process window investigated. This is illustrated
in Fig. 5 for pure Co films grown at 390ЊC. Additionally, accurate
chemical phase determination was achieved by examining the loca-
tion (E) of the high-resolution XPS Co 2p3/2 core peak as well as the
difference in binding energies (␦) between the Co XPS 2p3/2 and 2p1/2
core peaks. For example, a determination of the chemical phase pre-
sent in samples deposited at 390ЊC was performed by comparing
associated XPS data, with E ϭ 785.5 eV, ␦ ϭ 15.08 eV, to those of a
Co reference standard, with E ϭ 778.5 eV, ␦ ϭ 15.08 eV. The XPS
data thus showed that the films were a pure Co metal phase. Similar-
2ϩ
ly, the detection of Co XPS shake-up peaks in the samples deposit-
ed at 210ЊC, along with the observation of E ϭ 778.96 eV, ␦ ϭ
15.20 eV, indicate that the films were primarily a CoO phase, for CoO
standard, E ϭ 780.4 eV, ␦ ϭ 15.5 eV. The overall dependence of film
composition on substrate temperature is displayed in Fig. 2.
Hydrogen profiling indicated that hydrogen content in the films
was inversely proportional to substrate temperature, with 10 atom %
hydrogen incorporated in samples deposited at 210ЊC, while the
samples grown at 390ЊC were hydrogen-free.
The dependence of film texture on substrate temperature was also
investigated by low angle incidence XRD, as shown in Fig. 4a and
4
b. All films were polycrystalline, regardless of the chemical phase
present or substrate temperature used. Below a substrate temperature
of 250ЊC, the XRD spectra corresponded primarily to a CoO face-
centered-cubic (fcc) phase, as documented by the presence of the
CoO (111), (200), and (220) reflection peaks. No Co metal reflection
peaks were detected within the detection limits of XRD. However,
all peak positions were shifted from the reference CoO XRD pattern,
as obtained from the Powder Diffraction File (JCPDS PDF no. 9-
4
02). This shift was attributed to the existence of a nonstoichiomet-
ric oxide phase. In addition, lattice constant measurements, as
derived from the (111) peak positions, yielded values of 4.44 and
4
.48 Å for samples deposited at 210 and 240ЊC, respectively. Alter-
natively, the lattice constant for the reference CoO phase is 4.26 Å.
As a result, it is suggested that the higher lattice constant observed
for the CoO samples could be attributed to the presence of a signif-
icant concentration of defects in the CoO matrix. In particular, oxy-
gen concentrations were only about 15-20 atom % in the CVD CoO
films grown between 200 and 250ЊC, vs. 50 atom % in the reference
CoO sample.
For substrate temperatures between 250 and 300ЊC, the XRD
spectra corresponded primarily to a Co hexagonal-close-packed
(
(
hcp) phase, as documented by the presence of the hcp Co (100),
002), and (101) reflection peaks. The locations of these peaks are in
agreement with the reference hcp Co XRD pattern, as obtained from
the JCPDS PDF no. 5-727. Corresponding lattice constants for the
hexagonal structure were a ϭ 2.16 Å and c ϭ 4.06 Å. Alternatively,
in the substrate temperature range 300-350ЊC, XRD signaled the
presence of a mixture of the hcp and fcc phases, as documented by
the detection of the additional Co(111) and (200) fcc reflection
peaks. The intensity of the Co fcc peaks increased gradually with
higher substrate temperature, while the height of their hcp counter-
parts decreased in a similar fashion. This behavior indicated that the
fcc phase was becoming the dominant phase with increased thermal
budget. Finally, all films grown above 450ЊC consisted entirely of a
Co fcc phase. The corresponding lattice constant for the cubic struc-
ture was a ϭ 3.54 Å, in excellent agreement with the reference Co
XRD pattern, as obtained from the JCPDS PDF no. 15-806.
Film resistivity values were consistent with the observed compo-
sitional trends across the entire process window investigated, as
summarized in Table II and plotted in Fig. 6. As expected, film resis-
Figure 4. Low angle incidence XRD patterns of Co films deposited at substrate
temperature in the range (a, top) 200-350ЊC and (b, bottom) 360-500 ЊC.
Downloaded on 2015-01-31 to IP 129.81.226.78 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, 146 (6) 2139-2145 (1999)
2143
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
Surface grain morphology and associated roughness were also
highly dependent on substrate temperature. In particular, films
grown in the temperature window 200-250ЊC exhibited a dense sur-
face grain morphology, with average grain size around 63 nm
(Fig. 7a). Corresponding rms surface roughness ranged between 6
and 7 nm for ϳ45 nm thick films, or ϳ1.5% of total film thickness.
Conversely, films deposited between 250 and 350ЊC showed a
markedly different surface grain morphology, with frequently well-
separated, irregularly shaped grains, as shown in Fig. 7b. Average
surface grain size was ϳ135 nm.
For films grown at processing temperatures in the range 350-
4
50ЊC, the films exhibited smooth and dense surface morphology.
For example, the rms surface roughness and average grain size were
2.5 and 75 nm, respectively, for a 140 nm thick film grown at 390ЊC
Fig. 7c). Above 450ЊC, films exhibited significant increase in sur-
face roughness, from an rms value of 37 nm for films grown at
50ЊC, up to 110 nm for their 480ЊC counterparts. Corresponding
1
(
4
Figure 6. CVD Co film resistivity as a function of substrate temperature.
surface morphology consisted primarily of an underlying layer of
large grains, with sizes ranging from ϳ0.5 to ϳ3.0 ␮m, covered
with a matrix of smaller grains, approximately 10 nm diam, as
shown in Fig. 7d.
tivity was elevated in the oxide phase, with the resistivity increasing
with higher contamination. Alternatively, resistivity decreased from
values in the range 15-40 ␮⍀ cm for substrate temperatures of 250-
Evolution of film characteristics as a function of deposition
time.—The evolution of film structure and morphology as a function
of processing time was examined by performing deposition runs for
3, 5, 10, 15, 20, and 30 min, and examining the structural and com-
positional characteristics of the resulting films by XPS, XRD, AFM,
SEM, RBS, and four-point resistivity probe. For these runs, substrate
3
50ЊC, to values in the range 8.5-10.0 ␮⍀ cm for substrate temper-
atures of 350-450ЊC. In this respect, the increase in intensity and
associated decrease in full-width-at-half-maximum (fwhm) of the
XRD (111) reflection of the Co metal fcc phase, as shown in Fig. 4b,
indicates a more textured Co phase with increased thermal budget.
Figure 7. AFM images of Co films deposited at substrate temperature of (a, top left) 210, (b, top right) 330, (c, bottom left) 390, and (d, bottom right) 480ЊC.
The scale in (d) was magnified 10 times in order to provide a better illustration of associated film morphology.
Downloaded on 2015-01-31 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).

2
144
Journal of The Electrochemical Society, 146 (6) 2139-2145 (1999)
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
temperature, precursor flow, hydrogen reactant flow, and reactor
pressure were maintained at 390ЊC, 0.5 sccm, 750 sccm, and
flow. At 1.5 sccm, however, XPS indicated the presence of ϳ1.5 atom
% oxygen. This contamination is attributed to oxygen inclusion upon
film exposure to air, as supported by XPS and AFM. In particular, the
high-resolution Co 2p3/2 core peak position, as well as the difference
in binding energies (␦) between the Co 2p3/2 and 2p1/2 XPS core
peaks, corresponded to a pure Co phase. Additionally, AFM micro-
graphs of the surface morphology of the Co film showed the presence
of frequently well-separated, irregularly shaped grains. These find-
ings seem to suggest that diffusion from air through intergranular
gaps might have led to the incorporation of oxygen at Co grain
boundaries. Corresponding resistivity was ϳ15.0 Ϯ 1.5 ␮⍀ cm.
1
.5 Torr, respectively.
RBS and SEM findings indicated that film thickness was a linear
function of deposition time, as shown in Fig. 8. The resulting line
could be extrapolated directly through the origin of the x-y axes,
which indicated the absence of an incubation period and the occur-
rence of Co growth immediately upon exposure of the substrate to
precursor flux. The corresponding slope yielded an average growth
rate of ϳ10 nm/min.
XRD analyses showed the presence of a Co hcp phase for films
grown within the first 5 min of substrate exposure to precursor flux,
or film thickness below ϳ50 nm. The intensity of the Co fcc peaks
increased gradually with longer run times, while the height of their
hcp counterparts decreased in a similar fashion. This behavior result-
ed in the detection of mixed hcp and fcc Co phases for film thick-
ness between 50 and 300 nm. Finally, only the Co fcc phase could
be seen for run times above 30 min, or film thickness above 300 nm.
As expected, all films were pure Co, within the detection limits
of XPS and RBS. Corresponding resistivity was ϳ15.0 Ϯ 1.5 ␮⍀
cm for ultrathin films with thickness below 50 nm. This value
decreased to ϳ9.0 Ϯ 1.0 ␮⍀ cm for films with thickness in the range
Effects of hydrogen flow on film properties.—It is well estab-
lished that hydrogen plays a key role in the decomposition process
2
4
of metal carbonyls, leading to the deposition of pure metal. In par-
ticular, it was reported that hydrogen greatly facilitates the removal
of reaction by-products, thereby preventing C and O contamination
2
7
in Co deposition from carbonyl precursors.
Our findings are in agreement with these observations, as docu-
mented by a study which varied hydrogen flow rates between 500 and
1000 sccm, which corresponded to hydrogen partial pressures in the
range ϳ1.0-2 Torr. For these runs, substrate temperature and precur-
sor flow were maintained at 390ЊC and 0.5 sccm, respectively. In par-
ticular, films deposited at the lower hydrogen flow rate of 500 sccm
were found to contain up to 2 atom % oxygen and exhibited higher
resistivity values in the range 15 to 16 Ϯ 1.5 ␮⍀ cm. This is consis-
tent with earlier studies by the present investigators which showed
that films grown with hydrogen flows in the range 250-350 sccm
were contaminated with up to 50 atom % oxygen, leading to the for-
mation of an insulating CoO phase. Conversely, films grown at hy-
drogen flows above 500 sccm were pure Co, as documented by XPS,
with resistivities in the range 8.0 to 10.0 Ϯ 1.0 ␮⍀ cm.
Interestingly, growth rates were found to decrease with higher
hydrogen flow at a constant precursor flow, as illustrated in Fig. 10.
This trend could be attributed to the role of increased hydrogen flow
in reducing precursor residence time in the reaction zone, which
decreases the probability for precursor reaction and generates a
reduction in growth rates. Another possibility is the saturation of a
higher density of substrate surface sites with hydrogen at elevated
flow, thus causing a reduction in the availability of surface sites for
precursor adsorption and subsequent reaction. The net effect is a
reduction in Co growth rates. In this respect, our studies indicated
that the rate of rise in pressure was not linear with increased hydro-
gen flow, a trend that tends to support the reduction in precursor res-
idence time with higher hydrogen flow.
5
0-200 nm and ϳ7.0 Ϯ 0.5 ␮⍀ cm for thicker films. All films exhib-
ited smooth and continuous surface morphology, with rms surface
roughness being consistently below 10.0 Ϯ 2.0% of film thickness
and average surface grain size ϳ70.0 Ϯ 3.0 nm.
Effects of precursor flow on film properties.—For this study, pre-
cursor flow was varied between 0.25 and 1.5 sccm, which corre-
sponded to precursor partial pressures in the range 0.5-3 mTorr. For
these runs, substrate temperature, hydrogen reactant flow, and reac-
tor pressure were maintained at 390ЊC, 750 sccm, and 1.5 Torr,
respectively. Based on an initial set of screening studies, deposition
times were adjusted to obtain a constant film thickness of ϳ100 nm
for the various precursor flows used. The purpose was to eliminate
any thickness-induced variations in film properties.
In this respect, Fig. 9 plots film growth rates as a function of pre-
cursor flow. As can be seen in the figure, growth rates exhibited a lin-
ear dependence on precursor flow, with growth rates increasing near-
ly 80% in response to a 100% rise in precursor flow. This behavior
could indicate that virtually all the precursor is being consumed in
the CVD decomposition reaction within the process window inves-
tigated. Alternatively, the linear dependence on flow rate can possi-
bly be attributed to first-order kinetic dependence on precursor pres-
sure, since this variable appears to change linearly with precursor
flow rate.
XRD analyses showed the presence of mainly a Co fcc phase,
mixed with a small fraction of the Co hcp phase, for flows below
For precursor flows below 1.5 sccm, all films were pure Co, with-
in the detection limits of XPS and RBS. Corresponding resistivity
increased from 9.0 to 11.0 Ϯ 1.0 ␮⍀ cm as a function of precursor
1
000 sccm. For the latter, the films consisted solely of the fcc phase.
Figure 9. CVD Co film growth rate as a function of Co(CO) NO flow.
Figure 8. CVD Co film thickness as a function of deposition time.
3
Downloaded on 2015-01-31 to IP 129.81.226.78 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, 146 (6) 2139-2145 (1999)
2145
S0013-4651(98)08-015-X CCC: $7.00 © The Electrochemical Society, Inc.
Figure 11. Typical conformality within the process window investigated for
CVD Co films in Sematech’s nominal 0.25 ␮m, 4:1 aspect ratio, trench
structures.
Figure 10. CVD Co film growth rate as a function of hydrogen reactant flow.
Co conformality within the process window investigated.—The
References
conformality of Co films was investigated in Sematech’s nominal
1. The National Technology Roadmap for Semiconductors, Semiconductor Industry
Association, San Jose, CA (1997).
0
.25 ␮m, 4:1 aspect ratio, trench structures. For these studies, the
2
3
.
.
K. Maex, Semicond. Int., 18, 75 (1995).
process window consisted of substrate temperature, hydrogen reac-
tant flow, precursor flow rate, and reactor pressure values of 390ЊC,
S. L. Hsia, T. Y. Tan, P. Smith, and G. E. McGuire, J. Appl. Phys., 72, 1864 (1992),
and S. L. Hsia, T. Y. Tan, P. Smith, and G. E. McGuire, J. Appl. Phys., 70, 7579
(1991).
7
50 sccm, 0.5 sccm, and 1.5 Torr, respectively. It was found that the
4.
5.
6.
7.
G. Bai and R. Stivers, Mater. Res. Soc. Symp. Proc., 402, 215 (1996).
M. L. A. Dass, D. B. Fraser, and C.-S. Wei, Appl. Phys. Lett., 58, 1308 (1991).
K. Maex, Mater. Sci. Eng., R11, 53 (1993).
A. Vantomme, M.-A. Nicolet, G. Bai, and D. B. Fraser, Appl. Phys. Lett., 62, 243
(1993).
ratio of film thickness at the bottom of the trench to that in the field
was ϳ55%, as shown in Fig. 11. Alternatively, the ratio of film thick-
ness at the bottom of the trench to that on the sidewalls was ϳ110%.
Further studies are underway to establish an understanding of the
underlying mechanisms that drive film conformality in aggressive
device topographies and derive optimized process parameters for
enhanced step coverage in subquarter-micron device structures.
8
9
.
.
S. Nygren and S. Johansson, J. Appl. Phys., 68, 1050 (1990).
E. G. Colgan, C. Cabral, Jr., and D. E. Kotecki, J. Appl. Phys., 77, 614 (1995).
1
1
0. B.-S. Chen and M.-C.Chen, IEEE Trans. Eletron Devices, ED-43, 258 (1996).
1. S.-L. Zhang, J. Cardenas, F. M. d’Heurle, B. G. Svensson, and C. S. Peterson, Appl.
Phys. Lett., 66, 58 (1995).
1
1
2. J. R. Creighton and J. E. Palmer, Crit. Rev. Solid State Mater. Sci., 18, 175 (1993).
3. S. Sivaram, Chemical Vapor Deposition: Thermal and Plasma Deposition of Elec-
tronic Materials, Van Nostrand Reinhold, New York (1995).
4. J.-O. Carlsson, in Handbook of Deposition Technologies for Films and Coatings,
R. F. Bunshah, Editor, p. 374, Noyes Publications, Park Ridge, NJ (1994).
Conclusions
This paper reported the development of a thermal CVD process
for the deposition of pure Co from the source precursor cobalt tri-
1
carbonyl nitrosyl, Co(CO) NO. For this purpose, a study of the reac-
tion kinetics involved in the CVD process indicated that the forma-
tion of pure Co occurs in a mass-transport-limited regime above
15. S. W.-K. Choi and R. J. Puddephatt, Chem. Mater., 9, 1191 (1997).
3
1
1
1
6. T. Maruyama and T. Nakai, Appl. Phys. Lett., 59, 1433 (1991).
7. D. K. Liu, U.S. Pat. 5,171,610 (1992).
8. R. S. Dickson, P. Yin, M. Ke, J. Johnson, and G. B. Deacon, Polyhedron, 15, 2237
3
50ЊC. Additionally, systematic studies were implemented to estab-
(1996).
lish functionality curves for the dependence of key film properties,
namely, purity, texture, resistivity, and morphology on critical pro-
cess parameters. These parameters were identified as precursor flow,
hydrogen reactant flow, substrate temperature, and deposition time
19. K. L. Hess, S. W. Zehr, W. H. Cheng, J. Pooladdej, K. D. Buehring, and D. L. Wolf,
J. Cryst. Growth, 93, 576 (1988).
2
2
0. K. L. Hess and S. W. Zehr, U.S. Pat. 5,045,496 (1998).
1. C. J. Smart, S. K. Reynolds, C. L. Stanis, A. Patil, and J. T. Kirleis, Mater. Res. Soc.
Symp. Proc., Vol. 282, p. 229, Materials Research Society, Pittsubrgh, PA (1993).
(
thickness). Subsequent microstructural and microchemical analyses
22. G. J .M. Dormans, G. J. B. M. Meekes, and E. G. J. Starling, J. Cryst. Growth, 114,
64 (1991).
3
identified an optimized process window for the growth of pure Co
with resistivity of 9.0 Ϯ 2.0 ␮⍀ cm, smooth surface morphology,
and rms surface roughness at or below 10% thick.
2
3. G. A. West and K. W. Beeson, Appl. Phys. Lett., 53, 740 (1988).
4. C. F. Powell, in Vapor Deposition, C. F. Powell, J. H. Oxley, and J. M. Blocher, Jr.,
Editors, p. 249, John Wiley & Sons, New York (1966).
2
2
5. Strem Chemicals, Inc., catalog no. 17 (1997-99).
2
6. M. E. Gross, K. S. Kranz, D. Brasen, and H. Luftman, J. Vac. Sci. Technol., B6,
Acknowledgments
1
548 (1988).
2
7. J. P. Candlin, K. A. Taylor, and D. T. Thompson, Reactions of Transition-Metal
Complexes, Elsevier Publishing Co., New York (1968).
8. G. J. M. Dormans, J. Cryst. Growth, 108, 806 (1991).
The work was supported by the New York State Center for
Advanced Thin Film Technology (CAT), and Gelest, Inc. This sup-
port is gratefully acknowledged.
2
29. T. Maruyama, Jpn. J. Appl. Phys., 36, L705 (1997).
3
3
0. Landolt-Börnstein, 6th ed., Vol. 2, Part 2a, p. 49, Springer, Berlin (1960).
1. J. A. Mattern and S. J. Gill, in The Chemistry of the Coordination Compounds, J. C.
Bailar, Jr., Editor, p. 509, Reinhold Publishing Co., New York (1956).
The State University of New York at Albany assisted in meeting the pub-
lication costs of this article.
32. J. L. Roustan, Y. Lijour, and B. A. Morrow, Inorg. Chem., 26, 2509 (1987).
Downloaded on 2015-01-31 to IP 129.81.226.78 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).