Copper iminopyrrolidinates: A study of thermal and surface chemistry

Article abstract of DOI:10.1021/ic3021035

Several copper(I) iminopyrrolidinates have been evaluated by thermogravimetric analysis (TGA) and solution based ¹H NMR studies to determine their thermal stability and decomposition mechanisms. Iminopyrrolidinates were used as a ligand for copper(I) to block previously identified decomposition routes of carbodiimide deinsertion and β-hydrogen abstraction. The compounds copper(I) isopropyl-iminopyrrolidinate (1) and copper(I) tert-butyl-iminopyrrolidinate (2) were synthesized for this study, and compared to the previously reported copper(I) tert-butyl-imino-2,2- dimethylpyrrolidinate (3) and the copper(I) guanidinate [Me₂NC( ⁱPrN)₂Cu]₂ (4). Compounds 1 and 2 were found to be volatile yet susceptible to decomposition during TGA. At 165 C in C ₆D₆, they had half-lives of 181.7 h and 23.7 h, respectively. The main thermolysis product of 1 and 2 was their respective protonated iminopyrrolidine ligand. β-Hydrogen abstraction was proposed for the mechanism of thermal decomposition. Since compound 3 showed no thermolysis at 165 C, it was further studied by chemisorption on high surface area silica. It was found to eliminate an isobutene upon chemisoption at 275 C. Annealing the sample at 350 C showed further evidence of the decomposition of the surface species, likely eliminating ethene, and producing a surface bound methylene diamine.

Full text of DOI:10.1021/ic3021035

Article
pubs.acs.org/IC
Copper Iminopyrrolidinates: A Study of Thermal and Surface
Chemistry
Jason P. Coyle,† Peter J. Pallister,† Agnieszka Kurek,† Eric R. Sirianni,‡ Glenn P. A. Yap,‡
and Sean T. Barry*,†
́
†Department of Chemistry, Carleton University, 1125 Colonel By Drive, Ottawa, Canada
‡Department of Chemistry & Biochemistry, University of Delaware, Newark, United States
ABSTRACT: Several copper(I) iminopyrrolidinates have
been evaluated by thermogravimetric analysis (TGA) and
1
solution based H NMR studies to determine their thermal
stability and decomposition mechanisms. Iminopyrrolidinates
were used as a ligand for copper(I) to block previously
identified decomposition routes of carbodiimide deinsertion
and β-hydrogen abstraction. The compounds copper(I)
isopropyl-iminopyrrolidinate (1) and copper(I) tert-butyl-iminopyrrolidinate (2) were synthesized for this study, and compared
to the previously reported copper(I) tert-butyl-imino-2,2-dimethylpyrrolidinate (3) and the copper(I) guanidinate
[Me2NC(iPrN)2Cu]2 (4). Compounds 1 and 2 were found to be volatile yet susceptible to decomposition during TGA. At
165 °C in C6D6, they had half-lives of 181.7 h and 23.7 h, respectively. The main thermolysis product of 1 and 2 was their
respective protonated iminopyrrolidine ligand. β-Hydrogen abstraction was proposed for the mechanism of thermal
decomposition. Since compound 3 showed no thermolysis at 165 °C, it was further studied by chemisorption on high
surface area silica. It was found to eliminate an isobutene upon chemisoption at 275 °C. Annealing the sample at 350 °C showed
further evidence of the decomposition of the surface species, likely eliminating ethene, and producing a surface bound methylene
diamine.
INTRODUCTION
■
Copper metal remains an interesting topic for chemical vapor
deposition (CVD) and atomic layer deposition (ALD) because
of its use in microelectronics, primarily as an interconnect.1
One class of particularly well-studied precursors are the
amidinates, which originated from the Gordon group in the
early 2000s.2 Copper amidinates have utility in ALD2a and
CVD3 processes between temperatures of 150−240 °C when
using hydrogen as a reducing agent. Copper(I)-N,N′-di-
isopropylacetamidinate has been shown to undergo CVD
deposition as low as 140 °C,4 and produces free amidine,
acetonitrile, propene, and iminopropane when allowed to
thermally decompose in the absence of a reducing agent.5
These findings have been corroborated in the surface study
work of copper(I) acetamidinates on nickel whereabove 130
°Cthe self-limiting nature of the monolayer is compromised
and continuous uptake of the copper precursor is observed.
Similar ligand fragments were identified in the surface work as
in the gas phase work.
We have previously investigated the thermal decomposition
of copper(I) guanidinates, which are similar to amidinates
except that the exocyclic group is an amide, rather than an
alkyl.6,7 Using a guanidinate with isopropyl groups on the
chelating nitrogens, we found two distinct thermal decom-
position mechanisms (Figure 1). In solution at lower
temperatures, the guanidinate deinserted carbodiimide (CDI)
and produced the parent amine of the exocyclic amide group.6
In the gas phase at greater than 150 °C, the methylene carbon
Figure 1. Two thermal decomposition pathways for a copper(I)
guanidinate species.
of the isopropyl group on the chelating nitrogen loses a
hydrogen and produces an oxidized guanidine. Hydrogen gas is
also found as a thermolysis side product.7
Both of these thermal decomposition mechanisms are
troubling from the point of view of developing a copper
precursor for ALD. In the first case, the elimination of the
oxidized guanidinate by β-hydrogen abstraction produces a
copper hydride, which readily eliminates dihydrogen or parent
guanidine to produce copper metal. This circumscribes an ALD
process in favor of a CVD process. CDI deinsertion also greatly
limits the use of this precursor in thermal processes. The
formation of a (transient) copper-amide bond creates a species
Received: September 27, 2012
Published: January 3, 2013
© 2013 American Chemical Society
910
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
that will readily decompose,8 thus also preventing ALD in favor
of a CVD process.
imino-2,2-dimethylpyrrolidinate (3),10 and copper(I) N′,N″-
diisopropyl-N,N-dimethyl guanidinate (4). A thorough explora-
tion of the thermal chemistry of these four compounds was
made to demonstrate how the redesign of this ligand has
influenced thermal behavior. We have investigated the thermal
stability of this series in the solid phase using a variety of
thermogravimetric (TG) methods, as well as in solution phase
A redesign of the basic amidinate framework is necessary to
remove the known thermal decomposition pathways of this
ligand family. Our group has developed and extensively
characterized a novel iminopyrrolidinate ligand family that
has evidenced good thermal stability. Recently, we have
reported a family of volatile, stable aluminum compounds.9
This ligand has a five-membered ring linking one chelate
nitrogen to the “bridgehead” (i.e., quaternary) carbon of the
amidinate, thus preventing CDI deinsertion. Likewise, the
ligand has been synthesized to allow control over the number
of β -hydrogens, resulting in a series of four related copper(I)
iminopyrrolidinates (Scheme 1). Herein we introduce two
1
using high resolution H NMR of sealed tubes that have been
treated at elevated temperatures to study decomposition
kinetics. Finally, because of the superior thermal stability of
3, we have undertaken an extensive study of the nature of the
chemisorbed species produced on silica by 3 as deposited at
275 °C and when annealed at 350 °C.
RESULTS AND DISCUSSION
■
Scheme 1. Series of Amidinate-Type Ligands with Differing
Numbers of ß-Hydrogens, Shown As Their Corresponding
Copper(I) Compounds
The copper compounds 1 and 2 were simply made by salt
methathesis of copper(I) chloride with the in situ generated
lithium iminopyrrolidinates. These compounds were isolated as
white crystalline materials in moderate to high yields. Similarly,
a recent paper describes trimeric, heteroleptic copper(I)
guanidinates isolated from the attempted synthesis of [Me2NC-
(NtBu)2Cu]2.11 We believe the ring of the iminopyrrolidinates
reported herein was necessary to slightly reduce the steric bulk
and permit the isolation of 2 as a dimeric species (see below).
1
The H NMR of 2 was simple, suggesting a symmetrical
oligomeric solution structure, similar to known copper
guanidinate dimers.6 However, compound 1 showed a more
complex 1H NMR. The doublets for the methyls of the
isopropyl moiety in 1 suggested several environments. It is
unclear if these multiple environments were due to an
equilibrium of oligomers or from asymmetry in the molecule,
but the relative integrations of these doublets were dependent
on the concentration of 1. Serial dilution of an NMR sample
nor heating of the sample to 60 °C did not yield a trivial set of
1
peaks similar to the spectra of 2. The data from the H NMR
suggest that the species might be in an oligomeric equilibrium
where the exterior proton environments are disordered in the
structure, while the interior ring protons are more disordered.
Fortunately, 1 could be isolated as X-ray quality crystals.
The structure of 1 shows a tetrameric arrangement in the
solid state with an orthorhombic Ccca space group (Figure 2,
Table 1). There are 16 tetramers in the unit cell, but only two
that are crystallographically distinct. These molecules each have
novel copper iminopyrrolidinates (copper(I) isopropyl-imino-
pyrrolidinate, 1 and copper(I) tert-butyl-iminopyrrolidinate, 2)
to complement our previously reported copper(I) tert-butyl-
Figure 2. Structure of (a) 1 and (b) 2, shown with the hydrogens removed for clarity. The thermal ellipsoids are at 30%.
911
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
Considering the ligands oriented parallel to each other and
on the same side of the copper plane, it is obvious that more
crowding exists here. The ligands bend out slightly from one
another, on average 26.5° from parallel. As well, there is a slight
twist of about 6° across each ligand with respect to the line
defined by the coppers which bind them.
The metallocycle C−N bonds are all about equivalent in this
molecule (1.31 Å−1.34 Å), suggesting complete delocalization
of the double bond, similar to 3. The ligand’s bridging N−C−N
angle is about 123°, which is slightly larger than 3 and 4.6 This
is quite understandable in the case of the tetramer, however,
since the ligand is bridging a significantly large Cu−Cu
distance.
The structure of 2 is more typical of the expected dimeric
structures, considering 3 and 4 (Figure 2, Table 1). It
crystallized in the monoclinic P21/c space group with two
molecules in the unit cell centered on inversion centers, where
both molecules in the unit cell are identical. The Cu−Cu
distance is ∼2.5 Å, which is significantly closer than in 1. The
copper geometry is very close to linear, contorting by 4.7° to
lengthen the Cu−Cu bond. The metallocycle core is planar
with the core N−C bonds all essentially equivalent (1.33 Å−
1.34 Å), suggesting that the inherent double bond character is
delocalized between these two bonds.
The thermolyses of 1 and 2 are quite encouraging.
Compound 1 showed low residual mass using thermogravi-
metric (TG) analysis, and a single feature in weight loss,
suggesting that the species volatilized easily. This low residual
mass was surprising. Not only does this compound possess β-
hydrogen atoms for abstraction, but the tetrameric structure
was expected to yield a lower vapor pressure from increased
molecular mass. It is possible that there exists an oligomeric
equilibrium that allows 1 to volatilize as a lower order oligomer
(mass spectral analysis shows a dimer), thus providing a larger
vapor pressure at lower temperature. This also explains why 1
has a similar onset of volatilization to the dimeric compounds 3
and 4.
Table 1. Selected Bond Lengths and Angles for 1 and 2
compound 1
compound 2
selected bond
length (Å)
selected bond
length (Å)
Cu1−Cu2
Cu1−Cu1a
Cu1−N1
Cu1−N3
Cu2−N2
Cu2−N4
N1−C4
2.76
3.00
1.87
1.86
1.89
1.89
1.31
1.34
1.31
1.33
1.44
1.46
1.46
1.49
Cu−Cu
Cu−N1
Cu−N2
N1−C4
N2−C4
2.48
1.86
1.89
1.34
1.33
N2−C4
N3−C11
N4−C11
N1−C1
N2−C5
N3−C8
N4−C12
compound 1
compound 2
selected angle
angle (deg)
selected angle
angle (deg)
Cu2−Cu1−Cu2
Cu2−Cu1−Cu1
N3−Cu1−N1
N2−Cu2−N4
N1−C4−N2
114.2
57.1
N1−Cu−N2
N1−C4−N2
175.3
122.1
170.5
170.7
123.8
123.6
N3−C11−N4
a core of four copper atoms that lay in a rhombohedral plane,
with the rhombus of one molecule centered perpendicular to a
C2 axis (with a minor torsion angle of 1.31°), and the other
parallel to a different C2 axis (with a torsion angle of 0°). The
closest Cu−Cu contact in the unit cell is 2.76 Å, which suggest
there is no significant Cu−Cu interactions.12 This is supported
by the fact that the N−Cu−N bonds only deviate from linearity
by about 10°, which is likely caused by steric hindrance in the
ligands (see below). Since the connectivity of the molecules in
the unit cell is very similar, only one molecule will be discussed
in depth.
Interestingly, the two ligands bonded to any specific copper
are bonded through the same type of nitrogen. For example
Cu1 is bonded to the ring nitrogens N1 and N3, where one
might expect the ligands to alternate. Additionally, the
isopropyl groups of the ligand are all oriented toward the
copper plane. These observations suggest that there is a fine
balance of steric interference across the copper plane allowing a
tetramer to form. This structure type has been seen previously
for group 11.13 The Cu−N bonds in this molecule have an
average length of 1.88 Å, which is the same as in 3.10
Compound 2 showed more complex behavior. The residual
mass was 28.1%, which was close to the percent mass of copper
(31.3%) in this compound. The residual mass was confirmed to
be metallic copper by powder X-ray diffraction, suggesting that
2 undergoes a low temperature reduction to produce Cu(0).
This has previously been observed in the case of both 4 and the
copper amidinates.4 Because of the presence of the exocyclic
pyrrolidine ring, the more favorable path for this thermal
decomposition is β-hydrogen elimination rather than CDI
deinsertion. This can be rationalized considering the crystal
structure of 2. The distortion of the dimer core and lack of
Figure 3. Thermal stress test of compounds (a) 3 showing no change in residual mass with increasing sample mass, and (b) 1, showing increasing
residual mass with increasing initial sample mass.
912
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
steric protection exposes the copper and enables it to interact
with the β-hydrogen atoms of neighboring molecules.
hydrogens and the robust nature of the core of this ligand with
respect to CDI deinsertion.
Using TG analysis, the evaporation kinetics were evaluated
for 1−4 (Figure 5). Compound 4 showed the highest
These compounds were also investigated by a thermal stress
test using TG analysis. Essentially, a compound is measured
using the same temperature ramp rate (10 °C/min) but using
different sample masses. Thus, more sample would see higher
temperatures because of the kinetics of thermolysis or
volatilization. This valuable test can be used to gauge the
behavior of the compound with respect to thermal handling
during a deposition process, whether in the source bubbler or
during the volatilization and entrainment to the deposition
zone. A good example of this is the thermal stress test of 1
(Figure 3b). Compound 1 was shown to have a higher residual
mass as the initial pan loading of the TG was increased. This
indicated that it underwent decomposition as more compound
was exposed to higher temperature. Compound 3 (Figure 3a)
was seen to undergo no decomposition as pan loading was
increased, demonstrating a superior thermal stability.
Figure 5. Evaporation kinetics of 1−4 by TG analysis.
evaporation at the lowest temperatures. Compounds 1 and 2
showed similar evaporation kinetics within the temperatures
studied. This is unsurprising since 1 and 2 are expected to both
volatilize as dimers and have similar molecular masses (a mass
difference of only two methyl groups). Surprisingly, 3 showed
the slowest evaporation kinetics even though it has a mass close
to that of 4. It is possible that the measured evaporation
kinetics of the copper(I) iminopyrrolidinates are slower than 4
because of the more rigid framework of the iminopyrrolidinate
ligand. The rigidity of the ligand would cause a compact, planar
structure and allow for stronger intermolecular attraction within
the solid. Additionally, the copper(I) iminopyrrolidinates would
be expected to have less entropy gain during volatilization than
4: the rigid ligand has fewer bonds that are free to rotate in the
gas phase than in the solid state, compared to the relatively less
constrained guanidinate of 4. From the isothermal TG analysis
and using the Langmuir equation,14 the temperatures at which
1 Torr of vapor pressure was obtained was estimated to be 158
°C, 161 °C, 187 °C, and 125 °C for 1−4, respectively.
To further illuminate the thermal chemistry of these
compounds, they were each sealed in heavy walled NMR
tubes and heated in an oven at 165 °C over a period of days. An
1H NMR was collected each day to observe their thermal
decomposition (Figure 6, Table 2). Compound 4 revealed
decomposition to produce diisopropylcarbodiimide, as pre-
viously seen.6 The decomposition of 4 that followed first order
decomposition kinetics and had a calculated half-life of 33.8 h.
The parent guanidine was also observed as a product of
thermolysis, and its origin of production is discussed below.
The residual mass trends of each compound 1−4 can be seen
in Figure 4. It should be noted that the top sample mass that
Figure 4. Thermal stress test trends for compounds 1−4.
could be accommodated was about 65 mg; this filled the sample
pan in all cases. With the exception of 1, the stress trends are
quite linear with an R2 of 0.99. Compound 1 showed a higher
deviation from linearity with an R2 of 0.97. This might be due
to a higher preponderance of a dimeric rather than tetrameric
species at higher temperature, caused by a shift in oligomer
equilibrium.
Compounds 1 and 4 show the most stress, deviating from a
negligible residual mass at low loadings to quite high residual
masses (13.7% for 1 and 9.0% for 4) at high loadings. It is not
surprising that these two compounds show similar thermal
stress, since they both bear an isopropyl group for β-hydrogen
elimination although CDI deinsertion cannot be ruled out in
the case of 4.
Compound 2 again showed poor thermal behavior with very
high residual masses. Indeed, the highest loading was not
attempted for 2 since the 39.5 mg of sample gave 31.35%,
which is the mass percent of copper in that compound
(31.32%). This was not surprising, given the high residual mass
seen previously. This reactivity stands out because 2 can be
considered of intermediate reactivity with respect to number of
β-hydrogens within this family. However, it has a dimeric
structure (unlike 1, which exists in the solid as a tetramer) and
so is less sterically protected at the copper centers. This would
lend it worse thermal stability.
Compound 3 showed very good thermal stress resistance,
rising only to 0.56% residual mass at 63.0 mg sample mass. This
supports our hypothesis that 3 will be the most thermally stable
of this series of compounds because of its lack of reactive
Figure 6. Thermal decomposition of 1−4 at 165 °C over 21 days.
913
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
analysis, and high-resolution NMR (HR-NMR) to determine
the initial monolayer and surface chemistry of 3 on high-surface
area silica. SS-NMR has been shown to be a useful tool in
analysis of initial adsorption complexes on high-surface area
substrates.16,17
Table 2. Kinetic Data for the Thermal Decomposition of 1−
a
4 at 165°C
rate constant
(s−1
calculated half-life
(h)
correlation
coefficient
compound
)
1
2
3
4
1.06 × 10−6
8.14 × 10−6
181.1
0.902
0.944
Approximately 1 g of high surface area silica was loaded into
a reactor and annealed under vacuum at 350 °C for about 16 h
prior to exposure to precursor. This ensured a consistent
hydroxyl density at the surface between each experiment.
Under these conditions, silica is expected to have a hydroxyl
density of 2−3.5 OH/nm2.18 A bubbler temperature of 165 °C
has been show to produce a sufficient vapor pressure for 3 so
this temperature was used for all experiments.10 A deposition
temperature of 200 °C was attempted but it was determined
that the precursor did not react with the silica. When the silica
that was exposed to vaporous 3 was washed with D2O, only
unreacted, dimeric precursor was found at this temperature,
suggesting that 3 was simply physisorbed. This was quite
interesting, since it demonstrated the stability of 3: it was stable
in aqueous solution for several hours, which was not an
anticipated characteristic of this precursor. As well, it existed in
a physisorbed state on the silica surface for up to 17 h without
decomposing or significantly desorbing.
It was found that a temperature of 275 °C was needed to
chemisorb 3. There are three main experiments that will be
discussed: sample a is unmodified silica, sample b has
undergone chemisorption of 3, and sample c has undergone
chemisorption of 3 with a subsequent anneal at 350 °C for 4 h.
Energy-dispersive X-ray spectroscopy (EDX) showed the
presence of copper on samples b and c. No copper was
detected on sample a, as expected. Interestingly, when b was
allowed to sit in air, it changed color to turquoise-green within
15 min. This is likely a conversion of the as-deposited surface
species to a hydrated copper oxide.
23.7
no measurable decomposition
33.8
5.69 × 10−6
0.995
a
The “correlation coefficient” refers linear fit in Figure 10.
Compound 1 showed much better thermal behavior than 4,
as was expected from the TG data. The compound
decomposed following first order kinetics with a half-life of
181.1 h, and there was very obvious plating of copper on the
NMR tube walls. The main byproduct from the thermolysis was
the parent ligand, isopropyliminopyrrolidine. There was no
evidence of CDI nor any evidence of oxidized ligand. There was
an obvious flocculant in the NMR tube after the thermolysis;
this might be the oxidized ligand wherein a proton was lost.
Compound 2 showed similar decomposition to 1, but much
more quickly (t1/2 = 23.7 h). The NMR tube was again
obscured with plated Cu(0) and had a flocculant precipitate,
and the 1H NMR showed free, protonated ligand. It is possible
that since 1 forms a tetramer, this imparts thermal stability to
the compound, perhaps because the tetramer has to dissociate
to a smaller oligomer to thermalize. This would also account for
the relatively poor linear fit to first order kinetics: if the
tetramer forms at higher concentration, then there would be a
slightly higher rate of decomposition as the overall concen-
tration of 3 dropped and the dimer-tetramer equilibrium began
to favor the dimer.
The full mechanism that produces the protonated ligand
remains unclear. We propose β-hydrogens as the most likely
source for protons for the origin of free ligand. Surface work by
Gordon and Zaera4 found that butene elimination occurs from
copper(I)-N,N′-di-sec-butylacetamidinate leaving an N−H
moiety, which might be another source for protons. However,
1H/X Cross-polarization magic angle spinning solid-state
NMR (CP/MAS SS-NMR) was used as the primary method of
characterization for the samples a−c. The benefit of this
technique is that the observed signals originate only from nuclei
that are in close proximity to protons; in these samples, this
means surface species only. The 29Si CP/MAS SS-NMR spectra
of samples a−c showed typically silica signals (Figure 7).19 In a,
there are signals at around −101.6 ppm, −110.5 ppm and a
shoulder at −92 ppm. The signal at −101.6 ppm is attributed to
silanol groups (∥-Si−OH) at the surface while the signal at
−110.5 ppm is attributed to fully dehydroxylated silicon near
the surface of the bulk sample (∥-Si). The shoulder at −92 ppm
arises from silandiol species at the surface (∥-Si(OH)2).
1
no H signals were observed in the NMR tube thermolysis
experiment that suggest a similar mechanism occurring here.
Evidence for disproportionation of Cu(I) acetamidinates has
also been demonstrated in surface studies on SiO2 substrates.15
Since disproportionation of Cu(I) to Cu(II) and Cu(0) is a
common thermolysis route for Cu(I) compounds, we offer the
following insight. We observed a blue color imparted to 2
1
during the solution based H NMR thermolysis study (Figure
10), suggesting the formation of a Cu(II) species. Attempts to
isolate a Cu(II) compound by crystallization failed and resulted
only in the isolation of crystals of 2 with a blue impurity.
Interestingly, we previously had prepared oxidized guanidine
(i.e., the product of β-hydrogen elimination) from the reaction
between guanidine and a Cu(II) compound.7 Therefore, a
Cu(II) intermediate might be involved in β-hydride abstraction
thermolysis for this class of compound. We are continuing to
study the thermolysis mechanisms of these compounds to gain
a fuller understanding of this complicated thermal behavior.
Compound 3 showed no decomposition at 165 °C over 21
days. It was expected that its ligand (with no β-hydrogens and
without the CDI deinsertion pathway) would impart the best
thermal stability; however, this marked difference was
surprising. To further explore this, 3 was subjected to a
rigorous surface study. This study used a combination of solid-
state nuclear magnetic resonance (SS-NMR), elemental
Figure 7. Solid-state 29Si {1H} CP/MAS NMR of samples a−c.
914
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
a
The middle trace in Figure 7 shows the 29Si CP/MAS SS-
NMR spectrum for b. It is apparent that the only signals
present are similar to those of the unmodified silica. Indeed, the
lack of other signals demonstrated that chemisorbed precursor
interacts with the silica through an oxygen atom, rather than
forming direct precursor-silicon bonds. That is, the precursor
likely chemisorbs through a ∥-Si−O−Cu interaction, rather
than a ∥-Si−Cu or ∥-Si−N. As well, the relative intensities of
the various Si species present at the surface differ from that of a.
The silidiol shoulder vanished in b, and the relative intensity of
the silanol peak decreased in b compared to a. Thus, the active
surface hydroxyl groups are being consumed as precursor is
taken up by the silica surface. However, the continued presence
of a hydroxyl signal showed that saturation has not been
reached under these experimental conditions.
Table 3. Elemental Analysis Samples a−c
sample a
sample b
sample c
C
H
N
wt %
<0.3
2.71
7
4.16
7
mole ratio
wt %
<0.3
<0.3
0.59
17
0.79
16
mole ratio
wt %
0.95
2
1.40
2
mole ratio
a
The data for sample a were below the limit of detection.
appeared to be intact on the surface. However, the carbon ratio
is lower than expected for a completely intact ligand, which can
be explained by the loss of a fragment as large as C4 from the
ligand upon surface adsorption.
To further elucidate the adsorption mechanism a small
amount of sample b was washed with D2O to etch off the
chemisorbed surface species for study using HR-NMR (Figure
9, bottom trace). This relatively clean spectrum shows that all
The 13C CP/MAS SS-NMR spectra of samples a−c were
diagnostic of the nature of the surface species (Figure 8). As
Figure 8. Solid-State 13C {1H} CP/MAS NMR of high surface area
silica samples a−c.
1
Figure 9. High-resolution H NMR of D2O after it was washed over
samples b and c.
expected, a did not show any carbon signals, and highlights the
lack of carbon impurities at the silica surface before exposing
the system to precursor. The trace for b gave four clear signals
centered at 28.5, 35.8, 65.9, and 176.5 ppm. The signal at 176.5
ppm was attributed to the quaternary carbon between the two
chelating nitrogens on the iminopyrrolidinate ligand. The
broadening of this peak was likely due to the rigidity of the
ligand in the chemisorbed surface species. This caused an
incomplete averaging of the dipolar and scalar coupling at the
4.5 kHz MAS spinning speed used. The signal at 65.9 ppm
came from the quaternary carbon in the iminopyrrolidinate
ring. The quaternary carbon from the tert-butyl substituent
likely contributed to this peak as well. This peak has a similar
broadening to the previous peak. The sharp signals 28.5 ppm
and 35.8 ppm were due to the methyl substituents attached to
the ring and the methylene carbons that are part of the ring in
the iminopyrrolidinate ligand respectively. The methyl groups
on the tert-butyl substituent would appear in the same range as
these peaks and so were likely buried therein. The assignment
of these peaks agreed with the 13C HR-NMR reported for the
pure compound 3.10
Elemental analysis for carbon, hydrogen, and nitrogen was
also performed on samples a−c (Table 3). Given the assumed
hydroxyl density at the surface of 2−3.5 hydroxyls/nm2, the
elemental analysis showed about 60% saturation of the surface,
corroborating evidence from the 29Si SS-NMR. To give more
insight into the nature of the surface species present, a molar
ratio of each element was determined relative to nitrogen,
which was fixed at 2 to represent the stoichiometry of nitrogen
in the ligand. Given the molar ratios, the majority of the ligand
the peaks from the surface species match what is known for the
iminopyrrolidinate ligand, with an absence of the signal for the
methyl groups from the tert-butyl substituent. This concretely
demonstrate that the tert-butyl substituent was eliminated upon
initial surface adsorption at 275 °C. The somewhat higher
hydrogen ratio likely had a contribution from the silica hydroxyl
groups, which dehydrated during the combustion analysis.
The thermal stability of the initial chemisorbed species was
examined by annealing b at 350 °C for 4 h under vacuum,
producing sample c. The 13C CP/MAS SS-NMR for c gave a
spectrum that was somewhat similar to that of b (Figure 8).
However, the quaternary carbon signal originally at 176.5 ppm
was not present, and a new quaternary carbon signal appeared
at 160 ppm. The other quaternary carbon signal present in b
also appeared for c with a considerably weakened intensity from
the annealing process. Two new carbon signals have appeared
at 10 and 45 ppm, confirming that a new surface species forms.
The 29Si CP/MAS SS-NMR spectrum for c was very similar to
unmodified silica (Figure 7). Even at temperatures as high as
350 °C, no alkylation of the silicon occurred at the surface, and
any thermolyzed chemisorbed species were still bonded
through a ∥-Si−O-R type interaction. Again, there was a
change in the intensity of the peaks originating from each silica
environment. Similar to the signal from b, the intensity of the
silanol species at the surface decreased with respect to the
signal for completely dehydroxylated silicon. This is likely due
to the participation of silanol protons in the decomposition
mechanism of the chemisorbed species from b.
915
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
Similar to the study done for b, a small of amount of c was
solution experiments, and is an excellent candidate for copper
deposition experiments.
1
washed with D2O and examined via H HR-NMR (Figure 9,
top trace). Similar to the 1H HR-NMR of b, the spectrum for c
was very clean and exhibited only one species. There were two
doublets and two septets, representing iso-propyl groups. There
was an isolated singlet which integrated to two protons relative
to the isopropyl group. This singlet is too shielded to be an
An extensive surface reactivity study of 3 was undertaken,
revealing some interesting thermal behavior. Compound 3 lost
an alkyl group on chemisorption at 275 °C, but was stable to
thermolysis up to 350 °C, demonstrating that the design of the
ligand system did indeed protect the monolayer from
thermolysis. At 350 °C, the surface species underwent further
thermolysis by an unknown mechanism to produce a surface
i
imine group (like PrN=CH2), and thus it could be that both
nitrogen atoms remain in the surface species to form a
methylene diimine. A mass spectral analysis of a (non-
deuterated) aqueous solution found a peak at 89 amu, which
is the mass of this surface species with an additional proton.
Using a hybrid quadrupole time-of-flight mass spectrometer,
the fragmentation of this parent ion supported the formulation
iPrN(H)CH2NH2 (Figure 10). Thus, the ring methyls become
i
species suspected to be the methylenediamine PrN(H)-
CH2NH2.
EXPERIMENTAL SECTION
■
General Considerations. All manipulations involving the syn-
thesis and handling of copper(I) compounds were performed in a
nitrogen filled drybox. The chemicals CuCl and 2.5 M BuLi in hexanes
were purchased from Aldrich Chemical Co. and used as received. All
solvents used in manipulation of copper(I) compounds were ACS
grade and purified from an Mbraun Solvent Purifier System. All other
solvents were ACS grade and used as received. Nuclear Magnetic
Resonance was done on 300 MHz Avance 3 and 400 MHz Bruker
AMX. Canadian Microanalytical Service Ltd. performed combustion
analyses. Thermogravimetric analysis was performed on a TA
Instruments Q50 apparatus located in an MBraun Labmaster 130
Drybox under a nitrogen atmosphere. Isopropyl-iminopyrrolidine and
tert-butyl iminopyrrolidine were prepared by literature procedures.9
Copper(I) Isopropyl-iminopyrrolidinate (1). Isopropyl-imino-
pyrrolidine (5.11 g, 40.49 mmol) was partially dissolved in 60 mL of
Et2O and cooled to 0 °C in an ice bath. A 16.2 mL portion of butyl
lithium was added dropwise, and a suspension formed after stirring for
2 h. In a separate flask, CuCl (4.2 g, 42.42 mmol) was suspended in 60
mL of tetrahydrofuran (THF) and cooled to 0 °C in an ice bath. The
suspension of lithium isopropyl-iminopyrrolidine was added via
cannula to the cooled suspension of CuCl and was then allowed to
warm to room temperature and stirred overnight. The volatiles were
removed from the reaction flask by reduced pressure, and the
remaining solid was stirred in 100 mL of toluene for 10 min. The
cloudy solution was filtered, and the clear filtrate was concentrated
under vacuum and kept at −35 °C for 2 days. A mass of small, needle
crystals was collected by filtration, washed with pentane, and then
Figure 10. Possible thermolysis of the surface species formed by 3 on
high-surface area silica.
the methyls of an isopropyl group, and the quatenary carbon is
protonated to produce a methylene. Interestingly, Chabal et al.
also suggest a diamine as a stable surface species during ALD of
Cu metal on SiO2 substrates.15 Combustion analysis of c did
not clarify the formulation, giving very similar carbon,
hydrogen, and nitrogen molar ratios to that of b (Table 3). A
more rigorous surface study is planned to clarify the 350 °C
thermolysis mechanism.
1
dried under vacuum to afford 5.03 g, 65.8%. Mp 138 °C. H NMR
(400 MHz, C6D6): δ 3.49, δ 3.24 (m, 1H, NCH(CH3)2); δ 3.34, δ 3.26
(t, 2H, NCH2CH2CH2C); 2.04 (t, 2H, NCH2CH2CH2C); δ 1.58
(quintet, 2H, NCH2CH2CH2C); δ 1.487, δ 1.304, δ 1.208, δ 1.158 (d,
6H, NCH(CH3)2). 13C NMR (300 MHz, C6D6) δ 179.84, δ 175.38, δ
54.19, δ 53.30, δ 53.15, δ 51.19, δ 50.48, δ 50.00, δ 29.02, δ 28.69, δ
28.65, δ 27.70, δ 27.55, δ 27.39, δ 27.36, δ 24.99, δ 24.72, δ 24.64.
Combustion analysis, found (calculated): C, 44.63(44.55); H,
7.11(6.94); N, 14.84(14.84).
CONCLUSION
■
A series of copper(I) iminopyrrolidinates were synthesized and
evaluated by thermogravimetric analysis and in C6D6 solutions
by 1H NMR to determine their thermal stability and suitability
as precursors in CVD processes. The employed asymmetric,
cyclic amidinate ligand proved to impart excellent thermal
stability when no β-hydrogens were present. A direct
comparison of the number of β-hydrogens to thermal stability
was complicated by the tetrameric structure of 1, which was
thought to influence thermal stability.
Copper(I) tert-Butyl-iminopyrrolidinate (2). tert-Butyl imino-
pyrrolidine (1.72 g, 12.26 mmol) was dissolved in 60 mL of THF. A
4.9 mL portion of 2.5 M butyl lithium was added, and the solution was
stirred for 2 h. CuCl (1.25 g, 12.62 mmol) was added in one portion,
and the suspension was stirred overnight. Volatiles were removed
under reduced pressure, and the remaining solid was taken up in THF
and filtered. The filtrate was concentrated and then kept at −35 °C for
1 day. Colorless, block crystals were collected by decanting the
solution, washing with pentane, and then drying under vacuum;
A valuable thermal stress test by TGA was devised and
provided complementary thermal stability trends to solution
based thermolysis studies. Advantages of the thermal stress test
by TGA were faster data collection and use of conditions more
pertinent to CVD processes. The solution based thermolysis
studies identified the parent amidine and copper metal as
thermolysis products of 1 and 2. The amidines were produced
through an intramolecular hydrogen abstraction, and a β-
hydride abstraction mechanism was proposed. Compound 3
showed negligible decomposition in both the TGA and the
1
obtained 1.59 g, 63.9%. Mp 148 °C. H NMR (300 MHz, C6D6): δ
3.25 (t, 2H, NCH2CH2CH2C), δ 2.19 (t, 2H, NCH2CH2CH2C), δ
1.63 (quintet, 2H, NCH2CH2CH2C), δ 1.32 (s, 9H, NC(CH3)3). 13C
NMR (300 MHz, C6D6): δ 180.01 (NCH2CH2CH2C), δ 52.45
(NCH2CH2CH2C), δ 33.72 (NC(CH3)3), δ 33.26 (NC(CH3)3), δ
31.64 (NCH2CH2CH2C), δ 26.33 (NCH2CH2CH2C). Combustion
analysis, found (calculated): C, 47.08(47.39); H, 7.49(7.46); N,
13.69(13.82).
Crystallography. X-ray structural analysis for 1 and 2: Crystals
were selected and mounted on plastic mesh using viscous oil flash-
916
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917

Inorganic Chemistry
Article
cooled to the data collection temperature. Data were collected on a
Bruker-AXS APEX CCD diffractometer with graphite-monochromated
Mo−Kα radiation (λ = 0.71073 Å). Unit cell parameters were
obtained from 60 data frames, 0.3° ω, from three different sections of
the Ewald sphere. The systematic absences in the data and the unit cell
parameters were uniquely consistent to Ccca for 1 to P21/c for 2. The
data sets were treated with SADABS absorption corrections based on
redundant multiscan data.20 The structures were solved using direct
methods and refined with full-matrix, least-squares procedures on F2.
The compound molecules were each located on an inversion point in
2. Two symmetry unique but chemically identical molecules of the
compound in 1 are each located on a 2-fold rotation axis: in one case
the 2-fold axis is parallel to the N−Cu−N axis and in the other case
the 2-fold axis is perpendicular to the N−Cu−N axis bisecting the Cu
atoms on the opposite distal positions of the tetracopper rhombus.
One THF solvent molecule of crystallization per two tetrameric
complexes in 1 was located severely disordered and treated as diffused
contribution.21 One isopropyl group in 1 was located disordered with
a 23/77 refined occupancy ratio. Chemically equivalent bond distances
and angles in the disordered group were restrained to average values
with equal atomic displacement atomic parameter contraints on
equivalent atoms. All non-hydrogen atoms were refined with
anisotropic displacement parameters. All hydrogen atoms were treated
as idealized contributions. Atomic scattering factors are contained in
the SHELXTL 6.12 program library.20
Surface Exposure Experiments. The exposure experiments were
performed in a home-built reactor. The reaction chamber consisted of
a stainless steel ring support covered in 200 stainless steel mesh with a
plug of glass wool to prevent loss of silica powder. The system had one
inlet from a heated bubbler and one inlet for He (purity of 99.999%).
All fittings used in this system were either CF or VCR to ensure a
high-vacuum seal. The system was leak checked using a gas thermal
conductivity/leak detector (Gow-Mac Instrument Co.) and an
overpressure of He. For the exposure experiments, typically about 1
g of high surface area SiO2 powder (EP10X; PQ Corporation; 300 m2/
g S.A.; 1.8 cm3/g P.V.; 24 nm P.S.; 100 μm P.D.) was used. The
powder was pretreated in the reactor at 350 °C for 16 h under vacuum
before exposure to the precursor. The reactor and lines were heated to
temperature and allowed to equilibrate for 1−2 h before introduction
of the precursor. The precursor (typically 0.6−0.8 g) was then
vaporized and transported to the substrate with the system under 10−‑3
Torr vacuum. The substrate was exposed to volatilized precursor for
17 h before the system was cooled to room temperature for handling.
Both precursor and substrate were handled in inert atmosphere.
Annealing experiments were performed in a tube furnace while
under vacuum. Samples were loaded into the furnace under a blanket
of nitrogen gas. Samples were annealed for 2 h.
Electron dispersive X-ray spectroscopy (EDX) was performed on
the modified silica samples as qualitative proof for the presence of
copper. Samples were mounted on an aluminum support using carbon
tape and loaded into a Tescan Vega II SEM equipped with an Oxford
Inca 200 EDX for analysis.
AUTHOR INFORMATION
Corresponding Author
*Tel: +1 613-520-2600, ext. 2244. E-mail: sean_barry@
carleton.ca.
■
Notes
The authors declare no competing financial interest.
REFERENCES
■
(1) 2011 Edition of the International Technology Roadmap for
Semiconductors; http://www.itrs.net/Links/2011ITRS/Home2011.
htm.
(2) (a) Li, Z.; Rahtu, A.; Gordon, R. G.. J. Electrochem. Soc. 2006, 153,
C787. (b) Li, Z.; Barry, S. T.; Gordon, R. G. Inorg. Chem. 2005, 44,
1728.
(3) Krisyuk, V.; Aloui, L.; Prud’homme, N.; Sysoev, S.; Senocq, F.;
́
Samelor, D.; Vahlas, C. Electrochem. Solid-State Lett. 2011, 14, D26.
(4) (a) Ma, Q.; Guo, H.; Gordon, R. G.; Zaera, F. Chem. Mater. 2010,
22, 352. (b) Ma, Q.; Guo, H.; Gordon, R. G.; Zaera, F. Chem. Mater.
2011, 23, 3325.
(5) Turgambaeva, A.; Prud’homme, N.; Krisyuk, V.; Vahlas, C. J.
Nanosci. Nanotechnol. 2011, 11, 8198.
(6) Coyle, J. P.; Monillas, W. H.; Yap, G. P. A.; Barry, S. T. Inorg.
Chem. 2008, 47, 683.
(7) Coyle, J. P.; Johnson, P. A.; DiLabio, G. A.; Barry, S. T.; Muller, J.
Inorg. Chem. 2010, 49, 2844.
(8) Tsuda, T.; Watanabe, K.; Miyata, K.; Yamamoto, H.; Saegusa, T.
Inorg. Chem. 1981, 20, 2728.
(9) Wasslen, Y. A.; Kurek, A.; Johnson, P. A.; Pigeon, T. C.; Monillas,
W. H.; Yap, G. P. A.; Barry, S. T. Dalton Trans. 2010, 39, 9046.
(10) Coyle, J. P.; Kurek, A.; Pallister, P. J.; Sirianni, E. R.; Yap, G. P.
A.; Barry, S. T. Chem. Commun. 2012, 48, 10440.
(11) Willcocks, A. M.; Robinson, T. P.; Roche, C.; Pugh, T.;
Richards, S. P.; Kingsley, A. J.; Lowe, J. P.; Johnson, A. L. Inorg. Chem.
2012, 51, 246.
(12) Cotton, F. A.; Daniels, L. M.; Feng, X.; Maloney, D. J.; Matonic,
J. H.; Murilio, C. A. Inorg. Chim. Acta 1997, 256, 291.
(13) (a) Abdou, H. E.; Mohamed, A. A.; Fackler, J. P.; Jiang, X.;
Bollinger, J. J. Cluster Sci. 2007, 18, 630. (b) Baik, M.-H.; Lee, D.
Chem. Commun. 2005, 1043. (c) Beck, J.; Strahle, J. Angew. Chem., Int.
̈
Ed. Engl. 1986, 25, 95.
Characterization of Surface Species. Solid-state NMR experi-
ments were performed at 4.7 T on a Bruker Avance III console. All
spectra were obtained using a Bruker 7 mm 1H/X/Y probe. 13C (ν0 =
50.3 MHz) cross-polarization magic angle spinning (CP/MAS)
experiments were collected at a spinning rate of 4.5 kHz using a 3.4
μs 90° proton pulse with a contact time of 2 ms where the contact
(14) Kunte, G. V.; Shivashanker, S. A.; Umarji, A. M. Meas. Sci. Tech.
2008, 19, 025704.
(15) Dai, M.; Kwon, J.; Halls, M. D.; Gordon, R. G.; Chabal, Y. J.
Langmuir 2010, 26, 3911.
(16) Haukka, S.; Lakomaa, E. L.; Root, A. J. Phys. Chem. 1993, 97,
5085.
(17) Wasslen, Y. A.; Tois, E.; Haukka, S.; Kreisel, K. A.; Yap, G. P. A.;
Halls, M. D.; Barry, S. T. Inorg. Chem. 2010, 49, 1976.
(18) Bergna, H. E. The Colloid Chemistry of Silica; American
Chemical Society: Washington, DC, 1994.
(19) Maciel, G. E.; Sindorf, D. W. J. Am. Chem. Soc. 1980, 102, 7606.
(20) Sheldrick, G. M. Acta Crystallogr. 2008, A64, 112.
(21) Spek, A. L. J. Appl. Crystallogr. 2003, 36, 7.
1
pulse was ramped on the H channel. A relaxation delay of 2 s was
sufficient to prevent saturation and typically total acquisition times
were 16−30 h. Glycine was used as an external secondary reference for
the 13C chemical shift scale. Spectra were treated with 40 Hz line
broadening during processing. 29Si (ν0 = 39.7 MHz) CP/MAS
experiments were collected at a spinning rate of 4.5 kHz using a 3.85
μs 90° proton pulse with a contact time of 10 ms where the 1H
channel contact pulse was ramped. The relaxation delay was 2 s and
typically required 2−8 h acquisition times. TMSS was used as an
external reference for the 29Si chemical shift scale. Spectra were treated
with 30 Hz line broadening during processing. All spectra were
obtained with high power proton decoupling during acquisition.
Samples were prepared for High Resolution NMR (HR-NMR) by
adding a small amount, typically 40−50 mg, of modified silica powder
to 2 mL of D2O (Sigma), agitating, and allowed to sit for 15 min. The
D2O solution was then decanted and studied via HR-NMR. d4-TSP
was used as an internal reference.
917
dx.doi.org/10.1021/ic3021035 | Inorg. Chem. 2013, 52, 910−917