Synthesis of Al(N3)3 and the deposition of AlN thin films

Article abstract of DOI:10.1021/jp9625873

Al(N₃)₃ is produced by the stoichiometric reaction between Al(CH₃)₃ and excess HN₃ at room temperature. The reaction is thought to proceed by the addition of HN₃ to Al(CH₃)₃ followed by elimination of CH₄, repeated three times to produce the fully azidified Al(N₃)₃. The product Al(N₃)₃ is nonvolatile and condenses as a film on the walls of the reaction vessel. The reaction products were observed in the gas phase and in low-temperature argon matrices by FTIR spectroscopy. Ab initio methods were used to compute the geometry and frequencies of Al(N₃)₃, and the results are in good agreement with experimental data. The films produced upon condensation of Al(N₃)₃ contain the Al-N₂ complex and AlN as well as the azide. Heating the films to 400 K removes the azide and the Al-N₂, leaving AlN. This method may be useful as a low-temperature route to the synthesis of AlN thin films.

Full text of DOI:10.1021/jp9625873

1602
J. Phys. Chem. B 1997, 101, 1602-1608
Synthesis of Al(N3)3 and the Deposition of AlN Thin Films
C. J. Linnen, D. E. Macks, and R. D. Coombe*
Department of Chemistry and Biochemistry, UniVersity of DenVer, DenVer, Colorado 80208
ReceiVed: August 21, 1996; In Final Form: NoVember 19, 1996X
Al(N3)3 is produced by the stoichiometric reaction between Al(CH3)3 and excess HN3 at room temperature.
The reaction is thought to proceed by the addition of HN3 to Al(CH3)3 followed by elimination of CH4,
repeated three times to produce the fully azidified Al(N3)3. The product Al(N3)3 is nonvolatile and condenses
as a film on the walls of the reaction vessel. The reaction products were observed in the gas phase and in
low-temperature argon matrices by FTIR spectroscopy. Ab initio methods were used to compute the geometry
and frequencies of Al(N3)3 , and the results are in good agreement with experimental data. The films produced
upon condensation of Al(N3)3 contain the Al-N2 complex and AlN as well as the azide. Heating the films
to 400 K removes the azide and the Al-N2, leaving AlN. This method may be useful as a low-temperature
route to the synthesis of AlN thin films.
I. Introduction
subsequent thermal or photolytic dissociation to produce BN
films. The synthesis was based on the stoichiometric reaction
of BCl3 with HN3. This reaction is thought to occur by initial
formation of a BCl3:HN3 adduct, which can spontaneously
eliminate HCl because of the relative weakness of the N-H
bond in HN3. In this paper, we present the results of a study
of an analogous gas phase reaction between Al(CH3)3 and HN3,
which produces Al(N3)3 by the elimination of CH4. This
compound and its adducts were originally prepared in 1954 by
Wiberg and Michaud,16 who reacted gas phase HN3 with a
solution of aluminum hydride in ether at 77 K and then warmed
the solution to room temperature. This method produced an
extremely energetic association of loosely bound azide com-
plexes which exhibited a marked tendency toward explosion.
Perhaps because of this, compounds of fully azidified group
III elements have received very little attention until the recent
work in our laboratory. We present the synthesis of Al(N3)3,
experimental and computational studies of its structure and
vibrational characteristics in the gas phase and in low-temper-
ature matrices, and a preliminary evaluation of the utility of
this species as a precursor for AlN thin films.
Aluminum nitride thin films have become increasingly
important to the field of optoelectronics. Because of the
piezoelectric properties, wide bandgap, and high thermal and
chemical stability of this material, AlN films have been
suggested for applications ranging from dielectric insulators1
to surface acoustic wave devices2 to second harmonic genera-
tion.3 Methods for synthesis of these films have typically
involved inconvenient reagents in extreme conditions, however,
with precursor dissociation temperatures on the order of 900
K. Because of the nature of VLSI and ULSI microcircuit
technology, high temperatures can be a considerable detriment
in device fabrication. As a result, an established trend in AlN
thin film synthesis has been toward lower deposition temper-
atures, achieved by many different techniques including ion
beam sputtering,4-6 molecular beam epitaxy,7,8 laser ablation,9
and a host of chemical vapor deposition (CVD) methods.
MOCVD methods for the deposition of AlN were introduced
in 1971 by Manesevit and co-workers,10 who generated AlN
by the reaction of Al(CH3)3 with NH3 at high temperature. This
reaction produces AlN and CH4 at 1500 K. It is thought to
proceed via the formation of an Al(CH3)3:NH3 adduct, with the
high temperature needed for breakage of the strong N-H bonds.
Lower deposition temperatures (1000 K) were achieved by
Gaskill and co-workers,11 by replacing the NH3 reagent with
the considerably more reactive species hydrazine. Interrante
and co-workers12 investigated the use of organometallic amides
as precursors to AlN films. The approach was to synthesize
and dissociate species similar to the intermediates in the Al-
(CH3)3 + NH3 reaction, the most volatile of which was found
to be the [(CH3)2Al(NH2)]3 trimer. This species had the
additional advantage of fixing the Al:N stoichiometry at 1:1.
In 1989, Boyd and co-workers13 reported the deposition of AlN
by the dissociation of metal-organic aluminum azides, species
which are still more energetic. Using this approach, deposition
temperatures as low as 700 K were achieved, but the syntheses
of these compounds were complex and the films incorporated
significant amounts of CH4 and CN impurities. Very recently,
deposition temperatures as low as 400 K were obtained through
the use of organometallic amide complexes like Al2[N(CH3)2]6.14
Recent research in our laboratory15 has demonstrated a simple
gas phase synthesis of fully azidified boron, B(N3)3, and its
II. Experimental Methods
Gaseous HN3 was synthesized17 by reacting NaN3 with an
excess of stearic acid at 373 K. The middle fraction of the
HN3 generated was collected in a pyrex bulb and diluted 10-
fold with He. The purity of the HN3 in the mixture was
examined by using FTIR spectroscopy, and its absolute con-
centration was determined from the well-known UV absorption
at 268 nm.18 Al(CH3)3 was obtained from Alfa Products and
was purified by successive freeze-thaw cycles at 273 K to
remove the methane impurity.
Aluminum azides were generated by the spontaneous room
temperature reaction between gaseous HN3 and Al(CH3)3 vapor.
The gases were handled in a stainless steel vacuum system, and
pressures were measured with two MKS Baratron pressure
transducers. It was necessary to passivate the system with Al-
(CH3)3 for at least 30 min prior to the experiments to obtain
stable pressures of this compound in the system. The passi-
vation process clearly produces CH4, features of which are
observed to grow into the IR spectrum of the gas. This result
suggests that methane is released as the aluminum binds to the
metal surface, with methyl groups from the original Al(CH3)3
X Abstract published in AdVance ACS Abstracts, February 15, 1997.
S1089-5647(96)02587-4 CCC: $14.00 © 1997 American Chemical Society

Synthesis of Al(N3)3
J. Phys. Chem. B, Vol. 101, No. 9, 1997 1603
combining with H atoms displaced from the surface. Experi-
ments showed that the passivation layer would react with HN3
just as would gaseous Al(CH3)3. Consequently, it was difficult
to distinguish aluminum azides created by wall reaction from
those made in the vapor phase.
The reagents were mixed in an 11 cm cell fitted with either
KCl or fused silica windows for FTIR or UV analysis,
respectively. FTIR spectra were recorded with either a Nicolet
5DXC spectrometer or a Mattson Polaris spectrometer. UV-
vis spectra were recorded with a Milton-Roy 3000 diode array
spectrometer. Grazing incidence X-ray diffraction spectra of
solid reaction products were recorded with a Siemens diffrac-
tometer.
Low-temperature matrix isolation experiments were per-
formed by using an RMC Cryosystems LTS-22 closed cycle
refrigerator. For these experiments, flowing streams of the
reagents mixed with Ar were passed through a 5 L reaction
vessel prior to being deposited onto the cold window of the
apparatus. The flow rates of the reagents were measured with
MKS mass flow meters.
III. Results and Discussion
1. Reaction of Al(CH3)3 with HN3. By analogy with the
reaction of BCl3 with excess HN3, Al(CH3)3 is expected to react
stoichiometrically with HN3 to generate Al(N3)3 and methane:
Al(CH3)3 + 3HN3 f Al(N3)3 + 3CH4
(1)
There is a significant difference between this system and the
reaction of BCl3 with HN3, however, in that BCl3 exists as a
monomer at room temperature whereas Al(CH3)3 is primarily
dimeric below 573 K.19 The formation of an Al(CH3)3:HN3
adduct would require the breaking of the bridging methyl-
aluminum bond in the dimer. Because of the additional energy
needed, reaction 1 was expected to be much slower than the
BCl3 analog, if in fact it would occur at all at room temperature.
The initial experiments showed clearly that reaction was
occurring, however. Figure 1a shows the IR spectrum of vapor
phase Al(CH3)3 present in the reaction cell at 1.0 Torr, after
thorough passivation. The spectra indicated that the compound
was stable in the cell at this pressure for extended periods of
time (hours), in the absence of air or atmospheric H2O. Figure
1b shows the IR spectrum of 3.0 Torr of HN3 (at 1:10 in He)
in the cell, exhibiting the characteristic absorptions at 2150 and
1164 cm-1 (stretches in the N3 group) and 3324 cm-1 (the N-H
stretch). Figure 1c shows the IR spectrum obtained from a gas
mixture in which 1.0 Torr of Al(CH3)3 was allowed to react
with 3.0 Torr of HN3 (1:10 in He) for a period of 8 min. The
spectrum clearly shows the loss of the features corresponding
to Al(CH3)3 and HN3, and the evolution of CH4 is evident from
Figure 1. Gas phase mid-IR spectra of (a) 1 Torr of pure Al(CH3)3
after passivation of the cell for ∼30 min, (b) 30 Torr of a 10% mixture
of HN3 in He, and (c) mixture of Al(CH3)3 with 3 equiv of HN3 after
allowing 8 min for reaction. A 10 cm long stainless steel cell fitted
with KCl windows was used for all IR spectra.
3017 and 1306 cm-1, and the appearance of the “new” N3
stretching features at 2156 cm-1
. This feature is a single
absorption peak, clearly different from the “doublet” appearance
of the HN3 feature in this region. Though significantly
overlapped by methane and residual Al(CH3)3 features, new
absorptions at 1250 and 741 cm-1 grow in and appear to be
correlated with the 2156 cm-1 peak. At reaction stoichiometries
above 2.5:1, as the reaction approaches completion (3:1), these
new features at 2156, 1236, and 741 cm-1 disappear almost
completely. No other peaks appear to be growing in. Assuming
that the features are attributable to aluminum azides (as
discussed below), the disappearance might be attributed to the
formation of Al(N3)3 as the stoichiometry approaches 3:1, as
this species is expected to be nonvolatile at room temperature16
and should condense out on the vessel walls. This inference
was supported by the observation of the formation of a deposit
on the walls for the 3:1 stoichiometry.
its characteristic absorptions at 3017 and 1306 cm-1 20
The
.
spectrum of the reaction mixture still shows a feature near 2160
cm-1, as in the HN3 spectrum in Figure 1b, but the total absence
of the HN3 features at 1164 and 3340 cm-1 indicates that this
is not attributable to HN3. Since virtually all azides have an
absorption in the 2050-2150 cm-1 region (corresponding to
asymmetric stretching in the N3 group), this result strongly
suggests the formation of a new azide species.20,21
The stoichiometry of reaction 1 was determined by analysis
of IR spectra recorded for varying Al(CH3)3:HN3 ratios, at a
constant initial Al(CH3)3 pressure of 1.0 Torr. Spectra recorded
for a variety of initial HN3 pressures are shown in Figure 2.
Qualitative scrutiny of the spectra clearly reveals the occurrence
of the reaction. Perhaps most striking is the disappearance of
the Al(CH3)3 spectrum, the growth of the methane features near
A quantitative measure of the stoichiometry was obtained by
measurement of the relative absorbance intensities of the reactant
and product features as the mixing ratio increased. These data
are shown in Figure 3 as a plot of absorbance vs the initial
HN3:Al(CH3)3 ratio. In the data shown, the HN3 density was
tracked by the N3 stretching vibration of this molecule at 1164
cm-1. The density of Al(CH3)3 was monitored by observing
the features at 700 and 775 cm-1. The data show that the Al-
(CH3)3 density decreases with added HN3, falling to zero at a
mixing ratio just below 3:1. Further, residual unreacted HN3
appears in the spectrum at mixing ratios above 3:1. These data

1604 J. Phys. Chem. B, Vol. 101, No. 9, 1997
Linnen et al.
Figure 2. IR spectra obtained by varying the stoichiometric mixtures of HN3 (1:10 in He) with 1 Torr of Al(CH3)3.
bridging methyl groups) or that a 6:1 stoichiometry would result
if HN3 reacts only with monomers (as expected) and the gas
phase dimer T monomer equilibrium slowly responded to
replenish the monomers. Clearly, neither of these possibilities
is in accord with the data. It seems likely, particularly in view
of the slow rate of the reaction (see below), that reaction at the
walls is important. For example, HN3 might react with the
dimer at the walls, releasing a gaseous AlN3(CH3)2 moiety
(capable of reacting with two more HN3 molecules), leaving
the residual Al(CH3)3 monomer strongly bound to the metal
wall. Such wall-bound monomers would react only very slowly,
giving rise to the slow evolution of CH4 observed for HN3
densities in excess of the 3:1 stoichiometry.
In the course of recording the spectra shown in Figure 2, it
was indeed apparent that the reaction between Al(CH3)3 and
HN3 is very much slower than the analogous BCl3 + HN3
reaction studied earlier. In the latter case, the reaction appeared
to be immediate; i.e., the reactants were consumed and products
fully formed in the time necessary to record the spectra. For
the reaction with Al(CH3)3, complete reaction required on the
order of 10 min after mixing, for normal conditions (1.0 Torr
of Al(CH3)3 at 298 K). In an effort to quantify these observa-
tions, spectra were recorded at fixed times after mixing, and
the data were converted to time profiles of the reactant and
product species. Figure 4 shows such time profiles recorded
for HN3 and CH4 at a 3:1 reactant mixing ratio. The symmetry
between the loss of HN3 and the growth of CH4 is apparent.
Note that the CH4 density continues its slow growth at long
times, as described above. An effective rate constant can be
obtained from fitting data such as these to rate expressions which
assume second-order kinetics. Good fits are obtained, and these
indicate an effective reaction rate constant on the order of 10-20
Figure 3. Plot of absorbance versus HN3:Al(CH3)3 ratio. Varying
pressures of a 10% mixture of HN3 in He were added to 1 Torr of
Al(CH3)3 to obtain the reagent ratios shown. (+) HN3 peak at 1164
cm-1, (*) Al(CH3)3 peak at 775 cm-1, and (0) Al(CH3)3 peak at 700
cm-1
.
clearly indicate the 3:1 stoichiometry and suggest that reaction
1 proceeds to completion in our system.
Monitoring the stoichiometry via IR features of the products
of the reaction was problematic. As noted, the aluminum azide
features clearly grow into the spectrum with increasing HN3
density but abruptly disappear near the 3:1 stoichiometric “end
point”, likely because of condensation of the fully azidified Al-
(N3)3. The methane product was easily tracked and clearly grew
in as expected for mixing ratios below 3:1, but though the
absorbance flattened somewhat at higher ratios, its growth
continued. This may be attributable to HN3 reaction with Al-
(CH3)3 in the passivation layer on the vessel walls, with the
product aluminum azide remaining on the wall but the methane
released to the gas phase. Hence the methane absorbance did
not show a sharp end point at the 3:1 gas phase mixing ratio.
Nonetheless, the end point is clear from the data in Figure 3. It
is interesting to consider the fact that complete reaction with
the Al(CH3)3 dimers initially present would give rise to a 4:1
stoichiometry (assuming that the dimers retained the two
cm3 s-1
. The very small magnitude of this effective rate
constant certainly underscores the notion that the walls are
important in the reaction mechanism. Calculations of the
energetics of the BCl3 + HN3 reaction suggested that the first
step in the reaction sequence (formation of Cl2BN3) was likely
to be rate limiting. While this may also be true for the reaction
with Al(CH3)3, the very slow rate of this process certainly

Synthesis of Al(N3)3
J. Phys. Chem. B, Vol. 101, No. 9, 1997 1605
TABLE 2: Calculated Frequencies (cm-1) for Al(N3)3
ν(AlNNN)
(rel int)
δ(N3)
(rel int)
ν
sym(N3)
ν
asym (N3)
level of theory
(rel int)
(rel int)
RMP/6-31 G(d)
568 (21) 737 (251) 1366 (210) 2192 (1011)
obsd: gas phase
obsd: matrix studies
741
730
1236
1360
2156
2160
bond (74% covalent); fully ionic azides are generally linear.
The NR-Nâ and Nâ-Nγ bond lengths as well as the NR-Nâ-
Nγ angle are very similar to those calculated for B(N3)3. The
value of 1.79 Å for the Al-NR bond length is considerably
longer than that of the B-NR bond (1.44 Å) as expected, but is
somewhat shorter than Al-N bonds in other compounds
(approximately 1.9 Å). This shorter Al-NR bond length may
reflect back-donation of the azide π electron density into low-
lying unoccupied d-type orbitals on the Al atom.
Frequency calculations were performed for the optimized
geometry at the MP2/6-31G(d) level of theory, and the results
are shown in Table 2. Of the 24 normal modes of Al(N3)3,
only four were found to have appreciable IR intensity in the
mid-IR region. As shown in the table, the calculated results
are in reasonable agreement with experiment. The most intense
peak is predicted to lie at 2192 cm-1, only 36 cm-1 to the blue
of the experimental result. The calculation predicts less intense
lower frequency features at 1366 and 737 cm-1. These too agree
with experimental frequencies, although these features are
obscured by Al(CH3)3 and CH4 absorptions in our experimental
spectra. We note also that perfect agreement is not expected
since the experimental spectra may well contain features
corresponding to partially azidified molecules (as Al(N3)3 is
expected to be condensed on the vessel walls). Further
clarification was sought in the matrix isolation experiments
described below.
Using the minimum energy calculated for the optimized Al-
(N3)3 geometry, and calculations of the energies of Al(2P) and
N2(1Σg+) done at the same level of theory, the heat of formation
of Al(N3)3 (corrected for zero-point energies) was calculated
to be ∆Hf,298 ) 84.7 kcal/mol. While this value is substantially
larger than that calculated for B(N3)3 (21.2 kcal/mol),15 the
stability of methane relative to HCl leads to an overall
exothermicity for reaction 1 of -164 kcal/mol, quite similar to
the value calculated for the analogous formation of B(N3)3 from
BCl3 and HN3. Using average bond energies, the initial Al-
(CH3)3 + HN3 reaction is calculated to be exothermic by 29
kcal/mol, adding further support to the notion that the rate is
controlled by the dimer/monomer equilibrium.
3. Low-Temperature Matrix Isolation Studies. Matrix
isolation experiments were performed in an effort to better
resolve features in the IR spectra corresponding to reactants and
different possible products. In particular, it was hoped that the
matrix experiments would allow better resolution of the
aluminum azide features from HN3 features in the 2150 cm-1
region and resolution of aluminum azide from methane in the
low-frequency region.
Following passivation of the system, Al(CH3)3 and argon (at
a 100-fold excess of argon) were flowed through the system
and onto the cold window until the characteristic IR features
of the Al(CH3)3 were clearly visible and the growth of methane
features (a passivation product) reached a minimum. A flow
of HN3 was then added to achieve the 3:1 stoichiometric mixing
ratio. For typical conditions, the residence time in the reaction
bulb (i.e., prior to deposition of the mixture) was on the order
of 10 min, such that complete reaction was assured. The system
was very responsive to the HN3 flow rate, and the reaction
stoichiometry was easily manipulated. For a 3:1 flow ratio, no
excess HN3 was evident in the IR spectrum of the matrix.
Figure 4. Time profile showing (a) the decay of HN3 and (b) the
associated CH4 growth.
TABLE 1: Calculated Geometry Parameters for Al(N3)3
parameter
RHF/6-31G(d)
RMP2/6-31 G(d)
Al-NR (Å)
1.768
1.215
1.099
90.0
134.1
90.1
90.1
86.7
88.2
120
1.787
1.229
1.175
90.0
133.4
90.1
90.3
84.4
87.4
120
NR-Nâ (Å)
Nâ-Nγ (Å)
a
X1 -Al-NR (deg)
Al-NR-Nâ (deg)
a
X1 -Al-NR-Nâ (deg)
b
X2 -Nâ-NR (deg)
Al-NR-Nâ-X2b (deg)
b
X3 -Nâ-NR (deg)
NR-Al-NR (deg)
a X1: dummy atom placed on Al. b X2, X3: dummy atoms placed
on Nâ.
implies that the rate is somehow limited by the dimer T
monomer equilibrium.22 The system may be further complicated
by the documented tendency of partially azidified intermediates
(e.g., (CH3)2AlN3) to trimerize.13 It seems unlikely, however,
that trimerization would compete with reaction given the greater
densities of HN3 always present. Indeed, the spectra shown in
Figures 1 and 2 indicate the complete removal of methyl-
aluminum species.
2. Calculations of the Structure and Frequencies of Al-
(N3)3. Ab initio calculations of the structure and vibrational
frequencies of monomeric Al(N3)3 were performed in an effort
to more clearly characterize the IR spectra described above. The
calculations were performed by using the GAUSSIAN 92 family
of programs23 at the HF/6-31G(d) and MP2/6-31G(d) levels of
theory. In each case, partial geometry optimizations were
performed which assumed the equivalent geometry within each
of the N3 groups. For similar calculations done on B(N3)3, this
assumption produced a geometry for which calculated frequen-
cies agreed very well with experimental IR spectra. The results
of the geometry optimizations for Al(N3)3 are shown in Table
1. The RHF and MP2 calculations show considerable consis-
tency. The value of 133° calculated at the MP2 level for the
Al-NR-Nâ bond angle is substantially larger than the analogous
angle calculated for B(N3)3 (120°). This may indicate greater
inclusion of s atomic orbital character into the bonding orbitals
on NR. In another sense, it may reflect the lesser formal
covalency of the Al-N bond (53%) with respect to the B-N

1606 J. Phys. Chem. B, Vol. 101, No. 9, 1997
Linnen et al.
Figure 6. Difference spectrum obtained by subtraction of a single-
beam scan taken of clean KCl windows from a single-beam scan of
the same windows with film residue created from the reaction of HN3
with Al(CH3)3 at 3:1 stoichiometry.
Figure 5. IR spectra of (a) Al(CH3)3 (1:100 in Ar), (b) HN3 (1:100 in
Ar), and (c) mixture of Al(CH3)3 with HN3 at a 3:1 stoichiometric flow
rate (see section III.3) isolated in a ∼10 K argon matrix.
Figure 7. IR spectra of (a) 1900-2300 cm-1 region of the spectrum
shown in Figure 6, (b) 1900-2300 cm-1 region of the film represented
by Figure 6 after exposure to heat (∼100 °C) for 20 min, and (c)
difference spectrum obtained by subtraction of (a) from (b).
Raising the HN3 flow rate by only 3% clearly introduced HN3
features into the spectrum.
Parts a and b of Figure 5 show spectra of matrices containing
the reagents Al(CH3)3 and HN3, respectively. Both are in very
good agreement with similar spectra presented in the literature.24
The extensive passivation requirement for the Al(CH3)3 matrix
is reflected in the presence of intense CH4 features at 1305 and
3024 cm-1. The HN3 spectrum (Figure 5b) exhibits features
corresponding to trapped CO2, a byproduct of the HN3 synthesis.
Figure 5c shows the spectrum of a matrix made by deposition
of the 3:1 stoichiometric mixture as described above. While it
shows evidence of unreacted Al(CH3)3 and HN3 (artifacts of
the deposition method: the Al(CH3)3 was deposited during
passivation, and the HN3 appeared only when the mixing ratio
was adjusted to be slightly in excess of 3:1), it also clearly shows
features corresponding to the products CH4 and aluminum azide,
the latter marked with an asterisk. The high-frequency N3
stretch in the aluminum azide (2160 cm-1) is clearly separated
from the analogous feature in HN3 (2137 cm-1). Further, the
lower frequency features at 1361 and 730 cm-1 are easily
resolved.
These data lend credence to the assignments from the gas
phase spectra described above. Further, from the sensitivity to
the stoichiometry and the fact that the matrix environment does
not distinguish between volatile and nonvolatile products, the
trapped species is expected to be the fully azidified Al(N3)3.
Indeed, the observed frequencies are in better agreement with
those calculated for this molecule (as above) than are the gas
phase results.
cell, a persistent absorbance was apparent in the background
spectrum (i.e., in the spectrum of the evacuated cell). Figure 6
shows a spectrum of this absorbance, obtained by subtraction
of the spectrum of a clean KCl window from the spectrum of
the cell after contact with several reaction mixtures. A broad
absorbance in the 500-1000 cm-1 region, in good agreement
with well-known AlN phonon modes,25 is clearly evident. In
addition, two sharper features are seen at 2043 and 2111 cm-1
.
These features might be attributable to N3 stretching vibrations
of Al(N3)3 in the film (as noted above, this should be a
nonvolatile reaction product) or perhaps to the Al-N2 complex
in the film formed by dissociation of the azide in the generation
of the AlN. Mazur and Cleary5 have observed such complexes
in the IR spectra of AlN films deposited by ion beam sputtering.
To test these hypotheses, the evacuated cell was gently heated
with the thought that heating would lead to further decomposi-
tion of the azide and generation of more N2. The cell was heated
to a maximum temperature of 373 K for 30 min. Figure 7 shows
the resulting IR spectra in the 1900-2300 cm-1 range. The
difference spectrum (obtained by subtraction of the spectra
before and after heating) clearly shows that heating leads to
diminished intensity in the 2043 cm-1 band and increased
intensity in the 2111 cm-1 band. On the basis of these data,
we tentatively assign the 2043 cm-1 band to the N3 stretching
vibration in Al(N3)3 and the 2111 cm-1 band to the Al-N2
complex in the film.
4. Deposition and Dissociation of Al(N3)3. After several
successive gas mixtures were allowed to react within the IR

Synthesis of Al(N3)3
J. Phys. Chem. B, Vol. 101, No. 9, 1997 1607
boron analog. We note again that this is also reflected in the
much larger Al-NR-Nâ bond angle as given by calculation.
In recording spectra such as that shown in Figure 9, great care
was taken to ensure that background absorptions associated with
the spectrometer or the windows were adequately subtracted.
The absorption onset is in very good agreement with the spectra
of AlN films reported in the literature and corresponds well
with the known bandgap of this material at 6.2 eV.8
A number of experiments were performed in an effort to
obtain X-ray diffraction spectra of the AlN films. Powder
samples were prepared by scraping the film from the interior
surfaces of the reaction cell. These experiments were incon-
clusive, however, as the X-ray spectra obtained showed small
features attributable to AlN but were dominated by features
attributable to alumina (Al2O3) and NH3. It is well-known that
these products arise from reaction of fresh AlN with atmospheric
water:26
2AlN + 3H2O f Al2O3 + 2NH3
(2)
The films were exposed to the atmosphere in the course of
measurement of the diffraction spectra. Indeed, upon exposure
to moist air, the IR spectra of the films (as in Figures 7 and 8)
changed to include a broad high-frequency feature just above
3000 cm-1, characteristic of ammonia in the film, and features
near 950 cm-1 attributable to Al-O stretching vibrations.27
Figure 8. (a) Spectrum of film represented in Figure 6 after prolonged
(∼1 h) exposure to heat (∼100 °C). (b) Spectrum of AlN thin film
created by laser ablation. Reprinted with permission from ref 9.
Copyright 1992 American Institute of Physics.
IV. Summary and Conclusions
The data presented have shown that aluminum azides may
be prepared by the spontaneous room temperature reaction
between Al(CH3)3 and HN3. The reaction is quite slow at 298
K, likely because Al(CH3)3 exists primarily as a dimer at this
temperature. From studies of the stoichiometry of the reaction,
the infrared spectra of the reaction products, and comparison
of these data with computed spectra, it is apparent that in the
presence of sufficient HN3 the reaction proceeds to the fully
azidified Al(N3)3. This molecule is not volatile at 298 K,
however, and condenses as a film on the walls of the reaction
cell.
In the course of its preparation and condensation at room
temperature, some of the Al(N3)3 decomposes to form Al-N2
complexes and AlN. These species are observed in IR spectra
of the films produced by the reaction, along with undissociated
Al(N3)3. Gentle heating removes the azide and the Al-N2
complex from the film, leaving what appears to be amorphous
AlN. These results suggest that good quality AlN films might
be generated by continuous flow of the stoichiometric 3:1 Al-
(CH3)3:HN3 mixture over a substrate heated to perhaps 400 K.
The gentle heating would serve two purposes: to accelerate the
reaction and to drive off Al(N3)3 and free N2 from the film. We
are currently testing these thoughts by exploring the temperature
dependence of the reaction.
Figure 9. UV absorption spectrum of AlN film prepared by thermal
dissociation of Al(N3)3 created by mixture of 1 Torr of Al(CH3)3 with
30 Torr of 1:10 HN3 in He.
On further heating of the evacuated cell, to a maximum
temperature of 400 K for 60 min, both the 2143 and 2111 cm-1
features disappeared completely. The broad AlN absorbance
below 1000 cm-1 remained, however, as shown in Figure 8a.
This spectrum is nearly identical to that reported in the literature
for amorphous AlN produced by laser ablation, as shown in
Figure 8b.9
Acknowledgment. This work was supported by the U.S.
Air Force Office of Scientific Research under Grant F49620-
93-1-0631 and by the National Science Foundation under Grant
CHE-9300383. Special thanks go to Dr. Julanna Gilbert at the
University of Denver for her assistance in low-temperature
matrix isolation experiments.
Figure 9 shows the UV absorption spectrum of an AlN film
produced by reaction of 3:1 gas mixtures in a cell with fused
silica windows. The cell was not heated and should therefore
contain both AlN and Al(N3)3. The spectrum shows little or
no absorbance at wavelengths above 230 nm, but the absorbance
rises steeply below 210 nm. Gas phase B(N3)3 exhibits a strong
absorbance,15 believed to be the n f π* transition in the azide
moiety, at 230 nm. The lack of such an absorbance feature in
the spectrum of Figure 9 reflects the considerably different
bonding characteristics involved in Al(N3)3 with respect to the
References and Notes
(1) Baier, H.-U.; Monch, W. J. Appl. Phys. 1990, 68, 586.
(2) Pearce, L. G., Gunshor, R. L., Pierret, R. F. Appl. Phys. Lett. 1981,
39, 878.
(3) Lundquist, P. M.; Lin, W. P.; Xu, Z. Y.; Wong, G. K.; Rippert, E.
D.; Helfrich J. Al; Ketterson, J. B. Appl. Phys. Lett. 1994, 65, 1085.

1608 J. Phys. Chem. B, Vol. 101, No. 9, 1997
Linnen et al.
(4) Kubota, K.; Kobayashi, Y.; Fujimoto, K. J. Appl. Phys. 1989, 66,
2984.
(16) Wiberg, E.; Michaud, H. Z. Naturforsch., Teil B 1954, 9, 495.
(17) Schlie, L. A.; Wright, M. W. J. Chem. Phys. 1990, 92, 394.
(18) McDonald, J. R.; Rabalais, J. W.; McGlynn, S. P. J. Chem. Phys.
1970, 85, 290.
(19) Kvisle, S.; Rytter, E. J. Mol. Struct. 1984, 117, 51.
(20) Nakamoto, K. Infrared and Raman Spectra of Inorganic and
Coordination Compounds; John Wiley & Sons: New York, 1986.
(21) Patai, S. The Chemistry of the Azido Group; John Wiley & Sons:
New York, 1971; Chapter 1.
(22) Smith, M. B. J. Organomet. Chem. 1971, 46, 31.
(23) Frisch, J. M.; Trucks, G. W.; Head-Gordon, M.; Gill, P. M.; Wong,
M. W.; Foresman, J. B.; Johnson, B. G.; Schlegel, H. B.; Robb, M. A.;
Replogle, E. S.; Gomperts, R.; Andres, J. L.; Rachavachari, K.; Binkley, J.
S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; Defrees, D. J.; Baker, J.; Stewart,
J. J.; Pople, J. A. GAUSSIAN 92; Gaussian, Inc.: Pittsburgh, PA, 1992.
(24) Kvisle, S.; Rytter, E. Spectrochim. Acta 1984, 40A, 939. Kvisle,
S.; Rytter, , E. J. Mol. Struct. 1984, 117, 51.
(5) Mazur, U.; Cleary, A. C. J. Phys. Chem. 1990, 94, 189. Wang, X.
D.; Hipps, K. W., Mazur, U. J. Chem. Phys. 1992, 96, 8485.
(6) Zarwasch, R.; Rille, E.; Pulker, H. K. J. Appl. Phys. 1992, 71, 5275.
(7) Yoshida, S.; Misawa, S.; Fujii, Y.; Takada, S.; Hayakawa, H.;
Fonda, S.; Itoh, A. J. Vac. Sci. Technol. 1970, 16, 990.
(8) Yim, W. M.; Strofko, E. J.; Zanzucchi, P. J.; Pankove, J. I.,
Ettenberg, M.; Gilbert, S. L. J. Appl. Phys. 1973, 44, 292.
(9) Seki, K.; Xu, X.; Okabe, H.; Frye, J.; Halpern, J. Appl. Phys. Lett.
1992, 60, 2234.
(10) Manesevit, H. M.; Erdmann, F. M.; Simpson, W. I. J. Electrochem.
Soc. 1971, 118, 1864.
(11) Gaskill, D. K.; Bottka, N.; Lin, M. C. J. Cryst. Growth 1986, 77,
418.
(12) Interrante, L. V.; Carpenter, Jr., L.; Whitmarsh, C.; Lee, W.;
Garbaukas, M.; Alsck, G. A. Mater. Res. Soc. Symp. Proc. 1986, 77, 418.
(13) Boyd, D. C.; Haasch, R. T.; Mantell, D. R.; Schulze, R. K.; Evans,
J. F.; Gladfelter, W. L. Chem. Mater. 1989, 1, 119.
(14) Hoffman, D. M.; Rangarajan, S. P.; Satish, D. A.; Economou, D.
J.; Liu, J.-R.; Zheng, Z.; Chu, W.-K. J. Vac. Sci. Technol. A 1996, 14, 306.
(15) Mulinax, R. L.; Okin, G. S.; Coombe, R. D. J. Phys. Chem. 1995,
99, 6294.
(25) Collins, A. T.; Lightowlers, E. C.; Dean, P. J. Phys. ReV. 1967,
158, 3, 833.
(26) Mazur, U., Langmuir 1990, 6, 1331.
(27) Sipachev, V. A.; Grigor’ev, A. I. Russ. J. Inorg. Chem. 1970, 15,
905.