Stepwise conversion of a single source precursor into crystalline AlN by transamination reaction

Article abstract of DOI:10.1016/j.jssc.2007.12.026

Ammonolysis of the monomeric, base-stabilized trisaminoalane Me₃N-Al[N(H)Dipp)]₃ (Dipp=2,6-ⁱPr₂-C₆H₃) yielded Al-N oligomers, which were characterized in detail by solid state NMR spectroscopy (¹H, ¹³C, ¹⁵N, ²⁷Al) and TGA/DTA. Pyrolysis of as-prepared oligomers at different temperatures in an argon steam yielded carbon-containing black solids, whereas pyrolysis under a steady flow of NH₃ produced pure aluminum nitride (AlN). The role of the pyrolysis temperature and the influence of NH₃ on the formation of crystalline materials were investigated. As-prepared AlN was characterized by solid state NMR spectroscopy (¹⁵N, ²⁷Al), X-ray diffraction (XRD), transmission electron microscopy (TEM) and electron energy loss spectroscopy (EELS). Theoretical calculations were performed in order to identify potential reaction intermediates.

Full text of DOI:10.1016/j.jssc.2007.12.026

ARTICLE IN PRESS
Journal of Solid State Chemistry 181 (2008) 530–538
www.elsevier.com/locate/jssc
Stepwise conversion of a single source precursor into crystalline AlN
by transamination reaction
Ã
Stephan Schulza, , Tillmann Bauerb, Wilfried Hoffbauerb, Jorn Schmedt auf der Gunnec,
¨
¨
Markus Doerrd, Christel M. Mariand, Wilfried Assenmachere
aInstitute of Inorganic Chemistry, University of Duisburg-Essen, Universita¨tsstrasse 5-7, 45141 Essen, Germany
bInstitute of Inorganic Chemistry, University of Bonn, Gerhard-Domagk-Str. 1, 53121 Bonn, Germany
cDepartment of Chemistry and Biochemistry, University of Munich (LMU), Butenandtstr. 5-13, D2.076, 81377 Mu¨nchen, Germany
dInstitute of Theoretical and Computational Chemistry, University of Du¨sseldorf, Universita¨tsstr. 1, 40225 Du¨sseldorf, Germany
eInstitute of Inorganic Chemistry, University of Bonn, Ro¨merstr. 164, 53117 Bonn, Germany
Received 18 October 2007; received in revised form 19 December 2007; accepted 27 December 2007
Dedicated to Prof. A. Muller on the occasion of his 70th birthday
¨
Available online 1 January 2008
Abstract
Ammonolysis of the monomeric, base-stabilized trisaminoalane Me3N-Al[N(H)Dipp)]3 (Dipp ¼ 2,6-iPr2–C6H3) yielded Al–N
oligomers, which were characterized in detail by solid state NMR spectroscopy (1H, 13C, 15N, 27Al) and TGA/DTA. Pyrolysis of as-
prepared oligomers at different temperatures in an argon steam yielded carbon-containing black solids, whereas pyrolysis under a steady
flow of NH3 produced pure aluminum nitride (AlN). The role of the pyrolysis temperature and the influence of NH3 on the formation of
crystalline materials were investigated. As-prepared AlN was characterized by solid state NMR spectroscopy (15N, 27Al), X-ray
diffraction (XRD), transmission electron microscopy (TEM) and electron energy loss spectroscopy (EELS). Theoretical calculations
were performed in order to identify potential reaction intermediates.
r 2007 Elsevier Inc. All rights reserved.
Keywords: Single source precursor; Ammonolysis; AlN; Solid state NMR
1. Introduction
well as bulk AlN uses so-called single source precursors,
which contain the elements of the desired material (e.g. Al
and N) connected by a stable chemical bond [8]. In fact, the
very early study of Wiberg et al. [9] more than 60 years ago
on the reaction of H3Al and NH3 already demonstrated
that the formation of bulk AlN proceeds via formation of
reaction intermediates such as the Lewis acid-base adduct
H3Al–NH3, the aminoalane [H2AlNH2]x and iminoalane
[HAlNH]x. In addition, the reaction of AlH3 with an excess
of NH3 yielded Al(NH2)3 [10], which also was found to be
a suitable single source precursor for AlN (Scheme 1).
Unfortunately, the thermal stability of as described
compounds, which are ideal single source precursors for
the formation of AlN due to the lack of any carbon-
containing organic substituents, is rather low. Consequently,
kinetically stabilized precursors such as amine-alane adducts
R3Al–NR3, aluminum amides (Al[NR2]3, [R2AlNR2]x,y)
Aluminum nitride (AlN) is a ceramic material that has
received considerable interest in the last decade due to its
excellent combination of high thermal conductivity, low
thermal expansion coefficient, excellent chemical and
thermal stability and good mechanical strength. Thin films
are typically obtained from vapor deposition techniques
(MOCVD) [1], whereas bulk materials are usually synthe-
sized by high-temperature processes such as direct nitrida-
tion of Al metal or Al2O3 with N2 or NH3 [2], carbothermal
reduction/nitridation [3] and SHS processes [4] as well as
by reaction of AlCl3 with NH3 [5] or NaN3 [6]. An
alternative, low-temperature approach to AlN films [7] as
Ã
Corresponding author. Fax: +49 5251 603423.
E-mail address: stephan.schulz@uni-due.de (S. Schulz).
0022-4596/$ - see front matter r 2007 Elsevier Inc. All rights reserved.
doi:10.1016/j.jssc.2007.12.026

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
531
2. Experimental
All manipulations were performed by standard Schlenk
techniques under an Ar-atmosphere. Solvents were freshly
distilled from sodium benzophenone ketyl or Na/K-alloy
and NH3 was dried over KOH prior to use. Me3N–
Al[N(H)Dipp]3 I was synthesized by literature method [16],
15NH3 was commercially available from Aldrich. MAS-
NMR experiments (27Al, 15N) were carried out on a Varian
Infinity+spectrometer equipped with a commercial 4 mm
T3 MAS triple-resonance probe. The magnetic field
strength was 9.4 T. Both 15N and 27Al spectra were
acquired using continuous wave decoupling with a field
strength of 100 kHz. The given chemical shift values refer
to CH3NO2 (15N) and a 1.1 M solution of Al(NO3)3 (27Al)
as chemical shift standards. 15N NMR spectra were
acquired with a ramped 15N{1H} cross-polarization experi-
ment for the ammonolysis products and/or single-pulse
excitation for some of the pyrolysis products. The spectra
typically are the average of a few thousand transients
acquired with a repetition delay of 1–3 s and a rotor
spinning frequency of 10 kHz.
XRD-patterns of the materials were recorded on a
Philips PW1050 diffractometer using Cu Ka radiation
with an AFM secondary monochromator. The materials
were also characterized by means of transmission electron
microscopy (TEM) using a Philips CM30ST (LaB6-
cathode) at 300 keV equipped with a parallel electron
energy loss (EEL) spectrometer (Gatan P666) or a Philips
CM300UT (FEG) at 297 keV equipped with a 2k ꢀ 2k
SlowScan-CCD (Gatan MSC) and an imaging filter (Gatan
GIF200) for recording EEL-spectra and energy-filtered
images with a second 1k ꢀ 1k CCD at the end of the filter.
A HP-Ge EDX-detector (Noran Voyager) was present on
both transmission electron microscopes for EDX-analysis.
C–K, N–K, O–K and Al–K lines were used for quantifica-
tion of the EEL-data. The samples were prepared on
perforated carbon foils without further grinding.
Scheme 1. Reaction of AlH3 with NH3 as reported by Wiberg et al.
and imides [RAlNR]x, which are accessible by standard
reactions such as alkane elimination and salt metathesis
reactions, were investigated. Transamination reactions were
found to produce bulk AlN in a multistep reaction process
via formation of oligomeric reaction intermediates, which
yield AlN powders under pyrolysis conditions at higher
temperatures. For instance, the reaction of [Al(NMe2)3]2 and
NH3 initially yielded a mixed Al-amide-imide oligomer
[Al(NH)NH2]x, which still contained NMe2 groups to some
extent [11]. Pyrolysis of this oligomer in an Ar flow resulted
in the formation of a black solid material indicating a high
level of carbon (due to organic residues), whereas a light
gray, nanocrystalline AlN powder was formed in a steam of
NH3 between 900 and 1100 1C. Comparable results were
obtained with [Ga(NMe2)3]2 [12], which reacts with NH3
with formation of gallium imide [Ga(NH)3/2]2 and hetero-
cyclic aminogallane [H2GaNH2]3 [13]. Both precursors were
pyrolyzed under a flow of Ar or NH3 at 500 and 6001C,
respectively, yielding mixtures of cubic and hexagonal GaN.
In addition, heterocyclic Al-amides ([Et2AlNH2]3) [14] and
Al-imides ([HAlNR]x R ¼ Et, iPr) [15] have been successfully
decomposed to nanocrystalline AlN at temperatures of
1000 1C and higher under a steady flow of ammonia. In
contrast, pyrolysis under Ar or N2 yielded materials with
high carbon levels. NH3 was not only found to enhance the
decomposition pathway toward a more complete alkane
elimination and, as a consequence, to lower the C content of
the resulting material but also to significantly increase the
decomposition rate of [Et2AlNH2]3.
2.1. Ammonolysis studies
Herein we report on the results obtained from ammo-
nolysis reactions of the base-stabilized trisaminoalane
Me3N–Al[N(H)Dipp]3 I (Dipp ¼ 2,6-iPr2–C6H3) in liquid
ammonia and from pyrolysis studies of the resulting
oligomeric compounds in a steam of NH3. Different
analytical techniques were used to monitor the progress
of the reactions, including thermal analysis, solid-state
NMR and powder X-ray diffraction (XRD). Since our
interpretation of the 15N NMR spectra relies on quantum
mechanical predictions of isotropic chemical shift values,
we present a set of benchmark calculations and experi-
mental data of compounds similar to I in the Supporting
Information.1
A solution of 3.07 g (5.0 mmol) finely powdered Me3N–
Al[N(H)Dipp]3 I in 30 mL of liquid NH3 and 30 mL of
n-hexane was heated to reflux at ꢁ38 1C for the desired
time, followed by a 2 h NH3 boil-off. The resulting white
slurry was filtered and washed two times with 20 mL of
n-hexane, yielding a colorless powdery material that was
evacuated overnight.
2.2. Pyrolysis studies
Materials obtained after ammonolysis of I for 12 h (IIc)
were used in pyrolysis experiments. Pyrolysis experiments
under a steam of N2 and Ar yielded carbon-containing
black products. Therefore, pyrolysis experiments were
performed in a quartz boat under a steady steam of
ammonia at 600, 800, 900, 1000 and 1150 1C, respectively.
1Electronic Supplementary Information (ESI) available.

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
532
The precursor was typically preheated at 200 1C (1 h),
ramped (15 min) to the desired temperature, and held for
3 h. Light gray (600 1C) and white powders (800, 900, 1000,
1150 1C) were formed, which were analyzed by X-ray
powder diffraction (XRD), 15N and 27Al NMR spectro-
scopy. In addition, the sample obtained at 1150 1C was also
characterized by TEM and EEL spectroscopy.
transfer the calculated chemical shifts to the CH3NO2
standard, the experimental value dref,exp of a saturated
aqueous solution of NH4Cl was used which was experi-
mentally determined to be ꢁ353 ppm [30].
3. Results and discussion
Reaction of three equivalents of DippNH2 and
H3Al–NMe3 yielded the base-stabilized trisaminoalane
Me3N–Al[N(H)Dipp]3 I, which was obtained as colorless
crystalline solid after crystallization at ꢁ10 1C [16]. I has
been previously characterized in detail by mass and
(solution) NMR spectroscopy (1H, 13C) as well as by
single crystal XRD. In order to examine structural changes
resulting from ammonolysis and thermolysis reactions, we
characterized I by solid state 15N NMR spectroscopy. The
15N-CP-NMR spectrum of I shows a sharp peak at
ꢁ355.1 ppm due to the coordinating NMe3 group and a
broad signal due to the amino substituents (N(H)Dipp) at
ꢁ326.2 ppm, respectively. The broadening is probably
caused by quadrupolar-dipolar cross terms to coordinated
27Al-atoms.
2.3. Computational details
All calculations were performed using modules of the
program package turbomole [17]. Geometries were opti-
mized at the DFT level of theory employing the local BP86
functional [18–22] and using the valence triple zeta basis
set with polarization functions on all atoms (TZVP) from
the turbomole library [23]. Calculations were sped up
by applying the RI approximation for computation of
electronic Coulomb interactions [24].
NMR chemical shifts were calculated at the coupled-
perturbed Hartree–Fock level of theory using the modules
dscf and mpshift of the turbomole package [25,26] and
employing the TZP basis set from the turbomole library
[27]. The choice of the method was governed by the rather
large size of the systems which made the application of
higher-level methods like MP2 for the calculation of the
chemical shifts intractable. Experience with chemical shift
calculations on boron and silicon nitrides has shown,
however, that already at the Hartree-Fock level good
results may be obtained [28]. On the other hand it is also
well known that there are cases where electron correlation
plays a decisive role and the Hartree–Fock method badly
fails [29].
3.1. Ammonolysis studies
Me3N–Al[N(H)Dipp]3 I was dissolved in n-hexane and
ammonolyzed in boiling 15NH3 at ꢁ38 1C for different
times (3 h IIa, 6 h IIb, 12 h IIc, 48 h IId). Thereafter, NH3
was allowed to boil off at ꢁ10 1C, yielding a white
suspension, from which colorless crystalline solids were
isolated by filtration [31]. XRD diagrams of these solids did
not show any distinguished peaks indicating the complete
transformation of the crystalline starting material I (black
curve Figs. 1 and 2) into amorphous materials (red curve)
under these specific reaction conditions [32].
TGA studies of as-prepared solid materials were
performed in a steam of Ar. The observed mass loss was
found to steadily decrease with increasing ammonolysis
time (85% (3 h), 61% (6 h), 37% (12 h), 25% (48 h)), clearly
demonstrating a strong relationship between the degree of
In the framework of our approach the components of
the chemical shielding tensor rA of an atom A are given by
the mixed second derivative of the total energy E with
respect to the nuclear magnetic moment lA of atom A and
the external magnetic field B (a and b refer to cartesian
coordinates).
ꢀ
ꢁ
q2E
qmAaqBb
sAab
¼
.
Before the calculated values of the isotropic chemical
shielding
À
Á
1
3
sA;iso
¼
sxx þ syy þ szz
can be compared with experimental data, they have to be
converted to the chemical shift scale. To this end the
computed shielding sref,calc of a reference substance is
related to its experimental chemical shift dref,exp according
to the following equation:
dsample;calc ¼ dref;exp þ sref;calc ꢁ ssample;calc
with the indices denoting a calculated (calc) or experi-
mental (exp) value, the substance of interest (sample) and a
reference substance (ref). The reference substance used in
our calculations was the ammonium cation NH+4 . To
Fig. 1. XRD patterns of precursor I (black; calculated from single crystal
X-ray data), ammonolyzed sample IId (red), TGA residue (blue) and of
crystalline AlN (green, JCPDS 25-1133).

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
533
transamination of Me3N–Al[N(H)Dipp]3 and reaction
time.
demanding organic substituents (Dipp) kinetically stabilize
the trisaminoalane and hence significantly reduce its
reactivity toward transamination reactions. Wells et al.
[11] reported on the transamination reaction of the
sterically less protected trisaminoalane Al(NMe2)3 in liquid
NH3, yielding [Al(NH)NH2]x after 10 h [33]. XRD spectra
of the black residues of the TGA studies showed very
broad reflexes due to the formation of 2H-wurtzite-type
AlN as shown in Fig. 1 (blue curve).
Solid state 15N{1H} CP MAS NMR spectra of the
oligomeric material were recorded in order to qualitatively
monitor structural changes under ammonolysis conditions
(see Fig. 3). These NMR spectra show very broad peaks
around ꢁ350 ppm with large half widths indicating the
formation of amorphous oligomeric species featuring N
atoms in different chemical environments. Even if the
sample is ammonolyzed for a long time no intermediate
crystalline phases containing hydrogen and nitrogen
are formed.
The mass loss observed for each sample starts at 100 1C
and is finished at about 500 1C. The calculated mass loss
for the transformation of Me3N–Al[N(H)Dipp]3 into AlN
is 93.3%. Assuming that ammonolysis of Me3N–Al[N(H)-
Dipp]3 yields the completely transaminated oligomeric
aluminum-amide-imide [Al(NH)NH2]x, the expected mass
loss for the formation of AlN is 29.4%. According to
this rough calculation and the experimental TGA results,
Me3N–Al[N(H)Dipp]3 has to be ammonolyzed for at
least 20 h to form a polymeric material, most likely
[Al(NH)NH2]x, that is transaminated to a considerable
extent, whereas shorter ammonolysis reactions yield poly-
meric materials, which still contain organic groups
(NHDipp) to some extent. Obviously, the sterically
3.2. Computational studies
A series of quantum chemical computations of 15N
chemical shifts of several amino- and iminoalanes with
bulky NHDipp substituents were performed in order to
gain insight into NMR chemical shifts in chemical
environments present in these molecular compounds. This
should help to understand structural changes that occur
under ammonolysis conditions as indicated in the 15N
Fig. 2. TGA curves of ammonolyzed samples (black: 3 h, green: 6 h, blue:
12 h, red; 48 h).
Fig. 3. 15N{1H} CP MAS NMR spectra of the starting compound Me3N–Al[N(H)Dipp)3 and ammonolyzed samples (1, 3, 7, 12, 20 h).

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
534
Fig. 4. 15N chemical shifts in amino- and iminoalanes containing bulky NHDipp substituents. Chemical environments are classified using different
symbols. NR3 coordinated to Al (’); NHCar bonded to Al and coordinating to another Al (.); NHCar bonded to Al (c); NCar bonded to two Al (m); N
bonded to two aliphatic and one aromatic C atoms (K). (1) Diagonal ligands cis and (2) diagonal ligands trans.
NMR spectra and to interpret the NMR spectra of the
early stages of the transformation of the molecular
precursor I. More detailed information about the valida-
tion of the methods for calculating chemical shifts is
provided as supporting information.
Whenever available (compounds I, IV and V), experi-
mentally determined crystal structures served as starting
points of the geometry optimizations. Conformational
differences between the optimized structures and the crystal
structures were only small. Errors of calculated bond
lengths of covalent Al–N bonds are only 2–3%, whereas
errors of up to 5% were found for Al–N donor–acceptor
bond distances. The latter finding is in accord with earlier
observations that the bond strengths of dative bonds are
typically underestimated by density functional theory [34].
The calculated 15N NMR chemical shifts and the structures
of the molecular compounds are displayed in Fig. 4.
Symbols are used for classification of different chemical
environments. Taking into account the computational
uncertainties of the results, it is difficult to distinguish
between nitrogen atoms bonded to one phenyl group (R)
and to one or two Al atoms and/or H. Clearly separated
from these shifts and appearing at higher ppm values are
the 15N chemical shifts in NR3Al environments.
Thereafter, we performed theoretical calculations on
hydrogen-saturated clusters of crystalline AlN in order to
derive a correlation between the 15N chemical shift and
the number of H atoms in chemical environments of the
general type NAl4ꢁxHx. Since the details of these calcula-
tions are beyond the scope of this paper [35], only a rough
description is given here. Clusters of different sizes were
generated starting from cut-offs of crystalline wurtzite and
zinc-blende AlN. Dangling bonds on the surfaces were
saturated by hydrogen atoms. It turned out that it is

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
535
necessary to carefully choose the number of electrons of the
clusters. This was achieved in the following way: H atoms
in positions of coordinative-bonded N atoms contributed
two electrons, H atoms in positions of coordinative-bonded
Al atoms contributed no electron, and all other H atoms
contributed one electron. This scheme consequently
resulted in the formation of charged clusters. The charges
were kept as small as possible by suitable choice of the
cluster shape. The positive or negative charges of all
clusters did not exceed 73. Calculation of the chemical
shifts of the central atoms of the largest clusters confirmed
that the electronic structure resembles that of solid AlN.
The structures of all clusters were optimized under
symmetry restrictions (C3v for wurtzite AlN clusters and
Td for zinc-blende AlN clusters). For obtaining trends of
the chemical shifts we examined all nitrogen atoms of these
clusters. With decreasing number of H atoms bonded to
the N atom, d was found to be shifted to higher (less
negative) ppm values. Fig. 5 gives an overview of all
calculated N chemical shifts [36]. The symbols used for
Fig. 5. 15N chemical shifts in different chemical environments of amino-
and iminoalanes (lower part of the diagram) and in NAl4ꢁxHx
environments of AlN clusters (upper part of the diagram). The symbols
mark the mean values of the chemical shifts, bars denote the range of the
chemical shifts plus an uncertainty of the calculated data of 6.8 ppm.
Fig. 6. 15N MAS NMR spectra of samples pyrolyzed for 5 h in a stream of NH3 at the given temperatures; the samples pyrolyzed above 800 1C were
acquired with a Hahn-echo since only the samples pyrolyzed at 600 and 800 1C showed a cross-polarization 15N{1H} signal.

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
536
classification of different chemical environments are the
same as in Fig. 4.
The computational results allow a qualitative interpreta-
tion of the broad 15N NMR signals recorded at different
stages of the ammonolysis reaction. The broad shoulder at
high ppm values (left side of the peak), which can be
ascribed to the presence of NAl2C, NHAlC and NHAl2C
units in the samples, decreases with increasing ammono-
lysis time whereas the shoulder at low ppm values (right
side of the peak), which can be attributed to the formation
of NAlH3 and NAl2H2 moieties, increases in intensity.
Even though an exact differentiation of the different
NAl2(R,H)1,2 environments solely based on the 15N
NMR data is not possible since their calculated chemical
shifts fall into roughly the same range, the data strongly
indicate ongoing substitution reactions of the starting
NHDipp and NMe3 groups by NH2 and NH3 groups
(transamination reaction) with increasing ammonolysis
time. These findings also perfectly agree with the results
obtained from TGA/DTA studies.
Fig. 7. Central part of 27Al MAS NMR spectra of samples pyrolyzed for
5 h in a stream of NH3 at the given temperatures; asterisks mark rotational
sidebands.
3.3. Thermolysis studies
Oligomeric material obtained from ammonolysis reac-
tions of Me3N–Al[N(H)Dipp]3 for 12 h (IIc) were pyro-
lyzed for 5 h at 600 and 800 1C, respectively, in an Ar flow.
Black, amorphous solids were obtained in both cases,
which did not show any distinguished reflexes in XRD
experiments. The black color indicates the presence of
sufficient amounts of carbon in the materials, which was
verified by EEL spectroscopy [37]. The formation of
carbon-contaminated AlN has been previously observed
in pyrolysis reactions of single source precursors such as
[HAliPr]6 and [R2AlNH2]3 (R ¼ Me, Et), respectively, in a
steady flow of N2 or Ar. Since the addition of NH3 is
known to enhance the quality of the resulting AlN by
significant reduction of the carbon content, to increase the
rate of decomposition and to improve the crystallinity of
the resulting materials [14], we investigated the pyrolysis of
IIc in a steady steam of NH3. Samples of IIc were
pyrolyzed at 500, 800, 1000 and 1150 1C for 5 h, yielding
colorless materials.
15N MAS NMR spectra of the resulting materials
significantly change with increasing pyrolysis temperature.
Owing to the low NMR sensitivity of the isotope 15N, the
spectra were recorded in a non-quantitative manner. The
materials formed at 600 and 800 1C show a broad peak at
about ꢁ350 ppm as was observed for the ammonolyzed
material (Fig. 6). However, the center is shifted to the left
side to higher ppm values. In remarkable contrast, samples
pyrolyzed at 1000 and 1150 1C show strong sharp
resonances at ꢁ299 and ꢁ296 ppm, respectively. These
correspond very well with the 15N chemical shift observed
for crystalline AlN (ꢁ295 ppm) as determined by a
15N{27Al} FAM-CP-MAS experiment [38,43]. Small differ-
ences in the chemical shifts can be explained by suscept-
ibility changes of the surrounding matrix. Obviously,
crystalline AlN grows out of the amorphous polymeric
AlN matrix at temperatures higher than 1000 1C.
27Al MAS NMR spectra also show significant changes
(Fig. 7). The spectrum of the material obtained at 600 1C
shows a very broad resonance at 90 ppm, whereas the
materials pyrolyzed at 800 and 900 1C show very sharp
resonances at 111 ppm and some distinguished peaks at
higher field (63, 32, 4 ppm, 800 1C sample) or a broad
shoulder at the left side of the main signal (900 1C). In
sharp contrast, samples formed at 1000 and 1150 1C only
show sharp single resonances at 112 and 113 ppm,
respectively, which also agree very well with the 27Al
resonance of crystalline AlN (112.7 ppm).
XRD powder patterns clearly demonstrate the forma-
tion of crystalline AlN at pyrolysis temperatures above
1000 1C (Figs. 8 and 9). Patterns of the materials formed
below 1000 1C did not show any distinguished peaks
proving the resulting material to be amorphous. In
contrast, the material obtained at 1150 1C gave sharp
reflexes in the XRD, clearly revealing the formation of
crystalline 2H-wurtzite-type AlN.
TEM investigations on the crystalline AlN obtained at
1150 1C showed agglomerates of plate-like thin crystals
with an average diameter of the crystallites of 10 nm. The
crystals are randomly distributed and show no preferred
orientation. Selected area electron diffraction (SAED)
yielded Debye–Scherrer ring patterns, which could be
simulated with wurtzite-type AlN (Fig. 8c) [39]. The lattice
parameters obtained from these patterns (a: 310(2); c:
499(3) pm) agree perfectly with literature data [40]. It is
hardly possible to bring the nanocrystals in distinguished
zone axis orientation. Crystals lying close to zone axis

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
537
conditions by chance showed distances of the lattice fringes
of AlN. The AlN nanocrystals are covered by a small
amorphous seam. EEL-spectra of the material clearly
revealed the presence of aluminum and nitrogen as well as
small traces of oxygen and silicon. Si contamination occurs
from the use of joint grease during the ammonolysis and
pyrolysis experiments. There is no carbon detectable in
sample regions overhanging holes of the underlying carbon
films. The N/O-ratio determined by quantification of the
EEL-spectra decreases with decreasing thickness of the
transmitted area. The thickness was determined from
the ratio of in-elastically scattered electrons to elastically
scattered electrons. Thus, the amount of oxygen is caused
by a surface layer and the thicker the sample, the lesser the
amount of oxygen relatively to the nitrogen concentration.
The ELNES fingerprint of the N–K edge compares well
with literature values as found in the EEL and X-ray
database [41]. In addition, they are in good agreement with
values assigned to non-oxidized AlN as reported by
McKenzie and Craven [42].
4. Conclusions
Ammonolysis of Me3N–Al[N(H)Dipp)]3 yielded amor-
phous, oligomeric materials, which can be pyrolyzed under
a steady flow of NH3 to give pure AlN. Theoretical
calculations were performed in order to identify potential
reaction intermediates. Solid-state NMR spectroscopy
(15N, 27Al) and XRD studies clearly demonstrated the
formation of crystalline AlN at temperatures higher than
1000 1C.
Fig. 8. Calculated and experimental powder X-ray diffraction pattern of
crystalline AlN obtained at 1150.
Fig. 9. TEM bright field images of crystalline AlN nanoparticles obtained at 1150 1C (a, b). Electron diffraction pattern of AlN and simulation (c) and
ELNES of the N–K edge (d).

ARTICLE IN PRESS
S. Schulz et al. / Journal of Solid State Chemistry 181 (2008) 530–538
538
Acknowledgments
[18] P.A.M. Dirac, Proc. Roy. Soc. A 123 (1929) 714.
[19] J.C. Slater, Phys. Rev. 81 (1951) 385.
[20] S. Vosko, L. Wilk, M. Nusair, Can. J. Phys. 58 (1980) 1200.
[21] A.D. Becke, Phys. Rev. A 38 (1988) 3098.
[22] J.P. Perdew, Phys. Rev. B 33 (1986) 8822.
We gratefully acknowledge financial support of this
work by the Deutsche Forschungsgemeinschaft.
[23] A. Schafer, C. Huber, R. Ahlrichs, J. Chem. Phys. 100 (1994) 5829.
¨
Appendix A. Supporting Information
[24] K. Eichkorn, F. Weigend, O. Treutler, R. Ahlrichs, Theor. Chem.
Acc. 97 (1997) 119.
[25] M. Haser, R. Ahlrichs, J. Comp. Chem. 10 (1989) 104.
¨
Supplementary data associated with this article can be
found in the online version at doi:10.1016/j.jssc.2007.12.026.
[26] M. Haser, R. Ahlrichs, H.-P. Baron, P. Weis, H. Horn, Theor. Chim.
¨
Acta 83 (1992) 455.
[27] A. Schafer, H. Horn, R. Ahlrichs, J. Chem. Phys. 97 (1992) 2571.
¨
[28] M. Doerr, C.M. Marian, Sol. State Nucl. Magn. Reson. 30 (2006) 16;
C.M. Marian, M. Gastreich, Sol. State Nucl. Magn. Reson. 19 (2001)
29.
References
[29] J. Gauss, J.F. Stanton, Adv. Chem. Phys. 123 (2002) 355.
[30] M. Witanowski, L. Stefaniak, S. Szymanski, H. Januszewski,
J. Magn. Reson. 28 (1977) 217.
[1] D. Ruhela, M. Ritala, R. Matero, M. Leskela, J. Jokinen,
¨
¨
¨
P. Haussalo, Chem. Vap. Deposition 2 (1996) 277.
[2] Y. Qui, L. Gao, J. Eur. Ceram. Soc. 23 (2003) 2015;
E. Kroke, L. Loffler, F.F. Lange, R. Riedel, J. Am. Chem. Soc. 85
¨
[31] The filtrate gave an oily residue after evaporation of the solvent,
1
which was shown by H NMR spectroscopy to consist of DippNH2,
(2002) 3117.
[3] T. Suehiro, J. Tatami, T. Meguro, S. Matsuo, K.J. Komeya, Eur.
Ceram. Soc. 22 (2002) 521.
clearly demonstrating the occurrence of transamination reactions.
[32] The powder X-ray diffraction pattern of I was calculated from the
single crystal X-ray data.
[4] R. Fu, K. Chen, X. Xu, J.M.F. Ferreira, Mater. Lett. 59 (2005) 2605.
[5] M. Pouget, J.P. Lecompte, J. Eur. Ceram. Soc. 16 (1996) 521.
[6] See the following and references cited therein: K. Sardar,
C.N.R. Rao, Solid State Sci. 7 (2005) 217.
[33] The overall reaction time reported for the reaction of Al(NMe2)3 in
liquid NH3 was 8 h, followed by a 2 h NH3 boil-off. The calculated
weight loss for [Al(NH)NH2]x is 29.4%, but the TGA curve showed a
mass loss of only 20%.
[7] For review articles see: A.C. Jones, S.A. Rushworth, J. Auld, J. Cryst.
Growth 146 (1995) 503;
A.C. Jones, Chem. Soc. Rev. (1997) 101;
[34] S. Reinhardt, C.M. Marian, I. Frank, Angew. Chem. Int. Ed. Engl.
133 (2001) 3795;
R. Ahlrichs, F. Furche, S. Grimme, Chem. Phys. Lett. 325 (2000) 317.
[35] The details will be presented in a separate publication: M. Doerr,
C.M. Marian, Solid State NMR, in preparation. At present they are
available as part of a PhD thesis: Doerr, M. PhD thesis, University of
A.V. Timoshkin, H.F. Schaefer III, Chem. Rec. 2 (2002) 319.
[8] L. Baixia, L. Yinkui, L. Yi, J. Mater. Chem. 3 (1993) 117.
[9] E. Wiberg, A. May, Z. Naturforsch. B 10 (1955) 229.
[10] E. Wiberg, A. May, Z. Naturforsch. B 10 (1955) 231.
[11] J.F. Janik, R.L. Wells, J.L. Coffer, J.V. St. John, W.T. Pennington,
G.L. Schimek, Chem. Mater. 10 (1998) 1613.
Dusseldorf 2006. URL: /http://docserv.uni-duesseldorf.de/servlets/
¨
DocumentServlet?id=3461S.
[36] Note that the ranges of the chemical shifts are based on different
numbers of data points, which explains the different widths of the
shift intervals in Fig 5. The largest number of data points (37) was
available for the chemical shifts in NAl4 environments. Smaller
numbers of data points were available for chemical shifts in
environments NAl4ꢁxHx with larger numbers of H atoms because
such environments are only found at the surfaces of the clusters. The
interval of the chemical shifts in NAlH3 environments based on only
four data points.
[12] J.F. Janik, R.L. Wells, Chem. Mater. 8 (1996) 2708.
[13] J.-W. Hwang, J.P. Campbell, J. Kozubowski, S.A. Hanson,
J.F. Evans, W.L. Gladfelter, Chem. Mater. 7 (1995) 517.
[14] F.C. Sauls, W.J. Hurley, L.V. Interrante, P.S. Marchetti,
G.E. Maciel, Chem. Mater. 7 (1995) 1361.
[15] Y. Sugahara, T. Onuma, O. Tanegashima, K. Kuroda, C. Kato,
J. Ceram. Soc. Jpn. 100 (1992) 101;
Y. Saito, S. Koyama, Y. Sugahara, K. Kuroda, J. Ceram. Soc. Jpn.
104 (1996) 143;
[37] EEL spectroscopy studies clearly revealed the presence of carbon in
the material.
S. Koyama, H. Takeda, Y. Saito, Y. Sugahara, K. Kuroda, J. Mater.
Chem. 6 (1996) 1055;
[38] P.K. Madhu, A. Goldbourt, L. Frydman, S. Vega, Chem. Phys. Lett.
307 (1999) 41.
Y. Sugahara, S. Koyama, K. Kuroda, Key Eng. Mater. 159,160
(1999) 77.
[39] P. Stadelman, Jems Electron Microscopy Software, EPFL Lausanne,
Switzerland, 1999–2007.
[16] T. Bauer, S. Schulz, H. Hupfer, M. Nieger, Organometallics 21 (2002)
2931.
[17] R. Ahlrichs, M. Bar, H.-P. Baron, R. Bauernschmitt, S. Bocker,
[40] PDF2 file #025-1133 JCPDS-International Centre for Diffraction
Data, USA.
¨
¨
P. Deglmann, M. Ehrig, K. Eichkorn, S. Elliot, F. Furche, F. Haase,
M. Haser, C. Hattig, H. Horn, C. Huber, U. Huniar, M. Kattannek,
[41] EELS and X-ray Database. /http://www.cemes.fr/ꢂeelsdbS.
[42] M. MacKenzie, A.J. Craven, J. Phys. D: Appl. Phys. 33 (2000) 1647.
More detailed EELS data are supplied with the supplement material.
[43] M. Eden, J. Grins, Z. Shen, Z. Weng, J. Magn. Reson. 169 (2004)
279.
¨
¨
A. Kohn, C. Kolmel, M. Kollwitz, K. May, C. Ochsenfeld, H. Ohm,
¨
¨
A. Scha
¨
fer, U. Schneider, M. Sierka, O. Treutler, B. Unterreiner,
¨
M. von Arnim, F. Weigend, P. Weis, H. Weiss, Turbomole (vers. 5–6),
2002.