Solution chemistry of cyclotrigallazane: supercritical ammonia and lewis base catalyzed dehydrogenation to produce poly(imidogallane)

Article abstract of DOI:10.1021/ic981298l

The polymeric hydrido gallium imide [HGaNH]_n, or poly(imidogallane), was isolated as an intermediate in the conversion of cyclotrigallazane, [H₂GaNH₂]₃, to gallium nitride in supercritical ammonia. Poly(imidogallane) can be converted to nanocrystalline GaN via solid-state pyrolysis or by extended thermolysis in supercritical ammonia. Characterization by IR spectroscopy, elemental analysis, and X-ray and electron diffraction suggests that the structure of poly(imidogallane) is a layered solid with a coherent length of approximately 15 nm. The interaction of [H₂GaNH₂]₃ and various Lewis bases leads to the isolation of [HGaNH]_n and gallane - Lewis base adducts, H₃Ga·base, rather than the expected gallazane adducts, H₂GaNH₂·base. Experiments involving the use of ND₃ resulted in the isolation of [H₂GaND₂]₃ and suggest that the donor - acceptor complex H₂Ga(NH₂)(ND₃) was present in solution. It was proposed that these gallazane - base adducts equilibrate via a ligand redistribution reaction and that the formation of [HGaNH]_n results from NH₃ elimination from a diamidogallium hydride. The proposed mechanism is used to explain the catalytic role of ammonia in the conversion of [H₂GaNH₂]₃ to [HGaNH]_n.

Full text of DOI:10.1021/ic981298l

2726
Inorg. Chem. 1999, 38, 2726-2733
Solution Chemistry of Cyclotrigallazane: Supercritical Ammonia and Lewis Base Catalyzed
Dehydrogenation To Produce Poly(imidogallane)
Jolin A. Jegier, Stuart McKernan, and Wayne L. Gladfelter*
Department of Chemistry and Center for Interfacial Engineering, University of Minnesota,
Minneapolis, Minnesota 55455
ReceiVed NoVember 10, 1998
The polymeric hydrido gallium imide [HGaNH]n, or poly(imidogallane), was isolated as an intermediate in the
conversion of cyclotrigallazane, [H2GaNH2]3, to gallium nitride in supercritical ammonia. Poly(imidogallane)
can be converted to nanocrystalline GaN via solid-state pyrolysis or by extended thermolysis in supercritical
ammonia. Characterization by IR spectroscopy, elemental analysis, and X-ray and electron diffraction suggests
that the structure of poly(imidogallane) is a layered solid with a coherent length of approximately 15 nm. The
interaction of [H2GaNH2]3 and various Lewis bases leads to the isolation of [HGaNH]n and gallane-Lewis base
adducts, H3Ga‚base, rather than the expected gallazane adducts, H2GaNH2‚base. Experiments involving the use
of ND3 resulted in the isolation of [H2GaND2]3 and suggest that the donor-acceptor complex H2Ga(NH2)(ND3)
was present in solution. It was proposed that these gallazane-base adducts equilibrate via a ligand redistribution
reaction and that the formation of [HGaNH]n results from NH3 elimination from a diamidogallium hydride. The
proposed mechanism is used to explain the catalytic role of ammonia in the conversion of [H2GaNH2]3 to [HGaNH]n.
Introduction
Experimental Section
General. All manipulations were carried out with the rigorous
exclusion of oxygen and moisture using standard Schlenk and drybox
techniques (Vacuum Atmospheres Company, Dri-Train Model 40-1).
Tetrahydrofuran was distilled from sodium benzophenone ketyl im-
mediately prior to use. Anhydrous NH3, ND3, and NMe3 were dried
over a sodium mirror before being vacuum transferred to reaction
vessels. Cyclotrigallazane, [H2GaNH2]3,4 and the gallane adducts H3-
Due to their unique electronic properties, there has been
increased interest in the syntheses of nanocrystalline 13-15
semiconducting materials.1 Unfortunately, the current synthetic
methods offer little control over particle size and dispersion,
especially for the nitrides. Cyclotrigallazane ([H2GaNH2]3, 1)
is one of the known precursors to nanocrystalline GaN.2 Its
dehydrogenation to GaN involved a topochemical conversion
of 13 resulting from a network of unconventional hydrogen
bonds in the crystal structure of 1.4 We became interested in
developing a more controlled, solution-based route to GaN
utilizing cyclotrigallazane. The solvent of choice was super-
critical NH3, which has been found to be useful in the syntheses
of solid-state compounds, especially metal amides, imides, and
nitrides.5-7 During the course of our investigations, we have
isolated poly(imidogallane) (2), a polymeric intermediate in the
conversion of [H2GaNH2]3 to GaN in ammonia solution.8 This
paper describes the full synthesis and characterization of 2 and
delves more deeply into the reactions of 1 with a variety of
Lewis bases in solution. These studies have indicated that
cyclotrigallazane possesses an unusual reactivity in solution and
have led to an explanation for the formation of 2.
Ga‚quinuclidine,9 H3Ga‚NMe3,10 and H3Ga‚PCy3 were prepared by
11
literature methods.
Infrared spectra were acquired as KBr pellets or in solution using a
Nicolet Magna-IR 560 spectrometer and are reported in cm-1. Proton
NMR spectra were recorded on a Varian Inova 300 spectrometer in
C6D6 solvent and are referenced to the residual protons in the solvent
(7.15 ppm). Elemental analyses were obtained from Schwarzkopf
Microanalytical Laboratories, Woodside, NY.
High-Pressure Experiments. Reactions involving the use of NH3
at elevated temperatures were conducted in a steel autoclave (Parr
Instrument Corporation, Model A1120HC, 50 mL capacity) equipped
with a Pyrex sleeve and ethylene/propylene O-rings. The autoclave was
heated using an aluminum block, and the temperature was controlled
by means of an Omega temperature controller. Pressure was monitored
with an Omega electronic pressure transducer.
Synthesis of [HGaNH]n (2). In the drybox, the autoclave was
charged with 1 (2.15 mmol, 565 mg). The autoclave was sealed,
removed from the drybox, and attached to a vacuum line, where
ammonia (approximately 15 mL) was condensed into the autoclave at
-78 °C. The autoclave was then sealed and placed in the heating block
where it was heated to 150 °C over a period of 30 min. The reactor
was held at this temperature for 30 min, with a sustained pressure of
63 atm, and then cooled to 25 °C. The NH3 was vented, and the
(1) See for example: Wells, R. L.; Gladfelter, W. L. J. Cluster Sci. 1997,
8, 217 and references therein.
(2) Hwang, J.-W.; Hanson, S. A., Britton, D.; Evans, J. F.; Jensen, K. F.;
Gladfelter, W. L. Chem. Mater. 1990, 2, 342.
(3) Hwang, J.-W.; Campbell, J. P.; Kozubowski, J.; Hanson, S. A., Evans,
J. F.; Jensen, K. F.; Gladfelter, W. L. Chem. Mater. 1995, 7, 715.
(4) Campbell, J. P.; Hwang, J.-W.; Young, V. G., Jr.; Von Dreele, R. B.;
Cramer, C. J.; Gladfelter, W. L. J. Am. Chem. Soc. 1998, 120, 521.
(5) Jacobs, H.; Schmidt, D. Cur. Top. Mater. Sci. 1982, 8, 381.
(6) Young, D. M.; Schimek, G. L.; Kolis, J. W. Inorg. Chem. 1996, 35,
7620.
(9) Atwood, J. L.; Bott, S. G.; Elms, F. M.; Jones, C.; Raston, C. L. Inorg.
Chem. 1991, 30, 3793.
(10) Shriver, D. F.; Shirk, A. E. Inorg. Synth. 1977, 17, 42.
(11) Atwood, J. L.; Robinson, K. D.; Bennett, F. R.; Elms, F. M.;
Koutsantonis, G. A.; Raston, C. L.; Young, D. J. Inorg. Chem. 1992,
31, 2673.
(7) Purdy, A. P. Chem. Mater. 1998, 10, 692.
(8) Jegier, J. A.; McKernan, S.; Gladfelter, W. L. Chem. Mater. 1998,
10, 2041.
10.1021/ic981298l CCC: $18.00 © 1999 American Chemical Society
Published on Web 05/14/1999

Solution Chemistry of Cyclotrigallazane
Inorganic Chemistry, Vol. 38, No. 11, 1999 2727
autoclave was flushed with a stream of dry N2. The product recovered
from the autoclave was placed under vacuum for 24 h, resulting in the
isolation of 2 as a slightly off-white solid (507 mg, 92%). Mp: >270
°C. IR (KBr): νN-H 3280 (m); νGa-H 1883 (s), 1817 (m); ω,τN-H 962
(s), 906 (s). Anal. Calcd for [GaH2N]n: C, 0.0; H, 2.35; N, 16.34; Ga,
81.31. Found: C, <0.10; H, 2.34; N, 16.59; Ga, 81.25.
Conversion of [H2GaNH2]3 to GaN in Liquid Ammonia. The
procedure described above for the preparation of 2 was followed using
1 (1.14 mmol, 300 mg) and NH3 (approximately 30 mL). The autoclave
was maintained at 150 °C and 65 atm for 5 h. The product was isolated
as an off-white solid (211 mg, 74%) and was shown by XRD to be
GaN. Conducting the reaction at slightly lower pressures (52 atm) had
no visible effect on the XRD of the product. XRD (2θ): 35.1°, 57.8°,
69.0°.
Attempted Synthesis of [HGaNH]n (2) in the Solid State. In the
drybox, a quartz boat was loaded with 1 (0.569 mmol, 150 mg). The
boat was placed in a pyrolysis tube and sealed. The tube was placed in
a furnace and connected to a nitrogen line and heated to 150 °C under
a flow of N2 (approximately 30 mL/min). The tube was kept at this
temperature for 30 min and then cooled to 25 °C. The product was
isolated as a gray black powder (133 mg) which was shown to be
amorphous by XRD. IR (KBr): 3242 (w), 1875 (w), 956 (w).
Synthesis of [H2GaND2]3 (3) and [H2GaN(H)D]3 (4). ND3 (3 mL)
was condensed onto 1 (0.855 mmol, 225 mg) at -196 °C. The flask
was slowly warmed to -33 °C, resulting in a clear, colorless solution.
The flask was allowed to warm to 25 °C over a period of several hours,
and the solvent and byproducts were evaporated under a flow of dry
N2, resulting in the isolation of a white solid in essentially quantitative
yield. This solid was found to be 3, [H2GaND2]3, containing small
amounts (<5%) of 1, [H2GaNH2]3, and 4, [H2GaNHD]3. IR for 3
(KBr): νN-D 2470 (s), 2461 (s), 2391 (s), 2385 (s); νGa-H 1890 (s),
1865 (s), 1834 (s); δN-D 1122 (m). IR for 1 (KBr): νN-H 3312 (w),
3299 (w), 3255 (w), 3246 (w); νGa-H 1890 (s), 1865 (s), 1834 (s). IR
for 4 (KBr): νN-H 3286 (w), 3272 (w); νN-D 2428 (w), 2420 (w), νGa-H
1890 (s), 1865 (s), 1834 (s); δN-H 1339 (w).
Reaction of [H2GaNH2]3 with Tetrahydrofuran. To a flask charged
with 1 (0.950 mmol, 250 mg) was added THF (10 mL). The flask was
sealed and the slightly turbid solution was allowed to stand undisturbed
for 5 days at 25 °C, resulting in the deposition of a white film at the
bottom of the flask. The solvent was decanted off, and the solid film
was washed with THF (5 mL). The solid was dried under vacuum,
resulting in the isolation of 2 in approximately 20% yield. IR (KBr):
3324 (w), 3277 (m), 1878 (s), 960 (s), 905 (s). In a separate experiment,
1 (0.569 mmol, 150 mg) was dissolved in THF (10 mL). After the
solid had completely dissolved, an aliquot was removed and the IR
spectrum was obtained (νGa-H 1858 cm-1). Over the course of several
days the νGa-H at 1858 cm-1 diminished and a new peak at 1817 cm-1
grew.
Reaction of [H2GaNH2]3 with Quinuclidine. To a mixture of 1
(0.950 mmol, 250 mg) and quinuclidine (2.85 mmol, 317 mg) was
added THF (20 mL). The resulting clear, colorless solution was stirred
at 25 °C. After 1 h, an aliquot was removed (approximately 0.4 mL)
and characterized by IR spectroscopy (νGa-H 1821 cm-1). After 2 h
the solution became hazy. The mixture was stirred for 48 h, resulting
in a colorless solution and a white solid. The mixture was filtered on
a fine glass frit, and the white solid was washed with an additional 10
mL of THF. The solid was dried under reduced pressure, yielding 110
mg (90%) of 2. IR (KBr): 3327 (w), 3278 (m), 1879 (s), 960 (s), 905
(s). The volatiles were removed from the filtrate under reduced pressure,
yielding 214 mg (82%) of a white crystalline solid which was identified
as H3Ga(quinuclidine). IR (KBr): νGa-H 1818 (independently prepared
sample of H3Ga(quinuclidine) 1818; lit.9 1810). 1H NMR (C6D6): δGa-H
4.85 (br s) (independently prepared sample of H3Ga(quinuclidine) 4.87
(br s); lit.9 4.80 (br s)).
Reaction of [H2GaNH2]3 with Trimethylamine. Trimethylamine
(approximately 25 mL) was condensed onto 1 (1.52 mmol, 400 mg) at
-196 °C. The flask was warmed to -10 °C and stirred for 1 h. The
excess NMe3 was allowed to evaporate at 0 °C under a stream of N2,
resulting in an opaque oil which solidified at 25 °C. The volatiles were
sublimed into a cold trap, leaving a white, nonvolatile solid (294 mg),
with the physical and spectroscopic properties of 2. IR (KBr): 3316
(w), 3258 (m), 1880 (s), 956 (s). The crystalline solid collected in the
cold trap (68 mg, 23%) was determined to be H3Ga(NMe3). IR
(benzene): νGa-H 1831 (independently prepared sample of H3Ga(NMe3)
1831; lit.12 1832, 1852).
Reaction of [H2GaNH2]3 with Tricyclohexylphosphine. To a flask
charged with 1 (1.52 mmol, 400 mg) and tricyclohexylphosphine (4.56
mmol, 1.278 g) was added THF (35 mL). The slightly hazy solution
was stirred at 25 °C for 1 h and an aliquot (0.4 mL) was removed and
its IR spectrum recorded (νGa-H 1848, 1818 cm-1). The mixture was
stirred for 48 h, resulting in a colorless solution and white precipitate.
The mixture was filtered through a fine frit, yielding a white solid and
a clear filtrate. The filter cake was washed with an additional 10 mL
of THF and dried under vacuum, yielding 2 as a white solid (145 mg,
75%). IR (KBr): 3327 (w), 3279 (m), 1880 (s), 958 (s), 905 (s). The
volatiles were removed from the filtrate under reduced pressure,
resulting in the isolation of a white crystalline solid which was identified
as a mixture of PCy3 and H3Ga(PCy3) (1.334 g, 92% mass recovery).
IR (KBr) νGa-H 1844 (m), 1825 (s), 1801 (s) (independently prepared
sample of H3Ga(PCy3) 1846, 1824, 1801; lit.11 1800 (br) (no medium
given). 1H NMR (C6D6) δGa-H 4.26 (br s) (independently prepared
sample of H3Ga(PCy3) 4.26 (br s); lit.11 4.32 (br s)).
X-ray Diffraction Studies. XRD data were collected on a Siemens
D5005 diffractometer or a Siemens Microdiffractometer equipped with
a Hi-Star 2-D area detector using monochromatic (graphite) Cu KR
radiation. Samples of GaN were prepared as powders dispersed on glass
slides (Siemens D5005), while powdered samples of 2 were sealed
under nitrogen in glass capillaries (Siemens Micro-diffractometer). The
coherent lengths of the samples were estimated using the Scherrer
equation and the Jade 3.0 software package.
Transmission Electron Microscopy and Electron Diffraction
Studies. TEM images and electron diffraction patterns were obtained
on a Philips CM30 TEM operating at 300 kV. Samples were prepared
by sprinkling the powder over lacy carbon films supported on Cu mesh
grids. Particles overhanging the carbon support were used for analysis.
Despite the use of a liquid nitrogen cooled sample stage, the electron
beam degraded the samples of 2, as evidenced by the rapid decay of
the diffraction patterns.
Results
Synthesis and Characterization of Poly(imidogallane).
Cyclotrigallazane ([H2GaNH2]3, 1) was found to be soluble in
liquid ammonia. Heating an ammonia solution of [H2GaNH2]3
at 150 °C for 30 min in an autoclave led to the isolation of a
slightly off-white solid following removal of the ammonia. This
solid was found to be involatile and insoluble in all solvents
(including NH3) except for strong mineral acids with which it
reacted with gas evolution. The solid was also found to be
relatively stable with regard to attack from atmospheric oxygen
and moisture.
The IR spectrum of this solid (Figure 1) was simplified
compared to that of 1 and suggested that the product was a
hydrido gallium imide of the formula HGaNH (2). The single
peak at 3280 cm-1 was assigned to an imide (NH) rather than
an amide (NH2) group. The Ga-H vibration appeared as a
strong, broad peak at 1883 cm-1 and a sharper, less intense
band at 1818 cm-1. The peaks at 962 and 906 cm-1 were
attributed to the wagging motion of the N-H. Notably absent
was any peak in the region around 1510 cm-1 due to the
scissoring vibration of NH2 groups.
The formulation of this solid as a hydrido gallium imide was
supported by elemental analysis. Drying a sample of 2 under
dynamic vacuum for 24 h resulted in the isolation of a solid
which gave satisfactory elemental analyses and H:N:Ga ratios
of 1.99:1.02:1. Analysis of a sample prior to drying showed
that it was rich in nitrogen (calcd 16.34% for pure [HGaNH]n;
(12) Shriver, D. F.; Amster, R. L.; Taylor, R. C. J. Am Chem. Soc. 1962,
84, 1321.

2728 Inorganic Chemistry, Vol. 38, No. 11, 1999
Jegier et al.
Figure 1. Infrared spectra (KBr pellet) of (a) [H2GaNH2]3 (1) and (b)
[HGaNH]n (2).
Figure 2. X-ray powder diffraction pattern for [HGaNH]n (2) produced
from (a) the thermolysis of 1 in NH3 and (b) the reaction of 1 and
PCy3 in THF.
found 18.13%) and consequently gallium poor. This was
attributed to the presence of approximately 10% residual NH3
entrained in the solid, HGaNH‚0.1NH3. Reports of the analogous
aluminum compound, [HAlNH]n, suggested that as-prepared
samples “tenaciously” entrained solvent.13
These results were most consistent with the formulation of
the product as HGaNH. The poor solubility and involatility of
this solid led to the conclusion that 2 was most likely poly-
(imidogallane), [HGaNH]n. Reports of the analogous aluminum
compound suggested that [HAlNH]n was also polymeric.13,14
Conversion of [H2GaNH2]3 to GaN in Ammonia. Reaction
times as short as 10 min to as long as 1.5 h at 150 °C resulted
in the isolation of 2 as the only product. Lower temperatures
and times (i.e., 10 min at 100 °C) led to an incomplete reaction
and gray powders indicative of the decomposition of 1. Poly-
(imidogallane) was also the only product isolated at a variety
of pressures ranging from 8.4 to 71 atm. The products isolated
from higher reaction pressures ranging from 112 to 124 atm
exhibited shifts in νN-H from 3279 to 3254 cm-1 and νGa-H
from 1883 to 1869 cm-1 as well as a coalescence of the peaks
at 962 and 906 cm-1 to a single peak at 955 cm-1. These shifts
suggested that 2 underwent further reaction. Extended reaction
(5 h) under these conditions resulted in GaN powders with an
average coherent length of 3 nm.
Figure 3. TEM and electron diffraction (insert) of poly(imidogallane)
(2). The tabular morphology is evident by the characteristic image
density present over large regions of the particle, which is comprised
of many small crystallites (not individually visible.)
Reaction of [H2GaNH2]3 with ND3. Dissolution of 1 in ND3
followed by slow evaporation of the solvent resulted in the
isolation of a white, crystalline powder which was identical in
appearance to the starting material. However, the IR spectrum
(Figure 4) of this solid indicated >95% conversion of 1 to [H2-
GaND2]3 (3) and lesser amounts of 1 and [H2GaN(H)D]3 (4).
The IR spectrum of 3 exhibited N-D vibrations at 2470, 2460,
2391, and 2385 cm-1, while the spectrum of 4 consisted of N-H
vibrations at 3286 and 3272 cm-1 and N-D vibrations at 2428
and 2420 cm-1. The νGa-H region of the spectrum remained
unchanged. The assignment of these bands compared well those
lazane also reacted with neat THF and trimethylamine and with
quinuclidine and tricyclohexylphosphine in THF solution. These
reactions did not yield the expected H2GaNH2‚base complexes,
but resulted in the isolation of poly(imidogallane) and Lewis
base adducts of gallane, H3Ga‚base. The poly(imidogallane)
products were isolated in good yields (75 and 90%, respectively,
for the reactions with PCy3 and quinuclidine) based on the
stoichiometry discussed below. A greater than theoretical yield
of 2 was obtained for the reaction of 1 with NMe3. This was
believed to be due to the presence of excess NMe3 entrained in
the product. A low (approximately 20%) yield of 2 was obtained
from the reaction of 1 with THF after 5 days.
4
of the fully deuterated analogue [D2GaND2]3 as well as with
the vibrational energies predicted by theory.4,15
Interaction of [H2GaNH2]3 with Various Lewis Bases. [H2-
GaNH2]3 was found to exhibit moderate to good solubility in
other polar solvents such as THF, trimethylamine, and diethyl-
amine, but little or no solubility in diethyl ether. Cyclotrigal-
The gallane-Lewis base adducts were also isolated in high
yield (82% for H3Ga‚quinuclidine). Due to difficulties encoun-
tered in separating PCy3 and H3Ga‚PCy3, no yield could be
determined for this reaction. However, the total mass recovered
was 92% of that expected for the mixture of the two compounds.
The low isolated yield of H3Ga‚NMe3 was due to its high
volatility, resulting in the substantial loss of product during
(13) Janik, J. F.; Paine, R. T. J. Organomet. Chem. 1993, 449, 39.
(14) Wiberg, E.; May, A. Z. Naturforsch. B 1955, 10, 229.
(15) Cramer, C. J. Unpublished results.

Solution Chemistry of Cyclotrigallazane
Inorganic Chemistry, Vol. 38, No. 11, 1999 2729
Table 1. X-ray Diffraction Data for [HGaNH]n Prepared from
[H2GaNH2]3 and PCy3
expl 2θ (deg) d (Å)
I/Io
tentative hkl calcd 2θ difference
2.893
5.369
7.103
8.285
9.523
10.846
11.838
14.046
24.504
25.721
28.400
37.919
30.52
16.45
12.44
10.66
9.280
8.150
7.469
88.2
4.9
0.7
1.3
1.4
0.8
0.6
001
002
2.710
5.415
-0.1861
0.0464
003
004
8.127
-0.1581
-0.0030
10.843
6.265 100
100
110
113
200
210
14.050
24.460
25.814
28.317
37.760
0.0041
-0.0441
0.0935
-0.0828
-0.1594
3.630
3.445
3.140
2.371
3.1
0.5
4.9
3.1
prepared from the reaction of 1 with quinuclidine was identical
to that seen in Figure 2b. This eliminated the possibility that
any of the reflections were due to residual PCy3 or H3Ga‚PCy3.
On the basis of the line broadening, the coherent length of the
crystallites was 150 Å. Several of the reflections exactly
coincided in both position and relative intensity for both
samples; however, there were striking differences below 2θ
values of 14°. In Figure 2a a broad maximum is centered at 3°,
while an intense, well-defined reflection is observed in Figure
2b at 2.8°. Table 1 summarizes the results shown in Figure 2b.
Due to the small number and breadth of the reflections, it was
not possible to assign a crystal system and unit cell unambigu-
ously. Analysis of the d spacings suggested that the four, well-
defined reflections above 14° fit a pattern consistent with a
hexagonal cell. Unfortunately, none of these reflections con-
tained a component along the c-axis, so the only unit cell
dimension obtained was limited to the crystallographic a-axis
(7.27 Å). If the intense reflection at 2.8° was assigned to 001,
most of the additional reflections below 14° could be assigned
as 002 through 004. These values were consistent with a c-axis
of 32.6 Å. Using these two cell constants, the reflection located
at 25.7° was assigned to 113. In Figure 2a the breadth of the
low-angle reflections may be attributed to increased disorder
along the c-axis caused by incorporation of trace quantities of
NH3 between the layers.
Figure 4. Infrared spectra of (a) [H2GaNH2]3 (1) and (b) [H2GaND2]3
(3). Spectrum b indicates the presence of small amounts (approximately
5%) of 1 and [H2GaN(H)D]3 (4).
evaporation of the excess NMe3. No attempt was made to isolate
the presumed H3Ga‚THF.
Spectroscopic characterization of the gallane-base adducts
compared well with independently prepared samples and with
literature values (see Experimental Section).9-12 The IR spectra
of the poly(imidogallane) prepared by these reactions were
essentially identical to those of samples prepared in NH3.
However, there was a weak band at approximately 3325 cm-1
which was not present in samples of 2 prepared from ammonia
solution. The strong Ga-H vibration was generally shifted from
1883 to 1880 cm-1 and the weak band at 1818 cm-1 was absent.
The greatest deviation was seen in the sample prepared from
NMe3. The N-H vibration was shifted by approximately 20
cm-1 to 3258 cm-1. The Ga-H vibration was shifted to 1869
cm-1 in the as-prepared sample. This was believed to be due to
the presence of excess NMe3 entrained in the product, resulting
in the isolation of a greater than theoretical yield of poly-
(imidogallane). Drying a small sample of this solid under
dynamic vacuum for 24 h led to a shift in νGa-H from 1869 to
1879 cm-1, but no change in νN-H
.
This assignment was supported by TEM analysis of 2.
Detailed analyses of the particles was not possible because the
sample rapidly degraded in the electron beam, as evidenced by
the rapid decay of the diffraction patterns. However, bright field
images of the particles showed a broadly tabular morphology
(Figure 3). The sharp diffraction rings frequently showed strong
intensity only over small arcs, rather than being uniformly
intense around the whole ring. A few samples did show
complete rings, with intensity strongly peaked every 60° around
the rings. Within the rings there was a streaking of the diffraction
spots around the ring, which suggested that within the platelets
there were areas which were rotationally misaligned. There was
also an appreciable intensity peak corresponding to a rotation
of 90° within the plane. These observations suggested that the
particles were planar agglomerations, with most of the particles
oriented with the basal plane normal to the electron beam (to
give complete rings) or slightly off normal (to give the arc
patterns). The particles were aligned reasonably well within the
layer, with a small fraction aligned 90° out of phase. The intense
peaks were indexed as a hexagonal (or pseudohexagonal) lattice
with a plane spacing of 6.4 Å and a unit cell dimension within
the plane of 7.5 Å, which was entirely consistent with the X-ray
data. No data could be determined for the unit cell dimensions
normal to the plane.
In the reactions of 1 with THF, quinuclidine, and PCy3,
aliquots were removed from the reaction mixtures and the
progress of the reactions was monitored by IR spectroscopy.
In THF solution, 1 exhibited νGa-H at 1858 cm-1. Aliquots were
taken over a period of several days. The initially clear solution
grew turbid and a white solid began to precipitate as the peak
at 1858 cm-1 diminished and a new peak at 1817 cm-1
appeared. When three equivalents of quinuclidine were added
to a THF solution of 1, the IR spectrum consisted of a single
absorption at 1821 cm-1 before the onset of precipitation. The
reaction of 1 with PCy3 in THF appeared to be the intermediate
case. The IR spectrum of an aliquot of the slightly hazy reaction
mixture after 1 h displayed vibrations at 1848 and 1818 cm-1
.
The νN-H region of all of these spectra were similar and
exhibited broad bands at 3238 cm-1 with less intense shoulders
ranging from 3300 to 3312 cm-1
.
X-ray and Electron Diffraction Studies on 2. Figure 2
shows the X-ray powder diffraction patterns of samples of 2
prepared by the reaction of [H2GaNH2]3 in supercritical am-
monia (2a) and by the reaction of [H2GaNH2]3 with PCy3 in
THF (2b). Measurement of the pattern of an empty glass
capillary tube verified that the intensity of background reflec-
tions was insignificant. The XRD obtained from a sample of 2

2730 Inorganic Chemistry, Vol. 38, No. 11, 1999
Jegier et al.
Scheme 1
Discussion
inconsistent with a layered structure, as shown in Scheme 1e,
it has not been possible to reconcile the lattice constants with
the proposed molecular structure. Such an arrangement is related
to the proposed structure of carbon monofluoride, (CF)n,
prepared by the direct fluorination of graphite.28,29 Taking a 001
slice of hexagonal GaN as a model layer to compare with 2,
we would predict the centerpoint distance between adjacent six-
membered rings of a hexagonal net would be 3.18 Å; a value
somewhat less than half of the in-plane lattice constant (7.27
Å) found in 2. Assuming that the crystal structure requires a
doubling of this distance, the agreement between 6.36 and 7.27
Å is still unsatisfactory. Considering that the Ga-N bond
distance in GaN is nearly identical to the corresponding distances
in [H2GaNH2]3 and related compounds,30-32 a 14% lengthening
of the lattice constant seems unrealistic. It is also hard to
rationalize the length of the c-axis based upon a structure with
stacked layers, as shown in Scheme 1e; an estimated interlayer
spacing would be approximately 5.2 Å. The observed lattice
constant (32.6 Å) would require approximately six such layers
within each unit cell. One possible explanation is that [HGaNH]n
formed from the PCy3-induced reaction crystallizes in a
particular polytype that exhibits an elongated c-axis. This is
observed in other layered structures including CdI2 and various
micas.33
One of the attractive aspects about the planar model is that
the long-range ordering can apparently be disrupted by the
intercalation of small molecules such as NH3. It is believed that
the additional peaks observed in the X-ray diffraction pattern
for samples of 2 prepared from the reactions of 1 with PCy3
and quinuclidine are due to increased ordering in the alignment
of the two-dimensional sheets (the crystallographic c direction).
This ordering is disrupted in the samples of 2 prepared from
ammonia, resulting in only a broad diffraction peak in this
region. This is supported by the observation that samples of 2
prepared in ammonia were nitrogen rich and had to be subjected
to vacuum before correct elemental analyses could be obtained.
What alternatives exist? A molecular cluster that is large
enough to be insoluble or possibly a more complex (perhaps
microporous) solid-state structure. In either case we can say
little about the molecular structure. We conclude that despite
the problems, the best candidate for the structure of poly-
(imidogallane) is the planar hexagonal network.
Poly(imidogallane). An attempt to synthesize GaN in su-
percritical ammonia led to the isolation of hydridogallium imide.
The spectroscopic and elemental analytical data and the physical
characteristics of the solid are consistent with a polymeric
species having the formula [HGaNH]n. The polymeric nature
of 2 is consistent with other related group 13 imides and amide
imides such as [HAlNH]n,13,14 [Ga(NH)3/2]n,16 and [Al(NH2)-
(NH)]n.17,18
Poly(imidogallane) was found to be crystalline by both X-ray
and electron diffraction methods. There is precedent for forming
crystalline polymeric precursors to solid-state materials. Poly-
(borazylene), which contains linked borazine rings, exhibits
some long-range ordering and a diffraction pattern similar to
those observed for turbostratic boron nitride.19 Reports of
[HAlNH]n have not suggested any evidence of a crystalline
nature.13,14
The simplicity of the spectroscopic and diffraction data
suggests that 2 is composed of one discreet compound, and
Scheme 1 depicts several possible molecular and solid-state
structures. Structures a-d are based upon related, structurally
characterized organoaluminum and organogallium imides.20-27
Six-membered rings containing the heavier group 13 elements,
such as the type depicted in structure 1a, have only been isolated
when bulky substituents have been utilized.26 While structures
b-d cannot be eliminated from consideration, we suggest that
a two-dimensional solid-state network such as shown in Scheme
1e is most consistent with the volatility and solubility of 2.
The connection between the crystal and molecular structure
of 2 is complicated by the uncertainty in the crystal structure,
the assignment of the observed reflections, and, therefore, the
true lattice constants. While the XRD and ED results are not
(16) Janik, J. F.; Wells, R. L. Chem. Mater. 1996, 8, 2708.
(17) Wiberg, E.; May, A. Z. Naturforsch. B 1955, 10, 230.
(18) Maya, L. AdV. Ceram. Mat. 1986, 1, 150.
(19) Fazen, P. J.; Remsen, E. E.; Beck, J. S.; Carroll, P. J.; McGhie, A.
R.; Sneddon, L. G. Chem. Mater. 1995, 7, 1942.
(20) Cucinella, S.; Salvatori, T.; Busetto, C.; Perego, G.; Mazzei, A. J.
Organomet. Chem. 1974, 78, 185.
(21) Cesari, M.; Perego, G.; Del Piero, G.; Cucinella, S.; Cernia, E. J.
Organomet. Chem. 1974, 78, 203.
(22) Hitchcock, P. B.; Smith, J. D.; Thomas, K. M. J. Chem. Soc., Dalton
Trans. 1976, 1433.
(23) Del Piero, G.; Cesari, M.; Dozzi, G.; Massei, A. J. Organomet. Chem.
1977, 129, 281.
(24) Del Piero, G.; Cesari, M.; Perego, G.; Dozzi, G.; Cucinella, S.; Cernia,
E. J. Organomet. Chem. 1977, 129, 289.
(25) Amirkhalili, S.; Hitchcock, P. B.; Smith, J. D. J. Chem. Soc., Dalton
Trans. 1979, 1206.
(26) Waggoner, K. M.; Hope, H.; Power, P. P. Angew. Chem., Int. Ed.
Eng. 1988, 27, 1699.
(28) Rudorff, V. W.; Rudorff, G. Z. Anorg. Allg. Chem. 1947, 253, 281.
(29) Palin, D. E.; Wadsworth, K. D. Nature 1948, 162, 925.
(30) Harrison, W.; Storr, A.; Trotter, J. J. Chem. Soc. A 1966, 1554.
(31) Greenwood, N.; Ross, E.; Storr, A. J. Chem. Soc. A 1966, 706.
(32) Rendle, D.; Storr, A.; Trotter, J. J. Chem. Soc., Dalton Trans. 1973,
2252.
(27) Belgardt, T.; Roesky, H. W.; Notlemeyer, M.; Schmidt, H.-G. Angew.
Chem., Int. Ed. Eng. 1993, 32, 1056.
(33) Verma, A. R.; Krishna, P Polymorphism and Polytypism in Crystals;
John Wiley and Sons: New York, 1966; Chapter 4.

Solution Chemistry of Cyclotrigallazane
Inorganic Chemistry, Vol. 38, No. 11, 1999 2731
Scheme 2
equilibrium with the gallane base adduct H3Ga‚base and a
proposed diamidogallane adduct, HGa(NH2)2‚base. The latter
compound appears to be unstable with regard to the loss of
ammonia and Lewis base, yielding poly(imidogallane). This last
step is related to the transamination mechanism proposed for
the isotope exchange experiment described above. The loss of
NH3 from complexes containing multiple NH2 groups to form
metal amide imides such as [M(NH2)(NH)]n (M ) B, Al)14,17,34,35
16
and metal imides such as [Ga(NH)3/2 n
]
is well-documented.
The decomposition of the diamidogallane adduct, HGa(NH2)2‚
base, would shift the equilibrium to the right, resulting in the
isolation of H3Ga‚base and poly(imidogallane) as the only
products. Thus the overall reaction for this process can be written
as
The Nature of [H2GaNH2]3 in Ammonia Solution. There
was a large difference in the reactivity of cyclotrigallazane in
the solid state and in ammonia solution. Whereas heating an
ammonia solution of 1 at 150 °C for 30 min produced white,
crystalline poly(imidogallane) 2, heating 1 at 150 °C for the
same duration in the solid state resulted in the isolation of an
amorphous gray-black solid which contained only small peaks
due to N-H and Ga-H vibrations in its IR spectrum. The gray
color, presumably due to the formation of small amounts of
Ga0, and lack of crystallinity are typical of the solid-state
conversion of 1 to GaN at low temperatures.2,3 This difference
in reactivity may be due to a difference in the degree of
association between the two states.
To gain insight into the degree of association of cyclotrigal-
lazane in ammonia solution, [H2GaNH2]3 was dissolved in ND3,
resulting in a nearly quantitative conversion to [H2GaND2]3.
The H/D exchange in this transamination reaction most likely
occurred directly on the donor-acceptor complex H2GaNH2‚
ND3 (Scheme 2). Similar transamination reactions are common
for the group 13 metals.16,34,35 This result suggests that in
ammonia solution, [H2GaNH2]3 exists as (or is in equilibrium
with) the monomeric ammonia adduct of gallazane, H2GaNH2‚
NH3.
The Interaction of [H2GaNH2]3 with Other Lewis Bases.
While the ammonia adduct of gallazane, H2GaNH2‚NH3, was
not isolable, an attempt was made to isolate various other
donor-acceptor complexes of the general formula H2GaNH2‚
base. A summary of these reactions and their products can be
seen in Scheme 3. In all cases, the reaction of 1 with a Lewis
base led to the isolation of poly(imidogallane) (2) and Lewis
base adducts of gallane (H3Ga‚base) rather than to H2GaNH2‚
base. It is interesting to note that H3Ga‚NMe3 is isolated from
the reaction of cyclotrigallazane and NMe3. This is the reverse
of the reaction used to synthesize [H2GaNH2]3, where H3Ga‚
NMe3 is reacted with ammonia.4 The inability to isolate the
species H2GaNH2‚base is surprising in light of the isolation and
stability of the related chlorogallane adducts, HnGaCl3-n‚base
(where n ) 1, 2 and base ) quinuclidine, PCy3).36,37
(1)
In an attempt to observe one or more of the proposed
intermediates, the progress of several of the above reactions
was monitored by IR spectroscopy. In the reaction of [H2-
GaNH2]3 with THF, the νGa-H region of the spectrum exhibits
a single peak at 1858 cm-1. This peak could be due to H2-
GaNH2‚THF or intact [H2GaNH2]3. Over the course of several
days, the peak at 1858 cm-1 slowly disappeared and was
replaced by a new peak at 1818 cm-1 as the solid began to
precipitate.
The addition of 3 equiv of quinuclidine to a solution of [H2-
GaNH2]3 in THF resulted in a single νGa-H at 1821 cm-1. This
represents a shift of nearly 40 cm-1 from that of [H2GaNH2]3
in THF. While it is tempting to assign this peak as H2GaNH2‚
quinuclidine, its similarity to the peak eventually observed for
the reaction of [H2GaNH2]3 with THF, as well as the peak
observed at 1818 cm-1 for the reaction of [H2GaNH2]3 with
PCy3 in THF, would suggest it is due to a common as yet
unidentified species. This assertion is supported by the nearly
identical N-H vibrations observed for all three reactions.
Further studies are aimed at identifying this unknown.
There was no evidence for the presence of a species such as
Ga(NH2)3‚base or [Ga(NH2)3]n in these reactions. Wells and
Janik attempted to synthesize [Ga(NH2)3]n from [Ga(NMe2)3]2
and NH3. However, this reaction produced the imide [Ga(NH)3/2]n,
which possessed a strong νN-H at 3150 cm-1 16
. This vibration
is not present in any of the spectra of our samples, and is not
the source of the N-H vibration at approximately 3325 cm-1
,
which is seen in the spectra of poly(imidogallane).
The above results suggest that the formation of 2 in ammonia
occurs via the elimination of H2 and NH3. The mechanism of
this ammonia-assisted dehydrogenation can be seen in Scheme
5. The first step involves cleavage of the cyclotrigallazane ring
to yield the transient ammonia adduct of gallazane, H2GaNH2‚
NH3. This species equilibrates with H3Ga‚NH3 and HGa(NH2)2‚
NH3, and the diamidogallane-ammonia adduct eventually
eliminates two molecules of NH3 to yield poly(imidogallane).
The gallane-ammonia adduct, H3Ga‚NH3, which is presumed
to be present in the initial synthesis of cyclotrigallazane,
eliminates H2 to form H2GaNH2‚NH3. This cycle continues until
all of the gallazane is consumed. The overall equation for this
process is shown in eq 2, which emphasizes the catalytic role
of NH3.
We propose a possible mechanism to account for the isolation
of these products (Scheme 4), beginning with cleavage of the
cyclotrigallazane ring to form the desired gallazane adduct H2-
GaNH2‚base. This adduct could participate in a Schlenk-type
(34) Kimura, Y.; Kubo, Y.; Hayashi, N. J. Inorg. Organomet. Polym. 1992,
2, 231.
(35) Janik, J. F.; Wells, R. L.; Coffer, J. L.; St. John, J. V.; Pennington,
W. T.; Schimek, G. L. Chem. Mater. 1998, 10, 1613.
(36) Elms, F. M.; Koutsantonis, G. A.; Raston, C. L. J. Chem. Soc. Chem.
Commun. 1995, 1669.
(37) Luo, B.; Young, V. G., Jr.; Gladfelter, W. L. J. Chem. Soc. Chem.
Commun. 1999, 123.
(2)
Poly(imidogallane) appears to be an intermediate in the
conversion of cyclotrigallazane to GaN in ammonia solution.

2732 Inorganic Chemistry, Vol. 38, No. 11, 1999
Jegier et al.
Scheme 3
Scheme 4
Scheme 5
In various syntheses of 2, increasing the reactor pressure leads
to changes in the IR spectra of the isolated solids, indicative of
2 undergoing further reaction. Increases in reaction time (5 h)
and reactor pressure result in the complete conversion to
nanocrystalline GaN powders (coherent length ) 3 nm).
Increasing the reaction time to 24 h resulted in larger (4 nm)
coherent lengths. While the above results suggest that 1
dissociates into monomers in liquid NH3, it is unlikely that such
a reaction would be important for the conversion of 2 to GaN.
This is supported by the lack of solubility of 2 in NH3 and the

Solution Chemistry of Cyclotrigallazane
Inorganic Chemistry, Vol. 38, No. 11, 1999 2733
observation that 2 precipitates from solution when 1 is reacted
with quinuclidine in liquid NH3. Details of the syntheses of GaN
powders from the thermolysis of cyclotrigallazane in super-
critical ammonia and a study of the effect of temperature and
pressure on this conversion will be described in future publica-
tions.
elimination of NH3. This chemistry has been exploited to
synthesize poly(imidogallane) in high yield in liquid NH3, where
the NH3 acts as both solvent and catalyst. Poly(imidogallane)
is an intermediate in the conversion of cyclotrigallazane to GaN
in supercritical NH3. This new, low-temperature route to
nanocrystalline GaN will be discussed in future publications.
Acknowledgment. The authors gratefully acknowledge
support from the National Science Foundation (CHE-9616501).
The authors would also like to thank Prof. Chris Cramer for
normal mode calculations of 4 and Linda Sauer for assistance
in obtaining X-ray diffraction patterns of 2.
Conclusions
The reaction of [H2GaNH2]3 with a variety of Lewis bases
results in the isolation of gallane-Lewis base adducts and the
novel hydrido gallium imide poly(imidogallane). Isolation of
these products is due to the presence of the NH2 group, which
undergoes ligand redistribution reactions and the eventual
IC981298L