Characterization of metallurgical-grade aluminas and their precursors by27Al NMR and XRD

Article abstract of DOI:10.1139/V07-106

The structure of metallurgical- or smelter-grade aluminas (MGAs) is complex and poorly understood. Ultra-high-field solid-state ²⁷Al NMR results on industrial as well as on laboratory-prepared aluminas are discussed in relation to XRD results. It is demonstrated that high-field NMR can effectively be used to quantify the proportion of the thermodynamically stable alpha-alumina phase in these materials. The results demonstrate that ²⁷Al NMR is a vital adjunct to XRD methods to quantify the transition alumina phases that invariably dominate the MGAs. The nature of the disorder in these materials, determined by ²⁷Al NMR, is also compared with literature data, such as XANES and EXAFS studies, on typical laboratory-prepared materials. The utility of ²⁷Al NMR studies to provide new insight into the structural complexity of metallurgical aluminas is shown.

Full text of DOI:10.1139/V07-106

8
89
Characterization of metallurgical-grade aluminas
27
1
and their precursors by Al NMR and XRD
Linus M. Perander, Zoran D. Zujovic, Tania Groutso, Margaret M. Hyland,
Mark E. Smith, Luke A. O’Dell, and James B. Metson
Abstract: The structure of metallurgical- or smelter-grade aluminas (MGAs) is complex and poorly understood. Ultra-
high-field solid-state 27Al NMR results on industrial as well as on laboratory-prepared aluminas are discussed in rela-
tion to XRD results. It is demonstrated that high-field NMR can effectively be used to quantify the proportion of the
2
7
thermodynamically stable alpha-alumina phase in these materials. The results demonstrate that Al NMR is a vital ad-
junct to XRD methods to quantify the transition alumina phases that invariably dominate the MGAs. The nature of the
2
7
disorder in these materials, determined by Al NMR, is also compared with literature data, such as XANES and
2
7
EXAFS studies, on typical laboratory-prepared materials. The utility of Al NMR studies to provide new insight into
the structural complexity of metallurgical aluminas is shown.
Key words: solid-state magic-angle-spinning NMR, metallurgical-grade alumina, transition alumina, gamma-alumina,
coordination number.
Résumé : La structure des alumines de qualité métallurgique (AQM) ou pour électrolyse est complexe et mal
2
7
comprise. Opérant à champ élevé et à l’état solide, on a déterminé les spectres RMN du Al d’alumines industrielles
et d’autres préparées en laboratoire et on discute de ces spectres en relation avec les résultats obtenus par diffraction
des rayons X (DRX). On a démontré que la RMN à champ élevé peut effectivement être utilisée pour quantifier la
proportion de la phase thermodynamiquement stable d’alumine alpha qui est présente dans ces matériaux. Les résultats
démontrent que la RMN du 27Al est une addition absolument nécessaire aux méthodes de DRX pour quantifier les
phases d’alumines de transition qui dominent invariablement les AQM. On a aussi comparé la nature du désordre dans
2
7
ces matériaux, telle que déterminée par RMN du Al, avec les résultats déjà publiés dans la littérature et obtenus par
des études à l’aide de la spectroscopie d’absorption X de type XANES ou EXAFS sur des matériaux typiques préparés
2
7
en laboratoire. On démontre donc l’utilité de la RMN du Al à fournir de nouvelles indications sur la complexité
structurale des alumines métallurgiques.
Mots-clés : RMN à l’état solide et à l’angle magique, alumine de qualité métallurgique, alumine de transition, alumine
gamma, indice de coordination de l’aluminium.
[
Traduit par la Rédaction] Perander et al. 897
Introduction
able interest for a number of reasons; there has, for instance,
been interest in the relationship between properties and the
behavior of the alumina when used in gas scrubbing systems
and when fed to the reduction cell (1, 2). Furthermore, this
material can also provide insights into the complex relation-
ship within the family of the so-called “transition aluminas”
Over 50 million tonnes of metallurgical-grade alumina
MGA) is produced annually, largely as a raw material for
(
the electrolytic production of aluminum. This material is
produced almost exclusively by the Bayer process to a rea-
sonably uniform specification across the industry, in terms
of specific surface area, particle size and strength, residual
hydrogen or moisture as well as impurity and alpha-alumina
contents. The microstructure of this material is of consider-
produced in the calcination of gibbsite Al(OH) (3).
3
In the Bayer process, the calcination step plays an impor-
tant role in the evolution of microstructural features that can
be related to the properties and the performance of the MGA
Received 28 June 2007. Accepted 21 August 2007. Published on the NRC Research Press Web site at canjchem.nrc.ca on
October 2007.
2
L.M. Perander, Z.D. Zujovic, and T. Groutso. Department of Chemistry, University of Auckland, Private Bag 90219, Auckland,
New Zealand.
M.M. Hyland. Department of Chemical and Materials Engineering, The University of Auckland, Private Bag 92019, Auckland,
New Zealand.
M.E. Smith and L.A. O’Dell. Department of Physics, University of Warwick, Coventry CV4 7AL, UK.
2
J.B. Metson. Department of Chemistry, University of Auckland, Private Bag 90219, Auckland, New Zealand; MacDiarmid
Institute for Advanced Materials and Nanotechnology.
1
This article is part of a Special Issue dedicated to Professor G. Michael Bancroft.
Corresponding author (e-mail: j.metson@auckland.ac.nz).
2
Can. J. Chem. 85: 889–897 (2007)
doi:10.1139/V07-106
© 2007 NRC Canada

8
90
Can. J. Chem. Vol. 85, 2007
in the aluminum smelter. For the calcination step, produc-
tion technology has progressively moved from rotary kilns
to the more energy-efficient gas suspension or fluid-bed-
type rapid-flash calciners, in some cases with capacity in ex-
cess of 4 000 tonnes per day (4). Due to the extended resi-
dence time, rotary kilns produce a more-homogenous and
structurally ordered material. However, the literature and the
understanding of the processes operating under flash-
calcination conditions are lacking when compared, for ex-
ample, with the typical gamma-aluminas widely used in ca-
talysis.
Fig. 1. Scanning electron micrograph of “typical” metallurgical-
grade alumina particles.
The most-authoritative description of the interrelationship
of the alumina phases or “forms” during calcination pro-
cesses is that of Wefers and Misra (3). The speed of the
modern calcination process, together with the variation of
particle size, velocity, and residence time, produces a far-
from-equilibrium phase distribution both between and within
individual alumina grains. MGAs can thus be viewed as
granular materials comprised of aggregates of intergrown
alumina crystals; generally, with a crystallite size of 10–
4
0 µm. Figure 1 shows a scanning electron micrograph of a
typical assembly of MGA particles. The precipitation condi-
tions of the precursor gibbsite are selected to maximize yield
in the critical size range, thus causing the dominance of
heavily intergrown aggregates.
The disorder and complexity of MGA is influenced by the
calcination process as well as the gibbsite precipitation
stage. The calcination strategy is governed first and foremost
by the need to meet a specification for a specific surface
area of 70–80 m g , while maintaining a low residual-
moisture content, necessary to limit HF generation when fed
into the molten cryolite electrolyte in the reduction cell. Sur-
face area is also critical for effective performance as a gas
scrubbing medium and for optimal dissolution in the reduc-
tion cell. Surface area is maximized by controlling the tran-
sition sequence to limit maximum temperature and residence
time. With increasing temperature, the phase sequence
Further heating results in the conversion to delta- and then
to theta-alumina, which are also spinel-structured transition-
alumina forms. Gamma-, delta- and theta-alumina all have a
face-centred cubic packing of the oxygen anions with vari-
ous amounts of aluminum cations in either the octahedral or
tetrahedral interstices. The proportion of tetrahedral sites in-
creases with temperature (10), having 25%–30% of the
octahedral sites occupied for gamma-alumina and 50% for
theta-alumina (11, 12), until the abrupt disappearance of
tetrahedrally coordinated aluminum upon the formation of
alpha-alumina (corundum), which marks the end of the
transformation sequence. Alpha-alumina has a hexagonal
close packing of the oxygen anions, and all of the octahedral
aluminum sites are occupied (13).
The mix of “forms” and high levels of disorder in MGAs
present significant challenges in understanding their struc-
ture with relation to the performance as a raw-material feed
for the smelting process (14). The X-ray methods by which
we might follow these phase transformations are constrained
by the limited long-range order in these materials. Thus,
methods, which do not rely on long-range crystallographic
order, can provide new insights and improved understanding
of these materials. Kato et al. have reported X-ray absorp-
tion near-edge studies (XANES) of the Al K-edge in transi-
tion aluminas (15). In the manner of NMR studies, this
approach appears to provide significant insight into the oc-
cupancy of octahedral and tetrahedral sites, i.e., key to un-
derstanding the forms and transition points in the alumina
samples examined. However, the XANES approach has yet
to be applied to MGAs.
2
–1
moves from the hexagonal gibbsite, Al(OH) , through the
3
oxyhydroxide boehmite, AlOOH, into the cubic system rep-
resented by the spinel structured gamma-alumina (5)
(
Fig. 2). Neutron diffraction studies indicate that the face-
centred cubic oxygen sublattice in the spinel-based transition
aluminas tends to be highly ordered in these materials; how-
ever, the aluminum ions are quite disordered, leading to the
broadened lines typical of their powder X-ray diffraction
(
XRD) patterns (6).
In gibbsite, aluminum atoms occupy two thirds of the oc-
tahedral positions, with hydroxyl groups octahedrally coor-
dinated in double layers. This gives gibbsite a monoclinic
symmetry (space group P2 /n) with a hexagonal close pack-
1
ing of the hydroxyl groups. Upon heating, the hydroxyls are
progressively lost through the structural channels formed in
the gibbsite lattice by the formation of water by proton ex-
change between neighbouring hydroxyls (7–9). This is ac-
companied by the structural rearrangement of the crystal
lattice, a pseudomorphic transformation (i.e., the original
morphology of the parent gibbsite is retained) with little
change in the external dimensions, and as a consequence, a
large internal porosity develops. The long-range ordering in
gibbsite or boehmite is lost, and an amorphous transition-
alumina form, gamma-alumina, appears.
A new phase, described as gamma-prime-alumina (γ′-
Al O ) was introduced by Paglia et al., which significantly
2
3
improved the fitting of diffraction data on aluminas from
fluid-bed and gas-suspension calciners (16). Ashida et al. de-
veloped a model used for fitting XRD data to better accom-
modate the degree of structural disorder observed for MGAs
©
2007 NRC Canada

Perander et al.
891
Fig. 2. Thermal-decomposition pathways of gibbsite, adapted from ref. 3.
by including the contribution of theta-alumina to the back-
ground (14). However, a significant challenge is the broad
and diffuse diffraction patterns of amorphous transition alu-
mina phases, making quantitative or structural determina-
tions difficult.
ranges for the chemical shifts reflects the range of possible
Al–O bond lengths and angles, and next-nearest-neighbour
2
7
species. A recently published Al NMR study of a sol–gel
manufactured theta-alumina reported the NMR parameters
2
7
of its two aluminum sites (21). Several Al studies have
made use of the high sensitivity and the relatively minor
contribution of particle and crystallite size effects of this
technique to explore the transition aluminas (10, 11, 20–25).
Furthermore, MAS NMR has also been beneficial in settling
some of the ambiguity concerning the onset temperatures of
the transition reactions; since XRD relies on the develop-
ment of nuclei of sufficient size or long-range order for the
detection of the phase, the technique does not always give
reliable onset-temperature data. However, NMR has not
been frequently applied to MGAs because of the low spec-
tral resolution caused by the co-existence of different transi-
tion aluminas and the influence of the second-order
quadrupolar interaction causing significant spectral broaden-
ing. With higher magnetic fields now becoming available,
some of these problems may be overcome (26).
Ingram-Jones et al. compared flash calcining with soak
calcining of gibbsite in laboratory trials (17). They used
2
7
XRD and Al magic-angle spinning (MAS) NMR to exam-
ine the dehydroxylation sequence, as a function of both the
calcination temperature and particle size. They suggested a
cross-over of the calcination pathways shown in Fig. 2,
where chi-alumina would transform into gamma-alumina for
flash-calcined fine gibbsite particles (<0.5 µm) and partially
for coarse (~14 µm) particles as well. The remainder of the
coarse gibbsite would react via the previously discussed
spinel-based (fcc) transformation sequence (i.e., boehmite –
gamma – delta – theta alumina) before converting to alpha-
alumina. Whittington et al. proposed a similar mechanism
based on quantitative X-ray analysis of laboratory aluminas
(
18). The authors noted that the careful control of the calci-
This study examines the use of multiple-field 27Al MAS
NMR on samples formed from the reaction of model hy-
droxides and MGAs, as compared to XRD data. Solid-state
MAS NMR spectra of quadrupolar nuclei, such as 27Al
(nuclear spin I = 5/2), are broadened by the second-order
quadrupolar interaction, a mechanism characterized by the
nation time allows the phase distribution (although perhaps
not the degree of structural disorder) of both rotary kiln and
fluid-bed-calcined aluminas to be reproduced in the labora-
tory. Interestingly, Ingram-Jones et al. also found the chi-
alumina originating from soak and flash-calcined gibbsites
to have different short-range structures.
Solid-state MAS NMR studies can provide significant in-
sight into the occupancy of octahedral and tetrahedral sites
quadrupolar coupling constant C , in MHz (26). This
Q
second-order quadrupole interaction dominates the usually
observed central transition (26) and has an angular depend-
ence that is only partially averaged by conventional MAS.
The interaction scales inversely with the applied magnetic
field, so that high magnetic fields are normally advanta-
geous, and the magnetic-field dependence can also be used
to constrain the fits. In disordered solids, a distribution in
bond lengths and angles for a particular site leads to a distri-
(
the chemical shifts being characteristic of the oxide (or hy-
droxide) coordination (AlO /Al(OH) )) (19), which is key to
n
n
understanding the forms of alumina in the samples tested.
2
7
Al MAS NMR chemical-shift ranges for aluminum in dif-
ferent coordinations with oxygen (or hydroxyls) are typi-
cally –10 to 20 ppm for octahedrally coordinated aluminum
(
AlO ), 30 to 40 ppm for pentahedrally coordinated alumi-
6
num (AlO ), and 50 to 80 ppm for tetrahedrally coordinated
bution in C , resulting in asymmetric, rather featureless
spectral lineshapes with a tail on the low frequency side. Re-
5
Q
aluminum (AlO ) (19, 20). The variation in the reported
4
©
2007 NRC Canada

8
92
Can. J. Chem. Vol. 85, 2007
cording the spectra from samples at multiple magnetic fields
can greatly constrain the fitting even of featureless and over-
Fig. 3. Rietveld refinement of XRD data for a fluid-bed calcined
MGA, illustrating the wide and weak diffraction patterns stem-
ming from the amorphous transition alumina phases present in
the material.
lapping lines. The distributions in C in such materials can
Q
be simulated as a Gaussian around a central (mean) value,
and the FWHM of this distribution can be used as a quanti-
tative measure of the amount of disorder for a particular
atomic site (27). In this study, three magnetic fields are ap-
plied, including 18.8 T. The aluminum-site distribution in
2
7
two calcined samples and an MGA are compared using Al
MAS NMR and discussed in the context of the XRD data.
Experimental
Materials
Two Bayer gibbsites from different refineries but with
very similar specifications (in terms of particle and crystal-
lite size, morphology, and purity) were calcined in a fluid-
bed-type reactor, with electrically controlled heating to
ensure a uniform temperature distribution. The calcination
time was 20 min, and the temperature was 1020 °C; X-ray
analysis reveals that the disorder in the material is similar to
that of an industrially calcined gibbsite, although the higher
temperature and longer retention time than normally encoun-
tered in an industrial fluid-bed calciner produces a slightly
more-crystalline material, indicated by the narrower and
more-developed peaks in the X-ray diffractograms, Figs. 3–
Fig. 4. X-ray diffractogram and Rietveld refinement of XRD
data for alumina A (calcined for 20 min at 1020 °C).
5
.
XRD
X-ray diffractometry was performed on a Bruker D8 appa-
ratus using Cu Kα X-rays, a 2θ range of 10–80°, a stepwidth
of 0.02°, and a counting time of 10 s. To obtain quantitative
information about the phase composition of the samples,
subsequent Rietveld refinement of the X-ray diffractograms
was performed using the FullProf software (28–30).
NMR
27Al experiments were carried out using Varian
Chemagnetics Infinity (8.45 T), Bruker Avance II (14.1 T),
and Varian Infinity Plus (18.8 T) spectrometers. All MAS
NMR experiments on Varian spectrometers were performed
2
7
using the Spinsight program. The spectra at 8.45 T ( Al
Larmor frequency 93.81 MHz) were obtained with a 3.2 mm
probe at 10 kHz spinning frequency. The pulse width used
Results and discussion
2
7
27
was 1 µs. Al spectra at 14.1 T ( Al Larmor frequency
56.37 MHz) were obtained with a 3.2 mm probe at 15 kHz
spinning frequency using the TopSpin program. The pulse
1
In the last few decades, significant progress in the under-
standing of the complex structure and phase evolution of
MGA has been made possible by advances in the fields of
MAS NMR and improvements in structural modelling of dif-
fraction data. However, quantitative measurements are still
especially challenging, since the transformation reactions are
continuous with several phases co-existing during the rapid
heating encountered in modern calciners. Analyses of MGA
show that the material is dominated by the poorly defined
gamma phase(s); understanding the structure of these phases
is key to understanding the properties of MGA. The rela-
tively ill-defined and broad diffraction patterns are exempli-
fied in Fig. 3, where the result from a Rietveld refinement
on XRD data from a MGA is shown. In fact, even the struc-
tures of most transition aluminas are not fully understood (3,
2
7
27
width was 2 µs. Al spectra at 18.8 T ( Al Larmor fre-
quency 208.24 MHz) were recorded with a 2.5 mm probe at
2
5 kHz spinning frequency. The pulse width was 1 µs. These
pulse lengths meant that the tip angle was in the linear-
response regime (26). The repetition delay for all experi-
ments was 5 s, which was sufficient to produce relaxed spec-
tra. The spectra were referenced against YAG (Yttrium
Aluminium Garnet) with the signal corresponding to the
octahedrally coordinated aluminum site set to 0.7 ppm, so
that chemical shifts reported are relative to the primary shift
3
+
reference of 1 mol/L [Al(H O) ] at 0 ppm. The QuadFit
2
6
program was used to simulate the spectra and fit the peaks
31).
(
©
2007 NRC Canada

Perander et al.
893
Table 1. Results from the Rietveld refinement (quantitative phase analysis (%)) of the XRD data for the laboratory-prepared aluminas,
A and B, used in this study and for an industrial fluid-bed-calcined MGA sample.
Sample
Gamma-Al O3
Gamma-prime-Al O
Total gamma-Al O
Theta-Al O
Alpha-Al O
2
2
3
2
3
2
3
2
3
A
B
MGA
28
26
32
42
45
44
70
71
76
22
25
22
8
4
2
Fig. 5. X-ray diffractogram and Rietveld refinement of XRD
a distribution in C in the theta form present here suggests it
Q
data for alumina B (calcined for 20 min at 1020 °C).
is more disordered than the sol–gel formed sample (21, 26).
This disorder appears to be present in the corundum phase
too, which could explain why the C value is slightly larger
Q
here than the commonly accepted value of 2.38 MHz (33).
The peaks at 71 and 16 ppm have much larger ∆C , corre-
Q
sponding to a broader distribution of aluminum environ-
ments, as would be expected from a more disordered
alumina phase. Hence, these peaks probably belong to the
more-disordered gamma-alumina. Wefers et al. considered
theta-alumina to be an intermediate between the low-
temperature transition aluminas and the energetically fa-
vored alpha-alumina (3). Levin et al. showed that a disorder-
ing of the gamma phase is required to obtain the lattice
parameters for theta-alumina (34). Furthermore, numerous
XRD studies show that theta-alumina has narrower and
more-developed peaks than gamma-alumina, also indicating
a more-crystalline, or ordered structure (e.g., Stumpf et al.
(
35)). It is therefore assumed that gamma-alumina is more
disordered than theta-alumina, and consequently that
gamma-alumina has more asymmetric peaks (i.e., a broader
distribution of C ) and larger mean C .
Q
Q
1
2, 32). The XRD technique is limited by the need for long-
The peak centred at a chemical shift of approximately
38 ppm indicates the presence of pentahedrally coordinated
range ordering in the material. Supplementary information,
as produced from Al MAS NMR, can be very useful in
better defining such systems.
2
7
aluminum (AlO ) in accordance with previously reported
5
values for the chemical shift (20, 23, 36). The presence of
penta-coordinated aluminum is consistent with the disor-
dered nature of the MGAs and is thought to originate from
more highly distorted (disordered) sites (components) in an
X-ray amorphous part. More disordered or even amorphous
aluminas, if present in significant quantities, limit the accu-
racy of XRD methods for quantitative phase measurements
of MGAs. In gamma-alumina, penta-coordinated aluminum
has also been speculated be present and would originate
from aluminum at the external surface and be associated
with the Lewis acidity and catalytic activity of the materials
In Figs. 3–5, the results from the XRD measurements and
the Rietveld refinements for the MGA sample and alumina
A and B are displayed. The numerical data from the quanti-
tative phase analysis (using the Rietveld method) are listed
in Table 1.
The 27Al MAS NMR spectra obtained from alumina A at
three fields are shown in Fig. 6, along with the individual
simulated peaks for each aluminum environment. The NMR
parameters and relative intensities of each site for both
2
7
aluminas A and B are shown in Table 2. The Al MAS
NMR spectrum of the MGA sample is shown in Fig. 7. Inte-
gration of the various lineshapes revealed that this sample
consists of a much-higher fraction of the gamma-alumina
phase (see Table 3). The key to the robustness and unique-
ness of these fits, despite significant spectral overlap, is that
the same parameters are used to fit the spectra at all three
fields. The ability to fit all three fields accurately with spuri-
ous parameters would be highly coincidental and is hence
unlikely. Also, the positions (and parameters) of several
peaks are not arbitrary, as they are largely determined by the
known phases that are present in this sample (i.e., gamma-,
theta-, and alpha-alumina).
(
24).
The only remaining question is the small peak at 23 ppm
where ∆C is only 4 MHz. This could be associated with a
Q
more “ordered” site within the disordered gamma or a small
amount of penta-coordinated aluminum perhaps within the
theta phase.
The distribution of aluminum cations between octahedral
and tetrahedral positions for the transition aluminas has been
investigated in several studies. Considering the ideal spinel
structure as a basis for the gamma-alumina structure, the
fraction of octahedral to tetrahedral aluminum is 0.67. The
ᮀ
This assignment of the peaks was made by comparison to
the chemical-shift positions and NMR parameters previously
reported for a sol–gel theta alumina and corundum (21). The
deformed spinel structures (usually represented as Al8 O ,
12
ᮀ
where represents a vacant Al), either with octahedral or
tetrahedral cation vacancies, gives a theoretical fraction of
octahedral aluminum of 0.63 and 0.75, for octahedral or tet-
rahedral vacancies, respectively. Based on the results in this
chemical shifts and C values for the theta phase are in good
Q
agreement with the published data, although the presence of
©
2007 NRC Canada

8
94
Can. J. Chem. Vol. 85, 2007
Fig. 6. Fitted 27Al MAS NMR spectra of alumina A (calcined
for 20 min at 1020 °C). The black line is the experimental spec-
trum, and the red line is the sum of the individual fitted curves.
that no hydrogen is included in the structure), leaving tetra-
hedral vacancies in the spinel-based framework (6, 24). A
recent study, however, suggests a preferential distribution of
aluminum over tetrahedral sites for gamma-alumina (37). In
another study, conducted by Lee et al., the distribution of
octahedral and tetrahedral aluminum in gamma and alpha
alumina was examined using MAS NMR as well as quantum
mechanical calculations; this reported a fraction of 0.7 ±
0
.02 (11). It should be noted that in some of the earlier
work, the lower magnetic-field strengths and lower MAS
frequencies employed result in significant interference of the
octahedral spinning sidebands with the peaks stemming
from tetrahedrally and pentahedrally coordinated aluminum.
This makes accurate quantification of the spectra very diffi-
cult. Another possible reason for the significant variation in
the observed distribution of aluminum between different
sites is the influence of different preparation techniques and
precursor materials in the various studies.
Zhou et al. examined eta-, gamma-, and theta-aluminas
using profile refinement of neutron diffraction data (6). They
suggested that, in gamma-alumina, a fraction of the alumi-
num cations would also occupy non-spinel sites, aptly
labelled as quasi-octahedral, as they lie so close to the octa-
hedral aluminum sites; the distance being only 0.373 Å.
They ascribe these sites mainly on the surface of the
gamma-alumina where they also would contribute to the cat-
alytic activity of the material. The occupation of non-spinel
sites in gamma-alumina is supported in the study by
Wolverton et al., who propose that not only surface sites
would show this abnormal coordination but also atoms in the
bulk (12). One possible explanation of the small 23 ppm
peak observed here is that it is the quasi-octahedral alumi-
num resulting from a distorted gamma-alumina framework
resulting from the incomplete transformation or complex-
phase composition, which puts further stress on the struc-
ture. If the peak at 23 ppm is a quasi-octahedral site belong-
ing to the gamma-alumina, this would correspond to 11%
and 8% of the “octahedrally” coordinated aluminum atoms
in alumina A and B, respectively. However, the deformed
spinel-based crystal structure for gamma-alumina derived by
Zhou et al. suggests that 25% of the aluminum atoms would
be in quasi-octahedral coordination.
Theoretical and computational studies strongly indicate
that cation vacancies in octahedral sites would be energeti-
cally favorable in gamma-alumina (12, 38). Krokidis et al.
propose an equilibrium structure for gamma- alumina with
6
9%–75% of the aluminum ions in octahedral positions. The
fractions of octahedrally coordinated aluminum (Table 3)
show some agreement with this structure. However, the short
heating times and the presence of theta- and alpha-alumina
could distort the gamma-alumina structure, resulting in an
incomplete lattice framework, and perhaps accounts for the
AlO observed. Data obtained on crystallographically pure,
5
single phase systems cannot be directly compared to the data
obtained for the structurally disordered systems represented
by the samples used in this study. Nevertheless, important
information on the coordination environment can be deduced
from the results.
For the theta-alumina, the ideal structure would give a
50:50 distribution for aluminum over tetrahedral and octahe-
dral sites (12). The experimental results for aluminas A, B,
study (Table 3), it cannot be unambiguously deduced
whether octahedral or tetrahedral vacancies are favored. It is
generally considered, and also experimentally supported,
that aluminum would favor octahedral coordination in
gamma-alumina under normal conditions (usually meaning
©
2007 NRC Canada

Perander et al.
895
Table 2. Peak parameters used in the curve fitting of the NMR data, collected at 18.8 T, for alumina A (calcined for 20 min at
1
020 °C) and the resulting peak intensities. The relative abundances are approximate, since the spinning sideband intensities were not
included in the integration.
27Al NMR
Peak
Chemical Shift
(±2 ppm)
Mean CQ
(±1 MHz)
∆CQ
(±1 MHz)
Alumina A
Intensity (%)
Alumina B
Intensity (%)
Alumina
Phase
Tet 1
Tet 2
Pent 1
77
71
38
23
16
14
10
5.5
8
8
5
7
3
6
7
4
9.5
3
3.5
15
15
2
4
33
8
20
12
2
3
37
5
Theta
Gamma
b
—
a
Oct
Gamma
Gamma
Alpha
Oct 1
Oct 2
Oct 3
a
3.5
3
23
22
Theta
Possibly a quasi-octahedral peak (see discussion).
Small quantity of AlO from some amorphous component of gamma- or other transition alumina (such as the X-ray amorphous chi- or rho-alumina).
b
5
Table 3. Total intensities and fraction of octahedrally coordinated aluminum for the various transition alumina phases, for aluminas A
and B and the MGA sample.
Fraction of
Fraction of
Total gamma-Al O3
(±2%)
Total theta-Al O
(±2%)
Total alpha-Al O
(±2%)
octahedral Al in
gamma-Al O
octahedral Al in
theta-Al O
2
2
3
2
3
Sample
2
3
2
3
A
B
MGA
50
51
66
38
42
28
8
5
3
0.71
0.67
0.82
0.61
0.52
0.61
Note: The quasi-octahedral peak (at a chemical shift of 23 ppm) is included in the gamma-alumina.
Fig. 7. The 27Al MAS NMR spectrum of MGA obtained at
aluminas and their transformation reactions (15, 39–42) as
well as some aluminum-containing minerals and zeolites
1
8.8 T, with simulated peaks (same colors as Fig. 6).
(
43, 44). It seems that by examining the Al K-edge XANES
or NEXAFS features, information on the coordination envi-
ronment can be obtained. The methods clearly show some
promise for analyzing disordered systems, such as MGAs.
Quantification is made in a similar manner as with the NMR
results in this study; i.e., by curve fitting of the obtained
spectrum. The results generally give a distribution of octahe-
dral and tetrahedral aluminum closer to unity regardless of
the calcination temperature (15, 40), which does not agree
well with NMR data nor has there been any report of
pentahedrally coordinated aluminum for which there is
strong experimental evidence in this and other studies.
It is clear that there still exists considerable ambiguity
over several issues relating to the transition aluminas, the
variations in structure that occur, and the accurate character-
ization of these structures. The true structure and the distri-
bution of aluminum cations and vacancies for some of the
transition aluminas remain yet to be discovered (3, 32). An-
other outstanding question is the location of the residual
hydroxyl groups, initially thought to be located on the exte-
rior of the alumina structure and at grain boundaries. In gas-
suspension or fluid-bed calcined MGAs, the very high heat-
ing rates are expected to result in a far more random distri-
bution. Computational studies indicate that for the formation
of cubic spinel-based transition alumina forms, protons may
have a stabilizing effect on the structure (45), but contradic-
tory results are obtained in neutron diffraction studies (6,
12). It has also been suggested that hydrogen would be pres-
ent as water from the decomposition reactions of gibbsite
and boehmite and located within the amorphous structure (3,
46).
and the MGA show some agreement with the ideal structure.
The theta-alumina in alumina B appears to be very close to
the ideal structure. In alumina A and the MGA, the
complex-phase mixture and severe disorder in the material
results in the lineshapes of the associated resonances show-
ing much more disorder than in a crystalline alumina. The
“
theta” phase present in these two samples may actually be
closer to the more-disordered delta-alumina phase, which
would explain the lower tetrahedral to octahedral ratio.
Recently, a number of authors report on the use of various
synchrotron radiation sources to investigate the transition
©
2007 NRC Canada

8
96
Can. J. Chem. Vol. 85, 2007
The MGAs must be viewed as a kinetically “frozen” ma-
10. K.J.D. MacKenzie, J. Temuujin, and K. Okada. Thermochim.
terial in which the different phases exist not in a simple bi-
nary mixture, but in a continuum of evolving and
disappearing phases. Local variations between particles and
even within particles further contribute to difficulties and un-
certainties with interpreting the data. The structural domains
transcend both short-range nanostructures and longer-range
ordering; a combination of techniques such as MAS NMR
and XAS with XRD will be necessary for a more-complete
understanding of these important materials.
Acta, 327(1–2), 103 (1999).
1
1
1. M.H. Lee, C.-F. Cheng, V. Heine, and J. Klinowski. Chem.
Phys. Lett. 265, 673 (1997).
2. C. Wolverton and K.C. Hass. Phys. Rev. B: Condens. Matter.
6
3, 024102/1 (2001).
1
1
3. M.L. Kronberg. Acta Metall. 5, 507 (1957).
4. T. Ashida, J.B. Metson, and M.M. Hyland. In Light metals.
Warrendale, PA, United States. 2004. pp. 93–96.
1
5. Y. Kato, K.-i. Shimizu, N. Matsushita, T. Yoshida, H. Yoshida,
A. Satsuma, and T. Hattori. Phys. Chem. Chem. Phys. 3, 1925
(
2001).
Conclusions
16. G. Paglia, C.E. Buckley, A.L. Rohl, R.D. Hart, K. Winter, A.J.
Studer, B.A. Hunter, and J.V. Hanna. Chem. Mater. 16, 220
It has been shown that 27Al MAS NMR spectroscopy can
be used to quantify the distribution of aluminum between
different sites in the component phases of MGAs. The use of
multiple applied magnetic fields, and especially the avail-
ability of higher magnetic fields, is crucial to constrain the
simulations. The results also clearly demonstrate that the
complementary information obtainable by XRD significantly
aids quantification. Amorphous aluminas, if present in sig-
nificant quantities, can however limit the accuracy of XRD
methods for quantitative phase measurements of MGAs. An
NMR peak at 23 ppm was detected and could be due to alu-
(
2004).
1
7. V.J. Ingram-Jones, R.C.T. Slade, T.W. Davies, J.C. Southern,
and S. Salvador. J. Mater. Chem. 6, 73 (1996).
18. B. Whittington and D. Ilievski. Chem. Eng. J. (Amsterdam,
Netherlands), 98, 89 (2004).
19. M.E. Smith. Appl. Magn. Reson. 4, 1 (1993).
20. R.C.T. Slade, J.C. Southern, and I.M. Thompson. J. Mater.
Chem. 1, 563 (1991).
1. L.A. O’Dell, S.L.P. Savin, A.V. Chadwick, and M.E. Smith.
Solid State Nucl. Magn. Reson. 31, 169 (2007).
2. C.S. John, N.C.M. Alma, and G.R. Hays. Appl. Catal. 6, 341
2
2
(
1983).
minum in a quasi-octahedral coordination, relating to a dis-
torted gamma-alumina form. In conclusion, 27Al MAS NMR
23. G. Kunath-Fandrei, T.J. Bastow, J.S. Hall, C. Jaeger, and M.E.
Smith. J. Phys. Chem. 99, 15138 (1995).
has been shown to be an extremely useful part of experimen-
tal methodologies to examine the commercially significant
and structurally interesting transition aluminas and their pre-
cursors.
2
2
2
4. C. Pecharroma, I. Sobrados, J.E. Iglesias, T. Gonzalez-
Carreno, and J. Sanz. J. Phys. Chem. B, 103, 6160 (1999).
5. R.C.T. Slade, J.C. Southern, and I.M. Thompson. J. Mater.
Chem. 1, 875 (1991).
6. K.J.D. MacKenzie and M.E. Smith. Multinuclear solid-state
nuclear magnetic resonance of inorganic materials. Pergamon
Materials Series 6, Pergamon-Elsevier, Oxford. 2002.
7. L.A. O’Dell, S.L.P. Savin, A.V. Chadwick, and M.E. Smith.
Appl. Magn. Reson. In Press.
Acknowledgements
Experimental assistance by the solid-state NMR group at
the University of Warwick, which thanks Engineering and
Physical Sciences Research Council (EPSRC) for funding, is
gratefully acknowledged as is financial support from the
MacDiarmid Institute for Advanced Materials and
Nanotechnology and Outotec Aluminum Technologies.
2
28. H.M. Rietveld. Acta Crystallogr. 22, 151 (1967).
29. H.M. Rietveld. J. Appl. Crystallogr. 2, 65 (1969).
30. J. Rodriguez-Carvajal. FULLPROF [computer program]: A
program for Rietveld refinement and pattern matching analy-
sis. In Abstracts of the Satellite Meeting on Powder Diffrac-
tion of the XV Congress of the IUCr. Toulouse, France. 1990.
References
27
3
1. T.F. Kemp. High-field solid-state Al NMR of ceramics and
glasses. Masters thesis. The University of Warwick, UK. 2004.
2. I. Levin and D. Brandon. J. Am. Ceram. Soc. 81, 1995 (1998).
3. H.J. Jakobsen, J. Skibsted, H. Bildsoee, and N.C. Nielsen. J.
Magn. Reson. 85, 173 (1989).
1
. R.G. Haverkamp. Surface studies and dissolution studies of
fluorinated alumina. Ph.D. thesis. The University of Auckland.
Auckland, New Zealand. 1992.
3
3
2
. P.N. Østbø. Evolution of alpha phase alumina in agglomerates
upon addition to cryolitic melts. Ph.D. thesis. Norwegian
University of Science and Technology, Faculty of Natural
Sciences and Technology, Trondheim, Norway. 2002. p. 252.
. K. Wefers and C. Misra. Eds. Oxides and Hydroxides of alu-
minum. Alcoa Technical Paper No. 19. Aluminum Company
of America: Pittsburgh, PA. 1987.
3
3
3
3
3
3
4
4
4. I. Levin, L.A. Bendersky, D.G. Brandon, and M. Ruhle. Acta
Mater. 45, 3659 (1997).
5. H.C. Stumpf, A.S. Russell, J.W. Newsome, and C.M. Tucker.
J. Ind. Eng. Chem. 42, 1398 (1950).
6. T.R. Barton, T.J. Bastow, J.S. Hall, B.J. Robson, and M.E.
Smith. Light Metals, 71 (1995).
7. H.J. Kim, H.C. Lee, and J.S. Lee. J. Phys. Chem. C, 111, 1579
3
4
5
. B.E. Raahauge. Aluminum (Isernhagen, Germany), 83, 40 (2007).
. B.C. Lippens and J.H. de Boer. Acta Crystallogr. 17, 1312
(
2007).
8. X. Krokidis, P. Raybaud, A.-E. Gobichon, B. Rebours, P.
Euzen, and H. Toulhoat. J. Phys. Chem. B, 105, 5121 (2001).
9. A.W. Chester, R.P.A. Absil, G.J. Kennedy, P. Lagarde, and
A.M. Flank. J. Synchrotron Radiation, 6, 448 (1999).
0. J.F. Larue, B. Moraweck, A. Renouprez, and P. Lagarde.
Springer Series in Chemical Physics, 27, 32 (1983).
1. B.A. Latella, and B.H. O’Connor. J. Am. Ceram. Soc. 80,
(
1964).
6
. R.S. Zhou and R.L. Snyder. Acta Crystallogr., Sect. B: Struct.
Sci. 47, 617 (1991).
7
8
. F. Freund. Ber. Deut. Keram. Ges. 42, 23 (1965).
. J. Rouquerol, F. Rouquerol, and M. Ganteaume. J. Catal. 36,
9
9 (1975).
. J. Rouquerol, F. Rouquerol, and M. Ganteaume. J. Catal. 57,
22 (1979).
9
2
941 (1997).
2
©
2007 NRC Canada

Perander et al.
897
4
4
4
2. K.-I. Shimizu, Y. Kato, H. Yoshida, A. Satsuma, T. Hattori,
and T. Yoshida. Chem. Commun. (Cambridge), 17, 1681 (1999).
3. P. Ildefonse, D. Cabaret, P. Sainctavit, G. Calas, A.M. Flank,
and P. Lagarde. Phys. Chem. Miner. 25, 112 (1998).
45. K. Sohlberg, S.J. Pennycook, and S.T. Pantelides. J. Am.
Chem. Soc. 121, 7493 (1999).
46. G. Paglia, E.S. Bozin, and S.J.L. Billinge. Chem. Mater. 18,
3242 (2006).
4. J.A. Van Bokhoven, H. Sambe, D.E. Ramaker, and D.C.
Koningsberger. J. Phys. Chem. B, 103, 7557 (1999).
©
2007 NRC Canada