eLearning

the samples used in the measurements. The minor disagree-
ment in the Al-L edges with experiment can also be ex-
plained by the possible surface effect where the surface Al
atoms are fivefold bonded.
18
This could change the intensity
ratio of the two leading peaks in the measured spectrum.
Calculations of the XANES spectra for other transition
alumina97
show that the Al-L edge and the O-K edge differ
from those of
-Al2O3 even when their local coordination is
the similar. In Fig. 15, the experimental data for the O-K
edge15 are presented for comparison. As can be seen, the
agreement in the asymmetric shape of the spectrum with the
calculation is very satisfactory. The experimental spectrum

has much less energy resolution so some of the features
present in the calculated spectrum cannot be resolved.
VI. DISCUSSION AND SUMMARY
The above rather comprehensive list of the calculated
physical properties of
-Al2O3 based on the proposed struc-
ture in Ref. 30 and their good agreement with the available
experimental data give credence to this structure for

-Al2O3. In the cases where the agreement is less than de-
sirable, it can be attributed to possible sample variations
since
-Al2O3 is not in the form of a single crystal and may
not even be totally stoichiometric. Various previous studies
indicated the importance of the surface effect in porous

-Al2O3. The present calculation is for a bulk
-Al2O3
model with a unit cell with 40 atoms. When performing a
comparison with the measured physical properties, this fact
must be taken into consideration.
Our main conclusions are that
-Al2O3 should “not” be
considered to have an underlying crystalline structure
cubic

spinel or otherwise accompanied by defects
cation vacan-
cies. Rather, it should be viewed as an amorphous random
network structure of Al–O bonds where Al can be either
tetrahedrally or octahedrally bonded, and the corresponding
O will have to be either threefold or fourfold bonded. In such
a structure, the notion of cation vacancy in a regular spinel
lattice loses its meaning entirely. Indeed, a detailed simula-
tion studies by Wolverton et al.,
25
Krokidis et al.,
34
Digne et
al.
35
and Paglia et al.
28
already show that the nonspinel
model where no spinel sites are occupied by Al atoms is the
best representative for the structure of
-Al2O3. The calcu-
lated electronic structure of
-Al2O3 has significant differ-
ences when compared to -Al2O3 and -Al2O3
Ref. 40
calculated using the same method. This is mainly due to
different percentages of Altet
,Aloct
,O3-fold, and O4-fold. Table
IV lists the percentages of these local units for -Al2O3,

-Al2O3, and the other two transition alumina -Al2O3
Refs.
6 and 40 and -Al2O3.
6
Hence, an appropriate model for

-Al2O3 probably should be that of a glasslike random net-
work structure with well defined local short-range order but
no long-range order. Additional modeling work in this direc-
tion is highly desirable. The structure determined by Ref. 30
clearly represents the smallest unit cell for an amorphous
network that accounts for all the essential characteristics of

-Al2O3. It is highly desirable to extend the modeling effort
to larger amorphous models for
-Al2O3 and even those with
internal voids and surfaces which can rationalize the differ-
ence with some experimental observations. In these amor-
phouslike models, there could be some variations in the per-
centages of Altet
and Aloct
as well as O3-fold and O4-fold.
The above viewpoint is further supported by several facts.
First, there exist highly localized states at the top of the VB,
characteristic of an amorphous solid. Second, the variations
in the Mulliken effective charges and BO values in the
present model also indicate that the atomic structure within
the 40-atom cell is quite disordered. This is true for both the
cations and the anions and within groups of atoms with the