Você está na página 1de 14

THE ROLE OF NMR SmFfS IN STRUCTURAL STUDIES OF GLASSES, CERAMICS

AND MINERALS.
R. DUPREE, S.C. KOHN, C.M.B. HENDERSON, A.M.T BELL
University of Warwick
Physics Department
Covelltry CV4 TAL
U.K.

University of Manchester
Geology Department
Manchester M13 9PL
U.K.

SERC
Daresbury Laboratory
Warrington WA4 4AD
U.K.

ABSTRACT. Experimental data on chemical shifts in a variety of crystalline and amorphous silicates
and ceramics are reviewed and the published correlations between 29Si chemical shifts and structural
parameters are critically examined. None of the correlations accurately predict the 29Si spectra of
tridymite or K2MgSi 50 12 leucite, but those of Sherriff and Grundy (1986) and Sternberg (1988) are
superior to the simpler geometric functions. However all except the Sherriff and Grundy correlation can
be used to provide some information on bond angle distributions in simple glasses whose medium range
order is poorly understood. 15 N, 29Si and 13C shifts in ceramics are potentially very informative, but
no correlations between shift and structure in these materials are currently available. Much further
theoretical work on the calculation of chemical shi fts is required before quantitative structural information
can be reliably deduced from NMR shifts in the materials discussed here.

Introduction

An experimentalist uses the chemical shift as a probe of structure. Whilst appreciating that it
is not possible at present, to calculate the shift from first principles let alone characterise a
structure by one number (or three from the shift tensor) he would like some simple functional
form for the variation of shift with some simple structural parameter. In this paper we will
attempt to show how even limited knowledge of the connection between shift and structure can
be of value and to point out how the limitations in our current knowledge restrict the
information content of an NMR spectrum. Some suggestions for future calculations will be
made. We shall restrict our discussion to three classes of materials, glasses, framework silicate
minerals and analogues, and ceramics, and will concentrate (mostly) on 29Si NMR.

Background

Since the first reasonably high resolution NMR measurements on silicates correlations between
the shift and various structural parameters have been suggested. The large majority of these
have been concerned with 29Si tetrahedrally coordinated to oxygen since this is by far the most
common structural unit in these materials. Some work has been done on six-coordinated silicon
(Stebbins and Kanzaki (1991 and on 27 AI (e.g. Phillips et al. (1989 but the complication of
the quadrupolar contribution to the position and the broader lines has limited the amount of
421
J. A. Tossell (ed.), Nuclear Magnetic Shieldings and Molecular Structure, 421-434.
1993 Kluwer Academic Publishers.

422

data. There has been ver little attempt to quantitatively correlate the shifts of other readily
accessible nuclei such as 3Na or lIB with details of structure. The situation up to 1986 has
been reviewed by Engelhardt and Michel (1987) and a more complete discussion is given there.
The simplest correlations are with mean Si-O-Si bond angle, Ci ,or mean Si-O bond length.
The first of these has apparently given a reasonable agreement with experiment for instance the
chemical shift, 0, being given by
0= -0.6192Ci-18.68

(1)

for a number of Si02 polymorphs and Q4(OAI) zeolites (Engelhardt and Radeglia (1984. A
less good but still significant correlation was found between the bond length and the shift (e.g.
Smith and Blackwell (1983. These correlations are entirely empirical but show that both these
parameters are related to the shift. More physically based correlations have been suggested by
a number of authors in particular Engelhardt and Radeglia pointed out that for the typical net
charges for tetrahedral Si04 units the chemical shift varies approximately linearly with the
orbital electronegativity of the oxygens in the SiO bonds. This in tum can be related to the
degree of s hybridization, p, of the oxygen bond orbitals by p=cosClt/(cosClt-1). Thus for
tetrahedral units a correlation such as given in eqn 2 might be expected.
(2)
In fact the agreement with experiment is of similar quality to that of eqn 1. Another physically
based correlation which gives a slightly better agreement with experiment is based on the mean
secant of the four Si-O-Si bond angles (Smith and Blackwell (1983. More recently the effect
of the second coordination sphere on the shift has been calculated by Sternberg (1988) and
Hallas and Sternberg (1989) using a point charge approximation. The chemical shift is given
by the sum of two contributions, ofs the first sphere shift of an isolated unit and the shift caused
by the Si-O (or AI-O) bond polarization which is dependent on the Si-O-X bond angle and on
X (eqn 3).

o = ofs

Apol~ (VSiO)i

(3)

Apol is dependent on the bond polarity and is determined from experiment. For Si02
polymorphs and dealuminated zeolites this also gives a nonlinear variation of shift with bond
angle and only differs significantly from previous correlations at large (> 160) and small
140) angles for which there is little experimental data. A further correlation between the
chemical shift and structural parameters for silicate minerals was suggested by Sherriff and
Grundy (1988). This was based on the magnetic anisotropy and valence of the bond between
the oxygen and the second neighbour cation to the silicon. The expression for the magnetic
anisotropy is given by
(4)

where R is the distance between each silicon atom and the midpoint of the oxygen-second

423

Figure 1 Diagram showing the angles and


lengths used in eqn (4). (after Sherriff and
Grundy (1988

Si

neighbour cation bond, r the length of this bond and e the angle between them (see fig 1).
The bond valence Sj is given by Sj = exp[ (ro -ri)/0.37] where ro' the bond valence parameter,
was taken from Brown and Altermatt (1985). A total of 76 data points were used to give

/) = 359.11 X' - 44.37

(5)

A somewhat better agreement with experiment was found if the hybridization of the bridging
oxygen was included in which case eqn 4 becomes
(6)

giving the chemical shift as

/) = 650.08 X" - 56.06

(7)

Over the range of samples used this probably gives the best overall correlation although it is
less amenable to direct stuctural interpretation and perhaps is best used to discriminate between
different structural models. It includes the hybridization term (as does that of Sternberg) and
has a distance parameter R which is related to that of Sternberg. One problem with all the
correlations is that the quality of the crystal structures may not be very good, small variations
in bond angle/length between one determination and another can easily give a calculated shift
difference of -1ppm.
All of the shift correlations discussed above assume that a partial chemical shift per
silicon oxygen bond can be defined. The total shift being the sum of these partial values
(8)

424

Thus the bonding and chemical shift contribution at each corner of the tetrahedron is assumed
independent of that at adjacent corners; whilst this is a reasonable first order approach there
is some evidence (presented later) that it may not always apply.
3

Glasses

It is difficult to obtain much structural information about glasses because the lack of long range
order means that scattering techniques such as XRD etc are of limited applicability. In
particular information beyond the nearest neighbour distance is limited. The different
coordination polyhedra which can be present in some glasses are readily distinguished using
NMR since for e.g. 29Si typical shift ranges for Si04 units are in the range -6Oppm to
-12Oppm and for Si06 between -180ppm and -22Oppm. For 27 Al the situation is more
complicated with AlOs polyhedra occurring much more frequently than in crystalline materials.
These usually occur with a peak position - 30-40ppm compared with 70-4Oppm for Al04 and
< 20ppm for Al06 . However the shift can be quite strongly affected by the second
coordination sphere, e.g. replacement of Al by P causing a negative shift of -2Oppm for both
tetrahedral and octahedral AI. For aluminoborate glasses this has led to Al peaks at 30ppm
being assigned to AI[OB 41by some authors and to Al[OBsl by others (Dupree et a1. (1985a),
Hahnert and Hallas (1987), Hallas and Sternberg (1989) and Bunker et a1. (1991. A central
issue in simple binary glasses has been the distribution of non bridging oxygens which is
related to the amount of "order" in the glass. Fortunately for N~O-Si02 and L~O-SiOj
glasses Q4, Q3 and Q2 units are resolvable, however the shift difference between Q and Q
etc decreases with electronegativity, see fig 2, so that it becomes more difficult to resolve the
different Q species for glasses containing other alkalis or alkaline earths. The influence of the
second coordination sphere on the silicon shift can give useful information and examples are

..

-90

I-t: -80
Figure 2 29Si shifts for different Qn species
v electronegativity for alkali silicate glasses
(after Dupree et al.( 1986

II II

[51.1][ 48.5]

~ -70

(j

[67]

crystalline
x

at compositions
indicated in brackets
[mole ~

[60]
-60~~--~--~--~--~

0.7

0.8

0.9

1.0

1.1

1.2

ALLRED-ROCHOW ELECTRONEGATIVilY

425

-20
-18
-16

-94

-14
-12

6Si -92
ppm-SO

-10
SNa -8

I~

(a)

-88+---+------4
100
80
60
40

ppm

-6

-4
-2
33.3(Na20/AlzO;a) 666 SiO~

o
2+---~----~--~

40

60

80

100

Figure 3 Effect ~f substitution of Al 20 3 for Na20 in a (Na20.AI203>o.33(Si~)O.66 glass on (a) the 29Si
shift and (b) the _3Na shift. (after Dupree et al (1985b)
-110

Q.
Q.

;:: -100
I&.

:i:
en
..J

()

is

iii
I!l

-90

-80+-~~~----+------~------~---~------+-----4------~

02

04

06
51
51+AI

08

10

Figure 4 Variation of 29Si shift with Si/(Si + AI) ratio for framework aluminosilicate glasses
(after Dupree (1988) and Oestrike et al.(1987

426

shown in figs 3 and 4. Figure 3a shows the effect on the 29Si shift as Na20 is replaced by
AI 20 3 (Dupree et al. (1985b. The change in the shift -4ppm for 50% AI 20 3 is comparable
with that observed in zeolites when a silicon in the second coordination sphere is replaced by
aluminium. Much stronger eftects are seen in the sodium position (fig 3b) which changes by
-18ppm for a sodium to aluminium ratio of one. The true chemical shift for 23Na is moved
by -2ppm upfield from the peak position in Na20.2Si02 glass because of the second order
quadrupolar shift. Whilst there may be some change in the quadrupolar interaction as
aluminium is added to the glass most of this difference in position must come from a large
change in shift. Whilst confirming that [A1041- INa + association occurs in these glasses, there
do not appear to be calculations of 23Na shifts even in simple materials. Figure 4 shows the
29Si shift v Si/(Si + AI) ratio for a number of framework (i.e. all bridging oxygen) glasses
(Oestrike et al. (1987), Dupree (1988. The modifier cation has only a small effect and the
dominant cause of the shift changes is the increasing number of AI atoms in the next neighbour
shell. For an SiI(Si + AI) ratio of < 0.5 it is no longer possible for aluminium avoidance to be
maintained as at a ratio of 0.5 each Si would have 4AI neighbours. The smaller slope indicates
that there is only a small increase in the number of AI next nearest neighbours beyond this
composition whilst the second neighbour Al content increases. The width of the lines, which
would be very broad for a statistical distribution of aluminium over the tetrahedral sites, also
gives evidence for a restricted range of environments.
The width of the line can also be used in some systems to derive the Si-O-Si bond angle
distribution, an important parameter in describing the intermediate range order in the glasses.
We have discussed this several times before (Dupree and Pettifer (1984), Pettifer et al. (1988
so only brief details will be given here. If the bond angle distribution is V(o:), then:-

V(O!)

X(r\) , do:,
dF

(9)

where X(Oj) is the distribution function of the partial shifts. The distribution function of the
total shift W(O) is then given by:(10)

where FT denotes Fourier Transformation and W(O) is the measured NMR line. For SiOz this
is approximately Gaussian, thus X(Oj) can be readily determined and hence V(o:) using one of
the correlations described earlier. Figure 5 shows the bond angle distribution obtained for
three different correlations together with the distribution determined using X-rays. Whilst there
is qualitative agreement in that the most probable bond angle determined from the NMR data
lies between 142 0 and 151 0 compared with 146 0 from the X-ray data and all the NMR results
give a narrower distribution function than that deduced using X-rays, there are considerable
differences in the details. Very recently Farnan et al. (1992) have used 2D DAS on 170 in
some potassium silicate glasses to derive information on the Si-O-Si bond angle distribution.
This has the advantage of involving only one angle so that it does not rely on the additivity
assumed (eqn. (8 above. The correlation with angle is with the electric field gradient at the
oxygen site and with its asymmetry parameter 71 rather than the chemical shift which seems to
vary only slightly. They found that the bond angle distribution in this material was narrower
than that deduced from X-ray measurements on SiOz. Preliminary results for Si02 using 170

427

V(a)
/"

XRD-'
.I'
./

10

ci

"

.I'

~.",-

",'.

',~
.
,',

./

..

,'~.',
"

./~linear

140

" ".

"~'. xpoint-chOrge

120

'-.....

.I'

----

".'

"

'.'" ,".

'" .'"',".
"

", ~"",',

160

~~

180

Si-O-Si

Figure 5 The Si-O-Si bond angle distribution in Si02 using :- (i) the linear correlation, (ii) the point
charge model, (iii) the secant correlation and (iv) the distribution deduced using X-rays (after Pettifer
et al. (1988))

DAS are much closer to the distributions deduced from 29Si NMR than the X-ray V(a) and lie
somewhere between a and b on figure 5 (Farnan and Grandinetti pers. corom.).
4

Minerals

It is of interest to compare the predictions of these different correlations with the spectrum
observed for tridymite as we did some time ago (pettifer et al. (1988)) but this time including
the Sherriff and Grundy (1988) predictions. This is shown in figure 6(A) and (B). As we noted
before none of the models match the spectrum well, the linear model is worst with a peak at
-116ppm which is not observed and no peak - -109ppm. The secant model predicts too
narrow a spectrum and has only a low intensity at -114.Sppm and -109ppm. The point
charge model produces three peaks in roughly the correct positions but their relative amplitudes
are in poor agreement with experiment. The Sherriff and Grundy model predicts the high
frequency part of the spectrum well, but on the low frequency side there is a difference
between experiment and the model which increases as the shift becomes more negative. Thus
there is a peak predicted at -116.Sppm compared with the most negatively shifted peak of the
spectrum at -114.2ppm a difference of 2.3ppm. The peak at -11Sppm may correspond to
the experimental peak at -113.2ppm a difference of 1.8ppm etc. However some caution is
required since the calculated peak positions may not come in the same order as experiment, it
is only by knowing which site corresponds to each part of the spectrum that a truly valid
comparison can be made.
A case where this is known and where it is of interest to compare the different
correlations is in potassium magnesium leucite (K2MgSisOd which is based on the framework
topology of the mineral leucite (KAlSi 2 0 6) but with Mg and Si in tetrahedral sites. Natural

428

105

110

115

PPM

120

/\

\
\
. , \\
" \

.'.\:'\,'

...

..... \ \

".'

...........

'':-''

... - ':~""=-"';:~"""

105

110

PPM

115

120

Figure 6 The 29Si spectrum of tridymite, 6(A) lower. Experiment (a) compared with simulations based
on (b) the linear correlation, (c) the secant correlation, (d) the s-hybridisation correlation, and (e) the
point charge model; 6(B) upper, Experiment compared with the simulation (t) using eqns (6) and (7).

429

1\

L_.,-,-" ~~ ~'---=-'-':::-'-~~~-=':'::-,-~~~'---=f'=-,,"~~~'.J-'~'~~'--'-',-L'~~_',-----,,---l'

-80

-85

-90

-95

PPM

-100

-105

-110

Figure 7 The spectmm for KMg leucite (K 2MgSi 50 12) (lower), compared with the simulation using
eqns (6) and (7), (upper).

430
leucite is tetragonal with three distinct tetrahedral sites in the unit cell which contains 48
tetrahedral cations and aluminium and silicon are partially disordered over these sites. KMg
leucite prepared hydrothermally is of lower symmetry than naturalleucite and the 29Si spectrum
consists of 10 narrow lines spread over 20ppm (Kohn et a1. (1991 indicating that the material
is fully ordered with 12 distinct T sites in the unit cell. The chemical shift anisotropies
obtained from static and slow spinning MAS experiments show that two of the lines are due
to silicon linked to four other silicon sites (Q4(4Si and the other 8 were due to Si linked to
3Si and IMg (Q4(3Si. As the structure was unknown and the XRD complex a 29Si COSY
spectrum was run to determine the connectivity of the sites. The structure has now been
determined by Bell et al. (1992) enabling us, when combined with the connectivity data from
the COSY experiment, to make a quantitative comparison with the different correlations.
Figure 7 shows the experimental spectrum together with that calculated using equations 6 and
7. Overall there is a considerable degree of similarity between them, the theoretical spectrum
being slightly compressed compared with experiment. The worst agreement with experiment
is for the peak at -91.0ppm in the experimental spectrum, which is from the Q4(4Si) site with
relatively small bond angles (from 130 to 140.6), it is predicted to be at -95.6ppm a

-80

E
c..

-90

-100

:(I)
E
'"C
CD

~
~
()

-110
-110

-100

-90

-80

measured chemical shift (ppm)


Figure 8 Experimental 29Si shifts v calculated shifts for KMg leucite. Squares using eqn (6) and (7),
circles point charge model, triangles mean secant correlation.

431
difference of 4.6ppm. The order of the peaks around -89ppm is changed emphasising the
need for caution in the assignment of peaks if the connectivity is not known. The other
correlations give a somewhat worse agreement with experiment with that of Sternberg being
the best provided that Apol in eqn.3 is changed from -284ppm/Hartree to -200ppm/Hartree
and Dfs from -53ppm to -59.5ppm. for the sites with one Mg neighbour. This value for ~1
is close to that suggested by Sternberg, -190ppm/Hartree, for use with second sphere atoms
other than silicon. Peak 5 the Q\4Si) peak at -91ppm is predicted to be at -96ppm, the
other peaks which deviate most from experiment are peak 8 (at -96.4ppm) which is calculated
to be 2.7ppm more positive than is observed and peak 3 (at -89.2ppm) for which the
calculation gives -91.9ppm. Figure 8 shows a comparison of the calculated and measured
shifts for these two correlations together with a mean secant correlation where the slope and
constant for the Q4(3Si) lines has been adjusted to give a best fit to the data (i.e. to -31 and
-135ppm from -55.8 and -176.7 respectively). No correlation is able to fit peak 5.
Assuming the crystal structure is correct, this may be because of dynamic effects so that the
X-rays see a mean bond angle that is different from that seen by NMR or because the
correlations are extended beyond their normal range. All other Q4(4Si) sites known to us have
mean bond angles;;::: 139. This site also has the largest variation of Si-O distance amongst the
4 neighbours (0.026A) so this too may have an effect on the correlations. The other peak
which is not fitted very well (except for the Sherriff and Grundy correlation where it is in no
worse agreement with experiment than several other peaks) is peak 8 at -96.4ppm. This has
the largest range of bond angles from 119.8 to 160.5 and may be an indication that simple
additivity, eqn.8, is breaking down. It should be noted that the simple average angle
correlation (eqn (1 predicts completely unrealistic mean bond angle of 116.8 for peak 5
compared with the experimental value of 134.9. It is therefore recommended that this
correlation is not used outside the narrow angular range where it was first formulated, even
there caution is required as it predicts -109.6ppm instead of -104.5ppm for the Q\4Si) peak
10 of our KMg leucite sample.
5

Ceramics

The fairly large amount of NMR work on ceramics seems not to have ~roduced any successful
correlations between the shift and structure. Table 1 lists some 29Si, 1 C and lsN shifts in the
polymorphs of Si3N4 and some polytypes of SiC. In some sense these are good candidates for
shift calculations since they contain only two, light, elements and the shifts of both are known
to high accuracy. The 29Si shifts are all less negative than for silicon bonded to oxygen and it
is clear that the correlations discussed earlier are not applicable since for example eqn.6 and
7 would require Si-X-Si angles of < 100 to change the sign of the (1-3cos2S) term which is
neccessary for shifts more positive than -56ppm. The lsN shift is more sensitive to its
surroundings than either l3C or 29Si as evidenced from the 4 different sites in a-Si3N 4 and also
from work in sialon systems (Kruppa et al. (1991.
Figure 9 shows some 29Si shift ranges that occur in oxynitride ceramics, there is
considerable overlap of the different coordination tetrahedra and the shifts for SiN4 extend from
- 20ppm (sialon polytypoid 21 R) (Smith (1992 to -65ppm (LaSi3NS) (Dupree et al. (1989.
It is of interest that the shifts in lanthanum containing phases are more negative by -ISppm
than in the yttrium based phases. The 29Si shifts in a number of phases in the Si~-Si3N4AIN-AI 20 3 systems e.g. ,s'-sialon (Si6_zAlzNs_zOz) are remarkably insensitive to z. Two

432
TABLE 1. 29Si, l3C and 15N shifts
in some simple ceramics
5<;;lppm

5r /ppm

ref

~-SiC

-17.2

23.7

1,2

4H
SiC

-19.7
-22.5

14.7
2l.5

6H

-14.3
-20.4
-24.9

15.2
20.2
22.7

4,5

-14.6
-20.5
-24.1

16.0
20.7
22.7

SiC
15R
SiC

1,4,5

5N*

l.
3.
5.
7.

a-Si 3N4

-49.0
-47.1

5l.6
53.4
64.0
74.6

,6-Si3N4

-48.5

50.9
68.4

6,7,8

Si,N"O

-6l.6

40.3

6,7,8

6,7,8

G R Finlay et al. (1985), 2. G W Wagner et al. (1989)


M Leach (1990), 4. J R Guth et al. (1987)
J S Hartman et al. (1987), 6. R K Harris et al. (1990)
D Kruppa et al. (1991), 8. R Dupree et al. (1985c)

explanations for this have been proposed, either the silicon coordinates to four nitrogen atoms
at all levels of substitution (the mixed coordinations being restricted to aluminium tetrahedra)
(Dupree et al. (1988, 1989)), or the chemical shift remains unaltered for the SiN4-xOx (x=O-4)
tetrahedra when an AI-O group replaces an isoelectronic Si-N unit since the groups have similar
electronegativities (Leach (1990). Theoretical work on this type of system might clarify this.

433
Tetrahedron

SiD.

SiOJN

Si().j42

SiONJ

SiN.
-20

-40

-60

-81)

chemical shift in ppm

-100

-120

Figure 9 29Si shift ranges in oxynitrides

Conclusion

All the correlations of shift with structure discussed in this work are somewhat imprecise. The
formulation of Sherriff and Grundy (1988) seems to give the best overall correlation for the
29Si shift in silicates. Simpler shift correlations based solely on the Si-O-Si bond angle are able
to give useful structural information about glasses. Extreme care is needed if these
correlations, particularly the linear shift vangie correlation, are used outside the range of
angles/environments for which they were formulated. There is a need for successful
shift/structure correlations in ceramic materials and for calculations of the shift for nuclei other
than 29Si e.g. for 23Na, 15N and 13C in the materials discussed above.

Acknowledgement

This work was supported by the S.E.R.C. and the N.E.R.C.

References
Bell, A.M.T., Fitch, A.N., Cernik, R.I., Champness, P.E., Henderson, C.M.B.,
Redfern, S.A.T., and Kohn, S.C. (1992) Acta Cryst B (submitted).
Brown, I.D., and Altermatt, D. (1985) Acta Cryst. B41, 244-247.
Bunker. B.C., Kirkpatrick, R.I., Brow, R.K., Turner, G.L., and Nelson, C. (1991)

434

1. Am. Ceram. Soc. 74, 1430-1438.


Dupree, R., andPettifer, R.F. (1984) Nature 308, 523-525.
Dupree, R., Holland, D., and Williams, D.S. (1985a) Phys. Chern. Glasses 26, 2.
Dupree, R., Holland, D., and Williams, D.S. (1985b) J. Physique Colloque C8, 46,
119.
Dupree, R., Lewis, M.H., Leng-Ward, G., and Williams, D.S. (1985c) J. Mater. Sci.
Lett. 4, 393-395.
Dupree, R. (1988) Experimentelle Technik der Physik 36,315-325.
Dupree, R., Lewis, M.H., and Smith, M.E. (1988) J.AppI.Cryst. 21, 109-116.
Dupree, R., Lewis, M.H., and Smith, M.E. (1989) 1. Am. Chern. Soc. 111,5123.
Engelhardt, G., and Michel, D. (1987) High resolution solid state NMR of silicates
and zeolites. 1. Wiley & Sons.
Engelhardt, G., and Radeglia, R. (1984) Chern. Phys. Lett. 108, 271-274.
Farnan, 1., Grandinetti, P.J., Baitisberger, J.H., Stebbins, J.F., Werner, U.,
Eastman, M.A., and Pines. A. (1992) Nature 358,31-35.
Finlay, G.R., Hartman, J.S., Richardson, M.F., and Williams, B.L. (1985) J. Chern.
Soc. Chern. Commun. 159-161.
Guth, J.R., and Petuskey, W.T. (1987) 1. Phys. Chern. 91,5361-5364.
Hahnert, M., and Hallas, E. (1987) Rev. Chim. Miner. 29, 221.
Hallas, E., and Sternberg, U. (1989) Mol. Phys. 68, 315-326.
Hartman, 1.S., Richardson, M.F., Sherriff, B.L., and Winsborrow, B.G. (1987) J.
Am. Chern. Soc. 109,6059-6067.
Harris, R.K., Leach, M.J., and Thompson, D.P. (1990) Chern. Mater. 2,320.
Kruppa, D., Dupree, R., and Lewis, M.H. (1991) Materials Letters 11, 195.
Kohn, S.C., Dupree, R., Mortuza, M.G., and Henderson, C.M.B. (1991) Phys.
Chern. Minerals 18, 144-152.
Leach, M. (1990) Ph.D. thesis University of Durham.
Oestrike, R., Yang, W.H., Kirkpatrick, R.J., Hervig, R.L., Navrotsky, A., and
Montez. B. (1987) Geochim. Cosmochim. Acta, 51 2199.
Pettifer, R.F., Dupree, R., Farnan, I., and Sternberg. U. (1988) J. NonCryst. Sol. 106, 408-412.
Phillips, B.L., Kirkpatrick, R.J., and Putnis, A. (1989) Phys. Chern. Minerals 16,
591-598
Sherriff, B., and Grundy, H.D. (1988) Nature 332,819.
Smith, 1.V., and Blackwell, C.S. (1983) Nature 303,223-225.
Smith, M.E., (1992) J .Phys. Chern. 96, 1444
Stebbins, J.F., and Kanzaki, M. (1991) Science 251,294
Sternberg, U. (1988) Mol. Phys. 63, 249.
Wagner, G.W., Na, B.K., and Vannice, M.A. (1989) J. Phys. Chern. 93, 5061-5064.

Você também pode gostar