American Mineralogist, Volume 70, pages 537-548, 1985
High resolutionaluminum-27and silicon-29nuclearmagneticresonance
spectroscopic
study of layer silicates,includingclay minerals
Rosrnr A. KrNsnv
Schoolof ChemicalSciences,505SouthMathewsAuenue
Uniuersityof Illinois at Urbana-Champaign,
Urbana,lllinois 61801
R. Jeuns KnrpernrcK, JorrNHownnl
Departmentof Geology,I30l West GreenStreet
Uniuersityof Illinois at Urbana-Champaign,
Urbana, Illinois 61801
Kenm ArN Surur eNo Enrc Oronrr,o
Schoolof ChemicalSciences,
505SouthMathewsAuenue
Uniuersityof Illinois at Urbana-Champaign,
Urbana, Illinois 61801
Abstract
High-resolution solid-state magic-angle sample-spinning nuclear magnetic resonance
(NMR) spectroscopyof layer silicates,including clay minerals,indicatethe following relationships betweenthe NMR parametersand the structuresand compositionsof the minerals.
These relationshipsare potentially useful in understandingthe crystal chemistry of layer
silicatesand the chemicaland structural changesthat take place in them during diagenesis
and metamorphism.(1) There is a systematicdeshieldingat silicon (lessnegativechemical
shift) with decreasingSi/(Si + A(4). (2) There is a trend of more positive chemicalshifts for
tetrahedrally-coordinatedaluminum with decreasingSi/(Si + A(4)). (3) Thereis no systematic
relationship between the chemical shift of octahedrally-coordinatedaluminum and Si/
(Si + Al(4)).(4) There is moderatelygood agreementbetweenA(4yA(6) ratios determinedby
NMR and by chemicalanalysis.(5) The silicon-29NMR peak breadthsof mineralswith no
Al(4) or with orderedA(4ySi arrang€mentsare relatively narrow, while those of most of the
other minerals are broader. (6) Paramagneticimpurities (probably iron) in the natural samples causeextensivespinning side-bandswhich complicatespectralanalysis.(7) Higher magnetic field strengthsgreatly improve spectralresolutionfor aluminum-27.
They are also qualitatively consistentwith the relationship
between
silicon-29chemicalshift and cation-oxygenbond
High-resolution magic-angle sample-spinning(MASS)
strength
developed
by Smithet al. (1983).
nuclear magneticresonance(NMR) spectroscopyof solids
The resultsallow, for the first time, the use of solid-state
is a potentially powerful tool for understandingthe structure of materials such as clays and fine grained zeolite MASS NMR to estimateSi/(Si + Al(4)) ratios, the ratio of
catalysts(seee.g.,Klinowski et al., 1982).In this paper,we tetrahedral to octahedral aluminum. and to some extent
present high-resolution solid-state aluminum-27 and the degreeof Al/Si order in the tetrahedrallayers of sheet
silicon-29MASS NMR data for a variety of layer silicates, silicatesof unknownstructure.
including clay minerals. Our objectivesare to determine
Methods
how the NMR parametersof layer silicatesvary with comNMR
spectroscopy
position and structure,investigatingfor example,the parModern Fourier transform NMR spectroscopy is well described
titioning of aluminum betweentetrahedraland octahedral
sites,Si/(Si + A(4) ratios, and the extent of AllSi order in in the book by Farrar and Becker (1971). Applications to solid
silicates and aluminosilicates and brief reviews of the theory and
tetrahedralsites.
methods have been discussed by Lippmaa et al. (1980, l98l),
Our results are in good agreement with the small
Mtiller et al. (1981), Smith et al. (1983), and Kirkpatrick et al.
amount of mostly silicon-29MASS NMR data that has (198s).
already been published for layer silicates(Lippmaa et al.,
1980,19E1;Barron et al., 1982,1983;Smith et al., 1983). Spectrometers
Spectrawererecordedon threehome-builtFouriertransform
Introduction
I Deceased.
0003-o04x/85/050ils37$02.00
(Meadowset al., 1982;Smith et al., 1983).
NMR spectrometers
517
538
K/NSEY ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYER SILICATES
The first one usesa 3.52Tesla4.0-inchbore high-resolutionsuperconductingsolenoid(Nalorac Cryogenics,Concord, CA) together
with a variety of digital and radiofrequencyelectronics,including
a Nicolet (Madison, WI) model 1280computer, 293B pulse programmer,and a Control Data Corporation (Oklahoma City, OK)
9427H disc system.The second instrument uses an 8.45 Tesla
3.5-inchbore solenoid(Oxford), a Nicolet 1180data systemwith
NIC 2090 dual channeltransient recorder,and a Diablo (Diablo
Systems,Inc., Haywood, CA) model-44Bdisc system.The third
instrument usesan 11.7Tesla 2.0-inchbore solenoid(Oxford) and
a Nicolet l28Ol293Bdata systemwith Control Data Corporation
disc. Both the 8.45 and 11.7Tesla instrumentsuse Amplifier Research(Souderton,PA) model 200L amplifiers for final rf pulse
generation,while the 3.52Tesla spectrometeris equippedwith an
Arenburg (JamaicaPlain, MA) model PG 650-C transmitter. All
instrumentsuse Andrew-Beamstype magic-anglesample-spinning
rotor assemblies(Andrew, 1971).At 3.52 Tesla, correspondingto
an aluminum-27 resonance frequency of 39.1 MHz, sample
volume is about 0.2 rnl. and the rotation ratesare from about 3 to
5 kHz. At 8.45Tesla,correspondingto an aluminum-27resonance
frequencyof 93.8 MHz and a silicon-29 resonancefrequencyof
71.5MHz, samplevolumeis either 0.2 ml (aluminum-27)or 0.7 ml
(silicon-29),and rotation rates are about 3-5 kHz (aluminum) or
about 2-3.5 kHz (silicon).At 11.7Tesla,conespondingto an aluminum resonancefrequencyof 130.3MHz or a silicon-29 resonancefrequencyof 99.3 MH4 samplevolume is 0.2 ml, and rotation rates are from about 3 to 5 kHz. The magic angle was set
using 51VrO, or K7eBr, and for the aluminum-27 spectra was
generallyoffsetby 0.5' to eliminatesatellitetransition background
(Oldfield et al., 1982).The 90' pulse widths vary from 1-8 psecfor
aluminum and from 20-30 plsecfor silicon. Chemical shifts are
reported in ppm from external TMS for silicon-29,taken as 6.6
ppm upfield from hexamethyldisiloxane(in C(acac). doped
CHrClJ, and for aluminum-2?are in ppm from the A(HrO):*
ion in a lM solution of AI(HrO).CI. in water at 20"C. Upfield
shielding)are negativefor both nuclci,and the
shifts(increased
on the
errorsare *0.1 to I ppm for silicondepending
estimated
I ppmor greaterfor
linewidthandresolution,
and approximately
aluminumdepending
on linewidth.
Samples and sample preparation
The samplesexamined(Table 1) were selectedto be representativeof both di- and tri-octahedral2:1 phyllosilicates,although kaolinite and gibbsite were also examined.
Major-constituentchemicalcompositionsare availablefor
most samplesand are listed in Table 1. The samplesinclude exampleswith aluminum in octahedralcoordination
only, tetrahedral coordination only, and in both coordinations.Most natural mineralswerefrom the University of
Illinois Mineralogical Collection. Fluorophlogopite is a
syntheticmaterial kindly provided by Dr. Hatten S. Yoder
of the CarnegieInstitute of Washington,GeophysicalLaboratory. Synthetic mica-montmorillonite was kindly provided by ProfessorD. Hendrickson,University of Illinois.
Gibbsite (A(OH)3) was the kind gift of the Alcoa Chemical
Corporation (Bauxite, AK). All clay mineral samples
(exceptthe smectite,which was run as the bulk bentonite)
were particle-sizeseparatedinto the <l pm fraction by
centrifugalsedimentationto removenon-clay minerals.All
sampleswere examinedas finely ground powders (ground
with an agatemortar and pestle)in Delrin rotors. No background probe-Al signals(generallydue to Al in glass)were
visible on any of our spectrometers.All samples were
characterizedby powder X-ray diffractrometry using a
Phillips X-ray diffractometer.Electron paramagneticresonance spectra were obtained at 9 GHz and 23'C for all
sampleswith a Varian (Palo Alto, CA) E-4 spectrometer.
Table 1. Compositions and localities of samples analyznd
Conposltlon
oxygens)
+l
SaupIe
Pyrophyl|1
te
Glbbslte-?
Kaollnitq
Soectlte
(atons/10
Locall
2
Phlogoptte2
FluorophlegoplteIlectgrite
Talclr{uacovlte
Illtte
MlcaMontaorlllonlteMarSarlte
I I I i te / Snect Lte s
A1
3IH
34H
39H
ty
lrVr
Robblns,
NC
Syn the tlc
Washlngton Co.,
Crook Co., WY
,
2
2
t.97
r,47
cA
F.+3
Syflthetic
Unknom
Klnnekulle,
Klnnekulle,
Klnnekulle,
Klnnekulle,
F.42
Mg
0.03
0.18
ET
0.02
0.32
orru
1.99
3.94
o97
1.85
f.66
0.lI
0.03
0.0r
0.05
r.63
1. 6 0
1.44
0.09
0.08
0.10
0.17
t70
3
3.91
4
0.09
0.06
0.30
3.O2
3.43
0.98
0.57
3.37
22
0.67
0.41
0.31
0.33
0.43
tr
tr
0.20
tr
0.12
1
I
J
WY
K
0.0f
0.06
:
I.99
2
Sreden
Sweden
Sweden
Sweden
sl
4
Franklin,
l{J
Synthetlc
Hector,
CA
orange Co., NC
l'lethaef, l\rp,,
Ont.
Wlnd Rlver
Canyon,
Lt*l
3.67
3.79
3.88
o27
0.84
0,77
0.09
0.0r
0.01
0.01
070
I
o.27
0.33
0.21
0.r2
o.43
0.48
o.32
o.22
glbbslte,
atoEs per 5 oxygens;
atons per 6 oxygens.
stolchiometrlc
composltlon.
fron Van Olphen,
It. and Frlplat,
J. J. (1979)
and other
Non-Metalllc
Data Handbook for CLay Mlnerals
Pergaoon Press,
Oxford.
LMlnerala,
'Analysea
gibbsite
froo Van Olphen and Frlplat:
subtracted
CaCOa subcracted
froo reported
malysls
of hectorlce;
analyels
of Dlca-oontnorlllonlte.
.reported
-Analy6ls
(
1
9
7
6
)
froo Roble et al,
J. Res. U.S. ceol.
Survey, 4,631-644.
0.09
o.ll
0.14
o.z5
]f.oltnla.,
iAssuoed
-Analyses
froD
KINSET ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYER SIUCATES
ttsi MASS
Magnetic susceptibilitieswere determinedusing a CAHN
Research(Cerritos, CA) magnetic susceptibility System
6600-1,using the Faradaymethod.
Results
Silicon-29
We show in Figures1,2 and 3 silicon-29MASS NMR
spectra of our samples and summarize in Table 2 the
chemicalshift results.We observea rangefrom - 76.3ppm
for margarite, to -98.0 ppm for talc, in good agreement
with the results of Lippmaa et al. (1980,1981),who found
-98.1 ppm for talc, and -95.0 ppm for pyrophyllite (see
also Barron et al., 1983).The kaolinite value of -92.1 ppm
and the margarite value of -76.3 ppm agree with the
values obtained by Miigi et al. (1981) and Barron et al.
(1983).We do not resolvetwo peaks for our kaolinite, as
do Barron et al. for theirs. Our value of -86 ppm for
ttsi
0
-101
MASS
2:a) a
15C
'2CC
-i-ffi
-100
PPMlromTMS
2c0
539
o
- too
PPM fron TMS
-200
0
- roo
-2@
PPM fiom TMS
Fig. 2. Silicon-29MASS NMR spectraof layer silicatesat 8.45
and ll.7 Tesla. A, Synthetic fluorophlogopite (11.7 T), 4.0 kHz
MASS, 1130scansat l0 sec recycletime, 3.6 psec23 pulse excitation, 21096
data points, zero-filledto 8192,50 Hz linebroadening
due to exponentialmultiplication. B, Phlogopite(11.7T), 4.0 kHz
MASS, ,1471scansat 10 secrecycletime, 3.6 psec90' pulse excitation, 4096data points, zero-fillcdto 8192,50 Hz linebroadening
due to exponentialmultiplication. C, Ilite (8.45T), 2.7 kHz MASS
175 scansat 300 sec recycle time, 16 psec 90" pulse excitation,
2048data points, zero-filledto 8192,50 Hz linebroadeningdue to
exponential multiplication. D, Muscovite (8.45 T), 2.98 kHz
MASS, 13005scansat 168 msecrecycletime, 16.1psec90" pulse
excitation, 4096 data points, zero-filled to 8192, 50 Hz linebroadeningdue to exponentialmultiplication.
muscovite, which we obtained with several different samples, is slightly different from that of Lippmaa et al. (1980),
who reported a poorly split peak with chemical shifts of
- 84.6 and - 86.7 ppm.
Except for the fluorophlogopite, the smaller peaks on
either side of the tall peaks are spinning side-bands. These
c
1CC
PPM from TMS
-204
t"si
MASS 8457
Fig. l. Silicon-2gMASS NMR spectraof layer silicatesat E.45
I L L I T E- S M E C T I T E
and 11.7Tesla.A, Pyrophyllite(8.45T),2.7 kHz MASS,2l scans
at 600 sec recycle time, 30 psec 9O" pulse excitation, 2048 data
points, zero-filled to 8192, 25 Hz linebroadening due to exponential multiplication. B, Talc (8.45 T), l.l4 kHz MASS, 17
scansat 300 sec recycletime,21 psec 90" pulse excitation, 2O48
data points, zero-filledto 8192,25 Hz linebroadeningdue to exponential multiplication. C, Margarite (8.45T), 2.45 kHz MASS,
1325 scansat l0 sec recycletime, 5.3 psec 53" pulse excitation,
4096data points, zero-filledto 8192,25 Hz linebroadeningdue to
exponentialmultiplication. D, Kaolinite (8.45T), 2.66kHz MASS,
3616scansat 1 secrecycletime, 20 psec90' pulseexcitation,2048
data points, zero-filledto 8192,25 Hz linebroadeningdue to exponentialmultiplication. E, Hectorite(8.45T), 2.44kllz MASS, 50
scansat 300 sec recycletime,23 psec 90' pulse excitation, 2048
data points, zero-filledto 8192,25 Hz linebroadeningdue to exponential multiplication. F, Mica-montmorillonite (11.7 T), 4.0
Fig. 3. Silicon-29 MASS NMR spectra of illite-smectites at 8.45
kHz MASS, 2140 scansat l0 sec recycletime, 3.6 psec 23" pulse Tesla. The samples were from a mineralogically zoned Swedish
excitation, 4096 data points, zero-filled to 8192, 50 Hz line- bentonite. Spectral conditions all basically as in Fig. 2{C), except
broadeningdue to exponentialmultiplication.
the number of scans varies.
K/NSEY ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYERSILICA'|ES
Table 2. Siicon-29 (8.45 Tesla) and Aluminum-2? (11.7 Tesla)
MASS NMR chemicalshiftsof samolesexamined
SlLicon-29
Chenical
Shifta
(pp.)
Sanple
Pyrophyllite
-95.0
Kaolinite
Snectite
-92.1
N.D.
-a1
- 4 5, 2 , - 4 9 . 2 , - 9 2 . 7
-95.3
-98,0
Phloaopit€
Fluorophlogopite
Talc
I l11te
Mica-MontnorllIonite
Margari te
Mixed layered
Illi te-SDec tltes
A-l
31H
34tr
39{
a1n pf
froD
D1n ppm from
Terrahedrat
Aluninun-27
chentcal
Shifrb
( PpnJ
external
external
-86
-9r
_33c,_87c,_93
-76,1
-93
-94
-95
-95
Me4S1.
lM Ar(H2o)6clj
II,7 T
.'AI
OcEshedral
Alunlnun-27
Chenical
Shlftb
( ppo)
MASS
A t ( O H) 3
69
69
P y r o p h yIil f e
t2
70
70
14
a t 2 s i 4 o t o( o H) 2
2,-10
70
69
69
69
in
B2O,
ariseDiathe magic-anglespinning process,and can be distinguished from the true centerband resonancesbecause
their positionsand intensity vary with spinning speed.The
small peaks for fluorophlogopite are all centerbandresonances,indicating that nonequivalentsilicon sitesare present.
Spinningside-bandsfor silicon-29may be causedby the
presenceof a chemicalshift anisotropy(CSA)(Smith et al.,
1983),by various types of magneticimpurities (most commonly iron either in solid solution or in a separatephase
Oldfield et al., 1983),or by dipolar interactions,as interpreted by Grimmer et al. (1983)for an Fe2*-dopedolivine.
The increase in sideband intensity with magnetic field
strength suggeststo us that the effect of segregatedmagnetic domains dominatesour samples,although small contributions from substitutionalimpuritiesare difficult to rule
out. The presenceof substitutionalparamagneticimpurities
is expectedto affectisotropic chemicalshifts(with a Curielaw dependence),
thus, great care is needeedin interpreting
small (1-2 ppm) differencesin chemicalshift for naturally
occurringmineral phases.The effectof paramagneticimpurities would be to reduce the shielding (i.e., cause more
positive or lessnegativechemicalshifts.The inherent CSA
also probably makesa contribution to the side-bands,but
in most instancesthe side-bandsare quite prominent and
correlatewell with the presenceof large magneticsusceptibilities. Basedon our frequencydependencedata, we thus
believethat they ariseprimarily from the presenceof small
amounts of magneticimpurity domains and that their appearanceis unlikely to aflect the value of the isotropic
chemicalshift (oldfleld et al., 1983).
Aluminum-27
We show in Figures 4 through 7 the aluminum-27
MASS NMR sp€ctra of our samples,and summarizein
Table 2 the chemicalshifts observed.Figure 4 shows the
20c
o
-204
PPM frorn AlCl3
Fig. 4. Aluminum-27 MASS NMR spectra of layer silicates
with octahedrally coordinated aluminum, and gibbsite, at 11.7
Tesla. A, Gibbsite, 4.2 kHz MASS, 500 scans at 2 sec recycle time,
3 psec 90' pulse excitation, 4O96 data points, 50 Hz linebroadening due to exponential multiplication. B, Pyrophyllite, 4.5
kHz MASS, 200 scans at 100 msec recycle time, 3 psec 90" pulse
excitation, 8192 data points, 20 Hz linebroadening due to exponenrial multiplication. c, Kaolinite 4.2 kHz MASS, 1000 scans
at 15 sec recycle time, 3 psec 90" pulse excitation, 4096 data
points, zero-filled to 8192, 50 Hz linebroadening due to exponential multiplication.
ll.7 T aluminum-27 MASS NMR spectra of gibbsite
(A(OH)3), pyrophyllite (Al2Si4Oro(OH)r) and kaolinite
(Al2Si2O5(OH)o),all of which contain only octahedrally
coordinatedaluminum (A(6). Gibbsite showsa singleresonance,consistentwith only one type of aluminum site.
The chemicalshift of 7.8 ppm is consistentwith previously
reported values for octahedrally coordinated aluminum
(Mtiller et al., 1981;de Jonget al., 1983).Magneticsusceptibility measurements
on this sampleshow it to be diamagnetic, and it exhibits no EPR resonanceat 9 GHz and
23'C. The large sidebandsare presumably due to Al-Al
and AIOH dipole-dipole interactions. Field dependence
data (not shown)confirm this.
The aluminum-27 MASS NMR spectra of pyrophyllite
and kaolinite also show spectrawith significant sideband
structure(Fig. a). Thesesidebandsbecomelesspronounced
at lower fields (data not shown)and thus might be attributable to chemicalshift anisotropyeffectsrather than dipole
or second-orderquadrupoleinteractions.However,both of
thesenatural mineral samplesexhibit modestmagneticsus-
KINSEY ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYERSILICATES
ll,7 T
27Al
MASS
A
P h l o g o pr ei
K M e . ( A l S i . O , o( )O H ) ,
F- Phlogopite
541
exclusivelytetrahedralcoordination,in accordwith the acceptedcrystal structurefor this material, best viewed of as
being derivedfrom talc with partial Si+ Al(4) substitution,
and the presenceof charge-balancinginterlayer potassium.
The chemical shift of 69 ppm from AI(H2O)!+ for fluorophlogopite is also consistentwith solely tetrahedralaluminum coordination (Fyfe et al., 1982;Miiller et al., 1981;
de Jonget al., 1983;Kirkpatrick et al., 1985).For hectorite,
the peak at 66 ppm is also consistentwith tetrahedrally
coordinated aluminum, indicating that much of the trace
( A l S i sq o ) F 2
KMS3
r r , zr 2 1 rMASS
66,.
Heclorile
(Mg,Li )s S14oto(OH)aN%.3(HzO)4
B
?oo
o
-zoo
PPM from AlCl3
Fig. 5. Aluminum-27 MASS NMR spectra of layer silicates
containing tetrahedrallycoordinatedaluminum, at 11.7 Tesla.A,
Phlogopite,4.1 kHz MASS, 4155scansat 100msecrecycletime, 3
psec90' pulse excitation,4O96data points, 50 Hz linebroadening
due to exponentialmultiplication. B, Syntheticfluorophlogopite,
4.4 kHz MASS, 600 scansat 30 secrecycletime, 3 psec90" pulse
excitation, 8192 data points, 25 Hz linebroadeningdue to exponentialmultiplication. C, Hectorite,4.1 kHz MASS, 6400scans
at 100 msec recycletime, 3 psec 90" pulse excitation, Sl92 data
points,25 Hz linebroadenigdue to exponentialmultiplication.
c
ceptibilities (Xu of 2.0 and 0.78 x 10-6 cgs units for pyrophyllite and kaolinite, respectively), and both have weak
EPR absorptions, due at least in the case of kaolinite to
Fe3* and lattice defects (Pinnavaia, 1982). As we shall
show below, these results, together with the approximate
equivalence of the overall aluminum-27 and silicon-29
spectral breadths for pyrophyllite (Ao approximately 60
ppm) strongly suggests for both materials a bulk magnetic
susceptibility broadening, with concomitant sideband formation upon magic-angle sample-spinning (Oldfield et al.,
1983).Nevertheless,the 11.7 Tesla aluminum-27 chemical
shift values of 4.0 and 4.4 ppm are consistent with purely
octahedral Al coordination in both kaolinite and pyrophyllite.
We show in Figure 5 the 11.7 Tesla aluminum-27
MASS NMR
spectra of the minerals phlogopite
(KMg.AlSi.Oro(OH)r) and hectorite (a lithium montmorillonite, (Mg,Li)3Si4Olo(OH)r.Nao..(HrO)o with only a
small amount of tetrahedrally coordinated Al) together
with
the spectrum
of
synthetic
fluorophlogopite
(KMg.AlSirO,oFr).
The aluminum-27 results for phlogopite indicate Al in
E
D
?oo
o
-200
PPM fron AlC13
Fig. 6. Aluminum-27 MASS NMR spectra of layer silicates
containing mixed aluminum coordination, at ll.7 Tesla. A, Synthetic mica-montmorillonite,3.4 kHz MASS, 900 scansat 1 sec
recycletime, 3 psec90" pulse excitation,4@6 data points, 50 Hz
linebroadeningdue to exponentialmultiplication. B, Smectite,4.3
kHz MASS, 4000 scansat 100msecrecycletime, 3 psec90" pulse
excitation, 4096 data points, 100 Hz linebroadeningdue to exponential multiplication. C, illite, 3.6 kHz MASS, 3400 scansat
100 msec recycle time, 3 psec 90' pulse excitation, 4@6 data
points, 50 Hz linebroadeningdue to exponentialmultiplication.D,
Margarite, 3.7 kHz MASS, 3900scansat 100msecrecycletime, 3
psec90' pulse excitation, 4O96data points, 50 Hz linebroadening
due to exponentialmultiplication. E, Muscovite,3.6 kHz MASS,
700Oscansat 100 msec recycletime, 3 psec 90" pulse excitation,
4O96data points, 50 Hz linebroadeningdue to exponentialmultiplication.
542
KINSET ET AL.: NUCLEAR MAGNETIC RESON/NCE SPECTROSCOPY
OF LAYERSILICATES
lllile- Smectite
?oo
100
0
PPI'I ftm
-100
-zoo
AlCl3
Fig. 7. Aluminum-27 MASS NMR spectraof mixed layer illitesmectites,
at 11.7Tesla.A, A-1, 3.6kHz MASS,2000scansat 100
msecrecycletime, 3 psec90'pulse excitation,4096data points, 50
Hz linebroadeningdue to exponentialmultiplication. B, 3lH, 3.6
kHz MASS, 2000scansat 100msecrecycletime, 3 psec90. pulse
excitation, 4096 dzta points, 50 Hz linebroadeningdue to exponential multiplication. C, 34H, 3.3 kHz MASS, 3000 scansat
100 msec recycle time, 3 psec 90" pulse excitation, 4096 data
points, 50 Hz linebroadeningdue to exponentialmultiplication. D,
39H, 3.5 kHz MASS, 2000 scansat 100 msecrecycletime, 3 psec
90' pulse excitation,4@6 data points, 50 Hz linebroadeningdue
to exponentialmultiplication.
aluminum must occupy four-coordinatepositions,i.e. that
it substitutesfor Si rather than (Mg,Li). The shoulderat 59
ppm is probably due to a small amount of feldspar impurity (Kirkpatrick et al., 1985).
A dramatic differencebetweenour resultsfor phlogopite,
fluorophlogopiteand hectoriteis their very large difference
in spinning-sidebandstructure.We believethat such differencesoriginate in the different levelsof magneticimpurities
found in the natural mineral samplesand the synthetic
material. We find magneticsusceptibility(X")valuesof 2.7
x 10-6 cgsunitsfor phlogopite,andO.6T
x-10-6 cgsunits
for hectorite, together with clear evidenceof Fe3+ and
Mn2* EPR absorptionsin phlogopite,and a broad,weak
absorption at a g-value of about 2 for hectorite, but no
appreciablemagneticsusceptibilityor EPR absorption for
the fluorophlogopite. Thus, there appears to be a good
correspondence
betweenthe presenceof spinning-sideband
structure and the presenceof magnetic impurities, as we
observedfor the octahedrallycoordinatedaluminum (Fig.
4). Moreover,our resultsindicatea closesimilaritv between
the actual magnitudesof the aluminum-27 and silicon-29
magnetic shift anisotropiesof both phlogopite and heptorite, consistentwith this type of broadeningmechanism.
Figure 6 showsthe 11.7T aluminum-27MASS NMR
spectraof phyllosilicateswith both octahedral and tetrahedral aluminum. These spectra clearly show both Al(6)
resonances,in the range 2.5 to 4.2 ppm, and Al(4) resonances,in the range7O.2to 74.5ppm (Miiller et al., l98l;
de Jong et al., 1983).The smectitedoes not show a well
defined Al(4) resonancebecauseof the low Al(4) concentration. The spinning side-band structures are quite
variable.The extremesare the muscovite,with many, and
the synthetic mica-montmorillonite,with few. All natural
samplesdo, however,have large magnetic susceptibilities,
x8, ranglng from about 2.0 to 5.6 x 10-6 cgs units. There
are also EPR absorptionsat 9 GHz and 23"C for muscovite and smectite.Theseresultsagain suggesta large magnetic susceptibility broadening contribution to the observedsidebandstructurein the natural minerals.
We show in Figure 7 the ll.7 T aluminum-27 MASS
NMR spectraof four mixed-layeredillite-smectitesfrom a
mineralogicallyzoned Swedishbentonite. The proportion
of illite in theseillite-smectitesrangesfrom 650/oin A-l to
4O% in 39H. There is a concomitant change in composition, including the amount of 4l(6), as indicatedby structural formula calculations (Table 1). All spectra show a
large peak due to octahedralaluminum,and a smallerpeak
of varying rclative intensity due to tetrahedral aluminum.
Once again, the sidebandsarise from magneticimpurities,
as judged from 1, values and EPR spectra (data not
shown).
In order to obtain the most accurateA(4yAl(6) coordinate ratios, it is necessaryto understandthe origins of all
spectral featuresobserved,and to answer the questionwhat magnetic field strength is best for such determinations?We shall now considerthis matter briefly.
For quadrupolar nuclei, the most likely source of
linebroadeningis the second-orderquadrupole interaction
(Nolle, 1977; Meadowset al., 1982;Miiller et al., 1981),
which decreases
with increasingmagneticfield strengthand
with samplespinningalthough to only about ll4 to 1/10of
the breadth of the spectrum of the static sample. The
secondmost likely sourcesof static linebroadeningare dipolar and chemical shift anisotropy broadening, with
averageto zero under very fast magic-anglerotation. In
addition, as we discussedabove, bulk susceptibility(and
also potentially dipolar) broadeningand sidebandformation may also contribute for natural mineralshaving magnetic impurities. This effect increaseslinearly with increasing field strengthwhen magneticdomainsare present(Oldfield et al., 1983).Thesepoints are illustrated by the sctsof
aluminum-27MASS NMR data taken at ll.7 and 3.52
Tesla shown in Figure 8. These spectra show decreased
peak widths at high field (decreasedresidualsecondorder
breadth), but more intense spinning sidebands,due to a
large susceptibilitybroadening.In addition, the chemical
shifts are observedto be field dependentdue to a second
order quadrupolar shift, with the effect being largest for
543
KINSET ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPYOF LAYER SILICATES
II,7 T
,'AI
MASS
3.5 T
Mico - Montmoiillonlle
Mutcovile
200
a
PPMtt@Alcb
-200
200
0
PPMl@ ,{C!
-2ao
Fig. 8. Aluminum-27 MASS NMR spectraof layer silicatesat
3.5 and 11.7Tesla.A, Pyrophyllite,(11.7T), 4.5 kHz MASS, 5m
scansat 2 sec recycletime, 3 psec90" pulse excitation,4096data
points, 50 Hz linebroadeningdue to exponentialmultiplication. B,
Pyrophyllite (3.5T), 3.1 kHz MASS, 3600scansat 250 msecrecycle time, 3 psec 90' pulse excitation, 4@5 d,atapoints, 100 Hz
linebroadening due to exponcntial multiplication. C, Micamontmorillonite (11.7T\,3.4 k}Iz MASS, 900 scansat 1 secrecycle time, 3 psec 90" pulse excitation,4096 data points, 50 Hz
linebroadening due to exponbntial multiplication. D, Micamontmorillonite (3.5 T), 3.9 kHz MASS, 8mm scansat 100 msec
recycletime, 3 psec90' pulseexcitation,4096data points, 100 Hz
linebroadeningdue to exponentialmultiplication. E, Illite (11.7T),
3.6 kHz MASS, 3,!00 scansat 100 msec recycletime, 3 psec 90'
pulse excitation,4A96 dala points, 50 Hz linebroadeningdue to
exponentialmultiplication. F, Illite (3.5 T), 3.4 kHz MASS, 10000
scansat 100 msec recycletime, 3 psec90" pulse excitation,4096
data points, 100 Hz linebroadeningdue to exponentialmultiplication. G, Muscovite(11.7T), 3.6kHz MASS, 7000 scansat 100
msecrecycle tirne, 3 psec 90" pulse excitation,4@6 data points,
100Hz linebroadeningdue to exponentialmultiplication. H, Muscovite (3.5 T), 3.9 kHz MASS, 10100scansat 100 msec recycle
time, 3 psec 90' pulse excitation,4096 data points, 100 Hz linebroadeningdue to exponcntialmultiplication.
as Ho. For our materials,the overall effectis an increasein
resolution with increasingfield strength, albeit with the
production of more intensespinningsidebands.
The resultsin Figure 8 also indicatea slight differencein
the responseof A(a) and A(6). The apparent chemical
shifts of both A(4) and A(6) increasewith increasingfieldstrength,due to the decreaseof the second-orderfrequency
shift (Abragam,1961),but the effectis about twice as large
for Al(4) as for Al(6). This differencein behavior,together
with the increase in relative peak height (decreased
breadth)for the Al(4) resonancesindicatesa larger nuclear
quadrupole coupling constant for the tetrahedral sites.
Comparion with data obtained at 8.45 Tesla (data not
shown)indicatesthat the 11.7Teslachemicalshiftsare very
closeto the unperturbedvalues.
The results in Figure 8 clearly indicate that optimum
resolution of Al(4) from Al(6) is obtained at the highest
magneticfield strengthspossible--€venin samplescontaining magneticimpurities.The number of linesis greaterbut
they are narrower. Notably, however, even at high field
considerablecare needsto be taken to choosea spinning
speedand spinning angle that yields optimum resolution.
Figure 9 comparesa static aluminum-27NMR spectrumof
illite-smectiteA-1, a spectrum taken exactly "on angle",
and a spectrum"off angle" by 0.5". Optimum resolution is
obtained 0.5' off angle,which reducesthe number of satel-
l l l i t e - S m e c t i t eA - l
A
B
3,85 kHz
200
- 100
0
1'J0
PPMfrom AlCl3
-20a
'on angle", and MASS "off anglc"
Fig. 9. Static, MASS
of lllite-SmectiteA'1, at ll.7 Tesla.A'
NMR
spectra
aluminum-27
Al(4). The linewidths for Al(4) also seemto decreasemost
Static. 6000scansat I00 msecrecycletime, 3 psec90" pulse excion increasingfield strength.
tation, 4096data points, 50 Hz linebroadeningdue to exponeiltial
Theseresultsare all consistentwith the dominanceof a multiplication. B, MASS "on angle", 3.85kHz MASS, 20fi) scans
second-orderquadrupolar linebroadening mechanism at at 100 msec rccycle time, 3 psec 90' pulse excitation' t1096data
low field (3.52T), and a magneticsusceptibilityanisotropy points, 50 Hz linebroadeningdue to exponentialmultiplication.C,
one at high field (11.7T). Resolutiondue to the dpcreasein MASS "off angle",3.53kHz MASS, 6000scansat 100 msecrecythe second order broadening increasesat H3, while the cle time, 3 psec 90' pulse excitation, 21096data points, 50 Hz
decreasein resolution due to magneticimpurities increases linebroadeningdue to exponentialmultiplication.
544
KINSEY ET AL.: NUCLEAR MAGNETTC RESONANCESPEC.TROSCOPY
oF LAYERSILICATE|
lite side-bands,and at a spinning speed which puts the
Al(4) peak betweenthe side-bandsof the large Al(6) resonance.Spinning slightly off angle broadensthe side-bands
of the satellitetransitions,eliminating them from the spectrum, as discussedby Oldfield et al. (1983).
TetrohedrolAluminum
E
o
a
o
l"u
T
I
Discussion
E
lA-l
-lura
l r L l l t,,
T
The major objectivesin this work are to determinehow
3rHll34
r-pri-l3sH
rf
the aluminum-27 and silicon-29chemicalshiftsand silicon29 peak breadthsof well characterizedlayer silicatesvary
Ire
r{
with structure and composition,and to begin to use these
resultsto understandthe structuresand crystal chemistries
of unknown materials. Our results to date indicate that
o,,o]'*,,
MASS NMR can be usedto examinethe following crystal
Fig. 11.Plotof 11.7Teslaaluminum-27
MASSNMR chemical
chemicalfeaturesof layer silicates:Al/Si ratio in the tetra- shiftsof tetrahedral
aluminumversusSi/(Si+ A(4)),a measure
of
hedral sites,Al/Si ordering in the tetrahedralsites,and the thecomposition
of thetetrahedral
layer.
partitioning of aluminum between tetrahedral and octahedral sites.
(lessnegativechemicalshift) with increasingAl(4). The ocTetrahedral Al and Si: chemical shifts
tahedral cation appearsto have little systematiceffect,beThe AfSi ratio in the tetrahedral sites appearsto affect causealthough pyrophyllite and talc are 3 ppm apart, musboth the silicon-29chemicalshift and that of aluminum-27 covite and phlogopite have nearly the samechemicalshift.
in tetrahedralsites.Figure 10A shows the relationship be- Much of the scatteris probably due to inaccuraciesin the
tween the silicon-29chemicalshift and Si/(Si + Al(4)) ratio chemicalanalyses.
obtained from chemicalanalysis,or the stoichiometricforThe trend of Figure 10A is qualitatively consistentwith
mula if the material has not been analyzed.The data for the model of Smith et al. (1983) that relates increased
kaoliniteare not includedbecauseit doesnot havethe 2:1 shielding (increasinglynegative silicon-29 chemical shifts)
structure of the other materials.The resultsin Figure 10A to increasingvaluesof the sum of the cation-oxygenbond
show a clear decreasein the shieldingat the silicon nulceus strengthsof the four oxygenatoms surroundingthe silicon.
It is also similar to the deshieldingefect observedwith
increasingaluminum content in framework silicates(Lippmaa et al., 1981).Unfortunately, accuratestructure refinements are not availablefor most of thesematerials.so the
cation-oxygen bond strength sums cannot be calculated.
o!
$6
As with the framework silicates,however,increasingtetra8i
dil
hedral aluminum content leads to decreasing bondI
strength sums and, therefore, to less negative silicon-29
chemicalshifts.
The Si/(Si + A(4) ratios also appear to be correlated
Si+AH)
with the chemicalshift of aluminum(4).Resultsare plotted
in Figure 11.Note that the aluminum-27chemicalshifts of
the trioctahedrallayer silicatesapparentlyfall about 3 ppm
below the trend for the dioctahedral layer silicates.Although a tentative conclusion,it appears that Al in the
qd
ia
octahedral sites causesa larger deshieldingat Al(4) than
does Mg. It also again appears that aluminum in tetrahedral coordination causesan overall deshieldingeffectthis time at Al(4) itself. The relationship in Figure 11
should be usedwith care,however,becausechemicalshifts
Fig. f0. Correlation betweensilicon-29chemicalshift and cryswere obtained at a magnetic field strength of 11.7 Tesla
tal chcmicalstructure.A, Plot of silicon-29MASS NMR chemical
will be lower (lesspositive)at lower field strengths.In
and
shifts versusSi/(Si+ A(a)), a measureof the composition of the
addition,
if the margarite point is eliminated,and the error
tetrahedrallayer. For fluorophlogopiteand mica-montmorillonite
bars increasedby I ppm, the correlation weakensconsideronly the largestpeak is plotted. B, Plot of silicon-29MASS NMR
chemical shifts versus (interpreted) number of next-nearest- ably.
Finally, our results show that there is no systematicreneighbor tetrahedrally-coordinated
aluminums.For fluorophlogopite and mica-montmorillonite all three peaks or shouldersare lationship betweenthe aluminum-27chemicalshift of octaplotted. For illite, the two shifts from a spectral simulation arc hedrally coordinated aluminum and the Si/(Si + Al(4))
ploned.
ratio. Although octahedralaluminum in margariteis slight-
s5
I
OF LAYER SILICATES
KINSEY ET AL,: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
ly more shielded,the chemical shifts of octahedral aluminum in the other materialsare indistinguishable.
Tetrahedral Al and Si: order and disorder
One questionabout layer silicatesthat has beendiffrcult
to answer using diffraction methods is the extent of Al/Si
ordering in the tetrahedralsites.This question can be addressedusing silicon-29 NMR, becausethe presenceof
multiple peaks, and to some extent of increased peak
breadths,indicate a range of shieldingsat the silicon nucleus,and thereforea range of chemicaland structural environments. Becauseof poorer spectral resolution, such
questionsare more diflicult to answerwith the quadrupolar nuclei,suchas aluminum-27.
Such relationshipsare well understoodin feldspars.For
ordered low albite and microcline the three crystallographically-distinct silicon sites appear as narrow, resolved peaks with full-widths at half-height (FWHH) of
about I ppm. Thesesiteshaveeither one or two aluminum
neighbors(Lippmaaet al., 1980;Smith et al.,
next-nearest
1983; Kirkpatrick et a1.,1985). In disorderedhightemperaturesanidine,however,the resonancesare broadened into one peak with two poorly defined maxima,
having a FWHH of about 10 ppm. Similar phenomenadue
to Si/Al disorderingseemto occur in the layer silicates.
The narrowestsilicon-29peaksare for the layer silicates
with no tetrahedralaluminum (talc, pyrophyllite,and kaolinite). For thesesamplesthe FWHH's are 2 to 4 ppm. The
slightly increasedbreadth of theseresonancesover thoseof
the feldsparsmay be due to broadening from magnetic
impurities. For margarite, which has Si/(Si + Al(a)) of
about 0.5,the FWHH is almost as small (4 ppm), and there
is only one peak, which seemsto imply that margarite
contains an ordered arrangementof aluminum and silicon
in the tetrahedral sites. Thus, each silicon has three aluminum next-nearestneighborsand no silicon next-nearest
neighbors(a Q3(3Al) site in the nomenclatureof Lippmaa
et al., 1981).BecauseSi/(Si + A(a)) is 0.5 for margarite,this
arrangementimplies that no tetrahedral aluminum has a
tetrahedral aluminum next-nearestneighbor, and that the
aluminum avoidanceprinciple,which statesthat aluminum
atomsavoid eachother as next-nearestneighbors(Loewenstein, 1954),is obeyed.Farmer and Velde (1973)have previously provided evidenceof tetrahedral ordering in margarite on the basisofsharp infrared spectraand the lack of
Al-O-Al vibrations, and Guggenheimand Bailey (1975)
have refined a margaritein the spacegroup Cc, indicating
full ordering.flowever, Guggenheimand Bailey (1975)also
found that for their sample of margarite the two tetrahedral sheetswithin the 2:l layer werenot compositionally
identical, having Al(4) contents of 1.006and 0.877.If this
were true of our sample, the silicon-29 spectrum would
have an additional peak attributable to Q3(2Al) silicon
sites,which it doesnot. To ensurethat we are not missinga
peak with a very long relaxationtime (Tr) we ran a silicon29 spectrumof margarite assuminga T, of 4(X)0seconds.
The spectrumis identicalto that in Figure lC.
For muscovite and phlogopites, the silicon-29 NMR
data suggestthat there is more than one silicon site Present
and, therefore,that Al/Si ordering in the tetrahedrallayers
is not perfect.Extremelystrong evidenceof this is the spectrum of syntheticfluorophlogopite(Fig. 2A). The spectrum
consistsof a central peak at -90.2 ppm and smaller,but
equal sizedpeaksat -85.2 and -92.7 ppm. Theseresonancesare not spinning side-bands,becausethey are not
disposedat the spinning frequencyand do not shift position with spinning speed. Becausephlogopite has Si/
(Si + Al(4)) : O-75,in a perfectlyorderedsampleall silicons
would have one aluminum next-nearestneighbor(Q11AD).
We interpret the large peak at -89.2 ppm as due to these
Qt(lAl) sites,and by analogywith the previouslyobserved
deshieldingeffect of Al(4) in zeolites and feldspars,the
small peak at -92.1 ppm as due to Q3(0Al)sites,and the
small peak at -85.2 ppm as due to Q3(2Al)sites.The small
peaksare equal in height,becausetheir abundancemust be
the samein order to maintain an averageQt(lAl) composition. The presenceof Q3(0Al) and Q3(2Al) sites in this
sample is very strong evidencefor a lack of perfect Al/Si
ordering in the tetrahedral sites. From the simulation of
the spectrumin Figure 2A, it appearsthat about 46% of
the silicon sites are not Q3(1Al).The spectrumof fluorophlogopiteis very well resolved(FWHH about 2 ppm) due
to its lack of magneticimpurities.
The silicon-29 MASS NMR spectra of the natural
phlogopite and muscovite(Figs. 2B and 2D) are also consistent with some Al/Si disorder, although becausethe
large peak breadths observed could arise entirely from
broadeningdue to magneticimpurities, it is impossibleto
be more definitethan this.
The spectrum of the synthetic mica montmorillonite
(Fig. lF) also shows signalsfrom three silicon sites.This
sample has Si/(Si + A(4) between 1.00 and 0.75 so the
"average" silicon site should be between Q3(0AD and
Qt(lAl), and the material must contain both types of sites'
We interpret the largest peak, at -93 ppm, to be from
-87 ppm to be from Qt(lAD
Qt(OAl)sites,the shoulderat
sites,and the small shoulderat -83 to be from Qt(2Al)
sites. In this spectrum the presenceof the Q3(2Al) sites,
which are not required by stoichiometry,indicatessignificant Al/Si disorder in the tetrahedral layers. (Again, we
assumeno macroscopicheterogeneity.)
The silicon-29MASS NMR spectraof the Swedishillitesmectites(Fig. 3) are also consistentwith the presenceof
multiple silicon sites,and potentially Al/Si disorder on the
tetrahedral sites.These clays have Si/(Si + Al(4)) between
O.97 and 092 and, consequently,should have mostly
Q10Al) silicon sitesand someQs(lAl) sites.
We have attemptedto simulateeachof the illite-smectite
spectra shown in Figure 3. Many combinations of peak
height and breadth gave reasonablefits, but all successful
simulationscontain a large peak at about -93 to -96
ppm (Q3(0Al)), and smaller peaks at about -87 ppm
The Q3(1Al)peakwas
(Q'(1Al))and at -83 ppm (Q3(2Al)).
always larger than the Qt(2Al) peak. Theseresults again
suggestthe presenceof Al/Si disorder in the tetrahedral
sites.
546
KINS.EY ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYER SILI1ATEY
The silicon-29 MASS NMR spectrum of hectorite,and
its simulation,indicate mostly e3(0Al) and a small number
of Q3(lAl) silicon sites.The material has Si/(Si + Al(4)) of
about 0.95,requiring almost all Q10Al) sites,in agreemenr
with the NMR data.
Becausea number of the phaseswe have examinedcontain more than one type of silicon site, we have replotted
the silicon-29chemicalshifts versusnumber of next-nearest
neighbor aluminums in the tetrahedral layers (Fig. l0B).
This plot contains the three sites in the synthetic fluorophlogopite and mica-montmorilloniteand the two sitesin
illite (Fig. 2C).The correlationis excellent,and is similar to
that of Lippmaa et al. (1981)for tectosilicates.This presentation of the data also explains why the micamontmorillonite falls off the trend of the plot of silicon-29
chemicalshift versusSi/(Si + A(a)) (Fig. 10A).That is, the
largestpeak,which is the one plotted in Figure l0A, correspondsto a Q3(0Al)site and not to the bulk composition.
Al(4)lAl(6)
| 1,7 T
2hl
i l l i f e - S m e c t i f eA - l
ratios
Another important structural parameter for layer silicates is the A(a)/Al(6). Until now, this ratio, like Si/
(Si + Al(4)),could only be determinedby chemicalanalysis
and allocation of the componentsinto the availablesitesby
calculatingidealizedstructural formulae.
Using spectral simulations and integrations of the
aluminum-27MASS NMR spectra,however,it is now possible to determinethe relative amounts of tetrahedraland
octahedral aluminum. Spectra obtained at 11.7 T (or
above)should give the best results,becausethey have the
best resolution.Our simulationswere done using the Nicolet curve fitting routine, CAP. This program allows for
input of up to 28 separatelines of varying intensity,width,
and position. Lorentzian, Gaussian, or Voight linebroadeningfunctions are available.The program is interactive, and gives an rms %oerror as a measure of the
goodness-of-fitof the simulation.
The spectraof fluorophlogopite,pyrophyllite, kaolinite,
and phlogopite,all of which have only one aluminum site,
were simulated to help determinetypical variations in the
amount of Lorentzian or Gaussiancharacterof the lines.
Figure 12 showsa typical aluminum-27MASS NMR spectrum for illite smectiteA-1, together with its simulation.
The simulation uses a 100% Lorentzian linebroadening
and has an rms error of 4.3%. The largest discrepancy
betweensimulation and experimentappearsto be that the
experimentallines are not exactly symmetrical,generally
being skewed to the low ppm side, due presumablyto a
small residualsecondorder broadeningeffect.The simulation in Figure 12 yields an A(a)/Al(6) ratio of 6.2,in good
agreementwith the value of 5.7 determined by chemical
analysis.
Similar simulations have been carried out for all the
samplesthat contain both tetrahedraland octahedralaluminum. Table 3 lists the samples studied and the
4l(4)/4l(6) ratios determinedby chemical analysisand by
NMR. The NMR determined numbers are the average
values obtained from simulations of experimentalspectra
PPT
Fig. 12.Aluminum-27
MASSNMR spectraand simulationof
illite-smectite
A-1.The sp€ctrumwastakenwith 3.8kHz MASS,
4000scansat 100 msecrecycle,4O96data points and 50 Hz
linebroadening
dueto exponential
multiplication.
The simulation
used100%Lorentzianlinebroadening.
The numbersindicatethe
orderof spinning-sideband
for Al(6).
taken at severalspinning speeds,and with differentspectral
phasings.For most, the resultsfrom two or three spinning
speedswereaveraged.
As may be seenfrom Table 3, the agreementbetweenthe
A(4yA(6) ratios determinedby chemicalanalysisand by
NMR is rather good. The only sampleswith significant
disagreementare the margarite and the 31H illite-smectite.
With theseexceptions,the rms errors in the percentAl(4)
determinedby NMR rangefrom 2 to 4%. The lowesterror
is for the syntheticmica montmorillonite, which has only a
very small spectral overlap between the octahedral and
tetrahedr4lresonances.
The highesterror is for muscovite,
Table 3. ComparisonbetweenA(4)/A(6) ratios determinedby
chemical
analysis
andby NMR
Ar(4)/A1(6)
Ar(4)/Ar(6)
Chenical
Analyala
Muacovlte
lU lte
lllce-MontnoEllLonite
HerS6rlte
Mlx€d layered
I1l1te-S[ec t ites
A-I
318
34rl
39H
1.9
2.9
1.00
1.9
3.0
0.56
4.9
10.1
12.0
OF LAYER SILICATES
KINSEY ET AL.: NUCLEAR MAGNETIC RESON/4NCESPECTROSCOPY
for which both sidebandpatterns are very broad (about
100 ppm), and thus overlap considerably.Changing the
percent Gaussiancharacter of the peaks does not significantly affect the 4(4)/A(6) ratios determined.In general,
the rms errors rangefrom 3.5 to 9Vo,with 7% being about
average.
The margaritesampleoffers somepuzding resultsin addition to the poor agreement between the NMR and
chemicalcomposition.Careful selectionof spinning speed
permits resolution of two octahedrallines, one at 4 ppm,
the other at - l0 ppm. The - 10 ppm line and its sidebands account for about 7"/oof the total aluminum intensity. In the determinationof the Al(4)/,{l(6) ratios, this intensity was included in the amount of octahedral aluminum. The chemicalshift of - 10 ppm is unusual,being
much more shieldedthan the octahedrallines for aluminas
(R. A. Kinsey, unpublisheddata) and other sheetsilicates.
The relationshipsbetweenSi/(Si + A(a)) and silicon-29
and aluminum-27 chemicalshifts (Figs. 10 and 11)and the
ability to calculateA(4yA(6) ratios offer an opportunity to
very rapidly and accuratelyobtain a greatdeal ofstructural and compositional data for layer silicates. Because
Si + Al(4) must total four (neglectingFet*) in a ten-oxygen
formula unit, NMR uniquely defines the composition of
that layer.
An important potential application of aluminum-27
NMR spectroscopyof sheetsilicatesis in the examination
of the clay size fraction of shalesundergoingburial diagenesis.Hower et al. (1976)have shown that the percent of
illite layers and AlrO. in mixedJayer illite-smectitesfrom
the U.S. Gulf Coast increasein a complex way with depth.
This change in aluminum content and any associated
changesin A(a)/Al(6) partitioning should be detectableby
NMR methods.
Finally, we note that, perhapsevenmore than usual with
NMR, considerablecare must be exercizedin the acquisition and interpretation of aluminum-27 MASS NMR
spectrawhen quantitativeresultsare required.For the materials we have investigated,high field operation yields
chemical shifts that are essentiallythose which would be
obtained in the absenceof any second-orderquadrupole
shift-the samecannot be said for resultsobtainedat lower
field, e.g., with 200 MHz (rH equivalent) instruments,
which are presentin many laboratories.Moreover, we are
fortunate that the quadrupole coupling constantsin clays
appearto be fairly small (about l-2 MHz), but are not too
small. For example,in AI(H2O).CI., all three transitions
may be excitedby an r.f. pulse,causingintensity anomalies
(data not shown,seee.9.,Schmidt,1971)while with many
other minerals(Ghoseand Tsang, 1973),e'zqQ/hvaluesare
very large (up to 10 to 15 MHz), resulting in complex
sideband patterns for the central transition in a MASS
NMR experiment.Clearly,it is necessaryto obtain spectra
for Al(4)/4l(6) quantitation at severalpulse width settings
in order to ensure that both sites behave in the same
manner. When very complex mixtures containing second
phases,especiallyfeldspars and other clays, are investigated, such detailed studies becomeeven more essential.
547
Nevertheless,when such detailedinvestigationsare carried
out in combination with silicion-29 MASS NMR, unique
insight into the structuresof layer silicates,including clay
minerals.can be obtained.
Acknowledgments
This paper is dedicatedto the memoryof ProfessorJohn
of
the usefulness
Hower.John wasamongthe first to recognize
silicateminin investigating
MASSNMR spectroscopy
solid-state
erals and was instrumentalin bringing us all togetherto work on
the problem.We will misshis insight and guidance.
This work was supportedin part by the National ScienceFoundation Grants EAR8207260and EAR8,108421and the Texaco
Corporation, and has benefittedfrom facilitiesmade availableby
the University of lllinois-National ScienceFoundation Regional
NMR InstrumentationFacilitY.
References
Abragam,A. (1961)The Principlesof Nuclear Magnetism.Clarendon Press.Oxford.
Andrew, E. R. (1971)The narrowing of NMR spectraof solids by
high-speedspecimenrotation and the resolution of chemical
shift and spin multiplet strudures for solids.Progressin NMR
S, 1-39.
Spectroscopy,
Barron, P. F., Wilson, M. A., Campbell, A. S., and Frost, R. L'
(1982) Detection of imogolite in soils using solid state 2eSi
NMR. Nature,299,61ffi18.
Barron, P. F., Frost, R. L., Skjenitad,J. O., Koppi, A. J. (1983)
Detection of two silicon environmentsin kaolins by solid state
2esiNMR. Nature, 302,49-$.
de Jong, B. H. W. S., Schramm,C. M. and Paniale, V. E. (1983)
Polymerizationof silicate and aluminate tetrahedrain glasses,
melts, and aqueous solutions-IV. Aluminum coordination in
glassesand aqueoussolutionsand commentson the aluminum
avoidance principle. Geochimica et Cosmochimica Acta, 47,
t223-1236.
Farmer, V. C. and Velde, B. (1973)Effect of structural order and
disorder on the infrared spectraof brittle micas.Mineralogical
Magazine,39,282-288.
Farrar, T. C., and Becker,E. D. (1971)Pulse and Fourier trans'
form NMR: introduction to theory and methods. Academic
Press,New York.
Fyfe, C. A., Gobbi, G. C., Klinowski, J., Thomas, J. M., and
Ramdas,S. (1982)Resolvingcrystallographicallydistinct tetra'
hedral sitesin silicaliteand ZSM-5 by solid-stateNMR. Nature,
296,53V536.
Ghose,S., and Tsang,T. (1973)Structural dependenceof quadrupole couplingconstante2qQ/hfor 27Al and crystal field parameter D for Fe3* in aluminosilicates.American Mineralogist,58,
748-755.
Grimmer, A. -R., von Lampe,F., Megi, M. and Lippmaa, E. (1983)
High resolution 2esi NMR of solid silicates:influenceof Fe2+
in olivines.Zeitschriftliir Chemie,23,343-344.
Guggenheim,S., and Bailey,S. W. (1975)Refinementof the margarite structure in subgroup symmetry. American Mineralogist,
ffi, to23-1o29.
Hower, J., Eslinger,E. V., Hower, M. E. and Perry, E. A. (1976)
Mechanismof burial metamorphismof argillaceoussediment:
1. mineralogical and chernicalevidence.Bulletin of the Geological Societyof America,87,725-737.
Kirkpatrick, R. J., Kinsey, R. A., Smith, K. A., Henderson,D. M',
and Oldfield, E. (1985)High resolution solid state sodium-23,
aluminum-27,and silicon-29nuclear magneticresonancespec-
548
K/NSEY ET AL.: NUCLEAR MAGNETIC RESONANCESPECTROSCOPY
OF LAYER \ILI1AT,ES
troscopic reconnaissanceof alkali and plagioclasefeldspars.
AmericanMineralogist,70, lM-123.
Klinowski, J., Ramdas,S., Thomas, J. M., Flb, C. A. and Hartman, J. S. (1982)A re-examinationof Si, Al ordering in zeolites
NaX and NaY. Journal of the Chemical Society, Faraday
TransactionsII. 78. 1025-1050.
Lippmaa, E., Megi, M., Samoson,A., Engelhardt,G., and Grimmer, A. -R. (1980)Structural studiesof silicatesby solid-state
high-resolution 2esi NMR. Journal of the American Chemical
Society,102,4889-4893.
Lippmaa, E., Megl, M., Samoson,A., Tarmak, M., and Engelhardt,G. (1981)Investigationof the structure of zeolitesby
solid-statehigh-resolution2esi NMR spectroscopy.Journal of
the ArnericanChemicalSociety,103,49924996.
Loewenstein,W. (1954)The distribution of aluminum in the tetrahedra of silicatesand aluminates.American Mineralogist, 39,
92-96.
Megi, M., Samoson,A., Tarmak, M., Engelhardt,G., and Lippmaa, E. (1981)Studiesof the structure of silicatesand zeolites
by solid-state high-resolution silicon-29 NMR spectroscopy.
Doklady AkademiiNauk SSSR,261,tl69*tt't4.
Meadows, M. D., Smith, K. A., Kinsey, R. A., Rothgeb, T. M,
Skarjune,R. P., and Oldfield, E. (1982)High-resolution solidstate NMR of quadrupolar nuclei.Proceedingsof the National
AcademyofScienceofthe United States,79,l35t-1355.
Miiller, D., Gessner,W., Behrens,H. J., and Scheler,G. (lggt)
Determination of the aluminum coordination in aluminum-
oxygen compoundsby solid-statehigh-resolution 27Al NMR..
ChemicalPhysicsLetters,79, 59-62.
Nolle, A. (1977)Second-orderquadrupole splittings of the esMo
NMR signal in Mo(CO). and their reduction by samplespinning. Zeitschriftfiir Physik A,280,231-234.
Oldfield, E., Kinsey, R. A., Smith, K. A., Nichols, J. A., and Kirkpatrick, R. J. (1983)High-resolutionNMR of inorganic solids,
influence of magnetic centerson magic-anglesample-spinning
lineshapesin some natural aluminosilicates.Journal of Magnetic Resonance,51,3215-t27.
Pinnavaia,J. (1981)Electron spin resonancestudiesof clay minerals.Developmentsin Sedimentology,34, 139-161.
Robie, R. A., Hemingway,B. S., and Wilson, W. H. (1976)The
heat capacitiesof calorimetry conferencecopper and of muscovite KAlr(AlSi3)Olo(OH)2,pyrophyllite AlrSinOro(OH)r,and
illite Kr(AlrMglSir4Al2)O4o(OH),between15 and 375 K and
their standardentropiesat 298.15K. Journal of Researchof the
U. S. GeologicalSurvey,4,631-64.
Schmidt, V. H. (1971)Pulse responsein the presenceof quadrupolar splitting. Proceedings of the Ampere International
SummerSchoolII, Yugoslavia,7i-83.
Smith, K. A., Kirkpatrick, R. J., Oldfield, E., and Henderson,D.
M. (1983) High-resolution silicon-29 nuclear magnetic resonance spectroscopicstudy of rock forming silicates.American
Mineralogist,68, l20Gl2l5.
Manuscriptreceioeil,December13,I9E3;
accepted
for publication,N ooember19, 1984.