American Mineralogist, Volume 70, pages 106-123, I9E5

Iligh resolution solid-state sodium-23, aluminum-27, and silicon-29 nuclear magneticresonance spectroscopic reconnaissance of alkali and plagioclase feldspars

R. JeMes Ktnrpernrcr

Department of GeologyUniversity of lllinois at Urbana-Champaign

l30l West Green Street. Urbana. Illinois 61801

Ronnnr A. KrNsev. KeneN ANN SuIrn

School of Chemical SciencesUniversity of lllinois at Urbana-Champaign

505 South Mathews Avenue, Urbana, Illinois 61801

Donelo M. Hr,NoBnsoN

Department of GeologyUniversity of lllinois at Urbana-Champaign

1301 West Green Street. Urbana. Illinois 61801

eNo Enrc OlorrBI-o

School of Chemical SciencesUniversity of Illinois at Urbana-Champaign

505 South Mathews Avenue. Urbana, Illinois 61801

Abstract

We present in this paper high-resolution solid-state magic-angle sample spinning sodium-23, aluminum -27 , and silicon-29 nuclear magnetic resonance (NMR) spectra of a variety ofplagioclase and alkali feldspars and feldspar glasses. The results show that MASS NMRspectra contain a great deal of information about feldspar structures. In particular, weobserve the following. (l) AUSi order/disorder and exsolution effects in alkali feldspars arereadily observable. (2) To our present level of understanding the published structurerefinements of anorthite are not consistent with the NMR of spectra of anorthite. (3) Thespectra ofthe intermediate plagioclases are interpretable in terms ofe- and /-plagioclases,with much of the signal apparently coming from strained sites between albiteJike andanorthite-like volumes in the e-plagioclases. (4) The nearest-neighbor atomic arrangementsin albite and anorthite glasses are very similar to those in the corresponding crystals,although the range of cation-oxygen distances is larger in the glasses. (5) There are noaluminum-27 peaks in the range of 0 ppm that would indicate octahedrally coordinatedaluminum in either glass.

Introduction

We present in this paper high-resolution solid-statesodium-23, aluminum-27, and silicon-29 nuclear magneticresonance (NMR) spectroscopic results for a variety ofplagioclase and alkali feldspars and feldspar glasses. Theobjective of this work is to explore the applicability ofhigh-resolution solid-state NMR to the examination of thestructures of complex silicate materials, such as feldsparsand feldspar glasses. The emphasis in this paper is on the

spectroscopic aspects of this problem' Future work willexamine the mineralogical and petrological applicationsin more detail. Both magic-angle sample spinning (MASS)and variable-angle sample spinning (VASS) techniqueshave been used.

The samples discussed here include low albite, micro-line perthite, cryptopenhite, sanidine, a range of plagio-clase compositions, and albite and anorthite glasses. Thefeldspar samples are listed in Table 1.

In this paper we briefly summarize feldspar nomencla-

m03-004)vE5/0 102-0 r06$02.00 106

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS t07

Alb l te(Ab99)

AIb l te C lass(Abr00)

San ld in€(or60Ab38An2)

Mlc roc l lnePer th l te

Cryp toper th lEe( Spencer R)( o r 4 3 , 4 b , 5 3 , k 4 )

01 lgoc la6e( An23 )

Labradot l te(Ans2 )

Ey toml te(An70)

( h 9 2 )

(Anr00)

Anor th l te G lass(hroo)

&el ta , V l rg ln la

Synthe t lc

Cyonlng

Nomay)

Bakersv l l le ,North Carollna

Sydney Ene6,kbrado!

Crys ta l ky lMlnnesota

Pacaye Volceno

GuatgMla

Synthe t ic

Synrherlc

poor ly separa ted

Ab T2trT2bT2o + Ab T2oTlnAb T lo

Ab T2EA b D o + 0 r T 2Or Tl

Ab T ID

Shou lder (An Fak)

Ab peak

Cor le la t lon to

Flrst tro peakspoor ly reeo lved

Cor te la t lon topub l lshed Et ruc turereflnerentB poor

Table l. Silicon-29 NMR chemical shifts of feldspars andfeldspar glasses

CHEMI6LruTERIAL LOCALIN SHTF1 COffiNTS

spar mineralogy are given in the compendia of Smith(1974) and Ribbe et al. (1983).

For the benefit of those unfamiliar with the feldspars,we summarize their main features (following Smith, 1974and Ribbe et al., 1983), state some of our assumptions,and define some of our nomenclature and symbolism.

l. Feldspar structures are based on a framework of corner-linked SiO+ and AlOa tetrahedra. The ratio of aluminum tosilicon ranges from l:3 to l:1. We use the notation Q4(nAl) todescribe the silicon sites, where the superscript indicates thefour bridging oxygen atoms and the integer n (0 to 4) denotes thenumber of aluminum (AlOo) tetrahedra to which a silicon (SiO+)tetrahedron is connected.

We assume, as is customary in dealing with tetrahedral siliconand aluminum occupancies, that the "aluminum avoidance rule"(Loewenstein, 1954; Goldsmith and Laves, 1955) holds for allfeldspars. That is, no aluminum tetrahedron is linked to anotheraluminum tetrahedron, only to four silicon tetrahedra. A// alumi-num sites, therefore, should be A4 (4Si). We will show, however,that for our synthetic anorthite this is not completely true.

2. Feldspar frameworks have inherent monoclinic or pseudo-monoclinic symmetry. Because AlO4 tetrahedra are slightlylarger than SiO4 tetrahedra, some patterns of AllSi distributiongive monoclinic frameworks, whereas others cause slight distor-tions and triclinic symmetry. The pattern of tetrahedral siteoccupancy leads to the important concept of AVSi tetrahedralsite order/disorder. Other things being equal, high AVSi disorderpromotes monoclinic frameworks, whereas increasing orderpromotes triclinic frameworks having increasing cell obliquity(indicated by the magnitude ofthe departure ofthe B-axis fromB*).

The large, irregular cages in the tetrahedral framework areoccupied by potassium, sodium, or calcium ions. When occupiedby the larger potassium ions, the cages are stretched and theframework tends to retain its monoclinic symmetry if it has aproper AUSi distribution. Ifthe AUSi pattern is not appropriate,the symmetry is triclinic. When occupied by the smaller, sodiumor calcium ions, the cages crumple a little and distort theframework slightly so that the symmetry is triclinic, regardless ofthe AVSi order/disorder (except for the highest temperaturesodium feldspar, monalbite).

3. Feldspar compositions are conveniently described in termsof percentages of the endmember components KAlSi3Os (Ortho-clase, Or), NaAlSi3Or (Albite, Ab), and CaAl2Si2Os (Anorthite,An). The rock-forming feldspars occur in two series, the alkalifeldspars (Or-Ab) and plagioclases (ALAn).

4. The principal potassic feldspars include, from low to hightemperature forms, microcline (triclinic, Cl), orthoclase or lowsanidine (macroscopically monoclinic; monoclinic by polarizedlight microscopy but likely to be triclinic (low obliquity) by X-rayand electron diffraction), and high sanidine (monoclinic, C2im).The principal sodic feldspars are low (temperature) albite andhigh albite, both triclinic, Cl; high albite has the lower cellobliquity. High sanidine and high albite form a continuous solidsolution series despite their different symmetries.

5. In the triclinic alkali feldspars (microcline, low albite, highalbite) there are four sets of similar but not identical tetrahedralsites conventionally labeled Tlo, Tlm, T2o, T2m. At hightemperatures, the aluminum and silicon ions are randomlydistributed over these sites. If a high temperature feldspar isquenched, this disorder is retained. In the lowest temperature

- 9 6 . :- 1 0 4 . 4

-98

-97- I O I

-93- 9 6 . 0

- l 0 l , 8- 1 0 6 , 0

- 9 6 - o- 9 8 . 0

- 1 0 3 , 5

-47-93.2- 9 7 . 2

- 1 0 5 . 3

l2nT20T l o

-83-88-91

- 1 0 1- 1 0 7

-84-89

- 1 0 1

-83.0-84.6-89.2

-42.9-84.8- 8 9 ' 6

- 8 7

ture and structure and previous NMR work on feldspars,review the theory of MASS NMR (especially as it per-tains to the study of quadrupolar nuclei), and present anddiscuss our results. These results are, we believe, only abeginning of the application of MASS NMR to feldsparmineralogy, and we present our conclusions with therecognition that they are in some instances tentative. It isclear from the work we have done, however, that MASSNMR spectra of alkali and plagioclase feldspars contain agreat deal of information about Al/Si orderidisorder in thetetrahedral sites, about chemical bonding in the sodium,aluminum and silicon sites, and also, we believe, aboutdistortion of these sites due to exsolution. Our resultsindicate that solid-state MASS and VASS NMR. whencombined with X-ray and electron difraction and trans-mission electron microscopy (TEM) can provide newinsights into the structures of a wide range of mineralphases.

Previous work

Feldspar structure and nomenclature

Feldspars are the most abundant minerals in the earth'scrust and have received corresponding attention frommineralogists and petrologists. Excellent reviews of feld-

KIRKPATRICK ET AL,: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

forms, maximum low albite and maximum microcline, AVSiorder is, ideally, complete, with all the aluminum in the Tlo sitesand all the silicon in the other three sites.

6. In the monoclinic alkali feldspars there are two sets ofsimilar but not identical tetrahedral sites conventionally labeledTl and T2. High sanidine ideally has one-quarter of an aluminumand three-quaxters of a silicon in each site on the average(complete Al/Si disorder). Orthoclase (low sanidine) is moreordered, with equal numbers of aluminum and silicon atomsrandomly occupying the Tl sites and silicon only occupying theT2 sites.

7. Non-endmember, single phase, high temperature alkalifeldspars exsolve into intergrowths called perthites when cooledslowly enough. Cryptoperthite is perthite in which the inter-growth occurs on a submicroscopic scale. In the finer crypto-perthites, the lamellae of the potassium and sodium-rich phasesare thought to maintain coherency across the interface causingstrain of the sites near the interface. In the coarser perthites,microperthites and macroperthites, the boundaries between thelamellae are noncoherent, and there probably is no significantstraining of the sites near the interfaces.

8. The plagioclases are even more complex. The albite end-members have been described. The calcic endmembers are P-anorthite (triclinic, Pl) and /-anorthite (triclinic, 1l), which occurbelow and above 250'C, respectively. On cooling, the 1 to Ptransition is not quenchable. P- and /-anorthite have doubled ccell edge lengths of about 144 compared to the 7A of the otherfeldspars. Even the P-anorthite structure has a high degree of1-pseudosymmetry (Ribbe in Ribbe, 1983).

P-anorthite has 16 sets of similar but not identical tetrahedralsites. Eight are occupied by silicon and eight by aluminum. 1-anorthite has 8 sets, four occupied by silicon and four byaluminum. Both kinds of anorthite ideally have complete Al/Sisite order. Each silicon tetrahedron is connected to four alumi-num tetrahedra, Qo(4Al), and vice versa, Aa(4Si).

9. Standard compositional ranges of the plagioclases of in-termediate compositions are labeled by the names oligoclase,Anro-ro, andesine, Arro-so, labradorite, Anso-ro, and bytownite,AnTs-e6 (see Table l). These names refer to bulk compositions,not phases. The intermediate plagioclases are all triclinic.

At the highest temperatures there is extensive solid solutionbetween the Ab and An components. At lower temperatures,intermediate composition plagioclases exist as exceedingly fineintergrowths of albite-like and anorthite-like domains on severalscales. The complexity of these plagioclases arises from theincompatibilities of the AVSi site orderings in albite and anorthiteand from the fact that subsolidus atomic rearrangement process-es are so sluggish.

We shall accept the e-plagioclase model as described by Ribbeand by Smith in Ribbe (1983). The intergrowths in plagioclasesoccur on at least two scales. The finer type, e-plagioclase, isthought to consist of a modulated structure composed of coher-ent albite- and anorthite-like domains of the order of a few tensof A across. e-plagioclase seems to be restricted to the rangeAnzo-ro.

A second category of intergrowth occurs on the micrometerscale and is associated with the so-called peristerite (-Anz-ro),BQggild (-An+7-se) and Hiittenlocher (-An6s-q) gaps. Perister-ite intergrowths are considered here to be composed of albite(-An2) and an e-plagioclase (-An25). Bpggild ones are believedto be composed of two e-plagioclases (-Anor and -56) andHiittenlocher intergrowths are thought to be composed of an e-plagioclase (-Anoa) and /-anorthite (-Anso). This /-anorthite is a

type that appears to be body centered by diffraction, even atroom temperatures. It is considered to have a composite struc-ture made from domains of P-anorthite. Adjacent domains are"out-of-step" to one another (antiphase domains) with respect toAVSi ordering, which, coupled with the earlier mentioned 1-pseudosymmetry of each domain, causes systematic difiractionabsences. These diffraction absences give an apparent 1l symme-try for the average structure.

NMR of feldspars

Feldspars have been investigated by single crystalsodium-23 and aluminum-27 wide-line NMR at low mag-netic field strengths since the early 1960's. Brun et al.(1960), Hafner etal. (1962), and Hafner and Laves (1963)investigated alkali feldspars and were able to resolveseparate peaks from different phases. Hafner and Hart-mann (1964) determined electric quadrupole couplingconstants of a variety of compositions. Staehli and Brink-mann(1974a,b) examined the aluminum-z7 NMR of anor-thite using wide-line methods and were able to separatethe eight aluminum sites, determine that the P to Itransition takes place between 220" and 250'C, and againmeasure the values of the electric quadrupole couplingconstants. The quadrupole coupling constants of feld-spars and many other minerals have been summarized byGhose and Tsang (1973).

Silicon-29 chemical shifts obtained by high-resolutionMASS NMR have been presented by Lippmaa et al.

Table 2. Average Si-O bond distances, total cation-oxygenbond strengths, and predicted silicon-29 chemical shifts for

anorthite

sEczsl-rsrSt11cot rS l t e

S1-O Bond StrengthA s u ( v . U . )

k s a w e t a l . ( 1 9 6 2 )

Tr(000)r r ( 0 0 r )r r ( M 0 )T l ( r z t )

r2(0zo)T2(OzL)T2(n00)12(n01)

T l ( 0 0 0 )r l ( o0r)r l ( D z o )Tl (nzl )

1 2 ( 0 2 0 )T2(021)T2(D00)T2(n0r)

Tr (000)r r ( 0 0 1 )r r ( E z o )r l ( M r )

T2( 020)T2(ozt)T2(oo0)T3(Eot)

I . 5 I 3t . 5 1 6L608r , 6 2 6

I . 5 1 3I . 6 1 0t . 6 0 21 . 5 2 8

8 , 0 5 88 . 0 8 38 . 0 3 67 . 9 9 0

8 . 1 8 78 , 2 6 48 . I 5 t7 . 8 7 9

- 1 . 4 3 8- I . 4 7 1- r . 2 7 |-1.266

- 1 . 4 4 1-1.361- t . 3 4 6-1.469

-1.444-t ,243- t . 4 5 1- 1 . 2 6 8

- l , 4 4 6

-t .442

-r.445- 1 . 4 5 8-t .252- r , 2 1 3

- r . 4 6 0

-1.345- 1 . 4 4 5

I . 6 1 61 . 6 t 3I . 6 1 3I . 6 1 3

I . 6 r 7r . 6 0 8I . 5 1 5I . 6 1 4

I . 6 1 8I . 6 1 3t . 6 t 4I . 5 1 3

I . 6 1 7r . 6 1 3I . 6 1 4I . 6 I 6

8 . 1 4 68 . r 3 08 . 0 7 51 .939

8 . r 2 98 . 1 4 38 . 0 0 98 . 0 6 l

8 . I 4 48 . I 2 38 , 1 4 37 .926

8 . r 8 38 . 0 0 08 . 0 8 6

klus ( 1978)

KIRKPATRICK ET AL.: NI]CLEAR MAGNETIC RESONANCE OF FELDSPARS l(D

(1980) for what they call anorthite, albite, sanidine,orthoclase, and adularia. They presented spectra foralbite and sanidine, discussed structural interpretations,and proposed peak assignments for the albite peaks.Oldfield et al. (1983) have discussed the effects of ironimpurities on the aluminum-27 and silicon-29 MASSNMR spectra of a variety of feldspars. K. A. Smith et al.(19E3) have discussed silicon-29 MASS NMR spectrosco-py of a number of silicates, including some feldspars.Recently, J. V. Smith et al. (1984) have presented highresolution silicon-29 MASS NMR data for alkali feldsparsand discussed the peak assignments. Solid-state alumi-num-27 and silicon-29 MASS NMR have probably foundtheir most important application to date in the study ofzeolite catalyst materials (e.g., Klinowski et al., 1982).

Experimental methods

Spectrometers

The spectra presented here were recorded using three "home-built" Fourier transform NMR spectrometers using the methodsdescribed by Meadows et al. (1982) and Smirh et al. (1983). Thethree spectrometers are based on 3.52, 8.45. and ll.7 Teslasuperconducting magnets and use Nicolet ll80 or 1280 comput-ers for data acquisition. All are equipped with Andrew-Beamstype magic-angle sample-spinning rotor assemblies (Andrew,l97l), which can also be used to obtain static (non-spinning)spectra. Chemical shifts are reported in ppm relative to 3M NaClfor sodium-23, in ppm relative to lM Al(H2O)l+ for aluminum-27, and in ppm relative to tetramethylsilane for silicon-29. Morepositive chemical shifts correspond to decreased shielding.

Sample preparation and characterizationAll samples were ground to a fine powder using an agate (SiO2)

pestle and mortar and were tightly packed into Delrin rotors.Most samples were pure minerals and were simply ground andrun. Pacaya anorthite was purified by hand picking and magneticseparation. The Wyoming sanidine was purified by hand pickingand shaking in water in a small mill to remove the soft matrix.

The anorthite glass was prepared from 99.D9Vo pure SiO2,CaCO3, and Al2O3 by melting for about one hour at 1600.C, threetimes, with grinding in an agate mortar between meltings.Synthetic anorthite crystals were grown by crystallizing the glassat about 1500"C for about four hours and then quenching thecrystals. The albite glass was prepared by the Coming GlassTechnical Staffs Division of Corning Glass Company.

All samples were characterized by powder X-ray diffractionand optical microscopy.

Theoretical background

A useful introduction to NMR spectroscopy is given inthe book by Farrar and Becker (1971). Magic-anglesample-spinning (MASS) in general, and silicon-29 MASSNMR spectroscopy in particular, have been discussed byLippmaa et al. (1980) and by Smith et al. (19E3). We notehere only that a given nuclide ordinarily has differentnuclear resonance frequencies in different kinds of sites,whether in diferent materials or in the same material.even if the sites are as similar to each other as the

tetrahedral sites in feldspars. This is because they areshielded from the imposed magnetic field by slightlydifferent electron distributions. Silicon-29 has a nuclearspin I : ll2, and, as described elsewhere (Andrew, l97l;Smith et al., 1983), spinning of a powdered sample on anaxis inclined at 54.7" to the applied magnetic field (the"magic-angle") at kilohertz frequencies greatly narrowsthe NMR peak breadth and increases resolution. This isdue to an efective averaging of the nuclear dipolar andchemical shift anisotropy interactions. Thus, nonequiva-lent silicon sites in an ordered crystal each give a singlenalrow peak. Peaks for glasses and disordered crystalsare generally broader, because a range of magneticallynonequivalent sites is present.

Sodium-23 and aluminum-27, unlike silicon-29, havespins I : 312 and 5/2 respectively and, therefore, exhibitquadrupolar behavior. Although MASS averages the firstorder quadrupole interaction, for sites with large electricfield gradients (and thus large quadrupole interactions)second-order terms in the nuclear quadrupolar Hamilto-nian, Hq, become important, and may cause a significantlinebroadening. Such efects have been discussed in detailby Abragam (1961), Kundla et al. (1981), Meadows et al.(1982), Ganapathy et d. (1982), and Schramm and Old-field (1982). We review here those aspects of theorynecessary to understand the spectra presented.

A nucleus of spin I has (2I + l) different nuclear energylevels when placed in a magnetic field, as shown forsilicon-29 (l = ll2, two energy levels) and sodium-23 (I =312, 4 energy levels) in Figures lA and D. In NMRspectroscopy the energies (or frequencies) of the transi-tions (1/2, -ll2), (112, 312), etc. are measured. In theabsence ofany other magnetic interactions, only a singleresonance frequency would be observed. However, inreal samples, there are a variety of interactions whichcause slight changes in these nuclear energy levels. Forsilicon-29 the main efect is due to an orientation-depen-dent shielding of the main magnetic field at the nucleus.This is called the chemical shift anisotropy (CSA) andyields a broad resonance in powdered samples (Fig. 1B).In MASS this interaction is effectively averaged by rapidsample rotation in a sample rotor spun by an air-bearingat frequencies, 2,, typically from about 2,000 to 6,000 Hz.The result is a sharp-line spectrum with a centerband(CB, Fig. lC) at the same frequency as in the absence ofthe CSA, together with a number of spinning sidebands(SSB), spaced at the spinning frequency (2.).

For quadrupolar nuclei (sodium-23, | = 312, aluminum-27, | = 5 12), there are a number of transitions. In general,the resonances of these quadrupolar nuclei in powdersare broadened by the interaction of the nuclear electricquadrupole moment (eQ) with the electric field gradient atthe nucleus (eq). This interaction is described in terms ofthe quadrupole coupling constant, e2qQ/h. For sodium-23, the first order quadrupole interaction leads to twobroad absorptions due to the satellite transitions (312, ll2;-lD, -3/2) together with a relatively sharp central reso-

1 1 0 KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

B cA

ZEEMAN

DQUADRUPOLAR

-roo -r50 -200P P M

E

-roo -r50 -200PPM

FcB ( t /2 , - t /21

t , " ' t " " t " " t " "200 r50 roo 50 0

KHz H

200 t50 loo 50 0KHz

G

Fig. 1. Energy level diagrams, static spectra, and MASS NMR spectra of a spin I = 1/2 nuclide (A,B,C), a quadrupolar nuclide (spin| : 312) with a small quadrupole coupling constant (D,E,F), and a quadrupolar nuclide with a large quadrupole coupling constant(D,G,H). (A) Energy level diagram of spin I = l/2 nuclide showing Zeeman splitting into ll2, and - l/2 spin states and small additionalsplitting due to chemical shift anisotropy (CSA). (B) Simulation of a static spectrum of a spin I : l/2 nuclide (silicon-29 in Na2Si2O5)(C) MASS NMR spectrum of spin I : l/2 nuclide (silicon-29) in Na2Si2O5 showing the center band (CB, the true resonance) andspinning sidebands (SSB) spaced at the spinning speed, zr. (D) Energy level diagram of a spin I : 3/2 quadrupolar nucleus showingZeernan splitting and a small additional splitting due to the quadrupolar interaction. (E) Simulation of a static spectrum of aquadrupolar nuclide (I : 3/2 in this case) which has a small quadrupole coupling constartt. (F) Simulation of a MASS spectrum of an I= 3/2 nuclide showing the centerband (CB), which is the central (+ll2 + - l/2) transition and the spinning sidebands (SSB) mappingout the shape of the satellite spectra. (G) Simulation of the static spectrum of a quadrupolar nuclide with a large quadrupolar couplingconstant (sodium-23 in Na2MoO+). Only the central (+ 12+ -ll2)transition is shown. (H) Simulation of the VASS NMR spectrum ofsodium-23 in Na2MoOa showing the narrowing possible at 75'. The MASS spectrum of this compound shows less narrowing.

nance (112, -ll2), as shown in Figure lE. In general, withMASS, these interactions are again averaged, and asharper central line (112, -ll2) together with a number ofsidebands due to the satellite transitions are observed,Figure lF. When the quadrupole coupling constant isabout 1 MHz in frequency units, only the central (l/2,-ll2) resonance is readily detected, since there is a largeresidual broadening of the satellite lines, due to smallfluctuations around the magic-angle due to wobbling ofthe sample rotor.

When the quadrupole coupling constant is larger (great-

er than about 3 MHz), additional structure generallyappears on the central line due to the higher order effectswhich arise when the quadrupole interaction is a signifi-cant fraction of the actual resonance frequency. Forexample, for 23Na2MoOa the quadrupole coupling con-stant is approximately 2.6 MHz, and at all fields in therange 3.52 to 11.7 Tesla there is a so-called second-orderbroadening of the central transition, Figure lG. Thisinteraction has a different angular dependence than thatcausing the first order chemical shift or quadrupolarbroadenings. In order to reduce this interaction it is

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS l l t

necessary to spin the sample at a different angle in themagnetic field (variable-angle sample-spinning, VASS,Ganapathy et al., 1982).

Thus, for silicon-29, or for quadrupolar nuclei in highlysymmetric environments (where electric field gradientsare small)-such as sodium-23 in NaCl, magic-angle(54.7") sample-spinning yields the highest resolution spec-tra, Figures lC and F. For quadrupolar nuclei in highlyasymmetric sites, such as sodium-23 in Na2MoO4, vari-able-angle spinning, at 75" in this case, yields the narrow-est line width, Figure lH. Both techniques are used here.

As discussed in detail elsewhere (Meadows et al.,1982), the second-order quadrupole interaction is a field-dependent interaction, and results in different apparentchemical shifts for different applied magnetic freldstrengths (Abragam, 1961). As a result, there may besignificant discrepancies in ihe chemical shifts of a qua-drupolar nucleus when results obtained at different mag-netic field strengths are compared. All data for sodium-23and aluminum-27 presented here were obtained at thethree different magnetic field strengths to help determinethe true chemical shifts. Because silicon-29 has spinI : ll2, its chemical shifts are not field-dependent.

For the quadrupolar nuclei it is often useful to simulate(calculate) the spectra using the methods of Ganapathy etal. (1982). The parameters needed for the simulation arethe quadrupole coupling constant (e2qQ/h), the asymme-try parameter ofthe electric field gradient tensor (4), andthe spinning angle (0). The asymmetry parameter is ameasure of the deviation of the electric field gradient atthe nucleus from cylindrical symmetry (Abragam, 196l).Such simulations are used here in conjunction with pub-lished quadrupole coupling constants and asymmetrypa"rameters to determine isotropic chemical shifts, and tolook for contributions from exsolved phases.

For the silicon-29 NMR spectra we report the chemicalshifts of narrow peaks to +0.1 ppm, but subscript the lastplace because accuracy and reproducibility from labora-tory to laboratory is probably only about 0.3 ppm. This isprimarily due to magnet drift over the relatively longtimes (up to 12 hours) needed to collect some silicon-29spectra. For broader silicon-29 peaks and shoulders wereport the shifts to only -r I ppm.

For the quadrupolar nuclei we report the values ofmostof the singularities or peak maxima to + 1 ppm. This isbecause even at ll.7 Tesla the peaks are at least 10 ppmwide. We report some of the narrowest singularities to-+0.1 ppm with the last place subscripted.

Results and discussion

Silicon-29

We present in Figures 2 and 3 the silicon-29 MASSNMR spectra of some representative feldspars and feld-spar glasses obtained at a magnetic field strength of 8.45Tesla. Table I summarizes the chemical shifts obtained

-96'.-92'"

ALBITE

,rr,1/t/\"t"fdli

f - . -l -- L 1 0 - 1 0 0 - L O U

- q o

\

- - - f - ]- t l o - t o o - 1 6 0 - 1 0 c - 1 6 0

E

- + o

CRYPTOPERTHITE

- + o

t 'si 8,45 T MASS

-,oc,, A

i J J - 1 6 0

-ror D

l --T-----_lI U U . I O U

P P M

Fig. 2. 8.45 Tesla 2eSi NMR spectra (corresponding to aresonance frequency of71.5 MHz) ofnatural alkali feldspars anda synthetic albite glass. (A) Albite (Amelia, Virginia), 2.4 kHzMASS, 3l scans at 600 sec recycle time, 8.6 kHz spectral width,20 prsec 90'pulse excitation, 4096 data points, zero-filled to E192,25 Hz linebroadening due to exponential multiplication. (B)Microcline perthite (Delaware Co., Pennsylvania), 2.3 kHzMASS, 3000 scans at l0 sec recycle time,8.6 kHz spectral width,2.6 psec 23' pulse excitation, 4096 data points, zero-filled to8192,5 Hz linebroadening due to exponential multiplication. (C)Synthetic albite glass, 2.4 kHz MASS, 87 scans at 600 secrecycle time, 8.6 kHz spectral width, 30 psec 90" pulseexcitation, 2048 data points, zero-filled to 8192, 100 Hzlinebroadening due to exponential multiplication. (D) Sanidine(Wyoming), 2.4 kHz MASS, 2000 scans at l0 sec recycle time,8.6 kHz spectral width, 2.6 p,sec 23" pulse excitation, 4096 datapoints, zero-filled to 8192, 25 Hz linebroadening due toexponential multiplication. (E) Cryptoperthite ("Fredriksvarn",Norway),2.7 kHz MASS,65 scans at 6ff) sec recycle time, 8.6kHz spectral width,22 psec 90" pulse excitation, 204E datapoints, zero-filled to 8192, 25 Hz linebroadening due toexponential multiplication.

from these spectra, together with sample locality and

composition.The spectra of all the crystalline samples contain

considerable structure arising from silicon sites with

different electronic environments. The spectra of albiteand anorthite glass are both broad, single peaks with

t12 KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

discuss the alkali feldspars whose spectra are shown inFigure 2.

The MASS NMR of low albite (Fie. 2,{) was firstexamined by Lippmaa et al. (1980) and has also beendiscussed by K. A. Smith et al. (1983) and J. V. Smith etal. (1984). Because low albite has three particularly wellresolved peaks, it has become a standard material inmany NMR laboratories. The assignments of the threepeaks to the three silicon sites in albite (Harlow andBrown, 1980) listed in Table I are those of Lippmaa et al.(1980), based on qualitative arguments, and elaborated byK. A. Smith et al. (1983), based on a procedure that usescation-oxygen bond strength. The peak at -92.2 ppm isassigned to T2m, which has two aluminum next-nearestneighbors and two silicon next-nearest neighbors, Qa(2Al), and those at -96.3 and -104.7 ppm to T2o andTlm, which are Q4 (tlt) sites. These assignments are thesame as those of J. V. Smith et al. (1984) based on themean secant of the Si-O-T angle.

The silicon-29 NMR spectrum of microcline perthite(Fig. 28) exhibits five resolved peaks. The sample is atypical microcline perthite from a pegmatite. The micro-cline phase has a 13l-l3l powder X-ray peak separationof -0.7E" (CuKa) and is a maximum microcline. Our siteassignments are based on the NMR measurements ofcation-exchanged feldspars by J. V. Smith et al. (1984),which are diferent from those of K. A. Smith et al. (1983)based on the structure refinement of Brown and Bailey(196/.). The new assignments of J. V. Smith et al. give thesame order of peaks as for low albite. We assign the peaksat -96.0, -98.6 and -101.8 ppm in Figure 28 to T2m,T2o and Tlm, respectively, of maximum microcline. Thesmall peak at -93, the added intensity under the peaks at-96.0 and -98.6 ppm and the small peak at -106.0 ppmbelong to the small amount of intergrown low albite in theperthites.

The spectra of low albite and microcline perthite clear-ly demonstrate the high degree of Si/Al tetrahedral siteordering that has been deduced for these minerals by X-ray difraction. The fact that there is no recognizablesilicon-29 signal for the Tlo site in either spectrum is alsoconsistent with the accepted models for these mineralswhich propose that Tlo is occupied entirely by alumi-num.

The spectrum of cryptoperthite in Figure 2E has somesimilarities to those of albite and microcline perthite.There are four resolved peaks, but the background underthese peaks is higher than for the other two minerals.

This cryptoperthite comes from "Fredriksvarn" (nowStavern) in the well known Larvik region in southernNorway. Many of the cryptoperthites from this region aremoonstones. Our sample is the much-studied "SpencerR" (Spencer, 1937). The cryptoperthites from the Larvikregion are highly sodic, vary from place to place and arelikely to be zoned compositionally and texturally (Spen-cer,1937; Oftedahl, 1948; Muir and Smith, 1956; Smithand Muir, 1958). Spencer R has a bulk composition (molepercent) of about Or+rAbsrAnr (Spencer, 1937). The

- L + O - 1 u U - 1 6 0 - + 0 - l J J

ttsi 9,45 T MASS

- r lo - t o u - { o - l U U - L O U

- t o u

-82't

SYNTHETICANORTHITE

;fiA*"-;

I

monrxrre IGLASS i

I

II

I

- t + 0 - 1 0 0 - 1 6 0PPIY

r-- -f-- r t o - 1 0 0 - 1 6 0

PPIY

Fig. 3. 8.45 Tesla 2eSi MASS NMR spectra (corresponding toa resonance frequency of 71.5 MHz) of natural and syntheticplagioclase feldspars and a synthetic glass. (A) Oligoclase(Bakersviile, North Carolina),2.7 kHz MASS, E2 scans at 300sec recycle time, 8.6 kHz spectral width, 30 psec 90" pulseexcitation, 2048 data points, zero-filled to 8192, 25 Hzl inebroadening due to exponential mult ipl icat ion. (B)Labradorite (Sydney Mines, Labrador), 2.4 kHz MASS, 52scans at 600 sec recycle time, 8.6 kHz spectral width, 9 psec 90'pulse excitation, 4096 data points, zero-filled to 8192,75 Hzlinebroadening due to exponential multiplication. (C) Bytownite(Crystal Bay, Minnesota),2.3 kHz MASS, 313 scans at 8.6 secrecycle time, l0 kHz spectral width, 2.3 psec 23' pulseexcitation, 4096 data points, zero-filled to 8192, 25 Hzlinebroadening due to exponential multiplication. (D) Anorthite(Pacaya Volcano, Guatamala), 2.7 kHz MASS,64 scans at 600sec recycle time, 8.6 kHz spectral width, 22 psec 90' pulseexcitation, 204E data points, zero-filled to 8192, 25 Hzlinebroadening due to exponential multiplication. (E) Syntheticanorthite, 2.5 kH.z MASS, 3l scans at 6C)0 sec recycle time, 8.6kHz spectral width, 25 trr,sec 90o pulse excitation, 2048 datapoints, zero-filled to 8192, 25 Hz linebroadening due toexponential multiplication. (F) Synthetic anorthite glass, 2.4 kHzMASS, 9l scans at 600 sec recycle time, 8.6 kHz spectral width,25 g,sec 90'pulse excitation, 2048 data points, zero-filled to 8192,100 Hz linebroadening due to exponential multiplication.

essentially no structure. Because silicon-29 has spin I =

ll2 and does not suffer from quadrupolar broadening,these broad, structureless peaks imply a continuum ofelectronic environments around the silicon sites. We first

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS l l 3

aggregate perthite of our sample has monoclinic optics.Albite exsolution lamellae, though very fine, are clearlyvisible under the polarizing microscope, even at lowmagnifications. Nonetheless, transmission electron mi-crographs of Spencer R (e.g. Bollmann and Nissen, 1968)show albite lamellae less than 1000A across.

Stewart and Wright (1974) report the potassic phase ofSpencer R to be monoclinic (orthoclase) by powder X-raydifraction and to have a composition (mole percent) ofOrs2. They also report it to have the second largestamount of "strain" (Aa) of any potassic feldspar forwhich they have data.

It is to be noted that even the alkali feldspars from theLarvik area have significant calcium contents rangingfrom around Ana to at least Anle (Spencer, 1937; Ofte-dahl, 1948; Muir and Smith, 1956; Smith and Muir, 1958).Ribbe (1983, p. 243), for example, reports that the ex-solved plagioclase lamellae in cryptoperthite from Larvik(presumed by us to be similar to Spencer R) have acomposition in the range Anrs_ra (oligoclase). In line withthe bulk composition Ora3Ab53Ana for Spencer R, trans-mission electron micrographs of this and similar perthitesfrom the Larvik area show the cross-sectional area oftheplagioclase lamellae to exceed slightly that of the ortho-clase.

The NMR spectrum (Fig. 2E) of the cryptoperthite isdominated by sodic plagioclase peaks at -92.3, -96.0and -103.5 ppm because ofthe high degree ofexsolutionand the high proportion ofsodic feldspar. These peaks arealmost identical to the ones for low albite in Figure 2,4,and 28, despite the probably higher An content andhigher temperature state of the plagioclase lamellae inSpencer R.

Only two peaks attributable to the potassic feldsparphase are apparent: the added height ofthe peak at -96.0and the small peak at -98.9 ppm.

In monoclinic alkali feldspars (orthoclase and sanidine)there are two sets of tetrahedral sites, Tl andT2.In idealorthoclase, the Tl sites are thought to be randomlypopulated by aluminum and silicon ions; the T2 sites arethought to contain only Si ions. Such a phase shouldexhibit two silicon-29 NMR peaks, the one correspondingto T2 having about twice the area ofthe one correspond-ing to Tl. In ideal sanidine, on the other hand, the Tl andT2 sites are both thought to be randomly populated byone-quarter aluminum and three-quarters silicon, on theaverage. In this case, one expects two NMR peaks ofabout equal area.

Unfortunately, the unavoidable interference by theplagioclase peak at -96.0 ppm makes it diffcult toestimate the relative areas of the two peaks for orthoclaseat -96.0 and -98.9 ppm. We infer that the low peak at-98.9 is from the Tl site because such a site is morelikely than a T2 site to resemble the Tlm site of micro-cline having a peak at -101.8 ppm. The added height ofthe peak at -96.0 ppm, therefore, is assigned to the T2site. This is close to the peak at -98.6 ppm (Fie. 28) fromthe T2o site of microcline.

By qualitative comparison of the -92.3, -96.0, and-103.5 ppm peaks of Spencer R (Fig. 2E) with those forlow albite in Figure 2A,we believe that the component ofthe peak at -96.0 ppm (Fig. 2E) contribured by the T2site of orthoclase is appreciably larger than the peak at-98.9 ppm contributed by the Tl site. This is consistentwith the characterization of the potassic phase of thecryptoperthite as orthoclase.

We chose this cryptoperthite for study partly because itis thought to represent a potassic feldspar having a largenumber of strained sites in the coherent boundary regionsbetween lamellae (Stewart and Wright, 1974).We expect-ed to observe considerably broadened NMR peaks arisingfrom ranges of distorted silicon sites in these boundaryregions. The individual peaks in the cryptoperthite spec-trum are just as sharp and narrow as those for low albiteand microcline perthite. The unresolved background inthe cryptoperthite spectrum, however, is much larger.We interpret this background as signal arising fromstrained sites in coherent boundary regions. The signal inthe narrow peaks is coming from the unstrained volumesaway from the lamellae boundaries.

The silicon-29 NMR spectrum of sanidine (Fig. 2D)consists of a broad doublet. This feldspar (Or6sAb36An2)occurs as clear, glassy, I to 5 mm phenocrysts in rhyolitefrom Wyoming. Our interpretation of the sanidine spec-trum is the same as for the orthoclase in the cryptoperth-ite. The two overlapping peaks represent the two sets oftetrahedral sites, Tl and T2, which are the only twopresent in monoclinic alkali feldspars. We infer that thepeak at - 101 ppm for sanidine corresponds to the peak at-98.9 ppm for orthoclase and arises from the Tl site, andthat the peak at -97 ppm in sanidine corresponds to thepeak at -96.0 ppm for orthoclase and arises from T2.Again, this is consistent with the low albite and micro-cline data, for which T2m and T20 are less shielded (lessnegative) than Tlm.

The peaks for sanidine are much broader than any ofthose discussed so far. Because silicon-29 has spin| = l/2, there is no broadening from quadrupolar interac-tions. Moreover, there should be no peak broadeningfrom field inhomogeneity because our magnetic fieldinhomogeneity is less than I ppm. Consequently, weinterpret the peak broadening to arise from AVSi disorderin the occupancy ofthe two tetrahedral sites. The appar-ently equal areas under the two overlapping peaks indi-cates an approximately equal distribution of silicon andaluminum between the Tl and T2 sites, as expected for ahigh sanidine. Because ofthis disorder, for each ofthesesites there are probably Q4(0Al), Q4(lAl), Q4(2Al), andQ4(3Al) sites for both Tl and T2. Because of the overlap-ping of the peaks from each of these eight sites, andbecause of KA.{a disorder, they cannot be resolved. Thewings in Figure 2D in the range of 105 and 88 ppm areprobably due to Q410Al) and Q4(3Al) sites respectively.A significant concentration of Q4(4Al) sites would not beexpected for a phase with a SiiAl ratio of 3/1 (Klinowskiet al., 1982).

l l 4 KIRKPATRICK ET AL,: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

We consider now the silicon-29 spectra of our plagio-clases (Fig. 3). We have already dealt with the mostcommon Na endmember, low albite. We consider first thecalcium end-member, anorthite, and then the intermedi-ate plagioclases.

The spectrum (Fig. 3E) of the synthetic anorthite hasthree well resolved peaks: at -82.9, -84.8 and -89.6

ppm and a small peak at about 95 ppm.The spectrum of the natural anorthite (Ansz), Figure

3D, is essentially identical to that of the synthetic anor-thite, except that the peaks at -83.0 and -84.6 ppm arenot as well resolved in the natural sample. The intensityin the range of -95 ppm may be due to Q4(3Al) sites,although the signal intensity is at about the noise level.

Anorthite at room temperature ("P-anorthite") haseight nonequivalent Si sites (Megaw et al., 1962; Wain-wright and Starkey, l97l; Staehli and Brinkmann,l974a,b; Kalus, 1978). The peaks in the anorthite spectracan be assigned using a correlation between the secant ofthe mean Si-O-Al angle for each site and the silicon-29chemical shifts, similar to those for the silica polymorphsand alkali feldspars developed by J. V. Smith and Black-well (1983) and J. V. Smith et al. (1984). This correlationgives three clusters of points using all three availablestructure refinements (Megaw et al., 1962; Wainwrightand Starkey, 1967; and Kalus, 1978), while the bondstrength sum and mean Si-O bond distance correlationsof K. A. Smith et al. (1983) and Higgins and Woessner(1982) do not. The only discrepancy is the reversal ofTl(ooi) and Tl(mzo) in the Wainwright and Starkeystructure, probably due to a long Al-O distance for theirrefinement. The variations in the bond strength sum andbond distance are small (0.2 V.U. and 0.005A, respective-ly), and the predicted values are in the range of Q4(4Al)sites (Lippmaa et al., 1981; K. A. Smith et al. (1983).

. The small peak for the synthetic anorthite at about -95ppm is probably due to Q4(3Al) sites, which indicate alack of perfect AVSi order in this rapidly crystallizedsample. The poorer resolution of the peaks in the naturalanorthite (AnqJ is probably due to the albite component.

Many of the peaks of the spectra of the intermediateplagioclases in Figures 3,4,, 38, and 3C can be assigned tosilicon sites similar to those in albite or anorthite. Fornone of these samples, however, can all the peaks of bothalbite and anorthite be recognized. Consequently, a sim-ple model involving only domains of albite and anorthitestructures and compositions cannot account for the ob-served spectra. We can, however, account for the spectrain terms of the earlier-mentioned low albite, e-plagioclaseand /-anorthite structures. We attribute much of thesignal in each spectrum to sites in strained volumes in e-plagioclase.

The spectrum of oligoclase in Figure 3A, has manysimilarities to that of albite (Fig. 2A). The principaldifferences are that the peaks at -93.2, -97.2, and-105.3 ppm for oligoclase are broader, overlap consider-ably more, and have diferent relative heights, and that

there is a small peak at -87 ppm. Because the chemicalshifts are identical to those of albite, we assign the peaksat -93.2, -97.2, -105.3 ppm to the T2m, T2o and Tlmsites of albite. We believe the low peak at -87 ppm arisesfrom Q4(4Al) sites like those in anorthite. W. -H. Yang(1984 personal communication) has found a very similarspectrum for an An23 plagioclase.

It is possible to interpret the oligoclase spectrum interms of a single e-plagioclase of composition An23, acomposition between the peristerite and Bpggild gaps.Mclaren (1974) gives transmission electron micrographsof an oligoclase from the same locality as ours whichshow cross-hatched lamellae having a spacing of about1004. This is about the right order of magnitude for thespacings (204) that Smith (1974) says are to be expectedfor the modulated structure of Ab-rich and An-rich do-mains in this compositional range.

We believe that the peak broadening and overlap andthe extra intensity in the -90 to -100 ppm range comefrom a range of silicon site environments in strainedregions between the Ab-rich and An-rich domains in themodulated structure. The fact that the cbll volume ofalbite (660A3) is less than the comparable half-cell volume(6704) of anorthite suggests that the bond angles aredistorted and that the silicon tetrahedra of the albitedomains may be slightly stretched whereas those of theanorthite domains may be slightly compressed in thestrained volumes around the coherent boundaries. Thelonger Si-O bonds in the outer parts of the albite domainsare related to decreased shielding and less negativechemical shifts. The shorter Si-O bonds in the anorthiteare related to increased shielding and more negativechemical shifts. There are probable changes in T-O-Tbond angles also. The spectrum for this composition,therefore, is albiteJike because the "albite" domains arelarger. Because of this the ratio of strained to unstrainedvolumes of the albiteJike material is relatively low.Because the volume of An-rich domains is small, all ormost of this material is strained. This accounts for thelack of anorthite signal in the -83 ppm range. In addition,there could be a significant number of Q4(3Al) sites alongthe boundaries between the domains contributing signalin the range -90 to -95 ppm.

The small peak at about -108 may be due to Q4(0Al)sites in the boundaries between domains or to quartzimpurity, which was not detected optically or by X-ray.

The spectrum of labradorite (Fig. 3B) shows rather ill-defined peaks or shoulders at -94, -101, and -107 ppmwhich may be assignable to Q4(zat) and Q4(lAl) albite-like sites, but more signal than the oligoclase at lessnegative chemical shifts, a poorly resolved peak at -88

ppm, and a shoulder at -83 ppm. The significant amountof signal near -83 ppm indicates that the most deshielded

Q4(4Al) anorthitelike sites are present in this sample.This labradorite (AnsJ falls in the B/ggild composition-

al gap, and probably consists of two e-plagioclases. Itseems likely that the albite-like peaks arise mostly from

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

Q4(lAl) and Q4(2Al) sites in albite-rich volumes in themore albite-rich e-plagioclase. As for the oligoclase, theextrapeak height in the -93 to -Ifi) ppm range probablycomes from sites in the strained regions around themargins of the albite volumes. If the calcium-rich phase inthis material is also an e-plagioclase, some contribution tothe albite-like peaks must also come from albite volumesin this material. The intensity at less negative chemicalshifts (the peak at -88 ppm and the shoulder at -83 ppm)comes from Q4(4Al) anorthite-like sites, which occur inboth e-plagioclases. As for the oligoclase, some signal inthe central part of the spectrum could be due to e4(3Al)sites at the margins of albite and anorthite-rich volumes.The relatively shielded value of the peak at -107 proba-bly indicates some contribution from e4(OAt) sites.

The bytownite spectrum (Fie. 3C) shows poorly re-solved peaks at -84 and -89 ppm assignable to e4(4Al)anorthitelike sites, shoulders at -95 and -l0l ppmassignable to Q4(2Al) and e4(lAl) albitelike sites, andlittle signal in the range of - 105 ppm, although there is asmall shoulder that may be real.

This bytownite (An7s) probably consists of (volumesof) I-plagioclase and e-plagioclase. The e4(4Al) sites,which we believe to cause the peaks at -84 and -89 ppm,are in both materials. The Q4(2Al) and e4(lAl) sires arein only the e-plagioclase. It seems likely that no - 105ppm peak is present (or it is very small) because essential-ly all of the albite volume is strained, causing all of thesilicon sites in the albite to become deshielded. The peakat -84 ppm is probably due to overlap of the -82.9 and-84.8 peak ofpure anorthite.

The silicon-29 NMR spectra of the albite and anorthiteglasses (Figs. 2C and 3F) consist of single broad peaks,with no apparent structure. They are each approximately12 to 15 ppm broader than the total peak breadth of thecorresponding crystalline phases. The peak for albite hasa maximum at -98 ppm and covers the range from about-85 ppm to -ll5 ppm. The peak for anorthite has amaximum at -87 ppm and covers the range from about-72 to - 105 ppm.

Assuming, as discussed below, that all the aluminum inthese glasses is in tetrahedral coordination, both musthave a fully polymerized framework structure. This isbecause the ratio of tetrahedral cations (Si + AI) tooxygen is l/2. All the silicon sites must, therefore, beQa(nAl). One question is what values does n have. W. -H.Yang (1984, personal communication) has examined thesilicon-29 MASS NMR behavior of analbite produced byheating Amelia albite at 1073'C for 120 days andfound that detectable amounts of e4(3Al), e4(2Al),Q4(lAl), and Q4(0Al) are present. The signal in rhisspectrum covers the range from -85 to -115 ppm.Because the spectrum of albite glass extends to even lessshielded values, it seems likely that detectable amounts ofall five possible types of silicon sires et4Al) throughQo(Oetl are present in albite glass.

For anorthite glass the average silicon site must be

Q4(4Al). Because of the range of sites in albite glass, itseems likely that a range of sites are present in anorthitealso. This is consistent with the presence of Q4(3Al) sitesin crystalline anorthite. Because the amount of signal inthe silicon-29 spectrum of anorthite glass at values moreshielded (more negative) than - 100 ppm is small, theconcentration of Qa(0Al) sites must be quite small. Thepresence of a significant amount of signal at values lessshielded (less negative) than -80 ppm must be due todistorted Q4(4Al) sites. Some of these may be caused bythe presence ofAl in the second tetrahedral coordinationshell. In perfectly ordered anorthite all ofthese would beoccupied by Si, but in AVSi disordered material some ofthese must be occupied by Al.

Aluminum-27

Figures 4 and 5 show the ll.7 Tesla aluminum-27MASS NMR spectra of our samples. Table 3 lists the

AI I I ,7T MASS5 7

M I C R O C L I N EP E R T H I T E

I

I o c o o r 5 o c

Fig. 4. ll.7 Tesla 27Al MASS NMR spectra (corresponding toa resonance frequency of 130.3 MHz) of alkali feldspars andalbite glass. (A) Albite (Amelia, Virginia), 3.5 kHz MASS, l0scans at 25 sec recycle time, 15 kHz spectral width, 3 trr,sec 90'pulse excitation, 8192 data points, 10 Hz linebroadening due toexponential multiplication. (B) Microcline perthite (DelawareCo, Pennsylvania), 3.5 kHz MASS, l0 scans at 25 sec recycletime, 15 kHz spectral width, 3.5 psec 90'pulse excitation, 8192data points, l0 Hz linebroadening due to exponentialmultiplication. (C) Albite glass (synthetic), 5 kHz MASS, 200scans at I sec recycle time, 15 kHz spectral width, 3 psec 90'pulse excitation, Sl92 data points, 50 Hz linebroadening due toexponential multiplication. (D) Sanidine (Wyoming), 3.2 kHzMASS, 158 scans at 25 sec recycle time, 15 kHz spectral width,3.5 ;r.sec 90o pulse excitation, 4O96 data points, 50 Hzl inebroadening due to exponential mult ipl icat ion. (E)Cryptoperthite ("Fredriksvarn", Norway), 3.2kHz MASS, 76scans at 5 sec recycle time, 15 kHz spectral width, 4 psec 90'pulse excitation, 4096 data points, 50 Hz linebroadening due toexponential multiplication.

116 KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

t 'At l t ,7 T MASS1 6 0 A / \ . o B

l ,O L I G o C L A S E , L A B R A D O R I T E I r ,

r ' l 'I t f'-' \,./\

--'^-j \----..- d \n

at 59.3 ppm and a full-width at half-height of about 8 ppm.We assign this peak to the Tlo site.

Figure 6 shows observed and simulated aluminum-27and static and MASS NMR spectra for albite at 11.7,8.45, and 3.5 Tesla. The great narrowing of the spectrawith increasing magnetic field strength demonstrates thatthe peak shape is dominated by the second-order quadru-pole interaction. The resolution increases as the square ofthe field strength (Meadows et al., 1982). The simulationsin Figure 6 are in very good agreement with the experi-mental spectra. These simulations assume one kind ofaluminum tetrahedral site with a quadrupole couplingconstant of 3.29 MHz and an asymmetry parameter of0.62. These values are based upon and in good agreementwith the earlier single crystal NMR measurements foralbite by Hafner and Hartman (1964). The simulationsyield an isotropic chemical shift of 62.710.3 ppm. This isin the range of chemical shifts expected for tetrahedralaluminum (Mtiller et al., l98l).

The spectrum of microcline perthite (Fig. 48) consistsof a doublet having maxima at 54.3 and 57 .3 ppm. Singlecrystal aluminum-27 NMR studies by Hafner and Hart-mann (1964) have shown that microcline has one type ofaluminum with a quadrupole coupling constant of 3.27MHz and an asymmetry parameter of 0.21. Figures 7Cand E show the observed aluminum MASS NMR spec-trum of microcline perthite and a simulation assuming theHafner and Hartmann values. The agreement is not verygood. There is additional intensity in the experimentalspectrum around 60 ppm. We believe that this is due toaluminum in the exsolved albite. Figure 7D shows a two

Table 3. Aluminum-27 NMR chemical shifts of feldspars andfeldspar glasses

l O C ; c 0PPtl 116 AlCl3

t o c 5 0PPlil frm AlCl3

Fig. 5. I 1.7 Tesla 27Al MASS NMR spectra (corresponding toa resonance frequency of 130.3 MHz) of natural and syntheticplagioclase feldspars and anorthite glass. (A) Oligoclase(Bakersville, North Carolina),36kH2 MASS, 36 scans at l0 secrecycle time 15 kHz spectral width, 5 psec 90'pulse excitation,4096 data points, 50 Hz linebroadening due to exponentialmultiplication. (B) Labradorite (Sydney Mines, Labrador), 3.3kHz MASS, 1065 scans at 5 sec recycle time, 15 kHz spectralwidth, 4 psec 90' pulse excitation, 4096 data points, 50 Hzlinebroadening due to exponential multiplication. (C) Bytownite(Crystal Bay, Minnesota), 3.4 k}lz MASS, 549 scans at 5 secrecycle time, 15 kHz spectral width, 4 p.sec 90'pulse excitation,4096 data points, 50 Hz linebroadening due to exponentialmultiplication. (D) Anorthite (Pacaya Volcano, Guatamala), 3.3kHz MASS, 500 scans at 5 sec recycle time, 15 kHz spectralwidth, 4 pcsec 90'pulse excitation, 4096 data points, 50 Hzlinebroadening due to exponential multiplication. (E) Syntheticanorthite, 5.1 kHz MASS, 1300 scans at 10 sec recycle time, 15kHz spectral width, 4 psec 90'pulse excitation, 4096 data points,50 Hz linebroadening due to exponential multiplication. (F)Synthetic anorthite glass, 4.9 kHz MASS, 1889 scans at I secrecycle time, 15 kHz spectral width, 4 psec 90'pulse excitation,4096 data points, 50 Hz linebroadening due to exponentialmultiplication.

observed chemical shifts. Aluminum-Z7 has spin | : 512and, therefore, the aluminum-27 NMR spectra are affect-ed by peak broadening and frequency shifts due toquadrupole interactions as discussed above. This quadru-polar broadening prevents us from obtaining as muchdetailed structural information as with silicon-29. AII ofthe observed aluminum-27 chemical shifts are consistentwith aluminum in four-fold coordination (Miiller et al.,1981; Fyfe, et al.,1982).

We consider first the aluminum-27 NMR spectra of thealkali feldspars (Fig. 4). The spectrum for low albite (Fig.4A) consists of a single asymmetric peak with a maximum

' lFor quadr rpo la r nuc le l appareot chemlca l sh i f t changes c i th f le ld s t rength .

V a l u e a t 1 1 . 7 T e s l a l s e s s e n t t a l l y t l e t r u e l s o c r o p i c c h e n t c a l s h i f t .

UTERIALF I I L D

STRENGTH( TESI ,A)

lp"mutlCHEMICAL

STII FT

AIb t te (Ab99)

A lb t te C lass (Ab lOO)

San ld ine ( -0160)

i l i c roc l lne Per th l te

C r y p t o p e r t h i t e

0 l igoc lase (An23)

L a b r a d o r t l e ( A n 5 2 )

Bytocnt te (An70)

A n o r t h i t e ( A n 9 2 )

Anor th t te (An l00)

Anor th i te C lass (An lo0)

a . 4 5

3 , 5I . 4 5

I I . 7

T I . 7

8 . 4 5

1 t . 7

I t . 7

l I . 7

l I . 7

L l . 7

8 . 4 5t L . 7

8 , 4 51 1 . 7

- 2 4 , 2 , Z O , t 35 3 , 5 6

59

5 45 5

58

5 0 , 5 55 4 , 5 7

58

60

60

6 2 , 5 6

6 1 , 5 5

4 558

6 2 , 5 5

5 5

60

| 7

'hr aLBrtEmss

20a o a -zao

Fig. 6. 27Al static and MASS NMR specrra of albite alongwith the MASS NMR computer simularions at I1.7, g.45 and, 3.5Tesla. (A) ll.7 Tesla, static, 30 scans at 25 sec recycle time, 7gkHz spectral width, 3 psec 90" pulse excitation, 4096 data points,50 Hz linebroadening due to exponential multiplication. (B) I 1.7Tesla, 4.0 kHz MASS, l0 scans at 25 sec recycle time, jg kHzspectral width, 3 psec 90o pulse excitation, 8196 data points, l0Hz linebroadening due to exponential multiplication. (C) ll.7Tesla, MASS simulation, 65 Hz linebroadening. (D) g.45 Tesla,static, 50 scans at 25 sec recycle time, 56 kHz spectral width, 3psec 90o pulse excitation, 4096 data points, 50 Hz linebroadeningdue to exponential multiplication. (E) 8.45 Tesla, 4 kHz MASS,50 scans at 25 sec recycle time, 56 kHz spectral width, 3 !.sec 90opulse excitation, 8196 data points, l0 Hz linebroadening due toexponential multiplication. (F) 8.45 Tesla, MASS simulation. 150Hz linebroadening. (G) 3.5 Tesla, static, 300 scans at 5 secrecycle time, 24 kHz spectral width, 3 psec 90. pulse excitation,{D6 data points, 100 Hz linebroadening due to exponentialmultiplication. (H) 3.5 Tesla, 4 kHz MASS, 300 scans at 5 secrecycle time, 24 kHz spectral width, 3 psec 90o pulse excitation,zOlX data points, 100 Hz linebroadening due to exponentialmultiplication. O) 3.5 Tesla, MASS simulation, 100 Hzlinebroadening.

component simulation assuming 75Vo microcline and25Voalbite. Figures 7A and B show the observed albite spec-trum and a simulation at ll.7 T for comparison. Theagreement between the two component simulations andthe observed spectmm is excellent. The simulation (Fig.7E) yields an isotropic chemical shift for microcline of58.5+0.5 ppm. As for the albite, the aluminum-27 NMRdata for microcline clearly indicate that essentially all thealuminum is in one tetrahedral site.

The spectrum for the cryptoperthite (Spencer R, Fig.4E) consists of a single symmetric peak having a maxi-mum at 57.9 ppm. As mentioned earlier, the bulk compo-sition of this sample is about Ora6: there is a dominantoligoclase and a less abundant orthoclase. The quadrupo-lar broadening masks the splitting of the peaks from theAl sites in the orthoclase and oligoclase.

The spectrum of sanidine (Fig. aD) is also a singlesymmetric peak, but is slightly broader than those of theother alkali feldspars (12 ppm wide at half-height). The

maximum is at 58.5 ppm, intermediate between low albiteand cryptoperthite. Single crystal NMR studies of sani-dines (Brun et al., 1960) have not yielded discrete rota-tional patterns for different types of aluminum. The highresolution aluminum-27 MASS NMR lineshape of sani-dine (Fig. 4D) agrees with this and is consistent with arange of quadrupole coupling constants, asymmetry pa-rameters and/or isotropic chemical shifts. The Ho fielddependence ofthe lineshape (data not shown) shows thatthe average quadrupole coupling constant is about 3MHz, because the location and breadth of the centerbandis similar to those of albite and microcline.

We consider now the aluminum-27 spectra of theplagioclases. The spectrum of synthetic anorthite (Fig.5E) is more complicated that those of the alkali feldspars.It consists ofa peak at 61.5 ppm and a broad shoulder atabout 55 ppm, and is strongly asymmetric.

Anorthite at room temperature has eight slightly differ-

'ht n,7T MASS

P P M

Fig. 7. ll.7 Tesla, 27Al MASS spectra and computersimulations of albite and microcline perthite. (A) Albite, 4.0 kHzMASS, l0 scans at 25 sec recycle time, 4 kHz spectral width, 3g,sec 90" pulse excitation, E196 data points, l0 Hz linebroadeningdue to exponential multiplication. (B) Albite, computersimulation using quadrupole coupling constant of 3.29 MHz andasymmetry parameter of. 0.62, 65 Hz linebroadening. (C)Microcline perthite, 4.0 kHz MASS, 20 scans at 15 sec recycletime, 4 kHz spectral width, 3 g,sec 90o pulse excitation, El96 datapoints, l0 Hz linebroadening due to exponential multiplication.(D) Microcline perthite computer simulation using sum of albitesimulation (B) and a second component having a 3.22 MHzquadrupole coupling constant and asymmetry parameter of 0.21,75 Hz linebroadening due to exponential multiplication. (E)Computer simulation using 3.22 MHz quadrupole couplingconstant and asymmetry parameter of 0.21 and 75 Hzlinebroadening due to exponential multiplication.

--r[.

' io

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

SPECTRUM SIMULATION

l l E KIRKPATRICK ET AL.: NIJCLEAR MAGNETIC RESONANCE OF FELDSPARS

ent aluminum tetrahedral sites. Single crystal NMR stud-ies yield eight different electric field gradient tensors withquadrupole coupling constants ranging from 2.66 to 8.42MHz (Staehli and Brinktnann, l974a,b). The broadenedpeak in the MASS spectrum is due to overlap of thesignals from the different Al sites.

The aluminum-27 spectra of the four natural, intermedi-ate plagioclases (Figs. 5A, 58, 5C, and 5D) all have peakmaxima between 60.2 and 61.5 ppm. The peaks becomebroader and more like those of the synthetic anorthitewith increasing An-content. Even for the Pacaya anor-thite, the large peak and shoulder are less resolved thanfor the synthetic anorthite. The quadrupole broadeningprevents resolution of the many closely spaced peaks'

The aluminum-27 NMR spectra of albite and anorthiteglass (Figs. 4C and 5F) consist of broad, structurelesspeaks (about 18 and'28 ppm wide, respectively) withmaxima at 55.0 and 59.6 ppm, respectively. Quantifica-tion of the total aluminum content in albite glass using theheight of the free-induction decay yields about 75 percentof the expected intensity. Once again we interpret thesespectra as indicating that the aluminum sites in the glasshave a wide range of quadrupole coupling constants,asymmetry parameters, and/or isotropic chemical shifts.

The anorthite glass spectrum is skewed to less positivechemical shifts, with a very broad underlying component.Quantification of the aluminum indicates that only about50Vo of the expected intensity is observed. As for thealbite glass, we interpret this low intensity as indicatingthat the aluminum in the glass is in a wide range ofelectronic environments, many of which have large quad-rupole coupling constants.

Little is known about the effects of Al-O bond dis-tances or angles, total cation-oxygen bond strength, orother structural parameters on aluminum-27 chemicalshifts and quadrupole coupling constants, so it is notpossible at the present time to give a more detailedanalysis of our results in structural terms. It seems likely,however, that at least part ofthe peak broadening is dueto a wide range of electric field gradients around alumi-num in the glasses. There is no signal in the chemical shiftrange near 0 ppm, which would indicate octahedrally-coordinated aluminum, although because we do not ob-serve the full intensity expected on the basis of aluminumcontent, we cannot rule out the presence of highlydistorted, six-coordinated sites. X-ray radial distributionstudies (Taylor and Brown, 1979) and Raman spectrosco-py (Mysen, et al., 1982) do, however, suggest that 6-coordinated aluminum is unlikely to be present in theseglasses.

Sodium-23

We show in Figures 8 and 9 the ll.7 T sodium-23MASS NMR spectra for all of our sodium-containingsamples. Table 4lists the chemical shifts at the singulari-ties. Sodium-23 has spin I = 312 and like aluminum-27suffers from quadrupolar broadening. Because sodium-23

'uNo n,7 T MASS

- 5 0

Fig. 8. 23Na MASS NMR spectra of natural alkali feldspars

and synthetic albite glass at ll.7 Tesla (coresponding to a

resonance frequency of 132.3 MHz)' (A) Albite (Amelia,

Virginia), 4.0 kHz MASS, 30 scans at 25 sec recycle time' 15 kHz

spectral width, 4 (psec 90o pulse excitation, 8192 data points, l0

Hz linebroadening due to exponential multiplication. (B)

Microcline perthite (Delaware Co., Pennsylvania), 4.0 kHz

MASS, 100 scans at 25 sec recycle time, 15 kHz spectral width, 3

;.r.sec 90o pulse excitation, 8192 data points, l0 Hz linebroadeningdue to exponential multiplication. (C) Albite glass (synthetic)'

4.7 kHz MASS, 1ffi3 scans at I sec recycle time, 15 kHz spectral

width, 4 psec 90'pulse excitation, 4096 data points, 150 Hz

linebroadening due to exponential multiplication. (D) Sanidine(Wyoming), 4 kHz MASS, 159 scans at 5 sec recycle time, 15

kHz spectral width, 3.5 psec 90'pulse excitation' 4096 datapoints, 50 Hz linebroadening due to exponential multiplication.(E) Cryptoperthite ("Fredriksvarn", Norway), 4 kHz MASS'

708 scans at 5 sec recycle time, 15 kHz spectral width, 4 psec 90opulse excitation,4096 data points, 50 Hz linebroadening due to

exponential multiPlication.

has a smaller nuclear spin than aluminum-27, for the samequadrupole coupling constant, sodium has a larger sec-

ond-order quadrupole breadth and hence broader lines(Meadows et al., 1982).

The sodium-23 spectrum of low albite (Fig. 8A) con-

sists of a doublet with peaks at -13.1 and -21.9 ppm'

The split peak is not due to the presence of two sodium

sites in albite, but is caused by the second order quadru-

pole interaction, which is only partially averaged by

MASS. Single crystal sodium-23 NMR studies of low

albite have yielded a quadrupole coupling constantof 2'62

M I C R O C L I N EP E R T H I T E

KIRKPATRICK ET AL.: NT]CLEAR MAGNETIC RES0NAN1E oF FELDSPARS l l 9

n - c n

PPM from NoClo _ E n

PPM frodt NoCt

Fig. 9. 23Na MASS NMR spectra of natural plagioclasefeldspars at IL7 Tesla (corresponding to a resonance frequencyof 132.3 MHz). (A) Oligoclase (Bakersville, North Carolina). 3.6kHz MASS, ll8 scans at l0 sec recycle time, 15 kHz spectralwidth, 5 psec 90' pulse excitation, 4096 data points, 50 Hzlinebroadening due to exponential multiplication. (B)Labradorite (Sydney Mines, Labrador), 3.3 kHz MASS, 725scans at 5 sec recycle time, 15 kHz spectral width, 4 psec 90.pulse excitation, 4(D6 data points, 50 Hz linebroadening due toexponential multiplication. (C) Bytownite (Crystal Bay,Minnesota), 3.4kHz MASS, 314 scans at 5 sec recycle time, 15kHz spectral width, 4 ,rsec 90o pulse excitation, 4096 data points,50 Hz linebroadening due to exponential multiplication. (D)Anorthite (Pacaya Volcano, Guatamala), 3.4kHz MASS, 1000scans at 5 sec recycle time, 15 kHz spectral width, 4 psec 90.pulse excitation, 21096 data points, 100 Hz linebroadening due toexponential multiplication.

MHz and an asymmetry parameter of 0.25 (Hafner andHartmann, 1964). The static and MASS lineshapes as afunction of magnetic field strength (Fig. l0) are consistentwith the lineshape being dominated by the quadrupoleinteraction. The 3.52 T MASS spectrum is very complex,containing many spinning sidebands, due to the fact thatthe spinning speed is much less than the spectral breadth,when expressed in frequency units.

We have obtained an excellent simulation of the I1.7 TMASS spectrum of albite (Fie. il) assuming a singlesodium site with a quadrupole coupling constant of 2.5gMHz and an asymmetry parameter of 0.25, and yieldingan isotropic chemical shift of -6.8-{-0.3 ppm, in goodagreement with the value of -7.3 ppm obtained byLippmaa et al. (1980). The agreement between the experi-mental and simulated spectra clearly indicates that thereis only one sodium site in albite.

The values of the quadrupole coupling constant andasymmetry parameter of sodium-23 in albite are con-firmed by variable angle sample-spinning (VASS) NMR,

Figure I I, which generates different lineshapes dependingon the spinning angle, when the second order quadrupoleinteraction dominates (Oldfield et al., 1983;Ganapathy etal.,1982). As expected, better averaging is obtained near40o or 75". Note that at 75' the first order spinningsidebands are absent (Ganapathy et al., 1982).

The sodium-23 spectrum of microcline perthite at ll.7T (Fig. EB) shows peaks at -13.1, -21.9 and -26.5 ppm.From the Ho magnetic field dependence of the lineshape(data not shown) there appear to be two types of sodium,one in exsolved low albite (-13.1 and -21.9 ppm) andone in microcline (-26.5 ppm). The -26.5 ppm line is notappreciably shifted in going from 8.45 T to ll.7 T,suggesting that the sodium site in microcline has a smallerelectric field gradient and, hence, is in a more symmetricenvironment than in albite. Spectra of microcline ob-tained at ll.7 T simulate well with an albite componentand an additional llVo contibution from a line with aquadrupole coupling constant of about 0.5 MHz, thisadditional intensity being from sodium in the microcline.The low quadrupole coupling constant is in agreementwith the small magnetic field dependence of the chemicalshift. However, because equally good spectral simula-tions can be obtained using two powder patterns havinge2qQ/h of about 3 MHz, further work is needed in order tosubstantiate this very tentative assignment.

The sodium-23 MASS NMR spectrum of the crypto-perthite (Fig. 8E) shows a doublet due to the second-order quadrupole interaction. These peaks are very simi-lar to those of albite, except that they are less wellresolved. This spectrum is in agreement with the silicon-29 and aluminum-Z7 data, which indicate that some Si/Alordering has occurred in this material, but that neither isas complete as in low albite or microcline perthite.

Table 4. Sodium-23 NMR chemical shifts of feldspars andfeldspar glasses

FrdLD p"anett tSTRENGTH frE{I(;AL MMNTS( tEsLA) SqISl

tuNo n,7 T MASS

STERIAL

Alb t te (Ab99)

A lb lEe c lasB (Abroo)

sanrd tne ( "o r60)

I l c roc l tne Per rh t re

Cr tp toper th t re

o l t8oc lase (An23)

Labrador t t€ (An52)

Bytoh l te (An70)

Ano(Eh l te (e92)

Quad(upo le Spt t t

Qudrupore Sp l l r

Quadrupo le Sp l l r

r t . 7

r l . 7

- I 1 . 1- 2 1 . 9

- 1 3

-21

l I . 7

1 1 . 7

l l . 7

l r . 7

rFor quadrupo lar nuc le t apparedr cheo lca t sh l r r chanS€s d i !h f te ld s r reErh . I f

peak le oo t sp l l t by second-order quadrupo lar e f fecrs r va lue ar t t .7 Tes la tse€e€nt ta l l y th€ t3or rop lc chet r tca l sh t f t .

- I l

-26

- r 3-22

- 1 5

- I 6

- 1 4

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARSr20

, T.t 5C3

PPM t?6 NoCl

ttNo ALBTTE

STATIC

il,7 7

8,457

Fig. 10. 23Na static and MASS NMR spectra of albite at 11.7,

8.45 and 3.5 Tesla. (A) 11.7 Tesla, static,30 scans at 25 secrecycle time, 132 kHz spectral width, 3 psec 90'pulse excitation'4096 data points, 50 Hz linebroadening due to exponentialmultiplication. (B) 1l .7 Tesla, 4.0 kHz MASS, 30 scans at 25 secrecycle time, l3zkHz spectral width, 3 psec 90" pulse excitation,El96 data points, l0 Hz linebroadening due to exponentialmultiplication. (C) 8.45 Tesla, static, 100 scans at 25 sec recycletime, 95 kHz spectral width, 3.5 psec 90'pulse excitation' 4096data points, 50 Hz linebroadening due to exponentialmultiplication. (D) E.45 Tesla, 4.8 kHz MASS, 100 scans at 25

sec recycle time, 95 kHz spectral width, 3 psec 90' pulse

excitation, El92 data points, 50 Hz linebroadening due toexponential multiplication. (E) 3.5 Tesla, static, 1000 scans at 25

sec recycle time, 39 kHz spectral width, 3.5 psec 90' pulse

excitation, 40!)6 data points, 100 Hz linebroadening due to

exponential multiplication. (F) 3.5 Tesla, 3'5 kHz MASS' 1000scans at 15 sec recycle time, 39 kHz spectral width, 3 psec 90opulse excitation,4096 data points, 100 Hz linebroadening due to

exponential multiplication.

The sodium-23 MASS NMR spectrum of sanidine (Fig.

8D) consists of a single broad peak at -21'2 ppm, which

is slightly broader than those of albite and microcline

perthite. The lack of a well resolved quadrupolar splitting

for sanidine suggests to us, once again, that Al/Si disorder

results in a range of sodium sites with different quadru-

pole coupling constants, asymmetry parameters' and/or

chemical shifts. The presence of a split sodium-23 peak in

alkali feldspars appears to be a good indicator of AUSi

ordering.The sodium-23 MASS NMR spectra of the intermediate

plagioclases (Fig. 9A-9D) all show single, relatively sharp

centerband resonances in the range of -13.0 to -15.9

ppm, that are similar in shape and breadth to the sanidine

spectrum. As for sanidine, this probably indicates a range

of sites with different quadrupole coupling constants,

asymmetry parameters, and/or isotropic chemical shifts'

In the plagioclases, much of this is probably due to

distortion of the sodium sites near the margins of albite

and anorthite volumes. The relatively large spinning

sidebands in these spectra, like the aluminum-27 NMR

spectra of these same samples (Fig. 5), are probably due

to the presence of magnetic impurities (Oldfield et al.'

1983).The sodium-23 NMR spectrum of albite glass (Fig' 8C)

is very broad and featureless, again indicating sites with a

range of quadrupole coupling constants, asymmetry pa-

rameters. and/or chemical shifts.

SummarY and conclusions

The foregoing results demonstrate that MASS-VASS

NMR spectroscopy complements well the other main,

reasonably direct methods of investigating the structures

of feldspars-single-crystal X-ray and neutron diffraction,

electron diffraction, transmission electron microscopy(TEM) and single crystal NMR spectroscopy' Each of

these methods has its strengths and its weaknesses' Theparticular strength of NMR spectroscopy is that in many

ttNo 8,45 T VASS

SPECTRUM SIMULATIONA

750

DSTATIC

l o J J . , " - 1 J J r co o r r l l - 1oc

anele), (C) 40', (D) static'

KIRKPATRICK ET AL.: NLICLEAR MAGNETIC RES2NANCE oF FELDSPARS 12l

instances it provides direct evidence about the electronicenvironment of individual kinds of atomic sites, even forthose ditrering as little from one another as the Tlm, T2mand T2o silicon sites in the triclinic alkali feldspars.Moreover, MASS-VASS NMR allows direct and easycomparison of sites in crystalline and noncrystallinematerials. MASS-VASS NMR spectroscopy uses easilyobtained powder samples whereas single-crystal NMRspectroscopy requires difficult-to-obtain, large, high_quality single crystals. The major limitations on MASS-VASS NMR samples are that they must be relatively freeof paramagnetic impurities and that at least about l-10 mgfor sodium-23 and aluminum-27 andabout 100-500 mg forsilicon-29 must be available.

Although MASS-VASS NMR spectrometers currentlyare quite expensive and not numerous, this is likely tochange for the better in the near future.

We now summarize for each type of material we haveexamined the results presented above. Again, we empha_size that while our interpretations are consistent with theavailable knowledge of and ideas about feldspar struc-ture, they must at this time be considered tentative. Ourinterpretations of the intermediate plagioclase spectra arethe most uncertain.

Alkali feldspars

Albite. Because it has an ordered and well-understoodstructure, the NMR spectra of low albite are the leastambiguous to interpret. The silicon-29 NMR spectrumhas three well-resolved peaks of nearly the same intensitydue to the three different silicon sites. The aluminum-27NMR spectrum is a single relatively narrow peak, due tothe one tetrahedral aluminum site. Both the silicon-29 andaluminum-27 data clearly support the high degree of Al/Sitetrahedral site ordering deduced by X-ray diffraction.The sodium-23 spectrum consists of a sharp, well-re-solved doublet due to the second-order quadrupolar inter-action, indicative of a rather asymmetric sodium site. Allof these conclusions are consistent with the well-knownstructure of albite.

Microcline perthite. The spectra of this sample reflectthe presence of both microcline and low albite. Thesilicon-29 NMR spectrum contains three peaks which areassignable to the three silicon sites in microcline, and twosmall peaks due to the small amount of intergrown albite.The third albite peak apparently underlies two of themicrocline peaks. The aluminum-27 NMR spectrum ofthe perthite consists of a quadrupole split doublet fromthe microcline, and extra intensity due to the albite. Thesodium-23 NMR spectrum consists of a quadrupolar splitdoublet, due to low albite, and a second component dueto sodium in the microcline. The sodium site in themicrocline is tentatively thought to be in an envlronmentmore symmetric than that in albite.

Cryptoperthite. The NMR spectra of the cryptoperthite(Spencer R) are consistent with partiat AVSi ordering andseparation into potassium and sodium-rich volumes. The

silicon-29 NMR spectrum is relatively well separated intoa number of peaks. Three of these can be assigned toalbiteJike sites and to Tl and T2 sites in orthoclase (lowsanidine). The resolution ofthe peaks is, however, not asclear as for the microcline perthite. There is only onealuminum-27 resonance, again indicating that neitherphase has a composition approaching an end member.The sodium-23 NMR spectrum shows a somewhatsmoothed out doublet due to quadrupolar splitting, againsuggestive of at least enough Al/Si ordering to causedistortion of the sodium site. There is no sodium-23resonance specifically assignable to the orthoclase.

Sanidine. Because sanidine is highly disordered, itsspectra are also relatively easy to interpret. The twobroad, poorly split silicon-29 NMR peaks are due to arange of silicon sites caused by AUSi disorder in the Tland T2 sites and to Or-Ab solid solution. These peakscover the chemical shift range of the albite spectrum andindicate a continuum of electron distributions in e4(0Al),Q4(1Al), Q4(2Al), and e4(3Al) sires. The aluminum-27NMR spectrum consists of a single peak which is broaderthan that of albite, also suggesting a range of aluminumenvironments in Tl and T2. The sodium-23 resonance isnot split by quadrupolar effects, suggesting a range ofsodium environments and probably a more symmetricenvrronment.

Plagioclases

Anorthite. The silicon-29 NMR spectrum of syntheticanorthite consists of three Q4t+et) peaks that are assign-able to the eight silicon sites in anorthite plus a small peakthat we assign to Q4(3AD sites. The silicon-29 NMRspectrum of the natural anorthite (Ansr, a phenocrystfrom Pacaya volcano (Guatamala), is similar but less wellresolved. The aluminum-27 NMR spectrum of both anor-thites are broad, with one well resolved peak, and ashoulder at less positive (more shielded) chemical shifts.MASS NMR is apparently not able to resolve the eightdifferent aluminum sites. The sodium-23 spectrum ofnatural anorthite is discussed below along with those ofthe other intermediate plagioclases.

Intermediate plagioclases. The spectra of the interme-diate plagioclases seem to be consistent with structuralmodels involving e- and /-plagioclases with e4(4Al) anor-thite-like sires, e4(lAl) and e4(2Al) albite-like sites, anddistorted equivalents ofall three in strained regions wherethe albite-like and anorthite-like volumes meet. e4(3Al)sites may also be present at the boundaries.

The silicon-29 NMR spectra of the oligoclase (An3)and labradorite (An5) exhibit three peaks that can bereadily assigned ro Q4(1Al) and e4(2Al) sites in albite.The peaks do not all have the same height, as they do inlow-albite. We have attributed the extra height in the -93and - lfi) ppm range to deshielding of the albite-like sitesand increased shielding of the anorthite-like sites causedby coherency strain of the silicon sites in e-plagioclaseand, perhaps, to Q4(3AD sites at the margins. In the

t22 KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RESONANCE OF FELDSPARS

bytownite (Anro) the most shielded silicon peak is notpresent or is very small, perhaps suggesting that all ofthealbite-like volumes have strained silicon sites. The natu-ral anorthite (Anqz) has no peaks attributable to albite-likesites.

Peaks or shoulders in the silicon-29 NMR spectrawhich are attributable to anorthite-like sites increase inintensity with increasing An-content. For the oligoclase,there is only one shoulder at a relatively shielded chemi-cal shift (-88 ppm), which we interpret as arising fromstrained Qo(4Al) anorthite-like sites. The peaks assign-able to Qo(4Al) sites are larger in labradorite. We haveattributed these peaks to anorthite-like sites in e-plagio-clases. The peaks attributable to anorthite-like sites in thebytownite are very large. We attribute these peaks pri-marily to sites in /-anorthite, although some of the signalmust come from sites in the e-plagioclase. As notedabove, the silicon-29 NMR spectrum of the natural anor-thite is similar to the synthetic anorthite, but less wellresolved.

The aluminum-27 NMR spectra of the intermediateplagioclases consist of broad peaks that become more likethat of anorthite with increasing An content.

The sodium-23 NMR spectra show single, unsplit peaksof about the same breadth as in sanidine. Because thesilicon-29 NMR peaks indicate significant Al/Si order, wetentatively attribute the non-split sodium-23 resonancesto variable amounts of strain of the sodium sites, whichincreases the range of electronic environments aroundsodium.

Glasses

The NMR spectra of the albite and anorthite glassesconsist of single, essentially structureless peaks that arebroader than the spectra ofcrystalline albite or anorthite.The silicon-29 resonances are broader by about 20 ppm'The aluminum-27 resonance for albite glass is about 20ppm wider than that of low albite, whereas the aluminum-27 resonance of anorthite glass is only a few ppm widerthan that of synthetic crystalline anorthite. This wouldseem to indicate a much larger increase in the range ofaluminum sites in albite glass. The sodium-23 resonanceof albite glass is about 20 ppm broader than that of lowalbite, again indicating a somewhat larger range of elec-tric field gradients at the sodium sites in the glassy phase.

Acknowledgments

We wish to thank Drs. Glenn Buckley, David Stewart, andJohn Higgins, who supplied us with some of the samples andcompositions, W. -H. Yang and Richard Oestrike for usefuldiscussions, and an anonymous reviewer for many useful sugges-tions. This work was supported in part by the Solid-StateChemistry and Experimental and Theoretical Geochemistry Pro-grams of the U.S. National Science Foundation (Grant DMR8311339 and EAR 8207260 and 8408421) and in part by the U.S'National Institutes of Health, (Grant HL-19481) and has benefit-ted from facilities provided by the University of lllinois-National Science Foundation Regional NMR Instrumentation

Facility. E. O. is a USPHS Research Career Development

Awardee, 1979-1984.

References

Abragam, A. (1961) The Principles of Nuclear Magnetism'

Clarendon Press. Oxford.Andrew, E. R. (1971) The narrowing of NMR spectra of solids by

high-speed specimen rotation and the resolution of chemicalshift and spin multiplet structures for solids' Progress in NMR

Spectroscopy, E, 1-39.Bollmann, W. and Nissen, H. -H' (196E) A study of optimal

phase boundaries: the case of exsolved alkali feldspars. Acta

Crystallograph ica, A24, 546-5 57 .Brown, B. E. and Bailey, S. W. (1964) The structure of maxi-

mum microcline. Acta Crystallographica, 17, 1391-1400'Brun, E., Hartmann, P., Staub, H. H., Hafner, S., and Laves, F'

(1960) Magnetische kernresonanzz:ur beobachtung des Al' Si-

ordnungs/unordnungsgrades in einigen feldspiiten. Zeitschriftfiir Kristallographie, I 13, 65-76.

Farrar, T. C. and Becker, E. D. (1971) Pulse and Fourier

Transform NMR: introduction to theory and methods. Aca-

demic Press, New York.Fyfe, C. A., Gobbi, G. C., Klinowski, J., Thomas, J. M., and

Ramdas, S. (1982) Resolving crystallographically distinct tet-

rahedral sites in silicalite and ZSM-5 by solid-state NMR.

Nature, 296,530-516.Ganapathy, S., Schramm, S., and Oldfield' E. (1982) Variable-

angle sample spinning high resolution NMR of solids. Journal

of Chemical Physics, 77, 43ffi-4365.Ghose, S. and Tsang, T. (1973) Structural dependence of quadru-

pole coupling constant e2qQ/h for 27Al and crystal field param-

eter D for Fe*3 in aluminosilicates. American Mineralogist'

58, 748-755.Goldsmith, J. R. and Laves, F. (1955) Cation-order in anorthite

(CaAlzSizOs) as revealed by gallium and germanium substitu-

tions. Zeitschrift fiir Kristallographie, 106, 213-226.Hafner, S. and Hartmann, P. (1964) Elektrische Feldgradienten

und Sauerstof-Polarisierbarkeit in Alkali-Feldspiiten (NaAl-

SirOs) und KAlSi3Os). Helvetica Physica Acta, 37, 348-360'

Hafner, S., Hartmann, P., and Laves, F. (1962) MagnetischeKernresonanz von Al27 in Adular. Zur Deutung der Adular-

struktur. Schweizerische Mineralogische und PetrographischeMitteilungen, 42, 277 -294.

Hafner, S. and Laves, F. (1963) Magnetische Kernresonanz von

AI27 in einigen Orthoklasen. Schweizerische Mineralogische

und Petrographische Mitteilun gen, 43, 65-69 -

Harlow, G. E. and Brown, G. E. (1980) Low albite: an X-ray and

neutron diffraction study. American Mineralogist' 65' 986-

99s.Higgins, J. 8., and Woessner, D. E. (19E2) 2esi, 27A1, and 23Na

spectra of framework silicates. (abstr.) Transactions of the

American Geophysical Union (EOS)' 63,1139.Kalus, C. K. (1978) Ph.D. Thesis, Universitv of Munich. (Title

unavailable.)Klinowski, J., Ramdas, S., Thomas, J. M., Fyfe' C. A. and

Hartman, J. S. (1982) A re-examination of Si, Al ordering in

zeolites NaX and NaY. Journal of the Chemical Society,

Faraday Transactions II, 78, 1025-1050.Kundla, E., Samoson, A', and Lippmaa, E.' (1981) High-

resolution NMR of quadrupolar nuclei in rotating solids.

Chemical Physics Letters, 83, 229-232.Lippmaa, E., Magi, M., Samoson, A., Engelhardt, G., and

KIRKPATRICK ET AL.: NUCLEAR MAGNETIC RES2NANCE oF FELDSPARS 123

Grimmer, A. -R. (1980) Structural studies of silicates bv solidstate high-resolution 2eSi NMR. Journal of the AmericanChemical Society, 102, 48E9-4893.

Lippmaa, E., Magi, M., Samoson, A., Tarmak, M., and Engel-hardt, G. (1981) Investigation of the structure of zeolites bvsolid-state high-resolution 2esi NMR spectroscopy. Journal ofthe American Chemical Society, l03,49924996.

Loewenstein, W. (1954) The distribution of aluminum in thetetrahedra of silicates and aluminates. American Mineralogist,39,92-96.

Mclaren, A. C. (1974) Transmission electron microscopy of thefeldspars. In W. S. MacKenzie and J. Zussman, The Feld_spars, p. 378-423. Manchester University press.

Meadows, M. D., Smith, K. A., Kinsey, R. A., Rothgeb, T. M.,Skarjune, R. P., and Oldfield, E. (19S2) High-resolution solid_state NMR of quadrupolar nuclei. proceedings of the NationalAcademy of Science of the United States, 79, 135l_1355.

Megaw, H. D., Kempster, C. J. E., and Radoslovich, E. W.(1962) The structure of anorthite, CaAl2Si2O8. II. Descriptionand discussion. Acta Crystallographica 15, 1017-1035.

Muir, I. D. and J. V. Smith (1956) Crystallisation of feldspars inlarvikites. Zeitschrift fiir Kristallographie, 107, l g2_195.

Mtiller, D., Gessner, W., Behrens, H. J., and Scheler G. 09gl)Determination of the aluminum coordination in aluminum-oxygen compounds by solid-state high-resolution 27Al NMR.Chemical Physics Letters, j9, 59-62.

Mysen, B. O., Virgo, D., and Seifert, F. A. (19g2) The structureof silicate melts: implications of chemical and physical proper_ties of natural magma. Reviews of Geophysics and the SpacePhysics, 20, 353-383.

Oftedahl, C. (1948) Studies on the igneous rock complex of theOslo region. IX. The feldspars. Skrifter utgit av Det NorskeVidenskaps-Akademi i Oslo I. Mathematisk-Naturidenskapel_lig Klasse. 1948. No. 3. Naturvidensk Kl. 3.

Oldfield, E., Kinsey, R. A., Smith, K. A., Nichols, J. A., andKirkpatrick, R. J. (1983) High-resolution NMR of inorganicsolids. Influence of magnetic centers on magic-angle sample-spinning lineshapes in some natural alumino silicates. Journalof Magnetic Resonance, Sl, 325-327.

Ribbe, P. H., (1983) Feldspar Mineralogy, Vol. 2. Reviews inMineralogy, 2nd Edition. Mineralogical Society of America,Washington, D.C.

Schramm, S. E. and Oldfield, E. (19E2) High resolution solidstate NMR of quadrupolar nuclei: quadrupble induced shifts invariable-angle sample spinning of a borosilicate glass. Journalof the Chemical Society, Chemical Communications, 980-81.

Smith, J. V. (1974) Feldspar Minerals, volume l. Springer-Verlag, New York.

Smith, J. V. and Blackwell C. S. (1983) Nuclear magneticresonance of silica polymorphs. Nature, 223-225.

Smith, J. V., Blackwell, C. S., and Hovis, G. L. (1984) Nuclearmagnetic resonance of albite-microcline series. Nature. 309.r$-t42.

Smith, I. V. and Muir, L D. (1958) The reaction sequence inlarvikite feldspars. Zeitschrift f0r Kristallographie, I 10, I l-20.

Smith, K. A., Kirkpatrick, R. J., Oldfield, E., and Henderson,D. M. (1983) High-resolution silicon-29 nuclear magnetic reso-nance spectroscopic study of rock forming silicates. AmericanMinerafogist, 68, l20Ll2l5.

Spencer, E. (1937) The potash-soda feldspars L Thermal stabil-ity. Mineralogical Magazine 24, 453494.

Staehli, J. L. and Brinkmann, D. (1974a) A nuclear magneticresonance study of the phase transition in anorthite, CaAl2Si2Os. Zeitschrift fiir Kristallographie, 140, 3601373.

Staehli, J. L. and Brinkmann, D. (1974b) Assignment andstructural dependence of electric field gradients in anorthiteand simple field gradient calculations in some aluminosilicates.Zeitschrift ftir Kristallographie, 140, 37 4-392.

Stewart, D. B. and Wright, T. L. Onq AUSi order andsymmetry of natural alkali feldspars and the relationship ofstrained cell parameters to bulk composition. Bulletin de laSocidte Frangaise de Mineralogie et de Cristallographie, gT,356-377.

Taylor, M., and Brown, G. E. (1979) Srructure of mineralglasses. I. the feldspar glasses NaAlSijO8, KAlSi3Os, CaAl2Si2Os. Geochimica et Cosmochimica Acta, 43, 61J7.

Wainwright, J. E., and Starkey, J. (1971) A refinement of thestructure of anorthite. Zeitschrift fiir Kristallographie, 133,75-E4.

Manuscript received, October 26, 1983;accepted for publication, September 4, 1984.