Clay Minerals (1986) 21, 341-360

NICKEL-BEARING

CLAY

II. INTRACRYSTALLINE NICKEL: AN X-RAY A.

MANCEAU

MINERALS:

DISTRIBUTION OF ABSORPTION STUDY AND G.

CALAS

Laboratoire de Mindralogie-Cristallographie, UA CNRS 09, UniversitdsParis 6 et 7, 4 place Jussieu, 75230 Paris Cedex 05, and Laboratoirepour l'Utilisation du Rayonnement Electromagndtique (LURE), CNRS, 91405 Orsay, France (Received 7 February 1986) A B S T R A C T: The mechanism of Ni-Mg substitution has been studied by X-ray absorption spectroscopy in phyllosilicates belonging mainly to the lizardite--nepouite and kerolite-pimelite series. Two types of information were obtained: (1) Analysis of nickel K-edge spectra under high resolution confirmed that Ni atoms were substituted for Mg atoms. There was no evidence for 4-fold coordinated Ni. (2) Extended X-ray absorption fine structure (EXAFS) was sensitive to atomic pair-correlations and gave access to the radial distribution function around Ni atoms. For all samples, this function gave two peaks. The first one was related to the (O,OH) coordination shell and analysis confirmed that Ni atoms were 6-fold coordinated. The amplitude of the second peak was very sensitive to the atomic composition of the Ni-Mg second shell. It is shown that the intracrystalline distribution of Ni is never random within the octahedral sheet; Ni atoms are segregated into domains, the minimum size of which has been calculated. In the kerolite--pimelite series the mean domain size is at least 30 ./~ and EXAFS could not exclude the existence of pure Ni sheets. X-ray dispersive spectroscopy combined with TEM suggested that the minerals in this series have pure Ni layers associated with pure Mg layers. In the lizardite-nepouite series, Ni atoms are segregated into specific Ni-enriched areas, the exent of which depends on the specific chemical constitution of the sample. Distribution patterns are discussed with respect to the formation mechanisms of these ore minerals.

Spectroscopic techniques are now widely used for characterizing the crystal chemistry of transition elements in crystals. Classically these techniques give information on the site occupancy and oxidation state of the particular element. Several methods provide information on intracrystaUine order, and among them vibrational and nuclear magnetic resonance (NMR) spectroscopy are now used to characterize the distribution of cations within the octahedral sheet of phyUosilicates (Gerard & Herbillon, 1983; Sanz & Stone, 1983). X-ray absorption spectroscopy is a selective probe sensitive to the local environment around a chosen atom and it therefore provides structural information on both short- and medium-range order by contrast with optical spectroscopy which is mainly c o n c e r n e d with the first c o o r d i n a t i o n shell (cf. M a n c e a u et al., 1985). In the present study an a t t e m p t is m a d e to c h a r a c t e r i z e the intracrystalline distribution o f nickel in s o m e phyllosilicates. EXPERIMENTAL

METHOD

AND

ANALYSIS

Experimental conditions M e a s u r e m e n t s w e r e p e r f o r m e d at the L U R E s y n c h r o t r o n radiation facility, using the 9 1986 The Mineralogical Society

A. Manceau and G. Calas

342

X - r a y beam emitted by the D C I storage ring (positron energy of 1.72 GeV). The absorbance of the sample is recorded in a transmission mode as a function of the incident photon energy: ~(E) = Ln I~ /l(g) where I 0 = intensity of the incident beam and 11 --- intensity of the transmitted beam. The apparatus has been described by Raoux et al. (1980). The monochromator is a 'channel-cut' single crystal of silicon, the 400 reflection being used. The spectral resolution, which is limited by the geometry of the apparatus (source size and collimation) and by the 1s-level width, is about 1 eV at the Ni K-edge (8331 eV). For study o f the Ni K-edge structure, the energy was scanned at 0.25 eV steps, the reproducibility of the spectra being assessed by comparison with a reference compound every two records. For E X A F S data acquisition, the energy was scanned with 2 eV steps from 300 eV below the absorption edge of Ni to 900 eV above. The Ni-bearing clay minerals investigated were those previously studied by optical absorption spectroscopy (Manceau et al., 1985). The samples belong to the kerolitepimelite and lizardite-nepouite series: chemical data for these and a Ni-chlorite and Ni-sepiolite are given in Table 1.

X-ray absorption spectroscopy Although the technique of X-ray absorption spectroscopy has been known for about half a century, it has only been used as a structural tool for the last decade due to the TABLE 1. Mg, Ni and Fe atomic contents of the octahedral sheets (mol %) of the samples. Complete chemical analyses are given by Manceau et al. (1985).

Pimelite P2 P1 Kerolite K5 K4 K3 K2 Nepouite N1 Lizardite L8 L7 L6 L5 L4 L3 L2 LI Amesite AI* Sepiolite SP Chlorite CL*

Mg

Ni

Fe

12 39 58 73 73 86 10 65 75 85 89 92 88 89 95 95 69 64

88 60 42 27 27 14 89 31 21 9 8 6 6 4 3 5 30 30

-1 ----1 4 4 6 3 2 6 7 2 -1 6

* Octahedral AI added to Mg.

Intracrystalline distribution of nickel

343

availability of synchrotron radiation (Calas et al., 1984). Three distinct regions are classically considered at an absorption edge; these contain different kinds of information (Fig. 1). (1) The K (or L) absorption edge itself is due to the transition of core-level electrons towards excited bound states of the absorbing atom. Structure and energy position of the edge carry information about oxidation state, coordination number, site distortion and metal-ligand covalency. (2) The region just above the edge up to ~100 eV is related to multiple scattering processes and constitutes the X-ray absorption near-edge structure (XANES). It may be used to investigate the local geometry, including bond angles, but is presently understood only in simple compounds. (3) The oscillations of the absorption coefficient on the high-energy side of an absorption edge constitute the extended X-ray absorption fine structure (EXAFS). It extends up to 1000 eV above the edge and results from an interference effect between the outgoing photoelectron wave and its fraction which is backscattered by the neighbouring atoms. EXAFS carries structural information about the nearest coordination shells surrounding the absorbing atom, i.e. interatomic distances, number and nature of the constituent atoms. In this study only structural information provided by the absorption edge and EXAFS were exploited, with the aim of improving the precision of location of Ni atoms within the sheet silicates and the Ni-Mg intracrystalline distribution.

Outline of EXAFS theory The theory of EXAFS is now well established and quantitative comparison between calculated and experimental spectra is possible (see reviews of Raoux et al., 1980; Lee et

K-EDGE

I

9

•

I

I

I

I

I

Ni-Talc

II 8200

I

I

8500

I

I

I

I

8800

PHOTON ENERGY(eV) FIG. 1. X-ray absorption spectrum of Ni-talc at the Ni K-edge.

I

A. Manteau and G. Calas

344

al., 1981; Hayes & Boyce, 1982). The EXAFS signal z(k) may be described as a function of the wavevector k by a sum of damped sinusoids, each one attached to a j atomic shell =

1

Nj

Fj(k, z0 sin (2kRj(k)

+

Cj(k))

(1) where: Rj = distance between the absorbing atom and the jth shell, Ns. = coordination number of thejth shell, Fj(k,n) = backscattering amplitude function corresponding to the atomic species in this shell, sin (2kRj + Cj(k)) = oscillatory term including a phase shift Cj(k) due to both the central and backscattering atoms, Aj(k) = amplitude factor including (i) a damping term similar to a Debye-Waller factor e x p ( - 2 a ] k 2) where aj(A) is the standard deviation of the Rj distances, and (ii) a damping term exp (-2RjF/k) where F(A -2) is a mean free path parameter. For a given element, the magnitude of z(k) is directly proportional to the number of atoms in the shell. It depends also on the absolute value F(k, 7t), varying with the atomic number of the backscattering atoms (Teo & Lee, 1980). As a consequence, knowledge of the coordination number gives access to the nature of the backscatterer. In fact, the nearly equal scattering factors of Si, AI and Mg, on the one hand, and of Fe and Ni on the other, preclude any further distinction between atoms inside these groups on the basis of intensity measurements. Thus, it is only possible to recognize if a given position is occupied either by Ni, Fe atoms or by Mg and equivalent atoms. The period of the interference factor depends on interatomic distances Rj and on the phase shift r of the scattering atom (and thus on its atomic number): knowledge of one of these two parameters gives access to the other. In clay minerals, the cation-cation distance may be known from XRD measurements (b/3), which permits determination of r from x(k): this procedure provides another means of determining the nature of the backscattering atoms. This latter procedure is particularly sensitive for the distinction of light atoms such as Mg from the 3d transition elements as the difference in r is nearly n-radians. Hence, the occurrence of these two species of cations in one single shell results in partial destructive interferences of the backscattered photoelectric wave.

Analysis of EXAFS spectra This consists of a standard procedure for pre-edge and post-edge background removal, extraction of the EXAFS signal x(k), Fourier transform of k•(k) and inverse transform to isolate the EXAFS contribution from a selected region in real space (Hayes & Boyce, 1982). All the treatment parameters are kept fixed to ensure comparison between spectra. The EXAFS x(k) obtained after subtraction of the post-edge background and subsequently normalized to the edge absorption is presented in Fig. 2a for Ni-talc. A Fourier transform performed on kz(k) using a flat window in the range 3.7-12 /~-1 terminated on both sides by a cosine function yields a radial structure function F(R) in real space (Fig. 2b). Phase corrections are not included in the Fourier transform so that the apparent position of each peak is shifted with respect to its actual position and cannot be

Intracrystalline distribution of nickel

345

a .003

0.1O.

/

:

9

A

N i - t a l cb

.002

VVV

u.

-0.1-

.001 1

l

I

I

4 6 8 10 WAVEVECTOR K(~,)-1 ~ .02

I

12

1

2

3

4

R(/i,)

AEXPERIMENTAL C -I./ A SPECTRA A v

-.02

I

l V : .: .yv

:,c

I

4 6 8e - 1 WAVEVECTOR K(A)

10

I

4 6 8 WAVEVECTOR K(A) -1

12

I

10

I

12

FIG. 2. Treatment of X-ray absorption spectra and EXAFS data of reference compounds. (a) Normalized EXAFS spectrum. Comparison of F(R), (b) and x(k) functions (c) corresponding to the second peak for Ni-talc and Ni(OH)v (d) Theoreticalg(k) functions corresponding to 6Ni at 3.04/k and 4Si at 3.21 A. Phases are clearlyconstructive.

used to determine bond distances directly. Usually each peak of F(R) is associated with a definite shell; the backtransform of a given structural peak yields a partial EXAFS spectrum (filtered kz(k)) (Fig. 2c), i.e. the damped sinusoid corresponding to the particular atomic shell j (see equation (1)). The structural parameters are then calculated by a least-squares fitting. This fitting procedure assumes that the chemical composition of the shell as well as its structural parameters (N, R, o, F) are known; a model spectrum is then calculated (see Fig. 2d) using equation (1) and theoretical scattering factors (Teo & Lee, 1980), the validity of which has been tested with bunsenite (NiO) and Ni(OH) 2 (Table 2). The best fit is obtained by minimizing the agreement factor Q defined by: : Q=

~(Zth(k) -- Zobs(k))2 ~(Xobs(k))2

AS mentioned on p. 344, knowledge of the RJ values permits determination of the nature and number of each atomic species in an unknown shell. The determination of N requires knowledge of F which has been determined from reference compounds (Ni-talc, nepouite) and held constant during fitting of all the other spectra. The absolute accuracy of N determination is estimated to be _+1-0 for the first shell and +0.5 for the second shell. For R, absolute accuracy is estimated to +0.01 ,~ for the first shell and +0.03 A for the others. Nevertheless, a relative precision of about 0-01 ~, is attained by comparing spectra studied under identical treatment conditions.

346

A. Manceau and G. Calas NI K - E D G E

SPECTROSCOPY

These spectra include two distinct regions (Fig. 3a): (i) the 'pre-edge' peak corresponds to transitions of ls electrons towards bound empty levels of partly 3d character; (ii) the main part of K-edge is due to dipole-allowed ls-np transitions.

Pre-edge features The pre-peaks are isolated following the procedure described by Calas & Petiau (1983a) and subsequently normalized to the main edge intensity. The normalized pre-edge of the kerolite K5 is shown in Fig. 3b and compared with three reference compounds: (i) nickel hydroxide where Ni(II) atoms occupy a regular 6-fold site, (ii) Nio.gZno.lCr204 where Ni(II) is 4-fold coordinated and (iii) LaNiO 3 as a reference for Ni(III). The absorption maximum of LaNiO3 is shifted by 2 eV towards higher energies compared to the other compounds. Increasing oxidation state is known to shift the pre-peak in the direction of higher energy (chemical shift). The energy position of the pre-peaks confirms that no oxidation state of Ni other than divalent is present in clay minerals. The pre-peak intensity provides information about the site geometry. The ls-3d transitions are dipole forbidden but the absence of inversion centres favours the mixing of 3d andp orbitals. This increases the transition probability: the pre-peak intensity is enhanced in the tetrahedral symmetry relative to the octahedral one, which allows estimation of the 4-fold to 6-fold ratio (Calas & Petiau, 1983a). When nickel occurs in tetrahedral sites, the pre-peak intensity is increased by a factor of six against octahedral Ni (Fig. 3b). Attention is focussed on kerolites because Tejedor et al. 0983) assumed some nickel to be tetrahedral. The normalized pre-peak absorbance of kerolites is nearly equal to that of the 6-fold references (Table 3). These data clearly establish that 4-fold Ni(II) does not exist in significant

TABLE 2. Structural parameters of reference compounds. Scattering phases aad amplitudes from Teo & Lee (1980). Distance (A) Reference compounds

Shell atom

NiO

O Ni OH Ni O,OH Si Ni O,OH Si Ni

Ni(OH) 2 Ni-talc

Nepouite N1

XRD 2.09 2.95 2.07i" 3.12

* Fixed parameter t Brindley & Chih-chun Kao (1984)

EXAFS

a (A)

F (A 2)

N*

Q2

2.08

0. I 1

2.45

0-012

2.05 3.09 2.04 3.21 3.00 2.03 3.20 3.06

0.09 0.09 0.09 0.08 0.09 0.10 0.06 0.10

2.25 1.30 2.10 1.05 1-00 2.15 0.80 0.80

6.0 12.0 6.0 6.0 6.0 4.0 6.0 6.0 2.0 6.0

0.038 0.021 0-023 0-005 0.012 0.018

Intracrystalline distribution of nickel

347 a

Ni-CHROMITE

NiSO4(H20)

6

pfepeak

KEROLITE

lOeV I

I

I

ENERGY(eV)

8350

8320

O.09

-".

b

Ni-CHROMITE

\

Z 5.0. EXAFS results for the lizardite-nepouite series. The amplitude of the second peak continuously diminishes with decreasing Ni content; in the amesite A1 (5% Ni), this peak is weak and slightly enlarged (Fig. 4). Of structural significance is the stability of this peak position. By comparison with simulations, it is concluded that each Ni atom is surrounded on average by at least 3Ni in lizardites. This trend in the series is also observed on the x(k) functions (Fig. 9). The amplitude of these wave functions decreases with the Ni content but they remain always in phase as in the kerolite-pimelite series, which means that Ni atoms are never surrounded by 5-6 Mg atoms. Structural parameters were determined by the fitting procedure (Table 4) keeping fixed the parameters relative to the tetrahedral sheet for the same reason as for the 2:1 series. By contrast with the former series, increasing Mg content leads to a diminution of the Ni second neighbours, but Ni atoms are nevertheless always heterogeneously distributed in the octahedral sheet (Fig. 8). The total number of octahedral Ni second neighbours found for amesite is 1ow--3-9 instead of ~6. This discrepancy is explained by the distinct structure and chemistry of this mineral compared to those of the kerolite and lizardite series. EXAFS results for other phyllosilicates. EXAFS studies were also performed on a Ni-chlorite (CL) and a Ni-sepiolite (SP). Their Fourier transform (Fig. 4) is close to that of kerolites and the number of Ni second neighbours is > 5 (Fig. 8; Table 4).

INTRACRYSTALLINE

DISTRIBUTION

PATTERNS

OF NICKEL

On account of their similar atomic number, Ni and Fe atoms cannot be distinguished (cf. p. 344) and will be considered together below, EXAFS results show that Ni atoms are never randomly distributed. Two models are possible. (1) If it is assumed that Ni atoms are clustered into pure (Ni,Fe) areas of circular shape and devoid of Mg atoms, it is possible to obtain a mean area diameter using the determined average number of Ni second neighbours is given by EXAFS (see Fig. 10) provided that the number of second Ni neighbours is 5, model (1) leads to Ni clusters larger than 30/k as in the kerolite-pimelite series. In both models, the existence of a nepouite-type local structure must be ruled out because of the dissimilarity of the optical spectra between lizardite and nepouite, which is based on structural differences (Cervelle & Maquet, 1983; Manceau &

358

A. Manceau and G. Calas Mg/Si 9

Lizardite (L4)

0.5.~

Talc Luzenac eK2 K4 K4 K4

0.25-

K4

ols

1.O

K2

N1 1.5

Ni/Si

FIG. 11. XDS analysis of particles.

Calas, 1985; Manceau et al., 1985). Two reasons may be invoked to explain this discrepancy: (i) lizardites are also enriched in Fe atoms (Table 1) which are of different size from Ni and Mg atoms; Ni clustering certainly includes Fe atoms which would inhibit the structural transformation lizardite-nepouite, (ii) the coexistence of domains of nepouite and lizardite within a given layer is not structurally possible. DISCUSSION New data have been obtained on the actual structure of Ni-containing phyllosilicates by the use of the X-ray absorption spectroscopy. High-resolution Ni K-edge spectra confirm the 6-fold coordination of Ni atoms as already shown by optical measurement. EXAFS is shown to be a powerful method for investigation of the intracrystalline distribution of cations within the octahedral sheet of phyllosilicates, as all the other available structural data are taken into account. Indeed, it is necessary to consider carefully structural constraints imposed by the mineral under study to determine without ambiguity the numerous parameters provided by EXAFS. Ni atoms are never randomly distributed but clustered into Ni-rich domains, the average size of which is more dependent on the total (Ni, Fe) content of the bearing phase than on its nature. The behaviour of Ni in these two series may be considered in terms of their geological formation conditions. The main geochemical features leading to Ni ores in New Caledonia may be described briefly as follows (Pelletier, 1983). The most common parent rock is an olivine Fo90-rich harzburgite.

Intracrystalline distribution o f nickel

359

Hypogenic hydration of the ferromagnesian silicates results in serpentinites. Serpentinization of the p r i m a r y silicates is related to direct pseudomorphing o f olivine by serpentine giving rise to a typical mesh-texture. The chemical composition is characteristic of this p r i m a r y serpentine and shows some associated iron oxides (4-8%). The largest amount o f Ni occurs in the olivines (0.15% NiO) and serpentines (0-3% NiO). In the saprolitic level just above the fresh rock, the p r i m a r y serpentine is transformed: it loses Mg and is enriched in Fe and Ni whose concentration m a y increase by 5 - 1 0 % . Inside cracks, the transformed serpentine occurs with neoformed (solution-precipitated products) materials often called 'garnierites'. These latter mainly comprise kerolite and lizardite and contain < 1 % Fe. This lizardite is of s e c o n d a r y origin and m a y be clearly distinguished from the p r i m a r y serpentine by its iron content. In that respect, the Fe-content of the present lizardites confirms their p r i m a r y origin. By contrast, the neoformed origin o f nepouite is apparent from both its chemical analysis and its exceptional crystallinity which is u n c o m m o n for weathering products. F r o m examination o f Fig. 8, it is clear that the neoformed products (kerolite, sepiolite) are more enriched than the transformed products. Furthermore, E X A F S and optical d a t a demonstrate the existence of a N i - F e - l i z a r d i t e mineral whose structure is not nepouite-like. This result lends credence to a structural memory o f the formation mechanism in the 1:1 series: N i - F e TO layers enriched b y a transformtion process conserve their original lizardite structure (inherited structure), neoformed TO products possess a nepouite-like structure. ACKNOWLEDGMENTS The authors are particularly grateful to J. Petiau for helpful discussions and suggestions. We wish to thank also M. Cre~pin, A. Decarreau, M. Lenglet and B. Velde for suppling specimens and the staff of LURE laboratory for synchrotron availability. Sampling of the material from New-Caledonia was facilitated by Drs Escande, Esterle, Pelletier and Troly of the Soci&~ Le Nickel (SLN) who helped also with fruitful discussions. This work was supported by MST/MIR grant, Valorisation des Ressources du Sous-Sol, Contract N o82D-0811. REFERENCES BISHD.L. (1981) Cation ordering in synthetic and natural Ni-Mg olivine. Am. Miner. 66, 770-776. BRINDLEY G.W. & CHIH-CHUN KAO (1984) Structural and IR relations among brucite-like divalent metal hydroxides. Phys. Chem.Miner. 10, 187-191. CALASG. & PE~AU J. (1983a) Coordination of iron in oxide glasses through high-resolution K-edge spectra: informations from the pre-edge. Solid St. Com. 48, 625-629. CALAS G. & PETIAU J. (1983b) Structure of oxide glasses: spectroscopic studies of local order and crystallochemistry. Geochemical implications. Bull. Miner. 106, 33-55. CALAS G., BASSET W.A., PETIAU J., STE~BERG M., TCHOUBARD. & ZARKAA. (1984) Mineralogical applications of synchrotron radiation. Phys. Chem.Miner. 11, 17-36. CERVELLE B. & MAQUETM. (1983) Cristallochimie des lizardites substitu6es Mg-Fe-Ni par spectrom6trie visible et infrarouge proche. Clay Miner. 17, 377-392. GERARDP. & HERamLONA. (1983) Infrared studies of Ni-bearing clay minerals of the kerolite-pimelite series. Clays ClayMiner. 31, 143-151. HAVEST.M. & BOYCEJ.B. (1982) Extended X-Ray Absorption Fine Structure Spectroscopy. Solid St. Phys. 37, 173-351. LEE P.A., CITRINP.H., EISENBERCERP. & KrNCAIDB.M. (1981) Extended X-ray Fine Structure: its strength and limitations as a structural tool. Rev. Modern Phys. 53, 769-806. MANCEAU A., CALASG. & DECARREAUA. (1985) Nickel-bearing clay minerals: I. Optical study of nickel crystal chemistry. ClayMiner. 20, 367-388. MANCEAU A. & CALAS G. (1985) Heterogeneous distribution of nickel in hydrous silicates from New Caledonia ore deposits. Am. Miner. 70, 549-558.

360

A. Manceau and G. Calas

PELLETIER B. (1983) Localisation du nickel dans les minerais 'garni+ritiques' de Nouvelle Cal~donie. Sci. GdoL M~m. 73, 173-183. RAOUX D., PETIAUJ., BONDOT P., CALAS G., FONTAINEA., LAGARDE P., LEVITZ P., LOUPIASG. & SADOC A. (1980) I'EXAFS appliqu6 aux d&erminations structurales de milieux d6sordonn~s. Rev. Phys. Appl. 15, 1079-1094. SANZ J. & STONE W.E.E. (1983) NMR applied to minerals: IV local order in the octahedral sheet of micas: Fe-F avoidance. ClayMiner. 18, 187-192. TEJEDOR-TEJEDOR M.I., ANDERSON M.A. & HERBILLON A.J. (1983) An investigation of the coordination number of Ni 2+ in nickel bearing phyllosilicates using diffuse reflectance spectroscopy. J. Sol. St. Chem. 50, 153-162. TEO B.K. & LEE P.A. (1980) Ab initio calculation of amplitude and phase function for Extended X-Ray Absorption Fine Structure (EXAFS) Spectroscopy. J. Am. Chem. Soc. 101, 2815-2830. WAYCHUNAS G.A., APTED M.J. & BROWN G.E. (1983) X-ray K-edge absorption spectra of Fe minerals and model compounds. Near-edge structure. Phys. Chem. Miner. 10, 1-9.