Bonding-state characterization of constituent elements in phyllosilicate

Jun 1, 1988 - Bonding-state characterization of constituent elements in phyllosilicate minerals by XPS and NMR. A. R. Gonzalez-Elipe, J. P. Espinos, ...
0 downloads 0 Views 761KB Size
3471

J . Phys. Chem. 1988, 92, 3471-3476

Bonding-State Characterization of Constituent Elements in Phyllosilicate Minerals by XPS and NMR A. R. Gonzilez-Elipe,**+J. P. Espin6sJ G. Munuera: J. Sanz,$ and J. M. Serratosa* Institufo de Ciencias de Materiales de Sevilla, Centro mixto CSIC- Universidad de Sevilla, and Departamento de Quimica Inorgcinica, PO Box 1 1 15, c/ Prof. Garcia Gonzcilez s/n. 41071 Sevilla, Spain, and Instituto de Fhco-Quimica Mineral, CSIC, c/ Serrano 117, 28006 Madrid, Spain (Received: April 15, 1987; In Final Form: November 13, 1987)

Two sets of natural phyllosilicates-dioctahedral and trioctahedral-having different degrees of isomorphic substitution (either in the tetrahedral or octahedral sheets) have been studied by X-ray photoelectron spectroscopy (XPS). Si, 0, and A1 ions at the tetrahedral sheets show binding energies (BE) that linearly decrease with increasing excess of negative charge generated by such substitutions at the silicate framework. A1 or Mg ions, at the octahedral sheet, are also subjected to a weaker but systematic effect. A good correlation is observed between the XPS binding-energy shifts for Si and A1 and the previously recorded 29Siand 27AlNMR chemical shifts. A tentative explanation for the differences in the shifts observed in NMR and XPS spectra is given in terms of the hollow-sphere model for shifts in photoelectron binding energies.

Introduction

Although the general structure of silicate minerals is wellknown, there are specific aspects that only recently have been explored by the use of new spectroscopic techniques such as XPS or N M R . Thus, the bonding state of their constituent elements has been systematically examined by XPS by Wagner et al.’ comparing different silicates from isolated S i 0 4 units (nesosilicates) to a three-dimensional Si04 network (tectosilicates). More recently Seyama and Soma,* analyzing a wide number of minerals with different degree of S i 0 4 condensation, have concluded that photoelectron binding energies of Si, 0, and tetrahedrally coordinated AI ions in the silicate framework of these minerals decrease with increasing negative charge on their framework, suggesting that the negative charge should be delocalized over all the silicate tetrahedral structure. By contrast, they found that octahedrally coordinated A1 and M g ions, present in their silicate minerals, are not subject to a strong effect from the negative charge of the framework. On the other hand, Lippmaa et al.334have recently shown that the two predominant features determining isotropic 29SiN M R chemical shifts are (a) the degree of S O 4 anion condensation, which leads to the formation of distinct chemical-shift ranges according to the number of neighboring silicon-oxygen tetrahedra, and (b) the relative content of four-coordinated aluminum in the tetrahedral framework. Effects of other cations and peculiarities in the geometry of the tetrahedral lattice are additional factors that cannot be i g n ~ r e d . ~In our view, phyllosilicates represent an ideal choice for a comparative assessment of both XPS and N M R conclusions owing to the possibility of modifying the negative charge a t the tetrahedral framework by substitution of Si by Al, without any change in the degree of S O 4condensation. Phyllosilicates consist of layers made up by condensation of a central octahedral sheet and two tetrahedral sheets, one on each side. In the tetrahedral sheet, individual S i 0 4 tetrahedral units are linked with three neighboring tetrahedra to form hexagonal sheets (O/Si atomic ratio = 2 . 5 ) . Octahedral compositions conform closely to either a full occupancy of the sheet if cations are M2+ (trioctahedral minerals) or an occupancy of two-thirds of the available positions if cation are M3+ (dioctahedral minerals). Isomorphous replacement of A1 for Si in the tetrahedral sheets or octahedral M2+ or M3+by less-charged cations confer to the full layer a negative charge that is compensated for by interlayer cations. For these minerals Lippmaa et aL5 have found that octahedral and interlayer cation effects on the 29SiN M R chemical shift are small, while mutual adjustments of the parallel tetrahedral and ‘Instituto de Ciencias de Materiales de Sevilla. 3 Instituto de Fkco-Qulmica Mineral. 0022-3654/88/2092-3471$01.50/0

octahedral layers has a significant influence on the 29SiN M R chemical shifts in addition to the effect of four-coordinated aluminum. In this sense, Sanz et aL6 have recently shown, using ”Si and 27Almagic-angle spinning (MAS) N M R spectroscopy, that A1 ions in octahedral and tetrahedral positions in these phyllosilicates have different “chemical shifts” and that Si ions surrounded by a different number of A1 ions in the nearest cation positions in the tetrahedral framework (Le., Si-(SinAI3J, n = 0-3) can be distinguished by their resonance lines. This has allowed to establish, for a given sample, the cation distribution in the tetrahedral sheets, showing that Lowestein’s rule (avoidance of tetrahedral A1-0-A1 linkages) is obeyed’ and the charge is homogeneously distributed. In this paper the photoelectron binding and X-ray-induced Auger kinetic energies (KE) of Si, 0, Al, and M g of a series of phyllosilicates previously studied by Sanz et a1.6 by N M R have been determined, so that the results of XPS could be related with their detailed structural assessment by N M R . Experimental Section

Materials and Methods. Seven phyllosilicates have been used with origins and compositions given in Table I. The samples are well-characterized specimens of trioctahedral and dioctahedral subgroups in which increasing substitution of AI for Si confers to the layer a negative charge per formula unit that goes from 0 in talc and pyrophyllite to 1 in phlogopite and muscovite and -2 in margarite. One exception in this series is hectorite, where charge is generated in the octahedral sheet by replacement of Mg ions by Li ions. In addition, a sample of clintonite, with idealized chemical composition Ca0.99(Si,,~~A1~,~~),(Mg~.3A~0.6~)o010(0H)~, has been used to complete the trioctahedral series of minerals studied in this work (lamellar charge -2). For this sample there are no available N M R data, and no comparison between N M R and XPS can be tried. Except for clintonite, the mineral samples were ground in an agate mortar, and the powder was compacted under low pressure in a sample holder made of aluminum that was placed directly onto the transfer stab of the spectrometer.

-

(1) Wagner, C. D.; Passoja, D. E.; Hillery, H. F.; Kinisky, T. G.; Six, H. A,; Sansen, W. T.; Taylor, J. A. J. Vac. Sci. Technol. 1982, 21, 933. (2) Seyama, H.; Soma, M. J. Chem. SOC.,Faraday Tram. 1 1985,81,485. ( 3 ) Lippmaa, E.; Magi, M.; Samoson, A,; Engelhardt, G.; Grimmer, A. R. J . Am. Chem. SOC.1980, 102, 4889. (4) Lippmaa, E.; Magi, M.; Samoson, A,; Tarmak, M.; Engelhardt, G. J . Am. Chem. SOC.1981, 103, 4992. (5) Magi, M.; Lippmaa, E.; Samoson, A,; Engelhardt, G.; Grimer, A. R. J . Phys. Chem. 1984,88, 1518. ( 6 ) (a) Sanz, J.; Serratosa, J. M. J . Am. Chem. SOC.1984, 106, 4790. (b) Herrero, C. P.; Sanz, J.; Serratosa, J. M. J . Phys. C: Solid State Phys. 1985, 18, 13. (c) Herrero, C. P.; Sanz, J.; Serratosa, J. M. Solid State Commun. 1985, 53, 151. (7) Lowenstein, W. Am. Mineral. 1954, 39, 92.

0 1988 American Chemical Society

3472

The Journal of Physical Chemistry, Vol. 92, No. 12, 1988

Gonzdez-Elipe et al.

TABLE I: Structural Formulas of Phyllosilicates Samples" tetrahedral octahedral mineral Si AI AI Ti Fe3+ Mg Fe2+ pyrophilliteb 4.00 1.92 0.04 0.07 muscovite' 3.16 0.84 1.95 0.03 0.04 margarited 2.11 1.89 1.99 0.03 0.01 talc' 4.00 2.98 0.02 phlogopite P&' 2.84 1.16 0.09 0.06 2.82 0.02 vermiculiteg 2.89 1.11 0.08 0.02 0.06 2.82 hectorite' 4.00 2.65

Li

Em

0.35

2.03 2.02 2.03 3.OO 2.99 2.98 3.00

interlayer elements Na Ca

K 0.79 0.01

0.87 0.03 0.08

E:,,,

0.04 0.19

0.01 0.81

0.84 1.01

0.06

0.03 0.46*

0.96 0.49 0.35

0.27

"All values refer to an 01,,(OH)2formula unit. bNorth Carolina. Orcel, J.; Chaillere, S.; Henin, S. C. R.Hebd. Seances Acad. Sci. 1957, 244, 1383-1386. 'Miyori, Tochigi, Japan. Kodama, H.; Gatineau, L.; Mering, U. J . Clays Clay Mineral. 1971, 19, 405-413. dChester, M. S.; Guggenheim, S.; Bailey, S . W. A m . Mineral. 1975, 60, 1023-1029. 'Tijola, Spain. Alverez-Estrada, D. Talcos espaiioles, Monografia, CSIC: Madrid, 1952. fGreenville an Wakefield, Quebec, Canada. Rousseaux, J. M.; Rouxhet, P. G.; Vielvoye, L. A.; Herbillon, A. Clay Mineral. 1973, 10, 1-16. Ellano, Country, TX; Norrish, K. Proc. Int. Clay Con5 1972 1973, 417-432. "The natural vermiculite contains Mg2+as exchangable cations, but,

for convenience, chemical analysis was done in a calcium-exchanged sample. 'In the case of hectorite the values refer to an O,,(OH)F formula unit; Sand, UT. TABLE II: Elements Composition (atoms percent) of Phyllosilicate Samples from XPS Spectra framework elementsovb mineral 0 F Si AI Mg pyrophyllite 65.3 (66.9) 23.3 (22.3) 11.3 (10.7) muscovite 65.6 (63.8) 16.8 (16.8) 13.8 (14.8) margarite 63.5 (63.2) 11.0 (11.1) 21.5 (20.4) talc 63.4 (63.2) 20.7 (21.1) 15.7 (15.7) phlogopite 62.3 (60.6) 15.9 (14.4) 7.3 (6.3) 10.7 (14.2) vermiculite 65.2 (61.9) 14.2 (14.9) 7.9 (6.1) 12.3 (14.5) hectorite 59.3 (57.8) 5.3 (5.2) 20.9 (21.0) 12.5 (13.9) clintonite 63.3 (60.2) 11.8 (6.8) 13.7 (16.4) 7.2 (11.5)

K

interlayer elements"*b Na Ca

3.0 (4.2)

0.2 (0.2) 0.7 (1.0)

3.2 (4.2)

3.6 (4.4) 1.8 (2.3) 0.9 (1.4) 3.3 (4.9)

0.6 (0.0)

'Values in parentheses corresponding- to data reported in Table I. b O ( l s ) ,F(ls), Si(2p), A1(2p), Mg(ls), K(ls), Na(ls), and Ca(2p) levels have been used for calculations. Photoelectron- and X-ray-induced Auger electron spectra were recorded with a Leybold-Heraeus LHS- 10 instrument equipped with magnesium and aluminum X-ray sources. The electron signals were accumulated on a Hewlett-Packard IOOOE computer to improve the signal-to-noise ratio. Satellite and base-line subtraction, area calculation, and fitting procedures were used to calculate the composition of the samples. For this purpose the empirical sensitivity factors supplied with the instrument (except for Mg) were used. These factors do not vary significantly from those reported in the literature determined for other instruments.* A residual pressure of ca. 5 X lo-' Torr was always attained in the analysis chamber before XPS recording to remove most of the adsorbed water in the sample, so that framework oxygen content could be evaluated properly. Except for the wide-scan spectra, the energy analyzer was set in the pass energy constant mode a t 50 eV. The binding energy (BE) values were referred for all the samples to the C( Is) level taken a t 284.6 eV. In one case (phlogopite), where the C(1s) peak was rather broad so that an uncertainty higher than io.1 eV in BE could be expected, the Ar(2p) peak recorded after Ar+ incorporation by sputtering for ca. 30 s with an argon gun was used as reference.' For this purpose the BE of the Ar(2p) peak referred to the C ( l s ) peak in talc after a similar sputtering was used. *'A1 and *'Si high-resolution M A S N M R spectra of phyllosilicate samples in powder form were recorded on a Brucker CXP 300 high-power spectrometer at 78.2 and 59.6 MHz, respectively? The spinning frequency was in the range 4-4.5 kHz. All measurements were carried out at room temperature with A1(H20),3' and tetramethylsilane (TMS) used as external references. Cross polarization and proton decoupling were not used. Time intervals between succesive accumulations were chosen to avoid saturation effects and were 0.5 s for the 27Al and 4 s for the 29Sisignal. Accumulations amounted to 50 and 500 fid's, respectively. The mean error in the measured isotropic chemical shifts was -0.5 PPm. (8) Wagner, C. D.; Davis, L. E.; Zeller, M. V.; Taylor, J. A,; Raymond, R. H.; Gale, L. H. SIA, Surf. Interface Anal. 1981, 3, 211. (9) Kohiki, S.;Ohmura, T.; Kusao, K. J . Elecrron Spectrosc. Relar. Phenom. 1983, 31, 8 5 .

-> >

Y 0

'250

1003

750

Binding

500

250

0

E n e r g y , (eV1

Figure 1. Wide-scan photoelectron spectrum of Margarite excited by Mg Ka radiation.

Results Chemical Analysis of the Samples by XPS. The determination of the elements present in the minerals was made by recording survey XPS spectra as shown in Figure 1 for margarite. Only those elements (Fe, Ti, etc.) at very low concentration could not be easily detected in such spectra, and no detailed analysis was tried for them by XPS. Due to the surface-sensitive character of the XPS, any discussion about structural properties of the samples is possible only if the composition of the more external layers examined by this technique is representative of the bulk. Table I1 summarizes the compositions for the studied elements calculated (as atoms percent) from the recorded XPS spectra, compared with data given elsewhere6 from wet chemical analysis. For framework elements (0,Si, AI, Mg) both sets of data are in agreement within lo%, except for Mg in most samples, where a sensitivity factor from the literature determined with other instruments was used,8 and for Si in clintonite, where a lower

The Journal of Physical Chemistry, Vol. 92, No. 12, 1988 3413

XPS and N M R of Phyllosilicate Minerals

TABLE 111: BE Auger Kinetic Energy (AKE), and Modified Auger Values of Framework Elements of Phyllosilicate n Sib

mineral pyrophillite muscovite margarite talc phlogopite vermiculite hectorite clintonite

06s) 532.7 531.7 531.2 532.4 531.6 53 1.6 532.1 531.3

AKE"

' a

Si(2p)

AKE'

506.6 507.2 507.9 506.6 507.5 507.7 507.3 508.1

1039.3 1038.9 1039.1 1039.0 1039.1 1039.3 1039.4 1039.4

103.5 102.6 102.0 103.3 102.5 102.6 103.0 102.1

1608.2 1609.3 1609.9 1608.8 1609.4 1609.7 1608.8 1610.1

' a

1711.7 1711.9 1711.9 1712.1 1711.9 1712.3 1711.8 1712.2

75.6 74.5 74.1

1385.8 1386.9 1387.4

1461.4 1461.4 1461.5

73.9 74.2

1387.2 1387.1

1461.1 1461.3

74.1

1387.3

1461.4

1304.6 1304.2 1304.0 1304.3 1303.7

1179.9 1180.3 1180.5 1180.2 1180.8

2484.5 2484.5 2484.5 2484.5 2484.5

Oa,modified Auger parameter, calculated from the KL23L23 Auger peaks in the case of Si, AI, and Mg, and KVV for oxygen. bThe Auger peaks for A1 and Si were recorded by using the Bremstrahlung radiation of Mg Ka and AI Ka sources, respectively. CTheAl(2p) BE values in this table have been measured from the spectra before the fitting with two peaks for octahedral and tetrahedral aluminum ions (see Figure 5 and discussion in text).

5344 104

533-

I

J /

1 I

Pyrophyllite

4

Hectorite

101

102

103

-4

! 2 3 4 5 6 O/Si atomic ratio

-

Figure 3. Relationship of Si(2p) binding energy and O/Si atomic ratio obtained from XPS spectra.

(10) Wagner, C. D. J . Chem. SOC.,Faraday Discuss. 1975, 60,291.

minerals, thus confirming their previous conclusion that structures involving oxygen, silicon, and aluminum have a very low polarizability, contributing with similar extraatomic relaxation energies to the photoemission process. In the case of aluminum they also found that tetrahedral A1 ions have a smaller Auger parameter than do octahedral ones. Data in Table I11 show for this parameter the same trend with slightly lower values for the minerals containing only tetrahedral A1 ions (vermiculite and phlogopite). However, the important point in our case is that the fairly constant values of the Auger parameter for all the sheet atoms allow us to conclude that the observed changes in experimental photoelectron BE values are actually due to initial-state effects and therefore correspond to changes in the lamellar charge throughout the two phyllosilicate series. In most of the phyllosilicates considered here, the substitution of A1 for Si confers the net negative charge to the whole tetrahedral sheet, which can be assessed by using the experimental O/Si atomic ratio. The influence of this replacement on the Si(2p) BE values is clearly shown in Figure 3, where this parameter has been plotted against the experimental O / S i atomic ratios calculated from XPS data in Table 11. A similar plot can be obtained by using values of the lamellar charge, calculated from the mineral formulas in Table I, as shown in Figure 4 for Si(2p) and Al(2p) BE values of Si and AI ions a t the tetrahedral sheet. In the latter case, the BE values have been determined after a fitting of the Al(2p) peak as shown in Figure 5 following the procedure given in the next section. For these representations the relationship for the Si(2p) BE is even better, and clintonite and hectorite fit on the same line in spite of the fact that in the latter the charge is due to M g replacement by Li a t the octahedral sheet and in clintonite replacement of Si by A1 in the tetrahedral sheet is in part compensated by the replacement

3474 The Journal of Physical Chemistry, Vol. 92, No. 12, 1988 0 Si

OAl(2p) tet

(2pl

GonzBlez-Elipe et al. Mqilrl

a

i 75

b

0 All2pl

;;i ' -t

w 761 m

!

w

l m

5 103 w

I 02

101

1 00

'06

Si(2pl B E W -

lintonite

102

\

t

Margarite

I

00

0.5

Lamellar

I5

IO

I 0' t h q 4

Figure 6. Relationship between Al(2p) and Mg(1s) binding energies for octahedral ions and (a) Si(2p) binding energy and (b) lamellar charge (note the slope close to 0.7 in a).

-

a

I

05

Lamellar Charge ( u p

A1 I.:

b

0 A 1
I

1.o 15 20 Charge (unit of charge)

Figure 4. Relationship of Si(2p) and AI(2p) binding energies of tetrahedral ions to the lamellar charge. ID in

--

"(*"'f\

A Binding E n e r g y (eV) Figure 5. Si(2p) and AI(2p) spectra for different samples showing best fitting for octahedral and tetrahedral A1 ions.

of M g by AI in the octahedral sheet. These facts clearly suggest that the Si(2p) (or Al(2p)) and O(1s) BE values depend only on the net charge of the whole silicate layer, whatever its origin. BE Shifts in the Octahedral Sheet. Differences in BE values and Auger parameters for A1 ions with tetrahedral and octahedral environments should be expected due to their distinct Madelung potentials, and in fact these differences have been observed for several minerals.'," Thus it seems of interest to examine if the excess of charge induced by substitution a t the tetrahedral (or octahedral) sheets affects the BE values of A1 or M g ions in octahedral coordination. In our set of samples, phlogopite and vermiculite have A1 ions only in the tetrahedral sheets, and pyrophyllite only in the octahedral sheet, while muscovite and margarite have A1 ions with both types of coordination. A careful measurement of the full widths a t the half-maximum (fwhm) of the Al(2p) peaks for (1 1) Wagner, C. D.; Six,H. A.; Jansen, W. T.; Taylor, J. A. Appl. Surf. Sei. 1981, 9, 203.

132 Sil2p)

~

103 9 E INI-

~ I04

_

i

\

60,PO.

1L

_

15

AIOpI B E ieV'-

Figure 7. (a) Relationship between 29SiNMR chemical shifts and Si(2p) binding energies. (b) Relationship between 2'A1 NMR chemical shifts for octahedral and tetrahedral AI ions and AI(2p) binding energies.

dioctahedral samples (1.6 eV for pyrophillite, 1.9 eV for muscovite, and 2.1 eV for margarite) suggests the presence of more than one component in the recorded signals of muscovite and margarite. As there exist only two coordination types for AI, a fitting of the AI(2p) peaks in muscovite and margarite was carried out with two elemental curves of the same width as that of Al(2p) in pyrophyllite and with an intensity ratio, Alt/A1,, equal to that of the mineralogical formulas of these minerals in Table I. The results of the fittings, shown in Figure 5, allow us to determine the Al(2p) BE values in octahedral and tetrahedral coordinations for both minerals. When Al(2p) BE values for the octahedral ions are plotted against the values of the lamellar charge of these silicates (Figure 6), a good correlation is observed, though in this case the shift of the BE values for the octahedral A1 ions was only ca. 70%of the value for the shift of the tetrahedral cations, Si(2p) (or A1(2p)), as shown in the figure. So, we may conclude that A1 ions in the octahedral sheet are also subjected to the variations on the lamellar charge. Moreover, when Mg(1s) BE values (Table 111) for clintonite and trioctahedral minerals are plotted in the same figure, the effect of the layer charge on the BE values of these magnesium octahedral ions also follows the same trend as the octahedral AI ions. In this figure, the similar representations obtained by plotting Al(2p) and Mg(1s) BE values against either Si(2p) or lamellar charge confirm that Si(2p) BE can be taken as a direct experimental measurement of this lamellar charge in ph yllosilicates. XPS and N M R Shifts. From our XPS study it seems clear that the layer charge created by isomorphous substitutions through the sample series in the tetrahedral or octahedral sheet affects the BE values of all ions in the layered structure in a similar way but with different intensities. As N M R is also sensitive to the density of charge on the ions, a correlation should be expected between N M R chemical shifts and BE values as previously observed for discrete molecules.'* However, since Si(2p) peaks, contrary to 29SiN M R peaks, do not show any fine structure due (12) (13)

Lindberg, B. J. J. Electron Spectrosc. Relat. Phenom. 1974, 5 , 149. Barron, P. F.;Frost, R. L. Am. Mineral. 1985, 70, 7 5 8 .

XPS and N M R of Phyllosilicate Minerals

The Journal of Physical Chemistry, Vol. 92, No. 12, 1988 3475

differences in electronegativity between the two atoms involved), thus being a direct measurement of bond covalency.18 For the phyllosilicates studied in this work, Figure 8 shows Si(2p) BE values, taken as a measure of the lamellar charge, plotted against both the mean Si(A1)-0 bond lengths, obtained from X-ray diffraction data,I9 and Si(A1)-0 bond strengths, calculated by using the universal parameters of Brown and Shannon.I8 As can be seen, good linear, though opposite, correlations are obtained in both cases, indicating the existence in the tetrahedral framework of an inverse correspondence between bond length and covalency as the lamellar charge increases. This observation agrees with our XPS results, which show similar BE shifts of Si, Al, and 0 ions in the tetrahedral sheets, suggesting that excess of charge is delocalized over all the silicate tetrahedral Figure 8. Relationship of Si(2p) binding energies and (a) mean Si-0 framework and distributed in the sheets between all the ions. This bond lengths and (b) Si-0 bond strengths (covalency). excess of negative charge would induce a net electron repulsion to atoms in different environments, the measured Si(2p) peaks between ions, leading to framework expansion and therefore to must represent in each case a mean value of all the Si ions in each a lower overlapping (Le., loss of covalency). Lippmaa et aL5 draw sample. Owing to this fact, the most consistent N M R value with a similar conclusion based on low-field 29SiN M R chemical shifts which to make the comparison would be a mean chemical shift induced by progressive substitution of SiO, tetrahedra by A104 (isotropic shift) that can be calculated from the original 29SiN M R tetrahedra into a two-dimensional framework of phyllosilicates, spectra after giving a specific weight to each resonance frequency, where, provided that lattice geometry and cation effects are not according to their intensities in the spectra. The corresponding prevailing, the decrease of covalent character in A1-0 bonds is plot in Figure 7 shows a good correlation between the calculated parallel to the magnitude of the shift. mean N M R chemical shift and the corresponding Si(2p) BE On the other hand, in the octahedral sheet the variation with -,yalues for the two sets of phyllosilicates studied here, thus conthe charge of the BE values of A1 and Mg is smaller than in the firming our previous assumption. The same trend can also be tetrahedral sheet. This behavior can be explained in terms of the observed when the values of the chemical shift corresponding to hollow-sphere model for BE shifts.20 According to this model, each particular environment of Si (e.g., Si surrounded by three the actual binding energy of a given ion, E,, in respect to an energy Si atoms) are compared in samples with a different ~ h a r g e . ~ . ' ~ reference (E,,,) is a function of the effective charge on the ion (qJ A similar correlation is also found for A1 in tetrahedral and and of the charge on the neighboring ions (4,) separated from it octahedral coordination. Also in Figure 7, it can be observed that by a distance rIj(Madelung or ionic potential) according to the slope of the straight line defined by the points corresponding too the tetrahedral A1 ions is close to that of Si ions, while the I slope for the set of octahedral AI ions is much smaller (notice the From X-ray data of different minerals Sachaki et aL2' have change of scale in the figure). Though the values of the chemical derived the covalency of different M-0 bonds, showing that this shifts in this case should be corrected for quadruplar second-order bond is less covalent for octahedral aluminum than for tetrahedral interactions, we have found that such corrections for tetrahedral silicon. Therefore, in the series of phyllosilicates studied here the A1 are very similar in all our samples and therefore would not ionicity of A1-0 and Mg-0 bonds in the octahedral sheet is affect the correlation between XPS and N M R results. However, expected to be greater than for the M - 0 bonds in the tetrahedral in the case of octahedral Al, such corrections are less reliable, sheet (M = Si and Al), so that the negative charge will be mainly which together with the fact that the N M R shifts are much less localized in the oxygens and the other ions of the tetrahedral sheet. sensitive than those of XPS gives a higher uncertainty for these Thus, the contribution on the second term in (1) will be less data. important for octahedral cations, and changes in the Madelung Discussion potential (third term in (1)) due to neighbor oxygen ions will be the predominant factor that affects the BE values of inner Al(2p) The previous considerations on the effect of the excess of charge and Mg(1s) levels measured by XPS for the different phyllogenerated on the phyllosilicate framework by ion replacements silicates. However, in this case the N M R chemical shifts of the have a phenomenological character. However, they must have octahedral 27Al,mainly associated with the electron density in a correspondence at the scale of particular bonds (bond length, the own cations, should be little affected by the extra charge covalency, etc.) in the silicate framework. It could therefore be located in the oxygens, which explains the small variation in this interesting to compare the experimental magnitudes (XPS BE parameter in Figure 7 . values and mean N M R chemical shifts) with such parameters. Comparisons of this type have been previously made in an early Conclusions work by Urch and Murphy,IS who observed a correspondence Photoelectron binding energies of Si, 0, and tetrahedrally between the Si(2p) BE values and Si-0 bond length for a series coordinated A1 ions in the phyllosilicate framework decrease with of silicates, while Smith et a1.,I6using isotropic 29SiN M R chemical increasing excess of negative charge generated by replacement shifts, conclude that a better fit could be obtained with the average of ions in either the tetrahedral or octahedral sheets. In addition, bond strength as the structural parameter. This parameter, first A1 or Mg ions at the octahedral sheet are also subjected to a defined by Pauling17 as the valence of the cation divided by its slightly weaker but systematic effect of the negative charge. A coordination number and modified by Brown and Shannon18 to good correlation is observed between the XPS binding energy shifts introduce in its calculation the bond length, shows a good linear for Si and A1 and the corresponding N M R chemical shifts correlation with the covalent character of bonds (calculated from measured for the same set of samples and of these two parameters (14) Weiss, Ch. A.; Altaner, S. P.; Kirkpatrick, R. J., private communication. (15) Urch, D. S.; Murphy, S. J . Electron Spectrosc. Relat. Phenom. 1974, 5, 167. (16) Smith, K. A.; Kirkpatrick, R. J.; Oldfield, E.; Henderson, D. M. Am. Mineral. 1983, 68, 1206. (17) Pauline, L. In The Nature o f t h e Chemical Bond, 1st ed.; Cornel1 University Press: Ithaca, NY, 1939. (18) Brown, I. D.; Shannon, R. D. Acta Crystallogr., Sect. A 1973, ,429, 266.

(19) (a) Lee, J. H.; Guggenheim, S. Am. Mineral. 1981, 66, 350. (b) Rothbauer, R. Neues Jahrb. Mineral. Abh. 1971, 143. (c) Guggenheim, S.; Bailey, S. W. A m . Mineral. 1978, 63, 186. (d) Perdikatsis, B.; Burzlaff, H. 2.Kristallallogr. 1981, 156, 177. (e) Joswig, W. Neues Jahrb. Mineral., Abh. 1972, 1. (0Shirozu, H.; Bailey, S. W. A m . Mineral. 1966, 51, 1124. (20) Carlson, T. A. In Photoelectron and Auger Spectroscopy; Plenum: New York, 1975; p 168. (21) Sachaki, S.; Fujino, K.; Takenshi, Y.; Sadanaga, R. Acta Crystallogr., Sect. A 1980, ,436, 904.

J . Phys. Chem. 1988, 92, 3476-3483

3476

and the bond lengths and strengths of the Si(A1)-0 bonds at the tetrahedral sheets. The trends observed in the BE values of both octahedral and tetrahedral cations can be explained in terms of the hollow-sphere model in which variation of the effective charge of the ions, as a consequence of the lamellar charge, is taken into account.

Acknowledgment. W e thank C A I C Y T for financial support and dedicate this work to Prof. F. GonzBlez-Garcia on his 70th birthday. Registry No. Pyrophyllite, 12269-78-2; muscovite, 1318-94-1; margarite, 1318-86-1; talc, 14807-96-6; phylogopite, 61076-94-6; vermiculite, 12251-58-0; hectorite, 12173-47-6.

Photoelectrochemistry of Cadmium Sulfide. 1. Reanalysis of Photocorrosion and Flat-Band Potential Dieter Meissner,* Rundiger Memming, ISFH, Institut fur Solarenergieforschung, Hannouer. Sokelantstrasse 5, 0-3000 Hannover 1, FRG

and Bertel Kastening Institut fur Physikalische Chemie der Universitat, Bundesstrasse 45, 0-2000 Hamburg 13, FRG (Received: April 30, 1987; I n Final Form: September 30, 1987)

The photoelectrochemical behavior of cadmium sulfide depends strongly on the pretreatment of the electrodes. Sulfur, the main photocorrosion product of CdS, changes the potential drop across the Helmholtz double layer and shifts the flat-band potential of CdS in the anodic direction. Removal of sulfur from the surface is possible by prepolarizing the electrode at about -1.1 V(SCE) in the presence of oxygen. The flat-band potential of the clean (0001) Cd surface has been determined to about -1.8 V(SCE) in the dark. Illumination of the electrode leads to surface-state charging due to formation of S'-, the intermediate in the formation of both possible photooxidation products: So in the absence of oxygen and SO-; in the presence of oxygen.

Introduction Cadmium sulfide is one of the best characterized semiconductors for use in solar-energy conversion systems. Far more than 500 papers concerning the photophysical and photochemical behavior of this material have been published. One main reason for the interest in this material is the position of the band edges in aqueous solution, which is believed to be more suitable than that of nearly all other known semiconductors for the direct splitting of water into hydrogen and oxygen by solar energy,' which is the main goal of many photochemists and photoelectrochemists.2 According to the literature the flat-band potential of CdS in aqueous solution seems to be ideal for the water-splitting reaction. T h e position of the conduction band a t the surface of C d S is located a t sufficiently cathodic potential to reduce water if a catalyst is used to reduce the overpotential of this reaction. Hydrogen evolution from C d S is easy to observe, as has been proven many times.3 Furthermore, the reported flat-band position of CdS means that there should be a sufficient overpotential for holes in the'valence band to oxidize water. Oxygen production from C d S has indeed been reported a few timesS4 However, the (1) Memming, R. Electrochim. Acta 1980, 25, 77. (2) Porter, G. Nature (London) 1980, 288, 320. (3) (a) Nozik, A. J. Appl. Phys. Lett. 1977, 30, 567. (b) Frank, S. N.; Bard, A. J. J . Phys. Chem. 1977, 81, 1484. (c) Darwent, J.; Porter, J. J . Chem. SOC.,Chem. Commun. 1981, 145. Darwent, J. J. Chem. Soc.,Faraday Trans. 2 1981, 77, 1703. (d) Harbour, J. R.; Wolkow, R.; Hair, M. L. J. Phys. Chem. 1981, 85, 4026. (e) Frank, A. J.; Honda, K. J. Phys. Chem. 1982, 86, 1933. ( f ) Borgarello, E.; Kalyanasundaram, K.; Gratzel, M.; Pelizetti, E. Helv. Chim. Acta 1982,65, 243. Houlding, V.;Geiger, T.; Kolle, U.; Gratzel, M. J. Chem. SOC.,Chem. Commun. 1982, 681. (9) Hubesch, B.; Mahieu, B. Inorg. Chim. Acra 1982, 654, L65. (h) Aspnes, D. E.; Heller, A. J . Phys. Chem. 1983,87,4919. (i) Matsumura, M.; Saho, Y.; Tsubomura, H. J . Phys. Chem. 1983,87, 3807. Matsumura, M.; Hiramoto, M.; Iehara, T.; Tsubomura, H. J . Phys. Chem. 1984, 87, 4919. G) Rajh, T.; Micic, 0. I. Glas. Hem. Drus. Beograd 1983.48, 335. (k) Mau, A. W.-H.; Huang, C. B.; Kakuta, N.; Bard, A. J.; Campion, A,; Fox, M. A.; White, .IM.; . Webber, S . E. J . A m . Chem. SOC.1984, 106, 6537. Matsumoto, M.; Hiramoto, M.; Iehara, T.; Tsubomura, H. J . Phys. Chem. 1984, 88, 248. (1) Biihler, N.; Meier, K.; Reber, J. F. J . Phys. Chem. 1984, 88, 3261.

0022-365418812092-3476$01.50/0

stoichiometric production of hydrogen and oxygen from catalyst-loaded C d S reported by Gratzel and c o - w o r k e r ~was ~ ~surprising, especially because this meant a complete suppression of the CdS photocorrosion. Our own attempts to split water using catalyst-loaded particle^,^ monograin membranes,6 or electrodes' failed completely: no oxygen was evolved. Depositing catalysts on cadmium sulfide exclusively caused a n increase of photocorrosion by accelerating the rate of hydrogen production.* However, in several experiments the amount of Cd2+ found in the electrolyte after the experiment remained less than that calculated by assuming the well-known photocorrosion reaction CdS

+ 2H20

light

H,

+ Cd2++ S + 20H-

In addition the p H increase expected for this reaction was never achieved when air was allowed to get into the cell. These dis(4) (a) Kalyanasundaram, K.; Borgarello, E.; Gratzel, M. Helu. Chim. Acta 1981, 64, 362. Kalyanasundaram, K.; Borgarello, E.; Duonghong, D.; Gratzel, M. Angew. Chem. 1981, 93, 1012. (b) Frank, A. J.; Honda, K. J . Phys. Chem. 1982, 86, 1933. (c) Thewissen, D. H. M. W.; EeuwhorstReinten, M.; Timmer, K.; Tinnemans, A. H. A,; Mackor, A. Int. ConJ Solar Energy Convers. Storage. [Proc. In?. Con/.], 4 t h 1982, 261. (d) Khan, M. M. T.; Bhardwaj, R. C., Iadhev, C. M. J . Chem. Soc., Chem. Comml*n.1985, 1690. (5) (a) Meissner, D.; Kastening, B.; Memming, R. Photochem. Conuers. Storage Sol. Energy, [Proc. In?. Conf.],4th 1982. (b) Meissner, D.; Kastening, B.; Memming, R. Sol. Energy R&D Eur. Community, u2 1983, 110. (c) Meissner, D.; Memming, R.; Li, S . ; Jesodharan, S.; Gratzel, M. Eer. Bunsen-Ges. Phys. Chem. 1985, 89, 301. (6) Meissner, D.; Kastening, B.; Memming, R. Chem. Phys. Lett. 1983, 96, 34. (7) (a) Meissner, D.; Memming, R.; Kastening, B. 35th Meeting of the International Society of Electrochemistry, Berkeley CA, 1984. (b) Meissner, D.; Memming, R.; Kastening, B. Chem. Phys. Lett. 1986, 127, 419. (c) Meissner, D.; Benndorf, C.; Memming, R. Appl. Surf. Sci. 1987, 27, 423. (8) McEvoy, A. J.; Meissner, D.; Etman, M.; Memming, R.; Kastening, B. Report EUR 10108, Commission of the European Communities, Brussels, 1985. (b) Meissner, D. Diplomarbeit; Fachbereich Chemie der Universitat, Hamburg, 1983.

0 1988 American Chemical Society