9385
J . Phys. Chem. 1991, 95, 9385-9391 parameters but also the specific kinetic parameters for each elementary step measured on the same system. This is not the case we analyzed here, and then we can say only that the parameters determined through the model are physically meaningful only to the extent in which their values are in the range of the observed values for the same elementary steps in similar but different systems (for.example, as we said before, V H H , VH,, and VH, are in the range of H-H interactions on perfect metal surfaces observed by LEED). It is hoped that our analysis may help to stimulate experimentalists to design experiments suitable for the better understanding of the elementary microscopic processes together with the global effects.
Appendix Fundamentals and applications of the Monte Carlo simulation method are exhaustively treated in ref 18, and its use in the analysis of thermal desorption spectra for surface processes is discussed in ref 7-9. Here we describe in a concise form the algorithms for the hydriding and dehydriding processes that may help the reader to write his own computer program. The particle is simulated by a cubic lattice of L X L X L = N sites. L can be fixed according to the capacity and speed of the computer, but with L = 30 good statistics can be obtained and the program could run on a PC. Each site may be an a, a*, B, H,, or H, site. Hydriding Process. All N sites in the lattice are initially a-sites. For a given value of nu we have the following sequence of steps: 1, Choose a site i at random in the lattice (by generating three random numbers x, y , z , each uniformly distributed in the interval (I&). 2. If i is a @-site, return to step 1; otherwise proceed to 3. 3. If i belongs to the external surface, then proceed to 4; otherwise go to step 5. 4. Choose a random number [ uniformly distributed in the interval (0,l) and compare it with Ph;, given by eq 2. If [ < Pb,, site i becomes a &site; otherwise it does not. Go to step 6. 5. Test for the neighborhood of i : (a) If i has no @-sitesin its nearest neighborhood (it does not belong to the a-@ interface), return to step 1. (b) If i belongs to the a-B interface, choose a ( I 8) Binder, K.;Heermann, D. W. Monte Carlo Simulation in Statistical Physics; Springer-Verlag: New York, 1988.
random number [, If [ < P;.,, change site i to a O-site, otherwise do not. Proceed to 6. 6. If the fraction of @-sites is less than nu, repeat from step 1.
Dehydriding Process. For each n, the temperature is increased from an initial value To at constant heating rate v by intervals AT. This temperature interval is chosen in such a way that probabilities Pl,, = Pb.AT/u and P!, = PdAT/uare much smaller than 1. Now in the initial state we have (Nn,) @-sitesand [N(1 - nu)] a-sites. The sequence of steps is the following: 1. Set T = To. 2 . Choose a site i at random. 3. If i is an a-site, return to step 1. 4. If i is a @-site,then: (a) If the @-sitebelongs to the a-@ interface in the bulk, then get a random number 5. If [ < P;,., transform the @-siteinto an a-site and sum 1 to a counter of H atoms (H accumulator); otherwise go to 6. (b) If the @-sitebelongs to the a-@ contour on the external surface, then get a random number [, If $, < P?,., transform the 0-site into an H, site; otherwise go to 6. (c) If the @-sitedoes not belong to the a-8 interface at all, to to 6. 5. The selected site is either an H, or an H, site located at the external surface. It is tested for neighbors: (a) If i has a nearest-neighbor s i t e j that is also an H, or an H, site, then get a random number [. If [ < Pb,then i and j are "desorbed" and converted into a-sites or @-sites(depending on being H, or H, sites), otherwise go to 6. (b) If i has no nearest-neighbor H, or H, sites then proceed to 6. 6. If the number of sites tested is less than N , then go to step 2; otherwise proceed to 7. 7. Take as many H atoms as possible from the H accumulator and distribute them on surface sites that were initially &sites, converting them into H, or H, sites. 8. Increase the temperature in AT and go to step 2. The entire sequence is repeated until all Nn, sites have been converted into a-sites. Acknowledgment. This research was made possible by financial support from CONICET and Fundacion Antorchas. The donors of the Petroleum Research Fund, administered by the American Chemical Society, supported the publication of this research. Registry No. Hydrogen, 1333-74-0; magnesium hydride, 7693-27-8.
Defect Centers Induced by Evaporation of Alkali and Alkaline-Earth Metals on Magnesium Oxide: An EPR Study Elio Ciamello,* Andrea Ferrero, Salvatore Coluccia, and Adriano Zecchina Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali, Uniuersitb di Torino, Via P. Giuria 7, 10125 Torino, Italy (Received: October 22, 1990)
Defect centers created at the surface of polycrystalline, high-area magnesium oxide by evaporation of lithium and magnesium have been investigated by EPR. Both paramagnetic F: and diamagnetic F, centers, respectively consisting of one and two electrons trapped in surface ionic vacancies, are formed. The first type of center is preferentially formed with lithium, whereas the second prevails with magnesium. A fraction of the surface sites acting as surface electron traps is anion vacancies existing at the surface of activated MgO and a second fraction consists of "reconstructed sites" involving cations derived from the metal ionization.
Introduction The color centers in MgO and other alkaline earth oxides have been widely studied in the past. In particular various types of F centers, all consisting of one or more electrons trapped in anionic *Correspondence concerning this paper should be addressed
to
this author.
vacancies, have been identified in single crystals.',* Among the various experimental techniques employed, a predominant role (1) Wertz, J. E.; Orton, J. W.; Auzins, P. Discuss. Faraday Soc. 1961, 31,
140. (2) Henderson,
B.;Wertz, J. E. Ado. Phys. 1968, 17, 749 and references
therein.
0022-3654/91/2095-9385$02.50/0 0 1991 American Chemical Society
9386 The Journal of Physical Chemistry, Vol. 95, No. 23, 1991
has been played by ultraviolet-visible spectroscopy and, in the case of paramagnetic centers, by EPR. F centers in polycrystalline finely divided alkaline earth oxides were also actively investigated, mainly with the aim of clarifying the mechanism of formation and annihilation of the centers and their role in the surface reactivity of the oxides3+ or in other phenomena such as the luminescence of the Two basic methods are adopted to generate F-defective centers in MgO: (i) y-rays or neutron irradiation of the solid. In the former case electrons excited from the valence band are thought to be trapped in anion vacancies already present in the lattice before irradiation, whereas in the second case additional vacancies are formed upon neutron impact with the solid.'J (ii) Addition of a metal excess. This method consists in treating the oxide with vapors of metals having low ionization energy (typically alkali and alkaline-earth metals). The excess cations introduced in the system by ionization of the metal are balanced by the electrons removed from the metal itself, which are trapped in anionic vacancies, Le., in the form of F centers. This method was first adopted, in the case of MgO monocrystals, by Weber2 and then by Chen et al.9 For the polycrystalline solids, the interaction of vapors of sodium and other alkali metals with Na+/Y zeolites leads to the ionization of the metal atoms, and the released electrons are trapped in the sodalite cage by forming with preexisting Na+ cations ionic Na43+ clusters.I0-l2 Systems prepared by evaporation of alkali metals on polycrystalline MgO were indeed studied, in the past, by Malinowski and co-worker~.'~In that case the structural investigation of the metal-oxide interaction was not the principal aim of the work but the method was adopted to prepare the so-called superbasic catalysts active in some catalytic reactions such as isomerization and dehydrogenation of hydrocarbons. The reactivity of the system prepared by evaporating metallic sodium or lithium is clearly different from that of the pure oxide. However, despite a great deal of work being done on the systems, detailed spectroscopic characterization was not carried out. More recently, in our laboratory, systems obtained by contact between MgO and metallic vapors of magnesium or lithium have been reinvestigated with the aim of studying the basic features of the metal-oxide interaction. Particular attention was paid to (i) the spectroscopic characterization of the systems; (ii) a comparison of the systems obtained by this procedure with those prepared by y-irradiation and neutron irradiation of the oxides; and (iii) the investigation of the surface reactivity and, in particular, of the electron donor properties toward small simple molecules. A preliminary paper from our groupI4 reported that evaporation of metallic magnesium on high-area MgO leads to the appearance of a UV-visible spectrum and an EPR signal. The parallel behavior upon various treatments exhibited by the spectra allowed the assignment of both spectra to the same type of center, namely, surface F centers (F,+) formed by ionization of the metal and
(3) Nelson, R. L.: Tench, A. J.; Harmsworth, B. J. Trans. Faraday SOC.
1967. 63. 1427.
(4) Nelson, R. L.; Tench, A. J. Trans. Faraday SOC.1967, 63, 3039. (5) Tench, A . J.: Nelson, R. L. J. Colloid Interface Sci. 1968, 26, 364. (6) Lunsford, J. H.; Jayne, J. P. J . Phys. Chem. 1965, 69, 2182. (7) Coluccia. S.; Tench, A. J.; Segall, R. L. J . Chem. Soc., Faraday Trans. I 1979, 75, 1769. (8) Shvets, V. A.; Kuznetzov, A. V.; Fenin, V. A,; Kazansky, V. B. J . Chem. SOC.,Faraday Trans. I 1985, 81, 2913. (9) Chen, Y.; Sibley, W. A,: Srigley, F. P.; Weeks, R. A.; Hensley, E. B.; Kroes, R. L. J. Phys. Chem. Solids 1968, 29, 863. (IO) Kasai, P. H. J. Chem. Phys. 1965. 43, 3322. (11) Kasai. P. H.: Bishop, R. J. J . Phys. Chem. 1973, 77, 2308. (12) Edwards, P. P.; Harrison, M. R.; Klinowsky, J.; Thomas, J. M.; Johnson, D. C.; Page, C J. J . Chem. Soc., Chem. Commun. 1984, 982. (13) Kijenski, J.; Malinowski, S . J . Chem. Soc., Faraday Trans. I 1978, 74, 230. (14) Zecchina, A.; Scarano, D.; Marchese, L.; Coluccia, S.; Giamello, E. Surf Sci. 1988, 194, 513.
Giamello et al.
Figure 1. X-band EPR spectrum a t IO-mW microwave power of the system Mg/MgO ( T = 298 K, the mark "ref." indicates the resonance of the standard at g = 2.0028).
electron trapping at the surface. This paper deals with the analysis of the EPR spectra obtained when magnesium or lithium are evaporated on polycrystalline MgO. Despite the intrinsic complexity of the powder spectra (with more than one species resonating in a small range of magnetic field), the EPR technique seems to be the most suitable one for the analysis of the structure of the paramagnetic centers formed on the solid. A quantitative estimation of the number of centers and a qualitative description of the whole process occurring at the surface by metal-oxide interaction are also allowed by the EPR technique. The description of the surface reactivity of the Mg/MgO and Li/MgO systems will be reported in forthcoming contributions.
Experimental Section Two types of MgO samples were employed in this research, commercial MgO containing traces of Mn2+ and a high grade purity MgO (Johnson and Matthey). Both samples were previously hydroxylated to Mg(OH)2 by overnight treatment in water at 350 K. In both cases Mg(OH)2 was introduced in a quartz cell also containing the EPR tube and slowly decomposed under vacuum at 523 K to obtain a high surface area oxide. The oxide was then activated under vacuum at 1073 K to remove adsorbed impurities and to reach a controlled degree of dehydroxylation. After this product was cooled at room temperature a small ribbon of metallic Mg or Li (kept in a side chamber of the cell during the preparation and activation of the oxide) was brought in contact with the activated oxide. The cell was then heated under vacuum at 850-870 K in the case of Mg and 570-590 K in the case of Li. At this temperature vaporization of the metal occurred and the oxide sample turned from white to deep blue-violet. Finally the powder was transferred, always under high vacuum, into the silica tube for EPR measurements, keeping the residual metal out of the EPR tube in order to avoid cavity perturbations and consequent decrease of the cavity Q factor. The two systems under study will be hereafter indicated by the labels Mg/MgO and Li/MgO respectively. X-band EPR spectra were recorded on a Varian E-109 machine, equipped with a dual cavity, usually at room temperature since at 77 K the high concentration of trapped electrons causes disturbance around the g, value. Furthermore saturation of the signals of the trapped electrons occurred, at 77 K, at very low microwave power. The EPR machine was connected to a CS-EPR Stelar data station, allowing spectra recording and a careful double integration for spin counting. The standard sample for g-value calibration was Varian pitch (g = 2.0028). Standard samples for spin counting were DPPH (2,2-diphenyl-l-picrylhydrazil)solutions
Defect Centers Induced by Evaporation
The Journal of Physical Chemistry, Vol. 95, No. 23, 1991 9387
Figure 3. Saturation curves (intensity versus square root of the microwave power) for the four main lines of the EPR spectrum of Mg/MgO.
Figure 2. EPR spectra of the system Mg/MgO at three different microwave power levels (a, 0.1 mW; b, 10 mW; c, 50 mW). The weak line observed at high field in spectra b and c is one of the six due to Mn2+, present as an impurity in MgO.
in diethyl ether, prepared in the range between 1015and 10l8of total spin number. Results and Discussion EPR Spectra of Mg/MgO: The High-Field Components. The MgO samples doped with Mg vapors exhibit an intense asymmetric EPR spectrum in the region of the free-electron g value. The reproducibility of the EPR spectra recorded in different experiments is fairly good, even though small differences are observed depending on which of the two types of magnesium oxide was employed. As the differences observed are essentially due to small variations of the relative intensity of signals related to different species, this topic will not be discussed further. A typical spectrum of Mg/MgO recorded at room temperature and IO-mW microwave power is reported in Figure 1, where the main components are labeled with letters A-D. The spectrum in Figure 1 is clearly due to more than one paramagnetic center. This is deduced on the basis of the variation of the spectral shape as a function of the microwave power, which is illustrated in Figure 2, where the spectra recorded at 0.1, 10, and 50 mW are compared. The detailed analysis of the saturation behavior of the various components (intensity versus square root of the microwave power) is also given in Figure 3. The components A and B (g = 2.019 and g = 2.01 3, respectively) show a parallel, quasi-linear trend and, at 50" microwave power, are still unsaturated. The components C and D reach the maximum intensity at about 1 mW and, a t higher powers, exhibit behavior typical of inhomogeneous saturation. Spectra recorded at low microwave power (0.1 mW, Figure 2) are thus the most suitable for characterization of the signal due to C and D components as, under these conditions, interference from A and B signals is avoided. The signal observed at low microwave power (Figure 2a) can be easily assigned to single electrons trapped in anion vacancies located at the surface, usually called F,+ centers, on the basis of the similarity with the spectra of the same centers obtained by y-irradiation3 and also with their saturation behavior. On the other hand, the formation of F, centers by evaporation of alkali
r 7
1.9992
Figure 4. Expanded view of the EPR spectrum of Mg/MgO at 0.1-mW
microwave power.
and alkaline-earth metals is well-known2J3 and is confirmed, in the present case, by the color of the system after evaporation and by its UV-visible spectrum.I4 The surface location of the centers in particular is confirmed by the instantaneous reactivity of the solid with a variety of molecules admitted into the gas phase." The expanded spectrum of the centers, recorded a t 0.1-mW microwave power and reported in Figure 4, outlines the complexity of the pattern, which exhibits an axial (or quasi-axial) signal with g,,= 2.0020 and g, = 2.0005 and an additional line a t higher fields labeled E. We assign this signal to the:F centers since its g values are very close to those given by Tench and co-workers, which are g,,= 2.0016 and g, = 2.0003, re~pectively.~ However, the particular shape of the signal (Le., an unexpected crossing of the baseline between the parallel and perpendicular feature) indicates the possible presence of a second signal buried in the D component and due to a minor species whose nature is not easy deducible on the basis of the data collected up to now. The E line falls in the same position as the high-field component of a doublet observed by TenchSand assigned to the F:(H) center (previously S,) constructed by a trapped electron interacting with the proton of a nearby surface hydroxyl. In the present case the low-field component is not observed (Figure 4); we think, however, that this is due to the fact that the number of the F,+(H) centers is, in our samples, rather low, as the surface has been thoroughly dehydrated at 1073 K before the evaporation of the metal. In
9388 The Journal of Physical Chemistry, Vol. 95, No. 23, 1991 the spectrum reported by Tench, the low-field component, though overlapped with the F,+ line-, was clearly visible due to the high density of F,+( H) centers obtained by y-irradiation under a hydrogen atmo~phere.~ Some attempts to confirm the assignment of the E line to the F,+ centers coupled with a proton were performed: in particular the use of less dehydroxylated oxides, obtained either by activation at lower temperature (773 K) or by water adsorption onto a normally prepared Mg/ MgO sample, gave unsuccessful results since in the former case paramagnetic F,+ centers were almost not produced, whereas in the latter one water reacted with the surface-trapped electrons, destroying completely the paramagnetic centers. A significant result was instead obtained by adsorption of hydrogen onto a Mg/MgO sample: in fact a weak but unambiguous increase of the E component was observed after H2dosage, thus confirming the assignment of the E signal to F,+(H) centers at the surface of our systems. This result is probably due to the formation of a small amount of surface hydroxyls via heterolytic dissociation of hydrogen, which is known to occur on some very reactive surface sites of Mg0.15 In conclusion, the E line can be definitely assigned to the F,+(H) centers (spectral parameters estimated from the spectrum: g = 1.9992and A = 2 G). EPR Spectra of Mg/MgO: The Low-Field Components. The assignment of the low-field components of the spectrum of Mg/MgO (lines A and B) is rather cumbersome. First of all the idea that A and B could be the external lines of a hy rfine sextet due to the interaction of an unpaired electron with a FMg nucleus (I = 5/2, the only nonzero nuclear spin present in the system) must be excluded. In fact 25Mgrepresents about 10% of the total magnesium atoms: the hyperfine structure (if any) would be thus accompanied by an intense line at the center of the structure (ca. g = 2.0040)due to the 24Mg(I = 0) nuclei that, in effect, is not observed. Furthermore the high-field external line of the sextet, which, in principle, would not be overlapped with the F,+ signal, is not observed either. It has also to be noticed that, when F2’s are generated by y-irradiation or neutron irradiation in the absence of hydrogen, no hyperfine 2sMg structure is ~ i s i b l e . ~ Some alternative assignments of the signals can be advanced on a purely speculative basis although, at present, no one seems to be clearly prevalent. The first one involves the presence of two or more monoelectronic F,+ centers in mutual interaction. Indeed, pairs of monoelectronic centers in a singlet ground state were observed in alkali halides (F2centers, ref 16) and in calcium oxide: while, to our knowledge, no information is available on arrays of monoelectronic centers composed of more than two centers in alkaline-earth oxides. In the case of Mg/MgO the ionization of the two electrons of the metal could promote the formation of grouped vacancies, allowing interactions among trapped electrons. Of course the same would not apply in the case of lithium ionization (one electron): it is a matter of fact that the low-field spectral features are not observed in this case. A second hypothesis involves the formation of low nuclearity paramagnetic metal clusters (MnX+),because of the incomplete ionization of the metal itself. These systems, however, are expected to give rise to symmetric EPR signals without appreciable g tensor anisotropy,’*I2 scarcely resembling the signals in discussion. Also the spectra of small metallic particles” are usually symmetric, unstructured, and wider than our signal. Furthermore, although the hypothesis of an incompleted ionization of magnesium due to the high energy required for the whole process (737 and 1451 kJ/mol for the first and second ionization, respectively) cannot, in principle, be neglected, the spectral features in discussion are not related to Mg+ ions, whose spectral behavior is very different.I8 (1 5) Coluccia, S.; Tench, A. J. Proc. Inr. Congr. Card., 7rh 1980 B, I 1 54. (16) Henderson, B. Defects in crystulline solids; Edward Arnold Ltd.: London, 1912.
(17) Edmons, R.N.; Edwards, P. P.; Guy, S.C.; Johnson, D. C. J . Phys. Chem. 1984.88, 3164. (18) Koh, A. K.: Miller, D. J. Ai. Dum Nucl. Data Tubles 1985, 33, 235.
Giamello et al.
I
I
Figure 5. X-band EPR spectra of Li/MgO at IO- (a) and 0.1-mW (b) microwave power level. The expanded trace in (a) is overamplified by a factor of IO.
Finally the hypothesis of oxygen paramagnetic species must be considered. This hypothesis has been suggested by Wisocki in a paper where the y-irradiation proces of MgO is reexamined and EPR spectra exhibiting lines very similar to A and B are reported.lg The author assigns these latter lines to 02and 01species originating from the extraction of molecular oxygen from the damaged surface and successive readsorption of the molecule onto surface defects: the observed increase in oxygen pressure during irradiation and its successive decrease are indicated by the author as the main evidence in favor of the proposed mechanism. Actually the 02radical ion is by far the most abundant product of the interaction between surface trapped electrons and diHowever, the fingerprint of this species (Le., the g,, value resonating a t values higher than 2.05-2.06) is absent in Wisocki’s spectra and in ours also: consequently the subject deserves further discussion. In conclusion, the nature of A and B signals obtained by Mg evaporation remains uncertain and requires further work for clarification. EPR Spectra of Li/MgO. Typical spectra of the Li/MgO system, recorded at 10 and 0.1 mW, are shown in Figure 5. They present some differences from the spectra of Mg/MgO. The spectrum recorded a t 10 mW (Figure 5a) is centered a t g = 2.0013 and appears to be more symmetric than that recorded a t 0.1 mW (Figure 5b). Its intensity is by far higher than those typical of Mg/MgO and the line width is about twice. The traces of a broad and unresolved hyperfine structure are observed around the main signal and the low-field lines indicated with A and B in the spectrum of Mg/MgO are absent. The low-power spectrum (Figure Sb), instead, clearly recalls the corresponding one for Mg/MgO (Figure 2a) and the two components of the axial spectrum of the F,+ centers (g,,= 2.0025, g, = 2.0003, very close to those observed in the case of Mg/MgO) are easily observed. The large line width of the Li/MgO signal partially obscures the typical line of F,+(H), which, however, is (19) Wisocki, S.J. Chem. Soc., Furuday Trans. I 1986,82,715. (20) Che, M.;Tench, A. J. Ado. Carol. 1983, 32, I . (21) Giamello, E.; et al., unpublished results.
Defect Centers Induced by Evaporation visible as a shoulder of the main signal (Figure 5b). The large line width of the signal is probably related to its intensity, which is 1 or 2 orders of magnitude higher than that of Mg/MgO (vide infra): the higher density of paramagnetic centers observed for Li/MgO is likely to increase the extent of the dipolar interaction and cause line broadening. Although the external lines are weak an scarcely resolved, the fact that they are centered around the main signal and observed in both spectra of Figure 5 indicates that they belong to one or more hyperfine structures. This is not surprising since (opposite to the case of Mg/MgO) the lithium atoms added to the oxide have an isotopic abundance in 7Li ( I = 3/2) of 93% (the remaining 7% is 6Li with I = I ) . The low intensity of the lines belonging to the 7Li hyperfine structure indicates, however, that the corresponding centers represent a relatively small fraction of the whole paramagnetic sites: most of the unpaired electrons giving rise to the spectra in Figure 5 have therefore no coupling with Li nuclei. Furthermore, due to the large line width of the central line, the inner lines of the hyperfine multiplet(s) are not visible and the existence of two hyperfine multiplets can be tentatively proposed on the basis of the position of the external lines only. The first multiplet with coupling constant A IO G is likely a seven-line structure due to the interaction of an unpaired electron with two equivalent ’Li nuclei; the second line is a quadruplet related to the interaction with one 7Li nucleus and with the same coupling constant. In the former case the seven lines should have an intensity ratio of the type 1:2:3:4:3:2:1,whereas in the latter one the four lines should be equally intense. The stick diagrams corresponding to the present interpretation are reported in Figure 5 . The presence of hyperfine structures and the relatively high line width typical of Li/MgO spectra very likely determine the shape of the EPR spectra and their lower anisotropy with respect to that observed for Mg/MgO, particularly visible on comparing the spectra recorded at 10 mW (Figures 1 and 5a). Intensity of the Spectra and Mechanism of the Metal-Oxide Interaction. The experiments reported above indicate that F,+ paramagnetic centers are formed at the surface of MgO by interaction with vapors of both magnesium and lithium. All the data collected here, as well as those from literature, agree in the indication of a model where the metal atoms ionize and both the released electrons and the positive ions are stabilized on the ionic surface. Since the ionization of a metal atom is endothermic, the driving force of the observed process derives from the stabilization energy of electrons and ions at the surface of the ionic MgO lattice. The surface ionization process will spontaneously occur upon metal atom-surface interaction when the overall gain in Madelung energy will equalize or overcome the ionization energy in the gas phase. By EPR spectroscopy the fate of electrons can be explored; however, we must always keep in mind that the positive counterpart of the ionization process (the cations) play a symmetric role (although not often directly observable by spectroscopic techniques). In other words also the cations resulting from ionization of Mg and Li should occupy those positions at the surface where a maximum Madelung stabilization is attained. A marked difference exists between the number of centers produced in the two cases: the samples treated with lithium, in fact, exhibit an average value of 3 X IO” spin/g against the value of 4 X 1OIs spin/g shown by Mg/MgO. The values recorded on each single sample fall in a range within *IO% of the average value. The intensity and shape of the EPR spectra are not affected by thermal treatment below 473 K. At this temperature a net decrease in the signal intensity of Mg/MgO occurs with some changes in the spectral profile and the number of spin decreases from 4 X 1015 to 1 X 1015spin/g. The spectral changes are accompanied by a partial loss of color. The marked difference between the concentrations of the one-electron centers observed employing lithium (one ionized electron) and magnesium (two ionized electrons) suggests that in the latter case a fraction of the electrons released by the metal could be present as neutral, diamagnetic F, centers with two trapped electrons in the same anion vacancy: this possibility would
-
The Journal of Physical Chemistry, Vol. 95, No. 23, 1991 9389 TABLE I: Number of Paramagnetic Centers for Various Treatments no. of centers, sample tme of samvle svin/g 4 x IO” 1 Mg/MgO 3.5 x 10’6 2 0ion 1 3 Mg/MgO annealed at 473 K 1 X 10l5 3 x 10’6 4 02-on 3 3 x 10’1 5 Li/MgO 2.5 X loll 6 0; on 5 be favored on the basis of the electrical neutrality of the crystal. This possibility has also been advanced in a paper by Chen et a1.9 on the basis of a comparison between optical absorption and EPR intensities of variously treated MgO samples. In order to investigate this and other points in greater depth, adsorption of molecular oxygen has been performed onto the different samples (Mg/MgO and Li/MgO, both freshly prepared and annealed). Adsorption of oxygen at room temperature leads to the formation of the superoxide radical anion 02-,which is adsorbed on the positive ions of the surface and is stable in a wide range of temperature. Oxygen adsorption brings about the instantaneous disappearance of the blue violet color of the sample, which turns white. The EPR spectrum typical of the surface centers is replaced by the well-known asymmetric spectrum of the superoxide ion, whose intensity is measured and compared with that of the starting sample. The choice of oxygen to measure the one-electron donor properties of the sample could seem, at first sight, rather strange, since in the past several acceptor molecules such as tetracyanoethylene, trinitrobenzene, and nitrobenzene have been employed to measure the same properties on various solids. We think, however, that oxygen is a suitable electron acceptor molecule for the following reasons: (i) the complete and instantaneous disappearance of the color of the sample indicates a thorough reactivity of the solid; (ii) the product of the interaction is exclusively the superoxide ion, since other products formed by capture of more than one electron such as, for example, the peroxide ion (Ol-, diamagnetic and thus escaping EPR detection) are thermodynamically unfavored under our experimental conditions;2o (iii) the knowledge recently attained concerning the chemistry of MgO indicates that the solid is able to cleave the C-H bond of several organic molecules (surface metallation reactions), so initiating a chain of surface reactions that can ultimately generate negatively charged species (negative radical anions included). These species, in principle, lead to serious interferences in the analysis of the electron-transfer process.22 The result of the spin counting is reported in Table I. Inspection of Table I clearly indicates three facts: ( I ) The vast majority of the electrons at the surface of Mg/MgO, capable of transfer toward oxygen, are present in EPR silent forms. This is deduced by comparison of the number of spin of Mg/MgO (sample 1) with that of 02-(sample 2), which represents, according to our hypothesis, the total number of surface available electrons and is about 1 order of magnitude higher. As already mentioned, we think that at least a fraction of the surface electrons on MgO is trapped in the form of diamagnetic F, centers, each containing two electrons, previously observed on the same oxide by Chen et al.9 These centers, on the basis of the present results, function as single electron donors. A further possibility that cannot be discarded is that of the presence of small diamagnetic clusters of unionized Mg atoms. The presence at the surface of such clusters would not be, in principle, very surprising because of the relatively high ionization energy of the metal. However, in this case, the contact with oxygen very likely would lead to the total oxidation of the metal to MgO (highly exothermic) and not to the intermediate state of superoxide formation. We thus suggest that the electrons revealed by the oxygen adsorption as 02-mainly come from F, centers. (2) The annealing causes migration and coupling of single electrons from the F,+ centers to give further F, neutral centers, (22) Giamello, E.; Ugliengo, P.;Garrone, E. J . Chem. SOC.,Faraday Trans. I 1989, 85, 1373.
9390 The Journal of Physical Chemistry, Vol. 95, No. 23, I991 as deduced by comparing samples 1,3, and 4. This mechanism of electron pairing seems to be in contrast with what was observed on y- and neutron-irradiated MgO samples for which it has been observed3 that annealing a t 523 K causes the loss of reactivity of the samples with disappearance of trapped electrons. In that case, migration of the trapped electrons in the bulk and a recombination with positive holes (V centers) was hypothi~ed.~ This is probably true on samples obtained by the irradiation method (i.e., without adding matter to the oxide), for which separation between negative charges and positive holes is needed. The opposite exists in the case of the present work; the surface negative charge derives from the ionization of the metals (Mg, Li) and is balanced by the residual positive ions themselves (Mg2+,LP). It is not, therefore, surprising that under the conditions of thermal treatment migration occurs without charge recombination. (3) The evaporation of lithium produces a density of trapped electrons higher than that of magnesium (samples 1-5). The relatively low energy required for the one-electron ionization of lithium (520 kJ/mol), compared with that for the double ionization of magnesium (737 and 1451 kJ/mol for the first and second, respectively), is the reason for this result. Moreover, as the results for samples 5 and 6 are very close to one another, it can be suggested that nearly all the surface-trapped electrons of Li/MgO are in the form of monoelectronic, paramagnetic F: centers. Divalent magnesium atoms, therefore, mainly generate electron pairs, whereas monovalent lithium atoms give rise mainly to single trapped electrons. A last comment is devoted to the nature of surface anion and cation vacancies capable of trapping the electrons and localizing the cations. On (100) faces the anion vacancies are usually thought of (since they ideally derive from the octahedral bulk vacancies) as being limited by a square-pyramidal array of five positive Mg2+ ions, four of them being in a 4-fold coordination state.
Giamello et al.
f
+
b
+
/ a\
\
Figure 6. Schematic view of the MgO surface. I: Before metal vaporization. 11: After metal vaporization. The label a indicates a surface anion vacancy and the label b a reconstructed site at a step of the surface. The defects illustrated in I1 are F,+ centers. An analogous scheme applies for F,. The dashed symbols represent the positive ions (in this case Li+) resulting from metal evaporation.
++I++ ++ Symmetrically, the (100) cation vacancies are thought of as being constituted by a pyramidal array of 02-ions (four in a fourfold coordination state). The origin of the neutral anion vacancies has probably to be found, as discussed in ref 23, in the process of extended dehydroxylation occurring upon evacuation of MgO samples at high temperature. It has been noticed, however, that this kind of vacancy probably occurs not only at the (100) faces but also at the edges of the microcrystal^.^^ On high-area thermally activated MgO, in fact, it is known that a high amount of the surface Mg2+and 02-ions (about 25% of the total) are located at morphological defects (edges, kinks, steps, terraces, etc.), where they are mainly in 4-fold but also in 3-fold coordination state^.^)-^^ Consequently, also electrons trapped in sites containing ions in an even lower coordination state (and hence with lower electrostatic stabilization energy) could exist at the surface, in agreement with the slight heterogeneity observed in the EPR spectra. It has been reported that, in the case of y-irradiation, only the anion vacancies already present at the surface of the activated MgO are filled by electrons to give F,+ centers, because this kind of irradiation process has not sufficient energy for their generation. In contrast, neutron irradiation is sufficiently powerful to create oxygen vacancies and hence a higher number of surface F: centers3 The process occurring at the MgO surface upon contact with Li or Mg vapors differs from both of the irradiation processes, since it is based on the ionization of the metal atoms and on the (23) Coluccia,S . ; Lavagnino. S.: Marchese, L. Mater. Chem. Phys. 1988, 18. 445. (24) Zecchina, A.; Spoto, G.; Coluccia. S. J . Mol. Catal. 1982, 14, 351. (25) Ugliengo, P.; Borzani, G.;Viterbo, D. J . Appl. Crystallogr. 1988, 21, 75.
Figure 7. View of two 'reconstructed sites" for electron trapping obtained by addition of one and two extra ions. Large spheres: 02-ions. Small spheres: Mg2+ ions. Black spheres: positive ions resulting from metal ionization. Dotted spheres: Mgz+ ions belonging to the site. The drawings have been obtained by the computer program MOLDRAW."
stabilization of both the electrons and the metal ions formed. The electrons released from the metal can thus be hosted, in principle, in two types of sites: (i) anion (Schottky) vacancies present on activated MgO (as in the case of y-irradiation), according to the following schemes where the metal considered is magnesium
+ Mgo 20, + Mgo 0,= F+ :
-
-
0,
+ Mg2' 2F,+ + Mg2+ F,
= preexisting Schottky anion vacancy
whereby the formed Mg2+ions are located in the preexisting cation Schottky vacancies; (ii) newly formed vacancies generated during the adsorption process following the very general scheme Mgo
-
Mg2+ + 2F,+
whereby the Mg2+ and F,+ are stabilized by the Madelung potential of the surface, possibly through surface reconstruction at morphological defects (vide infra). The number of F,+ centers per square meter produced by lithium evaporation (assuming a surface area of 100 m2/g) is about
J. Phys. Chem. 1991, 95, 9391-9396 This means, taking into account the lattice parameter, that
on a (100) face about 0.18 of the anionic sites should be vacant. This percentage is probably too high to be due exclusively to preexisting vacancies (i); the role of the second class of vacancies must thus be considered. A confirmation of this idea comes from the hyperfine structure observed for a fraction of the paramagnetic centers on Li/MgO and due to one or two ’Li ions in the first coordination sphere of the trapped electron. Preliminary computer simulation results indicate that the fraction of sites exhibiting hyperfine coupling with one or two Li nuclei is about 25%. The whole process occurring at the surface can be, in conclusion, described as the ionization of a metal that releases one or two electrons. Some electrons are stabilized in surface vacancies deriving from the thermal treatment and the remaining in new anion vacancies formed by a (partial) reconstruction of the surface or, better, of the cationic sublattice of the surface.
9391
A schematic view of the process is shown in Figure 6. The surface reconstruction begins, very likely, at the morphological defects of the surface such as kinks, steps, or terraces, because in these points the electrostatic potential (or, in other words, the coordination number attained by the added ion) is higher than on a flat ( 1 00) face. Moreover, the number of additional ions required to complete a 5-coordinated anion vacancy at the mentioned morphological defects is limited to one or two, whereas the reconstruction of the same site on a flat (100) face needs four additional ions in vicinal positions and is therefore less probable. A picture of two possible “reconstructed” sites, involving one and two additional ions, is shown in Figure 7. The hyperfine structure shown by some F,+ centers on Li/MgO, and involving the presence of one or two Li nuclei in the site, once again confirms the above description of the surface. Registry No. MgO, 1309-48-4;Li, 7439-93-2;Mg, 7439-95-4.
Structure and Dynamics of Amino Functional Silanes Adsorbed on Silica Surfaces Hye-Jung Kang and Frank D. Blum* Department of Chemistry and Materials Research Center, University of Missouri-Rolla, Rolla, Missouri 65401 (Received: October 22, 1990)
Deuterated (aminopropy1)triethoxysilane(DAPES) and deuterated (aminobutyl)triethoxysilane (DABES) have been adsorbed on high-surface-area silica (CabOSil), dried, and probed with 2H (wide-line)and %i (CP/MAS) solid-state NMR techniques. When adsorbed, the DABES reacted to form siloxane linkages with the surface or other DABES molecules. The amount of DABES adsorbed on silica was dependent on the concentration of the silane coupling agent solution. The molecular motion of the adsorbed silane coupling agents on silica have been probed as a function of the amount and type of coupling agent. The slow-motional theory of Freed was used to simulate the resulting ZHspectra. The spectra were simulated by a superposition of a rigid and a mobile species. The C-D bonds of the mobile species undergo anisotropic diffusion (d, = 70-500 kHz and dxy = 10-100 kHz) plus faster two-site jumps about a diffusion axis tilted at 54.7’ (rate = 1-2 MHz). When treated with a solution resulting in approximately monolayer coverage, the molecular motion of the coupling agent was the slowest. From treating solutions resulting in coverages above and below monolayer coverage, the spectra of the coupling agents on the surface were indicative of more mobile species. At the higher concentrations, the increase mobility a was similar to that of a polysilsesquioxane layer which was indirectly bonded to the surface was observed. Surfaces treated to be at submonolayer coverages were also less restricted. The 2HNMR spectra for DAPS and DABS on the silica surface also showed that the difference in the alkyl chain length of silane coupling agents did not significantly affect the molecular motions of silane when they were adsorbed on silica surfaces for these short (C, and C,) chains.
Introduction Silane coupling agents are widely used to improve the bonding between silica and polymeric resins by altering the properties of the interfacial layer. In addition to the bulk properties of the component species, the properties of the interfacial region are very important in determining the properties of the composite materials. For example, in glass fiber/polyester composites, the treatment of glass with silane coupling agents has been shown to affect the mechanical properties of the composite as a function of both the amount and type of coupling agent used.’ Since the interfacial layers are very thin, it is necessary to probe their behavior on a molecular level. Microscopic interfacial properties can conveniently be classified as probing either structure or dynamics. These two are clearly interrelated, but different experimental techniques are usually sensitive to different kinds of properties. A structural property might be the conformation of a molecule on a solid surface or the mechanism by which it was attached to the surface. A dynamic or motional property might be the rate and manner in which the molecule moves while attached to the surface. In order to relate the microscopic interfacial properties to the mechanical properties in composite systems, a good deal more needs to be known about interfacial species. To whom correspondence should be addressed.
A significant amount of work, based various techniques, has been done to elucidate the structure of aminosilanes in solution and on substrates. Chiang et al.* have used FTIR spectroscopy to study the structure of (y-aminopropy1)triethoxysilane (APS) in aqueous solution and on glass surfaces. They reported that the hydrolyzed amino functional silane exists in two structural forms, internal cyclic ring and non-ring-extended structures. It was found that the hydrolyzed aminosilane adsorbed onto high-surface-area silica approximately as a monolayer and that the amino group is hydrogen bonded to the surface. Johansonn et aL3 used radiochemical tracers and found that APS formed uniform films on E-glass surfaces with covalent bonding a t the interface. Schrader4 has proposed a combination of physisorbed silane and two types of chemisorbed species on silica surfaces. The physisorbed silane coupling agents were not reacted with the E-glass surface and were removed with solvent washing. The two chemisorbed silanes were distinguishable by their ability to be removed with boiling water. The nonremovable species corresponded to ( 1 ) Eckstein. Y . J . Adhes. Sci. Technol. 1989, 3, 337. (2) Chiang, C. H.; Ishida, H.; Koenig, J. L. J . Colloid Interface Sci. 1980, 74, 396. (3) Johansonn, 0. K.; Stark, F. 0.;Vogel, K. M.; Fleischmann, R. M. J . Compos. Mater. 1967, I , 278. (4) Schrader, M . E. J . Adhes. 1970, 2, 202.
0022-3654/91/2095-9391%02.50/00 1991 American Chemical Society
