Structural and Induced Heterogeneity at the Surface of Some Si02

also due to the species originated by ita dissociation on strained bridges, during ita .... (24) Lefebvre, Y.; Jolicoeur, C. In R.K. Zler Memorial Sym...
0 downloads 0 Views 2MB Size
Langmuir 1993,9, 2712-2720

2712

Structural and Induced Heterogeneity at the Surface of Some Si02 Polymorphs from the Enthalpy of Adsorption of Various Molecules Bice Fubini,' Vera Bolis, Albert0 Cavenago, Edoardo Garrone, and Piero Ugliengo Dipartimento di Chimica Inorganica, Chimica Fisica e Chimica dei Materiali, Universith: di Torino, Via P. Giuria 7, 10125 Torino, Italy Received November 9, 1992. I n Final Form: July 1, 199P Surface heterogeneity of both amorphous (Aerosil) and crystalline polymorphs of silica (a-quartz and a-cristobalite)has been studied by measuringthe heat of reversible adsorption of water, ammonia,methanol and tert-butyl alcohol as a function of coverage and through the comparison with ab initio results on cluster models. The adsorption of tert-butyl alcohol is rather insensitive to both crystallinity and degree of dehydration, being largely due to nonspecific dispersive interactions. Water, ammonia, and methanol revealstructuralheterogeneity,though to adifferent extent; a peculiar induced heterogeneity transmitted through H-bondingof interactingsilanoh is evidenced by the adsorption of ammonia. The most dehydrated sample studied is the high-temperature treated Aerosil: the heat of adsorption of water reveals some heterogeneity because of the different role of isolated and geminal silanols; ammonia shows heterogeneity also due to the species originated by ita dissociation on strained bridges, during ita previous contact with the surface. On mildly dehydrated surfaces, patches of silanols are present: heat of adsorption of water distinguishes between such hydrophilic patches and hydrophobic ones. Extensively hydrated samples exhibit large patches of silanols, interacting with one another through H-bonds: adsorption of ammonia reveals that the H-bondingstrength of the terminal hydroxyl increases with the size of the patch. Rupture of H-bonds among adjacent SiOH groups occurs with increasing ammonia coverage, revealed by a linear decrease of the heat of adsorption over a very wide range of coverage: accordingly, the experimental isotherm follows the Temkin model. Ab initio calculations on chains of interacting silanols fully confirm such a picture.

Introduction Structural and induced heterogeneity in adsorption sites are often present on finely divided materials, the former arising from morphological features and the latter from electronic effects transmitted through the solid. With highly covalent solids, such as silica, structural heterogeneity only is expected, as no electronic effects are supposed to occur through the solid. Structural features causing intrinsic heterogeneity on silicas are, with crystalline samples, the kind of exposed crystal planes and other sorts of geometrical defects like steps, kinks, and edges; with amorphous specimens, heterogeneity arises only from different chemical functionalities at the surface. Three main kinds of adsorption sites are found on a silica surface: (i) silanols, either isolated, geminal, or H-bonded to each other in complex arrays;IA (ii) siloxane bridges, which may occur as strained ones, and thus are very reactiveFp6 or as regular ones, which are mainly unreactive;7*8(iii) and surface radicals related to cleaved

* To whom correspondence should be sent. e Abstract published in Advance ACS Abatracta, August 15,1993.

(1) KnOzinger, H. In The hydrogen bond; Schueter, P., Zundel, G., Sandorfy, C., Eds.; North Holland Amsterdam, 1978; p 1263. (2) Zhdanov, S. P.; Kosheleva, L. S.;Titova, T. I. Langmuir 1987,3,

960.

(3) Brinker, C. J.; Tallant, D. R.; Roth, E. P.; Ashley, C. S. J. NonCryst. Solids 1986,82,117. (4) Brinker, C. J.; Kirkpatrick, R. J.; Tallant, D. R.; Bunker, B. C.; Montez, B. J. Non-Cryst. Solids 1988,99,418. (5) Morrow, B. A,;Cody, I. A. J. Phys. Chem. 1976,80,1995. Morrow, B.A.;Cody,I.A.;~,L.S.M.J.Phys.Chem.1976,80,2761,andreferences therein. (6) Michalske, T. A.; Bunker, B. C. J. Appl. Phys. 1984,56,2686. (7) Bolis, V.; Fubini, B.; Marchese, L.; Martra, G.;Costa, D. J.Chem. SOC.,Faraday Trans. 1991,87,497. (8) Fubini, B.; Bolis, V.; Cavenago, A,;Ugliengo, P. J. Chem. SOC., Faraday Trans. 1992,88, 277.

Si-0 bonds and further reaction with the atmosphere, typically present on particles obtained by mechanical grinding.+" The distribution at the surface of morphological irregularities and of these various sites determines the heterogeneity, and hence the reactivity, of both polycrystalline and amorphous powders, as well as their pathogenic potential when inhaled.6J2 In spite of massive work on silicas in both past13J4and recent y e a r ~ , 2 Jthe ~ ~reactivity ~ of these various functionalities has not yet been fully clarified, and the influence on it of crystallinity and of thermal and mechanical history of the sample is still unclear. The knowledge of these facta is of paramount importance in different fields, from material science to pathogenicity of inhaled dusts in both such fields, crystalline polymorphs (whichhave been much less investigated than the amorphous ones) are directly involved. (9) Radtzig, V. A,;Bystrikov, A. V. Kinet. Katal. 1978,19,713. (10) Steinicke, U.; Henning, H. P.; Richter-Mendau, J.; Kretzchmar, U. Cryst. Res. Technol. 1982, 17, 1585. Steinicke, U.; Kretzchmar, U.; Ebert, I.; Henning, H. P. React. Solids 1987,4,1. (11) Fubini, B.; Giamello, E.; Puglieee, L.; Volante, M. Solid State Ionic8 1989,32-33,334. V. Toxicol.Ind.Health (12) Fubmi,B.;Giamello,E.;Volante,M.;Bolis, 1990, 6, 571, and references therein. (13) Iler,R. K. The ChemktryojSilica; WileyInterscience: NewYork, 1979; (a) Chapter 7, p 769; (b) Chapter 6, p 622, and references therein. (14) Zettlemoyer, A. C.; Micale, F. T.; Klier, K. Water in dispersed Systems; Franks, F., Ed.; Plenum: New Yolk, 1975; Vol. 5. Texter, J.; Klier, K.; Ittlemoyer, A. C. Prog. Surf. Membrane Sci. 1978,12,327. (15) Legrand,A.P.;Hommel,H.;Tuel,A.;Vidal,A.;Balard,H.;Papirer,

E.;Levitz,P.;Czemichowski,M.;Erre,R.;VanDamme,H.;Gallas, J.P.; Hemidy, J. F.; Lavalley, J. C.; Barres, 0.; Burneau, A.; Grillet, Y. Ado. Colloid Interjace Sci. 1990,33, 91, and references therein. (16) Papirer, E.; Keseaiseia, Z.; Balard, H. Bull. SOC.Chim. Fr.1980,

- _-.

Ad1

(17) Zhuravlev, L. T. Langmuir 1987,3, 316. (18) Ligner, G.; Vidal, A.; Balard, H.; Papirer, E. J . Colloid Interface Sci. 1989,133, 200, 1990,134,486. Zaborski, M.; Vidal, A,;Ligner, G.; Balard, H.; Papirer, E.; Burneau, A. Langmuir 1989,5, 447. (19) Scherer, W. J. Non-Cryst. Solids 1988,100, 77.

0 1993 American Chemical Society 0743-7463/93/2409-2712$04.~/0

Surface Heterogeneity of Silica

Work has been carried out by some of us in the past years on the surface reactivity of crystalline polymorphs of Si02 in comparison with the amorphous ones, in order to give a detailed picture of the distribution of the chemical functionalities at the silica surface, depending on its origin and thermal history. Surface radicals, which are a minor feature from a quantitative point of view, have been discussed in previous papers by some of us.12 The reactivity of strained siloxane bridges in dissociating various molecules has also been recently reported, and it has been shown that ammonia and water are irreversibly dissociated on these sites, whereas they are reversibly coordinated onto silanols through H - b ~ n d i n g .Silanols ~ ~ ~ and geminal species, on the other hand, have been extensively studied by some of us by means of ab initio quantum mechanical calculations based on cluster The present paper deals with the heterogeneity related to the distribution of silanols a t the surface and focuses on two aspects: the role of the probe molecule in the detection of structural heterogeneity and the occurrence of an induced heterogeneity-via collective effects through H bonds-shown by the differential heat of adsorption of ammonia.8 Adsorption on various silica samples has been investigated by measuring adsorbed amounts as well as related heats as a function of coverage of four probe molecules; water, methanol, tert-butyl alcohol, and ammonia, all acting as H acceptors with silanols; the first three also acting as H donors. Reasons for the choice of water are evident, methanol is a common eluent in chromatography and ammonia is the strongest base and has negligible tendency to act as H donor. tert-Butyl alcohol is interesting because its latent enthalpy of liquefaction is very close to that of water (44 kJ*mol-'), and comparison with methanol may give evidence of the effect of methyl groups in dispersive interaction with the s u r f a ~ e . ~ ~ ~ ~ ~ The surface of all samples has been examined in somewhat extreme conditions, i.e., either fully hydrated or deprived, by thermal treatment in vacuo, of most surface hydroxyls. In the case of ammonia, the heat of adsorption was measured systematically after a large set of pretreatment temperatures. The probe molecules employed are known to yield both reversible and irreversible adsorption on silica, the latter arising from the reopening of siloxane groups created by dehydration. This means that some changes are brought about with respect to the state of the surface produced by thermal treatment in uacuo. Reversible adsorption, measured from an adsorption run successive to a first adsorption-desorption cycle, characterizes the state of the surface after irreversible a d ~ o r p t i o n . ~ ? ~ To discuss structural heterogeneity, recourse has been made to the computational results obtained by some of us concerning the interaction of water, ammonia, and methanol with suitable model clusters, i.e., the silanol molecule H3SiOH (which mimics the isolated silanol at the real surface) and silanediol H2Si(OH)2 (modeling the geminal (20) Ugliengo, P.; Saunders,V.; and Garrone, E. J. Phys. Chem, 1990, 94, 2260.

(21) Ugliengo, P.; Saunders, V.; Garrone, E. Surf. Sci. 1989,224,498. (22) Ugliengo, P.; Bleiber, A.; Garrone, E.; Sauer, J.; Ferrari, A. M. Chem. Phys. Lett. 1992,191, 537. (23) Ferrani,A. M.; Ugliengo, P.; Garrone, E. J. Phys. Chem. 1993,97, 2671. (24) Lefebvre, Y.; Jolicoeur, C. In R.K. Zler Memorial Symposium on the Chemistry of Silica; American Chemical Society: Washington, DC, in press; Lefebvre, Y.; Jolicoeur, C. Colloids Surf., in press. (25) Lefebvre,Y. Etude phyeico-chimique des proprietes de surfacede ailice polymorphes. Thesis, Sherbrooke University, Canada, 1992.

Langmuir, Vol. 9, No. 10,1993 2713 species). Moreover, to study collective effects in silanol clusters, novel calculations have been carried out concerning the interaction of ammonia with rows of H-bonded silanols differing in length, as well as the interaction of water with a couple of adjacent silanols.

Experimental Section Materials. Most of the silicas employed have been already described in previous papers. The amorphous sample was a relatively low surface area (50 m2 g-1) pyrogenic silica (Aerosil50), especially chosen in order to facilitate comparison with the intrinsically low surface area crystalline polymorphs. Some of ita surface properties have been reported in refs 26 and 27. Crystalliie Si02 samples were as follows: (i) a-cristobalite, or micrometer-sized particles, from C. & E. Mineral Corp., kindly provided by Dr. Hemenway (Universityof Vermont). It exhibited in various in vivo and in vitro testa a high fibrogenicpotential.The specificBET surface area was 6.2m2g-1. Some of its surface properties are reported in refs 8 and 29. (ii) a-quartz, a finely divided powder obtainedby grinding very pure (99,999% ) quartz chips (from Atomergi~);~ BET surface area, 10 m2 gl. Water, methanol and tert-butyl alcohol were distilled several times in vacuo and rendered gas-free by several freeze-pump thaw cycles. Ammonia gas was Specpure from Matheson. Methanol and tert-butyl alcohol were from Merck. Experimental Methods. The simultaneous measurement of the heat of adsorption and the adsorbed amount was performed by means of aTian-Calvetmicrocalorimeter (Setaram)connected to a volumetric apparatus, following a well-established procedure?*" The samples, placed in the calorimetric cell, were pretreated in uacuo at the chosen temperatureand subsequently transferred into the calorimeter without further exposure to the atmosphere. Adsorption was performed in all cases at 303 K small doses of the adsorptive were subsequently admitted onto the sample, the pressure being continuously monitored by means of a 0-100 Torr (1 Torr = 101325/760 Pa) transducer gauge (Baratron MKS). With all probe molecules,adsorption occurred also on the frame glass walls. When this process becomes competitive with the adsorption on the actual sample, the volumetric data are overestimated, with consequent errors in both volumetric isotherms and differential heata, while calorimetricisotherms(heat evolved us equilibriumpressure) are virtually unaffected. This occurs when dealing with low surfacearea samples, in particular (as extensively discussed in ref 8) at relatively high pressures. Adsorption runs were therefore stopped below 8 Torr for water, 50 Torr for ammonia, and 20 Torr for methanol and tert-butyl alcohol, as only above these pressures does the adsorption on glass become important with respect to what is adsorbed on the sample. Accordingly, comparisons between the various samples were made at 5 Torr for HzO, at 40 Torr for NHs, and at 10 Torr for methanol and tert-butyl alcohol, where the isotherms are reasonably error-free. The vapor pressure @o) values at 303 K for water, methanol, and tert-butyl alcohol are 32,200,and 60 Torr, respectively; for NHs it is 10 X 109 Torr. Under our experimental conditions, adsorption of the three hydroxyl-containing species takes place at p/poup to 0.245;the p/po value for ammonia does not exceed 5 X 10-9, i.e., the adsorption process is by no means relatable to liquefaction. For this reason the latent heat of liquefaction (ML) value for ammonia is not reported as reference in the heat of adsorption vs coverage plots. This is instead done in the case of the otherthree adsorptives,for which the experimental conditions are not too far from those of liquefaction. The values of -ML (26)Bolis. V.: Marchese,. L.:. Coluccia, S.: Fubini, B. Adsorpt. Sci. Technol. 1988,5, 239. (27) Bolis, V.; Fubini, B.; Giamello, E. Mater. Chem. Phys. 1991,29, 1C n

IDS.

(28) Absher, M.; Hemenway, D. R.; Trombley, L.; Rigatti, M.; Vacek, P. Am. Rev. Respir. Dis. 1988, 137, 316. Absher, M.; Trombley, L.; Hemenway, D. R.; Mickey, R. M.; Leslie, K. 0.Am. J. Pathol. 1989,134, 1243. (29) Hemenway, D.; Absher, M.; Fubini, B.; Vacek, P.; Volante, M.; Cavenago, A. Ann. Occup. Hyg., in press. (30) Fubini, B. Thermochim. Acta 1988,135, 19.

2714 Langmuir, Vol. 9, No. 10, 1993

Fubini et al.

Table I. Quantitative and Energetic Data for the Adsorption of Water, Ammonia, Methanol, and &&-Butyl Alcohol on Pyrogenic Silica and Cristobalite Variously Outgassed

A50-423 A50-1073 CRIS-423 CRIS-1073

C0.2 0.4 1.3 2.6

2.0 0.4 8.1 3.8

40 30 53 47

C0.2 0.3 1.6 1.1

2.9 1.5 8.6 5.7

for water, methanol, and tert-butyl alcohol are 44,38, and 44 kJ mol-l, respectively. The procedure for each adsorption cycle was as follows: (i) adsorption of successive doses of the adsorptive, up to the established equilibrium pressure limit, on the fresh sample (Ads I); (ii) desorption by direct evacuation of the sample until no deviation on the calorimetric baseline was detectable; (iii) adsorption of doses similar to those in Ads I up to the same equilibrium pressure, in order to evaluate the fraction of the adsorbate which is reversible at the adsorption temperature adopted (Ads 11). As already said, in most cases Ads I comprises both irreversible and reversible processes. These latter, the ones relevant to the present paper, are thus described by Ads II; unless otherwise stated, all curves and data reported refer to the reversible adsorption (Ads 11). Thermal Treatments. Prior to adsorption, the samples were outgassed at temperatures chosen on the basis of previous experience:lm423 K, in order to removecontaminatingadsorbed species and weakly adsorbed water; 773 K, when a relatively large amount of surface silanolswill have condensed intosiloxane bridges; 1073 K, when most amorphous silicas (exposing after this treatment isolated silanols and siloxanes bridges only) are virtually hydrophobic, whereas crystalline polymorphs exhibit just a marked decrease in the surface activity toward water; 1573 K, in order to have also a crystalline specimen virtually hydrophobic. All thermal treatments lasted 2 hand were performed in uacuo, with the only exception of that at 1573 K which was carried out for 4 h in a flow of inert gas.8 Nomenclature. The samples will be hereafter referred to as A-50 for Aerosil 50, CRIS for cristobalite, and QRZ-p for the pure quartz, followedby the outgassing temperature; e.g., CRIS1073 indicates a sample of cristobalite outgassed at 1073 K. Ab Initio Computations. Full details on the cluster models are given in refs 20-23. We recall here that calculations have been run by means of the GAUSSIAN 92 package.al Already published results have been obtained adopting a double zeta basis set plus polarization functions, correcting the energetic data for the basis-set superposition error (BSSE)and takiig into account electronic correlation by means of the MP2 perturbative approach. As they concern rather large systems, new results reported in the present paper have been obtained with the MINI-1 basis seta2which, though being a minimal basis set, is known to be suitable for evaluatingreasonable H-bonding binding energies, when BSSE is taken into account. Calculations have been carried out on IBM-320H/RS6000and INDIGO Silicon Graphics workstations.

Results and Discussion

Figures 1-4 report isotherms and differential heat of adsorption versus coverage plots (obtained as described in ref 30) for the reversible adsorption (Ads 11) of water, ammonia, methanol and tert-butyl alcohol, respectively, on cristobalite and Aerosil, in two extreme conditions, i.e. outgassed mildly at 423 K and strongly at 1073 K. The (31) Frisch, M. J.; Tmcke, G. W.; Head-Gordon, M.; Gill, P. M. W.; Wong, M. W.; Foresman, J. B.; Johnson,B. G.; Schlegel,H. B.; Robb, M. A.; Replogle, E. S.;Gomperta, R.; Andrea, J.L.; Raghavachari,K.; Binkley, J. S.; Gonzalez, C.; Martin, R. L.; Fox, D. J.; Defrees, D. J.; Baker, J.; Stewart, J. J. P.; Pople, J. A. Gaussian 92, Revision B;Gaussian Inc.: Pittaburgh PA, 1992. (32) Tatewaki, H.; Huzinaga, S. J. Comput. Chem. 1980,1, 205.

48 40 52 45

c0.2 0.3 0.8 0.8

3.5 1.2

60 60

5.5 1.3

67 65

c0.2 C0.2 0.3 0.5

0-0

0

2

0

2

4

6

0

4

6

8

p / Torr

0

4.4 4.1 3.0 2.4

2

4

70 62 77 67

6 % /"or

0

1 m-2

0

-0 E, \

e

p/Torr

Figure 1. Volumetric isotherms and differential heat of adsorption of H2O at 303 K on CRIS-423 (0)and CRIS-1073 (c]), sections a, a', and on A50-423 (0)and A50-1073 (n),sections b, b'; latent enthalpy of liquefaction of H20, -AHL= 44 kJ-mol-1.

effect of the thermal treatment is thus shown by comparing data on the same sample outgassed at the two different temperatures, whereas the effect of crystallinity is highlighted by comparing the amorphous and the crystalline sample at the same outgassing temperature. For a general comparison among the four adsorbates, Table I reports adsorbed amounts and relevant molar heat (Q) measured for each adsorbate in the range in which the isotherm exhibits a relatively small slope and the differential heats ( q ) have attained a plateau. Table I also reports for the four cases under study the amounts irreversibly adsorbed during Ads I these measure the changes in the surface composition caused by the primary adsorption run with respect to the situation created by the thermal treatment. Though such study is not a primary objective of the present work, it must be noticed that the extent of irreversible adsorption may be sizable, in particular for CRIS. It appears from curves in Figures 1-4 and data in Table I that, on the one hand, each adsorbate exhibits a different adsorption characteristic and that, on the other hand, thesensitivity to both crystallinity and therm1 treatment markedly depends upon the adsorbate. More detailed information is gained examining in particular the behavior of differential heat as a function of coverage: from the extrapolated zero-coverage values, the energy of interaction with the strongest sites is evaluated; from the plateau value

Langmuir, Vol. 9, No. 10,1993 2715

Surface Heterogeneity of Silica

no Of CH+,COHmolQulos/nrri2

-

00

-

p /Torr

0

0

2

4

6

8

1

0

rg/pmol in*

44.8

0

5

1 0 1 5 x ) p /Torr

p /Torr

Figure 2. Volumetric isotherms and differential heat of adsorption of NHs at 303 K on CRIS-423 (0) and CRIS-1073 (01, sections a, a', and on A50-423 ( 0 )and A50-1073 (a),sections b, b'.

3

c

'4

0

I

2

4 6 8 na/pmol m q

10

Figure 4. Volumetric isotherms and differential heat of adsorption of (CHs)aCOH at 303 K on CRIS-423 ( 0 )and CRISand A50-1073 ( 0 ) , 1073 (o),sections a, a', and on A50-423 (0) sectionsb, b'; latent enthalpy of liquefaction of (CHahCOH,-AHL = 44 kJamol-1.

noof CYOH molocules/nm-2

10

a)

,AHL-

4 0

5

?

3

1

5

2

I

0

p /Torr

N

I

I s

Figure 3. Volumetric isotherms and differential heat of adsorption of CHaOH at 303 K on CRIS-423 ( 0 )and CRIS-1073 (n),sections a, a', and on Am-423 (0) and A60-1073 (01,sections b, b'; latent enthalpy of liquefaction of CHaOH, -AHL = 38 kJ.mol-1.

(if present), the energy of interaction in multilayers or on homogeneous predominant patches on the surface; the extent of coverage in which the heat value decreaaesbefore reaching the plateau, measures the extent of heterogeneity toward a particular adsorbate, whereas the extension of the plateau reveals homogeneity.

A clear feature in the results is the decrease in the heat of adsorption of the strongest sites for increasing pretreatment temperature, particularly evident for ammonia: this fact is worth of note, because with solids more ionic than silica the strongest sites (which are usually surface Lewis acid sites) become available only after harsh thermal treatments, whereby coordinatively unsaturated ions are exposed as a consequenceof the removal of surface hydroxyl groups, or other impurities. In highly covalent solids, like the present ones, the hydroxyls themselves are the sites for interaction: it so appears that dehydration weakens the hydroxyl strength, i.e., that H-bonded hydroxyls are more acidic than isolated ones. Such phenomenon, revealing induced heterogeneity among adsorption sites, is discussed in the second part of the paper for ammonia. Indeed, ammonia shows here other peculiar features. We did not succeeded in measuring any plateau in the differential heat of adsorption us coverage plot, because ammonia is adsorbed at very low relative pressures @/PO < 5 X 10-9, far from the liquefaction conditions. The span, however, in the differential adsorption heat values is the largest,so that the same probe molecule shows the highest and the lowest value in the same run: accordingly,the isotherms span over a much larger pressure range than for the other probemolecules. Also this feature is amenable to a cooperative behavior of silanol patches, and is discussed in the second part of the present paper. Ageneral feature concerning the other three adsorbates is the tendency of the heat of adsorption us coverage plot toward a plateau value, which appears to be below the latent enthalpy of liquefaction for water and above it for the two alcohols. The coverage values attained on our amorphous silica are below those expected for the adsorption of one molecule per silanol on the basis of the values reported in ref 17 (-4.7 SiOH nm-2 on hydrated samples, =l SiOH nm-2 on highly dehydrated ones).

Fubini et al.

2716 Langmuir, Vol. 9,No. 10,1993 H

I

AEC -47.5 kJ/mole Figure 6. Cluster model for the double interaction of a water molecule with adjacent silanols: structure and BSSE corrected binding energy at SCF-MINI-1 level.

0 10

I00

Figure 5. 1:l molecular complexes of water, ammonia, and methanol with the silanol molecule; 1:l complex of ammonia with silanediol; 2:l complex of water with.silanediol: structures and BSSE corrected binding energies at MPZ-DZP//SCF-DZP level (refs 16-19).

Considering that a much higher population is expected on crystalline formsF3in this case, too, the maximum coverage attained should be still below the silanol population. Other features are related to the presence of definite structures on the surface to which interaction occurs; to this purpose we will discuss separately the four adsorbates. Computationalresults, obtained in previous are gathered in Figure 5 and will serve as reference; the corresponding experimental results are the adsorption enthalpies extrapolated at vanishing coverage, for samples pretreated at high temperature. Structural Heterogeneity. The Water-Silica System. Water is by far the most extensively investigated adsorptive on silicas.l,5J4 Previous ~ 0 r k ~ , 8 ,has 3 ~ pro,~~ posed that water is adsorbed by H-bonding either to a single SiOH with one hydrogen bond as shown in Figure 5 or to two SiOH groups appropriately located by two hydrogen bonds, as shown in Figure 6, where water acts as both H acceptor and H donator. In the first case the enthalpy of adsorption is currently presumed to be below the latent enthalpy of liquefaction (44 kJ-mol-l), whereas in the latter case the enthalpy is thought to be above that value. The presence of a substantial fraction of geminal (33)Boehm, H. P. Adu. Catal. 1966,16, 226. (34) Fubini, B.; Bolis, V.; Bailes, M.; Stone, F. S. Solid State Ionics 1989,32133,258. (35) Garrone, E.; Ugliengo, P. Mater. Chem. Phys. 1991, 29, 287.

Garrone, E.; Ugliengo, P.; Fenari, A. M. In New Trends in Physical Chemistry; Edited by the Council of Scientific Research Integration: Trivandrum (India), in press.

Figure 7. Hydroxyl arrangement at (100) (right) and (010) (left) crystal planes on a-quartz.

species (up to 30% of the hydroxyl population)16has also to be taken into account: the interaction of water-geminal silanols is shown in Figure 7. Highly dehydrated surfaces, basically made up of siloxanes and isolated geminal species, are therefore virtually hydrophobic, in that the enthalpy of adsorption, except at the lowest coverages, is below the latent enthalpy of liquefaction,7,8J4as observed in Figure 1. Heterogeneity, in this case, is only ascribable to a few hydrophilic sites ( A 3 > AHL)originated by a limited dissociation of water on still reactive bridges and to the differences between isolated and geminal silanols, both hydrophobic ( A a < A&,). As the strongest sites are the first to react, the enthalpy of adsorption is higher than the liquefaction one at low coverage but well below the latter when coverage increases. An estimate of the binding energies (BE) related to the various adsorption modes in Figures 5 and 6 comes from the ab initio work carried out by some of As expected, from the lower electronegativity of hydrogen with respect to oxygen, ab initio calculations35 show that the silanol molecule (which does not exist as such, because condensation to disiloxane takes place) is less acidic than the actual SiOH group at the surface. This fact causes a systematic underestimation in the BE relative to a molecule interacting with the silanol itself. The ab initio value of 32.0 kJomol-l (obtained using a double zeta plus polarization functions basis set, inclusive of both electron

Surface Heterogeneity of Silica correlation, evaluated by the MP2 technique, and basis set superposition error) for the adsorption of one water molecule onto an isolated silanol is well below the heat of liquefaction of water (44 kJ-mol-9. When a geminal site involved in the adsorption of one water molecule is considered, owing to its ability to simultaneously act as both H donor and acceptor, the ab initio value of the BE per water molecule rises to 42.5 kJ-mol-l, which is now close to the limit for a hydrophilic site (44 kJ.mol-l).'J4 Recent ab initio calculations have also shown that up to two water molecules may be adsorbed per geminal ~enter.~3 This picture allows the interpretation of data concerning A50-1073, where only few adjacent silanols are present, and heterogeneity is mainly ascribable to the presence of both isolated and geminal species. As to other modes of interaction, the structure in Figure 6 envisaging adjacent hydroxyls shows a binding energy for water of 47.5 kJ-mol-l, calculated ab initio at MINI-1 level inclusive of BSSE correction. Previous work has shown that the hESCF(MINI-l)values correctedfor BSSE are very close to the value obtained at SCF with DZP basis sets, inclusive of BSSE corrections. The dispersive forces (not accountedfor by the SCF method) are currently estimated to be some 25% of the SCF binding energies. If this correction is made to the SCF-BE value, a final value of 60 kJ-mol-l, typical of a hydrophilic interaction, is obtained. It is interesting to note that such binding energy is slightly higher than the s u m of the two separate interactions (23 and 32 kJ.mol-' for water acting toward a single silanol as H acceptor or donor, respectivelym). This indicates that the three species interacting through H bonds undergo an extrastabilization arising from intermolecular charge-polarization effects. It is worthy of note that a similar effect has been evidenced by calculating the BE of chains of interacting silanols, as will be described below. Such a result allows us to understand the behavior of A50-423 and CRIS-1073, which are characterized by a larger hydroxyl population, compared to A50-1073. Adsorption occurs first on the most energetic sites, by two H bonds, followed by adsorption on the hydrophobic part of the surface. The fact that adsorption does actually proceed on the hydrophobic part of the surface also in the presence of hydrophilic patches, where the enthalpy would certainly be higher, is due to entropic factors, the building up of "columns" of water on the few hydrophilic patches being unlikely in the presence of a large part of bare, although hydrophobic,surface. At coverages much above those attained in the present paper, a rise in the heat of adsorption might be found, due to adsorption occurring preferentially on the water adsorbed on the most energetic sites, as previously reported by Zettlemoyer." It has however to be pointed out that the fact that the heat values on plateaus below 44 kJmmol-l are not the same is probably due to the proceeding of adsorption in parallel on both kind of sites: the more extended the hydrophilic patches, the higher the heat of adsorption measured at the plateau. If the surface is extensively hydrated, i.e. mainly made up of silanols close to each other, water may interact with SiOH pairs: the resulting heat values (>44 kJ-mol-l) indicate that the surface is hydrophilic. In this case, the heterogeneity found is due to differences in the location of the pairs of silanols. With two H-bonded silanols, adsorption will only occur on the terminal one: in analogy with what is discussed in the second part of this paper (cooperative effects found in the adsorption of ammonia), it is well possible that the resulting binding energy in this case would be higher than on the isolated SiOH. If the

Langmuir, Vol. 9, No. 10,1993 2717 two silanols are located at an appropriate distance, adsorption of one water molecule may occur with formation of two hydrogen bonds. According to the results in Figure 6, the ideal distance between the two silanols, for binding a water molecule through two H bonds, is 4.9 A. Clearly, the probability of finding this kind of site will vary from one polymorph to the other and from one crystal plane to the other. This may account for the differences in adsorption capacity found for the various polymorphs and, particularly, for the enormous difference between amorphous and crystalline materials, demonstrated in Figure 1and Table I. The regular arrays in crystalline polymorphs result in a possible periodic presence of vicinal silanols at appropriate distance: these will adsorb as many water molecules. These sites will only occur occasionally and randomly on amorphous silicas, particularly on Aerosil50, which is of pyrogenic origin in a very hot flame,16with consequently a much lower surface adsorption capacity toward water.26 Figure 7 reports the expected distances between silicon atoms-and thus between silanols at the open surface-on two of the most commonly exposed crystal planes of a-quartz, (100) and (010). It is easily seen that while on the former many silanols 4.8A apart (close to the optimum calculated value of 4.9 A) might be present, on the latter the sites will be mainly geminal ones. The (100) face will thus be basically hydrophilic and the (010) face hydrophobic. Obviously, an actual powdered sample is made up of several different planes, depending on the way the crystal was converted into a powder, which renders any prediction rather difficult. The data in Figure 1 are thus interpreted as follows: upon heating both crystalline and amorphous materials, pairs of silanols are condensed into a siloxane bridge with loss of one water molecule. The surface is progressively converted from hydrophilic to hydrophobic, with consequent loss both in adsorption capacity and in the enthalpy crystalline silica of adsorption. As already reported,7*n*g0 is much less easy to dehydroxylate than the amorphous material. A50-1073 exhibits indeed a very low adsorption capacity and is fully hydrophobic; CRIS-1073 is only partly hydrophobic, with more than one molecule per nm2 still located on strong hydrophilic patches. The Ammonia-Silica System. The peculiarity of the adsorption of ammonia on silicas, in particular the linear trend of the heat us coverage plot, has already been reported by some of ussand other authors.l8IM The second part of this paper deals with such aspect. Here we discuss the aspects of the data in Figure 2 and Table I concerning the interaction of the probe molecules with definite surface structures. On the fully hydrophobic surface (A50-1073) ammonia does not exhibit the linear trend but the heat us coverage plot exhibits however a marked variation, much larger than with water. Ab initio data in Figure 5 may serve for comparison with the initial value on A50-1073. Because the silanol molecule is less acidic than the actual hydroxyl group, all computed values are somewhat underestimated with respect to the experimental ones on silica. It is however reassuring to note that a difference of about 10 kJ-mol-' between the values of ammonia and water is found in both computed binding energy and molar heat. The most reliable computationalvalue for the adsorption of ammonia is 43.5 kJ-mol-1, which compares favorably with the initial value of about 60 kJ-mol-'. Note that, in contrast to water, the interaction of ammoniawith geminal species is only marginally stronger than that with isolated (36)Cardona-Martinez, N.;Dumesic, J. A. J. Catal. 1991,128, 23.

Fubini et al.

2718 Langmuir, Vol. 9,No. 10,1993 hydroxyls (46.8 kJ-mol-1). The rest of the plot, therefore, has to be probably interpreted as due to the structures that are formed by the irreversible reaction during Ads I (adjacent couples of hydroxyls and sylylamino groups). The effect of thermal treatments (see e.g. Table I, CRIS423 and CRIS-1073) as well as of crystallinity (see e.g. Table I, CRIS-423 and A50-423) on the uptakes are both less marked with ammonia than with the water: again, because of its stronger basicity,ammonia binds onto single silanols where binding of water is less favored. The Alcohol-Silica System: Methanol and tertButyl Alcohol. Although the results obtained for the adsorption of the two alcohols reveal remarkable differences, they are better discussed together because of their chemical similarity. A feature is immediately evident tert-butyl alcohol appears almost insensitive, as far as adsorption capacity is concerned,to crystallinity and also to thermal treatments, as the decrements found on the samples outgassed at 1073 compared with the 423 ones, are relatively small by comparison with ammonia and water. In contrast, the thermal treatment reduces the number of sites adsorbingmethanol even to a larger extent than it does for water, particularly on cristobalite (see Table I, on CRIS-1073less than 25% of the adsorbing sites on CRIS-423were active). As to crystallinity, more methanol is adsorbed on CRIS than on A50 but at a much smaller level than water or ammonia. As long as it is correct to draw a comparison between uptakes of the four adsorbates in the conditions chosen for Table I (i.e., equilibrium pressure in the plateau of heat us coverage plot), the following are apparent: (i) more water and ammonia are adsorbed on cristobalite than methanol or tert-butyl alcohol particularly on the fully hydrated sample (CRIS-423), whereas the situation is somehow reversed on A50 as more alcohol is adsorbed than water or ammonia; (ii) both of these effects are more evident on the samples outgassed at 423 K than on the 1073 K ones. These facts suggest that the sites behaving differently with the two kinds of adsorbates have to be located in the hydrophilic patches. More contradictory information is drawn from the enthalpy of adsorption values. Methanol adsorbs on hydrated or dehydrated samples in different amounts but with the same interaction energy. Inspection of the heat us coverage curves (Figure 3) reveals the presence of few more energetic sites on CRIS-423 than on CRIS-1073, which probably accounts for the small difference found in the molar values; the plateaus, however, are both at 60 kJ-mol-l. On A50 even less differences are found between the hydrated and dehydrated samples, and the final plateaus are somewhat lower (55 kJ-mol-1) than on CRIS. A detailed explanation of all these findings would require more experiments with various alcohols on differently hydrated silicas and goes beyond the scope of this paper; it appears, however, that some sites, regularly distributed on cristobalite, where water acts both as H acceptor and H donator, are not able to bind methanol in a similar way. Conversely, on A50 either similar sites preferentially bind methanol or a methyl/siloxane interaction favors adsorption, as compared to water. The differences in the calculated binding energies (Figure 5 ) on single silanols for water and methanol are more pronounced than those observed as molar values (Table I) or at the plateaus (Figures 1 and 3); this, too, suggests a substantial component arising from dispersion forces in the experimental heat values for methanol, which is accounted for only to some extent in the ab initio calculationsdue to the limited size of the clusters adopted.

a)

n. /POI

b)

% m'*

2

4

6

8

rc/pmol m-*

Figure 8. Differential heat of adsorption of NHa at 303 K on freshly ground pure quartz evacuated at room temperature (ID) and heated at 1573 K (v)(sectiona) on CRIS-423(O), CRIS-773 (A), CRIS-1073 (GI,and CRIS-1573 (v)(section b).

Dispersion forces acting to a much larger extent have to be invoked to explain the results obtained with tertbutyl alcohol. The main feature in this case is the basic insensitivity-both in uptakes and enthalpies of adsorption-to crystallinity and thermal history of the sample, Le., to quantity and distribution of silanols. This suggests an adsorption mechanism whereby the OH group of the alcohol is hindered by the three voluminousmethyl groups, sothat H-bonding is not easy;the molecule is consequently attracted to the surface by dispersion forces, no matter whether it is hydrophilic or hydrophobic. The rather high molar heat found in this case-the highest of the four adsorbates-has to be ascribed to dispersive forces. Detailed inspection of the curves in Figure 4, however, reveals some differences in the isotherm shape for the samples outgassed at 423 K and the 1073K, in very similar fashion for the crystalline and the amorphous sample. This suggests an adsorption governed by H bonding on the hydroxylated surfaces while it will be mainly due to dispersive forces on the dehydroxylated ones. At high coverage, the dispersive forcesprevail and the two isotherm curves merge. Correspondingly,differential heats (Figure 4) on the hydroxylated surfaces are higher than the correspondingones on the dehydroxylated samples. Thus, althoughto amuch lesser extent than the other adsorbates, also tert-butyl alcohol distinguishes hydroxylated and dehydroxylated surfaces. Induced Heterogeneity. The linearity found in the differentialenthalpy curves for the adsorption of ammonia on hydroxylated silicas has been found by us on more than 20 samples differing in crystallinity and in hydroxyl content (refs 7 and 27 and unpublished results). Figure 8 reports as an example the differential heat us coverage curves for a very pure quartz powder (outgassed at room temperature and at 1573K)and for cristobalite (outgassed at 423,773,1073,and 1573K). It is evident that linearity holds as long as the surface remains, at least partially, hydrophilic. As noted above, the extrapolated zerocoverage differential heat decreases with increasing pretreatment temperature. In a previous papere the following model has been proposed. Ammonia is a proton acceptor molecule, which interacts with silanols via the nitrogen lone pair. In contrast to water, ammonia will thus adsorb onto only one silanol, either isolated,geminal, or terminal of SiOH chains. When an ammonia molecule is asdorbed on such terminal silanols, the formation of the 0-He-N bond increases the polarity of the other 0-H bonds, with consequentincrease in the strength of the H-bonding between all interacting SiOH groups. This was hypothesized long ago by Folman

Surface Heterogeneity of Silica

Langmuir, Val. 9, No. 10, 1993 2719 Table 11. Ab Initio Energetic Data for Strings of p Silanols Both as Such and in Interaction with Ammonia P 1 2 3

AE@P

AE@)*

-22.3

4

-83.1

-22.3 -29.2 -31.6

-51.5

AE@NC -30.0 -40.2 -43.4

-45.1

a aE@)= E @ ) - p E(1). AE@) = E @ ) - E @ - 1)-E(1). c AE@N) = E @ M - E ( M - E @ ) . E @ ) =totalenergyforastringofpsilanols; E ( M = total energy for isolated ammonia. E @ N ) = total energy for a string of p silanols interacting with ammonia.

4

I O/H

-1.648

II /"t\l.,,,,

..J.657 -.-O/H

/",,,,

I

I /",,',

I

,, ' ,",

I

3 4N

0.997 H

0/

I

I 0'H..1.718 *._

2

1.737

H,

P

1

Figure 9. Models for silanol (1) and ita oligomers (2, 3, 4). Structures were calculated at SCF-MINI-1 level.

and Yates'937 for one couple of interacting SiOH, but it can apply to the whole cluster of silanols in mutual interaction. The more extended the cluster or chain, the higher the overall enthalpy value measured upon adsorption of NH3: this effect may be regarded as a cooperative one, Le., an aspect of induced heterogeneity. As adsorption proceeds, the clusters are progressively destroyed and the heat of adsorption consequently decreases: the amount of energy involved may be small because rupture of H bonds between SiOH groups occurs. Hence the large span in heat us coverage plots, which are observed to be linear: this feature is related to the inductive effects operating at a highly covalent surface such as the silica one. In order to confirm this model, we have tested the Temkin isotherm38 on our experimental results, on the one hand, considering that it implies a linear decrement of heat of adsorption us coverage: indeed, a very satisfactory agreement is found between experimental isotherms and values calculated by means of the formula N,, = K1 ln(1 K2p) (ref 39). On the other hand, always within a cluster approach envisaging the silanol molecule, we have calculated ab initio the effect of H-bonding between silanols on the acidity of the terminal one, interacting with an ammonia molecule. Results in Figure 9 concern the H-bonding among a row of silanols, which have been given an alternate structure, not really feasible as a model for the silica surface, but avoiding unwanted repulsions between SiH3 groups. It is evident that the intermolecular 0-H-0 distance for the terminal hydroxyl shrinks with increasing number of interacting units from 1.718 to 1.645 A: this is evidence that a strengthening of

+

(37) Folman,M.; Yaks, D. J. C . J. Phys. Chem. 1959,63, 183. (38) Tomkins, F.C. In Chemisorption of gases on metalp, Academic Press: London, 1978; p.13. (39) Garrone, E.; Bolls, V.; Fubini, B.;Morterra, C. Langmuir 1989, 5, 892.

3N

2N

I

-

\

1N

Figure 10. Model for the adsorption of ammonia on silanol oligomers. Structures were calculated at SCF-MINI-1 level.

H bonding occurs, already before interaction with ammonia. Data in Table I1 show the BSSE-corrected energy differences between the various oligomers: these are likely overestimated, as the distances 0-H-0 (one as low as 1.584 A) are definitely too short, due to inadequacy of the MINI-1 basis set. The energy of interaction is roughly proportional to the number of H-bonds, but indeed increases faster than this latter. This is evidence of extrastabilization arising from a cooperative behavior. This is reflected in an increased acidity of the terminal silanol group, as shown by the interaction withammonia, reported in Table I1 and illustrated in Figure 10. Such data give full support to the above mechanism of adsorption. The binding energy has a sudden increase when passing from the isolated to the pair of interacting silanols, and further limited increases are then observed. Such a model also explains the low values of adsorption energy a t high coverages. The strongly basic molecule can open the chain of interacting silanols, according to the schemes depicted in Figure 11: related energy changes are probably small, as a H-bond between silanols has to be broken. Indeed, by the use of the ab initio data it is possible to estimate the energy of adsorption of a second molecule of ammonia in a row of silanols by the cleavage of a silanol/silanol H bond, such as the reactions depicted

Fubini et al.

2720 Langmuir, Vol. 9, No. 10, 1993

borne in mind, however, that in the present calculations, the energy of interaction of ammonia is underestimated (because silanol is not sufficiently acidic), and the interaction between silanols is overestimated,because they are excedingly free to maximize the interaction. Thus, computational results support satisfactorily the adopted model.

-

-

AE(A1) = 2 AE(1N) AE(2N) AE(2) = 2.4

kJ/mole

-

AE(A2) = AE(2N) + AE(l N) - AE(3N) AE(3) = 2.2 kJlmole

Figure 11. Model for NHs adsorption on a chain of interacting silanole with disruption of the chain. Standard energy of reaction A1 and A2 evaluated from the binding energies of the complexes as indicated. For symbols refer to Figures 9 and 10 and Table 11.

in Figure 11. Simple considerations allow calculation of the standard enthalpy of reaction from the values listed in Table 11. The values arrived at indicate even that such processes are endothermic by about 2 kJ-mol-l;it must be

Conclusions Structural heterogeneity is an intrinsic property of a given surface; when, however, adsorption of a probe molecule is used to prove such heterogeneity, the choice of the molecule determines the results obtained. In the present case, the four molecules, all capable of H bonding, showed quite different behaviors, when adsorbed on the same set of solids. Iuduced heterogeneity is not strictly confined to conductor or semiconductor solids; cooperative effectsthrough hydrogen bonds have been made evident by the adsorption of ammonia, which exhibits the same characteristics (linearityin the enthalpy of adsorption us coverage curves) on all silicas, provided that it bears sufficiently extended hydrophilic patches. Acknowledgment. The Italian Consiglio Nazionale delle Ricerche, CNR (Progetto finalizzato “Materiali Speciali per Tecnologie Avanzate”), is acknowledged for financial support. The authors are indebted to Dr. D. Hemenway (University of Vermont) for providing the cristobalite samples.