J. Chem. SOC., Faraday Trans. 1, 1985, 81, 847-855 Nuclear Magnetic Resonance and Dielectric Relaxation Investigations of Water Sorbed by Spherisorb Silica BY PETER G. HALL,* RUTH T. WILLIAMS AND ROBERT C. T. SLADE Department of Chemistry, University of Exeter, Stocker Road, Exeter EX4 4QD Received 13th February, 1984 The Spherisorb silica/sorbed water system has been investigated at coverages 0.018 < O/g,,og~~o, < 0.36 using lH n.m.r. in the temperature range 190 < T/K < 300 and also at coverages 0.013 < O/gHz0 g;to2 < 0.033 using dielectric techniques at ambient tempera- ture. In cases where multilayer adsorption occurs there is a discontinuity in the temperature dependence of the n.m.r. linewidth accompanying a freezing phenomenon. Water molecules making up the first loosely packed adsorbed layer (complete at 0 = 0.06 gHzOg;~oz) remain mobile at low temperatures at allcoverages.Dielectric measurements reveal another discontinuity at a coverage of ca. 0.023 gHzOg;toz. Below that coverage the strength of binding of sorbed water increases with decreasing coverage, while at higher coverages additional water leads to a marked increase in the mobility of the first layer. The properties of water at interfaces have been the subject of considerable interest for many years. Among the various applications, colloid properties and stability, corrosion science and heterogeneous nucleation in the atmosphere are noteworthy. Various types of experimental study have been used to investigate the properties of liquids at interfaces. Most of the early work has been summarised in a review by C1ifford.l We have used infrared,2 dielectric3 and neutron-scattering (inelastic4 and quasielastic5) techniques to investigate the water/silica (or clay3) system.Neutron scattering from adsorbed molecules has been reviewed,s and recently a neutron- diffraction study of water in meso- and micro-pores has been reported.' The general picture which has emerged is that the modification caused by the interface is restricted to a range of the order of 10 A or less. Within this range clear differences from bulk water are observed; for example, a shift in a peak maximum in the inelastic spectrum4 for water/silica corresponds to a reduced degree of hydrogen bonding in the adsorbate. The specific interaction of water with silica manifests itself as lower freezing points and lower diffusion coefficients of the adsorbed water compared with bulk water.The strength of the interaction is a function of both (a) type and extent of porosity and (b) the degree of hydroxylation of the silica surface. In different silicas either (a) or (b) can be the dominant influence, and failure to recognise this has led to apparently contradictory conclusions in previous n.m.r. investigations of silica/water systems.8-11 There is widespread interest in the Spherisorb silica/water system in particular and in this paper we report results obtained using n.m.r. and dielectric techniques. Spherisorb SW silica (marketed by Phase Separations Ltd for h. .l.c. applications) 6-9 and 20pm for S7W and S20W, respectively. From the ratio of H,O to N, B.E.T.surface areas, Ow, of 0.33, Pidduck12 deduced that Spherisorb silicas fall into the category of partially dehydroxylated mesoporous silicas (according to the classification is a well characterised mesoporous silica of mean pore diameter 89 8: and particle sizes 847848 N.M.R. AND DIELECTRIC STUDY OF H,O/SiO, of Baker and Sing13). However, the similarity between the uptakes at saturation of water’, and nitrogen14 (i.e. Gurvitsch behaviour15) indicates that the loose packing in the first layer of sorbed water does not persist through the higher layers. In this work, variations in the lH n.m.r. absorption spectrum linewidth have been measured as a function of both coverage and temperature. The dielectric properties at ambient temperature of the silica/water system having coverages of the order of a monolayer have also been investigated.EXPERIMENTAL N.M.R. MEASUREMENTS Ca. 0.1 g of oven-dried (383 K) Spherisorb S7W contained in an 8 mm ‘ taperlok’ n.m.r. tube (supplied by Fluorochem Ltd) was outgassed to a dynamic vacuum of < mmHg* prior to the adsorption of water vapour. The solid was equilibrated with water from a reservoir containing triply distilled water freed from dissolved gases by three successive freeze-thaw cycles. The effects of fluctuations in the ambient temperature during equilibration with water vapour were minimised by surrounding the sample with a water jacket contained in a Dewar. Equilibration times of the order of two days (following Pidduck12) were followed with a silicone- oil manometer.The concentration of sorbed water was determined with respect to the dry mass of the silica. Spectra were recorded for coverages in the range 0.018-0.36 gHzogi/oz, i.e. 0.86 to 17 monolayers (based on a monolayer coverage, i.e. OHzO = 1, of 0.021 gHzOg&,,),l at temperatures from ambient to 190 K at intervals of 20 K. The reversibility of the process (i.e. recording spectra from 190 K up to ambient) was not investigated. lH n.m.r. absorption spectra were recorded using a JEOL PS/PFT 100 Fourier-transform spectrometer. Pulse lengths were typically 15 ps and the instrumental dwell-time was < 50 ps. Methylene dichloride contained in a glass capillary within the sample provided a trigger signal. No field-frequency lock could be employed and a block-averaging technique was used to overcome any problems associated with magnetic-field drift.Five transients were accumulated per block prior to transformation and twenty blocks in the frequency domain were sufficient to produce satisfactory spectra, i.e. a total of 100 f.i.d. (pulse spacing 1.5 s) were recorded per spectrum. Lineshapes observed were symmetrical and approximately Lorentzian. A y , the linewidth (f.w.h.m.), was measured for each spectrum. DIELECTRIC MEASUREMENTS The main features of the dielectric cell were four 1 mm thick concentric stainless-steel cylinders 6 cm in height and of outside diameter 1.9, 2.5, 3.1 and 3.7 cm, respectively, rigidly held in a polypropylene base. The cylinders formed three capacitors which were connected in parallel (by metal tabs) with a resulting capacitance of 85 pF (measured in uacuo).The cell was encased in an evacuable Pyrex housing having an outer jacket of water. The electrical connections were made via tungsten rods pinch-sealed into the Pyrex housing. All measurements were made at ambient temperatures; sample outgassing was also restricted to this temperature. A capacitance bridge (model 716-C) and guard circuit (model 716-P4), supplied by the General Radio Co., were used for measurements in the frequency range 200 Hz-100 kHz. For measurements at frequencies > 100 kHz, a Q meter (model T2A), frequency range 100 kHz-100 GHz, supplied by Advanced Electronics, was used. The adjustable source was provided by a power supply (model 1203-BQl8) and r.c.oscillator (model 1210-C) capable of frequencies from 20 Hz to 500 kHz. A tuned amplifier and null detector (model 1232A) was used for balancing at the required frequency in the range 20 Hz-100 kHz. All capacitance measurements were made using the substitution method ; no readings could be made using the bridge circuit if the capacitance of the combined cell and dielectric exceeded 1000 pF. The estimated errors for the calculated values of E’ and E” are? E‘ (dielectric constant), f0.007 (capacitance bridge) and & 0,012 (Q meter); E” (dielectric loss), 0.005 (capacitance bridge) and 50.008 (Q meter). * 1 mmHg x 133.3 Pa.P. G. HALL, R. T. WILLIAMS AND R. C. T. SLADE 849 E’ and E” are the real and imaginary parts of the complex relative permittivity of the dielectric material and are given by E’ = T,/To and E“ = &IDx, where Tx is the capacitance of the cell and material, To the capacitance of the empty cell and D, the dissipation factor.The dielectric cell was filled with oven-dried S7W silica (16.8 g) and outgassed at room temperature for 24 h (when a dynamic vacuum of < lop5 mmHg was recorded) prior to the adsorption of water vapour. Water sorption was carried out by (successive) 12 h exposure of the sample to the cooled reservoir followed by two days equilibration time with the reservoir isolated from the sample (equilibration was monitored by recording the change in capacitance of the cell against time). The water coverage, calculated according to the mass lost from the reservoir, ranged from 0.01 3 to 0.033 gHaO g;{oz based on the dried mass of the silica.The frequency scan covered the range 200 Hz-7 MHz. COVERAGE Following Pidduck,12 the coverage of the Spherisorb SW silica surface with sorbed water, OHzO, is, in the present work, based on a monolayer coverage of 0.021 gHnOgg~02, equivalent Thus OHnO = 0.320,,, based on the nitrogen surface area1* of 232m2 g-l, i.e. A concentration of 0.032 gHnO gito, sorbed water is equivalent to a coverage of one water molecule adsorbed per surface hydroxyl group, i.e. O,, = 1, given the surface hydroxyl group concentration of 4.5 nm-2 l6 (and nitrogen surface area of 232 m2 g-l) for Spherisorb SW silica. Thus for coverages of the order of one monolayer (eN2) of sorbed water on Spherisorb SW to OH20 = 1.ON, E 0.065 gH,O giilo,. silica : @HzO z 0.65 00, z 0132 ONz. The importance of these ‘ different monolayer coverages ’ becomes apparent when considering the sub-structure of the loosely packed monolayer (ON,). RESULTS AND DISCUSSION lH N.M.R. OF H20/S7W SILICA The temperature dependence of Av, as a function of temperature and coverage is illustrated in fig. 1. The freezing properties of water sorbed by Spherisorb S20W silica have been investigated in this laboratory using differential scanning calorimetry (d.s.c.).17 S20W differs from the silica in this work only in having a larger particle size. No melting transitions were observed in the temperature range 200-280 K at coverages < 0.06 gHzO gglo0,. At higher coverage, the freezing-point depression varied from ca.26 K (0 = 0.15 gH20 ggtoz) to ca. 12 K (at 0 > 0.2 gHZ0 g;lo2). The quantity of water freezing was equal to the coverage less ca. 0.1 gHzO g;to,. In this work the n.m.r. signal from protons in frozen water (solid) will be too broad to be detected by the available instrumentation, i.e. only mobile water contributes to the observed spectra. The results display a number of interesting and perhaps surprising features. It is evident from fig. 1 that the temperature dependence of Av; is strongly dependent on coverage for 0 > 0.06 gHzO g;lo,. For these coverages there are discontinuities in Avj as the temperature is decreased, and the behaviour at lower temperatures is close to that for lower coverages. A coverage of 0.06 gH,Og&, is equivalent to 0.9 ON,, i.e.approximates to the loosely packed first layer. The observed discontinuities in Av; parallel the freezing properties discussed above and only those water molecules remaining unfrozen contribute to the observed spectra at lower temperatures. The data are consistent with the following model: (1) the first layer of sorbed water is subject to surface ordering effects, and for the coverages investigated (0 2 OHZO) this region of mobile water is dynamically similar in all cases, and (2) at coverages in excess of the first loosely packed layer (at 0 = ON2) the water in higher layers freezes as temperature is decreased, with the first layer remaining diffusionally mobile.850 N.M.R. AND DIELECTRIC STUDY OF H,O/SiO, A 180 200 220 240 260 280 300 TIK Fig. 1. Temperature dependence of Av; for H20/S7W silica: 0, 0.9 OHlo; x , 1.6 OHzO; 0, 3.0 @ ~ * o ; H, 4.6 @H20; A, 4.9 O H 2 0 ; A, 7.6 OHzo; 0, 17.0 OHzo.The similarities in linewidth observed at sub-freezing temperatures for the non- freezing component at high coverages are not surprising. That these similarities should also be observed for lower coverages (where there is insufficient water for freezing to occur) is, however, surprising. Assuming a fast-exchange model at high temperatures (above freezing), and that the relaxation rate appropriate to the first layer is independent of the presence (or otherwise) of additional layers, the minimum observable ratio of the linewidths at high and low coverages would be nfirst/na?,, where ni is the population of layers i. The data in fig.1 are not consistent with this model (e.g. for 0 = 17.0 OHtO the minimum would be 3/17 = 0.18, whereas the observed ratio is ca. 0.14). It is, therefore, apparent that the presence of multilayers does alter the mobility of the first layer. The observation of Lorentzian absorption lineshapes in this work is consistent with rapid exchange of mobile waters, every mobile molecule experiencing all chemical environments open to it in a time shorter than the relaxation time (T,) characteristic of all such environments, as has been found by Pearson and Derbyshire18 in related studies. Under fast exchange conditions, inhomogeneous broadenings (such as have been found in other systems because of distributions of chemical ~ h i f t s l ~ , ~ ~ ) do not contribute to the linewidth.The observed n.m.r. spectrum is thus an average for all environments open to mobile waters within the pores, i.e. the linewidth depends onP. G. HALL, R. T. WILLIAMS AND R. C. T. SLADE 851 f” 1 0 \ 2 3 4 5 6 7 log,, WHz) Fig. 2. Dielectric loss of H20/S7W silica at room temperature: 0, 0.62 OHzo; 0, 0.71 OHzO; A, 0.81 OH20; X , 1.10 OHzO; 0, 1.60 OHzO. the relative populations of the mobile regions. Waters in layers other than the first will, at temperatures above the freezing transition, be less affected by surface-ordering effects and hence more mobile than those in the first layer. Following Pearson and Derbyshire18 it can be predicted that, as mesopores fill with adsorbate from their walls inwards, at high temperatures Av; should decrease with increasing coverage for multilayer adsorption.The results in this work are consistent with this prediction. The n.m.r. measurements in this work reveal information for coverages 0 b OHzO, while the dielectric measurements extend to lower coverages and reveal structure within the first sorbed water layer. DIELECTRIC PROPERTIES OF H20/S7W SILICA The dielectric measurements refer to ambient temperature and to coverages less than those studied using n.m.r. and reveal details within the first sorbed water layer. The Debye21 equations relate the real and imaginary parts of the complex permittivity, E’ and E ” , to the macroscopic relaxation time, z, of an applied alternating freyuencyf= o/2n. The variation of E’ and E” with logfshows dielectric dispersion and the maximum of the dielectric loss; the latter corresponding to an absorption of energy in the dielectric due to phase loss between the polarisation and applied field.For systems where a distribution of relaxation times exist, the plot of E” against E’852 10 9 8 7 6 E’ 5 4 3 2 1 0 N.M.R. AND DIELECTRIC STUDY OF H,O/SiO, I 2 3 4 5 6 7 log,, CfIHz) Fig. 3. Dielectric constant of H20/S7W silica at room temperature: symbols as in fig. 2. (Cole-Cole plot,,) results in an arc, in contrast to the semi-circle obtained for a single relaxation time. In the present work the results of the frequency scans are shown in fig. 2 and 3. The frequency scans showed a dielectric-absorption peak, a low-frequency loss, a high-frequency loss and dielectric dispersion. The dielectric-absorption peak (fig.2) was not apparent until a coverage of 0.015 gHzog~~oz, i.e. OHZ0 of 0.71 (based on the H,O monolayer amount of 0.021 gH,Ogg~o,) or ON, of 0.24 (OH,, = 0.33 ON,). The position of this peak was a function of water-vapour coverage, moving to higher frequencies with increasing coverage up to of 1.1, i.e. 0.023 gHPO gktoz. The height of the maxima remained relatively constant at values of E” = 1.3-1.4. This is in agreement with the work of Hall and Rose3 on water/clay systems (although values of E” for the latter were smaller and of the order of 0.10). The variation of E’ with logf(fig. 3) showed dielectric dispersion, i.e. a decrease in E’ with increasing frequency. A low-frequency loss was observed at coverages > 0.6 Or,o.. This was attributed to direct-current conductance across the terminals ; the auto-ionisation of the adsorbed water provides a conduction path. A small high-frequency loss was observed which increased with increasing coverage; this is probably due to inaccuracies resulting from the poor sensitivity of the Q meter in the range 6.0-7.0 MHz.At coverages > 1.1 the low-frequency loss overwhelmed the dielectric-absorption peak; thus no attempt was made to investigate coverages > 1.6 OHZ0 for this system.P. G. HALL, R. T. WILLIAMS AND R. C. T. SLADE 853 Table 1. Characteristic frequency of water sorbed by Spherisorb silica coverage 0.62 0.013 - - 0.71 0.01 5 900 2.23 0.8 1 0.0 17 3 5-40 3.65 k0.05 1.1 0.023 3.5-4.5 4.65 -t- 0.05 1.6 0.033 ice (298 K)a 3 4.72 - - a rice (298 K) is the value obtained by extrapolating to 298 K, the linear relationship between log z and l/T(in the temperature range 207.2-272.9 K) for ice as obtained by Auty and A broad peak in the E” against logfplot indicates a spread of relaxation times.However, the mean value of a symmetric distribution was determined from the value of the frequency,f,.,,,, corresponding to the peak maximum, according toychar = 1 /2m. The results are summarised in table 1 . From table 1 it can be seen that the z values are very large at values of O H z O < 0.7. At 0.8 OHzO the relaxation time is of the same order as that expected of ice at 298 K.23 This suggests that at coverages of OHzO < 0.8, i.e. < 0.017 gHzO g;loz, the adsorbed water is very tightly bound to the silica surface.As more water molecules adsorb, the adsorbate z becomes ice-like at a coverage of 0.023 gH,O g;toz, i.e. 1.1 QHzO. Although first apparent at 0.7 OHzO (0.015 gHpO ggto,), the conduction effect increases until it overwhelms the system at 1.6 OHZ0 (0.033 gHBO ggto,). This suggests that the mobility within the adsorbed phase becomes significant after the majority of hydroxyl adsorption sites on the surface have each become occupied by a single water molecule, i.e. at a coverage of 0.021-0.032 g,zog&O, (aHzO to OoH). Only for coverages of 0.015 and 0.017 gHzO ggto2 do the E’ against E” plots (fig. 4) approximate to Cole-Cole behaviour. The arcs apparent in these cases are indicative of distributions of relaxation times for adsorbed water. This is no surprise as the adsorbent surface is energetically heterogeneous.At higher coverages no such behaviour was observed. This, and the deviation from Cole-Cole behaviour at low frequency in the plot for 0.017 gH20g&oz coverage, is likely to be caused by conductance effects. ’The hydroxyl-group concentration on the surface determined by Gray16 was ca. 4.6 nm-2; ca. 1.2 nmW2 were attributed to be isolated and ca. 3.5 nmP2 to be hydrogen bonded. Assuming that, initially, one water molecule adsorbs per OH site, then the water coverage corresponding to each of the two OH sites is: OOH (isolated) = 0.008 gHzO gg!oo, and OoH (H bonded) = 0.024 gH,o g;;oo,. In the present work coverages in the range 0.008 gHZOg~{oz were not investigated. However, the dielectric measurements indicate that, at a coverage of 0.023 gHzO ggto2, the adsorbate has a well ordered structure, having a relaxation time similar to that expected of ice (at 298 K).At coverages < 0.023 gH20gg~oz the adsorbed phase appears to be more strongly bound, as indicated by the long relaxation times compared with ice (298 K). At a coverage > 0.023 gH20g&& a significant increase in the conductance of the system suggests increased mobility within the adsorbed phase. No dielectric relaxation time could be obtained for a coverage of 0.013 gHpO gglo,, probably because z was too long, i.e. z > 1.5 ms. The largest value of z = 0.9 ms (at8 54 3 - “i N.M.R. AND DIELECTRIC STUDY OF H,O/SiO, 0 1 2 3 4 5 6 7 E’ 0 ? 2 3 4 5 6 7 E l Fig. 4. Cole-Cole plots of H,O/S7W silica at room temperature: (a) 0.7 OHzO and (b) 0.8 OHzO.0.015 gH,og;:o,) compared with z = 35 ps (at 0.017 gH,Og&) suggests that the adsorbate becomes less strongly bound as the adsorbed layer increases. These results agree with the findings of Hall et a1.495 from inelastic neutron-scattering and time-of-flight measurements. On the basis of a simple theoretical treatment of the results, they concluded that the Spherisorb results were consistent with the formation of two-dimensional surface clusters of sorbed water containing a predominance of doubly hydrogen-bonded molecules. CONCLUSIONS The n.m.r. measurements have revealed details of the behaviour of sorbed water at coverages in the range 0.018-0.36 gHzO gg:oz at temperatures above and below those leading to freezing of water in the pores for samples with multilayer adsorption.The n.m.r. results reveal differences in behaviour for the first layer of water and higher layers. Water forming the first layer (complete and loosely packed at 0 = 0.06 gH20g;~oa z ON2) remains diffusionally mobile at low temperatures, while any higher layers freeze on cooling. The dielectric measurements mostly refer to coverages lower than those in the n.m.r. study and reveal structure within the first sorbed layer. Measurements concerned coverages < ON* but up to OoH (0.033 gH,og&). At coverages < 0.023 gH,Og~~oO,, decreasing coverage leads to more strongly bound water. A coverage of 0.023 gH20 g&,P. G. HALL, R. T. WILLIAMS AND R. C. T. SLADE 855 corresponds to OHZ0, and in that case the relaxation time observed is of the same order as that anticipated for ice at 298 K.Water sorbed in excess of OHZO leads to increased mobility within the adsorbed layer before the first loosely packed layer is completed (0.06 gHZ0g;tOp z ON?). In cases where Cole-Cole behaviour can be observed, a distribution of relaxation times is evident. The results described in this paper may be compared with recent quasi-elastic neutron-scattering rneasurement~,~~ which show the coexistence of two phases of sorbed water molecules with different dynamics in monolayer and near-monolayer films on the surface of Spherisorb silica. One component is immobile on the experimental timescale (< 10-lo s) and the other gives quasielastic broadening, which is fitted by a model of consecutive uniaxial rotation and two-dimensional jump translation.We thank Dr V. Sik for assistance with n.m.r. absorption spectra and the referees for their constructive comments. J. Clifford, in Water: A Comprehensive Treatise, ed. F. Franks (Plenum Press, New York, 1975), vol. 5. P. G. Hall and P. B. Barraclough, J. Chem. Soc., Faraday Trans. I , 1978, 74, 1360. P. G. Hall and M. A. Rose, J. Chem. Soc., Faraday Trans. I , 1978, 74, 1221. P. G. Hall, A. Pidduck and C. J. Wright, J. Colloid Interface Sci., 1981, 79, 339. P. G. Hall, A. J. Leadbetter, A. Pidduck and C. J. Wright, Neutron Inelastic Scattering 1977 (Interna- tional Atomic Energy Agency, Vienna, 1978), vol. 11, p. 5 l l . P. G. Hall and C. J. Wright, Surface and Defect Properties of Solids (The Chemical Society, London, 1978), vol. 7, p. 501. G. K. Rennie and J. Clifford, J. Chem. SOC., Faraday Trans. I, 1977, 73, 680. P. A. Sermon, J. Chem. Soc., Faraday Trans. I , 1980,76, 885. 25, p. 174. V. V. Morariu and R. Mills, Z. Phys. Chem. (Frankfurt am Main), 1972, 79, 1. A. J. Pidduck, Ph.D. Thesis (University of Exeter, 1980). ' D. C. Steytler, J. C. Dore and C. J. Wright, Mol. Phys., 1983, 48, 1031. I" J. Clifford and S. M. A. Lecchini, in Wetting (Society of Chemical Industry, London, 1967), monogr. l 3 F. S. Baker and K. S. W. Sing, J. Colloid Interface Sci., 1976, 55, 605. l4 M. J. Holdaway, AERE Harwell Report no. AERE-M2749 (Atomic Energy Research Establishment, l5 S. J. Gregg and K. S. W. Sing, Adsorption, Surface Area and Porosity (Academic Press, New York, l6 R. A. C. Gray, Ph.D. Thesis (U.M.I.S.T., 1976). l 7 J. Clare, B.Sc. Research Project (University of Exeter, 1978). l9 D. Geschke, Z. Naturforsch., Teil A, 1968, 23, 339. 2o J. H. Pickett and L. B. Rogers, Sep. Sci., 1970, 5(1), 11. 21 P. Debye, Polar Molecules (Chemical Catalog Co., New York, 1929). :12 K. S. Cole and R. H. Cole, J. Chem. Phys., 1941, 9, 341. 24 J. Clark, P. G. Hall, A. J. Pidduck and C. J. Wright, J. Chem. SOC., Faraday Trans. 2, in press. Harwell, 1976). 2nd edn, 1982), p. 113. R. T. Pearson and W. Derbyshire, J. Colloid Interface Sci., 1974, 46, 232. R. P. Auty and R. H. Cole, J. Chem. Phys., 1952, 20, 1309. (PAPER 4/252)