J. Phys. Chem. 1981, 85, 3628-3635
3628
For small xi ( x i = xl for R1 IR2, xi = x2 for R2 IR J the solution is xi(xj) = [-ki(r)[Rj(r) + xi + Ri(r) cos e] + kj(r)xj + ((ki(r)[Rj(r)+ xj + Rl(r) cos e] - kj(r)xjI2- 4ki(r) x kj(r)xj(Ri(r) + [Rj(r) + xj cos e]) cos e)1/2]/(2ki(r)cos 8 ) (A71 One finds the solution of eq A1 by numerically solving eq A2 for xi, with xj calculated from eq A7, i.e.
d E[xi,xj(xi)]/dxi = 0
(A8)
Since xi is small compared to Ri(r), one may simplify eq A5 to
Hence
Thus, one may alternatively find approximate solutions to eq 14 by solving numerically the sum of eq A2 (i = 1, 2) for xi, with x calculated from eq A10. Numerical testa of both ways of!solving eq 14 show that, for H + H2 and F + H2reactive systems (0 = O”, 30°,60°,90°, and 120°), the approximate relation, eq A10, gives practically identical solutions (up to three decimal figures).
Adsorption of Hydrocarbons on Silica-Supported Water Surfaces Gllles M. Dorrls and Derek G. Gray” Pulp and Paper Research Institute of Canada and Depatfment of Chemlstty, McGlll Unlvers#y,T Montreal, Quebec H3A 2A7, Canada (Received: March 24, 798 7; In Final Form: June 30, 798 1)
A gas-chromatographic technique was used to measure the adsorption from the vapor phase of a series of n-alkanes, at zero and finite coverages on the surface of water-coated silica having a uniform large pore size. The presence of one to two water monolayers preadsorbed on the glass was sufficient to reduce considerably the London force field of the glass. At water loadings greater than 3.8% by weight, the effect of the solid on the adsorption of n-alkanes appears to vanish, and the surface properties reach a steady state, as revealed by adsorption isotherms and by the differential heats of adsorption at zero coverage. Contrary to some results in the literature, the heats of adsorption of n-alkanes are found to be smaller than the heats of vaporization. Thus, the model postulating water surface restructuring by hydrocarbon molecules is not supported by the GC data. Thermodynamic functions agree well with some other reported data, but the surface excesses and surface pressures for the n-alkanes are larger in the present study. The surface of the water-coated glass has a London component of the surface free energy of 23 mN m-l, compared to 22 mN m-l for bulk water.
Introduction The adsorption of hydrocarbons has been measured by gas chromatography at zero and finite coverages on the surface of water-swollen cellulose’ and film.2 The surface properties did not approach those of pure water at the highest accessible water contents, probably because cellulose absorbs water to form a water-swollen gel. Because the properties of water at the liquid-vapor interface are of great interest in various contexts, our previous studies are here extended to a simpler system. The adsorption of hydrocarbon vapors is measured on a nonswelling macroporous glass/water system, which at high water contents should provide a surface approaching that of pure water. The adsorption of a number of hydrocarbon vapors on water surfaces has been studied by measuring the change in surface tensions with partial pressure of organic vap o r ~ . ~ - ’ Adsorbates ~ such as n-alkanes, branched alkane~,*~**~’ and aromatic hydrocarbon^^^^ invariably yield type-I11 isotherms, indicating that water acts as a lowenergy surface toward nonpolar vapors.13 This conclusion is also consistent with the high interfacial tensions and small works of adhesion at the liquid hydrocarbon-water interfaces. According to Fowkes,14J5the work of adhesion at a hydrocarbon-water interface is governed mostly by London force interactions so that the active part of the 3420 University Street. 0022-3654/81/2085-3628$01.25/0
surface tension of water (Le., its London component) is 22 mN m-’ at 20 “C. Adamson et al.16-20have questioned this apparently (1)Dorris, G.M.; Gray, D. G. J. Chem. SOC.,Faraday Trans. 1 1981, 77,713,725. (2)Katz, S.; Gray, D. G. J. Colloid Interface Sci. 1981,82,339. (3)Hayes, E. I.; Dean, R. B. J. Phys. Chem. 1953,57,80. (4)Cutting, C. L.;Jones, D. C. J. Chem. SOC.1955,4067. (5)Jones, D. C.; Ottewill, R. H. J. Chem. SOC.1955,4076. (6)Jones, D. C.; Ottewill, R. H.; Chater, A. P. J. h o c . Int. Congr. Surf. Act., 2nd, 1957 1957,1, 199. (7)Blank, M.; Ottewill, R. H. J. Phys. Chem. 1964,68, 2206. (8) Hauxwell, F.; Ottewill, R. H. J. Colloid Interface Sci. 1968,28,514. (9)Hauxwell, F.; Ottewill, R. H. J. Colloid Interface Sci. 1970,34,473. (10)Massoudi, R.;King, A. D., Jr. J. Phys. Chem. 1974, 78, 2262. (11)Massoudi, R.; King, A. D., Jr. In “Colloid and Interface Science”; Kerker, M., Ed.; Academic Press: New York, 1976; Vol. 3,p 331. (12)Jho, C.; Nealon, D.; Shogbola, S.; King, A. D., Jr. J. Colloid Interface Sci. 1978,65,141. (13)Vidal-Madjar, C.; Guiochon, G.; Karger, B. L. J. Phys. Chem. 1976,80,394. (14)Fowkes, F. M. Ind. Eng. Chem. 1964,56,40. (15)Fowkes, F. M. In “Chemistry and Physics of Interfaces 11”;Ross, S., Ed.; ACS Publications: Washington, DC, 1971;p 153. (16)Adamson, A. W.; Dormant, L. M.; Orem, M. J.Colloid Interface Sci. 1967,25,206. (17)Orem, M. W.; Adamson, A. W. J. Colloid Interface Sci. 1969,31, 278.
(18)Adamson, A. W.; Shirley, F. P.; Kunichika, K. T. J. Colloid I n terface Sci. 1970,34,461. (19)Adamson, A. W.; Orem, M. W. In “Progress in Surface and Membrane Science”; Cadenhead, D. A., Danielli, J. F., Rosenberg, M. D., Eds.; Academic Press: New York, 1974;Vol. 8, p 285.
0 1981 American Chemical Society
Adsorption of Hydrocarbons on Water Surfaces
simple behavior at hydrocarbon-water interfaces. From adsorption studies on ice,16J7it was suggested that below -35 " C ice is inert toward nitrogen and n-alkanes and behaves as a normal low-energy surface. However, above this transition temperature, the surface of ice resembles liquid water and is restructured by some adsorbates such as n-alkane~.l'-~~ The main evidence was that the isosteric heat of adsorption of n-hexane at low surface coverages was 43-16 kJ mol-l greater than the heat of vaporization. This was unexpected since the adsorption isotherms of n-hexane were of type I11 and the surface of ice was believed t~ be free of heterogeneities. The effect was ascribed to a rearrangement of surface water molecules to accommodate the hydrocarbon molecule, leading to larger heats of adsorption.lg Other evidence for perturbation of the surface was reported by Ottewil121for the adsorption of n-hexane on water and by Hauxwell and Ottewil18for the adsorption of toluene on water where unexpectedly high heats of adsorption at low coverages were also observed. On the other hand, in some other studies, the anomaly in the heat of adsorption at close to zero coverage is ab~ent.~JJ& Recently, l~ Guiochon et al.13have calculated the adsorption potential of hydrocarbons at the gas-liquid interface of water. The computed values were low because of the nonspecific character of the interactions and the low density of attractive force centers. They also treated the case where the hydrocarbons are surrounded by water molecules and found that a surface-perturbation model cannot be used to explain the high heats of adsorption observed for n-hexane and toluene on the surface of water and n-hexane on ice. Thus, despite these interesting observations on water restructuring, the evidence is sparse and not totally convincing. Part of the problem resides in the difficulty of getting precise heats of adsorption at very low coverages. In addition to the very small adsorption of n-alkanes on ice at low relative vapor pressures, the surface area of ice is both time2zand temperature16v2zdependent, particularly below -35 "C. On liquid water, adsorption studies are difficult, because of the very small decrease in surface tensions caused by hydrocarbons at low pressures and because of the volatility of water. This led Karger et al.23-26 to propose a chromatographic technique as an alternative method to measure the adsorption of hydrocarbons on water. The main advantage of the chromatographic technique over conventional methods in adsorption studies is that significantly lower coverages may be attained. Because of the very high sensitivity of flame ionization detectors, injections may be made so small that adsorption takes place in the linear part of the isotherm (i.e., in the Henry's law region). It was f o ~ n d ~that, ~ * ~for~ av large ~ ~ number of saturated and aromatic hydrocarbons, the heats of adsorption at zero surface coverage were always smaller than the heats of vaporization, which obviously conflicts with Adamson's model of water surface perturbation. Admittedly, the chromatographic approach is not free from problems. Unlike surface-pressure measurements, (20) Adamson, A. W. "Physical Chemistry of Surfaces", 3rd ed.; Interscience: New York, 1976; p 108. (21) Ottewill, R. H. Doctoral Thesis, University of London (Queen Mary College), London, England, 1951. (22) Jellinek, H. H. G.; Ibrahim, S. H. J. Colloid Interface Sci. 1967, 25. 245. (23) Karger, B. L.; Sewell, P. A.; Castells, R. C.; Hartkopf, A. J. Colloid Interface Sci. 1971, 35, 328. (24) Karger, B. L.; Castells, R. C.; Sewell, P. A,; Hartkopf, A. J . Phys. Chem. 1971, 75, 3870. (25) King, J. W.; Chatterjee, A.; Karger, B. L. J.Phys. Chem. 1972, 76. 27fi9. (26) Hartkopf, A.; Karger, B. L. Acc. Chern. Res. 1973,6, 209. . - I
The Journal of Physical Chemistry, Vol. 85,No. 24, 1981 3629
the surface of adsorption is not planar since the water is coated on a macroporous solid. Moreover, the solid may exert an ordering effect on water, and thus the adsorption of hydrocarbons may take place on a modified water surface. Other potential complications include secondary retention processes in the column, the volatility of water in the column, and the influence of the carrier gas. Apparently, none of these factors influenced the r e ~ u l t s . ~ ~ * ~ @ A major discrepancy between the results of surface pressure and chromatographic measurements is worth mentioning. The chromatographic results25for the surface excess of n-alkanes on water are about twice as large as those from surface-pressure measurements. Since in the GC experiments the surface area was estimated and not measured, Karger and Hartkopfz6ascribed the difference either to a low estimate of the surface area or to inaccuracies in the surface-tension measurements. More recently, Okamura and Sawyerz7 have determined Henry's law constants of n-alkanes on the surface of water coated on silica and Chromosorb W. The surface area at any water loading was measured from a dynamic nitrogen adsorption technique by freezing the column packing. With this approach, the Henry's law constants are larger than in the previous GC work and thus are in even greater disagreement with the surface-tension measurements. In this work we first investigate whether the surface mea in the chromatographic data is responsible for the disagreement between the static and chromatographic experiments. The adsorption of n-alkanes at zero and finite coverages is then measured. The thermodynamic functions of adsorption of a series of n-alkanes on the water-coated silica are determined to see whether the model of surface perturbation holds. Experimental Section Apparatus and Solid Support. The chromatography and the modified nitrogen carrier gas line have been described in detail previously.1p28 The temperature of the column was maintained to f0.03 O C by a water circulating bath which replaced the air oven of the chromatograph. A 0.6-m long glass column with an inside diameter of 4 mm was used in this study. The empty column was first washed with chromic acid and was extensively rinsed with water doubly distilled from alkaline potassium permanganate. The support material was CPG-10 controlled-pore glass and was supplied by Electro-Nucleonics, Inc. The glass sample is nearly pure silica with traces of B2O3 and NazO. The following physical properties were given by the manufacturer: mesh size, 120/200; mean pore diameter, 142.2 nm with a narrow pore distribution (A 4.5%); pore volume, 1.02 X lo4 m38-l; surface area, -17.5 mz g-l. The material as received was soaked in hot concentrated nitric acid for 12 h. During the washing, the glass was stirred occasionally to remove any gas bubbles. Deionized water free from organic contaminants was used to rinse the glass sample until neutral pH was reached for the water rinses. The glass sample was next dried for 24 h at 95 "C, followed by a heating period of 3 h at 205 "C. The CPG-10 sample (2.880 g) was then packed in the column. The BET surface area of the acid-washed silica, determined from nitrogen adsorption, was found to be 18.9 m2 g-l. Water Loadings. In this work, five different water loadings were studied: 0.6%, 1.04%, 3.80%, 21% and 47% by weight water on dry silica. The three low water contents were achieved by equilibrating the glass with the carrier (27) Okamura, J. P.; Sawyer, D. T. Anal. Chem. 1973,45, 80. (28) Dorris, G. M.; Gray, D. G. J. Colloid Interface Sci. 1979, 71,93.
3630
The Journal of Physical Chemistry, Vol. 85, No. 24, 1981
gas at three different partial pressures of water. The desired relative vapor pressure or relative humidity was attained by passing the carrier gas in a water saturator a t a controlled temperature. It was shown previously1that the relative humidity is governed not only by the temperature of the saturator and of the column but also by the pressure drop across the column. The packing causes a resistance to the gas flow, resulting in an inlet pressure larger than column outlet pressure, and thus in relative humidity larger at the inlet than at the outlet of the column. The magnitude of this effect has been treated previous1y.l In this work, at low water loadings, the inlet pressure was 1.25 larger than the outlet (= atmospheric) for a measured flow rate of 20 cm3min-'. Experimentally, the three lower water loadings correspond to average relative humidities of 25.7%, 62.3%, and 88%. The loadings were determined after the experimental runs by passing a dry nitrogen carrier gas at a column temperature of 150 "C and collecting the water in a U-tube of molecular sieves connected to the column outlet. Thus, the water layer thickness on the glass varies with the position in the column but the system is an equilibrium one, and, for a given temperature and flow rate, the water loading will remain the same from day to day. Higher water contents on the glass of -21% and -47% by weight were obtained by injecting directly liquid water at each end of the column. The system was equilibrated afterward for a period of a few days under a water-saturated carrier gas. Again, because of the back-pressure effect, the water tends to redistribute slowly on the solid. A set of measurements must be made in a relatively short period (1day) to prevent this problem. It will be shown later that this water redistribution does not play a role in the retention, provided that the surface properties are independent of the water layer thickness. Solutes. All of the solutes used in this study were analytical grades with a purity greater than 99% for all alkanes. They were obtained from Polyscience Corp. and were used as received. Measurements. Finite Concentration. Karger et al.24925 have shown that saturated hydrocarbons are retained solely by adsorption on the surface of water-coated solids. Thus, the net retention volume, VN, required to elute a solute, whose partial pressure in the gas phase is p (assumed to behave ideally), is given by VN = RT(dF/dp)A = RT(dq/dp)w (1) where R is the gas constant, T the absolute temperature, r is the surface concentration expressed per unit area, and A is the total area of the water film. If the surface area is not known, q may replace r in eq 1and represents the number of moles adsorbed per gram of packing, and w is the total weight of packing in the column. Thus, VNdepends on the gradient of the distribution isotherm, and conversely the adsorption isotherm may be calculated from eq 1 by integration, yielding
Dorris and Gray
-
The variation of VN with p is obtained from different injection sizes of the solute. The height of the peak may easily be converted to p by calibration of the detector sensitivity.l In this work the peak-maxima method1Ia was used to generate the chromatographic envelope 03 vs. V,) to be integrated graphically by eq 2. The correct calculation of VN when the compressible nitrogen carrier gas is water saturated and the flow measuring device at the column outlet is a soap-bubble meter has been reported previously.'
V&m3 Figure 1. Chromatographic peaks at 15 O C of n-heptane on silica exposed to 25.7% relative humidity (RH) (0.6% by weight water loading). The circles indicate the peak maxima; the GC envelope is obtained by joining the peak maxima (dashed line).
Zero Coverage. When Henry's law holds for the adsorption process, the gradient d r / d p in eq 1 becomes a simple ratio r/p, and eq 1 may now be expressed as KSA (3) where Ks is the Henry's law constant for the partition of the solute between the surface and the gas phase. Equation 3 is applicable when symmetrical peaks are obtained and when the retention volume is independent of sample size. Moreover, from Gibbs' adsorption equation29 VN =
(I'RT/P)r-o = (T/P)r-.o
(4)
so that, from eq 1, 3, and 4 Ks = T/P
(5)
where T is the surface pressure and represents the reduction of surface tension arising from the adsorption of the solute at a pressure p. From Ks or V,, determined at various temperatures, the differential enthalpy of adsorption at zero coverage, AHA, may be calculated from eq 6. The standard free energy
of transferring a mole of adsorbate vapor from an ideal gas state to a standard state on the surface is given by eq 7. AGA = -RT In (BKs) (7) When De Boer's definition of the surface pressure in the adsorbed standard state is adopted, B = 2.99 X lo8 m-laB The entropy of adsorption can finally be obtained from the following well-known expression: AGA = AHA- T MA.
Results and Discussion Adsorption of n-Alkanes at Finite and Zero Coverages. Adsorption isotherms at 15 "C of n-heptane on silica at 0.6%, 1.04%, and 3.80% water loadings were determined. As mentioned previously, the low loadings of 0.6%, 1.04%, and 3.80% correspond to average relative humidities of 25.7%, 62.3%)and 88.0%, respectively. In Figures 1-3a, chromatograms of n-heptane at the three low water con(29) Dorris, G. M.; Gray, D.G. J. Colloid Interface Sci. 1980, 77,353.
The Journal of Physical Chemistry, Vol. 85,No. 24, 1981 3631
Adsorption of Hydrocarbons on Water Surfaces
25.7% R.H. -‘01.5- 2.1.3. 62.3% R.H. 88.0% R.H.
-
x
-
0.1
Flgure 2. Chromatographic peaks at 15 ‘C of n-heptane on silica exposed to 62.3% relative humidity (1.04% by weight water loading).
VN/crn3
vN/cm3
Flgure 3. Chromatographic peaks at 15 ‘C of n-heptane on silica exposed to 88% relative humidity (3.80% by weight water loading): (a) finite concentration; (b) Henry’s law region, which corresponds to extremely small injection sizes.
tents are presented. In all cases, there is an excellent coincidence of the leading edges of the large asymmetrical peaks, reflecting equilibrium at all partial pressures. Moreover, the rear boundaries are vertical until close to the base line, indicating that band broadening due to kinetic effects is minoram In Figure 3b, peaks obtained from extremely small injections of vapor mol) are presented. The peaks are characterized by a good symmetry with VN independent of sample size, indicating that the Henry’s law region is indeed reached. At all water loadings and for all solutes, similar peak shapes for infinitely small injections were observed. Note that in Figures 1-3a the points for VN when p 0 are in the Henry’s law region. The shape of the chromatographic envelopes, obtained by joining the loci of the peak maxima, changes with the relative humidity (or the moisture content). In Figures 1and 2, VN first decreases at low pressures while, at intermediate and large p, V, increases continuously. On the other hand, at 88% relative humidity (Figure 3a), except in the Henry’s law region, V, increases with p. This last behavior is typical of retention governed by a type-I11 isotherm. At larger moisture contents on the glass and for other solutes, the GC envelopes are similar to n-C, at 88% relative humidity. By integrating the envelope according to eq 2, one may calculate a smooth isotherm. The method has been de-
-
(30) Conder, J. R.;Young,C. L. “Physico-ChemicalMeasurements by Gas Chromatography”;Wiley-Interscience: Chichester, 1979; Chapter 9.
Q2
0.3
0.4
0.5
Figure 4. Adsorption isotherms (qvs. p I p o ) at 15 ‘C for n-heptane on silica exposed to 25.7% relative humidity (0.6% water loading), 62.3% relative humidity (1.04% water loading), and 88.0% relative humidity (3.80 % water loading).
scribed but it should be noted that eq 2 takes into account only the gas compressibility factor but neglects Conder’s correction^^^ for the sorption effect and for gas-phase imperfections. Of the two corrections, the sorption effect is the most important. However, for the solutes of low volatility used here, this correction term is less than 2% at the largest partial pressure and is much less at lower p . In Figure 4 the adsorption isotherms at 15 “C, expressed as q (mol g-l) vs. p / p o ,for n-heptane at 25.7%, 62.3%,and 88% relative humidities are reported. Values of p o were calculated from Antoine’s coefficients listed in the API ~ o m p i l a t i o n . ~The ~ reproducibility of an isotherm at a given relative humidity was very good, the values of q at any p / p o varying by less than 2% from the mean. The three isotherms of n-C7 are typical of adsorption taking place on low-energy surfaces. As the relative humidity (or loading) increases, the adsorption decreases; water adsorbed on the surface must lower the London force field of the silica. Except for the highest humidity, the isotherms possess a certain type-I1 character as indicated by the slight concavity at low relative vapor pressures. (Conder and Young30 point out that the elution method is inaccurate for isotherms with a point of inflection, but the error should be minor for the very slight effect observed here.) The determination of the BET monolayer coverages from the isotherms was next attempted. The BET plot of n-heptane at 25.7% relative humidity was not linear over the usual range of applicability of the BET equation. From the small range where the plot was reasonably linear (0.18 p / p o < 0.28), the monolayer capacity, qm, was determined. A value of 2.52 pmol g-’ was found and, taking the area of n-C, to be 0.66 nm2 per molecule,l this yields a surface area of 10.0 m2 g-l. As will be shown later, this value is too low. (Capillary condensation of water in the large pores is absent at this low relative humidity, and thus the surface area should be nearly identical with the surface area of the dry samples.) At the two other relative humidities, the BET plots were also found to be curved and the surface areas even more unreasonable. It is not known why the adsorption isotherms of n-alkanes on moist cel(31) Conder, J. R. Chrornatographia 1974, 7,387. (32) “Selected Values of Properties of Hydrocarbons and Related Compounds”; American Petroleum Institute Research Project 44, Thermodynamics Research Center: College Station, TX, 1974.
3032 The Journal of Physical Chemistry, Vol. 85, No. 24, 198 1
lulose1p2yield good BET plots and acceptable surface areas while on water-covered silica the BET model becomes inapplicable. This, however, points out the difficulty of surface-area determination when the isotherms do not exhibit a well-defined point B. It appears that in some cases the BET analysis when applied to type-I11 isotherms yields good monolayer capacities, while in other cases the procedure is ~ n r e l i a b l e . ~ ~ B ~ Since the BET area cannot be used, an alternative method to convert q into r will be adapted. When Kelvin’s equation is used, it is possible to estimate at what relative humidity capillary condensation in the large pores of the silica starts to take place. The silica pores of these glass samples possess a cylindrical shape, and water wets clean glass (i.e., the contact angle of water on the walls of the pore is zero). Thus Kelvin’s equation simplifies to the form35 In (P/p0) = -2vr/(rRT)
(8)
where u is the molar volume of the liquid, y the surface tension, and r the radius of the pores. The average pore radius of our CPG-10 being 71.1 nm, a value of 0.985 is calculated from eq 8 for the relative vapor pressure of water at which hysteresis begins. It is evident that at 25.7% and 62.3% relative humidities the water is solely physically adsorbed and, hence, the surface area of the water film may safely be taken as the BET area of the dry sample. On the other hand, at 88% relative humidity the surface area on which n-heptane adsorbs is more difficult to determine. Indeed, because of the pressure drop across the column, an average relative humidity of 88% corresponds to relative humidities of 98% and 77% at the column inlet and outlet, respectively. Thus, it remains possible that close to the column inlet some pores will be filled. However, this is very unlikely to decrease significantly the total area of the film-covered glass in the column. As an approximation, the surface area at 88% relative humidity (3.8% loading) will be taken as the area of the dry glass. If one assumes multimolecular adsorption of water on glass and takes the molecular area of water in the monolayers to be 0.122 nm2,36the three low loadings correspond to 1.3,2.2, and 8.2 average statistical layers of preadsorbed water on silica. From the known surface area of the support and from Figure 4, n-heptane at o.4p/p0 forms 0.5,0.29, and 0.14 of a monolayer at the three relative humidities. Since n-heptane interacts with surfaces by London forces, it is evident from Figure 4 that preadsorbed water on glass reduces considerably the London force field component of glass. To determine how the adsorption potential of n-alkanes on water-coated glass varied with the water loading, we measured the enthalpies of adsorption at zero surface coverage. It was first found that at very low loadings, 0.5% and 1.04% by weight, the heats could not be determined because the loading could not be held constant when the column temperature was varied. Thus, only three water loadings were studied, 3.8%, 21%, and 47%, which correspond to 8.2,45, and 100 water layers, making the gross assumption of uniform water coating at the two higher loadings. In Table I are presented the differential en(33) Gregg, s. J.; Sing, K. s. W. “Adsorption, Surface Area and Porosity”; Academic Press: London, 1967; pp 100-8. (34)Brunauer, S.;Copeland, L. E.; Kantro, D. L. In “The Solid-Gas Interface”; Flood, E., Ed.; Marcel Dekker: New York, 1967; Vol. 1, Chapter 3, p 17. (35) Gregg, S. J.; Sing, K. S. W. “Adsorption, Surface Area and Porosity”; Academic Press: London, 1967; p 136. (36) Ferguson, C. B.;Wade, W. H. J. Colloid InterfaceSci. 1967,24, 366.
Dorris and Gray
TABLE I: Differential Enthalpies o f Adsorption at Zero Surface Coverage for a Series o f n-Alkanes on Water-Coated Silica at Three Water Loadings
H,O loadings, adsorbates n-pentane n-hexane n-heptane n-octane n-nonane n-decane
wt %
- a H * / ( k J mo1-I)
3.8
21
47
25.0 28.4 33.5 36.7 41.7 46.0
24.1 28.6 33.1 31.1 41.4 45.4
33.8 31.8 41.2 45.2
thalpies of adsorption for a series of n-alkanes at the three loadings. The values of AHAwere calculated from eq 6 by a linear regression analysis of the straight-line plots of In VN vs. 1/T. The temperature range was 5-20 “C, and, to minimize water evaporation or redistribution, we simultaneously raised both the temperature of the column and that of the saturator. Each V N for a given adsorbate at a given temperature was an average of at least four injections at infinite dilution. The standard deviation in the slope was as much as 4% in some cases. This indicates the difficulty of determining very precise heats of adsorption when the liquid stationary phase is volatile and the column back pressure nonnegligible. Slight variation in the liquid loading during the runs may change the surface area by a few percent, which is sufficient to introduce a significant error in AHAvalues. Probably because of lower pressure drops and smaller temperature ranges, Karger et al.23p24achieved slightly better precision than in this work. From Table I, within the limits of precision ( 1-1.5 kJ mol-l), the enthalpy of adsorption for a given n-alkane is independent of water loading. Thus, it appears that the surface properties of the column packing are independent of the water content on the glass, at least in the range 3.8-47% water by weight. If so, then the distribution constant, Ks, for a given n-alkane at a given temperature should also be loading independent. The surface area at a water content, x , greater than 3.8% may now easily be determined by injecting an infinitely small amount of a reference n-alkane. In this work nheptane was selected because of the reasonably large V N at 3.8% and higher loadings. It follows from eq 3 that the surface area at a loading x may be obtained from Ax = A3.8(VN)x/(VN)3.8 (9) since the total area at 3.8% loading, A3.8,and VN for n-C7 are known. Using the above method to estimate surface areas as a function of water content, one may determine the Henry’s law constant, Ks, for a series of n-alkanes on silica covered by different amounts of water. In Table I1 the values of Ks at 15 “C are listed. For a given n-alkane, Ks decreases with the water content until, at 3.8% loading, it reaches a constant value within experimental error (f2-3%). From eq 7, the free energy of adsorption at zero coverage may be calculated from Ks values. It was found that, at all loadings, the free energy increases regularly with the number of carbon atoms. The significance of the increments per methylene group in the free energy of desorption, -AGACH2, has been discussed p r e v i o ~ s l y . It ~ ~was shown that, like the heat of adsorption, -AGACH2 is a measure of the London force field of the adsorbent. We proposed that the London force contribution to the surface free energy of a phase 2 (in the present case the watercoated silica) may be calculated from the expression N
YzL = ( - A G A ~ ~ ~ / N ~ C H , ) ’ / ~ Y C H ~ (10)
where N is Avogadro’s constant, U C H ~is the area of a CH2
The Journal of Physical Chemistty, Vol. 85, No. 24, 1981 3033
Adsorption of Hydrocarbons on Water Surfaces
TABLE 11: Effect of Water Loading on the Adsorption Coefficients and Properties of Moist Silica at 15 "C 1O'Ks /m
H p
loadings,
wt % 0.6 1.04
3.80 21 47
n-C,
n-C,
n-C,
n-C,
n-C,
n-c,,
3.06 1.56 0.850 0.858
8.82 4.06 2.02 2.03
25.5 10.4 4.76 4.76 4.76
70.5 26.8 11.2 11.2 11.1
200 67.9 26.7 26.0 25.9
568 173 62.4 61.3 60.9
group lying flat on a surface (uCH = 0.06 nm2), and y is the surface tension of a pure C€&surface (35.9 mN mat 15 "C). Values of -AGACH2and yZLat the five water loadings are presented in the last two columns of Table 11. The lower affinity of the surface toward a CH2 group as the loading increases is very well reflected by -AGACH2 values. Moreover, over 3.8% loading -AGACHz remains constant. However, the most interesting quantity of Table I1 is yZLsince it quantifies the surface of the adsorbent. According to Fowkes,S7a thick film of water on glass would for dry glass to 22 mN tend to reduce yZLfrom 76 mN m-l for bulk water. He argues that a single monolayer of water should allow some of the force field of the glass to operate for an adjacent phase and estimated that yZLfor glass with an adsorbed water monolayer should be 30 mN m-l. On the other hand, Zisman et al. studied experimentally the effect of adsorbed water on the wetting properties of various silicaBJgand metal@surfaces. It was concluded40that the critical surface tension for wetting, yc (where yc = yZLwhen nonpolar liquids are used), depends on the surface concentration of adsorbed water but is relatively insensitive to the chemical nature of the underlying solid. From their contact-angle measurements, the value of yzLfor glass covered by ca. one monolayer was found to be -30-35 mN m-l at 20 "C. This is certainly of the same order as the values of yZLobtained here at 0.6% and 1.04% water loadings, which correspond to ca. one and two preadsorbed water layers on glass. Some care must be taken in the interpretation of yzLobtained from GC infinite dilution measurements through eq 10. Firstly, because of the column back pressure, the water layer thickness varies with the position in the column. Thus, particularly at 0.6% and 1.04% loadings, the possibility exists that some bare patches of glass are present at the column outlet. The value of yzLwould then represent some weighted averages of various yZLfor different energy sites in the column. Secondly, eq 10 is an empirical expression whose apparent success in determining yZL(ref 29 and 1) may reside in the assumed value for uCH and on the small range of y2Lwhere it has been tested. hevertheless, the results at low loadings are reasonable. At the three higher water loadings, the influence of the glass surface appears to vanish, as indicated by the constant value of 22 mN m-l which is indeed the value ascribed by F ~ w k e to s ~bulk ~ water. Because of the uncertainties in eq 10 and in the value of yZLfor bulk water, which according to Aveyard*l is -20 mN rather than 22 mN m-l, the results in Table I1 cannot be taken as positive proof that, at loadings greater than 3.8%, the surface of water on glass is identical with the bulk water surface. However, the data suggest a surface structure closely related to bulk water. (37) Fowkes, F. M.J . Colloid Interface Sci. 1968,28, 493. (38) Shafrin, E. G.; Zisman, W. A. J.A m . Ceram. SOC.1967,50,478. (39) Bernett, M. K.; Zisman,W. A. J. Colloid Interface Sci. 1968,28, 243. (40) Bernett, M. K.; Zisman, W. A. J . Colloid Interface Sci. 1969,29, 413. (41) Aveyard, R. J. Colloid Interface Sci. 1975,52, 621.
c 2 6
N
- AGACHa/
rz=/
(mNm-l)
(kJ mol-')
2.50 t 2.26 t 2.05 f 2.04 i 2.04 *
0.02 0.02 0.03 0.02 0.02
33 27 22 22 22
1. n-heptane 0 3.8% H 2 0 0
P
21% H 2 0 Data of Karger
.B
2. n-octane
g4 X
L
3 2 1
Oo
0.1
0.2
a3
0.4
0.5
0.6
1
7
PIR (r
Flgure 5. Adsorption isotherms at 15 ' C vs. PIP,,)for n-heptane on silica at 3.8% and 21 % water loadings and for n-octane at 21 % water loading. Some data points of Karger et at.'' for n-heptane and n-octane on water-coated Porasil D are also shown for comparison.
From a comparative study between chromatographic and surface-tension data at finite and zero coverages, it will be shown that the water-coated glass appears to differ slightly from bulk water. Figure 5 presents the adsorption isotherms at 15 "C for n-heptane at 3.8% and 21% by weight water loadings and for n-octane at 21% loading. The peak-maxima method yields smooth calculated isotherms, so that the points for the n-C, plot represent values of r calculated at arbitrary intervals of p / p @ At 21 % loading, the surface area is much smaller than at 3.8%, and, to convert q to I', we determined the area from eq 9. Finally, some data points obtained chromatographicallyby Karger et aLZ5for n-C7and n-C8at 12.1 "C on 20% by weight of water coated on Porasil D are also shown. The isotherm points of n-C, obtained in this work at two loadings superimposed within 1.5% at all PIPo, indicating again that the surface properties reach a steady state over -3.8% water content. As on ~ a t e rthe , ~isotherms ~~ for n-C7 and n-C8 are type I11 with decreasing adsorption as the number of carbon atoms in the n-alkane chain increases. When compared to the other chromatographic determination, the results for both solutes agree to better than -3% at all coverages. Note that Karger et al.23reported isotherms for the same two n-alkanes in earlier work, but with the values of r roughly 25% lower than in the second study. The reason for the discrepancy between their two sets of data is not clear since the surface area was estimated in the same manner in both studies. Our results agree very well with the more recent determination in view of the significant uncertainty in the surface area of the packing for both cases. The comparison between dynamic and static adsorption measurements at finite coverage is best made by converting
3634
The Journal of Physical Chemistry, Vol. 85, No. 24, 1981
''
I n-C7 1
n-C7 2. n-C8 o o
This Work
n-C8
/
Jones and
/
Ottewill
P/R Figure 6. Surface pressure vs. relative vapor pressure for the adsorptlon of n-heptane and noctane at 15 O C on silica coated with 21 % by weight water. The dashed portion of the two curves represents the extrapolation range. The data points of Jones and Ottewil16 for the same n-alkanes at 15 O C on bulk water are also shown.
surface concentration, l", to surface pressure, a,by means of Gibbs' adsorption equation a = RT
1"r d(ln 0
p)
(11)
As previous1y,2t28GC adsorption isotherms were first fitted to a polynomial of the form
r = j=l 5 ajp'
(12)
using a regression technique. The a's of eq 12 represent the coefficients of the polynomial, and the isotherm is assumed to pass through the origin. The best fits for n-heptane and n-octane were obtained for five- and sixterm expansions, respectively. Values of r calculated from the polynomial agreed to better than 1 % with the experimental isotherm at all partial pressures. Moreover, values of Ks determined from the initial slope of the fitted plot agree to 1%with the values in Table 11. The surface pressure, a, at a given p , could thus be calculated by combining eq 11 and 12, yielding eq 13. Figure 6 presents n ?r
= RT C ( u j / j ) # j= 1
(13)
smooth a vs. p / p o plots for n-heptane and n-octane at 15 "C on 21% water-coated glass. The dotted part of the lines represents the extrapolation of the polynomial fit to regions where the isotherms could not be measured. This extrapolation technique is believed to introduce in both cases an error of less than 10.3 mN m-l in the values of a at po, TO. For comparison, the tabulated values of (a,p)for n-C, and n-C8at 15 "C obtained directly from surface-tension measurements by Jones and Ottewil15are also plotted. The results in both studies indicate that the adsorption of these n-alkanes on water or on water-coated glass causes a very slight reduction of surface tension. Despite similar shapes for the plots, the GC values of a at any p / p o for the two n-alkanes are consistently larger than the surface-tension measurements. A similar discrepancy was obtained in other chromatographicwork,%and Hartkopf and Karger= ascribed the difference either to an erroneous estimate of their packing surface area or to imprecisions in the static measurements. It appears from the more recent surfacepressure data of Hauxwell and Ottewillg that the static
Dorris and Gray values of a in Figure 6 are unequivocally accurate and reproducible. On the other hand, to force agreement between dynamic and static data, the surface area of our packing at 3.8% water loading would need to be larger than the area of dry glass. This is clearly unreasonable. As mentioned previously, the wet surface area is probably lower than that of the dry glass because of capillary condensation at the column inlet. Thus, in this work, surface area cannot be responsible for the difference with static data. This difference may also be quantified by determining the London force component of the surface free energy from aovalues rather than from -AGACH2. If the n-alkane spreads on water and obeys Antonoff s rule, yZLof water may be c a l ~ u l a t e dfrom '~~~~ YzL = (ao4- 2?'1)~/471 (14) where y1is the surface tension of the spreading n-alkane. For n-heptane (which wets the water surface), y1 = 20.6 mN m-l at 15 0C.32 From Figure 6, ao is 2.5 mN mF1in this work and 1.75 mN m-l in the static experiments, and, from eq 14, this yields 78 values of 23.2 and 22.4 mN m-l, respectively. The values for yZLcalculated from -AGACHg (Table 11) and from aoagree well. It is also quite evident that the surface properties of the water-coated glass approach very closely the surface properties of bulk water, as indicated by a y2Lvalue only 0.8 mN m-' larger in the present study. This slight difference is, however, sufficient to cause 'I or Ks values to be almost twice as large in the GC experiment. If a systematic error in the surface-tension measurements may be discounted, then the adsorption of n-alkanes in the GC experiment appears to take place on a slightly larger energy surface than bulk water. Karger and cow o r k e r ~ have ~ ~ -discussed ~ ~ ~ ~ various ~ factors which may affect the retention behavior of hydrocarbons on watercoated macroporous solids. Examples of potential sources of error are the following: the nonattainment of equilibrium in the GC process; surface curvature effects on the water in the pores of the solid; the washing of the support; and the adsorption or solution of the nitrogen or helium carrier gas on or in water. In all cases, the influence was found to be negligible. The constancy of the adsorption data from 3.8% to 47% loading in this work would tend to add weight to their conclusions. An alternative explanation is that the silica surface may be partially dissolved by water. This possibility has also been advanced by Okamura and Sawyer.21 Massoudi and King43reported a slightly enhanced adsorption of butane on the surface of water containing sodium chloride. Moreover, from a theoretical study, Richmond et al.44have predicted that n-alkanes which normally form a lens on water may wet the surface when sufficient quantities of salt are dissolved in water. Thus, it is possible that the concentration of dissolved salts in the thin layers of water adjacent to glass is high enough to increase yZLslightly. However, more experimental work is required to confirm this hypothesis. Despite this anomaly in the GC results, since the water-coated glass approaches very closely the surface properties of bulk water, the thermodynamic functions at zero surface coverage should also be similar to those for bulk water. In Table 111, Ks values and the thermodynamics of adsorption at 20 "C for a series of n-alkanes on silica coated by 21% water by weight are presented. Also (42) Chatterjee, A. K.; King, J. W.; Karger, B. L. J. Colloid Interface Sci. 1972, 41, 71. (43) Massoudi, R.; King, A. D., Jr. J. Phys. Chern. 1976, 79, 1670. (44) Richmond, P.; Ninham, B. W.; Ottewill, R. H.J.Colloid Interface Sci. 1973, 45, 69.
Adsorption of Hydrocarbons on Water Surfaces
The Journal of Physical Chemistry, Vol. 85, No. 24, 198 1 3635
TABLE 111: Thermodynamic Functions of Adsorption at Zero Surface Coverage for a Series of n-Alkanes on Water-Coated Glass (21%Loading) at 20 "C
10'Ks Im
deg-' mol-')
(kJ mol-')
-AH,"/ ( k J mol-')
-AHA( theor)b/
( k J mol-')
n-C, n-C, n-C, n-C, n-C,
0.710 1.64 3.74 8.49 19.28 43.55
7.45 9.49 11.50 13.49 15.49 17.48
56.8 65.2 73.7 80.5 88.4 95.2
24.2 28.6 33.1 37.1 41.4 45.5
26.8 31.7 36.6 41.5 46.5 51.4
24.8 29.3 33.8
n-C10 a
-
adsorbates
From ref 32 at 25 "C.
-AS,l(J
-AHA/
( k J mol-')
Theoretical values from ref 13.
shown are the heats of liquefaction of the n-alkanes, AHL, and theoretical estimates of A H A from Guiochon et all3 The most obvious feature is that, for all of the n-alkanes, A H A is significantly smaller than the heat of liquefaction. This is perfectly consistent with the type-I11 isotherms obtained previously. Thus, the surface acts as an inert, low-energy surface, and there is obviously no sign of surface perturbation by the n-alkanes. Moreover, the values of AHA agree surprisingly well with the theoretical estimates. When compared to the other GC work,%our A H A values are slightly larger, but both sets of data agree within experimental error. A larger discrepancy is found when the heats are compared to the results of Jones and O t t e ~ i l l . ~ Note however that, in contrast to the GC experiment, where the Henry's law region is easily reached, the surface-pressure measurements have to be extrapolated to ?r 0. It is evident from Figure 6 that the extrapolation is very long so that Ks,and to a large extent AHA,values are subject to a large error. All of the functions of Table I11 are seen to vary regularly with the number of carbon atoms. For each additional methylene group, Ks values increase by a factor of 2.28, while the increments per CH2group in AGA, AHA,and MA are -2.0 kJ mol-', -4 kJ mol-', and -8 J deg-' mol-', respectively. Similar incremental quantities were obtained from surface-pressuremeasurements by King et al.'*l2 for lower homologues of the n-alkane series and also by Clint and c o - ~ o r k e r for s ~ ~a series of n-alkyl alcohols. Consequently, from the thermodynamics of adsorption, the surface properties of water, coated onto glass, resemble very closely those of bulk water with a difference in y2, of only -1 mN m-l. Any striking difference in the heats of adsorption on water-coated glass and on bulk water seems improbable. The observed heats of adsorption are close to the heats of liquefaction of the hydrocarbons and rule out a restructuring of water by the hydrocarbons on water-coated glass, and probably also on bulk water. We feel that our results are evidence against surface restructuring of water and that the contrary conclusions result from the experimental difficulties in determining heats of adsorption at very low surface coverages by conventional techniques.
-
Summary In contrast with water-swollen cellulose,'V2 the watermacroporous silica system does not yield a reliable BET surface area with n-alkane adsorbates. However, the well-defined and rigid pore structure of this silica sample allowed an alternative approach to surface-area estimation. Unlike cellulose-water gels, which undergo structural modifications as the relative humidity increases, large changes in surface area are unlikely to occur until pore (45) Clint, J. H.; Corkill, J. M.;Goodman,J. F.;Tate, 3.R.J . Colloid Interface Sci. 1968,28, 522.
filling by capillary condensation takes place. The following line of reasoning was used to estimate surface areas of the silica over the complete range of moisture contents. Firstly, the surface area of the dry macroporous silica was found to be 18.9 m2 g-I from nitrogen adsorption. The appearance of pore filling in this sample was estimated from Kelvin's equation to occur only at high relative humidities (Le., >98.5%). At lower relative humidities, only physisorption takes place, and hence the surface area is taken to be the same as that for the dry sample. When the water loading was increased from 3.80% (the loading corresponding to 88.0% relative humidity) to 47% by weight, the enthalpies of adsorption at zero surface coverage for a series of n-alkanes remained constant within experimental error. Because the enthalpies are independent of surface area, this was taken as evidence that beyond 3.80% loading the nature of the surface remained constant, so that marked changes in measured retention volumes for n-alkanes reflected variations in surface area due to pore filling. The surface areas at high water contents were determined with n-heptane as the probe, according to the equation VN = &A, where Ks was assumed to remain constant for loadings greater than 3.80% by weight. Although somewhat convoluted, the methods yield surface areas which are reasonable and which allow interpretation of the adsorption data with some confidence. Below 3.8% loading, the adsorption of the n-alkanes decreases markedly as the amount of water adsorbed on the silica increases. If the surface area is constant in this range of loadings, then these changes must result from a decreasing energy of interaction with the surface. This argument is supported and quantified by the values for the incremental free energy of adsorption and the London component of the surface free energy reported in Table 11, which are surface-area independent. Above 3.8% loading, the water on glass acts as a homogeneous, low-energy surface, as shown by type-I11 shapes for the n-alkanes, heats of adsorption significantly smaller than the heats of vaporization, and a yZLclose to 23 mN m-l. These findings suggest that water is inert toward hydrocarbons, and consequently the model for surface perturbation of the water surface by some adsorbates does not appear to hold. However, the water coated on glass appears to possess a London force field of attraction slightly larger than that of the plane surface of bulk water. This resulted in surface excesses and surface pressures for n-alkanes larger than the values determined by surface-tension measurements. Thus, a total reconciliation between dynamic and static data is not yet possible. Acknowledgment. G.M.D. thanks the Government of Quebec for a post-graduate scholarship. The work was supported in part by a grant from the Natural Sciences and Engineering Research Council of Canada.
