SURFACE AREAOF WATERPREADSORBED ON POWDERED SUBSTRATES
AGO values obtained by e.m.f. measurements with those of the literature.
Table I AGO,
AGO,
OK.
this work
Lit. values
800 1000 1200 700 850 800 1000 1200 800 1000 1200
-40.07 -35.58 -31.10 -211.34 - 198.87 - 108.07 -98.43 -88.79 - 104.35 -95.85 -87.35
-39.61” - 34. 90a -30.19” -210226b -197.93b -102.328 - 95.00’ -87.68a - 106. 148 -98. IOa -90. 14a
T, Oxide
KiO
Fez04
woz Moon
a 0. Kubaschewski and E. L. Evans, “Metallurgical Thermochemistry,” Pergamon Press, London, 1958, pp. 336-343. * 0. Kubaschewski and JI. A. Catterall, “Thermochemical Data of Alloys,” Pergamon Press, London, 1956, p. 173.
1029
Remarks In some runs it was noted that the metal and metal oxide may penetrate into the intermediate electrolyte. The total amount of metal and metal oxide which penetrated was greater in the case of W and Mo than in the case of Fe. Nickel did not show any penetration. This effect was less pronounced if the metallic pellets were previously kept in contact with other pellets of intermediate electrolyte or used in preceding runs. l4 It also decreased as the sintering temperature of the intermediate electrolyte increased and the working temperature decreased. Because of this effect, the measurements on chains [3], [4],and [5] were performed above 650°, a t increasing temperatures and at a relatively fast rate, even if this implies an imperfect stabilization of e.m.f. values to be measured. Thus, the variation of the conduction properties of the intermediate electrolyte during the measurements has been minimized as much as possible. (14) S. Aronson and J. Belle, J. Chem. Phgs., 29, 151 (1958).
The Surface Area of Water Preadsorbe(don Powdered Substrates
by William H. Wade Department of Chemistry, The Universitg of Texas, Austin 13, Texas
(Received September 27, 1963)
The surface areas of powdered :solidspre-equilibrated with various amounts of water have been measured by nitrogen adsorption. The data are best interpreted in terms of a dual mode of adsorption, requiring the adsorbate to be held as both a uniform film and as capillary condensate in the contact zone of touching particles.
Introduction Many liiodels have been employed to explain what is loosely referred to as “multilayer, physical” adsorption and lllost have fallen into disuse for diverse reasons. The B.E.T. model and its accompanying theoretical foriiiulation would sippear to be the sole survivor, not so much because of the correctness of the
underlying assumptions, but because of its general utility in providing reasonable values for the surface area of high specific surface area solids. Of probable secondary iinportance is its ability to fit yualitabively a variety of isotherm shapes. Several apparently valid objections to the B.E.T. model as applied to nonideal surfaces have been raised and even for ideal Volume 68,Number 6 M a y , 1964
WILLIAMH. WADE
1030
surface interactions the B.E.T. model must be an oversimplification. For instance, Karnaukhov and Kiselevl have calculated that during surface area measurements by gas adsorption, the adsorption process itself leads to apparent changes in B.E.T. surface areas. They assumed the growth of a uniform film on spherical adsorbent particles. One undeterminable parameter that is not independently evaluable by them is the number of physical contacts made by any one adsorbent particle and its neighbors. The surface area on the outside of an adsorbed gas film depends strongly on this parameter as does the volume in the contacting zones inaccessible to adsorbate molecules of finite size. The work discussed here is addressed to this point and arose from studies on the validity of HarkinsJura “absolute” area measurements. Two series of measurements had previously been performed in this laboratory. The first series2was an intercomparison of surface area measurements between B.E.T. analyses of krypton, argon, and nitrogen adsorption isotherms and Harkins-Jura calorimetric measurements for a series of alumina powders reported still earlier.a The surface areas by gas adsorption were found to be internally consistent, with an average deviation of approximately 4%. However, the Harkins-Jura areas obtained from heats of immersion data on the same samples equilibrated with water vapor a t p / p o = 0.95 uniformly yielded lower surface areas. For particulate samples with no internal pore structure, the discrepancy is 10-150j0, but for the one gel sample used, the Harkins-Jura specific surface area is 4 mb2/g.whereas by gas adsorption it was 221 m.”/g. As noted elsewhere,*the Harkins-Jura method does measure only the surface area of the adsorbategas interface but discrepancies of even 15% should not be overlooked in precise measurements. A second series of experiments5 was initiated a year ago to measure differential heats of adsorption by calorimetric immersion techniques but became sidetracked when it was found that heats did not assympotically approach the surface enthalpy of the liquid used in the calorimeter (water, in this case). Rather, it was found that immersional heats for particulate alumina samples eventually would fall to 100-110 ergs/cm.2 rather than the ‘(theoretical” limit of 118 ergs/cm.2. What is needed is an independent evaluation of the surface area of these powdered solids as a function of the water preadsorbed on the samples. The only apparent experimental approach is equilibration with water vapor, freezing in situ, followed finally by a gas adsorption surface area measurement a t liquid nitrogen temperatures. Karasz, et al. ,6 carrying out such measurements on a Ti02 sample, apparently The Journal of Physical Chemistry
did not consider that the preadsorption of water significantly altered the surface area.
Experimental Volumetric adsorption isotherms were obtained for nitrogen on four samples-three of Al2O3and one of Si02. Additional information on these samples can be obtained elsewhere. The previously reported surface areas, Z, of these samples are: A1203, 2.72, 65.4, and 104 m.2/g; Si02, 188 n 2 / g . Areas were estimated on the basis of the following molecular areas: N2, 16.2 Ar, 16.0 k 2 ; and Kr, 20.0 Samples were outgassed a t 200”. Water vapor was added slowly with the water sample bulb initially a t -80” and allowed to slowly warm to a final temperature consistent with the required pressure. This was done to preclude any possible irreversible pore condensation caused by ‘(shocking” the sample with a transient pressure higher than the ultimate pressure.* This equilibration was considered complete when a pressure drop of >> t ) ; (c) introduce the value of rm from (b) along with the appropriate values of R, n, 6, 0, and V (the latter obtained from the experimental water isotherm a t the p/pO set in (a)) ; (d) solve eq. 5 for t; and (e) introduce t, rm,R, 6, and n into eq. 6 and obtain the surface area, 2.
_- nrm2 V‘(R + t
+
rm2)
- ~2 sin-1
--
nR02
+ n02 d 2 R D +
Q2
sin-‘
R Ri-t+r, -
2-} (cc./g.)
R+0
(5)
The last two terms are now added to correct for the volume excluded to the water molecules. It is also possible to write an expression for the surface area of the adsorbate-gas interface
t
1
Figure 3. Dual adsorption mechanism consisting of a uniform film of thicknescl t with capillary condensation superimposed.
To facilitate the preceding calculations, the Control Data Corp. 1604 computer a t the University of Texas was programmed for these calculations and simultaneous evaluation of the corresponding surface areas. The calculations were performed for the four samples of Fig. 1 other than the compressed 104-m.2/g. A1203. It is apparent that compression of this latter sample increased the number of contacts, causing drastic modification in the surface area when large volumes of water are adsorbed but only minor changes under totally outgassed conditions. Even on the basis of this latter more refined model, quantitative agreement could not be obtained. However, the qualitative features are as required: for high specific area samples, a rapid drop in surface area primarily due to capillary condensation is observed, followed by a slow decrease due to both modes of water adsorption. Calculations for the uncompressed 104m.2/g. alumina are given in Fig. 4. The best fit is obtained n = 5 a t small volumes adsorbed and n L= 3 for large volumes adsorbed. In addition to calculating surface areas, the C.D.C. 1604 computer also tabulated the volumes of “film” water and “capillary” water. For n = 4 (best average value) the 104-ni.2/g. sample has approximately 1% of Vads (total) held as capillary condensate a t low p /pol Volume 68,Number 6 M a y , 1964
WILLIAMH. WADE
1034
""1 100.
I
0.01
0.02
0.03
0.04
V
0.05
0.06
0.07
0.08
0.09
(cc/gm)
Figure 4. Comparison of experimental and theoretical surface areas with volume of water preadsorbed on the bask of the dual mode model.
increasing to approximately 25% a t p / p o = 0.95 with an intermediate value of approximately 370 a t p / p o = 0.30. This indicates that a sinal1 percentage of capillary condensate is responsible for the majority of the decrease in surface area a t low and intermediate relative pressures (since the simple uniform film model would predict little or no change in surface area for n = 4). The best fit for Alon C (65.2 nx2/g.) is also obtained with n = 5 a t low p / p o and n = 3 a t high p / p o , and is very siiiiilar to the 104 m.2/g. sample with regard to the percentage of water held by capillarity, ranging from 1 to 30% for n = 4. The Cab-0-Si1 sample (188 m.2/g.) has a best fit for n = 3, but the calculated percentage of capillary water is much higher-20% a t low p / p o and up to 30% a t high p/pO, consistent with the almost type 3 shape of the water adsorption isotherm. The one low area sample (2.72 m.2/g.) shows no initial rapid drop of the surface area. The best fit to the experimental data is for n = 9 and the calculations for all values of n show no initial steep drop in the surface area; the calculations for n = 9 do apportioii only 0.1% to capillary condensate a t low relative pressures. Several possible explanations of the quantitative lack of agreement are possible. (1) For real saniples, n and r must be average values and correctly weighed distributions of n and r values would need to be folded
The Journal of Physical Chernistru
into the calculations. (2) The validity of the Kelvin equation for "meaningless" menisci radii of molecular dimensions is obviously open to question. This point was repeatedly discussed in the literature of 30 to 40 years ago, and no firm conclusions have ever been reached, although its validity is repeatedly assumed in obtaining pore size distributions from hysteresis loops. (3) Packing geometries may need to be considered for large adsorbed volumes as previously discussed. (4) A uniform adsorbate density is assumed without any independent check. ( 5 ) With regard to the present study, it is uncertain that the B.E.T. analysis of ISz adsorption isothernis nieasures the actual surface area. The following points can be made supporting the assumption that the B.E.T. analysis still measures the area. (1) Values of n = 3 or 4 predict little or no change in area for a uniform film (B.E.T.) model. ( 2 ) For type 2 isotherms (such as all the K2 isotherms), the i v z adsorbate held in contact zones by capillarity is only a few per cent of the total. (3) The nitrogen isotherms were measured over a restricted p / p o range (0.04 to 0.22) where there is not a large rariation in the NZ adsorbate coverage (less than a 30YG total variation). (4) To a first approximation, the void volume problem introduces a constant but usually small constant error and does not affect relative areas for a given sample. In any event and regardless of the correctness of the model here pictured, the Harkins-Jura areas are usually low by approximately lOgj.,, which is the same approximate percentage lowering of the surface area during equilibration with water at high relative pressures, and it should be noted that these corrections may need to be made for samples in the micron size range. In conclusion, this study has reaffirmed the existence of commonly overlooked complications encountered in measurements in the multilayer region. Though Harkins and Jura themselves were aware of such possible sources of error, later workers have a t times excessively minimized their importance. Acknowledgment. The author wishes to thank the American Petroleum Institute and The Robert A. Welch Foundation for their partial support of this study. Also, appreciation is expressed to Dr. William C. Gardiner for assistance in programing the computer, Dr. Hilton Mollenhauer for taking electron microscope pictures of the samples, and Mr. Arnold Charles Falk for perforniing many of the surface area measurements.