784
COSWAY PIERCE A S D R . NELSOS SMITH
ADSORPTIOK-DESORPTIOS HYSTERESIS I S RELATIOS TO CAPILLARITY O F ADSORBESTS' COSWAY P I E R C E
AND
R. NELSON S M I T H
Department of Chemzstry, Pomona College, Claremont, Calzfornia Received July 19, 1949
The phenomenon of hysteresis in the desorption? of vapors from solid surfaces is generally assumed to be related to the emptying of capillaries. I n fact, the occurrence of hysteresis is usually taken as evidence for capillary condensation. Several theories have been advanced to account for a difference in the manner in which the vapor condenses in and evaporates from capillaries. A11 but the more recent ones of Juhola and Wiig (lo), Hill (7), and Gleysteen and Deitz (6) have been summarized by Brunauer (3). In recent studies with some nonporous adsorbents we have found hysteresis to occur in desorptions from surfaces where there is evidence against the existence of capillaries. This has led us to inquire into a possible mechanism for hysteresis that does not require capillary condensation of the adsorbate. A new hypothesis has been developed which we believe gives an adequate explanation of hysteresis both from plane surfaces and from capillaries. I t also takes into account the occurrence of hysteresis with some adsorbates while other adsorbates show none from the same capillary pores. This hypothesis is herein presented. EXPERIMENTAL
Adsorption and desorption isotherms were determined for ethyl chloride and water on graphite KC-1, Graphon, and activated charcoal S84. Graphite SC-1 was obtained from the National Carbon Company? It has a nitrogen area of 4 sq. m./g. and has less than 0.001 per cent ash. This is the same sample as was used for other studies in this laboratory (11, 12, 13). Graphon is partially graphitized Spheron Grade 6 carbon black, obtained from Godfrey L. Cabot C ~ r n p a n yI. t~is from the same lot used by Beebe, Biscoe, Smith, and Wendell ( 2 ) and by Joyner and Emmett (8). The nitrogen area is given by the latter as 80 sq. m./g. 584 is a highly activated charcoal prepared in this laboratory by carbonization of sheet Saran a t 600°C. and steam activation a t 900°C. until there is 84 per cent weight loss of t,he carbonized material. I t has an adsorptive pore volume near 1.1 ml./g. 1 This is a progress report of work conducted under Contract 5 8 onr 54700 with the Office of S a v a l Research, United States Kavy Department. 2 In this paper v e are speaking only of reversible hysteresis. Brunauer (3) defines hysteresis as reversible if on repetition of the experiment the adsorption isotherm is completely reproduced. When the adsorbent itself is altered by adsorption and desorption of vapor the explanation must be s o m m h a t different than x h e n there is no alteration of the adsorbent structure. 8 Through the courtesy of D r . Lester L. Winter. 4 Through the courtesy of D r . Walter R . Smith.
ADSORPTIOS-DESORPTIOS
785
HYSTERESIS
All isotherms were determined gravimetrically by removing and weighing the sample bulb after equilibrium was attained at each pressure. Experimental details are given in previous papers. When the isotherm temperatures are below room temperatures, there is alternate desorption and readsorption of vapor as the sample tube is removed for Neighing and then re-cooled in the bath. This amounts to “scanning” the hysteresis loop in desorptions and may have the effect of diminishing the apparent amount of hysteresis. Vhen the isotherm temperature is below room
FIG. 1
FIG. 2
F I G . 1. Ethyl chloride a t 0°C. on Graphon. 0 , desorption. At 0.997 p , a volume of 293 ml. is adsorbed. F I G . 2 . K a t e r on Graphon at 28.9”C.; 0 , desorption
temperature it is necessary to reequilibrate the sample to a pressure near p,, after each removal for n-eighing. The isotherms are shown in figures 1 to 4. DISCUSSIOX
The ethyl chloride isotherm in figure 1 indicates, as discussed later, that the adsorbent, Graphon, has no capillaries. The water isotherm of figure 2 shows, however, a pronounced desorption hysteresis. If xi-e are coirect in concluding that Graphon has no capillaries, it follows that none of the previous theories (all of which are based upon capillary condensation) are adequate to explain the water desorption hysteresis. We have therefore sought an explanation which is applicable for adsorption-desorption in capillaries and on plane surfaces.
i8B
CONWAY PIERCE AND R. NELSOS SMITH
FIG.3 FIG.4 FIG.3. Ethyl chloride a t -iSnCc. (left) and water a t 2 8 . 9 T . on Charcoal SS1. Plotted a8 liquid volume adsorbed. 0 , desorption.
FIG.4 . Water on graphite SC-1 at 28.9"C.;
0
0 , desorption
e
1
FIG.5
FIG.6
FIG. 5 . Schematic representation of water adsorption on plane carbon surface. T h e droplets first formed a t active sites 9 and Y ( A ) later merge (B)as pressure is increased. FIG.6. Schematic representation of adsorption in large capillary. A multilayer film is first formed on each wall (A) and then merges at higher pressure (B)to fill the capillary. PROPOSED THEORY
If there is stable equilibrium in the desorption part of the cycle (which we believe to be the case), it means that the adsorbed molecules are more strongly held during desorption than a t the corresponding degree of saturation during
ADSORPTIOX-DESORPTION
HYSTERESIS
787
adsorption. I n other n-ords, a change has occurred in the nature of the adsorbed layer. Such a change is readily accounted for by the assumption that adsorption occurs in isolated clumps that later merge. When two clumps merge the forces on all molecules are increased, as specifically discussed below. First, we consider adsorption on a plane surface of nonuniform activity, as illustrated in figure 5-1.Two active sites, X and Y, each hold a multilayer clump of molecules. Xs adsorption proceeds the clumps may meet and merge, as in figure 5B. When this happens all molecules of the combined clump are held by long-range forces from both active sites, whereas before merging the molecules in each clump were held only by forces from the active site on which the clump is adsorbed. Thus the vapor pressure is lowered after the clumps merge, and desorption hysteresis vi11 occur. This hysteresis will persist until a point is reached at which the clump breaks up into two separate clumps; at this point the desorption branch of the isotherm rejoins the adsorption branch, since conditions are no\v exactly as they were at this degree of saturation in adsorption. A condition similar t o that described above exists when Jvide capillaries are filled by multilayer adsorption on opposite walls, folloned by eventual merging of the films. The conditions before and after merging are illustrated in figures 6-1and 6R. Prior to merging, the layers are held by surface forces from a single TTall, but after merging surface forces from both 1)-alls affect all molecules. This reduces the fugacity and there is desorption hysteresis rhich persists either until the capillary contains only a monolayer on each 11-all or until the remaining adsorbate breaks up into two separate multilayer films. In figure 6 we h a w , for simplicity, represented thecapillary as a fissure betiyeen two plane parallel valls. This is, of course, an oversimplification but is chosen to shon- that one need not assume constrictions, bottle necks, or closed-end capillaries to account for hysteresis. The single multilayer films on oppoEite walls before merging (figure 6.1) are represented as of variable thickness, in keeping with our Tie\\- that the most active sites will hold more molecules than less active ones. The present theory is based on two assumptions: (a) that many surfaces are of nonuniform activity and ( b ) that long-range forces from a surface may estend to great distances through a liquid-like film but not to the same distances in the absence of a film. There is much evidence from heats of adsorption for the first assumption; in fact, ive know of no heat of adsorption data xhich indicate uniform surfaces. As to the second, it is obvious from multilayer adsorption data that surface forces can extend t o great distances in a film. The ethyl chloride isotherm of figure 1 s h o m adsorption of some 32 statistical layers before saturation is reached. This means that syface forces can l o w r adsorbate vapor pressures through a film of some 150 h. in depth. It appears reasonable then to assume that when tm-o films merge all molecules are acted upon by additional forces. since all now have contact through the film \vith additional surface sites. In absence of a film it is generally assumed that van der Kaals forces fall off inversely v i t h the seventh pon-er of distance. Consequently the surface forces exert negligible effect in space at distances of the order of several molecular
788
CONTVAY PIERCE AND R. XELSON SMITH
diameters. I n the situation of figure 6 d the film on one wall is not affected by forces from the other nall until the t v o films merge. APPLICATIONS
I n applying the present theory to experimental data it must be kept in mind that hysteresis occurs only when the mechanism of adsorption is such that discrete clumps or films merge. This is the determining factor rather than the presence or absence of capillaries. I n fact, as sho~vnlater, given capillaries may with some adsorbates yield isotherms that exhibit desorption hysteresis while with other adsorbates there 1s no hysteresis. We consider now various examples of hysteresis. 1. Water shows hysteresis in desorption from Graphon but ethyl chloride does not. The difference is due to the way in which the t x o adsorbates are held. Ethyl chloride gives a typical Type I1 isotherm, with a monolayer apparently completed a t lovi relative pressure. This must mean that the entire surface can hold ethyl chloride molecules firmly, even though some sites have greater activity than others. The first layer is completed before there is appreciable multilayer adsorption. Therefore the condition of figure 5 does not apply; no change takes place in the nature of the adsorbed film as adsorption proceeds and there is no desorption hysteresis. Water, on the other hand, is poorly held by carbon. A relative pressure of 0.99 is needed to adsorb a statistical monolayer and at 0.997 PO only two statistical monolayers are held. We think that this behavior indicates a condition like that of figure 5 . Only the more active sites hold adsorbate, and multilayers build up as discrete clumps a t these sites. At high relative pressure some of these clumps merge, thereby changing the condition of the adsorbed molecules, and hysteresis is shown on desorption. Our assumption that adsorption of water occurs only at certain active sites is in keeping with known properties of the carbon-water system. First, as previously discussed (12), Tvater lacks mobility on a carbon surface, while other adsorbates appear to have such mobility There can be surface mobility only when all sites can hold adsorbate; otherwise the adsorbate molecule evaporates when it reaches a site that cannot retain it. Second, the heat of emersion data of Basford, Jura, and Harkins (1) indicate that a saturated graphite surface is not completely covered by a mater film. When a saturated sample of titanium dioxide is immersed there is a heat evolution of 119 ergs/sq. cm. due to disappearance of water surface. But when a nearly saturated graphite sample is immersed, approximately 86 ergsisq. cm. is absorbed; in other words, there is not a complete water film over the surface. From the adsorption isotherm and these other considerations then, we believe that water is adsorbed in clumps rather than over the whole surface. 2. Desorption from capillaries may or may not give hysteresis. This is shown in the isotherms of xater and ethyl chloride on charcoal S8-1, in figure 3. Here the amount adsorbed is plotted as liquid volume. As discussed in a previous paper (12), if water is held by capillary condensation then ethyl chloride must
ADSORPTION-DESORPTIOS HYSTERESIS
789
also be held by condensation in the same capillaries; the point B volume5 of ethyl chloride is greater than the saturation volume of ivater. Yet the water isotherm shows hysteresis and the ethyl chloride isotherm does not (no Type I isotherms give desorption hysteresis). The difference is readily explained by the theory given above. Ethyl chloride does not fill the capillaries by first forming a multilayer film on opposite walls and eventual merging of these films. >Tost of its adsorption occurs at lox relative pressures, too low for appreciable multilayer adsorption. This is seen by comparing the isotherms of Graphon and S84. It must be, then, that ethyl chloride fills the capillaries by adsorption at a meniscus which starts at the narrowest places, as discussed in our previous paper. If so, there is a t all times a n-all-to-mall film and the process of figure 6 does not apply. On desorption all molecules are held exactly as they were on adsorption and there is no hysteresis. Rater, on the other hand, is poorly held by carbon. Discrete clumps are adsorbed a t the most active sites and these later merge to fill the capillaries. This merging leads to hysteresis for water desorption. 3. Desorption from wide capillaries is always attended by hysteresis, regardless of the adsorbate employed. When the walls are too far apart to cause the capillary to fill at low relative pressure there is multilayer adsorption on the separate walls at high relative pressure and eventual meeting and merging of the films, as shown in figure 6. Whenever this happens the isotherm is of Type IT.‘. Such isotherms shov the rise beginning at the relative pressure region near 0.4 that is characteristic of multilayer adsorption. It is only when capillaries fill in the higher relative pressure region that multilayer adsorption on separate walls can occur. We know from many Type I1 isotherms that appreciable multilayer adsorption does not begin until the relative pressure is near 0.4 or greater. Gleysteen and Deitz (6) gave some years ago a clear explanation of hysteresis from capillaries that fill by multilayer adsorption. They were the first to “attribute hysteresis to some fundamental change in the adsorbed layer that takes place at a relative pressure near saturation” and to point out that molecular clusters may form at many points on a surface and then be modified by merging. In essence our explanation is like theirs but we differ in some important details. First, after shon4ng that the heat of desorption is greater than that of adsorption for isotherms that show hysteresis, they related this excess heat to the heat of adsorption of the last layer, Q, first proposed by Brunauer, Deming, Deming, and Teller (4). According to our viev, the excess heat of desorption is not the heat of adsorption of the last layer, as incorporated in the B.D.D.T. treatment. Rather it is due to the fact that after the separate films merge all molecules within the film are more strongly held than before merging. Second, the Gleysteen and Deitz explanation applies only to hysteresis in desorption from capillaries, not to desorption from a plane surface. Finally, they take no account of capillaries that may be emptied without hysteresis. Joyner and Emmett (9) have determined isosterically the excess heat of des By point B we mean the beginning of the linear portion of the isotherm, which is generally assumed t o be closely related to the completion of the first layer.
790
COSWAY PIERCE AXD R. NELSON SMITH
sorption for nitrogen on a porous glass sample. Over most of the hysteresis loop the excess heat is some 200-250 cal./mole, but it is not constant. Even with these small heat values there is obtained quite a large hysteresis loop, Le., a much lower equilibrium pressure for desorption than for adsorption. We believe their data to be quite in keeping with our theory that after the capillary has filled by merging of separate films all molecules are more tightly held than when the films are separate; the small excess heat values shoT7 that a small additional force on the molecules may appreciably lower their equilibrium vapor pressures. CLASSIFICATIOX O F CAPILLARIES
I t is convenient to classify the adsorption-desorption characteristics of capillaries according to their method of filling and their hysteresis properties. Class I: h capillary 1% hich fills at the low relative pressures usually associated with monomolecular adsorption and which shows no desorption hysteresis is designated as Class I. Such capillaries probably do not exceed a few molecular diameters in width. The walls are so close together that their cooperative effect lowers the vapor pressure of contained molecules to the relative pressure region usually associated with adsorption in the first layer on plane surfaces. The isotherms for Class I capillaries are of Type I. An example is the ethyl chloride isotherm of figure 3. In the past the interpretation of such isotherms has been that the capillaries are so narrow that the adsorption of a monolayer on each wall will fill the opening. But, as shown in a previous paper (12), there is considerable evidence that such capillaries are wider than two molecular diameters; in the case of charcoal S84 the width must be a t least 12-15 molecular diameters. (The preceding paper should be consulted for details.) Class 11: When a capillary fills by first adsorbing separate multimolecular clumps or films which later merge it shows desorption hysteresis. Such capillaries are designated as Class 11. They fill only in the higher relative pressure region, since multilayer films are formed only a t higher relative pressures. Their isotherms are usually of Type IV or V. Some capillaries may act as Class I with well-adsorbed molecules and Class I1 with poorly held molecules. The capillaries of charcoal S8-1 are examples of such. Toward ethyl chloride they behave as Class I but toward water as Class 11, as shown in the isotherms of figure 3. CRITERIA FOR DETECTION O F CAPILLhRY ADSORPTIOX
It has been assumed that when there is capillary condensation of adsorbate there is desorption hysteresis. Conversely, the existence of hysteresis is believed to indicate capillarity. Since, however, Class I capillaries do not give isotherms that exhibit hysteresis and since hysteresis may occur in the absence of capillaries, me should reexamine the criteria by which it is deduced whether a given adsorbent is porous. There are several lines of evidence that should be considered for such an evaluation.
ADSORPTIOPi-DESORPTIOS
HYSTERESIS
791
1. If an isotherm is of Type I it is almost conclusive evidence that there are narrow, Class I capillaries. We believe that such capillaries are several molecular diameters in width. 2. The existence of a Type IT’ or Type V isotherm indicates capillaries 11-hose filling prevents further multilayer adsorption and causes the isotherm to flatten at high relative pressures. Such isotherms show desorption hysteresis in all cases. 3. In Type I1 isotherms a comparison of the T‘,-value (point B ) with the adsorption near po, V8, may sometimes indicate the presence of small Class I capillaries that fill at low relative pressure. As shown by figure 1, multilayers often build up to as many as 20-30 or more statistical layers, as p , is approached. If there are Class I capillaries they add to the low-pressure adsorption and make the apparent value of T’, much higher than the true monolayer value. Consequently, the ratio VJT’, is made to become too small. When this ratio is smaller than 10-20 there is reason to suspect Class I capillaries. Some adsorbents appear to possess both Class I and Class I1 capillaries. They yield a Type IV isotherm and such an isotherm does not indicate whether the low-pressure adsorption is due to formation of a monolayer or to capillary condensation. This is the situation for the charcoals studied by Gleysteen and Deitz (6). We suspect, because of the high apparent T’,-values, that these charcoals have small pores that become filled at IOTV relative pressure and that the true surface area is much smaller than that estimated from the adsorption a t point B. 4. Use of different adsorbates may sometimes aid in detecting the presence of capillaries. For example, capillaries in charcoal which behave as Class I towards most adsorbates may be Class I1 with respect to water and thereby their presence can be recognized. One might use Tvater adsorption to show whether or not the low-relative-pressure nitrogen adsorption found by Gleysteen and Deitz is due to capillaries. 5 . The existence of hysteresis in the high-pressure region of a Type I1 isotherm indicates the presence of very wide capillaries. A case in point is graphite SC-1. Its ethyl chloride isotherm (11) shows some hysteresis at very high relatire pressures. Severtheless the isotherm does not flatten as p , is approached. Therefore 11-e reason that either there are few capillaries and most of the adsorption is on exposed surface, or else that the capillaries are so wide that they fill only at pressures very near p , . The suspected presence of large capillaries in SC-1 was further confirmed by determination of a water isotherm, shorn in figure 4.This isotherm shows more hysteresis than was found for Graphon, i.e., the hysteresis loop persists to lower relative pressure than was the case with Graphon. Both the r a t e r and the ethyl chloride isotherms therefore lead us to suspect, contrary to our previous opinion, that sample YC-1 does have some large capillaries. Graphon, however, is thought to be nonporous. Sone of its isotherms indicate porosity.
792
CONWAY PIERCE AND R. XELSOK SMITH APPLICABILITY O F THE KELVIN EQUATION TO PORE-SIZE DISTRIBUTION
Sumerous estimations of capillary pore sizes have been made by application of the Kelvin equation p - -2uu cos e In- rRT Po
to adsorption or desorption isotherms. Here e is the angle of wetting, u is the surface tension, c the molar volume, r the capillary radius, R the gas constant, and T the absolute temperature. Previously (12) we questioned the application of this equation for menisci of molecular dimensions. At the present time we are even more skeptical, particularly as to use of the Kelvin equation for water isotherms of charcoals. As shown above, water and ethyl chloride fill the same capillaries. If the Kelvin equation were applicable to either, one would think it more so to ethyl chloride than to water isotherms because the former apparently wets all the surface, the latter only certain active sites. I t is implicit in the equation that adsorbate molecules be strongly held at all surface atoms, since the equation relates vapor pressure lowering to the curvature of a meniscus, which is caused by the attraction of surface for adsorbate. Ethyl chloride is much more strongly attracted by carbon than is water, as shown by comparison of their isotherms. But it happens that application of the Kelvin equation to ethyl chloride or similar isotherms leads to computed radii smaller than one molecular diameter, while application to water isotherms does give computed radii of the order of 3-5 molecular diameters. This we believe to be misleading. There is more justification for using the Kelvin equation for wide capillaries that fill by merging of multilayer films from opposite walls, if an adsorbate is selected which wets the surface well. Here the meniscus is wide in comparison to the diameter of a single molecule and its radius of curvature has more meaning than for capillaries only a few molecular diameters in width. Emmett and Cines ( 5 ) have studied glasses with wide capillaries, comparing the pore diameters computed by the Kelvin equation with those computed by the surface-volume ratio. The isotherms are all of Type IV. They find that radii computed by the Kelvin equation applied to the steepest part of the adsorption isotherm are in fairly good agreement among the various adsorbates used, all of Tvhich appear to wet the surface well. The Kelvin radii tend to be somewhat larger than r , , the radius computed from the surface-volume ratio. One cannot, however, say that the agreement proires the correctness of the estimated radii. There are two factors which are not taken into account. One is that the Kelvin equation is for cylindrical pores. I t must be modified for pores of other shapes. The other factor, of perhaps greater importance, is that the r,-values are based on the assumption that all of the adsorption at point B is monomolecular, i.e., that V , is determined by the point B adsorption. If there are any Class I capillaries which are filled at low relative pressure the effect is to make the V,-value too large and therefor? the r,-value too small. One cannot tell from the isotherms whether such Class I capillaries are present or not.
ADSORPTIOK-DESORPTION
HYSTERESIS
793
R e feel, therefore, that there is not yet any reliable method for estimation of pore sizes and it is possible that all present estimates are not even correct as to the order of magnitude. Indeed, if one considers the possible mechanisms for pore formation in such materials as charcoal and silica gel it appears that pores must be highly irregular openings as to both size and shape. It may well be that except for saturation values, n-hich are determined by the total pore volume, other properties of isotherms are determined by pore vidths at constrictions and not by average pore vidths. If so, the concept of average width ceases to have meaning. SCAXXING O F ISOTHERMS
The proposed theory affords a satisfactory qualitative explanation of the phenomena observed when a hysteresis loop is scanned. For simplicity we shall consider first a single large pore of uniform cross section and surface activity. At the start of adsorption there is formation of the first layer, at 1017 relative pressure; the isotherm is just like that for a plane surface. As pressure is increased, a multilayer is formed on each wall-again just as for a plane surface. At the relative pressure where the multilayers on opposite walls become thick enough to meet, the pore becomes filled. S o further adsorption can occur, and the isotherm now becomes flat. At the point where the films meet there is a decrease in the vapor pressure of all molecules in the pore. If now pressure is reduced no evaporation occurs until the equilibrium pressure of the filled pore is reached. At this point an infinitesimal decrease in pressure causes evaporation and the desorption branch drops almost vertically until it meets the adsorption branch. This may occur when only a monolayer remains on each vall or the meniscus may break at a point when the adsorbed film on each wall is somewhat greater than monomolecular. If the ideal capillary is partially emptied by desorption, readsorption will follow the desorption branch. That is, condensation will occur a t the meniscus until the capillary is again filled; the vapor pressure at the meniscus is lower than that required to form separate films on each wall, since the meniscus molecules are subjected to forces from both walls. In other words, for a single uniform capillary we do not abtain the usual type of scanning loop. Now, in an actual sample there is a distribution of pore widths and of surface activity and each pore has irregular v-idth. The adsorption branch tends to become flatter than for a nonporous adsorbent because the smaller pores fill and their surface disappears while there is still multilayer adsorption on the walls of wide pores. Eventually all pores fill and the isotherm becomes flat. Now on desorption the drop is not completely vertical, for the larger pores have highest vapor pressures and start to desorb before the narrower ones. Severtheless one can obtain isotherms in nhich there is quite a sharp drop, such as shown by Emmett and Cines (5). If after partial desorption one now begins a readsorption, there will be two types of pores present. Some will be empty, except for a monolayer on each wall; others will still have a wall-to-wall meniscus. The empty pores will adsorb
794
CONWAY PIERCE AND R. NELSON SMITH
separate multilayer films on opposite walls, a t the equilibrium pressures that correspond to the adsorption branch of the isotherm. Those pores that still have a wall-to-wall meniscus will have vapor pressures corresponding to the desorption branch. The resultant effect will be that found in scanning a loop; the new adsorption branch will lie between the original adsorption and desorption branches. As the partially empty pores again become full the isotherm approaches the original adsorption branch, which it finally meets. On desorption scanning, a reverse situation is true; some pores are filled, others have unjoined films. Here too the scanning branch lies between the adsorption and desorption loops. SUMMARY
It is shown that desorption hysteresis does not require prior capillary condensation, but may also occur when adsorbate is held on a plane surface. Hysteresis is found when adsorption occurs by merging of clumps of molecules on separate sites. After clumps merge all molecules are held by forces from all active sites touched by the merged clump. Forces are therefore stronger than when the clumps were separate; the vapor pressures of all molecules are lowered and there is desorption hysteresis. The most common case of hysteresis is when capillaries are filled by the meeting of multilayer films on opposite walls. After the films meet, all molecules are held by surface forces from both walls. Small capillaries may fill by condensation nhich starts at the narrowest places, owing to effects of both walls on lowering the vapor pressure in the meniscus. Such capillaries undergo no change of the adsorbed layer as condensation proceeds and show no hysteresis. They are designated as ClassI. Capillaries which fill by meeting of multilayer films on opposite walls are called Class 11. They show hysteresis. Criteria are suggested for the recognition of capillary condensation in adsorption. I n some cases it is not possible to tell from isotherms xhether or not there is capillary condensation. The validity of using the Kelvin equation to compute pore sizes is discussed. An explanation is offered for the phenomena observed in scanning a hysteresis loop. REFEREKCES (1) BASFORD, JURA, A S D HARKINS: J. Am. Chem. SOC. 70. 1444 (1948). (2) BEEBE,BISCOE,SMITH, A N D WENDELL: J. .h.Chem. SOC. 89, 95 (1947). (3) BRCNAUER: The A d s o r p t i o n of Gases a n d V a p o r s . Princeton University Press, Princeton, Yew Jersey (1943). (1) BRUNACER, DEMING,D E ~ I I N G ASD , TELLER:J. Am. Chem. SOC.62, 1723 (1940). ( 5 ) EMMETT A N D CINES:J . Phys. & Colloid Chern. 61, 1248 (1917). A N D DEITZ:J. Research S a t l . Bur. Standards 36, 285 (1945). (6) GLEYSTEEX (7) HILL: J. Chem. Phys. 16, 767 (1947). J. Am. Chern. Soc. 70, 2353 (1948). (8) JOYKER A N D EMMETT: (9) JOYNER AND EXMETT: J. .h. Chem. SOC. 70, 2359 (1918). (10) J U H O L A ASD WIIG: J. Am. Chem. Soc. 71, 2069 (1949). (11) PIERCEIND SMITH:J. Phys. & Colloid Chem. 62, 1111 (1948). WILEY, A N D SwTH: J . Phys. & Colloid Chem. 63, 669 (1949). (12) PIERCE, (13) SMITHA N D PIERCE:J. Phys. & Colloid Chem. 62, 1115 (1948).
