MILTONMANESAND I,.J. E. HOFER
584 studiese of silver chloride solutions in pyridine which yielded a value of 8.4 X for K A ~ con~ the , assumption of simple dissociation only, are not corroborated in the present work.
donors of the Petroleum Research Fund, administered by the American Chemical Society, for partial support of this research.
Acknowledgment. Acknowledgment is made to the
(6) S.Bruckenstein and
J. Osugi, J. Phys. Chem., 65,1868(1961).
Application of the Polanyi Adsorption Potential Theory to Adsorption from Solution on Activated Carbon by Milton Manes1& and L. J. E. Hofer’b Mellon Institute, Pittsburgh, Pennsylvania
15813
(Receioed August 8 , 1968)
Liquid-phase adsorption isotherms at 25’ on an activated carbon have been determined, over a wide range of concentrations, for the following systems: Sudan I11 (benzeneazo-p-benzeneazo-&naphthol) in acetone, cyclohexane, carbon tetrachloride, benzene, and carbon disulfide, and Butter Yellow (p-dimethylaminoazobenzene) in methanol, acetonitrile, acetone, 2-propanol, cyclohexane, heptane, benzene, and carbon disulfide. Except for the high capacity range, most of the data can be fitted to a correlation curve determined for the same carbon from gas-phase adsorption measurements, as predicted by the Polanyi adsorption potential theory. The experimental link between liquid and gas-phase adsorption and the relative constancy of the solvent effect, on adsorption (measured in appropriate units) appear to introduce a measure of predictability to at least some liquid-phase adsorption isotherms. Adsorption tends to be weakest in solvents of highest refractive index.
Introduction The Polaiiyi adsorption potential theory2and modifications thereof have been widely applied to gas-phase Ry contrast , application of the theory lo liquid-phase adsorption has been used only in modified form for adsorption of binary liquids8 and apparently not at all for solutes. Since t>henature of the forces 011 adsorbed molecules may be expected to be independent of their state of aggregation, it seemed reasonable to expect that the adsorption isotherms of at least properly chosen solute-solvent systems should conform to a significant degree to the Polanyi theory. Such conformation has been found for the adsorption isotherms of two solutes (Sudan I11 and p-dimethylaminoazobenzene) in a wide variety of solvents. The results provide the expected experimental link between gas-phase and liquid-phase adsorption on activated carbon and, to the extent that they will be confirmed by continuing work, introduce the possibility of predicting adsorption isotherms on activated carbons for a wide variety of systems from minimal data.
Theoretical Section The Polanyi adsorption potential theory for gases may be summarized as follows: within the range of the The Journal o/ Physical Chemistry
attractive forces of the fiolid surface (the “adsorption space”) the potential energy of a given gas is reduced, relative to its value at infinity, by an amount e (the adsorption potential) that for a given gas depends on proximity to the solid surface. Onc can imagine points of equal e t o be joined to form equipotential surfaces that together with the solid surface enclose a volume ~ ( t ) . The plot of ~ ( e ) against E (the “characteristic curve”) depends on the structure of the adsorbent, and no attempt is made to derive it from theory; it is independent of temperature. When the adsorbent, initially under vacuum, is exposed to increasing pressures of gas, (1) (a) Professor, Department of Chemistry, Kent State University, Kent, Ohio: (b) Head, Adsorption Fellowship (Sponsored by Pittsburgh Activated Carbon Division of t h e Calgon Corporation) Rlellon Institute, Carnegie-Mellon University, Pit,tsburgli, Pa. (2) (a) M.Polanyi, Verh. Deut. Physik. Ges., 16, 1012 (1914): 18, 55 (1916): Z . Elektrochem., 26, 370 (1920); (b) M. Polanyi, Z . Physik, 2 , 111 (1920). (3) M.M.Dubinin, Chem. Rev., 60, 235 (1960). (4) W. K. Lewis, E. R . Gilliland, B. Chertow, and W. P. Cadogan, I n d . Eng. Chein., 42, 1319 (1950). (5) R. J. Grant, 31. Manes, and S. B. Smith, A . 1 . C h . E . J., 8 , 403 (19G2). (6) R. J. Grant and M. Manes, I n d . Eng. Chem. Fundamentals. 3 , 221 (1964). (7) R.J. Grant and M. Manes, ibid., 5, 490 (1966). (8) R. S. Hansen and W. V. Fackler, J . P h y s . Chein., 57, 634 (1053).
ADSORPTIONFROM SOLUTION ON ACTIVATEDCARBON the attractive forces of the solid for the gas molecules reinforce their attraction for each other, with the result that the gas liquefies between the solid surface and that equipotential surface for which e(v)
=2
RT In p s / p
v=/@ where 0 is his “affinity coefficient.” Dubinin and Timofeevg have compared the experimental affinity coefficients of a set of gases with the corresponding molar volumes (1’) and with several different estimates of the niolar polarizabilities, and have concluded that the affinity ratios of different pairs of gases are best approximated by the ratio of their molar volumes. Lewis, Gilliland, Chertow, and Cadogan4 and Grant Manes, and Smith5 found that within the homologous series of saturated hydrocarbons on activated carbon, the plots of volume adsorbed vs. E/TT collapsed to a single curve (the “generalized correlation curve”) with considerable accuracy. However, Dubinin and Timofeev found that the discrepancy between molar volume ratios and affinity coefficient ratios could be as much as 20yo (some of which could have been due to specific chemical effects). Similarly, Grant, Manes, and Smith5 found that at equal adsorbate volumes E / V was some 10% higher for carboii disulfide than for the hydrocarbon series. Anticipating that these differences turn out to be quite significant, we now consider adsorption from solution. Polanyi2”originally supposed that adsorption of solid solutes from solution mould be analogous to the adsorption of gases with precipitation of solid taking the place of liquefaction of gas, and with the adsorption potential now estimated as
RT 111 cO/c
solute would have to be accompanied by the desorption of an equal volume of solvent. As a result, eq 1 would have to be modified to read (with a slight change in notation)
(1)
where p is the (equilibrium) pressure of the gas and p , the vapor pressure of the corresponding liquid at the equilibrium temperature. Given an adsorption isotherm over some capacity range, one can calculate the characteristic curve over the same capacity range by use of eq 1 and an estimate of the density of the presumed liquid adsorbate. Having determined the characteristic curves one can now calculate adsorption isotherms at other temperatures. If the adsorptive forces for different gases are of the same nature, then one may expect that the characteristic curves for different gases on a single adsorbate should all be the same except for a constant factor multiplying the adsorption potential. Dubinin3 expressed this in the equation
eS(u) =
585
(3) where c0 and c are the saturated and the equilibrium concentrations and es is the adsorption potential of the pure solute adsorbing as a vapor. However, in a subsequent paper,2bhe noted that the adsorption of a solid
(4) where €1 is the adsorption potential of the solvent and V , and VI, the molar valumes of (solid) solute and solvent. Polanyi concluded from eq 4 that adsorption of a solute would be weakest from solvents with the highest values of elVg/V~(or el/Vl for constant VO). He then drew attention to an observation of Freundlich to the effect that adsorption tends to be weakest in those solvents that are themselves most strongly adsorbed from solution. He noted further that since the various solvents would differ only slightly in their values of elVs/V1, one would expect that the equilibrium concentrations a t equal capacities for various solvents would differ largely because of differences in solubilities, as had been found by Davislo in studies on the adsorption of iodine on carbon. We shall return to these points later. Let us now consider some of the consequences of the Polanyi adsorption potential theory for solutions in terms of plots of volume adsorbed vs. the adsorption potential per unit volume, ie., in terms of “generalized correlation curves.” We can rewrite eq 4 as (5)
which we can again rewrite somewhat more compactly as N,1
=
Ng
-
a1
(8)
where each CY is the corresponding e/T’. If we assume, with Polanyi, that the characteristic curves for the solvent and the solute have the same functional form (except for n constant factor in the adsorption potential), then the CY’S are directly proportional to the affinity coefficients of D u b i n i i ~ ~Considcr nom some of the consequence of eq G by reference to Figure 1, which shows schematic plots of volume adsorbed vs. adsorption potential on a scale that corresponds roughly to some experimental observations on activated carlx~n. XI1 of the curves in Figure 1 are drawn so that they can be collapsed to a single curve by application of a singlc abscissa scale faclor. The deviation from unity of the scale factor required to make two such curves coincidc is a measure of the extent t o which the adsorption potential deviates from proportionality to the niolar volume. We assume that the scale factor cy (or the affinity coefficient P ) of the solute is signif(9) M. M. Dubinin and D. P. Timofeev, Compt. Rend. Acad. Sci. U R S S , 54, 701 (1946). (10) 0 , C. M. Davis, Trans. Chem. S O L , 91, 1966 (1907).
Volume 75,Number S March la69
MILTONMANESAND L. J. E. HOFER
586
-
CORRELATION CURVES (SCHEMATIC)
-
Or YSl
=
Y E
-
71
(8)
where ySl = a,1/a1,, etc. If the reference substance is the solvent, t>henys = ysl 3- 1.
Experimental Section
icantly higher than that of any of the solvents (otherwise the Polanyi theory would predict stronger adsorption of the solvent). The correlation curves for the gas-phase adsorption of two solvents and the corresponding correlation curves for adsorption of solute in the liquid solvents are also shown, The larger the a1 for the solvent, the smaller the asl for the solution. Now PolanyiZbpointed out that one could verify eq 4 for volatile solutes (such as iodine) by separately determining agand 011 by gas-phase adsorption followed by the determination of aSlin liquid-phase experiments. However, most solutes of interest are not volatile, and one cannot therefore determine as independently. A somewhat less direct check on the theory would be to determine the differences between the asl for a single solute in different solvents and to see whether these differences would correspond to the differences in a1 as determined from vapor-phase measurements. Moreover, in the absence of the requisite vapor-phase data (which might be impossible to determine for solvents of low volatility) , one could determine a,l for a number of solutes in a set of solvents and observe whether the differences in ael between solvents (for a given solute) would remain constant for all solutes. Instead of expressing the relative values of the a’s at some stated adsorptioii volume, we can consider the ratios between the a’s and ah, where a h is the scale factor for some standard reference substance (or substances, if they all have the same a ) , in which case we can rewrite eq G as
The Journal of Physical Chemistrz,
Since, as noted earlier, all of the foregoing discussion is based on the assumption that the expected link between liquid-phase and gas-phase adsorption in fact exists, an initial objective mas to gather sufficient liquid-phase and gas-phase adsorption data on an activated carbon for comparing the resulting correlation curves. The principal problem was to find systems for which one could determine the liquid-phase adsorption isotherms down to the low capacity range, in order to permit a critical comparison of the resulting correlation curves. Oil-soluble dyes of relatively low molecular weight were originally sought as candidate solutes because of ease of spectrophotometric analysis over a wide range of concentrations (typically five decades), solubility in a wide variety of solvents, and expected relative ease of purification. The work reported here is based on the liquid-phase adsorption data for two azo dyes, Sudan I11 (benzene-azo-p-benzene-azo-pnaphthol) and Butter Yellow (p-dimethylaminoazobenzeiie) . Materials. Commercial Sudan I11 was recrystallized from acetone-toluene. Butter Yellow was recrystallized from ethanol; treatment of the concentrate with activated carbon was necessary to achieve the desired spectral purity. Successive recrystallizations did not significantly change the optical extinction coefficients. The largest experimental error due to impurities was expected to be in the possible change in composition of the equilibrium solutioii on adsorption; therefore, initial experiments were carried out in which the optical absorption curve for the solution remaining after adsorption of 9970 of a sample on activated carbon TTas indistinguishable from a similar curve determined on a 100: 1 dilution of a similar sample. The adsorbent in all cases came from a single batch of Pittsburgh Activated Carbon grade CAT, activated carbon, which was pulverized to pass 200 mesh. This carbon is normally used for removal of color bodies from solutions (e.g., cane and invert sugar) ; it \vas chosen for maximum availabilit!: of its surface area i o the solute dyes. Except for oven drying before weighing, the carbon was used as received. The surface area (BET) was 1140 m2/g. The solvents used, methanol, acetonitrile, acetone, 2-propanol, n-heptane, cyclohexane, benzene, carbon tetrachloride, and carbon disulfide, were in every caw Fisher Spectranalyzed reagents. Procedures. Equilibration took place in 125-1111 screw-capped erleixneyer flasks, the screw caps being sealed with Teflon gaskets. The flasks were shaken a t
ADSORPTIONFROM SOLUTION ON ACTIVATED CARBOX least 16 hr at 25" in a thermostated shaker bath; check experiments at longer shaking times established that the shaking time sufficed for equilibration. After equilibration the carbon was allowed to settle out and a portion of the supernatant liquid v a s cleared by centrifuging. The cleared equilibrium solution, after suitable dilution, was analyzed in a Cary spectrophotometer. The sample size and initial concentration of solution were so adjusted that no more than 99% of the original sample was adsorbed. The relative amounts of carbon, dye, and solvent were so chosen that the adsorption of solvent did not introduce a significant error into the calculations. Solubilities of the dyes in the individual solvents were determined by spectrophotometric analysis of saturated solutions. Addition of a small amount of activated carbon t o the equilibrium mixture improved the reproducibility of the solubility measurements. The densities of the pure dyes (necessary for calculating correlation curves) as determined at 23" by helium displacement mere: Sudan 111, 1.302 g/cc; Butter Yellow, 1.208 g/cc. The gas-phase correlation curve was determined from adsorption isotherm data for ethane, propane, and butane on a :\IcBain balance, using the methods described by Grant and
587
I
I I I P-DIMETHYLAMINOAZOBENZENE ON C A L CARBON
50.0 -
? -
-
>E 10.0 -5
u
$
50
ile
m 0
0 \
0.1 I 10-3
I
10-4
I 10-3
I
10-2
I
1 0 ' '
100
Relative Solubility (C/Cs)
Figure 3. Isotherms of Butter Yellow on CAL carbon from various solutions. (C, is the solubility and C is the Concentration of dye in equilibrium with the carbon for the loading indicated.)
Data and Results Figures 2 and 3 give the adsorption data for Sudan I11 and Butter Yellow, plotted on a log-log plot as g of dye/100 g of carbon us. the relative concentration c/c,. 100.0c
I
L
50.0
1
1
4
1
S U D A N I U ON C A L CARBON
I
$ 10.01 -c
I
C
