L.R. SNYDER
2344
volume of a salt depends only on the total ionic strength of the solution is valid only if d$z/dpw‘/z= d$3/dp,1/2, and a, p, and y are negligible. Iii this case eq. 8 reduces to: Os = 43 ( ~ ‘ / ~ /(d+s/dpcw1/2), 2) and g3 is independent of the value of m2. I n some cases the correction terms partially compensate for the difference in slopes for one member of the pair of electrolytes. For example, in l m solution the partial molal volume of XaC1 is 19.49 ml. in pure KaCl and 19.30 ml. in pure NaC104 solution, while the partial molal volume of NaC104 is 45.65 ml. in pure KaC104 aiid 45.61 in pure XaC1 solution. I n 4.17 m solution the corresponding values for KaC1 are 22.52 and 21.82 ml.; for NaC104, they are 48.34 and 47.92 ml. Choice of Variable.-Before beginning this work, the possibility was considered that there would be smaller deviations from Young’s rule if molar rather than molal concentrations were used. Preliminary experiments using direct density measurements indicated that the volume contractions on mixing two solutions of equal molarity were of the same order of magnitude as in mixing solutioiis of equal molality. Since such contractions do occur, the resulting solution no longer has the same molarity as the components and further corrections would be required. For practical reasons, molal concentrations were used. With the aid of eq. 6 it is possible to calculate by an iterative procedure the deviations between the value of @ a t a given volume concentration and that calculated from Young’s rule using volume concentrations. For example, in a 50-50 mixture of KaC1 and HC1, where the total molarity is 3.8317, the calculated deviation
+
Vol. 67
is -0.110 ml./mole, or almost exactly the same as for 4.1724 m solutions. (A molarity of 3.8317 corresponds to m = 4.1724 in pure NaC1 and m = 4.1586 in pure HCl.) For a NaC104-HC104 mixture a t a molarity of 3.4841 (corresponding to 4.1724 m NaC104 and 4.1268 m Hclod), the deviation is -0.027 ml./mole for a 50-50 mixture, which is a third larger than for equimolal solutions, but the deviations were more nearly symmetrical about a mole fraction of 0.5 (maximum near 0.57 mole fraction of HC104). Calculation of Density.-The density of a solution containing any possible combination of the four electrolytes, NaC1, HCl, NaC104, and HC104, can be obtained at 25” in the range 0-4 m by use of the equation
a=
+
+ + + + +
1000 (mzMz madla . .) 1002.93 (mz m3 . .)@
If @ is evaluated by Young’s unmodified rule, using eq. 2 and the constants in Table IT’, the maximum error in the density in a most unfavorable case (a 50-50 mixture of KaC104-HC1 in 4 m solution) would be about 2 parts per 1000. If eq. 6 with the constltnts in Tables I11 and IV is used to evaluate CP, then the maximum error would be of the order of 2 parts in 10,000 (0.02%). Acknowledgment.-The authors wish to thank Mr. Jeffrey Greenhouse for preparing the computer program for many of the calculations. This work was supported in part by the National Science Foundation, Crant G-14623.
ADSORPTION FROM SOLUTION. 111. DERIVATIVES OF PYRIDINE, ANILINE, AND PYRROLE ON ALUMINA BY L. R. SNYDER Union Oil Company of California, Union Research Center, Brea, California Received M a y 23, 19623 Linsar isotherm free energies of adsorption from n-pentane onto 3.6% HzO-AlzOa are reported for 66 nitrogen compounds related to pyridine, aniline, or pyrrole. A previously developed theoretical model permits the calculation of the nitrogen group adsorption energy for each adsorbate, free from the “normal” contributions to total adsorption energy by other adsorbate groups. I t is concluded that the nitrogen group in the pyridines and rinilines adsorbs with n-electron transfer to an adsorbent site, while the pyrrole nitrogen group adsorbs with proton transfer to the alumina surface. The localization or anchoring of strongly adsorbing adsorbate groups on the adsorbent surface is also discussed.
Introduction Recent communication^^-^ have drawn attention to certain regularities in the adsorption on alumina of the substituted pyridines and related aza aromatics. The contribution of the nitrogen atom in these adsorbates to total adsorption energy is markedly sensitive to the steric environment about the nitrogen atom, adsorption energy decreasing with increased crowding of the nitrogen. This observation has led Klemm2 to post,ulate that the nitrogen atom(s) in the less crowded aza aromatics serves as an “anchoring” group (a concept first introduced by Zechmeister4), with the remainder of (1) L. R. Snyder, J . Chromatog., 6,22 (1961). (2) L. H.Klemm, E. P. Antoniades, G. Capp, E. Chiang, and E. Y . K. Mak, ibid., 6,420 (1961). (3) L. R. Snyder, ibid., 8 , 319 (1962). (4) L.Zechmeister, Ddscusskone Faraday Soc., 1 , 54 (1949).
the adsorbate only loosely attached to the adsorbent surface. The sensitivity of the nitrogen atom adsorption energy to crowding by adjacent substituent groups is regarded by Klemm as resulting from the interference by such groups to a preferred tilted or edgewise configuration of the adsorbate relative to the plane of the adsorbent surface (presumably for optimum interaction of nitrogen and surface site). Klemm also has proposed that the interaction between nitrogen and adsorbent is the result of charge-transfer complex formation involving the nitrogen n electrons, on the basis of spectral evidence for the greater polarizability of the nitrogen n electrons and by analogy with the previously postulatedK a--complexation of aromatic hydrocarbons adsorbed on alumina. A previous study of the variation of adsorption energy
Nov. , 1963
ADSORPTION OF PYRIDINE, AXILINE,AND PYRROLE DEELIVATIVES ON ALUMIKA
2345
on alumina with adsorbate structure3 appears to verify the concept of a n anchoring group in adsorbates possessing sufficiently strongly adsorbing groups. It was shown that adsorbates with no strongly adsorbing groups have adsorption energies which are the simple sum of adsorbate group contributions ( i e . , no anchoring). I n the case of adsorbates with one or more strongly adsorbing groups, anchoring or “localization” of the strongest adsorbate group k occurs, and the resultant “delocalization” of remaining adsorbate groups (from a “normal” semilocalized or localized state) decreases their individual contributions to adsorbate adsorption energy. Reduction in adsorption energy of the delocalized groups i becomes appreciable a t a threshold value of the adsorption energy of k (corresponding to the onset of localization of k), increases with further increase in the adsorption energy of k, and finally levels off at a constant factor when the adsorption energy of k has increased to about four times the threshold value (corresponding to complete localization of k). This localization or anchoring behavior is typical both of the anilines and pyridines, as well as of other anchoring groups k (e.g., nitro, acetyl, etc.). While the anchoring group postulate of Klemm and Zechmeister thus appears in agreement with the adsorption of the aza aromatics and other strongly adsorbing compounds, the configuration of the “anchored” adsorbate is more equivocal. Solvent variations tudies3J suggest that the aza-aromatics and other compounds are adsorbed parallel (flat) to the adsorbent surface, rather than tilted or perpendicular. Similarly, analysis of the adsorption energies of certain of the aza-aromatics3 leads to the same conclusion, flat adsorption; all similar groups i in the adsorbed aza aromatics contribute an equal increment to the total adsorption energy (although this iiicrement is smaller in localized adsorbates). This cannot be reconciled with significant tilting of the adsorbate, since nonlocalizing groups i would not then be equidistant from the adsorbent surface as their equal (for similar groups) adsorption energies seem to require. Klemm’s postulate of a vertical adsorbate configuration appears to be based on the interference of substituent groups around the nitrogen with the adsorbent surface in this position. Actually, the sensitivity of the nitrogen adsorption energy to crowding by adjacent groups cannot be used per se to conclude anything about the adsorbate configuration, since both flat and perpendicular situations can be visualized in which steric hindrance to adsorption is possible. Many molecular complexes and adducts of pyridine show similar steric effects (n-complexes with iodine,’ adducts with trimethylaminesls even proton saltss). The existence of analogous n-complexes of pyridine with other electron acceptors’ supports an n-complex structure for the adsorbed pyridine derivatives, as do the apparently related inductive3 and steric effects in these different systems. Arguments of a similar nature have been advanced as evidence for r-complex
formation in the adsorption of the aromatic hydrocarbons, however, while two preceding papers in this serieslO,llhave shown the latter theory to be incorrect. It is, therefore, unclear that the analogous arguments on behalf of an n-complex structure for the adsorbed aza aromatics are valid. I n addition to the uncertainties with respect to both configuration and bond type in the adsorbed pyridine derivatives, similar questions are suggested with respect to the related aniline and pyrrole derivatives. The adsorption energy of the nitrogen group in uncrowded members of these three series of compounds is approximately constant (3.8-4.1 kcal./mole for adsorption from pentane onto 3.i’yO Ha0-A1203),3suggesting the same type of adsorption bonding for the nitrogen in each of these aza aromatic types. A similar adsorption mechanism for the pyridines and pyrroles, however, would appear to rule out an n-complex mechanism, in view of the greatly differing basicities of pyridine and pyrrole; n-complex formation should be promoted by basicity of the nitrogen atom.’ Previous papers in this series and elsewherelsEhave demonstrated the practicality of acquiring adsorption energy data a t low surface coverages (in the linear isotherm region) in various solvent systems and of extrapolating such data to a common solvent basis. This permits the acquisition of adsorption energies for numerous related adsorbates in a single standard state and prepares the way for a precise analysis of the relationships between adsorption affinity and adsorbate structure. The initial intention of the present study was the review and further acquisition of quantitative energy data for adsorption on alumina of a wide range of pyridine, aniline, and pyrrole derivatives. I n addition to clarifying the nature of the configuration and bonding type in the adsorbed aza aromatics, it was hoped that this investigation would provide additional insight into the related problem of adsorbate localization on alumina, as well as furnish further verification of previously proposed concepts in the general area of adsorption from solution. Experimental
15) L. H. Klemm, D. Reed, L. A. Miller, and B. T.Ho, J . OTQ.Chem., 24, 1468 (1959). (6) L. 12 Snyder, J. Chromatog., 8 , 178 (1962). (7) J. N. Chavdburi and S.Basu, Trans. Faraday Xoe., 66, $98 (1939). ( 8 ) H. C. Brown and G K. Barbaras, J . Am. Chem. Soc., 69, 1137 (1947). (9) A. Streitwieser .Jr., “Molecular Orbital Theory for Organia Cbemista,” J o h n I T ‘ley and Solla, Inr.. S e a Y s r k , N. Y . .1061.pp. 419-424.
function,’ and A , is the solute surface volume (Z&).’S~ The eluent strength-adsorbent activity product a e o for the pure solvents was calculated from previously tabulated parameters1.5: 0.000 for n-pentane, 0.118 for CCla, 0.209 for benzene, 0.275 for
All of the presently reported data were derived from chromatographic retentioxz volume measurements carried out as previously,6 with a few exceptions involving the determination of inconveniently large (>lo0 ml./g.) retention volumes. In the latter cases, solute was partially eluted through a column by passage of a large volume V of the eluent under study, the column dissected into several equal segments, and solute in each of the segments determined (by extraction into a very strong eluent and meaaurement in the usual manner). The point on the column dividing the solute band into halves could then be determined by interpolation, and a corresponding Rovalue determined straightforwardly. All retention volume measurements were carried out in the linear isotherm region, with adsorbent loading by sample never exceeding 5 X g./g. Values of the linear equivalent retention volume Ro (ml./g.) were corrected t o a pentane solvent basis (Rp) by means of the previously used relationship
RP --ROIOmEoA~ a
is the adsorbent activity function,’
eo
(1)
is the eluent strength z
(10) L. R. Snyder. J . I’hys. Chena., 67, 234 (1963). (11) L. R. Snyder, thzd., 67, 240 (1963).
L. R. SNYDER
2346
Vol. 67
CHzClz, and 0.411 for dioxane. The corresponding values for the various binary solvents used were evalupted as previouslyaJO from experimental RO values of several solutes in both a pure solvent of known 01eO and the binary solvent to be studied, using eq. 1. Experimental values found were 0.085 for 10% volume CHpClz/pentane, 0.160 for 25% volume CHzClr/pentane, 0.216 for 50% volume CHzClz/pentane, and 0.361 for 50% volume dioxane/pentane. The solute surface volumes A , were calculated as previouslya for all but the pyrrole derivatives. Studies reported in the following section show that the contribution of the pyrrole type -NH- group to the total solute A , value is 2.5 units larger than previously predicted from the area of this group. Free energy of adsorption data from pentane AFp were calculated from the relationship
hence available. The two aromatic hydrocarbons, 1,2,4,5-dibenzpyrene and benzcoronene, had experimental A, values (13.6 and 15.7) in close agreement with calculated values (15.0 and 17.0), as expected. The experimental A , values for the two pyridine derivatives, quinoline (8.0) and acridine (9.8), were in almost exact agreement with calculated values (8.0 and 10.0) for the flat configuration. An edgewise configuration for these pyridine derivatives would have given A, values only slightly larger than half that for flat adsorption. The standard deviation of experimental AF,,values for these two adsorbates in the seven solvents of Table I is only *0.09 kcal./mole, well within AFp = -(log Rp/V,)/2.31RT (2) the experimental uncertainty of these measurements. Va is the previously definedlJO adsorbent surface volume. For The data of Table I thus lend considerable support to the present adsorbent, chromatographically standardized1 3.6% the flat a,dsorption of the pyridine derivatives. HzO-Al208 (Alcoa F-20) prepared and tested as previously,lJO As indicated in the Experimental section, the experiVa was assumed equal to 0.017.' Previously reported retention volume data for adsorption of mental A, values of pyrrole, indole, carbazole, and the several solutes on alumina of similar chromatographic activities two benzcarbazoles (8.1, 10.4, 12.4, 14.3, and 13.3) were (3.7-5.0% H&-Alz03) were used in some instances. These consistently higher than the calculated values (5.5, data were corrected to a 3.6% H&-&03 basis as described 7.5, 9.5, 11.5, and 11.5) for flat adsorption. The previously' average difference in these calculated and experimental log RI = log Vi (ai/a2) log (R2IV2) (3) A, values is both experimentally significant and constant (2.6 & 0.4 std. dev.). Since the flat configuraHere, RI and Rz are retention volumes corresponding to adtion gives the largest theoretical A, value, an explanasorbents 1 and 2, Vl and VZare the respective adsorbent surface volumes V,, and 011 and 012 are the respective adsorbent activities tion other than configuration must be invoked. Pre01. The accuracy of this extrapolation of retention volume values vious studies of the adsorption of various solutes on from one adsorbent to another can be estimated from duplicate both silical2 and Florisilia have shown that strongly determinations of AFp values using the present 3.6% H&A~zO~ adsorbing groups generally exhibit larger contributions for twelve AFp values calculable as above from previously reto A , on these adsorbents than predicted from their ported retention volume data for 3.7% H&Aldh. The standard deviation of these twelve corrected (eq. 3) pairs of data was areas, and that this effect is linked to some sort of anf0.14 kcal./mole. Furthermore, choring or localization phenomena. The accuracy of eq. 1 in the present study was estimated the effect is particularly pronounced in the case of the from multiple determinations of A F p on the same solute using pyrrole nitrogen group. I n the case of experimental two or more solvents. For 36 different solutes involving an average of 3.7 different solvents each, the standard deviation of and calculated A, values for 9-methylcarbazole (10.0 vs. individual values of AFp for each solute from the average was 10.5), the effect vanishes, suggesting a key role of the f0.17 kcal ./mole. hydrogen in this anomalously large A , effect. Assuming that the -NH- group in pyrrole and its derivatives Eluent Variation Studies for some reason contributes an unusually large increMeasurement of Rofor a particular solute-adsorbent (3 units) to A,, the experimental data of Table I ment combination with two or more different eluents permits for the pyrroles support flat adsorption for this series of the experimental evaluation of A, for the adsorbate compounds as well. from eq. 1. The value of A , associated with various possible configurations of the adsorbate relative to the Absorbate Localization adsorbent surface can also be calculated, so that comPrevious studiesi*3~10 of adsorption on alumina show parison of experimental and calculated A, values that the adsorption energies of compounds with no offers the possibility of determining the actual constrongly adsorbing groups can be expressed as the figuration of the adsorbate.6 As already discussed, simple sum of contributions Fi from each adsorbate previous eluent variation studies tentatively suggest a group i flat configuration for the adsorbed aza aromatics. i Inasmuch as the confidence associated with an experi-AFp = Fi (4) mental A , value increases with the range in aeovalues covered and the number of eluents studied, and since I n the case of adsorbates with one or more strongly previous Ro data for both the aza aromatics and referadsorbing groups, a localization term must be added to ence hydrocarbons are largely confined to only two (4) eluent systems (benzene, CCIJ differing in eo by but
+
0.13 unit, additional experimental confirmation of this conclusion is desirable. For nine adsorbates of interest as regards configuration in the adsorbed state, Ro values were obtained for a t least six of the seven solvent systems of Table I, and values of A , were derived by a least squares application of eq. 1. The derived AF, values for these nine adsorbates are summarized in Table I. Since eo varies from 0.00 to 0.42 for these seven solvfmts, an accurate determination of A , and of adsorbate Configuration is
i f k
i
-AFp
=
Fi - f ( F k )
Fi
(5)
The function f ( F k ) varies from zero for small values of Fk to 0.45 for sufficiently large values, where localization is presumed complete. The original derivation of (5) was qualitative rather than rigorous, although it appears to give as good a description of relevant experimental data3 as is possible a t the present time. (12)
(13)
L. R. Snyder, J . Ch~omatog..11, 195 (1963). L. R. Snyder,