(6) Beroza, Morton, Sarmiento, Rafael, ANAL.CHEM.35, 1353 (1963). (7) Cowan, C. B., Stirling, P. H., “Gas Chromatography,” 2’. J. Coates, et al., eds., p. 165>Academic Press, Sew York, 1958. (8) navies, A. D., Homrd, G. A,, J . A p p l . Chem. 8, 183 (1958). 19) Ilrawert. F.. Feleenhauer. R. Anoew. Chem. 72,555 (19Gi). (10) Drnaert, F , Reuther, K. H., Chem. Ber. 93, 3068 11960). (11) Hall, FV. K Emmett, P. H., J . A m . Chem. SOC.79, 2091 (1957). (12) Haslam, J., Jeffs, A. R., Rillis, H. A,, Analyst 86, 1018 (1961). ~
(13) ,Ingold, C. K., “Structure and Mechanism in Organic Chemistry,” p. 182, Cornel1 Cniv. Press, Ithaca, S . Y., 1953. (14) Janak, J., Xature 185, 684 (1960). (15) Keulemans, A. I. AT., 1-oge, H. H., J . A m . Chem. SOC. 63, 476 (1959). (16) Knight, H. S., Weiss, F. T., A I ~ A L . CHEM.,34,749 (1962). (17) Kokes, R . J., Tobin, H., Jr., Emmett, P. H., J . Am. Chem. SOC. 77, 5860 (1955). (18) Miller, D. O., Ari.4~.CHEM.35, 2033 (1963). (19) Mourgues, L. de, Chim. ,4naI. 45 (3), 103 (1963).
(20) Okamoto, T., Tadamasa, O., Chem. Pharm. Bull. 1 1 , 1086 (1963). (21) Poe, R. W.,Kaelble, E. F., J . Am. 022 Chemists’ SOC.40, 347 (1963). (22) Radell, E. A , , Strutz, H. C., ANAL. CHEM.183, 1671 (1959). (23) Sykes, P., “4 Guidebook t o Mechanism in Organic Chemistry,” p. 10, Wiley, New York, 1961. (24) Zlatkis, A,, Oro, J. F., Kimball, A. P., A N A L . CHEM. 32, 162 (1960). R E C E ~ V EforD review February 25, 196.1. Accepted May 5 , 1964. Mention.of a proprietary product does not constitute an endorsement by the U. S.Department of Agriculture.
Study of Solid Support and Partition Liquid Interactions In Gas Chromatographic Separation of Ethanol-Methanol Mixtures EDGAR D. SMITH, JUNIOR L. JOHNSON,’ and J. M. OATHOUT* University o f Arkansas, Graduate Institute o f Technology, Little Rock, Ark. The use of independently determined solid support and partition liquid selectivities has been studied in connection with the separation of ethanol and methanol. Superior columns for this separation were readily achieved, with either alcohol being selectively retained through the proper choice of the solid support and partition liquid. The separations reported should be of value in analyzing trace quantities of either alcohol in a preponderance of the other, though problems still remain from residual gas-solid adsorption effects. Typical sensitivity limits obtained in this study were 10 p.p.m. for ethanol in methanol, and 100 p.p.m. for methanol in ethanol. Use of a more sensitive detector and improvements in instrument design would undoubtedly permit these limits to be lowered.
I
K AX E.4RLIER ARTICLE, methods were
described for the independent determination of relative values of substrate and partition liquid selectivities ( 7 ) . I t was shown that superior columns for the separation of 2- and 3-pentanone could be prepared by choosing a solid support, and partition liquid which acted in unison t,o bring about the desired separation. The order of elution of these close boiling pentanone isomers could be reversed with practically base line separation by either sequence. The present work was undertaken to 1 Present address, Dowsmith, Inc., Little Rock, .4rk. 2 Present address, Hendrix College, Convay, Ark.
175 0
ANALYTICAL CHEMISTRY
determine whether similar results could be achieved using ethanol and methanol as model compounds. Since ethanol boils about 14’ C. higher than methanol, it is, of course, a simple matter to develop packings capable of separating these alcohols in order of their boiling point. Many such packings have been reported; inert types of substrates generally are used to avoid the excessive tailing that is usually noted with more active solid supports (3, 4,9 ) . Since this work was completed, Rombaugh and Thomason have reported a separation of these alcohols in the reverse order of elution ( 1 ) . d highly selective liquid partition phase was employed along with an especially acetylated grade of Chromosorb IV. Tailing of the alcohols was practically eliminated on this acetylated support, thus allowing the detection of parts-per-million quantities of ethanol in methanol. Since this article did not report relative retention data for the alcohols on the bare solid supports, it is not possible to sag whether their inherent selectivity characterihtics were also affected by this treatment. In the present work, the selectivity characteristics of both the solid supports and the partition phases were evaluated so that the interaction of these two important variables could be assessed. EXPERIMENTAL
-1 Perkin-Elmer Model 154-L vapor fractometer having a thermistor-type detector mas used with a Leeds & Sorthrup Speedomax H recorder. The vapor fractoineter was modified by connecting a small ponerstat in the heating circuit to control independently
the injection block temperature. Helium was used as the carrier gas. One-fourth-inch copper tubing, 2 meters in length, was used for all screening columns with the exception of some of the bare solid support columns, where intense adsorption made the use of 1meter lengths necessary. The solid supports investigated were commercial materials sold under the names of firebrick, Chromosorb P, Chromosorb IT-> and Gas-Chrom Z. The first three were 60- to 80-mesh material purchased from Wilkens Instrument and Research, Inc., while the’last named was 80- to 100-mesh material purchased from Applied Science Laboratories, Inc. Partition liquids were obtained from various commercial sources and used without purification. Fisher reagent grade methanol and anhydrous USP ethanol (undenatured) were used as solutes. d Hamilton 701-SCH microsyringe equipped with a Chaney adapter was used to inject these solutes. The conditioning and screening procedures used in this work were the same as those described in reference 7 . RESULTS A N D DISCUSSION
Table I summarizes the relative selectivity characteristics of the various treated and untreated solid supports. These supports are listed in the order of decreasing alpha values (corrected retention time ratios) of the untreated support. In all cases, injections \yere reported a t exact 10-minute intervals until reproducible value3 of retention times were obtained. Column lengths and temperatures were varied as necessary to obtain retention times which could be measured with reasonable precision (see reference 7 ) . These data
show two major points of interest: there is a marked difference in the selectivity characteristics of firebrick and of Chromosorb P although these are often considered to be equivalent materials; and the effect of adding Carbowax-400 to either of these active supports is to decrease the retention of both alcohols, but to increase the relative retention of ethanol. Since Carbowax-400 itself is less selective for ethanol than either solid support alone (CY = 1.31,see Table 11),this increased retention can best be explained as a result of the covering up of the more active sites, which are methanol-selec tive. Table I1 lists the relative selectivity characteristics of the various partition liquids investigated in order of decreasing alpha value. In this tabulation, the retention times given are all measured from the injection point so that values of S, &, and R could be calculated. Since methanol tailed appreciably more than ethanol on many of these packings, the Phillips expression for R was used along with the corresponding modifications for S and & suggested by Jones and Kieselbach (3, 6 ) . Alpha values were, however, calci,ilated from corrected retention times as before to make these values more sensitive to small differences in the retention times of the alcohols. Inspection of Tables I and I1 suggests that the best separation of ethanol and methanol to elute ethanol last should result from a combination of Chromosorb P modified with 1% Carbowax-400 and one of the partition liquids a t the top of Table 11. The first three of these partition liquids were eliminated because of their well known inability to "cover up" active adsorbent sites and because of their low partition coefficients. Dodecyl alcohol combines the desired features of high selectivity, good covering power, and high partition coefficient. I t was therefore chosen for the separation of the two alcohols in order of their boiling points. Table 111 summarizes the data obtained t o confirm this choice. .U1 of the Chromosorb P columns \\-ere considerably more efrective than the Gas-Chrom Z column, and these data show a slight but significant advantage for the modified Chrornosorb P column. Visual inspection of the chromatograms also showed that the symmetry of both alcohol peaks was improved by the Carbowax 400-addition. An increase in column length increases separation if the flow rate and temperature are held constant. However, it was felt worthwhile to determine if this were also true when the temperature was increased to obtain approximately the same overall analysis time. Table TV provides data on this point for 1- and 2-meter columns and shows that the longer column is preferable. Comparison of this data with that of Table 1x1
Table 1.
Relative Selectivities of Various Support Materials
Support material Chromosorb P, regular Chromosorb P 1y0C.W. 400b Firebrick, regular Firebrick lY0 C.W. 400b Chromosorb W, regular Gas-Chrom Z
+
+
column Retention times," length, minutes meters Methanol Ethanol 1 1 1 1 2 2
Temp.,
2.16 1.42 5 66 1 41 0 45 0 07
1.22 0.66 4 06 0 95 0 36 0 06
c.
a
50 50 50 50 30 30
1.77 2.15 1 39 1 48 1 25 1 17
a Corrected for retention time of air. -411 columns operated a t 80 ml. per minute helium flow. * 1% Carbowax-400 added by dlssolving a weighed quantity in excess methanol, slurrying with a calculated weight of the solid support, then evaporating the methanol.
Table 11.
Relative Selectivities of Representative Partition Liquids.
Partition liquid
Retention time, minutes Methanol Ethanol Air
Sniialane 1 5;S D.C. 200 fluid 1.21 Poly-m- henyl ether, 6-ring 1.54 Dodecy ?alcohol 9.33 Dinonvl Dhthalate 2.85 Flexol"858 5.30 Armeen SD 4.77 l-h~ethyl-5-(2-rnethoxyethyl)tetrazole15.18 C arbow ax-20 M 4.36 Carbowax-400 6.78 Triethglene glycol 23.08 @"-Oxydipropionitrile 12.02 Diglpcerol 16.20 Glycerol 6.36 Sorbitolb 1.60 I
~
2 83 1.98 2.65 18.69 5.20 9.51 8.23 21.43 5.64 8.71 30.84 15.04 16.15 5.48 1.20
a
0 .i0 2 22
0.57 2.20 0.52 0.50 0.55 0.50 0.50 0.55 0.60 0.51 0.54 0.55 0.55 0.57 0.78
2.09 2.06 2.02 1.88 1.81 1.43 1.34 1.31 1.30 1.26 0.99 0.84 0.51
S
Q
0 585 0.483 0.530 0.667 0.585 0.568 0.535 0.341 0.256 0.250 0.288 0,224 0.003 0.149 0.286
5 43
R
3 17 8.39 4.05
3.12 6.28 7.00 5.89 2.21 7.12 4.84 5.84 6.66 8.07 1.61 3 21 2.14
1.65 4.18 4.09 3.35 1.17 2.42 1.24 1.46 1.92 1.81 0.01 0.48 0.61
All data obtained on 2-meter columns containing 2.73 grams of the indicated liquid on Gas-Chrom Z. Operating conditions, 40" C. and 80 ml. per minute helium flow except where noted. Data obtained a t 115" C. and 40 ml. per minute helium flow (see text). 0
Table 111.
Column Efficiency Data for Dodecyl Alcohol on Various Support Materials"
Chromosorb Chromosorb Chromosorb P lY0 GasP Pb Carbowax-400 Chrom Z
+
Retention time, methanol Retention time, ethanol S
B
11.2 21.7 0.63 9.5 6.0
10.7 20.5 0.63 9.5 6.0
11.4 21.9 0.63 9.9 6.2
9.3 18.7 0.667 6.28 4.18
Each column, 2 meters in length, conhined 2.73 grams of dodecgl alcohol on the indicated support material. A411data obtained at 40" C. and 80 ml. per minute helium flow. * Chromosorb P boiled with an aqueous solution containing 2.73 grams of dodecyl alcohol and evaporated to dryness.
also shows that the separation of the alcohols is better at 40" than a t 57" C. The effect of carrier gas flow rate on the column efficiencies was studied next, and an optimum flow rate for this separation was found to be 90 ml. per minute based on both the H E T P and on the resolution of the columri. Figure 1 is a direct tracing of the separation obtained when the column was operated a t the optimum conditions of 40" C. and 90 ml. per minute. This separation
appears to be markedly superior to those previously reported in the literature. For example, the R value obtained by Crone and Katnik was 4.0 us. 6.2 for the separation shown in Figure 1
(8). For the reverse separation of the alcohols to elute methanol last, sorbitol is the obvious choice for the partition liquid (see Table 11). Gas-Chrom Z appears to be the best choice for the solid support (see Table I) but even it is VOL. 36, NO. 9, AUGUST 1964
1751
0
2
4
6
a io isT I M E IN MINUTES
14
16
ia
Figure 1. Typical gas chromatogram of an ethanolmethanol mixture on a 2-meter column of 16.3% dodecyl alcohol on modified Chromosorb P under optimum conditions Ethanol eluted last a = 2.04 S = 0.65 Q = 9.55
PO
0
2
4
Methanol a = S = Q =
R
Table IV.
Effect of Column Length on Separation"
Column length 1 2 meter meters Retention, time, methanol 3.34 3.57 Retention time, ethanol 6 , 4 1 6,32 S 0.63 0.56 7.5 9.9 4.7 5.6 One-meter column operated at 100 ml. per minute and 40' C. Two-meter column operated at same flow rate but at 57" C. to obtain comparable retention time for ethanol. Both columns contained 16.3% dodecyl alcohol on Chromosorb P 1% Carbowax-400.
B
0
+
Table V.
surface area as Chromosorb P and a lower ethanol selectivity, should be the better of these two supports. Table V summarizes data taken to confirm these points. For the final separation, the column length was increased to 4 meters, and the flow rate and temperature optimized a t 40 ml. per minute and 117' C., respectively. Figure 2 is a direct tracing of the chromatogram thus obtained. This separation is roughly equivalent to that obtained by Bombaugh and Thomason on a column nearly three times as long. Finally, one of the reasons for attempting this separation in the two sequences of elution orders vias to obtain columns suitable for analyzing trace quantities of each alcohol in the presence of a large predominance of the other. I t is a generally held opinion that, in such cases, it is desirable that the trace contaminant be eluted ahead of the main constituent to avoid interferences caused by flooding of the detector with this main constituent. This idea was briefly tested by injecting 10-~1.samples
Column Efficiency Data for Sorbitol on Various Support Materials"
Firebrick Chromosorb +I% P+1% Carbowax- Carbowax GasFirebrick 400 400 Chrom Z 4.92 4 92 4.92 2.73 Grams of sorbitol per 2-meter column 3 19 1 60 2 56 2 48 Retention time, methanol 1 20 1 77 2 24 1 86 Retention time, ethanol s 0 35 0 29 0 32 0 34 2 1 4 6 3 7 4 5 1 4 1 6 1 3 0 61 a All data obtained on 2-meter columns operated at 115' C. and 40 ml. per minute helium flow.
1752
ANALYTICAL CHEMISTRY
io
is-
14
Figure 2. Typical gas chromatogram of an ethanolmethanol mixture on a 4-meter column of 3070 sorbitol on modified firebrick under optimum conditions
R = 6.24
actually selective for ethanol. Thus, the separation would have to be brought about by the partition phase and the suppression of the natural selectivity of the solid support. Relatively high liquid loadings were therefore indicated and, since sorbitol is an effective tailing reducer, the more active, higher surface area supports could be considered. Firebrick, having practically the same
a
6
TIME IN MINUTES
-
eluted last
0.60
0.34 6.67 2.20
of ethanol and methanol containing trace quantities of the other alcohol. With the sorbitol column, the assumption outlined above wai correct and about 100 p p.m. of ethanol in methanol could be clearly detected whereas 1000 p.p.m. of methanol in ethanol were required for positive detection. With the dodecyl alcohol column, however 100 p.p.ni. of methanol in ethanol was required for positive detection whereas 10 p.1i.m. of ethanol in methanol could be readily detected. The explanation of these facts may be that small quantities of the alcohols are irreversibly adsorbed by the solid supports, despite the large quantities of liquid coating. Since methanol i q more strongly ad.orbed than ethanol, it is easier to detect ethanol regardless of the column used and the elution sequence. The so-called "repeater" or "ghosting" effect was easily demonstrated in this work and, in this instance, appeared to be due to adsorption on the column rather than in the preheater (6). Thus, these columns do not represent a final answer to the problem of analyzing trace quantities of ethanol and methanol but they do represent a significant advance in this direction. Improved sensitivities should be readily achieved through the use of a more senqitive detector and improved instrument design. The preheater design of the Perkin-Elmer instrument is particularly poor for use with the dodecyl alcohol system a t 40" c. since condensation of large samples may occur in the capillary line leading from the preheater to the column. The answer to this problem is, of coursc, obvious but no attempt was made to correct this difficulty in the present work.
LITERATURE CITED
(1) Bombaugh, K. J., Thomason, W. E., AKAL.CHEM.35, 1452 (1963). (2) Jones, W. L., Kieselbach, R., Ibid., 30, 1590 (1958). ( 3 ) Littlewood, A. B., ,J.Gas Chromatog. 1 , 6 (May 1963). (4)Parker, K. D., Fontan, C. R., Yee,
J. L., Kirk, P. L., ANAL.CHEM.34, 1234 (1962). (5) Phillips, C. S. G., Second Symposium on Gas Chromatography, Amsterdam, May 22, 1958. (6) Smith, E. D., Gosnell, A. B., ANAL CHEM.34,646 (1962). ( 7 ) Smith, E . D., Johnson, J. L., Zbid., 33, 1204 (1963).
(8) Crone, P., University of Colorado,
Boulder, Colo., private communication to E. D. Smith, 1963. (9) Urone, P , , Kathik, R. J., ANAL. CHEM.35,767 (1963). RECEIVED for review December 16, 1963. Accepted May 27,1964. Presented at the Southwest Regional Meeting, ACS, Houston, Texas, December 1963.
Studies of the Liquid Phase Mass-Transfer Term in Gas Chromatography JAMES
K. BARR'
and
DONALD T.
SAWYER
Department of Chemistry, University of California, Riverside, Calif. 92502 The functional dependence of the liquid phase mass-transfer term in gas chromatography has been studied as a function of liquid loading with helium, nitrogen, and argon a s carrier gases. Carrier gas velocities up to 300 cm. per second have been used in these studies. The long stainding assumption that C I is independent of carrier gas has been shown to b e incorrect. With 3-pentanone as the solute, Carbowax400 as the solvent, and helium as the k)* carrier gas the expeicted k / ( l dependence of CI has been verified. However, the use of argon and nitrogen under the same conditions has shown essentially no k dependence for C iexcept a t extremely low loadings. A limiting rate of solute transfer across the gas liquid interface qualitatively accounts for theunusually large lband broadening observed for argon or nitrogen as the carrier gas. The effective depth of the liquid phase appears to be essentially independent of loading when Chromosorb W is used as a support material.
+
I
N GAS C H R O M A T O G R ~ ~ P H I C mass-trans-
fer formulations .the liquid phase mass-transfer contribution is generally the most important at higher carrier velocities and represents the limit,ing factor in the resolut,ion of adjacent elution bands. The band-broadening contributions to t,he t,heoret'ical plate height' are reasonably well understood except for the liquid. phase variance. Purnell (25) summari:zes two forms of the liquid phase t,erni, C i , in which a homogeneous film and a droplet distribution of the liquid are proposed. He suggests t'hat the actual Ci contribution may be some combination of these two models. Giddings, more realistically, has derived expressions for C l which depend on the geometric shape of the Present address, Engineering Physics Laboratory, E. I. du Pont de Xemours and Co., Inc., Wilmington, Del. 19898
pores in t,he solid support and has proposed eight separate expressions which depend on t'he surface st,ructure of the support ( 5 , 6, 8, 24). These expressions range from the uniform liquid-film model to a model in which any known porr distribution and size are allowed. However, no experimental evidence or opinion is given for a preference of one model over the others. More recently Perrett and Purnell ( M j , in a review art'icle, have stated that the form of the expression used t.0 calculate Ci is unkriolvn. Finally, Dal Nogare and Chiu ( 3 ) have presented experimental evidence that the capacityratio ( k j dependence of the C l term coincides with t,he uniform thickness model or the random-size pore model of Giddings (14). In mass-transfer theory, the number of sorpt,ions or desorpt,ions of a solute entering or leaving the liquid phase is inversely proportional to the theoretical plate height. In addition, t,he transfer of the solute across the gas-liquid interface and the desorption step, in the case of surface adsorption, are assumed to be kinetically faster than the limiting diffusion in the gas and liquid phases. This implies that the liquid phaAe masstransfer contribution to plate height will depend only on the partition coefficient of the solute in the liquid phase, the interdiffusion coefficient of that solut'e in the liquid phase, and the amount of the liquid present. Therefore, no rat,e constants appear explicitly in this formulation. Several suggestions exist' in the literature that the form of the rate equation also may contain a kinetically controlled gas liquid interfacial resistance to masstransfer term. Although no experimental evidence was given, Giddings (10) first proposed that simple diffusion steps alone are unlikely without a series of kinetic complications. Later papers by Giddings (11, 1%)expanded this concept t o include all possible rate steps
under chromatographic conditions ( 5 , 9, 13, 15). Recently, a series of measurements using helium carrier gas and glass beads as the support material has been made (16); the authors conclude that there appears to be no interfacial contribution under these conditions. Khan (28) has derived a term for the interfacial contribution to band-broadening in which interfacial transfer rate constants appear explicitly. The present, discussion summarizes research directed a t isolating the liquid phase mass-transfer term from the other terms, studying the functional dependence of Ci on column parameters, deciding which of the many liquid distribution models is realistic under chromatographic conditions, and determining if interfacial transfer really contributes to plat'e height. Several considerations have been important in the accomplishment of this goal. To st.ay in the Henry's law range and obtain a constant partition coefficient, extremely small sample sizes have been used. The practical Ilroblem of band spreading due t'o the finite rate of vaporization of a liquid sample has been avoided by introducing the sample as a vapor of minimum volume. To avoid longitudinal diffusion corrections, all measurements have been performed a t high carrier gas velocit,ies where liquid phase mass transfer and interfacial mass transfer should predominate. Finally, an extremely wide range of immobile phase loadings has been used to study the functional dependence of t'he Ci term upon the capacity factor, k . EXPERIMENTAL
Reproducible amounts of vapor were introduced into the chromatographic column by means of a borosilicate glass vacuum sampling system (26) in conjunct,ion with a gas Pampling valve ( 2 4 ) . The sampling system consisted of two separate parts. The first way connected directly to the vacuum pump through V O L . 36, NO. 9, A U G U S T 1964
1753
