Separation of aliphatic hydrocarbon mixtures by gas chromatography

Mar 11, 1991 - network and facilitate the input of only the most meaningful ... (1) Lyon, R. F.; Loeb, E. P. Proceedings International Conference on. ...
2 downloads 0 Views 477KB Size
Anal. Chem. 1991, 63, 1201-1264

network and facilitate the input of only the most meaningful information from the sample data. LITERATURE CITED (1) Lyon, R. F.; Losb, E. P. Rocedings Intematkmai Conference on Acoustlcs Speech and Slgnal Rocesslng ICASSP-87. Dallas, 1987. (2) Burr, D. J. I€€€ Trans. Acoust., Speech, Sbnai process. 1988, 36, 1162-1 185. (3) Saund, E. Neural Comput. 1990, 2 , 138-151. (4) Llppmann. R. P. I€€€ ASSP Mag. 1987, 3. 4-22. (5) thamsen, J. U.;Meyer, 8. J . Megn. Reson. 1989, 8 4 , 212-217. (6) Long, J. R.: Gregorlou, V. G.; Gemperline, P. J. Ami. Chem. 1990, 62, 1791-1797. (7) Rumelhart,D. E.: McClelland, J. L.; PDP Research Group Para&/ U s trlbuted procesSng: Expk~tbnsIn the Mkrostruchm,of CcgnMon;

1261

MIT Press: Cambridge, MA, l Q 8 6 Part 1.

(8) Wold, S.; Sjostrom, M. In c h e ” e M c s : Theory and AppuCethm; Kowalski, 8. R., Ed.; American Chemical Socbty: Washington, D.C.,

1977 .

DD C

242-282. r

~ ~

~

(9) Kowalski, 8. R.; Bender, C. F. J . Am. (2”.Soc. 1872, 04, 5632. (10) Burris, D. R.; MacIntyre. W. (3. Envkon. Toxlcd. Chem. 1985, 4 ,

371-377. (11) Puyear, R. L.; Fleckenstein, K. J.; Montz, W. E., Jr.; Brammer, J. D. Bull. Environ. Conhminants Toxlcol. 1984, 13, 171-178. (12) Mayfield, H. T.; Bertsch, W. J . Comput. Appk. Leb. 1985. 7, 130- 136.

RECEIVEDfor review November 26,1990. Accepted March 11, 1991.

Separation of Aliphatic Hydrocarbon Mixtures by Gas Chromatography Using Serial Liquid-Phase and Solid-Phase Columns J. Howard Purnell* a n d M.H. Wattan Department of Chemistry, University College of Swansea, Singleton Park, Swansea SA2 8PP,U.K.

The theory and procedures previously developed by us are applled to a system comprlsing a GLC and a GSC column serlally connected. A C,-C, allphatlc hydrocarbon mlxture Is w e d as an example where a GLC-GSC comblnatlon can provlCktunable~vHy.Thetheoryisagalncodbmedand proddm accurate predktbn of retentlomi In the serlal system from data for the lndlvldual columns. A base-llne separation of the mlxture components Is achleved followlng wlndow analysk that defined the necessary relatlve and total column lengths. The partlcular rultablllty of dual-temperature operatlon for GLC-GSC systems Is outllned.

We have recently (I) shown how serial column gas chromatographic theory (2-6) can provide precise prediction both of retention in serial open-tube systems and of the relative and total column lengths needed for optimized separation of the sample components. We pointed out (I) that only in certain special circumstanceswould serial operation of packed GLC columns offer advantage over the use of the corresponding mixed packings in a single column. Where, however, separation would be facilitated by use of a GLC and GSC substrate pair,a mixed packing would suffer the disadvantage of progressive relocation of the liquid substrate and consequent continuing modification of the adsorptive characteristics of the GS component. For GLC-GSC combinations, therefore, serial column operation will normally be demanded. A typical situation wherein a GLC-GSC system can provide analytical advantage is in the separation of alkanes and alkenes of similar carbon number. In general, GLC substrates can be found that will provide separation of both the isomeric alkanes and alkenes, but it is commonly found that, in mixtures of any degree of complexity, overlap of alkenes and alkanes occurs. In contrast, polar GSC substrates such as alumina or silica separate alkanes and alkenes well but provide little, if any, discrimination of alkane isomers. The combination of these selectivities provides opportunity for complete resolution of relevant mixtures, and we describe here an example 0003-270019 1/0363-1261$02.50/0

of the separation of a typical C5-CBmixture by serial GLCGSC. PROCEDURE The retention of any sample component in a serial system differs from a simple average for the two columns on account of the carrier gas compressibility, which, in addition, causes retentions to depend on the sequencing of the columns. The general expression for the overall capacity factor ( k ’ ) for elution from a serial pair (F = front; B = back) operating over the pressure regime, pi p potwhere p is the junction pressure, and p i and p o are the inlet and outlet pressures, respectively, is

--

k’=

Pk$ + k b P+1

(1)

k $ and k are the capacity factors for a sample component with front or back column and P ( = t m / t a ) ,where td represents the dead time (retention time of unsorbed material), is calculable via the equation

where LB and LF are the column lengths, V , and V , are their respective dead (free) volumes, and Rm and Rm are the resistance to flow factors defined by

(3) where p and ii are the compressibilityaveraged pressure and carrier gas velocity, respectively, and pii = piui = pouo. For a serial system, p o = p for the front column and pi = p for the back column. RF for any column is calculable via (4) 0 1991 Amerlcan Chemlcel Society

1262

ANALYTICAL CHEMISTRY, VOL. 63, NO. 13, JULY 1. 1991

and values of td with fully corrected flow rates, F,,provide the value of V,. Finally, the junction pressure for the serial system is evaluated for any given values of pi and povia the general equation

P? - 1 F b ? - (vh&FF/ vhfFRFB)P21 (5) 1- 1 ~ [ 1 (V~RFF/VMFRFB)I where 1~ is the length fraction of the front column [LF/(LF + LB)]. The foregoing explains the procedure allowing calculation of capacity factors for all sample components via eq 1for any desired values of LF and LB and of pi and po,to within the accuracy of the assumption of ideal (carrier) gas behavior. However, we cannot predict the optimum conditions for analysis. To achieve this, we need to develop the theory further. It is clear from eqs 2 and 1 that P, and hence k’, are not linear functions of column length fraction. We now introduce the functions fF and fB (fF + fB = 1) to provide a linear dependence of k‘, viz. p2

Table I. Capacity Factors (k’)for Named Sample Components Eluted by Nitrogen at 63 OC from (a) a 180- X 0.3-cmColumn of (5% w/w KOH) Deactivated Alumina (F20) and (b) a 702- X 0.3-om Column of Squalane (17.5% w/w)on 60-80-mesh Chromosorb G-AWDCMS

=

k’ component

alumina

squalane

1,isopentane 2, pent-1-ene 3, n-pentane 4, 2-methylpentane 5, 3-methylpentane 6, hex-1-ene 7, trans-hex-2-ene 8, cis-hex-2-ene

25.7 48.3

8.06 8.90 10.9 21.9

y

[-I ?I2

= LBRFB LFRFF

=

e] 2

[g][

(8)

the superior bars indicating values of RF and VM per unit length of column. Finally, the length fraction corresponding to fF a t fixed pi and po is given by

To evaluate lF at the optimum value of the fF that can provide complete base-line resolution of the mixture components, we proceed as follows. Since k’is linear in fF, we can, illustratively, construct plots of k’against fF (0 1)by connecting the values of k iand k i for each component. This allows us to recognize by inspection those component pairs that present the greatest difficulty of separation. We then ratio k’ for all pairs in this group (k)2/k’l = a, relative retention) and construct a window diagram (7-9) of a against fP The lower envelope defines the minimum a in the group at each fF, and from the highest window, we identify the highest minimum a (a:$,), Le., most favorable value for the most difficult pair to separate, and the corresponding value for fF* Finally, knowing the most favorable a and evaluating k’for the second eluted of the most difficult pair to separate from the k’/fF plot, we use the familiar equation (IO)

-

to determine the number of theoretical plates required for full base-line resolution of all components. From a knowledge of the minimum theoretical plate height (H-) and velocity (am& we then calculate both the total column length (LF+ LB) and the value of pi needed to generate amin.This value of pi,in turn, defines 1F for optimized separation (lyt)via eqs 7-9. The manual approach to finding lYt described can be more conveniently done via a simple computer program, of course.

156.2

25.0 25.0 29.0 31.0

Table 11. Resistance to Flow Factors (Rp)and Column Dead Volumes ( V , ) for Columns of Table I 10-7~~,

lo4&,

VM,

column

N s m-s

VM,mL

Nsm-‘

mLm-’

alumina

3.50 f 0.07 0.740 f 0.002

6.25 f 0.16 11.37 k 0.04

19.4 1.05

3.47 1.60

squalane

(7)

30.6 81.1 81.1 143.6 137.2

EXPERIMENTAL SECTION The mixture used as the example is detailed in Table I and the substrates chosen for use were alumina (F20)and squalane, both supplied by Phase Separations Ltd. (Queensferry,U.K.).The alumina was modified by addition of 5% (w/w) KOH before packing into a 1Wcm X 3-mm-0.d. stainless steel tube. To match the very considerable retention provided by the alumina (DA), the squalane (SQ) packing contained 17.5% (w/w) on 60-70-mesh ChromosorbG AWDMCS, the column being 702 cm X 3 mm 0.d. Nitrogen was used as the carrier gas, and the data refer to elution at 53 ‘C. The chromatograph used was a Perkin-Elmer F33 with flame-ionization detection. Solutes were injected either singly or in easily separable mixtures, sample sizes being in the range 0.2-1.0 p L of liquid. The values of k’cited represent the averages of numerous elutions as did the dead times that were determined from the elution times of methane. The flow rates were measured with a bubble meter and corrected for water vapor pressure and to column temperature. RESULTS Table I lists the values of k’ measured for the sample components with the DA and SQ columns. We see that, as anticipated earlier, the DA column provides no separation of 2-methylpentane and 3-methylpentane and only incomplete separation of hex-1-ene and trans-hex-2-ene, while, correspondingly, the SQ column offers no prospect of separating 3-methylpentane from hex-1-ene. Since, however, the columns fail with different components, a serial solution is feasible. Plots according to eq 4 yielded excellent straight lines for both columns, and the data derived for RF and VM for each are listed in Table 11. van Deemter plots provided the minimum values: DA (H- = 0.06 cm; a- = 2.0 cm 8-l) and SQ (H- = 0.15 cm;a- = 1.5 cm 9 3 . The poor value of HA for the squalane column reflects both the relatively coarse particle size of the solid support chosen for use because of the column length and, more particularly, the high solvent loading required to provide high k’. The data presented allow us to calculate the values of k’ to be anticipated when the test columns are simply linked together in either mode, SQ(F)/DA(B) or DA(F)/SQ(B). Table III lists a comparison of the caiculated and experimental data. The correspondence is remarkably good and given the likely accuracy of the measurements involved is essentially exact. The validity of the theory for a GLC-GSC system is thus confirmed, as is the quality of the basic data.

ANALYTICAL CHEMISTRY, VOL. 83, NO. 13, JULY 1, l 9 Q l

1263

Table 111. Compariron of Experimental and Calculated Valuer of k’for Linked Column8 of Table I mode: SQ(F)/DA(B)

mode: DA(F)/SQ(B)

1~ = 0.78; pq 14.83 psi pi = 34.8 PSI p = 30.0 psi, P = 2.52

= 34.8 psi p = 23.05 psi, P = 0.843

component 1 2 3 4 5 6 7 8

IF = 0.24; po

14.8 psi

pi

mode 1 k’(exp) k’(calc) 13.0 20.0 16.6 39.3 41.4 58.8 58.8 67.4

mode 2 k ’(exp) k ’(calc)

13.0 20.1 16.5 38.7 41.0 58.7 59.7 66.6

16.1 26.5 19.4 48.7 50.5 79.5 79.5 89.4

16.1 26.9 19.9 49.0 50.7 79.3 78.5 88.3

We now move to optimize the separation by constructing the window diagram of Figure 1 where values of a,derived either by use of k’against fF (SQ= F) plots or via calculation through eq 6, are plotted against fp There are four values of fF having a = 1 (totalseparation impossible at any length), viz. fF = 0 and 1 and at 0.61 (components 6 and 7 have a = 1) and at 0.90 (components 2 and 3 co-elute). Thus, we have three windows. That at fF = ca. 0.13 provides a very low window peak value of a and so can be discarded. Those at fF 0.780 (a = 1.073)and fF = 0.932 (a = 1.060) both offer the possibility of separation with columns of practical length. The former, at first sight, appears more attractive since azk corresponds to a value of Nq H 8o00, while the latter window indicatea a value of Nq 12000 theoretical plates. However, the k’of the last eluted component in the fF = 0.78 window, which fixes the total analysis time, is almost double that in the second window. For this reason, and because the choice allows us to illustrate another important feature later, we thus choose to illustrate the method by use of the window at fF =

0.932. Since fF(SQ) is so large, the theoretical plate height of the column combination is dominated by the efficiency of the SQ column. For operation at H,, (0.15cm), we calculate that L A + LB = 1780 cm, which requires a value of piof ca. 35 psi to generate the necessary average velocity. Having set this value, we can now calculate 1F from the optimum fF as previously described. The results for the two column modes are set out below: Mode 1: SQ(F)/DA(B). fp = 0.932; P = 13.05; y = 0.260; p o = 14.8 psi

pi = 34.8 psi; p = 22.8 psi;

IF = 0.951

L , = 1690 cm; LDA = 90 cm Mode 2 DA(F)/SQ(B). fF

P = 0.073;

= 3.84; Po = 14.8 psi pi = 34.8 psi; p = 32.3 psi; 1~ = 0.023 LDA= 40 cm; LQ = 1740 cm

= 0.068;

1.01

Figure 2 illustrates the first-time chromatograms obtained with the column systems and operating parameters defined above. As the theory demands, the chromatograms are identical while bese-line reeolution is also achieved. Calculated and experimental values of k ’are listed in Table IV and agree to within any reasonable estimate of the attainable accuracy.

DISCUSSION The theory and procedures described are clearly as valid in GLC-GSC seriel column operation as previously shown for

fP

Flgure 1. Window diagram for components of the eight-component mixture (seeTable I) eluted at 53 OC by nltrogen from the test columns of Table I. Values of a on the envelope represent the lowest value of relative retention for any pair in the mixture at comsponding f , (F = SQ, B = DA).

r1rs

.

Flgure 2. Chromatograms of eight component mixture (Table I) obtained with the optimized serial column system in both modes. Upper: SQ(F)/DA(B); L , = 1890 cm, L O A = 90 cm. Lower = DA(F)/SQ(B); L , = 40 cm, Lsq = 1740 cm. Inlet pressure: 34.8 psi. Outlet pressure: 14.8 psi in each mode.

Table IV. Comparison of Experimental and Calculated Capacity Factors (k’)for Serial Column Systems Optimized for Base-Line Separation of Sample Componentso k’(exp) mode 2‘

component

mode 1*

1 2 3 4 5 6 7 8

9.80 11.7 12.7 26.3 29.8 33.2 37.0 40.1

9.88 11.7 12.6 26.0 29.0 33.5 37.0 40.0

k ’(calc)

either mode 9.25 11.6 12.2 25.9 28.8 33.0 36.4 38.6

“Column and operating parameters as in text and Figure bSQ(F)/DA(B).‘DA(F)/SQ(B).

2.

packed GLC-GLC and open-tube pairs, and our ability to predict retentions and optimize separations thus opens up a new avenue to the tuning of selectivities, in aliphatic hydrocarbon separations at least. Several practical points that emerge are worth making. First, since fF in mode 1 was so close to unity, it would have been tempting to equate it to 1F and assume the same value of a at the window peak. In fact, the sensitivity of a to fF within the window is such that this would have been disastrous since, in this case, the window (see Figure 1) is so narrow. Setting 1F = 0.932(which corresponds approximately to f~ = 0.91)reduces a (see Figure 1) to around 1.025,which would

1204

Anal. Chem. 1991, 63, 1264-1270

increase Nm,i.e., the column length required for full resolution by a factor approaching 5. Since the assumed a would still have led us to the same overall column length, the resulting chromatogram would have shown very unacceptable resolution of the mixture components. The conversion of f F to 1~ is thus crucially important in this case and the example shows us also that, when a narrow window offers the best option, the envelope around the window maximum demands careful definition. Second, it is clear that the column systems used here by no means represent the best choice in terms of substrate loading, operating temperature, and 80 forth since they provide poor efficiencies and excessively long retentions so that not only is the optimized column combination inconveniently long but analysis time is unacceptably protracted. However, their choice has allowed a clear demonstration of the fact that, in a serial system, when the more retentive column is at the front, the length of this column required is far less than when it is at the back. Moreover, the effect is not trivial; the lengths of the DA columns needed differ by over a factor of 2. In the present instance, the overlong retention time is of no concern since the objective was to demonstrate the potential of the GLC-GSC serial column technique and certain aspects of serial column operation. In practical terms, a much more lightly loaded GLC column would be desirable to provide much higher efficiency. Then, however, the resulting k 'would be dominated by the values for the GSC column, which are intrinsically much greater. A higher level of KOH doping of the DA would reduce k'but also the desirable selectivity while

the possibility of reducing the diameter of the GSC column further to reduce the amount of adsorbent is impractical. The logical answer is to operate such systems in a dual-temperature mode, rather than isothermally, with the GSC column held at higher temperature. Fortunately, the theory and procedures are exactly the same, although the identification of the best pair of temperatures adds to the amount of basic data and of the volume of calculation needed. Given the data, of course, the optimization procedure is again amenable to window analysis and, with suitable modification to the computer program, is no more complex than that described here. Registry No. Isopentane, 78-78-4; pent-1-ene, 109-67-1; npentane, 109-66-0;2-methylpentane, 107-83-5;3-methylpentane, 96-14-0; hex-1-ene, 592-41-6; trans-2-hexene, 4050-45-7; cis-2hexene, 7688-21-3. LITERATURE CITED (1) Jones, J. R.; Purrall, J. H. Anal. 0 . 1990, 62, 2300. 1884, 292, 197. (2) Purnell, J. H.: Williams, P. S. J . Chmrm-. (3) Pwnell, J. H.: Williams, P. S. J . chrometugr. 1885, 321, 249. (4) Purnell, J. H.; Williams, P. S. J . Chfomatu&T. 1885, 325, 1. (5) Purnell, J. H.; Rodrlquez, M.; Williams, P. S. J . Chrometogr. 1886,

358, 39. (6) Purnell, J. H.; Jones, J. R.; Wattan, M. H. J . Chrometcg. 1887, 399, 99. (7) Laub, R. J.: Purnell, J. H. J . Chrometogr. 1975, 112, 71. (8) hub, R. J.; Purnell, J. H.: Williams, P. S. J . chrometogr. 1877, 134, 249. (9) Laub, R. J.; Purnell. J. H. J . Chfomtogr. 1878, 161. 59. (10) Purnell, J. H. J . Chem. Soc. 1880. 1268.

RECEIVED for review December 17,1990. Accepted March 5, 1991.

Consecutive Gas Chromatograms from Parallel Columns Using a Single Injection and a Common Detector Pardeep K. Gupta and John G. Nikelly* Department of Pharmaceutics and Department of Chemistry, Philadelphia College of Pharmacy and Science, Philadelphia, Pennsylvania 19104

A method ol t w d i " l o n a l chromatography k described In whlch two consecutive chromatograms are obtained In a dngk run using two parallel capillary columns connected to a dngk dotector. The two fused dllca capillary columns, whlch have OV-1 and Carbowax 20 as the llquld phase, Mer In length, fllm thickness, Indde diameter, or combinations of these. It was lwnd that the resunlng ditferences In carrler vebclth, capactty factom, and retention times are wfflclent to create a gap between the chromatograms from each column. Depondlng on the number of components In the Injected sample and the range of thelr polarities and bolllng points, dbtlnc( chromatogram are obtained that may be used In determining the Identity of the components.

INTRODUCTION The method of identifying compounds based on their gas chromatographic retention times has been used from the very beginning of the technique. But because of the pcesibility that more than one compound may have the same or sufficiently

* Corresponding author. 0003-2700/91/0363-1264$02.50/0

similar retention index on a given liquid phase, a confiitory procedure is often necessary. Aside from the method of two-dimensional gas chromatography, i.e., confirming the identification by obtaining a second set of retention times on a dissimilar liquid phase (I,2))other techniques include using a selective detector, splitting the effluent into separate detectors, and others (3-5). The most definitive of these is GC/MS in which the retention data are combined with mass spectral fragmentation patterns and molecular ion masses (5). Retention on Two Columns. There is little likelihood that two compounds will have the same relative retention on columns of different liquid phases. For this reason, and because of its simplicity, this method, often referred to as two-dimensional chromatography, is the method of choice when better identification techniques such as GC/MS are not readily available. The two most widely used liquid phases in two-dimensional gas chromatography are poly(methylsi1oxane) (OV-1, SE-30) and poly(ethy1ene glycol) (Carbowax 20M). Not coincidentally, these phases are a t opposite ends of the polarity scale. In the simplest if not most common procedure of two-dimensional gas chromatography, the two chromatograms are obtained in separate runs, i.e., using one column a t a time. (Alternatively, the two chromatograms may be obtained at 0 1991 Amerlcen Chemical Society