( i )Lovelock, J. E., ASAL. CHEV. 33,
162 (1961). (8) Lovelock, J. E., Institute of Petroleum, Gas Chromatography Discussion Group, Oxford, England, May 1957. (9) Lovelock, J. E., J . Chromatog. 1, 25 (1958). (10) Lovelock, J. E., Gregory, S . L., “Gas Chromatography,” S . Brenner, ed., p. 219, New York, hcadeniic Press, 1962.
(11) Lovelock, J. E., Lipeky, S. Ii., J . Am. Chem. Soc. 8 2 , 431 (1960). (12) Lovelock, J. E., Shoemake, G. R.,
Zlatkis, .4., A N A L . CHESI.35,460 (1963). (13) Lovelock, J. E., Simmonds, P. G., VandenHeuvel, W. J. h., lVature, in press. (14) Lovelock, J. E., Zlatkis, A . , ANAL. CHEJI.33, 1958 (1961). (15) Lovelock, J . E., Zlatkis, h.,Becker, R. S., Xature 193, 540 (1962).
(16) Shulz, G . J., J . C h e w Phys. 33, 1661 (1960). (17) Smith, V. N., Merritt, E. J., ASAL. CHEX.34, 1476 (1962). (18) Wentworth, W. E., Becker, R. S., J . Bm. Chem. SOC.84,4263 (1962). RECEIVEDfor review February 4, 1963. hccepted February 13, 1963. Presented a t International Symposium on Advances in Gas Chromatography, University of Houston, Houston, Texas, January 21-24 1963.
Gas Liquid Chromatography Temperature Changes during Passage of a Solute through a Theoretical Plate R. P. W. SCOTT W. G. Pye & Co., Itd., Cambridge, England
b This paper considers the change in temperature of a theoretical plate when a solute i s absorbed and desorbed during its passage through it. An equation i s derived that relates the temperature of the plate to the total flow of carrier gas. The effect of this temperature change on the column efficiency and peak symmetry i s considered and b y means of the derived equation conditions are predicted that will minimize or eliminate it. From the equation relating plate temperature to the total flow of carrier gas, theoretical curves are given which are substantiated b y experiment.
D
t’he passage of a solute through a theoretical plate adsorption and desorption of the solute in the liquid phase take place, accompanied by evolution and absorption of t’lie heat of solution. This heat exchange between the solute-solvent system and its surroundings d l result in a change in temperature of the contents of the theoret’ical plate. Ray ( 7 ) suggested temperature effect might form of a det’ecting system and Klinkenberg considered it might affect peak s\-mmetrp (6). This phenomenon has been employed by Claston ‘8 3 detecting system for liquid solid chromatography ( I ) , Recent work by Smith (8) has s1ion.n that under certain circumstance.: this temperature change e m be particularly significant when lwge charges :Ire placed on the column. This paper de,wihes a theoretical ini-estigxtion of the t~emperaturrchanges of the t’heoretical plate during the passage of a solute through it. An equation is given describing the change in tenil)erature of the plate relative to its surroundings in terms of the various
operating parameters. Tn o variables in the system are considered. the temperature and the total flow of gas through the plate. Experimental support t o the theoretical equations is included and the significance of the heat effect on column performance and peak symmetry is discussed. Reaults are also considered with a iiew to modifying column design in order to reduce or eliminate the temperature changes, particularly with re-pect to preparative scale columns THEORY
URISG
plate is insignificant compared with the heat capacities of the liquid phase and support. The concentrations of solute in the plate are a t all times maintained at levels m here the adsorption isotherm is linear and KX,,, = XL,,.
It is required to derive an equation relating the excess temperature of the plate to its surroundings (e) with a suitable operating variable. For convenience the variable chosen r i l l be v, the volume of gas passed through the plate in terms of “plate volumes” (5).
or
Consider the nth theoretical plate of a chromatographic column and let the follon-ing designate its propertie-.
V, VL Vs
Ss
= volume of gas in plate =
volume of liquid in plate
= volume of support in plate = specific heat of support in
plate specific heat of liquid phase in plate ps = density of support pL = density of liquid XLn= concentration of solute in liquid phase in the nth plate X,, = concentration of solute in gas phase i n the nth plate e = excess temperature of plate above its surroundings Carrier gas flow rate = Q ml. per second
SL
=
The following postulates will be made : The gas flow rate through the plate is constant. The temperature of the surroundings of the plate is constant. The heat capacity of the gas in the
6t =
Vo
+ KVL sv Q
Consider the heat balance of the plate. Heat capacity of plate X rise in temperature = heat evolved in plate heat conducted from plate Consider this balance during the passage of ga; volume 6v in time 6t Then
+
(VLPLSL vspsss)Ge .= heat evolved in plate heat conducted from plate radially
(2)
As the solute is distributed over a total of Gl/nplates ( 5 ) ,the temperature of the plates adjacent t o the one under consideration may be considered the same and thus conduction of heat along the axis of the column may be considered negligible compared with heat conducted radially from the column. Xow, change in concentration in gas 6 XPn phase during passage of 6v = 60 6V Thus, change in concentration in VOL. 35, NO. 4, APRIL 1963
481
-7rr---
liquid phase during passage of 6v = K.6XQn
O
4
O
r
-
13-o
~
V6
Therefore Heat evolved in plate
=
hK6Xon 6v (3) ~
6V
where h is the heat of solution of the solute in the liquid phase given in the appropriate units. Heat conducted from the plate in time 6t = A 2 .6te, which from Equation 1 becomes AZ (v,+ KV&ue Q
(4)
where d is the surface area of the plates and 2 is a function of the thermal conductivity of the contents of the plate and is given in appropriate units. Then from Equations 2 , 3 . and 4,
where Kh ff=and ( V L P L S L VspsSs)
+
-6.0
-50
-w-
10
20
OFKWHERE W: (1 - R ) v-n
30
Relative changes in 0 with w
Measured in units of
dB -dv ++e=U--
0
-10
-20
I N UNITS
Figure 1.
i.e.,
-30
-40
fi
dXon
dv Then
This is a standard differential equation, the solution of which can be seen b y differentiation to be:
J.
pv Xgoe- w 1 / 2 n dw ff8 4 2 4 1 - 8)"
+E
= __
1-8
(w
+ n)
and
Thus, Integrating by parts,
S o w from the equation of the elution curve ( 5 )
Hence
As n is large, when v = 0, w
~-
=
Thus
By replacing v by
nP + n [ - p JL w + __ 1-p I - p
e
=
2
0
Thus it' p i-: 0.10 or less $ = e@x
+ nB2/2
and Equation 8 becomes
E ii
noting n is large, and applying Sterlings' theorem, Equation 7 can be modified ta
482
0
+ n ) + n log ( 1 - (3)
Log $ =
1 - P
~
ANALYTICAL CHEMISTRY
-
&+T+;.? np3 no4
dw and =
(w
--
- n and
e @ = 1 and
ThusE
P
Log $ =
Under adiabatic conditions,
B = O
sild the temperature curve is of the bame form as the elution curve, with a maximum when o = n. With the use of standard statistical e - w2/2n t:t\)les to evaluate _ _ and 42%
dw and ascribing suitable 1:ilues to p, the form of Equation 9 c u i be plotted. In standard tables the variance is taken 8'. unity and therefore values for the function and integral of the function are in terms of d F Thus Equation 9 becomes of the form,
Referring to Equation 9 i t will be aeen that d i m 11) = +32/, the function within the brackets become> equal to - p and all the solute has passed through the plate. Under these conditions e may be negative and the temperature of the plate is below that of its surroundings. During the recovery of the plate temperature to that of its burroundings Equation 9 can be modified to
e
-aXo,&-Pu~ -
=
nP/2
portional to the change in temperature in degrees centigrade per unit change in concentration of the solute in the liquid phase, assuming no loss of heat b y conduction from the plate. I t follows t h a t the change in temperature of the plate will increase with the charge size and the heat of solution of the system concerned. The partition coefficient, K , and the heat capacity of the plate, however, determine the value of both CY and p. To assess the effect of these parameters on the plate temperature, i t is necessary to compare the relative effect of factors CY and p. may be considered as the thermal loss of the plate. It is directly proportional to the surface area of the plate, thermal conductivity of the contents of the plate, and plate capacity (V, K V L ) (8). p varies inversely as the thermal capacity of the plate and the column floiv rate. In practice p takes values between 0 and about 0.4. For this range of value, the effect of p is predominantly an attenuating function and reduces the
+
value of e at any value of w. Kithin the squared brackets 0 also reduces the Gaussian integral and thus tends, in this position, to increase the value of e. However, the overriding effect of p is t o reduce e for all values of w. Therefore, increase in the surface area, A , or the conductivity, 2, will result in a fall in the value of e for all values KVL) of w. Plate capacity ( V , will be nearly proportional to K , as V , is normally small compared with K V L , so both 01 and p are proportional to K and inversely proportional to the thermal capacity of the plate. However, the net effect on the excess plate temperature e of CT and p due to changes in partition coefficient and thermal capacity of the plate is rather c7omples and experimental results indicate that the temperature change in the plate will decrease as K increaces. It would therefore be expected, assuming no overloading, that the more symmetrical peaks would be eluted late in the chromatogram. The form of the curves given by
+
.4
.3
(12)
and Log
e
= log
-
BU
\\here p =
O I ~ Xe-nB2/2 C,
'Thus by plotting log e against w a itraight line will be obtained, the slope of which n 4 l be - p and will allow experimental determination of the thermal loss of the plate.
c
2!= . 2 w I
a Y
DISCUSSION OF THEORY
It may be seen from Equation 9 that e is a complex function of the plate volume flow, column efficiency, n, the initial concentration of solute placed on the first plate, XBo,and two constants, O( and p. Xp, will vary directly with the sample size, and thus the greater the charge the greater will he the yariations in plate temperature. Referring to Equation 5 i t mill be seen that CY is inversely proportional to the heat capacity of the plate, and directly proportional to the heat of solution of the solute-solvent system under consideration and the partition coefficient of the solute r i t h respect to the solvent. CY can then be considered as the gross thermal gain of the plate and is pro-
.l
VOL. 35, NO. 4, APRIL 1963
483
Equation 9 is shown in Figure 1, where the function relative to the plate temperature is plotted against w in units of z/la plate volumes of carrier gas passed through the plate. The different curves are for selected values of p between 0 and 0.2. After the solute has passed through the plate, the temperature of the plate is below that of its surroundings and measurement of plate temperature at different flow volumes during the period in which the plate temperature is returning to that of its surroundings, permits the experimental determination of p with the aid of Equation 11. Under adiabatic conditions \Then p = 0, the temperature curve of the plate follon-q the elution curve. Generally the higher concentrations in the peak are at a higher temperature than the lower concentrations in the peak. Furthermore, the concentrations of solute in the front of the peak are all at a temperature above that of the column surroundings, whereas the solute in the tail of the peak is often below the temperature of its surroundings. Kow the speed of a solute band through a chromatographic column is inversely proportional to the partition coefficient. The partition coefficient exponentially decreases with temperature; thus from Figure 1 it can be deduced that the high concentration in the front of the peak will travel more rapidly through the
GQbPU OF R b T E T E M P E R b ' U R E
&*INST
column than the low concentrations at the t'ail of the peak. This effect must produce peak asymmetry. Assuming that the maximum temperature reached by the plate above its surroundings ie 1" C., that the column contains 2500 theoretical plates, and that the partition coefficient is reduced by 1% per degree centigrade ( A H = 4000 cal. mole-') the displacement of the peak can be calculated for the different values of shown in Figure I. A series of elution curves obtained for various values of p is plotted in Figure 2. The lower the value of p and the higher the consequent change in plate temperature, the more the peak is deformed, ivith a sharp front and a sloping tail. These calculations have been carried out assuming the same temperature change in each plate. This, of COUI'S:E, is not true and it can be seen from Equation 9 that as n increaies e )vi11 lie reduced for any given value of w. Thus the change in temperature of the plate as the solute passes through it lvill be reduced as the iolute progresses towards t81ieend of the column. In fact, in an ordinary analytical column 3 c~r5 feet long, it viould appear that the ninjor deformation of the peak due to thi.: temperature change occurs in the first quarter of the column length. Another interesting aspect is that partition ctrefficients or retention volumes, ruppovtily measured a t the
TIME FOR DIFFERENT
ELUTION CLIRVES OaTAl%ED
AT
thermostated temperature, are, in fact, measured a t a temperature slightly abo\-e this. It follows also that the difference between the effective chromatographic temperature and the thermostated temperature will vary from one column system to another and from one set of operating conditions t o another. I n the theoretical treatment the effect of the carrier gas has not been included. However, the thermal conductivity of the plate will depend to some extent on the thermal conductivity of the carrier gas employed, and the specific heat of the gas will affect the thermal capacity of the plate. Consequentlr, if a series of experiments is carried out meaiuring partition coefficients or retention volumes using different carrier gases or different carrier gas pressures, in general, one would elpect the yalues obtained for the partition coefficient t o decrease with the thermal conductivity of the carrier gas employed or n-ith increase in column pressure The effect of radius has also not been considered in this theoretical treatment. However, it can be logically deduced that for a packed column the change in temperature of the plate \vi11 be greatest a t the center. This means that the peak will not only be deformed, but it will also be spread as a result of the temperature variations across the radius of the column. The d e f o r n d llraks
31FFEREh- FLOW R b - E S
FLOW R b l E S DURING T Y E ELUTION OF 2 p ) W CHWROFOPH
I
C H A R l SPEED
I\
ll
"t I
FLOW RATE
o : 1 84
ASYMMETRY R A T I O
't
:2
10
mljmin
14
z
mi(mn
1 84
e 5rn1/min
4 5 rnl min
161
? 34
-TIME
Figure 3.
Effect of carrier gas flow rate on plate temperature and peak shape
Graph of plate temperature against time for different flow rates during elution of 2 p'. of chloroform Right. Elution curves obtained a t different flow rates left.
484
ANALYTICAL CHEMISTRY
GRAPH OF PLATE TEMPERATURE AGAINST
ELUTION CURVES FOR DIFFERENT
T I M E FOR DIFFERENT CHARGES O F CHLOROFORM
CHARGES O F CHLOROFORM
C IN LlOUlD
FORM 1
COLUMN CONDITIONS COLUMN LENGTH COLUMN RADIUS
51t GLASS 2rnrn
COLUMN PACKING 15% SOUALANE ON 100-120 3 5 MESH CELITE COLUMN TEMPERATURE 7 5 ' C POSITION OF THERMOCOUPLE l O c r n BELOW IhlJECTION POINT' SAMPLE
0 2 5 TO 10.u OF LlOUlD C H L O R O r 3 R U
Y w 3
t W
c w
5 v1
W X
SAMPLE SIZE
1!J
ASYMMETRY RATIO r
Figure 4.
i
2
o
075111
0 5>1
02541
140
120
109
Effect of charge size (in liquid form) on plate temperature and peak shape left. Graph of plate temperature against time for different charges of chloroform Righf. Elution curves for different charges of chloroform (in liquid form)
shown in Figure 2 will be familiar to workers in the field of preparative scale gas chromatography. Low column flow rates are frequently employed in preparative scale gas chromatography in order to maintain peak symmetry. Low values of Q increase p and thus reduce the temperature changes in the coluni n. EXPERIMENTAL
To obtain the curves shown in Figures 1 and 2 experimentally, it is necessary to conduct a series of experiments in which a solute is chromatographed under conditions where factor fl changes. From examination of Equation 5 it may be seen that p varies inversely as the flow rate through the column. Thus, if a sample is chromatographed on a given column a t various flow rates (all other operating parameters being kept constant) and the temperature of a point in the column is recorded and plotted against total volume flow, a series of curves will result, similar in form to those shown in Figure 1. The X axis of the curves in Figure 1 is in units of plate volume flow and these will remain constant whatever the flow rate, providing the efficiency of the column does not significantly change. It is possible to approximate these conditions by operating the column over the flat portion of the HETP curve. I n order t o produce a series of curves on the recorder chart having the same scale of plate volumes flow, when operating at different column flow rates, the chart speed of the recorder must be increased in proportion to the flow rate used. Under these conditions a series of curves obtained at different
gas f l o ~rates would be relative to the same scale of flow in plate volumes and would thus be similar to those shown in Figure 1. If both the temperature of a section of the column and the concentration of solute eluted from the column were monitored simultaneously, a set of curves similar to those shown in Figure 2 ~vouldbe obtained. APPARATUS A N D METHOD
The apparatus used was a Pye Panchromatograph with its ancillary electronic equipment. A 5 - f O O t glass column, 4 mm. in diameter, packed with 15% equalane on 100 to 120 €33. mesh Celite was employed in conjunction with a molecular entrainer and a macroargon detector. The molecular entrainer was in effect a splitting device that permitted up to 10 pl. of charge to be placed on the column without overloading the macroargon detector. Ten centimeters from the column inlet a small hole was blown in the column wall and into this was sealed a copperconstantan thermocouple. The cold junction of the thermocouple was situated in a thermos flask, exterior to the chromatograph. The output of the thermocouple was fed to a Pye chopper amplifier which had a full scale deflection of 100 pv., and the output from the amplifier was connected to a potentiometric recorder. The column was thermostated a t 75" C. and flow rates of 4.5, 8.5, 14.2, and 18.4 ml. per minute were employed in conjunction with chart speeds of 6, 12, 18, and 24 inches per hour. Charges of 2 J. of chloroform were injected into the column, the resultant elution curves and temperature curves being shown in Figure 3. T o examine the effect of charge size on plate temperature and peak shape, various charges of chloroform, as liquid
sample, T\ ere injected onto the column operated with a carrier gas flow rate of 18.4 ml. per minute and a chart speed of 24 inches per hour. The elution curves and temperature curves monitored are shoxn in Figure 4. To eliminate the effect of cooling due to vaporization of the charge, different, volumes of air, saturated with chloroform vapor a t room temperature, lvere injected onto the column in place of the liquid sample. The elution curves and temperature curves obtained from the samples of chloroform in air are shown in Figure 5 . Under identical operating conditions, plate temperature and elution curves obtained from various charges of dichloromethane, ether, cyclohexane, benzene, and methanol are shown in Figure 6. DISCUSSION
OF
RESULTS
The temperature curves shown in Figures 3, 4, 5 , and 6 are superimposed on the same plate volume scale and are accompanied by the corresponding elution curves. A11 temperature curve3 and peaks are tracings from the original charts. Because of the difficulty in measuring temperature of porous structures having lorn thermal conductivity, the actual temperature change of the plate will be considerably in excess of that given by the thermocouple. The deformation of each peak was measured by its asymmetry ratio. This asymmetry ratio was similar to the skew factor described by Desty ( 2 ) and was measured a t half peak height. It was taken numerically as equal to the ratio of the distance between the normal through the peak maximum and the VOL. 35, NO. 4, APRIL 1963
485
GRAPH OF PLATE EUPEIIANR AGAINST TIME FOR DIFFERENT CUARGES bF CHLOROFORM
ELUTION CURVES FOR DIFFERENT CUARGES O F CHLOROFORM VAPOUR
COLUMN CONDITIONS COLUMN LENGlH 5 f t GLASS COLUMN RADIUS 2mm COLUMN PACKING 1 5 % SOUALANE
.lo
ffl 100-120 BS MESH CELITE
COLUMN TEMPERATURE 75.C
L
POSlTlffl
OF THERMOCOUPLE l O c m BELOW INJECTION POINT
SAMPLE CHLOROFORM VAPOUR AT 25'C
SAMPLE ( V A W R I N AIR)
ASYMMETRY RATIO r
Figure
5.
5ml
xn
Effect of charge (in vapor form)
on plate temperature and peak shape
Leff. Graph of plate temperature against time for different charges of chloroform Righf. Elution curves for different charges of chloroform (vapor)
-051
I
ASYMMETRY RATIO r
s
$59
Figure
486
ANALYTICAL CHEMISTRY
I18
6.
105
io9
Effect of partition coefficient on plate temperature and peak shape
25
CHARGE 5 0 ~ 1 OF NORMAL ALCOHOLS FLOW R A T E 2 0 0 m i / m l n
ASYMMETRY RATIO r
0.31
:
063
147
236
ture change of the plate for methanol is very small and so the asymmetry due to temperature changes will be negligible compared with the effect of adsorption. Although both phenomena produce asymmetric peaks, the asymmetry due to adsorption is different from that due to changes in plate temperature. Generally, asymmetry due to adsorption increases with the molecular weight of a homologous series. However, the deformation due to temperature changes decreases with the molecular weight and partition coefficient through the same series. The effect of changing p by the use of different flow rates with preparative scale columns is shown in Figure 6. Even though the charge has been reduced, the effect of change of flow rate from 200 to 400 ml. per minute has resulted in a greater deformation of the early peaks and improved the symmetry of the overloaded peaks later in the chromatogram.
CHARGE 12gl OF NORMAL ALCOHOLS F L O W RATE 4 0 0 r n l l m i n
CONCLUSIONS
SYMMETRY RATIO r
:0
69
128
173
250
Figure 7. Chromatograms obtained from preparative scale column at different carrier gas flow rates
trailing edge of the peak to the distance between the normal and the leading edge of the peak. The temperature curves obtained at different flow rates and consequently different values of B (Figure 3) are identical in form to the theoretical curves shoFvn in Figure 1 and support the validity of Equation 10. The asymmetry ratio of the peaks increases with the flow rate, as would be expected from the theoretical curves shown in Figure 2. According to Equation 10 the peak asymmetry and the amplitude of the temperature curves should also increase with sample size. This effect is confirmed by the results shown in Figure 4. These results, however, were obtained by using liquid samples, and i t might be argued t h a t the deformation of the peaks \vas due
to the adsorption of heat during the volatilization of the charge. The same effect, however, was observed when different charges of chloroform vapor in air were used, demonstrated by the results in Figure 5 . The effect of partition coefficient on plate temperature and peak symmetry is shown in Figure 6. Taking into account the different charges used for each substance, i t may be seen that the change in plate temperature and the peak asymmetry decrease with increasing partition coefficient. The peak for methanol is of particular interest, as it shows the effect of adsorption on peak symmetry and allows a comparison to be made between the forms of asymmetry due to the different effects. The tempera-
It has been theoretically deduced and experimentally verified that a significant temperature change can occur as a solute passes through a theoretical plate. This temperature change results in the deformation of a peak, producing a sharp front and a sloping tail. For a given set of operating conditions, peak asymmetry due to this effect becomes less, as the retention time increases. The change in temperature of the plate will vary with the operating conditions and nill affect the over-all efficiency of the column and consequently the height of the theoretical plate; it may account for some of the divergence, reported by many workers, between the theoretical and experimental values of the HETP of a given column. It will also affect the accuracy of absolute measurements of partition coefficient. Because of thermal effects in the column system, values of partition coefficients quoted in the literature may have been determined two or three degrees above the operating temperature. As the partition coefficient of a substance varies exponentially with the inverse of the absolute temperature, the published figures for partition coefficients may be less than their true values. The temperature change in a theoretical plate during the elution of a solute through it nill not be confined to conventional columns only. A1though charge sizes, plate areas, and plate capacities are several orders smaller in capillary columns, the thermal capacity of the plate is reduced1to the same extent. Therefore, chromatograms obtained from capillary columns should show peak asymmetry when thermal effects are present, in a manner VOL. 35, NO. 4, APRIL 1963
487
similar to those obtained from conventional columns. Peak deformation appears to be present in nylon capillary columns and, as far as can be judged from published chromatograms, also in glass, copper, and stainless steel capillaries. I n view of this, the small changes in partition coefficient attributed b y Desty, Goldup, et al. (5) to gas phase interaction may be in part due to this thermal effect. Data are not available to account for the effect of the charge size and flow rate used in their experiments or any thermal changes in the column. The general trend of their results shows that the partition coefficient falls with decrease in conductivity of the gas and increase in gas pressure. These results would have been expected if there had been thermal effects present in the column. I n their work examining the interfacial resistance to mass transfer between heptane and dinonylphthalate using nylon capillary columns, Hazeldean and Scott ( 4 ) found poor agreenient between the experimental and theoretical values for the resistance to mass transfer in the liquid phase. They also identified a third factor which they termed “the resistance to mass transfer a t the surface of the column.” Their experiments were carried out using different gas pressures and different film thicknesses of liquid phase, thus altering the thermal capacity and the plate capacity simultaneously. I t is difficult to assess the
total effect of these parameters on any temperature changes that might have occurred in the column. However, it is likely that either or both effects were partly due to thermal effects of the nature described in this paper. The thermal change in a partition column during the passage of a solute through it will have its most detrimental effect on the performance of preparative scale columns. It is possible that the design of these columns may have to be completely changed. The effect of column radius will be considered in another article, and until this has been done it is not possible accurately to deduce the form of a column that will reduce this temperature effect to a minimum. However, it can be generally said that the heat exchange between the column and its surroundings must be greatly improved. This will necessitate the use of high conductivity materials for column construction and, if possible, an increase in conductivity of the packing of the column. It may be necessary to use annular-shaped, or possibly laminarshaped, columns which should be designed in such a manner that the optimum gas velocity is as low as possible. Whether thermal effects are solely responsible for the lower performance of preparative scale columns relative to their analytical counterparts remains to be seen, but present work indicates that temperature variations during the transit of a solute through a theoretical plate can cignifi-
cantly reduce the performance of all gas liquid chromatographic columns. ACKNOWLEDGMENT
The author thanks C. G. S.Phillips, 3Ierton College, Oxford, for reviewing part of this paper and Dr. Crank, head of the Mathematics Department, Brunel College of Technology, for checking the mathematics. Thanks are also due to I. Fom-lis and T. Young for assistance in experimental work and P. G. W, Scott for supplying the chromatograms for the preparative scale column. LITERATURE CITED
(1) Claxton, G. C., J . Chromatog. 2 , 136
(1959). (2) Desty, D. H., ed., “Gas Chromatography,” p, 200, Academic Press. New York, 1958. (3) Desty, D. H., et al., “1962 Symposium on Gas Chromatography,” M. Van Swaay, ed., t o be published. ( 4 ) Hazeldean, G. S. F., S:ott, R. P. W., J . Inst. Petrol. 467, 380-1 (1962). (5) Keulemans, A. I. M., “Gas Chromatography,” p. 122, Reinhold, New York, 1957. (6) Klinkenberg, A., “Vapor Phase Chromatography,” D. H. Desty, ed., p. 51, Butterworths, London, 1057. ( 7 ) Ray, 3’.H., private communications; Symposium on Gas Chromatography, London, 1956. (8).Smith, J., Rubber Research Association, Welwyn, England, private communications. RECEIVED for review January 29, 1963. Accepted January 29. 1963. International Symposimn on Advances in Gas Chromatography, University of Houston, Houston, Texas. January 21 to 24, 1963.
Recirculated Column for Preparative Scale Gas Chromatography M. J. E. GOLAY,
H. 1.
HILL, and
S. D. NOREM
The Perkin-Elmer Corp., Norwalk, Conn.
b A new chromatographic system i s described which permits selection of a narrow cut and elimination of all faster and slower components of a complex sample mixture. The selected components may then b e recirculated through a set of columns until sufficiently high effective theoretical plate values have been achieved to provide the desired separation. A practical embodiment of the system i s described, and experimental results are given. The theoretical performance of the system i s discussed and compared with experimental data.
A
chromatographic process is scaled up to permit preparation of useful quantities of very pure mates
488
THE GAS
ANALYTICAL CHEMISTRY
rials, the column engineer is faced with a different design problem than that encountered in analytical columns. I n the latter, resolution is purchased ivith analysis time and pressure drop. As column size is increased, however, i t becomes appropriate to measure performance in terms of throughput per unit time per weight of expensive packing material. I n addition, the pressure drop acquires added significance in terms of the pneumatic work (or, indirectly, the expense) required to provide large quantities of carrier gas a t high pressure. A numbei of workers (2-4, 6) have noted the fact that chromatography suffers in comparison with true counter current processes such as distillation in making efficient use of the available
plates. The counter current process can proceed until the sample occupies virtually all the plates in the system, while in a once-through chromatographic column, the sample occupies only a small multiple of the square root of the number of plates traversed. Martin proposed recirculation of the sample between two columns until the band or bands of interest occupied all the plates in one column (6, 6). Independently, Saunders developed a practical recirculation system employing large-diameter columns s n d used it for preparing pure isomers for further chemical studies ( 8 ) . An alternative approach, that of pumping the sample around a single closed loop, has been described by Porter and Johnson ( 7 ) . The above schemes, while useful
