Role of Background Detector Response in Quantitative Gas Chromatography Josef Novak, Jarmila Gelbi;ova-RGi6kov&, Stanislav Wicar, and Jaroslav Janak Institute of Instrumental Analytical Chemistry, Czechoslovak Academy of Sciences, Brno (Czechoslovakia)
A theoretical analysis of the effects of the background detector response on the net response to a substance was performed. The effects were investigated experimentally with the use of two stationary phases of different volatilities, employing the flame ionization detector. The change in relative response factors caused by bleeding of the stationary phase and the systematic errors incidental to the use of a volatile stationary phase in quantitative PTGC are demonstrated,
ANYGAS CHROMATOGRAPHIC DETECTOR, operated under usual conditions in a conventional set-up, produces a certain background response. Despite attempts to compensate for this response, it usually enhances the noise level and should therefore be kept as low as possible. There are some exceptional cases in which the response t o the presence of the component eluted (net response) is given by a decrease in the background response [e.g. the electron capture detector ( I ) ] or in which the positive response is proportional to the background response (2, 3) [e.g., the alkali flame ionization detector (4, 5 ) ] . These special cases obviously require a compromise in the setting of the detector sensitivity to attain optimum performance. There is still another problem associated with the background detector response. AS shown below, in spite of various methods of compensating for the background response, it may cause a decrease in the net response. In this context, it is expedient to distinguish between two types of background response: one due t o the signal of the column effluent (carrier gas plus stationary-phase vapor), and a second type due to a gas stream fed directly into the detector (e.g., hydrogen in a FID). Each of the above types may manifest itself in various ways, depending on the type of detector. Both kinds of background response may produce a shift of the concentration range of the solute closer to or even beyond the upper limit of the overall range of linearity of detector response. This would obviously result in a tendency of the net response to become reduced. However, the first type of background response renders a n additional effect in that it causes a defined decrease of the net response even when the detector is operated well within the linear range. This effect, which seems t o have been given scant attention up to now, should apply t o any type of detector. In the present work particular attention was devoted to the above-noted special effect of the background signal of the column effluent on the net response. The background response incidental t o the additional gas stream was assumed constant and small as compared with the other sources of (1) J. E. Lovelock and S. R. Lipski, J. Amer. Chem. Soc., 82, 431 (1960). (2) M. Dressler and J. Janiik, J. Chromatogr., 44, 40 (1969). (3) M. Dressler and J. Janhk, J. Chromatogr. Sci., 7 , 451 (1969). (4) A. Karmen, ANAL.CHEM.,36, 1416 (1964). (5) A. Karmen and L. Giuffrida, Nature, 201, 1204 (1964). 1996
response. Since the flame ionization detector (6, 7) has perhaps been the one most thoroughly studied (8-13), it was chosen for this work. THEORETICAL
Since the chromatographic zone respresents a n open system, it could hardly constitute in the column, after some time of chromatographic development, a field with the state conditions differing appreciably from those existing in the given column segment when the zone is absent. It may therefore be assumed that the solute molecules are substituting those of the mixture of carrier gas and stationary phase vapor rather than being added to them during the chromatographic process. This situation has obviously no reference t o a gas stream being introduced after the column outlet-Le., directly into the detector. With the flame ionization detector, the overall background response may be assumed to be composed of the responses to the mixture of the vapor from stationary phase bleeding (b) and the carrier gas (o), Rho, and to the hydrogen being added after the column outlet, R,. Propided the above individual response components are additive, the background response (Rboa)is given by the sum Rbo R,. When the vapor of solute (i) appears in the column effluent, the overall response (Rtbo,) will be composed of the response to the solute-stationary phase vapor-carrier gas mixture, R,bo, and that to the hydrogen, Le.. Rzhoa = R,,, R,. The net response, R , , is the difference R,boa - Rboa, so that
+
+
Ri
=
Rz~o- Rho
(1)
The responses Rihoand Rbo can be expressed by means of theequations R l h o = k(f;nl -k fbnb’ -!- fono’) and Rho = k(f,nb+ fano)where k is a n apparatus constant, n and n‘ stand for the rate (in moles per second) of supplying the individual substances into the detector, and the f ’ s are the respective molar response factors [sums of the effective carbons (IO)]. Since the rate of supply of a given component of a gaseous mixture is given by the product of the rate of supply of the mixture and the mole fraction of the component, the last two equations may be rewritten to read R t h o = kn(fiy2 4(6) J. Harley, W. Nel, and V. Pretorius, Nature., 181, 177 (1958). (7) I. G. McWilliam and R. A. Dewar, ibid., p 760. (8) L. Ongkiehong, Ph.D. Thesis, University of Eindhoven, The Netherlands, 1960. (9) D. H. Desty, C. J. Geach, and A. Goldup, in “Gas Chromatography 1960,” R. P. W. Scott, Ed., Butterworths, London, 1960, p 46. (10) J. C. Sternberg, W. S. Gallaway, and D. T. L. Jones, in “Gas Chromatography,” N. Brenner, J. E. Callen, and M. D. Weiss, Ed., Academic Press, New York, 1962, p 231. (11) P. BoEek, J. Novrik. and J. JanBk, J. Chromatogr., 43, 431 (1969). (12) P. BoEek, J. NovBk, and J. JanBk, J . Chromatogr. Sci., 8, 226 (1970). (13) J. Noviik, P. BoEek, L. Keprt, and J. JanBk, J. Chromatogr., 51, 385 (1970).
ANALYTICAL CHEMISTRY, VOL. 4 3 , NO. 14, DECEMBER 1 9 7 1
20
c
! u 80 60
70
90
1.
100
I 0-9
temperature ( " C )
Figure 1. Dependence of peak area on column temperature fbyb'
+ foro')and Roo = knCfayb + &yo) where n denotes the
rate of supply of the mixture (column efffuent) and y and y ' stand for the mole fractions. Hence, it may be written for the net response :
Ri
=
kn[hyt
- f b b b - ybf) - &(yo
- Yo')]
+
ybf = yb(1
- Yi)
(3)
The substitution of yb' in Equation 2 by the RHS of Equation 3 results in
Ri
=
k4j-i
- fbydyt
(4)
Upon integrating Equation 4, the respective relationship may be expressed in terms of the peak area and the total amount of the substance present in the chromatographic zone. Since the chromatographic peak area, A i , is proportional to the integral
l
Ridt, and the number of moles of substance i,
J:
Ni, present in the chromatographic zone is equal to
nyidt
(tl and t2 stand for the times of the beginning and the end of elution of the zone), the peak area is given by
where K is again an apparatus constant. It can be inferred from Equation 5 that: the peak area corresponding to a given amount of the substance chromatographed will be decreased by a value proportional to the product of the mole fraction and molar response factor of the signal-producing component in the carrier gas. The absolute decrease of the peak area is proportional to the amount of the substance chromatographed, Le., the relative decrease of the peak area is independent of the amount of the substance chromatographed. The relative decrease of the peak area of a given amount of the substance chromatographed is the larger the smaller is the response factor of this substance. As long as the rate of supplying all the signalproducing components (solute stationary phase vapor additional gas) into the flame is kept within the limits compatible with the linear range, the decrease of the net response
+
apiezon K
70
80
90
100
temperature ( 'C
)
Figure 2. Dependence of background response on column temperature
(2)
Since fo is practically zero for carrier gases conventionally used with the flame ionization detector, the term &(yo - yo') will be omitted hereafter. The value of yo' is a variable depending on the given value of y b and the instantaneous value of yi. Owing to the above mentioned properties of the chromatographic zone, there holds y i ybf yo' = ~6 yo = 1 and y d y o = yb'/y0', which gives, after a simple rearrrangement,
+ +
60
- .
+
due to that of the background does not make the net response nonlinear. The background response may also affect appreciably the values of experimentally determined relative response factors. The relative molar response of a substance i, expressed with the use of a reference substance r , RMRtr, is defined by RMRi, = ( A i / A r ) (Nr/Ni),which equation may be rewritten by virtue of Equation 5 to read:
RMRir
=
(h- f b ~ d / W- f b ~ d
(6)
It is apparent from Equation 6 that the actual RMRir value may depend on the degree of purity of carrier gas, kind of stationary phase, and column temperature, particularly with substances of smallfvalues. EXPERIMENTAL
The experiments consisted in analyzing at various temperatures defined amounts of a model mixture on two different stationary phases while keeping all the working conditions of the detector strictly constant. The chromatograms obtained were evaluated quantitatively and the results correlated with temperature. In order to make the above-indicated effect sufficiently distinct, model substances differing appreciably in their response factors and stationary phases of substantially different volatilities were used. The model mixture was a solution of dichloromethane and hexane in benzene (0.0266 and 0.0131 gram/ml, respectively). All the chemicals mentioned were analytical-grade products of Lachema, N. E., Czechoslovakia. The stationary phases employed were Apiezon K (Associated Electrical Industries Ltd., England) and hexadecane (The British Drug Houses Ltd., England). Both stationary phases were deposited in amounts of 20Z by weight on Chromosorb W 60-80 mesh (Carlo Erba, Italy). The measurements proper were carried out on a HewlettPackard 402 gas chromatograph, employing the dual FID version. With both kinds of stationary phase, 6-foot long glass columns of 3-mm i.d. were used in both the operating and reference channels. The weight of the Apiezon and the hexadecane packings in the operating columns were 5.26 and 5.21 grams. In all cases, the flow rates of the carrier gas (NJ and hydrogen in the operating channel were 0.56 and 0.33 ml/sec as measured at the burner jet outlet and expressed at 25 "C and 735 mm Hg. Except in the case of studying the dependence of the response on the air flow rate (see below),
ANALYTICAL CHEMISTRY, VOL. 43, NO. 14, DECEMBER 1971
1997
A (
I
I
i.c. 1
30
20
10 I
3
I
I
I
]
6 9 12 air flow rate ( ml/sec 1
2
Figure 3. Dependence of peak area on air flow rate
the latter was always 4.17 ml/sec (at 25 "C and 735 m m Hg). The flow rates of the gases in the reference channel were adjusted to obtain a complete compensation of the background response without the use of the electrical bucking off system. The columns were operated at temperatures of 60, 70, 80, 90, and 100 "C while the detectors were kept at 120 "C. Since differential flow controllers (installed ahead of the column inlet) were used for stabilizing the carrier gas flow rate while maintaining the detector temperature constant, the flow rate of the gas through the jet orifice was believed t o be practically independent of the changes in the column temperature. Hamilton 7001 and 7005 syringes were employed for injecting the samples. The peak areas were evaluated with a disc and ball integrator (Disc Instruments, Inc., U.S.A.). RESULTS AND DISCUSSION
The extent to which the peak area may be affected by the stationary phase volatility is illustrated in Figure 1. Curves 1 and 1' represent the dependences on the column temperature of the peak areas of dichloromethane and hexane analyzed on the Apiezon K column, whereas curves 2 and 2' show a n analogous situation in the case of the work with the hexadecane column. At each temperature quoted, four chromatograms of 1 pl of the model mixture were run a t a sensitivity attenuation of 4 X l o 3 ; the points in the plots represent the averages of the four determinations. The peak areas are expressed in the number of integrator counts (is.), and the standard deviation of a single area determination varied within 0.5-1.5 i.c. Figure 2 shows the dependences on the column temperature of the background detector response (background ionization current) incidental to the work with the Apiezon K and hexadecane columns. The background ionization current was measured after each chromatographic run and the values obtained were averaged. The plots in Figures 1 and 2 provide for a direct correlation of the peak area decrease with the background detector response; the fall of the peak area upon increasing the background response is evident from the figures. It was also thought that the rate of air supply might be an important factor a t relatively high concentrations of the vapor of stationary phase in the flame. If the observed 1998
4 sample volume ( yl 1
Figure 4. Dependence of peak area on sample amount injected
reduction of peak area was due to the above factor, one could expect the dependences of the peak area on the air flow rate t o display different courses with the Apiezon K and the hexadecane columns. Figure 3 shows such dependences; the data were obtained from chromatograms of 1-111 samples of the model mixture, run at a temperature of 100 "C. Curves 1 and 1' correspond to dichloromethane and hexane chromatographed on Apiezon K and curves 2 and 2' refer to the same pair of substances analyzed on hexadecane. All the curves are similar to each other as t o their courses; the curves corresponding to the measurement on the hexadecane column display the same tendency to reach a plateau a t higher air flow rates than those obtained with the Apiezon K column. Hence, the air flow rate may be supposed to be a noncritical factor in the problem studied. The possibility that a narrowing of the utilizable range of the detector response linearity due t o the presence of the stationary phase vapor was responsible for the reduction of peak area was checked by analyzing various amounts of model mixture on the hexadecane column at 100 "C and plotting the peak areas against the sample amounts injected. This is shown in Figure 4. The sample charge was varied over about a n order of magnitude (0.4-4 ~ 1 attenuation , factor 16 X lo3) without causing the response t o become apparently nonlinear. This would hardly be so if the range of the sample amount had fallen into a nonlinear region of the detector response. An interesting picture is obtained when plotting the peak height expressed in the units of detector response (ionization current) against the corresponding rate of mass-supply into the flame. The rate of supplying the solute from the column into the detector may be varied by changing the carrier gas flow rate, sample amount injected, and column temperature. Provided the carrier gas flow rate through the detector is kept constant, a change of the mass-supply rate caused by changing the column temperature a t a constant sample amount and the same change brought about by changing the sample amount at a constant column temperature may be equivalent in their effects on the net response only if the background response due to column bleeding is not appre-
ANALYTICAL CHEMISTRY, VOL. 43, NO. 14, DECEMBER 1 9 7 1
Table I. Comparison of Basic Quantitative Data Obtained in Isothermal and Temperature Programmed Runs Quantity determined Operating conditions Ad Ah ad ah had Aah A&* Ash* RMRdh Isothermal run (100 "C) ApK 14.2 44.7 24.1 75.9 ... ... ... ... 0.175 HD 9.7 32.5 23.0 77.0 ... ... -4.65 1.48 0.166 ApK 15.6 48.3 24.4 75.6 1.24 -0.39 ... ... 0.179 Temperature programming (7.5 "C/min; 60-100 "C) HD 14.8 41.0 26.5 73.5 15.4 -4.58 8.64 -2.79 0.200
Ri
(amp.)
I
analyzed on hexadecane) correspond to the cases where the rate of solute supply was controlled by the column temperature (cf. Figure 1). The dashed curves (3 and 3' refer to dichloromethane and hexane analyzed on the hexadecane column kept at 100 "C) refer to the case where the solute supply rate was controlled by the sample amount injected (cf. Figure 4). Figure 5 is a persuasive verification of the above presuppositions. Owing to a strong dependence of stationary phase bleeding on the column temperature, the effect of the background response of the net one may be the source of an appreciable systematic error in quantitative programmed temperature gas chromatography if the conventional technique of internal normalization is used to evaluate the chromatogram. This situation is apparent from the data given in Table I where A stands for the absolute peak area (in the i.c. units) and a for the respective area fraction, Aa denotes the systematic error incidental to temperature programming, and Aa* denotes the error associated with the volatility of stationary phase in general. The subscript d denotes dichloromethane and h hexane, and ApK and H D are abbreviations for the Apiezon K and hexadecane columns. The quantities a, Aa, and Aa* are defined by 1 0 0 A t / ~ A , ,100[a,(PT) -
I
20
,
I
I
; 15 -
I I
10
5
0.05 0.10 mass supply rate (*
0.15 )x
10-6
z
Figure 5. Dependence of net response on rate of solute supply
ciable. With columns containing volatile stationary phases, a fall-off of the net response should occur at higher mass rates when the latter is increased by raising the column temperature at a constant sample amount, whereas a straight line, with the slope inversely proportional to the column temperature, should be obtained if the net response is plotted against the mass rate varied by changing the sample amount at a constant column temperature. This situation is shown in Figure 5. All the data necessary to construct the plots in this figure were obtained by processing the experimental material used for constructing the plots in Figures 1 and 4. The rate of solute supply (in moles per second) corresponding to the peak maximum, v, was calculated from the known weight amount (w) of the component injected into the gas chromatograph, time interval (At) of the elution of the zone, and the carrier gas flow rate (Fo) at the column outlet; the approximative relation v = 2w/(At)M was used where M stands for the molecular weight of the given substance. The At values were calculated by (14) d(2/2/21n2)/c where d is the peak width measured at the half-height and c is the recorder chart speed. Curves 1 and 1' as well as 2 and 2' (1 and 1' refer to dichloromethane and hexane analyzed on Apiezon K, and 2 and 2' refer to the same pair of substances (14) H. Purnell, "Gas Chromatography," Wiley, New York, N.Y., 1962, p 105.
at(IT)l/at(IT), and 1 0 0 M H D ) - at(ApK)l/at(ApK), ( P T ) and ( I T ) denoting programmed temperature and isothermal runs, respectively; RMRdh is the relative molar response of dichloromethane, expressed with the use of hexane as the reference compound. In the PT runs, the initial temperature and the rate of temperature rise were 60 "C and 7.5 "C/min with both stationary phases. The retention temperatures of dichloromethane and hexane were 73 and 82 "C with the Apiezon K column and 74 and 95 "C with the hexadecane one. The experimental findings concerning the effect of the background response on both the absolute and relative values of the net response are qualitatively in good agreement with the above-stated theoretical concepts. However, since the vapor pressure of hexadecane is only about 1 mm Hg at 100 "C, the actual decrease of the net response appears to be larger than that which may be predicted by virtue of Equation 5. Owing to the fact that the net response was practically linear within the employed range of sample amount, the discrepancy could hardly be explained by the narrowing of the overall range of response linearity due to the presence of hexadecane vapor alone in the flame. Namely, the mole fraction of hexadecane in the column effluent amounted to about 0.0014 at 100 "C, which concentration was comparable with those of dichloromethane and hexane (0.0017 and 0.0046) in the respective zone centers when analyzing l-pl samples of the model mixture on the hexadecane column at 100 "C. However, even when the concentration of stationary phase vapor in the column effluent is well below the upper limit of the linear range of detector response, the concentration of
ANALYTICAL CHEMISTRY, VOL. 43, NO. 14, DECEMBER 1971
1999
the combustion products of the vapor may already be so high as to bring about alterations in the properties of the flame and, consequently, in the magnitude of the net response. This may also constitute a factor contributing to the decrease of the net response. If the decrease were due to some variations in the working parameters (detector temperature, flow rate of the gases), it should be the same with both types of stationary phase. The above account indicates that the enhancement of the
decrease in the net response due to the background one is likely a case applying especially to the flame ionization detector. Since this effect cannot be suppressed by any method of compensating for the background response, it is advisable that the use of stationary phases producing a high background signal be avoided in accurate quantitative analysis by gas chromatography. RECEIVED for review July 19, 1971. Accepted August 30, 1971.
Evaluation of Dendritic Salt as a Support for Gas-Liquid Chromatography R. D. Schwartz, R. G . Mathews, J. E. Rountree, and D. M. Irvine Pennzoil United, Inc., P.O.Box 1407, Shreoeport, La. 71158 A commercially available dendritic salt has been evaluated as an inert support for packed column gas-liquid chromatography. The results obtained indicate that efficient columns may be prepared with this material. Further, polar compounds may be separated with nonpolar liquid phases when salt i s used as the support. The pressure drop of salt columns is low and no tailing, loss by adsorption, or reaction of polar compounds with this support were noted.
SILICIOUS MATERIALS, of high pore volume and low surface area, are currently employed for nearly all packed column gas-liquid chromatography ( I ) . These silicas, often treated with acid and/or alkali, as well as with silanes, are called “inert solid supports.” Their function is to hold the liquid phase utilized to provide the desired separation. In practice, the silicas have certain disadvantages, particularly when polar molecules are being separated. The surface of silica is reactive and can yield tailing peaks, nonquantitative recovery, and chemical decomposition or reaction of labile molecules. These problems, could, perhaps, be minimized if a nonpolar inert solid support were available. Inorganic salts such as sodium chloride have been utilized as supports and as modifiers of other adsorbents (2,3). Recently, a dendritic form of sodium chloride became commercially available. Dendritic crystals are branched or starlike in form. This particular Torm of sodium chloride is a six-pointed star crystal with an extremely jagged and irregular surface and an interior containing millions of cavities of only several microns diameter ( 4 ) . The material has the ability to hold liquids in its pores. An investigation was initiated to study the gas-liquid chromatographic behavior of packed columns prepared with dendritic salt as an inert solid support. EXPERIMENTAL
Materials. Dendritic sodium chloride-obtained as a sample from the Morton Salt Co.-was screened prior to use ( I ) D. M. Ottenstein, J . Gus Clrromutogr., 1, 11 (1963). (2) F. R . Cropper and A. Heywood, Ncrture, 174, 1063 (1954). ( 3 ) C. G. Scott and C . S. G . Phillips 111 “Gas Chromatography 1964,” A. Goldup, Ed., Institute of Petroleum, London, 1965,
p 266. “The Unique Physical Properties of Star Flake Dendritic Salt,” Morton Salt Company, Chicago. 111. 60606.
(4)
2000
C6 t a C i 0 t Benzene
i n j e c t i o n Paint
Tlme in Mlnutes-
7. 0
1
2
Figure 1. Column B. Room temperature; 20 psi helium. Cgto CloNparaffins plus benzene to 60-80 mesh and dried at 120 “C. Gas Chrom Q, 60-80 mesh, a silanized silicious support material, was obtained from Applied Science Laboratories, State College, Pa. The tubing used for the columns was l/s-in. 0.d. Type 316 Stainless Steel, from Handy & Harman Co., Norristown, Pa. Reagents. Squalane was obtained from Eastman Chemical Co. Carbowax-400 was a sample from Union Carbide Chemical Co. PZ-110, a polyimide, (5) was prepared at Pennzoil United, Inc. Igepal CO-210 was a sample from General Aniline and Film Corp. The chloroform used in preparing the column packings was Baker Analyzed Reagent Grade. Column Preparation. The screened support materials, Dendritic Salt and Gas-Chrom Q, were coated by slurrying with a chloroform solution of the liquid phase and subsequent ( 5 ) R . G . Mathews, R. D. Schwartz, M. Novotny, and A. Zlatkis, ANAL.CHEM., 43, 1161 (1971).
ANALYTICAL CHEMISTRY, VOL. 43, NO. 14, DECEMBER 1971
