3topper and electrodes into the cell as shown in Figure 1. Pass the sample through the cell at a rat? sufficient to prevent the solid carbonate from settling out until a constant p H is obtained. This may require an hour, if the carbon dioxide content of the sample is low. Measure the equilibrium p H and determine the carbon dioxide content of the sample from the calibration corresponding to the carbonate. the cell solution, and the temperature used for the measurement. RESULTS AND DISCUSSION
A cylinder of pure carbon dioxide, a cylinder of ethylene containing 4000 p.p.m. of carbon dioxide (by mass spectrometer analysis), and a source of compressed air which contained 300 p.p.m. of carbon dioxide were used. A cylinder of nitrogen !vas also used to compare different calibrations. Although its exact carbon dioxide content n-as not knonn, consistent results n-ere obtained with the calibrations by assuming it to contain 0.7 p.p.m. of carbon dioxide. Measurements using these gases and slurries of calcium, strontium, and lead carbonates are summarized in Table I. The results obtained with lead carbonate proved to be erratic and not reproducible, as shown in Table I. This may have been caused by precipitation of basic carbonates in the lead carbonate solid phase. As the result of these measurements, the xork Kith lead carbonate was discontinued. Results obtained with four separate runs using three different strontium carbonate slurries are reproducible to within the precision of the meter readings in every case. The same reproducibility of results was obtained n i t h a series of calcium carbonate slurries. I n
Figure 2, these data are plotted to shorn the linear relationship betri-een p H and log concentration of carbon dioxide. The effect produced by a soluble salt of the same metal as the carbonate being used may be seen by comparing the two sets of measurements made with calcium slurries. The presence of calcium chloride reduces the carbonate ion concentration in the solution and increases the acidity measured a t any given concentration of carbon dioxide. A linear plot of log concentration of carbon dioxide t's. p H is still obtained (Figure 2 ) , but its slope differs from those obtained with distilled water as the cell solution. The effect of temperature on these measurements may be observed by comparing the values obtained with a strontium slurry at 0" and a t 25" C. The higher acid readings a t 0" C. are the result of increased solubility of carbon dioxide in water as the temperature is loviered. The reproducibility of the results obtained with several batches of a given carbonate under the same experimental conditions indicates that it is possible to prepare ahsolute calibrations which can be used without rechecking for each new batch of carbonate or cell solution. However, the composition of the cell solution affects the position and the slope of the calibration. For this reason, distilled water is preferable as the cell solution, rather than a salt solution whose roniposition may be affected by evaporation during long periods of operation. The lower slope of the calibrations made with distilled n ater increases the scnsitivity of the method. Therefore, for routine application of this method, strontium carbonate or calcium carbonate in distilled water a t
0" or 25" C. is reconiinended for the determination l p.p.m. to 100% of carbon dioxide in a gas stream. This method has been used principally for measuring the carbon dioxide content of cylinder gases by passing the gas through the cell for several hours and then reading the p H of the solution in the cell. The rate of attaining equilibrium u-as not studied in detail, but it was observed that equilibrium )vas reached more rapidly when the carbon dioxide content of the sample n as high, and when the equilibrium p H !vas approached from the high side. Application to Continuous Process Control, The method should provide a continuous indication of the carbon dioxide content of a gas stream. By addition of recording and controlling equipment, it could be adapted to continuous measurement and control of carbon dioxide content. i i n instrument based on this principle lvould not provide the rapid response of an infrared instrument, but its simplicity, low cost, and freedom from calibration problems should make it worthy of consideration for process control applications. ACKNOWLEDGMENT
The authors wish to express their appreciation to the Phillips Petroleum Co. for permission to publish this paper. LITERATURE CITED
(1) hlaxon, W. D., Johnson, SI. J., ;ZSAL. CHEM.24, 1641-5 (1952).
(21 Quinn. E. L.. .Jones. C. L.. "Carbon Dioxide," ACS Monograph 72, p. 121 ff ., Ken, York, Reinhold, 1936. \
,
I
RECEIVED for review April 8, 1957. Accepted July 26, 1957. Pittsburgh Conference on Analytical Chemistry and .ipplied Spectroscopy, 1957.
Derivative Thermometric Titrations S. T. ZENCHELSKY and P. R. SEGATTO Ralph G. Wright Laborafory, School of Chemistry, Rutgers Universify, New Brunswick, N. f.
b Thermometric titration curves resemble conductometric titration curves. It is frequently difficult to locate the end point by extrapolation procedures. Derivative titration curves accentuate the end point change. Several methods for differentiating a voltage-time function are available, but the low output voltage of the thermistor bridge and the slow attainment of thermal equilibrium within the titration vessel pose special problems. An apparatus for obtaining first and second derivative curves is described. Recorder plots of the derivative curves are presented, with the results of several titra1856
ANALYTICAL CHEMISTRY
tions used to ascertain the accuracy and reproducibility of the method.
I
N A x n m m of recent papers on thermometric titrations (3, 4, 6 ) the end point has been selected as in a conductometric titration, because the data plots are similar for the two types of titrations. The inherent disadvantages of these curvestemperature-volume, conductancevolume-for end point location lies in the fact that the end point region may be considerably "rounded" or the difference in slopes of the linear
portions of the curve, on either side of the end point, niay be rather small. B y successively differentiating the original data curve, the end point region is accentuated, with a resultant gain in the precision of selection. 1Ialmstadt and Roberts ( 5 ) discuss this procedure in connection with photometric titrations, where the curves resemble those of thermometric and conductometric titrations. I n thermometric titrations a thermistor-bridge assembly is commonly employed as the transducer for converting temperature changes into voltage variations. The resultant recorder plot
c3
BRIDGE
MECHANICAL
SERVO POTENTIOMETER
INPUT
AMPLIFIER
11 1 c4
R4
R5
I I
FIRST
SECOND
D IFFE REN T I AT 0 R
DEWAR
It
I RECORDER
Figure 1.
DIFFERENTIATOR
FILTER
Block diagram of apparatus
i f FILTER
Figure 2.
Filter and differentiator networks
R,, R2. R3 0.5 megohm R4, Rs.1 megohm CI, CZ.10 mfd., Pyranol C,, C4. 0.5 mfd., paper
I B
A
Figure 3.
I M
ACID
C
Recorder plot of first derivative curves
0.6 ML.
A . Ii-ithout filter B. With filter of smaller time constant than that of Figwe 2 C. With filter as in Figure 2
is one of voltage 1’s. time or titrant volume, when the titrant is dclivered a t a constant rate. This voltage-time function may, in principle, be differentiated by several procedures (6),the most common being the use of an R-C circuit or a tachometer generator. For thermometric titrations, two factors make the differentiation difficult: The signal level is lon-several millii oltsand the rate of voltage change is small. The binglc level niay be raised by ainplification and the rate of voltage change niay be increased by faster delivery of titrant. Hoirerer, even n i t h esticmely efficient stirring (j),heat traiiqfer problem- pre\-ent titration rates greater than about 2 nil. per minute. -4s the successi1-e differentiatiom r d u w the signal, great amplification i- I equired. The desired gain i b not c a i i l ~ a c l i i c ~ed through the use of direct current amplifiers, as to reach the desired output voltage level would require cum1)ersome power supply coniponente. Another method of amplification was therefore used. EXPERIMENTAL
Amplification. Mechanical aniplification (Figure 1) was achieved by mechanically coupling a helical potentiometer t o t h e slidewire of a servo
(self-balancing) potentiometer. Thus the output of the potentiometer n a s proportional to its rotation (input signal), the proportionality constant being determined by the fixed i-oltage across its end terminals. For a 2.5-mv. span recorder and a 250-volt battery across the helical potentiometer, the voltage gain is 105. The servo potentiometer TTas of conventional design (I), with a Brnv n conrertcr and amplifier. Filtering. Random tempeiature fluctuations. inherent in thermometric titration, arise from the slow attainment of thermal equilibiium in the solution. The effect is more pronounced a t high titration rates, and is diminished by rapid stirring, but it could not be eliminated eiren by extrcmely efficient stirring ( 5 ) . As a result, the signal-noice ratio is small. I n the voltage-time plot, these fluctuations are hardly noticeable; but their effect is accentuated by the successive differentiations, and the signalnoise ratio decreases because the noise frequency, and thuq its rate of change, is greater than that of the signal. To diminish this effect, a filter was used betn een the meclianical amplifier and differentiator (Figure 1): an R-C network (Figure 2 ) whose time constant was sufficient to increase the
0.5 MV.
0.5M
A C I D
0.25M
A C I D
Figure 4. Recorder plot of second derivative curves
signal-noise ratio without appreci:ibly reducing the signal level. Differentiation. Although R-C networks (Figure 2) d o not produce a n output signal exactly proportional t o the derivative of the input signal (Z), the result is sufficiently close to the desired function. Moreover, signal fidelity has already been sacrificed by inclusion of the filter network. RESULTS
Typical first derivative curves (Figure 3) illustrate the effect of filtering. Although the signal-noise ratio is improved, a response lag is introduced determined by the time constant of the filter network, and the end point does not coincide with the inflection point of the curve. Second derivative curves (Figure 4) were obtained by using the filter of Figure 2. Here the end point coincides with the origin of the deflection, VOL. 29, NO. 12, DECEMBER 1 9 5 7
1857
rather than with the tnaximum, as would be expected if the true second derivative of the thermometric titration curve were obtained. The required filtering, however, prevents presentation of the true second derivative. This does not diminish the usefulness Qf the procedure, for the end point may be selected precisely and accurately if the origin, rather than the maximum, of the deflection is chosen. The results of several titrations are presented in Table I. End point was selected, in each case, a t the origin of the deflection in the second derivative curve. Titrant delivery rates were 1 ml. per minute.
Table 1. Titration of Sodium Hydroxide with Hydrochloric Acid
Base Taken, Mmoles
Acid Concn., N
1.25 2.50
0.25 0.50
Relative Mean Error,
%
No. of Detns.
0.4 0.2
16 6
port of a part of this work by a Cottrell Grant of the Research Corp. LITERATURE CITED
( 1 ) Brown Instruments Division, Engi-
neering Dept., Minneapolis-Honeywell Regulator Co., Philadelphia, “Adaptability of the Measuring
ACKNOWLEDGMENT
The authors wish to acknowledge sup-
Circuit, Input Circuit, and Amplifier of the Brown Electronik Potentiometer,” Tech. Bull. B15-10
(1950). (2) Greenwood, I. A., Holdam, J., Jr., MacRae, D. Jr., “Electronic Instru~
ments,” Radiation Laboratories Series, MIT, Vol. 21, pp. 64-78, New York, McGraw-Hill, 1948. (3) Jordan, J., Alleman, T. G., ANAL. CHEM.29, 9 (1957). (4) Linde, H. W., Rogers, L. B., Hume, D. N., Ibid., 24, 1348 (1952). (5) . , Malmstadt. H. V.. Roberts. C. B..
Zbid., 28,‘ 1408 (1956).. (6) Zenchelsky, S. T., Penale, J., Cobb, J. C., Ibid., 28, 67 (1956).
RECEIVED for review June 6, 1957. Accepted August 8, 1957. Division of Analytical Chemistry, Beckman Award Symposium Honoring Ralph H. Muller, 131st Meeting ACS, Miami, Fla., April 1957.
Distribution of n-Paraffins and Separation of Saturated Hydrocarbons from Recent Marine Sediments E. D. EVANS, G. S. KENNY, W. G. MEINSCHEIN, and E. E. BRAY Field Research laboratory, Magnolia Petroleum
The hydrocarbons in recent marine sediments are of interest because of their possible relationship to petroleum. Certain differences in sediment and crude oil hydrocarbons have been observed. In the investigation of these differences alumina and silica gel chromatographic separations were used and significant variations were found in the fractions obtained from these adsorbents a t high gel-sample ratios. To facilitate the study of the separation on the alumina and silica gel columns, cholestane and n-hexacosane were added to a saturated hydrocarbon fraction of a recent marine sediment. Mass spectrometric analyses of the chromatographic fractions established the efficiency of the separation of naphthenic and paraffinic hydrocarbons on alumina columns a t different adsorbent-sample ratios, and showed that n-paraffin mixtures are separated on a molecular weight basis on alumina a t certain adsorbent to sample ratios. Silica gel was not so selective as alumina for the described separations. The n-paraffins in recent marine sediment extracts, unlike petroleum paraffins, contain higher concentrations of odd-carbon-number molecules than of their even-carbon homologs.
T
HE COMPOSITION of the organic extracts of recent marine sediments is of interest because most crude oils
1858
ANALYTICAL CHEMISTRY
Co., Dallas,
rex.
are found in marine environments and are believed by many authorities to have originated in marine sediments. Evidence in support of this view was who showed the provided by Smith (I), presence of liquid hydrocarbons in the extracts of recent marine sediments, Similar extracts have been studied in the authors’ laboratory, and as reported by Stevens, Bray, and Evans (6) the hydrocarbons in shallow, recent marine sediments in the Gulf of Mexico differ in certain respects from those found in crude oils. Mass spectra indicate, for example, that the n-paraffins in crude oils are evenly distributed between molecules of even and odd carbon number while those from the recent marine muds have higher concentrations of molecules of odd carbon number. The greater concentration of odd-carbon n-paraffins suggests that these hydrocarbons in sediment extracts more closely resemble plant and animal paraffins than petroleum paraffins. This paper presents conclusive proof that the n-paraffins in some recent marine sediments have a preference for odd-carbon-number molecules. Silicagel and alumina chromatography were used to fractionate the sediment extracts and differences were observed in the fractions obtained from these adsorbents. I n particular, the mass spectra of the n-heptane eluates from alumina did not show the odd-carbon preference that was apparent in these
fractions from silica gel. Because of the general need for an efficient means to separate saturated hydrocarbons, the separations of these hydrocarbons on alumina were investigated. It was found that n-paraffis, cycloalkanes, and n-paraffins of different molecular weights can be separated on alumina a t high adsorbent-gel ratios. REAGENTS
Silica gel, commercial grade, 100 to 200 mesh, Davison Co. Activated prior t o use a t 42.5” C. for 6 hours. Alumina, activated powdered catalyst grade AI-OlOlP, Harshaw Scientific Co. Activated prior to use a t 343” C. for 15 hours. Extraction and chromatographic solvents. All solvents were carefully distilled and checked to contain less than 0.3 y per ml. of impurities which were not volatile under the conditions employed for sample recovery. Urea, C.P. or equivalent, recrystallized from ethyl alcohol. n-Hexacosane and cholestane, obtained from Pennsylvania State University. PROCEDURES
Silica G e l Chromatography. The marine sediment extracts were fractionated on two sizes of columns. Large samples, 0.3 to 1 gram, were separated on 120-gram gel columns 24 cm. in length and 28 mm. in diameter. Small samples, 20 mg. to 0.3 gram, were fractionated on 9-gram gel columns 18 cm. in length and 9 mm. in diameter.