Comparison of Gas Chromatographic Method and pH-Alkalinity Method for Determination of Total Carbon Dioxide in Sea Water SIR: Swinnerton, Linnenbom, and Cheek (2) determined dissolved gases in aqueous solutions by gas chromatography. They used an all-glass sample chamber in which dissolved gases were stripped from solution by an inert carrier ga.. We recently extended their method to determine total carbon dioxide in sea water, and compared our results with those of the widely used pH-alkalinity method (1). Apparatus and procedure were essentially the same as those described by Swinnerton, Linnenborn, and Cheek ( 2 ) . However, we used only one silica gel column to resolve the carbon dioxide chromatogram. To convert dissolved bicarbonate and carbonate ions in sea water into carbon dioxide, a 3-ml. sea water sample was mixed with 2 ml. of O.lLITHC1 in the sample chamber, followed by stripping of the gas formed. One- to 3niA11per liter solutions of sodium carbonate were used as standards. Reproducibility of the chromatographic method was good, with a relative standard deviation of +0,7% for five replicate det.erminations. To further test the validity of the chromatographic method, we compared it with the pH-alkalinity method (1) a t various salinities. Low salinity samples for this test were prepared by diluting sea water with distilled water. The two methods agreed well within a
Table 1. Comparison of Two Methods of Total Carbon Dioxide Determination
Total carbon dioxide,
mM per liter Gas SampHchrople Salinalka- matonum- ity, linity graphic Differber p.p.t. method method ence 1 2 0 019 020 001 2 5 2 045 041 002 3 10 3 0 76 0 80 0 04 4 15 5 1 10 1 12 0 02 5 21 1 1 44 1 42 0 02 6 26 4 1 75 1 76 0 01 7 31 3 2 05 2 01 0 04 8 33 8 2 19 2 16 0 03 9 36 4 2 33 2 31 0 02
maximum difference of 0.04m.V per liter (Table I). The pH-alkalinity method relies on measurement of pH, alkalinity, salinity, inorganic boron concentration, and temperature of sea water (1). Furthermore, we need to know dissociation constants of carbonic acid and boric acid in sea water to calculate the total carbon dioxide. Cnder ideal conditions relative standard deviation obtainable from this method is about +1y0. The chromatographic method offers a direct and rapid means of determining total carbon dioxide in sea water. I t takes not more than 5 minutes to analyze a sample. Further tests show
that the method is reliable aboard ship, and it in applicable to rain, river, and lake waters. Therefore, the modified method of Swinnerton, Linnenbom, and Cheek ( 2 ) should prove more useful than the pH-alkalinity method in routine oceanogrsphic, limnological, and water pollution work>. ACKNOWLEDGMENT
The authors thank Magdalena Catalfomo for completing a portion of the laboratory work, and J. P. Sullivar., c‘. S. S a v a l Oceanographic Office, and J. W.Swinnerton, U.S. Naval Research Eaboratory. for their technical assistance. LITERATURE CITED
(l),Harvey, H. IT., “Chemistry and Fertility of Sea-Waters,’! p. 172, Cambridge Univ. Press, Cambridge, Mass., 1960. ( 2 ) Sw-innerton, J. W., Linnenbom, V. J., Cheek, C. H., AXAL. CHEM.34, 483 (1962). KILHOP ~ R K GEORGEH. KENNEDY HI-GHH. I > O B S O S Department of Oceanography Oregon State University Corvallis, Ore. 97331 Work supported by Xational Science Foundation grants GP-622 and GP-2232, and by the Ofice of Naval Research contract Xonr 1286(lo), Project, ?;R 083- 102.
L n d iu m
Enca psula tion Technique for Introducing Weighed Samples in Gas Chromatography SIR: Obtaining known amounts of representative samples of volatile liquids for gas chromatography by injection techniques is difficult (4). Variations in volume injected are a source of error (8). I3ecause these variations depend on the depth and time of insertion of the needle into the preheat zone, selective vaporization may be a significant part of the error. Select,ive va1)orization is also likely when samples are stored before analysis. As a result, the accuracy of the analysis is uncertain because of sampling errors. Encapsulation ensures introduct,ion of a known amount of a representative , even after prolonged storage. Errors arising from variations in volume and manipulation of the syringe are avoided. Because thr entire sample within the capsule (tntcrs the gas chromatography in1686
ANALYTICAL CHEMISTRY
strument, selective vaporization is eliminated. Weighing the encapsulated sample eliminates the need to add a reference compound. Therefore, accuracy of gas chromatography may be determined without the uncertainty that sampling errors may be significant. To enbure representative sampling, an indium encapsulation technique originally developed for mass spectrometry ( 2 ) has been adapted for gas chromatography ( 7 ) . Indium encapsulation is faster and easier (3, 6 ) than glasb encapsulation ( I ) ; it avoids the flame sealing and the larger vaporization chamber usually required with glass encapsulation. In testing the technique, encapsulated samples were repeatedly weighed to determine weighing accuracy and the effectiveneqs of retaining volatiles during 4 weeks of storage. The rela-
tionship between weight and detector response wab established and applied to the gas chromatography analyses of encapsulated samples. Weighed samples were analyzed to determine the effectiveness of the approach in eliminating the need for a reference compound and retaining voltatile components. EXPERIMENTAL
The crimping device and a typical capsule are shown in Figure 1. The crimper consists of two jaws Lvelded to tweezers. The upper jaw is brass, the lower i3 iron. A hampie is encapsulated in the following manner. : I tared section is positioned against the aligning stop on the lower jaw of the crimper so that the ends of the tubing are under the two projections in the upper jaw. The magnet is placed on the tubing to hold the position qhown b y the dashed outline. Khile held on the crimper jaw.
Table I.
Time after encapsulation, days Start
Variations in Weight of Encapsulated 2-Methylpentane Samples over Four Weeks
Control 0 001 0 001 0 003
A 1 707 1 0 1 708 1 709 1 2 1 709 6 0 000 1 704 27 -0 010 - 0.001 1 ,708 Average 0,0017 U 0.0046 a Average t,are weight = 300 mg.
Weights of encapsulated samples, mg." B C D E 1 703 1 692 1 701 1 679 1 667 1 699 1 704 1 692 1 698 1 676 1 710 1 688 1 701 1 671 1 690 1 704 1 703 1 698 1 699 1 669 1.705 1.692 1.698 1.672 0.0030 0.0037 0.0055 0.0050
the tubing is touched t o the sample and filled by capillary action. The jaws of the crimper are pressed together to crimp and self-weld the ends of the tube to form the capsule. escess sample is wiped from the outside wall with soft tissue paper, and the filled capsule is weighed. As in conventional weighing practices, the capsules are handled with linen gloves to avoid contamination with dirt or grease. For the evaluation of the method, the capsules were made from 0.01-inch i.d. indium tubing cut into 5.3-cm. lengths with a razor blade. Care was taken to avoid distortion of the tubing during cutting and handling. The analyses were carried out in a gas chromatography unit equipped with a gas density detector and the vaporization chamber shown in Figure 2. T o minimize peak broadening, a small vaporization chamber of less than half the recommended value of 10 ml. ('7) was used. For high efficiency columns even smaller volume must be used. The chamber, of 3/le-inch i.d. stainless steel tubing, was connected with silicone tubing to a glass air lock that could be closed with clamps. Heating tape attached to a Variac was wrapped around the chamber to maintain its temperature a t 200' C. I n preparation for a typical analysis, clamp -4 was removed and the encapsulated sample was dropped into the air lock. Then clamp d was repltlqed and clamp B was removed to permit the capsule to drop into the heated vaporization chamber. As the indium melted, the vaporizing sample was swept from the chamber into the column by the carrier gas. After .everal samples had been analyzed, clamp C was removed and the melted indium which qettles to the bottom of
-INDIUM ,
I
r: Ol5CM
Figure 1 .
5 0 CM
-4
CAPSULE
(CRIMPED TUBE1
OlSCM
Indium tube crimper
-G
F 1 710 1 705
1 1 1 1 1
687 700
701 691 689 1.694 0.0065
1 708 1 703 1 698
1.705 0.0047
the chamber was removed with a glass capillary. For each analysis, the area under the chromatogram was related to the weight of the components chromatographed. With the gas density detector, this relationship is predictable from the following equation (6):
fl:lR
Average 1 697 1 696 1 697 1 696 1 694
LT
0.011
0.014 0 01R
0 oi3 0 012
LOCK
UMN
where It' is weight in milligrams, A is the area of the chromatographic peak in square centimeters, Jf is molecular weight, k is a conversion factor in milligrams per square centimeter, and the subscripts z and c refer to component and carrier gas, respectively. Capsules of 2-methylpentane were used to determine k , which is a constant for the gas density detector and applies to all components. T o determine whether the capsules were leak-tight, seven capsules of 2methylpentane were weighed five times over a 4-week period. Handling effects were studied by including a n empty capsule as a control. Known mistures of volatile components were encapsulated, and were analyzed immediately and 24 hours later, to determine whether the concentration of the sample would change with time.
).
Figure 2.
Table
Table I shows that the weight of the encapsulated 2-methyl pentane samples did not change significantly over four weeks. Small deviations arise from the weighing and handling of the indium capsule, but the data show that samples may be easily weighed to within about 0.5'%. The deviations between the weights of different samples are only two to three times larger and thus indicate that the encapsulated volumes are reproducible to within about 1%. Therefore, the option of introducing known volume or known weight of sample without danger of selective vaporization is feasible with this technique. Results in Table I1 are based on detector response in terms of weight and the weight of the mixture added to the column. Weight of the remainder
Sample introduction system
It. Analysis of Encapsulated Mixture
Wt., mg. Com~ponent Added Found Ether 0.260 0.271 2-Methylpentane 0 467 0.470 Hexene-1 0.467 0 464 Remainder of sample 0,936 0,925" Total 2.13 a By difference: 2.130 0.925. Table 111.
RESULTS A N D DISCUSSION
5CM
._._
Erior +0.011
+O 003 -0 003
-0.01 1 1.205 =
Stability of Encapsulated Mixture ___ Analyses _-
-
ImnieAfter Component diately 24 hours Isobutane 0 12 0 12 n-Butane 0 23 0 24 Isopen tam 16 82 17 37 n-Pentane 9 io 9 58 11 45 2,2-Dimethylhutane 11 66 2,3-Dimethylbutane 7 68 7 80 2-Methylpentane 21 40 21 46 3-Methylpentane 12 84 12 6 2 n-Hexane 14 35 14 16 Methylcyclopentane 1 26 1 28 Cyclohexane 3 90 3 04 Sample recovery 99 96577 100 OF,,
not chromatographed is the differcncc between weight added and responv of components chromatographed. The approach illustrates how accurate rcsults may be obtained on a w i g h t basis without recourse to adding a VOL. 36, NO. 8, JULY 1964
1687
reference compound or chromatographConsequently, ing to completion. saniplm such as naphthas in catalytic cycle oil and gasoline diluent in motor oil may be analyzed by this approach. Table 111 shows the results on successive days with encapsulated samples of a highly volat,ile mixture. Agreement b-tween the two analyses is within the experimental error to be expected in the measurement' of the peaks. 130th the individual determinations and the total recoveries show complete retention of all components. Wilkrns Instruments and Research,
Inc., has been licensed by Standard Oil Co. (Indiana) to manufacture commercial models of the indium tube introduction system. The commercial models are constructed of metal and can be operated a t 250' C. LITERATURE CITED
(1) Dimbat, M.,Porter, P. E., Stross, F. H., ANAL.CHEM.28, 290 (1956). (2) Ehrhardt, C. H., Grubb, H.. M., Moeller, W. H., L. S. Patent 3,103,277, Sept. 10, 1963. (3)rGrubb, H. M., Ehrhardt, C. H., 5ander Haar, R. W., Moeller, W. H., presented before ASTM Committee
E-14 on Mass Spectrometry, Los Angeles, California, May 1959. (4) KeulFmans, A., "Gas Chromatography, pp. 61-8, Reinhold, New York, 1957. (5) Meyerson, S.,Grubb, H. M.,I-ander Ham, R. W., J . Chem. Phys. 39, 1445 (1963). (6)h'erheim, A. G., ASAL. CHEM.35, 1640 (1963). ( 7 ) Sei-heim, A . G., U. S. Patent 3,063,286, Xov. 13, 1962. (8) Williams, A . F., Murray, R . T., Talanta 10,937 (1963). A. G. NERHEIM
Research and Development Department, American Oil Co. Whiting, Ind.
Infra red Determination of Hydroxyl Content of Epoxy Resins SIR: The reactive groups in epoxy resin are hydroxyl (-OH) and terminal 0 epoxide
/\
(-C-CH),
Inasniuch
as
H H hydrosyl is a functional group and inasmuch as the equivalents of hydroxyl per unit weight vary with molecular weight, a rapid, accurate, reproducible, standard method for determining hydroxyl should be available. Even though the resin was invented in the early 1940's no standard, mutuallyagreed-upon method is available. Classical methods of determining hydrosyl using acetic anhydride, acetyl chloride, or phthalic anhydride are well known (4)but these reagents are not applicable for determinat,ion of hydroxyl in epoxy resin because of the interference of the terminal epoxide group ( 5 ) . One approach that has been taken to eliniiriate epoxide interference is determination of reactive hydrogen and calculation of hytlroxyl since 1 mole of hJ-drogen ib equiralent to 1 hydroxyl. Steiirnark and Weiss ( 5 ) determined thc hydrosj~lcontent of epoxy resin by liberating the active hydrogen using lithium aluminum hydride as reagent. The amount of liberated hydrogen was determined by measuring its volume. 1Zartin and Jay (3) employed diborane as reagent to liberate the act'ive hydrogen and measured the amount liberated by pressure increase in the systeni. This laboratory used the diborane procedure of Martin and Jay. I t was concluded that the procedure was not applicable in o w laboratory for use by technicians because of the toxic, exploaire, and unstable properties of diborane and the meticulous attention to all details necessary to get a reliable answer. The lithium aluminum hydride procedure of Stenmark and Weiss offered no advantage. The near infrared procedure report'ed by Dannen1688
ANALYTICAL CHEMISTRY
berg ( 1 ) using the absorption band a t 1.456 microns was unsatisfactory because it was necessary to calibrate the instrument using either an internal standard or an epoxy resin of known hydroxyl content as determined by the lithium aluminum hydride procedure. This correspondent report's a procedure using conventional infrared equipment with NaCl optics. Hazards and terminal epoxide interference are eliminated and the determination can be performed by a technician. Nonpolar solvents with a window in the OH stretching region will not dis3olvt. cposy resin. Several of the solvent* used in the commercial application of eposy resin either absorb in the OH stretching region or contribute to intermolecular hydrogen bonding to some degree so these were not applicable. Pyridine was chosen as a solvent bccause it will dis;jolve all commercially available epoxy resins and because it possesses one other property of importance contributing to the success of the determination. The interaction between pyridine and alcohol molecules give an associated hydroxyl band exclusively according to the work of Valladas (6). Kabasakalian, Townley, and Yudis (2) employed pyridine as a solvent on 27 difficultly soluble hydroxyl containing compounds and found absorbance to be linear with concentration for the associated band which appears near 3.05 microns and is essentially independent of the type of hydroxyl group. Thus, pyridine as a solvent eliminates the complications of solubility i n d partial hydrogen bonding. This inforniat,ion also suggests a marked simplification of calibration compared to near infrared because it should be possible to use a high purity hydroxyl containing compound soluble in pyridine as a primary standard. From the data of Kabasakalian et al., it appeared glycerol was satisfactory for calibrating the instrument.
EXPERIMENTAL
Materials. The pyridine used was Eastman Kodak 1214 dried over KOH pellets before use. Shell Chemical Co. synthet'ic glycerol 99.9% was used as primary standard for calibration. The samples of epoxy resin tested had a weight per epoxide of 180-220 and a viscosity of 10,00016,000 centipoises a t 25' C. The resin was manufactured by the JonesDabney Co., Division of Devoe & Raynolde Co., Inc., and is marketed under the t'rade name of Epi Rez 510. Apparatus. All measurements were performed using a Llodel 21 PerkinElmer infrared spectrophotometer with NaCl opt'ics. The sodium chloride fixed cell had a sample path length of 0.07'o mm. Scans were made from 2.5 to 3.5 micronz in 1 minute with the instrument set at a resolution of 927, response 1, gain 5.8> suppiwsion 3, and a sodium chloride wafer in the reference beam. Procedure for Calibration. Two calibration curves are needed. For hydroxyl prepare five solutions of known concentration of glycerine in pyridine ranging from approximately 2.8 grams to 14 grams per liter. Fill cell with each solution and scan 2.5 to 3.5 microns. Read absorbance a t 3.08 microns. Fill cell with pyridine, scan, and read a t 3.08 microns to determine hackground. Plot net absorbance c s . grams OH for 10 ml. of solution. I n a similar fashion, prepare a calibration for water using five known solutions of water in pyridine
Table I. Hydroxyl Content in Equivalents Hydroxyl per 100 Grams Resin
Sample B-1339 B-1370 B-1434 B-1435 B-1529 B-1542
Infrared method Trial 1 Trial 2 0 048 0 047 0 052 0 052 0 041 0 043 0 041 0 044 0 045 0 043 0 050 0 048
Diborane method Trial 1 Trial 2 0 053 0 054 0 043 0 045 0 039 0 039 0 039 0 045 0 039 0 041 0 050 0 052