2.1 y per liter. The determination of iodate in sea water samples could be performed directly, thus minimizing the chances of iodine loss or contamination of the sample. Ferric and cupric ions in concentrations up to 280 y iron per liter and 320 y copper per liter ( 5 7 atoms per liter in each case) do not interfere. The determination of the total iodine content of sea water samples could also be done amperometrically, but the preliminary oxidation of the sample with bromine is time-consuming and introduces the possibility of contamination or iodine loss. However, the standard deviation of the total iodine determination mas found to be the same as that for the iodate determination, and the values obtained agreed well with those from the catalytic method. The principal contribution of the amperometric method is that it provides a rapid means for determining the form in which iodine occurs in sea mater samples. The catalytic method for the determination of iodine in sea water mas usable for minimum iodine concentrations of about 5 y per liter, with a standard deviation of 1.3 y per liter. The determination is direct, with no interferences present in sea water except the other halides, whose concentrations can be easily controlled. It is the most direct and rapid method for total inorganic iodine determinations in the range of concentrations encountered in sea water samples. The catalytic and the amperometric methods for iodate and total iodine were
checked for accuracy from time to time during determinations on sea water samples by adding internal standards to duplicates of samples which had already been analyzed. Iodate or iodide standards were added a t random but in fixed amounts which increased the concentrations by 40 y per liter. For the amperometric determinations, the error was less than two standard deviations on over 90% of the determinations; for the catalytic determinations the error was greater but within three standard deviaations for 90% of the samples. The increased error in the catalytic method was attributed to the marked increase in reaction rate, which made measurement difficult. Table I1 gives results obtained by the two methods in total iodine on samples of sea water from various depths a t two stations, one in the northeast Pacific Ocean and the other in an embayment north of Puget Sound. Table 111 shows data obtained on samples from the same stations analyzed for total chemically combined iodine and for the iodine in the form of iodate. LITERATURE CITED
(1) Anderson, R. C., Lasater, J. A., Lippman, D., J . Am. Chem. SOC.71, 2577-8 _ . - (1949). (2) Browning, P. B., Cutler, W. D., 2. anorg. Chem. 22, 303-5 (1900). (3) Cameron, A. T., Contribs. Can. Biol. ‘ and Fisheries 1, 75-80 (1922). (4) Dubravicic, bl., Analyst 80, 295-300 (1955). (5) Evans, Lacey H., Ph.D. dissertation, Chemistry Department, University of \ - -
~
-
I
Washington, Seattle, Wash., pp. 38-86, 1932, (6) Feigl, F., [[Chemistry of Specific, Selective, and Sensitive Reactions,” Academic Press, Xew York, 1949. (7) Grossman, A., Grossman, G. F., J .
Clzn. Endocrinol. and Metabolism 15,
354-61 (1955). (8) Hahn, F. L., Adler, V., Proc. Am. Sci. Congr., 7, 1942, 169-175, 8th Congr., Wash., D. C., Department of State, Washington, D. C. (9) Johannesson, J. K., Xature 182, 251 (1958). (10) Lang, R., 2. anal. u. allgem. Chem. 152,206 (1926). (11) Potter, E. C., White, J. F., J. A p p l . Chem. ( L o n d o n ) 7, 309-28 (1957). (12) Sandell, E. B., Kolthoff, I. M., J . Am. Chem. SOC.56, 1426 (1934). (13) Shaw, T. I., Cooper, L. H. N., Nature 180. 250 11957). 114’1 252‘(1958l. (14) Ibid.. Ibid., 182. 182, ~-~ 252 (1958). (15) Sugawara, K., Terada, Teraza, K., J . Earth Sca. Nagoya liniv. 5 , 81-102 (1957). (16) Sugawara, K., Ternda, K., Natzire 182, 250 (19%). (17) Sverdrup, H. U., Fleming, R. H., .Johnson. M. W.. “The Oceans.” Prentici-Hall, Kem’York, 1912. (18) Thompson, T. G., Karpi, E., J . Marine Research 5, 28-36 (1942). (19) Upson, U. L., ANAL. CHEJI. 25, 977-9 (1953). (20) Vinogradov, -4.P., “The Elementary of Marine Chemical Composition Organisms,” Memoir No. 11, p. 647, Sears Foundation for Marine Research, Yale University, Nev Haven, Conn., 1953. (21) Winkler, L. W.,Z. angew. Chem. 29, 205-7 (1916). RECEIVEDfor review July 21, 1959. Accepted October 30, 1959. Contribution No. 233 from the Department of Oceanography, University of Washington, Seattle, Wash. Research supported in part by the National Science Foundation and the Office of Naval Research, Contract 477( 10). \--,
~
I
~~
I
Quantitative Analysis of Acetic Acid-Acetic Anhydride Mixtures in the Near-Infrared Region A Statistical Study J. E. FERNANDEZ, RALPH T. McPHERSON, G. K. FINCH, and CARL
D. BOCKMAN
Tennessee Eastman Co., Kingsport, Tenn.
,This work was undertaken to develop a rapid and reliable spectrophotometric method for the determination of acetic acid in binary mixtures with acetic anhydride in the range 90 to 100% anhydride. The analysis is performed on the undiluted, untreated sample and requires approximately 6 minutes per sample. The effects of different analysts, instrument warmup time, and concentration of acetic anhydride on the method were also studied. This method has a lower variance (s2 = 0.026) than either the
158
0
ANALYTICAL CHEMISTRY
aniline (s2 = 0.1 6 ) or the triethylamine (s2 = 0.044) methods. Although the range studied was 90 to 100% acetic anhydride, the results at 90% anhydride did not fit in with the remainder of the data. The cause of this is unknown but suspected to b e an error in technique. The data are analyzed both with and without these ambiguous results. HE determination of acetic anT h y d r i d e by spectrophotometric methods has received very little atten-
tion (3). However, producers of large quantities of this chemical are interested in a fast, reliable, routine analysis for acetic acid in binary mixtures with anhydride. One of the established methods now in use is the aniline method, which involves weighing and titrating two samples (7‘). A second method involves titrating the acetic acid -with triethylamine (6). The nonacid material is assumed t o be all anhydride. The method described here is much easier and faster to carry out and is accurate and precise.
I
"I
90
92
96
9J
Acetic Anhydride Concn
100
98 Oo
Figure 2. Calibration curve of transmittance vs. per cent acetic anhydride
r,)
ACETIC ANHYDRIDE C O N C E N T R A T I O N
Wove length, M
Figure 1.
Spectra of acetic anhydride and acetic acid WARM-UPAHALYST B
I n the industrial laboratory, routine analyses involve considerable expense. To minimize this expense and to obtain the most accurate :tnd reliable results are frequently the objects of considerable study. A statistical analysis such as the following usually affords a highly desirable approach. An excellent discussion of one method of evaluating accuracy and precision is given by Linnig et al. ( 5 ) . The present work involves primarily x determination of the effects of various factors on accuracy. Although the analysis described here is for acetic acid, the results are given throughout this papc'r in per cent acetic anhydride. This does not affect the precision and is more consistent with Tennessee Eastman practices of reporting analyses of this sort in terms of the major constituent. ANALYTICAL METHOD
Reagents.
The acetic anhydride and acetic acid used in these experiments were commercial grades produced a t Tennessee Eastman Co. and were used without further treatment. The anhydride assayed 99.68% by the standard aniline method (average of five analyses), The acid contained 0.03% water by the temperature rise method. These values were used t o correct all calculations involving per cent acetic anhydride, Procedure. This analytical procedure is based on the spectrophotometric determination of acetic acid concentration in acetic anhydride solution. Figure 1 shows the superimposed spectra of acetic anhydride and acetic acid. A Beckman DK-2 recording spectrophotometer mas used in the 1600- to IGO-mW region ( 4 ) . The wave length a t which measurements were made was 1505 mp, corresponding to an absorbance minimum for acetic anhydride, The band a t this wave length in the acetic acid spectrum is undoubtedly due to intermolecularly
bonded hydroxyl groups, within an acetic acid dimer or a more highly associated species. The settings were as follows: scanning time = 5 minutes, scale expansion = 2 X , 7 0 T, time constant = 0.1, sensitivity = 0.020, and range = 0 to l O O ~ , T. Before each sample was run, the clean, dry, Corex 1.0-cni. cells ivere placed in the sample and reference compartments and the instrument mas set a t 0 and about 99% T a t 1600 mp. The range 1600 to 1450 mp was then scanned to obtain the background against which the sample was to be compared. The acetic anhydride sample \vas placed in the sample cell and the same region scanned using the same spectrophotometer settings. The next sample was analyzed only after rinsing the cell four times with carbon tetrachloride, drying with compressed air dried with calcium chloride, and then repeating the 99% T background scanning. The average time required to analyze one sample including the calculation of results was 6 minutes. Blanks on unfilled cells are not usuallv proper corrections for liquid samples in the same cells. However, samples and standards are run in the same manner and it is assumed that errors in cell correction are very small. This simplification in procedure was found desirable, especially when a large number of samples are run routinely and time becomes an important factor. Calibration Curve. Samples of acetic anhydride containing known amounts of acetic acid were prepared by measuring the calculated amounts of each component directly into the sample cell from automatic-filling burets protected from atmospheric moisture by drying tubes. These samples mere run exactly as described above. Five separately prepared samples a t each concentration were run and the concentrations plotted against average absorbance and transmittance a t 1505 mp. The Beer's lam equation was found not to be linear; therefore, for simplicity, the transmittance us.
+HOUR
ANALYST
y
y
y
x
x
x
y
x
x
x
WARMYPANILYST
anhydride concentration plot was used as calibration curve. This calibration curve is shown in Figure 2. By assuming that the analysis of acetic anhydride a t 99.68% by the aniline method was correct, the concentration of acetic acid in the standard anhydride was computed and corrections were made in the calibration curve. Calculations. The transmittance for each sample mas calculated from t h e spectrogram by correction for 1007c T as follows: % T (sample) x 100 % ' T (corrected) = yo T (air-filled cells) The per cent acetic anhydride was then read from the corrected calibration curve using the above per cent T .
Statistical Design. This experiment v a s designed to determine the effect of acetic anhydride concentration, instrument warm-up time, and different analysts on the accuracy of the method as well as t o obtain the precision. A further purpose was t o eliminate the possibility of changes in anhydride concentration with time due t o hydrolysis by atmospheric moisture. The design is shown in Figure 3. Each treatment combination was replicated and the entire 40 analyses were randomized. Each sample was prepared in the cuvette just before analysis by measuring the calculated amounts of each reagent from burets as described above. The reagents used were acetic anhydride of 99.687, purity and a solution containing 80% anhydride and 20% acetic acid, Analyst A prepared the samples VOL. 32, NO. 2, FEBRUARY 1960
159
Table 1.
Determination of Per Cent Acetic Anhydride
90 %
Analyst 2-hour warm-up
Totals
iz
B
h
B
- 0.95
x
0.50 0.40 0.90 0.27 -0.41 -0.14 0.76
2 40 2.70 5.10 2.45 2.83 5.28 10.38
2.98 2.93 5.91 2.95 2.95 5.90 11.81
5 80 5.77 11.57 6.12 5.65 11.77 23.34
5.75 5 57 11.3% 5.50 5.85 11.35 22.67
6.99 7.38 14.37 7.20 7.27 14.47 28.84
7 22 7.25 14.47 7.36 7.17 14.53 29.00
9.60 9.47 19.07 9.65 9.68 19.33 38.40
II.
Analysis of Variance
per cent acetic anhydride -90) Degrees Mean Symbol of Freedom Squares 0.1102 -4 1
Source Analyst Concentration Warm-up time Interactions
=
C T AC AT
4
1 4 4 20 39
ACT
-
Experimental error Toial Significant at 99% confidence level. * Significant at 7570 confidence level.
T w o - w a y Table Showing AT Interaction
per cent acetic anhydride -an\ "-,
2-Hour
8-Hour
4 97 5.20 5.08
5.11 5.09 5.10
warm-cp 1varm-up&lean Analyst A Analyst B Mean
112.3375 0.0026 0 0806 0 1690 0 0061 0 0639 0 0869
4 1
CT
=
5.04
5.14
Table IV.
F Ratio
=
1.2i 1292a 0.03 0.93 1 94b 0 07 0 74 s2, s = 0 3
to be analyzed by Analyst E. and B prerared the sani1:les to be analyzed by A in an attempt to remove personal bias. Each analyst anal!-zed tn o saiiiples at 1O:OO A.M. and tn-o a t 4:OO P.L e v e n other day. I n this way the e \ ; p m i e n t 01 as extended over a period of 2 n-erks. The analysis of variance was niade by the sums of squares method ( 1 ) . The per cent acetic anhydride found n a s ~ used as the response. I n this c : the
Comparison of Variances
Per Cent Anhydride 90 0 5040 0 1032 0 3036
2 hours 8 hours
Mean
93 0 0701 0 0561 0 0631
96 0 0108 0 0719 0 0413
97.5 0 0263 0 0071 0 0167
Mean
100 0 0145 0 0011 0 0078
0 1251 0 0478
F ratio = 0'1251 - 2,63a ~
a
0.0478 Significant a t 9570 confidence level.
Table V.
Analysis of Variance
(Per cent anhydride = 93 to 100) Sum of Degrees of Mean Squares Freedom Squares Crude Total bo
Time ( T ) Analyst ( A ) Concentration (C)
AT CT AC A X C X T (by diff.)
1487.0310 1288.7964 0.0128 0.0420 197.4465 0.0221 0,0014 0.2801 0.0105 0.4192
Error 5 Significant a t 95y0 confidence level.
160
100%
B
(Response
(Response
97.51,
A
Table
Table 111.
96%
B
+0.50 -0.45 0.27 0 0.27 -0.18
8-hour warm-up
93 7 0
h
ANALYTICAL CHEMISTRY
32 1 1 1 3 1 3 3 3 16
0.0128 0.0420 65.8155 0.0221 0.0005 0.0934 0.0035 0.0262
F
Ratio
1.60
3 , 5ga s
=
0.16
B 9.76 9.65 19.41 9.60 9.63 19.23 38.64
Totals 101.67
101.99 203.66
response was coded by subtracting 90 from each observation. This has no effect on the mean squares and niakes calculation simFler. DISCUSSION OF RESULTS
The data obtained in this PSI eriinent are tabulated in Table I. These values are per cent acetic anhydride minus 90. The analysis of variance is shown in Table 11. The highly significant effect of concentration is expected because the concentrations were made up significantly different. This effect, however, has been removed from all other effects by variance analysis. Although the effects of A and T seem to be nonsignificant as shown by the F ratios, consideration of the possibly significant effect of the AT interaction could cast doubt on this conclusion. I n this particular example the interaction is significant only a t the 75% confidence level. However, Table I11 shons that the effects of d and T are more important than the F ratios indicate and that a t 2-hour warm-up time there appears t o be a difference between analysts. This demonstrates the necessity of considering all the data together and points out that a n effect may be important and significant, but may be concealed in the analysis of variance table by a large interaction. Because the difference between analysts a t 8 hours is essentially zero, the difference a t 2 hours is probably a measure of the reproducibility of the instrument with insufficient warm-up time rather than a difference between analysts. A comparison of variances for 2-hour n arm-up times vs. 8-hour narm-up times shon s that this is correct and that a 2-hour n arm-up time gives a significantly larger variance than a n 8-hour n arm-up time. Therefore, even though the average d u e (bias or accuracy) does not vary significantly n i t h warm-up timr, the precision appears better with the 8-hour n:iim-up time, These varianccs were computed for each concentration and 71 arm-up time and are shon n in Table IT'. Examination of the data in Table IV also reveals an apparently large difference betn-een the variance obtained on 90% concentration samples and the
variances at higher concentrations. One of the basic assumptions inherent in analysis of variance is t h a t the variance is the same throughout the experimental region. Cochran’s test for homogeneity ( 2 ) was therefore run with the results shown below: Largest s2 ZS2
=
bo
T w o - w a y Table Showing AC Interaction
rlnalyst
93
Per Cent Anhydride_ _ 96 97.5 100
10 38 23.34 28 8-1: 38 40 11 81 22 67 29 00 38 64
A B
0.3036 = 2.35 0.1289
This ratio is highly significant and these variances are found not t o be homogeneous. Bartlett’s test can also be used (1). Examination of the data in Table IT indicates that the lack of homogeneity is caused by the variance of the 90% concentration samples. Elimination of these samples leaves four variances t h a t are homogeneous. Indcpendent n ork in these laboratories indicates t h a t the variance ib essentially constant in the range 50 to 100% acetic anhydride concentration. Therefore, the large variance of the 90% anhydride samples may be due t o faulty techniquc or to a n error of the alpha tyl;e. The analysis of vsiriance omitting the 90% concentration samples is presented in Table V. A comparison of this table with Table I1 shows t h a t two m t a n squares have changed appreciably. The mean square or variance for error is much smaller, and a n analystconcentration interaction is nom significant. This interaction, the cause of n hich is difficult t o exlilain, is shoivn in Table VI. Furthermore, the over-all variances for 2- and 8-hour narm-up times are now 0.0304 and 0.0340. respectively. These are now not significantly different. A further point of interest is whether the quantity, observed anhydride concentration minus calculated anhydride concentration, varies with concentration. This was studied by making a linear regression analysis for the 2arid 8-hour warm-up times of the observed-calculated per cent anhydride ovtr the range 93 t o 1007,. The slope of the regression line and the regression analysis art. given in Table VII. Table T’II s h o w that the linear model is adequate and that for either warm-up time this response does not dclpend on concentration in the range 93 t o 100% acetic a n h j dride. Another analysis \vas performed t o drtermine if a change in the instrument calibration had occurrpd with time. A regression analysis n as performed using the model: Y
Table VI.
+ btxt + btt~’r
where the b’s = cocfficients to be determined, z L = time since calibration, zZt = time squared, and y = obscrred anhydride - calculated anhydride. The data are given in Table T’III and
Table VII.
Tennessee Eastman Co. t o conil~arc~ three methods of ana1:sis. Table S shows the results of these experiments and affords a comparison of the precision of these methods. As a n example of the use of s in the analysis of anhydride by the spectrophotometric method, the confidence
Regression Analysis
(Observed - calculated per cent anhydride us. concentration) 8-Hour Warm-Uo ?-Hour Warm-Uo Degrees Degrees of free- Sum of Mean of free- Sum of Mean dom squares squares Source of Variance doni squares squares Total 16 0.3920 16 0.4242 1 0.0000 Due to mean 1 0.0183 1 0.0001 0.0001 Due to regression 1 0.0069 0.0069 12 0.4090 0.0341 Error 12 0.3650 0.0304 2 0.0151 Residual (lack of fit) 2 0 0018 0.0009 0 0075 Table VIII.
Days, Found - from Calculated Calculated Zero 1 89.73 -0.68 95.77 0.03 1 ‘7 92.75 -0.30 2 97.24 -0.04 95.77 -0.20 97.24 -0.02 89.73 -0.14 95.77 -0.27 4 92.75 -0.35 4 92.75 -0.05 4 95.77 -0.12 4 97.24 0.03 97.24 0.01 8 99.68 0.08 8 92.75 0.20 8 97.24 -0.07 8 95.77 0.00 9 97.57 0.58 9 9 92.92 -0.09 99.68 0.00 9
7%-4~20, Found 89.05 95,80 92.45 97.20 95.57 97,22 89,59 95,50 92.40 92.70 95,65 97.27 97 . . 2.5 ~-
99.76 92.95 97.17 95.77 96.99 92.83 99.68
a
r0 AczO,
Original Data
Da?-s,
yo A4ce0, % AceO, Found - from Found 92.98 92.93 92.95 95.85 90.00 96 12
90 99 90 99 99 90 90 99 97 99 97 99 90 95
50 47 27 63
63
27 50 65 38 GO
36 60 40 75
Calculated 9’7.75 92.75 92.75 95.77 89.73 95 77 89 73 99 68 89 73 99 68 99 68 89 73 89 73 99 68 97 24 99 68 97 24 99 68 89 7‘3 95 77
Calculated Zero 10 0.23 10 0.18 10 0 20 10 0.08 11 0.27 11 0 35 15 0 77 15 -0 21 15 0 54 15 -0 05 16 -0 OX 16 0 54 16 0 77 16 -0 03 17 0 14 17 -0 08 17 0 12 17 -0 08 18 0 67
Table IX. Regression Analysis (Observed - calculated per cent anhydride us. time) Degrees of Sum of Mean Total Freedom Squares Squares Total 32 0.8162 bo 1 0.003 bl 1 0.0007 btt 1 0 1222 Residual 29 0.6930 Error 20 0.3624 0.0181 9 0.3306 0.0367 Lack of fit Significant a t 9594 confidence level. Significant a t 90r0 confidence level
are treated by the Forn-ard-Doolittlc method. The regression analysis table is shown in Table IX. A study of the regression analysis indicates that a second-order time trend exists. This trend, although significant, is not large enough t o necessitate correcting the calibration curve during this time. Designed experiments independent of the prpsent work have been performed on plant samples of anhydride a t
18
-0 02
F Ratio
6 76. 2.03b
Table X. Comparison of Three Methods of Acetic Anhydride Analysis
Method
Degrees of Freedom
A*
S
Aniline Triethylamine
192 114
0.16 0.044
68
0.014
0 40 0.21 l2
Sear-infrared
VOL. 32, NO. 2, FEBRUARY 1960
161
limits for duplicate samples a t the 99% confidence level are found by the formulaz i -% to b e 3 =t0.23. N is
d3
the number of samples = 2, t is obtained from the t distribution tables a t the 99% confidence level and at 68 degrees of freedom, and s = 0.12. CONCLUSIONS
I n the range 93 t o 100% anhydride, neither the accuracy nor the precision of the method is dependent on the anhydride concentration. Also, in this range, the over-all standard deviation for the method is 0.16 with 16 degrees
of freedom and there is no significant difference in results between 2- or 8hour warm-up times. The instrument calibration was found not to change linearly over a 3-week period. The precisions of the standard aniline, triethylamine, and spectrophotometric methods are compared and the spectrophotometric method with a standard deviation of 0.16 is found to be more precise than the other two. LITERATURE CITED
(1) Davies, 0. L., “Design and Analysis of Industrial Experiments,” Chap. 7 ,
Hafner, New York, 1956.
(2) Dixon, W. J., Massey, F. J., “Intro-
duction to Statistical Analysis,” McGraw-Hill, New York, 1951. (3) Goddu, R. F., Le Blanc, N. F., Wright, C. M., ANAL.CHEM.27, 1251 (1955). (4) Holman, R. T., Edmondson, P. R., Zbid.. 28. 1533 f 1956). (5) Lihig: F. J:, Mandel, J., Peterson, J. M., Zbid., 1102 (1954). (6) McClure, J. H., Roder, T. M., Kinsey, R. H., Zbid., 27, 1599 (1955). (71 Nicolas. L.. Burel, R., Chim. anal. 33,
RECEIVEDfor review March 17, 1959. Accepted November 16, 1959. Work presented a t the Southeastern Regional Meeting, ACS, Gainesville, Fla., December 1958.
Direct Amino Acid Analysis by Gas Chromatography ALBERT ZLATKIS, JOHN F. ORO, and A. P. KIMBALL Deportment of Chemistry, University of Houston, Houston, l e x .
b Mixtures of amino acids that yield volatile aldehydes can b e rapidly analyzed by injecting a small sample of an aqueous solution of amino acids into a specially designed reactor-gas chromatographic unit. The amino acids are oxidized to aldehydes in a microreactor which is a part of a continuous flow system. The aldehydes a r e chromatographically separated, catalytically cracked, and analyzed in a thermal conductivity cell. The complete analysis of a mixture of seven amino acids can b e performed in less than 1 hour.
T
utilization of procedures whereby amino acids may be converted to volatile compounds offers a convenient method for their analysis by gas-liquid chromatography (14). Hunter, Dimick, and Corse (8) converted leucine, isoleucine, and valine into the corresponding aldehydes by oxidation with ninhydrin. These were collected in cold traps and then sampled for introduction into the gas chromatographic apparatus in a closed system. Bayer, Reuther, and Born (I, 2) esterified several amino acids with methanolic hydrochloric acid, extracted the esters with ether, and then chromatographed the ether extract. Similar techniques based on the chromatography of aldehydes, amines, hydroxy acid esters, and N-acylated amino acid esters have been reported more recently by Bier and Teitelbaum (3, 4,Liberti (8), and Youngs (12). The approach presented in this paper makes use of a special reactor-gas HE
162 *
ANALYTICAL CHEMISTRY
chroniatographic unit. The method is simple and rapid because it requires one single operation. An amino acid solution is injected into a continuously flowing system where oxidation, separation, reduction, drying, and analysis take place in a relatively short time. A heated niicrorcactor containing ninhydrin effects the oxidation of the amino acids to aldehydes. Separation of the latter takes place in a chromatographic column. After leaving the column, the aldehydes are cracked to methane and water over a nickel catalyst. The water is then removed by a drying column and the analysis is made by a thermal conductivity cell. This procedure has been applied to mixtures of amino acids, and in general it may be used for any amino acid which can be oxidized to a volatile aldehyde. APPARATUS AND REAGENTS
A schematic diagram of the system is shown in Figure 1. Reactor A consists of a 6 X l/4 inch glass U-tube containing 30% ninhydrin (Nutritional Biochemicals Corp.) on C-22 firebrick (Johns-Manville Co.) , a graded diatomaceous earth. The tube was heated with a heating tape and the temperature kept a t 140’ C. The chromatographic column, 10 foot X inch copper tubing, contains 30- t o 60-mesh aqua regia-treated C-22 firebrick ( I S ) coated with 10% of an equal mixture of ethylene and propylene carbonates. Reactor B was a 12 X ’/4 inch glass U-tube containing a nickel-kieselguhr catalyst of 30 to 60 mesh. The drying column was a 1 foot X inch glass U-tube filled with a Molecular Sieve of 10 t o 30 mesh.
This reactor-chromatographic train was tied into a Perkin-Elmer Vapor Fractometer, Llodel 154B, through its gas-handling system. PROCEDURE
d ninhydrin-amino acid solution was prepared by mixing 1 part of a saturated aqueous solution of ninhydrin with 1 part of a 0.28Jf solution of amino acids and kept in an ice bath until ready for sampling. Fifteen microliters of this solution were injected into reactor A with a Beckman liquid sampler. Reactors A and B were at 140’ and 4 2 5 O C., respectively; the detector cell, chromatographic column, and drying column were at 25’ C. Hydrogen mas used as the carrier gas at a flow rate of 100 ml. per minute. The analysis of amino acids by the reactor-gas chromatographic technique takes place in a flowing stream through a series of steps.
Oxidation of Amino Acids. The amino acids are oxidized with ninhydrin t o volatile aldehydes and carbon dioxide. T o ensure a complete reaction, a n excess of ninhydrin is used by keeping the reagent in the injecting solution and having the firebrick coated with 30% ninhydrin. A temperature of 140’ C. in reactor A ensures a n almost instantaneous oxidation. Lower temperatures generally result in incomplete reactions. The ninhydrin-firebrick material in reactor A is effective for about five sample injections. Mixtures of amino acids such as alanine, a-amino-*butyric acid, valine, norvaline, leucine, isoleucine, and nor-