V O L U M E 2 7 , NO. 5, M A Y 1 9 5 5 in the gravities of the salt diluents may be tolerat'ed without appreciably affecting the results, as the volume of salt contributes little to the total volume of the mull.
D. 11. Lewis of the Sabine River, Orange, Tex., plant has found the following alternative met.hod satisfactory in analyzing adipic-glutaric acid mixtures. .4 sample containing about 2 to 4 grams of adipic acid is weighed into a 100-ml. volumetric flask, and neutralized to the phenolphthalein end point with 25% sodium hydroxide solution, and the contents of the fla.sk are diluted to volume with water. After thorough mixing, a 10-ml. aliquot (pipet) of the solution is added, with vigorous stirring, t,o 250 ml. of acetone in a 600-ml. beaker. The precipitated disodium salts are filtered onto a tared, sintered-glass filter and dried to constant weight at, 110" C. in a vacuum oven. Esactly 0.200 gram of the dried salts is finely pulverized in a mortar, then mulled with 2.00 mi. of white mineral oil. A portion of the mull is transferred to a fixed 0.2-mm. disodium chloride cell. A second sodium chloride cell (about 0.05 mm. less in t'hickness) is filled wit,h the white mineral oil. Both cells are placed in t h e Perkin-Elmer Model 21 infrared spectrophotometer (dpuble-beam) and the sample is scanned between 10.4 and 11.5 niicrons. The absorbance a t 11.03 microns is determined by the stmdard base line technique.
741
The concentration of disodium adipate is determined by comparison with the absorbance of a mull of a weighed amount of di-odium adipate or by a calibration curve. LITERATURE CITED
(1) Badger. R. SI,,and Bauer, S. H., J . Chern. P h y s . , 5, 605-8 (1937). (2) I h i d . , pp. 369-70. (3) Davies, SI. SI., a n d S u t h e r l a n d , G . 13. €3. SI.. I b i d . , 6 , 755-70 (1938). (4) D u v a l , C.. Lecomte, J., and Douvilli., F.,Ann. p h y s . , 17, 5-71 (1942). ( 5 ) D u v a l , C., Lecomte. J.. and Doui-illi5. V., Bull. soc. chim.. 9, 263-74 (1942). (6) Duval. C., Lecomte, J . , and Douvilli.,. F., C o m p t . rend., 212, 95Z-6 (1941). ( 7 ) Fiuhter, K.,and Wehrli, SI., H e h . P l i y s . Acta. 13, 217-18 (1940). (8) Schieldt, U., 2. AYaturforscli., 76, 270 (1952). (9) Stimson, 11.hI., J . A m . Cltrru. S O C . . 74, 1805 (1952). (10) Wehrli, >I., a n d Schonmann. E.. Heli.. Phus. Acta. 15, 317-19 (1942). R E C E I V E Dfor review September 8 . 1954. Acceoted December 2 i . 1954.
Automatic Spectrophotometric Titration with Coulometrically 6enerated Titanous Ion Determination of Vanadium in Titanium Tetrachloride H. V.
MALMSTADT and C. B. ROBERTS
Noyes Chemistry Laboratories, University o f Illinois, Urbana,
This investigation was undertaken to find a simple method of sample preparation and of determining vanadium in titanium tetrachloride which could be carried out rapidly to & \ e precise and accurate results. Titanium tetrachloride was h>drolyzed without loss of kanadium by pipetting a sample into a hydrochloric acid solution in a flask fitted with a reflux condenser and water trap. The vanadium in the hydrolyzed titanium sample was determined directl) by a rapid and accurate titrimetric procedure using automatic spectrophotometric end point detection and coulometric generation of titanous ion. Ten-milliliter samples containing about 0.01 to 0.2% of vanadium in titanium tetrachloride were determined with an average deviation of about 0.00016~0. The results indicate the possihle general usefulness of roulometric generation of titrant for automatic spectrophotometric titrations, and the specific application of the method for the rapid and accurate titrimetric determination of vanadium in titanium tetrachloride.
1
THE importance of titanium tetrachloride as an intermediary
in the manufacture of titanium sponge has led to its analysis foi impurities. Vanadium iq a principal impurity in titanium tetiachloride, and it is important to have a rapid and accurate method for its determination. d method of sample preparation wa; first developed so that the titanium tetrachloride sample could be rapidly hydrolysed without loss of sample. The hydrolp i * was performed by adding the sample t o a hydrochloric acid solution in a flask immersed in an ice bath and fitted with a reflux condenser topped with a water trap. The resulting solution of titanium and its impuritig.s was about 7.5111 in hydrochloric acid, and contained the vanadium as both vanadium(1V) and vanadium(V). Goddu and Hume ( 2 ) developed a good photometric titration
111.
procedure for the deterniinatioii of' vatindium i n steel, \ ) u t the permanganate titration was not directly applicable to the himediate titration of the strong h>-drochloric acid Rolutions of hydrolyzed titanium tetrachloride. Arthur and Donahue ( 1 ) used coulometrically genemted t,itanoup ion for the titrimetric determination of iron, and it appeared that this procedure would tie espedient for the determination of vanadium in titanium tetrachloride because the hydrolyzed sample might he directly analyzed without external addition of reagents. Since the titsniuni tetrachloride and its hydrol!-ned solution contained the vanadium as both vanadium(V) and van:dium(IV), it waa neceassry either to oxidize all the variadiuni to vanadium(\'), xhich did not prove successful iri strong hydivchloric acid, or t o determine the amount of titanous ion used in going from the first end point [vanadium(T-) t o vanadiuni(IT')] to the second end point [vanadium(IV) to ~ a n a d i u m ( I I I ) ] . Potentiomet,rir determination (3)of the end point,s in the hydrochloric*wid solution was not too successful bec:tusr the second end point \vas not sharp and was impossihle t o deterinine accurately. .in investigation of spectrophototlietrio det,ection of the end point3 yielded precise and arcburate results for the det crmination of vanadium in titanium tetrachloride and indicated the possible general usefulness of coulometi~icgeiiwttion of titrant for automatic spectrophotoinetric titrations. Several excellent spectrophotometi~ictitration procedures have been developed within the past few years. Sweetser and Bricker (10, 11)used the Beckinan Models DU and B for manual spectrophotometric titrations in both the visible and ultraviolet spectral regions. Reilley and Schweizer (6) have cvtended its use to nonaqueous solvents, and Underwood (18, I S ) h a p used a similar procedure to determine iron, copper, and hieinuth. Malmstadt and Gohrbandt ( 4 ) used a Cary spectrophotometer for automatic spectrophotometric titrations to determine thorium, In all of these procedures the titrant was a standard solution which was added hy a buret. In the present work no standard Polution
742
ANALYTICAL CHEMISTRY
was used; titanous ion was generated electrolytically, the number of equivalents added being calculated from measurements of current and time. Wise, Gilles, and Reynolds (14) developed an automatic coulometric titration instrument which furnished constant current and automatically stopped the titration by means of photometric detection of the end point but their method depended on a large color change a t the end point and did not give a continuous recording of color change. The procedure described herein has the advantage of permitting extrapolation to an end point when there is only a slight change of absorbance a t the equivalence point because of dilute solutions or the nature of the color change itself. Because the Cary spectrophotometer can automatically record time versus absorbance, it is necessary to have only a suitable titration cell, a motordriven stirrer, a constant current source, and a means by which to measure the current. By using Faraday’s law, the number of equivalents of titrant added can be calculated a t any time. Thus a recorded plot of absorbance us. equivalents is easily obtained. This method has all the advantages which have been mentioned in previous literature on spectrophotometric titrations ( 2 , 4, 6). In addition, it has no dilution error, as the volume remains constant during the titration. The same precautions for coulometric titrations, such as 100% electrode efficiency for the titrant and rate of attaining equilibrium, must be observed. Arthur and Donahue ( 1 ) used a gold-plated platinum electrode in approximately 0.7M hydrochloric acid as the generator cathode for their titration of iron with titanous ion. They experienced difficulty in cleaning and maintaining their electrode so that it would not evolve hydrogen. In the present investigation it was possible to use a titanium metal rod as the cathode in about 1 to 1ZAVhydrochloric acid solutions with 100% current efficiency in generating titanous ion and no significant attack on the titanium metal was observed. Platinum was also used as the cathode d t h 100% efficiency for generation of titanous ion in hydrochloric acid solutions greater than 7 M , but hydrogen was evolved at lower acid concentrations. In all cases with both electrodes the current did not exceed 100 ma. Samples containing about 0.01 to 0.2% vanadium in titanium tetrachloride were determined with an average deviation of about 0.00016%, The possible interference of iron in the redox titration of vanadium by titanous ion was studied and it was found possible to determine the two simultaneously. APPARATUS
The Cary spectrophotometer titration vessel was essentially the aame as used by Malmstadt and Gohrbandt ( 4 ) in their determination of thorium. The only changes made were in the height of the cell and the cover for the cell compartment. The cell was increased in height until it touched the bottom of the cover through which a 2-inch hole was cut. A No. 11 stopper was fitted into this so that it formed a tight seal with both cover and cell. I n this way hydrochloric acid fumes were kept from the inside of the instrument and light was excluded from the cell compartment. Three holes were bored in the stopper to permit passage of the stirrer, cathode, and anode. Coulometric Electrodes. Two difTerent cathodes were used, and both gave the same results in strong hydrochloric acid solutions. One cathode consisted of a piece of 75A titanium rod, 4 mm. in diameter and 13 cm. long. The rod extended into the titration cell so that approximate1 0.5 inch was in contact with the solution. The other cathoak was a square platinum foil connected to a platinum wire sealed in a glass tube. The anode was a piece of platinum foil 1.5 by 1.5 cm. It was isolated from the solution that was being titrated by inserting it into a glass tube 1.0 cm. in diameter and 10 cm. long, with a medium porosit sintered-glass disk sealed a t the bottom end. The tube was filed to a height of about 1 inch with 1N sulfuric acid, and the platinum electrode was immersed in this. The part of the tube extending above the stopper was painted black in order to exclude light from the titration cell compartment. Stirrer. The stirrer was the same as that used by Malmstadt and Gohrbandt ( 4 ) . I t passed through the center of the stopper with the coulometric electrodes on either side.
Constant Current Source. Constant current was provided by a circuit similar to that described by Reilley, Cooke, and Furman (6). External Circuit. The external circuit was the same as that shown bv Reillev and coworkers (6) for macrothation& except that a 110-volt, double-pole, double-throw relay operated in place of the double-pole, double-throw shorting type switch. The relay and a Time It stop clock were connected through a singlepole, single-throw switch to the 110-line voltage, and a dummy resistor was used when the cell was not in operation. The voltage drop across a 3.513-ohm standard resistor waa measured by a Leeds and Northrup potentiometer. Apparatus for Hydrolysis of Titanium Tetrachloride. A 24-inch water-cooled reflux condenser was fitted with a rubber stopper so that it could be easily attached and removed from one of the side holes of Figure 1. T o p a 500-ml., three-hole, round-bottomed of condenser flask. A solid rubber stopper was placed and water trap in the larger center hole, and the other for hydrolysis side hole was fitted with a I-hole, No. 3 stopper through which passed a 10-ml. A . 60-ml. seuaratory pipet. A 2-hole rubber stopper was placed B. Side arm in the toD of the reflux condenser (see C. Top of reflux Figure 1); and a 60-ml., straight-sided, condenser borosilicate glass separatory funnel and B side arm of 6-mm. diameter glass tubing were fitted into the stopper. The loop in the side arm was made approximately the same height as the top of the separatory funnel. In this way all volatile materials which did not condense in the reflux condenser bubbled through the water in the separatory funnel. The water was used to wash the condenser and bring the reaction mixture in the flask to the correct acid concentration for the titration. PROCEDURE AND EXPERIMENTAL RESULTS
Preparation of Sample. Samples hydrolyzed in the open did not give consistent results for vanadium from one sample to another. Therefore. the following method was used to prepare samples so that all the vanadium was retained and gave a correct acid concentration. Approximately 60 ml. of concentrated hydrochloric acid (12M) were placed in the round-bottomed flask, the flask was attached to the condenser, and 40 ml. of distilled water were placed in the separatory funnel. Ice was placed in a pan under the flask to remove some of the heat of hydrolysis. A 10-ml. sample of titanium tetrachloride was pipetted slowly into the flask through one side arm. The stopper on the pipet was immediately put in place to prevent the loss of volatile material. Most of the hydrogen chloride and volatile impurities were either condensed in the reflux condenser or dissolved in the water in the water trap, although a small amount of material in the form of white fumes passed through the water. The fumes were not determined because the accuracy and reproducibility of the results showed that no vanadium passed through the combination of reflux condenser and Kater trap. After all the titanium tetrachloride was added from the pipet, the finger was held tightly over the end of the pipet and the stopcock of the separatory funnel was opened momentarily. Tl ; presence of water in the condenser caused a partial vacuum by dissolving hydrogen chloride gas. This caused a rapid removal of water from the separatory funnel through the side arm and mashed the condenser free of all condensate. Simultaneously with this operation, the finger was removed from the pipet and all of its remaining contents were drawn into the reaction flask. The stopcock on the separatory funnel was then opened and the remaining water was allowed to drain into the reaction flask. The solution was transferred directly to the titration vessel, and the reaction flask was washed with 10 ml. of concentrated hydrochloric acid which were added to the solution in the titration vessel. A sample prepared in this manner mas brown in color and contained a small amount of chlorine and oxygen which gave high results if not removed. This was doni by covering the titration cell with a Teflon shield to prevent possible loss of material, and bubbling tank carbon dioxide through the solution vigorously for about 10 minutes.
V O L U M E 27, NO. 5, M A Y 1 9 5 5
743
Spectrophotometric End Point Detection. The reduction of vanadium by titanous ion takes place in two steps. Vanadium which is present in the vanadium(V) oxidation state is first reduced t o vanadium(1V) and this in turn is reduced to vanadium(II1). Reduction of vanadium stops a t this point and purple titanous chloride appears. The presence of vanadium( IV) in the hydrolyzed titanium tetrachloride solutions was indicated by the smaller number of equivalents of titanous ion required to reach the first end point than between first and second end points. Attempts were made to oxidize all the vanadium to vanadium(V) so that only one end point would be necessary, but these failed because of the high concentration of chloride ion. An accurate method resulted when the time of reduction for vanadium was taken as the difference between the vanadium(V)-vanadium( Is') end point and the vanadium(1V)-vanadium(II1) end point, and did not require any addition of reagents prior to titration. One wave length which was used successfully for the titration was 490 mp. This was decided upon after scanning the solution a t different stages of electrolysis. The vanadium(V) complexes with chloride in strong hydrochloric acid solutions to give strong absorption of light in the region of 490 mp. Figure 2, A , shows a typical titration curve which was obtained a t this wave length. The current in this and all other titrations was approximately 67 ma., and was determined accurately for each run. The acid concentration is rather critical when the titration 1s performed a t a wave length of 490 mr, because it determines the slope of the vanadium(V)-vanadium(1V) absorption line. If the acid concentration is too great the line is very steep and the pen is off scale evcept in the region of the end point; if the concentration is too low the slope is too small and the end point Is more difficult to locate. Best results were obtained when 10 ml. of titanium tetrachloride were hydrolyzed in the equivalent of 110 ml. of 7.5M hydrochloric acid. Acid and water can be measured accurately enough in a graduated cylinder, The quantities given iri the procedure for hydrolysis will give this concentration. A wave length of 760 mp, which Sabatini, Hazel, and McXahh ( 7 ) determined as the maximum for vanadyl ion, was also used to determine the spectrophotometric end point. The titration curve a t this wave length is shown in Figure 2,B. The results obtained a t 490 and 760 mp are comparable in preclsion and accuracy. When only a small amount of vanadium(V) is present a sharper first end point is obtained a t a wave length of 490 mp. It was found that equilibrium was quickly attained whenever the electrolysis was stopped at intervals during the titration. By using the stop clock and plotting absorbance against time, the name results were obtained, as when the titration w a p automatically recorded. The stoichiometry of the method was tested with known quantities of vanadium. Known amounts of 99.97% pure vanadium pentoxide (V20,) were dissolved in concentrated hydrochloric acid and added to the reaction flask. Ten-milliliter aliquots of
0.5-
a
(V+5,V+4)
4 v t 4 . v +3) Figure 2.
Titration of vanadium with coulometrically generated titanous ion A . Wave length 490 mp B. Wave length 760 m p
purified Crane titanium tetrachloride were then added to the flask and the acid was brought to the proper strength as previously described. The results for the spectrophotometric titration of known amounts of vanadium are shown in Table I. Determinations 1 to 6, inclusive, were run a t a wave length of 760 mp, and determinations 7, 8, and 9 were run a t a wave length of 490 mp. The average deviation for nine runs was 0.03 mg.
Table I.
Spectrophotometric Titration of Known Amounts of yanadium M .ofV 'faken 4.06 4.06 4.06 8.01 8.01
Run NO.
1 2 3 4 5
39.56 39.56 39.56 39.56
6 7 8 9
Mg. of V Found
Difference, hfg.
4.06 4.09 4.12 8.02 8.04 39.42 39.58 39.55 39.57
=to.oo +0.03 + O 06 + 0 01 + O 03 - 0 14 + 0 02 - 0 01 +o 01
The total amount of vanadium taken was the sum of the vanadium in the vanadium pentoxide and the small amount contained in the purified Crane titanium tetrachloride. Spectrographic analysis showed the purified titanium tetrachloride to contain about 0.0005% of vanadium. This was determined more accurately by taking different known amounts of vanadium pentoxide and determining by titration the total quantity of vanadium present. It was calculated by difference that the Crane purified contained 0.0006% vanadium, or 0.12 mg. per 10 ml. of titanium tetrachloride.
Table 11.
Determination of Vanadium in Titanium Tetrachloride Samples
Calcd. Vanadium, Vanadium Source % Found, % 1 Synthetic 0.2293 0,2292 0.2291 Synthetic 2 0.2292 0.2284 Synthetic 3 0.2292 Svnthetio 4 0.22')3 0.2292 C"rane Co. crude 5 0 1147 Crane Co. crude 6 0 1150 Fisher technical 7 .... 0 0896 Fisher technical 8 .... 0 0898" 9-16 Cenco technical .... 0 0731 Cenco technical. 5 ml. 16 0 0366 0 0367 Crane pur., 5 ml. 17 Cenco technical 5 ml. 0.0366 0.0364 Crane pur., 5 mi. 18 Cenco technical, 2 ml. 0.0127 0.0124 Crane pur., 10 ml. 19 Cenco technical, 2 ml. 0.0127 0.0124 Crane pur., 10 ml. a Average value, average deviation = 0.00016%. Run
.... ....
Deviation +0.0001 0.0001
-
+0.0001 - 0.0008
.... .... .... ....
:
+0 0001
-0.0002 -0.0003
-0.0003
Analyses for vanadium were made on the following commercial sources of titanium tetrachloride: Crane Co. crude, Fisher technical grade, Cenco technical grade, and mixtures of different proportions of Cenco technical grade and Crane purified. Table I1 shows these results together with results on four synthetic samples of high vanadium content. These were the same samples as 6 to 9 in Table I, but the calculations were presented as percentage of vanadium in titanium tetrachloride. The amount of vanadium in runs 16 t o 19 was calculated by averaging runs 9 to 15, inclusive, calculating the amount of vanadium in the volume of Cenco technical titanium used, and correcting for residual vanadium. The weight of the titanium tetrachloride sample was determined for each 10-ml. aliquot by calculating the specific volume according to the formula given by Sagawa (8, 9):
VI
= 0.56773 (1
+ 9.645 X lO-'t6.026+ x 10-7t* + 5.94 x 10-t~)
A N A L Y T I C A L CHEMISTRY
744 where t is temperature in degrees Centigrade. This wag found to be accurate within the experimental limits of the method. INTERFERENCES
Table 111. Determination of Iron and Vanadium in Titanium Tetrachloride
2
3
Source
2 nil. 0 0513.21 FeCla 10 ml. Cenro tech. TiCL 4 ml. 10 ml. 1 ml. 10 ml.
Iron, M g . Calcd. Found 5 73 5.7
0 0196.W FeCla Cenco tech. TiClr
4.38
0.0196.WFeCls Cenco tech. TiClr
4 38
I
5 00
The only interference studied was that of iron, as it was considered the most probable impurity besidee vanadium. S o iron was found in the commercial samples of titanium tetrachloride, EO synthetic samples were prepared by adding a standard solution of ferric chloride to a hydrolysed solution of Cenco technical titanium tetrachloride. Acid strengths were the Pame as those used for the determination of vanadium alone.
Run 1
.-I‘
-1.3 4.3
_-
Vanadium. lfg, Calcd. Found 12.62 12.6 12.62 12.62
12.8 12.8
Figure 3, A , shoJvs the titration curve a t a wave length of 190 mp, and Figure 3, B, s h a m the titration curve a t 760 mp. The presence of iron causes a flattening of the vanadium(1V)vanadium(II1) absorbance line near the second end point, thereby making the end point ~ O I difficult P to determine. By using t h e straight portion of the lines near the vanadium( 1V)-vanadium (111) and iron(II1)-iron(I1) end points, fairly accurate results for both elements can be obtained although they are not so accurate as for vanadium alone. .1 possible improvement muld be made by adding a reagent which would form a ferrous compIex of high absorbance. Table I11 shows the results for analyses of per cent vanadium and iron in the same titanium tetrachloride solution. Titrations 1 and 2 were made a t 490 my and titration 3 was made a t 760 my. The iron solutions wei e standardized n-ith potassium dichromate.
em
m IO.
51
703
800
KO
400
333
ar,
loo
3 INCHES
205-
-
0
-(Fet:Fet2)
_ -~
-
(V ‘4V t3 ) - __..-
-(Vf4V+4) -_ -_
ACKNOWLEDGMENT
The authors are indebted to Cramet, Inc., Chattanooga, Tenn., for financial support of this project. LITERATURE CITED
(1) Arthur, P., and Donahue, J. F., ANAL.CHEM.,24, 1612 (1952). I b i d . , 22, 1314 (1950). (2) Goddu, R. F., and Hume, D. N., (3) Kolthoff, I. M.,and Furman, N. H., “Potentiometric Titrations,” 2nd ed., p. 373, Wiley, New York, 1931. {4) Malmstadt, H. V., and Gohrbandt, E. C., ANAL.CHEM.,26, 442 (1954). (5) Reilley, C. N., Cooke, W. D., and Furman, N. H., Ibid., 23, 1030 (1951). (6) Reilley, C. N., and Schweizer, B., Ihid., 26, 1124 (1954). (7) Sabatini, R., Hazel, J., and McNabb, W., A n a l . Chim. A c t a , 6, 368 (1952). (8) Sagawa, T., Complete Abstr. J a p a n . Chem. Lit.,6, 341 (1932). (9) Shirrer, G., Johnston, H. L., and Beckett, C., “Titanium and Its Compounds,” p. 43, Herrick L. Johnston Enterprises, Columbus, Ohio, 1954. (10) Sweetser, P. B., and Bricker, C. E., ANAL. CHEM.,24, 409 (1952). (11) Ihid.. 25. 263 (1963). tl2j Underwdod, A: L., Ibid., 25, 1910 (1953). (13) I h i d , , 26, 1322 (1954). (14) Wise, E. N., Gilles, P. TI7., and Reynolds, C. A., Jr., Ibid., 25, 1344 (1953). RECEIVED for review September 20. 1954. Accepted December 30, 1QZ-1.
Determination of Erythromycin by Ultraviolet Spectrophotometry JOHN
B. TEPE
and
C. V. ST. JOHN
Eli Lilly & Co., Indianapolis, lnd. I
The development of an accurate method for the determination of erythromycin w-as based on the observation that alkaline hydrolysis produced a material that had a characteristic ultraviolet absorption. An acid-inactivated blank was employed to correct for the ultraviolet absorbing impurities and degradation products of erythromycin. The assay has shown good reproducibility, sensitivity, and correlation with a turbidimetric microbiological assay.
E
RYTHROMYCIN ( 7 )is a basic antibiotic that is used estensively for the treatment of various bacterial infections. An accurate method of assay was desired for processing and formulation, during the manufacturing of this antibiotic. A colorimetric assay for erythromycin reported by Ford and coworkers ( 8 ) employed sulfuric acid to develop a color that absorbed at 485 mp. Acid-hydrolyzed erythromycin (8) reacted with NelEon’s reagent to develop a blue color. These methods do not appear to be specific for erythromycin.
During the course of t,his investigation for a chemical method of analysis, many color reactions were studied. Those that gave positive reactions with pure erythromycin are listed with comments as to their usefulness in analyzing samples of varying purities that are encountered in the fermentation and purification of the antibiotic. The procedures and conditions employed were the same as described in the literature. 1. Erythromycin when heated with a mixt,ure of glacial acetic acid and perchloric acid (varying proportions) gave a color that absorbed strongly at 485 my. Reproducibility was good but the test gave higher values on process samples than microbiological assay or the ultraviolet assay. 2. Erythromycin mas oxidized by periodic arid and the resultant, formaldehyde vas determined by chromotropic acid ( 2 ) . The reproducibility was good but the test was not sufficiently sensitive for dilute solutions. 3. The tryptophan-perchloric acid reagent. ( 3 ) reacted with erythromycin t o produce a red color. This method had good sensitivity and reproducibility; however, degradation products of erythromycin in impure samples reacted with the reagent. 4. The xanthydrol reagent ( 9 ) reacted with the erythromycin to give a red color that absorbed strongly at 540 my. The sensi-
