perometric methods for the determina- tion of halide ions recovered from pe- troleum products are discussed by Parks and Lykken (8).

Titrations were performed a t about 5 C. to take advantage of the lorver solu- bility of d y e r chloride (5) at lower temperatures. This allows greater ac- curacy in the titration and improves the appearance of the titration curve.

Four of these apparatus in use in this laboratory operate in pairs from the same set of gas cylinders and air- purification systems. Individual vacuum pumps are attached to each unit. For more than one year these

units have been in continuous use for the determination of micro- and milli- gram quantities of sulfur and chlorine.

ACKNOWLEDGMENT

The author wishes to express his thanks to hlildred 0. Corley for dili- gently operating the apparatus during the course of this investigation.

LITERATURE CITED

(1) Agazzi, E. J., Peters, E. D., Brooks, F. R., ANAL. CHEW 25,237 (1953).

(2) .4m. SOC. Testing Materials, Phila- delnhia. Pa.. “Standards on Petro- leuk P6d;cts and Lubricants,” p. 23, 1955.

(3) Granatelli, L., ASAL. CHERI. 27, 266 (1955).

(4) Kolthoff, I. I f . , Kuroda, P. K., Ibid., 23, 1306 (1951).

(5) Kolthoff, I. M., Sandell, E. B., “Textbook of Quantitative In- organic Analysis,” 3rd ed., p. 60, Macmillan, New York, 1952.

( 6 ) Ibid., p. 541. (7) Lieeett. L. M.. ANAL. CHEW 26. 748 . I .I.,

(1954).

1444 (1950).

cox, G. W., Ibid., 22,311 (1950).

21, 721 (1949).

(8) Parks, T. M., Lykken, L., Ibid. , 22,

(9) Pecherer, B., Gambrill, C. If., Wil-

(10) Wear, G. E. C., Quiram, E. R., Ibid. ,

RECEIVED for review -4ugust 9, 1956. Ac- cepted November 23, 1956.

Determination of Trace Amounts of Chlorine in Naphtha J. G. BERGMANN and JOHN SANIK, Jr.

Research Department, Standard Oil Co. (Indiana), Whiting, Ind.

,A method has been developed for determining less than 10 p.p.m. of chlorine in naphthas. Lamp combustion or sodium biphenyl reduction i s used to convert organic chlorine to chloride ion; for most naphthas, lamp combus- tion i s preferred. Of several ways investigated for determining less than 100 y of chloride, only a ferric thio- cyanate colorimetric procedure pro- vided sufficient accuracy. With lamp combustion, the method has a standard deviation of 0.1 p.p.m, and an analysis can be completed in 3 hours.

MONG the trace elements found in A light petroleum fractions are ar- senic, iron, nickel, vanadium, chlorine, nitrogen, and sulfur. hlethods for determining some of these elements in the parts-per-million range have been published (3, I S ) , but no method has been available for determining less than 10 p.p.m. of chlorine in naphthas within 0.5 p.p.m.

Two types of methods are possible: direct and indirect. In the first, the chlorine is determined directly in the naphtha; in the second, it is first sepa- rated from the naphtha and then determined in aqueous solution. Direct methods include neutron activation, x-ray fluorescence, and flame photom- etry. Neutron activation analyses of several naphthas containing known amounts of chlorocyclohexane shot\-ed an accuracy within 1.5 p.p.m., which was unsatisfactory for the purposes of this investigation. However, the analyses showed the presence of only negligible

quantities of halogen other than chlorine. Other direct methods either do not apply to trace amounts of chlorine or apply only to specific chlorine compounds. Hence, procedures were needed for sepa- rating chlorine from the naphtha as chlo- ride ion, and for determining trace amounts of chloride ion in aqueous solution.

Two general procedures exist for converting organic chlorine to chloride ion: oxidation of the entire sample (1, W), and reduction with sodium in various media (5, 14-16, 18). Most of the procedures require excessive reac- tion time for the more refractory chlo- rine compounds or cannot accommodate large enough samples. Exceptions are lamp combustion ( 1 ) and sodium bi- phenyl reduction (14, 16). Both are rapid and permit the use of 10 to 15 grams of sample-enough to provide sufficient chloride ion for accurate measurement.

Procedures for determining small quantities of chloride ion in aqueous solution include nephelometry ( I I ) , amperometry (IO, l a ) , differential potentiometry (4) I mercuric nitrate titrimetry (6) , and a microdiffusion- colorimetric method (‘7, 8, 17). Each procedure was investigated for measur- ing less than 100 y of chloride-the amount obtained from 10 to 15 grams of naphtha. However, accuracies were not sufficient, particularly below 50 y , because the amount of chloride ap- proaches the sensitivity limits of these procedures. A new and more sensitive procedure (9) depends upon displace- ment of thiocyanate ion from mercuric

thiocyanate by chloride ion; in the presence of ferric ion, a highly colored ferric thiocyanate complex is formed,

2C1- + Hg(SCN)t + 2Fe+++ -+

HgClz + 2 Fe(SCN)++

the color of which is stable and pro- portional to the original chloride ion concentration. The procedure can de- tect as little as 0.5 y of chloride and is accurate within 0.5 y in the range of 1 to 100 y .

A method has been developed for de- termining trace amounts of chlorine in naphtha. Organic chlorine is con- verted t o chloride ion by lamp com- bustion or sodium biphenyl reduction. After either method of isolation, the ferric thiocyanate colorimetric proce- dure determines the chloride ion.

LAMP-COMBUSTION PROCEDURE

Combustion is carried out in appa- ratus identical in principle to the ASTbl lamp sulfur apparatus (1) . It consists essentially of a means for burning the sample from a wick in a synthetic atmosphere, composed of 70% carbon dioxide and 3001, oxygen, and for absorbing the hydrogen chlo- ride from the combustion products. The apparatus is designed to accom- modate a blank and one or more samples.

Two modifications to the apparatus were made: inclusion of Fischer-Porter C-type rotameters to indicate the gas flow to each burner, and the use of water, rather than hydrogen peroxide, in the

VOL. 29, NO. 2, FEBRUARY 1957 241

scrubber, chimney-manifold manometer, and absorber. The rotameters (cali- brated from 0 to 0.2 cubic foot per min- ute) aid in the initial adjustment of the apparatus and improve the control of gas flow during combustion. Water quantitatively absorbs the hydrogen chloride produced and simplifies the subsequent determination of chloride. All water used is ion-exchanged; dis- tilled water passed through Amberlite MB-3 resin is satisfactory. To avoid contamination from the hands, lamp wicks are handled only with clean for- ceps.

Preliminary adjustments of the ap- paratus are made to provide the proper pressures ( I ) , and the burner vacuum valve is set so that the rotameter indi- cates a flow rate of 3 liters (0.1 cubic foot) per minute of the carbon dioxide- oxygen mixture through the absorber. A weighed sample of about 20 ml. is taken for combustion. With minor ad- justments of the control valves to pro- vide a stable and nonsmoking flame, the rotameter shows a flow rate of about 1.5 liters per minute. After nearly all the naphtha has burned, combustion is stopped and the flask is reweighed to determine the amount of naphtha burned. A blank must be run concur- rent with the combustion (1).

The aqueous solutions from both ab- sorbers are transferred to 50-ml. volu- metric flasks. Absorbers, chimneys, and spray traps are rinsed three times with small quantities of water. The washings are added to the absorber solutions, and final solutions are adjusted to volume.

SODIUM BIPHENYL PROCEDURE Sodium biphenyl reagent is prepared

by the reaction of 14.5 grams of sodium and 98 grams of biphenyl in 480 ml. of ethylene glycol dimethyl ether (14). These concentrations of sodium and bi- phenyl are lower than those used by Liggett (14) and give a reagent that is easier to pipet and provides more uni- form blanks.

Ten to 20 ml. of the naphtha in a separatory funnel is shaken for 30 sec- onds with sufficient sodium biphenyl to maintain a green color (usually no more than 10 ml.). The excess reagent is de- composed with 10 ml. of water; 20 ml. of hexane is added, and the mixture is agitated and allowed to settle. The aqueous layer is withdrawn into a 60-ml. nickel beaker (Nickelware is used be- cause it is alkali-resistant and because nickel does not interfere in the colori- metric procedure.) The funnel and contents are washed with 10 ml. of water, which is added to the initial ex- tract. To remove interfering substances extracted by the sodium biphenyl, the combined extracts are adjusted to a pH of 10 to 12 with concentrated nitric acid, evaporated to dryness on a hot plate, and ignited in a muffle furnace a t 500' to 550' C. for 30 minutes. (Hydrogen peroxide or ammonium persulfate oxi- dizes most interfering substances except thiophene and its homologs, which have been found in some naphthas.) The residue is dissolved in a minimum amount of water, and the solution is transferred to a 25-ml. volumetric flask

Table 1. Precision and Accuracy with Lamp Combustion Chlorine, P.P.M.

Mid-continen t Treated Mid- n-Heptane Kaphtha continent Naphtha

Added 0 .0 8.65 0.0 4.07 0 .0 3.98 Found 2.12 10.64 1.79 5.87 0.77 4 93

2.31 10.94 1.88 5.96 0.87 4 95 2.33 11.01 1.90 5.97 0.87 4.96 2.43 11.09 1.97 5.99 1.20 5 24

Av . 2.30 10.92 1.89 5.95 0.92 5.02 1.89 0.92

Recovered 8.62 4.06 4 10 - - 2.30 -

Table It. Comparison of Combustion and Sodium Biphenyl Reduction Chlorine. P.P.M.

Naphtha n-Heptane plus added C1 hfid-continent lus added C1 Mid-continent ~160'-390' F.) Mid-continent ~100"-360' F.)

Sodium bi henyl Lamp re Juction combustion 10.6, 10.8 5.6, 5 .9 6.1, 6 . 3 1 .9 . 2 . 1

Mid-continent (190'-400' F.) 1.9; 2 . 4 Mid-continent (200"-370" F.) 0.9, 1 . 5 Mid-continent, treated (190"-360" F.) 0 .9 , 1 . 9 Gulf Coast, naphthenic (23Oo-41O0 F.) 1.1, 1 . 4 Mid-continent, treated ( 180"-380' F.) 1 .3 , 1 . 6

1 0 . 9 , l l . O 5.9, 6 . 0 6 .6 , 6 . 7 1.8, 1 . 9 1.4, 1 . 5 1.5, 1.8 1.0, 1 3 1.0, 1 .2 0.8, 0 . 9

and diluted to volume with water. A blank is carried through the entire pro- cedure.

COLORIMETRIC PROCEDURE

A 20-ml. aliquot of the chloride solu- tion obtained by either lamp combustion or sodium biphenyl reduction is trans- ferred to a 25-ml. volumetric flask. Two milliliters of a 0.25M ferric am- monium sulfate [Fe(NH4) (S04)2.12H20 ] solution in 9M nitric acid is added, followed by 2 ml. of a saturated solu- tion of mercuric thiocyanate in ethyl alcohol. The solutions are mixed, di- luted to volume with water, and mixed again. The blank is treated in the same manner. Ten minutes after developing the color, the absorbances of both solu- tions are measured in a spectrophotom- eter against water in a 5-cm. cell a t 460 mp; a Beckman Model R is satis- factory. (Because any chloride present in the air is slowly absorbed, these measurements are made as promptly as possible, and water is used in the refer- ence cell instead of the blank.) The amount of chloride in the aliquot cor- responding to the difference between the two absorbances is then obtained from a calibration curve, and the original con- centration of chlorine in the naphtha is calculated,

To prepare the calibration curve, a t least five aliquots of a standard sodium chloride solution containing 10 y of chloride per ml. are taken to cover the range from 0 to 50 y; color development and absorbance measurement follow the outlined procedure. dbsorbance is plotted against micrograms of chloride.

DISCUSSION OF RESULTS

The accuracy and precision of the

method were tested in three ways. Lamp combustion was used to study the recovery of chlorine added tu naphthas. The two methods of isolat- ing chloride were compared with the same naphthas. Several naphthas of different origins were analyzed in duplid cate with lamp combustion.

Table I shows the virtually complete recovery of chlorine from naphthas to which chlorine had been added as chloro- cyclohexane. Four burnings were made on each sample. The standard devia; tion of the results is 0.1 p.p.m., and the average difference is 0.1 p.p.m.

Kine naphthas with chlorine con. tents established by lamp combustion were analyzed by sodium biphenyl reduction. The results in Table I1 show a standard deviation of 0.3 p.p.m. and an average difference of 0.3 p.p.m.

Duplicate routine analyses of typical naphthas are shown in Table 111. Lamp combustion was used for isolating chloride. The results show a standard deviation of 0.1 p.p.m. of chlorine, The method thus gives as precise re- sults with naphthas of widely different origins as it does with synthetic blends.

To obtain accurate results, two pre- cautions must be observed. The anal- yses must be conducted in a room from which fumes of interfering substances are excluded. The lamp-combustion apparatus and all glassware used must be reserved solely for determining trace chlorine.

CONCLUSION

Kith either lamp combustion or sodium biphenyl reduction, the method

242 ANALYTICAL CHEMISTRY

Table 111. Duplicate Analyses of Typical Naphthas with Lamp Corn bustion

Chlorine, P.P.M.

Alid-continent (l6O0-39O0 F.) 6 59,6 70 Mid-continent (140"-390' F.) 4 00,4 28 Mid-continent (10Oo-39O0 F.) 3 46,3 58 Arkansas 2 32,2 39 West Texas coker 1 23 , l 63 Mid-continent (250°-3800 F ) 1 11 , l 27 Gulf Coast. naphthenic

(210"-360" F.) 1 04,l 17 Mid-continent (210"-390" F ) 0 76,O 89 Kuwait 0 58,O 67 Mid-continent (200"-310" F ) 0 56,O 58 Wyoming, treated 0 23,O 23

is well suited to multiple analysis by nontechnical operators.

The absorber solutions from lnmp combustion are ideally suited t o the colorimetric procedure because of the absence of interfering ions. A 15- gram sample can be burned in about 2 hours, and the total analysis can be completed in about 3 hours. Lamp combustion is therefore recommended for isolating chloride from naphthas.

Although chloride can be separated from the naphtha in 10 minutes by

sodium biphenyl reduction, the time advantage is negated by the subsequent manipulations required. These steps in- crease the chances of chloride cont,am- ination and make the determination of chloride by the sodium biphenyl procedure less precise and accurate than by lamp combustion. Sodium biphenyl reduction is therefore recommended only for materials that do not burn cleanly in a lamp.

The method should be applicable to trace-chlorine analysis of other petro- leum fractions. Some heavier materials can be burned satisfactorily when blended with a low-chlorine diluent; others can best be analyzed by sodium biphenyl reduction.

( 5 )

LITERATURE CITED

Am. SOC. Testing Materials, ".1STM Standards on Petroleum Products and Lubricants," D 1266-551', p. 679 (1955).

(1955).

(1955).

Ibid. , ASTlI D 808-52T, p. 308

Barney, J. E., ANAL. CHEM. 27, 1283

Blaedel,' W. J., Lewis, W. B., Thomas, J. W., Ibid. , 24, 509 (1952).

Chabloy, E., Ann. chim. 1, 469 (1914).

Blaedel,' W. J., Lewis, W. B., Thomas, J. W., Ibid. , 24, 509 (1952)

Cl (1914).

Determination of Vanadium in Titanium Tetrachloride and Titanium Alloys WILLIAM H. OWENS', CHARLES L. NORTON, and J. ALFRED CURTIS

Cramet, Inc., Chattanooga, Tenn.

(6) Clarke, F. E., ANAL. CHERI. 22, 553 (1950).

( 7 ) Conway, E. J., "Micro Diffusion Analysis and Volumetric Error," rev. ed., chap. XXII, C. Lock- wood, London, 1947.

(8) Gordon, H. T., ANAL. CHEm 24, 857 (1952).

(9) Iwasaki, I., Utsumi, S., Ozawa, T., Bull. Chem. SOC. Japan 2 5 , 226 (1952).

(10) Kolthoff, I. bl., Kuroda, P. K., ANAL. CHEW 23, 1306 (1951).

(11) Kolthoff, I. M., Yutzy, H., J . Am. Chem. SOC. 5 5 , 1915 (1933).

(12) Laitinen, H. *4., Jennings, W. P., Parks, T. D., IND. EKG. CHEM.,

(13) Lake, Y. R., XlcCutchan, P., Van Meter, R., Neel, J. C., ANAL. CHEX. 23, 1634 (1951).

(14) Liggett, L. M., Ibid., 26, 748 (1954). (15) Lohr, L. J., Bonstein, T. E., Frauen-

felder, L. J., Ibid. , 25, 1115 (1953).

(16) Pecherer, B., Gambrill, C. XI., Wil- cox, G. W., Ibid., 22,31,f (1950).

(17) Rodden, C. J., Ed., Analytical Chemistry of the Manhattan Proj- ect," p. 297, McGraw-Hill, New York, 1950.

(18) Umhoefer, R. R., IXD. ENG. CHCM.,

ANAL. ED. 18, 355 (1946).

ANAL. ED. 15, 383 (1943).

RECEIVED for review June 27, 1956. -4c- cepted September 20, 1956. Division of Petroleum Chemistry, 130th Meeting, ACS, Atlantic City, N. J., September 1956.

A spectrophotometric procedure has been developed for the determination of trace amounts of vanadium in titanium tetrachloride. It is also appli- cable to the determination of vanadium in alloying amounts in titanium-base alloys. The procedure i s very simple, rapid, and accurate to within 270 relative. Chromium offers the only interference found in a limited in- vestigation.

RAPID and accurate control method A is necessary in the production of titanium tetrachloride of high purity. Most methods for determining vana- dium, including the peroxide and phos- photungstic acid (2 , 3) methods, are sub- ject to interference from titanium. Volumetric methods ( 1 ) are suitable for alloying amounts of vanadium but are

1 Present address, Yational Refearch Metals Corp., Cambridge, Mass.

not accurate enough for 0.100% and less. Lengthy procedures involving chemical separations are, of course, un- desirable.

Upon evaporation of a sulfuric acid solution of titanium tetrachloride con- taining vanadium with nitric acid. an intense yellow color was produced, which became more intense on further evaporation. As the color nas found to be reproducible if the evaporation was carried to fumes of sulfuric acid, it was used as a basis for determining vana- dium in titanium tetrachloride. The method was also applicable to the de- termination of vanadium in titanium alloys.

REAGENTS

Standard vanadium solutions, 0.05

Sulfuric acid, concentrated and 1 to 1. Nitric acid, concentrated. Vanadium-free titanium tetrachloride.

and 0.50 mg. per ml.

APPARATUS

Beckman 1Iodel DU or Model B

Glass hypodermic syringes, 2-mi. and spectrophotometer with 5-mm. cells

10-ml.

PROCEDURES

1. Vanadium in Titanium Tetrachlo- ride. Pipet a 2-ml. aliquot of vana- dium-free titanium tetrachloride into a clean, dry 250-ml. beaker. Slowly and carefully add 40 ml. of 1 t o 1 sulfuric acid. Rinse down the sides of the beaker with dilute sulfuric acid and then add 10 ml. of nitric acid. Cover and place on a hot plate. When most of the water and nitric acid have evaporated, remove the tvatch glass and continue heating. After all the water and nitric acid have evaporated, there is a short time before the fumes of. sulfuric acid appear. As soon as these fumes appear, remove the beaker from the hot plate. Allow the beaker to cool to about 50" to 60' C. and then

VOL. 29, NO. 2, FEBRUARY 1957 243