316 I N D U S T R I A L A N D E N G I N E E R I N G CHEMISTRY Vol. 17, No. 5

be possible, however, to standardize the time and temperature of contact, so that results agreeing with those obtained by a refer- ence method can be duplicated.

LITERATURE CITED ( 1 ) Almy, E. G., Griffin, W. C., andwilcox, C. S., TND. ENG. CHEar.,

( 2 ) .Isuoc. Official Agr. Chem., Official and Tentative Methods of

( 3 ) Ibid., p. 335. (4) Bryant, W. M. D., Smith, D. M., Mitchell, J., Jr., and Ashby, E.

C., J. Am. Chem. SOC., 63, 2924-26 (1941). (5) Fischer, K., Angew. Chem., 48, 394-6 (1935). (A ) Gilman and Blatt, “Organic Syntheses”, Collective Volume I,

2nd ed., pp. 219-49, New York, John U‘iley and Sons, 1941.

A N A L . ED., 12, 392-6 (1940).

.4nalysis, 5th ed., p. 308 (1940).

(7) Kaufmann, H. P., and Funk. S., Fette u. Seifen, 44, 345 (1937). (8) McKinney, D. C., Jr., and Hall, R. T., IXD. ENG. CHEM., ANAL.

(9) Makower, B., and Myers, S., Proc. Inst. Food Tech. 1943, 156-

(10) Mitchell, J., Jr., IND. ENG. CHEY., ANAL. ED., 12, 390-1 (1940). (11) Porter, 1%‘. L., and Willets, C. O., J . Assoc. Oficial Agr. Chem.,

(12) Prater, A. N., Johnson, C. M., Pool, M. F., and Mackinney, G.,

(13) Reeve, R . M., Food Research, 8, 128-36 (1943). (14) Richter, Angezc. Chen. , 48, 776 (1935). (15) Smith, D. M., Bryant, W. M. D., and Mitchell, J., Jr., J . Am.

(16) Wernimont, G., and Hopkinsop, F. J., IND. ENG. CHEX.. A N ~ L .

ED., 15, 460-2 (1943).

64.

27, 179-194 (1944).

IND. EIG. CHEX., ANAL. ED., 16, 153-7 (1944).

Chem. SOC., 61, 2407-12 (1939).

ED., 15, 272-4 (1943).

Determination of Water in Liquid Petroleum Fractions 0. T. AEPLP AND W. S. W. McCARTER

Porocel Corporation, 260 South Broad St., Philadelphia, Pa.

Small amounts of water (0 to 60 p.p.m.) in unleaded gasoline, naphtha, and electrical insulating oils can be determined b y the Karl Fischer reagent through the use of a specially desisned apparatus which effects the total exclusion of atmospheric moisture during the sampling and titration.

’ ” ’hT work (4 ) on the desiccant properties of Driocel activated bauxite) required the accurate determination of

very small amounts of water in liquid hydrocarbons. Prelimi- nary investigation of the available methods indicated that the Karl Fischer reaction could be most satisfactorily adapted to the needs. This reagent (2, 5) is quantitatively specific for water and its use for the determination of water in alcohols, carboxylic acids, saturated, ethylenic, and aromatic hydrocarbons (5), and petroleum fractions (1) has been reported. Early in the au- thors’ study it became evident that for the small amounts of water involved in their work (0 to 60 parts per million) repro- ducibility could be obtained only through the use of a sampling and analytical technique which would effect the complete exclu- nion of atmospheric moisture. Publication of the apparatus and method used in this laboratory is prompted by the increasing significance of moisture content in many processes of the chemi- cal and pet,roleum industries.

R y . -

REAGENTS

Directioiis for the preparation of the Karl Fischer reagent appear in the literature (2,5). Since the concentration of the re- : y w t generally decreases rapidly over several days immediately tollowing its preparation (Figure l), it is advisable to make the solution at least 24 hours prior to use and to standardize it fre- queritly thereafter until the concentration has become stable. Thtx need for frequent standardization of the reagent is further emphasized by the three curves which show that the rate of de- cwase in concentration is not predictable. The reagent was standardized against a solution of water in methanol containing I rng. of water per milliliter.

APPARATUS

’fhe sampling bulb shown in Figure 2 was designed to secure 150-ml. ortions of the liquid to be analyzed, either from stream 01‘ from Eulk, without exposure to the atmosphere. Stream sam- ple+, as from a drying tower, may be secured by attaching the hulh to the effluent tube of the tower by means of the ground- gla- connection, fi. With h3 closed, stopcocks hl and hz are oprried to allow the bulb to fill. Samples from bulk may be ob- taintad by attaching at f i an extension tube long enough to dip N ell into the body of liquid. With hl closed, h~ and h3 are opened :mi the bulb is filled by applying slight vacuum at hs. Before

1 Present address, Pennsylvania Salt Manufacturing Co., Philadelphia, Pa.

taking a sample for analysis, the dried bulb should be rinsed with the liquid to adjust the adsorbed water concentration on the walls of the vessel.

Titrations were carried out in the totally enclosed apparatus shown in Figure 2. A motor-driven glass stirrer, a, enters the 500- ml. flask through a mercury seal, b, attached by means of a 34/45 ST ground-glass joint. The flask is provided with three side necks: c, a 10/30 ST joint through which enters the tip of a 50- ml. Schilling buret for standard water-in-methanol solution; d, a 14/35 ST joint to hold a platinum electrode assembly; and e , a 10/30 ST joint to accommodate in turn joint f2 of the sample bulb, the tip of a 10-ml. Schilling buret (not shown) for Karl Fischer reagent, and a desiccant tube, g.

The “dead stop” end-point method (3, 6) &-as adopted in preference to the more difficultly detectable and less precise visual end point. The electrodes and wiring diagram required by this method are shown in Figure 2. By means of a battery and variable resistance an e.m.f. of 0.5 to 1.0 millivolt is imposed across the platinum electrodes which dip into the Karl Fischer reagent. The lower e.m.f. requirement as compared to Kerni- mont’s value is probably related to the immiscibility of the re- agent in petroleum fractions. \Then the end point is reached one of the electrodes becomes polarized and current cease> to flow. This condition is indicated by the gradual return of the galvanometer spot to the zero position. A Leeds & Northrup enclosed lamp and scale galvanometer, catalog S o . 2420-B (sensitivity, 25 microvolts per millimeter; period, 3 seconds; roil resistance, 23 ohms), is a suitable instrument.

METHOD

For gasoline and naphthas, which are normally lighter than Karl Fischer reagent, the electrodes extend to the bottom of the task. With heavier immiscible liquids, such as electrical insu- lating oils, the electrodes are constructed to contact the upper

= I - --I & 5

AGE OF SOLUTION (HOURS)

Figure 1. Decomposition of Representative Solutions of Karl Fischer Reagent

May, 1945 A N A L Y T I C A L E D I T I O N 317

cocks h, anti h:; the sainple ia drained into tlic’ titration flask. The sample holder is then quicklj- replaced by the buret for Karl Fischer reagent, and from 7 to 9 ml. of reagent are added to thr flask. As this buret is removed the drying tube is quickly slipped into place a t e . The emptied sample holder is reweighed to determine t,he amount of liquid transferred to the flask. The addition of 5 to 6 ml. of standard vvater-in- methanol solution at c helps to prevent adherence of the reagent to the walls of the flask after subsequent agitation. After mechanical stirring the excess Karl Fischer reagent is titrated with standard xater-in-methanol solution. In this back-titration each addition of water-in-methanol solution is followed by a brief period of stirring, after which the mixture is allowed to separate in- to layers. At the completion of the titration the sample is drawn off a t j and the titration flask, without further preparation, is ready for the next

9 ’

analysis.

The effect of time of stirring following the addi- tion of the Karl Fischer reagent to the hydrocar-

u Figure 4.

bon is shown in Table 11. It appears that 15 minutes is adequate to bring about substantially complete reaction.

Titration Flask, Sampling Bulb, and Electrode Assembly

REPRODUCIBILITY

The results of duplicate determiriatioils of the water content of unleaded gasoline, naphtha, and inhultLtirig oil samples are

90. Titration Titration Titration given in Table 111. The time of stirring in each determination 1 5.00 5.43 5 . 4 4 was 15 minutes.

The agreement between the tlvo determinations on each 2 3 . 2 1 3 . 7 7 3 .77 3 2 . 7 1 3 .59 3.60 4 2 .38 2 . 6 1 2.61 sample is within 2 p.p.m. and in some cases, especially in the 5 2 .19 2 . 5 4 2.54 6 1 . 1 4 1 .81 1 . 8 0 lower concentrations of water, t,he deviation is within 1 p.p.m.

Table I. Standardization of Karl Fischer Reagent OIg. of 1320 ~~ 1.00 ml. of KF reagent)

Solurioti First Second Third

Table II. Effect of Stirring Time (Parts per niillion of water)

Time, riileaded Gasoline Insulating Oil LIiii, 1 2 Naphtha 1 2

Stirring

30.5 . . 36.2 , . . . . . ..

Table 111. Reproducibility of M e t h o d (Parts per million of water)

ITtileaded t iawline Naphtha Insulating Oil 1 2 1 2 1 2

liquid layer. In the latter case, for obvious reasons, approxi- mately the same volume of liquid hydrocarbon should always be taken for analysis.

For standardization, 6 to 7 ml. of the Karl Fischer reagent are drained from the buret into the titration flask at e . The buret is removed and immediately replaced by the drying tube, 9, be- fore titrating with the standard water-in-methanol solution.

As shown in Table I, the first titration of the day gives a lower value than subsequent titrations. The agreement between the second and third titers indicates that adequate protection is afforded by the apparatus.

Following the standardization, the reaction mixture is drained from the flask through the capillary stopcock, j, with the drying tnhe still in place at e . The apparatus is then ready for actual analysis.

A sample to be analyzed for water is collected in the sample bulb and weighed. Drying tube g is removed from the titration flask at e and quickly replaced by the sample bulb. The drying tubr 1s then attached to the sample bulb at f i . By opening stop-

ACCURACY

Because of the difficultieb and uncertainties involved in prepar- ing water-in-hydrocarbon solutions of known concentration in the range under ronsideration, an evaluation CJf the “absolute” accuracy of the method would be questionable. As an alterna- tive, the acruracy was estimated by an indirect method with solutions prepared through the medium of a third component These solutions were prepared by mixing amall weighed quanti- ties of a water-in-pyridine solution of known concentration (determined by titration) with weighed quantities of exhaus- tively dried naphtha. The results show no interference by the hydrocarbon:

Parts per million of a a t e r Present Found (AY.)

9 . 3 20.8 3 0 . 7 3 3 . 6 60.7

104.9

8.8 20.5 30 .2 3 3 . 3 60 .4

104.4

The method is not applicahle to qasolint~ containing tetraethyl- lead. sample of gasoline> known to have a water conterit o f 19.3 p.p.m. showed an apparent viater content of 243 to 247 p.p.m. after thr addition of 4 ml. of Ethyl Fluid per gallon. Ob- viously thtl reagent is consumed by thtl Ethyl Fluid, giving erroneously high value\.

LITERATURE CITED

( 1 1 C’ayell, K. G., I inero , 11. C. . anti Slooie, J . I\ .. Phem. & . \ I t / .

(2) E‘isvher, K., A n g e w . chem. . 48, 394 (1935). (3) Foulk, C. IT., a n d B a w d e n , .\. T , J . Am. C‘hem. boc., 48, 2045

Eng., 50, 107 (1943).

(19261. \ - - - -,

(4) L a Lande , R. .i., M c C a r t e r . W. 9. IT., a n d Sanborn . .J. B.,

(5j Smi th , D. M., B r y a n t , IT. M. TI., and Mitchel l , J. J., J . dm. IND. ENG. CHEM., 36, 99 (1944).

Chem. SOC., 61, 2407 (1939). (t i l Werrii inont, G., and Hopkinson, 8. .J.. IKD t : ~ b CHEM.. . ~ Y \ L

r.:D.. 15 ,272 (1943).