Improvements in Karl Fischer Method for Determination of Water E. D. PETERS and J. L. JUNGNICKEL Shell Development Co., Emeryville, Calif. A substantial gain in stability can be achieved by substituting methyl Cellosolve for methanol in the formula for Karl Fischer reagent. In addition to its greater stability, the modified reagent extends the applicability of the method by permitting an appropriate choice of sample solvent. Methanol is an undesirable solvent in some cases because of interfering side reactions with the sample. A mixture of ethylene glycol and pyridine permits the direct titration of water in ketones and some aldehydes and improves the direct titration of water with Fischer reagent to the reverse dead-stop end point.
Add, in small increments and with constant swirling, 70 ml. of anhydrous liquid sulfur dioxide from a graduated cylinder and mix thoroughly. The water equivalence of this reagent is approximately 6 mg. of water per ml. of reagent. The relatively greater stability of Fischer reagent prepared with methyl Cellosolve compared with that prepared with methanol is shown in Figure 1. The curves show the percentage decrease in strength from the original standardization on standing.
Methyl Cellosolve Reagent
F
OR the quantitative determination of water in various materials, the Karl Fischer reagent ( 5 )has found continually expanding application. The principal advantages of this versatile reagent lie in the broad applicability and the rapidity with which titrimetric determinations can be made. Because eubstances which normally interfere in the method often react quantitatively or can be converted to inert derivatives, and because many compounds other than water can be determined by means of stoichiometric water-producing or water-consuming reactions with other substances, the method has a wide scope. One disadvantage of the reagent is its relative instability, which necessitates frequent standardization. It has become common practice to prepare Fischer reagent with methanol, which serves the dual purpose of acting as diluent and reacting with the pyridine-sulfur trioxide formed in the primary reaction of water with iodine, sulfur dioxide, and pyridine (7). Although this methanolic reagent is satisfactory for most purposes, a reagent prepared with the monomethyl ether of ethylene glycol (methyl Cellosolve) in place of methanol has several advantages, whereas no difficulty attributable t o this substitution of diluents has been observed in 15 years of use by the authors. Reagent prepared with methyl Cellosolve is appreciably more stable than that prepared with methanol, is less subject to interfering side reactions, and has some advantage in the direct titration with Fischer reagent to the reverse deadstop end point. Simple, but effective, apparatus for dispensing the reagent and conducting the titrations also have been thoroughly tested. MODIFIED FISCHER REAGENT
The reagent suggested here is essentially the same as that employed by Karl Fischer ( 5 ) , except that methyl Cellosolve is substituted for methanol and the quantity of solvent used is decreased so that the effective strength of the reagent is approximately three times as great. The molar ratio of iodine to pyridine to sulfur dioxide is maintained a t 1:10:3. It is essential that the ingredients of the reagent be pure and free of water. However, commercial methyl Cellosolve (Carbide & Carbon Chemicals Co.) and C.P. pyridine (J. T. Baker Chemical Co.) can generally be used without further purification. If the water content of the methyl Cellosolveis appreciably greater than O.l%, it may be decreased by distilling off about 5% through a small column and using the remaining 95%, but this has seldom been necessary.
For each liter of solution, dissolve 133 grams of C.P. iodine in 425 ml. of C.P. pyridine in a dry, glass-stoppered bottle, add 425 ml. of anhydrous methyl Cellosolve, and cool in an ice water bath.
Methanal Reagent (Inltial Strength 6 . 0 mg.d water per ml)
e
ci & *
4Q-
E
50-
L
70 0
I 10
I
LO
I
I
I
10
40
50
Time of Standing. days
Figure 1.
Stability of Karl Fischer reagent prepared with methanol and methyl Cellosolve
The stability of both the methyl Cellosolve reagent and the methanol reagent is improved by decreased concentration. Stability measurements on more dilute reagents, prepared with larger amounts of alcoholic solvent than recommended above, are given in Table I. In these tests, the water contents of the components of both reagents were identical, such that the theoretical strength of each reagent was calculated to be 4.3 mg. per ml., based on the weight of iodine added and allowing for the known water contents of the solvents employed. The results clearly show the greater stability and more nearly theoretical strength of the methyl Cellosolve reagent. Although the more dilute reagent is somewhat more stable than the 6 mg. per ml. reagent described above (compare Table I with Figure l ) , t h e stronger reagent tends to give sharper end points and readily permits the titration of as much as 0.2 to 0.3 gram of lyater in samples and in standardizations against weighed amounts of water. However, the choice of reagent strength is largely dependent upon personal preference and the particular application for which the reagent is to be used.
Table I. Stability of Dilute Fischer Reagent Prepared with Methanol and Methyl Cellosolve Methanol Reagent
Methyl Cellosolve Reagent
%
%
Age, % of drop from Yo of drop from Days hlg./ml. theory0 first value Mg./ml. theorya first value 2.84 66 .. 3.63 84 0.3 3.54 63 3 5 2.69 82 2.66 82 6 62 6 3.52 10 11 3.45 80 2.53 59 79 14 2.43 14 57 3.39 20 2.33 3.35 54 78 18 51 30 2.21 22 3.31 77 0. Theoretical strength of each reagent = 4.3 mg./ml.
V O L U M E 27, NO. 3, M A R C H 1 9 5 5 Mitchell and Smith (8) have found it advisable to prepare a very stable stock soluton of iodine, pyridine, and methanol, and to add the sulfur dioxide only a few days prior to use; they further recommend that the mixture be prepared in small batches-e.g., 3 liters a t a time. However, owing to the greater stability of the reagent prepared with methyl Cellosolve, it is practical to prepare 9-liter quantities of finished reagent. This reagent usually decreases in strength over a period of several months only to the extent of 1 mg. of water per ml. of reagent. Because of the increaEed stability of the reagent, it need be standardized only once during an 8-hour period of use for highest accuracy demands. Moreover, the reagent prepared with methyl Cellosolve permits the elimination of methanol from the titration mixture; methanol is sometimes undesirable because of side reactions with components of the sample. BURET ASSEMBLY
While several commercial, all-glass, automatic buret assemblies are suitable for storing and dispensing Fischer reagent, the specially constructed system illustrated in Figure 2 is suggested for long, continuous service. This asPembly consists of a 9-liter serum bottle, a modified 50-ml. precision buret, and connecting glass tubings fitted with spherical joints, constructed to afford ease of assembly and considerable flexibility. 911 stopcocks in the connecting lines are spring-loaded with light tension springs in order to minimize leakage; a large drying tube containing Indicating Drierite affords adequate protection against atmospheric moisture. Sisco 300 lubricant (Swedish Iron and Steel Co., 17 Battery Place, S e w York, S.Y . ) and Apiezon iX (Shell are satisfactory as lubricants for Oil Co., New York, K. I-.), the ground-glass surfaces.
Figure
2.
All-glass, siphon-type buret assembly
Modification of t h e buret involves the addition of a spherical joint to the top of the buret and addition of a side arm below the 50-ml. graduation of the buret; the side arm also terminates in a spherical joint. I n this laboratory it has been found convenient to mount burets by means of two spring-loaded clamps, one above and one below the graduated portion of the buret. The lower one is clamped to the buret in the area below the graduation but above .the side arm; to accomplish this without modification of the buret requires that the clamps be trimmed to avoid masking the .extreme lower portion of the graduations. The clamps in turn .are fastened to horizontal metal strips which extend over the
451 Table 11. Effects of Solvents on the Determination of Water in Ketones with Fischer Reagent (Visual end point procedure) Solventa Titration R ~ Sample ReTemp., ery, Analyzed agent Sample o C. % 1.17% water in R I M 25 130 acetone Ihz Rf 0 107 M P 25 104 M P 0 98 hIC P 0 101 G-P G-P
methyl ketone
P
25 0 25 0 25 0 0
101 100 115 105 103 101 98
G-P G-P
25 0
101 100
M M
1.15% water in
ethyl
P P
1IC hIC
Kature ~ of ~ E n d~PointColor change Stability Good Fades very rapidlyGood Fades rapidly Poor Fades Poor Fades sloivly Very Stable poor Good Fairly stable Good Stable Good Fades rapidly Good Fades Poor Fades slowly Poor Fairly stable Very Stable poor Good Fairly stable Good Stable
a 1\1 = methanol, M C = methyl Cellosolve, P = pyridine, G-P = ethylene glycol T pyridine (4: 1).
length of the titration stand. Behind each buret and fastcned to the horizontal metal strips there is a vertical strip of white plastic or metal with a white ename! finish to aid in reading the buret. TITRATION SOLVENT
rllthough methanol usually is a highly satisfactory solvent for the sample, as well as for the reaction products, in the titration of qyater with Fischer reagent, it has certain disadvantages. By a study of several sample solvents it was found that a mixture of ethylene glycol and pyridine (4 to 1, by volume) is satisfactory and widely applicable. Pyridine reduces the viscosity of the ethylene glycol to a considerable extent; it also has good solvent properties and in some cases avoids side reactions which may otherwise occur. When methanol is used as a solvent for samples containing carbonyl compounds it is necessary to convert these compounds to the relatively inert cyanohydrins (IO) in order to avoid interfering water-producing reactions between methanol and the carbonyl compound To determine water directly in the presence of ketones. other investigators have recommended the use of a reagent high in pyridine content (IO,1 2 ) or an excess of pyridine instead of methanol as solvent for the sample (IO). Results obtained by these modifications are reasonably accurate, but the end points still fade slowly and are rather indistinct. When the glycol-pyridine solvent mixture is employed in conjunction with Fischer reagent prepared with methyl Cellosolve, the titration can be performed in the presence of free ketones without interfering side reactions (provided the titration mixture is maintained near 0' C. throughout the titration), and the color change a t the end point is good. Results with various solvents in the presence of ketones are shown in Table I1 Large amounts of the lower aldehydes (except formaldehyde) interfere even when glycol-pyridine sample solvent a t 0" C is employed with the methyl Cellosolvereagent. Such samples must first be reacted with hydrogen cyanide to form the cyanohydrin. (IO). However, the interference from aromatic aldehydes and from CSand higher aliphatic aldehydes is slight when less than 1 or 2 grams of such aldehydes in 10 ml. of glycol-pryidine a t 0" C. are titrated with Fischer reagent prepared with methyl Cellosolve. Although most organic acids do not cause serious errors in water titrations by virtue of the relatively slow reaction between the acid and methanol, possible interference by acids is also eliminated by means of the reagent and solvent described. If ketones and organic acids are known to be absent, the titration may be carried out a t room temperature, and methanol or other suitable solvents for the sample may be employed as the titration solvent.
ANALYTICAL CHEMISTRY
452 Table 111. Solubility of Water in Hydrocarbons Determined by Means of Fischer Reagent (Visual end point procedure, using 15 ml. of ethylene glycol as the lower phase a n d side-well flask) Toluene" Diisobutylenea Propylene Tetramers Temp., Sample Water, Sample Water, Sample Water, C. size, g. wt. 3 '% size, g. wt. 7c size, g. wt. R 0 257 0.0293 152 0.0052 148 0.0091
a
267
0.0293
25
..
..
27
172 176
0.0605 0.0599
50
80 81
0.113 0.114
75
62 62 63 64 64 69
0.239 0.243 0.246 0.247 0.231 0.231
157 170
0.0053 0.0051
156
0.0099
149 200
0.0149 0.0154
,.
..
,
..
154 161
0.0176 0.0176
163 200
0.0388 0,0384
123 127
0.0281 0.0287
110 115 154
0.079 0.078 0.081
.
75 81
0 072 0.068
Saturated with water a t temperature indicated
ELECTROMETRIC DEAD-STOP TITRATION
In the reverse dead-stop method for detecting the end point in the direct titration of water with Fischer reagent, temporary drifting deflections of the galvanometer are frequently observed before the true end point when methanol is employed as the sample solvent. To avoid this difficulty, the addition of excess Fischer reagent and back-titration with standard water in methanol solution to the dead-stop end point is frequentlyrecommended (9, 1 2 ) . Modifications of the dead-stop apparatus, employing increased potential across t h e electrodes ( 3 ) or increased current (3),have been reported to be satisfactory for the direct titration with Fischer reagent, and an automatic titration apparatus incorporating a timer mechanism which distinguishes between transitory currents produced by temporary excesses of unconsumed Fischer reagent and the true end point has been described (6) and marketed. Although these instruments appear to be satisfactory for most direct titrations, a further decrease in tendency for temporary currents in advance of the true end point can be obtained by the use of the glycol-pryridine solvent mixture and the modified Fischer reagent described above. Under these conditions, the true end point can be quickly determined by direct titration, even when a dead-stop apparatus with low potential (20 mv.) across the electrodes and a low current (10 pa.) a t the end point is used. Apparently, a slightly larger concentration of Fischer reagent (and consequently larger concentrations of iodine and sulfur dioxide resulting in more rapid reaction with water) is required to depolarize the electrodes in glycol-pyridine solution than in methanol. The sensitivity of the end point is slightly lower in glycol-pyridine than in methanol, but is adequate for nearly all analytical purposes, giving more precise results than the visual end point procedure. An alternative solvent mixture, consisting of 1M sulfur dioxide and 1144 pyridine in methanol, also permits direct titration with Fischer reagent t o the dead-stop end point when the presence of methanol is not objectionable because of side reactions. The use of this solution as the titration solvent gives just as high endpoint sensitivity as the use of methanol alone and decreases the tendency for temporary, drifting deflections before the end point by increasing the rate of the reaction with mater.
A convenient titration vessel for reverse dead-stop titrations is shown in Figure 3 (left) (this apparatus is available from Rankin Glassblowing Co., 3920 Franklin Canyon Road, blartinez, Calif.), An interchangeable platinum electrode pair is inserted into a 250-ml. volumetric flask through a side arm cnrrying a standardtaper ground-glass joint. A volumetric flask is used because the long, narrow neck provides a gradient between the dry air in the flask and the moist atmosphere outside. When the moisture content of the atmosphere is not too high, the flask also protects
the buret tip while it is inserted in the neck. As a precaution, any moisture which collects on the buret tip n-hi!e it is exposed to the atmosphere during a titration should be frequently removed with a dry cloth or tissue by wiping with a domnn-ard motion. A flexible closure, such as a rubber dam with a hole for introducing the buret tip, between the neck of the flask and the buret is desirable in humid climateP. The flask is swirled by hand during titration. In laboratories nhich employ Fischer reagent for various purposes-for example, determination of organic functional groups by means of water-producing or water-consuming reactions-these titration vessels are more convenient than stationary titration vessels in permanent setups because they can be used as the containers in which preliminary reactions are carried out in water baths, ovens, etc., and they permit simultaneous preparation of a number of samples. One electrode pair is sufficient for many flask?, as the side arm can be kept closed with a standard-taper glass plug until just before the titration step.
For the highest accuracy, Fischer reagent should be standardized in the same manner (visual or dead-stop end point procedure) as it is to be used. The reverse dead-stop technique has been found to yield a standardization value approximately 1% higher than the value by t h e visual end-point procedure. Choice of titration solvent appears to make very little, if any, difference in standardization values, provided the reagent (or the solvent) contains a reactive alcohol TITRATION IN TWO-PHASE SYSTEM
For the determination of water in materials such as hydrocarbons which, because of ]OK water content, require a large sample, a procedure whereby all of the nater in the sample is concentrated in a relatively small vdume is obviously desirable. Solvent systems which give a large quantity of homogeneous solutions ( 1 1 ) do not yield the highest accuracy because of the indefinite nature of the end point obtained in a highly diluted titration mixture. Extraction of the Tvater from the sample with methanol or ethylene glycol (11)is generally satisfactory. Ethylene glycol (without pyridine) is to be preferred because it is less miscible with hydrocarbons than is methanol and is an efficient water-extracting agent. A marked improvement in resulta and greater simplicity are attained by performing the extraction and titration in the same ves5el.
2 cm.
H
Figure 3.
Titration flasks
L e f t . Dead-stop titration flask R i g h t . Side-well titration flask
For this purpose, the special titration flask shown in Figure 3 (right) has been employed (this apparatus is available from Rankin Glassblov-ing Co., 3920 Franklin Canyon Road, Martinez, Calif.). This flask is made from a borosilicate glass 500-ml. volumetric flask by adding a side well of 20-ml. capacity as shown in the figure. T h e iodine color is retained in the lower (glycol) phase and is easily observed in the side well. The glycol phase is prc-
V O L U M E 27, NO, 3, M A R C H 1 9 . 5 5 titrated with Fischer reagent to the end-point color of a matching
portion of the neck of the &sk does not become wet with gl$col. The mixture is then titrated with Fischer reagent until the loa er phase is near the end point color. The flask is again stoppered and shaken as before. The phases are allowed to separate and the titration is then completed with small increments of titrant until the color of the lower phase, as observed in the side well, matches the original color standard. When dense or viscous oils fraction before are analyzed, the addition of a light the pretitration of the glycol phase assists in the separation of phases. The application of this procedure for the determination of solubility of water in hydrocarbons is shown in Table 111. Replicate determinations indicated excellent precision of the method, except for the experiments a t 75' C., where difficulties in sampling caused somewhat poorer precision. The reverse dead-stop endpoint procedure also can be employed, using the dead-stop titration vessel (Figure 3, left). As long as the electrodes are immersed in the lower (glycol) phase, the end point can be readily determined when the flask is gently swirled. CONCLUSIONS
The usefulness of the reagent, solvents, and apparatus described has been demonstrated in the laboratories of the authors and others who have adopted several of these techniques but have
453 not included complete details in their publications (1, L ) . It
ACKR-OWLEDGMENT
The authors are indebted to W. B. M i l k a n for the design of the all-glass, siphon-type buret assembly and to R. H. Smith for Some of the experimental mrork involved, The Use of glycolWas suggested by H. R. Mcpyridine mixture as Combie, of Shell Chemical CO., Pittsburg, Calif. LITERATURE CITED (1) Brochmann-Hanssen, E., and Pong, P., J. Am. Pharm. d s s o c . , Sci. E d . , 41, 177 (1952). (2) Carter, R. J., and Williamson, L., Analyst, 70, 369 (1945). (3) Cornish, G. R., Plastics (London), 10, 99 (1946). (4) Davis, F. E., Kenyon, K., and Kirk, J., Science, 118, 276 (1953). (5) Fischer, K., Angew. Cheni., 48, 394 (1935). (6) Frediani, H. A., AXAL.C H m f . , 24, 1126 (1952). (7) Mitchell, J., Jr., and Smith, D. Rl., "Aquametry," p. 42, Interscience, Xew York, 1948. ( 8 ) Ibid.? p. 65. (9) I b i d . , p. 86. (10) I b i d . , pp. 146-54. (11) I b i d . , pp. 122, 162-8. (12) Jt7ernimont,G., and Hopkinson, F. J., ISD. ENG.CHEM.,ANAL. ED., 15, 272 (1943).
RECEIVED July
16, 1954.
dccepted October 22, 1954.
Apparatus for the Pyrohydrolytic Determination of Fluoride and Other Halides C. D. SUSANO, J. C. WHITE, and 1. E. LEE, JR. Analytical Chemistry Division, O a k Ridge N a t i o n a l Laboratory, O a k Ridge, Tenn.
A n apparatus for the pyrohydrolytic determination of fluoride and other halides is constructed entirely of nickel and stainless steel rather than platinum and quartz. The apparatus is economical, compact, and easily manipulated. .4pproximately 5000 determinations were made in 22 months before replacement of the reactor tube was necessary. The tube had been maintained at 1000" C. for approximately 3500 hours with daily cooling and reheating.
T
HE method of pyrohydrolysis for the determination
of fluorides was first exploited by Warf and coworkers ( I ) , who developed a method for the determination of fluoride in which steam is passed over a heated sample in a platinum apparatus. The volatile products of the pyrohydrolysis are then condensed and titrated: MF2 H20 -+ 110 2HF This reaction, in the case of many fluorides, is slow and is usually conducted a t temperatures of the order of 1000" C. This high temperature and the known corrosive action of solutions of hydrochloric, hydrofluoric, and hydrobromic acids on metals dictates the use of extremely resistant structural materials. Rarf and coworkers (1) used platinum and quartz in the fabrication of their apparatus. This report concerns the use of nickel for this purpose. I t was desired that the apparatus be constructed of material of similar durability to platinum and require less initial expenditure. Xickel n-as chosen a8 the metal most likely to meet these requirements. I n order to reduce the cost of the apparatus still further, stainless steel was to be used wherever possible. In the
+
+
design of the apparatus, prime consideration was given to these four factors: cost, durability, ease of manipulation, and compactness. DESCRIPTIO3 OF APPAR4TUS
The apparatus, in its final design, is shown in Figure 1. Ita over-all dimensions are 28 inches high, 21 inches deep, and 15 inches wide. Thus, it is possible to place two sets of apparatus in a 6-foot fume hood and still have ample space for titration and any other necessary operation. Steam is generated in a 1-liter flask, which is heated by an electric hot plate. The steam is passed through a line made of 1-inch (inside diameter) 316 stainless steel, which is heated by a 4 2 0 - ~ a t t5-inch , furnace operated a t 1000" C. A ball joint is used to connect the steam line to the reaction tube, which is made of nickel. The ball joint was fabricated by the machine shop and was patterned after the familiar glass ball joints. Fabrication was from nickel metal. Pertinent dimensions for the apparatus shown are: male joint, lQ/ls-inch outside diameter in TT idth; female, lj/lG-inch outside diameter broadening to 15/&ch outside diameter. The seal is made tight by the weight of the reactor tube. The reaction tube is heated by a 9-inch, 580-watt furnace. The condenser jacket is made of 316 stainless steel, as is the handle. The joint is made leakproof by means of the position and aeight of the handle M hich is 7.25 inches in over-all length, 4 inches of which is a I-inch bar and 3.25 inches of 0.25-inch 316 stainless steel rod. The handle Ti-eighs approximately 1.4 pounds. A photograph of the ball joints and ball joint loclr is sholvn in Figure 2. The apparatus is positioned so that the handle of the door lock is directlv in front of the operator. Thus, the receiver vessel for the condensate is located directly to the rear. This position is the most advantageous, a s it makes for simplicity in ihtroducing the sample into the tube and permits a full vien- of the ball joint and thus virtually eliminates any inadvertent loss of samples due to escape of hydrolysis products through an improper fitting of the loclr. The apparatus was designed t o use steam either generated by a water boiler or obtained from the plant steam line A condensing
