ANALYTICAL CHEMISTRY
1638 metal percentage may be calculated. If not, the oxides may be calculated as the predominating element, which, in most cases, is cerium. Alternate Procedure When Thorium Is Absent. Add 3 drops of bromophenol blue indicator to the prepared solution, and adjust to the blue color with ammonium hydroxide (1 to 4). Heat t o boiling, remove from the hot plate, and allow to stand for 5 minutes with occasional stirring. Filter on medium porosity (No. 40 Whatman) paper, and wash thoroughly with hot water. If zinc is present, wash well with 20 ml. of 1 to 1 ammonium hydroxide solution and hot distilled water. Remove and save the filtrate containing most of the rare earths. Dissolve the zirconium hydroxide, and any rare earths ahich may have been occluded, into the original beaker with 10 ml. of nitric acidperoxide solution and hot water. Boil down to approximately 25 ml., and save. T o the filtrate containing the rare earths, add 10 grams of ammonium chloride, and proceed as directed in the paragraph entitled, “Separation of the Rare Earth Elements.” Remove the rare earth oxides from the furnace, cool, and carefully u:ash the contents of the crucible into the beaker containing the zirconium. Heat on a hot plate until dissolution is complete. (An additional amount of hydrogen peroxide may be needed a t this point.) Follow ;,he procedure entitled, “Determination of Rare Earth Elements. EXPERIMEhTAL
Solutions of the rare earths were made by dissolving didymium, lanthanum, and mischmetal in hydrochloric acid. Didymium is an alloy produced from a rare earth mixture containing mainly
neodymium, praseodymium, and lanthanum. Mischmetal contains cerium, lanthanum, neodymium, and praseodymium. Standard solutions of thorium were prepared from thorium nitrate. These solutions were standardized by precipitating the rare earth or thorium oxalates and weighing the oxides after ignition. Standard solutions of magnesium were prepared by dissolving resublimed magnesium in hydrochloric acid and distilled water. Zirconium was added as a solution of the oxychloride. The data in Tables I to I11 were obtained by analyzing mixtures of the above standard solutions by the given procedures. LITERATURE CITED
(1) Hopkins, B. S., ”Chapters in the Chemistry of Less Familiar Elements,” T’ol. I, Chap. 6, p. 32, Champaign, Ill., Stipes
Publishing Co., 1939. (2) Leontis, T. E., J . M e t a l s , 4, 287 (1952). (3) Moeller, Therald, and Brantley, J. C., ANAL. CHEM.,22, 433
(1950). Kelson, K. E., and Strieter, F. P., T r a n s . Am. Foundrymen’s Assoc., 58,400 (1950). (5) Itodden, C. J., J . Research ,VatZ. Bur. Standards, 26, 557 (1941). (A) Venkataramaniah, M., Rao, C. L., and Rao, B.S.V.R., A n a l y s t 77, 103 (1952). (4)
R E C E I V Efor D review June 19, 1952.
Acwpted 4ugust 2 2 , 1952.
Improved Protective Colloid for use with Dichlorofluorescein Adsorption Indicator ROBERT B. DEAR’;, WILMER C. WISER, GEORGE E. MARTIN,
AND
DENNIS W. BARR’UM
University of Oregon, Eugene, Ore.
The authors have investigated 45 synthetic protective colloids for this titration and find that the best protective agent is a condensation product of ethylene oxide sold as Polyethylene cator is a standard procedure in most textbooks of analytical glycol 400 by the Carbide and Carbon Chemicals Corp. The chemistry. Unskilled students find it easier to recognize this principal components of this material have the formula HOend point than that of the classical blohr procedure. Most (CH&H20),H, where n varies from 7 to 12 with an average authors recommend the addition of 5 ml. of 1 or 2% dextrin as a value near 9. Results on other protective agents will be reported protective colloid t o keep the silver chloride precipitate in suselsewhere. Polyethylene glycol 400 is as good a suspending pension, although they do not refer t o the original literature for agent as dextrin or gum arabic, it is stable and not attacked by this procedure. microorganisms, and a 50% solution in water is a solvent for the indicator. Three to 5 drops of a 0.1% solution Table I. Titration of Chloride i n Pure Sodium Chloride (60,6670 C1) a n d i n of dichlorofluoroescein in 50% Polya hIixed Unknown ethylene glycol 400 provide optimum Method, Adsorption Indicator amounts of the indicator and the protecProtective Nohr Polyethylene Dextrin Dextrin Gun1 Agent Xone glycol A n arabic tive agent in a single solution which keeps 60.66 38.64“ 60.66 38.64 38.61 38.64 60.66 CI taken, %b 60.66 mdefinitely. This mixed indicator has 38.72 60 66 38 64 38.68 60.70 38.65 60.54 C l f o u n d , yc 60.63 +0.08 0.00 0.00 +0.04 +0.04 +0.01 - 0 12 -0.03 Deviation, been used in a small elementary class 0.024 0.174 0.029 0.067 0.039 0.024 0 051 0.095 30av, %b in quantitative analysis with results 17 6 17 18 11 17 3 No, of samDles 11 l-alue repor:ed by manua % of chloride in the mixed unknown determined by Xlohr titration. that compare favorably lyith those ohfacturer is 0.057, C1 lower and is believed t o be i n error. tained using dextrin. Because most protective agents interfere to some extent with the end point, the authors have tested in more detail the behavior of Kolthoff (1) seems to have been the first to recommend the Polyethylene glycol 400, ti\-o preparations of dextrin, and one use of dextrin as a protective colloid for absorption indicators, of gum arabic with dichlorofluurescein. The silver nitrate solualthough he did not recommend it for this system. I n fact, he tions were standardized gravimetrically and Mohr titrations were specifically pointed out ( 2 ) that protective colloids such as gelarun as controls. I n the first series samples of pure sodium tin interfere with the use of dichlorofluorescein as an adsorption chloride were weighed individually for each titration. I n the indicator. Several other protective colloids have also been resecond series a master solution of a commercial sample of sodium ported as interfering with the accuracy of this method ( 3 ) . chloride mixed with an inert diluent was prepared and 50-ml. Recently, comparative tests on four naturally occurring proaliquots were taken for analysis. The results are reported in tective colloids were reported ( 4 ) . Gum arabic was found to be Table I as the averages of the indicated number of replications. the most satisfactory and dextrin gave end points which were 1 Three times the standard deviation of the average (3gat)is also to 2 parts per thousand low. Because solutions of naturally given. The average f 3 u a , is a conservative confidence range occurring colloids are subject to attack by microorganisms and for the results. Deviations from the true value greater than f 3 fungi, these protective agents must be made up fresh for each uav are considered to be significant. day’s use. Polyethylene glycol ( 5 drops of a 50% solution containing 0.1% 1 Present address, Chemical Division, Borden Co., Bainbridge, S . Y .
HE determination of soluble chlorides by direct titration with
Tsilver nitrate using dichlorofluorescein as an adsorption indi-
~~
V O L U M E 24, NO. 10, O C T O B E R 1 9 5 2
1639
dye) gave results which are not significantly different from the true values. T h e two batches of dextrin, A and B (3 mil. of 2y0 solutions), gave results which differ from the true values by significant amounts but in opposite directions. This is not surprising, as the name “dextrin” covers a number of different types of starch degradation products and different dextrins may well have different properties. Gum arabic (3 ml. of 1% solution) gave results which were significantly high, although the differences were not greater than 2 parts per thousand. I t ir concluded that Polyethylene glyrol 400 is superior to other protective colloids for use with dichlorofluorescein from considerations of both accuracy and convenience. The use of
5 drops of a 50% solution in water containing 0.1% dichlorofluorescein is recommended. LITERATURE CITED
(1) Kolthoff. I. hf., J . Anal. Chem.. 71, 235 (1927). (2) Kolthoff, 1. M., Lauer, IT.31..and Sunder, C. J., .I. Am. Chem. Soc., 51,3273 (1929). (3) Santos-Ruiz, A., and Portillo, R., Anale. soc. espail. fis. qutm. (Madrid),39,91 (1940). (4)Stalser, R. F., Dillon, E. S., and Vosburgh, TV,C., ANAL.CHEW, 2 2 , 9 5 2 (1950). R E C E I V Efor D review J u n e 17, 1961. Accepted August 16, 1962. Presented a t the Pacific S o r t h a e s t Regional Xleeting, Seattle, Wash.. 1951.
Photometric Analysis Error W . 4. E. I\JcBKYDE, D e p a r t m e n t of C h e m i s t r y , University of Toronto, Toronto, Canada
i S AX effort t o show the regions of concentration which may
be
determined 15ith comparatively low relative analysi;: error, Ringborn (a), Alyres( I ) , and others have advocated an unusual method of representing photometric data-namely, a plot of (1-T) against log c. T h e region of steep slope on such a plot discloses the range of concentrations which may be determined with fair precision. T h e uriter feels that this type of working curve for transforming instrument leadings to concentrations suffers from certain practical drawbacks and, except for cases of gross deviation from Beer’s law, provides little information that cannot be obtained by other means. I n the first place it is much more difficult t o fit a skew-shaped curve through experimental points without sacrificing some accuracy than to draw a straight or nearly straight line: the Ringbom plot yields a working curve of the former sort whereas the conventional absorbancy-concentration plot is usually nearly a straight line. I t is also more difficult to interpolate concentration values on a logarithmic than on a linear scale. Furthermore, because many chemists routinely work with absorbancies in preference to transmittancies, several commercial in3truments have scales which &her read in, or are proportional to, absorbancies; accordingly it becomes highly artificial to convert these to absorpt,ancies (1- 2’). The purpose of the Ringboni plot is t o reveal regions of concent,ration where the relative analysis error is comparatively small, i a and yet, when 13eer’s law is strictly obeyed, this error function R known funct,ion of concentration. Thus, f o r d In c/dT t o be less than 5, t h e absorbancy must lie between 0.112 a n d 1.105 (transmittancy between 0.772 and 0.0785). There are various absorbing systems known t o analysts for which Beer’s law does not quite hold valid, and the deviation becomes more noticeable the higher the absorbancy. For such cases we may express the relationship h t w e e n absorbancy and concentration as
A
=
abc
+ bf(c)
where j is a correction term which changes in some waj. with c. The symbols here are given in accordance \&h recently adopted .4merican practice (3). B y differentiating t o obtain the expression for the error function we have
_-.
0.4343dT -~ ahdc + bdf/dc dc Tlog T abc bj abc bf
+
+
And if bf is small in comparison with abc, this becomes 0.4343dT -~ T log I’ dclc
dl’
-
0.4343 1’ log 2’ (1
&c c (1
also b r small compared H ith a , 30 that the error function does not differ greatly from the value< derived when Beer’s law holds exactly. For systems which fail to conform to the absorption law a t low absorbancies or show gross deviations therefrom, the foregoing derivation is not applicable, and information concerning analysis error is available only from the Ringboni plot. The analyst who is concerned with colorimetric determination of traces of constituents is concerned with the shape of the absorbancy-concentration curve and the error function when concentration is small. A glance a t many of the procedures employed for trace anal?& shows that they nearly all conform t o the absorption law or show only minor deviations in the region of low concentration. I n most of these cases the analjsis error function exceeds 5 for absorbancies less than 0.1 12 or therabouts. For smaller absorbancies, the relative error function hecomes approximately 1 / 1 - T (6). It is sometimes possible to increase the absorbancy to be measured by keeping the volume of the final solution, containing the colored constituent, to a minimum, and occasionally by increasing the depth of the absorbing vessel. The restriction in the latter case is thst, with the sccessories supplied with many commercial instruments for increasing the light path through the absorbing solution, the increase in the value of b is almost offset by dilution of the solution in order to fill the cell. It seems probable that very small amount3 of a substance cannot be determined with great precision in most cases, since the error on a relative basis of any measurement becomes large as the quantity t’obe measured is made small. I t is perhaps relevant to point out, however, that the value of dc/dT varies inversely as T , so that the absolute error in measuring c is actually slightly smaller as T approaches unity. These remarks are intended to offset the impression which appears to have been created that photometric analysis should never be extended to solutions whose absorbancies fall outside the range corresponding to a small error function, An extreme case of this att,itude came to the writer’s attention recently. He was asked to review a manuscript prior t o its publication in which the author, who was intere3ted in trace analyses performed with a spectrophotometer, made the suggestion that precision might be improved by adding to each sample a known amount of the constituent to be determined. The amount of added constituent was to be of such a size as to bring the absorbancy of the final solution into the region of low relative analysis error. A similar proposal has evidently been made before ( 2 , 4 ) . The writer endeavored t,o show by the following argument that this proposal would, if anything, increase the relative analysis error. I n this demonstration the absorption law was assumed to be applicable. Suppose that the original solution concentration was ci, absorbancy A , . Suppose that there is added to this the constituent to be determined until the new concentration is ca =
df /dc +
+ ip)
-7)
0.4343 d d /d T log T
(+)
If the preceding assumption is true, then dfldc will pro!)ably
