Determination of High Percentages of Copper - Analytical Chemistry

---

ANALYTICAL CHEMISTRY

972 Table IV. Titration of Iodide with Permanganate (Iodine Monochloride End Point) in Presence of Cyanide and Sulfate Expt. No.

Permanganate (0.1061~ N) Used, hll.

1 2 3 4 5

44.05 44.03 43.85 44.07 44.18

Potassium Iodide, G r a m Taken Found

0.3881 0.3879 0.3862 0.3883 0.3891

0.3880 0.3880 0.3863 0.3882 0.3892

results of these titrations are shown in Table 111, (Experiments 3, a, b, c, and d) and indicate that the manganese is without noticeable effect. Weighed samples of potassium iodide were titrated with permanganate after addition of amounts of cyanide and sulfate equivalent to those formed in the titration of thiocyanate. The results, tabulated in Table IF7, show no evident interference. The procedure was that described above. These data show that under the above conditions iodide can be accurately titrated by permanganate to the iodine monochloride end point, and that Lhe presence of cyanide and sulfate is without effect. The reason for the error in the titration of thiocyanate viith permanganate by the above procedure is therefore still uncertain. Titration with Ceric Sulfate. Preliminary experiments indicated, and the data shown in Table I11 (Experiments 6, a, b, and c) confirm that the titration with ceric sulfate is not stoichiometric; negative errors are again observed. Titration of iodide solutions with ceric solutions with and without addition of cy-

anide and sulfate again indicated that, as with iodate and permanganate, these constituents were without effect. The titrations with ceric sulfate were more critical on the final hydrochloric acid concentration than were these with iodate or permanganate, and in all cases required longer for attainment of apparent stability. Negative errors of around 10% were obtained when the fmal acid concentration was 3 F. ACKNOWLEDGMENT

The authors are indebted to David Pressman for a series of preliminary experiments carried out in 1934-35, to Miyoshi Ikawa for the experiments on the determinations of copper, and to Dale hIeier for helpful comments on the manuscript. LITERATURE CITED

Jamieson, J . Am. Chem. Soc., 40, 1036 (1918). Jamieson, J.Ind. Eng. Chem., 11,296 (1919). Jamieson, Levy, and Wells, J . Am. Chem. SOC.,30,760 (1908). Kolthoff, I. M., and Furman, N. H., “Volumetric Analysis,” Vol. 11,p. 300, New York, John Wiley & Sons, 1929. Kolthoff, I. M., and Lingane, J. J., J. Am. Chem. SOC.,57, 2126 (1935). Lang, Rudolph, 2.anorg. allgem. Chem., 142, 289 (1925). Oesper, “Newer Methods of Volumetric Chemical Analysis,” p. 85, Sew York, D. Van Nostrand Co., 1938. “Scott’s Standard Methods of Chemical Analysis,” N. H. Furman, ed., 5th ed., Vol. l , p. 375, New York, D. Van Sostrand Co., 1939. Stokes and Cain, Bull. U.S. Bur. Standards, 3,157 (1907). Swift, J. A m . Chem. SOC.,52,894 (1930). I

RECEIVED June 18, 1948. Contribution 1301 from the Gates a n d Crellio Laboratories of Chemistry, California Institute of Technology,

Determination of High Percentages of Copper With a Beckman Spectrophotometer ROBERT BASTIAN Sylvania Electric Products, Inc., Kew Gardens, N. Y . A spectrophotometric method for the determination of high percentages of copper with an accuracy and precision of 1 to 3 parts per thousand utilizes the color of the cupric ion contained in 10% perchloric acid solution. The stated precision is obtained by working with copper solutions possessing extinctions greater than 2.0. Such high densities are not read in the normal way, which would involve a large percentage error, but a differential method is used. The optical density scale is set for zero with a solution

T

HE colorimetric or spectrophotometric determination of major constituents has been delayed because of the limit of accuracy imposed by conventional methods. If one proceeds in the normal way, it can be shown that the precision of measurement of the intensity of a given color does not increase indefinitely with increasing concentration of the color. Rather, for colors that obey Beer’s law, the rnaximum precision is obtained when the extinction of the given colored solution is 0.434, corresponding to a transmittancy of 36.8%. Actually, the error remains nearly constant in the range of 20 to 60% transmittancy ( 5 ) ,and above or below this region it increases rapidly. The error in the determination will depend upon the accuracy with which the extinction measurement can be made. Assuming an error of 0.1 division in the scale reading (based on 100 scale divisions) a t an optical density of 0.434, the concentration error is 0.27%. In practice, however, this is difficult to achieve. Sandell says ( 5 ) , “The error in setting the microammeter needle

containing 1.5000 grams oficopper per 100 ml. rather than with distilled water. This is accomplished by working at a much larger slit width than is normally employed. Higher concentrations of copper are then read against this zero. The commonly occurring colored metal ions, cobalt, iron, chromium, and nickel, in concentrations up to 49‘0 each, do not interfere with the determination. This method has been applied to a lead brass, a phosphor bronze, and a synthetic sample.

at 100 and making the transmittancy or absorption reading should not exceed 0.2 scale division, which represents an error of 0.6% in concentration at 50% transmittancy, if it is assumed that the standard curve is not in error.” At the optimum point. this scale error corresponds to a concentration error of 0.54%, which is too high for the determination of major constituents. Moreover, this treatment assumes that no error is present in the standard curve; in actual practice, a concentration error of 1% is not unusual. Methods for decreasing this error have been described. Ringbom (4) has shown that much better results can be obtained if an unknown concentration of a given color contained in one cuvette is matched by adding a standard solution to a second cuvette containing the color-forming reagents only. This is done with the galvanometer a t full sensitivity. Under such conditions, which amount t o a colorimetric titration, an error of as little as 0.15% at a transmittancy of 50% is indicated (6).

973

V O L U M E 2 1 , NO. 8, A U G U S T 1 9 4 9 I

monochromaticity and yet a reasonably high concentration of copper. A concentration of 1.5000 grams of copper per 100 ml. was finally selected for the zero point, and a slit width of 0.34 mm. was employed. This corresponds to a band width of 25.5 millimicrons. Attempts to use the slit wide open (2.0 mm.) and 3.0grams of copper per 100 ml. yielded a curved rather than a straight line.

CU----l

SELECTION OF WAVE LENGTH

0 WAVE LENGTH, IN mu Figure 1. Extinction-Wave L e n g t h Curves for bletals in 10% Perchloric Acid Solution Copper concentration, 0.25 gram per 100 ml. per 100 ml.

Other metale, 1.0 gram

Kortum (2) describes a method for comparing a color with a similar standard and obtaining a n error of 0.2%. Rabinovitch and Wood (5) have been able to distinguish differences in concentration of 0.002% using an arc source and special apparatus. The above methods indicate that the problem can be solved but they are too inconvenient or tedious for general use. The method described below employs a model DU Beckman spectrophotometer and is almost as simple to use as conventional methods. For doing this work, copper was employed. The blue color of the cupric ion itself contained in a 10% perchloric acid solution was selected because of its expected stability, and because its spectral characteristics are such that interference from other colored metal ions is low.

All the work was done using 1.000-cm. Corex cells a t a wave length of 870 millimicrons. The selection of the wave length is based upon the absorption data shown in Figure 1 for the commonly occurring colored metal ions plotted against a distilled water zero with the instrument used in normal fashion. The slit widths varied from about 0.02 to 0.06 mm. The concentration of copper shown is 0.25 gram per 100 ml., the other metals 1.0 gram per 100 ml. MATERIALS

The sources of these metals were 99.857, c. P. iron wire, c . P. nickel, c. P. cobalt chloride, c . P. chromic acid, and 99.99 7, oxygen-free, high conductivity copper. The materials were dissolved in nitric acid, treated with 10 ml. of 607, perchloric acid, taken to fumes, and fumed for a few minutes. Such treatment leaves iron and chromium in oxidized states. After cooling, 50 ml. of water were added, and the solutions were boiled to remove chlorine, cooled, and diluted to 100.0 ml. in volumetric flasks.

+

Xickel is the only metal besides copper that shows appreciable absorption in the region shown on the graph. Because its absorption is nearly a t a minimum a t 870 millimicrons while the copper absorption is still high, this wave length was selected for the subsequent work.

PRINCIPLE O F METHOD

Assurrie that a concentration of 0.2 gram of copper per 100 ml. can be determined by the normal method with an accuracy of 1%. If the color obeys Beer’s law at higher concentrations, a difference in concentration of 0.2 gram of copper per 100 ml. will always give the same difference in extinction. Yet if an attempt is made to take ten times the concentration of copper in the hope of getting ten times as much accuracy, the density reading will be off the scale; and even if such a high density could be read, the percentage error in making such a reading would be greater than the error in reading the lower concentration. This is all based on the assumption that the optical density scale is set for zero (or the transmittancy scale for 100%) using distilled water. Sumose, however, instead of using distilled water for this purpose, asolution containing 1.8 grams of copper per 100 ml. were used. A solution containing 2.0 grams of copper per 100 ml. read against this standard should then give the same scale reading that a concentration of 0.2 gram of copper per 100 ml. gives against distilled water. If it is possible to determine the 0.2-gram difference a t the higher concentration to an accuracy of ly,, the total concentration is automatically determined to an accuracy of 0.170. Such a process could be carried farther. If 19.8 grams of copper per 100 ml. were taken for the zero setting, and all the above conditions still applied, an accuracy of 0.017, could be obtained. The above treatment assumes that there is no loss in accuracy in measuring differences in concentrations a t higher concentrations, provided Beer’s law is obeyed. KOcomparative experimental data are given to prove this in this paper. However, it is clear that if the error in determining such differences does not increase by a factor of 10 or more as the concentration is increased ten times, the method will prove advantageous. A limitation to the method is that some provision must be made for getting enough light through the high concentration of colored material to make the necessary zero setting. On the Beckman instrument this is accomplished by working a t a much wider slit width than is normally employed. This naturally increases the band width of light which emerges. Artually, a compromise was reached so as to provide reasonable

PREPARATION O F STANDARD CURVE

To prepare the standard curve, 1.5000 grams of oxygen-free, high conductivity copper and varying greater quantities were weighed and placed in 250-ml. beakers. (In all this work the 1.5000-gram standards varied in weight from 1.4995 to 1.5005 grams. Corrections were applied for the slight deviations.) After solution in 20 ml. of 1 to 1 nitric acid, 10 ml. of 60% perchloric acid were added and the samples were taken to heavy perchloric acid fumes. After fuming, the samples were cooled, diluted with 50 ml. of water, and boiled for 2 minutes to remove chlorine. The samples were cooled to room temperature, transferred to 100-ml. Exax volumetric flasks, and diluted to the mark.

Cu METAL G I 100 MI. Figure 2. S t a n d a r d Copper Curie 870 mp, slit at 0.34 m m .

1.000-cm. Corer cells

The 1.5000-gram standard was placed in two adjacent cells, the slit opened to 0.34 mm., and the wave length set a t 870 millimicrons. The dark current adjustment was made and then the galvanometer zeroed using the sensitivity knob alone. Under these conditions the knob was used a t from 2 to 4 turns from the clockwise end, thus giving satisfactory sensitivity. After this the companion cell was slid into position and if any difference in reading was observed (generally 0.002 to 0.005 on the density scale), a corresponding correction was made on the solktions measured in that cell. The standard solutions were then read in the second cell, always against the 1.5000-graq

ANALYTICAL CHEMISTRY

974 -

Table I. Effect of Colored Ions [Solutions contalned 0.08 gram each of iron, nickel, chromium, and cobalt, plus indicated amount of copper per 100 nil.) Cu Taken, Cu Found. Error, Grains Grams Parts per Thousand +1.4 1.6618 1.6495 -2.1 1.6443 1.6478 +2.2 1,6503 1.6460 -2.3 1.6403 ,6443 AV. $ 6471 1.6467 12 0

Error = -0.2 part per thousand Computed error = + 2 . 4 parts per thousand \laximum error = -2.6 parts per thousand

standard as a zeio point and adjusting for zero n-ith the sensitivity control only. Figure 2 (squares) shows the results obtained. The following day a new 1.5000-gram standard was prepared and the triangular points were obtained against this standard. The deviation of the points from the line in parts per thousand ai concentration are indicated on the graph. The maximum deviation is t 3 . 9 parts per thousand and the average deviation

1.5 parts per thousand, including the zero point in the average. Assuming that the increase in accuracy over the normal method is directly proportional to the increase in copper concentration over that required to give an extinction of 0.434 @gainst a distilled water zero, this method should yield about six times the normal accuracy over the range shown on the graph. E F F E C T O F O T H E R COLORED IONS

To test the effect of other colored ions, four samples of copper were analyzed, to which 0.08 gram each of iron, nickel, cobalt, and chromium were simultaneously added. On the weight of copper taken these amounts represent about 497, each of the metals. These samples were treated exactly like the standards and read against a freshly prepared 1.5000-gram copper standard using Figure 2. The values are given in Table I. Computing the interference from the other ions on the assumption that the absorption a t the wider slit is about the same as ,under the conditions in Figure 1, and assuming that the colors obey Beer’s law and are additive, the result should be 2.4 parts per thousand high. The maximum error under these conditions would be -2.6 parts per thousand. From the above, it may be concluded that nickel, chromium, iron, and cobalt in amounts up to 4% each do not appreciably interfere. Inasmuch as these concentrations are well above the normal amounts contained in many copper-base alloys, the method should be excellent for such materials. Any tin present in such materials will be precipitated as metastannic acid, and this must be filtered off. If appreciable amounts of tin are present, the precipitate ail1 retain copper, but this situation arises in the standard electrolytic method as well. R E S U L T S ON COPPER-BASE ALLOYS

Table I1 gives the results obtained on two copper-base alloys.

. Accurately weighed 2.7-gram samples were employed for sample B. Because a trace of tin was present, causing a faint turbidity,

the samples were diluted to the mark and then filtered into the absorption cells before reading. I n the case of sample A, accurately weighed 2-gram samples were dissolved in 20 ml. of 1 to 1 nitric acid. After decomposition, 50 ml. of hot water were added, the solutions were filtered, and the residue was washed twice with hot water. The filter papers were placed in the original beakers and treated with 20 ml. of nitric acid and 10 ml. of perchloric acid. After heating to fumes to destroy the paper, the solutions mere cooled somewhat, and a little water was added, followed by 20 ml. of 48% hydrobromic acid (1). The solutions were heated to fumes @gaint o remove tin. This was repeated three times. The small residue of antimony and tin remaining was then filtered off and washed thoroughly and the filtrate was combined with the original filtrate. This was boiled to fumes and diluted to 100 ml. in the volumetric flask as in the other methods. The solution

was very faintly turbid (probably owing to the incomplete removal of tin, antimony, or silica) and therefore was filtered into the cell before reading.

It is felt that the errors indicated here are higher than those which need occur under the best conditions. Because the original 1.5000-gram standards used in plotting the points were discarded, the author had to prepare fresh ones for subsequent analyses. Any error in the standard, therefore, was added to the sample error. In routine analysis it would be much better to preserve this standard and eliminate that source of error. In addition, the volumetric flasks were not calibrated. Some fluctuations due to this source are likely. (Cpon the suggestion of one of the reviewers, some time after this work was done, all thirteen 100ml. volumetric flasks possessed by the laboratory were calibrated. Barring the possibility of breakage, these Fould include all the flasks used. The maximum difference between any two flasks was 1.0 part per thousand and the average deviation from the mean 10.3part per thousand. I t is probable that the error introduced was between these two limits.)

Table 11. Determination of Copper i n Copper-Base Alloys Sample A. Bureau of Standards Phosphor Bronze 63b0 cu present, % Cu found, % 77.96 78.30 77.88 78.08 77.85 Av. 78.03 * 2 . 1 parts per thousand Error = f0.9 part per thousand

*

Sample B. Lead Brass Cu found, electrolytic, Cu found, % colorimetric, % 61.266 61.50 61.30 61.38 61.50 61.42 t 1 . 3 parts per thousand Error = 4-26 parts per thousand (taking electrolytic value as correct) Provisional analysis 77.96% Cu 9.35% Pb 9 78% Sn 0.71% Zn 0.54% Sb, 0.47% Fe, O.k4% P, 0.33% Ni, 0.17%’ S,‘O.12% s‘i, 0.05% Al: 0.04y0 Ag, and 0.01nL7, A b Anorox h a t e comidii’iion. 3.4% Pb, 61 t o 627, Cu, 0.13% Fe, traoe of Sn, remain1ier Zn. C Averag e of 3 determinations.

-~. 1

This paper is intended only to illustrate a method of approach in colorimetric work. The method described should be applicable to other metals and in cases where the colors are stable enough to the more sensitive color reactions. If very stable colors can be found, it should be possible to obtain very high precision on small samples by applying the method. A more intense source of light than is a t present contained on the Beckman would be very helpful because it would permit the use of even higher concentrations of color, a t the same time maintaining narrow slit widths. ACKNOWLEDGMENT

Many thanks are expressed to Richard Weberling for performing the electrolytic analysis of copper (sample B), to Frank Bassani and his drafting department for preparation of the graphs, and to M.J. Rafale and Xrs. -4.H. Childers for the;r assistance in checking over the manuscript. LITERATURE C I T E D

(1) Am. Soc. Testing Materials, “Methods of Chemical Analysis of Metals,” p. 186, 1946. (2) Korttlm, G., Angew. Chem., 50, 193 (1937). (3) Rabinovitch, E., and Wood, W., Trans. Faraday Soc., 32, 547

(1936). (4) Ringbom, A., 2.anal. Chem., 115, 332 (1938/39). ( 5 ) Sandell, E. B., “ColorimetricDetermination of Traces of Metals,” p. 53, New York, Interscience Publishers, 1944. (6) Idid., p. 54. RECEIVED September 17, 1948. Presented before the Division of Analytical and Micro Chemistry a t the 114th Meeting of t h e AVERICAXCHE~IICAL SOCIETY, St. Louis, Mo.