This old and well-known method of determining silver is, in bullion work, so far superior to the furnace-assay that it is looked upon with reverential awe by many, if not by most, users, and its ease of execution, with proper equipment, commends it highly where much commercial bullion work on silver is required.

The method is so thoroughly well known that a description of it here may seem to be unnecessary, but many thousands of determinations are annually made by it in the U. S. Mint, Service and its practice is there reduced to an art. This is necessary both for the sake of economy of time, and because high-grade work with it requires constant practice. It is useless for a person who uses it only occasionally to expect to attain high accuracy with it, without spending more time upon a determination than is warranted in a busy commercial laboratory.

It is always employed in the Mint service whenever possible, but its chief field is in the determination of silver in standard silver (ingots and coin), which is 900 fine in silver and 100 fine in copper, and I shall first briefly outline its use on this metal and then take up various points in detail, especially as related to the accuracy of the results obtained.

For this determination 1,115 mg. of standard metal is weighed and transferred to a glass-stoppered bottle, the metal is dissolved in nitric acid, and 100 cc. of a standard solution of NaCl run in. The bottle is then vigorously shaken and a measured portion of a decimal salt solution added, and again shaken, if necessary. These operations are repeated until the silver is precipitated.

Standard metal being 900 fine, 1,115 mg. should carry 1,003.5 mg. of silver, which is a convenient figure to work with. Legally, standard metal may vary 3 in fineness above or below 900. Practically, however, it seldom runs belows 898 or above 901. Therefore the variation in actual silver present above or below 1,003.5 mg. is not excessive. For testing the standard salt solution we weigh up 1,004 mg. of proof silver. For convenience and accuracy we have single weights of 1,115 mg. and 1,004 mg.

The bottles should be of perfectly white glass, carefully made and well annealed; 8 oz. is a convenient size often used; the bottles are most conveniently handled in a circular frame or basket holding 10. For dissolving the silver, both the amount and the strength of nitric acid used may vary considerably without apparent effect upon the accuracy of the results, 8 cc. at 1.30 specific gravity to 25 cc. at 1.20 specific gravity being allowable. Some operators heat the bottle to remove nitrous fumes, others do not.

The standard salt solution is designed to have 100 cc. precipitate exactly 1 g. or 1,000 mg. of silver, but it seldom “shows” this exact strength. The word “ shows ” is used instead of “ is ” because the equivalent of the solution depends upon other factors besides its composition.

In general, two proofs should be used in every set of 10 bottles, unless several sets are to be run in rapid succession, when one proof in each set may answer. Some operators run an independent proof occasionally during the day and omit the proofs in the sets, but this proceeding is objectionable.

The next step in the method is the addition of 100 cc. of the standard salt solution (often improperly called “ normal ” salt solution). This is a very simple operation, but it requires the utmost care and constant attention to details if high accuracy is desired in the results. Only a minute variation in the amount of solution added will make a serious variation in the fineness of silver shown.

The Stas pipette is the one universally used in the Mint service. It is a simple pipette open at both ends and securely mounted on a wall bracket. The upper end is drawn out to a fine opening and is provided with a collar-cup to catch the drip. The lower end is comparatively large and must have a free and smooth discharge. The lower end is connected through a removable rubber tube, provided with a pinch-cock, with an elevated tank containing the salt solution. No stop-cocks, floats, or graduations of any kind can be used on or in the pipette if rapid work is to lie done. It is also questionable if any such arrangement can be as accurate as the simple filling of the Stas pipette and letting its contents run directly into the silver brittle.

In operating the Stas pipette the rubber tubing is slipped over the lower end and the pinch-cock opened. As soon as the solution comes out of the upper end of the tube, it is closed by the first finger of the left hand and the pinch-cock closed. The operator must now be sure that there are no air bubbles in the pipette. If such should appear they must, be allowed to collect at the top of the pipette; the pinch-cock must be opened, the finger momentarily removed from the upper end of the pipette, and the pinch-cock closed again. When the pipette is full of solution and the pinch-cock closed, the rubber tube is withdrawn from the lower end of the pipette. This end must now be carefully examined to see that there is no surplus solution adhering to it or that the air has not commenced to ascend the tube.

If the lower end of the tube is in proper condition the silver bottle is now placed directly under it, the finger removed from the upper end and the solution allowed to flow into the bottle. The solution should flow out rapidly in a smooth, solid stream. Just as soon as the flow stops, the bottle must be removed from under the pipette. It is absolutely fatal to accuracy to attempt any adjustment of the drip of the pipette. There should be only just easy clearance between the bottom of the pipette and the top of the bottle.

The pipette is supposed to contain 100 cc. and on various accounts it is desirable that it should be fairly accurate, but it is not at all necessary that it should be absolutely accurate. The absolutely essential point about it is that it should, deliver exactly the same amount of solution to each one of the 10 bottles composing a set. The amount delivered may be a trifle more or a trifle less than 100 cc., or it may vary slightly from a hot day in summer to a cold day in winter, but it should not vary between the first and last bottles of a set. It is a good plan to fill and empty the pipette a few times before beginning to fill the set. After withdrawing the bottle from, under the pipette the stopper is dipped in distilled water and inserted into the neck of the bottle with care.

Having filled the set, the carefully stoppered bottles are placed in a shaking machine and agitated for from 3 to 5 min. to settle the precipitate. The bottles are next placed upon a black shelf, technically called a “ board,” with a black background about as high as the shoulder of the bottle and about 3 in. back of the bottle, the whole being installed in a window, preferably with a northern exposure.

Up to this point in the method, the procedure is substantially the same in all the laboratories, but from here on there are slight differences in the manipulations. In common with many volumetric methods there is difficulty in this one in determining the end-point. In many descriptions of the method the operator is directed to add decimal salt solution in small measured amounts, with agitation between the additions, until no more precipitate is formed on adding the salt solution. Too much salt has now been added, and this excess must be determined by the addition of small measured amounts of decimal silver solution, and the amount of silver present in the metal determined by balancing these amounts.

This method is open to two serious objections. When exactly the proper amount of salt has been added to precipitate the silver present, a condition of equilibrium in the solution results, which is disturbed by the addition of either reagent with the separation of a precipitate. This obscures the end- reaction. These alternate dosings and shakings consume too much time for rapid work, and, after many shakings, the solution does not clear well.

In the Mint service, therefore, the operation known as ” reading the cloud ” has been substituted for these alternate dosings of the solution. This operation is far more rapid, but it requires a great deal of skill and constant practice to yield the best results. In reading the cloud, after the addition of the 100 cc. of standard salt solution and shaking, a measured amount, 1 or 0.5 cc., of decimal salt solution is added to each bottle. The delivery end of the pipette is placed against the neck of the bottle as far down, as possible and the solution allowed to flow gently down the side of the bottle so that it will remain on the surface of the solution in the bottle with the minimum amount of mixing; by a slight rotary motion of the hand the decimal solution is then mixed with the upper portion (about 1/3) of the bottle solution. This produces a cloud of AgCl in the solution, and the next step is based upon the appearance of this cloud. Here the skill and visual condition, together with the personal equation, of the operator are of the utmost importance.

If the cloud be very heavy, two or more portions of the decimal salt solution are added, the amount depending upon the density of the cloud, and the bottles shaken in the machine again. If the cloud is light, only one dose of decimal is used. If the cloud is very light the bottle is again shaken by hand to bring more of the solution into reaction and the cloud again examined. As the result of this treatment one dose of decimal may be used, or the bottle may be shaken by hand again to bring the balance of the solution above the precipitate into the reaction. Here again a dose of decimal may be used, or the final reading of the cloud may take place. In the final reading the operator estimates from the density of the cloud what portion of the dose of decimal solution was consumed in precipitating the silver. Many operators estimate to 0.25 cc., others claim to be able to estimate to 0.1 cc. The results of the investigation given beyond indicate that on standard metal estimating to 0.1 cc. is not profitable under present conditions.

All the bottles are eventually brought down to the final cloud, and at the end the amount of decimal solution added to each bottle is recorded. The records of the assays are then compared with the proofs and the fineness of the samples determined. The actual fineness of a sample is shown entirely by the amount of decimal solution used as compared with the amount of decimal required by the proof. It is entirely independent of the amount of standard solution used. Therefore in determining the fineness of a sample we simply compare the amount of decimal used with the amount used in the proof, making allowance for the 0.5 mg. more of silver in the proof above the silver in 1,115 mg. of metal at the exact standard of 900, and taking into consideration the weight of the sample used. For instance, if the sample required 0.5 cc. less than the proof the sample would be reported at 900. This is not strictly exact because 0.25 cc. of decimal salt solution equals 0.25 mg. of silver, but 0.25 mg. is only 0.22 fine on 1,115 mg. and 0.5 cc. would be only 0.44 fine. However, this variation is too slight for practical consideration, and in general the finenesses are read directly from the difference between the sample and proof. Generally, also, the differences are read in 0.25 of a cc.

Practically we deduct 2/4 from the amount of decimal solution used on the proof, and call it standard. Then for each quarter’s difference from standard we add or subtract, as the case may be, from 900 in accordance with the following tabulation:

A second method of determining the end-point, which appears to be even more exact, but which requires more time, consists in adding only 0.25 cc. of the decimal solution after machine shaking, then estimating the number of quarters the solution will stand. These are added and the bottles shaken in the machine. This is continued until the quarter added after shaking produces no precipitate, in which case the last quarter is not counted, or else such a slight precipitate is produced that the quarter is counted but the bottle is not shaken again. The reading of the results is the same as before described.

In both these methods great care is exercised to avoid the addition of so much salt as to require back-titration with decimal silver. Back-titration is regarded with much disfavor.

In the regular work of the Mint Bureau it became desirable to ascertain just how much reliance could be placed upon the results obtained by the Gay-Lussac method as ordinarily executed in the service and an extensive investigation was undertaken for this purpose.

Six samples representing three melts of ingots were distributed among four laboratories in the service without informing any of the institutions just what the samples were. Table I. summarizes the results reported.

Notwithstanding the great preponderance of results from 898.9 to 899.1 shown here, the ingot melts were reported by the assayer of the mint where they were made as follows:

No. 1…………………………………….898.5 fine in silver.
No. 2……………………………………898.25 fine in silver.
No. 3……………………………………898.5 fine in silver.

Table I. was submitted to the various assayers for criticism and comment and suggestions for improvements in the method.

Five reasons were advanced to explain, in part at least, the differences shown, as follows :

1. Difference in samples taken for assay.
2. Non-homogeneity of the metal.
3. Working on a sample on different days.
4. The personal equation of the operator.
5. Accidents.

Two assayers suggested that perhaps under the actual, everyday working conditions of our service the method is not so accurate as it is supposed to be.

It was quite impossible at this time to deal with the personal equation of the operator or the chapter of accidents, but, in order to meet the first two of these reasons, a small bar of standard silver was prepared in the Bureau laboratory with especial care to avoid segregation of the metal. It was then rolled out and cut into small oblongs of which 30 to 35 were required to weigh 1,115 mg. To meet the third reason, four samples of the oblongs, each weighing approximately 3.75 g., sufficient for three determinations of the silver, were sent to each one of the same four laboratories at one time. Subsequently one laboratory ran two more samples. In all 66 assays were made and practically all of the bar was used for these assays. Table II. summarizes the results reported.

Table II. was also submitted to the four assayers for comment and criticism, and the investigation was continued upon six sets of ingot samples and six special samples. These samples were sent out to the four institutions for comparative test assays, in a long series. In no case was any institution informed as to the source of the metal, or given the results of any previous assays. The samples were sent out in lots of four in all cases but one. In several cases all four samples in a lot were the same, and in two cases this fact was stated in sending them out. In many cases two or three samples in a lot would be the same, but this fact was not disclosed. In one case six samples representing the three regular samples of a melt of ingots were sent out with this statement, but it was not shown which were the duplicate samples.

In the ordinary operations of melting ingots and taking samples from the melts, it is quite possible that there might be some slight difference in the samples themselves. Tables III. and IV., however, indicate that in the present condition of the method such differences would be masked by differences in the results due to the method itself.

The regular samples from a melt of ingots are generally too small to allow a sufficient number of test assays to be made. From three melts one mint supplied larger samples and they were sent to the four laboratories for test assays, and Table III. summarizes the results reported, together with the original assays made at the mint of origin of the samples.

This table indicates that there was a decided tendency at this mint to report their ingots high.

Another mint supplied two sets of ingot samples that were sent out to the four laboratories in the general series of samples, and Table IV. summarizes the results reported and gives the original assays made at the mint.

A third set of samples from the second mint was sent out in a set of six and the instructions informed that they represented a melt of ingots, but the duplicates were not indicated; Table V. summarizes the results reported together with the original mint assays.

The general tendency of the original reports on these three melts appears to be a trifle low, and this is supported by the low test assays on Melt No. 176, made at the mint of origin.

In Tables III., IV., and V. all the results reported have been given. Undoubtedly, some of these were vitiated by accidents. These tables also indicate that there is a general laboratory equation similar to the personal equation of an operator.

In order to eliminate differences in the samples as a source of difference in the assay results, one mint prepared two special samples, in regard to which the assayer wrote: “ I have prepared two samples of standard silver (6 oz. each). I used coin-ingot melted a number of times with stirring; finally cast in a closed mold that had been chilled with ice; poured silver at as low temperature as possible. Resulting bars were very free from oxidation. They were, however, thoroughly cleaned, and rolled into strips and cut.”

The samples cut into oblongs, of which it took about 30 to weigh 1,115 mg., were forwarded to the Bureau, where they were further mixed and portions of 3.75 g. weighed out into small envelopes, giving ample material for three determinations of silver in each envelope. Ten envelopes were sent at various times to each one of the four institutions. Each laboratory made 30 determinations of silver on each of the larger samples, making a total of 120 assays on each, but in tabulating the results reported a certain amount of discretion has been exercised. In general, results at either extreme in fineness have not been included in the table unless they were reported at least twice by two laboratories, or five times by one laboratory; Table VI. summarizes the results together with the original assays at the mint of origin.

Another mint (Laboratory No. 3 above) prepared two special granulation samples, in regard to which the assayer wrote: “ I am submitting two samples of 6 oz. each of the most homogeneous standard silver that I have been able to prepare. These samples were prepared as follows: Standard silver ingot granulations assaying from 899.1 to 899.8 and weighing in the aggregate a little over 30 oz. were melted with frequent stirring and the entire melt granulated. Two samples of about 8 oz. each were then segregated and 18 assays made from the different portions of the sample. 17 of these assays went 899.5.”

These samples were divided into portions weighing 3.75 g. each at the Bureau. It was not noticed at first that they were really only one sample, so that 11 envelopes of No. 1 and 9 envelopes of No. 2 were sent to each one of the four institutions. Table VII. summarizes the results reported, together the original assays at the mint of origin.

In this case Laboratory No. 3 is the one that supplied the special bar samples. The laboratory supplying these special granulation samples is the same one that forwarded ingot samples Nos. 132-3-8 and the same tendency to obtain high results is again shown here.

More than 15 years ago, in the article entitled, The Actual Accuracy of Chemical Analysis, in speaking of the “ degree of accuracy exhibited in actual every-day practice” I said: “In estimating this, little weight will be given to the evidence afforded by the agreement of duplicate or multiple determinations by the same chemist; for I am convinced that such agreement is a delusion and a snare.” The results reported on these granulation samples afford a striking illustration of the soundness of this position. This metal was assayed 257 times. The laboratory preparing it reported if 56 times at, 899.5 fine, but the other laboratories reported this figure only 3 times and one of these determinations was questioned by the laboratory reporting it. On the other hand, the work of the other laboratories shows that this metal could not have been over 899 fine.

Another small bar was prepared in the Bureau laboratory with especial care, and cut up into oblongs requiring from 30 to 35 to weigh 1,115 mg. These oblongs were thoroughly mixed and small samples of 3.75 g. put into envelopes. Ten of these envelopes were sent to each one of the four institutions, and Table VIII. summarizes the results reported.

The laboratories here are numbered the same as in Table VII., and No. 4 is the mint of Table VII.

A final sample was made up at the Bureau by uniting the oblongs left over from the regular coin work of the Bureau, which were of widely varying finenesses, and, without any special mixing, weighing up samples of 3.75 g. Ten envelopes were sent to each one of the institutions and Table IX. summarizes the results reported. As this sample had no claim to uniformity all the results reported are given.

The laboratory numbers here are the same as in Table VIII.

As a final test of the method when practically applied to ingot work, three large samples were taken from each one of nine melts of ingots and assayed in various laboratories. Each sample was assayed in duplicate five times and in triplicate once, making a total of 39 determinations on each melt. Table X. summarizes the results reported.

As a general conclusion from the elaborate tests herein given it may be stated that two operators, working upon identical samples of standard silver and making four determinations each, may differ as much as 1 fine in their reports. Having thus established the capacity of the method as at present carried on as a commercial operation, attention is now being given to improving the method so as to reduce the allowable limit of difference.

It may seem unnecessary to many of my readers, but my experience shows that there is much confusion and uncertainty

in stating the composition of precious metal bullion. This is generally done in parts per thousand or fineness. It is similar to stating composition by percentage, but in bullion work there is much misunderstanding regarding the decimal point. Intelligent people who would never think of expressing fifty per cent, by .50 per cent, habitually write five hundred fine as .500 fine and incorrectly use the decimal point before the numerals in stating fineness. Whenever “ fine ” or “ fineness ” or an equivalent expression is used, in stating the composition of bullion, the decimal point should not be used unless the figures following it express a quantity of less than 1 fine. Again, some people incorrectly use a preliminary “ 0 ” when the fineness is less than 100. They would not think of writing 05 per cent. but habitually do write 052 fine.

Again, some people mistake fineness as being the exact equivalent of percentage, but it is only one-tenth of percentage, and in considering the accuracy here shown by the Gay-Lussac method it should be noted that 1 fine is the equivalent of 0.1 per cent.