TABLE II.-Component minerals from certain of the above igneous rocks.

Of the igneous rocks specimens were so selected as to represent not only many widely separated localities, but also numerous varieties from the least siliceous up to those high in silica, in order to ascertain whether a preconceived opinion that the vanadium accompanied chiefly the less siliceous rocks was well founded or not. The choice was, however, confined largely to those rocks analyzed in this laboratory. within the past three or four years of which a supply of powder remained after the original analyses had been completed, and hence the list is perhaps not fully representative. Nevertheless it permits of drawing certain conclusions, the chief of which is that the vanadium predominates in the less siliceous igneous rocks and is absent, or nearly so, in those high in silica. The inference, based on the existence of the mineral roscoelite, classed as a vanadium mica, at once suggests itself, that the ultimate source of the vanadium may be one or more of the heavier silicates such as the biotites, pyroxenes, and amphiboles, and a few tests on all the available mineral separation products lend strong support to this view. For instance, the amphi bole-gabbros 7 and 11 show 0.038 per cent and 0.02 per cent V2O3, while the amphiboles 7 and 11a separated from them give 0.062 per cent and 0.037 per cent; the pyroxenic gneiss 23 shows 0.083 per cent against 0.127 from its contained biotite 23; the diorite 29 with 0.031 per cent contains an amphibole 29a with 0.066 per cent; from 0.011 per cent in the quartz-mica-diorite 45 and 0.012 per cent in the quartzmonzonite 46 the percentages rise to 0.048 and 0.066 in their separated

biotites 45 and 46a. The pyroxene 21a shows, however, practically the same amount as its mother rock, the syenite-lamprophyre 21.

In most of these cases, notably the last one, the vanadium in the separated mineral is not sufficiently in excess of that in the rock from which it was taken to account for all of that found in the latter. Hence,

3

if the determinations are correct it must also be a constituent of some other mineral than the one analyzed. In roscoelite the trivalent condition of vanadium corresponding to the oxide V2O, is now recognized as probable, although Roscoe's analysis reports V2O, and Genth's an oxide intermediate between V2O3 and V2Os. This assumption seems to be necessary if the mineral is to be regarded as a mica, and it is then doubtless the equivalent of trivalent iron or aluminum. It would then be natural to look for it in the aluminous or ferric silicates of igneous rocks, certain biotites, pyroxenes, and hornblendes, and its absence in such minerals as serpentine and chrysolite, as shown by analyses 3 and 60, appears natural enough.

Few and inconclusive as the above comparisons are, they seem to favor strongly this view as to the source of the vanadium, and in a measure are confirmatory of the observation of Hayes (Proc. Am. Acad. Arts Sci., Vol. X, 1875, p. 294), who rather indefinitely associates it with phosphorus and proto-salts of iron and manganese, which are usually more prominent components of basic than of acid rocks.

We are probably justified by the evidence in tabulating the vanadium as V2O, in analyses of igneous and some metamorphic rocks which have undergone little or no oxidation, but with sandstones, clays, limestones, etc., which are of more or less decided secondary origin, this is probably not the case. The probabilities are there largely in favor of its acid character and the existence of various vanadates of calcium, iron, aluminum, etc., in which case it should appear in analytical tables as V2Os.1

In the regular course of analysis vanadium will be weighed with alumina, iron, titanium, etc., since it is precipitated by both ammonia and sodium acetate in presence of those and other metals, hence the alumina percentages in nearly all rock analyses heretofore made are subject to correction for the vanadium the rock may have held. This correction is of course to be made in terms of V2O and not of V2O3.

All determinations of iron are likewise affected by its presence, whether as V2O, or V2O3. As V2O, it will make the FeO appear too high in the proportion V2O3: 4FeO, or 150.8: 288, an error which becomes appreciable in some of the basic rocks and amounts to 0.25 per cent in the biotite 23. As V2O, the FeO will not be affected, but in either condition the Fe2O will need correction and to a different extent, according as the titration of iron is made after reduction by hydrogen sulphide or by hydrogen. If the former is used, as should always be the case in presence of titanium, the vanadium is reduced by it to V2O4, which in its action on permanganate is equivalent to 2 molecules of FeO representing 1 of Fe2O3, or only one-half as great as the influence on the FeO titration of the same vanadium as V2O3. An example will make this clear:

Found 2.50 per cent apparent FeO in a rock containing 0.13 per cent V2O3.

Deduct 0.25 per cent FeO, equivalent in its action on KMnO4 to 0.13 per cent V.03.

Leaving 2.25 per cent FeO corrected.

Found 5 per cent apparent total iron as Fe2O, in the same rock.

Deduct 0.14 per cent Fe2O3, corresponding to 0.13 per cent V2O3.

Leaving 4.86 per cent corrected total iron as FeO3.

Deduct 2.50 per cent Fe2O3, equivalent to 2.25 per cent FeO.

Leaving 2.36 per cent Fe2O, in the rock.

Failure to correct for the vanadium in both cases would have made the figures for FeO and Fе,03, respectively, 2.50 and 2.22 instead of 2.25 and 2.36 as shown above.

It was not until the greater part of the above tests had been finished that any careful attempt was made to identify molybdenum as well as vanadium. From the evidence gathered during the latter part of the work it would seem that molybdenum, when it does occur, is a much less important constituent quantitatively than vanadium, and that unlike the latter it accompanies the more acid rocks. Molybdenite is a well-known accessory constituent of some granites, etc., but in the above instances its amount was extremely small and no hint was obtained as to its state of combination.

CHEMICAL METHOD EMPLOYED.

In conclusion it is proper to outline the method by which the foregoing tests were carried out, and to indicate the precautions that must be observed in order to insure good results.

Quite a number of workers have busied themselves with the problem of vanadium estimation in ores and rocks, particularly magnetites and other iron ores, and the methods used have been often diverse in parts if not altogether. There is nothing absolutely novel in the following except that chromium and vanadium when together need not be separated, but are determined, the former colorimetrically, the latter volumetrically, in the same solution as detailed elsewhere (p. 44).

Five grams of the rock are thoroughly fused over the blast with 20 of sodium carbonate and 3 of sodium nitrate. After extracting with water and reducing manganese with alcohol it is probably quite unnecessary, if the fusion has been thorough, to remelt the residue as above, though for magnetites and other ores containing larger amounts of vanadium than any of these rocks, this may be necessary, as Edo Claassen has shown.' The aqueous extract is next nearly neutralized by nitric acid, the amount to be used having been conveniently ascertained by a blank test with exactly 20 grams of sodium carbonate, etc., and the solution is evaporated to approximate dryness. Care should be taken to avoid overrunning neutrality because of the reducing action of the nitrous acid set free from the nitrite, but when chromium is present it has been my experience that some of this will invariably be retained by the precipitated silica and alumina, though only in one case have I observed a retention of vanadium, it being then large. The use of ammonium nitrate instead of nitric acid for converting the sodium carbonate into nitrate did not seem to lessen the amount of chromium retained by the silica and alumina.

As a precautionary measure, therefore, and always when chromium was to be estimated also, the silica and alumina precipitate was evaporated with hydrofluoric and sulphuric acids, the residue fused with a little sodium carbonate and the aqueous extract again nearly neutral

Am. Chem. Jour., Vol. VIII, p. 437.

ized with nitric acid and boiled for a few moments, the filtrate being added to the main one.

Mercurous nitrate was now added to the alkaline solution in some quantity so as to obtain a precipitate of considerable bulk containing chromium, vanadium, molybdenum, tungsten, phosphorus, and arsenic, should all happen to be in the rock, and also an excess of mercurous carbonate to take up any acidity resulting from the decomposition of the mercurous nitrate. Precipitating in a slightly alkaline instead of a neutral solution renders the addition of precipitated mercuric oxide unnecessary for correcting this acidity. If the alkalinity, as shown by the formation of an unduly large precipitate, should have been too great, it may be reduced by careful addition of nitric acid until an added drop of mercurous nitrate no longer produces a cloud.

After heating and filtering, the precipitate is ignited in a platinum crucible after drying and removing from the paper to obviate any chance of loss of molybdenum and of injury to the crucible by reduction of phosphorus or arsenic. The residue is fused with a very little sodium carbonate, leached with water, and the solution, if colored yellow, filtered into a graduated flask of 25 cubic centimeters or more capacity. The chromium is then estimated accurately in a few minutes by comparing with a standard alkaline solution of potassium monochromate (p. 37). Then, or earlier in absence of chromium, sulphuric acid is added in slight excess and molybdenum and arsenic together with occasional traces of platinum are precipitated by hydrogen sulphide, preferably in a small pressure bottle. If the color of the precipitate indicates absence of arsenic, the filter with its contents is carefully ignited in porcelain and the delicate sulphuric acid test for molyb denum is applied.

The filtrate, in bulk from 25 to 100 cubic centimeters, is boiled in a current of carbon dioxide to expel hydrogen sulphide, and titrated at a temperature of 70-80° C. with a very dilute solution of permanganate representing about one milligram of V2O, per cubic centimeter as calculated from the iron strength of the permanganate, one molecule of V2O5 being indicated for each one of Fe2O3. One or two checks are always to be made by reducing again in a current of sulphur dioxide gas, boiling this out in a current of carbon dioxide again, and repeating the titration.

As shown in a previous paper (p. 45), the presence of even thirty times as much Cr2O, as V2O, does not prevent a satisfactory determination of the vanadium if the precautions therein given are observed, provided there is present not less than one half to 1 milligram of V2O5 in absolute amount. In absence of chromium less than half a milligram can be readily estimated. The phosphoric acid almost invariably present does not affect the result.

From a sulphuric solution the separation of molybdenum by hydrogen sulphide is much more rapid and satisfactory than from a hydrochloric solution.

In case the volume of permanganate used is so small as to make doubtful the presence of vanadium, it is necessary to apply a qualitative test which is best made as follows: The solution is evaporated and heated to expel excess of sulphuric acid, the residue is taken up with 2 or 3 cubic centimeters of water and a drop or two of dilute nitric acid, and a couple of drops of hydrogen peroxide are added. A characteristic brownish tint indicates vanadium. Unless the greater part of the free sulphuric acid has been removed the appearance of this color is sometimes not immediate and pronounced, hence the above precaution.

The above is a surer test to apply than the following: Reduce the bulk to about 10 cubic centimeters, add ammonia in excess and introduce hydrogen sulphide to saturation. The beautiful cherry-red color of vanadium in ammonium sulphide solution is much more intense than that caused by hydrogen peroxide in acid solution, but the action of ammonia is to precipitate part or all of the vanadium with the chromium or aluminum that may be present or with the manganese used in titrating, and ammonium sulphide is unable to extract the vanadium wholly from these combinations. Usually, however, the solution will show some coloration, and addition of an acid precipitates brown vanadium sulphide, which can be collected, ignited, and further tested if desired.

SUMMARY OF RESULTS.

Vanadium occurs in quite appreciable amounts in the more basic igneous and metamorphic rocks, up to 0.08 per cent or more of V2O3, but seems to be absent or nearly so from the highly siliceous ones. The limited evidence thus far obtained points to the heavy ferricaluminous silicates as its source-the biotites, pyroxenes, amphiboles. As opportunities offer further evidence will be accumulated and it is hoped that other chemists will lend their aid.

Limestones and sandstones appear to contain very small amounts of vanadium, as shown by analyses of a composite sample of each, aggregating over 700 different occurrences.

From the few tests of molybdenum it appears as if this element were confined to the more siliceous rocks. It is present in no observed case in amount sufficient for quantitative measurement when operating on 5 grams of material.

NOTE. Since the above was written a few tests have been made on minerals of which powdered samples were at hand. A phlogopite from Burgess, Canada, gave 0.007 per cent V2O3. Mica from Laurel Hill, Georgia, gave 0.026 per cent V2O3. Protovermiculite from Magnet Cove, Arkansas, gave 0.04 per cent V2O3. Hallite from Chester County, Pennsylvania, gave 0.01 per cent V2O3. Jeffersonite from Franklin Furnace, New Jersey, gave none, and a nonferruginous amphi bole from St. Lawrence County, New York, gave a faint trace.