It is well known how useful for our knowledge of the metals in gadolinite has been the observation by Berlin, made twenty years ago, that the nitrates are differentially decomposed when the temerature is raised; it still continues to enrich science in new results. By applying that observation, Berlin himself succeeded in obtaining the white earth yttria perfectly free of the rose-colored earth that Mosander discovered in it in 1843; since then, Bahr and Bunsen, and later Hoglund, purified the rose-colored earth from yttria. This earth, which had been named "erbine," was recently shown by Marignac to be a mixture of two different earths: one of pure rose color and marked absorption bands, the real erbine; the other white, to which he gave the name "ytterbine." He had too little of the new substance at his disposal to prepare it in a pure state, but be was led to think that ytterbine is perfectly white and does not absorb light and that its molecular weight reaches 131, calculated for the formula YbO.

For lack of material, Marignac had to abandon the further study of the new earth, and he invited chemists who might have sizable quantities of erbine to continue this research. I have devoted myself to this study for some time, the more willingly since I was at the point of reviewing the molecular weight which Hoglund assigns to this earth (129.7); because this chemist mentions, among others he investigated, four that have a higher molecular weight (131.2, 130.4, 129.9, and 129.8). Therefore, I thought that it should be possible to push the decomposition of the nitrate from erbine still further and finally arrive at a higher constant number.

When I began this work, I bad in my possession 63 g. of erbine with a molecular weight of 129.25; 1 had obtained it from gadolinite as well as from euxenite by following exactly the method described by Marignac; however, I stopped heating the molten material as soon as red fumes started to come out, and I thus always obtained crystallized lower nitrates with increased erbine contents. At first, I tried to apply the same procedure for extracting ytterbine from erbine; I found that the molecular weight of the earth that crystallized out as the lower nitrate slowly rose to 130.0, 130.3, 130.4, up to 130.57 (for a very small quantity). This work was so long and exhausting that it seems questionable whether a completely pure ytterbine could ever be reached in this way.

I then used Marignac's process without modification and was successful. After thirteen series of decompositions, heating to complete solidification of the nitrates, there remained a lower nitrate which in the form of the molten nitrate showed only two weak rays of absorption in the green and the red. The solution was precipitated with oxalic acid, evaporated, and gave 3.5 g. of a white earth with a scarcely perceptible rose tint. The determination of the molecular weight gave me the following results:

I. 1.0238 g. of the earth gave 1.6656 g. sulphate (RO = 127.62).
II. 1.0302 g. of the earth gave 1.6758 g. sulphate (RO = 127.66).

I could not explain the results of these two determinations giving the low numbers of 127.66 or 127.62, unless there was a mixture with another earth of a molecular weight below that of ytterbine. Now the problem was to demonstrate the presence of such an oxide and, if possible, to isolate and characterize it. The memoir that follows contains the results of this effort.

Meanwhile, having obtained the molecular weight of 127.6 instead of Marignac's 131, 1 examined the solutions from which the insoluble nitrates had precipitated.

The mother liquors of fractions 9-17 all contained an earth of molecular weight above 131. They were combined, subjected to the partial decomposition of Marignac's method, and after eight series of operation they gave about 3.5 g. of an earth . . . with a molecular weight of 131.63. By decomposing its nitrate in such manner that the heating was stopped as soon as the initially molten mass became pasty, the last traces of erbine were easily removed. The deposit of basic nitrates contained an earth of molecular weight 131.92 and 132.17 in two different operations:

I. 0.7503 g. of earth gave 1.2053 g. of sulphate (RO = 131.92).
II. 0.7119 g. of earth gave 1.1428 g. of sulphate (RO = 132.17).

[In the original occur misprints of 0.2053 and 0.1428.]

The molten nitrate of this earth shows no trace of light absorption; it is, therefore, perfectly pure ytterbine. Marignac's supposition that the new earth would show no absorption is thus completely proved true.

About Scandium, a New Element

The preparation of ytterbine, described in the foregoing note, had furnished me with an earth that was deposited as an insoluble basic nitrate; by extracting the heated mass with boiling water, the molecular weight was found to be 127.6, and not 131, as it should have been according to Marignac. I concluded that the analyzed product should be a mixture with an earth of a lower molecular weight than 131. Thalén, who examined its spectrum, found that its chloride gave some rays not occurring in the known elements. In order to isolate this substance, I carried out several partial decompositions and determinations of the molecular weight of the earth deposited in the insoluble residues containing the new substance.

After the last series of decompositions, the molecular weight had dropped 26 units below 132, the weight of ytterbine; nevertheless, the examined product still contained this earth as an impurity. It was impossible for me to carry out any more partial decompositions of nitrates so as to obtain the new substance, perhaps, in perfect purity. Actually, I did not need to have it for demonstrating that a hitherto unknown element was mixed with ytterbine, because the spectrum of this substance, like that of impure ytterbine, sufficiently showed the character of a new element . . . .

For the element thus characterized I propose the name "scandium," which will bring to mind its presence in gadolinite or euxenite, minerals that have so far been found only in the Scandinavian Peninsula.

About its chemical properties, I know at present only this: It forms a white oxide and its solutions show no bands of light absorption. When calcined, it dissolves only slowly in nitric acid, even at boiling, but more readily in hydrochloric acid. It is completely precipitated from the solution of the nitrate by oxalic acid. This salt is very easily and completely decomposed at the temperature at which ytterbium nitrate is partially decomposed. With sulphuric acid it forms a salt that is as stable on heating as the sulphates from gadolinite or cerite and, like these, can be completely decomposed by heating with ammonium carbonate. The atomic weight of scandium = Sc is less than 90, calculated for the formula ScO. . . .

It would certainly be premature to discuss the affinities of the new substance or its place among the other elements; nevertheless, I cannot refrain from making some observations on this subject, guided by the chemical properties that are now known.

Since scandium nitrate decomposes so easily on heating that an almost pure ytterbine was obtained in the decompositions 13-21 of the preceding note, while scandine remained completely in the insoluble residues, it is not possible that the oxide has the formula ScO.

. . . The composition Sc₂O₃ for the earth material is supported by the following facts:

1. Scandine is present in the minerals, together with other rare earths R₂O₃
2. Solutions of scandium and ytterbium (salts) behave in the same way to oxalic acid.
3. There is much analogy between the behavior of the nitrates of scandium and ytterbium at high temperatures.
4. The double salt of sandium sulphate with potassium sulphate shows that scandium belongs to the same group of metals as those of gadolinite and cerite; all give salts of the same composition.
5. The insolubility of the same salt in potassium sulphate saturated solution indicates that scandium belongs to the cerite group.
6. In the composition of the selenites, the new earth shows much analogy with Y₂O₃, Er₂O₃, Yb₂O₃, on the one hand, giving neutral selenites, and on the other hand Al₂O₃, In₂O₃, Ce₂O₃, La₂O₃, which give very analogous acidic salts, as I have previously shown; I have also obtained a selenite of the same composition from Gl₂O₃
7. The atomic weight of scandium is 44; this is the value Mendeleev assigned to the predicted eka-boron. . .
8. The specific heat and the molecular volumes of the earth and of the sulphate place scandium between glucine and yttria.