barium salt were, as he admits, not satisfactory, while the analyses of the silver salt were made with a crude preparation, for the purity of which there is no guaranty. I have myself found that barium trimetaphosphimate varies considerably in composition, contains water, and is hygroscopic. Sodium trimetaphosphimate is the only well-defined substance I have been able to obtain by decomposing the chloronitride by alcoholic soda, and the salt is so characteristic as hardly to be mistaken. It seems quite possible, therefore, that Gladstone actually had crude trimetaphosphimic acid in hand and failed to recognize its true nature only because of the unfortunate selection of salts which are amorphous, hydrated, and difficult to obtain pure. In a paper of much later date Mente' described several acids obtained by the successive action of ammonium carbamate and water on phosphorus oxychloride. Among them is one to which he gave the PO.OH name diimidodiphosphoric acid and the formula NH<>NH, and which PO.OH he regards as identical with Gladstone's pyrophosphodiamic acid. If Mente's formula is correct, it is the second or dimetaphosphimic acid. I have been unable to repeat Mente's work, and his data are too meager to admit of a positive conclusion, either as to its molecular weight or even its empirical composition, but it appears not to be identical with trimetaphosphimic acid. Constitution of trimetaphosphimic acid.—Although trimetaphosphimic acid is an uncrystallizable and unstable body, it forms stable salts, several of which crystallize in characteristic forms. Three atoms of hydrogen are replaceable by alkali metals, while silver is able to replace either 3 or 6. The following are the most noteworthy: P3N3O6H3Nas+4H,O-Rhombic prisms. P3N3OHзNa Ba+1}H2O—Rhombohedra. The first, third, and fifth serve to identify the acid. The tertiary silver salt, being anhydrous and easily obtained pure, establishes its empirical formula. Those salts which contain water of crystallization do not lose it completely at any temperature short of decomposition, leaving open the question whether the acid may not have the formula PзN3O7H8. The constitution of trimetaphosphimic acid depends on that of the chloronitride P3N3Cl. The structural formula of the latter has not yet been definitely established, but the following data are available. It is reasonably certain that in the chloronitride phosphorus atoms are united by nitrogen atoms. Its formation from phosphorus pentachlo 1Ann. Chem. (Liebig), Vol. CCXLVII, 1888, pp. 239, 244. ride and ammonia is best explained on this assumption, as is its decomposition into orthophosphoric acid and ammonia. If phosphorus were united to phosphorus and nitrogen to nitrogen, the formation of reducing phosphoric acids or of hydrazine might be expected. Neither is it likely that chlorine is united to nitrogen, for in this case hydroxylamine might be expected to result. Several structural formulas are possible which meet this requirement, the simplest being that in which the nucleus consists of a symmetrical ring of 3 phosphorus and 3 nitrogen atoms: PCL, N or a similar one with diagonal or "centric" union. Direct replacement of chlorine by hydroxyl would then give The labile nature of the hydrogen atoms in nitrogenous bodies, as observed in many organic compounds, makes it by no means improbable that an acid of this form may of itself, or under the influence of reagents, undergo transformation into the tautomeric form: An acid of the form (a) may be expected to give two sodium salts in which, respectively, 3 and 6 atoms of hydrogen are replaced. A salt with 6 atoms of sodium can not be produced, however, by any method which I have tried. The only salt besides the ordinary one with 3 atoms of sodium is one with 4 atoms, and this is formed in the presence of a large excess of caustic soda; it is very unstable and is reconverted into the 3-atom salt by repeated precipitation from aque ous solution by alcohol. It is not obvious why an acid of the form (a) should give such a salt. If we assume that (b) represents the correct formula, and that the acid is the symmetrical inner anhydrid or lactam of an open chain acid, we may regard this salt as being formed by addition, thus: the latter being a salt of amido-diimidotriphosphoric acid.' Unfortu nately I have been unable to prove this with certainty. A silver salt of the composition P3N3O,H.Ag, would render it extremely probable, as all silver salts of phosphorus nitrogen acids hitherto observed are free from crystal water. A salt with 4 atoms of silver can be obtained from the above 4-atom sodium salt, but it is very unstable, and its composition does not agree with sufficient sharpness with that of the cyclic or open chain form to establish either formula definitely. I am therefore inclined to give preference to formula (b). It seems likely, however, that salts of triphosphonitrilic acid may also exist. The hexa-silver salt has been observed in two well-defined forms, as well as a third, possibly intermediate one. One of these is white, amorphous, and soluble in ammonium nitrate. It is easily converted into the other of identical composition, which is red, crystalline, and insoluble in the same reagent. The colorless salt probably contains silver united to oxygen only, while in the other one-half is united to nitrogen. Decomposition products of trimetaphosphimic acid.-An aqueous solution of trimetaphosphimic acid, or of one of its salts acidified with one of the stronger mineral acids, decomposes slowly in the cold, rapidly on heating, the ultimate products being orthophosphoric acid and ammonia. If, however, the action be limited, there results a mixture of intermediate acids. The analysis of this mixture, which involved many experimental difficulties, proved the presence of the following: 1. Unchanged trimetaphosphimic acid, P3N3O6H6. 2. Diimidotriphosphoric acid, P3N2O,H7. 3. Imidodiphosphoric acid, P2NOHË. 4. Pyrophosphoric acid, P2O7H4. 5. Orthophosphoric acid, PO,H3. 1 The acids PO(OH)2.O.PO(OH), and PO(OH)2.O.PO(OH).O.PO(OH), being frequently designated as di- and triphosphoric acid, I call those in which the linking oxygen is replaced by imide, NH, imido-di- and dümidotriphosphoric acid. This acid has the same composition and is probably identical with Gladstone's azophosphoric or pyrophosphamic acid. Jour. Chem. Soc. London [2], Vol. VI, p. 66, etc. Assuming formula (b) for trimetaphosphimic acid, its decomposition would be thus represented: The second acid is important as a connecting link, proving that trimetaphosphimic acid actually has the triple formula (PNO2H2);. Most remarkable is the formation of pyrophosphoric acid, which occurs directly from imidodiphosphoric acid. The facts are these: If sodium imidodiphosphate be weakly acidified with acetic acid and boiled from five to ten minutes, not more than 15 per cent is converted into pyrophosphoric acid and the remainder into orthophosphoric acid. Control experiments with sodium pyrophosphate showed that under the same conditions but a small portion is converted into orthophosphoric acid, 90 per cent of the pyrophosphate being recovered. The pyrophosphoric acid is therefore not an intermediate product, but the imidodiphosphate is converted for the greater part directly into orthophosphoric acid, and in a much less degree into pyrophosphoric acid. The above formula of imidodiphosphoric acid containing the group P—NH-P, follows directly from that of trimetaphosphimic acid. In view of the ease with which the condensed phosphoric acids split up in aqueous solution into orthophosphoric acid, while the reverse never occurs, it is difficult to see how the action of water or acids on a body of the assumed formula could give pyrophosphoric acid. If we adopt Gladstone's formula' O POKOH pyrophosphoric acid, but must at the same time assume that somewhere in the series the group P-N-P has been converted into P-O-P, for in the chloronitride, P3N3Cl, phosphorus can be united only by nitrogen. If the acid were really the amide of pyrophosphoric acid, boiling with acetic acid would first convert it into pyrophos phoric acid which, as above shown, is fairly stable under the observed 1 Jour. Chem. Soc. London, [2], Vol. VI, p. 70. conditions, and should be found present to the extent of at least 90 per cent instead of only 15 per cent. On this ground, it seems to me, we must reject the amidopyrophosphoric formula, and regard the body as imidodiphosphoric acid, This formula does not account for the pyrophosphoric acid. We may, perhaps, conceive that owing to the greater affinity of phosphorus for oxygen than for nitrogen, a group, PO.(OH)2, is able to change places with a hydroxylic hydrogen atom: the latter, in common with other amido phosphoric acids, being unstable, and passing easily into pyrophosphoric acid. It may be noted that this transformation seems to be promoted by heat, apparently more pyrophosphoric acid being formed by decomposing trimetaphosphimic acid in hot than in cold solution. The second acid of the series, diimidotriphosphoric acid, seems really to have the constitution, PO(OH),.NH.PO(OH).NH.PO(OH), implied by the name. It forms an acid salt with 3, and a neutral salt with 5 atoms of silver, both of which are colorless. If it were a diamide of triphosphoric acid, the penta-silver salt would have 2 atoms united to nitrogen, and such a salt, to judge from all phosphorus-nitrogen salts, where this is known to be the case, would be distinctly colored. Still another intermediate acid is possible, preceding imidodiphosphoric acid, viz, amidoimidodiphosphoric acid, I have been unable to detect this acid thus far, but the too high percentage of nitrogen invariably found. in imperfectly purified imidodiphosphoric acid, may be due to the presence of small amounts of this body. In giving formulas to these bodies, I wish to be understood as using them with full appreciation of their uncertainty. A much more exhaustive study is necessary before they can be regarded as estab lished as firmly as those of many organic compounds. There are several directions in which such a study might be of value. Besides the question of tautomerism, the stereochemistry of nitrogen might be attacked with profit from this side. The analogy of nitrogen and phosphorus suggests the possibility of steric phenomena being shown by the latter also. The formula (b) suggested for trimetaphosphimic acid, for example, may imply a cis- and cis-trans-form depending on the posi |