Jean Charles de Marignac (1817-1894)

M. Stas has just published a résumé of the researches to which he has devoted himself for a great number of years, researches made with the object of verifying whether simple relations really exist amongst the atomic weights of various elements.

Through the determination of the atomic weight of carbon made in 1839 and 1840 conjointly with M. Dumas,[1] he was initiated into the difficulties of such work and the minute precautions necessary for arriving at rigorously exact results, and he has neglected nothing to give these new researches every requisite degree of accuracy. He has thought too, and rightly thought, that it was not enough to have adopted all these precautions; he has described them in detail in the account of his experiments in order that every chemist should be in a position to judge of the degree of confidence to be accorded to the results he has obtained.

We must signalise this memoir, not only because of the importance of the subject treated, but also for the numerous lessons it affords to chemists who wish to devote themselves to work requiring extreme accuracy. They will find in it most valuable instructions and a model to follow.

But we cannot in this extract linger over these details, which should be read in the memoir of the author, and we must content ourselves with discussing the principal results at which he has arrived.

Although these researches have been extended to a considerable number of elements, the memoir cited only contains those relative to chlorine, sulphur, nitrogen, silver, potassium, sodium, and lead, these elements having been chosen as the best known, forming the most stable compounds, and generally because they have been considered as obeying Prout's law. The aim of the author was not so much to determine their atomic weights as to ascertain by the study of the most sharply defined actions that can be carried out between them, if there exists amongst these weights the simple ratios assumed by Prout's law.

To attain this end the author has performed the synthesis of the chloride, sulphide, and nitrate of silver, of the nitrate and sulphate of lead, the analysis of chlorate of potash and of sulphate of silver, and finally he has determined the ratios of silver to the chlorides of potassium, sodium, and ammonium, and those of nitrate of silver to the chlorides of potassium and ammonium. Each of these determinations has been made as far as possible by several different methods, and in every case with the aid of products prepared or purified by different procedures described with all necessary detail.

The results arrived at by the author in these twelve series of determinations are summarised in the following table. The table contains the gravimetric ratios of the two substances compared in each series, confronted with the ratio calculated from Prout's hypothesis, and in addition the record of the maximum and minimum values obtained in the various series, so that the reader can judge from the deviation of these two figures, to what extent the differences between the experimental results and the calculated values can be attributed to errors of observation. The calculation has been made by adopting the number 35.5 for chlorine and 103.5 for lead.

	min. 32.841
Ag:Cl = 100:	max. 32.850	[M. 32.854][2]
	mean 32.8445	(calc. 32.870)
	min. 14.849
Ag:S = 100:	max. 14.854
	mean 14.852	(calc 14.814)
	min. 157.463
Ag:AgO,NO5 = 100:	max. 157.481	[M.157.455]
	mean 157.473	(calc. 157.404]
	min. 69.099
Ag:KCl = 100:	max. 69.107	[M. 69.098]
	mean 69.103	(calc. 68.981)
	min. 54.206
Ag:NaCl = 100:	max. 54.209
	mean 54.2078	(calc. 54.166)
	min. 49.590
Ag:NH4Cl = 100:	max. 49.600	[M. 49.556]
	mean 49.5944	(calc. 49.537)
	min. 43.864
AgO,NO5:KCl = 100:	max. 43.885	[M. 43.878]
	mean 43.876	(calc. 43.823)
	min. 31.486
AgO,NO5:NH4Cl = 100:	max. 31.490
	mean 31.488	(calc. 31.470)
	min. 159.959
Pb:PbO,NO5 = 100:	max. 159.982
	mean 159.969	(calc. 159.903)
	min. 146.419
Pb:PbO,SO3 = 100:	max. 146.443
	mean 146.427	(calc. 146.376)
	min. 60.838
KO,ClO5:KCl = 100:	max. 60.853	[M. 60.839]
	mean 60.846	(calc. 60.8163)
	min. 69.197
AgO,SO3:Ag = 100:	max. 69.209
	mean 69.203	(calc. 69.230)

We see that the differences between the results obtained by M. Stas and those which Prout's law requires, although inconsiderable, are yet very much greater (six to eight times, and in two cases even thirteen or fourteen times greater) than the greatest differences between the results found in each series of experiments.

On the basis of these observations M. Stas calculates for the elements with which he has experimented, the following atomic weights referred to that of oxygen as 8. I place beside them the numbers which I deduced from my experiments of 1843 for some of the same elements, and also the numbers adopted by the partisans of Prout's law.

	Stas.	Marignac.	Prout.
Silver	107.943	107.921	108
Chlorine	35.46	35.456	35.5
Potassium	39.13	39.115	39
Sodium	23.05		23
Ammonium	18.06		18
Nitrogen	14.04	14.02	14
Sulphur	16.037		16
Lead (from sulphate)	103.453		103.5
Lead (from nitrate)	103.460		103.5

The atomic weight of silver has been derived in this calculation from all the experiments involving this metal, chlorine, and potassium, to which oxygen is related by the analysis of chlorate of potash. But it may also be obtained by starting from another series of experiments completely independent of these, namely from the synthesis of the sulphide of silver and the analysis of sulphate of silver. We then find

Silver	107.924
Sulphur	16.029

M. Stas observes that there results from his experiments a difference between the atomic weights of NH₄ and of nitrogen which is equal to 4.02 instead of 4. He does not believe that this difference can result from the imperfections of the experimental methods; he concludes that the ratio of the atomic weights of hydrogen and oxygen is not exactly 1:8, and proposes to submit this question to further study. This is perhaps exaggerating a little the degree of confidence to be placed in determinations of this kind, for the difference does not greatly exceed that which exists between the atomic weight of sulphur obtained by the author in two distinct series of experiments.

If I have recorded alongside the numbers of M. Stas those which I had previously obtained, it is not merely for the purpose of drawing attention to their close agreement; it seems to me that an important conclusion may be drawn from them. I recognise fully, after having studied the splendid work of this savant, that he has exercised in his experiments infinitely greater care than I thought to be necessary not only in the purification of the substance used in the researches, but also in the accuracy of his weighings, and in taking all imaginable precautions for the purpose of eliminating errors. His results therefore offer much greater guarantees of accuracy than mine; yet we see how little they differ from them, and it is specially noticeable that in the mean they are no nearer than mine to the numbers calculated from Prout's law. It seems to me permissible to conclude that if, after fresh improvements have been made in the methods of purification of the substances or in the experimental methods, some chemist at a future date takes up the same series of experiments with still greater guarantees of accuracy, the difference which may be found between his results and those of M. Stas will probably be of the same order as that which exists between the latter and my own, and that no greater accordance with Prout's law will be found.

This being my opinion on the question, it may perhaps be found astonishing that I do not entirely agree with the conclusions of M. Stas that "we must consider Prout's law as a pure illusion, and regard the undecomposable bodies of our globe as distinct entities having no simple relation by weight to one another." I may be permitted some observations on these conclusions which appear to me to be too absolute; they will bear on two separate points.

First, I confess that I shall not be convinced of the accuracy of an atomic weight determination (or rather, I shall not be able to form a clear opinion of the degree of confidence to be accorded to it) unless this weight has been obtained by several absolutely independent methods, based on the analysis of a number of compounds altogether distinct from one another. I recognise that this may be a very troublesome condition to fulfil, but I believe it to be a point which should above all be attended to by chemists who desire to work anew at these difficult questions. Even in the experiments of M. Stas I see two methods of arriving at the atomic weight of silver; one gives 107.943, the other 107.924. But that is not all. I should like, for example, that this savant, exercising all the care and perspicacity which he devoted to his researches, might take up the analysis of silver salts of organic acids which gave me the figure 107.968 as the atomic weight of silver.[3] Perhaps he might find yet other methods of arriving at the same end, and we should then see if the results yielded by the various methods differ from one another by a less amount than the differences found between these results and those calculated by means of Prout's hypothesis.

I expressly add that by different methods, I mean methods which depend on the analysis or synthesis of absolutely distinct compounds, and not merely those which differ only in the means of bringing the same substances into chemical action. Thus when in my first research I invoked as a proof of accuracy the coincidence of the ratio observed between silver and chloride of potassium on the one hand, with, on the other, the same ratio calculated from other experiments giving first the direct ratio of chlorine to silver, and second that between chloride of silver and chloride of potassium, or when M. Stas invokes as a control of the synthesis of nitrate of silver the experiments by which he determined the ratio between this nitrate and chloride of silver, which is itself directly connected with silver, I only recognise in these a confirmation of the exactness with which the experiments were made, and in no wise a confirmation of the experimental method itself. Indeed, if for any reason the nitrate of silver prepared under normal conditions does not contain its elements rigorously in the proportion of their atomic weights, the most exact methods applied to its analysis or its synthesis will show the same error in the ratio of these weights.

This is indeed the principal cause of the doubt which still exists in my mind. I do not regard it as absolutely demonstrated that many compounds do not contain, constantly and normally, an excess of one of their elements, an excess doubtless very minute, but still sensible in very delicate experiments. To make this idea clear I will cite a case in which this tendency is shown, though only to such a degree that it could scarcely pass unrecognised.

It has long been held that monohydrated sulphuric acid[4] is a perfectly stable compound, as might be expected from the powerful affinity of its two components. Yet I have shown[5] that it is on the contrary not at all stable. The slightest rise of temperature leads to the disengagement of the vapours of anhydrous sulphuric acid, and only when it has been brought by this process to contain a slight excess of water (about 1 per cent.) does it become perfectly stable and unaltered by distillation.

Are we justified in affirming that it is not in the same way a normal condition, so to speak, of the existence of a compound, one of whose components is very volatile or even gaseous, always to contain an excess, often perhaps scarcely appreciable, of the fixed component, the affinity of that free portion being necessary to counterbalance the elasticity of the gaseous substance?

Or perhaps conversely, just as hydrated sulphuric acid retains about 300 degrees a slight excess of water, which is itself vaporisable at 100 degrees, may it not be that a compound such as the sulphide or nitrate of silver should invariably contain, even at very elevated temperatures, a slight excess of sulphur or of nitric acid?

Lastly, are we sure, when we weigh a substance, that we have obtained really and uniquely the weight of that substance? We know that fused silver and fused litharge retain in solution a perceptible amount of oxygen. We assume that they lose it completely at the moment of solidification, but I do not believe that anyone has ever assured himself of this by rigorous experiment; and in any case this has not been done for all other substances. Even less are we assured of the impossibility of their having condensed atmospheric nitrogen.

These possible causes of error, to which we could probably add others, lead me, although I am satisfied that the experiments of M. Stas are perfectly exact, to be not absolutely convinced that the differences observed between his results and those required by Prout's law cannot be explained by the imperfect character of the experimental methods.

But assuming now that these objections are without foundation and that the experiments give us in reality the ratios of the atomic weights of the elements, there remains another objection which prevents me from concluding from them that Prout's law is nothing but a pure illusion, and that the elements are entities necessarily distinct.

One need, in fact, only glance at these weights to see that if they do not coincide absolutely with the numbers of Prout, they are so close to them that it is impossible to consider this fact as accidental. Thus for the nine determinations resulting from this work of Stas, the difference is as a mean 0.056, say 1/18 of the equivalent of hydrogen, or 1/9 of the half-equivalent. It is known besides that a concordance at least equal is exhibited for other elements whose atomic weights seem to have been determined with some certainty, such as iron, calcium, carbon.

We may thus say of Prout's law what must be said of the laws of Mariotte and of Gay-Lussac regarding variations in gaseous volumes. These laws, for long considered absolute, have been recognised as inexact when the experiments have been carried out with the degree of precision attained by M. Regnault, M. Magnus etc. Nevertheless they will always be considered as expressing natural laws not only from the practical point of view, since they allow us to calculate in most cases with a sufficient approximation the changes in volume of gases, but even from the theoretical point of view, for they very probably express the normal law of these changes of volume, apart from certain disturbing influences whose effect we shall perhaps also be able at a later date to compute.

We may believe that it is the same with Prout's law: even if it is not rigorously confirmed by experience, it none the less appears to express the ratios between the atomic weights of the elements with sufficient accuracy for the practical calculations of chemistry and perhaps also the normal ratios which should exist between these weights, apart from some perturbing causes the elucidation of which may in future exercise the sagacity or the imagination of chemists.

While preserving the fundamental principle of that law, that is to say, while adopting the hypothesis of the unity of matter, could we not make for example the following supposition, to which I do not otherwise attach importance except as showing that we might explain the discordance that appears to exist between the results of observation and the immediate consequences of that principle? Could we not suppose that the cause (unknown but probably different from the physical and chemical agencies familiar to us) which has determined certain groupings of the atoms of the sole primordial matter so as to give rise to our chemical atoms, by impressing on each of these groups a special character and particular properties, should not at the same time have exercised an influence on the manner according to which these groups of primordial atoms would obey the law of universal attraction, in such wise that the weight of each group might not be exactly the sum of the weights of the primordial atoms composing it?

To conclude, I must not forget an important observation due to the illustrious chemist who has so forcibly called attention to these interesting questions. The fundamental principle which led Prout to formulate his law, namely the idea of the unity of matter, and all the more or less brilliant conceptions which have been based on this principle, are altogether independent of the magnitude of the unit which serves as common divisor to the atomic weights of the elements, and which might therefore be considered as expressing the weight of the atoms of the primordial matter. Whether this weight be that of a single atom of hydrogen, or of a half or a quarter atom or whether it be an infinitely smaller fraction, say a hundredth or a thousandth, all these considerations would none the less preserve for it the same degree of probability. There would only result less simple ratios of composition amongst the various elementary components.

C.M.

[2][Figures in brackets following an M. represent results Marignac had obtained earlier. He quoted these results in a footnote in this paper. --CJG]

[3]Bibl. Univ. 1846 (Archives), vol. 1, p. 57. [Original note.---CJG]

[4]Marignac means what we should now call simply sulphuric acid (HO,SO₃ in the contemporary notation). [Alembic Club note.--CJG]

[5]Bibl. Univ. 1853 (Archives), vol. 22, p. 225. [Original note.--CJG]