the proportion by weight in which it combines with some other element, taken as a standard. There is no need, before this distinguished audience, of emphasizing the importance of the familiar table of atomic weights; but a few parenthetical words about their character is perhaps not out of place. As has been more than once said, the atomic weights of the relatively permanent elements, which constitute almost all of the crust of the earth, seem to be concerned with the ultimate nature of things, and must have been fixed at the very beginning of the universe, if indeed the universe ever had any beginning. They are silent, apparently unchanging witnesses of the transition from the imagined chaos of old philosophy to the existing cosmos. The crystal of quartz in a newly hewn piece of granite seems, and probably is, as compact and perfect as it was just after it was formed, eons ago. We can not imagine that any of its properties have essentially changed during its protracted imprisonment; and, so far as we can guess, the silicon and oxygen of which it was made may have existed for previous eons, first as gas, and then as liquid. The relative weights in which these two elements combine must date at least from the inconceivably distant time when the earth" was without form and void."

Although, apparently, these numbers were thus determined at the birth of our universe, they are, philosophically speaking, in a different class from the purely mathematical constants such as the relation of circumference to the diameter of a circle. 3.14159 . . . is a geometrical magnitude entirely independent of any kind of material, and it therefore belongs in the more general class of numbers, together with simple numerical relations, logarithmic and trigonometric quantities, and other mathematical functions. On the other hand, the atomic weights of the primeval elements, although less general than these, are much more general and fundamental than the constants of astronomy, such as the so-called constant of gravity, the length of the day and year, the proper motion of the sun, and all the other incommensurable magnitudes which have

been more or less accidentally ordained in the cosmic system. The physicochemical constants, such as the atomic weights, lie in a group between the mathematical constants and the astronomical "constants," and their values have a significance only less important than the former.

In the lead from uranium, we have a comparatively youthful elementary substance, which seems to have been formed since the rocks in which it occurs had crystallized. Is the atomic weight of this youthful lead identical with that of the far more ancient common lead, which seems to be more nearly contemporary as to its origin with the silicon and oxygen of quartz?

The idea that different specimens of a given element might have different atomic weights is by no means new-it far antedates the discovery of radioactivity.

Ever since the discovery of the definite combining proportions of the elements and the ascription of these proportions to the relative weights of the atoms, the complete constancy of the atomic weights has occasionally been questioned. More than once in the past investigators have found apparent differences in the weights of atoms of a single kind, but until very recently all these irregularities have been proved to be due to inaccurate experimentation. Nevertheless, even thirty years ago the question seemed to me not definitively answered, and careful experiments were made with copper, silver and sodium, obtained from widely different sources, in the hope of finding differences in the atomic weights, according to the source of the material. No such differences whatever were found. More recently Professor Baxter, of Harvard, compared the atomic weights of iron and nickel in meteorites (from an unknown, perhaps inconceivably distant source) and the same terrestrial metals. In these cases also the results were negative. Thus copper, silver, sodium, iron and nickel all appeared to be perfectly definite in nature, and their atoms, each after its own kind, all alike.

The general question remained, nevertheless, one of profound interest to the theoretical

chemist, because it involved the very nature of the elements themselves; and in its relation to the possible discovery of a difference between uranium lead and ordinary lead, it became a very crucial question.

Early in 1913, when the hypothesis of radioactive disintegration had assumed definite shape, Dr. Fajans's assistant, Max Lembert, journeyed to Cambridge, bringing a large quantity of lead from Bohemian radioactive sources, in order that its atomic weight might be determined by Harvard methods, with the precision attainable there. The Carnegie Institution of Washington gave generous pecuniary assistance toward providing the necessary apparatus, in this and subsequent investigations.

The most important precautions to be taken in such work are worthy of brief notice, because the value of the results inevitably depends upon them. The operation consists in weighing specimens of a salt of the element in question, and then precipitating one of the constituents in each specimen, determining the weight of the precipitate, and thus the composition of the salt. In the first place, each portion of substance to be weighed must be free from the suspicion of containing unheeded impurities, otherwise its weight will mean little. This is an end not easily attained, for liquids often attack their containing vessels and absorb gases, crystals include and occlude solvents, precipitates carry down polluting impurities, dried substances cling to water, and solids, even at high temperatures, often fail to discharge their imprisoned contaminations. Especial care was taken that each specimen was as pure as it could be made, for impurity in one would vitiate the whole comparison.

In the next place, after an analysis has once begun, every trace of each substance to be weighed must be collected and find its way in due course to the scale pan. The trouble here lies in the difficulty in estimating, or even detecting, minute traces of substances remaining in solution, or minute losses by evaporation at high temperatures.

In brief, "the whole truth and nothing but

the truth" is the aim. The chemical side of the question is far more intricate and uncertain than the physical operation of weighing. The real difficulties precede the introduction of the substance into the balance case. Every substance must be assumed to be impure, every reaction must be assumed to be incomplete, every measurement must be assumed to contain error, until proof to the contrary can be obtained. Only by means of the utmost care, applied with ever-watchful judgment, may the unexpected snares which always lurk in complicated processes be detected and rendered powerless for evil.

After all these digressions, made in order that the problems concerned should be clearly recognized, let us turn to the main object of our quest. In the present case, each form of lead was first weighed as pure chloride, and the chlorine in this salt after solution was precipitated as silver chloride, the weight of which was determined. Precautions too numerous to mention were observed. Thus the weight of chlorine in the salt was found, and by difference the weight of the lead. From the ratio of weights, the atomic weight of lead was easily calculated.

The outcome of the first Harvard trials, published in July, 1914, brought convincing evidence that the atomic weight of the specimen of uranium-lead from Bohemia is really less than that of ordinary lead, the value found being 206.6, instead of 207.2-a difference of 0.3 per cent., far beyond the probable error of experiment. Almost simultaneously preliminary figures were made public by Drs. Hönigschmid and St. Horovitz and Maurice Curie, pointing toward the same verdict.

This result, interesting and convincing as it was, was only a beginning. Other experimenters abroad have since confirmed it, especially Professor Hönigschmid, who had studied at Harvard and understood the necessary refinements of analysis; and many new determinations have been made at the Wolcott Gibbs Memorial Laboratory, with the assistance of Dr. Charles Wadsworth, 3d, and Dr. Norris F. Hall, upon various samples of lead

from radioactive sources in widely separated parts of the world. Messrs. E. R. Bubb and S. Radcliff, of the Radium Hill Company, of New South Wales, kindly sent a large quantity of lead from their radium mines, and a particularly valuable specimen prepared from selected crystals of pure mineral was put at our disposal by Professor Gleditsch-not to mention other important contributions from others, including Professor Boltwood and Sir William Ramsay. Each of these samples gave a different atomic weight for the lead obtained from them, and the conclusion was highly probable that they contained varying admixtures of ordinary lead in the uranium-radiumlead. This was verified by the knowledge that in at least some cases the uranium ore actually had been contaminated with lead ore. The purest Norwegian specimen thus acquired especial importance and significance, because it was only very slightly, if at all, vitiated in this way. As a matter of fact, it gave 206.08 for the atomic weight in question-the lowest of all. Here are typical results, showing the outcome; many more of similar tenor were obtained.

Hönigschmid, from similar pure material, had found figures (206.05) agreeing almost exactly with the last value. One can not help believing that this last specimen of lead is a definite substance, probably in a state almost pure, because of the unmixed quality of the carefully selected mineral from which it was obtained.

A further question now arises: is it a permanent substance-really an end-product of the disintegration? Soddy's hypothesis assumes that it is. The only important fact militating against this view is the observation that uranium-lead is always radioactive, and hence might be suspected of being unstable.

In various impure specimens, however, the radioactivity is not proportional to the change in the atomic weight; hence the radioactivity is probably, at least in part, to be referred not to the lead itself, but rather to contamination with minute, unweighable amounts of intensely radioactive impurities-other more transitory products of disintegration.2 If weighable, such impurities would almost certainly increase, not diminish, the atomic weight; hence their presence could not account for the low value.

Let us compare the actual result for the atomic weight of this kind of lead with the theory of Soddy and Fajans. If this theory is sound, the simple subtraction of eight times the atomic weight of helium from that of uranium, or five times the atomic weight of helium from that of radium, should give the atomic weight of the lead resulting from the disintegration, as follows:

The agreement is remarkably good. Each of the individual calculated values shows less than 0.05 per cent. deviation from the average, and the average itself shows essential identity with fact a striking confirmation of the theory. This is perhaps the most successful

2 For this reason the term "radio-active lead'' although it describes the fact, is perhaps not from a theoretical point of view the best designation of either uranium or thorium lead; but the term is convenient because it distinguishes between these two forms and common lead.

8 This is the Harvard result. If Hönigschmid's value is given equal weight, the average observed value would be 206.07, exactly identical with the hypothetical value.

attempt on record to compute an atomic weight from hypothetical assumptions. Usually we are wholly at a loss as to the theory underlying the precise relationships, and must determine our values by careful experiment alone.

The value 206.08 for the atomic weight of lead has further support in the fact that it is more nearly half way between thallium, 204, and bismuth, 208, the two neighboring elements in the periodic system, than is the atomic weight 207.2 possessed, by ordinary lead.

It appears, then, that 206, the value pertaining to uranium-lead, is a very reasonable value.

But, as has been repeatedly pointed out, ordinary lead, constituting the vast bulk of the lead in the world, has without doubt a much higher atomic weight, 207.2, not to be expected from either of the lines of reasoning just given. In order to test the uniformity of this circumstance, Professor Baxter, of Harvard, with the help of one of his assistants, investigated ordinary lead from nonuraniferous ores from many parts of the world, and discovered that the constancy of its quantitative behavior is as striking as that of copper or silver. His figures agreed very closely, within the limit of error of experimentation, with those obtained as a part of the present comparison of the two kinds of lead, so that there could be no question as to lack of identity of methods or precautions. Before leaving the subject of the relative atomic weights of these two types of lead, it is not without interest to note the exact absolute weights of the atoms. If, as we have excellent reason for believing on the basis of the brilliant work of Professor R. A. Millikan, a so-called gram-atom (the atomic weight in grams) contains 606.2 sextillion actual atoms, the weights of the atoms of the two kinds of lead must be respectively 342 and 340 septillionth of a gram. Their extreme smallness, as regards bulk, may perhaps best be inferred from the consideration that the smallest object visible as a point in the common microscope has a diameter probably about one thou

sand times as great as an atom of lead.*

Evidently, on the basis of the quantitative results just exhibited, we must admit that there is at least one real difference between radioactive lead and the common metal. Are there other differences?

A question as to the density of each substance, and therefore as to the bulk occupied by the respective atoms, at once arises. Since the atom of uranium-lead weighs less than the other, it must occupy less space, supposing that it has the same density; or else it must have less density, supposing that it should occupy the same space. The identity of the chemical behavior of the two types of lead suggests the probability of the latter alternative, and this was therefore assumed by Soddy; but experimental proof was evidently desirable. Therefore an extended investigation of the density of the various kinds of lead was carried out likewise in the Gibbs Memorial Laboratory. As a matter of fact, the densities of the several specimens were found to be very nearly proportional to their atomic weights; that is to say, the bulk of the atom of radioactive lead is almost exactly the same as the bulk of the atom of ordinary lead, although the weights of these atoms are so markedly different.

A distinctive property of elementary substances, which has always been supposed to be concerned more or less definitely with the atomic weight, is the spectrum, depending upon the wave-lengths of light emitted by the vapor. But, surprisingly enough, the spectrum lines produced by these two sorts of lead, when heated to the high temperature of the electric arc, are so precisely alike, both as to

4 If the smallest object visible in a microscope could be enlarged to the width of this printed page, the atoms in it would appear about the size of the dots on the letters i, or the periods, in the type above.

their wave-lengths and their intensities, that no ordinary spectrum analysis shows any difference whatever. This has been proved by careful experiments at Harvard and elsewhere, and is made obvious by the photographs now thrown on the screen. A and B were from two different specimens of radioactive lead, C from ordinary lead, all very carefully purified. The range covered is about from 3,000 to 2,000 wave-length-far in the ultra-violet. Very recently Professor W. D. Harkins, of Chicago, and two assistants, have detected, with a very extended grating spectrum, an exceedingly minute shift (0.0001 per cent. of the wave length-an amount far too small to be shown by the spectra exhibited) of one of the lines. The wonder is, not that there should be a difference, but rather that they should be so very nearly identical. Evidently the very considerable difference in the atomic weight produces only a barely perceptible effect on the wave-lengths of light emitted by the several isotopic forms of a given element, although a less difference in atomic weight between two different elements (for example, cobalt and nickel) is concomitant with utterly divergent spectra.

Another very interesting question, involving the relations of substance both to light and to weight (or rather density) is its refractive index. All the formulæ relating to molecular refraction involve the density of the substance concerned. In the case under consideration, do the differing weights of the atoms, and therefore the differing densities of the same compounds of the two kinds of lead, affect the refractive indices of the salts? It the refractive index of a given salt of radio lead identical with that of the same salt of ordinary lead? Evidence on this point would go far to decide whether density or atomic volume is the more important thing in determining refractive index. A very careful study carried out with the help of Dr. W. C. Schumb at Harvard has within the past few months shown that as a matter of fact the refractive index or ordinary lead nitrate is identical with that of the nitrate of radiolead within one part in nearly twenty thousand, a result which shows that density is a less important factor

in deterining refractive index than had been previously assumed.

Both of these conclusions concerning lightthat drawn from the spectra and that drawn form the refractive indices-have a yet more far reaching interest, for they give us a further clue as regards the innermost nature of the atom. That part of the atom which determines its weight seems to have, at least in these cases, very little effect on that part of the atom which determines its behavior toward light.

Immediately connected with the question of density of the solid salts is the question as to the densities of their saturated solutions, as well as to the extent of saturation. Fajans and Lembert had recently obtained results probably indicating that the molecular solubility of each kind of lead is the same, and that the densities of the solutions are different, the density of the radiolead solution being less to an extent consistent with its smaller molecular weight. These results, however, left much to be desired in the way of accuracy, and needed verification. Therefore a very careful investigation, begun at Harvard with the assistance of Schumb, before the appearance of Fajan's publication, furnished valuable knowledge on this point.