CONTEMPORARY ADVANCES IN PHYSICS 317 



to disperse again. It amounts to about 115 millions of electron-volts, 

 and this is not an unwelcome figure, for had the value been much 

 smaller we might expect oxygen nuclei to be easily disrupted, which 

 is not the case. 



This evidently makes an extra reason for measuring atomic masses 

 with the utmost care: not only are these masses important in them- 

 selves as constants of Nature, they may also be used as indices of the 

 stability or the fragility of the various kinds of nuclei. Aston's first 

 apparatus enabled him to measure them to one part in a thousand, 

 an accuracy which may be valuable among the lightest elements but 

 not among the heavier, where the uncertainty rises to one fifth of a 

 unit of mass. His second apparatus proved itself competent to one 

 part in ten thousand, and with its completion in 1925 the second period 

 of isotope-analysis began. Bainbridge in measuring the ratio of He* to 

 H^ pushed onward to a precision severalfold greater, claiming a 

 probable error of only one part in a hundred thousand. With such 

 data as these, it is necessary at times to take account of the fact that 

 what is measured is the ratio of masses of two ions, the unknown and 

 the O^® ion ; what is tabulated is usually the ratio of the corresponding 

 atoms; but what is required for nuclear theory is the ratio of the 

 masses of the nuclei. Even with contemporary accuracy, though, 

 the correction is still trivial unless the very lightest atoms are in- 

 volved.^** It should be mentioned here that band-spectra occasionally 

 permit the ratio of the nuclear masses of two or more isotopes of the 

 same element to be evaluated, with an accuracy which may attain 

 (in the case of the ratio C^^C^^) one part in ten thousand. 



Not nearly all of the known kinds of atoms have had their masses so 

 precisely measured. Suitable data exist for nearly all of the isotopes 

 of the first ten elements; beyond these there are but twenty-four 

 elements of which even a single kind of atom has been measured, and 

 deplorable gaps between them. 



How best to plot these data? This is a difificult problem. Con- 

 sidering the inchoate state of nuclear theory, it would probably be 

 best to plot the measured masses directly, as in Fig. 6 — were it not 

 that then the graph would have to be as large as a wall-map. It is 



1" This is due not entirely to the smallness of the electronic mass, but partly to 

 the fact that the ratio of nuclear mass (in standard units) to number-of-orbital- 

 electrons is always between 2 and 3 for all kinds of atoms excepting H^ for which 

 it is about one. 



Aston until 1930 published his estimates of atomic masses coupled not with their 

 probable errors, as the custom is, but with the extreme limits outside of which 

 (in his opinion) the value of the mass in question cannot possibly lie — an unusually 

 conservative policy, because of which some people who have used his values have 

 underestimated their probable accuracy. The ratio of these "uncertainties" to the 

 probable errors is commonly taken, with Aston's concurrence, as three. 



