318 



BELL SYSTEM TECHNICAL JOURNAL 



much more convenient to plot the differences between each measured 

 mass M and the nearest integer, which latter is the "mass-number" 

 A of the kind of atom in question. Aston prefers to use the quantities 

 10^(M — A) I A, the differences aforesaid divided by the corresponding 

 mass-numbers and "expressed in parts per ten thousand"; these he 

 calls the "packing fractions" in allusion to the principle that ele- 

 mentary particles suffer changes in mass when they are clustered or 

 packed closely together. 



If one plots either {M — A) or the packing fraction against yl, it is 

 immediately obvious that the values of either do not jump about at 

 random as one progresses along the procession of the atoms in order 

 of atomic mass. The packing fractions lie pretty closely along the 

 sweeping curve in the lower part of Fig. 8, with its odd bifurcation 

 (ordinates on the left!). Not all the available data are represented 



100 120 140 



MASS NUMBERS 



Fig. 8 — Deviations of atomic masses from mass-numbers (upper curve) and 

 packing-fractions (lower curve, with points of observation). Curves from the report 

 of a Royal Society discussion of 1929, with subsequent observations filled in from 

 Aston and from Bainbridge. 



by dots, as some would fall too close to others to be distinguishable on 

 this scale. The curve of {M — yl) is the full curve sketched above 

 (ordinates on the right !).^^ 



As the trend of either curve makes clear, the masses of the atoms 

 near either end of the procession, the "light" end and the "heavy" 

 end, exceed their nearest integers; while all through the middle (and 



^^ The packing fractions from one end of the curve to the other are mostly un- 

 certain by from one to three units, excepting those of H^, H^ and a few other very 

 light atoms. The uncertainty of {M — A) increases steadily with A, as the reader 

 will easily understand; it is indicated by the space between the dashed curves. This 

 is a reason for preferring packing-fraction to {M — /I) as a quantity for plotting. 



The data omitted in Fig. 8 are: - 5 for Cs"3, -f 2 for TPos, - ^ for Se^". 



