584 Prof. 0. W. Richardson and Mr. K. T. Compton on 



Having thus obtained the two end points, an estimate of the 

 mean energy can be obtained by assuming that the shape of 

 the curves is the same as that of those given by other sub- 

 stances which terminate at the same points. This method 

 was tested in a number of other cases, where all the data 

 were known, and found to give results for the mean energy 

 which agreed with the direct determinations. Moreover, in 

 almost all of the very large number of cases tried the curves 

 between the same terminal points were identical to within 

 the limits of experimental error. The curve which showed 

 the worst agreement with this rule is the aluminium curve 

 for \=31'3, which runs across from the platinum curve at 

 X=23 to that at \ = 25. This is shown in PL XIII. fig. 2. 

 It is to be remembered, however, that a wave-length A, = 31*3 

 gives very small currents even with aluminium, so that the 

 curves in this part of the diagram cannot be determined very 

 accurately. The more typical behaviour is shown by the 

 curves for the wave-length 21 for Pt and 25*4 for Al, which 

 are almost identical throughout their course (PL XIII. fig. 2) . 



The Maximum and Mean Energies. Comparison 

 with Theory. 



The maximum and mean energies of the electrons emitted 

 under the influence of light of a given frequency are readily 

 obtained from curves like those shown in PL XIII. fig. 3. The 

 maximum energy T m , expressed in equivalent volts, is clearly 

 equal to the intercept on the voltage axis between the point 

 of intersection of the curves with this axis and that of the 

 true axis of zero volts. The mean energy in the same units 

 T r is equal to the area which is bounded on the left by the 

 curve, on the right by the true axis of zero volts, and below 

 by the voltage axis, divided by the maximum current. The 

 values of the maximum energy in this way are collected 

 together in the following table (p. 585). 



Leaving out of account for the present certain possible 

 sources of systematic error, which will be considered more 

 fully below, and which are likely to affect both the deter- 

 minations of the mean and of the maximum energy in almost 

 equal proportions, we are inclined to place more reliance on 

 the measurements of the mean than on those of the maximum 

 values. In some cases the current-voltage curves (PL XIII. 

 figs. 2 & 4-8) appear to approach the voltage axis quite 

 gradually, so that the determination of the exact point of 

 intersection is a very difficult matter. Our electrostatic 

 arrangements were quite sensitive and worked very satis- 

 factorily ; but the currents in this region are extremely 



