found by taking the square root of the total number of counts contained in a rectangular area whose width is determined by the width of the Gaussian at the point where it ceases to coincide with the machine-drawn peak. to be the actual area compared. This area is presumed The statistical error found in this way varies widely (because of the great range in count rate), but for the peaks tested it varles betweon 2 and 6 percent. A hindrance to accurate superimposition of the Gaussian curves on the total absorption peaks, contributing further = to the error, may be the slight non-Gaussian character of the total absorption peak caused by the lack of perfect homoseneity -- 4 of the crystal's light production. | | To get some idea of how much of each spectrum occurred in analyzable peaks,the total areas of several spectra were measured with the planimeter and compared with the sums of tne total areas associated with each listed energy (found by | dividing ratio). | each peak area by the appropriate peak-to-total It was found by this method that from 70 to 90 per- cent of a spectrum was accounted for in the peak-by-peak analysis. The errors associated with the photon intensities shown in Table 2 are the result of estimates of the cumulative errors discussed above. They range from 3 percent for peaks with good resolution, good count rate,and good fit with the 23 TreeOe TER TS ere me en |