MICROWAVE PARAMAGNETIC RESONANCE ABSORPTION 477 



with more than 10^ spins agreed within a few percent with the calcu- 

 lated value. The calculated value being based on the known amount 

 of D.P.H. and the measured value being referred to the known amount 

 of CUSO4 -51120. D.P.H. samples with less than lO'^ spins had all a 

 smaller number of effective spins than calculated. The discrepancy was 

 more pronounced the smaller the sample. There Avas also evidence that 

 the smaller D.P.H. samples deteriorated with time. As a typical example 

 we quote a sample which started out as 10 " effective spins and was 

 reduced after 4 wrecks to 4 X 10 effective spins and another one which 

 initially had 10^^ spins, deteriorated in the same time interval to 10^^ 

 spins. Since only the smaller samples were noticeably affected, this 

 deterioration seems to be associated with a surface reaction. It was also 

 observed that the line width between inflection points of the D.P.H. 

 samples with less than 10 spins increased from 1.8 oersteds to 2.7 

 oersteds. This broadening probably arises from a reduction in the ex- 

 change narrowing mechanism due to the spreading out of the sample. 



S. Comparison of experimental results with theory 



In checking the sensitivity of the equipment D.P.H. samples were 

 used and the signal to noise was estimated from the recorded output. 

 The experimental points thus obtained are shown in Fig. 7 and Fig. 9. 

 We believe that the results are significant to within a factor of 2, the 

 main error arising from the estimate of the RMS noise. The band width 

 of the lock-in detector was Av = 0.03 sec~\ Qo = 4,000, and the field 

 modulation used was 3 oersteds p.t.p., 100 c.p.sec. for the barretter 

 schemes and 1,000 c.p.sec. for the crystal schemes. This large modula- 

 tion field somewhat distorts the line, but, as mentioned earlier was done 

 in order to get the full signal. The D.P.H. samples were calibrated 

 against CuS04-5H20 before each run. Even so it was not felt safe to 

 use samples which had less than 10^^ spins. 



Referring to Fig. 8 we see that for the straight barretter detector the 

 experimental points agree fairly well with the predicted value, l)ut in the 

 balanced mixer scheme fall short by about a factor of 4. A possible ex- 

 planation of this discrepancy is that the barretters were not completely 

 matched in which case the noise from the local oscillator would not be 

 compensated for. 



Fig. 9 shows the experimental points for the crystal schemes. For 

 powers between 10~ W and 10"^ W the system used fell between the 

 simple straight detection scheme and the one utilizing optimum RF 

 bucking. The reason is that it was very easj^ to obtain a certain amount 

 of positive bucking (+9 db) by merely adjusting one arm of the magic T. 



