MICROWAVE PARAMAGNETIC RESONANCE ABSORPTION 461 



The latter was used in the experimental part of this paper. From (28) 

 we see that with a 1 cm setup the number of observable electrons should 

 be approximately 40 times less than with a 8-cm setup. However, in 

 most practical cases one is not limited by the amount of a\-ailable 

 sample, since one usually can increase the sample size at longer 

 wavelengths. Therefore a better criterion is the minimum number of 

 electrons per unit volume 



Keeping now the filling factor Vc/Vs constant we see that the sensi- 

 tivity of a 3 cm setup as defined in (29) is only -y/s worse than of a 1 cm 

 setup. In addition the power outputs of .3-cm klystrons are usually 

 sufficiently higher than those of 1-cm klystron to overcome even the 

 -\/3 advantage. If RF saturation comes in, the power argument is not 

 valid, but one has to consider the RF magnetic field Hi inside the cavity 

 which for a given power and Qo is prop, to co. Thus at the higher fre- 

 quencies (1 cm) saturation effects become more pronounced reducing 

 again the advantage of a 1-cm over a 3-cm setup. In deciding the choice 

 of the frequency in special cases (e.g., when Aco is a function of the mag- 

 netic field, or the sample is larger than a skin depth) (29) should be used. 

 There are, of course, considerations, other than those of max. sensi- 

 tivity, which have to be taken into account. For example one would 

 always like to satisfy the condition Aco/co <<C 1 which favors higher fre- 

 quencies. On the other hand, for very narrow lines (say less than 0.1 

 oersteds) fractional field instabilities and inhomogeneities will favor 

 low magnetic fields, i.e., lower frequencies. One also might encounter 

 samples which exhibit an excessive loss in a given frequency band which 

 therefore has to be avoided. 



3. Optimum Amount of Sample to be Used 



The output voltage AF is proportional to the sample volume and 

 the unloaded Qq of the cavity, see (26). If the sample is lossy an increase 

 in its size will reduce the Q and therefore reduce the signal. We may 

 roughly distinguish two limiting cases. In one case the losses are pro- 

 portional to E- (e.g., high resistivity samples having dielectric losses), 

 in the other case they are proportional to Hi (e.g., low resistivity sam- 

 ples in which the losses are due to surface currents). 



a. Losses Proportional to E 



The paramagnetic sample is placed in the region of max. RF magnetic 

 field for instance at the end plate of a rectangular cavity resonating in 



