924 J, A. POLLOCK, E. M. WELLISCH AND A. B. B, RANCLAUD. 
symbols. All the observations were taken with the same 
pair of carbons so the curves are strictly comparable. 
Case 4 is the one just discussed in section 3. It is seen 
that for small time intervals between the break and the 
make of the circuit it requires greater potential differences 
to restart the arc in the cases 1, 2 and 3 than in that of 4, 
and that cases 2 and 3 approximate to that of 4 for large 
intervals of time. Considering that the are develops from 
a small non-luminous current of negative ions, an idea 
which we think must form the basis of any explanation of 
the critical relighting, in searching for a description of the 
differences between the curves it has to be noticed that in 
some cases the negative electrode, on the remake of the 
circuit, is hotter than the positive, in others the reverse; 
it is also essential to recognise that in some instances the 
negative stream of ions is opposed by the convection current 
of hot gas, while in others it is helped by it, as it has been 
shown in the previous paper that a change in the relative 
directions of the stream and current considerably affects 
the potential difference necessary for arcing. 
The feature of the relations is the evidence, shown by the 
curves for cases 2 and 3, of a critical change in the con- 
ditions for relighting occurring when the potential differ- 
ence attains the value of 90 volts. After reaching this 
value the minimum potential difference requisite to start 
the are remains for some time practically constant in spite 
of the fact that for increasing time intervals between the 
break and the make of the circuit the temperatures of the 
carbons are diminishing. 
The excess of the potential differences for relighting 
required in case 1 over those in case 4, for the same time 
intervals, may be accounted for, perhaps, by the much 
lower temperature of the negative electrode on the reclos- 
ing of the circuit in the former instance, but we have been 
