12(3 DE. W. Cr. DUFFIELD, MESSES. THOS. H. BUENHAM AND A. H. DAVIS ON 
current of 8 amperes and 3 mm. arc length, a deflexion of 30 degrees was recorded. 
When the current was turned oft’ and the poles were hot, balance was obtained when 
the deflexion in the opposite direction was 17 degrees. The deflexion corrected for 
convexion"^ is therefore 47 degrees, corresponding to 1’22 dynes. The couple due to 
convexion is in this experiment approximately 36 per cent, of the total couple upon 
the poles, and it corresponds to a negative pressure of — 0‘44 dyne. 
It had been hoped that the arrangement of an arc at each end of the suspended 
carbon (double-arc method, fig. 4), besides eliminating the electromagnetic effects, 
would obviate the convexion current difficulty, but this was not the case, since Davis 
found that moving the vertical arc from the centre to the end did not necessitate 
more than a minor change in the current strength necessary to produce a given 
deflexion. No doubt the currents about the vertical arc were very different from 
those about the horizontal arc at the tip of the carbon rod. 
The above experiments enable us to fix the lower limit to the effect of the 
convexion currents rising from a 7-8-ampere arc at — 0'5 dynes. 
Since it is a matter of considerable difficulty to measure convexion currents in the 
manner described, a comparative investigation of their value over the range of 
current strengths used in the main research was made thus :— 
A light paper vane was pivotted upon a vertical axis 60 cm. above a horizontal 
arc, so that the rising convexion currents caused it to turn, the number of revolutions 
of the vane in 1 minute providing a measure of their velocity. Arcs of constant 
length and varying current were employed, and the mean curve C in fig. 21 was 
obtained as the result of a number of experiments, whence it is clear that over 
the range 2 to 10 amperes there is an increase in the velocity of the convexion 
currents. For a given curvature of pole the suction upon it is proportional to the 
square of the velocity of the air moving past it, the curve D has therefore been 
drawn with ordinates proportional to the square of those of curve C, but on a 
different scale, it represents, therefore, the pressure effect of the convexion currents. 
We can approximately fix the scale of curve D from the knowledge that the 
convexion is approximately O'5 dynes at 8 amperes, remembering however that this 
is an underestimate. 
* Also radiometer action if it were effective, see p. 1.30. 
