470 Royal Society : — 



of about 40 inillionths of an atmosphere, when the repulsive force 

 is near its maximum, a spark, whose striking-distance at the normal 

 pressure is half an inch, will illuminate a tube having aluminium 

 terminals 3 millimetres apart. When I push the exhaustion further, 

 the |-inch spark ceases to pass ; but a 1-inch spark will still illu- 

 minate the tube. As I get nearer to a vacuum more power is 

 required to drive the spark through the tube : but at the highest 

 exhaustions I can still get indications of conductivity when an in- 

 duction-coil actuated with five Grove's cells, and capable of giving 

 a 6-inch spark, is used. 



When so powerful a spark is employed there is great danger of 

 perforating the glass, thus causing a very slight leakage of air 

 into the apparatus. The log. dec. now slowly rises, the repulsive 

 force of the candle increases to its maximum, and then slowly 

 diminishes to zero, the log. dec. continuing to rise till it shows 

 that the internal and external pressures are identical. With a 

 line perforation several days are occupied in going through these 

 phases, and they take place with such slowness and regularity as 

 to afford opportunities for getting valuable observations. 



The improvements now added by Mr. Grimingham to the pump 

 render it so easy to obtain high exhaustions, that, in preparing 

 experimental radiometers, I prefer to exhaust direct to one or two 

 millionths of an atmosphere. By keeping the apparatus during 

 this exhaustion in a hot-air bath heated to about 300° C. for some 

 hours, the occluded gases are driven off from the interior surface 

 of the glass and the fly of the radiometer. The whole is then 

 allowed to cool, and attenuated air from the air-trap is put in in 

 small quantities at a time, until the McLeod gauge shows that the 

 best exhaustion for sensitiveness is reached; if necessary, this point 

 is also ascertained by testing with a candle. Working in this way, 

 I can now do in a few hours what formerly required as many days. 

 In this manner, employiug hydrogen instead of air for the gaseous 

 residue, and using roasted mica vanes set at an angle with the 

 axis, as described further on, I can get very considerably increased 

 sensitiveness in radiometers. I am still unable, however, to get 

 them to move in moonlight. The statements made by an observer 

 nearly a year ago, that he obtained strong rotation by moonlight, 

 must therefore be considered erroneous. My most sensitive torsion- 

 balance will, however, move easily to moonlight. 



The above-mentioned facts, in addition to what has already been 

 published, leave no reasonable doubt that the presence of residual 

 gas * is the cause of the movement of the radiometer. But few 

 theories are sufficiently strong not to require reinforcement ; and 

 in the present case very much remains to be ascertained as regards 



* It is a, question whether the residual gas in the apparatus, when so 

 highly attenuated as to have lost the greater part of its viscosity, and to 

 be capable of acquiring molecular movement palpable enough to overcome the 

 inertia of a plate of metal, should not be considered to have got beyond the 

 gaseous state, and to have assumed a fourth state of matter, in which its pro- 

 perties are as far removed from those of a gas as this is from a liquid. 



