1890.] The Spectroscopic Properties of Dust. 439 



due to the dust, be ascribed to mere faintness in their light, for we 

 took photographs of the spectrum of B, and found that even length- 

 ened exposures produced no evidence of rays due to the dust ; nor 

 could it be ascribed to the character of the discharge in B, for the 

 discharge was varied ; sometimes A and B were in the same circuit ; 

 sometimes the discharge in B was from a separate coil, and even the 

 powerful discharge from a large coil stimulated by a De Meritens' 

 magneto- electric machine, was tried. 



That abundance of dust was formed by the sparking in A was 

 proved not only by the deposit in the tube, but by allowing the 

 stream of gas at atmospheric pressure from the tube h (of course 

 disconnected from the pump) to impinge on a flame, when the charac- 

 teristic flame-spectrum of the electrodes in A was at once manifest. 

 When the gas used was hydrogen, and it was burnt in oxygen, the 

 spectrum of the electrodes was particularly well seen ; also when 

 the gas was oxygen and led into a hydrogen flame. 



That the dust was of extreme fineness and capable of being carried 

 by a stream of gas to a great distance was proved as follows : A 

 stream of hydrogen, at ordinary pressure, was passed through the 

 sparking tube with magnesium electrodes, and then through more 

 than 100 feet of metal tube in a coil, and, finally, burnt as it issued. 

 Before the sparking began there were no signs of magnesium in the 

 flame ; but when sparks had been passing between the magnesium 

 electrodes for a short time, the magnesium spectrum was seen in the 

 flame. It took 55 seconds for the gas to carry the dust througn the 

 long pipe, and when the sparking ceased it was again about the same 

 time before the magnesium disappeared from the flame. It always 

 appeared and disappeared sharply in correspondence with the spark- 

 ing. Similar experiments, but with a shorter tube, were made with 

 other metals, iron, sodium, lithium, &c., always with like results ; 

 also a current of oxygen was passed through the sparking tube and 

 into a flame of hydrogen, and produced similar effects. Even alu- 

 minium, which does not usually show any part of its spectrum when 

 used as an electrode in a vacuous tube, gave, when sparked in oxygen, 

 dust which, when carried into a hydrogen flame, showed the charac- 

 teristic bands of alumina. 



Considering that a sensible amount of dust was deposited in the 

 bulbs of B, we should have expected that some would be deposited on 

 the electrodes e, e in that tube, and that the discharge from electrodes 

 so coated would give the spectrum of the metal on their surface. 

 There is no doubt that when no discharge was taking place in B the 

 electrodes e, e did receive their share of dust ; and, if it had been 

 allowed to accumulate so as to form a coherent crust, it would have 

 given its characteristic spectrum on first passing sparks in B. But, 

 so long as the dust is loose, the passage of a discharge instantly clears 



