2 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 66 



tiiuially decreased for as much as a minute thereafter. The decrease 

 in 30 seconds was a very considerable part of the who^e sky-deflection. 

 Our original method of reading thus proved quite arbitrary, for there 

 was no reason to suppose the reading at 30 seconds after exposure 

 was better than at 20 or 40 seconds. A similar drift of the galvanome- 

 ter but very small relatively to the whole deflection, was observed 

 after the lapse of about 20 seconds after exposure to the sun through 

 the glass hemisphere. Whenever the glass hemisphere was removed, 

 whether observing the sky by da\ or by night, or the sun, the deflection 

 increased gradually for about 20 or 30 seconds, as with the heating 

 current. 



A clue was soon found. Generally if the sky is observed by day in 

 Washington, with glass removed, almost no deflection occurs. The 

 gain of heat to the blackened strips from scattered sun rays at Wash- 

 ington is almost equal to the loss of heat by emission of long-wave 

 rays toward the sky. But if the oTiservation is made on Mount 

 Wilson, a large negative deflection occurs. At sunmier temperatures 

 of observation the scattered sun rays at Mount Wilson are by no 

 means equal in energy to the long-wave rays emitted toward the sky. 

 Now it is well known that glass is a nearly perfect absorber of these 

 long-wave rays, and hence is a nearly perfect radiator of them as well. 

 But, on the other hand, the brightly polished metal cover, used as a 

 shutter to the pyranometer, radiates almost nothing, being a nearly 

 perfect reflector for long-wave rays. But the nickel plated cover 

 absorbs about 30 per cent of the shorter- wave solar rays which meet 

 it, and thereby is warmed, and warms the glass close below it by air 

 convection. Now when the cover is removed the glass can cool 

 rapidly by radiation, and as it is almost completely transparent to 

 solar rays, it is hardly warmed at all by them. Hence the glass after 

 exposure grows cooler than before, and as it subtends a full hemi- 

 sphere, it tends strongly to reduce the temperature of the blackened 

 strips below, thus causing the gradual decrease of the galvanometer 

 deflection. 



Having discovered the cause of error, the remedy was seen to lie 

 in shortening the period of exposure so much that there would not 

 be time for the glass to become appreciably cooled on the inside. We 

 therefore began to investigate the behavior of the galvanometer with 

 a view to observing the first swing instead of the permanent deflection. 



As is well known to many readers, the time of swing of a moving 

 coil galvanometer is shortest on open circuit, and increases as the 

 external resistance in closed circuit diminishes, until at length no 



