74 AKSEL S. STEEN. TERRESTRIAL MAGNETISM. [NORW. POL. EXP. 



show that a cannot be constant, but, as we have already supposed, must 

 depend upon the temperature. If we call this t, we may put 



and equation (6) becomes 



2P" i "i a l " 



if t and t' have anything like the same value. This equation affords an 

 opportunity for the determination of t and 2 by the method of least 

 squares, as we obtain a series of equations in the form 



B 



a, + Acts = =-, 



where 



After a careful study of the forthcoming observation-data, which deter- 

 mined me in excluding two of the above-mentioned normal days, namely 

 June 18th, 1896, and March 2nd, 1897 (Hamburg) --it being likely that the 

 observations on these days were more affected by disturbing forces than on 

 the other normal days - - I finally retained 12 corresponding values of A and 

 B, employing the mean value of the constant fi used in the provisional 

 calculation of a. By the aid of these 12 values of A and B, I have been 

 able to draw up the following normal equations: 



12 , 99-9 2 = 0-0043426 



- 99-9 o, 4- 3462-95 a 2 = 0-045605, 

 which give 



Ol = 0-00033198 

 2 = 0-00000359, 

 and thus a = 0'00033198 0-00000359 r. 



By this formula, which I consider to be the final one, I have calculated 

 a table for a, for magnet V, with the temperature as argument, for every 

 degree from 40 to -j- 20 C. 



I have been compelled to abandon a similar examination of the tempe- 

 rature-coefficient for magnet VI, that magnet having, as previously stated, 

 been comparatively so seldom used, that the observations that we have do 



