CONTINUOUS ELECTRIC CALORIMETRY. 
77 
In order to throw further light on the question, Mr. Glazebrook undertook a 
determination of the temperature-coefficient of No. 4086 by heating it up to 20° C. 
and allowing it to cool slowly, the case being full of oil. Taking only the morning 
observations, which Mr. Glazebrook considers the most reliable, we obtain the 
following mean results on the two days :— 
Temperature-coefficient of No. 4086 at the National Physical Laboratory. 
Date. 
Mean temperature. 
Mean value from 3715. 
January 4, 1901 . . 
o 
17-2 
1-00024 
)> 5 ,, 
9-4 
1-00005 
These observations give a mean value of 1 '00015 at 13°‘3 C., which agrees very 
closely with the December comparisons, but is slightly lower, as though the diminu¬ 
tion of resistance were still in progress. The value of the temperature-coefficient 
deduced is + '000024, which is a very probable value for this kind of manganin for 
temperature changes of 10° in one day. 
(9.) Hysteresis in Manganin Coils. 
From my own experience of the behaviour of manganin, I am inclined to explain 
these discrepancies as follows:—The temperature-coefficient of a manganin coil 
frequently depends to some extent on the past history and on the rate of heating or 
cooling. It may exhibit a kind of lag or hysteresis, the resistance continuing to 
increase gradually for some time after a rise of temperature, so that the value of the 
coefficient found by short-period observations is often smaller than the value which 
applies to changes of long period. Taking the July and December observations at 
the National Physical Laboratory as being correct, we find a long-period temperature- 
coefficient for No. 4086 of +'000047, which is about twice as large as the short- 
period coefficients '000027 and '000024, deduced from the observations from day to 
day. It was for this reason that I discontinued the use of manganin for accurate 
work in platinum thermometry some years ago, preferring to use platinum-silver coils 
annealed at a red heat and compensated for temperature by means of platinum 
similarly annealed. The defect is of a kind that would escape notice except in very 
accurate work over long periods with coils exposed to considerable changes of 
temperature. It also appears to depend on the manner in which the wire has been 
annealed. 
It is worthy of notice, as a confirmation of this view, that a small effect of this 
nature was observed by Marker and Chappuis in their platinum thermometry 
