Mons. C. Despretz on the Maximum Demity of Liquids. 5 



with water at a lower or higher temperature than the sur- 

 rounding air, according to the experiment, it was closed with 

 a porcelain cover; after waiting a few moments the tempera- 

 ture of the thermometer was noted down from minute to 

 minute. A curve of the temperatures was then drawn, the 

 time being taken as abscissae, the temperature as ordinates. 



It is known that, under its maximum, the water below is 

 warmer than that above, and vice versa when above the maxi- 

 mum. We might thence have concluded that curves of the 

 temperatures would bisect one another at one point, which 

 would be the temperature of the maximum ; this however was 

 not the case. Near the 4th degree the curves cut one an- 

 other at many points. I obtained the maximum in the fol- 

 lowing manner: I took, 



1 . The mean of all the temperatures where the curves 

 suddenly change their direction. 



2. The mean of the temperatures corresponding to the 

 points of intersection. 



S. The mean of the points where the curve drawn with the 

 mean temperatures intersected the other four points. 



4. The mean of these three results. 



We see that the method of Tralles, thus modified, ought 

 to conduct to a result more certain than those hitherto ob- 

 tained. 



The mean of the two heating experiments is 4°'058. But 

 the thermometers being graduated in a vertical position, and 

 observed here in a horizontal one, a small correction is ne- 

 cessary on account of the pressure of the mercury ; as was also 

 a second, proceeding from the action of the air on the stem 

 of the thermometers : these two corrections, the influence of 

 which had been estimated by previous experiments, reduce 

 this mean to 3°-969. 



The two cooling experiments gave 3°'995 for the corrected 

 mean. The common mean is 3°*982. The difference, 0°'026, 

 is in the direction in which it ought to be; for in a state of 

 motion, i. e. whilst being warmed or cooled, the temperature 

 of a liquid is not indicated exactly by a thermometer. If the 

 fluid is cooling the thermometer indicates too much, if be- 

 coming warm too little. The retardation in question will be 

 greater in proportion to the rapidity of motion of the heat. 

 Moreover, the less the partial results from which the true re- 

 sult is drawn differ among each other, the greater will be the 

 probability of the exactitude of a series of experiments. This 

 condition then seems to be fulfilled by our experiments, since 

 the difference between the greatest and smallest result amounts 

 only to 0''-026. 



