The metal used in the experiment should not oxidise readily, and should possess a high melting point. Platinum is most suitable, but the cost of a piece sufficiently large would considerably exceed that of a thermo-electric or other outfit. Nickel is next best in these respects, and is now generally used for the calorimetric method, up to 1000° C. The specific heat varies to some extent in different specimens, but can be determined for the ranges involved in practical use. This may be done by heating a given weight to known temperatures and plunging into water, the result being obtained as in the foregoing example, _t_{0}_ in this case being known and the specific heat calculated. From a series of such determinations, a curve may be plotted connecting specific heat and temperature range, from which intermediate values may be read off. [Illustration: FIG. 65.—SPECIFIC HEAT OF NICKEL OVER RANGES FROM 0° C.] Regnault, who first suggested the calorimetric method for high temperature measurement, attempted to measure the specific heat of iron over different ranges, with a view to using this metal in the process. Owing to the absence of reliable means of determining the experimental temperatures, however, Regnault’s values were considerably in error. For the range 0 to 1000° C. he gave the average specific heat of iron as 0·126, a figure much below the truth. Thus, if a piece of iron be heated to 970° C., as measured by the thermo-electric method, and dropped into water, the temperature calculated from an assumed specific heat of 0·126 will be found to be 1210°, or 240° too high. The values now employed are obtained by experiments with a thermo-electric pyrometer, so that temperatures deduced by the calorimetric method agree, within the limits of manipulative error, with those of the standard scale. The accompanying curve, fig. 65, shows the average specific heat of nickel over all ranges between 0° and 1000° C., and from this curve the correct figure to use in the calculation for any range may be determined. Thus for a furnace between 800° and 900° C. the specific heat would be taken as 0·136; and although the choice of the value to be taken involves a knowledge of the temperature within 100°, no difficulty arises in practice, as it is easy to judge this limit by experience at temperatures below 1000° C In the most approved forms of calorimetric pyrometers for industrial purposes the temperature of the hot metal may be read directly from a scale, prepared in accordance with the value applying to the specific heat at various ranges.