Copper and iron are still used to a limited extent in these pyrometers, but lose continuously in weight by oxidation, the scales of oxide falling off when quenched, necessitating weighing before each test to ensure accuracy. Nickel oxidises very little below 1000 C., and as the thin film of oxide which forms does not readily peel off, the weight may increase slightly. Quartz would probably be more suitable than metals, not being altered by heating and quenching, but does not appear to have been tried for this purpose. Another possible material is nichrom, which resists oxidation below 1000° C. The weight of the solid should be at least 1/20 of that of the water, in order to ensure a tangible rise in temperature, and the thermometer should be capable of detecting 1/20 of a degree C. The rise in temperature should not be so great as to cause the water to exceed the atmosphere in temperature by more than 4° or 5° C., as otherwise radiation losses would have a marked effect. The limits of accuracy of the method will be shown by reference to examples. _Example I._—A piece of nickel, weighing 100 grams, is placed in a furnace, and after heating dropped into 2000 grams of water at 10° C., contained in a vessel of water equivalent 50 grams. The temperature rises to 16·25° C. The specific heat of nickel for the range is 0·137. To find the temperature of the furnace and the limits of accuracy, the thermometer being readable to 1/20° C. Equating heat lost by the nickel to that gained by the water and vessel:— 100 × 0·137 × (_x_ - 16·25) = 2050 × (16·25 - 10·0) from which _x_ = 952° C. If the error in each thermometer reading amounted to 1/40° the maximum difference in the above calculation is obtained by introducing the altered values as under:— 100 × 0·137 × (_x_ - 16·225) = 2050 × (16·225 - 10·025) when _x_ = 944° C. The maximum error due to a possible incorrect reading of 1/40° is therefore less than 1 per cent. _Example II._—The loss of heat by radiation in transferring 100 grams of nickel at 927° C., possessing a surface of 30 square centimetres, and with radiating power 0·7 of a black body, may be shown by the fourth-power law to be 50 calories per second (see page 139). If two seconds were occupied in the transfer, the error from this cause would be 1 in 130; and adding this to the thermometric error, the total is less than 2 per cent.