Since the calorific power of a coal is determined by the number of degrees Fahrenheit which a given quantity of water is raised in temperature by a known weight of fuel, it follows that every care should be taken that the experiment be performed under similar atmospheric conditions. The oscillation of barometric pressure does not appear to affect the working, but the temperature of the room in which the work was done, and especially that of the water, are most important considerations. It has been observed by some who have used this apparatus--and I have frequently noticed it myself--that the lower the temperature of the water is under which the fuel is burnt the higher is the result found. This has been explained on the assumption that the colder the water used, the greater is the difference between the temperature of the room and that of the water; hence it would be expedient that in all cases when such experiments are made the same difference of temperature between the air in the room and the water employed should always exist. For example, if the temperature of the room were 70°, and the water at 60°, then the same coal would give a like result with the water at 40° and the room at 50°. This has been regarded as the more evident, because the gases passing through the water escape under favorable conditions of working at the same temperature as the water, and are perfectly deprived of any heat in excess of that possessed by the water. Under these circumstances it would seem only reasonable that this assumption should be correct. It was, however, found after a large number of experiments upon the same sample of coal that this was not the case. 30 grammes of coal which raises the temperature of the water 13.4°, when the water at starting was 60° and the room at 70°, gives 13.7° rise of temperature with the water at 40° and the room at 50°. Conversely, when the water is at 70° and the room at 80°, a lower result is obtained. The explanation appears to be this: The gas which escapes from the water was not in existence in the gaseous form previous to the experiment, and the heat communicated to the gas being a definite quantity it follows that the more the gas is cooled the greater the proportion of chemical energy in the shape of heat will be utilized and recorded as calorific power.

In order, therefore, to make the experiment more simple and workable at all temperatures, a sample of coal was selected, which should be perfectly manageable and readily consumed. Appended is an analysis of the coal employed (from Ebbw Vale, Monmouthshire):

 Composition per cent. 
Carbon...............................88.33 Hydrogen............................. 5.08 Oxygen............................... 3.28 Nitrogen............................. 0.55 Sulphur.............................. 0.70 Ash.................................. 1.26 Water (moisture)..................... 0.80 ----- 100.00

In the following experiments the standard temperature of the water was taken as 60° F., and as the coal gave 13.4° of rise of temperature, 67° F. was selected as the standard room temperature. The reason for this room temperature is obvious, for, whatever heating effect the higher temperature of the room may have upon the water in the cylinder during the time occupied by the first half of the experiment, would be compensated for by the loss sustained during the second half of the experiment, when the temperature of the water exceeded that of the room. The mean of numerous trials gave 13.4° F. rise of temperature, equal to 14.74 lb. of water per lb. of coal. When the water was at 50° and the room at 57°, the mean of several experiments gave 13.5° rise of temperature. When the water was 40° at starting and the room at 47°, 13.65° was the average rise of temperature. Trials were made at intermediate temperatures, and the results always showed that higher figures were recorded when the water was coldest. With a view of getting uniformity in the results it was thought well to make experiments, in order to find out what temperature the room should be at, so that this coal might give the same result with the water at 50°, 40°, or at intermediate temperatures. Without going much into detail, it was found that when the temperature of the room was at 40° and that of the water 40°, and the experiment was rapidly and carefully performed, 13.4° rise of temperature was given; but this result could be obtained without special effort when the room was 42° and the water 40° at starting. It is evident that the cooling effect of the air in the room upon the water cylinder is very appreciable when the water has reached 13° above that of the room. When the water was at 50° and the room at 55°, the coal gave 13.4° rise with ease and certainty, and it would not be out of place to remark here that with those coals which burn well in Thompson's calorimeter, the results of several trials are remarkably uniform when properly performed. With the water at 70° and the room at 80°, a like result was worked out. Experiments at intermediate temperatures were also carried out (see table in sequel). It is true that the whole difference of temperature we are dealing with in making these corrections is only 0.25, but 0.2 in the result, when multiplied by 537 to bring it into calories, as is done by the authorities in Italy, makes more than 100 heat units--a serious difference when 5d. per ton fine is attached to every 100 calories lower than the number guaranteed.

Taking the latent heat of steam as 537° C., and multiplying this number by 14.74, the evaporative power of the coal used in these experiments, its equivalent in calories is 7,915. From the analysis of this coal, disregarding the nitrogen and deducting an equivalent of hydrogen for the oxygen present, the total heat units given by Favre and Silbermann's figures for carbon (8,080) and hydrogen (34,462) will be 8,746. It will be seen, therefore, that the calorific power, as determined by Thompson's apparatus, gives a much lower result when multiplied by 537 than the heat units calculated from the chemical composition of the coal. When I used Thompson's apparatus in the chemical laboratory at Turin to determine the evaporative power of various cargoes of South Wales coal, it was agreed by mutual consent that the temperature of the water at starting should be 39° F. (the temperature at which the heat unit was determined). The temperature of the room was about 60°, but this varied, as the weather was somewhat severe and changeable. Under these conditions, with the water at 39° and room 60°, the coal which gives 14.74 lb. of water per lb. of coal, will give as high as 15.88 lb. of water per lb. of coal. This result multiplied by 537=8,496 calories, approaching much more nearly to the theoretic value. This method of working is still practiced abroad, but experience has shown that very widely differing results follow when working in this manner, especially if the temperature of the room is changeable, as it naturally is where ash determinations and other chemical work is proceeding simultaneously. The time the experiment lasts, taking the reading on a quickly rising thermometer and other considerations, render the experiments anything but trustworthy when 0.2 of a degree makes a difference of more than 100 calories. In the instructions supplied with Thompson's calorimeter nothing is said as to the temperature of the room in which the experiment is performed, but simply that the water shall be at 60° F. If, with the water at 60°, a room were at 50°, as it often is in winter, a good coal would give 14 lb. of water per lb. of coal as the evaporative power; but if in summer, the room were at 75° and the water at 60°, the same coal would give 15 lb. of water per lb. of coal. If further evidence were needed of the effect of temperature consideration of the experiments already referred to will show how necessary it is that some general rule shall be adopted. Considerable stress is laid (in the instructions) upon the quantity of oxygen mixture used being determined by rough experiments. This I have found leads to erroneous conclusions unless a number of experiments are tried in the calorimeter, as it often happens that the quantity which appears to be best adapted is not that which yields a trustworthy result. There are many samples of South Wales coal, 30 grains of which will require 10 parts of oxygen mixture in order to burn completely, but since a little oxygen is lost in drying and grinding, and few samples of chlorate are free from chloride, it is not safe to use less than 11 parts of oxygen mixture, but this amount is sufficient in all cases, and never need be exceeded. I have made numerous experiments with various coals (anthracite, steam, semi-bituminous, and bituminous, including a specimen of the ten yard coal of Derbyshire), and find that with 11 parts of chlorate and nitrate of potash, they are all perfectly manageable and yield the best results. It is quite clear that the excess of chlorate is decomposed in all instances, and the latent heat of the oxygen evolved, but those coals which are best to experiment with did not yield results that differed when the quantity of oxygen mixture was reduced to nearly the limit required for combustion of the coal. Under these circumstances, therefore, the constant use of 11 parts of oxygen mixture--a suitable quantity for all coals exported--would enable operators to obtain similar figures, and make the test uniform in different hands.