must now be passed into the water, and as soon as the temperature of the water and upper plate begins to rise, and the needles of the galvanometer to be deflected, the direction of the deflections must be noted every few minutes until the temperature of the water reaches 205° or 210° F. If a verification of the result is required, the water may be allowed to cool gradually, and the deflections noted as it cools, or what is much better, the experiment may be wholly repeated. The current of steam should be of such an amount as to raise the temperature of the water from 60° to 210° F. in two and a half or three minutes.

15. The time required to make a single examination of a liquid varies from ten minutes to half an hour, according to the facility of manipulation; liquids which act strongly upon the washers or plates, such as strong acids or alkalies, or those which exhibit irregular phænomena, require much longer time, and need several examinations to verify the result.

16. Many precautions are necessary to be observed in the construction and use of the apparatus in order to exclude every interfering circumstance; the washers should be of such a nature as not to act chemically upon the metal plates; they should also not give way and be forced out by the combined influence of the heat and pressure; vulcanized india-rubber washers, covered with a thin sheet of ordinary india-rubber on the side next the plates by means of a solution of india-rubber in bisulphide of carbon, answer admirably after having been used a few times; or vulcanized india-rubber ones alone succeed well with some metals after much use, but not in examining strong nitric acid, because of its action upon them. The only washers I have found to withstand this liquid with the heat and pressure, were formed of a layer of asbestus-fibre stuck upon the ends of the cylinder by means of a little shell-lac varnish. Washers formed of ordinary india-rubber or of gutta-percha only last during a single experiment, the heat and pressure combined spoil them. Leather washers soaked in shell-lac varnish and dried, succeed for a few experiments, but soon become hard and brittle; and they are unfit for examining strong mineral acids or alkalies, being acted upon by them. Washers of calico, linen, or flannel, soaked in varnish and dried, quickly become hard and break. Of whatever material the washers may be composed, they should be perfectly alike, and be well cleaned from any previous liquid before each experiment.

17. The metal plates should be as homogeneous and alike as possible, and if made of cast metal, each must be cast at the same pouring; each should be smooth and clean, polished with the same materials and in the same manner; the most scrupulous care must be taken to have them equally smooth, and the

most perfect cleanliness is necessary in their preparation. The more oxidable metals require to be cleaned at each experiment; and they should be used within a few minutes after being cleaned, otherwise the deflections of the needles in the ensuing experiment will be considerably decreased. Platinum plates only require to be occasionally cleaned in ordinary experiments; but in all the comparative or more delicate experiments they should be heated red-hot all over before each time of using. In comparative experiments it is of the highest importance that the quantity of water in the cistern be equal in each case, also that the current of steam be of uniform strength, and that the periodof time occupied in raising the temperature of the water from the lowest to the highest point be accurately noted.

18. The inside of the vessel should be occasionally examined during experiments, and invariably when the water has risen in temperature to nearly 200° F., to see if any air-bubbles remain in the liquid, or are set free by the action of heat, or any gas evolved by the action of the liquid upon the plates, or if any action of the liquid has taken place upon the washers. With all liquids, more or less of the atmospheric air dissolved in them is set free against the upper plate at the higher temperatures; but this, with the metals I have used (aluminium, antimony, bismuth, zinc, tin, iron, copper, silver, and platinum), only slightly affects the amount of deflection at those temperatures, and disappears after a few examinations of the liquid, the air being expelled. I have repeatedly tried the effect of previously well boiling the liquid, and cooling it in a closed glass flask quite filled with it; but the deflections produced were but slightly different to those when the liquid was not so prepared; for delicate or comparative experiments the liquid should always be pre-boiled. I have been unable to examine those metals and liquids which evolve gas by their mutual contact, although I have tried various contrivances for the immediate removal of the gas; or those in which the metal in solution (i. e. with reducible metallic solutions) is reduced upon the plates by ordinary chemical action.

19. Care must be taken to keep the outside of the apparatus quite dry when in use, that the insulation of the plates from each other may be as perfect as possible.

INFLUENCE OF VARIOUS CIRCUMSTANCES UPON THE CURRENTS. 20. I have proved that the ordinary thermo-electric property of the metal has little or no influence upon the currents under consideration, by interposing in the circuit with the galvanometer the foregoing apparatus fitted with silver plates, and charged in one instance with a pre-boiled solution of carbonate of potash, and in another instance with an acidulated solution of sulphate

of copper; also interposing between the two instruments two platinum wires 20 inches long each, terminated by two discs, 2 inches wide, of sheet platinum; heating one of these circles redhot, and suddenly bringing it in contact over its whole surface with the other and cold one, not the slightest electric current arising from this action passed through the circuit, the liquid completely arresting it.

21. Contracting the column of liquid to an hour-glass form, either by means of a perforated diaphragm of gutta-percha, or by a glass vessel of that shape (7,note) instead of the cylinder, does not at all diminish the quantity of the current, thereby proving that the currents under consideration do not arise from any action of heat or other force in the body of the liquid.

22. It makes no difference in the direction of the currents whether the heat be applied to the liquid near the metal or to the metal itself, i. e. whether the heat passes from the liquid to the metal, or vice versa; a number of careful experiments clearly established this point.

23. The effect of constant heat was to produce a constant or nearly constant current, even when the two plates were maintained at uniform but different temperatures for several hours.

24. Increasing the depth of the liquid column to about 8 inches had no appreciable effect upon the quantity of the current, there being abundance of conductivity for so small a force; and increasing the thickness of the plates had no other effect than that of causing the currents to be more slow in manifesting themselves, in consequence of the heat being longer in reaching the liquid surface.

25. To ascertain the influence of size and number of the plates, I constructed four apparatuses similar to the one already described (7), but with square copper plates 7x7 inches, and charged them with an acidulated solution of sulphate of copper. I connected them in one instance intensity fashion, i. e. as four successive pairs, and in the other instance quantity fashion, i. e. all the hot plates as one plate, and all the cold ones as another; and tried various comparative experiments with the galvanometer and variable resistances, and found that the intensity of the current increased with the number of pairs, and the quantity with the amount of surface in each pair. No visible electrolytic effects were produced by several hours' action of this larger apparatus upon good conducting solutions, although the current, even of one pair, possessed abundant intensity to traverse long columns of those liquids.

26. Diminishing the area of liquid contact of the lower plate had the same or nearly the same effect in decreasing the quantity of the current as diminishing that of the upper one, thereby

proving the currents to be a result of a differential action between the plates, and not of a specific thermic action at the upper plate.

Influence oF CHEMICAL ACTION.

27. To ascertain whether the currents under consideration were solely due to chemical action, I examined several liquids in the foregoing apparatus (7), using platinum plates to exclude, as far as possible, chemical influences, and observed if any currents were produced: the following are the particulars of the experi

ments:

28. Took 4 ounces of distilled water, added to it a quarter of an ounce of pure sulphuric acid, boiled the mixture in a glass flask to expel dissolved air, and cooled it in the closed flask; prepared the apparatus with extreme care, and when the liquid was of the temperature of the atmosphere, introduced it into the cylinder; then tested the junction of the upper platinum plate with its platinum connecting wire by heat, to ascertain if any electric current could arise from the heat extending to the soldered junction; but no effect was produced, neither was there any permanent deflection of the galvanometer needles produced by any difference of cleanliness, temperature, smoothness or otherwise, of the plates or washers. Steam being now passed steadily into the cistern of water, the needles began to be deflected immediately the thermometer indicated a rise in temperature, the cold plate being positive; and the deflection gradually increased with the rise in temperature, being 0 at 65° F. (the temperature of the atmosphere and lower plate); 1.5 at 80°; 2 at 90°; 2.5 at 100°; 4 at 110°; 8·0 at 120°; 11.5 at 130°; 14-5 at 140°; 17.5 at 150°; 19.5 at 160°; 21 at 170°; 23 at 180°; 25 at 190°; 26 at 200°; and 26.75 at 208° F. I then took a mixture of 4 ounces of distilled water and a quarter of an ounce of pure nitric acid, and performed a similar experiment; the cold plate was again strongly positive to the hot one, and increasingly with the increased difference of temperature.

29. These currents did not arise from difference of chemical action at the metallic surfaces, that being excluded by the careful preparation of the plates, and proved by the previous quiescence of the galvanometer; nor from chemical action developed by heat,-1st, because there was not the slightest sign of such action upon either of the plates after either of the experiments; and 2ndly, because if such action had occurred, the current produced would have been in an opposite direction, as will hereafter be proved in the present paper.

30. If a minute degree of chemical action is the cause of the currents obtained with platinum plates, then I can at present imagine with any reasonable degree of probability only the fol

lowing cases:-1st, that under the influence of heat, acid liquids act upon metals differently to alkaline ones; either the hot plate dissolving most in the case of acids, and the cold one in the case of alkalies, or vice versa, the plate most freely acted upon being positive in all cases; or 2nd, that the hot plate dissolves most in both cases, but with acids the dissolving metal acts as a base and is positive, whilst with alkalies it acts as an acid and is negative. 31. If either hypothesis be true, experiment ought to accord. To test them, I mixed together a quarter of an ounce of sulphuric acid and 8 ounces of a solution composed of 220 grains of sulphate of copper dissolved in 10 ounces of water, and then accurately balanced two equal quantities (about 3 ounces each) of the mixture. I put the two portions in separate large glass flasks with two pieces (one in each) of copper foil, each 2 x 2 inches, of the same thickness, perfectly clean, and cut from contiguous parts of the same piece of metal; kept each exactly ten hours in its liquid, one being at about 65° F., and the other maintained at a temperature of about 200° F. the whole of the time, the small amount of water which evaporated from the hot one being occasionally replaced. The hot liquid in a few hours became manifestly deeper in colour, and the piece of metal in it gradually blackened, whilst the other solution and its metal did. not visibly alter. The pieces of metal were taken out at the end of the ten hours, rinsed in water, dried and weighed; the one from the cold liquid, which previously weighed 23.7 grs., had lost only 0.1 gr., whilst the other, which weighed 24.0 grs., had lost 64 grs., and was quite dark in colour, the cold one being but moderately tarnished. When the hot portion of the liquid had cooled, I mixed it with the other portion, filtered, and introduced it into the apparatus (7) with copper plates, and tried it in the usual manner. The cold plate was positive, the deflection of the galvanometer needles being 54 with the top plate at 208° F., and the lower one at 65° F. I repeated this experiment with the same liquid and very thin copper plates; the cold plate was again positive, the deflection being 51 with the upper plate at 206° F. In both of these experiments the upper plate alone was blackened on taking the apparatus to pieces.

32. In these cases there was doubtless more chemical action at the upper plate than at the lower one; yet notwithstanding this the lower plate was positive, which I attribute to the action of some other cause than chemical action. These instances are exceedingly clear ones, and the results are invariably the same with all solutions of sulphate of copper, acidulated or not, or of various degrees of dilution, when examined either with copper, silver, or platinum plates.

33. The following experiments also show that the currents