found impossible to avoid this disturbance altogether; and accordingly the following mode of procedure was adopted :--

Direct magnetic disturbances.-We first carefully investigated whether there was any direct magnetic effect on the galvanometer owing to the currents in the apparatus; this was done by simply short-circuiting the galvanometer. No such effect could be detected. Being assured of this, we always operated as follows:-Threw in the galvanometer by pressing down the button, then allowed the needle to come to rest with the small permanent deflection due to the thermoelectric current. If now, on pressing down the treadle for an instant, there was no motion of the spot, we concluded that there was a balance. It is to be noticed that since we are near balance the battery-circuit is conjugate to the galvanometer-circuit, and that, therefore, making or breaking the battery-circuit does not alter the effective resistance opposed to any electromotive force, thermoelectric or other, in the galvanometer-circuit. (Of this we also assured ourselves by direct experiment.) Another advantage of this method is that it ensures the least possible use of the battery, and thus avoids disturbances from heating. During our final experiments both of us had acquired by considerable practice an acquaintance with the indications of the galvanometer, which enabled us to adjust the balance quickly, and thus secure in greater measure the advantage above mentioned.

Self- and mutual induction.—It is also worth remarking that from the way the B.A. unit coils are wound, and from the general arrangement of the apparatus, neither self- nor mutual induction could have any sensible disturbing effects in our experiments

Method of using bridge for finding coefficients of variation &c.-In finding the variation-coefficients of the coils the bridge arrangement was used in the way described in the Report on Electrical Standards, 1864 (p. 353, &c.); but in finding the difference between the resistances of two coils, the method described by Prof. Foster (in the Journal of the Society of Telegraph Engineers, October 1874) was used. In this method the bridge is first read with the normal coil and the coil to be compared with it in one position, and then the coils are interchanged; the difference of the bridge-readings gives the required difference of resistance in units of the bridge.

Bridge-units. The unit in which we shall state our results further on is the resistance of a tenth of a millimetre of the bridge-wire, which is a metre long and has a resistance of about 075 ohm.

Calibration of bridge and thermometer. The wire was carefully calibrated, but no errors were found large enough to affect our results.

The thermometer used was also compared with a standard thermometer belonging to the laboratory, and the corrected temperatures are in every case given. The degree of accuracy attained in this last comparison was probably about 05 Centigrade.

Description of coils in the case.-In the case containing the coils there are altogether fourteen coils. Five of these are multiples of the unit, viz. 2, 3, 5, 8, and 10, and have brass labels on them; but the inscriptions have never been completed by filling in the last two figures of the temperature at which they are equal to the standard. We have not been able to get any description of these whatever, and have therefore not measured them. Besides these there accompany the box two coils marked A and B, which are not units, and a flat coil described as a normal coil, besides a set of

Another precaution of less importance was to cover the platinum-iridium wire of the bridge with pieces of wood to screen it from dust and radiation from the body of the observer.

tubes for mercury units. The flat coil we used and found very convenient, both from its shape and on account of its small variation-coefficient, which was only 34 per deg. Cent. in the above-mentioned units. The case contains altogether nine unit coils, viz.:

2 Pt Ir
2 Au Ag
2 Pt

3 Pt Ag

--

Of the first six, all except 57, which we have not measured, are mentioned at p. 146 of the Reports, but none of them have proper labels. All, however, were marked in some way or other so as to be identifiable. Of the last three all have labels, which are complete in 6 and 43. Nos. 6 and 29 do not appear in the Reports. The temperature on 43, which does appear, agrees with that given on p. 146. We used 29 as a companion middle coil to 43, because its variation-coefficient was small and nearly equal to that of 43; but otherwise we have not bestowed much care on it.

Coils measured.-The coils which we have measured are, therefore, 2, 3, 58, 35, 36, 29, 43. These we call for convenience A, B, C, D, E, F, G. The normal coil is the flat coil at 10° Centigrade. This temperature is chosen because it was the lower limit of the temperature of the tap-water, which varied on different days from 10° to 12°, though it was very constant during a good part of any one day. As far as our experience went, the use of a stream of tap-water was the best as well as most convenient way of reducing the coils to a known temperature *.

Results of comparison: the first statement. The following Table exhibits our results in the way which lies nearest the method by which they were obtained :

R stands for resistance of flat coil at 10° C.

of the respective coils A, B, &c. at 10° C. variation-coefficients.

Second statement.-The above is the most convenient form of representing our results; but for the sake of comparison we give also the following (Y now stands for the resistance of the coils A, B, C, &c., at the temperatures, or at some one of them, given at p. 483, B.A. Report, 1867 †) :

* One of us, in endeavouring to find the conductivity of paraffin, has since found that the temperature of a wire imbedded in a much greater thickness of paraffin than there is in the B.A. coils, reaches the temperature of the tap-water in considerably less than an hour, the paraffin-jacket having been at a temperature of about 30° throughout to start with. + Reprint, p. 146.

1876.

C

It will thus be seen that B and C are practically equal at the temperatures given, while A does not differ very much from these. D and E are not very different inter se, but differ somewhat from the first three; while G, considering its small coefficient, is considerably out.

Statement of standard temperatures.-If we consider B and C to be right at the temperatures given above and reduce the others so as to be equal to them, we should get the following Table of standard temperatures :

Results of control experiments. In the next place we give the results of our control experiments, in which the several coils were nursed to temperatures very near those given in the Report, and then compared with each other. The small deviations from the temperatures in the Report arise from thermometer corrections. The differences thus found are given side by side with those calculated from the data given above; the differences are given in the next column, and in the last the greatest possible difference, owing to an error of 0°.1 C. in temperature determination.

It appears, therefore, that the differences between the observed and calculated values are always less than what would arise in the most unfavourable case, owing to an error of 0°1 C. in the temperature determinations.

Rough comparison of coefficients with Matthiessen's.-It is perhaps worth while to give the following rough comparison between the results for the variation-coefficients which we have obtained in the neighbourhood of 10° C. with the mean results of Matthiessen.

There is no striking difference except in the case of Pt Ir, where the alloy of which the coil is made must approach much nearer a pure metal than Matthiessen's alloy (33.4 per cent. iridium) did.

Discrepancy in Coil G with former measurements.-The only other point to which we have to call attention is the discrepancy between former and present measurements in the coil G, whose resistance seems to have gone down since it was last tested.

In conclusion we venture to suggest two alterations in the construction of standard coils, which, as far as our experience goes, would be improvements. First, to make them flat instead of cylindrical. This would facilitate stirring when the coils are immersed in any liquid.

Secondly, to insert as near the wire as possible a properly insulated junction of a thermoelectric couple, the other junction of which should be fastened on the outer case of the coil. Several of these fitted to each coil would do away with a great deal of the trouble and uncertainty attending the temperature determinations required in comparing and copying standards.

Third Report of a Committee, consisting of Prof. A. S. HERSchel, B.A., F.R.A.S., and G. A. LEBOUR, F.G.S., on Experiments to determine the Thermal Conductivities of certain Rocks, showing especially the Geological Aspects of the Investigation.

THE object originally proposed by the Committee was to arrange and classify the most commonly occurring rocks experimentally according to their powers of conducting heat; and it has hitherto been so far successfully attained that the thermal conductivities of an extensive series of ordinarily occurring rocks have been shown to differ from each other on a very strongly marked scale of gradation, which it was endeavoured to represent graphically in the Committee's last Report by a series of ascending steps of absolute thermal resistance, or resistance to the passage of heat offered by the different rocks. To every 200 units of this ascending scale a new letter of the alphabet, starting with A for the interval 0-200 of absolute resistance, was assigned; the values of the resistances were shown graphically, and the various rocks that arrange themselves under the several classes so formed could be readily discerned. By adopting this graphical mode of representation the values of certain

thermal resistances observed during the past year and communicated in this Report may be exhibited with equal clearness, and an easy comparison may by this means be made of the values found in this and last year's series of experiments where the same rock-specimens, or specimens of very closely allied kinds of rock, were submitted in the former and in this year's series to examination. A slight change, however, is here introduced in briefly describing the results obtained numerically, by employing, instead of the significant figures of those results (as was done in the last Report), the tenth part of them as a brief expression for the absolute thermal conductivity. Thus the absolute thermal conductivity of galena in the present list being 0.00705 in centimetre-gramme-second units, hitherto described for brevity by its significant figures 705, will be spoken of in this Report as 70-5, to which the meaning may conveniently be attached that 70·5 gramme-degree units of heat per second pass through a plate of galena one centimetre thick, having an area of one square metre, for a temperature-difference of one degree between its faces.

The method of investigation without the use of a thermopile has hitherto proved unsuccessful, no soft material capable of effecting a close junction with the rocks having yet been found of sufficiently constant resistance to afford a useful standard of comparison with them when the rocks are introduced between its layers; but the progress of the investigation has shown that a simple water-film (if it could be preserved from drying off with porous rocks) effects a complete junction between them and any impervious surface, as that of caoutchouc, against which they are pressed. A similar film of oil, it appears from some experiments recorded in the present list, is less effective for the purpose; and to ensure a constant water-film in which the thin wires of the thermopile could be placed, pieces of well-soaked bladder kept soft in water rendered antiseptic with carbolic acid were laid on the india-rubber faces of the boiler and cooler, so as to press the thermopile-wires against the rock with a constantly moist and uniformly wet surface. The duration of an experiment and the temperature to which they were exposed (usually between 100° and 120° F.) were never so great as to cause the bladders to approach dryness before the termination of the experiment. The proportion of moisture absorbed by the rocks (when sensibly porous) was ascertained, and it was always such a small fraction of that imbibed by the same rocks thoroughly soaked in vacuo that it probably exercised a scarcely sensible influence on the results. Its amount, and that of the full quantity of water absorbable by the porous rocks tested, is stated in the list; and from the corresponding alteration of the observed conductivity some idea of the probable correction necessary to be applied for the presence of moisture in some of the porous rocks during the process of the experiment may be obtained.

The two chief defects of the thermopiles used hitherto had been their thickness (making them intrude too far from the rock-surface into the badly conducting strata with which it is in contact), and the false thermoelectric currents proceeding from irregularities of material and internal condition of the wires subjected to great varieties of temperature along their length. To diminish the former source of error, wires less than half a millimetre (0·40 millim., or inch) in diameter were used and neatly soldered at the junctions; and to counteract as far as possible the remaining evil, they were chosen of the most dissimilar metals (iron and German silver), and twelve junctions above and twelve below the rock-plate formed a continuous circuit giving a very strong thermoelectric current. The whole resistance of the circuit (including the 20 ohms usually added to bring its indications conveniently within the scale of a Thomson's reflecting galvanometer) was 40