;

in the magnetizing force experienced by the iron under test the latter being contained in a zinc box C, which was filled and surrounded with cotton-wool and placed between the poles DD.

The test-piece consisted of a bundle of fine iron wires lying parallel to the lines of force, into the centre of which was introduced one end of a "line" thermopile. The other end of the pile was surrounded by a coil of silk-covered copper wire all in one length, which was similar in size and shape to the iron test-piece.

The actual size and shape of the mould in which the separate plates of the alloys forming the pile were cast is shown in fig. 2. The plates were about millim. thick, and were composed alternately of the alloys 32 parts bismuth + 1 part antimony, and 12 parts bismuth +1 part tin. They were placed side by side with mica to separate them, and were soldered at their pointed extremities, which were then covered with goldbeater's-skin. Fourteen pairs were used to form the pile.

By this means the junctions formed a blunt knife-edge and were brought well into the centre of the iron wire; the mass of the pile being at the same time kept small at that point, while it was large enough outside to prevent undue resistance to the thermal current.

The galvanometer, connected through a mercury commutator with the pile, was at a distance of nearly 20 yards from the electromagnet, and its circuit was provided with a small variable E.M.F. to counterbalance accidental thermal currents. The scale was read by a telescope, and a motion of a tenth of a division could be easily detected.

Results of the Experiments.

The constant saturating field between DD (fig. 1), produced by exciting the main coils A, was first measured by the earth-inductor method and found to be about 2800 C.G.S.; and a separate current was then sent through the coils B of such a strength that on reversing it the main field was altered by 166 C.G.S. units (H in equation iii.). The currents required to produce these fields were also measured, so that they might be reproduced for the experiment.

After passing current through the exciting coils for a couple of hours to obtain a steady temperature, the method of experiment was as follows:

Galvanometer-readings were taken every 20 seconds, the current in B being at the same time reversed every 5 minutes,

*Boys, 'Cantor Lectures,' 1889, p. 18.

so as to alternately heat and cool the iron if the effect looked for existed. In this way fifteen readings were obtained between every pair of reversals. It was found, however, that a small inductive effect on the galvanometer circuit occurred at each reversal, and could not be quite got rid of. The needle always kicked from 0.5 to 1 and oscillated. On the other hand, it so happened that the period of oscillation of the needle was almost exactly 60 seconds. We therefore divided the fifteen readings into five groups of three; and taking the mean of each three we assumed it to represent the average position of the needle during one minute. Five reliable

readings were thus obtained at intervals of 0.5, 1.5, 2.5, 3.5, 4.5 minutes after each reversal of the B current. Nine such reversals formed 66 set." At the end of a set the connexions of the galvanometer were reversed, to eliminate a possible direct magnetic effect on the needle of the galvanometer, and the set was repeated. Two such sets thus constituted a complete experiment.

In the following table the mean values of the differences between the readings in the strong and in the weak field, taken at 0.5, 1.5, 2.5, &c. minutes after alteration of the field, are given for the four sets of two complete experiments. The final mean values of these differences are thus each dependent on 108 separate readings of the galvanometer.

The effect of commutating the galvanometer upon the sign of the differences has been allowed for in the table by reversing the signs of Set 2 in each experimert. The sign to every difference except one means that, except in that case, the cross-wires stood more to the left of zero when the pile was in the strong field than when it was in the weak one.

To determine the thermal meaning of this fact, we next sent a small momentary current through the copper-wire coil mentioned above as being upon the other end of the pile. The galvanometer was connected as for Set 1. A motion of the

cross-wires to the right resulted. Hence, if the differences. observed were due to the iron, an increase of the magnetic field must have warmed it—just the reverse of what theory required.

IH

The volume and resistance of the copper in the coil upon the pile were known. By sending the proper current through the coil for one second, it was possible to produce in the copper the number (0.006) of calories per cub. centim. which ought to have been produced in the iron by weakening the field, viz. The result was a deflexion which reached a maximum value of 3.8 at the end of about 75 seconds after the copper had been warmed, and died away in less than 6 minutes. As the two ends of the pole were as nearly alike as possible, this value gave a rough idea of the deflexion to be expected. The observed effect of 0.44 was thus about nine times too small and in the wrong direction.

These results were obtained at the end of the Christmas vacation, and at this point we were obliged to remove our apparatus for the term's work.

Two possible causes for the effect we had observed occurred to us. One, that the alteration in field-strength had altered the resistance of the pile, and therefore of the current passing through it. The other, that the iron had not been quite saturated throughout its length, and that hysteresis effects had been superposed upon the one we were looking for. We therefore undertook a fresh series of measurements this summer with better appliances. The galvanometer was rendered absolutely dead-beat with a large mica vane, so that each reading was complete in itself. The distance between the pole-pieces was decreased from 3.0 to 2.65 centim., by which means the strength of the main field was increased to 3200 C.G.S., and the length of the iron test-piece was increased until there was only 0.5 millim. clearance between its ends and the pole-faces, instead of 7 millim. as before.

was

Direct measurement now showed that the iron thoroughly saturated, and that the value of I for it was 1640. A further improvement consisted in the fact that the iron wire was itself silk-covered and all in one length, so that the artificial heating for calibration could be performed upon the specimen itself, and a far more accurate indication obtained of what to expect from the variations of the magnetic field.

Slight movements of the test-piece on altering the field were now found to give rise to small thermal effects, due doubtless to an alteration in the flow of heat from the exciting coils through the pole-pieces into the pile brought about

thereby. The pile and test-piece were therefore together embedded in a slab of paraffin wax, 2:55 centim. thick, through which passed two brass pins 2.65 centim. long and 0.3 centim. in diameter, which were melted into the wax on either side of the test-piece, and served to prevent any relative motion between it and the pole-pieces.

*

On trying this soon after the paraffin had set, the following readings were obtained in the right direction; that is to say, they pointed to a cooling of the iron when the field was strengthened.

00 +031 +0.83 +1.29 +1.26 +1.26 +1:35 +1:37

Here t stands for the time in seconds after altering the field between the values 3310 and 3090 either way (H=220).

The readings (which correspond to the final mean differences of Table I.) are plotted in curve (a). Curve (h) in

*The

sign is used in what follows to indicate that the readings are in accordance with theory, and vice versa.

calories

the same diagram represents the effect of heating the iron wire by means of a current flowing through it for one second, 1640 × 220 and of the proper strength to produce J per cub. centim. of the iron. The points which determine this curve are the mean values of several consistent observations.

The following day the readings were in the wrong direction (curves b and c), nothing having been altered in the mean time. Thinking that the alterations in the field-strength might still be capable of producing some relative movement of the pile and the strands of the test-piece, the effect of which had been altered by the hardening of the paraffin during the night, we caused the paraffin to penetrate the interstices of the iron wire by means of the air-pump, and the apparatus was then placed in position and subjected to a field of about 3000 units for 5 hours while it cooled. Readings taken next day were still negative, and to about the same extent as before (curves d and e). The substitution of a solid cylinder of iron for the iron wire did not affect matters. Relative motion of test-piece and pile seemed therefore to be excluded.

The clearance between pole-faces and test-piece was next increased from 0.5 to 3.5 millim. without altering the result; from which we inferred that the values obtained could not be due to a slight yielding of the paraffin slab when the field was altered, the distance from slab to poles being too great to be affected thereby in this case. Neither could

they be due to a direct effect of the field upon the resistance or E.M.F. of the pile, for we found that they were independent of the direction of any permanent current which might be flowing through the pile during the observations.

It only remained therefore to suppose that our results were attributable to a change in the conductivity for heat of the system pile-testpiece; the flow of heat through it from the exciting current being alternately checked and accelerated as the field was altered. That a flow was actually occurring from the iron into the pile and on through it to its other end, was shown by the direction of the E.M.F. required in the galvanometer circuit to balance the permanent effect of the pile. It was due to the fact that the pile was too long to allow of its being wholly between the poles at the same time that the test-piece was in a uniform part of the field.

The correctness of this view was proved by showing that the sign of the effect changed with the direction in which the