of soda, graphite, and asbestos, with perhaps other ingredients, were used for packing and insulation under the flanges of the main bars and within the cylindrical apertures adjacent.

To prevent leakage at the ends of the numerous guard-ring bars we depended solely on lead washers squeezed hard between the brass cap nuts c and the walls of the pots; and to lessen the danger of straining the apparatus and producing rupture of the joints around the main

bars at the tightening of these nuts, we used on or in each guard-ring bar just outside the walls of pots, nuts or iron pins, as Figure 1 shows. Although in most cases the lead washers thus used were entirely satisfactory, we shall see later that through one or two joints leakage finally occurred.

K and K' are the junctions, 1 cm. apart, of a copper-German-silver thermo-electric couple wound spirally in a groove made in the asphaltum coating of the bar a, as shown in Figure 5 of our former paper, which represents a part of the surface of the bar laid out in a plane. L L', M M', N N', OO', and P P' are similar spirally placed couples. Each such

couple is to indicate the gradient of temperature in its part of the bar on which it lies. Cemented by asphaltum to the bar a, midway between K and K'is a, one junction of a copper-German-silver couple, the other junction of which is kept at a known temperature in a Crafts 4 boiler containing water or naphthalin, which apparatus is not shown in the figure. The object of this couple is to indicate the mean temperature of that part of a which lies between K and K'. The junctions a, a, a ac, Ba, Bb, ẞc, ẞa, and Be, have each a function similar to that of a; for each of them there is a corresponding junction in one or another of four Crafts boilers. The bar G, carries five copper-German-silver junctions, 6, 6b, etc., and the bar Gis a like equipment, 16, 16, etc., corresponding in position and function to the junctions a, Ba, etc., already noted, on the main bars. In every case the junction indicated in the figure is connected with a mate, which can be kept at a definite temperature in one of the Crafts boilers. Each junction attached to a guard-ring bar is really lodged in a small hole drilled in the bar, insulation being secured by means of thin-walled glass tubing, and fixity of position by means of asphaltum varnish. Bars G1 and G11, shown in Figure 2, are similarly provided. We have, accordingly, on the main bars and the guard-ring taken together, six a junctions, six b junctions, etc., each set of six lying approximately, though by no means accurately, in one cross-sectional plane. The effect of displacements from such a plane, though for the most part eliminated in the end by the practice of reversing the gradient of temperature in the apparatus, will require notice in some cases.

In every case the distance from a to b, from b to c, etc., was intended to be 4 cm. The distance from pot to pot was very nearly 21 cm.

The copper and German-silver wires, all about 0.02 cm. in diameter, leading up from the various thermo-electric junctions which have been described, were insulated, where insulation seemed necessary, by passing them through short, thin-walled glass tubes. The wires from the main bars ran almost straight up, as Figure 2 indicates, while those from the guard-ring bars were made to follow the contour of the inner guard-ring for a greater or less distance before leaving the packing with which the guard-rings are surrounded. The wires from bar G1, at the top, were carried down to the bottom of the inner ring, as Figure 2 shows, and then brought up on its outside. The object of these precautions was to make sure that the gradient of temperature in the wires leading from

4 See a paper On Thermo-electric Heterogeneity in Certain Alloys, especially German-Silver, by the authors of the present paper, these Proceedings, March, 1906, for an account of the calibration of thermo-electric couples by use of the Crafts apparatus.

any junction should be very small in the neighborhood of that junction. All the wires, after leaving the packing surrounding the bars, ran nearly straight upward to a wooden platform about 30 cm. above the outer guard-ring. Here all the copper wires were connected with larger copper wires leading off toward the du Bois and Rubens Panzergalvanometer with which all the thermo-electric currents were measured, while all the German-silver wires were continued to the Crafts boilers, about 1 meter distant.

To obstruct the lateral flow of heat from the main bars outward, the outer guard-ring was covered, except at the top, where the wires came through, by a wrapping of magnesia and asbestos about 2.5 cm. thick, and the space within was stuffed rather loosely with wisps or wads of asbestos fibre. Concerning this packing something will be said later.

The space between the square pots was closed on both sides by thick asbestos board, between which and the packing outside the outer guardring there was at mid-level an interval about 2 cm. wide.

The space beneath the packed bars was walled in on all four sides by similar asbestos board, which was intended to prevent, to a great degree, circulation of air, and especially of hot gases from the large Bunsenburner flames used beneath the pots, from passing up past the bars. The space above the bars, up to the wooden platform already mentioned, was boxed in in the same way, though there were, before April 13, two gaps near the top of this enclosure, through which escape of air from the interior was facilitated, it being at first considered unsafe to close in completely the space under the wooden platform.

STUDY OF THE LATERAL FLOW.

The first observations of this kind were made January 26-27, 1906. Both pots contained water boiling at atmospheric pressure, and the outer junctions of all the thermo-electric couples a, b, c, d, and e connected with the main bars and the guard-ring bars were placed. in two of the Crafts boilers containing water boiling at atmospheric

pressure.

The time at which boiling began is not recorded in this case. Observations to test the approach of the bars to a constant condition of temperature began about 6 P. M., January 26, and were continued about two hours, the indications being that a fairly stable condition had been attained by 8 P. M. The boiling was continued, however, till past midnight before any of the thermo-electric observations finally used were made.

The final operations were substantially as follows, though not made in the precise order here given :

1. The couples 1., 6., 11., and 16,, all the guard-ring couples belonging to the cross-section a, were joined in series, and the current sent by this combination through a known resistance was measured. This made it possible to find the difference between the boiling-point of water in the Crafts boilers and the mean temperature of the points a on the guard-rings, and so to find this mean temperature itself.

Similar observations were made with respect to the sections b, c, d, and e of the guard-rings.

2. The couples a, ae, Ba, Be, on the main bars, joined in series, were placed in opposition to the couples 1., 6., 11., and 16,, joined in series, and the current sent by this combination through a known resistance was measured. This gave the means of finding the excess of the mean temperature of the points a and e on the main bars above the mean temperature of the points a on the inner guard-ring.

Similar observations gave the means of finding the excess of the mean temperature of the same points a and e on the main bars above the mean temperature of the points e on the inner guard-ring.

Further observations of a like character gave the excess of the mean temperature of the points b and d on the main bars above the mean temperature of the points b and the points d, respectively, on the inner guard-ring.

The excess of the mean temperature of the points c on the main bars above the mean temperature of the points c on the inner guardring was found in like manner.

3. The spiral couples K K', MM', N N', P P', on the main bars, were joined in series, and the current sent by this combination through a known resistance was measured. This made it possible to find the mean gradient of temperature at the points a and e on the main bars.

From (1) and (2) above we get the following summary of temperatures and temperature differences :

Main bars,

Guard-ring,

Differences,

:

Proceeding with these data, we find by a graphical method that the difference between the mean temperature of the main bars between

The barometer read about 77 cm. corrected, making the temperature of boiling water about 100.4° at the top of the pot, and perhaps 100.7° at the level of the main bars.