The distributor is heated at two opposite points on a diameter by the burners, _b b_, above which are the chimneys, _e e_. The cooling of the alternate section is aided by the circulation of cold air, which is effected by means of the draught in the chimneys, _e e_. At the points of lowest temperature a jet of air or water is maintained. The cross arms are insulated with mica or asbestos at the points where they extend from the armature to the distributor. It will now be evident that while the distributor is entirely cool, many of the lines of force pass from N to S without entering the armature core; but if heat is applied at the points 1 and 2 in the figure, so as to increase the magnetic resistance at these points, then a great portion of the lines will leave the distributor, and pass through the armature core. Under these conditions, so long as heat is applied at two points equidistant from N and S, we might, if we so pleased, cause the armature to be rotated by an external source of power, and we should then have an E.M.F. generated in the armature coils--that is to say, the machine would work as an ordinary dynamo, and the power expended in driving the armature would be proportionate to the output. Suppose next that the points of heating, and with them the alternate points of cooling 90 deg. apart, are shifted round about 45 deg., so that the two hot regions are no longer symmetrically situated in respect to each pole of the field. The distribution of the magnetization has therefore become unsymmetrical, and the iron core is no longer in equilibrium in the magnetic field. We have, in fact, the conditions of Schwedoff's experiment upon a larger scale, and if the forces are sufficient to overcome the frictional resistance, a rotation of the ring ensues in the endeavor to restore equilibrium. The regions of heating and cooling being fixed in space, this rotation is continuous so long as the difference of temperature is maintained. The ring in rotating carries with it the armature coils, and of course an E.M.F. is generated in the same way as if the motive power came from an external source. In this respect the machine therefore resembles a motor generator, and the rotation is entirely automatic. The armature coils are connected with a commutator in the usual way, and the field may, of course, be excited either in shunt or in series. M. Menges says that the residual magnetization is sufficient in his machine to start the rotation by itself. When the machine is to be used as a motor, it is evident that the windings on the armature core need only be sufficient to supply current to excite the field, or by the use of permanent magnets they may be dispensed with altogether. M. Menges has further designed a large number of variations on the original type, varying the arrangement of the several parts, and employing armatures and fields of many different types, such as are already in use for dynamos.