would occur at nearly the same moment. Heat this bar to redness, one end of the red-hot bar being firmly fixed (Fig. 41), and sling a weight not sufficient to bend it to the free end, which is lengthened by the addition of a reed to magnify any motion that may take place. As the bar is red hot, it ought to be at its softest when it is freshly withdrawn from the furnace, and, if the weight was ever to have power to bend it, it would be then; but, in spite of the rapidity with which such a thin bar cools down in the air and becomes rigid, points of molecular weakness come when the iron changes from ẞ to a, and the carbon passes from hardening carbon to carbide carbon; at that moment, at a temperature much below that at which it is withdrawn from the furnace, the bar will begin to bend. It has been found experimentally that this is the point at which, according to Osmond's theory, molecular change takes place. Coffin takes advantage of this fact to straighten distorted steel axles.*

Barus traces the connection of this singular minimum of the viscosity of hot iron and the interpretation given of Maxwell's theory of viscosity, and he points out that "when iron passes through the temperature of recalescence, its molecular condition is for an instant almost chaotic. This has now been abundantly proved by Hopkinson. The number of unstable configurations, or, more clearly, the number of configurations made unstable because they are built up of disintegrating molecules, is therefore at a maximum. It follows that the viscosity of the metal must pass through a minimum. Physically considered, the case is entirely analogous to that of a glass-hard steel rod suddenly exposed to 300°. If all the molecules passed from Osmond's ẞ state to his a state together, the iron or steel would necessarily be liquid. This extreme possibility is, however, at variance with the well-known principles of chemical kinetics. The ratio of stable to unstable configurations cannot at any instant be zero. Hence the minimum viscosity in question, however relatively low, may yet be large in value as compared with the liquid state.”

Mr. Anderson has recently urged that, "when, by the agency of heat, molecular motion is raised to a pitch at which incipient fluidity is obtained, the particles of two pieces brought into contact will interpenetrate or diffuse into each other, the two pieces will unite into a homogeneous whole, and we can thus grasp the full meaning of the operation known as 'welding.' It is, however, possible to obtain evidence of interchange of molecular motion, as has been so abundantly shown by Spring, * Trans. American Soc. Civil Engineers, vol. xvi. (1887), p. 324. + Nature, vol. xli. (1890), p. 369.

a

even at the ordinary temperature, while in the case of steel it must take place far below incipient fluidity, indeed, at a comparatively low temperature, as is shown by the following experiment on the welding of steel. Every smith knows how difficult it is to weld highly carburised hard tool steel, but if the ends of a newly fractured inch square 18 steel rod, a (Fig. 42), are covered with platinum foil, b, so as to exclude the air, and if the junction is heated in the flame of a Bunsen burner, c, the metal will weld, without pressure, so firmly that it is difficult to break it with the fingers, although the steel has not attained a red heat.*

FIG. 42.

The question now arises, what is the effect of the presence of other metals in steel, of which much has been heard recently? Take the case of manganese. This metal enables steel to harden very energetically, as is very well known. If much of it be present, 12 to 20 per cent., in iron, no break whatever is observed in the curve which represents slow cooling. (See line marked "Manganese Steel," Fig. 39.) That is, the iron never shows such a change as that which occurs in other cooling masses of iron. Then such a material should be hard, however it is cooled. So it is. There is one other important point of evidence as to molecular change connected with the addition of manganese. Red-hot iron is not magnetic, and Hopkinson † has recently stated that the temperature of recalescence is that at which iron ceases to be magnetic. It may be urged that ß iron cannot therefore be magnetised. Steel containing much manganese cannot be magnetised, and it is therefore fair to assume that the iron present is in the B form. Hadfield has given metallurgists wonderful alloys of iron and manganese in proportions varying from 7 to 20 per cent. of manganese. Professor Ewing and others have specially examined the magnetic properties of this material, and Ewing concludes that "no magnetising force to which the metal is likely to be subjected, in any of its practical applications, would produce more than the most infinitesimal degree of magnetisation " in Hadfield's manganese steel. It has been seen that quantities of manganese above 7 per cent. appear to prevent the passage of ẞ iron into the a form. In smaller quantities, manganese seems merely to retard the conversion, and to bring the two loops of the diagram nearer together. With regard to the effect of other elements on

* Trans. American Soc. Mechanical Engineers, vol. ix. (1888), p. 155. + Proc. Roy. Soc., vol. xlv. (1889), pp. 318, 445, and 457. Proc. Inst. Civil Engineers vol. xciii. 1888, part iii.

steel, it need only be added that tungsten possesses the same property as manganese, but in a more marked degree. Chromium has exactly the reverse effect, as it enables the change of hard Biron to soft a iron to take place at a higher temperature than would otherwise be the case, and this may explain the extreme hardness of chromium steels when hardened in the same way as ordinary steels.

The disappearance of the magnetic properties of iron on heating is of much interest in relation to the allotropism of the metal. Gilbert appears to have been the first to demonstrate, in his treatise de Magnete, published in 1600, that red-hot iron is not magnetic; and nearly half a century later Sir Thomas Browne,* with frequent reference to Gilbert's work, states that masses of iron, "by the fire, omit whatsoever they had received from the earth or loadstone," and he gives evidence of being aware that what is now called the " magnetic permeability" of iron and steel is affected by heating and cooling the metal. These facts have been recognised as being of vital importance in modern research, and they derive new interest from the sharp identification of the loss of magnetism with the temperature at which a molecular change in the iron takes place, and from Hopkinson's recent discovery that an alloy of iron with 25 per cent. of nickel is magnetisable if it be previously cooled (by solid carbonic acid) to a very low temperature.

Working of Steel.-There are a few considerations relative to the actual working of steel which can but briefly be dealt with, notwithstanding their industrial importance. The points a and b adopted in the celebrated memoir of Chernoff, to which reference has already been made, change in position with the degree of carburisation of the metal. It is useless to attempt to harden steel by rapid cooling if it has fallen in temperature below the point (in the red) a, and this is the point of "recalescence" at which the carbon combines with the iron to form carbide carbon; it is called V by Brinell. In highly carburised steel, it corresponds exactly with the point at which Osmond considers that iron, in cooling slowly, passes from the ẞ to the a modification. Now with regard to the point b of Chernoff (or W of Brinell). If steel be heated to a temperature above a, but below b, it remains fine grained, however slowly it is cooled. If the steel be heated above b, and cooled, it assumes a crystalline granular structure, whatever the rate of cooling may

*Pseudodoxia Epidemica: or Enquiries into Vulgar Errors. Second Edition, 1650, p. 45.

be.

The size of the crystals, however, increases with the temperature to which the steel has been raised.

Now the crystalline structure, which is unfavourable to the steel from the point of view of its industrial use, may be broken up by the mechanical work of forging the hot mass; and the in

vestigations of Abel, of Maitland, and of Noble, have shown how important "work" on the metal is. When small masses of hot steel are quenched in oil, they are hardened just as they would be if water were used as a cooling fluid. The diagram, Fig. 43, shows the way in which the tenacity of steel containing varying

Н

amounts of carbon is increased by oil-hardening,* while at the same time the elongation rapidly diminishes. With large masses the effect of quenching in oil is different. Such cooling of large hot masses appears to break up this crystalline structure in a manner analogous to mechanical working. If the mass of metal is very large, such as a propeller shaft, or tube of a large gun, the change in the relations between the carbon and the iron, or true" hardening" produced by such oil treatment is only effected superficially, that is, the hardened layer does not penetrate to any considerable depth, but the innermost parts are cooled more quickly than they otherwise would have been, and the development of the crystals, which would have assumed serious proportions during slow cooling, is arrested. It depends on the size of the quenched mass whether the tenacity of the metal is or is not increased, but its power of being elongated is considerably augmented. This prevention of crystallisation probably constitutes the great merit of oil quenching, which, as regards large masses of metal, is certainly not a true hardening process.

There has been much divergence of view as to the relative advantages of work on the metal and of oil-hardening, but it will be possible to reconcile these views if the facts so briefly stated be considered.

The effect of annealing remains to be dealt with. In a very complicated steel casting, the cast metal probably contains much of its carbon as hardening carbon, and the mass, which has necessarily been poured into the mould at a high temperature, is crystalline. The effect of annealing is to permit the carbon to pass from the "hardening" to the "carbide" form, and, incidentally, to break up the crystalline structure, and to enable it to become minutely crystalline. The result is that the annealed casting is far stronger and more extensible than the original casting. The carbide carbon is probably interspersed in the iron in a finely divided state, and not in crystalline pallets. It would obviously be impossible to "work"-that is, to hammer-complicated castings, and the extreme importance of obtaining a fine crystalline structure by annealing, with the strength which results from such a structure, has been abundantly demonstrated by Mr. J. W. Spencer, of Newcastle.

The effect of annealing and tempering is, in fact, very complicated, as is shown by the long series of researches which Barus and Strouhal have conducted in recent years. They consider that annealing is demonstrably accompanied by chemical change

* This was well shown in Prof. Akerman's celebrated paper on "Hardening Iron and Steel," Journ. Iron and Steel Institute, 1879, part ii. p. 504.