Heat Treatment of Steel

Heat Treatment of Steel

This paper may be regarded as a report of progress of an investigation into the influence of prior exposure to different high temperatures, under different conditions, on the properties of steel after it has been cooled completely. The following are the chief divisions:

  • II. —Experiments on toughening manganese-steel by sudden cooling.
  • III. —A study of the critical points of common or carbon-steel, chiefly by examining the spontaneous retardations which occur in heating and cooling it, and by examining the ductility of pieces of steel which have been cooled suddenly from different temperatures by quenching them in water or brine.
  • IV. —A brief note on Osmond’s theory.
  • V. —A study of the critical points by examining the ductility of pieces of steel which have been cooled slowly from different temperatures at, above, and below them.
  • VI. —A description of the furnace used, with notes on thermoelectric pyrometry.

It remains to throw further light on the questions studied by ascertaining the condition of the carbon and the hardness proper, and by studying the microstructure and the fracture of the many pieces treated. I hope to present later the results of experiments in some of these directions.

The temperature of the pieces treated has been very carefully regulated, and has been measured with accuracy by means of the thermo-electric couple devised by Le Chatelier, using the dead-beat galvanometer of Deprez and d’Arsonval. In cases where extreme accuracy is especially desirable, I believe that the error in measurement rarely exceeds 3° C. I here refer not to the error in the absolute temperature, but to the error in the difference between temperatures, the effects of exposure to which are to be compared closely, i.e., the relative error.

As the present study aimed chiefly at a rough survey or reconnaissance of the field, extreme accuracy has, however, at times been purposely sacrificed to expedition. First let us learn the general features of our laws, and later study with accuracy the important ones which we have thus rapidly discovered.

Nearly all the heatings have been carried on in a furnace which I designed for work of this kind, and which I call a “ tube-muffle.”

Toughening Manganese-Steel by Sudden Cooling

The remarkable toughening which manganese-steel undergoes when cooled suddenly is now generally known under the name of “ water-toughening.” Some attempts to learn the conditions under which this toughening occurs are recorded in Tables 5 and 6. The former treats chiefly of variations in the rate of cooling as affecting unforged cast manganese-steel; the latter treats of variations in the temperature at which cooling begins, and of the persistence of the toughness due to quick cooling, in case of forged manganese-steel.

I adopt the bending test because it gives comparable results so very quickly and at such slight cost. I bend the treated bars between dies under a hand press till rupture occurs, or till they are bent double.

Variations in the Rate of Cooling and in the Quenching Temperature

In their natural state, direct from the mould, after grinding roughly with emery, cast bars 5/8-inch square and 6¾ inches long bent only 6° on rupture; yet, after very fast cooling, they bent 212° without breaking through.

It was to be inferred from Hadfield’s experiments that, as we raise the quenching-temperature and as we increase the rapidity of cooling, or, in short, as the cooling becomes more violent, so does the toughness of the quenched metal increase. This we may call the direct effect of the violence of cooling. We may discriminate sharply between this direct toughening effect and the effect of residual stress which rapid cooling should set up, and which we should expect would be very marked in a material which conducts heat so slowly, because this stress is due to difference in the rates at which different layers cool and hence contract. Such residual stress should in itself tend to make the steel brittle. Suddenness of cooling then should have two opposite effects on toughness; directly it increases toughness; indirectly, through causing stress, it should lessen toughness.

We might fear that when the violence of cooling reached some determinate limit, the enbrittling stress-born effect of any further increase of violence might outweigh its direct toughening effect. This would be especially dreaded in case of pieces most liable to receive stress on rapid cooling, e.g., thick pieces and those of suddenly-varying sections. Experience shows that it is possible, under favoring conditions, to crack such pieces by violent quenching. I infer unhesitatingly that we are even more liable to induce injurious stress, outweighing the direct toughening effect, than thus to break up continuity.

In the present experiments, however, made on thin pieces of regular constant section, every increase of violence caused a further increase of toughness; in other words, no limit was found beyond which further increasing the violence of cooling failed to increase toughness further.

The influence of the rapidity of cooling is shown in (3 a) by comparing bars quenched from the same temperature in media varying in conductivity; and, the medium being constant, by comparing in (3 c) to (3 g) bars quenched intermittently, or after partial cooling in the air, with those held in the cooling bath till cold.

Influence of Different Quenching-Media

Taking these media in the order of conductivity, iced brine first, then cold water, boiling water, oil, and molten lead, we find in Tables 1 and 5 that the ductility of the quenched metal, as measured by the degree through which it bends, stands in the same order, viz.: 212°, 170° and 110° (flawed piece), 125°, 106°, 100°, for bars A21, E, F, G, A22, and A20 respectively, which were quenched from about the same temperature. So with bars 4, 8 and 15, which were quenched from a common temperature. The first two, quenched in cold water, bent 78° and 80°, respectively ; the last, quenched in boiling water, bent only 56°. Bars 14 and 16, though quenched from this same temperature in iced water, bent only 48° and 58° respectively ; but this may be explained by their breaking at flaws.


Neither here nor in Hadfield’s tests does iced water give better results than water at the common temperature.

It Is Best to Quench from a Very High Temperature

By comparing bars 1, 4, and 8, which broke at from 78° to 83° after quenching from 952° and 992° C., with those quenched in water from above 1050° C., which (when not injured by flaws) bent at least 135°, and, in one case, 210°, we see how important a high quenching-temperature is. Even those quenched intermittently from 1050° C. + , bent better than those quenched uninterruptedly from 992° and 952°. Group IV. of Table 5 shows us how rapidly the benefit caused by quenching decreases with the quenching temperature. Taken in connection with bar A20, which bent 100° after quenching in molten lead, they indicate that the toughening is due chiefly to a rapid passage through the upper ranges of temperature, say from a moderate yellow heat to dull redness; and that rapidity of the further cooling from dull redness down, while beneficial, is much less important. It is certainly very striking that, though metal quenched from 952° C. will bend 80°, if we allow it to cool to 743° before quenching, it will bend only 26° ; and if we further defer quenching till it cools from 952° to 592° it will bend but 18°, or only 12° more than an untoughened casting.

Table 2 condenses some results which bear on this subject.

Slightly Deferring the Immersion is Practically Harmless

This is shown by bars H and J, group III., Table 5, which were held 15 seconds and 30 seconds, respectively, in the air before quenching. In 30 seconds the exterior of the bar had cooled from a low


“Maximum Aperture” Defined.—Many bars can, without breaking, be bent till their ends touch, but yet break when we press them closer together and before they are completely flattened down. For comparing such bars with each other, the angle through which they have bent is no just measure of their relative ductility. Another measure suggests itself. The more ductile the bar, the closer can its two halves be pressed together before it breaks, and the smaller will be the distance between those halves when rupture occurs. I call this distance the “ maximum aperture.” Roughly speaking, the ductility is inversely as the maximum aperture.

white heat to a moderate red heat. The interior of the bar, doubtless, had lost very little heat. These two bars bent 180° and 205° respectively before cracking, and 200° and 210° respectively, before breaking. This result is what we should anticipate. The material conducts heat so slowly that its interior is not considerably cooled during moderate exposure to cold air. These bars actually bent farther than those (E and F) which were quenched immediately and uninterruptedly ; but the latter were injured by flaws.

Slight Retardation of Cooling Does but Slight Harm

This is exemplified by bar A25, which, instead of being held in water until it was cold, was simply immersed for two seconds, held in the air for two seconds, then re-immersed for two seconds, and so on until it was cold. This procedure retarded the cooling very markedly, yet the bar bent 165° before cracking, and 193° before breaking apart.

Great Retardation of Cooling Can be Tolerated

Carrying this simple idea farther, bars A23 and A24 give like results. They were held alternately in water and in air for 3-second periods in each medium in case of A24, and for 5-second intervals in case of A23. They bent 128° and 135° respectively. Bar A22 was in like manner dipped intermittently into oil, held in the air, re-immersed, and so on. It bent 106° before breaking. Bar G was held in boiling water till cool; it then bent 125°.

The toughness given by this greatly retarded cooling is certainly much less than that given by uninterrupted cooling in water; nevertheless I believe that it is more than sufficient for most purposes.

Rapidity of Cooling not Imperative Below Dull Redness

Bar A20 was immersed in molten lead, withdrawn, and re-immersed repeatedly, until its exterior had sunk to about dull redness. The temperature was not clearly noted, but it must have been considerably above the melting point of lead. The bar was then allowed to cool in the air completely. It bent 70° before cracking, and 100° before breaking. It is thought that this toughness is greater than will often be needed.


The application of these results to the case of thick castings, and of those of irregular cross-section, is obvious. Though none of the bars experimented on were noticed to crack, even on the most sudden cooling, yet such a result is to be feared in case of very thick castings. An obvious precaution would be to cool relatively slowly, as for instance in a jet of air or a spray of steam, or of mixed air and water, or by intermittent quenching. We may infer, also, that no special precautions for getting the water inconveniently cool are needed, or for extreme and precipitate haste in conveying the metal to the quenching tank, which, moreover, may be located in any convenient place, without reference to extreme rapidity of transfer from the furnaces to it. In considering these results we must not forget that, as is shown in a special column of Table 5, some of the castings broke at flaws.

Castings Should be Heated Slowly

Because manganese-steel conducts heat extremely slowly, if it be heated suddenly its outside will become very hot and will expand greatly while its inside is still cool and has expanded but slightly. Though this unequal expansion may do no direct harm to such a tough substance as forged and toughened manganese steel, yet it may crack the relatively brittle untreated castings; and even if it does not it may defeat sudden cooling. This, to be effective, must start from a very high temperature. If either castings or forgings of manganese-steel be heated quickly, at the time when their outside is as hot as it can safely be, their inside may yet be so cool as to receive no important benefit from the sudden cooling.

This is illustrated by group VI. of Table 5. Bars B 1, B 2, C and D, heated to whiteness very rapidly and quenched in cold water, bent in four cases less than 46°. The fifth bar indeed bent 70°, after being held for 30 seconds in the air before quenching. This suggests that during the 30 seconds that the bar was held in the air, although the outside became visibly much cooler, and therefore should be less benefited by the sudden cooling, nevertheless, this injury was outweighed by the fact that the heat from the outside soaked into the interior of the bar, raised its temperature, and thus increased the benefit of quenching. But this bar may have been heated less quickly than the others.

Even this rapid heating was not without benefit. A bend of from 29° to 45° is certainly very much better than the bend of only 6° which the untreated metal gave.

If rapid heating thus lessens the benefit of quenching such thin pieces, how much more must it be avoided in case of thick pieces, whose interior can be but slightly warmed by the time that the outside, if heated quickly, is as hot as it can safely be.

Influence of Repetition of Quenching

The results condensed in Table 3 from Tables 5 and 6 do not show that repeating the heating and quenching increases the toughness of the material, though final conclusions cannot safely be drawn from such scanty data. Their showing is what we should expect; for the effects of previous quenching seem to be obliterated by a subsequent slow cooling, and those of slow cooling seem to be effaced by later quenching, as I shall now try to show. Hence we should expect that the reheating for the purpose of requenching would in itself efface the benefit of the previous quenching. The injury done by slow cooling, and illustrated by group IV. of Table 5, and to a smaller extent by bar 23 of Table 6, has previously been well established. Bar A of Table 5 further shows that this injury is readily cured, simply by reheating and quenching. It was first cooled slowly from whiteness, which must have made it extremely brittle. It was then reheated to a little above 1050° C. and quenched in water; it was then as tough as any of the other bars tested, and did not break quite across when bent 208°.

Quenching Effaces the Injury done by Previous Slow Cooling

Boiled bar 10, Table 6, after cooling slowly was reheated to 952° and quenched in cold water, when it was bent double, and did not break till it was hammered together so closely that the maximum aperture between its sides was 0.37 inches. Thus it was actually


slightly tougher than bar 1 which had been quenched from the same temperature, 952° C., without previous slow cooling.

Reheating Destroys the Benefit of Previous Quenching

Table 4, condensed from Tables 5 and 6, shows this. We note that in each


case the quenched bar bent farther than that which was cooled slowly.

The bar which was cooled slowly from 664° C. seems to present an anomalous result. If we except it, the following inferences suggest themselves, but they must be taken cautiously because the cases are so few. The higher we reheat, the more we efface the toughening caused by the previous quenching, and hence the more brittle the metal, if we now cool it slowly. If, however, we again quench it, this second quenching will to a certain extent remove the injury caused by reheating, and the more so the higher the temperature


The bars which are said to have been heated to 1050° were heated in a coke-fired muffle till a lump of copper which stood among them melted.
The bars of group VI. were heated in direct contact with a hot coke fire.
All the rest were heated in the tube-muffle.
Bars A23, A24 and A25 were quenched intermittently, as follows: A25 was drawn from the muffle, immersed and moved up and down in cold water for two seconds, withdrawn and held in the air for two seconds, re-immersed for two seconds, etc., etc. Bars A24 and A23 were treated in the same way, except that the intervals were of three seconds and five seconds respectively.
Bar A22 was quenched intermittently in oil.
Bar 12 was heated to 952° C., quenched in cold water, reheated to 952° C., and requenched in cold water.
Bar 13 was thus heated and quenched three times.


[NOTE TO TABLE 6.—Many bars did not break when they were bent so far that their ends met. These bars were then further flattened in the press till they broke, and the widest aperture was then measured and its width recorded in column 7. The smaller this aperture, the tougher is the bar.]

from which we now quench; hence the injury in this case reaches a maximum with a certain degree of reheating, between 542° C. and 808° C. If the temperature to which we reheat be above that of maximum harm, the benefit caused by the re-quenching increases faster with further rise in temperature than the injury caused by the reheating itself does; and hence the higher we reheat the better.

If, however, the metal is to be reheated to some relatively low temperature, some temperature below that of maximum harm, then the lower this temperature the better. For in this range below that of maximum harm, the injury due to reheating increases faster with the temperature, than the benefit due to the quenching does.

Here we have another suggestion that it is in the higher ranges of temperature rather than in the lower, that slow cooling injures manganese-steel and sudden cooling benefits it.


These experiments brought out strongly one of the striking peculiarities of maganese-steel, its non-fissibility ; cracks do not propagate across it readily. It often happened in the bending test that the bar would crack at many places long before rupture. The first crack, instead of causing rupture, would open a little way, and then seem to cease opening, while other cracks would open later. I have often noticed this in the regular bending tests in the manufacture of this material, at the Taylor Iron and Steel Works; but a careful study of this peculiarity has not yet been made.

Study of the Critical Points of Carbon-Steel by Quenching and by Thermal Curves