Steel Hardening by Quenching

Steel Hardening by Quenching

The Hardening Power Apparently Lags

In the cooling- curves of the five series of steel which we have just discussed, running from the softest basic steel to very high-carbon tool-steel, we note that each has a marked retardation which reaches a maximum in the neighborhood of 660° C. (1220° F.). These we may temporarily class together as the V-retardation. Now the preceding discussion of the influence of the quenching-temperature on the properties of the quenched steel shows that, certainly in four and apparently in all five series, by far the most sudden and marked change in these properties is caused by the descent of the quenching-temperature through the lower slope of the V-retardation curve. In other words, as measured by the tenacity (where observed) and ductility of the quenched metal, the loss of the hardening power for all these steels is most marked after the evolution of heat has passed its maximum, and hence, presumably, after the V-change has begun to slacken; in short, it seems to lag behind the retardation, and hence behind the evolution of heat which underlies it. Indeed, the tool- steels acquired absolutely no ductility till after quenching had been deferred till the crest of the recalescence was reached.

This fact, if shown only by ductility tests, might be explained away on the carbon theory in the way indicated in (28). But in Series 9 the tensile strength also gives indications of like lagging. At any rate, for the present I insist only on the fact that, if there be any discrepancy between the thermal curves and the loss of the hardening-power, it is that this loss lags behind the retardation; certainly there is no evidence suggesting that it outruns or precedes the V-retardation.

Tipper Retardations

Next, when we come to the softer steels, which have noticeable upper retardations, the loss of the hardening-power which accompanies those upper retardations is not greater in proportion to their size than the loss at the lower or V-retardation. Let us consider this for Series 16, soft steel, and Series 15, very soft basic steel.

(37a) Series 16 (Table 22 and Fig. 6).—The fact that there is no loss but even a slight apparent gain of the hardening-power as the quenching-temperature sinks from 794° to 756° agrees with the fact that there is no considerable retardation in this range (Fig. 6).

The loss of the hardening-power implied by the increase of the permanent bend from 5° to 26° as the quenching-temperature sinks 92°, from 756° to 664°, which is near the beginning of the marked V-retardation, harmonizes with the very noticeable retardation in this range at about 734°. The further great loss of the hardening-power implied by the increase of the permanent bend from 26° to 190° as the temperature further sinks 22° and passes V, corresponds to the relatively great retardation at V. Now, while the upper retardations are much less sharp than that at V, yet they seem to imply at least half as much actual retardation as it does. But it can hardly be claimed that the loss of the hardening-power in these upper ranges, increasing the permanent bend only from 5° to 26°, is more than half of the loss in the V-range, which increases this bend from 26° to 190°.

(37b) Series 15.—In the cooling curve of the soft basic steel, Fig. 8, there certainly seem to be two considerable upper retardations; the one in the neighborhood of 790° is very noticeable. But when we turn to Table 23, we do not find a corresponding loss of the hardening-power; this loss seems to occur wholly in the neighborhood of V. On examining the rate of cooling of other bars of this same steel, I find that the retardation at this upper point is often faint, and sometimes not to be clearly detected at all. This irregularity in the amount of retardation in the upper ranges goes to explain the apparently aimless way in which the tenacity wanders about, varying from 90,319 pounds to 100,705 as the quenching- temperature descends. For instance, bar 24, which seems to have lost none of its hardening-power though the quenching-temperature had been lowered to 661°, shows no recognizable retardation in the upper ranges.

But, while further study must decide whether the loss of the hardening-power does tally closely with the upper retardations, our present evidence indicates that, if there be any discrepancy, it is that the loss of the hardening-power is greater in proportion to the retardation at V than in the upper ranges.

Relation Between Carbon-Content and Rapidity of V-Change

In Tables 9 and 10 I condense some of these results, in order to study the relation between the proportion of carbon in the steel and the rapidity with which the V-change takes place.

Time Needed For V-Change

Up to this point in our study of the progress of the V-change between Va and Vb, we have considered the effect of temperature only.

But the brittleness of steel quenched at, or even below V, evidently arises, at least in part, from too quick cooling through this range of temperature; in moderately slow cooling the metal does



not stay long enough at each degree of temperature to reach the maximum toughness attainable at that degree, i.e., to permit this V-change to proceed as far as it can at that given degree.

Thus Table 19 shows that, when the soft tool-steel of Series 9 was quenched after staying for ten minutes at 1° below V (after the end of the recalescence), it was not only very much tougher than when quenched immediately after cooling moderately slowly to practically the same point relatively to V, bending 37° in the former against only 3° in the latter case; but even slightly tougher than when quenched after slow cooling to 7° below V. In this last case it took sixty-four seconds to cool through these last 7°.

Like results are given in Table 17, and recondensed in Table 11; but they are less conclusive, because the V-point was not noted.


Here, simply by retarding the rate of cooling, the permanent bend of metal quenched immediately after cooling to 657° C. is increased from 18° to 117°, and that of metal quenched immediately on cooling to 635° is increased from 93° to 193°. Further, bar 50, held at 678° for 30 minutes before quenching, bent double, and was thus very much tougher even than bar 41 cooled slowly 21° lower to 657°, and much tougher than bar 47, which was cooled 43° lower before quenching.

So bars 55 and 56, held 3 and 5 minutes respectively at their quenching-temperatures before immersion, were much tougher than the corresponding bars 48 and 47, quenched from the same temperatures without being thus held there.

But, while we thus see that time plays a role second only to that of temperature in the loss of the hardening-power, we have not yet touched the questions raised in (23). We do not know how far it is possible for the loss of the hardening-power to go at any given temperature, nor do we know whether it would become complete at V, and if not at V then at some point but slightly below V, if the metal were only held long enough there; or whether the V- change can complete itself only to a certain extent at V and just below it, so that the metal must cool considerably below V before this change can become complete.

It is not unlikely that, while the soft steels can lose a very considerable proportion of their hardening-power far above V and even above W, a proportion which increases as the temperature descends towards V, the hard steels can lose none of theirs till the temperature falls very nearly to V.

Hardening-Power Gained at W

Let us next study the effects of the W-change by noting the ductility of bars quenched from W and from neighboring temperatures. The results condensed in Table 12 indicate that the change at W is accompanied by a great loss of ductility; for, although the permanent bend decreases in general as the quenching-temperature rises, yet this decrease is extremely marked as we pass W.

The results in group II. of Table 26, for hard tool-steel, indicate that the acquisition of the hardening-power on quenching is due to the temperature, not as such, but as the cause of some change—the W- change—in the metal itself. For whereas, on the one hand, both here and in Table 12, the ductility decreases progressively as the quenching-temperature rises, provided that this temperature be at or above W; yet, on the other hand, if it be below W, the ductility bears no clear relation to the quenching-temperature. I believe that the following explanation suffices. As we shall see, the temperature to which this hard steel has been exposed affects its properties profoundly, even if the exposure be followed by slow cooling. So does


the cool rolling which the metal receives in the rolling-mill. One or perhaps both these influences, which are not fully effaced by thus heating to points below W, affect the ductility far more than the quenching from below W does, even though this quenching be from a fair red heat.

The bars of this steel which received no preliminary heating to above W are placed together in column A of Table 26; those in columns B to E inclusive were preheated and cooled slowly in order to remove the effects of previous cool-rolling. The bars of group II., column A, are much less ductile than most of those in the other columns when quenched from below W.

But even slight variations in the quenching-temperature, once it closely nears or passes W, influence the ductility of the quenched metal to an astonishing extent. Thus, the permanent bend of one of the hard tool-steels (Series 10) falls from 50° to 4° on raising the quenching-temperature 5° C.; that of the other (Series 18) falls from 108° to 5° on raising the quenching-temperature about 5° ; that of the soft tool-steel (Series 9) falls from 85° to 12° and then to 0° on raising the quenching-temperature first by 10° C., then by 5° more (relatively to W); that of the rail-steel (Series 14) falls from 160° to 49° and then to 6° on raising the quenching-temperature first by 5° and then by 12° more; that of the rail-steel (Series 8) is greatly and suddenly lowered; and even that of the soft steel (Series 16) falls from 197° to 36° on raising the quenching-temperature by 13°. I here tabulate these losses:


(41) Series 14.—The most complete set is that of the rail-steel of Series 14, Tables 12 and 21, which shows in eight successive stages an unbroken loss of ductility, from which we infer a continuous increase of the hardening-power, as the quenching-temperature rises from just below W (711° C.) to 730° C.

(42) Series 10.—We have seen in (16) that Wa, in case of the hard tool-steel of Series 10, lies between W—6° and W—10°. In fair accordance with this is the fact that while bar 25, quenched at about W—7° or W—8°, had not gained the hardening-power measurably, i.e., it was as ductile as those quenched from lower temperatures, and even more ductile than many of the bars of this same steel which had been cooled slowly completely, yet bar 26, quenched from a temperature only 2° higher, bent considerably less, 50° against 60° ; and a further rise of the quenching-temperature by only 5°, probably past W, practically destroyed the remaining very considerable ductility, reducing the permanent bend from 50° to 4°. These bars were held for about five minutes at their respective quenching-temperatures.

Bar 106 of this series was quenched five seconds after the temperature had reached W. According to our earlier observations, from 14 to 40 per cent, of the total retardation should have occurred before reaching W ; and therefore, if the acquisition of the hardening- power be proportional to the retardation, this bar should have gained rather more than this proportion of its total hardening-power, and according to equation [1] should be at least as much hardened as steel of 14 x 1.20/100 = 0.17 per cent, of carbon, and perhaps as much as steel of 40 x 1.20/100 = 0.48 per cent, of carbon is when fully hardened. Its bend of 34° indicates that the amount of hardening which had occurred is nearer the lower of the above limits. I believe that steel of 0.17 per cent, of carbon should bend not far from 34° when quenched from above Wb. A direct experiment is needed to establish this.

The record of the heating of this particular specimen shows that but little retardation occurred before reaching W. At the rate of heating here employed, the retardation at W is normally about 100 seconds; hence the slight hardening which this particular bar underwent agrees fairly with the slight amount of retardation which had actually occurred before quenching.

Ductility Lost Suddenly by High-Carbon VS Low-Carbon Steels

As the quenching-temperature rises past W, the harder steels lose their ductility more suddenly than the softer ones. In other words, thus measured, the acquisition of the hardening power is more gradual in case of the soft steels than in case of the harder steels. Table 12 shows us that while the hard tool-steels, Series 10 and 18, leap in a few degrees from great ductility to great brittleness, the hardening-power is gained more gradually and progressively by the rail-steel of Series 14.

Here two suppositions suggest themselves:

  1. That the change which causes this loss of ductility, whatever be its nature, takes place more suddenly in the hard than in the soft steels;
  2. that, if this change be a change in the condition of carbon, a transfer of part, say of one-half of the carbon of a high-carbon steel, to the hardening state, suffices to remove all the ductility, so that a further transfer of the last half of the carbon at a higher temperature to the hardening state causes no effect which a bending test detects. B

ut the change of the last half of the carbon to the hardening state should affect the ductility of a low-carbon steel greatly. Indeed, both these suppositions are probably true. The heating curves (Figs. 1 to 8) support the former supposition. The hard tool-steel (Fig. 1) has a single very sharp retardation at W ; in the soft steel (Fig. 5) the retardation is further spread out, and is seen to continue for at least 100° C. above the crest of W, and, indeed, to grow more marked at a point far above W; while in the very soft basic steel a second and probably a third retardation occur.

Position of W Probably Constant

I remark in passing that the great critical point in heating, W, seems to occur at practically the same temperature in all steels. The variations which have appeared are hardly more than we can readily explain by certain instrumental errors which I may discuss hereafter. In short, the marked hardening-point of all steels seems to lie nearly if not quite at the same temperature; but a further elevation of the quenching-temperature affects the soft steels very much more than the hard ones. Yet I think it probable that, between the limits of perhaps 0.80 per cent, and 2 per cent, of carbon, exactly the same temperature suffices and is necessary for hardening all true carbon steels. I know that this is contrary to tradition, and I give it simply as a conclusion towards which the present data point. Naturally, a needlessly high temperature, which would crack a very hard tool-steel, can be tolerated in case of a softer one.

Be it remembered that the W of Series 10 varied strangely in the abnormal cases in which T min was not below V.

Should we find by further observations that W is as constant as it now appears to be, it would be a most convenient point for calibrating thermo-electric pyrometers, and, indeed, all which permit continuous observations.

Hardening-Power Even in Hard Steels

We have seen that the harder steels lose their ductility more quickly, as the quenching-temperature rises, than the softer steels do. I now describe some experiments which show that even in the case of the harder steels the acquisition of the hardening- power, though rapid, is far from instantaneous.

If we heat a hard tool-steel very slowly, its temperature seems to remain perfectly stationary at W spontaneously for a considerable time. Of course, it is clear that the faster we supply heat to the metal from without the shorter will this arrest be. In these experiments it sometimes nearly reached two minutes. During this time the hardening-power is acquired very rapidly. Three bars of hard tool-steel, Series 10, were quenched while at W, but after remaining there for different periods. The first was quenched five seconds after it reached W (714° C.); its bend on rupture was 34°. The second was quenched after staying for 30 seconds at W; its bend on rupture was only 8°. The third was quenched after it had remained spontaneously at W for 55 seconds. Its bend on rupture was only 4°.

So bars 25 and 31 were quenched at 707°, which is probably between Wa and W. The former was held at this temperature 5 minutes before quenching, the latter for 30 minutes. The longer stay in this range seems to have lessened the ductility of the quenched bar 31, i.e., to have increased the hardening-power. But this last pair of results must be received with caution.
These results are condensed in Table 13:


Neutral Zone

It is extremely probable that we cannot induce the W-change to complete itself, at least in case of the softer steels, simply by holding the metal long enough at W. I have shown, indeed, that this change begins at some point appreciably below W; but it probably cannot go far, no matter how long the metal be held below W. At W the change goes much farther, but still probably cannot complete itself, at least in the softer steels. As the temperature rises higher, this change becomes more and more nearly complete. But whatever be the temperature, an appreciable length of time seems to be needed to permit the W-change to complete itself to the full extent which is possible at that temperature. So with the V-change.

The incompleteness of the change at any given temperature seems, however, to be due partly to what I have already in (24) called inertia. Though the W-change can occur to only a very slight degree below W, yet if we heat the metal highly, and so allow this change to complete itself, it does not correspondingly reverse itself when we again cool the metal to a point slightly below W, at least in case of the harder steels. So, too, though the V-change can occur to only a moderate extent if the metal is cooled to a point just above V and below W; yet if we complete the V- change by cooling the metal far below V and then reheat it to that same point slightly above V, the V-change is not correspondingly reversed. This is shown for Series 10, 9 and 16 by Tables 9 and 10. They show that, if quenched after heating up to 707° C., all three of these steels are relatively ductile; if quenched at even lower temperatures after slow cooling from above W, all three are relatively brittle.
I here condense the results :


Note that steel of Series 9, quenched on heating up to 707° actually bent 50° ; yet it bent only 3° when quenched after cooling down to a point 49° lower; at the lower temperature it was very greatly hardened, at the upper one it was only very slightly hardened.

One might suppose that the reason why the bars quenched after heating up to 707° were much less hardened than those quenched after cooling to even lower temperatures, was that in the former the acquisition and in the latter the loss of the hardening-power had not sufficient time to proceed to the extent corresponding to the temperature reached.

Indeed, it is readily conceivable that if each had been held longer at its quenching-temperature before immersion, the difference in ductility might have been somewhat less; yet I think it most unlikely, at least in case of the harder steels, that this difference can be eliminated or even very greatly diminished in this way. It is most improbable that a longer exposure to 707°, the temperature up to which the high-carbon tool-steel of Series 10 had been heated before quenching, would have caused hardening and thus lessened the ductility ; for this particular bar which bent 60° was held for five minutes at 707° before quenching, and yet when quenched was more ductile than those quenched from a lower temperature. Indeed, as the quenching-temperature rose successively from 667° to 707°, the ductility progressively increased. I here summarize.


With a slight farther rise of the quenching-temperature to 709° the ductility again decreases. This suggests that between 667° and 707°, instead of acquiring the hardening-power, the metal was annealing; but that Wa lies between 707° and 709°, so that here the metal begins to acquire the hardening-power. This is not surprising, for W lies only slightly higher, between 714° and 717°.

To clinch this, I quenched two bars of Series 18, hard tool-steel, after holding them at one and the same temperature, 700° C. (about 19° below W, and about 30° above V) for several minutes, one after simply heating it to 700° C. from the cold, the other after cooling it to 700° C. from 880° C., which is far above W. When quenched, the former was at least as soft as if it had been roughly annealed; the latter was dead hard. This, then, seems a neutral zone,—the hardening power seems neither to be lost nor acquired in it. I here summarize these results in Table 14, giving with them the bends of two bars of the same lot just as received from the rolling-mill, and of two others which had been cooled slowly from 837° C. (1539° F.) in the tube-muffle, and would therefore usually be called annealed.


The fact that, in case of the harder steels, the W-change, including the acquisition of the hardening-power, does not occur till the metal has been heated to a temperature considerably above that at which the V-change occurs with its loss of the hardening-power, reminds us of the fact that the points at which freezing and melting set in do not necessarily coincide. Under favorable conditions water may be cooled considerably below 0° C. (32° F.) without freezing, and this phenomenon is much more marked in case of some other substances; that is to say, there seems to be a certain inertia or reluctance on the part of the substance to pass from the liquid condition to the solid. To those chemists who regard freezing and melting as essentially chemical changes from one allotropic or isomeric state to another, the resemblance is particularly striking. Such a substance can be cooled somewhat below the freezing-point without freezing, but if it be once frozen by further cooling and reheated to this same temperature just below the freezing- point, it does not melt again even in part. So while hard tool-steel does not lose its hardening-power when cooled to 700° C., even though it be held there, yet if it once be deprived of the hardening- power by further cooling, it will not regain it, nor apparently will it even begin to regain it, on reheating to 700° C., nor even on long exposure to that temperature. There seems to be this difference, however, in case of the freezing or melting of water or other simple substance: If the temperature is low enough to permit any freezing, the whole will freeze; if it be high enough to permit any melting, the whole will melt if we allow sufficient time; but, at least in case of the soft steels, this does not seem to be the case with the W-change and the V-change.

There is the further difference that, whereas when pure water thus overcooled begins freezing, the heat evolved raises the temperature to a fixed and absolute maximum-point, the true freezing-point, 0° C. (32° F.), at which freezing now completes itself; in the case of steel the crest of the recalescence (the maximum point to which the heat evolved in the V-change raises the temperature) shifts with the position of V.

In both these respects, however, the behavior of saline solutions resembles that of steel. If such a solution be over-cooled, so that its water does not begin freezing till the temperature has fallen below its normal freezing-point, the maximum-point to which the heat now evolved raises the temperature is not fixed, but shifts with the degree of over-cooling which occurred before freezing began, and is in general the lower the greater the amount of over-cooling. Further, part only of the water actually freezes at this maximum-point, and in order to freeze the remainder we must again depress the temperature below this point. As we lower it thence progressively, more and more of this remainder freezes.

Here we have a striking resemblance to the varying position of V, and of the crest of the recalescence, which, as we have seen, shifts with V ; and, further, a resemblance to the gradual progress of the V-change as the temperature sinks from V to Vb.

(48) The limits of this neutral zone, in which hard tool-steel will neither acquire much of the hardening-power if it be not previously acquired nor lose much if it has been acquired, remain to be determined. It extends neither to W nor to V, for we have noticed that bars quenched when heated nearly, but not quite, to W were somewhat hardened, and that those quenched when cooled to points somewhat above V were somewhat softened; and Table 7 shows us that when this steel is heated nearly, but not quite, to W, there was a slight retardation on again cooling past V, and that if we cooled it nearly, but not quite, to V, there was a slight retardation on again heating past W. So bars 13 and 19, of Series 18, Tables 14 and 26, were treated in the same way as bars 52 and 27, except that they were held at a point 22° lower, viz., at 678° instead of 700° C. Bar 19 took a permanent bend of 43°; if quenched from 700° it should have taken no permanent bend, as the behavior of bar 27 shows. Hence it apparently lost much of its hardening- power between 700° and 678° ; hence the neutral zone did not reach down to 678°.

It is too early to insist that this zone is absolutely neutral, for bar 27 may have undergone part of the V-change, yet not enough to give it any measurable ductility. For instance, if we assume that the changed condition of carbon causes the loss and acquisition of the hardening-power at V and W, then in bar 27 part of the carbon may have passed back from the hardening to the non-hardening state; yet enough may have remained in the hardening state to prevent the bar from taking any measurable permanent set on rupture. As to this, I hope to report later.

Experiments on the Loss/Gain of the Hardening

Power, and on the W- and V-Changes.—It thus appears that the V- change and the W-change are simply the opposite phases of the same transformation, the W-change conferring the hardening-power, the V-change removing it. The hardening-power is gained rapidly during the retardation at W, and is rapidly lost during the retardation at V.

The marked retardation which occurs at W, and which shows that some deep-seated change is occurring, does not take place unless the metal, since last rising past W, has been cooled so far as to induce the V-change. Nor does the well-marked retardation or even recalescence at V occur unless the metal, since last cooling below V, has meanwhile been heated highly enough to induce the W-change.

Further, so far as our experiments go, the amount of retardation in either phase (heating or cooling) is proportional to the amount which, by controlling the temperature reached in the opposite phase, we have permitted to occur in that opposite phase.

Neither change is instantaneous. The absorption of heat in rising past W, like the evolution of heat in cooling past V, occupies an appreciable time; and during this time the hardening-power is progressively acquired in one case and progressively lost in the other. So, part of the hardening-power is gained before the temperature has risen quite to W, i.e., between Wa and W, and part of it is lost before the temperature has sunk quite to V, i.e., between Va and V.

The acquisition and loss of the hardening-powers seem to go on pari passu with the retardations. We have noted only two suggestions of discrepancy in this respect, that the loss of the hardening- power seems to lag behind the V-retardation, and in case of the softer steels which have noticeable upper retardations, it seems to be if anything unduly concentrated into the V-range, and disproportionately small in the ranges where those upper retardations occur.

The distribution of the V- and W-retardations varies considerably, and after unknown laws. The positions of V and of Wb vary considerably; that of Vb probably more: but there is much to suggest that those of W, Wa and Va are nearly and perhaps quite fixed.

W is not far from 713° C. (1315° F.). In case of the hard tool- steel of Series 10, Wa lay near 706° C., and Va near 679° C. There was thus a range of some 27° C. (49° F.) between Wa where the W-change begins in heating and Vb where the V-change begins in cooling. Here Wb normally lies near 723°, and Vb between 645° and 674° ; the W-range covers from about 14° to about 20°, and the V-range from about 5° to about 34°; but further and lower retardations may be related to it.

This range between Wa and Va seems to be relatively and possibly completely neutral, at least for high-carbon steels : that is to say, high-carbon steel which by heating above W has acquired the hardening-power, apparently will not lose it in this range, nor can much of it here be acquired by high-carbon steel which has lost it by previous cooling to below V. In other words, steel quenched in this range will be hardened thereby if it has just previously been above W; but it will not be hardened thereby if it has just previously been below V.


Note on Osmond’s Theory.—He holds that sudden cooling hardens steel, not as has been believed by retaining its carbon in a special hardness-giving condition, but by retaining the iron itself in a purely hypothetical special hard allotropic state of β-iron.

Let us glance roughly over the field. It was known that quenching made steel hard and brittle; that slow cooling made it soft and ductile; that the degree of hardening caused by sudden cooling was roughly proportional to the percentage of combined carbon in the steel; that the carbon in quenched steel was combined with the iron in a condition radically different from that of its combination with iron in slowly cooled steel; and that the passage of carbon from the state in which it exists in quenched steel (called the “hardening” state) to that in which it exists in slowly cooled steel (called the cement slate) occupied a very appreciable length of time, at least under certain conditions.

All of these facts, except perhaps the last, were beyond dispute. In view of the fact that carbon passes spontaneously to the hardening state as the temperature rises through redness, and spontaneously but not instantaneously back to the cement state as the temperature sinks past redness, it was reasonably inferred that the reason why quenched steel was hard was that the quenching offered insufficient time for the carbon to pass from the hardening to the cement state.

It was further known that, as the temperature of steel rich in carbon sinks below redness, its fall arrests itself suddenly and spontaneously, the metal grows visibly hotter, and then again cools off completely; and further, that this recalescence takes place only in steel containing much carbon, and that if the steel be quenched when far above this recalescence-point it becomes glass-hard, but not if it be quenched from far below this point. It was reasonably inferred that this recalescence was due to evolution of heat accompanying the passage of the iron and carbon from their hardening to their non-hardening state of combination.

This Osmond practically admits, holding that it is well established that the recalescence is due chiefly to a change in the relations between iron and carbon, and referring it to the heat which their combination evolves.

Then Osmond found that, during the slow cooling of steels poor in carbon, there were several distinct retardations, such as Figs. 6, 7, and 8 show; one of them at about the temperature of the recalescence of steel rich in carbon, the others at higher temperatures. Each retardation was, of course, due to some change which evolved heat. Were they all due to the same kind of change? Osmond assumed that they were not. Two explanations suggest themselves. The first is that the upper and lower retardations represent successive fractions or instalments of a single change. This change might be composite in its nature, yet its components might accompany each other in more or less constant proportions in the different retardations, just as when we turn the slats more and more we cut off equally the different components of white light.

The second is that the upper retardations are due to a change essentially different from that which causes the lower. M. Osmond chose the latter explanation. His view seems to be that the single retardation of the hard steels at V is due to a group of changes which, though in essence distinct, here occur simultaneously.

These changes are (1) that of the carbon from its hardening to its cement state, and (2) that of the iron from the hypothetical β- to the ∝- state. In the soft steels these changes, he believes, occur separately, causing the different, retardations, of which the upper ones are due to the β—∝ change, the lower to the carbon-change.

This assumed, the question remains: “ Is quenched steel hard because sudden cooling denies time for the known change of carbon from the hardening to the cement state, or, because it denies time for the change from β- to ∝- iron?” M. Osmond chooses the latter supposition.

He reconciles this with the fact that the hardening-power is roughly proportional to the percentage of carbon in the steel, by supposing that the presence of carbon clogs or retards the change from β- to ∝- iron. He supposes that this change occurs nearly instantaneously if carbon be absent; hence, the softness of low-carbon steel even after sudden cooling. The change, he further supposes, takes place more and more slowly the more carbon the steel contains.

Testing Osmond Theory

Admitting for argument’s sake half of Osmond’s theory, let us test by it the other half; i.e., admitting for this purpose that the upper retardations in cooling represent the passage from β- to ∝- iron, and the V-retardation the change from hardening to cement carbon, let us directly test his theory by ascertaining to what degree the hardening-power is lost as the steel, cooling slowly, undergoes each of these successive retardations. To do this we have but to quench, at different points in the range which includes these retardations, a series of pieces of this soft steel which, by heating to above W, have acquired this hardening-power; and then to ascertain how much each is hardened. This test we have already applied to the soft steel of Series 15 and 16; and we have seen that the loss of the hardening-power, far from being confined chiefly to the upper retardation-points as Osmond’s theory requires, is, if anything, disproportionately small there, and unduly concentrated into the lower or V-range. The evidence which I have offered on this point is thus directly opposed to Osmond’s theory in its present form. I propose to apply still more accurate tests shortly.

It would have been well to apply this simple, direct and crucial test rigorously before publishing this theory, which now strikes me as untenable. How often does the investigator greatly reduce the debt which the practitioner owes him for his investigations, by publishing prematurely theories which, by their number and instability, confuse the beneficiary.

Test by Color-Carbon

Not content with this, I next sought an answer to the question: “If I quench a set of pieces of this soft steel of Series 16, which has these several retardation points, quenching each piece from a temperature a little below that at which I quench its neighbor, and if I thus obtain a series of quenched pieces, each less brittle and more ductile than its preceding neighbor, will the condition of the carbon in the quenched pieces correspond to their hardness, and like it to their quenching-temperature? i.e., as the quenching-temperature descends, does the percentage of hardening-carbon progressively diminish and that of cement-carbon progressively increase in harmony with the progressively increasing ductility, slowly as we pass the upper retardation points, quickly as we pass V ?

Closely pressed for time, I could do no more than determine the “ color-carbon,” by the Eggertz method, in the soft steel of Series 16, Table 22. I condense and rearrange the results in Tables 15 and 16. The Eggertz method cannot, I believe, be expected to give quite harmonious results. It simply indicates in a rough way the proportion of cement-carbon present, because part of the hardening-carbon is volatilized by the attack of the acid.

Arranging the cases in Table 15 in the order of temperature, we find very striking anomalies, a bend of 192° with a quenching-temperature of 732°, and bends less than 27° with temperatures from 664° to 704°. These anomalies nearly disappear if we take the condition of carbon into account; for the abnormally ductile bars have an abnormally high proportion of cement-carbon, because they were quenched in water instead of brine. But if we group these cases according to the carbon-condition, we find in Table 16 that, as far as the results go, the ductility agrees roughly with the proportion of cement-carbon, and that both increase markedly and simultaneously as the quenching-temperature sinks past V.

Moreover, in group IV. of Table 22, the change in the carbon- condition, like that in ductility, is, “ if anything, unduly concentrated into the V-range and disproportionately small in the ranges where the upper retardations occur.” Indeed most of each change occurs between 664° C. and 642° or 650°, which, as Fig. 6 shows, is just about the range covered by the V-retardation.

It is not to be supposed that the condition of carbon, if, as I be-



lieve, it be the chief, is therefore the sole element which determines the hardness and ductility. I repeat that the effects of stress and structure cannot be ignored.

We see in (54) how the properties of steel appear to be affected by variations in the temperature from which slow cooling occurs ; that varying this temperature by 28° C. reduces the bend on rupture from 166° to 130° (Table 20); that by varying it some 182° C. reduces the elongation from 32.81 per cent, to 12.50 per cent. (Table 18). Yet these variations in the temperature from which slow cooling takes place can hardly be expected to influence seriously either the proportion of hardening-carbon, or that of β-iron.

Even copper, which, in view of its low elastic limit and great thermal conductivity, should be affected but slightly by stress caused by sudden cooling, is shown in Table 27 to be stronger and far more ductile after quenching than after slow cooling.

Apparent Discrepancy Between Carbon & Ductility

I must now call attention to a discrepancy between bars 6 and III. of Table 22. Their color-carbon is alike, but the latter, quenched from a lower temperature, is much more ductile than the former. Such a single case may be due to error : but, if true, it would mean that, in slow cooling, some ductility-giving change continues after the carbon has fully changed from hardening to cement, i.e., that the restoration of ductility lags behind or at least continues after the carbon-change. This wholly unexpected phenomenon reminds us startlingly of another equally unexpected, which we saw in (36),that the loss of the hardening-power seemed to lag behind the V-retardation. Of this lagging I do not remember hearing before now the faintest suggestion. But if further experiments should verify it, and prove that, while part of the restoration of ductility accompanies, another part of it succeeds the carbon-change, is this second part a true loss of the hardening-power, or is it due to underlying changes like those which cause such enormous differences in the properties of slowly cooled steel, and which contain little suggestion of hardening? In other words, does any considerable loss of hardness proper, of resistance to abrasion, accompany that portion of the restoration of ductility which thus seems to occur after the end of the carbon-change ?

Osmond is estopped from asserting that this apparent lagging is due to a change from β- to ∝- iron, because he has demonstrated to his own satisfaction that this change from β- to ∝- iron causes the upper retardations, not the lower : while this suspected change below V, which causes this apparent lagging, is one that, if it exists at all, seems to cause no retardation and occurs below V.

I do not think that we are yet justified in drawing strong inferences as to the nature of the changes which cause the upper retardations ; and I deplore the haste and positiveness that have been shown in defining it and in declaring it freed from all hypothesis.