Prof. A. N. DOBROVIDOV

COLD BRITTLENESS

OF CARBON STEEL

(brittleness of steel at low temperatures)

United Scientific and Technical Publishing House
Sverdlovsk—Moscow
1937

PREFACE

It has long been known that ordinary carbon steel is very brittle at temperatures below zero. When operating in the open air, machine parts break on winter days more often than in summer.

In recent years engineers have had to deal with temperatures close to absolute zero. The use of liquefied and compressed gases is rapidly expanding, and the demand is growing for apparatus operating at low temperatures. And whereas at the beginning of this century interest in the question under consideration was not very great, in recent years it has increased enormously.

The author of the present study was faced with the task of finding a way to reduce the brittleness of ordinary carbon steel at low, natural temperatures—those at which, for example, rails and parts of railway rolling stock operate in regions with a severe climate.

The investigation was begun at the Siberian Physico-Technical Institute by the honored scientist, Doctor of Physical and Mathematical Sciences Prof. V. D. Kuznetsov, the author, and members of the Institute staff (1930–1931); it was then continued by the author at the Siberian Institute of Metals (1931–1935) and completed at the Tomsk Industrial Institute.

As will be seen from what follows, an increase in the toughness of ordinary steel at temperatures of the order of −40–60° is achieved by the appropriate heat treatment—full hardening and tempering to sorbite.

In the first instance, in accordance with the directives of the NKTP, rail steel was subjected to such treatment. Despite the seeming simplicity of the shape of a rail in comparison, for example, with a twist drill or a milling cutter, hardening it is not simple. Special and rather complex methods and devices are required to carry it out. Rail steel is very diverse in composition and properties and is not as pure as tool steel or even structural steel that is heat treated. This in itself introduces a certain complication into the course of heat treatment.

The rolling temperature is not constant (additional heating of the rail before hardening would make the entire process considerably more expensive), and the possibility of sound hardening when the heating temperature is variable becomes, in general, doubtful.

Rail steel ages, changes its properties over time; during cold straightening a rail acquires plastic deformations, which in turn should intensify the effect of aging. Obviously, the straightening of an already sorbitized rail cannot pass without a trace; it must certainly affect the properties of the metal at low temperatures.

The question itself of the increased brittleness of certain metals at low temperatures is of great interest from the purely theoretical side.

In what follows, the expression “brittleness at low temperatures” (on the order of −50°) will be replaced by the term “cold brittleness.”

Fig. 1

In winter, when the air temperature reached −50°, mass fractures of rails, tires, and car axles were observed.

As is generally known, the largest number of rails goes out of service as a result of wear; other reasons for removal are determined chiefly by fracture. Usually defects of the latter kind may be reduced to approximately the following groups: 1) transverse fracture, 2) cracks, and 3) spalls. An especially large number of defects was accounted for by transverse fracture, and somewhat fewer by other kinds of defects.

In Fig. 1 two curves are given: 1) the number of rails removed per month \((B)\), and 2) the mean monthly temperature \((A)\), according to data from the Tomsk

Fig. 2

meteorological station for the period from October 1928 to April 1930. For greater clarity, temperatures below zero are plotted upward, and those above zero downward.

According to Dr. N. N. Davidenkov (1)*, “the parallel character of both curves—the temperature curve and the accident curve—proves to be striking.” Indeed, the greatest number of rails removed occurs in the months with the lowest mean temperature.

Of course, mean monthly temperatures are not a characteristic of the thermal regime for a given month. Likewise, two different months with the same mean temperature may have a very different distribution of mean daily temperatures over the individual days. Under the conditions of Siberia, the daily amplitude of temperature fluctuation in January in some cases reaches 53°.6 (2). This circumstance must be specially noted as a peculiarity of the Siberian winter.

The easiest defect to detect in track is a defective rail with a transverse fracture. The case of transverse fracture most closely corresponds to the conditions of the Charpy impact test, and this method of determining impact toughness was adopted by us as the principal one throughout our work.

Figure 2 gives the number of type III and IIIA rails and type 32.2 kg rails removed as a result of transverse fracture, as a function of the mean (curve B), minimum (curve A), and maximum (curve C) temperatures in the same years (cf. Fig. 1).

The diagram presented in Fig. 2 emphasizes still more sharply the dependence of the number of broken rails on temperature.

Table 1

	February 1928	December 1929	January 1930	February 1930
Mean monthly temperature (in °C)	−15.4	−27.8	−18.3	−18.0
Minimum temperature (in °C)	−48.8	−47.9	−47.1	−35.3
Percentage of rails removed, of the two types, out of the total number of rails removed	37.8	25.0	42.3	11.6

From the comparisons made in Table 1 it is evident that, despite the approximately identical mean temperatures of January (−18°.3) and February (−18°.0) 1930, the number of rails removed is different (42.3 and 11.6%). But if, in addition, one takes into account the difference in the minimum, rather than in the mean, temperatures of those same months, it becomes clear that rail fracture depends chiefly on the minimum temperatures. Indeed, the lowest temperature

* The figures in parentheses denote references to the literature sources placed at the end of the book.

was −47°.1, and February −35°.3. The stronger the frost, the more rails break.

On the other hand, December 1929 was considerably colder than January 1930 in terms of the mean temperatures (−27°.8 and −18°.3), and differed very little in terms of the minimum temperatures (−47°.9 and −47°.1). The number of rails removed in these months was the same. It is also necessary to take into account the possibility that rails which actually burst at the end of December or the beginning of January were removed, but in the statistics were assigned to the January defective rails.

If we knew the exact distribution of temperatures along the entire railway and the number of rails removed at specific locations, then a more definite dependence between air temperature and the number of broken rails would undoubtedly be revealed.

The graphs presented (Figs. 1 and 2) provoked a very lively discussion, in which engineers of the track service of the Tomsk Railway, line personnel, and participants in the meeting on sorbitizing rails held in Kharkov in 1931 took part. It is interesting that railway engineers considered the chief cause of rail breakage to be the peculiar condition of the track in the winter months, characterized by the presence of frost heaves. There is no doubt that all track defects will have their greatest effect in the cold season, when the metal of the rails acquires an abnormally increased brittleness, irrespective of the general condition of the superstructure.

Incidentally, F. N. Rybinskii (3) points out that the majority of the brittle rails he investigated that broke in service (22 out of 32) broke precisely in winter. He also notes that the large number of rail breakages on German railways during the severe winter of 1928/29 was explained in German specialized journals by the prolonged and severe frosts.

I. LITERATURE DATA

V. Kirpichev (4) writes in his course Strength of Materials: “Rails frozen to \(-15^\circ\) often prove to be very brittle and break from a single blow of the ‘drop hammer,’ although at ordinary temperature they withstand several blows without breaking and acquire a noticeable deflection, indicating the plasticity of the metal.”

T. Andrews showed that at \(-18^\circ\) iron axles are very brittle and their fracture surface proves to be coarse-crystalline; meanwhile, at ordinary temperature a fibrous fracture was obtained, indicating the toughness of the metal.

As Körber and Pomp (5) assert, the most sensitive indicator of brittleness is the test for impact bending of notched specimens. One of the first experiments on the influence of temperature on impact toughness was carried out by Charpy (6). He investigated, among other things, two grades of very mild steel: with a carbon content of 0.04% (Fig. 3, curve A) and 0.14% (curve B). From the soundest part of the ingot, by forging and rolling, bars of square section with a side of 30 mm were made. To eliminate possible work hardening during the preparation of the specimens, the bars were annealed at a temperature of about \(900^\circ\), quenched in oil, and tempered (further details of the heat treatment are not indicated). As is known, German authors call such heat treatment “refining.”

Fig. 3

Specimens with the dimensions indicated in Fig. 3 were tested on a pendulum impact machine of 200 kgm at various temperatures: from \(-80\) to \(+500^\circ\). The portions of the curves relating to specimens that bent without fracture are given by thin lines.

The character of the curves is very similar: both materials showed a sharp drop in toughness at low temperatures, coinciding at \(-80^\circ\) practically at one point. It is somewhat unclear why the softer metal (0.04% C, curve \(A\)) gave a somewhat lower value of impact toughness at room temperature than the harder one (C—0.14%, curve \(B\)). One may assume the presence of defects in the material or some irregularities in the process of the preceding hot and heat treatments.

Fig. 4

Of less interest is the work of Guillet and Revillon (7) (Fig. 4), who investigated a steel with a carbon content of 0.218%. Specimens annealed at \(850^\circ\) were cooled in air (curve \(A\)). Steel artificially spoiled by overheating (30 min at \(1000^\circ\)) gave a considerable decrease in impact toughness (curve \(B\)) (down to 2 kgm/cm² as against 19 kgm/cm² for normal steel) at room temperature.

The curves of Charpy and Guillet have the same character, despite the great difference in the dimensions of the specimens and in the magnitude of the impact machines.

Goerens and Hartel (8) investigated, on a 75 kgm impact machine, two grades of steel, the chemical composition of which is given in Table 2.

Specimens were made from a rolled strip of section \(30 \times 10\) mm, the dimensions of which are given in Fig. 5. As is seen from Fig. 5, the impact-toughness curves have a clearly pronounced maximum near room temperature. Brittleness increases very rapidly when moving toward lower temperatures. The head part of the ingot (curve \(A\)—steel 1; curve \(B\)—steel 2) did not give a large difference in toughness in comparison with the lower part (curve \(C\)—steel 1; curve \(D\)—steel 2). At low temperatures both curves merged almost into one.

Table 2

Steel grade	Part of ingot	C	Mn	S	P	Si
Steel 1	Head part of ingot	0.050	0.46	0.098	0.094	—
Steel 1	Lower part of ingot	0.089	0.38	0.039	0.059	—
Steel 2	Head part of ingot	0.084	0.38	0.064	0.061	Traces
Steel 2	Lower part of ingot	0.085	0.38	0.060	0.066	Traces

Reingold’s experiments (Reingold) (9), who worked with specimens of the same dimensions and of approximately suitable chemical composition (C—0.08%; Si—0.24%; Mn—0.36%; P—0.050%; S—0.040%), gave results similar to Terence’s data (Fig. 6).

Fig. 5

Strauss and Fry (10) investigated the following steel: C—0.51%; Mn—0.51%; P—0.025%; S—0.013%.

Fig. 6

The first series of tests used metal rolled at a high temperature (“hell gewalzt”), followed by one-hour annealing at 900° (Fig. 7, curve A); the second series of tests used metal rolled at a lower temperature (“dunkel gewalzt”), annealed after rolling at 700° for 1 hour and then bent cold along an arc of radius 250 mm. As can be seen (curve B), such deformation considerably reduced the impact toughness in general, and especially at low temperatures. At higher test temperatures (200–300°), the preliminary plastic deformation had no effect at all on the impact toughness.

Langenberg (11) tested large specimens (160 × 30 × 30 mm) on a 300 kgm Charpy pendulum. The specimens, whose shape and dimensions are given in Fig. 8, were made of steel of the following composition: C—0.176%; Mn—0.69%; Si—0.053%; P—0.053; S—0.036%; Ni—0.011%; Cr—0.016%. The steel was first subjected to annealing (curve A) and to quenching-and-tempering treatment (curve B). For annealing, the specimens were heated to 925° and, after a half-hour hold at this temperature, slowly cooled together with the furnace. The quenching-and-tempering treatment consisted of a half-hour hold at 900° followed by oil quenching, tempering for one hour at 760°, a second oil quench from this temperature, and finally tempering at 200°. This complex heat treatment increased the impact toughness at room temperature more than tenfold and produced no reduction in brittleness at −62°, the lowest temperature used by the author.

Richardson and Mac Nutt (12) found a strong decrease in the toughness of steel (C—0.18%; Mn—0.57%) deformed by bending at 300°. As can be seen from Fig. 9,

steel, annealed at 925° and cooled with the furnace, showed at +20° an impact toughness of 2 kgm/cm², and at +100°—12 kgm/cm² (curve B). Work-hardening by bending (curve A) reduced the value of the impact toughness at the same test temperatures, respectively, to 0.4 and 2 kgm/cm².

Comparing Figs. 7 and 9, we note that deformation of cold metal and of metal heated to 300° affected the impact toughness at 200° differently. According to Strauss (see Fig. 7), the toughness of material deformed in the cold state and of undeformed material is the same at 200°; according to Richardson, work-hardening at 300° reduces the toughness at +200° by half.

Attention is drawn to the absence of a decrease in toughness at −70° (Richardson, see Fig. 9) after work-hardening. Perhaps this can be explained by the low toughness at low temperatures of the original metal, or perhaps by insufficient sensitivity of the impact testing machine.

Fig. 7

Terence (13) did not find a great difference in the impact toughness of soft Siemens-Martin steel in the annealed (curve A) and improved (curve B) states (Fig. 10). His data appear to be in sharp contradiction with the data of the authors cited above.

According to Mailänder (14), annealed iron (Fig. 11), (curve A) shows lower impact toughness, especially at 0°, than improved iron (curve B).

Kerber and Pomp (15) carried out an extensive investigation on mild steel in various states. From steel with a content of C—0.05%; Mn—0.33%; Si—0.02%; P—0.020%; S—0.028%, specimens of dimensions 120 × 15 × 15 mm were prepared (notch—3 mm in diameter, live section of the specimen—8 × 15 mm) and were tested on a Charpy impact machine of 75 kgm at temperatures from −80° to 500°.

Five series of experiments were conducted on one and the same material, but treated in different ways. The test results are summarized in Fig. 12.

The first series (curve A) included specimens improved by heating to 950°, followed by quenching in water and tempering for half an hour at 650°.

The second series (curve B) included specimens of rolled metal without any additional treatment (“as delivered”).

Fig. 8

For the third series (curve C) of tests, the specimens were heated for 2 hours in a muffle furnace at 1250° and cooled together with the furnace.

In the specimens of the fourth series (curve D), a coarse-grained structure was produced by recrystallization. The steel was first annealed for 2 hours at 930° and slowly cooled; after cooling, it was deformed by compression by 10% and annealed, in order to carry out recrystallization, a second time for 2 hours at 850°. And, finally, the metal was deformed under blue heat (curve E). For this purpose, specimens annealed at 930° for 2 hours and slowly cooled were rolled at 200°. In this process the thickness of the strip was reduced by 10%.

Under the influence of such varied treatment, the specimens differ—

...series had different ferrite grain sizes. Direct measurement of the mean grain area gave, in the improved specimens, \(256 \ \mu^2\); in the overheated specimens, \(242\,000 \ \mu^2\); in the rolled specimens (as delivered), \(563 \ \mu^2\); in those deformed under blue heat \((200^\circ)\), \(1462 \ \mu^2\); and in those recrystallized, from \(37\,772 \ \mu^2\) (middle of the specimen) to \(63\,532 \ \mu^2\) (edge of the specimen).

The improved specimens gave the maximum value of impact toughness (see Fig. 12). As is evident from the diagram shown in Fig. 12, the finest-crystalline specimens without any treatment (grain size \(563 \ \mu^2\)), and also the improved steel (grain size \(256 \ \mu^2\)), proved to be the toughest at low temperatures \((-50^\circ, -0^\circ)\). Under the same conditions, the most brittle was the iron rolled under blue heat. The improved specimens broke only at \(-70^\circ\) and \(+500^\circ\); at the other temperatures they merely bent.

Fig. 9

In Fig. 13 the left-hand part of the diagram shown in Fig. 12 is given on an enlarged scale; it is evident that at \(-20^\circ\) the improved steel (curve \(A\)) is fourteen times tougher than the overheated steel (curve \(B\)) and tougher than the rolled steel in the as-delivered condition (curve \(C\)). The absolute values of Charpy impact toughness are very greatly reduced at \(-70^\circ\), and at this temperature there is almost no effect on toughness from such different consequences of heat treatment as overheating, grain growth during recrystallization (curve \(E\)), and rolling at the temperature of blue heat (curve \(D\)).

Schematically, the results of the experiments of Kerber and Pomp may, as they themselves do, be represented in the form of three curves (Fig. 14).

Curve \(a\) gives an idea of the change in the impact toughness of improved mild steel as a function of the test temperature—

...; curve b is for steel spoiled by improper heat treatment; curve c is for steel in its original condition (rolled metal).

As will be seen below, curves of this kind can be interpreted in somewhat greater detail by introducing the concept of the critical range of impact brittleness.

Harder steels, approaching rail steels in carbon content, according to experiments by the same authors (16), also show a very considerable decrease in impact toughness with decreasing test temperature (Fig. 15). The type of specimen and the dimensions of the specimens were the same as in work (5). The steels investigated had the following chemical composition (Table 3).

Fig. 10

As is evident from Fig. 15, the rolled steel (as delivered, curve A) has a low impact toughness at temperatures of −70 and −40° in comparison with its toughness at room temperature. The toughness is especially reduced in the case of overheated steel (curve B). Quenching and subsequent tempering to sorbite, while increasing to a considerable degree

Table 3

Steel grade	C	Mn	Si
Steel 1	0.40	0.78	0.17
Steel 2	0.58	0.72	0.18

the impact toughness at room temperature, only insignificantly reduces brittleness at temperatures of the order of −70° (curve C).

Attention is drawn to the strong influence of carbon content on impact toughness. Thus, at \(+20^\circ\), changing the carbon content from 0.58 to 0.40% changes the impact toughness of untreated steel from 3 to 7.5 kgm/cm², i.e., by 2½ times. One must assume some threshold in steel 2, all the more so since heat-treated steel gives approximately the same result.

The strength of steel castings at low temperatures was investigated by Walle (17). For his tests he used steel of two grades, indicated in Table 4.

The steel was cast at \(1600^\circ\). The specimens were annealed at \(900^\circ\) in a furnace for 3½ hours and cooled over a period of 8 hours (Fig. 16). Owing to the inhomogeneity of the cast material, each point on the curve was obtained from 3–6 tests; when significant scatter appeared, additional specimens were taken.

In Fig. 16 one can see the change in the impact toughness of steel castings as a function of temperature. The specimens had dimensions of \(10 \times 10 \times 55\) mm, with a rounded notch of 2 mm. As can be seen, the value of the impact toughness at \(-60^\circ\) reaches only 14% of the toughness value at \(+20^\circ\).

Fig. 11

Table 4

Steel grade	C	Si	Mn	P	S
Steel 1	0.23	0.27	0.59	0.013	0.026
Steel 2	0.20	0.36	0.64	0.013	0.023

Corelli (Corelli) (18) carried out tests on steel containing 0.6% C. He used a Charpy pendulum machine (pendulum weight — 22.710 kg, lift — 1.321 m), applying Mesnager specimens with dimensions \(40 \times 10 \times 10\) mm. The curve constructed on the basis of his data is presented in Fig. 17. Attention is drawn to the insignificant change in the impact toughness of the rather hard steel with which Corelli worked, in the interval \(-40^\circ\) to \(+25^\circ\), barely reaching at \(-40^\circ\) 22% of the value at \(+25^\circ\). Such a phenomenon is not clear.

The question of the brittleness of welded seams at low temperatures is very topical, since the majority of welded structures operate in frost, while the ordinary seam of electrode welding in general does not possess high impact toughness.

Fig. 12

Smirnov and Fal’kevich (19) came to the conclusion that the impact toughness of the seam at \(-183^\circ\) decreases by 88.5% in comparison with the toughness at \(+20^\circ\).

Voznyak (20), studying a seam deposited with an electrode having the following composition: C — 0.1%; Mn — 0.17%; Si — 0.015%; P — 0.009%; S — 0.001%, says that the impact toughness of all welded specimens at \(-50^\circ\) is negligibly small. At \(-25^\circ\) all cast seams possess the same insignificant impact toughness as at \(-50^\circ\); the base metal and seams made by the fusion method give a fairly high toughness at \(-25^\circ\), equal to approximately 50% of the impact toughness at room temperature.

In Fig. 18 a portion of the curves included in Voznyak’s cited work is given.

Curve \(A\) refers to the base metal, whose toughness at \(-25^\circ\) is still sufficiently high, but already at \(-50^\circ\) it falls to an insignificant value—\(0.47\ \mathrm{kgm/cm^2}\).

The greatest toughness at room temperature is shown by the weld metal produced by the Langmuir method (\(14.77\ \mathrm{kgm/cm^2}\); curve \(B\)) and welding with an electrode having a thick LIM coating (curve \(C\)).

The impact toughness of a seam made with a bare electrode (with chalk coating), with direct current (curve $D$) and alternating current (curve $E$), is negligible.

Strokopytov (21) indicates that the impact toughness of an ordinary electric-arc seam at $-55^\circ$ reaches only 20% in comparison with the toughness at $+18^\circ$, and can be increased only to a very small extent by normalization.

In recent years, in connection with the development of chemical apparatus construction in the Union, a number of Soviet works have appeared on the study of impact toughness at low temperatures. In these works the range of investigation has been extended toward low temperatures down to the temperature of liquid air. Thus, Falkevich and Smirnov (22), studying, along with other materials, steel of the grade “steel 5” on a Charpy pendulum of 20 and 30 kgm, found that the impact toughness of this steel at $-183^\circ$ amounts to only 8.4% of the toughness at $+20^\circ$ (1.06 and 12.45 kgm/cm²).

Fig. 13

Experiments by V. D. Kuznetsov and the author generally confirmed (23) the proposition that becomes evident from a review of the cited works: carbon structural steel, possessing normal impact toughness at room temperature, becomes brittle at temperatures only slightly below zero.

For preliminary experiments an open-hearth rail was taken of the following composition: C — 0.42%; Si — 0.20%; Mn — 0.49%; P — 0.05%; S — 0.048%. From the place of the rail head where the de—

where the segregation zone is inconspicuous, longitudinal strips were cut out on a milling machine and used for making specimens tested on a Charpy pendulum with a capacity of 10 kgm (Shopper system). The dimensions of the specimens were \(10 \times 10 \times 100\) mm; the notch was a drilled hole 2 mm in diameter. The net cross section of the specimen was \(5 \times 10\) mm. The tests were carried out at temperatures of \(-183\), \(-80\), \(-40\), \(-30\), \(-20\), \(0\), \(+20\), \(+50\), \(+100\), and \(+150^\circ\). For each point at the indicated temperatures, 3–4 specimens were taken. To maintain the constancy of the temperature of the liquid (ligroin or ethyl alcohol) in which the specimens were cooled, the latter was stirred continuously. The vessel with the liquid was cooled by means of liquid air.

Fig. 14

The results are presented in Fig. 19. Here, as usual, the test temperatures are plotted along the abscissa axis, and along the ordinate axis—the specific work of fracture (impact toughness); thus from 4.5 kgm/cm² at \(150^\circ\) the latter decreases to 0.3 kgm/cm² at \(-40^\circ\) and then remains practically constant down to \(-183^\circ\). For the rail steel tested, the temperature \(-40^\circ\) is the temperature at which brittleness reaches its greatest value.

The macrostructure of the fracture at low temperatures (\(-40^\circ\), \(-183^\circ\)) differs sharply from the appearance of the fracture at \(+125^\circ\). The fracture at low temperatures is crystalline, shiny, and has no signs of residual deformation. The fracture at elevated temperature is fibrous, with sharply expressed plastic changes in shape. Owing to the presence of the latter, it is difficult to compare with each other

it is impossible to [fit] the half of the specimen broken at a high temperature,—large gaps always remain between them. The parts of a brittle specimen fit one another without any clearances; after the fragments have been put together, their axes serve as a continuation of one another.

If all the specimens that served for constructing the curve shown (Fig. 19) are arranged in a row, it will turn out that the specimens to the left of the point \(-40^\circ\), all as one, gave a uniform crystalline fracture, without the slightest signs of residual deformation.

Fig. 15

If now, after making this—admittedly far from complete—survey of what we know about the brittleness of carbon steel at low temperatures, we ask the question: on what does the increase in the number of broken rails in the winter months depend, then the answer, obviously, suggests itself. In winter the rails are considerably more brittle than in summer, rail defects show themselves more strongly, and the track has a different elasticity. In his critique of Larsen’s (24) work, Meyer (25) speaks of the need to take into account the fact that the resistance of very hard rails to impact loads is, at \(-20^\circ\), about one third, and at \(-40^\circ\) barely only one fifth, of the resistance at \(+20^\circ\).

Considering the graphs of the dependence of the impact toughness of steel on temperature, we always note the presence of a certain temperature interval near \(0^\circ\) in which the specimens become brittle. This interval is always observed, irrespective of the size of the specimen, the shape of the notch, and the magnitude of the striker. But it is sharply expressed only in some iron alloys. In many alloys (chiefly

…chiefly with a face-centered lattice) it is not manifested. Thus, when the impact-test temperature is changed within the range from \(+20\) to \(-190^\circ\), Lautal reduces its impact toughness by no more than 15–20% (26).

Tin’gaev (27) showed the independence from temperature of the impact toughness of raw, quenched, and aged duralumin.

Fig. 16

If, for the time being, we set aside the question of replacing ordinary carbon steel in parts suffering from cold brittleness by alloyed steel, then there remains only one path—to try to shift the dangerous range of impact brittleness into a region of temperatures not encountered during the operation of the given part. In most cases such a region will be temperatures below \(-60^\circ\). And only in special cases (compression of gases) will it have to be shifted farther—to absolute zero.

The curves showing the relation between air temperature and the number of rails removed owing to fracture (see Figs. 1 and 2) confirm that the operating conditions of rails in winter accidentally coincide with the operating conditions of a small specimen on the Sharpy pendulum (our experiments).

We say “accidentally” because the critical temperature \(-40^\circ\) is not something entirely definite, independent of the testing method. Under impact, bodies that are usually plastic sometimes become brittle; conversely, brittle bodies—for example, rock salt—under a slowly increasing load may show signs of flow, of plasticity.

As was indicated, the ductile fracture of steel passes into brittle fracture in a definite temperature interval, sometimes a very narrow one. According to Mailänder (28), the temperature of the transition region for impact bending is higher than for static bending; for a sharp notch it is higher than for a rounded one; for a wide specimen it is higher than for a narrow one, and so on. In short, the absolute value of the impact toughness of a specimen cannot be separated from its shape and size.

Rybinsky (29), determining impact toughness by Charpy, comes to the conclusion “that this latter type of test is very indicative in testing brittle rails, for in all cases we obtained, in carrying out this work, either very low average figures for the specific work of impact, or very fluctuating figures, varying for one and the same rail over individual specimens within very wide limits.” This once again confirms the proposition that the determination of impact toughness by Charpy is a very sensitive method for detecting brittleness of material in parts subjected, during service, to impacts at low temperatures.

“Fluctuating figures” were obtained by Rybinsky evidently when the beginning of the cold-brittleness interval for the given rail steel lay close to the test temperature. If he had attempted to determine the impact toughness in this case at $\pm 0^\circ$ or below, he would obviously have obtained a minimum value of the impact work and, probably, a brittle fracture.

Fig. 17

On the basis of what has been set forth, it seems to us possible to assert that, in passing from a specimen for the Charpy impact test to an entire rail, the character of the dependence “impact toughness—temperature” remains unchanged; only the interval in which ductile fracture passes into brittle fracture is shifted in one direction or another, to the right or to the left along the temperature axis.

If a steel specimen for impact testing of a certain size exhibits cold brittleness at some temperature $t_1^\circ$, then any product made of this steel will also, at some temperature $t_2^\circ$, become entirely brittle; but whether $t_2^\circ$ will be above or below $t_1^\circ$, and by how much, is difficult to say. For rail steel such a temperature ($t_1^\circ$) in some cases lies at about $-40^\circ$.

Before proceeding to the next part of our work, we posed the question: does fracture of a rail in track at low temperatures occur from a single blow, as in the Charpy impact test, or

is the fracture of the rail a prolonged process, akin to fatigue?

The literature sources (for example, N. N. Davidenkov (1)) indicate that the relation between the impact toughness determined by fracturing notched specimens with a single blow and with several hundred (250–500) blows is very simple. As the toughness determined with a single blow increases, the toughness found in testing with many blows also increases.

Fig. 18.

Prishchepa (30) fractured Mesnager-type specimens on a Krupp machine and found, up to 5350 blows, that with an increase in the Charpy impact toughness the number of blows required to fracture the specimen on the Krupp machine also increases.

But with an increase in the number of blows such a simple dependence begins to break down and, at \(10^5\) blows, sometimes becomes directly opposite: with an increase in the Charpy toughness, the toughness found at a very large number of blows decreases.

The impact fatigue limit, of which one has to speak in this case, varies parallel to the static fatigue limit and is in no way connected with the impact toughness of notched specimens.

The fracture surface may be very different, depending on the number of blows. At 300 blows the entire fracture of a steel specimen, for example, has a crystalline structure, but already at 1740 blows there appear regions having the fibrous character typical of fatigue fracture. As the number of blows increases, these fibrous regions begin to occupy the entire area of the cross section of the specimen.

A rail operates under extremely complex conditions and experiences very considerable stresses. When a train passes, impacts occur at the joints; the force bending the rail increases almost instantaneously, since the consist moves at a speed of 20–25 m per second. Owing to the play of the springs, the excessive mass of the counterweights, the inertial forces of the moving parts, etc., additional vertical forces appear.

These dynamic phenomena increase with the speed of motion and with the increase in the weight of train consists, in particular locomotives.

The condition of the track superstructure also has an enormous influence on the stresses in the rails.

Moreover, it should not be forgotten that when the head is crushed, surface cracks may arise. Severe skidding during braking causes the release of a large amount of heat over a short interval of time, which leads to local heating to such a temperature that quenching of some surface areas to martensite is possible (31). Under the influence of friction, the running surface absorbs nitrogen from the air (32) and becomes abnormally brittle.

Fig. 19

Cracks, hardened and abnormally brittle places have the significance of a notch on an ordinary specimen for impact testing.

At the same time, very considerable stresses arise on the surface of the rail head, reaching up to 40 kg/mm², while the usual permissible stresses in rails are generally taken as 30 kg/mm². At the same time, the yield point of ordinary rail steel rarely exceeds 40 kg/mm².

As we have seen, at temperatures below zero ordinary steel possesses very considerable brittleness, but the quantities characterizing static strength change only very slightly.

Thus, in the case of steel casting (17), the elastic limit of steel 1 (its composition is given in the description of Vallé’s experiments) was, at −60°, 19.6% higher than at +20°; the ultimate strength was higher by 16.7%, and the relative elongation at rupture by 11%. Only the impact toughness decreased by 86%.

According to Falkevich and Smirnov (19), for steel grade “steel 5” (C—0.32%) the ultimate strength at −20° is 59.1 kg/mm², and at −183°—93.9 kg/mm²; the relative elongation at the same temperatures is 27.9 and 25.9%, and the impact toughness is, respectively, 12.45 and 1.06 kgm/cm².

According to Wentrup (34), with a lowering of the testing temperature the fatigue limit of rail steel increases. Thus, at +18° the fatigue limit for steel of composition: C—0.62% and Mn—0.93% is 32 kg/mm², while in the interval from −14 to −18° it rises to 35 kg/mm². The fatigue limit of axle steel (0.41% C) at +18° is 27 kg/mm², and at −18°—31 kg/mm².

Kerber and Hempel (35) assert that plastic deformation and aging practically do not change the fatigue limit and have a very strong effect on the impact toughness according to Charpy (steel composition: C—0.39%; Mn—0.72%). The impact toughness decreases tens of times, while the fatigue limit remains within the usual limits, changing by only a few percent.

Taking into account the influence of temperature on these mechanical characteristics, we arrive at the conclusion that in frost such a difference will manifest itself still more strongly.

Since the static strength and the limit of static fatigue of carbon steel at low temperatures are higher than at normal temperature, it can hardly be assumed that any changes characteristic of fatigue specifically at low temperatures will appear in the rail metal. It is also unlikely that cracks will form in a rail as a result of a small number of impacts at an elevated temperature; cracks of this kind will more likely arise at low temperatures, when the impact strength (determined from one or several impacts) is several times lower than at temperatures above zero.

Of course, fractures also occur as a result of pure fatigue, but, as Skakov and Makhanek (31) state, “fatigue fracture does not often occur on our railways.”

Having investigated more than a dozen rails that had broken in service in mo—

for days on end, we never once found signs of a fatigue fracture. On the transverse fracture of such rails, even with the most careful examination, it was impossible to detect the presence of typical changes in structure; it was always a crystalline, homogeneous fracture.

If a fatigue crack arises in a rail in summertime, then such a rail must break on the first frosty day—not as a result of the slow development of the crack, not because the material will gradually become fatigued, but will break at once, from a single impact of sufficient force, owing to the considerably reduced impact toughness at low temperatures.

On the basis of the considerations set forth, we settled on determining the impact toughness by the Charpy method as a method entirely suitable for determining the cold brittleness of steel.

Everything indicates that this method of testing metal gives precisely that characteristic of the material which determines its suitability for the manufacture of parts operating at low temperatures and, moreover, sustaining specifically impact loading.

II. REDUCING THE COLD BRITTLENESS OF CARBON STEEL BY HEAT TREATMENT

The impact resistance of steel parts can be improved in several ways.

The simplest is to increase the dimensions of the part. But, obviously, this method can never find wide application, since such a solution to the problem would entail the aimless expenditure of scarce metal. In the USA there are railways with very heavy rails (up to 74 kg per running meter). Of course, the manufacture of such rails is of little expediency, chiefly because, by lowering the basic stresses in the rail through an increase in cross section, we are not in a position to reduce the local stresses from crushing of the running surface. Statically, such a rail is much stronger than our ordinary one (weighing 45.3 kg per running meter); its moment of inertia is several times greater. But the contradictory data of American practice indicate that not everything is clear in the matter of heavy rails. Most likely, the question of super-heavy rails arose as a result of the peculiarities of the capitalist economy in the pre-crisis period. For large quantities of cheapening metal it was necessary to find a market, and one was found by making the rails heavier and proving the expediency of this with technical considerations of doubtful quality and with the original assertion that steel would become more expensive after the crisis and that every dollar invested in track would, in 2–3 years, turn into three dollars.

The composition of ordinary steel can be changed by introducing, for example, nickel as an alloying addition. As we shall see below, nickel steels possess very low sensitivity to impact at low temperatures. But again, not every part, for reasons of...

economically may be made of expensive and often scarce materials, although in some cases the manufacture of special apparatus from highly alloyed steels is the only solution.

In particular, the question of alloyed rails is also a mature and very urgent one. But it is not yet possible to speak of the wide distribution among us of rails alloyed with the element that increases impact toughness in frost (chiefly nickel).

By changing the ordinary composition of carbon structural steel, in particular rail steel, it is possible to force the dangerous interval of impact brittleness to shift on the “temperature—impact toughness” diagram toward lower temperatures. But such a change in composition will consist primarily in reducing the carbon content, which is, of course, inadmissible, since it will lead to a decrease in strength and hardness. The latter circumstance is of paramount importance for rails.

In studying the literature it could be noted that the purity of steel affects cold brittleness: the purer the steel, the tougher it is. But even the improvement of steel has, at the present day, some quite real limit. Moreover, according to the experience of the Swiss railways, the best rails (not heat-treated) made of carbon steel prove unsuitable under an axle load of 23 t.

The influence of phosphorus on the cold brittleness of steel at low temperatures is somewhat unclear in the case when its content does not exceed 0.1%. Rybinskii (29) emphasizes that in 32 tested brittle rails the average phosphorus content was 0.093%, reaching a maximum of 0.135%, and considers the principal cause of the brittleness of these rails to be an abnormally large amount of phosphorus. It must be stressed that the German specifications for rails allow a phosphorus content from 0.06 (open-hearth metal) to 0.08% (Thomas metal), while the American specifications allow 0.04% for all grades of steel.

Fratkin (36) considers an elevated phosphorus content to be one of the causes of rail brittleness.

Vladimirov (37) arrives at the same conclusion.

Baboshin (38), in his report, says: “From the data of the Rail Commission of 1899–1906 it was concluded that if the phosphorus content in rails is more than 0.1%, the carbon content in them should be no more than 0.3%; with a higher carbon content, the ductility of phosphoric metal noticeably decreases (elongation drops below 6%). Moreover, phosphoric metal is very sensitive to impact in the cold (frost).”

To bring greater clarity to the question, we selected an open-hearth steel of approximately the same composition in the content of all elements, except phosphorus, as the steel used for making those specimens on the basis of whose investigation ...

our “impact toughness—temperature” curve was constructed (see Fig. 19). In Table 5, for comparison, we give the composition of both steels.

Table 5

Grade of steel	C	Si	Mn	S	P	Impact toughness at: +0°	Impact toughness at: −60°
Steel 1	0.42	0.20	0.49	0.048	0.051	1.0	0.4
Steel 2	0.43	0.23	0.43	0.040	0.099	2.5	0.5

As in the first series of experiments, each point was obtained by testing four specimens. We are far from ignoring the harmful influence of phosphorus on the properties of steel. But there is also no basis for fearing it excessively when it is present in amounts within the limits permitted by the standard. We think that the presence of phosphorus above 0.1% is especially undesirable and dangerous for increasing brittleness. Of course, in saying this we do not touch upon other harmful aspects of the influence of phosphorus (segregation, etc.).

Even if it were found that the brittleness of rails depends chiefly on the phosphorus content, it would still be difficult to proceed along the path of substantially reducing its content. For this it would be necessary to abandon the existing methods of steel melting, since the phosphorus content in rail steel permitted by the technical specifications is, evidently, its average content in normal, sound metal.

If a radical change in the chemical composition of rail steel is for the time being impossible for the reasons given, then the only remaining path is to change the structure of the steel by heat treatment. A whole series of the data already presented shows how a reduction in grain size or the formation of a sorbitic structure shifts the dangerous interval of transition from ductile fracture to brittle fracture toward lower temperatures.

Whereas the heat treatment of machine parts made of carbon steel has long been a matter of common knowledge, the hardening of rails has to this day remained something disputed, not generally accepted. Hardening and tempering specimens of rail steel and then testing them on the impact machine, as we did, made sense only in anticipation of a subsequent transition to the hardening of rails.

Until now we have made specimens for the small Charpy impact machine (10 kg·m), cutting them from the head of the rail in places free from segregation. But, obviously, with such a method of preparing specimens we depart very far from the conditions in which the rail exists as a complex profile having a segregated zone. The material of the rail has a chemical composition that is not constant over the cross-section, a variable structure, and diverse mechanical properties.

Therefore, in what follows we changed the method of preparing specimens as follows: from a piece of rail, a section was cut off on a milling—

the head was turned on a lathe, and then it was faced off to remove all traces of the wall. The resulting section was forged under a pneumatic hammer (1 t) at a temperature of 950–1000° into a square of \(13 \times 13\) mm. The heating temperature of the head and the temperature at the beginning and end of forging were carefully monitored with a “Pyropto” optical pyrometer. The forged bars were cut into pieces about 1.1 m long. From each bar an end section was taken, and shavings were removed from the end for analysis. The bars were then cut into specimens 105 mm long, which by milling or planing were brought to a size of \(11.5 \times 11.5 \times 100\) mm. The specimens were then heated in a muffle electric furnace, with an automatic thermoregulator and a self-recording pyrometer, to 850°, held at this temperature for 1 hour, and cooled in air. Such normalization destroyed, to some extent, the internal stresses arising during forging and gave a more or less homogeneous grain. The temperature of 850° was taken as the minimum temperature at the end of rolling. It was above the \(Ac_3\) point for this steel.

To protect the specimens from decarburization at high temperature, during all operations involving heating they were placed in iron boxes (10–12 pieces per box) and covered with cast-iron filings with a small amount of charcoal powder. The percentage of carbon in the mixture was selected experimentally so that carburization of the specimens would not proceed during heating. In addition to the pyrometer in the muffle furnace, a thermocouple was inserted into the box for more accurate temperature control.

After the appropriate heat treatment (for example, sorbitizing), the excess 1.5 mm along the cross section of the specimen was removed on a lathe, and a notch with an angle of 45° and a rounding radius of 0.25 mm was made in the specimen. For this operation a carefully profiled milling cutter was used.

We changed from a specimen with a drilled hole to a specimen with a milled notch because of the greater simplicity of manufacture and control.

As a rule, the notch did not reach the segregation zone, which was carefully checked.

We paid particular attention to ensuring that the cross section of the specimen should at least remotely resemble, in the distribution of segregation zones, the cross section of a normal rail, bearing in mind the interest that the question of the significance of segregation in a rail has always aroused. Without going into a detailed study of the question, we shall cite the concluding words of Zhukovskii (39), with which we are, in the main, in full agreement: “A rail works in service as a beam, and, as such, the rail must be tested upon acceptance with all the harmful and harmless inclusions contained in it; therefore, testing the rail for dynamic (and, if desired, static) bending, with determination of the deflection, together with the still-practiced abrasion test—these are the proper methods—”

...rail tests.” Testing not the rails themselves, but specimens cut from them, would have yielded, in Zhukovsky’s opinion, conclusions just as false as a conclusion about the mechanical properties of granite based on investigation of mica crumbling between the fingers, which is included in its composition along with other structural constituents.

The specimens were broken on a Charpy pendulum impact tester, constructed in the metallographic laboratory of the Siberian Institute of Metals, with a capacity of 10 kgm. In comparison with a new Charpy impact tester of the Schopper firm of the same size, it showed the same accuracy as the latter. Owing to the considerable distance between the uprights, it was more convenient to work with. The distance between the supports was 70 mm, as in the Schopper pendulum.

Comparative tests of specimens made, as described (forged), normalized, and of specimens cut from the head and base of the rail and likewise normalized, did not show any great difference in the value of impact toughness. But, as a rule, the forged and normalized specimens proved 8–10 percent tougher under all test conditions.

Fig. 20

The first specimens used for constructing the curve in Fig. 19 were frozen in a test tube filled with toluene or alcohol and cooled in a Dewar vessel with liquid air. Although this method gave good results, it was troublesome when one had to deal with a large number of...

specimens. Therefore we developed the following method of freezing the specimens at temperatures down to −60°.

In order to maintain low temperatures for a long time, a refrigerator was constructed, consisting of a series of boxes fitting one inside another. The outer box was made of wood (Fig. 20), the inner boxes (B and D)—of zinc, with double walls. The space between the walls was filled with cotton wool to reduce convection. The specimens were placed in the inner box (A) and covered with absolute alcohol. The wooden box was filled with ice or with a mixture of ice and common salt; solid carbon dioxide, calcium chloride with ice, or another cooling mixture was placed in boxes B and D.

For measuring the temperature in the inner box (A) we used three copper-constantan thermocouples, connected in series and enclosed in a common sheath. The zero junctions of all the thermocouples were in melting ice. The electromotive force of this triple thermocouple was approximately 10.35 mV per 100°, which made it possible to take readings on the galvanometer with an accuracy of ±0.1. The thermocouple was calibrated by constant points: the melting of ice, mercury, and eutectic mixtures of KCl and NaCl with water. The calibration curve is given in Fig. 21. For preliminary measurement of temperatures an ordinary alcohol or toluene thermometer was used.

Fig. 21

We paid serious attention to the rate of warming of the cooled specimens during testing on the pendulum impact machine and attempted to determine the temperature gradient then established from the surface of the specimen, which warms first, to its middle. For this purpose two ordinary Charpy specimens (10 × 10 × 100 mm) were made; in one of them a hole 2.5 mm in diameter was drilled along the long axis to the middle of the specimen. A thermocouple was inserted into the hole and sealed with wax. To the surface of another identical specimen was carefully soldered ...

the same thermocouple, but with a flat junction. These thermocouples were calibrated by comparison with a standard.

Specimens with thermocouples, into whose circuit the self-recording galvanometers of Acad. Kurnakov’s system were connected, were cooled in the thermostat to the corresponding temperature, removed with pre-cooled tongs, and placed on the supports of the pendulum machine, which were provided with wooden stops to facilitate the correct positioning of the specimen. With some practice this operation required from 2.5 to 4.0 sec.

During this interval of time the specimens warmed at different rates, depending on the initial temperature and on the medium in which they were kept in the thermostat. Kerosene gave good results in the sense of long preservation of a constant temperature, since, thickening at temperatures close to \(-60^\circ\), it enveloped the specimen with a thin layer and thereby protected it from heating. But, splashing apart on impact, the kerosene contaminated both the instruments and the laboratory. Therefore we settled on absolute alcohol, which does not appreciably change its coefficient of internal friction down to \(-80^\circ\). The curves recorded under these conditions by Acad. Kurnakov’s apparatus (thermocouple on the surface and inside the specimen) show the constancy of the temperature inside the specimen for 4–5 sec. (Fig. 22, curve \(A\)); the end of the interruption coincides with the moment at which the specimen is removed from the thermostat. The surface of the specimen begins to warm more rapidly (curve \(B\)), but during the maximum interval of time necessary for fracturing the specimen, the temperature rises (at a test temperature of \(-60^\circ\)) by only \(2^\circ,4\).

Fig. 22

We have dwelt in some detail on the freezing of specimens because, although the value \(2^\circ,4\) may be neglected, taking into account the not very high accuracy of the pendulum test in general, at \(-183^\circ\) the heating in 4 sec. reaches \(10^\circ\).

The first question, without resolving which it is impossible to proceed to the heat treatment of rail steel, concerns the choice of quenching temperature.

In hardening most parts, such a question raises no doubts: one must harden at the temperature which ensures complete hardening for the given steel and, after tempering, gives the optimal set of required properties (strength, toughness, etc.).

The problem is not solved so simply when one turns to the rail; the rail must be hardened from the temperature at which it leaves the last pass of the mill, or, more precisely, the hot straightening. Any other temperature would require special furnaces for heating, uniform cooling, or reheating of the rails subject to hardening. Obviously, the introduction of such furnaces into the technological process of rail manufacture will complicate and increase the cost of rail production. The problem of heat treatment of rails cannot be solved without taking the above circumstances into account—otherwise the approach obtained will be too academic, not engineering.

Preliminary experiments were carried out with the two grades of steel indicated earlier.

As our observations showed, a rail having a temperature after hot straightening of 950° retains, on cooling to 850°, sufficiently uniform heating along its length and cross section, although the ends of the base already begin to cool appreciably. Therefore the preliminary heat-treatment experiments were carried out specifically with hardening from 850°.

As the material for the study we took steel whose composition is given in Table 6.

Table 6

Steel grade	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)
Steel grade	C	Mn	Si	P	S
Steel 2 . . . . .	0.43	0.43	0.23	0.099	0.040
Steel 3 . . . . .	0.40	0.60	0.15	0.054	0.030

Figure 23 gives the results of Charpy impact tests of heat-treated specimens made of steel 2. The investigation was carried out at temperatures of −183, −78, −40, −20, ±0°; three specimens were taken for each point. Curve A refers to steel without heat treatment; the impact toughness at 0° is 2.5 kgm/cm². It falls rapidly to 0.6 kgm/cm² at −40°. The change in toughness of specimens quenched in oil and tempered at 640° (tempering duration—2 hours) as a function of temperature is given by curve B. After tempering, the specimens were cooled in air. Curve C refers to specimens also heat-treated, but with a different holding time during tempering—4 hours. At 0° the impact toughness is 5.4 kgm/cm²; at −60° it is about 1 kgm/cm². As can be seen from Fig. 23, by means of heat treatment it is possible to increase the impact toughness of steel at −40°, in comparison with the impact toughness of untreated steel also at −40°, by six to seven times, and at −60° by a factor of two.

Fig. 23

Figure 24 also gives curves for steel III. As can be seen, the raw steel (curve A) gives at \(0^\circ\) an impact toughness of about \(2.3\ \mathrm{kgm/cm^2}\), and at \(-40^\circ\) only \(0.6\ \mathrm{kgm/cm^2}\). There is practically no difference from steel 2 (Fig. 23). But the heat-treated specimens considerably increased the impact toughness at low temperatures. As in the preceding case, for each point \((-60,\ -40,\ -20,\ \pm 0^\circ)\) 3–4 specimens were taken.

The specimens quenched at \(850^\circ\) and tempered for 4 hours (curve B) showed at \(-40^\circ\) an impact toughness of about \(6\ \mathrm{kgm/cm^2}\), and at \(-60^\circ\), \(2\ \mathrm{kgm/cm^2}\).

Curve C belongs to steel treated in the same way, but with a tempering duration of 2 hours.

Comparing Figs. 23 and 24, one may perhaps conclude that, if within limits up to \(0.1\%\) phosphorus does not exert any particular influence on the cold brittleness of steel, then after quenching and tempering its effect nevertheless becomes apparent. But in our subsequent work we were still unable to detect a clearly expressed influence of phosphorus in this direction. Toughness, as we shall see later, is very strongly affected by slag inclusions, the abnormality of the metal, and many other factors, which in many cases mask the direct action of phosphorus.

Even these preliminary experiments on quenching and tempering (“through sorbitization”) of rail steel showed that the impact toughness at winter temperatures of heat-treated steel is higher than that of untreated steel, by several (3–10) times. The heat treatment was applied by Strokopytov (21) to increase the impact toughness of electric-arc welds. Normalizing at 900° gave good results. The specimen was held at 900° for 10 minutes and cooled in air.

Fig. 24

Table 7 gives the dependence of impact toughness on the test temperature.

As can be seen from Table 7, the impact toughness increased very substantially at ordinary temperature; beginning from 0° this increase in toughness is not very large. The increase in impact toughness in this case occurs owing to a reduction in the size of the weld grain, and mainly as a result of coagulation of the nitride needles, which acquire a spherical form after such heat treatment.

As we have already noted, rail rolling is finished at the most varied temperatures—from 850 to 1150°. In the literature (40) there are indications that an increase in the quenching temperature, if followed by high tempering to sorbite, does not affect the impact toughness of steel. From the experiments it followed that, when the quenching temperature of carbon steel (C—0.52%) was raised from 845 to 1040°, the impact resistance increased rather than decreased. This increase was the more intense, the higher the temperature of tempering to sorbite. Thus, for example, the impact toughness of a specimen quenched at 845° and tempered at 450° was 8.3 kgm/cm², whereas that of a specimen quenched at 1040° and tempered at the same temperature, 450°, was 10.8 kgm/cm². A rather considerable increase in toughness is observed (by 54%). Specimens quenched at the same temperatures, but tempered

Table 7

Test temperature (in °C)	Impact toughness (in kgm/cm²), as-welded seam	Impact toughness (in kgm/cm²), normalized seam
+18	1.00	2.47
0	0.50	0.57
−20	0.31	0.41
−40	0.19	0.27
−55	0.19	0.23
−183	0.16	0.22

to a higher degree (675°), gave impact toughnesses of 13.4 and 25.4 kgm/cm²; the increase, as we see, is 90%. The conclusion, at first glance paradoxical, is that the more strongly the steel is overheated during heat treatment to sorbite, the better it resists impact.

As Efremov and Gorskii (41) showed, overheating during quenching is harmless only if the subsequent tempering is no lower than 400–550°. The higher the quenching temperature, the stronger the tempering must also be if it is desired to obtain a part with high impact resistance. The authors dealt with a steel containing 0.32% C and 0.60% Mn.

Foley showed that the impact toughness of steel (C—0.52%), quenched at high temperatures (950–1040°) and tempered to sorbite, is considerably higher than that of steel quenched at the normal temperature for this steel. In Fig. 25 the toughness of steel quenched from a temperature of 845° and tempered at 450° is conventionally taken as unity. Obviously, the toughness of steel quenched at the same temperature and tempered at 675° will be higher (1.7 kgm/cm²). With an increase in the quenching temperature, the toughness of steel tempered at 450° rises only slightly, reaching a maximum at a quenching temperature of 980°. The toughness of steel tempered at 675° increases very strongly with increasing quenching temperature and again reaches its maximum value at 1000°.

To determine the influence of quenching temperature during sorbitization on the impact strength of rail steel, an unused open-hearth rail of the following composition was taken: C—0.43%; Mn—1.04%; Si—0.24%; S—0.014%; P—0.079% (steel 4). The steel was selected with a somewhat elevated manganese content, since at the time the experiments were carried out the rolling shop of the Kuznetsk Works named after Stalin was being put into operation, and the shop was to roll precisely medium-manganese rails.

The structure of this steel before forging, after forging, and after normalization (Fig. 26, magnification 500, etched with nitric acid)

Fig. 25

presented no special features. It was a fairly clean, fine-grained steel, without a large amount of slag inclusions. Forging and normalizing were carried out as already described. In order to obtain a more complete picture of the influence of the quenching temperature on the properties of this steel, the following quenching temperatures were selected: 900, 950, 1000, and 1050°.

Fig. 26

We had every reason to suppose that the impact toughness of sorbitic rail steel at low temperatures is also affected by the rate of cooling after tempering. According to the data of Minkevich (40), “carbon steel with a manganese content of less than 1.1% and nickel steel are indifferent to the rate of cool-

after tempering, i.e., they do not acquire brittleness on tempering with slow cooling, whereas steel with a high content of manganese and phosphorus exhibits this property.”

Bennek (Bennek) (42), investigating various steels, came to the conclusion that in special medium-manganese steels additions of phosphorus within the limits of 0.03–0.10% cause brittleness on tempering. In Fig. 27 the impact toughness of two steels is given (steel 2: Mn—1.43%; P—0.018%; C—0.28%; steel 6: Mn—1.69%; P—0.12%; C—0.27%). The specimens (their size and shape are shown in Fig. 28) were quenched at 820° in oil and tempered at 650° (curve A). The course of the curves shows the clear influence of phosphorus on the toughness of similar steels and the greatly increased brittleness of the steel containing a high percentage of phosphorus when cooled after tempering with the furnace (curve B).

Fig. 27

In Fig. 28 are presented curves of the dependence of impact toughness on temperature for steels with approximately the same, comparatively moderate manganese content and different phosphorus contents (steel 1: Mn—0.50%; P—0.012%; C—0.03%; steel 5: Mn—0.61%, P—0.11%, C—0.28%). A considerable phosphorus content generally lowered the impact toughness of the steel, but did not affect the increase in brittleness under slow cooling. The preliminary treatment of the specimens was the same as in the preceding case, i.e., some specimens were cooled after tempering in oil (curves A), while others were cooled with the furnace (curve B).

In modern rail steel the manganese content reaches up to 1%. Extensive investigations carried out by various authors in this direction have not, as far as we know, touched upon the behavior of steel quenched and tempered at various temperatures, with different cooling rates after tempering at low temperatures. All toughness tests were carried out at room temperature. The data cited by Bennek concern a special case.

To investigate the effect of the cooling rate after tempering and of the duration of tempering on brittleness, we, after hardening a series of specimens at 900°, tempered them at 660° for half an hour, 2 and 4 hours, and cooled them with the furnace. The number of specimens was chosen so that each point on the curve would correspond to 3–4 specimens. Cooling with the furnace from 660° to 400° lasted \(3 \tfrac{1}{2}\) hours, i.e. at a rate of about 1°.3 per minute, and from 400 to 200°—about 4 hours, i.e. 0°.8 per minute.

Fig. 28

The duration of tempering in the case of cooling together with the furnace (43) had a very considerable effect on the toughness of the steel. Thus, steel tempered for 4 hours showed, at \(+15^\circ\), an impact toughness of about \(10\ \text{kgm}/\text{cm}^2\), while that tempered for 2 hours and \(1/2\) hour showed, respectively, 4 and \(2\ \text{kgm}/\text{cm}^2\). This difference decreases somewhat at \(-40^\circ\), when the toughness values are: 2.0; 1.8 and \(1.0\ \text{kgm}/\text{cm}^2\), and becomes zero at \(-60^\circ\), when all specimens cooled after tempering with the furnace, regardless of the duration of tempering, have the same toughness—\(0.8\ \text{kgm}/\text{cm}^2\).

Very interesting results were obtained when the cooling rate after tempering was changed. In Fig. 29 are given the results of experiments carried out under the following conditions: hardening temperature—900°, tempering—640°, tempering duration—4 hours. As can be seen, cooling after tempering in oil (mineral oil, of the light cylinder-oil type, at a temperature of \(+20^\circ\)), in air, and in water gave almost identical results. This is quite understandable, since for small specimens the cooling rates produced by water, oil, and air are higher than that critical rate at which brittleness during tempering manifests itself. In all three cases the toughness at \(-60^\circ\) was (in comparison with normalized steel) quite satisfactory—\(2.1\text{–}3.2\ \text{kgm}/\text{cm}^2\). The specimen that cooled slowly with the furnace proved to be fairly tough at room temperature and only slightly more brittle than the one cooled in oil (11.2 and \(12.3\ \text{kgm}/\text{cm}^2\)), but at \(-60^\circ\) it became three times more brittle than

rapidly cooled. In Fig. 29, for comparison, the toughness values of a normalized specimen of steel 4 are given.

The conclusion is obvious: the cooling rate after tempering of steels with a moderate manganese content (apparently, up to 1%) does not affect their impact toughness at ordinary testing temperatures, but it very strongly affects the toughness at temperatures below zero.

In Fig. 30 is shown the effect of the quenching temperature on the impact toughness of the same steel at the former tempering temperature (640°) and the same duration (4 hours). Cooling after quenching was carried out in air, since the preceding data (see Fig. 29) had already made it possible to draw the corresponding conclusion about the effect of the cooling rate after tempering on impact toughness at low temperatures.

Fig. 29

As can be seen, the quenching temperature does not affect the toughness of sorbitic steel in the range of testing temperatures from −60° to +20°. It is somewhat unclear why the curve of the dependence of impact toughness on temperature for quenching at 1000° went below the corresponding curve for 1050°. This can hardly be explained by inaccuracy of the experiments.

In any case, it may be asserted that at the temperatures most often observed in the Siberian region (−40°), the toughness of sorbitic rail steel quenched at various temperatures (from 850 to 1050°) is more than ten times higher than the toughness of raw steel. It is quite possible to make steel parts tougher at winter temperatures. Sorbitic steel is frost-resistant, but sorbitization must be carried out with due regard for a whole series of conditions that are comparatively unimportant for parts operating at temperatures above zero.

At the same time, we also determined the temporary resistance, relative elongation, and reduction of cross section for steel 4, tempered for various times and cooled after tempering at different rates. From Fig. 31 it is evident that increasing the duration of tempering makes the steel

less strong, but statically more ductile. Different cooling rates have only a slight effect on the static strength of the steel at ordinary temperature (curves A—cooling after tempering together with the furnace; curves B—cooling in air).

Fig. 30

Table 8 gives two series of tests. As can be seen, the scatter of the experimental data obtained lies within permissible limits. The significant notches at \(-60^\circ\) fully characterize the essence of the phenomenon, since at this temperature the critical interval of transition from ductile dynamic fracture to brittle fracture begins. It is characteristic that at \(-40^\circ\) brittle fracture is not obtained at all. Fig. 32 gives a photograph of specimen No. 115. As can be seen, its toughness (at \(-40^\circ\)) proved to be very high; the specimen did not break on the pendulum and was then additionally bent (almost to \(180^\circ\)) in a vise.

In studying microphotographs of sorbitic steels, no clearly expressed connection between structure and toughness was found.

In Figs. 33 and 34 are shown, at a magnification of 1000 times, polished sections of specimens Nos. 113 and 115. The sections were etched for 50 min. in a sodium picrate solution heated to \(100^\circ\). Specimen No. 115 is certainly more ductile than specimen No. 113. The former (No. 115) did not break at \(-40^\circ\), i.e., its toughness is in any case equal to the full work of the pendulum, referred to the corresponding cross section (\(15\ \mathrm{kgm}\)); the latter (No. 113) broke at \(-20^\circ\), although it still showed sufficient toughness. No defects that could have reduced the toughness of this specimen were found, even under the most careful examination. The material throughout the entire section is sufficiently good; there are no slag inclusions, cracks, etc.; the structure is quite homogeneous. True, at a magnification of 1000 times, on the polished section of specimen No. 113 the sorbite retained the structure of martensite, and it is as though possible even to distinguish two differently colored phases, whereas the material of specimen No. 115 is more homogeneous. But in other microphotographs the reverse pictures are observed, and we refrain from drawing a definite conclusion.

Table 8

Specimen No.	Hardening and tempering temperature (in °C)	Test temperature (in °C)	Impact toughness by Charpy (in kg·m/cm²)	Temporary tensile strength (in kg/mm²)	Relative elongation (in %)	Reduction of area (in %)	Notes
82	950	+20	11,18	—	—	—
83	640	+20	11,29	—	—	—
84	640	−20	11,61	—	—	—
85	640	−20	>15,00	—	—	—	Did not break
86	640	−20	14,04	—	—	—
87	640	−40	4,20	—	—	—
88	640	−40	5,45	—	—	—
89	640	−60	2,67	—	—	—
90	640	−60	0,89	—	—	—
91	640	−60	5,02	—	—	—
92	—	+20	—	67,5	14,6	60,1
93	—	—	—	70,4	19,5	66,1
110	1050	+20	10,39	—	—	—
111	640	+20	>15,00	—	—	—	Did not break
112	640	−20	9,18	—	—	—
113	640	−20	8,96	—	—	—
114	640	−40	7,32	—	—	—
115	640	−40	>15,00	—	—	—	Did not break
116	640	−40	3,52	—	—	—
117	640	−60	1,70	—	—	—
118	640	−60	3,97	—	—	—
119	640	−60	2,62	—	—	—
120	640	+20	—	70,2	16,7	60,0
121	640	—	—	71,4	16,3	54,0

Fig. 31.

The brittleness of steel during tempering is a phenomenon related to aging, and we shall have to touch upon this question again when speaking of the aging of rail steel.

Before proceeding to the heat treatment of specimens cut from rails, it was decided to check the depth to which sorbitization extends (44).

As is known, the depth of hardening is connected with the abnormality of the metal (in Ehn’s old terminology) or with the size of the austenite grain in the steel. Therefore, this part of the work was carried out taking into account the possible abnormality of the steel of those rails that were to be heat-treated.

From the Kuznetsk Metallurgical Plant named after Stalin, nine rails of different heats were obtained. Here we present data for three heats, owing to the coincidence of the results for the other six. The composition and mechanical properties of the steel are given in Table 9.

Fig. 32

Table 9

Rail No.	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Temporary resistance (kg/mm²)	Relative elongation (in %)	Impact toughness (kgm/cm²)
	C	Mn	Si	S	P
1	0.58	0.85	0.25	0.030	0.048	81.4	12.6	0.78
9	0.55	0.67	0.23	0.027	0.018	77.7	9.3	0.77
14	0.54	0.61	6.28	0.039	0.027	68.5	16.7	0.68

Specimens for determining the mechanical properties, as before, were cut from those places in the head where no segregation was observed. As can be seen, the impact toughness of these steels is low and differs little for steels of different heats. Its variation for all the rails studied lay within the range of 0.68–0.97 kgm/cm².

After etching by Baumann and with the Le Chatelier reagent, rails were selected that were almost without a segregation zone (No. 1) and with sharply expressed segregation (No. 14). The Sharpy specimens subjected to heat treatment were cut from the middle of the head and had only a small allowance (0.75 mm per side) for final machining; and also whole heads, from the middle of which Sharpy specimens were cut only after quenching and the corresponding tempering.

Fig. 33

In making specimens from the heat-treated head, careful attention was paid to ensuring that they were cut from the same place in the profile as the specimens made from the untreated rail.

The heat treatment of the head and of individual specimens consisted in quenching from 875° and tempering at 650° for 4 hours.

The data on the determination of impact toughness at +20° (specimen 100 × 10 × 10 mm; triangular notch—45°; radius—0.25 mm; depth—2 mm) are given in Table 10.

From the data of Table 10 it is evident that the middle of the head of a whole rail “does not yield” to quenching, i.e., is not through-hardened. On the other hand, a series of preceding experiments is confirmed, from which it followed that

Table 10

Rail Nos.	Toughness of a heat-treated specimen from an untreated head (in kgm/cm²)	Toughness of a specimen from a heat-treated head (in kgm/cm²)
1	9.17	1.33
1	8.84	0.64
14	10.30	1.71
14	13.57	1.81

that the metal from the segregation zone does not always have low toughness. The specimens (rail No. 14) cut from the very segregation region proved to be even tougher than those taken from sound metal.

Fig. 34

Table 11 gives data on determining the impact toughness at \(+20^\circ\) of sorbitized specimens taken from different places in the web and foot of the rail (Fig. 35), after the rail had been quenched and tempered (the temperature conditions of the heat treatment were the same as in the preceding case: rail No. 9; specimen \(10 \times 10 \times 100\) mm; triangular notch). From Table 11 it is seen that the foot is hardened well, the web likewise; only at their junction is the quenching, as microscopic examination also shows, incomplete.

A similar result was, of course, expected. It could not be assumed that the rail head, sufficiently massive in type II-A, would be uniformly through-hardened to the middle.

Fig. 35

The determination of the impact toughness of specimens prepared in this way was also carried out at low temperatures (\(-20\) and \(-40^\circ\)). Figure 36 presents data on the toughness of specim-

Table 11

Specimens	Impact toughness (in kgm/cm²)
1	8.27
2	5.28
3	0.50
4	2.13
5	7.93

of specimens cut from the head and heat-treated (the lines labeled “h.-t. specimens”), and the toughness of specimens cut from the middle of the heat-treated head (the lines labeled “h.-t. head”). The designation “raw” refers to specimens made from the head in the original pearlitic condition, from which it is evident that the raw material possesses an almost unchanged and very low toughness at normal and reduced temperatures. Sorbitizing greatly increases toughness in frost. Curves A refer to rails No. 1, and curves B to rails No. 14.

Fig. 36

If the rail is sorbitized not through and through, but all around, along the entire perimeter of the section, to a depth of at least 6–10 mm, which is entirely possible, then even in that case the resistance to impact and static bending of such a shell alone will be many times greater than that of the remaining pearlitic-ferritic core. In this case the sorbitic rail may be compared with a thick-walled tube, in which the moment of inertia differs little from the moment of inertia of a solid bar of the same outside diameter. Although, under impact loading of specimens of different size, the law of simi—

brittleness cannot be applied, but nevertheless, on the basis of experimental material, one must assume that dynamic fracture of the metal occurs first of all at the surface, and that it is precisely the surface layers that at the first moment resist the dynamic bending moments. For this reason, all local surface defects, in the form of various kinds of scratches, hardened and work-hardened spots, are so dangerous to strength under impact bending load.

In determining the grain size and abnormality (by the method of cementation and by the precipitated ferrite network) it was found that the metal is normal (according to the old terminology of Ehn), and the grain size may be designated as number three to four.

An interesting dependence between the manganese and carbon content and the magnitude of impact toughness was found by Grdin (44). In the further development of our study on the hardenability of rails, several rail steels from different plants, of very varied composition, were taken. Table 12 gives both the composition and the impact toughness at \(+20^\circ\) of specimens cut from the heads of these rails and heat-treated (quenching and tempering—\(680^\circ\)).

Table 12

Rail No.	Chemical composition (in %) C	Chemical composition (in %) Mn	Ratio \(\dfrac{\mathrm{Mn}}{\mathrm{C}}\) (in %)	Impact toughness (in kgm/cm²)	Note
1	0.70	0.89	1.27	3.90
2	0.66	0.89	1.35	5.50
3	0.38	0.63	1.65	3.18	The steel contains much slag
4	0.35	0.92	2.62	13.19	The steel contains much slag
5	0.38	0.93	2.45	8.45
6	0.45	1.14	2.53	\(>15.00\)	The specimen did not break

The sulfur and phosphorus content in these steels did not exceed the norm of our technical specifications.

We supposed that such variation in the values of impact toughness could be explained by the influence of manganese on the grain size in the steel, and at the same time on the uniformity, depth of hardening, and character of the final sorbite structure. After quenching the heads cut from them, the specimens showed the greater toughness the more manganese there was in the steel.

The data of Table 12 are presented by us in a diagram (Fig. 37). Three curves are plotted on the diagram, showing the change in impact—

toughness as a function of the carbon and manganese contents and of the ratio of their percentage contents. It has long been known that an increase in carbon content lowers the impact toughness of carbon steel, both in the as-rolled and in the heat-treated condition. If rail 3 is not taken into account (points 3 in Fig. 37), owing to its heavy contamination with slags, it is clearly seen that steel with 0.7% C has a toughness of 3.90 kgm/cm². Of course, having at our disposal five points, we could outline only the general course of each curve.

Fig. 37

A moderate increase in the impact toughness of sorbitized steel is also observed as the manganese content is increased. This, as already mentioned, is quite understandable from the standpoint of the influence of manganese on the grain in steel (the austenite and ferrite grain).

But what is especially noteworthy is the course of the curve giving the dependence of impact toughness on the ratio of the percentage content of manganese to the percentage content of carbon. Of course, this curve cannot give absolute values, but the increase in the impact toughness of steel indicated by it with increasing manganese content or, conversely, with decreasing carbon content at a given manganese content is very interesting. In this connection it must be borne in mind that the optimum toughness is determined, of course, simultaneously by both the manganese content and the carbon content. At a ratio of 2.5 the maximum toughness is attained;

with further increase of this ratio, it apparently remains constant. It is noteworthy that the maximum impact toughness (8.45; 13.19 and 15.0 kgm/cm²) was found in steels with a medium and even, perhaps, low carbon content—0.35–0.45% C. This corresponds to a manganese content of 0.92–1.14%. It follows from this that the tendency of some plants to bring the carbon content in rail steels up to 0.8% at a manganese content of 0.8–0.9% should lead to the production of hard rails, but with moderate impact toughness in the sorbitized condition.

The noted relationship, of course, cannot be regarded as universal, since the connection between manganese content, grain size in steel, and its tendency toward abnormality is complicated by a whole series of secondary circumstances.

Probably the influence of zirconium on the grain size in steel also explains the high impact toughness of zirconium steels in the normalized condition (46). As can be seen from Fig. 38, steel with 0.07% C and 0.20% Zr shows a considerably higher toughness at low temperatures compared, for example, with SAE 1020 steel.

Fig. 38

If, in steel, the value of the ratio of manganese content to carbon content falls below unity, then such a steel must possess low toughness both in the as-cast and in the sorbitized condition and—depending on the use of aluminum in deoxidation—one or another degree of abnormality.

As an example, one may cite a rail encountered in our laboratory with a content of C—0.50% and Mn—0.23%; the percentage ratio

\[ \frac{\mathrm{Mn}}{\mathrm{C}} \]

was equal to 0.46.

The material had no noticeable slag inclusions and was distinguished by great purity. The Charpy impact toughness in the as-rolled state at all test temperatures (from \(+20^\circ\) to \(-60^\circ\)) was practically the same and fluctuated around \(0.5\ \mathrm{kgm/cm^2}\). Heat treatment did not make the specimens more ductile (\(0.5\text{—}0.6\ \mathrm{kgm/cm^2}\)).

From the work of Herty (45) it follows that boiling mild steel, deoxidized only with manganese (C—0.14%), showed at \(-40^\circ\) an impact resistance (Charpy test, Izod specimen) of 2 foot-pounds. The same steel, thoroughly killed and strongly deoxidized with silicomanganese, and then with an excess of aluminum, gave 110 foot-pounds under the same conditions. Steel killed with silicon and deoxidized with manganese, as well as steel deoxidized with manganese with a small amount of silicon and aluminum, gave intermediate values of impact toughness.

On the basis of the foregoing, we believe that the impact toughness of rail steel can be increased by heat treatment—quenching over a very wide temperature interval—with subsequent tempering at 620—680°. For successful heat treatment it is necessary to take into account the character of the steel production and the conditions of its deoxidation.

Heat treatment of poor-quality steel, however carefully it is conducted, will not give satisfactory results.

During one of our reports to the engineering and technical personnel of the Kuznetsk Metallurgical Combine on the heat treatment of rails, some comrades who spoke pointed out the possibility of a more considerable warping of the rails during quenching than we had assumed. Such a circumstance would have a strong effect on increasing the required hearth area of the tempering furnace, since the more strongly a rail is bent, the worse the furnace will be filled; the rails will not lie tightly.

It is true that, as subsequent experiments showed, one-and-a-half-meter specimens gave a deflection during quenching on our second quenching machine of the order of 1—2 mm, but to vouch that the deflection will increase in proportion to the increase in length of the quenched specimen, i.e., in the case of quenching a rail 12.5 m long will not exceed 8—16 mm, is impossible.

With further development of the experiments it became clear that a well-adjusted machine gives minimal warp—

…but we were interested in the possibility of applying to rails the method of isothermal hardening, with straightening of them on a roller machine immediately after hardening to austenite.

It is obvious that, with such hardening and straightening, the rails should no longer warp in the course of the subsequent decomposition of austenite. Nor should there be any reason for distortion during tempering.

Therefore, we first carried out experiments, if one may put it so, of a qualitative nature, in order to ascertain how much the sorbite obtained after tempering steel hardened by the Lewis method differs in its mechanical properties from ordinary sorbite, especially at low temperatures.

Investigations by a number of authors on carbon steel have shown (47) that steel with a carbon content in the range of 0.42–1.01% can, during isothermal hardening, retain an austenitic structure for several minutes.

The author carried out experiments with several grades of open-hearth steel with a C content of about 0.45%. The specimens were hardened from 920° in a salt bath, which had a temperature of about 230°. After hardening they were cooled in air and tempered at 680°. No particular difference in impact toughness was observed in steel treated in this way as compared with ordinary steel heat-treated to sorbite under the same temperature conditions of hardening and tempering.

We then returned to this subject a second time, deciding to investigate the quantitative aspect of the phenomenon (48). The material for the investigation was KMK rail steel of the following composition: C—0.53%, Mn—0.86%, Si—0.30%.

The transformation of austenite was followed by means of a magnetoscope. Despite its simplicity and seeming primitiveness, the magnetoscopic method gives very good results and quite comparable data.

The specimens were heated in a furnace to 850° and transferred for 30 sec. into a salt or oil bath, in which the required constant temperature was maintained in the interval from +20 to 450°. The specimens were quickly transferred from the bath into the magnet furnace, heated to the temperature of the bath (230°). There the specimens were held at constant temperature until the austenite transformation process was complete. At this time, every minute, the amount of transformed austenite was measured. The measure of the amount of austenite present in the specimen was the number of amperes in the electromagnet at which the specimen began to be attracted. The current strength in this case is a quantity proportional to the amount of transformed austenite.

The summary curve for this steel is given in Fig. 39. The points from which the curve was constructed were determined every 50°.

Fig. 39

The process proceeds most slowly in the region of 220–300°. And, probably, it is precisely at these temperatures that one may try to carry out isothermal hardening of rails.

Unfortunately, we have so far not succeeded in doing this even for half-meter pieces.

III. SEMI-PLANT-SCALE TESTS

In order to verify the considerations set forth above by hardening and tempering half-meter rail ends intended for standard impact tests, in the winter of 1932/33 a team was sent by the Siberian Institute of Metals to Dneprodzerzhinsk, to the Dzerzhinsky Plant.

By agreement with the plant, it was decided to conduct the sorbitizing tests with Bessemer rails of type II-A, and then with open-hearth rails.

Pieces of rails 1.5 m long were selected for testing according to scheme (49), which had been developed by the team at the plant (Fig. 40).

The work plan provided for possible variations in heat treatment: 1) hardening of the specimens immediately after cutting; in this case the hardening temperature usually coincided with the temperature of the rail at the moment it left the last pass, and fluctuated within the range 780–1000°; 2) hardening of a specimen that had cooled after rolling and was reheated in a furnace; hardening was followed by tempering at 630–670°.

The cut pieces of rails, 1.5 m long, after stamping, were gripped at both ends with special tongs and immersed horizontally in a rectangular tank with running water. The dimensions of the tank were \(2 \times 0.45 \times 0.4\) m.

It was not possible to harden a rail by this method. If quench cracks did not appear immediately, at the very moment of hardening, they were found during storage after hardening. Raising the temperature of the quenching bath did not change the essence of the matter.

The appearance of quench cracks was explained both by the complexity of the rail profile and by the diversity of temperatures at different points of the cross section.

After a number of attempts, the conclusion was reached that it was necessary to give the rail an incomplete hardening, i.e., to cool the rail not to the temperature of the bath, but to a temperature of 200–350°.

This was accomplished by immersing the rail for a certain definite interval of time, found by a series of experiments. Sometimes a similar result was obtained after several brief immersions of the specimen in the quenching tank. The time of residence of the specimen in water varied within the limits of 15–30 sec. The surface temperature was measured with a special strip thermocouple and rose on the surface of the head to 200–350°, and on the base to 150–200°. The surface of the rail section reached this temperature only 1–1.5 min after it had been removed from the bath.

After such hardening, the specimens, having received a slight self-tempering to troostite, could withstand storage for many days even in frost.

Some of the specimens were slowly cooled after cutting and reheated for hardening in the flame furnace used for annealing wagon axles. The hardening of these specimens was carried out by the method already described.

Tempering of the hardened pieces of rails was carried out in a chamber annealing furnace for tempering tires during sorbitizing. Upon reaching the required temperature (650–670°), the rails, after the corresponding holding period, were cooled with the furnace; for this purpose special ventilation openings located in the roof of the furnace and connected with a special exhaust pipe were opened.

Testing of the heat-treated rails was carried out on the plant drop hammer with a “tup” weighing 500 kg and a lift height of 7.3 m. In testing rails frozen to −40°, a “tup” weighing 1000 kg with a lift height of 5 m was used. A temperature of −20° was provided by a cooling mixture of snow and common salt, and −40° by a mixture of solid carbon dioxide and denatured alcohol. At the same time, two pieces of rail were frozen in a specially made bath.

The time required for through-cooling of the rail was found by determining the course of the temperature–time curve, taken with a pyrometer placed in a hole drilled in the middle of the head of the test rail. At −40° it was necessary to hold the rail for 40–45 min. Since the drop-hammer tests were carried out at the temperature of the surrounding air (about

Fig. 49. Scheme for selecting Bessemer rail specimens for heat treatment and testing at low temperatures

Hard heats, head end of the rail

\(C = 0.35\text{–}0.40\%\)

• As rolled
• \(+20^\circ\): 3
• \(-20^\circ\): 3

• \(-40^\circ\): 3

• Quenched directly from under the saw

• Tempered at \(630^\circ\)
  • 3
  • 3
  • 3
• Tempered at \(670^\circ\)
  • 3
  • 3
  • 3

Hard heats, middle of the rail

• As rolled, 9 pcs.
• \(+20^\circ\): 3
• \(-20^\circ\): 3

• \(-40^\circ\): 3

• Quenched directly from under the saw

• Tempered at \(630^\circ\)
  • 3
  • 3
  • 3
• Tempered at \(670^\circ\)
  • 3
  • 3
  • 3

Medium heats, head end of the rail

\(C = 0.40\text{–}0.45\%\)

• As rolled
• 3
• 3

• 3

• Quenched directly from under the saw

• Tempered at \(630^\circ\)
  • 3
  • 3
  • 3

• Tempered at \(670^\circ\)

  • 3
  • 3
  • 3

• Secondary heating at \(850^\circ\)

• 3
• 3
• 3

Soft heats, head end of the rail

\(C = 0.45\text{–}0.50\%\)

• As rolled
• 5
• 5

• 5

• Quenched directly from under the saw, tempered at \(630^\circ\)

• 5
• 5

• 5

• Secondary heating at \(850^\circ\), tempered at \(630^\circ\)

• 3
• 3
• 3

Medium heats, middle of the rail

• As rolled
• 3
• 3

• 3

• Quenched directly from under the saw, tempered at \(630^\circ\)

• 3
• 3

• 3

• Heated to \(1050^\circ\), tempered at \(630^\circ\)

• 3
• 3
• 3

—10°), and since there was no warm shed at the pit at that time, during the 20–40 sec. required for placing the rail on the anvil of the pit and carrying out the test itself, the heating of the specimen did not exceed 2–3°.

Without considering all the material on the testing of Bessemer rails, we shall confine ourselves to presenting only one diagram (Fig. 41), constructed on the basis of testing at normal temperature of soft rails, both as-rolled and sorbitic. As may be seen, the sorbitic specimens (curve A), for the same number of blows, gave a smaller deflection than the as-rolled ones (curve B). The deflection after the fourth blow for the sorbitic rail is equal to 130 mm, and for the as-rolled one, 260 mm.

Fig. 41

The structure of the heat-treated rail in the base, web, and surface of the head is sorbitic; in the center of the head—sorbite-like pearlite and ferrite.

But at −40°, both as-rolled Bessemer and sorbitic rails were brittle and broke from a single blow.

Better results were obtained with open-hearth rails, which, in the as-rolled state at normal temperature, withstood on average three blows without fracture. Sorbitized open-hearth rails, quenched immediately after cutting and tempered at 630° and 670°, did not break even after five blows. In the frozen state (at −40°), the open-hearth rails exhibited quite exceptional toughness. Under blows of a “drop hammer” weighing 1000 kg, raised to a height of 5 m, of six as-rolled open-hearth rails from the head and from the middle part of the ingot, five pieces broke from a single blow, and only one (from the head), having withstood the first blow, broke on the second. Twelve open-hearth head rails, quenched and tempered (630 and 670°), withstood at the same temperature (−40°) three blows each.

Table 13

No. in order	C	Mn	Si	S	P	Steel grade: Bessemer or open-hearth	Rail in relation to the place it occupied in the ingot	Hardening temperature (°C)	Tempering temperature (°C)	Yield point (kg/mm²)	Tensile strength (kg/mm²)	Relative elongation (%)	Reduction of area (%)	Test results on impact at ordinary temperature (+20°)
1	0,35	0,87	0,08	0,066	0,086	Bessemer	Head	850	630	46	70,6	16,6	62,1	Fracture on the first blow
2	0,46	0,81	0,09	0,074	0,084	Bessemer	Middle	Raw	—	32	70,4	16,0	38,4	Deflection of the bend on the fourth blow
3	0,54	0,85	0,24	0,045	0,039	Open-hearth	Middle	780	630	59	88,7	13,3	46,2	Fracture on the seventh blow
4	0,54	0,85	0,24	0,045	0,039	Open-hearth	Head	820	630	52	82,7	14,0	59,0	Severe distortion of shape on the fifth blow
5	0,58	0,87	0,27	0,029	0,035	Open-hearth	Middle	Raw	Raw	30	79,6	11,3	20,1	Fracture on the first blow

Most of them did not break even on the third blow. After three blows the profile of the unbroken rails was very strongly distorted, and it was no longer possible to test them further. Only two specimens broke on the second blow. One of the specimens was tested with a “baba” blow of 1 m from a height of 10.7 m and broke only on the third blow (test temperature −40°).

These experiments show that open-hearth sorbitic rails are not cold-brittle at −40° and surpass, in toughness, identical raw rails under ordinary temperature conditions (about +20°).

Table 13 gives data characterizing the chemical composition and mechanical properties of some Bessemer and open-hearth rails with which the work was carried out.

Unfortunately, the winter of 1932/33 was marked at the plant by numerous troubles with Bessemer rails, and we are inclined to attribute the failure of their heat treatment chiefly to their poor quality. Bessemer rails are deoxidized differently from open-hearth rails, and this circumstance was bound, of course, to affect the quality of the sorbitized metal, though not to such a degree.

In these rails there was a large quantity of slag inclusions and a strongly developed segregation zone. A typical picture giving an idea of their distribution is shown in Figs. 42 and 43. Fig. 42 (magnification 100; unetched) relates to a Bessemer rail, and Fig. 43 (magnification 100; unetched) to an open-hearth rail. It should be noted that in Bessemer rails from other plants there are normally far fewer slag inclusions. Nor can one pass over in silence the circumstance that the head rails are especially contaminated with slags (Fig. 44, magnification 100; unetched), but there is also much slag in rails from the middle part of the ingot.

Fig. 42

As is known, many investigators have studied the influence of nonmetallic inclusions on the mechanical properties of metal (50). From all investigations of this kind it may be concluded that the decisive influence on the quality of the metal is exerted not only by the quantity of inclusions, but chiefly by their size and by the character of their distribution in the main mass of the metal.

Fig. 43

In the work of Kinzel and Grafts (51), the measure adopted for the contamination of steel by nonmetallic inclusions was the total length of the inclusions visible on the polished section (in mm), referred to 1 mm² of the area of the polished section. The minimum size of inclusions taken into account is considered to be 0.005 mm. The authors divide all steels into three groups. The first group includes those steels in which the total length of inclusions is less than 0.5 mm per 1 mm²; this is the cleanest steel. For steels of the second group this value ranges from 0.5 to 0.7 mm per 1 mm²; this is steel of medium cleanliness. The third group includes heavily contaminated steel, for which the total length of inclusions is greater than 0.7 mm. Steel

Fig. 44

the third group, according to their investigations, possessed very low toughness.

If, by Kinzl’s method, one calculates the length of inclusions visible at a magnification of 100 times for those Bessemer rails of the Dzerzhinsky Works with which our brigade worked, the value obtained is 5–15 mm per 1 mm². The same value for open-hearth rails ranges from 1–2 mm per 1 mm².

In the extensive practice of the Siberian Institute of Metals and the Tomsk Industrial Institute in the study of rails in the course of expert examinations for the Tomsk Railway, there was once encountered a rail from the Krupa Works, which broke in January 1932. Specimens from this rail, tested on a Charpy machine at temperatures from +20 to −60°, showed a constant impact toughness, independent of the testing temperature, on the order of 0.7–0.5 kgm/cm². Sorbitized specimens from the same rail gave, at +20°, an impact toughness of about 3 kgm/cm², and at −20, −40, and −60° again a constant toughness within the limits of 0.7–0.6 kgm/cm². The remaining static characteristics were within the standard range.

Analysis showed the usual composition of Thomas metal: C — 0.38%; Si — 0.10%; Mn — 0.63%; S — 0.023%; P — 0.065%. The segregation zone did not exceed the usual limits, and specimens cut from it revealed no special abnormalities. But in determining the total length of slag inclusions, 6 mm per 1 mm² was obtained.

The foregoing gives us the right to consider that the unsatisfactory impact toughness of the Bessemer rails of the Dzerzhinsky Works depends chiefly on their considerable contamination.

It is characteristic that a large number of rails supplied by this works to the Tomsk Railway in 1931–1933 developed transverse cracks.

Experience in sorbitizing one-and-a-half-meter pieces of rails in Dneprodzerzhinsk showed that, for successful hardening, it is necessary to have sound metal and to cool the rail very carefully. The very complex profile of the rail, moreover, after rolling has different temperatures at different points, which complicates matters. On the other hand, the different parts of the profile, owing to their different thicknesses, will cool during hardening at different rates; martensite will not form at one and the same moment, and the change in volume will be very varied at individual points of the section. The brigade working in Dneprodzerzhinsk, having encountered such a difficulty, found a solution by applying self-tempering of the rail at the expense of the heat retained inside the profile, chiefly in the head. During hardening the rail was cooled for a precisely measured time and then, owing to this internal heat, was again heated more or less uniformly to 200–350°.

Naturally, such a circumstance would complicate the hardening of rails of normal length.

The author therefore conceived the idea of hardening long rails in water spray. Obviously, this variety of hardening is more flexible than the ordinary method, and makes it possible to remove heat from any points lying on the surface of the rail, at a desired rate—within definite limits, of course—by regulating the amount of atomized water supplied.

For this reason the author and Engineer Andreev (52) carried out preliminary work and then designed two hardening machines for hardening rails 1.5 m long. In designing the second machine, provision was made for the possibility of adapting it, without special alterations, for hardening rails of normal length. The preliminary work mentioned consisted in determining the cooling rates over the rail profile during hardening. Knowledge of these rates, which depend chiefly on the peculiarities of the rail profile and on the conditions of temperature distribution in it, made it possible to arrange the sprayers in the hardening machine with sufficient rationality.

The cooling rates of the rail during hardening were recorded with a self-recording galvanometer of Academician Kurnakov, with a drum carrying light-sensitive paper and running at a speed higher than normal. During rapid changes of the emf in the thermocouple circuit, the instrument has considerable inertia; however, as the preliminary experiments showed, the inertia of the instrument manifests itself only in smoothing out the bends on the cooling curves and in a certain displacement of the initial moments of cooling.

Since the muffle of our furnace did not accommodate the entire rail profile, the first experiments were carried out separately on the head and on the base with web. For this purpose a piece of rail 150 mm long was cut lengthwise. Rails of small profile entered the furnace whole. In a specimen of the indicated length, holes 4 mm in diameter and 60 mm deep were drilled from the end; their arrangement is shown schematically in Fig. 45. Nickel–nichrome thermocouples, carefully insulated with asbestos cord, were inserted into these holes. Special attention was paid to ensuring that the metal of the thermocouple junction had good contact with the metal of the rail and that water could not enter the hole when the rail was wetted. The latter was achieved by additional coating with clay at the point where the thermocouple entered.

The prepared specimen (C—0.40%; Mn—0.30%) was carefully heated to the hardening temperature (850°) for about 2 hours. It was then removed from the furnace with tongs and immersed in a tank of water at a temperature of +20°, to a depth of 200 mm. During hardening the specimen was vigorously moved in all directions.

The result of one of the experiments is given in Fig. 45 in the form of a family of curves. It should be borne in mind what we said at the beginning: the character of the lines was affected by the inertia of the instrument, and they therefore have, of course, only relative significance. Their appearance, nevertheless,

Fig. 45

explains the appearance of internal stresses and cracks during the hardening of such a complex profile as an ordinary rail.

On the basis of the course of the curves one may conclude that the maximum amount of water spray should be supplied to the head. The foot cools very vigorously (curve 9), and its ends cool especially rapidly (curve 7). The middle of the web also cools at a very considerable rate (curve 6); heat is removed most slowly from the middle of the head (curves 4, 5). The cooling rate of the middle of the foot (curve 9) fluctuates around \(30^\circ\) per sec. Taking into account the lag of the galvanometer readings in comparison with the actual temperatures, and the cases, observed as a rule, of through-hardening to martensite of the rail foot and web, it may be considered that, under the conditions of our experiment, this rate (\(30^\circ\) per sec.) is precisely such a rate at which, on the French diagram, the points \(A'r_1\) and \(A''r_1\) merge.

One of the problems that must be solved in designing a hardening machine consists in the proper regulation of the cooling rates by supplying the corresponding quantities of atomized water.

The first machine, designed by engineer D. S. Gruzdev and the author, consisted of 7 rows of pipes arranged parallel to the length of the specimen. On each of the lower pipes 9 atomizers (with helical channels) were fastened; in the upper pipes holes (9 each) of 3 mm diameter were drilled.

In the attached schematic sketch (Fig. 46) the arrangement of the pipes is clearly visible.

For moving the rail along the hardening machine, a primitive roller table made of gas pipes was constructed.

This quenching machine, although it made it possible to quench several dozen one-and-a-half-meter lengths of rail at the Kuznetsk Metallurgical Plant, had a number of very substantial shortcomings, of which the most important were: 1) lack of regulation of the water supply through the individual pipes; 2) the need to move the rail by hand, which did not guarantee a uniform quenching rate for all specimens; 3) considerable water consumption.

Fig. 46

Therefore, in the heat-treatment laboratory of the Siberian Institute of Metals, by Engineer D. S. Gruzdev and the author, with the participation of a consultant (head of the hydraulic laboratory of the Tomsk Industrial Institute, Docent N. A. Balakin), this machine was redesigned with all the observed defects taken into account. The designed machine was built in the workshops of the Siberian Institute of Metals and tested at the Kuznetsk plant. In designing it, the intention was to have a machine which, if increased in length a number of times, would be able to quench a rail of normal length.

As can be seen from Figs. 47 and 48, the machine is equipped with a roller table driven by a motor through a worm gear. The roller table had (without changing the number of revolutions of the motor) three speeds, forward and reverse motion. The cooling system consists of 8 pipes. Each pipe has its own valve for regulating the quantity of water and is provided with atomizers. The number of atomizers on the upper pipes reaches 40 per 1.5 m of pipe length. Water from the water main is supplied through a 120 mm pipe into two welded annular headers, into which the pipes with atomizers are screwed. There are 2 headers (one at each end of the machine); this promotes a more uniform distribution of pressure along the length of the pipes and, consequently, more proper operation of the atomizers.

On the basis of Docent Balakin’s experiments, the design of the ordinary atomizers with screw channels was changed, which made it possible to obtain, instead of a circular cone of atomized ...

Fig. 47

Fig. 48

water, an elliptical cone, with the long axis of the ellipse parallel to the motion of the rail in the machine. At a pressure of 4 atm, such sprayers produced a fine water mist possessing a high impact energy of the individual droplets.

By regulating the speed at which the one-and-a-half-meter rail ends passed through the machine, the water pressure in the individual pipes, and the number of operating sprayers, it was possible to obtain an entirely straight, crack-free specimen for impact tests on the factory drop hammer. The deflection of the quenched one-and-a-half-meter specimen did not ex—

rose by 2–3 mm, and in some cases was practically equal to zero.

But when quenched specimens were stored without tempering (especially at low temperatures in the rolling shop), the appearance of cracks was sometimes observed. In some cases they were detected in the first 25–30 min after quenching, and sometimes after a day or more. Most quenching cracks ran along the web, parallel to the head. Cracks of this kind arose during the quenching process, chiefly in head rails, while the rail was passing through the machine. During storage of quenched specimens, transverse cracks were sometimes obtained. There were many specimens which, despite the low temperature in the shop (below 0°), when stored for several days showed no traces of cracks at all. This indicates the possibility of conducting the quenching process in such a way that the internal stresses arising in the steel during quenching will be brought down to minimum values, and then no cracks at all will appear in the rail.

If the quenched pieces of rails were tempered during the first 30 min after quenching, no rejects due to quenching cracks were obtained at all. The first tempering trials, until an electric tempering furnace was ready, were carried out by placing the quenched specimens between hot rail cuttings of suitable length. Despite the seeming primitiveness of the method, it proved possible to temper several one-and-a-half-meter specimens quite satisfactorily at a temperature of 500–550° (cooling in air). The tempering temperature was checked with pyrometers. During these trials the rolling temperature varied within the range 860–1040°; the specimens, however, were quenched in the interval 950° ± 50°. The drop-weight test was conducted at −15°. The height of drop was 7.3 m, and the weight of the “drop hammer” was 500 kg.

Table 14 gives the results of the drop-weight test of these sorbitic rails (53).

Table 14

Heat No.	Rail in relation to the position occupied in the ingot	Number of blows: 1	Number of blows: 2	Number of blows: 3	Number of blows: 4	Number of blows: 5	Number of blows: 6	Notes
1	Head	30	60	89	—	—	—	Head delaminated
1	Middle	29	56	68	85	—	—	Broke on the fifth blow
1	Middle (as-rolled)	37	67	95	—	—	—	Broke on the fourth blow
1	Middle	24	32	63	75	98	—	Broke on the sixth blow
2	Head	31	60	89	177	—	—	Broke on the fifth blow
2	Middle (as-rolled)	40	70	103	—	—	—	Broke on the fourth blow

The chemical composition of the heats is given in Table 15.

Table 15

Heat Nos.	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)
Heat Nos.	C	Mn	Si	S	P
1	0.54	0.85	0.26	0.013	0.013
2	0.65	0.85	0.27	0.012	0.013

The Brinell hardness of the rolling surface of the untreated specimens ranged from 255 to 262; heat-treated rails at the same locations gave: 269–277 (heat No. 2) and 286–302 (heat No. 1).

Microscopic examination showed that, with the exception of the middle of the head and the middle of the base (at the transition into the web), where islands of ferrite were noticeable, the material has a sorbitic structure.

In the winter of 1934/35, work on the sorbitizing of rail specimens 1.5 m long for full-scale tests was continued by Engineer Gruzdev (54) at the Kuznetsk Metallurgical Works named after Stalin. The plan for carrying it out had first been worked out by the author jointly with Gruzdev. The intended number of specimens (250 pcs.) was hardened, but it has not yet been possible to test them all at −40°, since there was no refrigerating installation at the plant, and despite all attempts it proved impossible to obtain liquid carbon dioxide or liquid air in Stalinsk. Therefore, in the full-scale tests it was necessary to confine ourselves only to those temperature conditions determined by the temperatures of the winter months. A total of 150 specimens were tested; according to the plan, the remaining 100 had to be broken at −40°, and they are still being stored.

Hardening was carried out in the hardening machine described, immediately after cutting, without any additional heating.

Tempering was conducted in a nichrome muffle furnace designed by the author specifically for this purpose and permitting the simultaneous tempering of five ends of type III-A rails. The furnace consisted of a rectangular-section muffle, 2 m long, welded from 6-mm boiler iron, with a winding of nichrome strip (10 × 0.5 mm). Along the length of the muffle, the winding consisted of three sections for connection to a three-phase-current circuit. The furnace provided uniform heating up to 700°; the temperature along the length of the muffle was controlled by several thermocouples.

As already mentioned, 250 one-and-a-half-meter specimens of various composition and from different parts of the ingot were heat-treated. During hardening, owing to the ease of adjustment of the hardening machine, the specimens did not warp and did not develop cracks, except for rails from the head part of the ingot with traces—

with spots, remnants of shrinkage cavities, and similar defects, the presence of which makes it altogether impossible to harden the steel. Tempering was carried out immediately after hardening, since hardened specimens, as already indicated, when stored for a long time in a poorly heated shop, sometimes developed cracks.

Tables 16 and 17 give data on impact tests of several specimens.

Table 16

No. of heats	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Heat treatment: hardening and tempering	Heat treatment: hardening and tempering	Heat treatment: hardening and tempering
No. of heats	C	Mn	Si	S	P	Tempering temperature (in °C)	Duration of tempering (in hours)	Cooling after tempering
1	0.55	0.62	0.18	—	—	620	3	Air
2	0.64	0.82	—	—	—	680	2	»
3	0.58	0.84	—	—	—	680	3	»
4	0.62	0.76	—	—	—	560	2	Water
5	0.58	0.67	0.24	0.03	0.02	560	2	Air

Note. In all specimens the combined content of phosphorus and sulfur did not exceed 0.06%.

Normal rails from the middle or lower part of the ingot proved, as a rule, much better than head rails in the impact test. At \(-24^\circ\), head rails in the as-rolled condition usually broke with a single blow; sorbitized head rails withstood two or three blows, no more, breaking on the third or fourth. Sorbitized specimens from normal, non-head rails withstood 5–6 blows, and some did not break even on the seventh blow, becoming so deformed that further testing could not be carried out.

It should be noted that sorbitizing not only increases the number of blows withstood by a rail, but it also affects the change in the deflection after each blow. Let us take, for example, the rail of the first heat¹. Its composition: C—0.55%; Mn—0.62%; Si—0.18%; the combined content of S and P does not exceed 0.06%. The hardened rail was tempered at \(620^\circ\) and, after tempering, cooled in air. A specimen cut from the rail in the as-rolled condition, after the first blow on the impact machine, gave a deflection of 47 mm; the sorbitized specimen, taken from the same rail, showed a considerably smaller deflection—only 33 mm. The deflection of the as-rolled specimen is 43% greater than that of the heat-treated

Table 17

No. of heats	No. of specimens	Shear cutting force after impacts (mm), 1	2	3	4	5	6	7	Increase in shear cutting force after impacts (mm), 1	2	3	4	5	6	7	Test temperature (°C)	Notes
1	1	33	67	101	135	196	220	—	33	34	34	34	41	24	—	−24	Sorbitic, did not break
1	2	47	88	128	173	—	—	—	47	41	60	45	—	—	—	−24	Gray
1	3	33	67	100	135	177	200	—	33	34	34	35	41	23	—	−24	Sorbitic, did not break
1	4	46	85	123	170	—	—	—	46	39	38	47	—	—	—	−24	Gray
2	1	30	67	97	130	163	197	—	30	37	30	33	33	34	—	−2	Sorbitic
2	2	37	67	95	—	—	—	—	37	30	28	—	—	—	—	−2	Gray
2	3	34	66	119	—	—	—	—	—	—	—	—	—	—	—	−2	Head, sorbitic, cracked lengthwise
2	4	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	Head, raw, broke from a single blow
3	1	35	68	110	—	—	—	—	35	33	42	—	—	—	—	−24	Head, sorbitic
3	2	—	—	—	—	—	—	—	—	—	—	—	—	—	—	—	» raw
4	1	25	48	72	96	120	—	—	25	23	24	34	24	—	—	−21	Sorbitic
4	2	38	70	—	—	—	—	—	38	32	—	—	—	—	—	−21	Gray
4	3	38	—	—	—	—	—	—	38	—	—	—	—	—	—	−21	»
5	1	30	60	86	114	142	—	—	30	30	26	28	28	—	—	−21	Sorbitic
5	2	45	82	118	160	—	—	—	45	37	36	42	—	—	—	−21	Gray

worked. After the second blow the deflection of the as-rolled rail increased to 88 mm, i.e., by 88%, while that of the sorbitic rail reached 67; although it increased by 100%, in absolute value it nevertheless remained less than the deflection of the as-rolled rail by 21 mm, or by 31%. The fourth blow at −24° was the last for the as-rolled rail; after it the deflection was 173 mm, whereas under the same conditions the sorbitic specimen gave only 135 mm and did not fail at all, but merely deformed. The deflection after the sixth blow reached 220 mm, and the ordinary design of the drop-hammer supports did not permit any further increase in it.

This indicates that sorbitized rails withstand blows better at low temperatures not only in the sense of a reduction in impact brittleness, but also from the standpoint of the possibility of residual deformations appearing beyond the elastic limit. It is known that the elastic limit, both in compression and in tension, is higher in sorbitic steel than in as-rolled steel, and this also affects the magnitude of plastic deformation under impact loading.

As was indicated in its proper place, we sought to lower the tempering temperature during sorbitization, in order to obtain a rail that would be both frost-resistant and at the same time hard, so as to reduce wear of the running surface.

As can be seen from Tables 16 and 17, tempering at 560° gives a sufficiently satisfactory impact toughness in the drop-hammer test of an entire rail. Heats Nos. 1 and 5 are practically identical in chemical composition, as is the test temperature (−24 and −21°), but tempering to 560° made the specimens of heat No. 5 less tough, though at the same time considerably stronger. The deflection of the rail tempered at 620°, after the sixth blow, proved equal to 220 mm, and its increase during the first blows was 34 mm; while the deflection, when the tempering temperature was lowered to 560° under the same conditions, falls to 142 mm and its increment per blow to 30 mm. At the same time, the rail tempered at 560° nevertheless possesses considerable toughness. At −21° it withstood five blows and broke on the sixth.

But in this connection it is necessary to take the following into account. In heat-treating a small laboratory specimen we obtained a uniform sorbitic structure throughout the entire cross section. The dimensions of the profile of a modern rail and its complexity do not permit, even in the case of a suitable composition of carbon steel, any guarantee of a truly through hardening. On the other hand, it may be asserted that the same complexity of the profile will probably lead to the fact that the region of impact brittleness for a rail taken as a whole and tested in impact without a notch, under the same temperature conditions of tempering to sorbite as for a small specimen, shifts toward lower temperatures. More simply put, as is evident from the foregoing, the rail may be tempered, in order to obtain sufficient impact toughness, not to 620–680°, as a small specimen ...

only to 550°, thereby increasing the surface hardness, which is desirable for ensuring a certain wear resistance of the head.

Unfortunately, in drop tests of unnotched rails the conditions are somewhat different than in the fracture of specimens with a notch for a laboratory-type drop hammer; moreover, dynamic tests on factory-type drop hammers provide, properly speaking, only a qualitative characterization of the metal.

We had the idea of testing one-and-a-half-meter rail ends in such a way as to obtain not only a qualitative but also a quantitative characterization of the impact resistance of the rail in the form of a complete profile. But, unfortunately, for certain reasons this attempt could not be carried out. In view of the importance, as it seems to us, of such a question, we shall try to describe this attempt briefly.

The practice of accepting rails by foreign railways from supplier plants gives the following picture. In all countries (according to materials from 1933), with the exception of France, a drop test is carried out on whole, unnotched pieces of rail of a definite length. According to the French technical conditions of acceptance, a recess with a radius of 500 mm is cut out on the head of the rail. The purpose of this operation is to take account of the future wear of the rail, when the height of the head will decrease, and not at all to obtain a specimen with a notch, like a Charpy drop-hammer specimen.

In drop tests the rail must not break, but only bend. If the rail nevertheless breaks, then the heat is considered rejected if fracture occurred on the first (or second) blow.

Under certain temperature conditions (testing at low temperatures) it may be assumed that in many cases the rail will break from a single blow; but when working on all existing factory drop hammers we still cannot know the measure of the impact toughness of the metal. When struck by a “baba” of normal weight (500 kg) from the usual height, determined by a known formula for some type of rail (for example, 7.3 m), the rail broke; but it cannot be said that precisely \(7.3 \times 500 = 3650\) kgm, or some definite fraction of this value, was expended on its destruction.

This led us to propose the following method for testing unnotched profiles, primarily rails, for cold brittleness.

As already mentioned, a one-and-a-half-meter specimen can always be cooled to such a temperature at which it becomes so brittle that it breaks from a single blow. With the development of the industry of bound nitrogen, the easily obtained such temperature is the boiling temperature of liquid oxygen. The temperature of various mixtures with solid carbon dioxide is also accessible. Obviously, the magnitude of the energy of the blow that destroys

a rail at such a constant low temperature, at which all rails are brittle, may serve to a certain degree as a measure of the toughness of the metal. But in order to obtain such a characteristic it is necessary to be able to measure that part of the total impact energy which actually went into the fracture of the specimen. Proceeding in this way, we in effect replace the notch, which imparts impact brittleness even to a tough metal, by the action of the reduced test temperature. The number of kilogram-meters required to fracture a rail of a definite section on an anvil of ordinary factory-type construction, with a falling “tup,” and with a distance between supports of 105 cm, may be taken as a quantitative characteristic of impact toughness at known low temperatures, analogous to the Charpy, Izod, and other numbers. Obviously, the temperature conditions must then be constant and must ensure fracture from the first blow only.

We immediately rejected the idea of notching the rail in order to take account of the influence of subsequent wear of the head and to concentrate stresses so as to cause obligatory fracture from a single blow, since we assumed that any notch in a specimen of such a complex cross-section would introduce into the process of impact deformation too many additional factors, difficult or altogether impossible to take into account.

The author, together with Docent N. A. Balakinin, designed two instruments which, in laboratory tests, gave excellent results, but which at that time it proved impossible to apply under factory conditions on a “real” rail anvil.

The idea of the first instrument was as follows. If on the falling “tup” of the anvil there is a luminous point executing oscillations of constant period in a direction perpendicular to the trajectory of the “tup” during its fall, then, when the flying “tup” is photographed, this point will give on the negative the image of a sinusoid with variable wavelength. By measuring the wavelength of this sinusoid and knowing the number of periods of oscillation of the point per second, one can find the velocity of fall of the “tup” at any moment along its path, and with very high accuracy.

Now suppose that the “tup,” with a velocity depending on the height of lift, struck the piece of rail under investigation and broke it. Obviously, its velocity will change after this, since it has expended part of its energy on fracturing the specimen. But we shall determine this changed velocity from the wavelength of the sinusoid. Knowing how much the velocity of the “tup” has changed, it is evidently very simple to calculate also the number of kilogram-meters expended on fracture.

This is the simple idea of the instrument. To realize it we caused a steel spring \((A)\) (Fig. 49) to be attracted to a coil with an iron core \((B)\), supplied with alternating current. If the number of periods of oscillation of the spring is equal to the number of periods of the alternating current, then the oscillations will be continuously sustain—

will occur, resonance will set in, and the amplitude of the oscillations can be made sufficiently large—depending on the dimensions of the spring. Since large electric power stations operate so uniformly that synchronous electric clocks are connected to their network, the period of oscillation of the spring will serve as a reliable time meter and will ensure the accuracy of the experiment.

Fig. 49

An electric bulb from a pocket flashlight was fastened to the free end of the spring; it was supplied either by batteries or by several turns of the coil, as from an autotransformer.

With a sufficient length of the spring (about 200 mm) and with synchronism obtained by selecting the cross-section of the spring, the amplitude of its end—and consequently the path of the bulb—can easily be brought to 25–30 mm. This is quite sufficient to obtain entirely clear curves on a 13 × 18 plate when photographing from a distance of 2–4 m.

The entire apparatus, whose arrangement is shown in Fig. 49, is protected by a casing made of boiler iron, with a small opening to make it possible to photograph the bulb.

The velocity of the “hammer block” at the moment of impact against the rail need be determined only once, since it does not change, being set by the height

lifting, by friction in the guides, and by air resistance. Obviously, for a given drop hammer these quantities may be regarded, with no great error, as constant. Therefore the camera is set up so as to photograph the small lamp after the “ram,” having broken the specimen, flies downward. Since the distance from the specimen to the shock absorber (usually a log) lying on the anvil is not large in many drop hammers, it is necessary to raise the supports in order to bring the distance to 400–500 mm.

The apparatus was first tested and adjusted in the laboratory; for this purpose we built a model of a drop hammer with wooden guides and a wooden “ram,” with the apparatus mounted on it. The height to which the “ram” was raised reached 4 m. Wooden specimens were broken on the drop hammer, and Fig. 50 gives an idea of the character of the curve recorded by the camera. Obviously, the work expended in fracture can be determined from the formula:

\[ T = mgH\left(1-\frac{\lambda_2^2}{\lambda_1^2}\right), \]

where \(m\) is the mass of the “ram,” \(H\) is the height to which it is raised, \(g\) is the acceleration due to gravity, and \(\lambda_1\) and \(\lambda_2\) are the wavelengths of the sinusoid before and after fracture.

In Dneprodzerzhinsk, during work on sorbitizing, it was not possible to use this apparatus, since the specimens did not break from a single blow. It was installed on the “ram” of the drop hammer, and the experiments showed its complete suitability and the correctness of its design. Fig. 51 shows the “ram” with the mounted apparatus (the casing removed).

Fig. 50

The second apparatus, designed for the same purpose, is based on a different principle. If the falling “ram” of the drop hammer strikes, with its side, the rim of a wheel freely mounted on the axle, the latter will be set in motion, and its circumferential velocity, if contact with the “ram” occurred without slipping, will be equal to the linear velocity of the “ram.” If the wheel rotates easily (mounted on ball bearings), it will make more than a dozen revolutions before stopping. Taking into account the damping of the rotational speed of the wheel during one revolution, one can very easily find the initial circumferential velocity at the moment of contact, and consequently also the velocity of the “ram.”

After preliminary experiments with the same wooden drop hammer discussed above, we adopted the following design (Fig. 52): a very light aluminum wheel of small inertia is mounted on two ball bearings. Its rim is wrapped with a rubber strip 3 mm thick. On the side of the “ram” a similar rubber strip is stretched over two blocks. This strip engages

Beyond the wheel rim, when the “drop weight” flies past it, it imparts to it practically without any slipping its own speed. So that, if there is “play” in the guides of the ram, the wheel will not suffer from too severe a blow, it is mounted on two springs. Every quarter revolution the wheel closes the chronograph contact. From the record on the chronograph tape one can calculate both the mean peripheral speed of the wheel and the nature of its damping with time, and determine the actual speed of the “drop weight” after the blow.

Fig. 51

Of course, the whole apparatus must be well protected by a suitable casing against damage by fragments of the rail.

In tests on the above-mentioned laboratory ram, the apparatus gave excellent results in terms of the accuracy of the readings, but to use it in pro—

Fig. 52

Diagram labels in Fig. 52:

• Rubber tape
• Drop weight
• Contact (to the chronograph)
• Apparatus wheel

under production conditions, owing to the fact that the “anvil” of the drop hammer at the Dzerzhinsky Plant had clearances in the guides of up to 60 mm. This, of course, ruled out any possibility of bringing the wheel of the instrument into contact with it, since the springs on which it was mounted could allow oscillations of no more than 30 mm.

In describing these instruments and proposing our method for investigating the impact toughness of standard specimens for the factory drop test (1500 mm long) at low temperature, which necessarily ensures fracture, we believe that under more favorable circumstances this method should certainly be implemented.

Of the two described methods for determining the work expended in destroying the specimen, we, on the basis of personal experience, consider the first (with the oscillating lamp) more convenient, since it is almost entirely independent of the accuracy of assembly of the drop hammer, of clearances between the “anvil” and the guides of the frame.

Incidentally, it did not prove possible to apply to the factory drop hammer the method described in his time by Druzhinin (55), precisely because of the presence of clearances between the “anvil” and the guides, since this ruled out any possibility of counting on good contact, or even on the existence of contact at all. For the same reason, other, more modern methods, based in one way or another on electrical contact between the “anvil” and the instrument—for example, a change in the charge of a capacitor, etc.—could not be used.

IV. THE INFLUENCE OF AGING ON COLD BRITTLENESS

During one of the author’s trips to the Kuznetsk Metallurgical Combine, the head of the combine’s central laboratory informed the author of an observed change in the mechanical properties of rail steel over time. The matter was evidently one of aging.

Two varieties of aging may be distinguished: mechanical aging, which appears after plastic deformation, and spontaneous aging, when the properties of an undeformed metal change over time. Since in the present case there was a certain complication in the form of cold straightening on a roller machine, one first had to consider a combination of spontaneous and mechanical aging. During straightening, the rail does not undergo uniform plastic deformations along its entire length; some sections are deformed very strongly, while other places remain almost completely unchanged. A change in mechanical properties over time was also observed in the axle blank, which, in general, is not subjected to any straightening after rolling.

But if such a phenomenon was observed in the case of raw rails, then subsequently, during the sorbitizing of rails, it could hardly have been avoided. On the half-meter rail ends, during quenching on our machine, no especially significant warping was observed; but it certainly cannot be guaranteed that rails 12.5 and even 25 m long will not become distorted during sorbitizing. Obviously, cold straightening of such rails on a roller machine will remain the final operation, and the plastic deformation imparted by it may undoubtedly subsequently cause mechanical aging.

True, straightening could have been carried out in the course of the quenching itself, by quenching according to the Lewis method. Unfortunately, however, this entirely possible method of rail quenching has been studied by us under laboratory conditions and has not been tested at the plant.

The question of the influence of cold straightening on the quality of rails is rather old. As early as 1906 Zhukovskii raised it and came to the conclusion that cold straightening is a very dangerous stage in the technological process of rail manufacture. More recent investigations (57) give some grounds for doubting the harmfulness of cold straightening in general.

Aging of steel was observed at the Kuznetsk Metallurgical Combine under the following circumstances. Tensile testing of specimens left over from old heats, rejected several months earlier because of the low toughness of the metal of various heats, revealed that these heats had been rejected in vain, since the specimens showed quite satisfactory elongation and reduction of cross-section. From Table 18 it is clearly seen that, between the static toughness of specimens tested at the beginning and at the end of 1934 and taken from the same heats, there is a very considerable difference. However, the static strength over this same time changed almost not at all.

Table 18

Nos. of heats and time of testing	Temporary resistance (in kg/mm²)	Relative elongation (in %)	Reduction of cross-section (in %)	Hardness $H_B$
2494 5/II-34	85.5	10.0	16.4	228
2494 16/XII-34	85.0	13.3	24.8	241
4416 30/III-34	82.9	8.7	15.0	217
4416 16/XII-34	85.1	11.3	22.5	248

Measurement of impact toughness revealed an increase of 14–16% after the specimens had lain for 30 days at laboratory temperature. The impact toughness, determined on the Amsler pendulum machine (15 kg·m), immediately after preparation of the specimens ranged from 0.7 to 1.6 kg·m/cm², and after 30 days its value rose, respectively, to 0.8–2.0 kg·m/cm².

Holding the steel at +100° for 24 hours also changed its mechanical properties very strongly, and in precisely the same direction as did aging in the laboratory (Table 19). With such accelerated aging, as is evident from Table 19, all grades of rolled products manufactured in 1934 by the Kuznetsk Metallurgical Combine underwent changes in their properties.

Table 19

Heat Nos. and steel grade	C	Mn	Si	S	P	Before aging: tensile strength (kg/mm²)	Before aging: elongation (%)	Before aging: reduction of area (%)	After aging at 100° (24 h): tensile strength (kg/mm²)	After aging at 100° (24 h): elongation (%)	After aging at 100° (24 h): reduction of area (%)
1064, rail steel	0.55	0.78	0.23	0.03	0.01	80.1	8.6	11.6	80.2	15.0	36.5
1066, axle billet	0.32	0.71	0.24	0.03	0.03	56.1	17.5	27.8	56.2	22.4	47.4
8344, tractor shoe	0.41	0.70	0.27	0.02	0.02	73.9	12.0	16.7	74.0	15.9	28.2

The change in the mechanical properties of all Kuznetsk steels proceeds in one direction: strength increases, and ductility increases considerably.

There is a very extensive literature on the aging of steel, beginning with the work of Bauschinger, who first drew attention to this phenomenon in 1881.

Recently an especially significant number of works devoted to aging has appeared. In most cases the authors distinguish aging after plastic deformation from aging after hardening of steel from low temperatures, and aging of rolled metal that has not been subjected to any effects after rolling.

As applied to rails, these phenomena cannot, in our opinion, be considered separately. The process of manufacturing rails includes, as an indispensable operation, straightening in the cold state, and subsequently some rails also undergo additional bending when being laid on curves. Cold straightening does not deform the entire material of the rail uniformly along its length, but causes stresses that vary in magnitude, since before straightening the rail usually does not have a constant radius of curvature at different points. Therefore, in what follows it will be necessary to touch upon all types of aging of rail steel.

One of the recent works on aging is the review article by Galibourg (58), in which he distinguishes aging of steel after quenching below the point \(A_1\) and aging after cold working. As is known, alloys that exhibit aging phenomena after quenching contain in their structure solid solutions whose concentrations depend on temperature. Owing to the change with temperature in the solubility of such solid solutions, quenching fixes a composition different from the normal one for the given temperature conditions. This necessarily occurs in such cases, and only the insensitivity of the methods of investigation does not always reveal it.

The return to a state of more stable equilibrium occurs as a result of the precipitation of a phase that is more soluble at high temperatures. Sometimes this precipitation process is similar in its consequences to quenching, since the hardness of the alloy increases.

In reviewing a number of works, Galibourg comes to the conclusion that the aging of steel is influenced by carbon, nitrogen, and oxygen, which are precipitated in one form or another from supersaturated solutions in \(\alpha\)-iron. The solubility of carbon in \(\alpha\)-iron is very small and reaches \(0.03\%\) near the point \(A_1\), decreasing to \(0.01\%\) at room temperature. Such a change in solubility as a function of temperature can, of course, give rise to aging phenomena. Batt confirmed that decarburized iron does not age (58). Köster (59) showed that the mechanical properties of steel quenched below the point \(A_1\) changed with time at \(+20^\circ\). Hardness, elastic limit, and tensile strength gradually increased; relative elongation and reduction of area changed in the opposite direction.

Like carbon, nitrogen also has a variable solubility in iron depending on temperature. According to the equilibrium diagram, steel containing more than \(0.015\%\) nitrogen can age after sufficiently rapid cooling. Degassing of specimens in hydrogen at a definite temperature for 4 days destroys the capacity for aging. During aging due to the decomposition of the solid nitrogen–iron solution, strength and hardness increase, while the static and dynamic toughness fall sharply.

Apparently, oxygen also affects aging; its solubility changes (Galiburg) from 0.10% at 1000° to 0% at ordinary temperature. But the data on its action are contradictory (60).

The aging of which Galiburg writes changes the mechanical properties of steel in a somewhat different direction than was observed at the Kuznetsk Metallurgical Combine. As a result of aging, the Kuznetsk steels improved. With the time resistance unchanged, hardness increased and toughness rose. Such a change in properties is favorable for all steels, and especially for rail steel. An increase in hardness with a simultaneous increase in toughness is a circumstance particularly valuable for rails.

The causes producing such aging (61), the workers of the central laboratory of the Kuznetsk Metallurgical Combine considered to be:

1) residual internal stresses of thermal origin, arising during nonuniform cooling of the rolled metal;

2) the presence of alloying elements in the metal;

3) the influence of the gaseous constituent, which is of great importance because of the peculiarities of smelting open-hearth metal in large-capacity furnaces.

Such an explanation, like the influence of carbon, nitrogen, and oxygen considered above, evidently does little to help one understand the essence of the phenomenon.

Incidentally, in some months of 1935, up to 90% of the metal from the Kuznetsk Metallurgical Combine rejected by the NKPS inspector was accepted by him after aging of this same metal.

In one of his works, Kester (62) speaks of a phenomenon partly resembling the one described. In studying the influence of tempering at 100° on slowly cooled Thomas steel (composition: C — 0.09%; Si — 0.01%; Mn — 0.88%; P — 0.053%; S — 0.096%), he found a small increase in relative elongation, of the order of 1%, an increase in reduction of area by 5%, an increase in the number of twists in the torsion test by 45%, and a decrease in strength by 10% (Fig. 53). The ultimate strength after 14 hours of tempering at the indicated temperature decreased from 43 to 38.5 kg/mm², but the elastic limit fluctuated within very small limits. The curves shown in Fig. 53 indicate that the strength of the metal falls, while the plasticity increases somewhat.

Kester attributes all these changes to the precipitation of a nitrogen-containing phase from the solid solution, asserting at the same time that the precipitation of 0.01% N increases the number of twists by 25% and reduces the ultimate strength by 2.3 kg/mm². In Kester’s opinion, the nature of the precipitation of the nitrogen-containing phase, which separates from the solid solution at 100°, corresponds, taking into account certain not very substantial differences, to the conditions of precipitation

in hardened steel of carbon at 250°. Determination of the number of twistings is, according to Kester, a test especially sensitive to the formation and decomposition of solid solutions. It must be noted that phenomena of this kind were observed mainly in Thomas steel.

It is interesting to note the considerable influence of aging at 100° (without preliminary plastic deformation, as before) on impact toughness at low temperatures. The highest value of the impact toughness of the same steel after heating for one month at 100° rose from 9.3 to 13.4 kgm/cm², or increased by 44% (at +20°). At the same time the toughness at low temperatures changed little.

Temporary resistance

Elastic limit

Elongation (in %)

Reduction of cross-section (in %)

Hardness

Number of twistings

Duration of tempering (in hours)

Fig. 53

It seems to us that the following explanation of the differing influence of aging on the mechanical properties is the most plausible. We believe that the change in the properties of the metal in one direction or another depends not only on the very fact of precipitation of some phase from a supersaturated solid solution, but chiefly on the dispersity of the precipitating phase, the size of its particles, their shape, and the character of their arrangement with respect to those crystallographic directions along which slip occurs during plastic deformation. Obviously, there exist certain critical degrees of dispersity which do not very strongly affect the blocking of slip surfaces. At the same time, precipitation of a highly dispersed phase may reduce internal stresses, decrease the distortion of the crystal lattice, unavoidable in the formation of supersaturated solid solutions of the interstitial type. Removal of a certain fraction of such lattice stresses will evidently lead to an increase in static and impact plasticity. With continued aging, the precipitated particles will begin to coagulate and, being arranged along many slip planes, will reduce the possibility of shears, and brittleness will increase. Finally, when further growth of the particles leads to the formation of non-

of a large number of comparatively large crystals, the blocking effect of the latter on the slip planes again diminishes (a smaller number of possible slip planes will be blocked), the toughness will begin to increase, and the static strength to decrease. This will occur with an increase in the aging temperature.

Obviously, the form of the precipitated phase will also play no small role.

At a low aging temperature, coagulation, because of hindered diffusion, will not produce crystals of large size, but part of the lattice stresses will be eliminated, and this will lead to an increase in plasticity.

Kerber and Mehovar (Mehovar) (63) observed a change in the properties of rail steel after it had lain at ordinary temperature for several weeks. They investigated Thomas and Martin rails of approximately the same composition, with a tensile strength of 70–77 kg/mm² and of one and the same profile (Fig. 54). Preliminary experiments established that the condition of the surface of the specimens plays no role in the change of mechanical properties.

As can be seen from Fig. 54, the change in properties proceeded in the same direction as in the steel of the Kuznetsk works. Fig. 54 gives the results of experiments with Thomas steel. The left-hand diagrams refer to rails not straightened after rolling, and the right-hand diagrams to rails straightened on a roller machine. As is evident from the upper diagrams, the tensile strength both of specimens cut from rails and subjected to aging (curves B), and of specimens cut from rails already aged (curves D), is practically almost the same. The fluctuations in the values of tensile strength are small. The same is obtained on heating specimens (curves A, after tempering of the specimen) and rails (curves C₁) after tempering the rail to various temperatures; only heating at 500–600° somewhat lowered the strength. This decrease reaches 5%. The same also applies to the elastic limit. True, the latter does not give such abrupt values as the tensile strength, showing a uniform decrease with increasing duration of soaking. Heating a specimen from a straightened rail to 300° increases the elastic limit from 40.0 to 41.5 kg/mm². Heating to 600° lowers it to 36.0 kg/mm² (curve A).

If the strength under such aging does not undergo any particular fluctuations, then the plasticity—the relative elongation at rupture and the reduction of the transverse section—changes very strongly.

A specimen cut from a straightened rail gives a value of the relative reduction of the transverse section of about 16%; after a month of lying at ordinary temperature it increases to 40% (curve B). Heating acts in the same way (curve A). It may be considered that aging for 5 days is quite replaced by heat-

to 100°; the effect is the same: the value of the reduction of area in both cases is 25%. Aging for 30 days gives the same result as heating to 600°; the reduction of area in the first case is 40%, and in the second, 38%.

Fig. 54

The relative elongation, from 13% in a specimen from a straightened and aged rail (curve D), after 30 days of aging reaches 15.5%. When a specimen taken from a straightened rail is aged, the relative elongation changes from 13 to 17.5% (curve B).

As can be seen from these examples, the mechanical properties of this steel change almost in the same way as those of forged metal. It is more than probable that the causes producing this phenomenon are identical. Kerber at first considered the cause of such aging to be the destruc-

deformation in the lattice with the passage of time or under the influence of an elevated temperature.

Something similar was obtained in work with various steels by Drescher and Schäfer (64). Assuming that the increase in relative elongation and reduction of cross section occurs at the expense of removal from the metal of hydrogen dissolved in it, the content of which in steel decreases with time, they carried out, in addition to tests of ordinary mild steel after heating to various temperatures, also an investigation of steel remelted in a hydrogen atmosphere. The conclusions from their work were confirmed in the ensuing discussion by Körber and Metowar, who presented Table 20.

Table 20

Treatment	Temporary resistance (in kg/mm²)	Elongation (in %)	Reduction of cross section (in %)	Hydrogen content in 1 cm³ per 100 g of steel
Initial state . . . . . .	88.0	9.1	10	2.13
Heating for 1 hour at 200° . . . .	88.7	13.1	20	0.10
After lying for 7 weeks (specimens) . . . .	89.0	12.9	21	0.10
Same (whole rail) . . . .	86.0	12.3	19	0.36

Note: Changes have been made relative to the original—the average data from two tests have been taken.

But, it seems to us, it cannot be asserted unconditionally that the aging of Kuznetsk rails depends only on hydrogen. Undoubtedly, a number of factors not yet taken into account influence the change in properties, such as a change in the content of carbon dissolved in $\alpha$-iron, nitrogen, and, perhaps, certain alloying elements present in very small quantities.

There are a very large number of works on aging after cold working; we have no possibility of citing them and shall limit ourselves to a few which, in our view, are the most interesting.

It is necessary to note Schmidt’s (65) indication of the complete absence of the phenomena of mechanical aging in the case of very pure iron of the following chemical composition: C — 0.001%; Mn — 0.008%; P — 0.0008%; S — 0.0028%; O — 0.0028%; N — 0.0000%. Cold working was imparted to the specimens by changing the height of the cross section from 11 to 10 mm by a hammer blow, and artificial

Natural aging was carried out by heating deformed specimens for 1 hour in an oil bath at 350°. As can be seen from Fig. 55, a considerable increase in the brittleness of the annealed specimen (curve A) is observed at temperatures below −50°, and of the work-hardened one already from −20° (curve B). Aging did not increase the brittleness of the deformed metal (curve C).

Fig. 55

Schmidt also shows that the increase in brittleness after mechanical aging, and especially the decrease in impact toughness of aged specimens at low temperatures, is very strongly affected by oxygen and sulfur, and considerably more weakly by carbon, nitrogen, and phosphorus. As the author himself emphasizes, his conclusions are in some contradiction with the data of other investigators.

Studying heats of steel with varying carbon content, Eilender (66) and his coworkers came to the conclusion that it is nitrogen that increases the tendency of mild steel to mechanical aging, whereas an increase in the carbon content at constant nitrogen content somewhat reduces the consequences of aging. Oxygen has no influence in this direction on mild steel.

In determining the sensitivity to mechanical aging of steel castings (with 0.2% C), Knipp (67) found an especially strong decrease in impact toughness at low temperatures. Fig. 56 shows that at +20° the impact toughness of the aged specimen is on average half that of the annealed one (at 920°, curve A), and at −60° even ten times lower. The greatest increase in brittleness at low temperatures was found in specimens that had lain for twenty-four hours at room temperature after 10% tensile deformation (curve B), and not in those artificially aged by heating at 250° (curve C), as might have been expected.

The impact toughness of tensile-work-hardened specimens of the same steel (not aged) begins to fall noticeably at room temperature only at degrees of deformation greater than 2%,

is reduced by half in comparison with the toughness of the annealed material, while the plastic elongation increases to 9.5%.

In the work already mentioned by us, Kester pointed out a considerable drop in the dynamic toughness of mild steel, chiefly at ordinary temperature, after mechanical aging. The change in impact toughness as a function of the test temperature is presented in Fig. 57. Line A refers to steel normalized and held for 14 days at 100°; the impact toughness of steel that was only normalized is given by line B. Lines C show the changes in impact toughness after aging for 1 hour at 250°. Work-hardening before artificial aging was produced by compressing the specimen by 10% of the height of the cross section.

Fig. 56

Prolonged heating without preliminary deformation of this same steel at 100° for 3, 14, and 28 days led to an increase in toughness at temperatures above 0° and had almost no effect on brittleness at −50° (Fig. 58).

Fig. 57

According to Kester’s supposition, the nitrogen present in technical iron remains, on slow cooling, in a supersaturated solid solution and may gradually precipitate

for a very long time. In his opinion, for the complete removal of all nitrogen at \(100^\circ\), about 14 days are required. The nitrogen-containing phase is precipitated at \(100^\circ\) in the form of small dots, and at higher temperatures—in the form of needles.

It seems to us that Kester is hardly right, since, in elucidating the question, one cannot ignore the influence of other impurities and, as we have already seen, of hydrogen, with such categorical certainty.

According to Walzello (68), the tendency of open-hearth steels to mechanical aging is in very close dependence on the amount of pig iron in the charge, the character of segregation in the ingot, and the size of the ferrite grain in the steel.

Fig. 58

As a result of work with steel containing \(C—0.056—0.073\%\) and \(Mn—0.37—0.41\%\), he came to the conclusion that a reduction in the amount of pig iron in the charge intensifies aging of the steel after cold working, causing a deterioration of toughness and a shift of the transition interval from ductile fracture to brittle fracture toward higher temperatures. The influence of ferrite grain size in steel on the impact toughness of aged steel at all temperatures is especially strong. In Fig. 59 a series of curves is plotted showing the dependence of the impact toughness of steel with different ferrite grain areas on temperature in the initial (curves \(A\)) and aged states (curves \(B\)). It is especially interesting to trace the displacement of the boundary of the onset of cold brittleness with the growth of the ferrite grain. Steel with a grain area of \(583\ \mu^2\) has quite respectable toughness even at \(-40^\circ\); only at \(-60^\circ\) does a fairly considerable impact brittleness appear. An increase of the grain area by a factor of three (to \(1701\ \mu^2\)) shifts the brittle-fracture zone to \(+20^\circ\). At \(+20^\circ\) the toughness of this steel differs scarcely at all from that of fine-grained steel, but at \(-10^\circ\) it amounts to only 12% of its former value. Further coarsening of the grain to \(6020\ \mu^2\) leads to the appearance of abnormal brittleness even in the interval \(+30—40^\circ\).

Mechanical aging acts on this material in a completely exceptional manner. It was carried out, as usual, by heating to \(250^\circ\) after cold working imposed by compression with a deformation of 10%. As can be seen in Fig. 59, aging of the most ductile metal with a medium grain (\(1701\ \mu^2\)) turned it into an extraordinarily-

extraordinarily brittle, having at \(+20^\circ\) an impact toughness of only about \(2\ \mathrm{kgm/cm^2}\); the same steel before work hardening and aging had an impact toughness at \(+20^\circ\) of about \(45\ \mathrm{kgm/cm^2}\). Aging had an especially strong effect on coarse-grained steel; the transition from a ductile fracture to a brittle one shifted to \(+110^\circ,\ +150^\circ\).

Fig. 59

It is evident from Fig. 60 that if the steel after rolling (curve \(A\)) is heated to \(400^\circ\) for 5 hours, this has a favorable effect on the impact toughness (curves \(B\)). This confirms the data of Körber and Megewahr.

Wilhelm and Jahn (69) give an approximately similar picture for steel with a carbon content from 0.325 to 0.435%. In their opinion, increasing the scrap in the charge to 100% makes the steel less prone to aging in comparison with ordinary open-hearth steel.

Stumper (70) obtained very suitable results for Thomas rail steel in his very interesting work (Fig. 61).

As is evident from Fig. 61, the grain size increases rapidly with increasing rail-rolling temperature. At \(950^\circ\) the grain area fluctuates around \(4000\ \mu^2\), while at \(1150^\circ\) it reaches \(24\,000\ \mu^2\). It fluctuates within approximately the same limits as in Walpelya. At \(+25^\circ\) the impact toughness of specimens made from rails rolled at \(850^\circ\) (grain area—about \(4500\ \mu^2\)),—

17.5 kgm/cm², and in the case of rolling at 1150° (grain size—23,000 μ²)—10 kgm/cm². At ±0° the toughness of all the rails is the same; the rolling temperature for some reason has no effect on the toughness at this temperature.

Fig. 60

At −10° the impact toughness of the fine-grained rail is 10 kgm/cm², that of the coarse-grained rail—7.5 kgm/cm² (Fig. 62).

Of very great interest to us is the work of Krüger (Krüger) (71), who investigated the aging process of seven machine-building steels with a carbon content from 0.07 to 0.28%. Annealed steel strips were given tensile deformation of 10%; following such cold working, aging was carried out at 200–300°. Without citing all the steels, we shall dwell only on two; one of them is close in composition to rail steels. The chemical composition is given in Table 21.

Table 21

Steel designation	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)	Chemical composition (in %)
Steel designation	C	Si	Mn	Cu
в	0.11	—	0.48	0.29
48	0.28	0.20	0.80	0.22

Table 22 gives the limits of change in the impact toughness of these steels (in kgm/cm²) after the indicated treatment within the following limits.

Table 22

Steel	After normal annealing	After normal annealing and stretching by 10%	After normal annealing and stretching by 10%
Steel	After normal annealing	After 8 days of natural aging	Artificially aged steel
в	6.29	1.00	0.64
48	4.20	1.34	0.92

Impact toughness was determined on small specimens, measuring \(10 \times 10 \times 55\) mm, with a notch 3 mm deep and a rounding diameter of 2 mm.

Fig. 61

Most interesting in the cited work is the attempt to reveal the influence of aging, carried out after heat treatment of the steel, on its impact toughness at various temperatures (down to \(-80^\circ\)). In Fig. 63, curve \(A\) refers to steel 48, quenched above \(A_{c3}\), tempered to 600° with cooling after tempering in water, then deformed by stretching by 7% and artificially aged. Curve \(B\) belongs to the same steel, normally annealed, deformed by 10%, and likewise artificially aged. Although the results given by these two curves, generally speaking, cannot be compared, since the degree of strain hardening in the two cases is not the same, it is nevertheless quite possible to agree with the author’s assertion that sorbitic steel is less prone to aging in comparison with raw, thermally untreated steel.

Indeed, we can be still more convinced of this if we examine the curves in Fig. 64. The curves in this diagram have the following meaning: \(A\) refers to steel in the initial condition (“as supplied”); \(B\)—to normally annealed steel; \(C\)—to the preceding steel, deformed in tension by 10% and artificially aged; \(D\)—to normally annealed, defor-

deformed by 10% and artificially aged. Curve B in Fig. 63 corresponds to curve D in Fig. 64.

The highest value of toughness according to curve A (see Fig. 63) will be equal to 7 kgm/cm², and according to curve A (see Fig. 64)—6 kgm/cm².

Fig. 62

Aged sorbitic steel has greater toughness than raw steel not subjected to mechanical aging; however, despite the increase in impact toughness of the aged sorbitic

Fig. 63

steel, it possesses cold brittleness to a stronger degree than the raw steel. The beginning of the cold-brittleness interval for the latter lies around +20°, and for the aged steel—around +60°.

In Fig. 65 there are several curves for mild steel. Curve b gives an indication of the impact toughness of quenched

without tempering the steel. The carbon content is low, and therefore the toughness of the metal quenched without tempering is still sufficiently high at low temperatures. Quenching, while somewhat reducing the toughness at \(+50^\circ\), very substantially decreased the brittleness at \(-50^\circ\) in comparison with annealed steel.

Fig. 64

Curve 3 gives an idea of the same steel in the annealed condition. Curve 4 refers to metal quenched and tempered at \(600^\circ\), then subjected to artificial aging after 7% deformation, while curve 1 shows the effect of artificial aging after 10-percent deformation on the toughness of annealed steel. As we see, quenching without tempering of mild steel shifts the transition from ductile fracture to brittle fracture toward lower temperatures, which is obviously very advantageous, while quenching with subsequent tempering makes the steel, compared with the raw steel, practically non-aging.

Steel quenched without tempering, deformed by 5%, and aged proves to be more ductile at all temperatures than simply annealed steel (curve 2). Quenched steel tempered at \(400^\circ\) also gives, after aging (curve 5), very respectable values of impact toughness.

Rötscher and Fink (72) worked with a whole series of structural steels; deformation of the specimens before aging was produced by compression (10%). The carbon content in their metal varied within the limits of 0.04–0.19%; in some of the steels there was a certain amount of copper, reaching 0.60%.

They came to the conclusion that mechanical aging very strongly reduces the impact toughness of the steels investigated under all conditions. Thomas steel, on the basis of their observations, proved, despite its lower manganese content compared with open-hearth steel, to contain

carbon content (0.04 and 0.15%), more cold-brittle in comparison with open-hearth steel, both in the annealed state and after artificial aging.

On the basis of the cited literature material, some general conclusions may be drawn, formulating them approximately as follows: natural aging (without prior, deliberately imposed plastic deformation and without prior quenching) increases the ductility of normally annealed and slowly cooled steel at all testing temperatures.

Artificial and natural aging after plastic deformation (“mechanical aging”) increases the dynamic brittleness of steel, especially at low temperatures, shifts the cold-brittleness interval toward higher temperatures, sometimes by hundreds of degrees (Walzel).

Fig. 65

Heat treatment of structural carbon steels to sorbite reduces their tendency to acquire impact brittleness as a result of aging after work hardening. But just how sorbitization in this case affects the position of the transition from tough fracture to brittle fracture is difficult to say.

Returning to rails as-rolled with a pearlite-ferrite structure, it may be said that properly performed straightening in the cold state on a roller machine probably does not damage the material, since the plastic deformations of the metal in this operation are small. This conclusion is confirmed by experiments of the central laboratory of the Kuznetsk Metallurgical Combine, by Kerber and Megowar, and others.

But as to how sorbitized rails behave with respect to this unavoidable operation, we found no clear indications in the literature, and therefore came to the conclusion that it was necessary to carry out a series of experiments to determine the effect of cold straightening on the properties of heat-treated rail steel.

Wishing to introduce greater practical definiteness into these investigations, we decided to induce in specimens for impact testing, as far as possible, a degree of work hardening of the same order as that observed in rails when passing through a roller machine. Under these de—

deformations is, apparently, nothing other than bending of alternating sign: the rail is bent first to one side and then to the opposite side, alternately, several times. Therefore, deformation was imparted to the Charpy specimens before aging by fourfold bending in a die, followed by straightening between parallel plates under a press.

The die (Fig. 66) had a circular recess in the shape of the specimen. The specimen was placed in this recess, and pressure was applied by the cover of the die. The upper part of the die had the same radius of curvature as the lower part, namely 200 mm. The specimen underwent plastic deformation, turning into an arc of a circle of the same radius. Then the specimen was turned through 180° and again bent in the opposite direction and then straightened under the press. After straightening, the specimen was again bent twice under the press in the die, but this time the plane of bending was at 90° to the original one. The last operation was straightening again under the press.

Fig. 66

The dimensions of the specimen were thereby preserved almost unchanged, and if such an operation changed the shape of the cross section, then in any case only within a few hundredths of a millimeter, as control tests showed. We assumed that this kind of cold working is fully equivalent to the sum of those deformations which a rail experiences when passing through a roller-straightening machine. It is necessary to note the assertion of some authors (73) that, as a result of alternating stresses in the rail beyond the elastic limit during straightening on a roller machine, the cold working acquired by it in this case differs from cold working caused by deformation of one sign.

The radius of the circle described by the surface of the die was found from the following considerations. During straightening on a roller machine, the rail usually experiences bending in one plane, but in two mutually opposite directions. In the machine of the rail-straightening shop of the Kuznetsk Metallurgical Combine, 7 rollers are installed, and therefore the part of the rail located—

between the two middle rollers, may be regarded as a beam with fixed ends, loaded by a concentrated force at the center. It is rather difficult to determine this force, since the deformation it produces will consist of a plastic and an elastic part; but, with some error, the formulas of the theory of elasticity may be applied to determine the smallest radius of curvature of the bent rail. Knowing the radius of curvature, one can find the magnitude of the relative deformation of the fiber farthest from the neutral axis. In the roller-straightening machine of the Kuznetsk Metallurgical Combine, the smallest distance between the axes of the rollers is 1220 mm, and the maximum deflection caused by the pressure of the roller located at the center is about 6 mm.

Having determined the equation of the elastic line for this case, it is easy to find the value of the radius of curvature \((r)\) at the middle of the beam, at the place where the curvature will obviously be greatest.

We obtain:

\[ r = \frac{l^2 + 16 f^2}{32 f}, \]

where \(l\) is the distance between supports, \(f\) is the deflection. For the case cited by us we find: \(r = 76\,000\) mm.

The relative elongation of the stretched fiber, for example of the rolling surface of the head, and the compression, equal to it in magnitude, of the opposite part of the base, is found from the formula:

\[ \lambda = \frac{r_2 - r_3}{2r_1} = 100\%, \]

where \(r_1\) is the radius of curvature of the neutral axis of the beam, and \(r_2\) and \(r_3\) are the radii of curvature of the stretched and compressed fibers, respectively.

Let us determine \(\lambda\) for the most common rail section II—A:

\[ \lambda = \frac{140}{2 \cdot 76000} = 0.92\%, \]

The maximum relative elongation, as well as the compression, is equal to 0.92%. This value includes both plastic and elastic deformation.

If the temporary resistance of hard rail steel is about \(8000\ \text{kg}/\text{cm}^2\), then the elastic elongation corresponding to the elastic limit \((6000\ \text{kg}/\text{cm}^2)\) is found from Hooke’s law:

\[ i = \frac{\sigma}{E} \simeq 0.3\%, \]

and, thus, about 0.6% remains as the share of plastic deformation.

It is necessary once again to emphasize the approximate nature of these calculations, for although the deformations in question are small, they nevertheless lie beyond the elastic limit, where the formulas of the theory of elasticity cannot be applied. Direct measurements, however,

also gave values agreeing with the calculated values to tenths of a percent.

With a radius of the circumference whose arc defines the die equal to 200 mm, a specimen deformed in the die undergoes relative compression and elongation of the opposite sides of the profile of the order of:

\[ \lambda=\frac{202.5-197.5}{2\cdot200}=1.25\%. \]

The residual deformations will obviously be approximately 0.9%.

We took a deformation one and a half times greater than in reality, in order to reveal more strongly the effect of straightening on the course of aging, and also because in some cases, when straightening a strongly bent rail, such a magnitude of deformation is quite possible.

D’yakov (74) carried out a series of experiments to determine the effect of deformation of this kind on the course of aging of sorbitic rail steel from the Kuznetsk Metallurgical Combine, of the following composition: C—0.51%; Mn—0.82%; Si—0.31%; P + S < 0.07%. The tensile strength of this steel was 79.6 kg/mm².

From the rail head, bars of cross section 6 × 6 mm were forged. After the same annealing as in all our preceding experiments, the specimens were quenched from 900° in oil (+20°) and tempered for 2 hours at 650°.

A preliminary experiment showed that, owing to the small dimensions of the specimens, cooling in oil and in air after tempering gave the same impact toughness at ordinary and low temperatures. In the course of the heat treatment of this steel, the measures already described for reducing oxidation of the specimens were applied.

After sorbitizing, the specimens were brought to dimensions of 5 × 5 × 50 mm, deformed by four bendings in the die, and straightened, as described above.

Aging was carried out by heating for half an hour in an oil bath at 250°. The temperature was carefully monitored, and the oil was continuously stirred.

We chose 250° for artificial aging as an almost standard temperature used for this purpose by many authors. Under the action of this temperature, according to the literature, the maximum change in the mechanical properties occurs in work-hardened steel.

Natural aging took place while the specimens lay in the laboratory at ordinary temperature (from 16 to 22°) for three months.

The small dimensions of the specimens raised some doubts that some defect might accidentally be present in the dangerous section of the specimen, the presence of which would distort the average result,

if a small number of specimens is taken for each point of the temperature—impact-toughness curve. In view of this, we took 6 specimens for each point at temperatures \(+20\), \(-40\), and \(-183^\circ\). We selected only these 3 points, assuming that in the present case one could hardly expect any anomalies in the course of the impact-toughness—temperature curve. Despite our apprehensions, the scatter of the points proved to be very slight, as is seen from Table 23 (for two temperatures). The data of this table refer to sorbitic, artificially aged specimens.

Table 23

Test temperature (in °C)	Impact toughness (in kgm/cm²)	Mean value of impact toughness (in kgm/cm²)	Test temperature (in °C)	Impact toughness (in kgm/cm²)	Mean value of impact toughness (in kgm/cm²)
+20	2,33	2,51	−182	0,33	0,27
+20	2,52	2,51	−182	0,31	0,27
+20	2,45	2,51	−182	0,31	0,27
+20	2,56	2,51	−182	0,36	0,27
+20	2,71	2,51	−182	0,09	0,27
+20	2,60	2,51	−182	0,21	0,27

The results of all the tests are given in Fig. 67.

As can be seen, sorbitic steel at \(-40^\circ\), irrespective of aging, gave an impact toughness six times lower than at \(+20^\circ\). This indicates the existence of a critical interval of cold brittleness around \(-40^\circ\).

Broken line \(A\) gives an idea of the change in impact toughness of sorbitic, deformed steel and steel naturally aged for three months. Line \(B\) gives the same dependence for artificially aged steel (\(1/2\) hour at \(250^\circ\)), while line \(C\) refers to control specimens which, after sorbitizing but without prior cold working, had lain for 3 months in the laboratory.

The static strength (in tension) of artificially aged rail steel, previously sorbitized under the former heat-treatment conditions, differed hardly at all from the strength of steel only sorbitized and not aged. Thus, the average of several tests gave an ultimate tensile strength for the unaged steel at \(-180^\circ\) of \(83.5\ \text{kg}/\text{mm}^2\) and a relative elongation of \(2.6\%\); for the aged steel, \(82.8\ \text{kg}/\text{mm}^2\), and an elongation of \(2.8\%\).

It would seem possible to acknowledge the absence of a harmful influence

on the quality of rail steel of ordinary normal straightening in the cold state after sorbitizing. Of course, in the presence of slag interlayers and remnants of shrinkage cavities, in the case of straightening the heads of Bessemer rails, a deformation of even 0.5% will prove fatal and the rail may develop cracks; but such rails should not be allowed onto railways at all.

It appears to us possible, without any danger of spoiling sorbitic rails or lowering their toughness at normal and low temperatures, to straighten them cold on a roller machine.

Fig. 67.

The conclusions from the work cited were confirmed by Skobennikov (75). He deformed specimens of sorbitized steel of approximately the same composition by tension, imposing deformations of 5 and 7%. It turned out that ageing of sorbitic steel after such cold work has no noticeable influence on impact toughness at reduced testing temperatures.

As in the preceding work, Skobennikov used specimens whose final size was \(5 \times 5 \times 50\) mm.

Tensile deformation in his experiments had a number of advantages. Tension beyond the elastic limit guaranteed a more uniform stress over the whole cross-section of the specimen than bending or compression. In compression, to say nothing of bending, the deformation will not be uniform over the cross-section of the specimen, since, with the small height of the upset specimen, the magnitude of plastic deformation is affected by the action of the planes between which compression takes place.

According to Knipp’s data (76), in deforming steel of medium hardness, a residual deformation of 1.5% still does not cause a decrease in impact toughness at ordinary temperature (Fig. 68). Only an elongation of about 2% gives an increase in brittleness.

An increase in plastic deformation to 5% gives a decrease in impact toughness by \(1.7\ \text{kg}\ \text{m}/\text{cm}^2\), and with elongation of

at 7% the impact toughness reaches 9.5 kg m/cm². On this basis a deformation within the limits of 5–7% was chosen.

Fig. 68.

As can be seen from Fig. 68, a deformation of 10% reduces the impact toughness by more than half. In straightening rails under normal conditions, such deformations are not encountered. Such deformation would lead to the use of almost the entire reserve of static toughness possessed by the rail metal.

We think that the results obtained in experiments on the aging of sorbitic steel could have been anticipated.

During sorbitization the temperature regime of the process is such that one may expect a very good equalization of internal stresses and coagulation of the crystals precipitated during the decomposition of the aging solid solution. This should be especially promoted by prolonged tempering at 600–680°.

The crystals precipitated at this time—the products of decomposition of the solid solution—will probably serve, upon a subsequent lowering of the temperature, as centers of crystallization for the newly precipitating portions of the products of decomposition of the aging solid solution.

It is necessary here to bear in mind that the results of the two latest works, in their numerical expression, cannot be compared with our investigations, which were carried out on specimens for a Charpy machine of medium size. The numerical results are affected by the size of the specimen and the size of the notch. As has already been mentioned, the specimens had a net section of 4 × 5 mm.

Of course, this in no way diminishes the value of the results of the investigations cited, as we showed at the beginning when discussing methods for determining impact toughness.

Changing the size of the specimen did, of course, affect the position of the transition point from ductile fracture to brittle fracture on the temperature axis, and did not affect the character of the relationship impact toughness—temperature.

On the basis of what has been set forth, it may be concluded that straightening on a roller machine, when it is operating normally and when the bullhead rails being straightened are not too severely bent, should not have a harmful effect on the mechanical properties of bullhead rails or lead to the appearance of particular brittleness, as has sometimes been asserted.

What, then, is the essence of the changes that occur in rail steel during aging? We studied very carefully the structures of bullhead steel mechanically aged under various conditions, but were unable to draw indisputable conclusions. In the concluding part of his work (71), already cited by us, Kruger attempts to make certain generalizations from his microscopic investigations in this field. We could not draw indisputable conclusions in this part of our investigation, since after aging at comparatively small degrees of deformation we did not obtain the structural changes that Kruger observed.

We then attempted to determine the change in the lattice parameter of $\alpha$-iron in bullhead steel during aging. Although Debye-method photographs were taken by us with a large Siemens—Reiniger camera with a cylindrical cassette diameter of 114 mm and were measured carefully, we detected no changes in the parameter. In this case, specimens in the form of cylinders 0.6–0.7 mm in diameter were used. They were turned on a lathe, heat-treated to sorbite, deformed, and subjected to aging.

The problem of the aging of steel, despite its undoubted importance, had only secondary significance for us in the present work. Therefore, naturally, in this chapter we could not give an exhaustive survey of it or discuss it in all details. But it should be of interest to those concerned with rail steel.

It seems to us that, first of all, it is necessary to take up the question of the aging of raw rail steel as a result of work hardening on the running surface.

The diversity and contradictory nature of some results of certain investigations, and the presence of factors whose influence the designer and the consumer of steel have very limited ability to take into account—for example, the influence on the character of aging of the scrap content in the charge of the open-hearth furnace—indicate the obscurity of many phenomena and the poor substantiation...

the correctness of some of the conclusions and insufficient consideration of certain circumstances.

We should like to draw attention to one such point, not taken into account in many works. After all, the aging process is strongly influenced by the deoxidation conditions and, consequently, by the grain size in steel as determined by the Ehn method. As Gerty (77) showed, well-deoxidized steel changes its impact toughness only very slightly after aging.

The grain size of steel, determined by the Ehn method, in Dyakov’s experiments varied within the limits of 8–12 grains per 1 dm² with a 100-fold magnification. Such steel, as we have already seen, did not give, at small degrees of preliminary deformation, any noticeable change in mechanical properties after aging.

Apparently, the size of the austenitic grain in steel has a great influence on the course of the aging process only when the steel has undergone considerable deformation.

It seems possible to us to bring together a number of phenomena which at first sight are quite dissimilar: natural aging, mechanical aging, and the brittleness of steel on tempering. Probably at the basis of all these processes lies one common phenomenon—the decomposition of a solid solution which has become supersaturated at the given temperature and which separates out from itself some crystalline phase. This phase may remain unchanged in composition throughout the entire aging period, but it may change its chemical composition at certain moments. Thus, Bischof (78) proves the possibility of the transformation of carbides precipitated during tempering into others poorer, for example, in manganese and chromium.

Deformation of any kind, by affecting the elastic changes of the metal lattice, will evidently cause a displacement of the solubility limits of the components in the given solid solution.

Experiments (79) to determine the influence of external pressure on the course of aging processes confirm this opinion to some extent.

All the properties of the metal after aging will be in very close dependence not only on the chemical composition of the phase precipitating from the supersaturated solution, but also on the dispersity of its particles and on the site of its precipitation in the structure of the alloy.

Solid solutions with limited solubility are constructed according to the principle of insertion into the lattice of the solvent of atoms of the additive (the solute). Crystals of the phase containing atoms of the solute, separating within the crystals of the base metal, will occupy a definite position with respect to the principal crystallographic directions of the latter. As we showed in one of our works (80), the nitrogen-containing phase in the iron–nitrogen alloy, the so-called “nitride needles,” is precipitated in the iron crystal along the edges of the rhombic dodecahedron. A similar phenomenon is probably...

has a general character. We think that any crystalline phase precipitating from a supersaturated solid solution must stand in some crystallographic relation to the lattice of the solvent and precipitate along definite crystalline surfaces.

The influence of a phase precipitated in this way on the mechanical properties of the metal will depend on the extent to which it impedes the course of plastic deformation, the intracrystalline slips in the grain of the alloy. It is, of course, impossible to speak of the purely mechanical action of such a segregate (in Hanemann’s expression), representing it as keys in a composite beam, which prevent the parts of such a beam that are being joined from sliding relative to one another.

Its action will depend very strongly on the size of the individual crystals, and not on the total mass of the precipitated segregate. So long as it is a highly dispersed phase located on many slip surfaces of the crystal, it can, by its presence, exert a deforming influence on neighboring sites of the crystal lattice of the base metal. As the crystals of this phase grow, the number of deformed sites in the parent crystal will decrease, and the stress of the lattice will gradually disappear. Therefore, beginning at some moment, the two-phase system that has formed (base metal and segregate) may be regarded as a mechanical, more or less uniform mixture. The analogy with keys blocking some small portion of all possible slip planes can now be accepted. As a consequence of the freeing, from blocking crystals of the segregate, of a considerable part of the initially blocked slip planes, the viscosity of the material increases.

The quantity of the precipitated phase will be determined by the change in its solubility in the base metal at the given temperature difference and by the existing concentration gradient at the point under consideration. In addition, the solubility will be affected by the degree of plastic deformation of the metal, in the sense of the elastic deformation of the lattice caused by it.

The optimal aging temperature and the appropriate concentration of the supersaturated solid solution can produce precipitates in the crystal of such a dispersity, and blocking the slip surfaces so much (retardation of slip due to mechanical obstruction and as a result of local distortion of the spatial lattice, bending the slip paths), that the maximum reduction of the possible magnitude of plastic deformation of the entire metallic aggregate and a special increase in impact brittleness are obtained.

Raising the aging temperature beyond a certain temperature limit, when the precipitated crystals rapidly dis—

here, or the precipitation of an impurity that distorts the lattice and is included in the crystals of the base metal, in the form of a gas (hydrogen), will improve toughness because an increase in the size of the crystals of the precipitated phase in the first case will entail the relief of part of the internal stresses and a decrease in the number of blocked slip surfaces. In the second case, the dissolved gas will simply leave the metal for the atmosphere.

Which impurities cause these phenomena is, in my opinion, at present quite difficult to say with certainty. It may be assumed that these are—first of all, in rail and structural carbon steel in general—hydrogen and nitrogen, and then, more probably, carbon, phosphorus, and oxygen. The possibility is not excluded of the influence on Kuznetsk steel of such impurities as zinc—an inevitable companion of many Kuznetsk ores—or arsenic, which contaminates Kerch metal. The question, despite the great interest in it, has as yet been too little studied.

V. THE INFLUENCE OF CERTAIN ELEMENTS ON THE COLD BRITTLENESS OF STEEL

In Schmidt’s work already cited (65), there is an attempt to clarify the influence on the cold brittleness of steel—or, more precisely, of very soft iron—of such elements as oxygen, sulfur, phosphorus, and nitrogen. The author took very carefully prepared iron, in which the carbon content was kept within the limits of 0.001–0.0017%, and added to it the above-mentioned elements.

With the minimum oxygen content attainable under the conditions of Schmidt’s experiments, 0.0028%, the toughness of the metal remains at an approximately constant level down to −50° (Fig. 69). An increase in the oxygen content to 0.112% causes, at 0°, a decrease in impact toughness from 17 to 11 kgm/cm², i.e., by almost 30%. It is interesting that at −80° oxygen has absolutely no effect on the impact toughness of steel, which at this temperature has a value quite considerable in comparison with +50°.

Fig. 69

The influence of phosphorus on the cold brittleness of very soft steel is shown in Fig. 70. The upper curve should be taken into account

hardly has to be considered,—material with such a low content of this element is encountered extremely rarely. More realistic will be the toughness values for steel with 0.061% phosphorus (the middle curve). As we see, increasing the phosphorus content above all permissible limits does not very greatly increase the brittleness of mild steel at ordinary temperature, and still less at \(-80^\circ\). The impact toughness of steel containing 0.061% P at \(-80^\circ\) will be equal to \(3\ \text{kgm}/\text{cm}^2\), and with a content of 0.124% P—

Fig. 70

\(2.5\ \text{kgm}/\text{cm}^2\). Of course, the picture will change with a simultaneous increase in the carbon content.

Nitrogen very strongly increases the cold brittleness of mild steel. The toughness at \(-80^\circ\) is reduced by half when the nitrogen content is increased from 0.0000 to 0.0121% (Fig. 71).

It has almost no effect on the cold brittleness of such very pure iron: sulfur.

Fig. 71

Carbon has a very strong effect on impact toughness (Fig. 72), so long as its content is not large. The impact toughness of iron practically free of carbon (0.001% C) at \(0^\circ\) is \(17.2\ \text{kgm}/\text{cm}^2\), whereas when the carbon content is increased to 0.11% it falls to \(12.2\ \text{kgm}/\text{cm}^2\), i.e. by almost 30%. A further increase in the carbon content by more than four times (0.54%) reduces the impact toughness at the same temperature only to \(9\ \text{kgm}/\text{cm}^2\), i.e. by only 25%. In Fig. 72 it is evident that at \(-80^\circ\) carbon practically does not affect the cold brittleness of steel.

Schmidt’s experiments, judging by the description given by the author, were carried out very carefully. He worked with an exceptionally pure metal. The conclusion that suggests itself is that all the unavoidable impurities in technical steel increase its brittleness. In welded seams there is especially much nitrogen and oxygen, and in their carbon content they approach Schmidt’s metal, and, in our opinion, this may explain their particularly severe cold brittleness.

Fig. 72

Despite the great value of this investigation, it is nevertheless difficult to transfer its results directly to ordinary steel, owing to the considerably larger amount in it of all impurities, both those taken into account and those not taken into account (for example manganese and silicon) by the author.

To clarify the influence of carbon content on the toughness of ordinary commercial steel, containing normal impurities, at low temperatures, we, together with K. A. Bessonov and P. S. Grdin (81), decided to investigate steel with the minimum possible carbon content, with 0.45% carbon, and eutectoid steel.

Obtaining steels with different carbon contents within the indicated limits and with a constant content of all the other impurities (manganese, silicon, phosphorus, sulfur, etc.) is extremely difficult. Therefore we decided to take rail steel, reduce its carbon content to a minimum by decarburization, and bring its content up to 0.9% by cementation.

As the starting material, open-hearth steel of the following composition was taken: C — 0.43%; Mn — 0.14%; S — 0.014%; P — 0.089%.

The rail head was rolled at a temperature of 900–1050° to a size of \(6 \times 6\) mm. The square bars obtained were decarburized in an atmosphere of moist hydrogen at a temperature of 870° for 120–140 hours (82). For this purpose the specimens were placed in a glazed porcelain tube in an electric furnace. Chemically pure hydrogen (from Kahlbaum) was passed through the tube, saturated with water vapor by passing through a Winkler bottle filled with hot (95°) water. In order to

to avoid condensation of steam on the way to the furnace; the connecting tubes were heated by nichrome coils.

After such treatment the carbon content in the steel was reduced to 0.022%; the content of the other impurities changed hardly at all. Analysis of the steel gave the following results: C — 0.022%; Mn — 1.09%; P — 0.086%; S — 0.023%. The increase in the sulfur content is explained by the difference in the places where the samples were taken. The most important circumstance, which was not taken into account by Schmidt, in our opinion, is the constancy of the amount of nonmetallic inclusions and, what is still more important, the invariability of their distribution and shape in comparison with the original metal. Hydrogen had a very weak effect on the majority of slag inclusions; they remained in their places and did not change in shape or size.

To remove the hydrogen dissolved in the metal, the specimens were annealed in vacuum at a temperature of 1000° for 10 hours, slowly cooled in vacuum, and held for 20 days at ordinary temperature.

As Messkin (83) indicates, the hydrogen present in steel is removed to a considerable extent when the metal is heated for 4–6 hours at 800°; complete removal is possible only by annealing in vacuum.

According to the data of Luckmeyer and Schenk (84), especially strong evolution of hydrogen from iron is observed at the point \(Ar_3\), owing to the different solubility of the gas in \(\alpha\)-iron and \(\gamma\)-iron.

On the basis of these data we assumed that the treatment adopted by us for specimens decarburized in hydrogen would lead to the removal of all dissolved hydrogen.

To obtain eutectoid steel, forged rods were subjected to cementation with charcoal containing 30% barium carbonate. Cementation was carried out for 6 hours at 1000°. The thermal regime and the composition of the mixture were selected after a series of experiments. A microsection of the transverse section of a specimen after cementation clearly showed the three layers usual in such cases. After additional annealing for 12 hours in the range 1000–1050°, as a result of the diffusion that occurred, the microsection of the steel presented a uniform pearlitic field, and the carbon content at all points of the section, according to analysis data, fluctuated around 0.82%. To prevent decarburization during annealing, the specimens were placed in iron boxes and covered with a mixture of quartz sand and charcoal. Analysis of the specimens after cementation gave the following content of constituents: C — 0.83%; Mn — 1.12%; P — 0.088%; S — 0.015%.

After annealing, all specimens were normalized by cooling in air from 800°.

Subsequently we wished to obtain large single crystals, but unfortunately this could not be done—the maximum grain area obtained in the recrystallization of our decarburized ...

of carburized metal, fluctuated around 1 mm². In the study of coarse- and fine-crystalline specimens it became clear that the cold brittleness of soft iron is manifested the more strongly, the coarser the grain.

To obtain single-crystal specimens we took Soviet Armco iron, kindly sent to us by V. A. Mozharov (Kharkov Institute of Metals).

Coarse-crystalline and single-crystal specimens of Armco iron were prepared by rolling on a laboratory mill, followed by stretching by 3–5% and annealing. The maximum dimensions of the single crystals thereby obtained were 5.5 × 5.5 × 65 mm. In the machined state the specimens had dimensions of 4.5 × 4.5 × 45 mm, with a triangular notch 1 mm deep, an angle of 45°, and a rounding radius of 0.25 mm. They were tested on a pendulum impact machine of 0.5 kgm capacity.

The test results are given in Fig. 73. Single-crystal specimens of Armco iron (line A) gave a very considerable toughness, whereas specimens with an average grain size (with an area on the microsection of about 1 mm²) showed significant brittleness (line B). The single crystals gave, on the average, at −20° an impact toughness of about 6 kgm/cm², while the coarse-crystalline specimens gave only 0.6 kgm/cm²; the difference was tenfold. At −100° the toughness of all specimens became equal, falling to 0.3 kgm/cm². It is interesting that at −60° the single crystals still showed a very considerable toughness—2.4 kgm/cm². Some single-crystal specimens tested without a notch on a Charpy pendulum, at the temperature of liquid air, did not break, but only bent. This is at variance with the data of some authors (85) on the extreme brittleness of iron single crystals at the temperature of liquid air.

Fig. 73

The lateral faces of single-crystal specimens fractured on the pendulum impact machine were etched with Fry’s reagent; this clearly revealed slip lines. By the method of pressing a needle into the lateral

on the surface of the specimen by Tammann’s method, it was found that in all the single crystals the cube plane lies almost perpendicular to the axis of the specimen.

Fracture of a single crystal at the temperature of liquid air proceeded along a well-defined plane. Parallel to this plane a plate 0.1 mm thick was cut out and examined by Laue’s method. Study of the X-ray photograph showed that the cleavage plane coincides with the \([001]\) plane, i.e., is a face of the cube.

The number of slip lines revealed by etching by Fry’s method on the grains of coarse-crystalline specimens broken at room temperature was considerably greater than in the grains of similar specimens tested at the temperature of liquid air. Since such etching reveals not only slip lines but also twins, it may be asserted that the number of twins formed also increases with increasing test temperature.

Rail steel and eutectoid (case-hardened) steel were tested in the normalized condition and in the sorbitized form at temperatures \(+20\), \(-20\), \(-60\), \(-200\), and \(-180^\circ\). For each point, as in the preceding case, 2–3 specimens were taken, unless there were large discrepancies in the values of impact toughness. In this case, control specimens were taken.

The heat treatment to sorbite consisted of hardening and tempering at 620 and \(670^\circ\) (Fig. 74).

Fig. 74

As is evident from Fig. 74, the greatest toughness at room temperature is possessed by raw steel with a content of 0.45% C (line 1, tempering at \(670^\circ\); line 2, tempering at \(620^\circ\); line 5, annealed steel), and the lowest, as was to be expected, by steel with 0.9% C (line 3, tempering at \(670^\circ\); line 4, tempering at \(620^\circ\); line 6, annealed steel). True, the absolute magnitude of the toughness at all test temperatures is small, and at \(-60^\circ\) both steels are equally brittle.

Sorbitizing very greatly increases the impact toughness in

in the temperature interval from \(+20^\circ\) to \(-60^\circ\). Steel with a content of \(0.45\%\) C, quenched from \(900^\circ\) and tempered at \(670^\circ\), gives a very high value of impact toughness, equal at \(+20^\circ\) to \(6\ \text{kgm}/\text{cm}^2\). Steel treated in this way retains, even at \(-60^\circ\), a quite satisfactory impact-toughness value of the order of \(1.7\ \text{kgm}/\text{cm}^2\), and only at \(-100^\circ\) does this value fall to \(0.7\ \text{kgm}/\text{cm}^2\).

At \(-60^\circ\), sorbitized eutectoid steel differs little from sorbitic steel with a content of \(0.45\%\) C. It may apparently be asserted that a carbon content above a certain limit has little effect on the toughness of sorbitic steel. A change in the tempering temperature in any case has a stronger effect on the change in toughness than fluctuations in the carbon content by several tenths of a percent. On the basis of these experiments it may be considered that sorbitization reduces the brittleness of steels at low temperatures independently of the carbon content (of course, within certain limits, beginning, apparently, with \(0.4\%\) C).

The question of the considerable toughness of single-crystal specimens requires additional discussion.

As was mentioned at the beginning of this work, we were interested chiefly in structural carbon steel and, as a particular case, rail steel. But in working with ordinary steel we became convinced that carbon steel, in any structural state, at a sufficiently low temperature is deprived of toughness; brittleness at low temperatures can be shifted toward lower temperatures by means of heat treatment, but it cannot be eliminated altogether. At the same time, a whole series of alloys is known which are not brittle at the boiling temperature of liquid air and even below it.

There are many special steels that are much less susceptible to impact at frost than ordinary carbon steel.

Like simple carbon steel, alloy steels increase their static strength as the test temperature is lowered. The ultimate strength increases, and the elastic limit rises.

Manganese is already such an element which, while increasing the strength of steels at all temperatures, at the same time weakens cold brittleness.

In Fig. 75 are given curves (86) showing the influence of manganese in this direction. The upper curves refer to normalized steels, and the lower ones to annealed steels. As is seen from comparison of the left (Si — 0.07, C — 0.25, Mn — 0.49%) and right (Si — 0.06, C — 0.25, Mn — 1.24%) diagrams, the influence of manganese is manifested most strongly in steels in the normalized state. The toughness of both steels at ordinary temperature (\(+20^\circ\)) is practically identical. The impact toughness of normalized specimens fluctuates around \(30\ \text{kgm}/\text{cm}^2\), while that of simply annealed ones—

about 20 kgm/cm². At test temperatures below \(+20^\circ\), the influence of manganese begins to appear more markedly. At \(-40^\circ\), steel with a manganese content of 0.49%, regardless of the indicated heat treatment, shows an impact toughness of about 3 kgm/cm² (limits of variation \(-2.21\)—3.59 kgm/cm²), whereas annealed steel with a manganese content of 1.24% at \(-40^\circ\) gives only a slightly more significant toughness. But the same steel, normalized, shows at \(-40^\circ\) an impact toughness of about 15 kgm/cm². At \(-50^\circ\), the impact toughness of both steels under consideration is the same.

Vertical axis: Impact toughness (in kgm/cm²).
Horizontal axis: Temperature (in °C).

Fig. 75

Obviously, such an action of manganese can be explained by its influence on the size of the austenite grain in steel, and also on the structure of pearlite. Sorbite-like pearlite in manganese steels is known to be distinguished by high toughness.

As we mentioned, fine-grained steel is less prone to cold brittleness than coarse-grained steel, and therefore it should be expected that alloying elements which reduce the size of the austenite grain will increase the impact toughness of steel at low temperatures. Such additions include, as is known, titanium, zirconium, vanadium, and aluminum.

Etan points out that it is precisely zirconium steels that possess very considerable impact toughness at low temperatures (see Fig. 38).

Let us cite several data from the work of him and his collaborators (87). In Table 25 the Izod toughness is given for certain steels in the normalized state at various test temperatures. The specimen is the standard Izod specimen.

Table 25

Grade of steel	Izod toughness at temperature (in °C)	Izod toughness at temperature (in °C)	Izod toughness at temperature (in °C)	Izod toughness at temperature (in °C)
	+20°	−50°	−80°	−183°
Armco iron . . . . . . .	64	9	3	—
SAE 1020 (0.20% C) . . . .	65	7	2	—
SAE 1030 (0.30% C) . . . .	49	9	5	—
0.16% C; 0.66% Mn . . .	87	71	34	2.5
0.07% C; 0.20% Zr . . . .	70	78	11	2.3

From this table it is evident that zirconium does indeed significantly reduce the cold brittleness of steel. Zirconium steel with 0.20% Zr and 0.07% C, approaching Armco iron in carbon content, has at −50° a toughness considerably exceeding the impact toughness of Armco iron at ordinary temperature. We think that the specific action of zirconium can be explained by its influence on increasing the number of centers of crystallization of austenite. The addition of zirconium affects the size of the austenite grain in the same way as the addition of aluminum.

There are many literature data on the beneficial influence of nickel in reducing the cold brittleness of steel.

According to Morrison (88), and also Aldridge and Shepherd (89), a comparatively cheap medium-carbon (0.15–0.20% C) nickel steel with a nickel content of 2.0–2.5% has sufficient impact toughness in the interval from 38° to −59°. Normalizing such steel further increases the impact toughness in the indicated temperature range.

In order to be able to compare our data, obtained in testing many hundreds of specimens of various carbon steels, with the character of the cold brittleness of alloy steels, we investigated the two most common grades of chromium-nickel steels from the Elektrostal plant (90). Table 26 gives the plant data on the chemical composition of steels E6 and E10 and the results of analysis of the specimens in our possession, carried out by the chemical laboratory of the Siberian Institute of Metals.

The steel was obtained from the plant in round bars (diameters 18 and 40 mm) and was then forged into a square of 12×12 mm. The forging was carried out, as with rail steel, under a pneumatic hammer in the temperature interval 1050–850°. The temperature was carefully controlled throughout the work. After heat treatment, the specimens were brought on a milling machine to the final dimensions (10×10×100 mm) and were provided with tre-

Table 26

Steel grade	C	Ni	Cr	Mn	Si	V	S	P	Application
36	0.25–0.33	3.30–4.00	0.70–1.10	0.25–0.50	0.40	0.15	0.015	0.030	Factory data
36	0.34	3.45	0.95	0.44	0.40	—	0.014	—	SIM analysis
Э10	0.33–0.41	3.00–3.70	1.20–1.60	0.25–0.50	0.40	0.15	0.015	0.030	Factory data
Э10	0.48	3.33	1.40	0.45	0.23	—	0.016	0.01	SIM analysis

Table 27

Steel grade	Hardness by Brinell: in the initial state	Hardness by Brinell: after forging	Hardness by Brinell: after annealing
36	166	255	187
Э10	228	311	270

with an angular notch of the same shape as in the preceding experiments. The tensile-test specimens were made from the same forged bars. The diameter of the tensile specimens was 10 mm, and the length of the cylindrical part was 80 mm. Their remaining dimensions were maintained according to the German standard.

As control experiments showed, the specified allowance of 1 mm per side was quite sufficient to protect the specimen, within the main outlines of the cross section \((10 \times 10\ \text{mm})\), from the slightest oxidation. With holding times during heat treatment equal to triple the normal values, decarburization penetrated to a depth of several tenths of a millimeter, so that during machining the altered layer was completely removed from the specimen.

Specimens of E10 steel tempered at low temperatures had to be finished by grinding, since the milling cutter took them with difficulty. The grinding allowance was left at 0.1 mm. Therefore specimens of this steel were protected from oxidation by a special coating, the composition of which was determined experimentally (91). Its approximate composition was: white clay, borax, soda, and water glass. It protected the specimens well from furnace gases and was easily removed after quenching.

After forging, specimens of E6 and E10 steel were heated to 780°, held at this temperature for 20 min, and cooled together with the furnace at a rate of about 4° per minute in the range 780–300°. The hardness of the specimens changed in this case, as is evident from Table 27.

As is known, chromium-nickel steels are used exclusively in the quenched and tempered-to-sorbite condition. We selected the following heat treatment: quenching from the customary temperature—820° (in oil)—and tempering at 350, 500, and 650°. The mechanical properties of these steels change with the tempering temperature more or less smoothly, without jumps, and therefore we limited ourselves to only three tempering temperatures. The tempering duration was 40 min, and after tempering the specimens were cooled in water.

For testing impact toughness at low temperatures, the previously described method was used. The experiments were carried out on a Charpy pendulum of 10 kg·m capacity at \(+17\), \(-20\), \(-60\), and \(-100^\circ\). The results are presented in the form of curves in Fig. 76 (curves A, steel grade E10; curves B, steel grade E6). For comparison, the same diagram gives a curve for rail sorbitic steel (thin solid line). At normal room temperature, chromium-nickel steels show no particular advantages in impact toughness compared with ordinary sorbitic rail steel \((\mathrm{C}—0.6\%;\ \mathrm{Mn}—0.8\%)\), tempered at 640–660°. Even at \(-20^\circ\) no especially marked difference is observed between them. Steel E10,

tempered at 650° has lower toughness, not to mention steels E6 and E10, tempered at 500°. A difference in favor of chromium-nickel steel is observed beginning at \(-40^\circ\) and becomes especially noticeable at \(-60^\circ\).

The toughness of our chromium-nickel steels at this temperature is higher than the toughness of sorbitic rail steel, if one takes into account that a tempering temperature of 350° is not used in practice. Normal tempering for these steels begins only at 500°. Steel E6 is distinguished by particular toughness, possessing at a temperature of \(-100^\circ\) an impact toughness of about \(4.4\ \mathrm{kgm/cm^2}\) after quenching and tempering at 650°. When tempered at a considerably lower temperature (500°), this steel shows at \(-100^\circ\) a fairly high toughness, of the order of \(3\ \mathrm{kgm/cm^2}\). Evidently, for this chromium-nickel steel, heat-treated to sorbite, a temperature of \(-100^\circ\) is still not in the region of unconditional cold brittleness. The fracture of specimens at \(-100^\circ\) is sharply fibrous, characteristic of tough steel. At this temperature, for steel E6 tempered at various temperatures, there is observed a comparatively very slight scatter of the points that served for constructing the curves shown in Fig. 76. This circumstance confirms that at lower temperatures the region of cold brittleness lies somewhere to the right of \(-100^\circ\). The nature of the scatter is seen from Table 28.

Fig. 76.

Before the imperialist war, ordinary gun steel with \(5—7\%\) Ni or chromium-nickel steel with \(3—4\%\) Ni and \(1.5\%\) Cr was in wide use.

Table 28

Specimen Nos.	Test temperature (°C)	Tempering temperature (°C)	Impact toughness (kgm/cm²)
20	−100	350	1.78
34	−100	350	1.71
35	−100	350	1.13
10	−100	500	2.92
32	−100	500	2.92
33	−100	500	2.68
30	−100	650	4.61
36	−100	650	4.12
37	−100	650	—

As Schvinning (92) emphasizes, this steel had sufficient impact toughness not to be destroyed by the explosion of a shell in the gun barrel as a result of detonation, only if it had been well made by the metallurgists. The Germans, compelled during the war to economize on nickel, compensated for the reduction of its content in the metal by very careful heat treatment to sorbite and by especially attentive manufacture of the steel. It was extremely difficult to obtain the same explosive toughness of sorbitic barrels made of ordinary carbon steel under severe cold conditions.

An increase in nickel content leads to the formation of an austenitic structure; the space lattice of the alloy changes from cubic space-centered to face-centered, similar to the lattice of nickel, and then the steel retains very considerable toughness even at the boiling temperature of liquid air (93).

As is evident from all that has been set forth, the problem of developing a method for eliminating the excessive brittleness of carbon-steel parts when working in frost has been solved, more or less satisfactorily, by heat treatment to sorbite.

VI. RESISTANCE OF RAIL STEEL TO WEAR

For such a part as a rail, special requirements are imposed with respect to resistance to wear. The rail must not be strongly crushed and abraded during service; its rolling surface—

must possess a definite, and at the same time fairly high, hardness.

On the basis of our experiments it may be said that the hardness of sorbite, for a given grade of steel, depends on the tempering temperature and on the duration of the latter. The Brinell hardness decreases as the temperature and duration of tempering increase. True, a considerable Brinell hardness is still not an unconditional indicator of the resistance of rail steel to crushing and mechanical wear. Prof. A. L. Baboshin believes that the notion of the great wear resistance of hard rails is an oversimplification, and maintains that the quality of a rail depends first of all on “high elasticity with sufficient plasticity.”

Experiments to determine the resistance of rail steel to wear were carried out in the mechanical laboratory of the Siberian Institute of Metals and in the central laboratory of the Kuznetsk Metallurgical Combine by Engineer Konvisarov (92). The work was conducted first on a machine designed by him, and then on an ordinary machine of the Amsler system. In what follows we shall refer only to the latter experiments. For testing on the Amsler machine, specimens were cut from the head of the rail across the direction of rolling and heat-treated. The specimens, in the form of round washers, were made from rail steel of the Kuznetsk Metallurgical Combine, with a content of C—0.53% and Mn—0.77%.

The Brinell hardness of these specimens, after the corresponding heat treatment, is shown in Table 29.

Table 29

Quenching conditions	Tempering duration (hours)	Tempering temperature (°C)	Brinell hardness
900°, in water	0.25	500	447
(+20°)	0.25	600	345
(+20°)	0.25	650	266
(+20°)	0.25	680	232
(+20°)	4	600	293
(+20°)	4	650	256
(+20°)	4	680	197
Unheated	—	—	273

These specimens were tested for wear: 1) by pure sliding without rolling; 2) by rolling with 3% sliding; 3) by rolling with 10% sliding.

The pressure values (50; 80.5; 121 kg) between the standard washer and the washer under study were selected in order to obtain a friction moment not exceeding the maximum for a machine of this type. The maximum moment was chosen in order to obtain measurable results in a shorter time. At the same time, attention was paid to the inadmissibility of general overheating of the washers. Therefore, the number of revolutions during the experiments was taken as 190–200 per min.

After each experiment the specimen was carefully measured. Observations during the investigation showed that at first abrasion of the metal occurs, rather quickly passing into a process whose essence consists in the separation of oxidized black particles. These particles have the form of very thin scales. At the point of contact of the rubbing surfaces, a large amount of heat is released, evidently playing a significant role in the formation of oxidized metal. There is no doubt that the properties of the steel, determined by the preceding heat treatment, are thereby changed in some way.

As a measure of wear, the size of the imprint (area) after 25,000 revolutions of the control washer was adopted. Table 30 gives comparative values of the imprint for steel with different heat treatments. Obviously, the smaller the imprint, the better the steel resists wear under pure sliding.

Table 30

Quenching at 900° in water and tempering at (°C)	Duration of tempering (in hours)	Imprint area (in mm²)
500	0.25	51.32
600	0.25	73.50
650	0.25	62.46
680	0.25	65.14
600	4	86.42
650	4	70.05
680	4	111.83
Untreated	4	68.60

As can be seen, all sorbitic steels with short-time tempering resist wear better than untreated steel. Of the steels tempered for 4 hours, appreciable resistance was found in steels tempered at 650°. As additional experiments showed, reducing the duration of tempering to 1½–2 hours increases the abrasion resistance of sorbitic steel in comparison with untreated steel and with steel tempered for a longer time.

In the described series of experiments, there is evidently a very great similarity to the slipping of locomotive wheels along a rail, which is often encountered in practice.

The second and third series of experiments more closely approximate the conditions of actual rail operation, when almost pure rolling is observed, with slight slipping of the locomotive wheels under weak braking.

Obviously, the measure of wear in these cases will be something different, and not the magnitude of the imprint, since, owing to the motion of both washers—the test washer and the standard—the imprint, as an indicator of wear, is not obtained at all under these conditions. Therefore, as the measure of wear we took the change in the diameter of the specimen after 60 and 30 thousand revolutions of the standard washer. In the testing of metals for wear there is still, unfortunately, no generally accepted criterion for determining the degree of wear. In all such investigations it is also necessary to determine the magnitude of the plastically deformed volume, the loss in weight of the specimen, and the change in the area of the cross section. These quantities may be referred either to the work of friction or to the total path over which wear takes place. As the measure of wear in our case, the change in the linear dimensions of the specimen was adopted. We assumed that it is precisely the change in the linear dimensions of the rail, and not the loss in weight or volume, that is primarily the practical indicator of the degree of actual wear. The indicator of a rail’s resistance to wear in the track, in our opinion, is the number of trains (ton-axles) of a certain weight that have passed over the track before the first perceptible signs of wear appear—the decrease in the height of the rail.

As in pure friction without rolling, at the very beginning of the process the surface of the specimen remains shiny for some time. But after 15–20 minutes the outer layer, without losing its luster, begins gradually to darken. Then dark matte spots gradually form on it; this phenomenon is accompanied by jumps of the dynamometer pointer, since the friction force at this time increases in a stepwise manner. Finally, products of wear begin to separate from the rubbing surface in the form of shiny flakes of metal and a black fine powder. The surface then remains matte, and thereafter gradual uniform wear begins.

In these cases the sorbitic specimens with a short tempering time (15 min.) showed less wear than raw steel; the specimens with the maximum tempering duration (4 hours) proved less wear-resistant.

On the basis of the foregoing, a number of conclusions may be drawn. Wear by pure sliding, at considerable specific pressures, causes the development of large quantities of heat at the point of contact of the rubbing surfaces. This leads to a considerable local increase in the temperature of the specimen and, as a consequence, to the nullification of the effects of heat treatment. The same,

phenomena of approximately the same order were observed by Thum and Wunderlich (95). These authors believe that, when steel surfaces are rubbed, the following are observed: 1) the formation of red or black oxidized metal particles in the form of small flakes and powder; 2) local deep pitting of the surface, which very often (but not always) is associated precisely with the appearance of oxidized particles; 3) shiny polished spots having a red tint.

Schottky and Hiltenkamp (96) demonstrate the possibility of atmospheric nitrogen being absorbed by rubbing steel surfaces in cases where these surfaces are strongly crushed and scored. By chemical analysis the authors found an enrichment in nitrogen of surfaces rubbing under such conditions, which was also confirmed by metallographic examination. There is no doubt that such a change in rubbing surfaces will lead to the appearance of local surface brittleness of very considerable magnitude; to the formation, during the further service of the rail in the track, of surface cracks as a result of increased brittleness, the danger of which under dynamic and alternating stresses will be especially great. Wister (97) also found absorption of nitrogen by the surface of mild steel during grinding and polishing. Nitrogen is absorbed most strongly during coarse grinding, and its content in the surface layer may increase from 0.004 to 0.100%.

Bishop (78), quenching in water a steel containing 0.25% C and 0.63% Mn, heated to 650°, found an impact toughness immediately after quenching equal to 10.8 kgm/cm². After holding at room temperature for 14 days, the impact toughness fell practically to zero. At the same time the hardness increased from 143 to 181 Brinell units. Before such heat treatment the steel had been heated to 900° and cooled together with the furnace. It seems to us that aging with such results cannot in any way be ignored when discussing the changes caused by the movement of a braked train over the surface of a rail. Heating of the running surface of rails to 600–650° during service is entirely possible, as is rapid cooling.

Of the numerous literature data on the wear resistance of steel, we shall dwell on only a few works.

Pilz (98) gave an interesting relation between tensile strength and the wear of rails with a running surface sorbitized by the Neveu-Maison method. The abrasion tests were carried out by this author on an Amsler machine at 10% slip and a pressure of 90 kg. Fig. 77 gives the dependence of loss in weight on tensile strength for untreated rails (dashed curve) and rails with a sorbitized surface (solid line). As is evident from Fig. 77, rails with a sorbitized surface are more resistant to abrasion than ordinary ones only up to a certain limit. If the magnitude of the tensile resist—

…reaches 90 kg/mm², then sorbitic rails offer no advantages in comparison with untreated ones possessing the same strength. This is evident from the fact that both curves converge at approximately one point when the ultimate strength is equal to 90 kg/mm². As a measure of wear, Pilz adopted the loss in weight (in g) of the washer under investigation after 40,000 revolutions. According to Pilz, improvement in the quality of rails should proceed mainly by increasing the silicon and carbon content.

Fig. 77

According to Thibodier and Viteaux (99), American firms have long been trying to combat rail wear by raising the carbon content to 0.80 and even 0.88%.

As Ros and Eichinger (100) assert, the least wear is observed in the case of rails homogeneous in composition, with a strongly improved (sorbitized) head. Inferior to them in wear resistance are rails with considerable overall hardness; the worst in terms of wear are weakly hardened (more precisely, strongly tempered) rails.

In analyzing the question of the possible wear of rails sorbitized by our method, it must be pointed out that a high tempering temperature (680°) of long duration for steel quenched to martensite reduces the resistance to wear, as determined on the Amsler machine. Satisfactory results are obtained with a somewhat lower tempering temperature (640°) and shorter duration (2 hours). But in this connection it is necessary to take into account that all our experiments on determining the impact toughness of one-and-a-half-meter rail ends showed the possibility of lowering the tempering temperature to…

of the rail web even up to \(500\)—\(550^\circ\). With such tempering limits, half-size specimens still show very considerable toughness when tested on the plant impact machine. Taking this into account, it may be asserted that, by reducing brittleness at low temperatures through through-sorbitizing, it is possible at the same time to increase the steel’s resistance to wear. In any case, this question can be resolved with complete definitiveness only after trial rails have been laid on an experimental section.

Obviously, for the overwhelming majority of machine parts this question does not have such pressing significance. Only from rails do we demand properties that are as yet incompatible for ordinary steel—high impact toughness and increased hardness, ensuring at least to some degree resistance to wear.

VII. CONCLUSION

On the basis of the experimental data presented, although they do not, of course, claim to be exhaustive, the decrease in impact toughness of ordinary grades of carbon and low-alloy steel may be regarded as an entirely normal phenomenon. All grades of technical iron–carbon alloys become brittle at low temperatures; the transition from a tough state to a brittle one occurs abruptly within rather narrow temperature limits.

The magnitude of the temperature interval within which this rapid fall in impact toughness occurs, when related to the determined magnitude of the decrease in the value of impact toughness, may serve as a measure of the sensitivity of the given material to impact at low temperatures, as a characteristic of the degree of cold brittleness of the metal.

Let us take two steels \(A\) and \(B\) with diagrams constructed for them: impact toughness—temperature (Fig. 78). From the drawing it is seen that the aforementioned decrease in the impact toughness of steel \(A\) by some quite definite amount \(\rho\), expressed in \(\mathrm{kgm}\), occurred over the temperature interval \(t_1^\circ\), while for steel \(B\) it occurred over a temperature interval of another magnitude \((t_2^\circ)\). In the present case it is not yet possible to decide definitively which of the two steels is better in terms of resistance at low temperatures.

The magnitude \(t_1\) or \(t_2\) still gives no idea of the sensitivity of the metal to impact at low temperatures. What characterizes the metal more definitely in the direction of interest to us is the position of the interval \(t_1\) or \(t_2\) on the temperature axis, and also the position of point \(m\) on the temperature axis and its tendency, under the influence of one circumstance or another, to shift to the right or to the left.

As we have repeatedly pointed out, the magnitude of the interval \((t_1,t_2)\) in degrees and the position of point \(m\) on the temperature axis (Fig. 78) depend entirely on the nature of the alloy (the features of its structure ...).

lattice), its purity, the nature and quantity of impurities, and the structural state of the metal, determined both by mechanical and thermal treatment and by the special processes that give rise to the whole variety of aging phenomena. A work-hardened metal

Fig. 78

is usually more brittle at a suitable temperature than the same metal that has not undergone plastic deformation in the cold state; an aged metal is often less tough than a well-annealed one; a fine-grained metal is less brittle than a coarse-crystalline one. All these features are best revealed at reduced temperatures under dynamic testing.

It must be noted here that some authors (for example, Corelli), working with mild steel, did not find excessive brittleness at fairly low temperatures. The author of the present work (together with K. A. Bessonov) obtained single crystals from Soviet Armco iron that showed no brittleness even at the temperature of liquid air, which contradicts the conclusions from Sauerwald’s work (85).

Let us try to consider the behavior of an ideal metallic crystal deformed beyond the elastic limit. Under the influence of an increasing load, residual plastic deformations appear in the material. These deformations (101) represent, at the initial stage, phenomena of shear, of the sliding of some layers of the crystal parallel to others; this is sliding (translation) along a crystallographic plane, definite for the crystalline system of the given substance, in a definite direction. Externally, such a process is expressed in the appearance of slip lines on the polished or natural surface of the deformed crystal.

But even the most perfect single crystal cannot be represented as some ideal formation. This applies especially--

...with single crystals obtained by recrystallization of technical, albeit very pure, iron. In such a crystal, besides local distortions of the space lattice, there are the most ordinary slags, all sorts of nonmetallic inclusions, and simply voids and cracks of submicroscopic size.

In many cases such deformation, elementary in its nature, is accompanied by a special kind of shift: the formation of twins. Face-centered lattices have a great tendency toward twinning, but crystals of iron also readily form twins under suitable conditions.

The plasticity of crystals is very closely connected with the possibility of the occurrence of slip planes. Theoretically, every atomic layer can, without destroying the crystal, move parallel to a neighboring layer. There can be as many slip planes in a crystal as there are atomic layers perpendicular to the direction in which the shifts occur. The number of possible planes will be equal to the dimension of the crystal, perpendicular to the slip surface, divided by the distance between the slipping layers.

In a crystal of $\alpha$-iron, with a size of $1\ \mathrm{mm}^3$, if slip proceeds along the face of the cube, there will be 3.5 million such planes.

The minimum path of slip in this case cannot be, in magnitude, less than the lattice parameter (the edge of the cube), and the increase in the length of our crystal in the direction $[100]$ will be equal to $1\ \mathrm{mm}$, provided that all possible planes have come into action and each of the layers has shifted relative to its neighbor by the magnitude of the lattice parameter.

As direct observations show, shift by no means occurs along all possible slip planes of a given direction. The active slip surfaces are spaced from one another not by a distance of several Å, but by such a distance that the intervals between them are visible under a microscope. Likewise, the magnitudes of individual shifts are not limited by the magnitude of the lattice parameter, but reach hundreds of Å.

This is explained precisely by the fact that, in reality, we do not have ideal crystals with a mathematically regular lattice. Even crystals that are extremely pure in chemical composition always contain small amounts of impurities, which by their presence cause a local distortion of the lattice.

If all impurities, without exception, formed solid solutions with the material of the crystal, then even then the inhomogeneity of the solid solution within one and the same crystal would create conditions favorable for the occurrence of slip along only a portion of the possible planes. But even such a case is not encountered in reality; a crystal always contains a large number of impurities insoluble in the solid crystal. Often these are ultramicroscopic particles entering into...

the crystal lattice during crystallization and deforming it at those places where they were caught by crystallization of the base material.

As we have shown (80), certain inclusions in crystals of $\alpha$-iron are arranged along quite definite crystallographic surfaces. Thus, we found (80) that nitride needles, formed when crystals of $\alpha$-iron are heated in a stream of ammonia, are arranged along the faces of a rhombic dodecahedron.

Masing (102) showed that, in tin single crystals, impurities are arranged along the slip planes and, by their presence, hinder translation, reduce the number of active slip surfaces, and increase the strength and brittleness of the single crystal.

Impurities precipitating during the crystallization of salts are distributed in approximately the same way (103).

It may be asserted that a reduction in the number of possible slip surfaces in a metallic crystal lowers plasticity and leads to an increase in strength.

If we consider not one direction of slip, but all possible directions, then we shall obviously come to the conclusion that, when there is a very considerable number of slip surfaces, the possibility of obtaining the slip plane most favorably oriented with respect to the acting forces will also be greater than when their number is small.

In the case of a large number of different slip directions, several systems of them may come into action.

Technical iron possesses considerable impact toughness at room temperature and usually becomes catastrophically brittle at comparatively not very low temperatures. On the basis of what has been presented, it may be asserted that the purer the iron, the tougher it is even at low temperatures, the farther to the left point $m$ (see Fig. 78) is shifted, and the larger for it is the temperature interval $t_1^\circ$.

Impurities of a mechanical character—slags, nonmetallic inclusions—everything that hinders the occurrence of plastic deformations, also increase the cold brittleness of steel. It is necessary to note here that high toughness at normal temperature does not always indicate low brittleness at low temperature. By comparing the impact-toughness–temperature curves presented in our work, this is easy to establish.

One should also take into account the very harmful influence of certain nonmetallic inclusions having a shape especially favorable to the development of local overstresses in the material under impact. An inter-slag layer elongated perpendicular to the axis of the Mesnager specimen will play a role very harmful to the dynamic strength of the notched specimen.

But the influence on toughness of inclusions, whatever their character may be, depends both on their shape and on their size. Somewhere

there must be a limit at which very small particles no longer exert a particularly harmful influence on toughness.

Likewise, any distortion of the lattice that impedes translation will lead to an increase in the static strength of the crystalline aggregate. But this increase in strength will also be accompanied by an increase in impact brittleness. It is therefore understandable that the aging observed by Körber and by the workers of the central laboratory of the Kuznetsk Metallurgical Combine, and consisting in the removal of lattice stresses as a result of hydrogen removal, led to an increase in the impact toughness of steel.

Mechanical aging (after plastic deformation of considerable magnitude) is characterized by the precipitation of certain highly dispersed particles and, conversely, leads to a decrease in toughness, especially impact toughness. But aging after small deformations, as we have shown, does not impair steel. Probably, in the latter case the precipitating particles have very small dimensions and do not affect the mechanical properties.

Sorbitic steel is especially insensitive to aging, and this insensitivity increases with the duration and temperature of tempering, because the structure becomes ever more uniform, while the fine dispersed particles, by coagulating, increase very greatly in size with time.

The latter circumstance may be explained by the fact that the particles precipitating during aging can block slip surfaces in different ways, impeding translation. From particles of molecular order the blocking is small. As they grow, the blocking increases. But when their size reaches that of sorbitic cementite, visible under the microscope, the blocking action again weakens.

With respect to the brittleness of more or less pure iron, there exist, as we have indicated, contradictory opinions. We explain this by the fact that, in the interval in which cold brittleness appears (Fig. 78), the slightest nonmetallic inclusions of suitable form in the crystals of the metal create local lattice stresses, zones of overstress, and at the same time hinder the possibility of translation. But by chance a specimen may occur that has such defective sites in minimal quantities. Apparently we had just such a case when some single-crystal specimens (without notch) made from Soviet Armco iron did not break in impact tests at \(-183^\circ\). But this exceptional case by no means indicates that this iron is not cold-brittle and possesses toughness at all temperatures, like copper or aluminum. A single crystal that bent at \(-183^\circ\) without the formation of cracks will evidently nevertheless break at a lower temperature. The point \(m\) (see Fig. 78) for such a material lies to the left of \(-183^\circ\).

We think that the cold brittleness of iron is explained by the smaller number of possible slip directions at low

temperatures and with a decrease in the number of possible slips per unit length of the crystal as the temperature falls. As we pointed out earlier, the number of slip lines on the lateral surface of single crystals deformed by impact decreases as the temperature of the experiment is lowered. On an iron single crystal broken by impact at \(-183^\circ\), there are almost no slip lines.

Cold acts in the same direction as the impurities entering the crystal lattice; in both cases translation is made difficult. Very pure iron at \(-183^\circ\) gives a certain number of slip lines, which was also found on specimens that did not fracture at this temperature.

Along with the almost complete absence of slip lines on iron single crystals broken at the temperature of liquid air, there was observed—though not in all cases—the appearance of twins, which to this day are called “Neumann lines.”

N. N. Davidenkov (1) explains the appearance of twins by the fact that the resistance to twinning is less than the resistance to simple slip, and he points out that there is a connection between the brittleness of iron and its twinning. In his opinion, during deformation very considerable overstresses arise in the crystal around the twins, facilitating the destruction of the crystal.

It is hardly possible to attribute the cold brittleness of iron and some of its alloys exclusively to the formation of twins. Twins are formed with particular ease precisely in metals that are not afraid of impact at low temperatures.

Sakharov (104) notes the appearance of twins (Neumann lines), numbering 2–3 per grain, in steel with a content of \(0.2\%\) in the case of deformation at \(-183^\circ\). At the same time he also observed ordinary slip lines in the form of indistinct, curved lines.

In one of the works by the author and V. D. Kuznetsov (23) it was shown that the energy \(U\) of the Charpy pendulum hammer expended in fracturing a specimen consists of the following parts: \(U_1\)—the energy spent on producing elastic deformations in the specimen; \(U_2\)—the energy causing residual deformations of the specimen; \(U_3\)—the energy required to increase the surface of the specimen during its destruction; \(U_4\)—sound energy. In that work it was proved that the terms \(U_3\) and \(U_4\) may be neglected without great error, and the total energy for fracture consists of only two parts, \(U = U_1 + U_2\).

Under the influence of external forces, elastic deformations arise in the specimen and pass over into plastic ones. The work of the external forces that cause elastic deformations during impact bending is determined, in the final analysis, by the value of the modulus of elasticity. But Young’s modulus changes little with temperature; to a first approximation it may be considered constant in the interval \(+20\) to \(-183^\circ\).

Below a certain temperature (for example, \(-40^\circ\) in some of our experiments), destruction of the specimen is not accompanied by resid-

by exact deformations, sliding surfaces are almost not formed, and the elastic limit remains above the strength limit. The work of fracture of the specimen below this temperature, which may be called critical \((t_k^\circ)\), remains practically constant.

Consequently, in the interval \(t_k^\circ — 0^\circ\) (abs.) \(U_1\) must be constant, while \(U_2\) is equal to zero.

Above the critical temperature the specimen passes from the brittle state into the plastic state, and the work of plastic deformation must increase. \(U_2\) grows more rapidly than the decrease of \(U_1\), and the impact toughness—temperature curve goes upward; but from a certain temperature it must fall, reaching the temperature axis at the melting point of the metal.

VIII. CONCLUSIONS

1. We have confirmed, by a number of numerous experiments, that the impact toughness of iron–carbon alloys and of low-alloy steels decreases very strongly, and in some cases quite catastrophically, with a lowering of the test temperature.

2. The temperature interval in which the appearance of cold brittleness (brittleness under impact) is observed depends on the carbon content in the steel, on unavoidable impurities, on the method of producing the steel, on the preceding hot, cold, and heat treatment, and on the test method adopted.

3. It is possible to increase very considerably the impact toughness at low temperatures of ordinary structural and low-alloy steel by heat treatment to sorbite. The higher the tempering temperature to sorbite and the longer the tempering time, the farther the interval of cold brittleness is shifted toward lower temperatures, and the tougher the steel is at low temperatures.

4. Of especially great importance is the reduction of brittleness under impact at low temperatures, as a result of the proposed method of heat treatment, for rails operating under the conditions of Siberian winters.

5. For quenching rails, as a result of a series of investigations, a quenching machine has been designed, manufactured, and tested which makes it possible to quench, without cracks and practically without warping, rail specimens for the drop test, 1.5 m long.

It is possible, by increasing the length of this quenching machine by the appropriate number of times, to quench in it rails of normal length.

1. A number of investigations have shown that rails may be quenched at the temperature at which they leave the last stand, without reheating or equalizing their temperature in special furnaces.

The temperature fluctuations at the end of rolling that are usual under production conditions do not affect the final result.

1. Straightening sorbitized rails in the cold state on a roller straightening machine does not impair the mechanical properties of a rail sorbitized by our method, even after artificial aging.

2. The purer the iron and steel, the less they are susceptible to impact at low temperatures.

For heat treatment by the method we have proposed, good-quality open-hearth rails are suitable, primarily those with an appropriate carbon and manganese content.

The question of the possibility and profitability of such heat treatment of Bessemer and Thomas rails is not clear; most likely, no particular benefit can be expected from their treatment.

1. The sizes of the austenitic grain, within the limits in which they vary in normal open-hearth rail steel, have no special influence on the results of heat treatment for shifting the cold-brittleness interval into the region of lower temperatures.

2. The possibility has been demonstrated of hardening rail steel with isothermal decomposition of austenite. By applying such a hardening method, it is possible to carry out the straightening operation on a roller machine during the hardening process.

3. A new method has been developed for drop-weight testing of rails cooled to a low temperature, with accurate accounting of the energy expended in fracturing the normal specimen for the factory drop-weight test.

4. Certain alloy steels have been investigated for cold brittleness.

5. An attempt has been made to explain the cause of the cold brittleness of iron and its alloys with carbon as a function of their internal structure.

LITERARY SOURCES REFERRED TO IN THE WORK

1. N. N. Davidenkov, Dynamic Testing of Metals, pp. 7–8, 1936.

2. G. K. Tyumentsev, General Conclusions from 37 Years of Meteorological Observations, Tomsk, 1912.

3. F. N. Rybinskii, “Brittle Rails,” collection of TsNIIM NKPS, 1931.

4. V. Kirpichev, Strength of Materials, part 2, pp. 68–69, 1918.

5. F. Körber and A. Pomp, “Mitt. aus dem K-W-Inst. für Eisenforsch zu Düsseldorf,” 1925.

6. G. Charpy, “Intern. Verband f. d. Materialprüf d. Technik,” Bruss. Kongr., 1906.

1. L. Guillet and L. Revillon, Revue de Metallurgie, 1906, p. 96.
2. P. Goerens and G. Hartel, Zs für anorg. Chemie, 1913, p. 430.
3. O. Reingold, Ferrum, 1916, p. 97.
4. B. Strauss and A. Fry, Kruppsche Monatsheft, 1921, p. 126.
5. F. C. Langenberg, J. Iron a. Steel Inst., 1923, p. 95.
6. N. Richardson and E. K. Mac Nutt, Trans. Am. Soc. for Steel Treat., 1924, p. 348.
7. P. Goerens, ZVDI, 1924, p. 41.
8. R. Mailänder, Kruppsche Monatshefte, 1924, p. 16.
9. F. Körber and A. Pomp, Mitt. aus dem K.-W.-Inst. für Eisenforsch zu Düsseldorf, 1925.
10. F. Körber and A. Pomp, Mitt. aus dem K.-W.-Inst. für Eisenforsch., vol. VII, issue 4, 1935.
11. R. Waile, St. u. Eisen, 1932, pp. 479–488.
12. M. Corelli, Congrès Intern. pour l’essai de Mat., vol. I, 1928, pp. 479–88.
13. B. A. Smirnov and A. S. Falkevich, Avtogennoe delo, 1935, No. 4.
14. V. I. Voznyak, Avtogennoe delo, 1935, No. 4.
15. V. I. Strokopytov, Heat Treatment of an Electric-Arc Weld, diploma thesis (carried out under the author’s supervision), Tomsk State University, 1935.
16. A. Falkevich and B. Smirnov, Khimstroi, 1935, No. 7.
17. A. N. Dobrovidov and V. D. Kuznetsov, Proceedings of the Siberian Physico-Technical Institute, 1932, vol. I, issue II; Stal’, 1932, No. 4.
18. Larsen, Dr.-Ing. Diss., Berlin, 1930.
19. H. Meyer, St. u. Eisen, 1931, p. 1006.
20. W. A. Güldner, ZS für Metallkunde, 1930, pp. 258–60 and 412–16.
21. V. A. Tinyayev, Brittleness of Aluminum and Its Alloys, diploma thesis (carried out under the author’s supervision), Tomsk State University, 1935.
22. R. Mailänder, St. u. Eisen, 1935, pp. 749–51.
23. F. N. Rybinskii, “Brittle Rails,” collection of the TsNIIM of the NKPS, pp. 37–74, 1931.
24. Report on practical training at the Chelyabinsk Tractor Plant, Tomsk State University, 1934.
25. A. I. Skakov and A. S. Makhanek, On the Mass Investigation of Rails, collection of the NIIM of the NKPS, p. 29, 1933.
26. H. Schottky and H. Hiltenkamp, St. u. Eisen, 1936, pp. 444–46.
27. A. P. Chekmarev, Rails, Their Service, Quality, and Production, GNTIU, 1934.
28. H. Wentrup, ZVDI, 1934, p. 565.
29. F. Körber and M. Hempel, Mitt. aus dem K.-W.-Inst. f. Eisenforsch., Düsseldorf, 1935, vol. XVII, p. 292.
30. Fratkin, Bulletin of the Dzerzhinskii Plant, No. 9–10, 1932.
31. L. P. Vladimirov, Domez, 1930, No. 5.
32. A. L. Baboshin, “Investigation of Rail Practice in the USSR,” collection, part I, pp. 48–65, 1928.
33. S. L. Zhukovskii, ZhRMO, 1912, No. 3, pp. 329–336.
34. N. A. Minkevich, Properties, Heat Treatment, and Application of Steel and Cast Iron, parts I and 2, pp. 269–279, 1932.
35. I. N. Efremov and V. N. Gorskii, Bulletin of Uralmet, 1930, No. 3 and 4; Metallurg, 1931, No. 6–7.
36. H. Bennek, Arch. f. d. Eisenhüttenwes., 1935/36, pp. 147–153.
37. K. A. Bessonov, The Influence of Quenching Temperature, Tempering Duration, and Cooling Rate after Tempering on the Impact Toughness of Steel, technical report (work carried out under the author’s supervision), Siberian Institute of Metals, 1932.

1. P. S. Grdina, Anomalousness of rail metal, technical report, Siberian Institute of Metals, 1932.
   P. S. Grdina, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV, pp. 119–150.

2. S. M. Herty, “Grain Size in Steel,” collection, ONTI NKTP, 1935.

3. J. J. Egan, W. Crafts and A. B. Kinzel, “Metal Progress,” 1933, Sept., pp. 18–20.

4. I. Mirkin and A. Rozanov, Collection of the Moscow Institute of Steel named after I. V. Stalin, 1934.

5. M. P. Prishchepa, Properties of isothermally decomposed austenite, diploma thesis (carried out under the author’s supervision), Tomsk State University, 1936.

6. K. A. Bessonov, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV, pp. 17–59.

7. Inclusions in Steel, collection edited by Prof. S. S. Shteinberg, ONTI, 1933.

8. Kinzel and Crafts, “St. u. Eisen,” 1931, p. 778 and p. 1529.

9. G. I. Andreev, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV.

10. D. S. Gruzdev, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV, pp. 59–71.

11. D. S. Gruzdev, Preliminary technical report, Siberian Institute of Metals, 1935.

12. S. I. Druzhinin, “Investigation of Rail Steel” (collection), pp. 55–91, 1906.

13. Inclusions in Steel, collection edited by Prof. S. S. Shteinberg, pp. 158–171. ONTI, 1933.

14. S. Gallik and G. Hoensch, “ZVDI,” 1935, p. 1480.

15. M. J. Galibourg, “Revue de Mét.,” 1936, pp. 73–85.

16. W. Köster, “Archiv f. d. Eisenhüttenwes,” February 8, 1929.

17. W. Eilender, A. Fry and A. Gottwald, “St. u. Eisen,” 1935, pp. 555–64.

18. V. I. Strokopytov, Report on practical work at KMK, Tomsk State University, 1934.

19. W. Köster. “Arch. f. d. Eisenhüttenwes.,” 1930, 10, pp. 633–55.

20. F. Körber and J. Mehovar, “Mitt. aus. d. K.-W-Inst. zu Düsseldorf,” 1935. 7. pp. 89–105.

21. C. Drescher and R. Schäfer, “Archiv f. d. Eisenhüttenwes.,” 1935/36, pp. 327–32.

22. G. Schmidt, “Archiv. f. d. Eisenhüttenwes.,” 1934/35, pp. 263–69.

23. W. Eilender, H. Cornelius and H. Knüppel, “Archiv f. d. Eisenhüttenwes.,” 1934/35, pp. 507–9.

24. E. Knipp, “St. u. Eisen,” 1934, pp. 777–8.

25. R. Walzel, “Archiv f. d. Eisenhüttenwes.,” 1932, pp. 257–62.

26. H. Wilhelm and L. Jahn, “Archiv f. d. Eisenhüttenwes.,” 1933/34, pp. 201–3.

27. R. Stumper, “St. u. Eisen,” 1929, pp. 177–87.

28. A. Krüger, “Archiv f. d. Eisenhüttenwes.,” 1930, pp. 721–30.

29. F. Rötscher and M. Fink, “ZVDI,” 1932, No. 2.

30. T. N. Lyashenko, “Kuznetskstroy,” 1935, No. 1–2, pp. 60–65.

31. V. I. Dyakov, Aging of sorbitic steel, diploma thesis (carried out under the author’s supervision), Tomsk State University, 1935.

32. S. G. Skobennikov, Aging of steel, unpublished work. Tomsk Industrial Institute, 1936.

33. E. Knipp, “St. u. Eisen,” 1934, no. 24.

34. S. N. Herty, “Grain Size in Steel,” collection, ONTI NKTP, 1936, p. 363.

35. W. Biscohf, “Arch. f. d. Eisenhüttenwes.,” 1934/35, pp. 293–303.

36. L. V. Wert, “Trans. Amer. Soc. of Metals.,” 1935, pp. 633–44.

1. A. N. Dobrovidov, Impurities in Single Crystals, report at the Siberian Institute of Metals, 1929.

2. “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV, pp. 163–172.

3. C. A. Edwards and L. B. Pfeil, “J. Ir. a. Steel Inst.,” 1925, pp. 79–100.

4. W. S. Messkin and J. M. Margolin, “Arch. f. d. Eisenhüttenwes.,” 1934, pp. 399–405.

5. L. Luckmeyer-Hasse and T. Schenk, “Arch. f. d. Eisenhüttenwes.,” 1932, pp. 209–14.

6. Sauerwald, Schmid, and Dickenthal, “ZS f. Phys.,” 1930, vol. 61, p. 130.

7. H. W. Hiemke and W. C. Schulte, “Metals a. Alloys,” 1934, Febr., p. 31.

8. J. J. Egan, F. Grafts, and A. B. Kinzel, “Metal Progress,” 1933, Sept., pp. 18–20.

9. J. F. Morrison, “Eng.,” 1934, p. 550.

10. B. G. Aldridge and G. C. Shepherd, “Metals a. All.,” 1936, Juni, pp. 147–52; Juli, pp. 185–192.

11. A. N. Dobrovidov and D. S. Gruzdev, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV.

12. A. V. Smirnov and A. V. Beloruchev, Oxidation and Deoxidation of Steel, ONTI, 1934.

13. W. Schwinnig, “ZVDI,” 1935, pp. 35–40.

14. E. W. Colbeck, W. E. MacGillivray, and W. R. D. Manning, “Inst. Chem. Eng.,” 24 May, 1933.

15. D. V. Konvisarov, “Proceedings of the Siberian Institute of Metals,” 1934, vol. I, issue IV, pp. 85–112.

16. A. Thum and F. Wunderlich, “ZS f. Metallkunde,” 1935, pp. 277–80.

17. H. Schottky and H. Hiltenkamp, “St. u. Eisen,” 1936, pp. 444–46.

18. H. J. Wiester, “Arch. f. d. Eisenhüttenwes.,” 1935/36, p. 527.

19. O. Pilz, “St. u. Eisen,” 1927, pp. 1645–51.

20. L. Thibodier and H. Viteaux, “Revue de Mét.,” 1926, pp. 65–81.

21. M. Ros and A. Eichinger, “ZVDI,” 1934, p. 893.

22. G. Sachs, Plastische Verform., 1930, p. 20.

23. N. V. Myshak, Distribution of Impurities in a Single Crystal, diploma thesis (carried out under the author’s supervision), 1935.

24. D. D. Saratovkin, “Reports of the Academy of Sciences of the USSR,” 1935, vol. IV (IX), No. 8–9, pp. 307–310.

25. P. S. Sakharov, “Journal of Technical Physics,” 1936, pp. 1389–92.

Contents

Preface ........................................................................ 3

I. Literary Data .............................................................. 9

II. Reduction of the Cold Brittleness of Carbon Steel by Heat Treatment .... 27

III. Semi-Works Experiments .................................................. 53

IV. Influence of Aging on Cold Brittleness .................................... 74

V. Influence of Certain Elements on the Cold Brittleness of Steel ............. 101

VI. Resistance of Rail Steel to Wear .......................................... 113

VII. Conclusion .............................................................. 119

VIII. Findings ............................................................... 125

Literature Sources ........................................................... 126

Responsible editor: Prof. M. A. Usov.
Technical editor: N. Dugina.

Submitted for production: 28/VI 1937.
Signed for printing: 15/IX 1937.

Format 60×92 1/16. Printed sheets: 8.
Characters per printed sheet: 49,776.

Sverdlovsk branch of ONTI, No. 267. Authorized by Sverdloblita, B—3904.

Print run: 1150. Order No. 8384.

Printing house of the newspaper-and-journal publishing house of the Sverdlovsk Regional Committee of the VKP(b), “Uralskii Rabochii.”
47 Lenin St.

1. “First heat” is arbitrary numbering, not the mill’s numbering. ↩