Cold Brittleness of Steel

A. N. Dobrovidov
Professor, Doctor of Technical Sciences

I. Review of the Literature

Since the publication of our work [13] at the end of 1937, interest in questions connected with the cold brittleness of steel has not diminished. This is evidenced by the large number of works devoted to the problem of impact brittleness at low temperatures.

A very interesting monograph is the book by S. E. Belyaev [2], devoted to the question of the strength of aviation metals at low temperatures. In this extensive study the author gives a review of the literature and presents the results of his interesting and large-scale experiments.

S. E. Belyaev divides all the alloys he studied into three principal groups according to the structure of their crystal lattice. To the first group he assigns alloys possessing the space lattice of a body-centered cube—Armco iron, steels U4 and U9, chromansil, chromium-molybdenum steel (KhTM), and chromium-nickel steel (Kh3N). To the second group belong those having a face-centered cubic lattice (copper, aluminum, duralumin, stainless austenitic steel, silumin, and some other alloys). And, finally, to the third group belong those characterized by a hexagonal space lattice (pure zinc and cadmium, electron VIAM-95). Tests (for static strength, impact toughness, torsion, fatigue under alternating bending, and investigation of the strength of a welded seam) were carried out in the temperature interval from \(+10^\circ\) to \(-195^\circ\).

The conclusions reached by the author may be summarized briefly as follows. Armco iron—a representative of the metals of the first group—showed an extraordinary decrease in impact toughness (standard specimens with a Ménagé notch were used) at \(-70^\circ\). The value of the impact toughness falls from \(24\ \mathrm{kgm/cm^2}\) (at \(+20^\circ\)) to \(1\ \mathrm{kgm/cm^2}\) (at \(-70^\circ\)).

As the author indicates, in specimens subjected to dynamic testing on a pendulum at \(+20^\circ\), slips in plastically deformed crystals are visible (at 100-fold magnification). At \(-70^\circ\) and \(-195^\circ\) the fractures of the specimens are completely brittle: in the microphotograph, plastic deformation of the crystals, characterized by the appearance of slip lines, is not noticeable. In the specimens tested at \(-40^\circ\), mixed fractures are observed.

The impact toughness of steel U4 likewise decreases very strongly with a decrease in the testing temperature. However, the critical temperature of brittleness of this steel is considerably lower than that of Armco iron (about \(-100^\circ\)).

Chromansil steel (composition \(0.30\%\ \mathrm{C}\), \(1.1\%\ \mathrm{Mn}\), \(0.97\%\ \mathrm{Cr}\), \(0.97\%\ \mathrm{Si}\)) gave, after quenching and low (\(225^\circ\)) tempering, a slower decrease in impact toughness as the testing temperature was lowered.

The experimental data of S. E. Belyaev concerning metals of the first group agree in the main with the results of other authors who investigated carbon steels.

The impact toughness of metals of the second group (copper, aluminum, and some of its alloys, austenitic steels) depends only weakly on the test temperature in comparison with metals of the first group. It is interesting that alloys with a face-centered cubic lattice have the greatest impact toughness in the temperature range from −20° to −100°. Some cast aluminum alloys somewhat reduce their impact toughness at temperatures from −40° to −70° (by 15–20%) in comparison with the impact toughness at room temperature.

The impact toughness of metals of the third group changes with decreasing temperature in the same way as that of metals of the first group. The only difference is that the region of impact brittleness occurs at temperatures higher than room temperature.

S. E. Belyaev explains this by the fact that the hexagonal crystal lattice has only one possible slip plane (the basal plane), in contrast to cubic space lattices, which possess a considerably larger number of possible slip planes.

In static tensile tests it was found that, in metals of the first group, as a rule, the following increase with decreasing test temperature: the ultimate strength, the yield and proportional limits. Under the same conditions the relative elongation either remains unchanged or increases somewhat. The reduction of area decreases.

In metals of the second group, the resistance to fracture increases with decreasing temperature more rapidly than the yield point. The relative elongation increases, although in the case of annealed aluminum it decreases. Cast aluminum alloys give varied results: both an increase and a decrease in strength and toughness are observed at lowered temperatures, depending on the composition and the structural features of the various cast aluminum alloys.

In metals of the third group, their mechanical properties change analogously to the change in the properties of metals of the first group—the yield point increases more rapidly with falling temperature than does the change in the ultimate strength. But at the same time the relative elongation and the reduction of area decrease very sharply.

A series of fatigue tests at −70° gave very interesting results. As already mentioned in our work, and as should be expected on the basis of theoretical premises, the fatigue limit of pearlitic steels increases when the test temperature is lowered. Indeed, the fatigue limit almost always increases with increasing elastic limit, and this is precisely what is observed when the test temperature is lowered.

In Chapter IV S. E. Belyaev considers the influence of heat treatment, duration of cooling, the influence of repeated cooling, and the size of the austenitic grain on the properties of certain alloys and steels at low temperatures.

The author comes to the conclusion that a sorbitic structure, obtained by quenching and high tempering of pearlitic steels (U4 and Kh3N), does not exhibit significant brittleness at low temperatures. Prolonged and repeated cooling to low temperatures has absolutely no effect on the properties of alloys at ordinary temperatures, if the cooling does not cause any structural transformations (decomposition of supercooled austenite, for example).

A change in the size of the natural grain of steel has a rather strong effect on impact toughness at low temperatures. Fine-grained steel (grain No. 8) shows, at −70°, an impact toughness 40% higher than that of coarse-grained steel (grain No. 2).

The study of the kinetics of austenite decomposition led the author in Chapter V of the work to the conclusion that, in semiaustenitic steels, after quenching and subsequent cooling (to \(-40\) and \(-195^\circ\)), decomposition of austenite is observed, which leads to an increase in strength and a decrease in impact toughness.

From this brief account of S. E. Belyaev’s very interesting and thorough work, it is evident that, on a large body of experimental material, he essentially confirmed the conclusions we made in the first part of our work on the cold brittleness of carbon steel.

S. E. Belyaev, having at his disposal extensive experimental material, did not set himself the task of elucidating the causes of cold brittleness or finding the mechanism of brittle fracture. He merely indicates that “the nature of the metal is the principal factor determining its behavior under various temperature conditions.”

As was mentioned in the presentation of the cited work, S. E. Belyaev presented several microphotographs of steel near the fracture. Analyzing two of them, he comes to the conclusion that, on the polished section of Armco iron (Fig. 14 in S. E. Belyaev’s book), slip lines on the crystals are clearly visible, whereas in brittle fracture at \(-195^\circ\) the metal grains do not undergo plastic deformation (Fig. 15). Unfortunately, these microphotographs are printed on poor paper, and it is impossible to examine many details of the structure. But it seems to us that in Fig. 14 bent twins (“Neumann lines”) are quite clearly noticeable, and in Fig. 15 there is a considerable number of twins formed in large ferrite crystals. As we shall see below, such an assumption is quite natural in view of the structural features of the metal under consideration and the character of the deformation to which it was subjected.

In the review of the literature given in S. E. Belyaev’s monograph, references are made to a large number of works published before 1938. On the basis of this review the author comes to the conclusion that “the experimental data of many investigations do not make it possible to explain the nature of brittle fracture of metals at low temperatures.”

It may be noted with great satisfaction that, during the last five years, the most important and theoretically interesting works have belonged to Soviet investigators.

A particularly large number of fundamentally important studies during this time were published by N. N. Davidenkov and his collaborators. In his book [11], N. N. Davidenkov gives an exhaustive review of the literature on the question under consideration.

As a result of these experiments, all the main propositions expressed in our investigation [13] are likewise confirmed.

The temperature of brittle fracture of metals, whose impact toughness is strongly influenced by the temperature regime of the test, depends, under dynamic loading, on their composition, structure, the size and shape of the specimen, the condition of its surface, and the rate of deformation. If the combined influence of all these factors is directed in a favorable direction with respect to lowering the temperature of brittle fracture, then this temperature may shift (and quite considerably) toward lower temperatures. As N. N. Davidenkov indicates in his monograph [11] (p. 74), changing the conditions of surface treatment of specimens for dynamic testing can shift the mean temperature of the critical brittleness interval from \(-81^\circ\) to \(-163^\circ\).

In general, N. N. Davidenkov points to the dependence of cold brittleness on:

a) factors determining the shape and dimensions of the part;

b) mechanical factors determining the type of stressed state of the part;

c) the factor of the rate at which the given part is deformed;

d) technological factors determining the mechanical and thermal treatment and the condition of the surface of the part;

e) chemical factors determining the composition of the alloy.

It is obvious that, for a given part operating under definite conditions (characterized by the constancy of the shape factors and of the mechanical factors—a and b), under loads of a more or less known character (the rate factor—c), only factors d and e remain variable.

If it is accepted that the condition of the surface is determined by the method of manufacture of the given part and by its mechanical (and partly thermal) treatment, then the only variable factors are the composition and structure of the steel.

In our book [13] (pp. 127–128, Vol. VII) it is stated that the work “makes an attempt to explain the cause of the cold brittleness of iron and its alloys with carbon as a function of their internal structure.”

In the proposed study, which is the second part of the work, this attempt is somewhat broadened on the basis of materials that have appeared recently, and of our own experiments.

II. Hypotheses on the Nature of Cold Brittleness

N. N. Davidenkov [11] says that, “despite the abundance of experiments devoted to questions of brittleness, we still cannot make a definitive choice among the three hypotheses that are proposed for its explanation.” These hypotheses were based on experiments with crystals that in most cases have ionic lattices (for example, rock salt), and a direct transfer of these hypotheses to metallic (electronic) lattices is obviously impossible. But, undoubtedly, the study of nonmetallic crystals can serve as a first approximation in solving the basic question—the cold brittleness of iron crystals, which possess an electronic lattice.

1. A. F. Ioffe’s Hypothesis

The hypothesis proposed by A. F. Ioffe is based on the fact that brittle strength ($\sigma_t$) does not depend on temperature. The brittle strength of rock-salt crystals, as A. F. Ioffe’s experiments showed, is practically constant over a very wide temperature range: from −190° to 700°.

M. Georgiev and E. Schmid [31] found that the brittle strength of bismuth, having a fracture plane (111), is practically constant within the temperature range from −80° to +20°. The same was obtained by Sauerwald, Schmid, and Kremer [34] for iron single crystals at −90° and −191°.

On the other hand, the strong influence of temperature on the yield point $\sigma_s$ is known. In rock-salt crystals the value of $\sigma_s$ varies from 900 g/mm² to 20 g/mm² in the temperature range from 0° to 800°. The yield point of metallic crystals also depends very strongly on temperature.

In temperature–stress coordinates, the brittle strength of a rock-salt crystal ($\sigma_t$) will be represented as a straight line parallel to the temperature axis, while the yield point ($\sigma_s$) on the same graph is represented by a curve falling in the direction of increasing temperatures (Fig. 1).

The temperature at which brittle fracture begins to occur is determined by the point of intersection (s) of these curves. In this connection it is necessary to bear in mind the definite dependence of brittle strength on temperature, demonstrated by later studies [12]. Therefore, strictly speaking—

on the contrary, the values of $\sigma_T$ in Fig. 1 are expressed not by a straight line, but by a certain curve decreasing toward higher temperature; however, this fall is not great and does not strongly affect the displacement of the point of brittle fracture ($s$) to the right.

A. V. Stepanov [2] somewhat developed A. F. Ioffe’s hypothesis. The low strength of ionic-type crystals is explained, as is known, by A. F. Ioffe by the presence on the surface of crystals of defects, chiefly cracks. The cracks were formed in the process of growth of the crystal, during the preparation of specimens from it, or arose in the process of deformation itself. The now classic experiment of removing these surface flaws from the surface of a rock-salt specimen by dissolving the specimen in water showed that the strength of a crystal can be increased many times over when they are absent or their number is reduced.

A. V. Stepanov considers such specific surface defects, arising on the specimen, to be surface disturbances that appear as a result of slip planes emerging at the surface. But these defects of the specimen surface, formed as a result of its intersection by slip surfaces, cannot by themselves cause the appearance of cracks—the beginning of fracture—if previously formed cracks or scratches were not already present on this surface. It is also very probable that such nuclei of incipient fracture of the surface layer at the place where slip surfaces emerge may be distortions of the crystal lattice due to one cause or another (growth defects, mechanical actions). A number of experiments proved that such an addition to A. F. Ioffe’s original assumptions fully corresponds to reality. But it is still impossible, in the absence of sufficient experimental material, to say quite definitely whether this scheme is applicable in the case of deformation specifically of iron crystals.

Fig. 1

The second hypothesis used in an attempt to explain the cold brittleness specifically of iron crystals is:

2. The twinning hypothesis

Metal crystals with a body-centered cubic lattice are prone to the formation of mechanical twins, whereas rock-salt crystals under identical experimental conditions do not produce twins.

O’Neill [33], determining the slip surfaces in single crystals made of iron containing 1.8% Si, came to the conclusion that the slip planes in iron are the planes $[011]$. Near the place of rupture on the surface of the specimen he found the emergence of twins or, as they are still customarily called, Neumann lines. In O’Neill’s opinion the twinning plane is the plane $[112]$.

Twins arise during deformation in iron crystals that have become brittle for one reason or another. Brittle under com-

at room temperature are large crystals of iron containing silicon in amounts greater than usual (for example, transformer iron). Phosphorus acts similarly to silicon. In such brittle crystals, and under static loading, twins are readily formed. In large crystals of more or less pure iron, twins appear as a result of static deformation only at low temperatures. The larger the crystal, the higher, relatively, is the temperature at which static deformation leads to the appearance of twins. In polycrystalline pure iron, twins are obtained only under impact loading. Such a comparison of the conditions of twinning suggests the existence of a quite definite connection between the brittleness of iron crystals and the appearance of mechanical twins in these crystals. Evidently, there is a dependence between cold brittleness, defined as the brittleness of metals at lowered temperatures as a result of impact loading, and the appearance of twins. Mechanical twinning of iron crystals is facilitated by lowering the temperature of the experiment.

Cold brittleness is found in metals inclined to the formation of mechanical twins (iron, zinc, tungsten). Crystals with a face-centered lattice do not give mechanical twins so readily; at the same time they are not brittle at low temperatures.

Thus, there is a similarity in the tendency toward brittleness at low temperatures between the cold-brittle ionic crystals of rock salt and the cold-brittle crystals of iron.

Externally this similarity will appear still more complete if one considers the course of the curves $\sigma_s$ and $\sigma_T$ for these materials. But a more attentive acquaintance with the phenomenon, a more detailed analysis of the deformation process, show that the similarity remains only superficial. In the deformation of rock salt in the cold-brittle state the principal deformation will be slip, whereas in the deformation of iron in the cold-brittle state the principal deformation is twinning.

III. Fracture of Iron Crystals and Mechanical Twinning

Single crystals of iron can be made brittle under impact as a consequence of lowering the temperature or increasing the rate of deformation. Brittleness is intensified when the composition of the iron is changed (addition to the metal of phosphorus, silicon, nitrogen, hydrogen, etc.). The fracture of single crystals under dynamic deformation proceeds along the plane (100), which is the cleavage plane for $\alpha$-iron [26]. K. A. Bessonov [3] investigated the brittle fracture of large ($5 \times 5 \times 50$ mm) single crystals prepared by recrystallization from Soviet Armco iron. Fracture was produced on a Charpy-type pendulum at the temperature of liquid air and in all cases occurred along the plane [100].

P. A. Bryantsev [6] determined the fracture plane on coarse-crystalline specimens of transformer steel (with Si content $= 4.1\%$) and low-carbon steel. In the latter, as a result of overheating, a very large and brittle grain had grown. The crystallites in these specimens were so large, and the fractures so well expressed, that it was entirely possible to measure with a goniometer the angles between the fracture surfaces.

On the basis of such measurements and determinations of the indices of the faces, P. A. Bryantsev comes to the conclusion that fracture of $\alpha$-iron crystals occurs predominantly along the cube plane and occasionally along the face of the rhombic dodecahedron and the plane (123). The planes of slip and fracture in this case do not always coincide. Sometimes coincidence is observed of the slip and fracture planes with indices (110) and (123). Along the cube face (100) slip does not occur, but it is in the majority of cases

(according to P. A. Bryantsev’s data) as the plane of fracture. Of all the cleavage planes in crystals of transformer iron, 89% fall on the (100) plane, and correspondingly 76% in the case of recrystallized specimens.

E. Shevandin [23] investigated the development of a brittle-fracture crack in coarse-crystalline (average grain size 0.1 mm) iron containing 0.1% C and 0.25% P, and in iron containing 0.6 P (grain size 0.75 mm). To obtain maximum brittleness, notched specimens were fractured at the temperature of liquid air. The crystallographic direction of the crack, which intersected a number of crystals, was not determined, but the author made some observations on the position of the crack and on the changes caused by it in individual grains. In the fracture of an unnotched specimen in liquid air, as was indicated in the work of P. S. Sakharov [20], the crack was accompanied by twins and slips in the crystals. E. Shevandin found in the crystals near the crack only twins. These were, evidently, mechanical twins arising as a result of the impact. From the works of E. Shevandin and P. S. Sakharov there follows the basic conclusion that a brittle-fracture crack passes through crystals, and not between them.

In the microphotograph presented by E. Shevandin it is seen that the crack has a discontinuous character, being divided into separate sections that occupy several grains. At the end it passes into small cracks lying nearby and situated within the limits of one grain. Twins are found near the crack, often running parallel to the crack. In some cases it makes a series of bends parallel to the twins. The author supposes that some twins may turn into a crack. This is quite probable, since sometimes, during twinning in various directions, there may appear peculiar cracks—channels that increase the volume of the twinned crystal [10]. In some crystals the crack intersects twins. Therefore E. Shevandin presents the mechanism of formation and development of the crack as follows. First, during deformation, twins are formed; then one of them turns into a crack. With further deformation the cracks first formed in the grain cross a peculiar barrier—the boundary between grains—and turn into one large crack. A twin gives “life” to the first crack, and thereafter it can proceed without its companion—the twin. That is why twins were not observed along the entire length of the crack. The author, having constructed such a scheme, nevertheless acknowledges that he cannot say whether the twins appear earlier than the formation of the brittle crack, but believes that the twins create overstress in the metal. The appearance of a crack may be due to inhomogeneity of the material (local segregation, foreign inclusions), but, probably, at the site of such inhomogeneity a twin is first formed, producing stresses that already cause the appearance of the crack. Under ordinary conditions deformation in iron proceeds by way of slips; when slips are impossible, twins arise, and they are what cause brittle fracture.

As a result the author arrives at the following conclusions: twins are indispensable companions of brittle fracture of iron; twins are probably formed earlier than the brittle crack; brittle fracture in a polycrystal propagates by a discontinuous, not a continuous, front.

The twinning plane in crystals of $\alpha$-iron is the plane (112), i.e., the face of the tetragonal dioctahedron [26].

The twins, first discovered in meteoritic iron by Neumann, leave traces on the plane of the polished section—straight thin lines, and these lines are sometimes called Neumann lines. Although it should be noted that there are still some disagreements about the nature of Neumann lines. There are indications that Neumann lines may form parallel to the plane

of the cube [7]. Although the orientation of Neumann lines within a single crystallite is not random, their symmetry with respect to the principal crystal—the first indication of a “true” twin—has not yet been proved.

K. P. Bunin used for his work welding iron with a content of 0.275% P. Etching of polished sections with a solution of nitric acid made it possible to reveal, on individual crystallites, etch figures in the form of squares. If on such a grain there are Neumann lines—traces of twins on the plane (100), then obviously they must coincide with the trace of the plane (112) on the plane (100). On the face of the cube there may be four directions of such traces, intersecting at definite angles with respect to the square etch figure. On the basis of his investigations K. P. Bunin comes to the conclusion that the Neumann lines are apparently symmetrical with respect to the principal crystal.

P. A. Bryantsev [5] does not agree with such an identification of Neumann lines with traces of twins on the plane of the polished section cutting the given ferrite crystallite. As the author says, these line-like formations, which do not disappear upon repolishing, are “traces of planes running inside the grain, along which the grain has been destroyed.” As the author asserts, Neumann lines differ from actual twins in that they break off inside the grain. This testifies to an identical crystallographic orientation of both the principal grain and these lines. A “true” twin, if it does not reach the boundary of the grain, has its own boundary separating it from the principal crystallite. Twins are colored differently during etching than is the principal grain. The slip lines arising during plastic deformation change their direction when they meet twins, and do not change direction when crossing Neumann lines. Rejecting, on the basis of the considerations cited, the twin character of Neumann lines, P. A. Bryantsev refers to the opinion of E. O. Paton and P. P. Bushtedt [18], who regard Neumann lines as deformation lines having nothing in common with slip lines caused in the metal by stresses above the yield point, and also to the opinion of P. V. Sakharov [19], who calls such structural formations slip lines.

As a result of such lack of clarity in the definitions cited, P. A. Bryantsev considers it necessary to propose a uniform terminology and a correct interpretation of the phenomenon described. In his opinion, it is necessary to call slip lines those lines which arise during plastic deformations and disappear upon repolishing of the specimen; Neumann lines—“lines which do not disappear upon repolishing of the specimen and testify to destruction of the specimen along certain planes”; and, finally, twin lines—lines corresponding to twin planes present inside the grains. The author at the same time makes the observation that “the crystallographic lattice has the same arrangement on both sides of slip and Neumann lines and a different arrangement in twin formations.”

Incidentally, if P. V. Sakharov [19] (p. 6) quite definitely calls Neumann lines slip lines, then E. O. Paton and P. P. Bushtedt [18] (p. 138), after citing several contradictory statements by different authors on the nature of these lines, say that deformation lines have much in common with slip lines arising in iron under stresses above the elastic limit. Their main difference from slip lines, in the authors’ opinion, consists in the impossibility of destroying them by repolishing the specimen. In addition, deformation lines are more rectilinear in comparison with slip lines and are arranged in parallel bundles. The authors were unable, despite all their efforts, to obtain deformation lines in wrought iron, whereas slip lines are obtained very easily.

R. Schaefer [35] presents a photograph (Fig. 2) in which “sharp lines extending from grain boundaries, called ‘slip lines,’ are clearly visible.” The author further notes that there is no consensus regarding the cause of the appearance of these lines, and that their influence on the properties of iron is unknown.

The presence of deformation lines indicates certain changes in the properties of the metal, namely: the tensile strength increases, while the relative elongation decreases. According to the experiments of E. O. Paton and P. P. Bushtedt, the average tensile strength for specimens with deformation lines is 34.3 kg/mm², whereas for specimens without these lines it is 32.3 kg/mm². Correspondingly, the yield point increases from 23.4 kg/mm² to 24.7 kg/mm², while the relative elongation decreases from 10.5% (specimens without lines) to 8.9% (specimens with deformation lines). Studying 29 specimens of weld iron that had not been in service, the authors found deformation lines in 10 cases, namely in large ferrite crystals, and concluded that the appearance of deformation lines is a consequence of the peculiarities of the process of manufacturing bridge iron, the result of mechanical and temperature effects. As we have already mentioned, the authors were unable experimentally to obtain these lines on iron specimens. Cold forging under a hammer produced shear lines, but not deformation lines. As the authors say in the conclusion to their interesting study, “we were unable by any mechanical or thermal treatments artificially to produce such lines.”

Fig. 2 ×80

Unfortunately, the cited study contains no chemical analysis of the bridge iron. Perhaps a probable explanation of the causes of the appearance of deformation lines lay in a possible anomaly of the chemical composition.

D. B. Gogoberidze, in his monograph [10] (p. 20), says that Mathewson and Esmond, on the basis of a study by Laue’s method of a ferrosilicon crystal, established the identity of Neumann lines and twin formations. The twinning plane is the plane [112].

As can be seen from the statements cited regarding twins in iron and their identity with Neumann lines, or deformation lines, there is no unanimous opinion concerning their nature, the mechanism of their formation, or their influence on the properties of iron.

It is quite possible that this situation is explained by the fact that the original terminology was borrowed from mineralogy, where Neumann lines and the Widmanstätten structure were first discovered in sections of meteoric iron containing up to 80% nickel. And the expression [4] “one also distinguishes Neumann (or Braunau) lines, caused by thin lamellar twin intergrowths along the trioctahedron [211] into the principal individual” apparently applies equally to α-ferrite and to austenite.

Some authors (Taylor and Elam [36], Carpenter [21]) believe that in general it is impossible to speak quite definitely about slip planes in iron crystals, and that the mechanism of slip itself is not sufficiently clear.

All this, taken together, makes the twinning hypothesis, as a hypothesis claiming to explain the essence of the phenomenon of cold brittleness, somewhat unclear.

Since, nevertheless, in the brittle fracture of iron there are observed both slip elements and, apparently, twinning, while fracture proceeds along a cube face, the addition made by A. V. Stepanov to A. F. Ioffe’s hypothesis concerning the causes of the brittleness of iron at low temperatures is obviously also affected by a certain uncertainty. A. V. Stepanov regards as the place where cracks arise in ferrite crystals the traces of plastic deformation passing through the crystal on its outer surface and the interaction of these traces with already existing surface defects. There is a certain uncertainty in the conception of the plastic deformation of the iron crystal, and therefore the very nature of these interactions becomes unclear.

A. F. Ioffe’s hypothesis, represented graphically in the form of the curves $\sigma_t$ and $\sigma_s$, is very simple and clear. Surface defects, scratches, lattice defects, impurities pass into a fracture crack, and the fracture crack in an $\alpha$-iron crystal subsequently proceeds along the cube face [10¹].

In our investigation [13] we carefully examined a large number of sections prepared from the regions of brittle fractures of various steels tested at very diverse temperatures, and found all those consequences of plastic deformation of the iron crystal which are discussed in this chapter. We drew attention to the appearance of a considerable number of Neumann lines in cold-brittle crystals of $\alpha$-iron. At the same time we attempted to explain, for example, the influence of aging on the mechanism of cold brittleness as the blocking of certain planes in the crystal, important for the unhindered development of plastic deformations, by finely dispersed particles of a segregate precipitating from the decomposing solid solution of a number of components in $\alpha$-iron. We suggested that the appearance of cold brittleness could also be explained by the emergence of a smaller number of possible slip planes at low temperatures in comparison with higher temperatures, and also by a decrease in the number of arising shears on a definite region of the crystal as the temperature falls. The influence of low temperature and the blocking of the crystal as a result of decomposition of the solid solution are added together, and blocking is to some extent equivalent to a lowering of temperature. Accepting, in essence, A. F. Ioffe’s hypothesis, it is necessary to present the actual mechanism of brittle fracture.

Despite the contradictions and obscurities in the literature cited, it is nevertheless possible to accept certain basic propositions concerning the plastic deformation of $\alpha$-iron crystals.

Metallographers call the real crystals of $\alpha$-iron, visible under the microscope and essentially representing crystals of a solid solution of various components in $\alpha$-iron, ferrite. We shall use this term below. Ordinary ferrite even of very pure technical iron (for example, Armco iron) is a solid solution of carbon, manganese, phosphorus, silicon, nitrogen, hydrogen (and often chromium and nickel) in $\alpha$-iron. The lattice of this solid solution, with a comparatively small content of the listed impurities, remains basically the lattice of $\alpha$-iron (a body-centered cube).

At low rates of deformation beyond the elastic limit and at ordinary room temperature of the experiment, the principal type of plastic change of form is slip. Most investigators [10], apparently, consider the slip planes to be the planes (101), (112), and (123),

and the direction of slip [111]. We [14] determined in ferrite crystals containing about 0.1% N that the slip plane is the face of the rhombic dodecahedron (101). O’Neill [33] gives another slip plane for silicon iron—(112). In O’Neill’s experiments silicon iron consisted of ferrite single crystals, and all the silicon was in solid solution. Undoubtedly, silicon, like nitrogen in our ferrite crystals, could change the character of the plastic deformation of the ferrite crystal, and the results of these investigations cannot be unconditionally transferred to pure ferrite crystals.

It is impossible to cause the appearance of Neumann lines in pure ferrite by static deformation at ordinary temperature. On O’Neill’s specimens they were found under ordinary static tension, and he discussed rather thoroughly the question of their direction, their influence on the character of fracture, etc. On the other hand, E. O. Paton [18] could not obtain them either under static or under dynamic deformation, working with iron that undoubtedly contained not very much silicon. It is very likely that the slip lines near the impression of a ball on a large crystal of silicon iron (with 4% silicon) cited by P. Ferens [34] are in fact traces of twins on the (100) plane, judging by their character and arrangement (Fig. 3). Under static deformation and at room temperature in the experiment, more or less pure ferrite crystals do not form twins and give only slip.

Fig. 3

Wishing to determine the temperature at which twins appear in polycrystalline iron in the case of impact pressure by a ball, E. M. Shevandin [24] struck a specimen with a ball at a speed of 2 m/sec at various temperatures. It turned out that around the impression obtained under static pressure by a ball, twins appear only at temperatures below −170°, whereas under dynamic pressure at a speed of 2 m/sec they arise at ordinary temperature. Objecting to P. A. Bryantsev, who asserted that Neumann lines are not traces of twins on the polished surface of a crystallite, he proves that one is dealing with real twins and gives a number of convincing photographs.

Thus it may be said that, along with slips and independently of them, in ferrite crystals of suitable chemical composition, either at a considerable rate of deformation or at low temperature, twins appear (mechanical twins). It is somewhat unclear when twins appear—simultaneously with slip or, preceding brittle fracture along the ferrite crystal plane (100), they preclude by their appearance the occurrence of simple slip. And finally: in the most perfect crystals with respect to cleavage (for example, ionic—rock salt, calcite, or electronic—bismuth, antimony), the cleavage plane formed during the destruction of the crystal is not an ideally mirrorlike plane. Its shape is affected both by the imperfection of the real crystal and by the conditions of its formation. If the nucleation of the developing cleavage plane in ferrite crystals really begins at the place where a twin formation emerges on the surface of the specimen, then this inevitably affects its configuration.

All these considerations about twins and their role in solving the problem of the cold brittleness of steel prompted us to carry out certain experiments, described in the following sections.

IV. Selection of material for specimens and preparation of single-crystal and coarse-crystalline specimens

In steel consisting of pearlite and ferrite, the “carrier of cold brittleness,” if one may put it this way, is ferrite—a multicomponent limiting solid solution with the lattice of $\alpha$-iron. The crack of brittle fracture passes through the crystals of this solid solution. But the critical temperature of cold brittleness in pearlitic–ferritic steels depends very strongly on the chemical composition of the steel. Thus S. E. Belyaev [2], we, and some other authors [30] have shown that an addition of 1% manganese to steel shifts the temperature of brittle fracture by several tens of degrees toward lower temperatures, thereby reducing the cold brittleness of the metal. Vanadium, the combination of chromium and silicon (chromansil), nickel, nickel and chromium together, molybdenum, and other alloying elements act very beneficially in this direction. All these additions form solid solutions with $\alpha$-iron within the limits in which they are usually added to pearlitic structural steels. Under the influence of these additions the structure of pearlite also changes; it becomes finer-grained, approaching sorbite in character. But in pearlitic and sorbitic steels the brittle-fracture crack likewise passes through the ferrite grains, and the decrease in cold brittleness under the influence of alloying elements is due to a decrease in the brittleness of ferrite at low temperatures. The size of the austenitic grain of the steel has a strong influence on the change in cold brittleness; steel with a fine natural grain is less cold-brittle than coarse-grained steel. However, it is not clear how the size of the austenitic grain affects the nature of deformation of the actual ferrite crystal at low temperatures.

Therefore, for the investigation we selected steels with various alloying additions (Table 1), and some of them with different austenitic grain sizes.

Table 1

Steel designation	C %	Mn %	Si %	Cr %	V %	S %	P %	Ni %	Remarks
1	0,19	0,88	0,10	—	—	0,08	0,16	—
2	0,02	0,60	0,12	—	—	0,06	0,05	—
3	0,20	0,15	—	—	—	0,06	0,04	—
4	0,60	0,80	2,0	—	—	0,05	0,04	—
5	0,17	0,59	0,25	0,75	—	0,03	0,04	1,90
6	0,50	—	—	0,95	0,29	—	—	—

We wished to investigate the phenomenon, as far as possible, in its pure form and to try to determine what role slip and the appearance of twins play in the occurrence of brittle fracture. For this purpose, from all the steels listed we attempted to prepare single crystals and coarse-crystalline specimens.

It is in principle possible to prepare single crystals from a two-phase alloy in the case where one of the phases is present in a small amount compared with the other. We succeeded in obtaining single crystals from tin and zinc with an admixture of components forming with tin and zin—

by the eutectic method, but by Shubnikov’s somewhat modified method; however, from pearlitic steel, even with a low carbon content, single crystals of any considerable size do not grow. To make single crystals from the steels listed, it was necessary first to decarburize them to such an extent that there would be no carbides in the structure (neither structurally free nor in pearlite), and that the carbon remaining after decarburization would all be in the solid solution of ferrite. In the ferrite lattice of ordinary steel at room temperature there may be 0.015% C. As experiments have shown [9], after prolonged decarburization in hydrogen, carbon remains in the solid solution—ferrite—in an amount of about 0.02%. Decarburization of steel deprives it of pearlite, but does not change the carbon content in ferrite and, evidently, does not affect the composition and properties of this structural constituent. There are no exhaustive data on the carbon content in the ferrite of alloyed steels of the pearlitic class, but it may be assumed that it fluctuates within approximately the same limits.

A very convenient method for obtaining ferrite single crystals is the method of Edwards and Pfeil [9]. The steel is first decarburized for 48 hours at 875° in a stream of moist hydrogen, then held at 1000° for twelve hours and cooled over a period of 12 hours. After this the specimens are given a tensile deformation of 3.25% and recrystallized for 72 hours at 880°. The authors developed this procedure for ordinary steel with a content of 0.13% C and a normal content of other impurities (only Si was less than 0.02%). By this method O’Neill makes single crystals from iron containing 18% Si.

At 875° carbon can combine directly with hydrogen, forming chiefly methane. The process of decarburization of steel in hydrogen proceeds with the formation of methane. The process takes place at the interface between the solid and gaseous phases. Methane is obtained on the surface of the specimen, to which carbon atoms from the interior of the specimen continually arrive. It is hardly possible to suppose that methane is formed inside the steel by diffusion of hydrogen into the metal. Moreover, the large methane molecule (CH\(_4\)) cannot easily make its way through the lattice of \(\alpha\)-iron to the surface and join the gas phase. For rapid decarburization it is necessary to have a gas phase sufficiently rich in hydrogen. From the equilibrium condition [27] of mixtures of H\(_2\)—CH\(_4\) over austenite (at atmospheric pressure) it follows that the optimum temperatures for decarburization of a homogeneous austenite system lie on the line GOS, and the completeness of decarburization increases with increasing carbon content in the solid solution (Fig. 4).

Fig. 4

Considering the diagram (Fig. 4), we come to the conclusion that, at a content of C = 0.08% and a temperature of 880°, the value of \(\log \frac{P^2_{\mathrm{H}_2}}{P_{\mathrm{CH}_4}}\) is 2.52, and this corresponds to a partial pressure of hydrogen in the gas phase of 0.9945 atm at a total pressure of 1.0000 atm. Thus in the gas pha-

it is necessary to have 99.45% \(H_2\) in order to obtain, when decarburizing steel in a stream of hydrogen, a carbon content of about 0.08%.

To accelerate the process of decarburization of steel specimens and to reduce the final carbon content in them, it is necessary to use as rich a hydrogen gas mixture as possible and to lower the temperature in order to increase the concentration of carbon in austenite. But at low temperature the diffusion of carbon in iron slows down, and the decarburization process becomes excessively prolonged. At very low temperatures a continuous network of ferrite will surround islands of austenite and will prove a serious obstacle for carbon atoms moving toward the periphery to meet hydrogen.

True, one may suppose that at low carbon concentrations in \(\alpha\)-iron (within the range of formation of the solid solution \(\alpha\)), the rate of diffusion of carbon in it is the same as in the \(\gamma\)-solution (at the same low concentration), but at low temperatures this rate will in general be insignificant.

The diffusion coefficient of carbon in iron varies, according to the data of B. D. Averbukh and P. I. Chufarov [28], as follows: \(D_{600^\circ}=1.42\cdot10^{-8}\ \text{cm}^2/\text{sec}\), \(D_{800^\circ}=1.69\cdot10^{-8}\ \text{cm}^2/\text{sec}\), and \(D_{900^\circ}=7.4\cdot10^{-8}\ \text{cm}^2/\text{sec}\). Therefore decarburization has to be carried out at temperatures near the GOS line, although in doing so the value of \(\log \frac{P^2_{H_2}}{P_{CH_4}}\), which determines the consumption of hydrogen, also increases.

The decarburizing action of moist hydrogen, according to the data of many authors [25], is stronger than that of dry hydrogen. Hydrogen saturated at \(18^\circ\) with water vapor decarburizes faster than dry hydrogen, and the rate of decarburization increases with an increase in the water-vapor content. But in the decarburization of some alloy steels the picture changes markedly; the presence of water vapor in the hydrogen retards the removal of carbon from silicon transformer steels. B. D. Averbukh and G. I. Chufarov [1] explain this by the formation on the surface of the metal of a protective oxide film (\(\mathrm{SiO_2}\)) or iron silicate.

We moistened the hydrogen with water vapor either directly or by admixing oxygen to the gas in such quantities that, after combustion of the mixture in the furnace, the required amounts of water vapor were obtained.

In agreement with the literature data, steel No. 4 was decarburized in dry hydrogen faster than in moist hydrogen; steel No. 5 behaved in the same way, and in general it was not possible to decarburize it by heating for 300 hours. Evidently in this case a surface layer of \(\mathrm{Cr_2O_3}\), strongly impeding diffusion, is of great importance. The poor decarburization of steels with a moderate chromium content is indicated by V. Baukloh and G. Guthmann [28]. According to their data, in steel containing 0.57% Cr and 0.42% C, during decarburization in hydrogen at \(1000^\circ\), the carbon content decreased by only 1%, whereas carbon steel with 0.24–0.60% C was decarburized under the same conditions by 90–93%, vanadium steel (0.19% V and 0.11–0.13% C) by 84%, silicon steel (3.05% Si) by 87%, and manganese steel (1.03% Mn) by 96%. We were not able to decarburize steel No. 5 to a sufficient degree and to prepare single-crystal specimens from it.

For decarburization the specimens were placed in a U-shaped tube of mild steel; the bent part of the tube was located in the muffle of an electric furnace with automatic temperature control. The tube was protected from oxidation by a protective coating. The specimens were forged from bars of the corresponding steels, machined on a milling machine to the required dimensions \((5 \times 5 \times 140\ \text{mm})\), and ground. Specimens of steel No. 4 were made cylindrical \((d=6\ \text{mm},\ l=120\ \text{mm})\) with heads for

gripping by the clamps of the tensile-testing machine. Purified hydrogen entered one limb of the U-shaped tube; from the other there emerged a mixture of hydrogen and methane, discharged into the air through a water seal.

After a number of experiments we settled on the following decarburization regime, which gave very good results: heating for 120 hours at 900° and then 80 hours at 800°; cooling at a rate of about 40° per hour. The amount of hydrogen passed through was calculated on the basis of the considerations given above and was expressed in hundreds of liters per 100 grams of specimen weight. No traces of carbides were found in the steels after decarburization.

After annealing in hydrogen, no carbon analysis was made, since on the basis of previous investigations [9] its content in the metal could be assumed to be less than 0.02%. In any case, all the carbon was dissolved in the ferrite, and etching with sodium picrate did not reveal traces of cementite.

The process of decarburization of steel takes place at its surface; but hydrogen, dissolving readily in ferrite, remains in the specimens after they have cooled, and all the properties of the specimens will therefore be considerably altered.

Moreover, according to G. Schenk [25], the pressure of hydrogen evolved from a solid solution in iron increases inversely proportional to the square of the equilibrium constant, which decreases as the temperature falls.

The pressure of evolution of hydrogen from the solid iron–hydrogen solution increases during cooling and, if the hydrogen content in the metal is sufficiently high, may even exceed the strength of the steel. Considerable internal stresses will arise in the specimens, and internal cracks may even appear. It is not without reason that many investigators connect the formation of flakes with the presence of hydrogen in steel. At the same time, there is the opinion [25] that prolonged heating of steel, accompanied by plastic deformation, promotes the evolution of gas from ferrite and a reduction of internal stresses. Nor can the possibility of the formation, in the ferrite lattice, of methane molecules be ignored.

According to Bodenstein [4], molecular hydrogen does not dissolve in steel and does not diffuse through steel. If hydrogen does not diffuse, then diffusion of the complex molecules of methane, which may form inside the metal when dissolved hydrogen meets carbon atoms, is hardly possible. This circumstance, as well as formation through the reduction of oxides of water molecules, undoubtedly also leads to an increase in internal stresses.

As the experiments of Gr. Weinberg and S. I. Proshutinsky [9] showed, hydrogen retained in “Armco” iron during rapid cooling reduced the relative elongation from 31.8% to 19.9%. The authors observed that the effect of hydrogen on specimens becomes the stronger the longer they had been in it (at 1000°). If this time is short (2–4 hours), then exposure to hydrogen has almost no effect on the properties; but, beginning with 8 hours, it is manifested very strongly and increases with increasing duration of heating.

True, the effect of hydrogen on the properties of “Armco” iron is very small after slow cooling of the specimens in a hydrogen atmosphere together with the furnace; but we were dealing with alloy steels, whose relation to hydrogen, judging from the formation of flakes in them precisely under the influence of hydrogen, is somewhat different from that of “Armco” iron.

There are a number of indications that correct annealing promotes complete removal of hydrogen from steel. Therefore, after decarburization the specimens were annealed for 10 hours at 950° in vacuum (0.01 mm mercury col-

6a); and in the same vacuum they were slowly cooled. Since a very intense evolution of hydrogen occurs during the transformation \(\gamma \to \alpha\), the interval \(950^\circ\)—\(750^\circ\) was passed through at a rate of \(30^\circ\) per hour. Cooling from \(750^\circ\) to room temperature lasted about 24–30 hours. In order to be certain that there was no harmful influence of hydrogen dissolved in the specimens, the control specimens, after annealing, were normalized (from \(950^\circ\)) and tested in tension. The relative elongation of specimens of steels 1 and 6 (over a length of 10 diameters) proved to be 30–31%, which evidently indicates the absence of any influence of dissolved hydrogen on the ductility.

After annealing, the ferrite grain in the alloy steels proved to be of a different size compared with the grain which Edwards and Pfeil [29] consider normal for obtaining single crystals by their method. Vanadium steel, in particular, gave a very fine grain. For a single crystal to be able to grow during recrystallization, it is necessary that the initial grain size in the specimen be as uniform as possible. In all specimens of one grade of steel the grain, after the treatment described, was very uniform. But the grain size differed in the different steels, and this made it impossible to establish, for all specimens, the same optimum residual elongation required in order to obtain the smallest possible number of centers of recrystallization during annealing. The finer the grain, the greater the deformation required for the onset of recrystallization at a given temperature. We now have very extensive experimental material on the production of single crystals by recrystallization (see, for example, the review by E. Yakovleva [27]) from more or less pure iron, but there are no exhaustive data on the recrystallization of solid solutions of \(\alpha\)-iron with various components (as in our case). We therefore had to impose plastic deformation by tension within the limits from 2.5 to 6.0%, i.e., such as ensures the best formation of single crystals in technical iron with a low carbon content. In specimens with optimum deformation, crystals grew that occupied the entire cross-section of the specimen and had a length of several tens of millimeters.

Specimens in which no single crystals were found after recrystallization were normalized from \(950^\circ\), annealed to relieve internal stresses at \(850^\circ\) for 10 hours with subsequent slow (30–20° per hour) cooling. Then the already known optimum plastic deformation by tension was again imposed on them, and they were annealed for recrystallization.

In this way we obtained an almost one-hundred-percent yield of specimens in which the grown crystals occupied the entire cross-section.

Annealing was carried out in vacuum, and recrystallization in hermetically sealed metal tubes (the tube closure was outside the furnace), with the free space of the tube filled with metal shavings inert with respect to the specimens, which absorbed the oxygen remaining in the tube.

Incidentally, in order to ascertain whether hydrogen was evolved from the specimens after their prolonged annealing in vacuum, a polished plate of pure iron, preliminarily heated until a violet temper color was obtained, was placed next to the specimens during recrystallization. Removed after annealing together with the recrystallized specimens, the plate proved to be white; the violet color had disappeared. Evidently, the iron oxides responsible for the temper color had been reduced to metal by the hydrogen that continued to be evolved. Hydrogen was evolved despite the fact that the specimens had already been annealed in vacuum, cooled from \(850^\circ\) to \(+50^\circ\) over 30 hours, and then, in addition, had lain in the laboratory for about 20 days. We repeated this experiment with the plate during the secondary recrystallization of specimens; [[unclear: text continues on next page]]

that had occurred during the first recrystallization of the single crystals, but the plate remained violet. This once again emphasizes how slowly the last portions of dissolved hydrogen are removed from steel.

The process of recrystallization lasted each time for 70–100 hours. Heating during recrystallization was not interrupted. Some specimens were subjected to triple recrystallization, to preliminary annealing in vacuum to remove hydrogen, were normalized twice, and annealed twice after normalization. In all, counting also the annealing in hydrogen, some specimens were held at a temperature of 900° for about 120 hours, at 800° for about 80 hours, at 860° for about 250 hours, and at 850° for about 20 hours; that is, they were held in the temperature interval 800–900° for a total of approximately 400–450 hours. We shall have occasion to return to this circumstance later. Such prolonged heating evidently caused a very complete diffusion of the components dissolved in ferrite, and undoubtedly affected some properties of the metal. Cooling after recrystallization was as slow as after annealing.

On the surface of the specimens there was often, but not always, a fine-crystalline layer about 0.1 mm thick, covering the single crystals. This layer was dissolved in weak nitric acid and the specimens were ground. The work-hardened layer produced by grinding on emery paper was dissolved in acid; the specimens were then reground on finer (00000) paper and polished on wet cloth with alumina. During etching in acid, hydrogen was liberated and was partly absorbed by the specimens. The hydrogen was removed by heating the metal at 150° for 10 hours. According to literature data, such an operation ensured the evolution of hydrogen from the specimens.

V. Investigation of the Cold Brittleness of Ferrite Crystals

As we noted at the time, we consider A. F. Ioffe’s hypothesis to explain well the essence of the phenomenon. This hypothesis does not aim to present the mechanism of the phenomenon of cold brittleness specifically in ferrite crystals.

The twinning hypothesis, as it seems to us, can in the present case provide a picture of the behavior of crystals in the state of impact brittleness. To obtain the necessary experimental material, it was necessary to break the prepared specimens under conditions in which brittle fracture arises and to study the character of deformation of the specimen.

There are some grounds for assuming that the rate of formation of twins is close to, and perhaps even equal to, the speed of sound in the given crystal [10]. The speed of sound in steel is about ~5000 m/sec [12]. It is obviously impossible to deform a specimen at such a speed. We decided to achieve the maximum rate of deformation attainable under our conditions, namely the velocity of a bullet at the moment it leaves the barrel of a rifle.

The initial velocity of the bullet in the weapons we used is as follows [21]: a Nagant revolver gives 280 m/sec, a rifle (model 1891)—860–870 m/sec, a small-caliber rifle 320 m/sec. The Gerlich rifle gives a velocity on the order of 1400–1800 m/sec. We chose the small-caliber Walter rifle with an increased charge. The impact velocity of its bullet (on the order of 350 m/sec) exceeds the velocity of the striker of all existing pendulum hammers severalfold, but it is nevertheless considerably less than the velocity of propagation of the elastic-deformation wave and the probable rate of formation of twins in a ferrite crystal. The velocity of the small-caliber bullet is not so very different from the velocity of the bullet of a more powerful weapon—a military three-line rifle. We did not have at our disposal a Schiøtz instrument (or a similar one) for determining the velocity of the bullet, and we estimated

the constancy of the bullet velocity in individual shots by the uniformity of the weight of the cartridges and by control shots. From a box of identically loaded cartridges, 40% (i.e., 20 pieces) were selected, and with these selected cartridges shots were fired into wooden boards of, as far as possible, identical construction. From the penetrative power of the bullets they judged the constancy of the charge and the magnitude of the initial energy of the bullet. As a rule, the depth of penetration did not differ noticeably in individual shots, and this gave grounds for assuming that all the remaining cartridges in the box also possessed identical ballistic properties.

The live force of a bullet of weight \(P\) can be calculated by the formula

\[ A=\frac{PV^2}{2g}\ \text{kgm}. \]

With a bullet weight of 2.59 g, the initial velocity was (with the increased charge) not less than 350 m/sec. Therefore \(A \simeq 16.4\) kgm, which is quite sufficient for fracturing a normal specimen. It is also more than sufficient for our single crystals. According to S. A. Buturlin [8], the initial velocity of a bullet weighing 2.59 g with a normal powder charge is 335 m/sec.

The rifle was fastened by means of clamps with rubber gaskets in a stand which, when the screw \((E)\) was turned, could be moved somewhat in the vertical plane (Fig. 5). The stand was fixed on a table, at the other end of which parallel vises were installed for

Fig. 5

fastening the supports for the specimens. The distance from the end of the barrel to the specimen was 800 mm. Obviously, its value is of no great importance, and it was adopted as the minimum distance at which the powder gases do not reach the specimen. Between the barrel and the specimen there was placed a wooden shield with a hole (10 mm) for aiming and for passage of the bullet. Behind the vises, boards of soft wood were installed to catch the bullet and fragments of the specimen.

In the vises, clamps of various shapes were fastened, making it possible to carry out fracture by the Charpy method (the specimen lies freely on two supports—the blow is to the middle of the specimen) and by the Izod method (the specimen is clamped on one side—the blow is to the free end). In Figs. 6 and 7 it is seen what position the specimen \((A)\) occupied with respect to the flying bullet¹ in the Charpy and Izod arrangements. To cool the specimens to the temperature of liquid air, small cups made of paper impregnated with paraffin were put on them. To reduce heat transfer through the supports, a thin paper gasket was placed between the latter and the specimen. The steel supports were case-hardened, quenched, and ground.

Liquid air was poured into the cup fitted onto the specimen. At first very intense boiling took place; then the vigorous boiling subsided and continued with a uniform evolution of small gas bubbles.

Fig. 6

Fig. 7

The loss of liquid air in the cup was replenished continuously. Ten minutes after the cessation of vigorous boiling, the specimen was deformed by firing.

As control experiments showed, the temperature of the ends of the specimen protruding from the cup was of the order of \(-175^\circ\). The temperature of the clamped part of the vertical specimen below the support was of the same order.

After the bullet impact, the cup shattered into pieces 2–5 mm in size. The bullet, retaining a considerable part of its energy, penetrated into the board to a depth of up to 60 mm. Fragments of the specimens, striking the wood, entered it to a depth of up to 30–40 mm, which indicates that their velocity at the moment of impact on the board was of the order of 100–150 m/sec. Upon striking the board the specimens were not deformed, but in order to protect their polished surface from damage a trap made of several layers of cotton wool was provided.

Series of specimens with and without a notch were tested at room temperature and at the temperature of liquid air. The notch was made with a saw 0.23 mm thick and had a rectangular section 0.5 mm deep, with unrounded corners. The distance between the supports was taken as 20 mm for the following reasons. The diameter of the bullet after firing was 5.6 mm; upon impact it may increase in diameter to 6 mm. The maximum thickness of the specimen is 5 mm, and therefore, for the joint free passage of the bent specimen and the bullet, the distance between the supports must be at least 16 mm. The length of the specimen was 25 mm. In fracture according to Izod, in order to preserve—

under approximately the same conditions, the protruding part of the specimen was 13 mm long. In this case the bullet struck at a distance of 4 mm from the end of the specimen. At the place where the bullet struck, an indentation remained on the specimen, about 4 mm in diameter and 0.1–0.2 mm deep. Specimens from single crystals were tested according to the following scheme. The single crystal was cut into two parts. One part was tested at room temperature; the other half, crystallographically oriented in the same way as the preceding one with respect to the line of flight of the bullet, was tested at the temperature of liquid air. One test (two specimens at different temperatures) was carried out on specimens with a notch and another without a notch. If four specimens could not be obtained from a single crystal, then identically oriented specimens (two) were tested at different temperatures.

In many specimens the fourth-order axis coincided very well with the axis of the specimen, and the bullet strike often fell on the (100) plane. Specimens in which the fourth-order axis deviated from the axis of the specimen were oriented in such a way that the fourth-order axis was inclined in the direction of the bullet’s flight. The round specimens of steels Nos. 4, 6, and 7 were oriented in the same way.

As a rule, single-crystal and coarse-crystalline specimens from all decarburized steels, with and without notches, did not break at room temperature, but only bent when tested by Charpy and Izod. The angle between the bent parts of the Charpy specimens was about 100–120°. Upon impact, the support left on the specimen an indentation 0.1–0.3 mm deep. In the Izod test, the specimen that had been struck by the bullet bent by inertia until it came into contact with the support. In doing so it struck the support so strongly that a clearly visible imprint of the support remained on its surface, and even the surface defects of the support (scratches from grinding) could be seen on this imprint.

At the temperature of liquid air the picture was different. Most of the notched and unnotched specimens broke. But some specimens of steels 3 and 6 without notches merely deformed without fracture. The specimens that did not break were those in which large crystals grew only after threefold annealing for recrystallization, that is, those which had been heated for a total of 450–500 hours and, apparently, as a result of such prolonged annealing had reached a structurally equilibrated state. Annealing of such duration, accompanied also by sufficiently slow cooling, evidently also largely relieved the internal stresses that had been present. In addition, in the case of vanadium steel, the vanadium had a beneficial effect on the reduction of cold brittleness by causing the formation of a fine austenitic grain. Most often the specimens of steels 3 and 6 did not break. It should be noted that the austenitic grain size in these steels, determined by the McQuaid-Ehn method, was respectively 5 and 7. Steels 1 and 2 had an austenitic grain size of 3–4. Steel 1 was heavily contaminated with slags, which did not prevent the growth of large grains up to and including single crystals, but evidently shifted the impact brittleness toward higher temperatures. However, a number of single-crystal specimens from these steels also did not break at the temperature of liquid air; such specimens, however, were few.

VI. Discussion of the Experimental Results

The fracture plane of all single-crystal and coarse-crystalline specimens was, in all the cases we investigated (more than 100 fractures), the cube face (100). This plane was determined by the method of pressure figures, etch figures, and by taking Laue diagrams.

Figure 8 gives a photograph of the fracture along the (001) plane of a sharp type III specimen made from a single crystal (steel No. 4); the direction of the blow is shown in the photograph by an arrow. The fracture is not a mathematical plane, just as the fracture surfaces of minerals, even those with very perfect cleavage, are not mathematical planes; all the more so since, at the temperature of liquid air, plastic deformations are observed near the place of rupture. The fracture consists of a series of steps several hundredths of a mm high. Despite these steps, the fracture reflects light in a strictly definite direction, like a mirror surface. This indicates the parallelism of the surfaces forming these steps. In some places (for example, a) the development of such surfaces is noticeable, parallel to the main fracture plane (001) and standing away from it at a distance of several hundredths of a mm. In some places, thin sheets of metal are separated from the fracture surface, in places bending over the surface. In the photograph these sheets, reflecting light because of their bend in a different direction compared with the main plane (001), appear, under oblique illumination, either dark or light. The lateral boundaries of these sheets, as well as of the steps, are the planes \((\bar{1}10)\), \((1\bar{2}1)\), \((010)\), and \((100)\).

Fig. 8 × 12

The most frequently encountered steps are those with a direction coinciding with the planes \((010)\) and \((100)\).

Thus the fracture of a large crystal of \(\alpha\)-ferrite at low temperatures proceeds along the faces of the cube—cleavage planes—but planes \((1\bar{1}2)\) and \((110)\) take part in the formation of the fracture surface. The traces of the plane \((1\bar{1}0)\) and of the plane \((112)\) on the cube face \((001)\) have the same direction, as does the trace of the plane \((111)\). But careful investigation under the microscope and measurement of the angles, insofar as it could be performed on a face less than 0.1 mm wide, force one to incline to the view that we are dealing with traces of planes \((1\bar{1}2)\).

On the fracture surface, parallel lines that shine under suitable illumination are very clearly noticeable. They may be taken for traces of twins or steps of slip lines. These lines coincide with the traces of planes \((112)\) on the cube face. In some places the steps inclined to the plane of fracture \((100)\) have a direction coinciding with the direction of the traces of the surface \((112)\).

For greater clarity, Fig. 9 gives a schematic axonometric drawing of the fracture surface whose photograph is presented in Fig. 8. In the drawing, the fracture plane is combined with the face \((001)\) of the drawn cube, the other faces of which \((100)\) and \((010)\) coincide with the mentioned steps. The traces of twins or slip lines on the plane \((001)\) are given in the form of hatching.

Figure 10 shows the fracture (Fig. 8) with pressure figures obtained by pressing into the surface a conical steel needle with an angle

at the apex is \(45^\circ\). The cruciform pressure figures are parallel to one another, and their orientation makes it possible to determine rather well the position of the fourth-order axes. To make it possible to photograph the pressure figures, the direction of illumination in Fig. 8 and Fig. 10 was chosen differently, and therefore in Fig. 10 certain details appeared which are not visible in Fig. 8.

Fig. 9

The slip planes in crystals of \(\alpha\)-iron are the planes \((101)\), \((112)\), \((123)\). Thus, as a result of studying the fractures of steel No. 4, it may be considered that fracture proceeds mainly along the cleavage plane \((001)\), but that the slip planes and, probably, the twinning plane \((112)\) participate in the formation of the complex fracture surface. Considering only the fracture surface, it is difficult to say whether on this surface we see the emergence of twins or ledges of slip surfaces. O’Neill asserts that under static tension of a single crystal

Fig. 10

Fig. 11. × 8

Fig. 12. × 8

twins along the planes (112) were observed in silicon iron. Similar fracture patterns were obtained for single crystals of other steels (Nos. 2, 3, and 4). The principal plane of brittle fracture in these crystals is the plane (100). The complex fracture surface is formed by planes (100), spaced from one another by several hundredths of a mm. The transition from one plane (100) to another identical plane, located at a distance of several hundredths of a mm from the first, is effected by surfaces whose traces on the plane (100) are planes with indices (100) and (112).

In Fig. 11 a photograph is given of the fracture surface of a single crystal of steel No. 3, and in Figs. 12–13 of steel No. 4.

Parallel to the fracture surface of the specimens, polished sections were prepared at distances of 0.1; 0.5; 10 mm. After etching with an alcoholic solution of nitric or picric acid, parallel lines were found on the surfaces of the specimens, representing traces of the planes (112) on the plane (100)—the plane of the polished section. In Figs. 13 and 14 such a polished section is shown, prepared from the specimen given in Fig. 8, at a distance of 0.2 mm from the fracture surface. The polished section was photographed twice under different illumination.

Fig. 13. ×10

Fig. 14. ×10

As is known, the slip planes and twinning planes in a ferrite crystal are unambiguously (112). This circumstance does not make it possible to say definitely whether on the fracture surface we see traces of twins or traces of slip planes. Both would give a series of similar ledges on the plane (100).

P. A. Bryantsev [5] found that, besides the most frequently occurring plane (100), fracture surfaces may also be the planes (110) and (123). He observed fracture of large ferrite crystals at room temperature. In the case of fracture of single-crystal specimens at the temperature of liquid air, we did not observe fracture along these planes. The principal fracture surface was always the plane (100).

Slip surfaces are clearly visible on the surface of a polished and then deformed specimen in the form of a peculiar wrinkling, a series of parallel bands. These bands are traces of slip planes on the outer polished surface of the specimen. After renewed polishing these traces disappear and are not revealed again by ordinary etching, unless cracks have formed at the sites of the slips.

...of the crack. Careful examination (Figs. 13 and 14) established the presence of one such crack, the direction of which coincides with the axis of the fourth order lying in the plane of the drawing. In Figs. 13 and 14 this crack is clearly visible in the form of a line coinciding with the horizontal axis of the fourth order. This crack extended to a depth of 0.3 mm. When the polished section was turned with respect to the light source, the crack, with its edges exposed as a result of the dissolving action of the etchant, did not change its character and did not acquire the peculiar luster of a twin. This is clearly seen when comparing Figs. 13 and 14. At the same time, the bands that we take to be traces of twins disappear, change the character of their coloration and shade, or shine differently when the position of the polished section with respect to the light source is changed. A thin plate (0.3 mm), cut parallel to the fracture surface, broke under slight bending precisely along the crack. It was not possible, either in this case or in others, to break the plate so that the fracture would pass through one of the twin bands. The plate was bent through 180°, showing no signs of fracture. In Fig. 14 it is noticeable how a number of twin bands, parallel to the trace of the plane (211), are somewhat displaced on passing through the crack.

When the illumination is changed, some of the bands in question disappear, but new ones appear that were not visible before. In Fig. 13 the bands with direction (121) are clearly visible; they can hardly be detected in Fig. 14. But in Fig. 14 lines with direction (211) appeared; moreover, in Fig. 12 in the lower left corner several bands with direction (110) are visible. Evidently, one of the planes (112) can give, on the plane (100), traces with a direction corresponding to the plane (110).

On the fracture surface we found traces of the planes (110) and (112). Traces of these same planes we also found on the surface of the polished section parallel to the fracture surface. All these directions are traces of the faces of a tetratri-octahedron on the face of a cube. These traces are not the emergence of possible slip surfaces on the fracture surface, since they would not have been visible on a specimen that had been reground and polished several times. They reflect light in one definite direction. This argues that their crystal lattice is oriented differently in comparison with the basic crystal. It must be admitted that these bands are traces of the emergence of mechanical twins, formed in the process of impact deformation, along the faces of a trioctahedron onto the plane (100).

At the place where they occur most frequently there are about 15 twins per 0.5 mm. Over 0.03 mm there is one twin. Polished sections prepared parallel to the one considered, at a distance of up to 3–5 mm from the fracture surface, retain the same character in the arrangement and appearance of the twins. Essentially the same picture is given by other fracture specimens investigated in a similar manner.

In Fig. 15 an axonometric drawing is given of the fracture whose photograph is presented in Fig. 12 (steel No. 4). Wishing to trace the direction and arrangement of twins in the single crystal, we took a single crystal of the same steel No. 4, broken at the temperature of liquid air. The fracture in this crystal passed along two mutually perpendicular surfaces. These two surfaces, intersecting at an angle of 90°, were examined and proved to be cube faces. Each of these surfaces was ground down to a depth of 0.5 mm. Then the crystal was cut perpendicular to the line of their intersection (the edge of the cube). This cut produced a third face of the cube. The artificially obtained third face of the cube was ground and polished.

The accuracy in producing such a cube with dimensions \(6 \times 3 \times 4\) mm fluctuated within 2–3° deviation of the cube edges from their theoretical direction. All three faces were etched with picric acid (spir-

...solution), and photographed at a magnification of 5 times. The photographs were pasted onto the faces of a cube model (Fig. 16).

In Fig. 16 it is clearly seen how the traces of one and the same twin pass along two and three faces of the cube. The black lines on the faces of the cube

Fig. 15

mark the traces of the planes (112). If the upper face of the cube is taken as the plane (001), then the left side face will have the index (100), and the right one (010). The direction of the planes along which the twins are arranged will, evidently, be (112).

A similar photograph may be seen in Fig. 17. In this case the single crystal was made of steel No. 3. As in the preceding case,

Fig. 16

in the fractured single-crystal specimen a fracture was found which ran along two faces of the cube; the third face was obtained artificially.

A single crystal that had undergone a rapid impact at \(-185^\circ\) is entirely penetrated by twins. Polished sections of crystals about 25 mm long, broken in liquid air by the method of Charpy and Izod and prepared at different places along the cross section of the specimen, were covered with a continuous network of twins.

In Fig. 18 there is shown (at a magnification of 100) a longitudinal section, etched with picric acid, of a specimen of steel No. 2. The section was made at a distance of 10 mm from the fracture surface. On an Izod-type specimen (with a notch), after fracture no plastic deformation was noticeable; the specimen was completely unbent.

Fig. 17

In the photograph, twins running in two directions are visible. The boundaries of the twins are very sharply expressed. The twins (more precisely—their traces on the surface of the section) have a certain thickness, of the order of several microns. It is scarcely possible to take these bands for anything else—cracks, traces of slip.

Fig. 18          Fig. 19

The bodies of the twins are colored a darker shade than the background—the main crystal. This indicates their different orientation with respect to the surface of the section, as compared with the orientation of the main crystal.

In Figs. 19 and 20, at the same magnification, are shown sections of single crystals of steels Nos. 1 and 3 broken in liquid air. In Fig. 19, twins are visible, arranged on the plane of the section in four directions. Of these, two directions are almost mutually perpendicular. This indicates that the surface of the section is almost parallel to a face of the cube.

After regrinding (with subsequent polishing) and etching of the polished section, essentially the previous picture is repeated.

In Fig. 21 a polished section of a single crystal near the fracture site is shown (steel No. 3). The lower boundary of the specimen is the fracture line. Parallel to the fracture, three small lines—\(a\), \(b\), and \(c\)—are noticeable on the polished section. These small lines are cracks parallel to the \((100)\) plane. On regrinding they do not disappear and are clearly visible on the unetched surface of the polished section. During etching their edges are strongly eaten away, and they become thicker the longer the etching has continued. When the specimen is bent, they open up.

The appearance of cracks even near the place of failure is observed infrequently and in small numbers. Determination of the specific gravity of single-crystal specimens before fracture and after it did not give any substantial difference. The specific

Fig. 20

Fig. 21

gravity of a crystal with twins and without twins is the same. The latter circumstance also indicates that deformation at the temperature of liquid air, in the case where brittle fracture is obtained, is not accompanied by noticeable work hardening. The presence of work hardening and cracks would be reflected in the density of the metal.

On the basis of the preceding material, it seems to us, some conclusions may be drawn. Under rapid impact deformation (impact velocity on the order of 350 m/sec), single crystals of ordinary deoxidized steel and of silicon-manganese steel at the temperature of liquid air are at the boundary of impact brittleness. Homogeneous single crystals with very small internal stresses may, under such conditions, deform plastically (bend) without failing. When a single crystal fails, the fracture surface is a plane—the face of a cube when the direction of impact (the trajectory of the bullet) is within \(60\)—\(90^\circ\) to the axis of the fourth order (that is, to the axis of the fourth order closest to the geometrical axis of the prismatic or cylindrical specimen). The fracture surface often represents an almost mirror-like plane. In places there are ledges with a depth of several hundredths (up to 0.1) mm. The direction of these ledges coincides with traces of the \((100)\) and \((112)\) planes on the fracture surface. The latter direction is most often encountered in single crystals of deoxidized ordinary steels, and the former in steel with a considerable silicon content. It is highly probable that these ledges coincide with traces of possible slip planes \((123)\), but such an assumption is hardly correct. In single crystals that have undergone impact at considerable velocity, numerous twins appear, located along a plane of the order \((112)\).

During annealing of a deformed crystal, recrystallization begins within a definite temperature range. Recrystallization in pure iron begins at a temperature of about 350°. If, in the given specimen subjected to recrystallization, there is a deformed and an undeformed part, then the formation of new crystals occurs at the place where they are in contact [17].

When mechanical twins have formed in a single crystal, the appearance of new crystallization centers is to be expected at the boundary between the twin and the principal crystal [16]. This will be the case only if, during the twinning process, the crystal has not undergone plastic deformation of the usual kind. The formation of mechanical twins creates a nonuniformity in the distribution of energy, caused by the special state of the atoms at the boundary of the adjoining atomic layers.

The principal crystal and the twin are oriented quite symmetrically with respect to the twinning plane. Therefore there are no special grounds for expecting the twins to disappear during recrystallization. The principal crystal and the twin are crystallographically “equal in rights.” During recrystallization the principal crystal may increase in size, absorbing the twin, but the twin too is capable of growing at the expense of the principal crystal.

We performed the following experiment: a polycrystalline specimen with large crystallites (of steel No. 3), with a previously polished side surface, was broken by a shot at the temperature of liquid air. No signs of bending were found in either half of the fractured specimen. After fracture, all the crystals were penetrated by a large number of twins. The broken specimen was annealed in vacuum at a temperature of 870° for 30 minutes. No changes were observed on the polished surface. The number of twins neither increased nor decreased; no new crystallites appeared on the surface of the crystal. The surface of the microsection was ground down to a depth of 0.05 mm, and on it, after etching, the same structure was revealed as on the polished surface of the single crystal. After examination of the microsection, the specimen was again subjected to annealing at the same temperature, but now for three hours. After polishing (without grinding) and etching, the former twins were found on the surface, and at the twin–crystal boundaries new crystallites appeared as a result of recrystallization (Fig. 22).

In Fig. 22 (magnification 5), the fracture surface is the lower face of the specimen. As can be seen, it is perpendicular to the right lateral surface of the microsection. The fracture surface is identified as a cube face. The right lateral surface coincides quite satisfactorily with the second face of the cube, and the plane of the microsection under consideration with the third.

Some places of this microsection were photographed at a magnification of 200 (Figs. 23 and 24). The twins in Fig. 23 have a thickness of the order of 0.005–0.01 mm. In the lower left corner of the photograph, part of a large crystallite is visible, formed during the second, three-hour anneal. Judging by its position and coloration, this crystallite retained the orientation of the twin. Evidently this crystallite is a twin that has grown at the expense of the principal crystal. This is also indicated by the straightness of part of the crystallite boundaries, parallel to the traces of the twins on the microsection. In the upper part of the crystallite an offshoot—a twin—is noticeable. Probably it was from this twin that the crystallite in question grew.

In Fig. 24 a crystallite is seen that formed as a result of the growth of a twinned crystal. Increasing in size, it joined two neighboring twins that were identically oriented. The orientation of all the twins of this direction and of this crystallite is completely identical. The specimen was several times carefully reground and etched with picric acid and by oxidizing the surface upon heating to temper colors;

brittleness. In this case the dimensions of the crystallite were established in a direction perpendicular to the plane of the polished section, as well as its unaltered connection with two neighboring twins and its identical orientation with them (as indicated by identical coloration).

This same specimen was then subjected to heating for another \(6\frac{1}{2}\) hours. The total duration of heating was ten hours. After this annealing, newly formed crystallites were again found on the polished section,

Fig. 22            Fig. 23

but the character of the twins as a whole had not changed (Fig. 25, magnification 200, etched with picric acid).

Similar experiments were repeated with coarse-crystalline specimens from other steels (steels Nos. 1, 2, and 4). The results obtained did not differ from those described.

Evidently, we are in fact dealing with typical mechanical twins.

To clarify the role of slip in impact deformation, square specimens without a notch, with a polished and etched surface, were prepared from mono- and coarse-crystalline specimens of the steels investigated.

Fig. 24            Fig. 25

They were bent by impact according to the Charpy scheme. On specimens that did not break at the temperature of liquid air, striations were clearly visible, such as usually appear during the static deformation of polished

specimens. Such striation on the polished surface of crystals is considered to be traces of slip surfaces. In Figs. 26 and 27 such striation is clearly visible on the surface of specimens of steel No. 3 deformed by impact at the temperature of liquid air. Fig. 26 and Fig. 27 are photographs of one and the same specimen under different illumination (magnification 5). The bright bands perpendicular to the axis of the specimen and the less noticeable dense striation in the upper part of the specimen, almost parallel to its axis, are clearly visible (Fig. 26). In Fig. 27 the lines perpendicular to the axis have disappeared, but a series of bands inclined to the axis has appeared (lines running from the concave surface). After photographing, a layer about 1 mm thick was removed from the specimen and the resulting plane, parallel to the side surface of the specimen that we had just examined, was polished and etched with picric acid. After etching, the polished section was photographed at the same magnification as Figs. 26 and 27 (Figs. 28 and 29). These photographs were taken under the same illumination as the preceding ones (Figs. 26 and 27). As can be seen, each line on the surface of the specimen corresponds to a line on the polished section. Obviously, these lines are traces of twins on the surface of the specimen, and not slip lines.

Fig. 26

Fig. 27

Fig. 28

Fig. 29

Examining the polished section at sufficiently high magnifications, we found that these lines, which did not disappear during subsequent regrindings of the specimen, are nothing other than twins. But the twins in this case proved to be bent. Despite a careful examination of the specimens, no traces of cracks were found in them. Examination of a series of ...

other similar specimens gave analogous results. At the same time, we were unable to detect traces of sliding surfaces on the surface of the specimen.

Bent twins had been observed by us earlier in single crystals deformed by impact, and we shall dwell somewhat on this interesting phenomenon. In Fig. 30 (magnification 100, section etched with picric acid) a large number of bent twins is visible at the boundary of two large crystals (steel No. 3). The specimen was bent by a rapid impact at room temperature. The photograph was taken not far from the place of greatest deformation. The radius of curvature of the bent twins varies approximately around 4 mm. In Fig. 31 (magnification and etching as in the preceding case) is given a photograph of a section of a single crystal of steel No. 1, bent by impact at ordinary temperature. The radius of curvature of some twins in this case is approximately 0.2 mm. The bending experienced by them is rather large. The thickness of the twinned bands, determined at higher magnification, is of the order of 3–4 microns. Not all twins experienced very strong bending (in Fig. 31): the vertically running twins retained their parallelism to some degree. A similar picture is also observed in Fig. 30. Fig. 32 shows a photograph of part of a coarse-crystalline specimen of steel No. 4, bent (according to Charpy’s scheme) at room temperature. The photograph shows the part of the specimen that rested on the support, which passed vertically with respect to the plane of the section. The depression visible at the left is the trace of the impact against the support. The twins near the place of deformation were bent in a very peculiar manner. However, no cracks were found in the metal either when examined under the microscope or after deep etching of the specimen, cut into several parallel sections. Twins formed at liquid-air temperature are usually rectilinear. Apparently, the twins arise at the initial moment of deformation and develop extremely rapidly. The rate of deformation of the metal at the place of contact with the support (Fig. 32)

Fig. 30

Fig. 31

Fig. 32

is equal to or, in any case, very close to the velocity of a bullet, i.e. it is of the order of 300–350 m/sec. The amount of compression of the specimen near the support is about 0.1 mm, and the time during which plastic deformation proceeded was of the order of millionths of a second. Apparently the rate of formation of twins is equal to the speed of sound.

Fig. 33

The answer to the question of whether the presence of twins is a sign of the dynamic brittleness of steel will be obtained as a result of studying the properties of a crystal cut through by twins. We took several coarse-crystalline specimens (steel No. 3) and broke them by shooting at the temperature of liquid air (Izod scheme).

Fig. 34

Figure 33 shows such a specimen. The bullet struck from left to right in the plane of the photograph, hitting the upper part of the specimen, from which it rebounded. Near the fracture site the specimen is slightly bent. In the rest of it there are no noticeable plastic deformations; the specimen is perfectly straight. Figure 34 gives a photograph of the same specimen, but under different illumination. Before photographing, the lateral surface of the specimen was ground down to a depth of 0.3 mm, polished, and etched with nitric acid. All the crystals of the specimen are densely mottled with traces of twi-

There is every reason to suppose that the mechanical properties of the specimen were not changed, since there was no work-hardening. A slight work-hardening exists only near the place of fracture. In the usual sense there is no work-hardening; if there is any plastic deformation, it is the result of twinning. Such a specimen, in each crystallite of which there are many twins of various orientations, was again deformed by a shot at room temperature (according to the Izod scheme). The specimen, deformed a second time, is shown in Fig. 35. Before photographing, a layer of metal about 0.2 mm thick was removed from it (from the same surface as that shown in Figs. 33 and 34); it was polished and etched with picric acid. At the place of maximum deformation, the transverse—

Direction of impact

Fig. 35

—section of the specimen changed very strongly, becoming an irregular trapezoid. The stretched place (the left part of the specimen, at the top in the photograph) became thinner, and removal of 0.2 mm did not eliminate the depression visible in the photograph. At the place where the bullet struck, in the upper part of the specimen, a dent about 0.2 mm deep was formed. The number of twins in all crystallites of the specimen after bending increased greatly, and both the old twins and those newly formed became curved. In spite of the very considerable plastic deformation, no traces of destruction of the metal of the specimen were observed. Even after prolonged deep etching, no cracks were found. Several specimens of steels Nos. 1, 2, 3, and 4 were tested in the same way. In all cases the results were similar. It seems to us that it may be asserted that the presence of mechanical twins in iron does not cause any particular brittleness of the metal. The actual rate of deformation at the place of greatest bending was, of course, less than 350 m/sec—the velocity of the end of the specimen at the moment of impact by the bullet—but nevertheless the order of this velocity is 50–70 m/sec.

On the basis of the foregoing, it may be assumed that at very low temperatures slip in ferrite crystals is absent. At considerable rates of deformation, even at ordinary temperature, twins arise in ferrite crystals at a rate close to the—

of sound. In the course of far-advanced plastic deformation, the twins that form are bent. Twins that have appeared in ferrite crystals, even in very large numbers, do not reduce the impact toughness of the crystals—at least at room temperature.

Cracks do not appear in the crystallites even in stretched regions, where the relative elongation reaches 30%.

In the investigation of single crystals and polycrystalline specimens of decarburized steels containing Mn and V, a somewhat different picture was observed. Fracture at the temperature of liquid air in all cases occurred along the (100) plane. At the temperature of liquid air, only notched single crystals always broke. Unnotched polished specimens, especially those of steel containing vanadium, were often bent without cracks. The fracture surfaces differed little from the fracture surfaces of crystals of ordinary unalloyed steel. Twins emerging onto the surface were always clearly visible on the fracture surface.

Fig. 36

Fig. 37

In Figs. 36 and 37 are given photographs (at a magnification of 6) of the fracture and of the polished section of steel No. 6, etched with picric acid. The polished section was made parallel to the fracture surface of a notched single-crystal specimen at a distance of 0.2 mm from the fracture. Figs. 36 and 37 were photographed with different directions of illumination relative to the surface of the polished section. Although the single crystal was broken by impact at the temperature of liquid air, a large number of curved twins is noticeable in the photograph. In the right-hand part of Fig. 36 a series of curved twins is visible. Curved twins of two directions are also visible in Fig. 37.

The fracture surface of the specimen coincides with the (001) plane, and the direction of the curved twins before their deformation apparently coincided with (112). The strong curvature of the twins very considerably hinders an exact determination of their initial orientation. Along with the twinning direction (112), these photographs also show clearly one more twinning direction, (123), which we had not previously encountered in studying the deformation of single crystals of other steels. In studying crystals of steel No. 6 deformed at low temperatures, twins with traces corresponding to the (123) plane were almost always encountered together with twins lying in the (112) plane. This twinning direction was checked by the method already described: cutting out cubes correctly oriented with respect to the fracture surface of the single crystal. Apparently, this twinning plane (123) is characteristic precisely of crystals of the solid solution iron—vanadium—chromium, possessing the lattice of α-iron.

In one or two single crystals made of decarburized steel No. 1 (with an increased manganese content), the same twinning planes $(123)$ were found in a small number.

As we have shown, as the temperature of impact testing is lowered, the number of slip surfaces decreases and the number of twins formed increases. It may be said that, as the temperature is lowered, plastic deformation changes its character. Twinning produces a certain residual change in the shape of the crystal. A twinned crystal changes its external shape to a certain extent. Even very numerous twins, if their thickness is small, cannot produce as strong a change of shape as ordinary slip and the corresponding rotation of the crystallites. Nevertheless, the appearance of a new twinning direction—namely $(123)$ in the present case, in addition to the already existing $(112)$—is capable of producing some additional change of shape and thereby of providing additional plastic deformation.

This fact can explain the greater impact toughness at low temperatures of steels with additions of vanadium and manganese as compared with steels of ordinary composition.

The presence of a certain toughness in steel No. 6 at the temperature of liquid air is confirmed by the formation of bent twins. The curvature of twins in itself indicates plastic deformation at the given place. Twins may be deformed as a result of subsequent (after the formation of the twins) slip of individual elements of the crystal at an angle to the twinning plane. Bending of twins will also occur if twins of different directions are formed not simultaneously, but in a certain sequence. It may be assumed that twins with the direction $(112)$ arise first, and then $(123)$; or initially the most favorable (for the given arrangement of the crystal axes and the direction of impact) twinning direction appears, say $(211)$, and then $(121)$. Such an assumption is difficult to prove, but there is nothing improbable in it, and this assumption explains the curvature of twins without the participation of ordinary shears. If twins are formed with the propagation velocity of an elastic wave in the crystal (the velocity of sound), then, evidently, this velocity of their formation must depend on the value of the modulus of elasticity of the first kind in the given direction. As is known, the modulus of elasticity changes rather strongly with a change in direction in a given crystal. Therefore, it is probable that twins along the surfaces $(112)$ are formed first, and then twins along the planes $(123)$ appear. Twinning along the latter planes may cause curvature of the twins that appeared initially. It may be said that curvature of twins may also cause subsequent slip and subsequent twinning.

The study of the impact fracture of fine-crystalline metal essentially repeated the picture described above. The fracture line passing through the crystallites is accompanied by the appearance of twins. The larger the grains of the metal, the more twins they contain. It was extremely difficult to determine the orientation of the twins, as well as the crystallographic significance of the fracture surfaces of individual crystallites. Nevertheless, a whole series of indirect methods of investigation (etch figures, direction of reflection of greatest brightness) confirmed that the fracture planes are faces of the cube, while the twins are arranged along the surfaces $(112)$. In individual crystallites there is also observed a very considerable curvature of the twins formed under impact.

VI. Conclusion

On the basis of the foregoing, several general conclusions may be drawn. In dynamic fracture at low temperatures and at high velocity, in single crystals, coarse-crystalline and fine-grained specimens of more or less pure ferrite, and ferrite with small additions of manganese, silicon, and vanadium (simultaneously with chromium), the fracture surface is the cube face (100).

If the fracture surface is not a mirror plane, which is observed in the fracture of single crystals of small cross section (about one square millimeter), then a stepped fracture is usually obtained. The height of the parallel steps that form is very small—of the order of hundredths of a millimeter. The transition from one parallel plane to another, from one step to the adjacent one, takes place along the planes (100), (112), and, more rarely, along (110).

One may suppose that it is possible (under suitable experimental conditions) to obtain the main fracture along one of these planes. In our investigations, however, this was not observed.

An unquestionably brittle fracture, without noticeable residual deformations, at the boiling temperature of liquid air is obtained only in notched single-crystal specimens. Unnotched round, well-polished single crystals sometimes merely bend without traces of fracture. The more homogeneous the structure of the single crystal, the more completely diffusion has proceeded as a result of prolonged (hundreds of hours) annealing, and the smaller the internal stresses (removed by annealing with very slow cooling), the lower the brittle-fracture temperature. In this case the brittle-fracture temperature may lie far below the boiling temperature of liquid air.

On the fracture surface, in addition to steps, traces of the emergence of twins are found. The planes on which the twins are situated are the (112) planes. At low temperatures and high rates of deformation, ordinary slip begins to disappear, and plastic deformation at low temperatures can occur only through twinning shears of the crystals.

In ferrite crystals alloyed with vanadium and chromium, and also with manganese, besides the twinning plane (112), a new plane appears on which twins are situated: (123). As a result, the possibility of plastic deformation through twinning of the crystal becomes still more probable. In alloyed ferrite crystals, the disappearance of ordinary slip is observed at a lower temperature than in ordinary crystals. It is difficult to say why alloying additions in such comparatively small amounts promote the preservation of slip at low temperatures and the appearance of additional twinning planes. It is quite probable that in this case the better deoxidation of steels alloyed with vanadium, their greater freedom from submicroscopic inclusions, has an effect. A smaller quantity of the finest particles precipitated during aging or slow cooling facilitates the development of slip and the formation of twins, and facilitates the course of plastic deformation.

The appearance of twins in ferrite crystals does not in itself “spoil” the metal. Twinned (without the appearance of slip) large ferrite crystals exhibit considerable dynamic toughness at room temperature. Under rapid impact deformation they undergo very considerable plastic deformations.

There is observed a curvature of twins formed earlier, due to slip, and the appearance, in a definite sequence, of new twinning planes.

It may be asserted that twins, as such, are not the cause of the cold brittleness of ferrite crystals. By their appearance, twins provide a certain plastic deformation, replacing the slip that is absent at low temperatures, and thereby reduce brittleness.

LITERATURE

1. Averbukh B. D. and Chufarov G. I.—Metallurg, 7, 48. 1939.
2. Belyaev S. E.—Mechanical properties of aviation materials at low temperatures. Oborongiz. 1940.
3. Bessonov K. A.—Izv. Sib. Ind. Metallov, I, 163. 1934.
4. Boldyrev A. K.—Course of descriptive mineralogy. Leningrad. 1925.
5. Bryantsev P. A.—Vestnik metalloprom., 12, 70. 1937.
6. Bryantsev P. A.—Vestnik metalloprom., 6–10. 1938.
7. Bunin K. P.—Metallurg, 1, 97. 1937.
8. Buturlin S. A.—Rifled hunting weapons. Vsekhotsoyuz. 1929.
9. Weinberg Gr. and Proshutinsky S. I.—Metallurg, 9, 3. 1938.
10. Gotoberidze D. B.—Mechanical twinning. GONTI. 1938.
11. Davidenkov N. N.—The problem of impact in metal science. Academy of Sciences of the USSR. 1938.
12. Davidenkov N. and Shevandrin E.—Zh. Eksp. teor. fiziki, 6, 261. 1936.
13. Dobrovidov A. N.—Cold brittleness in medium-carbon steel. ONGGI. 1937.
14. Dobrovidov A. N. Zh. tekhn. fiz., IV, 1, 27. 1927.
15. Carpenter G.—Advances in physics. ONTI. 1933.
16. Kolesnikov A. F., Voevodin V. N. and Bogomolova M. N.—Proceedings of the Physico-Technical Institute, 3, 201. 1938.
17. Kuznetsov V. D.—Physics of the solid state. I. 1932.
18. Latun E. O. and Bushtedt P. P.—Metallographic study of welded iron. OGIZ. 1931.
19. Sakharov P. V.—Investigation of the question of annealing boiler iron. 1927.
20. Sakharov P. S.—Zh. tekhn. fiziki, VI, 1388. 1937.
21. Filatov N.—Brief information on the fundamentals of shooting. Gosvoenizd. 1936.
22. Khvolson O. D. Course of physics. II, 33. 1923.
23. Shevandrin E.—Zh. tekhn. fiziki, VIII, 411. 1938.
24. Shevandrin E. M.—Zh. tekhn. fiziki, X, 402. 1940.
25. Schenk—Physical chemistry of metallurgical processes. ONTI, 1. 1935.
26. Schmidt E. and Boas W.—Plasticity of crystals. GONTI. 1938.
27. Yakovleva E.—Metallurg, II, 68. 1935.
28. Bainklon H. and Guthmann H.—Arch. f. d. Eisenhüttenwesen, 4, 201. 1935.
29. Edwards C. A. and Pfeil L. B.—Journ. Ir. a. Steel Inst., 109, 129. 1924.
30. Hanel R.—V. D. I., 410. 1937.
31. Georgieff M. and Schmid E.—ZS f. Physik, 5, 607. 1930.
32. Goerens P.—Handbuch d. Exper. Physik, 5, 607. 1930.
33. O’Neill.—Journ. Ir. a. Steel Inst., 113, 417. 1926.
34. Sauerwald F., Schmidt K. and Krämer G.—ZS f. Physik, 67, 179. 1931.
35. Schäfer R.—Die Konstruktionsstähle, 172. 1923.
36. Taylor G. I. and Elam C. F.—Proc. R. Soc., 112, 337. 1926.

Noted Misprints

Page	Line	Printed	Should read
23	14 from top	at point a	at point 8
24	Table 2, column 2, first row	0	50
32	18 from top	elektroderzhatel	elektrododerzhatel
33	16 from bottom	moduritsirovanie	modifitsirovanie

Proceedings of TPI, vol. 61, issue 4

1. The direction of flight of the bullet is indicated by an arrow. ↩