On the Experimental-Graphical Study of the Bessemer Process

(With 4 tables of zincographs and 1 table of diagrams.)

Diploma work of L. Krasin, a student of the mechanical department of the Tomsk Technological Institute.

I. Introduction.

Bessemerizing is a highly interesting and distinctive metallurgical process. In addition to the organic distinction, which consists in obtaining the heat necessary for the operation from the refined metal itself, the Bessemer process is accompanied by a whole series of very beautiful and striking phenomena. Among the latter, on the one hand, are the external physical features consisting in the coloration of the flame bursting from the mouth of the converter, its spectrum, and sound effects; on the other hand, Bessemerizing gives no less curious a picture of the structural transformations of the metal, studied by microanalysis, and also reveals a very interesting regularity in the internal complex thermochemical relationships disclosed by the graphical method. In metallurgical literature, since the announcement of Henry Bessemer’s patent in 1855, no few studies and investigations of this original invention have appeared, but the majority of inquiries have concerned foreign pig irons. Among these works, the authors’ attention was devoted to a sufficient degree to diagrams of the burnout of elements obtained on the basis of chemical analyses of samples taken from the converter, at definite intervals, over the entire period of the heat. All these diagrams, although they demonstrate one and the same metallurgical process, proceeding in siliceous pig irons, and reveal quite vividly the general characteristic features of the curves, nevertheless each diagram possesses its own particular, individual characteristics.

Among the earliest diagrams must be counted the one obtained by E. Barker in the seventies of the last century at one of the English works and thus characteristic of English Bessemerizing¹. (Diagram No. 1).

The greatest attention in this diagram is attracted by the carbon-burning curve \(C\). In the first four minutes the curve rises upward and indicates a relative increase of this element in the metal. This phenomenon, strange at first glance, is a fictitious enrichment of the bath in \(C\), due to the rapid oxidation of \(Fe\) and its transition into slag in the form of \(FeSiO_3\). The energetic and uniform burning out of \(Si\) in the first 6 min. contributes to this characteristic feature of English Bessemerizing. With the decrease in the percentage content of \(Fe\) in the bath, there naturally occurs a relative increase in \(C\), which only after the seventh minute begins to burn out rapidly, when the temperature of the bath is already sufficiently high. The \(Mn\) curve indicates its very insignificant presence; and although English pig irons, generally speaking, are poor in \(Mn\), nevertheless, in the diagram presented the curve of this element is somewhat doubtful, as, among others, Prof. Pavlov also mentions in his work Metallurgy of Steel of 1907.

A diagram characterizing Swedish Bessemerizing was given by Sogasson and Magnuson¹. (Diagram No. 2). This diagram does not have a fictitious enrichment of the bath with carbon, owing, firstly, to the excessively high percentage content of it in the metal, up to 4.5%, as a result of which the carbon begins to burn out from the very first minutes of the blast, and, secondly, this distinctive feature of English Bessemerizing is also absent here because the low \(Si\) content conditions a cold course of the operation. The \(Si\) curve in the first two minutes falls rapidly downward, as does the \(Mn\) curve, which is most characteristic of the Swedish process.

American Bessemerizing was recorded in the diagram of Prof. Howe, obtained by him in Chicago at the plant of the Illinois company in recent years¹. The curves of the diagram (Diagram No. 3) are very close to the curves of Swedish Bessemerizing and are almost analogous to the latter. A fictitious enrichment of the bath with carbon is likewise not observed here, in view of the fact that the \(Si\) content is not high, about 1%, and therefore the operation at first proceeds with insufficient vigor, and there is not that considerable transfer of oxidized iron into the slag which was observed in the English process. During the first four minutes almost all the \(Si\) burns out, but over this interval of time the temperature of the bath is already so high, and the heat imparted to it by the burned silicon so sufficient, that the operation proceeds to the end with a hot course, and by the tenth minute complete refining of the metal is already achieved. In Germany the Thomas process is most widely disseminated, and therefore there is no direct necessity to present here Finkener’s diagram³, obtained by him in the study of the refining of phosphoric pig irons.

In Russia the market is dominated by Ural, and in recent times chiefly by South-Russian siliceous pig irons.

Ural Bessemerizing was investigated by Prof. Grum-Grzhimailo, who obtained at the Nizhne-Saldinsk Works a diagram of the burning-out of the elements of Ural pig iron.¹) The diagram differs considerably from all the diagrams of foreign investigators cited above. Curve C (Diagram No. 4) is here concave toward the axis of abscissas, whereas in all the other diagrams it is convex. The burning-out of carbon proceeds very vigorously at the beginning of the process and slows somewhat toward the end. Si burns out extremely slowly: its curve, up to the fifteenth minute, descends almost in a straight line toward the axis of abscissas, at a very slight inclination, and only at the end of the process drops sharply downward. Prof. Pavlov, illustrating his course “Metallurgy of Steel” with the diagram under consideration, gives the following explanation of this somewhat peculiar character of the curves in the diagram of Ural Bessemerizing. Thus, regarding curve C Prof. Pavlov says that its concavity toward the axis of abscissas is explained by the fact that “the energy of oxidation of carbon falls as the operation proceeds, which in turn indicates a fall in temperature.”

Further, explaining the very slow burning-out of Si in Prof. Grum-Grzhimailo’s diagram, Prof. Pavlov says: “this depends on two causes: the high ratio C : Si, which in Russian (Ural) Bessemerizing reaches its maximum, and the excessively high temperature at the beginning of the process.”

Thus, the diagram considered pertains to the Ural charcoal industry.

As for the study of the refining of pig irons of Southern Russia, unfortunately no works have yet been published that give a diagram of the burning-out of the elements entering into their composition. This is why obtaining a diagram of the burning-out of the elements in South-Russian Bessemerizing served as the initial task of the work presented here. But the experimental data collected for this purpose, especially in the thermochemical and metallurgical fields, proved so interesting that the scope of the initial task was considerably expanded, as a result of which the very method of the work was modified. Whereas its first part was conducted chiefly by a deductive path, whereby it was necessary to break down and detail the moments composing a very complex process abounding in chemical and physical phenomena, on the contrary, in the second half of the investigation induction became the guiding thread of those generalizations and conclusions which inevitably emerged as the result of the preceding analysis.

¹) Prof. Grum-Grzhimailo. Metallurgy of Steel. The Bessemer Process. 1910.

2. Description of the plant experiments and observations.

All the experimental data necessary for the present work were obtained at the Aleksandrovsk Metallurgical Works of the Bryansk Joint-Stock Company in the city of Ekaterinoslav, with the very kind assistance of the plant administration. We consider it our pleasant duty to express our deep appreciation and gratitude to the assistant director of the works, mining engineer P. T. Golyshev, who, being in charge of the Bessemer and open-hearth shops, kindly permitted the experimenters to carry out a whole series of experiments necessary for the investigation, up to and including stoppages of the operating converter, during both day and night heats.

In conducting the Bessemer operations under investigation and in making the observations, in addition to the foreman on duty, the crew of workers, and the author of the present work, students of the senior courses of the Kiev Polytechnic Institute—Messrs. Olimpiev, Kutsenok, Kryzhanovskii, and Golyshev—took direct part.

The organization of the work took the following form. One of the participants in the investigation remained continuously at the air-blowing and hydraulic machines, monitoring the adjustment of the latter; two jointly observed the rise in the temperature of the converter flame by means of Wanner’s photographic pyrometer; one supervised the workers who took samples of metal and slag from the converter every two minutes; and, finally, a fifth participant in the work, making observations through the spectroscope, kept a record of all the manipulations performed and at the same time had as his duty the control signaling that coordinated the work of all the others. These observations were carried out on the blowing of metal in one of the three large Bessemer converters operating at the works and each holding on average up to 12 tons of material. The pig iron required for the Bessemer shop was obtained from a mixer, into which it in turn came from blast furnaces working on the known Krivoi Rog ores. To judge the general composition of these ores, we give the ordinary analysis of ore from the mine belonging to the works and situated in Sukhaya Balka: Fe—65%; SiO₂—4%; Al₂O₃—1.5%; CaO + MnO—0.03%; Ph—0.03%; S—traces.

The results of observations on the rise in temperature of the Bessemer flame for five heats gave the following results:

Table No. 1

Table No. 2

Table No. 3

Table No. 4

Table No. 5

With particular care, samples of metal were taken from the converter. Every two minutes, from the beginning of the blow to the end of the heat, samples were taken from the middle of the metal bath by specially assigned experienced workers, with the greatest attention being observed-

Thus, with these operations the experimenter’s attention was directed toward the cooling of the metal. Samples from the receiver were poured into a cast-iron mold, which was kept at all times in a moderately heated state. The latter circumstance eliminated hardening of the material and placed all the samples under identical cooling conditions. This precaution was not slow to show itself in the subsequent metallographic investigations: of all the polished sections, only the next-to-last one, representing almost already pure iron, had to be heated in an electric furnace to \(900^\circ\) C, so that the subsequent very slow cooling might produce a more contrasting crystallization, breaking up the coarse grains of iron¹.

From the samples collected in this way, first, small pieces of metal, about \(1\frac14\) cubic centimeters in size, were taken for preparing polished sections from them; and second, fine turnings (chips), weighing from 10 to 15 grams, were planed off for chemical analysis. The cutter of the machine, moving automatically, planed not only the core of the fracture of the sample piece, but also those parts of the metal which directly adjoined its periphery, not reaching by 1 mm the outline where the material was already covered with scale. All these conditions had to be observed because Si is usually present in greater quantity in the middle parts of castings and in smaller quantity at the periphery, while the presence of scale could significantly alter the results of the chemical investigations.

The analyses of the samples from one of heats No. 1, carried out in the plant laboratory, gave the following results.

The pig iron charged into the converter had the composition:

\[ \begin{aligned} \mathrm{C} &— 3.69\% \\ \mathrm{Si} &— 1.48\% \\ \mathrm{Mn} &— 0.88\% \\ \mathrm{S} &— 0.05\% \\ \mathrm{Ph} &— 0.104\%. \end{aligned} \]

Metal samples:

Table No. 6.

Slag samples:

Table No. 7.

In addition to these analyses, subsequently, already in the metallurgical laboratory of the Tomsk Technological Institute, it was necessary to carry out a further series of control analyses for Si on one of the heats, in order to obtain a second curve for the burning out of this element; but these results will be described in detail in Chapter 5, which sets forth the thermochemical aspect of the process.

3. Diagram of the burning out of the elements in South-Russian Bessemerizing.

The data of Table No. 6, plotted on coordinate axes, with time in minutes laid off on the abscissae and the percentage content of the elements C, Si, and Mn on the ordinates, give a diagram characterizing South-Russian Bessemerizing (Diagram No. 5). Let us consider the diagram in connection with those internal chemical phenomena of the process which are recorded by its curves. First of all, attention is drawn to the fictitious enrichment of the bath with carbon, so characteristic of English Bessemerizing and so clearly expressed on the diagram obtained.

In the first two minutes of the blow the carbon curve rises very steeply upward, while the Si and Mn curves descend at an acute angle. This is the moment of energetic slag formation. The temperature of the bath at this time is not yet high enough to bring C out of its inert state, and therefore the oxygen of the blown-in air chiefly oxidizes Fe, Si, and Mn. The ferrous oxide forming very abundantly partly dissolves in the iron, and partly again oxidizes Si and Mn, which are thereby burned to SiO₂ and MnO. Rising to the top of the metallic bath, owing to its lower specific gravity, SiO₂ combines on its way either with FeO or with MnO and floats to the surface of the metal as a black acid slag in the form of ferrous metasilicate FeSiO₃ and in the form of the double silicate 3MnO · 2SiO₂ ¹). Such slag formation makes the metal poorer in Fe, Si, and Mn, and this is indirectly recorded by the ordinate of the second minute of the diagram, which establishes an increase of C from 3.69% to 4.03%.

The second minute is also of interest because it is the culminating point of the fictitious enrichment of the bath in C, since with the further course of the operation the curve begins to descend. The latter circumstance indicates that the affinity of C for the oxygen of the air is already sufficiently strong for its burning out to begin. The role of reducing agent for ferrous oxide gradually passes from Si and Mn to C, and this moment is very clearly visible in the third and fourth minutes of the diagram: the C curve descends steeply to its original level, while the Si and Mn curves, becoming flatter, indicate a slower burning out of these elements.

¹) Prof. Pavlov. Metallurgy of Steel. 1907.

The first period of the process, characterized by vigorous slag formation, gradually gives way to the second period, with its clear sign—the struggle between carbon and silicon for oxygen. With the rise in temperature, and accumulating ever greater energy in its affinity for oxygen, C readily reduces ferrous oxide and manganous oxide, as a result of which their burning out during the second period almost ceases; this is again very clearly confirmed by the Mn curve, which during the 5th, 6th, 7th, 8th, 9th, and partly the 10th minutes runs parallel to the axis of abscissas. This constancy in the second period of the percentage content of iron and manganese in the bath gives grounds for stating that in it the concentration of only two elements—C and Si—is changing. Therefore the refining of pig iron in the second period of the Bessemer process, on the basis of the law of Guldberg and Waage¹, may be regarded as a bimolecular reaction. The rate of this latter changes continuously, and therefore at any given moment it is the derivative of the concentration \(W\) of the medium under consideration with respect to time \(t\):

\[ \frac{dW}{dt} \]

But since the rate of reaction of a medium of complex composition, according to the law of molecular concentration, must be proportional to the concentration of each individual substance, then, denoting the concentration of carbon in the bath by \(W_A\) and that of silicon by \(W_B\), we find the corresponding expression satisfying the condition indicated above:

\[ -\frac{dW}{dt} = K \cdot W_A \cdot W_B \tag{1} \]

The negative value of the derivative is taken because, as time increases, the concentration decreases. \(K\) is the proportionality factor, a quantity constant for a given temperature, which therefore may be called the reaction constant.

To clarify the chemical meaning of this factor, let us suppose that in a monomolecular reaction:

\[ -\frac{dW}{dt} = KW \]

\(W = 1\), then

\[ K = -\frac{dW}{dt} \tag{2} \]

i.e., \(K\) denotes the rate for that particular case when the amount of reacting substance is equal to 1 mole (gram-molecule).

Considering expression (1), we see that each given increment of time must correspond to a change in the magnitude of \(W_A\) and \(W_B\), and, moreover, if one of them decreases more vigorously, then the other relatively increases, and conversely.

Returning to diagram No. 5, we see how this relation between the concentrations of the solutions of C and Si in iron is sharply reflected throughout the entire second period of the process. Thus, the ordinate of the third minute indicates the vigorous burning-out of carbon and, at the same time, the cessation of the oxidation of Si. Conversely, beginning with the ordinate of the fourth minute, C stops burning out, extending the curve parallel to the axis of abscissas throughout the whole of the 5th and 6th minutes, while the Si curve again descends vigorously. The seventh minute marks the beginning of the secondary burning-out of C and a certain bend in the Si curve, indicating the sensitivity of the rate of burning-out of Si to this activity of carbon. During the next two minutes, namely, the 8th and 9th, there is established in the bath, as it were, a certain equilibrium in the intensity of burning-out of these two elements struggling with each other, so that after 10 minutes there occurred a decisive turning point in the whole internal character of the process that had continued up to that time. At this critical moment C takes possession of the oxygen and its curve drops rapidly downward, whereas the Si curve, especially after 11 minutes, sluggishly stretches to the right, becoming ever more parallel to the axis of abscissas.

The second period, characterized by the simultaneous, but varying in intensity, burning-out of C and Si, has, from the chemical point of view, one more very distinctive feature, which consists in the reducing action of C on ferrous oxide according to the reaction:

\[ \mathrm{FeO} + \mathrm{C} = \mathrm{Fe} + \mathrm{CO} \tag{3} \]

Soon after the curves pass through the tenth minute, when the greater part of C has already burned out, the third period of the Bessemer process begins—the shortest one, characterized by the appearance, among the gaseous products of combustion, of brown vapors of FeO¹. Gaseous ferrous oxide appears as the result of the finishing refining of the metal and, by virtue of this, of the very decisive oxidation of Fe by the oxygen of the blast. The FeO and \(\mathrm{Fe_2O_3}\) that form in abundance saturate the bath and, carried along by the gases, surround the flame of the converter with brown smoke. The blast, a few seconds after the appearance of the smoke, is stopped and spiegel is added.

In the heat under study, 45 poods of the latter, of the following composition, were thrown into the bath for 677 poods of liquid cast iron: C—4.79%; Si—1.03%; Mn—14.83%; S—0.02%; Ph—0.171%.

After the charging of the spiegel is completed, the blast is again turned on for a very short time, and this concluding moment of the process is

the fourth period of it. From the chemical point of view, the last period differs sharply from all the preceding ones. While throughout all three preceding periods the bath is decarburized through the burning out of C, in the fourth the opposite occurs: a carbon-rich addition carburizes the soft iron of the bath into steel of the desired hardness. No less important a role is played by Mn in the spiegel, which is a very energetic reducing agent, both for iron oxide dissolved in the metal and for iron disilicate present in the slag. This reduction of Fe proceeds according to the reactions:

\[ \mathrm{FeO} + \mathrm{Mn} = \mathrm{Fe} + \mathrm{MnO} \tag{4} \]

\[ \mathrm{FeSiO_3} + \mathrm{Mn} = \mathrm{Fe} + \mathrm{MnSiO_3} \tag{5} \]

Table No. 7 very clearly illustrates the role of the manganese in the addition of the fourth period: the eighth slag sample gives 14.10% MnO, while the last, the ninth, contains MnO in the amount of as much as 23.96%.

From the side of the physical phenomena, bessemerizing is characterized by the following features: temperature, spectrum, color of the flame, and sound effects.

In the first period, when the temperature of the bath is still not high, about 1300°, and when the spectroscope shows only one yellow line of Na, the flame has a reddish-yellow hue and is accompanied by an abundance of sparks. Gradually, with the further course of the operation, the temperature rises; the flame grows lighter, taking on a slightly bluish reflection, and, whereas it was short in the first period, it now becomes longer. The spectrum, besides the bright yellow line, gives four weak green lines, indicating the beginning of a more or less considerable decarburization of the bath.

The process thus enters the second period, whose temperature is on average equal to 1400°. Very characteristic of the second period is the roaring sound of the flame, whereas in the first period it was bubbling. Toward the end of the first half of this period the color of the flame becomes dazzlingly white; in the spectrum, in addition to the bright green lines and the yellow one, the appearance of further blue lines is observed. The second period is the longest. The third, on the contrary, lasts only a few seconds in all and is characterized chiefly by a temperature that has already risen to 1450–1500° and by the disappearance of the green lines of the spectrum, indicating the end of the burning out of C. The flame of the third period, while remaining, in its color, still dazzlingly white, quickly shortens, becoming enveloped in brown smoke from gaseous FeO and iron oxide Fe₂O₃. The fourth period of the Bessemer process is separated from the third artificially by the addition of spiegel. Differing chiefly in its chemical aspect from all…

of the others, this period possesses the same physical characteristics as the preceding one, with the sole exception of the temperature, which rises to 1600° and more.

Comparing the diagram of South-Russian Bessemerizing with the diagrams characterizing foreign processes, we see that it is quite akin to them and belongs to the same category. Diagram No. 5 occupies, as it were, an intermediate position between the diagrams of English (No. 1) and American (No. 2) Bessemerizing. Indeed, the Si curve of all these diagrams, burning out very energetically in the first half of the operation, becomes flat in its second stage. The Mn curve, descending steeply in the first minutes, thereafter runs parallel to the axis of abscissae, both in American Bessemerizing and likewise in Swedish and South-Russian. The burning-out of C in the first half of the process, similarly to the English diagram, reveals a fictitious enrichment of the bath in carbon, whereas the second half of the curve for this element falls rapidly downward, as was seen in the diagram of American and Swedish refining. But the feature most individual to South-Russian Bessemerizing is the arc of fictitious enrichment of the bath with C, lasting only four minutes in all, while in the English diagram this phenomenon covers the first seven minutes of the process. These characteristic features of the diagrams under consideration are explained chiefly by the different chemical composition of the pig irons. First, the ratio C : S in the English heat was 1.58, whereas in the South-Russian it was 2.5; second, the English pig iron contained 2.26% silicon, the South-Russian only 1.48%. Both these factors contribute to the English diagram prolonging the phase of fictitious enrichment of the bath with C, while the South-Russian shortened it and thereby came closest to the diagram of American Bessemerizing.

4. Metallographic investigation of samples from the heat.

The study of the structure of Bessemer metal consisted in the microscopic examination of 10 polished sections of samples taken from the converter every two minutes throughout the entire process. These metallographic observations give an interesting and vivid picture of the genetic modification of the metal, which, during the period of its refining, passed through a series of stationary moments.

The raw pig iron, charged into the converter from the mixer, under the influence of the increasing temperature of the bath and of the chemical processes, is transformed into soft iron, changing, in the individual phases of its refining, both the eutectic and the form and size of the precipitated crystals.

Fig. 1 (v = 100) represents the structure of the gray pig iron charged into the converter. On the polished section it is possible to discern quite clearly...

but an abundant precipitation of graphite in the form of hairlike shoots, in places concentrated into a spiderlike form.

Fig. 2 \((v = 100)\) already represents the structure of the metal subjected to blowing in the converter for two minutes. The former character of the structure, so characteristic of gray cast iron, is preserved in this polished section only in a few places as large egg-shaped formations, which from the periphery are penetrated by bright-white needles of cementite \((\mathrm{Fe}_3\mathrm{C})\), surrounded in turn by a fine eutectic. Whether these large spots of gray cast iron owe their appearance to a secondary crystallization of graphite, owing to favorable cooling conditions¹, or whether the process of whitening the gray cast iron did not complete, in this time, the cycle of its constitution, cannot be said with certainty. This question has not yet been sufficiently developed in the metallographic literature and has not received a definitively established classification. Considering further the eutectic, a dark mass dispersed between the white needles of cementite, we see that in places it gathers into very considerable, rounded clusters, often arranged in a regular row, as may be seen on the right side of Fig. 2. At a magnification of 1000 times (Fig. 3) the eutectic clearly reveals the character of its structure, consisting of an aggregate of mixed crystals (Mischkristalle), small ovalized crystals. In Fig. 3, in its central part, there is placed a large isolated white crystal of cementite, into the mass of which numerous ellipsoids of pearlite have penetrated. The lower quadrant of the right semicircle gives the most relief-like picture of the transformation of cementite into eutectic under the influence of the internal thermo-chemical process.

The following, then, Fig. 4 \((v = 100)\), shows how the area occupied by the eutectic is continually increasing, while the large accumulations of cementite reveal a coarse-grained structure. At a magnification of 1000 times (Fig. 5) this porous structure, in its dark places, consists of lamellar pearlite, which is most clearly visible in the large dark band of eutectic situated in the left semicircle of the photograph. The polished section of the second sample (Fig. 4) is of further interest in that, in some dark spots of eutectic, along the line of their contact with the white crystals of cementite, there appears a light-gray, smoky medium, representing, as it were, a transitional stage from cementite to pearlite. This hazily represented eutectic is visible quite clearly in the lower quadrant of the left semicircle of Fig. 4.

Fig. 6 \((v = 100)\) represents a photograph made from the polished section of the preceding second sample (Figs. 4 and 5), but etched with picra—

of potassium. In contrast to the preceding etching (weak \( \mathrm{NO_3H} \)), the pearlite in Fig. 6 is colored white, and the cementite dark gray. The photograph shows two large white pearlite nuclei, revealing their lamellar structure quite distinctly. But the structure of the pearlite is most prominently shown in the third, somewhat grayish nucleus, of elongated form, situated in the lower quadrant of the right semicircle. All three of these pearlite formations are surrounded by a dark border separating them from the cementite. In the same dark consistency there float, as it were, smaller inclusions of pearlite scattered through the mass of cementite.

Fig. 7 is the same polished section (Fig. 6), but the upper layer, colored with potassium picrate, has been removed from it by rubbing with crocus on cloth. Prepared in this way, the polished section reveals the relief of its structure with the utmost clarity. The upper convex contour of the photograph bounds the cementite; the second, lower surface shows the unmistakable character of pearlite, whose lamellar structure has not yet been completely destroyed by gentle polishing; and, finally, the dark areas of the photograph, lying on the third, very lowest level, constitute the intermediate eutectic dissolved by potassium picrate. The eutectic that is dark in Fig. 7, black in Fig. 6, and light smoky in Fig. 4 is nothing other than ferro-silico-carbide. This termolecular eutectic is formed most abundantly in the structure of the third sample of Bessemer metal (Fig. 8), corresponding to the sixth minute of the process, when \( \mathrm{Si} \) burns out most vigorously. In Fig. 8, in the lower half of the disk, a considerable quantity of termolecular eutectic is observed, as also in Fig. 4, appearing as a smoky-white medium amid the dark formations of bimolecular eutectic (cementite, ferrite). At a magnification of the polished section of 1000 times (Fig. 9), this termolecular eutectic stands out especially demonstratively. Thus, in the lower quadrant of the left semicircle, to the left of a powerful precipitation of cementite, one can see two small whitish crystallites surrounded by a smoky mass of termolecular eutectic. The greater part of the photograph is occupied by pearlite, colored dark gray.

Fig. 10 (\(v = 1000\)) likewise detects the presence of termolecular eutectic, which here surrounds a small white crystal situated in the region of a large, somewhat irregularly shaped disk of bimolecular eutectic (upper quadrant of the left semicircle), surrounded in turn by a mass of spongy cementite.

That the whitish, smoky mass is a termolecular eutectic consisting of ferro-silico-carbide is confirmed by the following arguments.

In the classic work of Prof. F. Wüst and L. Laval, Experimentelle Untersuchung des Thomasprocesses¹, the authors, incidentally, say the following: “Ob auch das Eisencarbid etwas Silicium enthalten kann, ist nicht mit Sicherheit nachgewiesen, doch erscheint die Existenz eines Eisensilicocarbides nicht unmöglich.”

The somewhat indecisive tone of this quotation is not justified by the experimental evidence available to us. First, returning to diagram No. 5, we see that metal sample No. 3, the polished section of which has just been examined by us in Figs. 8, 9, and 10, corresponds to the ordinate of the sixth minute of the process, when C and Mn cease their burning out, extending the corresponding curves parallel to the abscissa axis, while the Si curve at this time gives a steep descent downward, indicating its vigorous burning out. Consequently, the influence of silicon inevitably must have affected the structure of the metal, given the inertness of C and Mn. The increase in the pearlite of the smoky-clouded mass records this activity of Si. Secondly, potassium picrate, dissolving Si sufficiently well in itself, gives a relief (Fig. 7) on which the dark places are at the very lowest level and thus indicate the presence of silicon in them².

Fig. 11 ($v = 100$) represents the structure of the fifth metal sample. The dark area of bimolecular eutectic has increased considerably in comparison with the preceding photographs, and this clearly indicates that the process of decarburization of the metal is proceeding at full speed. The white crystals of cementite now lose their spongy structure and are constituted into a fine but dense network. Only the most insignificant quantity of thermomolecular eutectic remains, which fully accords with the diagram of South-Russian Bessemerizing, whose Si curve indicates that there is already very little of this element in the bath.

Fig. 12 ($v = 100$) is of interest for its beautiful and original crystallization of cementite. Very little of the latter remains, and it appears in two varieties: as irregular, elongated crystals with a very jagged configuration, and in the form of regular, thin, needle-like formations. The black spots surrounding the cementite nuclei are sorbite. All the remaining space of a grayish shade is filled with homogeneous bimolecular eutectic.

Fig. 13, in its structure, bears the evident character of eutectoid steel. This polished section stands, as it were, on the boundary separating the crystallization of cementite and ferrite. In the indistinct, elongated whitish accumulations the contour of ferrite is already outlined, which in Fig. 14 of the following sample appears as a powerful network. The process of decarburization is approaching its end, and in Fig. 15 we encounter a very

characteristic structure of the iron. Carbon remains in the metal, as is evident from diagram No. 5, only in an insignificant quantity.

Finally, Fig. 16 gives us the structure of the finished material. From the fine structure of the polished section, in which the fragmented white grains of ferrite alternate with dark accumulations of pearlite, we may judge the character of the rail steel obtained.

5. The Thermo-Chemical Aspect of the Process.

The pyrometric observations presented in the second chapter of this work, when coordinated into a system in which temperature is plotted along the ordinate and time along the abscissa, give a picture of the rise in temperature of the Bessemer bath. The curves obtained in this way indicate that the temperature of the metal rises continuously throughout, while exhibiting very peculiar fluctuations.

However, is it possible to identify these temperature curves, the data for which were collected by photometric observations with Wanner’s pyrometer over the flame of the Bessemer converter during the melting period, with the actual temperatures possessed at the corresponding moments by the metal being refined? The initial points prove to be below the melting point of cast iron. The oxidation process in the initial stage is carried out entirely within the metallic bath; the heat liberated in considerable quantity by the combustion of Si and Mn goes to heating the metal; C, as was clarified above, almost does not burn out, owing to which a very insignificant content of CO₂ is found in the combustion products, whereas at that time the flame shows the maximum percentage content of nitrogen¹ and the maximum content of free oxygen, reaching 50%². All these factors naturally affect the lag of the flame temperature behind the temperature of the metal. Nevertheless, the flame-temperature curve of the period under consideration is equidistant from the actual curve of the rise of the metal temperature and thus undoubtedly gives a parallel picture of the thermal effects of the bath itself.

With the further course of the operation the oxidation process proceeds ever more vigorously, the temperature rises, a considerable appearance of CO₂ is observed in the combustion products (from 0.85 to 9.1 in 5 minutes from the beginning of the blow, by volume)¹, the unused oxygen becomes ever less (from 14.3 to 0.18 by volume in the same 5 minutes), and the combustion taking place in the metal is not completed there as a whole, but part of it is already completed in the region of the flame. The temperature of the latter and of the bath becomes identical throughout the second and third periods of the blow, while

¹ Prof. Pavlov. Metallurgy of Steel, p. 40.
² Journal of the Russian Metallurgical Society, No. 2, 1910. Article by Capt. Davydov, “Investigation of Certain Phenomena in Bessemerizing with Surface Blowing.”
³ Journal of the Russian Metallurgical Society, 1910, No. 2, p. 48.

in the last minutes, there is reason to suppose that the temperature of the flame is even somewhat higher than the temperature of the bath, owing to the fact that, because of the bad reaction and splashes of metal, a considerable percentage of the particles of the gaseous combustible is mechanically carried by the blast jet out of the bath and burns already outside the converter¹.

Having plotted the temperature curve of heat No. 1, we had to note the striking connection and harmony in the fluctuations between the obtained curve of temperature rise and the curve of silicon burning, corresponding to diagram No. 5. Although a priori one had to suppose a close connection between these two curves, since the chief agent of heat in the Bessemer process is Si, nevertheless the results obtained, when both curves were plotted on a common coordinate system with one and the same scale, surpassed the expectations of theoretical assumptions.

Indeed, in diagram No. 6 this mutual dependence of the curves is very clearly discernible. Both the one and the other have the form of decreasing sinusoids, with the curve T going upward and the curve Si, conversely, downward.

Both curves consist of phases turned now convexly, now concavely toward the axis of abscissas. The phases of the curves are correspondingly harmonious with one another in an inversely proportional relation: if the phase of the temperature-rise curve is convex with respect to the axis of abscissas, then the phase of the silicon-burning curve corresponding to it in time is concave. During the first three minutes the curve T is convex toward the axis of abscissas, while the Si curve is concave. The ordinates of the third minute give points of inflection on both curves. The curve T now becomes concave, and the curve Si convex toward the axis of abscissas. This second phase lasts until the ordinate of 5½ minutes, where both curves, again passing simultaneously through an inflection point, change the character of their curvature.

Further we note the points of the ordinates of 8⅕ minutes, 10 and 12⅓, at which the curves again change their curvatures with respect to the axis of abscissas.

This wave-like character of the curves is a very vivid illustration of that internal thermochemical process which takes place in the bath and which indicates the struggle between Si and C for oxygen, the result of which is the oscillations in the rise of temperature. No lesser confirmation in these curves is also obtained by the law of molecular concentration, which theoretically substantiates that peculiar and regular process which cements into one harmonious, integral complex the three principal agents of Bessemerizing: temperature, the rate of burning of Si, and the rate of burning of C.

Since the increase of temperature in the process under consideration, during its second and third periods, depends entirely on the rate of combustion of silicon, and this latter in turn reveals, through the temperature, a dependence on the rate of combustion of C, we are able to discern a certain functional relation in these phenomena.

Denoting by \(v\) the rate of combustion of Si, and by \(w\) the rate of increase of temperature of the Bessemer bath, we have

\[ w = f(v) \text{ and } v = \varphi(w) \tag{6} \]

Substituting the first into the second, we find

\[ w = f[\varphi(w)] \]

Differentiating the expression obtained,

\[ w' = \{ f[\varphi(w)] \}' \]

\[ 1 = f'(v)\cdot \varphi'(w) \]

or

\[ \varphi'(w) = \frac{1}{f'(v)} \tag{7} \]

The last expression is a property of inverse functions, asserting that the derivative of the inverse function is equal to the reciprocal value of the derivative of the direct function. Thus, the property of the curves of diagram No. 6, establishing the harmony of their phases with one another in an inversely proportional relation, is also confirmed by equation 7.

The desire to remove the observed property of the curves of temperature increase and silicon combustion from the realm of chance was the reason for new analyses for silicon in one of the following heats, in order to obtain a control curve.

For this purpose, metal samples from heat No. 3 were subjected to analysis to determine their percentage content of Si. In view of the substantial importance of the control curve, it will be appropriate to describe briefly here the method of determination by means of which the analyses of the samples from heat No. 3 were carried out, already in the metallurgical laboratory of the Tomsk Technological Institute.

A weighed portion of metal was taken in the amount of 1 gram, which was dissolved in 3 cubic centimeters of strong hydrochloric acid (1.19). After sufficient dissolution of the metal and evaporation of HCl on a sand bath, 5 cubic centimeters of strong sulfuric acid, diluted in 5 cubic centimeters of distilled water, are added to the solution. A second evaporation on a sand—

the acid bath was continued until the precipitate was completely dry and until the evolving white vapors of \(\mathrm{SO}_3\) had, as far as possible, weakened.

After the beaker had cooled, 15 cubic centimeters of HCl and a double quantity of water were again added. The heated solution, after slight boiling, was filtered. The residue was washed on the filter with heated hydrochloric acid \((1:2)\).

Next the filter was washed until the reaction for iron disappeared (tested with rhodanammonium). After this operation, the precipitate, together with the filter, was slightly dried and ignited in a platinum crucible until snow-white \(\mathrm{SiO}_2\) was obtained. The latter was weighed and the silicon was calculated.

Analyses carried out in this way on samples from heat No. 3 gave the following results:

Table No. 8.

Plotting the data of Tables No. 3 and No. 8 on one and the same coordinate system, we obtain diagram No. 7 with two curves: temperature and silicon. This control diagram, at first glance, reveals a complete analogy with the preceding one (No. 6), emphasizing no less decisively the same phase oscillations and the same functional dependence that was established above.

Indeed, in the first two minutes from the beginning of the blow, the curve \(T\) is concave toward the axis of abscissas, while the curve Si is convex. The ordinates of the second minute simultaneously give points of inflection for both curves, which change their curvature, lasting until the fifth minute, where again an inflection point is observed on both the one and the other curve. The following phases of the curves are bounded by the inflection points of the ordinates at 8 and \(10^{1}/_{3}\) minutes.

At 12 minutes, an addition of spiegel was made. To obtain the control curve, samples of the third heat were selected because the temperature curve of this latter gives the greatest amplitude of the oscillations of decreasing phases, and, consequently, the burn-out curve of the residual Si should have displayed the greatest wave-like character.

Thus, formulating the properties discovered above for the curves of diagrams No. 6 and No. 7, we obtain the following three propositions:

a) The rise of the curve of increase in the temperature of the Bessemer bath is proportional to the rate of burning-out of silicon;

b) The phases of the oscillations of the silicon burn-out curve are harmonic in an inversely proportional relation to the oscillations of the corresponding phases of the temperature-rise curve;

c) The points of inflection of the curves \(T\) and \(Si\) coincide on the ordinates with respect to time.

In addition to these features of the curves being studied, they possess one further, very interesting property. Thus, when examining control diagram No. 7, a tendency was found for the ordinates of the curves to maintain, in sum, one and the same constant magnitude throughout the entire second and third period of the process. If we denote the ordinate of the temperature-rise curve by \(l_1\), and the ordinate of the silicon burn-out curve by \(l_2\), then from diagram No. 7 we have, in centimeters:

\[ \begin{aligned} l_1^{\mathrm{III}} + l_2^{\mathrm{III}} &= 6{,}6\\ l_1^{\mathrm{IV}} + l_2^{\mathrm{IV}} &= 6{,}7\\ l_1^{\mathrm{V}} + l_2^{\mathrm{V}} &= 6{,}8\\ l_1^{\mathrm{VI}} + l_2^{\mathrm{VI}} &= 6{,}7\\ l_1^{\mathrm{VII}} + l_2^{\mathrm{VII}} &= 6{,}6\\ l_1^{\mathrm{VIII}} + l_2^{\mathrm{VIII}} &= 6{,}5\\ l_1^{\mathrm{IX}} + l_2^{\mathrm{IX}} &= 6{,}7\\ l_1^{\mathrm{X}} + l_2^{\mathrm{X}} &= 6{,}7\\ l_1^{\mathrm{XI}} + l_2^{\mathrm{XI}} &= 6{,}7 \end{aligned} \]

Generalizing, we find:

\[ l_1^i + l_2^i = \mathrm{Constans} \tag{9} \]

Diagram No. 6, although it also shows the same tendency of the curves to preserve constancy in the sum of their ordinates, does so to a somewhat lesser extent. Unfortunately, the limitation of the present investigation by the framework of a diploma thesis did not allow us to dwell in greater detail on the development of the noted properties of the curves being compared, and especially on the proof of equation 9. Meanwhile, if further investigations confirm the validity of the propositions outlined, then the question may be transferred from the theoretical realm into the sphere of practical metallurgy and realized in plant practice. The point is that equation 9 gives the idea of determining the percentage content of \(Si\) in the bath immediately by the pyrometric method, avoiding chemical analysis: knowing the temperature at a given moment of the process, and consequently its ordinate on the corresponding scale, it would be possible, from tables compiled in advance, to find the percentage content of \(Si\) in the bath.

Let us now consider jointly all the available pyrometric observations entered in the tables of the second chapter and encompassing...

five Bessemer heats. Plotting, for this purpose, the data of Tables No. 1, No. 2, No. 3, No. 4, and No. 5 on a common system of coordinates, we obtain five curves of Diagram No. 8.

The curves of the first four heats proceed as interlacing rays of a bundle, and only the curve of heat No. 5 shows a certain deviation from the general character of the others. The dotted lines that break the curves represent the addition of spiegel iron. As is seen from the diagram, the bath lowers its temperature during the stoppage of the blast for the addition by a quite negligible amount, with the exception of heat No. 1, in which the temperature fell by 170°. This jump in temperature is explained by the fact that the process of this heat was artificially prolonged somewhat longer than was usually done, in order to give a complete picture of the burning-out of the elements. The considerable temperature, which reached 1635°, naturally contributed to a more energetic cooling of the bath than in ordinary heats, with the temperature at the moment of addition being only about 1480° (No. 2 and No. 3).

The curves of the rise in temperature of the Bessemer heats, plotted in this way on a common coordinate system, revealed a very interesting feature—to intersect at one point, corresponding to an ordinate of \(10\tfrac{1}{3}\) minutes from the beginning of the process. The only exception was the curve of heat No. 5, and thereby, as it were, it revealed a certain anomaly. Indeed, the initial moment of this heat gives the lowest temperature of all the five curves under consideration. The process proceeds in a sluggish manner and, despite the fact that the curve from the 9th minute, precisely obeying the general law of temperature curves to intersect at one point, shows a considerable effort to rise upward, nevertheless it passes significantly to the right of the common point of intersection.

Drawing the abscissa of the common intersection of the curves, we find that this point corresponds to 1420° C.

What, then, is the special character of this moment from the standpoint of internal thermochemical phenomena? To clarify this question, let us turn to Diagram No. 5 of the burning-out of the elements. The ordinate of the point under consideration, whose abscissa is a time equal to \(10\tfrac{1}{3}\) minutes, marks on this diagram that decisive moment of the second period of the process when the intense struggle that has lasted until then between C and Si for oxygen ends in favor of the first element. From this moment the carbon curve rapidly descends, while the Si curve, finishing its energetic burning-out, slowly begins to extend parallel to the axis of abscissas.

At this point we are dealing, as it were, with the decisive moment of the heat, and it will be natural to propose calling the intersec-

...of the temperature curves of the various heats—the critical point of the Bessemer process¹.

The temperature curve of heat No. 5, by not passing through the critical point of the process, thereby confirms, and does not deny, the existence of this critical moment in the Bessemer heat. Indeed, from the observation records it is apparent that, at the beginning of the blow, the foreman conducting the heat charged a considerable quantity of scrap, in view of the somewhat higher percentage content of Si in the pig iron supplied than had been the case earlier, wishing thereby to prevent an excessively hot heat. But it turned out that this precaution was excessive and became the cause of the opposite phenomenon: the heat became cold before the addition, and after it the curve reveals a high rise in temperature, reaching 1635°. Exactly the same picture of a strong increase in temperature after the addition is also shown by the curve of heat No. 4, the cupola of which had been charged somewhat early: the curve had not yet passed the critical point.

On the other hand, heats No. 2 and No. 3 must be regarded as quite normal and successfully conducted. As their temperature curves show, the addition was made 1½–2 minutes after they had passed the critical point, i.e., at the moment when Si and C in the bath had already completed their energetic burning out and their presence in the metal did not exceed—of the former, hundredths of a percent, and of the latter, tenths. Owing to this there occurs a calm rise in temperature after the introduction of the spiegeleisen, and the curves give a maximum temperature within the limits from 1560° to 1610°.

The material from heats conducted according to this scheme is entirely of good quality, as is shown by tensile tests in the mechanical laboratory, which gave the following results:

Table No. 9.

Diagram No. 8 draws attention by one more very curious phenomenon. This latter consists in the fact that the curves of temperature increase, at the end of the process, show a stoppage and for some time give horizontal straight lines.

Thus, melting curve No. 3 is horizontal during the 23rd and 24th minutes; curve No. 2 is horizontal in the second half of the 21st minute and throughout the 23rd and 24th; curve No. 4 is horizontal during the 21st minute; and No. 1 during half of the 15th and 16th minutes. During these periods the metal in the bath is boiling exactly, while maintaining its temperature constant.

Whether this stage of the end of the process is the natural and desirable outcome of the metal cannot be said, in the absence of more extensive experimental data. This question can be answered by further experimental data from the mechanical and metallographic laboratories, depicting both the degree of resistance of the material to destructive forces and the structural changes of form in the metal and the character of its crystallization.

Turning to the diagrams of foreign Bessemer processes, cited in Chapter 1 of the work, we can see, from the character of the curves, at what moment these heats pass their critical point. The earliest passage is given by Swedish Bessemerizing (diagram No. 3), whose critical point lies on the ordinate of the second minute, and this last moment is the most characteristic for the Swedish process.

The American diagram (No. 3) of the burning out of elements indicates that the critical point of its temperature curve must be between the 3rd and 4th minute from the beginning of the blow. The English process passes its critical point between the 7th and 8th minutes. As for the Ural Bessemerizing, it entirely lacks this characteristic moment of the Bessemer process, and the critical point is absent here.

6. Conclusion.

Concentrating all the experimental data that characterize the chemical and physical phenomena of the Bessemer process into a single scheme, we obtain the following table (see p. 23).

From all that has been set forth above, it is apparent that the thermal factor in the Bessemer process plays a cardinal role, decisively influencing all its physical and chemical phenomena.

The curve of temperature increase is in close organic connection with the rate of silicon burning, showing a regular, functional dependence with respect to the latter.

Metallographic investigations confirm the conclusions of thermochemical analysis. The significance of pyrometric control observations, therefore, during a Bessemer heat becomes beyond doubt.

Neither personal experience of many years, nor spectroscopic observations, nor the color and character of the flame can at any moment of the process give that true picture of the thermochemical transformations taking place within the bath which the pyrometer reveals.

Table No. 10.

Its indications apply to all the above-enumerated methods for determining the moment of addition and the end of the refining of the metal, just as an objective criterion relates to subjective estimates.

In every rationally equipped metallurgical plant, at the Bessemer converters there should be a pyrometer, as their indispensable companion.

In conclusion to this work, we may say that both personal, very modest experience and, to the greatest extent, the investigations of recent years on Bessemerizing and Thomasizing by foreign¹ and Russian² experimenters indicate that technology has by no means yet taken from Bessemer’s patent all that it can yield.

1. March.

L. Krasin.

¹ Wüst, Howe, Campbell, Wiborgh, Le-Chatelier, Laval, Hartig, Shelby, and others.
² Pavlov, Weinberg, Davydov, and others.

To L. Krasin’s article.
Table I.

Fig. 1. \(v = 100\)

Fig. 2. \(v = 100\)

Fig. 3. \(v = 1000\)

Fig. 4. \(v = 100\)

For the article by L. Krasin.

Table II.

Fig. 5. \(v = 1000\)

Fig. 6. \(v = 1000\)

Fig. 7. \(v = 1000\)

Fig. 8. \(v = 100\)

For the article by L. Krasin.

Table III.

Fig. 9. $v = 1000$

Fig. 10. $v = 1000$

Fig. 11. $v = 100$

Fig. 12. $v = 100$

Concerning the article by L. Krasin.

Table IV.

Fig. 13. \(v = 100\)

Fig. 14. \(v = 100\)

Fig. 15. \(v = 100\)

Fig. 16. \(v = 100\)

For Lafrasit’s article: “On the question of an experimental-graphical investigation of the Bessemer process.”

Plate 5. Diagrams.

1.

2.

3.

4.

5.

6.

7.

8.

1. The author considers the experiments on the basis of which the critical point was established to be far from sufficient. Therefore, this question is put by him not in the form of an assertion, but only of a proposal. ↩↩↩↩↩↩↩↩↩↩

2. The polished section was kept in a boiling solution of potassium picrate for 20 minutes. ↩

3. Mittheilungen der Königl. techn. Versuchsanstalt zu Berlin (1883). ↩