The Technology of Low Temperature Carbonization


Frank M. Gentry

[ Chapter 7: Operation, Design, & Materials of Construction ]

Table of Contents

Preface & Table of Contents
Chapter I ~ Fundamentals
Chapter II ~ Low Temperature Coal Gas
Chapter III ~ Low Temperature Coal Tar
Chapter IV ~ Low Temperature Coke
Chapter V ~ Nitrogenous & Other By-Products
Chapter VI ~ Processes of Low Temperature Carbonization
Chapter VII ~ Operation, Design, & Materials of Construction
Chapter VIII ~ Economics & Conclusion
Name Index
Subject Index

Chapter VII

Operation, Design & Materials

Operation of Retorts
Design of Retorts
Materials of Construction
Refractory Retorts
Properties of Refractories
Metallic Retorts
Properties of Cast-Iron
Properties of Steel
Heating of Retorts
Convection & Radiation

General ~

Lander conceives the successful development of a commercial carbonization process as passing through four distinct stages first, laboratory investigation of the method under close control and with accurate measurement; second, work on an intermediate scale in a plant able to deal with several hundred pounds of raw material daily; third, erection of a full-scale unit with a daily capacity of not less than 5 tons, which will allow multiplication in number; and fourth, a commercial battery which consists of a number of full-scale units erected in a locality which will test its economic possibilities.

In the foregoing chapters, we have dealt at length with the phenomena accompanying the transition of coal under application of heat, so that the principles to be noted during the first and second stages of development have been laid down. The successful solution of any low temperature process depends upon the utmost attention to and thoughtful working out of details. While a knowledge of the physico-chemical reactions accompanying the thermal breaking down of the coal conglomerate is the sine quo non of any successful process, of quite as much importance is a knowledge of the properties of the materials of construction, of the difficulties of operation, of the principles of design, and of the economies which justify its existence. It is the purpose of this and the next chapter to consider these phases of the subject which are of particular interest during the third and fourth stages of commercial development.

Operation of Retorts ~

Caracristi (256) has given a very frank discussion of the operating difficulties of low temperature carbonization, as has Curtis (213) and his associates). To preserve the primary character of the low temperature products, it is necessary to avoid localized heating, which causes decomposition of the hydrocarbons and adversely affects the process, both by virtue of loss of valuable products and by the deposition of amorphous carbon, which is detrimental to the continuous operation of the plant. The deposition of amorphous carbon, sufficiently hard to disrupt the driving mechanism, was experienced in the Carbocoal primary retort, resulting in the adaption of a cutting blade to the paddles, as illustrated in the McIntire modification.

Temperature regulation is of great importance for a number of different reasons, principally among which is the sensitivity of the quality of the low temperature products to the temperature of distillation and the danger of exceeding a temperature which will injure the retort, if it be constructed of metal. Of course, the lower the temperature consistent with complete carbonization, the greater is the thermal efficiency of the process. Great care must be taken to prevent infiltration of air, either to the combustion chamber or to the carbonization chamber, in order to maintain close temperature regulation, in the first case, and to prevent destruction of the hydrocarbons, in the second.

Tempering of the semi-coke to prevent spontaneous combustion is a serious matter. Of course, this can be effected by water-quenching the material, as is done in some instances, but this is not good practice, for it breaks up the lumps, producing excess breeze, and lowers the net value of the fuel by virtue of its moisture content. Furthermore, it is desirable to recover the sensible heat of the coke, as far as is possible, in order to increase the thermal efficiency of the process. This can be accomplished by means of heat recuperators of by dry quenching. The spontaneous combustion of semi-coke can be prevented easily by crushing and pulverizing it, provided it can be sued in this form, for the finely divided material permits the absorption of sufficient oxygen to prevent further spontaneous oxidation.

It is common experience for low temperature processes to encounter the building up of a thick layer of carbon on the inner wall of the retort shell, due partly to cracking of hydrocarbons and partly to the sticking of fusing coals. This is a serious condition, to be avoided in all events, for aside from its interference with the operation of moving parts, it greatly reduces heat transfer because of its thermal insulating character, and increases the likelihood of injuring a metallic retort through overheating in an effort to attain the necessary heat transfer to carbonize the coal properly. The carbon layer bakes harder with age and is difficult to remove.

It seems that a great deal of trouble has been found in handling fusing coals in a rotary retort, because they stick to the walls and build up a thick heat-insulating layer. To avoid this difficulty, recourse has been had to various mechanical and chemical devices. Thus, Hutchins introduced a star-shaped breaker into his Fusion process to chip the coke from the inner walls, while the K.S.G. process destroys the sticking quality by preheating the coal before carbonization. No serious mechanical difficulties are encountered in carbonizing non-coking materials when in motion, but because of the absence of binding resins, the residuum after distillation is pulverent in nature and of little value as fuel, unless further processed by briquetting or by pulverization.

Another source of trouble in practice is the clogging of the raw coal feed pipes. This arises from the condensation of hydrocarbons or water vapor on the cool incoming coal to form a sticky paste. The Piron-Caracristi process partially remedied this difficulty by admission of steam with the coal, but finally completely avoided it by projecting the feed pipe into the hot oven. This expedient was adopted also in the McEwen-Runge process, where the pulverized coal is introduced through pipes extending several feet into the carbonization shaft.

In those processes in which molten lead is used as the heating medium, Caracristi (256) reports that there is no difficulty in holding the bath of lead in a refractory structure. Except for mechanical defects, such as cracks, bricks, which had been in use over a period of several months, shoed no indication of lead infiltration upon examination. Furthermore, the loss of lead may be effectively prevented by a judicious location of traps and cooling pipes.

Moving metallic parts, especially those made of cast-iron, which are subjected to high temperature, such as chain conveyers, are subject to growth. This subject will be treated more fully later, but it should be pointed out that some provision must be made for taking up the slack or bringing the moving parts into register.

McEwen and Runge depend upon convection to transfer the heat from their recirculated hot gas to the pulverized coal particle. Gentry (208) has pointed out that the transfer of heat by convection in such circumstances is a function of a power of the relative velocity of the fluid and the particle. The laws of falling bodies in non-vortical fluids are well known and it is safe to conclude from them, even for a case of turbulent flow, that there are certain definite limits to heat transfer between a particle and a gaseous fluid in motion at certain gas velocities, due to the reduction to zero of the relative velocity of particle and fluid. This sets a maximum limiting velocity of gas circulation, which depends, among other things, on the size of the particles, density of the gas, etc., beyond which the efficiency of heat transfer is reduced and beyond which difficulties will be encountered with the gas carrying the pulverized coal out of the retort. One very important aspect of low temperature carbonization in pulverized form is found in the fact that, both by convection and radiation, heat transfer depends directly upon the area of the absorbing body. Consequently, carbonization of small particles should and does effect an enormous reduction in the tie of distillation.

In the early days of low temperature carbonization, the first attempts to develop a continuous process naturally led to an internal spiral conveyer. Most of these efforts utterly failed, because the inventors could not prevent the charge from plugging in the retort. The Greene-Laucks process seems to have successfully surmounted this difficulty, and probably the provision for heating the inner surface of the screw has a great deal to do with the elimination of this trouble. In the Marshall-Easton process, which also uses internal spiral conveyors, the problem has been solved in an ingenious mechanical way by providing a nest of interlocking spirals, so that no rotary motion whatever is imparted to the charge, which moves only vertically through the retort.

Rotary retorts have received a good deal of attention in Germany, where several large-scale plants have been in operation for a number of years. Rotary retorts, heated externally or internally, have certain well known advantages. In the first place, they are continuous processes and avoid high maintenance, inevitably associated with a discontinuous process, where the retort is periodically heated and cooled. Furthermore, the rotary retort permits bulk treatment of the charge, which requires less handling and correspondingly lowers the cost of processing per unit of throughput. The constant tumbling of the charge during carbonization assists heat transfer enormously by presenting fresh surfaces to the hot wall, in the case of external heating, and to the hot gas in the case of internal heating. As a consequence of this stirring up of the coke, heating is very uniform and likewise the volatile content of the semi-coke. On the other hand, there is a great diversity in the size of the lumps of semi-coke produced. Some pieces are very large and must be broken, others are just right for domestic use, and there is a relatively large amount of breeze, which must be briquetted for the market. Many experimenters with rotary retorts, however, have had a god deal of trouble with dust raised in the retort by the tumbling charge and carried over into the tar main by the gas. This reduces greatly the value of the tar and causes much trouble in the by-product plant. In certain instances, even after the installation of special dust-catchers, 2% dust was carried over into the tar.

Design of Retorts ~

Despite the basic simplicity of the low temperature carbonization process, reduction to practice in commercial installations introduces a number of complications of a serious nature. Caracristi (256) summarizes them as follows:

(1) The necessity for large tonnage throughput, per unit of time and cost.

(2) Low heat conductivity of coal in mass.

(3) The difficulty of constructing an apparatus in which heat losses are minimized to a point where heat input is not prohibitive.

(4) The formation of gases which condense even at relatively high temperatures into sticky resins or tars.

The use of any structures similar in design to those employed in high temperature distillation is entirely precluded in primary carbonization by the first two difficulties listed above. It is necessary to design the structure in such a manner that the evolved gases pass from their point of origin to a region no higher in temperature than that at which the gases were evolved, in order to prevent cracking them. At the same time, it is equally important to prevent their passage into regions which are cool enough to cause their condensation, a condition which injures the oils through redistillation, as they trickle again into the hot zones, and a condition which tends to block the operation of the retort by the formation of a pasty mass.

Caracristi (256) apparently believes that the third factor listed above is the dominating problem in low temperature carbonization. Unless due effort is made to reduce heat losses, their cost may reach a value which will entirely outweigh the gain in value of the recovered by-products. Reduction of heat loss can be accomplished by proper thermal insulation or by an increase in plant throughput at a given temperature. The only effective heat is that which goes into the charge and a large part of this is carried away as sensible heat of the products.

The fourth consideration, that of the formation of sticky resins, creates an operating situation which is difficult to overcome in commercial practice. The sticky property of coals, of course, disappears with the volatile content, so that the trouble from this source decreases in a given coal as coking proceeds.

We have witnessed in Chapter VI the variety of structure of widely different design in which low temperature carbonization has been effected. These structural differences arise from an understanding of the fundamentals of the coking process and form ingenious solutions of its various problems, particularly those of dealing with fusing coals, of securing the necessary heat transfer, and of obtaining a high throughput. The first of these problems has been discussed already under the subject of operation of retorts. The question of heat transfer has more or less resolved itself into the principle of external heating in thin layers, the principle of internal heating, or the principle of stirring the charge. These problems have been discussed at various times, particularly under the subject of heating and resolves itself into multiplication of small units, in the case of heating externally by the principle of carbonization in thin layers, or mass distillation, in the case of internally heated or of rotary retorts.

Mass throughput requires continuous operation. As pointed out before, so long as the fuel is shale or other non-coking material, no trouble is experienced with moving parts, but the swelling and sticking of caking fuels renders the retort inoperative and is a difficult property of coal to surmount. Since the production of a domestic fuel requires the manufacture of a coherent product, this type of fuel is the class most usually met in practice. According to Simpkin (257), continuity of operation has been effected in low temperature carbonization processes in the following manners:

(1) Retorts constructed upon the principle of the tunnel kiln with provision to transport the coal through the carbonization chamber in suitable containers.

(2) The vertical shaft retort in which the charge is transported through the carbonization chamber under its own weight.

(3) Retorts in which the charge is transported by rotation of the structure itself.

(4) Retorts in which the fuel is transported by agitation from internal mechanical devices.

Intermittent working is severe on metal retorts, which grow hot when empty and thereafter cool rapidly when filled by the incoming charge. A semi-intermittent operation, accomplished by dropping the charge of vertical retorts a few feet at a time, such as was adopted in the Fuel Research Board narrow vertical retorts, apparently provides a satisfactory compromise.

It has been observed that much enterprise and ingenuity has been expended by those responsible for the development of the various low temperature processes and naturally the question arises as to which method is the most satisfactory. Due consideration must be given to a number of factors in reaching such a decision, as outlined by Simpkin (257):

(1) The character of the material to be processed.

(2) The relative importance attached to the products of carbonization.

(3) Efficiency, reliability, and simplicity of the method.

(4) The initial and operating costs per unit of throughput.

In consideration of the potential fuel resources residing in the great variety of carbonaceous materials, ranging all the way from mine refuse and shales to high grade bituminous coal, designers of low temperature retorts have been justified in constructing certain retorts with the intent of treating only one variety of material. At the same time others have set for themselves the more ambitious task of designing a retort of greater flexibility in the type of materials which it can treat.

The solid residuum from the distillation of shales is practically worthless, while that from colliery waste is of doubtful value. In certain circumstances the residuum from carbonization of the latter might be used as a producer fuel. With such materials as these, the best process is one which gives special attention to the production of tar and gas. On the other hand, when the object in view is to process better grades of fuel to extract their by-products or to modify the nature of the solid material, the selection of the best process would rest upon the special character ascribed to the various products, means at hand for their disposition, and finally the market condition for by-products in that locality. Obviously, if the primary object is the production of a domestic fuel, no process can be considered which delivers the semi-coke for the most part in a finely divided form. On the other hand, if the purpose of carbonization is the production of a power char, less discrimination can be made in the selection of the raw fuel for distillation and the delivery of the finished product in a finely divided state may be more of an asset than a deterrent to the process. If the solid product is desired in a firm or lumpy condition, there is little doubt that the oven type of retort is the best, but if throughput of material is the first consideration, without regard to the conditions of the semi-coke, either the vertical shaft or rotary retort should be given preference. It has already been pointed out in Chapter VI, under the subject of adaptability of processes, that no single retort will yield a maximum of even two of the most important products. The production of large yields of oil is ordinarily accomplished by a friable coke, and likewise, if a large volume of gas is wanted, the coke is somewhat less desirable and the oils are low both in quantity and quality.

The economic efficiency of a retort resides not alone in its thermal efficiency. The most desirable process for the treatment of a given fuel to yield predetermined products of known relative importance is that which delivers the semi-coke, tar, and gas at the minimum cost per unit. Thermal efficiency is, of course, a factor in the attainment of this end, but of even greater importance is the reduction of overhead and operating costs. Simplicity of design will usually contribute not only to a minimum initial expenditure, but to reduction of operating expense, through elimination of maintenance and attention that is required in more complicated operations. On the other hand, the extra capital requirements to effect continuity and automation will often be more than offset by savings in the cost of labor.

Evans (84) conducted a comprehensive series of experiments to determine the nature of the carbonization of coal at high temperatures in vertical retorts. His work is of equal interest in low temperature carbonization, because particular attention was given to the inward movement of the plastic layer and its effect on the path of travel of the gases. We have already seen in Chapter I, under the subject of plastic layer, that coking coals fuse somewhere in the vicinity of 400 C and form a plastic zone which progresses towards the center of the retort as coking proceeds. It was demonstrated that this plastic layer was approximately one inch thick and that its resistance to the flow of gas was of an order several thousand times that of raw crushed coal and several hundred times that of medium temperature coke. Evans found that the maximum pressure developed shortly after the passage of the plastic layer, at a temperature of roughly 425 C, and that the coke solidified shortly thereafter at approximately 440 C.

According to Evans (84), the difference in pressure across the plastic layer is dependent upon the percentage of voids in the charge and therefore upon the size to which the coal is crushed. Thus, he found a difference in pressure equal to 60 inches of water when using an unscreened coal with a small percentage of voids was used. He made the remarkable observation that the resistance of gas was dependent upon its direction. Thus the pressure difference between two points about 5 inches apart was 30 inches of water, when the gas flowed across the plastic zone toward the core of the retort, and 40 inches of water when the direction of flow was toward the retort wall. This shows that the plastic layer is more resistant to gas flow when backed by coke than when backed by green coal.

While in the past there has been a good deal of difference of opinion as to the relative quantities of the volatile products which pass upward inside and outside of the fusion zone, this question seems to have been pretty well established by Evans' experiments (84). Since it is evident that the flow of gas will follow the path of least resistance, due consideration must be given to the character of the original coal in determining whether the gas passes through the cold core of green coal or through the hot ring of semi-coke. Obviously, the coarser the raw coal and the greater the voids, the greater will be the proportion of the volatile products which pass upward through the core. In high temperature carbonization Evans estimates that roughly 90% of the gas is evolved after passage of the plastic condition. In low temperature carbonization, probably 25% of the gas is evolved inside of the plastic envelope and 75% outside. In addition to the amount of gas liberated in the respective regions, the quantity which is removed by the different passages depends on their relative resistance. At the beginning of distillation, the area within the plastic envelope is large and the area of the annular passage through the semi-coke is small, so that the resistance to gas flow is roughly inversely proportional to the area of the passages. Consequently, some of the gas liberated outside the plastic envelope breaks through the fusion zone and escapes through the core. As carbonization progresses, the inner passage becomes smaller and the outer passage correspondingly larger, so that a smaller and smaller percentage of the gas flows through the plastic zone and out through the core passage. During the first 3 hours of high temperature carbonization, Evans (84) found that approximately equal quantities of the volatile products passed outward through the raw coal and outward through the coke, but after that period, flow through the interior practically ceased. In low temperature externally heated vertical retorts, it is estimated that approximately half of the gas takes each passage. As the fused layer migrates inward, the region of maximum pressure, which is always outside of the plastic envelope, follows, but the distance between the two regions constantly becomes smaller. In general, the maximum pressure decreases in value as it moves towards the core and finally drops rapidly when the plastic envelope reaches the center and disappears.

In vertical retorts, a certain taper is necessary to facilitate discharge of the coke. Evans (84) has pointed out the very interesting fact that, for a retort of given area and given taper, there is a given rate of heating which will permit the plastic layer to reach the center of the retort at the bottom before it does so at the top. If this rate if heating is exceeded, the fusion layer will close first at the top of the retort and trap all volatile products inside of the plastic envelop at the bottom. This causes enough pressure to build up to force the gas through the fusion zone into the free passage outward through the hot coke. It might also be pointed out that the resistance to flow of gas along the retort walls, as compared to that through the core, is greater at the bottom of the a vertical retort than at the top. As a consequence, a larger percentage of the volatile products which are evolved at the bottom of the retort will flow through the core, than is the case for those evolved at the top.

In the design of low temperature retorts, it is well to bear in mind the various stages in which carbonization proceeds. These have been outlined by Fulweiler (258) as follows:

(1) A preliminary decomposition which begins as soon as the coal has acquired a certain fairly definite temperature. As this stage is quite strongly endothermic and approaches a fusion, the temperature remains fairly constant until completion.

(2) The products from the first stage, consisting principally of higher members of the alipathic hydrocarbons, suffer considerable molecular rearrangement. In general, compounds containing less than 3 atoms of carbon are formed. This stage may be regarded as a continuation of the simplification in which every distillation results.

(3) The gaseous vapors, resulting from the second stage, when evolved from the protecting influence of the actual coal particles, are acted upon by the conducted and radiant heat of the more highly heated portions of the charge proper, of the sides of the retort, and of the superheated regions above the coal.

The first two stages take place more or less simultaneously within the charge itself. The reactions which take place in the third stage are very complicated, depending as they do upon the time of exposure and upon the temperature. We have seen in the foregoing chapters that the mechanism of the third stage consists of the spitting up and breaking down of the alipathic hydrocarbons and their reunion into complex carbo-cyclic compounds. The benzene hydrocarbons may be further decomposed with the liberation of hydrogen and carbon and the formation of still higher cyclic derivatives. It is the third stage of distillation that is tremendously affected by the method of carbonization, and therefore by the conditions under which it occurs.

Fulweiler (258) also has listed the 6 factors which influence carbonization. These general conditions are:

(1) Size of the coal particles.

(2) Moisture content of the coal.

(3) Temperature of carbonization.

(4) Volume ratio of charge to retort.

(5) Time of carbonization.

(6) Pressure permitted during carbonization.

All of these conditions can be simplified into 3 effects of time, temperature and pressure by tracing them back to their origin. In the foregoing chapters, each of these factors has been treated at length, as have others indirectly derived from them.

Materials of Construction ~

It is of historic interest to note that the gas industry was born in a metallic retort, for Murdock's first efforts were confined to the use of a cylindrical iron pipe. Since gas was the primary product in the early days of the carbonization industry, progress in the art naturally led to higher and higher temperatures until the shortcoming of metallic retorts became so conspicuous that they were entirely superceded by those made of refractory materials. As regards the preference of refractory or of cast-iron retorts, Lander and McKay (186) see no basis of superiority of one above the other, all things being taken into consideration. However, two particular advantages can be associated with a metallic retort: first, its excellent heat conductivity and high thermal diffusivity, thus giving better thermal efficiency to the process and reducing the period of carbonization by virtue of the high rate of heat transfer; and second, the facility with which the system can be kept gas-tight. Its one drawback is the ease with which a metallic retort can be injured by overheating. The advent of low temperature carbonization again made their use a possibility and brought into prominence the great advantages of metallic retorts, as compared with those built of refractory. Of course, this applies only to externally heated processes and it cannot be said for internally heated processes that metal has any superiority over refractory as a material for the construction of carbonization chambers, aside from the consideration of leakage.

As a criterion by which to judge the speed of heat transfer through the retort, the thermal diffusivity is more important than the thermal conductivity in low temperature carbonization, where the periods of heating are relatively short. This is especially true in the case of intermittent processes. The truth of this fact can be made clear by the following considerations. The general case of simple linear propagation of a thermal disturbance is represented by Fourier's law of linear diffusion:

[26]     dT / dt = k ( d2T2 / dx )

where k is the diffusivity of the substance, T the temperature, t the time, and x the distance within the material. In general, this equation cannot be solved for any but the simplest cases, so that the solution for such a complex shape as a retort is quite beyond possibility for practical application. Fourier's law of diffusion is worthy of consideration, however, from the standpoint of heat transfer, because it demonstrates that the diffusivity of the material is the physical constant which is of importance during the process of heating up the retort, rather than the thermal conductivity, which is the limiting factor only after the steady state of heat transfer has been attained. Of course, the diffusivity is itself related to the thermal conductivity in the following way:

[27]     k = k / cp

where k is the thermal conductivity, c is the specific heat of the material, and p is its density. When, however, the steady state of heat transmission has been reached, then dT / dt = 0, which is to say, that the temperature gradient within the retort wall is uniform. Under this condition, Fourier's diffusion equation can be solved to give the well known equation for the quantity of heat, Q, flowing between two parallel isothermals at temperatures T1 and T2, respectively:

[28]     Q = k ( T2 - T1 ) A t / x

where A is the area considered perpendicular to the direction of heat flow, and x is the distance between the two isothermals. It will be recognized that this is the integrated form of Newton's law given in Chapter I, equation [11], under the subject of heat transfer.

We can now summarize with the statement that, in the transient thermal condition, the greater the diffusivity of the material, the faster is the propagation of temperature by conduction, and in the steady state, the greater the thermal conductivity of the material, the higher is the heat transfer. The diffusivity of cast-iron is roughly 40 times greater than that of fireclay. The thermal conductivity, and hence the diffusivity, is constant only in a limited sense. It has long been known that both of these properties are functions of the temperature, as will be shown later.

Aside from such materials as carborundum, which is occasionally employed for special purposes in the construction of carbonization plants, the refractories ordinarily used fall into 3 classes: fireclay brick, which contains not more than 75% silica; siliceous brick, which contains 80 to 92% silica, and silica brick, which contains over 92% silica. Their respective properties differ materially. The fireclay or aluminous retorts are usually made from a mixture of plastic fireclay, flint fireclay, and "bats", while siliceous retorts are composed of a ganister with clay as a binding agent. All refractories should be fired at a temperature higher than will be reached in practice, so that there will no be excessive shrinkage or expansion with use. This is a point which requires little attention in low temperature carbonization, so long as the bricks are well fired for high temperature work.

Refractory retorts may be of several types: the one-piece hand-molded, machine-molded cast retorts, and segmented retorts. The molded retorts are usually built of fireclay, while the segmented retorts may be built of shapes made of aluminous fireclay, of siliceous, or of silica material. Their thickness varies generally from 2.5 to 4 inches. They should be able to withstand abrasion, as the deposits of carbon and ash which accumulate are usually removed by scraping an scurfing the refractory. Fireclay has the disadvantage, compared to the other materials, of being especially susceptible to the corrosive action of salt in the coal.

According to Cole (259), molded retorts tend to crack after they have been in use a short time, due to the absence of joints to relieve heat strains which have been set up. On the other hand, the joints which are provided in segmented retorts permit full adjustment to the temperature conditions and prevent distortion. If the shapes are properly jointed and cemented, it cannot be said that the segmented retorts are subject to any greater leakage than molded retorts. The gas-tightness of the joint is further improved by the deposition of carbon by the cracked hydrocarbons. The segmented shapes should be designed to give maximum strength to the wall structure. Porter (260) has noted that joints in the retort wall offer a relatively high resistance to heat transfer, to such extent that considerations of thermal conductivity of different retort materials become of lesser importance when joints are present.

The requirements for a refractory cement, as established by Gill (261), are that it must be highly refractory, non-contracting, chemically inert at the working temperature, and must be capable of maintaining gas-tight joints. In addition to these qualities, it should have good adhesive and plastic properties. Generally, the more nearly alike are the brick and the cement, with respect to both their chemical and physical properties, the more satisfactory will be the result.

American silica bricks contain as much as 98% silica, which is somewhat higher than that found in European samples. Silica brick has a considerably higher thermal conductivity and diffusivity than most other refractory materials, as will be shown later, although this question has been one subject to controversy in the past. It is usual to attribute the greater average coking rate and throughput per oven in American high temperature practice to this property of silica refractories, which are so extensively used in the United States. However, Porter (260) feels that this is due more to the low degree of distortion and thermal expansion of this material, which allows ovens to be operated at a higher temperature, than to the higher rate at which it transmits heat. As far as low temperature carbonization is concerned, it will be presently seen that the high diffusivity if silica brick gives it a great advantage over most other refractories and its comparatively great strength gives an additional advantage in permitting the use of thinner walls, These advantages, however, are in part offset by the tendency of this material to spall when subjected to frequent temperature changes, especially those at low temperatures.

Cole (259) points out that when steam is introduced into vertical retorts, erosion occurs in both fireclay and quartzite or siliceous retorts, but not in those constructed of silica material. The action is similar to spalling, which is generally considered to be the cause of the failure, although some authorities attribute it to a chemical reaction.

In addition to their use in the retort proper, refractories play an important part in thermally insulating the retort setting to prevent heat losses. The heating gas, consumed in carbonization, is conserved in proportion as the losses by radiation and convection are reduced. Aside from the increased thermal efficiency gained thereby, working conditions are made more comfortable for the men employed in operating the plant. Brick if a porous non-conducting nature, or diatomaceous earth, either in the form of bricks or powder, are usually employed for purposes of insulation. The use of refractories is also extensive in the construction of regenerators, combustion chambers, and other elements of the heating system, but this subject will be treated later under the discussion of heating of retorts.

Refractory Retorts ~

Chief among the characteristics of refractories, which are used in the construction of ovens, should be immunity to injury from sudden fluctuations in temperature and freedom from volumetric changes. For general purposes, grog bricks, composed of a mixture of fireclay and grog, or silica bricks of moderate porosity and high refractoriness, which have been well burnt to prevent excessive expansion when in use, are perhaps the most suitable. Siliceous refractories have an advantage over fireclay material for combustion chamber construction, in that higher temperatures can be used. With bricks of this type, temperatures of 1350 C can be worked continuously. With few exceptions, firebricks will show signs of squatting at 1325 C and siliceous bricks at 1400 C, when loaded to 50 psi. In actual practice, however, the load on the refractory will seldom exceed 30 psi. Silica bricks can be used continuously at 1450 C or even higher. As far as internally heated low temperature carbonization processes are concerned, these are consideration of great importance, but the temperatures are so low in externally heated retorts that no particular weight need by attached to questions of refractoriness, except in certain parts of the combustion chamber which are likely to become excessively heated.

Some of the physical properties of the refractories which are commonly used in the construction of coke ovens are summarized in Table 116, after Gardner (262). Except in the instance of silica brick, it should be observed that the difference in refractoriness of the various bricks under their own weight and under a load of 50 psi is very great, approaching as much as 300 C. Consequently, the test of a material for refractoriness, unless it be made under loaded conditions, is of little value when used for the purpose of design. The high rigidity of silica is due in part at least to the purity of the rock used in its manufacture.

Searle (263) admits that it is controversial whether fireclay or silica materials are to be preferred in the construction of coke ovens. Their relative advantages and disadvantages are numerous. Unless exceedingly well burnt, fireclay refractories contract in use and are liable to crack, while silica refractories, on the contrary, tend to expand and must be very carefully heated during the initial stages of starting up, as well as during any subsequent temperature fluctuations below 600 C. On the other hand, silica materials are less likely to fail from overheating than those of fireclay, because they do not soften until very near the fusion point. In addition, the former are less subject to the corrosive action of salt and ash in the coal. In the past, there has been some question as to whether silica refractories have better thermal properties than those of fireclay. Later, it will be shown that, while at low temperatures there is in reality not a great difference, at higher temperatures the thermal conductivity and diffusivity of silica brick are considerably greater.

Gardner (262) states that aluminous, siliceous and silica refractories are all capable of withstanding the rapid fluctuations in temperature which are occasioned by the introduction fo wet coal into a hot retort. In silica materials there is a certain range of temperatures whose upper critical point is in the vicinity of 600 C where there is a great expansion for a relatively small increase in temperature. Below this critical point, which is well within the limits of low temperature carbonization by external heating, silica refrractories are liable to suffer severely from spalling. For that reason they are not recommended for use in retorts operated intermittently at temperatures below 600 C. When the temperature is maintained as high as 1000 C, however, sufficient heat is stored in the material to prevent its temperature falling to the critical contraction point when the retort is charged. In any event, the material should be well burnt and its porosity should be such as to give god mechanical strength. Silica brick, which are soft-burnt and which contain a high percentage of quartz and a poorly developed bond, spall less readily than hard-burnt bricks. When sued at high temperature, however, or when accidentally overheated, soft-burnt silica bricks expand greatly and distort the structure. For that reason, it is better to use a well burnt refractory and choose a material less subject to spalling.

Harvey and McGee (264) investigated the problem of abrasion in silica brick. Where the product of only one manufacturer was concerned, they found that the resistance to abrasion was inversely related to the porosity of the sample. However, for silica brick in general, they were able to find no particular connection between any characteristic property of the brick and its resistance to abrasion, but they concluded that it was influenced by any or all of the following factors: porosity, degree of burn, quality of ganister used as a raw material, percentage of lime used a s a bond, and workmanship.

The action of salt is commonly regarded as one of the causes of trouble with refractories. This is not of importance in low temperature carbonization when externally heated retorts are used because the sodium chloride present in coals, often to the extent of 0.5%, does not begin to volatilize until about 800 C is reached. However, in internally heated processes by the method of partial gasification, as in the Maclaurin process, where the coal passes through an incandescent zone, this may be a matter of consequence.

An explanation of salt erosion, as given by Gardner (262), is that the volatilized sodium chloride mixes with the as in the retort and is further decomposed, at least partially, into sodium carbonate and hydrochloric acid by reaction with carbon dioxide and water vapor. This reaction has already been discussed in Chapter V under the subject of ammonia decomposition. These vapors do not attack the surface of the brick, but penetrate the material to some extent and the corrosive vapors are deposited in the pores. The depth at which the reaction takes place depends upon the temperature within the refractory. The sodium carbonate fumes react with the alumina and silica to form sodium aluminum silicate. This compound has a low melting point and tends to convert the refractory to a porous and friable mass which disintegrates rapidly if the salt action is serious. In turn, the hydrochloric acid reacts with the iron in the coal ash to form ferric chloride, which also penetrates the refractory and decomposes to deposit iron oxide within the pores. The iron oxide tends to increase the fluxing action by combining with the sodium aluminum silicate. In the case of operation at low temperatures, the alkali fumes cause the refractory surfaces to become crazed after several years of work, but this action rarely interferes with the operation of the retort until decomposition begins to occur.

A large amount of finely divided ash is carried over into the combustion chamber from the gas producer which is used to furnish gas for heating the retort. This hot ash exhibits a high affinity for refractories which contain more than a small percentage of alumina. At high temperature, this causes rapid slagging, especially if the material is one of a porous nature. As far as the retort itself is concerned, this slagging action of molten ash is not of great importance in low temperature plants, because the temperatures are kept at a minimum. In certain parts of the combustion chamber, however, the refractories are exposed to injury from this source. This is also true, in particular, for those low temperature processes in which heating by partial gasification is adopted, for here the fine particles of hot ash come in direct contact with the brick lining of the carbonization shaft. The extent to which ash slagging of the refractory takes place depends greatly upon the nature and composition of the material. In this respect, silica brick seems to stand up better than other types of refractories. Apparently, the absence of alumina, iron, and alkalies in this material, together with the large size of the silica grains, renders the attack of the ash much less severe.

Quite a lot has been published on the disintegrating effect of carbon monoxide on brick containing iron oxide. This is of particular importance in low temperature processes involving partial gasification. Booze (265) has pointed out two obvious solutions of this problem: first, the use of brick free form iron oxide and second, the use of refractory that has been sufficiently hard-burnt to effect the union of iron with the slag to from silicate. It has been demonstrated, in the latter case, that disintegration does not take place, due either to the fact that the bricks are structurally stronger and able to withstand the disintegrating action or to the fact that iron in the form of a silicate is no longer able to function as a catalyst for the action of the carbon monoxide. While it has been objected that refractories, which are burnt sufficiently to effect formation of iron silicate, are so brittle that they spall readily, it has also been maintained that this is not necessarily the case and, even if they do spall, the injury is not so serious as the disintegrating effect of carbon monoxide.

Aside from aluminous, siliceous, and silica materials, there are many other refractories which are available for retort and setting construction, but their cost or other peculiarities usually eliminate them from consideration. Carborundum bricks or shapes have many special advantages, particularly in the form of high thermal diffusivity and mechanical strength, but their initial cost is many times above that of fireclay products. Magnesite bricks, although highly desirable from the standpoint of their thermal properties, have not been satisfactory in coke ovens because of their tendency to spall. The practical question of cost is also a deterrent to the use of refractories made of this and such other materials as chromite and zircon.

It is not good practice to use a single refractory throughout in setting construction, or even in the retort itself. By a prudent selection of materials, the contraction of one can be matched against the expansion of another, which with proper arrangement of expansion joints will prevent thermal distortion of the structure. It is best to segregate various section of the structure and consider the requirements for the refractory in each location with respect to temperature, load, heat transfer, expansion and contraction, abrasion, spalling, and salt and ash erosion.

Leakage in refractory retorts is far more serious in low temperature than in high temperature carbonization practice for two reason: first, because the temperature control is more important in the former and high calorific gas, permeating into the combustion chamber, will render proper temperature regulation impossible, and second, because of the relatively small gas yield in primary distillation, escaping gas represents a far greater percentage loss. The size of the pores greatly affects the permeability of gas through refractories. There is a double advantage, therefore, in selecting material with fine pores, for not only does such a refractory have the minimum gas leakage, but also the maximum structural strength.

Properties of Refractories ~

Data on the thermal conductivity of refractories, and therefore on their thermal diffusivity are very scarce, and such as are available are in great disagreement among the authorities. Wologdine (266) made some measurements which are often quoted, but they were not or a rage of specific temperatures and were made on under-burnt refractories. For these reasons, they are open to such criticism that they will not be repeated here. Not only with regard to the absolute values of thermal conductivity is there great uncertainty, but much confusion also exists as to the relative heat conductivity of fireclay, of siliceous, and of silica refractories. Summarizing the situation, Gardner (262) concluded that, at moderate temperatures, there was probably very little difference between any of them, but at high temperatures, silica was undoubtedly superior to either fireclay or siliceous bricks. He attributes the high thermal conductivity of silica to a property of the material itself, though probably assisted by radiant heat transmission through the relatively numerous and large pores of that material.

The measurements by Dudley (267), as well as those by Dougill, Hodsman, and Cobbs (268), apparently sustain the view taken by Gardner (262) that the thermal conductivity of silica is greater at high temperatures than that of other common refractories. In CGS units, Dudley reports the following conductivities at 100 C: fireclay 0.0016, quartzite 0.0020, and silica 0.0022; which increased at 1000 C to: fireclay 0.0034, quartzite 0.0034, and silica 0.0043. He also gave the mean thermal conductivity of magnesite, between 445 C and 830 C, as 0.013. Analysis of the fireclay showed it to contain 52.9% silica and 42.7% alumina. The quartzite brick analyzed 73.9% silica and 22.9% alumina. Formerly this brick was frequently used in by-product oven construction. The silica brick, which contained 95.9% silica, is extensively used in by-product oven construction. The magnesite brick was dead-burnt material, containing 86.5% magnesite and 7% ferric oxide.

Hersey and Butzler (269) examined a Georgia fireclay brick and reported a true thermal conductivity of 0.00187 at 370 C and 0.00263 at 910 C. Dougill, Hodsman and Cobb (268) measured the mean specific conductivity of a number of materials over the range extending approximately from 350 C to 1350 C. The average of their results gave 0.0034 for fireclay, 0.0025 for siliceous, 0.0036 for silica, and 0.0124 for magnesite brick. They reported the true thermal conductivity of magnesite to be 0.0194 at 250 C, 0.0138 at 500 C, 0.0107 at 750 C, and 0.0092 at 1000 C, thus showing a decrease in conductivity for this material as the temperature rises. The same authors found the true thermal conductivity for fireclay to be 0.0020 at 250 C, extending linearly to 0.0040 at 1000 C. Dougill, Hodsman, and Cobb quote some data by Heyn and Bauer, which are approximately in agreement with their measurements. These determinations at 500 C were: 0.0028 for fireclay, 0.0024 for silica, and 0.014 for magnesite; and at 1000 C they were: 0.0040 for fireclay, 0.0046 for silica, and 0.0085 for magnesite.

The disagreements of the measurements by Dudley (267), by Dougill, Hodsman, and Cobb (268), by Hern and Bauer, and by Hersey and Butzler (269) may not be looked upon as serious, considering the difficult nature of the experiments, but the results of this group of experimenters depart so much from the measurements obtained by Wologdine (266) and by Green, as hereinafter described, that all data on the thermal conductivity and diffusivity of refractories at high temperatures must be accepted with caution.

By far the most extensive data on thermal conductivity have been gathered by Green (270-273). The curves in Figure 73 have been plotted from his figures, which should be used with care since they are lower than those quoted above by other authorities. As the thermal conductivity depends greatly upon the texture of the material, a short description of each sample will be given. Sample 1 and 2 were magnesite bricks of very close texture. Sample 1 had an apparent specific gravity of 2.56 and a true sp gr of 3.38, while those of Sample 2 were 2.63 and 3.29, respectively. Sample 3 and 4 were of fireclay retort material. Sample 3 had a very close texture and evenly graded grog. Its apparent sp gr was 1.91 and its true sp gr was 2.54. Sample 4 had a very open texture and unevenly graded grog. It contained an abundance of small fissures. The apparent sp gr was 1.85 and the true sp gr 2.45. Sample 5 and Sample 6 were of fireclay brick. Sample 5 had a close structure with very few fissures and evenly graded silica grains. Its apparent sp gr was 2.03 and the true sp gr 2.46. Sample 6 had a very open texture and poorly adhering unevenly graded grog. Its apparent sp gr was 1.92 and its true sp gr 2.46.S ample 7 and Sample 8 were of silica brick. Sample 7 had a uniform close fine-grained texture, but it was quite porous and friable. The apparent sp gr was 1.51 and the true sp gr 2.20. Sample 8 had a very open texture with abundant large fissures. Its apparent sp gr was 1.77 and its true sp gr 2.31. Sample 9 and Sample 10 were from the same batch of siliceous brick. Sample 10 had seen long use in a coke oven, while Sample 9 was unused. The texture of the unused material was very close.

The thermal diffusivities for the same materials as illustrated in Figure 73 are shown in Figure 74, from the measurements by Green (270, 272). The CGS unit used to measure thermal diffusivity is the rise in temperature in one cubic centimeter of material by one calorie in one second through one square centimeter of a layer one centimeter thick by a temperature difference of one degree Centigrade. It will be observed that the relative thermal diffusivities of the various materials have only slight similarity to their relative conductivities.

The silica brick used by Green (270) was of very poor quality, when judged from the standpoint of American practice, Silica bricks used in the US contain a larger percentage of silica, are denser, and not nearly so porous. This undoubtedly accounts for the unusually poor thermal values given for silica in Figure 73 and Figure 74. That the thermal properties of refractories depend greatly upon the temperature at which they were fired has been observed by Wologdine (266) and has been confirmed by others. This is especially true of silica brick, which has a remarkably low thermal conductivity when poorly fired. Green (271) concluded, regarding the relative thermal qualities of fireclay and silica products, that superior well-fired silica bricks are better heat conductors than those of fireclay below 800 C and that most silica bricks, except those that have been ineffectively fired, are better conductors than fireclay at high temperatures. However, many silica bricks and firebricks have equal conductivities at low temperatures. Even though magnesite has a high conductivity, its high specific heat and high density reduce its diffusivity and, in fact, make it notably less than that of other materials at high temperatures. The thermal conductivity of used siliceous brick is form 10% higher at 500 C to 22% higher at 1300 C than that of most unused material. Although little change was observed in the specific gravity, Green (272) attributes this increase wit use to alterations in the texture of the material by changes in its constitution and porosity. The thermal conductivity of most all refractories, except that of magnesite, which is a particularly dense material, rises rapidly above 1000 C, probably because the transmission of heat by radiation across the pores of the material becomes sensibly appreciable, as compared with the heat transmission through the material by pure conduction.

The mean specific heat of refractories is a function of the temperature, as may be seen in Figure 75. The curves have been plotted from data determined principally by Wilson, Holdcroft, and Mellor (274), by Bradshaw and Emery (276), and by Dudley (267), supplemented by the data collected by Wilkes (15). It is interesting to note the small difference between the mean specific heats of silica and fireclay refractories. In all of the common refractory materials, the mean specific heat increases practically linearly up to 1000 C, after which there is an apparent falling off. Very few determinations have been published on the specific heat of carborundum, but Wilkes (15) reports a mean specific heat of 0.201 at 100 C and 0.187 at 1000 C for this material, which seems to indicate that its heat capacity decreases with the temperature, as contrasted with the increase observed in other materials.

Heretofore, the author has made frequent reference to the spalling of refractories, particularly of silica brick, without any consideration being given to the cause of this phenomenon. According to Searle (263), spalling is the tendency of refractories to flake caused by the appearance of numerous fine cracks when the material is subjected to frequent sudden heating. It naturally results in a reduction of strength of the material and finally is responsible for its complete failure by disintegration. The cause of these cracks is attributed to the sudden transformation of quartz into tridymite or cristobalite, due to its incomplete conversion during burning. This situation is particularly liable to occur when a flame plays directly on the refractory. It can be avoided by carefully heating the material gradually so that the conversion of the quartz to its other allotropic form may occur slowly. Repeated heating of silica brick to 800 C at a slow rate will reduce the tendency to the material to spall. Tridymite bricks are well known to spall less than those of quartz or of cristobalite, and this procedure favors the formation of tridymite. The average loss by spalling of a brick is about 30%, but it varies directly with the fineness of the material. Consequently, a coarse-grained brick should be used where it is desirable to reduce spalling to a minimum, unless other properties possessed by such a material make its selection unwise. If properly manufactured, machine-made bricks spall no more than the hand-made ones.

An example of the effect of heating below 700 C on the spalling of silica brick has been furnished by Ross (276), who found that the specimen invariably spalled when heated at the rate of 270 C in 20 minutes, 520 C in 40 minutes, and 690 C in 60 minutes. When, however, the rate of heating was reduced to 50 C in 15 minutes, up to 500 C, and 100 C in 15 minutes, above that temperature, spalling seldom occurred. He concluded, however, that the changes in the brick structure which caused spalling took place below 500 C.

The composition of all refractories consists largely of a mass of crystals which may have several allotropic forms, each of which is stable only over a limited range of temperature. A large change in volume usually accompanies the inversion of one allotropic form into another and these changes are sharply defined in the expansion curve for the material. In this respect, silica is the chief constituent of refractories to be reckoned with. The curves in Figure 76 give the linear thermal expansion of some common refractories, as determined by Norton (277). These data agree with the less precise measurements of Bogitch (278) for magnesite up to 1400 C and for chromite up to 900 C.

The silica sample was cut from an ordinary commercial brick containing about 97% silica. Its thermal expansion rose rapidly to 1.09% at 260 C, at which volume it remained constant until 620 C was reached. The length of the sample then suddenly increased to 1.29%. Since only about 5% tridymite was present, there is no indication of inversion of that allotropic form of silica at 110 C. At 260 C, the conversion of a-crystobalite to B-cristobalite, together present to the extent of about 70%, is clearly defined. Quartz was present in an amount approaching 25% and its inversion at 610 C is very sharp. Beyond 620 C, the brick contracted uniformly of about 1400 C, and then expanded rapidly to a maximum at 1525 C, due to the conversion of quartz to cristobalite. At about 1525 C, the brick again began to shrink and started to soften at 1700 C.

The composite curve for kaolin is plotted from 3 samples fired at different temperatures, ranging from 1430 C to 1620 C, and represents approximately the characteristic of a sample burnt at 1500 C. The expansion of a kaolin brick, fired at 1430 C, is about 21% above the composite curve and that for a sample fired at 1620 C is approximately 21% less than the composite curve. Kaolin brick expands more or less uniformly up to about 1400 C, beyond which a permanent shrinkage takes place. In general, it may be said that permanent contraction begins to appear just below the temperature at which the brick was burnt. There are slight irregularities in the expansion curve for kaolin, not showing the composite curve, but these are unimportant and occur principally from inversions of small amount of silica present as a bond or as an impurity.

The composite curve for fireclay is constructed from 4 samples of brick from Missouri, Pennsylvania, and Maryland. The specimen from Missouri contained 53.1% silica, that from Pennsylvania 54.2%, that from Colorado 62.6%, and that from Maryland 62.3%. Each individual sample departed more or less from the composite curve to the extent of perhaps 20%, at points, due to silica inversions. On the whole, however, the increase in linear expansion for fireclay can be considered approximately uniform up to about 1000 C, beyond which there was a contraction in every case, except that of the Pennsylvania sample, which began to increase rapidly in length at 1250 C, finally reaching 4.8% expansion at 1600 C. The Maryland brick began to contract rapidly at 1000 C, and continued to do so until it fell to a total expansion of only about 0.11% a 1400 C, beyond which it again increased in a manner similar to the specimen from Pennsylvania, reaching finally 2.2% at 1600 C.

The expansion of carborundum was almost linear, was less than that of the other refractories. And showed no sign of contraction up to 1700 C. The zirconia brick contained 27.3% silica, which accounts for the fluctuations of its expansion curve. It will be observed that these variations are closely related to those in the curve for silica brick. The general trend in expansion of the magnesite brick was a uniform increase up to 1440 C, after which an irreversible shrinkage occurred, due probably to inversion of periclase. Except for small irregularities, due no doubt to a small amount of free silica, the chromite brick expanded uniformly to about 900 C, beyond which there was a tremendous change in size up to 1000 C. Thereafter there was a slow contraction up to 1500 C, after which a rapid contraction developed. The similarity of the diatomaceous earth insulating brick expansion curve with that of silica shows the former to be composed largely of that material. The insulating brick shrank rapidly beyond 1100 C.

Silica bricks expands as much as 20% of their original size when fired and this is accompanied by a reduction in specific gravity from about 2.60 to about 2.30. In fact, the specific gravity furnishes a good indication of the extent to which the brick has been burnt. This expansion is due to a conversion of quartz, whose specific gravity is 2.65, to the allotropic modifications of crystobalite and tridymite with specific gravities of 2.33 and 2.27, respectively. However, since the conversion is never complete, even though the brick be twice burnt, there is always a further permanent expansion which takes place during use and must be allowed for by providing expansion joints or openings at intervals in the brickwork. In practice, approximately 0.25 inches per linear feet is allowed for this purpose.

Norton (279) considers the tendency of a brick to spall as being directly proportional to its coefficient of linear expansion, other things being equal. Since this coefficient varies with the temperature, it is necessary to pick out the temperature at which spalling occurs. This has been determined as between 300 C and 700 C. In the light of this theory, the curves in Figure 76 fully account for the observation that silica and chromite spall very easily.

The crushing strength of a refractory depends upon the amount of cementing material which has been fused during burning to bind the particles of the material together. The strength, consequently, is contingent upon the temperature and duration of the burn. When the refractory is subsequently heated intensely, a part of this cementing material fuses and lessens the strength of the brick, so that it deforms easily under load. We have already seen in Table 116 an illustration of the way the squatting temperature is reduced under load. Another example ha been reported by Mellor (280), who examined a number of British fireclays. In a number of samples, whose unloaded melting point average 1640 C, with a range extending from 1580 C to 1690 C, the average melting point under a load of 54 psi was 1415 C, with a range of 1380 C to 1435 C. When the load was increased to 72 psi, the average fusion point was 1395 C, with a range of 1350 C to 1395 C. Mellor has suggested that the relationship between the bending temperature, T, and the pressure applied may be represented by the formula:

[29]     T = T0 E -CW

where T0 is the bending temperature without load; W is the pressure in psi, and C is a numerical constant depending upon the clay, mode of manufacture, etc.

There is much discrepancy between the crushing strengths of different samples of the same refractory, tested at various temperatures, as will be observed from Table 117. The data in this table are all taken from the measurements by Bodin (281), except those for the third sample of silica brick, which are by Le Chatelier (282). The variation between the respective samples of a given material arise from such factors of manufacture as the kind and quantity of bond used, firing temperature of the kiln, duration of burn, pressure used in molding, etc. Thus Sample 1 of bauxite was fired at 1500 C and Sample 2 at 1300 C. The figures of Bodin are probably somewhat high for commercial bricks, as they were made on small test specimens. The crushing strength of common refractories decreases with rise in temperature up to about 800 C. Strange as it may seem, Bodin (281) found that thereafter the strength increases rapidly until approximately 1000 C is reached, due probably to a transition in the structure of the material. Thereafter, the crushing strength falls rapidly with rise in temperature. Magnesite and chromite refractories were exceptional in that no increase in strength was observed between 800 C and 1000 C.

Metallic Retorts ~

In the past, cast-iron retorts were used extensively in gasworks, as we have seen, but they were later supplanted by refractory retorts for a number of reasons. For high or moderate temperature carbonization, the refractory retorts had the advantage of being cheaper, more durable, and more refractory, which is to say, that they were capable of operating without injury at a higher temperature. On the other hand, for such operating temperatures, clay retorts have one serious disadvantage, as compared to those of cast-iron; that is, they do not stand cooling down, invariably contracting and subsequently cracking. But neither is cast-iron perfect in this respect, as we shall see later, for it is subject to growth upon repeated heatings. As far as low temperature carbonization is concerned, there are two big advantages to metallic retorts, such that, if other difficulties can be surmounted, they are vastly superior to refractories. As already pointed out, these advantages are principally, first, the highly desirable thermal properties of metals, and second, the absence of gas leakage. Of somewhat minor importance is the lesser retort breakage and the adaptability of metal retorts to the use of moving parts which may be used to effect the transportation and discharge of the material that is being carbonized.

Cast-iron retorts have certain advantages over other types of metallic retorts. In the first place, it is the cheapest metal; second, it is almost as strong as ordinary cast steel, within the temperature range of low temperature carbonization; third, it has excellent corrosion-resisting qualities, and fourth, it lends itself to cheap quantity production. However, the designer of a cast-iron retort is confronted with a number of difficult problems. In addition to the usual expansion, which can be provided for when designing the retort setting, and the problem of warping, which can be taken care of by strengthening the retort with carefully placed ribbing and proper design of the setting to prevent local overheating, there is the question of the growth in cast-iron after repeated heatings, especially when superheated steam is present.

Carpenter (283) has demonstrated, as we shall see later, that cast-iron growth is greatest in casting of high silicon content and least in castings of low silicon. He attributes this to the fact that the silicon is present in the iron as iron silicide, which reacts with the graphite present to form oxidizing gases, which in turn attack the iron chemically. Without the presence of graphite the silicon has no effect. Carpenter (284) fully recognized that, while there is little difficulty in getting suitable cast-iron to stand temperatures up to 625 C, it was another matter, and a very difficult one at that, the get a casting which would withstand use at 650 C. he suggested that the use of chromium as an alloy, in preference to manganese, was good practice in many such cases.

The Fuel Research Board has had a rather successful experience in the use of mild steel retorts in their horizontal setting. Mild steel does not suffer from growth like cast-iron, but it is easily overheated to produce softening slightly above 600 C. Moreover, mild steel has a low elastic limit, within the temperature range used for primary distillation, and suffers a creep, after a time, at loads below the elastic limit. This creep has been particularly noted in low temperature rotary retorts which have been supported on trunnions only at the ends of the heated cylinder. In such cases, there has been a creep, sometimes mistaken for growth similar to cast-iron, of sufficient magnitude as to cause the retort to become inoperative. This difficulty has been ingeniously overcome in the K.S.G. retort by supporting the entire load from an inner cooler cylinder, as has been pointed out already in Chapter VI.

Efforts have been made to discover ferrous and other alloys which would withstand usage at high temperatures and these investigations have in part been successful. There are a number of metals of various compositions which have the desired properties at elevated temperatures for primary carbonization retorts. Unfortunately, however, these alloys are all so expensive, at the present time, as to render their use entirely prohibitive from a commercial standpoint.

Properties of Cast-Iron ~

Principal among the properties of cast-iron is its tensile strength and the manner in which this characteristic varies with the temperature. One of the features to be observed in this metal is the fact that the change in tensile strength of cast iron up to 850 F is comparatively small. There are some indications of a maximum between 650 F and 900 F. Above the latter temperature the metal softens rapidly and the tensile strength accordingly drops sharply.

The curves in Figure 77 give the tensile strength of cast-irons, as a function of temperature, for a variety of different alloys from measurements by a number of authorities. Specimen 1, after Schwartz (285), was a malleable cast-iron containing from 2.8% to 3.5% carbon and 1.1% to 2% silicon. Specimen 2 was also a malleable cast-iron from another source (286) and of unknown composition. Specimen 3 was a semi-steel of unknown composition (286). Specimen 4 was a semi-steel, as measured by Harper and MacPherran (287). It contained 1.84% silicon, 0.106% sulfur, 0.52% phosphorus, and 0.64% manganese. Specimen 5 was a semi-steel which contained 2.8% total carbon, of which 1.6% to 2% was graphite, and 1.1% to 1.4% silicon. Specimen 6, also a semi-steel, contained 3.1% total carbon, of which 2.1% to 2.4% was graphite, and 1.3% to 2% silicon. Specimen 7 was a cast-iron which contained 3% to 3.3% total carbon, of which 2.7% was graphite, and 1.1% to 2% silicon. Specimens 5a, 6a, and 7a were all annealed, while Specimens 5b, 6b, and 7b were all as cast. Specimens 8 and 9 are for cast-irons form the results obtained by Smalley (289). Specimen 8 analyzed 3.3% total carbon, of which 2.8% was graphite, and 2.2% to 2.7% silicon, while Specimen 9 showed 3.7% total carbon, with 3.5% graphite, and 2.9% silicon. From these curves, the difficulty of working cast-iron, or indeed any other metal, continuously over a temperature range from 1000 F to 1200 F may be easily appreciated. But we shall se later that the whole trouble does not stop here, in the case of cast-iron, which has the unfortunate habit of growth with time and temperature.

Kennedy and Oswald (290) studied the effects of various percentages of silicon, nickel, and chromium on the strength and other properties of cast-iron. With about 1.4% silicon and no chromium present, the tensile strength reached a maximum increase of 52%, over the base metal, with the addition of 3.88% nickel, 45% of which increase was obtained with 1.23% nickel. With about 1.4% silicon and 0.51% chromium the highest value of 57% increase in tensile strength was obtained with the addition of 1.1% nickel. With higher silicon content, the effect of nickel in increasing the strength of cast-iron was not so noticeable. Silicon apparently destroys the effectiveness of nickel in strengthening iron, unless means are taken to counteract it, as by the addition of chromium. Thus, with the presence of about 2.6% silicon and increasing amounts of nickel and chromium, the maximum increase in tensile strength of about 35%, was obtained with the presence of 0.85% nickel and 0.59% chromium.

The specific gravity (291) of wrought iron varies from 7.8 to 7.9, that of white cast-iron extends from 7.58 to 7.73, while gray cast-iron ranges for 7.03 to 7.13 at ordinary temperatures. The mean specific heat of cast-iron, from 20 C to 100 C, is 0.1189 according to Schmitz (292), but Honda, as quoted by Hatfield, Wollman, and Priest (15), gives as the mean specific heat from 0 C to 100 C the value of 0.131 to 0.142 for the material as cast and 0.116 to 1.139 when the material was annealed. Nichol (293) made some measurements on the mean specific heat of wrought iron over several short temperature ranges. From 15 C to 100 C he found it to be 0.1152; at 500 C he gave it the value 0.176, and from 1000 C to 1200 C he measured the mean specific heat of wrought iron as 0.1989. The mean specific heat of a rather pure sample of iron from 0 C to various temperatures has been accurately determined by Harker (294). His data have been plotted in Figure 75, along with the mean specific heats of refractories. It will be observed from the illustration, that the results are not greatly different from those for zirconia brick. Schwartz (285) has calculated the true specific heat of malleable cast-iron and found it to be approximately 0.108 at 0 C; 0.135 at 250 C, and 0.180 at 500 C.

Very little information is available on the thermal conductivity of cast-iron. Kaye and Laby (295) gave 0.149 as the mean conductivity between 0 C and 100 C. Some data by Jaeger and Diesselhorst (291) give the thermal conductivity of wrought iron as 0.144 at 18 C and 0.143 at 100 C, with a decrease of 0.00008 for each degree Centigrade increase in temperature. It can be deduced, however, that the thermal conductivity decreases with temperature, for the case of electrical conductivity in metals and the two phenomena are closely related. Without an adequate knowledge of the thermal conductivity at high temperatures, it is impossible to calculate the thermal diffusivity, except for ordinary temperatures. Ingersoll and Zobel (296) report the thermal diffusivity of cast-iron at ordinary temperatures as 0.121 and of wrought iron as 0.173, which data are sufficient to indicate that, in order of magnitude, the thermal conductivity and diffusivity are several hundred fold greater than those found in refractories.

Schwartz (285) has measured the thermal expansion of malleable cast-iron up to 1100 F, as shown in Figure 80. This is the same sample of material whose tensile strength was reported as Sample 1 in Figure 77. His results are in agreement with the measurements of Souder and Hidnert (297), who carried their temperatures up to 1650 F, where vast changes in the structure of the material took place. The sample tested by them contained 3.08% carbon and 1.68% silicon. Reference to Figure 80 shows the expansion and contraction which accompanied the heating of this sample of cast-iron. The linear growth upon cooling amounted to 0.92% for one heating.

The growth of some cast-irons may be tremendous. Thus, Regan and Carpenter (298) observed a volume increase of 63% after 32 heatings to 850 C in a cast-iron containing 3.38% carbon, 6.14% silicon, 0.30% manganese, and negligible quantities of sulfur and phosphorus. This was, of course, exceptional, but volume increases of half this amount are not extraordinary. Since the volume changes more rapidly than the weight during growth, the specific gravity of cast-iron decreases with repeated heating and with it there is a decided change in the physical properties of the material. Its mechanical strength and thermal characteristics suffer accordingly. The action can proceed so far that a failure in tensile strength may occur, or even advance to a stage where the casting can be crumpled between the fingers.

Outerbridge (299, 300) conducted the first systematic examination of cast-iron growth. He observed that white cast-irons do not grow to the same extent as gray cast-irons and established that the breakup of the carbides through the separation of graphitic carbon was insufficient, in itself, to account for the phenomenon. Rugan and Carpenter (298), however, were the first to demonstrate that the volumetric changes were accompanied by a gravimetric increase, form which they were able to advance a theory for the action which has been accepted rather generally. They laid particular stress on the influence of silicon, to which exception has been taken by others. While it cannot be denied that the amount of silicon present is of great import, its influence is apparently of an indirect nature, in that it favors the separation of large graphite flakes which make the alloy ductile and assist the premature appearance of cracks.

The effect of temperature, in accelerating the linear growth of cast-iron, is well illustrated in Figure 78, which is plotted from measurements by Schwinning and Flossner (301). The sample in question was composed of 3.49% carbon, 2.56% silicon, 0.46% manganese, 0.735% phosphorus, and 0.135% sulfur. The growth began slowly at about 450 C and was strong at 500 C. Microscopic examination revealed that the first period of slow growth up to 18 heatings, of 3-hour duration, was accompanied by a slow change of pearlitic structure into a mixture of pearlite and cementite. No decomposition of the carbide was noted until the second stage of growth, after the 18th heating, where the cementite was destroyed with the release of graphitic carbon, which sets in veins and grew in thickness until all the material was ferrite. At 650 C the early stages of growth were entirely eliminated and destruction of the carbide set in at once.

The usefulness of cast-iron in the construction of low temperature retorts, or its appurtenances, has been restricted by the tendency of this material to undergo permanent volume changes, when the castings are repeatedly heated or used continuously in the vicinity of 650 C or even lower temperatures. The phenomenon depends not only upon the factors of time and temperature, but upon the composition. In general, white cast-irons and gray cast-irons grow, but Rugan and Carpenter (298) observed that white cast-irons, which contain more than 3% carbon, tend to deposit temper carbon upon prolonged heating, which in turn contributes to the metal's expansion. They found that white cast-iron, which contained about 3% carbon and negligible amounts of the other constituents, and especially one in which the silicon content did not exceed 0.3%, remained practically constant in volume when heated many times to 900 C.

The effect of the various elements on the growth of cast-iron was studied also by Rugan and Carpenter (298) and later by Carpenter (302). They found the phosphorus tends to reduce growth, to the extent that the presence of 0.3% phosphorus lessens the growth about 3%. Manganese retards the rate of growth and in most cases diminishes the absolute amount. The small quantity of sulfur found in commercial cast-iron has only a minor effect and this contributes to the action's retardation. The presence of dissolved gas may cause an expansion of 1% to 2% when the silicon content is 1.75% to 3%, but with the silicon present exceeding the higher figure, occluded gas has no influence. When the silicon amounts to less than 1%, dissolved gases may cause a growth to the extent of as much as 10%. Broadly speaking, the percentage volume growth is proportional to the percentage of silicon present, which according to Carpenter (284) amounts to a 15% volume increase with 1% silicon, 27% with 2% silicon, and 37% with 3.5% silicon.

Rugan and Carpenter (298), as a result of their investigations, recommended the use of a special semi-steel for high temperature work. Its composition consists of 2.66% total carbon, all of which should be combined, 0.59% silicon, and 1.64% manganese, with 0.01% sulfur and phosphorus. Such a casting showed no growth, but on the contrary, a slight contraction of 0.13% after 150 heatings to 850 C. This sample had a tensile strength of 50,200 psi initially, which increased to 55,300 psi after conclusion of the heatings. Analysis showed 2.37% combined carbon and 0.25% temper carbon at the end of the experiment, representing a conversion of only 9.6%.

The mechanism of cast-iron growth is only partially understood. It is generally agreed that the phenomenon is due in part to the deposition of carbon as a result of the breaking down of iron carbide, and in part to oxidation of the elements present. While occlusion of gasses in the metal, together with casting stresses, may also be factors contributory to growth, opinions differ widely among the authorities regarding the relative importance of each, and, in fact, even in regard to the sequence of events leading up to expansion. It is generally accepted, however, that the presence of silicon is highly undesirable and that the rate at which the metal deteriorates is roughly proportional to the amount of silicon present.

Carpenter (283) definitely concluded that silicon was chiefly responsible for swelling in the gray cast-irons examined by him, whether the cause of growth be superheated steam above 150 C, hot furnace gases, hot air, or repeated and prolonged heating at high or low temperatures. According to his view, silicon is preset as dissolved iron silicide, but this is absolutely innocuous unless graphite is present, by thermal decomposition of cementite or otherwise, so that the breaking down of iron silicide into silicon and iron oxide permits the penetration of oxidizing gases, which attack the metal chemically. The solution of the problem naturally resides in using an alloy which contains no free carbon and which further does not deposit it upon prolonged or repeated heating. Consequently, Carpenter (283) concluded that no gray cast-iron and only a few white cast-irons were suitable for use under such conditions. The choice of a suitable material thereby leads to a consideration of the many varieties of mild, medium, and hard steels, all of which have no free carbon, do not deposit it upon prolonged or repeated heating, and which are low in silicon content. The presence of manganese retards the deposition of temper carbon and toughens the alloy.

On the other and, Oberhofer and Piwowarsky (303), as well as Honegger (304), conclude that the mere presence of silicon does not necessarily cause rapid growth in cast-iron, but that everything depends upon the size and arrangement of the graphite plates. Even if relatively low in silicon, an iron tends to grow rapidly, if during solidification it is cooled at such a rate as will cause the production of coarse graphite plates.

Pearson (305) examined the growth of cast-irons of low silicon content, including among other two samples made by the hot-mold process. Repeated heating to 900 C for one-hour periods showed a constant increase in length at the end of about 40 heatings, the volume increase being about 5.47%, as compared with 16% to 37% observed in ordinary commercial gray cast-irons. The sample in question contained 3.33% carbon and 0.68% silicon. Similar results for hot-molded low silicon cast-iron were obtained by Donaldson (306), except that his tests were carried only up to 550 C. In many instances, where a white cast-iron could be successfully used as a heat-resisting metal, the difficulty of producing the castings without internal strains, liable to cause cracks, is a practical foundry consideration which renders its use impractical. The hot-mold process, however, deserves consideration on this score in that it gives a means, hitherto unavailable, of producing a machinable gray cast-iron from pigs of such low silicon content as would yield a white or mottled iron under ordinary conditions.

Andrew and Hyman (307) made a comprehensive study of the effect of various alloys on cast-iron growth, using a base material containing 3% to 4% carbon, 0.4% to 1.6% silicon, and 0.5% to 1.8% manganese. After 50 heatings to 900 C, the sample containing nickel had increased in length about 9.5%, the samples containing chromium about 5% and 3.5% respectively. It appears, therefore, that aluminum and nickel have an action similar to silicon in favoring the deposition of temper carbon, which is detrimental to the casting from the standpoint of growth. Kennedy and Oswald (290) investigated the effect of other alloys and confirmed the conclusion of Rugan and Carpenter (298) that high phosphorus has a marked influence in slowing the rate of growth. They also observed that an iron deoxidized with titanium grew much more slowly than one of the same composition with that element absent.

Properties of Steel ~

The tensile properties of steel depend greatly upon its composition and quite as much upon the treatment which it receives during manufacture. The variations in composition are so infinite and the special treatments by rolling, forging, annealing, and quenching are so numberless that it is quite beyond the scope of this book to undertake a detailed discussion of this subject. It can be said, in general, regarding the tensile strength of carbon steels, that for temperatures below 850 F the cast sample has the lowest value, the cast and annealed specimen is somewhat stronger, the rolled sample has even a higher value, and finally, the rolled and heat-treated specimen has the highest tensile strength of all. However, above 850 F and extending to beyond 1400 F, a statistical examination of the determination by many authorities has convinced the present author that in this temperature region there is no particular advantage to either heat-treatment or rolling in the average case, as will be observed from Figure 79. This statement does not apply to the case of mild steels, with 0.15% and less of carbon, for the results of this study indicate that cast mild steel possibly has a stronger tensile strength between 800 F and beyond 1400 F than the rolled variety, so far as conclusions may be drawn with limited data. Neither does the statement apply to high carbon steels with carbon contents exceeding 0.9%, in which case the difference between the tensile strengths of the rolled an cast samples is quite indistinct over the entire range of temperatures. This generalized conclusion is a statistical statement and hence great departures are found in individual samples.

The data in Figure 79 are composite curves which represent the average result for carbon steels containing 0.4% to 0.8% manganese, 0.1% to 0.6% silicon, 0.005% to 0.04% phosphorus, and 0.02% to 0.005% sulfur. The ultimate short-time tensile strength, as a function of the temperature, is given for each of 3 cast steels, of carious carbon contents, and also for 3 rolled steels. The measurements for alloy steels are for individual samples. Sample 1 represents a cast steel containing 0.15% carbon. It is plotted principally from the results obtained by Dupuy (308). Sample 2 represents a cast steel containing 0.30% carbon. It is the composite result of measurements by Malcolm (309) and by Dupuy (308), together with another sample of unknown composition (286). Sample 3 represents a cast steel with 0.55% carbon. It is the composite curve from data gathered by Dupuy (308), by Tapsell and Clenshaw (310), and by Malcolm, as he is quoted by French and Tucker (311). Sample 4 is the characteristic curve of a rolled steel containing 0.15% carbon. It is the composite result of measurements by French and Tucker (311, 312), by White and Clark (313), and by Dupuy (308). Sample 5 represents a rolled steel with 0.35% carbon. It is the composite result of determinations by Fahrenwald (314), by Lynch, Mochel, and McVetty (315), by French and Tucker (311), by French (316), by Perrine and Spencer (317), and by Dupuy (308). Sample 3 represents the characteristic of a rolled steel which contains 0.6% carbon. It is plotted from the results obtained by Tapsell and Clenshaw (310), by French and Tucker (311), by Dupuy (308), and by Dickenson (318).

As compared with the cast-irons, semi-steels, and malleable irons illustrated in Figure 77, it can be said that the maximum allowable tensile stresses above 900 F of those materials do not differ much from the maximum allowable tensile stresses of the carbon steels illustrated in Figure 79, considering the tendency of the latter to creep. It has already been pointed out that the main advantage of steel over cast-iron in the construction of low temperature retorts is the absence of growth in the former, but on the other hand, steel tends to creep under load, at temperatures above 500 F, and this phenomenon generally requires the use of a much lower stress than the ultimate strength of the material would indicate. As a matter of fact, according to Mellanby and Kerr (319), at 1000 F the creep limit of 0.35% carbon steel is only 20% of its ultimate tensile strength, whereas with a chrome-nickel steel, the creep limit was approximately 45% of the ultimate strength of that temperature. It can be said, in general, that the higher the temperature, the smaller is the creep limit when computed as a percentage of the tensile strength.

The tensile strength of alloy steels have been pretty well investigated at elevated temperatures, though the measurements have rarely been carried higher than 1100 F. Determinations of the tensile strength have been made by MacPherran (320), who investigated a great variety of alloys, by Bregowsky and Spring (321), by Lynch, Mochelm and McVetty (315), by French and Tucker (312), and by Malcolm and Welter, as they are quoted by French and Tucker (311). Mockel (322) also examined 15 samples of high chromium stainless steels. The tensile strengths of some of these specimens of ferrous alloys are included in Figure 79 for purpose of comparison with carbon steels.

The properties of alloy steels doubtlessly render them of great value for high temperature use, but their cost is practically prohibitive for the construction of low temperature retorts. Like carbon steels, alloy steels are also subject to creep. Malcolm (323) feels perfectly assured that cast chrome-nickel alloy steel will stand severe service without deformation over long periods, provided the following maximum allowable stresses are not exceeded: 26,000 psi at 600 F; 18,000 psi at 850 F, and 6,000 psi at 1100 F. For ordinary cast carbon steel he recommends the use of only about 60% of these allowable stresses.

The thermal expansion of a manganese steel, containing 0.49% carbon, 1.21% manganese, and 0.12% silicon, as determined by Souder and Hidnert (297), is given in Figure 80, along with the thermal expansion of cast-iron, which has already been discussed. As a matter of fact, there is very little difference between the thermal expansion of steels of different composition. The effect of alloys is noted mostly during contraction, particularly in the temperature region at which the anomalous changes occur and in the permanent shrinkage observed on cooling. Slightly above 1200 F nearly all ferrous alloys undergo a slight shrinkage during further increase of temperature over a region extending 50 F to 200 F. Thereafter, uniform expansion with temperature increase is observed, but the rate of increase in expansion is always greater thereafter, than before the period of contraction. Upon cooling the steel, the reverse phenomenon takes place, but usually the sample never completely regains its length and a permanent contraction results. During the period of anomalous expansion all the physical properties of the material are affected, including its thermal as well as its elastic characteristics.

Like cast-iron, very little is known regarding the thermal properties of steel at high temperatures. The mean specific heat for a sample of steel containing 0.30% carbon, after Fahrenwald (314), is plotted in Figure 75 up to 500 C. According to these data, the mean specific heat of steel is slightly higher than that of pure iron and has a somewhat greater rate of increase. The mean specific heat depends a great deal upon the composition and heat-treatment of the sample, as does the mean thermal conductivity. Thus, the mean specific heat form 0 C to 100 C has been reported (315) as 0.123 to 0.113 for carbon steels, 0.129 to 0.115 for chromium steels, and 0.124 for a nickel steel. In general, the lower figures apply to quenched samples and the higher figure to annealed samples, the same source gives the mean thermal conductivity from 0 C to 100 C as 0.143 to 0.044 for carbon steels, 0.098 to 0.031 for chromium steels, and 0.060 to 0.039 for chrome-nickel steels. Some data by Jaeger and Diesselhorst (291) give the thermal conductivity for 1% carbon steel as 0.108 at 18 C and 0.107 at 100 C, with a decrease of 0.0001 for every degree Celsius rise in temperature. Few measurements likewise have been made on the specific gravity of steels at high temperatures. At ordinary temperatures, this constant ranges form 7.60 to 7.80. With such meager data on the thermal properties of steel, it is not surprising to find a dearth of information on its diffusivity. Ingersoll and Zobell (296) give the thermal diffusivity of mild steel as 0.173 and of 1% carbon steel as 0.121.

Heating of Retorts ~

The setting for cast-iron retorts should be as simple as possible. No great heat control is needed below 650 C and the simpler flues give equally as good results as the more expensive constructions. There are but two criteria to be observed; first, that the air is best admitted separately for the burners to avoid highly localized temperatures, as is often obtained with the Bunsen type of burner, and second, the burners should be placed in such a way that the metal retort will not be touched directly by the heating flame. By observing these details, the flue will be heated by a long lazy flame and the retort will be heated by a combination of radiation from the setting walls and convection from the combusted gases, thus giving the retort a uniform temperature which tends to minimize warping of the casting. The Fuel Research Board carefully observed these two principles in both their horizontal and vertical ovens and therein lies a large part of the secret of their success with metal retorts.

When intense local heating of flat metallic surfaces occurs repeatedly, buckling or cracking is certain to take place. Fahrenwald (314) attributes this to cumulative plastic deformation of the material, caused by excessive alternating temperature stresses. This phenomenon is quite distinct from bulging or sagging, with which it is often confused. The latter arise from the failure of physical strength at elevated temperatures. The plastic deformations, which result in buckling, are due to thermal expansion stresses which exceed the elastic limit of the material and which arise from non-uniform heating. More failures in metallic retorts may be attributed to unequal temperature distribution than to all other design influences, while the necessary cooling down of the retort periodically is the operating factor which limits its serviceable life more than any other cause, aside from overheating.

Except for certain heat losses which occur in the retort setting, practically all the heat imparted to the coal goes to raise the charge to the desired temperature of carbonization and is later discharged from the retort as sensible heat of the coke, as sensible heat of the volatile products, and as latent heat of the condensable vapors.

While there are periods of endothermic reactions, there are also periods of exothermic reaction, and the two tend to counteract each other. So, from a thermal standpoint, the reactions occurring in low temperature carbonization can, in the aggregate, be considered as slightly exothermic. The major proportion of the heat which is imparted goes to raise the temperature of the charge to a point where the desired transformations are effected, and barring heat losses, it is an irreducible quantity, but the other, or minor part, which raises the temperature of the exit vapors, varies with the efficiency of the retort design.

From the standpoint of thermal economy, the use of an insulating material to prevent heat losses is highly important, for by reducing the surface losses by radiation and convection s much as possible, the amount of heating gas required for carbonization can be reduced accordingly. Cole (259) recommends diatomaceous earth to be used for this purpose. With a furnace temperature of 1800 F, the heat loss through a wall consisting of 9 inches of firebrick and 8.5 inches of redbrick is of the order of 20 BTU/sq ft/hr, with an outer surface temperature of 370 F, whereas a wall of the same total thickness, consisting of 9 inches of firebrick, 4 inches of redbrick, and 4.5 inches of insulation, has a heat loss of about 190 BTU/sq ft/hr, with a surface temperature of roughly 190 F.

After the heat has been transmitted by conduction through the retort wall it must be transmitted into the interior of the charge. The problem is a simple one as far as the first layers of the charge are concerned, but thereafter, when the coal has been converted to porous semi-coke, the problem is of quite a complicated nature. Practically no data are available on the thermal conductivity and diffusivity of coke at the temperature at which the process takes place. Even if these data were at hand, the difficulty of treating the subject of heat conduction in the charge has not been overcome, because the hot vapors rising through the porous coke and through the voids of the coal on each side of the plastic layer carry with them much of the heat to the upper levels of the charge and entirely remove a large proportion as sensible and latent heat of a non-recoverable nature.

The rapidity with which the heat can be transmitted through the retort wall is not the limiting factor of heat transfer, even in refractory retorts, much less than in metallic retorts, for in either case the rate at which heat can be transferred through the walls is considerably greater than it can be transmitted into the charge. Hence, the speed at which heat is passed from the retort wall into the charge determines the rate at which heat can be imparted to the retort wall from the flue. In the very early stages of carbonization, the conductivity of the retort wall may be of importance, but it becomes relatively less and less of a factor as the plastic layer progresses inwardly, leaving an ever increasing envelope of non-conducting semi-coke behind, through which heat passes with the utmost difficulty. If heat be imparted to the retort wall from the flue faster than the charge can transmit it inwardly, the retort wall will rise in temperature gradually until, in the case of metallic retorts, the casting will be seriously injured.

Within the retort, heat transfer takes place primarily by conduction which argues for a full retort. However, since the evolved gas must pass out, a certain percentage of voids must be permitted to remain. We have already seen that the passage of hot gases through the voids may be an important source of heating by means of convection, and indeed, this expediency has been adopted entirely in certain internally heated retorts. It follows, therefore, that for maximum rate of heat penetration there should be an optimum size of the coal lump. This has been established by Guegnen, as quoted by Fulweiler (258), as being 2.5 inches when the period of carbonization is 4 hours.

A high velocity of flue gas is desirable, for not only does it increase the rate of heat transfer, but it promotes uniform heating of the setting. The generation of heat occurs, of course, at the point where the heating gas is combusted and this is largely at the entrance of the gas to the flue. A high velocity permits the heat to be carried rapidly away ad uniformly distributed over the retort and setting. This is of particular importance, where low temperature gas is being used to heat the retort, because of its high thermal value. As a matter of fact, it is not good practice to sue the low temperature gas, even in part, for heating purposes. Lower calorific producer gas should be used instead, as it heats the retort more uniformly, and because the more valuable gas of higher thermal content should be reserved for sale where there is a market for its disposition.

The purpose of checker-brick in regenerators is to absorb as much heat as possible from the flue gases before they are vented to the atmosphere, to store the absorbed heat for a short period of time, and finally to yield that heat to a cooler gas entering the combustion chamber. The transfer of heat from and to the gases in the regenerator takes place primarily by convection and conduction, although radiation has been observed to play an unimportant part. As we shall see later, under the subject of convection heat transfer, a thin film of stationary gas surround the checkers, its thickness depending upon the velocity of the gas. Heat transfer takes place from the moving gas through this film to the surface of the brick by conduction, so that, since the film thickness is a function of the gas velocity, heat transfer in the regenerator depends greatly upon the gas velocity.

For use in the regenerator checkerwork, bricks with a high heat-absorbing capacity are, of course, most desirable. As many bricks as possible should be used in the regenerator, so far as it is compatible with a large exposed surface. Because of the continual reversal from hot gas to cold air and vice versa, regenerator bricks are subject to spalling, so that particular attention must be paid to this property of the refractory. In addition to these properties, other characteristics of the material must be considered in selecting refractories for checkerwork in regenerators, such as their crushing strength and resistance to abrasion and corrosion. The design of the regenerator usually follows the practice established by long experience and the use of standard shapes and high mechanical strength should largely be considered in its construction. Heat is transmitted to the surface of the brick mainly by convection and is transferred to the interior by conduction at a rate which depends upon the thermal diffusivity of the material. Consequently, the rate at which heat is stored or delivered by the regenerator depends upon the thermal conductivity, density, and specific heat of the refractory, while the quantity of heat stored depends primarily on the specific heat capacity and the temperature to which the brick is raised.

Ingersoll and Zobel (296) developed a mathematical expression for the amount of heat absorbed by a brick heated from two sides, where the temperature of the two surfaces is held constant, viz:


where H is the heat flow in gram calories through one square centimeter of surface area in t seconds of time; p is the thickness of the brick in cubic millimeters; h2 = k / cp is the diffusivity, where k is the thermal conductivity in gram calories per centimeter cube per second per degree Centigrade; p is the apparent density; c is the specific heat; and finally, where Oo = Os - Ob, in which Os is the temperature of the brick surface and Ob is the average initial temperature of the brick. For a period greater than 5 minutes only the first three exponential terms need to be considered, but the fractional coefficient must be retained, the summation of the fractional series being 1.232.

This equation holds only for the particular case of parallel walls heated from two sides, which approximates the condition in some regenerators. Hougen and Edwards (324) have shown that, when the bricks are placed as checkers in the regenerator, a shape factor amounting to 0.97 must be introduced into the above equation, when the average temperature of the refractory is 750 C and when the free cross-sectional area of the regenerator passage is equal to the cross-sectional area of the brick.

Hougen and Edwards (324) calculated the time required for bricks of different thickness and made of different refractory materials to reach 95% thermal saturation. Their results are shown in Table 118. The time required for a brick of given thickness to attain a proportion of its maximum heat absorption varies as the square of the thickness. From the standpoint of thermal properties, it will be seen from the table, that for checker-brick, magnesia is the most desirable and carborundum next in order, but the former is eliminated by its tendency to spall and the latter, in ordinary commercial practice, by its cost.

According to Brown (325), a good high duty firebrick for use in checkerwork should contain less than 70% silica and should not soften below 1680 C, while siliceous brick should contain more than 70% silica and should not soften below 1630 C. Searle (263) states that chemical action and dust erosion are the chief causes of low durability in refractories which are used in regenerators and recuperators and their refractoriness is of relatively minor importance. Although costly, basic bricks are preferable to fireclay and grog materials because the basic nature of the gases causes rapid chemical action on siliceous bricks. On this point, as well as others, magnesia is again the most desirable material, but is ruled out of consideration because of its cost and spalling tendency. The latter fault is also found with silica brick, although in some cases they may be used in the parts of the regenerator where the temperature does not fall below the critical spalling temperature. When silica can be so used, its desirable thermal properties are highly advantageous. Phelps, as quoted by Booze (265) has apparently established beyond question that, contrary to the opinion of many operating men, there is a distinct advantage in using machine-pressed brick for the regenerators. He established that the heat-absorbing capacity of brick had a distinct relation to its specific gravity. The difference in this respect may be as much as 6% in favor of the machine-pressed sample.

Convection & Radiation ~

Heretofore, we have dealt only with the transmission of heat through the retort wall and through the charge. We have seen that this takes place principally by conduction. It remains now to treat the problem of heat transfer from the heating flue to the retort wall, this is accomplished in two ways, by convection and by radiation, the latter being effected both by gaseous and by solid radiation.

The transfer of heat by convection is a function of gas velocity, size of passage, and temperature difference, but it is almost entirely independent of the gas composition. Reynolds investigated heat transfer by convection as early as 1874, giving special attention to the matter of fluid turbulence above the critical velocity, which marks the transition from streamline or non-vortical flow. Above the critical velocity, turbulence becomes an important factor, and hence, heat transfer becomes a function of the fluid velocity. Reynolds attributed this phenomenon to increased molecular bombardment, but it is at present well established that the increase in heat transfer is due to a thin stationary fluid film whose thickness varies with the fluid velocity.

When gas flows past a stationary surface there is a frictional drag which causes the gaseous layers next to the surface to lag, to such an extent that the layer immediately in contact with the surface is at rest and is known as the stationary film. The transfer of heat through this film from the moving gas to the stationary surface takes place by pure conduction, and hence follows the law established in Equation [28], where the film coefficient, h, replaces the coefficient of thermal conductivity divided by the thickness. The film coefficient depends largely upon the unknown thickness of the film, which is a function of the gas velocity and its physical properties, as we shall see presently.

Reynolds (326, 327) deduced a formula for turbulent flow convection heat transfer of the following form:

[31]     h = a + b p V = a + bW / A

where a and b are constants, p is the density, V is the velocity of the fluid, and h is the film coefficient of heat transfer. The equation also can be written in the form which involves the mass flow, W, and the cross-sectional area, A, of the flue passage. Although his theory of molecular bombardment has now been displaced by the theory of a stationary film, Reynolds' equation still finds favor. The constant, a, is small, so that the heat transfer is roughly proportional to the velocity. Stanton (328), in verifying Reynolds' equation experimentally, came to the conclusion that the heat transfer varied as a power of the velocity somewhat less than unity.

Among the mathematical investigations of forced convection, the work of Boussinesq (329, 330) deserves particular consideration, but this work is based upon the premises of streamline flow with inviscid and incompressible fluids. Rayleigh (331) continued the study and demonstrated, from the principle of similitude, that the kinematical viscosity of the fluid must be considered. The fully developed Boussinesq equation for heat transfer by forced convection with turbulence can be written as:


where h is the film coefficient in British thermal units per square foot per hour per degree Fahrenheit; D is the inside diameter of the flue; K is the thermal conductivity of the gas in British thermal units per hour per foot cubed per degree Fahrenheit; V is the velocity of the gas in feet per second; p is the specific gravity referred to water, u is the kinematical viscosity in poises; Cp is the specific heat of the gas at constant pressure, and K, alpha and are constants to be determined experimentally.

Many other equations have been proposed for heat transfer from gases in turbulent motion during forced convection. Those derived by Nusselt (332, 333) and by Rice (334) are particularly noteworthy, but both of them have been shown by Cox (335) to be only special cases of the more general Equation [32] derived by Boussinesq and extended by Rayleigh. Thus, Rice (334) found that K = 53.5, alpha = 5/6, and = 1/3; and Nusselt (332) for simplicity placed, alpha = , thus eliminating u, and found that alpha = 0.786, combining the alpha-powers of D, k, and Cp with the constant, K, to give a new constant. According to Royds (336), the Nusselt formula can be put in the form:

[33]     h = e ( V p )0.786 = e ( W / A )0.786

where e is a constant and the other symbols are the same as used before. When the film coefficient of heat transfer, h, is expressed in BTU/sq ft/hr per degree Fahrenheit, then e = 4.10 for air, e = 3.39 for carbon dioxide, e = 11.01 for coal gas, and e = 7.67 for superheated steam. Later, Nusselt (333) extended his formula in the following manner:

[34]     h = 12 k / D  (D / L )0.054 ( V p D Cp / k )0.786

for the heat transfer from a flowing gas inside of a tube of length, L. It will be recognized that, since W = Vp, there is no essential difference in form between Equation [33] and Equation [34]. Nusselt's formula has its limitations and is not applicable to temperatures much above 1000 F, because of gaseous radiation, not for velocities below the critical, which according to Nusselt is located at about 3.3 ft/sec, but which, according to Lent (337) is, in the light of recent investigations, more nearly located at 13 to 16 feet per second.

Royds (336) feels that the experiments of Nusselt are confirmatory of Reynolds' Equation [31], and in fact it will be noted that Equation [33] differs from the former only in dropping the constant, a, which is small, and in raising the second term to a fractional power, as previously established by Stanton (328). The experiments of Jordan (338) on air in pipes may also be taken as confirmatory of the work of Reynolds. Jordan's data evaluated the constants in Equation [31] as a = 0.0015, and b varies from 0.00055 to 0.00090, depending upon the hydraulic radius, that is, upon the ratio of the area of flow to the perimeter of the passage, and upon the temperature of the stationary film. The smaller the value of the hydraulic radius and the greater the film temperature, the greater is the numerical value of b and hence the rate of heat transmission. The constant, a, in Reynolds' equation is supposed to represent the heat transferred by conduction and hence should increase with the temperature difference, but Jordan's data, which shows no increase, is inconclusive on this point.

As far as the present author is aware, while there have been limited investigations on forced convection for gases flowing outside of conduits, they have been confined entirely to transverse flow, and no data whatever are published on forced convection for gases in external longitudinal flow, which is the case found in a coke oven flue. The research of Carrier (339) is applicable only to air flowing transversely across small wrought iron pipes and does not hold when the gas flows longitudinally to a plane surface of either longitudinally or transversely to a large cylinder. The experiments of Nusselt (332), as well as of Bell (340, 341), and of Jordan (338), are all applicable only to the case of forced convection of gases inside of pipes. In the latte case, the measurements have been correlated by Weber, as quoted by Walker, Lewis, and McAdams (342), into the following formula, which holds for pipes with diameters up to 2 inches, gas temperatures up to 2000 F, mass flows up to 20 pounds of gas per second of free area, and for gases with molecular weights varying from 17, as in the case of illuminating gas, to 44, as in the case of carbon dioxide:

[35]     h = 0.88 W0.8 T0.5  S0.2 Cp  /  M0.3

where h is the film coefficient expressed in BTU/hr/sq ft per degree Fahrenheit; W is the mass velocity in pounds of gas per second per square foot of free area; T is the absolute arithmetic mean gas temperature in degrees Fahrenheit; S is the surface factor, that is, the heat transfer area in feet divided by the volume of the open gas passage; and m is the average molecular weight of the gas. The mass velocity is equal to the average linear velocity multiplied by the density of the gas.

According to the laws of radiation established by Stefan and Boltzman (343), the total radiant emission of a solid body is proportional to the fourth power of the absolute temperature. This is exact for an absolute black body, that is, for a perfect radiator, but is sufficiently accurate for other materials to be used in ordinary calculations. As it is usually the net heat transferred by radiation that is of interest, the Stefan-Boltzman law becomes for the case of two infinite planes:


where Q is the net heat transferred, A is the area, t is the time, T1 and T2 are the absolute temperatures of the hot and cold bodies, respectively, and p1 and p2 are the relative emissivities of the respective surfaces referred to the black body. The most probable value of the total radiation constant, C, is the mean of the corrected observations of many authorities as reported by Coblentz (344), who gives C = 1.37 x 10-12 gram calories/sec/sq cm/C4. In practice the receiving and emitting radiant surfaces are always peculiarly related, in regard to their shape and relative positions, so that it is necessary to introduce a proportionality shape factor, O, to provide the necessary correction from the special case of infinite parallel planes.

In most every case for metals, the total relative emissivity increases slightly with rise in temperature up to about 1500 C, after which the increase is rapid. The manner in which the relative emissivities of oxidized cast-iron and steel vary with the temperatures is illustrated in Table 119, form the measurements of Randolph and Overholzer (345). The relative emissivity of a body is the measure of its radiant energy absorbing power and represents the percentage of the incident radiation which enters the body and is manifest as heat, the remaining energy which impinges being reflected into surrounding space by opaque bodies and being transmitted by translucent bodies. A good radiator is also a good absorber. Although the radiating power of a body varies with the wavelength of the radiation, as established by Kirschoff (346), an average value for the emissivity can be used in engineering calculations with sufficient accuracy to make consideration of the spectral selectivity unnecessary. There is hardly any published data on the emissivity of refractory surfaces at high temperatures, but Green (347) has reported an average value of 0.72 for a firebrick surface heated from 200 C to 500 C.

Flames are luminous or non-luminous, the luminosity depending on the amount of solid material held in suspension and heated to incandescence. Helmholtz in 1890 was the first to investigate radiation from flames and to observe that the radiation from luminous flames was more extensive than that from non-luminous flames. His student, Julius, found that the radiation from combustion products could be attributed almost entirely to carbon dioxide and water vapor and he established that the radiation was highly selective, occurring in sharp bands. Callender (348) continued the investigation and found that non-luminous flames radiate about 10% of their heat of combustion, as compared to as much as twice that for luminous flames. The radiation from each individual incandescent particle in a luminous flame undoubtedly follows the Stefan-Boltzman law and hence the radiation varies as the fourth power of the absolute temperature. Consequently, as gaseous radiation varies with temperature at a power somewhat less than the fourth, it follows that the more incandescent particles there are suspended in the flame, and hence, the greater its luminosity, the more nearly it approaches the true Stefan-Boltzamn relation. Haslam and Boyer (349) have conducted experiments which lead them to conclude that the radiation from luminous flames of methane, ethylene, and acetylene is 25 to 30% of the combustion heat and that this radiation is perhaps fourfold that from a non-luminous flame.

As distinguished from convection heat transfer, gaseous radiation is a function primarily of temperature, the gas composition, and the shape of the flue. It is independent of gas velocity, except insofar as the presence of fresh gas maintains the temperature. From a consideration of the intimate nature of radiation, it can be demonstrated theoretically and verified by experiment that a body is capable of absorbing its own radiations. Consequently, it is only from near the surface of a flame that radiation is emitted, since radiation from the interior is absorbed by the surrounding gas and does not penetrate great depths of the flame. However, the greater the mass of the flame, the more easily it can preserve its radiating temperature.

No all gases have ability to absorb radiation easily and it so happens that the products dealt with in combustion are particularly endowed with this ability, both carbon monoxide and dioxide, as well as water vapor and hydrocarbons, all being unusually efficient radiators, while oxygen and nitrogen lack the property almost entirely. Consequently, the radiating efficiency of a given gas mixture falls off as its content of oxygen and nitrogen increases, due to reduction in the concentration of the active gases. The luminosity of the flame, of itself, has nothing whatever to do with its power to radiate but the luminosity of a flame due to suspended particles greatly increase the heat radiated. This, however, is solid radiation, as distinguished from gaseous radiation, an entirely distinct phenomenon. According to its content of suspended particles, such as soot, slag, etc., a flame can pass through all the stages from pure gaseous to solid radiation. Indeed, Lent and Thomas, as quoted by Schack (350), noted that the addition of a little benzol to blast furnace gas, to render it strongly illuminating by particles of soot without increasing its temperature, increased its radiation fourfold.

Until the thickness becomes sufficient, the gas radiates energy at a power of the temperature somewhat below the fourth. Table 120, after Schack (350), gives the radiation from water vapor and carbon dioxide as a function of temperature and thickness. The radiating and absorption power of a gas is highly selective for energy of particular wavelengths and is confined entirely to the regions indicated in its absorption spectra. Thus, the radiation and absorption of energy by gases takes place in spectral bands, and the capacity for radiation in each band may be decidedly different. Carbon dioxide radiates from 3 bands and available measurements indicate that a 12.5% concentration of that gas radiates from the second band, in wavelengths from 4.1 microns to 4.5 microns, practically as a black body when the thickness is only about 3.2 inches, whereas the gaseous layer must be a score or more times that thickness before the radiation for the first band, with wavelengths of 2.6 microns and 2.8 microns, approaches black body conditions. It will be readily appreciated that the effect of gaseous radiation from the carbon dioxide second band may be appreciable even in thin layers. The thickness required of the gas to approach black body conditions in the third band, extending from 13 microns to 167 microns, is intermediate of the other two. Radiation form water vapor also takes place in 3 bands of wavelengths, the first extending from 2.6 to 2.9 microns, the second from 5.6 to 7.6 microns, and the third from 12 to 25 microns.

Haslam, Lovell, and Hunneman (351) investigated the radiation from non-luminous flames and found that it amounted to 14.9% of the total heat of combustion for methane, 13.8% for illuminating gas, and 10.4% for carbon dioxide, when they were burned with the theoretical amount of air, the radiation decreasing with excess air, as established by Helmholtz and by Callendar. They also confirmed the results of Helmholtz that the radiation from a given thickness of flame decreases with the preheat temperature of the primary air. For small thicknesses, of less than 30 cm, the radiation varies almost linearly with the flame thickness in non-luminous flames, but beyond a thickness of 100 cm it becomes practically constant.

Lent and Thomas (337) measured the heat radiated from a cylindrical gas column 48 inches in diameter, containing 21% carbon dioxide and 2.4% water vapor. They found 2,500 kilogram calories per square meter of flame area per hour radiated at 300 C; 6,000 kg calorie/m2 at 500 C, and 18,000 kg/m2 at 800 C. A method of calculating heat transfer by gaseous radiation has been proposed by Schack (352) and has been applied by Broido (353) and by Wohlenberg and Lindseth (354) to the problem of heat transfer in boiler furnaces. For a more detailed discussion of the method, reference should be made to the original articles quoted.

An exact calculation of gaseous radiation can be carried out from Planck's law, and Hottel (355, 356) has used this method to compute charts by which the radiation from carbon dioxide and water vapor can be determined easily for flues of various types, for which he has computed the necessary shape factors. His calculations show that, for a heat absorbing body whose surface is maintained at 800 F, the heat transfer by gaseous radiation is of the same order as that transferred by convection in a 6 inch square flue with an average gas temperature of 1358 F and with the gas containing 20% carbon dioxide. Since these are temperatures somewhat similar to those obtaining in low temperature carbonization practice, it appears that gaseous radiation plays an important part in heat transfer. It is in the order of 965 BTU/sq ft/hr.

Goebel, as quoted by Schack (350), observed that, due to the different radiating power of the gases, the coefficient of heat transfer between regenerative checkerwork and the flue gas was higher than between air and the checkerwork during reversal. In designing a heating flue, it is necessary to consider the transfer of heat by both radiation and convection, so that the maximum benefit can be gained from the predominating influence at the zone in question. Thus, at points where the temperature of the gases is extremely high, the gas velocity may be retarded, as radiation is independent of that variable, but where convection heat transfer is dominating, the passages should be narrow and a high gas velocity maintained. The present author is of the opinion that in low temperature carbonization, where the flue temperatures may reach 760 C, heat transfer by gaseous radiation may amount to from 10% to 20% of the total heat transferred, depending upon the design of the particular setting.