Thermodynamics of Clinker Formation

On many pages in this website, I emphasise that it is about History, and not a Cement Technology Course. This is particularly true of this page. This page is another prompted by re-reading Martin's 1925 report on The Theory of the Rotary Cement Kiln. Peter Jackson, writing in 1999, described Martin's 1932 re-write of the report in these terms:

. . . it is understandable that the book in question did not become a standard text book in the UK cement industry - which is a pity because it contained much useful material. What copies (that) were available enjoyed a similar secretive status on cement works as that adopted at schools in the 1930/1940s to copies of "Lady Chatterley's Lover".

Considering the circulation of Martin's work (in its brown paper wrapper) on cement plants throughout the country, passed furtively from hand to hand, it is obvious that it actually was a standard text book, and continued to be so throughout the thirty years that I was employed by cement companies. It is interesting to speculate that it was its cabalistic status that maintained its credibility long after its contents had been de-bunked.

I recently got hold of copies of his much-pirated works, and could track down the origins of some retrogressive misconceptions that continued to infect the industry during the rest of the 20th century. Prominent among these was the systematic misunderstanding of the chemical reactions that take place in turning a cement rawmix into clinker.

martin graph 3

All continuous cement kilns have similar temperature profiles as shown right. The gases (red line) travel right-to-left and are boosted in temperature by the flame at the hot end of the kiln. The feed, eventually converted to clinker (blue line) travelling left-to-right, rises in temperature in three distinct stages, first fast, then slow, then fast again. In the middle stage, the rate of temperature rise is depressed because energy is being absorbed by the endothermic decomposition of calcium carbonate. Heat transfer from the gas to the feed along the kiln is driven by the temperature difference between the two flows. The shape of the feed temperature curve means that there is always a "pinch point", where the temperature difference is small, at the cold end of the decarbonating zone. This critically affects the overall heat exchange effect of the kiln, and therefore controls its output and thermal efficiency.

That cement kilns were subject to this constraint was recognised by Martin in the early 1920s. Evidently, quantification of this effect requires an understanding of the temperature range over which decarbonation takes place, and the energy absorbed by decarbonation. Martin set about obtaining information about the decomposition of calcium carbonate.

The first item of evidence was the temperature variation of the pressure of carbon dioxide in equilibrium with calcium oxide. By 1921, several scientific determinations of this had been done. The determination is difficult because equilibrium is hard to establish, and impurities in the calcium carbonate can cause large disturbances. Martin used Johnstone's data (Note 1) of 1910. This deviates somewhat from modern data, the curve being slightly shallower, crossing the modern curve at around 850°C. The temperature at which the pressure reaches 1 standard atmosphere (101.325 kPa) was 899°C in the Johnstone data, and 896°C on the modern curve.

martin graph 3

If the partial pressure of CO2 in the surrounding atmosphere exceeds the decomposition pressure, no decomposition can take place (and, in fact, calcium carbonate will re-form). In the cement kiln, the atmosphere at the hot end of the decarbonation zone consists of combustion gases containing carbon dioxide, while at the cold end of the decarbonation zone, it contains also the carbon dioxide from decarbonation. This latter condition therefore defines the temperature at which decarbonation begins.

Martin simply assumed that the CO2 content was 24% by volume at this point - equivalent to a partial pressure of 24.3 kPa - and read off the temperature of 805°C from the Johnstone curve. The CO2 content he used was simply typical of the early (pre-1922) inefficient kilns with high fuel consumption and high excess air. A recalculation for various energy consumptions and 6% excess air, using modern decomposition pressure data, is shown right.

Martin also gave a long list of literature references giving information on decomposition temperatures. These variously gave the temperature range of decomposition, or the equilibrium temperature at atmospheric pressure, some referring to pure calcite, some to lime manufacture, and some to cement manufacture. In general, those referring to pure calcium carbonate decomposition agreed with his temperature, while those referring to cement kilns gave lower values. In those cases (and there were several) where temperatures below 600°C were suggested, Martin said these "must be too low" (i.e. it doesn't agree with the value he has so irrefutably demonstrated).

Martin also considered the decomposition of MgCO3, making the entirely unjustified (and still common) assumption that all the magnesium in the clinker derives from magnesite. Particularly in south-east England (which was the focus of his attention), less than half the magnesium in clinker derived from carbonates.

In addition to providing the temperature of onset of decarbonation, the decomposition pressure also yields - by means of the Clapeyron equation - the enthalpy of decomposition. The standard enthalpy increase for the reaction:

CaCO3 → CaO + CO2 - Equation 1

was already fairly well established. He gave a value of 420 kCal/kg (= 1.757 MJ/kg). The modern value is 178.31 kJ/mol or 1.7815 MJ/kg. A low value (1.766 MJ/kg) was still in use by Blue Circle in 2002.

Now, in this page, which is on a subject that can be broadly described as Chemistry (i.e. what chemists do), the above is the first mention of a chemical equation. Calcium oxide forms, and at a much higher temperature, sintering allows this to react (exothermically) with silica, alumina and iron oxide. But - wait a minute! (I hear you say) - is that the reaction that actually takes place in the decarbonation zone? Surely (you will say, if you're a chemist), it's this:

CaCO3 + 0.5 SiO2 → 0.5 Ca2SiO4 + CO2 - Equation 2

This reaction is also endothermic - but less so. As with the first reaction, the standard enthalpy change here can be calculated using the table provided. This time it's 115.08 kJ/mol (of CaCO3) or 1.1498 MJ/kg.

Poor Martin was entirely unaware of data on the enthalpies of formation of the silicates (Note 2), and in 1925 the nature of the minerals in clinker was by no means settled - many believed (and conclusively proved!) that alite did not exist. In the absence of a consensus on the chemistry, the default position was to assume (as non-technical 19th century manufacturers had always believed) that Portland cement was just a diluted form of lime, hence the use by Martin of data applicable only to the lime industry.

Most respectable modern texts on cement manufacturing chemistry admit that at least part of the calcium silicate is produced by direct reaction. There is reticence in admitting that (as is in fact the case) nearly all the silica in the rawmix reacts directly with calcium carbonate, possibly because, for the academic writer, it is difficult to replicate this reaction in a one- or two-gram sample in the laboratory. However, experience in real cement kilns is unequivocal. Both Lea and Taylor quote work published by Paul Weber (Note 3) in 1963, showing that this is the dominant reaction. In this instance, the process was studied by taking axial samples from operating kilns. In view of the fact that Weber's findings were not accepted by Blue Circle people, it is curious to learn that the same finding was established conclusively by Blue Circle as early as 1953.

The two models are as follows:

  1. Calcium carbonate decomposes according to equation 1 (and this can't occur below about 800°C): clays decompose to free oxides: at a much higher temperature, the lime then combines exothermically with the clay oxides to form the clinker minerals.
  2. Calcium carbonate undergoes a concerted reaction with clay silica (Equation 2) - commencing at a temperature much lower than 800°C, until all the silica is used up. The remaining calcium carbonate then decomposes forming calcium oxide according to Equation 1 in the range 900-1050°C. Finally, above 1300°C, the calcium oxide reacts with belite to form alite - a slightly endothermic reaction.

martin graph 4

In distinguishing between these models, the key parameter to watch is the free lime, as shown right. Model 1 is shown by dotted lines, and Model 2 by solid lines. The obvious difference is that, if free lime remains un-reacted until a high temperature, the entire calcium oxide of the clinker will be free before that temperature is reached. On the other hand, in Model 2, although decarbonation begins at a lower temperature, no free lime appears until the calcium carbonate is 70% decomposed, and the final free lime peak value is only about 30% of that in Model 1.

Suffice it to say, that in all real-world kilns (Note 4), the pattern observed is always that of Model 2.

The fact that this is the case was demonstrated in an impressively unambiguous way by Blue Circle in 1953. The experiment was conducted in an unusually objective manner (Note 5). Just before midnight on 31st January 1953, the North Sea flood put nine cement plants out of action. These were (E-W) Crown & Quarry, Cliffe, Bevans, Swanscombe, Wouldham, Johnsons, Kent, West Thurrock and Metropolitan. A total of 31 kilns were crash-stopped, simultaneously and without warning. All but one of the plants, and 25 of the kilns, were owned by Blue Circle. Making the best of a bad job, while the plants were being dried out and prepared to restart (about a week), laboratory personnel were required to take samples at 3 m intervals along every kiln. These were analysed for moisture, CO2 content by calcimeter, and free lime. This yielded 660 data sets, and constituted a "warts-and-all" snapshot of 25 kilns in operation - kilns were running smoothly or upset, too hot or too cold. Free lime values peaked, at the rear of each burning zone, typically around 15%. Only three free limes exceeded 20%. CO2 values decreased below 10% before significant free lime was detected.

A theoretical objection to Model 2 is that it requires solid calcium carbonate to react with solid clay oxides, and that this should occur at low temperature. The temperature issue is easy to deal with. I used to do a trick in training sessions. I pose the question: "At what temperature does calcium carbonate decompose?" Anyone who has done O-level chemistry will say 900°C. I then produce a beaker with a few grams of calcium carbonate in it, and add some hydrochloric acid. "There you are: room temperature!!" The point is that, in the presence of an acid, the temperature of decomposition is depressed. In the kiln, the acid is microcrystalline silica. Furthermore, conversion to the silicate renders the decomposition much less reversible compared with the simple thermal process in Equation 1. The reaction in Equation 2 commences at a considerably lower temperature than that calculated by Martin - typically 600-700°C - hence the low values that were already in the literature in 1920, dismissed out of hand by Martin.

The mechanism of the low temperature reaction requires that a certain amount of liquid should be present. Only a trace is required in order to weld reacting particles together and this is provided by fine silica along with recirculating sodium, potassium, sulfur and chlorine. Even in short, early kilns, recirculation took place between the burning zone and the cold end. A few percent of liquid, together with the rotation of the kiln, causes the powdery rawmix to begin to sinter together. Anyone familiar with the form of mid-kiln samples at around 600°C will recognise this condition. Other mechanisms may exist, but regardless of the explanation, reaction at 650°C is a fact.

An interesting independent confirmation of the acidic effect of the clay component came from Germany during WWI (Note 6). Anhydrite was being investigated as a sulfur source in a process that ultimately became the Anhydrite Process of cement/sulfuric acid manufacture. Initially, the focus was purely on acid production. Anhydrite only decomposes thermally at about 1600°C, but it was found that the temperature could be reduced by adding other minerals. Vapour pressures were studied over a wide temperature range. Anhydrite alone produced a partial pressure of 20 kPa at 1455°C, whereas addition of silica reduced this to 1135°C, and addition of kaolin reduced it to 986°C. The situation is entirely analogous. Clearly, the presence of clay minerals in a cement kiln increases the CO2 partial pressure. A typical decomposition pressure of calcite in the presence of microcrystalline clay oxides is 27 kPa at 650°C, 56.3 kPa at 700°C, and reaches atmospheric at 745°C.

In overestimating the temperature of onset of decarbonation by 150°C, Martin got the heat exchange constraints in kilns wrong, but this was of little importance in future developments. Much more important was the erroneous concept of the exothermic reaction that he set in motion. In order to understand this, it is necessary to return to the two reaction sequence models.

For simplicity, I consider only the reaction of silica: this is much more electronegative than alumina or iron oxide, and so reacts much more energetically with calcium. For alumina and iron oxide the effect is similar but less pronounced. The enthalpies are modern standard values, at 25°C.

Model 2 shows that the formation of clinker is endothermic at all stages, whereas Model 1 suggests that, at high temperature, the reactions to form the silicates are highly exothermic. There is still some controversy about the reaction between belite and calcium oxide to form alite, possibly because of the many combinations of polymorphs that might be involved, and some sources say the reaction is slightly exothermic. Furthermore, using the standard values given, the endotherm diminishes with temperature, and reaches zero at 1430°C (Note 7). However, whether endo- or exothermic, the enthalpy change is near zero, and occurs against the backdrop of melt formation, which is strongly endothermic (typically 65 kJ/kg clinker).

Because every stage of the process is endothermic, Model 2 suggests that the formation of clinker is an unrelenting arduous "uphill" process until the peak temperature is achieved. Any drop-off in heat input results in an immediate downturn in liquid formation and deterioration of the clinker "degree of burning". That this is indeed the case is the daily experience of every kiln operator.

On the other hand, Model 1 implies a massive "exothermic reaction" which, once started, is sufficient to cause the feed to heat up spontaneously by 200-300°C. Starting at 1300°C, this would produce molten clinker (and probably molten brick and kiln shell). That this is not the case is obvious to any objective observer who has looked inside a kiln. And yet, throughout the 20th century, this was the "standard model". In the manner of Hans Christian Andersen's "Emperor's New Clothes", trainee operators were encouraged to look in the kiln and see something that wasn't there.

Of course, in Martin's time, the exact magnitude of the heat of formation of the silicates from the oxides was not known - some even suggested it was thermally neutral, in line with the old idea that a cement kiln is just a kind of lime kiln. During the 1920s, the BPCRA group made bizarre attempts to derive the heat of reaction by difference from heat balances. As might be expected, such attempts are doomed to failure, even in relatively accurate modern heat balances - the accumulated errors are far too great. William Gilbert in 1930 said (Note 8):

The actual kiln tests do not show (by difference) a steady value for the exothermic reaction during clinkering. All the essential heat quantities except the exothermic reaction are measured during a test, and the practice has been to denote by "exothermic reaction" the amount by which the heat which is calculated to be necessary exceeds the heat due to the coal which is actually burnt in the kiln.
On various tests the exothermic reaction determined in this manner has ranged from +2.75 to -1.79 per cent (Note 9) on the clinker. Generally speaking, a positive value of the exothermic reaction has been found on the longer kilns and a negative value on the 100-ft. kilns. This discrepancy is due to causes not yet fully understood.

The first thing to understand would be error theory. A. C. Davis (or more likely one of his anonymous minions) gave an excellent review of the available data in 1929 (Note 10). The accounts fall into two distinct camps: those who were pursuing, with gradually increasing accuracy the value of the heat generated when silicates are formed from the oxides: and those who were making observations on real-world kilns. Among the latter was kiln pioneer Henry S. Spackman, who in America in 1905 said

In general it is assumed that heat is liberated (during the sintering process) but experiments, not yet completed, in my laboratory show that this reaction is only weakly exothermic or probably endothermic.

Note the careful use of language - "assumed", not "observed". The assumption is that of one who knows a little chemistry, but has never seen a cement kiln. Prefiguring my remarks above, the German Friedrich Carl Wilhelm Timm remarked in 1906:

Richards and Naske have recently put forward the view that the sintering of cement proceeds exothermically, and on that account consider that the sintering zone is too short in comparison with the CO2 zone. In the first place, this conclusion was due to an erroneous gas analysis, and in the second place the strong exothermic reaction assumed by Richards and Naske does not agree with Le Chatelier's work, which was based on a study of certain constituents of cement. Observations on other cement kilns show that at most there is only a very small exothermic reaction, and my own calculations confirm this. An immediate consequence of a strong exothermic reaction would be the development of a process of burning such that if the raw materials were placed in a chamber and heated to 1175 deg. C by means of grate-firing, and were then played upon at one point with a blowpipe flame, the entire mass would spontaneously ignite and its temperature would increase to 1450 deg. C!

So it's not as though the world was ignorant of the true facts, even in the first decade of the 20th century. It's just that the British industry chose to ignore them. In 1932, William Gilbert, by then knowing the thermochemistry of alite and belite, was capable of saying in 1934 that the feed temperature in the Swanscombe kilns rose spontaneously from 1010°C to 1343°C (the peak temperature). This continued throughout the rest of the century. Although people in touch with chemistry (e.g. Taylor, Lea) wrote learned works that skated around the issue (while not, of course, actually saying the Emperor was naked!), the completely mythical exothermic reaction remained part of the commonplace narrative of cement kilns. Promoters of this old nonsense are now mostly retired, but can still get their daft ideas published (I won't name them! Just Google!)

NOTES

Note 1. "The thermal dissociation of calcium carbonate", JACS, 32, 1910, p 938.

Note 2. Le Chatelier (1901) had determined the heat of formation of the relatively-accessible wollastonite (CaSiO3), the formation from the oxides being exothermic: -81 kJ/mol, so it was evident that the value for dicalcium silicate would be more than this. Tschernobaeff (Rev. Metallurgie, 2, 1905, pp 729-736) used Le Chatelier's direct calorimetric method to obtain values for various mixtures of silicates, and assigned heats of formation from the oxides of -119 kJ/mol for both Ca3SiO5 and Ca2SiO4 - values fairly close to the modern ones. Determinations on the pure phases were finally done in 1934: Johannson and Thorvaldson, JACS, 56, 1934, p 2327.

Note 3. Paul Weber, "Heat transfer in rotary kilns with due regard to cyclic processes and phase formation" in Zement-Kalk-Gips 9, Special Edition, 1963, Bauverlag.

Note 4. Here I am talking about the vast majority of kilns that use a single inter-ground and pre-blended rawmix. Where, as in a few modern installations, the rawmix components are introduced as separate streams at different points in the system, the considerable thermal advantage offered by a combined rawmix is sacrificed for special process reasons. In these instances, the pattern may be near to model 1.

Note 5. By which I mean that there was no opportunity to design the experiment in such a way as to yield a pre-ordained politically acceptable result.

Note 6. Subsequently published: Neumann, Z. angew. Chem., 39 1926, p 1537.

Note 7. E. C. Jøns, Cem. Concr. Res., 10, 1980, p 103.

Note 8. "The Rotary Kiln in Cement Manufacture: VII" in Cement and Cement Manufacture, 3, Nov 1930, p 1479.

Note 9. This is % Standard Coal, equivalent to a range of +773 to -503 kJ/kg. The actual value is around +420 kJ/kg.

Note 10. "The Reaction in Burning Cement" in Cement and Cement Manufacture, 3, Nov 1929, pp 303-314.