Although this page is often accessed as a stand-alone piece, it is part of a work on the history of the British and Irish cement industries, and where statements are made about the historical development of techniques, these usually refer only to developments in Britain. Formulae in bold are Cement Chemists' Notation.
History of the understanding of Portland clinker
The early history of cement is dominated by the non-scientific approach of its then non-scientist manufacturers, and even today, an understanding of the constitution and chemistry of the product is quite unnecessary for the manufacturer, at least until something goes wrong. A few simple "rules of thumb" serve well.
The first milestone in the understanding of clinker was the mere recognition that its presence was required in Portland cement. Early manufacturers of pre-Portland products, including Joseph Aspdin, had always accidentally made quantities of dark, glassy, "over-burned" material, which they discarded as useless because it was hard to grind and slow to react. It was William Aspdin, presumably some time in the 1830s, who recognised that it could be used to make relatively high strength cement. I. C. Johnson made the discovery independently in 1845, but not before he had submitted a sample of Aspdin's cement to a chemist, who pronounced it to be calcium phosphate. Such was chemistry's first contribution to the subject! Nevertheless, as early as 1856, Winkler in Germany was proposing that cements contain basic silicates that hydrate to lime and lower silicates. Such theories gained ground as the ability to perform accurate chemical analyses improved, but remained speculative until the 1870s, when Le Châtelier first applied petrographic techniques - specifically thin section transmission microscopy - to the study of clinker.
Le Châtelier's description of clinker composition, presented as early as 1883, was essentially that of today, and subsequent history has been one of rather minor refinement of its details. Many other researchers took up his petrographic methods, although not all had his scientific ability. The Swedish researcher Törnebohm independently repeated Le Châtelier's work and, finding four different crystal types in clinker, named them alite, belite, celite and felite. The ternary system CaO-SiO2-Al2O3 was the first such 3-oxide system to be completely characterised, by Rankin and Wright at the Carnegie Institute in 1915, showing the importance of cement research at that time, and also showing the shift in the centre of gravity of research effort to the other side of the Atlantic. In Britain, this sort of academic research took a long time to penetrate, a situation exacerbated by research during the 1920s conclusively proving that alite does not exist. It was only when XRD techniques were applied to clinker analysis by Brownmiller and Bogue (1930), unambiguously identifying and quantifying the phases, that the Le Châtelier model of clinker became mainstream. After this, rather belatedly, the British industry began to employ scientific principles, with research conducted, not by the industry, but by the likes of Lea at the Building Research Station.
Clinker Phases
Portland cement clinker is a "phase assemblage" in which the main mineral crystalline species are:
All these chemical formulae are "rough" because the phases are actually solid solutions, and the atoms of the formula are substituted to a small but significant extent by other elements. For instance, some of the calcium locations in the belite crystal can be occupied by magnesium, and some of the silicon locations can be occupied by aluminium, phosphorus, sulfur, iron etc. Furthermore, the crystals of the phases are usually highly defective, with discontinuities in their structure and missing atoms, this being a desirable characteristic because defects increase the crystal's reactivity.
The above are the four essential phases of the "Portland clinker system". In practice, various other phases may be present in small quantities:
Calcium oxide - CaO
Periclase - MgO
Salt phases - various combinations of sodium, potassium and calcium cations with sulfate and chloride anions, such as:
Arcanite - K2SO4
Calcium Langbeinite - K2Ca2(SO4)3
Aphthitalite - K3Na(SO4)2
Sylvite - KCl
Low temperature phases - various intermediate chemical species that have escaped further thermal processing, such as:
Spurrite - Ca5(SiO4)2(CO3)
Ternesite - Ca5(SiO4)2(SO4)
Ellestadite - Ca10(SiO4)3(SO4)3(OH)2
Ye'elimite - Ca4(AlO2)6(SO4)
A typical modern good-quality general purpose grey clinker might contain 72% alite, 9% belite, 7% tricalcium aluminate, 10 % tetracalcium aluminoferrite, 1% salt phases and 1% free calcium oxide, but depending on the properties desired, clinkers with markedly different compositions may be made.
Alite and belite are the "active ingredients" of cement, producing strength when hydrated. They are "primary phases" that remain solid throughout the clinkering process. Tricalcium aluminate and tetracalcium aluminoferrite contribute no useful properties, and can be to some degree deleterious. They are "interstitial phases" that crystallize from the liquid formed by partial melting during clinkering, and fill in the spaces between the silicate crystals. They are present purely for the convenience of the manufacturer, because, without the liquid from which they form, the clinker can't be formed at an economically viable temperature.
Although modern manufacturers often fine-tune the chemistry of their clinker by minor additions to the rawmix of "tertiary materials", historical rawmixes have usually been made from limestone and the most conveniently sourced local clay or shale. It is therefore no surprise to find that elements such as silicon, aluminium, iron, alkalis, titanium, phosphorus and manganese are present in much the same proportions as are found in the Earth's average continental crust, and trace elements, if non-volatile, also follow this rule. The minor and trace elements are usually present in amounts sufficiently small that they can be accommodated in solid solution in the four main clinker minerals.
Early clinker composition
The phase diagram for the CaO-SiO2 (C-S) system (shown right) was first accurately determined by Day, Shepherd and Wright in 1906. This shows that there are four stoichiometrically distinct calcium silicates (each with multiple polymorphs): CS, C3S2, C2S and C3S. Portland cements contain mixtures of C2S and C3S. The zone containing Portland cements is shaded.
Early cement manufacturers knew nothing of this; they only knew that certain ratios of limetone to clay in their rawmix produced acceptable cement, and others didn't. They only determined the calcium carbonate content of their rawmixes (if they did any chemistry at all) to control their composition. To simplify the understanding of the system, it is helpful to consider mixtures containing silica and calcium carbonate only. Such a mixture can (with a few mineralogical tricks) be made to react to give a mixture of calcium silicates. As an example, consider a mixture containing 20 g SiO2 and 70 g CaCO2. Using the chemistry data in the molecular mass and minerals pages, the results of the reaction of these can be calculated. The carbonate, on calcination, leaves 44.82283 g of CaO. So we have 0.799303 mol of CaO and 0.332866 mol of SiO2, and the CaO:SiO2 molar ratio is 2.40128. It's clear, therefore, that the reaction product will be a mixture of C2S and C3S; in fact it will contain 59.872 molar % of C2S and 40.128 molar % of C3S. Converting to mass percentages, we get 52.954% C2S and 47.046% C3S.
Repeating this calculation across the composition range, we get a relationship as shown. In this simplified model, the limits of the Portland cement range are 76.91% and 83.33% calcium carbonate. Below the lower limit, we get mixtures of C2S and the non-hydraulic rankinite. Above the upper limit, we get mixtures of C3S and expansive free calcium oxide. This immediately emphasises the difficulty experienced by the early Portland cement manufacturer; successful operation involved staying within a composition window only 6% wide, while dealing with variable natural raw materials, and without the benefit of any chemical analysis, and in many instances, without any knowledge of chemistry. In practice, crossing the lower limit led to lower strengths, while crossing the higher limit caused fatally destructive expansion, so it is not surprising that the compositions of early cements, although widely scattered, cluster around the lower limit.
In real world raw material mixtures, the effect of other oxides, and particularly alumina and iron oxide, come into play, but a similar picture emerges. In clinker made from mixtures of chalk and the Medway clay originally used by William Aspdin, the alite content falls to zero below a CaO content of 61%. Early chemical analysis data is extremely unreliable, but CaO contents below this value - as low as 57% - are fairly common. Portland cement, by definition, contains alite, but it should not be assumed that cements with lime content below 61% are non-Portland. This was because of the variability of the product. The cement of the time was made from a blend of clinkers. Because of the crude methods of compounding the rawmix, the clinker varied vastly from day to day, from batch to batch, from lump to lump, and even within the individual lump. The clinker of the time was, like the curate's egg, good in parts. A clinker with an average CaO content of 61%, half of it might to Portland, and half sub-Portland. Parts of it (with CaO 6% above the average) might even be over-limed.
Having said this, it must soon have become apparent that a higher standard of rawmix blending was required, and that this allowed a somewhat closer approach to the upper CaO limit, with increased strength. By the 1860s, the "barn-door" target emerged, and for the next 20 years (until German competitors started to increase CaO contents with the aid of chemical analysis) most clinker had an average CaO content safely midway between the upper and lower limits (and therefore with equal amounts of alite and belite). After 1880, CaO contents gradually rose, and by 1970 they reached a practical maximum, giving clinkers in which the silicate was nearly all alite.
There now follows a discussion of the individual minerals.
Approximately Ca3SiO5. Alite is the main mineral (>50%) in most modern clinker, and is the characteristic mineral of Portland cement. Alite only becomes stable above 1250°C, and can't form unless the burning temperature gets above that at some stage. Thus, in early cements, that were burned at a lower temperature, alite was not formed (except accidentally), and it was William Aspdin who first produced it, when he invented "Portland cement as we know it".
Le Châtelier first identified it as the main phase in clinker, and assumed it was tricalcium silicate because, under the microscope, it was visually identical to the mineral in "grappiers" - un-slakeable silicate nodules in over-burned lime, which contained so little of other species that a bulk chemical analysis was sufficient to establish its CaO:SiO2 ratio (nearly 3).
Alite forms distinctly hexagonal crystals. The crystal unit cell of pure tricalcium silicate contains 27 calcium ions, 9 oxide ions and 9 orthosilicate ions. The oxide ions in its structure confer on it a high reactivity, which causes it to develop "early strength" (i.e. strength developed in mortar or concrete during the first seven days). Its hydration is complex but can be represented very roughly by the equation:
2 Ca3SiO5 + 6 H2O → Ca3Si2O4(OH)6 + 3 Ca(OH)2
The hydrate, of fairly variable composition, is usually called calcium silicate hydrate (C-S-H).
Alite has several polymorphic modifications, all of which, as mentioned above, are unstable below 1250°C, decomposing into dicalcium silicate and calcium oxide. In practice, provided that small additions of other elements are present, and cooling from the peak temperature is very rapid, the alite is preserved down to room temperature as a meta-stable phase. Its meta-stable nature contributes to its reactivity.
The composition of real-world alite departs from that of tricalcium silicate, with a wide variety of "foreign" atoms in solid solution. Taylor quotes a "typical" composition:
Na0.01Ca2.90Mg0.06Fe0.03Al0.04Si0.95P0.01O5.00
Clearly, the amount of each "foreign" ion depends upon the amount of the element in question in the bulk composition, but aluminium and iron are always present. The alite and belite accommodate nearly all the phosphorus in the system, and any sulfur not tied up in salt phases. It is because the phase compositions deviate so widely from the "ideal" composition that the conventional "Bogue" calculation of phase composition is so wildly inaccurate.
Approximately Ca2SiO4. Dicalcium silicate can be made at low temperatures, and its "hydraulic" forms (i.e. those that react and set with water) form above about 820°C. It was the strength-giving constituent of early cements.
Belite forms crystals that lack well-formed angular surfaces and appears as globular masses under the microscope. The crystal unit cell contains 8 calcium ions and 4 orthosilicate ions. Belite reacts with water much more slowly than alite, and is typically only half reacted after a month. Its hydration (also complex) can be represented very roughly by the equation:
2 Ca2SiO4 + 4 H2O → Ca3Si2O4(OH)6 + Ca(OH)2
The hydrate is essentially the same as that produced by alite. Both reactions produce calcium hydroxide as a by-product. The alite paste is more diluted by calcium hydroxide (which produces no strength), and so its "ultimate strength" (strength after an infinitely long curing period) is lower than that of belite. The calcium hydroxide makes the mortar or concrete highly alkaline, which is of benefit in reinforced concrete - the alkaline conditions prevent steel from corroding.
If a mixture of calcium carbonate (CaCO3) and silica (SiO2) are mixed together in molar ratios between 2:1 and 3:1, very finely ground and heated to 1400°C, then a mixture of alite and belite forms, the amount of each being related to the composition of the original mixture. There is no calcium silicate with higher calcium content than alite, so mixtures with molar ratios over 3:1 will form a mixture of alite and free calcium oxide. If the molar ratio is less than 2:1 then a mixture of belite and lower, non-hydraulic silicates such as rankinite forms. This process exemplifies the way in which rawmix composition is used to control the clinker composition, although the real-world problem is much more complex due to the presence of many other elements in the mix. It also exemplifies the sensitivity of the system to relatively small changes in rawmix composition. The mixture to make pure belite contains76.91% by mass of calcium carbonate, whereas that for pure alite contains 83.33%. So the product is entirely transformed in the course of a 6.4% rawmix composition change.
Belite also has multiple polymorphs. γ-dicalcium silicate (the low-temperature form) differs from the others in that it is non-hydraulic. Its structure is completely different from that of belite (it is lime olivine) and the inversion of the β- to the γ-form involves so complete an atomic re-arrangement that the crystal falls to dust. This "dusting" was an often observed (and feared) process in the early industry. Decomposition of the belite usually set off simultaneous decomposition of the alite, and the resulting dust was useless. As with alite, the high temperature forms are preserved as meta-stable phases at room temperature by "doping" the silicate with "foreign" atoms, and by rapid cooling. This is easily achieved in modern kiln systems, but it was difficult to get good cooling in - for example - a bottle kiln.
The composition of real-world belite departs from that of dicalcium silicate, to an extent somewhat greater than that of alite. Taylor quotes a "typical" composition:
Again, the amount of each "foreign" ion depends upon the amount of the element in question in the bulk composition, but aluminium and iron are always present. In clinkers more modern than that on which this analysis is based, where an excess of sulfur over alkalis is found, significant amounts of sulfate ion are also present, with an SO42- ion and two AlO45- ions substituting for three SiO44- ions.
Approximately Ca3Al2O6. Of the many calcium aluminate minerals, this has the highest calcium content, and is the only one normally present in Portland clinker, although early cements could contain mayenite (roughly Ca6Al7O16.OH). As an interstitial phase, its crystals grow to fit into the available gaps between the silicate crystals.
A number of polymorphs exist and are stable over the entire temperature range: which polymorph predominates depends upon the amount of minor elements - particularly alkalis - entering the structure. Pure tricalcium aluminate has only one crystal form - a cubic structure in which each unit cell contains 72 calcium ions and eight ring-shaped Al6O1818- ions. The structure has many vacant cation sites, and so it can accommodate alkali by two alkali metal ions replacing a calcium ion. In the pure Ca3-nNa2nAl2O6 series, new polymorphs appear as n increases:
n
Crystal
Name
0-0.04
Cubic
CI
0.04-0.16
Cubic
CII
0.10-0.20
Orthorhombic
O
0.20-0.25
Monoclinic
M
n=0.25 is the maximum substitution, corresponding to a Na2O content of 5.7%. The often-quoted compound NaCa4Al3O9 (n=0.33, Na2O 7.6%) and its potassium analogue are mythical. However the tricalcium aluminate in cement always contains many other elements. In particular, substitution of silicon in the aluminate structure allows more cations to be accommodated, and for the isomorphs in the table, the values of n are all considerably higher in real-world aluminate phases.
As would be expected from their higher basicity, the alkali-loaded forms of tricalcium aluminate are progressively more reactive with water and because this makes setting of cement more difficult to control, the production of the orthorhombic and monoclinic forms is avoided in good manufacturing practice. This is usually accomplished by ensuring that there is sufficient sulfate in the clinker to tie up the alkalis in salt phases.
In the absence of sulfate, tricalcium aluminate reacts rapidly and exothermically with water to form hydrogarnet:
Ca3Al2O6 + 6H2O → Ca3(Al(OH)6)2 Hydrogarnet
This gives rise to the phenomenon of "flash set", obtained when cement is made by grinding clinker without gypsum addition. The cement/water paste sets within a few minutes of adding water. Addition of slowly-soluble sulfate in the form of gypsum (or, strictly speaking, its partially dehydrated form, bassanite) provides sufficient sulfate ion concentration to modify the hydration reaction:
The very insoluble ettringite forms a thin cohesive waterproof layer over the surface of the tricalcium aluminate crystal, and prevents further reaction - the aluminate is "passivated". By this means the final hydration of the aluminate is delayed until after the alite has undergone its much more leisurely setting reaction, after an hour or two. This delay, of course, is crucial for the effective use of cement, since the paste must remain fluid until the mortar or concrete has been placed.
The final reaction of the aluminate involves slow conversion of the ettringite into "monosulfate" - an "AFm" phase
This inconvenient behaviour is purely a property of the tricalcium aluminate. Cements with no tricalcium aluminate set normally, even with no sulfate present. Furthermore, the eventual hydration of the aluminate contributes little or nothing to the strength of the cement. All in all, tricalcium aluminate makes no useful contribution to the properties of cement, and can cause major problems.
In addition to alkali metal ions, the aluminate phase accommodates a large amount of "foreign" ions. Taylor quotes a "typical" composition for the cubic form:
Again, the amount of each "foreign" ion depends upon the amount of the element in question in the bulk composition, but silicon and iron are always present.
Very approximately Ca4Al2Fe2O10. Of all the phases, this has the most variable composition. In addition to its variable aluminium/iron ratio, it can take up large amounts of "foreign" elements. It functions as the "garbage can" of the system. Taylor quotes a "typical" composition for ferrite in clinkers with higher alumina contents:
Again, the amount of each "foreign" ion depends upon the amount of the element in question in the bulk composition. Transition metals Ti-Zn easily substitute for iron and the "ferrite" phase is the main destination of these in clinker. These elements, particularly titanium, may exceed iron in the "ferrite" of white clinkers. Mg2+ can also substitute for Fe3+, with charges balanced by substitution of an AlO45- ion with a SiO44- ion. This provides a home for MgO in excess of the 1.5% that can be accommodated in alite and belite, although clinkers with more than 3% MgO always contain the free oxide (periclase) as a separate phase.
In the absence of other elements, the "pure" mineral is brownmillerite, and this consists of a very broad solid solution series of formula Ca2Al2nFe2-2nO5 where n=0 to 0.7. The crystals are orthorhombic. The composition produced in real clinker depends upon the aluminium / iron ratio of the clinker as a whole, but is also affected by cooling rates and the presence of minor elements.
The ferrite, like tricalcium aluminate, hydrates to hydrogarnet:
In this case, the precipitation of insoluble iron oxide on the crystal surface protects it from further reaction - the ferrite is "self-passivating". Thus, although the ferrite hydrates fairly energetically, it does not produce flash set, and is not as problematic as tricalcium aluminate. It provides no strength, and affects cement strength properties only as a diluent.
In one significant property it is important: nearly all the transition metal ions in the clinker end up in the ferrite. The other phases are nearly colourless, but the ferrite has a dark greenish-grey colour, and is responsible for the overall colour of the clinker. It is for this reason that in making white clinker, the quantity of ferrite is minimised by restricting the amount of transition metals in the rawmix.
Free lime and periclase both have a deleterious effect on cement properties and manufacturers minimise the amount produced, but it is not possible to eliminate them altogether. Both tend to hydrate when cement mixes with water, and the resulting hydroxide occupies more space than the original, dense oxide. This is fine as long as hydration occurs while the concrete mix is still fluid or plastic, but, once structure and strength have started to form, the formation of hydroxides tends to have a destructive effect - and more so if it occurs after strength development is complete. The onset of an expansive reaction after the cement has gained strength is known as “unsoundness”, and has been a feature of poorly-manufactured Portland cement throughout its history.
Free lime exists in the clinker if the finishing reaction of lime with belite to form alite is not completed, if there are large unreactive particles of calcium carbonate in the rawmix, or if the mix contained too much lime. In early manufacture, both the control of burning conditions and the chemical control of the mix were hit and miss. The usual technique to deal with defective clinker involved grinding the cement with some water addition, and storing (“maturing”) the cement in contact with damp air for long periods – often months. The lime – which is the most reactive phase present – preferentially reacts with water, and is thus already hydrated before the cement is used. However, this did not entirely cure the problem, since the free lime only rapidly hydrates if it is on the surface of a particle. Any free lime deeply buried within the cement particles would be protected from hydration, and would cause unsoundness to emerge only after the outer part of the particle had hydrated away, during curing. Aside from this, addition of water at the manufacturing stage inevitably destroys some of the alite.
Control of free lime improved in the 20th century due to better mix control and the arrival of the rotary kiln, which allowed much better control of burning conditions, so that for the last century or so, unsoundness due to free lime has become an entirely avoidable condition.
Periclase was not much of a problem in the early British industry, because most of the manufacturing sites had raw materials with very low magnesia content. However, with higher-magnesia materials, the effect of periclase is more insidious than that of free lime, because it is much less reactive than free lime, and its hydration is much more likely to take place over a long period at a late stage in curing. Magnesia dissolves in all the four main phases (and particularly the ferrite) to a limited extent. Once this limit (which may be in the range 1.5-3.0%) is exceeded, periclase starts to form as a separate phase. The magnesia is somewhat more soluble in the clinker melt than in the solid minerals, and so periclase tends to crystallise out. When magnesia content is high, the most usual means of mitigating the problem is to ensure the clinker gets very rapid cooling, so that the periclase solidifies as very small – and therefore more rapidly soluble – crystals. The formation of large periclase crystals during slow cooling can lead to bad expansion of mature concrete. Periclase formation is also exacerbated by burning the clinker in reducing conditions (caused by insufficient fuel combustion air, or un-ground fuel). Under these circumstances Fe(III) is reduced to Fe(II), which substitutes for Mg(II) and raises the effective magnesia content.
It is the alite and belite that contribute strength to the finished cement. It is often claimed that aluminate and ferrite make small contributions, but there is no scientific basis for this. In particular, the observation that increasing aluminate boosts early strength is mainly due to the increased alkalis which usually accompany an increase in rawmix alumina.
In a simple comparison of alite and belite, alite reacts more rapidly (contributing "early" strength), but has a lower "ultimate strength" (i.e. strength after infinite curing time). That the latter should be the case is reasonable in view of the larger amount (see hydration equations above) of calcium hydroxide formed, which acts as a diluent. The more rapid reaction of alite is mainly due to its greater basicity, resulting in a more vigorous reaction with water. However, part of the difference observed in real cements may be due to alite's friability. On grinding clinker, alite is brittle and breaks easily. Belite, on the other hand, is tougher (although not necessarily much harder) and tends only to deform on impact. The aluminate and ferrite are also somewhat harder. For this reason, breakage of clinker particles occurs mainly through the alite crystals, so that polymineralic cement particles have mainly alite exposed on the surface, whereas belite tends to be buried in the middle of the particle. Its hydration is therefore delayed until the surface alite has reacted. For this reason, in cements consisting mainly of belite, the apparent reactivity of the belite is greater.
Micrograph of alite-surfaced cement grains. Etch colours: alite = brown, belite = blue, aluminate and ferrite = white.
Strength is also affected somewhat by the chemistry and mineralogy of the phases. Preservation of the higher-temperature polymorphs by rapid cooling enhances reactivity, particularly of belite. For both alite and belite, production of small, ill-formed crystals by rapid burning at the minimum temperature, produces the highest reactivity, while extended burning and/or excessive temperature allows the crystals to grow, healing up defects and thereby reducing reactivity. Chemistry also has an effect. Increased alkalis increase early reactivity while suppressing late strength. An excess of sulfur over alkalis can, as mentioned above, be accommodated by paired substitution in the belite, and belite with high sulfate levels displays a reactivity equal to that of alite. Certain minor elements, e.g. phosphorus, zinc, strontium and barium, cause substantial reductions in reactivity (and therefore strength) by stabilising the more reactive polymorphs.
When cement includes (or is mixed with) ground granulated blastfurnace slag or pozzolan, as is normal in recent years, the additive develops strength mainly by reaction with calcium hydroxide or with alkalis. Since alite hydration produces much more calcium hydroxide, cements high in alite perform best with these additives.
