Historic cements: effect of temperature and composition

This was an investigation that I did in 1986 (Note 1). It was prompted by my interest in early cements, having read Major Francis’ book, and in particular his discussion of Roman cements (Note 2), and follow-up reading of the lectures by Dr Halstead (Note 3) and Professor Skempton (Note 4). It becomes apparent that there are question marks over the analytical data quoted and the phases calculated in the compositions of the various historic building materials of the first half of the nineteenth century. This coincided with the recent arrival of a high-temperature furnace at Shoreham, where I was working at the time, and the availability of relatively precise XRF analysis methods. The analyses, although good by British standards, were subsequently (1995) repeated at Atlanta, where a system with state-of-the-art accuracy and precision was available.

In this study, the effects of uncertainty about particle size and inhomogeneity of raw materials was avoided by deliberately selecting a highly combinable pair of components, and making all mixes from these. The two components were both obtained from Pitstone (Note 5).

These materials were combined in various proportions in twenty 40 g lots: the powders were slurried with 30 mL water, well mixed and dried. The dried material was roughly hand-ground and re-mixed, then pressed at 200 kN into a cylindrical pellet of dimensions 40 mm diameter by about 11 mm thick. The pellet, in a 50 mm platinum dish, was placed in the furnace set at 600ºC, and the temperature was raised at 15ºC per minute (Note 6) to either 1150ºC or 1380ºC. The lower target temperature was reached in 45 minutes and the higher in 88 minutes. In both cases total furnace time was 120 minutes, after which the crucible was removed and rapidly cooled by suspension in water, seated in a splash guard. On no occasion was there any indication of “dusting” (Note 7), although the pellets, intact up to that point, usually warped and cracked into a few large pieces on cooling. The material was rapidly hand ground to <100 μm and free lime was determined by selective solution in ethane-1,2-diol. Phases were then calculated, using Taylor’s phase compositions (Note 8).

The chemical analyses of the two components (ASTM C 114, revised 1995) were:

AnalyteChalkClay
SiO20.8466.37
Al2O30.1517.01
Fe2O30.105.87
CaO54.681.44
MgO0.561.33
SO30.060.57
LoI95043.513.82
Na2O0.0050.48
K2O0.0171.74
SrO0.0650.01
TiO20.0080.83
P2O50.0040.41
Mn2O30.0230.13

The phases in the ten samples fired at 1150ºC were as follows:
roman graph 1150
The phases in the ten samples fired at 1380ºC were as follows:
roman graph 1380

The lower temperature series shows the break between hydraulic limes and “Roman” cement at around 25-26% clay content. The hydraulic limes lack reactive aluminates and are slow setting. In the range 19-25% clay content, hydraulic limes have less than 20% free lime. These would slake reluctantly and need to be ground prior to use: they correspond to Vicat’s typical product. Among the “Roman” cements, rapid set is brought about by rising content of reactive aluminates, but belite content falls rapidly with rising clay content, and probably become non-viable above 30% clay content.

The higher temperature series yields only Portland cement as a viable product, with clay contents between 19 and 25% clay content. The lower limit is governed by the free lime content. Below 19% clay content, we have a mixture of alite and dead-burned lime which develops strength rapidly, but becomes unsound. Above 25% clay, alite (which characterises Portland cement) disappears and the product becomes indistinguishable from that of “Roman” cement, although because of the high temperature, the product vitrifies. It is because of having to grind this hard material, and the waste of kiln fuel, that “Roman” cement manufacturers avoided higher temperatures. It is quite incorrect to suggest that, had they ground their discarded clinker, “Roman” cement manufacturers would have “discovered” Portland cement, since their clay content was outside the Portland range. On the other hand, manufacturers of artificial hydraulic limes of the Vicat type – and there were none of these in Britain – might well have done so.

The last point also disproves the frequently-repeated error, that Joseph Aspdin “discovered” modern Portland cement simply by burning harder. The innovation also involved a deliberate increase in lime content, and can’t be seen as accidental. This fits with the narrative of the product’s invention by William Aspdin.

NOTES

Note 1. It was with some amusement that I learned, much later, that Frost had done much the same experiment in 1836 and published it (Journal of the Franklin Institute of the State of Pennsylvania, XIX, April 1837, pp 277-280). He also narrowly missed inventing calcium aluminate cement.

Note 2. Francis, p 28.

Note 3. Halstead, pp 51-53

Note 4. Skempton, pp 130 ff./p>

Note 5. I suppose I should retrospectively thank Castle Cement and their successors for their involuntary donation of these samples on a quiet Sunday afternoon. I’m sure they didn’t miss them.

Note 6. At the higher temperature (1380ºC) the heat-up rate tailed off to about 5ºC per minute towards the end.

Note 7. i.e. β-γ inversion of belite and/or other phase changes that cause the material to fall to a powder.

Note 8. Taylor (1st ed), p 10

Note 9.