The formation of ettringite within cementitious systems can be seen as either a boon or a bane. Ettringite can be used as the structural hydrate within mine pack grouts and small amounts for shrinkage compensation in more substantial grouts. Unfortunately, the expansive nature of ettringite formation can also produce structural  disruption of the grout, cracking and  finally failure. For some time a number of ettringite formation mechanisms have been proposed. However, none have ever successfully be able to describe the observed behaviour.

It is proposed that the study of high yield grouts which are currently used in mine packing can be used to give an uncluttered view of the mechanism of ettringite formation. These grouts can be simply constructed¹ and can give uncluttered, easily prepared samples for analytical examination. 

The currently accepted mechanism for the formation of ettringite is based on super saturation, a similar type of mechanism which is accepted for the formation of CH and CSH hydrates in the hydration of Portland cement. It is  suggested that the anhydrous raw materials, when introduced to water, dissolve and form a super saturated solution within the pour solutions. The crystals of ettringite then form from solution within the pores and between the cement grains. As the density of ettringite is lower when compared to the bulk density of the wet grout, expansion occurs.

The laboratory preparation of ettringite crystals can be achieved using classical synthetic methods such as Strubles method². These methods are useful for the formation of quantities of very fine crystals for XRPD and other crystallographic investigations as the chemical compositions can be readily predetermined. The methods, however, are unsuitable for the preparation of large single crystals.  The morphology of the crystals prepared using Strubles method are radically different when compared to those found in cementitious high yield grouts³ where the morphology of the large lath-like ettringite crystal is used as a structural hydrate. 

Considerable work has been undertaken to grow lath-like ettringite crystals where by chemistry is used to selectively poison facets of the ettringite crystal and encourage a lath-like morphology. Intriguingly these elegant methods are absent from simple aluminate cement/calcium sulphate mixtures which readily produce lath-like ettringite crystals.

mechanism

For the purpose of this discussion it is assumed that the CH/CSH type of mechanism is not correct and that ettringite is not formed as an amorphous mass which then develops the familiar lath-like crystal morphology over time.   

Ettringite crystals are regularly observed in cement systems during formation and are also associated with cracking and the decay of mature concrete matrixes. Recent research⁴ suggests that the formation of ettringite is closely associated with the formation of a seed site.

The concept of a seed site would suggest that ettringite crystals could be placed directly into Strubles mixture and subsequent crystal growth would then occur. Unfortunately, this is not the case.  

The concept of the seed site must be better defined prior to further discussion of the hypothetical mechanism. There are two basic types of seed site mechanisms which can be considered here. The seeding of a concentrated solution with small crystals of material desired onto which lattice atoms aggregate from solution to form massive crystals. This type of mechanism, as previously mentioned, does not work in the case of ettringite and should be discarded.

However there is a second type of seeding mechanism where the seed site does not have the parent lattice of the target crystal. Seeding is controlled by the charge on the surface of the seed site causing localised environments which facilitate the construction of the crystal lattice. It is proposed that this seeding mechanism forms part of the mechanism for the growth of ettringite.

Having suggested a seeding mechanism for the development of the crystal, a discussion regarding the general overall solution chemistry should be made.

To do this, the solution chemistry of a PC/SAB/CaSO₄ system will be considered. Assuming that each of the components are in equal proportion, then there should be no limiting component from solution to control the growth of ettringite in the short term. When each of the individual components are introduced to water, the concentration of the various ions, Ca²⁺, SO₄^2- Al(OH)₃ can be determined by ICP-AES⁵. This gives the theoretical reservoir size of the ions and their contribution to solution concentration.

Early age expression (metod 2.17) m g/ml

Expression results at 12 hr intervals (methods 2.18 ) /ppm

Grout 2120 (Results 3.2, grout series 20)

The ions required for the formation of the ettringite crystal can bee seen to occur throughout the solution and realistically one could expect that ettringite crystals may be generated evenly through the solution. General observations of many series of grouts indicate that this is not the case. Suggesting that the concentration of ions essential to the formation of ettringite is not homogeneous but heterogeneously distributed throughout the grout in a series of micro environments. When the PC / SAB / CaSO₄ grout is examined using scanning electron microscopy of samples freeze clamped in liquid nitrogen and fractured before vacuum drying, the ettringite can be seen to be associated with the SAB cement particles or the source of the aluminium.

Figure 1 Electron micrograph of 2 hour hydrate produced by grout 2120^{(Results 3.2, grout series 20)}

Crystals can also be seen to be radiating from a central point in the upper part of the electron micrograph (A). It was not possible to find an angle to take an elemental analysis of this particle as the ettringite crystals completely covered the surface of the particle. However, ordinary Portland cement (B) and Anhydrite (C) particles were identified and appeared to be relatively clear of crystals. It is suggested here, that a calcium sulpho-aluminate cement particle is at the centre of the crystal mass A.

Solution experiments confirm that during early hydration the aluminium reservoir is rapidly depleted. There is now a condition of heterogeneous distribution of the elements within the grout matrix.     

The depletion of the aluminium ions form solution is suggested to be a key indicator for the existence of non bulk precipitation or super saturation – precipitation mechanism as the reaction would essentially stop, or become highly granular in nature with clear inner and outer hydration spheres. This behavure is not observed, in reverse the formation of ettringite continues rapidly in a lathlike conformation. Expansion is also associated with the formation of ettringite suggesting that a reduction in the density of the hydration product when compared to the starting materials. This is certainly the case for Portland cement, however, these grouts have an enormous water content, being 71% water, thus the overall density of the ettringite must be considerably lower than 1 to cause such expansion. When the grout is examined, large inter crystal voids can be seen to exist which can be many times greater than the physical dimensions of the crystal. One would normally predict that the expansion of a low density crystal would occur into these voids and a null gross dimensional change would be observed. 

The nature of the initial gel formed within these grouts is more of a fractal dam effect seen in thixotropes rather than the merging of diffuse hydration spheres. The fractal dam can be visualised simply as the void filling effect of a number of  lathlike objects randomly orientated in 3 dimensions. The volume of the random distribution is considerably greater than that of the parent solid material. This is rather like comparing a pile of randomly orientated needles with the parent steel ingot. Traditional grouts operate at much higher hydrate density and it is not possible to observe the morphology of individual crystals.

Having hypothesised that the gel of these systems occurs at the point when the individuals crystals “bridge” the gap between the cement particles and other crystals suggests that the crystals are growing during the hydration. The implications are that this is the point at which the ettringite crystal is preferentially generated in the lath like form.

Reviewing the hypothesised mechanism made so far, it is a reasonable suggestion that the crystals are growing at the point where all the ions necessary for crystal formation occur in concentration. Conspicuously aluminium is depleted form the general pore solution and is present in the SAB cement grain or close to the surface of the cement grain. Suggesting that the ettringite crystals grow out from the cement grain from a specific “point”. This “point” would suggest that the seed site occurs on the surface of the cement grain.

Lithium ions are regularly used as accelerators for aluminate cements. Lithium  aluminates are insoluble and it is hypothesised that these form on the surface aluminium source facilitating the formation of ettringite from the cement grains.

One further consideration must be made before a detailed examination for the formation ettringite can be undertaken. The dissolution of aluminium into solution is considerably greater than the rate of aluminium consumption by ettringite. The excess aluminium would be expected to dissolve into the pore solution and diffuse away form the source. It is hypothesised that the aluminium dissolution rate is modified by the formation of unstable, sparingly soluble aluminium hydroxide polymers which are pH sensitive. These polymers would effectively act as a short term reservoir of aluminium, effectively removing the aluminium from solution. The slow release of the aluminium hydroxide back into solution would allow a constant but small development of ettringite over time. This can be seen from the thermal profile of a PC/SAB/CaSO₄ hydrating over time. The initial exotherm tapers off to a small but significant level for a period of several days, finally coming decaying almost 76 hours after the initial hydration. 

The levels of other components required for ettringite can be seen also continue to fall during this time. 

This mechanism also predicts that grouts with this type of hydrate matrix will expand even though the matrix consists of fine crystals surrounded by voids. As the crystals elongate they mechanically force apart and solid materials between which the crystals are growing.

The hypothetical seeding mechanism for the formation of ettringite predicts that the seed is based on the charge required to form hydroxide bridges to the aluminium ions in solution. It also predicts that the backbone elongates from the surface of the lithium aluminate by sequential insertions of aluminium-calcium ion pairs.

Theoretically, the aluminium in the ion pair is hypothesised to form an OH bridge with the lithium ion. The calcium from the ion pair then forms an OH bridge to the last aluminium in the back bone. The induced charge on the backbone aluminium is slightly reduced attracts the oxygens from two sulphate counter ions. The calcium from the sulphate forms an OH bridge with the adjacent hydroxide already present on the six coordinate aluminium. This charge redistribution reduces the attraction for the OH-Li bridge which then breaks. This effectively elongates the ettringite backbone by a Ca-Al ion pair unit.

The formation of ettringite is hypothesised to be a continual process, the rate being controlled by the rate at which counter ions can be coordinated within the formation cycle. A prediction of this mechanism is that the inclusion of different counter ions will effect the rate of formation of ettringite or other members of the solid solution series.

Another prediction of the mechanism is that single crystals may vary in composition in sequential bands along the long axis of the crystals depending on the components available at the time of formation. There is some evidence for this from sewer linings where some crystals have a banded composition.

The hypothetical mechanism essentially suggests that the formation of ettringite is independent of the saturation of the pore solution. This is observed in practice where ettringite can be formed in grouts containing enormous water contents.  

Further the hypothesis suggests that the formation of ettringite will be susceptible to variation in the pH. Should the pH be too low, then the OH environment would be incorrect for this model, and should the pH be too high then the formation mechanism would fail due to the solubility of the aluminium hydroxide within strongly alkaline solutions. In this case the ettringite would occur as a non-lathlike ettringite or as has been described in earlier published work, spherical ettringite, closely associated with the aluminium source.  

A further prediction of the hypothesis regarding the nature of the seed site refers to the acceleration effect of fluorides. Soluble fluorine salts can be seen to show considerable acceleration effects in mine pack grouts. Here the hypothesis suggests that the fluorine ion is not the ion which is directly associated with the X-OH-Al bridge. However, fluorine forms insoluble calcium fluorides, and the induced charge on the calcium due to the proximity of the fluorine facilitates the formation of Ca-OH-Al bridges in a similar manner to the Li-OH-Al bridges in the hypothesised mechanism.  

Refs