Chloride-induced corrosion of reinforcement is the most common cause of premature deterioration in concrete structures exposed to seawater or de-icing salts (Neville, 1995). The usual way of describing it follows Tuutti (1982): an initiation period, during which chloride moves through the cover until its concentration at the steel exceeds a threshold, and a propagation period, during which the steel corrodes. Most of the design effort goes into lengthening the first period, and the most effective single measure is the choice of binder (Bertolini et al., 2013).

Limestone calcined clay cement, usually written LC3, is a ternary binder in which clinker is partly replaced by calcined kaolinitic clay and limestone. In the common formulation known as LC3-50 the clinker content is about half of the binder, with roughly 30 per cent calcined clay, 15 per cent limestone and the remainder gypsum (Scrivener et al., 2018). The idea of combining metakaolin and limestone so that the two react together was established by Antoni et al. (2012), and the binder has since attracted attention because suitable clays are abundant where fly ash and slag are not (Sharma et al., 2021).

Laboratory work on saturated specimens has consistently reported low chloride diffusion and migration coefficients for LC3 concretes. The exposure classes in which chloride attack is most severe are not saturated ones, though. The European concrete standard reserves its highest chloride classes for tidal, splash and spray zones and for surfaces subject to cyclic wet and dry conditions (CEN, 2013). This review asks what the published evidence allows one to say about LC3 concrete under that kind of exposure, and where the evidence stops.

The article is a narrative review of the published literature. It does not report new measurements, and it makes no claim to have searched the literature exhaustively. Sources were chosen to cover three bodies of work: the established understanding of chloride transport in partially saturated concrete, the studies that characterise hydration, pore structure and chloride behaviour of LC3 and closely related metakaolin and limestone blends, and the test standards on which reported coefficients depend. Where the literature on LC3 itself is thin, as it is for long cyclic exposure, the review says so and reasons from the properties that are documented instead of filling the gap with conjecture.

In permanently submerged concrete the pores are full and chloride moves by diffusion down its concentration gradient, a process first treated with Fick's second law by Collepardi et al. (1972). When the surface dries, the outer pores empty. On the next wetting, salt water is drawn in by capillary suction, which is far faster than diffusion, and carries chloride with it to the depth the water front reaches. During the following dry period water evaporates and the salt stays behind. Repetition raises the chloride concentration in the outer layer well above that of the wetting solution, and the profile often shows a maximum some millimetres below the surface instead of at it.

Figure 1 sets out the zones that result. The outer layer in which moisture content fluctuates is generally called the convection zone. Behind it the concrete stays close to saturation and transport is again diffusive, but it is driven by the elevated concentration at the inner edge of the convection zone. Saetta et al. (1993) gave an early numerical treatment of coupled moisture and chloride movement in partially saturated concrete, and the fib service life model handles splash-zone exposure by applying the diffusion solution from the inner boundary of a convection zone of specified depth (fib, 2006).

The balance between wetting and drying matters a great deal. In the cyclic ponding experiments of Hong and Hooton (1999), longer drying periods led to deeper chloride penetration, because a drier surface layer absorbs more solution on rewetting, and penetration grew with the number of cycles. Polder and Peelen (2002) followed concretes with different binders through salt wetting and drying using electrical resistivity, and found that blended cements with high resistivity showed both less chloride penetration and lower corrosion activity. Two material properties therefore govern cyclic performance in addition to the diffusion coefficient: the rate of capillary absorption, and the depth to which the concrete dries in the time available.

Four properties of LC3 are relevant: how fine and how connected its pores are, how much chloride its hydrates can bind, what its pore solution contains, and how it responds to carbon dioxide. The first three are taken in turn here and the fourth in the next section.

Drying opens the outer pores to carbon dioxide, and carbonation proceeds fastest at intermediate humidity. Binders with little portlandite have less buffer against it, and the lower carbonation resistance of high-replacement blends, LC3 included, is acknowledged in the literature on the binder (Sharma et al., 2021). Under purely atmospheric exposure this is a matter of cover depth and curing. Under salt wetting and drying it interacts with chloride. Friedel's salt is not stable in carbonated paste, so carbonation of the outer cover releases chloride that had been bound, and the fall in pH lowers the amount the silicate hydrate can hold (Yuan et al., 2009). The released chloride is then free to be carried inward on the next wetting.

Whether this combined action is significant in practice depends on how deep the carbonation front travels between wettings. In a splash zone that is wetted daily the surface seldom dries far enough for much carbonation, and the fine pore structure of LC3 works in its favour. On a bridge deck or parking structure that receives de-icing salt in winter and dries for months in summer, the conditions for combined attack are much better met. The published evidence on LC3 does not yet settle the question, and it is the point at which extrapolation from saturated tests is least safe.

Almost all published chloride data for LC3 come from a small group of standard tests, which are summarised in Table 1. None of them reproduces cyclic exposure. The rapid chloride permeability test measures charge passed and is sensitive to the conductivity of the pore solution, so it tends to flatter binders with low alkalinity. Migration testing after NT BUILD 492 gives a coefficient from the penetration depth under an applied voltage, with no contribution from absorption. Bulk diffusion tests are closest to natural transport but are run on saturated specimens.

Cyclic regimes in the literature differ in the length of the wet and dry periods, the drying temperature and humidity, the salt concentration and the age at first exposure. Each of these changes the result, as Hong and Hooton (1999) demonstrated for the drying period alone. Drying at elevated temperature shortens a test but can alter the hydrate assemblage of a binder that contains carboaluminates, and it accelerates carbonation unless carbon dioxide is excluded. A comparison between LC3 and a reference concrete under one regime therefore says little about their ranking under another, and still less about a real structure.

Three gaps stand out. The first is the absence of an agreed cyclic test. A regime that fixes the wet and dry durations, the drying climate and the atmosphere, and that reports absorption depth and carbonation depth alongside the chloride profile, would let results from different laboratories be compared. The second is field data. LC3 has been produced at industrial scale only recently, and exposure-site results over periods comparable with a design life do not yet exist. Service life estimates rest on short-term coefficients and on assumptions about how those coefficients age.

The third gap concerns the threshold. Service life models multiply the uncertainty in the diffusion coefficient by the uncertainty in the critical chloride content, and for low-alkalinity binders the second is the larger. Threshold measurements on LC3 under partially carbonated and cyclically wetted conditions would be of more practical value than further migration coefficients. Models that couple moisture, chloride and carbonation in the manner begun by Saetta et al. (1993) exist, but they need binder-specific sorption and binding data that have not been published for LC3 in any systematic form.

The published evidence supports a qualified confidence in LC3 concrete under cyclic chloride exposure. Its fine pore structure reduces capillary absorption and diffusion, the two transport processes that operate under wetting and drying, and the reduction is large enough to outweigh a lower chloride threshold in the service life calculations published so far. The reservations concern chemistry. The binder carbonates more readily than Portland cement, carbonation releases bound chloride, and exposures with long dry periods bring the two together. Until an agreed cyclic test and long-term exposure data are available, designers would be prudent to pair LC3 with thorough curing and unreduced cover in the most severe chloride classes.

To be declared by the author(s) before publication.

To be declared by the author(s) before publication.