While high performance concrete (HPC) outperforms conventional concrete in nearly every performance category, one Achilles heel is its performance when exposed to a fire. Sporadically, such as during the recent fire in the Channel Tunnel , HPC fails rapidly and dramatically due to the explosive spalling of the concrete's surface layer. This failure mode is typically not observed in conventional concretes of different mixture proportions (higher w/c ratio and greater volume fraction of fine aggregate) and whose binder component is based solely on portland cement . The mechanism of failure for HPC during a fire is not yet well understood. One possibility is that the spalling is due to the buildup of strain energy in the specimen due to thermal incompatibilities between the cement paste and aggregates . During exposure to a fire, the aggregate expands, while after an initial expansion, the cement paste actually contracts due to the loss of moisture and the generation of (drying) shrinkage type stresses . In HPC, the interfacial transition zones (ITZs) between aggregate and cement paste are much denser and could result in a higher stress concentration in the ITZ regions at elevated temperatures then in a conventional concrete where the more porous ITZ region may act as a sort of (thermal) shock absorber due to its higher porosity. While this mechanism is certainly plausible, in itself, it is somewhat difficult to reconcile with the fact that the addition of about 0.2 % by volume of polypropylene fibers is able to significantly improve the fire resistance of HPC . Addition of the fibers at this low volume fraction should not mitigate the thermal incompatibility problem. Furthermore, the saturation state of the concrete has been shown to effect its fire performance , which would be unexpected if thermal incompatibility were the sole mechanism responsible for failure.
A second hypothesis concerning the failure mechanism is that the explosive spalling is due to the buildup of very high pore pressures within the HPC, due to the liquid-vapor transition of the capillary pore water as well as that bound in the cement paste component of the concrete. An excellent review of this "moisture clog spalling" process can be found in Ref. . A large portion of this water is released between 100 ºC and 250 ºC when the calcium silicate hydrate gel ( C-S-H) begins to degrade. This release is compounded in typical high performance concretes due to both their higher cement factor and the presence of silica fume which produces pozzolanic C-S-H gel from the calcium hydroxide formed during hydration. A w/c=0.5 cement paste with 10 % silica fume can release about 50 % more water in this temperature range than a reference paste with no silica fume . If this water vapor can not escape from the specimen, significant pressures will develop and may eventually cause spalling of the concrete. If the pressure buildup follows that of saturated water vapor, for example, pressures of 0.5 MPa, 1.5 MPa, and 3.9 MPa will be generated at temperatures of 150 ºC, 200 ºC, and 250 ºC, respectively . Recently, pore pressures on the order of 3 MPa have been measured in saturated cement mortars subjected to radiant heating .
Thus, in this failure scenario, the permeability of the concrete is one critical parameter (others being the saturation state of the concrete and the heating rate) as it will regulate the rate at which the generated saturated vapor can escape from within the interior of the concrete specimen. The permeability of conventional concretes is one to two orders of magnitude higher than that of their component cement pastes . Simulation studies and experimental evaluation using mercury intrusion porosimetry have indicated that this permeability increase is likely due to the percolation of the porous ITZs surrounding each aggregate particle  and air void . Further evidence for this percolation of ITZs in conventional concretes has been provided by the Wood's metal intrusion and subsequent scanning electron microscopy evaluations of concrete by Scrivener and Nemati , who suggested that the ITZ regions were indeed percolated and estimated an ITZ thickness on the order of 20 µm in ordinary concrete, consistent with the value suggested by the concrete microstructural model employed by Winslow et al. .
In HPC and even in conventional w/c (0.45 or so) concretes containing silica fume, the thickness of the ITZ is reduced to about 10 µm or less [13,14,15,16,17,18]. In these studies, the thickness of the ITZ has been determined based on measurements of either porosity distributions using SEM analysis [14,15,17] or the orientation index of the calcium hydroxide in the ITZ and bulk paste [16,18], for w/c ratios ranging from 0.23 to 0.5. For a given ITZ thickness, the volume fraction of ITZ paste is mainly dependent on the surface area of the aggregates. Thus, depending on the specific gradation and volume fraction of aggregates employed in a concrete, the ITZ regions in an HPC may or may not be percolated. This percolation concept, which will subsequently be illustrated more clearly, could thus explain the inconsistency of the performance of HPC under fire testing (sometimes explosive spalling is observed  and other times not ). When the ITZs are themselves depercolated, the addition of polypropylene fibers (which also vaporize during the fire exposure) could perhaps provide pathways between locally percolated ITZ clusters to allow for the escape of water vapor before a significant pressure buildup produces spalling behavior. In support of this, Toutanji et al. have measured increases of up to a factor of nearly 3 in the rapid chloride permeability of concrete containing 0.3 % fibers on a volume basis . Furthermore, Alonso et al. have measured an increase in the room temperature gas permeability of more than three orders of magnitude for an ultra-high performance concrete containing fibers when first heated to 300 ºC relative to one heated to only 200 ºC (presumably due to the disappearance of the polypropylene fibers) . In this paper, a 3-D fiber-reinforced concrete microstructure model is presented and applied to examining the percolation of the ITZ regions in conventional and high performance concretes with and without polypropylene fibers.