Concrete is a complex random composite material. In addition, concrete is also a multi-length scale material. From the nanometer to the millimeter scale, concrete is a different random composite at each length scale [1]. Concrete is not only different at each length scale, it is also an interactive composite, where the amount and properties of one phase (aggregate volume and surface area) affects the properties of other phases (bulk and interfacial transition zone cement paste) [2]. Percolation processes [3-5] and composite interactions [6-8] play key roles in the performance of concrete, and help to explain the overall dependence of transport properties like ionic diffusivity [9] and fluid permeability [10] on the microstructure.
Because of this randomness and complexity, analytical methods for quantitatively relating microstructure and properties of concrete are generally ruled out from the start. That is not to say that analytical equations are not useful, but they can only be used by "smearing" out some aspect of the complex random microstructure. Therefore, they cannot be used to directly relate microstructural details to properties. By default, sophisticated computer models are necessary to describe the microstructure and transport properties.
But having sophisticated computer models is not enough. As will be seen below, the computational materials science of concrete (CMSC) has been intimately wrapped up with the development of computer processing speed and the growth of computer memory over the last 15 years. For further advances to be made, both the algorithm development and the hardware necessary for proper implementation of the algorithm must be available simultaneously. Computer hardware development goes on independently of concrete research. But the development of models for predicting the microstructure and properties of concrete is dependent on having researchers available to carry out the work.
What we mean by the term "computational materials science of concrete" is this: computer-based models of microstructure at the relevant length scale, operated on by algorithms that give accurate measures of physical properties. The microstructure models ideally do incorporate much of the basic physics and chemistry of the microstructure formation process. However, they do not have to, as long as the eventual microstructure reflects reality. This working definition is similar to the definition of "fundamental models" that we have previously contributed [11]. We should therefore mention that this short paper is not in any sense a comprehensive review of the field, but rather is an informal, personal look at this field in terms of where it has come from, where it is at currently, and where it might be going, with a focus on work we have done ourselves and with others.
Since this is a symposium to honor Francis Young on his retirement, we do want to mention that his encouragement and advice over the last 11 years have been invaluable. At the start of ACBM in the beginning of 1989, CMSC was just in its infancy. The support of ACBM, which was in no small part due to Francis´ personal support of the work, played a large role in our part of the development of CMSC.