4. Future Trends

To discuss the future of CMSC is an exercise in speculation. The most we can do is to extrapolate some trend lines, and state where the field seems to be going and what the field needs to be doing. Whether the field ever gets there, or gets somewhere else altogether different, depends on future trends in computing power, available researchers, funding opportunities, and interest by industry. At present, there is still only a limited number of researchers in this field. However, this situation seems to be changing, as this field becomes more accepted, both by academic and by industrial users, and as computer power becomes more easily available. NIST is also dedicated to making computational tools accessible on the Internet as they become available and as we have manpower to document them.

One item that should remain fairly constant is the rough divide that exists in the field of CMSC between those researchers who think in terms of "modelling" and those who think in terms of "theory." These terms are overly broad, but they are meant to point out the difference between those who wish to replace experiments with modelling and those who wish to explain experiments with computer-based theory.

We think that the primary function of CMSC is to explain and suggest experiments, much like the use of theory in physics. As was stated above, the relationships between microstructure and properties for concrete are too complex for analytical methods to be very useful. Computational materials science provides the theory that is needed to understand current experiments and help plan new experiments. However, once the validity of a computer model is unequivocally demonstrated, one can then progress to replacing some experiments with models.

Those who want to replace experiments with modelling wish to save time and money by replacing expensive experiments with cheap (once the cost of development is past!) computer simulations. If the materials science of some aspect of concrete is well-understood, and the models developed give results that have been well-validated by careful experiments, then at least some of the experimental work involved can indeed be replaced by computer modelling. At NIST, we are in the process of developing an industrial consortium called the Virtual Cement and Concrete Testing Laboratory [31], which is intended to replace many standard tests on cement and concrete with well-validated computer models. At least some of the models developed so far in CMSC are at this stage. This will hopefully speed up the R&D process to develop new cement-based materials. We invite cement and concrete companies to join with us in this development. The future of this activity will also depend on advances in computer power, both at the research end, so that models can do more, and also at the user end, so that the average user can take better advantage of the ability of models to replace experiments.

As for the future of CMSC in explaining experiments, this will depend on both computer processing speed and memory improvements, although perhaps more on processing speed improvement. The memories available on the large shared-memory clusters (32 or more processors, 1 gigabyte per processor) are large enough to do many computations that were impossible in the past. In general, however, doing large computations, involving 10 or more gigabytes, is still too slow. Parallel processing will help, but often many processors are not available at one time. Also, the current models that are used the most – cement hydration and finite element electric and elastic codes – have not been optimized for parallel processing. They can be, but this has not been done to date. Intrinsic processor speed must also increase greatly to allow many problems to be solved.

The following is a partial list of the problems that will challenge CMSC in the next decade or so.

• The viscoelasticity of concrete is a difficult problem, but very important for understanding real concrete mechanical properties, both at early and later ages. Regular linear viscoelasticity would be hard enough, but concrete is a non-linear, aging viscoelastic composite. The aging is due both to changes in the amount of constituents over time due to hydration, and changes within the phases themselves (aging of C-S-H). Coupling moisture flow to viscoelasticity is also a difficult and important problem. The correct modelling of viscoelasticity of course affects all other properties and processes that involve mechanical stresses.

• Simulating the rheology of concrete suspensions will also become an active area, making use of modern developments in computational fluid dynamics. Among the parameters that need to be studied are the effect of the size distribution and shape of the aggregates, and the effect of chemical admixtures.

• Developing a predictive model for the nanostructure of C-S-H will be a focus of much effort in the next decade. Structural information at the atomic level from NMR and molecular dynamic modelling [32, 33] will need to be coupled in to achieve a truly predictive model of C-S-H. This will need also to be able to understand and predict the change of structure and properties of C-S-H with time.

• The hydration of blended cements and cements with mineral admixtures, along with the effect on properties, is an important problem that will definitely be worked on computationally over the next decade. We believe that the future of the cement and concrete industry is closely tied to the future of blended cements.

• The challenge of correctly modelling transport properties like chloride diffusivity, fluid permeability, and sorptivity, especially in unsaturated concrete, will not go away for some years to come. Here is a case where better experiments, which are beginning to be made, drive the development of better (and more complicated) models.

• Nearly all performance properties of concrete change as chemical and physical degradation processes take place. Computational materials science, in conjunction with careful experiments, should be able to analyze and predict these changes. This will be crucial for a proper understanding of the service life of concrete, which is determined by the coupling of properties to environment factors.

• The simulation of the degradation processes occurring in concrete, with application to the prediction of service life, is a case where models should replace experiments, in order to compress 50-100 years of service life into one computer run. Even longer times are necessary for concrete used for the long-term containment of nuclear and hazardous waste. A special challenge is the growth of cracks and deterioration of mechanical properties during the degradation.

• The prediction of strength, either tensile or compressive, which involves the application of fracture mechanics to crack growth processes, is a challenge that will occupy CMSC for some time to come.

• Finally, temperature affects hydration mechanisms, microstructure, and degradation processes. A successful CMSC must be able to take the role of temperature quantitatively into account.

All of these areas will demand new algorithmic developments, as well as increased hardware speed and capacity. People who combine skills in basic materials science and programming will be essential in new algorithmic development.

We want to re-emphasize the important need for close cooperation between experimental and computational materials science. This is how important advancements will be made in concrete science and technology in the future. One example of what can be accomplished via this kind of cooperation is in the work on the electrical properties of plain and fiber-reinforced cement paste and mortar, and the effect of electrode shape and roughness, which has been carried out collaboratively between Northwestern University and NIST [34-46]. Close cooperation means experimentalists designing experiments to make the job of modelling easier, and modellers forcing themselves to model real systems, not just simple systems. In this way, the models can help explain the experiments, and the experiments can show the shortcomings of the model.

In spite of the many problems that remain to be solved, which will keep this field scientifically fruitful for years to come, we also expect to see important applications of CMSC to concrete practice within the next decade. We invite those interested in this important field to collaborate with us to advance CMSC.