Cementitious materials In this way, it is guaranteed that the material microstructure remains virtually unaltered during each individual creep test, while subsequent creep tests refer to dierent microstructures. In order to minimize possible material damage, the compressive loads are restricted to at most 15% of the uniaxial compressive strength reached at the time of testing. The loading protocol consists of quasiinstantaneous compressive loading and unloading steps as well as a three minutes long holding period in between. As for experiments on cement pastes, three dierent material compositions are investigated, de ned in terms of initial watertocement mass ratios amounting to 0.42, 0.45, and 0.50, respectively. Precise representation of the measured compliances by means of a powerlaw expression including elastic and creep moduli, as well as a creep exponent, while requiring the elastic and creep strains to be compressive at all times, yields concavely increasing time evolutions of elastic and creep moduli, as well as slightly decreasing or quasiconstant evolutions of the creep exponent. Combination of these results with calorimetrybased evolutions of the degree of hydration yields linear elasticityhydration degree and overlinear creep modulushydration degree relations, while the creep exponents slightly decrease with ongoing hydration. Notably, the herein quasistatically determined elastic moduli agree very well with those determined ultrasonically on the same cement pastes. This impressively underlines the fundamental characteristics of the elastic properties being related to an energy potential, independently of loading paths and corresponding strain rates. Conclusively, Young's moduli which are either determined from loading or unloading paths only, may not exclusively refer to elastic material behavior, but also to dissipative phenomena. The measured creep properties of cement pastes result from the viscoelastic behavior of the hydration products. We here identify a corresponding single isochoric creep function characterizing wellsaturated Portland cement hydrates, through downscaling of 500 dierent nonaging creep functions obtained from vi the aforementioned three minutelong tests on dierently old cement pastes with three dierent initial watertocement mass ratios. A twoscale micromechanics representation of cement paste is used for downscaling. At a scale of 700 microns, spherical clinker inclusions are embedded in a continuous hydrate foam matrix. The latter is resolved, at the smaller scale of 20 microns, as a highly disordered arrangement of isotropically oriented hydrate needles, which are interacting with spherical water and air pores. Homogenization of viscoelastic properties is based on the correspondence principle, involving transformation of the timedependent multiscale problem to LaplaceCarson space, followed by quasielastic upscaling and numerical backtransformation. With water, air, and clinker behaving elastically according to well accepted published data, the hydrates indeed show one single power lawtype creep behavior with a creep exponent being surprisingly close to those found for the dierent cement pastes tested. The general validity of the identi ed hydrate creep properties is further corroborated by using them for predicting the creep performance of a 30 years old cement paste in a 30 days long creep test: the respective model predictions agree very well with results from creep experiments published in the open literature. Focusing nally on predicting the mechanical properties of mortars and concretes, it is important to note that customary micromechanics models for the poroelasticity, creep, and strength of concrete restrict the domain aected by the hydration reaction, to the cement paste volume; considering the latter as thermodynamically closed system with respect to the chemically inert aggregates. Accordingly, such micromechanical models typically rely on the famous Powers hydration model, in order to quantify volume fractions of clinker, cement, water, and aggregates, as functions of the hydration degree. The situation changes once internal curing occurs, i.e. once part of the present water is absorbed initially by the aggregates, and then soaked \back" to the cement paste during the hydration reaction. For this case, we here develop an extended hydration model, introducing water uptake capacity of the aggregates on the one hand, and paste void lling extent on the other, as additional quantities. Based on constant values for just these two new quantities, and on experimentally determined creep properties of cement pastes as functions of an eective watertocement mass ratio (i.e. that associated to the cement paste domain, rather than to the entire concrete volume), a series of three minute creep tests on dierent mortars and concretes can indeed be very satisfactorily predicted by a standard microviscoelastic twoscale model. This further extends the applicability range of micromechanics modeling in cement and concrete research, and it concludes the present thesis which combines innovative macroscopic material testing and stateoftheart multiscale modeling from submicrometric hydrate needles to decimetersized specimens of mortars and concretes.
