As hydration of cement is a water-consuming process, in absence of external water supply, evolution of hydration desaturates the cement paste: this phenomenon is called self-desiccation. Like desiccation by drying, the self-desiccation also decreases the relative humidity in the pore network of cement paste. We are interested in this decrease of relative humidity, as it leads to the apparition of capillary forces and induces autogenous shrinkage, even in the absence of drying. We aim in this work at predicting the long-term kinetics of autogenous shrinkage due to this self-desiccation.

In the first part, we performed a bibliographical study and found out all experimental data that provide evolutions of relative humidity with respect to time under autogenous condition, i.e., in absence of any water exchange between the sample and its environment. Since hydration stops below a certain relative humidity, we expect that the relative humidity reaches an equilibrium value in the long term. We fitted this equilibrium value for all available data. By plotting the equilibrium value of relative humidity as a function of water-to-cement ratio, we obtained a range of values toward which relative humidity must drop due to self-desiccation.

The second part is dedicated to estimate long-term kinetics of autogenous shrinkage. We considered concrete as a multi-scale material. Knowing the creep modulus of calcium-silicate-hydrate gel, we computed creep modulus of cement paste (excluding all aggregates from concrete) by two steps of upscaling. We computed then the capillary force due to self-desiccation as the product of capillary pressure and saturation degree, using Kelvin's law and Power's model. In the end, we predicted the long-term kinetics of autogenous shrinkage of cement paste as that of creep under the action of this capillary force. Comparing the predicted kinetics of autogenous shrinkage with the results of experimental data that are available in literature, we found the prediction satisfactory.