Assessing long-term carbonation resistance of blended concretes from short-term natural exposure

Abstract

The use of blended cements by partially replacing ordinary Portland cement (CEM I) with supplementary cementitious materials (SCMs) is a major approach for mitigating the massive CO2 emissions from the concrete industry. Such cements usually have better engineering properties but are more vulnerable to carbonation. Most standard carbonation studies involve carbonating cementitious samples in an accelerated environment. However, such accelerated tests distort the carbonation mechanism in relation to natural carbonation, providing results that are not ideal for service life prediction modelling for cement-based materials. Demonstrating that novel blended cements are durable is a major barrier to their commercial adoption. As such, there is an urgent need to develop new approaches for rapidly assessing the durability of blended cements. This thesis aims to assess the long-term carbonation resistance of cement-based materials from short-term exposure to natural carbonation. The main methodology involves exposing blended samples to short-term natural carbonation, characterising the evolution of shallow carbonation depth through high-resolution techniques, and extrapolating to longer-term carbonation resistance. Another major focus of this research is to understand the effects of natural carbonation on the microstructure and mass transport properties of concretes containing blended cements. It is expected that the research will lead to improved service life prediction models and facilitate the development of low-carbon sustainable cementitious materials. Confocal Raman microscopy (CRM) was explored as a novel method for high-resolution characterisation of shallow carbonation depth after relatively short-term natural exposure. To enable this, a parametric optimisation of CRM for spectral mapping of cement-based materials was performed. The effects of sample preparation and various operating parameters on data quality were examined. Microstructural changes and damage induced by prolonged laser irradiation were also investigated. Recommendations on optimal settings were made based on the obtained results. The feasibility of CRM for monitoring early surface carbonation of hardened cement pastes was studied in real-time for up to 7 d. Samples were exposed to natural carbonation (440 ppm CO2) and accelerated carbonation (4% CO2), and the evolution of calcium carbonate (CaCO3) polymorphs, portlandite, ettringite, C-S-H gel and unreacted cement particles was followed. Results demonstrated that CRM is a valuable tool for non-destructive real-time imaging of surface carbonation in cement-based materials. The reliability of CRM for determining carbonation depth was investigated by comparing with the conventional phenolphthalein spray testing and various profile-grinding analysis methods such as thermogravimetric analysis, X-ray diffraction, Fourier transform infrared spectroscopy, and Raman spectroscopy. Results showed that the carbonation fronts observed with CRM were distinct and matched well with those observed with the phenolphthalein method under accelerated carbonation conditions. However, the difference in the measured carbonation depths by the two methods was increased after short-term (1-3 months) natural carbonation due to the ambiguous colouration of phenolphthalein when measuring shallow carbonation depth in some samples. Furthermore, profile-grinding analysis measures only the ‘apparent’ maximum carbonation depth, which is larger than the depths measured with CRM profiles. The discrepancies were found to be related to the spatial variation of the carbonation front and sampling interval adopted. A simple method was proposed to estimate the average carbonation depth from the carbonation profile, and the results obtained for profile-grinding analysis agreed well with those determined by CRM image analysis, with the average difference being only 1.1%. Overall, this chapter demonstrates that CRM is a reliable technique for measuring the true carbonation depth of cement-based materials and is especially powerful for measuring shallow carbonation depth (e.g. < 3 mm) after short-term natural carbonation. Using CRM and phenolphthalein, the evolution of average carbonation depth in CEM I and blended cements paste, mortar and concrete samples after short-term (1-3 months) and longer-term (6-24 months) natural exposure was studied. The main variables include water/binder ratio (0.45 and 0.6), curing duration (3d and 122d), aggregate content (0% for paste, 20% and 50% for mortar, and 70% for concrete), and binder type (CEM I, CEM I + 30% fly ash and 50% slag). The effect of these variables on natural carbonation was examined, and the changes in mass transport properties and microstructure with the progress of carbonation were investigated. The derived results were employed to assist the carbonation modelling study in the next chapter. Artificial neural networks (ANNs) were applied to model the long-term carbonation depths in concrete exposed to natural exposure for up to 20 years. Approximately 3300 pieces of data records were compiled from 33 papers, including this study. Thirty-three influential variables were selected as input neurons to the ANN model. Two critical parameters relating to the architecture of ANNs were optimised to reduce the errors in the prediction. By fitting the required input variables to the trained ANN algorithm, prediction of carbonation depths can be conveniently obtained. Results show a good correlation between the predicted and measured carbonation depths in the test dataset with an average root mean squared error of 1.97 and an average percentage error of 19.2%. The prediction accuracy is much higher for larger carbonation depths and the obtained error was reduced to merely 4.3% when predicting a carbonation depth of 50 mm. Since larger carbonation depth is usually detected in blended cements due to their higher vulnerability to carbonation, ANNs are thus deemed suitable for rapidly assessing the long-term carbonation resistance of blended cements with good accuracy.