Cement and Concrete Research 42 (2012) 1534–1548 Contents lists available at SciVerse ScienceDirect Cement and Concrete Research journal homepage: http://ees.elsevier.com/CEMCON/default.asp Rietveld refinement of the structures of 1.0 C-S-H and 1.5 C-S-H Francesco Battocchio a, Paulo J.M. Monteiro b,⁎, Hans-Rudolf Wenk a a Department of Earth and Planetary Sciences, University of California, Berkeley, CA 94720, USA b Department of Civil and Environmental Engineering, University of California, Berkeley, CA 94720, USA article info abstract Article history: Low-Q region Rietveld analyses were performed on C-S-H synchrotron XRD patterns, using the software MAUD. Received 1 November 2010 Two different crystal structures of tobermorite 11 Å were used as a starting model: monoclinic ordered Merlino Accepted 26 July 2012 tobermorite, and orthorhombic disordered Hamid tobermorite. Structural modifications were required to adapt the structures to the chemical composition and the different interlayer spacing of the C-S-H samples. Refinement Keywords: of atomic positions was done by using special constraints called fragments that maintain interatomic distances and Calcium-silicate-hydrate (C-S-H) (B) orientations within atomic polyhedra. Anisotropic crystallite size refinement showed that C-S-H has a nanocrys- X-ray diffraction (B) Crystal size (B) talline disordered structure with a preferred direction of elongation of the nanocrystallites in the plane of the Ca Crystal structure (B) interlayer. The quality of the fit showed that the monoclinic structure gives a more adequate representation of C-S-H, whereas the disordered orthorhombic structure can be considered a more realistic model if the lack of long-range order of the silica chain along the c-direction is assumed. © 2012 Elsevier Ltd. All rights reserved. 1. Introduction different from these crystal structures, basically for the following reasons: Calcium silicate hydrate (C-S-H), the main binding phase in Portland cement matrix, constitutes up to 70% in weight of hardened • The Ca/Si ratio of about 1.75 is higher than that of jennite and much ordinary cement pastes. Despite the large number of studies and the higher than that of 14 Å tobermorite. vast amount of literature available on cementitious materials, the • Both tobermorite and jennite have long tetrahedral chains, whereas atomic scale structure of C-S-H is still partly unknown owing to its in C-S-H the chains have lengths of 2, 5, 8, … (3n−1) tetrahedra. high complexity. The optimization of strength and durability of cement This pattern results from the repetition of two paired tetrahedra can be obtained through adjustments of the structure of C-S-H at a connected by the bridging tetrahedron. A particular case occurs for nanometric level [1,2], but this is subject to a detailed knowledge of n=1, when only two bridging tetrahedra are present and as will be the C-S-H crystal structure. discussed later, they are referred to as dimer. Several models have been proposed for structure of C-S-H. It is • The average chain length of C-S-H increases with age: 29Si NMR recognized that it has a multilayer structure composed of calcium experiments on C3S paste cured at 25 °C show that the mean chain layers and interrupted tetrahedral chains on both sides. Various studies length after 1 day is 2.1 tetrahedra, 2.6 after 1 month, 3.3 after have indicated structural relationships to tobermorite 14 Å (Ca5Si6O17 1 year, and 4.8 after 26 years [7]. However, when the Ca/Si ratio is 9H2O–C5S6H9, plombierite) [3,4] and jennite (Ca9Si6O21 10H2O– low, dimer chains (length=2) occur even in mature paste. C9S6H10) [5,6]. Both structures contain linear silicate chains of the “dreierkette” form in which the silicate tetrahedra are arranged in In 1986, Taylor proposed a model [3] for C-S-H that consisted of a such a way as to repeat a kinked pattern after every three tetrahedra. disordered layer structure, whereby the majority of the layers were Two of the three tetrahedra share O–O edges with the central Ca–O structurally similar to those of jennite and others were related to part of the layer; these are linked together and are often referred to as 14 Å tobermorite. In both types of layer, the structures were modified ‘paired’ tetrahedra (P) (Fig. 1, I). The third tetrahedron, which shares by the omission of silicate tetrahedra. This is an effective solution because an oxygen atom at the pyramidal apex of a Ca polyhedron, connects it is possible to obtain the expected chain length and correct the Ca/Si the two paired tetrahedra and is called “bridging” (B) [2].However, ratio of C-S-H. The tobermorite-type structure was found to be more C-S-H formed by hydration of portland cement paste, is significantly suitable to describe the lower Ca/Si ratio C-S-H, whereas jennite-type structure was more suitable for the high Ca/Si ratio C-S-H. In 1992, Richardson and Groves proposed a generalized model for the nanostructure of C-S-H [8] that accounted for the chemical differences be- ⁎ Corresponding author. tween C-S-H and the structures of tobermorite and jennite. The model in- E-mail address: [email protected] (P.J.M. Monteiro). cluded the chemical neutrality of the structure by the protonation of 0008-8846/$ – see front matter © 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.cemconres.2012.07.005 F. Battocchio et al. / Cement and Concrete Research 42 (2012) 1534–1548 1535 O2− anions, resulting from deleting bridging tetrahedra. This led to silanol groups in place of the oxygen atoms shared by paired and bridging tetrahedra. In addition, different amounts of calcium polyhedra were located in place of the deleted bridging tetrahedra to respect the real Ca/Si ratio of C-S-H under consideration. Later on, Richardson suggested detailed structural representations of C-S-H, derived from crystal structure data of tobermorite, for three different levels of proton- ationforCa/Siratiosof1.0,1.25and1.5[8,9]. Recently, new understanding of the C-S-H structure has been I) achieved using a variety of methods: X-ray diffraction [10–13], total scattering methods using the pair distribution function (PDF) [14], and molecular dynamics (MD) [15,16]. The analysis of Skinner's work [14] had been performed on a sample of C-S-H 1.0 aged 4 months, which revealed that the nanostructure of synthetic calcium hydrate re- sembled the crystal structure of natural tobermorite 11 Å, as refined by Merlino et al. [17]. This conclusion was obtained by simulating the diffuse X-ray scattering due to the nanostructured features of the material by a Gaussian shape broadening of the structure factor of tobermorite 11 Å, tobermorite 14 Å, and jennite, that were supposed to resemble the nanostructure of C-S-H in real portland cement concrete. Their results show that the structure of tobermorite 11 Å is strikingly similar to that of C-S-H (I). The loss of coherent scattering on the C-S-H sample above about 3.5 nm might reflect the maximum crystallite size of the material. In addition, a Monte Carlo refinement of tobermorite 11 Å was done to obtain the best fit of the PDF to the synthetic C-S-H sample. After the refinement it is still possible to see the distinct multi- layer structure made by a stacking of calcium layers with tetrahedral chains on both sides, and calcium ions in addition to water molecules II) in the interlayer space. X-ray patterns can be analyzed in two ways: One is to determine the distribution of atom pairs. This method is mainly applied to amorphous materials with no long-range order or highly disordered structures [18]. The second method is based on the structure factor, which relates the diffraction pattern to lattice planes in the long-range ordered crystal [19]. Interestingly C-S-H is in between. In this study we will use the standard crystallographic Rietveld method [20] to further quantify the nanostructure of this material. Two different samples of C-S-H aged 4 months are used. The first is the same used by Skinner et al. [14] for C-S-H with a Ca/Si ratio of 1.0, and the second refers to C-S-H with Ca/Si of 1.5. Two crystal structures for C-S-H were used as a starting model in the Rietveld analysis: (a) the monoclinic tobermorite 11 Å refined by Merlino et al. [17] and successfully applied in Skinner's work [14], characterized by high structural order, and (b) the orthorhombic version of the tobermorite structure refined by Hamid [21] that, due to the smaller unit cell, presents a lower degree of order and therefore is more likely to capture the real structure of C-S-H. The detailed description of the two tobermorite crystal structures is out of the scope of this presentation and we refer to the papers by Merlino et al. [17] and Hamid [21] and references within them for further details. However, for clarification, we note that tobermorite is a mineral that occurs with a substantial degree of disorder in its X-ray diffraction pattern that is expressed by diffuse reflections along the c-axis. Hamid attributed this feature to the different positions and orientations that the silicate chains can take. In particular, it was pro- III) posed that a chain can be statistically displaced by b/2 and that each tetrahedron of the chain can be statistically tilted in two alternative Fig. 1. I) Schematic diagram showing dreierkette chains present in tobermorite: central positions.