UNIVERSITY OF ABERDEEN DEPARTMENT OF CHEMISTRY

Thermodynamics of Cement Hydration

by

Thomas Matschei

Dipl.-Ing., Bauhaus University Weimar

A Thesis presented for the degree of

Doctor of Philosophy

at the University of Aberdeen

Aberdeen, 06 December 2007 Declaration

This Thesis is submitted to the University of Aberdeen for the degree of Doctor of Philosophy. It is a record of the research carried out by the author, under the supervision of Professor F.P. Glasser. It has not been submitted for any previous degree or award, and is believed to be wholly original, except where due acknowledgement is made.

Thomas Matschei

Aberdeen, December 2007

Abstract 3

Abstract

The application of thermodynamic methods to cement science is not new. About 80 years ago, Bogue wrote a series of equations describing the relationship between clinker raw meal chemical composition and the mineralogy of the finished clinker. These enabled the amounts of minerals to be calculated from a bulk chemical composition. Fundamental to the equations was a precise description of the high temperature equilibrium achieved during clinkering. Bogue admitted four oxide components into the calculation; lime, alumina, silica and ferric oxide and assumed that equilibrium was attained (or very nearly attained) during clinkering. This approach, which is, with modifications, still a widely used tool to quantify cement clinkering, was one of the main motivations of this work. Thus the overall aim of this Thesis is to provide a generic toolkit, which enables the quantification of cement hydration. The use of thermodynamic methods in cement hydration was often doubted, as the water-cement system was considered to be too complex. Furthermore metastable features occur, e.g. C-S-H, which lead to the conclusion cement hydration is a “non-equilibrium” process. Nevertheless pioneering works, by Damidot and Glasser, as well as from other groups e.g. Reardon et al. and Berner et al. prove that cement hydration follows the basic principles of physical chemistry by minimisation of the free energy of an isochemical system. Hence these studies demonstrated the usefulness of thermodynamic equilibrium models in cement hydration. However the success and the accuracy of these predictions are strongly linked to a reliable thermodynamic database, including the standard state properties of the aqueous species and the cement hydrates. Whereas the thermodynamic properties of the aqueous ions are well described in the literature, the dataset for cement hydrates is incomplete or inconsistent, or both. Thus the main goal of this Thesis was to develop a consistent thermodynamic database, which enables the assessment of the constitution of hydrated Portland cements. Because hydrated concretes are exposed to different service temperatures, data were obtained in the range ~1°C to 99°C. The database is developed for commonly-encountered cement substances including C-S-H, Ca(OH)2, selected AFm, AFt and hydrogarnet compositions as well as solid solutions. Literature data were critically assessed and completed with own experiments. The tabulated thermodynamic properties were derived by a harmonisation of the available data. The new database enables the hydrate mineralogy to be calculated from the bulk chemical composition of the system: most solid assemblages, the persistence of C-S-H and failure to nucleate siliceous hydrogarnet apart, correspond closely to equilibrium. This realisation means that hydrate assemblages can be controlled. The development of a thermodynamic approach also enables a fresh look at how mineralogical changes occur as a function of cement composition as well as in response to environmentally-conditioned reactions. According to a literature review the constitution of the AFm phase in Portland cement is very sensitive with respect to its chemical environment. Except for limited replacement of sulfate by hydroxide, AFm phases do not form solid solutions and, from the mineralogical standpoint, behave as separate phases. Therefore, in dependence of the bulk chemical composition, many hydrated cements will contain mixtures of AFm phases rather than a single AFm solid solution. Relative to previous databases, sulfate-AFm is shown to have a definite range of stability at 25°C thus removing long-standing disagreement with theory about its persistence in hydrated cement pastes. - 2- Carbonate is shown to interact strongly with AFm and displaces OH and SO4 at species activities

Abstract 4

commonly-encountered in cement systems across a broad range of temperatures ≤50°C. Many of the predicted reactions were confirmed by focussed experiments and literature studies.

Possible anion substitutions in the AFt phase were investigated. Non-ideal thermodynamic models for SO4-CO3-AFt and ettringite-thaumasite solid solutions were derived from solubility experiments. Whereas at 25°C only minor anion substitution is likely, low temperatures tend to stabilise carbonate substituted AFt phases. Possible pathways of thaumasite formation were developed. It was concluded that there is no single route of thaumasite formation, but several pathways for thaumasite formation may occur simultaneously. Limestone, mainly consisting of calcite, is a permitted additive to Portland cements up to a 5 wt.-% limit under EN 197. The final chapter, on the impact of calcite addition upon cement hydration, enables a quantitative approach to its interaction with cement phases and prediction of space filling properties of pastes. The distribution of sulfate in AFt and AFm is much affected by the presence of carbonate. In the presence of portlandite the stabilisation of carboaluminate results in changes of the amounts of both portlandite and AFt: specimen calculations are presented to quantify these changes. Calculations of the specific volume of solids as a function of calcite addition suggest that the space filling ability of the paste is optimised when the calcite content is adjusted to maximise the AFt content. Additional calculations show how sulfate and carbonate distribution are affected by temperature. Carboaluminates become increasingly unstable at elevated temperatures, ≥ 50°C, whereas carbonate substitution in AFt is favoured at low temperatures in the presence of calcite. The resulting consequences of thermal cycles on the space filling properties of hydrated cements are discussed.

Keywords: thermodynamics, thermodynamic data, modelling, cement hydration, AFm, AFt, sulfate, carbonate

Acknowledgement 5

Acknowledgement

Several people contributed to the successful completion of this Thesis, to whom I am grateful and indebted:

Professor Fred Glasser for his excellent supervision of this project during my time in Aberdeen. I am most grateful for the advice and support he gave me throughout the duration of this Thesis. I enjoyed our long-lasting, motivating and academically stimulating discussions, mainly related to cement -surely one of the most fascinating manmade materials of this world-, but also to several other aspects of “daily life”. Dr. Barbara Lothenbach, EMPA Dűbendorf, my Thesis co-supervisor, for invaluable assistance with questions about thermodynamic modelling and guidance with GEMS-PSI. Her enthusiasm contributed to the completion of the database and related applications. My industrial advisors, Dr. Ellis Gartner, Lafarge Central Research, France, and Dr. Duncan Herfort, Aalborg Portland Group, Denmark, for stimulating discussions and guidance during this work. Nanocem, a research network of European cement producers and academic institutions, for funding this work and for giving me the opportunity to present and discuss the results at several intern meetings as well as at international conferences. Special thanks to Professor Karen Scrivener, representing members of Nanocem and the Nanocem steering-committee, for valuable discussions and helpful critics during the preparation of publications related to this Thesis. I would like to thank Marie-Alix Dalang-Secrétan for her help with administration throughout this project and for assistance with the preparation of the Workshop “Thermodynamic Modelling”. Dr. Dmitrii Kulik, PSI, Switzerland, for troubleshooting and assistance with GEMS-PSI and for helpful advice during the preparation of the thermodynamic database. In that respect I would also like to thank Dr. John Gisby, NPL, UK, for his comments. The staff of the Chemistry Department, University of Aberdeen, for technical assistance. I would like to thank Professor Jőrg Feldmann and his TESLA-team for invaluable guidance and introduction in “analytical methods for civil engineers” as well as for giving me the opportunity to participate in “various” group meetings. Thanks to Professor Donald Macphee for inspiring discussions about cement science, especially with respect to thaumasite formation. I enjoyed the refreshing discussions with the recently formed “cement-group” as well as with my colleagues from office “G 85”. The staff at EMPA Dűbendorf, for great technical and intellectual support during my stay in Switzerland. I would like to thank my examiners, Professor Denis Damidot, Ecole des Mines de Douai, France, and Professor Donald Macphee, University of Aberdeen, for critically reviewing this work.

Finally, special thanks go to

Kristina,… for her infinite patience with this “cement guy”

my friends, whose support I appreciated throughout the years ein besonderes Dankeschőn gebűhrt meiner Familie in Deutschland, insbesondere meinen Eltern, Grosseltern, sowie allen Verwandten, die mich űber Jahre hinweg trotz mancher privater Rűckschläge uneigennűtzig mit viel Geduld und Aufmunterung unterstűtzt und einen entscheidenden Anteil am erfolgreichen Abschluss dieser Arbeit haben

.

Table of contents 7

Abstract ...... 3 Acknowledgement...... 5 1. Introduction...... 10 2. Status of database development...... 11 3. Analytical methods...... 12 3.1. Preparation of samples ...... 12 3.2. X-ray diffraction...... 12 3.3. Thermal analysis ...... 12 3.4. Microscopic examinations...... 13 3.5. Analysis of solutions ...... 13 3.5.1 Calcium, aluminium and sodium...... 13 3.5.2 Sulfate ...... 13 3.5.3 Silicon ...... 14 3.5.4 Carbon ...... 14 3.5.5 Measurement of pH...... 14 4. Development of a thermodynamic database for cement hydrates ...... 15 4.1. Synthesis of relevant cement hydrates ...... 15 4.1.1 Hydrogarnet...... 15 4.1.2 AFm phases...... 16 4.1.3 AFt phases...... 17 4.2. Solubility determinations ...... 18 4.3. Methods used to derive and manipulate thermodynamic data ...... 19 4.3.1 Software and standard databases...... 19 4.3.2 Estimation of heat capacity ...... 19 4.3.3 Solubility based estimation of standard molar thermodynamic properties ...... 21 4.3.4 Thermodynamics of solid solutions and the use of Lippmann phase diagrams ...... 23 4.4. Results ...... 27 4.4.1 Hydrogarnet...... 27 4.4.2 AFm phases...... 30 4.4.3 AFt phases...... 41 4.4.4 C-S-H ...... 44 4.5. Discussion ...... 46 4.5.1 Data accuracy ...... 46 4.5.2 Relations between equilibrium and kinetics...... 48 4.5.3 Applications of the database...... 49 4.6. Concluding remarks ...... 52

Table of contents 8

5. The AFm phase in Portland cement...... 53 5.1. Literature review ...... 53 5.1.1 General remarks on the structure and formation of AFm phases...... 53 5.1.2 Stability of AFm phases ...... 54 5.1.3 Solid solutions between AFm phases...... 56 5.1.4 Summary and conclusions from the literature...... 59 5.2. Preparation of solid solutions...... 61 5.3. Formation of AFm phases and AFm solid solutions...... 62 5.3.1 Monosulfoaluminate-hydroxy-AFm solid solutions ...... 62 5.3.2 Monosulfoaluminate-monocarboaluminate...... 73 5.3.3 Monocarboaluminate-hydroxy-AFm...... 74 5.4. Ternary phase relations between sulfate-, carbonate- and hydroxy-AFm...... 75 5.4.1 Metastable phase assemblages at 25°C ...... 75 5.4.2 Stable phase assemblages at 25°C...... 76 5.5. Discussion of the results...... 78 5.5.1 Extent of solid solution ...... 78 5.5.2 Transformation mechanisms ...... 79 5.5.3 Solubility data and its interpretation ...... 79 5.6. Conclusions ...... 80 6. The AFt phase in Portland cement...... 82 6.1. State of the art of science ...... 82 6.1.1 General remarks on the structure and composition of AFt...... 82 6.1.2 Solid solutions between AFt phases...... 83 6.1.3 The formation of thaumasite and related solid solutions...... 84 6.2. Solid solutions between sulfate- and carbonate-AFt...... 87 6.2.1 Synthesis of solid solutions and investigations of the solid phase ...... 87 6.2.2 Thermodynamic modelling of the solid solution formation...... 88 6.3. Investigations on thaumasite ...... 92 6.3.1 The need for investigations ...... 92 6.3.2 Synthesis of thaumasite...... 92 6.3.3 The temperature-dependent stability of thaumasite in aqueous solutions...... 94 6.3.4 Solid solution formation between ettringite and thaumasite ...... 100 6.4. Formation of thaumasite and ettringite solid solutions in hydrated cements ...... 104

6.4.1 Carbonate substitution in SO4-AFt at 25°C...... 104 6.4.2 Phase assemblages including thaumasite and related solid solutions...... 106 6.5. Discussion ...... 112 6.5.1 Interpretation of phase diagrams ...... 112 6.5.2 Formation of AFt solid solutions...... 112 6.5.3 Pathways of thaumasite formation ...... 114

Table of contents 9

7. The influence of limestone addition on cement hydration ...... 116 7.1. Literature review ...... 116 7.1.1 Limestone addition to Portland cement...... 116 7.1.2 Influence of temperature on the mineralogy of hydrated (limestone blended) Portland cement: 25°C and above...... 118 7.2. Implications of limestone addition to cement hydration ...... 121

7.2.1 Experiments on the role of carbonate and sulfate on C3A hydration ...... 121 7.2.2 Phase relations between AFt-AFm phases relevant to Portland cements at 25°C...... 124 7.2.3 “Reactive” vs. “filler” calcite ...... 128 7.2.4 Quantification of phases...... 131 7.2.5 Space filling by cement paste solids...... 133 7.2.6 Composition of the aqueous phase...... 135 7.3. Applicability to Portland cement systems ...... 136 7.3.1 Experimental validation of phase changes ...... 136 7.3.2 Space filling vs. engineering properties ...... 139 7.4. The role of temperature on Portland cement hydration...... 142 7.4.1 Thermally induced mineralogy changes...... 142 7.4.2 Space filling vs. temperature...... 148 7.4.3 Experimental verification of thermally induced phase changes...... 151 7.4.4 Summary ...... 160 7.5. Discussion ...... 161 7.5.1 Limitations of the methodology and of the database...... 161 7.5.2 Kinetic factors: availability of sulfate, carbonate and alumina in the course of hydration . 166 7.5.3 Volume changes ...... 168 7.5.4 Thermally induced phase changes...... 169 8. Summary and conclusions...... 171 8.1. Thermodynamic quantification of cement hydration ...... 171 8.2. Implications for cement hydration ...... 172 References...... 174 Abbreviations ...... 187 Nomenclature in cement chemistry...... 187 Abbreviations used in calculations...... 187 Other abbreviations...... 188 Figures and Tables...... 189 Figures ...... 189 Tables ...... 194 Appendix ...... 195

Introduction 10

1. Introduction

One of the unsolved problems in the application of Portland cement is to quantify the performance lifetimes of concrete constructions. The problem affects nuclear waste containments and, increasingly, long-lived infrastructure developments where quantification has failed to keep pace with the expectation of stakeholders.

Although a wealth of empirical evidence on the performance of historic concretes information on their formulation is available, emplacement and exposure history is often incomplete and, moreover, the nature of cements supplied today will almost certainly have changed since the original construction. Empirical studies and historic examples have however yielded much useful qualitative information on the aggressivity of various service environments. Numerous tests and test methods have been used as indicators of durability but they do not yield generic conclusions and their predictive capabilities are limited. As a consequence, designers of long-lived constructions have at present to rely on received wisdom, as interpreted by experts and incorporated into codes of practice.

The changing nature of cements is also of concern. Cement producers are under pressure to lower the specific energy requirements of cement production and reduce gaseous emissions. These goals are presently addressed by a combination of methods; partly by optimisation of process technology, including the use of alternative fuels and raw materials (the effect of which are beyond the scope of this study), and partly by reliance on supplementary cement materials to lessen the need for energy- rich cement. Although the use of supplementary materials is generally regarded as beneficial in terms of strength and durability as, for example highlighted by developments in cement and concrete standards, long term performance is not fully understood. Supplementary materials presently in use include industrial by-products such as slag, fly ash, silica fume, etc. as well as natural materials, e.g. ground limestone, natural pozzolanic and semi-synthetic pozzolans such as metakaolin. Each of these materials has a complex but distinctive chemistry, mineralogy and granulometry. Moreover, each type of material ranges in composition and performance. Studies of their behaviour in blended cements under controlled conditions are confined to selected compositions and short term (1-5 years) laboratory measurements, perhaps supplemented by observations on actual constructions, for which conditions may not be well controlled.

The complexity of blended cement systems and the wide-ranging nature of supplementary cementing materials have meant that guesses -sometimes well informed- have to be made at the outset about what aspects of behaviour should be studied. But the number and complexity of the resulting systems are such that results are often confined to measurement of a few of the many parameters affecting performance. Arguably the most serious question arising from the results of empirical testing is how to extend or extrapolate the results to other compositions and formulations, or to conditions other than those measured, or both. At present one cannot address these issues, except qualitatively.

If quantification of performance is to be achieved, a new paradigm is needed and a key to the development of a successful paradigm must be to concentrate on generic approaches. Thermodynamics provides a consistent framework for the analysis of complex systems. Given an adequate database to support calculations, its strength lies in its generic nature; user-defined compositions and conditions can be selected for calculation. This realisation is not new although previous attempts to apply thermodynamics have had only limited success.

Introduction 11

From an industrial point of view it could be argued that thermodynamic approaches to cement durability are too theoretical and the calculations too difficult to perform. However, the latter is no longer true: geochemists, faced by similar problems of treating complex systems, have developed and validated computer-based protocols capable of being implemented on a PC. Thus reliable protocols are available, many of which are in the public domain. Furthermore these protocols often couple to other modules, for example, enabling mass to be conserved in the course of reaction while maintaining a system of balanced equations. Selected physical properties can also be calculated, for example by enabling the specific volumes occupied by the constituent solid phases of mixtures to be tabulated.

Thermodynamics is most readily applied to isochemical systems, i.e., to systems having a constant composition, whereas many cement deterioration reactions involve transport of species into or out of the matrix (or both). But computer-based calculations also enable more complex conditions to be imposed on the system.

It is not argued that the primary output achieved by application of thermodynamic methods necessarily enables the durability and performance of cements and concretes to be quantified. But it is believed that, in the search for quantification, a sound quantitative understanding of cement paste mineralogy and of the ability to calculate features and processes arising from the interaction of cement with potentially aggressive agents introduced from the environment, with additional possibilities for calculating physical functions and the introduction of kinetic variables, constitutes a great step forward. Other necessary links to develop integrated models of cement performance will be anticipated in the discussion.

2. Status of database development

After years of development, Babushkin et al. published the first reasonably comprehensive compilation of thermodynamic data for cement substances. Their book [12] also gives numerous application examples but, as these are pre-computer, the examples selected for calculation tend to be rather simplistic and at first sight, do not afford significant advance over empirical conclusions. However a serious problem is that referenced data in this compilation have proved difficult if not impossible to trace to source.

Other databases adding to our knowledge of cement substances have been produced subsequently, e.g., by Atkins et al. [9][10][21], Damidot [44], Reardon [158][159], Lothenbach and Winnefeld [125] by the Lawrence Livermore National Laboratory [38] and by ANDRA, the French National Agency for Radioactive Waste Management [30]. Studies of phase equilibria, cited subsequently in the text, have also added data on the thermodynamic properties of specific substances.

Experimental studies have shown that temperature is an important parameter with respect to the formation of phase assemblages in the course of cement hydration. Pioneering calculations by Damidot et al. [43][44][45][46][47][48][49] disclosed that thermodynamics enables to understand phase transitions due to temperature changes as well as due to changes of the chemical composition. Returning to the present state of development of performance-based thermodynamic models, calculation is handicapped by lack of a consistent database applicable to a wide range of temperature and cement compositions. Thus the aim of this work was to make progress towards the development of a database which is sufficiently reliable and inclusive to sustain and support calculations as well as experiments.

Analytical methods 12

3. Analytical methods

A programme of the acquisition of thermodynamic data was undertaken. This involved synthesis, characterisation, analysis and data integration. New data were obtained through synthesis of phase pure substances with subsequent solubility measurements. Focussed experiments were designed to substantiate calculated results in the course of application of the database.

3.1. Preparation of samples

Cement hydrates are generally very sensitive to decomposition by carbonation. Therefore many syntheses and solubility experiments must be performed under N2-atmosphere to minimise access of atmospheric CO2. Generally, if not stated otherwise, the synthesised solids, aged and stored in inert HDPE or PTFE ware, have been vacuum-filtered with Whatman #540 filter paper and washed several times with ultra pure degassed water to remove alkalis, if present. Subsequently the solids were dried over saturated CaCl2 solution at 37% r.h. for 2 weeks. Immediately before X-ray analysis a light disaggregation of the dried solids was necessary, using an agate mortar, to obtain a homogeneous powder.

Solutions were obtained by filtration of 15 ml aliquots (30 ml for carbon determination) of the excess solution through a 0.22 μm alkali-resistant MF-millipore membrane syringe filter unit. Part of the solution (10 ml) was acidified with 1 ml 0.1 M HCl for cation analysis; the remainder was used to measure pH and determine anions, e.g. sulfate or carbonate. Solutions awaiting analysis were stored briefly in polypropylene (PP) centrifuge tubes.

3.2. X-ray diffraction

Mineralogical examination of the dried solid was made by X-ray diffractometry (XRD) using a PANALYTICAL X’PERT PRO diffractometer with CuKα-radiation at room temperature, ~25°C; the angular scan was between 5-80° 2θ with a step size of 0.02 and a count time of 1 s per step.

To characterise products of the temperature-dependent reactions in cementitious systems (chapter 7) a BRUKER D8 ADVANCE powder diffractometer was used for X-ray analysis. XRD-data of the samples have been collected using CuKα radiation at room temperature, ~25°C. The angular range was set between 5-40° 2θ with a step size of 0.04 and a count time of 2 s per step. Powder samples were filled sideways to the sample holder to minimise preferred orientation effects, particularly important for samples containing platy crystals, e.g. AFm and/or portlandite.

3.3. Thermal analysis

A simultaneous TGA-SDTA apparatus TGA/SDTA 851 by METTLER-TOLEDO INC. was used for thermal analysis. Simultaneous TGA/SDTA collects complex thermo- and differential-thermo- gravimetric weight changes (TGA and DTG) as well as thermal effects (DTA) during heating of the sample. In the current investigations thermal analysis was used to determine the state of hydration of the investigated solids. Furthermore significant weight losses during thermal decomposition may be used for identification and distinction of the solids in phase mixtures. The observed temperature range was between 30°C to 980°C; the rate of heating was 10°C/min. All measurements were done in N2-atmosphere.

Analytical methods 13

3.4. Microscopic examinations

An environmental scanning electron microscope (ESEM) FEI XL 30 was used for microscopic observations. In contrast to ordinary SEMs, no additional sample preparation, e.g. coating, was necessary and samples could be investigated in a defined gas atmosphere. Thus the possibilities of experimental artefact formation, e.g. changes of morphology due to coating and dehydration, were reduced. In the current work the GSE (gaseous secondary electron) or BSE (back scatter electron) detectors were used for observations in a low vacuum water vapour atmosphere (pressure: ~1 Torr). The acceleration voltage was 10 to 25 kV depending on the investigated sample.

3.5. Analysis of solutions

3.5.1 Calcium, aluminium and sodium Aqueous calcium, aluminium and sodium were analysed by atomic absorption spectrometry using a VARIAN SPECTRAA 10 flame AAS. A nitrous oxide / acetylene flame was used for calcium and aluminium and an air / acetylene flame for sodium. Calcium was measured at a wavelength of 422.7 nm. Aluminium was determined at a wavelength of 309.3 nm whereas sodium was analysed at 589 nm. Standards were prepared from 1000 mg/l TITRISOL (VWR chemicals). A linear calibration curve for calcium was obtained in the range from 0 - 5 mg Ca2+/l and for sodium from 0 - 2.5 mg Na+/l; 2500 mg K+/l, as analytical grade KCl, was added to suppress ionisation. Due to minor sodium impurities in KCl the analytical limit of detection of sodium was ~0.5 mg Na+/l. For calcium and sodium analysis samples were diluted with 2500 mg K+/l solution to suppress ionisation prior to analysis. Aluminium was usually measured in undiluted samples; no further suppressing agents were added. Standards for aqueous aluminium calibration were prepared in the range between 0-100 mg Al3+/l. High calcium concentrations (~400 mg Ca2+/l) may interfere with aluminium determinations. To evaluate this matrix effect, a standard of 40 mg Al3+/l was diluted by 2+ a factor of 2 with a saturated Ca(OH)2-solution (initial concentration 850 mg Ca /l). The subsequently measured Al3+ concentration, 17.4 mg/l, was slightly lower than the theoretical reference value of 20 mg/l. A sensitivity analysis has shown that the analytical errors from this slight depression do not significantly affect the subsequent thermodynamic calculations.

3.5.2 Sulfate Aqueous sulfate concentrations were determined by ion chromatography with a DIONEX DX-120 IC. An ion exchange analytical column IONPAC AS 4A 4mm equipped with a guard column was fitted for sulfate analysis. The analyte was injected into a 25 μl sample loop; the applied pressure was set between 1000-1100 psi (67-74 bars). The eluent used was 1.8 mM Na2CO3 / l.7 mM

NaHCO3. The eluent conductivity was suppressed by an ASRS Ultra self regenerating suppressor with deionised water (> 18 MΩ cm) regenerant. The chromatograph had a linear calibration in the 2- 2- range 0-10 mg SO4 /l and standards were prepared from 1000 mg SO4 /l TITRISOL solution. The 2- 2- achieved detection limit was ~0.25 mg SO4 /l at a background activity of ~14 μS. Due to low SO4 concentrations, most measurements were taken on undiluted samples and the mean value of three independent analyses used to minimise the analytical error.

Analytical methods 14

3.5.3 Silicon A spectrophotometric method based on the molybdenum blue method was adopted to determine the silicon concentration: Ramachandran and Gupta [156]. A CAMSPEC 301 spectrophotometer was used for the measurements. The optimum wavelength for silicon determination was 810 nm. Standards were prepared from 1000 mg Si4+/l TITRISOL solution. A linear calibration was achieved in the range from 0 to 5 mg Si4+/l.

3.5.4 Carbon Carbon was analysed using a LABTOC Analyser by PPM. Ultra-violet-promoted persulphate oxidation is used to determine the contamination of dissolved organics present in the sample. The content of total inorganic carbon (TIC) was determined as difference from the concentrations of total carbon (TC) and total organic carbon (TOC) present in the solution. TC was first determined as the sum of TIC- and TOC-content. In a second step, TOC was determined from the catalytically oxidised CO2. Inorganic carbonates were automatically removed by sample pre-treatment using an acid sparge. A sample with reagent was then injected in to the reaction vessel, the vessel being continuously sparged by a carrier gas (typically nitrogen), to liberate any CO2 gas generated through oxidation. The resulting evolved CO2 is delivered to an infrared detection system. The CO2 plot is integrated and compared against a calibration curve to determine the reportable value. The detection limit for a reliable TIC measurement using this method is ~ 1 mg/l; problems arose due to the low concentrations of TIC in the solutions. In most cases the concentrations of carbon in solution were lower than the limit of detection. Thus it was decided to estimate the values with the help of thermodynamic calculations based on assumed equilibria with phases with known thermodynamic properties, e.g. calcite.

3.5.5 Measurement of pH The pH was measured by a METTLER TOLEDO system equipped with a combination pH electrode INLAB 413 for simultaneous determination of pH and temperature of the solution. The pH-meter was calibrated with a 3-point calibration with pH 4.01, 7.00 and 9.21 buffers (at 25°C).

Additionally calibration was checked with a saturated Ca(OH)2 solution (pH~12.48 at 25°C) as an external buffer. The pH was automatically corrected by the pH-meter for the measured temperature. To minimise carbonation effects, the pH was measured immediately after filtration of the sample solutions.

Development of a thermodynamic database for cement hydrates 15

4. Development of a thermodynamic database for cement hydrates

A thermodynamic database will include many substances for which standard compilations already provide adequate data: it is not necessary to start totally afresh. For example, the thermodynamic properties of water and of many aqueous ions and complexes are well known. Database development focussing on cements is therefore mainly concerned with the properties of solids that are abundant in cements but uncommon or absent in nature. Thus a comprehensive database can be compiled by focussing on relatively few substances.

4.1. Synthesis of relevant cement hydrates

The synthesis of the relevant cement hydrates required several solid precursors. These were made from analytical grade (AR) reagents. C3A was prepared from a 3:1 molar ratio of CaCO3 and

Al2O3. The Al2O3 had a high fineness, dmax < 10 μm, to enhance its reactivity. To eliminate adsorbed water, the Al2O3 and the CaCO3 were dried previously at 950°C or 100°C overnight, respectively. The starting materials were mixed to a homogeneous paste with water in an agate mortar. A sufficient viscosity was necessary to avoid segregation. Afterwards the mixture was dried for 2 hours at 50°C. Then it was placed in a platinum crucible and heated to 950°C in a muffle furnace to decarbonate the CaCO3. After 4 hours the temperature was increased to 1400°C for another 6 hours. Then the sintered material was cooled down, ground to a fineness < 75 μm and reheated to 1400°C for 6-10 hours. This procedure was repeated at least four times. Afterwards the material was checked for purity by XRD. No significant impurities e.g. free lime, CaO, were found following the described procedure.

Lime, CaO, was obtained from decarbonation of analytical grade CaCO3 at 900°C overnight. The purity was checked by XRD and showed no impurities. Due to its sensitivity to hydration and carbonation only freshly prepared CaO was used.

Anhydrite, CaSO4, was used as sulfate source in the experiments. It was prepared by dehydration of gypsum in a muffle furnace at 750°C for 5 hours.

4.1.1 Hydrogarnet “Hydrogarnet” is usually defined in the cement literature as the silicon-free composition

Ca3Al2(OH)12. However silicon is a main constituent of Portland and blended Portland cements and the existence of solid solution between katoite, Ca3Al2(OH)12, and grossularite, Ca3Al2Si3O12, is well-known both in the laboratory and from natural occurrences. Siliceous hydrogarnet thus impacts significantly on silicon distribution in cements. To enable thermodynamic calculations, two different compositions were synthesised. Ca3Al2(OH)12 was prepared by mixing previously synthesised C3A with boiling water and subsequent ageing at 105°C in sealed PTFE bottles for 7 days. A siliceous composition was also prepared with the target composition Ca3Al2SiO4(OH)8, starting from stoichiometric amounts of CaO, Na2Si2O5⋅2H2O, NaAlO2 and water. A slurry of

Na2Si2O5⋅2H2O and NaAlO2 was prepared with an appropriate amount of water. In a separate operation, CaO was suspended in boiling ultra pure water and the slurry containing mixed

Na2Si2O5⋅2H2O, NaAlO2 added. Subsequently the preparation was aged for 4 weeks with periodic agitation at 105°C in sealed PTFE bottles until filtration.

Development of a thermodynamic database for cement hydrates 16

3 grossular (C3AS3; PDF 1-74-1087 )

hibbschite (C3AS2.3D1.4; PDF 1-84-2016)

hibbschite (C AS H ) 2 3 2 2; PDF 1-73-1654

hibbschite (C3AS2H2; PDF 31-250)

sil. hydrogarnet (C3AS1.25H3.5; PDF 45-1447)

1

sil. hydrogarnet (C3AS0.8H4.4) Garnet silicon content [mol] content silicon Garnet

C3AH6 (PDF 24-217) 0 11.8 11.9 12 12.1 12.2 12.3 12.4 12.5 12.6 Unit cell size [Å] Fig. 4.1: Estimation of the silicon-content of siliceous hydrogarnet; data marked PDF are from the Powder Diffraction File. Data not marked were obtained in the course of the title study (solid compounds). The composition of the synthesised solid solution (open diamond) is estimated by fitting to the curve shown by a dashed line.

A literature review showed that the synthesis and characterisation of siliceous hydrogarnet is more complicated than the silicon-free variant. Jappy, et al. [95] synthesised hydrogrossular solid

solutions Ca3Al2(SiO4)3-x(OH)4x and encountered two different hydrogarnet phases in most preparations. A miscibility gap was postulated to exist at low silica substitutions. However in the title study, one synthesis yielded a single hydrogarnet phase. According to subsequent XRD analysis this solid solution had a lower silicon-content than the target composition,

Ca3Al2SiO4(OH)8. Its silicon-content was estimated assuming a linear relation of the unit cell lattice

parameter between C3AH6 (a0~12.58 Å, PDF 24-217) and grossular, Ca3Al2(SiO4)3 (a0~11.85 Å,

PDF 39-368). The unit cell size of the synthetic (a0 = ~12.39 Å) was calculated by refining its XRD-pattern by least squares minimisation on 14 reflections with the software CELREF using

silicon, a0= 5.4308 Å as an internal standard (see Fig. 4.1). Accordingly, its formula was corrected

to Ca3Al2(SiO4)0 8(OH)8 8. The XRD-pattern also contained reflections attributed to C-S-H and most of the “missing” silica and part of the alumina are believed to be present as minor C-S-H impurity.

4.1.2 AFm phases Although the stability relations of the AFm phases are known to be sensitive to temperature, few relevant data are available. To enable an estimation of thermodynamic data, the following

preparation route yielded suitable material. “Monosulfoaluminate”, Ca4Al2(SO4)(OH)12·6H2O,

becomes more stable at temperatures > ~40°C. So a 1:1 molar mixture of C3A and CaSO4 was suspended at 100°C in initially ultra pure water and thereafter kept at 85°C for 7 days to synthesise phase pure monosulfoaluminate.

“Monocarboaluminate”, Ca4Al2(CO3)(OH)12·5H2O, was prepared by mixing C3A and CaCO3 in a 1:1 molar ratio with previously degassed ultra pure water at 25°C and stored with agitation in HDPE-bottles for 14 days until filtration at 25°C. A second source of monocarboaluminate was

prepared by mixing stoichiometric amounts of CaO, CaCO3 and gibbsite (Al(OH)3) with a 0.1 M KOH solution. The suspension was agitated periodically and stored at 50°C for 4 weeks, with subsequent washing.

Development of a thermodynamic database for cement hydrates 17

“Hemicarboaluminate”, Ca4Al2(CO3)0 5(OH)13·5.5H2O, was made by addition of C3A, CaCO3 and CaO in stoichiometric quantities to previously degassed ultra pure water at 25°C and stored in HDPE-bottles to achieve a successful synthesis. The mixture was aged with stirring for 14 days before filtration.

C4AHx, Ca4Al2(OH)14⋅xH2O, was synthesised according to the method of Atkins et al. [9]. C3A was mixed with CaO in a 1:1 molar ratio at 5°C using degassed ultra pure water (w/s ~ 10). Afterwards the preparation was stirred for 72 hours and periodically agitated, still at 5°C. After 3 weeks at 5°C the solid was vacuum filtered under N2-atmosphere.

Several methods for the preparation of strätlingite, Ca2Al2SiO2(OH)10⋅3H2O are described in the literature (e.g. sol-gel, preparation from glasses, etc.). These routes were pursued but the best preparations were obtained by starting from a stoichiometric mix of CaO, Na2Si2O5⋅2H2O, NaAlO2 and water (w/s ~ 10) at 25°C. First, all chemicals were separately suspended in ultra pure water at 25°C. The slurry containing sodium aluminate solution was added to the previously prepared portlandite solution with stirring. Finally the sodium silicate solution was added and the resulting suspension stirred for 4 weeks at 25°C prior to filtration. HDPE-bottles were used in all stages of the preparation. Unlike other cement hydrates, which are uniformly white, the strätlingite preparation had a pale bluish colour.

4.1.3 AFt phases

The preparation and determination of the solubility of SO4-AFt, ettringite, has been the subject of numerous investigations [8][9][147][192]. The literature data show generally good agreement and it was therefore concentrated on its less well characterised carbonate analogue, CO3-AFt or

“tricarboaluminate”, Ca6Al2(CO3)3(OH)12⋅26H2O, which was synthesised using a modification of the method of Carlson and Berman [36] by precipitation from a stoichiometric mixture of CaO,

NaAlO2 and Na2CO3 in a 10% w/v sucrose solution. The previously-prepared slurries of sodium aluminate and sodium carbonate were added to the sucrose-portlandite mixture (total w/s ~ 10), stirred for 3 days and periodically agitated at 25°C until filtration after ~ 2 weeks.

Development of a thermodynamic database for cement hydrates 18

4.2. Solubility determinations

Solubility determinations were made at various temperatures between 5°C and 105°C. Two series of reaction mixtures were prepared to derive solubility data for katoite, Ca3Al2(OH)12. In the experiment from undersaturation, previously synthesised dry Ca3Al2(OH)12 powder was redispersed in water (water/solid-ratio ~ 30) at 25°C and then stored in HDPE-bottles (at temperatures < 70°C) or PTFE-bottles (at temperatures ≥ 70°C) isothermally at well-spaced intervals between 5°C to

105°C prior to analysis. In a second series, a slurry of Ca3Al2(OH)12, prepared from C3A as described at 105°C, was divided into several samples with the aid of a syringe. The sample bottles

(HDPE or PTFE) were then filled with boiled water, sealed under N2-atmosphere and slowly cooled to the desired temperature and held for ~ 4 weeks prior to analysis. The water/solid mass- ratio of the second set of samples was deliberately set to ~ 1000 to enable direct comparison with the results of Wells et al. [193], which were also obtained at this ratio.

Experiments were conducted from both super- and undersaturation for monosulfoaluminate, monocarboaluminate and hemicarboaluminate at temperatures between 5°C and 110°C. Stoichio- metric mixtures of C3A with either CaSO4 or CaO and CaCO3 (see paragraph 4.1) were used at various temperatures in experiments from supersaturation (water/solid-ratio ~ 30). Solubilities from undersaturation were determined by redispersing previously synthesised and characterised single phase solids in ultra pure degassed water (water/solid-ratio ~ 30). The samples were analysed after 4 - 6 weeks reaction time in both cases. Additionally a second dataset for hemicarboaluminate at

25°C was generated by reaction of monocarboaluminate with synthetic C4AH13 (previously prepared at 5°C) at a 1:1 molar ratio (water/solid-ratio ~ 30): complete reaction at 25°C required ~ 3 weeks.

The synthesis of siliceous hydrogarnet, strätlingite and the second source of monocarboaluminate required the addition of alkalis and synthesis was thus achieved at high aqueous pH. Most of the alkalis could be removed by flushing the filtrates several times with ultra pure degassed water prior to drying. Residual alkali contents are commented subsequently. Solubilities determined from undersaturation were obtained by redispersing powders of each mineral in ultra pure, degassed water (water/solid-ratio ~ 30) and undertaking analyses at well-spaced time intervals.

Development of a thermodynamic database for cement hydrates 19

4.3. Methods used to derive and manipulate thermodynamic data

4.3.1 Software and standard databases Chemical thermodynamic modelling consists of calculating the chemical speciation (i.e. amounts or concentrations of chemical components in all phases present in equilibrium state) from total bulk composition of the system and thermodynamic data for components. In the GEM (Gibbs free energy minimisation) method and GEMS-PSI code [109], the total Gibbs energy of the system is minimised at given temperature and pressure; accordingly, for each component, the standard molar Gibbs energy at the temperature of interest must be provided. Calculations require a database of thermodynamic properties of components (substances), a correct statement of the problem, and a solver of chemical equilibria. In this work GEMS-PSI [109] was used - a software package including a GEM solver, a built-in thermodynamic database [93] and a graphical user interface for easy extension of the thermodynamic database to user-defined “projects”. This was convenient because not all cement minerals are included in standard databases such as Nagra-PSI [93] supplied within GEMS-PSI package. This database was initially designed in “logK format” for application to codes that use law of mass action algorithms at standard conditions (1bar and 25°C); to include it in GEMS, the logK values were converted into standard molar Gibbs energies and merged with the slop98.dat database [96][172], which was originally developed for the SUPCRT92 code [96]. For aqueous species, this dataset is based on the HKF (Helgeson-Kirkham-Flowers) equation of state which can be used to calculate temperature and pressure corrections up to 1000°C and 5 kbar; the necessary parameters for aqueous species relevant for cementitious systems are given in [172][179] and are summarised in Table A.1. The heat capacity coefficients needed for temperature corrections for most of the minerals in the GEMS databases are also given in the slop98.dat dataset. The database, included in the current software package, GEMS version 2.2, is documented in [184] and is in the public domain [93][96][172]. Raw data for minerals obtained in the title study have been converted into standard molar thermodynamic properties and added to the GEMS-PSI database in order to perform modelling calculations. Temperature corrections for thermodynamic properties of condensed substances (e.g. minerals) used in GEMS are based on the well known standard integration of the heat capacity function (e.g. [143]) as discussed below.

4.3.2 Estimation of heat capacity The heat capacity function for solids (at constant pressure) was calculated according to Eq. 4.1 where a0, a1, a2 and a3 are empirically derived, temperature independent parameters characteristic of each solid. o −2 − 50 210 +++= 3TaTaTaaCp (4.1) The heat capacities can be measured experimentally or, as was done here, estimated by using a reference reaction with a solid having a known heat capacity and similar structure. As shown by Helgeson et al. [89], this principle was successfully applied to estimate the heat capacity of silicate minerals by formulating a reaction involving a structurally-related mineral of known heat capacity. Gu et al. [86] used a similar approach to predict equilibrium constants for reactions related to aqueous species. Nevertheless, Helgeson et al. [89] pointed out that this method has limitations due to the differing thermodynamic properties of “water”, variously bound loosely as hydrate water or structurally, as OH-groups. To minimise errors associated with the varying strengths of bonding for “water”, care was taken to formulate reference reactions so as not to involve “free” water as a substituent in reactions unless appropriate to do so. Table 4.1 shows the coefficients to determine the heat capacity of reference solids to estimate heat capacity data of the relevant cement hydrates.

Development of a thermodynamic database for cement hydrates 20

Table 4.1: Standard molar thermodynamic properties of cement hydrates at 25°C, 1 bar 0 0 0 II Phase log KS0 ΔfG ΔfH S a0 a1 a2 a3 V° Ref 2 0.5 3 [kJ/mol] [kJ/mol] [J/(mol⋅K] [J/(mol⋅K)] [J/(mol⋅K )] [J⋅K/mol] [J/(mol⋅K )] [cm /mol] hydrogarnet I I C3AH6 -20.84 -5010.1 -5540 419 292 0.561 0 0 150 t.s.

C3AS0 8H4 4 -29.87 -5368.0 -5855 369 109 0.631 -1.95e+06 2560 143 t.s. AFt

I I C6AsH32 -44.90 -15205.9 -175351900 1939 0.789 0 0 707 [123]

C6AcH32 -46.50 -14565.7 -167921858 2042 0.558 -7.78e+06 0 650 t.s. AFm

I I C4AsH12 -29.26 -7778.5 -8750 821 594 1.168 0 0 309 t.s.

C4AcH11 -31.47 -7337.5 -8250 657 618 0.982 -2.59e+06 0 262 t.s.

C4Ac0 5H12 -29.13 -7336.0 -8270 713 664 1.014 -1.30e+06 -800 285 t.s.

C4AH13 -25.40 -7326.6 -8300 708 711 1.047 0 -1600 274 t.s.

C2AH8 -13.56 -4812.8 -5433 438 392 0.714 0 -800 184 t.s.

C2ASH8 -19.70 -5705.1 -6360 546 438 0.749 -1.13e+06 -800 216 t.s. C-S-H jennite-type -13.17 -2480.8 -2723 140 210 0.120 -3.07e+06 0 78III [123] (C1 67SH2 1) tobermorite - type -8.0 -1744.4 -1916 80 85 0.160 0 0 59III [123] (C0 83SH1 3) supplementary data water (H2O) -237.2 -286 70 75 0 0 0 18 [93]

CAH10 -7.50 -4622.4 -5320 501 151 1.113 0 3200 194 t.s. IV SiO2 (amorph) -848.9 -903 41 47 0.034 -1.13e+06 0 29 [104] IV gypsum (CaSO4⋅2H2O) -1797.8 -2023 194 91 0.318 0 0 75 [93] IV anhydrite (CaSO4) -1322.1 -1435107 70 0.099 0 0 46 [93] IV portlandite (Ca(OH)2) -897.0 -985 83 187 -0.022 0 -1600 33 [93] lime (CaO) -604.0 -635 39 49 0.004 -6.53e+05 0 17 IV [93]

IV calcite (CaCO3) -1129.2 -1207 93 105 0.022 -2.59e+06 0 37 [93] IV gibbsite (Al(OH)3) -1151.0 -1289 70 36 0.191 0 0 32 [93] clinker phases

C3S -2784.3 -2931 169 209 0.036 -4.25e+06 0 73 [12]

β-C2S -2193.2 -2308 128 152 0.037 -3.03e+06 0 52 [12]

C3A -3382.3 -3561205 261 0.019 -5.06e+06 0 89 [12]

C4AF -4786.5 -5080326 374 0.073 0 0 130 [12] t.s. - data obtained in title study Isee Ederova et al. [57] IIcalculated from unit cell parameters given in Taylor [180] if not stated otherwise IIIsee Lothenbach et al. [123] IV see GEMS-PSI-dataset [93][184]

Development of a thermodynamic database for cement hydrates 21

Table 4.2: Reference reactions used to estimate unknown heat capacities of cement minerals Phase Formula and reference reaction

Siliceous Ca Al (SiO ) (OH) + 1.6 Ca(OH) Æ Ca Al (OH) + 0.8SiO + 1.6CaO hydrogarnet 3 2 4 0.8 8.8 2 3 2 12 2 Ca Al (CO )(OH) ·5H O + 0.5CaSO ⋅2H O + 0.5CaSO Æ Ca Al (SO )(OH) ·6H O + Monocarboaluminate 4 2 3 12 2 4 2 4 4 2 4 12 2 CaCO3 Ca Al (CO ) (OH) ·5.5H O + 0.25CaSO ⋅2H O + 0.75CaSO Æ Ca Al (SO )(OH) ·6H O + Hemicarboaluminate 4 2 3 0.5 13 2 4 2 4 4 2 4 12 2 0.5CaCO3 + 0.5Ca(OH)2

Hydroxy-AFm Ca4Al2(OH)14·6H2O + CaSO4 Æ Ca4Al2(SO4)(OH)12·6H2O + Ca(OH)2

C2AH8 2Ca2Al2(OH)10·3H2O + CaSO4 Æ Ca4Al2(SO4)(OH)12·6H2O + Ca(OH)2 + 2Al(OH)3

CAH10 CaAl2(OH)8⋅2H2O + CaSO4 + 2Ca(OH)2 Æ Ca4Al2(SO4)(OH)12·6H2O 2Ca Al SiO (OH) ⋅3H O + CaSO Æ Ca Al (SO )(OH) ·6H O + 2SiO + Ca(OH) + Strätlingite 2 2 2 10 2 4 4 2 4 12 2 2 2 2Al(OH)3

Tricarboaluminate Ca6Al2(CO3)3(OH)12·26H2O + 3CaSO4 Æ Ca6Al2(SO4)3(OH)12·26H2O + 3CaCO3

The coefficients of the heat capacity function (Eq. 4.1) of the relevant cement hydrates were calculated according to reference reactions given in Table 4.2. Experimentally-determined heat capacities by Ederova and Satava [57] and data from the built-in GEMS database [93][184] were used as the basis for the estimation of the unknown coefficients (see Table 4.1). Analogue structures were used in the calculations, e.g. for “unknown” AFm phases monosulfoaluminate was used as the model. Assuming the reference reaction Eq. 4.2; Eq. 4.3 shows the principle way of calculating the necessary coefficient, an,A, for component A with the aid of the known coefficients an,B and an,C with y moles of component B and z moles of component C.

A Æ yB + zC (4.2)

an,A = y⋅an,B + z⋅an,C (4.3)

Table 4.1 summarises the resulting coefficients and estimated standard heat capacities at 25°C and 1 bar pressure calculated from Eq. 4.1.

4.3.3 Solubility based estimation of standard molar thermodynamic properties Explanation of the basic principles of the formulation, calculation and manipulation of solubility products is given in textbooks [143]. In this Thesis, activity coefficients of the relevant species were calculated using the extended Debye-Hűckel Eq. 4.4:

2 − i IAz log i =γ + bI (4.4) α+ i IB1 where γi is the activity coefficient of ion i, A and B are Debye-Hűckel solvent parameters dependent on the dielectric constant of water and temperature, zi is the ionic charge, αi is a parameter dependent on the size of ion, i, taken from [93][184]; b is a semi-empirical parameter (~0.064 at 25°C) and I is the effective ionic strength. Aqueous ion activities and speciation were calculated using the GEMS database appropriate to the particular calculation. Finally, temperature-dependent solubility products were calculated from the activities obtained according to the dissolution reactions in Table 4.3.

Development of a thermodynamic database for cement hydrates 22

Table 4.3: Dissolution reactions used to calculate solubility products

Mineral Dissolution reaction

2+ - - C3AH6 Ca3Al2(OH)12 Æ 3Ca +2AlO2 + 4OH + 4H2O 2+ - - - Siliceous hydrogarnet Ca3Al2(SiO4)0 8(OH)8 8 Æ 3Ca +2AlO2 + 0.8HSiO3 + 3.2OH + 2.4H2O 2+ - 2- - Monosulfoaluminate Ca4Al2(SO4)(OH)12·6H2O → 4Ca +2AlO2 + SO4 + 4OH + 10H2O 2+ - 2- - Monocarboaluminate Ca4Al2(CO3)(OH)12·5H2O → 4Ca +2AlO2 + CO3 + 4OH + 9H2O 2+ - 2- - Hemicarboaluminate Ca4Al2(CO3)0 5(OH)13·5.5H2O → 4Ca +2AlO2 + 0.5CO3 + 5OH + 9.5H2O 2+ - - C4AH13 Ca4Al2(OH)14·6H2O → 4Ca +2AlO2 + 6OH + 10H2O 2+ - - C2AH8 Ca2Al2(OH)10·3H2O → 2Ca +2AlO2 + 2OH + 7H2O 2+ - CAH10 CaAl2(OH)8·6H2O → Ca +2AlO2 + 10H2O 2+ - - - Strätlingite Ca2Al2SiO2(OH)10⋅3H2O Æ 2Ca +2AlO2 + HSiO3 + OH +7H2O

2+ - 2- - Tricarboaluminate (CO3-AFt) Ca6Al2(CO3)3(OH)12·26H2O Æ 6Ca +2AlO2 + 3CO3 + 4OH + 30H2O 2+ - 2- - Ettringite (SO4-AFt) Ca6Al2(SO4)3(OH)12·26H2O → 6Ca +2AlO2 + 3SO4 + 4OH + 30H2O 2+ - - Jennite-type C-S-H Ca1 67SiO2(OH)3 33⋅0.43H2O Æ 1.67Ca + HSiO3 + 2.33OH + 0.43H2O 2+ - - Tobermorite-type C-S-H Ca0 83SiO2(OH)1 67⋅0.5H2O Æ 0.83Ca + HSiO3 + 0.67OH + 0.5H2O

0 The Gibbs energy of reaction ΔrG T at temperature T was computed using Eq. 4.5:

0 Tr −=Δ KlnRTG T (4.5) where R = 8.31451 J/(molK) is the universal gas constant and KT is a thermodynamic equilibrium constant (=equilibrium solubility product) at a given temperature.

From the solubility products calculated at each temperature, the standard molar thermodynamic properties of each solid were computed with the help of GEMS-PSI using the built-in three-term temperature extrapolation [106][108] to obtain a temperature-dependent “logK” function, which was fitted to the previously calculated solubility products. This function was estimated using Eq. 4.6 and the relationships shown in Eqns. 4.6 to 4.12.

−1 20T ++= 3 TlnATAAKlog (4.6) and

4343.0 0 0 A [ Tr r +Δ−Δ⋅= 0 )Tln1(CpS ] (4.7) 0 R 0 T0

4343.0 A 0 Δ−Δ⋅−= 0 )TCpH( (4.8) 2 R Tr 0 r T0 0

4343.0 0 A 3 Δ⋅= r Cp (4.9) R T0

0 0 0 T Tr Tr Δ+Δ=Δ r lnCpSS 0 T0 (4.10) T0

0 0 0 −Δ+Δ=Δ )TT(CpHH Tr r T0 r T0 0 (4.11)

0 0 0 Tr Tr Δ−Δ=Δ STHG Tr (4.12)

Development of a thermodynamic database for cement hydrates 23

0 0 The heat capacity effect of reaction, ΔrCp T = ΔrCp T0 = Δao, was assumed to be constant over the temperature range 0-100°C. Two parameters were adjusted to obtain a best visual fit to the experimental data:

0 1) ΔrG T at the reference state (25°C and 1 bar pressure) was estimated according to Eq. 4.5 using the experimentally-derived solubility product at 25°C.

0 2.) ΔrH T at the reference state was estimated to obtain the best visual fit between extrapolated solubility products according to Eq. 4.6 and calculated solubility products from experimentally- derived solubilities.

0 ΔrS T was subsequently calculated using Eq. 4.12. Then the related standard molar thermodynamic properties were calculated according to dissolution reactions given in Table 4.3 using standard state 0 0 properties of the aqueous species Table A.1 and the earlier estimated parameters ΔrG T, ΔrH T and 0 ΔrS T. Combined with previously estimated Cp(T) coefficients (Table 4.1) and the known HKF parameters of the aqueous species, the individual temperature dependency of Cp(T) was subsequently calculated for each hydrate phase. Thus, in a final step, the solubility products were recalculated by GEMS using built-in parts of the SUPCRT92 program [96] to derive temperature- dependent values. Differences arising between the first approach, using the three-term temperature 0 extrapolation with assumed constant ΔrCp T, and the second calculation, taking into account temperature-dependent heat capacity coefficients according to Table 4.1 and using values from the GEMS standard database [93][184], are marginal and lie within limits of other errors over the temperature range 0 to ~100°C [106].

To check the internal consistency of the thermodynamic database, the experimentally-derived solubility data were predicted using the derived thermodynamic database and the observed phase assemblages. This seems like a circular argument and indeed is not intended to prove that the data are correct, only to demonstrate that internal self-consistency was achieved. Generally the best agreement between recalculated and experimental solubility data of AFm, AFt and hydrogarnet phases was observed by suppressing the formation of gibbsite in the computations. The experimentally-derived solubility data for each phase are also listed in the appendix A of this Thesis, leaving the reader free to perform other calculations, if desired.

4.3.4 Thermodynamics of solid solutions and the use of Lippmann phase diagrams Solid solutions are frequently encountered in cementitious systems. The molar Gibbs free energy

ΔGss of a solution between different end members i can be calculated according to Eq. 4.13:

0 Δ+Δ=Δ GGXG ss ∑ i if M (4.13) i

=Δ+Δ=Δ − )KlnKlnX(RTGGG M id ex ∑ ii ss (4.14)

=Δ XlnXRTG id ∑ ii (4.15)

=Δ lnXRTG γ ex ∑ ii (4.16)

Development of a thermodynamic database for cement hydrates 24

The first term of Eq. 4.13 describes the free energy of a mechanical mixture of the end members i of the solid solution and is calculated using the mole fraction Xi = ni/Σni (ni is the mole amount of 0 the end member i; ΣXi = 1) and ΔfGi - the standard molar Gibbs energy of formation of end member i. The second term of Eq. 4.13 expresses the molar Gibbs energy of mixing, ΔGM for a given composition of the solid solution series and is computed according to Eq. 4.14 as the sum of the Gibbs energy of mixing of an ideal solid solution, ΔGid, and the excess Gibbs energy of mixing,

ΔGex, for the solid solution. ΔGid is calculated as described in Eq. 4.15 (with R, the universal gas constant, and T as the temperature of interest). The excess Gibbs energy of mixing, ΔGex, is only needed to compute thermodynamic properties of non-ideal solid solutions and can be calculated according to Eq. 4.16, where γi is the activity coefficient of the end member i. In the case of an ideal solid solution, all γi equal 1 and thus ΔGex = 0. In the title study, an ideal solid solution model

(ΔGex = 0) was used to describe the thermodynamic properties of single phase calcium-silicate- hydrate (C-S-H). A more detailed explanation of this solid solution model can be found in [104].

As shown in Eq. 4.14, ΔGM can be calculated independently, from the deviation of a linear function of the sum of the partial solubility products Ki of the end members i of the solid solution and the actually calculated solubility product Kss of the solid solution at stoichiometric saturation as described by Glynn [80].

Several cementitious phases form non-ideal solid solutions, but only over a limited range of compositions. Hence, miscibility gaps will be observed. In the case of non-ideal mixing the Gibbs energy of the solid solution, ΔGss, is calculated with Eq. 4.13 and the activity coefficients γi ≠ 1 and the excess Gibbs energy of mixing, ΔGex, of a binary solid solution are calculated according to Eq. 4.17:

ex =Δ γ+γ 2211 )lnXlnX(RTG (4.17) where X1= n1/(n1+n2) and X2 = n2/(n1+n2) (n1 and n2 are the amounts of end members; X1+X2=1).

The GEMS-PSI code has several built-in functions for non-ideal solid solutions [105]. In this Thesis, a semi-empirical model first suggested by Guggenheim and later developed by Redlich and Kister [80][81][82] was used to estimate the excess Gibbs free energy function of non-ideal binary solid solutions (Eq. 4.18):

2 ex =Δ 21 2110 212 +−+−+ ...))XX(a)XX(aa(RTXXG (4.18)

The empirical interaction parameters a0, a1, … are dimensionless. As shown by Glynn [81], knowledge of two fitting parameters a0 and a1 is sufficient to estimate the excess Gibbs energy function with reasonable accuracy. The activity coefficients γi of the end members i can be derived by fitting a0 and a1 to Eq. 4.18 and estimated according to Eqns. 4.19 and 4.20:

2 21 [ −+=γ 2110 )XX3(aaXln ] (4.19)

2 12 [ −−=γ 1210 )XX3(aaXln ] (4.20)

In the title study, the software MBSSAS [81] was used to derive the fitting parameters a0 and a1 based on experimentally-observed compositional boundaries of the miscibility gap in the binary solid solution series. A detailed description of MBSSAS is given in [81]: Kersten [101] applied a similar approach to estimate thermodynamic data for C-S-H.

Development of a thermodynamic database for cement hydrates 25

Lippmann phase diagrams Lippmann [120][121][122] developed a mathematical algorithm to construct phase diagrams to display all possible equilibrium states of a binary solid solution B1-xCxA and its related aqueous phase composition. The diagrams are based on the law of mass action equilibrium according to Eqns. 4.21 and 4.22 ([81][82]):

−+ XK]A][B[ γ= BABABA (4.21)

−+ XK]A][C[ γ= CACACA (4.22)

- + + - + + where [A ], [B ], [C ] are the aqueous activities of the ionic species A , B and C ; KBA and KCA are the solubility products of the end members BA and CA of the binary solid solution series; XBA and

XCA are the mole fractions of BA and CA in the solid and γBA and γCA are the activity coefficients of the BA and CA members of the solid solution series; γBA and γCA are calculated according to

Eqns. 4.19 and 4.20 respectively, with X1 = XBA and X2 = XCA.

To enable the construction of this phase diagram Lippmann [121] introduced the “total solubility product” ΣΠ as sum of Eqns. 4.23 and 4.24 and it is expressed as follows:

+− + ∑ =Π + ])C[]B]([A[ (4.23) or

XKXK γ+γ=Π ∑ sd,eq BABABA CACACA (4.24) Eq. 4.24 is used to derive the solidus curve of a Lippmann diagram in dependence of the solid- phase composition of the solid solution series.

To enable a complete description of the binary solid solution aqueous solution (SSAS) system a mathematical function needs to be derived to relate the aqueous composition to the “total solubility product” ΣΠ. Thus Lippmann [121] derived the solutus-equation according Eq. 4.25:

⎛ X X ⎞ =Π 1 ⎜ aq,B + aq,C ⎟ ∑ sl,eq ⎜ ⎟ (4.25) ⎝ K γ BABA K γ CACA ⎠ + where XB,aq and XC,aq are the “aqueous activity fractions” [82][121] of the substitutable species B and C+ calculated according to Eqns. 4.26 and 4.27:

+ ]B[ XK γ BABABA X = = (4.26) aq,B + + + ]C[]B[ ∏ ∑ sd,eq

+ ]C[ XK γ CACACA X = = (4.27) aq,C + + + ]C[]B[ ∏ ∑ sd,eq

As stated by Glynn [81] the activity coefficients γBA and γCA are dependent on the solid-phase composition (compare with Eqns. 4.19 and 4.20) and therefore are not a function of the aqueous phase composition except in the case of an ideal solid solution with γBA = γCA = 1.

Development of a thermodynamic database for cement hydrates 26

To enable the construction of a Lippmann phase diagram the dimensionless parameters a0 and a1 have to be derived to calculate the activity coefficients γBA and γCA according to Eqns. 4.19 and 4.20. Glynn [81] developed the software MBSSAS which uses the previously described mathematical relations to calculate thermodynamic equilibrium states of binary SSAS systems and to construct Lippmann phase diagrams. In dependence of the input parameters, MBSSAS is able to calculate the fitting parameters a0 and a1 or computes the location of miscibility gaps if the fitting parameters are estimated. In cement science MBSSAS was applied by Kersten [101] to estimate thermodynamic data for C-S-H. Further information about the software and possible applications are given in [80][81][82][101][105]. In the current work MBSSAS was used to calculate the fitting parameters based on experimentally- estimated compositional boundaries of the miscibility gap in different binary solid solution series.

Lippmann diagrams were used to refine the parameters a0 and a1 to experimentally-derived solubility data. A more detailed description and an example of the use of Lippmann phase diagrams will be given in chapter 5.3.1.

Development of a thermodynamic database for cement hydrates 28

C 3 AH 6 -20 10 4) 1) 5) -21 8 5) calcium 3) 5) 6 2) -22 8) 6) 7) 8) 4 7) 1) Carlson 1958 5) Atkins 1991 aluminium -23 2) Roberts 1969 6) D'Ans 1953 2 3) Butler 1958 7) Wells 1943 Concentration [mmol/l] 4) Nacken 1934 8) Peppler 1954 Calc. solubility product log Ksp-24 0 0 20 40 60 80 100 120 0 20406080100 Temperature [°C] Temperature [°C]

Fig. 4.3: Calculated solubility products of C3AH6 from Fig. 4.4: Recalculated solubility data for solubility experiments (lines show calculated results hydrogarnet, C3AH6, based on fitted thermo- as best fit) dynamic data from Fig. 4.3 (markers show experimental values from the title study; lines show calculated results; formation of gibbsite suppressed in calculations.)

The resulting standard molar thermodynamic properties of C3AH6 are summarised in Table 4.1. 0 0 The data agree well with those reported by Babushkin et al. [12] (ΔfG = -5014.1 kJ/mol, ΔfH = 0 0 -5548 kJ/mol) as well as with experimentally-determined ΔfH by Berman [23] (ΔfH ~ -5561 0 kJ/mol) and Schönitz et al. [168] (Δ fH ~ -5551.5 kJ/mol): all lie within the expected range of analytical errors.

Fig. 4.4 shows the recalculated temperature-dependence of calcium and aluminium solubilities from C3AH6 redispersed in pure water using the experimental conditions described in paragraph 4.3.3. The calculated data show good agreement with averaged measured concentrations. Consideration of other solubility data (see Fig. 4.3), especially values at lower temperatures, results in slightly lower concentrations compared to the data obtained in this Thesis. Hydrogarnet was predicted as the only stable CaO-Al2O3-H2O solid at this composition over the temperature range from 5 to 100°C at 1 bar pressure.

Siliceous hydrogarnet, Ca3Al2(SiO4)0.8(OH)8.8 As shown in Fig. 4.5, phase pure siliceous hydrogarnet could not be synthesised using the procedure described in 4.1.1. Small amounts of a C-S-H phase, coprecipitated during the initial synthesis of siliceous hydrogarnet, persisted at all temperatures between 5°C to 85°C despite three dispersions in the course of solubility experiments. As noted in section 4.1.1, it is likely that the siliceous hydrogarnet had a lower silicon-content than the target. Thus provisional solubility products were calculated using the composition estimated from XRD data, Ca3Al2(SiO4)0 8(OH)8 8.

Table A.3 shows provisional solubility data for the siliceous hydrogarnet. Comparison with the

C3AH6 data shows that silica substitution leads to a significant reduction of solubility, indicating stabilisation of the hydrogarnet phase by silica substitution. The calculated solubility products of siliceous hydrogarnet are considerably lower than those of C3AH6.

Development of a thermodynamic database for cement hydrates 35

As shown in Fig. 4.15, the intensities of the XRD-reflections of hemicarboaluminate decrease regularly over the temperature range between 5°C and 40°C while at the same time, those of

C3AH6 and monocarboaluminate increase in intensity in the range 5° to 40°C. But hemicarbo- aluminate was absent at both 50°C and 85°C and the phase constitution was instead dominated by the coexistence of monocarboaluminate, C3AH6 and portlandite.

Solubilities of hemicarboaluminate were obtained from super- and undersaturated solutions; the results are shown in Table A.6. The resulting calculated solubility products, Table A.6, were used to fit the standard thermodynamic properties of the compound (Fig. 4.16 and Table 4.1). The amounts of dissolved carbonate were always below the detection limit of the analytical technique but in this instance, carbonate concentrations could not be calculated by assuming calcite saturation because hemicarboaluminate is not stable in the presence of calcite. But, since XRD-investigations showed hemicarboaluminate to be compatible with monocarboaluminate, saturation with respect to the latter was assumed and used to fix the carbonate concentration.

Hemicarboaluminate solubilities are strongly incongruent. As can be seen in Fig. 4.17, calcium concentrations increase with rising temperatures to ~50°C, at and above which hemicarboaluminate is predicted to be unstable. The derived solubility products from this study agree reasonably well with values calculated from solubility data by Hobbs [91] (using carbonate concentrations estimated by assuming monocarboaluminate saturation) and by Damidot et al. [49]. Hemicarboaluminate decomposes progressively with increasing temperatures. As calculated, (Fig. 4.18), and in agreement with XRD-analyses (Fig. 4.15), the amount of hemicarboaluminate decreases regularly with increasing temperatures while increasing proportions of C3AH6 and monocarboaluminate develop. It is tentatively concluded that the upper thermal stability limit of hemicarboaluminate is 45 ± 5°C.

monocarb. + calcite + hemicarb. + monocarb. portlandite + portlandite + C AH C AH + C 3 AH 6 3 6 3 6 -28 18 16 14 1) calcium -29 12 10 2) 8 -30 6 1) 4 carbonate 1) Hobbs 2001 2 aluminium 2) Damidot 1994 [mmol/l] Concentration -31 0 Calc. solubility product log Ksp product log solubility Calc. 0 2040600 20406080100 Temperature [°C] Temperature [°C]

Fig. 4.16: Calculated solubility products of hemi- Fig. 4.17: Recalculated solubility data for carboaluminate from solubility experiments (line hemicarboaluminate based on fitted thermo- shows calculated best fit) dynamic data from Fig. 4.16 (markers show experimental values from the title study, lines show calculated fits). Predicted solid phases are shown at the top.

Development of a thermodynamic database for cement hydrates 37

-5

-6

-7 4) 1) Carlson 1958 1) 2) Roberts 1969 -8 3) Butler 1958 2) 4) Percival 1958 3)

Calc. solubility product log Ksp Ksp log product solubility Calc. -9 Fig. 4.20: Calculated solubility products of 0204060 CAH10 from literature solubility data (line Temperature [°C] shows calculated best fit)

The stability of C4AHx relative to mixtures of portlandite and C3AH6 is in dispute. Broadly the literature offers two interpretations; citations [61][52][193] report that C4AHx decomposes rapidly to mixtures of C3AH6 and portlandite at and above 20 - 25°C. On the other hand, Carlson [35], Roberts [162], Seligman and Greening [170] and van Aardt and Visser [1] report the opposite:

C3AH6 and portlandite react with formation of C4AHx, at low temperatures, 1 - 5°C.

Furthermore Butler and Taylor [32] and Carlson [35] show the apparent stable formation of CAH10 at low Ca/Al ratios and temperatures ≤ 5°C. Butler and Taylor report that CAH10 is a stable phase in the system CaO-Al2O3-H2O and thus C2AH8 should decompose to a mixture of CAH10 and

C4AH13 at low temperatures. However subsequent investigations by Percival and Taylor [145] have led to the conclusion that CAH10 is not stable at temperatures above 21°C.

Only Peppler and Wells [155] gave solubility data for C4AH13 at higher temperatures, > 25°C.

Analysis of the phase composition showed increasing amounts of C3AH6 formed upon prolonged ageing at temperatures ≥50°C, which indicates decomposition of C4AH13 and C2AH8 at elevated temperatures. However the early-age values of C2AH8 and C4AH13, before the onset of significant decomposition, are close to the values reported here. D’Ans [52] observed the same stability trend for prolonged ageing at 25°C. Hobbs [91] showed significant differences between solubility data obtained from undersaturation (aged ~ 515 d) and supersaturation (aged ~170 d). Major differences in the aluminium concentration caused large deviations between the solubility products calculated from under- and supersaturation. Nevertheless, relative to the many sources of data, there is generally good agreement between the various solubility datasets. The fitted value of C4AH13 for 0 0 ΔfH (-8302 kJ/mol) agrees reasonably with the calculated data by Babushkin et al. [12] (ΔfH ~ - 0 8318 kJ/mol). But ΔfG of C4AH13 (-7326.6 kJ/mol) is significantly higher than was calculated by 0 0 0 Babushkin et al. (ΔfG ~ -7347.8 kJ/mol). The data for C2AH8 (ΔfG ~ -4813 kJ/mol, ΔfH ~ -5432 0 0 kJ/mol) are in very good agreement with Babushkin et al. [12] (ΔfG ~ - 4818 kJ/mol, Δ fH ~ -5436 0 0 kJ/mol). The estimated values for CAH10 (ΔfG ~ -4622 kJ/mol, Δ fH ~ -5320 kJ/mol) agree very 0 0 well with the calculated data by Babushkin et al. [12]: (ΔfG ~ -4618 kJ/mol, Δ fH ~ -5320 kJ/mol).

Development of a thermodynamic database for cement hydrates 42

CO3-AFt, Ca6Al2(CO3)3(OH)12⋅26H2O

The existence of a carbonate analogue of SO4-AFt was first described by Carlson and Berman [36].

However, as shown in [36] and subsequently confirmed by Pöllmann et al. [152], synthesis of CO3- AFt has only been achieved in a solution containing sucrose as well as sodium carbonate. This leads to a strong increase of the solubility of calcium in conjunction with high carbonate concentrations. Other synthesis methods under conditions likely to occur in cementitious system were unsuccessful and gave mixtures of monocarboaluminate and calcite, suggesting that CO3-AFt is metastable with respect to CO3-AFm in Portland cement pastes. This contention was later proved by Damidot et al. [49] who published the first thermodynamic data for CO3-AFt.

800

600

400

Intensity [cps] calcite 200

0 5 10152025303540 [2Θ] CuKα

Fig. 4.27: XRD-pattern (left) and scanning electron micrograph (right) of CO3-AFt used for solubility determinations.

A fresh set of solubility experiments was initiated to determine the solubility of CO3-AFt made using the sucrose method at 25°C, as described in section 4.1.3. Fig. 4.27 shows the XRD-pattern of the dry CO3-AFt powder and its typical needle-like morphology. Mixtures of previously dried

CO3-AFt with ultra pure water were aged up to 9 months at 25°C. Despite the supposed metastability of CO3-AFt with respect to CO3-AFm, no obvious mineralogical changes were observed: CO3-AFt continued to coexist with small amounts of calcite. The solubility data of successive extractions are given in Table A.9.

Although solubility data for CO3-AFt have only been determined at 25°C the temperature- dependent behaviour of CO3-AFt can be estimated using the Helgeson approach [89]. Data can be approximated for other temperatures by formulating a reference reaction with compounds of similar structure and known entropies. The reference reaction for CO3-AFt is listed in Table 4.2. With the known entropies of anhydrite and calcite and the previously estimated value for ettringite 0 (see Table 4.1) the standard absolute molar entropy of CO3-AFt was estimated assuming ΔS r = 0. 0 Subsequently ΔfH was calculated using the Gibbs-Helmholtz relation (Eq. 4.12), the known 0 0 standard state entropies of the elements [109] and ΔfG . In the literature only one value for ΔfH 0 was found for CO3-AFt [12]; ΔfH ~ -16217 kJ/mol for the 30 H2O hydrate (which corresponds to ~ 0 -16789 kJ/mol for the 32 H2O hydrate) in very good agreement with the estimated value (ΔfH ~

-16792 kJ/mol) for the formula with 32 H2O. The resulting estimated standard molar thermodynamic data of CO3-AFt are summarised in Table 4.1. The computed composition of the aqueous phase in equilibrium with CO3-AFt and calcite at 25°C showed good agreement with the averaged solubility data for CO3-AFt (Table A.9).

Development of a thermodynamic database for cement hydrates 43

-40 230

1) C3A + CaCO3 + 11H2O --> C4AcH11 -42 210 -44

-46 190 [kJ/mol] r -48 0 G C A + 3CaCO + 32H O Δ 170 3 3 2 - -50 --> C6Ac3H32 1) Damidot 1994 150

Calc. solubility products log Ksp log products solubility Calc. -52 0 20406080100 020406080100 Temperature [°C] Temperature [°C]

Fig. 4.28: Calculated solubility product of CO3-AFt Fig. 4.29: Free energy plot of CO3-AFt and mono- from solubility experiments. The solid diamond shows carboaluminate the single value experimentally-determined in the title study. The solid line shows the calculated solubility product: see text and estimated thermody- namic data

As noted, although CO3-AFt is metastable with respect to monocarboaluminate at ~25°C, free energy plots are a useful instrument to derive data about relative stabilities of phases as a function of temperature. Fig. 4.29 compares free energy plots for the formation of monocarboaluminate and

CO3-AFt from C3A and calcite over the temperature range 1 to 99°C. As expected, monocarboaluminate is thermodynamically more stable than CO3-AFt between 1 to 99°C. But with decreasing temperatures CO3-AFt is increasingly stabilised and is close to being more stable than monocarboaluminate at 1°C. Due to lack of experimental data and uncertainties in the estimation of thermodynamic parameters, the present state of the data must be regarded as provisional. 2- 2- Nevertheless low temperatures tend to stabilise CO3-AFt and substitution of SO4 by CO3 in the ettringite structure is clearly favoured at low temperatures, < 25°C.

Thaumasite is another compound containing essential carbonate and having an AFt type structure.

In contrast to CO3-AFt and SO4-AFt, aluminium is replaced by silicon in the thaumasite structure.

According to the literature [2][40], and in contrast to CO3-AFt, thaumasite has a stability range in the cement system and is preferably formed at low temperatures, < 10°C. This is in agreement with own calculations, where it was shown that CO3-containing AFt structures are stabilised at low temperatures. Furthermore Macphee and Barnett [127] and others [103][126][185] have shown that

SO4-AFt forms solid solutions with thaumasite. Macphee and Barnett [127] published solubility data for the solid solutions but not for the thaumasite end member. Bellmann [20] obtained solubility data for natural thaumasite, but only at 8°C. Thus until recently no complete thermodynamic dataset for thaumasite was available. In the course of this work a new dataset was prepared to (i) predict the temperature-dependent stability of thaumasite and (ii) to enable the modelling of the solid solution behaviour between thaumasite and SO4-AFt (see chapter 6).

Development of a thermodynamic database for cement hydrates 45 amorphous silica and C-S-H (Ca/Si = 0.83) at low initial bulk Ca/Si ratios < 0.83, and also enables unambiguous calculation of the range of Ca/Si ratios of most interest to cement paste to be encompassed. These conclusions are supported by recent calculations by Kulik [107].

The C-S-H phase may sorb sulfate, alumina and alkalis. At present, these chemical variants are not included into a thermodynamic model. However, Hong et al. [92] give distribution coefficients of the sorption of alkalis into C-S-H, which can be used to estimate the amount of sorbed alkalis in C-S-H in dependence of the alkali concentration of the pore solution of the hydrated cement paste. Similarly, the amount of sorbed sulfate can be approximated. Corresponding temperature- dependent values for the adsorption of sulfate to C-S-H are given in [15][54][130][160]. Richardson and Groves [160] also describe aluminium for silica substitution in the C-S-H structure. The amount of substituted aluminium increases with decreasing Ca/Si ratios and can be estimated using the relations described in [160].

Development of a thermodynamic database for cement hydrates 46

4.5. Discussion

4.5.1 Data accuracy The data for individual substances vary in quality. The reasons for this have been developed in the text but it is still difficult to establish error limits. Firstly, the quality of synthesis is variable: it is difficult if not impossible to establish the absolute phase purity of the preparations. Only the use of single crystals would enable a high confidence in the absolute purity of the preparations used. Secondly, the substances themselves vary in crystallinity with as yet unknown consequences to the numerical values of the data derived from their use. An example is strätlingite, X-ray patterns of which always exhibit line broadening associated with low crystallinity, small crystallite size and internal disorder, perhaps with respect to the stacking of successive layers: no attempt was made to deconvolute the causes of line broadening or the impact of crystallinity on thermodynamic properties. But it could be argued that the AFm phase, particularly strätlingite occurring in commercial cements, also exhibits disorder and thus the synthetic product is representative of “real” strätlingite.

The results, expressed in tables and diagrams, comprise a working set of data for subsequent calculations. It is not claimed that the database is perfect and indeed have noted the shortcomings of particular datasets as appropriate. In a very few cases where data are incomplete, there is also noted work in progress. It is the author’s hope that, as the use of thermodynamic methods becomes routine, others will add to and improve the accuracy of the compilation.

A general source of experimental uncertainty concerns the analytical data. Some of the very low concentrations, e.g. the aluminium concentrations of C3AH6 in equilibrium with portlandite, have numerical values close to the limit of accurate analysis. Carbonate concentrations generally had to be estimated assuming equilibrium with other carbonate phases whose thermodynamic properties were known, e.g., calcite. However, as shown in this work, the solubility data generally pass two tests: of self-replication; that is when used as inputs for calculations they reproduce reliably the input data, usually in the form of solubilities, and second, that they are derived from approach to equilibrium from both oversaturation and undersaturation. This latter test, while not infallible, is generally regarded as substantive proof of the attainment of equilibrium. Moreover the data presented are broadly comparable with those for other substances recorded in the literature. Care was taken to assess all main solubility investigations recorded in the accessible literature. However, as many of the investigations go back to the 1930’s, when analytical techniques were often not advanced sufficiently to analyse low concentrations of species, these add another source of uncertainty.

Impurities present in solutions can significantly influence the solubility data. For example a large data scatter was observed for the solubilities of hydrogarnet, C3AH6. The older datasets by Wells et al. [193] and Peppler et al. [155], used to derive thermodynamic values in previous databases, differ significantly from data obtained in the course of the title study. Wells et al. [193] and Peppler et al. [155] used glass flasks in the course of the solubility determinations. Thus it is likely that silicon was dissolved in the course of the experiment. As shown in section 4.4.1, silicon can be bound in the hydrogarnet structure and the resulting siliceous hydrogarnet, Ca3Al2(SiO4)3-x(OH)4x, will be significantly stabilised by silica and have lower solubility than the silica-free composition.

Atkins et al. [7], who determined solubilities of C3AH6 assemblages, confirmed this hypothesis. In the silica-free samples the solubility product of C3AH6 at 25°C was close to the values obtained in the title study whereas in assemblages containing silicates, e.g. C-S-H and strätlingite, the

Development of a thermodynamic database for cement hydrates 48

4.5.2 Relations between equilibrium and kinetics In the course of preparing this study, there were numerous comments that, while equilibrium is studied, kinetics is ignored. This comment contains an element of truth: it is not possible to advance both topics given constraints on resources. But kinetics is not ignored. For example in the system CaO-SiO2-H2O, the stable phases tobermorite and jennite are suppressed in favour of metastable C-S-H: C-S-H is assigned a definite set of thermodynamic properties and is allowed to participate in reactions with other phases, including some that are thermodynamically stable under the conditions of the calculation. This was done in the knowledge that the C-S-H phase is persistent. While accepting that the kinetics are important, relatively little progress has been made in their quantification. A thermodynamic approach is immediately successful in showing why this is so. Reactions can be divided into two types: internal, in which overall composition does not change (isochemical) and externally-induced, where cement solids react with their service environment. The latter class of reactions are intrinsically more difficult to quantify because the bulk composition of solids changes in the course of reaction. But thermodynamics immediately shows that, in seeking to understand the diffusion of a single species into (or out of) cement, it is in fact necessary to constantly recalculate mass balances: diffusion of a single species in isolation is insufficient to account for the overall reactions. A thermodynamic approach, which automatically tabulates mass balances, provides a mass balance template for subsequent kinetic studies and usefully highlights the underlying mechanisms and their probable importance. The experimental studies give direct information on reaction kinetics. Many examples are given in the text of reactions which occur relatively rapidly (hours to days) while others occur slowly or not at all. An example of slow reactions is the stability of siliceous hydrogarnet at low temperatures and at low carbonate and sulfate activities (see paragraph below). There also exists a third class of reactions: those predicted to occur but where observation and experiment are inadequate to determine whether or not the relevant prediction is confirmed. It is thus suggested that the kinetics are best approached in conjunction with thermodynamic equilibrium calculations rather than as a separate exercise: the driving forces responsible for change have a thermodynamic basis and, moreover, comparison of calculation and experiment shows that with few exceptions, the principal of local equilibrium is maintained, or nearly so, in the course of environmentally-conditioned reactions. Many of the verification experiments performed in the course of this thesis and in other supporting studies disclose that reaction occurs relatively rapidly. Of course the experimental conditions were selected to facilitate reaction, as for example, by using high water/solid-ratios. Nevertheless further work may disclose that phase changes in cement mineralogy driven by temperature changes occur relatively rapidly. It is of course a well-known generalisation that reaction rates become more rapid at higher temperatures, but perhaps surprisingly, some mineralogical conversion reactions were found to proceed rapidly (days or weeks) at 5°C. Thus while kinetics are important, much evidence indicates that, with the notable exception of C-S-H, the internal constitution of hydrated Portland cements tends to approach a thermodynamic equilibrium.

Development of a thermodynamic database for cement hydrates 49

4.5.3 Applications of the database Inspection of data tables does not directly reveal relationships between chemistry and mineralogy and its variation with temperature – to see the effects, much thermodynamic modelling calculations are needed. But five features stand out. Firstly, the assemblage of portlandite and C-S-H dominate the products of cement hydration. With few exceptions -thaumasite being an important one-, these phases persist in most of the assemblages and are rather insensitive in composition to the presence or absence of other phases and, although the C-S-H phase can vary widely in its Ca/Si ratio, its ratio is effectively fixed at or near its highest attainable value (1.5 - 1.9) in the presence of portlandite. Secondly, the minor phases of hydrated Portland cement (AFm, AFt,...) are very variable in composition and structure. By not forming complete solid solution even within a structural family, the nature of the minor solids remains very sensitive to changes in bulk composition and temperature: note for example, the chemical-mineralogical complexity of the AFm and AFt families. Thirdly the temperature dependence of the nature of the stable phases is quite remarkable and will be explored in a subsequent chapter (see sections 6 and 7), following additional calculations. Fourthly, changes between assemblages, particularly in respect of the minor phases, require mass transport and structural reconstitution. This creates a series of buffering reactions within the cement paste matrix. For example internal buffering systems exist for hydroxyl, sulfate and carbonate: while the first of these is relatively well-known, the existence of buffers for other species and the consequences of buffering with respect to alteration has been less well recognised. The workings of these buffer systems and their interactions, as well as their consequences to environmentally-conditioned reactions of cement systems, will undoubtedly assist the future development of chemical/ mineralogical models of cement performance. Finally, the database derived in this Thesis redefines the stability of sulfoaluminate hydrates and shows that sulfate interacts strongly with carbonate. The undoubted persistence of sulfate-AFm amongst the hydration products of commercial cements has long been at variance with its supposed instability at < 40ºC: it has been necessary to attribute its persistence to metastability. However, this work shows monosulfoaluminate hydrate to be more stable than hitherto supposed, with a lower limit of thermal stability at ~5±5ºC in the system CaO-Al2O3-CaSO4-H2O. This revision immediately resolves a long-standing conflict. However the importance of treating sulfate in conjunction with carbonate is emphasised and the sharp temperature dependence of sulfate-carbonate phase relations is highlighted.

Albert, et al. [4], have recently suggested that calcium monosulfoaluminate was preserved to lower temperatures, <40ºC, by the reduced activity of water in hydrated cement paste. While they direct attention to an important factor, it is not in fact necessary to invoke special conditions to explain the stability of sulfate-AFm according to the revised data. However, it is agreed with these authors about the importance of extending knowledge of cement hydration to regimes characterised by having reduced water activities, as for example may occur in alkali-activated systems or in normal compositions at low w/s-ratios.

The relative thermodynamic stabilities of the cement hydrates are very sensitive to their chemical environment. Care has to be taken that the calculated temperature-dependent stability regions for each mineral depend on the total chemical composition of the system and cannot be extrapolated to more complex chemical environments, such as those in commercial cement. Two examples will be discussed briefly:

Development of a thermodynamic database for cement hydrates 51 may help explain why enhanced ettringite formation appears to be a precursor to thaumasite: reaction proceeds in two stages by replacement of sulfate in ettringite by carbonate followed by silicate replacement of aluminate, with conversion of carbonate-AFt to thaumasite.

The influence of silica on the stability of hydrate phases containing calcium and aluminium Silica is one of the main constituents of Portland cements. Fig. 4.33 a) shows free energy plots of the reaction of C3A, with gypsum and water. As noted, in the silica-free system CaO-Al2O3-CaSO4-

H2O, with an initial molar bulk SO3/Al2O3 ratio = 1, monosulfoaluminate is more stable than the phase assemblage of ettringite (C6AsH32) and C3AH6 at temperatures > ~5°C (reactions (2) and (3), Fig. 4.33 a). But if silicon is added to the system, according to reaction (1) the phase assemblage of

C6AsH32 and C3AS0 8H4 4, which represent a member of the solid solution series

Ca3Al2(SiO4)3-x(OH)4x, has a lower Gibbs free energy of reaction and is therefore thermo- dynamically more stable than monosulfoaluminate (reaction (2)) at temperatures from 1 to 99°C. In the system CaO-Al2O3-CaCO3-H2O, addition of silicon causes an analogous thermodynamic stabilisation of the hydrogarnet phase (Fig. 4.33 b)). Whereas in the silicon-free system, monocarboaluminate is calculated to be more stable than the assemblage of C3AH6 and calcite at temperatures below ~85°C, if silica is added according to reaction (4), C3AS0 8H4 4 and calcite become more stable than monocarboaluminate. Thus calcite is predicted to be an essentially inert material in the system CaO-Al2O3-CaCO3-SiO2-H2O at all temperatures, 1-99°C. a) system CaO-Al2O3-CaSO4-SiO2-H2O b) system CaO-Al2O3-CaCO3-SiO2-H2O

280 280 (1) C A + CaSO .2H O + 0.533 SiO +11.6H O 3 4 2 2 2 (4) C3A + CaCO3 + 0.8 SiO2 +4.4H2O --> 0.33C As H + 067C AS H 260 260 6 3 32 3 0.8 4.4 --> C3AS0.8H4.4 + CaCO3 240 240 (5) C A + CaCO +11H O --> C AcH (2) C A + CaSO .2H O +10H O --> C AsH 3 3 2 4 11 3 4 2 2 4 12 220 [kJ/mol] [kJ/mol] r r 0

0 220 G

G 200 Δ Δ - - 200 (6) C3A + CaCO3 +6H2O (3) C3A + CaSO4.2H2O +12.666H2O 180 --> C3AH6 + CaCO3 -->0.33 C6As3H32 + 0.67C3AH6 180 160 0 102030405060708090100 0 102030405060708090100 Temperature [°C] Temperature [°C]

Fig. 4.33: Influence of silicon on the relative stabilities of aluminate hydrates. Probably due to kinetic reasons, siliceous hydrogarnet is rarely observed in hydrated Portland cement at room temperature. Numerous researchers have shown that AFm phases are abundant reaction products of Portland cements. In agreement with experimental results, sulfate- and carbonate-AFm phases are observed as persistent phases in Portland cement. It is not possible to extrapolate stability regions obtained from simplified model systems with two or three components directly to complex chemical systems, e.g. to commercial Portland cement, with numerous components present. However, with the knowledge of the thermodynamic properties of all potential reaction products and a suitable software, it is possible to undertake the additional calculations necessary to predict and quantify the phase assemblages. On the other hand, as shown above, the database cannot be used as a “black-box” approach and fundamental knowledge about the processes of cement hydration, including knowledge of reaction pathways and

Development of a thermodynamic database for cement hydrates 52 metastable vs. stable phase assemblages -the latter based on experimental observations- is necessary to ensure that the calculations are appropriate.

4.6. Concluding remarks

A short review of the obtained dataset showed that the composition of the AFm phase varies with chemical environments. The stability of the different AFm phases is temperature-dependent. Furthermore the stability of the AFm phases with respect to other cement hydrates, e.g. AFt and hydrogarnet needs to be investigated.

Thus an extensive work programme was set up to investigate the AFm and AFt phase assemblages in Portland cement. Using the presently available thermodynamic data it is assumed that C-S-H and portlandite, the main constituents of Portland cement, do not interfere with AFm and AFt phases, except with respect to sulfate sorption and aluminium substitution in C-S-H, and in the formation of thaumasite. In the case of thaumasite formation, a correction has been applied (see section 6.3) to accommodate this.

Systematic investigations, presented in subsequent chapters of this work, dealt with the following aspects:

• Investigation of AFm phase assemblages in Portland cement. The possibility of solid

solution formation between AFm phases in the system CaO-Al2O3-CaSO4-CaCO3-H2O at 25°C and the impact on Portland cement hydration was investigated. Generic paradigms to the distribution of sulfate and carbonate in dependence of the chemical composition of Portland cement were developed.

• The formation of solid solutions between SO4-AFt, CO3-AFt and thaumasite was examined.

• Another important scope of this work was to investigate the impact of calcite (limestone) additions on the hydration of Portland cement. By calculating specific volume changes, the investigations are closely linked to mechanical and physical properties of the hydrated cement paste.

• The influence of elevated temperatures in the range 5°C to 85°C on phase assemblages in hydrated Portland cement containing limestone was subject of investigation.

The AFm phase in Portland cement 54

Different anions can be integrated into the AFm interlayer with the general composition - - [X⋅mH2O] where X denotes a singly-charged e.g. OH , or half of a doubly charged anion, e.g. ½ 2- - 2- SO4 . The most important anions present in hydrating OPC and blended cements are OH , SO4 2- and CO3 , to form hydroxy-AFm C4AHx, monosulfoaluminate C4AsHx and monocarboaluminate - 2- 2- C4AcHx respectively [180]. Anions like Cl , CrO4 and SeO4 can be bound in the crystal structure of AFm phases [19][112][146] but are not included in the scope of this investigation. Pöllmann [149] also shows that organic anions e.g. carboxylic acid groups can be incorporated into an AFm- like structure although the layer spacing is vastly increased to accommodate the organic chains. In the presence of amorphous silica or silica-containing blends, e.g. blast furnace slags or metakaolin, the formation of C2ASH8, strätlingite, or gehlenite hydrate, is observed [199]. The crystal data [111] of strätlingite show it to be an AFm phase with an aluminosilicate anion in the interlayer. Due to changing hydration state in response to the relative humidity of the environment, a careful interpretation of experimental results with regard to the initial hydration state is necessary. To enable phase analysis without carbonation, e.g. by X-ray diffraction, the solid may need to be dried and, as a consequence, the water content of the interlayer may be affected. For example, C4AHx is likely to reduce its water content from x = 19 to x = 13 water molecules at relative humidity < 88% and ∼20°C [180]. As many experiments were carried out at relative humidity < 100% the initial hydration state of C4AHx is often referred to x = 13 and therefore an artificially low water content may be observed. On the other hand, the high ionic strength of the pore solution of hydrated OPC decreases the activity of water and favours a lower hydration state of the AFm phase [79]. Thus care is required to preserve the initial hydration state during subsequent analysis of the solid; if this is not possible any changes of the hydration state during the analysis should be recorded. The ability to form solid solutions is dependent on mechanism as well as upon the thermodynamics. It is well known that AFm phase compositions are sensitive to anion exchanges with their environment and formation of solid solutions between AFm end members cannot be neglected. One of the most important AFm phases in hydrated OPC is “monosulfoaluminate”, with stoichiometric composition C4AsHx. But several investigations describe formation of solid solutions between monosulfoaluminate and other AFm phases in hydrated OPC, e.g. hydroxy-AFm, C4AHx, or monocarboaluminate, C4AcHx.

5.1.2 Stability of AFm phases The absolute stability of selected AFm compositions is shown in Table 5.1. The Table needs to be used in conjunction with supplementary information. For example, cements normally contribute a high hydroxyl content to the mix water so OH- is invariably present at high concentrations. Most modern cements contain sulfate, added as calcium sulfate, to control set times. But in the cement environment sulfate is sparingly soluble and carbonate even less soluble at high pH. Hence these species, carbonate and sulfate, have low aqueous concentrations, but nevertheless compete for anion sites in AFm structures. Carbonate sources include i) alkali “sulfates” condensed onto clinker in the course of cooling or from added clinker kiln dust; these “sulfates” are in fact solid solutions with carbonate partly replacing sulfate [18] ii) carbonates gained by reaction with the atmosphere iii) carbonate added as a permitted extender, normally in the form of interground calcite and iv) carbonate present as impurity in gypsum. Damidot [49] showed that with rising carbonate activity, hydroxy-AFm was replaced, first by hemicarboaluminate and then by monocarboaluminate. As the carbonate content of the AFm solid increased stepwise, its stability at 25°C also increased.

The AFm phase in Portland cement 55

Table 5.1: Thermodynamic stabilities of selected AFm phases at 25°C Distinctive Formula Designation Stability at 25°C and 1 Reference Anion bar

- OH C3A·Ca(OH)2·xH2O hydroxy-AFm Unstable - decomposes to [193]

hydrogarnet C3AH6, and portlandite

- 2- OH ; CO3 C3A·Ca[(OH)(CO3)0 5]·xH2O hemicarboaluminate Stable - but not many [49] thermodynamic data available; unstable in contact with calcite

2- CO3 C3A·CaCO3·xH2O monocarboaluminate stable [49]

2- SO4 C3A·CaSO4·xH2O monosulfoaluminate Calculated to be stable [47] above 40°C - at lower temperatures decomposes

to AFt, C3AH6 and gibbsite

stable at 25°C This study

- [AlSi(OH)8] C2ASH8 gehlenite hydrate, stable [48] strätlingite

- Cl C3A·CaCl2·xH2O Friedel`s salt Stable, but not described in [45] this work. Mainly occurs as an alteration product of cement paste in chloride- rich environments

This pattern of stability means that, in addition to taking into account the bulk chemistry of cements, one must also note that some AFm compositions are metastable under all conditions and are prone to decomposition or reaction while others have a definite range of stability under conditions relevant to the internal environment achieved in commercial cements. The impact of small amounts of carbonate on the nature and stability of the AFm phase is noteworthy; Table 5.2 shows selected composition data for AFm phases, from which it can be seen that ideal hemicarboaluminate and monocarboaluminate contain only 3.9 and 7.7 wt.-% CO2 respectively.

Table 5.2: Sulfate- and carbonate-contents of minerals relevant to Portland cement

Carbonate content Sulfate content Mineral Formula CO2 [wt.-%] SO3 [wt.-%]

Gypsum CaSO4·2H2O - 46.5

Ettringite Ca6Al2(SO4)3(OH)12·26H2O - 19.1

Monosulfoaluminate Ca4Al2(SO4)(OH)12·6H2O - 12.9

Calcite CaCO3 44.0 -

Monocarboaluminate Ca4Al2(CO3)(OH)12·5H2O 7.7 -

Hemicarboaluminate Ca4Al2(CO3)0 5(OH)13·5.5H2O 3.9 -

The AFm phase in Portland cement 56

5.1.3 Solid solutions between AFm phases Solid solutions between monosulfoaluminate and hydroxy-AFm The literature reveals considerable divergence of interpretation concerning the existence and extent of solid solution between OH- and SO4-AFm phases, arguably the most important pair with respect to the constitution of fresh, carbonate-free Portland cements.

Jones [97] published results of an investigation of the system CaO-Al2O3-CaSO4-H2O at 25°C. Microscopic examination of the hexagonal calciumsulfoaluminate hydrates showed changing refractive indices of the solid phases with changing initial CaSO4/Al2O3 ratios. Therefore it was concluded that a complete solid solution series exists between monosulfoaluminate and hydroxy- AFm.

D’Ans and Eick [51] found similar results in their microscopic examination of the quaternary CaO-

Al2O3-CaSO4-H2O at 20°C. They concluded the existence of complete miscibility between both end members, C4AsHx and C4AHx.

Turriziani [190] synthesised a solid solution series between C4AsHx and C4AHx following the method used by D’Ans and Eick [51], with changing molar CaSO4/Al2O3 ratios between 0.22 to 1.0 and higher. The microscopic studies of samples cured for 30 days apparently confirmed the results of D’Ans and Eick [51] showing complete miscibility between end members. But in contrast to the optical properties, the X-ray data showed two immiscible phases with patterns similar to those of mixtures of sulfate- and hydroxy-AFm. A direct comparison to the results from Jones [97] and D’Ans and Eick [51] cannot be made as X-ray data were not obtained in D’Ans and Eick’s studies, but Turriziani’s data are internally inconsistent.

Roberts [163] published a detailed examination of the formation of AFm solid solutions. A solid solution series between C4AsHx and C4AHx was prepared by adding increasing amounts of gypsum to a suspension of C4AHx and water. After ageing the mixture for at least 7 days, the subsequent X- ray investigation showed a shift of the d-space of the main basal reflection of the solid solution d0001 from ~ 9.0 Å to 8.77 Å with decreasing molar CaSO4/Al2O3 ratios from 1 to 0.5. At CaSO4/Al2O3 ratios ≤ 0.5 two phases were observed: the limiting solid solution with a d-spacing of 8.77 Å and

C4AH19. Thus it was believed that there is a miscibility gap occurred between C4AHx and a limiting solid solution, having the approximate composition C4A⋅½SO3⋅xH2O.

Seligmann [169] synthesised a similar C4AsHx-C4AHx solid solution series by hydration of C3A with varying amounts of Ca(OH)2 and CaSO4⋅2H2O. In comparison with Roberts’ work, a limiting 3 solid solution composition of C4A⋅ /5 SO3⋅xH2O was reported.

Pöllmann [150] confirmed Roberts’ results. A C4AsHx-C4AHx solid solution series was prepared, as described above, by hydration of C3A or CA with Ca(OH)2 and CaSO4⋅2H2O for 8 months. A miscibility gap between CaSO4/Al2O3 ratios from 0 to 0.5 was concluded from least square refinements of X-ray patterns of the aged precipitates. The limiting solid solution with, a d0001 value of 8.76 Å, had the approximate composition C4A⋅½SO3⋅xH2O and was termed “hemisulfate”. Furthermore it was shown that there is a strong temperature dependence of the extent of solid solution between C4AsHx and C4AHx. The width of the miscibility gap increased at 60°C to include ratios between 0 to 0.8. However, at temperatures ≥ 45°C, C4AH19 was not stable and decomposed slowly to a mixture of C3AH6 and Ca(OH)2 in compositions within the supposed miscibility gap. At

The AFm phase in Portland cement 57

80°C no miscibility of C4AsHx and C4AHx was observed. The X-ray analysis of the precipitates showed a mixture of C4AsHx with C3AH6 and Ca(OH)2.

More recently, Zhang [197] synthesised solid solutions between sulfate- and hydroxy-AFm according to Pöllmann’s method, by mixing stoichiometric amounts of CA, Ca(OH)2 and

CaSO4⋅2H2O. Solids and solutions were analysed after ageing for ~16 months at 25°C. The XRD- pattern showed, in contrast to the results from Pöllmann and Roberts, C4AsHx and C4AHx occurring as two immiscible phases. Furthermore ettringite, C6As3H32, was found in sulfate-rich compositions and C3AH6 and Ca(OH)2 at the hydroxide-rich end of the notional solid solution series. Additionally the author measured the conductivity of the mother liquor of the solid solution with CaSO4/Al2O3 ratios of 1 and 0.5. After a reasonable time (100-200days) for equilibration the conductance of the solution was still decreasing and had not stabilised by the end of the investigations, after ~ 260 d. This behaviour was interpreted as arising from an ongoing phase transformation. Therefore the authors supposed that initially solid solutions formed but that these decomposed in the course of the experiments. But, due to missing phase analyses in the first stages of the experiment, this could not be proved.

Glasser et al. [79] prepared solid solutions between sulfate- and hydroxy-AFm by mechanical mixing of C4AH13 and sulfate-AFm and by mixing appropriate quantities of C3A, Ca(OH)2 and

CaSO4⋅2H2O. The samples were stored 9 months at 5°C and then analysed by analytical electron microscopy. A plot of the sulfate content of all analysed samples showed two population clusters, one in the range of 0-20 mol% sulfate and the other between 45-90 mol% sulfate. The results were interpreted as indicating incomplete solid solution between hydroxy- and sulfate-AFm end members: the hydroxy-AFm phase showed a mean sulfate content of 6 mol% and the sulfate-AFm phase had a mean composition of ~75-90 mol% sulfate. The substantial scatter of the analysis data was ascribed either to an intergrowth of sulfate- and hydroxy-type-AFm, or to equilibrium not having been obtained despite 9 months reaction time, or both.

Kapralik and Hanic [100] investigated the hydration products of sulfoaluminate clinker, C4A3s, with increasing addition of CaO. The mixtures were prepared with a water/solid-ratio of 10:1 and stored at 25°C for 18 days. X-ray patterns showed a compound with similar crystallographic properties

(d0001 ~ 8.79 Å) to those of the limiting solid solution C4A⋅½SO3⋅xH2O reported by Pöllmann and

Roberts. Furthermore, a second hydrate with a d0001 value of ~10.52 Å was found and attributed to a higher hydrated form of the solid solution.

Stark et al. [176] investigated the hydration of OPC at different temperatures and showed the formation of a tetracalciumsulfoaluminate hydrate solid solution. Additionally to AFm solid solutions, ettringite, portlandite and C-S-H were observed after 24 hours hydration. Due to the variance of the electron microprobe analysis an exact quantification of the amount of substituted sulfate in AFm was not possible. Further investigations at 62°C indicated no AFm solid solution formation; monosulfoaluminate and hydrogarnet, C3AH6, were observed among other hydration products: ettringite, portlandite and hydrotalcite. At 5°C C4AHx was the only AFm phase detected in hydrated OPC after 21 d. Thus the solid solution between C4AsHx and C4AHx will be formed over a limited temperature range from 5 to 65°C.

The AFm phase in Portland cement 58

The studies thus far described were made at self-generated pH values, typically close to 12. However Kalousek [99] has studied AFm reactions with added alkali. These may well be relevant to “real” cements in which the aqueous ratio OH/SO4 is affected by readily soluble alkalis. It was shown that significant substitution of sulfate by hydroxyl occurred in the presence of alkalis, effectively NaOH. For example, the sulfate content of the monosulfoaluminate-like phase decreased from 1 mol SO3 when no additional alkalis were present, to 0.22 mol in ~ 0.6 N NaOH-solution. Aqueous calcium concentrations decreased while aluminium and sulfate concentrations increased with rising aqueous alkali contents. As the amount of NaOH increased, analysis of the solids showed a nearly constant sulfate content of 0.66 mol, corresponding to the composition C4As0 66Hx in the range between 0.15 N NaOH to 0.27 N NaOH. A further increase of the NaOH concentration, 2- to 0.6 N NaOH, led to a reduction of the total sulfate in the solid phase to 0.22 mol SO4 . Kalousek [99] concluded that complete solid solution occurred between monosulfoaluminate and hydroxy-

AFm. But the corresponding XRD-patterns showed a rapid decrease of the d0001 values from 8.9 Å for 1 mol SO3, i.e. monosulfoaluminate, to ~8.0 - 8.2 Å at ~ 0.5 - 0.6 mol SO3 in the solid phase: no further decrease was observed, despite rising aqueous alkali concentrations. These values might be an indication of a miscibility gap at low sulfate contents, less than ~0.5 - 0.6 mol SO3 in the solid phase, although no second phase was identified. But the solids were investigated after only 48 hours of equilibration; their X-ray peaks were very broad: the solids were probably poorly crystallised and the accuracy of measuring d0001 correspondingly poor. Furthermore, Kalousek [99] worked with highly concentrated NaOH-solution. Dosch [55] showed the existence of the U-phase, a sodium substituted AFm phase, which may precipitate at high sodium concentrations. The separate identity of the U-phase - not known to Kalousek [99] at the time - may therefore affect the interpretation of these data.

AFm phases and related solid solutions in carbonate-containing systems According to EN 197, CEM I may contain up to 5% minor constituents; including limestone.

Furthermore other types of blended cements containing up to 35 % CaCO3 are permitted by EN 197. Therefore carbonate-containing systems are important and have to be considered in hydrate phase assemblages, even in the absence of external carbonation.

The most important carbonate-containing AFm phase is monocarboaluminate, C4AcHx. It is found as a reaction product of the hydration of C3A with CaCO3. The structure is similar to that of 2+ monosulfoaluminate; it is fully ordered and is composed of positively charged [Ca4Al2(OH)12] 2- 3+ 2+ main layers and negatively charged [CO3⋅5H2O] interlayers. The Al and Ca ions are six- and 2- sevenfold coordinated, respectively, by oxygen atoms. The planar CO3 groups are directly connected to the main layers and are tilted by 21.8 ° from the plane of the layers [68].

Investigations of possible solid solution formation between monocarboaluminate and hydroxy-AFm by Roberts [163] showed no solid solution. However at CaCO3/Al2O3 ratios < 1, a second compound with a d0001 value of ~8.2 Å, was found as an additional reaction product. According to its composition this hydrate was termed “hemicarboaluminate”, C4Ac0 5H11. The X-ray patterns of the solids showed a mixture of mono- and hemicarboaluminate up to CaCO3/Al2O3 ratios of 0.5; the amount of hemicarboaluminate increased until at a CaCO3/Al2O3 ratios of 0.5 only this hydrate occurred. At CaCO3/Al2O3 ratios <0.5, the amount of hemicarboaluminate again diminished and

C4AHx was present as an additional compound. Due to its fixed composition over a wide range of

CaCO3/Al2O3 ratios, hemicarboaluminate should be referred to as a distinct compound.

Furthermore, Roberts showed that the so-called α-C4AH13 with a d0001 value of ~8.2 Å, first described by Wells et al. [193] as a polymorph of C4AH13, is probably hemicarboaluminate.

The AFm phase in Portland cement 59

Later works by Fischer [63] verified Roberts’ results. Dosch et al. [56] obtained similar results but the stoichiometrically calculated composition of the 8.2 Å hydrate was given as C4Ac0 25H12. The discrepancy between Roberts’ and Dosch’s results may be explained by additional carbonation occurring during storage or X-ray examination, resulting in a subsequently higher CO3 content in the system than initially calculated.

Thermodynamic calculations by Damidot et al. [49] showed that hemicarboaluminate is only persistent below the self-generated carbonate activities of calcite. Thus the formation of hemicarboaluminate becomes unlikely in real systems where CaCO3 is present; the activity of carbonate is sufficient to stabilise monocarboaluminate.

Experimental results by Pöllmann [148] show that there is probably no solid solution between the monocarboaluminate and the monosulfoaluminate end member. Both hydrates coexisted with unchanged X-ray patterns over a range from 0

Kuzel et al. [117] investigated the hydration of C3A in the presence of Ca(OH)2, CaCO3 and

CaSO4⋅2H2O. At a molar composition C3A:CaO:CaSO4:CaCO3 = 1:1:1:2/3 the final hydration products consisted of ettringite, monocarboaluminate and portlandite after 720 h hydration. If the amount of CaCO3 was reduced to C3A:CaO:CaSO4:CaCO3=1:1:1:1/3, ettringite, hemicarboaluminate, small amounts of monocarboaluminate and portlandite were found as reaction products. Further reduction of the CaCO3 content, in connection with a reduced sulfate content to a molar composition of C3A:CaO:CaSO4:CaCO3=1:1:1/2:1/4, led to the formation of a mixture of monosulfoaluminate, hemicarboaluminate and portlandite. Monocarboaluminate was not observed as reaction product. Finally, if the sulfate content was diminished to

C3A:CaO:CaSO4:CaCO3=1:1:1/4:1/4, the limiting solid solution between monosulfoaluminate and hydroxy-AFm as well as hemicarboaluminate and portlandite occurred; formation of solid solutions between carbonate- and sulfate-AFm were not observed. It is therefore probable that monocarboaluminate and monosulfoaluminate will not occur together in real systems, as even very low CO2 contents are apparently sufficient to suppress formation of monosulfoaluminate in cements.

5.1.4 Summary and conclusions from the literature There has been a tendency to regard the constituents of a hydrated cement paste as consisting of four main hydrate phases; Ca(OH)2, C-S-H, AFt and AFm. This simplification, in which the AFm phase is regarded as a single phase with chemical variants, is perhaps satisfactory for a broad-brush approach, but does not conform to reality and is too simplistic to develop for a quantitative model of cement paste. However, the number and constitution of the AFm family of phases is subject to a range of partially conflicting conclusions. It is noted that the internal inconsistencies in interpretation of the evidence contained in some papers and, if the evidence is reinterpreted it is found that:

• The cation chemistry of AFm phases of cement and blended cement is relatively simple; the divalent ion is overwhelmingly calcium and the trivalent ion, aluminium.

The AFm phase in Portland cement 60

• The hydration state of AFm is variable, but at the moderately high ionic potential, hydroxyl ion concentration and pH values, typical of Portland cement, a satisfactory representation - of the formula unit is Ca4Al2(X)(OH)12.nH2O where x = 2 OH or one divalent anion, e.g., 2- 2- CO3 or SO4 and n= 5-6 H20.

• With few exceptions, AFm phases do not form significant solid solutions. Thus the AFm

phases are conveniently distinguished by means of a prefix, e.g., OH-AFm, SO4-AFm,

CO3-AFm or in the case of hemicarboaluminate, OH/CO3-AFm.

• An exception has to be made for SO4-AFm which forms extensive solid solutions with OH- AFm, extending to ~50 mol% replacement of sulfate. However, the resulting solid solutions may be labile and are not persistent over 1-3 year timescales at ~20ºC.

• The absolute order of thermodynamic stability of AFm variants at ~25ºC appears to be, in order of increased stability:

OH-AFm < SO4-AFm < CO3-AFm

• The anion constitution of AFm is sensitive to the chemistry of cement and responsive to interactions between cement and its service environment.

The AFm phase in Portland cement 61

5.2. Preparation of solid solutions

A focus of the current work was to determine if significant solid solution occurred between AFm end members. Phase relations in the following systems were considered as to be the most important for cement systems:

• Monosulfoaluminate - hydroxy-AFm • Monosulfoaluminate - monocarboaluminate • Monocarboaluminate - hydroxy-AFm Two approaches have been applied to synthesise solid solutions (see Table C.1 - Table C.5 for exact amounts used in the experiments): i) by mechanical mixing of previously synthesised AFm end members (monosulfoaluminate,

C4AsHx; monocarboaluminate, C4AcHx and hydroxy-AFm, C4AHx) ii) by precipitation from supersaturated solutions, formed by mixing appropriate amounts of

previously synthesised C3A, CaO, CaSO4 or CaCO3, respectively, depending on the desired phase or phase assemblage. Approach i) was used to clarify whether reaction occurs, i.e. solid solution formation occurred spontaneously between the AFm phases; also, where reaction was complete, to derive solubility data. Approach ii) was applied to solubility determinations and derivation of thermodynamic data and to conduct experiments on the extent of solid solution formation between AFm phases. Additional solubility data commencing from undersaturation were obtained by redissolving the reaction products obtained from the mixed end member method. The AFm solid solution experiments were made at 25°C. Water/solid-ratios were restricted to < 30 to reduce errors induced by incongruent dissolution. Solid and aqueous phases were analysed and prepared according to the experimental procedures described in chapter 3.

The AFm phase in Portland cement 64

27.0 5.79 780.0 [Å] ] [Å]

3 0 26.8 5.78 0 775.0 [Å C A

770.0 26.6 C 0 5.77 765.0 26.4 5.76 A 0 760.0 26.2 5.75

calc. cell volume 755.0 lattice parameter lattice parameter 26.0 5.74 750.0 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1

C4AHx C4AsH12 calc. S O 4/(S O 4+2OH) ratio C4AHx calc. S O 4/(S O 4+2OH) ratio C4AsH12

Fig. 5.5: Refined lattice parameter for the mono- Fig. 5.6: Calculated volume change for the sulfoaluminate type phase with changing monosulfoaluminate type phase with changing SO4/(SO4+2OH) ratio SO4/(SO4+2OH) ratio

Volume changes in the unit cell due to the solid solution formation were obtained using the software CELREF, to refine the obtained XRD-pattern by least squares minimisation. The calculated results (Fig. 5.5) show that the decrease of sulfate content tends to result in a decrease of the unit cell lattice parameters up to a nearly constant size at SO4/(SO4+2OH) < ~0.5 - 0.6. As shown in Fig. 5.6, the total volume of the unit cell decreased ~ 2.5 % with increasing replacement of sulfate by hydroxide. As no standard was used to calibrate measurements for experimental deviations of the 2θ values, these results are best used to indicate trends of changing unit cell size with increasing sulfate substitution.

ESEM micrographs of the monosulfoaluminate end member and its solid solutions, Fig. 5.7 (a) and (b) show well-crystallised hexagonal, platy crystals with diameters between 4 - 30 μm. In agreement with XRD data, needle-like AFt prisms with lengths up to ~5 μm were found randomly distributed in the monosulfoaluminate preparation, Fig. 5.7 (a). In contrast to the end member preparation, no ettringite was observed when ~10 mol% of the sulfate was substituted by hydroxide i.e., at a solid molar SO4/SO4+2OH ~ 0.9, Fig. 5.7 (b).

(a) (b)

Fig. 5.7: ESEM micrographs of the monosulfoaluminate end member and ettringite (a) and the monosulfo- aluminate type solid solution (~10 mol% of sulfate substituted by hydroxide) (b)

The AFm phase in Portland cement 65

Thermodynamic modelling of the solid solution formation

With the aid of the previously derived thermodynamic data of monosulfoaluminate and hydroxy- AFm (Table 4.1) and the Redlich-Kister non-ideal solid solution model (chapter 4.3.4) it is possible to estimate the thermodynamic properties of the solid solution.

The dissolution reaction for the non-ideal monosulfoaluminate-hydroxy-AFm solid solution series is shown in Eq. 5.1:

2+ - -2 - x424 − 2x214 ⎯⋅ ⎯→ 4CaOH6)OH()SO(AlCa 2AlO2 4 +++ 2x)OH-6(xSO + 2 O10H (5.1) In agreement with own experimental data and literature data by Roberts [163] and Pöllmann [150] ~ 50 mol% of sulfate in the monosulfoaluminate-phase can be replaced by hydroxide. However the data about sulfate substitution in the original hydroxy-AFm phase are not consistent. Pöllmann found no sulfate substitution. However according to investigations in this Thesis the pure hydroxy-

AFm end member tends to decompose relatively rapidly; C3AH6 is observed in the XRD-pattern Fig. 5.3 and ESEM micrographs (Fig. 5.8) after 4 weeks of hydration. Furthermore calcium concentrations of the aqueous phase are significantly higher and approach portlandite saturation compared to the preparation at low sulfate contents at SO4/(SO4+2OH)=0.05 (see solubility data Table A.11). This indicates that minor sulfate substitution stabilises hydroxy-AFm. However this amount of sulfate is too small to be quantified by the experimental methods used in this work. Thus the existing miscibility gap of the solid solution series (according Eq. 5.1) is provisionally defined as x1 ≤ x ≤ 0.5. But, as described below, x1 can be estimated with the help of Lippmann diagrams.

Fig. 5.8: ESEM micrograph of cubic C3AH6 crystals on the surface of a crystal of the hydroxy-AFm end member

As shown by Glynn [80], Lippmann phase diagrams are a useful instrument to fit the dimensionless parameters a0 and a1, necessary to calculate the activity coefficients of the members of the solid solution series. According to the mathematical algorithm described in chapter 4.3.4 the “total solubility product” of the monosulfoaluminate-hydroxy-AFm solid solution series can be derived from the dissolution reactions (see Table 4.3) of monosulfoaluminate and hydroxy-AFm (Eq. 5.2; [ ] denote activities of the aqueous species):

- 2− =Π [Ca + 42 2 − 4 10 +[OH]][SO]OH[][OH][AlO] 2 ∑ 2 2 ( 4 ) (5.2) Solubility data for the solid solution series need to be extracted to enable a calculation of the total solubility product. Solubilities (Table A.10) were obtained as described above from initially supersaturated and undersaturated solutions with respect to the solid solution series. As shown in Table A.10 and Fig. B.1 (see appendix B) the samples from undersaturation and mixed end members did partly carbonate, indicated by the presence of hemicarboaluminate. Thus only the

The AFm phase in Portland cement 67

The deviations of the calculated data points using the inbuilt non-ideal solid solution model of GEMS-PSI (white markers, dashed line, Fig. 5.9) can be explained by differences in the solid

SO4/(SO4+2OH)-ratio. The theoretical solidus of the Lippmann diagram (black solid line) is plotted as function of the theoretical solid solution composition; assuming that no other solids are present. However in the experiments as well as in the calculations using GEMS, AFt was calculated to be present at SO4/(SO4+2OH)-ratios ≥ 0.98 and C2AH8 was observed and calculated to be present at

SO4/(SO4+2OH)-ratios ≤ 0.70. To enable a direct comparison between calculated and experimen- tally derived total solubility products, both values were plotted against the initially calculated molar

SO4/(SO4+2OH)-ratio of the bulk composition. Thus the dashed line, Fig. 5.9, was used to fit the calculated total solubility products, including the a0 and a1 parameters, to the experimental values, as it was not possible to analyse accurately the composition of the solid solution.

With the help of the previously determined fitting parameters a0 and a1 and the known activities of the sulfate and hydroxide-species, it is now possible to calculate the experimentally-derived solutus of the Lippmann diagram (Fig. 5.9) using Eq. 5.3:

⎛ X ⎞ SO2− aq, X − ⎜ 4 aq,OH ⎟ sl,eq =Π 1 + (5.3) ∑ ⎜ K γ K γ ⎟ ⎝ C4AsH12 C4AsH12 C4AH13 C4AH13 ⎠ where KC4AsH12 and KC4AH13 are the solubility products of the end members, γC4AsH12 and γC4AH13 are the activity coefficients of the end members and XSO42- , aq and XOH-, aq are the “aqueous activity 2- - fractions” of the substitutable species SO4 and OH calculated using Eqns. 5.4 and 5.5 (where [ ] denote calculated activities of the aqueous species:

2− 4 ]SO[ X = (5.4) SO2− aq, 2− − 2 4 4 + ]OH[]SO[ − ]OH[ 2 X = (5.5) − aq,OH 2− − 2 4 + ]OH[]SO[ As shown in Fig. 5.9 experimentally-derived values (solutus markers, Fig. 5.9) agree well with the theoretically computed solutus using Eqns. 4.25-4.27. Originally Lippmann diagrams were used to express thermodynamic equilibrium states between the solid solution composition and the related aqueous phase. For example, one can determine the activity fractions of the sulfate and hydroxide species of the related aqueous phase by drawing a horizontal line from the solidus of the solid solution with known composition to the related solutus and read off the values for XSO4,aq on the abscissa. Thus, as shown above, Lippmann diagrams are a useful instrument to describe the thermodynamic behaviour of solid solutions and to derive missing parameters necessary to illustrate the thermodynamics of SSAS. However, due to existing uncertainties about the exact upper and lower limit of the miscibility gaps, the derived numerical values should be regarded as provisional because they are sensitive to changes of the compositional boundaries of the miscibility gap. Furthermore the sulfate concentration of many samples was 2- below the limit of detection (< ~0.002 mmol SO4 /l). Thus sulfate concentrations had to be estimated by applying the non-ideal solid solution model at SO4/(SO4+2OH)-ratios ≤ 0.8 . However 2- due to the supposedly very low values (<<0.002 mmolSO4 /l) this has only minor impact on the total solubility product of the solid solution.

The AFm phase in Portland cement 68

hydroxy-AFm type ss miscibility gap monosulfoaluminate type ss lim.monosulf.- and hydroxy-AFm type ss 1.0

0.5 ΔGex 0.0

-0.5

ΔGM -1.0 G (kJ/mol) Δ

-1.5 ΔGid -2.0

-2.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 C AH C AsH 4 x SO4/(2OH+SO4)-ratio 4 12

Fig. 5.10: Gibbs energy of ideal mixing, ΔGid, excess Gibbs energy of mixing, ΔGex, and resulting molar Gibbs energy of mixing, ΔGM, calculated from Eq.4.14 for the monosulfoaluminate-hydroxy-AFm solid solution series

With the help of the previously derived parameters a0 and a1, it is now possible to determine the changes of the (excess) Gibbs energies of mixing. Fig. 5.10 compares the functions of the Gibbs energy of ideal mixing, ΔGid, the estimated excess Gibbs energy of mixing ΔGex and the resulting estimated Gibbs energy of mixing of the monosulfoaluminate-hydroxy-AFm solid solution series,

ΔGM, according to Eqns. 4.14 - 4.16. The molar Gibbs energy of the solid solution can be estimated according to Eq. 4.13 as the sum of the partial Gibbs energies of monosulfoaluminate and hydroxy-

AFm and ΔGm.

The best fit of the calculated solubilities of the monosulfoaluminate-hydroxy-AFm solid solution series, relative to experimental solubility data at 25°C, was obtained with the assumed compositional boundaries of the miscibility gap and the resulting Guggenheim parameters

(a0=0.188 and a1=2.49). As shown in Fig. 5.11, comparison between experimentally-derived solubility data using averaged values from super- and undersaturation and calculated solubilities of the monosulfoaluminate-hydroxy-AFm solid solution at 25°C shows excellent agreement in the range 0.05 ≤ SO4/(2OH+SO4) ≤ 1, which accounts for the good internal consistency of the dataset. Only the solubility data for the pure hydroxy-AFm end member differ significantly from the calculated values. This behaviour can be explained by the metastability of C4AHx with respect to

C3AH6 and portlandite. As this decomposition reaction of C4AHx was only observed in the sulfate- free preparation this might indicate the stabilisation of the hydroxy-AFm end member by substitution of small amounts of sulfate in the structure.

The AFm phase in Portland cement 69

a) b) 20 12.5 1.E+00 pH 1.E-01 15 12.0 2- Ca 1.E-02 SO4 10 limit of detection pH 1.E-03 11.5 5 1.E-04

Al [mmol/l] concentration concentration [mmol/l] concentration

0 11.0 1.E-05 0 0.2 0.4 0.6 0.8 1 00.20.40.60.81

C4AHx calc. SO4/ (SO4+2OH)-ratio C4AsH12 C4AHx calc. SO4/ (SO4+2OH)-ratio C4AsH12

Fig. 5.11:Comparison of calculated (lines) and mean experimental (markers) solubility data for the SO4- AFm and OH-AFm solid solution series at 25°C

Fig. 5.11 b) shows that the aqueous sulfate concentrations are generally very low and at

SO4/(2OH+SO4) ratios ≤ 0.8 they lie below the analytical detection limit. Except for the hydroxy- AFm end member the predicted phase assemblages agreed well with the experimental observations. In agreement with XRD-analysis, small amounts of AFt were predicted to coprecipitate at

SO4/(2OH+SO4) ratios ≥ 0.98; also, traces of C2AH8 were both predicted and observed at

SO4/(2OH+SO4) ratios ≤ 0.70.

Investigations to the long-term stability of the solid solution series According to calculations of this Thesis (see Table 4.4), in agreement with experiments and observations by Wells et al. [193], D’Ans and Eick [52] and Faurie-Mounier [61], hydroxy-AFm is metastable with respect to a mixture of C3AH6 and portlandite at ~25°C. Thus to enable a recalculation of solubility data and solid distribution the metastability of C4AH13 with respect to

C3AH6 was suppressed in the previous computations. This is justified as no formation of C3AH6 was observed in the experiments with initial SO4/(SO4+2OH)-ratios > 0.05 aged for 28 days. On the other hand, Zhang [197] examined samples after prolonged ageing and suggested that initially formed solid solutions were probably unstable and decomposed to the end members:

“…In the sulfate-solid solution series, the enrichment of SO4 in hydrocalumite is likely achieved by disproportionating the initially formed complete solid solution into the OH- and SO4-rich phases.”

In their work the aqueous solubility was analysed at intervals, and was still declining in the final interval measured, it was considered possible that solid solutions might have formed during the early stages of reaction but were not present at the end of the experiments after ~560 d, the only point in time at which samples were subject to X-ray examination. Thus the literature reveals considerable divergence of interpretation concerning the existence and extent of solid solution between OH- and SO4-AFm phases. Hence another aim of this work was to assess the long-term stability of the solid solution series. Thus the composition of aqueous and solid phases of the preparations from supersaturation, which showed no signs of initial carbonation, were analysed again after 1 year and 2 years ageing, respectively.

The AFm phase in Portland cement 75

5.4. Ternary phase relations between sulfate-, carbonate- and hydroxy-AFm

5.4.1 Metastable phase assemblages at 25°C

Fig. 5.21 shows the ternary AFm system consisting of sulfate-, hydroxy-, and carbonate-AFm, calculated for 25°C with the aid of the thermodynamic data in Table 4.1 under consideration of partial solid solution formation between monosulfoaluminate and hydroxy-AFm. Fig. 5.21 is restricted to interactions between these three AFm phases; no other solid phases were admitted. The interactions and phase assemblages between these AFm phases were calculated with a total solid content of 1 mol AFm dispersed in 1 kg water, but are not expected to be sensitive to the relative masses of solid and aqueous phases except at high dilutions, such that incongruent dissolution affects the bulk composition of solids.

A monosulfoaluminate (Ms)

A'

Ms-type ss

limiting Ms-type s s monosulfoaluminate + Ms(ss)+Mc (~ C4As0.5Hx) Ms-type ss monocarboaluminate B + Hc

miscibility gap (limiting Ms-

type ss + C4AHx) Hc + Ms Ms + Hc + Mc

limiting M s-type ss

+ C4AHx+ Hc

Hc + C AH Hc + Mc C 4 x D E C4AHx hemicarboaluminate (Hc) monocarboaluminate (Mc)

Fig. 5.21: Calculated metastable phase assemblages between different AFm phases at 25°C. A possible range of stoichiometry of hemicarboaluminate and C4AHx is not shown. (Note: the region of two solid phases Ms type ss + Hc and Ms(ss) +Mc in which the monosulfoaluminate-type phase has a variable composition; the exact position of point A’ on A-B is strongly dependent from the chosen solid solution model) Fig. 5.21 is best used as a tool to understand the possible interactions between these three AFm phases in response to changes in anion activities. However in the first set of calculations the metastability of hydroxy-AFm and related solid solutions with respect to C3AH6 and portlandite is suppressed to present a complete picture of possible interactions between hydroxy-, sulfate- and carbonate-AFm at early ages of hydration. A minor substitution of sulfate in the hydroxy-AFm end member is neglected in Fig. 5.21 as this has only insignificant impact on the shown phase diagram. A second set of calculation will show the influence of the metastability of hydroxy-AFm and related solid solutions on the phase assemblages.

A significant substitution of sulfate due to the incorporation of hydroxide in monosulfoaluminate characterises region A-C-D, Fig. 5.21. The experimental results show that the sulfate content of the solid solution is variable in the two-phase region A-B-D; the limiting solid solution has a

The AFm phase in Portland cement 76

composition of ~ C4As0 5Hx The solid solution composition remains fixed at ~ 50 mol% in region

B-C-D. As Fig. 5.21 shows, the C4AHx type solid solution will only occur as a phase of essentially fixed composition together with the limiting monosulfoaluminate solid solution and hemicarboaluminate in region B-C-D. C4AHx cannot persist together with both ideal monosulfoaluminate and monocarboaluminate. In a carbonate-free system, C4AHx will react with monosulfoaluminate to form a solid solution according to Eq. 5.6 (with x+y = 1; y ≤ 0.5):

+ H 2O 424 12 2 +⋅ 24 14 2O)nH)OH(AlCa(yO)nH)OH)(SO(Alx(Ca ⎯⋅ ⎯→ +⎯ x424 − x214 ⋅ 2OnH)OH()SO(Aly)Ca(x (5.6)

The addition of monocarboaluminate will initially cause hemicarboaluminate to form as a result of the reaction of monocarboaluminate with C4AHx according to Eq. 5.7:

+ H 2O 24 14 2 +⋅ 324 21 2O)Hn)OH)(CO(AlCa(5.0O)nH(OH)AlaC0.5( ⎯⋅ ⎯→⎯ 5.0324 13 ⋅ 2OHn)OH()CO(AlCa (5.7)

The reaction of monocarboaluminate with C4AHx is generally thermodynamically preferred relative to formation of the monosulfoaluminate-C4AHx solid solution series. Thus, while C4AHx is still present, it will react first with monocarboaluminate to form hemicarboaluminate (Eq. 5.7); any remaining C4AHx will preferentially react with monosulfoaluminate to form a solid solution (Eq.

5.6). Moreover, in the presence of excess monocarboaluminate, all available C4AHx will preferentially react to form hemicarboaluminate; extensive solid solution formation between monosulfoaluminate and hydroxy-AFm will not be observed in the presence of monocarboaluminate. Therefore only minor hydroxide for sulfate substitution will be observed to occur in the monosulfoaluminate phase in the region A-D-E of Fig. 5.21, i.e., “monosulfoaluminate” will have close to its ideal composition (up to ~10 mol% calc. maximum sulfate substitution possible). Due to the reactions shown in Eqns. 5.6 and 5.7, C4AHx will not persist at higher activities of sulfate and/or carbonate, i.e., in the presence of ideal monosulfoaluminate and monocarboaluminate. According to the calculations based on the thermodynamic data compiled in this work, hemicarboaluminate is metastable in the shaded region A-A’-E (Fig. 5.21) with respect to monocarboaluminate. Thus in region A-A’-E monocarboaluminate is predicted to coexist with monosulfoaluminate-type solid solutions. However the exact position of point A’ is strongly dependent from the chosen solid solution model and deviations of the thermodynamic data of the related hydrate phases. But as shown in Fig. 5.21 region A-A’-E is relatively small compared to other stability fields and does not affect the main conclusions depicted in Fig. 5.21.

Pöllmann et al. [151] reported that monosulfoaluminate-hydroxy-AFm solid solutions may interact with hemicarboaluminate forming limited ternary solid solutions including carbonate, sulfate and hydroxide ions. However no solubility data for these solid solutions are available. Therefore these ternary solid solutions cannot as yet be integrated into the dataset to assess their thermodynamic stability and relations with other hydrate phases.

5.4.2 Stable phase assemblages at 25°C

As described earlier hydroxy-AFm type solid solutions and part of the monosulfoaluminate-type solid solutions are metastable with respect to hydrogarnet and portlandite and tend to slowly decompose to the latter with time. The resulting phase changes will mainly affect region A-C-D of the original metastable phase diagram Fig. 5.21. As a result of these phase changes Fig. 5.22 was developed to represent the stable phase relations between AFm phases and hydrogarnet at 25°C. The size of region A-B-D, where monosulfoaluminate-type solid solutions coexists with

The AFm phase in Portland cement 77 hemicarboaluminate, decreased markedly (compared to Fig. 5.21), due to the metastability of monosulfoaluminate-type solid solutions with x ≤ ~0.76 (Eq. 5.1) with respect to a mixture of

C3AH6, portlandite and the limiting monosulfoaluminate type solid solution with x ~0.76 (Eq 5.1) which coexists with hemicarboaluminate in region B-C-D.

According to experiments described in this Thesis metastable phase assemblages, as shown in Fig. 5.21, are likely to form in the initial stage of hydration and tend to be persistent. The decomposition of these metastable phase assemblages is generally very slow, on the order of years. Thus to assess short-term-experiments < 1 year it is recommended to use the partly metastable phase relations shown in Fig. 5.21. An exemption is the hydroxy-AFm end member which tends to decompose fairly rapid in a sulfate free environment. However as experiments did show, its persistence is significantly improved due to apparently minor sulfate substitution in its structure.

A monosulfoaluminate (Ms)

Ms-type ss1) A' limiting Ms-type ss (~ C4As0 76Hx) B M s- type ss + Hc1) Ms(ss)+Mc2) monosulfoaluminate + monocarboaluminate3) limiting Ms -type s s + C3AH6 + Ca(OH)2

Ms + Hc + Mc1) limiting Ms-type ss

+ C3AH6+ Ca(OH)2

+ Hc

Hc + C AH + Ca(OH) 1) C 3 6 2 D Hc + Mc E C3AH6 + Ca(OH)2 hemicarboaluminate (Hc) monocarboaluminate (Mc)

Fig. 5.22: Calculated stable phase assemblages between different AFm phases at 25°C. A possible range of stoichiometry of hemicarboaluminate and C4AHx is not shown. (Note: the region of two solid phases Ms type ss + Hc and Ms(ss) +Mc in which the monosulfoaluminate-type phase has a variable composition; due to 1) 2) 3) phase rule restrictions additionally traces of C3AH6 gibbsite gibbsite and AFt predicted in relevant stability fields; the exact position of point A’ on A-B is strongly dependent from the chosen solid solution model).

The AFm phase in Portland cement 78

5.5. Discussion of the results

5.5.1 Extent of solid solution Perhaps surprisingly, AFm phases form only limited mutual solid solutions under conditions likely to be obtained in cement environments. Thus, for example, hydroxy- and carbonate-AFm phases are not directly miscible but instead form an ordered solid solution, i.e., a definite compound, ideally with the 1:1 ratio. Part of the reason for lack of miscibility is associated with the difficulty of generating high component activities of some species, e.g., carbonate. In normal cement environments the solubility of carbonate is invariably low, being limited by the solubility of carboaluminates and calcite. Where it is possible to achieve a relatively broad spectrum of component activities, as for example can be achieved for sulfate and, especially, hydroxide, significant ranges of solid solutions may form in response to the expanded range of species activities. It is shown that at self-generated aqueous concentrations of sulfate and hydroxide, solid solution extends from sulfate-AFm to approximately 50 mol% of a C4AHx end member. This limit is not invariably agreed in experimental studies [150][163][197]. In some cases, data pointing to complete solid solution contain internal conflict and the interpretation is thus suspect. It is nevertheless possible to envisage conditions, for example where precipitation of AFm occurs from highly supersaturated solution, or in the presence of cations interacting only weakly with solids but promoting high concentrations of hydroxyl, e.g., sodium or potassium hydroxides. But the results reported in this Thesis are probably applicable to “normal” OPC compositions, i.e., those with pH in the range conditioned by portlandite (~12.5 at 25°C) and up to pH about one unit more.

Besides need for conditions favouring formation of solid solutions, kinetics needs also to be considered; solid solutions of exotic compositions may form, but will they be stable or persistent?

In this connection, it is worth recalling that C4AHx is not only metastable, it is also labile. Its labile behaviour is manifested by its comparatively rapid decomposition at ambient temperature to the more stable solid phases, hydrogarnet (C3AH6) and portlandite (Ca(OH)2). Sulfate-AFm preparations, on the other hand, while claimed to be thermodynamically metastable at <40°C [46][47], are nevertheless persistent at 25°C. Zhang [197] made preparations which might initially have contained AFm (sulfate/hydroxy) solid solutions but after prolonged ageing (560d at 25°C) were found to have decomposed to mixtures of sulfate-AFm and other solids, e.g., C3AH6. According to own investigations and calculations reported in this Thesis monosulfoaluminate and part of the monosulfoaluminate-hydroxy-AFm solid solution series (up to ~24 mol% sulfate replacement possible) are thermodynamically stable at 25°C. Experiments have shown that the hydroxide-substituted sulfate-AFm solid solutions are relatively persistent and, moreover, are observed to occur in “real” cements, it is important to include them in a phase development model. In this connection, it is noteworthy that it was possible rapidly to achieve hydroxide-substituted AFm by mechanically mixing the two solids, hydroxy- and sulfate-AFm, at 25°C, suggesting that - 2- this associates good persistence within a range of solid solutions: partial OH for SO4 substitution does not necessarily detract from the ability of these solid solutions to form and persist. As will be shown, a variety of mechanisms contribute to the ultimate decomposition of these phases.

The AFm phase in Portland cement 79

5.5.2 Transformation mechanisms It has always been assumed that changes in the anion chemistry of an AFm phase occur by ion + exchange. In this view, the [Ca2Al(OH)6] layers are relatively fixed, while water and charge- balancing anions in interstitial spaces are regarded as more mobile and hence exchangeable. Thus the chemistry of the interlayer ions adjusts itself to that of the pore fluid. So, for example, if a cement develops C4AHx but carbonate ions become slowly available, perhaps from dissolution of added calcite, a change in AFm composition will occur and it will proceed via ion exchange, first to hemicarboaluminate and eventually to monocarboaluminate. But the structures of the three phases differ in detail and, as shown, they do not form continuous solid solution. It is therefore difficult to envisage the transformation proceeding by other than by a reconstructive process; that is, one phase dissolves while another, differing in composition and structure, nucleates and crystallises. While transformation by ion exchange could permit the morphology and microstructure to persist throughout the conversion(s), nucleation and crystallisation are more likely to require a reconstitution of the microstructure. Putnis et al. [153] have recently described a novel mechanism for pseudomorphic replacement of one phase by another and this may have relevance for the mechanism of transformations amongst AFm phases driven by anion exchanges. The net result of this study is to create uncertainty concerning the actual mechanism of the transformation and consequently, the potential for changes in microstructure associated with transformations occurring amongst the AFm phase(s).

More accurate structure determinations of individual AFm phases would be helpful to elucidate the actual mechanism of transformation: the structures may have a common plan, but differ significantly in detail and these details are important in the present context. It is not impossible to envisage the transformation between AFm phases occurring by ion exchange but with some + simultaneous reordering and rearrangement within the [Ca2Al(OH)6] layers avoiding, or partly avoiding, need for a nucleation and crystallisation step, as described in [153]. However, the discontinuous nature of the AFm phase, requiring that it be treated as a series of discrete phase rather than a single solid solution must affect subsequent treatments of paste mineralogy.

5.5.3 Solubility data and its interpretation Interest in the solubility of AFm phases centres on the solid solution series between sulfate- and hydroxy-AFm end members: these data are shown in Table A.10 and Table A.11. Special problems arise in determining the solubility of C4AHx at 25ºC because other hydration states, notably the 19 H2O state stable at > 88% r.h. and the 13 H2O state at < 88 % r.h. [180]. However the impact of lower r.h. parallels the behaviour of solutions with lowered water activities corresponding to the high ionic strengths observed in cement pore fluids [79]. Consequently, it is difficult reliably to perform solubility measurements on a single hydration state, although the author believes, solubility data reported here are obtained on C4AHx, with x close to 19, after allowing for other anionic substituents. But, the preparation for X-ray diffraction, done in a dry protective atmosphere to avoid carbonation, results in some dehydration of C4AHx with the result that the XRD-data actually relate to the 13 H2O state, or 13-x if some sulfate solid solution occurs. The extent of solid solution, as well as its persistence, may be influenced by the hydration state of

C4AHx. Thus it is not surprising that solid solution would be incomplete between C4AsH12 or its highly hydrated form C4AsH14 and C4AH19. There is a shortage of direct and reliable information on AFm hydration states to progress this assessment. The possibility of using synchrotron radiation and examining AFm solids still in contact with solution offers hope that new data can be obtained to settle this question. The higher ionic potential of the aqueous phase of cements resulting from

The AFm phase in Portland cement 80

dissolution of alkalis, etc. may tend to favour the 13 H2O state [79] of the C4AHx phase or the 12

H2O state of monosulfoaluminate. Secondly, the lability of the C4AHx phase results in a difficult experimental choice: either prolong the time allowed for equilibration with an aqueous phase and risk decomposition to other solids, thereby influencing in solubility, or else use only brief equilibration times, thereby minimising decomposition, but perhaps failing to saturate the aqueous phase.

C4AHx almost certainly takes into solid solution a small amount of sulfate. Lippmann diagrams are a useful instrument to estimate the extent of solid solution; accurate experimental data are a precondition. Best estimates of the extent of solid solution are on the order of max. 3% according to the derived best fit to the theoretical Lippmann diagram Fig. 5.9. This small sulfate content apparently causes a drastic lowering of calcium concentrations. However the lowering is not necessarily a consequence of solid solution; if it were, it would imply strong crystallochemical stabilisation by sulfate. Addition of a small amount of sulfate significantly reduces the labile nature of C4AHx. The solubility measurements, supposedly on C4AHx, may actually relate to a self- generated spurious invariant point of an aqueous solution coexisting with a mixture of phases; hydroxy-AFm solid solutions as well as the stable monosulfoaluminate type solid solutions and

C2AH8. However decomposition has much less impact on the rest of the system and the solubility trends are consistent with a broad two (solid) phase gap and continuous solid solution ranging 2- between initially about 50 mol% SO4 and the monosulfoaluminate end member. The miscibility gap is likely to be widened with time. However experiments have shown that decomposition of the labile solid solutions is slow: it is likely to be several years until the final equilibrium stage is reached.

5.6. Conclusions

A basic thermodynamic description of the AFm phase(s) necessitating collection of new data, and its coupling to experimental investigations, has enabled new insights into the relations between the most important AFm phases likely to occur in Portland cement: The principal findings are:

- 2- 2- • AFm phases containing as principal anions OH , SO4 and CO3 are incompletely miscible with each other. Thus several AFm phases may coexist in a matured cement paste at ~25°C.

• An extensive solid solution with a miscibility gap may occur between monosulfoaluminate 2- - and C4AHx. The substitutable SO4 content, by OH is ~ 50 mol% of the initial sulfate content of monosulfoaluminate but is likely to decrease with prolonged ageing (1-2 years) at 25°C. Due to discrepancies in the literature, long term investigations and calculation were used to determine the thermodynamic stability of the solid solutions. According to the calculation, supported by experiments, hydroxy-AFm solid solutions and part of the monosulfoaluminate type solid solutions with ≥ ~24 mol% hydroxide substitution are thermodynamically metastable with respect to hydrogarnet, portlandite and sulfate-rich monosulfoaluminate type solid solutions (max. ~24 mol% hydroxide substitution). The - 2- amount of substitutable OH in hydroxy-AFm due to incorporation of SO4 is small (≤5%).

• Thermodynamics and experiment agree in support of the above conclusions. However hydroxide-substituted sulfate-AFm have only a small driving force for decomposition at ~25°C, as a consequence of which they are persistent.

The AFm phase in Portland cement 81

• Carbonate and sulfate provide thermodynamic stabilisation for the AFm phase. Thus monocarboaluminate and part of the monosulfoaluminate-hydroxy-AFm solid solution series are stable at 25°C.

• Hemicarboaluminate is a stable phase over a short range of carbonate activities at ~25°C. This range, although short, coincides with commonly encountered conditions in cements, e.g. the presence of minor carbonate even in fresh clinker, and for this reason it should not be excluded from consideration in a mineralogical model for paste hydration.

The AFt phase in Portland cement 83

Beside sulfate, different anions, X, have been reported to be bound into the ettringite structure. The current study concentrated on compositions relevant to cement, despite the existence of more “exotic” AFt phases.

Flint and Wells [67] observed the formation of a mixture of hexagonal platy AFm-like crystals with a composition of 3CaO⋅Al2O3⋅CaSiO3⋅12H2O and needle-like AFt crystals with a proposed composition 3CaO⋅Al2O3⋅3CaSiO3⋅xH2O (x in the range of 30 - 32) from a solution containing appropriate amounts of silica, calcium and aluminium. Furthermore the formation of hydrated hexacalcium aluminate with a composition 3CaO⋅Al2O3⋅3Ca(OH)2⋅30H2O, made by mixing a lime- sucrose solution with calcium aluminate solution, was described.

Carlson and Berman [36] and later Pöllmann et al. [152] repeated successfully the synthesis of

3CaO⋅Al2O3⋅3Ca(OH)2⋅30H2O according to the procedure given by Flint and Wells, but agreed that the compound is very sensitive to carbonation and converts with time to a carbonate analogue of ettringite with a composition 3CaO⋅Al2O3⋅3CaCO3⋅32H2O. Carlson and Berman [36] synthesised this carbonate analogue of ettringite (tricarboaluminate, 3CaO⋅Al2O3⋅3CaCO3⋅32H2O) using a sucrose solution and stoichiometric amounts of CaO, NaAlO2 and NH4HCO3. However tricarboaluminate could not be synthesised in the absence of sucrose; the formation of monocarboaluminate was observed instead. Hence it seems to be unlikely that tricarboaluminate will appear as a hydration product of Portland cement. Carlson and Berman [36] reinvestigated the silica-containing samples prepared by Flint and Wells [67] after ageing for more than 18 years. The XRD-pattern of these preparations showed a strong pattern similar to tricarboaluminate and diffuse signals corresponding to strätlingite and C-S-H. The solids lacked good morphology and were concluded to be poorly crystalline. The overall chemical composition of the solid was in agreement with the results earlier published by Flint and Wells [67]. However, in contrast to Flint and Wells, considerable amounts of CO2 (0.73 moles per mole of Al2O3) were found to be present in these samples, which suggests a formation of a solid solution between 3CaO⋅Al2O3⋅3CaSiO3⋅xH2O and

3CaO⋅Al2O3⋅3CaCO3⋅32H2O.

6.1.2 Solid solutions between AFt phases Carlson and Berman [36] investigated the possibility of solid solution formation between sulfate- AFt and tricarboaluminate. In a first approach, the procedure to form sulfate-AFt was modified by partial replacement of calcium sulfate by calcium carbonate. This resulted in the formation of a mixture of sulfate-AFt and monocarboaluminate. In a second trial the lime-sucrose method was applied but half of the ammonium bicarbonate was replaced by ammonium sulfate. The XRD- pattern showed the presence of two distinct AFt phases, one with reflections close to sulfate-AFt and the other close to tricarboaluminate. Thus Carlson and Berman concluded that if solid solution occurs, it is limited in extent.

In contrast to Carlson and Berman [36], Pöllmann [152] successfully applied the sucrose method to synthesise solid solutions between sulfate-AFt and its carbonate analogue by direct precipitation of the solids from a supersaturated lime-sucrose mixture by mixing solutions containing appropriate amounts of NaAlO2, Na2CO3 and Na2SO4. Lattice constants were refined by least square methods from the XRD-pattern of the solids. He found that almost 2/3 of the sulfate of AFt can be replaced by carbonate. However no sulfate substitution was found for the tricarboaluminate end member. Additional experiments demonstrated that ~50 mol% of the sulfate can be replaced by hydroxide, OH-, if sucrose solutions are used. Furthermore he showed that there is a continuous solid solution

The AFt phase in Portland cement 84

series between the hexacalcium aluminate hydrate, 3CaO⋅Al2O3⋅3Ca(OH)2⋅30H2O, and tricarboaluminate.

Barnett et al. [16] published an XRPD-fitting study of the solid solutions between sulfate-AFt and tricarboaluminate prepared according the procedures described by Pöllmann et al. [152]. In agreement with Pöllmann, only limited solid solution formation was found and about 2/3 of the sulfate could be replaced by carbonate in the original sulfate-AFt structure. In contrast to Pöllmann’s investigation significant amounts of carbonate were believed to be replaced by sulfate in the structure of tricarboaluminate, which resulted in a formation of a second solid solution series, though the determination of the chemical composition of the solid solution was not possible.

Midgley and Rosaman [131] investigated the time-dependent phase evolution of hydrated Portland cement. They found a systematic decrease of the d0001 value of ettringite in conjunction with a shift of the normalised peak temperature in DTA investigations of ettringite to higher temperatures. Solid solutions between sulfate-AFt and hexacalcium aluminate hydrate (hydroxy-AFt) exhibit similar results: with increasing hydroxide substitution; the d0001 value decreased and the normalised peak temperature in the DTA experiments increased. Thus Midgley and Rosaman concluded a progressive substitution of sulfate by hydroxide in the AFt structure as it occurs in Portland cement with increasing reaction time.

In solubility determinations Warren and Reardon [192] found significant deviations of the ettringite solubility product at pH values > 13 compared to the mean values obtained at lower hydroxide - 2- activities. A possible explanation is the substitution of OH for SO4 in the ettringite structure at higher hydroxide activities.

Recently Neubauer et al. [140] reported a significant decrease of the dimension of the c lattice constant of AFt in the early stages of hydration. It is believed that at the outset of hydration, a iron- and carbonate-rich AFt formed which was metastable and converted with time to an aluminium- and sulfate-rich AFt. However as long as no detailed information about chemical changes of the ettringite composition during hydration are available, it is difficult to apply these results.

6.1.3 The formation of thaumasite and related solid solutions The first known occurrence of thaumasite induced deterioration of concrete was reported by Erlin and Stark [59] in 1965. They showed the coexistence of thaumasite, Ca3Si(OH)6(SO4)(CO3)⋅12H2O, with calcite, gypsum and brucite in the affected regions of a sewer pipe. In another pipe, thaumasite, ettringite, gypsum and traces of calcite were found below a seriously deteriorated zone.

In 1975 van Aardt and Visser [2] found the formation of thaumasite in 1 year old autoclaved Port- land cement mortar containing dolomitic aggregates exposed to sodium sulfate solutions at 5°C.

Growing concerns about reports of concrete deterioration resulting from thaumasite formation led to the formation of a Thaumasite Expert Group in the UK in 1998. Since that time, more cases of thaumasite formation have been reported worldwide and the number of scientific publications dealing with the thaumasite form of sulfate attack (TSA) has increased. As this chapter deals more with the thermodynamic stability of thaumasite and related solid solutions, a review of the literature was restricted to these aspects. General overviews about thaumasite formation in concrete are given in [183] and in a special issue of Cement and Concrete Composites edited by Macphee and Diamond [128].

The AFt phase in Portland cement 85

The structure of thaumasite The structure of thaumasite is similar to the column- and channel-based structure of ettringite (Fig. 6.1); carbonate replaces part of the sulfate in the channel sites while silicon replaces aluminium in the columns (see [58]). Thus the coordination of silicon in thaumasite is sixfold, which is unusual in low pressure phases. Indeed thaumasite is the only known mineral with Si(OH)6 octahedra to be formed at low pressure and ambient temperatures.

Solid solution formation The existence of solid solutions between SO4-ettringite and thaumasite in nature is known [103][138]. Murdoch and Chalmers [138] described an ettringite-like mineral with considerable amounts of SiO2 and CO2 and designated it as woodfordite, Ca6Al1 5(SO4,SiO3,CO3)3(OH)10 5⋅

15H2O. The authors subsequently revised this claim on the grounds that woodfordite is not a new mineral but is closely related to ettringite and withdrew the name “woodfordite” [136][137]. Lukas [126] found the occurrence of ettringite-thaumasite solid solutions in deteriorated concrete in tunnels with similar crystallographic properties as described by Murdoch and Chalmers. Barnett et al. [17] synthesised several members of the solid solution series and showed that limited solid solution between both end members occurs and thus a miscibility gap in the solid solution series exists. Later several researchers identified thaumasite-ettringite solid solutions in concretes undergoing sulfate attack [28][31][185][102]. Thus the formation of solid solution between both minerals has to be considered as an important process in the progression of sulfate attack in concrete and Bensted [22] proposed the “woodfordite-route” as one possible step of the formation of thaumasite in concrete.

Torres et al. [185] found ettringite-thaumasite solid solutions in cement paste undergoing sulfate attack. The extent of solid solution formation was discussed. Due to divergent outcomes of the XRD-peak fitting procedure compared to results given by Barnett et al. [17], it was concluded that the solid solution behaviour of ettringite and thaumasite is more complex than initially believed. The results indicated the presence of sulfate- and/or carbonate-rich ternary solid solutions between thaumasite, ettringite and possibly, tricarboaluminate.

The temperature-dependent stability of thaumasite and related solid solutions It is generally believed that thaumasite preferentially forms in carbonate-containing service environments at temperatures below 15°C. However Diamond [53] recently described thaumasite formation in California at temperatures well above 15°C. Brown and Hooton [31] found thaumasite in concrete samples stored in MgSO4 and Na2SO4 solutions at 23°C for 21 years. Macphee and Barnett [127] obtained solubility data of ettringite-thaumasite solid solutions in the temperature range between 5°C - 30°C; no apparent decomposition of thaumasite and related solid solutions occurred after 6 months storage at 30°C, which suggests the persistence of thaumasite at temperatures at least up to ~30°C. Blanco Varela et al. [28] studied the effect of cement C3A content, temperature and storage medium on the formation of thaumasite in precarbonated mortars. Thaumasite and thaumasite solid solutions formed at both 5°C and 21°C. The kinetics of formation is not only dependent on temperature but also on the chemistry of the aqueous phase. Samples stored at 5°C in water formed thaumasite after 5 years, whereas the same samples stored at 21°C in a 2% gypsum solution formed thaumasite after 16 months.

The AFt phase in Portland cement 86

Giampaolo [76] estimated the upper thermal stability limit of thaumasite to 60±5°C from coupled dehydration and XRD experiments. In his experiments different batches of natural thaumasite were heated in an oven to 67, 77, 83 or 90°C. After 240h at 67°C no significant changes of the unit-cell parameters were detected, whereas thaumasite dehydrated completely to form amorphous phase(s) after 270h at 77°C. Subsequent rehydration experiments at 18°C and 60°C of the dehydrated samples resulted in the formation of gypsum and calcite from the initially amorphous dehydration products. Giampaolo concluded that the dehydration of thaumasite is irreversible at the applied experimental temperatures.

The AFt phase in Portland cement 88 in the sucrose method used in both [16] and [152]. As a result of the changed conditions, a more extensive solid solution formation will be observed. However it is difficult to determine the “real” limits of miscibility as in the mixed end member approach it could be argued that kinetical restraints retard a solid solution formation while on the other side, in the sucrose preparations, initially formed metastable solid solutions might decompose with time to more stable phase assemblages. However this study has shown, in agreement with the literature, that solid solutions between SO4-AFt and CO3-AFt form spontaneously.

6.2.2 Thermodynamic modelling of the solid solution formation As previously described in chapter 4.3.4 a binary non-ideal solid solution model was used to estimate the thermodynamic properties of the SO4-CO3-AFt solid solution series. The dissolution reaction of the solid solution series is given in Eq. 6.1:

2+ - -2 2− - −x33x426 12 2 ⎯⋅ ⎯→ 6CaOH26)OH()CO()SO(AlCa 2AlO2 4 3 ++−+++ 2 OH03OH4CO)x3(xSO (6.1) To fix phase boundaries, X-ray diffraction data need to be considered together with solubility data obtained from the same compositions. The investigation of the solid phases has shown that there is a divergence between the phase boundaries estimated from solid state investigations in this study and values reported in the literature. Lippmann diagrams have been shown to be a useful tool to estimate the extent of miscibility in a binary solid solution system (see chapter 5.3) and to show the relation between the chemical compositions of the solid and aqueous phases. The solubility data obtained on the solid solution series are listed in Table A.12. The carbonate concentrations were always below the limit of detection and had to be estimated by assuming an equilibrium with calcite. Subsequently activities of the relevant ions were computed using GEMS and the appropriate solubility data. Thus it is now possible to calculate the total solubility product of the AFt solid solution series according to Eq. 6.2; [ ] denote activities of the aqueous species:

- 2− 2− ]CO[]SO[]OH[]OH[]AlO[][Ca =Π + 62 2 − 4 30 3 + ]CO[]SO[]OH[]OH[]AlO[][Ca 3 ∑ 2 2 ( 4 3 ) (6.2) The solidus of the Lippmann phase diagram represents a series of total solubility products calculated from the aqueous phase composition in dependence of the changing chemical composition of the solid phase. As shown in chapter 5.3.1, Lippmann diagrams in conjunction with the software MBSSAS can be used to estimate the dimensionless Guggenheim parameters a0 and a1, which are necessary to calculate the activity coefficients of the sulfate and carbonate end members in dependence of the solid phase composition according to Eqns. 4.19 and 4.20. The theoretical total solubility product is then calculated with Eq. 4.24. To obtain realistic values of the miscibility gap boundaries, literature investigations as well as solubility and solid state investigations reported here have to be in agreement. Thus the miscibility gap boundaries should be in the range x1~0…~0.15 ≤ x ≤ x2~1 … ~1.5 according to Eq. 6.1. To harmonise the available literature data with results obtained in this study and to enable the estimation of x1, one end of the miscibility gap was fixed to x2 = 1.25, the mean value between previously reported results and data obtained in this study.

The AFt phase in Portland cement 90

To complement the Lippmann phase diagram (Fig. 6.3) and establish the relation between the constitution of the aqueous phase and the solid solution composition, the solutus function needs to be included. With the help of the previously determined fitting parameters a0 and a1 and the known activities of the sulfate- and carbonate-species, the experimental solutus can be calculated with Eq. 6.3:

⎛ X X ⎞ SO2− aq, CO2− aq, ⎜ 4 3 ⎟ sl,eq =Π 1 + (6.3) ∑ ⎜ K γ K γ ⎟ ⎝ C6As3H32 C6As3H32 C6Ac3H32 C6Ac3H32 ⎠ where KC6As3H32 and KC6Ac3H32 are the solubility products of the end members, γC6As3H32 and γC6Ac3H32 are the activity coefficients of the end members and XSO42- , aq and XCO32-, aq are the “aqueous activity 2- 2- fractions” of the substitutable species SO4 and CO3 calculated from Eqns. 6.4 and 6.5; [ ] denote calculated activities of the aqueous species:

− 32 4 ]SO[ X = (6.4) SO2− aq, − 32 − 32 4 4 + 3 ]CO[]SO[

− 32 3 ]CO[ X 2− = − 32 − 32 (6.5) CO3 aq, 4 + 3 ]CO[]SO[ As shown in Fig. 6.3, experimentally-derived values (triangular and cross markers) agree well with the theoretically computed solutus. 3.0 experimental 2.0 calculated 1.0

0.0 [kJ/mol] m

G

Δ -1.0

-2.0

mixing mixing -3.0 calc. misc. gap 1)

relative changes of the energy of -4.0 1)single datapoint -5.0 Fig. 6.4: Comparison between calculated 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 and theoretical changes of the relative energy of mixing of the SO4-CO3-AFt CO3-AFt molar bulk SO4/(S O 4+CO3)-ratio SO4-AFt solid solution series at 25°C

Fig. 6.4 shows a comparison between the theoretically calculated and experimentally-derived relative energies of mixing in dependence of the solid solution composition. The relative energy changes are independent of the initial solubility product of the end members. The theoretical energy of mixing is calculated using Eq. 4.14 and the previously derived dimensionless

Guggenheim parameters a0 and a1. The experimentally-derived values are calculated according to the following procedure:

Independently of the existence of a miscibility gap, the solubility product of a solid solution at stoichiometric saturation Kss [81] can be calculated based on applying the law of mass action to the dissolution Eq. 6.1 for ettringite solid solutions (Eq. 6.6; [ ] denote species activities):

+ 62 - 2 − 4 2− x 2− −x3 30 ]OH[]CO[]SO[]OH[]AlO[][Ca K ss = 2 4 3 2 ]OH[]CO[]SO[]OH[]AlO[][Ca (6.6)

The AFt phase in Portland cement 92

6.3. Investigations on thaumasite

6.3.1 The need for investigations Juel et al. [98] developed a thermodynamic model to predict the stability of thaumasite. This model is based on phase rule and mass balance restrictions and gives useful information about the qualitative and quantitative composition of phase assemblages compatible -or incompatible- with thaumasite. Experimental work in support of this model was confined to 5°C and both ettringite and thaumasite were assumed to have their ideal compositions. But Macphee and Barnett [127] published solubility data for ettringite-thaumasite solid solutions showing extensive solid solubility of thaumasite in ettringite and this Thesis supports their conclusions. To enable thermodynamic modelling of the solid solution formation it is necessary to know the thermodynamic properties of the pure end members of the solid solution series. Unfortunately, Macphee and Barnett [127] did not give solubilities for the thaumasite end member and it is thus difficult to estimate the thermodynamic properties of the solid solutions. However Bellmann [20] reports solubilities of natural thaumasite, but only at 8°C. The preparations by Macphee and Barnett [127], disclosing the existing miscibility gap between ettringite and thaumasite, can be used to estimate the thermodynamic properties of thaumasite. Damidot et al. [44] used their data to derive a solubility constant for thaumasite at 25°C at which temperature thaumasite was stable. Invariant points were calculated for phase assemblages including thaumasite in the system CaO–Al2O3–SiO2–CaSO4–CaCO3–H2O. However limitations of the data meant that calculations were confined to 25°C. Thus need arises to determine thermodynamic properties of the thaumasite end member and its temperature dependence in order to assess its stability and compatibility, especially in the range 0°C to 30°C.

6.3.2 Synthesis of thaumasite A critical review of the literature shows that the synthesis of thaumasite is relatively difficult compared to other AFt phases, e.g. SO4-AFt or CO3-AFt. Thaumasite forms slowly and low temperatures are preferred. According to the literature [3][17][177] the best synthesis is obtained using a sucrose solution instead of water. In the current study, approximately 25g batches of thaumasite were synthesised using the sucrose method described by Barnett et al. [17]: typically 4.34 g of Na2Si2O5⋅2H2O, 4.23g of Na2CO3, 5.65 g of Na2SO4 and 6.70 g of freshly burnt CaO and ~ 440 ml of a 10% w/v sucrose solution were mixed according to the following procedure:

1. The CaO was dissolved in to the 10% w/v sucrose solution and cooled to 5°C.

2. A slurry containing Na2Si2O5⋅2H2O, Na2CO3, Na2SO4 and 60g water was prepared and agitated for 10 min.

3. The sodium salt slurry was added to the lime-sucrose solution and stored for 6 weeks at 5°C. To accelerate reaction the slurry was stirred continuously for 2 weeks and agitated regularly thereafter.

4. After 6 weeks the solid was vacuum-filtered with Whatman #540 filter paper and washed several times with ultra pure degassed water to remove alkalis and sucrose. Subsequently

the solid was dried over saturated CaCl2 solution at 37% r.h. at 25°C for 3 weeks and analysed by XRD using a PANALYTICAL X’PERT PRO diffractometer; the operating parameters are described in chapter 3.2.

The AFt phase in Portland cement 94

The TG and DTG curve of synthetic thaumasite Fig. 6.8 showed a 4 step-weight loss: The first step (DTG maximum at 133°C) can be attributed to the loss of loosely bound water linking the columns (~37.3%). A second step (DTG maximum at ~450°C) is likely due to the decomposition of OH-- groups. The sum of both weight losses corresponds closely to the total amount of bound water; 44.3 wt.-% versus a theoretically calculated water content of 43.4 wt.-% using the formula

Ca3Si(SO4)(CO3)(OH)6⋅12H2O. The third step, at 720°C, is believed to be caused by a loss of carbon dioxide; the mass loss of ~7 wt.-% carbon dioxide, which agrees with the theoretical CO2 content (~7 wt.-%). The origin of the fourth step, a weight loss of ~3 wt.-% at ~910°C, remains less clear. It could be that this is an additional peak due to the loss of CO2, which would however imply a higher carbonate content than theoretically calculated. Götz-Neunhöfer [84] found a similar peak at the thermal analysis of ettringite, synthesised by the sucrose method. However, this peak only occurred in N2-atmosphere. Götz Neunhöfer [84] suggested that small amounts of organic carbon, 2- 2- probably present as contamination due to insufficient washing, reduced parts of the SO4 to SO3 which subsequently decomposed to SO2 (g) at ~ 900°C. If the latter hypothesis is correct the binding of sucrose to the solid needs to be explained. If the sucrose is structural (which appears unlikely on crystallochemical grounds) it would result in significantly lower water contents of thaumasite. However as the water content is close to the theoretical value it is believed that small amounts of sucrose were probably adsorbed on the surface of the thaumasite samples1.

6.3.3 The temperature-dependent stability of thaumasite in aqueous solutions Experimental To assess the stability of thaumasite and derive solubility data, 0.5g of previously synthesised thaumasite was redispersed in 60 g ultra pure degassed water. Several preparations were stored 4 weeks at 1, 5, 25, 40, 55 and 70°C in HDPE-bottles or PTFE bottles, respectively. Subsequently

~ 30ml of the aqueous phase were removed in a N2-atmosphere with a syringe and analysed as described in chapter 3.5. The remaining preparations were filled with ultra pure degassed water and two more sets of solubility data were obtained in 4 week cycles at 1, 5, 25, 40, 55 and 70°C, according to the procedure described above. At the end of the experiment the remaining solids were dried over saturated CaCl2-solution at 37% r.h. and subsequently analysed by XRD using a PANALYTICAL X’PERT PRO diffractometer according to chapter 3.2.

Additionally the solubility of a natural thaumasite sample from Akschal/Kazakhstan, was determined to compare these data with the data obtained from synthetic thaumasite and previously reported data in the literature [20][44][127]. Samples intended for solubility determinations were gently ground by hand and sieve sized to ≤ 40μm using an agate mortar and subsequently dispersed in ultra pure degassed water using a water/solid-ratio of ~100. The dispersions were stored for 180d at either 5°C or 25°C respectively. The aqueous phase was analysed as described above.

Investigation of the solid phases Fig. 6.9 shows the temperature-dependent XRD-pattern of the solids after a series of 3 redispersions (after 84d) of the originally phase-pure thaumasite.

1 Balonis [13] confirms that rigorous washing of ettringite, synthesised by the sucrose method, eliminates these effects. It is likely that sucrose is strongly retained at the surface of thaumasite.

The AFt phase in Portland cement 96 a) b) 100 2.0E-03 70°C 70°C 90 0.0E+00 -2.0E-03 80 55°C 55°C 55°C 1 - 40°C -4.0E-03 70 -6.0E-03 70°C 1 - 40°C 60 -8.0E-03 -1.0E-02 1 - 40°C 50 relative weight [wt.-%]relative -1.2E-02 40

differential weightdifferential loss [mg/K] -1.4E-02 0 200 400 600 800 1000 0 200 400 600 800 1000 Temperature [°C] Temperature [°C]

Fig. 6.12: Thermal analysis (TG data left; DTG data right; heating rate 10K/min; N2-atmosphere) of synthe- tic thaumasite and related decomposition products following redispersions at temperatures from 1 to 70°C.

The thermal analysis of the dried solids after 12 weeks annealing at temperatures between 1 to 70°C is shown in Fig. 6.12. Up to 40°C no differences were observed in the TG and DTG-patterns: the DTG pattern (Fig. 6.12 b) shows the typical peaks previously observed for synthetic thaumasite. The sample previously annealed at 55°C showed a significant decrease of the area of the main peak of thaumasite dehydration at ~100°C but the peak area at ~ 760°C resulting from decarbonation of calcite increased significantly. The sample stored at 70°C showed no thaumasite. A broad peak between ~90 to 140°C indicates the typical slow dehydration of C-S-H. Furthermore the peak at 460°C resulted from the decomposition of OH- groups of C-S-H. Due to absence of thaumasite and release of water, the relative amount of calcite increased, as evidenced by the increase of the calcite peak area at 760°C. The peak at ~900°C, typical of the samples which contained thaumasite, disappeared.

Thermodynamic modelling To derive a consistent thermodynamic dataset for the temperature dependence of thaumasite solubility, data obtained from undersaturation are listed in Table A.13. Due to divergent literature results, solubility data of natural thaumasite after 180d equilibration at 5°C and 25°C, respectively, are given together with the solubilities of synthetic thaumasite. As noted in Table A.13, carbonate concentrations were always below the analytical limit of detection (~0.1mmol/l); the aqueous carbonate concentrations had to be estimated by assuming a saturated solution with respect to calcite. As before (paragraph 4.4.2), although the presence of calcite does not necessarily confirm that this phase is in equilibrium with the solution, this approach has given good agreement between experimental and calculated data.

The theoretical principles of the solubility based estimation of thermodynamic properties of minerals are described in chapter 4.3.3. Based on the assumed dissolution reaction of thaumasite, Eq. 6.8, and the law of mass action, temperature-dependent solubility products of thaumasite were calculated according to Eq. 6.9:

2+ − 2− 2− − 3 46 3 2 ⎯⋅ ⎯→ Ca3OH12)CO)(SO()OH(SiCa HSiO3 4 3 +++++ 2 OH14OHCOSO (6.8)

+ 32 − 2− 2− − 14 3 4 3 ⋅⋅⋅⋅⋅ 2 ]OH[]OH[]CO[]SO[]HSiO[]Ca[ (6.9) where [ ] denote aqueous species activities.

The AFt phase in Portland cement 100

Thaumasite stable Th. unstable 120

1) CaO+SiO2+CaSO4 2H2O+CaCO3+13H2O --> C3SscH 15 (thaumasite) 110

100 3) CaO+SiO2+CaSO4 2H2O+CaCO3+1 49H2O

--> CaO SiO2 1 49H2O+ CaSO4 2H2O+CaCO3 [kJ/mol] r

0 90 G Δ - 2) CaO+SiO +CaSO 2H O+CaCO 80 2 4 2 3 --> CaO SiO2 1 49H2O+ CaSO4+CaCO3+0 51H2O

70 0 102030405060708090100 Temperature [°C]

Fig. 6.16: Free energy-plot of the relative thermodynamic stability of thaumasite in the system CaO-SiO2- CaCO3-H2O Additionally Fig. 6.16 shows the free energy plot of the formation of the phase assemblage gypsum, C-S-H and calcite (reaction 3)). But at temperatures above ~ 50°C, gypsum is thermo- dynamically metastable with respect to anhydrite and accordingly, gypsum is not predicted to be a stable decomposition product of thaumasite at elevated temperatures ≥68°C.

6.3.4 Solid solution formation between ettringite and thaumasite Until recently it was not possible to model this solid solution formation owing to i) a lack of a consistent temperature-dependent thermodynamic dataset for the thaumasite end member and ii) lack of solubility data for the solid solutions. However since Macphee and Barnett [127] published a solubility dataset for the members of the solid solution series at 5°C, 15°C and 30°C, and since datasets for the thaumasite and ettringite end members were derived in this study, it is now possible to model the solid solution behaviour of thaumasite and ettringite.

The calculations are based on the theory for non-ideal model binary solid solutions described in chapter 4.3.4. According to Macphee and Barnett [127] the dissolution reaction of the thaumasite- ettringite solid solution series can be described, assuming a doubled molar unit of thaumasite

(Ca6Si2(OH)12(SO4)2(CO3)2⋅24H2O), according to Eq. 6.11:

+⋅ ⎯⎯→ 2 − −++ − −+ 2− ++ SO)x2(CO)x22(HSiO)x22(xAlO2Ca6OH)x224()OH()CO()SO(SiAlCa 2− − − x223x24x22x26 12 2 2 3 3 4 (6.11) − ++++ 2OH)x228(OH)x22( Macphee and Barnett [127] report the degree of aluminium substitution in thaumasite is very small. However, depending on temperature, up to ~50 mol% of the aluminium of the ettringite end member might be replaced by silicon (see Fig. 6.17). To enable thermodynamic modelling the limits of the miscibility gap need to be fixed; as suggested by Macphee and Barnett [127] these were placed at 0.025 ≤ x ≤ 0.5 in Eq. 6.11.

The AFt phase in Portland cement 101

miscibility gap applied in the calculations

(1 ≤ nSioct ≤ 1.95) 30

25 thaumasite-type 1 phase phase 20 (SO4-AFt - thaumasite solid solution) 15 immiscible region with 2 phases

10 (lim. SO4-AFt-thaumasite-ss + thaumasite-type ss) Temperature [°C] Temperature 5

0 0 0.25 0.5 0.75 1 1.25 1.5 1.75 2 SO4-AFt Thaumasite nSi C6As 3H32 oct 2(C3SscH15)

Fig. 6.17: Temperature dependence of the phase relations between thaumasite and ettringite, SO4-AFt (according to Macphee and Barnett [127]) In the formal thermodynamic sense the thaumasite-ettringite solid solution series is not binary because a coupled substitution of silica and carbonate has to be considered in the original AFt structure. This becomes apparent by formulating a total solubility product, ∏, of the solid solution series according to Lippmann [121] by adding the solubility product of ettringite (Eq. 6.12) and the doubled solubility product of thaumasite (Eq. 6.9) according to Eq. 6.13:

+ 62 − 2 2− 3 − 4 30 2 4 ⋅⋅⋅⋅ 2 ]OH[]OH[]SO[]AlO[]Ca[ (6.12)

⎛ X Y ⎞ ⎜64744444 4864444 4744 4844 ⎟ Π = [Ca + 62 2− 2 − 2 28 − 2 2− − 2 2 + − 2 2− ][CO]HSiO[]OH[]OH[][SO]AlO[]OH[][OH][SO] 2 ∑ 4 2 ⎜ 2 4 2 3 3 ⎟ (6.13) ⎜ ⎟ ⎝ ⎠ where [ ] denotes aqueous species activities

Thus one limitation of the proposed binary model is, as shown by the two terms X and Y in parentheses (Eq. 6.13), that aluminium, sulfate and water have to be replaced simultaneously by silica and carbonate in the original ettringite structure and vice versa in the thaumasite structure. Thus for charge balance reasons a replacement of 1 mol aluminium requires the simultaneous substitution of 0.5 mol sulfate, 1 mol hydroxide and 1 mol water in the ettringite structure while incorporation of 1 mol silica requires a simultaneous substitution of 1 mol carbonate. It is questionable if in reality the phase boundaries of this simultaneous ion-substitution are rigidly fixed, as assumed in the model. To overcome this problem a ternary non-ideal solid solution model would have to be developed based on thaumasite, carbonate-AFt and sulfate-AFt. But due to the high complexity and unknown interactions between thaumasite and carbonate-AFt it is not been practicable to develop such a model for this Thesis; additional experimental would help to decide what priority should be attached to a higher-order solid solution model for AFt phases.

The solid solution behaviour between thaumasite and ettringite has been modelled by using the known miscibility gap boundaries and the binary non-ideal solid solution model described above. Due to the assumed simultaneous multiple substitution of aluminium, sulfate, hydroxide and water or silica, carbonate and hydroxide (Eq. 6.13) the construction of a Lippmann phase diagram [121],

The AFt phase in Portland cement 104

While reasons for this disagreement are not clear, the Rietveld analysis of the preparations by Macphee and Barnett [17][127] showed increasing amounts of amorphous phase present towards the thaumasite-end of the solid solution series. The preparation may contain impurities and C-S-H is probably also present. The Ca/Si ratio of the C-S-H is likely to decrease with increasing silica substitution, which is consistent with decreasing calcium and increasing silicon concentrations, as experimentally observed (see Fig. 6.19).

6.4. Formation of thaumasite and ettringite solid solutions in hydrated cements

Damidot et al. [49] have shown that the formation of tricarboaluminate is not likely in cement paste due to the greater stability of monocarboaluminate. However the thermodynamic data used by

Damidot et al. for CO3-AFt have been significantly revised and two non-ideal solid solutions i) between SO4-AFt and CO3-AFt and ii) between SO4-AFt and thaumasite were included in this study. Thus previous investigations and calculations raise the following question:

“Is extensive carbonate and silica substitution in AFt likely to occur in commercial cement paste?”

With the estimated thermodynamic properties of the ettringite-thaumasite and SO4-CO3-AFt solid solutions it is now possible to calculate the relevant hydrate phase assemblages including the extent of solid solution formation. Calculations were done by using GEMS-PSI (see chapter 4.3.1). The necessary thermodynamic data of relevant cement hydrates are given in Table 4.1 and Table 6.1. For practical reasons, the calculations were restricted to calcite saturated systems. This restriction is useful because either i) most cements are saturated in calcite by addition of calcite in the course of manufacture or ii), if not saturated during manufacture, they quickly become calcite saturated in the course of service. A second constraint is the suppression of siliceous hydrogarnet in the calculations: i) there are still uncertainties about the thermodynamic data and ii) according to data derived in this study, most of the AFm phases are metastable with respect to siliceous hydrogarnet in silica-containing systems. AFm phases, e.g. carboaluminates and monosulfoaluminate are commonly observed in hydrated cements to the exclusion or virtual exclusion of siliceous hydrogarnet. It seems therefore justified to suppress the formation of siliceous hydrogarnet as a first approximation to develop a consistent picture.

6.4.1 Carbonate substitution in SO4-AFt at 25°C As indicated in chapter 6.2.2, substitution of carbonate in AFt might lead to a stabilisation of AFt solid solutions. An experiment was done to test this hypothesis: C3A was mixed with Na2CO3 and

CaSO4 (molar ratios C3A:Na2CO3:CaSO4 = 1 : 1.2 : 1.8) and excess portlandite to obtain a SO4-

CO3-AFt solid solution with a SO4/(SO4+CO3) ratio of 0.6. Na2CO3 was used to achieve high carbonate activities during the initial stages of reaction. In contrast to the synthesis procedure of

CO3-AFt (chapter 4.1.3) deionised water (w/s ~ 30) was used instead of a sucrose solution to simulate similar conditions as encountered in cementitious systems. The mixture was reacted for 4 weeks in HDPE bottles. The solids were filtered and dried over saturated CaCl2-solution at ~37% r.h. and subsequently analysed by XRD using a PANALYTICAL X’PERT PRO diffractometer; see chapter 3.2.

As shown in Fig. 6.20 a phase mixture consisting of AFt with reflections close to SO4-AFt, mono- carboaluminate, calcite and portlandite has formed instead of a single SO4-CO3-AFt solid solution. The formation of monocarboaluminate and calcite suggest that extensive carbonate substitution is unlikely in SO4-AFt in hydrated Portland cements at 25°C.

The AFt phase in Portland cement 106

6.4.2 Phase assemblages including thaumasite and related solid solutions An increasing number of cases of the formation of thaumasite in concrete have been reported since the 1990’s. Growing concerns about the stability of typical cement phase assemblages with respect to thaumasite have led to increased research activities in the field of thaumasite formation. Several publications [28][44][126][127][185] note the importance of the formation of thaumasite-ettringite solid solutions in the course of sulfate attack. Thus this study concentrates mainly on phase assemblages occurring in Portland and blended cements undergoing “sulfate attack” including the formation of thaumasite-ettringite solid solutions. To avoid duplication the reader is referred to Damidot et al. [44] who calculated stable phase assemblages including ettringite and thaumasite in equilibrium with an aqueous phase in the system CaO-Al2O3-SiO2-CaCO3-CaSO4-H2O at 25°C. Possible pathways of the formation of thaumasite and ettringite-thaumasite solid solutions are discussed subsequently.

To enable a graphical representation only phase assemblages relevant to Portland and blended cements were considered, which include at least one of the following silica-containing constituents:

C-S-H, SiO2 (amorphous) and/or thaumasite. The calculations were restricted to calcite saturated systems as calcite is generally encountered in field studies and, in agreement with the restrictions in EN 197, many European Portland cements contain up to 5% limestone - itself mainly calcite.

As correctly stated by Damidot et al. [44] a graphical representation of the system CaO-Al2O3-

SiO2-CaCO3-CaSO4-H2O is difficult due to its high dimensionality. However by applying the above-mentioned restrictions it is possible to construct a 3D phase diagram. As shown in Fig. 6.22, the most useful graphical representation of the system is achieved using the molalities of sulfate, hydroxide and aluminate as variables. In general hydroxide is a dependent variable in alkali-free systems (the role of Na, K is discussed later) and the solubility of the cement solids is allowed to control the hydroxide activity. Fig. 6.22 can be interpreted as follows. All the labelled assemblages contain calcite (solid) and an aqueous phase in addition to those shown on the figure. Considerable supplementary information is presented in Table 6.2. Some of the solids, notably C-S-H, ettringite and thaumasite have variable compositions and Table 6.2 also gives key data on solid solution compositions. The phase relations are sensitive to temperature. Thus a second diagram constructed for 5°C is shown in Fig. 6.23.

Fig. 6.22 shows the stability fields of relevant phase assemblages at 25°C. The general sequence of phase formation does not differ significantly from that shown in [44] except that i) the formation of solid solutions between ettringite and thaumasite and ettringite and tricarboaluminate was taken into account and ii) that at higher silica activities, strätlingite will have a stability field; phase assemblages including siliceous hydrogarnet were not taken into account for the reasons mentioned above. C-S-H, although strictly not thermodynamically stable, has nevertheless been admitted into the calculation because of its well-known persistence in cement pastes. This does not present any difficulties with respect to calculations. Fig. 6.22 does not however enable calculation of the amount of phases present nor does the physical size of phase fields necessary represent their abundance or importance to the constitution of cement pastes.

The AFt phase in Portland cement 107

Fig. 6.22: 3D-representation of part of the system CaO-Al2O3-SiO2-CaCO3-CaSO4-H2O at 25°C and 1bar total pressure (solubility surfaces representative for solids in equilibrium with excess calcite; see Table 6.2 for additional carbonate and silicon concentrations at relevant invariant points) The results of calculation indicate two invariant points, one relevant to the phase constitution of ordinary Portland cement and the other to blended cements, containing sufficient slag or fly ash to react with portlandite; these are designated points I (with Ca(OH)2) and K (without Ca(OH)2) respectively. Because of their importance to the mineralogy of cement paste, the diagram is orien- ted so as to give prominence to these points. Table 6.2 shows the composition of the aqueous phase at these (and other) points from which it will be noted that the calcium solubility at point K is about an order of magnitude less, and the pH decreased by about 0.5 units, with respect to point I.

Solids at both points I and K coexist with an AFt phase and in both cases, the phase is ettringite- like. However the calculations predict the formation of a SO4-CO3-AFt solid solution (~9 mol%

CO3 substitution; AFt: Ca6Al2(SO4)2 73(CO3)0 27(OH)12⋅26H2O) but no significant silica substitution in AFt in point I, which agrees with the predicted carbonate substitution for the silica-free system (Fig. 6.21). However a simultaneous silica and carbonate substitution, i.e., the formation of a thaumasite-ettringite solid solution with a substitution of 20% thaumasite, is calculated for point K,

Fig. 6.22. The AFt phase at point K has a composition corresponding to Ca6Al1 6Si0 4(SO4)2 8

(CO3)0 4(OH)12⋅25.6H2O. The increasing silica and carbonate substitution is ascribed to the increase in silica and carbonate activities between the two points. The increase in silica activity also affects the constitution of C-S-H: note that the Ca/Si molar ratio of C-S-H, ~1.6 at point I, decreases to ~1.25 at point K. Thus the calculations indicate that silica and carbonate substitution in AFt is more likely to occur i) in silica-rich blended cements, assuming an excess of calcite is present, or ii) in portlandite-depleted cements which have undergone leaching.

The AFt phase in Portland cement 108

Table 6.2: Composition of aqueous phase of selected stable phase assemblages and invariant points in the

system CaO-Al2O3-SiO2-CaSO4-CaCO3-H2O at 1bar and 5 and 25°C (sorted in order of decreasing OH molalities, excess calcite present)

I I I I I II Pt. Temp pH OH Ca Al CO3 SO4 Si Ca/Si Solids [°C] [-] [mmol/kg] [-] A 25 12.48 34.44 19.37 0 0.0065 0 0.0287 1.57 Cc, CH, C-S-H 5 13.27 39.05 21.17 0 0.0055 0 0.0392 1.59 Cc, CH, C-S-H B 25 12.48 34.44 19.37 0.0074 0.0065 0 0.0287 1.57 Cc, CH, Mc, C-S-H 5 13.27 39.05 21.17 0.0015 0.0055 0 0.0392 1.59 Cc, CH, Mc, C-S-H

I 25 12.48 34.42 19.39 0.0074 0.0065 0.024 0.0287 1.57 Cc, CH, Mc, AFt(ss 9%CO3), C-S-H

5 13.27 39.05 21.17 0.0015 0.0055 0.004 0.0392 1.59 Cc, CH, Mc, AFt(ss33%CO3), C-S-H H 25 12.48 34.38 19.55 2E-04 0.0065 0.18 0.0287 1.57 Cc, CH, AFt(ss), Th, C-S-H 5 13.27 39.05 21.18 3E-04 0.0055 0.01 0.0392 1.59 Cc, CH, AFt(ss), Th, C-S-H G 25 12.47 34.38 19.53 0 0.0065 0.18 0.0287 1.57 Cc, CH, Th, C-S-H 5 13.27 39.04 21.18 0 0.0055 0.01 0.0392 1.59 Cc, CH, Th, C-S-H T 25 12.43 31.62 30.69 0 0.0065 12.47 4E-04 --- Cc, CH, Th., Gp 5 13.24 36.69 30.66 0 0.0055 10.56 4E-05 --- Cc, CH, Th., Gp S 25 12.43 31.62 30.69 3E-07 0.0065 12.47 4E-04 --- Cc, CH, AFt(ss), Th, Gp 5 13.24 36.69 30.66 8E-09 0.0055 10.56 4E-05 --- Cc, CH, AFt(ss), Th, Gp J 5 12.95 18.45 9.75 0.013 0.0060 0.005 0.110 1.37 Cc, Mc, AFt(ss), Th, C-S-H C 25 12.06 12.86 6.99 0.130 0.0070 0 0.098 1.25 Cc, Mc, Str, C-S-H 5 12.68 9.60 5.28 0.085 0.0065 0 0.166 1.10 Cc, Mc, Str, C-S-H K 25 12.06 12.85 7.01 0.130 0.0070 0.028 0.098 1.25 Cc, Mc, Str, AFt (ss 20%Th), C-S-H 5 12.69 9.85 5.28 0.085 0.0065 0.004 0.166 1.10 Cc, Mc, Str, Th, C-S-H L 25 11.80 6.96 3.89 0.140 0.008 0.084 0.145 1.00 CC, Str, AFt(ss), Th, C-S-H 5 12.66 8.21 4.88 0.110 0.007 0.006 0.113 --- Cc, Mc, Str, AFt(ss), Th

R 25 11.51 3.50 2.28 0.268 0.009 0.270 0.095 --- Cc, Str, AFt(ss), Th, AH3

5 12.38 4.32 2.61 0.205 0.008 0.019 0.054 --- Cc, Str, AFt(ss), Th, AH3

D 25 11.42 2.88 1.71 0.220 0.009 0 0.240 0.86 Cc, Str, AH3, C-S-H

5 12.21 3.10 1.86 0.140 0.009 0 0.200 0.85 Cc, Str, AH3, C-S-H

M 25 11.42 2.75 1.84 0.210 0.009 0.160 0.240 0.86 Cc, Str, Th, AH3, C-S-H

5 12.21 3.20 1.87 0.140 0.009 0.008 0.196 0.86 Cc, Str, Th, AH3, C-S-H

Q 25 10.23 0.190 15.74 0.015 0.008 15.62 0.004 --- Cc, AFT(ss), Th, AH3, Gp

5 10.61 0.077 14.19 0.004 0.007 14.18 3E-04 --- Cc, AFT(ss), Th, AH3, Gp

E 25 9.84 0.073 1.21 0.048 0.023 0 4.22 0.83 Cc, AH3, C-S-H, SiO2(am)

5 10.33 0.042 0.965 0.018 0.021 0 3.06 0.83 Cc, AH3, C-S-H, SiO2(am)

W 25 9.84 0.073 1.2 0 0.023 0 4.19 0.83 Cc, C-S-H, SiO2(am)

5 10.33 0.042 0.96 0 0.021 0 3.04 0.83 Cc, C-S-H, SiO2(am)

N 25 9.79 0.065 1.68 0.043 0.021 0.580 4.00 0.83 Cc, Th, AH3, C-S-H, SiO2(am)

5 10.33 0.041 1 0.017 0.021 0.054 3.03 0.83 Cc, Th, AH3, C-S-H, SiO2(am)

F 25 9.02 0.011 0.32 0.007 0.280 0 2.26 --- Cc, AH3, SiO2(am)

5 9.47 0.006 0.26 0.002 0.250 0 1.47 --- Cc, AH3, SiO2(am)

X 25 9.02 0.011 0.32 0 0.280 0 2.26 --- Cc, SiO2(am)

5 9.47 0.006 0.26 0 0.250 0 1.47 --- Cc, SiO2(am)

U 25 8.56 0.004 15.73 0 0.055 15.62 2.07 --- Cc, Th, SiO2(am), Gp

5 8.46 6E-04 14.27 0 0.13 14.18 1.26 --- Cc, Th, SiO2(am), Gp

O 25 8.56 0.004 15.73 0.003 0.055 15.62 2.08 --- Cc, Th, AH3, SiO2(am), Gp

5 8.46 6E-04 14.27 3E-04 0.133 14.18 1.42 --- Cc, Th, AH3, SiO2(am), Gp

V 25 8.46 0.003 15.72 0 0.068 15.63 2.04 --- Cc, SiO2(am), Gp

P 25 8.46 0.003 15.72 0.002 0.068 15.63 2.04 --- Cc, AH3, SiO2(am), Gp Iall molalities given as total concentrations independently from the individual speciation; II Ca/Si of C-S-H if present abbreviations: Cc - Calcite, CH - Portlandite, Mc - Monocarboaluminate, Str - Strätlingite, AFt(ss 8%Th) - ettringite-thaumasite solid solution with a calculated substitution of 8% Thaumasite, AFt(ss) - limiting ettringite-thaumasite solid solution with a calculated substitution of 50% Thaumasite (if Thaumasite is present), Th - Thaumasite (minor Al substitution possible), AH3 - gibbsite, Gp - Gypsum, SiO2(am) - amorphous Silica, C-S-H - Calcium-Silicate-Hydrate

The AFt phase in Portland cement 109

In the current model, C-S-H is treated as an ideal solid solution between tobermorite and jennite- type end members (see chapter 4.4.4). The incorporation of other species into C-S-H is not automatically calculated in the present state of the database. Richardson and Groves [160] describe aluminium for silica substitution in the C-S-H structure. The amount of substituted aluminium increases with decreasing Ca/Si ratios and can be estimated using the relations described in [160]. To assess the chemical constitution of C-S-H in cement undergoing “sulfate attack” it is important to consider the sorption of sulfate to C-S-H; data for temperature-dependent values for the adsorption of sulfate to C-S-H can be found in [14][15][54][160]. Furthermore Hong and Glasser [92] provide data for the sorption of alkalis to C-S-H. Thus the chemical constitution of C-S-H can in principle be corrected with these data.

As demonstrated in Fig. 6.22, AFt solid solutions are generally predicted to be stable at lower sulfate concentrations than are required for thaumasite precipitation. However it is interesting to note that this trend is reversed if the silica activity is further increased, e.g. in systems with C-S-H with a Ca/Si ratio <1 (between points L and N): thaumasite is predicted to form at lower sulfate concentrations than are needed to form ettringite solid solutions, which could markedly influence the pathway of thaumasite formation in relevant compositions. But in this context it has to be noted that, compared to Portland and blended Portland cements with lower silica activity, the threshold sulfate concentration to form thaumasite and/or AFt-thaumasite solid solutions (Points L and M) is up to 10 times higher.

As shown in Fig. 6.22 and Table 6.2, the AFm chemistry is significantly affected by changing hydroxide, silica and carbonate activities. Whereas at portlandite and calcite saturated conditions monocarboaluminate dominates, strätlingite will form instead at increased silica activities. At 25°C monocarboaluminate is predicted to coexist with C-S-H at Ca/Si ratios ≥ 1.25. Based on the currently available dataset, strätlingite is expected to coexist with C-S-H at ~0.9 ≤ Ca/Si ≤ 1.25. AFm phases become unstable if the sulfate activity is increased. As shown in Fig. 6.22 and in agreement with calculations by Damidot [43], strätlingite is predicted to persist to higher sulfate molalities than monocarboaluminate. This might be one possible explanation of the better perfor- mance of slag- or metakaolin-rich, strätlingite-containing cements to sulfate exposure. Thus in the region L-M-R of Fig. 6.22 it is interesting to note that strätlingite is predicted to coexist stably with thaumasite and calcite, whereas monocarboaluminate is incompatible with thaumasite at 25°C.

A further increase of the sulfate activity leads to an enrichment of silicate and carbonate in the AFt phase before the phase boundary (line H-L) is crossed and thaumasite coexists with the limiting

AFt solid solution with 50 mol% thaumasite substitution (Ca6(Al,Si)(SO4)2 5(CO3)(OH)12⋅25H2O). As demonstrated in Fig. 6.22 thaumasite is more stable than AFt solid solutions at low hydroxide activities (and increased carbonate and silica activities, see Table 6.2). For example at invariant point O the coexistence of thaumasite with gibbsite, amorphous silica, gypsum and calcite at pH ~8.56 is predicted, which indicates a high stability of thaumasite with respect to carbonation processes, whereas the AFt solid solution is only computed to be stable at pH ≥ 10.23 (point Q).

The practical implications of Fig. 6.22 are that, at calcite saturation and even in the presence of excess calcite, thaumasite will not appear at equilibrium in either plain or blended Portland cements at 25°C. As shown in Fig. 6.22 and Table 6.2, only minor carbonate substitution occurs in the AFt phase in hydrated Portland cements or limestone blended cements saturated with respect to calcite and portlandite. However in the presence of excess calcite the amount of AFt is less limited in blended cements by the available sulfate content, as increasing substitution by a thaumasite

The AFt phase in Portland cement 110 component, including silica and carbonate, becomes energetically possible. Moreover as will be explained in a subsequent section, temperatures <25°C enhance the stability of substitutions in ettringite. Furthermore in most cases where C-S-H has a Ca/Si ratio ≥ 1, the silicate- and carbonate- rich AFt solid solutions are precursors of the precipitation of thaumasite at 25°C in cements undergoing sulfate attack.

Fig. 6.23 shows the coexisting phase assemblages and invariant points of part of the system CaO-

Al2O3-SiO2-CaCO3-CaSO4-H2O at 5°C. In comparison to Fig. 6.22 it is obvious that the stability fields of thaumasite shift significantly to lower sulfate concentrations. A comparison of the relevant invariant points is given in Table 6.2. Furthermore, the stability field of the ettringite-type AFt solid solutions coexisting with C-S-H in region H-I-J decreases due to the higher stability of thaumasite. Not surprisingly, therefore, the conditions required for thaumasite formation at 5°C are attained with less alteration of the cement composition than at 25°C.

Fig. 6.23: 3D-representation of part of the system CaO-Al2O3-SiO2-CaCO3-CaSO4-H2O at 5°C and 1bar total pressure (solubility surfaces representative for solids in equilibrium with excess calcite; see Table 6.2 for additional carbonate and silicon concentrations at relevant invariant points)

The AFt phase in Portland cement 111

Carbonate substitution in the ettringite-type AFt phase is significantly increased at lower temperatures. For example the predicted carbonate substitution for point I, most relevant to ordinary Portland cements, increased from ~9 mol% at 25°C (AFt composition:

Ca6Al2(SO4)2 73(CO3)0 27(OH)12⋅26H2O) to ~33 mol% at 5°C (AFt composition: Ca6Al2(SO4)2(CO3)

(OH)12⋅26H2O). Another interesting feature of Fig. 6.23 is the predicted coexistence of monocarboaluminate with thaumasite in the region J-K-L which was an incompatible phase assemblage at 25°C. In this context small errors in the thermodynamic data become important. The free energy difference between the two competing phase assemblages i) monocarboaluminate+thaumasite or ii) ettringite+monocarboaluminate is small (see Fig. B.3, appendix B). Minor changes in numerical values of the thermodynamic parameters will influence the predicted phase assemblages significantly. However it is difficult to experimentally prove the calculations. Mixtures of monocarboaluminate+thaumasite or ettringite+monocarboaluminate might not react, due to the small thermodynamic driving force of reaction as indicated in Fig. B.3 in the relevant temperature range (~7.5±7.5°C). The calculations disclose no significant silica substitution in AFt at invariant points I at either temperature, 5 and 25°C. This result should be regarded as trend rather than as fixed overall valid 0 statement. For example a change of the ΔfG of thaumasite of ~-1300 J/mol at 25°C, involving a stabilisation of thaumasite, would change the result and predict a thaumasite-ettringite solid solution instead of SO4-CO3-AFt. This would imply simultaneous carbonate and silica substitution could occur at point I at both 5 and 25°C. Another possibility, the formation of ternary solid solid solutions involving ettringite, thaumasite and tricarboaluminate, was not considered in the calculations but is of potential importance.

The AFt phase in Portland cement 112

6.5. Discussion

6.5.1 Interpretation of phase diagrams In the interpretation of Fig. 6.22 and Fig. 6.23 phase boundaries have their usual significance: to cross a phase boundary, solids have to reconstitute and, to change the phase assemblage, considerable mass transport/ exchange must occur partially to reconstitute the solids: not only will the nature of the phases change, some appearing and others disappearing, but even those phases which persist may change in amount. Thus the relative size of the phase fields shown in Fig. 6.22 and Fig. 6.23 is not relevant to the importance of the assemblage. Instead, its importance depends on how much mass has to be transported to achieve change. This cannot be inferred directly from inspection of the Figures. However consideration of the mass balances with knowledge of the constitution of the phases generally discloses what has to change and how much mass is required to be transferred in order to drive a particular transformation.

6.5.2 Formation of AFt solid solutions The composition of ettringite has long been known to permit novel substitutions, for example of borate for sulfate, as occurs in charlesite. But under conditions prevailing in normal Portland cements, the composition of the AFt phase has generally been regarded as being close to that of ettringite, Ca6Al2(SO4)3(OH)12⋅26H2O. The present study is not complete with respect to the composition of AFt in cement but it does provide new information on carbonate and sulfate balances as well as on solid solutions between AFt and thaumasite.

The composition of AFt at constant temperature is essentially fixed by four variables: the activities of sulfate, hydroxide, carbonate and silicate. However while the actual numerical values of these activities range widely; different types of control affect the parametric values. In the presence of portlandite the pH is essentially controlled by the solubility of portlandite. Thus hydroxide is an abundant aqueous species, which is present in relatively high concentrations. However carbonate concentrations in the aqueous phase are numerically much lower and, if one wants to preserve cement phases such as C-S-H, AFt and portlandite, their stabilities are limited by the solubility of calcite to ~ 0.007 mmol/kg carbonate at 25ºC because at higher carbonate activities, these phases are susceptible to carbonation. Likewise the sulfate ion concentrations are effectively buffered by the solubility of AFt and are on the order of 0.02 mmol/kg; Table 6.2 shows exact values for the relevant aqueous concentrations at the relevant invariant points. An important additional factor affecting phase composition is temperature. The calculated phase relations are very sensitive to relatively small temperature differences in the range 0- 25ºC. Thus if Fig. 6.22 and Fig. 6.23, together with Table 6.2, are compared, several trends become evident. First the range of stable ettringite-type AFt solid solutions enlarges rapidly with decreasing temperature with respect to the maximum possible amount of carbonate and silicate substitution. In particular, the stability of carbonate within the AFt structure is greatly enhanced at lower temperatures with the result that extensive solid solution formation towards the “carbonate-AFt “ composition is more likely at low temperatures, e.g. at 5ºC. This solid solution is not however continuous, but can extend to a maximum of ~60 mol% carbonate end member in SO4-AFt, depending on the boundary conditions. This helps to explain the observation that a large increase in amount of ettringite seems to precede thaumasite formation; it is characteristically a low temperature feature, and carbonate can bulk out and reduce the amount of sulfate required for formation of an ettringite- like AFt phase.

The AFt phase in Portland cement 113

Table 6.3: Comparison of calculated activities for selected aqueous species relevant to point I, Fig. 6.22, in dependence of the alkali content of the aqueous phase at 25°C (see Table 6.2 for aq. composition)

activity ratio

-- 2- 2- pt. solids pH [OH ] [CO3 ] [SO4 ] [OH]/[SO4] [CO3]/[SO4]

AFt(ss, 9%CO ), Mc, I (no alkalis) 3 12.48 2.99E-02 4.71E-07 6.69E-06 4476 0.070 calcite, C-S-H, CH AFt(ss, 8%CO ), Mc, I’ (500 mmol Na+/kg) 3 13.58 3.60E-01 6.91E-05 1.11E-03 323 0.062 calcite, C-S-H, CH

While departures from the ideal ettringite composition were calculated not to exceed 10% at 25ºC, at 5ºC the solubilities of these other components in ettringite become too large to be ignored. The composition of AFt can be expressed in terms of four end members, ideal ettringite, the “carbonate ettringite” described by Carlson [36], hydroxy-substituted ettringite and ideal thaumasite. Note that in the first three cases, the mechanism can formally be defined in terms of simple isomorphous replacement, e.g., of sulfate by hydroxide or carbonate, but the replacement necessary to achieve thaumasite requires multiple substitutions.

Although the relevant activities are well defined for the alkali-free system (Table 6.2), commercial cements also contain alkalis -sodium and potassium- which perturb these numerical values and must be taken into account to apply the data to commercial cements. The main part of the alkalis in Portland cement is associated to alkali sulfates. As most of the sulfate is removed in the early stages of hydration, the role of alkalis is, most simply and directly, to elevate pH and thereby increase the hydroxide concentration. Hong and Glasser [92] showed that some alkali is also absorbed into solids. Sulfate remaining in solution plays a secondary role in satisfying the charge balance on alkali. But, at constant temperature, and as the alkali concentration increases, the soluble sulfate concentration also increases. This elevation of sulfate is also accompanied by rising silicate and carbonate concentration although these increases are less marked. But the order of concentration of anions in pore fluid -all potentially available for substitution in AFt- remains

[OH]>[SO4]> [SiO4] (or other appropriate silicate anion) > [CO3].

Thus at constant temperature, three factors govern the constitution of the AFt phase: i) the activity fractions of anion species ii) numerical values of the partition coefficient of anions between aqueous and solid for each relevant species and each relevant solid, iii) the possible occurrence of two AFt type phases and iv) the competition of other solids, e.g. AFm, often for the same anions.

First, at the self-generated pH imposed by portlandite, the carbonate activity generated by carboaluminates and/or calcite, and the silicate activity generated by high Ca/Si ratio C-S-H, the AFt phase can be expected to have close to the ideal ettringite composition, with not more than about 9 mol% of sulfate anions replaced by carbonate at 25ºC, point I Fig. 6.22. Thaumasite is not stable under these conditions. Secondly, if alkalis are present such that the pH rises above the - 2- portlandite threshold, both aqueous [OH ] and [SO4 ] activities in pore fluid increase. This relation is demonstrated in Table 6.3 for point I, Fig. 6.22 at 25°C. However, perhaps unexpectedly, the activity ratio [OH]/[SO4] decreases, as shown in Table 6.3, which indicates that despite an increase of the hydroxide activity resulting from alkali addition as (Na,K)OH, not much sulfate will be replaced in the cement solids by OH with increasing alkali content. Thus OH- substitution is expected to be of minor importance in ettringite even in the presence of increasing hydroxide contents. A similar relation can be derived for the expected carbonate substitution. The [CO3]/[SO4]

The AFt phase in Portland cement 114 activity ratio decreases slightly with increasing alkali content. As a consequence, the maximum carbonate substitution in AFt at I’ decreases also slightly to 8% (Table 6.3).

Finally, increasing the carbonate activity beyond that conditioned by the solubility of calcium carbonate is not practical without at the same time beginning to decompose the normal matrix- forming phases, e.g., C-S-H. Thus while carbonate activity affects the formation of thaumasite, not much latitude exists for increasing carbonate activities in relatively unaltered cement, although this restriction is clearly removed during the later stages of alteration, for example after portlandite has been depleted.

An important consequence of these considerations is that i) the composition of the AFt phase is likely to undergo reversible changes as functions of temperature and activity of potential anionic substituents, ii) the amount of AFt that can form is not necessarily limited by the sulfate content: other ions may substitute and thereby enhance AFt formation, especially at low temperatures and iii) as a result of the temperature dependence and composition of the constitution of AFt, and of other concomitant phase changes, the specific volume of the solids and the amount of bound water may exhibit significant changes over a narrow range of temperatures, 5°C -25ºC, with marked increases expected to occur at low temperatures.

6.5.3 Pathways of thaumasite formation Literature studies tend to distinguish between deterioration in cements arising from ettringite formation and those caused by thaumasite. The logic of this distinction needs to be explained. In the classic case, of the ettringite form of “sulfate attack”, the amount of ettringite is increased either (or both) by a temperature-dependent redistribution of sulfate (“internal sulfate attack”) and ingress of additional sulfate from the service environment. However the thaumasite form of sulfate attack (TSA) does not depend on sulfate alone: carbonate and silicate are required. As with enhanced ettringite formation, the sources of silica and sulfate could be internal or external, or both. Thus the required mass of species necessary to form thaumasite could arise from i) migration from the service environment of sulfate, carbonate and silicate, ii) leaching, iii) availability of one or more of sulfate, silicate and carbonate from aggregate, from fillers added to cement or from C-S-H, or iv) some combination of the above sources. Thermodynamic calculations do not reveal or distinguish between possible sources of sulfate, carbonate and silicate but they do suggest i) that, owing to the estimated theoretical stability of thaumasite at temperatures up to 40°C and possibly as high as 68°C, thaumasite formation is not necessarily and uniquely a low-temperature process and ii) that reactions involving the AFt phase are part of a continuum and that division into characteristic types, e.g., TSA, is operationally convenient but not fundamental.

Fig. 6.22 and Fig. 6.23 show that, considering plain Portland cement, several pathways are possible for thaumasite formation; there is no single route to form thaumasite. These pathways were outlined in detail by Damidot et al. [44] and, while the present study provides additional fine detail regarding the influence of temperature and solid solution formation, the basics remain unaffected. Thus Fig. 6.22 and Fig. 6.23 do not show all the possible pathways of thaumasite formation because it is constructed for the special condition that calcite saturation is maintained. However from published descriptions of field occurrences of TSA [41][42] it appears that this condition obtains in the majority of cases (in some examples uncertainties remain because the necessary data are not given). So Fig. 6.22 and Fig. 6.23 probably illustrate the most commonly-encountered reaction pathways to thaumasite formation.

The AFt phase in Portland cement 115

Depending on the composition of the service environment, it would be possible to form thaumasite at both 5°C and 25ºC either in the presence of portlandite and C-S-H (along G-H, Fig. 6.22 and Fig. 6.23) or if leaching and reaction removed free portlandite, by first forming a substituted AFt solid solution together with C-S- H (in region (H-I-K-L Fig. 6.22 or region H-I-J Fig. 6.23) and subsequently, mixtures of thaumasite with AFt. This route seems to be commonly encountered in field studies and has led to the belief that AFt is a necessary precursor to thaumasite. In the sense that AFt is a normal constituent of most cement, including cement not showing signs of alteration, this assertion is true. But as the diagrams show, i) a number of sequences are possible and ii) because of compositional variation in AFt, including partial substitution of sulfate by carbonate and aluminium by silicate, the amount of AFt phase may well increase significantly before thaumasite first appears. The extent of this increase will be more important at low temperatures and underlies the conclusion of several field studies, that often considerable amounts of AFt are reported in these samples.

However as shown in Fig. 6.22 and Fig. 6.23 it is possible that zones evolve in the course of reaction in which progressive decalcification due to leaching occurs. These leached zones are likely to be located close to the surface exposed to the aggressive solution. The silica activity is relatively high due to the formation of C-S-H with low Ca/Si ratios ≤ 1 as a consequence of leaching. In these zones, thaumasite is predicted to form directly, without AFt, along the connection between points L-M-N at 25°C and J-K-M-N at 5°C.

Thus the case study has shown that several pathways of thaumasite formation exist, rather than a single route, e.g. as suggested by Bensted [22]. It is likely that in most concretes exposed to sulfate- rich environments zoned microstructures will evolve and, as a consequence, several pathways to thaumasite formation occur simultaneously as sequence of different local equilibria. The initial formation of AFt solid solutions is an essential precursor to form thaumasite in the calcium-rich core zones whereas in the decalcified near surface zones thaumasite may form directly.

In conclusion, the thaumasite problem is undoubtedly complicated. But this is not because of any special properties of thaumasite itself, although its complex chemistry does increase the number of cement components required for its formation. All the evidence is that thaumasite forms “normally”, in response to the composition of cement and its interaction with its service environment. The physical chemical “rules” governing the formation of thaumasite are not different than for other cement phases. An ordinary Portland cement cured at ambient temperatures will never form quantities of either, or both, ettringite and thaumasite sufficient to cause subsequent physical expansion and cracking. But thermal cycling and chemical exchanges between cement paste and its service environment can cause problems.

The influence of limestone addition on cement hydration 116

7. The influence of limestone addition on cement hydration

7.1. Literature review

7.1.1 Limestone addition to Portland cement Cement manufacturers are under pressure to reduce costs and decrease emissions while at the same time, producing a quality product with high early strength and superior durability. Limestone is much used in this context as a partial replacement for cement and appears to be innocuous at least up to 5% replacement: it is a permitted additive under the EN 197 specification.

The role of calcite, as inert “filler” or as a reactive admixture (or both), has been much studied in the literature. Thus numerous investigations exist, dealing with aspects of limestone addition. Many of the studies use high levels of calcite, such that most of the calcite acts as a filler. Therefore the literature review was limited to publications relevant to this study, using mainly lower additions, 1-10 wt.-%. The review is selective, and concentrates on what, from present knowledge, appear to be the salient publications.

Mineralogical changes Following the discovery of monocarboaluminate by Bessey [27] and its successful synthesis by Carlson and Berman [36], there is little doubt about the reactivity of calcite in limestone blended Portland cements. However researchers argue about the extent of reaction and its consequences to Portland cement hydration. While Sprung and Siebel [175] concluded that only a small fraction of limestone is reactive, and therefore no effects on cement hydration will be expected, Seligman and Greening [169], Kuzel et al. [113][115][116][117] and others [29][37][166] observed significant changes in the mineralogy of hydrated Portland cements in the presence of carbonate or calcium carbonate, the main mineral component of limestone. Seligman and Greening [169] showed that the addition of 5% calcite to an originally carbonate- free Portland cement (44.5% C3S, 28.0% C2S, 12.2% C3A, 6.8% C4AF and 1.83% SO3, w/c=0.5) led to significant changes of the mineralogy of the hydrated cement paste. After 28d moist curing a monosulfoaluminate solid solution was detected in the carbonate-free control sample; ettringite peaks were absent in the XRD-pattern. However the calcite blended samples showed significant amounts of ettringite as well as monocarboaluminate after the same curing period. Hemicarboaluminate was found as an intermediate compound after 14d of hydration which converted to monocarboaluminate with time. The carbonate-free control sample was reinvestigated after 14 years of hydration. Monosulfoaluminate persisted as the main sulfoaluminate phase, but no formation of ettringite was observed.

Kuzel [113] showed the influence of the carbonate content on the hydration products of OPC. Even low carbonate contents in OPC led to changes of the paste mineralogy. In nearly carbonate-free systems (<0.1 wt.-% CO2) the amount of initially formed ettringite diminished and a sulfate-type

AFm was observed after 50 h hydration. In contrast, at only slightly higher CO2 contents of ~0.4 wt.-% CO2, more ettringite and less monosulfoaluminate were observed together with hemicarboaluminate. A further enhancement of the carbonate content to ~0.9 wt.-% CO2 led to the formation of mono- and hemicarboaluminate while the amount of initially formed ettringite persisted unchanged. Formation of monosulfoaluminate was not observed. Furthermore, Kuzel referred to the possibility of delayed ettringite formation if carbonate-free cement, previously hydrated for 310 days, was subsequently exposed to atmospheric CO2 in combination with

The influence of limestone addition on cement hydration 117

sufficient humidity. It was shown that access of CO2 led to a transformation of sulfate-AFm to monocarboaluminate and the sulfate thus released initiated secondary ettringite formation. A shift of the hk0 reflections of ettringite was observed and could be a sign of partial replacement of sulfate anions in the AFt structure, presumably by carbonate.

Bonavetti et al. [29] investigated the hydration of OPC blended with 20 wt.-% CaCO3. Ettringite and monocarboaluminate were detected after 3 days of hydration. After 28 days the amount of ettringite diminished and monosulfoaluminate coexisted with monocarboaluminate. In the course of hydration, monosulfoaluminate disappeared and monocarboaluminate and ettringite were found as apparently stable hydration products after 90 days. In agreement with thermodynamic calculations by Damidot et al. [49], hemicarboaluminate was not found in the presence of CaCO3.

Sawicz and Heng [166] found that limestone prevents the conversion of AFt (ettringite) into monosulfoaluminate. Monocarboaluminate and hemicarboaluminate are formed instead of mono- sulfoaluminate. On the other hand, Ramachandran and Zhang [157] concluded, but only from thermal analysis experiments, that the conversion of ettringite to monosulfoaluminate is accelerated by the presence of CaCO3. Thus the literature reveals divergent opinions about the influence of limestone on the hydration of Portland cement.

Physical and mechanical properties Ingram and Daugherty [94] review the many claims for the physical attributes of limestone. The opinions about possible benefits of the addition of limestone conflict. Studies by Bobrowski et al. (cited in [94]) indicate that the addition of 3-5% limestone does either improve, or has no effect on the engineering properties of limestone blended cement. On the other hand Adams and Race (cited in [94]) found that 2-5% addition of limestone significantly increased the 4d drying shrinkage of mortars. A cement with low C3A content was used in their experiments. Bedard and Bergeron

(cited in [94]) found no negative influence of 4.1 and 2.3 % limestone addition to C3A-rich cement. Concretes cured at ambient temperatures had compressive strengths similar to the carbonate-free control sample at 1-56 d. Ingram and Daugherty concluded that the addition of up to 5% limestone has at worst no effect on the engineering properties of the cement, but at best, has a positive influence. Sprung and Siebel [175] investigated the strength development and freeze thaw resistance of different limestone blended cement concretes. They concluded that concretes made from cements blended with up to 20% limestone have similar performance compared to Portland cement-only concrete. Soroka and Setter [174] investigated the effect of different fillers on the compressive strength of blended cements. It was shown that limestone blended cement samples exhibited the largest gain in strength, compared to other pozzolanic fillers, including Rhine trass. A maximum strength increase was found at ~10% limestone substitution of the cement especially in the early age samples. Soroka and Setter argued that the strength increase is mainly due to an acceleration of cement hydration due to the availability of crystallisation nuclei provided by the fillers rather than being caused by a chemically induced strength gain due to the formation of carboaluminates.

The influence of limestone addition on cement hydration 118

Tsivilis, et al. [186][187][188] published a series of papers to the effects of limestone addition on the engineering properties of Portland cements and concretes. They showed that addition of up to 10% limestone either increased or had no effect on the 1-28d compressive strength of hydrated mortars (w/c=0.5). Concrete investigations showed that up to 20% limestone addition led to a marginal decrease of compressive strength, while 35% limestone substitution caused a significant loss of compressive strength. In another study Tsivilis et al. [187] showed that limestone additions up to 15% did not significantly change the total porosity of concretes. Furthermore they showed a positive influence of increasing C3A contents on the total porosity and permeability of limestone blended cements.

Catinaud [37] investigated the role of limestone on the leaching mechanisms of hydrated cement paste and paste containing C3S, C3A and gypsum blended with calcite (w/c=0.5) at 25°C. An analysis of the porosity of the initial state of the samples before immersion in water for the leaching experiments showed an increase in porosity with increasing limestone addition (5, 10, 20 wt.-%

CaCO3). However samples were vacuum dried for 3h at 105°C before the porosity measurement. Initially portlandite, carboaluminate phases and AFt were present in the samples. No analysis was provided after drying at 105°C, so the extent of drying damage, if any, is unknown. The literature review has shown that there are divergent opinions about possible benefits of the addition of limestone. Claims of decreased porosity and strength increases could not be proven, although the majority of the literature on limestone addition showed no severe reduction of the physical and mechanical properties if low amounts of limestone were added (<10%). The present study, including AFt- AFm phase relations, concentrates on the impacts of calcite on paste mineralogy. With help of space filling calculations and mineralogical maps in dependence of the composition of reactive cement fractions, a systematic investigation of the role of limestone has been used to provide better understanding of the quantitative aspects of cement hydration in the presence of calcite additions.

7.1.2 Influence of temperature on the mineralogy of hydrated (limestone blended) Portland cement: 25°C and above The literature about temperature-dependent changes on limestone blended cements is limited. Therefore the review was extended and takes Portland cements into account as well as delayed ettringite formation (DEF), a widely studied subject relevant to heat cured cements.

Until recently no thermodynamic dataset was available to assess the temperature-dependent stability of carboaluminate phases, i.e. hemi- and monocarboaluminate. Fentiman [62] examined the hydration of carboaluminate cement (limestone blended high alumina cement) at different temperatures. In his work he showed that monocarboaluminate formed in cures at least up to 50°C.

At temperatures above 55°C C3AH6 dominated the XRD-pattern, at least up to 7d hydration. Furthermore he reported an increase of the amounts of monocarboaluminate at the expense of

C3AH6 and calcite on curing the carboaluminate cement paste up to 7d hydration at 65°C and 70°C, which is in dispute with his earlier findings (see [62]).

Kuzel and Baier [114] investigated the hydration of calcium aluminate cements in the presence of calcite at different temperatures. They showed that monocarboaluminate is persistent up to 90±5°C; at higher temperatures it converts to C3AH6 and calcite. Kuzel and Baier showed that the addition of β-C2S led to an apparent stabilisation of the hydrogrossular phase, C3AS3-xH2x, due to the incorporation of silica in the structure. Monocarboaluminate, C3AS3-xH2x, gibbsite and calcite

The influence of limestone addition on cement hydration 119 coexisted at least up to 6 months at 60°C. However no siliceous hydrogarnet appeared in aluminate cement blended with powdered quartz or silica and cured at 25°C or 60°C.

Sulfate was absent in the studies discussed above. Investigations of the hydration of sulfate resisting Portland cement containing small amounts of calcite at elevated temperatures by Lothenbach et al. [124] have shown that monocarboaluminate tends to convert to monosulfoaluminate at the expense of ettringite at temperatures ~50°C. A similar reaction was shown by Ghorab et al. [75]. A mixture of ettringite and monocarboaluminate (mole ratios 1:1 and 10:1) decomposed in boiling water (~100°C) to monosulfoaluminate, hydrogarnet and calcite.

Provided sufficient sulfate is available (SO3/Al2O3 ≥ 3), ettringite itself is stable up to ~115±5°C as experimentally shown by Zhou and Glasser [195] and Hall et al.[87]. This was confirmed in investigations by Shimada and Young [171] who have investigated the stability of ettringite in alkaline solutions at 80°C. Ettringite persisted for 12h at 80°C although the formation of the U- phase, a sodium substituted AFm phase, was found at initial sodium concentrations ≥ 0.25 N. Furthermore the coexistence of ettringite with C-S-H in alkaline solutions was investigated. A mixture of C3S and ettringite was heat treated for 12h in various concentrations of NaOH solutions. With rising NaOH concentrations, increasing amounts of monosulfoaluminate were observed. Ettringite peaks were absent in the XRD-pattern at ≥ 0.75 N. No formation of the U-phase was reported. However in contrast to the experiments made with pure ettringite, the ratio ettringite:solution was significantly increased from 1:2 to 1:8. The increasing amounts of alkalis resulted in a strong increase of the aqeuous sulfate concentration, especially at high temperatures, which could have caused the precipitation of monosulfoaluminate, as observed in the latter case. Thus the reformation of ettringite after cooling cannot only be attributed to an increased sulfate sorption to C-S-H.

The influence of alkalis on DEF was shown by Wieker and Herr [194] and later confirmed by Famy and Kelham (both cited in [182]): increasing amounts of alkalis led to an increase of expansion of heat cured mortars, provided sufficient sulfate was available. The formation of monosulfoaluminate has been reported in heat cured samples mainly under the aspect of DEF. But why does monosulfoaluminate form at elevated temperatures and is converted to ettringite at lower temperatures? Calculations by Damidot and Glasser [46][47] showed that monosulfoaluminate is apparently metastable with respect to C3AH6 and AFt at 25°C, becoming stable at temperatures above ~40°C. This would be one possible explanation of the conversion of monosulfoaluminate to ettringite due to cooling of heat cured mortars. On the other hand, Paul and Glasser [144] reported the formation of a sulfate-substituted siliceous hydrogarnet due to prolonged heat curing of Portland cement paste for more than 8 years at 85°C. No AFm or AFt phase was found in these samples and it was assumed that sulfate was substituted into hydrogarnet.

A second explanation of the mechanisms of DEF is the increased sulfate sorption of C-S-H at increased temperatures which might cause a destabilisation of ettringite and force precipitation of monosulfoaluminate, which subsequently transforms back to ettringite due to slow desorption of sulfate from C-S-H at ambient temperatures, as discussed by Fu et al. [70] and Famy et al. [60]. In that respect the possibility of increased sulfate sorption on C-S-H due to increased amounts of alkalis has to be mentioned, as discussed by Glasser et al. [78]. Barbarulo [14][15] has reported increasing sulfate sorption in C-S-H with increasing temperatures and alkali contents.

The influence of limestone addition on cement hydration 120

Kuzel and Strohbauch [118] provide a third theory to explain the conversion of monosulfoaluminate to ettringite at ambient temperatures as they showed that monosulfoaluminate present in initially carbonate-free cements will be transformed to ettringite and carboaluminate due to subsequent carbonation of the paste. A similar mechanism was described by Seligman and Greening [169].

The literature review has shown that there is a general lack of systematic studies on the influence of carbonate on the hydration of Portland cement at temperatures other than room temperature. Moreover carbonate would appear to interact with sulfate, so the role of carbonate cannot be studied in isolation. The upper thermal stability limit of carboaluminates could not be estimated due to a lack of thermodynamic data. No consistent conclusion of the influence of temperature on the stability of carboaluminates is possible based on literature reports. In hydrated Portland cements it is generally accepted that monosulfoaluminate does form at elevated temperatures and converts to ettringite due to cooling. But there are divergent opinions about the mechanisms of this process.

The influence of limestone addition on cement hydration 121

7.2. Implications of limestone addition to cement hydration

7.2.1 Experiments on the role of carbonate and sulfate on C3A hydration AFm and AFt phases are the main hydration products of the aluminate phases in Portland cement clinker (C3A and C4AF): they react with water and sulfate or, if available, carbonate. Studies of pure phases generally help to explain the hydration mechanisms. Kuzel and Pöllmann [117] published probably the most extensive experimental study to the hydration of C3A in the presence of lime, calcium sulfate and calcium carbonate. Their main findings are summarised in chapter 5.1.3.

Due to the existence of several literature studies on the hydration of C3A in the presence of sulfate and carbonate at room temperature it was decided to concentrate on two experimentals paths to confirm the literature studies: (i) investigation of the hydration of C3A with calcium sulfate in a carbonate-free system and (ii) investigation of the hydration of C3A with calcium sulfate in the presence of excess calcite. Excess portlandite (as free lime) was added to each experiments to buffer the pH of the aqueous phase to ~ 12.48 at 25°C. Tricalcium aluminate, anhydrite and lime were freshly prepared according to the procedure descri- bed in chapter 4.1. Analytical grade calcium carbonate was used as carbonate source in respective experiments. Previously boiled and degassed ultra pure water (water/solid-ratio = 5) was used in all experiments. HDPE bottles served as storage containers for the samples. The reaction temperature was kept constant at ~25°C. The samples were initially stirred for 7d using magnetic stirrers and regularly agitated afterwards. To avoid CO2-contaminations all experiments were carried out under

N2-atmosphere. In the “carbonate-free” experiment (i) a mixture of C3A, CaSO4 and CaO in a 1:1:2 molar ratio were used as starting materials, whereas in experiment (ii) a molar composition

C3A:CaSO4:CaCO3:CaO = 1:1:1:2 was hydrated. To assess the kinetic pathway of hydration, mineralogical changes in the samples were monitored qualitatively by XRD at selected points in time over 90d. Part of the suspension was removed using a syringe and vacuum filtered subsequently using G4 Pyrex glass filter funnels in N2-atmosphere. Hydration was stopped and remaining water removed by flushing the samples with acetone. The subsequent XRD analysis of the dry powder was done using a BRUKER D8 ADVANCE powder diffractometer. The instrumental setup is described in chapter 3.2. No attempts were made to scale the reported reaction times to commercial cement systems. Although realistic ratios of oxides were used, the kinetics are typical of suspensions, not pastes, Moreover it was not intended to produce powder size distributions typical of ground commercial clinker.

Fig. 7.1 shows the mineralogical changes during hydration of C3A in the presence of sulfate

(SO3/Al2O3=1) and absence of carbonate. As expected from the literature, ettringite formed as initial hydration product together with portlandite, which was invariably present. The amounts of

AFt increased with increasing reaction time, while the amounts of anhydrite and C3A decreased. Anhydrite was fully consumed after ~14d of hydration which corresponds to a maximum amount of AFt present. Once all anhydrite has been consumed, AFt starts to convert to monosulfoaluminate due to reaction of AFt and water with the remaining C3A between 14 and 28 days. The characteri- stic peaks of C3A are absent at 28d. No mineralogical changes were observed between 28d and 90d, which indicate complete reaction after 28d. The phase assemblage observed after 90d of hydration comprises mainly monosulfoaluminate, small amounts of AFt and portlandite. The peak positions of monosulfoaluminate agree well with that of the C4AsH12 reference pattern in the PDF- database. This indicates no extensive hydroxide substitution in the structure (see chapter 5.3.1).

The influence of limestone addition on cement hydration 124

A comparison of Fig. 7.3 a) and b) shows that the addition of limestone slightly accelerates the reaction of C3A with calcium sulfate. Whereas in the carbonate-free sample anhydrite was consumed after ~14d of hydration, anhydrite could no longer be detected after ~7 d of hydration.

7.2.2 Phase relations between AFt-AFm phases relevant to Portland cements at 25°C The aim of this study was systematically to analyse the effects of limestone additions on the chemistry and mineralogy of hydrated Portland cement paste. It is first necessary to understand phase changes which occur as a consequence of limestone and/or carbonate addition to Portland cements. The ternary phase relations between sulfate-, carbonate- and hydroxy-AFm have been studied in chapter 5, while AFt phases were studied in chapter 6. Therefore the next logical step was to investigate the relations between AFt and AFm. Related experimental results are available in the literature [29][117] and experiments on the hydration of C3A were described in the previous chapter. Based on the thermodynamic data of the main cement hydrates and the aqueous species, listed in Table 4.1 and Table A.1, and including the derived AFt and AFm solid solution models, the stable invariant points including AFt and AFm phases in the system CaO-Al2O3-CaSO4-CaCO3-H2O were calculated. Previous works by Damidot et al. [43][46][49] dealt with this subject and are a useful basis for the following calculations. However one important outcome of the database development is the possibility that monosulfoaluminate is stable at 25°C, in contrast to earlier studies. Thus it seems justified to revise previous calculations using the dataset developed in this Thesis.

The phase relations between AFm, AFt, calcite and gypsum at 25°C are shown as a pseudo-3D phase diagram in Fig. 7.4. As limestone blending is mainly relevant to Portland cements, calculations assume an excess of portlandite, which acts as a pH buffer (pH ~ 12.43 - 12.48). Thus calcium concentrations remain almost constant (~19.4 mmol/kg) except in phase assemblages containing gypsum where aqueous Ca ~ 30.7 mmol/kg. The aqueous phase composition at relevant invariant points is given in Table 7.1. The phase assemblages shown in Fig. 7.4 generally agree with those depicted in Fig. 5.22, except that, due to phase rule restrictions, a parallel coexistence of monocarboaluminate, monosulfo- aluminate and hemicarboaluminate with portlandite and AFt is not possible. The calculations indicate that monosulfoaluminate and monocarboaluminate are incompatible in portlandite saturated conditions at 25°C. However thermodynamically it is very difficult to assess this claim due to the small free energy difference of reaction between the phase assemblage monocarboaluminate and hemicarboaluminate or monocarboaluminate and monosulfoaluminate coexisting with AFt and portlandite, respectively. Experimental results by Kuzel and Pöllmann [117] showed the formation of hemicarboaluminate, monosulfoaluminate and ettringite during reaction of C3A with gypsum and small amounts of calcite (initial molar composition of the mixture C3A:CaO:CaSO4:CaCO3=1:1:1/2:1/4). Later observations on hydrated Portland cements [113][115][116] blended with small amounts of calcite agreed with these findings. Experiments (shown in chapter 5.3.2, Fig. 5.19 b)) disclose the formation of increasing amounts of hemicarboaluminate due to reaction of monosulfoaluminate and monocarboaluminate in the presence of Ca(OH)2. Thus, in agreement with calculations and experiments, the stable coexistence of monocarboaluminate and monosulfoaluminate in hydrated Portland cements at ~25°C is unlikely.

The influence of limestone addition on cement hydration 125

Excess portlandite present in all assemblages

Fig. 7.4: Phase relations between AFm-AFt phases as well as gypsum, hydrogarnet and calcite in the system

CaO-Al2O3-CaSO4-CaCO3-H2O at 25°C (Note: excess portlandite present in all assemblages, dashed lines in region K’-L’-M’-N’ represent stability field of ideal monosulfoaluminate without hydroxide substitution).

Fig. 7.4 shows that all AFm phases are predicted to coexist with AFt at appropriate aqueous compositions. The calculations indicate the possibility of the coexistence of monosulfoaluminate 2- with AFt at sulfate concentrations of ~0.01 mmol SO4 /kg. This is in contrast to calculations by Damidot et al. [46] who reported the metastability of monosulfoaluminate with respect to ettringite and C3AH6 at 25°C. However their calculations used solubility data by Wells et al. [193] to derive thermodynamic data for C3AH6, which significantly differ from the dataset derived in this study (see Fig. 4.3, chapter 4). Compatibility experiments by Atkins et al. [10] did not give consistent results: on one hand formation of C3AH6 was not observed in a sample containing initially mono- sulfoaluminate and portlandite stored 180d at 25°C but on the other hand, no formation of mono- sulfoaluminate was observed in a mixture of AFt and C3AH6 stored under the same conditions.

As shown in chapter 5, part of the sulfate in monosulfoaluminate may be replaced by hydroxide ions. This solid solution formation has been considered during construction of Fig. 7.4. The calculations show that in the presence of portlandite and C3AH6 (points M and L, Fig. 7.4) up to ~24 mol% of the initial sulfate of monosulfoaluminate can be replaced by hydroxide. Increasing carbonate and sulfate concentrations decrease the extent of hydroxide substitution in monosulfoaluminate. However it is interesting to note that with the estimated thermodynamic data for monosulfoaluminate-hydroxy-AFm solid solutions a substitution of 11 mol% hydroxide is predicted in the presence of portlandite and AFt in the system (points N and K, Fig. 7.4), whereas as shown in chapter 5.4, Fig. 5.22, calculated for lower hydroxide activities, monosulfoaluminate approaches its ideal composition while coexisting with AFt in the absence of portlandite.

The influence of limestone addition on cement hydration 126

Table 7.1: Composition of aqueous phase of selected stable phase assemblages and invariant points in the system CaO-Al2O3-CaSO4-CaCO3-H2O at 1bar and 25°C (excess portlandite present, comparison with reported points in the literature and effect of solid solution formation)

I I I I Pt. pH Ca Al CO3 SO4 Solids Reference [mmol/kg] [mmol/kg] [mmol/kg] [mmol/kg]

A 12.48 19.38 0.087 0 0 C3AH6, CH t.s. 12.52 21.95 0.008 0 0 [49]

B 12.48 19.37 0.087 8.3E-06 0 C3AH6, Hc, CH t.s. 12.52 21.95 0.008 1.6E-03 0 [49] C 12.48 19.36 0.037 0.0003 0 Mc, Hc, CH t.s. 12.52 21.95 0.006 0.0050 0 [49] D 12.48 19.35 0.007 0.0065 0 Mc, calcite, CH t.s. 12.52 21.95 0.005 0.0069 0 [49] E 12.48 19.34 0 0.0065 0 calcite, CH t.s. 12.52 21.95 0 0.0069 [49] F 12.43 30.70 0 0.0065 12.47 gypsum, calcite, CH t.s. --- 33.90 0 0.0069 12.40 [50] G 12.43 30.69 5.7E-07 0.0066 12.47 gypsum, AFt, calcite, CH t.s. 12.49 33.90 1.0E-07 0.0069 12.40 [50]

H 12.48 19.37 0.007 0.0065 0.024 AFt(ss, 9%CO3), Mc, calcite, CH t.s. H’ II 12.48 19.37 0.007 0.0065 0.025 AFt, Mc, calcite, CH t.s. 12.52 21.90 0.005 0.0069 0.010 [50] I 12.48 19.35 0.037 0.0003 0.009 AFt, Mc, Hc, CH t.s. 12.52 21.90 0.006 0.0050 0.009 [50]

K 12.48 19.36 0.038 0.0002 0.008 AFt, Hc, Ms(ss; 11%OH), CH t.s. AFt, Hc, Ms, CH K’ II 12.48 19.36 0.044 0.0001 0.0075 t.s.

12.52 21.96 0.018 4.7E-05 0.004 [50]

L 12.48 19.35 0.087 8.2E-06 0.001 C3AH6, Hc, Ms(ss, 24%OH), CH t.s. II L’ 12.48 19.35 0.087 8.2E-06 0.002 C3AH6, Hc, Ms, CH t.s.

M 12.48 19.35 0.087 0 0.001 C3AH6, Ms(ss; 24%OH), CH t.s. II M’ 12.48 19.37 0.087 0 0.002 C3AH6, Ms, CH t.s. N 12.48 19.35 0.038 0 0.008 AFt, Ms(ss; 11%OH), CH t.s. N’ II 12.48 19.35 0.044 0 0.0075 AFt, Ms, CH t.s. O 12.43 30.69 5.7E-07 0 12.47 AFt, gypsum, CH t.s. 12.47 31.30 0.0003 0 11.40 [46] --- 31.56 0.06 0 12.33 [97] P 12.43 30.69 0 0 12.47 gypsum, CH t.s. 33.80 0 0 12.40 [50]

Iall molalities given as total concentrations independently from the individual speciation II calculated for ideal stoichiometry of monosulfoaluminate (Ca4Al2(SO4)(OH)12⋅6H2O) and AFt (Ca6Al2(SO4)3(OH)12⋅26H2O) abbreviations:, CH - Portlandite, Mc - Monocarboaluminate, Hc - Hemicarboaluminate, Ms(ss; 11%OH) - monosulfoaluminate solid 2- -- 2- solution with a replacement of 11% SO4 by 2OH ; AFt(ss; 9%CO3) - ettringite solid solution with a calculated replacement of 9% SO4 2- by CO3 , AFt - SO4-ettringite; t.s. - calculated in the title study

The influence of limestone addition on cement hydration 127

The calculations shown in Fig. 7.4 are in accord with experiments by Seligmann and Greening [169]. As a result of their investigations they stated:

“It is therefore proposed that another invariant point occurs … that represents the liquid phase composition in equilibrium with Ca(OH)2, C3AH6 and the lowest sulfate form of calcium sulfohydroxyaluminate hydrate as the solid phases.”(at 25°C)

This point corresponds to the invariant M in Fig. 7.4 comprising the phase assemblage SO4-AFm

(solid solution), C3AH6 and portlandite. The good agreement between both calculated and experimentally-derived results is another indication that a range of monosulfoaluminate solid solution compositions are stable at 25°C.

Region K’-L’-M’-N’ shows the stability field of ideal monosulfoaluminate. It is apparent that due to the stabilising effect of the solid solution formation, the stability field of hydroxide substituted monosulfoaluminate is markedly increased compared to that of the ideal SO4-AFm end member. Increasing carbonate concentrations lead to the formation of carboaluminate phases in the system

CaO-Al2O3-CaSO4-CaCO3-H2O at 25°C. Hemicarboaluminate is stable in region B-C-I-K-L in Fig.

2- 7.4 corresponding to a range of 8.1e-6 mmol/kg ≤ CCO3 ≤ 3e-4 mmol/kg at 25°C. In agreement with calculations by Damidot et al. [49], hemicarboaluminate is not stable in the presence of calcite but does coexist with monocarboaluminate. Monocarboaluminate, in contrast to hemicarbo- and monosulfoaluminate, can coexist with calcite and is thus stable in region C-D-H-I Fig. 7.4 (3e-4

2- mmol/kg ≤ CCO3 ≤ ~7e-3 mmol/kg at 25°C). Thus the formation of monocarboaluminate is often reported in blended cements containing >10 to 20 wt.-% calcite. As shown in point H, Fig. 7.4 , 2- monocarboaluminate is stable up to 0.024 mmol SO4 /kg, whereas hemicarbo- and mono- sulfoaluminate are metastable with respect to ettringite at sulfate concentrations ≥ ~0.01 mmol/kg.

The previously-developed solid solution model for carbonate- and sulfate-AFt was also considered in Fig. 7.4. AFt is stable over a wide range of sulfate and carbonate concentrations at 25°C. However the resulting self-generated carbonate activities are generally too low to enable significant carbonate substitution in the AFt phase. Thus AFt is predicted to have essentially its ideal composition at invariant points I, K, N, O, G. Only at invariant point H -which is however probably 2- of most importance to limestone blended cements- minor substitution of ~9 mol% CO3 is predicted to occur for AFt coexisting with monocarboaluminate and calcite.

Previous experiments have suggested a reaction pathway of C3A hydration which is consistent with the phase diagram Fig. 7.4. Only AFt is thermodynamically stable at sulfate activities generated by excess gypsum i.e., at aqueous compositions along line O-G, Fig. 7.4. Thus independently of the addition of calcium carbonate, AFm is not predicted to form as long as a physical excess of calcium sulfate is present. The consumption of calcium sulfate leads to a decrease in sulfate activity. The precipitation of monosulfoaluminate and/or carboaluminates in presence of AFt should therefore lead to an aqueous phase composition along the line N-K-I-H. The influence of alkalis on the calculated phase assemblages is discussed in chapter 7.5.1.

In experiments with excess calcite present, the intermediate formation of hemicarboaluminate was observed. One possibility to explain this behaviour is a temporary decrease of the carbonate activity to a region where hemicarboaluminate becomes stable due to the slow dissolution rate of calcite in the course of carboaluminate precipitation. With progressive hydration carbonate activities increase again, hemicarboaluminate becomes metastable, and reacts to form monocarboaluminate with time.

The influence of limestone addition on cement hydration 128

To summarise the key findings, a close relationship exists between the composition of the aqueous phase and the nature of solid phases forming. At portlandite saturation the pH is buffered by portlandite so the construction of Fig. 7.4 is, very nearly, at constant pH. But the activities of other species, carbonate and sulfate are also buffered: sulfate by gypsum, AFt or AFm, and carbonate by calcite, mono- or hemicarboaluminate. At 25°C, phases that contain sulfate do not contain carbonate, and vice versa, as a good first approximation (calculation discloses only limited substitution of carbonate for sulfate in AFt). Hence the three buffering systems, one for pH and one of several for sulfate and carbonate, act essentially independently of each other. The highest carbonate activities are obtained in the presence of calcite and the highest sulfate activity, in the presence of gypsum. However, carbonate activities are numerically low and, not surprisingly, aqueous solutions may not always achieve carbonate saturation in the presence of calcite especially during the early stages of hydration, when aluminates are react rapidly. Regarding phase compatibility, each AFm phase has a stability field where it can coexist with portlandite and other AFm and AFt phases and/or hydrogarnet at 25°C. All AFm phases can coexist with AFt, whereas monocarboaluminate is the only AFm phase which is compatible with calcite. Increasing carbonate concentrations lead to the transformation of C3AH6 and/or monosulfoaluminate solid solutions to hemicarboaluminate and finally to monocarboaluminate and calcite. The composition of the AFt phase in the system CaO-Al2O3-CaSO4-CaCO3-H2O at 25°C is close to ideal SO4-AFt. The self- generated carbonate activities are generally too low for extensive carbonate substitution in the AFt 2- phase; maximum ~9 mol% CO3 may be substituted in the presence of monocarboaluminate and calcite in portlandite-buffered systems at 25°C.

7.2.3 “Reactive” vs. “filler” calcite The previous investigations and calculations show that, depending on the initial chemical composition of the mixture, several AFm phases might coexist: thermodynamic calculations demonstrated different possible phase assemblages among the AFm-AFt compositions included in the study. To apply these data to commercial cements it is necessary to couple the initial chemical composition of the cement to the previous thermodynamic calculations. That will (i) enable prediction of the phase assemblages expected during Portland cement hydration in the presence of calcite and (ii) to calculate amounts of phases and thereby to quantify these processes.

To apply the data to commercial cements, it is argued as follows: Ca(OH)2 is of constant composition (but not necessarily in constant amount) and it is assumed for present purposes that sulfate, carbonate and aluminium included in C-S-H require a small but constant correction of the chemical totals. Quantitative studies of sulfate incorporation in C-S-H are in progress and in due course, it will be possible to correct for sulfate sorbed by C-S-H (see discussion paragraph 7.5.2). The maximum activity of carbonate is controlled by an excess of calcite, but these are too low to carbonate either Ca(OH)2 or C-S-H. In the first approach, at 25°C, minor carbonate substitution in AFt is neglected, but considered in supplementary plots (see appendix B). This study concentrates on the iron-free phases. However work is also in progress to assess the stability iron substituted AFm and AFt phases [132][133].

Fig. 7.5 shows the calculated phase assemblages that will typically occur during hydration of C3A in hydrated OPC and the dependence on initial sulfate and carbonate contents, expressed as ratios of SO3/Al2O3 and CO2/Al2O3, subsequently abbreviated as sulfate and carbonate ratios respectively. This method of representation enables all the relevant chemical variables to be displayed on a two- dimensional diagram without error.

The influence of limestone addition on cement hydration 130

2OH-). But, with increasing carbonate contents, the stability field of monosulfoaluminate solid solution decreases markedly. Monosulfoaluminate and the monosulfoaluminate-type solid solutions are not stable in the presence of monocarboaluminate and portlandite and are unlikely to persist in cements with a carbonate ratio > 0.5.

• OPC with very low carbonate contents will form hemicarboaluminate as an additional AFm phase in the course of hydration. But hemicarboaluminate is not stable at 25°C if excess calcite remains.

• Fig. 7.5 agrees well with previous experiments on the hydration of C3A in the presence of either sulfate or carbonate, or both. Depending on the availability of sulfate, alumina and

carbonate, phase boundaries will be crossed during C3A hydration. At initial stages of hydra-

tion, SO3/Al2O3 is typically >3, conditioned by the presence of excess calcium sulfate and phase

assemblage VI is observed. Once free calcium sulfate is depleted, the SO3/Al2O3 ratio decreases and phase assemblages containing AFm phases will be observed and, depending on the availability of carbonate, monosulfoaluminate solid solutions and/or carboaluminates will form.

• C3AH6 has only limited stability in the system CaO-Al2O3-CaSO4-CaCO3-H2O. It will not

occur if the initial solid SO3/Al2O3 ratio is greater than 0.75 and it is not stable in the presence of monocarboaluminate.

• The diagram, Fig. 7.5, is constructed for ~25ºC. The phase relations are however sensitive to temperature and calculations for low (~5°C) and elevated temperatures up to 99°C reveal substantial changes of phase assemblage predictions. The effect of temperature on the AFt- AFm phase assemblages is investigated in chapter 7.4.

The impact of adding calcite on the paste mineralogy can be predicted, using Fig. 7.5, as follows: select a horizontal line corresponding to the effective sulfate ratio of the cement. The intersection of this line with a vertical construction line corresponding to the appropriate carbonate ratio will enable the phase assemblage to be calculated with respect to the AFm portion and associated phases, as necessary to preserve mass balances. As can be seen from the slopes of the relevant boundaries, the appearances of hemicarboaluminate and of monocarboaluminate are functions of cement sulfate, carbonate and alumina contents. Calcite totally reacts in regions I, II, III and IV, appearing in excess only in regions V and VI, Fig. 7.5. Since the carbonate content of AFm phases is low and the total content of the phases containing alumina (AFm and AFt) is low, not much calcite is required to attain region V; this quantity will be calculated subsequently. Aged hydrated cements are normally undersaturated with respect to gypsum, so the phase assemblage corresponding to region VI is not likely to be encountered in practice (altered cements are excepted from this presentation). The boundary between regions IV and V is shown in bold to indicate its special significance as a dividing line between compositions that are either undersaturated or oversaturated with respect to calcite (see second bullet point page 124). Compositions to the left of this line are undersaturated: that is, calcite behaves as a reactive admixture and is totally consumed, while those compositions to the right of this line will contain a permanent excess of free calcite; in these regions excess calcite behaves as a filler. Of course the amount of any excess will generally be much less than the total added because some calcite will have reacted. The position of the saturation line (bold) enables the numerical saturation value to be determined in terms of sulfate and carbonate ratios for a given cement composition. The figure does not depict the composition of the pore fluid coexist- ing with solids, but one notes that an insignificant proportion of the total solids are dissolved in the

The influence of limestone addition on cement hydration 132

As shown in Fig. 7.6, the amount of ettringite reaches a maximum amount when all sulfate in AFm has been replaced by carbonate, i.e., when monocarboaluminate first appears and hemicarbo- aluminate is maximized at a carbonate ratio equal to about 0.33. Thereafter, upon continued addition of calcite, hemicarboaluminate is progressively replaced by monocarboaluminate: the trend of portlandite consumption is reversed and it is now liberated by calcite additions at higher carbonate ratios in the range between 0.33 and 0.66, approximately. At carbonate ratios at or above 0.66, calcite saturation is achieved and increasing amounts of free calcite appear as the carbonate ratio continues to increase. This calcite effectively dilutes the other solids and their amounts gradually but slowly decrease as calcite amounts increase. However, because of the scale of the Figure, the dilution effect is only just perceptible.

Because of the capacity of the system to react with calcite, carbonate reacts completely at lower

CO2/Al2O3 ratios with aluminate paste components with formation of either or both hemi- and monocarboaluminate. However as the amounts of these phases change, mainly at the expense of monosulfoaluminate, mass has to be conserved and, as a result, the amount of portlandite is also affected as sulfate, progressively displaced from monosulfoaluminate, is increasingly consumed with formation of AFt.

Two noteworthy results of the calculations are (i) to highlight the importance of minor components (sulfate, carbonate) in controlling mineralogical reactions in cement paste, some of which consume portlandite and others of which liberate portlandite and (ii) the role of carbonate in increasing the amount of ettringite in certain composition ranges, despite the total amount of sulfate remaining constant.

10

SO /Al O =0.2 9 3 2 3 SO /Al O =0.4 8 3 2 3

SO 3/Al2O3=0.6 [wt.-%] 7 SO 3/Al2O3=0.8 6 SO /Al O =1.0 3,reactive 3 2 3

5 SO 3/Al2O3=1.2

4 SO 3/Al2O3=1.4 3 2 amount of CaCO 1 0 012345678910

amount of Al2O3 [wt.-%]

Fig. 7.7: Calculated amount of reactive CaCO3 in dependence of the initial amount of initial solid Al2O3 content and bulk SO3/Al2O3 ratio of the paste at ~25°C (data are expressed in weight units; the influence of minor carbonate substitution in AFt is not considered)

The influence of limestone addition on cement hydration 133

Of course not all cements will have a sulfate/alumina ratio equal to 1.0. Therefore Fig. 7.7 and Fig. 7.8 are presented to enable the boundary between reactive and filler calcite to be determined for other ratios; that is, to make the calculation generic. Because the industry tends to work in terms of weight-%, the alumina and sulfate ratios of Fig. 7.7 and Fig. 7.8 are shown in weight units. Fig. 7.7 shows how the maximum reactive wt.-% of calcite varies with amount of alumina and with changing sulfate ratio. It can be seen that for a fixed alumina content, the amount of reactive calcite decreases with increasing amounts of sulfate present expressed as SO3/Al2O3 ratio. Fig. 7.8 shows a similar plot, of reactive calcite shown as functions of the sulfate ratio and amount of sulfate, as sulfur trioxide. This plot makes clear that if the amount of sulfate is kept constant, increasing amounts of alumina (decreasing SO3/Al2O3 ratios), will increase the amount of reacted calcite significantly.

10 SO /Al O =0.2 9 3 2 3

8 SO 3/Al2O3=0.4

[wt.-%] 7 SO /Al O =0.6 6 3 2 3 3,reactive 3,reactive 5 SO /Al O =0.8 4 3 2 3

3 SO 3/Al2O3=1.0 SO /Al O =1.2 2 3 2 3

amount of CaCO SO 3/Al2O3=1.4 1

0 012345

amount of SO3 [wt.-%]

Fig. 7.8: Calculated amount of reactive CaCO3 in dependence of the initial amount of initial solid SO3 content and bulk SO3/Al2O3 ratio of the paste at ~25°C (data are expressed in weight units; the influence of minor carbonate substitution in AFt is not considered)

7.2.5 Space filling by cement paste solids It is instructive to compare the space filling ability of the solids arising in the course of hydration and the chemical bound water demand of the hydration products. As C-S-H is not directly involved in the reactions arising from calcite blending, it need not be considered in the following example calculation. Moreover, while the numbers selected for the calculations will be affected by the exact analysis of the cement as well as the water/cement-ratio, changing the water/solid-ratio affects the calculation only in the sense that the amount of free water is influenced. However the solid phase ratios remain essentially unaffected by the excess water content, up to w/s ~ 1, the calculated trends are of general applicability for all practical formulations. This additional information supplements chapter 7.3 of this Thesis.

The following calculation depicts changes in mineralogical balances expressed as specific volume changes as a function of the carbonate ratio. The calculation of Fig. 7.9 has been done by applying the same conditions as were used to generate Fig. 7.6. Data for the molar volume of phases have been compiled from the literature and are integrated in the database given in Table 4.1.

The influence of limestone addition on cement hydration 135

In commercial cements, where C-S-H develops, C-S-H will dilute the volume change effects arising from calcite reactions but its presence as a diluent does not affect the general conclusions described above. In general, these volume changes are not expected to be physically expansive because much of the reaction takes place before the cement has attained high strength.

7.2.6 Composition of the aqueous phase With the help of thermodynamic modelling the composition of the aqueous phase of relevant AFm- AFt phase assemblages (compare with Fig. 7.6 and Fig. 7.9) can also be calculated. The composition of the aqueous phase of the model system used in the previous paragraphs was calculated as functions of the sulfate and carbonate ratios. Results of a specimen calculation for a fixed sulfate ratio = 1.0 are shown in Fig. 7.10 as a function of the carbonate ratio.

Calcium and hydroxide are the most abundant aqueous species and exhibit constant concentrations independent of the CO2/Al2O3 ratio. The concentrations of aluminium, carbonate and sulfate evolve along stepped profiles. The steps correspond in position to the carbonate ratios of phase boundaries shown in Fig. 7.5 and are numbered accordingly. While the change from phase assemblage I to phase assemblage IV, including the disappearance of monosulfoaluminate and the precipitation of monocarboaluminate, does not significantly change the aqueous composition except for a slight in- crease of carbonate concentrations, the precipitation of calcite leads to a significant increase of the carbonate and sulfate concentrations in conjunction with a decreasing aluminium concentration.

Note that if the sulfate ratios were set to ~1 ≤ SO3/Al2O3 ≤ 3 these sequences remain unchanged, except in position. But if the sulfate ratio would be less than ~1.0, a new step, corresponding to region II of Fig. 7.5, would be generated, while at sulfate ratios < 0.76 another additional step would emerge, corresponding to region III of Fig. 7.5. Thus each region in diagram Fig. 7.5 is related to a distinctive aqueous phase composition - which is a characteristic of a buffered system.

Hc+Ms+AFt Hc+Mc+AFt +CH +CH Mc+AFt+Cc+CH I IV V 40 hydroxide 30

20 calcium

10 0.1 aluminium sulfate 0.01 concentration [mmol/kg] 1E-3 carbonate 1E-4 0.0 0.5 1.0 1.5 2.0 solid molar bulk CO2/Al2O3-ratio

Fig. 7.10: Change of composition of aqueous phases of a hydrated model mixture consisting of C3A, portlandite at a fixed sulfate ratio (SO3/Al2O3=1) in dependence of changing carbonate ratios (CO2/Al2O3) at 25°C (constant total amount of solids, C3A+CaSO4+CH+Cc = 3.25 mol, reacted with 500 g water; compare with Fig. 7.5, Fig. 7.6 and Fig. 7.9 showing solid phase distributions).

The influence of limestone addition on cement hydration 136

While the change in numerical values of pore fluid concentrations appears sometimes so low as to be of no practical importance, this is not so. Each aqueous phase change must be accompanied by a change in the nature of the coexisting solids. Transfer of a significant amount of chemical substance, in this instance mainly of carbonate, is required to complete the phase transformation. This requirement for mass transfer is typical of systems containing solid buffers. The amount of mass required to effect the reaction, and hence the buffering capacity, depends not only on the difference in compositions but also on mass of substance available to participate in reaction. But the principle is established that buffering reactions occur and provide a definite demonstration that cement systems have the ability to buffer aqueous concentrations of species other than hydroxide, OH. In this instance, the buffered species is carbonate but it is apparent that similar buffer systems can be defined for sulfate, e.g. as shown previously for calcium sulfate in the early stages of hydration.

7.3. Applicability to Portland cement systems

7.3.1 Experimental validation of phase changes Seligmann and Greening [169] published one of the first evidences of the formation of mono- and hemicarboaluminate in hydrated Portland cement blended with 5 wt.-% limestone. They also showed that addition of limestone inhibits the transformation of ettringite into monosulfoaluminate for the reasons explained in chapter 7.2.4. Other researchers [37][166] came to similar conclusions which are in agreement with the calculations. Kuzel et al. [113][115][116] provide further detailed experimental validation of the calculations, as he confirmed directly the relations shown in Fig. 7.5 by systematic investigations of hydrated Portland cements containing small amounts of carbonate. Indeed Kuzel and Pöllmann [117] derived similar phase relations as shown in Fig. 7.5 resulting from their interpretation of experiments.

Experimental setup The literature review indicates that the calculations are applicable to Portland cement. Thus no extensive experimental programme deemed necessary to confirm the calculations. However two experiments were done to verify the applicability of the calculations to hydrated iron-free Portland cement paste: i) a carbonate-free cement analogue and ii) a calcite blended cement paste.

To enable a direct comparison of experiments and calculations, and to ensure an initially carbonate- free system, it was decided to prepare a Portland cement analogue by mixing C3S, C3A, CaSO4 and

K2SO4. The batch chemical composition is given in Table 7.2. AFm phases are often poorly crystalline and hard to detect by XRD. Therefore the aluminium content is slightly higher than in common OPC, as this study is focussed on changes with respect to alumina-containing phases, to enable a detection of possible phase changes. The sulfate content was adjusted to a molar

SO3/Al2O3 ratio of ~0.8 which represents SO3/Al2O3 ratios generally encountered in modern

Portland cement compositions. The total alkali content is representative for OPC. K2SO4 was used as alkali sulfate as many cements contain typically more potassium than sodium.

Table 7.2: Chemical composition of experimentally used Portland cement analogue

CaO SiO2 Al2O3 SO3 K2O SO3/Al2O3weight SO3/Al2O3molar wt.-% 68.07 19.33 6.93 4.55 1.12 0.66 0.84

The influence of limestone addition on cement hydration 137

A batch of C3S was prepared from an aqueous slurry of calcite (analytical grade) and amorphous silica (Aerosil 300, Degussa) in a 3:1 molar ratio yielding ~30g C3S. After 15 min homogenisation using an agate mortar, excess water was evaporated by drying at 85°C. The mixture was then placed in a platinum crucible and heated to 950°C in a muffle furnace for decarbonation of the

CaCO3. After 1 hour the temperature was increased to 1450°C for a further 6 hours. The sintered material was cooled down, ground to a fineness < 75 μm and reheated to 1450°C for 6-10 hours. This procedure was repeated five times. Afterwards the material was checked for purity by XRD. No significant impurities, e.g. free lime, CaO, remained.

A mixture of initially 12g C3S, 3g C3A, 1g CaSO4 (anhydrite) was ground in an agate mortar to ≤

75μm. 0.34g K2SO4 were dissolved in 8.2g water (w/s=0.5). Subsequently the alkali sulfate solution was added to the powder and mixed in an agate mortar for ~4 min. Afterwards the paste was divided in small parts of ~1g which were wrapped in polyethylene-film to avoid additional

CO2-contamination. The samples were stored in centrifuge tubes filled with small amounts of water to ensure 100% r.h. during hydration. Subsamples were stored at 25°C, 50°C and 85°C and the temperature-dependent reactions were described in chapter 7.4.

Similarly a limestone blended mix was prepared by using the same procedure but with addition of

2g calcite to the dry mix resulting in a molar CO2/Al2O3 ratio = 1.8. The amount of water was slightly increased to 9.2g to keep the w/s-ratio constant at 0.5. The paste was prepared as described above and subsamples were stored at 25°C, 50°C and 85°C.

Hydration was stopped after 90d by grinding the sample in acetone with subsequent vacuum filtration. In a following step the mineralogy of the dried powder, ground to < 75 μm, was deter- mined by XRD using a BRUKER D8 ADVANCE powder diffractometer as described in chapter 3.2.

Results and discussion Fig. 7.11 shows the XRD-pattern hydrated cement paste analogues in dependence of the initial calcite addition to the mixture. The observed mineralogical changes are similar to the pure C3A pastes hydrated in the presence of sulfates and carbonates (Fig. 7.1 and Fig. 7.2). Thus in the presence of excess calcite, formation of monocarboaluminate and AFt is observed, whereas monosulfoaluminate formed in the calcite-free mixture.

The observed phase assemblage after 90d of hydration is in good agreement with the calculated phase assemblage using the phase relations of Fig. 7.5; for example, in the case of calcite blending at a molar CO2/Al2O3 ratio ~1.8 and a SO3/Al2O3 ratio 0.84, the phase assemblage AFt, monocarboaluminate, calcite and portlandite is predicted and was indeed observed as shown in Fig. 7.11. In contrast to the calculations shown in Fig. 7.5, silica is present in the Portland cement analogue, leading to the formation of C-S-H. C-S-H is of constant composition with respect to its Ca/Si ratio in the presence of portlandite. Divet [54] and Barbarulo [14] reported the sorption of sulfate to C-S-H, which would influence the mass balances calculations and affect the SO3/Al2O3 ratio to be applied in Fig. 7.5. However C-S-H will contain small amounts of aluminium as well as sulfate, as shown by Richardson et al.[160]. Thus the effect of sulfate sorption on the overall

SO3/Al2O3 ratio will be diminished to some extent and, according to own experimental results as well as experiments on hydrated Portland cement published by Kuzel [113][115][116], diagram Fig. 7.5 is still valid to estimate the expected mineral phase assemblage of limestone blended cements although sulfate and aluminium substitution is important and will eventually have to be considered in a quantitative hydration model.

The influence of limestone addition on cement hydration 139

7.3.2 Space filling vs. engineering properties Chapter 7.2 dealt mainly with the fundamental relations arising from limestone addition. In this section an example was chosen to show the direct impact of the previous findings on the engineering properties of limestone blended cements.

Model cement calculations Calculations are based on automated mass balance calculations using the GEMS-PSI routine and the database in Table 4.1. The formation of siliceous hydrogarnet was suppressed in the calculations (see discussion chapter 7.5). From the amount and molar volumes of the phases, the specific volume of solids and aqueous phase are calculated. The calculations use a model cement composition consisting of initially 69 wt.-% CaO, 22 wt.-%

SiO2, 4.5 wt.-% Al2O3 and 4.5 wt.-% CaSO4 (~2.6 wt.-% SO3) corresponding to 83.4 wt.-% C3S,

0.2 wt.-% C2S, ~11.9 wt.-% C3A and 4.5wt.-%CaSO4. 100 g of cement are blended with increasing quantities of calcite (CaCO3) as surrogate of limestone. Thus the initial amounts of several of the cement’s main chemical components (SiO2, Al2O3 and CaSO4) diminish due to dilution. The water/solid-ratio was kept constant at 0.5 by mass (including CaCO3). A hydration degree of 100% and a constant temperature of 25°C were assumed. Alkalis have been omitted from the scheme, as has iron oxide, to minimise the complexity of the system and due to unknowns with respect to a complete thermodynamic dataset for the iron phases. Omitting iron introduces a systematic error but does not necessarily change the principal results of the calculation, as it is believed that much of the iron in Portland cement is present after hydration as ferric oxide - an essentially inert diluent [180].

Results and discussion

The resulting molar SO3/Al2O3 ratio of the model cement is ~0.7. This ratio is broadly representative of modern cements and expresses the qualitative changes occurring upon addition of limestone. If ratios other than 0.7 are selected, the modified calculation takes only a few minutes to complete and enables mineral balances to be quantified.

Fig. 7.12 shows the mineralogical evolution and the specific volume of the solids of the hydrated model cement with increasing calcite contents at 25°C. A comparison with previous calculations for C3A hydration in the presence of calcite and/or sulfate gave similar results (see Fig. 7.6 and Fig. 7.9) except for the presence of C-S-H in the hydrated model cement and the estimation of a chemical shrinkage from the specific molar volumes of the constituents of the raw mix. Since C-S-H is not explicitly involved in the reactions, it behaves as an inert diluent for the other solids. Minor sulfate and aluminium substitution in C-S-H were not considered in this example. However the amount of C-S-H will diminish due to dilution by increasing the calcite content.

A true mass balance calculation, as has been presented here, is required to quantify mineralogical changes associated with the reaction of calcite and the model cement. It is obvious that mineralogical changes influence the solids volume significantly. It is important to note that the changes occurring upon calcite addition to cement do not just involve carbonate-containing phases, e.g. calcite and carboaluminates, but markedly affect the amounts of other solid phases, notably ettringite and portlandite, and of liquid: sulfate, displaced from monosulfoaluminate in the course of its conversion to carboaluminates, is incorporated into ettringite (AFt), a low-density, water-rich phase with high molar volume. This conclusion, demonstrated for one composition, is nevertheless generic for all Portland clinkers containing aluminium.

The influence of limestone addition on cement hydration 142

Thus controlled limestone4 blending shows promise to improve the physical space filling of pastes, reducing free porosity, while also decreasing the clinker content. Furthermore at least one example (Fig. 7.14) has shown that engineering properties, e.g. mechanical strength, can be optimised by limestone blending. Limestone has traditionally been regarded as inert filler and the strength- enhancement effect -if any- attributed to improved packing arising from optimised granulometric properties of the mix. However this study shows that the calcite component of limestone is reactive and that mineralogical changes have a significant impact on the space filling and related physical and mechanical properties of the hydrated matrix. Furthermore, it was shown that although only a small fraction of the total limestone is reactive (≤ ~ 5wt.-% depending on the cement and limestone compositions) the consequences of this chemical reaction, including the displacement of sulfate from AFm and the resulting formation of AFt, affects the performance positively at much higher levels of calcite additions than the original reactive calcite fraction. Besides the mineralogical effects discussed in this work, improved granulometric properties of the paste will contribute to an enhanced performance of limestone blended cements. Thus both physical and mineralogical aspects have to be considered to achieve optimum performance of limestone blended cement.

7.4. The role of temperature on Portland cement hydration

7.4.1 Thermally induced mineralogy changes Phase assemblages at low temperatures ~5°C

Fig. 7.15 shows the stable phase assemblages in part of the system CaO-Al2O3-CaSO4-CaCO3-H2O in dependence of the initial molar sulfate and carbonate ratios at 5°C. Phase assemblages I and IV are similar to the corresponding regions in the diagram for 25°C, Fig. 7.5, except that an increased 2- - substitution of SO4 by 2OH is predicted in the monosulfoaluminate phase. Whereas at 25°C, SO4-

AFm with ~11 mol% SO4 substitution is stable in presence of ettringite, hemicarboaluminate and portlandite (see Fig. 7.5), the maximum sulfate replacement is doubled to ~22 mol% at 5°C. In region II the extent of solid solution formation between monosulfoaluminate and hydroxy-AFm 2- - increased to a maximum substitution of 50 mol% SO4 by 2OH according to the initially fixed miscibility gap boundaries described in chapter 5.3.1. In contrast to the phase relations at 25°C, no

C3AH6 is predicted in region III at 5°C, Fig. 7.15; hydroxy-AFm is computed to be stable, co- existing with hemicarboaluminate, the limiting SO4-OH-AFm solid solution and portlandite at 5°C. A comparison of both diagrams at 25°C and 5°C shows considerable differences in the phase assemblages due to significant changes of the AFt composition in regions IV and V. As illustrated in Fig. B.4, only minor carbonate substitution in AFt is expected at 25°C. However as shown in Fig. 7.15, this substitution increases in importance with decreasing temperature: calculations based on the thermodynamic data in Table 4.1 and the SO4-CO3-AFt solid solution model derived in this study indicate the possibility of significant carbonate substitution in the AFt phase at ~5°C. In dependence of the chemical composition, up to 1/3 of the original sulfate in the AFt phase may be replaced by carbonate at low temperatures under the conditions depicted in Fig. 7.15. On the other hand essentially no carbonate substitution in AFt is predicted in regions I-IV, Fig. 7.15, due to very low carbonate activities. Once the phase boundary between region IV and IVa is crossed, hemicarboaluminate becomes unstable and carbonate substitution in AFt increases due to increasing carbonate activities resulting from increasing carbonate ratios.

4 “Limestone” is not necessarily 100% calcium carbonate and in assessing limestone, since the active fraction is CaCO3, a correction factor may need to be applied, which can be derived from most standard analyses reporting CO2 contents

The influence of limestone addition on cement hydration 144 relationships; thus the calculations are to some extent speculative and more experimental work is needed. Of course reactions depicted here may be complicated by the appearance of thaumasite, especially at lower temperatures. Fig. 7.5 and Fig. 7.15 do not contain silica and other types of calculation are therefore required to show the relationships. Thermodynamic calculations, in the presence of additional C-S-H, have indicated that the composition of the AFt solid solution does not significantly change from that shown in Fig. 7.5 and Fig. 7.15, i.e. no significant silica substitution is expected as long as carboaluminates, portlandite and calcite are present (see composition of point I in Fig. 6.22 and Fig. 6.23 and Table 6.2). However due to a lack of thermodynamic data of silica-substituted ettringite the results are tentative and have to be subject of further investigation. In portlandite saturated systems thaumasite and ettringite-thaumasite solid solutions are calculated to be incompatible with carboaluminate phases. Thus the positions of phase assemblages I-V, in which the formation of carboaluminate phases is predicted, remain essentially unchanged in the presence of C-S-H. But, ettringite-thaumasite solid solutions would become stable in regions Va and VI at higher sulfate activities and influence the existing phase assemblages as shown in Fig. 6.22 and Fig. 6.23.

Phase assemblages at elevated temperatures

The literature review has shown that elevated temperatures significantly alter the mineralogy of hydrated Portland cements. This study concentrated on the changes in the AFm-AFt chemistry with increasing temperature and in dependence of the reactive SO3/Al2O3 and CO2/Al2O3 ratio, respectively. In the first instance mineralogical changes at slightly elevated temperatures occurring in the range up to ~50°C were investigated. A mineralogical map as function of sulfate and carbonate ratios at 40°C, based on thermodynamic calculations using the dataset derived in this study, is shown in Fig. 7.16.

A temperature increase to about 40°C decreased the stability field of hemicarboaluminate significantly. A comparison with the phase assemblages at 25°C showed that the maximum amount of reactive calcite is not influenced (bold line, Fig. 7.16), except that carbonate substitution in AFt becomes negligible with increasing temperatures up to 40°C and the composition of the AFt phase is always close to ideal sulfate-ettringite. But, in contrast to the diagram at 25°C, regions I and IV, Fig. 7.5, disappeared and three new phase assemblages, designated as regions VII, VIII and VIIIa, develop at 40°C (Fig. 7.16). Regions VIII and VIIIa contain only AFm phase assemblages and up to three AFm phases may coexist at relevant sulfate and carbonate ratios in region VIII. Note that AFt is not stable in region VIII while the stability of monosulfoaluminate increases with increasing temperature. The limits of region VIIIa are strongly temperature-dependent. In agreement with Pöllmann et al. [150], the hydroxy-substituted AFm solid solutions become increasingly unstable at elevated temperatures. Thus the phase boundaries between regions III and VIIIa and VII and VIIIa will shift markedly. As a consequence, the compositional extent of region III increases at 40°C relative to 25°C (see Fig. 7.5 for comparison), while the released hydroxide is bound in C3AH6, which coexists with a monosulfoaluminate type solid solution with minor hydroxide-substitution and hemicarboaluminate in region III at 40°C.

The influence of limestone addition on cement hydration 147 form at sulfate ratios ≥ 1; instead a mixture of monosulfoaluminate, AFt, calcite and portlandite coexists in region IX independently of the carbonate content. Calcite will be consumed due to formation of monocarboaluminate in region X; sulfate is bound to monosulfoaluminate but excess

C3A will react with water to form C3AH6. A third conclusion applies to heat cured hydrated cements at temperatures >50°C which will be subsequently placed in a cooler service environment. Due to the reversion of Eq. 7.2, initially formed monosulfoaluminate is predicted to react with excess calcite to form monocarboaluminate and AFt upon cooling to temperatures <<50°C. Thus carbonate addition, or uptake from the service environment, or both, is another factor that needs to be considered to explain phase changes in the course of delayed ettringite formation. The experimental validation of this postulated mechanism of ettringite formation is given in chapter 7.4.2 and is confirmed in the literature. Kuzel and Strohbauch [118] showed the conversion of initially formed monosulfoaluminate to AFt and carboaluminate following exposure of initially heat cured cement at 90°C to carbonate-containing water at 22°C. Fig. 7.18 was developed to complete the overview of reactions occurring at 1bar total pressure up to ~99°C. As shown in chapter 4.4.2, monocarboaluminate is calculated to decompose to C3AH6 and calcite at temperatures > ~85°C. Thus phase assemblages X and XI, Fig. 7.17, become metastable at > 85°C and will be replaced by phase assemblage XII, Fig. 7.18 a), comprising monosulfoaluminate, C3AH6, calcite and portlandite. Finally at temperatures > 92°C (Fig. 7.18 b)) ettringite becomes unstable with respect to monosulfoaluminate and phase assemblage XIII (mono- sulfoaluminate, anhydrite, calcite) which will replace the AFt-containing phase assemblages VI and IX. This temperature is lower than the previously experimentally-derived thermal stability limits for pure ettringite given by Hall et al. [87] and Glasser et al. [195] (~115±5°C). However in both experiments ettringite decomposed to a mixture of monosulfoaluminate and hemihydrate instead of anhydrite. An estimation of the transition temperature using free energy plots, assuming the formation of hemihydrate and monosulfoaluminate from ettringite at 1bar total pressure, is shown in Fig. B.6 (appendix B). It gives 109°C which is in good agreement with the experimental values by Hall et al. [87] and Glasser et al. [195].

a) 86°C - 92°C b) >92°C -99°C excess portlandite present excess portlandite present 3.5 VI 3.5 AFt + anhydrite + calcite 3.0 3.0 XIII -ratio -ratio 3 2.5 3 2.5 O O 2 IX 2 /Al 2.0 /Al 2.0 3 3 monosulfoaluminate +anhydrite + calcite AFt + monosulfoaluminate + calcite 1.5 1.5 1.0 1.0 monosulfoaluminate + monosulfoaluminate + 0.5 XII 0.5 XII molar bulk SO bulk molar C3AH6 + calcite bulkSO molar C3AH6 + calcite 0.0 0.0 0 0.25 0.5 0.75 1 1.25 1.5 0 0.25 0.5 0.75 1 1.25 1.5

molar bulk CO2/Al2O3-ratio molar bulk CO2/Al2O3-ratio

Fig. 7.18: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at 90°C(left) and 99°C (right)

The influence of limestone addition on cement hydration 151

The second stepwise volume decrease is calculated to occur at 50°C. At and above 50°C, monocarboaluminate and AFt become unstable and react to form monosulfoaluminate and calcite according to Eq. 7.2. Due to the release of water and the formation of denser monosulfoaluminate and calcite, the specific volume of solids decreases significantly at ≥ 50°C. In contrast to Fig. 7.19, the amount and specific volume of AFt remains constant at low temperatures, ≤ 28°C, in Fig. 7.21. Calcite is absent under the conditions depicted to calculate Fig. 7.21. Therefore the activity of carbonate is too low to sufficiently stabilise carbonate substituted ettringite at low temperatures. Hence AFt approaches its ideal sulfate-AFt composition under the conditions applied to construct Fig. 7.21.

7.4.3 Experimental verification of thermally induced phase changes The previous chapters have shown that temperature has a major influence on the reactions of the aluminate phase in Portland cement in the presence of carbonate and/or sulfate. Reaction 7.2 and the phase transition AFt + monocarboaluminate ↔ monosulfoaluminate + calcite are especially important to heat cured systems. Until recently, there was insufficient evidence of this reaction and its reversibility in the literature. However, Lothenbach et al. [124] investigated the hydration of calcite blended cements and showed that monosulfoaluminate appeared at the expense of monocarboaluminate and ettringite at ~50°C. Later calculations based on the dataset derived in this study confirmed the experimental observations. Kuzel and Strohbauch [118] realised the influence of carbonate and its role in converting initially-formed monosulfoaluminate in heat cured cements to AFt and monocarboaluminate. However they investigated only the influence of atmospheric carbonation6 on the phase assemblages, by storing the heat cured samples in water in contact with air. Thus leaching and carbonation occurred simultaneously, which can significantly influence the mechanism of ettringite formation. But Kuzel and Strohbauch suggested that the transition temperature between AFt Æ monosulfoaluminate is at ~80+10°C which is well above the value calculated in this study (~50°C). Therefore the following experiments were set up to assess the previous calculations and to show their applicability to commercial cements.

Hydration of C3A The experiment is explained in chapter 7.2.1. Two test cases were evaluated: (i) hydration of a

1:1:2 molar mixture of C3A:CaSO4:CaO and (ii) of a 1:1:1:2 molar mixture of

C3A:CaSO4:CaCO3:CaO at 50°C and 85°C. The water/solid-ratio was kept constant at 5 in all experiments. In contrast to the experiments at 25°C, PTFE bottles were used to store the samples. Freshly boiled ultra pure water previously heated to 50°C and ~85°C was added to the solids, which were subsequently sealed and stored at 50°C and 85°C. Mixing and transfer were done under

N2-atmosphere to prevent CO2 contamination. To assess the kinetic pathways of hydration, mineralogical changes in the samples were monitored qualitatively by XRD at selected intervals over 90d. Sample preparation and setup for XRD measurements are described in sections 7.2.1 and 3.2, respectively.

6 As noted previously, adding CaCO3 contributes calcium and carbonate, while atmospheric carbonation adds CO2 but not calcium. Hence the two types of carbonation reactions differ in detail.

The influence of limestone addition on cement hydration 156

Fig. 7.26 shows the influence of temperature on the mineralogy of a hydrated Portland cement model composition (Table 7.2). As in the C3A samples, monosulfoaluminate and portlandite are the predominant crystalline phases in the XRD-pattern. No significant amounts of remaining clinker phases (C3A or C3S) could be detected which suggests that hydration is close to completion after 90d. Thus, while not detectable by XRD, considerable amounts of C-S-H are present. Siliceous hydrogarnet is absent at 25°C but was observed at 50°C and 85°C. It is interesting to note that the reflections of AFm decreased in intensity with increasing cure temperature. If similar systematic errors apply to all samples and a similar degree of crystallinity is assumed, these observations indicate decreasing amounts of AFm with increasing temperatures, which is consistent with the observation that part of the alumina becomes bound in a hydrogarnet-type phase; it is likely that part of the sulfate is either bound in hydrogarnet as suggested by Paul and Glasser [144] or sorbed by C-S-H, or both.

The number of time-dependent XRD measurements was relatively few, so it is not appropriate to give a graphical representation as was done for the C3A experiments, but the XRD-patterns, Fig. B.12 - Fig. B.14, give a qualitative view of the hydration kinetics at the relevant temperature. As expected, comparison of the XRD-patterns shows that increasing temperatures accelerate the hydration significantly. While reaction proceeds very slowly at 25°C and considerable amounts of

C3A and C3S were still present after 28d, at both 50°C and 85°C, most clinker reacted after 7d of hydration.

Ettringite was present until 28d of hydration, before it was converted due to reaction with the remaining C3A to monosulfoaluminate after 90d at 25°C. At 50°C, ettringite was found at relatively early stages of hydration in the 1 and 2d samples but was fully converted to monosulfoaluminate after 7d of hydration. Ettringite was absent at 85°C; monosulfoaluminate was detected instead after 24h of hydration. However the intensities of the monosulfoaluminate peaks in the XRD-pattern of the 85°C samples were always lower compared to the 50°C samples.

Siliceous hydrogarnet was invariably absent in the hydrated Portland cement analogue at 25°C. Traces of siliceous hydrogarnet were found in some samples at 50°C but at 85°C siliceous hydrogarnet forms progressively between 2d and 90d of hydration, as shown in Fig. B.14. This is again in agreement with the comparatively low intensity of the reflections of monosulfoaluminate at 85°C.

In a second set of experiments the influence of calcite on the hydration of Portland cement paste at

25, 50 and 85°C was investigated. The CO2/Al2O3 ratio was set to 1.8. Thus, as demonstrated by previous calculations (see chapter 7.2.2 and 7.4.1) excess calcite should be present at all hydration times. The related XRD-patterns after 90d of hydration are shown in Fig. 7.27. The results are similar to the investigations of C3A in the presence of sulfate and calcite (Fig. 7.24): ettringite, monocarboaluminate and portlandite dominate the pattern at 25°C, while monosulfoaluminate, portlandite and calcite are predominant at 50°C and 85°C. In agreement with the carbonate-free sample (Fig. 7.26), weak reflections of siliceous hydrogarnet are found at 85°C, but are absent at 25°C and 50°C. The intensities of the reflections of monosulfoaluminate at 85°C are lower than at 50°C.

The influence of limestone addition on cement hydration 158 monocarboaluminate and AFt are the predominant aluminate phases in calcite blended cements at 25°C, which is in accordance with numerous literature studies, e.g. [113]. But, in agreement with computations, monosulfoaluminate becomes predominant at curing temperatures ≥ 50°C. Thus under conditions likely to occur in commercial cement pastes (SO3/Al2O3,molar ≤ 1) hydrated in warm conditions, ettringite and monocarboaluminate will be absent at temperatures ≥ 50°C. On the other hand, previous calculations have shown that monosulfoaluminate and carboaluminates are metastable with respect to siliceous hydrogarnet if excess C-S-H is present. Experiments have shown that the formation of siliceous hydrogarnet must be kinetically restrained at ≤ 50°C. However at 85°C weak reflections of siliceous hydrogarnet were found in all samples. The experiments indicate that siliceous hydrogarnet forms slowly at the expense of monosulfoaluminate at 85°C (see discussion, chapter 7.5.1).

Phase transformations upon cooling of heat cured paste

Calculation and experiments show that elevated temperatures cause significant changes to the mineralogy of calcite blended cements. Heat curing is a common technique to accelerate the hydration of precast concrete elements. However problems have been reported due to destructive delayed ettringite formation (DEF) in samples cured above ~70°C. These observations have been subject of numerous investigations; e.g. Taylor et al. [182], give an excellent overview about possible reaction mechanisms causing destructive DEF. Because most of the reported experience relates to nominally “carbonate-free” cements, the impact of carbonate on DEF has not been assessed. But to evaluate possible changes in the paste mineralogy arising from calcite additions of initially heat cured cements subsequently placed at lower service temperatures, the following experiments were done:

The same experimental protocol applies as for previous investigations using the cement analogues (see chapter 7.3.1). However in contrast to previous investigations, selected samples of the heat cured, carbonate-free and calcite-containing cement analogues were removed after 24h storage at 85°C and subsequently stored at 5°C. Phase changes were monitored by XRD at selected time intervals.

The non-isothermal treatment implies a complex reaction mechanism could occur. To assess the experimental results one has to consider that: i) Hydration is incomplete during the brief heat treatment; anhydrous solids e.g. clinker are still present and continue to react at 5°C ii) The applied temperature change can cause a reconstitution of the initial phase assemblage due to reactions amongst the initially formed phases and/or possible interactions with the unreacted anhydrous phases.

Fig. 7.28 shows the XRD-pattern of the hydrated carbonate-free cement analogue composition after 24h of curing at 85°C and subsequent 180d storage at 5°C. Monosulfoaluminate is observed as the only sulfoaluminate phase in both patterns, beside portlandite and traces of clinker. After 24h at

85°C monosulfoaluminate formed, but unreacted C3S and C3A were still present. Anhydrite was absent. No AFt formation was observed upon prolonged ageing for 180d at 5°C. Indeed the intensities of the monosulfoaluminate reflections increase slightly, which is consistent with the disappearance of the C3A reflections after 180d at 5°C.

The influence of limestone addition on cement hydration 160

The experiments show that hydrated cements rapidly adjust their mineralogy as a function of their initial chemical composition and service temperature. Even allowing for only partial hydration in the initial 24h heat treatment at 85°C, the observed mineralogical changes agree generally well with the previously predicted phase relations. The approach to 5°C equilibrium demonstrates that most temperature-dependent mineralogy changes are reversible and occur within days or weeks.

7.4.4 Summary To summarise the findings of experiments and calculations, it was shown that temperature- dependent phase transformations significantly influence the mineralogy of calcite blended cements. Focussed experiments generally agree with the calculations and experiments reported in the literature [118][124]: monosulfoaluminate becomes increasingly stable while carboaluminates become unstable at temperatures ≥50°C. It was shown that carbonate apparently has an important influence on the conversion process of monosulfoaluminate to ettringite in initially heat cured cements, which are subsequently exposed to low temperatures <<50°C. Thus warm cure results in precipitation of monosulfoaluminate and calcite at elevated temperatures ≥50°C but with resorption of calcite and formation of carboaluminates and AFt in the course of cooling, or at some time thereafter. As very small amounts of carbonate are sufficient to trigger these reactions, it is believed that they are of importance to real cementitious systems even in nominally “carbonate- free” cements. While several investigations showed the importance of factors like sorption of sulfate to C-S-H in dependence of the alkali content, maximum amounts of ettringite in dependence of a pessimum SO3/Al2O3 ratio, etc. (see [182]), it is apparent that these factors may not be the only reasons which need to be considered to explain DEF. The lower temperature range is also important; predictions of the stable phases (amount, composition) differ depending on the choice of lower temperature, 5°C or 25°C.

The influence of temperature on the related specific volume distribution of the solids was calculated. It was found that high temperatures lead generally to a decrease of the total specific volume of the calcium-aluminate phases present in calcite blended cementitious systems mainly by conversion of bound water to free water. The transition temperatures are affected by unavoidable analytical errors in the process of the database development. Therefore it is suggested that the transition temperatures given be regarded as provisional values, at which temperature the relevant reaction becomes likely, not as numerical values which are absolutely fixed. Furthermore the calculations indicate that in the presence of calcite, extensive carbonate substitution in AFt is likely at low temperatures. As a consequence the amount of AFt increases and leads to an increase of the total specific volume of the system. The calculations indicate that this process applies mainly to cements cured at or exposed to temperatures ≤ 10°C. So far the calculated reactions are not supported by experiments. Thus more experimental evidence is desirable, particularly as the microstructural changes arising as a consequence of reaction cannot be predicted.

The influence of limestone addition on cement hydration 161

7.5. Discussion

7.5.1 Limitations of the methodology and of the database The influence of carbonate, sulfate and aluminium and of the resulting sulfate and carbonate ratios on the phase assemblages of hydrated Portland cements, have been extensively investigated by calculations as well as by focussed experiments. They are realistic in terms of (i) assuming an excess of Ca(OH)2 to be present and allowing portlandite to react as required and (ii) allowing phase relations to be depicted as functions of SO3/Al2O3 and CO2/Al2O3 ratios of the reactive fractions. Note that the definition of the “reactive fraction” is subject to uncertainties: corrections, some of which involves the fraction of clinker minerals reacted at a given time, and hence the availability of alumina, carbonate, etc., for reaction as well, as corrections for sorbed species, such as sulfate in C-S-H (see chapter 7.5.2). The previous calculations were initially done for the system

CaO-Al2O3-CaSO4-CaCO3. However in commercial cementitious systems at least three more components have to be taken into account i) silica, ii) iron oxide and iii) alkalis. Nevertheless own experiments reported here and practical experiences by Kuzel et al. [113][115][116] have shown that the results can be applied to hydrated Portland cements.

Formation of siliceous hydrogarnet While silica is mainly bound to C-S-H, which is inert with respect to reactions with calcite, and of constant composition, except for minor sulfate and aluminium substitution, as discussed above, the possibility of siliceous hydrogarnet formation and related phase assemblages has to be considered. Fig. 7.30 shows the predicted phase assemblages as a function of sulfate and carbonate ratios, analogous to previous calculations, but the formation of siliceous hydrogarnet has not been suppressed in the presence of C-S-H and portlandite. Fig. 7.30 a) shows the phase assemblages relevant to lower temperatures ≤ ~35°C, while Fig. 7.30 b) is constructed for temperatures between 35°C to ~65°C. As the main focus is on the formation of siliceous hydrogarnet, AFt and thaumasite were regarded as single phases in the calculations to avoid complex phase relations including solid solution formation. According to the calculations in the presence of a silica source, e.g. C-S-H, monosulfoaluminate and both carboaluminates (hemi- and mono-) are generally metastable with respect to siliceous hydrogarnet, Ca3Al2(SiO4)0 8(OH)8 8 - one member of the Ca3Al2(SiO4)3-x(OH)4x solid solution series which is currently implemented in the database. Siliceous hydrogarnet is predicted to be stable over a wide range of SO3/Al2O3 and CO2/Al2O3 ratios at various temperatures, as shown in Fig. 7.30. At low temperatures, up to 35°C (Fig. 7.30 a)) siliceous hydrogarnet would be accompanied by the formation of ettringite and/or thaumasite and ettringite, or solid solutions between both, in regions 1, 2 and 3. But, thaumasite becomes increasingly unstable at temperatures > 35°C (Fig. 7.30 b)). Thus phase assemblage 5 (AFt, siliceous hydrogarnet, calcite and C-S-H) dominates the cement relevant range of SO3/Al2O3 and CO2/Al2O3 ratios. At temperatures >~65°C (Fig. 7.30 c)) AFt, thaumasite and gypsum are not stable and phase assemblages 1, 4 and 5 are instead replaced by assemblage 6 comprising a mixture of siliceous hydrogarnet, anhydrite, calcite, portlandite and C-S-H, as appropriate.

The influence of limestone addition on cement hydration 162

a) ~1-35°C b)~35-65°C

excess portlandite and C-S-H (Ca/Si~1.6) present excess portlandite and C-S-H (Ca/Si~1.6) present 3 5 AFt+Th. +gypsum 3 5 AFt+Th +gypsum 1 1 4 AFt+Th +calcite 3 0 3 0 2 -ratio -ratio 3 2 5 3 2 5 O O 2 AFt+Thaumasite + 2 /Al 2 0 /Al 2 0 3 sil. hydrogarnet (C3AS0.8H4.4) 3 5 1 5 1 5 AFt+ sil. hydrogarnet (C3AS0.8H4.4)+ 1 0 3 1 0 +calcite molar bumolar k SO 0 5 Thaumasite +sil. hydrogarnet SO bulk molar 0 5 (C3AS0.8H4.4)+ calcite 0 0 0 0 0 025 05 075 1 125 15 0 025 05 075 1 125 15

molar bulk CO2/Al2O3-ratio molar bulk CO2/Al2O3-ratio c) ~66 - 99°C

excess portlandite and C-S-H (Ca/Si~1.6) present 3 5

3 0

-ratio 6

3 2 5 O 2 sil. hydrogarnet (C3AS0.8H4.4)+

/Al 2 0 3 +calcite 1 5 +anhydrite

1 0

molar bulk SO 0 5

0 0 0 025 05 075 1 125 15

molar bulk CO2/Al 2O3-ratio

Fig. 7.30: Calculated phase assemblages of a hydrated mixture consisting of C3A, C-S-H (Ca/Si~1.6), portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at different temperatures. Numbers of the assemblages are used in the text.

Siliceous hydrogarnet is apparently very stable over a wide range of sulfate and carbonate activities. However siliceous hydrogarnet is only rarely reported to occur in commercial cements, whereas carbo- and monosulfoaluminate are often described as hydration product of Portland cement at ambient temperatures, in agreement with own investigations. For example Gebauer and Harnik [74] observed the formation of monocarboaluminate in 84 year old hydrated concrete, showing it to be persistent, although no information was available to determine when it was initially formed. But the introduced metastability constraint with respect to siliceous hydrogarnet in the calculations presented in chapter 7 seems to be justified as it reproduces the actual hydration behaviour of Portland cements, as recorded from field observations.

The reason why siliceous hydrogarnet does not form is not yet clear. One possibility is that the thermodynamic data are not correct and overestimate the stability of siliceous hydrogarnet. Until recently there has not been a consistent dataset available which describes the solid solution properties of Ca3Al2(SiO4)3-x(OH)4x. Miscibility gaps in the solid solution series between C3AH6 and C3ASH4 siliceous hydrogarnet, as observed by Jappy et al. [95], could indicate non-ideal mixing behaviour. The limits of the miscibility gap could not be fixed exactly and related solubility data necessary to estimate the thermodynamic properties of this apparently non-ideal solid solution

The influence of limestone addition on cement hydration 164

The experiments suggest that siliceous hydrogarnet does form at the expense of monosulfoaluminate. However this process of conversion of monosulfoaluminate to hydrogarnet is relatively slow, as observed in experimental studies. If one assumes a partly decomposition of initially formed monosulfoaluminate to siliceous hydrogarnet two possible scenarios arise: i) Siliceous hydrogarnet is the thermodynamically favoured phase, as indicated by calculations, but its formation is kinetically restrained at low temperatures ≤ 50°C; monosulfoaluminate will decompose to hydrogarnet at elevated temperatures and the released sulfate is either bound in the hydrogarnet phase, as suggested by Paul and Glasser [144] or sorbed on C-S-H, or both. ii) The amount of monosulfoaluminate is affected by a strong increase of the sulfate sorption to C-S-H at high temperatures. Thus sulfate is increasingly removed from the pore solution and monosulfoaluminate has to dissolve to maintain saturation. Due to the sorption of sulfate on C-S-H an excess of calcium and aluminate arises and the resulting supersaturation with respect to siliceous hydrogarnet reaches a critical point which, once exceeded, nucleates and crystallises hydrogarnet from the pore solution. The presence of alkalis promotes this process as increasing amounts of alkalis lead to a strong increase of the sulfate solubility in the presence of ettringite and monosulfoaluminate as shown in Fig. B.7. Therefore the sorption of sulfate on C-S-H increases strongly with increasing temperature and alkali content.

Role of iron and alkalis

Ferrite (Ca2AlxFe1-xO5) is one of the four main components of Portland cement clinker. However compared to aluminate, C3A, the number of literature studies on the hydration of the ferrite solid solution type phase is limited. Although the kinetics of ferrite hydration are generally considered to be slow compared to aluminate, experiments by Seligmann and Greening [169] suggests that iron can be treated analogous to aluminium. As a consequence, solid solutions between aluminium and iron bearing AFt and AFm may form as ferrite hydrates. From a literature review, including an assessment of Mőssbauer spectra, Seligmann and Greening [169] suggested that most of the iron becomes bound into a SO4-AFm type phase. On the other hand a review by Taylor [180] led to the conclusion that the formation of iron hydroxides, amongst other phases, is likely.

In addition to the database derived in this Thesis, the derivation of thermodynamic data on iron- containing AFt and AFm phases is in progress. Preliminary calculations, based on solubility experiments by Möschner et al. [132][133], in agreement with the conclusions by Seligmann and Greening [169], indicate that iron stabilises the AFm phase whereas significant iron substitution in the AFt phase does not occur. This would imply a possible increase of the amounts of carboaluminate in connection with an increase of the reactive calcite content. However, although a complete dataset for Fe-ettringite and provisional values for monocarboferrite and monosulfoferrite hydrate were derived by Möschner et al. [132][133], a complete dataset for iron AFm and hydrogarnet phases including solid solutions with the Al-analogues was not available and experimental validation for Portland cement systems containing both iron and aluminium has not yet been undertaken. Due to these uncertainties an exact assessment of the state of iron in hydrated Portland cement paste is not possible at the moment.

The influence of limestone addition on cement hydration 165

Table 7.3: Influence of alkalis on the composition of aqueous and solid phases at selected invariant points at 25°C (see Fig. 7.4 for original phase diagram)

Pt. pH Na Ca Al CO3 SO4 Solids [Fig. 7.4] [mmol/kg] [mmol/kg] [mmol/kg] [mmol/kg] [mmol/kg]

M 12.48 0 19.35 0.086 0 0.001 C3AH6, Ms(ss, 24%OH), CH 13.58 500 0.58 1.3 0 0.350 C AH , Ms(ss, 22%OH), CH 3 6

H 12.48 0 19.37 0.007 0.0065 0.024 AFt(ss, 9%CO3), Mc, calcite, CH

13.58 500 0.61 0.11 0.86 6.98 AFt(ss, 8%CO3), Mc, calcite, CH

Alkalis in Portland cement, potassium and sodium, are partly present as alkali sulfates and partly substituted in the clinker phases, mainly in the C3A-type phase of Portland cement. Alkali sulfates are readily soluble and influence the composition of the pore solution of hydrated cements significantly. As part of the sulfate is removed in the course of hydration, due to formation of - ettringite and SO4-AFm, the pH of the pore solution rises quickly to values higher than 13 as OH is released to charge balance the positive alkali cations. Thus the hydroxide activity increases markedly. A recalculation of the aqueous and solid phase composition of the invariant points M and H (Fig. 7.4) in the presence of 500 mmol Na+/kg (Table 7.3) shows that sulfate, carbonate and aluminium concentrations and related corrected activities (not shown here) increase significantly while calcium concentrations decrease. However, despite these complex changes in the aqueous phase, the solid composition is not significantly affected. Hence the effects of increasing aqueous hydroxide or carbonate activities are balanced to some extent by the increasing sulfate activities. In hydrated cements made to low w/c-ratios the mass of excess pore solution is small compared to the mass of solids. Thus despite changes in the solubility of the solid phases only minor changes of the total solid mass balance result from the presence of alkalis: hence the phase relations shown in diagrams Fig. 7.5 and Figs. 7.15-7.18 are not significantly affected.

As shown in Table 7.3, the presence of high amounts of alkalis is accompanied by a significant increase in the aqueous sulfate, aluminium and carbonate concentrations. However, C-S-H is known to bind significant amounts of alkalis, sulfate and aluminium, dependent on the aqueous concentration of the related species. Thus increasing alkali contents can lead to an increased incorporation of these elements in C-S-H, either by surface processes e.g. adsorption or by substitution of ions in the bulk. Thus the amounts of sulfate and alumina, available to form AFm and AFt decrease and can result in significant mass balance changes (see chapter 7.5.2).

The influence of limestone addition on cement hydration 166

7.5.2 Kinetic factors: availability of sulfate, carbonate and alumina in the course of hydration

The foregoing calculations are based on the simplistic assumption that all chemical components of the cement, as revealed by a bulk chemical analysis, are available for reaction. This is probably not true for real cements. Clinker grains have a range of sizes and complex internal microstructures, with the result that different clinker phases hydrate at different rates and may be partly protected from hydration, perhaps by occlusion in hydrate envelopes with low permeability. Thus, for example, alumina in tricalcium aluminate may become available at an earlier stage of hydration than alumina in ferrite. Kuzel [115] suggested using only the C3A-content to estimate the available alumina content during hydration. Seligmann and Greening [169] showed a case where ferrite remained unreactive in hydrated cement paste (20% C3S, 51% C2S, 3.4% C3A , 15.2% C4AF and

1.98% SO3, at w/c = 0.5) for more than 14 years, which impacted on the effective SO3/Al2O3 ratio and the related hydration products of the cement. However, it is likely that the reactivity of ferrite is strongly influenced by minor element concentration e.g. Mg, Ti, as well as by its A/F ratio. The role of iron in the distribution of hydrate phase assemblages is discussed further in chapter 7.5.1, but at present, it is not possible to estimate ferrite reactivity, or to assume that it will remain constant from one clinker to another.

Sulfate, nominally in the form of added calcium sulfate and alkali sulfates, is typically available for reaction at an early stage of hydration, although it is not unknown for sulfates to be physically occluded within clinker grains only becoming available for hydration at longer ages [110]. Cements contain carbonate from many sources some of which are not obvious. For example, during clinker cooling in contact with kiln gases, sulfates may condense onto the clinker. Sulfur oxide is swept through the kiln together with alkalis, water, carbon dioxide, etc. As the hot gas stream cools, it may reach the dew point of alkali sulfates. Calculations on vapor - melt equilibria by Barry and

Glasser [18], assuming a realistic partial CO2 pressure in the kiln, disclose that these alkali-sulfates are actually alkali sulfate-alkali carbonate solid solutions. Thus vapour condenses onto cooling clinker, introducing significant carbonate as well as sulfate. Carbonate contributions may also arise in the same manner if clinker kiln dust is recovered and added to clinker. Moreover “gypsum” interground with cement often contains significant calcite. These several sources of carbonate are often sufficient to stabilise hemicarboaluminate in the course of hydration, even of fresh and nominally uncarbonated “limestone free” Portland cement.

Thus key components, including alumina, sulfate and carbonate are present in clinker in a range of forms and, while these range in dissolution rates, with few exceptions all sulfate becomes available early within the hydration process. Typically most of the gypsum interground with clinker is consumed within the first few days of hydration whereas some of the alumina, mainly in ferrite, becomes available only over much longer time scales. Therefore detailed and realistic calculations would have to make use of the reactive fraction of these phases: the “composition” that determines the hydrate mineralogy is not necessarily the bulk composition as determined from a total chemical analysis but is instead the sum of the reactive fractions of each component liberated from the clinker solid phases. To quantify these amounts requires a series of “snapshots” on the state or condition of the system with changing time. While the relevant hydration kinetics are not sufficiently well known- and indeed almost certainly vary with non-chemical factors such as clinker granulometry and microstructure of the individual grains- it is known in general that sulfate ratios tend to be ≥ 3 at the outset of hydration owing to readily soluble alkali sulfates and calcium

The influence of limestone addition on cement hydration 167 sulfates, i.e. gypsum and/or anhydrite. Thus, in agreement with previous calculations supported by focussed experiments it was shown that ettringite is the initially formed hydrate phase in cement systems at early ages while still saturated with respect to calcium sulfates at 25°C. AFm phases including (monosulfoaluminate type solid solutions and carboaluminates) are not stable under these conditions (SO3/Al2O3 ≥ 3) and are not observed if the system is saturated with respect to gypsum. However due to the consumption of calcium sulfate, the sulfate ratio decreases to ≤ 3 as hydration progresses. As a consequence monosulfoaluminate and/or carboaluminate tend to become stable as a function of changing carbonate and sulfate ratios; the influence of temperatures other than 25°C is discussed in chapter 7.5.4. Fig. 7.5 shows the phase relations between AFt-AFm phase assemblages in dependence of the sulfate and the carbonate ratio of the system. Thus as hydration approaches completion, ettringite and calcite will partly or fully react with the remaining alumina to form monosulfoaluminate and hemicarboaluminate in regions I-III (Fig. 7.5). However owing to the comparatively higher thermodynamic stability of the carboaluminates, the remaining aluminate tends to react preferably with calcite rather than ettringite. Thus at sulfate and carbonate ratios relevant to regions IV and V (Fig. 7.5), the conversion of ettringite to monosulfoaluminate is suppressed by the formation of carboaluminates in the progress of hydration.

As the clinker silicates react, large amounts of C-S-H dominate the mineralogy of hydrated cements. Owing to the high specific surface of C-S-H, significant amounts of sulfate can be taken up by C-S-H; the sorption depends on the Ca/Si and Al/Si ratio of the C-S-H, the sulfate content in the pore solution and the reaction temperature [14][54]. As shown by Richardson et al. [160], significant amounts of aluminium can be substituted in C-S-H. Thus, while the phase boundaries shown in Fig. 7.5 are broadly unaffected by sulfate and aluminium uptake of C-S-H, the amount of available sulfate and aluminium may be significantly lower than are disclosed by the total chemical analysis, which will affect subsequent mass balance calculations. A correction for sulfate and alu- minium uptake by C-S-H will be useful to improve the accuracy of the mass balance computations.

It is not known how carbonate ratios are affected by the kinetics of hydration and an important task for those who determine reaction kinetics is to quantify how reactive calcite fractions change as a function of time, as well as with changing bulk composition and formulation. The intermediate formation of hemicarboaluminate in systems with excess calcite present has been shown in own experiments as well as by Kuzel and Pöllmann [117], while calculations in this Thesis, in agreement with calculations by Damidot et al. [49], have shown that hemicarboaluminate is not stable in the presence of excess calcite. The precipitation of carboaluminates from a supersaturated solution involves the removal of carbonate ions from the solution. Apparently due to a disequilibrium between the kinetics of carboaluminate precipitation and a slower dissolution of calcite, the resulting carbonate activities in the pore solutions decrease to levels such that monocarboaluminate may become unstable and hemicarboaluminate is formed in the presence of portlandite as shown in Fig. 7.4. Data obtained in the course of this study also apply in principle to the reactivity of coarse limestone aggregate. In theory, limestone aggregate is predicted to be reactive with cement in the same way as is limestone with high specific surface. The same conditions apply to aggregate except that for coarse aggregate, the kinetics of reaction with cement are likely to be appreciably slower: not only is the surface area available for reaction much reduced but diffusion paths for other species, e.g., aluminium, tend to lengthen. When clinker is interground with limestone, the many observations in the literature show rapid formation of monocarboaluminate in hydrating pastes within the first days

The influence of limestone addition on cement hydration 168 or weeks of hydration. It is likely that the potential of coarse limestone aggregate to react with cement can be reduced or even eliminated by saturating the cement matrix in carbonate, as occurs when sufficient fine-grained reactive calcite is included in the formulation. Thus if the matrix is calcite saturated, the potential for subsequent reaction of cement with limestone aggregate can be reduced to zero, with consequences for the development of bond between aggregate and paste and spacefilling in the interstitial transition zone, i.e., at the interface between paste and aggregate.

Carbonation from external sources is not included in this study and will always follow a slightly different reaction path, because a mass unit of CO2 in the form of CaCO3 also adds a unit of

“CaO”; this equivalence does not however apply to atmospheric CO2. But the differences at low carbonate contents are minor and cements undergoing atmospheric carbonation develop carbonate gradients and may also experience the sequence of mineralogical changes described here. Spatially these zones are most likely to develop ahead of the main carbonation front. It is also noteworthy that adding carbonate in the form of calcium carbonate to cement slightly reduces its buffering capacity towards subsequent carbonation, for example in atmospheric carbonation. If protection of embedded steel is important, it is desirable that improvements in cement performance, for example reduced permeability arising from calcite addition, should outweigh loss of buffering capacity.

In summary, the foregoing calculations assume that the bulk chemistry is relevant but, almost certainly, kinetic controls operate somewhat to modify the picture presented here. Typically the effective sulfate ratio will be somewhat higher during early hydration than is assumed in the calculations. Reference to the figures shows that, other factors remaining unchanged, increasing the sulfate ratio decreases the potential for consuming calcite. Thus calcite will initially react in response to the effective sulfate ratio, leaving potential for slower secondary reactions to occur. These slow reactions will result in continuing calcite consumption with time as clinker hydration approaches completion.

7.5.3 Volume changes

Volume changes were calculated by attending calcite additions in the expectation that these calculations, or very similar calculations, will eventually be used to develop property-composition correlations for hardened cement pastes. The calculations implicitly assume that the volume changes attending reaction of calcite with cement components occur after paste has hardened but do not lead to changes in external dimensions of a paste monolith, i.e., they are not expansive. If this supposition is correct, calcite addition can be used creatively, to improve space filling by solid hydration products and thereby lower product porosity. Because the relationships between porosity and permeability are complex, lowering porosity will not necessarily reduce permeability and indeed, no consistent pattern emerges from a study of the literature. Nevertheless the space filling property-composition correlations developed here, as confirmed by at least one experimental example [90], should give fresh direction and focus to these studies leading, for example to concentration on the impact of additions in critical concentration ranges. The critical concentration may be low, a few wt.-% in the case of calcite. But each case depends on the cement composition and requires separate calculation.

The influence of limestone addition on cement hydration 169

7.5.4 Thermally induced phase changes Hydrated cements experience a wide range of service temperatures and may encounter strong self- heating early in their life cycle, as heat of hydration is not always readily dissipated. This study has shown that the temperature dependence of the thermodynamic properties of AFm-AFt phases impact significantly on phase development. It was shown that carboaluminates become increasingly unstable at elevated temperatures, whereas monosulfoaluminate, although stable at 25°C, is stabilised by increasing temperatures up to at least 99°C. Calcite is shown to be a reactive component and forms carboaluminates at temperatures ≤ 50°C, but increasingly behaves as inert filler at higher temperatures. At sulfate ratios ≤1 AFt is stabilised due to the formation of carbo- aluminates at temperatures ≤ 50°C. However, prolonged curing at elevated temperatures, ≥ 50°C, leads to the formation of monosulfoaluminate at SO3/Al2O3 ratios typical of Portland cements, ~ 1 or less, independently from the calcite content. Thus a subsequent lowering of the service temperature induces an extra potential for ettringite formation in calcite or carbonate-containing cements. Ettringite will form in increased amount due to partial displacement of sulfate by carbo- nate in the AFm structure at ambient temperatures ~20°C - 25°C. This process will increase the specific solid volume significantly. It should be noted that the approach used in this work is only able to predict phase changes and associated mass balance and specific volume changes, but not necessarily actual expansion. While phase changes are often a precondition of destructive processes, e.g. DEF, the resulting volume changes cannot be used directly to calculate %expansivity. To assess the implications of phase changes on the development of mechanical stress the thermodynamics of crystal growth, including the evolution of crystal pressures in dependence of the temperature changes, has to be taken into account as described by Scherer [167] and Flatt and Scherer [64].

Much interest centres on the role of heat curing in concrete. With the “discovery” that excessive heat cure results in subsequent re-formation of ettringite (DEF), the relations between upper cure temperature and duration of the warm cure period and of the thermal cycle to factors like cement composition and fineness have created a web of complexity, which has proven almost impossible to resolve into an action plan. The present calculations enable a method to elaborate a phase development model, independently from long-term experiments. Of course some relevant factors, e.g. cement fineness and the degree of hydration of the cement in the elevated temperature curve are either not included or have to be estimated in order to apply the method. But even so, there are a range of other inputs still needed. For example, in studies of DEF described in the literature, it is often not stated whether limestone (calcite) has been added to the cement. However the present study discloses that the calcite content has an important bearing on the phase development in the course of thermal cycling. Moreover, while many literature studies of DEF test different maximum temperatures it is usually assumed that the expansive potential can be measured at 25°C. The present study shows that this assumption is untentable: in a thermal cycle with an arbitrarily set maximum temperature, e.g. 60°C, the amount of mineralogical reconstitution needed to reequilibrate a cement paste at 10°C differs significantly from that at 25°C, e.g. due to the possible formation of SO4-CO3-AFt solid solutions. It is therefore concluded that the potential for DEF depends on both maximum cure temperature and on the service temperature.

The influence of limestone addition on cement hydration 170

The divergence in literature studies on the causes and avoidance of DEF is perhaps not surprising and quantitative modelling of DEF is as yet not possible. It is not claimed that this study solves the problems, but it provides an analytical tool to anticipate changes, to harmonise experimental observations and to design simplified experiments, which, although simplified, will unambiguously illustrate selected aspects. Some examples are given in the subsequent paragraph.

The reaction of monosulfoaluminate, calcite and water to form AFt and carboaluminates at temperatures <50°C can be an important factor, amongst others, to explain expansive processes in the course of DEF. For example, Kuzel and Strohbauch [118] have shown that initially formed monosulfoaluminate in heat cured samples subsequently converts to AFt and monocarboaluminate upon subsequent water storage in contact with atmospheric CO2 at 22°C. Furthermore self heating might occur in mass concretes. If the heat of hydration is not released quickly, prolonged cure at ≥ 50°C can result in a formation of monosulfoaluminate, which will be converted to carboaluminate and AFt, if the service temperature is << 50°C and if sufficient carbonate and water are available. The local availability of water can be another potentially rate limiting factor. Depending on conditions in warm cure, water liberated by phase transformations may be lost, i.e. the system is not isochemical. So the reverse reactions encountered in the post-warm cure phase may also be kinetically controlled by need to uptake water and its availability. Nevertheless, the calculations present another explanation of the occurrence of monosulfoaluminate at high temperatures and its transformation to ettringite at low temperatures.

The previous calculations have shown that phase changes are not limited to elevated temperatures, e.g. hydroxide substituted AFm phases, e.g. C4AHx, or SO4-OH-AFm solid solutions and hemicarboaluminate, are stabilised at lower temperatures. Furthermore AFt solid solutions, including SO4-CO3-AFt and ettringite-thaumasite solid solutions, are preferably formed at low temperatures. Service temperatures in the northern hemisphere are often less than 10°C. Although direct experimental evidence is absent, calculations indicate that a substitution of carbonate in AFt is favoured at low temperatures. Thus the maximum amount of AFt is not restricted by the cement sulfate content alone. The amounts of ettringite and its specific solid volume increase with decreasing temperature in the presence of calcite, which would significantly influence the space filling properties of the hydrated cement paste. As long as experimental evidence of this reaction is absent, the results should be regarded as tentative, but indicate that more research is needed to explore the influence of low temperatures on cement hydration.

The previous experimental and theoretical studies have shown that they have obvious implications for the dimensional stability, porosity and permeability of concretes in service and reflect on the binder design, formulation and thermal cure regime. Future studies need to link these factors. But it is possible to envisage tailoring cement compositions to optimise performance in specific service environments without dependence on purely empirical results.

Summary and conclusions 171

8. Summary and conclusions

8.1. Thermodynamic quantification of cement hydration

In the past decades applications of thermodynamic methods to cement hydration have long been considered as too theoretical and without any practical use for the cement industry and mineralogical evolution in the course of cement hydration was often regarded as too complex to model. However the development of advanced geochemical codes, e.g. GEMS-PSI, enabled the treatment of complex systems. Although these codes were originally developed to solve geochemical problems, they can readily be applied to other chemical systems, e.g. hydrated cements. But to attain meaningful results it is necessary to couple these geochemical codes with a comprehensive thermodynamic database including all main cement hydrates expected to form in cementitious systems. Available databases for cement hydrates are often incomplete and with few exceptions are limited to isothermal systems at 25°C. Therefore the main goal of this work was to develop a database, capable of dealing with the complexity of modern cement compositions, including reactions occurring over a broad range of temperatures to quantify the mineralogical evolution during cement hydration.

A database was developed for commonly-encountered cement substances including C-S-H, AFm phases -namely monosulfoaluminate, monocarboaluminate, hemicarboaluminate and hydroxy-

AFm, AFt -including SO4-AFt, CO3-AFt and thaumasite- and hydrogarnet compositions, C3AH6 and C3As0 8H4 4. The formation of solid solutions, if observed experimentally, was taken into account by appropriate models. The database is valid in a temperature range of ~1-99°C at 1 bar total pressure. Available literature data were critically assessed and completed with experimental determinations. The final thermodynamic properties were derived by a harmonisation of the available data.

The overall conclusion of this Thesis is that thermodynamics is a useful method to quantify cement hydration, provided an accurate database is available. A big advantage of thermodynamic calculations is its generic character. Thus once the model is setup correctly it can be applied to a wide range of cement compositions and temperatures. Another advantage of thermodynamic methods is that mass is conserved in all calculations and allows a quantification of cement hydration independently from existing detection limits often encountered in experimental methods. Of course as with every other method there are still limitations of the thermodynamic approach used in this study e.g.:

i) database development: there are still missing gaps in the database, e.g. the extent of alkali, sulfate and aluminium substitution and/or adsorption in C-S-H. However work is in progress to account for these processes.

ii) the lack of a kinetic model describing hydration as a function of time. However successful attempts are reported in the literature using a fractional reactivity approach to overcome this problem [123][125].

It is argued that the user of this method has to appreciate the significance of the calculated results. In this respect, the computer is still servant, not master and the need for a thorough understanding of cement science is strengthened, not reduced, by the advent of thermodynamic computations.

Summary and conclusions 172

8.2. Implications for cement hydration

This section summarises the main conclusions arising from thermodynamic calculations and experiments of this work. The main focus was to investigate hydrate phase assemblages, including the distribution of sulfate and carbonate between AFm and AFt phases, and to derive generic conclusions to assess the chemistry and mineralogy of commercially available hydrated Portland cement paste. To highlight the practical importance of this problem the impact of limestone addition, mainly as calcite, on cement hydration was systematically investigated.

The constitution of the AFm phase was assessed by thermodynamic calculations and focussed experiments at 25°C. Experiments and calculations reveal that AFm phases, containing as principal - 2- 2- anions OH , SO4 and CO3 , are incompletely miscible with each other, except for limited replacement of sulfate by hydroxide in monosulfoaluminate. Carbonate and sulfate provide thermodynamic stabilisation for the AFm phase. Thus monocarboaluminate, hemicarboaluminate and part of the monosulfoaluminate-hydroxy-AFm solid solution series (up to 24 mol% sulfate substitution) series are stable at 25°C. Therefore it is likely that many hydrated cements will contain mixtures of AFm phases in dependence of their initial chemical composition.

Although the existence of solid solutions based on ettringite has long been known, and syntheses of these solid solutions are reported in the literature, no assessment of the thermodynamic stability of these solid solutions was possible, due to lack of data and of appropriate models. Thus, to assess the formation of AFt solid solutions in cementitious systems, it was necessary to derive their thermodynamic properties. Based on a binary solid solution model and available experimental data, the non-ideal mixing behaviour of SO4-AFt and CO3-AFt, as well as of SO4-AFt and thaumasite were investigated. Subsequent calculations indicate that carbonate substitution in AFt is strongly dependent on temperature. Whereas at 25°C only minor carbonate substitution is likely in ettringite, low temperatures, ≤ 10°C, promote the incorporation of carbonate in the AFt structure.

Pathways of thaumasite formation have been discussed. According to the calculations there is no single route of thaumasite formation as often suggested in the literature. The simultaneous occur- rence of several pathways to form thaumasite is likely. The initial formation of AFt-thaumasite solid solutions is an essential precursor to form thaumasite in calcium-rich zones. In agreement with field observations the amount of AFt is likely to increase significantly before thaumasite first appears. The extent of this increase will be more important at low temperatures. On the other hand, in decalcified regions, e.g. zones undergoing extensive leaching, thaumasite may form directly without preliminary AFt formation. Hence the distinction between “conventional sulfate attack” and “thaumasite form of sulfate attack” seems not to be appropriate as carbonate is widely abundant in most service conditions and the “conventional sulfate attack” including the conversion of AFm to AFt is likely to be a precursor of thaumasite formation.

EN 197 specifications permit addition of up to 5% limestone, mainly calcite, in “Portland cement”, while much higher additions are often used in special cements. However concerns exist about the performance of limestone-rich cements. Thus another main part of this work was to systematically investigate the mineralogical consequences of carbonate addition on cement hydration. Thermodynamic calculations enable a qualitative and quantitative prediction of phase assemblages including AFm and AFt relations as a function of SO3/Al2O3 and CO2/Al2O3 ratios of the cement.

Summary and conclusions 173

At ambient temperatures, ~25°C, and in the presence of portlandite, carbonate stabilises carboaluminate phases while the sulfate thus displaced from AFm forms additional AFt. Specific volume calculations have shown that the space filling of hydrated cement paste can be optimised by adjusting the calcite content such that the AFt content is maximised. Experiments indicate that engineering properties respond to changes of the space filling properties. Controlled calcite addition can therefore be used to improve physical and mechanical properties of hydrated cement paste. Temperature has an important influence of the expected hydrate phase assemblages. While at temperatures ≤ 50°C, and in dependence of the chemical composition of the cement, considerable amounts of calcite react to form carbonate substituted AFm and/or AFt phases, calcite behaves increasingly inert, as a consequence of the instability of carboaluminates, at elevated temperatures; monosulfoaluminate is the preferred AFm phase at high temperatures. On the other hand, carbonate substituted AFt phases, which usually coexist with monocarboaluminate and calcite, are stabilised at low temperatures, preferably <10°C. In dependence of the service temperature up to ~50 mol% of the initial sulfate in AFt can be theoretically replaced by carbonate, which suggests that the maximum AFt content in hydrated cement is no longer restricted by the cement sulfate content. Hence temperature has important influence on the space filling properties of hydrated cement and a general conclusion arose from this work: the higher the temperature the lower the total specific volume of hydrated calcite blended cement paste; other factors being equal, e.g. the specific volume of C-S-H remaining approximately constant. On the contrary, if cements are initially cured at high temperatures ≥ 50°C, e.g. by heat curing or self-heating due to heat of hydration, solids with low specific volume tend to form, e.g. monosulfoaluminate, and calcite behaves inert. If these hydrated cements are subsequently exposed to low service temperatures << 50°C, carboaluminates and AFt will form as a consequence of this mineralogical reconstitution. The amounts of ettringite are predicted to increase with decreasing temperatures due to the possible substitution of carbonate in the AFt structure. Experiments have shown that these mineralogical changes occur relatively rapid, within weeks or months in laboratory simulations. Thus carbonate is another factor which has to be considered by assessing temperature-related phenomenoms, e.g. the occurrence of DEF in heat cured cement systems. This work has shown that carbonate, mainly present in cements as blended calcite, is an important variable, which has to be considered for a correct assessment of cement hydration. Most literature studies tend to neglect the influence of minor limestone addition on cement hydration, the outstanding experimental contributions by Kuzel and co workers or Seligmann and Greening being an exception. Calcite has two functions in hydrated cement, one as an active participant in the hydration process, the other as an inert filler. The mineralogical changes arising from calcite addition are significant and are likely to influence the performance of hydrated cement paste and therefore cannot be neglected. However first results indicate that controlled limestone or calcite addition can be used to optimise engineering properties, e.g. porosity or permeability, of hydrated cements.

References 174

References

[1] Aardt van, J.H.P.; Visser, S.: Some reactions of tricalcium aluminate hexahydrate at medium temperatures. Cement and lime manufacture, 1967, 40, 7-11

[2] Aardt van, J.H.P.; Visser, S.: Thaumasite formation: A cause of deterioration of portland cement and related substances in the presence of sulphates. Cement and Concrete Research, 1975, 5, 225-232

[3] Aguilera, J.; Blanco-Varela, M.T.; Vaszquez, T.: Procedure of synthesis of thaumasite. Cement and Concrete Research, 2001, 31, 1163-1168

[4] Albert, B.; Guy, B.; Damidot, D.: Water chemical potential: A key parameter to determine the thermodynamic stability of some cement phases in concrete? Cement and Concrete Research, 2006, 36, 783-790

[5] Allmann, R.: Die Doppelschichtstruktur der plättchenförmigen Calcium-Aluminium- Hydroxysalze am Beispiel des 3CaO⋅Al2O3⋅CaSO4⋅12H2O. Neues Jahrbuch fuer Mineralogie Monatshefte, 1968, 140-144

+ - [6] Allmann, R.: Refinement of the hybrid layer structure [Ca2Al(OH)6] ⋅[1/2SO4⋅3H2O] . Neues Jahrbuch fuer Mineralogie Monatshefte, 1977, 136-144

[7] Atkins, M.; Glasser, F,P.; Kindness, A.: Phase relations and solubility modelling in the CaO-SiO2-Al2O3-MgO-SO3-H2O-system: for application to blended cements. Materials Research Society Symposium, 1991, Proceedings Vol 212, , 387-394

[8] Atkins, M.; Glasser, F.P.; Jappy, T.G.; Kindness, A.; Moroni, L.P.; Constable, M.; Fenton, A.; Lee, D.J.; Jones, D.V.C.; Cook, A. Wilding, C.R.: Effect of elevated temperature on blended cement in radioactive waste disposal. UK DoE report, 1991, University of Aberdeen

[9] Atkins, M.; Glasser, F.P.; Kindness, A.; Macphee, D.E.: Solubility data for cement hydrate phases (25°C). DOE report No.: DoE/HMIP/RR/91/032, 1991, University of Aberdeen

[10] Atkins, M; Bennet, D.G.; Dawes, A.C.; Glasser, F.P.; Kindness, A.; Read, D.: A thermodynamic model for blended cements. Cement and Concrete Research, 1992, 22, 497- 502

[11] Atkins, M; Bennet, D.G.; Dawes, A.C.; Glasser, F.P.; Kindness, A.; Read, D.: A thermodynamic model for blended cements. DOE report No.: DoE/HMIP/RR/92/005, 1991, University of Aberdeen

[12] Babushkin, V.I.; Matveev, G.M.; Mčedlov-Petrosjan, O.P.: Thermodinamika Silikatov. Strojisdat, Moskau, 1965 (in Russian)

[13] Balonis; M.: personal communication, 2007 [14] Barbarulo, R.: Comportment de materiaux cimentaires: actions des sulfates et de la temperature. PhD-Thesis, University of Laval, Canada, 2002

[15] Barbarulo, R.; Peycelon, H.; Prene, S.: Experimental study and modelling of sulfate sorption on calcium silicate hydrates. Annales de Chimie Science des Matériaux (Suppl.), 2003, 28, S5-S10

[16] Barnett, S.J.; Adam, C.D.; Jackson, A.R.W.: An XRPD profile fitting investigation of the solid solution between ettringite, Ca6Al2(SO4)3(OH)12·26H2O, and carbonate ettringite, Ca6Al2(CO3)3(OH)12·26H2O. Cement and Concrete Research, 2001, 31, 13-17

References 175

[17] Barnett, S.J.; Macphee, D.E.; Lachowski, E.E; Crammond, N.J.: XRD, EDX and IR analysis of solid solutions between thaumasite and ettringite. Cement and Concrete Research, 2002, 32, 719-730

[18] Barry, T. I. and Glasser, F.P: Calculation of Portland cement clinkering reactions, Advances in Cement Research, 2000, 12, 19-28.

[19] Baur, I.; Johnson, A.C.: The solubility of selenate-AFt (3CaO⋅Al2O3⋅3CaSeO4⋅37.5H2O) and selenate-AFm (3CaO⋅Al2O3⋅CaSeO4⋅xH2O). Cement and Concrete Research, 2003, 33, 1741-1748

[20] Bellmann, F.: On the formation of thaumasite CaSiO3⋅CaSO4⋅CaCO3⋅15H2O: Part I.: Advances in Cement Research, 2004, 16, 55-60

[21] Bennet, D.G.; Read, D.; Atkins, M.; Glasser, F.P.: A thermodynamic model for blended cements II: Cement hydrate phases; thermodynamic values and modelling studies. Journal of nuclear materials, 1992, 190, 315-325

[22] Bensted J.: Thaumasite – direct, woodfordite and other possible formation routes. Cement and Concrete Composites, 2003, 25, 873-877

[23] Berman, H.A.; Newman, E.S.: Heat of formation of calcium aluminate monocarbonate at 25°C, Journal of Research of the National Bureau of Standards, A. Physics and. Chemistry. 1961, 65A , 197-207.

[24] Berman, H.A.; Newman, E.S.: Heat of formation of calcium aluminate monosulfate at 25°C, Journal of Research of the National Bureau of Standards, A. Physics and. Chemistry. 1963, 67A , 1-13.

[25] Berman, H.A.; Newman, E.S.: Heat of formation of tricalciumsulfoaluminate at 25°C. Proceedings of the IV. Intern. Symposium on the chemistry of Cements, Washington, D.C, 1964, p. 247-257

[26] Berner U.R.: Modelling the incongruent dissolution of hydrated cement minerals’. Radiochimica Acta, 1988, 44/45, 387-393.

[27] Bessey, G.E.: The calcium aluminate and silicate hydrates. Proceedings of the Intern. Symposium on the Chemistry of Cements, Stockholm 1938, pp 178 – 215

[28] Blanco-Varela, M.T.; Aguilera, J.; Martinez-Ramirez, S.: Effect of cement C3A content, temperature and storage medium on thaumasite formation in carbonated mortars. Cement and Concrete Research, 2006, 36, 707-715

[29] Bonavetti, V.L., Rahhal, V.F., Irassar, E.F.: Studies on carboaluminate formation in limestone filler-blended cements. Cement and Concrete Research, 2001, 31, 853-859

[30] Bourbon, X.: Chemical conceptual models for cement based materials. Mineral phases and thermodynamic data. Andra Technical report C.NT.ASCM.03.026., 2003

[31] Brown, P.; Hooton, R.D.: Ettringite and thaumasite formation in laboratory concretes prepared using sulfate-resisting cements. Cement and Concrete Composites, 2002, 24, 361- 370

[32] Butler, F.G.; Taylor, H.F.W.: The system CaO-Al2O3-H2O at 5°C. Journal of the Chemical Society. 1958, 2103-2110

[33] Buttler, F.G., Dent Glasser, L.S. & Taylor, H.F.W.: Studies on 4CaO⋅Al2O3⋅13H2O and the related natural mineral hydrocalumite. J. Amer. Ceram.Soc.,1959, 42, 121 - 126

References 176

[34] Carlson E.T.: written discussion in: Greene, K.T.: Early hydration reactions of Portland cement. Proceedings of the IV. Intern. Symposium on the chemistry of Cements, Washington, D.C, 1964, Vol. IV p. 375-376

[35] Carlson, E.T.: The system lime-alumina-water at 1°C. Journal of Research of the National Bureau of Standards, 1958, 61, 1-11

[36] Carlson, E.T.; Berman, H.A.: Some observations on the calcium aluminate carbonate hydrates. Journal of Research of the National Bureau of Standards, 1960, 64, 333-341

[37] Catinaud, S.; Beaudoin, J.J.; Marchand J.: Influence of limestone addition on calcium leaching mechanisms in cement-based materials. Cement and Concrete Research, 2000, 30, 1961-1968

[38] Clodic, L.; Meike, A.; Thermodynamics of Calcium Silicate Hydrates. Development of a Database to Model Concrete Dissolution at 25°C Using the EQ3/6 Geochemical Modeling Code. Lawrence Livermore National Laboratory 1997

[39] Courtois, A.; Dusausoy, Y.; Laffaille, A.; Protas, J.: Preliminary studies on the crystalline structure of ettringite. Comptes Rendus des Seances de l'Academie des Sciences, Serie D: Sciences Naturelles, 1968, 266, 1911-1913

[40] Crammond, N.J.: The thaumasite form of sulfate attack in the UK. Cement and Concrete Composites, 2003, 25, 809-818

[41] Crammond, N.J.; Collet, G.W.; Longworth, T.I.: Thaumasite field trial at Shipston on Stour: three-year preliminary assessment of buried concretes. Cement and Concrete Composites, 2003, 25, 1035-1043

[42] Crammond, N.J.; Sibbick, R.G.; Collet, G.W.: Thaumasite field trial at Shipston on Stour: three-year chemical and mineralogical assessment of buried concretes. Proceedings of the 7th CANMET/ACI International Conference on Durability of Concrete, Montreal, Canada, 2006, 539-559

[43] Damidot, D.: Improvement of calcium aluminate cement resistance to sulfate and carbonate attack by silica addition: a thermodynamic approach. in Corrosion of cement paste, Ed. by W. Kurdowski, Kracow, 1994, p. 49-63

[44] Damidot, D.; Barnett, S.J.; Glasser, F.P.; Macphee, D.: Investigation of the CaO-Al2O3- SiO2-CaSO4-CaCO3-H2O system at 25°C by thermodynamic calculations. Advances in cement research, 16, 2004, 69-76

[45] Damidot, D.; Birnin-Yauri, U.A.; Glasser, F.P.: Thermodynamic investigation of the CaO- Al2O3-CaCl2-H2O system at 25°C and the influence of Na2O. Il Cemento, 1994, 91, 243-254

[46] Damidot, D.; Glasser, F.P.: Thermodynamic investigation of the CaO-Al2O3-CaSO4-H2O- System at 25 °C and the influence of Na2O. Cement and Concrete Research, 1993, 23, 221- 238

[47] Damidot, D.; Glasser, F.P.: Thermodynamic investigation of the CaO-Al2O3-CaSO4-H2O- System at 50°C and 85°C. Cement and Concrete Research, 1992, 22, 1179-1191

[48] Damidot, D.; Glasser, F.P.: Thermodynamic investigation of the CaO-Al2O3-SiO2-H2O system at 25°C by thermodynamic calculations. Cement and Concrete Research, 1995, 25, 22-28

[49] Damidot, D.; Stronach, S.; Kindness, A.; Atkins, M.; Glasser, F.P.: Thermodynamic investigation of the CaO-Al2O3-CaCO3-H2O closed system at 25°C and the influence of Na2O. Cement and Concrete Research, 1994, 24, 563-572

References 177

[50] Damidot, D.; Glasser, F.P.: Thermodynamic investigation of the CaO-Al2O3-CaSO4- CaCO3-H2O system at 25°C and the influence of Na2O. Advances in cement research, 7, 1995, 129-134

[51] D'Ans, J.; Eick, H.: Das System CaO-Al2O3-CaSO4-H2O bei 20°C. Zement-Kalk-Gips,1953, 6, No. 6, 302-311

[52] D'Ans, J.; Eick, H.: Das System CaO-Al2O3-H2O bei 20°C und das Erhärten der Tonerdezemente, Zement-Kalk-Gips,1953, 6, No. 9 197-210

[53] Diamond, S.: Thaumasite in Orange County, Southern California: an inquiry into the effect of low temperature. Cement and Concrete Composites, 2003, 25, 1161-1164

[54] Divet, L.; Randrianbololona, R.: Delayed ettringite formation: The effect of temperature and basicity on the interaction of sulphate and C-S-H phase. Cement and Concrete Research, 1998, 28, 357-363

[55] Dosch, W.; zur Strassen, H.: Ein alkalihaltiges Calciumaluminatsulfathydrat (Natrium- Monosulfat). Zement-Kalk-Gips, 1967, 20, No. 9, 392-401

[56] Dosch, W.; zur Strassen, H.: Untersuchungen von Tetracalciumaluminathydraten. I; Die verschiedenen Hydratstufen und der Einfluss von Kohlensäure. Zement-Kalk-Gips, 1965, 18, No. 5, 233-237

[57] Ederova, J.; Satava, V.: Heat capacities of C3AH6, C4ASH12 and C6AS3H32, Thermochimica Acta, 31, 1979, 126-128

[58] Edge, R.A.; Taylor, H.F.W.: Crystal structure of thaumasite, [Ca3Si(OH)6⋅12H2O](SO4)(CO3). Acta Crystallographica, B27, 1971, 594-601

[59] Erlin, B.; Stark, D.C.: Identification and occurrence of thaumasite in concrete. Symposium on the effects of aggressive fluids in concrete. Highway Research Record, 1965, No. 11, 108-113

[60] Famy, C.; Scrivener, K.L.; Atkinson, A.; Brough, A.R.: Influence of the storage conditions on the dimensional changes of heat-cured mortars. Cement and Concrete Research, 2001, 31, 795-803

[61] Faurie-Mounier, M.T.: Contribution à l'étude du système Ca-Al2O3-H2O, Revue des matériaux de construction, 1966, 635-636

[62] Fentiman C.H.: Hydration of carbo-aluminous cement at different temperatures. Cement and Concrete Research, 1985, 15, 622-630

[63] Fischer, R.; Kuzel, H.J.: Reinvestigation of the system C4A⋅nH2O - C4A⋅CO2⋅nH2O. Cement and Concrete Research, 1982, 12, 517-526

[64] Flatt, R.J.; Scherer, G.W.: Thermodynamics of crystallisation stresses in DEF. Cement and Concrete Research, in press 2007

[65] Flint, E.P.; McMurdie, H.; Wells, L.S.: Hydrothermal and X-ray studies of the Garnet- Hydrogarnet series and the relationship of the series to hydration products of Portland cement. Journal of Research of the National Bureau of Standards, 1941, 26, 13-33

[66] Flint, E.P.; Wells, L.S.: Study of the system CaO-SiO2-H2O and of the reaction of water on the anhydrous calcium silicates. Journal of Research of the National Bureau of Standards, A. Physics and. Chemistry. 1934, 12, 751-783

[67] Flint, E.P. and Wells, L.S.: Analogy of hydrated calcium silicoaluminates and hexacalcium aluminate to hydrated calcium sulfoaluminates. J. Res. Natl. Bur. Stand., 1944, 33, 471–478

References 178

[68] Francois, M.; Renaudin, G.; Evrard, O.: A cementitious compound with composition 3CaO⋅Al2O3⋅CaCO3⋅11H2O. Acta Cryst. Sect. C, 1998, C54, 1214-1217

[69] Frias Rojas, M.: Study of hydrated phases present in a MK–lime system cured at 60 °C and 60 months of reaction. Cement and Concrete Research, 2006, 36, 827-831

[70] Fu, Y; Gu, P; Xie, P.; Beaudoin, J.J.: A kinetic study of delayed ettringite formation in hydrated Portland cement paste. Cement and Concrete Research, 1995, 25, 63-70

[71] Fujii, K.; Kondo, W.: Estimation of thermochemical data for Calcium Silicate Hydrate (C- S-H). Journal of the American Ceramic Society, 1983, 66, C 220 – C 221

[72] Fujii, K.; Kondo, W.: Heterogeneous equilibrium of calcium silicate hydrate in water at 30°C. Journal of the Chemical Society, Dalton transactions, 1981, issue 2, 645-651

[73] Gartner, E.M.; Jennings H.M.: Thermodynamics of calcium silicate hydrates and their solutions. Journal of the American Ceramic Society, 1987, 70, 743-749

[74] Gebauer, J.; Harnik, A.B.: Microstructure and composition of the hydrated cement paste of an 84 year old concrete bridge construction. Cement and Concrete Research, 1975, 5, 163- 170

[75] Ghorab, H. Y.; Heinz, D.; Ludwig, U.; Meskendahl, T.; Wolter, A.: On the Stability of Calcium Sulphate Hydrates in Pure Systems and in Cements. Proceedings of the 7th International Congress on the Chemistry of Cement, Paris, 1980, Vol. IV, 496-503

[76] Giampaolo, C.: Dehydration kinetics of thaumasite at ambient pressure. Neues Jahrbuch Mineralogie, Monatshefte, 1986, H. 3, 126-134

[77] Glasser F.P., MacPhee D.E. and Lachowski E.E.: Compositional model for calcium silicate hydrate (C-S-H) Gels, their solubilities, and free energies of formation. Journal of the American Ceramic Society, 1987, 70, 481-485

[78] Glasser, F.P.; Damidot, D.; Atkins, M.: Phase development in cement in relation to the secondary ettringite problem. Advances in Cement Research, 1995, 7, 57-68

[79] Glasser, F.P.; Kindness, A.; Stronach, S.A.: Stability and solubility relationships in AFm phases Part I: Chloride, sulfate and hydroxide. Cement and Concrete Research, 1999, 29, 861-866

[80] Glynn P.: Solid-solution solubilities and thermodynamics: sulfates, carbonates and halides. In: Sulfate Minerals: Crystallography, Geochemistry and Environmental Significance, Alpers C.N., Jambor J.L. and Nordstrom D.K. (eds.), Mineralogical Society of America and Geochemical Society, Reviews in Mineralogy and Geochemistry, 2000, v. 40, 481-511

[81] Glynn, P.D.: MBSSAS: A code for the computing of margules parameters and equilibrium relations in binary solid-solutions aqueous-solution systems. Computers & Geoscience, 17, 1991, 907-966

[82] Glynn, P.D.; Reardon, E.J.; Plummer, L.N.; Busenberg, E.: Reaction paths and equilibrium end-points in solid-solution aqueous-solution systems. Geochimica et Cosmochimica Acta, 1990, 54, 267-282

[83] Götz-Neunhöffer, F.; Neubauer, J.: Refined ettringite Ca6Al2(SO4)3(OH)12⋅26H2O structure for quantitative X-Ray diffraction analysis. Powder Diffraction, 2006, 21, 4-11

[84] Götz-Neunhöffer, F.; Neubauer, J.; Schweswig, P.: Mineralogical characteristics of Ettringites synthesised from solutions and suspensions. Cement and Concrete Research, 2006, 36, 65-70

References 179

[85] Greenberg, S.A.; Chang, T.N.: Investigations of the colloidal hydrated calcium silicates. II Solubility relationships in the calcium oxide-silica-water system at 25°C. The Journal of Physical Chemistry, 1965, 69, 182-188

[86] Gu, Y; Gammons, C.H.; Bloom, M.S.: A one-term extrapolation method for estimating equilibrium constants of aqueous reactions at elevated temperatures, Geochimica et Cosmochimica Acta, 1994, 58, 3545-3560.

[87] Hall, C; Barnes, P.; Billimore, A.D.; Jupe, A.C.; Turillas, X.: Thermal decomposition of ettringite Ca6[Al(OH)6]2(SO4)3·26H2O. J. Chem. Soc., Faraday Trans., 1996, 92, 2125 - 2129

[88] Hartman, M.R.; Berliner, R.: Investigation of the structure of ettringite by time-of-flight neutron powder diffraction techniques. Cement and Concrete Research, 2006, 36, 364-370

[89] Helgeson, H.C.; Delany, J.M.; Nesbitt, H.W.; Bird, D.K.: Summary and critique of the thermodynamic properties of rock-forming minerals. American Journal of Science, 1978, 278-A, 229 pp.

[90] Herfort, D.: Relationship between Strength, Porosity and Cement Composition. unpublished paper, Nanocem workshop on the “Relation between Microstructure and Mechanical Properties” Czech Technical University, Prague 2004

[91] Hobbs, M.: Solubilities and ion exchange properties of solid solutions between the OH, Cl and CO3 end members of the monocalcium aluminate hydrates. PhD-Thesis, University of Waterloo, Canada, 2001

[92] Hong, S.Y; Glasser, F.P.: Alkali binding in cement paste. Cement and Concrete Research, 1999, 29, 1893-1903

[93] Hummel, W.; Berner; U.; Curti, E.; Pearson, F.J.; Thönen, T.: Nagra/PSI Chemical Thermodynamic Data Base 01/01.Universal Publishers, Parkland, Florida, USA, 2002

[94] Ingram, K.D.: A review of limestone additions to Portland cement and concrete. Cement and Concrete Composites, 1991, 13, 165-170

[95] Jappy, T.G.; Glasser, F.P.: Synthesis and stability of silica-substituted hydrogarnet Ca3Al2Si3-xO12-4x(OH)4x. Advances in cement research, 1991/92, 4, 1-8

[96] Johnson, J. W.; Ölkers, E.H.; Helgeson, H.C.: SUPCRT92: A software package for calculating the standard molal thermodynamic properties of minerals, gases, aqueous species, and reactions from 1 to 5000 bar and 0 to 1000°C. Computers & Geosciences, 1992, 18, 899-947

[97] Jones, F.E.: The quaternary system CaO-Al2O3-CaSO4-H2O at 25°C. J. Phys. Chem., 1944, 48, 311-350

[98] Juel, I.; Herfort, D.; Gollop, R.; Konnerup-Madsen, J.; Jakobsen, H.J.; Skibsted, J.: A thermodynamic model for predicting the stability of thaumasite. Cement and Concrete Research, 2003, 33, 867-872

[99] Kalousek, G.L.: Sulfoaluminates of calcium as stable and metastable phases and a study of a portion of the five-component system CaO-SO3-Al2O3-Na2O-H2O at 25°C. PhD-Thesis, University of Maryland, 1941

[100] Kapralik, I.; Hanic, F.: Phase relations in the subsystem C4A3S-CSH2-CH-H2O of the system CaO-Al2O3-CS-H2O referred to the hydration of sulphoaluminate cement. Cement and Concrete Research, 1989, 19, 89-102

References 180

[101] Kersten M.: Aqueous solubility diagrams for cementitious waste stabilisation systems: I, The C-S-H solid solution system. Environmental Science & Technology, 1996, 30, 2286- 2293

[102] Köhler, S.; Heinz, D.; Urbonas, L.: Effect of ettringite on thaumasite formation. Cement and Concrete Research, 2006, 36, 697-706

[103] Kollmann, H, Struebel, G.: Ettringite-Thaumasit-Mischkristalle von Brenk (Eifel). Chemie der Erde, 1981, 40, 110-120

[104] Kulik D.A. and Kersten M.: Aqueous solubility diagrams for cementitious waste stabilisation systems: II, End-member stoichiometries of ideal calcium silicate hydrate solid solutions. Journal of the American Ceramic Society, 2001, 84, 3017-3026

[105] Kulik D.A.: Dual-thermodynamic estimation of stoichiometry and stability of solid solution end members in aqueous–solid solution systems. Geochemical Geology, 2006, 225, 189- 212

[106] Kulik, D.: Minimising uncertainty induced by temperature extrapolations of thermodynamic data: a pragmatic view on the intergration of thermodynamic databases into geochemical computer codes. Proceedings the use of thermodynamic databases in performance assessment, Barcelona, OECD, 2002, pp. 125-137.

[107] Kulik, D.: personal communication 2007 [108] Kulik, D.: Thermodynamic properties of surface species at the mineral–water interface under hydrothermal conditions: A Gibbs energy minimisation single-site 2pKA triple-layer model of rutile in NaCl electrolyte to 250°C. Geochimica et Cosmochimica Acta, 2000, 64, 3161-3169

[109] Kulik, D.; Berner, U.; Curti, E.: Modelling chemical equilibrium partitioning with the GEMS-PSI code. PSI Scientific Report 2003, Vol. IV, 109-122, http://gems.web.psi.ch.

[110] Kurdowski, A.: Role of delayed release of sulphates from clinker in DEF. Cement and Concrete Research, 2002, 32, 401-407

[111] Kuzel H.J.: Crystallographic data and thermal decomposition of synthetic gehlenite hydrate 2CaO⋅Al2O3⋅SiO2⋅8H2O. Neues Jahrbuch fuer Mineralogie Monatshefte, 1976, 319-325

[112] Kuzel H.J.: X-ray investigations of some complex calcium aluminate hydrates and related compounds. Proceedings of the V. Intern. Symposium on the chemistry of Cements, Tokyo, 1969, Vol. II pp 92 – 97

[113] Kuzel, H.J.: Initial hydration reactions and mechanisms of delayed ettringite formation in Portland cement. Cement and Concrete Research, 1996, 26, 195-203

[114] Kuzel, H.-J.; Baier, H.: Hydration of calcium aluminate cements in the presence of carbonate. European Journal of Mineralogy, 1996, 8, 129-141

[115] Kuzel, H.-J.; Meyer, H.: Effect of carbon dioxide on the hydration of pastes. Hydration and setting of cements, Edited by A. Nonat and J.C. Mutin, 137-149, Rilem, E&FN Spon, London, 1992

[116] Kuzel, H.-J.; Meyer, H.: Mechanisms of ettringite and monosulfate formation in cement and 2- concrete in the presence of CO3 . Proceedings of the 15th International Conference on Cement Microscopy, Dallas 1993, 191-203

[117] Kuzel, H.J.; Pöllmann, H: Hydration of C3A in the presence of Ca(OH)2, CaSO4⋅2H2O and CaCO3. Cement and Concrete Research, 1991, 21, 885-895

References 181

[118] Kuzel, H.-J; Strohbauch, G.: Reaktionen bei der Einwirkung von CO2 auf wärmebehandelte Zementsteine. Zement-Kalk-Gips, 1989, 42, No. 8, 413-418

[119] Kwan, S.; LaRosa, J.; Grutzeck, M.W.: 29Si and 27Al MASNMR Study of Strätlingite. Journal of the American Ceramic Society, 1995, 78, 1921-1926

[120] Lippmann F.: The solubility products of complex minerals, mixed crystals, and three-layer clay minerals. Neues Jahrbuch Mineralogie, Abhandlungen, 1977, 130, 243–263

[121] Lippmann, F.: Phase diagrams depicting aqueous solubility of binary mineral systems. Neues Jahrbuch Mineralogie, Abhandlungen, 1980, 139, 1-25

[122] Lippmann F.: Stable and metastable solubility diagrams for the system CaCO3—MgCO3— H2O at ordinary temperature. Bull. Minéral., 1982, 105, 273–279

[123] Lothenbach, B.; Matschei, T.; Möschner, G.; Glasser, F.P.: Thermodynamic modelling of the effect of temperature on the hydration and porosity of Portland cement. Cement and Concrete Research, 2008, 38, 1-18

[124] Lothenbach, B. Winnefeld, F.; Alder, C.; Wieland, E.; Lunk, P.: Effect of temperature on the pore solution, microstructure and hydration products of Portland cement pastes. Cement and Concrete Research, 2007, 37, 483-491

[125] Lothenbach, B.; Winnefeld, F.: Thermodynamic modelling of the hydration of Portland cement, Cement and Concrete Research, 2006, 36, 209-226

[126] Lukas, W.: Betonzerstörung durch SO3-Angriff unter Bildung von Thaumasit und Woodfordit. Cement and Concrete Research, 1975, 5, 503-518

[127] Macphee, D.; Barnett, S.: Solution properties of solids in the ettringite—thaumasite solid solution series. Cement and Concrete Research, 2004, 34,1591-1598

[128] Macphee, D; Diamond, S.: Thaumasite in cementitious materials. special issue Cement and Concrete Composites, 2003, 25, 805-1209

[129] Matschei, T., Glasser, F.P.: Interactions between Portland cement and carbon dioxide. Proceedings of the 12th ICCC, Montreal, 2007

[130] Matschei, T., Skapa, R.; Lothenbach, B.; Glasser, F.P.: The distribution of sulfate in hydrated Portland cement paste. Proceedings of the 12th Intern. Congress on the Chemistry of Cements, Montreal, 2007

[131] Midgley, H.G.; Rosaman, D.: The composition of ettringite in set Portland cement. Proceedings of the IV. Intern. Symposium on the chemistry of Cements, Washington, D.C, 1964, Vol. III-S2 p. 259-262

[132] Möschner, G.; Lothenbach, B.; Kretzschmar, R.; Figi, R.; Ulrich, A.: Löslichkeit von Fe- Ettringit. Proceedings of the 16th Ibausil, Weimar, 2006, Vol. I, 1-0461 - 1-0468

[133] Möschner, G.; Lothenbach, B.; Winnefeld, F.; Ulrich, A.; Figi, R.; Kretzschmar, R.: Solubility of Fe-ettringite and its solid solutions with Al-ettringite. Proceedings of the 12th Intern. Congress on the Chemistry of Cements, Montreal, 2007

[134] Moore, A.; Taylor, H.F.W.: Crystal structure of ettringite. Nature, 1968, 218, 1048-1049 [135] Moore, A.E.; Taylor, H.F.W.: Crystal structure of ettringite. Acta Crystallografica, 1970, B26, 386-393

[136] Murdoch, J.: Crestmore past and present. American Mineralogist, 1961, 46, 245-257

References 182

[137] Murdoch, J.; Chalmers, R.A.: Ettringite (“woodfordite”) from Crestmore. American Mineralogist, 1960, 45, 1275-1278

[138] Murdoch, J.; Chalmers, R.A.: Woodfordite a new mineral from Crestmore, California. Geological Society of America Bulletin, 1958, 69, 1620-1621

[139] Nacken, R.; Mosebach, R.: Untersuchungen am System CaO-Al2O3-H2O. Zeitschrift fuer anorganische und allgemeine Chemie, 1934, Bd. 225, 289-301

[140] Neubauer, J.; Götz-Neunhöffer, F.; Schmitt, D.; Degenkolb, M.; Holland, U.: In-situ Untersuchung der frühen PZ-Hydratation. Proceedings of the 16th Ibausil, Weimar, 2006, Vol. I, 1-0375 - 1-0382

[141] Nishikawa, T.; Sato, K.; Ito, S.; Suzuki, K.: Thermal and chemical stability of AFm-phase- isostructural group. Proceedings of the 9th Intern. Congress on the Chemistry of Cements,, 1992, Vol. I, 437 - 442

[142] Nonat, A.; Courault, A.C.; Damidot, D.: A new model describing the variation of C-S-H Ca/Si ratio with lime concentration in solution. Cement wapno beton, 2001, 5, 184-191

[143] Nordstrom, D.K.; Munoz, J.L.: Geochemical Thermodynamics. 2nd edition, Blackwell Scientific, Palo Alto, 1994

[144] Paul, M.; Glasser, F.P.: Impact of prolonged warm (85°C) moist cure on Portland cement paste. Cement and Concrete Research, 2000, 30, 1869-1877

[145] Percival, A.; Taylor, H.F.W.: Monocalcium Aluminate Hydrate in the system CaO-Al2O3- H2O at 21°C. Journal of the Chemical Society, 1959, notes 526, 2629-2631

[146] Perkins, R.B.; Palmer, C.D.: Solubility of chromate hydrocalumite (3CaO⋅Al2O3⋅CaSeO4⋅nH2O). Cement and Concrete Research, 2001, 31, 983-992

[147] Perkins, R.B.; Palmer, C.D.: Solubility of ettringite (Ca6[Al(OH)6]2(SO4)3·26H2O) at 5– 75°C. Geochimica et Cosmochimica Acta, 1999, 63, 1969-1980

[148] Pöllmann, H.: Die Kristallchemie der Neubildungen bei Einwirkung von Schadstoffen auf hydraulische Bindemittel. PhD-Thesis, University of Erlangen-Nuernberg, 1984

[149] Pöllmann, H.: Organische Derivate des TCAH mit Schichtstrukturen vom Typus + - Ca2[Al(OH)6] [RCOO⋅nH2O] Z. Krist., 1989, 186, 236 - 237

[150] Pöllmann, H.: Solid solution in the system 3CaO⋅Al2O3⋅CaSO4⋅aq - 3CaO⋅Al2O3⋅Ca(OH)2⋅aq. Neues Jahrbuch fuer Mineralogie Abhandlungen, 1989,161 27-41

[151] Pöllmann, H.: Syntheses, properties and solid solution of ternary lamellar calcium 2- 2- - aluminate hydroxi salts (AFm-phases) containing SO4 , CO3 and OH . Neues Jahrbuch fuer Mineralogie Abhandlungen, 2006, 182/2, 173-181

[152] Pöllmann, H.; Kuzel, H.J.; Wenda, R.: Solid solution of ettringites part I: incorporation of − 2- OH and CO3 in 3CaO.A12O3.32H2O. Cement and Concrete Research, 1990, 20, 941-947

[153] Putnis, C.V., Tsukamoto, K.; Nishimura, Y.: Direct observations of pseudomorphism: compositional and textural evolution at a fluid-solid interface. American Mineralogist, 2005, 90 1909-1912

[154] Quillin, K.C., Majumdar, A.J.: Phase equilibria in the CaO-Al2O3-SiO2-H2O system at 5 °C, 20 °C and 38 °C. Advances in Cement Research, 1994, 6, 47-56

[155] R.B. Peppler, L.S. Wells, The system of lime, alumina, and water from 50° to 250° C. Journal of Research of the National Bureau of Standards, 1954, 52, 75-92

References 183

[156] Ramachandran, R.; Gupta, P.K.: An improved spectrophotometric determination of silicate in water based on molybdenum blue. Analytica Chimica Acta, 1985, 172, 86-89

[157] Ramachandran, V.S.; Zhang, C.: Thermal analysis of the 3CaO.Al2O3-CaSO4.2H2O- CaCO3-H2O system. Thermochimica acta, 1986, 106, 273-282

[158] Reardon, E.J.: An ion interaction model for the determination of chemical equilibria in cement/water systems. Cement and Concrete Research, 1990, 20, 175-192

[159] Reardon, E.J.: Problems and approaches to the prediction of the chemical composition in cement/water systems. Waste Management, 1992, 12, 221-239

[160] Richardson, I.G.; Groves, G.W.: The incorporation of minor and trace elements into calcium silicate hydrate (C-S-H) gel in hardened cement pastes. Cement and Concrete Research, 1993, 23, 131-138

[161] Rinaldi, R.; Sacerdoti, M.; Passaglia, E.: Strätlingite: crystal structure, chemistry, and a reexamination, of its polytype vertumnite. European Journal of Mineralogy, 1990, 2, 841- 849

[162] Roberts M.H.: written discussion in: Schwiete, H.E.; Ludwig, U.: Crystal structures and properties of cement hydration products (hydrated calcium aluminates and ferrites). Proceedings of the V. Intern. Symposium on the chemistry of Cements, Tokyo, 1969, Vol. II pp 67-71

[163] Roberts, M.H.: Calcium aluminate hydrates and related basic salt solid solutions. Proceedings of the V. Intern. Symposium on the Chemistry of Cements, Tokyo, 1969, Vol. II pp 104 – 117

[164] Roller, P.S.; Ervine, G.: The system calcium oxide-silica-water at 30°C. The association of silicate ion in dilute alkaline solution. Journal of the Chemical Society, 1940, 62, 461-471

[165] Satava, V.: Determination of standard enthalpies and gibbs ernergies of formation for 6CaO·Al2O3·3SO3·32H2O (C6AS3H32) and 4CaO·Al2O3·SO3·32H2O (C4ASH12) by the DHA- method. Silikaty, 1986, 30, 1-8

[166] Sawicz, Z.; Heng, S.S.: Durability of concrete with addition of limestone powder. Magazine of Concrete Research, 1996, 48, 131-137

[167] Scherer, G.W.: Stress from crystallisation of salt. Cement and Concrete Research, 2004, 34, 1613-1624

[168] Schönitz, M., Navrotsky, A.: Enthalpy of formation of katoite Ca3Al2 [(OH)4]3: Energetics of the hydrogarnet substitution, American Mineralogist, 1999, 84, 389-391

[169] Seligmann, P.; Greening, N.R.: Phase equilibria of cement-water. Proceedings of the V. Intern. Symposium on the Chemistry of Cements, Tokyo, 1969, Vol. II pp 179 – 200

[170] Seligmann, P.; Greening, R.: New techniques for temperature and humidity control in X-ray diffractometry. Journal of the PCA Research and Development Laboratories, 1962, 4, 2-9

[171] Shimada, Y.; Young, J.F.: Thermal stability of ettringite in alkaline solutions at 80°C. Cement and Concrete Research, 2004, 34, 2261-2268

[172] Shock E.L.; Sassani, D.C.; Willis, M.; Sverjensky, D.: Inorganic species in geologic fluids: Correlations among standard molal thermodynamic properties of aqueous ions and hydroxide complexes. Geochimica et Cosmochimica Acta, 1997, 61, 907-950 slop98.dat database available from http://affinity.berkeley.edu

References 184

[173] Smolczyk H.G.: The ettringite phases in blastfurnace cement. Zement-Kalk-Gips, 1961, 14, No. 7, 277-284

[174] Soroka, I.; Setter, N.: The effect of fillers on strength of cement mortars. Cement and Concrete Research, 1977, 7, 449-456

[175] Sprung, S.; Siebel, E.: Beurteilung der Eignung von Kalkstein zur Herstellung von Portlandkalksteinzement (PKZ). Zement-Kalk-Gips, 1991, 44, No. 1, 1-11

[176] Stark, J.; Möser, B. Bellmann, F.: New approaches to cement hydration in the early hardening stage. Proceedings of the 11th Intern. Congress on the Chemistry of Cements, Durban, 2003, Vol. I pp 261-277

[177] Struble LJ. Synthesis and characterisation of ettringite and related phases. In: Proceedings of VIIIth International Congress on the Chemistry of Cement 1987, p. 582–8.

[178] Suzuki, K.; Nishikawa, T.; Ito, S.: Formation and carbonation of C-S-H in water. Cement and Concrete Research, 1985, 15, 213-224

[179] Sverjensky, D.; Shock E.L.; Helgeson, H.C.: Prediction of the thermodynamic properties of aqueous metal complexes to 1000°C and 5 kbar. Geochimica et Cosmochimica Acta, 1997, 61, 1359-1412

[180] Taylor, H.F.W.: Cement Chemistry (2nd Edition), Thomas Telford Publishing, London, 1997

[181] Taylor, H.F.W.: Hydrated Calcium Silicates, Part I: Compound formation at ordinary temperatures. Journal of the Chemical Society, 1950, notes 726, 3682-3690

[182] Taylor, H.F.W.; Famy, C.; Scrivener, K.L.: Delayed ettringite formation. Cement and Concrete Research, 2001, 31, 683-693

[183] Thaumasite Expert Group, The thaumasite form of sulfate attack: Risks, diagnosis, remedial works and guidance on new construction, Report of the Thaumasite Expert Group, Department of the Environment, Transport and the Regions: London, January, 1999

[184] Thönen, T.; Kulik, D. Nagra/PSI Chemical Thermodynamic Data Base 01/01 for the GEM- Selektor (V.2-PSI) Geochemical Modeling Code: Release 28-02-03. PSI Technical Report TM-44-03-04 (unpublished) about the GEMS version of Nagra/PSI chemical thermodynamic database 01/01. 2003 http://les.web.psi.ch/Software/GEMS-PSI/doc/pdf/TM-44-03-04-web.pdf

[185] Torres, S. M.; Kirk, C. A.; Lynsdale, C. J.; Swamy, R. N.; Sharp, J. H.: Thaumasite– ettringite solid solutions in degraded mortars. Cement and Concrete Research, 2004, 34, 1297-1305

[186] Tsivilis, S.; Batis, G.; Chaniotakis, E.; Grigoriadis, Gr.; Theodossis, D.: Properties and behavior of limestone cement concrete and mortar. Cement and Concrete Research, 2000, 30, 1679-1683

[187] Tsivilis, S.; Chaniotakis, E.; Kakali, G.; Batis, G.: An analysis of the properties of Portland limestone cements and concrete. Cement and Concrete Composites, 2002, 24, 371-378

[188] Tsivilis, S.; Tsantilas, J.; Kakali, G.; Chaniotakis, E.; Sakellariou, A.: The permeability of Portland limestone cement concrete. Cement and Concrete Research, 2003, 33, 1465-1471

[189] Turriziani, R.: The calcium aluminate hydrates and related compounds. Cited in Taylor, H.F.W.: Cement Chemistry, Volume I, p. 233 Academic press, London and New York, 1964

References 185

[190] Turriziani, R.; Schippa, G.: Riconoscimento all’ATD ed ai raggi X dei solidi quaternary CaO-Al2O3-CaSO4-H2O. La Ricerca Scientifica, 1955, 24, No. 11, 2356-2363

[191] Walker, C.: Characterisation and solubility behaviour of synthetic calcium-silicate hydrates. PhD-Thesis, University of Bristol, UK, 2003

[192] Warren, C.J.; Reardon, E.J.: The solubility of ettringite at 25°C. Cement and Concrete Research, 1994, 24, 1515-1524

[193] Wells, L.S.; Clarke, W.F.; McMurdie, H.F.: Study of the system CaO-Al2O3-H2O at temperatures of 21° and 90°C. Journal of Research of the National Bureau of Standards, 1943, 367-407

[194] Wieker, W.; Herr, R.: Zu einigen Problemen der Chemie des Portlandzements. Zeitschrift Chemie, 1989, 29, No. 3, 321-327

[195] Zhou, Q.; Glasser, F.P.: Thermal stability and decomposition mechanisms of ettringite at <120°C. Cement and Concrete Research, 2001, 31, 1333-1339

[196] Zhang, F.; Zhou, Z.; Lou, Z.: Solubility product and stability of ettringite, Proceedings of the 7th Intern. Congress on the Chemistry of Cements, 1980, Vol. II, 88– 93

[197] Zhang, M.: Incorporation of Oxyanionic B, Cr, Mo and Se into hydrocalumite and ettringite: Application to cementitious system. PhD-Thesis, University of Waterloo, Canada, 2000

[198] Zur Strassen, H.: Die chemischen Reaktionen bei der Zementerhärtung. Zement-Kalk-Gips, 1958, 11, No. 4, 137-143

[199] Zur Strassen, H.; Strätling, W.: Die Reaktion zwischen gebranntem Kaolin und Kalk in wässriger Lösung II. Zeitschrift fűr anorganische und analytische Chemie, 1940, 245, , 257- 278

References 186

Related publications by the author of this Thesis

Material of this Thesis has already been accepted for publication as follows:

Journals

Matschei, T.; Lothenbach, B.; Glasser, F.P.: Thermodynamic properties of Portland cement hydrates in the system CaO-Al2O3-SiO2-CaSO4-CaCO3-H2O. Cement and Concrete Research, 37, 2007, 1379-1410 (see chapter 4)

Matschei, T.; Lothenbach, B.; Glasser, F.P.: The AFm phase in Portland cement. Cement and Concrete Research, 37, 2007, 118-130 (see chapter 5)

Matschei, T.; Lothenbach, B.; Glasser, F.P: The role of calcium carbonate in cement hydration. Cement and Concrete Research, 37, 2007, 551-558 (see chapter 7)

Matschei, T.; Glasser, F.P.: Zum Einfluss von Kalkstein auf die Zementhydratation. Zement Kalk Gips, 59, 12/2006, 78-86 (see chapter 7)

Conference Proceedings

Matschei, T., Glasser, F.P.: Interactions between Portland cement and carbon dioxide. Proceedings of the 12th ICCC, Montreal, 2007

Matschei, T., Skapa, R.; Lothenbach, B.; Glasser, F.P.: The distribution of sulfate in hydrated Portland cement paste. Proceedings of the 12th Intern. Congress on the Chemistry of Cements, Montreal, 2007

Matschei, T.; Herfort, D.; Lothenbach, B.; Glasser, F.P.: Relationships of Cement Paste Mineralogy to Porosity and Mechanical Properties. Modelling of Heterogenous Materials with Applications in Construction and Biomedical Engineering, Prague, 2007, extended abstract Matschei, T.; Glasser, F.P.: New Approaches to Quantification of Cement Hydration. Proceedings of the 16th Ibausil, Weimar, 2006, Vol. I, 1-0389 - 1-0400 Matschei, T.; Glasser, F.P.: Carbonation - The early stages. International Conference on Accelerated Carbonation for Environmental and Materials Engineering, London, 2006, extended abstract Matschei, T.; Lothenbach, B.; Glasser, F.P.: The AFm phase in Portland cement. IoMMM 25th Cement and Concrete Science, London, 2005, extended abstract

Abbreviations 187

Abbreviations

Nomenclature in cement chemistry

C CaO A Al2O3 S SiO2 s SO3 c CO2 H H2O

AFm Al2O3-Fe2O3-mono phase AFt Al2O3-Fe2O3-tri phase

C-S-H Calcium silicate hydrate C3A Tricalcium aluminate

C3S Tricalcium silicate

Abbreviations used in calculations

A Debye-Hűckel solvent parameters dependent on the dielectric constant of water and temperature (A = 0.5114 at 25°C)

αi parameter dependent on the size of species, i, taken from Kielland`s table (cited in [143]) B Debye-Hűckel solvent parameters dependent on the dielectric constant of water and temperature (B = 0.3288 at 25°C) bi common extended Debye-Hűckel parameter (bi ~ 0.064 at 25°C) Cp0 standard molar heat capacity of species at T, P (J K-1 mol-1) 0 -1 -1 ΔrCp T standard molar heat capacity change of reaction at T (J K mol ) 0 -1 -1 ΔrCp T0 standard molar heat capacity change of reaction at T0 =298 K (25°C) (J K mol ) 0 -1 ΔfG T0 standard molar Gibbs energy (of formation from elements) at T0 = 298 K (25°C) (kJ mol ) 0 -1 ΔrG T standard Gibbs energy change in a reaction (kJ mol ) -1 ΔGex excess molar Gibbs energy of mixing for the solid solution series (kJ mol ) 0 ΔfG i standard molar Gibbs energy of formation of end member i of a solid solution series (kJ mol-1) -1 ΔGid molar Gibbs energy of mixing of an ideal solid solution (kJ mol ) -1 ΔGM molar Gibbs energy of mixing for end members i of the solid solution series (kJ mol ) -1 ΔGss molar Gibbs energy of a solution between different end members i (kJ mol )

γi Activity coefficient of species i 0 -1 -1 ΔrH T standard change of enthalpy of reaction at T (kJ K mol ) 0 -1 -1 ΔrH T0 standard change of enthalpy of reaction at T0 = 298 K (25°C) (kJ K mol ) 0 -1 -1 ΔfH T0 standard molar enthalpy at T0 = 298 K (25°C) (kJ K mol ) I effective molal ionic strength of aqueous solution

KT thermodynamic equilibrium constant of reaction at T ∏ total solubility product in Lippmann phase diagrams R universal gas constant (8.31451 J K-1mol-1) 0 -1 -1 ΔrS T standard entropy change in reaction at T (J K mol )

Abbreviations 188

0 -1 -1 ΔrS T0 standard entropy change in reaction at T0= 298 K (25°C) (J K mol ) 0 -1 -1 S T0 standard molar absolute entropy at T0 = 298 K (25°C) (J K mol ) T Temperature of interest (K)

T0 Reference temperature (298 K) V0 standard molar volume (cm3 mol-1)

Xaq,i aqueous activity fractions of the substitutable species i

Xi mole fraction of end member i in solid solution

Other abbreviations

BSE Backscattered electron DEF Delayed ettringite formation DTA Differential thermal analysis DTG Derivative thermogravimetry EN European standard ESEM Environmental scanning electron microscope GEMS Gibbs energy minimisation selector GSE Gaseous secondary electron HDPE High-density polyethylene MBSSAS Margules binary solid solution aqueous solution (computer code by Glynn et al. [81]) OPC Ordinary Portland cement PP Polypropylene PDF Powder diffraction file PE Polyethylene PSI Paul Scherre Institute PTFE Polytetrafluoroethylene (Teflon) SSAS Solid solution aqueous solution TG Thermogravimetry TSA Thaumasite form of sulfate attack w/c water/cement weight-ratio w/s water/solid weight-ratio XRD X-ray diffraction XRPD X-ray powder diffraction

Figures and Tables 189

Figures and Tables

Figures

Fig. 4.1: Estimation of the silicon-content of siliceous hydrogarnet ...... 16

Fig. 4.2: Comparison of XRD-patterns of C3AH6 following annealing at 5°C, 25°C and 105°C.27

Fig. 4.3: Calculated solubility products of C3AH6 from solubility experiment^ ...... 28

Fig. 4.4: Recalculated solubility data for hydrogarnet, C3AH6, based on fitted thermodynamic data from Fig. 4.3 ...... 28 Fig. 4.5: Comparison of XRD-patterns of siliceous hydrogarnet following ageing at temperatures in the range 5°C - 85°C...... 29

Fig. 4.6: Calculated solubility products of siliceous hydrogarnet, C3AS0 8H4 4, from solubility experiments...... 30

Fig. 4.7: Recalculated solubility data for hydrogarnet, C3AS0 8H4 4, based on fitted thermodynamic data from Fig. 4.6...... 30 Fig. 4.8: ESEM micrographs of synthetic monosulfoaluminate a) after synthesis at 90°C and b) after 4 weeks redispersion in deionised water at 25°C ...... 30 Fig. 4.9: Comparison of XRD-patterns of monosulfoaluminate after ageing at temperatures in the range 5°C - 90°C...... 31 Fig. 4.10: Calculated solubility products of monosulfoaluminate from solubility experiments.... . 32 Fig. 4.11: Recalculated solubility data for nominally monosulfoaluminate based on fitted thermodynamic data from Fig. 4.10...... 32 Fig. 4.12: Comparison of XRD-patterns of monocarboaluminate after annealing at temperatures in the range 5°C - 110°C...... 33 Fig. 4.13: Calculated solubility products of monocarboaluminate from solubility experiments ... . 34 Fig. 4.14: Recalculated solubility data for nominally monocarboaluminate based on fitted thermodynamic data from Fig. 4.13...... 34 Fig. 4.15: Comparison of XRD-patterns of nominally hemicarboaluminate after ageing at temperatures in the range 5°C - 85°C ...... 34 Fig. 4.16: Calculated solubility products of hemicarboaluminate from solubility experiments .... . 35

Fig. 4.17: Recalculated solubility data for nominally hemicarboaluminate based on fitted thermodynamic data from Fig. 4.16...... 35 Fig. 4.18: Calculated mass balance of initially hemicarboaluminate (2g hemicarb. reacted in 60 g water at selected temperatures between 1°C - 99°C)...... 36

Fig. 4.19: Calculated solubility products of C4AH13 (left) and C2AH8 (right) from literature solubility data...... 36

Fig. 4.20: Calculated solubility products of CAH10 from literature solubility data...... 37

Fig. 4.21: Solubility relations between C2AH8, C4AH13 and CAH10 in the system CaO-Al2O3-H2O at 5°C (left) and 25°C (right)...... 38 Fig. 4.22: Comparison of XRD-patterns of nominally strätlingite after ageing at temperatures in the range 5°C - 85°C...... 39 Fig. 4.23: ESEM image of strätlingite (left) and corresponding TG-and DTG curves (right)...... 40 Fig. 4.24: Calculated solubility products of strätlingite from solubility experiments...... 40

Figures and Tables 190

Fig. 4.25: Recalculated solubility data for strätlingite based on fitted thermodynamic data from Fig. 4.24 ...... 40 Fig. 4.26: Calculated solubility product of ettringite according to solubility experiments by Perkins &Palmer and other literature sources...... 41

Fig. 4.27:XRD-pattern (left) and scanning electron micrograph (right) of CO3-AFt used for solubility determinations...... 42

Fig. 4.28: Calculated solubility product of CO3-AFt from solubility experiments...... 43

Fig. 4.29: Free energy plot of CO3-AFt and monocarboaluminate...... 43 Fig. 4.30: Comparison of recalculated solubility of C-S-H at 25°C (solid line) with experimental values from literature (markers) at temperatures between 20°C -30°C ...... 44 Fig. 4.31: Comparison of solubility products calculated from the three-term temperature extrapolation (solid black line) and integrated Van’t Hoff extrapolation (dashed black line) with values calculated from the HKF equation of state (grey line) and experimentally-derived solubility products of siliceous hydrogarnet (markers)...... 47 Fig. 4.32: Influence of sulfate and carbonate on the relative stabilities of aluminate hydrates...... 50 Fig. 4.33: Influence of silicon on the relative stabilities of aluminate hydrates...... 51 + Fig. 5.1: Brucite-like [Ca2Al(OH)6] main layer projected onto the c-axis according to Allmann [5] .53 Fig. 5.2: Structure of monosulfoaluminate according to Allmann [5] ...... 53 Fig. 5.3: XRD-pattern of selected solid solutions between monosulfoaluminate and hydroxy- AFm aged 28d...... 62

Fig. 5.4: Observed interplanar distances d0001 of monosulfoaluminate type solid solution (top) and hydroxy-AFm type solid solution (bottom) ...... 63 Fig. 5.5: Refined lattice parameter for the monosulfo-aluminate type phase with changing sulfate ratio ...... 64 Fig. 5.6: Calculated volume change for the monosulfoaluminate type phase with changing sulfate ratio ...... 64 Fig. 5.7: ESEM micrographs of the monosulfoaluminate end member and ettringite (a) and the monosulfoaluminate type solid solution (~10 mol% of sulfate substituted by hydroxide) (b)...... 64

Fig. 5.8: ESEM micrograph of cubic C3AH6 crystals on the surface of a crystal of the hydroxy- AFm end member ...... 65 Fig. 5.9: Lippmann diagram for the monosulfoaluminate-hydroxy-AFm system at 25°C with

a0=0.188 and a1=2.49 for miscibility gap fractions 0.03 ≤ x ≤ 0.5 (equation 5.1) ...... 66

Fig. 5.10: Gibbs free energy of ideal mixing, ΔGid, excess Gibbs free energy of mixing, ΔGex, and resulting Gibbs free energy of mixing, ΔGM, calculated from Eq.4.14 for the monosulfoaluminate-hydroxy-AFm solid solution series...... 68 Fig. 5.11: Comparison of calculated (lines) and mean experimental (markers) solubility data for the

SO4-AFm and OH-AFm solid solution series at 25°C...... 69 Fig. 5.12: XRD-pattern of solid solutions between monosulfoaluminate and hydroxy-AFm aged 2 years...... 70

Fig. 5.13: Formation of C3AH6 in dependence of the initial molar bulk SO4/(SO4+2OH)-ratio of the solid solution preparations at 25°C ...... 70 Fig. 5.14: calculated thermodynamic metastable solid mass balance changes of the monosulfoaluminate-hydroxy-AFm solid solution series at 25°C...... 71

Figures and Tables 191

Fig. 5.15: Calculated thermodynamic stable solid mass balance changes of the monosulfoaluminate-hydroxy-AFm solid solution series at 25°C...... 71

Fig. 5.16: Changes of interplanar distances d0001 of monosulfoaluminate type solid solution with prolonged ageing up to 2 years at 25°C ...... 71

Fig. 5.17: Comparison of calculated (solid lines = stable phase assemblage including C3AH6; dashed lines metastable phase assemblage, formation of C3AH6 suppressed) and mean experimental (markers) solubility data for the SO4-AFm and OH-AFm solid solution series in dependence of the sample age at 25°C ...... 72 Fig. 5.18: Change of pH due to prolonged ageing of the solid solution preparations at 25°C (markers = experimentally measured; solid lines = calculated stable phase assemblage including C3AH6; dashed lines calculated metastable phase assemblage, formation of C3AH6 suppressed)...... 73 Fig. 5.19: Comparison of XRD-pattern of 1:1 molar mixtures of synthesised sulfate- and carbonate- AFm end members...... 73

Fig. 5.20: Comparison of d0001 spaces for C4AHx-C4AcH11 mixed solids at25°C...... 74 Fig. 5.21: Calculated metastable phase assemblages between different AFm phases at 25°C. A possible range of stoichiometry of hemicarboaluminate and C4AHx is not shown...... 75 Fig. 5.22: Calculated stable phase assemblages between different AFm phases at 25°C. A possible range of stoichiometry of hemicarboaluminate and C4AHx is not shown...... 77 Fig. 6.1: Crystal structure of ettringite ...... 82

Fig. 6.2: XRD-patterns of selected SO4-CO3-AFt phase solid solutions...... 87

Fig. 6.3: Lippmann diagram for the SO4-CO3-AFt solid solution series at 25°C with a0=-0.823 and a1=2.82 for miscibility gap fractions 0.12 ≤ x ≤ 1.25 (Eq. 6.1) ...... 89 Fig. 6.4: Comparison between calculated and theoretical changes of the relative energy of mixing of the SO4-CO3-solid solution series...... 90 Fig. 6.5: Comparison of calculated (lines) and mean experimental (markers) solubilities for the

SO4-AFt and CO3-AFt solid solution series at 25°C...... 91

Fig. 6.6: Comparison of the XRD-pattern of synthetic thaumasite with SO4-AFt and CO3-AFt. . 93 Fig. 6.7: ESEM micrograph of synthetic thaumasite at 5°C ...... 93 Fig. 6.8: Thermal analysis (TG and DTG data) of synthetic thaumasite...... 93 Fig. 6.9: XRD-pattern of synthetic thaumasite and related and related decomposition products upon redispersion at temperatures from 1°C to 70°C ...... 95 Fig. 6.10: ESEM micrograph of partly decomposed synthetic thaumasite embedded in a matrix of C-S-H and calcite at 55°C...... 95 Fig. 6.11: ESEM micrograph of C-S-H and calcite resulting from thaumasite decomposition at 70°C ...... 95

Fig. 6.12: Thermal analysis (TG data left; DTG data right; heating rate 10K/min; N2-atmosphere) of synthetic thaumasite and related decomposition products following redispersions at temperatures from 1°C to 70°C...... 96 Fig. 6.13: Calculated solubility products of synthetic and natural thaumasite samples from solubility experiments...... 97 Fig. 6.14: Recalculated solubility data for thaumasite based on fitted thermodynamic data (Table 6.1) and for other relevant cement hydrates (Table 4.1)...... 98 Fig. 6.15: Calculated mass balance changes due to the decomposition of thaumasite with increasing temperatures (water/solid-ratio = 120)...... 99

Figures and Tables 192

Fig. 6.16: Free energy-plot of the relative thermodynamic stability of thaumasite in the system CaO-SiO2-CaCO3-H2O ...... 100

Fig. 6.17: Temperature dependence of the phase relations between thaumasite and ettringite, SO4- AFt (according to Macphee and Barnett [99]) ...... 101 Fig. 6.18: Comparison of theoretically calculated (solid lines denote single solid phase regions) and dashed lines two solid phase regions), using the Guggenheim parameters a0 = -0.05 and a1 = 2.74) and experimentally-derived solubility products...... 102 Fig. 6.19: Comparison of calculated (lines) and experimental (single markers, taken from Macphee and Barnett [99]) solubility data for the ettringite-thaumasite solid solution series at 15°C ...... 103

Fig. 6.20: XRD-pattern of a mixture of initially C3A, CaSO4, Na2CO3 (molar ratio 1 : 1.8 : 1.2) and excess Ca(OH)2 reacted 4 weeks in deionised water (w/s~30) at 25°C...... 105

Fig. 6.21: Calculated stable assemblages in the SO4-CO3-AFt system buffered by excess portlandite at 25°C ...... 105

Fig. 6.22: 3D-representation of part of the system CaO-Al2O3-SiO2-CaCO3-CaSO4-H2O at 25°C, 1bar total pressure (excess calcite present) ...... 107

Fig. 6.23: 3D-representation of part of the system CaO-Al2O3-SiO2-CaCO3-CaSO4-H2O at 5°C 1bar total pressure (excess calcite present) ...... 110 Fig. 7.1: XRD-pattern showing mineralogical changes during 90d hydration of a 1:1:2 molar

mixture of C3A, CaSO4 and CaO (w/s = 5) at 25°C...... 122 Fig. 7.2: XRD-pattern showing mineralogical changes during 90d hydration of a 1:1:1:2 molar

mixture of C3A, CaSO4, CaCO3 and CaO (w/s = 5) at 25°C ...... 122 Fig. 7.3: Comparison of qualitative time-dependent mineralogical changes during hydration of C3A with lime and calcium sulfate (SO3/Al2O3=1) in carbonate-free systems (left) or in the presence of excess calcite (right) at 25°C ...... 123 Fig. 7.4: Phase relations between AFm-AFt phases as well as gypsum, hydrogarnet and calcite in

the system CaO-Al2O3-CaSO4-CaCO3-H2O at 25°C (excess portlandite present in all assemblages) ...... 125

Fig. 7.5: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at 25°C (the influence of minor carbonate substitution in AFt is not shown)...... 129

Fig. 7.6: Relative amount of hydrate phases of a hydrated model mixture consisting of C3A, portlandite and with fixed sulfate ratio (SO3/Al2O3=1) in dependence of changing carbonate contents (CO2/Al2O3) at 25°C (the influence of minor carbonate substitution in AFt is not shown) ...... 131

Fig. 7.7: Calculated amount of reactive CaCO3 in dependence of the initial amount of initial solid Al2O3 content and bulk SO3/Al2O3 ratio of the paste at 25°C (data are expressed in weight units; the influence of minor carbonate substitution in AFt is not considered) ... 132

Fig. 7.8: Calculated amount of reactive CaCO3 in dependence of the initial amount of initial solid SO3 content and bulk SO3/Al2O3 ratio of the paste at 25°C (data are expressed in weight units; the influence of minor carbonate substitution in AFt is not considered) ...... 133 Fig. 7.9: Specific volume changes of hydrate phases of a hydrated model mixture consisting of C3A, portlandite and with fixed sulfate ratio (SO3/Al2O3=1) in dependence of changing carbonate contents (CO2/Al2O3) at 25°C (the influence of minor carbonate substitution in AFt is not considered)...... 134

Figures and Tables 193

Fig. 7.10: Change of composition of aqueous phases of a hydrated model mixture consisting of C3A, portlandite at a fixed sulfate ratio (SO3/Al2O3=1) in dependence of changing carbonate ratios (CO2/Al2O3) at 25°C...... 135 Fig. 7.11: XRD-pattern showing the influence of calcite addition to hydrated Portland cement analogue (mixture of C3S, C3A, CaSO4 and K2SO4, SO3/Al2O3molar=0.84, w/s=0.5) at sealed conditions stored 90d at ~25°C...... 138 Fig. 7.12: Volume changes of hydrate phases and pore solution of a hydrated model cement with fixed sulfate ratio (SO3/Al2O3~0.7) in dependence of changing carbonate contents. .... 140

Fig. 7.13: Dependence of the amount of excess pore solution from the initial CO2/Al2O3 ratio of a hydrated model cement with fixed sulfate ratio (SO3/Al2O3=0.7;w/c = 0.5) at 25°C.... 140

Fig. 7.14: Comparison of relative changes of calculated porosity of model cement paste (SO3/Al2O3 ~ 0.7, w/s=0.5) and relative changes of measured 1 year compressive strength of mortars (SO3/Al2O3 ~ 0.7, w/c=0.5)...... 141

Fig. 7.15: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate ratios (CO2/Al2O3) at 5°C...... 143

Fig. 7.16: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at 40°C ...... 145

Fig. 7.17: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at 60°C ...... 146

Fig. 7.18: Calculated phase assemblages of a hydrated mixture consisting of C3A, portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at 90°C(left) and 99°C (right)...... 147 Fig. 7.19: Influence of temperature on specific volume changes of hydrate phases of a hydrated model mixture consisting of C3A, portlandite, CaSO4 and CaCO3 ; fixed sulfate ratio (SO3/Al2O3=1) and carbonate ratio (CO2/Al2O3 =1.25)...... 148 Fig. 7.20: Calculated maximum amount of carbonate substitution in AFt as a function of temperature; excess calcite, monocarboaluminate and portlandite present ...... 149 Fig. 7.21: Influence of temperature on specific volume changes of hydrate phases of a hydrated model mixture consisting of C3A, portlandite, CaSO4 and CaCO3 ; fixed sulfate ratio (SO3/Al2O3=1) and carbonate ratio (CO2/Al2O3 =0.33)...... 150 Fig. 7.22: XRD-pattern showing the influence of temperature on phase assemblages after 90d

hydration of a 1:1:2 molar mixture of C3A, CaSO4 and CaO (w/s = 5) at 25°C, 50°C and 85°C ...... 152 Fig. 7.23: Qualititative comparison of time-dependent mineralogical changes during hydration of a

1:1:2 molar mixture of C3A, CaSO4 and CaO (w/s = 5) at 50°C (left) and 85°C (right)153 Fig. 7.24: XRD-pattern showing the influence of temperature on phase assemblages after 90d

hydration of a 1:1:1:2 molar mixture of C3A, CaSO4, CaCO3 and CaO (w/s = 5) at 25°C, 50°C and 85°C ...... 154 Fig. 7.25: Qualitative comparison of time-dependent mineralogical changes during hydration of a

1:1:1:2 molar mixture of C3A, CaSO4, CaCO3 and CaO (w/s = 5) at 50°C (left) and 85°C (right) ...... 154 Fig. 7.26: XRD-pattern showing the influence of temperature on phase assemblages after 90d hydration of a Portland cement analogue composition (w/c = 0.5) at 25°C, 50°C and 85°C ...... 155 Fig. 7.27: XRD-pattern showing the influence of temperature on phase assemblages after 90d hydration of a Portland cement analogue composition blended with limestone (w/c = 0.5) at 25°C, 50°C and 85°C ...... 157

Figures and Tables 194

Fig. 7.28: XRD-pattern showing the phase assemblages of carbonate-free Portland cement ana- logue (w/c = 0.5) after 24h heat curing at 85°C and subsequent storage at 5°C for 180 d...... 159 Fig. 7.29: XRD-pattern showing the phase assemblages of a calcite blended Portland cement ana- logue (w/c = 0.5) after 24h heat curing at 85°C and subsequent storage at 5°C for 180d...... 159

Fig. 7.30: Calculated phase assemblages of a hydrated mixture consisting of C3A, C-S-H (Ca/Si~1.6), portlandite and varying initial sulfate (SO3/Al2O3) and carbonate contents (CO2/Al2O3) at different temperatures...... 162 Fig. 7.31: Schematic showing the probability of hydrogarnet formation as function of time and temperature (solid lines indicate that formation is likely, dashed lines indicate formation is uncertain)...... 163

Tables

Table 4.1: Standard molar thermodynamic properties of cement hydrates at 25°C, 1 bar ...... 20 Table 4.2: Reference reactions used to estimate unknown heat capacities of cement minerals ...... 21 Table 4.3: Dissolution reactions used to calculate solubility products...... 22

Table 4.4: Metastable and stable invariant points in the system CaO-Al2O3-H2O at 5 and 25°C ... 38 Table 5.1: Thermodynamic stabilities of selected AFm phases at 25°C ...... 55 Table 5.2: Sulfate- and carbonate-contents of minerals relevant to Portland cement...... 55 Table 5.3: Observed phase assemblages and their dependence of the initial solid carbonate ratio at 25°C...... 74 Table 6.1: Standard molar thermodynamic properties of thaumasite at 25°C...... 97 Table 6.2: Composition of aqueous phase of selected stable phase assemblages and invariant points

in the system CaO-Al2O3-SiO2-CaSO4-CaCO3-H2O at 1bar and 5 and 25°C...... 108 Table 6.3: Comparison of calculated activities for selected aqueous species relevant to .point I, Fig. 6.22, in dependence of the alkali content of the aqueous phase at 25°C ...... 113 Table 7.1: Composition of aqueous phase of selected stable phase assemblages and invariant points

in the system CaO-Al2O3-CaSO4-CaCO3-H2O at 1bar and 25°C ...... 126 Table 7.2: Chemical composition of experimentally used Portland cement analogue ...... 136 Table 7.3: Influence of alkalis on the composition of aqueous and solid phases at selected invariant points at 25°C ...... 165

Appendix 195

Appendix

A Thermodynamic data of the aqueous species and solubility data of cement hydrates ...... II

B Additional figures...... XVII

B.1 AFm solid solutions...... XVII B.2 Thermodynamic calculations and free energy plots ...... XVIII

B.3 Additional XRD-patterns C3A hydration - carbonate-free system...... XXI

B.4 Additional XRD-patterns C3A hydration - excess calcite present ...... XXII B.5 XRD-patterns Portland cement analogue - carbonate-free system ...... XXIII B.6 XRD-patterns Portland cement analogue - excess calcite present ...... XXV

C Additional details - AFm experiments ...... XXVII

Appendix II

A Thermodynamic data of the aqueous species and solubility data of cement hydrates

Appendix III

Table A.1: Standard (partial molal) thermodynamic properties of aqueous species used in GEM Calculations at 25°C, 1 bar 0 I) 0 0 0 0 IV) IV) IV) IV) IV) IV) IV) V) species ΔG ΔH S Cp V a1 a2 a3 a4 c1 c2 w Ref 3 [kJ/mol] [kJ/mol] [J/(K⋅mol)] [J/(K⋅mol)] [cm /mol] [cal/(bar⋅mol)] [cal/mol] [calK/(bar⋅mol)] [calK/mol] [cal/Kmol] [calK/mol] [cal/mol] Al3+ -483.71 II) -530.63 -325.10 -128.70 -45.24 -0.33802 -1700.71 14.5185 -20758 10.7 -80600 275300 [172] + + AlO ( + H2O = Al(OH)2 ) -660.42 -713.64 -112.97 -125.11 0.31 0.21705 -248.11 6.7241 -26763 -2.5983 -91455 95700 [172] - - AlO2 ( + 2H2O = Al(OH)4 ) -827.48 -925.57 -30.21 -49.04 9.47 0.37221 399.54 -1.5879 -29441 15.2391 -54585 174180 [172]

AlO2H ( aq) ( +2H2O = Al(OH)3 (aq ) -864.28 -947.13 20.92 -209.21 13.01 0.35338 84.85 5.4132 -28140 -23.4129 -132195 -3000 [172] AlOH2+ -692.60 -767.27 -184.93 55.97 -2.73 0.20469 -278.13 6.8376 -26639 29.7923 -3457 172470 [172] + AlSO4 -1250.43 -1422.67 -172.38 -204.01 -6.02 0.13869 -439.2 7.4693 -25974 -11.6742 -129914 117290 [179] - Al(SO4)2 -2006.30 -2338.40 -135.50 -268.37 31.11 0.68275 889.25 2.2479 -31466 -12.022 -161447 211990 [179] Ca2+ -552.79 II) -543.07 -56.48 -30.92 -18.44 -0.01947 -725.2 5.2966 -24792 9 -25220 123660 [172] CaOH+ -717.02 -751.65 28.03 6.05 5.76 0.27243 -113.03 6.1958 -27322 11.1286 -27493 44960 [172]

CaCO3 (aq) -1099.18 -1201.92 10.46 -123.86 -15.65 -0.03907 -873.25 9.1753 -24179 -11.5309 -90641 -3800 [179] + CaHCO3 -1146.04 -1231.94 66.94 233.70 13.33 0.3706 126.7 5.252 -28310 41.722 83360 30800 [96]

CaSO4 (aq) -1310.38 -1448.43 20.92 -104.60 4.70 0.24079 -189.92 6.4895 -27004 -8.4942 -81271 -100 [179] + + CaHSiO3 ( +H2O = CaSiO(OH)3 ) -1574.24 -1686.48 -8.33 137.80 -6.74 0.10647 -517.87 7.7785 -25649 30.8048 36619 58310 [179] Na+ -261.88 II) -240.28 58.41 38.12 -1.21 0.1839 -228.5 3.256 -27260 18.18 -29810 33060 [172] NaOH (aq) -418.12 -470.14 44.77 -13.40 3.51 0.22338 -232.87 6.6683 -26826 4.0146 -36863 -3000 [172] - VII) NaCO3 -797.11 -938.56 -44.31 -51.28 -0.42 0.23862 -195.21 6.5103 -26982 15.3395 -55686 178700 [184] VII) NaHCO3 (aq) -847.39 -929.50 154.72 200.33 32.32 0.6173 729.43 2.876 -30805 33.879 67193 -3800 [184] - VII) NaSO4 -1010.34 -1146.66 101.76 -30.09 18.64 0.47945 392.84 4.199 -29414 13.4899 -45256 126060 [184]

CO2 (aq) -386.02 -413.84 117.57 243.08 32.81 0.62466 747.11 2.8136 -30879 40.0325 88004 -2000 [172] 2- CO3 -527.98 -675.31 -50.00 -289.33 -6.06 0.28524 -398.44 6.4142 -26143 -3.3206 -171917 339140 [172] - II) HCO3 -586.94 -690.01 98.45 -34.85 24.21 0.75621 115.05 1.2346 -28266 12.9395 -47579 127330 [172] 2- II) SO4 -744.46 -909.70 18.83 -266.09 12.92 0.83014 -198.46 -6.2122 -26970 1.64 -179980 314630 [172] - HSO4 -755.81 -889.23 125.52 22.68 34.84 0.69788 925.9 2.1108 -31618 20.0961 -19550 117480 [172] OH- -157.27 -230.01 -10.71 -136.34 -4.71 0.12527 7.38 1.8423 -27821 4.15 -103460 172460 [172] - - HSiO3 ( + H2O = H3SiO4 ) -1014.60 -1144.68 20.92 -87.20 4.53 0.29735 -51.81 5.9467 -27575 8.1489 -73123 155110 [179] 0 -2 silica species (temperature correction using Cp(T) integration) a0 a1 a2 (Cp =a0+a1T+a2T ) [J/(mol⋅K)] [J/(mol⋅K2)] [J/(mol⋅K0.5)] VIII III) III) III) III) III) SiO2 (aq) ( + 2H2O = H4SiO4 (aq)) -833.41 -887.86 41.34 44.47 16.06 46.94 0.034 -1.13E+06 IX) -1 silica species (estimated temperature correction by one and two term temperature extrapolation) A0 [-] A1[-] A2[-] (LogK=A0+A1T+A2T ) - - IX) AlSiO4 ( + 3H2O = Al(OH)6SiO ) -1681.44 -1833.98 11.13 -4.58 25.53 0 0 1073.34 2+ 2+ AlHSiO3 (+ H2O = AlSiO(OH)3 ) -1540.55 -1717.55 -304.18 -215.90 -40.72 0 0 2206.31

CaSiO3 (aq) ( + H2O = CaSiO2(OH)2 (aq)) -1517.56 -1668.06 -136.68 88.90 15.69 0 0 1371.49 2- 2- SiO3 ( + H2O = H2SiO4 ) -938.51 -1098.74 -80.20 119.83 34.13 -10.0006 0 -3917.5 I) recalculated by Thönen et al. [184] based on log_k values given in Hummel et al. [93] if not stated otherwise II) taken from Shock et al. [172] III)Cp parameters calculated using reference reaction IV) V) SiO2 (aq) = SiO2(amorph) see Table 2 parameters to solve the HKF-equation of state; given in original calorimetric units (see [96][172][179] as used in GEMS references for HKF parameters 0 VII) VIII) IX) - - 0 0 and S (see [184] for additional information) HKF parameters predicted using program PRONSPREP [179] taken from slop98.dat (see [96]) AlO2 + SiO2(aq) → AlSiO4 (ΔrS =ΔrCp =0); 3+ - 2+ 0 0 2- + 0 0 2- 2+ 0 0 Al + HSiO3 → AlHSiO3 (ΔrS =ΔrCp =0); SiO2(aq) + H2O → SiO3 + 2H (ΔrCp =0; ΔrH taken from [93]); SiO3 + Ca → CaSiO3 (aq) (ΔrS =ΔrCp =0)

Appendix IV

Table A.2: Solubility data for C3AH6 at different temperatures 5°C - 105°C and ages

C3AH6 (samples cooled down from initially ~105°C)

Age Temp. Ca Al pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l] 28 5 6.57 4.17 12.63 -20.59 n.d. 56 5 7.20 4.10 12.68 -20.29 n.d. C AH , C AH , 84 5 7.11 4.77 12.65 -20.31 3 6 2 8 C4AcH11 28 25 6.91 4.24 11.91 (11.981)) -20.52 n.d. 56 25 7.29 4.80 11.92 (11.902)) -20.34 n.d. 3) 84 25 6.31 5.07 11.81 (11.79 ) -20.84 C3AH6 28 55 6.66 4.20 11.01 -20.91 n.d. 56 55 6.87 4.59 11.01 -20.80 n.d.

84 55 7.01 4.90 11.01 -20.73 C3AH6 56 70 6.34 4.44 10.60 -21.25 n.d.

84 70 6.39 4.24 10.62 -21.22 C3AH6 28 85 5.73 3.64 10.27 -21.80 n.d. 56 85 6.18 4.39 10.27 -21.55 n.d.

84 85 6.19 4.60 10.26 -21.55 C3AH6 28 105 5.64 3.44 9.89 -22.22 n.d. 56 105 5.70 3.76 9.88 -22.17 n.d.

84 105 5.81 4.44 9.85 -22.11 C3AH6

C3AH6 (from undersaturation) 28 5 6.79 4.63 12.62 -20.47 n.d. 56 5 7.03 5.07 12.64 -20.36 n.d.

84 5 7.01 4.40 12.65 -20.37 C3AH6, C2AH8 28 25 7.20 4.73 11.91 (11.961)) -20.38 n.d. 56 25 7.34 4.55 11.93 (11.902)) -20.32 n.d. 3) 84 25 7.31 3.94 11.95 (11.91 ) -20.36 C3AH6 28 55 6.93 4.33 11.03 -20.78 n.d. 56 55 6.84 4.04 11.03 -20.83 n.d.

84 55 7.16 4.44 11.04 -20.68 C3AH6 56 70 6.39 3.97 10.63 -21.23 n.d.

84 70 6.31 4.37 10.61 -21.26 C3AH6 28 85 6.15 4.14 10.28 -21.57 n.d. 56 85 6.13 3.88 10.29 -21.59 n.d.

84 85 6.11 4.27 10.27 -21.59 C3AH6 28 105 5.57 3.67 9.87 -22.24 n.d. 56 105 5.63 4.46 9.83 -22.21 n.d.

84 105 5.66 4.07 9.86 -22.19 C3AH6 1) temperature of solution 24°C at time of measurement 2) temperature of solution 26°C at time of measurement 3) temperature of solution 25°C at time of measurement

Appendix V

Table A.3: Solubility data of siliceous hydrogarnet (C3AS0.8H4.4) at different temperatures 5°C - 85°C

1) C3AS0 8H4 4 (from undersaturation )

Temp. Ca Al Si Na pH Log_Ksp Phases [°C] [mmol/l] [mmol/l] [mmol/l] [mmol/l] 5 0.46 0.29 0.10 0.83 11.84 -30.05 n.d. 5 0.52 0.23 0.08 0.35 11.75 -30.42 n.d. 3) 5 0.48 0.22 0.06 0.20 11.65 -30.90 C3AS0 8H4 4 , C-S-H 25 0.56 0.26 0.11 0.87 11.19 (11.132)) -29.58 n.d. 25 0.55 0.24 0.10 0.30 11.00 (10.942)) -30.22 n.d. 2) 3) 25 0.57 0.26 0.08 0.17 10.97 (10.90 ) -30.29 C3AS0 8H4 4 , C-S-H 55 0.72 0.28 0.15 0.96 10.41 -28.94 n.d. 55 0.72 0.29 0.16 0.26 10.23 -29.43 n.d. 3) 55 0.70 0.30 0.17 0.13 10.16 -29.64 C3AS0 8H4 4 , C-S-H 70 0.70 0.33 0.18 0.87 10.02 -28.97 n.d. 70 0.67 0.30 0.18 0.26 9.84 -29.64 n.d. 3) 70 0.65 0.34 0.17 0.10 9.74 -29.90 C3AS0 8H4 4 , C-S-H. 85 0.67 0.30 0.18 0.87 9.72 -29.22 n.d. 85 0.62 0.30 0.19 0.30 9.52 -29.87 n.d. 3) 85 0.60 0.31 0.22 0.16 9.43 -30.14 C3AS0 8H4 4 , C-S-H 1)all samples analysed after 4 week stored at the given temperature. After each analysis the removed solution (~20 ml) was replaced with ultra pure degassed water. 2) temperature of solution 26°C at time of measurement 3) estimated composition

Appendix VI

Table A.4: Solubility data of nominally monosulfoaluminate (C4AsH12) at temperatures 5°C - 100°C

C4AsH12 Ag Temp. Ca Al SO 2- pH Log_K Phases e 4 sp [d] [°C] [mmol/l] [mmol/l] [mmol/l] C AsH , AFt, 42 5 us 4.70 2.92 0.005 12.49 -29.92 4 12 C4Ac0.5H11.5 C AsH , AFt, 42 5 ss 14.6 <0.02 9.300 12.63 n.d. 4 12 C4Ac0.5H11.5, C3AH6 1) 28 25 ss 5.51 2.91 0.009 11.84 (11.90 ) -29.24 C4AsH12, AFt 1) 56 25 ss 5.06 3.13 0.009 11.79 (11.83 ) -29.48 C4AsH12, AFt 1) 84 25 ss 5.14 3.29 0.010 11.77 (11.82 ) -29.43 C4AsH12, AFt 1) 373 25 ss 5.01 3.71 0.012 11.73 (11.77 ) -29.43 C4AsH12, AFt 2) 750 25 ss 4.47 3.26 0.012 11.69 (11.80 ) -29.66 C4AsH12, AFt 3) 28 25 us 4.87 3.93 0.014 11.71 (11.70 ) -29.44 C4AsH12, AFt 1) C AsH , AFt, 56 25 us 4.95 3.42 0.011 11.75 (11.80 ) -29.45 4 12 C4Ac0.5H11.5

56 40 us 3.95 3.31 0.240 11.10 -29.26 C4AsH12, AFt 4) 42 50 us 4.40 2.84 0.398 10.91 -28.84 C4AsH12, AFt

42 50 ss 4.49 2.64 0.253 10.97 -28.86 C4AsH12, AFt 56 65 us 4.09 3.13 0.520 10.44 -29.29 n.d. 3) 42 70 ss 4.88 2.57 1.264 10.37 -28.79 C4AsH12, AFt

42 70 us 4.37 2.18 1.053 10.36 -29.17 C4AsH12 56 85 us 4.62 2.63 1.310 9.99 -29.33 n.d.

42 90 ss 5.19 2.57 1.788 9.91 -29.12 C4AsH12

42 90 us 5.19 1.87 2.333 9.87 -29.47 C4AsH12 56 100 us 4.45 2.67 1.510 9.62 -29.86 n.d. 1)measured at 24°C, 2)measured at 23°C 3)measured at 24°C 4)very weak signal

Appendix VII

Table A.5: Solubility data of monocarboaluminate(C4AcH11) at temperatures 5°C - 110°C

C4AcH11

2- 1) Age Temp. Ca Al CO3 pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l] [mmol/l]

42 5 us 2.58 1.60 0.008 12.24 -32.28 C4AcH11, Calcite 42 5 us 2.52 1.56 0.008 12.24 -32.36 n. d. 2) 42 5 us 2.35 1.63 0.008 12.18 -32.60 C4AcH11, Calcite 3) 56 25 us 3.19 2.32 0.008 11.56 (11.59 ) -31.66 C4AcH11, Calcite 3) 84 25 ss 3.73 2.16 0.008 11.67 (11.62 ) -31.12 C4AcH11, Calcite

42 50 us 4.42 2.34 0.007 11.01 -30.94 C4AcH11, Calcite 42 50 us 4.46 2.51 0.007 11.00 -30.89 n. d. 2) 42 50 us 4.30 2.40 0.007 10.99 -31.02 C4AcH11, Calcite

42 70 us 5.24 3.30 0.007 10.56 -30.78 C4AcH11, Calcite 42 70 us 5.35 3.42 0.007 10.56 -30.71 n. d. 2) 42 70 us 5.11 2.94 0.007 10.56 -30.88 C4AcH11, Calcite C AcH 4), Calcite, 42 90 us 6.54 5.02 0.006 10.17 n. d. 4 11 C3AH6. C AcH 4), Calcite, 42 110 us 6.05 4.56 0.005 9.78 n. d. 4 11 C3AH6. 1) 2) calculated values assuming equilibrium with calcite prepared from a stoichiometric mixture of Al(OH)3, CaCO3 and CaO in 0.1 mol KOH-solution at 50°C 3)measured at 25°C 4) weak signals

Appendix VIII

Table A.6: Solubility data of nominally hemicarboaluminate(C4Ac0.5H12) at temperatures 5°C - 85°C

C4Ac0 5H12

2- 1) Age Temp. Ca Al CO3 pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l] [mmol/l] C Ac H , C AcH , 28 5 ss 8.01 1.19 2.1E-06 12.84 -29.67 4 0.5 11.5 4 11 C3AH6 C Ac H , C AcH , 56 5 us 7.21 0.81 8.2E-06 12.80 -30.01 4 0.5 11.5 4 11 C3AH6. C Ac H , C AcH , 28 25 ss 13.08 0.12 1.5E-04 12.32 (12.312)) -29.35 4 0.5 11.5 4 11 C3AH6 C Ac H , C AcH , 56 25 us 11.43 0.40 3.3E-05 12.26 (12.242)) -29.08 4 0.5 11.5 4 11 C3AH6 C Ac H , C AcH , 28 25 ss 11.32 0.48 2.1E-05 12.25 (12.272)) -29.00 4 0.5 11.5 4 11 C3AH6

28 25 us 10.48 0.49 3.4E-05 12.22 (12.222)) -29.14 n.d.

3) 365 25 us 9.55 0.80 2.3E-05 12.18 (12.30 ) -29.12 C4Ac0.5H11.5, C3AH6

28 40 ss 15.02 0.20 1.4E-04 11.90 -28.82 n.d.

C Ac H , C AcH , 56 40 us 12.81 0.41 8.4E-05 11.83 -28.79 4 0.5 11.5 4 11 C3AH6 C AcH , C AH 28 50 ss 16.03 0.31 1.2E-04 11.64 n.d. 4 11 3 6, Ca(OH)2

56 50 us 17.02 0.30 9.2E-05 11.67 n.d. C4AcH11, C3AH6

28 70 ss 10.31 0.62 1.3E-03 10.97 n. d. n.d.

56 70 us 12.39 0.45 1.6E-03 10.97 n. d. C4AcH11, C3AH6

C AcH , C AH 28 85 ss 12.02 0.53 3.1E-03 10.71 n. d. 4 11 3 6, Ca(OH)2

56 85 us 10.53 0.62 1.1E-03 11.04 n. d. n.d.

84 85 us 10.80 0.53 5.1E-03 10.67 n. d. C4AcH11, C3AH6

1)calculated values assuming equilibrium with monocarboaluminate 2)measured at 25°C 3)measured at 23°C

Appendix IX

Table A.7: Solubility data of nominally hydroxy-AFm (C4AH13) at 25°C

C4AH13 (from supersaturation)

Age Temp. Ca Al pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l]

28 25 17.64 0.21 12.44(12.40) -25.60 C4AH13, C3AH6

56 25 20.11 0.20 12.49(12.46) -25.21 C4AH13, C3AH6 C AH Ca(OH) , 373 25 19.91 0.17 12.48(12.43) (-25.38) 3 6, 2 C4AH13 )trces C AH Ca(OH) , 750 25 19.96 0.09 12.49(12.49) (-25.93) 3 6, 2 C4AH13 )trces

Appendix X

Table A.8: Solubility data of strätlingite (C2ASH8) at temperatures 5°C - 85°C

1) C2ASH8 (samples from undersaturation )

Age Temp. Ca Al Si Na pH Log_Ksp Phases [°C] [mmol/l] [mmol/l] [mmol/l] [mmol/l] 28 5 1.78 0.47 0.05 <0.2 12.18 -19.89 n.d. 561) 5 1.38 0.28 0.06 <0.2 12.08 -20.46 n.d. 1) 84 5 1.44 0.17 0.05 <0.2 12.12 -20.92 C2ASH8, C-S-H 28 5 1.68 0.40 0.05 2.35 12.41 -20.08 n.d. 56 5 1.95 0.44 0.04 2.80 12.49 -20.03 n.d. C ASH C-S-H, 841) 5 1.32 0.31 0.06 0.91 12.20 -20.35 2 8, Calcite 28 25 1.75 0.32 0.07 <0.2 11.45(11.58) 2) -19.82 n.d. 28 25 1.75 0.33 0.07 <0.2 11.45(11.54) 2) -19.82 n.d. 2) 28 25 1.21 0.35 0.09 <0.2 11.26(11.35) -20.00 C2ASH8, C-S-H 561) 25 1.55 0.24 0.09 <0.2 11.40(11.40) 2) -20.05 n.d. 561) 25 1.58 0.24 0.09 <0.2 11.41(11.37) 2) -20.06 n.d. 1) 2) 56 25 1.75 0.23 0.10 <0.2 11.46(11.42) -19.95 C2ASH8, C-S-H 28 25 1.66 0.50 0.05 2.17 11.65(11.70) 3) -19.64 n.d. 56 25 1.94 0.45 0.05 2.60 11.73(11.75) 3) -19.61 n.d. C ASH C-S-H, 841) 25 1.23 0.33 0.07 0.90 11.41(11.48) 2) -20.14 2 8, Calcite 28 40 2.04 1.00 0.05 2.17 11.21 -18.91 n.d. 56 40 2.03 1.00 0.04 2.60 11.24 -18.99 n.d. C ASH C-S-H, 841) 40 1.42 0.53 0.08 1.05 11.02 -19.57 2 8, Calcite 28 50 1.93 0.50 0.09 <0.2 10.74 -19.23 n.d. 1) 56 50 2.14 0.43 0.11 <0.2 10.79 -19.14 C2ASH8, C-S-H 28 50 1.95 0.79 0.05 2.26 10.95 -19.13 n.d. 56 50 1.90 0.76 0.05 2.40 10.95 -19.13 n.d. C ASH C-S-H, 841) 50 1.49 0.49 0.08 1.14 10.78 -19.53 2 8, Calcite 28 70 1.98 1.56 0.04 2.28 10.13 -18.76 n.d. 56 70 2.03 1.60 0.04 2.20 10.14 -18.74 n.d. C ASH C-S-H, 841) 70 1.77 1.22 0.06 1.50 10.12 -18.87 2 8, Calcite 28 85 1.95 1.01 0.07 <0.2 9.89 -18.89 n.d. 561) 85 1.80 0.71 0.11 <0.2 9.89 -19.06 n.d. 28 85 2.09 1.50 0.04 2.30 10.13 -18.79 n.d. 56 85 2.07 1.55 0.03 2.30 10.12 -18.87 n.d. C ASH C-S-H, 841) 85 1.75 1.08 0.07 1.50 10.03 -19.02 2 8, Calcite 1) solids filtered and redissolved after previous extraction 2)measured values at 23°C 3) measured values at 24°C

Appendix XI

Table A.9: Solubility data of CO3-AFt at 25°C

C6Ac3H32

2- 1) 2) Age Temp. Ca Al CO3 pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l] [mmol/l]

35 25 us 6.04 3.59 0.007 11.86(11.93) -46.43 CO3-AFt, Calcite

70 25 us 6.65 3.94 0.007 11.90(11.83) -46.10 CO3-AFt, Calcite.

70 25 us 6.16 3.49 0.007 11.88(11.83) -46.37 CO3-AFt, Calcite

70 25 us 5.59 3.23 0.007 11.83(11.81) -46.70 n.d.

70 25 us 5.69 3.21 0.007 11.84(11.78) -46.64 n.d.

70 25 us 5.81 3.15 0.007 11.86(11.83) -46.58 n.d.

70 25 us 5.96 3.51 0.007 11.86(11.80) -46.47 n.d.

270 25 us 5.74 3.82 0.007 11.82(11.96) -46.59 CO3-AFt, Calcite

270 25 us 5.49 3.59 0.007 11.80(11.90) -46.74 CO3-AFt, Calcite

3) Recalculated values for the metastable dissolution of CO3-AFt

- 25 Calc. 5.90 3.93 0.007 11.82 -46.50 CO3-AFt, Calcite 1)calculated values assuming equilibrium with calcite 2)measured values at 25°C in parethesis3)metastability with respect to carboaluminates suppressed

Appendix XII

Table A.10: Solubility data of monosulfoaluminate-hydroxy-AFm solid solutions at 25°C (early age values)

C4AsxH14-2x

values from supersaturation

2- 1) Age x Ca Al SO4 pH Phases

[d] [°C] [mmol/l] [mmol/l] [mmol/l] 28 0.98 5.15 2.40 1.0E-02 11.83 (n.d.) ms-type (ss), AFt

28 0.95 6.69 2.30 3.0E-03 11.97 (n.d.) ms-type (ss)

28 0.90 8.28 1.01 3.6E-03 12.11 (12.12) ms-type (ss) 56 0.90 7.58 1.51 2.8E-03 12.06 (12.05) ms-type (ss) 28 0.90 7.36 3.30 1.6E-03 11.98 (12.05) n.d 28 0.85 7.72 3.50 1.0E-03 12.00 (n.d.) ms-type (ss) 28 0.80 8.36 2.51 1.7E-042) 12.07 (12.04) ms-type (ss) 56 0.80 8.56 2.73 1.7E-042) 12.08 (12.03) ms-type (ss) 28 0.80 8.31 2.22 1.7E-042) 12.08 (12.14) n.d 2) 28 0.70 9.73 1.60 9.3E-05 12.17 (12.16) ms-type (ss), C2AH8 2) 56 0.70 9.68 2.04 9.3E-05 12.15 (12.14) ms-type (ss), C2AH8 28 0.70 9.47 2.86 9.3E-052) 12.12 (12.16) n.d. 2) ) 28 0.60 10.73 1.56 5.7E-05 12.21 (12.18) ms-type (ss), C2AH8 2) 56 0.60 10.70 2.10 5.7E-05 12.20 (12.18) ms-type (ss), C2AH8 28 0.60 10.56 1.99 5.7E-052) 12.19 (12.25) n.d. 2) 28 0.50 10.85 1.84 4.0E-05 12.21 (12.18) ms-type (ss), C2AH8 2) 56 0.50 10.95 1.77 4.0E-05 12.22 (12.23) ms-type (ss), C2AH8 28 0.50 11.01 1.82 4.0E-052) 12.22 (12.25) n.d. ms-type (ss), hydroxy- 28 0.40 11.03 1.84 3.7E-052) 12.22 (12.19) AFm-type(ss), C2AH8 56 0.40 11.18 1.89 3.7E-052) 12.22 (12.21) ms-type (ss), hydroxy- AFm-type(ss), C2AH8 28 0.40 10.92 1.83 3.7E-052) 12.21 (12.26) n.d. ms-type (ss), hydroxy- 28 0.20 11.00 1.53 3.7E-052) 12.22 (12.22) AFm-type(ss), C2AH8 56 0.20 11.08 1.75 3.7E-052) 12.22 (12.25) ms-type (ss), hydroxy- AFm-type(ss), C2AH8 28 0.20 11.18 1.74 3.7E-052) 12.23 (12.26) n.d. 28 0.05 11.15 1.84 3.7E-052) 12.22 (12.24) hydroxy-AFm-type(ss) values from mixed end members 28 0.98 4.78 3.48 1.4E-02 11.70 (11.68) ms-type (ss), AFt 28 0.95 4.71 2.99 1.3E-02 11.72 (11.71) ms-type (ss), AFt 28 0.90 6.61 3.54 1.4E-02 11.75 (11.75) ms-type (ss) 28 0.85 7.15 3.90 2.6E-03 11.91 (11.85) ms-type (ss) 2) 28 0.80 7.82 4.05 1.7E-04 11.94 (11.87) ms-type (ss) 2) 28 0.70 8.86 3.34 9.3E-05 11.99 n.d. ms-type (ss) 2) 28 0.60 9.63 2.57 5.7E-05 12.08 n.d. ms-type (ss), hemicarb. 2) 28 0.50 10.19 2.04 4.0E-05 12.14 n.d. ms-type (ss), hemicarb. 2) 28 0.40 10.43 1.80 3.7E-05 12.18 n.d ms-type (ss), hemicarb. 2) 28 0.20 10.53 1.78 3.7E-05 12.19 n.d. ms-type (ss), hemicarb. 2) 28 0.05 15.12 0.39 3.7E-05 12.20 n.d hemicarb., C3AH6

Appendix XIII

(Table A.10 continued)

values from undersaturation 3) 2- Age x Ca Al SO4 pH Phases [d] [°C] [mmol/l] [mmol/l] [mmol/l] 28 0.98 4.97 3.43 1.2E-02 11.75 (11.75) ms-type (ss), AFt 28 0.95 5.09 3.17 1.2E-02 11.78 (11.76) ms-type (ss), AFt 2) 28 0.80 7.49 3.78 1.7E-04 11.97 (11.98) ms-type (ss) 2) 28 0.70 8.28 4.04 9.3E-05 12.02 (12.01) ms-type (ss) 2) 28 0.60 9.04 3.32 5.7E-05 12.09 (12.08) ms-type (ss), hemicarb. 2) 28 0.50 9.96 2.68 4.0E-05 12.15 (12.15) ms-type (ss), hemicarb. 2) 28 0.40 11.00 2.19 3.7E-05 12.21 (12.20) ms-type (ss), hemicarb. 2) 28 0.20 11.30 1.83 3.7E-05 12.23 (12.22) ms-type (ss), hemicarb. 2) 28 0.05 12.29 1.35 3.7E-05 12.27 (12.30) hemicarb., C3AH6 1)calculated values, measured values given in parenthesis 2)calculated values; experimental values below limit of 3) detection <~2.0E-3 mmol SO4/l time of experiment; original samples from supersaturation experiments aged 56 d

Appendix XIV

Table A.11: Solubility data of monosulfoaluminate-hydroxy-AFm solid solutions at 25°C (aged samples)

C4AsxH14-2x

values from supersaturation

2- 1) Age x Ca Al SO4 pH Phases

[d] [°C] [mmol/l] [mmol/l] [mmol/l] 373 0.98 4.62 2.87 1.2E-02 11.74 (11.82.) n.d.

373 0.95 4.97 2.15 1.2E-02 11.83 (11.87.) n.d.

373r 0.90 6.44 3.13 2.1E-03 11.92 (11.93) n.d. 373 0.80 8.16 4.05 2.1E-03 12.01 (12.02) n.d. 373 0.70 9.93 1.51 <2.0E-3 12.18 (12.12) n.d. 373 0.60 12.72 0.74 <2.0E-3 12.30 (12.23) n.d. 373 0.50 15.89 0.31 <2.0E-3 12.40 (12.38) n.d. 373 0.40 16.07 0.30 <2.0E-3 12.40 (12.40) n.d. 373 0.20 16.22 0.25 <2.0E-3 12.40 (12.40) n.d. 373 0.05 16.29 0.36 <2.0E-3) 12.41 (12.39) n.d.

750 0.95 4.67 1.85 2.1E-02 11.81 (11.84) ms-type (ss) 750 0.90 5.86 3.19 8.5E-03 11.86 (11.95) ms-type (ss) 750 0.80 7.36 4.08 4.2E-03 11.95 (12.01) ms-type (ss)

750 0.70 10.33 1.18 <2.0E-3 12.20 (12.15) ms-type (ss), C3AH6

750 0.60 13.47 0.42 <2.0E-3 12.33 (12.26) ms-type (ss), C3AH6

750 0.50 17.34 0.13 <2.0E-3 12.43 (12.36) ms-type (ss), C3AH6 ms-type (ss), hemicarb, 750 0.40 <2.0E-3 12.47 (12.43) 19.24 0.06 C3AH6 ms-type (ss), hydroxy- 750 0.20 <2.0E-3 12.47 (12.43) 19.09 0.07 AFm-type(ss), C3AH6 750 0.05 <2.0E-3 12.47 (12.41) hydroxy-AFm-type(ss), 19.19 0.15 C3AH6 1)calculated values, measured values given in parenthesis

Appendix XV

Table A.12: Solubility data of SO4-CO3-AFt solid solutions at 25°C

C6As3xc3-3xH32

2- 2- 2) 1) Age x Ca Al SO4 CO3 pH Phases [d] [mmol/l] [mmol/l] [mmol/l] [mmol/l] 28 1 2.00 0.60 1.460 --- 10.90(11.01) n.d.

35 1 2.20 0.64 1.500 --- 10.85(10.90) n.d.

84 1 1.93 0.58 1.174 --- 10.93 SO4-AFt

35 0.95 2.96 1.12 0.360 0.008 11.56(11.26) n.d.

84 0.95 1.95 0.65 0.974 0.010 11.07 SO4-AFt-type, calcitetraces

28 0.9 3.00 2.10 0.030 0.008 11.53(11.66) n.d.s

35 0.9 3.04 2.33 0.043 0.008 11.51(11.54) SO4-AFt-type, calcitetraces

28 0.8 5.20 3.46 0.004 0.007 11.78(11.87) n.d.

35 0.8 4.59 3.72 0.006 0.007 11.68(11.75) n.d.

84 0.8 2.80 1.88 0.070 0.008 11.51 SO4-AFt-type, calcite

84 0.8 2.72 1.70 0.070 0.008 11.51 SO4-AFt-type, calcite

35 0.7 4.99 3.75 0.005 0.007 11.73(11.77) n.d.s

28 0.6 6.00 3.20 0.003 0.007 11.88(11.90) n.d.

35 0.6 5.09 3.83 0.006 0.007 11.74(11.83) n.d.

84 0.6 5.14 3.39 0.005 0.007 11.77 SO4-AFt-type, calcite

84 0.6 5.03 3.21 0.009 0.007 11.77 SO4-AFt-type, calcite

28 0.5 6.15 3.50 0.003 0.007 11.87(11.92) n.d.

35 0.5 5.24 3.83 0.005 0.007 11.76(11.83) n.d. SO -AFt-type, CO -AFt-type , 84 0.5 5.59 3.23 0.003 0.007 11.83 4 3 traces calcite SO -AFt-type, CO -AFt-type , 84 0.5 5.61 3.44 0.004 0.007 11.82 4 3 traces calcite 28 0.4 6.10 3.40 0.002 0.007 11.88(11.93) n.d

35 0.4 5.09 3.83 0.006 0.007 11.74(11.83) n.d. SO -AFt-type, CO -AFt-type , 84 0.4 5.69 3.21 0.004 0.007 11.84 4 3 traces calcite SO -AFt-type, CO -AFt-type , 84 0.4 5.72 3.51 0.003 0.007 11.83 4 3 traces calcite 28 0.3 6.30 3.20 0.002 0.007 11.90(11.93) n.d. SO -AFt-type, CO -AFt-type , 84 0.3 5.81 3.15 0.003 0.007 11.86 4 3 traces calcite 35 0.2 5.31 3.33 0.004 0.007 11.80(11.87) n.d. SO -AFt-type, CO -AFt-type , 84 0.2 6.18 3.43 0.002 0.007 11.88 4 3 traces calcite 28 0.05 6.50 3.20 0.002 0.007 11.92(11.96) n.d.

35 0.05 5.31 3.50 0.004 0.007 11.79(11.88) n.d.

84 0.05 5.96 3.51 0.002 0.007 11.86 CO3-AFt-typetraces, calcite 1)calculated values, measured values given in parenthesis 2) calculated values assuming equilibrium with calcite

Appendix XVI

Table A.13: Solubility data of thaumasite from undersaturation at different temperatures 1°C - 70°C

Synthetic thaumasite (C3SscH15)

1) 2- 2- 2) Age Temp. Ca Si SO4 CO3 Na pH Log_Ksp Phases [d] [°C] [mmol/l] [mmol/l] [mmol/l] [mmol/l] [mmol/l] C SscH , 28 1 0.62 0.69 0.52 0.056 0.10 9.94 -27.09 3 15 Calcite traces

56 1 0.65 0.52 0.40 0.020 0.02 10.76 -26.10 n.d.

C SscH , 84 1 0.70 0.54 0.45 0.019 0.03 10.76 -26.00 3 15 Calcite traces C SscH , 28 5 0.80 0.74 0.62 0.029 0.10 10.20 -26.19 3 15 Calcite traces

56 5 0.67 0.50 0.38 0.018 0.02 10.90 -25.79 n.d.

C SscH , 84 5 0.70 0.53 0.39 0.017 0.03 10.91 -25.70 3 15 Calcite traces C SscH , 28 25 1.07 0.65 0.54 0.013 0.07 10.67(10.50)3) -24.78 3 15 Calcite traces

28 25 1.12 0.66 0.54 0.013 0.07 10.73(10.57)3) -24.69 n.d.

56 25 1.07 0.62 0.50 0.013 0.02 10.71(10.61)3) -24.79 n.d.

56 25 1.12 0.64 0.48 0.013 0.02 10.78(10.61)3) -24.69 n.d.

C SscH , 84 25 1.02 0.64 0.51 0.014 0.03 10.62(10.79)3) -24.89 3 15 Calcite traces C SscH , 84 25 1.07 0.66 0.49 0.013 0.03 10.70(10.88)3) -24.78 3 15 Calcite traces C SscH , 28 40 1.87 0.91 0.85 0.010 0.07 10.57 -23.90 3 15 Calcite traces

56 40 1.90 0.91 0.76 0.010 0.03 10.62 -23.88 n.d.

C SscH , 84 40 1.97 0.95 0.80 0.010 0.03 10.64 -23.80 3 15 Calcite traces

C3SscH15, 28 55 8.93 0.10 7.40 0.007 0.08 10.52 -23.14 Calcite, C-S-H 56 55 6.54 0.11 5.47 0.008 0.03 10.36 -23.47 n.d.

C3SscH15, 84 55 6.49 0.11 5.56 0.008 0.03 10.31 -23.52 Calcite, C-S-H Calcite, 28 70 13.75 0.07 11.04 0.007 0.07 10.38 n.d. C-S-H

56 70 5.54 0.12 4.57 0.008 0.04 9.97 n.d. n.d.

Calcite, 84 70 5.29 0.13 4.21 0.008 0.04 10.02 n.d. C-S-H

natural thaumasite (C3SscH15)

C SscH , 180 5 0.92 0.71 0.61 0.018 ~0.01 10.57 -25.61 3 15 Calcite traces C SscH , 180 25 1.32 1.00 0.96 0.020 ~0.01 9.94(9.68)3) -25.10 3 15 Calcite traces 1) solids filtered and redissolved after previous extraction 2) calculated values assuming equilibrium with calcite 3)calculated values, measured values at 25°C given in parenthesis

Appendix XXVII

C Additional details AFm experiments

Table C.1: composition of mixtures between monosulfoaluminate and hydroxy-AFm (supersaturation)

OH-SO4-AFm

SO4/(SO4+2OH) C3A CaSO4 CaO H2O [g] 1 2 1.0077 0.0000 125 0.98 2 0.9876 0.0083 125 0.95 2 0.9573 0.0208 125 0.9 2 0.8566 0.0623 125 0.85 2 0.9069 0.0415 125 0.8 2 0.8062 0.0830 125 0.7 2 0.7054 0.1245 125 0.6 2 0.6046 0.1660 125 0.5 2 0.5039 0.2075 125 0.4 2 0.4031 0.2491 125 0.2 2 0.2015 0.3321 125 0.05 2 0.0504 0.3943 125 0 2 0.0000 0.4151 125

Table C.2: composition of mixtures between monosulfoaluminate and hydroxy-AFm (mixed end members)

OH-SO4-AFm

SO4/(SO4+2OH) monosulfoaluminate hydroxy-AFm H2O [g] 1 2 0 60 0.98 1.4 0.0257 60 0.95 2 0.0948 60 0.9 1.8 0.1801 60 0.85 1.8 0.2860 60 0.8 1.6 0.3601 60 0.7 1.4 0.5402 60 0.6 1.2 0.7203 60 0.5 1 0.9003 60 0.4 0.8 1.0804 60 0.2 0.4 1.4405 60 0.05 0.1 1.7106 60 0 0 2.0000 60

Appendix XXVIII

Table C.3: composition of mixtures between monocarboaluminate and hydroxy-AFm (supersaturation)

OH-CO3-AFm

CO3/(CO3+2OH) C3A CaCO3 CaO [g] 1 2 0.7409 0.0000 0.8 2 0.5927 0.0830 0.5 2 0.3704 0.2075 0.2 2 0.1482 0.3321 0 2 0.0000 0.4151

Table C.4: composition of mixtures between monocarboaluminate and hydroxy-AFm (mixed end members)

OH-CO3-AFm

CO3/(CO3+2OH) monocarboaluminate hydroxy-AFm [g] 1 2.0 0.0000 0.8 1.6 0.3944 0.5 1.0 0.9860 0.2 0.4 1.5775 0 0.0 2.0000

Table C.5: composition of mixtures between monocarboaluminate and monosulfoaluminate (mixed end members)

SO4-CO3-AFm

CO3/(CO3+SO4) monocarboaluminate monosulfoaluminate [g] 1 2.0 0.0000 0.5 1.0 1.0951 0 0.0 2.0000