Fluoride Mineralization of Portland cement

Thuan T. Tran

FACULTY OF SCIENCE AARHUS UNIVERSITY

Fluoride Mineralization of Portland cement Fluoride Mineralization of Portland cement

Applications of Double-Resonance NMR Spectroscopy in Structural Investigations of Guest Ions in Cement Phases PhD Thesis

Front cover illustration. Stacked plot of a 2D 29Si{19F} HETCOR NMR spectrum for a synthetic sample of cuspidine (Ca4Si2O7F2).

THUAN T. TRAN PhD Thesis 2011

Interdisciplinary Nanoscience Center, 2011

Thuan Thai Tran PhD Thesis Instrument Centre for Solid-State NMR Spectroscopy Interdisciplinary Nanoscience Center (iNANO) Department of Chemistry, Faculty of Science and Technology Aarhus University, Denmark August 2011

Preface This thesis presents the results obtained during my PhD study from August 2007 to July 2011 at the Instrument Centre for Solid-State NMR Spectroscopy, Department of Chemistry, and the Interdisciplinary Nanoscience Center (iNANO) at Aarhus University. The PhD study has been a part of the FUTURECEM project (2007  2010), which is a joint collaboration between Aalborg Portland A/S, iNANO (Aarhus and Aalborg Universities) and the Geological Survey of Denmark and Greenland (GEUS), Copenhagen, with partly financial support from the Danish National Advanced Technology Foundation (“Højteknologifonden”).

Acknowledgments It is a pleasure for me to express my gratitude to all the people who have contributed with valuable assistances to my PhD project. First and foremost, I would like to thank my supervisor Associate Professor Jørgen Skibsted for his excellent support and scientific guidance, ever since I joined the Instrument Centre for Solid-State NMR Spectroscopy as a young, undergraduate student for almost eight years ago. He introduced me to solid-state NMR as well as cement research and gave me the opportunity to extend my studies at the Instrument Centre. His ongoing fruitful discussions on both subjects have been a great source of inspiration to this PhD project. Professor Hans Jørgen Jakobsen and Associate Professor Henrik Bildsøe are acknowledged for their general interest in the project and several fruitful discussions about experimental as well as theoretical solid-state NMR problems. They are also thanked for being available when there have been problems with the spectrometers or the computers. Mrs. Rigmor Søeberg Johansen and Mrs. Anne Birgitte Bundgaard Johannsen are thanked for their valuable assistance in conducting numerous of NMR experiments during my time at the Instrument Center. I also want to acknowledge all the members at the NMR laboratory, including PhD Søren Lundsted Poulsen and Phd students Søren Sørensen, Tine Fly Sevelsted, Maiken Rabøl Jørgensen and Nicklas Bromose Kolman, for creating a good scientific and social atmosphere. I would like to thank the researchers at the R&D, Quality and Technical Sales Support at Aalborg Portland A/S: Project Managers Lise Frank Kirkegaard and Mette Steenberg Glyø and Chief Scientist Duncan Herfort for a good collaboration and several interesting discussions on the cement chemistry. Finally, the Danish National Advanced Technology Foundation and the Faculty of Science and Technology, Aarhus University are acknowledged for their financial support of this PhD project.

Abstract The increasing applications of alternative raw materials, alternative fuels and supplementary cementitious materials (SCMs) in today’s cement production lead to a significant amount of impurities (e.g. fluoride, SO3, P2O5 etc.) in the final cement materials. These so-called minor components may also be added to the raw meal of starting materials to modify the clinkering processes. They introduce several important effects on the formation of the cement clinker phases as well as on their hydrational properties. In general, the structural characterization of the minor components is rather difficult due to their low quantities in cement. This PhD thesis focuses on two main subjects: (i) modification of the Portland clinker composition with the aim of increasing the clinker reactivity, enabling a 30 % replacement of the Portland clinkers by SCMs without a significant reduction in the strength properties of the blended cement, and (ii) development and implementation of solid-state double-resonance Nuclear Magnetic Resonance (NMR) techniques including Cross Polarization (CP), RotationalEcho Double-Resonance (REDOR) and Rotational-Echo Adiabatic-Passage Double-Resonance (REAPDOR) experiments in the structural characterization of cementitious materials. The mineralizing properties of calcium fluoride have been investigated in detail. The mechanism for fluoride mineralization is probed using different solid-state

19

F,

27

Al and

29

Si

NMR techniques. Furthermore, the impacts of Al3+ and Fe3+ ions on the fluoride mineralization and their site preferences in the calcium silicate phases of Portland cement are investigated. It is found that the reactivity of Portland cement and its strength properties may be improved substantially by the addition of fluoride, Al2O3 and Fe2O3 in appropriate amounts. A comprehensive study of the incorporation of fluoride ions in the calcium silicate hydrate (CSH) phases of hydrated Portland cement has been conducted. This work provides important structural information about the fluoride guest ions in the CSH, contributing to a further understanding of the retarding effect of fluoride on the hydration of Portland cement. Furthermore, a new solid-state NMR pulse sequence (Forth and Back Cross Polarization) has been developed during this work. This experiment provides a selective detection of

19

F

resonances that are dipolar coupled to 29Si which is used to investigate 19F29Si connectivities in the CSH structure. Finally, AlOSi connectivities in the open framework structures of alkali-activated materials and strätlingite (2CaO·Al2O3·SiO2·8H2O) have been investigated using

29

Si{27Al}

REAPDOR experiments. These materials are of interest in recent cement research since they have the potential of being alternative binders in concrete.

Resumé Den stigende anvendelse af alternative råmaterialer og brændstoffer i moderne Portland cement produktion tilfører en signifikant mængde af de såkaldte fremmede komponenter, såsom SO3, P2O5 og fluorider, i den fremstillede cement. Disse oxider kan fremme dannelsen af cement-klinker faserne og deres hydrauliske egenskaber, men de kan også have en negativ effekt ved at hæmme dannelsen af alit, som er den vigtigste hydrauliske komponent i Portland cement. Dette Ph.d. projekt har to hovedformål, hvor det første sigter mod en 30 % erstatning af Portland cement-klinker med alternative materialer, som har hydrauliske egenskaber, uden at reducere cementens styrke. Dette opnås bl.a. ved at optimere klinkernes kemiske sammensætning, således at der opnås en forøget reaktivitet for alit-fasen. Det andet mål i projektet har været at udvikle og implementere faststof dobbelt-resonans kerne magnetisk resonans (NMR) eksperimenter til strukturkarakterisering af cement-materialer. I projektet er der bl.a. blevet anvendt Cross Polarization (CP), Rotational-Echo DOuble-Resonance (REDOR) og Rotational-Echo Adiabatic-Passage DOuble-Resonance (REAPDOR) NMR eksperimenter til strukturkarakterisering af gæst ioner i calciumsilikat-faserne (alit og belit) i Portland cement. I studiet af optimeringen af Portland cement-klinkers kemiske sammensætning har der været særlig fokus på mineraliseringsegenskaberne af calciumfluorid. Mineraliseringseffekten af fluorid-ionerne er blevet undersøgt med forskellige faststof

19

F,

27

Al og

29

Si NMR teknikker.

Disse undersøgelser har også inkluderet indflydelsen af trivalente metal-ioner (såsom Al3+ og Fe3+) på flourid-mineraliseringsprocessen og deres strukturelle omgivelser i calciumsilikatfaserne. Hoved-konklusionen er, at tilsætningen af fluorid, Al2O3 og Fe2O3 i passende mængder kan signifikant forøge Portland cements reaktivitet og trykstyrke. PhD afhandlingen inkluderer også et omfattende studium af indbygningen af fluorid-ioner i calcium-silikat-hydrat (CSH) fasen dannet ved hydratisering af Portland cement. Resultaterne herfra bidrager med vigtig strukturel information om fluorid-ionernes placering i CSH fasen og afslører mekanismen for den retarderende effekt af fluorid-ioner på hydratiseringen af Portland cement. I disse studiet blev der udviklet et nyt faststof NMR eksperiment kaldet ”Forth and Back Cross Polarization”. Dette eksperiment blev anvendt til at studere SiF sammenkædninger i CSH stukturen. Til sidst er der foretaget en karakterisering af AlOSi netværkstrukturerne i alkali-aktiverede materialer og i den uordnede struktur for strätlingit (2CaO·Al2O3·SiO2·8H2O) ved hjælp af 29

Si{27Al} REAPDOR eksperimentet. Alkali-aktiverede materialer har i øjeblikket en stor

forskningsmæssig bevågenhed, da de potentielt delvist kan erstatte Portland cement-klinker, primært med henblik på at reducere CO2 emissionen fra cement produktionen.

Table of content

Futurecem

……………………………………………………………………………………..1

Chapter 1:

Cement Chemistry .................................................................................................3

1.1.

Portland cement manufacture ..................................................................................4

1.1.1.

Quarrying and preparing of raw materials...............................................................4

1.1.2.

Clinker production in the rotary kiln .......................................................................5

1.1.3.

Cement grinding ......................................................................................................6

1.2.

The main constituent phases of Portland clinker .....................................................8

1.2.1.

Tricalcium silicate ................................................................................................... 8

1.2.2.

Dicalcium silicate ....................................................................................................9

1.2.3.

Tricalcium aluminate, 3CaOAl2O3 .......................................................................10

1.2.4.

Ferrite, Ca2(AlxFe1-x)O5 .........................................................................................11

1.3.

Minor components .................................................................................................12

1.3.1.

Flux ........................................................................................................................13

1.3.2.

Mineralizers ...........................................................................................................14

1.4.

Hydration chemistry for Portland cement .............................................................14

1.4.1.

Hydration of calcium silicate phases .....................................................................14

1.4.2.

Hydration of tricalcium aluminate.........................................................................15

1.4.3.

Hydration of Portland cement ...............................................................................16

1.5.

Structure of the main hydration product CSH...................................................18

1.5.1.

Tobermorite 14-Å, Ca5Si6O16(OH)27H2O ............................................................19

1.5.2.

Jennite, Ca9Si6O18(OH)68H2O ..............................................................................20

1.5.3.

Portlandite, Ca(OH)2 .............................................................................................22

Chapter 2:

Applications of solid-state NMR in cement research .......................................23

2.1.

NMR theory ...........................................................................................................24

2.1.1.

Chemical shift interaction......................................................................................25

2.1.2.

Quadrupolar coupling interaction ..........................................................................26

2.1.3.

Dipolar coupling interactions ................................................................................27

2.2.

Solid-state NMR techniques in cement research ...................................................29

2.2.1.

29

Si MAS NMR .....................................................................................................29

2.2.2.

27

Al MAS NMR .....................................................................................................32

2.2.3.

19

F MAS NMR ......................................................................................................32

2.2.4.

43

Ca MAS NMR ....................................................................................................34

2.2.5.

Inversion-Recovery (IR) MAS NMR ....................................................................36

2.2.6.

Cross Polarization (CP) .........................................................................................37

2.2.7.

Forth and Back Cross Polarization (FBCP)...........................................................38

2.2.8.

Rotational-Echo Double-Resonance (REDOR) ....................................................41

2.2.9.

Rotational-Echo Adiabatic-Passage Double Resonance (REAPDOR) .................42

2.2.10.

Multiple-Quantum (MQ) MAS NMR ...................................................................46

Chapter 3:

Fluoride Mineralization ......................................................................................49

3.1.

Site preferences of F in the calcium silicate phases of Portland cement .............50

3.1.1.

19

F MAS NMR ......................................................................................................50

3.1.2.

29

Si{19F} and 27Al{19F} CP/MAS NMR ...............................................................51

3.1.3.

29

Si{19F} CP-REDOR NMR..................................................................................53

3.2.

Mineralizing effects of calcium fluoride and calcium sulphate ............................57

3.3.

Secondary effects of fluoride mineralization.........................................................60

3.4.

Incorporation of Fe3+ ions in the calcium silicate phases ......................................65

3.5.

Hydration of fluoride-mineralized Portland cement..............................................71

3.6.

Application of fluoride mineralization ..................................................................76

3.6.1.

Preparation of test clinkers ....................................................................................76

3.6.2.

Strength performances of fluoride-mineralized Portland cement..........................77

3.7.

Summary................................................................................................................79

Chapter 4:

Fluoride-ion environments in CSH................................................................81

4.1.

A brief description of the CSH structure from the T/J viewpoint. ....................82

4.2.

Fluoride-ion environments in synthetic CSH and hydrated Portland cement ...84

4.2.1.

Site preferences of F ions in the CSH structure from 19F MAS NMR ............86

4.2.2.

Influence of fluoride ions on the CSH structure ...............................................89

4.3.

Influence of fluoride ions on the hydration of Portland cement ............................95

4.3.1.

Identification of CaF2 in hydrated Portland cement by 19F MAS NMR ...............95

4.3.2.

Hydration of fluoride-mineralized Portland cement from 19F MAS NMR ...........96

4.3.3.

Retarding mechanism of fluoride ions on the hydration of Portland cement ......101

4.4.

Summary..............................................................................................................103

Chapter 5:

Framework structures of alumino silicates from NMR .................................105

5.1.

SiOAl connectivities in alkali-activated materials ..........................................106

5.1.1.

Effects of Si/Al and Na/Al molar ratios ..............................................................108

5.1.2.

29

5.2.

Disorder in the double tetrahedral layer structure of Strätlingite ........................115

5.2.1.

29

Si MAS NMR ...................................................................................................116

5.2.2.

29

Si{27Al} REAPDOR NMR ...............................................................................117

5.3.

Summary..............................................................................................................119

Si {27Al} REAPDOR NMR ..............................................................................111

Conclusions …………….....………………………………………………………………....120 Reference

………………………………………………………………………………….123

Appendix 1

Sample preparations...………………………………………………………….143

Appendix 2

NMR Measurements and other analytical techniques …………………………146

Paper I

Site Preferences of Fluoride Guest Ions in the Calcium Silicate Phases............151

Paper II

Incorporation of Fluoride Guest Ion in the Calcium Silicate Phases..……...…155

Paper III

Characterization of Guest-Ion Incorporation…………………………..….……163

Paper IV

Characterization of Guest-Ion Incorporation (in russian)……………..…….…163

Paper V

Site Preferences of F, Al3+ and Fe3+ Guest Ions in the Calcium Silicate……..173

Paper VI

Characterization of the network structure of alkali-activated aluminosilicate..181

Manuscript I

Characterization of the aluminosilicate network………………………………..191

Manuscript II The distordered structure of Strätlingite.........………………………………..225 Manuscript III Fluoride mineralization of Portland cement…………………………………..…251

Introduction

1

FUTURECEM Sustainable cement for the future Cement is the essential glue in concrete, which is today’s fundamental building material with an annual consumption only surpassed by water. The cement manufacturing process is however associated with a severe CO2 emission. With an annual worldwide cement production of about 3.3 billion tonnes in 2010[1], the cement industry is responsible for about 5 % of the man-made CO2 emissions[2-4]. In fact, according to the increasing demand for concrete, as growth and modernization take place in the developing countries, the cement production is forecasted to double by the year 2050. Thus, identifying technology to reduce the CO2 emission intensity from cement production is urgent. The Getting the Number Right (GNR) Database System[5], which has registered data from 844 cement installations worldwide from 1999 to 2008, reveals that about 60 % of the direct CO2 emission originates from the chemical reactions during cement clinker formation, e.g. the calcination process, CaCO3  CaO + CO2. The fuel combustion to reach temperatures of 1450 C in the cement kilns, necessarily for the formation of clinker phases, is responsible for the remaining 40 %. Indirect emissions (e.g., from electric power consumption) only contribute about 10 % to the overall CO2 emissions. Based on current knowledge, the International Energy Agency (IEA) and the World Business Council for Sustainable Development (WBCSD) Cement Sustainability Initiative (CSI) have recently presented the Cement Technology Roadmap 2009[6], focusing on four distinct reduction levers available in cement production: (i) thermal and electric efficiency, (ii) alternative fuel use, (iii) clinker substitution and (iv) carbon capture and storage (CCS). The roadmap also emphasizes that none of the options alone can yield the necessary reduction in CO2 emissions. However, from an economical viewpoint, option (i) is less attractive since an optimization of the energy efficiency usually requires installation of new plants or upgrading of the old plants. CCS is a relatively new technology and it has not yet been utilized at the industrial scale in cement production. Thus, most of the ongoing developments and innovations focus on the use of alternative fuels and clinker substitutions by other minerals with hydraulic and/or pozzolanic properties[7,8]; those are usually referred to as Supplementary Cementitious Materials (SCMs). However, the use of SCMs is complicated by several factors, e.g. high water demand, poor workability, retention and a significant reduction in the early strength of the blended cement. Therefore, the clinker substitution is typically limited to an average clinker factor of 0.78, i.e., 22 % of the clinker is replaced by SCMs including gypsum[6].

Introduction

2

The FUTURECEM project was established in 2007 for a four year period (2007 - 2010) with the main purpose to develop new building materials produced from readily available Danish raw materials of low cost and low energy consumption, which should have the same performance as today’s concrete materials. The project is a joint collaboration between Aalborg Portland A/S, iNANO at the Aarhus and Aalborg University, and the Geological Survey of Denmark and Greenland (GEUS), with partly financial support from the Danish National Advanced Technology Foundation. The first principal goal is to obtain a 30 % replacement of the clinkers in concrete production by SCMs, resulting in an approx. 30 % reduction of the total CO2 emission. On a longer time scale, the FUTURECEM project also targets a larger replacement of clinkers utilizing geopolymeric materials, such as alkali-activated metakaolin, which possibly may result in a 70-80 % CO2 reduction. The research program for this PhD project is a part of FUTURECEM and focuses on modifications of the Portland cement clinker composition with the aim of increasing the clinker reactivity, which may enable a higher degree of clinker replacement. Particularly, the mineralizing effects of calcium fluoride, which is widely used in today’s Portland clinker production, are investigated. The presence of fluoride ions has shown several important impacts on the formation of the cement clinker phases as well as their hydrational reactivities[9,10]. The optimization of the mineralizing properties for fluoride, which seems to be dependent on the chemical composition of the clinkers, requires knowledge about the location of the fluoride ions and their potential couplings to other guest ions in the anhydrous as well as hydrated phases of Portland cement. Thus, another major goal of this project is to investigate the feasibility of solidstate Nuclear Magnetic Resonance (NMR) spectroscopy in obtaining such information for cementitious materials, which otherwise is difficult to be measured by other techniques. The present dissertation consists of five chapters. The first two chapters give a brief introduction to the relevant cement chemistry and the basic theory for NMR spectroscopy, respectively. Chapter 3 is devoted to the investigation of fluoride mineralization of Portland clinker while Chapter 4 concerns the influence of fluoride ions on the hydration of Portland cement and their site preferences in the calcium silicate hydrate phases. Chapter 5 demonstrates the potential applications of solid-state NMR in structural investigations of alumina-rich cementitious systems such as alkali-activated materials and strätlingite. The dissertation includes applications of NMR techniques such as Cross Polarization (CP), Rotational-Echo DOubleResonance (REDOR) and Rotational-Echo Adiabatic-Passage DOuble-Resonance (REAPDOR) on different nuclear-spin isotopes, including

19

F,

27

Al,

structural information for the studied cement phases.

29

Si and

43

Ca, to obtain complementary

Chapter 1. Cement Chemistry

3

1. Chapter

Cement Chemistry

Portland cement was patented by Joseph Aspdin in 1824[11]. The name refers to the similarity in the color and strength properties of the cement and Portland stone, i.e. a limestone quarried in Dorset, south west England. However, the mineralogy and hydrational properties of Aspdin’s cement differ significantly from today’s Portland cement. Nowadays, several cement types are produced throughout the world for a wide range of applications, although the majority has been developed for general constructional use[12]. Cements are named in accordance with standard specifications, which define the chemical composition, physical properties and performance requirements for the cement. The standard specifications also set the upper limits for the content of each of the minor components. For example, the upper limit for the MgO content in Portland cement is typically set to 5 %w because MgO levels above this limit result in uncombined MgO (i.e. MgO that is not incorporated in the clinker phases), which can cause destructive expansion in hardened concrete[13]. In fact, the standard specifications used by countries around the world can be slightly different and therefore, various names may also be given to the same class of cements[14]. For example, “Portland cement” in current European and British standards corresponds to “Types I and II Portland cement” in the American standard, i.e American Society for Testing and Materials (ASTM).

Chapter 1. Cement Chemistry 1.1.

4

Portland cement manufacture Portland cement is generally made from calcareous materials such as limestone,

marlstone or chalk. Furthermore, argillaceous materials including sand, clays or bauxite and iron ores may be needed to provide additional SiO2, Al2O3 and Fe2O3. Typically, the raw meal has a chemical composition of 67 %w CaO, 22 %w SiO2, 5 %w Al2O3, 3 %w Fe2O3 and 3 %w of minor components such as MgO, SO3, alkali oxides and halogenides[13]. Various engineering technologies are available for the Portland cement production. However, the short kiln with a multistage of preheater and a cooling system is the current most efficient technology[15,16]. The basic steps of the process will be briefly described below (for further reading see references [17 and 18]).

1.1.1. Quarrying and preparation of raw materials Cement plants are often located close to naturally occurring calcareous and argillaceous deposits to minimize the transport cost. The first step of the dry process involves crushing of the quarried lime stone and clay into small pieces. In order to ensure a high reactivity for the raw materials, they are milled together to produce a raw meal with more than 85 % of the particles having a diameter below 90 m. The quality of the cement depends largely on the chemical composition of the raw meal. Three important parameters[13] are widely used in clinker production to design the chemical composition of the raw meal

LSF 

%w CaO 2.8(%w SiO 2 )  1.2(%w Al2O3 )  0.65(%w Fe 2O3 )

(1.1)

SR 

%w SiO 2 %w Al2O3  %w Fe 2O3

(1.2)

AR 

%w Al2O3 %w Fe 2O3

(1.3)

The proportion of silicate phases in Portland clinker is determined by the silica modulus (SR) while the lime saturation factor (LSF) mainly governs the alite (Ca3SiO5) to belite (Ca2SiO4) ratio. For the pure CaOSiO2Al2O3Fe2O3 system, a LSF of 1.0 or above indicates that the formation of alite from belite and lime is saturated. For most productions, LSF is set in the range

5


of 0.92 – 0.98 to minimize the amount of uncombined CaO, i.e. the free CaO which are not incorporated in the principal clinker phases. The alumina modulus (AR) is important for the quantity of the clinker melt which is the only transport medium for the raw material in the rotary kiln. In general, the silica and alumina modulus are in the ranges of 2.0  3.0 and 1.0  4.0, respectively. However, this does not include special clinker types such as white Portland clinker which has much lower aluminum and iron oxide contents. Variations in LSF, SR and AR can significantly change the “burnability”, which is defined as the ease of combining belite and lime into alite. The burnability may also be affected by several other factors, including the quantity of the minor components and the grain size of the raw material. It is usually expressed by the quantity of free CaO present in the final clinker.

1.1.2. Clinker production in the rotary kiln

The fine-grained raw meal is dried in a preheater tower, which consists of a series of vertical cyclones, before it is fed into the rotary kiln. As the particles fall into the combustion chamber, which is placed at the bottom of the preheater above the kiln, they have typically reached a temperature of 800 – 900 C. In this temperature zone, approximately 90 % of the limestone decomposes to lime in accordance with the chemical reaction CaCO3(s)  CaO(s) + CO2(g)

(1.4)

This process is usually referred to as calcination and is responsible for 60  65 % of the direct CO2 emissions from clinker production. The remaining ~40 % originates from the combustion of fuel, which is fired directly into the discharge end of the kiln. As the tilted kiln slowly rotates, about 3  5 rotations per minute, the calcined material slides and tumbles towards the flame. The material is heated to partial fusion and several chemical reactions occur as a kiln temperature of about 1450 C is progressively reached. Due to the rotation, the clinker leaves the kiln as roughly spherical nodules of 1  3 cm in diameter. The relative proportions of clinker minerals formed as the material moves into different kiln-temperature zones are graphically illustrated in Figure 1.1. The chemical reactions for the clinker formation can be divided into three groups: (1)

The most important reactions occurring below ~1300 C are calcination, decomposition of the clay minerals and formation of belite (2CaOSiO2), aluminate and ferrite. The major phases at the end of this stage are belite, lime, aluminate and ferrite. Only a small

6


amount of clinker melt is formed at this temperature. The formation of belite, which is the second most occurring phase in Portland clinker, can be expressed as 2CaO(s) + SiO2(s)  2CaOSiO2(s) (2)

(1.5)

As the kiln temperature is raised above 1300 C, a melt phase is formed, consisting mainly of tricalcium aluminate (3CaOAl2O3) and ferrite (Ca2(AlxFe1x)2O5, where 0 < x < 0.7). The pure CaO  SiO2  Al2O3  Fe2O3 system has a eutectic point, which occurs at 1338 C for an alumina modulus of AR = 1.38. A deviation of the alumina modulus from this value requires much higher temperature to preserve the same level of clinker melt content. In the temperature zone of 1300 – 1450 C, the clinker melt is the only transport medium for belite and lime. Therefore, the quantity and properties (e.g., viscosity and surface tensions) of the clinker melt are decisively important for the formation of alite. Lime and belite are brought into reaction within the melt phase to form alite, in accordance with the chemical reaction CaO(s) + 2CaOSiO2(s)  3CaOSiO2(s)

(1.6)

To some extent, the presence of the minor components such as SO3, P2O5, F and alkali oxides may contribute to the quantity as well as the properties of the clinker melt. These minor components originate from either the quarried raw materials or the combustion fuel. However, they may also be deliberately added to the raw meal to modify the clinker towards certain application requirements. (3)

The hot clinker falls into a cooler where it is immediately quenched either by incoming air or by water. Upon cooling, the clinker melt re-crystallizes to aluminate and ferrite. Depending on the cooling rate, different polymorphic transitions will also occur for alite and belite.

1.1.3. Cement grinding

To produce cement, the clinker is mixed and finely ground with other hydraulic minerals and usually 4  5 %w gypsum. Depending on the application requirements, the clinker content can vary from 10 %w to 95 %w. For example, the clinker factor in blastfurnace cement CEM III (C) is only 0.5  0.19, although this cement type is only used for a few special constructions such as roads and tunnels.


7

Figure 1.1 Schematic illustration of the typical proportions of phases for the formation of Portland clinker minerals as a function of the progressive kiln temperature. The figure is adapted from the work by Wolter[19].

8


The main constituent phases of Portland clinker

The properties of Portland cement are mainly determined by the proportion of its four principal clinker phases which are the impure forms of Ca3SiO5 (alite), Ca2SiO4 (belite), Ca3Al2O6 (tricalcium aluminate) and Ca2(AlxFe1-x)2O5 (ferrite). Other phases such as periclase (MgO), quartz (SiO2), free lime (CaO), etc. may also be present in minor quantities, usually less than 1 %w.

1.2.1. Tricalcium silicate

Alite is the most important constituent clinker component and typically constitutes 50  70 %w of Portland clinker. This highly hydraulic clinker mineral is an impure form of tricalcium silicate, which can occur in seven crystal modifications: three triclinic[20] (T), three monoclinic[21] (M) and one rhombohedral[22] (R) polymorphs. These alite structures are built of Ca2+, O2 and SiO44 ions. They are all structurally similar with respect to the positions of Ca, Si and O atoms, although they differ significantly in the orientation of the SiO4 tetrahedra. Furthermore, the mean coordination number of the Ca2+ ions is different for the polymorphs. For example, it is 5.66 in the R form, but 6.21 in TI polymorph. On heating, tricalcium silicate undergoes a series of reversible phase transitions[23] o

o

o

o

o

o

600 C 920 C 980 C 990 C 1050 C 1060 C TI   TII   TIII   M I   M II    M III    R

In fact, pure tricalcium silicate is only thermodynamically stable at temperatures above 1300 C. On cooling, it becomes unstable between 1300 C and 1000 C with a maximum rate of decomposition into belite and lime at 1175 C. At room temperature, pure tricalcium silicate is meta-stable only in its TI form. The crystal structure for tricalcium silicate[20] in its triclinic form is shown in Figure 1.2; the unit cell corresponds to the formula unit Ca27Si9(Ob)36(Oi)9, where b and i denote the oxygen sites involved in covalent SiO bonds and the interstitial oxygen sites, respectively. The high-temperature polymorphs of tricalcium silicate can be stabilized at room temperature by the incorporation of a sufficient amount of impurities in its structure[24-27]. For example, alite is stabilized as the MIII polymorph by the incorporation of Mg2+ ions[26] in the Ca2+ sites while the rhombohedral form is stabilized by Zn2+ incorporation[28]. The impurities are often referred to as substituent oxides and their ions as guest ions. In Portland clinkers, alite occurs mainly in the MIII and/or MI forms, containing roughly 4 %w substituent oxides[29-31]. The

9


typical chemical compositions, including the most common encountered substituent oxides, for Portland clinkers are summarized in Table 1.1.

Figure 1.2 Graphical illustration of the crystal structure for the triclinic form of tricalcium silicate. The unit cell includes 9 non-equivalent SiO4-tetrahedra, 9 interstitial oxygens (Oi) sites and 36 oxygens (Ob) bonded directly to the Si atoms. () Ca2+, (): interstitial oxygen O2 and (blue tetrahedra): SiO44-. The crystal structure data is adapted from reference [20].

1.2.2. Dicalcium silicate

The second major constituent phase of Portland clinkers is the impure form of dicalcium silicate, denoted belite. Its content in Portland clinkers is typically 5 – 30 %w. The structure of dicalcium silicate is built of Ca2+ and SiO44+ ions and can occur in five different polymorph modifications: one , one  and three  forms[32]. At room temperature, dicalcium silicate occurs mostly in the  form. On heating, the mineral undergoes phase transitions to form the high temperature polymorphs  500 C o

630  680 C o

1160 C o

1425 C o

'     α 'High     α γ    β     α 'Low

10


The structure of -Ca2SiO4, which is similar to that of olivine, (Mg,Fe)2SiO4, is much less dense as compared to the other polymorphs. Therefore, the transition from the high temperature polymorphs to  on cooling causes the sintered material to crack and fall to a more voluminous powder. The transition    can be avoided by the incorporation of a sufficient amount of guest ions in the dicalcium silicate structure. It can also be prevented if the -Ca2SiO4 crystallites are sufficiently small. The arrangements of Ca2+ and SiO44+ ions in the high temperature phases are very similar. However, they differ significantly from the  form. Clinker belite typically contains 4  6 %w substituent oxides and occurs largely in the  form

[29-31]

. An illustration of the 

dicalcium silicate crystal structure is shown in Figure 1.3; the asymmetric unit contains a single tetrahedral SiO4 site and two distinct Ca2+ sites, which have seven and eight oxygen atoms within the first coordination sphere.

Figure 1.3 Graphical illustration of the crystal structure for the beta form of dicalcium silicate [33]. The asymmetric unit only includes a single SiO4-tetrahedral site and two distinct Ca2+ sites. (): Ca2+ and (blue tetrahedra): SiO44-.

1.2.3. Tricalcium aluminate

Pure tricalcium aluminate (3CaOAl2O3) has a cubic structure and does not exhibit polymorphism[34]; it is built of Ca2+ ions and rings of six AlO4 tetrahedra. However, a rather large amount of guest ions can be incorporated in the Ca2+ as well as the Al3+ sites, resulting in a series of different structures[35]. In general, the tricalcium aluminate phase of Portland clinker exhibits cubic and/or orthorhombic structures. These impure forms contains up to 13 %w and 20 %w substituent oxides, respectively. The most important guest-ion incorporations are Si4+ and Fe3+ substituted in the tetrahedral Al3+ sites. A large amount of other guest ions such as Na+, K+ and Mg2+ may also be incorporated in the Ca2+ sites. The tricalcium aluminate content in ordinary Portland clinker is in the range of 5  10 %w whereas it is much lower in white Portland clinkers.

11

Chapter 1. Cement Chemistry 1.2.4. Ferrite

The ferrite (Ca2(AlxFe1-x)O5) content in ordinary Portland clinker is usually 5  15 %w, but much lower in white Portland clinker. The grey color of ordinary Portland cement is mainly due to the presence of this iron-rich phase. In clinkers, ferrite is typically found closely mixed with tricalcium aluminate. Due to the similarity in the cell parameters for the tricalcium aluminate and ferrite phases, oriented intergrowth can occur for these phases on recrystallization from the clinker melt[36]. The structure of ferrite can be derived from that of perovskite (CaTiO3). It can occur with any of the compositions in the solid-solution series Ca2(AlxFe1-x)2O5, where 0 < x < 0.7[13]. Each Ca2+ coordinates to seven oxygen atoms. Furthermore, Al3+ and Fe3+ are both distributed between octahedral and tetrahedral sites. Clinker ferrite contains about 10 %w of substituent oxides, mainly SiO2 and MgO. Other ions such as Na+, K+ and S6+ are also present in minor concentrations. The guest ions are mainly incorporated by substituting for either the Al3+ or Fe3+ sites whereas Ca2+ exhibits almost no substitution [29,37].

Table 1.1 Typical chemical compositions for the four principal clinker phases of Portland cement [13]. CaO

SiO2

Al2O3

Fe2O3

MgO

Na2O

K 2O

SO3

P2O5

TiO2

Mn2O3

Alite

71.6

25.2

1.0

0.7

1.1

0.1

0.1

0.1

0.1

0.0

0.0

Belite

63.5

31.5

2.1

0.9

0.5

0.1

0.9

0.3

0.1

0.2

0.0

Aluminate

56.6

3.7

31.3

5.1

1.4

1.0

0.7

0.0

0.0

0.2

0.0

Ferrite

47.5

3.6

21.9

21.4

3.0

0.1

0.2

0.0

0.0

1.6

0.7


12

Minor components

In addition to CaO, SiO2, Al2O3 and Fe2O3, Portland clinker contains approximately 3 %w of other components such as MgO, SO3 and alkali oxides[13]; those are usually referred to as minor component in cement chemistry. They originate mainly from the quarried raw materials and the combustion fuel. The most encountered minor components in Portland clinker can be divided into two groups according to their general influences on the clinkering process [24,25]: (1)

Flux agents include compounds that modify the temperature of liquid formation, the

properties of the clinker melt and the crystal morphology of the principal clinker minerals. (2)

Mineralizer agents include compounds that influence the thermodynamic stability of the

principal clinker minerals. However, in many cases, it is only used to categorize the compounds that influence the reaction of alite formation, i.e. Ca2SiO4 + CaO  Ca3SiO5, by modifying the thermodynamic properties of one or both calcium silicates. Furthermore, some minor components can modify the hydration properties of Portland cement to a significant extent. They typically involve one of the processes: (1)

Introduction of structural defects by forming solid solution within the principal clinker mineral phases.

(2)

Stabilization of clinker minerals in their high temperature polymorphs.

(3)

Effects arising during cement hydration such as coating of the cement grains by insoluble basic salts.

In general, the first two effects increase the hydrational reactivity of the clinker phases whereas the last prevents the hydration of the anhydrous cement phases. Another important distinction that also requires careful attention is the volatility of the minor components. Appreciably volatile components typically have a higher concentration in the kiln atmosphere than that in the raw material. Furthermore, they tend to be deposited in the cooler part at the end of the kiln due to the recycling of air to reduce heat loss. For laboratory synthesis, the volatile components have a rather high loss on ignition and thus, their contents in the clinkers can be much lower than that added to the raw meal. SO3, K2O, Na2O, ZnO, and Cl2 are very volatile while F, V2O5 and As2O3


13

have low volatility. Components such as MgO, P2O5, TiO2 and NiO are essentially non-volatile in the clinkering process.

1.3.1. Flux

Since almost all of the clinker melt is formed immediately at the eutectic temperature, the reactants, e.g. lime and belite, are limited within “local volumes” by various proportions. Therefore, the ease of combining lime and belite to form alite is strongly controlled by the transport of the reactants through the clinker melt between such local volumes. The transport rate is dependent on the quantity, viscosity and surface tension of the clinker melt. Most of the minor components, when present in the raw meal, decrease the temperature for the first liquid formation and contribute to the quantity of the clinker melt by approximately the same amount as their content. Their influence on the viscosity and surface tension of the clinker melt is somewhat complicated. In general, ions of strongly electropositive elements such as alkali metal increase the viscosity while those of electronegative elements like Cl and F have a reverse effect[25]. The influence of the p-block elements on the surface tension also follows a similar trend, where the surface tension is decreased for an increase in the electronegativity. On the other hand, the surface tension is increased as the electronegativity of the s-block elements is increased. The application of minor components for controlling the morphology of alite and belite plays a very important role in clinker production. It is apparent from Table 1.1 that Mg2+, Al3+ and Fe3+ ions constitute the major part of guest ions in the calcium silicate phases. However, compounds including ions such as S6+, P5+, K+ and Na+ have an increased interest in modern Portland clinker production. For example, Mg2+ ions are incorporated in the Ca2+ sites and stabilize alite in its MIII polymorph[26]. S6+ ions, on the other hand, substitute for the tetrahedral Si4+ sites and promote the formation of alite in its MI form. Furthermore, S6+ ions, when incorporated in belite, stabilize this phase in its  form[38]. Alkali ions and P5+ are preferentially incorporated in belite, stabilizing the ’-belite polymorph, and they substitute for the Ca2+ and Si4+ sites

[24,39,40]

, respectively. The guest ions may also compete for specific structural sites in

the clinker phases and therefore, limit or increase the substitution level for one another. For example, Na+, K+ and Mg2+ ions substitute preferentially for the Ca2+ sites while Al3+, P5+ and S6+ ions compete for the tetrahedral Si4+ sites.


14

1.3.2. Mineralizers

According to the restricted definition of a mineralizer which only includes compounds that influence the reaction of alite formation[25], the free lime content in the clinker can be used to monitor the mineralizing effect. For example, the substitution of Si4+ by Al3+ ions in the calcium silicate phases effectively increases the bulk SiO2 content and therefore, increases the belite content. However, this is a mass-balance effect and Al3+ is not a mineralizer since the free lime content remains low. On the other hand, S6+ ion is a strong mineralizing agent on the formation of belite. It increases the thermodynamic stability of belite by forming solid solutions in this phase[38]. Therefore, the presence of a SO3 source in the raw meal simultaneously increases the amount of belite and the free lime content for the SO3-mineralized clinker. In modern Portland cement production, CaF2 and CaSO4 are by far the most used mineralizers, where F strongly facilitates alite formation and SO3 mineralizes belite [41-45].

1.4.

Hydration chemistry for Portland cement

The hydration of Portland cement is a process that includes many simultaneous chemical reactions between the clinker minerals, gypsum and water, where the water/solid ratio required in the cement mixture (paste) is typically 0.3  0.6 by weight[13]. Furthermore, the addition of supplementary cementitious materials (SCM) such as natural pozzolans may also have important impacts on the entire hydration process[46-50]. This section gives a brief description of the basic hydration reactions for the clinker minerals as individual phases, which is fundamental for the hydration of Portland cement in general. The description is based on references [13 and 51].

1.4.1. Hydration of calcium silicate phases

The main products formed from the hydration of alite and belite are calcium silicate hydrates (CSH = xCaOSiO2yH2O), which constitute approx. 60 %w of the entire hydration product of Portland cement. The relative proportions of phases formed from the hydration of Portland cement are graphically illustrated in Figure 1.4. Typically, about 40 % of alite has reacted within 1 day after mixing with water, 70 % within 28 days and virtually all in one year. For a given grain-size distribution and water/solid (w/s) ratio, alite sets and hardens in a similar manner to Portland cement. In this context, the term setting (or set) is used for stiffening without


15

increase in compressive strength while hardening (or harden) is the process of significant strength development. Furthermore, hydration times up to 28 days are usually denoted early hydration periods whereas the later hydration is referred to as long-term hydration. The hydration of alite can be divided into three principle stages[52]: (1)

Within the first four hours, alite is attacked by water and a honeycomb-like product of calcium silicate hydrates (CSH) is formed on the surface of the alite grains. Some portlandite, i.e., Ca(OH)2, is also formed. nCaO  SiO 2  (n  x  y)H 2 O  xCaO  SiO 2  yH 2 O  (n  x)Ca(OH) 2 (1.7)

(2)

During the middle stage, i.e. 4  24 hours, CSH and portlandite are rapidly formed. The morphology of CSH formed at this stage strongly depends on the available space for growth. Fibrous CSH has been observed as the outer product whereas the honeycomb-like material grows in the surface fractures of the alite grains. As the hydration proceeds after 24 hours, the inner product of CSH is also formed, which is characterized by a very fine pore structure.

Belite follows the same hydration process as alite, however, with a much slower hydration rate; approx. 30 % of belite has reacted during the early hydration period and about 90 % in one year. Thus, the hydration of belite only contributes little to the early strength development, but it is mainly responsible for the strength development during the long-term hydration.

1.4.2. Hydration of tricalcium aluminate

Tricalcium aluminate (C3A)1 possesses a much higher reactivity as compared to alite and belite. Initially, the meta-stable calcium aluminate hydrate phases, C4AH13 and C2AH8, are rapidly formed from the paste-liquid[53]. Subsequently, these phases are slowly converted to the thermodynamically stable C3AH6 phase. The overall hydration of C3A has finished within a few hours after the mixing. This excessive hydration of C3A results in a severe stiffening, which cannot be dispelled by remixing. This phenomenon is usually referred to as flash setting. In order to retard the hydration of C3A, gypsum is normally added to the Portland clinker at the end of the 1

The nomeclature C = CaO, S = SiO2, A = Al2O3, F = Fe2O3 and H = H2O is commonly used in cement chemistry.

However, it is only used to write the hydration reactions of tricalcium aluminate in this section and to abbreviate the calcium silicate phases throughout the thesis.

16


clinkering process[54]. In the presence of sulphate ions, the C3A phase reacts rapidly to form ettringite, which is an AFt phase. Fast

C3A  3CSH 2  16H  C 6 AS3H 32

(1.8)

Slow

2C3A  C 6 AS3H 32  4H  3C 4 SH12

(1.9)

When the sulphate ions are consumed by this reaction, ettringite slowly reacts with the exceeding aluminate phase to form monosulphates (AFm), equation (1.9). The general formula for

the

AFm

(Al2O3F2O3mono)

and

AFt

(Al2O3F2O3tri)

phases

are

Ca2(Al,Fe)(OH)6XwH2O and [Ca3(Al,Fe)(OH)612H2O]2X3wH2O, respectively[13,51]. Here, X in AFm denotes one formula unit of a singly charged anion such as OH or half a formula unit of a double-charged anion such as SO42 or CO32. On the other hand, in AFt it denotes one formula unit of a double-charged anion or two formula units of a single-charged anion. In addition to setting regulation properties, the sulphate ions have important impacts on the hydration rate of alite and the volume stability of the hydration products. The optimum content of gypsum, however, depends on several factors, such as the cement composition, hydration time and other conditions of hydration. The minimum content of SO3 required to control setting is approx. 2 %w for an ordinary Portland cement.

1.4.3. Hydration of Portland cement

The microstructure of hydrated Portland cement develops in a similar manner to that formed from the hydration of tricalcium silicate[55,56]. Soon after mixing, a gel-like layer containing amorphous alumina, silica, calcium and sulphate in varying amounts is formed on the cement grain surfaces. The AFt phase (e.g. ettringite) starts to precipitate within about 10 minutes after mixing. After about four hours, the cement grains are completely covered by a thickening layer of CSH. As the hydration proceeds, the CSH shells grow outward and inter-grow with surrounding adjacent grains. This happens about 12 hours after mixing and usually it is referred to as the cohesion point, where the setting has completed. A space filled with a highly concentrated ion solution is observed between the shell and the inside of the anhydrous material; the space can be up to 0.5 m wide. At this point, the CSH shell is still sufficiently porous and ions can migrate between the inner and outer space through the shell wall. The hydration at this


17

stage is driven by dissolution and precipitation mechanisms. At a later stage (>24 hours), the permeability of the shell is decreased significantly, preventing ion migration. CSH is formed and deposits on the inside wall. The concentration of SO42 ions inside the shells also drops rapidly and AFt slowly reacts with the remaining aluminate phase to form AFm. The strength development of Portland cement is mainly related to the formation of CSH, where the SiO4 monomers released from the hydration of alite and belite polymerize into dimers and longer chains of silicate tetrahedra.

Figure 1.4 Schematic representation of the hydration for the main Portland clinker phases and their resulting hydration products. The areas of the boxes represent roughly the typical relative proportions of these phases as reported in reference [13].

18


Structure of the main hydration product CSH

A calcium silicate hydrate (CSH) phase is the principal binding phases in hardened cement[13,51]. The nanostructure of CSH exhibits several similarities to the structure of the crystalline minerals tobermorite and jennite. However, CSH occurs with a wide range of compositions, morphology and degree of structural order. The dashes indicate that no particular composition is implied. The Ca/Si molar ratio in CSH formed from a nearly saturated alite paste can possess a value in the range from about 1.2 to 2.1[57]. The average Ca/Si ratio is, however, typically ~1.75[57-59]. An increase in the quantity of belite and/or SCMs results in a decrease in the Ca/Si molar ratio, which typically enhances the long-term strength and durability of the hydrated Portland cement, where the lower limit for the Ca/Si ratio in hydrated Portland cement is 0.7[57]. Several models have been proposed for the nanostructure of CSH phases that are formed from the hydration of calcium silicates, cement or related materials[60-64]. Many of them exhibit several similar features, although they provide different degree of flexibility and complexity. The models fall typically into two categories, which are derived from the structure of tobermorite 14-Å. The first category, denoted T/CH, includes building blocks of tobermoritelike structure (T) intermixed with layers of calcium hydroxide (CH). The second group, which is usually denoted as T/J, contains elements of tobermorite and jennite (J) structures. The structures of tobermorite, jennite and Ca(OH)2 are described in Sections (1.5.1)  (1.5.3), respectively. Richardson and Groves introduced a generalized model for CSH[62,65], which includes formulations that could be interpreted from both T/CH and T/J structural viewpoints. The model proposes the following general composition for the CSH

Ca 2n H wSi 3n 1O9n  2  OH w  n y  2  Ca n y  mH 2O

(1.10)

2

The number of silanol (SiOH) groups is given by w and the degree of protonation of the silicate chains by w/n. The Ca/Si ratio can be obtained as Ca/Si 

n4  y  23n  1

(1.11)

The chemical formula for the main layer is given by the braces. The main layer contains 2n Ca2+ ions surrounded by dreierketten silicate chains of mean length 3n  1. The charge balance is preserved by n  (w/2) interlayer Ca2+ ions of the (ny)/2 given outside the braces. From the T/CH viewpoint, the remainder of the (ny)/2 Ca2+ ions occur in Ca(OH)2 layers which


19

corresponds to the interlayer of tobermorite whereas, from the T/J viewpoint, they form a part of the main layer resulting in regions with jennite-like structure. This flexible structure model allows the Ca/Si ratio to vary from 1.0 to 2.5 by (1)

omission of part of the bridging SiO4 tetrahedra in the dreierketten silicate chains, resulting in an increased Ca/Si ratio. Ca2+ may also be substituted into the bridging sites leading to a further increase in the Ca/Si ratio,

(2)

varying the degree of protonation, w/n; since, the increased Ca2+ content can be balanced by a decrease in the SiOH content and vice versa,

(3)

incorporation of additional Ca2+ ions can be charge-balanced by additional OH. For the T-based CSH structure the OH ions are located in the interlayer whereas they are incorporated in the main calcium layer of the J-based part.

1.5.1. Tobermorite 14-Å

Tobermorite 14-Å, Ca5Si6O16(OH)27H2O, is the most hydrated phase of the tobermorite mineral family[66]. The name refers to the basal spacing between the principal layers which exhibit a chemical formula of [Ca4Si6O16(OH)2(H2O)2]2. The principal layer includes two structurally nonequivalent calcium ions in sevenfold coordination and three tetrahedral Si4+ sites. The SiO4 tetrahedra form dreierketten chains with periodicity of three silicon atoms, two Si on the pair sites and one Si in the bridging site. The Ca atoms are coordinated to six oxygen atoms and a hydroxyl group of the bridging SiO4 tetrahedron forming a central Ca-O sheet, where the dreierketten silicate chains are running on both sides. The CaO7 and SiO4 polyhedra form the principal layer by sharing their vertices and apices as illustrated in Figure 1.5. The principal layers are separated by an interlayer, which contains Ca2+, OH and a large amount of water molecules. The layers are held together by strong hydrogen bonds which are formed between the oxygen atoms from the principal layer and the hydrogen atoms from water molecules present in the interlayer.


20

Figure 1.5 Illustration of the layer structure for tobermorite 14-Å, Ca5Si6O16(OH)27H2O. The calcium oxide in the main layers is shown by green polyhedra. The dreierketten silicate chains shown in blue are formed from sequences of three SiO4 tetrahedra, one bridging and two pair sites. The interlayer consists of Ca2+ and OH ions, which are depicted in grey and red, respectively. The crystal structure data is adapted from reference [66].

1.5.2. Jennite

The crystal structure of jennite[67], Ca9Si6O18(OH)68H2O, is classified in the space group P-1. It exhibits a triclinic unit cell with a = 10.756 Å, b = 7.265 Å, c = 10.931 Å,  = 101.30,

 = 96.98 and  = 109.65. The structure, Figure 1.6, is built of three distinct modules: ribbons of two calcium octahedra sharing edges, dreierketten silicate chains and calcium octahedra on inversion centers. The ribbons are connected to each other by sharing vertices. This results in a zigzag layer of calcium layers, containing two types of ribbons: the first is sharing all its vertices with other polyhedra while in the second ribbon, the apices on both sides of the zigzag layer are


21

corresponded to water molecules. The ribbons are further firmly linked through the dreierketten silicate chains, which are present on both sides of the zigzag calcium layer. Furthermore, the silicon atoms on the bridging sites are strongly coupled to the protons on water molecules. Both calcium ribbons and silicate chains are running along the b axis. The composite octahedraltetrahedral layers are further connected through the additional calcium atoms in octahedral coordination occurring on the inversion centers.

Figure 1.6 Illustration of the layer structure for jennite[67], Ca9Si6O18(OH)68H2O. The calcium oxides in the main layers are shown by green polyhedra. The dreierketten silicate chains shown in blue are formed from sequences of three SiO4 tetrahedra, one bridging and two pair sites. The main layers are connected through the additional calcium atoms in octahedral coordination occurring on the inversion center.


22

1.5.3. Portlandite

Ca(OH)2 has a layer structure[68] (Figure 1.7) and belongs to the space group P-3m1. The layers consist of one Ca2+ sheet which is sandwiched by two sheets of hydroxyl groups (OH). Each calcium ion coordinates to six hydroxides forming a perfect octahedron with a CaO bond length of 2.37 Å. Under ideal conditions, Ca(OH)2 crystallizes with euhedral hexagonal shape. In hydrated Portland cement, however, it exhibits a more massive and indeterminate shape. Furthermore, a small amount of SiO2 can be incorporated in the calcium layer. Ca(OH)2 constitutes roughly 20 %w of the hydration products from Portland cement and this phase is usually referred to as portlandite.

Figure 1.7 The layered crystal structure for Ca(OH)2. The calcium ions, shown in grey, are coordinated to six hydroxyl groups. The oxygen atoms are shown in red and hydrogen atoms in blue. The crystal structure data is adapted from reference [68].

Chapter 2. Applications of solid-state NMR in cement research

23

2. Chapter

Applications of Solid-State NMR in Cement Research

The main advantage of solid-state NMR is related to its applicability in structural investigations of crystalline as well as amorphous or poorly crystalline materials. Complimentary to conventional analytical methods in cement research such as powder X-ray diffraction (XRD), which generally detect the bulk or long-range order structures, different NMR techniques can be used to obtain specific information about local structures. Generally, the individual NMR-active spin isotopes exhibit different resonance frequencies and their NMR interactions show a strong dependence on their specific site locations. Another important benefit of NMR in cement research is its high sensitivity for specific NMR isotopes, making it possible to investigate the structure of guest ions, which otherwise is very difficult by other techniques. This Chapter gives an introduction to the basic solid-state NMR theory and its applications in cement research. Furthermore, it includes relevant details for the NMR techniques, which have been employed in this PhD project.

Chapter 2. Applications of solid-state NMR in cement research 2.1.

24

NMR theory

For nuclear-spin isotopes with a non-zero spin-quantum number (I), when subjected to an external static magnetic field (B0), their nuclear magnetic moments orient either parallel or antiparallel with respect to the direction of B  0, 0, B0  . The interaction between the nuclear spin and the externally applied static magnetic field is the so-called Zeeman interaction, which splits the degenerated energy states of the nuclear spin into (2I + 1) states. The detected NMR resonance frequency () is related to the energy difference (E2  E1) between these Zeeman states by ν

E 2  E1  Δm  γ  B eff 

(2.1)

Here,  is the Planck’s constant (h/2) and m is the magnetic quantum number, which can take any of the values from I to I. In NMR, only single-quantum coherences, i.e. transitions with m = 1, can be detected directly. However, multiple-quantum coherences such as zero- and doublequantum coherences of a two spin system or triple-quantum coherences for I ≥ 3/2 quadrupolar nuclei may be indirectly detected in advanced NMR experiments. is the gyromagnetic ratio, which is a specific constant for each of the nuclear-spin isotopes in the periodic table. Beff is the effective magnetic field whose magnitude is determined by the externally applied magnetic field and the local electronic and magnetic environments at the specific nuclear site. In essence, the state of such a system is governed by quantum mechanics and can be fully described by the timedependent Schrödinger equation. However, solving the Schrödinger equation for a NMR experiment, which often includes a very large number of spins (> 1015), can be rather difficult and time consuming. Alternatively, the spin state in the NMR experiment is more conveniently expressed by a spin-density matrix formalism using the Liouville-Von Neumann equation, which in the absence of relaxation is given by[69]  ρ t   iH  t , ρ t  t H  H Z  H rf  H Q  H σ  H D  H J

(2.2)

(2.3)

where i is the complex imaginary unit. The density matrix, (t), is the statistical average for an ensemble of spins. H(t) is the nuclear-spin Hamiltonian describing the external and internal spin


25

interactions, where the external terms (HZ and Hrf) are applied to modify the nuclear-spin state whereas the internal interactions (HQ, H, HD and HJ) are nuclear-spin dependent and sensitive to the local environment of the observed nuclear spin. HZ = LIZ is the Hamiltonian for the Zeeman interaction. For a given static magnetic field, B0, the Larmor frequency L = B0 is specific for each nuclear-spin isotope. The Hrf term in equation (2.3) represents the pulsed radiation, applying an external rf field perpendicular to B, which in the laboratory-fixed frame may be expressed as H rf  2 1  cos  carr t  φ I x

(2.4)

where 1 = –B1/2 is the rf-field strength, carr (L) is the carrier frequency of the pulse and

is its phase relative to the direction of B1 . The J-coupling (HJ) is the direct spin-spin interaction, i.e. a through bond coupling, which is usually small and has not been of importance in the present study. A short description of the remaining internal interactions, i.e. the chemical shift interaction (H), dipolar couplings (HD) and the quadrupolar coupling interaction (HQ), is provided below. In general, these interactions are very important in solid-state NMR of inorganic diamagnetic solids.

2.1.1. Chemical shift interaction

The chemical shift interaction, including the isotropic chemical shift and chemical shift anisotropy (CSA), originates from the chemical shielding of the external magnetic field caused by the local electron distribution surrounding the observed nuclear spin. The Hamiltonian for the chemical shift interaction between the nuclear spin I and the external magnetic field ( B) is given by H  =  I    B , where  is the chemical shift tensor, which in the Principal-Axis System (PAS) can be expressed as[70] σ xx  σ 0  0

0 σ yy 0

0  σ iso  ½ σ 1   σ  0 0     0  0 σ iso  ½ σ 1    0  σ zz   0 0 σ iso    

(2.5)

Here, the PAS elements fulfill the condition |zz iso|  |xx iso|  |yy iso| and the other parameters are defined according to


 iso  



1 σ xx  σ yy  σ zz 3



 σ  σ iso  σ zz

ησ 

26 σ xx  σ yy

σ

(2.6)

iso = iso is the isotropic component, where the chemical shift () is positive to high frequency (i.e. down-field) and the absolute shielding () is positive to low frequency (i.e. up-field).  and

 give the magnitude and the asymmetry of the chemical shift interaction, respectively. Alternatively, the CSA can be characterized by the span ( = 33  11) and skew ( = 3(iso 

22)/) parameters, where the PAS elements are chosen as 11  22  33 and iso = 1/3(11 + 22 + 33). In general, the NMR resonances are referenced by their chemical shift values quoted in parts per million (ppm)

 sample 

ν sample  ν ref ν ref

 10 6

(2.7)

This normalization of NMR resonances by standard reference samples (ref) makes it possible to compare data independent of the externally applied static magnetic field. In liquid-state NMR,

iso is the only detectable component of the chemical shift interaction since the orientationdependent CSA is averaged out by molecular tumbling. A similar averaging can be achieved for solids using Magic-Angle Spinning (MAS)[71]. In this technique, the powder sample is rotated around an angle of θ  arctan 2 (~54.736) with respect to the external magnetic field. The geometrical dependency of the CSA becomes time dependent, where the time-dependent terms can be averaged out by employing a spinning frequency (R) larger than the width of the resonance in a static-powder experiment. For lower spinning speeds, the CSA results in a characteristic spinning sideband (ssb) pattern with an envelope that resembles the lineshape in a static NMR spectrum[72]; the positions of the spinning sidebands are separated by the value of R. The magnitude of  and , which reflect the local electronic structure of the probed spins, can be obtained from a simulation of the ssb pattern[73].

2.1.2. Quadrupolar coupling interaction

NMR experiments for nuclear spin isotopes with I > 1/2 exhibit an additional interaction, 

the quadrupolar coupling interaction ( H Q  I  Q  I ), which is the coupling between the nonspherical charge distribution of the nucleus and the electric-field gradients (EFGs) at the nuclear


27

sites created by its surrounding electron density[71]. The quadrupolar coupling tensor in its PAS is given as Vxx eQ  Q 0 2 I 2 I  1h   0

0 Vyy 0

0  0  ½  Q  1 CQ   0  0 ½ Q  1 2 I 2 I  1h   Vzz  0 0





0  0 1

(2.8)

where V(x,y,z) is the electric field gradient tensor which in its PAS fulfils |Vzz|  |Vxx|  |Vyy| and Vzz = eq is used for the principal EFG tensor element. CQ = e2qQ/h is the quadrupolar coupling constant, where eQ is the nuclear quadrupole moment, and Q = (Vxx  Vyy)/Vzz is the EFG asymmetry parameter (0  Q  1.0). The quadrupolar coupling can possess a magnitude of several MHz and thus, it cannot be averaged out by MAS. In general, it is sufficient to describe the quadrupolar coupling interaction by the two lowest orders of its effective Hamiltonian obtained by the Magnus expansion[69] ~

H

eff Q



~ ( 1) HQ 

~ ( 2) HQ

(2.9)

The two terms are usually referred to as the first-order and second-order quadrupolar coupling interactions. For quadrupolar nuclei with half-integer spin-quantum number, the first-order term introduces a frequency modulation on the satellite transitions (m ↔ m – 1, where m  1/2) resulting in a manifold of ssbs while it does not affect the central transition (1/2 ↔ 1/2). The first-order interaction can in principle be averaged out by MAS, although the achievable spinning speed are often much to low for a complete averaging. On the other hand, all transitions are affected by the second-order quadrupolar coupling interaction, which is only reduced by MAS. For strong quadrupolar couplings, the centerband for the central transition exhibits a characteristic lineshape with singularities, from which the quadrupolar coupling parameters CQ and Q can be determined. The quadrupolar coupling parameters may also be obtained from a simulation of the full spectrum of ssbs observed for the satellite transitions[74].

2.1.3. Dipolar coupling interactions

The dipolar coupling is a through space interaction between two adjacent nuclear spins (I and S), including two possible cases: (i) homo-nuclear dipolar couplings where I and S are


28

identical spin nuclei and hetero-nuclear dipolar couplings where I and S are different spins. The Hamiltonian for these interactions in a tensorial form is given as[71] 

HD  I DS

(2.10)

where the second-rank tensor for the dipolar coupling takes the following form in its PAS D xx  D 0  0

0 D yy 0

0   d / 2 0 0    0  0  d / 2 0 D zz   0 0 d 

(2.11)

Here, d 0hIS82r3is the dipolar coupling constant, where 0 is the permeability in vacuum and r is the internuclear distance for the IS spin pair. It is apparent from equation (2.11) that the tensor is traceless with the largest PAS element (Dzz) oriented along the dipolar coupling axis. The absence of an isotropic component implies that dipolar couplings cannot be observed directly in liquids. In solid-state NMR, the first-order Hamiltonian for the dipolar coupling, when considered as a perturbation to the Zeeman interaction, takes the form





H D  t   ν D 3I z Sz  3I  S

(2.12)

where I  S  I x Sx  I y S y  I z Sz and D is the frequency modulation caused by the dipolar coupling interaction. For a pair of hetero-nuclear spins (i.e. I  S), the Hamiltonian can be further reduced to H  t   ν D 2 I z Sz

(2.13)

assuming that (L,I – L,S) >> D. This non-zero dipolar coupling interaction results in a frequency modulation, which in the magic-angle spinning axis system is given by νD 





d cos2ν R t  α  sin 2 β   2 cosν R t  α  sin 2β  2 6

(2.14)

Here,  and  are the azimuthal and polar angles describing the orientation of the internuclear vector between I and S within the MAS frame. As the dipolar coupling constant is inversely proportional to the cube of the distance between the spins (1/rIS3), the condition of R > D is


29

often achievable for hetero-nuclear dipolar couplings. However, the dipolar couplings between high- spins such as 1H and 19F and an observed nuclear spin of lower Larmor frequency can be very strong and therefore, they are not necessarily averaged out by MAS. Alternatively, different homo- and hetero-nuclear decoupling sequences can be applied to eliminate or partly reduce the line-broadening caused by the dipolar couplings on the observed resonances[75-77].

2.2.

Solid-state NMR techniques in cement research

2.2.1.

29

Si MAS NMR

29

Si is probably the most studied nuclear spin for cementitious materials, owing to the

high bulk SiO2 content and the decisive role of the silicate phases in anhydrous as well as hydrated cements. In general, 29Si MAS NMR experiments are rather time-consuming due to the low natural abundance of the the

29

29

Si spin of 4.7 %. Furthermore, when present in cement phases,

Si spins may exhibit a long spin-lattice relaxation time[78], which usually requires a

relaxation delay of 30 s at 7.1 T. From pioneering

29

Si MAS NMR studies of zeolites and

amorphous materials such as glasses and cements[79-82], it has been demonstrated that the

29

Si

isotropic chemical shift for silicate mainly depends on the condensation of SiO4 tetrahedra (Figure 2.1), where an increased condensation (Qi, 0  i  4) corresponds to an up-field shift, i.e. more negative (29Si) value. In addition, the 29Si isotropic chemical shift reflects the number of Al atoms incorporated in the second-coordination sphere of the probed silicon sites, denoted Qi(nAl, 0  n i), for which each Si4+ Al3+ substitution leads to a down-field shift of approximately 5 ppm. However, this implies that the chemical shift regions for different Qi units overlap with each other, complicating the interpretation of the detected 29Si resonances. In accordance with the structure of triclinic tricalcium silicate determined by XRD, the 29

Si MAS NMR spectrum of a synthetic sample of this mineral identifies nine distinct silicon

environments with isotropic chemical shifts in the region for isolated SiO4 tetrahedra (Q0)[49], Figure 2.2. For the monoclinic forms MI and MIII, which are the commonly encountered forms of alite in Portland cement,

29

Si MAS NMR shows broadened lineshapes from the overlapping

resonances of their 18 silicon sites[83], covering the same spectral region. As it can be seen from Figure 2.2, the MI and MIII forms of alite are distinguishable by their characteristic

29

Si MAS

NMR lineshapes. In addition to the alite resonances, the 29Si MAS NMR spectrum of anhydrous Portland cement contains a narrow resonance at (29Si) = 71.33 ppm from belite in its 


30

form[84]. The isotropic chemical shifts for the Q0 tetrahedra of the alite and belite structures can be correlated with their corresponding mean SiO bond lengths, according to the linear relationship[85]

  29 Si   316.7d Si  O  Å   445.3

(2.15)

Figure 2.1 Schematic representation of the chemical shift regions for tetrahedrally coordinated silicon in silicates[79]. Q0 corresponds to the silicate monomers, Q1 denotes the dimers and chain-end groups, Q2 are the middle groups in a silicate chain while Q3 are chain-branching silicates and Q4 are cross-linked framework silicates.

The fact that the

29

Si resonances for SiO4 tetrahedra with different degrees of

condensation appear in distinguishable spectral regions can be utilized to follow the hydration process for the silicate phases of Portland cement using

29

Si MAS NMR. Generally, the

hydration processes involve condensation of the Q0 units to form dimer (Q1) and polymer (Q2) silicates. The degrees of hydration for the individual phases (e.g. alite and belite) at each hydration stage can be obtained by

Chapter 2. Applications of solid-state NMR in cement research  I t   H t   1   I0  

31

(2.16)

where I0 and I(t) are the normalized

29

Si intensities of the actual phase in the anhydrous and

hydrated samples, respectively. The fraction of each individual phase (e.g. alite, belite and CSH) can be quantified from a deconvolution of the

29

Si MAS NMR spectra[83]. In this project, the characteristic

lineshapes for alite in its monoclinic forms (Figure 2.2) have been simulated satisfactorily by including nine

29

Si resonances. For the CSH phases covering the spectral region from –75

ppm to –90 ppm, the deconvolution includes two resonances at 76 ppm and 79 ppm for the Q1 components and two resonances for the Q2(1Al) and Q2 units at –81 ppm and –85 ppm, respectively. Additionally, a resonance at 83 ppm has been included, accounting for protonated Q2 units, which have been identified for CSH samples cured in a CO2 atmosphere (Chapter 4).

Figure 2.2 29Si MAS NMR spectra (7.1 T and R = 7.0 kHz) of synthetic samples of alite in the triclinic (a) and monoclinic MI (b) and MIII (c) forms. Nine resonances have been observed for the triclinic Ca3SiO5, for which the rather broad peak at –69 ppm includes two overlapping resonances while the intensity of the peak at –73.44 ppm is twice as large as the remaining resonances. The diamond () at 71.33 ppm indicates an impurity of belite.

Chapter 2. Applications of solid-state NMR in cement research 2.2.2.

27

27 29

Si,

27

32

Al MAS NMR

Al MAS NMR is another very commonly applied tool in cement research. In contrast to

Al has a quadrupolar nuclear spin with the spin-quantum number I = 5/2, which in the

presence of an external magnetic field gives rise to six non-degenerated energy levels. The 27Al MAS NMR spectrum may be rather complex if all transitions are observed; the satellite transitions of the

27

Al spin are substantially broadened by the first-order quadrupolar coupling

interaction. On the other hand, the central transition is only perturbed by the second-order quadrupolar coupling interaction. Thus, in general, only the

27

Al MAS NMR spectrum for the

centerband of the central transition is considered in the analysis of the 27Al MAS NMR spectra. In ordinary Portland cement, the main part of the bulk Al2O3 content is present in tetrahedral coordination, which exhibits

27

Al resonances in the spectral region from about 40

ppm to 90 ppm at 14.1 T[82,86]. In general, the main aluminate phase (i.e. tricalcium aluminate) appears as a broad resonance covering from about 40 ppm to 85 ppm. As a result of the high natural abundance for the 27Al spin of 100 %, it is also possible to detect Al3+ guest ions that are incorporated in the calcium silicate phases, despite their rather low concentration. The Al3+ guest ions from alite and belite appear as a rather narrow centerband at about 82 ppm with a shoulder at 84 ppm, respectively. A small amount of Al2O3 is also present in the ferrite phase, but these 27

Al spins cannot be observed due to their strong dipolar coupling with the electron spins of the

Fe3+ ions. Generally, the hydration of Portland cement involves a conversion of tetrahedral AlO4 into octahedrally coordinated species (e.g. ettringite and monosulphate) and therefore, it may easily be followed by 27Al MAS NMR[82]. The 27Al resonances from ettringite and monosulphate appear at approx. 13.0 ppm and 9.0 ppm at 14.1 T, respectively. Furthermore, a resonance from the tetrahedral Al3+ guest ions in the CSH phases has been observed at about 60 ppm.

2.2.3.

19

F MAS NMR

The increasing application of fluoride mineralization in Portland cement production has been the major motivation for the application of

19

F MAS NMR in structural investigations of

cementitious materials in this project. Although several 19F MAS NMR studies are available in the literature for fluoride structures in cement related materials such as glasses[87-90], only a single study on Portland cement[91] has been reported earlier. 19F is a spin I = 1/2 nucleus with a natural abundance of 100 %. Due to the high gyromagnetic ratio of the

19

F spin, close to the


33

value of 1H, the method is very sensitive, enabling fluorine to be detected even when present in very small concentrations. Previous studies of several different metal fluoride compounds and fluorine minerals have demonstrated that the (19F) chemical shift is strongly affected by the type of counter ions[92], and distributed over a spectral region of about 200 ppm with -PbF2 at high frequency (20 ppm) and NaF on the other low-frequency side at approximately 222 ppm. The (19F) chemical shifts for different SiF and AlF covalent bonds of silicon and aluminum in octahedral, tetrahedral and five-fold coordination are predicted to appear in distinct spectral regions[90]. However, the presence of different types of counter ions, such as Ca2+ and Na+, may introduce additional frequency shifts to the observed

19

F nucleus, merging the chemical shift regions for

the different fluorine environments of the central cation coordination states. This is illustrated in Figure 2.3 by 19F MAS NMR spectra for fluoride with different counter ions.

Figure 2.3 19F MAS NMR spectra (7.1 T, R = 8.0  12.0 kHz) of NaF (a), SrF2 (b), Na2SiF6 (c) and MgSiF6 (d). Spectra (a) and (b) demonstrate the effect of counter ions on the chemical shift of fluoride ions while (c) and (d) show the effect of counter ions on fluorine in covalent SiF bonds with the silicon atom in octahedral coordination state. The asterisk shown in (a) indicates an artifact.

Chapter 2. Applications of solid-state NMR in cement research 2.2.4.

43

34

Ca MAS NMR

The application of 43Ca NMR spectroscopy has been very limited since just about twenty 43

Ca NMR studies of organic as well as inorganic compounds have been reported[93-101]. Only

few of these studies concerns the application of 43Ca MAS NMR in structural investigations of Portland cement or its related materials[93,94]. The major difficulties in obtaining 43Ca MAS NMR spectra with an acceptable signal to noise (S/N) ratio are related to the low natural abundance of the 43Ca spin isotope (0.135 %) and its low gyromagnetic ratio, e.g. its Larmor frequency is only 40.39 MHz at 14.1 T[95]. Furthermore, since 43Ca is a quadrupolar spin (I = 7/2), when present in asymmetric environments such as in Ca(OH)2, its centerband from the central transition is severely broadened by the strong second-order quadrupolar interaction. These difficulties are illustrated in Figure 2.4 for a sample of Ca(OH)2. The 43Ca MAS NMR spectrum was recorded at 14.1 T using a 7.5 mm PSZ rotor (450 l sample volume). To get a decent signal to noise ratio (S/N = 11.9), the experiment employed 196,608 scans with a relaxation delay of 2 s, corresponding to a spectrometer time of 111 hours. The quadrupole coupling parameters ((43Ca) = 63 ppm, CQ = 2.65 MHz and Q = 0.22) for the 43Ca spins in Ca(OH)2 were determined from a spectral simulation of the second-order quadrupolar lineshape (Figure 2.4) using the STARS software package[74]. This strong quadrupolar coupling leads to a broad resonance for the centerband of the central transition, covering a region of nearly 40 ppm.

Figure 2.4

43

Ca MAS NMR spectra (14.1 T) of Ca(OH)2 (a, R = 5.0 kHz) and a synthetic sample of cuspidine

Ca4Si2O7F2 (b, R = 3.0 kHz). The spectra were obtained using a solid /2-pulse of 2 s, a relaxation delay of 2 s and 196,608 and 128,000 scans, respectively. The simulation of the second-order lineshape for the Ca(OH)2 is shown below the spectrum (a).

43

Ca spins in


35

Another example, shown in Figure 2.4, is for a synthetic sample of cuspidine (Ca4Si2O7F2), in which calcium is distributed among four distinct crystallographic sites. The 43

Ca MAS NMR spectrum (14.1 T, R = 3.0 kHz) was acquired over 74 hours, i.e. 128,000 scans

for a relaxation delay of 2 s. The cuspidine sample was packed in a thin wall 7 mm Si3N4 rotor (310 l sample volume). This experiment results in a

43

Ca MAS NMR spectrum with S/N =

10.6. Due to the severe overlap of resonances, covering a spectral region from approximately 30 ppm to 30 ppm, in conjunction with the low S/N ratio, the different Ca environments in the cuspidine structure cannot be distinguished. Alternatively, for Ca sites with weak quadrupolar couplings such as in CaF2, the CrossPolarization sequence may be applied to enhance the sensitivity in the

43

Ca NMR experiment.

This is demonstrated in Figure 2.5 for a synthetic sample of CaF2. The standard single-pulse 43Ca MAS NMR spectrum was recorded with 17,280 scans, corresponding to 73 hours and the spectrum has a S/N ratio of 58.8. However, a significant increase in the S/N ratio is obtained for the

43

Ca{19F} CP/MAS spectrum, which has S/N = 140.3 and was recorded with only 6,144

scans (~26 hours).

Figure 2.5

43

Ca MAS (a) and

43

Ca{19F} CP/MAS NMR spectra (14.1 T, R = 3.0 kHz) of a synthetic sample of

CaF2. The sample was doped with a small amount of NiO in order to reduce the spin-relaxation time for the 19F and 43

Ca spins. The spectra were recorded using a 15-s relaxation delay and 17,280 and 6,144 scans, respectively. The

signal to noise (S/N) ratios in the spectra are 58.8 and 140.3, respectively. The vertical scale is expanded by a factor of 11.6 in (a) relative to (b).


36

2.2.5. Inversion-Recovery (IR) MAS NMR

A reliable quantification of the intensities in NMR spectra generally requires that the spectra are recorded with a relaxation delay of at least d1  5T1, where T1 is the spin-lattice relaxation time. Determination of the magnitude for T1 can be obtained by the InversionRecovery (IR) pulse scheme[102] (Figure 2.6), where a systematic increase of the recovery time ( results in a series of spectra, for which the observed intensity is governed by the exponential relation[103]   t  M t   M 0 1  1     exp    T1  

(2.17)

Here, M0 is the equilibrium magnetization, i.e. M(t = ), and  is a constant related to the pulse imperfections, where  = 1 for ideal pulses.

Figure 2.6 Illustration of the Inversion-Recovery (IR) pulse scheme. Systematic increase of the recovery time ( results in a series of spectra, from which the T1 or T1’ relaxation time can be determined from a fit of the experimental data to equation (2.17) or (2.18).

The IR experiment can also be used to study paramagnetic ions indirectly. For dilute spin systems, the spin-lattice relaxation time constant of the observed spin (I) may be dominated by the strong dipolar interactions between the I spin and the free electrons spin (S) of the paramagnetic ions. The inversion-recovery magnetization for such systems can be fitted to a ‘stretch exponential’ equation[103,104]   t  M t   M 0 1  1     exp   '   T  1  

(2.18)

where the spin-lattice relaxation time, usually denoted by T1’, which in the absence of spin diffusion is related to the average concentration of paramagnetic ions (NP) by

Chapter 2. Applications of solid-state NMR in cement research 1 16   3 N 2P CI , ' 9 T1

CI 

e 2  e I S( S  1) 5 1   I e

37 (2.19)

I and I are the gyromagnetic ratio and the Larmor frequency of the I spins while e and e are the gyromagnetic ratio and the spin-lattice relaxation time of the electron spins, respectively.

2.2.6. Cross Polarization (CP)

The CP experiment (Figure 2.7) involves a magnetization transfer from one spin (I) to another (S) via the IS dipolar coupling obtained by matching the rf-field strengths applied on the I and S spins[105]. In structural investigations of cementitious materials, the CP experiment may be useful as a result of two main features[106]. First of all, according to the fact that the dipolar coupling is inversely proportional to the cube of the internuclear distance (1/r3IS), the CP pulse sequence can be used as a filter which only detects IS spin pairs with internuclear distances of usually less than 5 Å. Secondly, CP may be used to enhance the sensitivity for NMR experiments of nuclei (S) with low gyromagnetic ratio, low natural abundance and/or long spinlattice T1 relaxation times, such as 29Si and 13C[107,108]. In this case, the magnetization from an I spin, typically with a high natural abundance, large gyromagnetic ratio and short T1 relaxation time (e.g. 1H and 19F), is transferred to the dilute S spin. The sensitivity for the detection of the S spins in such an experiment is enhanced by a theoretical factor of I/S since the magnetization is initiated from the I spins. Furthermore, as the relaxation delay depends on the T1 relaxation time of the I spins, a substantial reduction in the spectrometer time may be achieved. The magnetization transfer requires that the rf-field strengths ( applied on the I and S spins in the spin-lock period fulfill the Hartmann-Hahn (HH) matching condition[109,110]: I(I  1 )  m´(m´  1 )ν1I 

S(S  1 )  m(m  1 )ν1S  nν R

(2.20)

where n = 1, 2, R is the spinning frequency, I and S are the spin-quantum numbers of the I and S spins, and m are the magnetic-quantum numbers defining the transitions between the Zeeman states (e.g. m  m  1). The HH-matching condition is simply reduced to 1I = 1S  nR for spin-systems where both I and S are spin-1/2 nuclei. If quadrupolar nuclei (S > ½) are involved in the spin system, the CP transfer during the spin-lock period become rather complex since the satellite transitions of S are substantially perturbed by the first-order quadrupolar interaction[111]. Moreover, MAS


38

causes the quadrupolar frequency to oscillate back and forth between Q. Depending on the magnitude of Q = 3CQ/[2S(2S  1)], the quadrupolar frequency experiences two or four zerocrossings per rotor cycle. These oscillations result in a population interchange between the 2S + 1 Zeeman states of the quadrupolar nucleus and thereby, a modulation of the HH-matching. Consequently, the CP transfer from the I to S spins is only efficient for the central transition of the quadrupolar nucleus. Three regimes have been observed for the CP behavior during the spinlock period[112]: adiabatic passage (>> 1), intermediate passage (~ 1) and sudden passage 2 /(ν Q ν R ) . Efficient spin-lock can be achieved (> 1S and (ii) ν1I  ν1S  nν R for Q D is fulfilled, the modulated dipolar frequency given by equation (2.14) becomes zero when integrated over a full rotor period (Tr) of MAS. In the REDOR experiment[117,119], the -pulses applied in the middle of each rotor cycle result in a sign-reversal of the dipolar frequency in the following half-rotor cycle (Figure 2.10). Consequently, the effective dipolar coupling becomes finite and gives rise to a dipolar dephasing of the magnetization; the attenuated signal is denoted by S. The dipolar coupling between the nuclear spins is monitored through a series of different experiments where the evolution period is increased systematically in steps of nTr. To account for T2 relaxation, a spin-echo experiment, i.e. when the I-channel (19F in Figure 2.10) is turned off, is performed on the S-spins (29Si) to produce the full signal (S0) for each evolution period. The REDOR fraction is obtained by subtracting the attenuated signal from the full signal (S = S0 – S). A plot of S/S0 as a function of the evolution period (nTr) gives the universal REDOR curves with distinct slopes and modulations, reflecting the dipolar coupling constant for the probed I-S spin pairs. The analytical solution for the dipolar dephasing of an isolated spin-pair, in which both I and S are spin-½ nuclei, can be expressed by an expansion using Bessel functions Jk(x)



ΔS  1  cos 4 2dnTr sin  β  cos β  sin α  S0

 

ΔS  1  J0 S0

2dnTr



 2  2k1 16k12  1 J k 

(2.21)

2dnTr

 2

(2.22)

Here, k is the order of the Bessel function[120]. The appearance of the REDOR curve becomes very complicated when multiple-spin systems (InS) are considered[121]. Furthermore, the REDOR sequence is only appropriate for studying nuclei with spin-quantum number of 1/2 since


42

the -pulses are inefficient on the satellite transitions for quadrupolar spins[122,123]. The quadrupolar coupling also introduces an oscillation of the REDOR fraction as a function of the evolution time. The REDOR experiment has been applied with success in structural investigations of a wide range of crystalline as well as amorphous materials[98,124-126]. This experiment is frequently used to identify and distinguish between independent InS connectivities. Furthermore, it emerges as a useful technique for determination of the IS distance of isolated two-spin systems.

Figure 2.10 29Si{19F} CP-REDOR pulse scheme. The 29Si19F dipolar coupling is re-introduced by a string of pulses which are phase-cycled in accordance with the xy-4 sequence[127,128], i.e., (xyxy)n. The -pulses are separated exactly by one half rotor period. The initial CP period[108] is applied to enhance the sensitivity of the experiment and to selectively detect the 29Si19F spin pairs.

2.2.9. Rotational-Echo Adiabatic-Passage Double Resonance (REAPDOR)

The REAPDOR pulse scheme[129] (Figure 2.11) is somewhat similar to the REDOR sequence. However, this experiment has been developed for spin systems that involve quadrupolar spins. In the experiment, it is utilized that the population of the 2S + 1 Zeeman states can be interchanged due to the oscillations of the quadrupolar frequency during MAS, cf. Section (2.2.6). This transfer process is facilitated by a single adiabatic-pulse of length , which should be shorter than one rotor period (Tr) to keep the oscillations of the quadrupolar frequency at a low number of zero-crossings[122,130]. In our studies, the REAPDOR experiments have employed an adiabatic pulse of length TR/3, which maximizes the number of odd zero-crossings for the quadrupolar frequency. Consequently, the oscillations caused by the quadrupolar coupling are almost eliminated from the REAPDOR fraction. A recent study of the REAPDOR sequence shows that the appearance of its S/S0 curve is determined by the spin-quantum number of the probed quadrupolar nucleus[122]. In general, the dipolar dephasing, reflected by the


43

REDOR and REAPDOR curves, becomes much complicated for multiple-spin systems. However, for small dipolar dephasing, S/S0  0.3, the slope of the curves can be approximated by a second-order polynomial[131,132] 1 ΔS  nTr 2 M2 S0 I  I  I π 2

(2.23)

M2 is the dipolar second moment, which reflects the strength of the dipolar coupling 2 4 2 2 2  μ0   1  M 2   γ I γ S     6   I  I  1 15  4 π  n  rIS 

(2.24)

I

where I is the spin-quantum number of the non-observed nuclear spin and nI is the number of I spins that contribute significantly to the I-S dipolar coupling. The remaining parameters used in equation (2.24) have their usual meanings according to standard abbreviations in NMR spectroscopy[133].

Figure 2.11 29Si{27Al} REAPDOR pulse scheme. The 29Si27Al dipolar coupling is re-introduced by the adiabatic pulse on the 27Al channel using a pulse length of 1/3 Tr. The experiment employs 1H decoupling during the dipolar evolution and the detection periods in order to remove (or reduce) strong hetero-nuclear dipolar couplings between the observed nuclei and the 1H atoms present in the system.

As mentioned in Section (2.2.1), the assignment of 29Si resonances from standard singlepulse

29

Si MAS NMR may be somewhat uncertain due to the overlap of chemical shifts for

different SiO4 environments caused by the Si4+ Al3+ substitution in their second-coordination sphere. In this context, the REAPDOR experiment may be an appropriate tool to distinguish


44

between different SiOAl connectivities and thereby, provides new insight to the interpretation of the

29

Si MAS NMR spectrum[134,135]. From the general structural features of

aluminosilicates[136], it is reasonable to consider an average SiAl internuclear distance (rSi-Al) for SiOAl bonds in the Q4 network structures since the bond lengths and bond angles of SiO4 and AlO4 tetrahedra only vary slightly for different SiOAl connectivities. This assumption reduces equation (2.23) to ΔS  k  n Al  nTr 2 S0

(2.25)

where k represents all the constant parameters, including rAl-Si, and nAl is the number of Al atoms substituted in the second coordination sphere of the observed silicon site. The applicability of the 29Si{27Al} REAPDOR experiment is tested for a synthetic sample of NaY zeolite. The

29

Si MAS NMR spectrum of the aluminosilicate network structure of the

zeolite shows four well-separated resonances in the spectral region from –80 ppm to –105 ppm (Figure 2.12). The difference

29

Si{27Al} REAPDOR spectrum clearly demonstrates that the

resonance at (29Si) = –100.0 ppm is unaffected by the reintroduction of the

29

Si27Al dipolar

coupling and therefore, it is assigned to a Q4(0Al) site. The remaining resonances exhibit REAPDOR curves with distinct slopes, which are illustrated in Figure 2.13. The values of knAl for each of the resonances were obtained from a fit of the REAPDOR fractions, S/S0  0.3, to equation (2.24). The resulting values are knAl = 0.16, 0.30 and 0.44 for the resonances at –94.8 ppm, 89.2 ppm and –83.9 ppm, respectively. However, it should be noted that it was necessary to include a data point for S/S0 > 0.3 in the curve fit for the resonance at –83.9 ppm since it has only one data point that fulfils the condition S/S0 ≤ 0.3. The corresponding number of Al atoms in the second-coordination sphere for each silicon site (i.e. nAl = 1, 2 and 3) are obtained by taking the relative ratios of their knAl values, since k is approximately constant for all Qi(nAl) components in the network structure. Thus, according to their isotropic chemical shift values in conjunction with the number of Al in their second-coordination sphere, the resonances at –94.8 ppm, 89.2 ppm and –83.9 ppm are unambiguously assigned to the Q4(1Al), Q4(2Al) and Q4(3Al) sites, respectively. This assignment is consistent with the structure of NaY zeolite proposed from earlier studies[80,137].


45

Figure 2.12 29Si{27Al} REAPDOR spectra (14.1 T, R = 10.0 kHz, nTr = 1.2 ms) for a synthetic sample of NaY zeolite: (a) reference spectrum (S0), (b) REAPDOR spectrum (S) in which the 29Si−27Al dipolar couplings have been reintroduced and (c) the difference spectrum (S0 − S).

Figure 2.13

29

Si{27Al} REAPDOR curves for a synthetic sample of NaY zeolite. The experimental REAPDOR

fractions S/S0 as a function of evolution times (Tr = 0.1 ms) are shown for the resonances at (■) 83.9 ppm, (▼) 89.2 ppm, () 94.8 ppm and (♦) 100.0 ppm. The curve fits of the REAPDOR fractions for small dephasing (S/S0 0.3) using the function S/S0 = knAl(nTr)2 are shown as solid lines.


46

2.2.10. Multiple-Quantum (MQ) MAS NMR

The MQMAS NMR experiment[138] plays a very important role in several recent investigations of half-integer spin quadrupolar nuclei. In general, the analysis of NMR spectra for the central transition of such spin nuclei is complicated by a strong second-order quadrupolar interaction. This interaction may introduce a severe line-broadening of the observed signals, causing an overlap of resonances from structurally different sites. On the other hand, in a twodimensional (2D) MQMAS NMR experiment, the second-order quadrupolar line-broadening is effectively removed in the isotropic dimension (F1) whereas it is retained in the anisotropic dimension (F2). The elimination of the second-order quadrupolar broadening in the MQMAS experiment significantly enhances the NMR spectral resolution for quadrupolar nuclei, enabling more complex quadrupolar spin systems to be studied. Several modifications of the MQMAS experiment have been developed with the aim of improving the sensitivity of the experiment[139143]

. It has also been applied with success in combination with other rf pulse schemes such as in

the CP-MQMAS to obtain a high-resolution hetero-nuclear correlation (HETCOR) spectrum[144]. This section gives a brief description of the basic two-pulse sequence for the triple-quantum MQMAS experiment (Figure 2.14). However, the concepts may be applied to any other multiquantum excitation schemes. The triple-quantum MQMAS experiment utilizes the fact that both the single- (1/2   1/2) and the triple- (3/2   3/2) quantum transitions are affected by the second-order quadrupolar interaction but unaffected by the strong first-order quadrupolar interaction. In the MAS axis system, the time averaged precession frequency due to the quadrupolar interaction for such symmetric transitions (m   m) is governed by[145] (0) (2) (2) ν Q  I , m, θ   ν (0) Q C I  m   ν Q α, β C I  m  P2 cos θ  (4)  ν (4) Q α, β C I  m  P4 cos θ 





P2 cos θ  

1 3cos 2θ  1 2

P4 cos θ  

1 35cos 4θ  30cos 2θ  3 8



(2.26)



where  is the angle between the spinning axis and B, and C(i)I are constant factors which depend of m and the spin-quantum number I.  and  are the Euler angles describing the orientation of the quadrupolar PAS relative to the rotor-axis MAS system. P2(cos) and


47

P4(cos) are the second-order and fourth-order Legéndre polynomials. It can be seen from

equation (2.26) that the geometrical dependency (3cos2 1) of the second-order Legéndre function can be removed by applying θ  arctan 2 , i.e. the magic angle. Furthermore, the term depending on the fourth-order Legéndre polynomial may be eliminated by correlation of two different symmetric transitions that fulfils the condition C I( 4 )(m1 )t1  C I( 4 )(m2 )t 2  0

(2.27)

In practice, the MQMAS NMR experiment employs an increment of t1 and the signal is acquired as a function of t2. Therefore, m2 must be 1/2 whereas m1 can be freely selected by applying the rules of phases cycling. Figure 2.14 shows the basic MQMAS pulse sequence for triple-quantum excitation (i.e. m1 =  3) with the corresponding coherence transfer pathway for a half-integer spin quadrupolar nucleus. It can be seen from equation (2.27) that the spin evolution as triplequantum coherence under the effect of the second-order quadrupolar interaction for a time t1 is refocused after a time t2 = t1k (where k = 7/9 and 19/12 for I = 3/2 and 5/2, respectively) after the conversion of the triple-quantum coherence to single-quantum coherence.

Figure 2.14 Schematic illustration of the pulse sequence and the coherence transfer pathway for a triple-quantum MQMAS NMR experiment. The selection of the triple- and single-quantum coherences are obtained by the excitation and mixing pulses, which are phase-cycled in accordance with i = 2/2N, where  is the phase of the pulse and N is a multiple of p. The triple-quantum coherences excited by the first pulse are evolved for a time t1 before these transitions are converted to single-quantum coherence by the second pulse. The acquisition begins immediately after the second pulse to gain the best S/N ratio, although the echo occurs at a time t2 = t1k (where k = 7/9 and 19/12 for I = 3/2 and 5/2, respectively) after the second pulse. Furthermore, the phase sensitivity in the indirect dimension F1 is conducted by the hyper-complex phase-cycling method[139].


48

The advantage of the MQMAS NMR experiment is illustrated for a sample of gibbsite, Al(OH)3 (Figure 2.15). The projection in the F2 dimension, under ideal conditions, corresponds to a standard single-pulse

27

Al MAS NMR spectrum, in which the second-order quadrupolar

line-broadening has not been removed. The interpretation of this spectrum can be somewhat uncertain since the observed second-order lineshape can potentially be misinterpreted as an overlap of several resonances. On the other hand, the projection of the F1 dimension in conjunction with the contour plot clearly identify two well-separated resonances, where the shoulders at 4 ppm, –3.5 ppm and 9 ppm seen in the F2 dimension correspond to the singularities and shoulders for the Al site with the strongest quadrupolar coupling. The observation of two distinct crystallographic Al sites by MQMAS NMR is consistent with the crystal structure for gibbsite determined by XRD[146,147].

Figure 2.15

27

Al MQMAS NMR spectrum (9.4 T, R = 10.0 kHz) of gibbsite, -Al(OH)3. The spectrum was

recorded using the three-pulse z-filter sequence and a 144-steps phase cycle. Furthermore, 1H decoupling (TPPM) was employed during both the evolution of the triple-quantum coherence (t1) and the detection (t2) periods. The asterisks (*) indicate spinning side bands in the F1 dimension. Projections onto F1 and F2 dimensions, corresponding to summations, are shown on the left side and above the 2D spectrum, respectively.

Chapter 3. Fluoride Mineralization

49

3. Chapter

Fluoride Mineralization

Since its first application in clinker preparation, probably in the late 1800s by Michaelis, fluoride mineralization has got much attention in cement research[41,42,148]. Due to its high natural abundance, CaF2 (e.g., fluorspar) is by far the most widely used fluoride mineralizing agent in modern cement production. However, other fluoride-containing compounds such as NaF2, MgF2, Na2SiF6 and MgSiF6 have also shown similar properties[25,149]. In addition to the mineralizing properties, fluoride also exhibits fluxing properties and seems to be involved in several chemical reactions during clinker formation. At a kiln temperature of about 1100 C, fluoride reacts with the

raw

meal

components

forming

a

clinker

melt[150,151],

largely

consisting

of

11CaO7Al2O3CaF2. As belite and lime are brought together within the liquid phase, alite is slowly formed. The alite formation accelerates rapidly as the kiln temperature is raised just above 1200 C; approximately 40 % of alite is already formed at 1200 C and about 60 % at 1300 C. At kiln temperatures above 1300 C, 11CaO7Al2O3CaF2 decomposes to tricalcium aluminate and CaF2. It has been demonstrated in earlier studies of synthetic alite that a small amount of fluoride can be incorporated in this phase, most likely in accordance with an O2  F substitution. Charge balance can be preserved by the incorporation of a similar amount of Al3+ ions substituting for Si4+ in the alite structure, tentatively forming a solid solution with the chemical formula Ca3[Si1-xAlx][O5-xFx], where the upper limit for x is approximately 0.15[9]. This chapter presents the results from a study of the mineralizing effect of fluoride on the formation of the calcium silicate phases of Portland cement. Furthermore, the site preference of F and its coupled substitution mechanism with Al3+ and Fe3+ ions have been investigated using solid-state NMR. Finally, the results are applied in an optimization of the alite content in Portland clinker.


50

3.1.

Site preferences of F ions in the calcium silicate phases of Portland cement

3.1.1.

19

F MAS NMR

The

19

F MAS NMR spectrum of a commercial white Portland clinker (wPc) containing

only 0.04 %w F is shown in Figure 3.1. The spectrum exhibits a characteristic lineshape covering a spectral region from about –100 ppm to –125 ppm, which can only be satisfactorily simulated by including at least three 19F resonances with (19F) = –122.1 ppm, –116.5 ppm and –111.4 ppm (Figure 3.1 b-c).

19

F MAS NMR spectra of modified clinkers prepared from wPc,

but containing higher F and Al2O3 contents, show somewhat similar lineshape features with a centre of gravity at (19F)  114.9 ppm, cf. Figure 3.10. This indicates that all F atoms occur in similar structural environments, but in a less-ordered arrangement. Since the values of (19F) for fluoride ions cannot be distinguished from those for SiF and AlF covalent bonds, due to the strong effect of different types of counter ions on the 19F chemical shiftss, an exact identification of the fluorine environments in Portland clinker cannot be extracted from the 19F chemical shift alone. However, their chemical shifts exclude them from being considered as the crystalline phases Ca4Si2O7F2 ((19F) = –101.6 ppm and –106.1 ppm), CaF2 ((19F) = –105.9 ppm) or the alumina-rich phase 11CaO7Al2O3CaF2, which exhibits two different

19

F resonances with a

large spinning sideband intensities resulting from the CSA interactions. Furthermore, recent studies of aluminosilicate glasses demonstrated that AlF and SiF covalent bonds are only found in glasses with much higher F contents as compared to that in Portland clinker[152,153].

Figure 3.1

19

F MAS NMR spectrum (7.05 T, R = 10.0 kHz) of a commercial white Portland clinker (wPc)

containing 0.04 %w F. The experimental spectrum (a) can be simulated, shown in (b) and (c), by three resonances at

(19F) = –122.1 ppm, –116.5 ppm and –111.4 ppm.

Chapter 3. Fluoride Mineralization 3.1.2.

29

51

Si{19F} and 27Al{19F} CP/MAS NMR

The incorporation of fluoride ions in the calcium silicate phases of Portland clinker has been investigated for a series of fluoride-mineralized clinkers using

29

Si{19F} and

27

Al{19F}

CP/MAS NMR. Since the magnetization is transferred through the dipolar coupling (d  1/rIS3), the CP period acts as a filter which only allows the 29Si and 27Al spins that are located in the near vicinity to a

19

F spin to be detected. The studied clinkers were modified from a commercial

white Portland clinker to have a high Al2O3 bulk content of 4.3 %w, a lime saturation factor (LSF) of 0.90 and fluoride contents in the range of 0.04  1.0 wt. % F. The bulk fluoride contents in selected samples have been measured with an ion-selective electrode, after the samples was dissolved in a mixture of HCl and KAl(SO4)2. These clinkers are used as reference samples in the quantification of the fluoride content from

19

F MAS NMR experiments for the

remaining samples. Furthermore, the free lime content in all modified clinkers has been measured. The experimental details for sample preparations and analyses are given in Appendix 1 and 2, respectively.

Figure 3.2 29Si{19F} CP/MAS NMR spectra (7.1 T) of three selected fluoride-mineralized clinkers containing 0.23 %w F (a), 0.47 %w F (b) and 0.77 %w F (c). The spectra were recorded with a spinning speed of R = 3.0 kHz and a 2.0 ms CP contact time. The overall lineshape observed in these spectra resembles closely the spectrum of alite in its monoclinic MIII form, see Figure 2.

29

Si MAS NMR

Chapter 3. Fluoride Mineralization 29

52

Si{19F} CP/MAS NMR spectra of selected modified clinkers are shown in Figure 3.2.

The absence of the belite resonance at (29Si) = 71.33 ppm in these spectra demonstrates that the fluoride ions are only incorporated in the alite phase. Furthermore, they all exhibit broadened lineshapes with features resembling closely that of alite in the MIII form. In contrast to this work, most of the previous studies on the structure of fluoride-mineralized alite propose that this phase is stabilized in its rhombohedral form[9,154,155]. However, the polymorphic form of the alite phase when present in real Portland cement may as well be controlled by several other factors[26] such as the presence of other impurities, the cooling rate, etc. As considered earlier, the O2  F substitution should be charge-balanced by a similar amount of Al3+ guest ions substituting into the tetrahedral silicon sites of alite. The location of these Al3+ guest ions are, of course, important for the thermodynamic properties of the fluoride mineralization since a random distribution of F and Al3+ guest ions in alite yields the highest possible number of atomic configurations and thereby, maximizes the entropy of mixing for this phase. However, a random distribution of F and Al3+ guest ions will create local regions with formula units (Ca3SiO4F)+ and (Ca3AlO5), which are probably energetically unfavorable. On the other hand, the charge balance can be preserved locally if Al3+ guest ions are located in the proximity to the fluoride ions. Consequently, the incorporation of these Al3+ guest ions will not affect the entropy of mixing for the alite phase significantly. 27Al MAS and 27Al{19F} CP/MAS NMR spectra of a modified clinker containing 0.77 %w F are shown in Figure 3.3. The singlepulse spectrum (Figure 3.3) shows a sharp

27

Al resonance at approximately 75 ppm, which

originates from the tetrahedrally coordinated Al3+ guest ions in the calcium silicate phases. In addition to the fact that fluoride is only incorporated in the alite structure, the appearance of the resonance at (27Al) ~75 ppm in the 27Al{19F} CP/MAS spectrum (Figure 3.3 b), reveals that the fluoride ions are located in the near vicinity of the Al3+ guest ions. Furthermore, the HHmatching condition of F = 3Al + R applied in this experiment to achieve magnetization transfer reflects that the observed 27Al spins possess a large quadrupole coupling constant, which is consistent with the values reported previously for Al3+ guest ions in the alite phase. The observation clearly clarifies the essential role of an aluminum source in fluoride mineralization as proposed by Shame and Glasser[9], since charge balance is achieved locally by a coupled incorporation of F and Al3+ ions.


Figure 3.3 (a)

27

Al MAS and (b)

53

27

Al{19F} CP/MAS NMR spectra (7.1 T and R = 5.0 kHz) of a fluoride-

mineralized clinker containing 0.77 %w F. The magnetization transfer from the

19

F spins to the

27

Al spins is

obtained using a Hartmann-Hahn matching condition of F = 3Al + R and a contact time of 1.5 ms. The asterisks (*) indicate spinning sidebands.

3.1.3.

29

Si{19F} CP-REDOR NMR

The structures of the alite polymorphs include two different crystallographic oxygen sites, as described in section 1.2.1: the covalently bonded oxygen in SiO4 tetrahedra (SiOb) and the interstitial oxygen ions (SiOi). Their corresponding average SiO internuclear distances (rSi-O) are 1.63 and 4.32 Å, respectively, for the monoclinic MIII structure. The mineralizing properties of fluoride ions may strongly depend on which types of those oxygen sites that are accessible for O2  F substitution. In order to elucidate the site preference of fluoride ions in the alite structure, the average internuclear distance of the SiF spin pairs has been measured using 29

Si{19F} CP-REDOR NMR. The initial CP period acts as a 29Si19F filter by only detecting the

29

Si nuclear spins which are dipolar coupled to

19

F, i.e., roughly rSiF < 5 Å. Subsequently, the

dipolar re-coupling ability of the REDOR sequence allow a measurement of the 29Si19F dipolar coupling constant, from which the average SiF internuclear distance can be calculated. Furthermore, the CP period enhances the sensitivity of the experiment since the magnetization is initiated from the 19F spins, which have a high gyromagnetic ratio and a short relaxation delay of only 8 s.


Figure 3.4 29

29

54

Si magnetization (Mx(t) given in arbitrary units) as a function of the CP contact time obtained from

Si{19F} CP/MAS NMR experiments (7.1 T, R = 5.0 kHz) for a modified clinker containing 0.77 wt.% F (■) and

for cuspidine (). The experiments employed rf fields of SiB1/2 = 47.5 kHz and FB2/2 = 42.1 kHz, and a relaxation delay of 8 s. The solid curves illustrate the optimum fits to the experimental data for the clinker and cuspidine to equation (3.1) and (3.2), respectively.

The cross-polarization dynamics curves for a modified clinker containing 0.77 %w F and a synthetic sample of cuspidine are shown in Figure 3.4. Considering the cross-polarization curves for the applied contact times (t  8.0 ms), the cross-polarization time constant (TSiF), which reflect the strength of the SiF dipolar couplings, may be obtained by fitting the experimental data to[106] Mx(t) = [M0/ TSiF/T1,F exp(tcp/T1,F) [1  exp((1/ T1,F  1/TSiF) tcp)]

(3.1)

Mx(t) = M0[1  exp(  tcp/TSiF)]

(3.2)

and

for the clinker and cuspidine, respectively. In equation (3.1), it is assumed that the 29Si rotatingframe relaxation time is at least an order of magnitude larger than the corresponding value for 19

F (i.e., T1,Si >> T1,F). On the other hand, equation (3.2) assumes that both T1,Si and T1,F are

much larger than TSiF. M0 is the 19F magnetization immediately after the initial 90º pulse on the 19

F-channel and tcp is the CP contact time. An optimized fit corresponding to the parameters TSiF


55

= 4.61 ms and T1,F = 4.57 ms has been obtained for the modified clinker. On the other hand, the SiF2 three-spin spin system of cuspidine, with SiF distances of 3.88 Å and 4.00 Å, exhibits a CP time constant of only 1.20 ms. Owing to the low natural abundance of the 29Si nuclear spins (4.7 %) and the low fluoride content (0.77 %w) of the Portland clinker, it is reasonably to consider the

29

Si19F spin-pairs as isolated. In the absence of spin diffusion, the SiF spin

system may be approximated by a rigid lattice model, where the cross-polarization time constant TSiF becomes inversely proportional to the dipolar second moment M2, whose magnitude 6 ) and the number of F atoms involved in the dipole coupling depends on the SiF distance ( rSiF

(see equation 2.24). According to this assumption, the factor of nearly four between the values of the CP time constant for the SiF spin system in Portland clinker and the three-spin system, SiF2, in cuspidine indicates that the rSiF distances in alite is slightly longer than that for SiF in cuspidine structure. Results from the 29Si{19F} CP-REDOR experiment are summarized in Figure 3.5 and 3.6. The experimental data are plotted along with their optimized simulation and two simulated universal REDOR curves for the average SiOb and SiOi distances in monoclinic alite (Figure 3.6). As a result of the severe overlap of resonances and the low signal to noise ratio of the REDOR spectra (Figure 3.5), it has not been possible to distinguish between the different

29

Si

resonances from the eighteen crystallographic silicon sites of monoclinic alite. Therefore, they have been considered to be equal in the analysis of the REDOR spectra. A dipolar coupling constant of d = 285 Hz, corresponding to an average internuclear distance of 4.29 Å for the isolated SiF spin-pair in alite, was obtained from a numerical fitting of the experimental REDOR fractions to equation (2.21), see Appendix 2. The magnitude of the SiF internuclear distance as well as the slopes of the REDOR curves shown in Figure 3.6 clearly demonstrate that the fluoride anions are located in the interstitial oxygen site, i.e., the fluoride anions not involved in covalent SiF bonds. These data are obtained for an optimized CP contact time of 3.0 ms (Figure 3.4). However, selected REDOR experiments for CP contact times of 1.0 and 7.0 ms give S/S0 fractions of the same order of magnitude as observed for the 3.0 ms contact time. The experimental observation for the fluoride environments in the alite phase of Portland cement is consistent with a recent theoretical investigation of the coupled incorporation of F and Al3+ guest ions in the alite phase applying Density Functional Theory (DFT) calculation and structure optimizations[156]. These calculations of energies for the Si4+ + O2  Al3+ + F substitution propose that the F ion is preferentially substituted for the interstitial oxygen site, since it requires a much lower substitution energy as compared to the value for the formation of SiF covalent bonds.

56


Figure 3.5 29Si{19F} CP-REDOR NMR spectra (7.1 T, R = 10.0 kHz) of a fluoride-mineralized Portland cement containing 0.77 %w F. (Left) Full signal (S0), which is not affected by the SiF dipolar coupling. (Right) The attenuated signals (S), reflecting the intensity reduction caused by the SiF dipolar couplings. The evolution periods are (a): 4 Tr, (b): 16 Tr, (c): 28 Tr and (d) 40 Tr, where Tr = 0.1 ms.

Figure 3.6 Experimental REDOR fractions (S/S0) for the

29

Si–19F spin-pairs in the alite phase of a modified

Portland clinker (0.77 %w F). The experiments employ CP contact times of 1.0 ms (o), 3.0 ms () and 7.0 ms (x). Numerical fitting of the experimental data (solid line) results in a rSi-F distance of 4.29 Å. The numerical simulations for rSi-F of 1.6 Å and 4.3 Å are illustrated by the dashed lines, corresponding to the mean Si–O bond length of the SiO4 tetrahedra and the mean distance for the interstitial oxygen atoms (Oi) to their nearest Si atoms, respectively.

57

Chapter 3. Fluoride Mineralization 3.2.

Mineralizing effects of calcium fluoride and calcium sulphate

The main effect of fluoride mineralization may be addressed to the fact that the coupled incorporation of F and Al3+ ions increases the entropy of mixing ( S mix  k  ln(Ω ) ) for the alite phase. However, due to the local charge-preserving effect of F and Al3+, only the concentration of F contributes effectively to the increase in mixing entropy, where the total number of possible atomic configurations[157] is given by Ω  ninterstitial ! /( nF ! nO ! ) ; ninterstitial is the total number of the interstitial sites, i.e. nine in the triclinic unit cell, Ca27Si9(Ob)36(Oi)9, while nF and nO are the numbers of the interstitial sites occupied by F and O2, respectively. Thus, the

entropy of mixing associated with fluoride mineralization for the alite phase has its maximum when nF = 4.5, corresponding to the incorporation of 4.1 %w F in alite, or x = 0.1 for the formula Ca3[Si1-xAlx][O5-xFx]. However, this content is much lower as compared to the fluoride level of x = 0.15, reported in a previous study using the selective chemical dissolution method[9]. One possible explanation for this discrepancy is that at high fluoride contents, the F atoms may also form AlF and/or SiF covalent bonds. At sufficiently low fluoride levels, the ethalpy (H) for the alite phase may not be affected by the presence of fluoride ions. On the other hand, the Gibbs free energy (G) for alite formation from belite and lime will be more negative[157] ΔG  ΔH  ΔS  T Δ  alite  belite  lime ,

(3.3)   G, H , S

where T is the temperature at which the reaction takes place. Alternatively, for complete reaction between belite and lime, the amount of uncombined CaO, i.e. the free lime content, can be used to monitor the thermodynamic stability of alite relative to belite. For a given burning condition, a high free lime content indicates that belite is more favorable as compared to alite whereas a low free lime content indicates the reverse. Calcium sulphate is another widely use mineralizing agent in Portland clinker production[42,148,158], either alone or in combination with CaF2. In contrast to F, the S6+ ions are preferentially incorporated in the belite structure[38], in accordance with the tentative coupled substitution mechanism, 3Si4+  S6+ + 2Al3+. The mineralizing effect of CaSO4 has been reexamined for a series of modified clinkers prepared from raw meals including 4.3 %w Al2O3 (~12 %w tricalcium aluminate), a LSF of 1.0 and SO3 contents in the range 0  3 %w. The 29Si MAS NMR spectra for these samples are shown in Figure 3.7. It is apparent from the spectra that the belite resonance exhibits an increasing line-broadening as a function of the SO3 content,


58

supporting the assumption of an increased incorporation of S6+ in the belite structure. Furthermore, in addition to an increase in the belite intensity with increasing SO3 content (Figure 3.7), the amount of uncombined CaO is increased rapidly as a function of the SO3 content, Figure 3.8. This clearly demonstrates that the presence of a SO3 source in the raw meal increases the stability of the belite phase and thereby, reduces the quantity of alite. It is also apparent from Figure 3.8 that SO3 has a rather high volatility since the actual SO3 contents are much lower than the SO3 amount added to the corresponding raw meals. In a similar way, the mineralizing effect of calcium fluoride is demonstrated for a series of modified clinkers prepared from raw meals including a LSF of 1.0, 4.3 %w Al2O3, 3.0 %w SO3 (the actual SO3 content in the clinkers is approximately 2.3 %w) and fluoride contents in the range 0.04  1.0 %w F. The strong mineralizing effect of calcium fluoride is recognized by the rapid decrease in the amount of uncombined CaO as a function of the fluoride content, Figure 3.8. It can be seen that the stability of SO3-mineralized belite is effectively reduced by the addition of calcium fluoride. It only requires 0.6 %w F to retain the free CaO content at ~0.4 %w, i.e. the same level of uncombined CaO as in the clinker without using any mineralizing agents. The mineralizing effects of fluoride and SO3 on alite and belite, respectively, are graphically illustrated in Figure 3.9.

Figure 3.7 29Si MAS NMR spectra (7.1 T, R = 7.0 kHz) of SO3-mineralized clinkers prepared from a commercial white Portland clinker. The actual SO3 content in the clinkers is 0.48 %w SO3 (a), 1.09 %w SO3 (b) and 1.91 %w SO3 (c). Furthermore, the preparation of the clinkers used a LSF of ~1.0 and a bulk Al2O3 content of ~4.3 %w.


59

Figure 3.8 Mineralizing effects of CaSO4 and CaF2 reflected by the quantity of uncombined CaO (free lime) in the clinkers as a function of the SO3 and F contents, respectively. (Left) The graph shows the free CaO content (■) and the actual SO3 content (♦) in the mineralized clinkers (LSF ~1.0, 0.04 %w F) as a function of the SO3 quantity added to the raw meal. (Right) Free lime content as a function of the actual fluoride content in the mineralized clinkers (LSF ~1.0, 2.3 %w SO3). The actual SO3 contents of the clinkers have been determined by X-ray fluorescence.

Figure 3.9 Graphical illustration of the effects of fluoride and SO3 mineralization on the thermodynamic stability of alite and belite, respectively.


60

Secondary effects of fluoride mineralization

In order to explore non-thermodynamic effects of fluoride mineralization, a series of clinkers prepared from raw meals containing 4.3 Al2O3 (~12 %w tricalcium aluminate), a LSF of 0.9 and fluoride contents in the range 0.04  1.0 %w have been investigated. Experimental details including the sample preparation and analysis are given in Appendices 1 and 2. The 19F MAS NMR spectra of the clinkers show a single broad resonance with a centre of gravity at 114.9 ppm, Figure 3.10. However, the lineshape exhibits some variation as a result of the increased fluoride content. This observation confirms that for low concentrations (> 0.77 %w) the fluoride ions are only incorporated in the interstitial oxygen sites of alite, which is consistent with earlier discussed investigations[10,150], since excess phases of CaF2 ((19F) = –105.9 ppm), cuspidine ((19F) = –101.6 ppm and –106.1 ppm) or 7CaO11Al2O3CaF2 (two 19F sites with large CSA’s) are not formed in these systems. However, due to the moderate volatility of CaF2, the actual fluoride contents in the clinkers are somewhat lower than that used in the clinker preparation.

Figure 3.10 19F MAS NMR spectra (7.1 T, R = 10 kHz) of selected fluoride-mineralized clinkers modified from a commercial white Portland clinker to have a high bulk aluminate content of 4.3 %w (~12 %w tricalcium aluminate) and fluoride contents of 0.04 (a), 0.23 (b), 0.36 (c) and 0.77 %w F (d). The broad resonance with a centre of gravity at (19F)  114.9 ppm indicates that the fluoride anions have nearly identical structural environments, but they are present in a less-ordered arrangement.


Figure 3.11

27

61

Al MAS NMR spectra (14.1 T, R = 13.0 kHz) of selected fluoride-mineralized clinkers modified

from a commercial white Portland clinker (0.04 %w F, 2.13 %w Al2O3) to have a relatively high bulk aluminate content of 4.3 %w (~12 %w tricalcium aluminate) and fluoride contents of 0.04 (a), 0.32 (b), 0.48 (c), 0.69 (d) and 0.77 %w F (e). The asterisks (*) indicate the spinning sidebands for spectrum (a) while the diamond () marks the 27

Al centerband from an AFt phase (ettringite), which most likely reflects a small degree of pre-hydration.

The correlation between the incorporation of fluoride and aluminum ions in alite is apparent from the

27

Al MAS NMR spectra of the selected samples shown in Figure 3.11. The

increasing intensity for the 27Al resonance at ~82 ppm illustrates that the increased incorporation of fluoride ions in alite is accompanied by an increase in the quantity of Al3+ guest ions in this phase. On the other hand, the 27Al resonance from the tricalcium aluminate phase, which appears as a broad lineshape from approximately 30 ppm to 80 ppm, exhibits a decrease in its intensity

62


as the fluoride content is increased, indicating that this phase is partially consumed by the increased incorporation of Al3+ ions in alite. Therefore, the coupled substitution of F and Al3+ ions in alite may be described by the reactions Ca27Si9(Ob)36(Oi)9 + 0.5xCaF2  [Ca27Si9(Ob)36(Oi)9-x(Fi)x]x+ + 0.5xCaO

(3.4)

[Ca27Si9(Ob)36(Oi)9-x(Fi)x]x+ + 0.5x3CaOAl2O3  Ca27Si9-xAlx(Ob)36(Oi)9-x(Fi)x + xSiO2 + 1.5xCaO xSiO2 + 2xCaO  x2CaOSiO2

(3.5) (3.6)

The reactions only consider the incorporation of F and Al3+ guest ions, assuming that no other guest ions are present in the solid solution. The alite phase is expressed with the formula for one unit cell of its triclinic polymorph[20], in which it is discriminated between the covalently bonded (SiO) and interstitial (non-bonded) oxygen sites. In fact, the presence of fluoride may also results in incorporation of Al3+ ions in the alite phase by other mechanisms since Figure 3.11d indicates that the majority of the bulk Al2O3 content of 4.4 %w (~ 8.610-2 mol Al3+) is incorporated in this phase. On the other hand, the fluoride content in this sample only corresponds to 1.810–2 mol F. The distribution of Si atoms in the calcium silicate phases, quantified by the molar ratio of Si in alite and belite, nSi[alite/belite], is achieved from the

29

Si MAS NMR spectra using a

deconvolution procedure described previously. Although the broad sub-spectra for MI and MIII alite originate from an overlap of resonances from their eighteen different crystallographic silicon sites, a satisfactory simulation can be achieved by only including nine different

29

Si

resonances, Figure 3.12. In this work, the MI lineshape has only been employed for the reference clinker shown in Figure 3.12, which contains 0.04 %w F. The deconvolutions of the 29Si MAS NMR spectra for the remaining fluoride-mineralized clinkers include a sub-spectrum of alite in its MIII form, with small modifications only at the tails of the overall peak. The nSi[C3S/C2S] molar ratio, resulting from the deconvolutions, as a function of the actual fluoride content is illustrated in Figure 3.13. This graph clearly demonstrates that for a fixed chemical composition an increased incorporation of fluoride ions in alite tends to decrease the alite to belite molar ratio. However, this decrease in nSi[alite/belite] does not reflect a decrease in the actual alite content since the quantity of uncombined CaO in these clinkers is rather constant, i.e. approx. 0.4 %w free CaO.


63

Figure 3.12 Simulations of 29Si MAS NMR spectra for two modified clinkers with LSF ~0.9, 4.3 %w Al2O3 and 0.04 %w F (left) and 0.32 %w F (right), prepared from a commercial white Portland clinker. (Left) The spectrum simulation (b) includes one resonance at (29Si) = 71.33 ppm for the belite phase (c) and a broad sub-spectrum of alite in its MI form (d). (Right) For the fluoride-mineralized clinker (a), a sub-spectrum of the MIII form of alite is employed (d). Both simulations include only nine different 29Si resonances for the sub-spectra of alite.

The results from the 19F, 27Al and 29Si NMR experiments demonstrate first of all that the incorporation of fluoride ions in the interstitial oxygen sites of alite increases the amount of Al3+ guest ions substituting for Si4+ on the tetrahedral sites, for which the tricalcium aluminate phase acts as the aluminate source for the fluoride mineralization, equation (3.4) and (3.5). The coupled substitution of F and Al3+ guest ions stabilizes local regions with the composition Ca27Si9-xAlx(Ob)36(Oi)9-x(Fi)x in the alite structure. Secondly, belite is formed from the formally released quantities of SiO2 and CaO, equation (3.6), according to the fact that “free SiO2” has not been observed in the 29Si MAS NMR experiments and that the free CaO content in the clinkers is constant. Thus, the decreased nSi[alite/belite] molar ratio as a function of the fluoride content may account for the Si4+  Al3+ substitution on the tetrahedral sites of alite and an increase in the quantity of belite formed from the released SiO2 at the same time. Moreover, equations (3.4  3.6) reveal that the alite content is almost unaffected by the quantity of fluoride whereas the chemical composition of this phase is significantly changed by the fluoride mineralization. This secondary effect may be compensated by additional CaO according to the fact that fluoride mineralization increases the thermodynamic stability for alite only, Figure 3.8.


64

Figure 3.13 The molar ratio of Si atoms in the alite and belite phases, nSi[alite/belite], for the fluoride-mineralized clinkers. The graph is based on data obtained from deconvolutions of the 29Si MAS NMR spectra.

65


Incorporation of Fe3+ ions in the calcium silicate phases of Portland cement

The iron content in ordinary Portland cement is typically around 3 %w Fe2O3, but much lower in white Portland cement. The main constituent iron phase is ferrite (Ca2(AlxFe1-x)2O5), although a small amount of Fe3+ may also be present as guest ions in tricalcium aluminate and the calcium silicate phases[29,31,159]. Thus, these Fe3+ guest ions may be involved in the fluoride mineralizing process in a similar manner as the Al3+ ions. Considering the

27

Al MAS NMR

spectrum of the modified white clinkers shown in Figure 3.11 (e), it can be seen that the majority of Al2O3 must be incorporated in the alite phase, partly associated with the fluoride mineralization. Following this result, a series of fluoride-mineralized clinkers has been prepared from raw meals containing 1.0 %w F and a LSF of 0.9. Furthermore, the bulk Al2O3 content of 4.3 %w is partly replaced by additional Fe2O3, targeting Fe/Al molar ratios in the range of 0.05  0.8. The corresponding bulk Fe2O3 contents are in the range of 0.3  4.0 %w. The potential of Fe3+ ions in replacing Al3+ in the fluoride mineralizing process has been investigated for this series of modified clinkers using

29

Si MAS NMR,

29

Si IR NMR and XRD. This study also

reveals the influences of an increased incorporation of Fe3+ ions in the calcium silicate phases of Portland cement, with the main findings presented below.

Figure 3.14

29

Si MAS NMR spectra (7.05 T, R = 5.0 kHz) of two modified clinkers (LSF ~0.9, 0.9 %w F)

containing 0.38 %w Fe2O3 (a) and 3.33 %w Fe2O3 (b). The spectra demonstrate that the 29Si resonances from belite as well as alite are strongly affected by the increased bulk Fe2O3.

66


The 29Si MAS NMR spectra of two modified clinkers with different bulk Fe2O3 contents, shown in Figure 3.14, clearly demonstrate that the presence of paramagnetic Fe3+ ions in Portland cement introduces a significant line-broadening and intensity loss on the observed nuclear spins. Recently, it has been demonstrated that the spin-lattice relaxation process for the 29

Si spins in the calcium silicate phases of ordinary Portland cements is predominantly affected

by the Fe3+ ions incorporated as guest ions in these phases[78], rather than by the Fe3+ ions present in the separate ferrite phase as proposed earlier[160]. This observation been reexamined for two modified Portland clinkers containing 4.3 %w Al2O3, LSF ~ 1.0 and Fe2O3 contents of 1.5 %w and 2.0 %w. The aluminate and ferrite phases were removed from the clinkers using a selective dissolution method[161]. The validity of this method is supported by the absence of the 2reflexions at ~33.2 and ~33.8 in the XRD pattern of the clinkers after selective dissolution, Figure 3.15. On the other hand, the

29

Si IR NMR spectra shown in Figure 3.16 for the same

clinkers before and after selective dissolution show no significant differences, which

C3S

14000 12000 10000

C3A

6000 4000

C4AF

8000 C3S

Arbitrary Unit

C3S

C3S

16000

C3S

demonstrates that the ferrite and aluminate phases only have a small effect on the 29Si relaxation.

2000 0 28

29

30

31

32

33

34

35

2 Figure 3.15 XRD diffractograms of a modified clinker (4.3 %w Al2O3, ~2.0 %w Fe2O3 and LSF ≈ 1.0). The XRD pattern for the original clinker is shown by the solid line while the dashed line corresponds to the pattern after selective dissolution of the aluminate (C3A) and ferrite (C4AF) phases.

67


Figure 3.16 29Si IR NMR spectra (7.1 T, R = 5.0 kHz) of a modified clinker (4.3 %w Al2O3, ~2.0 %w Fe2O3 and LSF ≈ 1.0). The 29Si IR NMR spectra of the same sample before and after selective dissolution are shown in parts (a) and (b), respectively, where the recovery times (in seconds) are shown below the individual spectra.

Figure 3.17 shows the results from the 29Si IR NMR experiments, including experimental data together with their best fits to equation (2.18), for the alite and belite phases of three clinkers with bulk Fe2O3 contents of 0.38, 0.93 and 3.91 %w. The Mz(t) values for the alite and belite phases have been obtained for each of the

29

Si IR NMR spectra employing the

deconvolution procedure described earlier. It shows clearly that the relaxation times for the 29Si


68

resonances from alite as well as belite are affected by the increase in the bulk Fe2O3 content, reflecting an increased incorporation of the paramagnetic Fe3+ ions in the alite and belite structures. However, the actual concentration of Fe3+ ions incorporated in the alite and belite phases can not be obtained from these experiments since the spin-lattice relaxation time for the electron spins has not been determined, cf. equation (2.19). It is evident from the plot of

29

Si

spin-lattice relaxation times (T´1) for alite and belite as a function of the bulk Fe2O3 content, Figure 3.18, that the increase in the amount of Fe3+ guest ions (1/T´1  NP2) is very similar for these two phases at low bulk Fe2O3 concentrations (i.e. < 0.8 %w Fe2O3), indicating that the presence of F ions in alite has no significant impact on the Fe3+ guest-ion incorporation in the calcium silicate phases. Therefore, Fe3+ seems not to be involved in the fluoride mineralization. The slight change in slope for 1/T´1 for bulk Fe2O3 contents above 0.8 %w suggests that different amounts of Fe3+ ions are incorporated in alite and belite, potentially in different sites as mentioned by Taylor[13] who suggests that the Fe3+ ions enter the tetrahedral sites in belite by substituting for Si4+ ions, but occupy the octahedrally coordinated sites in alite obtained by substitution for Ca2+ ions.

Figure 3.17 Plots of Mz(t) values for alite (left) and belite (right) as a function of the recovery time obtained from the 29Si IR NMR experiments for three modified clinkers containing 0.38 %w Fe2O3 (■), 0.93 %w Fe2O3 (♦) and 3.91 %w Fe2O3 (●). The best fits of the experimental data to equation 2.13 are shown as solid lines.


69

Figure 3.18 A plot of the relaxation rate (1/T´1) for the 29Si spins in alite (■) and belite () versus the bulk Fe2O3 content for a series of modified clinkers. The clinkers were prepared from a commercial white Portland cement, targeting LSF ~0.9, 0.9 %w F and Fe/Al molar ratios in the range of 0.05  0.8. The relaxation time constant T´1 for each clinker was obtained from 29Si IR NMR experiments.

The proposed Ca2+  Fe3+ substitution is further supported by an evaluation of the alite and belite content for each of the studied modified clinkers by a plot of the molar ratio of silicon in alite and belite (nSi(alite/belite)) as a function of the bulk Fe2O3 content, Figure 3.19. For small concentrations (< 0.8 %w Fe2O3), the incorporation of Fe3+ ions in the calcium silicate phases results in an increase of the alite content, i.e. the nSi(alite/belite) ratio, for a fixed clinker composition apart from the variation in the Fe/Al molar ratio. Since the amount of free CaO is nearly constant (~0.4 %w) for all clinkers, the increase in the alite content may potentially reflect that the Fe3+ ions preferentially substitute for the Ca2+ sites in alite, corresponding to an effective increase in the bulk CaO content. For a further increase of the bulk Fe2O3 content above ~0.8 %w, the alite content decreases as a result of the formation of the ferrite phase which reduces the amount of CaO available for alite formation. This is supported by the observation of the ferrite phase for the clinkers with bulk Fe2O3 contents above 0.8 %w by the 2 reflection at 33.8 in the XRD diffractograms (not shown).


70

Figure 3.19 Plot of the molar fractions of silicon in alite and belite (), nSi(alite/belite), as a function of the bulk Fe2O3 content. The nSi(alite/belite) ratios were obtained for each of the modified clinkers by deconvolution of the 29

Si MAS NMR spectra.


71

Hydration of fluoride-mineralized Portland cement

In addition to the strong mineralizing effect, the quantity of fluoride ions also shows an important impact on the hydrational properties of the mineralized cement. It seems to have an adverse effect on the reactivity of the calcium silicates leading to a retarding effect on the hydration[154,162]. However, in combination with an optimized amount of SO3 containing mineralizers (e.g. CaSO4), fluorine shows improved effects on the early hydration process[10,41] with an optimum fluoride content, as monitored by strength measurements after one day, at about 0.2 %w for an ordinary Portland cement. Higher fluoride contents will reduce the compressive strength measured after one day while the long-term hydration, i.e. > 28 days, is almost unaffected.

Figure 3.20 Experimental (a) and simulated (b) 29Si MAS NMR spectra of a fluoride-mineralized cement hydrated for 28 days. The cement is prepared from a clinker containing 0.36 %w F, 4.3 %w Al2O3 and with a LSF of 0.9. The deconvolution of the spectrum includes the sub-spectrum for the MIII alite and resonances for belite and the CSH phase. Alite and belite are shown in (c) while the CSH spectrum (d) includes five 29Si resonances at (29Si) = – 76.7 ppm, –79.1 ppm, –81.5 ppm, –83.3 ppm and –85.4 ppm.

72


To investigate the effects of fluoride ions on the hydration process, hydrated cement samples have been prepared (water/solid = 0.5, nSO3/nAl2O3 = 1.3 and a relative humidity of RH = 100 %) for clinkers produced from raw meals including 4.3 %w Al2O3, a LSF of 0.9 and fluoride content in the range 0.04  1.0 %w. The hydration has been stopped and studied by 27Al and 29Si MAS NMR for each series at 1, 2, 7, and 28 days. The degrees of hydration, given by equation 2.11, for alite and belite at each hydration time were estimated from

29

Si MAS NMR spectra

using the common deconvolution procedure. Based on the observations for synthetic CSH presented in Chapter 4, five

29

Si resonances at (29Si) = –76.7 ppm, –79.1 ppm, –81.5 ppm,

–83.3 ppm and –85.4 ppm have been included in addition to the resonances from belite and alite in its MIII form in the deconvolution of

29

Si MAS NMR spectra the for hydrated samples. An

example is shown in Figure 3.20 for a modified cement (0.36 %w F) hydrated for 28 days. A plot of the degree of hydration as a function of the fluoride content for alite and belite is shown in Figure 3.20. This graph clearly demonstrates that the quantity of fluoride ions has an important impact on the hydration rate of the calcium silicate phases. A critical fluoride content seems to exist around 0.36 %w F. Below this value, the fluoride ions enhance the hydration rate for the alite phase, Figure 3.21. An increase in the fluoride content above 0.36 %w F significantly retards the early hydration rate of alite. However, the effects from an increased fluoride content on the hydration of alite are less apparent as the hydration proceeds and nearly cancel out at 28 days of hydration. The degree of hydration for belite at 28 days of hydration also tends to be reduced slightly by fluoride levels above 0.36 %w, Figure 3.21. 27

Al MAS NMR spectra of selected samples, investigating the effects of fluoride ions

on the hydration process of the Al-containing phases, are shown in Figure 3.22. It is apparent from these spectra that the quantity of fluoride ions strongly influences the hydration of the aluminate phases after one day of hydration. However, the effect from fluoride anions is less apparent as the hydration proceeds to seven day of hydration, Figure 3.21, and it nearly cancels out after hydration for 28 days (not shown). Considering the centers of gravity for the centerbands at cg(27Al)  13 and 9 ppm in Figure 3.21, it can be seen that the AFt phase (ettringite) is rapidly converted into an AFm phase (monosulphate) as the fluoride content is increased up to 0.36 %w. Finally, when considering the intensity of the

27

Al resonance at ~82

ppm (i.e., Al3+ guest-ions in alite), the fluoride content of 0.36 %w F also appears to be a critical limit for the reactivity of alite. Below this value, an increase in the fluoride content tends to enhance the hydration rate of the alite phases.


73

Figure 3.21 The degree of hydration for alite and belite in fluoride-mineralized cements prepared from clinkers containing different quantities of fluoride ions. The relative quantities of the silicate species are obtained from 29Si MAS NMR spectra (7.1 T, R = 7.0 kHz, 30-s relaxation delay) of the hydrated samples. (Top) the total degree of hydration for alite after 1 day (▲), 2 days (), 7 days (▼) and 28 days (■) of hydration. (Bottom) the degree of hydration for belite (▲) after 28 days of hydration. The lines are shown as a guide to the eyes.

74


The hydration experiments demonstrate that the quantity of fluoride ions in Portland cement has an important impact on the early hydration, most significantly after one and two days of hydration. In agreement with the previous investigation for the effects of fluoride ions on the compressive strength of Portland cement[10], the

27

Al and

29

Si MAS NMR experiments of the

hydrated samples have shown the existence of a critical fluoride content, although for this series of clinkers of the critical fluoride content is of ~0.36 %w F. Below the critical value, an increase in the fluoride content tends to improve the hydration properties of the hydrated Portland cement. Most importantly, the increasing fluoride content up to 0.36 %w F increases the reactivity of the alite phase. Hence, a larger amount of aluminate ions may be released from the hydration of alite into the paste-liquid, which probably is the reason for the rapid conversion of the AFt (ettringite) into a AFm phase (monosulphate) after one day of hydration, see Figure 3.22. Above the critical value of ~0.36 %w F, the fluoride ions reduce the degree of hydration for alite up to 28 days, although the most pronounced reduction is observed after one and two days of hydration. One possible reason for the existence of this critical fluoride content is that calcium fluoride has a very low solubility product of ksp = 3.71011 M3, which causes CaF2 to precipitate immediately on the cement grain surface as the fluoride ions are released from the hydration of alite (cf. Chapter 4). Above the critical fluoride content, the precipitating CaF2 forms a protective layer on the cement grain surface reducing the cement hydration. This is consistent with the observation by

29

Si and

27

Al MAS NMR revealing that above the critical

concentration, fluoride ions reduce the degree of hydration of alite and belite, and they affect the formation of all hydrated phases. Thus, the presence of fluoride ions in high concentrations may result in a decrease in the quantity of CSH phases and therefore, the early strength development for the cement is reduced. Furthermore, it also retains a large amount of aluminium ions in the anhydrous alite phase leading to a retarding effect on the conversion of AFt (ettringite) into AFm phases. The critical fluoride content may, however, vary slightly depending on the quantity and reactivity of the alite phase. Another effect that also needs consideration is the decreased SiO2 content in fluoride-mineralized alite as a consequence of the coupled incorporation of F and Al3+ ions (section 3.2). Obviously, this results in a reduction in the quantity of the CSH phase formed during the early hydration period, since only a small amount of belite has reacted with water at this time. As the hydration process proceeds, the cement grain will expand due to the gradual hydration of alite and belite. Hence, the protecting layer of CaF2 is broken and the retarding effects from the fluoride ions become less apparent with time.


Figure 3.22

75

27

Al MAS NMR spectra (14.1 T, R = 13.0 kHz) of selected hydrated fluoride-mineralized cements

after one (left column) and seven (right column) days of hydration. The cements were prepared from clinkers containing 0.04 %w F (a), 0.23 %w F (b), 0.36 %w F (c) and 0.65 %w F (d). The spectra demonstrate that the quantity of fluoride significantly affects the formation of the AFt (ettringite) phase and its conversion into AFm phases (monosulphate). The asterisks (*) indicate the spinning sidebands from the AFt phases.

76


Application of fluoride mineralization

3.6.1. Preparation of test clinkers

Based on the results presented in the previous sections, a fluoride-mineralized clinker, denoted test clinker, was prepared using a commercial white Portland clinker (HKL) as the main source. The chemical compositions of the original and modified clinkers are shown in Table 3.1. As it is shown by the table, only the contents of SiO2, Al2O3, Fe2O3 and F have been modified. The modification of the chemical composition leads to an increase in the lime saturation factor from 0.95 for HKL to approximately 1.00, although the actual CaO content is nearly constant for the clinkers. Furthermore, the actual fluoride content in the test clinker is 0.28 %w rather than 0.36 %w which is the optimum fluoride content observed in the hydration experiments (cf. section 3.5). This is due to the fact that CaF2 has a moderate volatility, which makes it difficult to control the actual content of fluoride in the final clinker. Quantitative Rietveld analysis of the test clinker for batch 1 reveals that it consists of 77 %w alite, 21 %w belite and 2 %w tricalcium aluminate whereas the unmodified HKL contains 69 %w alite, 24 %w belite and 5 % tricalcium aluminate. In fact, the alite content may be further increased by additional CaO, if it is desired; the saturation condition defined by equation (1.1), i.e. LSF = 1.0, may not be applied to this clinker type since the majority of aluminum is not present in tricalcium aluminate but rather as guest ions in the alit phase.

Table 3.1 Chemical composition of the original and the modified clinkers from XRF experiments. The bulk fluoride content and the quantity of free CaO for the clinkers were measured using the methods described in Appendix 2.

HKL CaO SiO2 Al2O3 Fe2O3 F SO3 MgO Na2O K2O Free CaO

70.3 25.4 2.13 0.37 0.04 0.12 0.63 0.19 0.07 1.7

Test clinker Batch 1

Batch 2

70.4 22.8 4.49 0.78 0.28 0.36 0.56 0.10 0.00 0.4

70.3 22.5 4.40 0.73 0.29 0.40 0.73 0.07 0.02 0.4


77

3.6.2. Strength performances of fluoride-mineralized Portland cement

The clinkers were milled together with gypsum to produce cements with a Blaine fineness of 480 m2/kg. The SO3 content for the clinkers was optimized by following the heat evolution of the hydration process up to eighteen hours after mixing for a series of cements with different SO3 contents. For white Portland cement (wPc) produced from the unmodified HKL, the optimum SO3 content is 2.1 %w whereas the test cements required a SO3 content of 4.2 %w to avoid flash setting, due to their rather high bulk Al2O3 content. Strength tests were performed for the white Portland cement, an ordinary Portland cement (oPc) and the two test cements (from batch 1 and 2) in accordance with the mini-RILEM method[163]. The compressive strengths for the hydrated cements were measured at 1, 2, 28 and 90 days of hydration with the results presented in Figure 3.23. A comparison of the hydration experiments for wPc and oPc clearly demonstrates the effect of the quantities of alite and belite on the strength development. For one and two days of hydration, the high alite content in oPc results in higher compressive strengths as compared to wPc. On the other hand, as the hydration proceeds the strength development is taken over by the hydration of belite and therefore, wPc exhibits higher strengths at 28-days and later hydration times. After one day of hydration, the test cements exhibit comparable strengths as the oPc due to their high alite content. However, the high belite content of the test cements as compared to the oPc results in a much faster strength development during the long-term hydration. Moreover, in comparison with the unmodified wPc, the test cements have shown improved compressive strengths for all hydration stages; after one day of hydration, the test cements exhibit an increase in the compressive strength corresponding to an average value of 36 % and approximately 10 % after long-term hydration. Strength tests have also been conducted for a series of blended cements produced from the test clinker of batch 1, employing a 30 %w replacement of the clinker by limestone filler (CaCO3), heat-treated bentonite or glass particles. The heat-treated bentonite and glass are developed in two other PhD co-projects of the FUTURECEM project. After one and two days of hydration, the test cement shows the best performance with limestone filler alone, Figure 3.24. However, its compressive strength is somewhat lower than that for pure wPc. At 28 days of hydration and later, the partly replacement by 10 %w CaCO3 and 20 %w glass results in compressive strengths which are roughly 5 % lower than that of pure wPc. The replacement of the clinker by heat-treated bentonite results in the lowest strength for all samples, except for that after one-day of hydration.

78


120.0

Compressive strength (MPa)

100.0

80.0

60.0

40.0

20.0

0.0 1

2

28

90

days

Figure 3.23 Compressive strength of blended cements measured in accordance with the mini-RILEM method. The test cements (■, ■) and white Portland cement (■) have a Blaine fineness of 480 m2/kg whereas it is 575 m2/kg for the ordinary Portland cement oPc (■). Thus, to be able to compare the results of test cements with oPc, the actual compressive strength of oPc has been adjusted relating to the difference in their fineness[18].

100.0 90.0

Compressive strength (MPa)

80.0 70.0 60.0 50.0 40.0 30.0 20.0 10.0 0.0 1

2

28

90

Days

Figure 3.24 Compressive strength of blended cements measured in accordance with the mini-RILEM method. The cements were prepared with a clinker substitution level of 30 %w, i.e. (■) 30 %w limestone filler, (■) 10 %w limestone filler and 20 %w pre-treated betonite and (■) 10 %w limestone filler and 20 %w glass particles. The unmodified wPc is shown in red (■).

79


Summary

Site preferences for F and Al3+ guest ions in the calcium silicate phases of Portland clinker have been investigated using

19

F,

27

Al, and

29

Si MAS NMR combined with double-

resonance experiments such as 27Al{19F} and 29Si{19F} CP/MAS, and 29Si{19F} CP-REDOR. It is demonstrated that fluoride mineralization can be represented by the coupled substitution Si4+ + O2  Al3+ + F in alite, where fluoride ions substitute for the oxygen on interstitial sites and Al3+ ions replace silicon on tetrahedral sites in the near vicinity of the F ions. The coupled substitution of F and Al3+ stabilizes local regions with the chemical composition Ca27Si9-xAlx(Ob)36(Oi)9-x(Fi)x, where the upper limit of x is 4.5. A series of laboratory-synthesized fluoride-mineralized Portland clinkers have been prepared and studied using solid-state NMR and other analytical tools such as XRD, XRF, free lime measurement, etc. This work demonstrates that the use of fluoride-mineralizing agents in clinker production promotes the formation of alite by two different mechanisms. The most important effect is that the incorporation of fluoride in alite results in an increase in the entropy of mixing, which corresponds to a reduction in the Gibbs free energy for this phase. Assuming that the enthalpy of alite is unaffected by the incorporation of fluoride, the optimum fluoride content in Portland cement is approx. 4.1 % by the weight of alite. Another important effect arises from the mass balance of the coupled substitution Si4+ + O2  Al3+ + F in alite, which results in an increase in the quantity of belite and thereby, further facilitates the forward reaction CaO + belite  alite. The incorporation of Fe3+ ions in calcium silicate phases of Portland clinker has been investigated using

29

Si IR NMR. It is evident that small amounts of Fe3+ ions are incorporated

into the alite as well as belite structures. However, the Fe3+ ions do not appear to be involved in fluoride mineralization since the presence of a high F content in alite does not promote the incorporation of Fe3+ ions in this phase. This study also demonstrates that the Fe3+ ions tend to be incorporated on the octahedral sites of alite by substituting for the Ca2+ ions. The hydration experiments for fluoride-mineralized cements have proven the existence of a critical bulk fluoride content of 0.36 %w. Below this value, an increase in the fluoride content tends to improve the hydrational properties of Portland cement, e.g. it promotes the hydration of alite and the conversion of AFt (ettringite) into AFm (monosulphate) phases. A further increase in the fluoride content above this critical value significantly reduces the degree of hydration for alite up to 28 days. Furthermore, it exhibits a retarding effect on the conversion of AFt into


80

AFm, although this is nearly cancelled out after seven days of hydration. The high fluoride level also seems to affect the degree of hydration for the belite phase. A test clinker has been prepared based on the results from the optimization of the fluoride content in Portland clinker. Although the CaO content in the test clinker is comparable to the unmodified white Portland clinker, i.e. 70 %w CaO, the test clinker has an increase in the alite content by approximately 8 %w whereas the content of belite and tricalcium aluminate is decreased with a similar amount. Tests on the strength performances of fluoride-mineralized cements prepared from the test clinker have been conducted in accordance with the mini-RILEM method. The test cements exhibit an increase in its one-day compressive strength by 36 % as compared to the unmodified white Portland cement. For longer hydration times, the test cement shows an increase of approximately 10 %. Blended cements with a 30% clinker replacement have been prepared based on the test clinkers. Three systems have been studied: (i) 30 %w lime, (ii) 10 %w limestone and 20 %w glass particles and (iii) 10 %w limestone and 20 %w pre-treated clay. The blended cement (i) shows the best early strength performance whereas system (ii) has the best performance after long-term hydration. In both cases, the blended cements exhibit a reduction of about 5 % as compared to the compressive strength of the unmodified white Portland cement.

Chapter 4. Fluoride-ion environments in CSH

81

4. Chapter

Fluoride-ion environments in synthetic CSH and hydrated Portland cement

The main aim of this work is to investigate the structural environments of fluoride guest ions in the calcium silicate hydrate (CSH) phases and their influence on the hydration of Portland cement. As it is demonstrated in a number of previous studies, in addition to the strong mineralizing effects, fluoride ions also affect the hydration properties of Portland cement; they lead to an extended setting time and a reduction in the early strength of hydrated Portland cement[10,41,154,162]. Despite the obvious important effects of fluoride, the mechanism by which fluoride ions influence the hydration of Portland cement is only poorly understood. Two main proposals have been presented. The first considers the retarding effect as an intrinsic property of fluoride-mineralized alite[155], which is possibly stabilized in its rhombohedral form. This polymorph has proven a prolonged dormant time but it also hydrates much faster than the remaining alite forms. The second proposal explains the retarding effect of fluoride as a result of the precipitation of CaF2 salt on the cement-grain surface, preventing the cement to hydrate[164].


82

This explanation seems to be more appropriate for the reduction in the degree of hydration of Portland cement with fluoride contents above the critical value (cf. Section 3.5). However, experimental identification of CaF2 forming from the hydrating Portland cement has not yet been possible, due to the very low fluoride content. This chapter is divided into two parts. The first focuses on the incorporation of fluoride and its influence on the structure of synthetic calcium silicate hydrates while the second part presents the results from a study on the fluoride environments in hydrated Portland cement. The observation contributes to a further understanding of the mechanism by which fluoride strongly modifies the early hydration properties of Portland cement.

4.1.

A brief description of the CSH structure from the T/J viewpoint.

Figure 4.1 shows a schematic two-dimensional representation of the principal layer of CSH structure based on the tobermorite (T) structure[62,66], consisting of a CaO layer sandwiched between two sheets of dreierketten silicate chains. The upper part of the figure represents an infinite silicate chain with periodicity of one bridging Q2B and two Q2P sites. If the principal layers are not moved too far away from each other, such as in tobermorite 11-Å, the Q2B silicate units from two adjacent layers can inter-link forming Q3 units[165]. Different types of

guest ions may be present in the CSH structure[39,65,166,167]. For example, the substitution of Si4+ by Al3+ ions on the bridging sites is usually accompanied by incorporation of monovalent alkali cations or Ca2+ in the interlayer, preserving the charge balance. As it is shown in the lower part of Figure 4.2, each Al3+ incorporated in the bridging sites results in two Q2(1Al) units. The principal layer consists of highly disordered finite silicate chains of length 3n  1, where n is an integer whose value depends on the Ca/Si molar ratio and the degree of protonation (w/n). The tetrahedral silicate units on the bridging sites and the defect sites are charge-balanced partly or fully by the H atoms present in the interlayer region, forming the silanonl SiOH groups. For high Ca/Si molar ratios, the bridging SiO4 tetrahedra may also be replaced by calcium ions, which are coordinated to seven oxygen atoms originating either from water molecules or hydroxyl groups, Figure 4.1. The figure is only for the purpose of demonstrating the distinctions between different SiO4 connectivities and the distribution of OH in the CSH structures; it is not meant to accurately represent the actual arrangement of the atoms.


83

Figure 4.1 Schematic two-dimensional representations of the CSH structure derived from the tobermorite/jennite (T/J) viewpoint. The upper part in (a) shows an infinite dreierketten silicate chain with SiO referring to as SiOSi and SiO groups that are charge-balanced by Ca2+ present in the interlayer (not shown). The lower part of (a) shows the incorporation of Al3+ ions on the bridging site leading to two Q2(1Al) sites for each Al3+ for Si4+ substitution. The diagram shown in (b) illustrates the distribution of hydroxyl groups between tobermorite- and jennite-based dimer structures. For the T-based CSH structure, OH occurs in the interlayer (i.e. the lower part) whereas it takes part of the principal layer (i.e. the upper part) for the jennite-based part.

Chapter 4. Fluoride-ion environments in CSH 4.2.

84

Fluoride-ion environments in synthetic CSH and hydrated Portland cement 29

Si MAS and

29

Si{19F} CP/MAS NMR spectra of a fluoride-mineralized Portland

cement and two synthetic CSH samples, revealing the presence of F guest ions in the CSH phase, are shown in Figure 4.2 and 4.3, respectively. For the hydrated Portland cement sample (Figure 4.2), the 29Si resonances identifying SiF connectivities from remainders of alite appear in the spectral region from about 65 ppm to 75 ppm while for the CSH phase, they are observed from 75 ppm to 90 ppm[82]. The

29

Si{19F} CP/MAS NMR spectrum clearly shows

the presence of the Q1 and Q2 components. However, a clear identification of the remaining silicate environments (such as Q2(1Al) and Q2B*) cannot be extracted from the spectrum due to the severe overlap of resonances from the disordered CSH structure formed by the hydration of Portland.

Figure 4.2

29

Si MAS (7.1 T, R = 7.0 kHz) and

29

Si{19F} CP/MAS NMR spectra (7.1 T, R = 10.0 kHz) of a

Portland cement hydrated for 28 days, prepared from a fluoride-mineralized clinker containing 0.32 %w F. The spectra were recorded with relaxation delays of 30 s for (a) and 4 s for (b). The total numbers of scans for the experiments are 2048 scans and 62400 scans, respectively. Furthermore, the 29Si{19F} CP/MAS NMR experiment used a CP contact time of 5.0 ms.


85

Figure 4.3 (a, c) 29Si MAS (7.1 T, R = 7.0 kHz) and (b, d) 29Si{19F} CP/MAS NMR spectra (7.1 T, R = 10.0 kHz) of two synthetic CSH samples prepared with Ca/Si molar ratios of 1.25 and 0.83, respectively. The F/Si molar ratio in both samples is 0.3. The spectra were recorded with relaxation delays of 30 s for (a, c) and 10 s for (b, d). The total numbers of scans for the experiments are 2048 scans and 22400 scans, respectively. Furthermore, the 29

Si{19F} CP/MAS NMR experiment used a CP contact time of 5.0 ms.

In contrast to the hydrated Portland cement sample, the synthetic CSH samples exhibit 29

Si MAS NMR spectra with three well-resolved peaks at 79.2 ppm, 83.3 ppm and 85.2 ppm,

Figure 4.3. Furthermore, the broad resonance covering a spectral region from 90 ppm to 95 ppm, Figure 4.3 (c), may be assigned to a Q3 site, which is the characteristic chain-branching silicate of the tobermorite structure[165,166,168]. According to previous studies, the resonances at 85.2 ppm may account for both Q2B and Q2P units while the resonance with (29Si) = 79.2 ppm should be attributed to the chain-end Q1 or dimeric silicate species. The resonance at about 83 ppm has been identified in several studies[63,168,169] and it is assigned to the single protonated bridging silicate, referred to as Q2B* in Figure 4.1; however, this assignment has not yet been fully proven experimentally. Additionally, the

29

Si{19F} CP/MAS NMR spectra reveal a

resonance of low intensity at 73.9 ppm, which may originate from monomeric silicate species, Q0(H), as a result of the edging effect[170].

To achieve further information about the site preference of the fluoride guest ions and their influence on the formation of the CSH structures, a series of synthetic CSH samples including molar ratios Ca/(Si + Al) = 0.83  1.75, where Al/Si = 0 and 0.05, and F/Si = 0  0.5 have been studied by 19F and 29Si MAS NMR, with the main findings presented below.


86

4.2.1. Site preferences of F ions in the CSH structure from 19F MAS NMR 19

F MAS NMR spectra of a synthetic CSH sample recorded at two different magnetic

fields, i.e. 7.1 T and 14.1 T, are shown in Figure 4.4. The spectra reveal two distinct fluoride environments with isotropic chemical shifts at 122.0 ppm and 101.4 ppm. In fact, the same resonances are observed for all of the synthetic CSH samples, although they exhibit different spectral intensities as a result of the variation in the Ca/Si molar ratio. The narrow peak at 122.0 ppm may originate from an ordered fluoride structure which exhibits a large CSA pattern covering a spectral width of nearly 250 ppm. This CSA pattern can be simulated by including two different

19

F sites with identical isotropic chemical shift of 122.0 ppm where one has a

large chemical shift anisotropy () and the other possesses a small CSA. However, the optimized simulations for the spectra recorded at 7.1 T and 14.1 T result in two distinct sets of CSA parameters, Figure 4.4. This observation indicates that these fluoride ions are located in nearly identical local structures with an ordered arrangement but with a slight distribution in their CSA tensors. The second 19F resonance observed at 101.4 ppm is shifted slightly towards higher frequency relative to the resonance observed for CaF2, (19F) = –105.9 ppm. Furthermore, it exhibits a symmetric spinning sideband pattern covering a spectral region from 180 ppm to 20 ppm (Figure 4.5), similar to that for CaF2, indicating that this resonance is affected by strong dipolar couplings such as

19

F19F or 1H19F. Furthermore, the rather large line width of the

centerband for this resonance indicates that these fluoride ions occur in a disordered local environment. According to the structure model for CSH proposed by Richardson and Groves[62,170], two distinct types of hydroxyl groups are available for replacement with F ions, i.e. the covalently bonded hydroxyl groups of the Q1 and Q2B sites and the free OH ions distributed in the interlayer of the T-based part or in the principal layer of the J-based part of CSH structure; those are sketched in Figure 4.1. In order to obtain information about the SiF connectivities in the CSH structure, the

29

Si{19F} CP hetero-nuclear correlation (HETCOR) experiment was

applied to a synthetic CSH sample containing Ca/Si = 1.25 and F/Si = 0.3. However, a reconstruction of the 2D spectrum has not been possible due to a short T2 relaxation for the 19F spins, which causes the

19

F magnetization to dephase completely within very short time under

the t1 evolution. Alternatively, the 29Si{19F} FBCP experiment, described in section (2.2.7.), was applied to identify

19

FSi connectivities. Since this experiment utilizes the dipolar coupling to

transfer magnetization between the 19F and 29Si spins, the absence of a resonance at 122.0 ppm in the 29Si{19F} FBCP spectrum, Figure 4.5 (b), indicates that these fluoride ions are located far


87

away from the dreierketten silicate chains. Thus, they are tentatively assigned to F ions that substitute for the OH groups in the principal layer of the J-based CSH structure. This assignment is supported by the large chemical shift anisotropy observed for this resonance, which may be a result of the ordered local structure in the principal layer.

Figure 4.4 19F MAS NMR spectra of a synthetic CSH sample prepared with a Ca/Si molar ratio of 1.75 and F/Si = 0.3. Spectrum (a) was recorded at 7.1 T using a spinning speed R of 10.0 kHz, a 10-s relaxation delay and 128 scans. Spectrum (b) was recorded 14.1 T with R = 13.0 kHz, a 15-s relaxation delay and 512 scans. The corresponding simulated spectra using the STARS program are shown in (b) and (d), respectively. The simulations included three 19F sites one with isotropic chemical shift at 101.4 ppm ( = 0) and two with identical isotropic chemical shifts of 122.0 ppm. However, the refined simulations (root-mean-square, rms < 1 %) of the spectra result in two different set of CSA parameters; these are  = 132 ppm and 48 ppm, = 0.02 and 0.08 with relative intensity ratio of 1.0 : 3.1 for (b) and  = 112 ppm and 57 ppm, = 0.13 and 0.34 with a relative intensity ratio of 1.0 : 2.4 for (d).


88

Figure 4.5 19F MAS (a) and 19F{29Si} FBCP MAS (b) NMR spectra (7.05 T, R = 10.0 kHz) of a synthetic CSH sample prepared using a Ca/Si molar ratio of 1.25 and F/Si = 0.3. Both spectra were recorded using a 10-s relaxation delay. The spectrum in (a) is achieved with 512 scans while (b) is recorded with 58482 scans. Furthermore, the 19

F29Si cross polarization transfers were achieved using an 8-ms CP contact time.

The resonance at (19F) = 101.4 ppm, which exhibit a 19F29Si connectivity (Figure 4.5), may originate either from F atoms that substitute for the covalently bonded OH groups of the Q1 and Q2B silicates or from the fluoride ions distributed in the interlayer of the T-based CSH structure. In this particular case, the

29

Si{19F} CP-REDOR experiment (Section 2.2.8) may be

applied to determine the average SiF distances for each of the Q units, facilitating an unambiguous assignment. However, due to the limited time available for this project, such experiments have not been conducted. Nevertheless, recent theoretical calculations as well as experimental investigations of the incorporation of fluoride in bioactive calcium/alkali-silicate glasses demonstrate that SiF covalent bonds in tetrahedral environments is unlikely to form in such systems, which is consistent with the absence of resonance from the Q2B* component ((29Si) = 83.3 ppm) in the

29

Si{19F} CP/MAS NMR spectra shown in Figure 4.3. Therefore,

the 19F resonance at 101.4 ppm is most likely associated with the fluoride ions distributed in the interlayer of the T-based CSH structure. This assignment is supported by the observation of a large linewidth and a large symmetric spinning sideband pattern for the

19

F resonance, which

may result from the disordered structure of the interlayer and the strong dipolar couplings between these fluoride ions and H atoms from the OH groups or the water molecules.


89

4.2.2. Influence of fluoride guest ions on the CSH structure 29

Si MAS NMR spectra for a series of synthetic CSH samples, illustrating the effect of

fluoride on the polymerization of the silicate chains, are shown in Figure 4.6. The CSH samples were prepared from raw mixes including a Ca/Si molar ratio of 1.25 and F/Si molar ratios in the range 0  0.5. The 19F MAS NMR experiments for these samples only show the two resonances at 101.4 ppm and 122.0 ppm which indicates that at the studied concentrations, fluoride are exclusively incorporated in the CSH phase. However, the distribution of fluoride ions between these two fluoride environments exhibits variations with increasing F/Si molar ratio, Figure 4.7. Overall, the increased bulk fluoride content seems to increase the amount of fluoride guest ions in the T-based CSH structure, reflected by the decreased I[-122 ppm]/I[101 ppm] ratio. However, a slight increase is observed at F/Si = 0.5.

Figure 4.6

29

Si MAS NMR spectra (7.1 T, R = 7.0 kHz) of synthetic CSH samples synthesized with a Ca/Si

molar ratio of 1.25 and F/Si molar ratios of (a) 0, (b) 0.1, (c) 0.2, (d) 0.3 and (e) 0.5. The spectra were recorded using a 30-s relaxation delay and typically 2048 scans.


90

As it can be seen from the overall features of the 29Si MAS NMR spectra shown in Figure 4.6, the increased bulk fluoride content leads to a slight decrease in the line width for all

29

Si

resonances indicating that a CSH with a higher locally ordered structure is formed. Moreover, it significantly increases the amount of the Q2 components at the expense of Q1 units. The average silicate chain length, = (Q2 + Q1)/½Q1, as a function of the F/Si molar ratios is determined from deconvolutions of the 29Si MAS NMR spectra and plotted in Figure 4.7. For all spectra, in order to obtain acceptable lineshape simulations, it includes two Q1 resonances (79 ppm and 80 ppm) and two Q2 resonances (83 ppm and 85 ppm) whereas the Q0(H) resonance at 74 ppm has been neglected. The plot of versus the F/Si molar ratio clearly shows that for a fixed Ca/Si molar ratio an increased incorporation of fluoride ions in the CSH phase tends to promote the polymerization of the SiO4 tetrahedra. In addition, the chemical shifts of the Q1 and Q2 sites experience a small shift towards more negative ppm values with increasing F/Si

molar ratio, Figure 4.8.

Figure 4.7 (Left) Plot of the I[122 ppm]/I[101 ppm] ratio, where I is the spectral intensities of the 19F resonances at 122 ppm and 101 ppm, as a function of the F/Si molar ratio. The plot reflects the distribution of fluoride ions in the Tbased part relative to J-based part of CSH structure. (Right) Plot of the average silicate chain length (i.e. = (Q1 + Q2)/½Q1) for the CSH versus the F/Si molar ratio. The data are obtained from NMR experiments at 7.1 T, respectively.

19

F MAS and

29

Si MAS


91

Figure 4.8 Plots of 29Si isotropic chemical shifts (iso) for the Q1 and Q2 sites as a function of the F/Si molar ratio (left) and the Ca/(Si + Al) molar ratio (right). The values of (29Si) are determined from experiments at 7.1 T. Furthermore, the graph shown on the right includes

29

Si MAS NMR

29

Si chemical shifts for two series of

synthetic CSH samples, i.e. with (Al/Si = 0.05) and without including Al2O3 which are shown in red and blue, respectively.

Similar trends are observed for two other series of synthetic CSH samples, which are synthesized with and without including an aluminum source (Al/Si = 0.05). Furthermore, the syntheses used a fixed F/Si molar ratio of 0.3 but with Ca/Si ratios varying from 1.75 to 0.83; the Ca/Si ratio of 1.75 is the typical average value for CSH in mature cement pastes while 0.83 and 1.5 are the ideal values for tobermorite 14-Å and jennite, respectively. The results from the 19

F MAS and 29Si MAS NMR experiments for these samples are summarized in Figure 4.8 and

Figure 4.9. As it can be seen from Figure 4.8, the increase in the Ca/Si molar ratio slightly shifts the Q1 and Q2 resonances towards positive ppm values. In fact, the same trend is reported in previous investigations[169,171,172], although for CSH without including fluoride. Evidence for the incorporation of Al3+ in the tetrahedrally coordinated environment is obtained from

27

Al

MAS NMR experiment on a synthetic CSH sample (Ca/Si = 1.25), Figure 4.10. The spectrum clearly shows two resonances in the spectral region for Al3+ ions in tetrahedral coordination, i.e. 67 ppm and 73 ppm. However, their total spectral intensity only account for half of the bulk Al2O3 content. The remainding Al2O3 occurs as octahedrally coordinated Al species such as the third aluminate phase[173] (4 ppm) and AFm phases (9 ppm). A minor amount of aluminum also occurs in penta-coordinated environments (36 ppm). As the 29Si chemical shift primarily reflects local structural features within the first and second coordination sphere, this Al3+ incorporation


92

in the bridging sites only introduces a shift of approximately 3 ppm towards positive ppm value on the two Q2P silicates, which are referred to as Q2(1Al) in Figure 4.1, whereas the remaining Q1 and Q2 resonances (including Q2B and Q2P) are nearly unaffected by presence of the Al3+ guest ions, Figure 4.8. A comparison of the reciprocal average silicate chain length for the two series of samples with a previously reported data set[174] for a series of CSH sample without fluoride ions is shown in Figure 4.9. Consistent with earlier observations, this figure shows that a decrease in the Ca/(Si + Al) molar ratio results in an increase in the average silicate chain length. However, in the presence of fluoride guest ions, CSH phases with much longer average silicate chains are formed. The effect of fluoride guest ions on the average chain length seems to be increased with decreasing Ca/(Si + Al) molar ratio. For Ca/(Si + Al) ratios above 1.25, the average silicate chain length of the fluoride-containing CSH phases tends to be increased by the presence of an aluminum source whereas it is reduced below this value. The distribution of fluoride ions between the J-based and T-based CSH structures is also affected by the presence of Al3+ guest ions, although a significant difference is only observed for samples with Ca/(Si + Al) ratios of 1.25 and 1.5. For all samples, only two

19

F

resonances with isotropic chemical shifts at 101.4 ppm and 122.0 ppm are observed. However, the distribution of fluoride over these two sites has shown a strong dependency on the molar ratios Ca/(Si + Al), Al/Si as well as F/Si. It is apparent from a plot of the I[-122 pm]/I[-101 ppm] ratio as a function of the Ca/(Si + Al) molar ratio that the ordered fluoride environment is predominant in samples with high bulk CaO content. As the Ca/(Si + Al) ratio is decreased, the resonance at 101.4 ppm exhibits an increasing intensity. The most significant change is observed when the Ca/(Si + Al) ratio is reduced just below 1.5, i.e. the ideal Ca/Si molar ratio of jennite. This observation strongly supports the assignment of the 19F resonances, (19F) = 101.4 ppm and 122.0 ppm, to fluoride ions incorporated in the T-based and J-based CSH structures, respectively. In the absence of an aluminum source, the intensity of the resonance at 101.4 ppm becomes much larger than that at 122.0 ppm as the Ca/Si molar ratio is decreased to 1.25. However, below this value, the I[101

ppm]/I[122 ppm]

ratio is slightly increased and the

distribution of fluoride in the two different environments become almost equal as the Ca/Si molar ratio is further reduced to 0.83. This observation is consistent with the fact that the J-based CSH structure is predominant at high Ca/Si ratios and therefore, a larger amount of fluoride ions are incorporated into this phase. As the Ca/Si ratio decreases, a larger amount of T-based structure will be formed within the CSH phase leading to an increased intensity for the fluoride resonance at 101.4 ppm.


93

Figure 4.9 (Left) Plot of the reciprocal average silicate chain length, 1, as a function of the Ca/(Si + Al) molar ratio for three series of synthetic CSH samples: (♦) F/Si = 0.3 and Al/Si = 0, (■) F/Si = 0.3 and Al/Si = 0.05, (●) F/Si = 0 and Al/Si = 0.05. (Right) Plot of the I[-122 ppm]/I[-101 ppm] ratio, where I is the spectral intensitiy of the

19

F

resonances at 122.0 ppm and 101.4 ppm, as a function of the Ca(Si + Al) molar ratio.

A possible explanation for the similarity in the shifting effects on the Q1 and Q2 resonances as a result of an increase in the fluoride content and the decreased Ca/Si molar ratios (Figure 4.8) is that the increased incorporation of fluoride in the T-based CSH structure is accompanied by a similar increase in the quantity of Ca2+ ions in the interlayer of this phase. This effectively reduces the Ca/Si molar ratio for the main calcium layer and therefore, it increases the average silicate chain and promotes the formation of the T-based CSH structure at the same time. Furthermore, the absence of the Q2B* resonance in the

29

Si{19F} CP/MAS

spectra shown in Figures 4.2 and 4.3 indicates that these fluoride guest ions most likely substitute for the OH groups associated with CaOH on the bridging sites of the silicate chains rather than distributed uniformly in the interlayer. Thus, a large amount of fluoride ions may be incorporated in the dimeric T-based CSH structure, which may predominate for Ca/Si ratios just below the ideal value of the jennite structure (i.e. 1.5). As the Ca/Si molar ratio is further decreased, SiO4 dimers polymerizes to form longer silicate chains (Figure 4.9), reducing the amount of Ca2+ ions on the bridging sites and the quantity of fluoride guest ions in the T-based CSH structure at the same time. A similar effect may result from the incorporation of Al3+ in the bridging sites. Furthermore, as it is apparent from Figure 4.10, the formation of calcium aluminate hydrate phases may reduce the Ca/(Si + Al) molar ratio for the CSH phase and


94

therefore, it affects the average silicate chain length as well as the distribution of fluoride between the T-based and J-based CSH structures.

Figure 4.10

27

Al MAS NMR spectrum (14.1 T, R = 13.0 kHz) of a synthetic CSH sample prepared using a

Ca/(Si + Al) molar ratio of 1.25 and Al/Si = 0.05. The fluoride content used in the synthesis corresponds to F/Si = 0.3. The spectrum was recorded using a 0.5-s excitation pulse, a 2-s relaxation delay and 27520 scans. The tetrahedrally coordinated Al3+ guest ions of the CSH phase are identified by the two resonances at about 73 ppm and 67 ppm. The Al3+ ions in penta-coordination appear at about 36 ppm while the octrahedral aluminum species, the third aluminate phase and AFm, are observed as a sharp peak at 4 ppm and a shoulder at 9 ppm, respectively.

Chapter 4. Fluoride-ion environments in CSH 4.3.

95

Influence of fluoride on the hydration of Portland cement

Based on the observations from the 29Si and 27Al NMR experiments of the hydration of fluoride-mineralized Portland cements presented in Section (3.5), the hydration of the unmodified white Portland cement (0.04 %w F) and three selected fluoride-mineralized cements, containing approximately 0.04, 0.29 and 0.9 %w F, have been examined using 19F, 27Al and 29Si MAS NMR. Hydrated samples with ages from 6 hours up to 200 days have been prepared using the procedure described in Appendix 1.

4.3.1. Identification of CaF2 in hydrated Portland cement by 19F MAS NMR

The ability to identify CaF2 formation during the early hydration of Portland cement may be decisive for the understanding of the retarding mechanism of fluoride. The

43

Ca{19F}

CP/MAS NMR experiment, described in Section (2.2.4.), may provide a clear distinction for the resonance from Ca2+ ions of the insoluble CaF2 salt from the remaining calcium-containing phases. However, due to the very low CaF2 content in combination with the low natural abundance of the

43

Ca nuclear spin isotope, it has so far not been possible to detect the

43

Ca

signal from this phase for hydrated Portland cement samples. Figure 4.11 demonstrates an alternative approach for the identification of CaF2 in hydrated Portland cement using 19F MAS NMR. The experiment utilizes the fact the 19F nuclear spins of CaF2 has a much longer spin-lattice relaxation time than that for F ions present in the CSH phase; the required repetition time for complete spin-lattice relaxation of the 19F spins in pure CaF2 is about 120 s at 14.1 T while it is only 15 s for the fluoride guest ions in CSH. Thus, the

19

F MAS NMR spectrum recorded using a 120-s relaxation delay may contain

19

F

resonances from CaF2 ((19F) = 106 ppm) as well as the CSH phases ((19F) = 103 ppm and –122 ppm). In a second experiment where the 19F MAS NMR spectrum is recorded using a relaxation delay of only 15 s, the 19F resonance of CaF2 only contributes to the first few scans and will be partly saturated afterwards. As demonstrated in Figure 4.11 for a Portland cement (0.9 %w) which has been hydrated for 200 days, the difference spectrum corresponds to the spectrum of CaF2. For this particular sample, the subtraction of spectra recorded with different relaxation delay seems not to be necessary. However, as demonstrated in the following section, the CaF2 content in samples corresponding to one-day of hydration is much lower and therefore, the subtracting procedure is necessary in order to identify the 19F resonance from CaF2.


Figure 4.11

19

96

F MAS NMR spectra (14.1 T, R = 13.0 kHz and 2048 scans) of a Portland cement (0.9 %w F)

hydrated for 200 days. The spectrum shown in (a) was recorded using a 120-s relaxation delay whereas it was 15 s in (b). The difference spectrum (c) shows the 19F resonance from the insoluble CaF2 salt. The asterisks (*) indicates the spinning sidebands from the CaF2 phase.

4.3.2. Hydration of fluoride-mineralized Portland cement studied by 19F MAS NMR

As it is demonstrated in Figure 4.12 – 4.15, the high sensitivity of 19F MAS NMR makes it possible to follow the hydration of fluoride ions in Portland cements, even for very small fluoride concentration of only 0.04 %w F, within reasonable spectrometer times; the spectra presented in Figure 4.12 and 4.13 was obtained using a 8-s relaxation delay and typically 8190 scans which corresponds to approximately 18 hours whereas the spectra in Figure 4.14 and 4.15 can be obtained within two hours applying a repetition time of 8 s and typically 512 scans. Since fluoride is entirely incorporated in the alite phase, the hydration of fluoride may somehow reflect the hydration of this phase. In general, the

19

F MAS spectra of hydrated Portland cements

contain resonances from the anhydrous phase centered at 114 ppm and two resonances at 122 ppm and 103 ppm, which have been identified as fluoride guest ions of the CSH phases as described above. These resonances experience variation in their intensities as the hydration proceeds. Furthermore, it can be seen that the release of fluoride ions from the anhydrous phase is nearly complete after 28 days of hydration. In addition, the insoluble CaF2 salt ((19F) = 106


97

ppm) has been identified in samples prepared from the modified cements containing 0.29 and 0.9 %w F after hydration for one day, Figure 4.16. Considering the two series of samples shown in Figure 4.12 and 4.13, they contain approximately the same fluoride concentration, but differ significantly in their Ca/Si ratios and Al2O3 contents; the unmodified white Portland clinker has a fluoride content of 0.04 %w, 70.30 %w CaO, 25.42 %w SiO2 and 2.13 %w Al2O3 while the modified clinker has 0.04 %w, 68.40 %w CaO, 24.98 %w SiO2 and 4.43 %w Al2O3. Thus, apart from the difference in the distribution of fluoride guest ions between the T- and J-based CSH structures, which is possibly due to the difference in their Ca/Si ratios, the fluoride species in the two cements show very similar hydrational properties. Both cements show clear hydrational activities at 12 hours after mixing with water, indicated by the decrease in intensity of the resonance at 114 ppm and the appearance of the resonance at 122 ppm. At hydration ages up to seven days, the resonance at 122 ppm exhibits the predominant intensity while the fluoride guest ions of the T-based CSH structure is only present in small amount. As the hydration proceeds, the resonance at about 102 ppm exhibits an increasing intensity with time whereas the fluoride guest ions of the Jbased CSH structure experience variation in quantity and finally decrease to nearly zero after 200 days of hydration. This is consistent with earlier observations for the CSH structures formed from the hydration of Portland cement[58,175], which identify tobermorite- as well as jennite-like structures for the CSH formed during the early hydration period while after longer hydration times, the CSH phase turns out to be suitably described by the T/CH viewpoint. For both cements, the fluoride guest ions in alite are nearly consummed after 28 days of hydration.


98

Figure 4.12 19F MAS NMR spectra (7.1 T, R = 10.0 kHz) of hydrated samples prepared from the original white Portland clinker (0.04 %w F) with the hydration times given below each spectrum. The spectra were recorded using an 8-s relaxation delay and typically 8192 scans. The 19F resonance from the anhydrous phase is observed at –114 ppm whereas the fluoride guest ions from the CSH phases appear at –102 ppm and –122 ppm.

Figure 4.13 19F MAS NMR spectra (7.1 T, R = 10.0 kHz) of hydrated samples prepared from a modified white Portland clinker (0.04 %w F and 4.3 %w Al2O3) with the hydration times given below each individual spectrum. The spectra were recorded using a 8-s relaxation delay and typically 8192 scans.


99




100

Fluoride in the cement prepared from the clinker containing 0.29 %w F exhibits a somewhat slower hydration rate within the first 12 hours after mixing with water; only minor changes are observed in the

19

F MAS NMR spectra recorded for the hydrated sample stopped

after 12 hours as compared to that of anhydrous cement. For longer hydration times, it shows similar hydration features to the two cements described above. After 24 hours of hydration, the samples clearly show a large proportion of fluoride ions incorporated in both the T- and J-based CSH structures, i.e. the clear appearances of resonances at 102 ppm and 122 ppm, respectively. Moreover, a minor amount of CaF2 salt is formed at this hydration stage, identified by the appearance of the 19F resonance at –106 ppm, Figure 4.14. It is clearly evidenced from the experiments that the quantity of CaF2 observed after one day of hydration increases with increasing fluoride content in the anhydrous cement. As the fluoride content in the cement is increased much above the critical value (i.e. 0.36 %w F in cement clinker), the alite phase seems not to react with water before 12 hours; the 19F MAS NMR spectrum of this sample closely resembles that of the anhydrous cement, Figure 4.15. Moreover, instead of being incorporated in the CSH phase, a large amount of fluoride ions precipitates as CaF2 within the first 12  24 hours after mixing, Figure 4.16. Furthermore, the absence of

19

F resonances at 122 and 102 ppm reflects that only a minor amount of

fluoride ions is incorporated in the CSH phase at this hydration stage.

Figure 4.16 19F MAS NMR spectra (7.1 T, R = 10.0 kHz) of a synthetic CSH sample (Ca/Si = 1.25 and F/Si = 0.5) and samples hydrated for one day, prepared from the three modified cements with fluoride contents shown below their corresponding spectra. The difference spectra clearly demonstrate that CaF2 is only formed in the hydrated modified cements, where the corresponding clinkers contain 0.29 %w F and 0.90 %w F, respectively.


101

4.3.3. Retarding mechanism of fluoride ions on the hydration of Portland cement

An understanding of the retarding mechanism of fluoride on the hydration of Portland cement may be derived when evaluating the results from the

19

F,

27

Al and

29

Si MAS NMR

experiments together with observations from previous investigations[155,162,164]. It is apparent from the 27Al MAS NMR spectra shown in Figure 4.18 that a significant amount of AFt phases (e.g. ettringite) is already formed within the first six hours after mixing, regardless the fluoride content. From six to twelve hours after mixing, the hydration of the aluminum species in cements with fluoride contents below the critical value (i.e. 0.36 %w F) proceeds, although with a very slow hydration rate. The conversion of AFt into AFm phases is already observed for the modified cement containing 0.04 %w F after 12 hours of hydration while in the modified cement containing 0.29 %w F, the formation of ettringite is still proceeding. On the other hand, the fraction of aluminum in the anhydrous and hydrated phases of the modified cement containing 0.9 %w F exhibits no significant changes from six to twelve hours of hydration. In conjunction with the observations by 19F MAS NMR, summarized in Figure 4.15, this strongly indicate that the alite phase in the cement containing 0.9 %w F is intact within this hydration period. This retarding effect might be an intrinsic property of the fluoride-mineralized alite[155] as proposed earlier since fluoride is not yet released from the alite phase and thus, CaF2 cannot be formed at this hydration stage.

Figure 4.17 The degree of hydration for alite and belite in fluoride-mineralized cements containing different quantities of fluoride: (■) 0.04 %w F, (▲) 0.29 %w F and (♦) 0.90 %w F. The relative fractions of the silicate species are obtained from deconvolutions of the 29Si MAS NMR spectra (7.1 T, R = 7.0 kHz, 30-s relaxation delay) of the hydrated samples, stopped at different hydration times up to 100 days.


102

Figure 4.18 27Al MAS NMR spectra (14.1 T, R = 13.0 kHz) of selected hydrated samples for three modified cements with the fluoride contents shown below the spectra. The tetrahedrally coordinated Al3+ guest ions of the alite and belite phases are identified by the resonance at about 82 ppm with a shoulder at 84 ppm, respectively. The tricalcium aluminate phase appears as a broad resonance from 30 to 85 ppm. The hydrated phases including the AFt and AFm phases are observed at 13 ppm and 9 ppm, respectively. The asterisks (*) indicate spinning sideband from the AFt phase.

For one day of hydration and later, the presence of fluoride in quantities above the critical value tends to affect the hydration of the entire cement. As demonstrated in Section 3.5, the quantity of fluoride ions strongly modifies the release of Al3+ from the alite phase and the conversion of AFt into AFm phases. The examination of the effect of fluoride on the hydration properties of the alite and belite phases of Portland cement using

29

Si MAS NMR also

demonstrates that below the critical value, the presence of fluoride ions only slightly affects the hydration of alite and belite, Figure 4.17; up to seven days, the presence of fluoride seems to increase the reactivity of alite, but it slightly decreases the degree of hydration of belite. When present in concentrations above the critical value, fluoride has shown a substantial retarding effect on the hydration of alite as well as belite. It significantly reduces the degree of hydration


103

of alite up to 28 days. However, for longer hydration times, the quantity of fluoride has only a marginal effect on the hydration of alite. According to the fact that fluoride is only incorporated in the alite phase but seems to affect the hydration of the entire cement when present in quantities above the critical content, a plausible explanation for this is the formation a protective layer of insoluble CaF2 salt on the cement grain surface, which prevents the cement to hydrate[162,164]. The formation of such layer is supported by the observation of CaF2 in the hydrated fluoride-mineralized cement, Figure 4.16.

4.4.

Summary

Site preferences of fluoride ions and their influence on the structure of CSH has been investigated for synthetic CSH and hydrated fluoride-mineralized Portland cements using 19F, 27

Al and

29

Si MAS NMR. These studies demonstrate that

19

F MAS NMR represents a unique

tool for the structural characterization of fluoride ions in Portland cement. The high sensitivity of the

19

F spins makes it possible to follow the hydration of fluoride species even when they are

present in quantities of as low as 0.04 %w F. From the study of synthetic CSH, it is demonstrated that for the studied fluoride contents (F/Si = 0 – 0.5), fluoride ions are almost exclusively incorporated in the CSH structure. The insoluble CaF2 salt is not formed in such systems. The

19

F MAS NMR

experiments reveal two different structural environments for the fluoride guest ions of the CSH phases. The first exhibits a

19

F resonance at (19F) = 122.0 ppm with a large CSA

spinning sideband pattern, which reflects an ordered local structure and therefore, it has been assigned to fluoride ions distributed in the principal layer of the jennite-based CSH structure. The second type of fluoride ions appear at (19F) = 101.4 ppm. It is evident from the 19F{29Si} FBCP/MAS experiment that this resonance is dipolar-coupled to the

29

Si spins from the

dreierketten silicate chains and therefore, it most likely originates from the fluoride ions incorporated in the interlayer for the tobermorite-based CSH structure. This is supported by the observation of a symmetric spinning side band pattern for this resonance, which covers nearly 160 ppm, indicating that these fluoride ions possess strong 1H19F dipolar couplings, possibly to the H atoms from the OH groups or the water molecules. The incorporation of fluoride ions in the CSH phase results in an increase in its average silicate chain length. This effect of fluoride tends to increase with decreasing Ca/Si ratios. Furthermore, it also seems to be affected by the simultaneous incorporation of Al3+ ions in the bridging site of the silicate chain. For Ca/Si ratios above the ideal Ca/Si ratio of the jennite


104

structure (i.e. Ca/Si = 1.5), the presence of Al3+ guest ions appears to increase the effect from fluoride ions on the average silicate chain length whereas below this value, it reduces the average silicate chain length of the fluoride-containing CSH structure. It is also observed that the increased incorporation of fluoride ions introduces a slight high-field shift, i.e. towards more negative (29Si) value, of the 29Si isotropic chemical shifts for the Q1 and Q2 silicates. The hydration of fluoride ions in Portland cement has been followed by 19F MAS NMR for four cements: a commercial white Portland cement and three modified cements prepared from clinker with a high bulk aluminium content (~ 4.3 %w Al2O3) and fluoride contents of 0.04, 0.29 and 0.90 %w F. The experiments show that the majority of fluoride is incorporated in the CSH phases, although the fraction of fluoride distributed between the jennite-based and the tobermorite-based CSH structures experience variation with hydration time. For hydration ages up to seven days, the fluoride ions of the jennite-based CSH is predominant while on longer hydration times, the fluoride ions are preferentially incorporated in the tobermorite-based CSH structure. For the cements prepared from the modified clinker with fluoride contents of 0.29 and 0.9 %w F, the CaF2 salt is formed already after one day of hydration. The content of this phase tends to increase as the hydration proceeds. However, it only constitutes a minor part of the bulk fluoride content. The hydration of the cements has also been followed by

27

Al and

29

Si MAS NMR. The

experiments for samples with hydration ages up to 12 hours reveal that the prolonged setting time might be an intrinsic property of the fluoride-mineralized alite, since the fluoride is not yet released at this hydration stage and therefore, the CaF2 cannot be formed. On the other hand, after one day of hydration, a significant amount of CaF2 is detected by 19F MAS NMR for the cement containing 0.9 %w F. Moreover, the 27Al and 29Si MAS NMR experiments demonstrate that the hydration of all the present phases are affected by the presence of fluoride, despite the fact that fluoride ions are only incorporated in the alite phase. Therefore, the consideration of the formation of a protective layer of CaF2 is plausible for Portland cements with a fluoride content above the critical value of 0.36 %w F (cf. Chapter 3). This layer prevents the cement grains to react with water and therefore, it reduces the degree of hydration for all cement phases during the early hydration period.

Chapter 5. Framework structures of aluminosilicate binders

105

5. Chapter

Framework structures of aluminosilicate binders from solid-state NMR spectroscopy

This chapter includes two applications of the 29Si{27Al} REAPDOR NMR experiment in structural investigations of the network structure of aluminosilicate binders. The first part concerns a study of the SiOAl connectivities in geopolymeric materials formed from alkali activation of metakaolin samples. The second part presents the results from a reexamination of the disordered layer structure of strätlingite using solid-state NMR. More details on both studies are presented in manuscripts 1 and 2, respectively.

Chapter 5. Framework structures of aluminosilicate binders 5.1.

106

SiOAl connectivities in alkali-activated materials

Recent developments of alternative building materials have resulted in a renewed interest in alkali-activated materials[176-179], often denoted geopolymers, as alternative binders, with emphasis to the generally much lower energy consumption and CO2 emission required for production of these materials as compared to conventional Portland cement[180]. Additionally, the materials provide other advantageous properties like high early strength development, long-term durability, fire-resistance and storage of hazardous inorganic waste[181-184]. The process of forming the aluminosilicate networks requires activation by a relatively high concentration of alkali hydroxides, typically NaOH and/or KOH. It involves three main steps[177,178,185] (1)

Release of monomeric Si(OH)4 and Al(OH)4- units from the starting materials activated by alkali hydroxide.

(2)

Reorganization and diffusion of ions to form small coagulated structures.

(3)

Termination by poly-condensation, i.e., the formation of SiOAl linkages to form a 3dimensional framework structure.

The structure of geopolymeric materials as well as their performance is dependent on several factors, including the concentration of alkali hydroxides, the bulk chemical composition and the quantity of soluble silicates and aluminates dissolved from the precursors[186-188]. Alternatively, the alkali-activation process can be applied to blended cements[189,190], in which the clinker substitution level can be increased above the current limit of approximately 30 % by weight. However, geopolymer cements have only been commercialized in small-scale facilities so far, but not in large-scale applications where the strength is critical. In the generally accepted structural model[185], geopolymers are described by a backbone structure of poly(sialates) with the empirical formula of Mn{(SiO2)z AlO2}nwH2O, where the SiO4 and AlO4 tetrahedra are linked by sharing all their oxygen atoms, corresponding to Q4 units. The presence of cations M such as Na+, K+, Li+, Ca2+, Ba2+, NH4+ and H3O+ located in the framework cavities is necessary to preserve charge balance of the incorporated aluminium ions in four-fold coordination. So far, the detailed characterization of those materials has been a major challenge, owing to the lack of long-range periodicity in their open-framework structures. Thus, alkali-activated materials are often considered as ‘X-ray amorphous’ and their XRD patterns provide very little structural information[177,185].


Figure 5.1 Experimental (a) and simulated (b)

107

29

Si MAS NMR spectra of an alkali-activated metakaolin sample

(molar Si/Al = 2.0, Na/Al = 1.5). The experimental spectrum (a) was recorded at 7.1 T using a spinning speed of R = 7.0 kHz and a 30-s relaxation delay. Spectrum (b) is a simulation of (a), in which the eight different resonances shown in (c) and (d) are included.

According to the general ability of NMR spectroscopy to probe local structural features independent of long-rang order,

29

Si MAS NMR spectroscopy appears to be a useful tool for

investigations of the open framework structures of alkali-activated materials. However, a limitation of the single-pulse 29Si MAS NMR experiment is apparent when a number Al3+ ions is substituted into the second-coordination sphere of the probed SiO4 site; each Al3+ for Si4+

108


substitution in the second-coordination sphere leads to a resonance shift of about 5 ppm towards higher frequency[79], causing resonances from different types of SiO4 condensation to overlap and preventing a straight-forward interpretation of the observed chemical shifts[191-194]. For example, the chemical shifts of Q3(0Al) falls in the same spectral region as those from Q4(2Al) and Q4(3Al) sites. This is illustrated for an alkali-activated metakaolin sample in Figure 5.1; the sample was synthesized using a Si/Al molar ratio of 2.0 and Na/Al molar ratio of 1.5. It is apparent from the deconvolution, shown in Figure 5.1 (b-d), that the 29Si MAS NMR spectrum consists of eight resonances having chemical shifts within the region from –115 to –75 ppm. The assignment of these resonances from single-pulse

29

Si MAS NMR experiment is somewhat

uncertain, since this chemical shift region may include any of the Q4(nAl) and Q3(nAl) components. In such cases, the

29

Si{27Al} REAPDOR experiment described in section 2.2.10

appears to be an appropriate tool for probing the SiOAl network. An unambiguous assignment of the 29Si resonances may be achieved from a combination of the 29Si chemical shifts and the corresponding number of Al atoms in their second-coordination spheres, obtained from 29

Si{27Al} REAPDOR experiments.

5.1.1. Effects of Si/Al and Na/Al molar ratios 29

Si MAS NMR spectra for a series of alkali-activated metakaoline samples, synthesized

using Si/Al = 1.0  3.0 and Na/Al = 1.0, are shown in Figure 5.2. As it can be seen from the spectra, the

29

Si resonances appear in the region from about 115 ppm to 75 ppm, which, in

principel, can be assigned to all types of tetrahedral SiO4 species. Considering these 29Si MAS NMR spectra, a very intense peak at about 85.0 ppm has been observed for the sample prepared using a Si/Al molar ratio of 1.0. However, four additional

29

Si resonances with isotropic

chemical shifts (29Si) at 78.6 ppm, 83.2 ppm, 87.5 ppm and 92.2 ppm can be identified from a deconvolution of this spectrum. As the Si/Al molar ratio is increased from 1.0 to 2.0, five well-separated 29Si resonances are observed at 87.5 ppm, 92.2 ppm, 97.0 ppm, 102.5 ppm and 107.1 ppm. A further increase in the Si/Al molar ratio results in locally disordered network structures, which is indicated by the smooth, broadened lineshape of the 29Si MAS NMR spectra shown in Figure 5.2 (d) and (e). Finally, when the Si/Al molar ratio becomes larger than 3.0, a significant amount of non-reacted metakaolin have been detected by the 29Si as well as the 27Al MAS NMR experiments, cf. manuscript 1.


109

Figure 5.2 29Si MAS NMR spectra (7.05 T, R = 7.0 kHz) of alkali-activated metakaoline samples, prepared using a Na/Al molar ratio of 1.0 and Si/Al = 1.0 (a), 1.5 (b), 2.0 (c), 2.5 (d) and 3.0 (e). The spectra were recorded employing a 30-s relaxation delay and typically 2048 scans.

As it is stated above, the use of alkali in a correct amount is vital for the synthesis of aluminosilicate network structure since the substitution of Al3+ for Si4+ on the tetrahedral sites has to be charge-balanced. Thus, deficient or excessive amounts of alkali may affect the formation of the aluminosilicate structures substantially, leading to several different SiO4 environments within the network structure. This is illustrated in Figure 5.3 showing selected 29Si MAS NMR spectra for another series of alkali-activated samples, for which the Si/Al molar ratio is 2.0 and the Na/Al molar ratio is varied from 0.8 to 2.15. However, only samples with Na/Al  1.5 have been studied; the alkali-activated product becomes gel-like and is very difficult to pack in a NMR rotor as the Na/Al molar ratio is raised above 1.5. As it is apparent from Figure 5.3, the spectra consist of partly overlapping

29

Si resonances with isotropic chemical shifts in the


110

range from 107 to 79 ppm. Furthermore, the 29Si resonances that appear at 85.0, 83.2 and 78.6 ppm are only observed with significant intensity for samples including an excessive amount of Na+ ions (i.e. Na/Al > 1.0). Altogether, when changing the raw mix composition (e.g. Na/Al and Si/Al molar ratios) used in the syntheses, eight 29Si resonances within a spectral region from –75 ppm to –110 ppm, i.e. the chemical shift region for silicon in tetrahedral coordination, have been observed for the alkali-activated metakaolin samples. The assignment of these resonances from their chemical shifts is somewhat uncertain since this chemical-shift region may include any of the Q1, Q2(nAl), Q3(nAl) and Q4(nAl) components.

Figure 5.3 29Si MAS NMR spectra (7.05 T, R = 7.0 kHz) of alkali-activated metakaolin samples, prepared using a Si/Al molar ratio of 2.0 and Na/Al = 0.8 (a), 10.0 (b), 1.3 (c) and 1.5 (d). The spectra were recorded employing a 30s relaxation delay and typically 2048 scans.

111

Chapter 5. Framework structures of aluminosilicate binders 5.1.2.

29

Si {27Al} REAPDOR NMR

In order to assign the eight samples presented above,

29

29

Si resonances, observed for alkali-activated metakaolin

Si{27Al} REAPDOR NMR has been applied to determine the

number of Al atoms substituted in the second-coordination sphere for each of the resonances. This study involves three selected alkali-activated metakaolin samples corresponding to the spectra shown in Figures 5.2 (a), 5.3 (c) and 5.3 (d). The first sample considered in this study contains Si/Al and Na/Al molar ratios of 2.0 and 1.3, respectively. Its 29Si MAS NMR spectrum, Figure 5.3, clearly shows five distinct resonances separated by approx. 5 ppm. Additionally, the 29Si{1H} CP/MAS NMR experiment (not shown) of this sample reveals a broad resonance of low intensity at about 85.0 ppm. However, this resonance has been neglected in the deconvolution of the 29Si{27Al} REAPDOR spectra owing to its very low intensity. It appears from the 29Si{27Al} REAPDOR spectra shown in Figure 5.4 that only four of the 29Si resonances are affected by the re-introduced SiAl dipolar couplings. They experience different rates of Si–Al dipolar dephase, reflecting that the numbers of Al atoms incorporated in their second-coordination sphere are different. On the other hand, the resonance at –107.1 ppm shows no significant change in its intensity (S0 ~ S) and therefore, it may be assigned to a Q4(0Al) unit. The individual

29

Si{27Al} REAPDOR spectra were deconvolved to

assess the intensities, S0 and S, for the different sites. From a plot of the REAPDOR fractions versus the evolution time, Figure 5.5, it is confirmed that only four 29Si resonances are affected by the Si–Al dipolar couplings. The knAl values associated with their dipolar dephases are obtained from the curve fits shown in Figure 5.5 (right). However, it has been necessary to include S/S0 > 0.3 in the curve fit for the resonance at –87.5 ppm, since it has only one data point that fulfils the condition S/S0 ≤ 0.3. The knAl values are summarized in Table 5.1. Complementary results have been achieved for another alkali-activated metakaolin sample corresponding to that with the 29Si MAS NMR spectrum shown in Figure 5.3(d). From the spectral deconvolution shown in Figure 5.1, it can be seen that the sample includes eight non-equivalent

29

Si sites: six

29

Si resonances having the same isotropic chemical shifts as

observed for the spectrum in Figure 5.3(c) and two additional resonances at 78.6 and –83.2 ppm with low intensities. The last two resonances have been neglected in the deconvolution of the 29Si{27Al} REAPDOR spectra for this sample due two their rather low intensities. The knAl values obtained from curve fits of S/S0 > 0.3 for the remaining Table 5.1.

29

Si resonances are listed in


Figure 5.4

29

112

Si{27Al} REAPDOR (14.1 T, R = 10.0 kHz) spectra of an alkali-activated metakaolin sample with

Si/Al and Na/Al molar ratios of 2.0 and 1.3, respectively. (Left) Full signal, S0, which is not affected by the SiAl dipolar couplings. (Right) The attenuated signals, S, reflecting the SiAl dipolar couplings. The evolution periods are (a): 4 Tr, (b): 8 Tr, (c): 12 Tr and (d) 16 Tr, where Tr = 0.1 ms.

Figure 5.5

29

Si{27Al} REAPDOR curves for an alkali-activated metakaolin sample with Si/Al and Na/Al molar

ratios of 2.0 and 1.3, respectively: (♦) 87.5, (■) 92.4, () 97.2 and ( ) 102.5 ppm. (Left) Experimental REAPDOR fractions, S/S0, as a function of the evolution times (Tr = 0.1 ms). (Right) Curve fits for the REAPDOR fractions at short dephasing (S/S0  0.3), using the function S/S0 = at2, corresponding to equation (2.21).

113


Finally, in order to assign the 29Si resonance at (29Si) = 83.2 ppm, the alkali-activated metakaoline sample containing Si/Al = 1.0 and Na/Al = 1.0 has been studied by REAPDOR. It is evident from the deconvolution of its

29

29

Si{27Al}

Si MAS NMR spectrum, Figure 5.6,

that the sample contains six non-equivalent 29Si environments. However, the resonances at about 79 and 97 ppm, Figure 5.6 (e), have been neglected in the deconvolution of the REAPDOR spectra due to their rather low intensity. Furthermore, the dipolar dephasing for the resonances at 87.5 ppm and 85.0 ppm were considered together. This assumption is reliable since both resonances are derived from the Q4(4Al) components and have shown to have similar magnitudes of dipolar dephasing, Table 5.1. The knAl values for the resonances obtained from a fit of their REAPDOR fractions are summarized in Table 5.1. It shows clearly that the resonance at (29Si) = 83.2 ppm possesses a dipolar dephase and an isotropic chemical shift corresponding to a Q3(2Al) component.


29

Si MAS NMR spectra of an alkali-activated metakaolin sample

(Si/Al = 1.0, Na/Al = 1.0). The experimental spectrum (a) was recorded at 7.05 T using a spinning speed of R = 7.0 kHz, a 30-s relaxation delay and 2048 scans. The deconvolution of the spectrum reveals six different resonances, which are shown in (c), (d) and (e).

29

Si

114


As it is apparent from the data in Table 5.2, the eight 29Si resonances observed in Figure 5.1 can be assigned unambiguously by considering their (29Si) values together with the knAl values obtained from the 29Si{27Al} REAPDOR experiments. So far, it has not been possible to assign the

29

Si resonance at (29Si) = 78.6 ppm owing to its very low

29

Si intensity in the

studied sample. However, its isotropic chemical shift implies that it may originate from a nonfully condensed tetrahedrally coordinated silicon environment.

Table 5.1 Isotropic chemical shifts and corresponding knAl values for the eight 29Si resonances observed for alkaliactivated metakaolin samples. The data are obtained from curve fits of the REAPDOR fractions at short dipolar dephase (S/S0  0.3), using the function S/S0 = at2, corresponding to equation (2.21).

29

Si isotropic chemical shift in ppm

102.5

92.0

97.2

Figure 5.2(a)

87.5

0.39

83.2

0.51

Figure 5.3(c)

0.16

0.31

0.43

0.73

Figure 5.3(d)

0.19

0.29

0.44

0.67

Table 5.2 An assignment of the eight

85.0

0.25

0.69

29

Si resonances observed for the alkali-activated metakaolin samples. The

assignment is made in accordance with the results achieved from the 29Si{27Al} REAPDOR experiments.

29

Si isotropic chemical shift in ppm

-107.1

-102.5

-97.2

-92.0

-87.5

-85.0

-83.2

Figure 5.2(a)

-

-

-

Q4(3Al)

Q4(4Al)

Q4(4Al)

Q3(2Al)

Figure 5.3(c)

Q4(0Al)

Q4(1Al)

Q4(2Al)

Q4(3Al)

Q4(4Al)

-

-

Figure 5.3(d)

Q4(0Al)

Q4(1Al)

Q4(2Al)

Q4(3Al)

Q4(4Al)

Q4(4Al)

-

115

Chapter 5. Framework structures of aluminosilicate binders 5.2.

Disorder in the double tetrahedral layer structure of Strätlingite

Strätlingite is a potential hydration product of aluminate-rich cements[195]. More recently, this mineral has also been identified in alkali-activated blended cements, in which alkali hydroxides are used to activate the hydration of supplementary cementitious materials[196,197]. Strätlingite exhibits the ideal composition 2CaO·Al2O3·SiO2·8H2O and crystallizes in the trigonal space group R3m. A previous single-crystal X-Ray diffraction (XRD) study[198] of a mineral sample revealed that the strätlingite structure consists of an octahedral brucite-type layer [Ca2Al(OH)6·2H2O]+ and a double tetrahedral layer [(T,□)4(OH,O)8· 0.25H2O]–. The octahedral sites are fully occupied by Al3+ while approx. 45% of the tetrahedral sites are vacant (□). The remaining 55 % are occupied by either Si4+ or Al3+ with an overall Si/Al molar ratio of 1:1. Supplementary information about the double tetrahedral layer of strätlingite was provided by another study[199] using 27Al and 29Si MAS NMR. That study identified four different tetrahedral 29

Si resonances from the

29

Si MAS NMR experiments (7.1 T), indicating four different SiO4

environments of the double tetrahedral layer. Based on their isotropic chemical shifts, the three almost equally intense 29Si resonances in the region from –88.0 ppm to –81.0 ppm were assigned to Q2(0Al), Q2(1Al) and Q2(2Al) components and a low intensity resonance at −110 ppm was ascribed to the Q4(0Al) unit. This assignment is, however, in contrast to the basic structure of strätlingite determined by XRD, which suggests that Al3+ and Si4+ occur mainly as Q3 units. In the same NMR study, a single tetrahedral AlO4 resonance from the double tetrahedral layer and an AlO6 resonance from the brucite-type layer were observed. The resonances occur at 60.4 and 8.4 ppm in the 27Al MAS NMR spectrum (7.05 T), respectively. In this work the local structure of silicon and aluminum of the octahedral brucite-type layer and the double tetrahedral layer have been reexamined using single-pulse MAS,

29

Si{1H} and

27

Al{1H} CP/MAS,

27

Al 3QMAS and

29

29

Si and

27

Al

Si{27Al} REAPDOR NMR

experiments. However, this section only presents the results from the

29

Si{27Al} REAPDOR

NMR experiments for a synthetic sample of strätlingite, from which the

29

Si resonances are

assigned. In order to enhance the 29Si sensitivity, a 29Si{1H} CP/MAS period was applied prior to the

29

Si{27Al} REAPDOR sequence as demonstrated by Figure 2.11 in Section 2.2.9. A

detailed discussion of the disordered layer structure of strätlingite is provided in manuscript 2.


29

116

Si MAS NMR

The

29

Si resonances of strätlingite shown in Figure 5.7 appear in the region from about

75 to 95 ppm, i.e. the isotropic chemical shift region for tetrahedral SiO4 environments. However, the overall lineshape of the spectrum differs somewhat from that presented in the previous study of strätlingite. A 29Si{1H} CP/MAS NMR experiment of the same sample clearly resolves the lineshape into four well-separated 29Si resonances with isotropic chemical shifts at

(29Si) = 90.8 ppm, 86.4 ppm, 82.1 ppm and 79.7 ppm. Furthermore, a comparison of Figure 5 (a) and (b) reveals an additional resonance at 84.9 ppm, which has to be included to obtain satisfactory simulations of the spectra. As discussed above, the assignment of these resonances from their isotropic chemical shifts alone appears to be uncertain, since the chemical shift region from 75 ppm to 95 ppm may include any of the Q2(nAl) and Q3(nAl) structural units. Furthermore, fully condensed Q4 sites with a high number of Al incorporated in the second-coordination sphere such as Q4(4Al) and Q4(3Al) may also exhibit isotropic chemical shifts within this spectral region.

Figure 5.7 29Si MAS (a) and 29Si{1H} CP/MAS (b) NMR spectra (9.4 T) of a synthetic sample of alkali-containing strätlingite. The spectra were recorded using spinning speeds R of 6.0 and 3.0 kHz, respectively. Furthermore, the 29

Si{1H} CP/MAS NMR experiment was employed a contact time of 5.0 ms

117


29


The assignment of the 29Si resonances from the double tetrahedral layer of strätlingite has been facilitated by supplementary information obtained using

29


experiments. The ability of the REAPDOR experiment to retain SiAl dipolar couplings is utilized to determine the number of Al atoms in the second-coordination sphere for each of the 29

Si resonances, as demonstrated in the previous sections. A plot of the REAPDOR fractions as a

function of evolution time for a synthetic sample of strätlingite is shown in Figure 5.8. The plot clearly demonstrates that all

29

Si resonances are affected by the re-introduced SiAl dipolar

couplings. The corresponding knAl values are determined from curve fits of REAPDOR fractions to equation (2.21) and are summarized in Table 5.3. According to their isotropic chemical shifts together with the knAl values, the five 29Si resonances representing five different SiOAl environments of the double tetrahedral layer of strätlingite can be assigned to Q3(1Al): 90.8 ppm, Q3(2Al): 86.4 ppm, Q3(3Al): 83.8 ppm, Q2(1Al): 82.1 ppm and Q2(2Al): 79.7 ppm. This assignment is consistent with the fact that the resonances at 79.7 and 82.1 ppm experience a more efficient

29

Si{1H} cross-polarization transfer than the remaining resonances

(cf. Figure 5.7), since they are linked to a larger number of OH groups.

Figure 5.8 29Si{27Al} REAPDOR curves for a synthetic sample of strätlingite: (▲) 90.8 ppm, () 86.4 ppm, (▼) 83.8 ppm, (■) 82.1ppm and (♦) 79.7 ppm. (Left) Experimental REAPDOR fractions, S/S0, as a function of evolution time (Tr = 0.1 ms). (Right) Curve fits for the REAPDOR fractions at short dephasing (S/S0  0.3), using the function S/S0 = at2, corresponding to equation (2.21).

118



29

Si MAS NMR spectra of a synthetic sample of strätlingite. The

experimental spectrum (a) was recorded at 9.4 T using a spinning speed of R = 6.0 kHz, a 30-s relaxation delay and 2048 scans. The deconvolution of the spectrum reveals five different resonances, which are shown in (c) and (d) with isotropic chemical shifts at (29Si) = 90.8 ppm, 86.4 ppm, 83.8 ppm, 82.1 ppm and 79.7 ppm.

Table 2.3 Isotropic chemical shifts, knAl values determined from 29Si{27Al} REAPDOR experiments and relative intensities achieved from a deconvolution of the single-pulse 29Si MAS NMR spectrum for a synthetic sample of strätlingite.

Site assignment

(29Si)

knAl

Normalized intensity

29

Si: Q3(1Al)

90.8 ppm

0.21

2.8 %

29

Si: Q3(2Al)

86.4 ppm

0.44

55.2 %

29

Si: Q3(3Al)

–83.8 ppm

0.65

10.3 %

29

Si: Q2(1Al)

82.1 ppm

0.25

22.7 %

29

Si: Q2(2Al)

79.7 ppm

0.48

8.9 %


119

The distribution of silicon on the five different SiO4 sites, reflected by their intensities in the

29

Si MAS NMR spectrum, is obtained from a deconvolution, Figure 5.9; their normalized

intensities relative to the total intensity of the spectrum are Q3(1Al): 2.8 % , Q3(2Al): 55.2 %, Q3(3Al): 10.3 %, Q2(1Al): 22.7 % and Q2(2Al) : 8.9 %. The result clearly demonstrates that the

double tetrahedral layer consists largely of Q3 components, where Q represents either Si or Al in tetrahedral coordination with three QOQ and one QOH bonds, forming a 2-dimensional layer structure, which is consistent with the layer structure of 6-membered rings for strätlingite reported from XRD[198]. Furthermore, this alumino-silicate network is disrupted by site vacancies leading to approx. 30 % Q2 units. 5.3.

Summary

It has been demonstrated that the double-resonance REAPDOR sequence represents a valuable tool in structural characterisation of cementitious materials. The 29Si{27Al} REAPDOR experiment has been applied to estimate the number of Al atoms incorporated in the secondcoordination sphere for SiO4 tetrahedra and it has been shown that it provides a clear discrimination between different Qi(nAl) environments. Unambiguous assignment of the

29

Si

resonances has been achieved by this approach for alkali-activated metakaolin samples and for strätlingite.

120

Conclusions

Conclusion In this project, the effects of fluoride ions on the formation of Portland clinker and its hydration have been investigated. The modified clinkers were prepared using a white Portland clinker from Aalborg Portland A/S as a main source. It is found that the mineralizing effect of fluoride on the clinker formation associates with a coupled substitution Si4+ + O2  Al3+ + F in the alite phase, promoting the formation of this phase by two different mechanisms. The first is a thermodynamical effect, where the coupled incorporation of F and Al3+ stabilizes local regions with the chemical composition Ca27Si9-xAlx(Ob)36(Oi)9-x(Fi)x in the alite structure, where i and b denote the interstitial and covalently bonded oxygen atoms, respectively. The second effect arises from the mass balance of the coupled substitution, resulting in formal increase in the quantity of belite which facilitates the reaction, CaO + belite  alite. For the studied fluoride contents (i.e. 0.04  0.77 %w), the mineralizing effects are increased with increasing fluoride quantity. For the hydration of fluoride-mineralized Portland cement, the existence of a critical bulk fluoride content of about 0.36 %w F has been demonstrated. Below this value, an increase in the fluoride content tends to increase the degree of hydration for the alite phase. However, it is found that above the critical value, fluoride substantially affects the early hydration (i.e. the hydration up to 28 days) for all cement phases by two tentative mechanisms. Firstly, the retarded setting during the first 24 hours after mixing seems to be an intrinsic property of the fluoridemineralized alite; its hydrational activity during this period is very small. Secondly, at 24 hours after mixing, the precipitation of the rather insoluble CaF2 salt has been identified for the fluoride-mineralized cements, where the CaF2 quantity is increased with an increased fluoride content in the anhydrous cements. This observation supports the model where the insoluble CaF2 salt precipitates on the cement grain surface. At sufficiently high concentrations (i.e. for fluoride contents above the critical value), a protective layer of CaF2 may be formed, which retards the cement hydration. However, this effect seems to be leveled out after 28 days of hydration. From the optimizations of the Al2O3, Fe2O3 and fluoride contents in Portland clinker, it is found that the alite content can be increased by about 8 %w without additional CaO, which implies that the CO2 emission from the chemical reactions of the clinker formation is remained at the same level as that of the unmodified white Portland clinker. Furthermore, hydration experiments of the cement produced from this modified clinker show an increase by 36 % in its

121

Conclusions

one-day compressive strength as compared to the unmodified white Portland cement. For longer hydration times, an increase of about 10 % in the compressive strength is observed for the modified cement. The strength performance of blended cements produced from the modified clinker using a 30 % replacement by different supplementary cementitious materials has also been investigated. A remarkable reduction in the one-day compressive strength relative to that of wPc is observed for all blended cements. However, for the long-term hydration (i.e. after 28 days of hydration), the blended cement containing 70 % modified clinker, 10 % limestone filler and 20 % glass particles (developed in another PhD project of FUTURECEM) shows a reduction in strength of only 5 % as compared to the white Portland cement. Solid-state NMR has proven to be a valuable tool for structural investigation of the guest ions (such as F, Al3+ and Fe3+) in anhydrous as well as hydrated phases of Portland cement. The high sensitivity of the 19F spins makes it possible to study the local structure environments of the fluoride ions and their hydration processes in Portland cement. Moreover, it has been demonstrated in this project that the couplings of fluoride to other guest ions may be investigated using a combination of different advanced solid-state NMR techniques. In particular, the site preferences for F and Al3+ guest ions in the calcium silicate phases of Portland clinker have been investigated using

27

Al{19F} and

29

Si{19F} CP/MAS, and

29

Si{19F} CP-REDOR

experiments. It is found for the studied fluoride contents (> 0.77 %w F) that fluoride ions substitute for the interstitial oxygen sites of the alite structure only. Furthermore, the incorporation of fluoride ions in the alite structure is charge-balanced by a Si4+  Al3+ substitution on the tetrahedral site in the near vicinity of F. This observation unambiguously reveals the coupled substitution mechanism for the incorporation of F and Al3+ ions in the alite structure, i.e. Si4+ + O2  Al3+ + F. A new solid-state NMR pulse scheme (Forth and Back Cross Polarization), which is a modification of the Cross-Polarization experiment, has been developed for studying the site preferences of F ions in the calcium silicate hydrate (CSH) phases of Portland cement. This experiment utilizes the dipolar couplings to transfer the magnetization from the 19F spins to 29Si spins and subsequently back to the

19

F spins, making it possible to selectively detect

19

F

resonances from 29Si19F connectivities in the CSH structure within reasonable spectrometer times. In this study, it is observed that a rather large amount of fluoride ions may be incorporated in the CSH structure. The fluoride ions are either incorporated in the principal layer of the jennite-like part or distributed in the interlayer of the tobermorite-like part of the CSH structure. The incorporation of fluoride ions in the CSH structure tends to increase the average silicate chain length for this phase.

122

Conclusions

The incorporation of the paramagnetic Fe3+ ions in the calcium silicate phases of Portland clinker has been investigated using the

29

Si Inversion-Recovery NMR experiment and X-ray

powder diffraction. These experiments show clear evidence for the presence of paramagnetic Fe3+ ions in the alite as well as the belite structures. It is also demonstrated that the Fe3+ ions tend to be incorporated in the octahedral sites of alite by substitution for the Ca2+ ions. Finally, the applicability of the

29

Si{27Al} REAPDOR experiment in structural

characterizations of the aluminosilicate networks structure has been demonstrated in this project for a series of alkali-activated metakaoline samples and a synthetic sample of strätlingite. This NMR experiment allows a determination of the number of Al atoms incorporated in the secondcoordination sphere for the probed silicon atoms and thereby, discriminating between different SiOAl connectivities.

References

123

References

[1]

K. Salazar and M.K. McNutt. Mineral commodity summaries 2011. U.S. Department of the Interior and U.S. Geological Survey. United States Goverment Printing Office, Washington, 2011.

[2]

Cement industry energy and CO2 performance "Getting the Number Right". Prepared by

the World Business Council for Sustainable Development - The Cement Sustainability Initiative, 2009. [3]

E. Gartner. Industrially interesting approaches to “low-CO2” cements. Cem. Concr. Res., 2004, 34, 1489-1498.

[4]

J.S. Damtoft, J. Lukasik, D. Herfort, D. Sorrentino and E.M. Gartner. Sustainable development and climate change initiatives. Cem. Concr. Res., 2008, 38, 115-127.

[5]

Global Cement Database on CO2 and Energy Information. Prepared by the World

Business Council for Sustainable Development - Cement Sustainability Initiative. www.wbcsdcement.org/co2data. [6]

Cement Technology Roadmap 2009 - Carbon emissions reductions up to 2050. Prepared

by the World Business Council for Sustainable Development and the International Energy Agency. https://www.wbcsd.org. [7]

M. Schneider, M. Romer, M. Tschudin and H. Bolio. Sustainable cement productionpresent and future. Cem. Concr. Res., 2011, 41, 642-650.

[8]

C. Shi, A.F. Jiménez and A. Palomo. New cement for the 21st century: The pursuit of an alternative to Portland cement. Cem. Concr. Res., 2011, 41, 750-763.

[9]

E.G. Shame and F.P. Glasser. Stable Ca3SiO5 Solid Solutions containing fluorine and aluminium made between 1050 and 1250 oC. Br. Ceram. Trans. J., 1987, 86, 13-17.

[10]

G.K. Moir. Improvements in the early strength properties of Portland cement. Phil. Trans. R. Soc. Lond. A, 1983, 310, 127-138.

[11]

J. Aspdin. Producing an artificial stone. 1824, British Patent BP 5022.

References

[12]

124

The European Stardard EN 197-1, 2000. Composition, specification and conformity criteria for common cements.

[13]

H.F.W. Taylor. Cement chemistry. 2nd ed, 1997, Thomas Telford Publishing.

[14]

V. Zetola. Differences in basic standards of common cements. Cem. Concr. Res., 2011, 41, 764-766.

[15]

A.K. Chatterjee. Chemistry and engineering of the clinkerization process - Incremental advances and lack of breakthroughs. Cem. Concr. Res., 2011, 41, 624-641.

[16]

F. Sorrentino. Chemistry and engineering of the production process: State of the art. Cem. Concr. Res., 2011, 41, 616-623.

[17]

J.I. Bhatty, F.M. Miller, S.H. Kosmatka and R.P. Bohan (eds.). Innovations in Portland cement manufacturing. Portland Cement Association in Skokie, 2004, Illinois, USA.

[18]

A.A. Phillip, C. Hung, T. Herman. The cement plant operating handbook. 5th ed. ICR, Tradeship Publications Ltd, 2007, Surrey, UK.

[19]

A. Wolter. Influence of the kiln system on the clinker properties. Zement-Kalk-Gips, 1985, 38 (10), 612.

[20]

N.I. Golovastikov, R.G. Matveeva and N.V. Belov. Crystal structure of the tricalcium silicate. Kristallografiya, 1975, 20, 721-729.

[21]

F. Nishi, Y. Takeuchi and I. Maki. Tricalcium silicate: The monoclinic superstructure. Z. Kristallogr., 1985, 172, 297-314.

[22]

A.M. Il'inets, Y.A. Malinovskii and N.N. Nevskii. The crystal structure of the rhombohedral modification of tricalcium silicate. Doklady Akademii Nauk SSSR, 1985,

281, 332-336. [23]

M. Bigaré, A. Guinier, C. Mazières, M. Regourd, N. Yannaquis, W. Eysbl. Polymorphism of tricalcium silicate and its solid solutions. J. Am. Ceram. Soc., 1967, 50, 609-619.

References

[24]

125

G.K. Moir and F.P. Glasser. Mineralizers, modifiers and activators in the clinkering process. Proc. 9th International Congress on the Chemistry of Cement, New Delhi, India, 1992, Vol. I, 125-152.

[25]

V. Johansen and Bhatty J. I. Fluxes and mineralizers in clinkering process. In: J.I. Bhatty, F.M. Miller, and S.H. Kosmatka (eds.), Innovations in Portland cement manufacturing, 2004, Section 3.5, 369-402.

[26]

I. Maki and K. Goto. Factors influencing the phase constitution of alite in portland cement clinker. Cem. Concr. Res., 1982, 12, 301-308.

[27]

A.B. Diouri A., J. Aride, F. Puertas and T. Vazquez. Stable Ca3SiO5 solid solution containg Manganese and Phousporus. Cem. Concr. Res., 1997, 27, 1203-1212.

[28]

W. Sinclair and G.W. Groves. Transmission Electron Microscopy and X-Ray Diffraction of doped tricalcium silicate. J. Am. Ceram. Soc., 1984, 61, 325-330.

[29]

A.M. Harrison, H.F.W. Taylor and N.B. Winter. Electron-Optical analyses of the phases in a Portland cement clinker, with some observations on the calculation of quantitative phase composisition. Cem. Concr. Res., 1985, 15, 775-780.

[30]

A. Ghose and P. Barnes. Electron Microprobe Ananlysis of Portland cement clinker. Cem. Concr. Res., 1979, 9, 747-755.

[31]

M. Kristmann. Portland cement clinker mineralogi and chemical investigations. Cem. Concr. Res., 1978, 8, 93-102.

[32]

I. Nettleship, K.G. Slavick, Y.J. Kim and W.M. Kriven. Phase transformation in dicalcium silicate: I, Fabrication of phase stability of fine-grained  phase. J. Am.

Ceram. Soc., 1992, 75, 2400-2406. [33]

K.H. Jost, B. Ziemer and R. Seydel. Redetermination of the structure of -dicalcium silicate. Acta Cryst., 1977, B33, 1696-1700.

[34]

P. Mondal and J.W. Jeffery. The crystal structure of tricalcium aluminate, Ca3Al2O6. Acta Cryst., 1975, B31, 689-697.

126

References

[35]

L. Gobbo, L. Sant Agostino and L. Garcez. C3A polymorphs related to industrial clinker alkalies content. Cem. Concr. Res., 2004, 34, 657-664.

[36]

I. Maki. Morphology of the so-called prismatic phase in Portland cement clinker. Cem. Concr. Res., 1974, 4, 87-97.

[37]

F. Puertas, M.T. Blanco Varela and R. Dominguez. Characterization of Ca2AlMnO5. A comparative study between Ca2AlMnO5 and Ca2AlFeO5. Cem. Concr. Res., 1990, 20,

429-438. [38]

L. Bonafous, C. Bessada, D. Massiot and J.P. Coutures.

29

Si MAS NMR study of

dicalciums silicate: The structural influence of sulfate and alumina stabilizers. J. Am.

Ceram. Soc., 1995, 78, 2603-2608. [39]

S.L. Poulsen, H.J. Jakobsen and J. Skibsted. Incorporation of phosphorus guest ions in the calcium silicate phases of Portland cement from 31P MAS NMR spectroscopy. Inorg.

Chem., 2010, 49, 5522-5529. [40]

M.N. Noirfontaine, S. Tusseau-Nenez, M. Signes-Frehel, G. Gasecki and C. GirodLabiance. Effect of phosphorus impurity on tricalcium silicate T1: From synthesis to structural characterization. J. Am. Ceram. Soc., 2009, 92, 2337-2344.

[41]

H.E. Borgholm, D. Herfort and S. Rasmussen. A new blended cement based on mineralised clinker. World cement reseach and development, 1995, August, 27-33.

[42]

M.T. Blanco-Varela, A. Palomo, F. Puertas and T. Vázquez. CaF2 and CaSO4 in white cement clinker production. Adv. Cem. Res., 1997, 35, 105-113.

[43]

W.A. Klemm, I. Jawed and K.J. Holub. Effects of calcium fluoride mineralization on silicates and melt formation in portland cement clinker. Cem. Concr. Res., 1979, 9, 489-

496. [44]

W.A. Klemm and I. Jawed. Mineralizers and fluxes in the clinkering process. Proc. 9th International Congress on the Chemistry of Cement, New Delhi, India, 1992, Vol. II, 150155.

127

References

[45]

Y.B. Pliego-Cuervo and F.P. Glasser. The role of sulphate in cement clinkering: subsolidus phase relations in the system CaO-Al2O3-SiO2-SO3. Cem. Concr. Res., 1979,

9, 51-56. [46]

J. Skibsted, M.D. Andersen and H.J. Jakobsen. Structural investigations of Portland cement components, hydration, and effects of admixtures by solid-state NMR spectroscopy. Proc. 16th Internationale Baustofftagung, IBAUSIL conference, Weimer, 2006, Vol. I, 0297-0313.

[47]

J. Ambroise, S. Maximilien and J. Pera. Properties of metakaolin blended cements. Adv. Cem. Based Mater., 1994, 1, 161-168.

[48]

H.F.W. Taylor, K. Mohan and G.K. Moir. Ananlytical study of pure extended Portland cement pastes: II, Fly ash- and slag-cement pastes. J. Am. Ceram. Soc., 1985, 68, 685-

690. [49]

J. Hjorth, J. Skibsted and H.J. Jakobsen.

29

Si MAS NMR studies of Portland cement

components and effects of microsilica on the hydration reaction. Cem. Concr. Res., 1988,

18, 789-798. [50]

B. Lothenbach, G. Le Saout, E. Gallucci and K. Scrivener. Influence of limestone on the hydration of Portland cements. Cem. Concr. Res., 2008, 38, 848-860.

[51]

J. Bensted and P. Barnes (eds.). Structure and Performance of Cements. 2nd ed., 2002, Spon Press, New York., p. 57-113.

[52]

H.M. Jennings, B.J. Dalgleish and P.L. Pratt. Morphological development of hydrating tricalcium silicate as examined by Electron Microscopy techniques. J. Am. Ceram. Soc., 1981, 64, 567-572.

[53]

E. Breval. C3A Hydration. Cem. Concr. Res., 1976, 6, 129-138.

[54]

M. Collepardi, G. Baldini and M. Pauri. Tricalcium aluminate hydration in the presence of lime, gypsum or sodium sulfate. Cem. Concr. Res., 1978, 8, 571-580.

[55]

E. Gallucci, P. Mathur and K. Scrivener. Microstructural development of early age hydration shells around cement grains. Cem. Concr. Res., 2010, 40, 4-13.

128

References

[56]

G.W. Groves. Portland cement clinker viewed by transmission electron microscopy. J. Mater. Sci., 1981, 16, 1063-1070.

[57]

I.G. Richardson. The nature of C-S-H in hardened cements. Cem. Concr. Res., 1999, 29, 1131-1147.

[58]

L.S.D. Glasser, E.E. Lachowski, K. Mohan and H.F.W. Taylor. A multi-method study of C3S hydration. Cem. Concr. Res., 1978, 8, 733-740.

[59]

H.F.W. Taylor and D.E. Newbury. Calcium hydroxide distribution and calsium silicate hydrate composition in tricalcium silicate and -dicalcium silicate pastes. Cem. Concr.

Res., 1984, 14, 93-98. [60]

J.D. Bernal, J.W. Jeffery and H.F.W. Taylor. Crystallographic research on the hydration of Portland cement. A first report on investigations in progress. Mag. Concr. Res., 1952,

4, 49-54. [61]

H.F.W. Taylor. Proposed structure for calcium silicate hydrate gel. J. Am. Ceram. Soc., 1986, 69, 464-467.

[62]

I.G. Richardson and G.W. Groves. Models for the composition and structure of calcium silicate hydrate (C-S-H) gel in hardened tricalcium silicate pastes. Cem. Concr. Res., 1992, 22, 1001-1010.

[63]

X. Cong and R.J. Kirkpatrick.

29

Si MAS NMR study of the structure of calcium silicate

hydrate. Adv. Cem. Based Mater., 1996, 3, 144-156.

[64]

I.G. Richardson. The calcium silicate hydrates. Cem. Concr. Res., 2008, 38, 137-158.

[65]

I.G. Richardson and G.W. Groves. The incorporation of minor and trace elements into calcium silicate hydrate (C-S-H) gel in hardened cement pastes. Cem. Concr. Res., 1993,

23, 131-138. [66]

E. Bonaccorsi and S. Merlino. The crystal structure of Tobermorite 14 Å (Plombierite), a C-S-H phase. J. Am. Ceram. Soc., 2005, 88, 505-512.

References

[67]

129

E. Bonaccorsi, S. Merlino and H.F.W. Taylor. The crystal structure of Jennite, Ca9Si6O18(OH)68H2O. Cem. Concr. Res., 2004, 34, 1481-1488.

[68]

D.M. Henderson and H.S. Gutowsky. A nuclear magnetic resonance determination of the hydrogen positions in Ca(OH)2. American Mineralogist, 1962, 47, 1231-1251.

[69]

R.R. Ernst, G. Bodenhausen and A. Wokaun. Principles of Nuclear Magnetic Resonance in one and two Dimensions. 1987, Clarendon Press, Oxford.

[70]

J. Mason. Conventions for the reporting of nuclear magnetic shielding (or shift) tensors suggested by participants in the NATO ARW on NMR Shielding Constants at the University of Maryland, College Park, July 1992. Solid State Nucl. Magn. Res., 1993, 2,

285-288. [71]

M.J. Duer (eds.). Solid-State NMR Spectroscopy, principles and applications. Blackwell Science Ltd, 2002, Great Britain.

[72]

J. Herzfeld and A.E. Berger. Sideband intensities in NMR spectra of samples spinning at the magic angle. J. Chem. Phys., 1980, 73, 6021-6030.

[73]

M.R. Hansen, H.J. Jakobsen and J. Skibsted. 29Si Chemical Shift Anisotropies in calcium silicates from high-field 29Si MAS NMR spectroscopy. Inorg. Chem., 2003, 7, 2368-2377.

[74]

J. Skibsted, N.C. Nielsen, H. Bildsøe and H.J. Jakobsen. Satellite transitions in MAS NMR spectra of quadrupolar nuclei. J. Magn. Res., 1991, 95, 88-117.

[75]

M. Lee and W. Goldburg. Nuclear-Magnetic-Resonance line narrowing by a rotating rf Field. Phys. Rev., 1965, 140, 1261-1271.

[76]

A.E. Bennett, C.M. Rienstra, M. Auger and L.V. Lakshmi. Heteronuclear decoupling in rotating solids. J. Chem. Phys, 1995, 103, 6951-6958.

[77]

A.J. Shaka, J. Keeler, T. Frenkiel and R. Freeman. An improved sequence for broadband decoupling: WALTZ-16. J. Magn. Res., 1983, 52, 335-338.

130

References

[78]

S.L. Poulsen, V. Kocaba, G.L. Saout, H.J. Jakobsen, K.L. Scrivener and J. Skibsted. Improved quantification of alite and belite in anhydrous Portland cements by

29

Si MAS

NMR: Effects of paramagnetic ions. Solid State Nucl. Magn. Res., 2009, 36, 32-44.

[79]

E. Lippmaa, M. Mägi, A. Samoson and G. Engelhardt. Structural studies of silicates by solid-state high-resolution 29Si NMR. J. Am. Chem. Soc., 1980, 102, 4889-4893.

[80]

J. Klinowski. Nuclear Magnetic Resonance studies of zeolites. Prog. NMR Spec., 1984, 16, 237-309.

[81]

G. Engelhardt. Silicon-29 NMR of solid silicates. In: D.M. Grant and R.K. Harris (eds.), Encyclopedia of Nuclear Magnetic Resonance, John Wiley & Sons, 1996, New York, p. 4398-4407.

[82]

J. Skibsted and H.J. Jakobsen. Characterization of the calcium silicate and aluminate phases in anhydrous and hydrated Portland cements by

27

Al and

29

Si MAS NMR

spectroscopy. In: P. Colombet, A.R. Grimmer, H. Zanni, P. Sozzani (eds.), Nuclear

Magnetic Resonance Spectroscopy of Cement-Based Materials. Springer-Verlag, 1998, Berlin Heidelberg, p. 3-45. [83]

J. Skibsted, H.J. Jakobsen and C. Hall. Quantification of calcium silicate phases in Portland cements by

29

Si MAS NMR spectroscopy. J. Chem. Soc. Faraday Trans., 1995,

91, 4423-4430. [84]

A.R. Grimmer, F. Von Lampe, M. Mägi and E. Lippmaa. High-resolution solids-state 29

[85]

Si NMR of polymorphs of Ca2SiO4. Cem. Concr. Res., 1985, 15, 467-473.

J. Skibsted, J. Hjorth and H.J. Jakobsen. Correlation between

29

Si NMR chemical shifts

and mean Si---O bond lengths for calcium silicates. Chem. Phys. Lett., 1990, 172, 279-

283. [86]

J. Skibsted, E. Henderson and H.J. Jakobsen. Characterization of calcium aluminate phases in cements by 27Al MAS NMR spectroscopy. Inorg. Chem., 1993, 32, 1013-1027.

[87]

A. Labouriau, Y.W. Kim, S. Chipera, D.L. Bish and W.L. Earl. A 19F Nuclear Magnetic Resonance study of natural clays. Clays and Clay Minerals, 1995, 43, 697-704.

131

References

[88]

R.E. Youngman and S. Sen. Structural role of fluorine in amorphous silica. J. Non-Cryst. Solids, 2004, 349, 10-15.

[89]

H. Koller, A. Wölker, L.A. Villaescusa, M.J. Días-Cabanas and Valencia, S. and Camblor, M. A. Five-coordnate silicon in high-silica zeolites. J. Am. Chem. Soc., 1999, 121, 3368-3376.

[90]

Y. Liu and H. Nekvasil. Si-F bonding in aluminosilicate glasses: Inferences from ab initio NMR calculations. American Mineralogist, 2002, 87, 339-346.

[91]

J. Skibsted, M.D. Andersen and H.J. Jakobsen. Applications of solid-state Nuclear Magnetic Resonance (NMR) in studies of Portland cement-based materials. Zement-

Kalk- Gips, 2007, 60 (6), 77-83. [92]

K.J.D. Mackenzie and M.E. Smith. Multinuclear Solid-State NMR of inorganic materials. Pergamon, Elsevier science Ltd, 2002, Oxford, UK, p. 550-563.

[93]

I.L. Moudrakovski, R. Alizadeh and J.J. Beaudoin. Natural abundance high field

43

Ca

solid state NMR in cement science. Phys. Chem. Chem. Phys., 2010, 12, 6961-6969.

[94]

G.M. Bowers and R.J. Kirkpatrick. Natural abundance

43

Ca NMR spectroscopy of

Tobermorite and Jennite: Model compounds for C-S-H. J. Am. Ceram. Soc., 2009, 92,

545-548. [95]

Z. Lin and M.E. Smith. Probing the local structural environment of calcium by naturalabundance solid-state 43Ca NMR. Phys. Rev. B, 2004, 69, 224107-1-224107-7.

[96]

K.J.D. Mackenzie, M.E. Smith and A. Wong. A multinuclear MAS NMR study of calcium-containing aluminosilicate inorganic polymers. J. Mater. Chem., 2007, 17, 5090-

5096. [97]

F. Angeli, M. Gaillard, P. Jollivet and T. Charpentier. Contribution of 43Ca MAS NMR for probing the structural configuration of calcium in glass. Chem. Phys. Lett., 2007, 440,

324-328.

132

References

[98]

A. Wong, D. Laurencin, R. Dupree and M.E. Smith. Two-dimensional

43

Ca-1H

correlation solid-state NMR spectroscopy. Solids State Nucl. Magn. Res., 2009, 35, 32-

36. [99]

K. Shimoda, Y. Tobu, Y. Shimoikeda, T. Nemoto and K. Saito. Muliple Ca2+ environments in silicate glasses by high-resolution 43Ca MQMAS NMR technique at high and ultra-high (21.8 T) magnetic fields. J. Magn. Res., 2007, 186, 156-159.

[100] D. Laurencin, C. Gervais, A. Wong, C. Coelho, F. Mauri and D. Massiot. Implementation of High Resolution

43

Ca Solid State NMR Spectroscopy: Toward the elucidation of

calcium sites in biological materials. J. Am. Chem. Soc., 2009, 131, 13430-13440.

[101] I.C.M. Kwan, A. Wong, Y.M. She, M.E. Smith, and G. Wu. Direct NMR evidence for Ca2+ ion binding to G-quartets. Chem. Commun., 2008, 682-684.

[102] R.L. Vold, J.S. Waugh, M.P. Klein and D.E. Phelps. Measurement of spin relaxation in complex systems. J. Chem. Phys., 1968, 48, 3831-3832.

[103] M.R. McHenry and B.G. Silbernagel. Nuclear Spin-Lattice relaxation in the La1-cGdcAl2 intermetallic compounds. Phys. Rev. B, 1972, 5, 2958-2972.

[104] S. Hayashi and E. Akiba. Nuclear spin-lattice relaxation mechanisms in kaolinite confirmed by magic-angle spinning. Solid State Nucl. Magn. Res., 1995, 4, 331-340.

[105] S.R. Hartmann and E.L. Hahn. Nuclear double resonance in the rotating frame. Phys. Rev., 1962, 128, 2042-2053. [106] W. Kolodziejski and J. Klinowski. Kinetics of Cross-Polarization in solid-state NMR: A Guide for Chemists. Chem. Rev., 2002, 102, 613-628.

[107] J. Skibsted, S. Rasmussen, D. Herfort and H.J. Jakobsen.

29

Si cross-polarization magic-

angle spinning NMR spectroscopy - an efficient tool for quantification of thaumasite in cement-based materials. Cem. Concr. Compos., 2003, 25, 823-829.

[108] G. Metz, X.L. Wu and S.O. Smith. Ramped-Amplitude Cross Polarization in MagicAngle-Spinning NMR. J. Magn. Res. A, 1994, 110, 219-227.

133

References

[109] M. Sardashti and G.E. Maciel. Effects of sample spinning on Cross Polarization. J. Magn. Res., 1987, 72, 467-474. [110] B.H. Meier. Cross polarization under fast magic angle spinning: thermodynamical consideration. Chem. Phys. Lett., 1992, 188, 201-207.

[111] A.J. Vega. CP/MAS of quadrupolar S = 3/2 nulcei. J. Magn. Res., 1992, 1, 17-32. [112] A.J. Vega. MAS NMR spin locking of half-integer quadrupolar nuclei. J. Magn. Res., 1992, 96, 50-68.

[113] J. Schaefer, E.O. Stejskal, J.R. Garbow and R.A. McKay. Quantitative determination of the concentrations of

13

C-15N chemical bonds by Double Cross-Polarization NMR. J.

Magn. Res., 1984, 59, 150-156. [114] M. Wilhelm, H. Feng, U. Tracht and H.W. Spiess. 2D CP/MAS

13

C isotropic chemical

shift correlation established by 1H spin diffusion. J. Magn. Res., 1998, 134, 255-260.

[115] S. Saburi, A. Kawahara, C. Henmi, I. Kuschi and K. Kihara. The refinement of the crystal structure of cuspidine. Mineral. J., 1977, 8, 286-298.

[116] A. Brinkmann and M.H. Levitt. Symmetry principles in the nuclear magnetic resonance of spinning solids: Heteronuclear recoupling by generalized Hartmann-Hahn sequences.

J. Chem. Phys., 2001, 115, 357-383. [117] T. Gullion and J. Schaefer. Rotational-Echo Double-Resonance NMR. J. Magn. Res., 1989, 81, 196-200.

[118] B.V. Rossum, H. Förster and H.J.M.d. Groot. High-field and high-speed CP-MAS

13

C

NMR heteronuclear dipolar-correlation spectroscopy of solids with frequency-switched Lee-Goldburg homonuclear decoupling. J. Magn. Res., 1997, 124, 516-519.

[119] T. Gullion. Introduction to Rotational-Echo, Double-Resonance NMR. Concepts Magn. Reson., 1998, 10, 277-289. [120] K.T. Mueller. Analytic solutions for the time evolution of dipolar-dephasing NMR signals. J. Magn. Res. A, 1995, 113, 81-93.

References

134

[121] C.A. Fyfe and A.R. Lewis. Investigation of the viability of solid-state NMR distance determination in multiple spin systems of unknown structure. J. Phys. Chem. B, 2000,

104, 48-55. [122] T. Gullion and A.J. Vega. Measuring heteronuclear dipolar couplings for I = 1/2, S > 1/2 spin pairs by REDOR and REAPDOR NMR. Progress in NMR Spec., 2005, 47, 123-136.

[123] M. Bertmer, L. Zuchner, J.C.C. Chan and H. Eckert. Short and medium range order in sodium aluminoborate glasses. 2. Site connectivities and cation distributions studied by Rotational Echo Double Resonance NMR Spectroscopy. J. Phys. Chem. B, 2000, 104,

6541-6553. [124] M.R. Hansen, H.J. Jakobsen and J. Skibsted. Structural environments for boron and aluminum in alumina-boria catalysts and their precursors from 11B and 27Al Single- and Double-Resonance MAS NMR Experiments. J. Phys. Chem. C, 2008, 112, 7210-7222.

[125] L. Zhang, C.C.d. Araujo and H. Eckert. Structural role of fluoride in aluminophosphate sol-gel glasses: High-resolution Double-Resonance NMR studies. J. Phys. Chem. B, 2007, 111, 10402-40412.

[126] C.A. Fyfe, D.H. Brouwer, R.H. Levis and J.M. Chzeau. Location of the fluoride ion in tetrapropylammonium fluoride silicalite-1 determined by H/F/Si triple resonance CP, REDOR, and TEDOR NMR experiments. J. Am. Chem. Soc., 2001, 123, 6882-6891.

[127] A.A. Maudsley. Modified Carr-Purcell-Meiboom-Gill sequence for NMR fourier imaging applications. J. Magn. Res., 1986, 69, 488-491.

[128] T. Gullion, D.B. Baker and M.S. Conradi. New, compensated Carr-Purcell sequences. J. Magn. Res., 1990, 89, 479-484. [129] T. Gullion. Measurement of dipolar interactions between spin-1/2 and quadrupolar nuclei by rotational-echo, adiabatic-passage, double-resonance NMR. Chem. Phys. Lett., 1995, 246, 325-330.

[130] E. Hughes, T. Gullion, A. Goldbourt, S. Vega and A.J. Vega. Internuclear distance determination of S=1, I=1/2 spin pairs using REAPDOR NMR. J. Magn. Res., 2002, 156,

230-241.

References

135

[131] M. Bertmer and H. Eckert. Dephasing of spin echoes by multiple heteronuclear dipolar interactions in rotational echo double resonance NMR experiments. Solid State Nucl.

Magn. Reson., 1999, 15, 139-152. [132] L. Zuchner, J.C.C. Chan, W. Muller-Warmuth and H. Eckert. Short-Range Order and Site Connectivities in Sodium Aluminoborate Glasses: I. Quantification of Local Environments by High-Resolution 11B, 23Na, and 27Al Solid-State NMR. J. Phys. Chem. B, 1998, 102, 4495-4506.

[133] R.K. Harris, E.D. Becker, S.M.C. de Menezes, R. Goodfellow and P. Granger. NMR Nomenclature. Nuclear Spin Properties and Conventions for Chemical Shifts (IUPAC Recommentdations 2001). Pure Appl. Chem., 2001, 73, 1795-1818.

[134] Q. Luo, J. Yang, W. Hu, M. Zhang, Y. Yue, C. Ye. Unambiguously distinguishing Si[3Si,1Al] and Si[3Si,1OH] stuctural units in zeolite by 1H/29Si/27Al triple-resonance solid-state NMR spectroscopy. Solid State NMR, 2005, 28, 9-12.

[135] S. Ganapathy, R. Kumar, V. Montouillout, C. Fernandez and J.P. Amoureux. Identification of tetrahedrally ordered Si–O–Al environments in molecular sieves by {27Al}–29Si REAPDOR NMR. Chem. Phys. Lett., 2004, 390, 79-83.

[136] J.B. Nicholas, R.E. Winans, R.J. Harrison, L.E. Iton, L.A. Curtiss and A.J. Hopfinger. Ab initio molecular orbital study of the effects of basis set size on the calculated structure and acidity of hydroxyl groups in framework molecular sieves. J. Phys. Chem., 1992, 96,

10247-10257. [137] C.A. Fyfe, K.T. Mueller and G.T. Kokotailo. Solid-State NMR Studies of Zeolites and Related Systems. In: A.T. Bell and A. Pines. NMR Techniques in Catalysis (eds.). Marcel

Dekker, Inc., 1994, New York. [138] L. Frydman and J.S. Harwood. Isotropic spectra of half-integer quadrupolar spins from bidimensional Magical-Angle Spinning NMR. J. Am. Chem. Soc., 1995, 117, 5367-5368.

[139] D. Massiot, B. Touzo, D. Trumeau, J.P. Coutures, J. Virlet, P. Florian. Two-dimensional magic-angle spinning isotropic reconstruction sequences for quadrupolar nuclei. Solid

State NMR, 1996, 6, 73-83.

136

References

[140] J.P. Amoureux, C. Fernandez and L. Frydman. Optimized multiple-quantum magic-angle spinning NMR experiments on half-integer quadrupoles. Chem. Phys. Lett., 1996, 259,

347-355. [141] J. Amoureux, C. Fernandez and S. Steuernagel. Z-Filtering in MQMAS NMR. J. Magn. Res. Ser. A, 1996, 123, 116-118. [142] A. Goldbourt and S. Vega. Signal ehancement in 5QMAS spectra of spin-5/2 quadrupolar Nuclei. J. Magn. Res., 2002, 154, 280-286.

[143] G. Wu, D. Rovnyak and R.G. Griffin. Quantitative multiple-quantum Magic-AngleSpinning NMR spectroscopy of quadrupolar nuclei in solids. J. Am. Chem. Soc., 1996,

118, 9326-9332. [144] S.H. Wang, S.M.D. Paul and L.M. Bull. High-resolution heteronuclear correlation between quadrupolar and spin-½ nuclei using Multiple-Quantum Magic-Angle Spinning.

J. Magn. Res., 1997, 125, 364-368. [145] A. Medek, J.S. Harwood and L. Frydman. Multiple-Quantum Magic-Angle Spinning NMR: A new method for the study of quadrupolar nuclei in solids. J. Am. Chem. Soc., 1995, 117, 12779-12787.

[146] H. Saalfeld and M. Wedde. Refinement of the crystal structure of gibbsite, Al(OH)3. Z. Kristallogr., 1974, 139, 129-135. [147] S.E. Ashbrook, J. McManus, K.J.D. Mackenzie and S. Wimperis. Multiple-Quantum and Cross-Polarized

27

Al MAS NMR of mechanically treated mixtures of kaolinite and

gibbsite. J. Phys. Chem. B, 2000, 104, 6408-6416.

[148] C. Gilioli, F. Massazza and M. Pezzuoli. Studies on clinker calcium silicates bearing CaF2 and CaSO4. Cem. Concr. Res., 1979, 9, 295-302.

[149] L. Kacimi, A. Simon-Masseron, A. Ghomari and Z. Derriche. Influence of NaF, KF, and CaF2 addition on the clinker burning temperature and its properties. C. R. Chimie, 2006,

9, 154-163.

References

137

[150] J. Mukerji. Phase equilibrium diagram CaO-CaF2-2CaOSiO2. J. Am. Ceram. Soc., 1965, 48, 210-213. [151] S. Giminez-Molina, M.T. Blanco, T. Marr and F.P. Glasser. Phase relations in the system Ca2SiO4-CaO-CaSO4-CaF2 relevant to cement clinkering. Adv. Cem. Res., 1991, 14, 81-

86. [152] J.K. Christie, A. Pedone, M.C. Menziani and A. Tilocca. Fluoride evironment in bioactive glasses: ab initio molecular dynamics simulations. J. Phys. Chem., 2011, 115,

2038-2045. [153] D.S. Brauer, N. Karpukhina, R.V. Law and R.G. Hill. Structure of fluoride-containing bioactive glasses. J. Mater. Chem., 2009, 19, 5629-5636.

[154] I. Odler and S. Abdul-Maula. Structure and properties of Portland cement clinker doped with CaF2. J. Am. Ceram. Soc., 1980, 63, 654.

[155] R.T.H. Aldous. The hydraulic behaviour of rhombohedral alite. Cem. Concr. Res., 1983, 13, 89-96. [156] B. Jansang, A. Nonat and J. Skibsted. Modelling of guest-ion incorporation in the anhydrous calcium silicate phases of Portland cement by periodic Density Functional Theory calculations. Proc. CONMOD 2010, 3rd RILEM Conference on Concrete

Modelling, 2010, Lausanne, Switzerland. p. 25-28. [157] A.R. West. Chapter 5: Crystal defects, non-stoichiometry and solid solution. In: Basic Solid State Chemistry, 2nd ed. John Wiley & Sons Ltd. 2000. [158] E.M. Gartner and F.J. Tang. Formation and properties of high sulfate Portland cement clinkers. il cemento, 1987, 2, 141-164.

[159] A. Crumbie, G. Walenta and T. Füllmann. Where is the iron? Clinker microanalysis with XRD Rietveld, optical microscopy/point counting, Bogue and SEM-EDS techniques. Cem.

Concr. Res., 2006, 36, 1542-1547. [160] C.L. Edwards, L.B. Alemany and A.R. Barron. Solid-state 29Si NMR analysis of cements: comparing different methods of relaxation analysis for determining spin-lattice relaxation

138

References

times to enable determination of the C3S/C2S ratio. Ind. Eng. Chem. Res., 2007, 46, 5122-

5130. [161] W.A. Gutteridge. On the dissolution of the interstitial phases in Portland cement. Cem. Concr. Res., 1979, 9, 319-324. [162] S. Rasmussen and D. Herfort. Mineralised Clinker and Cement. In: R.K. Dhir and TD Dyer (eds.). Modern Concrete Materials: Binders, Addititions and Admixtures. Thomas Telford, 1999, London. p. 59-67. [163] F.L. Smidth Denmark A/S. Report on the Mini-RILEM method. Interal Report APr/EP, 1969.

[164] R. Kondo, M. Daimon, E. Sakai and H. Ushiyama. Influence of inorganic salts on the hydration of tricalcium silicate. J. Appl. Chem. Biotechnol., 1977, 27, 191-197.

[165] X. Cong and R.J. Kirkpatrick.

29

Si and

17

O NMR investigation of the structure of some

crystalline calcium silicate hydrates. Advn. Cem. Based Mat., 1996, 3, 133-143.

[166] G.K. Sun, J.F. Young and R.J. Kirkpatrick. The role of Al in C-S-H: NMR, XRD, and compositional results for precipitated samples. Cem. Concr. Res., 2006, 36, 18-29.

[167] M.D. Andersen, H.J. Jakobsen and J. Skibsted. Incorporation of aluminum in the calcium silicate hydrate (C−S−H) of hydrated Portland cements: A high-field 27Al and 29Si MAS NMR investigation. Inorg. Chem., 2003, 42, 2280-2287.

[168] F. Matsushita, Y. Aono and S. Shibata. Calcium silicate structure and carbonation shrinkage of a tobermorite-based material. Cem. Concr. Res., 2004, 34, 1251-1257.

[169] J.J. Chen, J.J. Thomas, H.F.W. Taylor and H.M. Jennings. Solubility and structure of calcium silicate hydrate. Cem. Concr. Res., 2004, 34, 1499-1519.

[170] I.G. Richardson. Tobermorite/jennite- and tobermorite/calcium hydroxide-based models for the structure of C-S-H: applicability to hardened pastes of tricalcium silicate, dicalcium silicate, Portland cement, and blends of Portland cement with blast-furnace slag, metakaolin, or silica fume. Cem. Concr. Res., 2004, 34, 1733-1777.

References

139

[171] A. Popova, G. Geoffroy, M.F. Renou-Gonnord, P. Faucon and E. Gartner. Interactions between polymeric dispersants and calcium silicate hydrates. J. Am. Ceram. Soc., 2000,

83, 2556-2560. [172] H. Viallis, P. Faucon, J.C. Petit and A. Nonat. Interaction between salts (NaCl, CsCl2) and calcium silicate hydrates (C-S-H). J. Phys. Chem. B, 1999, 103, 5212-5219.

[173] M.D. Andersen, H.J. Jakobsen and J. Skibsted. A new aluminium-hydrate species in hydrated Portland cements characterized by 27Al and 29Si MAS NMR spectroscopy. Cem.

Concr. Res., 2006, 36, 3-17. [174] M.D. Andersen. Structural investigations of Portland cement-based materials by solidstate NMR spectroscopy. PhD Thesis 2004, University of Aarhus, Denmark.

[175] R. Taylor, I.G. Richardson and R. Brydson. Nature of C-S-H in 20 years old neat ordinary Portland cement and 10% Portland cement - 90% ground granulated blast furnace slag pastes. Adv. App. Ceram., 2007, 106, 294-301.

[176] A. Palomo, M.W. Grutzeck and M.T. Blanco. Alkali-activated fly ashes: a cement for the future. Cem. Concr. Res., 1999, 29, 1323-1329.

[177] P. Duxson, A. Fernández-Jiménez, J.L. Provis, G.C. Lukey, A. Palomo and J.S.J. van Deventer. Geopolymer technology: the current state of art. J. Mater. Sci., 2007, 45, 29172933. [178] J. Davidovits. Geopolymer Chemistry and Applications. Institut Géopolymère, SaintQuentin, 2008. [179] K. Komnitsas and D. Zaharaki. Geopolymer: A review and prospects for the minerals industry. Minerals Engineering, 2007, 20, 1261-1277.

[180] P. Duxson, J.L. Provis and J.S.J. van Deventer. The role of inorganic polymer technology in the development of 'green concrete'. Cem. Concr. Res., 2007, 37, 1590-1597.

[181] D.M.J. Sumajouw, D. Hardjito, S.E. Wallah and B.V. Rangan. Fly ash-based geopolymer concrete: study of slender reinforced columns. J. Mater. Sci., 2007, 42, 3124-3130.

140

References

[182] M. Sofi, J.S.J. van Deventer, P.A. Demdis and G.C. Lukey. Engineering properties of inorganic polymer concretes. Cem. Concr. Res., 2007, 37, 251-257.

[183] T.W. Cheng and J.P. Chiu. Fire-resistant geopolymer produced by granulated blast furnace slag. Minerals Engineering, 2003, 3, 205-210.

[184] J.S.J. van Deventer, J.L. Provis, P. Duxson and G.C. Lukey. Reaction mechanisms in the geopolymeric conversion of inorganic waste to useful products. J. Hazard. Mater., 2007,

139, 506-513. [185] J. Davidovits. Geopolymers, Inorganic polymeric new materials. J. Therm. Anal., 1991, 37, 1633-1656. [186] D. Khale and R. Chaudhary. Mechanism of geopolymerization and factors enfluencing its development: a review. J. Mater. Sci., 2007, 42, 729-746.

[187] P.D. Silva, K. Sagoe-Crenstil and V. Sirivivatnanon. Kinetics of geopolymerization: Role of Al2O3 and SiO2. Cem. Concr. Res., 2007, 37, 512-518.

[188] T. Bakharev. Geopolymeric materials prepared using class F fly ash and elevated temperature curing. Cem. Concr. Res., 2005, 35, 1224-1232.

[189] A.G. Gelevera and K. Munzer. Alkaline Portland and slag Portland cement. Proc. 1st International Conference on Alkali Cement and Concrete 1, 1994, Kiev, Ukraine: p. 173179. [190] Y.S.A. Fundi. Alkaline pozzolana Portland cement. Proc. 1st International Conference on Alkali Cement and Concrete 1, 1994, Kiev, Ukraine: p. 181 - 192. [191] A. Fernández-Jiménez, A. Palomo, I. Sobrados and J. Sanz. The role played by the reactive alumina content in the alkaline activation of fly ashes. Micropor. Mesopor.

Mater., 2006, 91, 111-119. [192] M.R. Rowles, J.V. Hanna, K.J. Pike, M.E. Smith and H.H. O'Connor. 29Si, 23

27

Al, 1H and

Na MAS NMR Study of the bonding character in aluminosilicate inorganic polymers.

Appl. Magn. Reson., 2007, 32, 663-689.

141

References

[193] P. Duxson, J.L. Provis, G.C. Lukey, F. Separovic and J.S.J. van Deventer. 29Si NMR study of structural ordering in aluminosilicate geopolymer gels. Langmuir, 2005, 21, 3028-

3036. [194] P.S. Singh, T. Bastow and M. Trigg. Structural studies of geopolymers by

29

Si and

27

Al

MAS NMR. J. Mater. Sci., 2005, 40, 3951-3961.

[195] J. Ding, Y. Fu and J.J. Beaudoin. Strätlingite formation in high alumina cement - silica fume systems: Significance of sodium ions. Cem. Concr. Res., 1995, 25, 1311-1319.

[196] I.G. Lodeiro, A. Fernández-Jiménez, A. Palomo and D.E. Macphee. Effect on fresh C-S-H gels of the simultaneous addition of alkali and aluminium. Cem. Concr. Res., 2010, 40,

27-32. [197] X. Pardal, I. Pochard and A. Nonat. Experimental study of Si–Al substitution in calciumsilicate-hydrate (C-S-H) prepared under equilibrium conditions. Cem. Concr. Res., 2009,

39, 637-643. [198] R. Rinaldi, M. Sacerdoti and E. Passaglia. Strätlingite: crystal structure, chemistry, and a reexamination of its polytype vertumnite. Eur. J. Mineral., 1990, 2, 841-849.

[199] S. Kwan, J. LaRosa and M.W. Grutzeck. 29Si and 27Al MASNMR study of strätlingite. J. Am. Ceram. Soc., 1995, 78, 1921-1926.

References

142

Appendixes

143

Appendix 1 Sample preparations Precursors

The white Portland clinker (HKL) was received from Aalborg Portland A/S, Denmark, and has the following chemical composition: 70.30 wt. % CaO, 25.42 wt. % SiO2, 2.13 wt. % Al2O3, 0.37 wt. % Fe2O3, 0.63 wt. % MgO, 0.41 wt. % P2O5, 0.12 wt. % SO3, 0.04 wt. % F, 0.065 wt. % K2O, 0.19 wt. % Na2O, 0.01 wt. % Cl, 0.094 wt. % TiO2, 0.0033 wt. % Cr2O3, 0.03 wt. % C and a Blaine fineness of approx. 395 m2/kg. Ca(OH)2: Sigma-Aldrich Laborchemikalien GmbH, Seelze, Germany. Gypsum: VWR International Ltd., Poole, England. CaF2: Sigma-Aldrich Laborchemikalien GmbH, Seelze, Germany. Na2SiF6: Merck KGaA, Darmstadt, Germany. AlOOH: Boemitte, CONDEA Vista Company, Houston, Texas. FeOOH: Riedel-De Haën AG, Seelze – Hannover. SiO2: Silicagel, Fluka Chemie GmbH, Buchs, Switzerland. Kaolinite: Fluka Chemie GmbH, Buchs, Switzerland. NaAlO2: Strem Chemicals, New Buryport, USA. NaSiO39H2O: Fisher Chemicals, Fair Lawn, New Jersey. 10 M NaOH solution: Merck KGaA, Darmstadt, Germany. KOH: Sigma-Aldrich Laborchemikalien GmbH, Seelze, Germany. Sucrose: Merck KGaA, Darmstadt, Germany.

Fluoride-mineralized Portland clinker/cement

Clinker preparation

The fluoride-mineralized clinkers were synthesized using a white Portland clinker as the main source. The desired bulk CaO, Al2O3, Fe2O3, SO3 and F contents were achieved by adding small quantities of Ca(OH)2, AlOOH, FeOOH, CaSO4 and CaF2. Nodules with diameters of about 1 cm were formed by hand from mixtures of typically 20 g powder and 7 ml water, corresponding to a water/solid (w/s) ratio of 0.35. The wet nodules were dried at 200 ºC for one day.

Appendixes

144

The clinkers were prepared by burning the dried nodules at 1450 ( 10) ºC for one hour in a high-temperature furnace. Subsequently, the clinkers were cooled to 1250 – 1300 ºC and immediately, quenched in water at ~25 ºC. Finally, the resulting clinkers were dried at 100 ºC for one day and afterwards stored in closed containers.

Preparation of hydrated samples

The fluoride-mineralized clinkers were ground and sifted to a grain size below 40 m prior to the hydration experiments. The hydrated samples were prepared from a mixture of water (w/c = 0.5), gypsum (SO3/Al2O3 = 1.3) and typically 10 g clinker. The paste was mixed by hand for 15 minutes and subsequently stored over water (RH = 100 %) in a sealed desiccator at room temperature. The hydration was stopped after 6h, 12h, 1, 2, 7, 28, 90 and 200 days. Initially, a small piece (~1.0 g) was knocked off the paste cylinder and ground to a fine powder. Subsequently, the powder was suspended in acetone for 15 minutes under slow rotation using a magnetic stirrer. Finally, the powder was separated from the acetone and dried at room temperature in a desiccator for 24 hours. The samples were stored in small sealed specimen tubes to avoid reactions with moisture and CO2.

Selective dissolution

A solution containing 15 g KOH, 15 g sucrose and 150 ml water was heated to 90 C. A quantity of 5 g clinker was added to this solution and stirred for 1 minute. The residue was then washed in 50 ml water and 100 ml methanol and dried at 60 C.

Preparation of CSH samples

The CSH samples were prepared from typically 2 g powder mixture of Ca(OH)2, SiO2, AlOOH and Na2SiF6. The solids, with the desired composition, were mixed with three-times distilled water using a water/solid ratio of 45. The mixture was stored at room temperature for three weeks in closed glass tubes with continuous stirring using a magnetic stirrer. The reaction was stopped by filtering the suspensions and the precipitates were washed with three-time distilled water. The resulting products were dried in a desiccator over silica gel at room temperature.

Appendixes

145

Alkali-activated metakaolin

Metakaolin was prepared by heat treatment of kaolinite at 600 oC for 4 hours. The powder was cooled to room temperature in a desiccator before adding the alkali-activating solution. The desired molar Si/Al and Na/Al ratios were obtained by adjusting the quantities of metakaolin, silica and the amount of 10 M NaOH solution. Silica was mixed with the 10 M NaOH solution prior to the addition of metakaolin. The resulting mixtures were cured at 80 oC in closed glass tubes and the reactions were stopped after one week by suspending the mixtures in acetone for ca. 15 minutes. Finally, the samples were dried in a desiccator, ground, and stored in a closed container to avoid reactions with moisture and CO2.

Synthetic strätlingite

Strätlingite were prepared from a mixture of NaSiO39H2O, Ca(OH)2 and NaAlO2 in accordance with its stoichiometry of 2CaO·Al2O3·SiO2·8H2O. The powder was homogenously mixed with water using a w/s ratio of 10. Subsequently, the mixture was cured in an ultrasound bath for 53 days at room temperature. Finally, the powder was filtrated from the solution and dried at 100 ºC for one day.

146

Appendixes

Appendix 2 NMR measurements and other analytical techniques The Instrument Centre for Solid-State NMR Spectroscopy has four Varian NMR spectrometers with different magnetic fields: Unity INOVA-200 (4.7 T), Unity INOVA-300 (7.1 T), Unity INOVA-400 (9.4 T) and Direct Drive VNMRS-600 (14.1 T).

29

Si MAS NMR experiments

The 29Si MAS NMR spectra (7.1 T) were recorded using a home-built CP/MAS probe for 7 mm o.d. rotors, a pulse width of 3 s for an rf-field strength of SiSiB1/2 = 58 kHz, a spinning speed of R = 7.0 kHz, a 30-s relaxation delay and typically 2048 scans. The

29

Si{19F} CP-REDOR NMR experiments (7.1 T) for the fluoride-mineralized

Portland clinkers used a home-built CP/MAS probe for 5 mm o.d. rotors, R = 10.0 kHz, a 8-s relaxation delay and 10560 scans. The HH-matching was obtained for Si = 38 kHz, F = 47 kHz and the CP part used a decreasing linear ramped amplitude (RAMP) on the 19F channel with an amplitude variation of 10 kHz and a CP contact time of 3.0 ms. The rf-field strengths (rf = Brf/2 applied in the REDOR sequence are Si = 38 kHz and F = 67 kHz, corresponding to pulses of 13.2 and 7.5 s, respectively. The

29

Si{19F} CP/MAS experiments for this series of

samples were performed with the same conditions but used a spinning speed of 5.0 kHz, Si = 47.5 kHz and F = 42.1 kHz. The

29

Si{19F} CP/MAS NMR spectra (7.1 T) of the synthetic CSH samples were

recorded using a home-built CP/MAS probe for 7 mm o.d. rotors, R = 10.0 kHz, a 10-s relaxation delay and 16,384 – 22,400 scans. The HH-matching was obtained for Si = 41.0 kHz, F = 31.2 kHz and the CP part used a RAMP pulse of 10 kHz on the 19F channel for a CP contact time of 5.0 ms. The

29

Si{27Al} REAPDOR NMR experiments (14.1 T) used a triple-resonance MAS

probe for 5 mm o.d. rotors from DOTY Scientific Inc. The spectra were acquired using rf-field strengths of Si = 49 kHz (/2-pulse = 5.1 s and -pulse = 10.2 s) and Al = 50 kHz, a spinning speed of R = 10.0 kHz, a 30-s relaxation delay and typically 960 scans for the individual spectra without (S0) and with (S) 27Al irradiation.

147

Appendixes

27

Al MAS NMR experiment 27

Al MAS NMR spectra (14.1 T) were recorded using a home-built CP/MAS probe for 4

mm o.d. rotors, a pulse width of 0.5 s for an rf-field strength of Al = 60 kHz to ensure quantitative reliability of the intensities observed for the

27

Al central transitions for sites

experiencing different quadrupole couplings. The 27Al NMR experiments typically employed 1H decoupling with H = 50 kHz, a spinning speed of 13.0 kHz, a 2-s relaxation delay and ~3,200 scans. Furthermore, the

27

Al MAS NMR spectrum of the probe itself with an empty spinning

PSZ rotor showed a broadened resonance, which was subtracted from the

27

Al MAS NMR

spectra of the samples prior to the evaluation of spectra. The triple-quantum

27

Al MQMAS experiment (9.4 T) was performed using the three-

pulse z-filter sequence with an rf-field strength of B1/2 = 60 kHz for the first and second pulses and B1/2 = 25 kHz for third selective 90 pulse. Furthermore, 1H decoupling (B2/2 = 40 kHz) was employed in the evolution and detection periods. The

27

Al{19F} CP/MAS NMR experiment (7.1 T) was performed using a home-built

CP/MAS probe for 7 mm o.d. rotors. The Hartmann-Hahn (HH) match was obtained for Al = 12.5 kHz and F = 35.8 kHz using a CP contact time of 1.5 ms, R = 5.0 kHz, a relaxation delay of 4 s and 72,384 scans. The CP experiment was further improved using a decreasing linear ramped amplitude on the 19F channel with an amplitude variation of 15 kHz on the 19F channel during the CP contact time.

19

F MAS NMR MAS experiments

The 19F MAS NMR spectra (7.1 T) were recorded using a home-built CP/MAS probe for 7 mm o.d. rotors, a pulse width of 5 s for an rf-field strength of F = 50 kHz, R = 10.0 kHz, a 10-s relaxation delay and typically 512 scans for the anhydrous samples, 2048 – 8192 for hydrated Portland cements and 128 for the synthetic CSH. The 19F MAS NMR spectra for the synthetic CSH samples recorded at 14.1 T used a home-built CP/MAS probe for 4 mm o.d. rotors, a pulse width of 5.6 s for an rf-field strength of 48 kHz, a R = 13.0 kHz, a 15-s relaxation delay and typically 512 scans.

148

Appendixes 43

Ca MAS NMR experiments

The 43Ca MAS NMR experiments were performed at 14.1 T used two different probes: a home-built CP/MAS probe for 7 mm o.d. rotors and a Chemmagnetic probe for 7.5 o.d. rotor. The Hartmann-Hahn (HH) match for the

43

Ca{19F} CP/MAS NMR spectrum was obtained for

AlB1/2 = 13 kHz and FB2/2 = 36 kHz using a CP contact time of 1.5 ms, R = 5.0 kHz, a relaxation delay of 4 s and 72,384 scans. The

19

F,

27

Al, 29Si and

43

Ca chemical shifts were referenced to external samples of neat

tetramethylsilane (TMS) using a secondary reference sample of -Ca2SiO4 (–71.33 ppm), a 1.0 M AlCl3.6H2O solution, CFCl3 using a secondary reference sample of Na2SiF6 (-149.3 ppm) and a 1.0 M CaCl2 solution, respectively. Other analyses including measurements of the free lime content, the bulk fluoride content and the chemical compositions by XRF and XRD Rietveld analyses have been conducted at the chemical laboratory at Aalborg Portland A/S.

Free lime content measurement

The free lime content was measured by dissolving the clinker in ethylene glycol. Subsequently, the solution was titrated with hydrochloric acid.

Bulk fluoride content measurement

The clinkers were dissolved in a solution of water, HCl and alum KAl(SO4)2. The fluoride concentration in the resulting mixture was measured by an ion selective electrode. The solvent were prepared by dissolving 16 g alum in 800 ml water and 800 ml conc. HCl and finally, diluted with water to 2.0 litres.

Appendixes

149

Procedure for fitting REDOR data in Mathematica Data data = {{x1,y1},{x2,y2}...,{xn,yn}} g1 = 2.518148*10^8 g2 = 5.3190*10^7 u = 4Pi*10^(-7) h = 1.0546*10^(-34) spin = 10000 k=

Fitting