Durability of Portland Cement - Calcined Clay

Zhenguo Shi

SCIENCE AND TECHNOLOGY AARHUS UNIVERSITY

Durability of Portalnd Cement - Calcined Clay - Limestone Blends

Front cover illustration. Carbonation front indicated by phenolphthalein for the P, L, ML and M mortars (40 mm × 40 mm) in clockwise order starting from the upper-left after 91 days of hydration.

Durability of Portland Cement - Calcined Clay - Limestone Blends PHD THESIS

Zhenguo Shi 石振国 PhD Thesis 2016

Interdisciplinary Nanoscience Center, 2016

DURABILITY OF PORTLAND CEMENT – CALCINED CLAY – LIMESTONE BLENDS

PhD thesis by

Zhenguo SHI

Interdisciplinary Nanoscience Centre (iNANO) Department of Chemistry Aarhus University Denmark

January 2016

“If your experiment needs statistics, you ought to have done a better experiment.”

– Ernest Rutherford

PREFACE This PhD thesis presents the research outcome of my three-year PhD study from January 2013 to December 2015 at the Department of Chemistry and the Interdisciplinary Nanoscience Center (iNANO) at Aarhus University. The study is based on the investigations conducted for Work Package 3 of the LowE-CEM project (2012 – 2017), which focuses on durability studies of white Portland cement – calcined clays – limestone blends. The LowE-CEM project funded by the Danish Strategic Research Council was launched as a joint collaboration between iNANO (Aarhus University, Denmark), Section of Chemistry (Aalborg University, Denmark), Norwegian University of Science and Technology (NTNU, Trondheim, Norway), Swiss Federal Laboratories for Materials Science and Technology (EMPA, Zürich, Switzerland), Cementir - Aalborg Portland A/S (Denmark), and FLSmidth A/S (Denmark). Taking full advantage of the different test setup and research experiences, the work presented in this thesis has been performed by the author with the assistance of colleagues at the different laboratories and institutes as summarized below: 

Materials and sample preparations (Chapter 3) . . . . . . . . . . . . Aalborg Portland A/S



Sample characterizations (Chapter 4) . . . . . . . . . . NTNU/EMPA/Aarhus University



Chloride ingress (Chapter 5, 8) and chloride binding (Chapter 6) . . .NTNU/SINTEF



Carbonation (Chapter 7 and 8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . EMPA



Sulfate attack (Chapter 8) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Aarhus University



Thermodynamic modelling (Chapter 6 and 7) . . . . . . . . . . . . . . . . . . . . . . . . . EMPA

The contributions from these colleagues, not only for assistance in the laboratory but also for the constructive discussions, must be acknowledged.

i

ii

ACKNOWLEDGEMENTS The Danish Council for Strategic Research is acknowledged for financial support to the LowE-CEM project. Without help from many people mentioned below, this three-year PhD would have been more difficult. Thanks to my supervisor Jørgen Skibsted for giving me the opportunity, independence and continued support to do this PhD project in Demark and abroad. I am deeply grateful for his guidance, supervision and discussions. Thank you for providing me with the best courses and many opportunities to participate in conferences, which are essentially helpful for my professional development. Thank you also for spending your precious time to make the quality of the manuscripts and this thesis much better. I can’t say enough to thank Barbara Lothenbach and Mette Rica Geiker, the cosupervisors on this project, for sharing their broad knowledge, invaluable comments, suggestions and encouragements. Thank both of you also for guidance when I stayed at EMPA and NTNU. Thank you Barbara for introducing me thermodynamic modelling and also invitations to workshops, parties and beers. Thanks to Mette for spending enormous time and effort on the discussion of the experimental plans, test methods and interpretation of the results. Thank you also for guidance to write manuscripts and for your hospitality whenever I come to Trondheim. Special and great thanks to Sergio Ferreiro Garzón from Aalborg Portland A/S for his assistance in the laboratory to prepare most of the samples described in chapter 5 and for comment on the manuscripts. Special thanks also to Klaartje De Weerdt for her generosity to share her test setup and methods to investigate the chloride ingress (chapter 5) and chloride binding (chapter 6). Thank you for your time and efforts, which are significantly above and beyond the call of duty, to make this project run smoothly at NTNU and SINTEF. Thanks to the colleagues from EMPA, Josef Kaufmann for the MIP measurements and discussions of the MIP data presented in chapter 4 and chapter 7. Andreas Leemann for the SEM measurements and discussion presented in chapter 7. Frank Winnefeld for discussions on chloride binding data presented in chapter 6. Nikolajs Toropovs from Riga Technical University for his great help on TGA measurement during his visit at EMPA. These data are presented in chapter 5 and chapter 7. Käppeli Marcel for measurements of carbonation depths presented in chapter 8. Walter Trindler for iii

providing a nice place to stay, and for taking care of my samples, organizing parties, being tour guide to Elm and Greifensee. Thanks to Luigi Brunetti, Boris Ingold, Stephan Geis and Daniel Käppeli for their hand in the laboratory. Thanks to the colleagues from SINTEF, Tone Anita Østnor for measuring the data used for the chloride binding isotherm and TGA quantification used in chapter 6. AnneKristin Mjøen for measuring the total chloride content presented in chapter 8 and for preparing the distilled water whenever I needed. Knut Lervik for organizing the chloride profile grinding for chapter 5 and chapter 8. Thanks to Ove Loraas from NTNU for taking care of my samples when I was not there and also for invitations to the laboratory parties. Thanks to Finn Bendixen from Aalborg Portland A/S for assistance to prepare mortar samples used in chapter 8. Thanks to my office colleague at Aarhus University, Wolfgang Kunther for help on the measuring part of the sulfate expansions presented in chapter 8 and sulfate profiles which is not included in the thesis. Thank you for being a nice office college and free consultant for many questions. Thanks to my former college Zhuo Dai for receiving me and providing me accommodation when I came to Denmark, and for sharing compressive strength and hydration data used in chapter 4. Kasper EnemarkRasmussen for the XRD quantifications presented in chapter 5. Thanks to Anne Birgitte Johannsen, Erika Vigna, Julian Holzinger, Malene Thostrup Pedersen, Cristina Ruiz-Santaquiteria and Michael Ryan Hansen for being nice colleges, for help, and for creating a nice research atmosphere. Thanks to my former master advisor Caijun Shi from Hunan University for introducing me to the world of cement, and for his persistent encouragement and advices to my professional development. Last, but not the least, I would like to express my greatest acknowledgement to my family for their unlimited support and encouragement, especially to my loving wife, Yan Yue, for her constant love and quiet patience to endure my absence. I apologize for not being able to mention everyone who I am sure I would have wished to thank.

Zhenguo Shi January 2016, Aarhus University

iv

ABSTRACT Clay minerals represent a promising source of new supplementary cementitious materials (SCMs) that receive increasing research interests because of their low carbon footprint and high abundance in the Earth’s crust. A combined utilization of calcined clays with limestone allows replacements of Portland cement as high as 35 wt.% without sacrificing compressive strengths of the concrete. However, the durability of these blends have not been explored so far and thus, their interactions with chloride, carbonate and sulfate ions need to be investigated prior to real applications of these materials on an industrial scale. This has been the principal aim of this PhD study. The main results have shown that all the studied calcined clay blends give an improved resistance to chloride ingress and sulfate attack, as compared to pure Portland cement, but a poor resistance against carbonation. In addition, there is no relationship between the durability indices and the compressive strengths for the studied calcined clay blends. The observed expansions by sulfate attack are in very good agreement with the mass variations. Comparison of the pore structures for the metakaolin and montmorillonite (or glass) blends reveals that the high sulfate resistance for the metakaolin blends is not solely attributed to the refinement in pore structures, since the pore structures for the montmorillonite (or glass) blends are similar to the mortars of pure Portland cement. A new method to quantify Friedel’s salt is developed based on thermogravimetric measurements and used in studies of chloride ingress in the mortars. Comparison of the measured chloride in Friedel’s salt with the total chloride content indicates that Friedel’s salt gives only a small contribution to the chloride binding. The high chloride binding observed for metakaolin blends is mainly attributed to the formation of Friedel’s salt. Moreover, additional calcium ions in the exposure solution can significantly increase the chloride binding in the calcium-silicate-hydrate (C-S-H) phase and promote the conversion of monocarbonate to Friedel’ salt. A two-step decalcification process of the C-S-H phase, resulting from carbonation, is predicted where calcium initially is released from the interlayer of the high-Ca C-S-H, resulting in the formation of a low-Ca C-S-H phase and a reduction in pH to 9.6. In general, the data obtained in the present study confirm that the CO2 binding capacity, the porosity, and the capillary condensation are the decisive parameters governing the carbonation depth in mortars exposed to carbonation.

v

vi

RESUME (IN DANISH) Lermineraler udgør en lovende ressource til fremstilling af nye 'supplementary cementitious materials' (SCMs), som tiltrækker stor forskningsinteresse på grund af deres lave CO2 fodaftryk og høje forekomst i jordskorpen. En udnyttelse af calcineret ler sammen med kalksten tillader op til 35 wt.% erstatning af Portland cement uden at dette har en betydning for styrken. Holdbarheden af disse blandinger er imidlertid ikke undersøgt i detalje, og det kræves derfor, at deres vekselvirkning med klorid-, carbonatog sulfat-ioner klarlægges inden at de kan anvendes på industriel skala. Dette har været det primære formål i dette PhD studium. Hovedresultaterne viser, at alle studerede blandinger med calcineret ler giver en forbedret modstand mod sulfat-angreb og kloridion indtrængning, sammenlignet med ren Portland cement, men en ringere resistens mod carbonatisering. Derudover er det vist, at der ikke er nogen sammenhæng mellem holdbarhedsindikatorerne og kompressionsstyrken for blandingerne med calcineret ler. De observerede ekspansioner ved sulfat-angreb er i meget god overensstemmelse med variationerne i elementernes masse. En sammenligning af pore-strukturen for blandingerne med metakaolin og montmorillonite (eller glas) viser, at den høje sulfatbestandighed for metakaolin-blandingerne ikke kun er relateret til en forfinet porestruktur, idet blandingerne med montmorillonite (eller glas) udviser den samme pore-struktur som elementerne fremstillet med ren Portland cement. En ny metode til kvantificering af Friedels salt er blevet udviklet baseret på termogravimetri og brugt i studier af klorid-ion indtrængning. En sammenstilling af det målte klorid-indhold bundet i Friedel’s salt med det totale klorid-indhold indikerer, at Friedel’s salt kun giver et lille bidrag til den samlede klorid-binding. Den høje kloridbinding for blandingerne med metakaolin kan imidlertid tilskrives Friedel’s salt. Derudover vil calcium-ioner i eksponeringsvæsken forøge klorid-bindingen i calciumsilikat-hydrat (C-S-H) fasen og fremskynde omdannelsen af monocarbonat-fasen til Friedel’s salt. En to-trins proces for de-calcificeringen af C-S-H fasen, forårsaget af carbonatisering, er blevet forudsagt, hvor calcium i starten frigives fra mellemlagsstrukturen af C-S-H faser med et højt calcium indhold, hvilket giver en C-S-H fase med et lavt calciumindhold samt en reduktion i pH til 9.6. De opnåede data i dette studium viser generelt at CO2 bindingskapaciteten, porøsiteten og kapillarkondensationen er afgørende parametre for carbonatiseringsdybden af elementer udsat for CO2. vii

viii

CONTENTS PREFACE ........................................................................................................................ I ACKNOWLEDGEMENTS ......................................................................................... III ABSTRACT .................................................................................................................... V RESUME (IN DANISH) ............................................................................................. VII CHAPTER 1: INTRODUCTION .................................................................................. 1 1.1 BACKGROUND .......................................................................................................... 2 1.2 THE LOWE-CEM PROJECT ....................................................................................... 4 1.3 OBJECTIVES .............................................................................................................. 5 CHAPTER 2: DURABILITY STUDIES OF PORTLAND CEMENT-BASED MATERIALS .................................................................................................................. 7 2.1 HYDRATION.............................................................................................................. 8 2.1.1 Portland cement and SCMs.............................................................................. 8 2.1.2 Hydration products .......................................................................................... 9 2.2 CHLORIDE IN HYDRATED CEMENTS ........................................................................ 12 2.2.1 Evaluation of the chloride resistance............................................................. 12 2.2.2 Chloride binding in AFm phases ................................................................... 12 2.2.3 Chloride binding in the C-S-H phase ............................................................. 14 2.2.4 Chloride binding in the other phases ............................................................. 18 2.2.5 Impact of cement type .................................................................................... 18 2.2.6 Impact of cations ............................................................................................ 19 2.2.7 Impact of pH................................................................................................... 19 2.2.8 Impact of SCMs .............................................................................................. 21 2.2.9 Improvement of the chloride resistance ......................................................... 21 2.3 CARBONATION ....................................................................................................... 23 2.3.1 Carbonation mechanism ................................................................................ 23 2.3.2 Carbonation front and profiles ...................................................................... 25 2.3.3 Change in pH ................................................................................................. 25 2.3.4 Change in microstructure .............................................................................. 26 2.3.5 Change in compressive strength .................................................................... 27 2.3.6 Factors affecting carbonation resistance ...................................................... 27 2.4 SULFATE ATTACK ................................................................................................... 30 2.4.1 Formation of ettringte .................................................................................... 30

2.4.2 Formation of gypsum ..................................................................................... 31 2.4.3 Formation of thaumasite ................................................................................ 32 2.4.4 Mitigation of sulfate attack ............................................................................ 33 CHAPTER 3: MATERIALS AND METHODS ........................................................ 35 3.1 MIXES WITH THE SAME WATER/BINDER RATIO (BATCH 1) ...................................... 36 3.1.1 Materials ........................................................................................................ 36 3.1.2 Mix design ...................................................................................................... 36 3.1.3 Mortar preparation ........................................................................................ 38 3.1.4 Paste preparation........................................................................................... 40 3.2 MIXES WITH COMPARABLE COMPRESSIVE STRENGTHS ........................................... 41 3.2.1 Materials ........................................................................................................ 41 3.2.2 Mix design and mortar preparation ............................................................... 41 3.3 METHODS ............................................................................................................... 44 3.3.1 Workability measurement .............................................................................. 44 3.3.2 Strength testing .............................................................................................. 44 3.3.3 PF method ...................................................................................................... 44 3.3.4 Mercury intrusion porosimetry ...................................................................... 45 3.3.5 AC impedance ................................................................................................ 45 3.3.6 Scanning electron microscopy ....................................................................... 45 3.3.7 Thermogravemetric analysis .......................................................................... 46 3.3.8 X-ray diffraction analysis .............................................................................. 46 3.3.9 Thermodynamic modelling............................................................................. 46 CHAPTER 4: SAMPLE CHARACTERIZATION ................................................... 49 4.1 PHASES ASSEMBLAGES ........................................................................................... 50 4.1.1 wPc – MK – SF – LS mortars ........................................................................ 50 4.1.2 wPc’ – MT –LS and wPc’ – G – LS mortars.................................................. 51 4.2 COMPRESSIVE STRENGTH ....................................................................................... 52 4.2.1 wPc – MK – SF – LS mortars ........................................................................ 52 4.2.2 wPc’ – MT – LS and wPc’ – G – LS mortars................................................. 53 4.3 PORE STRUCTURES ................................................................................................. 54 4.3.1 wPc – MK – SF – LS mortars ........................................................................ 54 4.3.2 wPc’ – MT – LS and wPc’ – G – LS mortars................................................. 58 4.4 CONCLUSIONS ........................................................................................................ 60

CHAPTER 5: ROLE OF FRIEDEL’S SALT ON CHLORIDE BINDING FOR MORTARS UNDER CHLORIDE INGRESS ............................................................ 63 5.1 INTRODUCTION ....................................................................................................... 64 5.2 EXPERIMENTAL ...................................................................................................... 64 5.2.1 Total chloride profile analysis ....................................................................... 65 5.2.2 Quantification of Friedel’s salt and portlandite by TGA ............................... 66 5.3 RESULTS AND DISCUSSION...................................................................................... 68 5.3.1 Total chloride profiles .................................................................................... 68 5.3.2 Friedel’s salt in mortars ................................................................................ 69 5.3.3 X-ray diffraction analysis............................................................................... 71 5.3.4 Effect of total chloride content on formation of Friedel’s salt ...................... 72 5.3.5 Role of Friedel’s salt on chloride binding ..................................................... 73 5.3.6 Evaluation of the chloride resistance............................................................. 74 5.4 CONCLUSIONS ........................................................................................................ 75 CHAPTER

6:

CHLORIDE

BINDING

IN

PORTLAND

CEMENT

–

METAKAOLIN – LIMESTONE PATES .................................................................. 77 6.1 INTRODUCTION ....................................................................................................... 78 6.2 EXPERIMENTAL ...................................................................................................... 79 6.2.1 Chloride exposure .......................................................................................... 79 6.2.2 Determination of the chloride binding isotherms .......................................... 79 6.2.3 Quantification of Friedel’s salt by TGA ........................................................ 80 6.3 RESULTS ................................................................................................................. 80 6.3.1 Chloride binding isotherms ............................................................................ 80 6.3.2 Chloride in Friedel’s salt as measured by TGA ............................................ 81 6.3.3 Thermodynamic modeling .............................................................................. 85 6.3.4 pH of the exposure solutions .......................................................................... 86 6.4 DISCUSSION ............................................................................................................ 87 6.4.1 Chloride binding in hydrated Portland cement ............................................. 87 6.4.2 Chloride binding in blended cements ............................................................. 88 6.4.3 Impact of pH on chloride binding .................................................................. 89 6.5 CONCLUSIONS ........................................................................................................ 90 CHAPTER 7: IMPACT OF CARBONATION ON THE MICROSTRUCTURE AND CHEMISTRY OF MORTARS .......................................................................... 93 7.1 INTRODUCTION ....................................................................................................... 94

7.2 EXPERIMENTAL ...................................................................................................... 94 7.2.1 CO2 exposure ................................................................................................. 94 7.2.2 Phenolphthalein spray method ...................................................................... 94 7.3 RESULTS AND DISCUSSION ..................................................................................... 95 7.3.1 Apparent carbonation depths ......................................................................... 95 7.3.2 Mercury intrusion porosimetry (MIP) ........................................................... 96 7.3.3 Scanning electron microscopy (SEM) ............................................................ 98 7.3.4 Thermogravimetric analysis (TGA) ............................................................. 100 7.3.5 Carbonation profiles .................................................................................... 102 7.3.6 Thermodynamic modeling ............................................................................ 105 7.3.7 Discussion .................................................................................................... 108 7.4 CONCLUSIONS ...................................................................................................... 112 CHAPTER 8: IMPACT OF COMPRESSIVE STRENGTH ON DURABILITY 115 8.1 INTRODUCTION ..................................................................................................... 116 8.2 EXPERIMENTAL .................................................................................................... 117 8.2.1 Chloride ingress and carbonation ............................................................... 117 8.2.2 Changes in mass and lengths ....................................................................... 117 8.3 RESULTS AND DISCUSSION ................................................................................... 118 8.3.1 Carbonation resistance ................................................................................ 118 8.3.2 Chloride resistance ...................................................................................... 119 8.3.3 Sulfate resistance ......................................................................................... 121 8.4 CONCLUSIONS ...................................................................................................... 126 CHAPTER 9: CONCLUSIONS ................................................................................ 127 REFERENCES ............................................................................................................ 131 APPENDIX I: HYDRATE PHASE ASSEMBLAGES ............................................ 145 APPENDIX II: PUBLICATIONS ............................................................................. 149

GLOSSARY OF NOTATIONS AND TERMS The following are the basic abbreviations used in cement chemistry: —

C

CaO

S

S

SiO2

C

CO2

A

Al2O3

F

Fe2O3

H

H2O

—

SO3

These abbreviations are used in the formulas for anhydrous and hydrates phases: C3 S

3CaO·SiO2

C2 S

2CaO·SiO2

C3 A

3CaO·Al2O3

C4AF

4CaO·Al2O3·Fe2O3

C-S-H

xCaO·ySiO2·zH2O

CH

CaO·H2O or Ca(OH)2

AFt = Al2O3-Fe2O3-tri

C3(A,F)·3CX· Hy or C6(A,F)X3·Hy or {Ca3(Al,Fe)(OH)6·12H2O}2·X3·xH2O

—

Ettringite {Ca3Al(OH)6·12H2O}2·(SO4)3·xH2O

C6AS3H32

or 3CaO·Al2O3·3CaSO4·32H2O AFm = Al2O3-Fe2O3-mono

C3(A,F)·CaX2·y H or C4(A,F)X2· H or {Ca2(Al,Fe)(OH)6}·X·xH2O

—

—

—

—

C4ASH12 or C3A·CS·H12 C4ACH11 or C3A·CC·H11 —

Monosulfate Monocarbonate

C4AC0.5H12

Hemicarbonate

C2ASH8

Strätlingite

Ca4Al2(Cl)2(OH)12·4H2O

Friedel’s salt

Ca4Al2(Cl)(SO4)0.5(OH)12·6H2O

Kuzel’s salt

Ca3{Si(OH)6·12H2O}(SO4)(CO3)

Thaumasite

{Mg0.75Al0.25(OH)2}(CO3)0.125(H2O)0.5

Hydrotalcite

The following abbreviations are also used in this thesis: TGA

thermogravimetry analysis

DTG

differential thermogravimetry

NMR

nuclear magnetic resonance

MIP

mercury intrusion porosimetry

XRD

X-ray diffraction

XRF

X-ray fluorescence

GEMS

Gibbs Energy Minimization Software

wPc or wPc’

white Portland cement

LS

limestone

MK

metakaolin

MT

montmorillonite

SF

silica fume

G

glass

SP

superplasticizer

SCMs

supplementary cementitious materials

w/b

water to binder ratio

b/s

binder to sand ratio

LOI

loss on ignition

RH

relative humidity

Chapter 1: Introduction

CHAPTER 1: INTRODUCTION Chapter 1 describes the background for the LowE-CEM project and the objectives of this thesis. The research problem and research needs are described in the background, followed by a description of Work Package 3 of the LowE-CEM project, on which the thesis is based. At last, the objectives are stated. The outline of the thesis is not provided in this chapter, since a short description is given at the beginning of each of the chapters.

1

Durability of Portland Cement – Calcined Clay – Limestone Blends

1.1 Background Portland cement is and will still remain the essential “glue” to form concrete, which is the unique and most widely used material to satisfy global demands for housing and modern infrastructure (Wray and Scrivener, 2012). A tremendous growth in Portland cement production has been made during the past several decades (Schneider et al., 2011), and today’s annual global cement production has reached 3.5 billion tons according to the European Cement Association Activity Report released in 2014 (CEMBUREAU, 2014). As a consequence of the large quantity produced and the very high temperatures (about 1450 oC) needed to produce cement clinker together with the energy needed for the grinding process, the cement production is responsible for 5 – 8 % of the global man-made CO2 emissions. This leads to challenges faced worldwide by the cement industries and cement scientists in conserving material and energy resources as well as reducing CO2 emissions. The most promising route to meet the challenges mentioned above is to reduce the cement clinker production by partially replacing it with supplementary cementitious materials (SCMs) without additional clinkering process, which are either introduced in blended cements or added separately in the concrete mixer (Lothenbach et al., 2011). Nowadays, a vast majority of SCMs such as fly ash and slag have been used to partially substitute Portland cement. This results in a substantial decrease in the world’s average percent clinker in cement from 85% in 2003 to 77% in 2010 (Schneider et al., 2011). However, the availability of the most widely used fly ashes and slags is limited since their annual global productions are approximately 1 billion tons and 360 million tons, respectively (Juenger and Siddique, 2015). This leads to a shift to exploration and development of other alternative SCMs. Clay minerals, which are abundant in the Earth’s crust, become a promising source for alternative SCMs with a lower environmental impact for its lower temperature (about 500 – 600 oC) for calcination. The utilization of calcined clay in the form of metakaolin as a SCM for concrete has received considerable interest (Sabir et al., 2001, Siddique and Klaus, 2009). Limestone represents another interesting material and it is commonly added in small amounts to Portland cements, where it increases the early strength, reduces the water demand and improves the rheology of the fresh concrete (Tsivilis et al., 1999, Herfort, 2004, Voglis et al., 2005, Lothenbach et al., 2008a). Furthermore, limestone provides nucleation sites for the formation and growth of the calcium-silicate-

2


hydrate (C-S-H) phase and it is also partially consumed during hydration, resulting in the formation of calcium monocarboaluminate hydrate (Ca4Al2(OH)12CO3·5H2O) (Herfort, 2004, Matschei et al., 2007, Lothenbach et al., 2008a). The interaction between limestone and the AFm phase, leading to formation of monocarboaluminate, is amplified as a result of additional aluminate brought in the system by alumina-rich SCMs (e.g. fly ash) (De Weerdt et al., 2011a, De Weerdt et al., 2011b). A synergetic effect between metakaolin and limestone has also been observed as seen by an increase in compressive strength (Steenberg et al., 2011, Antoni et al., 2012, Antoni, 2013). This finding enables both the calcined clays and limestone to be used in a high level replacement of the Portland cement in ternary blended cements. The most recent significant research efforts focusing on the development and characterization of these new ternary Portland cement blends is published in the Proceedings of the 1st International Conference on Calcined Clays for Sustainable Concrete (Scrivener and Favier, 2015) with the principal aim of reducing CO2 emissions associated with Portland cement production. Most studies of Portland cement – calcined clays – limestone blends have focused on the reactivity and optimization of the calcination temperature for obtaining the best compressive strength. Few studies focus on their durability, which is very important for reducing the cement production and CO2 emission associated with the potential restorations and reconstructions of housings and infrastructures (Mehta and Monteiro, 2006, Grantham et al., 2009). Furthermore, although significant efforts have been made to study the durability of Portland cement – SCMs blends, the deterioration mechanisms still remain partially unknown, e.g. (i) how carbonation affects the changes in pore structures and how the pore structures relate to the changes in phase assemblages as a result of carbonation? (ii) although alumina-rich SCMs can enhance the chloride binding, this is commonly ascribed to formation of Friedel’s salt and the role of C-S-H in these blends is usually not known. Even for the chloride binding in hydrated Portland cement, whether C-S-H or Friedel’s salt dominates the chloride binding is still not clarified. The answers for the (i) and (ii) questions are seen by discrepancies in published studies. (iii) Exploration of the key factor(s) (e.g. porosity, pore connectivity, compressive strength, chemical compositions of the raw materials etc.) is important for controlling the durability performance. Some studies have established relationships between durability indices and compressive strengths (Khan and Lynsdale, 2002, Baghabra Al-Amoudi et al., 2009, Rabehi et al., 2013, Lee and Yoon, 2014), which still 3


need to be evaluated. Obviously, a better understanding of the corresponding durability mechanisms is essential to explore the key factors controlling the durability performance.

1.2 The LowE-CEM project In order to meet the goal to reduce the CO2 emission associated with the production of Portland cement materials, the LowE-CEM project (2012 – 2017) – “Low-Energy CEMents for sustainable concrete” partially funded by the Danish Strategic Research Council was launched in a joint collaboration between iNANO (Aarhus University, Denmark), Section of Chemistry (Aalborg University, Denmark), Norwegian University of Science and Technology (NTNU, Trondheim, Norway), Swiss Federal Laboratories for Materials Science and Technology (EMPA, Zürich, Switzerland), Cementir Aalborg Portland A/S (Denmark), and FLSmidth A/S (Denmark). This Ph.D. thesis is based on the outcome from the Work Package 3 of the LowE-CEM project, focusing on the durability of blended cements where Portland cement clinker is partially replaced by calcined clays and limestone. The mix proportions investigated in this Ph.D. thesis are selected based on the investigation in Work Package 7 (Dai, 2015) of the “SCM” project (2010 – 2015) – “Manufacturing of highly reactive Supplementary Cementitious Materials for low-CO2 cement”. The “SCM” project characterized blends of pure white Portland cement with 35 wt.% (i) metakaolin (MK), (ii) limestone (LS), (iii) metakaolin and limestone, (iv) metakaolin and silica fume (SF), and (v) metakaolin, silica fume and limestone (Dai, 2015). The combined utilization of metakaolin and silica fume was used to mimic the composition for a 2:1 clay mineral (smectite/montmorillonite), whereas metakaolin represents a 1:1 clay. For each type of clays, a series of proportions have been designed by varying the MK/(MK+LS) ratio and the (MK+SF)/(MK+SF+LS) ratio. The compressive strengths of the mortars, characterization of the phase assemblages and reaction kinetics have been studied. It was found that the compressive strength of the blends with MK/(MK+LS) = 0.75 and (MK+SF)/(MK+SF+LS) = 0.75 give the best performance (Dai, 2015). This Ph.D. thesis evaluates the durability performance of the 0.75 blends, considering their interactions with chloride, carbonate and sulfate ions. The 0 and 0.94 blends were also investigated for comparison. The pure hydrated white Portland cement blend was

4


tested as reference. This forms the first part of the thesis where the samples were prepared with the same water/binder ratio of 0.5. The second part investigates the impact of compressive strength on the durability performance of mortars, where the mortars were designed to have comparable compressive strengths by varying the water/binder ratios and binder/sand ratios. The SCMs investigated in the second part the thesis are calcined montmorillonite (characterization of its reactivity is reported in (Garg, 2015) associated with Work Package 1 of the “SCM” project) and a laboratorysynthesized glass (by Aalborg University associated with Work Package 2 of the LowECEM project). The mix proportions are the same as investigated in the first part of the thesis.

1.3 Objectives This Ph.D. thesis is based on the investigations of the durability performance of the newly developed Portland cement – calcined clay – limestone blends from the “SCM” project (and glass blends from the LowE-CEM project). The objectives of the current study are: 

to investigate the durability performance of the aforementioned blends (developed in Work Package 7 (Dai, 2015) of the “SCM” project) with the best compressive strengths prior to an industrial application.



to document the durability performance of the blends with calcined montmorillonite (from the Work Package 1 (Garg, 2015) of the “SCM” project) and the laboratorysynthesized glass (from the Work Package 2 of the LowE-CEM project).



to improve our understanding of the durability mechanisms for Portland cement – SCMs blends, particularly on the chloride binding and carbonation.



to investigate relationship between compressive strength and durability performance.

5


6

Chapter 2: Durability Studies of Portland Cement-Based Materials

CHAPTER 2: DURABILITY STUDIES OF PORTLAND CEMENT-BASED MATERIALS Partial replacement of Portland cement with increasing amounts of supplementary cementitious materials (SCMs) is rapidly being developed to meet various requirements, which brings more pressure on ensuring durable concrete. Development of sustainable concrete requires knowledge concerning the hydration products of the newly developed binder materials, the microstructures of the final products and interactions of the resulting concrete with aggressive environments. For this reason, a literature-based investigation has been carried out prior to the experimental studies, i.e., chloride ingress, carbonation and sulfate attack. This review tends to narrow down the scope to an extent highly linked to the results obtained from the current studies for thorough discussion.

7


2.1 Hydration 2.1.1 Portland cement and SCMs Portland cement, used as a basic ingredient of mortar and concrete, is produced by heating a mixture of limestone and clays to a temperature of ~ 1450 oC. The resulting nodular product, the so-called cement clinker, is mainly composed of four different phases, i.e., 50 ~ 70 wt.% of tricalcium silicate (Ca3SiO4), 15 ~ 30 wt.% of dicalcium silicate (Ca2SiO4), 5 ~ 10 wt.% of tricalcium aluminate (Ca3Al2O6) and 5 ~ 15 wt.% ferrite (Ca4(AlxFe1-x)4O10). These phases are named and abbreviated in cement chemistry to alite (C3S), belite (C2S), aluminate (C3A) and ferrite (C4AF). Portland cement is made by mixing and grinding cement clinker with a small fraction of gypsum (CaSO4·2H2O) in order to control the rate of setting during hydration. Nowadays, supplementary cementitious materials (SCMs), such as fly ash, slag and silica fume, are widely employed for partial substitution of Portland cement to make concrete. The use of these industrial by-products has significant benefits to the waste management and reduction of energy consumption, since they are reactive and no additional clinkering process is involved for partially replacing Portland cement. Moreover, incorporation of these materials in Portland cement concrete may also improve the workability during mixing and enhance the sustainability of the final products. However, strong global demands of SCMs for producing concrete cannot be met as a result of the limited availability and production of these industrial by-products. This reflects the recent boosted research interests on utilization of calcined clays as an additional source of SCMs, since clays are widely abundant in the Earth’s crust and their thermal activation temperature is significantly lower compared to the temperature used for cement clinker production. In this thesis, calcined kaolinite, denoted metakaolin (ideal Si/Al = 1, represents a 1:1 clay), and calcined montmorillonite (ideal Si/Al = 2, represents a 2:1 clay) are investigated. Silica fume (amorphous SiO2) is also employed in conjunction with metakaolin, to target a Si/Al ratio of 2 to mimic a 2:1 clay. Limestone is already being used as a source of CaO in production of cement clinker. It is also used as a secondary cementitious material by substituting up to 5 wt.% of cement clinker. Grinding of cement clinker together with a fraction of limestone produces a Portland – limestone cement, which has an improved particle-size distribution, water demand and workability (Tsivilis et al., 2000, Hawkins et al., 2003, Chowaniec, 2012). 8


Furthermore, the small particles of limestone also provide extra nucleation sites for the formation and growth of hydration products. In addition to these benefits, several published studies have confirmed that the limestone is reactive (Matschei et al., 2007, Lothenbach et al., 2008a) and a combination of limestone with other SCMs seems to give a synergetic effect. For example, a synergetic effect between metakaolin and limestone has been observed in Portland cement – metakaolin – limestone blends, as indicated by an increase in compressive strength for such systems (Steenberg et al., 2011, Antoni et al., 2012). The weight percentages of the principle chemical components of Portland cement, limestone and SCMs are projected in a ternary diagram as shown in Figure 2.1 (Lothenbach et al., 2011). A generally lower amount of CaO is found for the SCMs (except for the limestone) as compared to the Portland cement.

Figure 2.1. CaO-Al2O3-SiO2 ternary diagram of cementitious materials (in wt.%) (Lothenbach et al., 2011).

2.1.2 Hydration products When Portland cement-based materials are in contact with water, a series of chemical reactions associated with the hydration immediately take place. It is this process that gives an adhesive property to cementitious paste and enables it to bind aggregates together to make concrete. Although nowadays an overwhelming majority of Portland cement is blended with SCMs, the early-age compressive strength is still mainly attributed to hydration of Portland cement clinkers. Later on the hydration process may also be modified by the filler effect of SCMs and their own reactions (Scrivener et al., 2015). In order to eliminate the impact of the hydration process on the durability performance of Portland cement-based materials, the pastes and mortars investigated in this thesis are cured in demineralized water for three months before investigation. Thus, the hydration process is not discussed and only the final hydration products relevant to 9


the studied materials are presented. The hydration products of Portland cement-based materials differ in type and quantities as a result of variations in the chemical compositions of the Portland cement – SCMs blends. The possible hydrate phases for these blends are projected into the CaO-Al2O3-SiO2 system as shown in Figure 2.2 (Lothenbach et al., 2011).

Figure 2.2. Hydrate phases from Portland cement – SCM blends in H2O-CaO-Al2O3SiO2 system (in wt.%) (Lothenbach et al., 2011).

 C-(A)-S-H and CH The most important hydrate in Portland cement, i.e., a family of calcium silicate hydrates (C-S-H), is formed as a result of the hydration of C3S and C2S. The CS-H phase family has a similar structure but widely different Ca/Si ratios and amounts of bound water. A minor uptake of Al by C-S-H may also occur, resulting in a C-A-S-H phase. The C-(A)-S-H phase is generally accepted to contribute most to the early-age compressive strength of concrete. Formation of the C-S-H phase in Portland cement is accompanied by the formation of another hydration product, i.e., portlandite (CH). The conceptual chemical reactions can be expressed as: 3CaO·SiO2 + zH2O → xCaO·Si(OH)y·(z+x-y/2-3)H2O + (3-x)Ca(OH)2

(2.1)

2CaO·SiO2 + zH2O → xCaO·Si(OH)y·(z+x–y/2-2)H2O + (2-x)Ca(OH)2

(2.2)

 Ettringite (AFt) —

Ettringite (C6A·S3·H32), representing an important phase among the AFt family (“Al2O3-Fe2O3-tri”), is formed by hydration of C3A in the presence of sufficient amounts of gypsum. A structurally similar phase including a small amount of

10


iron is formed by hydration of C4AF under similar conditions. The formation of ettringite can be expressed as: 2− 2+ 2Al(OH)− + 4𝑂𝐻 − + 26H2O → C6A·S3·H32 4 + 3SO4 + 6C𝑎 —

(2.3)

 AFm phases With an insufficient amount of gypsum in cement, monosulfate and its ironcontaining analogue representing the phases among the AFm family (“Al2O3Fe2O3-mono”) are formed through hydration of C3A and C4AF. The formation of monosulfate can be expressed as: 2− 2+ 2Al(OH)− + 4𝑂𝐻 − + 6H2O → C4A·S·H18 4 + SO4 + 4C𝑎 —

(2.4)

In general, AFm phases have the representative formula C4(A·F)X2·Hy, where X refer to a single-charged anion (e.g. OH- or Cl-) or a half double charged anion 2− (e.g. SO2− 4 or CO3 ). The AFm phases have a layered structure in which the

positively charged main layer contains [Ca2Al(OH)6]+ units and the negatively charged interlayers host the X- anions. One of the important AFm phases is Friedel’s salt, which is not a direct hydration product, but is well known to bind chloride. More discussion on this phase is given in section 2.2 in conjunction with transformation of other AFm phases, which may affect the chloride binding. When SCMs are introduced in Portland cement, the chemical composition of the above mentioned hydrates varies, e.g. C-S-H is formed with a lower Ca/Si ratio as compared to C-S-H formed in pure hydrated cement. In addition, the uptake of Al in the C-S-H is also increased as a result of the additional Al source from the SCMs. These changes take place with the reaction between SCMs and portlandite, the so-called pozzolanic reaction. In addition, with limited amounts of gypsum, ground together with the cement clinker, supply of additional Al from the SCMs tends to form monosulfate instead of —

ettringite as a result of an insufficient amount of gypsum. Hemicarbonate (C4AC0.5H12) —

is formed by adding smaller amounts of limestone, and later monocarbonate (C4ACH11) is formed by increasing the amount of limestone, followed by a depletion of portlandite. At the same time, ettringite is also stabilized with the formation of monocarbonate. When the amount of the Si source is in excess, e.g. by adding silica fume, it may lead to a decomposition of monocarbonate to release Al and form strätlingite (C2ASH8).

11


2.2 Chloride in hydrated cements Deterioration of reinforced concrete structures caused by the ingress of chloride ions is a common durability issue, which has to be considered in service life predictions for reinforced concrete structures exposed to sea water and de-icing salts. When the free chloride ions penetrate into the concrete cover and reach a critical level, a depassivation of the rebar occurs, followed by accumulation of the reaction products, and finally a degradation of the concrete. In chloride contaminated concrete, most chloride ions are free to transport in the pore solution, the rest are known to be chemically bound in Friedel’s salt or physically bound in the C-S-H phase. Moreover, it is generally accepted that only free chloride is harmful to reinforced concrete. Thereby, reducing the free chloride in the pore solution by increasing the chloride binding capacity and refinement of pore structures may be beneficial for improving chloride resistance.

2.2.1 Evaluation of the chloride resistance The resistance of chloride ingress of a given concrete is usually assessed by analyzing the total chloride profiles. They are obtained by applying a grinding process on cylinder samples where layers of increasing thickness are gradually removed from the exposed surface. The ground powder is collected for chloride analysis. The results are plotted as a function of sample depths, from which apparent chloride diffusion coefficients are obtained. However, the total chloride profiles do not differentiate between the bound chloride and free chloride ions, not even the physically bound and chemically bound chloride. Some studies (Yuan et al., 2009, Baroghel-Bouny et al., 2012, Bentz et al., 2013) reported that the interaction of chloride ions with cement hydrates may retard the chloride transport and change the shape of the chloride profiles. It is thus very important to understand the role of different cement hydrates on chloride binding.

2.2.2 Chloride binding in AFm phases It is accepted that the main contribution to chemical binding of chloride results from the formation of Friedel’s salt (Ca4Al2(OH)12Cl2·4H2O). Suryavanshi et al., proposed two mechanisms for the formation of Friedel’s salt: adsorption mechanism and anionexchange mechanism (Suryavanshi et al., 1996). In the adsorption mechanism bulk chloride ions present in the pore solution are absorbed into the interlayer of the principle layers [Ca2Al(OH)6]+ structure to balance the charge. In the case of NaCl addition, the equivalent amount of Na+ ions is removed from the pore solution to maintain the ionic 12


charge neutrality, and bound into the lattice of the C-S-H phase to balance the charge arising from replacement of Si4+ by Al3+ or Fe3+. In the anion-exchange mechanism, OH- ions in the interlayer of the principal layer are exchanged by chloride ions from the pore solution. Different from the adsorption mechanism, anion-exchange causes an increase of pH of the pore solution as a result of release of OH- ions. However, the adsorption mechanism seems to apply only when chloride salt is intermixed in the mixing procedure for sample preparation. Considering the penetration of external chloride, the AFm phase has already formed in hydrated cement, hence formation of Friedel’s salt is mostly likely ascribed to anion exchange. However, the anion exchange occurs not solely between OH- and chloride ions. Sulfate and carbonate ions may also participate in the anion exchange. In general, the stability equilibria of AFm phases in hydrated cements strongly depend on the surrounding environment such as the presence of an excess of one or another anion in the pore solution, the amount of water, pH, temperature, etc. Thus, a holistic picture describing the transformation of the AFm phases has to be considered when evaluating the chloride binding though formation of Friedel’s salt. As the current investigation on chloride ingress and chloride binding are conducted at 20 oC, the impact of temperature on the stability of AFm phases is not discussed. (Glasser et al., 1999) based on published studies established a diagram depicting the stability and characterization of AFm phases as shown in Figure 2.3, in which four types of anions, i.e. OH-, Cl-, 𝐶𝑂32− , and 𝑆𝑂42− are considered. Except for the presence of the distinct compounds, there are some solid solutions that may also be present which are not included in this diagram, but discussed by the authors in the text. Recently, (Galan and Glasser, 2015) proposed the buffering concept for hydrated cement systems, e.g. pH buffering action in hydrated cements depends on the presence and accessibility of portlandite rather than its amount. This concept was applied in a similar manner to AFm stability, particular for explaining critical concentrations for the formation of Friedel’s salt. The authors stated that “the action of chloride on AFm is not just a matter of considering the action of chloride on a single AFm phase, instead, its reaction with one of several possible phase assemblages has to be taken into account”. For instance, Kuzel’s salt (Ca4Al2(Cl)(SO4)0.5(OH)12·6H2O) may also form at lower chloride concentrations and at higher sulfate activities (Glasser et al., 1999). However, Portland cements usually contain insufficient amounts of sulfate to stabilize Kuzel’s salt, hence Friedel’s salt is favorably formed. 13


(Elakneswaran et al., 2009) also stated that physical adsorption of chloride onto Friedel’s salt occurs through adsorption of chloride ions onto the positive charged surface of [Ca2Al(OH-)]+ arising from dissolution of Friedel’s salt in water. Besides to the formation of Friedel’s salt as a result of the interaction between chloride ions and AFm phases, it is also reported that a chloro-complex similar to Friedel’s salt can be formed, but with iron instead of aluminum (3CaO·Fe2O3·CaCl2·10H2O). More details can be found in the review article by (Galan and Glasser, 2015) and (Yuan et al., 2009).

Figure 2.3. AFm compositions showing the principle anion types encountered in cement. Joins of interest are shown by straight lines (Glasser et al., 1999). Smaller numbers indicate references prior to the author’s work and are listed in the same paper.

2.2.3 Chloride binding in the C-S-H phase 

Experimental evidence of chloride binding in the C-S-H phase

Evidence that C-S-H can bind chloride ions has been documented in many published studies, and the role of C-S-H on chloride binding is normally considered secondary compared to the formation of Friedel’s salt (Zibara et al., 2008). (Ramachandran, 1971) studied the hydration of C3S in the presence of different concentrations of CaCl2 solutions. The author concluded that calcium chloride may exist in the states of (i) free chloride (soluble in water and alcohol), (ii) a chemisorbed chloride layer on the surface of C-S-H (not removable by alcohol leaching), (iii) chloride in the interlayer space of C-S-H (not removable by alcohol leaching, whereas removable by water leaching), and (iv) intimately bound chloride in the lattice of C-S-H (not removed by leaching with water).

14


(Lambert et al., 1985) studied the chloride binding in hydrated C3S when sodium chloride is present in the mixing water. The results showed no detectable capacity of the hydration products to bind chloride. (Beaudoin et al., 1990) studied the interaction of chloride and synthetic C-S-H phases with a wide range of Ca/Si and H2O/Si ratios. They found two types of bound chloride in C-S-H, i.e. “alcohol-insoluble” and tightly held. In fact, the “alcohol-insoluble” refers to the difference in bound chloride between ethanol washed C-S-H and C-S-H washed first in ethanol and then in water, which is in principle equivalent to the summation of chemisorbed chloride and that in the interlayer space of the C-S-H as described by (Ramachandran, 1971). The chloride remaining after washing with water, as referred to as “tightly held” chloride, is equivalent to the lattice chloride described by (Ramachandran, 1971). The results from (Beaudoin et al., 1990) show a linear dependence between alcohol-insoluble chloride and three variables, i.e. Ca/Si ratio, H2O/Si ratio and surface area, where alcohol-insoluble chloride increases with an increase of the Ca/Si ratio and H2O/Si ratio (note that H2O/Si generally increases with increasing Ca/Si ratio) and decreases with increasing surface area (large surface area in C-S-H occurs at low Ca/Si ratio near Ca/Si = 1 (Chen et al., 2004). Correlations of the tightly held chloride with the three variables are generally poor. However, the tightly held chloride tends to decrease with increasing Ca/Si ratio and H2O/Si ratio. The reason for this is not clear. However, the contribution of lattice chloride seems to be small, since the tightly held chloride was found to be an order of magnitude lower than alcohol insoluble chloride. Moreover, according to the authors (Beaudoin et al., 1990), the degree of layering of the C-S-H decreases with a decrease of Ca/Si ratio (or increase of surface area), thus more available chlorides are removable from the interlayer space for C-S-H phases with low Ca/Si, i.e. less chloride binding. (Luping and Nilsson, 1993) studied chloride binding in both mortars and pastes exposed to NaCl solution saturated with Ca(OH)2, and their results showed that the chloride binding capacity of concrete strongly depends on the content of C-S-H phase in the samples. However, it should be considered that the saturated Ca(OH)2 solution they used may contribute greatly to the chloride binding in C-S-H in their samples, thus such high levels of chloride binding may not be observed when the same samples are exposed to a NaCl solution only.

15


(Wowra et al., 1997) studied sorption of chlorides on hydrated cements and C3S pastes. They found that the chloride adsorption capacity is mainly determined by the specific surface area, the pH value and the associated cations. Chloride adsorption is obviously enhanced by adsorption of calcium ions. The authors attributed this to the reason that adsorption of Ca2+ ions increased the positive charge on the surface of the used materials and resulted in further adsorption of chloride. However, sorption of chloride was found to increase with increasing specific surface area of the hydrated C3S, which contradicts the previous study (Beaudoin et al., 1990). More interestingly, a gradual decline of the pH for the C3S paste suspensions was observed with increasing CaCl2 concentration as a result of sorption of Ca2+ ions. Moreover, they also found that chloride sorption was lower in the presence of 𝑆𝑂42− ions. In their study, they have also compared the chloride binding between hydrated cements (CEM I 32.5 R with 4.9 wt.% Al2O3, CEM III 32.5 NW/NA with 7.8 wt.% Al2O3) and hydrated C3S paste, and surprisingly, a little effect of the aluminate content on the chloride binding was observed. (Zibara et al., 2008) studied silica fume – lime mixtures and found that the chloride binding capacity of the C-S-H was significantly higher with a higher Ca/Si ratio of the C-S-H. When two parts of silica fume was mixed with one part of lime, the resulting CS-H (Ca/Si assumed to be higher than 0.42 but lower than 1.24) was found to have no capacity to binding chloride. (Plusquellec et al., 2012) investigated the impact of Ca/Si ratio of the C-S-H, the amount and types of salts (CaCl2 and KCl) on the interactions between Cl- ions and the C-S-H phase. They found that the uptake of Cl- by C-S-H only occurs in the presence of sufficient amounts of Ca2+ ions in the equilibrium solution, and adsorption of two chloride ions is accompanied by the uptake of one calcium cation. They also explained that the uptake of chloride ions occurs through the overcompensation of the negative surface charge of the C-S-H by calcium cations. 

Mechanisms for chloride binding in C-S-H

As a result of the combination of ionization, dissolution and ionic adsorption, the C-S-H phase develops an electrical charge on its surface when in contact with electrolyte solutions according to the following reactions (Elakneswaran et al., 2010). For high Ca/Si C-S-H, a solution in equilibrium with the C-S-H has a high pH as a result of

16


calcium leaching, hence the surface of the C-S-H reacts with OH- ions to form a negatively charged surface according to the reaction: ≡SiOH + OH- ↔ ≡SiO- + H2O

(2.5)

The negatively charged surface is soon overcompensated by Ca2+ present in the pore solution to form a positively charged inner-sphere (≡SiOCa+) according to the reaction: ≡ SiO- + Ca2+ ↔ ≡SiOCa+

(2.6)

which is equivalent to the combined reactions: ≡SiOH + Ca2+ ↔ ≡SiOCa+ + H+

(2.7)

H+ + OH- ↔ H2O

(2.8)

This phenomenon has been documented in many published studies as quoted by (Elakneswaran et al., 2010) and (Plusquellec et al., 2012). The surface charge cannot be directly measured but it is closely related to the surface potential, which can be indirectly obtained by the measured zeta potential (ζ). Figure 2.4 shows the evolution of the measured zeta potential and the calculated surface potential for a C-S-H phase in a calcium hydroxide solution. The results show that the surface charge becomes more and more positive with increasing pH and Ca2+ ion concentration of the solution. It should be mentioned that, a high pH is here equivalent to a high Ca2+ concentration, as the pH is maintained by Ca2+ ions by reaction: high-Ca C-S-H ↔ low-Ca C-S-H + Ca2+ + 2OH-

(2.9)

Figure 2.4. Measured zeta potential and calculated surface potential for a C-S-H phase as a function of pH (left) and calcium concentration (right) in calcium hydroxide solutions (Elakneswaran et al., 2009, Elakneswaran et al., 2010).

17


Similar to the above mentioned reaction, when chloride ions are present in the solution, a part of the chloride can be absorbed onto the surface of the C-S-H via the inner-sphere complexation reaction as presented by (Elakneswaran et al., 2010): ≡ SiOH + Cl- ↔ ≡SiOHCl-

(2.10)

However, it is difficult to understand why this reaction occurs because the silanol (≡ SiOH) is neutralized. Thereby, the so-called inner-sphere complexation reaction seems not to be convincible. (Plusquellec et al., 2012), and several other studies quoted in their work, proposed that the adsorption of chloride occurs due to the overcompensation of the negatively charged surface of C-S-H by Ca2+ ions, The observation that the uptake of one chloride ion by C-S-H is accompanied by adsorption of two Ca2+ ions confirmed this hypothesis (Plusquellec et al., 2012).

2.2.4 Chloride binding in the other phases (Hirao et al., 2005) reported that AFt and Ca(OH)2 had no capacity to bind chloride, but AFt can possible dissolve at high chloride concentration and form Friedel’s salt. On the contrary, (Elakneswaran et al., 2009) reported that chloride can be absorbed on the positive surface of dissociated portlandite [CaOH]+ to form CaOHCl crystals. (Galan and Glasser, 2015) reviewed that portlandite remains stable in contact with NaCl over a wide range of chloride concentrations, and that the pH of the saturated portlandite solution remains unaffected. However, there is a substantial increase in solubility compared to the solubility of portlandite in pure water when the chloride concentration increases up to 0.5 M, which is representative of the chloride concentrations in sea water. At much higher chloride concentrations, which may relevant to areas where sea water is evaporated, Friedel’s salt can also be destabilized to form chloride complex compounds. As this is not closely related to the current studies, details are not presented here, which can be found in (Galan and Glasser, 2015, Galan et al., 2015b).

2.2.5 Impact of cement type The cement type has a significant impact on chloride binding as reviewed by (Justnes, 1998) and (Yuan et al., 2009). The most dominant impact of the cement type is related to the quantity of the C3A phase, where increased chloride binding with increasing amounts of C3A has been thoroughly established. 18


2.2.6 Impact of cations (Tritthart, 1989) reported a higher chloride binding of cement pastes exposed to CaCl2 solutions compared to NaCl solutions. They explained that the higher chloride binding was attributed to a lower OH- concentration of the pore solution, caused by precipitation of Ca(OH)2 equivalent to the amount of added chloride salt. (Suryavanshi et al., 1996) reported that the amount of total bound chloride is significantly higher for CaCl2 solutions than for NaCl solutions. The authors suggested that the higher chloride binding with equivalent CaCl2 addition is caused by the lower degree of competition offered by the OH- ions during the bulk free chloride adsorption into the interlayers of the principal layers of the AFm phase. (De Weerdt et al., 2015) also reported an increased chloride binding by exposing samples to CaCl2 solutions compared to NaCl solutions. The authors attributed the difference in chloride binding to an increased chloride binding in the C-S-H, which essentially is governed by the pH of the exposure solution.

2.2.7 Impact of pH 

Impact of pH on the stability of Friedel’s salt

(Page and Vennesland, 1983) studied chloride binding of silica fume cement pastes and found that increasing percentage of silica fume causes a decrease in alkalinity of the pore solution and a reduction in the chloride binding. They attributed this to an increased solubility of Friedel’s salt with decreasing pH, in accordance with the equilibrium: 3CaO·Al2O3·CaCl2·10H2O ↔ 4Ca2+ + 2Al𝑂2− + 4OH- + 2Cl- + 8H2O

(2.11)

(Suryavanshi and Swamy, 1996) studied the stability of Friedel's salt in carbonated concrete and found that the solubility of Friedel’s salt increase with the degree of carbonation of concrete. The authors concluded that the stability of Friedel's salt is pH dependent, according to the following reactions, because the alkalinity of the pore solution drops as a result of carbonation. 3CaO·Al2O3·CaCl2·10H2O ↔ 4Ca2+ + 2Cl- + 4OH- + 2Al(OH)3 +5H2O

(2.12)

Al(OH)3 + 4OH- ↔ Al(𝑂𝐻)− 4

(2.13)

(Zhu et al., 2012) studied different types of chloride salts on chloride binding and reported that Na+ and K+ lead to a rise in pH and increase of the solubility of Friedel’s 19


salt, which contradicts the earlier conclusions. (Yuan et al., 2009) quoted the work from (Roberts, 1962) that an increase of the pH for the chloride solution increases the solubility of Friedel’s salt. It is known that sulfate ions have a negative effect on chloride binding, but it is also quoted that sulfate attack can increase the alkalinity of the pores solution (Geng et al., 2015). This clearly indicates that pH is not a decisive parameter for controlling chloride binding. 

Impact of pH on chloride binding in C-S-H

(Wowra et al., 1997) described that the decrease of pH in the presence of CaCl2 compared to a NaCl solution could be a result of an ion-exchange reaction on the C-S-H surface as described by: SiOH + Ca2+ ↔ SiOCa+ + H+

(2.14)

The surface charge of the C-S-H becomes increasingly negative with increasing pH (which is related to the increase of the Ca/Si) due to the ionization of the silanol groups. The negative surface leads to sorption of the Ca2+ ions, and at the same time an overcompensation of the negative surface occurs by sorption of Ca2+ ions (Labbez et al., 2006), which justify that it is possible for anions to be adsorbed by the C-S-H. 

Impact of pH on the total amount of bound chloride

(Lambert et al., 1985) argued that if a significant amount of chloride ions are removed from the pore solution of hydrated C3S pastes, the charge neutrality will necessitate the dissolution of other ions (mainly hydroxyl ions) rather than removal of an equivalent quantity of cations and hence increased the alkalinity of the pore solution, i.e. pH. (Tritthart, 1989) studied chloride binding of cement pastes, which were exposed to solutions with different pH (saturated Ca(OH)2 solution with pH of 12.5; 0.1 M NaOH solution with pH of 13.0; 0.5 M NaOH solution with pH of 13.7) prior to the chloride exposure. They reported that chloride binding increased with decreasing OHconcentration of the pore solution, through competition between Cl- and OH-. 

Summary of the impact of pH

According to the published studies discussed above, the impact of decreasing pH on chloride binding seems to be twofold: (i) it increases the solubility of Friedel’s salt, (ii) it increases the chloride binding in the C-S-H. However, the mechanism on how pH affects the chloride binding especially in the C-S-H is not well explained. In fact, pH is just a scale to indicate the alkalinity of the pore solution, which can be highly 20


influenced by different types of cations. Discussions on the impact of pH only on chloride binding, without taking into account the cations, is obviously not sufficient.

2.2.8 Impact of SCMs The impact of SCMs on chloride binding is believed to be related to the aluminate content as indicated by many studies reviewed by (Justnes, 1998, Yuan et al., 2009). Nilsson et al. (quoted in (Zibara et al., 2008)) suggested that silica fume will affect the chloride binding capacity in three ways: (i) increase of chloride binding through reduction of pH, (ii) reduction of chloride binding due to dilution of C3A, and (iii) increase of chloride binding due to an increase in the amount of the C-S-H. However, the reduction of chloride binding due to dilution of C3A is not fully convincible since a C-S-H phase with a low Ca/Si ratio formed in the presence of silica fume can also decrease the chloride binding of the C-S-H compared to the C-S-H formed in pure hydrated Portland cement. In order to eliminate the effect of the reduction of Ca/Si for the C-S-H formed in the presence of SCMs, mixtures of silica fume and metakaolin with lime rather than Portland cement were studied by (Zibara et al., 2008). These authors concluded that when alumina is present, the chloride binding is dominated by the formation of Friedel’s salt. However, the Ca/Al ratio varies the chloride binding capacity as a result of the formation of different hydrates. At higher Ca/Al ratio, the formation of monocarboaluminate is favored, whereas formation of stratlingite is favored when Ca/Si is lower. Both phases can be transformed into Friedel’s salt. However, the transformation of monocarbolauminate to Friedel’s salt occurs faster even at low chloride concentrations (less than 0.1 M), while transformation of stratlingite occurs much slower and at much higher chloride concentration. In the absence of alumina, the chloride binding in the C-S-H phase was obtained, which was found to decrease with decreasing of Ca/Si ratio as discussed earlier.

2.2.9 Improvement of the chloride resistance 

Via chloride binding

Since only free chloride in the pore solution is responsible for corrosion of the steel reinforcement, removal of chloride from the pore solution via chloride binding is known to be beneficial for increasing chloride resistance (Yuan et al., 2009, Baroghel-Bouny et al., 2012, Bentz et al., 2013). (Martın-Pérez et al., 2000) reported that chloride binding has an important effect on the rate of chloride ionic transport in concrete. (Birnin-Yauri 21


and Glasser, 1998) mentioned that chloride profiles may be affected by ion exchange and binding into the AFm phases. However, (Glass and Buenfeld, 2000) found that chloride binding reduces the free chloride content within the concrete, and that it may increase or decrease the total chloride content, depending on the distance from the concrete surface. (Yuan et al., 2009) mentioned in the introduction of their review that the formation of Friedel’s salt can result in a less porous structure and slow down the transport of chloride ions. However, the argument behind this is not clear, since formation of Friedel’s salt will not change the basic structure of the AFm phase via an ion-exchange mechanism. 

Via microstructure refinement

Refinement of the microstructure through incorporation of most of the SCMs has been widely accepted. (Luo et al., 2003) reported that ground granulated blast furnace slag can improve the pore structure of hydrated Portland cement and decrease the chloride diffusion coefficient greatly. (Loser et al., 2010) reported that the decisive parameter for chloride resistance is the permeability while the influence of chloride binding is less important. On contrary to the widely accepted conclusions, some studies as reviewed by (Elakneswaran et al., 2010) in the introduction of their work, suggest that the total pore volume does not make any relationship with the measured apparent diffusion coefficient of chloride in mature cement paste. (Elakneswaran et al., 2010) mentioned that chloride penetrates into concrete accompanied by other ions which directly influence the chloride transport. When the pore size is very fine, the ionic interaction may play a significant role on the chloride diffusion.

22


2.3 Carbonation Carbonation of concrete is a widely considered durability issue for reinforced concrete structures. Portland cement concrete maintains its pore solution at an intrinsically high pH, such that the alkaline environment has the capability to keep the reinforcements being passivated. However, when CO2 from the atmosphere penetrates into hydrated cement paste, it dissolves in the interstitial solution and thereby modifies the chemical balance between the solution and the hydrates. This leads to consumption of the calcium-bearing phases, e.g. portlandite (CH) and C-S-H, and followed by precipitation of calcium carbonates and a reduction of pH of the pore solution. When the pH is reduced to a certain level, the protective layer on the steel bars is destroyed, resulting in corrosion of the reinforcement and thereby damage of concrete. In this part, carbonation of hydrated Portland cement-based materials is reviewed. Carbonation of fresh concrete or clinker phases related to CO2 curing is not covered.

2.3.1 Carbonation mechanism Carbonation of Portland cement is commonly described by the following two simple reactions corresponding to carbonation of CH and C-S-H. In fact, it is not the gaseous CO2 that reacts directly with the Ca bearing hydrates. When CO2 penetrates into the cement matrix, it firstly dissolves in the pore solution to produce 𝐻𝐶𝑂3− and 𝐶𝑂32− ions, which react with Ca2+ ions leached from CH and C-S-H to the pore solution. —

—

CH + C  CC + H —

(2.15) —

CxSyHz + xC  xCC + ySHt + (z-yt)H

(2.16)

Because CH is thermodynamically less stable than C-S-H, CH will be firstly carbonated, followed by carbonation of the C-S-H. Sometimes in reality, even for maturely carbonated samples, small amounts of CH can still be detected. This is attributed to the formation of a CaCO3 shell around the non-carbonated CH particles (Groves et al., 1990, Groves et al., 1991, Morandeau et al., 2014). Regarding the carbonation of C-S-H, it was found that when the CO2 concentration is below 3%, a complete carbonation of the C-S-H seems unlikely to occur. Instead, a C-S-H with lower Ca/Si ratio is formed from decalcification of high-Ca C-S-H (Castellote et al., 2009). In addition, other Ca bearing phases may also be carbonated, since they can also buffer the Ca concentration in pore solution. It has also been established that carbonation of ettringite produces CaCO3, gypsum and an alumina gel (Nishikawa et al., 1992, Zhou and Glasser, 2000) . 23


However, in hydrated Portland cement, carbonation of these phases are less important since they are either present in smaller amounts (e.g. the AFm phases) or thermodynamically relatively stable (e.g. AFt phases). A major carbonation product is known as CaCO3 both for carbonation of CH and C-S-H. In addition, carbonation of CH also produces water and a complete carbonation of the C-S-H also gives an amorphous silica phase. The water initially bound in CH released during carbonation was experimentally highlighted by (Pihlajavaara, 1968). It is worth to mention that CaCO3 has several polymorphs, such as calcite, vaterite, aragonite and even amorphous CaCO3. These polymorphs can be distinguished by XRD diffractions and they are also reflected to some extent on DTG curves as presented in several studies (Šauman, 1971, Goto et al., 1995, Thiery et al., 2007, Villain et al., 2007, Borges et al., 2010, Morandeau et al., 2014).

Figure 2.5. DTG data obtained from samples at different depths of a cement paste after accelerated carbonation. Zoom in of the temperature range 450 – 1000 oC (Thiery et al., 2007).

Figure 2.6. DTG data and results of mass spectroscopy (H2O and CO2) determined in a 2 mm-deep sample (T1) of the cement paste studied in Figure 2.5 (Thiery et al., 2007)

(Thiery et al., 2007) proposed three modes regarding to the CaCO3 decomposition observed in DTG curves as shown in Figure 2.5. Mode I (780 °C < T < 990 °C) would be due to well crystallized CaCO3, namely calcite. Mode II (680 °C < T < 780 °C) is thermally less stable and can be related to metastable CaCO3 (vaterite and aragonite). Mode III (550 °C < T < 680 °C) is more difficult to explain, but it is definitely linked to an emission of CO2, instead of H2O, which has been confirmed by (Thiery et al., 2007). The anhydrous crystalline polymorphs calcite, vaterite and aragonite related to Mode I and Mode II, are known from the experiments (Plumber and Busenberg, 1982, Usdowski, 1982, Ogino et al., 1987, Brečević and Nielsen, 1989). In addition, amorphous calcium has been characterized previously (Brečević and Nielsen, 1989), 24


and it can be associated with mode III. The formation of amorphous silica is not clear, especially that the amount of remaining bound water in silica gel has not been determined yet.

2.3.2 Carbonation front and profiles It has been documented that carbonation of cement hydrates occurred beyond the phenolphthalein measured carbonation depth (Rahman and Glasser, 1989, Houst and Wittmann, 2002, Thiery et al., 2003, Villain and Platret, 2006, Thiery et al., 2007, Villain et al., 2007). These observations suggest that the carbonation front may not be sharp as suggested in earlier studies. (Papadakis et al., 1991) proposed that carbonation is a diffusion-controlled process, i.e., the chemical reactions are instantaneous, which should result in a sharp reaction front. However, this is not the truth, according to the earlier observations published by (Parrott and Killoh, 1989) in situ conditions and by (Rahman and Glasser, 1989) for accelerated conditions. A recent publication by (Thiery et al., 2007) explained the feature of a gradient carbonation profile by the fact that the kinetics of the chemical reactions become the rate-controlling processes rather than the diffusion of CO2. (Morandeau et al., 2014, Morandeau et al., 2015) also observed that the carbonation front is not sharp and support which the chemical reaction occurring during carbonation are not instantaneous with respect to CO2 diffusion through the porous network.

2.3.3 Change in pH The pH change induced by carbonation of concrete is ascribed to carbonation reactions of CH and the C-S-H phase. For pure phases, CH and C-S-H can maintain a pH of 12.5 and 13.0 respectively. Although the carbonation front can be analyzed by quantification of CH and CaCO3, the corresponding pH profiles are not known, since they are difficult to measure experimentally as a result of drying of the samples during carbonation. (McPolin et al., 2009) have measured the pH profiles after carbonation by pore solution expression and leaching methods. However, these methods can only provide an estimate of the real pH values because partially carbonated large crystals of portlandite are cracked during the measurements which may change the pH of the pore solution. In addition, the phenolphthalein method indicates a gradual change in color from colorless to fuchsia, reflecting pH changes from 8.2 to 10.0 in solution. This gradual color change has not been clearly documented for the carbonation front in concrete or mortars 25


measured by the phenolphthalein method. Thus, it is not clearly determined at which pH levels the phenolphthalein method reflects the carbonation depths in concrete.

2.3.4 Change in microstructure In addition to a change of the phase assemblages, carbonation of cement hydrates also cause a change of the molar volume of the solid, thereby changing the total porosity. Owing to the increase of solid volume from portandite to CaCO3, a decrease of the total porosity can be observed at lower degrees of carbonation. When the C-S-H phase is carbonated, an increase in total porosity can be detected as a result of a decrease of the solid volume. Another impact of carbonation on concrete is the changes in microstructure, originating from the differences in molar volumes of the hydrated and carbonated phases. Predictions become complex when the molar volume of the C-S-H phase changes with carbonation as a result of changes of the Ca/Si ratio (Morandeau et al., 2014, Morandeau et al., 2015, Sevelsted and Skibsted, 2015). Moreover, the different polymorphs of CaCO3 also exhibit different unit cell volumes (Smyth and McCormick, 1995). Several studies have reported a reduction of the porosity for pure Portland cement (Pihlajavaara, 1968, Houst and Wittmann, 1994, Ngala and Page, 1997, Thiery et al., 2003, Morandeau et al., 2014, Leemann et al., 2015), whereas an increase of the total porosity has been measured by different techniques for Portland cement – SCM blends (De Ceukelaire and Van Nieuwenburg, 1993, Leemann et al., 2015). (Morandeau et al., 2015) have reported that the total porosity decreases for a Portland cement paste with 60 vol.% of fly ash, however, its microstructure is rearranged and large capillary pores are created. The same work (Morandeau et al., 2015) also reviewed some earlier studies and stated that there is a shift in the porosity towards larger pore radii during carbonation as measured by mercury intrusion porosimetry. (Johannesson and Utgenannt, 2001) reported that the difference in pore-size distribution is more pronounced than the difference in specific surface area for carbonated and non-carbonated Portland cement mortars. In addition, the presence of some micro-cracks has also been noticed as a result of the volume increase of the solid during carbonation (Lange et al., 1996). Thus, the earlier studies show several discrepancies, uncertainties and lack of explanations for the interpretation of the microstructural changes caused by carbonation.

26


2.3.5 Change in compressive strength As a result of the change in phase assemblage and microstructures, mechanical properties may also change. When the carbonation degree is lower, the formation of CaCO3 will fill in the porosity, resulting in an increase of the compressive strength. When the carbonation degree increases over time, a significant decrease of the compressive strength can be observed due to increased total porosity.

2.3.6 Factors affecting carbonation resistance There are many factors influencing the carbonation resistance, e.g. CO2 concentration, RH, CaO content, portlandite content, microstructure, SCMs, etc. These factors together complicate the carbonation performance for a specific concrete. 

Impact of CO2 concentration

In many carbonation test methods, accelerated conditions are designed, one of which is increasing the CO2 concentration. Two critical considerations are: (i) whether the carbonation rates at higher CO2 concentration can be used to predict the carbonation rate at lower CO2 concentration? (ii) whether carbonation products will be different at different CO2 concentration? Both of these factors concern the transferability of the accelerated carbonation data to predict natural carbonation. (Leemann et al., 2015) compared the carbonation rates when concrete are exposed to a 1% and 4% CO2 environment and obtained a very satisfied linear correlation between the carbonation rates under these two environments. This indicated that 4% of CO2 can be used for accelerated carbonation tests with predictable carbonation rates at lower CO2 concentration. (Castellote et al., 2009) also conducted

29

Si NMR studies on the impact

of CO2 concentration on changes of the Si species and they found that when the CO2 concentration is not higher than 3%, the change of phases will be similar as for natural carbonation. They also observed that carbonation of C-S-H results in a decrease of the Ca/Si ratio to a low-Ca C-S-H, which cannot be further carbonated. However, at higher CO2 concentrations even low-Ca C-S-H is completely decomposed. 

Impact of RH

Relative humidity (RH) has a significant impact on THE carbonation rate by influencing the transport of CO2. Studies have found that when RH is about 57%, the carbonation of Portland cement is maximized. If the RH is very high, it will fill the capillary pores and reduce the diffusion rate of CO2. If the RH is too low, gaseous CO2 27


will not be able to produce sufficient 𝐻𝐶𝑂3− and 𝐶𝑂32− ions. (Leemann et al., 2015) studied the carbonation exposed to a natural environment and compared the carbonation rate for sheltered and non-sheltered conditions. A lower carbonation rate was observed for concrete exposed to a non-sheltered natural condition. 

Impact of pore sizes

Combination of another important factor with RH has a significant impact on CO2 diffusion, i.e. the pore sizes. At a specific RH, the finer the pore size is, the easier the capillary condensation occurs according to the Kelvin equation. For instance, when mortars are exposed to a 57% RH environment, a calculated Kelvin radius would be 4 nm. When the cement matrix has a larger fraction of pores below this pore size, CO2 diffusion will be very slow since pores with sizes below 4 nm will be filled by condensed water. 

Total CaO content

It is accepted that the total CaO has a significant impact on the carbonation rate, since higher total CaO has higher CO2 binding capacity. However, the carbonation rate is usually expressed by the slope of the carbonation depth as a function of the square root of time whereas the carbonation depth is usually measured by the phenolphthalein method, which is a pH indicator changing color when pH is above 8.2. Thereby, the total amount of CaO may not be a suitable parameter for correlations with carbonation rate, since non-carbonated phases can still be identified in apparently carbonated regions. 

Impact of SCMs

In modern concrete, SCMs are widely used for reducing the CO2 foot print and to improve concrete properties. However, the carbonation issue may become particularly important when SCMs are incorporated in cement blends. Several studies have reported that cement-based materials including SCMs exhibit poor carbonation resistance (Papadakis, 2000, Leemann et al., 2015, Morandeau et al., 2015). Hydrated Portland cement – SCM blends contain generally a smaller amount of CH as compared to hydrated pure Portland cements, since a part of the CH has been consumed by reaction with the SCMs forming C-S-H phase. This lower amount of CH may account for the faster carbonation process for Portland cement – SCM blends, compared to pure Portland cement, as mentioned in earlier studies (Papadakis, 2000, Leemann et al., 2015, Morandeau et al., 2015). However, it seems of minor importance for the 28


carbonation resistance under accelerated conditions, if the CaO buffering is present from CH or C-S-H (Leemann et al., 2015). The lack of CH cannot solely explain the variations in apparent carbonation depths, when the depths are measured by the phenolphthalein indicator method, since several of the hydrate phases are not completely carbonated within the carbonation depths, as revealed by thermogravimetric analysis (Chang and Chen, 2006, Thiery et al., 2007). Thus, it is necessary to compare the actual and potential CO2 binding of the different binders in order to better understand their carbonation performance. Recently, (Leemann et al., 2015) pointed out that a decisive material parameter for the carbonation resistance is the buffer capacity per volume of cement paste that can be expressed as the ratio between the mixing water and CaO reacting with CO2.

29


2.4 Sulfate attack Sulfates ions attack on concrete may occur when concrete structures are exposed to sulfate environments, such as sea water, ground water and natural soils. When sulfate ions penetrate into hardened cement paste of concrete, it will interact with the cement hydration products, which can form ettringite (3CaO·Al2O3·3CaSO4·32H2O) and gypsum (CaSO4·2H2O) (Tian and Cohen, 2000). In the presence of limestone and particularly at lower temperature, another type of sulfate attack may also occur as a result of formation of thaumasite (CaSiO3·CaCO3·CaSO4·15H2O). This type is commonly known as thaumasite sulfate attack (TSA). These reactions may result in expansion and softening of the hardened cement pastes, finally leading to cracking and loss of strength. In this part, sodium sulfate related sulfate attack is reviewed whereas other types like magnesium and potassium sulfate attack are not included.

2.4.1 Formation of ettringte In addition to being a major hydration product, ettringite is also the principle reaction product formed during sulfate attack though conversion of AFm phases as a result of ingress of external sulfate ions into hardened Portland cement. This process is believed to generate expansion, as indicated by the inverse correlation between the C3A content in Portland cement and the resistance of concrete to sulfate attack. For this reason, lowC3A cements are required in many sulfate environments. This leads to the development of ASTM Type II and V cements (Al-Dulaijan et al., 2003), which have limited C3A contents. During the past decades, several theories have been proposed for possible explanation of the expansion caused by formation of ettringite, such as swelling due to formation of colloidal ettringite (Mehta, 1973), topochemical reaction (Cohen and Richards, 1982, Odler and Colán-Subauste, 1999), increase in volume of the solid (Taylor, 1994), and crystal growth pressure (Ping and Beaudoin, 1992, Min and Mingshu, 1994, Scherer, 1999, 2004). These theories were also summarized by (Brown and Taylor, 1999), among which the expansion was often assumed to be related only to an increase of the solid volume. However, direct evidence showing that expansion is purely attributed to the increase in solid volume have never been observed. Recently, crystallization pressure theory has been supported by several observations (Yu et al., 2013). These observations confirm that the expansion caused by crystal growth

30


under supersaturation of the pore solution is the most likely mechanism for expansion caused by formation of ettringite during sulfate attack. (Yu et al., 2013) reported evidence for that the expansion is related to the transformation of monosulfate crystals to ettringite embedded in the C-S-H. During the sulfate ingress process, sulfate ions firstly react with aluminate hydrates in larger pores to form ettringite without expansion, and at the same time the concentration of sulfate ions is buffered by the readily available aluminate hydrates. Once these aluminate hydrates have almost fully reacted, the concentration of sulfate ions in the pore solution increases, as indicated by the rise of sulfate absorbed on the C-S-H. This leads to supersaturation of the pore solution and the presence of driving forces for precipitation of ettringite crystal within small confined pores (less than 0.1 μm) in the C-S-H, and thus to expansion of the cement matrix. (Müllauer et al., 2013) designed a specially constructed stress cell to measure the stress generated in thin-walled Portland cement mortar cylinders caused by sulfate attack. They concluded that the damage is caused by the formation of ettringite in small pores (10 – 50 nm), which generates stress up to 3 MPa exceeding the tensile strength of the binder matrix (3 to 4 MPa). Higher sulfate concentrations and C3A contents result in higher stresses. The formation of ettringite in larger pores also takes place, but its contribution to the expansion and damage is negligible as a result of insufficient pressure generated. The phenomena are explained based on the effect of crystal surface energy and size on supersaturation and crystal growth pressure.

2.4.2 Formation of gypsum Gypsum is formed during sulfate attack through interactions between sulfate ions and portlandite. Although sulfate attack can cause expansion of concrete, some damages of concrete are caused by softening and spalling, rather than expansion. Studies have indicated that the C3A content less than 5 wt.% for Type V cement may not prevent damages caused by sulfate attack (Gonzalez and Irassar, 1997, Lamond, 2006). In some cases, cement without C3A does not provide resistance to sulfate attack (Gonzalez and Irassar, 1997). Nowadays, it is generally accepted that the damage caused by sulfate attack cannot solely be explained by the formation of ettringite. Gypsum is often found to damage concrete through softening as well as mass and strength loss of concrete (Cohen and Mather, 1991). Whether gypsum formation led to any expansion has been controversial, for which reason (Tian and Cohen, 2000) reviewed the literature dealing with the deterioration and mechanism of sulfate attack 31


caused by gypsum formation. They also conducted relevant experiments on C3S hydrated cement paste and found that formation of gypsum also caused expansion, which indicated that the expansion and cracking of Portland cement concrete should probably not be exclusively attributed to ettringite formation (Tian and Cohen, 2000). The expansion was also reduced when C3S is partially replaced by silica fume. Later, (Santhanam et al., 2003) also studied sodium sulfate attack on C3S mortars and an increased expansion was measured with increased formation of gypsum.

2.4.3 Formation of thaumasite Thaumasite sulfate attack (TSA) may occur in hydrated Portland – limestone cements when they are exposed to sulfate environments. It is different from conventional sulfate attack because it is the C-S-H rather than aluminate hydrates and portlandite that reacts with sulfate and carbonate ions, forming a white mushy and non-cohesive substance called thaumasite. Thaumasite has a similar crystal structure as ettringite (Lachowski et al., 2003), and it is also a member of the AFt family. In general, thaumasite is preferentially formed at low temperatures (≤ 15 oC) in the presence of calcium silicate, sulfate and carbonate ions, and sufficient moisture. However, some researchers (Irassar et al., 2005, Blanco-Varela et al., 2006, Lee et al., 2008) have also reported the formation of small quantities of thaumasite in pastes and mortars at room temperature. Formation of thaumasite at room temperature is thermodynamically very slow, but it remains stable up to 30 oC once it is formed (Macphee and Barnett, 2004). Formation of thaumasite is affected by many factors, e.g. limestone, reaction time, C3A content, C-S-H content, portlandite, temperature and leaching. Thaumasite itself contains no aluminum, but its formation is influenced by the amount of aluminum. (Schmidt et al., 2008) found that the C3A content has no significant influence on the formation of thaumasite, but it can only be formed when the molar SO3/Al2O3 ratio in the cement exceeds 3, otherwise only ettingite is formed for lower amounts of SO3. Since the formation of thaumasite is kinetically very slow, gypsum may form and act as a source of sulfate for the precipitation of additional thaumasite. (Bellmann and Stark, 2007, 2008) stated that thaumasite can be formed from C-S-H and portlandite at very low sulfate concentrations whereas higher sulfate concentrations are needed in the absence of portlandite. It is suggested by the same authors (Bellmann and Stark, 2007, 2008) that a lower Ca/Si ratio requires higher sulfate concentration to form thaumasite. Formation of thaumasite is very complicated and further research is needed to achieve a 32


holistic picture of the reaction mechanism before understanding how the formation of thaumasite causes damage of concrete.

2.4.4 Mitigation of sulfate attack Although the utilization of low-C3A cement was proposed to reduce the risk of sulfate attack caused by formation of ettringite, a gypsum oriented sulfate attack characterized by a softening of the cement matrix may be amplified as a result of the presence of increased amounts of portlandite due to higher C3S contents in cement. This suggests partial substitution of Portland cement by SCMs for potentially enhancing the sulfate resistance. It has been documented that partial substitution of Portland cement with alumina-rich SCMs (Irassar et al., 1996, Al-Dulaijan et al., 2003, Bhatty and Taylor, 2006), such as slag (Yu et al., 2015), fly ash (Chindaprasirt et al., 2004) and metakaolin ((Al-Akhras, 2006), can mitigate conventional sulfate attack. Some researchers (AlDulaijan et al., 2003) attributed this to the consumption of portandite by pozzolanic reactions and the dilution of the C3A content. However, the presence of an additional alumina source from SCMs may cause formation of more ettringite. It was also reported that concretes with SCMs may exhibit a greater surface scaling (Irassar et al., 1996). The most plausible reason for improved sulfate resistance for SCM blended concrete is attributed to a refinement of the pore structures, and thereby a slower ingress of sulfate ions (Bhatty and Taylor, 2006, Najimi et al., 2011). Because of the different damage processes between SCM blends and Portland cement concretes, a detailed description of the damage process is needed for evaluation of the sulfate attack for the SCM blends in order to develop a performance based test method for these blends. Recently, a layer-by-layer degradation process and the refinement of the microstructure are likely to explain the higher sulfate resistance of mortars and concretes including slag. (Yu et al., 2015) conducted comparative experimental studies on slag blended mortars and pure Portland cement mortars, and found that the damage of slag blended mortars is governed by the loss of surface rather than expansion as compared to Portland cement mortars. This is explained by the major fixation of sulfate ions by the aluminate phases in a relatively narrow region close to the surface of the slag blend mortar due to the buffering effect by slag addition, consequently by refinement of the pore structures. When the aluminate hydrates on the surface have fully reacted, supersaturation of sulfate ions result in a driving force for crystal growth of ettringite. If ettringite precipitates in a sufficient amount, spalling of sulfate of the mortar occurs, and then 33


sulfate ions are readily to penetrate into the freshly exposed surface. This process can happen without detection of expansion. (Yu et al., 2015) also found that the ingress depth of sulfate ions do not depend on the external sulfate ion concentration as a result of the buffer effect of aluminate hydrates. However, higher sulfate ion concentrations can cause a higher crystallization pressure of ettringite, and thus an earlier occurrence of the spalling of the surface. In general, owing to the refinement of the pore structures and the higher buffer effect of aluminate hydrates, the process for damage is very slow compared to the damage for Portland cement mortars. Combination of sufficient amounts of SCMs with Portland – limestone cement was also found to have high resistance against thaumasite sulfate attack (Tsivilis et al., 2003, Skaropoulou et al., 2009, Ramezanianpour and Hooton, 2013). (Mirvalad and Nokken, 2015) also reported that an improved resistance to thaumasite sulfate attack can be achieved by increasing the amount of SCM, e.g. slag, metakaolin and fly ash. (Tsivilis et al., 2003) mentioned that the formation of thaumasite requires transportation of Ca2+, 𝐶𝑂32− , 𝑆𝑂42− ions, and the use of SCMs that refine the pore structure may therefore contribute to a better performance against thaumasite sulfate attack. Besides, (Bellmann and Stark, 2007, 2008) also suggested that silicon rich C-S-H, formed by pozzolanic reaction shows higher resistance against thaumasite formation.

34

Chapter 3: Materials and Methods

CHAPTER 3: MATERIALS AND METHODS This chapter describes the materials, mix designs and methods used in this thesis. The specific experimental procedures, e.g. for total chloride profile analysis, measurement of carbonation depth and determination of expansion caused by sulfate attack, are not described in this chapter but in the chapters where the specific procedures are employed. Despite of the types of durability issues investigated, two experimental batches are conducted in the thesis: (i) studies on blends with constant water to binder ratio (batch 1); (ii) studies on blends with comparable compressive strengths (batch 2). For the two batches of experiments, different materials and mix designs are adopted.

35


3.1 Mixes with the same water/binder ratio (batch 1) 3.1.1 Materials The binders used in this part were made from a white Portland cement (wPc, CEM I, 52.5 N) and three supplementary cementitious materials (SCMs): metakaolin (MK), white silica fume (SF) and limestone (LS). The wPc was produced by Aalborg Portland A/S from Aalborg, Denmark, and included 3.1 wt.% LS, 4.1 wt.% gypsum and 1.9 wt.% free lime. The MK was produced in the laboratory from kaolinite (Kaolinite Supreme TM from Imerys Performance Minerals, UK) by thermal treatment at 550 oC for 20 h. The SF was purchased from Elkem, Norway. The LS was a Maastrichtian chalk from Rørdal, Northern Denmark. The chemical, physical and mineralogical properties of the starting materials are given in Table 3.1 (Batch 1). The wPc contained 64.9 wt.% alite (C3S), and 16.9 wt.% belite (C2S) and 7.8 wt.% aluminate (C3A) (Dai et al., 2015), where the content of the silicate phases were determined by 29Si MAS NMR according to the method proposed by (Skibsted et al., 1995) and the quantity of the aluminate phase was determined by mass balance calculations. The ferrite (C4AF) phase was not taken into account as the small amount of iron is expected to be incorporated as guestions in the C3S, C2S and C3A phases. The sand used to for the mortars was a CEN reference sand (Normensand GmbH, Germany), which has a silica content of at least 98 wt.% and a density of 2650 kg/m3. A superplasticizer (SP, Glenium 27, BASF) was used to achieve similar flow for all mortars.

3.1.2 Mix design The compositions of the binders (Table 3.2) targeted a replacement of 35 wt.% of white Portland clinker by SCMs. Considering the small amounts of LS and gypsum in the wPc, this resulted in the actual binder compositions with 31.9 wt.% replacement of the wPc. The P mortars produced by pure wPc and the L mortars produced by a combination of wPc and 31.9 wt.% LS were prepared as reference mortars. The ML and M mortars were produced by replacing wPc with MK (1:1 clay) and/or LS, where the actual Si/Al ratio of 1.13 for the MK accounts for the 2 wt.% quartz in the MK (Dai et al., 2014). The MSL and MS mortars were produced by replacing wPc with MK, SF and/or LS, where a fixed combination of MK and SF was used to mimic a 2:1 clay (Dai, 2015). However, the actual Si/Al ratio = 2.36 was higher than the ideal ratio in order to account for the partial substitution of Al by Mg in the octahedral layers found in

36


Table 3.1. Chemical, physical and mineralogical properties of the raw materials (wt.%) Batch 1 Batch 2 (For chapters 4, 5, 6, 7, 8) (For chapters 4, 8) wPc LS #4 MK SF clinker MT #5 G SiO2 21.81 3.92 52.84 90.44 19.55 63.63 37.56 Al2O3 3.56 0.33 39.49 0.34 6.05 19.94 27.43 Fe2O3 0.24 0.14 1.42 0.03 3.32 3.90 0.043 CaO 66.13 53.73 0.22 1.37 65.98 1.20 27.24 MgO 1.1 0.35 0.48 0.93 0.92 4.62 K2O 0.43 0.05 1.00 1.87 0.52 0.15 < 0.02 Na2O 0.04 0.08 0.05 0.19 0.25 3.82 7.56 SO3 (fused bed) 3.37 0.05 0.06 0.3 1.57 0.73 TiO2 0.21 0.02 0.88 0 0.32 0.35 Cl 0.003 0.01 0.003 0.13 0.004 0.06 P2O5 0.04 0.1 0.11 0.55 0.33 0.04 MnO 0.009 Cr2O3 0.02 0 0.008 0 0.012 0.005 < 0.001 LOI 2.57 41.8 3.55 3.35 0.47 1.34 Total 99.52 100.58 100.11 99.51 99.3 99.78 #1 CRTOT (mg/kg) 124.47 0 55.29 21.21 36.88 Na2Oet 0.33 0.11 0.61 1.42 Carbon content (by LECO) 0.37 Limestone content 3.08 93.83 SO3 (by LECO) 3.31 Quatz content (by NMR) 2 Gypsum content 4.11 Free lime 1.91 3 Density (kg/m ) 3080 2700 2530 2240 3180 2595 2760 2 Blaine (m /kg) 287 1211 1891 330 758 690±10 45 microns residue 2~4 20 microns residue 2.7 26 #2 #3 C3S 64.9 74.6 #2 C2S 16.9 0 C3A (by mass balance) 7.8 10.43 C4AF (by mass balance) 0 10.09 #1

CRTOT is total chromium content in mg/kg, which is calculated considering Cr 2O3 content and molecular weight ratio between Cr and Cr2O3. #2

Measured by 29Si MAS NMR (Dai et al., 2015).

#3

Calculated based on Bogue calculation (Taylor, 1997).

#4

Some LS is also used in the batch 2 investigations

#5

Data for MT after calcination

37


montmorillonite (Garg and Skibsted, 2014). A previous study (Dai, 2015) has found a synergetic effect between LS and MK, and highest compressive strength for mortars with MK / (MK + LS) = 0.75 and (MK + SF) / (MK + SF + LS) = 0.75 in binary and ternary blends, respectively. The binders were used to produce mortars with a constant water to binder ratio (w/b = 0.5) and binder to sand ratio (b/s = 1/3), both ratios by weight. The dosage of superplasticizer (SP/(MK+SF) = 0.04 for the mortars including SF, and SP/MK = 0.07 for mortars without SF) was adjusted to achieve a flow within ± 5% of the flow of the reference P mortar. Table 3.2. Binder blend compositions of mortars (wt.%). id Ratio (g/g) wPc MK SF P 100 0 0

LS

Si/Al (mol/mol)

0

-

L

MK/(MK+LS) = 0

68.1 0

ML

MK/(MK+LS) = 0.75

68.1 25.5 0

6.4

1.13

M

MK/(MK+LS) = 0.94

68.1 31.9 0

0

1.13

MSL (MK+SF)/(MK+SF+LS) = 0.75 MS

(MK+SF)/(MK+SF+LS) = 0.94

0

31.9 -

68.1 15.6 9.88 6.4

2.36

68.1 19.5 12.4 0

2.36

3.1.3 Mortar preparation Mortars for different durability investigations are mixed following the same mixing procedures. Moreover, in order to prevent the agglomeration of silica fume, the mixing procedure for silica fume blends is slightly different from that for the reference and metakaolin blends. The speeds of the mixer blades specified in (EN196-1, 2005) in Table 3.3 are adopted. Table 3.3. Speeds of mixer blades specified in (EN196-1, 2005) Rotation (min -1) Gears Planetary movement (min -1) 1st gear: Low speed

140 ± 5

62 ± 5

2nd gear: High speed

285 ± 10

125 ± 10

Mixing procedure for the P, L, ML and M mortars (only at low speed) Binder blends for P, L, ML and M mortars are premixed and homogenized before mixing with sand. The mixing procedure is described as follows: 1) Dry mixing of all the materials (sand and blended cement) for 30 s. 2) Addition of demineralized water, saving some water for a better pouring of SP. 38


Note the time when water is added (to be considered as zero time). 3) Wet mixing for 30 s. 4) Addition of superplasticizer (Glenium 27) which has been dissolved in that amount of water mentioned above. 5) Wet mixing for 4 minutes and 30 seconds. Total mixing time: 5 minutes and 30 s Mixing procedure for MSL and MS mortars 1) Mixing of SF, sand and demineralized water (saving some water for a better pouring of SP) for 2 minutes in the 1st gear. 2) Then, mixing for 3 minutes in the 2nd gear. 3) Stop and addition of blended cement on top. 4) Wet mixing for 30 s in the 1st gear. 5) Then, addition of superplasticizer (Glenium 27) which has been dissolved in that amount of water mentioned above. 6) Wet mixing for 4 minutes and 30 seconds in the 1st gear. Total mixing time: 10 minutes. For different durability investigations, mortars are cast in different molds at the end of the mixing. These are described separately as follows together with curing, preconditioning and sample treatment if necessary. 

For chloride profile analysis (chapter 5)

Mortars were mixed and cast into polypropylene bottles (ø50 mm, 125 ml). A small amount of water was added on top of the mortar to keep the mortar saturated and to compensate for self-desiccation caused by early age chemical shrinkage (Geiker, 1983). Mortars were demolded after 24 hours and cured in demineralized water for 91 days in an airtight bucket with a water/mortar ratio of 3:1 by volume at 20 ± 1 oC. At the last week of curing, mortar cylinders were treated by removing about 5 mm mortar from both ends (wet cutting) in order to remove the paste-rich layer, the potentially carbonated layer and the leaching layer. Immediately after cutting, the mortars were placed on a table and exposed to the natural environment in the laboratory for 2 hours for surface drying, followed by a two-layer epoxy coating (1 mm thickness) on the bottom and circumference surfaces, leaving the top surface uncoated for exposure to the chloride solution after 91 days of curing. When the epoxy hardened, the uncoated

39


surfaces of mortars were submerged in small amount of demineralized water for resaturation until the end of the 91 days of curing. 

For carbonation investigations (chapter 7)

Mortars were cast into 40 × 40 × 160 mm3 molds and cured in a moist cabinet maintained at a temperature of 20 ± 1.0 oC and a relative humidity of not less than 90% for 24 hours. After demolding, the mortar bars were cured in demineralized water in sealed buckets with a water-solid ratio of 3:1 by volume at 20 oC for 91 days. 

For sulfate attack (chapter 8)

Mortars were cast into 20 × 20 × 160 mm3 molds with embedded gauge studs. They were cured in a moist cabinet maintained at a temperature of 20 ± 1.0 oC and a relative humidity of not less than 90% for 24 hours. After demolding, the mortar bars were cured in demineralized water in sealed buckets with a water-solid ratio of 3:1 by volume at 20 oC for 91 days.

3.1.4 Paste preparation 

For chloride binding investigations (chapter 6)

Selected binder blends (P, ML and M) from Table 3.2 are used for the chloride binding investigations. Paste samples were produced with the same water/binder ratio (w/b = 0.5) by weight. They were mixed according to the laboratory mixing procedure described in (Poulsen et al., 2009) and then cast and sealed in plastic bags. The fresh pastes in the bags were flattened to a thickness of about 5 mm to avoid formation of hydration shells around the hydrating cement grains (Kjellsen et al., 1991). All the pastes were sealed cured in a moist cabinet at 5 oC for the first three days for the same purpose, followed by curing in a moist room at 20 oC for about 2 months. The resulting cement paste plates were split into small pieces using a hammer and subsequently ground to 1 mm fine particles. The resulting powder was collected in a 1.0 liter polypropylene bottle and mixed with 30 wt.% distilled water by weight of the powdered cement paste. The new mixes with a resulting w/b ratio of 0.95 were stored in bottles and rotated slowly along the longitudinal axis for additional 7 days at 20 oC. The whole curing procedure aimed to maximize the degree of hydration of the cement pastes and minimize carbonation. Each moist cement paste was crushed and homogenized, giving samples that look like “moist sand”.

40


3.2 Mixes with comparable compressive strengths 3.2.1 Materials The binders used in this part were made from a laboratory produced white Portland cement (wPc’), calcined montmorillonite (MT), limestone (LS, same as used in the batch 1 experiments), and a laboratory produced glass (G). The chemical, physical and mineralogical properties of the starting materials are given in Table 1 (Batch 2). The clinker used to produce wPc’ contains high C3S (74 wt.%) and C3A (10.5 wt.%) contents and it has a density of 3180 kg/m3. The clinker was ground in a 20 kg laboratory scale mill with 0.03% triethanolamine (TEA, 50% water based solution) to get a Blaine fineness of 330 m2/kg and a 45 microns residue around 2 – 4 wt.%, which targets a compressive strength of approximately 65 – 67 MPa at 28 days. The clinker was then SO3 optimized firstly by adding hemihydrate, which was obtained by drying ground natural gypsum (Blaine fineness around 650 m2/kg) at 110 oC for 2 days. The target wPc’ is comparable to a CEM I 52.5 N. The calcined MT is produced in the laboratory from montmorillonite (Nanoclay, hydrophilic bentonite, SIGMA-ALDRICH, USA) by thermal treatment at 650 oC for 20 h. Each time 2 × 80 g montmorillonite is heated in two crucibles. The Blaine fineness of the calcined MT was 436 m2/kg and the 20 microns residue is 19.7%. The calcined MT was ground to a Blaine fineness of 758 m2/kg and 20 microns residue of 2.7% in order to increase its reactivity. The density of the ground calcined MT is 2595 kg/m3. The glass (G) was produced at Aalborg University, in a paralleled sub-project of the LowE-CEM project. The sand and superplasticizer (SP, Glenium 27, BASF) used are the same as for the batch 1 experiments described in the beginning of this chapter.

3.2.2 Mix design and mortar preparation For generic comparison, a similar mix design was applied as for batch 1 experiments. For example, SCM / (SCM + LS) = 0.75 and 1 are used, where SCM in this part represents MT and G. The reason for the ratio 1 used in this part is that wPc’ does not include any LS as compared to wPc. In addition, MT / (MT + LS) = 0.5 was also investigated. Moreover, all the mortars investigated in this part are designed to have comparable compressive strengths. Therefore, both the water to binder ratio (w/b) and the binder to sand ratio (b/s) are adjusted. For this purpose, some preliminary trial experiments to measure the 28 day compressive strengths were conducted. The results

41


were corrected to the same air content and then correlated to the w/b ratios. Thus, a constant, K, can be determined according to the empirical equation: 1 E[𝑓𝑐 ] = 𝐾 ∙ ( ) − 0.5 0.45 ≤ 𝑤/𝑏 ≤ 1.25 𝑤/𝑏 where E[𝑓𝑐 ] is the mean value of compressive strengths; 𝐾 is a constant, depending on the degree of hydration and cement type; 𝑤/𝑏 is the water to binder ratio. Based on the determined K value for each mix and targeted compressive strength, the w/b ratio can be calculated. Table 3.4. Binder blend compositions of mortars (wt.%). Mortar id

wPc'

P'

Mix design -

L'

0

66

MTL-0.5

MT/(MT+L) ratio = 0.5

66

MTL-0.75

MT/(MT+L) ratio = 0.75

MT-1

MT

G

LS

Sand

s/b

w/b

300

3

0.5

34

274

4.16

0.41

17

17

299

4.53

0.48

66

26

9

307

4.66

0.5

MT/(MT+L) ratio = 1

66

34

286

4.34

0.44

GL-0.25

G1/(G1+L) ratio = 0.25

66

9

26

297

4.5

0.48

GL-0.5

G1/(G1+L) ratio = 0.5

66

17 17

310

4.7

0.51

GL-0.75

G1/(G1+L) ratio = 0.75

66

26 9

324

4.91

0.55

G-1

G1/(G1+L) ratio = 1

66

34

305

4.62

0.5

100

Mixing procedure for all mortars (only at low speed) All binder blends are premixed and homogenized before mixing with sand. The mixing procedure is described as follows: 0 → 30s

Mixing of cement and water in 1st gear, reserve some water for SP addition.

At 30s

Addition of sand while mixing in 1st gear.

At 1.5 min

Addition of the SP mixed with the reserved water while mixing in 1st gear.

1.5 → 2 min

Mixing in 1st gear.

2 → 2.5 min

Mixing in 2nd gear.

2.5 → 4 min

Stop and place mortar adhering to the wall in the middle of the bowl

4 → 5 min

Mixing in 2nd gear.

Total mixing time 5 min 42


At the end of mixing, mortars were cast in different molds for the different durability investigations, following the same casting procedure and curing conditions for 24 h as used for the batch 1 experiments. Afterwards, all mortars are demolded and most of them are cured in demineralized water (water to solid ratio equal to 3) for 91 days. However, the glass mortars were wrapped in a wet cloth and then sealed in plastic bags to prevent alkali leaching. All the mortars were stored in a moist room with controlled temperature of 20 oC. The way to pretreat the mortars, if necessary prior to durability investigations, is the same as already described above.

43


3.3 Methods 3.3.1 Workability measurement The orkability of fresh mortar is measured immediately after the preparation of the mortars. A defined mold is placed on a defined flow table disc. Freshly made mortar is introduced in the mold in two layers, each layer being compacted by at least 20 short strokes of the tamper. During filling, the mold is firmly held on the disc. The excess mortar is skimmed off with a palette knife and the free area of the disc is wiped clean and dry. The mold is then slowly raised vertically. The mortar is spread out on the disc by applying 20 jolts. The diameter of the flowed mortar is measured in 4 directions at right angles to one another using calipers. The calculated mean values can be used to describe the workability.

3.3.2 Strength testing 

Strengths for batch 1 mortars

Compressive strengths of the mortars prepared for the batch 1 experiments were determined in the “SCM project”, as presented in the thesis by (Dai, 2015), on MiniRILEM mortar prisms (2 × 2 × 15 cm3). After the relevant time of hydration, the specimens were cut into three pieces of length 4.8 cm for the strength measurement. 

Strengths for batch 2 mortars

Finally, the compressive strengths of the mortars prepared for the batch 2 experiments were determined in this work at the end of the curing for 91 days. Due to the limited numbers of mortars and the targeted comparable strengths for all mortars, only one mortar bar (40 × 40 × 160 mm3) for each mix was cut into two pieces for strength test. The reported data is an average of the two measurements.

3.3.3 PF method The description of the PF method has been documented earlier by (Sellevold and Farstad, 2005) and recently modified and extended for its use by (De Weerdt et al., unpublished). The principle of this method is that the mass evolution over time of drying at 50 oC and 105 oC allows a differentiation of the coarse and fine pores. The weight of the specimens (ø50 mm, thickness: ~30mm) were recorded using a scale with a chamber with an accuracy of 0.001g. The scale allows weighing the specimens both in

44


air and under water. Pressure saturation was applied followed by again drying again at 105 oC. The “capillary” porosity is calculated according to the mass loss at 50 oC and the “gel” porosity is calculated according to the mass loss at 105 oC. The total porosity is the summation of these two porosities with air voids obtained from pressure saturation.

3.3.4 Mercury intrusion porosimetry Mercury intrusion porosimetry (MIP) is widely used for studies of hardened Portland cement-based materials. The concept and procedure of this method is well described in (Canut, 2012). Samples for MIP measurements are prepared by the isopropanol-type of solvent exchange for 2 days to remove the water in the pore structures, followed by drying at 50 oC in oven for 24 hours. In this study, two full subsequent intrusion cycles were performed using a combined Pascal 140/440 equipment on samples with a grain size of 2 – 3 mm, employing a contact angle of 140o and a surface tension of 0.48 N/m (Kaufmann, 2010). The maximum applied pressure was 200 MPa, corresponding to a minimum pore radius of 3.6 nm. The total porosity (intruded pore volume) and the pore connectivity (from the threshold pore size) can be obtained from the intrusion curves. The threshold pore size was obtained from the intersection of the two tangents in the intrusion curve, as described in an earlier study (Canut, 2012).

3.3.5 AC impedance AC (alternating current) impedance spectroscopy has been successfully used for studies of Portland cement-based materials (Gu et al., 1993). In the present study, the electrical resistance of the mortar cylinders (ø50mm, thickness: ~30mm) was measured using an AC impedance instrument. The used electrodes were two metal plates and the electrode contacts were two slices of wet cloth soaked with a curing solution. The frequency range is 10Hz ~ 850MHz. The conductivity can finally be calculated from the measured electrical resistance, according the relation: conductivity = resistance × cross section area / length.

3.3.6 Scanning electron microscopy In this study, scanning electron microscopy (SEM) was employed to study the microstructural changes of the samples exposed to carbonation in chapter 7. Images of the non-carbonated and the carbonated areas of the mortars were acquired with an 45


environmental scanning electron microscope (ESEM-FEG XL30) in the backscattering electron mode. Sections were impregnated in an epoxy resin and polished. The sections were studied in high vacuum (3.0·10-6 – 6.0·10-6 Torr) with an accelerating voltage of 12 kV and a beam current of 310 – 330 μA.

3.3.7 Thermogravemetric analysis Thermogravimetric analysis (TGA) was performed on powdered mortar samples which were used for the total chloride profile analysis (chapter 5), samples for chloride binding (chapter 6) and samples for carbonation investigations (chapter 7). A Mettler Toledo TGA/SDTA 851 instrument was used. About 50 mg of the powdered samples were loaded in a 150 µm alumina crucible. The weight loss of the samples was monitored while heating from 30 oC to 980 oC at a rate of 20 oC/min and purging with 50 ml/min N2. The weight loss and quantified data were reported as mass percentage of dried mortars at 105 oC for the chloride ingress samples, at 50 oC for the samples used for chloride binding tests, and at 800 oC for the samples for the carbonation test.

3.3.8 X-ray diffraction analysis X-ray diffraction (XRD) analysis was carried out with a CubiX3 diffractometer (PANalytical, NL) for Bragg angles (2Θ) of 5o to 65o, using a step size of 0.02o. The powdered mortar samples used both for the total chloride profile analysis and the TGA analysis in chapter 5 were prepared by mixing and grinding 1.80 g of sample and 0.20 g of high-purity nanocrystalline anatase (TiO2) in a planetary ball-mill for 45 seconds at 350 rpm. The mixed powder was then lightly pressed into a pellet with a smooth surface. The acquired diffractograms were simulated with the Rietveld procedure using starting parameters for the individual phases obtained from the Inorganic Crystal Structure Database (ICSD). Only the scale factor, unit cell dimensions and profile parameters were refined in the process. The addition of a known amount of anatase allowed for a possible quantification of the observed phases (e.g. Friedel’s salt and portlandite), despite the presence of amorphous phases and a large amount of quartz in the analyzed powdered mortar samples.

3.3.9 Thermodynamic modelling Thermodynamic modeling was carried out using the Gibbs free energy minimization program, GEMS3 (Wagner et al., 2012, Kulik et al., 2013), which calculates the 46


equilibrium phase assemblages in chemical systems from their total bulk elemental composition. The default databases were expanded with the CEMDATA07 database (Lothenbach et al., 2008b) and additional data for the C-S-H phase (Kulik, 2011). The data include solubility products of the solids relevant for cementitious systems and for C-S-H phases with different compositions. 

For hydration (Chapter 4 and Appendix 1)

The thermodynamic modelling calculates the complete reaction without considering the reaction kinetics, thus a restriction of the reaction was employed to predict the phase assemblages as a function of time. For this purpose, the hydration degree was varied with time to build up the restrictions imposed by hydration kinetics. The initial amounts of the C3S, C2S, MK and SF phases were obtained from NMR, whereas the C3A content was calculated by mass balance. The hydration degrees for C3S, C2S, MK and SF at different curing ages were quantified by using 29Si MAS NMR deconvolution technique as presented by (Dai, 2015). The hydration kinetics for the C3S, C2S, MK and SF phases to be implemented into the thermodynamic modelling were achieved by fitting the NMR results. These calculations have been presented in the thesis by (Dai, 2015), however, repeated calculations can also be found in Appendix I using different kinetic model. The predicted phase assemblages were adopted as a starting point to predict the phase assemblages upon exposure to chloride, CO2 and sulfate environments. 

For chloride ingress and binding (chapter 5, 6)

The thermodynamic modelling was also applied to predict the phase assemblages of both mortars and pastes exposed to chloride. The ingress of chloride for the mortar samples was mimicked assuming that the inner core of the mortars is not in contact with the chloride ions by adding very small volumes of chloride solution. On the other hand, the outer layers of the mortars were mimicked by increasing the amount of chloride used in the calculations. In addition to the chloride binding in Friedel’s salt, chloride can also be bound by the C-S-H phase, which is not considered in the calculations. The influence of chloride binding on the C-S-H has been neglected in the current calculations, as the aluminate phases are much more important than the other phases for the chloride binding (Zibara et al., 2002). 

For carbonation (chapter 7)

47


The changes in phase assemblages and total porosity upon exposure to CO2 were predicted for the different blends in chapter 7. The ingress of CO2 was mimicked by increasing the amount of air (constant CO2 concentration, 1% (v/v)) from the center to the surface of the mortars. The uptake of alkalis by the C-S-H phase is taken into account by employing an ideal solid-solution model between the C-S-H phase and two hypothetical alkali silicate hydrates ({(KOH)2.5SiO2H2O}0.2 and {(NaOH)2.5SiO2H2O}0.2) (Lothenbach et al., 2012) as proposed by (Kulik et al., 2007). 

For sulfate attack (chapter 8)

Thermodynamic modelling has been successfully used to illustrate the changes in the phase assemblages when prisms are exposed to Na2SO4 solutions in several studies (Schmidt et al., 2008, Schmidt et al., 2009, Lothenbach et al., 2010, Kunther et al., 2013). The ingress of sodium sulfate was mimicked assuming that the inner core of the mortars is not in contact with the sodium sulfate by adding very small volumes of sulfate solution. The outer layers of the mortars were mimicked by increasing the amount of solution used in the calculations. At 20 oC, formation of thaumasite is a very slow process (Lothenbach et al., 2008b) and the formed amount will be negligible and thus it is suppressed in the calculations.

48

Chapter 4: Sample Characterization

CHAPTER 4: SAMPLE CHARACTERIZATION Durability of concrete is strongly affected by the types of hydration product and the pore structures of the concrete. Chapter 4 describes the characterization of the studied mortars hydrated for 91 days, which includes phase assemblages and the analysis of compressive strengths and pore structures. The phase assemblages are calculated by thermodynamic modeling and evaluated by TGA measurements. The pore structures are evaluated using MIP, PF and AC impedance methods. These methods have already been described in chapter 3.

49


4.1 Phases assemblages 4.1.1 wPc – MK – SF – LS mortars The hydration kinetics for the wPc – MK – SF – LS pastes were derived from the data measured by

29

Si MAS NMR, and the results were presented in the thesis by (Dai,

2015), in which the data were also implemented in thermodynamic modeling for prediction of phase assemblages as a function of hydration time. The data has also been recalculated in the current study with a different kinetic model as described in Appendix I. The results show that the increase in hydration degree is less remarkable after 28 days of hydration. Based on this, the mortars investigated in this study are first exposed to the detrimental environments after curing for 91 days in order to eliminate the impact of hydration on durability. The phase assemblages after 91 days are presented in Figure 4.1. These calculations are later used for predictions of the phase assemblages for samples exposed to chloride, sulfate and CO2 environments.

Figure 4.1. Phase assemblages calculated with thermodynamic modeling for the wPc – MK – SF – LS blends based on hydration degree data at 91 day measured by (Dai, 2015) using 29Si MAS NMR. The results in Figure 4.1 show that the major hydration products for the pure Portland cement blend (P) are C-S-H, ettringite, portlandite, and monocarbonate. In addition, a minor amount of hydrotalcite is also predicted to accommodate for the magnesium content. The same hydration products are also predicted for the Portland cement – limestone blend (L), where most of the LS contribute with a dilution effect in the system since only a small fraction of the LS is needed to form monocarbonate as seen in the P blend. No portlandite is predicted for the other blends as a result of pozzolanic 50


reactions. However, more AFm phases are predicted for these blends as compared to the P and L blends. In the ML and MSL blends, the major AFm phase is monocarbonate formed in the presence of excess LS, whereas several AFm phases are predicted for the M and MS blends. The presence of these phases was also confirmed by XRD analysis as presented in (Dai, 2015). A comparison of the measured and predicted amount of portlandite in the hydrated mortars is shown in Table 4.1, which indicates a fairly good agreement between measurements and calculations, though minor amounts of portlandite are still observed experimentally. Table 4.1. Comparison of measured and predicted portlandite of the hydrated mortars (g/100g ignited mortar at 800 oC). Method

P

L

ML

M

MSL

MS

GEMs #1

8.2

5.2

0

0

0

0

TGA #2

7.4

5.2

0.6

0.3

0.5

0.3

#1

The data predicted from GEMs (no sand) are divided by a factor of 4 according to the binder/sand ratio of 3 for the mortars. #2

The samples for the TGA measurements were taken from the non-carbonated inner core of mortars after 91 days of carbonation.

4.1.2 wPc’ – MT –LS and wPc’ – G – LS mortars The DTG curves for the wPc’ – MT – LS and wPc’ – G – LS mortars and the quantified amount of portlandite are shown in Figure 4.2. The C-S-H phase, ettringite and AFm phases are observed in the low temperature range. Portlandite is observed to be present in all mortars in significant quantities as shown in the DTG curves at 480 oC. The results of quantification of the portlandite show that the amounts of portlandite present in the MT and G mortars are pronounced, which indicates that only a small fraction of MT and G reacts with a very low pozzolanic reactivity as compared to the MK mortars. The results are in line with that observed in the thesis by (Garg, 2015), where he compared the dissolution rates of MK and MT, and demonstrated the pronounced difference between the dissolution rates of the two calcined clays. The DTG curves in Figure 4.2 also show the presence of LS in all mortars, but the observation of LS in the P, MT-1 and G-1 mortars is not expected as no LS were added when producing these mortars. It should be noted that the samples used for the TGA measurements were taken from the surface of the mortars, which may potentially be carbonated during the curing period or transport. 51


Figure 4.2. DTG data for the wPc’ – MT – LS and wPc’ – G – LS mortars hydrated for 91 days: (left) DTG curves, (right) quantified portlandite contents from the DTG curves. (The samples were taken close to the surface of mortars, thereby slightly carbonated as seen in the P, MT-1 and G-1 mortars).

4.2 Compressive strength 4.2.1 wPc – MK – SF – LS mortars The development of the compressive strengths for the studied mortars is described in the thesis by (Dai, 2015), which shows a slow increase in strength for all mortars from 28 days up to 180 days. The results together with small changes in hydration degree presented in his thesis indicate that the mortars used in this study are well hydrated after curing for 91 days. Thus, the impact of minor continuous hydration on their durability performance is considered to be negligible. The 91-day compressive strengths for the mortars produced with the wPc – MK – SF – LS blends and the same w/b = 0.5 are shown in Figure 4.3. The results show that the compressive strength for the L mortar is substantially lower than for the P mortar as a result of the dilution effect, since only a small fraction of LS has reacted according to the thermodynamic calculations presented in the previous section. The compressive strengths for the four MK mortars are lower than for the P mortar, but significantly higher than for the L mortar. It is also seen that the ML mortar exhibits a better compressive strength than the L and M mortars, which indicates the presence of a synergetic effect between MK and LS as already reported by (Steenberg et al., 2011). Similar observations are also recorded for the MSL mortar, which exhibits a higher compressive strength than the L and MS mortars. 52


Figure 4.3. Compressive strengths for mortars with the same w/b = 0.5 and cured saturated in demineralized water for 91 days. (The reported data is actual compressive strengths determined on mini-RILEM mortar prisms). The data is from (Dai, 2015).

4.2.2 wPc’ – MT – LS and wPc’ – G – LS mortars In this section, the compressive strengths for the MT and G mortars are determined. With the purpose to reduce the time for the trial tests, we assumed a similar slow increase in compressive strengths from 28 days to 91 days for these mortars according to the previous data for the MK mortars. Thereby, the comparable strengths for the studied mortars are designed according to measurements after 28 days of curing. The final 91-day compressive strengths for the mortars produced with the wPc’ – MT – LS and wPc’ – G – LS blends are shown in Figure 4.4. As expected, similar comparable compressive strengths are obtained for all the studied mortars in this section. It should be noted that, each data in the figure is an average of two measurements, which are only taken on two halves of one mortar prism, and thus the data can only be used as an estimation.

Figure 4.4. Compressive strengths for mortars with designed comparable compressive strengths by varying the w/b and binder/sand ratios (b/s). 53


4.3 Pore structures 4.3.1 wPc – MK – SF – LS mortars 

Mercury intrusion porosimetry (MIP)

The MIP intrusion curves after the first and second intrusion cycles for the well hydrated mortars are shown in Figure 4.5. The samples were taken from the noncarbonated center of the mortars after 119 days of hydration (i.e., 91 days of curing in demineralized water and 28 days of CO2 exposure at 20 oC and 57% RH). Since the hydration and strength development of these mortars are very slow after 91 days of hydration, as mentioned before, the pore structures would not change with additional 28 days of exposure. The results show that the incorporation of MK in the mortars results in a refined microstructure with a lower threshold pore size (breakthrough pore size), as compared to that observed for the reference mortars (P and L). Partial substitution of MK with SF for the MSL and MS mortars does not make any difference in the pore structures compared to the ML and M mortars. Moreover, the total intruded porosities for the mortars, except for the L mortar, are almost identical as shown in Figure 4.6. The largest total intruded porosity is observed for the L mortar which also exhibits the highest threshold pore size. This is ascribed to a minor degree of reaction and a dilution effect of LS as mentioned above. The differences in threshold pore sizes reflect the poor and good pore connectivity for the four MK mortars and the L mortar, respectively.

Figure 4.5. Intrusion curves from the first and second MIP intrusion cycles for the wPc – MK – SF – LS mortars. The measurements were performed on samples prepared for carbonation investigations. The samples are taken from the center of the mortar bars after 91 days of saturated curing plus 28 days of carbonation. 54


Figure 4.6. Porosity and ink-bottle porosity derived from the first and second intrusion MIP curves for the wPc – MK – SF – LS mortars as shown in Figure 4.5. Ink-bottle pores are present in cement-based materials and they are connected to the external surface by smaller neck pores. These neck pore entrances may snap off the mercury during extrusion and for this reason two cycles of mercury intrusion/extrusion are used in the present work. All ink-bottle pores can be assumed to be filled at the beginning of the second intrusion cycle and the percolated (non-ink-bottle) pores will be filled during the second intrusion successively, according to Washburn’s equation linking pore radius to pressure (Kaufmann et al., 2009, Kaufmann, 2010). The volume of the ink-bottle pores and their access radius can be obtained by subtraction of the second intrusion (percolated pores) from the first intrusion (total pores) as shown in Figure 4.6. The results reveal that the incorporation of MK or a combination of MK and SF in the mortars results in a larger fraction of ink-bottle pores. 

Mass evolution measured by the PF method

The measurements of the mass evolution were performed on the mortar cylinders, which have already been hydrated for 91 days under saturated conditions. Two discs were taken from each cylinder, i.e. one close to the top of the cylinder and the other close to the bottom. The data were obtained following the procedure described for the PF method as summarized in chapter 3. The mass evolution curves of the mortars at different temperatures are shown in Figure 4.7. It is seen that water is continually removed when they are dried at 50 oC and 105 o

C. The weights of the mortars are stabilized at each temperature after a period of time.

The results show that the rate of weight loss is much higher for the P and L mortars than for the MK mortars. It is interesting that it takes nearly two months for the MK mortars 55


to dry out all the water from the coarse pores at 50 oC, whereas only 2 – 3 days are required for the P and L mortars. The results reveal that the pore structures of the mortars are highly refined by addition of MK or the combination of MK and SF, which is in very good agreement with the results obtained from the MIP measurements. Moreover, the results in Figure 4.7 also show the feature of inhomogeneity of the mortars, especially for the P and L mortars, where the samples from the top of the mortar cylinders have more coarse porosity (at 50 oC) and total porosity (at 105 oC) than the samples from the bottom. In addition, the air voids can also be evaluated through pressure saturation after the first 105 oC drying process. The fine porosity is obtained by subtracting the coarse porosity from the total porosity. The results are summarized in Figure 4.8. The total porosity is found to be twice the value measured by the MIP method, which is not surprising, because the drying method removes the water from the whole size range of pores, whereas the MIP method cannot determine the pores below 3 nm. Comparison of the different methods for characterization of the pore structures has been reviewed in the thesis by (Canut, 2012).

Figure 4.7. Normalized mass evolution for the wPc – MK – SF – LS mortars. Samples are extra mortar cylinders prepared for the chloride ingress tests. The zero time starts after 91 days of hydration under saturated conditions. The results in Figure 4.8 also show that the lowest total porosity is observed for the P mortar whereas it is very similar and higher for the MK mortars. However, this is not consistent with the total intruded porosities measured by MIP (Figure 4.6), which is explained by the very poor pore connectivity observed for the four MK mortars as shown in Figure 4.5 and Figure 4.7. In addition, the poor pore connectivity can be 56


explained by the expected higher gel porosity in the MK mortars, as compared to the P and L mortars, as shown in Figure 4.8.

Figure 4.8. Porosities derived from the water mass evolution curves in Figure 4.5. as determined by PF method. The second drying process was also conducted following the pressure saturation as shown in Figure 4.7, which might be relevant for comparison with the data presented in the earlier published studies, since at that time, total porosities were usually determined after pressure saturation. This procedure has been reported to be destructive to the pore structures of concrete as reviewed in thesis by (Canut, 2012). 

Conductivity measured with AC impedance

Disabilities such as chloride ingress and sulfate attack are highly affected by the ionic transport. The impact of the ionic interaction on transport of ions is minor in the coarse pores, but it may be amplified in the fine pores, according to the theory of the electrical double layer (EDL). It was reported that the surface of hydrated cement paste with a pore diameter of less than around 7 nm would be completely covered with the diffuse layer because the Debye length calculated from the ionic strength of the pore solution is nearly 7 nm (Elakneswaran et al., 2010). This hypothesis is positively supported by the conductivity calculated from the measured bulk resistance of the mortars with the AC impedance method as shown in Figure 4.9, where the conductivity for the MK mortars are more than 10 times lower than that for the P and L mortars.

57


Figure 4.9. Conductivity of mortars calculated from the measured bulk resistance of mortars with the AC impedance method.

4.3.2 wPc’ – MT – LS and wPc’ – G – LS mortars The intrusion curves from the first and second MIP intrusion cycles for the MT and G mortars together with those for the reference P’ and L’ mortars are shown in Figure 4.10 and Figure 4.11. The results show no obvious differences in threshold pore radius for all mortars, only the pore radii for the MT-1 and G-1 mortars tend to be slightly smaller. A similar refinement of the pore structure, as observed for MK mortars, is not observed for the MT and G mortars, which is not surprising as it is a result of the very low reactivity of MT and G as mentioned above.

Figure 4.10. Intrusion curves from the 1st and 2nd MIP intrusion cycles for the wPc’ – MT – LS mortars after 91 days of hydration.

58


Figure 4.11. Intrusion curves from the 1st and 2nd MIP intrusion cycles for wPc’ – G – LS mortars after 91 days of hydration. The total pore volumes from the first and second MIP intrusion cycles together with the volume of ink-bottle pores are summarized in Figure 4.12. The results show no pronounced trends in total porosities (1st intrusion cycle), but the percolated pores (2nd intrusion cycle) seem to decrease with increasing the amount of MT or G. The inkbottle porosities for all MT and G mortars are similar and a bit higher than for the reference mortars (P’ and L’).

Figure 4.12. Pore volumes derived from the 1st and 2nd MIP intrusion curves for the wPc’ – MT – LS mortars after 91 days of hydration. The weight losses for the MT and G mortars together with the reference mortars (P’ and L’) dried at 50 oC are shown in Figure 4.13. The results show similar rates of weight loss upon drying, indicating similar a pore connectivity for the differenct types of 59


mortars. In addition, the MT mortars show higher coarse porosity than the P’ and L’ mortars, whereas no difference in coarse porosity is observed between the G and reference mortars.

Figure 4.13. Mass evolution curves for the wPc’ – MT – LS and wPc’ – G – LS mortars after 91 days of hydration.

4.4 Conclusions For the mortars produced with the same water/binder ratio (wPc – MK – SF – LS mortars): (1) The highest and lowest compressive strengths are observed for the P and L mortars, respectively. No significant differences in compressive strengths are observed for the four MK mortars. (2) The pore structures of the four MK mortars are significantly refined as a result of the substitution of wPc with MK and the combination of MK and SF. No difference in pore connectivity is observed for the four MK mortars. (3) According to the development of compressive strength and hydration degree, all the mortars are considered to be well hydrated. Hence influence of continuous hydration on accelerated durability tests in the following chapters will not be considered. For the mortars produced with comparable compressive strengths (wPc’ – MT – LS and wPc’ – G – LS mortars): (1) As expected, there are no significant differences in compressive strengths after 91 days of hydration as observed for all mortars.

60


(2) Comparable pore connectivities are observed for all mortars as revealed by the different methods. (3) Quantification of portlandite from the DTG curves shows large fractions of portlandite present in the MT and G mortars, indicating a very low pozzolanic reactivity of MT and G used in this study.

61


62

Chapter 5: Role of Friedel’s Salt on Chloride Binding for Mortars under Chloride Ingress

CHAPTER 5: ROLE OF FRIEDEL’S SALT ON CHLORIDE BINDING FOR MORTARS UNDER CHLORIDE INGRESS Chapter 5 analyzes the total chloride profiles and chemically bound chloride profiles (i.e. chloride in Friedel’s salt) for the wPc – MK – SF – LS mortars after hydration for 91 days and then exposed a 2.8 M NaCl solution for 35 days. The amount of Friedel’s salt is quantified using a novel method developed in this chapter based on thermogravimetric analysis (TGA). The role of Friedel’s salt on chloride binding is evaluated by comparing the total chloride profiles with the chloride profiles from Friedel’s salt.

63


5.1 Introduction The chloride resistance of a given concrete is commonly evaluated by the total chloride profiles (Tang et al., 2011, Bertolini et al., 2013, De Weerdt et al., 2014), though it is agreed that only free chloride is harmful to reinforced concrete. Among the types of chloride binding, it has been widely accepted that the main binding of chloride arise from the C3A or C4AF phases through formation of Friedel’s salt or its analogue as reviewed by (Justnes, 1998, Yuan et al., 2009, Galan and Glasser, 2015). Several studies (Glass and Buenfeld, 2000, Martın-Pérez et al., 2000, Yuan et al., 2009, Baroghel-Bouny et al., 2012, Bentz et al., 2013) have stated that the chloride binding may retard the chloride transport. (Martın-Pérez et al., 2000) stated that the chloride binding had an important effect on the rate of chloride ionic transport in concrete and on the corrosion initiation of the steel reinforcement. (Glass and Buenfeld, 2000) also stated that chloride binding may affect the rate of chloride ingress and the chloride threshold level which in turn determine the time to the chloride induced corrosion initiation. More specifically, (Yuan et al., 2009) even mentioned that formation of Friedel’s salt can result in a less porous structure and hence slow down the transport of chloride ions. However, these statements are mainly based on model calculations and speculations, because it is difficult to make an experiment which can directly compare the chloride ion diffusion rates with and without chloride binding. Meanwhile it is also difficult to eliminate the impact of differences in pore structures on the chloride ion transport. On contrary, (Loser et al., 2010) reported that only a limited fraction of chloride is bound by the cement hydrates, whereas a larger fraction of chloride is in the pore solution. In this chapter, the chloride resistance of mortars made from wPc – MK – SF – LS blends is investigated. In order to understand the role of Friedel’s salt on chloride binding, the total chloride profiles are compared with chemically bound chloride profiles from Friedel’s salt quantified using TGA.

5.2 Experimental The materials, mix design and mortar preparations for the wPc – MK – SF – LS blends are described in chapter 3 (section 3.1). In order to eliminate the impact of continuous hydration on chloride ingress, the prepared mortars are cured saturated for 91 days and they are known to be well-hydrated according to chapter 4. The TGA and XRD techniques are used for the characterization of the studied mortars after chloride ingress, 64


and these methods are described in chapter 3 (see section 3.3.7 and section 3.3.8). The other methods used in this chapter are described as follows.

5.2.1 Total chloride profile analysis After 91 days of curing, three epoxy coated mortar cylinders from each mix were exposed to 700 ml 2.8 M NaCl solution in an airtight box for 35 days at 20 oC. The solutions were changed weekly in order to maintain an approximate constant chloride concentration. At the end of exposure, one of the three mortar cylinders was split along the chloride ingress direction. The chloride penetration depth was measured by spraying a 0.1 M AgNO3 solution on the fresh split surface of one half-cylinder. The remaining two full mortar cylinders and the one half-cylinder were profile ground with a layer thickness increasing from 1 to 4 mm. The total depth of grinding was determined based on the AgNO3 solution measured chloride penetration depth. The powder samples from each layer were collected for the total chloride profile analysis. The powders were firstly dried in an oven at 105 oC for 24 hours. The moisture content was calculated according to the mass loss after drying. The dried powders were dissolved in a 1:10 (v/v) HNO3 the solution at 80 oC. The volume of solution to be used to dissolve the paste was 20 ml for 2 g powder (from and close to the outer layer) and 40 ml for 4 g powder (for the inner layer), respectively. The powder-acid suspension was stirred and left to rest for 2 hours. The resulting solutions were subsequently filtrated, 15 ml filtrate was sampled, and 1 ml filtrate was titrated by a 0.01 M AgNO3 solution for samples from and close to the outer layer. For filtrates with low chloride concentration for samples from the inner layer, 10 ml of filtrate was titrated with acetic acid. The total chloride content was reported as mass percentage of the dried mortars at 105 oC. Base on the measured total chloride profiles, the effective diffusion coefficient (𝐷𝑒 ) and the surface chloride concentration (𝐶𝑠 ) can be obtained by fitting the total chloride profiles to the following equation (Build-NT-443), which is derived from Fick’s second law. 𝐶(𝑥, 𝑡) = 𝐶𝑠 − (𝐶𝑠 − 𝐶𝑖 ) ∙ erf(𝑥/√4 ∙ 𝐷𝑒 ∙ 𝑡) where, the total chloride concentration 𝐶(𝑥, 𝑡) is taken from the total chloride profiles, and 𝐶𝑖 is the initial chloride concentration of the mortars uncontaminated by chloride. The equation does not take into account the measured total chloride content on the surface of the mortars 65


5.2.2 Quantification of Friedel’s salt and portlandite by TGA TGA has been used to identify Friedel’s salt in cement pastes or mortars in several studies (Jain and Neithalath, 2010, De Weerdt et al., 2015, Geng et al., 2015). However, only one reference (Jain and Neithalath, 2010) was found to use TGA for the quantification of Friedel’s salt in Portland cement-based pastes. Unfortunately, the methodology for the quantification was not described. The first derivative of the TG curve (i.e. DTG) for synthesized Friedel’s salt, as reported in several other studies (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013, Lothenbach et al., 2016), is shown in Figure 5.1. The thermal decomposition of Friedel’s salt results in two major DTG peaks in the temperature ranges of 30 ~ 180 oC and 180 ~ 450 oC. It should be noted that two well-defined single peaks were observed by (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013) rather than a splitting of the second peak (180 ~ 450 oC) as observed by (Lothenbach et al., 2016). The reason for this discrepancy is not clear. Nonetheless, the weight losses associated with the lower and higher temperature intervals have the same ratio of 4:6 for the different studies. The fixed ratio between the two weight losses is ascribed to the composition and layered structure of Friedel’s salt, which contains positively charged main layers [Ca2Al(OH)6]+, comprising six hydroxyl groups and negatively charged interlayers [2Cl-,4H2O], containing four water molecules (Grishchenko et al., 2013). The obtained ratio of 4:6 allows for a quantification of Friedel’s salt in chloride contaminated Portland-cement based materials, using the weight loss of the main layer water in Fredel’s salt.

Figure 5.1. Characterization of synthesized Friedel’s salt using TGA (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013, Lothenbach et al., 2016). 66


Thermogravimetric analysis (TGA) was performed on powdered mortar samples which were used for the total chloride analysis according to the method described in chapter 3 (section 3.3.7). The weight loss and the quantified data were reported as mass percentage of dried mortars at 105 oC. According to the ratio (4:6) between the interlayer water and the main layer water, determined in the synthesized samples as mentioned above, the content of Friedel’s salt in the mortars can be quantified by firstly determining the content of the main layer water (230 oC ~ 410 oC) in Friedel's salt. Figure 5.2 shows the DTG curves for powdered mortar samples, obtained at a 1 ~ 2 mm depth, in the non-contaminated region (e.g. 8 ~ 12 mm), and for synthesized Friedel’s salt. The quantification was performed by drawing a baseline on the DTG curve for the chloride contaminated sample, which is related to the baseline for the reference sample non-contaminated by chloride. The area marked on the DTG curve of the chloride contaminated sample is then integrated, i.e. 0.12 wt%. The content of Friedel’s salt in the mortar can thus be calculated according to the following equation: 𝑚𝐹𝑠 =

𝑀𝐹𝑠 561.32 × 𝑚𝐻 = × 𝑚𝐻 (6 × 𝑀𝐻 )𝜇 (6 × 18.015)𝜇

Where mFs is the mass fraction of Friedel’s salt in the mortar (wt. %), mH is the TGA measured loss of water from the main layer (wt. %), MFs is the molar mass (561.32 g/mol) of Friedel’s salt with the chemical composition of Ca4Al2Cl2(OH)12·4H2O and MH is the molar mass (18.015 g/mol) for H2O. Since the temperature range (230 ~ 410 o

C) for the decomposition of the main layer water of Friedel’s salt in the mortar samples

does not fully account for the whole temperature range (180 ~ 450 oC) for the decomposition of the main layer water determined in synthesized Friedel’s salt as shown in Figure 5.2, a correction coefficient, 𝜇, is suggested to take into account the underestimation of the main layer water of Friedel’s salt in mortar samples, which can be expressed as: 𝜇=

𝑚230~410 × 100 𝑚180~450

where, 𝑚230~410 is the partial main layer water of synthesized Friedel’s salt in the temperature range 230 ~ 410 oC. This temperature range is determined according to the temperature range of main layer water of Friedel’s salt, observed in mortar samples in Figure 5.2. The 𝑚180~450 accounts for all the main layer water, which can be determined in the temperature range of 180 ~ 450 oC in the DTG curve for synthesized 67


Friedel’s salt. Therefore, in this study, about 80% of the main layer water of Friedel’s is able to be determined in mortar samples (i.e., 𝜇 = 80%). Similarly, the portlandite content is also quantified by determining the weight loss in the temperature range 400 – 500 oC. The portlandite content t is calculated according to the equation: 𝑚𝐶𝐻 =

𝑀𝐶𝐻 74.09 ×𝑚 = ×𝑚 𝑀𝐻 18.015

where mCH is the mass fraction of portlandite in the mortar (wt.%), m is the loss of water from portlandite (wt. %), MCH is the molar mass (74.09 g/mol) of portlandite in the form of Ca(OH)2, and MH is the molar mass (18.01 g/mol) of H2O.

Figure 5.2. Quantification of Friedel’s salt in a chloride exposed mortar cylinder (P) using TGA. The chloride contaminated samples was obtained at a mortar depth of 1~2 mm, whereas the uncontaminated reference sample is obtained from the same mortar at a depth of 8~12 mm.

5.3 Results and discussion 5.3.1 Total chloride profiles The total chloride profiles for all mortars exposed to the 2.8 M NaCl solution for 35 days are shown in Figure 5.3. The total chloride contents decrease with increasing chloride penetration depths. The L mortar shows the deepest chloride penetration depth and the highest total chloride content at any depth excluding the outer layer, whereas the MK mortars show the lowest chloride penetration depths. The total chloride profiles show no differences for the different MK mortars. The chloride diffusion coefficients (𝐷𝑒 ) derived from the total chloride profiles are shown in Figure 5.4. The results show substantially lower diffusion coefficients for the four MK mortars as compared the P and L mortars, indicating a very high chloride resistance for the MK mortars. 68


Figure 5.3. Total chloride profiles for the Portland cement-based mortars

Figure 5.4 Chloride diffusion coefficients obtained from the total chloride profiles

5.3.2 Friedel’s salt in mortars Figure 5.5 shows DTG curves for the samples taken at different depths of the mortars. Three major weight losses are observed for the samples not contaminated by chloride: C-S-H/AFm/AFt phases (30 ~ 300 oC), portlandite (400 ~ 520 oC) and CaCO3 (520 ~ 760 oC). When the samples are closer to the exposed surface, additional weight loss peaks (230 ~ 410 oC) caused by the decomposition of Friedel’s salt are detected. At the same time, another peak at a lower temperature range (below 150 oC), also associated with Friedel’s salt, is identified. Moreover, Figure 5.5 reveals that the formation of Friedel’s salt is accompanied by the consumption of the AFm phase, monocarbonate, in this study as predicted by thermodynamic modelling on the same blends in chapter 4. For the chloride non-contaminated samples at deeper depths, the DTG peak of monocarbonate is much clearer for the MK mortars than for the P and L mortars. This is explained by the increased formation of monocarbonate as a result of the additional alumina source from MK, as predicted by thermodynamic modelling. According to the method described in section 5.2.2, the amount of Friedel’s salt can be quantified. The Friedel’s salt profiles are plotted in Figure 5.6. The results show that the amount of Friedel’s salt decreases with increasing chloride penetration depth, excluding the outer layer. Higher amounts of Friedel’s salt are observed in the MK mortars (except for M mortar) as compared to P and L mortars. Friedel’s salts were observed at deeper depths for the P and L mortars as compared to the MK mortars, which is in agreement with the chloride ingress depths.

69


Figure 5.5. DTG curves of the chloride contaminated and non-contaminated samples from different depths of the Portland cement-based mortars. (Ettr. = Ettringite, Mc = Monocarbonate, Fs = Friedel’s salt, CH = Portlandite, Cc = CaCO3). Similarly, the quantified portlandite profiles are shown in Figure 5.7. Within a 2 mm depth from the exposed surface, a considerable decrease in the portlandite content is observed. This indicates that leaching has occurred during the chloride exposure. Figure 5.5 also shows the changes in the shape of the CaCO3 peak for the L mortar. It can be concluded that the changes in CaCO3 content within the range of 7.8 ~ 8.7 wt.% are too limited to justify the reduction in portlandite by carbonation. The phase assemblage for the samples from the outer layer has altered and is reflected by the reduction in the quantities of total chloride (Figure 5.3), Friedel’s salt (Figure 5.6), and portlandite (Figure 5.7). The Friedel’s salt profiles and portlandite profiles of the MK mortars

70


indicate an overlap between the chloride diffusion zone and the leaching zone. This complicates discussions of the chloride binding capacity of the MK blends.

Figure 5.6. Friedel’s salt (Fs) profiles of the Portland cement-based mortars, quantification of Fs using TGA.

Figure 5.7. Portlandite (CH) profiles for Portland cement-based mortars, quantification of CH using TGA.

5.3.3 X-ray diffraction analysis The diffraction patterns for the samples taken from 1~2 mm depths of the P and MSL mortars are shown in Figure 5.8. The recorded diffractograms are expected to be dominated by high-intense quartz reflections due to the presence of sand in the mortars. The reflections of Friedel’s salt and portlandite are observed at 11.4° and 18.3° 2θ respectively (Paul et al., 2015).

Figure 5.8. Quantification of Friedel’s salt by XRD on samples from the P and MSL mortars at 1 - 2 mm depths. (Fs = Friedel’s salt, CH = Portlandite). 71


Both phases have been quantified by Rietveld analysis and the results are summarized in Table 5.1. In general, the recorded diffractograms contain too strong reflections from sand to be properly analyzed. Therefore, the results can only be used as an estimation. It can be seen in Table 5.1 that a general agreement between XRD and TGA for the quantification of Friedel’s salt and portlandite is obtained. The results also show that the chloride-free AFm (e.g. monocarbonate) is hardly detected in the samples in Figure 5.8. Table 5.1. Summary of the amounts of Friedel’s salt and Portlandite quantified by TGA and XRD (wt.%) Friedel’s salt (wt.%)

Portlandite (wt.%)

Mortar

Sample depths

labels

(mm)

TGA

XRD

TGA

XRD

P

1~2

0.9 ± 0.1

0.8 ± 0.2

2.3 ± 0.3

2.2 ± 0.4

M

1~2

0.6 ± 0.1

1.4 ± 0.5*

< 0.1

< 0.2

MSL

1~2

1.7 ± 0.2

2.0 ± 0.3

< 0.1

0.5 ± 0.2

MS

1~2

1.2 ± 0.2

1.7 ± 0.3

< 0.1

< 0.2

* This sample does not contain anatase, and thus no quantitative Rietveld analysis can be performed. However, considering the intensity and width of the primary reflection from Friedel’s salt in this sample compared to the other three samples, 1.4 wt.% of Friedel’s in this sample can be estimated. Obviously, this value is not precise (± 0.5 wt.% at least).

5.3.4 Effect of total chloride content on formation of Friedel’s salt The Friedel’s salt contents as a function of the total chloride concentration for the P and L mortars are shown in Figure 5.9. The results show that the amount of Friedel’s salt formed in the L mortar is lower than in the P mortar for total chloride concentrations ranging from 0.6 wt.% to 1.1 wt.%. This can be explained by the dilution of C 3A by the incorporation of additional 31.9 wt.% limestone in the L mortars. However, the Friedel’s salt content is not as low as expected for the actual dilution. This indicates a higher binding capacity of Friedel’s salt per unit of C3A for the L mortar. This may be explained by the formation of additional AFm phases due to the increase in the degree of hydration, as presented in previous work (Dai, 2015). Figure 5.6 shows that the amount of Friedel’s salt seems to more or less at the same level from 2 mm to 5 and 10 mm depths for the P and L mortars, respectively. This indicates that the maximum chemical binding capacity in terms of formation of Friedel’s salt has been reached. The chloride ingress in the MK mortars is too limited to 72


detect such trends. Figure 5.6 shows generally that the incorporation of MK in Portland cement mortars increases the chloride binding in Friedel’s salt, possibly due to the additional alumina originating from the MK, as suggested in earlier studies (Yuan et al., 2009, Thomas et al., 2012, Ipavec et al., 2013). One exception is the lower amount of Friedel’s salt observed for the M mortar. However, this may be due to limited ingress and leaching. Leaching causes a reduction in the quantities of total chloride and Friedel’s salt. A reduction in the chloride content at the exposure surface has also been observed by others (Andrade et al., 1995, Jensen, 1999, De Weerdt et al., 2014). The chemical changes result in larger variations and therefore errors in the total chloride content as shown in Figure 5.3. Figure 5.9 also shows that Friedel’s salt is formed only after the total chloride content has reached a certain level. This is due to either a detection limit of the TGA method or that a critical total chloride content is needed for the initiation of the formation of Friedel’s salt. Galan and Glasser (Galan and Glasser, 2015) stated that a phase boundary has to be crossed before Friedel’s salt becomes stable and that the threshold chloride concentration to form Friedel’s salt varies due to different buffering actions of the AFm phases.

Figure 5.9. Effect of total chloride on the formation of Friedel’s salt for the Portland cement-based mortars P and L. Fs quantified using TGA.

5.3.5 Role of Friedel’s salt on chloride binding In order to highlight the role of Friedel’s salt on chloride binding, the bound chloride in Friedel’s salt and the total chloride profiles are plotted together in Figure 5.10 for the P and L mortars. It is found that the chloride bound in Friedel’s salt is about 0.1 wt.% for 73


both mortars, which is over 10 times less than the maximum total chloride content (about 1.12 wt% of mortar), excluding the surface layer data, revealing a weak contribution of Friedel’s salt to the binding of chloride. Based on the C3A content (7.8 wt.%) determined for the white Portland cement, the maximum amount of bound chloride in Friedel’s salt is calculated to be not more than about 0.45 wt.% of the P mortar, assuming that all C3A is converted into Friedel’s salt. Hence, only about 20% of all the C3A contributes to the formation of Friedel’s salt.

Figure 5.10. Comparison of the total chloride and chloride (Cl-) in Friedel’s salt (Fs) for the Portland cement-based mortars, P and L. Fs quantified using TGA. Y. Elakneswaran et al. (Elakneswaran et al., 2009) studied chloride binding isotherms of hydrated cement pastes, where the C3A and C4AF contents of the ordinary Portland cement were 9.67 wt.% and 8.82 wt.%, respectively. They found that the total bound chloride (by the equilibrium concentration technique) and bound chloride in Friedel’s salt (by XRD Rietveld analysis) are about 1.3 wt.% and 0.6 wt.%, when the sample was exposed to a 1 M NaCl solution. Such an amount of bound chloride in Friedel’s salt is about twice more than that measured in this study. However, it should be noted that a similar level of bound chloride, as observed in our study, was not observed in their work, which may underestimate the binding capacity of their cement.

5.3.6 Evaluation of the chloride resistance The minor contribution of Friedel’s salt on chloride binding for the P and L mortars tends to support the statement from (Loser et al., 2010) that the effect of chloride binding on chloride ingress is less important compared to the effect of permeability. On

74


this respect, the very high chloride resistance for the MK mortars is most likely dependent on the pore structures, because the amounts of Friedel’s salt, as shown in Figure 5.6, are not as high as observed for the P and L mortars. However, the total porosities for these mortars, presented in chapter 4 (section 4.3.1), do not make any relationship with the measured effective diffusion coefficients. A very good agreement is seen between the effective chloride diffusion coefficients and the pore connectivities for the studied mortars. This interpretation seems not to be strong enough, because it is very hard to image how the refined pore structures of the mortars perform as barriers which can strongly keep the chloride ions out of the mortar. An alternative explanation is that the ion-ion interaction in the ionic transport may play a significant role on chloride transport, especially when the pores are very fine (Elakneswaran et al., 2010). This explanation is nicely supported by the very low conductivity for the MK mortars as mentioned in chapter 4 (section 4.3.1).

5.4 Conclusions (1) A novel semi-quantitative method has been developed to quantify Friedel’s salt in chloride-exposed Portland cement-based mortars using TGA. The results agree well with those quantified by XRD Rietveld analysis. (2) The Friedel’s salt profiles have been obtained and compared with the total chloride profiles. The measured maximum chloride content in Friedel’s salt is found to be more than ten times lower than the highest total chloride content, indicating a low chemical contribution of Friedel’s salt to chloride binding. In addition, only 22% of the C3A seems to be available for the formation of Friedel’s salt. (3) For mortars with Portland cement only and the Portland cement – limestone blend, the amount of Friedel’s salt seems to reach a plateau from 2 mm to 5 and 10 mm depths, respectively. Close to the mortar surface, the amount decreases as a result of leaching. The plateau in the Friedel’s salt profiles indicates an exhausted chloride binding capacity in terms of formation of Friedel’s salt. The chloride ingress in mortars containing metakaolin was too limited to detect such a trend. (4) The highest chloride resistance is observed for the MK mortars, which is ascribed to a very low pore connectivity. No relationship between the total porosity and the effective chloride diffusion coefficient is obtained.

75


76

Chapter 6: Chloride Binding in Portland Cement – Metakaolin – Limestone Pates

CHAPTER 6: CHLORIDE BINDING IN PORTLAND CEMENT – METAKAOLIN – LIMESTONE PATES Chapter 6 investigates the chloride binding of the Portland cement – metakaolin – limestone pastes (P, ML and M). The chloride binding isotherms are determined for samples exposed to CaCl2 and NaCl solutions. The chloride bound in Friedel’s salt is quantified using the TGA-based approach developed in the chapter 5. Thermodynamic modeling is used to predict the phase assemblages and particularly to compare the relative quantities of the formed Friedel’s salt. Comparison of these results allows evaluation of the role of the C-S-H in chloride binding in hydrated Portland cement and in blended cements and the impact of cations.

77


6.1 Introduction Chloride binding in hydrated Portland cement has been widely studied. A major debate on the contribution of Portland cement hydrates to bind chloride arose from two statements about which of the phases Friedel’s salt and C-S-H dominate the chloride binding capacity. It has been claimed that the main binding of chloride arises from C3A or C4AF through formation of Friedel’s salt (3CaO·A12O3·CaCl2·10H2O) or its iron analogue as reviewed by (Justnes, 1998, Yuan et al., 2009, Galan and Glasser, 2015). This statement is usually drawn based on a simple relationship between the amount of the total bound chloride, determined by conducting the equilibrium experiment proposed by (Luping and Nilsson, 1993), and the amount of Al2O3 according to the chemical composition of the starting materials. However, (Luping and Nilsson, 1993) reported that the main binding of chloride for Portland cement is ascribed to adsorption of chloride in the C-S-H phase. None of the two statements were proposed based on a direct quantification of chloride bound in Friedel’s salt or adsorbed on the C-S-H. Chloride binding in hydrated Portland cement – SCMs blends has also been investigated. Substitution of Portland cement with alumina-rich SCMs has been largely reported to increase the chloride binding capacity, which is mainly attributed to the formation of Friedel’s salt (Luo et al., 2003, Thomas et al., 2012). In contrast, addition of silica fume in blended cement was reported to decrease the chloride binding (Arya et al., 1990, Arya and Xu, 1995, Thomas et al., 2012). Few published studies have focused on the chloride binding in the C-S-H in blended cements. (Zibara et al., 2008) studied the Ca/Si and Ca/Al ratios of the hydration products on the chloride binding capacity of lime-silica fume and lime-metakaolin mixtures. The authors concluded that the chloride is primarily bound in Friedel’s salt in presence of alumina. At a low Ca/Al ratio, the initially formed stratlingite converted to Friedel’ salt but at a much lower rate as compared to the conversion of monocarbonate formed at high Ca/Al ratios. In the absence of alumina, the chloride binding in C-S-H was obtained and found to increase with increasing Ca/Si ratio. A TGA based method for quantification of Friedel’s salt has been introduced in the chapter 5 in this thesis, from which the amount of bound chloride in Friedel’s salt can be quantified. However, the total bound chloride or the free chloride in the pore solution was not determined in chapter 5, and thereby the role of the C-S-H on chloride binding in the studied mortars was not obtained. Moreover, the chloride ingress depths for the

78


MK mortars were too shallow and the chloride diffusion region was completely overlapped with the leaching region, which may cause an underestimation of the chloride binding capacity of the MK mortars. In this chapter, the chloride binding in the P, ML and M pastes is investigated. The total bound chloride is determined by the equilibrium experiment, and the bound chloride in Friedel’s salt is estimated with the TGA method. The relative quantities and types of phases in the hydrated Portland cement – metakaolin – limestone pastes after chloride exposure are also predicted by thermodynamic modeling.

6.2 Experimental The experimental setup including the sample preparation, exposure, and determination of the chloride binding isotherms is based on (De Weerdt et al., 2014). The preparation of the P, ML and M pastes used in this chapter is described in chapter 3 (section 3.1.4). The laboratory grade salts NaCl and CaCl2·6H2O, together with distilled water, were used to prepare the exposure solutions with the following chloride concentrations: 0 (reference), 0.125, 0.25, 0.5, 1 and 2 mol/L. The actual concentrations were checked by titration and further used for the calculations.

6.2.1 Chloride exposure For the chloride exposure, 30.0 g of the hydrated cement paste (w/b = 0.95) was weighed into a 45 mL plastic centrifuge bottle and 15 mL of the chloride solution was added. As reference, a sample exposed to the same amount of distilled water was also included. The samples were shaken regularly, sealed and stored at 20 oC for 2 months prior to the analysis.

6.2.2 Determination of the chloride binding isotherms The free water content of the wet pastes (w/b=0.95) was determined using thermogravimetric analysis (TGA) prior to the chloride exposure. Approximately 400 mg of the wet paste was dried at 105 oC in the TGA equipment and purged with N2 for 4 h during which the weight of sample stabilized. The measured free water contents, i.e. the weight loss at 105 oC, were 36.6 wt.% (P), 39.9 wt.% (ML) and 39.1 wt.% (M) of the initial weight of the corresponding wet pastes.

79


The samples were shaken and subsequently centrifuged after 2 months of chloride exposure. The pH and the Cl content of the clear supernatant were determined with the same techniques as described in (De Weerdt et al., 2014). The detailed calculations of the chloride binding isotherms is also given in (De Weerdt et al., 2014).

6.2.3 Quantification of Friedel’s salt by TGA The chloride exposed samples were also investigated by TGA. For comparison, the sample stored in distilled water was also investigated. Approximately 300 mg of the sample was dried at 105 oC in the TGA equipment purged with N2 for 4 h, followed by heating up to 950 oC at rate 10 oC/min. The Friedel’s salt in chloride exposed mortar samples has been characterized by the first derivative of the TGA curve (DTG) in chapter 5. The results show that Friedel’s salt had two major dissociation peaks in the temperature ranges 100 – 150 oC and 200 – 400 oC, which agreed with the temperature range 30 – 236 oC and 236 – 403 oC determined for synthesized Friedel’s salt (BirninYauri and Glasser, 1998, Grishchenko et al., 2013, Lothenbach et al., 2016). The two DTG peaks reflect the release of water from the interlayer and main layer spaces of Friedel’s salt, which have a fixed water ratio (4:6). The weight loss due to the release of six main-layer water molecules can easily be quantified, since the DTG peak is nonoverlapping with the weight loss from other phases in the hydrated cements. More details about the quantification of Friedel’s salt with TGA can be found in chapter 5.

6.3 Results 6.3.1 Chloride binding isotherms The chloride binding isotherms for the well hydrated P, ML and M pastes exposed to NaCl and CaCl2 solutions of varying chloride concentrations are given in Figure 6.1. In general, the amount of the bound chloride increases with increasing chloride concentration. It is also seen that the amount of bound chloride in the P paste exposed to the NaCl solution is at the same level when the chloride concentration is over 1.0 mol/L. This indicates that the maximum chloride binding capacity for the P paste in the NaCl solution is reached. The results in Figure 6.1 also show that a partial substitution of the wPc with MK or MK/LS increases the chloride binding as compared to the P paste when they are exposed to same type of chloride solution. This observation is consistent with those 80


from earlier published studies (Coleman and Page, 1997, Thomas et al., 2012). Moreover, there is no major difference in chloride binding between the ML and M pastes, although LS can react with alumina from the MK, and subsequently form more monocarbonate (Antoni et al., 2012, Vance et al., 2013, Dai, 2015). In addition, a higher chloride binding is observed for the pastes exposed to the CaCl2 solution than to the NaCl solution, which has also been reported in several other studies (Tritthart, 1989, Arya et al., 1990, Wowra et al., 1997, De Weerdt et al., 2015). This observation implies that calcium in the exposure solution may play an important role in enhancing the chloride binding capacity.

Figure 6.1. Chloride binding isotherms for the well hydrated P, ML and M pastes exposed to NaCl (left) and CaCl2 (right) solutions.

6.3.2 Chloride in Friedel’s salt as measured by TGA The formation of Friedel’s salt through the conversion of the AFm phases in the studied pastes after exposure to NaCl and CaCl2 solutions is analysed by TGA, where the DTG curves are shown in Figure 6.2. The results show that the intensity of the DTG peaks representing the principal-layer water (280 – 405 oC) in Friedel’s salt increases with increasing the chloride concentration of the NaCl and CaCl2 solutions. The change of this DTG peak is accompanied by the changes in the AFm phases, observed at lower temperature around 200 oC in the DTG curves. According to the method for quantification of Friedel’s salt, as proposed in chapter 5, the chloride binding isotherm with respect to the chloride bound in Friedel’s salt can be obtained for samples exposed to the NaCl and CaCl2 solutions as shown in Figure 6.3.

81


Figure 6.2. DTG curves of the pastes exposed to the NaCl and CaCl2 solutions with varing chloride concentrations. It is reasonable that the quantified amount of chloride in Friedel’s salt is lower than the total amount of bound chloride. However, it should be noted that the baseline of the DTG curves in Figure 6.2 for the pastes containing Friedel’s salt is not as horizontal as those for the mortar samples presented in chapter 5 (see Figure 5.5). In addition, the temperature range representing the weight loss of the main-layer water of Friedel’s salt seems to extend to a temperature lower than 280 oC and thus cause some overlap with weight loss from the other AFm phases, i.e. mainly monocarbonate. Thereby, the quantification the Friedel’s salt based on the weight loss from 280 – 405 oC 82


underestimates its amount formed in the studied pastes. However, it is believed that the evaluation of on the trend of chloride adsorption and the comparison of the relative amounts are not a problem. The results in Figure 6.3 show that Friedel’s salt is formed when the chloride concentration reaches a certain level and then it increases with increasing chloride concentration. It is expected that more Friedel’s salt is formed for the ML and M pastes than for the P paste at high chloride concentrations. Moreover, it is interesting that more Friedel’s salt is observed to form in the ML and M pastes when they are exposed to CaCl2, whereas there is no difference in the Friedel’s salt content in the P pastes exposed to the NaCl and CaCl2 solutions.

Figure 6.3. Chloride binding isotherms with respect to bound chloride in Friedel’s salt for all the pastes exposed to NaCl (solid line) and CaCl2 (dash line) solutions. Comparison between the chloride bound in Friedel’s salt and the amount of the total bound chloride is presented in Figure 6.4 for the samples exposed to the NaCl and CaCl2 solutions. It is seen in Figure 6.4a that at least 50% of the chloride is bound in Friedel’s salt for the ML and M pastes when they were exposed to 2.0 mol/L NaCl solution, whereas nearly 100% of the chloride is bound in Friedel’s salt for the P paste exposed to 2.0 mol/L NaCl solution. The results (Figure 6.4a) reveal a dominant contribution of Friedel’s salt on the chloride binding in the case of NaCl exposure. However, the difference between the bound chloride content in Friedel’s salt and the total bound chloride content is enlarged with increasing chloride concentration when the samples are exposed to the CaCl2 solution as compared to the NaCl exposure. This observation is particularly evident for the P paste exposed to the CaCl2 solution as 83


shown in Figure 6.4(b). In addition to the formation of Friedel’s salt and the changes in the AFm phases, as identified in Figure 6.2, a minor decrease in the portlandite content is also diretly observed for the ML and M pastes. It is seen that the minor amounts of portandite present in these samples are depleted at higher chloride concentrations. The variations in the portlandite content in the P pastes are presented in Figure 6.5 based on a TGA quantification. Similar observations on the reduction of the portlandite content are found for the P pastes exposed to both the NaCl and CaCl2 solutions, however the overall decrease for both of them is clearly not siginificant.

Figure 6.4. Chloride binding isotherms with respect to the total bound chloride (solid line, as in Figure 6.1) and chloride bound in Friedel’s salt (dash line) for all the pastes exposed to the (a) NaCl and (b) CaCl2 solutions.

Figure 6.5. Portlandite content for the P pastes exposed to NaCl (solid line) and CaCl2 (dash line) solutions. 84


6.3.3 Thermodynamic modeling Thermodynamic modelling is employed to predict the phase assemblages for the studied pastes exposed to the NaCl and CaCl2 solutions as shown in Figure 6.6. It is predicted that the major hydrates in the P paste before chloride exposure are C-S-H, ettringite, portlandite, monocarbonate, calcite and hydrotalcite. Similar hydrates are predicted for the ML and M pastes but in the absence of portlandite as a result of the pozzolanic reactions. The predicted phase assemblages are in good agreement with the experimental data presented by (Dai, 2015). When the NaCl and CaCl2 solutions are added in the hydrated system with increasing chloride concentration, thermodynamic modelling predicts that a conversion of monocarbonate to Friedel’s salt occurs after the chloride concentration reaches a certain level for the P, ML and M samples. However, the conversion is facilitated for sample exposed to the CaCl2 solution as compared to the NaCl solution, which indicates that Ca plays a significant role for promoting the formation of Friedel’s salt possibly by accommodating the released carbonate ions from the monocarbonate phase. Thermodynamic modelling also predicts that a plateau of Friedel’s salt is observed for the P paste both in the NaCl and CaCl2 solutions. The maximum amount of Friedel’s salt is predicted to be similar regardless of the salt type, which is in good agreement with the TGA data (Figure 6.3). For the ML and M pastes, the plateau of Friedel’s salt is only observed for samples exposed to the CaCl2 solution, whereas the Friedel’s salt content keeps increasing for samples exposed to the NaCl solution. This observation is also in good agreement with the TGA quantification (Figure 6.3).

85


Figure 6.6. Phase assemblages for the P, ML and M pastes exposed to the NaCl (left) and CaCl2 (right) solutions.

6.3.4 pH of the exposure solutions The measured pH of the supernatants of the exposure solutions for the P, ML and M pastes are shown in Figure 6.7. The results show a lower pH for the ML and M pastes than for the P paste as already known for hydrated blended cements (Lothenbach et al., 2011). It is seen that there is minor increase in pH when the pastes are exposed to the NaCl solutions as compared to the pastes exposed to a similar amount of distilled water, which has also been observed by (De Weerdt et al., 2015). In line with the previous study (De Weerdt et al., 2015), no major change in pH with increasing chloride concentrations is observed for all the pastes exposed to the NaCl solution. However, a substantial decrease in pH is observed with increasing chloride concentrations when the 86


pastes are exposed to the CaCl2 solution as also reported by (Tritthart, 1989, Zibara, 2001, De Weerdt et al., 2015).

Figure 6.7. pH as a function of the chloride concentrations of the exposure solutions for the pastes exposed to NaCl and CaCl2 solutions.

6.4 Discussion The discussion on the chloride binding is based on a general agreement from published studies that the major phases binding chloride are Friedel’s salt and the C-S-H. The possible minor uptake of chloride ions by other phases (e.g. ettringite and portlandite as reported in the literature (Elakneswaran et al., 2009)) and the possible physical adsorption of chloride ions on the positively charged surface of Friedel’s salt (Elakneswaran et al., 2009) are considered to be negligible. Based on this fact, the amount of chloride bound in the C-S-H phase (𝐶𝑙𝐶−𝑆−𝐻 ) in principle can be calculated from the total bound chloride (𝐶𝑙total ) and the chloride content in Friedel’s salt (𝐶𝑙𝐹𝑠 ) through the equation: 𝐶𝑙𝐶−𝑆−𝐻 = 𝐶𝑙total − 𝐶𝑙𝐹𝑠 . However, since the TGA method used for quantification of Friedel’s salt underestimates the real amount of bound chloride in Friedel’s salt, the role of the C-S-H on the chloride binding will only be discussed qualitatively at this stage.

6.4.1 Chloride binding in hydrated Portland cement The maximum chloride binding capacity, as observed for the P paste (Figure 6.1a) exposed to the NaCl solution, is found to be determined by Friedel’s salt according to the TGA quantification (Figure 6.3) and the thermodynamic calculations (Figure 6.6a), 87


which is due to a smaller amount of monocarbonate initially present in the hydrated Portland cement (Figure 6.6a). Considering the underestimation of the quantified Friedel’s salt content and thus eventually small differences between the bound chloride in Friedel’s salt and the total bound chloride as shown in Figure 6.4a, it can be deduced that there is only a minor or no uptake of chloride in the C-S-H for the P samples exposed to the NaCl solutions. In contrast, a continuous increase of the chloride binding in the P paste with increasing CaCl2 concentration (Figure 6.1b) suggests that the increased chloride binding in the P paste exposed to the CaCl2 solution is dominated by chloride binding in the C-S-H phase, because the amount of chloride in Friedel’s salt is stabilized at low chloride concentrations as seen by the TGA quantification (Figure 6.3) and the thermodynamic calculations (Figure 6. 6b). These two conclusions reconcile the discrepancies presented in the published studies for the past decades about which phase (C-S-H or Friedel’s salt) that dominates the chloride binding capacity. The minor or no uptake of chloride in the C-S-H for tne P samples exposed to the NaCl solution observed in this study is also supported by the latest published studies on chloride adsorption in synthetic C-S-H samples (Plusquellec et al., 2012). The authors found that chloride uptake only occurs in the presence of sufficient Ca in the exposure solution and that adsorption of two chloride ions is accompanied by one Ca. (Luping and Nilsson, 1993) reported a dominant contribution of the C-S-H phase on the chloride binding for the hydrated Portland cement. However, it should be noted that their samples were exposed to a NaCl solution saturated with Ca(OH)2.

6.4.2 Chloride binding in blended cements Both the TGA quantification (Figure 6.3) and the thermodynamic calculations (Figure 6.6) show that the higher chloride binding for the ML and M pastes as compared to the P paste is ascribed to the formation of more Friedel’s salt. It is not caused by the possible adsorption of chloride in the C-S-H, because it is known that the Ca/Si ratio is lower for C-S-H in hydrated blended cements than in pure hydrated Portland cement (Lothenbach et al., 2011). Furthermore, it is known that a C-S-H with a lower Ca/Si ratio has a lower chloride binding capacity (Zibara et al., 2008, Plusquellec et al., 2012). Similar to the P paste, the ML and M pastes also exhibit higher chloride binding in the CaCl2 solution than in the NaCl solutions, which is partially attributed to the formation of Friedel’s salt as shown by thermodynamic calculations in Figure 6.6. (Ipavec et al., 2013) studied the effect of alkalinity on the chloride binding in the hydrated blended 88


cements and found that the presence of alkalis hindered the formation of chloroaluminate phases and decreased the chloride binding capacity. Besides, the possible enhanced chloride binding in the C-S-H in the ML and M pastes exposed to the CaCl2 solution may also contribute partly to the increased total chloride binding, because it is found that calcium can deprotonate the silanol sites in the C-S-H and overcharge the initially negatively charged surface of the C-S-H, and thus result in adsorption of chloride (Plusquellec et al., 2012, Plusquellec et al., 2013).

6.4.3 Impact of pH on chloride binding Addition of CaCl2 leads to a pH decrease as Ca is incorporated in the interlayer of the C-S-H phase by replacing protons (as Ca(OH)2) (Labbez et al., 2011, Plusquellec et al., 2012, Plusquellec et al., 2013, Lothenbach and Nonat, 2015): 2≡SiOH + Ca2+ => (≡SiO)2Ca + 2H+. The authors (Plusquellec et al., 2013) also found that no portlandite is observed from an XRD analysis after addition of CaCl2 into the C-S-H suspension with a Ca/Si = 0.8, which further confirmed that Ca ions in the solution were absorbed in the C-S-H phase rather than resulting in the formation of portlandite. There is no change in pH for the conversion of monocarbonate to Friedel’s salt, according to the reaction: CaCl2 + 3CaO·Al2O3·CaCO3·11H2O => 3CaO·Al2O3·CaCl2·10H2O + CaCO3 + H2O. Although the phase assemblages for the P pastes in the NaCl and CaCl2 solutions are constant, as predicted by thermodynamic modelling (Figure 6.6), a difference in the pH is observed for the two solutions. This may be explained by the following considerations: (i) For samples exposed to the CaCl2 solution, the presence of CaCl2 inhibits the Ca solubility of the C-S-H (same to portlandite) as a result of the higher ionic strength, thus no or only a minor Ca buffer effect to solution is expected from the C-S-H (same to portlandite). However, Ca can deprotonate Si-OH sites in the C-S-H, thus more H+ is released to the solution, which results in a lowering of pH. (ii) For samples exposed to the NaCl solutions, addition of NaCl increases the Ca solubility of the C-S-H and portlandite (Glasser et al., 2005). The released Ca is used for the AFm conversion to Friedel’s salt to accommodate the released carbonate, and thus, the initially Ca concentration in the exposure solution decreases until depletion of monocarbonate and then the Ca concentration starts to increase. The relationship between the bound chloride and the pH of the exposure solutions for all the pastes exposed to the NaCl and CaCl2 solutions are shown in Figure 6.2. Since the changes in pH of the NaCl exposure solutions are minor, the increased chloride 89


binding for the pastes exposed to the NaCl solutions is mainly driven by the formation of Friedel’s salt. Obviously, it is seen that the chloride binding for the pastes exposed to the CaCl2 solution is highly pH dependent with high correlation coefficients. Similar observations have been reported in earlier studies (Tritthart, 1989, Zibara, 2001, De Weerdt et al., 2015). (De Weerdt et al., 2015) studied the impact of the associated cations on chloride binding in hydrated Portland cement, and the authors concluded that the measured change in chloride binding depending on the cation type was mainly governed by the pH of the exposure solution and thereby the binding capacity of the CS-H. However, this conclusion does not fully agree with the results presented in Figure 6.8, especially for blended cements, because the increased chloride binding for the samples exposed to CaCl2 is mainly ascribed to the formation of Friedel’s salt as seen by the TGA quantification and thermodynamic modelling. It seems that the impact of cations on chloride binding is driven by the Ca concentration as proposed recently by (Plusquellec et al., 2012, Plusquellec et al., 2013). Thus, the pH dependency of the chloride binding in the case of CaCl2 exposure (Figure 6.8) may be explained by the intrinsic link between calcium concentration and pH of the pore solution.

Figure 6.8. Relationship between the bound chloride and the pH of the exposure solutions for all the pastes exposed to the NaCl and CaCl2 solutions.

6.5 Conclusions The chloride binding of the Portland cement – metakaolin – limestone blends exposed to different concentrations of NaCl and CaCl2 solutions has been investigated in this chapter. Based on the results and discussion, the following conclusions can be drawn: 90


(1) The use of metakaolin as an SCM increases the chloride-binding capacity compared to the P sample and to a similar extent for both the ML and M samples, regardless of the addition of limestone. This is mainly attributed to the formation of a larger quantity of Friedel’s salt in the ML and M samples rather than to the formation of more C-S-H with a lower Ca/Si ratio, as compared to the C-S-H in the P sample, since a lower Ca/Si ratio of the C-S-H phase is known to lower the chloride binding capacity (Plusquellec et al., 2012). (2) The chloride binding increases significantly for the samples exposed to a CaCl2 solution rather than to a NaCl solution. For the P sample, this is ascribed to the higher Ca/Si ratio of the C-S-H for samples exposed to the CaCl2 solution, since the amount of Friedel’s salt is independent on the types of cations according to the TGA data and the thermodynamic calculations. However, the thermodynamic modeling clearly shows that the increased chloride binding for the ML and M samples exposed to CaCl2 is ascribed to the formation of a larger amount of Friedel’s salt. Moreover, the higher Ca/Si ratio of the C-S-H for the ML and M samples exposed to the CaCl2 solution may also contribute to an increased chloride binding. (3) The pH is found to decrease with increasing CaCl2 concentration, whereas no major changes are observed with increasing NaCl concentration, in accordance with other studies (Plusquellec et al., 2012, De Weerdt et al., 2015). An earlier study of chloride binding in hydrated Portland cement samples (De Weerdt et al., 2015) has suggested that (i) the addition of CaCl2 can cause an increase in the Ca/Si ratio of C-S-H, which may lead to an increase in the adsorption of OH– ions on the C-S-H and hence a lowering of the pH; (ii) the measured change in chloride binding for different cations is mainly governed by the pH of the exposure solution and thereby the binding capacity of the C-S-H. However, this may be different for blended cements, since (i) no major changes in the Ca/Si ratio are predicted with increasing CaCl2 concentration. (ii) the increase in chloride binding for the ML and M samples with increasing CaCl2 concentration is mainly ascribed to formation of different amounts of Friedel’s salt according to the TGA data and thermodynamic modelling.

91


92

Chapter 7: Impact of Carbonation on the Microstructure and Chemistry of Mortars

CHAPTER 7: IMPACT OF CARBONATION ON THE MICROSTRUCTURE AND CHEMISTRY OF MORTARS Chapter 7 investigates the carbonation performance of the wPc – MK – SF – LS mortars after hydration for 91 days and exposure to 1% (v/v) CO2 at 20 oC / 57% RH for 280 days. The impact of carbonation on the microstructure and chemistry of the mortars are evaluated for pursuing the decisive parameter(s) that control the carbonation rates of the studied mortars.

93


7.1 Introduction It is generally accepted that addition of SCMs increases the risk of corrosion with respect to carbonation for reinforced concrete as reviewed in chapter 2. However, most of these studies focused on fly ash and slag, and only a few of them involved calcined clay blends (Antoni, 2013, Bucher et al., 2015). This chapter investigates the carbonation performance of the Portland cement – metakaolin – silica fume – limestone mortars. The study of the carbonation of these mortars, which have distinct phase assemblages and microstructures as already described in chapter 4, is conducted with the aim to form the basis for a better understanding of the carbonation mechanism for general Portland cement – SCM blends.

7.2 Experimental The carbonation performance of the wPc – MK – SF – LS mortars is investigated in this chapter. All the studied mortars are well hydrated for 91 days before CO2 exposure. The MIP and SEM methods, as described in chapter 3, are used to investigate the changes in microstructures of the mortars. The TGA and thermodynamic modeling described in the same chapter are employed to analyze the CO2 binding capacity of the studied blends.

7.2.1 CO2 exposure After 91 days of hydration under the saturated condition as described in chapter 3 (section 3.1.3), the mortar bars were exposed to a controlled atmosphere with 1% (v/v) CO2 in an incubator at 20 oC. The CO2 concentration of 1% (v/v) was chosen since the formed phases are expected to be the same as those formed during natural carbonation (0.035 % (v/v)), as indicated in an earlier study (Castellote et al., 2009). The relative humidity (RH) was 57 ± 1%, which matches the RH range of ~ 40 – 70% that maximizes the carbonation rate (Thiery et al., 2007).

7.2.2 Phenolphthalein spray method Carbonation depths were measured after 0, 7, 14, 21, 28, 56, 91 and 280 days of CO 2 exposure. Slices with a thickness of 15 – 20 mm were taken from the mortar bars by splitting, and the fresh surface was sprayed with a 1 wt.% phenolphthalein aqueous solution. Photographs of the sprayed slices were taken with a Sony camera (SLT-A65V) and the carbonation depths were measured on the photographs after the color had stabilized on the slices as demonstrated in Figure 7.1, using the commercial software 94


“GetData Graph Digitizer”. Each depth reported in this study is the average of 10 to 15 measurements excluding the corners. In addition, the top and bottom surfaces were omitted to minimize effects from different surface qualities and mortar inhomogeneities originating from the casting.

Figure 7.1. Illustration of the color change (0 – 66 s) after spaying the M mortar with phenolphthalein.

7.3 Results and discussion 7.3.1 Apparent carbonation depths The apparent carbonation depths, as measured by the phenolphthalein spray method, are shown as a function of the square root of time in Figure 7.2. No initial carbonation is observed for any of the mortars in accordance with the mortars being kept saturated prior to exposure. A measurable degree of carbonation is observed already after 7 days of exposure to 1% CO2 at 20 oC and 57% RH for the blended mortars but not for the P mortar. The reference mortar (P) shows a very slow progress in carbonation and thereby the highest resistance to carbonation whereas the L mortar, containing no MK and SF, is most vulnerable to carbonation. None of the four MK containing blends exhibits a high resistance to carbonation. The carbonation depths increase linearly as a function of the square root of time within 280 days of carbonation. Linear regression of the data in Figure 7.2 gives the carbonation rates (K) listed in Table 7.1 when a square root of time dependency (x = K t1/2) is assumed. When the carbonation depths were measured, the mortars showed different rates of color evolution after spraying with phenolphthalein as illustrated in Figure 7.1. For the P and L mortars, the change of color occurred immediately after spraying (within 1 second) whereas the change in color developed more slowly for the four MK mortars, which can be explained by the refinement in pore structures for the four MK mortars as presented in chapter 4 (section 4.3.1). The P and L mortars have relatively good pore connectivity, which causes an immediate contact between phenolphthalein and the alkaline environment. In contrast, the four MK mortars have refined pore structures with low connectivity, resulting in an increased 95


time of diffusion for phenolphthalein to reach the alkaline environment and thus, a slower color change. At the end, the pink color for all samples reached similar darkness.

Figure 7.2. Carbonation depths for the mortars indicated by phenolphthalein (1% (v/v) CO2 at 57% RH and 20 oC). The carbonation coefficients (K) determined from these plots are listed in Table 7.1. Table 7.1. Carbonation coefficients (K) for the studied mortars #1. Mortar id.

K (mm/d1/2)

R2

P

0.17

0.923

L

0.85

0.992

ML

0.62

0.995

M

0.61

0.983

MSL

0.61

0.993

MS

0.68

0.991

#1

The coefficients are obtained from linear fits of the carbonation depths as a function of the square root of time (x = Kt1/2), using the data in Figure 7.2. R2 is the correlation coefficient from the least-squares linear fits.

7.3.2 Mercury intrusion porosimetry (MIP) The impact of carbonation on the intrusion curves from the first and second intrusion cycles is illustrated in Figure 7.3 for all mortars. It can be observed that carbonation causes an increased volume of percolated pores (2nd intrusion) in the range 10 – 100 nm for the MK mortars. In contrast, a small decrease in the volume of percolated pores combined with an increase in pore size of the percolated pores is observed for the P and L mortars after carbonation.

96


Figure 7.3. Impact of carbonation on the changes in intrusion curves from the first and second MIP intrusion cycles for the studied mortars.

97


The impact of carbonation on the different mortars is compared in Figure 7.4, which summarizes the intruded pore volumes and threshold pore sizes derived from the data in Figure 7.3. The results show that the pore threshold for all types of mortars is larger after carbonation, and at the same time that no significant changes of the total intruded porosity are observed for any of the mortars. The intrusion curves for the second intrusion cycle reveal a coarsening effect of the threshold pore radius for all mortars and indicate a minor reduction of the larger intruded porosity for the P and L mortars, but an increase for the M and ML mortars. It has been reported, e.g. by (Morandeau et al., 2014), that the pore-size distribution in ordinary Portland cement paste is significantly reduced during accelerated carbonation (10% CO2, RH = 63%) as a result of clogging by the formed CaCO3 of the whole range of pores accessible by MIP (4 nm – 1 µm). This result is not consistent with the coarsening observed in the present work for all samples. The coarsening effect is expected to facilitate carbonation, however, this is not observed for the P mortar, where a very shallow carbonation depth was observed even after 280 days.

Figure 7.4. Impact of carbonation on the microstructure based on the first (left) and second (right) MIP intrusion cycles for the studied mortars.

7.3.3 Scanning electron microscopy (SEM) Although the molar volumes of the non-carbonated and carbonated phases differ, the MIP analysis does not reveal major changes in the total porosity due to carbonation. Low-magnification SEM BSE images of the P, L and MK mortars are shown in Figure 7.5, where the images of the ML mortars are presented since the images for all MK mortars are very similar. The images reveal that there is an increase in backscattering 98


Figure 7.5. SEM images for non-carbonated (left) and carbonated (right) polished sections of the mortars P (a, b), L (c, d) and ML (e, f). The images for the M mortar are very similar to those of the ML mortar, and thus not shown here. The measurements are performed after 91 days of carbonation for the L mortars and after 280 days for the other mortars. contrast for the carbonated cement paste of all samples. At the used magnification, the images indicate a reduction of the coarse porosity (pores > 5 µm) for the P and L mortars, and a small increase in coarse porosity for the ML mortars. This is in accordance with the changes in porosity derived from the 2nd MIP intrusion curves (Figure 7.4). However, it should be kept in mind that the low-magnification SEM 99


images in Figure 7.5 do not allow for identification of pore sizes in the size range measured by MIP. Changes in porosity resulting from changes of the molar volume of the solid phases will be further discussed based on predictions from thermodynamic modeling.

7.3.4 Thermogravimetric analysis (TGA) Figure 7.6 shows the first derivative (DTG) of the mass losses measured by TGA for samples collected at different depths of the mortars after 91 days of carbonation. The DTG curves show no differences below the apparent carbonation depths, indicating that none of the phases were carbonated in this part. Three principal DTG peaks can be resolved for these samples, i.e., C-S-H/AFt/AFm phases (30 – 300 oC), portlandite (400 – 500 oC) and CaCO3 (500 – 800 oC). The amounts of CaCO3 in the non-carbonated mortars are generally consistent with the initial binder compositions (Table 3.2 in Chapter 3), also considering that the wPc included 3.1 wt.% of limestone. Moreover, the DTG curves for the samples close to the mortar surfaces reveal a significantly higher amount of CaCO3 and a decrease in intensity for the thermal events associated with Ca(OH)2, C-S-H as well as the AFm and AFt phases. The TGA data recorded at different depths in the mortars indicate that portlandite is carbonated first, followed by monocarbonate and ettringite, while some of the C-S-H phase is still present. To which extent the C-S-H is carbonated is difficult to assess based on the TGA data, since only broad DTG peaks are associated with the C-S-H phase. However,

29

Si MAS NMR

studies indicate a progressive decalcification of the C-S-H phase as a result of carbonation (Castellote et al., 2009, Sevelsted and Skibsted, 2015). The DTG curves in Figure 7.6 also show that the decomposition of the CaCO3 that is formed close to the mortar surface, where the degree of carbonation is relatively high, is extended to lower temperature for all mortars. This effect is most clearly seen for the four MK mortars, where the lowest temperature goes down to about 250 oC. This feature has been reported in several other studies (Thiery et al., 2007, Borges et al., 2010, Rostami et al., 2012, Morandeau et al., 2014), and it is likely associated with the presence of a poorly crystalline CaCO3 (Šauman, 1971, Goto et al., 1995). Comparison of the DTG curves for the four MK mortars with that observed for the P mortar indicates that the formation of a poorly crystalline CaCO3 phase is associated with carbonation of the C-S-H or AFt/AFm phases, since only a small amount of portlandite is present in the M and ML mortars. The decomposition of CaCO3 extends to 100


Figure 7.6. DTG curves for samples collected from different depths of the mortars after 91 days of carbonation. The carbonation depths measured by the phenolphthalein spray method are given in each of the individual graphs and refer to the cut slice, which is a physical mixture of carbonated and non-carbonated sample. (Ettri. = Ettringite; Mc = Monocarbonate; CH = Portlandite; Cc = CaCO3; Strä. = Strätlingite.) temperatures as low as 250 oC for these two mortars. Such a broad temperature range has also been reported for Portland cement paste samples, where higher CO2 concentrations up to 10% (v/v) were employed (Morandeau et al., 2014), which reasonably results in higher degrees of carbonation of the C-S-H or AFt/AFm phases. This assumption is also supported by comparison of the P and L mortars, where the L 101


mortar can be seen as a diluted version of the P mortar, since only a minor part of the limestone has reacted during hydration (see section 4.1.1 in chapter 4). As found in chapter 4 (section 4.3.1), the L mortar has the highest intruded porosity and pore connectivity, which may provide a feasible access for CO2 diffusion into the L mortar, resulting in a higher carbonation degree compared to the P mortar. The curves in Figure 7.6 for the L mortar reveal that the decomposition of CaCO3 in the lower temperature range is more obvious than that observed for the P mortar, implying that a larger fraction of the C-S-H or AFt/AFm phases is carbonated for the L mortar. It should be noted that the samples were analyzed by TGA immediately after preparation and without any defined pre-drying. Thus, the first DTG peaks in Figure 7.6 may be affected by moisture in the samples, which is especially visible for the L mortar. However, the presence of small amounts of moisture will not cause any problems for the quantification of CH and CaCO3 from the DTA curves, as the amounts are normalized to the dry weight (TGA temperature at 800 oC). For the analyzed samples, the presence of portlandite particles with carbonated surfaces and non-carbonated centers has also been observed for the P mortar as shown in Figure 7.6. This phenomenon has also been reported in other studies and it is related to the formation of dense calcite layers coating the portlandite crystals, thereby preventing its further carbonation (Groves et al., 1990, Groves et al., 1991, Morandeau et al., 2014).

7.3.5 Carbonation profiles The carbonation profiles for the different mortars after 28 and 91 days of carbonation are illustrated in Figure 7.7. The quantity of formed CaCO3 is estimated by subtracting the amount of CaCO3 originally incorporated as added limestone in the blends from the total CaCO3 determined by TGA. The CaCO3 and Ca(OH)2 contents have been normalized as mol percentage of 100 g ignited mortars (TGA temperature at 800 oC) in order to eliminate the impact of the moisture content in the samples. The amount of CaCO3 is quantified from the DTG curves from about 300 to 850 oC, since the decomposition of CaCO3 is known to extend to lower temperature. The carbonation profiles may be either sharp or gradual. In the literature, it has been stated that (i) carbonation is a diffusion-controlled process, i.e., the chemical reactions are instantaneous, which would result in a sharp reaction front (Papadakis et al., 1991) and (ii) carbonation is a reaction-controlled process, which would give a gradual reaction front (Thiery et al., 2007). In the present study, the different binder types and distinct 102


Figure 7.7. Carbonation profiles obtained by quantification of CaCO3 and Ca(OH)2 from the TGA data for the different mortars. The results are presented as molar content per 100 g of ignited mortar (800 oC). The contents of CaCO3 originally present in the mortars have been subtracted from the total measured CaCO3 quantities.

103


Figure 7.8. Carbonation profiles after 91 days of CO2 exposure from Figure 7.7. The amount of CaCO3 formed by carbonation of other phases than Ca(OH)2 is calculated by subtracting the molar amount of reacted Ca(OH)2 from the total quantity of CaCO3.

104


pore structures of the mortars result in the observation of both sharp and gradual carbonation fronts. The Ca(OH)2 profiles appear sharp for all mortars (Figure 7.7), indicating that the carbonation of Ca(OH)2 can be considered as quasi-instantaneous for the present mortars. Thus, the gradual change in CaCO3 content between the surface and the apparent carbonation depth is ascribed to carbonation of other phases such as the CS-H and AFm phases. In earlier studies, where a 10 % (v/v) CO2 was employed (Thiery et al., 2007), a gradual change in the Ca(OH)2 content and even a decrease of the amount of Ca(OH)2 below the apparent carbonation depth were detected by TGA. These observations were explained by an insufficient number of available reaction sites to complete the carbonation reactions at the high CO2 concentration, resulting in CO2 that penetrates to deeper depths, i.e., below the apparent carbonation depth. Fig 7.8 shows the Ca(OH)2 and CaCO3 profiles after 91 days of carbonation, including the CaCO3 profiles attributed to carbonation of the C-S-H and/or the AFt/AFm phases rather than Ca(OH)2. The profiles show that most of the CaCO3 formed in the M and ML mortars originates from carbonation of the C-S-H and AFt/AFm phases. In addition, carbonation of the C-S-H and/or AFt/AFm phases is also observed for the P and L mortars in addition to the principal carbonation of Ca(OH)2. Thus, based on the TGA data, it appears that the reaction of portlandite is fast compared to the diffusion of CO2 while the decomposition of C-S-H and the AFm phases is slow, resulting in the gradual changes of the CaCO3 content close to the surface of the samples.

7.3.6 Thermodynamic modeling The phase assemblages caused by carbonation can be predicted by thermodynamic modeling by stepwise adding a certain amount of CO2 (1% (v/v)) as described in chapter 3 (section 3.3.9). The predicted data are presented in volume percentages of the original volume as a function of the volume of added air in Figure 7.9. It is predicted that Ca(OH)2 will carbonate before the C-S-H phase for the P and L blends. However, this is not always observed experimentally by bulk analyses, since some Ca(OH)2 particles carbonated only at the surface may be present (Morandeau et al., 2014). The modeling also shows that monocarbonate is carbonated along with the high-Ca C-S-H phase, the reaction products being strätlingite and calcite. Strätlingite is also predicted for the MK blends, in accordance with the experimental studies (Dai et al., 2015), and near the surface this phase is also carbonated. Moreover, thermodynamic modeling predicts a decalcification of the C-S-H phase from a high-Ca C-S-H to a low-Ca C- S-H 105


Figure 7.9. Changes in phase assemblages and total porosities predicted by thermodynamic modeling for the studied blends in contact with air containing 1% (v/v) CO2. The horizontal dashed lines show the original porosities of the samples prior to carbonation, whereas the dashed arrows indicate the changes in porosity after carbonation. The vertical dashed lines reflect the pH corresponding to decalcification of the high-Ca C-S-H phase, which is also relevant for the de-passivation of the reinforcement. 106


, i.e., C-S-H depleted from Ca2+ ions in the interlayer space (Kulik, 2011). This is in accord with a very recent NMR study of carbonated C-S-H samples (Sevelsted and Skibsted, 2015) which showed that the carbonation of C-S-H includes two steps: (i) calcium is gradually removed from the interlayer and the defect sites in the silicate chains until Ca/Si = 0.67 is reached and (ii) the Ca from the principal layer is then consumed, resulting in the final decomposition of the C-S-H and the formation of an amorphous silica phase. A progressive polymerization of the C-S-H phase as a result of carbonation was also observed by Castellote et al. (Castellote et al., 2009) for increasing CO2 concentrations (0.03%, 3%, 10% and 100%). In addition, a fraction of a C-S-H phase with a lower Ca/Si ratio than observed for the non-carbonated sample (Ca/Si=1.87) was detected by

29

Si MAS NMR when the samples were exposed to a

0.03% and 3% CO2 environment. These results are well supported by the predicted changes of the Ca/Si ratio for the C-S-H phases shown in Figure 7.10, where Ca/Si decreases gradually from the initial values of 1.63 (for the P and L mortars) and 1.29 (for M and ML mortars) to 0.67, corresponding to the decalcification of the high-Ca CS-H. The Ca/Si ratio of the C-S-H phase in the four MK mortars before carbonation are lower than the ratio for the C-S-H in the P and L samples, implying that the C-S-H phases in the P and L mortars have a higher CO2 binding capacity as compared to the low-Ca/Si C-S-H phases in the MK mortars, again in accordance with the recent study of carbonation of synthesized C-S-H phases with different Ca/Si ratios (Sevelsted and Skibsted, 2015).

Figure 7.10. Changes of the Ca/Si ratio for the C-S-H phase as a result of carbonation for different blends predicted by thermodynamic modeling. 107


7.3.7 Discussion 

Pore structures

Thermodynamic modeling can be used to predict changes in total porosity as a result of variations in the molar volume of phases after carbonation as shown in Figure7.9. It should be noted that the volume of the C-S-H phase does not include the gel porosity associated with this phase but only the interlayer water (Kulik, 2011, Muller et al., 2013). Thermodynamic modeling predicts a major increase of the total porosities for all mortars, when all phases are completely carbonated, corresponding to large volumes of CO2 containing air (e.g., 10–1.0 – 10–0.50 dm3/g of cement, Figure 7.9). At lower carbonation degrees a small decrease in total porosity is predicted for the P and L mortars as a result of the differences in molar volumes for portlandite (32.29 cm 3/mol) and CaCO3 (e.g. 36.92 cm3/mol (Smyth and McCormick, 1995)). Only an increase in porosity is predicted for the MK mortars, since virtually no portlandite is present in these mortars. The DTG curves (Figure 7.6) show that large fractions of the phases (30 – 300 oC) are not carbonated after 91 days of exposure. Thus, the thermodynamic stability of the phases towards carbonation (Figure 7.9) and the observed minor changes in total porosity measured by MIP (Figure 7.4) suggest that the major non-carbonated phase observed in Figure 7.6 most likely is a low-Ca C-S-H phase. This is in line with the earlier study of cement pastes exposed to difference CO2 concentrations where 29Si MAS NMR revealed the presence of a fraction of low-Ca C-S-H in the carbonated samples exposed to 0.03% and 3% CO2 (Castellote et al., 2009). The combination of the results from MIP and thermodynamic modeling indicates that the carbonation increases the percolation (threshold pores sizes). The total porosity is initially reduced but it increases at higher degrees of carbonation for the P and L mortars. Similar results can be derived for the MK mortars, except that these mortars reveal no decrease in total porosity (Figure 7.9). The reduced total porosity for the P and L mortars is explained by carbonation of portlandite. The percolation and later increased porosity for all mortars are probably related to carbonation of the C-S-H resulting in a partial decalcification of this phase. 

CO2 binding

The L mortar has the largest intruded porosity and a reduced amount of portlandite as compared to the P mortar, which reasonably well accounts for the lower carbonation resistance observed for the L mortar. The two metakaolin-containing mortars, which 108


include the same amount of Portland cement as L, show considerable better carbonation resistance as a result of the refinement and decrease of the porosity. Comparison of the carbonation resistance for the P and MK mortars reveals that the poorer performance for the MK mortars cannot be explained by the pore structure measurements, since the noncarbonated P mortar has a total MIP porosity comparable to those for the MK mortars. The ability of the hydrated cement to bind CO2 must also be taken into account and in this section the CO2 binding results obtained from TGA and thermodynamic modeling are analyzed. To predict the CO2 binding capacity based on the phases formed after 91 days of hydration, the amounts of the main calcium-bearing phases relevant to cement carbonation have been calculated by the GEMS software, using molar percentages rather than volumes, and related to the masses of ignited mortars (by division with a factor of 4 according to the binder/sand ratio of 3). The total amount of CaO present in the hydrates (i.e., in portlandite, C-S-H and AFm but not in limestone) can be used to predict the maximum CO2 binding capacity for each system. The amount of CaO from ettringite for all mortars and from initially formed strätlingite for MK blends are not taken into account, since these phases are predicted to be thermodynamically more stable than the high-Ca C-S-H phase. The measured CO2 binding was obtained by quantification of CaCO3 and portlandite by TGA and the results are summarized in Table 7.2. It should be noted that the samples used for the CaCO3 measurements were taken from the outer layer of the mortars, which have been in direct contact with CO2 in the air, whereas the samples used for the portlandite measurements were taken from the non-carbonated center of the mortars. The thermodynamic calculations indicate that the P mortar has a significantly higher available CaO content (Table 7.2), potentially reactive to CO2, allowing for a much higher binding as compared to the three other blends. However, the experimental data in Table 7.2 shows that the amount of CaCO3 bound in the different composite cement mortars is similar after 91 days of carbonation, indicating that carbonation proceeds at a similar reaction rate in the different mortars, although they have different total potential CO2 binding capacities and contain different phases (CH or C-S-H) to bind CO2. In this context, it should also be noted that the pore structure data in Figure 7.4 shows that the P and MK mortars exhibit similar microstructural changes during carbonation. The similar changes in the pore structure and the small variations in measured CO2 binding

109


data do not explain the observed differences in the apparent carbonation depths measured by the phenolphthalein method as shown in Figure 7.2, unless the pH profiles and capillary condensation are taken into account. Table 7.2. Comparison of the measured CaCO3 (carbonated samples close to the mortar surface) and portlandite (non-carbonated inner core) after 91 days of carbonation and the calculated total potential CO2 binding capacity (mol CaO/100g ignited mortar) for the hydrated mortars. Method

ML

M

MSL MS

0.11 0.07

0.00

0.00

0.00

0.00

C-S-H

0.17 0.14

0.13

0.14

0.14

0.14

Monocarbonate

0.06 0.003 0.055 0.030 0.046 0.028

Total CaO

0.28 0.20

0.18

0.17

0.19

0.17

0.13 0.12

0.13

0.13

0.12

0.12

Surface: Ca(OH)2

0.02 0

0

0

0

0

Core: Ca(OH)2

0.10 0.07

0.008 0.004 0.007 0.004

Phases

Calculated Ca(OH)2 by GEMS

Measured Surface: CaCO3 (carbonation) by TGA



P

L

pH profiles

Since the carbonation depth was measured with the phenolphthalein indicator, the differences between the P and MK mortars with respect to their apparent carbonation performances reflect changes in pH which are related to phase changes. Figure 7.11 shows the pH profiles predicted by thermodynamic modeling for the different blends. The pH decreases with increasing carbonation from an initial high value of 13.4 (P and L) and 12.8 (MK mortars). For the high pH values (pH > 12.8), no decrease in pH is predicted as a result of the carbonation of portlandite (P and L mortars) or the carbonation of monocarbonate and the formation of strätlingite (MK mortars) since the pH under these conditions is dominated by the alkali ions present in the pore solution. The first major decrease in pH (to a value of pH ≈ 9.7) is predicted to result from carbonation of the high-Ca C-S-H to a low-Ca C-S-H phase and calcium carbonate. The second major decrease of pH (to pH = 7.4) is predicted to occur when the low-Ca C-S-H phase is carbonated, which corresponds to a decalcification of the principal layer of the C-S-H phase, thereby decomposing into a silica phase and calcite. Based on this interpretation, the carbonation depths indicated by the phenolphthalein method reflect a 110


depletion of the high-Ca C-S-H phase. As very little portlandite is available for carbonation, due to the pozzolanic reaction with MK, the carbonation of the C-S-H phase from a high-Ca C-S-H to a low-Ca C-S-H and thus the observed pH decrease, take place for lower amounts of CO2 for the four MK mortars as compared to the P mortar. This explains the shallow apparent carbonation depths observed for the P mortar (Figure 7.2). A combination of the results for the similar measured CO2 bindings and the predicted pH profiles support the recent study by Leemann et al. (Leemann et al., 2015), which proposed that the concept of CO2 capture capacity per volume of cement paste is the decisive material parameter for carbonation resistance. This higher CO2 binding capacity can be used as an explanation for the observed highest carbonation resistance for the P mortar, since it can capture a higher amount of CO2 per volume of paste as compared to the other mortars. For the L and MK mortars, thermodynamic modeling predicts that the partial carbonation of the C-S-H phase will occur for higher amounts of CO2 for the L mortar than for the MK mortars. The deeper carbonation depth observed for the L mortar compared to the MK mortars is likely caused by differences in the pore volume accessible to CO2. According to the Kelvin equation, capillary condensation at 57% RH takes place at a pore radius of ≈ 2 nm. Below 2 nm the pores will remain to be water filled which slows down CO2 diffusion, while pore radii > 2 nm are expected to contain air. However, pores with a 2 nm radius are close to the detection limit of MIP of 3 nm. (Turcry et al., 2014) have reported the evolution of RH at different concrete depths as a function of drying time at a temperature of 45 ± 3 oC. When the RH of the concrete surface was dried to about 20%, a RH of 60% at a 10 mm concrete depth was still detected after 14 days, and at even deeper depth, the RH was 100%. The data presented in chapter 4 (Figure 4.7) have shown that the weight loss of the MK mortar during drying at 50 oC do not stop before 65 days of drying as a result of the very fine pores, as indicated by the MIP data (Figure 4.5 in chapter 4). Considering the slower drying process when the samples are exposed to 57% RH at 20 oC compared to the samples exposed to 50 oC in the oven, suggests that a higher RH inside the mortars can be present for a long time during carbonation, e.g., a RH of 80% which corresponds to a Kelvin pore radius of 4.8 nm. The MK mortars have considerably lower amounts of pores above the 4.8 nm radius than the L mortar (Figure. 4.5 in chapter 4), and thus the carbonation is slowed down for the MK mortars in comparison to the L mortar. Thus, the effect of capillary condensation has to be taken into account, when the performance of cements under accelerated conditions 111


with low RH is transferred to concrete structures, where the higher RH may have a strong cement-specific impact on the carbonation resistance (Leemann et al., 2015).

Figure 7.11. pH profiles related to the changes in phase assemblages predicted by thermodynamic modeling. The right bar indicates the gradual color change of phenolphthalein from fuchsia to colorless upon pH changes from 10 to 8.2. It should be noticed that phenolphthalein in solution shows a gradual color change from fuchsia to colorless upon a change in pH from 10 to 8.2. Such a gradual change in color has not been documented for cement-based materials and neither has the cement literature agreed on a single, specific value for the pH that causes the color change. Evidence for calcite coating of portlandite crystals, preventing them from buffering the pore solution, has been clearly demonstrated (Galan et al., 2015a). Similarly, indications of remains of non-carbonated C-S-H are observed in apparently carbonated samples (Figure 7.6). Here, the DTG peak in the range 30 – 200 oC is assumed to originate mainly from the low-Ca C-S-H phase.

7.4 Conclusions (1) The highest and lowest carbonation performance is observed for the P and L mortars, respectively, and the four MK mortars exhibit poor resistance to carbonation as compared to the P mortar. (2) Thermodynamic modeling predicts carbonation of Ca(OH)2 followed by carbonation of the C-S-H phase for the P and L mortars. In the blends with metakaolin, where virtually all Ca(OH)2 has been consumed by the pozzolanic reaction, carbonation of monocarbonate and the C-S-H phase is predicted. In addition, a two-step decalcification process of the C-S-H phase is predicted where calcium initially is released from the interlayer of the high-Ca C-S-H, resulting in the formation of a low-Ca C-S-H phase. In 112


the second step, Ca is released from the principal layers of the low-Ca C-S-H, leading to the final decomposition of the C-S-H phase, in agreement with recent results from

29

Si

MAS NMR studies (Castellote et al., 2009, Sevelsted and Skibsted, 2015). The TGA data recorded at different depths in the mortars confirm the sequence of carbonation predicted

by

thermodynamic

modeling,

i.e.,

portlandite

carbonates

before

monocarbonate and ettringite before the complete decomposition of the C-S-H phase to silica. (3) Carbonation of portlandite is predicted to have little impact on pH as long as some high-Ca C-S-H phase is still present. Based on these predictions, the carbonation of the high-Ca C-S-H phase reduces pH to 9.6, which is in the range from 10 to 8.2 where the phenolphthalein changes from fuchsia to colorless. The subsequent decalcification of the low-Ca C-S-H to a silica gel and calcium carbonate reduces pH further. Moreover, from the TGA measurements it is found that the amount of CaCO3 bound in the different composite cement mortars is similar after 91 days of carbonation, indicating that carbonation proceeds at a similar reaction rate in the different mortars, although they have different total potential CO2 binding capacities and contain different phases to bind CO2. (4) Thermodynamic modeling reveals a major increase of the total porosity when all phases are carbonated. Little changes in porosity is predicted for the P and L mortars before the low-Ca C-S-H phase is carbonated. This is confirmed by the experimental observations by SEM and the repeated MIP (2nd intrusion) which show a slight reduction of the percolated porosity in the P and L mortars. The volume and size of the percolated pores seem to have increased in the mortars containing metakaolin. More importantly, MIP showed a coarsening of the pore threshold for all mortar types, but only small variations of the total intruded porosity. (5) The P and MK mortars were found to bind similar quantities of CO 2 from TGA and to exhibit similar coarsening of the pores, as measured by MIP. The difference in apparent carbonation depths for these mortars is explained by the modelled impact of phase changes on pH. The partial carbonation of the C-S-H phase starts earlier for the MK mortars, resulting in an earlier pH decrease and thus a deeper apparent carbonation depth. For the L mortar, the amount of portlandite is higher than in the MK mortars. However, this does not improve the carbonation resistance, which is explained by a

113


coarser pore structure of the L mortar leading to a lower amount of pores filled with water due to capillary condensation and as such a better accessibility to CO2. (6) In summary, the data obtained in the present study confirm that the CO2 binding capacity, the porosity, and the capillary condensation are the decisive parameters governing the carbonation depth. A higher CO2 binding capacity, and thus a slower decrease of pH, is the main reason for the higher carbonation resistance observed for the P mortar. The difference in carbonation resistance for the limestone and the metakaolin containing mortars, which all have similar CO2 binding capacities, is related to the difference in the porosity. This is attributed to the presence of metakaolin which leads to less porosity and a refined pore structure that favors the capillary condensation at a given RH, as compared to the limestone mortar, and consequently results in a lower carbonation rate.

114

Chapter 8: Impact of Compressive Strength on Durability

CHAPTER 8: IMPACT OF COMPRESSIVE STRENGTH ON DURABILITY Chapter 8 investigates the impact of the 91-day compressive strengths presented in chapter 4 on the durability performance (i.e., carbonation, chloride ingress and sulfate attack) for the wPc’ – MT – LS and wPc’ – G – LS mortars. The results are compared with those obtained for the wPc – MK – SF –LS mortars, as partially presented in chapter 5 and chapter 7.

115


8.1 Introduction A concrete mix is often designed by the construction industry based on the 28-day compressive strength, or possibly on 91-day compressive strengths for concretes incorporating SCMs, because the determination of other properties, e.g., chloride diffusion coefficient, water permeability, etc. is time consuming and costly. In some published studies, it is assumed that a durable concrete is related to a higher compressive strength of the concrete (Baghabra Al-Amoudi et al., 2009), because it is widely accepted that compressive strength and durability both are influenced by the porosity of the concrete (Halamickova et al., 1995, Tumidajski et al., 1996, Khan et al., 2000, Kumar and Bhattacharjee, 2003, Poon et al., 2006). It is reported (Armaghani et al., 1992) that fly ash and silica fume concrete mixtures with equal compressive strengths do not necessarily produce equal levels of permeability. However, the lack of correlation continued until the strength reached about 8000 psi (55 MPa), beyond which a well-defined trend was observed. (Vu et al., 2001) reported a strength gain of kaolinblended Portland cement mortars and concrete accompanied by an improved durability. (Khan and Lynsdale, 2002) proposed a good correlation between carbonation depth and 28-day compressive strength over a wide range from 30 to 110 MPa. Similar relationships have also been established by (Khunthongkeaw et al., 2006, Rabehi et al., 2013). It is reported that the chloride diffusion coefficient decreased with an increase in compressive strength for all Portland cement – fly ash concretes with a substitution level from 0 to 50 wt.% and for all chloride exposure temperatures from 5 to 45 oC (Dhir et al., 1993). The relationship between the chloride diffusion coefficient and the compressive strength was also reported in several other studies (Lee and Yoon, 2014), where the chloride diffusion coefficient decreased with increasing compressive strength. (Baghabra Al-Amoudi et al., 2009) established very good correlations between the compressive strengths and selected durability indices, particularly the chloride permeability and the coefficient of chloride diffusion, irrespective of the mix design parameters. However, these correlations were observed to be dependent on the type of cement. (Loser et al., 2010) also reported a correlation between the chloride diffusion coefficients and the compressive strengths. However, they also stated that the correlation does not mean that the strength is the decisive factor for chloride resistance. (Leemann et al., 2015) found that the relation between carbonation and compressive strength was dependent of type of cement. Addition of siliceous SCMs in Portland cement-based mortars shifts the data away from the correlation curve between the 116


compressive strength and the carbonation coefficients. Based on the above mentioned studies, it seems to be too simple to correlate the compressive strengths with durability indices by just considering their common effects from porosity. This chapter will provide more evidence to show that the compressive strength may not be a good indicator for predicting the durability especially for blended cements with SCMs. For this purpose, the Portland cement-based mortars including montmorillonite (MT), glass (G) and limestone (L) are designed to have comparable compressive strengths and comparable volumes of paste per volume of mortar by varying the water/binder ratio and the binder/sand ratio. The durability performance of these mortars are investigated, and the results are compared with those obtained for the wPc – MK – SF – LS mortars with the fixed water/binder ratios as partially presented in chapter 5 and chapter 7.

8.2 Experimental The materials, mix design and mortar preparations for the wPc’ – MT – LS and wPc’ – G – LS blends are described in chapter 3 (section 3.2). The mortars were designed to have comparable compressive strengths and comparable volumes of paste per volume of mortar. In addition, sulfate resistance for the wPc – MK – SF – LS mortars prepared in (section 3.1) are also presented in this chapter.

8.2.1 Chloride ingress and carbonation The procedures for chloride ingress (e.g. chloride exposure and total chloride profile analysis) and carbonation (e.g. exposure and measurement of carbonation depths) experiments are the same as described in chapter 5 and chapter 7, respectively.

8.2.2 Changes in mass and lengths The mortar size (20 × 20 × 160 mm3) and the concentration of the Na2SO4 solution (16 g/L, i.e. 0.11 mol/L) are chosen according to the standard (CENTC51:WG12/TG1, 1994). After 91 days of curing, as described in chapter 3 (section 3.1.3), the initial mass and length of the mortars with gauge studs were measured. They were then exposed to the prepared Na2SO4 solution in separately sealed plastic boxes for each blend. The sulfate exposing mortars are stored in temperature-controlled chambers at 20 oC and 5 o

C, and the mortars submerged in demineralized water as reference are stored at 20 oC

under the same conditions. The subsequent length measurements and changes of the 117


solution were carried out weekly at the early exposure and monthly during the later exposure.

8.3 Results and discussion 8.3.1 Carbonation resistance The increase in carbonation depths over time for the P’, L’, MT and G mortars exposed to 1% CO2 and 57% RH are shown in Figure 8.1 and Figure 8.2. The highest carbonation resistance is observed for the P’ mortar, as expected to be similar to the P mortar as presented in chapter 7, which is attributed to the similar CaO content for wPc’ and wPc. The 91-day carbonation depth for the L’ mortar is reduced by half as compared to the L mortar, as a result of the reduction in w/b ratio from 0.50 to 0.41 and the increase in binder content. The 91-day carbonation depths for the MT and G mortars (except for G-0.25) are similar to the MK mortars, which are slightly affected by the varied binder contents. Although the P’, L’, MT and G mortars are designed to have similar compressive strengths, and the MIP data in chapter 4 (section 4.3.2) shows similar pore structures for all mortars, the carbonation resistance for the different mortars will not necessary be the same as shown in Figure 8.1 and Figure 8.2.

Figure 8.1. Carbonation depths for the P', Figure 8.2. Carbonation depths for the P’, L’ and MT mortars. L’ and G mortars. It is interesting that no major decrease in carbonation rates is observed for the MT and G mortars, as compared to the MK mortars, even though a large fraction of portlandite is still present in the MT and G mortars according to the data presented in chapter 4 (section 4.1.2). Thus, the carbonation rates for the studied mortars may be dependent on 118


the total reactive CaO per volume of cement paste rather than the type of phases. (Leemann et al., 2015) proposed a new parameter which can be expressed by the ratio between water added during production and the amount of reactive CaO present in the binder (w/CaOreactive). The amount of CaO present in the binder materials are subtracted from the total CaO content as it is not reactive. It should be mentioned that only the CaO in the cement hydrates is considered to be able to capture CO2, whereas those present in anhydrous cement or unreacted SCMs are not responsible for CO2 capture. However, the latter amounts were not subtracted in the calculations, as the degree of hydration was not measured, which might cause some scatter in the relationship between carbonation coefficient and w/CaOreactive as shown in Figure 8.3. In general, the results show an increase in carbonation coefficient with an increase in w/CaOreactive, which is consistent with the results reported in (Leemann et al., 2015). The authors (Leemann et al., 2015) also reported that there is no trend in the carbonation coefficients indicating a slower or faster reaction with CO2 based on the different C-S-H to portlandite ratios.

Figure 8.3. Carbonation coefficient as a function of the w/CaOreactive for the P’, L’, MT and G mortars (R2=0.63).

8.3.2 Chloride resistance The chloride resistance of the MT, G and reference mortars (P’ and L’) is evaluated by analyzing the total chloride profiles as shown in Figure 8.4 and Figure 8.5. The results are compared with the total chloride profiles for the wPc – MK – LS mortars presented in chapter 5 and their fitting curves based on Fick’s second law are repeated in Figure 119


8.4 and Figure 8.5. Since there are no differences in chloride resistance for the four MK mortars, only the total chloride profile for the M mortar is repeated for comparison. It is expected that the total chloride profiles for the P’ and P mortars are similar according to their similar chemical composition and pore structures. It is known that the P’, L’, MT and G mortars have comparable compressive strengths and pore structures as presented in chapter 4. Thus, similar to the impact of compressive strengths on the carbonation depths, there is no direct relationship between the compressive strengths and the chloride resistance of the studied mortars. In fact, the changes in total chloride profiles between different mortars are attributed to different chloride binding capacities. It seems that the incorporation of MT and G in the Portland cement mortars increases the chloride binding, hence retards the chloride penetration and increase the chloride resistance. It is known from chapter 6 that the increased chloride binding for the MK mortars is ascribed to the formation of Friedel’s salt, because the chloride binding capacity for the low Ca/Si ratio C-S-H formed in the MK mortars is lower than those formed in the P mortar. This may also be similar for the MT and G mortars as a result of the additional alumina source from MT and G. However, it is not clear how the formation of Friedel’s salt retards the chloride penetration. (Yuan et al., 2009) stated in their review article that the formation of Friedel’s salt may decrease the porosity. Unfortunately, the pore structures of the MT and G mortars after chloride exposure have not been analyzed.

Figure 8.4. Comparison between the total Figure 8.5. Comparison between the total chloride profiles for the wPc’ – MT – LS chloride profiles for the wPc’ – G –LS and and wPc – MK – LS mortars. wPc – MK – LS mortars.

120


8.3.3 Sulfate resistance 

wPc – MK – SF – LS mortars with the same water/binder ratios

The expansion curves for the wPc – MK – SF – LS mortars caused by sulfate attack at 20 oC and 5 oC are shown in Figure 8.6 and Figure 8.7, respectively. The results show that the expansions for the P and L mortars increase gradually and finally that the mortars degrade. The tortuosity of the mortars is reflected by the error bars shown in the figures, which increase with the exposure time. It should be noted that the solutions were changed weekly at the early-age exposure and changed monthly during the later exposure. The results show that the expansion for the L mortar at 20 oC develops faster than for the P mortar at the beginning, and finally reaches a similar level after 400 days. However, the expansion for the L mortar at 5 oC develops similar to the expansion for the P mortar at the beginning of the exposure, and the expansion for the L mortar increased dramatically after 91 days of exposure until falling apart. The results from both 20 oC and 5 oC show that the four MK mortars do not show any expansions throughout the measurements. There is not any degradation on the surface of the mortars, which indicates a very high resistance for the MK mortars against sulfate attack.

Figure 8.6. Length changes for the wPc – Figure 8.7. Length changes for the wPc – MK – SF – LS mortars exposed to the MK – SF – LS mortars exposed to the Na2SO4 solution at 20 oC. Na2SO4 solution at 5 oC. It is known that sulfate attack can lead to the formation of ettringite, gypsum, and/or thaumasite as described in chapter 2 (section 2.4). This may cause an increase in the weight of the mortars. It is also known that leaching of calcium from the hydration products will occur when hydrated cements are exposed to calcium-free solutions, 121


which will cause a loss of weight (Rozière et al., 2009). Since the studied mortars in this section were kept saturated in demineralized water and were proved to be well hydrated before exposing to the Na2SO4 solution, neither water absorption nor continued hydration would significantly affect the mass variation of the mortars. The changes in weight of the mortars exposed to the Na2SO4 solution will mainly be attributed to sulfate attack and leaching, and the results are presented in Figure 8.8 and Figure 8.9. The data are normalized as the weight percentage of the weight at 14 days of exposure rather than the weight of the mortars before exposure in order to eliminate the impact of initial ion uptake through diffusion. It should be noted that the impact of leaching is not subtracted, thus the recorded mass evolution curves are a combined effect of sulfate attack and leaching. The results show that the weight of the L mortar increases throughout the measurements at both temperatures, indicating a much stronger effect of sulfate attack than the effect of leaching. Comparison of the weight loss for the L mortar at different temperatures shows that the increased weight of the reaction products at 5 o

C is less than those formed at 20 oC. However, the L mortar at 5 oC degrades more

severely than at 20 oC, which reveals the different expansive natures of the reaction products between conventional sulfate attack and thaumasite sulfate attack. The weight changes for the P mortars at both temperatures show clearly the impact of leaching, which retards the increase in weight caused by sulfate attack especially for the P mortar at 20 oC. It is interesting that there is not any increase in weight recorded for any of the MK mortars. However, a major decrease in weight is recorded for these mortars. The sulfate ions seem to be blocked outside of the mortar and the mechanism for this is not clear, although many published studies have attributed this to the fine pore structures (Khatib and Wild, 1998, Bhatty and Taylor, 2006, Najimi et al., 2011). In general, the measured weight changes support the expansion data very well, and the absence of expansion for the MK mortars is clearly a result of almost no ingress of sulfate ions. The relationships between the sulfate expansion and mass variation for the P and L mortars at 20 oC and 5 oC are shown in Figure 8.10. The results show nicely an increase in expansion with a gain in mass, and an identical trend is observed for the same mortar regardless of the types of sulfate attack. However, it should be mentioned that the mortars at 5 oC degrade much more severely than the mortars at 20 oC even though their expansions are identical. It is interesting that the data show a shift towards a higher mass gain for the L mortars, which can be explained by the larger porosity for

122


the L mortar than for the P mortar as presented in chapter 4 (section 4.3), because the formed sulfate products tend to fill the larger pores before generating expansions.

Figure 8.8. Normalized mass variations for Figure 8.9. Normalized mass variations for the wPc – MK – SF – LS mortars exposed the wPc – MK – SF – LS mortars exposed to the Na2SO4 solution at 20 oC. the Na2SO4 solution at 5 oC.

Figure 8.10. Relationship between the sulfate expansions and the mass gain for the P and L mortars at 20 oC and 5 oC under sulfate attack. 

wPc’ – MT – LS and wPc’ – G – LS mortars with comparable compressive strengths

The evolution of the expansion caused by sulfate attack over time of exposure for the P’, L’, MT and G mortars are shown in Figure 8.11. The results show a gradual increase in expansions for the P’ and L’ mortars in the beginning and a substantial increase after about 60 days. On contrary to the observations for the P and L mortars, the expansions 123


for the P’ mortar developed faster than for the L’ mortar. In addition, both P’ and L’ mortars have expansions developing faster than for the P and L mortars, as a result of the higher C3A content in wPc’ as compared to wPc. Similar to the MK mortars, there are no major expansions recorded for the MT and G mortars, even though they have similar pore structures as compared to the P’ and L’ mortars. This seems to contradict with most of the conclusions from published studies (Khatib and Wild, 1998, Bhatty and Taylor, 2006, Najimi et al., 2011) which stated that the increased sulfate resistance for mortars including SCMs is attributed to refinement of the pore structures.

Figure 8.11. Length changes for the wPc’ – MT – LS and wPc’ – G – LS mortars exposed to the Na2SO4 solution at 20 oC. The changes in weight for the P’, L’, MT and G mortars exposed to the Na2SO4 solution and demineralized water are shown in Figure 8.12 and Figure 8.13. Similar to the MK mortars, the leaching phenomenon for the MT mortars exposed to the Na2SO4 solution is also recorded, but not for the G mortars, where no major changes in weight is measured. However, the MT and G mortars are measured to have a slight increase in weight, especially for the G mortars, which are not measured for the mortars exposed to the Na2SO4 solution. It is known that the pozzolanic reactivity of the MT and G compounds is much lower as compared to MK (see chapter 4). Thus, the gain in weight for the MT and G mortars might be ascribed to continuous hydration. However, the reason for the weight increases, that are not observed for the MT and G mortars exposed to Na2SO4, is not clear. Similar to the relationships between expansion and mass gain established for the P and L mortars as shown in Figure 8.10, the similar relationships for the P’ and L’ mortars are also observed as shown in Figure 8.14. It is seen that the two parameters correlate very well regardless of the rates of development in expansions. 124


Figure 8.12. Normalized mass variations for the wPc’ – MT – LS mortars exposed to the Na2SO4 solution and demineralized water at 20 oC.

Figure 8.13. Normalized mass variation for the wPc’ – G – LS mortars exposed to the Na2SO4 solution and demineralized water at 20 oC.

Figure 8.14. Relationship between sulfate expansions and mass gain for the P’ and L’ mortars at 20 oC under sulfate attack. 125


8.4 Conclusions In this chapter the impact of compressive strength on the durability has been investigated. The results show that the compressive strength is not a good indicator to predict the durability performance of the studied mortars (P’, L’, MT and G). The results are compared with those obtained for mortars with the same water/binder ratio (P, L and MK mortars). Because the P’, L’ MT and G mortars have comparable pore structures, their durability performances are dependent of their chemical properties. The following conclusions can be drawn: (1) The carbonation data confirms that the carbonation resistance of the P’, L’ MT and G mortars are governed by the CO2 buffer capacity as expressed by the added water over the total reactive CaO content (w/CaOreactive) as proposed by (Leemann et al., 2015) rather than the type of phases. (2) Addition of MT and G in Portland cement mortars retards the chloride penetration depth and change the shape of the total chloride profiles, which is attributed to an increased chloride binding through formation of more Friedel’s salt. (3) Addition of MT and G increase the sulfate resistance of the Portland cement mortars significantly as no expansions are observed for the MT and G mortars. The absence of expansions for the MT and G mortars are well explained by the mass variation data.

126

Chapter 9: Conclusions

CHAPTER 9: CONCLUSIONS The independent conclusions for each of the durability issues covered in this thesis have been outlined in the final parts of the previous chapters. This chapter tends to bring the related conclusions together in a logical structure with the aim of shaping the diverse durability issues together into a unified evaluation.

127


Durability of Portland cement – calcined clays and limestone blends have been investigated in this thesis considering their interactions with chloride, carbonate and sulfate ions. The study has been divided into two parts in terms of mix design: (i) Portland cement – metakaolin – silica fume – limestone blends, i.e. P, L and MK (ML, M, MSL and MS), prepared with the same water/binder ratio of 0.5. This part has mainly focused on the chloride ingress in mortars (chapter 5), chloride binding in pastes (chapter 6), carbonation in mortars (chapter 7), and performance tests with respect to sulfate attack (chapter 8). (ii) Portland cement – montmorillonite – limestone blends and Portland cement – glass – limestone blends, i.e. P’, L’ MT (MTL-0.5, MTL-0.75 and MT-1) and G (GL-0.25, GL-0.5, GL-0.75 and G-1), prepared with comparable compressive strengths by varying the water/binder ratios and binder/sand ratios. This part focuses on the impact of compressive strength on the durability performance (chapter 8). Since the hydration of the P, L and MK mortars are nearly stabilized after 91 days of hydration, the influence of the minor continuous hydration of these blends during the accelerated durability tests can be eliminated. Thermodynamic modeling predicts a complete pozzolanic reaction after 91 days, even though there is still minor portlandite observed from the DTG curves. The results indicate a high reactivity of the metakaolin and silica fume used in this study. The present work shows that the pore structures of the four MK mortars are significantly refined. As expected, no differences in compressive strengths after 91 days of hydration are observed for the P’, L’, MT and G mortars. Comparable pore connectivity is also observed for these mortars, as revealed by the MIP and PF methods. Quantification of portlandite from the DTG curves shows that a large fraction of portlandite is present in the MT and G mortars, indicating a very low pozzolanic reactivity of the montomorillonite and synthetic glass used in this study, as compared to the metakaolin and silica fume. For the wPc – MK – SF – LS system, the L mortar exhibits lowest performance in all the durability testes in this study. The P mortar shows poor resistance to chloride ingress and sulfate attack, but highest resistance to carbonation. The four MK mortars have the highest chloride and sulfate resistance, but a poor resistance towards carbonation. For the wPc’ – MT – LS and the wPc’ – G – LS blends, similar durability performances are observed for the P’, MT and G mortars, despite the variation in mix design. The

128

Chapter 9: Conclusions

performance for the L’ mortar is better than for the L mortar, as expected due to the lower water/binder ratio for the L’ mortar. It seems that the measured high chloride and sulfate resistance for the four MK mortars is related to the refinement in pore structures. However, the high chloride and sulfate resistance is also observed for the MT and G blends, which have similar pore structures as the P’ and L’ mortars. This finding contradicts most of the conclusions published in earlier studies that the addition of SCMs in concrete improves the durability performance via refinement of the pore structures. A speculative assumption is that the ionic interactions may affect the ion transport, which needs further research. Nevertheless, it is clear that the sulfate resistance can be well explained by the mass variation, which shows that no significant amounts of sulfate ions penetrate into the MK, MT and G mortars. Chloride binding data shows that the major chloride binding for the samples exposed to the NaCl solution is attributed to the formation of Friedel’s salt. The results show that the addition of MK in blended cement increases the chloride binding which is two times as high as in the P sample in the NaCl solution. However, the chloride profile results show that the chloride bound in Friedel’s salt is more than ten times lower than the maximum total chloride content for the P and L mortars, which quantitatively indicates a low contribution of chloride binding to remove the chloride from the pore solution. Thus, it can be assumed that the high chloride resistance for the MK mortars is mainly attributed to the refinement in pore structures. Moreover, the chloride binding does have an impact on the chloride ingress, which can be seen from the total chloride profiles for the MT and G mortars, as compared to the P’ and L’ mortars, because the impact of the similar pore structures can be eliminated. Regardless the type of mortars, the carbonation rates can be described by introducing the CO2 capture capacity proposed by (Leemann et al., 2015), which is expressed by the ratio between water added during production and the amount of reactive CaO present in the binder (w/CaOreactive). This empirical relationship relies on the fact found in the present study that carbonation of different mortars proceeds at a similar reaction rate in the different mortars, although they have different total potential CO2 binding capacities and contain different phases to bind CO2. In general, the data obtained in the present study confirm that the CO2 binding capacity, the porosity, and the capillary condensation are the decisive parameters governing the carbonation depth. 129


All the durability performance data show that the compressive strength is not a good indicator to predict the durability performance of the studied mortars. For the MK mortars, significant improvement in chloride and sulfate resistance is observed whereas the reduction in compressive strength for the MK mortars as compared to the P mortar is minor. For the P’, L’, MT and G mortars with comparable compressive strength, a diverse durability performance is observed between these mortars.

130

References

REFERENCES

131


Al-Akhras NM (2006) Durability of metakaolin concrete to sulfate attack. Cement and Concrete Research 36:1727-1734. Al-Dulaijan S, Maslehuddin M, Al-Zahrani M, Sharif A, Shameem M, Ibrahim M (2003) Sulfate resistance of plain and blended cements exposed to varying concentrations of sodium sulfate. Cement and Concrete Composites 25:429-437. Andrade C, Diez JM, Alaman A, Alonso C (1995) Mathematical modelling of electrochemical chloride extraction from concrete. Cement and Concrete Research 25:727-740. Antoni M (2013) Investigation of cement substitution by blends of calcined clays and limestone (PhD thesis). École Polytechnique Fédérale de Lausanne. Antoni M, Rossen J, Martirena F, Scrivener K (2012) Cement substitution by a combination of metakaolin and limestone. Cement and Concrete Research 42:1579-1589. Armaghani JM, Larsen TJ, Romano DC (1992) Aspects of concrete strength and durability. Transportation Research Record p.63 - 69. Arya C, Buenfeld NR, Newman JB (1990) Factors influencing chloride-binding in concrete. Cement and Concrete Research 20:291-300. Arya C, Xu Y (1995) Effect of cement type on chloride binding and corrosion of steel in concrete. Cement and Concrete Research 25:893-902. Baghabra Al-Amoudi OS, Al-Kutti WA, Ahmad S, Maslehuddin M (2009) Correlation between compressive strength and certain durability indices of plain and blended cement concretes. Cement and Concrete Composites 31:672-676. Baroghel-Bouny V, Wang X, Thiery M, Saillio M, Barberon F (2012) Prediction of chloride binding isotherms of cementitious materials by analytical model or numerical inverse analysis. Cement and Concrete Research 42:1207-1224. Beaudoin JJ, Ramachandran VS, Feldman RF (1990) Interaction of chloride and CSH. Cement and Concrete Research 20:875-883. Bellmann F, Stark J (2007) Prevention of thaumasite formation in concrete exposed to sulphate attack. Cement and Concrete Research 37:1215-1222. Bellmann F, Stark J (2008) The role of calcium hydroxide in the formation of thaumasite. Cement and Concrete Research 38:1154-1161. Bentz DP, Garboczi EJ, Lu Y, Martys N, Sakulich AR, Weiss WJ (2013) Modeling of the influence of transverse cracking on chloride penetration into concrete. Cement and Concrete Composites 38:65-74. Bertolini L, Elsener B, Pedeferri P, Redaelli E, Polder RB (2013) Corrosion of steel in concrete: prevention, diagnosis, repair: John Wiley & Sons. Bhatty J, Taylor P (2006) Sulfate resistance of concrete using blended cements or supplementary cementitious materials. Birnin-Yauri U, Glasser F (1998) Friedel’s salt, Ca2Al(OH)6(Cl,OH)·2H2O: its solid solutions and their role in chloride binding. Cement and Concrete Research 28:1713-1723.

132

References

Blanco-Varela M, Aguilera J, Martinez-Ramirez S (2006) Effect of cement C3A content, temperature and storage medium on thaumasite formation in carbonated mortars. Cement and Concrete Research 36:707-715. Borges PH, Costa JO, Milestone NB, Lynsdale CJ, Streatfield RE (2010) Carbonation of CH and C–S–H in composite cement pastes containing high amounts of BFS. Cement and Concrete Research 40:284-292. Brečević L, Nielsen AE (1989) Solubility of amorphous calcium carbonate. Journal of Crystal Growth 98:504-510. Brown P, Taylor H (1999) The role of ettringite in external sulfate attack. Materials Science of Concrete: Sulfate Attack Mechanisms, Marchand, J and JP Skalny, eds, The American Ceramic Society, Westerville OH 73-98. Bucher R, Cyr M, Escadeillas G (2015) Carbonation of Blended Binders Containing Metakaolin. In: Calcined Clays for Sustainable Concrete, pp 27-33: Springer. Build-NT-443 (1995) Concrete, hardened: accelerated chloride penetration. Nordtest Method. Canut MMC (2012) Pore structure in blended cement pastes (PhD thesis). Technical University of Denmark, Department of Civil Engineering. Castellote M, Fernandez L, Andrade C, Alonso C (2009) Chemical changes and phase analysis of OPC pastes carbonated at different CO2 concentrations. Materials and Structures 42:515-525. CEMBUREAU (2014) The European Cement Association Activity Report. CENTC51:WG12/TG1 (1994) Draft European Standard, Methods of testing cement determination of teh resistance of cements to attack by sulfate solution or by seawater. Document N40. Chang C-F, Chen J-W (2006) The experimental investigation of concrete carbonation depth. Cement and Concrete Research 36:1760-1767. Chen JJ, Thomas JJ, Taylor HF, Jennings HM (2004) Solubility and structure of calcium silicate hydrate. Cement and Concrete Research 34:1499-1519. Chindaprasirt P, Homwuttiwong S, Sirivivatnanon V (2004) Influence of fly ash fineness on strength, drying shrinkage and sulfate resistance of blended cement mortar. Cement and Concrete Research 34:1087-1092. Chowaniec O (2012) Limestone addition in cement (PhD thesis). École Polytechnique Fédérale de Lausanne. Cohen M, Richards C (1982) Effects of the particle sizes of expansive clinker on strength-expansion characteristics of type K expansive cements. Cement and Concrete Research 12:717-725. Cohen MD, Mather B (1991) Sulfate attack on concrete: research needs. ACI Materials Journal 88. Coleman NJ, Page CL (1997) Aspects of the pore solution chemistry of hydrated cement pastes containing metakaolin. Cement and Concrete Research 27:147154. Dai Z (2015) Solid-state 27Al and 29Si MAS NMR investigations of white Portland cement - metakaolin blends (PhD thesis). Aarhus University. 133


Dai Z, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015) Investigation of blended systems of supplementary cementitious materials with white Portland cement and limestone (manuscript in preparation). Dai Z, Tran TT, Skibsted J (2014) Aluminum Incorporation in the C–S–H Phase of White Portland Cement–Metakaolin Blends Studied by 27Al and 29Si MAS NMR Spectroscopy. Journal of the American Ceramic Society. De Ceukelaire L, Van Nieuwenburg D (1993) Accelerated carbonation of a blastfurnace cement concrete. Cement and Concrete Research 23:442-452. De Weerdt K, Colombo A, Coppola L, Justnes H, Geiker M (2015) Impact of the associated cation on chloride binding of Portland cement paste. Cement and Concrete Research 68:196-202. De Weerdt K, Haha MB, Le Saout G, Kjellsen K, Justnes H, Lothenbach B (2011a) Hydration mechanisms of ternary Portland cements containing limestone powder and fly ash. Cement and Concrete Research 41:279-291. De Weerdt K, Justnes H, Geiker MR (2014) Changes in the phase assemblage of concrete exposed to sea water. Cement and Concrete Composites 47:53-63. De Weerdt K, Kjellsen K, Sellevold E, Justnes H (2011b) Synergy between fly ash and limestone powder in ternary cements. Cement and Concrete Composites 33:3038. De Weerdt K, Muller A, Geiker M (unpublished) PF-Method for cementitious materials. unpublished. Dhir R, Jones M, Elghaly A (1993) PFA concrete: exposure temperature effects on chloride diffusion. Cement and concrete research 23:1105-1114. Elakneswaran Y, Iwasa A, Nawa T, Sato T, Kurumisawa K (2010) Ion-cement hydrate interactions govern multi-ionic transport model for cementitious materials. Cement and Concrete Research 40:1756-1765. Elakneswaran Y, Nawa T, Kurumisawa K (2009) Electrokinetic potential of hydrated cement in relation to adsorption of chlorides. Cement and Concrete Research 39:340-344. EN196-1 (2005) Methods of testing cement–Part 1: Determination of strength. European Committee for standardization 26. Galan I, Glasser F, Baza D, Andrade C (2015a) Assessment of the protective effect of carbonation on portlandite crystals. Cement and Concrete Research 74:68-77. Galan I, Glasser FP (2015) Chloride in cement. Advances in Cement Research 27:6397. Galan I, Perron L, Glasser FP (2015b) Impact of chloride-rich environments on cement paste mineralogy. Cement and Concrete Research 68:174-183. Garg N (2015) Structure, reactivity, and dissolution of calcined clays by solid-state NMR (PhD thesis). Aarhus University. Garg N, Skibsted J (2014) Thermal Activation of a Pure Montmorillonite Clay and its Reactivity in Cementitious Systems. The Journal of Physical Chemistry C 21:11464-11477.

134

References

Geiker M (1983) Studies of Portland cement hydration by measurements of chemical shrinkage and a systematic evaluation of hydration curves by means of the dispersion model (PhD thesis). vol. PhD: Technical University of Denmark. Geng J, Easterbrook D, Li L-y, Mo L-w (2015) The stability of bound chlorides in cement paste with sulfate attack. Cement and Concrete Research 68:211-222. Glass G, Buenfeld N (2000) The influence of chloride binding on the chloride induced corrosion risk in reinforced concrete. Corrosion Science 42:329-344. Glasser F, Kindness A, Stronach S (1999) Stability and solubility relationships in AFm phases: Part I. Chloride, sulfate and hydroxide. Cement and Concrete Research 29:861-866. Glasser F, Pedersen J, Goldthorpe K, Atkins M (2005) Solubility reactions of cement components with NaCl solutions: I. Ca (OH) 2 and CSH. Advances in cement research 17:57-64. Gonzalez M, Irassar E (1997) Ettringite formation in low C3A Portland cement exposed to sodium sulfate solution. Cement and Concrete Research 27:1061-1072. Goto S, Suenaga K, Kado T, Fukuhara M (1995) Calcium silicate carbonation products. Journal of the American Ceramic Society 78:2867-2872. Grantham M, Majorana C, Salomoni V (2009) Concrete Solutions: CRC Press. Grishchenko RO, Emelina AL, Makarov PY (2013) Thermodynamic properties and thermal behavior of Friedel's salt. Thermochimica Acta 570:74-79. Groves G, Rodway D, Richardson I (1990) The carbonation of hardened cement pastes. Advances in Cement research 3:117-125. Groves GW, Brough A, Richardson IG, Dobson CM (1991) Progressive changes in the structure of hardened C3S cement pastes due to carbonation. Journal of the American Ceramic Society 74:2891-2896. Gu P, Xu Z, Xie P, Beaudoin J (1993) Application of AC impedance techniques in studies of porous cementitious materials:(I): Influence of solid phase and pore solution on high frequency resistance. Cement and Concrete Research 23:531540. Halamickova P, Detwiler RJ, Bentz DP, Garboczi EJ (1995) Water permeability and chloride ion diffusion in Portland cement mortars: relationship to sand content and critical pore diameter. Cement and Concrete Research 25:790-802. Hawkins P, Tennis PD, Detwiler RJ (2003) The use of limestone in Portland cement: a state-of-the-art review. Herfort D (2004) Challenges of cement production. In: Anna Maria 2004 Workshop on Cement and Concrete Anna Maria Island, Florida, USA. Hirao H, Yamada K, Takahashi H, Zibara H (2005) Chloride binding of cement estimated by binding isotherms of hydrates. Journal of Advanced Concrete Technology 3:77-84. Houst YF, Wittmann FH (1994) Influence of porosity and water content on the diffusivity of CO2 and O2 through hydrated cement paste. Cement and Concrete Research 24:1165-1176.

135


Houst YF, Wittmann FH (2002) Depth profiles of carbonates formed during natural carbonation. Cement and Concrete Research 32:1923-1930. Ipavec A, Vuk T, Gabrovšek R, Kaučič V (2013) Chloride binding into hydrated blended cements: The influence of limestone and alkalinity. Cement and Concrete Research 48:74-85. Irassar E, Bonavetti V, Trezza M, Gonzalez M (2005) Thaumasite formation in limestone filler cements exposed to sodium sulphate solution at 20 oC. Cement and Concrete Composites 27:77-84. Irassar E, Di Maio A, Batic O (1996) Sulfate attack on concrete with mineral admixtures. Cement and Concrete Research 26:113-123. Jain JA, Neithalath N (2010) Chloride transport in fly ash and glass powder modified concretes–influence of test methods on microstructure. Cement and Concrete Composites 32:148-156. Jensen OM (1999) Chloride ingress in cement paste and mortar measured by Electron Probe Micro Analysis: Department of Structural Engineering, Technical University of Denmark. Johannesson B, Utgenannt P (2001) Microstructural changes caused by carbonation of cement mortar. Cement and Concrete Research 31:925-931. Juenger MC, Siddique R (2015) Recent advances in understanding the role of supplementary cementitious materials in concrete. Cement and Concrete Research. Justnes H (1998) A review of chloride binding in cementitious systems. Nordic Concrete Research Publications 21:48-63. Kaufmann J (2010) Pore space analysis of cement-based materials by combined Nitrogen sorption–Wood’s metal impregnation and multi-cycle mercury intrusion. Cement and Concrete Composites 32:514-522. Kaufmann J, Loser R, Leemann A (2009) Analysis of cement-bonded materials by multi-cycle mercury intrusion and nitrogen sorption. Journal of Colloid and Interface Science 336:730-737. Khan M, Lynsdale C, Waldron P (2000) Porosity and strength of PFA/SF/OPC ternary blended paste. Cement and Concrete Research 30:1225-1229. Khan MI, Lynsdale CJ (2002) Strength, permeability, and carbonation of highperformance concrete. Cement and Concrete Research 32:123-131. Khatib JM, Wild S (1998) Sulphate Resistance of Metakaolin Mortar. Cement and Concrete Research 28:83-92. Khunthongkeaw J, Tangtermsirikul S, Leelawat T (2006) A study on carbonation depth prediction for fly ash concrete. Construction and Building Materials 20:744-753. Kjellsen KO, Detwiler RJ, Gjørv OE (1991) Development of microstructures in plain cement pastes hydrated at different temperatures. Cement and Concrete Research 21:179-189. Kulik D, Tits J, Wieland E (2007) Aqueous-solid solution model of strontium uptake in CSH phases. Geochim Cosmochim Acta 71:A530.

136

References

Kulik DA (2011) Improving the structural consistency of CSH solid solution thermodynamic models. Cement and Concrete Research 41:477-495. Kulik DA, Wagner T, Dmytrieva SV, Kosakowski G, Hingerl FF, Chudnenko KV, Berner UR (2013) GEM-Selektor geochemical modeling package: revised algorithm and GEMS3K numerical kernel for coupled simulation codes. Computational Geosciences 17:1-24. Kumar R, Bhattacharjee B (2003) Porosity, pore size distribution and in situ strength of concrete. Cement and Concrete Research 33:155-164. Kunther W, Lothenbach B, Scrivener KL (2013) On the relevance of volume increase for the length changes of mortar bars in sulfate solutions. Cement and Concrete Research 46:23-29. Labbez C, Jönsson B, Pochard I, Nonat A, Cabane B (2006) Surface charge density and electrokinetic potential of highly charged minerals: Experiments and Monte Carlo simulations on calcium silicate hydrate. The Journal of Physical Chemistry B 110:9219-9230. Labbez C, Pochard I, Jönsson B, Nonat A (2011) CSH/solution interface: Experimental and Monte Carlo studies. Cement and Concrete Research 41:161-168. Lachowski E, Barnett S, Macphee D (2003) Transmission electron optical study of ettringite and thaumasite. Cement and Concrete Composites 25:819-822. Lambert P, Page C, Short N (1985) Pore solution chemistry of the hydrated system tricalcium silicate/sodium chloride/water. Cement and Concrete Research 15:675-680. Lamond JF (2006) Significance of tests and properties of concrete and concrete-making materials: ASTM International. Lange L, Hills C, Poole A (1996) The effect of accelerated carbonation on the properties of cement-solidified waste forms. Waste Management 16:757-763. Lee J-H, Yoon Y-S (2014) The effects of cementitious materials on the mechanical and durability performance of high-strength concrete. KSCE Journal of Civil Engineering 1-9. Lee S-T, Lee D-H, Kim D-K, Jung H-S, Park K-P, Kim S-S, Lee C-S (2008) Occurrence of thaumasite in lining concrete of old-tunnel structure. In: Geotechnical Engineering for Disaster Mitigation and Rehabilitation, pp 860865: Springer. Leemann A, Nygaard P, Kaufmann J, Loser R (2015) Relation between carbonation resistance, mix design and exposure of mortar and concrete. Cement and Concrete Composites 62:33-43. Loser R, Lothenbach B, Leemann A, Tuchschmid M (2010) Chloride resistance of concrete and its binding capacity–Comparison between experimental results and thermodynamic modeling. Cement and Concrete Composites 32:34-42. Lothenbach B, Bary B, Le Bescop P, Schmidt T, Leterrier N (2010) Sulfate ingress in Portland cement. Cement and Concrete Research 40:1211-1225. Lothenbach B, Durdzinski P, De Weerdt K (2016) Thermogravimetric analysis. Chater 5 in: A Practical Guide to Microstructural Analysis of Cementitious Materials.

137


Karen Scrivener, Ruben Snellings, and Barbara Lothenbach (Eds.): Taylor & Francis. Lothenbach B, Le Saout G, Gallucci E, Scrivener K (2008a) Influence of limestone on the hydration of Portland cements. Cement and Concrete Research 38:848-860. Lothenbach B, Le Saout G, Haha MB, Figi R, Wieland E (2012) Hydration of a lowalkali CEM III/B–SiO2 cement (LAC). Cement and Concrete Research 42:410423. Lothenbach B, Matschei T, Möschner G, Glasser FP (2008b) Thermodynamic modelling of the effect of temperature on the hydration and porosity of Portland cement. Cement and Concrete Research 38:1-18. Lothenbach B, Nonat A (2015) Calcium silicate hydrates: Solid and liquid phase composition. Cement and Concrete Research. Lothenbach B, Scrivener K, Hooton RD (2011) Supplementary cementitious materials. Cement and Concrete Research 41:1244-1256. Luo R, Cai Y, Wang C, Huang X (2003) Study of chloride binding and diffusion in GGBS concrete. Cement and Concrete Research 33:1-7. Luping T, Nilsson L-O (1993) Chloride binding capacity and binding isotherms of OPC pastes and mortars. Cement and Concrete Research 23:247-253. Macphee DE, Barnett SJ (2004) Solution properties of solids in the ettringite— thaumasite solid solution series. Cement and Concrete Research 34:1591-1598. Martın-Pérez B, Zibara H, Hooton R, Thomas M (2000) A study of the effect of chloride binding on service life predictions. Cement and Concrete Research 30:1215-1223. Matschei T, Lothenbach B, Glasser FP (2007) The role of calcium carbonate in cement hydration. Cement and Concrete Research 37:551-558. McPolin D, Basheer P, Long A (2009) Carbonation and pH in mortars manufactured with supplementary cementitious materials. Journal of Materials in Civil Engineering 21:217-225. Mehta P (1973) Mechanism of expansion associated with ettringite formation. Cement and Concrete Research 3:1-6. Mehta PK, Monteiro PJ (2006) Concrete: microstructure, properties, and materials: McGraw-Hill New York. Min D, Mingshu T (1994) Formation and expansion of ettringite crystals. Cement and Concrete Research 24:119-126. Mirvalad S, Nokken M (2015) Minimum SCM requirements in mixtures containing limestone cement to control thaumasite sulfate attack. Construction and Building Materials 84:19-29. Morandeau A, Thiery M, Dangla P (2014) Investigation of the carbonation mechanism of CH and C-S-H in terms of kinetics, microstructure changes and moisture properties. Cement and Concrete Research 56:153-170. Morandeau A, Thiéry M, Dangla P (2015) Impact of accelerated carbonation on OPC cement paste blended with fly ash. Cement and Concrete Research 67:226-236.

138

References

Muller A, Scrivener K, Gajewicz A, McDonald P (2013) Use of bench-top NMR to measure the density, composition and desorption isotherm of C–S–H in cement paste. Microporous and Mesoporous Materials 178:99-103. Müllauer W, Beddoe RE, Heinz D (2013) Sulfate attack expansion mechanisms. Cement and Concrete Research 52:208-215. Najimi M, Sobhani J, Pourkhorshidi A (2011) Durability of copper slag contained concrete exposed to sulfate attack. Construction and Building Materials 25:1895-1905. Ngala V, Page C (1997) Effects of carbonation on pore structure and diffusional properties of hydrated cement pastes. Cement and Concrete Research 27:9951007. Nishikawa T, Suzuki K, Ito S, Sato K, Takebe T (1992) Decomposition of synthesized ettringite by carbonation. Cement and Concrete Research 22:6-14. Odler I, Colán-Subauste J (1999) Investigations on cement expansion associated with ettringite formation. Cement and Concrete Research 29:731-735. Ogino T, Suzuki T, Sawada K (1987) The formation and transformation mechanism of calcium carbonate in water. Geochimica et Cosmochimica Acta 51:2757-2767. Page C, Vennesland Ø (1983) Pore solution composition and chloride binding capacity of silica-fume cement pastes. Matériaux et Construction 16:19-25. Papadakis VG (2000) Effect of supplementary cementing materials on concrete resistance against carbonation and chloride ingress. Cement and Concrete Research 30:291-299. Papadakis VG, Vayenas CG, Fardis MN (1991) Fundamental modeling and experimental investigation of concrete carbonation. ACI Materials Journal 88:363-373. Parrot L, Killoh D (1984) Prediction of cement hydration. In: Proceedings of the British Ceramic Society, p 41. Parrott L, Killoh D (1989) Carbonation in a 36 year old, in-situ concrete. Cement and Concrete Research 19:649-656. Paul G, Boccaleri E, Buzzi L, Canonico F, Gastaldi D (2015) Friedel's salt formation in sulfoaluminate cements: A combined XRD and 27Al MAS NMR study. Cement and Concrete Research 67:93-102. Pihlajavaara S (1968) Some results of the effect of carbonation on the porosity and pore size distribution of cement paste. Matériaux et Construction 1:521-527. Ping X, Beaudoin JJ (1992) Mechanism of sulphate expansion I. Thermodynamic principle of crystallization pressure. Cement and concrete research 22:631-640. Plumber L, Busenberg E (1982) The solubilities of calcite, aragonite and vaterite in CO2-H2O solutions between 0 and 90 °C, and an evaluation of the aqueous model for the system CaCO3-CO2-H2O. Geochimica et Cosmochimica Acta 46:1011-1040. Plusquellec G, Nonat A, Pochard I (2012) Anion uptake by calcium silicate hydrate. Plusquellec G, Pochard I, Nonat A (2013) Anions uptake by calcium silicate hydrates: influence of type of counter-ions and temperature. 139


Poon CS, Kou SC, Lam L (2006) Compressive strength, chloride diffusivity and pore structure of high performance metakaolin and silica fume concrete. Construction and Building Materials 20:858-865. Poulsen SL, Jakobsen HJ, Skibsted J (2009) Methodologies for measuring the degree of reaction in Portland cement blends with supplementary cementitious materials by 27 Al and 29 Si MAS NMR spectroscopy. IBAUSIL. Rabehi M, Mezghiche B, Guettala S (2013) Correlation between initial absorption of the cover concrete, the compressive strength and carbonation depth. Construction and Building Materials 45:123-129. Rahman A, Glasser F (1989) Comparative studies of the carbonation of hydrated cements. Advances in Cement Research 2:49-54. Ramachandran VS (1971) Possible states of chloride in the hydration of tricalcium silicate in the presence of calcium chloride. Matériaux et Construction 4:3-12. Ramezanianpour AM, Hooton RD (2013) Thaumasite sulfate attack in Portland and Portland-limestone cement mortars exposed to sulfate solution. Construction and Building Materials 40:162-173. Roberts M (1962) Effect of calcium chloride on the durability of pre-tensioned wire in prestressed concrete. Magazine of Concrete Research 14:143-154. Rostami V, Shao Y, Boyd AJ, He Z (2012) Microstructure of cement paste subject to early carbonation curing. Cement and Concrete Research 42:186-193. Rozière E, Loukili A, El Hachem R, Grondin F (2009) Durability of concrete exposed to leaching and external sulphate attacks. Cement and Concrete Research 39:1188-1198. Sabir BB, Wild S, Bai J (2001) Metakaolin and calcined clays as pozzolans for concrete: a review. Cement and Concrete Composites 23:441-454. Santhanam M, Cohen MD, Olek J (2003) Effects of gypsum formation on the performance of cement mortars during external sulfate attack. Cement and Concrete Research 33:325-332. Šauman Z (1971) Carbonization of porous concrete and its main binding components. Cement and Concrete Research 1:645-662. Scherer GW (1999) Crystallization in pores. Cement and Concrete Research 29:13471358. Scherer GW (2004) Stress from crystallization of salt. Cement and Concrete Research 34:1613-1624. Schmidt T, Lothenbach B, Romer M, Neuenschwander J, Scrivener K (2009) Physical and microstructural aspects of sulfate attack on ordinary and limestone blended Portland cements. Cement and Concrete Research 39:1111-1121. Schmidt T, Lothenbach B, Romer M, Scrivener K, Rentsch D, Figi R (2008) A thermodynamic and experimental study of the conditions of thaumasite formation. Cement and Concrete Research 38:337-349. Schneider M, Romer M, Tschudin M, Bolio H (2011) Sustainable cement production— present and future. Cement and Concrete Research 41:642-650.

140

References

Scrivener K, Favier A (2015) (Eds.). Calcined Clays for Sustainable Concrete: Proceedings of the 1st International Conference on Calcined Clays for Sustainable Concrete: Springer. Scrivener KL, Juilland P, Monteiro PJM (2015) Advances in understanding hydration of Portland cement. Cement and Concrete Research 78, Part A:38-56. Sellevold EJ, Farstad T (2005) The PF-method–A simple way to estimate the w/c-ratio and air content of hardened concrete. In: Proceedings of ConMat’05 and Mindess Symposium Vancouver, Canada: The University of British Colombia ISBN 0-88865-810-0. Sevelsted TF, Skibsted J (2015) Carbonation of C–S–H and C–A–S–H samples studied by 13C, 27Al and 29Si MAS NMR spectroscopy. Cement and Concrete Research 71:56-65. Siddique R, Klaus J (2009) Influence of metakaolin on the properties of mortar and concrete: A review. Applied Clay Science 43:392-400. Skaropoulou A, Tsivilis S, Kakali G, Sharp J, Swamy R (2009) Thaumasite form of sulfate attack in limestone cement mortars: A study on long term efficiency of mineral admixtures. Construction and Building Materials 23:2338-2345. Skibsted J, Jakobsen HJ, Hall C (1995) Quantification of calcium silicate phases in Portland cements by 29Si MAS NMR spectroscopy. Journal of the Chemical Society, Faraday Transactions 91:4423-4430. Smyth JR, McCormick TC (1995) Crystallographic data for minerals. Mineral Physics and Crystallography: A Handbook of Physical Constants 2:1-17. Steenberg M, Herfort D, Poulsen S, Skibsted J, Damtoft J (2011) Composite cement based on Portland cement clinker, limestone and calcined clay. In: 13th International Congress of the Chemistry of Cement, p 97 Madrid. Suryavanshi A, Scantlebury J, Lyon S (1996) Mechanism of Friedel's salt formation in cements rich in tri-calcium aluminate. Cement and Concrete Research 26:717727. Suryavanshi A, Swamy RN (1996) Stability of Friedel's salt in carbonated concrete structural elements. Cement and Concrete Research 26:729-741. Tang L, Nilsson L-O, Basheer PM (2011) Resistance of concrete to chloride ingress: Testing and modelling: CRC Press. Taylor H (1994) Sulfate reactions in concrete--microstructural and chemical aspects. Ceramic Transactions, 40 pp 61. Taylor HF (1997) Cement chemistry: Thomas Telford. Thiery M, Villain G, Dangla P, Platret G (2007) Investigation of the carbonation front shape on cementitious materials: Effects of the chemical kinetics. Cement and Concrete Research 37:1047-1058. Thiery M, Villain G, Platret G (2003) Effect of carbonation on density, microstructure and liquid water saturation of concrete. In: Proc 9th Eng Conf on Advances in Cement and Concrete, USA (Copper Mountain), pp 481-490. Thomas M, Hooton R, Scott A, Zibara H (2012) The effect of supplementary cementitious materials on chloride binding in hardened cement paste. Cement and Concrete Research 42:1-7. 141


Tian B, Cohen MD (2000) Does gypsum formation during sulfate attack on concrete lead to expansion? Cement and Concrete Research 30:117-123. Tritthart J (1989) Chloride binding in cement II. The influence of the hydroxide concentration in the pore solution of hardened cement paste on chloride binding. Cement and Concrete Research 19:683-691. Tsivilis S, Batis G, Chaniotakis E, Grigoriadis G, Theodossis D (2000) Properties and behavior of limestone cement concrete and mortar. Cement and Concrete Research 30:1679-1683. Tsivilis S, Chaniotakis E, Badogiannis E, Pahoulas G, Ilias A (1999) A study on the parameters affecting the properties of Portland limestone cements. Cement and Concrete Composites 21:107-116. Tsivilis S, Kakali G, Skaropoulou A, Sharp J, Swamy R (2003) Use of mineral admixtures to prevent thaumasite formation in limestone cement mortar. Cement and Concrete Composites 25:969-976. Tumidajski PJ, Schumacher A, Perron S, Gu P, Beaudoin J (1996) On the relationship between porosity and electrical resistivity in cementitious systems. Cement and Concrete Research 26:539-544. Turcry P, Oksri-Nelfia L, Younsi A, Aït-Mokhtar A (2014) Analysis of an accelerated carbonation test with severe preconditioning. Cement and Concrete Research 57:70-78. Usdowski E (1982) Reactions and Equilibria in the Systems CO2-H2O and CaCO3— CO2—H2O (0°-50 °C) - A Review. Neues Jahrbuch für Mineralogie Abhandlungen 144:148-171. Vance K, Aguayo M, Oey T, Sant G, Neithalath N (2013) Hydration and strength development in ternary portland cement blends containing limestone and fly ash or metakaolin. Cement and Concrete Composites 39:93-103. Villain G, Platret G (2006) Two experimental methods to determine carbonation profiles in concrete. ACI materials Journal 103. Villain G, Thiery M, Platret G (2007) Measurement methods of carbonation profiles in concrete: thermogravimetry, chemical analysis and gammadensimetry. Cement and Concrete Research 37:1182-1192. Voglis N, Kakali G, Chaniotakis E, Tsivilis S (2005) Portland-limestone cements. Their properties and hydration compared to those of other composite cements. Cement and Concrete Composites 27:191-196. Vu DD, Stroeven P, Bui VB (2001) Strength and durability aspects of calcined kaolinblended Portland cement mortar and concrete. Cement and Concrete Composites 23:471-478. Wagner T, Kulik DA, Hingerl FF, Dmytrieva SV (2012) GEM-Selektor geochemical modeling package: TSolMod library and data interface for multicomponent phase models. The Canadian Mineralogist 50:1173-1195. Wowra O, Setzer M, Setzer M, Auberg R (1997) Sorption of chlorides on hydrated cement and C3S pastes. Frost Resistance of Concrete 147-153. Wray P, Scrivener K (2012) Straight talk with Karen Scrivener on cements, CO2 and sustainable development. American Ceramic Society Bulletin 91:47-50. 142

References

Yu C, Sun W, Scrivener K (2013) Mechanism of expansion of mortars immersed in sodium sulfate solutions. Cement and Concrete Research 43:105-111. Yu C, Sun W, Scrivener K (2015) Degradation mechanism of slag blended mortars immersed in sodium sulfate solution. Cement and Concrete Research 72:37-47. Yuan Q, Shi C, De Schutter G, Audenaert K, Deng D (2009) Chloride binding of cement-based materials subjected to external chloride environment–a review. Construction and Building Materials 23:1-13. Zhou Q, Glasser F (2000) Kinetics and mechanism of the carbonation of ettringite. Advances in Cement Research 12:131-136. Zhu Q, Jiang L, Chen Y, Xu J, Mo L (2012) Effect of chloride salt type on chloride binding behavior of concrete. Construction and Building Materials 37:512-517. Zibara H (2001) Binding of external chlorides by cement pastes. Zibara H, Hooton D, Yamada K, Thomas M (2002) Roles of cement mineral phases in chloride binding. Cement Science and Concrete Technology 56:384-391. Zibara H, Hooton R, Thomas M, Stanish K (2008) Influence of the C/S and C/A ratios of hydration products on the chloride ion binding capacity of lime-SF and limeMK mixtures. Cement and Concrete Research 38:422-426.

143


144

Appendix I: Hydrate Phase Assemblages

APPENDIX I: HYDRATE PHASE ASSEMBLAGES Thermodynamic modeling of the interactions of the studied blends with chloride and carbonate ions is based on predicted phase assemblages after 91 days of hydration. The hydration degree for C3S, C2S, MK and SF at different hydration time were quantified using the NMR deconvolution technique as presented in (Dai, 2015). The hydration kinetics for C3S and C2S are obtained by fitting the NMR data with Parrott and Killoh model (Parrot and Killoh, 1984). The hydration kinetics for MK and SF are obtained by fitting the NMR data with a four parameter logistic model: F(x) = D + ((A – D)/(1 + ((x/C)^B))), where A and D controls minimum and maximum asymptotes, respectively. C controls the inflection point and B controls the steepness of the curves. The hydration kinetics for the C3S, C2S, MK and SF phases are implemented into thermodynamic modelling as restraints to predict the phase assemblages over hydration time.

145


Figure I.I. Hydration kinetics implemented as restraints in the thermodynamic modelling, which obtained by fitting the NMR results measured by (Dai, 2015).

146

Appendix I: Hydrate Phase Assemblages

147


Figure I.II. Hydration phase assemblages for white Portland cement – metakaolin – silica fume – limestone blends in volume percentage as a function of hydration time.

148

Appendix II: Publications

APPENDIX II: PUBLICATIONS The copyright to all publications attached below belong to the respective publisher whenever such a copyright was transferred. Following is the record of publications, both conference and journal articles, which have been published for dissemination of the research. Most of the data from the conference articles is reproduced in the subsequent journal articles. Thus, journal articles can be considered as the final version of the research that is appropriate for reference and citation.

149


Conference Article I Z. Shi, M.R. Geiker, K. De Weerdt, B. Lothenbach, W. Kunther, S. Ferreiro Garzon, D. Herfort, J. Skibsted. Durability of Portland cement blends including calcined clay and limestone: interactions with sulfate, chloride and carbonate ions, in: Proceedings of the 1st International Conference on Calcined Clays for Sustainable Concrete, K. Scrivener, A. Favier, (Eds.), Springer, 2015, pp. 133-141. Conference Article II Z. Shi, B. Lothenbach, M.R. Geiker, J. Kaufmann, S. Ferreiro, J. Skibsted. Carbonation of Portland cement mortars including metakaolin and limestone. 14th International Congress on the Chemistry of Cement, Beijing, October 13 – 16, 2015. Conference Abstract I (accepted) Z. Shi, M.R. Geiker, K. De Weerdt, T.A. Østnor, B. Lothenbach, F. Winnefeld, J. Skibsted. Impact of cations on chloride binding in hydrated Portland cement – metakaolin – limestone blends. 4th International Workshop on Mechanisms and Modelling of Waste/Cement Interactions, Murten, Swizerland, May 22 – 25, 2016. Journal Article I (submitted) Z. Shi, B. Lothenbach, M.R. Geiker, J. Kaufmann, A. Leemann, S. Ferreiro, J. Skibsted. Experimental and thermodynamic modeling studies on carbonation of Portland cement metakaolin - limestone mortars. (Cement and Concrete Research, submitted October 2015) Journal Article II (in preparation) Z. Shi, M.R. Geiker, B. Lothenbach, K. De Weerdt, S. Ferreiro Garzón, K. EnemarkRasmussen, J. Skibsted. Role of Friedel's salt on chloride binding in chloride exposed Portland cement-based mortars. Journal Article III (in preparation) Z. Shi, M.R. Geiker, K. De Weerdt, T.A. Østnor, B. Lothenbach, F. Winnefeld, J. Skibsted. Chloride binding of Portland cement paste including metakaolin and limestone.

150

Conference article I

151


152


153


154


155


156


157


158


159


160

Conference article II

161


162


163


164


165


166


167


168


169


170

Conference abstract

171


172

Journal article I

Experimental studies and thermodynamic modeling of the carbonation of Portland cement, metakaolin and limestone mortars

Zhenguo Shia, Barbara Lothenbachb, Mette Rica Geikerc, Josef Kaufmannb, Andreas Leemannb, Sergio Ferreirod, Jørgen Skibsteda*

a. Department of Chemistry and Interdisciplinary Nanoscience Center (iNANO), Aarhus University, 8000 C Aarhus, Denmark b. Laboratory for Concrete & Construction Chemistry, Swiss Federal Laboratories for Materials Science and Technology (Empa), 8600 Dübendorf, Switzerland c. Department of Structural Engineering, Norwegian University of Science and Technology (NTNU), Trondheim 7491, Norway d. Aalborg Portland A/S, Cementir Holding S.p.A., 9100 Aalborg, Denmark

____________________ * Corresponding author. Department of Chemistry and Interdisciplinary Nanoscience Center (iNANO), Aarhus University, DK-8000 Aarhus C, Denmark. Tel: +4587155946; Fax: +45 8619 6199. E-mail address: [email protected] (J. Skibsted). 173


Abstract The carbonation of Portland cement, metakaolin and limestone mortars has been investigated after hydration for 91 days and exposure to 1% (v/v) CO2 at 20 oC/57% RH for 280 days. The apparent carbonation depths have been measured by phenolphthalein whereas mercury intrusion porosimetry (MIP), SEM, TGA and thermodynamic modeling have been used to study pore structure, CO2 binding capacity and phase assemblages. The Portland cement has the highest resistance to carbonation due to its higher CO2 binding capacity. The metakaolin and limestone blends have a lower, but comparable CO2 binding capacity, where the better carbonation resistance of the metakaolin blends is related to their finer pore structure and lower total porosity. MIP shows a coarsening of the pore threshold for all mortars but only small variations in total intruded porosity. Overall, the CO2 binding capacity, porosity, and capillary condensation are found to be the decisive parameters governing the carbonation rate.

Keywords: Carbonation (C); Portland cement (D); Metakaolin (D); CaCO3 (D); Thermodynamic calculations (B).

1. Introduction Significant research efforts focus on the development and characterization of new Portland cement blends including calcined clays (Scrivener and Favier, 2015) as supplementary cementitious materials (SCMs) with the principal aim of reducing CO2 emissions associated with Portland cement production. Partial replacement of Portland cement by SCMs such as fly ashes and slags represent a common route to reduce CO 2 emissions. This interest in calcined clays reflects both that slags and fly ashes may not be available in sufficient quantities in the future and more importantly, that clays are widely abundant in the Earth’s crust. Limestone represents another interesting material, which is commonly added in small amounts to Portland cements, where it increases the early strength, reduces the water demand and improves the rheology of the resulting concrete (Tsivilis et al., 1999, Herfort, 2004, Voglis et al., 2005, Lothenbach et al., 2008). Furthermore, limestone provides nucleation sites for the formation and growth of the calcium-silicate-hydrate (C-S-H) phase and it is also partially consumed during 174

Journal article I

hydration, resulting in the formation of calcium monocarboaluminate hydrate (Ca4Al2(OH)12CO3·5H2O) (Herfort, 2004, Matschei et al., 2007, Lothenbach et al., 2008). The combination of limestone with other SCMs has been used to develop ternary cement blends (De Weerdt et al., 2011). For example, a synergetic effect between metakaolin and limestone has been observed in ternary Portland cement blends, as seen by an increase in compressive strength (Steenberg M et al., 2011, Antoni et al., 2012). Carbonation of concrete is one of the important age-limiting factors for the durability of reinforced concrete structures. Given the intrinsically high pH in the pore solution of concrete, a thin passive layer around the steel bars is formed which protects the steel reinforcement against corrosion. However, carbonation of the cement hydrates, e.g., portlandite (CH) and the C-S-H phase, can result in a significant reduction of pH and to a certain level, where the protective layer on the steel bars is destroyed. This problem may become particularly important when SCMs are incorporated in cement blends, since several studies have reported that cement-based materials including SCMs exhibit poor carbonation resistance (Papadakis, 2000, Leemann et al., 2015, Morandeau et al., 2015). This underlines the research needs for carbonation studies of the new Portland cement – calcined clay – limestone blends before an industrial realization of these materials can take place. The pH change induced by carbonation of concrete is ascribed to carbonation reactions of portlandite (CH) and the C-S-H phase, which result in the formation of CaCO3 and CaCO3 and/or amorphous silica, respectively. Hydrated Portland cement – SCM blends contain generally a smaller amount of portlandite compared to hydrated pure Portland cements, since a part of portlandite has been consumed by reaction with the SCMs forming C-S-H phase. This lower amount of portlandite may account for the faster carbonation process for Portland cement – SCM blends, compared to pure Portland cement, as mentioned in earlier studies (Papadakis, 2000, Morandeau et al., 2015). However, it seems of minor importance for the carbonation resistance under accelerating conditions, if the CaO buffering the ingression of CO2 is present in portlandite or C-S-H (Leemann et al., 2015). A decisive material parameter for the carbonation resistance is the buffer capacity per volume of cement paste that can be expressed as the ratio between mixing water and CaO reacting with CO2 (Leemann et al., 2015). Moreover, the lack of portlandite cannot solely explain the variations in 175


apparent carbonation depths, when the depths are measured by the phenolphthalein indicator method, since several of the hydrate phases are not completely carbonated within the carbonation depths as revealed by thermogravimetric analysis (Chang and Chen, 2006, Thiery et al., 2007). Thus, it is necessary to compare the actual and potential CO2 binding of the different binders in order to better understand their carbonation performance. The carbonation profiles for pure Portland cement and Portland cement – fly ash blends have been thoroughly studied by various techniques, revealing that the carbonation depths do not necessarily exhibit a sharp reaction front (Thiery et al., 2007, Morandeau et al., 2015). However, the corresponding pH profiles are not known, since they are difficult to measure experimentally as a result of drying of the samples during carbonation. McPolin et al. (McPolin et al., 2009) have measured the pH profiles after carbonation by pore solution expression and leaching methods. However, these methods can only provide an estimate of the real pH values because partially carbonated large crystals of portlandite are cracked during the measurements which may change the pH of the pore solution. In addition, the phenolphthalein method indicates a gradual change in color from colorless to fuchsia, reflecting pH changes from 8.2 to 10.0 in solution. This gradual color change has not been clearly documented for the carbonation front in concrete or mortars measured by the phenolphthalein method. Thus, it is not clearly documented at which pH levels the phenolphthalein method reflects the carbonation depths in concrete. Another impact of carbonation on concrete is the changes in microstructure, originating from differences in molar volumes of the hydrated and carbonated phases. Predictions become complex when the molar volume of the C-S-H phase changes with carbonation as a result of changes of the Ca/Si ratio (Morandeau et al., 2014, Morandeau et al., 2015, Sevelsted and Skibsted, 2015). Moreover, the different polymorphs of CaCO3 also exhibit different unit cell volumes (Smyth and McCormick, 1995). Several studies have reported a reduction of the porosity for pure Portland cement (Pihlajavaara, 1968, Houst and Wittmann, 1994, Ngala and Page, 1997, Thiery et al., 2003, Morandeau et al., 2014, Leemann et al., 2015), whereas an increase of the total porosity has been measured by different techniques for Portland cement – SCM blends (De Ceukelaire and Van Nieuwenburg, 1993, Leemann et al., 2015). Morandeau et al. (Morandeau et 176

Journal article I

al., 2015) have reported that the total porosity decreases for a Portland cement paste with 60 vol.% of fly ash, however, its microstructure is rearranged and large capillary pores are created. The same work (Morandeau et al., 2015) also reviewed some earlier studies and stated that there is a shift in the porosity towards larger pore radii during carbonation as measured by mercury intrusion porosimetry. Johannesson et al. (Johannesson and Utgenannt, 2001) reported that the difference in pore size distribution is more pronounced than the difference in specific surface area for carbonated and noncarbonated Portland cement mortars. In addition, the presence of some micro-cracks has also been noticed as a result of the volume increase of the solid during carbonation (Lange et al., 1996). Thus, the earlier studies show several discrepancies, uncertainties and lack of explanations in the interpretation of the microstructural changes caused by carbonation. The present work is a part of a series of durability investigations of Portland cement – calcined clay – limestone blends, all focusing on a replacement level of 35 wt.% of Portland clinker (Shi et al., 2015). This paper focusses on the carbonation of Portland cement – metakaolin blends, with and without limestone, with the goal of determining the carbonation resistance for these calcined clay blends and key factors associated with the carbonation processes. For comparison, a reference of pure Portland cement and a limestone Portland cement with a 35 wt.% of replacement are also investigated. Analysis of the distinct phase assemblages and microstructures in the blends may form the basis for a better understanding of the differences in carbonation performance for Portland cement – SCM blends. The phenolphthalein spray method is used to measure carbonation depths whereas mercury intrusion porosimetry (MIP) and scanning electron microscopy (SEM) are applied to monitor the changes in microstructures due to carbonation. Thermogravimetric analysis (TGA) and thermodynamic modeling are employed to investigate the actual and potential CO2 binding for the different blends. Thermodynamic modeling is also used to examine changes in total porosity and pH in relation to phase changes. Finally, the carbonation performances of the studied blends are evaluated in terms of pore-structural changes, CO2 binding and calculated pH profiles.

177


2. Experimental 2.1 Materials The binders used in this study were made from a white Portland cement (wPc, CEM I 52.5 N) and two SCMs: metakaolin (MK) and limestone (LS). The wPc was obtained from Aalborg Portland A/S, Denmark, and included 3.1 wt.% LS, 4.1 wt.% gypsum and 1.9 wt.% free lime. The MK was produced in the laboratory from kaolinite (Kaolinite SupremeTM from Imerys Performance Minerals, UK) by thermal treatment at 550 oC for 20 h. The limestone was a Maastrichtian chalk from Rørdal, Northern Denmark. The chemical compositions of the starting materials, determined by X-ray fluorescence (XRF), and their physical properties are given in Table 1. The wPc contained 64.9 wt.% alite,16.9 wt.% belite and 7.8% C3A where the content of the silicate phases was determined by 29Si MAS NMR and the quantity of the aluminate phase by mass balance calculations (Dai et al., 2015). The C4AF phase was not taken into account as the small amount of iron is expected to be incorporated as guest-ions in the alite, belite, and calcium aluminate phases. The sand used for the mortars was a CEN reference sand (Normensand GmbH, Germany), which has a silica content of at least 98 wt.% and a density of 2650 kg/m3. A superplasticizer (SP, Glenium 27, BASF) was used to achieve similar flow for all mortars.

2.2 Mortar preparations The compositions of the binders (Table 2) targeted a replacement of 35 wt.% white Portland clinker by the SCMs. Considering the small amounts of LS and gypsum in the wPc this resulted in actual binder compositions with 31.9 wt.% replacement of the wPc. The blends were used to produce mortars with a constant water to binder ratio (w/b = 0.5) and binder to sand ratio (b/s = 1/3), both ratios by weight. The dosage of superplasticizer (SP/MK=0.07) was adjusted to achieve a flow within ±5% of the flow of the reference P mortar. The mortar bars were cast into 40 × 40 × 160 mm3 molds cured in a moist cabinet maintained at a temperature of 20 ± 1.0 oC and a relative humidity of not less than 90% for 24 hours. After demolding, the mortar bars were cured in demineralized water in sealed buckets with a water-solid ratio of 3:1 by volume at 20 oC for 91 days. 178

Journal article I

The degrees of hydration for alite, belite and MK in similar paste samples of the different blends have been determined by 29Si MAS NMR in another study (Dai et al., 2015), see Table 3. In the same study, the compressive strengths of mortar bars were determined after 28, 91 and 182 days of hydration (Table 3). The degrees of hydration reveal that more than 80% of alite has been consumed for all cement pastes after 28 days. The degrees of reaction for belite and MK are lower and reach values of approximately 35% and 50%, respectively, for the ML and M blends after 182 days of hydration, where the changes in degree of hydration are minimal from 28 days to 182 days. The degree of hydration for belite in the P and L blends increases significantly from 28 days to 182 days. However, the associated small increase in compressive strength of the corresponding mortars from 91 days to 182 days indicates only a minor development in physical properties during hydration for this period. Thus, these data indicates that only a small fraction of the constituent phases reacts in the curing period from 91 to 182 days, thereby justifying that it is assumed that only a minor degree of hydration takes place during the carbonation tests of the studied mortars which begins after 91 days of hydration.

2.3 CO2 exposure After 91 days of hydration under the conditions given above, the mortar bars were exposed to a controlled atmosphere with 1% (v/v) CO2 in an incubator at 20 oC without any pre-drying of the mortar bars. The CO2 concentration of 1% (v/v) was chosen since the formed phases are expected to be the same as those formed during natural carbonation (0.035 % (v/v)), as indicated in an earlier study (Castellote et al., 2009). The relative humidity (RH) was 57 ± 1%, which matches the RH range of ~40 – 70% for maximum carbonation reactions (Thiery et al., 2007).

2.4 Methods 2.4.1 Phenolphthalein spray method Carbonation depths were measured after 0, 7, 14, 21, 28, 56, 91 and 280 days of CO2 exposure. Slices with a thickness of 15 – 20 mm were taken from the mortar bars by splitting and the fresh surface was sprayed with a 1 wt.% phenolphthalein aqueous 179


solution. Photographs of the sprayed slices were taken with a Sony camera (SLT-A65V) and the carbonation depths were measured on the photographs after the color had stabilized on the slices (Fig. 1), using the commercial software “GetData Graph Digitizer”. Each depth reported in this study is the average of 10 to 15 measurements excluding the corners. In addition, the top and bottom surfaces were omitted to minimize effects from different surface qualities and mortar inhomogeneity originating from the casting.

2.4.2 Mercury intrusion porosimetry The pore structure of the mortars was characterized by mercury intrusion porosimetry (MIP). Two full subsequent intrusion cycles were performed using a combined Pascal 140/440 equipment on samples with grain sizes of 2 – 3 mm. Cut slices with a thickness of about 2 mm of the carbonated surface and slices from the non-carbonated center of the mortars were sampled and dried by solvent exchange with isopropanol for 2 days, followed by drying in an oven at 50 oC for 24 hours. The total porosity (intruded pore volume) and the pore connectivity (from the threshold pore size) were obtained from the intrusion curves. The threshold pore size was obtained from the intersection of the two tangents in the intrusion curve, as described in an earlier study (Canut, 2012), using a contact angle of 140o and a surface tension of 0.48 N/m (Kaufmann, 2010). The maximum applied pressure was 200 MPa, corresponding to a minimum pore radius of 3.6 nm.

2.4.3 Scanning electron microscopy Images of the non-carbonated and the carbonated areas of the mortars were acquired with an environmental scanning electron microscope (ESEM-FEG XL30) in the backscattering electron mode. Sections were impregnated in an epoxy resin and polished. The sections were studied in high vacuum (3.0·10-6 – 6.0·10-6 Torr) with an accelerating voltage of 12 kV and a beam current of 310 – 330 μA.

2.4.4 Thermogravimetric analysis

180

Journal article I

Cut slices of mortar including the carbonated and non-carbonated regions were investigated by thermogravimetric analysis (TGA) in order to quantify the CaCO3 and Ca(OH) 2 contents and to analyse the actual CO2 binding. First, the top and bottom sides were removed using a saw with a thick blade (for fast cutting). The remaining part, where the CO2 diffusion can be considered one-dimensional, was then cut layer-bylayer using a diamond saw with a thin blade (blade thickness of about 0.2 mm). Each layer was ground into a fine powder for the TGA measurements. Approximately 50 mg of the powder was heated at 20 oC/min from 30 to 980 oC. Quantification of Ca(OH) 2 and CaCO3 was performed by the tangential method (Lothenbach et al.). The results were normalized as mass percentages of ignited mortars relative to the weight at 800 oC and converted to molar amounts per 100 g of ignited mortars. The amount of CaCO3 originating solely from carbonation was calculated by subtracting the limestone content in the blends from the measured total amount of CaCO3. Furthermore, an estimate of the amount of CaCO3 resulting from carbonation of other phases, mainly the C-S-H phase, was obtained by a subsequent subtraction of the molar amount of Ca(OH) 2

2.4.5 Thermodynamic modeling Thermodynamic modeling was carried out using the Gibbs free energy minimization program, GEMS3 (Wagner et al., 2012, Kulik et al., 2013), which calculates the equilibrium phase assemblages in chemical systems from their total bulk elemental composition. The default databases were expanded with the CEMDATA07 database (Lothenbach et al., 2008b) and additional data for the C-S-H phase (Kulik, 2011). The data include solubility products of the solids relevant for cementitious systems and for C-S-H phases with different compositions. The changes in phase assemblages and total porosity upon exposure to CO2 were predicted for the different blends. The ingress of CO2 was mimicked by increasing the amount of air (constant CO2 concentration, 1% (v/v)) from the center to the surface of the mortars. The uptake of alkalis by the C-S-H phase is taking into account by employing an ideal solid-solution model between the CS-H phase and two hypothetical alkali silicate hydrates ([(KOH)2.5SiO2H2O]0.2 and [(NaOH)2.5SiO2H2O]0.2) (Lothenbach et al., 2012) as proposed by Kulik et al (Kulik et al., 2007).

181


3. Results 3.1 Apparent carbonation depths The apparent carbonation depths, as measured by the phenolphthalein spray method, are shown as a function of time (square root scale) in Fig. 2 for the four different mortars. No initial carbonation is observed for any of the mortars in accordance with the mortars being kept saturated prior to exposure. A measurable degree of carbonation is observed already after 7 days of exposure to 1% CO2 at 20 oC and 57% RH for the L, M and ML mortars but not for the P mortar. The reference mortar (P) shows a very slow progress in carbonation and thereby the highest resistance to carbonation whereas the L mortar, containing no MK, is most vulnerable to carbonation. Neither of the MK containing blends (M and ML) exhibits a high resistance to carbonation. The carbonation depths increase linearly as function of the square root of time within 280 days of carbonation. Linear regression of the data in Fig. 2 gives the carbonation rates (K) listed in Table 4 when a square root time dependency (x = K t1/2) is assumed. When the carbonation depths were measured, the mortars showed different rates of color evolution after spraying with phenolphthalein as illustrated in Fig. 1. For the P and L mortars, the change of color occurred immediately after spraying (within 1 second) whereas the change in color developed more slowly for the M and ML mortars. At the end, the pink color for all samples reached similar darkness. Fig. 3 shows the carbonation front indicated by phenolphthalein for the four different mortars and illustrates that the carbonation front is well defined for the M, ML and L mortars, whereas it is irregular for the P mortar as reflected by the larger variations in carbonation depths for this mortar in Fig. 2.

3.2 Pore structure 3.2.1 Mercury intrusion porosimetry (MIP) The MIP intrusion curves after the first intrusion cycle for the non-carbonated mortars are shown in Fig. 4. The samples were taken from the non-carbonated center of the mortars after 119 days of hydration (i.e., 91 days of curing in demineralized water and 28 days of CO2 exposure at 20 oC and 57% RH). The results (Fig. 4) show that the incorporation of metakaolin in the M and ML mortars results in a refined microstructure 182

Journal article I

with lower threshold pore size (breakthrough pore size), as compared to that observed for the reference mortar (P). However, their total intruded porosities are almost identical. The largest total intruded porosity is observed for the L mortar which also exhibits the highest threshold pore size. The differences in threshold pore sizes reflect the poor and good pore connectivity for the M/ML mortars and the L mortar, respectively. This observation can explain the long time required for color setting of the M mortar as shown in Fig. 1 (a similar long color setting has been observed for the ML mortar). The P and L mortars have relatively good pore connectivity, which causes an immediate contact between phenolphthalein and the alkaline environment. In contrast, the M and ML mortars have refined pore structures with low connectivity, resulting in an increased time of diffusion for phenolphthalein to reach the alkaline environment and thus, a slower color change. Ink-bottle pores are present in cement-based materials and they are connected to the external surface by smaller neck pores. These neck pore entrances may snap off the mercury during extrusion and for this reason two cycles of mercury intrusion/extrusion is used in the present work. All ink-bottle pores can be assumed to be filled at the beginning of the second intrusion cycle and the percolated (non-ink-bottle) pores will be filled during the second intrusion successively according to Washburn’s equation linking pore radius to pressure (Kaufmann et al., 2009, Kaufmann, 2010). The volume of the ink-bottle pores and their access radius can be obtained by subtraction of the second intrusion (percolated pores) from the first intrusion (total pores). The impact of carbonation on the intrusion curves from the first and second intrusion cycles is illustrated in Fig. 5 for all mortars. The results reveal that the incorporation of MK in the mortars results in a larger fraction of ink-bottle pores. In addition, it can be observed that carbonation causes an increased volume of percolated pores in the range 10 – 100 nm for the M and ML mortars. In contrast, a small decrease in the volume of percolated pores combined with an increase in pore size of the percolated pores is observed for the P and L mortars after carbonation. The impact of carbonation on the different mortars is compared Fig. 6, which summarizes the intruded pore volumes and threshold pore sizes derived from the data in Fig. 5. The results show that the pore threshold for all types of mortars is larger after carbonation, and at the same time that no significant changes of the total intruded 183


porosity are observed for any of the mortars. The intrusion curves for the second intrusion cycle reveal a coarsening effect of the threshold pore radius for all mortars and indicate a minor reduction of the larger intruded porosity for the P and L mortars, but an increase for the M and ML mortars. It has been reported, e.g. by Morandeau et al. (Morandeau et al., 2014), that the pore-size distribution in ordinary Portland cement paste is significantly reduced during accelerated carbonation (10% CO2, RH = 63%) as a result of clogging by the formed CaCO3 of the whole range of pores accessible by MIP (4 nm – 1 µm). This result is not consistent with the coarsening observed in the present work for all samples (Fig. 5 and Fig. 6). The coarsening effect is expected to facilitate carbonation, however, this is not observed for the P mortar, where a very shallow carbonation depth was observed even after 280 days.

3.2.2 Scanning electron microscopy (SEM) Although the molar volumes of non-carbonated and carbonated phases differ, the MIP analysis does not reveal major changes in the total porosity due to carbonation. Lowmagnification SEM BSE images of the P, L and ML mortars are shown in Fig. 7, where images of the M mortar have been excluded since they are very similar to those of the ML mortar. The images reveal that there is an increase in backscattering contrast for the carbonated cement paste of all samples. At the used magnification, the images indicate a reduction of the coarse porosity (pores > 5 µm) for the P and L mortars, and a small increase in coarse porosity for the ML (and M) mortars. This is in accordance with the changes in porosity derived from the 2nd MIP intrusion curves (Fig. 6). However, it should be kept in mind that the low-magnification SEM images in Fig. 7 do not allow for identification of pore sizes in the size range measured by MIP. Changes in porosity resulting from changes of the molar volume of solid phases will be further discussed in section 3.3.3 based on predictions from thermodynamic modeling.

3.3 Chemical properties 3.3.1 Thermogravimetric analysis (TGA)

184

Journal article I

Fig. 8 shows the first derivative (DTG) of the mass losses measured by TGA for samples collected at different depths of the mortars after 91 days of carbonation. For those mortars with significant apparent carbonation depths, samples were collected from two slices above the apparent carbonation depths and two slices below the apparent carbonation depths. The DTG curves show no differences below the apparent carbonation depths (i.e., in the non-carbonated part of the mortars), indicating that none of the phases were carbonated in this part. Three principal DTG peaks can be resolved for these samples, i.e., C-S-H/AFt/AFm phases (30 – 300 oC), portlandite (400 – 500 o

C) and CaCO3 (500 – 800 oC). The amounts of CaCO3 in the non-carbonated mortars

are generally consistent with the initial binder compositions (Table 2), also considering that the wPc included 3.1 wt.% of limestone. Moreover, the DTG curves for the samples close to mortar surfaces reveal a significantly higher amount of CaCO3 and a decrease in intensity for the thermal events associated with Ca(OH)2, C-S-H as well as the AFm and AFt phases. The TGA data recorded at different depth in the mortars indicate that portlandite is carbonated first, followed by monocarbonate and ettringite, while some of the C-S-H phase is still present. To which extent the C-S-H is carbonated is difficult to assess based on the TGA data, since only broad DTG peaks are associated with the C-SH phase. However, 29Si MAS NMR studies indicate a progressive decalcification of the C-S-H phase as a result of carbonation (Castellote et al., 2009, Sevelsted and Skibsted, 2015) , as discussed in detail in section 3.3.3. The DTG curves (Fig. 8) also show that the decomposition of the CaCO3 that is formed close to mortar surface, where the degree of carbonation is relatively higher, is extended to lower temperature for all mortars. This effect is most clearly seen for the M and ML mortars, where the lowest temperature goes down to about 250 oC. This feature has been reported in several other studies (Thiery et al., 2007, Borges et al., 2010, Rostami et al., 2012, Morandeau et al., 2014), and it is likely associated with the presence of poorly-crystalline CaCO3 (Šauman, 1971, Goto et al., 1995). Comparison of the DTG curves for the M and ML mortars with that observed for the P mortar indicates that the formation of poorly-crystalline CaCO3 is associated with carbonation of the C-S-H or AFt/AFm phases, since only a small amount of portlandite is present in the M and ML mortars. The decomposition of CaCO3 extends to temperatures as low as 250 oC for these two mortars. Such a broad temperature range has also been reported for Portland

185


cement paste samples, where higher CO2 concentrations up to 10% (v/v) were employed (Morandeau et al., 2014), which reasonably results in higher degrees of carbonation of the C-S-H or AFt/AFm phases. This assumption is also supported by comparison of the P and L mortars, where the L mortar can be seen as a diluted version of the P mortar, since only a minor part of the limestone has reacted during hydration (see section 3.3.3). As found in section 3.2.1, the L mortar has the highest intruded porosity and pore connectivity, which may provide a feasible access for CO2 diffusion into L mortar, resulting in a higher carbonation degree compared to the P mortar. The curves in Fig. 8 for the L mortar reveal that the decomposition of CaCO3 in the lower temperature range is more obvious than that observed for P mortar, implying that a larger fraction of the CS-H or AFt/AFm phases are carbonated for the L mortar. It should be noted that the samples were analyzed by TGA immediately after preparation and without any defined pre-drying. Thus, the first DTG peaks in Fig. 8 may be affected by moisture in the samples, which is especially visible for L mortar. However, the presence of small amounts of moisture will not cause any problems for the quantification of CH and CaCO3 from the DTA curves, as the amounts are normalized to the dry weight (TGA temperature at 800 oC). For the analyzed samples, the presence of portlandite particles with carbonated surfaces and non-carbonated centers has also been observed for the P mortar as shown in Fig. 8. This phenomenon has also been reported in other studies and it is related to the formation of dense calcite layers coating the portlandite crystals, thereby preventing its further carbonation (Groves et al., 1990, Groves et al., 1991, Morandeau et al., 2014).

3.3.2 Carbonation profiles The carbonation profiles for the different mortars after 28 and 91 days of carbonation are illustrated in Fig. 9. The data are obtained by quantification of CaCO3 and portlandite from the TGA curves (Fig. 8). The quantity of formed CaCO3 is estimated by subtracting the amount of CaCO3 originally incorporated as added limestone in the blends from the total CaCO3 determined by TGA. The CaCO3 and Ca(OH)2 contents have been normalized as mol percentage of 100 g ignited mortars (TGA temperature at 800 oC) in order to eliminate the impact of moisture content in the samples. The amount

186

Journal article I

of CaCO3 is quantified from the DTG curves from about 300 to 850 oC, since the decomposition of CaCO3 is known to extend to lower temperature (see section 3.3.1). The carbonation profiles may be either sharp or gradual. In the literature, it has been stated that (i) carbonation is a diffusion-controlled process, i.e., the chemical reactions are instantaneous, which would result in a sharp reaction front (Papadakis et al., 1991) and (ii) carbonation is a reaction-controlled process, which would give a gradual reaction front (Thiery et al., 2007). In the present study, the different binder types and distinct pore structures of the mortars result in the observation of both sharp and gradual carbonation fronts. However, care should be exercised for the obtained TGA data for samples taken close to the carbonation depths measured by the phenolphthalein method, since they may be a physical mixture of carbonated and non-carbonated parts of the mortars, according to the standard deviations of the apparent carbonation depths in Fig. 2 and the irregular color boundaries illustrated in Fig. 3. Thus, identification of a gradual carbonation front for specific mortars or concretes may be misleading if the errors of the measurements are unknown. The Ca(OH)2 profiles appear sharp for all mortars (Fig. 9), indicating that the carbonation of Ca(OH)2 can be considered as quasi-instantaneous for the present mortars. Thus, the gradual change in CaCO3 content between the surface and the apparent carbonation depth is ascribed to carbonation of other phases such as the C-S-H or AFm phases. In earlier studies, for example where a 10 % (v/v) CO2 was employed (Thiery et al., 2007), a gradual change in the Ca(OH)2 content and even a decrease of the amount of Ca(OH)2 below the apparent carbonation depth were detected by TGA. These observations were explained by an insufficient number of available reaction sites to complete the carbonation reactions at the high CO2 concentration, resulting in CO2 that penetrates to deeper depths, i.e., below the apparent carbonation depth. Fig. 10 shows the Ca(OH)2 and CaCO3 profiles after 91 days of carbonation, including the CaCO3 profiles attributed to carbonation of C-S-H and/or the AFt/AFm phases rather than Ca(OH)2. The profiles show that most of the CaCO3 formed in M and ML mortars originates from carbonation of the C-S-H and AFt/AFm phases. In addition, carbonation of the C-S-H and/or AFt/AFm phases is also observed for the P and L mortars in addition to the principal carbonation of Ca(OH)2. Thus, based on the TGA data, it appears that the reaction of portlandite is fast compared to the diffusion of CO2 while 187


the decomposition of C-S-H and the AFm phases is slow, resulting in the gradual changes of the CaCO3 content close to the surface of the samples.

3.3.3 Thermodynamic modeling The phase assemblages of the hydrated samples have been predicted by thermodynamic modeling, employing the measured degrees of hydration for the principal phases (alite, belite, and metakaolin) in the paste samples, as determined by 29Si MAS NMR analyses of the samples after 91 days of hydration (Dai et al., 2015). The phase assemblages caused by carbonation of the hydrated samples can be predicted by thermodynamic modeling by stepwise adding a certain amount of CO2 (1% (v/v)) as described in section 2.4.5. The predicted data are presented in volume percentages of the original volume as a function of the volume of added air (exponential scale) in Fig. 11, and this representation shows both the changes in phase assemblages and in total porosity. For the P paste, the major hydrate phases prior to carbonation are C-S-H, ettringite, portlandite, and monocarbonate (left side of Fig. 11). Similar phases are predicted for the L paste, however, in smaller amounts as the cement is diluted by the addition of 31.9 wt.% of LS which only reacts to a small extent. For the M and ML samples, portlandite is predicted to be absent as a result of its pozzolanic reaction with metakaolin. The phase assemblages calculated by thermodynamic modeling are found to be in good agreement with experimental data (Dai et al., 2015), although, some small amounts of portlandite are still detected by XRD and by TGA after 91 days of hydration for the M and ML pastes (Lawrence et al., 2006). For the carbonated samples, thermodynamic modeling (Fig. 11) predicts that portlandite will carbonate before the C-S-H phase for the P and L mortars, implying that carbonation of the C-S-H phase will first occur when portlandite is depleted. However, this is not always observed experimentally by bulk analyses, since some portlandite particles carbonated only at the surface may be present (Morandeau et al., 2014). The modeling also shows that monocarbonate is carbonated along with the high-Ca C-S-H phase, the reaction products being strätlingite and calcite. Strätlingite is also predicted for the M and ML blends, in accordance with the experimental studies (Dai et al., 2015), and near the surface this phase also decomposes as a result of carbonation. It is also interesting that the thermodynamic modeling predicts a decalcification of the C-S188

Journal article I

H phase from the high-Ca C-S-H to the low-Ca C-S-H, i.e., that is C-S-H depleted from Ca2+ ions in the interlayer space (Kulik, 2011). This is in accord with a very recent NMR study of carbonated C-S-H samples (Sevelsted and Skibsted, 2015) which showed that the carbonation of C-S-H includes two steps. Firstly, calcium is gradually removed from the interlayer and the defect sites in the silicate chains until Ca/Si = 0.67 is reached and secondly, the Ca from the principal layer is consumed, resulting in the final decomposition of the C-S-H and the formation of an amorphous silica phase. A progressive polymerization of the C-S-H phase as a result of carbonation was also observed by Castellote et al. (Castellote et al., 2009) for increasing CO2 concentrations (0.03%, 3%, 10% and 100%). In addition, a fraction of C-S-H phase with a lower Ca/Si ratio than observed for the non-carbonated sample (Ca/Si=1.87) was detected by 29Si MAS NMR when the samples were exposed to a 0.03% and 3% CO2 environment. These results are well supported by the predicted changes of the Ca/Si ratio for the C-SH phases shown in Fig. 12, where Ca/Si decreases gradually from the initial values of 1.63 (for the P and L mortars) and 1.29 (for M and ML mortars) to 0.67, corresponding to decalcification of the high-Ca C-S-H. The Ca/Si ratio of the C-S-H phase in the M and ML mortars before carbonation are lower than the ratio for the C-S-H in the P and L samples, implying that the C-S-H phases in the P and L mortars have a higher CO2 binding capacity as compared to the low-Ca/Si C-S-H phases in the M and ML mortars, again in accordance with the recent study of carbonation of synthesized C-S-H phases with different Ca/Si ratios (Sevelsted and Skibsted, 2015). 4. Discussion 4.1 Pore structure Thermodynamic modeling can be used to predict changes in total porosity as a result of variations in the molar volume of phases after carbonation as shown in Fig. 11. It should be noted that the volume of the C-S-H phase does not include the gel porosity associated with this phase but only the interlayer water (Kulik, 2011, Muller et al., 2013). For small volumes of CO2 containing air (10–2.00 dm3/g of cement) added, the lowest porosity is observed for the P mortar whereas it is very similar and nearly a factor two higher for the mortars of the blended systems. However, this is not consistent with the total intruded porosities measured by MIP for the non-carbonated samples (Fig. 4) which may be explained by the very poor pore connectivity (small threshold pore 189


sizes) observed for the M and ML mortars in the same figure. In addition, the poor pore connectivity can be explained by an expected higher gel porosity and thus a relatively lower capillary porosity in the M and ML mortars as compared to the P and L mortars. Thermodynamic modeling predicts a major increase of the total porosities for all mortars, when all phases are completely carbonated, corresponding to large volumes of CO2 containing air (e.g., 10–1.0 – 10–0.50 dm3/g of cement, Fig. 11). At lower carbonation degrees a small decrease in total porosity is predicted for the P and L mortars as a result of the differences in molar volumes for portlandite (32.29 cm3/molar) and CaCO3 (e.g. 36.92 cm3/molar (Smyth and McCormick, 1995)). Only an increase in porosity is predicted for the M and ML mortars, since virtually no portlandite is present in these mortars. The DTG curves (Fig. 8) show that large fractions of the phases (30 – 300 oC) are not carbonated after 91 days of exposure. Thus, the thermodynamic stability of the phases towards carbonation (Fig. 11) and the observed minor changes in total porosity measured by MIP (Fig. 6) suggest that the major non-carbonated phase observed in Fig. 8 most likely is a low-Ca C-S-H phase. This is in line with the earlier study of cement pastes exposed to difference CO2 concentrations where 29Si MAS NMR revealed the presence of a fraction of low-Ca CS-H in the carbonated samples exposed to 0.03% and 3% CO2 (Castellote et al., 2009). The combination of the results from MIP, SEM and thermodynamic modeling indicates that the carbonation increases the percolation (threshold pores sizes). The total porosity is initially reduced but it increases at higher degrees of carbonation for P and L mortars. Similar results can be derived for the M and ML mortars, except that these mortars reveal no decrease in total porosity (Fig. 11). The reduced total porosity for the P and L mortars is explained by carbonation of portlandite. The percolation and later increased porosity for all mortars are probably related to carbonation of the C-S-H resulting in a partial decalcification of this phase. It should be noted that the carbonation front at a given exposure time shows some variations in depths (Fig. 2 and Fig. 3). Consequently, if the samples are taken close to carbonation front, it will contain carbonated and noncarbonated materials, which can result in misleading interpretations as also mentioned in section 3.3.2.

190

Journal article I

4.2 CO2 binding The L mortar has the largest intruded porosity and a reduced amount of portlandite as compared to the P mortar, which reasonably well accounts for the lower carbonation resistance observed for the L mortar. The two metakaolin-containing mortars (M and ML), which include the same amount of Portland cement as L, show considerable better carbonation resistance as a result of the refinement and decrease of the porosity. Comparison of the carbonation resistance for the P, ML and M mortars reveals that the poorer performance for the M and ML mortars cannot be explained by the pore structure measurements, since the non-carbonated P mortar has a total MIP porosity comparable to those for the M and ML mortars. The ability of the hydrated cement to bind CO2 must also be taken into account and in this section the CO2 binding results obtained from TGA and thermodynamic modeling are analyzed. To predict the CO2 binding capacity based on the phases formed after 91 days of hydration, the amounts of the main calcium-bearing phases relevant to cement carbonation have been calculated by the GEMS software, using molar percentages rather than volumes, and related to the masses of ignited mortars (by division with a factor of 4 according to the binder/sand ratio of 3). The total amount of CaO present in the hydrates (i.e., in portlandite, C-S-H and AFm but not in limestone) can be used to predict the maximum CO2 binding capacity for each system. The amount of CaO from ettringite for all mortars and from initially formed strätlingite for the M and ML blends are not taken into account, since these phases are predicted to be thermodynamically more stable than the high-Ca C-S-H phase. The measured CO2 binding was obtained by quantification of CaCO3 and portlandite by TGA (section 3.3.2) and the results are summarized in Table 5. It should be noted that the samples used for the CaCO3 measurements were taken from the outer layer of the mortars, which have been in direct contact with CO2 in the air, whereas the samples used for the portlandite measurements were taken from the non-carbonated center of the mortars. The thermodynamic calculations indicate that the P mortar has a significantly higher available CaO content (Table 5), potentially reactive to CO2, allowing for a much higher binding as compared to the three other blends. However, the experimental data in Table 5 and Fig. 10 shows that the amount of CaCO3 bound in the different composite cement mortars is similar after 91 days of carbonation, indicating that carbonation proceeds at a 191


similar reaction rate in the different mortars, although they have different total potential CO2 binding capacities and contain different phases (CH or C-S-H) to bind CO2. In this context, it should also be noted that the pore structure data in Fig. 6 shows that the P, M and ML mortars exhibit similar microstructural changes during carbonation. The similar changes in the pore structure and the small variations in measured CO2 binding data do not explain the observed differences in the apparent carbonation depths measured by the phenolphthalein method as shown in Fig. 2, unless the pH profiles and capillary condensation are taken into account (for an explanation, see section 4.3).

4.3 pH profiles Since the carbonation depth was measured with the phenolphthalein indicator, the differences between the P, M, and ML mortars with respect to their apparent carbonation performances reflect changes in pH which are related to phase changes. Fig. 13 shows the pH profiles predicted by thermodynamic modeling for the different blends. The variations in pH related to the changes of phases are plotted on the same scale of added volume of air as used for the predicted phase assemblages in Fig. 11. The pH decreases with increasing carbonation from an initial high value of 13.4 (P and L) and 12.8 (M and ML). For the high pH values (pH > 12.8), no decrease in pH is predicted as a result of the carbonation of portlandite (P and L mortars) or the carbonation of monocarbonate and the formation of strätlingite (M and ML mortars) since the pH under these conditions is dominated by the alkali ions present in the pore solution. The first major decrease in pH (to a value of pH ≈ 9.7) is predicted to result from carbonation of the high-Ca C-S-H to a low-Ca C-S-H phase and calcium carbonate. The second major decrease of pH (to pH = 7.4) is predicted to occur when the low-Ca C-S-H phase is carbonated, which corresponds to a decalcification of the principal layer of the C-S-H phase, thereby decomposing into a silica phase and calcite. Based on this interpretation, the carbonation depths indicated by the phenolphthalein indicator reflects depletion of the high-Ca C-S-H phase. As very little portlandite is available for carbonation due to the pozzolanic reaction with MK, the carbonation of the C-S-H phase from a high-Ca C-S-H to low-Ca C-S-H and thus the observed pH decrease, take place for lower amounts of CO2 for the M and ML mortars as compared to the P mortar. This explains the shallow apparent carbonation depths observed for the 192

Journal article I

P mortar (Fig. 2). Combination of the results for the similar measured CO2 binding and the predicted pH profiles support the recent study by Leemann et al. (Leemann et al., 2015), which proposed that the concept of CO2 capture capacity per volume of cement paste is the decisive material parameter for carbonation resistance. This higher CO2 binding capacity can be used as an explanation for the observed highest carbonation resistance for the P mortar, since it can capture a higher amount of CO2 per volume of paste as compared to the other mortars. For the L and M (and ML) mortars, thermodynamic modeling predicts that the partial carbonation of the C-S-H phase will occur for higher amounts of CO2 for the L mortar than for M and ML mortars. The deeper carbonation depth observed for the L mortar compared to the M and ML mortars is likely caused by differences in the pore volume accessible to CO2. According to the Kelvin equation, capillary condensation at 57% RH takes place at a pore radius of ≈ 2 nm. Below 2 nm the pores will remain to be water filled which slows down CO2 diffusion, while pore radii > 2 nm are expected to contain air. However, pores with a 2 nm radius are close to the detection limit of MIP of 3 nm. Turcry et al. (Turcry et al., 2014) have reported the evolution of RH at different concrete depths as a function of drying time at a temperature of 45 ± 3 oC. When the RH of the concrete surface was dried to about 20%, a RH of 60% at a 10 mm concrete depth was still detected after 14 days, and at even deeper depth, the RH was 100%. Additional investigations in our laboratory have shown that the weight loss of the MK mortar during drying at 50 oC do not stop before 65 days of drying as a result of the very fine pores as indicated by the MIP data (Fig. 4). Considering the slower drying process when the samples are exposed to 57% RH at 20 oC compared to the samples exposed to 50 oC in the oven, suggests that a higher RH inside the mortars can be present for a long time during carbonation, e.g., a RH of 80% which corresponds to a Kelvin pore radius of 4.8 nm. The M and ML mortars have considerably lower amounts of pores above the 4.8 nm radius than the L mortar (Fig. 4), and thus the carbonation is slowed down for the M and ML mortars in comparison to the L mortar. Thus, the effect of capillary condensation has to be taken into account, when the performance of cements under accelerated conditions with low RH is transferred to concrete structures, where the higher RH may have a strong cement-specific impact on the carbonation resistance (Leemann et al., 2015).

193


It should be noticed that phenolphthalein in solution shows a gradual color change from fuchsia to colorless upon a change in pH from 10 to 8.2. Such a gradual change in color has not been documented for cement-based materials and neither has the cement literature agreed on a single, specific value for the pH that causes the color change. Evidence for calcite coating of portlandite crystals, preventing them from buffering the pore solution, has been clearly demonstrated (Galan et al., 2015). Similarly, indications of remains of non-carbonated C-S-H are observed in apparently carbonated samples (Fig. 8). Here, the DTG peak in the range 30 – 200 oC is assumed to originate mainly from the low-Ca C-S-H phase, as discussed in section 3.3.3.

5. Conclusions The carbonation performance of well-hydrated mortars of pure Portland cement (P), binary blends with limestone (L) and metakaolin (M) and a ternary blend containing metakaolin and limestone (ML) has been investigated by the exposure to a 1% (v/v) CO2 concentration at RH = 57 % for up to 280 days. TGA, MIP, SEM and thermodynamic modelling have been used to study the changes in pore structure and CO2 binding capacity. The highest and lowest carbonation performance is observed for the P and L mortars, respectively and the M and ML mortars exhibit poor resistance to carbonation as compared to the P mortar. Thermodynamic modeling predicts carbonation of Ca(OH) 2 followed by carbonation of the C-S-H phase for the P and L mortars. In the blends with metakaolin, where virtually all Ca(OH)2 has been consumed by the pozzolanic reaction, carbonation of monocarbonate and the C-S-H phase is predicted. In addition, a two-step decalcification process of the C-S-H phase is predicted where calcium initially is released from the interlayer of the high-Ca C-S-H, resulting in the formation of a low-Ca C-S-H phase. In the second step, Ca is released from the principal layers of low-Ca C-S-H, leading to the final decomposition of the C-S-H phase, in agreement with recent results from 29Si MAS NMR studies (Castellote et al., 2009, Sevelsted and Skibsted, 2015). The TGA data recorded at different depths in the mortars confirm the sequence of carbonation predicted

by

thermodynamic

modeling,

i.e.,

portlandite

carbonates

before

monocarbonate and ettringite before the complete decomposition of the C-S-H phase to silica. 194

Journal article I

Carbonation of portlandite is predicted to have little impact on pH as long as some highCa C-S-H phase is still present. Based on these predictions, the carbonation of the highCa C-S-H phase reduces pH to 9.6, which is in the range from 10 to 8.2 where the phenolphthalein changes from fuchsia to colorless. The subsequent decalcification of the low-Ca C-S-H to a silica gel and calcium carbonate reduces pH further. Moreover, from the TGA measurements it is found that the amount of CaCO3 bound in the different composite cement mortars is similar after 91 days of carbonation, indicating that carbonation proceeds at a similar reaction rate in the different mortars, although they have different total potential CO2 binding capacities and contain different phases to bind CO2. Thermodynamic modeling reveals a major increase of the total porosity when all phases are carbonated. Little changes in porosity is predicted for the P and L mortars before the low-Ca C-S-H phase is carbonated. This is confirmed by the experimental observations by SEM and repeated MIP (2nd intrusion) which show a slight reduction of the percolated porosity in the P and L mortars. The volume and size of the percolated pores seem to have increased in the mortars containing metakaolin. More importantly, MIP showed a coarsening of the pore threshold for all mortar types, but only small variations of the total intruded porosity. The P, M and ML mortars were found to bind similar quantities of CO2 from TGA and to exhibit similar coarsening of the pores, as measured by MIP. The difference in apparent carbonation depths for these mortars is explained by the modelled impact of phase changes on pH. The partial carbonation of the C-S-H phase starts earlier for the M and ML mortars, resulting in an earlier pH decrease and thus a deeper apparent carbonation depth. For the L mortar, the amount of portlandite is higher than in the M and ML mortars. However, this does not improve the carbonation resistance, which is explained by a coarser pore structure of the L mortar leading to a lower amount of pores filled with water due to capillary condensation and as such a better accessibility to CO2. In summary, the data obtained in the present paper confirm that the CO2 binding capacity, the porosity, and the capillary condensation are the decisive parameters governing the carbonation depth. A higher CO2 binding capacity, and thus a slower decrease of pH, is the main reason for the higher carbonation resistance observed for the P mortar. The difference in carbonation resistance for the limestone and the metakaolin 195


containing mortars, which all have similar CO2 binding capacities, is related to the difference in the porosity. This is attributed to the presence of metakaolin which leads to less porosity and a refined pore structure that favors the capillary condensation at a given RH as compared to the limestone mortar, and consequently results in a lower carbonation rate.

Acknowledgement The Danish Council for Strategic Research is acknowledged for financial support to the LowE-CEM project. We thank Nikolajs Toropovs for assistance with the TGA measurements at the Laboratory for Concrete and Construction Chemistry, Empa, Dübendorf.

196

Journal article I

References Antoni M, Rossen J, Martirena F, Scrivener K (2012) Cement substitution by a combination of metakaolin and limestone. Cement and Concrete Research 42:1579-1589. Borges PH, Costa JO, Milestone NB, Lynsdale CJ, Streatfield RE (2010) Carbonation of CH and C–S–H in composite cement pastes containing high amounts of BFS. Cement and Concrete Research 40:284-292. Canut MMC (2012) Pore structure in blended cement pastes (PhD thesis). Technical University of Denmark, Department of Civil Engineering. Castellote M, Fernandez L, Andrade C, Alonso C (2009) Chemical changes and phase analysis of OPC pastes carbonated at different CO2 concentrations. Materials and Structures 42:515-525. Chang C-F, Chen J-W (2006) The experimental investigation of concrete carbonation depth. Cement and Concrete Research 36:1760-1767. Dai Z, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015) Investigation of blended systems of supplementary cementitious materials with white Portland cement and limestone (manuscript in preparation). De Ceukelaire L, Van Nieuwenburg D (1993) Accelerated carbonation of a blastfurnace cement concrete. Cement and Concrete Research 23:442-452. De Weerdt K, Kjellsen K, Sellevold E, Justnes H (2011) Synergy between fly ash and limestone powder in ternary cements. Cement and Concrete Composites 33:3038. Galan I, Glasser F, Baza D, Andrade C (2015) Assessment of the protective effect of carbonation on portlandite crystals. Cement and Concrete Research 74:68-77. Goto S, Suenaga K, Kado T, Fukuhara M (1995) Calcium silicate carbonation products. Journal of the American Ceramic Society 78:2867-2872. Groves G, Rodway D, Richardson I (1990) The carbonation of hardened cement pastes. Advances in Cement research 3:117-125. Groves GW, Brough A, Richardson IG, Dobson CM (1991) Progressive changes in the structure of hardened C3S cement pastes due to carbonation. Journal of the American Ceramic Society 74:2891-2896. Herfort D (2004) Challenges of cement production. In: Anna Maria 2004 Workshop on Cement and Concrete Anna Maria Island, Florida, USA. Houst YF, Wittmann FH (1994) Influence of porosity and water content on the diffusivity of CO2 and O2 through hydrated cement paste. Cement and Concrete Research 24:1165-1176. Johannesson B, Utgenannt P (2001) Microstructural changes caused by carbonation of cement mortar. Cement and Concrete Research 31:925-931. Kaufmann J (2010) Pore space analysis of cement-based materials by combined Nitrogen sorption–Wood’s metal impregnation and multi-cycle mercury intrusion. Cement and Concrete Composites 32:514-522. 197


Kaufmann J, Loser R, Leemann A (2009) Analysis of cement-bonded materials by multi-cycle mercury intrusion and nitrogen sorption. Journal of colloid and interface science 336:730-737. Kulik D, Tits J, Wieland E (2007) Aqueous-solid solution model of strontium uptake in CSH phases. Geochim Cosmochim Acta 71:A530. Kulik DA (2011) Improving the structural consistency of CSH solid solution thermodynamic models. Cement and Concrete Research 41:477-495. Kulik DA, Wagner T, Dmytrieva SV, Kosakowski G, Hingerl FF, Chudnenko KV, Berner UR (2013) GEM-Selektor geochemical modeling package: revised algorithm and GEMS3K numerical kernel for coupled simulation codes. Computational Geosciences 17:1-24. Lange L, Hills C, Poole A (1996) The effect of accelerated carbonation on the properties of cement-solidified waste forms. Waste Management 16:757-763. Lawrence RMH, Mays TJ, Walker P, D’Ayala D (2006) Determination of carbonation profiles in non-hydraulic lime mortars using thermogravimetric analysis. Thermochimica Acta 444:179-189. Leemann A, Nygaard P, Kaufmann J, Loser R (2015) Relation between carbonation resistance, mix design and exposure of mortar and concrete. Cement and Concrete Composites 62:33-43. Lothenbach B, Durdzinski P, De Weerdt K Chapter 5: Thermogravimetric analysis. In A Practical Guide to Microstructural Analysis of Cemetitious Materials, Edt. Karen Scrivener, Ruben Snellings, and Barbara Lothenbach, to be published byTaylor & Francis Group. Lothenbach B, Le Saout G, Gallucci E, Scrivener K (2008) Influence of limestone on the hydration of Portland cements. Cement and Concrete Research 38:848-860. Lothenbach B, Le Saout G, Haha MB, Figi R, Wieland E (2012) Hydration of a lowalkali CEM III/B–SiO2 cement (LAC). Cement and Concrete Research 42:410423. Lothenbach B, Matschei T, Möschner G, Glasser FP (2008b) Thermodynamic modelling of the effect of temperature on the hydration and porosity of Portland cement. Cement and Concrete Research 38:1-18. Matschei T, Lothenbach B, Glasser FP (2007) The role of calcium carbonate in cement hydration. Cement and Concrete Research 37:551-558. McPolin D, Basheer P, Long A (2009) Carbonation and pH in mortars manufactured with supplementary cementitious materials. Journal of Materials in Civil Engineering 21:217-225. Morandeau A, Thiery M, Dangla P (2014) Investigation of the carbonation mechanism of CH and CSH in terms of kinetics, microstructure changes and moisture properties. Cement and Concrete Research 56:153-170. Morandeau A, Thiéry M, Dangla P (2015) Impact of accelerated carbonation on OPC cement paste blended with fly ash. Cement and Concrete Research 67:226-236.

198

Journal article I

Muller A, Scrivener K, Gajewicz A, McDonald P (2013) Use of bench-top NMR to measure the density, composition and desorption isotherm of C–S–H in cement paste. Microporous and Mesoporous Materials 178:99-103. Ngala V, Page C (1997) Effects of carbonation on pore structure and diffusional properties of hydrated cement pastes. Cement and Concrete Research 27:9951007. Papadakis VG (2000) Effect of supplementary cementing materials on concrete resistance against carbonation and chloride ingress. Cement and Concrete Research 30:291-299. Papadakis VG, Vayenas CG, Fardis MN (1991) Fundamental modeling and experimental investigation of concrete carbonation. ACI Materials Journal 88:363-373. Pihlajavaara S (1968) Some results of the effect of carbonation on the porosity and pore size distribution of cement paste. Matériaux et Construction 1:521-527. Rostami V, Shao Y, Boyd AJ, He Z (2012) Microstructure of cement paste subject to early carbonation curing. Cement and Concrete Research 42:186-193. Šauman Z (1971) Carbonization of porous concrete and its main binding components. Cement and Concrete Research 1:645-662. Scrivener K, Favier A (2015) (Eds.). Calcined Clays for Sustainable Concrete: Proceedings of the 1st International Conference on Calcined Clays for Sustainable Concrete: Springer. Sevelsted TF, Skibsted J (2015) Carbonation of C–S–H and C–A–S–H samples studied by 13C, 27Al and 29Si MAS NMR spectroscopy. Cement and Concrete Research 71:56-65. Shi Z, Geiker MR, De Weerdt K, Lothenbach B, Kaufmann J, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015) Durability of Portland cement blends including calcined clay and limestone: interactions with sulfate, chloride and carbonate ions. In: Calcined Clays for Sustainable Concrete, pp 133-141: Springer. Smyth JR, McCormick TC (1995) Crystallographic data for minerals. Mineral Physics and Crystallography: A Handbook of Physical Constants 2:1-17. Steenberg M, Herfort D, Poulsen SL, Skibsted J, JS D (2011) Composite cement based on Portland cement clinker, limestone and calcined clay. In: 13th International Congress of the Chemistry of Cement, p 97 Madrid. Thiery M, Villain G, Dangla P, Platret G (2007) Investigation of the carbonation front shape on cementitious materials: Effects of the chemical kinetics. Cement and Concrete Research 37:1047-1058. Thiery M, Villain G, Platret G (2003) Effect of carbonation on density, microstructure and liquid water saturation of concrete. In: Proc 9th Eng Conf on Advances in Cement and Concrete, USA (Copper Mountain), pp 481-490. Tsivilis S, Chaniotakis E, Badogiannis E, Pahoulas G, Ilias A (1999) A study on the parameters affecting the properties of Portland limestone cements. Cement and Concrete Composites 21:107-116.

199


Turcry P, Oksri-Nelfia L, Younsi A, Aït-Mokhtar A (2014) Analysis of an accelerated carbonation test with severe preconditioning. Cement and Concrete Research 57:70-78. Voglis N, Kakali G, Chaniotakis E, Tsivilis S (2005) Portland-limestone cements. Their properties and hydration compared to those of other composite cements. Cement and Concrete Composites 27:191-196. Wagner T, Kulik DA, Hingerl FF, Dmytrieva SV (2012) GEM-Selektor geochemical modeling package: TSolMod library and data interface for multicomponent phase models. The Canadian Mineralogist 50:1173-1195.

200

Journal article I

Table 1. Chemical composition (wt.%), density and Blaine fineness for the starting materials wPc

LS

MK

SiO2

21.81

3.92

52.84

Al2O3

3.56

0.33

39.49

Fe2O3

0.24

0.14

1.42

CaO

66.13

53.73

0.22

MgO

1.10

0.35

0.48

K2O

0.43

0.05

1.00

Na2O

0.04

0.08

0.05

SO3

3.37

0.05

0.06

TiO2

0.21

0.02

0.88

P2O5

0.04

0.10

0.11

LOI

2.57

41.8

3.55

Density (kg/m3)

3080

2700

2530

Blaine fineness (m2/kg)

387

1211

1891

Carbon content

0.37

-

-

CaCO3

3.1

93.8

-

201


Table 2. Binder compositions for the produced mortars (wt.%). Mortar id.

wPc

MK

LS

P

100

0

0

L

68.1

0

31.9

ML

68.1

25.5

6.4

M

68.1

31.9

0

Table 3. Degrees of hydration for alite, belite and MK in hydrated paste samples and compressive strengths for the corresponding mortars after 28 – 182 days of hydration in demineralized water (Dai et al., 2015). degree of hydration (H, %)a

Compressive strengthb

Mortar and alite

belite

(MPa)

MK

paste id.

a

28d

180d

28d

180d

P

81

95

25

L

90

97

ML

83

M

86

Determined from

29

28d

180d

28d

91d

182d

63

68.7

78.7

80,6

57

82

42.6

50.2

49.4

86

34

36

48

54

67.3

73.1

73.5

86

34

34

38

50

62.6

70.9

73.9

Si MAS NMR (Dai et al., 2015) as H = (1 – I(t)/I(t=0)), where

I(t=0) and I(t) are the intensities of the individual phases before and after hydration for the time (t), respectively. b

The compressive strengths are normalized to 2% (v/v) air content.

202

Journal article I

Table 4. Carbonation coefficients (K) for the studied mortars. Mortar id.

K (mm/d1/2)

R2

P

0.17

0.923

L

0.85

0.992

ML

0.62

0.995

M

0.61

0.983

a

The coefficients are obtained from linear fits of the carbonation depths as a function of

the square root of time (x = K t1/2), using the data in Fig. 2. R2 is the correlation coefficients from the least-squares linear fits.

Table 5. Comparison of the measured CaCO3 (carbonated samples close to the mortar surface) and portlandite (non-carbonated inner core) after 91 days of carbonation and the calculated total potential CO2 binding capacity (mol CaO/100g ignited mortar) for the hydrated mortars. Method

Phases

P

Calculated

Ca(OH)2

by GEMS

L

ML

M

0.11 0.07

0.00

0.00

C-S-H

0.12 0.10

0.11

0.11

Monocarbonate

0.01 0.006 0.06

0.03

Total CaO (potential binding capacity) 0.24 0.18

0.17

0.14

Measured

Surface: CaCO3 (carbonation)

0.13 0.12

0.13

0.13

by TGA

Surface: Ca(OH)2

0.02 0

0

0

Core: Ca(OH)2

0.10 0.07

0.008 0.004

203


Figure 1. Illustration of the color change (0 - 66 s) after spraying the M mortar with phenolphthalein.

Figure 2. Carbonation depths for the four types of mortars indicated by phenolphthalein (1% (v/v) CO2 at 57% RH and 20 oC). The carbonation coefficients (K) determined from these plots are listed in Table 4.

Figure 3. Illustration of the carbonation front indicated by phenolphthalein for the four different mortars after 91 days of carbonation. 204

Journal article I

Figure 4. Intrusion curves from the first MIP intrusion cycle for the four different mortars The measurements were performed on non-carbonated samples taken from the center of the mortars bars after 91 days of saturated curing plus 28 days of carbonation.

205


Figure 5. Impact of carbonation on the changes in intrusion curves from the first and second MIP intrusion cycles for the four mortars (P, L, ML and M).

206

Journal article I

Figure 6. Impact of carbonation on the microstructure based on the first (left) and second (right) MIP intrusion cycles for the four studied mortars.

207


Figure 7. SEM images for non-carbonated (left) and carbonated (right) polished sections of the mortars P (a, b), L (c, d) ML (e, f). The images for the M mortar are very similar to those of the ML mortar, and thus not shown here. The measurements are performed after 91 days of carbonation for the L mortars and after 280 days for the other mortars.

208

Journal article I

Figure 8. DTG curves for samples collected from different depths of the four types of mortars after 91 days of carbonation. The carbonation depths measured by the phenolphthalein spray method are given in each of the individual graphs and refer to the cut slice, which is a physical mixture of carbonated and non-carbonated sample. (Ettri. = Ettringite; Mc = Monocarbonate; CH = Portlandite; Cc = CaCO3; Strä. = Strätlingite.)

209


Figure 9. Carbonation profiles obtained by quantification of CaCO3 and Ca(OH)2 from the TGA data for the four different mortars. The results are presented as molar content per 100 g of ignited mortar (800 oC). The contents of CaCO3 originally present in the mortars have been subtracted from the total measured CaCO3 quantities.

210

Journal article I

Figure 10. Carbonation profiles after 91 days of CO2 exposure, obtained by quantification of CaCO3 and Ca(OH)2 from the TGA data for four different mortars. The results are presented as molar content per 100 g of ignited mortars (800 oC). The amount of CaCO3 formed by carbonation of other phases than portlandite is calculated by subtracting the molar amount of reacted portlandite from the total quantity of CaCO3. The contents of CaCO3 originally present in the mortars have been subtracted from the total measured CaCO3 quantities.

211


Figure 11. Changes in phase assemblages and total porosities predicted by thermodynamic modeling for samples of the four blends in contact with air containing 1% (v/v) CO2. The horizontal dash lines show the original porosities of the mortars prior to carbonation, whereas the dash arrows indicate the changes in porosity after carbonation. The vertical dash lines reflect the pH corresponding to decalcification of the high-Ca C-S-H phase, which is also relevant for the de-passivation of reinforcement.

212

Journal article I

Figure 12. Changes of the Ca/Si ratios for the C-S-H phase as a result of carbonation for different blends predicted by thermodynamic modeling.

Figure 13. pH profiles related to the changes in phase assemblages predicted by thermodynamic modeling. The right bar indicates the gradual color change of phenolphthalein from fuchsia to colorless upon pH changes from 10 to 8.2.

213


214

Journal article II

Semi-quantification of Friedel's salt by thermogravimetric analysis in chloride exposed Portland cement-based mortars Zhenguo Shia, Mette Rica Geikerb, Barbara Lothenbachc, Klaartje De Weerdtb, Sergio Ferreiro Garzónd, Kasper Enemark-Rasmussene, Jørgen Skibsteda

a. Department of Chemistry and Interdisciplinary Nanoscience Center (iNANO), Aarhus University, 8000 C Aarhus, Denmark b. Department of Structural Engineering, Norwegian University of Science and Technology (NTNU), 7491 Trondheim, Norway c. Laboratory for Concrete & Construction Chemistry, Swiss Federal Laboratories for Materials Science and Technology (Empa), 8600 Dübendorf, Switzerland d. Aalborg Portland A/S, Cementir Holding S.p.A., 9100 Aalborg, Denmark e. Department of Chemistry, Technical University of Denmark, 2800 Kgs. Lyngby, Denmark

215


Abstract A novel method is presented to quantify the amount of Friedel’s salt in chloride exposed Portland cement-based mortars using thermogravimetric analysis (TGA). The method is adopted to obtain Friedel’s salt profiles in mortar cylinders exposed to 2.8 M NaCl solution through bulk diffusion. The Friedel’s salt profiles are compared for the first time with the total chloride profiles. The comparison of the results revealed that the chemically bound chloride in Friedel’s salt appears to be limited, i.e. less than 10 % of the maximum total chloride content. It is found that only 22 wt.% of the total amount of C3A contributes to formation of Friedel’s salt. In addition, the Friedel’s salt seems to have a limited capacity to bind chloride as the amount plateaued after reaching a certain total chloride concentration.

KEYWORDS Friedel's salt; thermogravimetric analysis; Portland cement; chloride binding; total chloride profiles; semi-quantification

1

Introduction

Chloride induced reinforcement corrosion is one of the major deterioration mechanisms for reinforced concrete structures. When concrete is exposed to seawater or de-icing salts, chloride may penetrate into the concrete cover over time. If a critical chloride level at the steel bar is reached, reinforcement corrosion may occur. In chloride-exposed concrete, chloride ions are either chemically bound or physically adsorbed by hydrates, or present in pore solution. The resistance to chloride ingress of a given concrete is usually assessed by analyzing the total chloride profile (Tang et al., 2011, Bertolini et al., 2013, De Weerdt et al., 2014). However, some studies (Yuan et al., 2009, Baroghel-Bouny et al., 2012, Bentz et al., 2013) stated that the interactions of chloride with cement hydrates may retard the chloride transport and change the shape of chloride profiles. On contrary, R. Loser et al. (Loser et al., 2010) found only a limited fraction of chloride bound by cement hydrates, whereas larger fraction of chloride was

216

Journal article II

in pore solution. They concluded that the effect of chloride binding on chloride ingress was less important compared to the effect of permeability. In addition to the discrepancy on the role of total chloride binding on chloride ingress, the main binding was deemed to arise from the C3A or C4AF through formation of Friedel’s salt (3CaO·A12O3·CaCl2·10H2O) or its analogue as reviewed by (Justnes, 1998, Yuan et al., 2009, Galan and Glasser, 2015). However, Neville (Neville, 1995a, Neville, 1995b) argued that the fact that more chloride ions are bound when the C3A content of the cement is higher may be true, when chloride ions are present at the time of mixing as a result of rapid reaction of chloride with C3A. The author (Neville, 1995a, Neville, 1995b) also stated that only a smaller amount of chloroaluminates is formed, and that they may become dissociated under some future circumstances. The aim of this paper is twofold. First a novel method is developed to quantify the amount of Friedel’s salt in Portland cement-based mortars using thermogravimetric analysis (TGA). This method is then applied to determine the contribution of Friedel’s salt to chloride binding in Portland cement-based mortars exposed to 2.8M NaCl solution for chloride ingress test. TGA has been used to identify Friedel’s salt in cement pastes or mortars by several studies (Jain and Neithalath, 2010, De Weerdt et al., 2015, Geng et al., 2015). However, only one reference (Jain and Neithalath, 2010) was found to use TGA for the quantification of Friedel’s salt in Portland cement-based pastes. Unfortunately, the methodology for quantification was not described. The first derivative of the TG curve (DTG) for synthesized Friedel’s salt as obtained by several other studies (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013, Lothenbach et al., 2016) is shown in Fig. 1. The thermal decomposition of Friedel’s salt results in two major DTG peaks in temperature ranges of 30 ~ 180 oC and 180 ~ 450 oC. It should be noted that two welldefined single peaks were observed by (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013) rather than splitting of the second peak (180 ~ 450 oC) as observed by (Lothenbach et al., 2016). The reason for this discrepancy is not clear. Nonetheless, the weight losses associated with the lower and higher temperature intervals have the same ratio of 4:6 for different studies. The fixed ratio between the two weight losses is ascribed to the layered structure and composition of Friedel’s salt, with positively charged main layers [Ca2Al(OH)6]+ comprising six hydroxyl groups and negatively 217


charged interlayers [2Cl-, 4H2O] containing four water molecules (Grishchenko et al., 2013). The Friedel’s salts in cement pastes or mortars is usually identified by the second DTG peak (De Weerdt et al., 2015, Geng et al., 2015). In this study, the proposed method for quantification of the Friedel’s salt in chloride exposed Portland cementbased mortars is based on the determination of the main layer water, finally taking into account of all the water from Friedel’s though a simple calculation. The applicability of TGA method for Friedel’s salt quantification is verified by comparison of the results with those obtained by Rietveld refinement of X-ray diffraction spectra for several selected samples. The method is applied to determine the Friedel’s salt profiles in chloride exposed Portland cement-based mortars. These profiles are compared with the total chloride profiles measured for the same samples. The comparison of the two profiles reveals the contribution of Friedel’s salt to the chloride binding. To the titled authors’ knowledge, this has not been presented before.

2

Experimental

2.1

Materials

The binders used in this study were made from a white Portland cement (wPc, CEM I 52.5 N) and three supplementary cementitious materials (SCMs): metakaolin (MK), white silica fume (SF) and limestone (LS). The wPc was produced by Aalborg Portland A/S, Denmark, and included 3.1 wt.% LS, 4.1 wt.% gypsum and 1.9 % wt.% free lime. The MK was produced in the laboratory from kaolinite (Kaolinite SupremeTM from Imerys Performance Minerals, UK) by thermal treatment at 550 oC for 20 h. The SF was purchased from Elkem. The LS was a Maastrichtian chalk from Rørdal, Northern Denmark. The chemical compositions of the starting materials determined by X-ray fluorescence (XRF) are given in Table 1. The wPc contained 64.9 wt.% alite (C 3S = 3CaO·SiO2) and 16.9 wt.% belite (C2S = 2CaO·SiO2) and 7.8 wt.% aluminate (C3A = 3CaO·Al2O3) (Dai et al., 2015), where the content of silicate phases were determined by 29Si MAS NMR according to the method proposed by Skibsted et al. (Skibsted et al., 1995) and the quantity of the aluminate phase was determined by mass balance calculation. The Ferrite (C4AF = 4CaO·Al2O3·Fe2O3) phase was not taken into account 218

Journal article II

as the small amount of iron is expected to be incorporated as guest-ions in the C3S, C2S and C3A phases. The sand used to for the mortars was a CEN reference sand (Normensand GmbH, Germany), which has a silica content of at least 98 wt.% and a density of 2650 kg/m3. A superplasticizer (SP, Glenium 27, BASF) was used to achieve similar flow for all mortars.

2.2

Mortar preparations

The compositions of the binders (Table 2) targeted a replacement of 35 wt.% of white Portland clinker by SCMs. Considering the small amounts of LS and gypsum in the wPc, this resulted in actual binder compositions with 31.9 wt.% replacement of the wPc. The P mortar produced by pure wPc and L mortars produced by combination of wPc and 31.9 wt.% LS were prepared as reference mortars. The ML and M mortars were produced by replacing wPc with MK (1:1 clay) and/or LS, where the actual Si/Al ratio of 1.13 for the MK accounts for the 2 wt.% quartz in the MK (Dai et al., 2014). The MSL and MS mortars were produced by replacing wPc with MK, SF and/or LS, where the combination of MK and SF was to mimic a 2:1 clay (Dai et al., 2014). However, the actual Si/Al ratio of 2.36 was higher than the ideal ratio in order to account for the partial substitution of Al by Mg in the octahedral layers found in montmorillonite (Garg and Skibsted, 2014). Previous study (Dai et al., 2014) found a synergetic effect between LS and MK, and highest compressive strength for mortars with MK/(MK+LS) = 0.75 and (MK+SF)/(MK+SF+LS) = 0.75 in binary and ternary blends respectively. The binders were used to produce mortars with a constant water to binder ratio (w/b = 0.5) and binder to sand ratio (b/s = 1/3), both ratios by weight. The dosage of superplasticizer (SP/(MK+SF) = 0.04 for mortars including SF, and SP/MK = 0.07 for mortars without SF) was adjusted to achieve a flow within ± 5% of the flow of the reference P mortar. Mortars were mixed and cast into polypropylene bottles (ø50 mm, 125 ml). A small amount of water was added on top of the mortar to keep the mortar saturated and to compensate for self-desiccation caused by early age chemical shrinkage (Geiker, 1983). Mortars were demolded after 24 hours and cured in demineralized water for 91 days in an airtight bucket with water/mortar ratio of 3:1 by volume at 20 ± 1 oC. At the last

219


week of curing, mortar cylinders were treated by removing (wet cutting) about 5 mm mortar from both ends in order to remove the paste rich layer, potentially carbonated layer and leaching layer. Immediately after cutting, mortars were placed on table and exposed to natural environment in laboratory for 2 hours for surface drying, followed by a two layer epoxy coating (1 mm thickness) on the bottom and circumference surfaces, leaving the top surface uncoated for exposure to chloride solution after 91 days of curing. When the epoxy hardened, the uncoated surfaces of mortars were submerged in small amount of demineralized water for re-saturation until the end of 91 days of curing.

2.3

Total chloride profile analysis

After 91 days of curing, three epoxy coated mortar cylinders from each blend were exposed to 700 ml 2.8 M NaCl solution in an airtight box for 35 days at 20 oC. The solutions were changed weekly in order to maintain an approximate constant chloride concentration in the exposure solution. At the end of exposure, one of the three mortar cylinders from each box was split along the chloride ingress direction. The chloride penetration depth was measured by spraying 0.1 M AgNO3 solution on the fresh split surface of one half-cylinder. The remaining two full mortar cylinders and the one halfcylinder were profile ground with a layer thickness increasing from 1 to 4 mm. The total depth of grinding was determined based on the AgNO3 solution measured chloride penetration depth. The powder samples from each layer were collected for total chloride profile analysis. The powders were firstly dried in an oven at 105 oC for 24 hours. The moisture content was calculated according to the mass loss after drying. The dried powders were dissolved in 1:10 HNO3 solution at 80 oC. The volume of solution to be used to dissolve the paste was 20 ml for 2 g powder (from and close to outer layer) and 40 ml for 4 g powder (from inner layer) respectively. The powder-acid suspension was stirred and left to rest for 2 hours. The resulting solutions were subsequently filtrated, 15 ml filtrate was sampled, and 1 ml filtrate was titrated by 0.01 M AgNO3 solution for samples from and close to outer layer. For filtrates with low chloride concentration for samples from inner layer, 10 ml of filtrate was titrated with the addition of acetic acid (EMSURE®). The total chloride content was reported as mass percentage of dried mortars at 105 oC.

220

Journal article II

2.3

Quantification of Friedel’s salt and portlandite by TGA

Thermogravimetric analysis (TGA) was performed on powdered mortar samples which were used for total chloride analysis. A Mettler Toledo TGA/SDTA 851 instrument was used. About 50 mg of the powdered samples were loaded in 150 µm alumina crucible. The weight loss of the samples was monitored while heating from 30 oC to 980 oC at a rate of 20 oC/min and purging with 50 ml/min N2. The weight loss and quantified data were reported as mass percentage of dried mortars at 105 oC. According to the ratio (4:6) between interlayer water and main layer water determined in the synthesized samples as mentioned in the introduction, the content of Friedel’s salt in mortars can be quantified by firstly determining the content of the main layer water (230 oC ~ 410 oC) in Friedel's salt. Fig. 2 shows the first derivative of TGA curve (DTG), for powdered mortar samples obtained at 1 ~ 2 mm depth and non-contaminated region (e.g. 8~12 mm), and for synthesized Friedel’s salt. The quantification was done by drawing a baseline on the DTG curve for the chloride contaminated sample, which is relevant to the baseline for the reference sample non-contaminated by chloride. The area marked on the DTG curve of chloride contaminated sample was then integrated, i.e. 0.12 wt%. The content of Friedel’s salt in the mortar can thus be calculated according to the following equation: 𝑚𝐹𝑠 =

𝑀𝐹𝑠 561.322 × 𝑚𝐻 = × 𝑚𝐻 (6 × 𝑀𝐻 )𝜇 (6 × 18.015)𝜇

Where m_Fs is the mass fraction of Friedel’s salt in mortar (wt. %); m_H is TGA measured loss of water from the main layer (wt. %); M_Fs is the molar mass (561.322 g/mol)

of

Friedel’s

salt

with

a

chemical

composition

in

the

form

of

Ca4Al2Cl2(OH)12·4H2O; M_H is the molar mass (18.015 g/mol) of H2O; Since the temperature range (230 ~ 410 oC) for the decomposition of the main layer water of Friedel’s salt in mortar samples does not fully account for the whole temperature range (180 ~ 450 oC) for the decomposition of the main layer water determined in synthesized Friedel’s salt as shown in Fig. 2, a correction coefficient, μ,is suggested to take into account the underestimation of the main layer water of Friedel’s salt in mortar samples, which can be expressed as: 𝜇=

𝑚230~410 × 100 𝑚180~450 221


Where, m_(230~410) is partial main layer water of synthesized Friedel’s salt in temperature range 230 ~ 410 oC. This temperature range is determined according to the temperature range of main layer water of Friedel’s salt observed in mortar sample in Fig. 2. The m_(180~450) accounts for all the main layer water, which can be determined in temperature range of 180 ~ 450 oC in DTG curve for synthesized Friedel’s salt. Therefore, in this study, about 80% of main layer water of Friedel’s is able to be determined in mortar samples. Similarly, the portlandite content was also quantified, by determining the weight loss in temperature range from 400 oC to 500 oC. It was calculated according to the equation: 𝑚𝐶𝐻 =

𝑀𝐶𝐻 74.09 ×𝑚 = ×𝑚 𝑀𝐻 18.015

Where m_CH is the mass fraction of portlandite in mortar (wt. %); m is loss of water from portlandite (wt. %); M_CH is the molar mass (74.09 g/mol) of portlandite in the form of Ca(OH)2; M_H is the molar mass (18.015 g/mol) of H2O.

2.3

X-ray diffraction analysis

X-ray diffraction (XRD) analysis was carried out with a CubiX3 diffractometer (PANalytical, NL) between Bragg angles (2Θ) of 5o and 65o using a step size of 0.02o. The powdered mortar samples used both for total chloride profile analysis and TGA analysis were prepared by mixing and grinding 1.80 g of sample and 0.20 g of highpurity nanocrystalline anatase (TiO2) in a planetary ball-mill for 45 seconds at 350 rpm. The mixed powder was then lightly pressed into a pellet with a smooth surface. The acquired diffractograms was simulated with the Rietveld procedure using starting parameters for the individual phases obtained from the Inorganic Crystal Structure Database (ICSD). Only the scale factor, unit cell dimensions and profile parameters were refined in the process. The addition of a known amount of anatase allowed for a possible quantification of the observed phases (e.g. Friedel’s salt and portlandite), despite the presence of amorphous phases and a large amount of quartz in the powdered mortar samples.

222

Journal article II

3

Results

3.1

Total chloride profiles

The total chloride profiles for all mortars exposed to 2.8M NaCl solution for 35 days are shown in Fig. 3. The total chloride contents decrease with increasing chloride penetration depths. The L mortar shows the deepest chloride penetration depth and the highest total chloride content at any depth excluding the outer layer, whereas the MK mortars show the lowest chloride penetration depths. The total chloride profiles show no differences for the different MK mortars. The highest and lowest chloride resistance observed for the MK and L mortars may be explained by the lowest and highest pore connectivity observed for the MK and L mortars respectively from the mercury intrusion porosimetry (MIP) data (Shi et al., 2015a, Shi et al., 2015b).

3.2

Friedel’s salt in mortars

Fig. 4 shows DTG curves for samples taken at different depths of mortar. Three major weight losses are observed for samples not contaminated by chloride: C-S-H/AFm/AFt phases (30 ~ 300 oC), portlandite (400 ~ 520 oC) and CaCO3 (520 ~ 760 oC). When the samples were closer to the exposed surface, additional weight loss peak (230 ~ 410 oC) caused by the decomposition of Friedel’s salt were detected. At the same time, another peak at a lower temperature range (below 150 oC) also associated with Friedel’s salt is identified. Moreover, Fig. 4 reveals that the formation of Friedel’s salt is accompanied by the change of AFm, which is monocarbonate in this study as predicted by thermodynamic modelling on same blends (Dai et al., 2015, Shi et al., 2015b). For the chloride non-contaminated samples at deeper depths, the DTG peak of monocarbonate is much clearer for the MK mortars than for the P and L mortars. This is explained by the increased formation of monocarbonate as a result of additional alumina source from MK as predicted by thermodynamic modelling (Dai et al., 2015, Shi et al., 2015b). According to the method described above, the content of Friedel’s salt can be quantified. The Friedel’s salt profiles are plotted in Fig. 5. The results show that the amount of Friedel’s salt decreases with increasing chloride penetration depth excluding the outer layer. Higher amounts of Friedel’s salt were observed in MK mortars (except for M mortar) compared to P and L mortars. Friedel’s salts were observed at deeper 223


depths for P and L mortars compared to MK mortars, which is in agreement with the chloride ingress depths. Similarly, the quantified portlandite profiles are shown in Fig. 6. Within a 2 mm depth from the exposed surface, a considerable decrease in the portlandite content is observed. This indicates that leaching has occurred during chloride exposure. Fig. 4 shows changes in the shape of the CaCO3 peak for the L mortar. However, the changes in CaCO3 content within the range of 7.8 ~ 8.7 wt.% are too limited to justify the reduction in portlandite by carbonation. The phase assemblage for the samples from outer layer has altered and is reflected in reduction in quantity of total chloride (Fig. 3) Friedel’s salt (Fig. 5) and portlandite (Fig. 6). The portlandite profiles and Friedel’s salt profiles of the MK mortars indicated an overlap between chloride diffusion zone and leaching zone. This complicates discussion regarding the chloride binding capacity of the MK blends.

3.3

X-ray diffraction analysis

The diffraction patterns for the samples taken from 1~2 mm depths of the P and MSL mortars are shown in Fig. 7. The recorded diffractograms are expected to be dominated by high-intense quartz reflection due to the presence of sand in the mortars. The reflections of Friedel’s salt and portlandite were observed at 11.4° and 18.3° 2θ respectively (Paul et al., 2015). Both phases were quantified by Rietveld analysis and the results are summarized in Table 3. In general, the recorded diffractions contain too much sand to be properly analyzed. Therefore, the results can only be used as estimation. It can be seen in Table 3 that a general agreement between XRD and TGA for quantification of Friedel’s salt and portlandite was obtained. The results also show that the chloride-free AFm (e.g. monocarbonate) was hardly detected in the samples in Fig. 7.

4

Discussion

The Friedel’s salt contents as a function of the total chloride concentration for the P and L mortars are shown in Fig. 8. The results show that the amount of Friedel’s salt formed

224

Journal article II

in L mortar is lower than in P mortars for total chloride concentrations ranging from 0.6 wt.% to 1.1 wt.%. This can be explained by the dilution of C3A through incorporation of additional 31.9 wt.% limestone in the L mortars. However, the Friedel’s salt content is not as low as expected by the dilution effect. This indicates higher binding capacity of Friedel’s salt per unit of C3A content for the L mortar. This may be explained by formation of additional AFm phases due to the increase in the degree of hydration as presented in previous work (Shi et al., 2015b). Fig. 5 shows that the amount of Friedel’s salt seems to more or less plateau from 2 mm to 5 and 10 mm depths for the P and L mortars respectively. This indicates that the maximum chemical binding capacity in terms of formation of Friedel’s salt has been reached. The chloride ingress in the MK mortars is too limited to detect such trends. Fig. 5 generally shows that incorporation of MK in Portland cement mortars increases chloride binding in Friedel’s salt possibly due the additional alumina originating from the MK. The results are generally in line with others (Yuan et al., 2009, Thomas et al., 2012, Ipavec et al., 2013). One exception is the lower amount of Friedel’s salt observed for the M mortar. However, this might be due to limited ingress and leaching etc. Leaching causes a reduction in quantities of total chloride and Friedel’s salt. A reduction in chloride content at the exposure surface has also been observed by others (Andrade et al., 1995, Jensen, 1999, De Weerdt et al., 2014). The chemical changes result in larger variations and therefore errors in the total chloride content as shown in Fig. 3. Fig. 8 also shows that Friedel’s salt is formed only after the total chloride content has reached a certain level. This is due to either a detection limit of the TGA method or critical total chloride content is needed for the initiation of the formation of Friedel’s salt. Galan and Glasser (Galan and Glasser, 2015) stated that a phase boundary has to be crossed before Friedel’s salt becomes stable, and the threshold chloride concentrations to form Friedel’s salt varied due to different buffering action of AFm phases. In order to highlight the role of Friedel’s salt on chloride binding, the bound chloride in Friedel’s salt and the total chloride profiles were plotted together in Fig. 9 for the P and L mortars. It was found that the chloride bound in Friedel’s salt is about 0.1 wt.% for both mortars, which is over 10 times less than the maximum total chloride content 225


(about 1.12 wt% of mortar) excluding the surface layer data, revealing a weak contribution of Friedel’s salt to chloride binding. Based on the C3A content (7.8 wt.%) determined for the white Portland cement, the maximum amount of bound chloride in Friedel’s salt is calculated to be no more than about 0.45 wt.% of P mortar, assuming that all C3A is converted to Friedel’s salt. Hence only 22% of all the C3A contributes to the formation of Friedel’s salt. Y. Elakneswaran et al. (Elakneswaran et al., 2009) studied chloride binding isotherm of hydrated cement paste, where the C3A and C4AF contents of the ordinary Portland cement were 9.67 wt.% and 8.82 wt.% respectively. They found that the total bound chloride (by equilibrium concentration technique) and bound chloride in Friedel’s salt (by XRD Rietveld analysis) are about 1.3 wt.% and 0.6 wt.%, when the sample was exposed to 1 M NaCl solution. Such amount of bound chloride in Friedel’s salt is about twice more than that measured in this study. However, it should be noted that a similar plateauing of bound chloride as observed in our study was not observed in their work, which may underestimate the binding capacity of their cement. The above discussion tends to supports the conclusion that the chloride resistance of the studied mortars are controlled by pore connectivity (Shi et al., 2015a). R. Loser et al. (Loser et al., 2010) also found only a limited fraction of chloride bound by cement, while larger fraction was in the solution. They concluded that the effect of chloride binding on chloride ingress was less important compared to the effect of permeability.

5

Conclusion

A novel semi-quantification method was developed to quantify Friedel’s salt in chloride exposed Portland cement-based mortars using TGA. The results agreed well with those quantified by XRD Rietveld analysis. The Friedel’s salt profiles were obtained and compared with total chloride profiles. The measured maximum chloride content in Friedel’s salt was found to be more than ten times lower than the highest total chloride content, indicating a low chemical contribution of Friedel’s salt to chloride binding. Only 22% of the C3A seems to be available to form Friedel’s salt.

226

Journal article II

For mortars with the Portland cement only and Portland cement – limestone blend, the amount of Friedel’s salt seems to reach plateau from 2 mm to 5 and 10 mm depths respectively. Close to the mortar surface, the amount decreased as a result of leaching. The plateauing of the Friedel’s salt profiles indicates an exhausted chloride binding capacity in terms of formation of Friedel’s salt. The chloride ingress in mortars containing metakaolin was too limited to detect such a trend.

6

Acknowledgements

The authors acknowledge the financial support from the LowE-CEM project (project No. 904544) funded by the Danish Research Council.

227


7

References

Andrade C, Diez JM, Alaman A, Alonso C (1995) Mathematical modelling of electrochemical chloride extraction from concrete. Cement and concrete research 25:727-740. Baroghel-Bouny V, Wang X, Thiery M, Saillio M, Barberon F (2012) Prediction of chloride binding isotherms of cementitious materials by analytical model or numerical inverse analysis. Cement and Concrete Research 42:1207-1224. Bentz DP, Garboczi EJ, Lu Y, Martys N, Sakulich AR, Weiss WJ (2013) Modeling of the influence of transverse cracking on chloride penetration into concrete. Cement and Concrete Composites 38:65-74. Bertolini L, Elsener B, Pedeferri P, Redaelli E, Polder RB (2013) Corrosion of steel in concrete: prevention, diagnosis, repair: John Wiley & Sons. Birnin-Yauri U, Glasser F (1998) Friedel’s salt, Ca2Al(OH)6(Cl,OH)·2H2O: its solid solutions and their role in chloride binding. Cement and Concrete Research 28:1713-1723. Dai Z, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015) Investigation of blended systems of supplementary cementitious materials with white Portland cement and limestone (manuscript in preparation). Dai Z, Tran TT, Skibsted J (2014) Aluminum Incorporation in the C–S–H Phase of White Portland Cement–Metakaolin Blends Studied by 27Al and 29Si MAS NMR Spectroscopy. Journal of the American Ceramic Society. De Weerdt K, Colombo A, Coppola L, Justnes H, Geiker M (2015) Impact of the associated cation on chloride binding of Portland cement paste. Cement and Concrete Research 68:196-202. De Weerdt K, Justnes H, Geiker MR (2014) Changes in the phase assemblage of concrete exposed to sea water. Cement and Concrete Composites 47:53-63. Elakneswaran Y, Nawa T, Kurumisawa K (2009) Electrokinetic potential of hydrated cement in relation to adsorption of chlorides. Cement and Concrete Research 39:340-344. Galan I, Glasser FP (2015) Chloride in cement. Advances in Cement Research 27:6397. Garg N, Skibsted J (2014) Thermal Activation of a Pure Montmorillonite Clay and its Reactivity in Cementitious Systems. The Journal of Physical Chemistry C 21:11464-11477. Geiker M (1983) Studies of Portland Cement Hydration by Measurements of Chemical Shrinkage and a Systematic Evaluation of Hydration Curves by Means of the Dispersion Model (thesis): Institute of Mineral Industry, Technical University of Denmark. Geng J, Easterbrook D, Li L-y, Mo L-w (2015) The stability of bound chlorides in cement paste with sulfate attack. Cement and Concrete Research 68:211-222. Grishchenko RO, Emelina AL, Makarov PY (2013) Thermodynamic properties and thermal behavior of Friedel's salt. Thermochimica Acta 570:74-79. 228

Journal article II

Ipavec A, Vuk T, Gabrovšek R, Kaučič V (2013) Chloride binding into hydrated blended cements: The influence of limestone and alkalinity. Cement and Concrete Research 48:74-85. Jain JA, Neithalath N (2010) Chloride transport in fly ash and glass powder modified concretes–influence of test methods on microstructure. Cement and Concrete Composites 32:148-156. Jensen OM (1999) Chloride ingress in cement paste and mortar measured by Electron Probe Micro Analysis: Department of Structural Engineering, Technical University of Denmark. Justnes H (1998) A review of chloride binding in cementitious systems. Nordic Concrete Research Publications 21:48-63. Loser R, Lothenbach B, Leemann A, Tuchschmid M (2010) Chloride resistance of concrete and its binding capacity–Comparison between experimental results and thermodynamic modeling. Cement and Concrete Composites 32:34-42. Lothenbach B, Durdzinski P, De Weerdt K (2016) Thermogravimetric analysis. Chater 5 in: A Practical Guide to Microstructural Analysis of Cementitious Materials. Karen Scrivener, Ruben Snellings, and Barbara Lothenbach (Eds.): Taylor & Francis. Neville A (1995a) Chloride attack of reinforced concrete: an overview. Materials and Structures 28:63-70. Neville AM (1995b) Properties of concrete: Fourth and Final Edition, 4th Edition. Wiley. Paul G, Boccaleri E, Buzzi L, Canonico F, Gastaldi D (2015) Friedel's salt formation in sulfoaluminate cements: A combined XRD and 27Al MAS NMR study. Cement and Concrete Research 67:93-102. Shi Z, Geiker MR, De Weerdt K, Lothenbach B, Kaufmann J, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015a) Durability of Portland cement blends including calcined clay and limestone: interactions with sulfate, chloride and carbonate ions. In: Calcined Clays for Sustainable Concrete, pp 133-141: Springer. Shi Z, Lothenbach B, Geiker M, Kaufmann J, Leemann A, Ferreiro S, Skibsted J (2015b) Experimental and thermodynamic modeling studies on carbonation of Portland cement mortars including metakaolin and limestone (submitted). Skibsted J, Jakobsen HJ, Hall C (1995) Quantification of calcium silicate phases in Portland cements by 29Si MAS NMR spectroscopy. Journal of the Chemical Society, Faraday Transactions 91:4423-4430. Tang L, Nilsson L-O, Basheer PM (2011) Resistance of concrete to chloride ingress: Testing and modelling: CRC Press. Thomas M, Hooton R, Scott A, Zibara H (2012) The effect of supplementary cementitious materials on chloride binding in hardened cement paste. Cement and Concrete Research 42:1-7. Yuan Q, Shi C, De Schutter G, Audenaert K, Deng D (2009) Chloride binding of cement-based materials subjected to external chloride environment–a review. Construction and Building Materials 23:1-13. 229


Table 1. Chemical compositions of the binders (wt.%). wPc

MK

SF

LS

SiO2

21.8

52.8

90.4

3.9

Al2O3

3.6

39.5

0.34

0.33

Fe2O3

0.236

1.4

0.03

0.14

CaO

66.1

0.22

1.37

53.7

MgO

1.1

0.48

0.93

0.35

K2O

0.43

1

1.87

0.05

Na2O

0.04

0.05

0.19

0.08

SO3

3.37

0.06

0.3

0.05

Cl

0.003

0.003

0.13

0.01

LOI (%)

2.6

3.55

3.35

41.8

Table 2. Binder blend compositions of mortars (wt.%). Ratio Mortar labels

Si/Al wPc MK SF

LS

(g/g)

(mol/mol)

P

-

100

0

0

0

L

MK/(MK+LS) = 0

68.1 0

0

31.9 -

ML

MK/(MK+LS) = 0.75

68.1 25.5 0

6.4

1.13

M

MK/(MK+LS) = 0.94

68.1 31.9 0

0

1.13

MSL


68.1 15.6 9.88 6.4

2.36

MS


68.1 19.5 12.4 0

2.36

230

-

Journal article II

Table 3. Summary of the amounts of Friedel’s salt and Portlandite quantified by TGA and XRD (wt.%) Friedel’s salt (wt.%)

Portlandite (wt.%)

Mortar

Sample depths

labels

(mm)

TGA

XRD

TGA

XRD

P

1~2

0.9 ± 0.1

0.8 ± 0.2

2.3 ± 0.3

2.2 ± 0.4

M

1~2

0.6 ± 0.1

1.4 ± 0.5*

< 0.1

< 0.2

MSL

1~2

1.7 ± 0.2

2.0 ± 0.3

< 0.1

0.5 ± 0.2

MS

1~2

1.2 ± 0.2

1.7 ± 0.3

< 0.1

< 0.2

* The sample does not contain any anatase, thus no quantitative Rietveld analysis can be performed. However, considering the intensity and width of the primary reflection from Friedel’s salt in this sample compared to the other three samples, 1.4 wt.% of Friedel’s in this sample would be estimated. Obviously this value is not precise (± 0.5 wt.% at least).

Fig. 1. Characterization of synthesized Friedel’s salt using TGA (Birnin-Yauri and Glasser, 1998, Grishchenko et al., 2013, Lothenbach et al., 2016) 231


Fig. 2. Quantification of Friedel’s salt in chloride exposed mortar cylinder (P) using TGA. The chloride contaminated samples was obtained at mortar depth 1~2 mm, the uncontaminated reference sample was obtained from the same mortar at depth 8~12 mm.

Fig. 3. Total chloride profiles of Portland cement-based mortars

232

Journal article II

Fig. 4. DTG curves of the chloride contaminated and non-contaminated samples from different depths of the Portland cement-based mortars. (Ettr. = Ettringite, Mc = Monocarbonate, Fs = Friedel’s salt, CH = Portlandite, Cc = CaCO3)

233


Fig.5. Friedel’s salt (Fs) profiles of the Portland cement-based mortars; quantification of Fs using TGA.

Fig.6. Portlandite (CH) profiles for Portland cement-based mortars; quantification of CH using TGA.

234

Journal article II

Fig. 7. Quantification of Friedel’s salt by XRD. Samples from P and MSL mortar at 1 2 mm depths. (Fs = Friedel’s salt, CH = Portlandite.)

Fig. 8. Effect of total chloride on formation of Friedel’s salt for Portland cement-based mortars P and L. Fs quantified using TGA.

235


Fig. 9. Comparison of total chloride and chloride (Cl) in Friedel’s salt (Fs) for Portland cement-based mortars P and L. Fs quantified using TGA.

236

Journal article III

Chloride binding of Portland cement paste including metakaolin and limestone Zhenguo Shi (a), Mette Rica Geiker (b), Klaartje De Weerdt (b, c), Tone Anita Østnor (c), Barbara Lothenbach (d), Frank Winnefeld (d), Jørgen Skibsted (a)

(a) Department of Chemistry and Interdisciplinary Nanoscience Center (iNANO), Aarhus University, DK-8000 Aarhus C, Denmark (b) Department of Structural Engineering, Norwegian University of Science and Technology (NTNU), 7491 Trondheim, Norway (c) SINTEF Building and Infrastructure, 7491 Trondheim, Norway (d) Laboratory for Concrete & Construction Chemistry, Swiss Federal Laboratories for Materials Science and Technology (Empa), 8600 Dübendorf, Switzerland

KEYWORDS Fridel’s salt, thermogravimetric analysis, composite cements, chloride binding,

237


Abstract The study investigates the chloride binding of the Portland cement – metakaolin – limestone pastes (P, ML and M). The chloride binding isotherms are determined for samples exposed to CaCl2 and NaCl solutions. The chloride bound in Friedel’s salt is quantified using TGA-based approach developed in the previous study. Thermodynamic modeling is used to predict the phase assemblages and particularly to compare the relative quantities of the formed Friedel’s salt. Comparison of these results allows revealing the role of C-S-H in chloride binding in hydrated Portland cement and blended cements and the impact of cations. It is found that calcium affect the chloride binding in two different ways: (i) it increases the chloride binding in C-S-H, (ii) it increases the chloride binding through formation of Friedel’s salt.

1

Introduction

Deterioration of reinforced concrete structures caused by chloride ingress is a common durability issue, which has to be considered in service life prediction for reinforced concrete structures exposed to sea water and de-icing salt. When the free chloride ions penetrate into the concrete cover and reach a critical level, a de-passivation of the rebar occurs, followed by accumulation of the reaction products, finally degradation of concrete. In chloride contaminated concrete, most chloride ions are free to transport in pore solution, the rest are known to be chemically bound in Friedel’s salt or physically bound on C-S-H. Moreover, it is agreed that only free chloride is harmful to reinforced concrete. Thereby, reducing the free chloride in pore solution by increasing the chloride binding capacity and refinement of pore structures may be beneficial for improving chloride resistance. Chloride binding in hydrated Portland cement has been widely studied. A major debate for the contribution of Portland cement hydrates to bind chloride arose from two statements about which phases between Friedel’s salt and C-S-H dominates the chloride binding capacity. It has been claimed that the main binding of chloride arose from the C3A or C4AF through formation of Friedel’s salt (3CaO·A12O3·CaCl2·10H2O) or its analogue as reviewed by (Justnes, 1998, Yuan et al., 2009, Galan and Glasser, 2015). This statement is usually drawn based on a simple relationship between the amount of

238

Journal article III

the total bound chloride determined by conducting an equilibrium experiment proposed by (Luping and Nilsson, 1993) and the amount of Al2O3 according to the chemical composition of the starting materials. However, (Luping and Nilsson, 1993) reported that the main binding of chloride for Portland cement is ascribed to adsorption of chloride in C-S-H phase. None of the two statements was proposed based on a direct quantification of chloride bound in Friedel’s salt or adsorbed in C-S-H. Chloride binding in hydrated Portland cement – SCMs blends has also been investigated. Substitution of Portland cement with alumina-rich SCMs has been largely reported to increase the chloride binding capacity, which is mainly attributed to the formation of Friedel’s salt (Luo et al., 2003, Thomas et al., 2012). In contrast, addition of silica fume in blended cement was reported to decrease chloride binding (Arya et al., 1990, Arya and Xu, 1995, Thomas et al., 2012). Few published studies have focused on the chloride binding in C-S-H in blended cements. (Zibara et al., 2008) studied the Ca/Si and Ca/Al ratios of the hydration products on the chloride binding capacity of lime-silica fume and lime-metakaolin mixtures. The authors concluded that the chloride is primarily bound in Friedel’s salt in presence of alumina. At low Ca/Al ratio, the initially formed stratlingite converted to Friedel’ salt but at a much lower rate as compared to the conversion of monocarbonate formed at high Ca/Al ratio. In the absence of alumina, the chloride binding in C-S-H was revealed and increased with increasing Ca/Si ratio. A TGA-based method for quantification of Friedel’s salt has been introduced in (Shi et al., 2015a), from which the amount of bound chloride in Friedel’s salt can be quantified. Although the evidence of chloride binding of the Portland cementmetakaolin-limestone blends of real mortars has been acquired, the influence of microstructure on very shallow chloride diffusion layer makes it difficult to quantify the binding capacity of Friedel’s salt. To highlight the interaction of chloride ions with the cement hydrates, well hydrated cement pastes with the same binder blend composition are demanded for chloride binding test. The purpose of this study is to investigate the chloride binding of Portland cement – metakaolin blends. As the limestone contributed to the formation of monocarbonate, a metakaolin blend including limestone was also studied.

239


2

Experimental

The experimental setup including the sample preparation, exposure, and determination of the chloride binding isotherms is based on (De Weerdt et al., 2014). 2.1

Materials

The binders used in this study were made from a white Portland cement (wPc, CEM I 52.5 N) and two supplementary cementitious materials (SCMs): metakaolin (MK) and limestone (LS). The wPc was produced by Aalborg Portland A/S, Denmark, and included 3.1 wt.% LS, 4.1 wt.% gypsum and 1.9 wt.% free lime. The MK was produced in the laboratory from kaolinite (Kaolinite SupremeTM from Imerys Performance Minerals, UK) by thermal treatment at 550 oC for 20 h. The LS was a Maastrichtian chalk from Rørdal, Northern Denmark. The chemical compositions of the starting materials determined by X-ray fluorescence (XRF) are given in Table 1. The wPc contained 64.9 wt.% alite (C3S = 3CaO·SiO2) and 16.9 wt.% belite (C2S = 2CaO·SiO2) and 7.8 wt.% aluminate (C3A = 3CaO·Al2O3) (Dai et al., 2015), where the content of silicate phases were determined by 29Si MAS NMR according to the method proposed by Skibsted et al. (Skibsted et al., 1995) and the quantity of the aluminate phase was determined by mass balance calculation. The Ferrite (C4AF = 4CaO·Al2O3·Fe2O3) phase was not taken into account as the small amount of iron is expected to be incorporated as guest-ions in the C3S, C2S and C3A phases. The sand used to for the mortars was a CEN reference sand (Normensand GmbH, Germany), which has a silica content of at least 98 wt.% and a density of 2650 kg/m3. A superplasticizer (SP, Glenium 27, BASF) was used to achieve similar flow for all mortars. The laboratory grade salts NaCl and CaCl2·6H2O with distilled water were used to prepare exposure solutions with the following chloride concentrations: 0 (reference), 0.125, 0.25, 0.5, 1 and 2 mol/L. The actual concentrations were checked by titration.

2.2

Sample preparation

The binder blend compositions (table 2) are the same as used for mortars in another study (Shi et al., 2015b), where the degree of hydration of the pastes can also be found. Three types of paste (P, ML and M) were reproduced with the same water/binder ratio (w/b = 0.5) by weight. They were mixed according to the laboratory mixing procedure 240

Journal article III

described in (Poulsen et al., 2009), and then cast and sealed in plastic bags. The fresh pastes in the bags were flattened to a thickness of about 5 mm to avoid formation of hydration shells around the hydrating cement grains (Kjellsen et al., 1991). All the pastes were sealed cured in a moist cabinet at 5 oC for the first three days for the same purpose and followed by curing in a moist room at 20 oC for about 2 months. The resulting cement paste plates were crashed by hammer and subsequently ground to 1 mm fine particles. The resulting powder was collected in an one liter polypropylene bottle and mixed with 30 wt.% distilled water by weight of the powdered cement paste. The new mixes with a resulting w/b ratio of 0.95 were stored in bottles and placed on a rotary machine at 20 oC for additional 7 days. The whole curing procedure aimed to maximize the degree of hydration of the cement pastes and minimize carbonation. Each moist cement paste was crushed and homogenized, thus a sample looking like “moist sand” was obtained.

2.3

Chloride exposure

For the chloride exposure, 30 g of the hydrated cement paste (w/b = 0.95) was weighted into 45 mL plastic centrifuge bottle and 15 mL of the chloride solution was added. As reference a sample exposed to same amount of distilled water was also included. The samples were shaken regularly and sealed stored at 20 oC for 2 months prior to analysis.

2.4

Methods

2.4.1

Determination of chloride binding isotherms

The free water content of the wet pastes (w/b=0.95) was determined using thermogravimetric analysis (TGA) prior to chloride exposure. Approximately 400 mg of the wet paste was dried at 105 oC in the TGA purged with N2 for 4 h during which the weight of sample had stabilized. The measured free water contents, i.e. the weight loss at 105 oC, were 36.6 wt.% (P), 39.9 wt.% (ML) and 39.1 wt.% (M) of the initial weight of the corresponding wet pastes. The samples were shaken and subsequently centrifuged after 2 months of chloride exposure. The pH and Cl content of the clear supernatant were determined with the

241


same techniques described in (De Weerdt et al., 2014). The detailed calculation of the chloride binding isotherms was described by De Weerdt et al (De Weerdt et al., 2014). 2.4.2

Quantification of Friedel’s salt by TGA

TGA samples prepared at NTNU Thermogravimetric analysis (TGA) was performed on moist paste samples. A Mettler Toledo TGA/SDTA 851 instrument was used. About 300 mg of the samples were loaded in 150 µm alumina crucible and dried at 105 oC in the TGA purged with N2 for 4 h, followed by heating up to 950 oC at rate 10 oC/min. The weight loss and later quantified Friedel’s salt were reported as mass percentage of dried mortars at 105 oC. TGA samples prepared at EMPA We stopped with isoprop 10 min + ether afterwards to remove the isoprop. The we dried a few minutes at 40 °C. Please keep in mind that your quantification by TGA (and also the GEMS data) should be normalized to 100 g dry solid (i.e. referring to mass ignited at 550 °C). Method for quantification Earlier study (Shi et al., 2015a) showed that Friedel’s salt had two major dissociation peaks in temperature ranges 100 – 150 oC and 200 – 400 oC, which agreed with the temperature range 30 – 236 oC and 236 – oC 403 determined on synthesized Friedel’s salt (Birnin-Yauri and Glasser, 1998, Dilnesa, 2011, Grishchenko et al., 2013). The two DTG peaks reflect the release of water from the interlayer and main layer spaces of the Friedel’s salt, which have a fixed water ratio (4:6). The weight loss due to release of six main layer water can be easily quantified, since the DTG peak is non-overlapping with the weight loss due to other phases in hydrated cements. According to the ratio (4:6) between interlayer water and main layer water, the content of Friedel’s salt in pastes can be quantified by determine the weight loss of water from the main layer (180 oC – 400 o

C). The Friedel’s salt content in pastes was calculated according to the following

equation: mFs =

242

MFs × mH 6 × MH

Journal article III

Where mFs is the mass fraction of Friedel’s salt in pastes (wt. %); mH is loss of water from the main layer (wt. %); MFs is the molar mass (561.322 g/mol) of Friedel’s salt with a chemical composition in the form of Ca4Al2Cl2(OH)12·4H2O; MH is the molar mass (18.015 g/mol) of H2O. Similarly, the portlandite content was also quantified with the same method, by determining the weight loss in temperature range from 400 oC to 500 oC. It was calculated according to equation: mCH =

MCH ×m MH

Where mCH is the mass fraction of portlandite in pastes (wt. %); m is loss of water portlandite (wt. %); MCH is the molar mass (74.09 g/mol) of portlandite in the form of Ca(OH)2; MH is the molar mass (18.015 g/mol) of H2O.

2.4.3

Thermodynamic modeling

Themodynamic modelling calculates the equilibrium phase assemblages in chemical systems from their total bulk elemental composition. The change in phase assemblages and pH due to interaction of hydrated cements with exposure solution were modelled using the Gibbs free energy minimization program, GEMS3 (Wagner et al., 2012, Kulik et al., 2013). The default databases were expanded with the CEMDATA07 database (Lothenbach et al., 2008b). The data includes solubility products of solids relevant for cementitious systems.

3

Results

3.1

Chloride binding isotherms

The chloride binding isotherms for the well hydrated P, ML and M pastes exposed to NaCl and CaCl2 solutions of varying chloride concentrations are given in Figure 1. In general, the amount of the bound chloride increases with increasing chloride concentration. It is also seen that the amount of bound chloride in the P paste exposed to NaCl solution is plateaued when the chloride concentration is over 1 mol/L. This

243


indicates that the maximum chloride binding capacity for the P paste in NaCl solution is reached. The results in Figure 1 also show that partial substitution of the wPc with MK or MK/LS increases the chloride binding as compared to the P paste when they are exposed to same type of chloride solution. The observation is consistent with those from earlier published studies (Coleman and Page, 1997, Thomas et al., 2012). Moreover, there is no major difference in chloride binding between the ML and M pastes, although LS can react with alumina from the MK, and subsequently form more monocarbonate (Antoni et al., 2012, Vance et al., 2013, Dai, 2015). In addition, the higher chloride binding is observed for the pastes exposed to the CaCl2 solution than to the NaCl solution, which has also been reported by several other studies (Tritthart, 1989, Arya et al., 1990, Wowra et al., 1997, De Weerdt et al., 2015). The observation implies that calcium in exposure solution may play an important role in enhancing the chloride binding capacity.

3.2

Chloride in the Friedel’s salt as measured by TGA

The formation of Friedel’s salt though conversion of the AFm phases in the studied pastes after exposure to NaCl and CaCl2 solutions is analysed with TGA, and the DTG curves are shown in Figure 2. The results show that the intensity of the DTG peaks representing the principle layer water (280 – 405 oC) in Friedel’s salt increases with increasing the chloride concentration of the NaCl and CaCl2 solutions. The change of this DTG peak is accompanied by the changes in AFm phases observed at lower temperature around 200 oC in the DTG curves. According to the method for quantification of Friedel’s salt as proposed in chapter 5, the chloride binding isotherm with respect to the chloride bound in Friedel’s salt can be obtained for samples exposed to NaCl and CaCl2 solutions respectively as shown in Figure 3. It is reasonable that the quantified amount of chloride in Friedel’s salt is lower than the total bound chloride. However, it should be noted that the baseline of the DTG curves in Figure 2 for the pastes containing Friedel’s salt is not as horizontal as those for mortar samples presented in (Shi et al., 2015a). In addition, the temperature range representing the weight loss of the main layer water of Friedel’s salt seems to extend to

244

Journal article III

a temperature lower than 280 oC and thus cause some overlaps with the other AFm phases, i.e. mainly monocarbonate. Thereby, the quantification the Friedel’s salt based on the weight loss from 280 – 405 oC underestimates its amount formed in the studied pastes. However, it is believed that the discussion on the trend of chloride adsorption and comparison of the relative amounts are not a problem. The results in Figure 3 show that Friedel’s salt is formed when the chloride concentration reaches a certain level and then increases with increasing chloride concentration. It is expected that more Friedel’s salt is formed for the ML and M pastes than for the P paste at high chloride concentration. Moreover, it is interesting that more Friedel’s salt is observed to form in ML and M pastes when they are exposed to CaCl 2, whereas there is no difference in the Friedel’s salt content in P pastes exposed to NaCl and CaCl2 solutions. Comparison between the chloride bound in Friedel’s salt and the amount of the total bound chloride is presented in Figure 4 for samples exposed to NaCl and CaCl2 respectively. It is seen in Figure 4a that at least 50% of total bound chloride is bound in Friedel’s salt for ML and M pastes when they were exposed to 2 mol/L NaCl solution, whereas nearly 100% of total chloride bound in Friedel’s salt for P paste exposed to 2 mol/L NaCl solution. The results (Figure 4a) reveal a dominant contribution of Friedel’s salt on chloride binding in the case of NaCl exposure. However, the difference between the bound chloride content in Friedel’s salt and total bound chloride content is enlarged with increasing chloride concentration when the samples are exposed to CaCl 2 solution as compared to the NaCl exposure. This observation is particularly evident for the P paste exposed to CaCl2 solution as shown in Figure 4(b). In addition to the formation of Friedel’s salt and changes in AFm phases as identified in Figure 2, a minor decrease in portlandite content is also diretly observed for the ML and M pastes. It is seen that minor amount of portandite present in these samples is depleted at higher chloride concentration. The variations in the portlandite content in the P pastes are presented in Figure 5 based on a TGA quantification. Simialr observation on reduction of portlandite content is observed for the P pastes exposed to both NaCl and CaCl2 solution, but the overall decrease for both of them is clearly not siginificant.

245


3.3

Thermodynamic modelling

Thermodynamic modelling is employed to predict the phase assemblages for the studied pastes exposed to NaCl and CaCl2 solutions as shown in Figure 6. It is predicted that the major hydrates in the P paste before chloride exposure are C-S-H, ettringite, portlandite, monocarbonate, calcite and hydrotalcite. Similar hydrates are predicted for the ML and M pastes but in the absence of portlandite as a result of pozzolanic reaction. The predicted phase assemblages are in good agreement with the experimental data presented by (Dai, 2015). When the NaCl and CaCl2 solutions are added in the hydrated system with increasing chloride concentration, thermodynamic modelling predicts that a conversion of monocarbonate to Friedel’s salt occurs after chloride concentration reaches a certain level for the P, ML and M samples. However, the conversion is facilitated for sample exposed to CaCl2 solution as compared to NaCl solution, which indicate that Ca play a significant role for promoting the formation of Friedel’s salt possibly by accommodating the released carbonate from nomocarbonate phase. Thermodynamic modelling also predicts that a plateau of Friedel’s salt is observed for the P paste both in NaCl and CaCl2 solution. The maximum amount of Friedel’s salt is predicted similar regardless of salt type, which is in good agreement with TGA data (Figure 3). For the ML and M pastes, the plateau of Friedel’s salt is only observed for samples exposed to CaCl2 solution, whereas the Friedel’s salt content keeps increasing for samples exposed to NaCl solution. The latter is also in good agreement with the TGA quantification (Figure 3).

3.4

pH of the exposure solutions

The measured pH of the supernatants of the exposure solutions for the P, ML and M pastes are shown in Figure 7. The results show a lower pH for the ML and M pastes than for the P paste as already known for hydrated blended cements (Lothenbach et al., 2011). It is seen that there is minor increase in pH when the pastes are exposed to NaCl solutions as compared to the pastes exposed to similar amount of distilled water, which has also been observed by (De Weerdt et al., 2015). In line with the previous study (De Weerdt et al., 2015), no major change in the pH with increasing the chloride

246

Journal article III

concentrations is observed for all the pastes exposed to the NaCl solution. However, a substantial decrease in pH is observed with increasing chloride concentrations when the pastes are exposed to CaCl2 solution as also reported by (Tritthart, 1989, Zibara, 2001, De Weerdt et al., 2015).

4

Discussion

The discussion on the chloride binding is based on a general agreement from published studies that the major phases binding chloride are Friedel’s salt and C-S-H. The possible minor uptake of chloride ions by other phases (e.g. ettringite and portalndite etc. as reported in literature (Elakneswaran et al., 2009)) and possible physical adsorption of chloride on positive charged surface of Friedel’s salt (Elakneswaran et al., 2009) are considered to be negligible. Base on the fact, the amount of chloride bound in C-S-H phase (ClC−S−H ) can in principle be calculated from the total bound chloride (Cltotal ) and chloride in Friedel’s salt ( ClFs ) though equation: ClC−S−H = Cltotal − ClFs . However, since the TGA method used for quantification of Friedel’s salt underestimates the real amount of bound chloride in Friedel’s salt, the role of C-S-H on chloride binding will be only discussed qualitatively at the moment.

4.1

Chloride binding in hydrated Portland cement

The maximum chloride binding capacity as observed for the P paste (Figure 1a) exposed to NaCl solution is found to be determined by the Friedel’s salt according to TGA quantification (Figure 3) and thermodynamic calculation (Figure 6a), which is due to smaller amount of monocarbonate initially present in the hydrated Portland cement (Figure 6a). Considering the underestimation of quantified Friedel’s salt content thus eventually small difference between the bound chloride in Friedel’s salt and total bound chloride as shown in Figure 4a, it can be deduced that there is only minor or no uptake of chloride in C-S-H for the P samples exposed to NaCl solution. In contrast, a continuous increase in chloride binding in the P paste with increasing CaCl2 concentration (Figure 1b) suggests that the increased chloride binding in the P paste exposed to CaCl2 solution is dominated by chloride binding in the C-S-H phase, because the amount of the chloride in Friedel’s salt is stabilized at low chloride 247


concentration as seen by TGA quantification (Figure 3) and thermodynamic calculation (Figure 6. 6b). These two conclusions reconcile the discrepancies presented in the published studies for the past several decades about which phase (between C-S-H and Friedel’s salt) dominates the chloride binding capacity. The minor or no uptake of chloride in C-S-H for P samples exposed to NaCl solution observed in this study is also supported by the latest published studies on chloride adsorption in synthetic C-S-H (Plusquellec et al., 2012). The authors found that chloride uptake only occurs in presence of sufficient Ca in exposure solution, and adsorption of two chloride ions is accompanied by one Ca. (Luping and Nilsson, 1993) reported a dominant contribution of the C-S-H phase on chloride binding for hydrated Portland cement. However, it should be noted that their samples were exposed to a NaCl solution saturated with Ca(OH)2.

4.2

Chloride binding in blended cements

Both TGA quantification (Figure 3) and thermodynamic calculations (Figure 6) show that the higher chloride binding for the ML and M pastes as compared to the P paste is ascribed to formation more Friedel’s salt. It is not caused by possible adsorption of chloride in C-S-H, because it is known that Ca/Si ratio is lower for C-S-H in hydrated blended cements than in pure hydrated Portland cement (Lothenbach et al., 2011). Furthermore, it is known that a C-S-H with lower Ca/Si has lower chloride binding capacity (Zibara et al., 2008, Plusquellec et al., 2012). Similar to the P paste, the ML and M pastes also exhibit higher chloride binding in CaCl2 solution than in NaCl solution, which is partially attributed to formation of Friedel’s salt as shown by thermodynamic calculations in Figure 6. (Ipavec et al., 2013) studied the effect of alkalinity on chloride binding in hydrated blended cements and found that the presence of alkalis hindered the formation of chloroaluminate phases and decreased the chloride binding capacity. Besides, the possible enhanced chloride binding in C-S-H in the ML and M pastes exposed to CaCl2 solution may also contribute part of the increased total chloride binding, because it is found calcium can deprotonate the silanol sites in C-S-H and overcharge the initially negatively charged surface of C-S-H, thus adsorption of chloride (Plusquellec et al., 2012, Plusquellec et al., 2013).

248

Journal article III

4.3

Impact of pH on chloride binding

Addition of CaCl2 leads to pH decrease as Ca is incorporated in interlayer of the C-S-H phase by replacing protons (as Ca(OH)2) (Labbez et al., 2011, Plusquellec et al., 2012, Plusquellec et al., 2013, Lothenbach and Nonat, 2015): 2≡SiOH + Ca2+ => (≡SiO)2Ca + 2H+. The authors (Plusquellec et al., 2013) also found that no portlandite is observed from XRD analysis after addition of CaCl2 into C-S-H suspension with Ca/Si 0.8, further confirmed the reduced Ca in solution were absorbed in C-S-H phase rather than formation of portalndite. There is no change in pH for conversion of monocarbonation —

—

to Friedel’s salt according to the reaction: CaCl2 + C4ACH11 => C4A(Cl)2H10 + CC + H2O. Although the phase assemblages for the P pastes in NaCl and CaCl2 solutions are constant as predicted by thermodynamic modelling (Figure 6), a difference in the pH is observed for the two solutions. This may be explained by the following considerations: (i) For samples exposed to CaCl2 solution, presence of CaCl2 inhibits the Ca solubility of C-S-H (same to portlandite) as a result of higher ionic strength, thus no or minor Ca buffer effect to solution is expected from C-S-H (same to portlandite). However, Ca can deprotonates Si-OH in C-S-H, thus more H+ released to solution, which reacts with OH, and then result in both lowering of H+ and OH- in the pore solution as compared, i.e. lowering of pH. (ii) For samples exposed to NaCl solution, addition of NaCl increases the Ca solubility of C-S-H and portlandite (Glasser et al., 2005), the released Ca is required for AFm conversion to Friedel’s to accommodate the released carbonate, thus initially Ca concentration in exposure solution decreased until depletion of monocarbonate, and then the Ca concentration keep rising. The relationship between the bound chloride and the pH of the exposure solutions for all the pastes exposed to the NaCl and CaCl2 solutions are shown in Figure 2. Since the changes in pH of the NaCl exposure solutions are minor, the increased chloride binding for the pastes exposed to NaCl solutions is mainly driven by formation of Friedel’s salt. Quite obvious, it is seen that the chloride binding for the pastes exposed to CaCl 2 solution is highly pH dependent with high correlation coefficients. Similar observation was also reported by earlier studies (Tritthart, 1989, Zibara, 2001, De Weerdt et al., 2015). (De Weerdt et al., 2015) studied the impact of associated cations on chloride binding in hydrated Portland cement, the authors concluded that the measured change in 249


chloride binding depending on the cation was mainly governed by the pH of the exposure solution and thereby the binding capacity of the C-S-H. However, this conclusion does not full agree with the results presented in Figure 8 especially for blended cements, because the increased chloride binding for samples exposed to CaCl 2 is mainly ascribed to formation of Friedel’s salt as seen in TGA quantification and thermodynamic modelling. It seems that the impact of cations on chloride binding is driven by the Ca concentration as proposed recently by (Plusquellec et al., 2012, Plusquellec et al., 2013). Thus, the pH dependency of the Cl binding in the case of CaCl2 exposure (Figure 8) may be explained by its intrinsic link between calcium concentrations of the pH of the pore solution.

5

Conclusions

The chloride binding of Portland cement – metakaolin – limestone blends exposed to varied concentrations of NaCl and CaCl2 solutions is investigated in this chapter. Based on the results and discussion, the following conclusions can be drawn: (1) The use of metakaolin as an SCM increases the chloride-binding capacity compared to P sample and to a similar extent for both the ML and M samples, regardless of the addition of limestone. This is mainly attributed to the formation of a larger quantity of Friedel’s salt in the ML and M samples rather than to the formation of more C-S-H with a lower Ca/Si ratio, as compared to the C-S-H in the P sample, since a lower Ca/Si ratio of the C-S-H phase is known to lower the chloride binding capacity (Plusquellec et al., 2012). (2) The chloride binding increases significantly for the samples exposed to a CaCl2 solution rather than to a NaCl solution. For the P sample, this is ascribed to the higher Ca/Si of the C-S-H for samples exposed to the CaCl2 solution, since the amount of Friedel’s salt is independent on the types of cations according to the TGA data and thermodynamic calculations. However, the thermodynamic modeling clearly shows that the increased chloride binding for the ML and M samples exposed to CaCl2 is ascribed to formation of a larger amount of Friedel’s salt. Moreover, the higher Ca/Si ratio of the C-S-H for the ML and M samples exposed to the CaCl2 solution may also contribute to an increased chloride binding.

250

Journal article III

(3) The pH is found to decrease with increasing CaCl2 concentration, whereas no major changes are observed with increasing NaCl concentration, in accordance with other studies (Plusquellec et al., 2012, De Weerdt et al., 2015). An earlier study of chloride binding in hydrated Portland cement samples (De Weerdt et al., 2015) has suggested that (i) the addition of CaCl2 can cause an increase in the Ca/Si ratio of C-S-H, which may lead to an increase in the adsorption of OH– ions on the C-S-H and hence a lowering of the pH; (ii) the measured change in chloride binding for different cations is mainly governed by the pH of the exposure solution and thereby the binding capacity of the C-S-H. However, this may be different for blended cements, since (i) no major changes in the Ca/Si ratio are predicted with increasing CaCl2 concentration. (ii) the increase in chloride binding for the ML and M samples with increasing CaCl2 concentration is mainly ascribed to formation of different amounts of Friedel’s salt according to the TGA data and thermodynamic modelling.

Acknowledgements The authors would like to acknowledge the LowE-CEM project for the financial support.

251


References Antoni M, Rossen J, Martirena F, Scrivener K (2012) Cement substitution by a combination of metakaolin and limestone. Cement and Concrete Research 42:1579-1589. Arya C, Buenfeld NR, Newman JB (1990) Factors influencing chloride-binding in concrete. Cement and Concrete Research 20:291-300. Arya C, Xu Y (1995) Effect of cement type on chloride binding and corrosion of steel in concrete. Cement and Concrete Research 25:893-902. Birnin-Yauri U, Glasser F (1998) Friedel’s salt, Ca2Al(OH)6(Cl,OH)·2H2O: its solid solutions and their role in chloride binding. Cement and Concrete Research 28:1713-1723. Coleman NJ, Page CL (1997) Aspects of the pore solution chemistry of hydrated cement pastes containing metakaolin. Cement and Concrete Research 27:147154. Dai Z (2015) Solid-state 27Al and 29Si MAS NMR investigations of white Portland cement - metakaolin blends (PhD thesis). Aarhus University. Dai Z, Kunther W, Ferreiro S, Herfort D, Skibsted J (2015) Investigation of blended systems of supplementary cementitious materials with white Portland cement and limestone (manuscript in preparation). De Weerdt K, Colombo A, Coppola L, Justnes H, Geiker M (2015) Impact of the associated cation on chloride binding of Portland cement paste. Cement and Concrete Research 68:196-202. De Weerdt K, Orsáková D, Geiker M (2014) The impact of sulphate and magnesium on chloride binding in Portland cement paste. Cement and Concrete Research 65:30-40. Dilnesa BZ (2011) Fe-containing hydrates and their fate during cement hydration: thermodynamic data and experimental study (thesis). École Polytechnique Fédérale de Lausanne. Elakneswaran Y, Nawa T, Kurumisawa K (2009) Electrokinetic potential of hydrated cement in relation to adsorption of chlorides. Cement and Concrete Research 39:340-344. Galan I, Glasser FP (2015) Chloride in cement. Advances in Cement Research 27:6397. Glasser F, Pedersen J, Goldthorpe K, Atkins M (2005) Solubility reactions of cement components with NaCl solutions: I. Ca (OH) 2 and CSH. Advances in cement research 17:57-64. Grishchenko RO, Emelina AL, Makarov PY (2013) Thermodynamic properties and thermal behavior of Friedel's salt. Thermochimica Acta 570:74-79. Ipavec A, Vuk T, Gabrovšek R, Kaučič V (2013) Chloride binding into hydrated blended cements: The influence of limestone and alkalinity. Cement and Concrete Research 48:74-85.

252

Journal article III

Justnes H (1998) A review of chloride binding in cementitious systems. Nordic Concrete Research Publications 21:48-63. Kjellsen KO, Detwiler RJ, Gjørv OE (1991) Development of microstructures in plain cement pastes hydrated at different temperatures. Cement and concrete research 21:179-189. Kulik DA, Wagner T, Dmytrieva SV, Kosakowski G, Hingerl FF, Chudnenko KV, Berner UR (2013) GEM-Selektor geochemical modeling package: revised algorithm and GEMS3K numerical kernel for coupled simulation codes. Computational Geosciences 17:1-24. Labbez C, Pochard I, Jönsson B, Nonat A (2011) CSH/solution interface: Experimental and Monte Carlo studies. Cement and Concrete Research 41:161-168. Lothenbach B, Matschei T, Möschner G, Glasser FP (2008b) Thermodynamic modelling of the effect of temperature on the hydration and porosity of Portland cement. Cement and Concrete Research 38:1-18. Lothenbach B, Nonat A (2015) Calcium silicate hydrates: Solid and liquid phase composition. Cement and Concrete Research. Lothenbach B, Scrivener K, Hooton RD (2011) Supplementary cementitious materials. Cement and Concrete Research 41:1244-1256. Luo R, Cai Y, Wang C, Huang X (2003) Study of chloride binding and diffusion in GGBS concrete. Cement and Concrete Research 33:1-7. Luping T, Nilsson L-O (1993) Chloride binding capacity and binding isotherms of OPC pastes and mortars. Cement and concrete research 23:247-253. Plusquellec G, Nonat A, Pochard I (2012) Anion uptake by calcium silicate hydrate. Plusquellec G, Pochard I, Nonat A (2013) Anions uptake by calcium silicate hydrates: influence of type of counter-ions and temperature. Poulsen SL, Jakobsen HJ, Skibsted J (2009) Methodologies for measuring the degree of reaction in Portland cement blends with supplementary cementitious materials by 27 Al and 29 Si MAS NMR spectroscopy. IBAUSIL. Shi Z, Geiker MR, Lothenbach B, De Weerdt K, Ferreiro Garzón S, Skibsted J (2015a) Quantification of Friedel's salt by thermogravimetric analysis in chloride exposed mortars with different composite cements (manuscript in preparasion). Shi Z, Lothenbach B, Geiker MR, Kaufmann J, Leemann A, Ferreiro S, Skibsted J (2015b) Experimental and thermodynamic modeling studies of on carbonation of Portland cement - metakaolin - limestone mortars (manuscript in preparation). Skibsted J, Jakobsen HJ, Hall C (1995) Quantification of calcium silicate phases in Portland cements by 29Si MAS NMR spectroscopy. Journal of the Chemical Society, Faraday Transactions 91:4423-4430. Thomas M, Hooton R, Scott A, Zibara H (2012) The effect of supplementary cementitious materials on chloride binding in hardened cement paste. Cement and Concrete Research 42:1-7.

253


Tritthart J (1989) Chloride binding in cement II. The influence of the hydroxide concentration in the pore solution of hardened cement paste on chloride binding. Cement and Concrete Research 19:683-691. Vance K, Aguayo M, Oey T, Sant G, Neithalath N (2013) Hydration and strength development in ternary portland cement blends containing limestone and fly ash or metakaolin. Cement and Concrete Composites 39:93-103. Wagner T, Kulik DA, Hingerl FF, Dmytrieva SV (2012) GEM-Selektor geochemical modeling package: TSolMod library and data interface for multicomponent phase models. The Canadian Mineralogist 50:1173-1195. Wowra O, Setzer M, Setzer M, Auberg R (1997) Sorption of chlorides on hydrated cement and C3S pastes. Frost resistance of concrete 147-153. Yuan Q, Shi C, De Schutter G, Audenaert K, Deng D (2009) Chloride binding of cement-based materials subjected to external chloride environment–a review. Construction and Building Materials 23:1-13. Zibara H (2001) Binding of external chlorides by cement pastes. Zibara H, Hooton R, Thomas M, Stanish K (2008) Influence of the C/S and C/A ratios of hydration products on the chloride ion binding capacity of lime-SF and limeMK mixtures. Cement and Concrete Research 38:422-426.

254

Journal article III

Table 1. Chemical compositions of the binders (wt.%). wPc

LS

MK

SiO2

21.81

3.92

52.84

Al2O3

3.56

0.33

39.49

Fe2O3

0.24

0.14

1.42

CaO

66.13

53.73

0.22

MgO

1.10

0.35

0.48

K2O

0.43

0.05

1.00

Na2O

0.04

0.08

0.05

SO3

3.37

0.05

0.06

TiO2

0.21

0.02

0.88

P2O5

0.04

0.10

0.11

LOI

2.57

41.8

3.55

Density (kg/m3)

3080

2700

2530

Blaine fineness (m2/kg)

387

1211

1891

Carbon content

0.37

-

-

CaCO3

3.1

93.8

-

Table 2. Binder compositions for the produced mortars (wt.%). Mortar id.

wPc

MK

LS

P

100

0

0

ML

68.1

25.5

6.4

M

68.1

31.9

0

255


Figure 1. Chloride binding isotherms for the well hydrated P, ML and M pastes exposed to NaCl (left) and CaCl2 (right) solutions.

256

Journal article III

Figure 2. DTG curves of the pastes exposed to the NaCl and CaCl2 solutions with varied chloride concentrations.

257


Figure 3. Chloride binding isotherms with respect to bound chloride in Friedel’s salt for all the pastes exposed to NaCl (solid line) and CaCl2 (dash line) solutions.

Figure 4. Chloride binding isotherms with respect to total bound chloride (soli line, same to Figure 1) and chloride bound in Friedel’s salt (dash line) for all the pastes exposed to (a) NaCl and (b) CaCl2 solutions.

258

Journal article III

Figure 5. Portlandite content for the P pastes exposed to NaCl (solid line) and CaCl 2 (dash line) solutions.

259


Figure 6. Phase assemblages for the P, ML and M pastes exposed to NaCl (left) and CaCl2 (right) solutions. 260

Journal article III

Figure 7. pH as a function of the chloride concentrations of the exposure solutions for the pastes exposed to NaCl and CaCl2 solutions.

Figure 8. Relationship between the bound chloride and the pH of the exposure solutions for all the pastes exposed to the NaCl and CaCl2 solutions.

261