UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Microfiller effect on rheology, microstructure, and mechanical properties of high-performance concrete Nehdi, Moncef 1998

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1998-272117.pdf [ 35.32MB ]
Metadata
JSON: 831-1.0050312.json
JSON-LD: 831-1.0050312-ld.json
RDF/XML (Pretty): 831-1.0050312-rdf.xml
RDF/JSON: 831-1.0050312-rdf.json
Turtle: 831-1.0050312-turtle.txt
N-Triples: 831-1.0050312-rdf-ntriples.txt
Original Record: 831-1.0050312-source.json
Full Text
831-1.0050312-fulltext.txt
Citation
831-1.0050312.ris

Full Text

MICROFILLER E F F E C T ON R H E O L O G Y , M I C R O S T R U C T U R E , A N D M E C H A N I C A L PROPERTIES OF H I G H - P E R F O R M A N C E C O N C R E T E  by M O N C E F NEHDI B . A . S c , Universite Laval, 1991 M . A . S c , Universite de Sherbrooke, 1993  A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF T H E R E Q U I R E M E N T FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY  in  T H E F A C U L T Y OF G R A D U A T E STUDIES (Department of Civil Engineering)  We accept this thesis as conforming to the required standard  T H E UNIVERSITY OF BRITISH C O L U M B I A APRIL 1998 © Moncef Nehdi, 1998  in  presenting  degree freely  at  this  the  thesis  in  partial  fulfilment  of  University  of  British  Columbia,  I agree  available for reference  copying  of  department publication  this or  thesis by  of this  for  his thesis  and  or  her  Columbia  requirements that  agree  may  be  It  is  representatives.  for financial  Department  DE-6 (2/88)  I further  scholarly purposes  permission.  The University of British Vancouver, Canada  study.  the  gain shall  not  that  the  an  permission for  granted  allowed  advanced  Library shall  by the  understood be  for  that  make it extensive  head  of  my  copying  or  without my written  11  ABSTRACT The objective of this study was to develop a fundamental understanding of the microfiller effect in high-performance concrete. Ultimately, this would help in the development of blended highperformance cements containing recycled materials and industrial byproducts, offering both significant economic advantages and environmental relief.  The mechanisms underlying the microfiller effect on the rheology were investigated in cement paste using a coaxial-cylinders viscometer, a mini slump test, the Marsh cone flow time, and a pressure bleed test. They were also studied in mortars using the A S T M flow-table test, and in concrete using a computer-controlled rheometer, a slump-flow test, the conventional slump test, and an induced bleeding test. It was found that microfillers enhance the superplasticizer efficiency because they increase the surface layer water and reduce the bulk water through a reduced void space in the particulate mixture. In the presence of a superplasticizer, microfillers also decrease the viscosity of concrete mixtures; the finer the particle size of the microfiller the greater the decrease. This seems to be due to a reduction of the mechanical interlocking between coarser particles. Ultrafine particles also decrease the bleed water, which reduces the occurrence of bleed channels and low density microstructural features at interfaces. As a result of the above, microfillers make the production of fluid and self-leveling concrete much easier. It was also demonstrated that triple-blended cements containing pozzolanic and non-pozzolanic fillers can achieve superior rheological properties.  The microfiller effect on mechanical properties was investigated in mortars and in concrete both at early and later ages. It was discovered that this effect depends on the initial porosity of the system. At very low w/b ratios, partial replacement of cement with non-cementitious fillers would not result in lower density hydration products because the initial porosity is already very low. The hydration reactions in fact yielded denser hydration products. Thus, up to 15% replacement of cement by a non-cementitious filler caused significant increases in strength. This was even more significant in triple-blended cements containing combinations of pozzolanic and non-pozzolanic fillers for which up to 30% partial replacement of cement resulted in significant strength increases.  Ill  Ultrafine carbonate fillers increased the very early age strength by about one order of magnitude, because certain microfillers appear to present energetically preferential substrates for the germination and growth of calcium hydroxide. Removal of calcium ions from the solution catalyzes the dissolution of C S in an attempt to regain equilibrium. This signals an earlier end of 3  the induction period and a faster rate of the hydration reactions at early ages.  Quantitative image analysis of backscattered electron micrographs was used to quantify the microfiller effect on the microstructure of high-performance concrete. Analysis was carried out on cement paste and concrete both at Id and at 28d. The acceleration of the hydration reactions at early ages due to carbonate microfillers was confirmed by this technique. Microfillers generally decreased the porosity and refined the microstructural features. This was accompanied by increased strength only when the ratio of inner hydration products to outer hydration products was increased. Densification of the paste-aggregate interface did not seem to necessarily increase the compressive strength.  The microfiller effect in high-performance concrete was studied from the standpoint of the theory of particle packing. A n insight into particle packing models, the effects of particle packing on rheology, and the effects of particle size distribution on hydration reactions was obtained. A new parameter, the microfiller efficiency factor was developed, based on an estimation of packing density and microfiller effect on hydration rate. A close correlation was found between the microfiller efficiency factor and compressive strength. In addition, a new model relating microstructure to strength was proposed. Most available models relate porosity to strength without accounting for the nature of the solid phase. The model proposed herein considers for the first time a quantitative value representing the nature of the hydration products to help estimate strength. This value is the ratio of the dense inner hydration products to the bulk of the rest of the hydration products. The proposed model achieved good estimations of strength.  Overall, this study proposes a new approach to achieving high-strength materials. Traditionally, high strength is obtained through increased cement content, reduced w/b ratios, and high rates of hydration. This work suggests that high strength can be achieved through high initial particle packing combined with a low rate of hydration, which causes less chemical contraction, less drying shrinkage and self dessication stresses, and a higher content of inner hydration products.  iv  ACKNOWLEDGMENTS This research was supported by Concrete Canada, The Network of Centers of Excellence on High-Performance Concrete. I acknowledge the Concrete Technology Lab at Sherbrooke University, and the Civil Engineering Materials Lab at the University of British Columbia in which experimental work was conducted. I sincerely thank my thesis adviser, Dr. Sidney Mindess, for his trust, unconditional support, and for the freedom that all of his graduate students enjoy. He has a quiet but expressive way of appealing to a student to challenge his own limitations. I equally thank my thesis co-adviser, Dr. Pierre-Claude Ai'tcin from Sherbrooke University. Through the years, Dr. Ai'tcin provided infallible support. I thank him particularly for his positive attitude and extraordinary optimism. His high esteem for his students is a demanding responsibility.  I am very grateful to Dr. Arnon Bentur who, during his one-year leave at the University of British Columbia, was a continuous source of inspiration. He provided genuine insights into research issues pertinent to this work. Dr. Sidney Diamond from Purdue University taught me backscattered-electron microscopy and image analysis. I am deeply grateful.  I have immensely benefited from the courses I took at UBC. I really appreciate the contribution of each professor who taught me. I also wish to thank all technicians, secretaries and personnel without whom this work would not be possible.  My stay at the University of British Columbia was a very fulfilling and enjoyable experience. I wish to thank all friends who helped that happen.  V  T A B L E OF CONTENTS  Abstract  ii  Acknowledgments  iv  Table of Contents  v  List of Figures  ix  List of Tables  xii  CHAPTERI INTRODUCTION  1  1.1 Incentive and Objectives 1.2 Presentation and Thesis Structure  1 3  CHAPTER II HIGH-PERFORMANCE CONCRETE AND MINERAL ADMIXTURES 2.1  2.2  2.3  2.4  6  High-performance concrete  6  2.1.1 Definition 2.1.2 Concepts Underlying the Production of High-Performance Concrete Mineral Admixtures in High-Performance Concrete 2.2.1 Need for Mineral Admixtures 2.2.2 Fly Ash 2.2.3 Ground Granulated Blast Furnace Slag 2.2.4 Silica Fume 2.2.5 Rice Husk Ash Potential of Inert Fillers and Triple-Blended Cements 2.3.1 Limitations of Supplementary Cementing Materials 2.3.2 Potential of Inert Fillers and Triple-Blended Cements References  6 7 10 10 11 12 13 14 15 15 16 18  CHAPTER III USE OF L I M E S T O N E POWDER IN C O N C R E T E  20  3.1  Introduction  20  3.2 3.3 3.4 3.5  Limestone Substitution for Gypsum as a Set Regulator Reactivity and Effect on Hydration Reaction Effect on Rheology of Fresh Concrete Effect on Properties of Hardened Concrete 3.5.1 Shrinkage 3.5.2 Heat Release 3.5.3 Mechanical properties  21 23 24 26 26 26 27  vi 3.6  3.7 3.8  Durability Concerns 3.6.1 Carbonation Potential in Limestone Filler Cements 3.6.2 Potential of Steel Corrosion in Limestone Filler Cement 3.6.3 Freeze-Thaw Performance 3.6.4 Sulfate-Attack: a New Concern Conclusions References  29 29 29 31 31 33 34  CHAPTER IV MICROFILLER E F F E C T O N R H E O L O G Y OF C E M E N T PASTES 4.1  Rheology of Dense Suspensions  4.1.1 Introduction 4.1.2 Rheology 4.1.3 Suspensions of Cement Particles in Water 4.1.4 Modeling Rheological Behavior of Cement Paste 4.2 Materials, Apparatus and Procedure 4.2.1 Materials 4.2.2 Apparatus and Procedure 4.3 Experimental Plan 4.4 Microfiller Effect on Superplasticizer Efficiency 4.5 Cohesion and Plastic Viscosity 4.6 Stability of Cement Pastes 4.7 Cement Paste Flow versus Mortar Flow 4.8 Correlation amongst Various Rheological Characteristics 4.9 Conclusions 4.10 References  39 39 39 40 40 41 43 43 44 48 54 55 58 58 60 65 65  CHAPTER V MICROFILLER E F F E C T O N R H E O L O G Y O F H I G H - P E R F O R M A N C E C O N C R E T E 5.1 5.2 5.3 5.4  5.5 5.6 5.7 5.8 5.9 5.10 5.11 5.12 5.13  Introduction Apparatus Materials and Procedure Applicability and Significance of Rheometric Tests for the Rheology of HighPerformance Concrete 5.4.1 Fluid High-Performance Concrete 5.4.2 Self-Leveling High-Performance Concrete Superplasticizer Requirement Flow Resistance Torque Viscosity Slump Loss Induced Bleeding Rheology of Concrete Made with Triple-Blended Composite Cements Ultrafine Particles and the Rheology of Cement Suspensions Conclusisons References  67 67 67 69 71 72 75 78 79 80 82 83 84 86 88 88  CHAPTER  VI  MICROFILLER EFFECT ON MECHANICAL PROPERTIES OF HIGH PERFORMANCE CONCRETE  90  6.1 Introduction 6.2 Investigations on Mortar 6.2.1 Materials and Procedure 6.2.2 Filler Effect on Mechanical Strength of Cement Mortars 6.3 Investigations on Concrete 6.3.1 Experimental 6.3.2 Filler Effect on Compressive Strength 6.3.3 Dependence of Filler Effect on w/b Ratio 6.3.4 Effect of Mixing Technique on Compressive Strength 6.3.5 Filler Effect on Modulus of Elasticity 6.3.6 Filler Effect on Flexural Strength 6.4 Microfiller Effect in Triple-Blended Composite Cements 6.5 Mechanisms of Microfiller Effect on Early Age Properties 6.5.1 Impermeable C-S-H Layer Theory 6.5.2 Osmotic Membrane Theory 6.5.3 Nucleation Mechanism Theory 6.5.4 Nucleation of C-S-H or Nucleation of CH? 6.5.5 Effect of Calcium Ions Dissolved from Calcite 6.5.6 Effect of Reaction Between Calcite and C A 6.6 Conclusions 6.7 References 3  CHAPTER  90 90 90 91 95 95 98 101 102 103 106 107 110 Ill 112 113 114 116 117 118 120  VII  MICROFILLER EFFECT ON MICROSTRUCTURE OF HIGH-PERFORMANCE CONCRETE 123 7.1 7.2 7.3 7.4 7.5 7.6 7.7  Introduction Background Specimen Preparation Image Acquisition Strategies Quantifying the Microstructure Microfiller Effect on Early Age Microstructure Microstructure of the Aggregate-Cement Paste Interface 7.7.1 Anhydrous Cement Phase 7.7.2 Porosity 7.7.3 Calcium Hydroxide 7.8 Microfiller Effect in Concrete Versus Cement Paste 7.9 Conclusions 7.10 References CHAPTER  123 123 126 128 129 131 133 133 136 137 138 140 141  VIII  PARTICLE PACKING, RHEOLOGY, AND MICROSTRUCTURE VS. PROPERTY RELATIONSHIPS: A QUANTITATIVE APPROACH 8.1 Towards a Quantitative Science of Cement Based Materials  143 143  viii 8.2 Microfiller Effect and Particle Packing 8.2.1 Theoretical Considerations of Particulate Mixtures a) Wall Effect b) Interference Effect 8.2.1 Modeling Particle Packing a) Linear Packing Density Model b) Model of Aim and Goff c) Model of Toufar, Klose and Born 8.3 Particle Packing and Rheology 8.4 Microfiller Effect and Hydration Reactions 8.4.1 Nucleation and Growth Model 8.4.2 Phase-Boundary Model '. 8.4.3 Diffusion-Controlled Model 8.4.4 Total Degree of Hydration and Effect of Particle Size Distribution 8.5 General Model for the Microfiller Effect 8.5.1 Initial Particle Packing 8.5.2 Rheology 8.5.3 Degree of Hydration 8.5.4 Model of the Microfiller Efficiency 8.5.5 Application of the Model 8.6 Microfillers and Microstructure vs. Property Relationships 8.6.1 Models for Microstructure-Strength Relationships 8.6.2 Mechanical Strength versus Porosity 8.6.3 New Model for Microstructure-Strength Relationships 8.7 Conclusions 8.8 References  144 144 145 145 146 146 148 149 150 152 152 153 154 154 157 158 159 159 159 160 167 167 167 168 176 177  CHAPTER IX  SUMMARY AND GENERAL CONCLUSIONS  180  ix  LIST OF FIGURES Fig. 2.1: Fig. 2.2:  Fig. 3.1: Fig. 3.2: Fig. 3.3: Fig. 3.4:  Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig.  4.1: 4.2: 4.3: 4.4: 4.5: 4.6: 4.7: 4.8: 4.9:  Fig. Fig. Fig. Fig.  4.10: 4.11: 4.12: 4.13:  Fig. 4.14:  Fig. Fig. Fig. Fig. Fig.  5.1: 5.2: 5.3: 5.4: 5.5:  Fig. 5.6: Fig. 5.7: Fig. 5.8: Fig. 5.9: Fig. 5.10:  Schematic representation of floes of cement particles Schematic representation of fresh and hardened concrete at high and low w/c ratios  Effect of limestone substitution for gypsum on (a) initial and final set, (b) water consistency, and (c) 28 d compressive strength Effect of limestone addition on (a) yield stress and (b) plastic viscosity Early age compressive strength for low-heat cement with different proportions of limestone filler Effect of partial replacement of cement by limestone on compressive strength  7 9  22 25 27 28  Schematic of the helical mixer Schematic of the modified Marsh cone Determination of the superplasticizer saturation dosage Schematic of the coaxial-cylinders viscometer Illustration of the uniform-precision experimental plan Rheological characteristics of OPC + L F cement pastes Rheological characteristics of OPC + 10% SF + L F cement pastes Shear thinning behavior of low superplasticizer dosage cement pastes Isoresponse curves for the temperature after mixing of the various cement pastes Non linearity of the Marsh cone flow time (w/b = 0.40, % L F = 12.5%) Correlations between rheological parameters at comparable shear rates Correlations between rheological parameters at different shear rates Induced bleeding versus (a) plastic viscosity; and (b) superplasticizer saturation dosage Correlation between superplasticizer saturation dosage and superplasticizer required  45 46 46 47 49 50 51 56  Illustration of the U B C Rheometer Mixing sequence Illustration of flow curves at different intervals of time for fluid HPC Maximum torque requirement for fluid HPC mixtures Correlation between (a) slump and g, and (b) slump and h for fluid HPC mixtures Illustration of typical flow curves for self-leveling HPC mixtures Superplasticizer requirement for a constant workability Flow resistance at various ages for (a) fluid, and (b) self-leveling HPC mixtures (filler = 15%) Torque viscosity at various ages for (a) fluid and (b) self-leveling HPC mixtures (filler =15%) Slump variation of concretes made with different proportions  68 70 73 74  57 59 61 62 63 62  75 76 78 80 81  X  Fig. 5.11: Fig. 5.12: Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig. Fig.  6.1: 6.2: 6.3: 6.4: 6.5: 6.6: 6.7: 6.8: 6.9: 6.10: 6.11:  Fig. 6.12: Fig. 6.13: Fig. 6.14: Fig. 6.15: Fig. 6.16: Fig. 6.17: Fig. 6.18: Fig. 6.19: Fig. 6.20:  Fig. 7.1: Fig. 7.2: Fig. 7.3: Fig. 7.4: Fig. 7.5: Fig. 7.6: Fig. 7.7:  Fig. 7.8:  Fig. 7.9:  of various fillers Microfiller effect on induced bleeding of fresh high-strength concrete Statistical modeling of various rheological responses for OPC-LF2-SF triple-blended composite cements Particle size distribution of the cement and limestone filler Isoresponse curves for compressive strength of mortars at various ages Comparison of filler effect in binary and ternary blended cements Illustration of the Omni Mixer Experimental set-up to measure the modulus of elasticity Illustration of the flexural strength test Filler effect on compressive strength (w/b = 0.33) Dependence of microfiller effect on the w/b ratio Effect of mixing technique on compressive strength Filler effect on modulus of elasticity at various ages Modulus of elasticity versus compressive strength for various HPC mixtures Filler effect on flexural strength Dependence of 28d flexural strength on fineness of the filler (w/b = 0.33, filler = 15%) Compressive strength iso-response curves at various ages for HPC made with LF2-SF-OPC triple-blended cements Typical isothermal calorimetric curve for the hydration of Portland cement Microfiller effect on early-age strength of HPC Compressive strength ratio at 12h of limestone HPC mixtures to hydrated lime HPC mixtures Illustration of consumption of C a C 0 during the hydration of cement paste Comparison of the effect of partial replacement of cement by limestone and hydrated lime on compressive strength (w/b = 0.33) Comparison of the effect of partial replacement of cement by limestone and hydrated lime on modulus of elasticity (w/b = 0.33) 3  Occurrence and detection of secondary and back-scattered electrons Illustration of micro-cracking due to severe oven drying during specimen preparation Illustration of the image acquisition system Example of phase segmentation in a hydrated cement paste Microfiller effect on relative area of microstructural phases in cement pastes at Id (w/b = 0.33) Microfiller effect on early age compressive strength Illustration of (a) general view of H P C at low magnification, (b) porosity of the paste aggregate transition zone, (c) gradient of microstructure at the interface, and (d) general view of the bulk cement paste Microfiller effect on relative area of anhydrous grains in a 50 pm transition zone, in the bulk cement paste matrix in concrete, and in an equivalent cement paste at Id (w/b = 0.33) Illustration of the wall effect in HPC  82 83 85 91 92 95 97 97 98 99 102 103 104 105 106 107 109 Ill 114 116 118 119 119  124 128 129 130 132 132  134  135 135  xi Fig. 7.10:  Fig. 7.11:  Fig. 8.1: Fig. 8.2.: Fig. 8.3: Fig. 8.4:  Fig. 8.5: Fig. 8.6:  Microfiller effect on relative area of porosity in a 50 urn transition zone, in the bulk cement matrix in concrete, and in an equivalent cement paste (w/b = 0.33, 28d) Microfiller effect on relative area of C H in a 50 urn transition zone, in the bulk cement matrix in concrete, and in an equivalent cement paste (w/b = 0.33, 28d)  Illustration of the wall effect Illustration of the interference effect Schematic illustration of the evolution of the heat of hydration vs. time, and corresponding hydration phases Illustration of (a) the particle size distribution of the cement and microfillers, and (b) the particle packing density of the binary-cement containing various proportions of each microfiller Correlation between packing density and torque viscosity for microfiller cements (microfiller = 15%, w/b = 0.33) : Illustration of the effect of initial time of nucleation on the degree of hydration at 12h (at K = 0.009) Illustration of the effect of rate of growth of hydration products on the degree of hydration at 12h Illustration of the combined effect of initial nucleation time (t ) and rate of growth of hydration products (K ) on the degree of hydration at 12h Illustration of the modeling of the microfiller efficiency factor at early ages as a function of the particle packing ratio and the degree of hydration Relationship between microfiller efficiency factor and compressive strength at 12h for cements containing 15% of various microfillers Schematical Illustration of the microstructure vs. strength model bse micrographs of 0.33 and 0.25 w/b concrete mixtures Bse micrographs illustrating the paste-aggregate interface in high and low w/b ratio concrete mixtures Calibration of the microstructure versus compressive strength model G  Fig. 8.7: Fig. 8.8:  N  Fig. 8.9:  Fig. 8.10:  Fig. 8.11: Fig. 8.12: Fig. 8.13: Fig. 8.14:  136  138  145 146 155  161 162 164 164  G  165  165  166 170 171 173 175  xii  LIST OF TABLES  Table 2.1Table 4.1Table 4.2Table 5.1Table 5.2Table 7.1Table 7.2Table 8.1-  Annual Production And Utilization Rates Of Silicious By-Products In Tons Various Models For The Rheology Of Cement Paste Chemical And Physical Properties Of The Cement And Microfillers Used Fineness And Average Particle Size Of The Cement And Fillers Mix Proportions Of Concrete Mixtures Mean Atomic Numbers, Backscattering Coefficients, And Gray Levels Usually Observed For The Major Components Of Hydrated Cement Paste Contrast Values For Pairs Of Various Adjacent Components Of Hydrated Cement Paste Microstructure And Strength Data To Calibrate Model Of Equation 8.32  15 .42 44 69 70 127 127 174  To my parents Larbi and Zohra in recognition and love. Their sacrifices, patience and affection are beyond description.  Chapter I  1  Chapter I  INTRODUCTION  1.1 I N C E N T I V E A N D O B J E C T I V E S  The study of high-performance concrete (HPC) has been an extremely active research area in recent years, and the material is now used widely in economically competitive field applications. Designers and contractors have grasped the special opportunities offered by this material to implement it in an increasing variety of structural applications. The strength, durability, and slenderness of structures thus achieved were dreams only a few years ago. Owing to this cumulative effort, a comprehensive technology has been developed around HPC. However, merely achieving ever higher strength HPC's has lost its excitement. The industry currently faces the practical difficulties of implementing the use of this material on construction sites. The real challenge is to bridge the gap between materials and structural engineers, and to introduce this new concrete technology as a design tool, though there are still concerns regarding the possibly enhanced brittlness of HPC, its relatively higher material cost due to a high cement content, and the more rapid loss of workability.  The high cement requirement of HPC is both an ecological and an economic concern. After aluminum and steel, cement is the most energy intensive construction material. The production of one ton of cement is not only costly in terms of fossil fuels, but also releases about one ton of CO2 into the atmosphere. The very low water content in HPC mixtures results in only a partial hydration of the cement grains. Thus, a significant proportion of the cement acts merely as a filler, and its potential hydraulic activity is not fully utilized. In the wake of a potential energy crisis in the next century, the threat of global warming, and the increasing cement consumption of a steadily growing world population, the use of cement has to become more rational. The traditional extra 100 to 2 0 0 kg of cement per cubic meter of HPC compared to conventional concrete has to be seriously reconsidered.  Furthermore, the high cement content  may cause a significant  chemical (le Chatelier)  contraction, and may generate an excess of heat of hydration, which could cause thermal  Chapter I  2  cracking in massive structural elements. The extra cement also causes serious rheological difficulties in the production of high-performance concrete. As an "esquisse", the author quotes his thesis co-director who used to say: 1  "at a water/cement ratio of 0.50 the compatibility  problems are submerged in the water, at 0.40 they start to resurface, and at 0.30 you ha cope with them ". The delicate equilibrium between the reactivity of  C 3 A and the solubility of the  sulfates is a major issue in very low water/cement ratio systems. The superplasticizer tends to compete with the calcium sulfate to neutralize the C3A, and rapid loss of workability is frequently reported. In addition, the angular cement grains which tend to flocculate do not provide an optimal particle distribution for the production of HPC. Anyone who has tried to mix a 0.25 water/binder (w/b) mixture with pure portland cement would understand these rheological difficulties.  In view of the above concerns, the concrete industry has resorted to the use of several supplementary cementing materials for the production of HPC. Most of these are recycled industrial byproducts, which allow fossil fuels to be saved, cement raw materials to be preserved, and hazardous emissions into the atmosphere due to cement production to be reduced by as much as 55% for a 60% blast-furnace slag cement. Most importantly, these materials impart to the concrete superior workability, strength, and durability characteristics. However, the mechanisms underlying these improvements are still a matter of a considerable controversy. Whether the positive effect of mineral additions is due to their pozzolanic activity or to a physical filler effect, or whether they improve the cement paste or the aggregate-paste bond, is still in dispute.  Once considered merely as waste products, some of these minerals have become more expensive than portland cement. Others have regional availability problems and high transportation costs associated with their use in concrete. The ideal situation would be to find a mineral for the partial substitution of cement which is available at all cement plants, which imparts the positive effects mentioned earlier, without compromising the fundamental characteristics of the  cement.  Limestone powder offers an attractive option in this regard. Although the effect of using limestone powder in conventional concrete has been extensively studied, its effect in low w/c ratio superplasticized concrete is not fully understood. Generally, the mechanisms underlying the  'This expression is attributed to Professor P-C. Ai'tcin (translation from French by the author).  Chapter I  3  effects of inert and nearly inert fillers in H P C have not received due investigation. This is especially true when they are used with other pozzolanic fillers in triple-blended binders.  Likewise, modern concretes seem to be shifting from having a granular packing to behaving as dense suspensions. The industry is developing a self-leveling material which consolidates under its own weight even in heavily reinforced structural elements. To achieve the desired flow properties and inhibit segregation and bleeding, such a material requires a high powder content, often without the need for ultra-high strengths. In this regard, the partial substitution of cement by available and less costly inert fillers presents an interesting scenario. Currently in Japan, the production of self-consolidating concrete involves a systematic addition of limestone powder.  The present study undertakes the task of clarifying the mechanisms of the action of microfillers in low w/b ratio superplasticized concrete. The effects of pozzolanic and non-pozzolanic ultrafine particles, whether used separately or in triple-blended binders, are examined. Understanding these mechanisms could ultimately allow the optimization of blended cements, leading to both ecological and economic benefits. A modest 10% partial replacement of cement by a non-burned filler would save 10% on the cost of cement production, and reduce hazardous emissions into the atmosphere by an equal amount. By the same token, the amount of binder would be increased by 10%, while providing cements with improved performance. This is certainly worth the investment.  1.2 PRESENTATION AND THESIS STRUCTURE The topics addressed in this study are presented in 9 chapters. Each chapter is written to be as self-contained as possible. In order to set the stage for the reader, a brief summary of chapters 2 to 9 is given below.  Chapter 2 illustrates the basic concepts underlying the production of high-performance concrete. The importance of reducing the water/cement (w/c) ratio, the effects of superplasticizers, the selection of materials, and the need for ultrafine particles are discussed. This is followed by a brief outline of the use of mineral additions in H P C . The most common minerals such as silica fume, fly ash, blast-furnace slag, and rice husk ash are discussed. The chapter provides an  4  Chapter I  introduction to the use of inert and nearly-inert fillers in HPC, both individually and in tripleblended cements.  Since this study deals primarily with the effect of limestone powder in HPC, Chapter 3 provides a state-of-the-art report on the use of limestone powder in concrete. First, the early attempts to substitute limestone for gypsum as a set regulator are presented. Then, there is a summary of the effects of limestone filler on the early hydration reactions of cement, on the rheological properties of concrete, and on its mechanical performance and hardened properties. Although durability aspects are not a part of this investigation, durability advantages and concerns associated with the use of limestone powder in concrete are presented in order to provide a general overview.  There is a question as to whether microfiller cements can provide rheological advantages in addition to their ecological and economic benefits. The microfiller effect on the rheology of cement pastes needs to be investigated because of the subsequent effect on the rheology of concrete, and the potential use of microfiller cements in grouting applications. Chapter 4 describes the statistical modeling of the particular effects of limestone microfiller on the rheology of superplasticized cement pastes both in silica fume and non-silica fume systems. A factorial experimental plan was used in conjunction with various rheological tests. This permitted the modeling of the various rheological responses over the selected experimental domain, and an examination of the correlations amongst the various rheological parameters.  Chapter 5 describes the investigation of the microfiller effect on the rheology of HPC. Unfortunately,  one  cannot  rely  on  traditional  workability  tests  for  the  rheological  characterization and quality control of HPC. Hence, the first part of this chapter deals with the applicability and significance of rheometric tests for this purpose. Then, there is a discussion of the microfiller effect on the superplasticizer requirement, flow resistance, torque viscosity, slump loss, and induced bleeding of HPC mixtures. The rheology of HPC made with triple-blended cements is also addressed.  In chapter 6, the mechanical properties of HPC made with various proportions of pozzolanic and/or inert fillers having various particle sizes are presented. The compressive strength was monitored from the age of 12 hours till 91 days. The modulus of elasticity and the flexural  Chapter I  5  strength were also measured for various mixtures. The effects of the nature of the microfiller and its fineness on the mechanical performance of HPC are discussed. This chapter also includes an insight into the mechanisms that underly the increase of the rate of early age hydration in microfiller cements.  In chapter 7, an insight into the microstructure of HPC made with microfiller cements is presented. Quantitative backscattered electron microscopy was used to compare the effects of pozzolanic and inert fillers in HPC. Measurements at various ages of unhydrated cement grains, porosity, and calcium hydroxide were carried out both in cement pastes and concretes. The microstructure of the aggregate/cement paste interface is also discussed, and the filler effect in cement paste versus concrete is addressed.  In chapter 8, an effort is made to unify the rheological, mechanical, and microstructural measurements of the previous chapters into single theoretical model that can, potentially, have predictive capabilities. Theoretical considerations for particulate mixtures along with different models for the packing density of granular materials are presented. A n attempt at models that can predict concrete properties from the knowledge of the fillers or predict mechanical strength from quantitative measurements of microstructure is presented.  In chapter 9, the general conclusions are presented.  Chapter II  6 Chapter II  HIGH-PERFORMANCE CONCRETE AND MINERAL ADMIXTURES  2.1 HIGH-PERFORRMANCE CONCRETE  2.1.1 Definition  In the 1950's, concrete with 35 MPa compressive strength was considered high-strength [1]. One decade later, 40 to 50 MPa concrete was commercially available. In the 1970's, 60 MPa concrete was being produced.  Due to recent developments in the cement and chemical admixtures technology, the strengths that can be achieved have been steadily increasing. On site, the 28-day strength reached 115-120 MPa in Two Union Square, Seattle [2], and a pedestrian overpass in Sherbrooke, Quebec, has just been constructed with a design concrete strength of 200 MPa. Indeed, newly developed cement based materials can now achieve strengths far beyond those which can be implemented using current structural design practice. Many concrete design codes still have an upper strength limit of 65 MPa for structural applications, although new codes such as the Norwegian standard have introduced an upper limit of up to 105 MPa. To date, the applications of HPC have been mainly in high-rise buildings, bridges and structures performing under severe exposure conditions. Yet, there are several other potential applications. HPC, with its relatively high compressive strength per unit weight per unit cost, is often the least expensive means of carrying compressive forces, and allows the construction of lighter and more slender members [3]. However, in practical applications, the interest has gradually shifted from the compressive strength towards other properties of the material, such as a higher modulus of elasticity, higher density, and better resistance to chemical and/or physical attack [2]. Thus, the material is now generally referred to as High-Performance Concrete. It is a versatile concrete with properties which can be "tailor made" to satisfy the increasingly demanding performance criteria.  Chapter II  7  2.1.2 Concepts underlying the production of H P C  a)  Water/cement ratio and superplasticiser effect  The strength of concrete is an inverse function of its total void content. Thus, the crucial criterion in the production of high-strength concrete is the use of low w/c ratio systems, coupled with an optimal consolidation and curing of the concrete. This was expressed by Feret [4] as early as 1897 as follows:  J  where f  c  0  \c + w + a  is the compressive strength, and c, w, and a are, respectively, the volume of cement,  water and air. Reducing the water content and properly consolidating the concrete have long been known to increase the strength. However, workability considerations generally impose a practical limit to the reduction of the w/c ratio. At one time, the w/c ratio could not be reduced below a value of about 0.40. In simple terms, the workability constraint arose from the tendency of cement grains to flocculate and thus to exhibit a higher shear resistance at lower water contents (Fig. 2.1). This flocculation is due to the unsaturated surface charges of the cement grains when they come into contact with a polar liquid such as water.  In order to alleviate the flocculation and reduce the amount of mixing water, certain organic molecules having dispersing properties were successfully developed. The first of these were lignosulfonates obtained from papermill waste [5]. However, it was soon discovered that the  Chapter II  8  dosage of these molecules could not be increased to high levels without the entrapment of large air bubbles and without retarding the setting of the cement due to the excess of sugars from the wood. The concrete industry remained at this stage for about 30 years.  The real development that led to the production of H P C was the manufacture of synthetic molecules having superior dispersive properties. These were condensate salts of sulfonated naphthalene formaldehyde and sulfonated melamine formaldehyde, which appeared almost simultaneously  in  Japan  and  Germany,  respectively.  These  new  products,  termed  superplasticizers, suffered from one major drawback: their action was limited to about 15 to 30 minutes. Progress has been made since then, and superplasticizers can now maintain the workability for up to around 45 minutes, and for longer times when coupled with retarders. However, the industry still faces difficulties such as cement/superplasticizer compatibility problems and the stability of the entrained air in superplasticized mixtures.  The effect of reducing the w/c ratio on the structure of concrete is illustrated schematically in Fig. 2.2. At a high w/c ratio, the cement grains are distant from each other. Larger and more crystalline hydration products are developed along with high porosity. Reducing the w/c ratio reduces the distance between the cement particles at the onset of hydration. A denser and less crystalline structure is thus produced.  b)  Selection of cement  The cement to be used in H P C has to meet two essential requirements: it must develop an appropriate strength, and exhibit an appropriate rheological behavior. Commercial cements meeting the A S T M C-150 Standard Specifications for Type I Portland Cement have successfully been used to make HPC. However, these cements vary considerably in terms of both their chemical composition and their fineness [6, 7]. Therefore, even though the cements generally meet the strength criterion, rheological difficulties in producing very low w/c ratio concrete are frequently encountered. The kinetics of the hydration of a cement that is too fine do not allow 1  enough time during which the workability of the concrete can be maintained. The interstitial phases of the cement (C3A and C4AF), the reactivity of the C3A, and the content of calcium sulfate and the proportions of its various forms in the ground cement (gypsum, hemihydrate, and  Even a previously compatible cement/superplasticizer couple can display compatibility problems when a different production run of the same brand is used.  Chapter II  9  Fig.  2.2  Schematic representation of fresh and hardened concrete at high and low w/c ratio [2].  Anhydrate) are crucial for the rheological behavior of cement. The selection of a superplasticizer with an appropriate solids content, length of molecular chain, and residual sulfates content is required for a satisfactory rheological performance.  c)  Selection of aggregates  In normal strength concrete (NSC), the aggregate requirements usually pertain to the workability and durability. These requirements also apply for H P C . The particle shape, particle size distribution, and surface area of the aggregates should be selected to minimize the water demand. For NSC, in which the strength depends mainly on the quality of the cement paste, the strength of the aggregate itself is not crucial. In fact, the strength achieved can even be higher than that of the  aggregate  itself  when  lightweight  aggregates  are used.  However,  for  H P C the  aggregate/cement paste bond must be strong enough to transfer significant stress to the aggregate.  Chapter II  10  Trans-granular failures going right through the aggregate particles are commonly observed on fracture surfaces of HPC specimens. Hence, the aggregate can become the weak link, and may constitute the limiting factor for the strength of the concrete [8]. The mineralogical nature of the aggregate affects the bond properties to the hydrated cement paste, and this can influence the strength and the elastic properties of HPC. Results of studies on the effect of the maximum aggregate size on the strength of HPC are mixed. On the one hand, larger aggregates can reduce the water demand, which increases strength. On the other hand, the size reduction process (crushing) can eliminate internal defects in aggregate particles, and results in a smaller cement paste/aggregate transition zone, which also enhances strength.  2.2 MINERAL ADMIXTURES IN HIGH-PERFORMANCE CONCRETE 2.2.1 Need for Mineral Admixtures  Contrary to the commonly held opinion, mineral admixtures such as silica fume are not required to produce high-strength concrete. Strengths of up to 90 MPa can be achieved using portland cement alone. However, rheology, durability, and economy requirements combine to impose the need for supplementary cementing materials in HPC.  At the low w/c ratios and high cement contents typically used in HPC, it is difficult to control the delicate equilibrium between the reactivity of the C A and the solubility of sulfates. A high 3  superplasticizer dosage is usually needed, and the workability is lost quickly. Also, due to the high cement content, the hydration reaction generates excessive heat. This results in thermal contraction stresses and cracking in massive structural elements. Furthermore, to achieve ultrahigh strengths, the reduction of the w/c ratio alone is not sufficient to reduce the porosity to the required value. The usual gradation of the cement grains cannot achieve an optimal particle packing, and a wall effect at the aggregate/cement paste interface will be observed. Finally, economic and environmental considerations call for the replacement of a part of the cement by recycled and/or unburned materials. A large variety of industrial or naturally occurring materials such as volcanic ash, zeolite, diatomaceous earth, metal powder, etc. can be used in concrete. However, only the most commonly used mineral admixtures to produce HPC are described in the following.  11  Chapter II  2.2.2 Fly Ash  Generally 15-20% of the coal burned in power plants is transformed to ash. Once emitted into the atmosphere and causing pollution, these fine powders can now be recycled. Depending on the type of coal, 70-85% of the ash produced is fly ash. This product was first used as a concrete admixture at Hungry Horse Dam in the United States [9], and is now one of the major mineral admixtures used worldwide in concrete. Fly ash is broadly classified into two classes: Class F fly ash is generally produced by burning anthracite or bituminous coal. It has only pozzolanic properties since it contains less than 5% calcium oxide. Class C fly ash is normally produced by burning lignite or sub-bituminous coal. Since its lime content can exceed 10%, class C fly ash has cementitious properties in addition to its pozzolanic properties. Most of the fly ash used in concrete is Type F; Type C fly ash is produced mainly in the United States and Canada.  Usually, class F fly ash includes 50-60% SiO , 20-30% A1 0 , as well as CaO, Fe 0 , carbon, z  2  3  2  3  and other impurities. High amounts of unburned carbon are undesirable since they tend to adsorb chemical admixtures such as air entrainers and inhibit their effect. The rapid cooling of the molten ash causes the silica phase to be amorphous, which gives the fly ash its pozzolanic reactivity with lime. CaO acts as an activator in class C fly ash, and C-S-H can thus be formed, but the rate of hydration is usually slow. Conversely, class F fly ash needs an activator such as alkalis, gypsum, and Ca(OH) for cementing properties to be developed. 2  The nitrogen B E T surface area of fly ash is usually in the range of 0.4-1 m /g. A special attribute 2  of fly ash is the spherical shape of its particles, which improves the flowability of concrete. However, the properties of fly ash can vary over time even during one day of production. The type of coal, type of boiler, burning temperature, and mode of operation combine to control the shape of particles, their surface area, amorphous nature, carbon content, etc. Non-spherical or non-amorphous ash, or ash that contains lumps of carbon may also be produced, but these are not suitable for use in concrete.  Owing to its spherical particles, fly ash can reduce the water demand of concrete mixtures when used to replace some of the cement. Bleeding and drying shrinkage can be reduced as a consequence. Fly ash is not highly reactive. When used to replace some of the cement in HPC, the heat of hydration can be reduced significantly. Therefore fly ash concrete is very suitable for  Chapter II  12  massive structures. The pozzolanic reaction in fly ash proceeds slowly, and the early age strength of fly ash concrete tends to be lower than that of conventional concrete. At later ages the strength of fly ash concrete usually exceeds that of pure portland cement (OPC) concrete; its density is higher, and its permeability is lower. However, longer curing is required to achieve these longterm benefits. Thus, underwater massive structures prove to be an excellent application for highperformance fly ash concrete.  Concrete made with a good quality fly ash and properly moist cured is denser than OPC concrete, and better resists the ingress of deleterious fluids. The corrosion resistance in sea water is enhanced compared to an OPC concrete, the expansion due to alkali-aggregate reaction is reduced, and the frost resistance is comparable for similar strength and air entrainment. The alkalinity of the concrete is reduced, but published data show that the depth of carbonation is similar to that of an ordinary concrete with a similar compressive strength.  2.2.3 Ground Granulated Blast-Furnace Slag  Slag is a byproduct separated from molten metals as a low melting point substance by adding agents such as limestone in the process of refining the metal. Slag obtained from the steel making process is referred to as blast-furnace slag, while other types are known as nickel slag, copper slag, lead slag and so on. When quenched rapidly with water or air to a glassy state and finely ground, slag has a high latent hydraulicity, and can be used as a mineral admixture in concrete. The first granulated blast-furnace slag (GBFS) was produced in Germany in 1923.  The chemical composition of GBFS depends on the impurities present in the iron ore, the ash cokes used as the reducing agent, and the limestone or dolomite added. It comprises 27-40% Si0 , 5-33% A1 0 , 30-50% CaO and 1-21% MgO. Minor constituents such as Fe 0 , MnO, S0 , 2  2  3  2  3  3  and Na 0 are usually present. Typically, the Blaine fineness of GBFS is around 300-400 m /kg, 2  2  similar to that of ordinary portland cement. The finer the slag, the higher its reactivity. Some standards [9] use the basicity, (CaO+MgO+AL 03)/Si02, as an index of the reactivity of GBFS. 2  Being a ground product, GBFS has highly angular and irregularly shaped particles. In contact with water, a layer having low permeability to water forms around the slag particles and inhibits the dissolution of ions. Thus, activators are usually added to slag to initiate and catalyze its  Chapter II  13  reaction. The most commonly used activators are alkalis, gypsum, Ca(OH) , and portland 2  cement.  The effect of GBFS on fresh concrete depends on its fineness and the proportions used. The water demand to achieve a constant workability is usually decreased, and the bleed water can be significantly reduced when the slag is finely ground to 6000-8000 cm /g. A delay in the initial 2  strength is usually observed with GBFS concrete, especially at low curing temperatures. The heat of hydration is usually decreased when slag is used, though there may not be a reduction at high curing temperatures (higher than 30 °C). High proportions of slag usually suppress the alkaliaggregate reaction in concrete, and reduce the corrosion of steel since the ingress of chloride ions is decreased. These benefits require, however, an adequate curing of the concrete.  2.2.4 Silica Fume  Silica fume (SF), also known as microsilica, is a byproduct of the production of silicon metal and ferro-silicon alloys. It usually has 85-98% Si0 content, a spherical shape with diameters 2  typically 100 times finer than those of cement grains (mean particle size between 0.1 to 0.2 pm) and an amorphous structure. The earliest mention of using SF in concrete was in a U.S. patent in 1944 [10]. However, SF-blended cements were first produced on a small scale in Iceland, Sweden and Norway only since 1976. The benefits of this new material were not fully understood until the early comprehensive studies at the end of the 1970's [11]. The amorphous structure of SF, its high Si0 content and large surface area (about 20 m /g) are 2  2  thought to make it very reactive with the calcium hydroxide Ca(OH) produced by cement 2  hydration, to form additional calcium silicate hydrate binder (C-S-H). This reaction is more significant at the paste-aggregate transition zone. In non-silica fume concretes, large crystals of calcium hydroxide (CH) usually form in abundance in the transition zone, and are preferentially oriented, acting thus as the weakest link of hardened concrete. This transition zone is also under the greatest stress due to the elastic mismatch between the cement matrix and the relatively stiffer coarse aggregate.  Silica fume also increases the internal cohesion of fresh concrete, reducing drastically some traditional local areas of weakness, such as bleed-water channels and the accumulation of bleedwater beneath the coarse aggregate particles. The CH formation at the paste aggregate interface  Chapter II  14  is consequently reduced. The thickness of the transition zone is significantly decreased as the amount of silica fume incorporated in the concrete mixture increases [12].  Moreover, owing to their small size and spherical shape, the SF particles tend to act as a filler. They fit into the void spaces between the relatively coarser cement grains which would otherwise be occupied by water, and which is then not free to fluidify the concrete. Conversely, SF particles tend to adsorb water because of their high surface area, increasing thus the water demand. The overall effect of SF on the water demand depends amongst other parameters on the w/(c+SF) ratio, the amount of SF added and the presence of a superplasticizer. Again, the physical filler effect of SF particles could be more significant at the paste-aggregate interfacial zone, whose densification is believed to improve the mechanical properties and durability of concrete. Finally, SF particles may act as nucleation sites to increase the hydration rate of cement, and to increase the homogeneity and fineness of the hydration products, through a socalled grain refinement process. 2.2.5 Rice Husk Ash  Rice husks are a byproduct of the rice milling industry. One fifth of the yearly world production of 500 million tons of rice is husks. Upon completion of combustion of one ton of husk, 200 kg of ash are obtained [13]. A controlled incineration of 500-700 °C is needed to obtain a noncrystalline ash of high pozzolanic value [14]. Rice husk ash (RHA) usually consists of 80-95% of Si0 , 1-2% alkalis (K 0 and Na 0), and 3-18% unhurried carbon. Its BET surface area ranges 2  2  2  from 60-100m /g. 2  Usually, the water demand of concrete containing RHA increases significantly, but this can be overcome by the use of superplasticizers. Bleeding is almost eliminated, but the heat of hydration is not decreased. RHA was shown to increase the strength by about 15% in HPC compared to a control OPC concrete [15]. Due to its high surface area, RHA increases the demand for superplasticizer and air entraining admixtures compared to OPC and silica fume concretes. RHA concrete also has excellent resistance to chloride penetration, to freezing and thawing cycles, and to scaling by deicing salts.  Chapter II  2.3  15  POTENTIAL OF INERT FILLERS AND TRIPLE-BLENDED CEMENTS  2.3.1 Limitations of Supplementary Cementing Materials  Table 2.1 lists the amounts of the major mineral admixtures produced as byproducts in various countries, and the level of utilization of these resources. There is still a great potential for the implementation of these materials in the concrete industry. In recent years, new research work has geatly expanded the use of mineral admixtures in concrete. For instance, up to 40% low calcium fly ash can now be incorporated into superplasticized concrete to produce high-strength concrete [13]. Also, ultra-low heat cements made of proportions of slag, fly ash, and portland cement have been used in large scale bridge construction in Japan [9]. Silica fume has became a familiar ingredient of HPC in leading North American cities where ready mix HPC is commercialized.  Table 2.1- Annual Production and Utilization Rates of Siliceous By-Products in Tons [16]  2  Country  Fly Ash x  10*  Blast Furnace Slag x 10*  Condensed Silica Fume x 10  3  Production  Utilization  Production  Utilization  Production  Australia  3.5  0.25  4.7  0.12  60  ' 20  Canada  3.3  0.8  2.9  0.2  23  11  China  35  7.2  22  16  None  None  1  0.45  None  None  None  None  France  5.1  1.5  10.4  1.9  60  None  Germany (Fed. Rep)  2.6  2.0  15  2.8  25  None  India  19  0.5  7.8  2.8  None  None  Japan  3.7  0.5  24  8.2  25  None  Netherlands  0.5  0.3  1.1  1  None  None  None  None  0.1  None  140  40  South Africa  12.9  0.1  1.5  0.6  43  0  Sweden  0.1  0.02  0.1  0.03  10  1  United Kingdom  13.8  1.3  1.5  0.25  None  None  United States  47  5  13  1  100  2  Denmark  Norway  Utilization  These figures have certainly changed since 1988. In particular, the use of silica fume has increased and a shortage of this material in the future is foreseen.  Chapter II  16  Nonetheless, the use of supplementary cementing materials is not without problems. The properties of these minerals are variable, and often little care is paid to their control during production since they are only byproducts. The quality control of concrete and its sensitivity to curing become more acute when mineral admixtures are used. The dosage of air entraining admixtures to achieve a certain air content increases dramatically in silica fume, fly ash, rice husk ash, and to a lesser extent in GBFS concretes. Furthermore, the early age development of strength is slower in slag/fly ash concrete. This can hinder their use in jobs which are restrained to a tight schedule for form removal, and can be particularly troublesome in winter concreting. In addition, silica fume and rice husk ash concretes exhibit an increased tendency for plastic shrinkage.  In many countries, no standards have yet been established to regulate the use of mineral admixtures in concrete. Data regarding such materials are often not available, and there has not been sufficient effort to transfer the technology of using mineral admixtures to developing countries. Many cement plants still do not include provisions for the production of blended cements containing mineral admixtures. The cost of transportation associated with the use of such materials dissuades some cement plants from pursuing this venture. Also, mineral admixtures are not available in all places. Some byproducts such as silica fume have become even more expensive than cement, and a future shortage of this product can be expected.  2.3.2 Potential of Inert Fillers and Triple-Blended Cements  Conventionally, mineral admixtures have been used to reduce the cost of concrete. However, with new developments  in the chemical admixtures technology, mineral additions are  increasingly being used to improve the performance of concrete. In this respect, the strength is generally not the most critical issue since the current strength values obtainable with HPC are beyond those that structural engineers require for the design of new structures. In many applications, large amounts of mineral admixtures are desired only as a fine powder (filler) rather than as a cementing material. With the increasing cost of labor, concrete is shifting towards a self-leveling material that consolidates under its own weight. Some heavily reinforced structures require a high fluidity concrete that can be placed easily without vibration. Unfortunately, material segregation and bleeding are characteristic of this kind of concrete. To cope with this situation, the water content should be reduced through the use of superplasticizers, and the fines  Chapter II  17  in the concrete mixture should be increased. For this purpose slag or/and fly ash can be used, with the additional benefits of lower heat of hydration and higher long term strength. Yet, these materials are sometimes not available, a high cost of transportation is associated with their use in concrete, and their effect on the early strength development can be detrimental. In these cases, inert fillers, such as limestone which is available in all cement plants , can impart the rheological 3  benefits and enhance the early age strength. Furthermore, if the mechanisms of action of these microfillers are well understood, they can even improve the strength of HPC.  Other opportunities for the concrete industry lie in the field of triple and multiply-blended cements. Often, single mineral admixtures have drawbacks. However, when combined with other mineral admixtures, a countermeasure for the unwanted behavior can be obtained. Some simple examples can illustrate this concept. Suppose a slow hydrating pozzolan has to be used in a HPC mixture, yet removal of formwork is an issue and early strength is desired. The addition of an inert filler such as limestone powder , which can enhance the early strength, could be a solution. By the same token, suppose that a large proportion of an inert filler has to be substituted for cement to reduce the heat of hydration, yet, the strength should not be compromised. In this case, incorporating a very efficient supplementary cementing material such as silica fume can solve 4  the problem. An excellent practical example for this is the use of belite cement (in which C S 2  content is higher than 50%) in conjunction with slag and fly ash to produce an ultra-low heat cement with an acceptable initial strength. This material has been used in Japan in submerged bridge abutments. The previous option for this application was to use normal cement with fly ash and slag, which also provides low heat. Yet neutralization of the material occurs quickly and durability problems can be expected.  In the following chapters, the rheological and mechanical performance of triple-blended cements including pozzolanic and non-pozzolanic fillers will be presented. Likewise, in view of the importance of limestone powder in this study, chapter 3 provides a state of the art analysis of the use of limestone powder in concrete.  In fact not all limestone powder can be used for this purpose. Requirements regarding the fineness, calcite content, presence of clay or organic matter in the limestone filler, etc., should be satisfied. 3  Chapter II  18  2.4 REFERENCES  [I]  Gjorv, O.E. [1992], "High strength concrete", Advances in concrete technology, Malhotra, V.M., ed., CANMET, pp. 21-57.  [2]  Ai'tcin, P.C. and Neville, A. [1993], "High performance concrete demystified", Concrete International, Vol. 15, No. 1, pp. 21-26.  [3]  Ahmad, S.H. and Shah, S.P. [1985], "Structural properties of high strength concrete and its implications for precast prestressed concrete", PCA Journal, Vol. 30, No. 6, pp. 92119.  [4]  Feret, R. [1987], Bull. Soc. Encour. Ind. Nat. Paris, No 2, p. 1604.  [5]  Ai'tcin, P.C. [1992], "High-performance concrete: from material to structure", Yves Malier, ed., E&FN Spon, London, pp. 14-33.  [6]  Mehta, P.K. and Ai'tcin, P.C. [1990], "Microstructural basis of selection of materials and mix proportions for high-strength concrete", Utilization of High-Strength Concrete Second International Symposium, ACISP-121, pp. 265-286.  [7]  Mindess, S. [1994], "Materials selection, proportioning, and quality control", HighPerfornance Concretes and Applications, S. P. Shah and S. H. Ahmad, eds, Edward Arnold, London, pp. 1-25.  [8]  Ai'tcin, P.C. and Mehta, P.K. [1990], "Effect of coarse-aggregate characteristics on mechanical properties of high-strength concrete", ACI Materials Journal, Vol. 87, No. 2, pp. 103-107.  [9]  Nagataki, S. [1994], "Mineral admixtures in concrete: state of the art and trends", Concrete Technology: Past, Present, and Future, V.M. Malhotra Symposium, ACI SP144, P.K. Mehta, ed., p. 447.  [10]  Roberts, L.R. [1989], "Microsilica in concrete, I", Materials Science of Concrete I, J. Skalny, ed., American Ceramic Society, pp. 197-222.  [II]  Khayat, K.H. and Ai'tcin, P.C. [1992], "Silica fume in concrete: an overview", Proceedings of ACI 4 International Conference on Fly ash, Silica Fume and Natural Pozzolans in Concrete, ACI SP-132, Vol. n, Istanbul, Turkey, pp. 835-871. th  [12]  Sellevold, E.J. [1987], "The function of silica fume in high strength concrete", Proceedings: Utilization of High-Strength Concrete, Stravanger, Norway, pp. 39-49.  [13]  Mehta, P.K. [1992], "Rice husk ash - a unique supplementary cementing material", Proceedings: ACI-CANMET International Symposium on Advances in Concrete Technology, Athens, Greece, V.M. Malhotra, ed., pp. 407-430.  Silica fume has an efficiency factor of about 3 to 4 when used in small percentages as a replacement for cement, i.e. 1 kg of SF can replace 3-4 kg of OPC and still result in a similar strength of concrete.  Chapter II  19  [14]  Malhotra, V.M. [1993], "Fly ash, slag, silica fume, and rice-husk ash in concrete: a review", Concrete International, Vol. 15, No. 4, pp. 23-28.  [15]  Hong, M. and Malhotra, V.M. [1996], "High-performance concrete incorporating rice husk ash as a supplementary cementing material", ACI Materials Journal, Vol. 93, No. 6, pp. 629-636.  [16]  RILEM Technical Report [1988], 73-SBC RLLEM Committee, "Final Report on Siliceous Byproducts for Use in Concrete", Materials and Construction, Vol. 21, No. 121, pp. 69-80.  20  Chapter III  Chapter III USE OF LIMESTONE POWDER IN C O N C R E T E  3.1 INTRODUCTION In the wake of the oil shortage of 1974, the United States Congress passed a law mandating cement producers to reduce their energy consumption. By 1985, 92% of the clinker produced in the U.S. was burned using coal or petroleum coke, as compared to only 36% in 1972. The energy crisis also stimulated a worldwide interest in adding fillers to portland cement as a means of saving energy. Since limestone is available at all cement plants and thus has no additional transportation costs associated with its possible use as a filler in cement, this subject has gained increased interest. The question of whether limestone additions should be permissible has stimulated a great deal of debate and research. Various national standards have adopted different positions. The French cement standards have allowed up to 35% filler additions to portland cement since 1979. The filler cements (known as CPJ) have developed considerably since then, and represent more than 60% of the total cement production in France [1]. The European pre-standard ENV 197 permits portland cement to contain 5% of additional constituents such as limestone, and defines a portland-filler cement which may contain up to 20% of high purity limestone. It was suggested that this upper limit be increased to 30% [2] before the standard was actually adopted in 1992.  The Canadian Standards Association has resolved the issue of limestone additions to cement in CSA-A5-M83, where it recognizes the existence of an optimum carbonate addition to cement, and allows a maximum of 5% limestone of specified quality to be added to Type 10 (normal) portland cement. This was later extended to Type 30 (high early strength) cement. The option of limestone addition is not used by all cement plants, since not all cements may contain carbonate additions. In the U.S. however, a 1985 proposal before the ASTM C-l committee to allow up to 5% limestone to be ground with the clinker has resulted in a considerable controversy. The proponents suggest that the cost and energy effectiveness can be improved without a degradation in the overall quality of cement, and even report some improvements in cement and concrete properties. The opponents claim that ground limestone acts merely as an adulterant and that its addition to cement should be rejected on ethical grounds [3, 4].  Chapter III  21  However, some new issues have arisen that have not yet received proper analysis. For instance, the chemical effects on the hydration reactions occurring upon the addition of limestone to cement and the subsequent effects on the rheology of fresh concrete have been extensively studied. However, the same effects in a low w/c ratio concrete in the presence of a superplasticizer are still not clearly understood. In addition, the use of carbonate additions to produce cements of different strength grades is well documented. Yet, the possible use of ultrafine limestone to produce fluid HPC, increase the early age strength, or/and alter the porosity and the interfacial properties in high strength mixtures, has not been fully investigated. In this chapter, the aim is to provide a state of the art analysis of the use of limestone powder in concrete. This should shed light on subsequent chapters which will focus on the different mechanisms of microfiller action in HPC, with a particular emphasis on limestone powder.  3.2 LIMESTONE SUBSTITUTION FOR GYPSUM AS A SET R E G U L A T O R Amongst the early attempts to use limestone in cement was limestone substitution for gypsum as a set regulator. Some cement plants experience seasonal problems related to the dehydration of gypsum due to high grinding mill temperatures, with subsequent false set tendencies. Since limestone is often mined at costs around 1/10 the cost of mining gypsum [5], the financial impact of such substitution is quite clear, especially in areas of the world where natural gypsum deposits are rare, which results in high transportation costs as well.  Regulations regarding the addition of gypsum and the permissible S0 content in cement go as 3  far back as the 1890's. Yet, as a result of the energy crisis, the cement industry resorted to fuels having higher sulfur contents, which brought about new concerns regarding this issue. The gypsum additions should be kept to a minimum to account for the additional clinker S0 content 3  and to keep the S0 level within specification limits. Many attempts have been made to find an 3  adequate replacement for gypsum [6], among which limestone has attracted the most interest.  Bobrowski et al. [5] found that with the correct limestone/gypsum ratio for a given clinker, the autoclave expansion and the setting time could be maintained at acceptable values, the concrete strengths were favorable except at early ages, and the severe false setting systems were altered dramatically. Later work by Bensted [7, 8] and by Negro et al. [6], has confirmed that a partial substitution of limestone for gypsum of up to 50% was possible without deleterious effects to the  Chapter III  22  cement performance (Fig. 3.1). However, it was also noted that each clinker-limestone-gypsum system should be individually investigated to ascertain the optimum replacement rate. It was also observed [9] that the optimum S0 decreases exponentially with an increase in the limestone 3  content.  In this regard, it would be of interest to investigate the set regulation effect of ultrafine limestone in modern low w/c ratio superplasticized concrete. In such a system, the rapid loss of workability due to the delicate equilibrium between the reactivity of C A and the solubility of the sulfates is a 3  major issue.  Fig. 3. 1 Effect of limestone substitution for gypsum on (a) initial and final set, (b) water consistency, and 28d compressive strength [8].  Chapter III  23  3.3 REACTIVITY AND E F F E C T ON HYDRATION REACTIONS Limestone has usually been considered as an inert additive in a hydrating cement paste. Although there is agreement that limestone is not pozzolanic, several studies have shown that it has significant reactivity. As early as 1938, Bessey [10] suggested the existence of calcium carboaluminates which result from the reaction of carbonates and calcium aluminate according to the chemical reaction: C A + CaC0 +11H 0 3  3  2  C A.CaC0 .llH 0 3  3  (3.1)  2  Manabe et al. [11] observed that calcium monocarboaluminate hydrate (C A.CaC0 .llH 0) was 3  3  2  formed in an equimolar mixture of C A and CaC0 and water. They also observed that in the 3  3  presence of sulfate and carbonate salts, both trisulfoaluminate and monocarboaluminate were formed. Feldman et al. [12] studied the influence of CaC0 on the hydration of C A. Mixtures of 3  3  different proportions were observed using differential thermal analysis, X-ray diffraction and infrared absorption. It was concluded that the hydration of C A was suppressed by CaC0 3  3  additions due to the formation of calcium carboaluminate on the C A grains. This effect was also 3  observed by Ramachandran and Chun-Mei [13] who noticed further that ettringite formation and its conversion to monosulfoaluminate were accelerated in the presence of CaC0 . Conversely, 3  Vernet [14] has suggested that the early formation of ettringite is similar to the condition in which no limestone is added. But when the gypsum is exhausted, the limestone reaction dominates,  and after  C A is 3  consumed  the  stable compounds  are  ettringite  and  monocarboaluminate.  Ramachandran and Chun-Mei [15] suggested that in the presence of CaC0 , the hydration of C S 3  3  was accelerated as the particle size became finer and the amount of CaC0 was increased. There 3  was some evidence of partial incorporation of CaC0 into the C-S-H phase, a slight modification 3  in the C/S ratio of C-S-H and an increase in the microhardness and the density of the hardened cement paste (hep). Husson et al. [16] and Barker and Cory [17] reported similar trends where limestone addition to portland cement enhanced the hydration of C S and the formation of 3  calcium hydroxide, probably because it offered nucleation sites for its growth. On the other hand, Ushiyama et al. [18] suggested that the addition of small amounts of carbonates retards the early hydration of alite, while the addition of large amounts accelerates its hydration. More recently, work by Evrard et al. [19] has highlighted the reactivity of calcite in an alkaline environment and has confirmed the hypothesis of Ramachandran and Chun-Mei. [15] that some reaction occurs  Chapter III  24  between calcite and C-S-H. Several other investigations dealing with the reactivity of limestone and the formation of carboaluminates have also been reported [20-23].  Other evidence of the reactivity of limestone comes from the use of limestone aggregates in concrete. Farran [24] studied the effect of the mineralogical nature of the aggregate on its bond with cement paste. He concluded that calcareous aggregates formed more intimate bonds compared to other minerals, and that calcite cannot be considered as truly inert in concrete. Buck and Dolch [25] examined the bond of different aggregates to cement paste in concrete; limestone aggregate offered the best bonding strength. It was observed more recently by Ai'tcin and Mehta [26] that high strength concrete made with limestone aggregate showed higher strength and elastic modulus than gravel concrete, due to interfacial reactions. Grandet and Ollivier [27] explained the lower calcite orientation at the cement paste-carbonate aggregate interface by the reactivity of limestone and the presence of monocarboaluminates in the interfacial zone. On the other hand, Monteiro and Mehta [28] explained the same reactivity between carbonate rock and cement paste by a possible reaction between calcite and calcium hydroxide, which resulted in the formation of a basic calcium carbonate hydrate.  Although the data on the reactivity of limestone are extensive, it is not clear how important this reactivity is, whether it occurs under field conditions, and how it varies in different concrete mixtures and in the presence of different admixtures.  3.4 EFFECTS ON T H E R H E O L O G Y OF FRESH CONCRETE  Generally, limestone fillers complement the deficiency in fine particles of the particle size distribution of cement (CaC0 having mean particle sizes of 5 um [1, 29] and 3 um [38] have 3  been used), which can enhance the stability of fresh concrete and mortar. In addition, the fine CaC0 grains can fill in between the relatively coarser cement grains. Beyond the lubricating 3  role they can play, the filler particles can reduce the room available for water, decreasing the water demand and obstructing the capillary pores at later ages [1, 29, 30].  Bombled [29] noted that when the effect of limestone filler on the rheology of fresh concrete is significant, it is usually in the sense of a higher cohesion, better plasticity and in some instances a pseudo-acceleration of the setting. Neto and Campitelli [31] used a two-point test to characterize the rheology of limestone filler cements. They observed that there was a reduction in the yield stress as the limestone addition increased, and also an increase of the plastic  Chapter i i i  25  viscosity beyond a certain fineness/limestone relationship (Fig. 3.2). Brookbanks [32] has presented the results of a comprehensive study on the effect of limestone additions to cement in the range of 5% to 28% on the properties of fresh concrete. He observed that the setting times were marginally reduced as the limestone addition increased. For the materials used, there was no discernible difference in the water demand at 5% replacement, while a marginal reduction was observed at higher filler levels. The limestone fillers greatly reduced the bleed water and seemed to have no effect on the air entraining properties. The mechanisms underlying the effect of limestone filler on the rheology of cement paste and concrete seem to be controlled by the filler particle size and its flocculation capacity. In most of the above studies, limestone was interground with clinker, which increases the overall fineness due to the lower grinding energy of limestone. Whether the use of blended ultrafine limestone when deflocculated with a superplasticizer can play a different role, similar to the filler effect of silica fume on the rheological properties of cement paste, was not fully investigated. Being a grinding mill product with a non-spherical shape and unsaturated surface electrical charges may limit the effect of ultrafine limestone on the rheological properties. Yet, it was noticed that the lubricating effect of limestone filler was dominant for the lower w/c ratio concretes [1, 33].  Fig. 3.2 Effect of limestone addition on (a) yield stress, and (b) plastic viscosity [31].  Chapter III  26  3.5 EFFECTS ON PROPERTIES OF HARDENED CONCRETE 3.5.1 Shrinkage  One concern about the addition of interground or blended limestone is the adverse effect on shrinkage. Adams and Race [34] observed appreciable increases in the 4-day shrinkage for all cases of both fine and coarse limestone fillers at different levels of addition. This effect was explained by the increased fineness and the possible increase in the C-S-H formation at early ages due to nucleation effects, which increases the gel content of the paste and subsequently the amount of shrinkage. It seems that cements with lower C A contents are more prone to shrinkage 3  upon limestone addition [4], and that the shrinkage increases with the fineness of the filler and occurs mostly within the first 24 hours [33]. Conversely, Escadeillas [1] suggested that the shrinkage of filler cements is not really different from that of OPC cements in nonsuperplasticized concrete, especially for long term behavior. For superplasticized concrete, the limestone filler cement concretes showed even lower shrinkage values than the OPC concretes. It is thus not yet clear whether limestone fillers will have any adverse effects on the shrinkage in low w/c ratio superplasticized concrete.  The possible nucleation mechanism due to the fine limestone particles may inhibit the formation of larger portlandite crystals, and subsequently reduce their possible shrinkage restraining effect. In addition, the structure and the amount of the C-S-H formed may itself be changed in the presence of microfillers. It is not yet clear to what extent each of these parameters can influence the apparent shrinkage. The effect of microfillers on shrinkage remains an open question. 3.5.2 Heat Release  Barker and Matthews [35] studied the heat release characteristics of limestone filler cements. Using isothermal conduction calorimetry, they observed that the addition of increasing amounts of limestone to portland cement progressively decreased the peak rate of heat evolution and the total heat released. When introduced by blending, limestone was shown to retard the main peak of heat evolution compared to OPC, while it resulted in an advance of the main heat evolution peak when inter-ground with cement, probably due to an increased specific area of cement. A 25% limestone addition was sufficient to reduce the peak temperature rise to about 8°C, as compared to 11°C for 25% fly ash cement.  Chapter III  27  Hooton [36] has compared the 7-day heat of hydration according to the A S T M C186 Test for Heat of Hydration of Hydraulic Cement for commercially available pure cements and limestone filler cements made from the same clinkers. There was no consistent effect of limestone additions on the heat of hydration. Accounting for the commercial levels of limestone addition and the enhancement of the cement fineness due to inter-grinding, no major effects on the heat of hydration can be expected. It was observed [1] that within the first hours, the heat release of filler cements is higher than the reference pure cements especially for the finer fillers. This was probably due to the acceleration of the hydration of C S because of the nucleation effect. At later 3  ages, the amount of heat release was reduced with increasing levels of fillerization, which is of interest for mass concrete applications.  Fumy a et al. [37] used limestone powder in triple-blended cement at addition rates of up to 150 kg/m of concrete. Large amounts of limestone were helpful in developing mass concrete with 3  higher workability, lower heat release and better stability. More recently, work by Kessal [38] has shown that it is possible to use limestone filler with low heat cements to obtain high early strengths (Fig. 3.3) without reducing the dormant period or increasing the 24-hour heat of hydration.  • 10% Limestone • 15% Limestone • 20% Limestone  14h  24h  Fig. 3.3 Early age compressive strength for low heat cement with different proportions of limestone filler [38].  3.5.3 Mechanical Properties  There is some agreement that, up to 5% limestone replacement by mass of cement, the compressive strength of concrete is not affected. There are some reports which claim that up to  Chapter III  28  15% of replacement [39, 40] and up to 20% of addition [4, 41, 42], the mechanical performance of cement paste is equivalent to pure OPC and even more favorable for some ages (Fig. 3.4). Livesey [43] reported the strength development characteristics of British cements containing a limestone filler. A 5% limestone addition seemed not to affect either the compressive or flexural strengths, while a 25% addition reduced the strength class of the cement. It was also observed that at a constant strength, increasing the amount of limestone fdler increased the modulus of elasticity [39]. A calcite having a disorganized crystalline network was shown to enhance the mechanical performance in a 75% OPC and 25% limestone cement better than either a well crystallized calcite or a quartz filler of the same fineness [44].  The European practice of dealing with the reduction of the compressive strength at high levels of cement replacement by carbonates, is to increase the fineness of the cement and/or to increase its C A content so that the formation of carboaluminates compensates a bit for the loss of strength. 3  In view of its fluidifying effect [1, 29, 30, 31], its reduction of the transition zone thickness [1], and its refinement of the grain size of the hydration products due to a possible nucleation effect, ultrafine limestone may play a different role in HPC. There is an increasing tendency to use multiply-blended binders having an optimal particle packing for the development of highstrength materials [45, 46]. In these systems, a portion even of the cement plays a filler role due to incomplete hydration.  Fig. 3.4 Effect of limestone partial replacement of cement on compressive strength [40].  Chapter HI  29  3.6 DURABILITY CONCERNS  3.6.1 Carbonation Potential in Limestone Filler Cement  The presence of Ca(OH) is primarily responsible for maintaining the high alkalinity level of 2  concrete. As C0 diffuses into the concrete and reacts with the Ca(OH) to form calcium 2  2  carbonate, the pH may be reduced to a level at which the passive oxide layer on the reinforcing steel becomes unstable. Therefore, oxidation of the steel can proceed. To what extent does the addition of carbonates to cement affect the carbonation depth and carbonation rate of concrete? Ingram and Daugherty [20] stated that: "One of the most interesting side effects of carbonate additions to cement is that atmospheric carbonation damage might be reduced". They proposed  two mechanisms for this possible improvement: The first is that the C0 and C0 react in 3  2  hydrating cement to produce the bicarbonate ion HC0 - In such a scenario, two moles of C0 are 3  2  consumed by one mole of incorporated carbonate before attack of the calcium hydroxide proceeds. The second mechanism is that C0 does attack the Ca(OH) normally, but the system 2  2  stays basic for a longer time owing to the presence of calcium carbonate. Closer examination of the published data regarding this issue reveals that such a statement is very optimistic. Research carried out at the Building Research Establishment [47] at three independent laboratories resulted only in two widely held views: The carbonation depth increases proportionally to the square root of time, and the carbonation rate is inversely proportional to the compressive strength of concrete. There was no evidence that the type of cement had any significant effect on carbonation. This trend applies also to the previous data acquired during the relatively long French experience with the CPJ cements [33]. 3.6.2 Potential of Steel Corrosion in Limestone Filler Cement  There are certain requirements for the corrosion of steel embedded in concrete to proceed. The first of these requirements is inherent to the steel itself and to the presence of microstructural and stress gradients within it. These are beyond the scope of this discussion. The second requirement is the reduction of the alkalinity of the concrete (usually around 12.5 to 13) to a critical value (usually below 11) which causes the passive oxide layer coating the steel rebars to become unstable and eventually be destroyed. The major reason for this reduction is the carbonation reaction discussed above. Based on the carbonation criterion alone, there is no evidence to believe that a drastic change in the corrosion behavior should be expected in limestone filler cements.  Chapter III  30  Nevertheless, the chloride ions have the special ability to destroy the passive oxide film of steel even if high alkalinity conditions prevail in the concrete. The attack of this passive layer can create an electrochemical process where the unprotected part of the steel can act as an anode (oxidation of iron) and the remaining steel can act as a cathode (reduction of water). For such an electrochemical process to be maintained, there is need for an electrolyte, which requires the presence of both oxygen and moisture. The difference in the corrosion potential of embedded steel in concrete made with OPC or limestone filler cement will be directly related to the resistance of the "skin" of the concrete to the mechanisms of mass transfer, that is, the diffusion of chlorides, oxygen and moisture. The mechanism of chloride incorporation in the hep, as a means of delaying the ingress of chlorides, is mainly related to the formation of Friedel salt, which in nam depends on the C A content of the 3  cement. This will not be discussed further here.  The French experience with the CPJ cements [48] has demonstrated that the diffusion of the chloride ions is primarily dependent on the strength grade of the cement, and is not related to the amount of carbonate which is incorporated. Research carried out at the Building Research Establishment [47] has shown no clear trends in the chloride diffusion of cements with or without carbonate additions. Although the 25% filler cement showed higher diffusion rates, these results are comparable to some OPC controls in the study. More recent work [49] showed that the addition of limestone filler reduces the diffusion coefficient of chloride ions. This was attributed to the effect of limestone filler particles on the tortuosity of the system. Relatively lean concrete prisms were also placed in a tidal zone exposure site [47]. The 25% limestone cements appeared to allow for greater corrosion rates than OPC. No difference was observed with respect to the 5% filler cements. Expressing the chloride concentrations relative to the weight of the OPC rather than the total weight of the binder, leads to the 25% filler cement lying on the same curve describing the behavior of the other cements. This indicates that the strength grade of concrete and its porosity, rather than its carbonate content, controls the diffusion of chlorides. The same research program showed that the filler cements tend to have a somewhat lower permeability to oxygen. Their porosity was shown to increase with the filler content, but this effect was very slight [43, 47].  Chapter III  31  3.6.3 Freeze-Thaw Performance It was shown [50] that for a constant water/binder ratio, the critical average spacing of air bubbles decreased with an increase of the carbonate addition to cement. If the comparison was carried out on the basis of the same w/c ratio instead of the w/b ratio, the behavior of the filler cement was similar to the OPC. Livesey [43] reported that the performance of cements in freeze/thaw tests was improved as the filler limestone replacement level was reduced, although the performance of a 25% limestone replacement was comparable to that of fly ash cements. It was also shown that below a replacement threshold of 15%, the freeze thaw resistance of cements was not an issue [33]. Matthews [51] has reported the results of a study on the frost performance of limestone filler cements using the ASTM C-666 Standard Test Method for Resistance of Concrete to Rapid Freezing and Thawing and an adaptation of the BS 5075 (Part 2 Appendix C) Test for Resistance to Freezing and Thawing. It was observed that the freeze-thaw resistance became poorer as the level of limestone addition increased, though some 5% limestone cements gave better results than the poorer control cements.  3.6.4 Sulfate Attack: a New Concern The sulfate resistance of cements made with finely ground carbonate additions has been studied by several investigators. Hooton [36] carried out a study on three pairs of cements produced with limestone additions and having medium and high contents of C A, using both the ASTM C452 3  Test for Potential Expansion of Portland Cement Mortars Exposed to Sulfate and the ASTM C1012 Test for Length Change of Hydraulic Cement Mortars Exposed to Sulfate Solution. Whether the specimens passed or failed for moderate sulfate resistance limits was unaffected by the carbonate additions. The French experience with CPJ cements supports the widely held opinion that limestone additions to cement increase its sulfate resistance, and this is more noticeable in cements rich in C A [33]. Ingram et al. [4], Livesey [43] and Matthews et al. [51] 3  reported the same effect. Only one study [52] was available in which cement pastes containing large proportions of limestone showed decreased resistance to sulfate attack as measured by expansion. The positive effect on the sulfate resistance of limestone filler cement is so well documented that a German patent was actually issued for sulfate resistant portland cement containing finely ground limestone [53].  The chemical reactions involved in the sulfate attack which have been identified are the transformation of portlandite (Ca(OH) ) to gypsum and calcium aluminate hydrate to ettringite 2  (3CaO.Al 0 .3CaS0 .32H 0). To account for these mechanisms, sulfate-resisting portland 2  3  4  2  Chapter III  32  cement (SRPC) having a low level of tricalcium aluminate was formulated. SRPC has been used satisfactorily where sulfate resistance has been specified. In recent years, however, cases of severe sulfate attack have been reported in concrete foundations specifically designed to provide good sulfate resisting properties [54, 55, 56]. A third chemical reaction was identified as responsible for these degradations. Where a wet and cold environment prevails and a constant source of sulfate and carbonate ions is provided, C-S-H reacts to form the mineral thaumasite (CaSi0 .CaC0 .CaS0 .15H 0). In this regard, SRPC is as vulnerable as normal cement. 3  3  4  2  Thaumasite formation is not simply an expansive reaction leading to the cracking of concrete; it affects in a more fundamental way the binding capabilities of the hydrated cement paste. The concrete undergoes a considerable softening, which in structural terms means that the concrete element loses its load bearing capacity. Comparing the conventional expansive reaction due to ettringite formation to the thaumasite form of sulfate attack, Taylor [57] states that: "The quantity of ettringite that can be formed from a cement is limited by the amount of Al 0 2  }  available, but assuming a continuing source of S0 ", thaumasite formation is limited only by the 4  available CaO and Si0 . It can therefore form in large quantities, reducing a mortar into a 2  mush".  A detailed list of the published papers and technical reports regarding thaumasite [58], its occurrence in modern structures [59] and a review of the analytical methods used to investigate its presence [60] are available. The thaumasite form of sulfate attack in concretes containing ground calcium carbonate was reported only for very limited cases and the extent of the problem in practice is not yet clear [56]. However, in most reported cases of thaumasite attack, an available carbonate source was identified. After the abundant literature on the beneficial effect of carbonate additions to cement on its sulfate resistance, the thaumasite reaction might be a surprise . To date, only a few precautions against this form of sulfate attack have been 1  recommended: one for concretes buried in sulfate soils in the UK [62], the other for concreting in cold regions in the Canadian Arctic [55]. Other cold and wet parts of the world with sulfate soils, including the Prairie Provinces of Canada and parts of the western U.S. present a possible scenario for the thaumasite reaction. Since the potential of this reaction with local cements, especially when carbonates are added, is not fully understood, more research is required.  1  In Greek, thaumasite literally means: to be surprised [61].  33  Chapter III  3.7 CONCLUSIONS The controversy regarding the use of carbonate additions to cement and the divergence of the national standards on this issue in recent years is related to the fact that the question has been addressed in a global manner: Should all cements contain carbonate additions, or should they not? A more pragmatic approach to the question within the present context is to look at the specific applications where carbonate additions can offer significant advantages. Owing to its low production cost, ground limestone of specified quality and fineness can be blended with cements and used in several new applications.  Because it complements the grain size distribution of cement, ground limestone can have positive effects on the stability and workability of fresh concrete mixes, and can be useful in the production of self-leveling HPC. In addition, due to its set controlling properties, it may improve the rheology of superplasticized low w/c ratio concrete mixtures, where the equilibrium between the reactivity of C A and the solubility of sulfates is difficult to achieve. This latter effect might 3  however be limited by the acceleration of the C S hydration due to the nucleation mechanism 3  imparted by the fine limestone particles.  Ultrafine limestone can be expected to play at least a part of the filler effect of more conventional microfillers such as silica fume. Although it does not have the optimal spherical shape and fineness that silica fume has, limestone microfiller can probably densify and reduce the thickness of the transition zone via a reduction of the wall effect, and refine the grain size of the hydration products via a possible nucleation mechanism. This may enhance the mechanical properties through a better composite behavior of concrete. Furthermore, limestone microfiller might find use in the new high strength systems where optimal particle packing is essential. Likewise, large amounts of ground limestone used in binary or ternary blends can help in developing high workability, low heat cements for mass concreting applications. These aspects have to be fully investigated for practical recommendations to be formulated. Research should also address the issue of the thaumasite form of sulfate attack in sulfate cold and wet environments when fine carbonates are added to cement.  The significant scatter in the published data regarding the use of ground limestone in concrete can be attributed mainly to the fact that different investigators used different materials. Fillers having different grain size distributions can probably play different roles in concrete because they may have different reactivities and different effects on the water demand, the stability of the concrete mix, the kinetics of the hydration reaction and the particle packing of the system.  Chapter III  34  Furthermore, a porous limestone having harmful impurities such as clay or organic matter can be detrimental to the concrete properties. There is a real need to optimize the limestone filler properties for its various intended uses. For this to be achieved, a better understanding of the filler effect on the rheology, microstructure, and mechanical properties, especially in HPC, has to be established. The following chapters are a contribution in this direction.  3.8 REFERENCES [I]  Escadeillas, G. [1988], "Les ciments aux fillers calcaires: Contribution a leur optimisation par l'etude des proprietes mecaniques et physiques des betons fillerises", Doctoral Thesis in Civil Engineering, Universite Paul-Sabatier, Toulouse, France, 143 p.  [2]  Corish, A.T. [1989], "European cement standards", Proc: Performance of Limestone Filled cements, BRE/BCA, Garston Seminar, pp. 2.1-2.25.  [3]  Mayfield, L.L. [1988], "Limestone additions to cement-An old controversy revisited", Cement Concrete and Aggregates, Vol. 10, No. 1, pp. 3-8.  [4]  Ingram, K.D. and Daugherty, K.E. [1991], "A review of limestone additions to portland cement and concrete", Cement and Concrete Composites, Vol. 13, No. 3, pp. 165-170.  [5]  Bobrowski, G.S., Wilson, J.L. and Daugherty, K.E. [1977], "Limestone substitutes for gypsum as a cement ingredient", Rock Products, Vol. 80, No. 2, pp. 404-410.  [6]  Negro, A., Abbiati, G. and Cussino, L. [1986], "Calcium carbonate substitute for gypsum as set regulator", II Cemento, Vol. 4, pp. 537-544.  [7]  Bensted, J. [1980], "Some hydration investigations involving portland cement- Effect of calcium carbonate substitution of gypsum", World Cement Technology, Vol. 11, No. 8, pp. 395-406.  [8]  Bensted, J. [1983], "Further hydration investigations involving portland cement and the substitution of limestone for gypsum", World Cement Technology, Vol. 14, pp. 383-392.  [9]  Campitelli, V.C. and Florindo, M.C. [1990], "The influence of limestone additions on optimum sulfur trioxide content in portland cements", Carbonate Additions to Cement, ASTM STP 1064, pp. 30-40.  [10] Bessey, G.E. [1938], "Calcium aluminate and calcium silicate hydrate", Proc. Symp. Chem. Cements, Stockholm, Vol. I, pp. 178-215. [II]  Manabe, T., Kawada, N. and Nishiyama, M. [1961], "Studies on 3CaO.Al 0 .CaC0 .nH 0 (Calcium monocarboaluminate hydrate)", CAJ Review, pp. 48-55. 2  3  3  2  [12] Feldman, R.F., Ramachandran, V.S. and Sereda, J.P. [1965], "Influence of CaC0 on the hydration of 3CaO.Al 0 ", Journal of the American Ceramic Society, Vol. 48, No. 1, pp. 2530. 3  3  3  Chapter III  35  [13] Ramachandran, V.S. and Chun-mei, Z. [1986], "Hydration and microstructural development in the 3CaO.AL0 --CaC0 -H 0 system", Materials and Structures, Vol. 19, No. 114, pp. 437-444. 3  [14]  3  3  Vernet, C. and Noworyta. [1992], "Mechanisms of limestone filler reactions in the system {C A-CSH -CH-CC-H}", 9 Int. Cong. On Chem. Cem., New Delhi, Vol. IV, pp. 430-436. Ul  3  2  [15] Ramachandran, V.S. and Chun-mei, Z. [1986], "Dependence of fineness of calcium carbonate on the hydration behavior of tricalcium silicate", Durability of Building Materials, Vol. 4, pp. 45-66. [16]  Husson, S., Gulhot, B. and Pera, J. [1992], "Influence of different fillers on the hydration of C S", 9 Int. Cong. On Chem. Cem., New Delhi, Vol. IV, III-A.013, pp. 83-89. th  3  [17] Barker, A.P. and Cory, H. [1991], "The early hydration of limestone filled cements", Proc: Blended Cements in Construction, Swamy, R.N. ed., University of Sheffield, pp. 107-124. [18]  Ushiyama, H. et al. [1986], "Effect of carbonate on early stage of hydration of alite", 8 Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. II, pp. 154-159. th  [19] Evrard, O. and Chloup, M. [1994], "Reactivite chimique des calcaires en milieu basique: application aux ciments et betons", Annales ITBTP, N. 529, Serie 316, pp. 83-87. [20] Ingram, K.E. [1992], "Limestone additions to cement: uptake, chemistry and effects", 9 Int. Cong, on Chem. Cem, New Delhi, Vol. Ill, pp. 180-186. th  [21] Vernet, C. [1986], "Sequence et cinetique des reactions d'hydratation de l'aluminate tricalcique en presence de gypse, de chaux et de fillers calcaires", 8" Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. Ill, pp. 70-74. 1  [22] Jambor, J. [1980], "Influence of 3CaO.Al 0 .Ca0 .nH 0 on the structure of cement paste", 7 Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. IV, pp. 487-492. 2  3  3  2  th  [23]  Klemm, W.A and Adams, L.D. [1990], "An investigation of the formation of carboaluminates", Carbonate Additions to Cement, ASTM STP 1064, pp. 60-72.  [24] Farran, J. [1956], "Contribution mineralogique a l'etude de l'adherence entre les constituants hydrates des ciments et les materiaux enrobes", Revue des Materiaux et Constructions, No. 490-491, pp. 155-172 and 191-209. [25] Buck, A.D. and Dolch, W.L. [1966], "Investigation of a reaction involving nondolomitic limestone aggregate in concrete", ACI Journal, Vol. 63, No. 7, July, 1966, pp. 755-763. [26] Ai'tcin, P.C. and Mehta, P.K. [1990], "Effect of coarse-aggregate characteristics on mechanical properties of high-strength concrete", ACI Materials Journal, Vol. 87, No. 2, pp. 103-107. [27]  Grandet, J. and Ollivier, J.P. [1980], "Etude de la formation du monocarboaluminate de calcium hydrate au contact d'un granulat calcaire dans une pate de ciment portland", Cement and Concrete Research, Vol. 10, pp. 759-770.  Chapter III  36  [28] Monteiro, P.J.M. and Mehta, P.K. [1986], "Reaction between carbonate rock and cement paste", Cement and Concrete Research, Vol. 16, pp. 127-134. [29] Bombled, J.P. [1986], "Rheologie du beton frais: Influence de l'ajout de fillers aux ciments", 8* ICCC, Rio de Janeiro, Vol. IV, pp. 190-196. [30] Sprung, S. and Siebel, E. [1991], "Assessment of the suitability of limestone for producing portland limestone cement (PKZ)", Zement, Kalk and Gibs, N. 1/1991, pp. 1-11. [31] Neto, S.N and Campitelli, V.C. [1990], "The influence of limestone additions on the rheological properties and water retention value of portland cement slurries", ASTM STP 1064, pp. 24-29. [32] Brookbanks, P. [1989], "Properties of fresh concrete", Performance of Limestone Filled Cements", Proc. BRE/BCA seminar, Garston, pp. 4.1-4.15. [33] Cochet, G. and Sorrentino, F. [1993], "Limestone filled cements: properties and uses", Mineral Admixtures in Cement and Concrete, Ghosh, S.N., Sarkar, S.L. and Harsh, S., eds, ABI Book Pvt. Ltd., New Delhi, India, Vol. 4, pp. 266-295. [34] Adams, L.D. and Race, R.M., "Effect of limestone additions upon drying shrinkage of portland cement mortar", Carbonate additions to cement, ASTM STP 1064, pp. 41-50. [35] Barker, A.P. and Mattews, J.D. [1989], "Heat release characteristics of limestone filled cements", Performance of Limestone Filled Cements, Proc. Building Research Establishment Seminar, Garston, pp 5.1-5.29. [36] Hooton, R.D. [1990], "Effects of carbonate additions on heat of hydration and sulfate resistance", Carbonate Additions to Cement, ASTM STP 1064, pp. 73-81. [37] Furuya, N., Saito, T., Tikamatsu, R. and Sogo, S. [1994], "Basic research on highly workable concrete with low heat type cement and large amount of limestone powder", Concrete Library of JSCE, No. 23, pp. 81-101. [38] Kessal, M. [1995], "Developpement d'un systeme cimentaire a haute resistance initiale a base de ciment de Type 20M", M.A.Sc thesis, Universite de Sherbrooke, 117 p. [39] Gegout, P. et al. [1986], "Texture et performance des ciments fillerises", 8 Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. I., pp. 197-203. th  [40] CRIC. [1987], "Les ciments portland au filler', In chapter III, Centre National de Recherches Scientifiques et Techniques pour lTndustrie Cimentaire, Report RA-f-1986/87, Belgium, pp. 8-28. [41] Soroka, I. and Setter, N. [1977], "The effect of fillers on strength of cement mortars", Cement and Concrete Research, Vol. 7, No. 4, pp. 449-456. [42] Soroka, I. and Stern, N. [1976], "Calcareous fillers and the compressive strength of portland cement", Cement and Concrete Research, Vol. 6, No. 3, pp. 367-376. [43] Livesey, P. [1991], "Performance of limestone filled cements', in: Blended Cements in Construction, R.N. Swamy, ed., Elsevier, pp. 1-15.  Chapter III  37  [44] Regourd, M. [1986], "Ciments speciaux et ciments avec addition: caracteristiques et activation des produits d'addition", 8 Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. I, pp 119-229. th  [45] Richard, P. and Cheyrezy, M.H. [1994], "Reactive powder concrete with high ductility and 200-800 MPa compressive strength", Concrete Technology: Past, Present, and Future, ACI SP-144, pp. 507-518. [46] Fidjestol, P. and Frearson, J. [1994], "High performance concrete using blended and triple blended binders", ACI SP-149, Singapore, V. M. Malhotra, ed., pp. 135-157. [47] Moir, G.K. and Kelham, S. [1989], "Durability", in: Performance of limestone filled cements", in [35], pp 7.1-7.67. [48] Cochet, G. and Jesus, B. [1991], "Diffusion of chloride ions in portland cement-filler mortars", Int. Conf. on Blended Cements in Construction, R. N. Swamy ed., University of Sheffield, pp. 365-376. [49] Hornain, H., Marchand, J., Duhot, V. and Moranville-Regourd, M. [1995], "Diffusion of chloride ions in limestone filler blended cement pastes and mortars", Cement and Concrete Research, Vol. 25, No. 8, pp. 1667-1678. [50] Gegout, P., Hornain, H. Thuret, B. and Regourd, M. [1986], "Resistance au gel des ciments aux fillers calcaires", 8 ICCC, Rio de Janeiro, Vol. VI, pp. 47-52. th  [51] Mattews, J.D. [1989], "Sulfate and freeze thaw resistance", Proc: BRE/BCA, Garston, pp. 8.1-8.16. [52] Marsh, B.K. and Joshi, R.C. [1986], "Sulfate and acid resistance of cement paste containing pulverized limestone and fly ash", Durability of Building Materials, No. 4, pp. 67-80. [53] Portland-Zementwerke Heiderberg, A.G. [1972], "Increasing sulfate resistance of Portland cement", German Patent 1646910, May, 1972, 3 pp. [54] Crammond, N.J. and Nixon, P.J. [1993], "Deterioration of concrete foundation piles as a result of thaumasite formation", 6 Int. Conf. on Durability of Building Materials, Japan, pp. 295-305. th  [55]  Bickley, J.A. et al. [1994], "Thaumasite related deterioration of concrete structures", Proc. Concrete Technology: Past, Present and Future, ACI SP: 144-8, pp. 159-175.  [56] Crammond, N.J. and Halliwell, M.A. [1995], "The thaumasite form of sulfate attack in concretes containing a source of carbonate ions: a microstructural overview", Advances in Concrete Technology, 2 CANMET-ACI Symposium,V.M. Malhotra, ed. nd  [57] Taylor, H.F.W. [1990], Cement Chemistry, Academic press, pp. 401-402. [58] Crammond, N.J. [1991], "Thaumasite: a detailed list of published papers and technical reports", BRE Internal Note N. 148/91,46 p. [59] Crammond, N.J. [1991], "The occurrence of thaumasite in modern constructions", BRE Internal Note N. 147/91, 8 p.  Chapter III  38  [60] Hjorth, L. [1991], "Thaumasite: review of analytical methods", EUREKA, Project EU-672, 16 p. [61] Berra, M. and Baronio, G. [1987], "Thaumasite in deteriorated concretes in the presence of sulfates", Concrete Durability, ACI SP-100, Vol. II, pp. 2073-2089. [62] BRE. [1991], "Sulfate and acid resistance of concrete in the ground", BRE. Digest 363.  Chapter IV  39  Chapter IV MICROFILLER E F F E C T ON R H E O L O G Y OF C E M E N T PASTES  4.1 R H E O L O G Y OF DENSE SUSPENSIONS  4.1.1 Introduction  The use of mineral admixtures and fine powders in concrete is likely to increase in the future. For instance, the recent ENV 197 European Standard specifies 5 main types and 25 varieties of cements, among which only portland cement is not made of two or more components [1]. These specified composite cements often have major economic and ecological advantages, without compromising the fundamental properties of the cement. The clinker is partially substituted for by unburned materials, and fossil fuels can therefore be saved. In addition, cement raw materials can be conserved when substituted for by recycled industrial byproducts. The environmentally hazardous emissions can also be reduced by as much as 12% for portland-limestone cement, and up to 55% for a 60% blast furnace slag cement [2].  With the expected worldwide implementation of such composite cements and their potential use in low w/c ratio superplasticized concrete, the question arises whether they can also provide economic and performance advantages from the rheological point of view. Despite their potential, there appear to be few published studies on the effect of mineral admixtures on the rheological behavior of cement paste and concrete. This is especially true for HPC. Some interesting work has been reported [3], but the effect of inert fillers and triple blended cements was not fully addressed. In this chapter, a statistical model of the microfiller effect on the rheology of superplasticized cement pastes in silica fume and non-silica fume systems is presented. The cohesiveness, plastic viscosity, stability, and superplasticizer efficiency are described for various composite-binder cement pastes. Without going into too much detail, pertinent information on the rheology of cement paste is first provided.  Chapter IV  40  4.1.2 Rheology  "Rheology" originates from the Greek word reo, meaning flow. It is the study of deformation and flow of matter, embracing elasticity, viscosity, and plasticity. The main concern of rheologists is to establish relationships between stress, deformation, and the deformation rate of fluids and suspensions. Fresh concrete falls into the category of dense suspensions, and its plasticity is only a transient state. However, the effective methods of handling, transporting, pumping, placing, finishing, and consolidating concrete depend on its flow properties. Traditional empirical methods have been used to characterize these technological aspects, but the increasing use of chemical and mineral admixtures and the production of HPC have made these techniques inefficient and unreliable. Thus, rheology has been introduced into concrete technology to provide a more fundamental approach to the characterization and quality control of fresh concrete.  4.1.3 Suspensions of cement particles in water  Without going into the details of intrinsic material charges, ionization, specific ion adsorption, electron double layer and zeta potential, and other concepts, we can simply say that cement grains tend to flocculate when mixed with water. Whether this is due to London-van der Waals forces or hetero-coagulation is not clear. Attractive van der Waals forces decrease as the cube of the distance between grains, and become appreciable only for particles smaller than 1 urn, which is the range of colloids. Cement can only be considered as an upper limit of a colloidal system [4] because it contains particles considerably larger than 1 um. Cement contains about 3% by weight of particles smaller than 2 urn and these represent about 20% of the number of total particles. Thus, a significant number of cement grains flocculate through van der Waals forces. On the other hand, hetero-coagulation results from electrostatic attraction of particles of different surface potential, and is not strongly dependent on particle size. Cement is a multiphase mineral, and compositional differences would typically cause electrostatic attraction between particles. Powers [5] describes cement paste as flocculent when concentrated, with particles forming a 3 dimensional network, as opposed to flocculated when dilute, with particles forming discrete clusters.  Chapter IV  41  The concentration of a cement suspension can be expressed as a proportion by volume by converting the w/c ratio to a volume proportion using Eq. 4.1; where (j) is the proportion of cement by volume, p is the density of water and p is the density of cement. w  c  1 — —  0> =  (4.1)  p (wlc) c  P  w  For a w/c ratio of 0.80 the volume proportion of cement is around 0.28, and for a 0.22 w/c ratio it is around 0.59. The maximum possible solid concentration for monosized particles is around 0.64, but the continuous gradation of cement particles can achieve higher values because the smaller particles can fit in between the larger ones. Considerable progress has been made to better understand the flow behavior of concentrated suspensions. This was achieved through model systems using dispersed suspensions. A quantitative relationship [6] between viscosity and particle packing has been developed (eq. 4.2):  JL  f  ^ \-W>» 1  (4.2)  where r\ is the suspension viscosity, r\ is the liquid phase viscosity, [r\] is the intrinsic viscosity 0  of the suspension, § is the concentration (volume fraction), and <t> the maximum concentration M  possible for the particular system. This relationship is routinely applied to colloid suspensions, and has been shown to be valid both for model systems and broad particle size and shape distributions [4]. However, its validity for cement paste has not been demonstrated.  4.1.4 Modeling the Rheological Behavior of Cement Paste A great deal of research has been aimed at developing models for the rheology of cement paste which can have predictive capabilities. This would permit one to specify certain technological characteristics for the flow behavior of cement based materials and to obtain them with a reasonable accuracy. Reviews of these models have been made by Malek and Roy [7] and by Papo [8]. A summary of various models is presented in Table 4.1.  Chapter IV  42  Table 4.1- Various Models for the Rheology of Cement Paste Model  Main Equation  1  Parameters  Features  Packine Models Mooney TJ  r| = viscosity, (j) = packing volume fraction, a and k = constants = packing volume fraction, x, = volume fraction of component.  = e^  Lee  j=n  7=1  De Larrard etal.  Pi = i a  Ball & Richmond  (~ i  ) ? j s(hM  +  Hi, ;)<D ,  +  l  a  I  j  Calculates shear viscosity from estimate of the particle packing.  Calculates packing volume fraction of idealized binary mixture of spheres as function of diameter ratio of composition. Mooney's Pi = packing density, <(>; = Extends constant, f and g = crowding concept to calculate the factors, <j>j = fractional solid packing density by including the particle volume of component j. distribution and packing density of various monosize classes of grains. n = viscosity, <|> Accounts for the effect maximum packing fraction of phase volume on viscosity. m a x  1-  Particle-Particle Interaction Models Elastic Floe Model Friction and Viscosity Model  n j-r  f  I2d,  5  r  1 =  B  r| = viscosity, n = number of bonds, y = shear rate, A and B = constant, d, = distance of max. attraction between particles, E, = zeta potential. 2/3 t| = viscosity, B = friction, n = number of particles, x = value proportional to repulsive l) potential, G = constant proportional to shear rate, t = time, p = constant.  2/3 3 " 3  x„ 2 e Gt  [  1 -  ( , P  +  /)( , G  2 +  Expresses the repulsive forces in terms of the surface potential,  Expresses repulsive forces in terms of Deybie radius,  Time Independent Models x = shear rate, T = yield stress, y = shear rate, r| = viscosity, A,B,C,k,k ,k ,a,p,y = constants Linear r = rj Newtonian 0  Bingham Hershel & Buckly  1  2  T  =  n>l Dilatant n<l Viscoplastic  V +ky  n  0  7 =  Robertson and Stiff  1  = A(  A{T-C)  B  R  +  c)  B  B>1 Dilatant B<1 Viscoplastic  References for the various models can be found in [7, 8].  (cont'd)  43  Chapter IV  Table 4.1 (cont'd)  Vom Berg  r = T +B.s'mh~ (y / C) i  Viscoplastic  0  Eying  r =  Ellis  Z A sinh  Pseudoplastic  -1  Pseudoplastic  r = K,T + K T  n  2  Casson  T  1 / 2  =  T  i ,2 0  +  k y  n  Viscoplastic  Time Dependent Models (Thixotropy) Atzeni al.  et  T  T = shear stress, x = equilibrium shear stress (at t = ° C ) , T = initial shear stress (at t = 0), t = time, B and T = constants, a = structural factor. e  M  +  a  Despite this considerable effort, the results achieved so far do not agree, either qualitatively or quantitatively. Certainly, using different mixing techniques and rheological instruments has something to do with this. However, results are sometimes difficult to reproduce even for the same investigator using the same equipment and methods. For example, for normal consistency cement pastes, some researchers reported yield stress values of 50-200 Pa, while others reported values around 2000 Pa [4]. The flow behavior of cement paste is not only shear rate and shear history dependent, but also time dependent. In view of this inherent complexity and the scatter of results mentioned above, a statistical approach to the modeling of the microfiller effect on the rheology of superplasticized cement pastes was selected in this work. Likewise, experimental conditions, shear history, timing, and other experimental details were maintained constant during this study.  4.2 MATERIALS, APPARATUS AND PROCEDURE  4.2.1 Materials  Ordinary ASTM Type I cement (OPC), high purity limestone microfiller (LF) having a 3 pm mean particle size, and silica fume (SF) were used. Their chemical and physical properties are summarized in Table 4.2. A naphthalene sulfonate superplasticizer and tap water were employed for the mixing. The different mortars were made with standard Ottawa sand (ASTM C778-91 Standard Specification for Standard Sand).  Chapter IV  44  Table 4.2- Chemical and Physical Properties of the Cement and Microfillers Used  Chemical Composition [%] Componenet  OPC  SF  LF  Si0  21.5  93.6  0.25  3  4.6  3  Physical Tests OPC Initial set time [min]  125  0.3  Final set time [min]  230  3.2  0.5  Autoclave expansion [%]  0.04  CaO (total)  63.2  0.3  Air content of mortar [%]  7  CaO (free)  0.6  S0  2.7  Compressive strength at 3d [MPa] 3d 7d 28d Passing 45 nm sieve [%]  24.6 30.0 38.5 87.7 346  2  A1 0 2  Fe 0 2  3  SF  LF  100  100  MgO  3.0  0.5  Specific surface Blaine [m /kg]  Na 0 + K 0  0.84  1.4  Specific surface BET [m/kg]  17500  2300  Loss on ignition  0.7  2.8  Particle size range (sedigraph) [nm]  0.04-0.28  0.2-0.12  Insoluble residue  0.2  0.18  3  2  2  Carbon CaC0  2  0.35 1.9  3  Mean particle size [|im]  14  Residue on 325 mesh [%] 98.2  3  MgC0  2  0.005  Specific gravity  3.16  CA C AF CS CS  7 10 51 23  2.23  2.71  1.2 3  Potential composition of cement [%]  4  3 2  4.2.2 Apparatus and Procedure The mixing procedure was identical for all the cement pastes. First, the superplasticizer was added to the mixing water which was at a constant temperature of 17 ± 1 °C. Then the binder was added over a period of 1 min while mixing the grout at a constant speed of 3500 rpm for 3 min, using a helical mixer (Fig. 4.1). A rest-period of 1 min followed during which the inner sides of the mixer pan were scraped down. Finally, the grout was mixed at the same velocity for 1 additional min, and its temperature was measured at the end of the mixing.  The flow time was measured just after the mixing using a modified Marsh cone with a 5 mm outlet (Fig. 4.2). A volume of 1.1 L of grout was placed in the cone while locking the outlet with the index. A chronometer was switched on while removing the index and the time was measured for each 100 ml flow of grout up to 1 L.  Chapter IV  45  For each mix, the percentage of superplasticizer was systematically varied to obtain the saturation dosage; that is the superplasticizer dosage beyond which no improvements in the fluidity of the grout were observed (Fig. 4.3). For the superplasticizer dosage which achieved a flow time of 110 ± 10 sec at 5 min, the grout was further characterized using a mini slump test, a rotational viscometer and an induced bleeding test. A flow time of 110 sec was selected because shorter flow times for the lower w/b ratio cement pastes were not achievable even at the HRWR (high-range water reducer) saturation dosage.  The yield stress and plastic viscosity were measured from the flow curve established using a coaxial-cylinders rotational viscometer (smooth cylinders, no serration). The test grout was contained within the annular space between the two cylinders. The outer cylinder (rotor, radius = 1.8415 cm) was rotated at 12 controlled speeds varying from 1 to 600 rpm, and the viscous drag exerted by the grout created a torque on the inner cylinder (bob, radius = 1.7245 cm, height = 3.80 cm). The torque was transmitted to a precision torsion spring whose deflection was measured and related to shear stress using the following formulae (Fig. 4.4):  Chapter IV  46  Fig. 4.3  Determination of the superplasticizer saturation dosage.  47  Chapter IV  Shear stress = k^Q  (4.3)  Shear rate = k N  (4.4)  3  where: 0 : viscometer reading, ki : torsion constant of spring per unit deflection [N-cm/degree], k : shear stress constant for the effective bob surface [cm ], k : shear rate constant [sec.i/rpm]. 3  2  3  The yield stress and plastic viscosity were obtained from linear regression of the shear stress versus shear rate data thus obtained. The slump of the cement pastes was measured using a miniature slump test as described in detail by Kantro [9]. The induced bleeding test consisted of pouring a representative sample of 200 mL of grout into an airtight vessel equipped with a Baroid type filter and a paper filter. After closing the cell, a graduated cylinder was placed under the outlet of the cell. A pressurization system was then connected to the cell to apply a constant nitrogen gas pressure of 0.55 MPa. The volume of induced bleeding was recorded at 15 and 30 sec, and then at every minute up to 10 min.  Chapter IV  48  A mortar corresponding to each grout was made using standard Ottawa sand. The mixing procedure for the mortars followed that of ASTM C305-94 (Standard Practice for Mechanical Mixing of Hydraulic Cement Pastes and Mortars of Plastic Consistency). A proportion of sand of 2.3 times the mass of the binder was generally found suitable to achieve a flow of 100 ± 10% on a flow table for the designed mortars. The flow measure of the mortars was carried out using ASTM C230-90 (Standard Specification for Flow Table for Use in Tests of Hydraulic Cement). The presented results were calculated as: [(d-d )/do x 100] where d is the initial diameter and d 0  0  is the diameter after spread.  4.3 E X P E R I M E N T A L  PLAN  The experiments were designed according to a 2-level uniform-precision factorial plan [10] as shown in Fig. 4.5. This approach was selected to limit the number of cement pastes to be investigated, while first and second order models could be used to fit the data. In addition, this method highlights the significance of the effect of the experimental variables and their interactions, and has a predictive capability for the response of other experimental points located within the experimental domain.  The experimental plan thus consisted of a 2-level factorial plan corresponding to mixes 1, 2, 3, and 4 in Fig. 4.5 (coded ± 1), augmented by 4 axial points corresponding to mixes 5, 6, 7 and 8 (coded ± a). The first set of experiments was used to fit a first-order model, whilst the second set allowed for a second-order model. Since a 2-level factorial plan does not allow for an estimate of the experimental error unless some runs are repeated, it is a common practice to augment the design with observations at the center of the experimental domain. Five central points are required for this uniform-precision plan, which affords more protection against bias in the regression coefficients as compared to the most widely used design to fit a second-order model: the centered-composite plan [10]. The sequence in which the experimental points were investigated was randomized to avoid any statistical significance of a blocking effect.  The quadratic model for the selected factorial plan is shown in equation 4.5:  Chapter IV  49  i1  (0, + a ) 1)  •a,  0)  •?•  1  I  (+a,  0)  ^+1, -1)  (-1.-1) (0. • a )  -I  -> LF [%]  Fig. 4.5 Illustration of the uniform-precision experimental plan. Y = a + ax 0  l  +a x  i  2  2  +a x x l2  l  2  +a x n  2 x  +a x 22  2 2  (4.5)  where: X i and x : experimental variables. Y: response of the experiment. &i, aij: coefficients of the model. 2  The two experimental variables were the water/binder ratio (w/b) and the proportion of limestone replacement by volume of cement (%LF). All other controllable parameters were kept constant. The same experimental plan was carried out on an OPC and an OPC with 1 0 % silica fume replacement of cement. The various responses of the designed experiments were the temperature of the cement paste just after the mixing, the flow time, the superplasticizer dosage to achieve a flow time of 110 ± 10 sec, the saturation dosage of superplasticizer, the mini-slump, the yield stress and plastic viscosity, the induced bleeding, and the mortar flow, for all the designed cement pastes.  The data corresponding to the various responses which resulted from the designed experimental program were analyzed and plotted using a statistics software package [11]. The rheological characteristics of the OPC + LF cement pastes are given in Fig. 4.6; those of the OPC + 1 0 % SF + LF are shown in Fig. 4. 7.  Chapter IV  50  (a) H R W R saturation dosage [%]  (b) H R W R required [%]  0.30  (c) M i n i slump [mm]  0.0  (d) Y i e l d stress [Pa]  Fig. 4.6 Rheological responses for O P C + L F mixtures. (cont'd)  Chapter IV  51  (e) Plastic viscosity [Pa.s]  (f) Induced bleeding [mL]  Fig. 4.6 (cont'd)  Chapter IV  52  (c) M i n i slump [mm]  (d) Y i e l d stress [Pa]  F i g 4.7 Rheological responses for O P C + L F + 1 0 % S F mixtures.  (cont'd)  Chapter IV  53  Fig. 4.7 (cont'd)  Chapter IV  54  4.4 MICROFILLER E F F E C T ON SUPERPLASTICIZER EFFICIENCY  The superplasticizer saturation dosage increased significantly when the w/b ratio decreased, and tended to decrease with increased levels of L F replacement of cement (Fig. 4.6-a). In the presence of 10% SF however, the superplasticizer saturation dosage decreased up to about 10 to 15% LF replacement of cement, and tended to increase slightly beyond this threshold value, probably due to an excess of microfines (Fig. 4.7-a).  The superplasticizer required to achieve a flow time of 110 ± 10 sec in the modified Marsh cone behaved approximately in the same manner as the superplasticizer saturation dosage only in the absence of SF (Fig. 4.6-b). The superplasticizer required increased significantly when the w/b ratio was reduced and tended to decrease with increasing levels of LF replacement of cement. In the presence of 10% SF however, the superplasticizer required increased when the w/b ratio decreased, but had only a slight tendency to decrease beyond 10 to 15% L F replacement of cement (Fig. 4.7-b).  The above results suggest that with increased LF replacement of cement, the superplasticizer efficiency is equal to or higher than that in the case of pure cement. Given the increased surface area of the binder due to the LF, one might expect the LF to inhibit the positive action of the superplasticizer, or at least reduce it by stopping and adsorbing its molecules. This did not seem to be the case for the above results, nor was it in work by Bombled [12], who observed no significant difference between an OPC and an OPC + LF made from the same clinker, though the increased surface area in the latter case was due to inter-grinding the LF which has a higher grindability with the clinker.  Pierre et al. [13] have determined the adsorption isotherms of naphthalene formaldehyde condensate (NFC) in CaC03 dispersions versus the pH, the ionic strength and the Ca** concentration. They observed an increase in the NFC adsorption when the pH decreased and when the Ca** concentration increased. In particular, the addition of Ca""" was efficient in reducing the yield stress and the plastic viscosity of CaC03 suspensions in the highly alkaline media. The conclusion was that in their model system, any factor that enhances the NFC adsorption leads to its increased efficiency. It seems that in the much more complex cementitious system, a LF partial replacement of cement caused a more efficient adsorption of the  Chapter IV  55  superplasticizer. This was probably enhanced by a possible slight increase of the Ca  ++  concentration due to the slow CaC03 solubility. This increase in the superplasticizer efficiency was significant enough to offset the negative contribution due to the increase in the surface area.  Another possible explanation would be that the increased superplasticizer efficiency was simply due to a reduction in the water demand of the mix. The fine CaC03 particles filled in between the relatively coarser cement particles, improving the particle packing of the system, and reducing the pore space which otherwise would have been filled with water. Although Brookbanks [14] suggested that this water reducing effect was not discernible at 5% LF replacement of cement and was only marginal at 28% replacement, Cochet et al. [15] suggested that when the filler is finely ground, its water reducing effect is higher especially for w/c < 0.4.  4.5 COHESION AND PLASTIC VISCOSITY Whether in the absence of SF (Fig. 4.6-d) or in its presence (Fig. 4.7-d), the yield stress for all the cement pastes decreased with a decreased w/b. It should be remembered that all the cement pastes under investigation were designed to have approximately the same 1 L flow time in the modified Marsh cone. The cement pastes having lower w/b ratios required higher superplasticizer dosages, and therefore their behavior was closer to that of a Newtonian fluid, while the higher w/b ratios cement pastes required less superplasticizer and underwent substantial shear thinning before approaching a Bingham behavior (Fig. 4.8).  The yield stress seemed to increase as the LF replacement of cement increased (Fig. 4.6-d). This result was reflected in the mini slump behavior (Fig. 4.6-c), a test also carried out at low shear rates, only at the low w/b ratios. The pat diameter decreased with an increase of the LF, which confirms an increase in the cohesion of the cement paste. In the presence of 10% SF however, the yield stress seemed to decrease with increased LF replacement up to about 10 to 15%, then tended to increase significantly beyond this threshold value (Fig. 4.7-d). Again, this is consistent with the results of the mini slump test only for the lower w/b ratio cement pastes (Fig. 4.7-c) where the mini slump decreased significantly beyond 10 to 15% LF replacement of cement. At higher w/b ratio however, an increase in the mini slump was observed with increased %LF.  Chapter IV  56  Fig 4.8 Shear thinning behavior of low superplasticizer dosage cement pastes.  Neto et al. [16], using w/c ratios of 0.65 and 0.70, suggested that there was a tendency of yield stress reduction with increased LF levels in cement. On the other hand, Bombled [12] showed that at least below LF levels of 10 to 15%, there was generally little difference with regard to the yield stress. It should be noted that for both these results, limestone was interground with the clinker, while in the present case a much finer LF was blended with the cement.  The plastic viscosity of the cement pastes behaved in a similar fashion, whether in the absence (Fig. 4.6-e) or presence of SF (Fig. 4.7-e). It increased with a reduction of the w/b ratio and tended to decrease as the LF replacement of cement increased up to about 10%. Beyond this value, the %LF had limited effect on the plastic viscosity. This latter behavior requires a special look at the relationship between the microstructure of the cement paste and its flow behavior.  From the standpoint of the theory of particle packing, the partial replacement of cement by the finer LF would increase the density of the powder. Krieger and Dougherty [6] quantitatively related the viscosity and the particle packing for dispersed particles as shown previously in Eq. 4.2. This relationship accounts for various effects related to the concentration, the particle size distribution and particle shape, and successfully describes the substantial increase in the viscosity when <j) approaches § M  Chapter IV  57  Another model relating viscosity and particle packing (reported in [4]) is based on the thickness of the water film separating adjacent particles in the cement paste. It predicts an inverse relationship between viscosity and water film thickness. The latter may be estimated by the hydraulic radius, i.e. the volume of water divided by the specific surface area. It is interesting to note that, contrary to the predictions of both of these models, the plastic viscosity seemed not to increase either with increased packing density or increased surface area. This suggests that more complex chemical and physical parameters intervene besides those accounted for by the models.  0.00  0.40  Fig. 4.9 Isoresponse curves for the temperature after mixing of the various cement pastes [°C].  Fig. 4.9 illustrates the isoresponse curves over the experimental domain for the temperature of the cement pastes just after the mixing. The temperature generally decreased with increasing w/b ratio and with increasing %LF. Although this can be attributed to a reduction in the friction due to a possible lubricating effect of the LF particles, it may also be considered that the presence of LF brought about a better control of the very early chemical reactions between cement and water. For instance, Feldman et al. [17] have concluded that the hydration of C3A was suppressed by a C a C 0 3 addition due to the formation of calcium carboaluminates on the C A grains. Several 3  attempts to substitute limestone for gypsum as a set regulator have also been reported [18]. Although a decreased temperature implies a slight increase in the viscosity of the. liquid phase, it also implies a slight decrease in the chemical activity.  Chapter IV  58  4.6 STABILITY OF C E M E N T PASTES  Fig. 4.6-f shows that the 10 min induced bleeding for the OPC + LF cement pastes was mainly controlled by the w/b ratio. Increased levels of LF replacement of cement reduced the induced bleeding only at high w/b ratios, and did not have a significant effect at lower w/b ratios. In the presence of 10% SF (Fig. 4.7-f), the induced bleeding depended mainly on the w/b ratio and did not seem to be affected by the LF either at high or low w/b ratio. A rather surprising tendency for the induced bleeding to slightly increase was observed at high levels of L F replacement of cement.  The induced bleeding test characterizes the ability of water to be transferred through the fresh cement paste under the effect of a pressure gradient. Therefore, it can be better understood when expressed using Darcy's law:  a-  - *  4  §  (4.6)  where Q is the volumetric flow [m /sec], p. is the viscosity of the fluid, A is the apparent area 3  [m ], and cP I di is the pressure gradient [N/m ]. The reduction of the permeability of the 2  2  system due to a reduction of the w/b ratio and the addition of very fine SF particles seemed to be the dominant factor. The addition of LF seemed to be of limited effect except for the non SF mixtures at high w/b, where the addition of fine L F particles is expected to reduce the permeability of the system and thus its induced bleeding.  4.7 CEMENT PASTE FLOW VERSUS MORTAR F L O W  The response curves for the 0.7 L flow time of the cement pastes at 5 min and the flow of mortars made with the same cement pastes are shown in Fig. 4.6-g and 4.6-h, and Fig 4.7-g and 4.7-h for the OPC + LF and OPC + 10% SF + LF systems, respectively. The 0.7 L flow was selected as a basis for comparison rather than the 1 L flow because of the nonlinear nature of the flow which appears beyond 0.7 L, due to an increased effect of friction, and also to the thixotropic character of the cement pastes (Fig. 4.10). It is clear that while the response curves for the cement paste flow exhibited a more horizontal behavior, implying that the w/b ratio was the dominant parameter, the response curves for the mortar flow exhibited a more vertical behavior, indicating  Chapter IV  59  that the LF was a more significant parameter. In the absence of SF, the mortar flow seemed to increase as the LF level increased. In the presence of SF however, the mortar flow was optimal at about 7 to 18% LF, then tended to decrease, probably due to an excess of microfines.  This observed difference in the paste and mortar flows raises the questions of how reliable are the rheological results obtained on cement paste when transposed to mortar and concrete, and how well can we predict the rheological behavior of concrete based on the fundamental rheological characteristics of cement paste. Tattersall and Banfill [19] have already pointed out that the effects of variations in cement paste are diluted by the aggregates and may not show up in concrete.  Volume of cement paste [mL]  Fig. 4.10 Non linearity of the Marsh cone flow time (w/b = 0.40, %LF = 12.5%). In this respect, a model describing the behavior of concrete or mortar based on the rheological characteristics of the cement paste should also include the effect on the cohesion of the Coulomb friction and the interlocking of the aggregates. This can be expressed by associating the Coulomb law and the Bingham behavior after the fashion of Bombled [12]:  F = r + K(p)tan <f>+ juG  (4.6)  where x is the cohesion of the paste, K(u) tan(<()) accounts for the internal friction, and uG accounts for the viscosity of the interstitial paste. It seemed in the results discussed earlier that  Chapter IV  60  the fine LF particles played a significant role in reducing the internal friction in the mortars, which increased their flow. This aspect will be discussed further for HPC in Chapter 5.  4.8 CORRELATIONS AMONGST VARIOUS RHEOLOGICAL PARAMETERS The flow behavior of cement paste is complex; it is not only shear dependent, but also time dependent. Therefore, the flow of cement paste is very sensitive to its shear history. The above experiments were carried out with extreme care in order to keep the shear history, the experimental procedures and their timing as constant as possible. Hence, it is interesting to assess the various possible correlations amongst the different tests carried out.  Fig. 4.11-a illustrates the correlation between two rheological values which are characteristic of the cement pastes at low shear rates: the mini slump and the yield stress. The correlation does not seem to follow a universal model for different cement pastes. The OPC + LF system seemed to follow a linear model while the OPC + 10% SF + LF followed a second order polynomial model. Fig. 4.11-b represents the correlation between two rheological values which are characteristic of the cement paste at higher shear rates: the plastic viscosity and the 0.7 L Marsh cone flow time. The relationship seemed to be unique for all the cement pastes examined, yet the scatter of the data was significant. Figs. 4.12-a and 4.12-b illustrate correlations amongst rheological parameters of the cement pastes at low and high shear rates: the 0.7 L flow versus the yield stress and the mini slump versus the plastic viscosity, respectively. It is shown that no clear trend was observed. The practical implication of these observations is that the cement paste should be Theologically characterized at shear rates which correspond to the particular intended application: injection, pumping, concrete workability, etc.  The correlations above are often affected by a significant variability, though the shear history, the experimental procedures and their timing were kept constant. Struble [4] has noted that the flow behavior of cement paste is highly variable, with various workers reporting quite different results and even individual studies showing different behavior on nominally the same material. Since such variability has also been reported for other materials, there is a question whether this is an inherent feature of dense suspensions.  Chapter IV  61  45  Mini slump [mm]  Fig. 4.11 Correlations amongst rheological parameters at comparable shear rates.  Chapter IV  62  45 40  • OPC+LF HOPC+10% SF+LF  35 + «  •  30 +  CL  8 o  25 +  !  20  2:  >  15  10 5 + 0  —h 50  55  60  65 0.7 L  70  flow  75  80  85  t i m e [sec]  0.16 0.14 0.12 0.1 0.08 + 0.06 0.04 • OPC+LF 0.02  » O P C + 1 0 % SF+LF  0  — I  90  100  110  1  120  130  140  M i n i s l u m p [mm]  Fig. 4.12 Correlations amongst rheological parameters at different shear rates.  Chapter IV  Fig. 4.13 Induced bleeding versus (a) plastic viscosity; and (b) superplasticizer saturation dosage.  Chapter IV  64  to the plastic viscosity since the high viscosity systems were the low w/b mixtures which required higher superplasticizer dosages. Yet, the OPC + LF and the OPC + 10% SF + LF data did not seem to follow the same pattern. The induced bleeding was also inversely proportional to the superplasticizer dosage and all cement pastes examined followed the same third degree polynomial model.  Fig. 4.14 shows the correlation between the superplasticizer saturation dosage and the superplasticizer dosage to achieve a flow time of 110 ± 10 sec for the various cement pastes. All the cement pastes investigated followed a unique polynomial model. The curve has an inflection point around a superplasticizer saturation dosage of 1%. This shows that beyond this limit, a small increase in the superplasticizer saturation dosage caused a larger increase in the superplasticizer dosage to achieve the same flow behavior of the cement paste. Although based on limited results, the practical implication of this behavior is that a rational and economic use of superplasticizers should not greatly exceed a dosage of 1%.  1.6  *  OPC+LF  •  OPC+10% SF+LF  0 0  0.2  0.4  0.6  0.8  1.2  1.4  1.6  HRWR saturation dosage [%]  Fig. 4.14 Correlation between superplasticizer saturation dosage and superplasticizer required.  Chapter IV  65  4.9 CONCLUSIONS  The effect of 0 to 25% limestone microfiller replacement by volume of cement on the superplasticizer efficiency and the rheology of 0.3 to 0.4 w/b ratio cement pastes was investigated in silica fume and non silica fume systems. The following conclusions can be drawn: •  The superplasticizer efficiency was improved by the LF replacement of cement and this contribution offset the negative effect due to the consequent increase in the surface area. The temperature of the cement pastes after mixing was slightly reduced in the presence of LF which implies a possible improved control of the very early reactions (before the dormant period) between cement and water.  •  The LF replacement of cement slightly increased the yield stress of cement paste and decreased its plastic viscosity, which implies a better stability and flowability of the cement paste. However, increased LF levels reduced the induced bleeding of cement paste only at high w/b ratios and did not seem to have a significant effect at low w/b ratios.  •  Mortars made with cement pastes of constant flow time had increased flow as the LF replacement of cement increased. This implies that beside the effect on the cement paste yield stress and plastic viscosity, the L F effect on the reduction of the friction and interlocking of aggregates can be important. This lubricating aspect may be more significant in low w/b ratio concretes.  4.10 REFERENCES [1]  European Pre-standard pr ENV 197-1 [1992], Cement-Composition, Specifications and Conformity Criteria - Part 1: Common Cements, Final Draft, pr ENV 197-1: 1992E.  [2]  Schmidt, M. [1992], "Cement with interground additives—Capabilities and environmental relief, Zement-Kalk-Gips, Vol. 45, No. 4. pp. 67-91 and No. 6, pp. 296301.  [3]  Roy, D.M., Skalny, J. and Diamond, S. [1982], "Effects of blending materials on the rheology of cement paste", Effect of Surface and Colloid Phenomena on Properties of Fresh Concrete, Proc: MRS Symposium, Boston, Nov. 1-4, J. Skalny, ed., pp. 152-173.  [4]  Struble, L.J. [1990], "The rheology of fresh cement paste", Advances in Cementitious Materials, S. Mindess, ed., American Ceramic Society, Vol. 16, pp. 7-26.  Chapter IV  66  [5]  Powers, T.C. [1968], The Properties of Fresh Concrete, John Wiley & Sons, New York, 664 p.  [6]  Krieger, I.M. and Dougherty, T.J. [1959], "A mechanism for non-newtonian flow in suspensions of rigid spheres", Trans. Soc. Rheol., 3, pp. 137-152.  [7]  Malek, R.I.A. and Roy, D.M. [1990], "Modeling the rheological behavior of cement pastes: a review", Advances in Cementitious Materials, S. Mindess, ed., American Ceramic Society, Vol. 16., pp. 31-41.  [8]  Papo, A. [1988], "Rheological models for cement pastes", Materiaux et Constructions, Vol. 21, pp. 41-46.  [9]  Kantro, D.L. [1981], "Influence of water-reducing admixtures on properties of cement paste - A miniature slump test", Cement, Concrete, and Aggregates, Vol. 2, No. 2, pp. 95-102.  [10]  Montgomery, D.C. [1984], Design and Analysis of Experiments, John Wiley and Sons, New York, pp. 261-292 and 460-470.  [11]  Design-Expert, Version 5.0.3, STAT-EASE Inc., 2021 East Hennepin Avenue, Suite 191, Minneapolis, MN 55413.  [12]  Bombled, J.P. [1986], "Rheology of fresh concrete: Influence of filler addition in the cements", 8* Int. Cong, on Chem. Cem., Rio de Janeiro, Vol. IV, pp. 190-196.  [13]  Pierre, A., Lamarche, J.M., Mercier, R. and Foissy, A. [1989], "Adsorption d'un fluidifiant du ciment sur le carbonate de calcium", Cement and Concrete Research, Vol. 19, pp. 692-702.  [14]  Brookbanks, P. [1989], "Properties of fresh concrete", in: Performance of Limestone Filled Cements, Proc. Building Research Establishment Seminar, pp. 4.1-4.15.  [15]  Cochet, G. and Sorrentino, F. [1993], "Limestone filled cements: properties and uses", Mineral Admixtures in Cement and Concrete, Ghosh, S.N., Sarkar, S.L. and Harsh, S., eds., ABI Book Pvt. Ltd., New Delhi, Vol. 4, pp. 266-295.  [16]  Neto, S.N. and Campitelli, V.C. [1990], "The influence of limestone additions on the rheological properties and water retention value of portland cement slurries", ASTM STP 1064, pp. 24-29.  [17]  Feldman, R.F., Ramachandran, V.S. and Sereda, J.P. [1965], "Influence of CaC0 on the hydration of 3CaO.Al 0 ", Journal of the American Ceramic Society, Vol. 48, No. 1, pp. 25-30. 3  2  3  [18]  Nehdi, M., Mindess, S. and A'itcin P-C. [1995], "Use of ground limestone in concrete: a new look", Building Research Journal, Vol. 43, No. 4, pp. 245-261.  [19]  Tattersall, G.H. and Banfill, P.F.G. [1983], The rheology of fresh concrete, Pitman Adv. Pub. Prg., London, pp. 283-284.  Chapter V  67  Chapter V  MICROFILLER E F F E C T ON R H E O L O G Y OF HIGH-PERFORMANCE CONCRETE  5.1 INTRODUCTION  The incorporation of microfillers and supplementary cementing materials in concrete as partial replacement of cement continues to increase. The properties of hardened concrete depend strongly on the early stages of concrete production. However, a fundamental understanding of the effect of these ultrafine particles on the rheology of HPC has not yet been achieved. On the other hand, the emergence of new special concretes such as fluid and self-leveling HPC has shown that concretes of the same slump may behave quite differently on the job. For these concretes, one cannot rely on the traditional workability tests for quality control and rheological characterization. Currently the concrete industry faces several questions which have not received due investigation, such as: Why is the production of HPC made  Theologically  easier when  ultrafine particles are added to the concrete mix? How is the superplasticizer requirement affected by adding microfillers? How do the ultrafine particles affect the slump loss of fresh HPC? What happens to the rheological characteristics of fresh HPC while it stiffens with time? Is it Theologically advantageous to use triple-blended binders containing microfillers of various particle sizes? This chapter constitutes a contribution towards a better understanding of these aspects.  5.2 APPARATUS  The rheometer used in this study was developed at the University of British Columbia [1]. A computer drives a motor from rest to the desired speed and then back to rest, which causes a planetary motion of a four-finger impeller. A tachometer is used to measure the impeller speed, and a torque-measuring device equipped with four strain gauges measures the torque from the deflection of a small beam in bending. The impeller speed and torque data are acquired by a data acquisition system. A computer program allows the user to customize test parameters, such as  Chapter V  68  the number of readings, the speed increment, the speed decrement, the sampling interval between the readings, etc. In this work the rheometric test consisted of an impeller speed loop starting from rest and going up to about 1.08 rev/sec in 10 increments. The speed was then gradually reduced to zero in 30 decrements. For each step, a total of 50 measurements of speed and torque were made in approximately 0.06 sec followed by a waiting period of 1.2 sec. The test duration was 2.8 minutes, a compromise between accuracy and the segregation due to longer tests. The rheometer is shown in Fig. 5.1.  The induced bleeding test consisted of applying a controlled pressure of 2.1 MPa on a cylindrical sample of concrete which was placed in an air-tight cell, and measuring the water collected from a bleed hole over time. The test set-up and procedure are described in detail elsewhere [1].  Fig. 5.1 Illustration of the UBC Rheometer.  Chapter V  69  5.3 MATERIALS AND PROCEDURE  Ordinary ASTM Type I portland cement (OPC) was used. Two high purity limestone fillers (LF1 and LF2), finely ground silica (GS) and silica fume (SF), were used as partial replacements for cement. The fineness and average particle size of the cement and various microfillers are given in Table 5.1. Washed gravel having 10 mm maximum particle size was employed as coarse aggregate, and siliceous sand having a fineness modulus of 2.3 was used as the fine aggregate. Tap water was employed for the mixing. A naphthalene sulfonate superplasticizer was used for the 0.33 w/b concrete mixtures. A more efficient superplasticizer from the same category was necessary for the 0.25 w/b concrete mixtures.  Table 5.1- Fineness and average particle size of the cement and fillers BET surface area  Average particle size [p;m]  [m /kg] 2  OPC  350  25.6  SF  18000  0.26  LF1  2300  LF2  10000  0.7  GS  1250  13.8  .  2.9  Proportions of 0, 5, 10, 15, and 20% of LF1, LF2, GS, and SF were individually dry blended with the cement. HPC mixtures were prepared with these composite cements at a w/b of 0.33 and a constant slump of 200 ± 20 mm. The mixing was carried out using a pan mixer, then the concrete was transferred to the rheometer. The mixing sequence is illustrated in Fig. 5.2. Extensive mixing was carried out to maximize the dispersion of the very fine particles. The rheometer test was carried out for these concretes at 15, 30, 45, 60, and 90 minutes, and the slump test was carried out at 15, 30, and 60 minutes (ASTM C143-90, Standard Test Method for Slump of Hydraulic Cement Concrete). In addition, self-leveling HPC mixtures having a w/b of 0.25 and incorporating 15% of various microfillers were prepared and tested in the rheometer at the same time intervals as above. The 15% cement replacement level was selected because it seemed optimal [2] in maintaining high compressive strengths.  Chapter V  70  Add coarse aggregate + 1 L of water Add cement + filler + sand Add water gradually Superplasticizer added till target behavior is achieved  Time [min]  12  10 Rest |  Mixing Fig. 5.2 Mixing sequence.  Table 5.2-  M i x proportions of concrete mixtures [kg/m ] 3  0.33 w/b fluid HPC mixtures Replacement rate [%] By volume Cement Sand GS LF1 LF2  or or or or  SF 10 mm aggregate Sand Water Superplasticizer  0.25 w/b self-leveling mixtures  0  5  10  15  20  0  15  500  475  450  525  400  550  468  0  21.0  42.0  63.0  86.0  —  —  0  21.0  42.1  63.1  84.1  0  69  0  21.5  43.0  65.0  86.0  0  71  0  21.5  43.0  65.0  86.0  0  71  0  17.7  35.4  53.1  70.8  0  58  1050 715 165 L Adjusted for 200 ± 20 m m slump  1050 695 128.5 L Adjusted for slump flow>51 c m  The slump flow test was conducted on these mixtures similarly to the regular slump test. The measured value is the mean base diameter of the concrete spread after the slump test. A lower bound of 51 cm slump flow value without segregation was used in this study as a criterion to define the self-leveling behavior. Both fluid HPC and self-leveling HPC were not agitated and were allowed to rest between consecutive rheological tests. They were also covered to prevent loss of moisture. The composition of the various concrete mixtures is given in Table 5.2.  Chapter V  71  Furthermore, in order to investigate the microfiller effect on the rheology of triple-blended cements, a uniform precision experimental plan (see 4.3) including 13 concrete mixtures was designed. The proportions of cement, silica fume and limestone powder were varied according to the factorial experimental plan requirements. Isoresponse curves for the superplasticizer demand, slump loss, flow resistance and torque viscosity at different ages can thus be obtained.  5.4 APPLICABILITY AND SIGNIFICANCE OF RHEOMETRIC TESTS FOR T H E R H E O L O G Y OF FLUID AND SELF-LEVELING HPC  A reliable measurement tool for the effect of microfillers on the rheology of fluid HPC mixtures is required. It has been recognized that the slump test is not a satisfactory workability criterion for such fluid mixtures which have slump values higher than 180 mm [3]. The compacting factor is inappropriate since fluid concrete will compact into a mold yielding a value close to 1. In addition, the suitability of the DIN flow table test for fluid concrete is questionable [4]. A test that can be used as an acceptance criterion for the rheology of fluid and self-leveling concrete, and that can differentiate between concretes meeting this criterion is therefore needed.  Since the flow properties of fresh concrete seem to approximate closely the Bingham model, at least in the range of shear rates which are considered more relevant in practice [5], considering the workability in terms of at least two constants would reveal rheological characteristics normally not depicted by conventional single point tests. However, others have found that the Bingham model, although useful, has its limitations, and have suggested that combined use of the slump test and a flow measure can provide more practical information [6].  For practical reasons, coaxial-cylinders viscometers are not considered appropriate for concrete: an inconvenient apparatus size to reconcile the conflicting requirements of a large gap between coaxial cylinders and a small ratio of the radii of the outer to the inner cylinder (ratio as close to 1.0 as possible), slippage at the cylinder walls, etc. Nonetheless, cylinder/cylinder shear [7-9] and plate/plate shear [10], amongst others, have been used to try to describe the rheology of fresh concrete using viscometers. An alternative procedure [5] is to base the workability tests on the torque requirement in mixing. Since the shear rate in a mixer varies significantly from point to point, it is difficult to carry out a consistent analysis. It is therefore assumed that an effective  Chapter V  72  average shear rate exists, which is proportional to the impeller speed (eq. 5.1). This involves the assumption that the flow of concrete in the mixer is laminar, and that the shear threshold is exceeded (no plug flow).  T = g + hN  (5.1)  where: T is the torque to drive the rheometer impeller (Nm), g is the flow resistance (Nm), h is the torque viscosity (Nm.s) and N is the impeller angular speed (rps). By means of suitable calibration, it is possible to convert the flow resistance and torque viscosity to yield value and plastic viscosity in fundamental units [5].  In the first part of this chapter, rheometer results based on the principle of the torque requirement in mixing are examined, and compared to the slump test results. An effort is made to highlight the relevant information in a torque/impeller-speed curve (referred to here as a flow curve), and to see whether it is sufficient to characterize the rheology of fluid and self-leveling HPC. The possibility of depicting the rapid stiffening behavior of fresh HPC and the effect of microfillers on its rheology using rheometric tests are also examined.  5.4.1 Fluid High-Performance Concrete  Fig. 5.3 illustrates flow curves at different intervals of time, which are typical for the rheology of fluid HPC mixtures having a w/b ratio of 0.33. For the total of 85 rheometric tests carried out on these concretes (5 tests per mixture), the correlation coefficient of the descending part of the flow curve to the Bingham model had a mean value (u) of 95.49% and a standard deviation (S ) x  of 3.13%. This shows how closely the rheology of HPC correlates to the Bingham model regardless of the amount of superplasticizer required and the type and proportion of microfiller added to the mixture. Generally, as the fresh HPC stiffened with time, the correlation coefficient to the Bingham model was slightly reduced, probably because the laminar flow assumption was no longer quite correct.  Chapter V  73  o*  1  !  1  1  ^  0  0.2  0.4  0.6  0.8  1  I m p e l l e r s p e e d [rev/sec] (a)  0 J  1  1  1  1  f.  0  0.2  0.4  0.6  0.8  1  I m p e l l e r s p e e d [rev/sec] (b)  Fig. 5.3 Illustration of flow curves at 15 min (a), and 60 min (b) for fluid HPC.  Whether the flow resistance (g) and the torque viscosity (h) were extracted from the ascending or the descending part of the flow curve yielded comparable results except at 15 min. However, whenever the initial low impeller speed region of the flow curve (below 0.4 rps) was considered alone, the ascending and descending parts of the flow curve yielded different values for g and h. Thus, measuring the rheological properties of this type of concrete based only on two tests at different shear rates would not provide a global view of the flow curve, and a reasonable estimation of the two rheological characteristics g and h would not be obtained. Each pair of measurements would yield different results depending on whether the two points were on the ascending part of the curve, the descending part, or at extreme impeller speeds for the same branch of the curve.  Chapter V  74  Fig. 5.3 shows an initial highlighted region on the flow curves where the torque requirement increased for a nearly zero angular speed. The number of highlighted points corresponds to the number of time increments for which the impeller speed was still equal to zero. These first points on the torque axis show the initial torque that has to be overcome before any measurable angular speed of the impeller can be observed. Since the concrete was allowed to rest between the tests, this maximum initial torque increased with time. Once this value was overcome, the behavior followed closely the linear Bingham model. This large increase in the initial flow resistance with time, which is an effect of the thixotropic nature of concrete, was observed recently by other investigators and termed the resting shear yield stress [11]. It is a value distinguished from the flow resistance obtained in a steady state regime.  The maximum torque requirement attained during the impeller speed loop tended to decrease when the finer and more spherical microfillers were substituted for cement, yet it did not show a clear trend with time (Fig. 5.4). Also, the correlation between the slump and the flow resistance, g, tended to be linear while the correlation between the slump and the torque viscosity did not show a clear trend (Fig. 5.5). Although some published data have suggested that there was a linear or curved relationship between the slump and the flow resistance [3], more recent work [12] has shown that this relationship breaks down at slumps in excess of 150 mm, a value beyond which the slump test is increasingly meaningless. It is also reported [12] that computer simulations have demonstrated that a change in the yield value affects the slump much more than a comparable change in the plastic viscosity.  Fig. 5.4 Maximum torque requirement for fluid HPC mixtures.  Chapter V  75  250  4  5  6  Torque Viscosity [Nm.s] (b)  Fig. 5.5 Correlation between slump and g (a)', and slump and h (b) for fluid HPC mixtures.  5.4.2 Rheology of Self-Leveling High-Performance Concrete  Self-leveling concrete can be defined as concrete having a yield stress small enough to be overcome by its own weight. The plastic viscosity only determines the rate at which the flow will proceed. Typical rheometer flow curves at 15 minutes and 60 minutes for the self-leveling HPC mixtures are illustrated in Fig. 5.6. For the 25 rheometric tests carried out on this type of concrete, the correlation coefficient of the descending part of the flow curve to the Bingham model had a mean value of 95.90% and a standard deviation of 2.13%.  1  g was obtained from linear regression of torque impeller speed data below 0.4 rps.  Chapter V  76  0  0.2  0.4  0.6  0.8  1  1.2  I m p e l l e r s p e e d [rev/sec] (a) 15 m i n  0 -| 0  1  1  i  1  I  0.2  0.4  0.6  0.8  1  1  1.2  I m p e l l e r s p e e d [rev/sec] (b) 60 m i n  Fig. 5.6 Illustration of typical flow curves for self-leveling HPC mixtures.  The self-leveling HPC mixtures were all thixotropic at all test ages. While the descending part of the flow curve was linear, the ascending curve consisted generally of an initial nearly linear part with a steep slope, followed by a plateau where the torque requirement was nearly constant. The linear part of the ascending curve became steeper with time. This means that the resting structure of concrete is significantly different from the steady state structure reached during a test. When this resting structure was broken down at higher shear rates, the behavior became closer to the descending part of the curve, which may explain the occurrence of the plateau. It is also suspected that this plateau may be partly an effect of some segregation of the concrete in the rheometer bowl.  Chapter V  77  The ascending curve did not approximate the Bingham model well. Hence, two rheological tests carried out at different shear rates could yield different values for the rheological characteristics g and h depending on the two shear rates selected. Therefore, measuring the rheological characteristics without accounting for the shear history of the fresh self-leveling HPC mixture may not provide accurate results. Researchers have recognized that certain time dependent phenomena affect the behavior of cementitious suspensions. A downward flow curve has been preferred, since the time to attain equilibrium at a given shear rate is shorter when going from a higher to a lower shear rate. It may be argued however that the structure of concrete is changed when descending measurements are carried out. On the other hand, using an upward curve with a much longer test would enhance segregation, which is an acute problem for rheometric tests.  The flow resistance g for the self-leveling mixtures measured from the ascending curve at 15 minutes (p = 0.63 Nm, S = 0.16 Nm) was generally lower than the values obtained for the fluid x  HPC mixtures at the same age (u = 1.75 Nm, S = 0.43 Nm). The self-leveling mixtures required x  very low stresses to initiate and maintain a plastic deformation, but showed greater resistance to higher shear rates. This is unlikely to be due to slippage or any other artifacts since it was consistent with operator observation when handling the concrete. The flow resistance did not reflect the workability of self-leveling concrete as it did for the 0.33 w/b fluid HPC mixtures. For example, at 15 minutes g was 0.38 Nm for the OPC-15% GS mix which was difficult to work, while it was 0.76 Nm for the OPC-15% SF mix which was easy to work. In addition, between 15 and 90 min, the flow resistance was nearly constant for the OPC, OPC-GS and OPC-LF1 mixtures which were difficult to work, while it more than doubled for OPC-LF2 and OPC-SF which were easily workable . 2  On the other hand, the torque viscosity, h, was nearly one order of magnitude higher for the selfleveling HPC mixtures (u = 21.23 Nm.s, S = 9.32 Nm.s, at 15 min.) compared to the normal x  HSC mixtures (u = 2.65 Nm.s, S = 1.08 Nm.s, at 15 min.). Contrary to the previous case of the x  fluid HPC mixtures, g for the self-leveling concrete mixtures was of little relevance for the evaluation of the workability, while h reflected the differences in the workability of the various concrete mixtures. However, h did not reflect the stiffening behavior of concrete with time. It  The term workability in this context refers to the subjective judgment of the operator in the handling of the concrete.  2  Chapter V  78  remained nearly constant for some mixes, while it even tended to decrease slightly for others; this may be due to segregation of the coarse aggregates.  The maximum torque requirement did not reflect the differences in workability for the various concrete mixtures. Thus, basing the rheological characterization on the maximum torque requirement did not seem to be effective. The impeller speed at which the maximum torque was reached (IS ma*) increased as the fineness of the microfiller substituted for cement was increased (OPC: 0.19, GS: 0.20, LF1: 0.23, LF2: 0.36, SF: 0.77 in (rev/sec), at 15 min.). However, IS^did not show a clear trend as the concrete stiffened with time.  5.5 SUPERPLASTICIZER REQUIREMENT  The superplasticizer requirement to achieve a slump of 200 ± 20 mm for the 0.33 w/b concrete mixtures is shown in Fig. 5.7. Replacing proportions of the cement by concrete sand reduced the superplasticizer requirement probably due to an increase in the w/b ratio and a reduction of the wettable surface area. Partial substitution of GS and LF for cement slightly reduced the superplasticizer requirement. However, SF replacement of cement increased the superplasticizer demand to achieve a constant workability.  Fig. 5.7 Superplasticizer requirement for a constant workability.  Chapter V  79  It is interesting to note that 20% replacement of cement by LF2 (mean particle size = 0.7 um) slightly reduced the superplasticizer requirement, while a 5% replacement by SF (mean particle size = 0.26 um) increased the superplasticizer requirement. This implies that the high surface area of SF may not be the sole factor affecting the increase in the superplasticizer demand for SF mixtures, and supports the idea that SF may have an affinity for multi-layer adsorption of superplasticizer molecules. Previous work [13] has also shown that while using a 3 um LF tended to reduce the superplasticizer requirement to achieve a constant flow for cement pastes, SF increased the superplasticizer requirement significantly.  5.6 FLOW RESISTANCE  The flow resistance, g, at various ages after mixing for different levels of partial replacement of cement by GS, LF1, LF2, and SF microfillers is illustrated in Fig. 5.7. For fluid HPC mixtures, g tended to increase with time reflecting the stiffening behavior of concrete with aging. Partially substituting various fine fillers for cement reduced g, although the surface area, and thus the wettable surface, were increased. The improved gradation of the binder and the lubricating effect imparted by the fine particles may have reduced the aggregate interlocking, and may have reduced g as a result.  The finer and the more spherical the filler, the more g was reduced. It may be argued, however, for the SF mixtures, that the superplasticizer requirement was higher and g might thus be expected to be lower. This was not true for the GS and L F mixtures which required less superplasticizer than the reference pure OPC mix. For self-leveling mixtures, g tended to increase with time only for mixtures containing ultrafine microfillers which required less superplasticizer to achieve a self-leveling behavior. The effect of the microfillers on g was possibly offset by the significant difference in the superplasticier dosage between the various mixtures.  Chapter V  80  • 15 min  • 30 min D 6 0 min  • 15 min • 30 min • 60 min  OPC  GS  LF1  LF2  SF  (b) Fig. 5.8 Flow resistance at various ages for (a) fluid, and (b) self-leveling HPC mixtures (filler = 15%).  5.7 TORQUE VISCOSITY The torque viscosity, h, for various mixtures at different ages is illustrated in Fig. 5.9. Generally, the finer the microfiller, the more h was reduced . This was true for both fluid and self-leveling 3  mixtures. Generally, higher proportions of filler replacement of cement were more effective in reducing the torque viscosity. It was observed earlier that higher filler proportions reduced the plastic viscosity of cement paste in both silica fume and non-silica fume systems [13]. Yet, this effect was more significant in concrete than in cement paste probably because the ultrafine particles played a more important lubricating role via a reduction of the aggregate interlocking.  Chapter V  81  This will be illustrated more clearly later, in the discussion of the results of the statistical modeling of the filler effect on the torque viscosity of HPC made with triple-blended composite cements.  • 15 min • 30 min • 60 min  OPC  GS  LF1  LF2  SF  (a)  • 15 min • 30 min D 60 min  OPC  GS  LF1  LF2  SF  (b)  Fig. 5.9 Torque viscosity at various ages for (a) fluid, and (b) self-leveling HPC mixtures (filler = 15%).  The torque viscosity did not, however, reflect the stiffening behavior of the fresh concrete with time. Since concrete was allowed to rest between consecutive rheological tests, the torque requirement for the low impeller speeds increased more than that for the high impeller speeds probably due to flocculation of binder particles, growth of hydration products and exhaustion of superplasticizer molecules in the complex chemical processes. This resulted in a reduction of the slope of the shear stress vs. shear rate curve, which may explain why the torque viscosity did not increase with time and tended to decrease slightly in some cases. Induced segregation may also contribute to this behavior. It should also be remembered that the torque viscosity is a 'dynamic' value as opposed to the flow resistance and slump test, which can be regarded as 'static' values.  It should be remembered that the mixtures under investigation contain a substantial amount of superplasticizer.  3  Chapter V  82  5.8 SLUMP LOSS  Although superplasticizers are effective in dispersing the cement particles through a deflocculation process, this action is time dependent and very low w/b concrete usually undergoes a rapid slump loss. There are two concerns regarding this issue: the cement should have the lowest rheological reactivity, that is, the amount of water fixed immediately after mixing should be minimal, and the superplasticizer should not compete with the calcium sulfate to neutralize the C A [14]. The changes of the slump loss behavior occurring upon partial 3  replacement of cement by microfillers of various mineralogies and mean particle sizes is not yet clear. Fig. 5.10 illustrates the slump variations of 0.33 w/b concretes at 15, 30 and 60 min. Results are shown for 0, 5, 10, 15 and 20% replacement of cement by GS, LF1, LF2, and SF. Starting at slumps of 200 ± 20 mm, all the mixtures had slumps higher than 60 mm after 60 min.  Fig. 5.10 Slump variation of concretes made with different proportions of various fillers.  Chapter V  83  The blended composite cements generally showed performance equal to or better than pure OPC. Generally, concretes which had the highest slumps at 15 min kept higher slumps at 60 min. It should be remembered that only the silica fume mixtures had higher superplasticizer dosages (Fig. 5.7). Therefore, the effect of the difference in the superplasticizer dosage on the slump loss is limited for the rest of the mixtures.  5.9 INDUCED BLEEDING  It has been known that the bleeding rate and bleeding capacity of cement paste are strongly dependent on the water content and the specific surface area of the cement [15]. Thus, it can be reasonably expected that partial replacement of cement by finer microfillers would reduce the induced bleeding of cement paste and concrete. The water thus retained may participate in lubricating the concrete mixture and improving the rheology. The induced bleeding of the various concrete mixtures at 5 and 30 min is illustrated in Fig. 5.11.  Fig. 5.11 Microfiller effect on induced bleeding of fresh high-strength concrete.  Chapter V  84  Replacing some of the cement by finer fillers reduced the bleeding both at 5 and 30 min. The finer the microfiller, the more the bleeding was inhibited, except for GS which, though being the coarsest microfiller used, was the most effective in reducing the bleeding at 5 min. The reason for this behavior is not clear. It may however be that the GS changed the electrolytic environment, which in turn increased the inter-particle attraction. This is known to reduce the bleeding capacity while generally having little influence on the bleeding rate [15].  5.10 R H E O L O G Y OF HPC M A D E WITH TRIPLE-BLENDED CEMENTS  The isoresponse curves resulting from the factorial experimental plan [16] for the rheological characteristics under investigation are illustrated in Fig. 5.12. The superplasticizer demand to achieve a 200 ± 20 mm slump (Fig. 5.12-a) was not markedly affected by the LF proportion at low SF rates. At high SF levels, the superplasticizer demand was reduced as the LF proportion increased. A previous investigation [13] also showed that increased LF levels reduced the superplasticizer amount required to achieve a constant flow time of low w/b ratio silica fume cement paste.  The slump loss of the fresh concrete mixtures is illustrated in Fig. 5.12-b. Increased levels of LF or SF reduced the slump loss of the concrete. A combination of moderate levels of LF and SF reduced the slump loss, while high levels seemed to enhance the slump loss, though a 10% LF10% SF binder is still advantageous compared to an OPC. However, the ability of the slump test to describe properly the effect of time on the rheology of this kind of concrete is questionable; the flow resistance and torque viscosity are more relevant.  The flow resistance, g, at 15 minutes decreased with higher SF levels, and increased with higher LF levels up to about 6%. Higher LF levels seemed to reduce g slightly. A combination of moderate levels of LF and SF reduced g, while high levels seemed to increase g. At 60 min, g increased with higher LF levels, yet a combination of high LF-SF levels did not seem to be deleterious as was suggested by the slump test.  Chapter V  85  (a) HRWR required [%]  (c) g at 15 min [Nm]  (b) Slump loss at 60 min [mm]  (d) g at 60 min [Nm]  Fig. 5.12 Statistical modeling of various rheological responses for OPC-LF2-SF triple-blended composite cements. (cont'd)  Chapter V  86  (e) h at 15 min [Nm.s]  (f) h at 60 min [Nm.s] Fig. 5.12 (cont'd)  The torque viscosity, h, was reduced as the LF and SF levels were increased, both at 15 and 60 min. The overall effect of replacing cement by a LF-SF combination on the workability was positive. It was rather surprising that h decreased from 15 to 60 min. It seemed that the increase of g with time and the torque requirement for the lower impeller speeds was not accompanied by an equal increase of the torque at higher impeller speeds. This resulted in a reduction of the slope of torque/impeller-speed curve, and an artificial decrease of viscosity as a consequence. It is possible that induced segregation may also contribute to this behavior. The results obtained above for concrete support similar results of previous work [13] where it was shown that increased LF levels resulted in an increased yield stress and a reduced viscosity of low w/b ratio cement pastes, both in SF and non SF systems.  5.11 ULTRAFINE PARTICLES AND R H E O L O G Y OF C E M E N T SUSPENSIONS  Various kinds of forces coexist in a cement suspension. First, there is the Brownian randomizing force [17] which influences the spatial orientation and arrangement of particles. This force is strongly size dependent and has a major influence below a particle size of 1 urn. The effect of  Chapter V  87  microfiller partial substitution of cement on this kind of force is beyond the scope of this investigation.  Second, there are forces of colloidal origin, which arise from mutual interactions between particles, and are affected by the polarizability of water. When the van der Waals attraction between cement grains and the electrostatic attraction between unlike charges on the surface of particles are dominant, the net result is an attraction and the particles tend to flocculate. However, in the presence of polymeric or surfactant materials on the surface of cement grains, the net result may be a repulsion and the particles remain separate. In this respect, the filler material can influence the electrostatic forces depending on its mineralogical nature and the state of its particle surface charges. Since colloidal forces depend also on the average distance between nearest neighbor particles, the interposition of finer filler grains between cement particles may affect their electrostatic attraction and thus their flocculated structure. Likewise, replacing cement by a material of different specific surface area would change the wettable surface area and the amount of water adsorbed. Some fillers having a certain solubility in water may modify the electrolyte solution and thus the electrostatic forces.  Third, there are viscous forces which are proportional to the local velocity difference between a cement particle and the surrounding water, and between an aggregate and the surrounding cement paste. Since cement based materials fall in the range of dense suspensions, particles have to move out of the way of each other, especially when floes are formed. The filler effect on the rheology will depend on its fineness, its particle size distribution and its particle shape. Broader particle-size ranges have a higher maximum particle packing because the finer particles fit into the gaps between the coarser particles. The viscosity of suspensions usually increases as the deviation from ideal grading increases. For a certain volume of water, the viscosity reaches a minimum at the most compact arrangement of particles [17]. In addition, any deviation from a spherical shape implies an increase in viscosity for the same phase volume. Thus, in the presence of superplasticizers, the finer and the more spherical the filler, the better the rheological properties. Filler materials can also have different efficiencies  in the adsorption of  superplasticizers. They can, if soluble, introduce certain ions which may affect the kinetics of the hydration reaction and the nucleation of hydration products.  Chapter V  88  The idea that results from the above discussion is that one can combine the effects of superplasticizers and suitable ultrafine particles to make HPC easier to place. The superplasticizers combat the electrostatic forces and ensure the dispersion of the ultrafine particles, while the microfillers reduce the amount of water required to fill the pores between particles, and the viscous forces and the Coulomb friction between aggregates.  5.12 CONCLUSIONS High-performance concrete can be made easier to place by substituting portions of ultrafine particles for cement. In the presence of superplasticizers, the finer the microfiller the lower the flow resistance and torque viscosity of the mixture. Up to 20% of ground silica or limestone powder did not increase the superplasticizer requirement to achieve a constant workability, even though one of these fillers had a surface area as high as 10 m /g. Silica fume, however, while 2  being the most effective filler from a rheological point of view, increased the superplasticizer requirement needed to achieve similar workability. This may suggest that a high surface area is not the sole parameter influencing the superplasticizer demand of silica fume mixtures and that silica fume may have a strong affinity for multi-layer adsorption of superplasticizer molecules. Microfillers did not seem to reduce systematically the slump loss of fresh HSC, and were advantageous in some instances in maintaining better workability over time. Microfillers were also successful in inhibiting the induced bleeding of fresh concrete. It also seemed possible to design triple-blended composite cements with different fillers which may achieve improved rheological characteristics.  5.13 REFERENCES [1]  Beaupre, D. [1994], "Rheology of High Performance Shotcrete", Ph.D. thesis, University of British Columbia, 250 p.  [2]  Nehdi, M., Mindess, S., and Aitcin, P-C. [1996], "Optimization of high-strength limestone filler cement mortars", Cement and Concrete Research, Vol. 26, No. 6, pp. 883-893.  [3]  Banfill, P.F.G. [1980], "Workability of flowing concrete", Magazine of Concrete Research, Vol. 32, No. 110, pp. 17-26.  Chapter V  89  [4]  Dimond, C.R., and Bloomer, S.J. [1977], "A consideration of the DIN flow table", Concrete, Vol. 11, No. 12, pp. 29-30.  [5]  Tattersall, G.H., and Banfill, P.F.G. [1983], The Rheology of Fresh Concrete, Pitman Adv. Pub. Prog, 365 p.  [6]  Smeplass, S. [1994], "Applicability of the Bingham model to high-strength concrete", Special Concretes Workability and Mixing, Proc. RILEM Int. Workshop, Bartos, P.J.M., ed., pp. 145-151.  [7]  Uzomoka, O.J. [1974], "A concrete rheometer and its application to a rheological study of fresh concrete", Rheologica Acta, Band 13, Heft 1, pp. 12-21.  [8]  Murata, J. and Kikukawa, H. [1974], "Studies on rheological analysis of fresh concrete", Fresh Concrete: Important Properties and their Measurement, Proc. RILEM Seminar, Leeds, pp. 1.2-1 to 1.2-33.  [9]  Wallevik, O.H. and Gjorv, O.E. [1990], "Development of a coaxial cylinders viscometer for fresh concrete", Properties of Fresh Concrete, Proc. RILEM Colloq., Hanover, pp. 213-224.  [10]  de Larrard, F., Szitkar, J.-C, Hu, C , and Joly, M. [1994], "Design of a rheometer for fluid concretes", In ref. 6, pp. 201-208.  [11]  Hu, H. and de Larrard, F. [1996], "The rheology of high-performance concrete", Cement and Concrete Research, Vol. 26, No. 2, 1996, pp. 283-294.  [12]  Tattersall, G.H. [1991], Workability and Quality Control of Concrete, E&FN SPON, pp. 81-82.  [13]  Nehdi, M., Mindess, S., and Ai'tcin, P-C. [1997], "Statistical modeling of the microfiller effect on the rheology of composite cement pastes", Advances in Cement Research, Vol. 9, No. 33, pp. 37-46.  [14]  Ai'tcin, P-C. [1992], "The use of superplasticizers in high performance concrete", High Performance Concrete: from Material to Structure, Yves Malier, ed., E & FN Spon, p. 14.  [15]  Powers, T.C. [1968], The Properties of Fresh Concrete, John Wiley & Sons Inc., p. 553.  [16]  Nehdi, M., Mindess, S., and Ai'tcin, P-C. [1996], "Optimization of triple-blended composite cements for making high-strength concrete", World Cement Research and Development, Vol. 27, No. 6, pp. 69-73.  [17]  Barnes, H.A., Hutton, J.F., and Walters, K. [1989], An Introduction to Rheology, Elsevier, p. 115.  Chapter VI  90  Chapter VI  MICROFILLER E F F E C T ON M E C H A N I C A L PROPERTIES OF HIGH-PERFORMANCE CONCRETE  6.1 I N T R O D U C T I O N  Considerable research has been carried out within the last 15 years on the use of ground limestone in ordinary concrete. In recent years, the emergence of high-strength cement based materials has sparked a new interest in microfillers. Coupled with the effect of superplasticizers, ultrafine powders can improve the particle packing of the cementitious system, its rheological properties in the fresh state, and its mechanical properties and durability. However, the fundamental mechanisms of action of these microfillers have not received due investigation. For instance, it is not clear how the addition rate, fineness, and mineralogy of the fillers affect the mechanical strength and elastic properties of HPC. The particular effect of inert and nearly inert fillers on the early age properties of HPC has not been fully addressed, and the use of tripleblended cements containing a combination of inert and pozzolanic fillers to make HPC needs further study. This chapter will focus on the effect of a variety of microfillers on the compressive strength, flexural strength, and modulus of elasticity of high-strength mortars and concretes, both in binary and triple-blended cements. An attempt is made to explain the particular effect of limestone powder on the early age mechanical strength. 6.2 I N V E S T I G A T I O N S O N M O R T A R  6.2.1 Materials and Procedure  Materials described in Chapter IV: ASTM Type I cement, limestone filler (mean particle size = 3 urn), and silica fume were used. The particle size distributions of the cement and the limestone filler are shown in Fig. 1. The experiments were designed according to a uniform-precision factorial plan as described in § 4.3. The limestone filler content was varied systematically from 0 to 25%, and the w/b ratio from 0.30 to 0.40 according to the experimental plan requirements.  Chapter VI  91  - - - -Cement Filler  | II  100  i  I  i  I  1  t  | I I I I I I  10  I  I  | I II I I I  1  I  I  |Q 0.1  Particle size [um]  Fig 6.1 Particle size distribution of the cement and the limestone filler. The experimental plan was carried out for a pure OPC and an OPC containing 10% silica fume. A mortar corresponding to each of the constant-flow cement pastes of Chapter IV was made. Standard Ottawa sand was used. A proportion of sand of 2.3 times the mass of the binder was found suitable to achieve a flow table value of 100 ± 10% for the designed mortars. The mixing procedure for the mortars followed that of ASTM C305-94: Standard Practice for Mechanical Mixing of Hydraulic Cement Pastes and Mortars of Plastic Consistency, and the flow measure was carried out using ASTM C230-90 Standard Specification for Flow Table for Use in Tests of Hydraulic Cement. Standard 50 mm cubes were cast, demolded at 24 hours, and cured in lime saturated water until testing. The compressive strength was measured at 1, 3, 7, 28, and 91 days. The values presented are averages obtained on three specimens for each mortar.  6.2.2 Filler Effect on Mechanical Strength of Cement Mortars  The data that resulted from this factorial experimental plan were analyzed and plotted using the statistics software package employed in Chapter IV. Fig. 6.2 illustrates the isoresponse curves corresponding to the compressive strength data over the experimental domain under investigation. The Id compressive strength decreased with increased w/b ratio. It seemed not to be drastically affected by the LF replacement of cement up to about 10 to 15%; then, the response curves became steeper, indicating a loss of strength at high levels of LF, with the SF mortars being somewhat more sensitive to this effect (Fig. 6.2-a and 6.2-b). For example, at a w/b ratio of 0.30, Id compressive strengths of 35 and 32.5 MPa could be achieved with mortars having 12% and 21% LF, respectively, while in the presence of 10% SF, the same strengths could be achieved only with 10% and 15% LF, respectively.  Chapter VI  92  0.00  0.40  (c) OPC + LF at 3d  (d) OPC + LF + 10% SF at 3d  Fig. 6.2 Iso-strength curves for mortars with LF proportions in SF and non-SF systems. (cont'd)  93  Chapter V I  (g) OPC + LF at 28d  (h) OPC + LF + 10% SF at 28d  Fig. 6.2 (cont'd)  Chapter VI  94  (i) OPC + LF at 91d  (j) OPC + LF + 10% SF at 91d Fig. 6.2 (cont'd).  At the age of 3d, the compressive strength was also not significantly affected by the LF replacement of cement up to about 10 to 15%. Then, the response curves again became steeper, and the strength decreased with higher levels of LF replacement (Fig. 6.2-c and 6.2-d). The SF mixtures started to outperform the non-SF ones at this age. A 3d compressive strength of 40 MPa could be achieved at a w/b ratio of 0.30 using 12% LF replacement of cement (Fig. 6.2-c), or using 10% SF and 21% LF replacement of cement (Fig. 6.2-d). At 7d and later ages, the response curves became steeper towards the vertical axis, indicating that some early age effects due to the presence of L F were no longer as significant as they were earlier . The differences in 1  compressive strength between the SF and non-SF mixtures became much larger at 7d and later ages (Fig. 6.2-e to 6.2-j).  One interesting aspect of the above results is the positive effect of replacing some of the cement by a combination of LF and SF. Fig. 6.3 shows the compressive strength at a w/b = 0.33 of pure OPC, OPC + LF, OPC + 10% SF and OPC +10% SF + LF mortars. Blending SF and LF simultaneously with cement brought about significant strength improvements compared to the  The early age effects of limestone powder in cement based materials will be discussed in detail later on in this chapter.  95  Chapter VI  1  21  41  61  81  Age [d]  Fig. 6.3 Comparison of filler effect in binary and ternary blended cements.  pure OPC and OPC + LF systems, and compared to the OPC + 10% SF system at some LF proportions. An OPC +18% LF + 10% SF outperformed a pure OPC cement. This opens up the possibility of partially replacing portland cement by industrial byproducts combined with nonprocessed fdlers. Pozzolanic industrial byproducts can provide long term strength while carbonate additions can increase the early age strength. This can perhaps decrease the required cement without compromising the overall performance of the cementitious binder.  6.3 INVESTIGATIONS ON CONCRETE 6.3.1 Experimental  ASTM Type I cement along with the microfillers of Table 5.1 were used in this investigation. A total of 78 concrete mixtures were examined. Only the most relevant results to illustrate the mechanisms of filler effect on the mechanical properties of HPC are presented herein. HPC mixtures were made at a w/b ratio of 0.33. Levels of 0, 5, 10, 15, and 20% filler replacement by volume of cement were used. In addition, a uniform precision factorial plan was designed to optimize an OPC-LF2-SF triple blended cement. The two experimental variables were the proportion of LF2 and the proportion of SF replacement of cement. Likewise, in order to  Chapter VI  96  investigate the influence of the w/b ratio on the filler effect, concrete mixtures made with binary and ternary-blended cements were made with a 0.25 w/b ratio. In all of these mixtures, a pan mixer was used according to the mixing sequence of Fig. 5.2. Since the filler effect depends on the degree of dispersion of the ultrafine particles, several concrete mixtures were produced using a much more efficient OMNI mixer (Fig. 6.4).  For all mixtures, tap water, washed 10 mm maximum particle size gravel, silica sand having a fineness modulus of 2.3, and a naphthalene sulfonate superplasticizer having 42% solid content were employed. Cylinders (150 mm x 75 mm) and beams (100 mm x 100 mm x 350 mm) were cast using a vibrating table. Specimens were demolded at 24 hours after casting (except for the 12-hour compressive strength tests) and cured in lime saturated water until testing. The compressive strength was measured at 12 h, 1, 3, 7, 28, and 91 days as per ASTM C39 using 75 mm x 150 mm cylinders with mechanically ground ends. For each age, three specimens were tested.  The modulus of elasticity was measured at 1, 3, 7, and 28 days according to ASTM C469 on 75 mm x 150 mm cylinders. The axial load was measured using a load cell, and the displacement was measured by means of two LVDT's placed vertically on each side of the specimen (using the middle 100 mm of the specimen as the gauge length, Fig. 6.5). Each specimen was loaded and unloaded at a cross head speed of 0.5 mm/min up to about 40% of the average compressive strength of the concrete mixture. The load cell and LVDT signals were recorded using a digital data acquisition system. The modulus of elasticity, E, was calculated from the stress (a) vs. strain (e) curve obtained from the load cell and the averaged LVDT signals. The first loading/unloading cycle was ignored, and the modulus of elasticity was taken as the mean of the two values obtained from the two subsequent loading cycles.  The flexural strength test was carried out at Id and 28d on 100x100x350 mm beam specimens by means of a third point loading test according to ASTM C78-84 (Fig. 6.6). A hydraulic digital Instron machine with a 150 kN capacity was used with a cross head speed of 0.05 MPa per second. A load cell and two LVDTs continuously measured the load and the corresponding deflection of the beam specimen, and the signals were acquired by a data acquisition system. The flexural strength represents the average value obtained on three specimens.  Chapter VI  Fig. 6.5 Experimental set-up to measure the modulus of elasticity of concrete specimens.  97  Chapter VI  98  Fig. 6.6 Illustration of the flexural strength test.  6.3.2 Filler Effect on Compressive Strength of HPC  The compressive strength values for the 0.33 w/b HPC mixtures at ages varying from 12h to 9Id are illustrated in Fig. 6.7-a to 6.7-d. Results are shown for 0 to 20% replacement of cement by various microfillers.  At 12h, the compressive strength depended strongly on the type of microfiller added to the cement. The pure OPC mixture had very low strength. The addition of 5 to 20% of ground silica did not seem to affect the 12h strength significantly. The addition of silica fume increased the 12h strength, and this effect was higher the higher the replacement rate. The 3 um mean particle size limestone filler was not effective in increasing the 12h strength at 5% replacement rate, but had a significantly higher effect at increased replacement rates. The much finer 0.7 um mean particle size limestone filler had a greater influence and increased the very early strength by an order of magnitude at 15% replacement rate. The limestone fillers had the greatest effect on the 12h compressive strength, and this was more significant the higher the fineness and the replacement rate. Section 6.5 will focus on the explanation of this behavior.  Chapter VI  99  OPC  GS  LF1  LF2  SF  (a) 12h  45  (c)3d  Fig. 6.7 Filler effect on compressive strength (w/b = 0.33).  (Cont'd)  Chapter VI  100  Fig. 6.7 - continued F i l l e r effect on compressive strength (w/b =  0.33).  Chapter VI  101  At Id the compressive strength of the limestone filler mixtures was still advantageous compared to a pure OPC mixture. At this age, the silica fume concrete started to outperform the rest of the mixtures except those containing the ultrafine limestone. Only the ground silica mixtures had lower strengths than the OPC mixtures at this age.  At ages higher than 3d, the OPC mixtures gradually reached the compressive strength of the limestone filler mixtures or outperformed them, especially at high replacement rates. Contrary to the early age behavior, there was not a major difference between the strengths of the concretes made with the various inert fillers. The silica fume mixtures achieved distinctly higher long-term strengths. It is worth mentioning that concretes with 80 MPa strength and good rheological properties were achieved using as much as 20% replacement of cement by inert fillers in 0.33 w/b mixtures.  6.3.3 Dependence of Filler Effect on w/b Ratio  Fig. 6.8 shows the compressive strength ratio at different ages for mixtures with various proportions of fillers to a reference OPC mixture. Results are illustrated for mixtures incorporating 15% of microfillers at w/b ratios of 0.33 and 0.25.  The silica fume mixtures had compressive strength ratios higher than 1 at both w/b ratios. However, for a w/b ratio of 0.33, the limestone filler mixtures had compressive strength ratios higher than 1 only at early ages; but these values were lower than 1 at later ages. At a w/b ratio of 0.25, the limestone filler mixtures had compressive strength ratios higher than 1 even at 91d. The practical implication of these results is that HPC mixtures can be made with as much as 15% of a grinding mill filler with improved strength both at early and later ages. The increased efficiency of the microfiller effect at very low w/b ratios will be discussed from a microstructural point of view in Chapter 7. At this point it is thought that at a very low w/b ratio, a significant part of the cement plays a filler effect due to the limited hydration. Partial replacement of cement with a finer filler may improve the particle packing of the system and reduce its porosity without much loss in the bond, thereby increasing the strength.  Chapter VI  102  • i  d  • 3d D  7d  • 28d • 91d  w/b = 0.33  w/b = 0.25 (a) 15% LF1  s1  £  n c  4  • 1d • 3d  1.3  Ji  12  <n 1.1  I  1  <fl  u  g  0.9 0.8  o  w/b = 0.33  o  •7d Q28d • 91 d  w/b = 0.25 (b) 15% LF2  1.3  11 d |3d • 7d • 28d 0.9 ?  • 91d  0.8 w/b = 0.33  w/b = 0.25 (c) 15% SF  Fig. 6.8 Dependence of the microfiller effect on the w/b ratio. 6.3.4 Effect of mixing technique on compressive strength  Fig. 6.9 compares compressive strength results obtained with HPC mixtures made using a regular pan mixer versus a much more efficient OMNI mixer. Results are shown for mixtures incorporating various microfillers at two different w/b ratios. It was observed that the compressive strength generally increased when the OMNI mixer was used, and this was more significant in mixtures containing silica fume.  Chapter VI  103  P a n Mixer  • • • • • •  OPC 10% G S 10% LF1 10% L F 2 10% S F 10% LF2-10% S F  • • • • •  OPC 15% 15% 15% 15%  OMNI Mixer (a) w/b  = 0.33  145 135 "125 cn  S ts  g  115  B 1 0 % LF2-10% S F • 15% L F 2 - 1 5 % S F  105  S5  g  95  5  85  GS LF1 LF2 SF  Pan Mixer  OMNI Mixer (b) w/b  = 0.25  Fig. 6.9 Effect of mixing technique on compressive strength.  Of particular interest in this figure is that the replacement of 30%  of the cement by a combination  of 15% limestone filler-15% silica fume yielded compressive strength values 20% higher than an OPC  mixture at 28d. This triple-blended binder also achieved strengths that outperformed (at  early age) and compared to (at later age) an OPC +15% silica fume binder.  6.3.5 Filler effect on modulus of elasticity The modulus of elasticity values measured at various ages for concrete mixtures incorporating various proportions of microfillers are illustrated in Fig. 6.10. At early ages, the modulus of elasticity values for limestone filler mixtures were higher than for the reference OPC mixtures.  Chapter V I  104  This was maintained up to 28d for filler replacement rates lower than 15%. For 20% filler replacement, the modulus of elasticity was lower than that of a reference OPC mixture. On the other hand, the silica fume mixtures generally had a lower elastic modulus at early ages. After 7d, these mixtures often had higher modulus values than either the OPC or limestone filler mixtures.  3d  7 d  28 d  (a) Filler = 5%  3d  (b) Fillers 10%  7d  28 d  28 d  (c) Filler = 15%  (d) Filler > 20%  Fig. 6.10 Filler effect on modulus of elasticity at various ages. The modulus of elasticity versus the compressive strength for the various mixtures followed a power model (Fig. 6.11) similar to the model suggested by Gardner and Zhao [1]. Fig. 6.11 also shows the estimation of the modulus of elasticity from the compressive strength data using the ACI 318 committee relationship (equation 6.3). The figure confirms what we already know that the ACI method tends to overestimate the modulus of elasticity for concretes with compressive strength values higher than 40 MPa.  E = 9000(f ) c  033  c  E =6640 (f' ) c  c  c  E =0.043p (f j 15  c  for f' > 27 MPa  042  0  5  (6.1)  (Gartner and Zhao)  (6.2)  (from results above)  (6.3)  (ACI 318 equation)  Chapter VI  105  80000 tti CO  ACI 318  60000  ^  0 £ 40000 1 TJ  y = 6639.2x R = 0.8816  04227  20000  o  2  40  60 80 100 C o m p r e s s i v e strength [MPa]  Fig. 6.11 Modulus of elasticity versus compressive strength for various HPC  The  120  mixtures.  modulus mainly depends on the nature and fraction of aggregates per unit volume, which  was constant in all the mixtures. However, the various mineral admixtures affected not only the modulus of the cement paste, but also that of the aggregate-cement paste transition zone. This transition zone adds up to a considerable volume [2]: about one third to one half of the hardened cement paste. It is generally observed that a layer of oriented crystalline Ca(OH) , about 0.5 urn 2  thick, covers the surface of the aggregate, behind which there is a calcium silicate hydrate layer, also about 0.5 um thick. The two layers constitute what is known as the duplex film. The main interfacial zone is the next 50 urn further away from the aggregate, which contains larger Ca(OH) crystals and a low unhydrated cement content. The significance of this is twofold. First, 2  the absence of unhydrated cement suggests a lower initial cement content and a locally higher w/c ratio. Second, the presence of large calcium hydroxide crystals implies a higher porosity at the interface. If fine pozzolanic microfillers such as silica fume are added in the mixture, an improved packing at the interface is achieved, and less Ca(OH) is formed. This may explain the 2  slightly higher modulus of the silica fume mixtures at 28d.  The  higher modulus of the limestone filler mixtures at early ages is due to the accelerated  hydration and the more rapid development of strength. These mixtures also had modulus values that compared to and even outperformed OPC  mixtures at later ages. It seems that the improved  packing at the interface due to a physical filler effect of the ultrafine limestone particles compensated for the reduced C-S-H non-pozzolanic filler.  gel resulting from the partial replacement of the cement by a  Chapter VI  106  A model was proposed by Lutz and Monteiro [3], in which the aggregate particles are modeled as spheres surrounded by a radially non-homogeneous matrix. A constant and a term that decays with the radius following a power law represent the modulus of the matrix. It was found that the modulus of the transition zone was 15-50% lower than that of the bulk cement paste, which may explain why concrete falls below the Hashin-Shtrikman bounds for a two component material.  OPC  LF2  SF  (b) 28d  Fig. 6.12 Filler effect on flexural strength.  6.3.6 Filler effect on flexural strength The modulus of rapture at Id and 28d of HPC mixtures made with various proportions of limestone microfiller and silica fume is shown in Fig. 6.12. At Id, the limestone filler mixtures had higher flexure strengths than a reference OPC mixture, perhaps due to the more advanced hydration and better development of bond. At this age, the silica fume mixtures had decreased flexural strengths with increased replacement levels, though the mixtures outperformed an OPC  Chapter VI  107  concrete up to 20% silica fume content. At 28d, the flexural resistance of the limestone filler mixtures compared positively to an OPC mixture up to 15% replacement. Beyond this level the flexural resistance tended to decrease with the addition of limestone powder. Escadeillas [4] has also observed that at 28 and 90d, the filler replacement of cement caused a decrease in the flexural strength.  On the other hand, the 28d flexural resistance of the silica fume concrete increased with increased replacement rates. The paste-aggregate bond represents the major factor affecting the resistance of concrete to tensile stresses. Thus, the densification of the transition zone by the silica fume particles and the consumption of portlandite in the pozzolanic reaction may explain this improvement in the flexural strength. It has been observed earlier [4] that at early ages, a higher adhesion was obtained with cements containing the finest fillers. This was explained by a decreased thickness of the paste-aggregate transition zone and a decreased orientation of the portlandite crystals. This was confirmed in this work; the flexural strength was higher the finer the microfiller (Fig. 6.13).  121<I  1  3  **  11  CC to o 0.  10 / ]  lus  i  a. 3  E  9  s  3  u o 5  8 7- s  OPC  LF2  SF  Fig. 6.13 Dependence of the 28d flexural strength on the fineness of the filler (w/b = 0.33, filler = 15%).  6.4 MICROFILLER E F F E C T IN TRIPLE-BLENDED COMPOSITE CEMENTS The investigation of mortars described earlier showed that OPC-limestone-silica fume tripleblended cements were advantageous. It was decided to investigate this aspect further in HPC mixtures. The iso-strength curves at various ages over the selected experimental domain are illustrated in Fig. 6.14. The proportion of LF in the triple-blended binder dominated the very  Chapter VI  108  early strength development (Fig. 6.14-a). The 12h compressive strength in a 0% LF binder was around 1 MPa, while it reached 9 MPa in a 10% LF binder. The effect of SF on the 12h strength was less significant. A 10% SF binder could achieve around 4.5 MPa at the same age. Optimal conditions were observed at combined high LF and SF levels. A 10% LF-10% SF triple-blended binder could achieve 12 MPa at 12h; about one order of magnitude higher than a pure OPC.  Kessal et al. [5] obtained similar results though they used fundamentally different materials. A low heat cement having lower C3A and C3S contents and lower fineness was employed with a coarser limestone filler (mean particle size = 3 urn). A melamine-based superplasticizer was selected to reduce the retarding effect. It was possible to develop a high early-strength concrete without increasing the heat of hydration, which is of interest for mass concrete applications.  The Id compressive strength was also significantly influenced by the LF proportion in the tripleblended composite cement (Fig. 6.14-b). The highest Id strength was obtained in a 6 to 10% LF binder with no SF. As the SF proportion increased, the Id strength at high L F proportions became lower, probably because of the low cement content and because much of the hydration of the SF had not occurred yet.  At 3d, the LF had only a limited effect on the strength for SF contents lower than 5%. For higher SF levels, the strength decreased as the LF proportion increased (Fig. 6.14-c). At 7d, the presence of SF was a dominant factor, and the compressive strength increased with increased SF proportions regardless of the LF levels (Fig. 6.14-d). The LF proportion did not seem to have a significant effect on the strength at this age. Optimal strength was obtained using a 4 to 7.5% SF2.5 to 7.5% LF triple-blended binder. Combined high levels of SF and LF were advantageous at this age as opposed to the age of 3d.  The 28d compressive strength increased with higher SF proportions and decreased with higher LF levels. A similar trend was observed at 91d. It is interesting to point out that the 28d strength obtained using an OPC can be outperformed using 20% replacement of the cement by a combination of 10% LF-10% SF. Fig 6.9 shows that a 10% LF2-10% SF binder would have a compressive strength 10% higher than that of a pure OPC binder at 28 d using a pan mixer.  Chapter VI  109  Fig. 6.14 Iso-strength curves at various ages for HPC made with LF2-SF-OPC triple-blended cements. (cont'd)  Chapter VI  110  (e)at28d  (f)at91d Fig. 6.14 (cont'd).  If an OMNI mixer with higher fine-particle dispersive capability is used, the strength of the composite binder can reach 16% higher than the pure OPC. At a lower w/b ratio of 0.25, a 15% LF2-15% SF binder would achieve a 20% higher strength than a pure OPC at 28 d. These kinds of triple-blended cements can achieve increased strength both at early and later ages, while saving energy, raw materials, reducing toxic emissions into the atmosphere, and improving the rheological characteristics of the HPC mixtures.  6.5 MECHANISMS OF MICROFILLER EFFECT ON EARLY AGE PROPERTIES It has been shown above that the presence of very fine calcite in concrete mixtures increases the early age strength development. It is intended in this section to discuss the possible mechanisms that could explain this effect. Some of this discussion is based on experimental results from the literature. Also, since some of the mechanisms have not been directly observed experimentally, part of the discussion will be based on theoretical grounds.  111  Chapter VI  6.5.1 Impermeable C-S-H Layer Theory  To better understand why a limestone microfiller seems to accelerate the early age strength development of cement based materials, it is pertinent to approach this topic from the point of view of the main theories for the early hydration of cement. Due to the complexity of the reactions between cement and water, most research in this field has been carried out on dilute suspensions of alite. Nonetheless, it has been observed that the mechanisms underlying the induction period and the onset of hydration are similar in alite and cement systems.  When cement comes into contact with water, a reaction that is characterized by a short and intense heat release occurs (Fig. 6.15). An induction period which extends up to several hours follows, in which low heat release and hydration activity take place. An acceleration phase involving high hydration activity is then initiated. Subsequent events are of little relevance for the early age behavior.  -.1  .  ...I-I  .-I.,l...l..l,-I.„,l..l  LJ,.J.,I„,I  .1 ,1 l - I . J , I - 1 , . J . J - , I , . J . . I , . I „ . I ,  l„,,1 I , I I  I, I I  Time [hours] Fig. 6.15 Typical isothermal calorimetric curve for the hydration of portland cement.  Stein and Stevels [6] amongst others have suggested that a layer of C-S-H forms on the surface of C S particles in the early stages of hydration. The permeability of this layer is assumed to be 3  sufficiently low to inhibit further hydration, which explains the onset of the induction period.  Chapter VI  112  Later on, it is thought that this layer would convert through a re-crystallization process into a more permeable structure. Thus, the rate of hydration would increase, which explains the initiation of the acceleration phase.  Several theories have been put forward to explain the transformation of the impermeable C-S-H layer into a more permeable one. For instance, Damidot and Nonat [7] believe that this transformation occurs when the lime concentration in the solution reaches a certain level (22 mM), and that the lime concentration is the most crucial parameter in the C3S hydration. On the other hand, Gartner and Jennings [8] support the idea that the conversion of this impermeable layer proceeds via a solid state reaction, but depends essentially on the concentration of CaO and Si0 in the solution. Experimental results have shown that the initial nucleation of CH and the 2  end on the induction period coincide with the maximum Ca** concentration. This observation has led Young et al. [9] amongst others to theorize that the induction period will end when CH starts to grow from the solution. This would occur when the solution is saturated in Ca** and OH' ions. The removal of these ions from the solution by means of CH nucleation would reactivate the C S 3  hydration, thus ending the induction period. There is no consensus for this explanation. Others have found that crystalline CH formed even before the end of the induction period and the Ca** concentration peak [10, 11]. Skalny and Young [12] explained this CH formation previous to the Ca  ++  peak by a possible nucleation of CH in the vicinity of the C S particles where the Ca  ++  3  concentration is locally higher than in the rest of the solution.  6.5.2. Osmotic Membrane Theory  Powers [13] and Double et al. [14] proposed the osmotic membrane theory to explain the occurrence of the induction period. It is believed that water can diffuse through the primary membrane that forms around C3S particles at the very early stages of the hydration of cement, and dissolves calcium and silicate ions. By virtue of their smaller size and better mobility, the calcium ions can diffuse out into the bulk solution, while the silicate ions are restrained by the hydration membrane. This theory is experimentally supported by the high concentration of calcium ions and the very low concentration of silicate ions in the bulk solution measured during the early stages of hydration. This differential diffusion process causes a gradient of concentration between the solution inside versus the solution outside the primary hydration layer. This generates osmotic pressures that would reach, by the end of the induction period, a level  Chapter VI  113  capable of bursting this hydration layer. The previously blocked hydro-silicate solution combines with the calcium ions to produce C-S-H. Thus, the rupture of the protective layer marks the end of the induction period and the onset of the acceleration phase.  Others [10] believe that crystal defects generated in the processing stage of the anhydrous cement manufacture determine to a large extent the subsequent hydration characteristics. These crystal defects constitute active sites for the growth of hydration products. Once the initial hydration nuclei reach a certain critical size, they start to grow rapidly, which marks the end of the induction phase. However, recent environmental electron microscope investigations [15] confirm the previous theories, which argue that a certain layer of hydration products surrounds the C S 3  particles at the early stage of the hydration reaction.  6.5.3 N u c l e a t i o n M e c h a n i s m T h e o r y Three mechanisms are believed to be involved in the process of hydration: a diffusion mechanism through the hydration layer surrounding the anhydrous grains, a phase-boundary interaction, and a nucleation and growth mechanism. It is conceivable that ultrafine particles may act as preferential substrates for the initiation of hydration nuclei. The nucleation and growth process might in fact be a predominant factor in determining the rate of hydration and strength development at early ages. At first it would seem that ultrafine particles provide more sites for the growth of hydration products, thus, increasing the early age strength. Previous results in this chapter showed that limestone microfiller enhanced the early age strength more than silica fume, though the latter is finer and would provide more numerous fine sites for nucleation (Fig. 6.16).  Jiang et al. [16] observed that limestone, titanium dioxide, and barium carbonate fillers accelerated the cement hydration, while quartz and alumina retarded the hydration. They suggested that the filler acceleration effect depends on the number and nature of interparticle contacts achieved in a cement-filler system. Only fillers with an acceleration effect would have coagulation contacts on which nuclei of hydration products will initiate.  Chapter VI  114  Fig. 6.16 Microfiller effect on early age strength of HPC (15% filler, w/b = 0.33).  However, rheological results in Chapter 4, in addition to Zeta-potential measurements by Kjellsen and Lagerblad [17] provide no evidence for such coagulation. On the contrary, the rheology of cement suspensions was improved when ultrafine limestone particles were added to the mixture, which contradicts any extra coagulation of the system due to calcite.  Likewise, Beedle et al. [18] have observed that graphite and cc-alumina did not affect the rate of hydration, while clays and y-alumina had a significant acceleratory effect. Being a surface mechanism, the nucleation will thus depend on the surface characteristics of the filler, namely: its chemical composition, atomic structure and surface morphology. However, calcite does not have a strong ionic nature, and would thus not precipitate ions through particularly high interatomic forces. Rather, it would constitute a preferential substrate for the germination and growth of hydration products, thus, accelerating the hydration process.  6.5.4 Nucleation of C-S-H or Nucleation of C H ?  Although there is abundant literature suggesting the nucleation mechanism as the primary cause for the microfiller acceleration of the early age hydration, there is a certain confusion as to what is nucleating: C-S-H or portlandite or both, and in what sequence?  For instance, Ramachandran and Zhang [19] suggested that calcite not only accelerates the hydration of C S, but that a certain percentage of calcite is consumed before Id in this process. 3  They explained the increased hydration rate by the nucleation of C-S-H around calcite particles, which would incorporate a part of the calcite in some kind of a composite. They supported this idea by Id SEM micrographs illustrating the growth of C-S-H around calcite grains. Kjellsen and  Chapter V I  115  Lagerblad [17] argued that the acceleration of the hydration is initiated during the induction period, before any notable long-range nucleation of C-S-H can be expected. Therefore they questioned the above explanation.  A rational explanation for the accelerating effect of microfillers should: (i) account for the fact that the mechanism responsible for such an acceleration is initiated within the induction period, and (ii) be compatible with the theories of hydration mentioned earlier. During the induction period, the solution is increasing in its concentration of calcium and hydroxyl ions. Calcite does not have a strong ionic character, and would not exhibit particularly strong ionic attractions to precipitate calcium and hydroxyl ions on its surface. But let us assume that calcite has an affinity to grow CH on its surface, or in other words, CH energetically prefers to germinate on calcite as a substrate. Thus, in the presence of calcite, the removal of calcium ions from the solution would start earlier, and would enhance further dissolution of calcium ions. By the same token, there will be a stronger concentration gradient for the migration of calcium ions through the hydration membrane that engulfs the cement particles.  The continued germination of CH on calcite implies an increased dissolution of C3S since the system will tend to restore the equilibrium between the surrounding solution and the hydration membrane. Furthermore, more osmotic pressure would have built up since more silicate ions are now trapped inside the hydration shell. This promotes the conditions for rupture of this shell, and thus, an earlier end of the induction period. By the same token, this would stimulate the appropriate thermodynamic conditions for a re-crystallization of the hydration shell around the C3S particles into a more permeable structure. This also signals an earlier end of the induction period. This explanation seems thus to be compatible with the hydration theories stated earlier. During the acceleration period, calcite may continue to be a preferential site for CH and C-S-H growth, which is compatible with the observations of Ramachandran and Zhang [19] and others.  It is conceivable that calcite can be a preferential substrate for the nucleation and growth of seemingly compatible calcium bearing CH and other hydration products. However, why would other non-carbonate microfillers also have an acceleration effect on the hydration of cement, though not to the same extent? Previous results in this chapter illustrated that silica fume also has a significant acceleration effect. As mentioned earlier, Beedle et al. [18] reported that microfillers of cc-alumina and graphite had no effect on the hydration rate, titania had little effect,  Chapter VI  116  while clays and y-alumina had significant effect. Their explanation for the acceleration of the rate of hydration is compatible with the CH nucleation theory above. Amorphous silica and clays would reduce the calcium ion potential in the solution through the introduction of silicate ions into the system. This is similar to the reduction of this potential by removal of calcium ions during CH germination and growth. Both mechanisms catalyze further dissolution of C3S because the system will tend to shift towards equilibrium. Hence, the acceleration of cement hydration due to some non-carbonate microfillers is compatible with the explanation based on the nucleation and growth of CH discussed earlier.  6.5.5 Effect of Calcium Ions Dissolved from Calcite  It might also be that calcite would dissolve in the limestone-cement-water system, providing extra Ca  ++  ions, thus modifying the kinetics of the hydration reactions. Accounting for the low  solubility of calcite in alkaline systems, this theory does not seem to be thermodynamically plausible. To investigate this aspect further, HPC mixtures were prepared with 5%, 10%, 15%, and 20% of limestone filler and hydrated lime powder. The hydrated lime is much more soluble in water than limestone powder and would provide many more Ca** ions in the solution, thus further increasing the rate of hydration. Fig. 6.17 shows the compressive strength ratio of limestone powder to hydrated lime mixtures at 12h. The limestone powder slightly outperformed the hydrated lime at all replacement levels. However, since the limestone powder had higher fineness, and that the affinity to the CH growth of the two microfillers might be different, the real contribution from the calcium ions dissolved from the calcite remains uncertain.  1.1 1.08 -  |  1.06 4-  £ .2  1.04 -  | o  102-  O  W  1 --  0  5  10  15  20  Replacement rate [%]  Fig. 6.17 Compressive strength ratio at 12 h of limestone to hydrated lime HPC mixtures.  Chapter VI  117  6.5.6 Effect of Reaction between Calcite and C A 3  The potential reaction between calcite and calcium aluminate in the presence of water to produce calcium carboaluminates (C3A.CaCO3.HH2O) was discussed in Section 3.3. The question is: could this reaction be responsible for the reduction of the induction period and the increase of the rate of the early hydration? Or is it possible that the formation of carboaluminates enhances the early age strength in the same way sulfoaluminates do?  The work reported in Section 3.2 regarding the attempt to substitute limestone for gypsum as a set regulator rather supports the idea that this reaction would form carboaluminates around the C A grains, thus preventing their further hydration. This is analogous to the set control action of 3  gypsum through the formation of ettringite. Also, based on the observation that calcite accelerates the hydration of pure C S (no potential formation of carboaluminates), the hypothesis 3  that the reaction between calcite and calcium aluminate is responsible for the acceleration of the hydration reactions is unlikely.  Using X-ray diffraction of limestone filler cement pastes, Escadeillas [6] observed that the percentage of calcite started to decrease after the first few hours of hydration. However, carboaluminates were not detected before 7d, either because they are too small to be detected by X-ray diffraction, or because the crystallization of carboaluminates is only possible at high carbonate levels in the solution. It was claimed that the transformation of aluminates to carboaluminates is practically complete at around 9 months. Fig. 6.18 shows the calcium carbonate variation up to the age of 180d in 5% and 25% limestone filler cement pastes. Since most of the consumption of CaC0 occurred at later ages, it is unlikely that the production of 3  calcium carboaluminates is responsible for the very early age strength enhancement.  Figs. 6.19 and 6.20 compare the compressive strength and modulus of elasticity values measured on HPC mixtures made with various proportions of limestone or hydrated lime powder. At early age, hydrated lime and limestone provided comparable effect on the rate of hydration and enhanced the early strength development. At later ages, the limestone microfiller led to higher strength and modulus values. This difference tended to be more significant the higher the replacement rate. It might be that the higher fineness of the limestone filler yielded a better filler  Chapter VI  118  effect. But it is also conceivable that carboaluminates could have enhanced the limetone effect at later ages.  25  5% Filler  25% Filler  Fig. 6.18  Illustration of consumption of CaCO"3 in the hydration of cement paste [20].  6.6 CONCLUSIONS Microfillers can have significant effects on the mechanical properties of HPC mixtures. A better understanding of the mechanisms underlying their action could lead to improvements in the design of HPC mixtures. The fineness of the microfiller, the degree of dispersion of the ultrafine particles, and the homogenization of the system greatly influence the mechanical properties. Limestone microfiller enhanced the early age strength, and did not significantly decrease the long-term mechanical strength up to 15% replacement of cement. On the other hand, silica fume increased the long-term strength significantly. Using these two microfillers in triple-blended cements provided binders with both higher early and long-term strengths. A 30% replacement of cement by a limestone-silica fume combination achieved a 25% increase in the 28d compressive strength. As shown in Chapter 5, these triple-blended mixtures also provided improved rheological properties.  Chapter VI  1  s:  100  cn c  90  0)  fm +" (Ji  „  80  > Q.  70  U) '  0)  60  Q.  E o O  50 40  — « — 5 % Lime 7-  110  - H - r - 5 % LF2  1 20  30  Age [d]  Fig. 6.19 Comparison of the effect of limestone and hydrated lime partial replacement of cement on compressive strength (w/b = 0.33).  Fig. 6.20 Comparison of the effect of limestone and hydrated lime partial replacement of cement on modulus of elasticity (w/b = 0.33).  Chapter VI  120  It was shown that the filler effect depends on the w/b ratio. The lower the w/b ratio, the more the ultrafine particles are beneficial. Replacing as much as 15% of the cement with a noncementitious filler at a w/b ratio of 0.25 provided increased strength. The microfiller effect is addressed from a microstructural and particle packing point of view in Chapters 7 and 8, respectively.  At very early ages, it appears that some microfillers, such as limestone, present energetically preferential substrates for the germination and growth of calcium hydroxide. The removal of calcium ions from the solution would catalyze the dissolution of C3S in an attempt to achieve equilibrium between the hydration layer that engulfs these cement particles and the surrounding solution. Ultimately, this would initiate the re-crystallization of this protective membrane into a more permeable structure, which signals an earlier end of the induction period. The filler surface continues to be a preferred substrate for the growth of hydration products, which increases the early age strength of limestone filler mixtures. Other non-carbonate fillers can also stimulate the early rate of hydration when they can reduce the calcium potential in the solution. This also enhances the dissolution of C S in order to move towards equilibrium. The microfiller effect on 3  later age properties will be discussed later in the light of the next chapters.  REFERNCES  [1]  Gardner, N.J. and Zhao, J.-W. [1991], "Mechanical properties of concrete for calculating long term deformations" Proceedings, Second Canadian Symposium on Cement and Concrete, Vancouver, pp. 150-159.  [2]  Monteiro, P.J., Maso, J.C. and Ollivier, J.P. [1985], "The aggregate mortar interface", Cement and Concrete research, Vol. 15, No. 6, pp. 953-958.  [3]  Lutz, M.P. and Monteiro, P.J.M. [1995], "Effect of the transition zone on the bulk modulus of concrete", MRS Sym. Proceedings, Microstructure of Cement-Based Systems/ Bonding and Interfaces in Cementitious Materials, Vol. 370, pp. 413-418.  [4]  Escadeillas, G. [1988], "Les ciments aux fillers calcaires: Contribution a leur optimisation par l'etude des proprietes mecaniques et physiques des betons fillerises", Doctoral Thesis in Civil Engineering, Universite Paul-Sabatier, Toulouse, France, 143 p.  [5]  Kessal, M., Edwards-Lajnef, M., Tagnit-Hamou, A. and Ai'tcin, P.-C, "L'optimization de la resistance a court terme des betons fabriques avec un ciment de Type 20M", Accepted for publication, Canadian Journal of Civil Engineering.  Chapter V I  [6]  121  Stein, H.N. and Stevels, J.M. [1964], "Influence of silica on the hydration of 3CaOSi0 ", Journal of Applied Chemistry, Vol. 14, August, 1964, pp. 338-346. 2  [7]  Damidot, D. and Nonat, A. [1991], "Investigations of the C S hydration process during the first hours of hydration", Proceedings: International RILEM Workshop on Hydration and Setting of Cements", Dijon, pp. 23-34.  [8]  Gartner, E.M. and Jennings, H.M. [1987], "Thermodynamics of calcium silicate hydrates and their solutions", Journal of the American Ceramic Society, Vol. 70, No. 10, pp. 743-749.  [9]  Young, J.F., Tong, H.S. and Berger, R.L. [1977], "Compositions of solutions in contact with hydrating tricalcium silicate pastes", Journal of American Ceramic Society, Vol. 60, No. 5-6, pp. 193-198.  [10]  Fierens, P. and Verhagen, J.P. [1976], "Hydration of tricalcium silicate in paste kinetics of calcium ions dissolution in the aqueous phase", Cement and Concrete Research, Vol. 6, No. 3, pp. 337-342.  [11]  Siegers, P.A. and Rouxhet, P.G. [1977], "The hydration of tricalcium silicate: calcium concentration and portlandite formation", Cement and Concrete Research, Vol. 7, No. 1, pp. 31-38.  [12]  Skalny, J. and Young, J.F. [1980], "Mechanisms of portland cement hydration", Proc: 7 International Congress on the Chemistry of Cement, Paris, Vol. 1, 45 pp.  3  th  [13]  Powers, T.C. [1961], "Some physical aspects of the hydration of portland cement", Journal of the Portland Cement Association, Vol. 3, No. 1, pp. 47-56.  [14]  Double, D.D., Hellawell, A. and Perry, S.J. [1978], "The hydration of portland cement", Proc. :R. Soc. Lond, Vol. A359, pp. 435-451.  [15]  Sujata, K., Bergstrom, T.B. and Jennings, H.M. [1991], "Preliminary studies of wet cement pastes by an environmental scanning electron microscope", Microbeam Analysis, pp. 195-198.  [16]  Jiang, S.P., Mutin, J.C. and Nonat, A. [1993], "Effect of fillers (fine particles) on the kinetics of cement hydration, Proc: 3 International Symposium on Cement and Concrete, Beijing, Vol. 3, pp. 126-131. rd  [17]  Kjellsen, K.O. and Lagerblad, B. [1995], "Influence of natural minerals in the filler fraction on hydration and properties of mortars", Swedish Cement and Concrete Research Institute, Report S-100 44, Stockholm, 41 pp.  [18]  Beedle, S.S., Groves, G.W. and Rodger, S.A. [1989], "The effect of fine pozzolanic and other particles on the hydration of C S", Advances in Cement Research, Vol. 2, No. 5, pp. 126-131. 3  [19]  Ramachandran, V.S. and Zhang, C-M. [1986], "Dependence of fineness of calcium carbonate on the hydration behaviour of tricalcium silicate", Durability of Building  Chapter VI  122  Materials, Vol. 4, pp. 45-66. [20]  Parker, A.P. and Cory, H.P. [1991], "The early hydration of limestone filled cements", Blended Cements in Construction, R.N. Swamy, ed., Elsevier Science Publishers, pp. 107-124.  Chapter VII  123 Chapter VII MICROFILLER EFFECT ON MICROSTRUCTURE OF HIGH-PERFORMANCE CONCRETE  7.1 INTRODUCTION The relationship between microstructure and engineering properties of cement-based materials is as yet only imperfectly understood, partly because there are no reliable quantitative measurements of the microstructure. Although enough is known to define which alterations are required to improve certain properties, we still do not have truly predictive capabilities. When it comes to designing cement blends with environmental, economic and rheological advantages, we really ought to try to provide a microstructural basis for the mechanisms underlying the microfiller effect. Thus, estimations of their effects on engineering properties can be made. In this chapter, an attempt is made to elucidate the microfiller effect on the microstructure of highperformance cement based materials. The quantitative image analysis of digitized backscattered electron micrographs was selected as the primary tool for this investigation.  7.2 BACKGROUND Secondary electrons (SE) are produced by inelastic (electron-electron) collisions between the bombarding electron beam and the specimen electrons. The released SEs have significantly lower energies than the primary electrons (typically below 50 eV), they follow a curved trajectory, and they are affected by electrostatic collection fields. Thus, SE detectors are often placed at a 90° angle relative to the optical axis. On the other hand, backscattered electrons (BSEs) are caused by elastic (electron-nucleus) collisions. A primary electron strikes the nucleus and rebounds with little loss of energy (typically around 20% below the primary electron energy), and follows a linear trajectory. Hence the orientation of SE detectors is not ideal for BSE collection, and the design of BSE detectors is based on the cosine distribution of BSEs around the primary beam. Fig. 7.1 illustrates the concept of SE and BSE occurrence and detection.  Chapter VII  124  Incident electron beam  Fig 7.1 Occurrence and detection of secondary and backscattered electrons.  The secondary electron emission is confined to an interaction volume near the beam impact area, allowing images to be obtained with high resolution. The three dimensional appearance of the images is obtained due to the contrast provided by the shadow relief effect of SEs. At low magnification, BSE images can give a better impression of the surface topography than SE images owing to the sharper shadow effects obtained when a BSE detector is employed. When ultrafine polishing of specimens is used, the topographic effect in BSE micrographs is reduced, and contrast in images is primarily due to the local difference in the average atomic number (Z) of the observed area. The probability of a backscattering event is higher the higher Z .  If I SE is the current resulting from the BSEs and I is the incident electron-probe current, the B  P  backscattering coefficient, r\, can be defined by equation (7.1). The higher n is the brighter the feature in a BSE micrograph:  Chapter VII  rj =  125  (7.D  The backscattering coefficient can be expressed in terms of the characteristics of the material such as its atomic number, Z, and atomic weight, A. For a thin film of thickness t and density p, r) is proportional to the number of atoms per unit area N pt/A, where N is Avogadro's number, N A  A  is the number of atoms per unit volume, doVdQ is the differential cross-section, 9 is the backscattering angle > nil, and c(E) is the backscattering constant:  r? = ^ )lS 2*sm0d0 = ™ "« A- f 4 (4ns ) AE 4Z2  2  Pt =c(E)NZ pt  (7.2)  2  0  Equation (7.2) illustrates how r\ increases with increasing atomic number. For a pure element, the backscattering coefficient (at 20 keV) has been experimentally expressed as a function of the atomic number Z [1]:  n = -0.0254 + 0.016Z - 1.86xlO" Z + 8.3xl0" Z 4  2  7  3  (7.3)  When the target material is heterogeneous, a simple rule of mixtures based on weight fractions of the various components applies [1]:  1 = SC,-^-  (7-4)  In fact, SEs are also dependent on the atomic number but to a smaller and less predictable extent than BSEs. When the detector is sensitive to BSEs only, such as a solid state detector, the contribution of SEs to the image is ignored. The contrast is then known as compositional (or atomic number) contrast, and can be defined using equation (7.5):  (7.5)  Chapter V I I  126  where: T] and n are the backscattering coefficients of the phases having the higher and lower 2  x  atomic numbers, respectively. Table 7.1 shows the mean atomic numbers corresponding to the various components in hydrated cement paste and the gray level associated with each component in a BSE image. Table 2 shows the theoretical contrast values for adjacent pairs of various components of hydrated cement paste calculated based on equations (7.4) and (7.5). The higher C is, the easier it is to segment the two adjacent phases in a BSE image.  7.3 S P E C I M E N P R E P A R A T I O N  Cement pastes and concretes having w/b ratios of 0.33 and 0.25 were made with pure OPC (Blaine fineness = 345 m /kg), and OPC having 15% replacement (by volume) of cement by 2  silica fume (BET surface area = 18000 m /kg) and limestone filler (BET surface area = 10000 2  m /kg), respectively. Cylindrical specimens (40x100 mm for cement paste and 75x150 mm for 2  concrete) were cast and cured in a curing room for 24 hours, then demolded and cured in lime saturated water. Cylindrical specimens for compressive strength measurements at different ages were also prepared.  Two centimeter square samples were obtained at Id and 28d from the middle third of the cylinders using a diamond saw, then oven dried under vacuum at 70 °C for 24 hours (drying at excessive temperatures would result in microcracks having triple-junction nodes at angles of about 120° (Fig. 7.2), a pattern considered typical of drying shrinkage). Samples were then vacuum impregnated using an ultra-low viscosity epoxy. After hardening, extra epoxy on the surface of specimens was removed using a precision low speed diamond-saw. The specimens were then ground using, successively, 180, 240, and 600 grit silicon carbide papers. A 10 pm particle size diamond pad was then used for finer polishing, followed by 6 pm and 1 pm diamond suspensions on suitable polishing clothes. No hydraulic lubricants were used during cutting and polishing of the samples. The specimens were then vacuum dried and coated with a gold conductive layer before SEM observation.  Chapter VII  127  T A B L E 7.1- Mean atomic numbers, backscattering coefficients, and gray levels usually observed for the major components of hydrated cement paste  Phase  z  cs 3  15.06  0.1716  bright  CS 2  14.56  0.1662  bright  CA  14.34  0.1639  bright  C AF  16.65  0.186  bright  CH (calcium hydroxide)  14.3  0.1618  light gray  CSHo.s (calcium silicate hydrate)  12.39  0.1413  gray  Ci. SH (calcium silicate hydrate)  12.78  0.1455  gray  C 3 A S 3 H 3 2 (ettringite)  10.76  0.1233  gray  CASH  11.66  0.1328  gray  3  4  5  2  3  1 2  (monosulfate)  Gray level  Voids  black  T A B L E 7.2- Contrast values for pairs of adjacent components of hydrated cement paste Phase  3  CS 2  CA 3  C AF 4  CH  S  2  C A S 3H32 3  C AS H 3  CS  0  0.0315  0.0449  0.0774  0  0.0138 0  3  cs  C,. SH  cs  n  2  CA 3  C AF  CH  CSH0.5  C1.5SH2  C3A S 3H32 C A S H  0.0571  0.1766  0.1521  0.2815  0.226  0.1065  0.0265  0.1498  0.1245  0.2581  0.201  0.1188  0.0128  0.1379  0.1123  0.2477  0.1897  0  0.1301  0.2403  0.2177  0.3371  0.286  0  0.1267  0.1007  0.2379  0.1792  0  0.0289  0.1274  0.0602  0  0.1526  0.0873  0  0.0715  4  3  0  1 2  Chapter VII  128  7.4 IMAGE ACQUISITION STRATEGIES  Specimens were examined using a Hitachi S-2300 SEM equipped with a GW backscatter detector and the Quartz PCI™ digital imaging system (Fig. 7.3). The advantage of this system is that the analog signal usually sent to the CRT of the SEM is digitized instantly and presented in real time as a raster display on a large screen computer. This allows one to focus almost exclusively on the microscopic features being displayed rather than on manipulations of the SEM.  Fig. 7.2 Illustration of micro-cracking due to severe oven drying (60x).  A working distance of 22 mm and a 25 keV voltage were used. Micrographs were acquired at 1024x839 pixels and 256 gray levels (0 for darkest and 255 for brightest). A constant magnification of 500x was selected. This was a tradeoff between resolution and a representative sample. Consistent settings for the contrast and brightness were maintained for all of the data acquired. At least six randomly selected fields were examined for each specimen and corresponding digital micrographs were stored for quantitative analysis. Occasionally, X-ray dot maps were used in conjunction with BSE images to identify some features. For consistency, all specimen preparation, image acquisition, and measurements were carried out by the author.  Chapter VII  129  Fig. 7.3 Illustration of the image acquisition system.  7.5 Q U A N T I F Y I N G T H E M I C R O S T R U C T U R E  The basis of BSE image analysis, the specimen preparation techniques, the experimental methods, the statistical aspects of the method, and its merits and limitations have been discussed elsewhere [2]. In this work, the boundary threshold levels for the various phases of the microstructure were imposed by the operator. Fig. 7.4 illustrates an example of phase segmentation in a hydrated cement paste. Porosity induced by entrained air bubbles or eventual shrinkage microcracks has not been considered in calculating the porosity of specimens.  Chapter VII  130  Two software packages, namely Adobe Photoshop and NTH Image were jointly used for the treatment and analysis of the S E M digital images. For cement pastes, the entire image was analyzed at once. For concretes, a series of 10 um wide bands equidistant from the aggregate surface were individually analyzed. Therefore, the amounts of the various microstructural phases can be represented as a function of the distance from the aggregate-paste interface. The relative area of each component in each band represents the averaged value obtained on six or more  Chapter VII  131  equally distant bands in different random fields in the specimen. In addition, fields representing the bulk cement paste in concrete (at least 50 um away from the interfaces) were analyzed.  7.6 MICROFILLER EFFECT ON EARLY AGE MICROSTRUCTURE Fig. 7.5 illustrates the relative areas of the various microstructural phases in BSE micrographs at Id for cement pastes (w/b = 0.33) made with OPC, OPC + 15% LF, and OPC + 15% SF. It may be seen that the SF and LF pastes had smaller areas of unhydrated cement than had the OPC. The porosities in the SF and LF pastes were lower than in the OPC, but the CH content was higher in the LF pastes. At the same time, particularly at 12 hours but even at 24 hours, concretes made with the SF and LF pastes had higher compressive strengths than concretes made with the OPC pastes, as shown in Fig. 7.6. Clearly, microfillers enhanced the early age mechanical performance of high-strength concrete. In addition to the initial improved particle packing due to the ultrafine particles, an acceleration of the hydration reaction is indicated by the microstructural analysis. As suggested in Chapter 6, the mechanism involved might be an increase in the precipitation of the hydration products due to the fine particles, which offer nucleation sites for the growth of CH and C-S-H gel solution [3-4]. As illustrated in Fig. 7.5, the CH content at early ages was higher in the limestone filler cement pastes. The CH content was apparently not higher in silica fume mixtures compared to OPC mixtures, perhaps because the CH particles formed in the presence of silica fume were too fine to be detected by BSE image analysis. There are disagreements amongst researchers regarding this possibility. For instance, the absence of calcium hydroxide peaks in X-ray diffractometers, commonly explained by the pozzolanic activity of silica fume, is seen by other researchers as a possible consequence of the formation of a poorly crystalline CH in the presence of silica fume. A similar explanation is given for the scanning electron micrographs, in which the CH crystals may not have enough contrast for resolution by SEM. Some reported differential thermogravimetry data supports the presence of CH in silica fume grouts, although it was not detected by X-ray diffraction and SEM [5].  Chapter VII  132  Fig. 7.5 Microfiller effect on relative area of microstructural phases in cement pastes at Id (w/b = 0.33).  15% S F  Fig. 7.6 Microfiller effect on early age compressive strength.  Chapter VII  133  7.7 MICROSTRUCTURE OF THE INTERFACE 7.7.1 Anhydrous cement phase Fig. 7.7 illustrates the difference observed between the interfacial transition zone and the bulk cement paste. Micrograph (a) shows a typical view of the packing of cement grains between sand and coarse aggregate particles. Point B in micrograph (a) was magnified in micrograph (b) and shows a cement paste between a coarse aggregate and a sand particle. The micrograph indicates that during mixing, the anhydrous cement grains tend to move out of the high shear regions adjacent to the coarse aggregate surfaces, thus a low anhydrous cement content is usually observed around interfaces. This effect causes a locally higher w/b ratio, and thus, a higher porosity in the vicinity of the aggregates. Point D in micrograph (a) was magnified in micrograph (c), which illustrates a gradient of anhydrous cement content around a coarse aggregate. Micrograph (d), which is a magnification of point C in micrograph (a), shows that even in the bulk cement paste, the presence of a small but elongated sand particle can locally disrupt the microstructure. Fig. 7.8 shows that the relative area of anhydrous cement grains decreases significantly towards the aggregate-paste interface. This wall effect (Fig. 7.9) is due to the fact that large cement grains, because of physical constraints, migrate away from the high-shear rate locations represented by the aggregate surfaces during mixing. This effect seemed to be somewhat reduced by the fine fillers, but it should be remembered that particles smaller than 0.5 pm cannot be identified on the BSE micrographs. The slightly improved packing of cement grains near the interface in the presence of microfillers is probably due to a rheological effect and a reduction of the interference between larger grains. It is mainly the particle size distribution of the cement that defines the packing of anhydrous cement grains in the interfacial zone. However, the predominant presence of fine particles near the interface (Fig. 7.9) suggests that ultrafine particles can achieve a much better particle packing around the aggregate surfaces in concrete. This can be confirmed by measuring the porosity gradient around interfaces.  Chapter VII  134  (c)  (d)  Fig. 7.7 Illustration of (a) a general view of HPC at low magnification, (b) porosity of the paste aggregate transition zone, (c) gradient of microstructure at the interface, and (d) a general view of the bulk cement paste.  The anhydrous cement content in the bulk cement matrix at Id was about 19% lower in LF concrete than in OPC concrete. This supports earlier results obtained on cement pastes, which suggest an acceleration of the hydration reaction in the presence of LF. At 28d, the anhydrous cement content had decreased significantly. Only the cores of large cement grains remained, and large hydration rims could be observed around them. It was also observed that the hydration sometimes occurred only in selective areas of the cement grains. Due to the initial presence of only fine cement grains in the vicinity of aggregates and the locally higher w/c ratio, the unhydrated phase in this area at 28d was very low.  Chapter VII  135  •  • • * -•—OPC -«—15% LF2 -A—15% S F  0  20  Distance from interface [nm]  40  CD  O  -a  5  c  •  0 3 1  Fig. 7.8 Microfiller effect on the relative area of anhydrous grains in a 50 um transition zone, in the bulk cement paste matrix in concrete, and in an equivalent cement paste at 1 d (w/b = 0.33).  Fig. 7.9 Illustration of the wall effect in HPC (OPC concrete, w/c = 0.33, magnification = lOOx). The gray level has been inverted and contrast and brightness adjusted to highlight unhydrated cement grains; A indicates the wall effect due to aggregates.  Chapter VII  136  7.7.2 Porosity Fig. 7.10 illustrates the porosity gradient around aggregates in the three different concretes under investigation. Generally, the porosity increased in the vicinity of aggregates. The initial packing of cement grains at the interface may have contributed to this trend. Less anhydrous material was available to react and thus fill the voids in this area. In addition, the fine anhydrous grains that are present at the interface would have hydrated quickly, leaving hydration shells (Hadley grains) that increased the porosity. Due to differences in hardness between the aggregate and the paste, a shadow effect could result from the erosion of the paste adjacent to the aggregate during polishing. This would artificially appear as porosity. However, careful polishing greatly reduced this artifact in the present investigation.  The porosity gradient in the interfacial zone was highest in OPC, followed by LF and SF pastes, respectively. This gradient was most significant in the first 30 pm layer adjacent to the aggregate surface, and was maintained even at 28d when the hydration was at a more advanced stage. SF was most efficient in densifying the interfacial zone. LF, due to its high fineness, had a similar effect, though lower than that of SF. Being a ground product, LF has orthorhombic particles, which would not allow the optimal packing achieved with the smoother, finer, spherical SF particles.  Fig. 7.10 Microfiller effect on the relative area of porosity in a 50 pm transition zone, in the bulk cement matrix in concrete, and in an equivalent cement paste (w/b = 0.33, 28d).  Chapter VII  137  7.7.3  Calcium hydroxide  The  relative area of calcium hydroxide at 28d for a 50 um transition zone adjacent to the  aggregate, for the bulk cement matrix in concrete, and for a corresponding cement paste is shown in Fig. 7.11. In spite of the porosity gradient near the interface, a higher CH  content was not  observed in this zone. This is in contrast to the results obtained from the analysis of fracture surfaces [6], which may indicate that fracture surfaces are not necessarily representative of the true physical interface, and that weak planes rich in CH might instead be exposed. The hydration reactions proceed faster in the transition zone compared to the bulk paste because of the finer cement grains which are present in this area and the locally higher w/b ratio. It might be that the early hydration of fine particles in the transition zone saturates the solution with ions and fills the voids with low-density hydration products which inhibit further growth of CH.  Previous work has shown that limestone fdlers enhance the formation of CH  at early ages [3].  Larger CH regions were more unevenly distributed throughout the LF paste, but at later ages the amount of CH in LF pastes was comparable to that in OPC  pastes. However, in [3] much coarser  fillers were used than the one used in this study. The lowest fineness of LF blended cements was about 578 m /kg, whereas the overall surface area of the 15% LF-cement blend used in this work 2  was about 1800 m /kg. In the earlier study, the fillers probably accelerated the hydration at early 2  ages, increasing the precipitation of CH  around unevenly distributed coarse filler particles (no  superplasticizer was used, w/b = 0.5). Therefore, larger and less evenly distributed CH particles were produced. In the present work, much finer LF particles were more evenly distributed in the matrix due to the use of a superplasticizer. More CH compared to an OPC  paste was measured at  early ages due to the acceleration of the hydration reaction. At later ages however, less massive CH was measured in LF cement pastes, perhaps due to a refinement of the hydration products by the ultrafine limestone particles.  In SF cement pastes and concretes, CH areas were finer, more difficult to tally, and more evenly distributed. This is probably due to the combined effect of the pozzolanic reaction that consumes CH, and a refinement effect of the hydration products due to their nucleation around very fine SF particles. However, there is disagreement amongst researchers regarding this . 1  This has been discussed earlier in section 7.6.  Chapter VII  138  12  Fig 7.11 Microfiller effect on the relative area of CH in a 50 um transition zone, in the bulk cement matrix in concrete, and in an equivalent cement paste (w/b = 0.33, 28d).  7.8 FILLER E F F E C T IN CONCRETE VERSUS CEMENT PASTE  When deflocculated by a superplasticizer, ultrafine particles can fit into gaps between relatively coarser cement grains, improving the particle packing of the anhydrous products. This is, according to Bache [7], the main physical effect of silica fume. This effect can be expected to occur in cement paste and concrete alike. However, several studies have suggested that silica fume does not enhance the strength of cement pastes as much as it does of concrete [8-11]. The reason might be that the shearing effect of aggregates during mixing enhances the distribution of fine particles in concrete. However, cement pastes extracted from concrete also did not seem to show the strength increase usually observed in silica fume concrete [11]. It was argued that the enhancement of the paste-aggregate bond via the densification of the interfacial zone is responsible for the strength improvements in silica fume concretes. In this regard, the physical filler effect of silica fume seemed to be at least as important as the pozzolanic effect, since an inert filler of similar physical properties (carbon black) seemed to have a comparable influence. Others, however, firmly support the contradictory opinion that the strength increase in silica fume concrete is due to an increase in the strength of the cement paste, and base this opinion on both experimental data and finite element modeling [12-14].  Chapter V I I  139  In the above controversy, an important aspect seems to have been overlooked. The relative fraction of cement paste in a cement paste sample (100%) is different from that in a concrete sample (25 to 35%). The process of hydration is accompanied by a chemical contraction (le Chatelier contraction, about 9%), which is proportional to the heat of hydration. Thus, a cement paste sample would develop higher internal stresses, coupled with the absence of the restraining effect of the aggregates. Added to this is the shrinkage due to a self-desiccation mechanism when the hydration reactions exhaust the capillary water. The latter is much more significant in a noncured silica fume cement paste.  Opposed to this is another effect ignored in the above controversy: the crystal growth mechanism. During the hydration reaction crystals form and grow, inducing a crystallization pressure. This is similar to the mechanisms of ettringite and portlandite in expansive cements. This pressure would cause shearing stresses at interfaces, which are proportional to the total volume of crystals and their sizes. In silica fume systems, microstructural evidence shows the crystalline phase to be both lower in volume and finer than in OPC systems.  Another aspect of the above controversy is to consider the wall effect between cement grains and silica fume particles in the same way that we consider the wall effect between coarse aggregates and cement grains. It is the particle packing of the whole system and the total porosity that matter for strength. The difference is that larger pores affect the strength more than the finer ones. It is such a global view that led to the production of the 200-800 MPa reactive powder concretes (RPC) [15]. In this system, the largest particle is about 400 um. The above mechanism of large interfaces is therefore absent. It is by achieving an optimal packing of the system, using various proportions and sizes of grains, and by applying a pressure that compensates for the chemical shrinkage mentioned earlier, that such materials are achieved.  Furthermore, improved homogenization and packing of a particulate system, even without the addition of fines, increases the strength significantly. Fig. 6.9 showed compressive strength results for HPC mixtures made using a regular pan mixer versus an Omni mixer. The mixtures made using the much more efficient Omni mixer had higher strengths.  The ideal situation for the development of strength is, first, to have an optimal particle packing. Then, the hydration reaction is needed mostly to bind the various discrete grains together, and not to fill the large porosity introduced by a poor packing. A very low water content would cause  Chapter VII  140  only the surfaces of the cement grains to hydrate to provide this binder. Denser inner hydration products and hydration rims around cement grains could form, instead of the porous C-S-H usually observed in ordinary systems. The limited hydration reduces the development of internal stresses due to shrinkage and crystal growth (w/b ratio in RPC is typically around 0.15). In this context the controversy about the filler effect in cement paste versus the aggregate-paste interface is irrelevant, since the particle packing of the system is viewed as a whole.  As an illustration of this idea, Fig. 6.8 compared the filler effect in 0.33 and 0.25 w/b ratio HPC. Although the filler improved the aggregate-paste interface in the 0.33 w/b mixtures (Fig 7.9 & 7.10), it did not improve the strength. In the 0.25 w/b mixtures, the non-pozzolanic filler also improved the interfacial properties, but in this case, it also improved the strength. The lower hydration rate in the latter case combined with the improved packing due to the filler was positive. In the former case, a higher hydration rate coupled with the higher initial porosity did not allow the non-pozzolanic filler to compensate for the filling action of the hydration products of a similar volume of cement. Thus 15% of coarser but reactive cement grains were more efficient than 15% of ultrafine but nearly inert filler.  7.9 CONCLUSIONS  The positive effects of microfillers in HPC start at very early ages, before the onset of the pozzolanic reaction. Ultrafine particles improve the homogeneity of the material through a rheological effect, and enhance the particle packing of the system. Early age strengths are increased significantly through a possible nucleation mechanism, which in the case of limestone reached one order of magnitude higher than OPC concrete at 12 hours. This is reflected in the microstructural analysis at Id through a higher CH content, fewer unhydrated grains, and lower porosity in LF systems.  A porosity gradient at the paste-aggregate interface was observed even at 28d in HPC. The welldocumented increase of CH towards the interface measured on fracture surfaces was not, however, observed in polished sections. This suggests that fracture surfaces may not be representative of interfacial properties, or that the CH crystals formed are mostly finer than 0.5 pm and thus, are non-detectable by BSE image analysis.  Chapter VII  141  Densifying the interface alone would not guarantee an increase in strength. For instance, adding a very fine inert fdler would densify the interface via a reduction of the wall effect and the interference between larger particles. However, it would increase the strength only if its filling action outperformed that of an equivalent volume of cement and the hydration products it yielded, and did not cause more porous hydration products to be produced. Therefore, in very low w/b systems, where cement grains hydrate only partially on the surface, an optimal packing of the system using inert fillers could save cement and increase the strength at the same time. The continued controversy regarding the effect of silica fume in cement paste versus concrete overlooks the internal stresses due to shrinkage and crystal growth, which act differently in a cement paste sample compared to a concrete sample. This controversy is less relevant when we look at the cement-aggregate wall effect as a larger scale version of the silica fume-cement wall effect. The aggregate-paste interface is only a special case of the global concern of achieving optimal particle packing of the system. There is an increasing belief that by achieving denser interfaces, a composite material can be obtained in which the strength of the aggregates is mobilized to outperform cement pastes. However, by considering the particle packing of the system as a whole, the large interfaces concept would, ideally, be eliminated altogether. An optimal packing of fine grains could offer a more homogeneous system having strength values that only the very best aggregates could approach.  7.10 REFERENCES [1]  Heinrich, K.F.J. [1966], "X-ray optics and microanalysis," 4* Int. Congress on X-Ray Optics and Microanalysis, Hermann, Paris, p. 1509.  [2]  Nehdi, M. and Mindess, S. [1997], "Quantitative microstructural analysis of cement based materials using backscattered electron microscopy", Proceedings: International Conference on Engineering Materials, A. Al-Manaseer, S. Nagataki, and R.C. Joshi, eds, CSCE, Ottawa, June 8-11, Vol. II, pp. 467-484.  [3]  Barker, A.P. and Cory, H. [1991], " The early hydration of limestone filled cements", Blended Cements in Construction, Swamy, R.N. (ed.), University of Sheffield, pp. 107-124.  [4]  Husson, S., Gulhot, B. and Pera, J. [1992], "Influence of different fillers on the hydration of C S", 9* Int. Cong. Chem. Cem., New Delhi, Vol. IV, III-A.013, pp. 83-89. 3  [5]  Khayat, K.H. and Ai'tcin, P.C. [1992], "Silica fume in concrete: an overview", Proceedings, 4* International Conference on Fly Ash, Silica Fume and Natural Pozzolans in Concrete, ACI SP-132, Vol. II, Istanbul, Turkey, pp. 835-87.  Chapter VII  [6]  142  Grandet, J. and Ollivier, J.P. [1980], "Nouvelle methode d'etude des interfaces ciment-granulats", 7 Int. Cong. Chem. Cem., Paris, Vol. VII, pp. 85-89. th  [7]  Bache, H.H. [1981], "Densified cement/Ultra fine particle based materials", presented at 2 International Conference on Superplasticizers in Concrete, Ottawa. nd  [8]  Cheng-Yi, H. and Feldman, R.F. [1985], "Influence of silica fume on the microstructural development in cement mortars", Cement and Concrete Research, Vol. 15, No. 2, pp. 285-294.  [9]  Goldman, A. and Bentur. A. [1994], "Properties of cementetious systems containing silica fume or non-reactive microfillers", Adv. Cem. Bas. Mat., Vol. 1, pp. 209-215.  [10]  Rosenberg, A.M. and J.M. Gaidis [1989], "New mineral admixture for high-strength concrete", Concrete International: Design & Construction, Vol. 12, No. 8, p. 31-36.  [11]  Goldman, A. and Bentur. A. [1993], "The influence of microfillers on enhancement of concrete strength", Cement and Concrete Research, Vol. 23, No. 4, pp. 962-972.  [12]  Cong, X., Gong, S, Darwin, D. and McCabe, S. [1992], "Role of silica fume in compressive strength of cement paste, mortar, and concrete", ACI Materials Journal, Vol. 89, p. 375.  [13]  Maher, A. and Darwin, D. [1977], "Microscopic finite element model of concrete", Proceedings, First International Conference on Mathematical Modeling, University of Missouri-Rolla, Vol. Ill, p. 1705.  [14]  Popovics, S. [1987], "Attempts to improve the bond between cement paste and aggregate", Materials and Structures, RILEM, No. 20, p.32.  [15]  Richard, P. and Cheyrezy, M.H. [1994], "Reactive powder concrete with high ductility and 200-800 MPa compressive strength", Concrete Technology: Past, Present, and Future, ACI SP-144, pp. 507-518.  Chapter VIII  143  Chapter VIII  PARTICLE PACKING, RHEOLOGY, AND MICROSTRUCTURE-PROPERTY RELATIONSHIPS: A QUANTITATIVE APPROACH  8.1 TOWARDS A QUANTITATIVE SCIENCE O F CEMENT-BASED MATERIALS  The various solid particles (aggregates, cement, mineral fillers) in cement-based materials are randomly distributed at all ranges of particle sizes. This randomness is made more complicated by the fact that the microstructure changes markedly with time due to the hydration process. To develop a quantitative scientific model with true predictive capabilities, it is required to: (i) understand and quantitatively measure the microstructure of these multisize, multiphase particulate systems, and (ii) quantitatively relate these measurements to engineering properties.  The prediction of the mechanical strength alone is the focus of this chapter. This property depends, amongst other things, on: the chemical nature of the ingredients, their physical properties, the initial arrangement of particles, the rheology of the mixture, technological properties related to the compaction and curing, and the nature of the bond that develops between the various phases. Unlike properties such as shrinkage and creep, the mechanical strength does not seem to be controlled by phenomena at the C-S-H level, a dimension not yet clearly understood. The major factors that affect strength were summarized by Mindess [1] as: (i) total porosity, (ii) pore size distribution, (iii) presence of "flaws" in the system; and (iv) homogeneity or heterogeneity of the system.  No attempt at comprehensive modeling of cement-based materials is made in this chapter. Following on the empirical investigations of the microfiller effect on rheology, mechanical properties and microstructure in previous chapters, there is an attempt here only to find a fundamental basis for understanding the microfiller effect in HPC. The reader can legitimately ask: Can we predict concrete properties from the knowledge of fillers? Can we rationally design a cement blend that contains a variety of microfillers? Can we design a concrete mixture to get a particular microstructure, and can we relate microstructure to properties?  Chapter VIII  144  Though these questions cannot all be addressed in a single thesis, this chapter will attempt to clarify the mechanisms underlying the microfiller effect on the properties of HPC. The microfiller effect on particle packing, rheology and hydration kinetics is discussed based on theoretical models and practical observations. Thus, even if the questions raised above are not all answered here, the chapter should pave the way for future investigations.  8.2 MICROFILLER E F F E C T AND PARTICLE PACKING  8.2.1 Theoretical Considerations of Particulate Mixtures  Perhaps the first article dealing with particle packing in concrete was published as early as 1892 by Feret [2]. The main issue then was to optimize concrete strength by choosing a convenient coarse aggregate, and to relate the porosity of the hardened mortar to the compressive strength of concrete. Most of the literature on particle packing was published in the 1930's and focused on the idealized packing of spheres. The topic gained renewed interest later in the 1950's and the 1960's owing to applications in atomic and space research [3]. Particle packing of concrete is, however, far from the idealized packing of spheres. Account should be taken of the shape, diameter, flatness and angularity of aggregates. For instance, it is shown that porosity increases from 40% to 60% when the angularity factor of aggregates doubles [4]. However, understanding the physics of particulate mixtures can help elucidate the microfiller effect in concrete mixtures and assist in the rational design of high-strength cement-based materials. For example, Roy at al. [5] reported that an increase of 40% in strength could be obtained just by adjusting the particle size distribution of the cement. As early as the 1930's, Caquot [6] formulated the volume v of voids in a binary mixture of aggregates in terms of the maximum and minimum diameters of the considered aggregates, D and d, respectively (eq. 8.1). Therefore, by adding finer components in an aggregate mixture, it should be possible to obtain higher packing densities [7]. According to this formula, adding a microfiller of mean particle size d to a cement having an average particle size D would become more effective in reducing the volume of voids of the dry mixture as d becomes smaller.  (8.1) where: v is an experimental constant. 0  Chapter VIII  145  a) Wall Effect One concept arising from the theory of packing of spheres is of great interest to concrete. That is the wall effect (figure 8.1). This effect is observed at the surface of larger aggregates, where the porosity is higher than in the bulk of a binary mixture. It has been shown [4] that a disturbed zone of higher porosity, whose thickness equals d/2, forms at the wall of the larger aggregate. This explains why adding a larger aggregate to a matrix of a finer aggregate will make the matrix less compact.  Fig. 8.1 Illustration of the wall effect. b) Interference effect Another important parameter is the interference effect [4]. When a quantity of larger aggregates is substituted for the same quantity of finer aggregates in a binary mixture, a threshold value of the substitution rate is observed. Below this value, adding larger aggregates does not disturb the arrangement of the finer aggregates (except for the wall effect). Above this value, a complementary interference effect, additional to the wall effect intervenes in the arrangement of fine aggregates (Fig. 8.2). The interference effect illustrates that there is a threshold value of the addition rate of a microfiller beyond which no reduction in the porosity of the dry mixture can be observed.  Considering the particle size distribution of the concrete components, their corresponding characteristics, the wall effect, the interference effect and the method of compaction, models for the packing density of concrete can be established. The optimization of the packing density is of great interest. It was found by Johansen and Peterson [3], that minimum porosity, minimum  Chapter VIII  146  permeability, maximum slump and maximum compressive strength were achieved for mixtures corresponding to an optimal packing density. For HPC, this implies that the predictive capabilities offered by a model can help in obtaining optimal properties, just by using optimal proportions of ingredients and adding an appropriate microfiller. In practice, a number of models for the packing density of concrete have been reported [3]. Those considered significant for this study are described below.  Binary aggregate mixture with no interference effect  Binary aggregate mixture with interference effect  Fig. 8.2 Illustration of the interference effect.  8.2.2 Modeling the particle packing a) Linear packing density model This model is based on the model for the viscosity of suspensions of randomly scattered particles developed by Mooney [7]. Particle packing is considered analogous to the limiting case of a suspension with an infinite viscosity. This assumption will be shown later to be one of the limitations of this approach.  When a large aggregate, i, is introduced into a finer aggregate matrix, j , the latter will be locally disorganized. A function called the interaction function lij is thus introduced to account for this effect. Knowing the individual packing density of each aggregate (Ci and Cj) and their respective diameters, the interaction functions lij and lji can be calculated. Thus, the evolution of the  Chapter VIII  147  packing of the binary mix can be calculated as a function of the aggregate proportions (Yj and Yj). If <j>i and <|)j are, respectively, the volumetric concentrations of the aggregates of class i and j in a unit volume, the class i imposes a volumetric restriction ^-ij-<Pj on the class j. The class j mutually imposes a volumetric restriction Ajj.^i on the class i. The assumption made by Mooney is that the interactions in a mix of n classes are additive, and hence the linear behavior of the model. The volumetric restriction imposed on the class i by all of the other classes in the mix is given by the expression: (8.2) 1*1  A better estimation of the interaction functions can be obtained from experimental binary mixtures. Considering the general case of a mix of n aggregates, and defining Y(x) as the cumulative size distribution of the mix (x being the particle size (d<x<D), and C the packing T  density of T size grains), the theoretical packing density C of the mix is given by :  C = min (d < T < D) \ (8.3) 1 -(l-C )\X .Y(x).dx- \X .Y(x).dx T  Tx  xT  Some criticism has been aimed at the linear model [8]. The model does not account for the overlapping of disturbed regions. Thus, it is only justified when the proportion of one component in the aggregate mixture is dominant. In addition, because of the linear assumption, the model does not account for second or higher order interactions. Judicious improvement of the model is possible by introducing non-linear terms. Because of its linear nature, the curves showing the relationship between proportions and packing density exhibit discontinuities in the vicinity of the optimal values. To account for this feature, which does not appear in practice, De Larrard and Sedran [8] have reconsidered the infinite viscosity assumption. Instead, Mooney's model is used, considering the random packing of particles as analogous to high, but finite viscosity. The packing density c(f) of any granular mixture is expressed in terms of the following implicit equation:  Chapter VIII  148  7j  = exp  1  r  2.5  ii.. V c  y(t)  (8.4)  1  c(t)J  P(t)  c(t) = -  where:  1 - jy(x).f (J-J  dx  dx - [1 - P(x)] Jy(x).g  n f : relative reference viscosity c(t): packing density y(t): volumetric size distribution of particulate mixture, t: size of grains d, D: respectively maximum and minimum sizes of grains f(z): loosening effect function, f(z) = 0.7 (1-z) + 0.3 (l-z) g(z): wall effect function, g(z) = (1-z) B(t): virtual specific packing density of t-size grains  and  r  l/2  13  b) Model proposed by Aim and Goff This geometrical model accounts for the wall effect when spherical particles are packed against a smooth wall, as well as variations in porosity in binary mixtures of spherical grains. The assumptions of the model are discussed elsewhere [9]. It considers that the porosity can be described for two distinct cases: (i) the volume fraction of fine particles is small, (ii) the volume fraction of fine particles is large. Assuming that the particles are spherical and packed with the same density, the volume fraction y* of fine particles that represents the boundary between these two cases gives the mixture with the maximum packing density:  1+0.9  a-£ y 0  (8.5)  y =• 1 + 0.9  2V  iJ  d-*o)  2  where d] and d are the diameters of the fine and coarse particles, respectively, and e the 2  0  porosity of the mixtures of the pure components. If (pi and q> are the experimental packing 2  densities of the fine and coarse particles, respectively, and assuming that the particles deviate from a spherical shape and have different porosities, then the volume fraction of fine particles to achieve a maximum density of the binary mixture is given by:  Chapter VIII  149  9x_  1 + 0.9  4,  •9x  V  M(  y -  —  (8.6)  <0  1 + 0.9 - •<Px +1 ^ dJ L  2  In the case where the volume fraction, yi, of the fine particles is small (yi < y*), the porosity and the packing density of the binary mixture are given respectively by:  e=  cp = l-s  =  (8.7)  i-y.  When the volume fraction of the fine particles is large (yi > y*), the packing density is given by:  <P =  1 + 0.9  <Px  -r-  d.  (8.8)  d-,  In practice, the replacement rates of cement are usually less than y*, and equation 8.7 will apply. Equation 8.7 does not directly depend on the diameter of the microfiller particles, and would give the same packing density for the various microfillers used in this study. Since practical rheological and microstructural observations have shown that the microfillers have different effects on the behavior of concrete, this model will not be used further here.  c)  Model proposed by Toufar, Klose and Born [101.  fill  This model calculates the packing density of multi-component granular mixtures as the weighted average of the total number of binary mixtures. Three limiting cases were considered: di » d , 2  di «  d and di* d . The authors accounted for the statistical probability of the number of 2  2  interstices between coarser aggregates that are free from smaller particles, and developed a formula to calculate the packing density of the binary mixture. To extend the same formula to the case of multi-component granular mixtures, the authors assumed that each pair of components  Chapter  VIII  150  would form a binary mixture. The total packing density C, is obtained as the summation of the contributions from all binary mixtures: 1 n  (8.9)  j - l  r  7=2 i=l  hij  where: r,  h = n-j +r , r _j = l - r f ^ '  a  n  t  H  d  r  H = rT-7, r  v\: volume fraction of grains i and: (8.10) ( 1  r. 1 +4 —  ^  -1  d.. + d,.  1-  1+ ^ £ : specific packing density of i-size grains di: diameter of grain i. Equation 8.10 can be used to estimate the particle packing of binary cements, while equation 8.9 can be used for triple and multiply-blended cements.  8.3 PARTICLE PACKING AND RHEOLOGY In a freshly mixed cement based material, the mixing water will usually be divided into: (i) bulk water which fills in the voids between the various discrete particles, and (ii) water that adsorbs on the surface of the particles. The bulk water does not enhance the flow of the mixture, and its volume depends on the packing density of the particulate system. The addition of a microfiller will fill in some of the voids between the relatively coarser cement grains, leaving less room for bulk water. Thus, based on a constant water content, more water will be available as surface layer water when a microfiller is partially substituted for cement.  Chapter VIII  151  However, a microfiller usually has a much higher surface area than that of cement. Thus, more water is adsorbed on the surface of the particles. Although the bulk water is decreased, the increase in the adsorbed water due to a higher surface area can result in an overall increased water demand.  A superplasticizer usually cannot decrease the amount of bulk water, and its effect in low surface area systems is limited. For instance, in a cement system without a microfiller, the superplasticizer saturation dosage is low, meaning that beyond this level, the superplasticizer is merely playing a filling role similar to that of water. In practice, the formulator will observe that making HPC at a very low w/c ratio without using a microfiller is  Theologically  very difficult,  regardless of the amount of superplasticizer used. On the other hand, in a cement-microfiller system, the bulk water is decreased, and most of the water is adsorbed on the surface of grains or entrapped inside floes of particles. Through its action on the surface charges, the superplasticizer liberates a significant amount of adsorbed water, which can lubricate the mixture through an increased thickness of the surface layer of free water around the particles. In practice, the higher the surface area of the microfiller, the higher the superplasticizer saturation dosage. This explains the results of Chapters 4 and 5, where HPC mixtures containing the finer microfillers had improved rheological characteristics and a higher efficiency of the superplasticizer.  Apart from these technological aspects, the dependence of viscosity on particle packing has led to the development of several mathematical models. Mooney [7] was the first to attempt to calculate the shear dependency of the viscosity from an estimate of the packing volume fraction of a solid:  a<j>  ri=e ~ * l  k  (8.11)  where: n = viscosity <() = packing volume fraction a and k = constants Some other models have followed the same approach. For instance, Ball and Richmond [12] suggested another relationship between viscosity and phase volume as follows:  Chapter VIII  152  f n=  1-  (8.12)  where (jw is the maximum packing fraction. These models however, consider primarily the effect of the solid phase volume on the rheology of suspensions. In cement based materials, particle-particle interactions usually prevail, and the packing volume fraction becomes shear dependent. Expressing the rheological characteristics in terms of a wide range of shear rates is complicated and impractical.  8.4 M I C R O F I L L E R EFFECT AND HYDRATION REACTIONS  The hydration reactions of cement can be divided into different stages; each controlled by a different mechanism: (0) Initial hydrolysis prevailing up to the end of the induction period. (1) A nucleation and growth stage which prevails during the initial hydration reactions. (ii) A phase-boundary reaction which controls the transition period between acceleration and deceleration of the hydration process. (iii) A diffusion-controlled process which dominates the hydration reactions at later stages. Several mathematical models have been proposed to try to predict the rate of hydration at a certain time t. These models are either based on one of the mechanisms above, or on a hybrid formulation combining them. Some of these models are discussed below, with special focus on the effect of particle size distribution. 8.4.1 Nucleation and Growth Model  This model is based on the assumption that each cement particle reacts independently of the others. The growth of hydration products is controlled by a surface chemical reaction. The nucleation is initiated during the induction period, and growth proceeds at later stages. This can be expressed by the model proposed by Avrami-Erofeev [13,14] as follows:  Chapter VIII  -Ln  153  [ l - ( « - a )] N  =  k (t-t ) G  n  (8.13)  N  where: a = degree of hydration oc = degree of hydration at time t t = time when nucleation starts ko = rate constant for growth of hydration products (hr") n = determinator of the product type, 0.5<n<4.0 n = (P/S+Q) where P = dimensionality constant for the growth of product (P=l for elongated shapes, P=2 for sheets, plates and foils, P=3 for polygons), S = limiting growth rate constant (S=l for phase-boundary control, S=2 for diffusion control), Q = nucleation rate constant (Q=0 for zero nucleation rate, Q=l for a constant nucleation rate). N  N  N  n  8.4.2 Phase-Boundary Model After a certain time, the hydration products are no longer coalescing through nucleation and growth, but proceed through a one-dimensional thickening mechanism. The nucleation and growth come to an end, but the hydration layers around cement grains are not yet thick enough for diffusion to be the main mechanism for further hydration. An intermediate mechanism between nucleation and diffusion, called the phase-boundary process, prevails. The phase boundary model for the hydration reaction is based on the assumption that the rate of hydration is proportional to the surface area of the unhydrated cement grains. The general expression of the Kondo and Ueda model [15] that describes this behavior is given by:  [l-d-«)  1 / 3  ]  N  (8.14)  where: k = rate constant [hr ], k = k/R where K is a mass-transfer coefficient and R is the radius of the particle. N = empirical constant (N=l when process is phase-boundary controlled, N=2 when the process is diffusion controlled). 1  p  P  Thus, after the nucleation and growth process has ended, the degree of hydration is given by:  a(t) = 1 + a-[i-(k (tP  p  )y ]  ,N 3  tp  (8.15)  Chapter  VIII  154  8.4.3 Diffusion-Controlled Model  The phase-boundary reaction proceeds for several hours till the thickness of the hydration layers around the cement grains reach levels that allow only for diffusion mechanisms to proceed. For a single particle reacting independently, the hydration process is then described by the relation proposed by Knudsen [16] (Eq. 8.16). The extension of this model to include the particle size distribution will be presented later.  In a =  (8.16)  where: m: describes the type of kinetic model (m=l for a linear process such as the phase boundary reaction, m=0.5 for a parabolic process such as the diffusion process).  8.4.4 Total Degree of Hydration and Effect of Particle Size Distribution  As stated earlier, the hydration reaction can be divided into four stages (Fig 8.3). The first stage (t<t ) includes an initial phase characterized by an intense chemical activity and a significant N  heat release, followed by a dormant period characterized by little chemical activity and heat release. Both of these phases are not considered here for the estimation of the degree of hydration. At a time t <t<t , the reaction is driven by a nucleation-growth process, at a time N  P  t <t<t the hydration reaction is controlled by phase-boundary reactions, and at a time t>t , the p  D  D  hydration reaction is essentially driven by a diffusion process. If the three models for the hydration rate above are integrated over the three corresponding periods of time, a complete model can be obtained. However, the effect of the particle size distribution of the cement has to be included for realistic results to be achieved.  Starting with the nucleation-growth phase, a particle with a diameter R at a time t would have a degree of hydration represented by a(t,R). The total degree of hydration of the cement dispersion can be written as:  (8.17)  Chapter VIII  155  where: W(R) is the weight fraction of the cement having a particle size R, and R^n and R ^ are the radii of the smallest and largest particles in the dispersion, respectively. As suggested by Pommersheim [17], at a certain time t, some particles with an original diameter smaller than R*(t) would have fully hydrated and would thus be considered "dead". The degree of hydration is then expressed as:  R^  R'(t)  a(t) =  jW(R).dR +  (0)  la(t,R).W(R).dR  (2)  (i)  (8.18)  (3)  I O  i  lt.,,1,,, I, „L„I,,„l,„,i,..,i „,li„l  tN  „„l„ ,li,„l„„l„„lii„l„t),i,l ,„l„.l„.,l„..l„„l,i„ l  l  tp  ...i.,,.I.. ,1 ...1... 1  I ,: I  ^ i  :  ;  i  I.,,,I,,,,I., ,1,.,,I, ,| ,„l „l,„l„„l„„l„, l  k> Time [hours]  Fig.8.3 Schematic illustration of the evolution of the heat of hydration vs. time and the corresponding hydration phases. To solve equation 8.18, two approaches can be used: (1) to introduce a mathematical function for the particle size distribution of the cement that best fits the actual experimental data ; or (2) to 1  introduce a model function for the particle size distribution of cement. The most widely used function for this purpose is the Rosin-Rammler distribution function given by:  Chapter VIII  156  n -\  "R '  R  {)  W(R) =  R  exp  (8.19)  where: W(R) = weight fraction of particles with a size R in the dispersion. n = shape parameter of the distribution function. Ro = mean particle size of the dispersion. R  Inserting equations 8.13 and 8.19 in 8.18 yields:  \ "R  a (r) = [l + a N  -exp(-k (t-t ) ) n  N  G  N  exp  f** V  f  + exp  — V  vR  0  J  - exp J  (R\t))  ( / )  -exp - •  ] J  Ro J t  "R^  (8.20)  —  J  V  The effect of the particle size distribution on the phase boundary reaction imposes mathematical difficulties beyond the scope of this dissertation. Fortunately, the short period of time during which this stage is the controlling one allows us to disregard this effect without much loss in accuracy [18]. On the other hand, for the diffusion-controlled reaction, Bezjak [19] has extended the Knudsen formula (eq. 8.16) to account for the particle size distribution by including the Rosin-Rammler distribution as follows:  -.R * n  a(t, R) = exp  (8.21)  <D"*(0  where the function O(t) is given by:  U/2  0(0 =  2k  + k (t 2  to)  2L  (8.22)  For this purpose curve fitting of sedigraph particle size distribution can be used.  Chapter VIII  157  where ki [um/hr] is a mass transfer parameter for the diffusion mechanism, and k [um /hr] is the 2  2  coefficient of diffusion. Considering Rmin(tD) and Rmax(tD) as the smallest and largest dimensions of particles that have not fully hydrated at time t when the phase boundary process has ended D  and the diffusion process started, the degree of hydration can now be estimated by:  a(t)=  \a't,R).W'R).dR Rutin  (8.23)  0) D  Substituting equation 8.17 And 8.21 into equation 8.23 yields:  a(t) = a + D  (0(0/*,)"  (o(o//? r+i.  exp  0  (8.24)  U " 0  The total degree of hydration can now be calculated in the fashion of Saglik [18] by adding the relative contribution of each phase (eq. 8.20 + eq. 8.15 + eq. 8.24):  «"(')  \w(R).dR +  a(t) =  «max('«)  \a(t,R).W(R).dR  «max(</>)  +  \a(t,R).W(R).dR+  i U ( ' o )  \a(t,R).W(R).dR  "min  (8.25)  8.5 GENERAL M O D E L FOR T H E MICROFILLER E F F E C T  From the above discussion, it would seem possible to estimate the initial particle packing of the cementitious binder, as well as the degree of hydration at a time t including the effect of the particle size distribution of the cement. The question is: would it be possible to combine this information to predict the performance of a microfiller cement at a certain time t? In clearer terms: can we model the performance of a microfiller cement so that we can predict the mecanical strength achieved at a certain time t and compare it to that of an ordinary portland cement?  Chapter VIII  158  Let us assume that MEF(r,t) is the microfiller efficiency factor at a replacement rate, r, of the cement and at time, t, after mixing the cement based material. MEF(r,t) will essentially depend on: (i) the initial particle packing of the system. (ii) the microfiller effect on rheological characteristics such as the water demand, superplasticizer efficiency, segregation, bleeding, etc. This will determine the yield stress and viscosity of the system, which in turn will affect the flow characteristics, the final packing of the system and its homogeneity. (iii) the microfiller effect on the kinetics of the hydration reactions. (iv) the reactivity of the microfiller (pozzolanic or other) and its effect on the nature of the resulting products of the hydration reactions.  These effects will largely determine what kind of microstructure will be obtained at a certain age. In particular, they will affect the porosity, unhydrated grains content, interfacial transition zone properties, and the type and amount of hydration products. Also, they will define the nature of the bond that will develop to "glue" the discrete grains in the system together and provide strength.  8.5.1 Initial Particle P a c k i n g In terms of initial particle packing, the MEF can be defined as:  (8.26)  where: cp = packing density of cement. (pf = packing density of cement with r% microfiller. D = mean particle size of the cement, d = mean particle size of the microfiller. c  Chapter VIII  159  8.5.2 Rheology It was shown earlier (Figs. 4.9 and 4.10) that the microfiller efficiency factor with respect to rheology of fluid and self-leveling HPC is a function of the viscosity of the mixture. According to Mooney's equation and later extensions (section 8.3), the viscosity is related to the packing density of the system, which in turn is a function of (d/D). Assuming a constant w/b ratio and an ideal dispersion of particles using a superplasticizer:  MEF (r,t ) = rh  r  (8  .27)  where: Uf = viscosity of a system made with cement and r% microfiller. Ut = viscosity of a system made with pure cement. t = time when the cement based material is poured into the molds. r  8.5.3 Degree of Hydration The efficiency of a microfiller at a time, t, and a replacement rate r, will depend on its reactivity and its effect on the kinetics of the hydration reaction:  MEF (r,t) =  (1-r) a (r,t) + rP(t)  dh  (  f  ^  (8.28)  where: cif(r,t) = degree of hydration of cement with r% microfiller at time t. a (t) = degree of hydration of pure cement at time t. P(t) = accounts for pozzolanic reaction of the microfiller at time t. c  At a very early age, a (t) and 0Cf(r,t) can be estimated using the nucleation and growth model (eq. c  8.13). At later ages the model should also include the phase boundary and diffusion reactions, and eq. 8.25 should be used.  8.5.4 Model for the Microfiller Efficiency  The overall microfiller efficiency at a time, t, and replacement rate, r, is proportional to the degree of hydration ratio and the packing density ratio, and inversely proportional to the w/b  Chapter VIII  160  ratio. Assuming a power relationship between the hydration and particle packing efficiency, and the overall efficiency factor, MEF can be written as:  a MEF(r,t) =  where:  (8.29)  : degree of hydration ratio of a filler cement to a pure cement.  : Particle packing ratio of a pure cement to afillercement, a, b, c and d are constants.  8.5.5 Application of the model  The model of equation 8.29 requires an evaluation of the particle packing ratio for r% replacement of cement by a microfiller, as well as the hydration efficiency factor at a certain time t. The particle packing ratio can be obtained using the Toufar et al. model (equation 8.10). The hydration efficiency factor can be estimated using the nucleation and growth model at early ages (equation 8.13), equation 8.25 for later ages, or directly from experimental hydration isothermal curves.  The Toufar et al. model was applied to the microfiller cements used in the present study. The replacement rates used were 0 to 20%. In total four different microfillers were evaluated. Fig. 8.4 shows the particle size distribution of the cement and fillers (a) and the packing density corresponding to each microfiller at increasing replacement rates (b). It may be observed that the packing density increases significantly with a reduction of the mean particle size of the microfiller. At 5% replacement, the filler effect is limited, but at 10% and higher rates, the microfiller effect on the packing density is significant. The results suggest that the microfiller LF2 should be able to play a significant part of the physical filler effect of silica fume. It should be remembered, however, that cement and filler particles are not monodisperse but polydisperse, and the particle size distribution should be taken into account for a more accurate estimation of particle packing.  Chapter VIII  161  Fig. 8.4 Illustration of (a) sedigraph particle size distribution of the cement and microfillers, and (b) the packing density of binary cement-microfiller combinations containing various proportions of each microfiller.  Chapter VIII  162  a) Microfiller effect on rheology Fig. 8.5 shows the relationship between the packing density and the torque viscosity for a 15% replacement of cement by various microfillers. The relation fits an exponential equation as suggested earlier by Mooney [7]. It is observed that the fine microfillers achieved a higher packing density and a lower torque viscosity. This reduction in the viscosity at increased packing density (and at a constant phase volume) is known as the Farris effect.  Combining particle sizes allows the viscosity of a suspension to be reduced whilst maintaining the same phase volume, or alternatively, the phase volume to be increased whilst maintaining the same viscosity. In practice, the particle size distribution of a cement or a blended cement (sedigraph curve) is a curve which often fits some empirical mathematical expression. It is not yet clear however, how the packing density and viscosity can be defined from such an expression in a reliable manner. A model that specifically considers the effects of temperature, hydration process, particle size distribution and chemical composition on rheology has not yet been formulated, nor do existing models account for the effect of chemical admixtures and mineral additions. Thus, it is still difficult to predict rheology and loss of flow from known properties such as the particle packing of the system.  3 .-9.6197X 9 6 1  y = 912.86e  R = 0.9515 2  m cr  SF  1 -  £ 0.5 1 0  0.58  0.6  0.62  0.64  0.66  0.68  0.7  Packing density Fig. 8.5 Correlation between packing density and torque viscosity for microfiller cements (microfiller = 15%, w/b = 0.33).  Chapter VIII  163  b) Microfiller effect on early age properties At 12 hours, and in normal temperature conditions, the hydration rate can be estimated using the nucleation and growth model without significant loss in accuracy [18]. There is still a question as to whether an earlier start of nucleation and growth and/or a higher rate of growth of hydration products is responsible for the acceleration of the strength development in microfiller cements. Figs. 8.5-8.8 illustrate plots of equation 8.13 keeping constant all parameters except t , the time N  when nucleation starts, and ko, the rate constant for growth of hydration products. The rest of the constants were selected from previous data [18].  Figs 8.6-8.8 below illustrate that either decreasing the initial time of nucleation alone, or increasing the rate of growth of hydration products alone, does not lead to drastic increases in the rate of hydration at 12 hours. However, when both the initial time of hydration is decreased and the rate of growth of hydration products is increased, the degree of hydration is significantly increased.  For a certain coarse aggregate, equation 8.29, which expresses the microfiller efficiency factor as a function of the particle packing ratio and the degree of hydration ratio, can be written as:  a MEF(r,t)  =  \  2  (8.30)  A. Fig. 8.9 illustrates the modeling of the microfiller efficiency factor (Eq. 8.30) at 12h as a function of the degree of hydration ratio and the particle packing ratio. At early ages, the microfiller efficiency factor increases significantly when the microfiller can accelerate the rate of the hydration reactions and when a better packing density of the cementitious binder is achieved. The acceleration of the hydration rate directly increases the strength through a more rapid development of bond. The improved particle packing increases the strength indirectly via an improvement of workability and compactness, a reduction of porosity and flaws in the cementitious system, and a reduction of the wall effect and the transition zone thickness.  Chapter VIII  164  Initial time of nucleation [hour]  Fig. 8.6 Illustration of the effect of the initial time of nucleation on the degree of hydration at 12 hours (at the rate of growth of hydration products K = 0.009). G  0.005  0.01  0.015  0.02  0.025  0.03  Rate of growth of hydration products  Fig. 8.7 Illustration of the effect of rate of growth of hydration products on the degree of hydration at 12hours.  Chapter VIII  165  iu  •  t [hr] N  Fig. 8.8 Illustration of the combined effect of the initial nucleation time (t ) and the rate of growth of hydration products ( K G ) on the degree of hydration at 12 hours. N  MEF  4  Fig. 8.9 Illustration of the modeling of the microfiller efficiency factor at early age as a function of the particle packing ratio and degree of hydration ratio.  Chapter VIII  166  Fig. 8.10 shows the relationship between the microfiller efficiency factor and the compressive strength at 12 hours for blended cements containing 15% of various microfillers. The microfiller efficiency was calculated from equation 8.30 which accounts for the microfiller effect on the rate of hydration reactions and the particle packing density of the binder, as well as the effect of the w/b ratio. A unique relationship relates the microfiller efficiency factor to the compressive strength with a high correlation coefficient. Since compressive strength depends on the nature of coarse aggregate used, research should be conducted to investigate how this model can be modified to account for aggregates varying from light weight to high-density ones.  These results show that it is possible to predict the early age compressive strength of microfiller cements based on a knowledge of the particle size distributions of the cement and the microfiller, and the microfiller effect on the rate of the hydration reactions of cement. This kind of model can be extended to include the effect on the rheological properties of cement based materials which in turn determine the flow properties of the mixture and the final compactness of the system, before the onset of the hydration reactions and the development of bond and mechanical strength.  0 -I 0  1  1  !  j 2  1  3  !  4  i  1  5  6  1  7  8  1 9  Microfiller efficiency factor  Fig. 8.10 Relationship between microfiller efficiency factor and compressive strength at 12h for cements containing 15% of various microfillers.  Chapter VIII  167  8.6 MICROFILLERS AND MICROSTUCTURE-PROPERTY RELATIONSHIPS 8.6.1 Models for microstructure- strength relationship It has been suggested [20] that the strength of hardened cement pastes depends mainly upon: (i) Inter-atomic forces within the hydration products particles, (ii) Interfacial atomic forces which bind together the particulate system, (iii) Size and morphology of the hydrated particles along with their inter-growth and agglomeration, (iv) Geometrical arrangement and density distribution of the phases in the microstructure. This traditional view is based on models that describe the microstructure as discrete particles of C-S-H bonded to each other and to unhydrated grains at their points of contact. As stated by Mindess [1], these "particulate" models do not provide an adequate representation of the microstructure, and thus, no satisfactory relationship between the mechanical strength and the above listed parameters has yet been established.  Another model [21] describes the structure of C-S-H as an irregular array of single layers of C-SH bonded together in a random pattern through solid-solid contacts, creating regions of interlayer space. It has been argued [22] that almost all models that have been proposed for the structure of C-S-H suggest a higher degree of order than really exists. C-S-H is in fact an amorphous, colloidal material containing large amounts of impurities. It has considerable variability both in composition and structure, which may be associated with the local conditions pertaining to its formation. The microstructure at the C-S-H level is a dimension not yet clearly understood, and no reliable quantitative measurement of the C-S-H structure has yet been achieved. Fortunately, the strength seems to depend on structural features at a coarser scale, and a detailed understanding of the C-S-H structure may not be required.  8.6.2 Mechanical strength versus porosity  For a lot of engineers and practicing designers, the strength of cement based materials is still perceived as being proportional to the cement content. However, as far back as 1897, Feret demonstrated that the compressive strength of a cement-based material depends to a great extent on its porosity. Several models have been proposed which try to relate strength to porosity, for instance the Powers and Brownyard model [23]:  Chapter VIII  cr  168 = AX"  (8.31)  whereCT is the compressive strength, X is the gel/space ratio, and n a constant with values c  ranging from 2.6 to 3.0. Other models have been reported in Mindess [1]. Although these models seem to be appropriate when applied to ordinary concrete, they oversimplify the system, the pore size distribution, the effect of humidity, and the nature of the solid material itself.  Indeed models relating strength to total porosity suffer from serious limitations. Regardless of the reliability of the models, there is a dispute regarding how accurately the available techniques can measure porosity [24]. Furthermore, the apparent volume of any pore is a strong function of the moisture content [25]. In practice, engineering properties of cement based materials such as strength, permeability, shrinkage, and creep are altered by wetting/drying. In addition, while there is a general agreement that microsilica increases the strength of cement paste, mortar and concrete, it is observed that the total porosity of microsilica-cement paste or concrete is similar to or greater than the total pore volume of a reference sample not containing microsilica [26]. There is much evidence that for the same total porosity, the pore size distribution has a considerable effect on strength. Indeed, the fracture mechanics approach suggests that the correlation between total porosity and strength is fortuitous. It is the largest capillary pores which constitute the "Griffith flaws" that determine strength. A reduction in the total porosity only suggests a reduction in the probability of occurrence of large flaws. Above all, the models relating strength to porosity ignore the nature of the solid phase itself which plays a fundamental role in determining mechanical strength.  8.6.3 New Model for Microstructure-Strength Relationship  Direct evidence shows that the nature of the solid phase is different in high-strength cementitious systems compared to ordinary systems. The structure is more amorphous, denser and more homogeneous. The features in the transition zone are finer and less crystalline. A true predictive model should account for these differences in relating microstruture to properties. The major difficulty is how to quantify such a random and diverse microstructure.  Chapter VIII  169  However, this very complicated microstructure is greatly simplified when observed in the backscattered electron mode in which details finer than 0.5 pm cannot be distinguished. The microstructure can simply be divided into the following phases:  •  Unhydrated cement grains: which are proportional to the cement content, and inversely proportional to the initial w/c ratio and the degree of hydration.  •  Inner hydration products: which develop within the boundaries of the original particles. This is typically a denser and more homogeneous phase which has higher bonding capabilities.  •  Outer hydration products: which surround the inner products and the unhydrated grains. They typically form in the original pore space or in spaces from which grains have dissolved away. This phase encompasses distinct morphologies, including porosity, crystalline CH, ettringite and other phases.  The microstructure-strength model suggested here is represented schematically in Fig. 8.11. It is assumed that as the original pore space is decreased, the hydration reactions will proceed preferentially through a phase boundary mechanism, yielding more inner hydration products. When the original pore space is high, long range diffusion of calcium and silicate ions will drive the hydration process, yielding lower density hydration products, and thus a higher outer hydration products content. This is because the calcium and silicate ions have to travel far from the cement grains under the concentration gradient between that in the vicinity of cement grains and that in the remote pore solution. This concept is illustrated in Fig. 8.12 which shows two bse micrographs of 0.33 and 0.25 w/b concrete mixtures. The first concrete shows a higher gray scale ground mass corresponding to a low density and more porous hydration products. The second shows a denser and lighter microstructure representing a higher inner hydration products content and less porosity.  According to this model, the ideal situation for the development of strength would be when the particle packing of the cement grains is optimum. The initial porosity is very low, and a minimal water content is present. The hydration of cement grains would be very limited. Only the surfaces of cement grains would react, providing a dense glue bonding the discrete particles together.  Chapter VIII  170  (a)  d V 4jf  o  Unhydrated cement Hydrated cement  Porosity  Filler + water  Calcium hydroxide Ettgingite  Groundmass  Fig. 8.11 Illustration of the microstructure-strength model: (a) initial state in high w/b ratio, (b) initial state in low w/b ratio, (c) final state in high w/b ratio-Mow inner hydration products content + high crystalline phase and porosity, (d) final state in low w/b ratio->high inner hydration products content + low crystalline phase and porosity. Certainly, this situation is difficult to achieve for both economic and rheological reasons. The hydraulic potential of the cement is not fully utilized, and this is not consistent with the ecological and energy requirements mentioned earlier in this study. Likewise, an ordinary pure cement binder is far from having an ideal particle packing. In practice, it is difficult to produce a HPC with pure portland cement at a w/c ratio lower than 0.30. Rheological requirements will impose a high water content. In addition, the thickness of the interstitial layer of water will be high, requiring the long-range diffusion of calcium and silicate ions, and thus, more outer hydration products.  Chapter VIII  (a) 0.33 w/b concrete with higher porosity and crystalline phase, and lower inner hydration products content.  (b) 0.25 w/b concrete with lower porosity and crystalline phase, and higher inner hydration products content  Fig. 8.12 BSE micrographs illustrating concepts of the microstructure-strength model.  171  Chapter VIII  172  If a microfiller is added to the mixture in the presence of a superplasticizer, the particle packing of the system will be improved, and less free water will be needed to achieve the rheological requirements. The initial porosity will be finer, and because of the higher surface area, more water will be adsorbed on the particle surfaces. In these conditions, the long-range diffusion of calcium and silicate ions to fill in the porosity will be limited. The hydration will proceed 2  mostly on the surface of particles, yielding denser inner hydration products. Furthermore, the fine particles will provide nucleation sites for the growth of the hydration products. The microstructure will be more homogeneous, and the presence of large crystalline CH inclusions, usually observed in lower density systems, will be limited. This situation will be improved even more if the microfiller itself is pozzolanic. It will consume CH, reducing the crystalline phase which has a lower binding capacity. In addition, paste-aggregate interfaces will be denser, containing finer and less crystalline features (Fig. 8.13).  Another advantage of the limited hydration of the cement grains pertains to shrinkage stresses. The process of hydration is accompanied by about 9% chemical contraction (le Chatelier contraction). In a high cement content system which also has a poor particle packing, the water content will be high, and hydration can proceed to the extent that chemical contraction can induce severe stresses. Added to this will be the self-desiccation shrinkage in non-cured cementbased materials due to the consumption of internal water by the hydration reaction. Furthermore, in this system, abundant growth of ettringite and portlandite crystals will occur, adding crystal growth stresses, which induce shear at interfaces and disrupt the microstructure.  The model relating microstructure to strength proposed here will account not only for the porosity of the system, but also for the ratio of inner hydration products to outer hydration products. The advantage is that porosity, inner, and outer hydration products can be quantified from a single bse micrograph, contrary to other models where porosity has to be measured by intrusion techniques, and the microstructure qualitatively observed using microscopy. The disadvantage is that the threshold of gray levels in a bse micrograph is not clear cut, and has to be defined by the operator. However, if consistent settings are maintained for all micrographs, satisfactory results can be obtained. In addition, porosity and hydration products have to be averaged on several fields in a bse micrograph, which is time consuming. However, automated image analysis systems can make this process more efficient. 2  The calcium and silicate ions repeatedly referred to in this chapter are simply Ca** and Si '. 4  Chapter VIII  173  (a) Interfaces in high w/b ratio concrete containing no microfiller illustrating low density, high porosity and crystalline features.  (b) Interfaces in very low w/b ratio concrete illustrating denser phases, low porosity and less crystalline features. Fig. 8.13 BSE micrographs illustrating paste-aggregate microstructure in low w/b concrete containing no microfiller versus a low w/b concrete containing a microfiller.  Chapter VIII  174  The strength is inversely proportional to porosity and directly proportional to the ratio of the inner to outer hydration products. As stated earlier, inner hydration products are denser products, which form within the cement particles, while outer products constitute the groundmass of cement paste including CH and excluding porosity. The model can be written as:  cr. = q exp  (8.32)  -1 + z v)  where: q and z: empirical constants. inner products/outer products, o: porosity measured in a bse micrograph at a 500x magnification. The model of equation 8.32 was calibrated using microstructural data measured on an OPC concrete and concretes containing 15% LF2 and 15% SF at two w/b ratios (Table 8.1).  Table 8.1- Microstructure and strength data used to calibrate model of equation 8.32. Mixture  I n n e r C S H [%]  O u t e r C S H [%]  P o r o s i t y [%]  e x p (§/v)  Strength [Mpa]  23.8 21.8 25.6 28.9 32.7 35.4  49.5 45.1 43.0 42.1 39.3 35.7  13.6 11.1 8.2 8.5 6.7 5.5  1.036 1.045 1.075 1.084 1.132 1.198  77.0 79.5 94.3 91.3 101.3 115.3  OPC-w/b=0.33 LF2-w/b=0.33 SF-w/b=0.33 OPC-w/b=0.25 LF2-w/b=0.25 SF-w/b=0.25  Fig. 8.14 shows that the microstructure-strength relationship fits the proposed model with a correlation coefficient of 95%. The advantage of this relationship is that, unlike previous models, it does not ignore the effect of the solid phase in relating strength to porosity. Both the nature of the C-S-H and the amount of porosity are combined to model the strength. The calibrated model can be written as:  a  c  =228.5 exp  -157  (8.33)  Chapter VIII  175  The rationale of this model is not to express the measured strength of concrete in terms of quantitative microstructure values measured using time consuming bse imaging and analysis. Rather, it is proposed that extensive measurements of porosity and C-S-H separated into inner and outer products, including the effects of various types and proportions of microfillers, can provide data that may improve our understanding of how to obtain high-strength concrete. A model having predictive capabilities both of microstructure and strength, based on the knowledge of the nature and particle size distribution of the ingredients, degree of dispersion of particles, and w/b ratio may be achieved. Based on the model proposed here and the experimental results obtained in this study, strength can be increased via better particle packing, combined with a better control of the nature of the hydration products. This can be achieved by the addition of one or more microfillers, the use of a superplasticizer, and a very low w/b ratio to limit the degree of hydration to the surface of the cement particles. This would limit the initial porosity, provide denser inner hydration products, and limit the shrinkage stresses due to chemical contraction and self- desiccation. Contrary to commonly held opinion, strength is not necessarily increased through costly and ecologically harmful increased cement contents.  Fig. 8.14 Calibration of the microstructure versus compressive strength model.  Chapter VIII  176  8.7 CONCLUSIONS  This chapter describes a particle packing approach to the design of cement blends. The formulation of an optimal mixture is treated as a problem of selecting the appropriate sizes and proportions of particulate materials to fill larger voids with smaller particles, forming smaller voids which in turn are filled with smaller particles, and so on.  The particle packing of a cement-based material depends mainly on its particle size distribution, the wall effect, the interference effect, the degree of dispersion of particles and the method of compaction. In this chapter, various models for the prediction of the packing density from the knowledge of the particle size distributions of the ingredients have been presented. Concrete materials are neither monosized nor spherical as the models often assume. Nevertheless, the concept of Caquot [6] in which the packing density increases with a wider particle size distribution has been confirmed. The packing density increases with the addition of microfillers; the finer the microfiller the higher the packing density. This stresses the important role of superplasticizers; it is futile to optimize a particulate mixture if an optimal deflocculation of the ultrafine particles is not attained.  Results seem to indicate that the best workability of HPC mixtures is obtained with the densest particle packing. In a dense particulate system water will not play only the role of filling voids; it will increase the thickness of the water layer around particles, which in turn decreases viscosity and improves flow properties. A superplasticizer can partially liberate the water adsorbed as a surface layer, but cannot decrease the amount of bulk water. Thus, a superplasticizer is more efficient in a dense particulate system with a high surface area, and has only a limited effect in a porous low-density system.  To develop models that can predict the strength of cement based systems from the knowledge of the characteristics of the particulate mixture, not only does the effect of the particle size distribution on the packing density have to be quantified, but also its effect on the hydration kinetics. Models that predict the hydration rate of cement grains are presented, including the effect of the particle size distribution. A microfiller efficiency factor has been defined based on the calculation of the microfiller.effect both on particle packing of the system and kinetics of the  Chapter VIII  177  hydration reactions. The model was applied to the prediction of the early-age strength of HPC mixtures and a close correlation was obtained.  Evidence from microstructural analysis shows that the microfiller effect is not limited to a reduction of porosity and a refinement of the pore size distribution; the nature of the solid phase can also be altered. Previous models relating microstructure to strength have almost exclusively been based on a porosity-strength relationship. The rest of the microstructural phases are rather considered qualitatively. The backscattered electron microscopy technique used in this work permits measurement of porosity, unhydrated grains, CH, and inner and outer C-S-H. It was observed in this study that when a non-pozzolanic microfiller is added to a high w/b ratio mixture, even though porosity was decreased and interfaces became denser, strength was not increased. However, if the same microfiller is added at a very low w/b ratio, strength is increased. It was found that when the particle packing is improved and the w/b ratio is low, more inner C-S-H is formed and the effect of the microfiller on strength is positive. On the other hand, at a high w/b ratio, long range diffusion of hydration products occurs, more outer hydration products form, and the microfiller effect is limited.  The chapter illustrates that if the mechanisms underlying the microfiller effect in HPC are understood, it is possible to replace 15% of the cement with grinding mill ultrafine particles, while at the same time improving significantly the rheological properties and increasing the strength. It is also well known that the single most important way to enhance the durability of cement based materials is a better, density of the cementitious matrix, including the pasteaggregate interface. It would be worthwhile to investigate the microfiller effect on durability in the light of the mechanisms explained in this chapter.  8.8 REFERENCES  [1]  Mindess, S. [1985], "Relationships between strength and microstructure for cementbased materials: an overview", Very High-Strength Cement-Based Materials, MRS Symposium, Vol. 42, J.F. Young, ed., pp. 53-68.  [2]  Feret, R. [1892], "Sur la compacite des mortiers hydrauliques", Annales des Ponts et Chaussees, 2 Semester, pp. 5-161. nd  [3]  Johansen, V. and Andersen, PJ. [1989], "Particle packing and concrete properties",  Chapter VIII  178  Materials Science of Concrete U, Skalny, J. and Mindess, S. eds., American Ceramic Society, pp. 111-147. [4]  Louvet, F. ,"Granularite des melanges de porosite minimale", Draft of Chapter 14 in an unpublished book, personal communication from: Ai'tcin, P-C.  [5]  Roy, B.E., Scheetz, B.E. and Silsbee, M.R. [1993], "Processing of optimized cements and concretes via particle packing", MRS Bulletin, August 1993, pp. 45-49.  [6]  Caquot, A. [1937], "Le role des materiaux inertes dans le beton", Memoires de la Societe des Ingenieurs Civils de France, pp. 562-582.  [7]  Mooney, M. [1951], "The viscosity of a concentrated suspension of spherical particles", Journal of Colloid and Interface Science, Vol. 6, pp. 3-20.  [8]  De Larrard, F. and Sedran, T. [1994], "Optimization of ultra-high-performance concrete by the use of a packing model", Cement and Concrete Research, Vol. 24,. No. 6, pp. 997-1009.  [9]  Aim, R.B. and Goff, P.L. [1967], "Effet de paroi dans les empilements desordonnes de spheres et application a la porosite des melanges binaires", Powder Technology, No. 1, pp. 281-90.  [10]  Toufar, W., Born, M., Close, E. [1976], "Beitrag zur optimierung der packungsdichte polydisperser korniger systeme", Freiberger forschungsheft A 558, VEB Deutscher Verlag fur Grundstoffindustrie, pp. 29-44.  [11]  Toufar W., Close, E., Born, M. [1977], "Berechnung der packungsdichte von Korngemischen", Aufbereitungs-Technic, 11, pp. 603-608.  [12]  Ball, R. and Richmond, P. [1980], "Dynamics of colloidal dispersions", Journal of Phys. Chem. Liquids, Vol. 9, pp. 99-116.  [13]  Avrami, M. [1939-1940], "Kinetics of phase change I-H-IU", J. Chem. Phys, Vol. 7 and 8, pp. 1103-1112, 212-224, 177-184.  [14]  Erofeev, B.V. [1946], "A generalized equation of chemical kinetics and its application in reactions involving solids", USSR Academy of Science, L l l (6), pp. 511-514.  [15]  Kondo, R. and Ueda, S. [1968], "Kinetics and mechanisms of the hydration of cements", Proc. 5 Int. Cong. Chem. Cem., Tokyo, Vol. U, pp. 203-255. th  [16]  Knudsen, T. [1984], "The dispersion model for hydration of portland cement, I. General concepts", Cement and Concrete Research, Vol. 14, pp. 622-630.  [17]  Pommersheim, J.M. [1982], "Effect of particle size distribution on hydration kinetics", Materials Research Society, Symposium Proceedings, Vol. 85., pp. 301-306.  [18]  Saglik, A. [1997], "Relationship between rate of hydration and physical and chemical characteristics of portland cement", International Conference on Engineering Materials, Ottawa, Almanaseer, A., Nagataki, S. and Joshi, R . C , eds., CSCE, pp. 181-196.  Chapter VIII  179  [19]  Bezjak, A. [1980], "An extension of the dispersion model for the hydration of portland cement", Cement and Concrete Research, Vol. 10, pp 260-264.  [20]  Idorn, G. [1968], In proceedings of the 5 Int. Cong. Chem. Cem., Tokyo, Japan, Vol. HI, p. 411.  [21]  Ramachandran, V.S., Feldman, R.F. and Beaudoin, J.J. [1981], Concrete Science, Heyden and Son Ltd., London, p. 3.  [22]  Taylor, H.F.W. [1979], in: Cement Production and Use, Engineering Foundation, Publication No. 79-08, New York, p. 107.  [23]  Powers, T.C. and Brownyard, T.L. [1948], "Studies of the physical properties of hardened portland cement paste", Bulletin 22, Portland Cement Association, Skokie, Illinois.  [24]  Feldman, R.F. [1989], "The porosity and pore structure of hydrated cement paste", Pore Structure and Permeability of Cementitious Materials, Roberts and Skalny, eds., MRS, pp. 59-73.  [25]  Jennings, H.M., Bhatty, J.L and Hodson, S.K. [1990], "Towards establishing relationships between microstructure and properties of cement-based materials", Advances in Cementitious Materials, S. Mindess, ed., American Ceramic Society, pp. 289-316.  [26]  Roberts, L.R. [1989], "Microsilica in concrete I", in Materials Science of Concrete I, J. Skalny, ed., American Ceramic Society, pp. 197-222.  th  Chapter IX  180 Chapter IX  SUMMARY AND G E N E R A L CONCLUSIONS  Although much information has been published on high-strength concrete during the last decade, the selection of materials and mix proportioning have been almost exclusively based on experience and empirical data. A fundamental understanding of the effect of the cementitious ingredients has not as yet been developed. It is still not possible to express microstructure as a "set of numbers" that can be related to engineering properties, nor to predict properties from a quantitative knowledge of the ingredients.  Despite its increasing use, High-strength concrete still faces considerable challenges. For instance, the material requires a high cement content compared to conventional concrete. At a low w/c ratio, rheological problems arise, which limit the workable time and make rheological characteristics variable and unpredictable. Excessive heat is often generated, combined with le Chatelier contraction and self-desiccation shrinkage. In these conditions most of the cement acts as a filler since only partial hydration is achieved due to the very low w/c ratios. In the wake of a potential energy crisis in the next century, the threat of global warming (the production of 1 ton of cement releases approximately one ton of CO2 into the atmosphere), and the increasing cement consumption of a steadily growing world population, the use of cement has to become more rational.  In view of the above considerations, the concrete industry has resorted to the use of several supplementary cementing materials for partial replacements of cement. Most of these are industrial byproducts or unprocessed materials, which make high-strength concrete not only more economic, but also more workable, stronger, and more durable. However, the mechanisms underlying these improvements are still in dispute. Whether the positive effect of mineral additions is due to their pozzolanic activity or to a physical filler effect is not yet clear. Furthermore, it is not as yet understood what role inert microfillers can play in HPC mixtures, and what would be their effect when used in triple-blended cements along with pozzolanic fillers.  Chapter IX  181  The motivation of this work was to define a more rigorous approach to the use of microfillers in making HPC. This can be achieved through understanding the mechanisms of the microfiller effect on the rheology, microstructure and mechanical properties. Thus, this work has mainly focused on explaining mechanisms. This study did not adopt the traditional experimental approach, which investigates local points within the experimental domain without determining the significance of interactions between parameters. Whenever possible, factorial experiments having predictive capabilities within the experimental domain were involved. Iso-response curves for the parameters under investigation were obtained rather than local response values. This provided a higher level of statistical reliability to the results.  The first mechanism investigated was the microfiller effect on the rheology of superplasticized cement pastes. The effect of 0 to 25% microfiller replacement of cement on the rheology of cement pastes was investigated both in silica fume and non-silica fume systems. The Marsh cone flow time, a rotational coaxial-cylinders viscometer test, the mini-slump test, induced bleeding, and flow of mortars have been carried out for various blended binders. It was found that the superplasticizer efficiency was increased when a microfiller was partially substituted for cement. This is explained by the fact that in a microfiller cement, the particle packing is denser and less water is needed to fill voids. More water is adsorbed on the surfaces of particles in a system which has a higher surface area. A superplasticizer can liberate surface layer water, but can decrease the bulk water only partly through a better dispersion of particles. In addition, the microfiller slightly increased the yield stress of cement paste and decreased its viscosity, which implies a better stability and flowability of cement paste. This was also reflected through a decreased induced bleeding and a better flow of mortars made with these cement pastes. Correlations amongst various rheological responses for cement paste showed that only those characteristics measured at comparable shear rates could readily be related by a model.  The second mechanism investigated was the microfiller effect on the rheology of highperformance concrete. A computer-controlled rheometer was used to measure the flow resistance and plastic viscosity of various HPC mixtures containing different proportions of various microfillers. In addition, the conventional slump test, the slump flow test, and induced bleeding were measured for the various concretes. First, the applicability and significance of rheometric tests for fluid and self-leveling concrete was addressed. It was demonstrated that rheometric tests can give reliable flow resistance and plastic viscosity measurements provided that a flow curve,  Chapter IX  182  rather than two tests at different shear rates, is obtained. Tests proved the existence of a resting shear yield stress which is higher than the steady state yield stress. There seemed to be a certain correlation between flow resistance and slump, while no clear relationship was observed between torque viscosity and slump.  Partial replacement of cement with ultrafine limestone did not increase the superplasticizer requirement even at high proportions, while silica fume replacement of cement increased the supersticizer requirement significantly. This suggests that the high surface area of silica fume may not be the sole factor affecting the increase in superplasticizer demand, and that silica fume may have an affinity for multi-layer adsorption of superplasticizer molecules. Microfiller replacement of cement reduced the flow resistance and torque viscosity of HPC  mixtures; the  finer the microfiller the greater the effect. Higher microfiller proportions were more effective in reducing the torque viscosity. It seemed that the improved gradation of the binder particles provided a lubricating effect through decreased aggregate interlocking.  The  flow resistance increased with time, reflecting the stiffening of the concrete, while the torque  viscosity did not show a clear trend with time.  Microfiller cements generally exhibited  performance equal to or better than a pure cement in terms of maintaining workability over time. In addition, microfillers reduced the induced bleeding of HPC  mixtures; again the finer the  microfiller the greater the effect. This should be reflected in fewer bleeding channels and fewer weak microstructural features at interfaces around coarse aggregates. It also seemed that the rheology of self-leveling mixtures could mainly be defined by the flow resistance, the torque viscosity only determines at what rate the flow is going to occur.  The  rheology of triple-blended cements containing ultrafine limestone and silica fume was also  investigated. It was demonstrated that it is possible to design triple-blended cements having superior rheological properties. A combination of an adequate superplasticizer to ensure optimal dispersion of ultrafine particles, and adequate microfillers to reduce the Coulomb friction between particles can dramatically enhance the workability of  The for  microfiller effect on mechanical properties of HPC  HPC.  was investigated both at early ages and  the long term. Investigations on mortar showed that up to around 10% to 15% replacement of  cement by limestone filler did not decrease the strength significantly. Triple-blended cements  Chapter IX  183  containing more than 18% limestone filler and 10% silica fume outperformed pure cement in terms of mechanical strength. Investigations on concrete confirmed that up to 15% filler replacement of cement did not decrease the long-term strength significantly. Indeed, at very low w/c ratios, using 15% ultrafine limestone increased the compressive strength. The finer the limestone filler, the more the strength was increased. Triple-blended cements containing 15% limestone microfiller and 15% silica fume had compressive strengths that compared to a binary cement containing 15% silica fume, and outperformed a pure cement by 20% at 28 days. This shows that up to 30% of the cement can be replaced by a combination of industrial byproducts and unprocessed grinding mill fillers, with considerable improvement in both workability and strength.  It was demonstrated that the microfiller effect depends on the w/b ratio. At a w/b ratio of 0.33, there is still significant porosity filled by water. When 15% of the cement is replaced by a microfiller, the improved particle packing may not be able to compensate for the reduction of the total hydration products. The result would be less dense hydration products, and a slight decrease in strength. At a very low w/c ratio of 0.25, porosity is lower and the improved packing of the system due to the microfiller is more significant compared to the reduction in the reactive binder. The net result would be an increased strength. When the non-pozzolanic filler is combined with a pozzolanic filler, both improved packing and compensation for the reactive cement are provided. The net result is a considerable increase in strength.  It was also demonstrated in this work that improvements imparted by microfillers are partly due to the increased homogeneity they provide to the cementitious system. Concretes mixed in a regular pan mixer and in a more efficient OMNI mixer were compared. All mixtures made using the OMNI mixer had significantly higher strengths compared to the reference mixtures. This was more significant in silica fume systems for which the degree of dispersion of the ultrafine particles seemed very important.  Limestone microfiller greatly increased the very early age strength of HPC, and this was more significant the finer the microfiller; 15% of a 0.7 um mean particle size limestone filler increased the strength by about an order of magnitude at 12h. This is very significant for winter construction and in cases where a quick removal of formwork is essential. One of the major drawbacks of fillers such as slag and fly ash is their detrimental effect on the early age strength,  Chapter IX  184  though their long-term effect is positive. If combined with a fine limestone filler, this problem can be overcome. Limestone filler would accelerate the early age strength, while fly ash and slag can compensate for the cement later on. Silica fume also increased the early age strength, though to a lesser extent.  It seems that at very early ages, some microfillers such as limestone represent energetically preferential substrates for the growth of calcium hydroxide. The removal of calcium ions from the solution would catalyze the dissolution of C3S in an attempt to achieve equilibrium between the hydration layers that surround the cement particles and the pore solution. Ultimately, this would initiate a re-crystallization of this protective membrane into a more permeable structure, which signals an earlier end to the induction period. The surfaces of the filler particles continue to be a preferred substrate for the growth of hydration products at later ages. The combination of these effects contributes to increasing the early age strength. Other non-carbonate fillers can also stimulate the early age hydration because they decrease the calcium potential in the solution, which also accelerates the dissolution of C3S.  The mechanisms of microfiller action in HPC were also addressed in this work from a microstructural point of view. Quantitative image analysis of back-scatter electron micrographs was carried out at early and later ages, both on cement pastes and concretes containing pozzolanic and non-pozzolanic microfillers. Unhydrated cement grains, porosity, calcium hydroxide, and in some instances inner and outer C-S-H were measured. It was shown that at early ages the unhydrated cement content was lower and the CH content was higher in limestone filler mixtures, which confirms the acceleration of the early age hydration due to carbonate additions. At later ages, the porosity was lower in microfiller cements. In addition, the calcium hydroxide particles were finer, perhaps due to a refinement effect of hydration products via their nucleation around fine filler particles.  The microfiller effect on the microstructure of the paste-aggregate transition zone was also analyzed. It was found that the unhydrated cement content at the vicinity of interfaces is lower. This is partly because larger cement grains migrate away from the high shear planes represented by the aggregates during mixing. Also the smaller cement grains in this region hydrate quickly due to a locally higher w/c ratio. There was a porosity gradient around interfaces, especially in the first 30 pm adjacent to the aggregate, and this was maintained even after 28 days. However,  Chapter IX  185  this effect was significantly reduced in microfiller cements due to a reduction of the wall effect and improved packing of the particulate system. Higher CH contents were not measured at interfaces, in contrast with results suggested by analysis of fracture surfaces, which may indicate that fracture surfaces are not representative of the true interface, and that instead weak planes rich in CH are exposed. It was also observed that densifying the interface alone might not increase strength. Strength is improved only when lower porosity is obtained without producing lower density hydration products.  This work also addressed the microfiller effect based on theoretical considerations of particulate mixtures. A microfiller efficiency factor was defined based on a calculation of the packing density of the binder particles and the microfiller effect on the kinetics of hydration reactions. A model which predicts the early age strength of HPC based on the knowledge of the microfiller efficiency factor was achieved and yielded satisfactory results.  Based on particle packing considerations, the author would recommend to move away from closed circle grinding of portland cement, which usually results in fairly uniform particle sizes, towards methods that result in a wider range of particle sizes and hence the potential for closer packing. For practical purposes, a simple air flow test on the dry cement powder would provide a suitable test to quickly estimate the packing density at cement plants.  Most previous models which relate microstructure to properties for cement based materials are based on relationships either between porosity and strength or between porosity and permeability. In this work, it was demonstrated that a decrease in the capillary porosity is not necessarily accompanied by an increase in the strength. The nature of the solid phase also seemed critical in determining mechanical properties. A model has been proposed to account for the nature of the hydration products in addition to porosity in predicting mechanical strength. Hydration products were divided into two distinct categories: inner hydration products which form within the space initially occupied by the cement grains, and outer hydration products. The model was applied to various HPC's and a close relationship was obtained.  Microstructural observations and model predictions of this study suggest that strength should not be developed through excessive addition of cement and a high rate of hydration. Rather, an initially dense packing of particles combined with a very low w/b ratio will lead to a low rate of  Chapter IX  186  hydration. Cement grains will only react on their surfaces to bind the discrete particles together. Hydration products will not migrate far away to fill in the porosity. Thus, denser products will be the result. Chemical contraction, self-desiccation and drying shrinkage will be limited, and lower internal stresses will be developed.  An effort has been made in this work to address the microfiller effect in high-performance concrete from a new perspective. An original approach has been adopted in trying to understand the effect of ultrafine particles, starting at the packing of the dry particulate mixture, going through the microfiller effect on rheology and flow properties, mechanical strength, and microstructural features. The most fundamental and quantitative tools available to the author have been used. Whenever it was feasible, quantitative relationships based on statistical experimental techniques have been established between the microfiller effect and properties.  However, research is a humbling experience; this work has raised more questions than it has answered, questions that clearly cannot all be addressed in a single thesis. Thus, this work should be pursued further. Concrete seems to be shifting from a particulate mixture to a dense suspension. The industry is targeting a self-leveling material that compacts under its own weight. To prevent segregation and bleeding, this material requires a high powder content. Significant microfiller volumes will be needed in the future for the production of this type of concrete. At the same time, on a planetary scale, generation of dust in mineral quarries, of ash in coal-burning power plants, of rice husks in agriculture, of slag in the metal industry, and of various other industrial wastes continues every day. When a true science based on the knowledge of the particulate nature of these wastes is developed, recycling these materials into concrete will not only be economic and environmental, but it can also bring about improvements in strength and durability.  The annual global production of concrete is about 5 billion tons. Mankind consumes only water in larger quantities. In view of demographic considerations and the human quest for sustainable development, the current cement production and concrete construction practices should be altered in the future to fulfill environmental requirements. The dream of high-performance multiblended cements containing recycled products can be achieved if a scientific understanding of the microfiller effect is reached. This work is a contribution towards achieving that dream.  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0050312/manifest

Comment

Related Items