Durability Performance of SCC Mortars Including Different ...Durability Performance of SCC Mortars Including DifferentTtypes of Metakaolin XII DBMC, Porto, PORTUGAL, 2011 3 0,000 1,000

Durability Performance of SCC Mortars Including Different Types of Metakaolin

Helena Figueiras 1 Sandra Nunes 2 Joana Sousa Coutinho 3

Joaquim Figueiras4 ABSTRACT The design of high performance and high durability concrete materials demands an adequate selection and combination of constituent materials. The present paper intends to study and compare the performance of two different types of metakaolin included in a ternary cement blend and to provide a comprehensive procedure for the design of mortar mixtures which are adequate for self-compacting concrete (SCC), with defined performance requirements in terms of durability. The methodology used holds firstly, an experimental phase conducted according to a central composite design; second, a statistical analysis and model fitting of data collected during the experimental phase and; third, a numerical optimization of mixture parameters using the derived models. Two different types of metakaolin were assessed in combination with cement, limestone filler and superplasticizer. Each type of metakaolin interacts in a different way with other constituents resulting in different fresh and hardened properties. Mortar characterization tests included the flow cone and V-funnel tests for mortar fresh state and the electrical resistivity test for the hardened material state. This study clearly shows the improvement in durability given by the incorporation of metakaolin in mortars. Nevertheless, it was found that both types of metakaolin perform differently. After building the regression models that establish the relationships between mix design variables and the responses (mortar tests results), numerical optimization was used to determine the mixture parameters which exhibit best overall performance (including economic), while maintaining self-compactability. KEYWORDS Metakaolin, Limestone filler, SCC mortars, Durability performance.

1 LABEST/FEUP Faculty of Engineering of the University of Porto (FEUP), Porto, PORTUGAL, [email protected] 2 LABEST/FEUP Faculty of Engineering of the University of Porto (FEUP), Porto, PORTUGAL, [email protected] 3 LABEST/FEUP Faculty of Engineering of the University of Porto (FEUP), Porto, PORTUGAL, [email protected] 4 LABEST/FEUP Faculty of Engineering of the University of Porto (FEUP), Porto, PORTUGAL, [email protected]

H. Figueiras, S. Nunes, J. Sousa Coutinho and J. Figueiras

2 XII DBMC, Porto, PORTUGAL, 2011

1 INTRODUCTION Currently, durability, and especially prevention of reinforcement corrosion, has become a critical aspect of reinforced concrete structures. The design of durable structures should focus on one hand, on the possibility of using additional protective measures, like the use of controlled permeability formwork (CPF) or self-compacting concrete (SCC), that could delay the corrosion process and the deterioration of concrete, on the other hand, the composition design should be appropriate for the exposure class. Sustainable design of highly durable concrete requires an adequate selection and combination of constituent materials, exploiting synergies between different cements and additions. Supplementary cementing materials have become an integral part of high strength and high performance concrete mix design. Pozzolanic materials including silica fume, fly ash, slag and metakaolin have been used in recent decades for developing high performance concrete with improved workability, strength and durability. Pozzolanic reactions change the microstructure of concrete and chemistry of the hydration products by consuming the released calcium hydroxide (CH) and production of additional calcium silicate hydrates (C-S-H), resulting in an increased strength and reduced porosity and therefore improving durability [Bentz & Garboczi 1991]. In recent years, there has been a growing interest in the use of metakaolin as a high reactive natural pozzolan. Metakaolin is a manufactured material produced by calcining kaolin clay at high temperature that essentially contains SiO2 (50-55%) and Al2O3 (40-45%). Metakaolin pozzolanic reactivity depends on the crystallinity, particle size and the amorphization degree. These aspects are interrelated and modifiable through the production technology. The efficiency of a metakaolin in a composition also depends on the type of cement used, in particular the fine particle size complementarity and the differential electrostatic forces between particles [Sampaio et al. 2001]. It has been reported that the use of metakaolin helps in enhancing the early age mechanical properties as well as long-term strength properties of concrete. The partial replacement of cement with metakaolin reduces the water penetration into concrete by capillary action. Metakaolin modifies the pore structure of the cement matrix and significantly reduces permeability, resulting in higher resistance to transportation of water and diffusion of harmful ions which lead to the deterioration of the matrix. Although total porosity may be increased by metakaolin blending, the process also causes a refinement of pore structure [Siddique & Klaus 2009]. This paper discusses the durability properties of self-compacting mortars with mineral admixtures. Portland cement, limestone filler and two different types of metakaolin were used in binary (two-component) and ternary (three-component) cementitious blends. A methodology was developed for the design of mortar mixtures which are adequate for SCC [Nunes et al. 2009]. This methodology was developed in three phases: first, the experimental phase conducted according to a central composite design; second, the statistical analysis and model fitting of data collected during the experimental phase and, third, the numerical optimization of mixture parameters using the models derived in the previous phase. The numerical models obtained were used to find optimal solutions that satisfy performance requirements, including: fresh state (deformability and viscosity), hardened state (durability properties) and economic requirements. 2 EXPERIMENTAL PROGRAMME 2.1 Materials characterization Mortars investigated in this study were prepared with binary mixtures, of Portland cement (CEM I 42.5R) and limestone filler, and ternary mixtures, including cement, limestone filler and metakaolin 1 (MTK1) or metakaolin 2 (MTK2). The specific gravity of cement, limestone filler, MTK1 and MTK2 was 3.11, 2.68, 2.20 and 2.55, respectively. Figure 1 shows the particle size distribution curves, in terms of volume, of each fine material used in this study. The mean particle size dimension was 14.61µm, 6.53µm, 7.12µm and 15.38µm for cement, limestone filler, MTK1 and MTK2, respectively.

Durability Performance of SCC Mortars Including DifferentTtypes of Metakaolin

XII DBMC, Porto, PORTUGAL, 2011 3

0,000

1,000

2,000

3,000

4,000

5,000

6,000

7,000

8,000

0,01 0,1 1 10 100 1000 10000

volu

me

(%)

particle size (µm)

CEM I 42.5Rfíler calcárioMTK1MTK2

Figure 1. Particle size distribution curves in terms of volume.

The main differences in particles shape of the two types of addition used in this study can be observed in Fig. 2. Metakaolin particles have a very irregular shape with a flaky structure (Figure 2right), while limestone filler particles are more angular (Figure 2 left). MTK1 metakaolin is produced in USA and is white in colour, making it particularly attractive for architectural applications, while MTK2 is produced in Portugal and has a pinkish tone. MTK1 is currently marketed in Portugal at a price about 3.3 times MTK2.

Figure 2. Backscattered electron images with a magnification of 5000x: limestone filler (left);

metakaolin MTK1 (right).

Siliceous round natural sand (0.08–2 mm) was used with a specific gravity of 2.63 and an absorption value of 0.3 %, which is a reference sand conforming to CEN EN 196-1. A polycarboxylate type superplasticizer with a specific gravity of 1.05 and 25.5% solid content was employed for binary mixtures and a polycarboxylate type superplasticizer with a specific gravity of 1.07 and 26.5% solid content was employed for ternary mixtures. 2.2 Experimental design To define the mortar composition three independent variables were selected, namely, water to powder volume ratio (Vw/Vp); water to cement weight ratio (w/c) and superplasticizer to powder weight ratio (Sp/p). For ternary mixtures it was necessary to include an additional variable to completely define the composition, which was metakaolin to cement weight ratio (mtk/c). In this study sand to mortar volume (Vs/Vm) was kept constant. The selection of these parameters, used widespread in the design of self-compacting concrete compositions, was based on the method developed by Okamura et al. [2000]. For binary mixes a 23 factorial statistical design [Montgomery 2001] corresponding to three parameters at two levels was carried out to establish models that describe key SCC properties and for the ternary mixes study a 24 factorial statistical design was adopted. To both experimental designs axial points and center points were added (Central Composite Design), allowing thereby a second-order model fitting. The effect of each factor was evaluated at five different levels −α, −1, 0, +1, +α. In order to make the design rotatable (i.e. the standard deviation of the predicted response is constant in all points at the same distance from the center of the design) the α value should be taken equal to



nf1/4, where nf is the number of points in the factorial part of the design [Montgomery 2001].

Experimental programs also included central point repetition in order to evaluate the experimental error associated to conditions and test procedures variability. Table 1 presents the experimental plans characterization for binary and ternary mixtures.

Table 1. Experimental plans characterization.

Designation Fine materials Experimental design

Independent variables

[-α; +α]

Experimental design A (ED0)

Cem I 42.5R limestone filler

23

α =1.692 nc=4 nf=8 na=6

Vw/Vp w/c Sp/p

[0.650; 0.900] [0.320; 0.510]

[0.550%; 0,650%]

Experimental design B (ED1)

Cem I 42.5R limestone filler MTK1

24

α =2.000 nc=6 nf=16 na=8

Vw/Vp w/c Sp/p mtk/c

[0.700; 1.000] [0.370; 0.490]

[0.985%; 1.030%] [0.50; 0.20]

Experimental design C (ED2)

Cem I 42.5R Limestone filler MTK2

[0.650; 0.950] [0.360; 0.480]

[0.978%; 1.023%] [0.05; 0.20]

nc: nº of central points; nf: nº of factorial points; na: nº of axial points

2.3 Response variables Given the main objectives of this study, the selected response variables were: flow diameter (Dflow); V-funnel test (Tfunnel) and resistivity at 28 days (Resistiv28d). The mortar flow test (Dflow) was used to assess deformability by calculating the flow diameter as the mean of two diameters in the spread area and the V-funnel test (Tfunnel) was used to assess the viscosity and passing ability of the mortar. These mortar tests using the flow cone and the V-funnel, with the same internal dimensions as the Japanese equipment were carried out to characterize fresh state (see [Okamura et al. 2000] for details on equipments and test procedures). The resistivity test was selected to evaluate the durability of mixtures because it is a very simple and rapid, nondestructive test which enables the microstructure characterization. The electrical resistivity provides indications on the pore connectivity and therefore, on the concrete resistance to penetration of liquid or gas substances, and so resistivity is a parameter which accounts for the main key properties related to concrete durability [Andrade 2004]. Resistivity is an intrinsic property of the material that relates to the ability of concrete to carry electric charge and it depends mainly on the nature and topography structure, moisture condition, temperature and on the ion concentration in the interstitial solution. Regarding the influence of the chemical composition of pore solution, Andrade [2004] stated that its impact in the total resistivity is small providing the concrete remains alkaline. At high pH values the pore solution resistivity varies from 0.3 e 1.0 Ω.m, which is comparatively very small taking into account that concrete resistivity after several days of hardening is in the range of several hundreds Ω.m. Mortar resistivity was assessed by imposing a current passing through a mortar prism, in saturated conditions, and measuring the electric resistance which can be used to compute resistivity as follows,

)/()/(/ ILAVALIVR ⋅⋅=⇒⋅== ρρ (1)

where R is the electric resistance, (Ω - Ohm); I, current (Amp.); V, voltage (Volts); ρ, electric resistivity (Ω⋅m); L, length (m); and A (m2) the cross area of the test specimen through which current passes. Resistivity measurements were performed on prismatic specimens (4×4×16cm), in which stainless steel networks were embedded to work as electrodes. Mortar prisms were demoulded one day after casting and kept under water in a chamber under controlled environmental conditions (Temp.=20ºC and HR=95-98%) until testing. Since all the specimens were at the same moisture



(saturated) and temperature conditions, the resistivity measure can be used to evaluate the porous structure, and therefore constitute a measure of the amount and interconnectivity of the cementitious matrix pores. 2.4 Response models For each response variable a quadratic model can be estimated from the central composite design data. The model parameters are estimated by means of a multilinear regression analysis. The generic form of a second order model is presented in Eq. 2, where y is the response of the material; xi are the independent variables; β0 is the independent term; βi, βii and βij are the coefficients of independent variables and interactions, representing their contribution to the response; ε is the random residual error term representing the effects of variables or higher order terms not considered in the model [Montgomery 2001].

ε+∑∑β+∑β+∑β+β=<==

jiji

ij2ij

k

1iiiij

k

1ii0 xxxxy

(2)

3 FITTED MODEL ANALYSES A commercial software (Design-Expert) was used to analyse the results for each response variable by examining summary plots of the data, fitting a model using regression analysis and ANOVA, validating the model by examining the residuals for trends, autocorrelation and outliers and interpreting the model graphically [Nunes et al. 2009]. Fitted models presented relatively high correlation coefficients (> 0.95) which indicates their good adjustment. The derived models also have good accuracy because the residual standard deviation and the standard deviation calculated from the central points (experimental error) were very close. The assumptions that the residuals are independent and normally distributed with zero mean and constant variance were also validated for all fitted models. Due to limitations of space, fitted models obtained for the fresh state (Dflow and Tfunnel models) will not be presented in the current paper. The results of the estimated models for the resistivity at 28 days (Resistiv28d), including the residual error term, along with the correlation coefficients, are given in Table 2.

Table 2. Resistivity fitted numerical models (coded values of parameters) and relative influence of each parameter on the variation of response.

Response variable ED0 - R28d (ohm.m) (%)

ED1 - R28d0.5 (ohm.m) (%)

ED2 - R28d-0.5 (ohm.m) (%)

Model terms: estimate independent 48,89 16.73 0.094

Vw/Vp -3.010 -20.6 -0.510 -9.8 0.003 11.8 w/c -8.518 -58.3 -0.903 -17.4 0.004 17.6 Sp/p -0.785 -5.4 NS - NS - mtk/c NS - 2.335 44.9 -0.015 -57.7

(Vw/Vp)×(w/c) 0.762 5.2 NS - NS - (Vw/Vp)×(Sp/p) NS - NS - NS - (Vw/Vp)×(mtk/c) NS - -0.513 -9.9 NS -

(w/c)×(Sp/p) NS - NS - NS - (w/c)×(mtk/c) NS - NS - 0.001 4.5 (Sp/p)×(mtk/c) NS - NS - NS -

(Vw/Vp) 2 NS - 0.170 3.3 NS - (w/c)2 1.525 10.4 NS - NS - (Sp/p)2 NS - NS - NS - (mtk/c)2 NS - -0.764 14.7 0.002 8.4

Residual error* std. deviation

1.189 0.039 0.002

R2 / R2Adj 0.986/0.980 0.977/0.971 0.991/0.989

(NS) non-significant terms; (*) error term is a random and normally distributed variable with mean zero



The model coefficients presented in Table 2 give an indication of the relative influence of each mixture parameter on the Resistiv28d response variable. Naturally, higher values indicate greater influence of this parameter in the response and on the other hand, a negative value reflects a response decrease to an increase in this parameter. Observation of the table shows that, in the case of the experimental design without metakaolin (ED0), the water/cement ratio (w/c) is the variable that has a greater influence on the resistivity, while in experimental designs including metakaolin (ED1 and ED2) the metakaolin/cement ration (mtk/c) is the most influential variable. It is noted, however, that these two variables have opposite effects on the response. In the experimental plan ED0, beyond the great influence of the w/c variable, Vw/Vp and Sp/p parameters have some influence on Resistiv28d, 20% and 5%, respectively. In the experimental plans including metakaolin as mineral admixture (ED1 and ED2) the influence of most relevant variables was quite similar. The range of resistivity results of the mortars obtained with the three experimental plans was significantly different, varying between 39.1 e 67.2 ohm.m in ED0, between 87.4 e 437.3 3 ohm.m in ED1 and between 55.7 e 200.7 ohm.m in ED2. In order to highlight the differences between compositions including MTK1(ED1), including MTK2 (ED2) and without MTK (ED0), the chart shown in Fig. 3 presents the values obtained in the mortars for the three experimental plans (measured values) and the predicted values based on the Resistiv28d model derived for the ED2 plan (predicted values).

0

50

100

150

200

250

300

350

400

450

500

0 50 100 150 200 250 300

mea

sure

d va

lues

(oh

m.m

)

predicted values (ohm.m)

ED0ED1ED2

Figure 3. Resistiv28d predicted values based on the model derived for the ED2 plan versus measured

values for plans ED0, ED1 and ED2. From the analysis of this figure it can be concluded that the behavior of the three mixtures, binary mixture (CEM + filler), ternary mixture 1 (CEM + filler + MTK1) and ternary mixture 2 (CEM + filler + MTK2), with respect to the resistivity is fairly different. The adjusted model for the compositions including limestone filler and MTK2 (ED2) is not adequate to predict the behaviour of the compositions including MTK1 (ED1) and, although not as evident, the same seems to happen in the compositions including only limestone filler (ED0). The resistivity behaviour of mixtures including only cement and limestone filler (binary mixtures) is different from the behaviour of mixtures including metakaolin (ternary mixtures). Besides, the two metakaolins analysed performed differently), achieving higher values of resistivity, therefore better durability, in compositions including MTK1. According to Siddique and Klaus [2009] the partial replacement of cement by a pozzolanic material, like metakaolin, causes substantial changes on the pore structure of the mortar and on the chemistry of the hydration products. The metakaolin contributes to the microstructure improvement by the filler effect (like limestone filler) and by the pozzolanic reaction of metakaolin with calcium hydroxide. The different performances achieved with MTK1 and MTK2 may be due to the different particle size distribution and different origins and production systems of these metakaolins (different chemistry).



4 MIXTURE OPTIMIZATION After building the regression models that establish relationships between mix design variables (Vw/Vp, w/c, Sp/p e mtk/c) and the responses (Dflow, Tfunnel and Resistiv28d) the numerical optimization technique was used to determine a mortar composition that meets certain performance requirements defined in advance. These requirements may be set in terms of fresh and/or hardened state properties. With regard to the fresh state, it was defined that the optimal mortars from the standpoint of self-compactability should exhibit a Dflow between 260 e 280 mm and a Tfunnel between 8 e 12 sec, simultaneously. If a requirement related to resistivity is also imposed, say greater than 120 ohm.m, then self-compactable mortars based on cement and limestone filler (ED0) never fulfil this requirement (Fig.3). The set of optimized solutions based on the defined performance requirements: 260 mm ≤Dflow≤280 mm; 8 s ≤Tfunnel≤12 s and Resistiv28d>120 ohm.m, for mixtures including metakaolin (ED2 and ED3), is presented in the graphs below (Fig. 4).

0,050

0,075

0,100

0,125

0,150

0,175

0,200

0,37 0,39 0,41 0,43 0,45 0,47 0,49

mtk

/c

w/c

0,050

0,075

0,100

0,125

0,150

0,175

0,200

0,36 0,38 0,4 0,42 0,44 0,46 0,48

mtk

/c

w/c (a) (b)

Figure 4. Range of values of the variables w/c and mtk/c for optimized mixtures: (a) ED1; (b) ED2. By adding an economic constraint (minimizing cost) to the optimization criteria mentioned above, one single solution is obtained for each metakaolin type, identified in Figure 4. Table 3 shows the final optimized composition for each case, the respective fresh and hardened properties and the relative cost. Although both solutions have quite different w/c ratios, their performance in terms of fresh state properties and resistivity (durability) is very similar (see table 3). Nevertheless, the mixture including MTK2 exhibits lower cost.

Table 3. Mixture composition, fresh and hardened properties and the relative cost of optimized compositions

Composition 1 (with MTK1) Composition 2 (with MTK2)

kg/m3 relative cost kg/m3 relative cost

CEM I 42.5R 587 48.8% 487 40.5% limestone filler 274 7.3% 316 8.4%

MTK1 32 21.9% - -

MTK2 - - 73 15.5%

superplasticizer 8.8 22.0% 8.6 21.3%

Relative total cost 100% 85.7%

Dflow 279.7 mm 279.9 mm

Tfunnel 12.0 sec 12.0 sec

Resistiv28d 120.2 ohm.m 123.0 ohm.m



5 CONCLUSIONS An experimental plan conducted according to a factorial design is useful to evaluate the effects of constituent materials and their interactions on SCC-mortar properties. Data collected during the experimental plan can be used to establish numerical models relating mixture parameters with the SCC properties of interest. Mortars including binary mixes (cement + limestone filler) and ternary mixes (cement + limestone filler + metakaolin) were study. Two different types of metakaolin were used in order to compare their performance. In this research, besides fresh state properties, deformability and viscosity, a durability related parameter (resistivity) was also analysed and mathematically modelled. The beneficial effect of metakaolin on durability indicators is clearly showed in this study. The variable mtk/c exhibited the greatest effect on resistivity. For mortars including only cement and limestone filler as fine materials, w/c was the mixture parameter that most influenced resistivity (durability). Furthermore, Vw/Vp also had a significant influence on resistivity. The superplasticizer to powder weight ratio (Sp/p) exhibited a reduce influence on resistivity. Other interaction and quadratic effects were found to be significant to explain SCC mortar durability property. The two types of metakaolin performed differently, which may be due to the different particle size distribution and different origins and production systems (different chemistry). Numerical models established for the mortar mixtures can simplify the test protocol required to optimize a given SCC mixture, namely, to select the combination of powder materials with admixtures and to design a SCC mixture with defined performance requirements, in terms of fresh properties, hardened properties and economic constraints. For the optimization example presented in this paper it can be concluded that including MTK2 in the composition it is possible to satisfy all the requirements of fresh and hardened state (durability) with a lower cost than using MTK1. Compositions where only limestone filler was included were unable to meet the durability requirement (resistivity greater than 120 ohm.m). ACKNOWLEDGMENTS This research was financed by FCT – Portuguese Foundation for Science and Technology Research Project PTDC/ECM/70693/2006 and supported by PhD Research Grant SFRH/BD/25123/2005. Collaboration and materials supplying by SECIL, SIKA, COMITAL and Cerâmica Condestável is also gratefully acknowledge. REFERENCES ANDRADE, C. 2004, Calculation of initiation and propagation periods of service life of reinforcement by using the electrical resistivity, International symposium on Advances in Concrete through Science and Engineering, RILEM.

Bentz D, Garboczi E., 1991, Simulation studies of the effects of mineral admixtures on the cement paste-aggregate interfacial zone, ACI Materials Journal 88(5):518-529.

Montgomery, D. 2001, Design and analysis of experiments. 5ª ed. Wiley, New York.

Nunes S, Milheiro Oliveira P, Sousa Coutinho J & Figueiras J. 2009, Interaction diagrams to assess SCC mortars for different cement types, Construction and Building Materials, 23:1401–12.

Okamura H, Ozawa K & Ouchi M. 2000, Self-compacting concrete. Structural Concrete, 1:3–17.

Sampaio J, Coutinho J S, Sampaio N, 2001, Melhoria do desempenho de betões pelo metacaulino, 43º Congresso Brasileiro do Concreto, Instituto Brasileiro do Concreto, Brasil.

Siddique R & Klaus J, 2009, Influence of metakaolin on the properties of mortar and concrete: A review, Applied Clay Science, vol. 43, n. 3-4, p. 392.

Documents

Durability Performance of SCC Mortars Including Different ...Durability Performance of SCC Mortars Including DifferentTtypes of Metakaolin XII DBMC, Porto, PORTUGAL, 2011 3 0,000 1,000