POZZOLANIC REACTIVITY OF GROUND GRANULATED BLAST FURNACE SLAG IN BLENDED CEMENT

Transcription

1 POZZOLANIC REACTIVITY OF GROUND GRANULATED BLAST FURNACE SLAG IN BLENDED CEMENT Will Hansen (1), Yanfei Peng (2), Claus Borgnakke (1), Yousef Nouri (1) and Joseph J. Biernacki (3) (1) University of Michigan Ann Arbor, MI 489, USA (2) US Gypsum Co., USA (currently) (3) Tennessee Technological University, Cookeville, TN 38, USA Abstract The pozzolanic reactivity of ground granulated blast furnace slag (GGBFS) with a Type I portland cement was investigated for different GGBFS contents (, and 7 percent by weight) and curing temperatures (ºC, 23ºC, ºC) and hydration times (. hours to 9 days). Evaluation was based on isothermal heat of hydration measurements and thermal gravimetric analysis (TGA). The pozzolanic reaction in the blended cement was found to be a substantial part of total heat development and uniquely related to hydration of the portland cement, increasing in a nonlinear manner with hydration level. Pozzolanic reactions were found to develop primarily at later stages of hydration. 1. INTRODUCTION Pozzolanic reactions in blended cements containing supplementary cementitious materials (SCM) are slow compared to the rate of hydration of Portland cement [1]. Therefore, concretes containing SCM s are more sensitive to curing conditions. Although the pozzolanic reactions can be activated by alkalis and/or heat, the pozzolanic reactivity of a given SCM is not so well-established as to allow prediction of performance without extensive testing [2]. The challenge in utilizing SCM s (e.g. ground granulated blast-furnace slag (GGBFS), fly ash, and silica fume) in blended cements is in predicting the pozzolanic reactivity of an SCM in a given portland cement, and for different SCM contents and different temperatures. In the classic work by W. Lerch, published in 1946, the optimum sulfate content in portland cement was clearly identified by isothermal calorimetry heat profiles [3]. Isothermal calorimetry has also been used by J. P. Skalny and J. F. Young to study the mechanisms of portland cement hydration [4]. In this study heat of hydration is used to quantify the contribution of GGBFS to the total heat of hydration and the pozzolanic heat of reaction in blended cement for a range of GGBFS contents and temperatures.

2 2. EXPERIMENTAL PROGRAM 2.1 Materials Blended cement systems with varying GGBFS contents (,,, and 7 percent by weight) were investigated. The fineness and chemical composition of the portland cement and the GGBFS are listed in Table 1. Table 1. Fineness and chemical composition % by weight Type I Portland cement GGBFS SiO Al 2 O Fe 2 O CaO MgO Na 2 O..28 K 2 O Cl.3 SO Total as Oxides C 3 S C 2 S C 3 A C 4 AF Total as N/A Clinker Phases Blaine (cm 2 /g) Isothermal calorimetric tests Heat of hydration measurements started about minutes after mixing for pastes of.4 water/binder (w/b) ratio, where the binder is the sum of the cement and slag weights. Isothermal heat evolution curves of the blended systems were measured at three temperatures (, 23, and ºC) continuously for up to three weeks using a TAM Air model by Thermometric Inc. This unit consists of eight dual channels. Each channel, holding a ml glass ampoule, consists of a sample vessel and a reference vessel. Prior to mixing, the cement, GGBFS, and water (deionized) were equilibrated at the test temperature. After mixing the preconditioned raw materials, a spatula was used to add the paste to the ampoules, which were then sealed and placed inside the measuring vessel. A reference ampoule, containing a dry cement sample of the same total heat capacity as the test sample, was placed in the reference vessel. The calorimetric aluminum block containing the 16 channels is housed in an air-thermostat maintained at ±.2 K. The heat flux drift of the Peltier sensing unit is less than ± µw over a 24-hour period. Two samples were run for each mix. A Mettler model 81 thermal analyzer (TGA/SDTA) was used to determine bound water content for the portland cement pastes versus time of hydration at three isothermal temperatures (ºC, 23ºC and ºC). Paste samples were cut from a block, then crushed in a

3 mortar and pestle and placed inside a sealed container in a methanol solution in order to terminate hydration. Again, two samples were tested, and the results were averaged. The degree of hydration in the portland cement system was calculated assuming a total bound water content of.23 g/g ignited cement corresponding to % hydration of portland cement. 3. RESULTS AND DISCUSSION 3.1 Heat of hydration of portland cement Calorimetric results in Figure 1-a demonstrate excellent reproducibility between two test samples and long-term sensor stability. Furthermore, the rate results follow expected trends for temperature effects, in that higher initial rates of hydration develop with increasing temperature, while at later ages a crossover is found. The crossover effect is due to termination in hydration associated with higher initial rates, probably due to particle-size effects. The rate curves span nearly three orders of magnitude over a time-period of hours (Figure 1-b). 3.2 Kinetics of hydration A number of kinetics models have been used to predict the rate of hydration versus time and temperature for the acceleration and deceleration stages of hydration. Amongst them are the Avrami Nucleation and Growth model and the Jander based diffusion models [-8]. Cumulative heat of hydration from calorimetric measurements can be modeled using the Freiesleben-Hansen and Pedersen three-parameter equation [9]: τ a Q = Qmax exp[ ( ) ] (1) t where Q, = heat of hydration Q max τ = time characteristic parameter a = curvature parameter Using the Q max predicted from the above equation, one can approximate the degree of hydration using the relative heat of hydration, Q / Qmax. Thus, acceleration followed by deceleration in the nucleation and growth stage can be explained by the mathematical model introduced by Avrami: where Q / Qmax max n = ( Q / Q ) [1 exp( kt )] (2) ter Q / Qmax = relative heat development representative of degree of hydration at time t ( Q / Qmax ) = termination value for nucleation-growth (Q/Q max.46) ter

4 Typical rate of hydration for portland cement paste (w/c =.4) ºC 23ºC ºC Time, hrs Sample A ºC Sample A 23ºC Sample A ºC Sample B ºC Sample B 23ºC Sample B ºC (a) 1.1 Typical rate of hydration for portland cement paste (w/c =.4) ºC ºC.1 Time, hrs Sample A ºC Sample A 23ºC Sample A ºC Sample B ºC Sample B 23ºC Sample B ºC k, n = modeling parameters t = hydration time The rate form of Avrami s model is: (b) Figure 1: Typical rate of heat evolution curves, for a.4 w/c ratio portland cement paste. dq dt = ( Q / Q max ) ter. k n t n 1 exp( kt n ) (3)

5 In order to explain the diffusion part of the cement hydration process, Jander s diffusion model is used: 1/ 3 [1 (1 Q / Q ) ] Kt N max = (4) where Kand N are modeling parameters. The modeled rate curves in Figure 2-a to 2-c are plotted as a function of the relative heat ratio, which is a quantitative measure of hydration. The curves are plotted as a function of time in Figures 2-d to 2-f. It can be observed that the selected models, starting with nucleation and growth and followed by diffusion, describe the major hydration processes well. The Nucleation and growth stage seems to be completed at about 4-% relative heat of hydration []. 3.3 Pozzolanic reactivity of slag in a portland cement. In the present approach, the observed heat evolution (Q) is assumed to be sum of the OPC contribution (Q OPC ), scaled by the weight fraction of OPC (f OPC ) in the blend, and the pozzolanic contribution to heat evolution (Q pozz ). Since the hydraulic heat contribution by the GGBFS was found to be less than 1% of that of the portland cement, it was ignored. Q= Q OPC f OPC + Q pozz - SCM () The primary assumption is that the presence of SCM does not accelerate or decelerate the OPC hydration process significantly. One might argue that scaling the OPC heat of hydration would imply a change in the w/b ratio. For instance, in a blended cement with % slag content and w/b=.4, if the slag is removed while the amount of water and OPC are kept constant, the resultant system would have a w/c ratio of.6. However, the effect of the w/b ratio (.6 and.9) on heat development was found to be insignificant in this range, so no correction for the w/b ratio was applied. The pozzolanic reactivity in blends of, and 7 percent GGBFS by weight was determined by scaling the hydration curves for the % portland cement by 7%, % and %, respectively. An example of the scaling method is shown in Figure 3. It was found that the slag pozzolanic reaction is thermally activated, and that the extent of the pozzolanic heat development is substantial compared to the OPC heat of hydration curves scaled for OPC content. Furthermore, at ºC there is an optimum slag content, while at lower temperatures an optimum cannot be observed as the pozzolanic reaction at lower curing temperatures is too slow to reach a maximum within 9 days. Using the scaling method, the pozzolanic heat development curves are plotted in Figure 4-a to 4 c) for the three slag contents (,, and 7%) and three temperatures (, 23 and o C). The modeled curves are shown in dashed lines. These curves illustrate the temperature sensitivity for the pozzolanic reaction and that they have a crossover effect similar to hydration of portland cement. Furthermore, it appears that the pozzolanic reaction is delayed as compared to the portland cement hydration curves shown in Figure. Whereas a major portion of the portland cement hydration develops within the first 1-2 days, relatively little pozzolanic reactivity is found in this time period. This suggests that a certain amount of hydration products is needed in order for the pozzolanic reaction to proceed. To further illustrate this, the degree of hydration was obtained from TGA measurements on the OPC system at the three different temperatures. The results in Figure 6 are consistent with expected trends. A-6% hydration level is achieved

6 Nucleation OPC Rate of Hydration at C Diffusion Nucleation OPC Rate of Hydration at 23C Diffusion Nucleation OPC Rate of Hydration at C Diffusion Relative Heat Development (Q/Qmax) Avrami Nucleation Model Jander Diffusion Model Relative Heat Development (Q/Qmax) Avrami Nucleation Model Jander Diffusion Model Relative Heat Development (Q/Qmax) Avrami Nucleation Model Jander Diffusion Model (a) (b) (c) OPC Rate of Hydration at C OPC Rate of Hydration at 23C OPC Rate of Hydration at C Nucleation Diffusion Nucleation Diffusion Nucleation Diffusion Time, hrs Avrami Nucleation Model Jander Diffusion Model Time, hrs Avrami Nucleation Model Jander Diffusion Model Time, hrs Avrami Nucleation Model Jander Diffusion Model (d) (e) (f) Figure 2: Typical rate of hydration curves for OPC at, 23 and o C and a w/c ratio of.4: rate of heat generation as a function of relative heat of hydration (a-c) and rate of heat generation as a function of time (d-f).

7 7% OPC +% GGBFS % OPC +% GGBFS % OPC +7% GGBFS 4 Pozzolanic Effect 4 Pozzolanic Effect 4 Q (J/g binder) Q (J/g binder) Q (J/g binder) Pozzolanic Effect.1 1 Time, days C.7xOPC ( C) 23 C.7xOPC (23 C).1 1 Time, days C.xOPC ( C) 23 C.xOPC (23 C).1 1 Time, days C.xOPC ( C) 23 C.xOPC (23 C) (a) (b) (c) Figure 3: Effect of pozzolanic reaction at 23 and o C shown as the difference between the total generated heat and the OPC hydration curve scaled by weight fraction Pozzolanic Slag Contribution for % Slag content Pozzolanic Slag Contribution for % Slag content Pozzolanic Slag Contribution for 7% Slag content.1 1 Time, Days.1 1 Time, Days.1 1 Time, Days C 23 C C C Predicted 23 C Predicted C Predicted C 23 C C C Predicted 23 C Predicted C Predicted C 23 C C C Predicted 23 C Predicted C Predicted (a) (b) (c) Figure 4: Heat development due to pozzolanic reaction for different slag contents at, 23 and o C (a-c) (the dashed lines are predicted using Friesleben-Hansen equation(4))[9].

8 within one to three days, depending on the temperature. This time period is consistent with the onset of major pozzolanic reactions, as seen from Figure 7, when plotting the heat of hydration for the OPC system and the pozzolanic heat development versus extent of OPC hydration for different slag contents. The fact that the different temperature curves for the pozzolanic heat development fall nearly on one curve, as do the OPC curves, suggests that pozzolanic reactivity has similar temperature sensitivity to the OPC hydration and that the pozzolanic reaction is controlled by the extent of the OPC hydration. Furthermore, the fact that pozzolanic reactivity is pronounced after 6% hydration suggests that this reaction is diffusion-controlled. Total Heat of Hydration for Type I Portland Cement Time, Days C 23 C C Figure : Cumulative heat of hydration for Type I portland cement paste (w/c =.4) at three different temperatures as a function of time Hydration of OPC (w/c =.4) Degree of Hydration (α TGA) Time, days ºC 23ºC ºC Figure 6: Changes of degree of hydration of portland cement (obtained from TGA) vs. time

9 4 Heat of Hydration of Portland Cement (w/c=.4) Q, J/gr Degree of Hydration (atga) ºC 23ºC ºC Figure 7: Heat of hydration for Type I portland cement (w/c =.4) at, 23 and o C as a function of degree of hydration Pozzolanic Slag Contribution at % Slag content Degree of OPC Hydration, percent (TGA wb/.23) C 23 C C Pozzolanic Slag Contribution at % Slag content Degree of OPC Hydration, percent (TGA wb/.23) C 23 C C Pozzolanic Slag Contribution at 7% Slag content Degree of OPC Hydration, percent (TGA wb/.23) C 23 C C (a) (b) (c) Figure 8: Heat development due to pozzolanic reaction for different slag contents at, 23 and o C

10 Results in Figures 7 and 8-a to 8-c suggest that the products of reaction are invariant with temperature, although the non-linearity in Figure 7 suggests that the stoichiometry of the reaction products, i.e. the concentration of CSH, changes slightly as a function of degree of hydration and that it is more pronounced at the later stages in the hydration process. 3.4 Optimum slag content in a portland cement. Heat development is directly correlated with the degree of cement hydration (Figure 7). Degree of hydration is in turn directly related to compressive strength development for a given concrete mix [11]. The pozzolanic reactions create more C-S-H. Thus, total heat of reaction in a blended cement can be a measure of its compressive strength. Meanwhile pozzolanic reactions not only improve strength but also improve alkali-aggregate interface durability [12]. Total heat curves follow expected trends for compressive strength development in blended slag-cement systems [12]. For the system of OPC and slag used in this study, a blended system with a slag content of to percent is expected to yield higher strength than the % OPC cement system at ºC within about 7 days due to thermal activation of the slagcement pozzolanic reaction, while at ambient temperatures and below, an optimum slag content was not found within 9 days, as seen from Figures 9-a to 9-c. The optimum slag content for pozzolanic contribution is closer to % at a high curing temperature (ºC). Without thermal activation no optimum can be found within 9 days as seen from Figures - a to -c.

11 Total Heat, J/g binder 4 C Total Heat, J/g binder 4 23 C Total Heat, J/g binder 4 C % % % 7% % Slag Content, % 1 DAY 3 DAYS 7 DAYS 28 DAYS 9 DAYS % % % 7% % Slag Content, % 1 DAY 3 DAYS 7 DAYS 28 DAYS 9 DAYS % % % 7% % Slag Content, % 1 DAY 3 DAYS 7 DAYS 28 DAYS 9 DAYS (a) (b) (c) Figure 9: Total heat of hydration for blended systems containing slag contents of to 7% at, 23 and o C Pozzolanic Slag Contribution (per gr-binder) at C Pozzolanic Slag Contribution (per gr-binder) at 23C Pozzolanic Slag Contribution (per gr-binder) at C 7 Slag Content, percent 9 Day 28 Day 21 Day 7 Day 3 Day 1 Day 7 Slag Content, percent 9 Day 28 Day 21 Day 7 Day 3 Day 1 Day 7 Slag Content, percent 9 Day 28 Day Day 7 Day 3 Day 1 Day (a) (b) (c) Figure : Pozzolanic contribution to heat at, 23 and o C and different slag contents (the dashed lines are predicted using Friesleben-Hansen equation (4))[12]

12 4. CONCLUSIONS The pozzolanic reaction between GGBFS and a portland cement was investigated for different blends (,, and 7 percent GGBFS by weight) and temperatures (, 23, and o C) using isothermal calorimetry and TGA for ages ranging from minutes to 21 days experimentally, and up to 9 days using predictions. The major conclusions are: Pozzolanic reactivity of the slag-cement systems was found to be a substantial portion of total heat development relative to the heat of hydration of the neat portland cement. The pozzolanic reaction was found to be uniquely related to the degree of hydration of the portland cement, increasing non-linearly with increasing hydration levels. The major pozzolanic reaction develops at hydration levels greater than about 6%. The pozzolanic conversion of hydration products is thermally activated in a similar way as the hydration process. ACKNOWLEDGEMENTS This work was sponsored by the National Science Foundation (NSF) project No. F12939 titled, Multi-Scale Kinetics-Based Model for Predicting Mechanical Property Development of Concrete Containing Supplementary Cementitious Materials. Nathan Banka, who received a fellowship from the NSF-REU program, is thanked for help in editing. REFERENCES [1] Detwiler, R. J., Bhatty, J.A., and Bhattacharja, Supplementary Cementing Materials for Use in Blended Cements, Portland Cement Association, Research and Development Bulletin RD112T, 1996, pp [2] Lawrence P., Cyr M., and Ringot, E., Mineral Admixtures in mortars effect of type, amount and fineness of fine constituents on compressive strength, Cement and Concrete Research,, pp [3] Lerch, W., The Influence of gypsum on the hydration and properties of portland cement pastes, Proceedings, Vol. 46, of the American Society for Testing Materials, [4] Skalny, J.P., and Young, J.F., Mechanisms of Portland Cement Hydration, 7 th International Congress on the Chemistry of Cement, Paris, 198, Vol. I., pp. II-1/1-1/4. [] Bezjak, A., Nuclei growth model in kinetic analysis of cement hydration, Cement Concrete Research, Vol. 16, 1986, pp [6] Gartner, E.M., Gaidas, J.M., Hydration Mechanisms, I, Materials Science of Concrete I, Edited by Jan P. Skalny, The American Ceramic Society, 1989, pp [7] Krstulovic, R., Dabic, P., A conceptual model of the cement hydration process, Cement and Concrete Research, Vol.,, pp [8] Van Breugel, K., Simulation of Hydration and Formation of Structure in Hardening Cement- Based Materials, Chapter, Cement Hydration and Formation of Microstructure Kinetics and Rate Formulae, Delft University Press, 2 nd Ed., [9] Freiesleben-Hansen, P. and Pedersen, E., J., Maaeleinstrument til control af betons haerdning, (in Danish), Nordisk Betong, Vol. 1, 1977, pp.1-21.

13 [] Peng Y., New models for predicting hydration and maturity development in blended cementitious systems PhD thesis, University of Michigan Ann Arbor, 6. [11] Byfors, J., Plain concrete at early ages, Swedish Cement and Concrete Research Institute, No. 8, 198, pp [12] Roy, D. M, and Idorn G. M. Hydration, Structure and Properties of Blast Furnace Slag Cements, Mortars and Concrete, ACI Journal, v 79, N 6, Nov-Dec, 1982, p