Cement & Concrete Composites 26 (2004) 417–426 www.elsevier.com/locate/cemconcomp Early-age heat evolution of clinker cements in relation to microstructure and composition: implications for temperature development in large concrete elements Y. Ballim *, P.C. Graham School of Civil and Environmental Engineering, University of the Witwatersrand, Private Bag 3, WITS, 2050 Johannesburg, South Africa Abstract This paper presents an assessment of the range and extent of variation of heat evolution of nominally similar cement clinkers from a range of cement production facilities in South Africa. Clinker samples were collected at nine cement plants and cements were prepared by grinding each clinker with a uniform quality of gypsum. X-ray fluorescence and optical microscope techniques were then used to characterise each clinker and cement in terms of chemical composition and cement compound morphology. Concretes were then prepared with the laboratory-manufactured cements and these were tested in an adiabatic calorimeter in order to determine the rate of heat evolution from each of the clinker samples. The results of these tests were related to the chemical and morphological characteristics of the corresponding cement clinkers. The results indicate a clear differentiation of clinker cements into low, medium and high heat cements. The relationships between this classification of the heat performance of the cements and the chemistry and morphology of the clinker is not clear at this stage. However, using a finite difference heat model, the paper presents an indication of the implications of the measured heat characteristics of the cement for early-age temperature distributions in large concrete elements. � 2003 Elsevier Ltd. All rights reserved. Keywords: Cement; Clinker; Chemistry; Morphology; Fineness; Heat; Microscopy; Modelling; Mass concrete 1. Introduction The thermal response of concrete due to hydration of cement is a predominant factor in the potential for early-age cracking of large concrete elements. An anal- ysis of this cracking potential requires an ability to quantify both the amount of heat that is evolved by the cement as well as the rate at which this heat is evolved [1]. Both these parameters are strongly influenced by the chemical and mineralogical composition of the cement, insofar as it affects the kinetics of the hydration reac- tions of cement. Furthermore, clinker morphology has been shown [2] to influence the compressive strength and, by inference, the hydration development of cement. Clearly, an ability to estimate the thermal response of cement in concrete, based on a knowledge of clinker characteristics would be of assistance to mass concrete *Corresponding author. Fax: +27-11-339-1762. E-mail address: ballim@civil.wits.ac.za (Y. Ballim). 0958-9465/$ - see front matter � 2003 Elsevier Ltd. All rights reserved. doi:10.1016/S0958-9465(03)00064-7 designers and cement manufacturers seeking to control temperature development in such structures. The kinetics of Portland cement hydration and as- sociated rates of heat evolution have been commented on by a number of authors [3–8]. Earlier work [3–6] favoured the view that the total heat liberated by a ce- ment is the mass weighted sum of the heat liberated by the individual pure crystallographic components. In this regard, Copeland et al. [5] comment that ‘‘the contri- bution of each component of a completely hydrated cement paste is, within experimental error, the heat of hydration of the pure components’’. However, later work by van Breugel [7] and Kishi et al. [8] argues that the concept of independent hydration of the clinker components requires correction to account for interac- tion between the cement components during hydration. This concept of component interaction during hydration has been adopted in much of the numerical models de- veloped to simulate the hydration process and associ- ated heat evolution [7–9]. The research reported in this paper was undertaken to quantify the range and extent of variation in the rates of mail to: ballim@civil.wits.ac.za Table 1 Chemical composition of the gypsum used in the cements and the amount of gypsum added to each clinker % by mass Clinker iden- tification % Gypsum added (by mass) CaO 30.20 A 3.683 SiO2 6.90 B 4.910 Fe2O3 0.44 C 5.031 Al2O3 0.60 D 5.010 MgO 0.80 E 5.147 TiO2 0.04 F 4.490 Mn2O3 0.02 G 4.984 K2O 0.19 H 5.120 Na2O 0.25 K 4.490 SO3 41.00 P2O5 0.00 LOI 18.25 418 Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 heat evolution of nominally similar cement clinkers in South Africa. As a secondary objective, the research was aimed at assessing if the observed variations in heat evolution characteristics can be explained by differences in clinker chemistry and mineralogy, based on chemical and microscopic evaluation of the clinkers. Finally, the objective was to assess the effects of variations in the heat evolution characteristics of South African cements on the early-age temperature development in typical mass concrete structures placed under controlled conditions. Nine clinker samples were obtained from different manufacturing facilities in the country. A cement was produced from each clinker by grinding with a con- trolled amount of the same gypsum. The chemical composition of the clinkers and the gypsum were de- termined using X-ray fluorescence (XRF) and samples of the raw clinkers were mounted, polished and etched for optical microscope study. These results were then used to characterise the crystallographic and chemical properties of the cements. Each cement was used to prepare a concrete with fixed mixture proportions and aggregate type. The concretes were tested in an adiabatic calorimeter in or- der to determine the rate and amount of heat evolved by the cements. Comparisons of the heat of the heat properties of the different cements were made on the basis of a maturity function rather than clock time. The results of the adiabatic tests were also used to indicate the implications of the different heat charac- teristics of cements for the time–temperature profiles in a large, high strength beam element and a low strength mass concrete element. A finite difference heat flow model was used in this analysis. 2. Clinker sampling and preparation 2.1. Sample collection Cement clinker samples were drawn from nine pro- duction plants belonging to three cement companies operating in South Africa. Each of these production plants manufacture Type 1 cement which satisfies the requirements of the local standard specification, SABS ENV 197 [10]. To ensure that the samples were repre- sentative of current production, the clinker was drawn from the moving stock just before entry into the grind- ing mill. Approximately 4 kg of clinker was drawn from each plant and 500 g of each sample was reserved for microscopy and chemical analysis. The remaining 3.5 kg was used to manufacture cement in the laboratory. 2.2. Gypsum for cement manufacture In order to manufacture the cements in the labora- tory, a sample of gypsum was obtained from one of the production plants and, based on a chemical analysis of the gypsum and the clinkers, the amount of gypsum added to each of the clinkers was adjusted so that each cement had an SO3 content of 2.3%. The same gypsum type and SO3 content was used to ensure that the rate of heat evolution of each clinker was compared on the basis of clinker characteristics only. Also, while an SO3 content of 2.3% is typical of South African cements, in reality, manufacturers would adjust the SO3 content in response to variations in aspects such as the C3A con- tent of the clinker. Table 1 shows the chemical compo- sition of the gypsum used, based on an XRF analysis, as well as the proportion of gypsum added to each of the cement clinkers. 2.3. Grinding and cement fineness Each clinker with the appropriate amount of gypsum was ground in a laboratory ball mill. The mill was pe- riodically stopped and the specific area of the sample was determined using the Blaine test [11]. Grinding was stopped when the measured specific surface area of the sample was 3200± 50 cm2/g. Using this procedure, the average fineness measured for the nine samples was 3198 cm2/g with a standard deviation of 24 cm2/g. 3. Test methods 3.1. X-ray fluorescence XRF was used to determine the chemical compound composition of the clinkers. The analyses were con- ducted using ground clinker samples and these results were also used to determine the required amount of gypsum for each cement. This analysis was conducted in the laboratory of one of the cement producers, where XRF is used for routine quality control purposes. The instrument is regularly calibrated and checked for op- erational accuracy. The samples used in this project were Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 419 placed in the automatic feed system of the instrument alongside the routine production sample analyses. 3.2. Optical microscopy A geological polarising microscope with reflected light capability, was used to assess the structure and morphology of polished and etched samples of each of the clinkers. The clinker samples were lightly crushed using a steel pestle and the )2.36 to +1.18 mm fraction was used to prepare the polished sections. The frag- ments were encapsulated in epoxy and one surface was polished, initially with 600 grit carborundum paper, followed by progressively finer abrasion systems, until a final polish with 0.25 lm diamond paste, on a lapping disk with a non-aqueous lubricant. The polished samples were then etched using a pro- cedure recommended by St John et al. [12]. This pro- cedure involves the following treatment of the polished surface of the sample: • 20 s immersion in a 10% solution of KOH in ethyl al- cohol to etch the C3A and C4AF phases; • rinse in ethyl alcohol; • brief rinse in isopropyl alcohol and light buff with pa- per towel; • 6 s immersion in a solution of 1.5 ml HNO3 in 100 ml isopropyl alcohol to etch mainly the alite (C3S) and the belite (C2S) phases; • rinse in isopropyl alcohol. Digital photo-micrographs were then obtained of the etched surfaces and an image analysis system was used to measure crystal sizes and to provide a qualitative description of the clinker morphology. 3.3. Adiabatic calorimetery 3.3.1. Expressions for the rate of heat evolution The adiabatic calorimeter proposed by Gibbon et al. [13] was used to determine the amount and rate of heat evolved by each of the laboratory cement samples. In essence, the calorimeter uses a 1 litre sample of concrete in a bath of water but separated from direct contact with the water by a 40 mm thick air space. A temperature probe in the concrete sample is monitored by a personal computer fitted with an analogue-to-digital input/output card. A heater element in the water bath is then turned on or off in response to the concrete temperature to maintain the water bath at the same temperature as that of the concrete. The concrete temperature is then mon- itored over time and the amount of heat per unit mass of cement (qt) is determined from: qt ¼ Cp � ðTt � T0Þ � ms mc ð1Þ where Cp is the specific heat capacity of the concrete, determined as the mass weighted average of the specific heat capacities of the concrete components and, as dis- cussed by Gibbon et al. [13], is assumed to be constant throughout the test; Tt is the temperature of the concrete sample at time t during the adiabatic test and T0 is the sample temperature at the beginning of the test; ms is the mass of the concrete test sample and mc is the mass of cementitious binder in the sample. The heat rate ( _qt) is conventionally determined by differentiating Eq. (1) with respect to time, so that: _qt ¼ dqt dt ð2Þ The time basis of Eq. (2) presents a problem because of the circular relationship between the rate of heat evolution, the extent of hydration, time and the tem- perature at which the hydration reactions take place. This is generally addressed by expressing the heat rate (as in Eq. (2)) as a function of maturity [7,14], rather than time. Hence, the heat rate applicable to a particular point in a structure at a given time is determined by the maturity (or time–temperature history) at that point. Stated differently, at a given point in time, the heat rate will vary across a concrete section in accordance with the variation in maturity across the section. In order to normalise the heat curves to an equivalent degree of advance of hydration, the time axis is ex- pressed in terms of Arrhenius maturity development [15], which, in its relative form, is written as: t20 ¼ Xi¼n i¼1 exp E R � � 1 293 �� � 1 273þ 0:5 � ðTi þ Ti�1Þ �� � ðti � ti�1Þ ð3Þ The reference temperature in Eq. (3) is taken as 20 �C and t20 is the equivalent maturity time (in h); E is the activation energy parameter (33.5 kJ/mol, taken as constant [16]); R is the universal gas constant (8.314 J/ mol �C); Ti is the temperature (�C) at the end of the ith time interval, ti. Noting that the rate of hydration and therefore, the rate of heat evolution, is dependent on the absolute temperature, Ballim and Graham [17] have shown that, since Eq. (1) accounts only for temperature differences, the rate of heat evolution determined from an adiabatic calorimeter test should be normalised with respect to maturity, rather than time. This means that, rather than the form shown in Eq. (2), the heat rate should be ex- pressed as: _qM ¼ dqt dM ð4Þ Expressing the maturity heat rate (Eq. (4)) as a function of the cumulative maturity, allows the heat rate curve to be normalised so as to be independent of the starting temperatures in the adiabatic test. 420 Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 It should be noted that normalising the heat rate curves using a maturity function as described above, effectively converts the adiabatic heat rate curve to an equivalent isothermal calorimetery curve. With this ap- proach, an error may be introduced by assuming a constant value for the activation energy parameter (E). Broda et al. [18] have shown that E varies with tem- perature during hydration but note that the variation is fairly small and that a single value would suffice. In an assessment of blended cements using isothermal calori- metry, Xiong and van Breugel [19] show similar varia- tions in the apparent activation energy with the progress of hydration. However, they also conclude that this variation ‘‘. . .may be less important in real engineering practice. . .’’. 3.3.2. Sample preparation and testing The nine laboratory cements were each used to make a concrete sample with the mixture composition as shown in Table 2. All the mixture components, includ- ing the water, were stored in the same room as the cal- Table 2 Mixture composition of the concrete used for the adiabatic calorimeter tests Laboratory cement 350 kg/m3 9.5 mm washed silica stone 850 kg/m3 Graded, washed silica sand 885 kg/m3 Water 233 l/m3 Table 3 Results of XRF analysis of the clinker samples and Bogue calculation of cr A B C D E % Composition of clinker CaO 66.50 68.16 66.50 65.57 SiO2 22.30 23.03 21.20 22.15 Fe2O3 3.61 2.28 3.03 2.98 Al2O3 3.80 3.82 4.20 4.51 MgO 1.10 1.48 3.30 2.64 TiO2 0.18 0.31 0.30 0.45 Mn2O3 0.05 0.17 0.61 0.22 K2O 0.62 0.13 0.25 0.16 Na2O 0.37 0.00 0.12 0.20 SO3 0.82 0.30 0.25 0.26 P2O5 0.15 0.00 0.02 0.00 Free lime 1.02 1.10 1.25 0.80 LOI 0.80 0.30 0.58 0.00 Total 100.20 99.98 100.30 99.14 Alumina ratio 1.05 1.67 1.39 1.51 Silica ratio 3.03 3.80 2.96 2.98 Crystalline phase composition (%) of clinker by Bogue analysis C3S 66 69 72 61 C2S 14 14 7 18 C3A 4 6 6 7 C4AF 11 7 9 9 orimeter at least 24 h before mixing. This allowed the temperature of the materials to equilibrate to the room temperature, which was controlled at 19 ± 1 �C. A 1.2 l sample of each concrete was prepared by manual mixing in a steel bowl and the adiabatic test was started within 15 min after the water was added to the mixture. All the tests were started at temperatures between 18 and 20 �C and temperature measurement in the calorimeter was continued for approximately four days. The silica sand used in the concretes was obtained in four size fractions and these were recombined as needed for the mixing operation to ensure a uniform sand grading for each concrete. 4. Results and discussion 4.1. XRF analysis Table 3 shows the results of the XRF analyses of each of the clinker samples. The table also shows the calcu- lated proportions of the four main cement mineral phases in each clinker based on the Bogue analysis [20]. It was not thought necessary to apply the modifications to the Bogue analysis as proposed by Taylor [21] since the analysis was conducted on clinker samples. In this case, the error incurred in the analysis would not be significantly improved for rapidly cooled clinkers such as those tested in this investigation. Table 3 also shows ystalline phase composition F G H K 68.40 67.60 67.60 65.06 65.6 22.10 22.20 22.80 22.26 21.90 4.26 3.33 1.90 3.05 1.57 3.80 4.40 4.30 3.98 5.00 0.50 0.80 1.40 2.04 3.70 0.19 0.21 0.28 0.25 0.26 0.09 0.06 0.24 0.99 0.72 0.39 0.57 0.58 0.53 0.32 0.19 0.18 0.06 0.14 0.11 0.20 0.48 0.27 0.21 0.48 0.16 0.13 0.03 0.00 0.10 0.82 0.61 1.57 0.97 0.35 0.64 0.12 0.98 0.90 0.16 100.80 100.80 100.40 99.41 99.92 0.89 1.32 2.25 1.31 3.16 2.77 2.90 3.70 3.19 3.36 75 70 64 61 63 6 11 17 18 15 3 6 8 5 11 13 10 6 9 5 Fig. 1. Sample photo-micrographs of the clinkers showing: (a) ragged edges of C2S––clinker B; (b) nests of C2S––clinker E; (c) Dendritic distribution of C2S––clinker K. Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 421 the calculated values of the alumina and silica ratios as defined in Ref. [12]. The following points are noted from the results pre- sented in Table 3: • The chemical compound analysis of the clinkers gen- erally lie within the limits proposed by Addis [22] for South African cements. This means that the nine clin- ker samples selected can be considered typical of South African clinkers intended for cement manufac- ture. • Clinker E shows an unusually high content of Fe2O3 which explains the high calculated C4AF content of this clinker. • Clinker K is unusual in the group because of the sig- nificantly higher MgO (periclase) content as well as the higher Al2O3 content. • Accepting that the Bogue analysis indicates likely or potential compound development, the results indicate that the higher alite content of clinkers B, C, E and F is likely to result in relatively high early-age heat re- sults because of the strong influence of this com- pound on heat liberation during early hydration [21]. • The relatively high C3A content of clinker K is also likely to add significantly to the early heat evolution of this clinker. • Considering the alkali content of the cement (K2O and Na2O), clinker B is particularly low in alkali con- tent, with these values falling below the minimum va- lue suggested by Addis [22] for South African cements (range 0.2–0.8%). 4.2. Optical microscopy Fig. 1 shows sample photo-micrographs of some of the clinkers viewed under reflected light after etching. In general, the C3S crystals appear large and angular to subangular, while the C2S crystals are smaller, rounded and often striated. The interstitial material between these crystals is mainly composed of C4AF and C3A. Table 4 presents a summary of the observations made during the microscopic assessment of the clinker. Note that, in this table, the term ‘‘dendritic’’ is used in its geological sense to mean a ‘‘tree-like’’ distribution of crystal growth, forming a cluster of small crystal nod- ules which are poorly connected to each other. 4.3. Adiabatic calorimetry Fig. 2 shows the results of the adiabatic calorimeter tests conducted with the concretes containing each of the laboratory cements. The results are presented as the total heat liberated by the cement (per unit mass), as calculated using Eq. (1). The specific heat (Cp) was taken as the mass weighted average of the concrete mixture components (as shown in Table 2) and was calculated to be 1270 J/kgK. The time axis is presented as equivalent maturity hours, determined using the Arrhenius func- tion shown in Eq. (3). There is clearly strong overlap of many of the curves and only those curves which are separately identifiable have the corresponding clinker cement indicated on Fig. 2. The adiabatic test was terminated when the sample showed no appreciable increase in temperature. This is the point at which the rate of hydration of the cement is so slow that the heat being evolved is not sufficient to increase the temperature of the sample to a level that is measurable by the temperature probes. For most of the cements tested, this point was reached before 600 ma- turity hours. However, in the case of clinker cement E, 0 100 200 300 400 500 0 200 400 600 800 1000 1200 Maturity (t20 hours) T ot al H ea t kJ /k g H EK F G A B/C/D Fig. 2. Heat curves determined from the adiabatic tests for the nine clinker cements tested. Table 4 Summary of observations made during the microscopic examination of the clinker samples Clinker C3S C2S Interstitial phase A Angular, euhedral crystal shape, generally hexagonal; crystal size from 20 to 60 lm Well distributed with occasional nests; rounded, oval or pear-shaped crystals; crystal size in the range 15–55 lm Generally thinly distributed, especially in areas where C3S and C2S are tightly packed B Angular, euhedral crystal shape, generally hexagonal; crystal size from 20 to 60 lm; most crystals have C2S inclusions Well distributed with occasional nests; rounded shapes with ragged edges; crystal size in the range 20–50 lm Well distributed matrix with approxi- mately equal aluminate and ferrite phases C Dominant crystal form; many crystals have joined but retain the angular structure; crystal size from 10 to 50 lm Sparse distribution; individual crystals rare and mainly occurs as flows between the C3S crystals; crystal size from 3 to 15 lm Well distributed matrix; aluminate clearly discernable and appears as localised patches or concentrated spots D Crystals usually angular with common pseudohexagonal shapes; crystal size from 10 to 50 lm; some large inclusions Well distributed in small nests; rounded, oval or pear-shaped crystals; crystal size in range 10–30 lm; narrow rims on C3S crystal edges Well distributed throughout; poor distinc- tion between phases; aluminates appear present as localised patches E Dominant crystal form; angular with pseudohexagonal shapes; crystal size from 10 to 55 lm; C2S inclusions common Well distributed in small nests; rounded to oval-shaped crystals; crystal size in range 10–35 lm; distinct lamellar structure Good distribution with occasional abutting C3S grains; aluminate in the form of narrow streaks––possibly dendritic F Tightly packed with much joining while retaining angular shape; small inclusions which are not common Well distributed in small nests; rounded to oval-shaped crystals; crystal size in range 20–30 lm; distinct lamellar structure Good distribution; aluminate in the form of small patches and narrow streaks–– possibly dendritic G Relatively small grains: 5–35 lm; angular with pseudohexagonal shapes; inclusions common Rounded to oval; small crystals: 5–40 lm; thin zones on edges of C3S Aluminate phase prevalent; large patches with small ferrite areas H Tightly packed grains; joined to form subhedral shapes; 10–50 lm; some inclu- sions Well distributed in small and large nests; rounded to pear shaped; 10–30 lm Aluminate phase in dendritic form K Crystals angular euhedral; often joined; 10 to 40 lm; inclusions small and rare Individual crystals rare; generally occurs in the matrix in dendritic crystals or as rims on C3S grains Well distributed and rich in aluminate; small amounts of periclase visible 0 1 2 3 4 5 0 10 20 30 40 50 Maturity (t20 hours) dq /d M (W /k g) K F A B E H G D C Fig. 3. Heat rate curves for the cements tested in the adiabatic calo- rimeter. 422 Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 this point was only reached at approximately 800 ma- turity hours and, in the case of clinker cement H, mea- surable temperature increases continued to be recorded after 1100 maturity hours. This trend was confirmed by re-testing of this cement. The reason for this behaviour are not clear at this stage and certainly require further investigation. The results shown in Fig. 2 were converted to the maturity heat rate form using Eq. (4) and Fig. 3 shows the calculated heat rates for the nine cements in this form. The units of the heat rate in this figure are un- conventional and are best described as ‘‘maturity watts’’ per kg of cement (J/maturity second/kg). Fig. 3 again shows considerable overlap and only the clearly discernable curves are identified. Expressed in this form, clinker cement K stands out in the brief but significantly larger heat rate at between 15 and 20 ma- turity hours. To improve the clarity, the curves in Fig. 3 are shown only for the first 50 maturity hours. Beyond this point, the curves bunch together and individual characteristics are not discernable. Based on the heat rate curves shown in Fig. 3, the clinker cements fall into three distinct groups in terms of the maximum heat rate as follows: • high heat––clinker cement K; • medium heat––clinker cements A, B, E, F and H; 0 1 2 3 4 5 0 25 50 75 100 Maturity (t20 hours) dq /d M (W /k g) K F D Fig. 4. Three clinker cements representative of the three proposed heat rate groups. Fig. 5. Comparison between the measured and predicted heat rate curves for clinker cements A, D and K using the Maekawa et al. model [27]. Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 423 • low heat––clinker cements C, D and G. For clarity, representative curves of each of these three groups are extracted from Fig. 3 and shown in Fig. 4. The essential difference between the three curves shown in Fig. 4 lies in rate of heat evolution between 10 and 25 maturity hours. Outside of this maturity range, the heat rates show only marginal differences. The av- erage maximum measured heat rates for each of the three groups are 2.2 J/maturity second/kg for the low heat group, 3.0 J/maturity second/kg for the medium heat group and 4.6 J/maturity second/kg for clinker ce- ment K. This indicates that the heat rate model pro- posed by Wang and Dilger [14] is only appropriate for the low heat group of clinker cements tested in this in- vestigation. Their model sets a maximum heat rate of 2.2 W/kg occurring at 10 maturity hours. 4.4. Relating heat rates to physico-chemical characteris- tics of the clinker Considering the results of the XRF analysis presented in Table 3, the qualitative description of the clinker morphology characteristics presented in Table 4 and the heat rate classifications presented above, the following points are noted: • The high heat rate performance of clinker K seems to be related to the high C3A content, which is also re- flected in a high alumina ratio. While working with significantly different cements at the time, Lerch [23] noted similar ‘‘spikes’’ in the rate of heat evolution of cements with high C3A contents. He also noted that the magnitude of the heat rate ‘‘spike’’ decreased and the time of its occurrence increased as the SO3 content of the cement increased. Macphee and La- chowski [24] note that the presence of periclase has the effect of reducing the hydration kinetics of the ce- ment. However, in the case of clinker K, it appears that the effect of the relatively high periclase content is overshadowed by the effect of the high C3A con- tent. Furthermore, given the finer distribution of the dendritic form of belite in this clinker, this phase may show earlier hydration than would normally be expected with more typical belite phases. However, the extent to which this contributes to early heat evo- lution will require further investigation. • There appears to be very few common chemical or morphological characteristics between clinkers A, B, E, F and H to indicate that they should belong in a joint heat rate category as measured in the adiabatic calorimeter tests. This is particularly so when con- sidering the clinkers that have been categorised as low-heat clinkers (C, D and G). These two groups of clinker are not distinguishable on the basis of the chemical analysis or the morphological descriptions. • Because of the relatively high alite contents, none of the clinkers assessed conform to the description of a ‘‘typical low-heat’’ cement as described by Jackson [25], who suggests a C3S content of 34% for this ce- ment classification. At this stage, we should speculate that the characteri- sation techniques used for the clinkers has not provided sufficient information to discriminate between the clinkers on the basis of the rate of heat liberation. It may be that a more detailed analysis of the polymorph [26] composition of the clinker phases is necessary to better understand the hydration kinetics of a particular phase. A further useful analysis would be to consider the phase distribution in the different size fractions of the ground cement. Cements with a higher proportion of alite in the finer fractions can be expected to show higher rates of heat evolution, even if the bulk alite content is not un- usually high. 4.5. Modelling the rate of heat evolution The composition heat rate model developed by Maekawa et al. [27] was used to assess the ability of such a model to predict the heat rates of the cements tested in this study. The model was adjusted to express the heat rate in terms of maturity time (t20 h). Fig. 5 shows a comparison of the measured and predicted heat rates for 424 Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 clinkers A, D and K which indicate the range of typical results. In some cases, the prediction was particularly good (clinker D), while in others, the model predicted the peak heat rate correctly, but the timing of the peak was time-shifted from that measured. In the case of clinker K, the model showed particularly poor predic- tion. It therefore seems that the parameters accounted for in the model are not always sufficient to describe the heat evolution characteristics of clinker cements. Fig. 6. Variation in the modelled temperature across a central section of a large concrete block at the time when the maximum temperature occurs, for the three cement heat rates assessed. 5. Effects of the measured heat rates on actual structures Our interest in the rate of heat evolution of cement derives primarily from concern for the maximum tem- peratures likely to occur in concrete structures. The implications of the different heat rates for the cements assessed in this programme was therefore investigated using a modified version of the finite difference heat model proposed by Ballim [28]. The model was run us- ing the three heat curves shown in Fig. 4 (D, F and K), representing the low, medium and high heat rate clinker cement groups. The intention here is not to try to reflect field conditions but rather to indicate the temperature effects of using the different cements in large concrete elements under controlled conditions. In reality, con- crete for large structures would normally have signifi- cant proportions of cement extenders (fly ash or blastfurnace slag) and include admixtures to reduce the rate of heat evolution. Two structures were assessed using the three heat rate curves as input into the model: • A concrete block, typical of that used for dam wall construction, measuring 4 m wide and 3 m deep; ce- ment content¼ 200 kg/m3; concrete specific heat ca- pacity¼ 1120 J/kg �C; concrete density¼ 2430 kg/m3. • A concrete beam measuring 500 mm square; cement content¼ 475 kg/m3; concrete specific heat capac- ity¼ 1200 J/kg �C; concrete density¼ 2385 kg/m3. • The thermal conductivity of the concrete is 2.7 W/ m �C as suggested by van Breugel [7] for granite ag- gregate concrete. Furthermore, the external and processing conditions used in the modelling exercise for both structures, were as follows: • the concrete is cast directly onto rock with a thermal conductivity of 1.2 W/m �C; • concrete is placed at a temperature of 17 �C and the ambient temperature varies sinusoidally between 16 and 20 �C; • the side formwork is held in place for 18 h after cast- ing, during which time the heat transfer coefficient from these surfaces is taken as 5 W/m2 �C; • upon removal of the formwork, the heat transfer co- efficient for the side surfaces is taken as 25 W/m2 �C, representing a surface directly exposed to a mild wind. Fig. 6 shows the modelled temperatures for large concrete block from the vertical surface to the centre of the block, across a section 1.5 m below the top surface and at the time when the maximum internal temperature occurs. The time to reach the maximum temperature varied for the different clinker cements and this is indi- cated in Fig. 6 in parentheses alongside the clinker ce- ment identification. In this application, Fig. 6 shows that the higher the maximum heat rate of the binder as measured in the adiabatic calorimeter, the higher is the maximum tem- perature obtained on the structure. For the three clinker cements assessed, the maximum temperature varied from 40 �C for clinker cement K, 38 �C for clinker ce- ment F and 36.5 �C for clinker cement D. Also, while the maximum temperature for clinker cements F and K occurred at approximately 93 h after casting, the maxi- mum temperature was only reached at 108 h for clinker cement D. In addition to the lower temperature gradient that occurs with clinker cement D, the later occurrence of the maximum temperature may further mitigate the potential for cracking, assuming that the creep capacity remains relatively unchanged during this period. Fig. 7 shows the modelled temperatures across the high strength beam from a vertical surface to the centre of the beam, across a section 250 mm below the top surface. The temperatures shown in Fig. 7 occur at 20 h after casting for each of the clinker cements assessed, which is when the maximum internal temperature oc- curs. In comparison with the results shown in Fig. 5 for the large concrete block, Fig. 7 shows that: • there is very little difference between the effects of clinker cements F and K in the maximum tempera- ture profile obtained in the beam––possibly because Fig. 7. Modelled temperatures across a central section of a concrete beam at the time when the maximum temperature occurs, for the three cement heat rates assessed. Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 425 the ‘‘spike’’ of high heat rate shown by clinker cement K is of too short a duration to have a significant effect in this application where heat dissipation is more ra- pid; • on average, the three cements show temperature dif- ferences between the surface and the centre of the beam of approximately 8 �C, which is significantly smaller than the corresponding value of approxi- mately 18 �C obtained for the concrete block; • using clinker cement D to construct the beam (Fig. 7) results in a maximum temperature which is 6 �C lower than that obtained when using clinker cements F or K. 6. Conclusions (1) Considering the maturity rate of heat evolution, the nine clinker cements tested can be categorised into high, medium and low heat cements based on the maximum measured heat rate. The average maximum heat rates obtained in each of these categories were: 2.2 J/maturity second/kg for the low heat group, 3.0 J/ma- turity second/kg for the medium heat group and 4.6 J/ maturity second/kg for the high heat cement. (2) Within this categorisation, only clinker cement K was placed in the high heat category, clinker cements A, B, E, F, and H were categorised as medium heat cements and clinker cements C, D and G were considered as low heat cements. (3) XRF analysis for chemical composition and qualitative microscopy for morphological characterisa- tion of cement clinker, in themselves, do not provide sufficient information to allow a heat-rate characterisa- tion of the cement derived from such a clinker. It is suggested that this information should be combined with further analysis of the clinker involving a thorough characterisation of the crystallographic forms and an assessment of the phase distribution in different size fractions of the ground cement. (4) Based on a finite difference heat model analysis, it appears that the categorisation of low, medium and high heat cements is more significant for the temperature development in low cement content, mass concrete structures than for high-strength, relatively small con- crete elements. Importantly, the maturity duration of the measured peak heat rate of the cement is an im- portant in determining the effect of maximum heat rate on temperature development in smaller concrete ele- ments. For the present study, because of the relatively short duration of the peak heat rate of clinker cement K, the effect of the medium and high heat rate cements were almost identical when considering the temperature de- velopment in a 0.5 m thick concrete beam-type element. References [1] Hughes BP, Mahmood AT. Laboratory investigation of early age thermal cracking of concrete. ACI Mater J 1988;(May/June). [2] Campbell DH. A summary of Ono�s method for cement quality control with emphasis on belite color. In: de Hayes SM, Stark D, editors. Petrography of cementitious materials. ASTM STP 1205. Philadelphia: American Society for Testing Materials; 1994. [3] Bogue RH. The chemistry of Portland cement. 2nd ed. New York: Reinhold Publishing Corporation; 1955. [4] Bye GC. Portland cement––composition production and proper- ties. Oxford: Pergamon Press; 1983. [5] Copeland LE, Kantro DL, Verbeck G. Chemistry of hydration of Portland cement. In: 4th International Symposium on the Chemistry of Cement. US Department of Commerce, Washing- ton, 1962. [6] Steinhour HH. The reactions and thermochemistry of cement hydration at ordinary temperature. In: 3rd International Sympo- sium on the Chemistry of Cement. Cement and Concrete Association, UK, 1952. [7] vanBreugel K. Prediction of temperature development in harden- ing concrete. In: Springenschmid R, editor. Prevention of thermal cracking in concrete at early ages. London: E&FN Spon; 1998. p. 51–75. RILEM Report 15 [Chapter 4]. [8] Kishi T, Shimomura T, Maekawa K. Thermal crack control design of high performance concrete. In: Proceedings of the International Conference ‘‘Concrete 2000’’Dundee, Scotland. London: E&FN Spon; 1993. [9] Garboczi EJ, Bentz DP. Computer based models of the micro- structure and properties of cement-based materials. In: Proceed- ings of the 9th International Conference on the Chemistry of Cement, vol. VI. New Delhi, 1992. p. 3–15. [10] SABS ENV 197-1:2000. Cement––Part 1: Composition, specifica- tions and conformity criteria for common cements. South African Bureau of Standards, Pretoria, 2000. [11] SABS SM 748:1971. Standard test method: specific surface of cement. South African Bureau of Standards, Pretoria, 1971. [12] St John DA, Poole AW, Sims I. Concrete petrography: a handbook of investigative techniques. London: Arnold; 1998. [13] Gibbon GJ, Ballim Y, Grieve GRH. A low-cost, computer- controlled adiabatic calorimeter for determining the heat of hydration of concrete. ASTM J Test Evaluat 1997;25(2):261–6. [14] Wang Ch, Dilger WH. Prediction of temperature distribution in hardening concrete. In: Springenschmid R, editor. Thermal cracking in concrete at early ages. London: E&FN Spon; 1994. p. 21–8. [15] Naik TR. Maturity functions for concrete during winter condi- tions. In: Naik TR, editor. Temperature Effects on Concrete. 426 Y. Ballim, P.C. Graham / Cement & Concrete Composites 26 (2004) 417–426 ASTM STP 858. Philadelphia: American Society for Testing and Materials; 1985. [16] Bamford CH, Tipper CFH, editors. Comprehensive chemical kinetics Vol 1: The practice of kinetics. London: Elsevier Publishing Company; 1969. [17] Ballim Y, Graham PC. A maturity approach to the rate of heat evolution in concrete. Magazine of Concrete Research 2003;55(3): 249–256. [18] Broda M, Wirquin E, Duthoit B. Conception of an isothermal calorimeter for concrete––determination of the apparent activa- tion energy. Mater Struct 2002;35:389–94. [19] Xiong X, van Breugel K. Isothermal calorimetry study of blended cements and its application in numerical simulations. Heron 2001;46(3):151–9. [20] Lawrence CD. The constitution and specification of Portland cements. In: Hewlett PC, editor. Lea�s chemistry of cement and concrete. 4th ed. London: Arnold; 1988 [Chapter 4]. [21] Taylor HFW. Cement chemistry. 2nd ed. London: Thomas Telford; 1997. [22] Addis BJ. Cementitious materials. In: Addis BJ, editor. Fulton�s concrete technology. Midrand, South Africa: Cement and Con- crete Institute; 2001 [Chapter 1]. [23] Lerch W. The influence of gypsum on the hydration and properties of Portland cement pastes. Proc ASTM––technical papers 1946;46:1252–97. [24] Macphee DE, Lachowski EE. Cement components and their phase relations. In: Hewlett PC, editor. Lea�s chemis- try of cement and concrete. 4th ed. Arnold: London; 1988 [Chapter 3]. [25] Jackson PJ. Portland cement: classification and manufacture. In: Hewlett PC, editor. Lea�s chemistry of cement and concrete. 4th ed. Arnold: London; 1988 [Chapter 2]. [26] Roode-Gutzmer QI, Ballim Y. Phase composition and quantita- tive X-ray powder diffraction analysis of Portland cement and clinker. In: Mindess S, Skalny J, editors. Materials science of concrete VI. American Ceramic Society; 2001 [Chapter 1]. [27] Maekawa K, Chaube R, Kishi T. Modelling concrete perfor- mance––hydration, microstructure formation and mass transport. London: E&FN Spon; 1999. [28] Ballim Y. A low-cost heat model and calorimeter to predict temperature profiles in mass concrete elements. In: Proceedings of the 1st International Conference on Concrete and Develop- ment. Building and Housing Research Centre. Tehran, Iran, May 2001. Early-age heat evolution of clinker cements in relation to microstructure and composition: implications for temperature development in large concrete elements Introduction Clinker sampling and preparation Sample collection Gypsum for cement manufacture Grinding and cement fineness Test methods X-ray fluorescence Optical microscopy Adiabatic calorimetery Expressions for the rate of heat evolution Sample preparation and testing Results and discussion XRF analysis Optical microscopy Adiabatic calorimetry Relating heat rates to physico-chemical characteristics of the clinker Modelling the rate of heat evolution Effects of the measured heat rates on actual structures Conclusions References