- Email: [email protected]
Shrinkage of self-compacting concrete made with blast furnace slag as fine aggregate
Construction and Building Materials 76 (2015) 1–9
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Shrinkage of self-compacting concrete made with blast furnace slag as fine aggregate M. Valcuende a,⇑, F. Benito b, C. Parra b, I. Miñano b a b
Department of Architectural Constructions, Universitat Politècnica de València, Camino de vera s/n, 46022 Valencia, Spain Department of Structures and Construction, Universidad Politécnica de Cartagena, Doctor Fleming s/n, 30202 Cartagena, Spain
h i g h l i g h t s Replacing sand by slag gives rise to mixtures with higher total pore volume. At early ages SCCs with slag show similar strength, but in the long term it increases. SCCs with slag show a higher autogenous shrinkage. Due to their higher porosity, SCCs with slag are less stiff and lose water faster. SCCs containing slag show higher total shrinkage.
a r t i c l e
i n f o
Article history: Received 18 March 2014 Received in revised form 2 October 2014 Accepted 12 November 2014
Keywords: Self-compacting concrete Shrinkage Compressive strength Granulated blast furnace slag Pore size distribution
a b s t r a c t The aim of this experimental work was to study shrinkage evolution with age in self-compacting concretes (SCC) in which part of the fine aggregate was replaced by granulated blast furnace slag (GBFS) as sand. Seven types of SCC were made with a w/c ratio of 0.55 and different slag contents. The results show that replacing sand by GBFS gives rise to mixes with higher pore volume but with slightly finer porous structure (smaller median pore and threshold diameters). At early ages slag SCCs have similar compressive strength to that of the reference concrete, although in the long term their strength increases as a result of slag reactivity. We also observed that the higher the slag content, the higher were both autogenous and drying shrinkage and consequently also total shrinkage. In comparison with the reference concrete, the increase in total shrinkage was found to be of the order of 4% and 44% when 10% and 60%, respectively, of the sand was replaced by slag. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Melted slag from a blast furnace can be subjected to various cooling techniques in order to obtain products with a wide range of end-uses. Granulated blast furnace slag (GBFS) is obtained by the rapid cooling of liquid slag. At least two-thirds of its mass consists of vitreous slag [1], which has hydraulic properties when activated and leads to the formation of C–S–H. It has a similar chemical composition to that of crystallized slag, although as it has been subjected to rapid cooling its ions are not given time to organize themselves into a crystalline network and solidify into a vitreous state. As is well known, the hydraulically active component resides in the vitreous phase and the crystalline phase can be considered as almost inert. It is generally thought that the greater its disordered structure the greater is its reactivity [2], although some authors ⇑ Corresponding author. Tel.: +34 963877450; fax: +34 963877459. E-mail address: [email protected] (M. Valcuende). http://dx.doi.org/10.1016/j.conbuildmat.2014.11.029 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
consider that the presence of a small proportion of the crystalline phase may be conducive to increased slag reactivity. The hydraulic capacity of slag is highly attenuated and is slow to appear. There are three methods of accelerating its hydration reactions: (i) by the use of chemical activators, (ii) increasing its specific surface and (iii) by raising its temperature. Portland cement is a good slag activation catalyst as it provides the three main compounds that activate slag: lime, calcium sulphate and alkalis [3]. Granulated slag is formed by small alveolar particles with sharp edges [1]. A number of studies have been published on the use of blast furnace slag as aggregate in concrete and some standards, such as ASTM C33, ASTM C989, EN-206 or EN-12620 also provide specifications for using this slag as aggregate. In many cases a problem arises when the authors do not specify whether they have worked with granulated or crystallized slag, which makes it difficult to reach general conclusions. Furthermore, in the studies which do indicate that granulated slag has been used, the results do not always coincide. According to Mosavinezhad and Nabavi [4] adding
2
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
GBFS to replace sand yielded increased compressive strength but gave lower flexural strength. Shi et al. [5] and Saito et al. [6] point out that at early ages the compressive strength of concrete using GBFS as fine aggregate is lower than concrete made with river sand, but at 91 days the strength is higher. Scandiuzzi and Battagin [7] also obtained better long-term strength due to the latent hydraulicity of the slag. On the contrary, Topçu and Bilir [8] point out that the compressive strength, flexural strength and modulus of elasticity are reduced due to the higher porosity of the mixture and that shrinkage cracking is also reduced. Yüksel et al. [9] and Yüksel and Genç [10] also maintain that replacing sand by GBFS may give rise to concretes of higher porosity and lower strength, according to the replacement ratio employed. However, when no more than 20% of the sand is replaced by slag, the concrete maintains a similar porous structure and GBFS has a positive effect on the concrete’s durability properties [11,12]. Replacing part of the aggregates by slag with hydraulic capacity could in theory give rise to concretes susceptible to higher shrinkage, partly due to the volume reduction of the cementitious materials that occurs during their hydration (chemical shrinkage). The lower the w/c ratio, the higher is this shrinkage, especially when the w/b ratio falls below 0.4 [13,14]. However, as can be seen from certain studies [6,8], using GBFS does not necessarily lead to higher shrinkage. GBFS reactions form additional C–S–H phases, which means that the pore network becomes more compact. Indeed, Arellano et al. [15] found that a dense and strong slag–paste interfacial zone was noted as a result of the blast furnace slag participation in the reactions. Increased matrix strength or density reduces material deformability [16] and thus restricts shrinkage. However, very few studies on the use of GBFS as fine aggregate have included the concrete shrinkage involved. Our objective in this experimental work was therefore to study shrinkage in self-compacting concretes (SCC) in which part of the limestone aggregate had been replaced by GBFS as fine aggregate and to compare the results obtained with those from concretes with no slag content. Due to the fact that shrinkage is directly related to the porous structure of the material, the microstructure of the different types of concrete was analysed in order to explain the behaviour observed. Using GBFS as aggregate firstly has the advantage of recycling waste and secondly the amount of natural aggregate used in the mix is reduced, with consequent savings in natural resources and the energy required for their extraction, together with a reduction of the amount of pollutants released into the atmosphere. 2. Experimental program
Fig. 1. Grading curves of slag and aggregates.
Fig. 2. Grading curves of cement, aggregate fines and limestone filler.
All the studied mixes were made with 375 kg/m3 of CEM II/B-M (S-L) 42.5R cement, and 25 kg/m3 of fly ash. The granular skeleton was composed of three fractions of crushed limestone aggregate: 4/12 mm coarse aggregate, 0/4 mm coarse sand and 0/2 mm fine sand, with a fines content (particle size <0.063 mm) of 1.0%, 12.0% and 17.1%, respectively (Fig. 1). The process of replacing sand by slag was carried out by firstly removing the fine sand only, and when this was exhausted (in concretes SCC-20–SCC-60), by also removing part of the coarse sand. Limestone filler was added to these mixes to make up for the lack of fines in the slag. The grading curves of the fines of aggregates and the filler were determined by means of laser diffraction (Fig. 2). As can be seen in this figure, the aggregate fines are finer than the limestone filler used. The basic GBFS used had high SiO2 and CaO contents, with a CaO/SiO2 ratio of 1.31. As regards its mechanical properties, the slag had a Micro-Deval index much lower than that of the limestone aggregates, and therefore had greater resistance to wear. The properties of the sand and slag are shown in Table 2 and the composition of the slag is shown in Table 3. The admixture used was a polycarboxylate-based superplasticizer (Viscocrete 3425). The concrete test specimens were made by filling moulds with a single uncompacted lift of SCC.
2.1. Concrete mixtures, materials and mixing procedure 2.2. Test program and methodology Seven types of concrete were made: a reference (SCC-0) and six with different percentages of fine aggregate replaced by slag (in weight): 10% (SCC-10), 20% (SCC20), 30% (SCC-30), 40% (SCC-40), 50% (SCC-50) and 60% (SCC-60). In order to obtain strengths approaching those normally used for building, all of them were manufactured with a 0.55 w/c ratio. The characteristics of each mix are shown in Table 1. All SCC mixes showed a slump flow of about 700 mm and good resistance to segregation in the segregation test.
2.2.1. Shrinkage test Prismatic specimens measuring 100 100 400 mm were used to determine autogenous shrinkage and total shrinkage. Temperature and weight loss were also measured in order to obtain further information on this phenomenon. Thermocouples were inserted into the centre of the specimens immediately after casting to determine the evolution of temperature rise due to the hydration reaction.
Table 1 Mixture proportions of concretes. Mix
Sand replacement ratio (%)
Cement (kg/m3)
Fly ash (kg/m3)
Water (l/m3)
WEA (kg/m3)
Coarse aggregate (kg/m3)
Coarse sand (kg/m3)
Fine sand (kg/m3)
Limestone filler (kg/m3)
Slag sand (kg/m3)
Paste volume (dm3)
SCC-0 SCC-10 SCC-20 SCC-30 SCC-40 SCC-50 SCC-60
0 10 20 30 40 50 60
375 375 375 375 375 375 375
25 25 25 25 25 25 25
220 220 220 220 220 220 220
4.4 4.2 4.4 4.6 4.8 5.2 6.0
656.6 656.8 656.6 656.5 656.3 656.0 655.4
829.7 829.6 818.4 691.4 564.4 437.5 310.5
260.3 124.8 0.0 0.0 0.0 0.0 0.0
0.0 26.9 53.6 71.4 89.1 106.8 124.4
0.0 102.7 205.4 308.0 410.6 512.9 614.7
410.9 412.1 413.9 415.1 416.2 417.6 419.3
(WEA: water enhancement admixture).
3
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9 Table 2 Properties of the aggregates. Aggregate
Micro-Deval index
Specific gravity (g/dm3)
Water absorption (%)
Coarse sand (0/4) Fine sand (0/2) Slag
43 43 14
2.60 2.60 2.45
1.0 1.0 8.0
Table 3 Chemical composition of the slag.
Slag
SiO2 (%)
SO3 (%)
CaO (%)
MgO (%)
(CaO + MgO)/ SiO2
CaO + MgO + SiO2 (%)
30.2
1.9
39.5
7.5
1.5
77.2
Fig. 4. Shrinkage test (age more than 24 h).
Table 4 Pore characteristics of concrete mixtures.
Total intrusion volume (10 3) (mm3/g) Total pore area (m2/g) Median pore diameter (lm) Threshold diameter (lm) Total porosity (%)
SCC-0
SCC-10
SCC-20
SCC-40
SCC-60
52.7 (21.50) 10.3 (30.97) 0.05 (11.86) 0.15 (14.23) 11.91 (17.62)
61.2 (8.09) 13.3 (3.64) 0.05 (10.22) 0.12 (7.96) 13.51 (9.09)
62.3 (3.59) 14.3 (4.02) 0.04 (13.64) 0.12 (4.23) 13.88 (2.87)
63.4 (8.59) 15.0 (4.57) 0.04 (7.41) 0.11 (5.01) 13.91 (7.22)
66.8 (12.28) 17.2 (22.16) 0.03 (14.27) 0.09 (16.53) 14.57 (10.27)
Numbers in parentheses are the coefficients of variation (%).
Fig. 3. Measurement of shrinkage before mould removal (age less than 24 h).
In order to measure shrinkage of the specimens until mould removal (24 h), the specimens were made in accordance with the recommendations of the Technical Committee on Autogenous Shrinkage of Concrete of the Japan Concrete Institute [17]. A polystyrene sheet (thickness 3 mm) was placed on the bottom and on both sides of the mould so that free movement of the specimen was not restricted by the mould. A polyester film was also placed over the polystyrene sheet on all sides of the mould and on the surface of the specimen. As the specimens were produced, containers measuring 25 cm diameter and 20 cm high were filled with the same concrete to determine setting time. According to Standard ASTM C 403-08, setting time is established on the basis of concrete hardening as measured by penetration resistance needles. At the time of initial setting, the end plates of the mould were removed and two dial gauges were put into place (Fig. 3). Later, at the age of 24 h, the specimens to be used for measuring autogenous shrinkage, total shrinkage and weight loss were removed from the moulds and were stored in a vertical position inside a climatically controlled chamber for one year at 20 °C and 50% RH (Fig. 4). To determine autogenous shrinkage, half of the specimens were sealed with various layers of plastic film to avoid moisture loss. Those used to measure total shrinkage were left unsealed. Shrinkage was measured by a dial gauge of 1 lm resolution placed on the upper surface of the specimen (Fig. 4). The specimens were weighed on a balance accurate to within 0.1 g. Three batches were made from each mix. One specimen for each type of test was made from each batch, the result of each test being the arithmetic mean of the three values obtained. Two 150 mm diameter 300 mm high cylindrical specimens were also made from each batch for compressive strength tests at 7, 28, 90 and 365 days (EN 12390-3:2003). Two extra specimens with the same dimensions were produced to determine the elasticity modulus (EN 1352:1997) at 28 days. 2.2.2. Mercury intrusion porosimetry (MIP) test Pore size distribution was determined using a Micromeritics AutoPore IV-9500 mercury porosimeter with a maximum pressure of 60,000 psia (414 MPa). At this pressure the smallest pore size into which mercury can be introduced is 3 nm. This test was carried out on small drilled cores, weighing approximately 6 g. The cored samples were obtained from 100 100 100 mm cubic specimens. The samples were first dried in an oven at 110 °C and were then immersed in mercury under gradually increasing pressure. As pressure rises, mercury is forced into the pores of the sample. By modelling the pores as cylindrical channels, the test pressure can be connected to the radius of these cylinders by the Washburn–Laplace law. Using this technique, a measure of the total sample porosity may also be obtained,
as well as the pore network surface area. However the MIP technique has certain limitations. In cement paste, high pressure is a source of error due to possible damage occurring during intrusion [18–20]. The microstructure of hardened cement paste is composed of pores that are separated by thin walls of hydration product. High pressures are required to measure smaller pores, so mercury enters into pores by breaking through the pore structure [18]. Olson et al. [20] found significant damage caused by relatively low pressures and much higher pore connectivity after intrusion. When the barrier between the pores is destroyed, a volume change is measured where a pore does not exist, and the data actually represent the pressure required to break through the barrier. Despite these limitations, the MIP technique is still an effective aid for comparing the pore structure and pore network characteristics of different types of cement-based materials. A sample was taken from each of the three batches of each type of concrete. The results shown in Table 4 are the arithmetic mean of the three values obtained.
2.2.3. Compressive strength tests of mortars containing slag In order to study the slag’s hydraulic activity and its evolution with time, four different mortar types were made with Portland cement without additions (CEM I), CEN standard silica sand (European Standard EN 196-1) and the same quantity of water: a reference mortar without slag (M-1) and with a mix proportion 1:3:0.5 (cement:silica sand:water), one with half the cement replaced by limestone filler (M-3), another with half the cement replaced by milled slag with a particle size similar to that of the filler (M-4), and lastly another with half the cement only, without adding either limestone filler or slag (M-2). The specimens were tested to compressive strength at 2, 7, 28, 90 and 365 days. Three batches were made from each mix. Two specimens were made from each batch, the result of each test being the arithmetic mean of the six values obtained.
3. Results and discussion 3.1. Reactivity of finely milled slag The slags used in this work present a CaO/SiO2 basicity index equal to 1.31 (Table 3) and are therefore potentially hydraulic [2]. This slag reactivity can also be seen in the X-ray diffraction test (Fig. 5), since the diffractogram obtained is characteristic of amorphous slag, with almost no crystalline mineral peaks but rather a hump located at a 2h angle around 30°. This angle corresponds to the principal peak of melilite, which is the main mineral of the
4
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 5. X-ray diffractogram of slag.
crystallized slag. The small peaks detected are of calcite (CaCO3) and mullite (3Al2O32SiO2). To test the slag’s hydraulic activity, various mortar types were made with different slag contents and were tested under compression at various ages. As can be seen in Fig. 6, the mortar made with 50% less cement (Mix M-2), as could be expected, shows much less strength than reference mortar (M-1), of the order of 77%. However, when the eliminated cement is replaced by limestone fines (M-3), the loss of compressive strength is not so pronounced, since the fines occupy the available space and give a denser cementitious matrix (filler effect). Some authors also point out that small amounts of limestone additions increase the volume of solids in hydrated cements [21]. On the other hand, it can be seen that less strength is lost at early ages, as limestone fines accelerate cement hydration, due to the fact that they provide additional surfaces for the nucleation and growth of hydration products [22,23]. At more advanced ages this effect is no longer relevant; for example at 2 days the loss of strength is more than 52% higher than that of the reference mortar, although at 28, 90 and 365 days it remains almost constant, around 62%. It can also be seen in Fig. 6 that the compressive strength of mortar containing slag (M-4) is considerably higher than the mortar with limestone fines. This difference increases with the age of the specimens. As ground slag is more reactive, finely ground slag acts as a cementitious material. In addition, powdered slag also fills up the gaps and creates nucleation centres in the same way as limestone fines. Slag hydration is quite slow, as at early ages most of the slag remains inert. The compressive strength of the mortar at 2 days old is around 50% of that of the reference mortar. As time passes the hydration reactions increase and hydrated calcium silicates are formed, so that after one year the strength reaches that of the reference mortar. However, it should not be forgotten that finely ground slag particles are more reactive than sand-sized slag particles.
Fig. 6. Evolution of compressive strength with time.
3.2. Concrete porous microstructure Fig. 7 shows the increase in mercury intrusion volume according to the equivalent pore diameter. In all the mixes, pore volume distribution follows a similar pattern, each curve having two peaks: one around pore sizes of 90 lm (peak no. 2) and another, much more pronounced, around sizes 0.05 lm (peak no. 1). The porous structure of all the concretes analysed is quite similar. The small differences between the different mixes are only apparent in the volume of the smallest capillary pores (0.01–0.1 lm).
Fig. 7. Pore size distribution.
The volume of pores larger than 0.1 lm in diameter is very similar in all the concretes, although the number of the smallest pores increases when the percentage of sand replaced by slag is higher (Fig. 8). This gives rise to a higher total pore volume in the concretes made with higher quantities of slag (Table 4). Since the
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 8. Cumulative mercury intrusion volume.
volume of cement paste (water + cement) is the same in all the types of concrete, this higher porosity is due to the slag particles being more porous than the grains of sand (Table 2) and probably to the higher fines content of the concretes made with slag (in which part of the sand is replaced by limestone filler). The higher fines content also creates a finer porous structure, since the fines fill up the small gaps (filler effect) and provide additional surfaces for the nucleation of hydration products (since there are more nucleation sites, the size of portlandite crystals is smaller and therefore the ITZ is not so thick [24]). Moreover, a small part of the CaCO3 present in the limestone filler promotes the reconversion of monosulfoaluminate into ettringite, leading to an increase of the total volume of the hydrate phase [25–28] (ettringite has a low density and therefore a relatively large volume per formula unit). On the other hand, thanks to reaction on the surface of the slag grains (the smaller the particles, the greater the reaction), the volume of hydration products increases with the passage of time (gradual formation of CSH) and the aggregate–paste interface zone becomes denser, which is the zone where the porosity is higher and the capillary pores are larger than those in cement paste [29,30]. In fact, the SEM images taken at different ages show the presence of some compounds in the slag–paste interfacial transition zone that are not present in the rest of the paste, as for example Mg (Figs. 9 and 10). The Mg content (coming from the slag [31]) is rapidly reduced as distance from the aggregates increases and tends to practically disappear at a distance of a few microns.
5
In samples analysed at 28 days (Fig. 9) it can be seen that the width of the zone where the Mg diffusion has been produced is about 2.5 lm. At 120 days, due to the slow slag reactivity, the thickness of this zone increases to around 5 lm (Fig. 10). It can also be seen that in this zone the concentration of Mg atoms is similar to the concentration in the slag itself. It can therefore be said that SCCs with slag as aggregate are slightly more porous but have a somewhat finer porous structure. This aspect is important since, according to Kumar and Bhattacharjee [32], the smallest pores have no influence on the concrete’s strength properties but do have a direct relationship with shrinkage. It can also be seen from Fig. 7 and Table 4 that the maximum pore diameter at which the continuous mercury intrusion rapidly increases varies slightly between mixes. This value, known as the threshold diameter, signals the pore diameter after which the highest number of pores are concentrated and is therefore a good indicator of the fineness of the porous structure. In this work, the threshold diameter is the smallest diameter with a differential intrusion volume >0.001 cm3/g [33]. Here again it is seen that as slag content increases the threshold diameter tends to diminish. Thus, for example, in SCC-0, SCC-20, SCC-40 and SCC-60 the threshold pore width is respectively equal to 0.15, 0.12, 0.10 and 0.09 lm. The pore size of the maximum pore concentration and median pore diameter are two other important parameters related to the fineness of pore structure. Fig. 7 shows that in SCC-0 the maximum concentration of pores (peak no. 1) tends to be around larger pore sizes than in SCCs with slag. For example, in SCC-0 the maximum concentration of pores is for a pore size of 0.065 lm and in SCC10, SCC-20, SCC-40 and SCC-60 for a pore size of 0.05, 0.05, 0.05 and 0.04 lm, respectively. Median pore diameter is also smaller in the SCCs containing slag, the size being in inverse proportion to the slag content (Table 4). 3.3. Compressive strength The compressive strength results are shown in Table 5. Although the results are very similar in all concretes, with no statistically significant differences between them, at early ages (7 days) those with the highest slag content tend to have lower compressive strength. This could be due to the higher porosity of these concretes, as observed in the MIP tests, and the greater volume of paste in their mixes (Table 1). At more advanced ages this behaviour changes. Since reactive slags were used in this work, as hydration proceeds CSH is formed around the particles (Figs. 9 and 10) and the aggregate–paste interface becomes denser, improving the aggregate–paste bond. In fact,
Fig. 9. SEM of slag–paste interface and element distribution of interfacial zone by linear scanning method at the age of 28 days (1300).
6
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 10. SEM of slag–paste interface and element distribution of interfacial zone by linear scanning method at the age of 120 days (1300).
Table 5 Mechanical properties of the concretes (MPa). Mix
SCC-0 SCC-10 SCC-20 SCC-30 SCC-40 SCC-50 SCC-60
Compressive strength
Elastic modulus
7 days
28 days
90 days
365 days
28 days
33.0 (1.26) 31.0 (2.93) 34.0 (0.34) 32.8 (2.73) 31.2 (1.37) 31.3 (0.28) 29.9 (0.71)
39.2 (3.14) 38.9 (2.13) 38.7 (1.49) 39.2 (3.53) 38.9 (3.09) 38.2 (1.39) 38.7 (2.05)
46.7 (0.39) 46.5 (0.49) 46.7 (1.46) 47.5 (4.88) 45.5 (1.60) 48.7 (2.02) 47.6 (0.58)
49.6 (0.96) 49.2 (0.25) 49.7 (1.88) 51.1 (2.92) 50.9 (1.22) 51.8 (1.67) 52.3 (0.59)
30,539 (1.53) 29,017 (0.26) 29,404 (0.15) 28,750 (5.69) 28,445 (2.30) 28,804 (1.31) 28,652 (2.41)
Numbers in parentheses are the coefficients of variation (%).
according to the MIP tests, the mixtures with the highest percentages of sand replaced by slag show a finer porous structure that is probably due to greater CSH formation. However, as has been seen in Section 3.2, such hydration reactions are quite slow, so that the improvement in mechanical behaviour is also slow. Thus, for example, at 28 and 90 days the compressive strength of all concretes is practically the same, with no statistically significant differences between them (P-value = 0.73 at 28 days and P-value = 0.19 at 90 days, for a 95% confidence level). However, at 365 days, the higher the aggregate slag content the higher the compressive strength of the concrete tends to be (P-value = 0.04). 3.4. Autogenous shrinkage Shrinkage strain and thermal strain from the heat generated during cement hydration are created simultaneously. The temperature of the concrete was monitored by thermocouples from casting up to the age of 28 h (Fig. 11). During this time period the temperature varied only slightly (less than 1.0 °C in all cases) due partly to the chemical composition of the cement (low C3A content) and partly to the small size of the concrete specimens. In Fig. 11 it can be seen that the rise in temperature is similar in all concretes and happens in the first 10–12 h, after which it drops to ambient temperature before reaching 24 h of age. In order to eliminate the temperature effect, the shrinkage values were
Fig. 11. Temperature evolution in the first 28 h.
corrected considering De = aDT, where De is the thermal strain, a is the coefficient of thermal expansion (1/°C) and DT is the temperature change (°C). The strain/time curve corrected for the temperature effect corresponds to the volumetric change of concrete under isothermal conditions. As in the case of many other study groups, a = 10 5/°C was selected. After 24 h the temperature variation in the studied mixes was small, so this effect was neglected. The evolution of autogenous shrinkage with time is shown in Fig. 12. In general, the autogenous shrinkage values obtained by other research groups are lower than those obtained in the present work. However, in agreement with the results obtained in other studies [34,35], this shrinkage forms a considerable percentage of total shrinkage. For example, at one month old this percentage varies between 36% and 44% in the seven concrete types analysed (Fig. 13). As the specimens lose water and reach hygrometric equilibrium with the atmosphere, the drying shrinkage decreases and stabilizes, although as the cement and the slag continue to hydrate, autogenous shrinkage due to self-desiccation inside the specimen still goes on. With the passage of time the autogenous shrinkage/ total shrinkage ratio therefore increases until it reaches values between 68% and 78% (at the age of twelve months), depending on the type of concrete, with similar behaviour in all types. The SCCs containing slag show higher autogenous shrinkage than the reference concrete, with higher shrinkage when the quantity of sand replaced by slag is increased. Thus, for example, at
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 12. Autogenous shrinkage.
7
Fig. 14. Drying shrinkage.
Fig. 13. Ratio between autogenous shrinkage and total shrinkage. Fig. 15. Weight loss.
12 months old SCC-0 has 11%, 14%, 21%, 26%, 28% and 33% less shrinkage than SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC60, respectively. This can be explained by several reasons; firstly, the concretes made with slag have higher volume of paste (Table 1), so that they are more likely to experience greater autogenous shrinkage [36]. However, some authors [37,38] point out that increasing only the content of limestone powder but keeping cement and water constant has only a minor effect on autogenous shrinkage. Secondly, as we have seen in the MIP tests, the higher the quantity of slag used the higher the total pore volume and consequently the concrete is more prone to deformation. In addition, the slags used are much more porous than limestone aggregates (Table 2). They are therefore probably less stiff and less likely to hinder the deformation of the concrete. In fact, in the elasticity modulus tests, the concretes with slag were seen to lose stiffness (Table 5). Thirdly, as the porous structure is also finer in concretes with higher slag content, they can be expected to show higher shrinkage, since the smaller the capillaries the higher is the tensile stress generated by the water menisci in the capillaries [35,39] (due to the surface tension of the water, as the capillary pores lose water, the force of attraction between pore walls increases and promotes shrinkage). The fourth factor to be borne in mind is the reactivity of the slag as seen in the different tests carried out (Xray diffraction, SEM and compressive strength of mortars). As a result of hydration of the surface of the slag particles, the amount of water in the pores is reduced, leading to self-desiccation of the concrete. In addition, as CSH is formed during the reaction, this produces a chemical shrinkage, since the volume of the hydrated products is less than the sum of the volume of the water plus the
initial anhydrous products. However, as the slag is not finely ground, only a small part of it reacts and the remainder acts as an inert aggregate that hinders the shrinkage of the paste. 3.5. Drying shrinkage Drying shrinkage cannot be measured. As many authors have pointed out, this shrinkage is obtained by subtracting autogenous from total shrinkage [36]. However, this process involves a small error, which can be seen in Fig. 14 in the form of a drop in the curve at advanced ages. This of course is impossible, since it would mean that the concrete stops shrinking and starts to swell. The error is probably due to the fact that self-desiccation is less pronounced in unsealed than in sealed specimens, since when the ambient RH is higher than that in the concrete, a small quantity of water vapour is absorbed by the concrete, which compensates for its self-desiccation. The more porous the concrete, the greater the error, since there is greater penetration of water vapour. In fact, the descending sides of the shrinkage curves tend to be more pronounced in the concretes with the highest slag contents, i.e. in the most porous. The higher the slag content in the SCC mix the higher the drying shrinkage (Fig. 14). Thus, for example at one year old, shrinkage in SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC-60 was, respectively, 4%, 11%, 15%, 19%, 24% and 36% higher than that of reference concrete SCC-0. Since the concretes made with slag are more porous, water is lost faster, as can be seen in Fig. 15. As the water
8
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Acknowledgements The authors of this paper would like to thank the Consejería de Universidades, Empresa e Innovación de la Comunidad Autónoma of the Region of Murcia for funding the SUE-ICI 07/02-0013 project, in which the Region of Murcia Construction Technology Centre (CTCON) and the companies HOLCIM and SIKA (Spain) also took part. References
Fig. 16. Total shrinkage.
evaporates the internal pressure in the capillary network increases and promotes drying shrinkage. Furthermore, since the porous structure is also finer in the concretes with the highest slag content, the attraction exerted by the capillary walls is greater and therefore shrinkage is higher. On the other hand, it has also been seen that the larger the quantity of slag used, the higher the volume of paste in the mixture and the lower its stiffness, so that there is less hindrance to the deformation due to shrinkage. 3.6. Total shrinkage Fig. 16 shows total shrinkage evolution with time. Since total shrinkage is autogenous plus drying shrinkage, in accordance with the results obtained in the previous sections, the higher the slag content in the SCC mix the higher the total shrinkage. At one year old shrinkage in SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC60 was, respectively, 4%, 10%, 17%, 19%, 26% and 44% higher than that of reference concrete SCC-0.
4. Conclusions In accordance with the tests carried out in the course of this work the following conclusions can be drawn: – Replacing sand by the type of slag used in this research gives rise to mixtures with higher total pore volume and slightly finer pore structure, with smaller median pore size and threshold diameter. – At early ages, slag SCCs show similar compressive strength to the reference SCC. However, at 365 days, due to slag reactivity, the higher the quantity of sand replaced by slag the higher the concrete’s compressive strength tends to be. – The SCCs with slag show a higher autogenous shrinkage, due to their higher concrete deformability, the higher self-desiccation generated by the slag hydration, and the chemical shrinkage caused by the slag reactivity. This increase is on average 11% and 33% when sand is replaced by 10% and 60% slag, respectively. – Due to their higher porosity, SCCs with slag are less stiff and lose water faster. This leads to higher drying shrinkage, which increases with the percentage of aggregate replaced by slag. – Due to higher autogenous shrinkage and drying shrinkage, the SCCs containing slag show higher total shrinkage. The increase is on average 4% and 44% when the sand replaced by slag is 10% and 60%, respectively.
[1] CEDEX; 2011. [accessed 01.2014]. [2] Puertas F. Escorias de alto horno: composición y comportamiento hidráulico. Mater Constr 1993;43:37–48. [3] Aïtcin PC. Binders for durable and sustainable concrete. London: Taylor and Francis; 2008. [4] Mosavinezhad SHG, Nabavi SE. Effect of 30% ground granulated blast furnace, lead and zinc slags as sand replacements on the strength of concrete. KSCE J Civil Eng 2012;16(6):989–93. [5] Shi D, Han P, Ma Z, Wang J. Report of experimented on compressive strength of concrete using granulated blast furnace slag as fine aggregate. Adv Mater Res 2012;575:100–3. [6] Saito K, Kinoshita M, Umehara H, Yoshida R. Properties of low-shrinkage, highstrength SCC using shrinkage-reducing admixture, blast furnace slag and limestone aggregates. In: Khayat KH, Feys D, editors. Design, production and placement of self-consolidating concrete, RILEM Bookseries 1; 2010. p. 283–93. [7] Scandiuzzi L, Battagin AF. A Utilização da escória granulada de altofornocomo agregado miúdo. Estudo Técnico n° 95. Associação Brasileira de Cimento Portland, São Paulo; 1990. [8] Topçu IB, Bilir T. Effect of non-ground-granulated blast-furnace slag as fine aggregate on shrinkage cracking of mortars. ACI Mater J 2010;107(6):545–53. [9] Yüksel I, Özkan Ö, Bilir T. Use of granulated blast-furnace slag in concrete as fine aggregate. ACI Mater J 2006;103(3):203–8. [10] Yüksel I, Genç A. Properties of concrete containing nonground ash and slag as fine aggregate. ACI Mater J 2007;104(4):397–403. [11] Yüksel I, Bilir T, Özkan Ö. Durability of concrete incorporating non-ground blast furnace slag and bottom ash as fine aggregate. Build Environ 2007;42(7):2651–9. [12] Bilir T. Effects of non-ground slag and bottom ash as fine aggregate on concrete permeability properties. Constr Build Mater 2012;26:730–4. [13] Aïtcin PC. Demystifying autogenous shrinkage. Concr Int 1999;21(11):54–6. [14] D’Ambrosia MD, Lange DA, Brinks AJ. Restrained shrinkage and creep of selfconsolidating concrete. In: Shah SP, editor. Proceedings of 2nd North American conference on the design and use of self-consolidating concrete (SCC) and the 4th international RILEM symposium on self-compacting concrete, Chicago; 2005. p. 921–8. [15] Arellano R, Burciaga O, Escalante JI. Lightweight concretes of activated metakaolin–fly ash binders, with blast furnace slag aggregates. Constr Build Mater 2010;24:1166–75. [16] Chopin D, Francy O, Lebourgeois S, Rougeau P. Creep and shrinkage of heatcured self-compacting concrete (SCC). In: Wallevik O, Nielsson I, editors. Proceedings of the 3rd international RILEM symposium on self-compacting concrete, RILEM Publications S.A.R.L., Reykjavik; 2003. p. 672–83. [17] Japan Concrete Institute. Technical committee on autogenous shrinkage of concrete. Taylor & Francis; 1999. [18] Feldman RF. Pore structure damage in blended cements caused by mercury intrusion. J Am Ceram Soc 1984;67(1):30–3. [19] Shi D, Winslow DN. Contact angle and damage during mercury intrusion into cement paste. Cem Concr Res 1985;15(4):645–54. [20] Olson RA, Neubauer CM, Jennings HM. Damage to the pore structure of hardened Portland cement paste by mercury intrusion. J Am Ceram Soc 1997;80(9):2454–8. [21] Matschei T, Lothenbach B, Glasser FP. The role of calcium carbonate in cement hydration. Cem Concr Res 2007;37(4):551–8. [22] Bosiljkov VB. SCC mixes with poorly graded aggregate and high volume of limestone filler. Cem Concr Res 2003;33(9):1279–86. [23] Ye G, Xiu X, De Schutter G, Poppe AM, Taerwe L. Influence of limestone powder as filler in SCC on hydration and microstructure of cement pastes. Cement Concr Compos 2007;29(2):94–102. [24] Leemann A, Münch B, Gasser P, Holzer L. Influence of compaction on the interfacial transition zone and the permeability of concrete. Cem Concr Res 2006;36(8):1425–33. [25] Kakali G, Tsivilis S, Aggeli E, Bati M. Hydration products of C3A, C3S and Portland cement in the presence of CaCO3. Cem Concr Res 2000;30(7):1073–7. [26] Bonavetti VL, Rahal VF, Irassar EF. Studies on the carboaluminate formation in limestone filler-blended cements. Cem Concr Res 2001;31(6):853–9. [27] Matschei T, Lothenbach B, Glasser FP. The role of calcium carbonate in cement hydration. Cem Concr Res 2007;37(4):551–8. [28] Lothenbach B, Le Saout G, Gallucci E, Scrivener K. Influence of limestone on the hydration of Portland cements. Cem Concr Res 2008;38(6):848–60. [29] Ollivier JP, Maso JC, Bourdette B. Interfacial transition zone. Adv Cem Based Mater 1995;2(1):30–8.
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9 [30] Valcuende M, Parra C, Marco E, Garrido A, Martínez E, Cánoves J. Influence of limestone filler and viscosity-modifying admixture on the porous structure of self-compacting concrete. Constr Build Mater 2012;28(1):122–8. [31] Zhao Q, Stark J, Freyburg E, Zhou M. The mechanism of ground granulated blast furnace slag preventing alkali aggregate reaction. J Wuhan Univ TechnolMater 2010;225(2):332–41. [32] Kumar R, Bhattacharjee B. Porosity, pore size distribution and in situ strength of concrete. Cem Concr Res 2003;33(1):155–64. [33] Feldman F, Beaudoin JJ. Pretreatment of hardened hydrated cement pastes for mercury intrusion measurements. Cem Concr Res 1991;21(2–3): 297–308. [34] Brooks JJ, Cabrera JG, Megat MA. Factors affecting the autogenous shrinkage of silica fume high-strength concrete. In: Tazawa E, editor. Autogenous shrinkage of concrete. Taylor & Francis; 1999. p. 195–202.
9
[35] Valcuende M, Marco E, Parra C, Serna P. Influence of limestone filler and viscosity-modifying admixture on the shrinkage of self-compacting concrete. Cem Concr Res 2012;42(4):583–92. [36] RILEM. Mechanical properties of self-compacting concrete. In: Khayat KH, De Schutter G, editors. RILEM state of the art reports; 2014. [37] Rozière E, Granger S, Turcry P, Loukili A. Influence of paste volume on shrinkage cracking and fracture properties of self-compacting concrete. Cem Concr Compos 2007;29(8):626–36. [38] Loser R, Leemann A. Shrinkage and restrained shrinkage cracking of selfcompacting concrete compared to conventionally vibrated concrete. Mater Struct 2009;42(1):71–82. [39] Tazawa E, Miyazawa S. Effect of constituents and curing conditions on autogenous shrinkage of concrete. In: Tazawa E, editor. Autogenous shrinkage of concrete. Taylor & Francis; 1999. p. 269–80.
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Shrinkage of self-compacting concrete made with blast furnace slag as fine aggregate M. Valcuende a,⇑, F. Benito b, C. Parra b, I. Miñano b a b
Department of Architectural Constructions, Universitat Politècnica de València, Camino de vera s/n, 46022 Valencia, Spain Department of Structures and Construction, Universidad Politécnica de Cartagena, Doctor Fleming s/n, 30202 Cartagena, Spain
h i g h l i g h t s Replacing sand by slag gives rise to mixtures with higher total pore volume. At early ages SCCs with slag show similar strength, but in the long term it increases. SCCs with slag show a higher autogenous shrinkage. Due to their higher porosity, SCCs with slag are less stiff and lose water faster. SCCs containing slag show higher total shrinkage.
a r t i c l e
i n f o
Article history: Received 18 March 2014 Received in revised form 2 October 2014 Accepted 12 November 2014
Keywords: Self-compacting concrete Shrinkage Compressive strength Granulated blast furnace slag Pore size distribution
a b s t r a c t The aim of this experimental work was to study shrinkage evolution with age in self-compacting concretes (SCC) in which part of the fine aggregate was replaced by granulated blast furnace slag (GBFS) as sand. Seven types of SCC were made with a w/c ratio of 0.55 and different slag contents. The results show that replacing sand by GBFS gives rise to mixes with higher pore volume but with slightly finer porous structure (smaller median pore and threshold diameters). At early ages slag SCCs have similar compressive strength to that of the reference concrete, although in the long term their strength increases as a result of slag reactivity. We also observed that the higher the slag content, the higher were both autogenous and drying shrinkage and consequently also total shrinkage. In comparison with the reference concrete, the increase in total shrinkage was found to be of the order of 4% and 44% when 10% and 60%, respectively, of the sand was replaced by slag. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Melted slag from a blast furnace can be subjected to various cooling techniques in order to obtain products with a wide range of end-uses. Granulated blast furnace slag (GBFS) is obtained by the rapid cooling of liquid slag. At least two-thirds of its mass consists of vitreous slag [1], which has hydraulic properties when activated and leads to the formation of C–S–H. It has a similar chemical composition to that of crystallized slag, although as it has been subjected to rapid cooling its ions are not given time to organize themselves into a crystalline network and solidify into a vitreous state. As is well known, the hydraulically active component resides in the vitreous phase and the crystalline phase can be considered as almost inert. It is generally thought that the greater its disordered structure the greater is its reactivity [2], although some authors ⇑ Corresponding author. Tel.: +34 963877450; fax: +34 963877459. E-mail address: [email protected] (M. Valcuende). http://dx.doi.org/10.1016/j.conbuildmat.2014.11.029 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
consider that the presence of a small proportion of the crystalline phase may be conducive to increased slag reactivity. The hydraulic capacity of slag is highly attenuated and is slow to appear. There are three methods of accelerating its hydration reactions: (i) by the use of chemical activators, (ii) increasing its specific surface and (iii) by raising its temperature. Portland cement is a good slag activation catalyst as it provides the three main compounds that activate slag: lime, calcium sulphate and alkalis [3]. Granulated slag is formed by small alveolar particles with sharp edges [1]. A number of studies have been published on the use of blast furnace slag as aggregate in concrete and some standards, such as ASTM C33, ASTM C989, EN-206 or EN-12620 also provide specifications for using this slag as aggregate. In many cases a problem arises when the authors do not specify whether they have worked with granulated or crystallized slag, which makes it difficult to reach general conclusions. Furthermore, in the studies which do indicate that granulated slag has been used, the results do not always coincide. According to Mosavinezhad and Nabavi [4] adding
2
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
GBFS to replace sand yielded increased compressive strength but gave lower flexural strength. Shi et al. [5] and Saito et al. [6] point out that at early ages the compressive strength of concrete using GBFS as fine aggregate is lower than concrete made with river sand, but at 91 days the strength is higher. Scandiuzzi and Battagin [7] also obtained better long-term strength due to the latent hydraulicity of the slag. On the contrary, Topçu and Bilir [8] point out that the compressive strength, flexural strength and modulus of elasticity are reduced due to the higher porosity of the mixture and that shrinkage cracking is also reduced. Yüksel et al. [9] and Yüksel and Genç [10] also maintain that replacing sand by GBFS may give rise to concretes of higher porosity and lower strength, according to the replacement ratio employed. However, when no more than 20% of the sand is replaced by slag, the concrete maintains a similar porous structure and GBFS has a positive effect on the concrete’s durability properties [11,12]. Replacing part of the aggregates by slag with hydraulic capacity could in theory give rise to concretes susceptible to higher shrinkage, partly due to the volume reduction of the cementitious materials that occurs during their hydration (chemical shrinkage). The lower the w/c ratio, the higher is this shrinkage, especially when the w/b ratio falls below 0.4 [13,14]. However, as can be seen from certain studies [6,8], using GBFS does not necessarily lead to higher shrinkage. GBFS reactions form additional C–S–H phases, which means that the pore network becomes more compact. Indeed, Arellano et al. [15] found that a dense and strong slag–paste interfacial zone was noted as a result of the blast furnace slag participation in the reactions. Increased matrix strength or density reduces material deformability [16] and thus restricts shrinkage. However, very few studies on the use of GBFS as fine aggregate have included the concrete shrinkage involved. Our objective in this experimental work was therefore to study shrinkage in self-compacting concretes (SCC) in which part of the limestone aggregate had been replaced by GBFS as fine aggregate and to compare the results obtained with those from concretes with no slag content. Due to the fact that shrinkage is directly related to the porous structure of the material, the microstructure of the different types of concrete was analysed in order to explain the behaviour observed. Using GBFS as aggregate firstly has the advantage of recycling waste and secondly the amount of natural aggregate used in the mix is reduced, with consequent savings in natural resources and the energy required for their extraction, together with a reduction of the amount of pollutants released into the atmosphere. 2. Experimental program
Fig. 1. Grading curves of slag and aggregates.
Fig. 2. Grading curves of cement, aggregate fines and limestone filler.
All the studied mixes were made with 375 kg/m3 of CEM II/B-M (S-L) 42.5R cement, and 25 kg/m3 of fly ash. The granular skeleton was composed of three fractions of crushed limestone aggregate: 4/12 mm coarse aggregate, 0/4 mm coarse sand and 0/2 mm fine sand, with a fines content (particle size <0.063 mm) of 1.0%, 12.0% and 17.1%, respectively (Fig. 1). The process of replacing sand by slag was carried out by firstly removing the fine sand only, and when this was exhausted (in concretes SCC-20–SCC-60), by also removing part of the coarse sand. Limestone filler was added to these mixes to make up for the lack of fines in the slag. The grading curves of the fines of aggregates and the filler were determined by means of laser diffraction (Fig. 2). As can be seen in this figure, the aggregate fines are finer than the limestone filler used. The basic GBFS used had high SiO2 and CaO contents, with a CaO/SiO2 ratio of 1.31. As regards its mechanical properties, the slag had a Micro-Deval index much lower than that of the limestone aggregates, and therefore had greater resistance to wear. The properties of the sand and slag are shown in Table 2 and the composition of the slag is shown in Table 3. The admixture used was a polycarboxylate-based superplasticizer (Viscocrete 3425). The concrete test specimens were made by filling moulds with a single uncompacted lift of SCC.
2.1. Concrete mixtures, materials and mixing procedure 2.2. Test program and methodology Seven types of concrete were made: a reference (SCC-0) and six with different percentages of fine aggregate replaced by slag (in weight): 10% (SCC-10), 20% (SCC20), 30% (SCC-30), 40% (SCC-40), 50% (SCC-50) and 60% (SCC-60). In order to obtain strengths approaching those normally used for building, all of them were manufactured with a 0.55 w/c ratio. The characteristics of each mix are shown in Table 1. All SCC mixes showed a slump flow of about 700 mm and good resistance to segregation in the segregation test.
2.2.1. Shrinkage test Prismatic specimens measuring 100 100 400 mm were used to determine autogenous shrinkage and total shrinkage. Temperature and weight loss were also measured in order to obtain further information on this phenomenon. Thermocouples were inserted into the centre of the specimens immediately after casting to determine the evolution of temperature rise due to the hydration reaction.
Table 1 Mixture proportions of concretes. Mix
Sand replacement ratio (%)
Cement (kg/m3)
Fly ash (kg/m3)
Water (l/m3)
WEA (kg/m3)
Coarse aggregate (kg/m3)
Coarse sand (kg/m3)
Fine sand (kg/m3)
Limestone filler (kg/m3)
Slag sand (kg/m3)
Paste volume (dm3)
SCC-0 SCC-10 SCC-20 SCC-30 SCC-40 SCC-50 SCC-60
0 10 20 30 40 50 60
375 375 375 375 375 375 375
25 25 25 25 25 25 25
220 220 220 220 220 220 220
4.4 4.2 4.4 4.6 4.8 5.2 6.0
656.6 656.8 656.6 656.5 656.3 656.0 655.4
829.7 829.6 818.4 691.4 564.4 437.5 310.5
260.3 124.8 0.0 0.0 0.0 0.0 0.0
0.0 26.9 53.6 71.4 89.1 106.8 124.4
0.0 102.7 205.4 308.0 410.6 512.9 614.7
410.9 412.1 413.9 415.1 416.2 417.6 419.3
(WEA: water enhancement admixture).
3
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9 Table 2 Properties of the aggregates. Aggregate
Micro-Deval index
Specific gravity (g/dm3)
Water absorption (%)
Coarse sand (0/4) Fine sand (0/2) Slag
43 43 14
2.60 2.60 2.45
1.0 1.0 8.0
Table 3 Chemical composition of the slag.
Slag
SiO2 (%)
SO3 (%)
CaO (%)
MgO (%)
(CaO + MgO)/ SiO2
CaO + MgO + SiO2 (%)
30.2
1.9
39.5
7.5
1.5
77.2
Fig. 4. Shrinkage test (age more than 24 h).
Table 4 Pore characteristics of concrete mixtures.
Total intrusion volume (10 3) (mm3/g) Total pore area (m2/g) Median pore diameter (lm) Threshold diameter (lm) Total porosity (%)
SCC-0
SCC-10
SCC-20
SCC-40
SCC-60
52.7 (21.50) 10.3 (30.97) 0.05 (11.86) 0.15 (14.23) 11.91 (17.62)
61.2 (8.09) 13.3 (3.64) 0.05 (10.22) 0.12 (7.96) 13.51 (9.09)
62.3 (3.59) 14.3 (4.02) 0.04 (13.64) 0.12 (4.23) 13.88 (2.87)
63.4 (8.59) 15.0 (4.57) 0.04 (7.41) 0.11 (5.01) 13.91 (7.22)
66.8 (12.28) 17.2 (22.16) 0.03 (14.27) 0.09 (16.53) 14.57 (10.27)
Numbers in parentheses are the coefficients of variation (%).
Fig. 3. Measurement of shrinkage before mould removal (age less than 24 h).
In order to measure shrinkage of the specimens until mould removal (24 h), the specimens were made in accordance with the recommendations of the Technical Committee on Autogenous Shrinkage of Concrete of the Japan Concrete Institute [17]. A polystyrene sheet (thickness 3 mm) was placed on the bottom and on both sides of the mould so that free movement of the specimen was not restricted by the mould. A polyester film was also placed over the polystyrene sheet on all sides of the mould and on the surface of the specimen. As the specimens were produced, containers measuring 25 cm diameter and 20 cm high were filled with the same concrete to determine setting time. According to Standard ASTM C 403-08, setting time is established on the basis of concrete hardening as measured by penetration resistance needles. At the time of initial setting, the end plates of the mould were removed and two dial gauges were put into place (Fig. 3). Later, at the age of 24 h, the specimens to be used for measuring autogenous shrinkage, total shrinkage and weight loss were removed from the moulds and were stored in a vertical position inside a climatically controlled chamber for one year at 20 °C and 50% RH (Fig. 4). To determine autogenous shrinkage, half of the specimens were sealed with various layers of plastic film to avoid moisture loss. Those used to measure total shrinkage were left unsealed. Shrinkage was measured by a dial gauge of 1 lm resolution placed on the upper surface of the specimen (Fig. 4). The specimens were weighed on a balance accurate to within 0.1 g. Three batches were made from each mix. One specimen for each type of test was made from each batch, the result of each test being the arithmetic mean of the three values obtained. Two 150 mm diameter 300 mm high cylindrical specimens were also made from each batch for compressive strength tests at 7, 28, 90 and 365 days (EN 12390-3:2003). Two extra specimens with the same dimensions were produced to determine the elasticity modulus (EN 1352:1997) at 28 days. 2.2.2. Mercury intrusion porosimetry (MIP) test Pore size distribution was determined using a Micromeritics AutoPore IV-9500 mercury porosimeter with a maximum pressure of 60,000 psia (414 MPa). At this pressure the smallest pore size into which mercury can be introduced is 3 nm. This test was carried out on small drilled cores, weighing approximately 6 g. The cored samples were obtained from 100 100 100 mm cubic specimens. The samples were first dried in an oven at 110 °C and were then immersed in mercury under gradually increasing pressure. As pressure rises, mercury is forced into the pores of the sample. By modelling the pores as cylindrical channels, the test pressure can be connected to the radius of these cylinders by the Washburn–Laplace law. Using this technique, a measure of the total sample porosity may also be obtained,
as well as the pore network surface area. However the MIP technique has certain limitations. In cement paste, high pressure is a source of error due to possible damage occurring during intrusion [18–20]. The microstructure of hardened cement paste is composed of pores that are separated by thin walls of hydration product. High pressures are required to measure smaller pores, so mercury enters into pores by breaking through the pore structure [18]. Olson et al. [20] found significant damage caused by relatively low pressures and much higher pore connectivity after intrusion. When the barrier between the pores is destroyed, a volume change is measured where a pore does not exist, and the data actually represent the pressure required to break through the barrier. Despite these limitations, the MIP technique is still an effective aid for comparing the pore structure and pore network characteristics of different types of cement-based materials. A sample was taken from each of the three batches of each type of concrete. The results shown in Table 4 are the arithmetic mean of the three values obtained.
2.2.3. Compressive strength tests of mortars containing slag In order to study the slag’s hydraulic activity and its evolution with time, four different mortar types were made with Portland cement without additions (CEM I), CEN standard silica sand (European Standard EN 196-1) and the same quantity of water: a reference mortar without slag (M-1) and with a mix proportion 1:3:0.5 (cement:silica sand:water), one with half the cement replaced by limestone filler (M-3), another with half the cement replaced by milled slag with a particle size similar to that of the filler (M-4), and lastly another with half the cement only, without adding either limestone filler or slag (M-2). The specimens were tested to compressive strength at 2, 7, 28, 90 and 365 days. Three batches were made from each mix. Two specimens were made from each batch, the result of each test being the arithmetic mean of the six values obtained.
3. Results and discussion 3.1. Reactivity of finely milled slag The slags used in this work present a CaO/SiO2 basicity index equal to 1.31 (Table 3) and are therefore potentially hydraulic [2]. This slag reactivity can also be seen in the X-ray diffraction test (Fig. 5), since the diffractogram obtained is characteristic of amorphous slag, with almost no crystalline mineral peaks but rather a hump located at a 2h angle around 30°. This angle corresponds to the principal peak of melilite, which is the main mineral of the
4
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 5. X-ray diffractogram of slag.
crystallized slag. The small peaks detected are of calcite (CaCO3) and mullite (3Al2O32SiO2). To test the slag’s hydraulic activity, various mortar types were made with different slag contents and were tested under compression at various ages. As can be seen in Fig. 6, the mortar made with 50% less cement (Mix M-2), as could be expected, shows much less strength than reference mortar (M-1), of the order of 77%. However, when the eliminated cement is replaced by limestone fines (M-3), the loss of compressive strength is not so pronounced, since the fines occupy the available space and give a denser cementitious matrix (filler effect). Some authors also point out that small amounts of limestone additions increase the volume of solids in hydrated cements [21]. On the other hand, it can be seen that less strength is lost at early ages, as limestone fines accelerate cement hydration, due to the fact that they provide additional surfaces for the nucleation and growth of hydration products [22,23]. At more advanced ages this effect is no longer relevant; for example at 2 days the loss of strength is more than 52% higher than that of the reference mortar, although at 28, 90 and 365 days it remains almost constant, around 62%. It can also be seen in Fig. 6 that the compressive strength of mortar containing slag (M-4) is considerably higher than the mortar with limestone fines. This difference increases with the age of the specimens. As ground slag is more reactive, finely ground slag acts as a cementitious material. In addition, powdered slag also fills up the gaps and creates nucleation centres in the same way as limestone fines. Slag hydration is quite slow, as at early ages most of the slag remains inert. The compressive strength of the mortar at 2 days old is around 50% of that of the reference mortar. As time passes the hydration reactions increase and hydrated calcium silicates are formed, so that after one year the strength reaches that of the reference mortar. However, it should not be forgotten that finely ground slag particles are more reactive than sand-sized slag particles.
Fig. 6. Evolution of compressive strength with time.
3.2. Concrete porous microstructure Fig. 7 shows the increase in mercury intrusion volume according to the equivalent pore diameter. In all the mixes, pore volume distribution follows a similar pattern, each curve having two peaks: one around pore sizes of 90 lm (peak no. 2) and another, much more pronounced, around sizes 0.05 lm (peak no. 1). The porous structure of all the concretes analysed is quite similar. The small differences between the different mixes are only apparent in the volume of the smallest capillary pores (0.01–0.1 lm).
Fig. 7. Pore size distribution.
The volume of pores larger than 0.1 lm in diameter is very similar in all the concretes, although the number of the smallest pores increases when the percentage of sand replaced by slag is higher (Fig. 8). This gives rise to a higher total pore volume in the concretes made with higher quantities of slag (Table 4). Since the
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 8. Cumulative mercury intrusion volume.
volume of cement paste (water + cement) is the same in all the types of concrete, this higher porosity is due to the slag particles being more porous than the grains of sand (Table 2) and probably to the higher fines content of the concretes made with slag (in which part of the sand is replaced by limestone filler). The higher fines content also creates a finer porous structure, since the fines fill up the small gaps (filler effect) and provide additional surfaces for the nucleation of hydration products (since there are more nucleation sites, the size of portlandite crystals is smaller and therefore the ITZ is not so thick [24]). Moreover, a small part of the CaCO3 present in the limestone filler promotes the reconversion of monosulfoaluminate into ettringite, leading to an increase of the total volume of the hydrate phase [25–28] (ettringite has a low density and therefore a relatively large volume per formula unit). On the other hand, thanks to reaction on the surface of the slag grains (the smaller the particles, the greater the reaction), the volume of hydration products increases with the passage of time (gradual formation of CSH) and the aggregate–paste interface zone becomes denser, which is the zone where the porosity is higher and the capillary pores are larger than those in cement paste [29,30]. In fact, the SEM images taken at different ages show the presence of some compounds in the slag–paste interfacial transition zone that are not present in the rest of the paste, as for example Mg (Figs. 9 and 10). The Mg content (coming from the slag [31]) is rapidly reduced as distance from the aggregates increases and tends to practically disappear at a distance of a few microns.
5
In samples analysed at 28 days (Fig. 9) it can be seen that the width of the zone where the Mg diffusion has been produced is about 2.5 lm. At 120 days, due to the slow slag reactivity, the thickness of this zone increases to around 5 lm (Fig. 10). It can also be seen that in this zone the concentration of Mg atoms is similar to the concentration in the slag itself. It can therefore be said that SCCs with slag as aggregate are slightly more porous but have a somewhat finer porous structure. This aspect is important since, according to Kumar and Bhattacharjee [32], the smallest pores have no influence on the concrete’s strength properties but do have a direct relationship with shrinkage. It can also be seen from Fig. 7 and Table 4 that the maximum pore diameter at which the continuous mercury intrusion rapidly increases varies slightly between mixes. This value, known as the threshold diameter, signals the pore diameter after which the highest number of pores are concentrated and is therefore a good indicator of the fineness of the porous structure. In this work, the threshold diameter is the smallest diameter with a differential intrusion volume >0.001 cm3/g [33]. Here again it is seen that as slag content increases the threshold diameter tends to diminish. Thus, for example, in SCC-0, SCC-20, SCC-40 and SCC-60 the threshold pore width is respectively equal to 0.15, 0.12, 0.10 and 0.09 lm. The pore size of the maximum pore concentration and median pore diameter are two other important parameters related to the fineness of pore structure. Fig. 7 shows that in SCC-0 the maximum concentration of pores (peak no. 1) tends to be around larger pore sizes than in SCCs with slag. For example, in SCC-0 the maximum concentration of pores is for a pore size of 0.065 lm and in SCC10, SCC-20, SCC-40 and SCC-60 for a pore size of 0.05, 0.05, 0.05 and 0.04 lm, respectively. Median pore diameter is also smaller in the SCCs containing slag, the size being in inverse proportion to the slag content (Table 4). 3.3. Compressive strength The compressive strength results are shown in Table 5. Although the results are very similar in all concretes, with no statistically significant differences between them, at early ages (7 days) those with the highest slag content tend to have lower compressive strength. This could be due to the higher porosity of these concretes, as observed in the MIP tests, and the greater volume of paste in their mixes (Table 1). At more advanced ages this behaviour changes. Since reactive slags were used in this work, as hydration proceeds CSH is formed around the particles (Figs. 9 and 10) and the aggregate–paste interface becomes denser, improving the aggregate–paste bond. In fact,
Fig. 9. SEM of slag–paste interface and element distribution of interfacial zone by linear scanning method at the age of 28 days (1300).
6
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 10. SEM of slag–paste interface and element distribution of interfacial zone by linear scanning method at the age of 120 days (1300).
Table 5 Mechanical properties of the concretes (MPa). Mix
SCC-0 SCC-10 SCC-20 SCC-30 SCC-40 SCC-50 SCC-60
Compressive strength
Elastic modulus
7 days
28 days
90 days
365 days
28 days
33.0 (1.26) 31.0 (2.93) 34.0 (0.34) 32.8 (2.73) 31.2 (1.37) 31.3 (0.28) 29.9 (0.71)
39.2 (3.14) 38.9 (2.13) 38.7 (1.49) 39.2 (3.53) 38.9 (3.09) 38.2 (1.39) 38.7 (2.05)
46.7 (0.39) 46.5 (0.49) 46.7 (1.46) 47.5 (4.88) 45.5 (1.60) 48.7 (2.02) 47.6 (0.58)
49.6 (0.96) 49.2 (0.25) 49.7 (1.88) 51.1 (2.92) 50.9 (1.22) 51.8 (1.67) 52.3 (0.59)
30,539 (1.53) 29,017 (0.26) 29,404 (0.15) 28,750 (5.69) 28,445 (2.30) 28,804 (1.31) 28,652 (2.41)
Numbers in parentheses are the coefficients of variation (%).
according to the MIP tests, the mixtures with the highest percentages of sand replaced by slag show a finer porous structure that is probably due to greater CSH formation. However, as has been seen in Section 3.2, such hydration reactions are quite slow, so that the improvement in mechanical behaviour is also slow. Thus, for example, at 28 and 90 days the compressive strength of all concretes is practically the same, with no statistically significant differences between them (P-value = 0.73 at 28 days and P-value = 0.19 at 90 days, for a 95% confidence level). However, at 365 days, the higher the aggregate slag content the higher the compressive strength of the concrete tends to be (P-value = 0.04). 3.4. Autogenous shrinkage Shrinkage strain and thermal strain from the heat generated during cement hydration are created simultaneously. The temperature of the concrete was monitored by thermocouples from casting up to the age of 28 h (Fig. 11). During this time period the temperature varied only slightly (less than 1.0 °C in all cases) due partly to the chemical composition of the cement (low C3A content) and partly to the small size of the concrete specimens. In Fig. 11 it can be seen that the rise in temperature is similar in all concretes and happens in the first 10–12 h, after which it drops to ambient temperature before reaching 24 h of age. In order to eliminate the temperature effect, the shrinkage values were
Fig. 11. Temperature evolution in the first 28 h.
corrected considering De = aDT, where De is the thermal strain, a is the coefficient of thermal expansion (1/°C) and DT is the temperature change (°C). The strain/time curve corrected for the temperature effect corresponds to the volumetric change of concrete under isothermal conditions. As in the case of many other study groups, a = 10 5/°C was selected. After 24 h the temperature variation in the studied mixes was small, so this effect was neglected. The evolution of autogenous shrinkage with time is shown in Fig. 12. In general, the autogenous shrinkage values obtained by other research groups are lower than those obtained in the present work. However, in agreement with the results obtained in other studies [34,35], this shrinkage forms a considerable percentage of total shrinkage. For example, at one month old this percentage varies between 36% and 44% in the seven concrete types analysed (Fig. 13). As the specimens lose water and reach hygrometric equilibrium with the atmosphere, the drying shrinkage decreases and stabilizes, although as the cement and the slag continue to hydrate, autogenous shrinkage due to self-desiccation inside the specimen still goes on. With the passage of time the autogenous shrinkage/ total shrinkage ratio therefore increases until it reaches values between 68% and 78% (at the age of twelve months), depending on the type of concrete, with similar behaviour in all types. The SCCs containing slag show higher autogenous shrinkage than the reference concrete, with higher shrinkage when the quantity of sand replaced by slag is increased. Thus, for example, at
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Fig. 12. Autogenous shrinkage.
7
Fig. 14. Drying shrinkage.
Fig. 13. Ratio between autogenous shrinkage and total shrinkage. Fig. 15. Weight loss.
12 months old SCC-0 has 11%, 14%, 21%, 26%, 28% and 33% less shrinkage than SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC60, respectively. This can be explained by several reasons; firstly, the concretes made with slag have higher volume of paste (Table 1), so that they are more likely to experience greater autogenous shrinkage [36]. However, some authors [37,38] point out that increasing only the content of limestone powder but keeping cement and water constant has only a minor effect on autogenous shrinkage. Secondly, as we have seen in the MIP tests, the higher the quantity of slag used the higher the total pore volume and consequently the concrete is more prone to deformation. In addition, the slags used are much more porous than limestone aggregates (Table 2). They are therefore probably less stiff and less likely to hinder the deformation of the concrete. In fact, in the elasticity modulus tests, the concretes with slag were seen to lose stiffness (Table 5). Thirdly, as the porous structure is also finer in concretes with higher slag content, they can be expected to show higher shrinkage, since the smaller the capillaries the higher is the tensile stress generated by the water menisci in the capillaries [35,39] (due to the surface tension of the water, as the capillary pores lose water, the force of attraction between pore walls increases and promotes shrinkage). The fourth factor to be borne in mind is the reactivity of the slag as seen in the different tests carried out (Xray diffraction, SEM and compressive strength of mortars). As a result of hydration of the surface of the slag particles, the amount of water in the pores is reduced, leading to self-desiccation of the concrete. In addition, as CSH is formed during the reaction, this produces a chemical shrinkage, since the volume of the hydrated products is less than the sum of the volume of the water plus the
initial anhydrous products. However, as the slag is not finely ground, only a small part of it reacts and the remainder acts as an inert aggregate that hinders the shrinkage of the paste. 3.5. Drying shrinkage Drying shrinkage cannot be measured. As many authors have pointed out, this shrinkage is obtained by subtracting autogenous from total shrinkage [36]. However, this process involves a small error, which can be seen in Fig. 14 in the form of a drop in the curve at advanced ages. This of course is impossible, since it would mean that the concrete stops shrinking and starts to swell. The error is probably due to the fact that self-desiccation is less pronounced in unsealed than in sealed specimens, since when the ambient RH is higher than that in the concrete, a small quantity of water vapour is absorbed by the concrete, which compensates for its self-desiccation. The more porous the concrete, the greater the error, since there is greater penetration of water vapour. In fact, the descending sides of the shrinkage curves tend to be more pronounced in the concretes with the highest slag contents, i.e. in the most porous. The higher the slag content in the SCC mix the higher the drying shrinkage (Fig. 14). Thus, for example at one year old, shrinkage in SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC-60 was, respectively, 4%, 11%, 15%, 19%, 24% and 36% higher than that of reference concrete SCC-0. Since the concretes made with slag are more porous, water is lost faster, as can be seen in Fig. 15. As the water
8
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9
Acknowledgements The authors of this paper would like to thank the Consejería de Universidades, Empresa e Innovación de la Comunidad Autónoma of the Region of Murcia for funding the SUE-ICI 07/02-0013 project, in which the Region of Murcia Construction Technology Centre (CTCON) and the companies HOLCIM and SIKA (Spain) also took part. References
Fig. 16. Total shrinkage.
evaporates the internal pressure in the capillary network increases and promotes drying shrinkage. Furthermore, since the porous structure is also finer in the concretes with the highest slag content, the attraction exerted by the capillary walls is greater and therefore shrinkage is higher. On the other hand, it has also been seen that the larger the quantity of slag used, the higher the volume of paste in the mixture and the lower its stiffness, so that there is less hindrance to the deformation due to shrinkage. 3.6. Total shrinkage Fig. 16 shows total shrinkage evolution with time. Since total shrinkage is autogenous plus drying shrinkage, in accordance with the results obtained in the previous sections, the higher the slag content in the SCC mix the higher the total shrinkage. At one year old shrinkage in SCC-10, SCC-20, SCC-30, SCC-40, SCC-50 and SCC60 was, respectively, 4%, 10%, 17%, 19%, 26% and 44% higher than that of reference concrete SCC-0.
4. Conclusions In accordance with the tests carried out in the course of this work the following conclusions can be drawn: – Replacing sand by the type of slag used in this research gives rise to mixtures with higher total pore volume and slightly finer pore structure, with smaller median pore size and threshold diameter. – At early ages, slag SCCs show similar compressive strength to the reference SCC. However, at 365 days, due to slag reactivity, the higher the quantity of sand replaced by slag the higher the concrete’s compressive strength tends to be. – The SCCs with slag show a higher autogenous shrinkage, due to their higher concrete deformability, the higher self-desiccation generated by the slag hydration, and the chemical shrinkage caused by the slag reactivity. This increase is on average 11% and 33% when sand is replaced by 10% and 60% slag, respectively. – Due to their higher porosity, SCCs with slag are less stiff and lose water faster. This leads to higher drying shrinkage, which increases with the percentage of aggregate replaced by slag. – Due to higher autogenous shrinkage and drying shrinkage, the SCCs containing slag show higher total shrinkage. The increase is on average 4% and 44% when the sand replaced by slag is 10% and 60%, respectively.
[1] CEDEX; 2011.
M. Valcuende et al. / Construction and Building Materials 76 (2015) 1–9 [30] Valcuende M, Parra C, Marco E, Garrido A, Martínez E, Cánoves J. Influence of limestone filler and viscosity-modifying admixture on the porous structure of self-compacting concrete. Constr Build Mater 2012;28(1):122–8. [31] Zhao Q, Stark J, Freyburg E, Zhou M. The mechanism of ground granulated blast furnace slag preventing alkali aggregate reaction. J Wuhan Univ TechnolMater 2010;225(2):332–41. [32] Kumar R, Bhattacharjee B. Porosity, pore size distribution and in situ strength of concrete. Cem Concr Res 2003;33(1):155–64. [33] Feldman F, Beaudoin JJ. Pretreatment of hardened hydrated cement pastes for mercury intrusion measurements. Cem Concr Res 1991;21(2–3): 297–308. [34] Brooks JJ, Cabrera JG, Megat MA. Factors affecting the autogenous shrinkage of silica fume high-strength concrete. In: Tazawa E, editor. Autogenous shrinkage of concrete. Taylor & Francis; 1999. p. 195–202.
9
[35] Valcuende M, Marco E, Parra C, Serna P. Influence of limestone filler and viscosity-modifying admixture on the shrinkage of self-compacting concrete. Cem Concr Res 2012;42(4):583–92. [36] RILEM. Mechanical properties of self-compacting concrete. In: Khayat KH, De Schutter G, editors. RILEM state of the art reports; 2014. [37] Rozière E, Granger S, Turcry P, Loukili A. Influence of paste volume on shrinkage cracking and fracture properties of self-compacting concrete. Cem Concr Compos 2007;29(8):626–36. [38] Loser R, Leemann A. Shrinkage and restrained shrinkage cracking of selfcompacting concrete compared to conventionally vibrated concrete. Mater Struct 2009;42(1):71–82. [39] Tazawa E, Miyazawa S. Effect of constituents and curing conditions on autogenous shrinkage of concrete. In: Tazawa E, editor. Autogenous shrinkage of concrete. Taylor & Francis; 1999. p. 269–80.