- Email: [email protected]
Cooling history of a wet-granulated blast furnace slag (GBS)
Journal of Non-Crystalline Solids 499 (2018) 344–349
Contents lists available at ScienceDirect
Journal of Non-Crystalline Solids journal homepage: www.elsevier.com/locate/jnoncrysol
Cooling history of a wet-granulated blast furnace slag (GBS) a
N. Pronina , S. Krüger a b
a,1
a
a,⁎
, H. Bornhöft , J. Deubener , A. Ehrenberg
T
b
Institute of Non-Metallic Materials, Clausthal University of Technology, Germany FEhS – Institut für Baustoff-Forschung e.V, Duisburg, Germany
A R T I C LE I N FO
A B S T R A C T
Keywords: Blast furnace slag Cooling rate Calcium aluminosilicate glass Thermal history Fictive temperature Hyperquenched glass
Granulated blast furnace slag (GBS) is a glassy by-product of the steel industry that is formed during quenching of the molten slag after the blast furnace in a water-jet. To elucidate the cooling history of GBS, calorimetric scanning in the glass transition range and viscometric experiments at temperatures above the liquidus were performed. The GBS studied was of industrial origin with a d50 = 700 μm and > 99% glassy. It is found that GBS glass is of high potential energy, showing a fictive temperature that is approx. 160 K higher than the glass transition temperature under standard cooling conditions. Using different viscosity models and the relationship between quench rate and shear viscosity at the fictive temperature it is calculated that GBS was formed at a cooling rate of approx. 2.6 × 105 K s−1 corresponding to a shear viscosity of 105.9 Pa s. Due to the high Tf and high amount of heat that is released during structural relaxation, GBS is assigned to the class of hyperquenched glasses.
1. Introduction Granulated blast furnace slag (GBS) is a non-metallic by-product of the steel industry (annual GBS production is approx. 280 million tons) that is formed when after leaving the blast furnace the molten slag is quenched using powerful water jets or fast rotating drums in air. GBS is a fine, granular, almost fully glassy material of calcium aluminosilicate composition. GBS material exhibits latent hydraulic (cementitious) properties [1–4] and furthermore as cement constituent or concrete addition it serves advantageous technical, environmental and economic benefits [5–8]. The reactivity of ground GBS (so-called GGBS) in cement is found to be influenced by the slag properties such as the glass content and its chemical composition as well as the mineralogical composition of residual crystal fraction [9, 10]. As the residual crystalline material is unreactive, small crystal fractions can be beneficial for blending cement. Thus, temperature-time-transformation (TTT) and continuouscooling-transformation (CCT) diagrams have been determined using the hot thermocouple method in which the tips of a thermocouple are welded in a U-type shape to heat and measure the temperature simultaneously [11, 12]. On the one hand TTT and CCT curves show that GBS glass is formed when the cooling rate exceeds 100–101 K s−1. On the other hand, the cooling rates of industrially manufactured GBS are generally unknown and only estimated to be above 102 K s−1 [11]. However, variation in operation parameter, such as molten slag
temperature, water jet flow and disc rotation speed may cause the formation of GBS glass with different cooling history and therefore different structure and properties [13]. As a consequence of these fluctuations, the performance of GBS from different production campaigns and sites becomes less predictable. Generally, the cooling history of glasses can be estimated through their fictive temperature Tf. Despite the fact that the dependence of Tf on cooling rate has been established for > 40 years [14–17], it has not been applied to GBS. The present work aims therefore to express the cooling history of GBS by Tf and cooling rate. The former is determined by calorimetric scanning, whereas the latter requires information on the viscosity-temperature curve. To provide viscosity data at temperatures above the liquidus, rotational viscometry of the molten granules was utilized. The implementation of Tf and cooling rate as parameters of the GBS production might allow assessment of samples independently of their composition and simplify the control and modification of the quenching process according to desired properties (e.g. reactivity in cementitious systems). 2. Material and methods 2.1. Material
⁎
Corresponding author. E-mail address: [email protected] (J. Deubener). 1 Present address: SCHOTT AG, Mainz, Germany. https://doi.org/10.1016/j.jnoncrysol.2018.07.054 Received 1 June 2018; Received in revised form 20 July 2018; Accepted 23 July 2018 Available online 27 July 2018 0022-3093/ © 2018 Elsevier B.V. All rights reserved.
The wet-granulated blast furnace slag (GBS) under investigation was
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
of industrial origin (moisture 10.6 wt%) and of chemical composition 0.03 P2O5, 33.4 SiO2, 1.98 TiO2, 11.4 Al2O3, 44.9 CaO, 5.7 MgO, 0.75 FeO, 0.23 MnO, 0.36 K2O and 0.17 Na2O (wt%) as analyzed by X-ray fluorescence (Σ = 98.92%; 100-Σ = 1.08% volatile species). Transmission optical microscopy of the immersed granules revealed a glassy volume fraction > 99%. Granulometry (sieve analysis) revealed a monomodal particle size distribution with a mean size d50 = 700 μm (not shown). The GBS glass was analyzed as received without any processing prior to analysis.
lg η = A 4 +
3. Experimental 3.1. Calorimetry
Using logarithmic scales, the cooling rate qc and the shear viscosity at the fictive temperature η(Tf) are related by a dimensionless shift factor K in the form
lg qc = K − lg η (Tf )
To compensate for possible inhomogeneity, three samples of the same lot were used for calorimetry. The first sample was measured in a DSC (404F3 Pegasus; Netzsch, Selb, Germany), whereas for samples 2 and 3 a TGA equipped with DSC sensor (TGA/DSC 3+, Mettler Toledo, Greifensee, Switzerland) was used. Each measurement was carried out with 20–30 mg GBS granules in a PtRh-crucible in a nitrogen environment. Sample 1 was subjected to the sequence of heating to 1095 K, with subsequent cooling to 313 K followed by reheating to 1273 K. Heating (qh1), cooling (qc1) and reheating (qh2) were performed with 10 K min−1. For calculation of heat capacities cp1 and cp2 of the first and second run, the baseline and a sapphire standard were measured as well. Samples 2 and 3 were subjected to a series of down- and upscans with rates 2.5, 5, 10, 20, 25 and 30 K min−1 with preliminary heating of 10 K min−1 for relaxation (heat release) and setting a desired cooling history. The cooling rate was always matching with following heating rate. Each experiment was conducted with upscans up to 1095 K (approx. 80 K above Tg) followed by cooling down to 723 K (approx. 290 K below Tg). All the upscans, except for the first, were used to determine the fictive temperatures Tf as extrapolated onset temperatures. These were computed graphically by determining the crossover temperature of two tangents aligned to the base and the decreasing flank of the endotherm, respectively. Tf of the as received GBS (first run) was also graphically computed but using the method of matching areas, which is described in detail elsewhere [15, 17]. To fit the heat capacity values in the glassy field best, a Maier-Kelley curve equation cpg = a + bT + cT−2 + dT−0.5 [25] was constructed. To obtain the total heat ΔHex that is released by the GBS, the area covered by cp2 and cp1 curves was calculated according to the left-hand side of Eq. (2).
(1)
with K = 11.3 [16] and K = 11.35 [18]. In order to determine the cooling history from Eq. (1), knowledge of both Tf and the viscositytemperature dependence are required. The fictive temperature of a glass of unknown thermal history can be determined from the difference in the heat capacity curves of a first upscan using the heating rate qh1 and a second upscan at the same heating rate (qh2 = qh1) upon cooling the glass with qc1 = qh2. The second upscan represents the thermally equilibrated glass, whereas the first upscan captures the relaxation (heat release) of the initial glass structure towards the liquid state. A hysteresis is evident between the heating curves of the first and second cycle, if the cooling rate of the as received glass and the cooling rate of the second cycle are different. Tf upon cooling can be graphically computed by matching the areas under the curves for which one has [15, 17]: Teq
c
(cp2 − cp1) dT =
∫T
Tf
g
(cpl − cpg ) dT
(2)
where Tc is the onset temperature at which the release of heat starts, Teq is the temperature at which the hysteresis closes (cp2 = cp1), Tg is the glass transition temperature (i.e. the fictive temperature of the standard cooling rate with qh2 = qc1 = 10 K min−1 [17]) and cpg, cpl are heat capacities of the equilibrated glass and liquid state, respectively. As Tf of a highly quenched glass can exceed Tg of the standard cooling by > 100 K (e.g., hyperquenched basalt glass fiber: ΔT = Tf Tg = 204 K; Tf = 1.22Tg [17]), a direct determination of the viscosity at Tf is complicated due to the proneness of GBS to crystallize in this temperature range. For that reason, viscosity data that capture the glass transition range are calculated by Eq. (1) using fictive temperatures which are received from DSC heat flow measurements at different cooling rates. On the other hand, data that capture the range above liquidus temperature TL, (at T > TL the melt is stable against crystallization) are gained directly by conducting rheometric experiments. Both sets of data are used to compute the viscosity-temperature curve of GBS. Frequently used equations representing the temperature dependence of viscosity of glass-forming liquids are: VFTH [19–21]:
lg η = A1 +
B1 T − C1
3.2. Rheometry Due to the proneness of GBS to crystallize the viscosity of GBS was measured by a Searle-type rotational rheometer (RV20, RC20, TP 1700, Haake, Karlsruhe, Germany) at temperatures above the liquidus. Approx. 68 g of GBS was placed into a PtRh10-crucible and heated up to 1730 K until the GBS was completely molten. Subsequently, a Pt-spindle was inserted into the molten GBS with the lower end of the rod about 2 mm above the bottom of the crucible. The spindle was then rotated and the shear stress as well as the shear rate were recorded at constant temperature of 1730 K. After the measurement the temperature was set subsequently to lower values and after temperature equilibrium the next measurement started until crystallization of the GBS was detected by strongly increasing values of the shear stress. The rheometer was calibrated by the standard glass G1 of the PTB (Physikalisch-Technische Bundesanstalt) [26]. For homogenous melts (crystal-free) the error in viscosity is ± 0.02 lg units [27].
(3)
AG [22]:
lg η = A2 +
(6)
where An, Bn and Cn (n = 1, 2, 3 and 4) are adjustable parameters. It is expected that the predicted viscosity at Tf will vary as the accurateness of the above viscosity models is different [24]. To account for the error that arises from the interpolating viscosities by Eqs. (3)–(6) all four models were used in our data analysis.
2.2. Methods
∫T
B4 C exp ⎛ 4 ⎞ T ⎝T ⎠
B2 T ln(T / C2)
(4) 4. Results
AM [23]:
B C3 lg η = A3 + ⎛ 3 ⎞ ⎝T ⎠
The thermal stability of GBS subjected to a first DSC upscan (heating rate 10 K min−1) is illustrated in Fig. 1. The DSC curve consists of three well-defined signals (glass transition, crystallization, melting) that are characteristic for a glass that easily crystallizes during heating. After the
(5)
and MYEGA [24]: 345
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
Fig. 1. DSC upscan (10 K min−1) of GBS. The onsets of the glass transition, crystallization and melting signals are indicated by the temperatures Tg, Tc and Tm, respectively.
Fig. 3. DSC heat flow of the sequence of upscans for qh = qc = 2.5, 5, 10, 20, 25 and 30 K min−1 of sample 3. The first upscan of the as received GBS is not shown.
endothermic glass transition an exothermic signal with a shoulder on the high temperature flank is evident which is assigned to the crystallization of GBS. The crystallization onset is at approx. 1170 K while the peak is centered at 1196 K. The endothermic signal, assigned to melting reactions, comprises 3 peaks at 1509, 1538 and 1606 K. Melting starts at 1466 K and is completed at 1663 K. The heat capacity curve of GBS of the first upscan cp1 reveals a noticeable hysteresis (different curve progression) due to a strong heat release (ΔHex = 54.95 J g−1) as compared to the second upscan cp2 upon cooling with the standard cooling rate (Fig. 2). The fictive temperature of the as received GBS was calculated by the matching-area method based on Eq. (2) to be 1175 K, whereas Tg determined as extrapolated onset temperature of cp2 was 1014 K. Fig. 3 illustrates the DSC signals of series of the upscans of GBS with defined cooling history (qc = qh). Faster cooling rates result in increased Tf of 1023 K at 30 K min−1, whereas at the lowest cooling rate of 2.5 K min−1Tf was 1005.2 K. The fictive temperature of the three samples was found to
Table 1 Fictive temperature of GBS for cycling with different cooling qc and heating rates qh. Standard deviations σ are from replicate analyses of each sample. qc/qh (K min−1)
Fictive temperature (K) Sample 1
2.5/2.5 5/5 10/10 20/20 25/25 30/30 Unknown/10
1014.0
σ
0.9
Sample 2
σ
Sample 3
σ
1004.7 1009.4 1016.0 1021.8 1022.8 1023.7
0.9 1.8 0.9 1.9 0.8 0.9
1005.2 1009.8 1013.6 1020.8 1021.8 1023.0
0.9 0.5 0.9 0.5 1.9 0.8
1175 ± 1⁎
Key: *calculated by area-matching.
Fig. 2. Comparison of the heat capacity traces of the first and second upscan of sample 1 (A) and total heat release during first upscan (B). The cp1 curve (red solid line) is the heat capacity of the as received GBS (unknown cooling history), whereas cp2 (black solid line) is the heat capacity of GBS glass upon standard cooling (qh2 = −qc1 = 10 K min−1). Dotted and dashed lines are constructed to match the areas of the hysteresis and the glass-liquid transition under standard conditions in accordance with Refs. [15, 17]. The temperature dependence of cpg (blue dashed line) is obtained using a Maier-Kelley curve equation cpg = a + bT + cT−2 + dT-0.5 [25] with a = 1.094, b = 3.5 × 10−5, c = −27,913.682 and d = −1.640 to obtain a best fit of cp2 values in the glassy field. The heat capacity of the liquid cpl (black dashed line) was assumed to be temperature-independent, i.e. a horizontal line upon cp1 = cp2 is constructed. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 346
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
5. Discussion The results show that a direct determination of the viscosity at 1175 K (= Tf of the as received sample) was impeded due to the proneness of GBS to crystallize. Several metrics are proposed for glass stability [28–30] including the most widely used Hrubý parameter TH = (Tc-Tg)/(TmeTc) [31]. Taking the onset temperatures of glass transition, crystallization and melting, which are determined as the intersection of tangents to the curve, traced on the baseline and on the peak side of Fig. 1 we find TH = 0.527, whereas for a good glass former values close to 2 are reported [29, 30]. We note that TH can serve only as a rough measure for the proneness of GBS to crystallize as crystal growth is assumed to proceed from the surface and Tc will depend on the size and surface properties of the particles used in thermal analysis. Further, it should be stressed here that GBS is a product of fast quenching and Tg of the first upscan is affected by thermal history. Therefore, Tg = 1014 K (Table 1) of the second upscan was used for the determination of TH. Besides crystallization issues, standard methods to determine the viscosity of glass-forming liquids at the glass transition range, such as beam bending or micro-sphere indentation [32, 33] were ruled out as the granular GBS requires re-melting to meet the geometric requirements of these experiments. Furthermore, re-melting bears the risk to loose volatiles such as hydrogen, sulfur and carbon containing species that are dissolved in the slag melt under the reducing conditions of the pig iron production in the blast furnace. In particular, traces of dissolved water (as OH-groups) are known to act as a fluxing agent for silicate melts. The effect of reducing viscosity is large at the glass transition range, whereas close to the liquidus temperature it is less pronounced [34]. Further, it has been shown that hydrous melts are relatively stable close to Tg, which allows their repeated cycling through the glass transition range without considerable loss of water [35]. Thus, to bypass a GBS re-melt, the calorimetric approach is used in this study. However, the actual concentration of hydrogen, sulfur and carbon in the GBS glass is not known and their effect on viscosity will be a subject of future studies. In general, different 3-parameter relations can be used to describe the non-Arrhenian viscosity-temperature behavior when the considered temperature range spans the boundary between glass transition and liquidus temperature. To interpolate viscosity between the high and low temperature data Eqs. (3)–(6) are utilized (Fig. 5). Fitting with each
Fig. 4. Viscosity of molten GBS. Vertical dashed line indicates onset temperature of crystallization by visual inspection. The solid line connecting the data points is intended as visual guide. Uncertainty of viscosity of the homogeneous melt is within the symbol size.
vary for each cooling rate only within the error of the measurement, which indicates homogeneous sampling of GBS (Table 1). The low viscosity of the slag melt at high temperatures (> 1730 K) and its crystallization below the liquidus temperature limited the accessible viscosity interval by rotational rheometry to a narrow temperature range (Fig. 4). At T ≤ 1630 K crystallization was observed through a mirror that reflects the image of the melt in the slit between the rotating spindle and the crucible wall. The heterogeneous melt showed sudden, irreproducible breakages of crystal chains and re-agglomeration during spinning accompanied with a strong increase in effective viscosity. Thus, these data were discarded from further evaluation, whereas the three data points of the homogeneous melt at 1730 K (η = 0.40 ± 0.02 Pa s), 1685 K (η = 0.52 ± 0.03 Pa s) and 1638 K (η = 1.00 ± 0.05 Pa s) were used to compute the viscositytemperature dependence, which is described in the following section.
Fig. 5. Viscosity vs. temperature of GBS liquid. Overview (A) and detail close to the fictive temperature (= 1175 K) of the as received GBS (B). Solid and dashed lines are the best fits through the data using Eqs. (3)–(6). The insert of (A) provides details of the dependencies close to the glass transition, where η(Tf)-data was calculated by Eq. (1) using the shift factors K = 11.3 and K = 11.35. 347
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
Table 2 Parameters of Eqs. (3)–(6) which fit the experimental η(T) and η(Tf) data best. Interpolated viscosity at the fictive temperature of the as received GBS and cooling rate during the water-jet quench (Eq. (1)). Viscosity parameters are for η in Pa s and T in K, whereas cooling rates are given in units of K s−1. R2 is the coefficient of determination that measures of how well the regression predictions approximate the real data points. VFTH K A B C R2 lg η(1175 K) lg qc
11.3 −6.11 5804.7 695.8 0.9999 6.01 5.29
AG 11.35 −6.16 5866.1 694.2 0.9999 6.04 5.31
11.3 −4.82 7118.3 670.3 0.9999 5.97 5.33
AM 11.35 −4.86 7205.1 668.4 0.9999 6.01 5.34
11.3 −2.11 1969.8 4.00 0.9999 5.79 5.51
MYEGA 11.35 −2.14 1978.4 3.98 0.9999 5.83 5.52
11.3 −3.45 1067.5 2732.3 0.9999 5.85 5.45
11.35 −3.49 1094.2 2713.4 0.9999 5.89 5.46
glass of the same composition cooled at about 103 K s−1. These reports let us believe that cooling history has to be taken into account in the interpretation of the reactivity of ground GBS in cementitious systems and that its energetic state, expressed by the fictive temperature, should be used to specify these materials in future in addition to glass content and chemical composition.
viscosity model was performed twice to include the small difference in the reported value of the shift factor K (11.3 and 11.35) into the calculations. Inspection of Fig. 5 shows that different viscosity values at 1175 K are received when using Eqs. (3)–(6). The calculated viscosities vary by 0.25 lg units. In particular, the VFTH model (Eq. (6)) results in the highest viscosity (106.04 Pa s), whereas the AM relation (Eq. (5)) led to the lowest viscosity (105.79 Pa s). The quality of the fits is high (coefficient of determination R2 > 99%) and about the same for all of the four viscosity-temperature models (Table 2). Good agreement is also found between the fictive temperature of the standard cooling rate (Tf(10 K min−1) = Tg ≈ 1015 K) and the calculated isokom temperature T12 (temperature at which η = 1012 Pa s ≈ 1016 K) which confirms the established reference point, i.e. T12 = Tg [36, 37]. Further, comparing the extrapolated infinite temperature viscosity, i.e. A (= lgη∞) we find the same order as for the high-temperature limit of the viscosity curves of about 1000 silicate liquids, which show mean values of −3.87 (VFTH), −1.74 (AM) and − 2.93 (MYEGA) [38]. In particular the values −6.11, −6.16 (VFTH), −3.45, −3.49 (MYEGA) and − 2.11, −2.14 (AM) are found to be at the lower boundary of the corresponding distribution. Thus, we believe that deviations in the predicted viscosity are originated by the physical foundation and accurateness of the different viscosity-temperature models. The predicted viscosity at 1175 K was utilized to approximate the cooling rate at which the molten GBS glass was frozen-in during the water-jet quench. Taking both shift factors into account, qc is calculated to be in the range from 105.29–105.52 K s−1 (Table 2). qc of GBS is found to be considerably larger than for an air-quenched alkaline earth aluminosilicate glass of same size. A cooling rate of approx. 101.5 K s−1 was calculated when utilizing data of a commercially drawn flat glass (Corning Jade®: thickness = 700 μm, ΔT = Tf - Tg = 1125 K 1055 K = 70 K [39], lgη∞ = −2.9 (in Pa s) [40] and kinetic fragility mvis = 36.8 [40]). This comparison clearly shows that the use of water as the coolant in combination with the high flow rate of the water-jet has the potential for an improved quenching efficacy. In addition, the total amount of released heat (ΔHex = 54.95 J g−1) together with the large difference between Tf and Tg (ΔT = 160.5 K) qualifies the GBS to be assigned to the class of hyperquenched glasses [17, 41]. It should be stressed here that the GBS processing let to granules of different sizes up to 4 mm. Thus, the fictive temperatures and the cooling rates calculated here are approximates representing mean sizes of samples 1–3. The extreme Tf measured in hyperquenched GBS glass is expected to affect specific volume-dependent glass properties, such as density, enthalpy and refractive index. On the one hand, a dependence of the Na+K+ ion-exchange under hydrothermal conditions on cooling history of aluminosilicate glasses has been reported [42], which can also be used to speeding up the procedures of chemical strengthening by achieving a certain DOL (depth of layer) in a shorter time period [43]. On the other hand, Striepe et al. [44] showed that static fatigue of aluminosilicate glasses and their environmental effects (humidity) decrease with the fictive temperature of the glasses. Moreover, Moesgaard et al. [45] reported the noticeable increase of the reactivity of the aluminosilicate glass that was prepared by about 106 K s−1 cooling if compared to the
6. Conclusions GBS is a by-product of pig iron production in blast furnaces that is manufactured in large quantities by the steel industry using different granulation processes. The GBS of this study was formed by wet-granulation through a water-jet. Under these conditions the glass was hyperquenched. GBS is of high potential energy and shows a fictive temperature that is approx. 160 K higher than Tg of the equilibrated glass. Taking into account the non-Arrhenian dependence of viscosity on temperature and the relationship between quench rate and shear viscosity at the fictive temperature it is calculated that GBS freeze-in at 105.9 ± 0.1 Pa s, which corresponds to a cooling rate of (2.6 ± 0.6) × 105 K s−1. A major topic of further investigations will be to clarify which relevance different cooling rates might have regarding different latent hydraulic reactivities of wet-granulated blast furnace slags. The overall results of this article clearly demonstrate the strong effects of cooling rates on the values of the fictive temperature and the energetic state of the frozen-in glass. Hence, authors are advised to be aware of this dependence and to report the cooling rate and fictive temperature of GBS used in their experiments. Funding The IGF research project 19416 N of the VDEh-Gesellschaft zur Förderung der Eisenforschung mbH was supported by the Bundesministerium für Wirtschaft und Energie based on a decision of the German Bundestag. References [1] F. Schröder, Slags and slag cement, Proceedings 5th Internal Congress Chemistry Cement, Tokyo, vol. IV, 1969, pp. 149–199. [2] H.G. Smolczyk, Slag structure and identification of slags, Proceedings 7th Internal Congress Chemistry Cement, Paris, vol. III, 1980, pp. 1/3–1/17. [3] C. Shi, J. Qian, High performance cementing materials from industrial slags – a review, Resour. Conserv. Recycl. 29 (2000) 195–207. [4] C.J. Shi, Steel slag – its production, processing, characteristics, and cementitious properties, J. Mater. Civ. Eng. 16 (3) (2004) 230–236. [5] A. Ehrenberg, CO2 emissions and energy consumption of granulated blastfurnace slag, Proceedings. 3rd European Slag Conference, 2 Euroslag Publication, Keyworth, 2002, pp. 151–166. [6] R. Rughooputh, J. Rana, Partial replacement of cement by ground granulated blast furnace slag in concrete, J. Emer. Trends Eng. Appl. Sci. 5 (2014) 340–343. [7] D. Suresh, K. Nagaraj, Ground granulated blast slag (GGBS) in concrete – a review, IOSR J. Mech. Civil Eng. 12 (2015) 76–82. [8] S. Samad, A. Shah, Role of binary cement including supplementary cementitious material (SCM) in production of environmentally sustainable concrete: a critical review, Intern. J. Sustain. Built Environ. 6 (2017) 663–674.
348
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
estimate glass-forming ability, J. Non-Cryst. Solids 320 (2003) 1–8. [29] M.L.F. Nascimento, L.A. Souza, E.B. Ferreira, E.D. Zanotto, Can glass stability parameters infer glass forming ability? J. Non-Cryst. Solids 351 (2005) 3296–3308. [30] A.F. Kozmidis-Petrović, Theoretical analysis of relative changes of the Hrubý, Weinberg, and Lu–Liu glass stability parameters with application on some oxide and chalcogenide glasses, Thermochim. Acta 499 (2010) 54–60. [31] A. Hrubý, Evaluation of glass-forming tendency by means of DTA, Czech. J. Phys. B 22 (1972) 1187–1193. [32] R. Brückner, G. Demharter, Systematische Untersuchung über die Anwendbarkeit von Penetrationsviskometern, Glastechn. Ber. 48 (1975) 12–18. [33] H.E. Hagy, Experimental evaluation of beam-bending method of determining glass viscosities in the range 108 to 1015 poises, J. Am. Ceram. Soc. 46 (1963) 93–97. [34] P. del Gaudio, H. Behrens, J. Deubener, Viscosity and glass transition temperature of hydrous float glass, J. Non-Cryst. Solids 353 (2007) (2007) 223–236. [35] S. Reinsch, C. Roessler, U. Bauer, R. Müller, J. Deubener, H. Behrens, Water, the other network modifier in borate glasses, J. Non-Cryst. Solids 432 (2016) 208–217. [36] P.K. Gupta, J.C. Mauro, Composition dependence of glass transition temperature and fragility. I. A topological model incorporating temperature-dependent constraints, J. Chem. Phys. 130 (2009) 094503. [37] Y. Yue, The iso-structural viscosity, configurational entropy and fragility of oxide liquids, J. Non-Cryst. Solids 355 (2009) 737–744. [38] Q. Zheng, J.C. Mauro, A.J. Ellison, M. Potuzak, Y. Yue, Universality of the hightemperature viscosity limit of silicate liquids, Phys. Rev. B 83 (2011) 212202. [39] S. Striepe, M. Potuzak, M.M. Smedskjaer, J. Deubener, Relaxation kinetics of the mechanical properties of an aluminosilicate glass, J. Non-Cryst. Solids 362 (2013) 40–46. [40] X. Guo, M.M. Smedskjaer, J.C. Mauro, Linking equilibrium and nonequilibrium dynamics in glass-forming systems, J. Phys. Chem. B 120 (2016) 3226–3231. [41] C.A. Angell, Y. Yue, L.M. Wang, J.R.D. Copley, S. Borick, S. Mossa, Potential energy, relaxation, vibrational dynamics and the boson peak, of hyperquenched glasses, J. Phys. 15 (2003) S1051–S1068. [42] R. Shiraki, J.T. Iiyama, Na-K ion exchange reaction between rhyolitic glass and (Na, K)Cl aqueous solution under hydrothermal conditions, Geochim. Cosmochim. Acta 54 (1990) 2923–2931. [43] M.J. Dejneka, A.J. Ellison, S. Gomez, Ion exchanged, fast cooled glasses, US Patent (2012) 8,232,218 B2. [44] S. Striepe, J. Deubener, M.M. Smedskjaer, M. Potuzak, Environmental effects on fatigue of alkaline earth aluminosilicate glass with varying fictive temperature, J. Non-Cryst. Solids Solids 379 (2013) 161–168. [45] M. Moesgaard, D. Herfort, M. Steenberg, L.F. Kirkegaard, Y. Yue, Physical performances of blended cements containing calcium aluminosilicate glass powder and limestone, Cem. Concr. Res. 41 (2011) 359–364.
[9] S.C. Pal, A. Mukherjee, S.R. Pathak, Investigation of hydraulic activity of ground granulated blast furnace slag in concrete, Cem. Concr. Res. 33 (2003) 1481–1486. [10] A. Bougara, C. Lynsdale, N.B. Milestone, Reactivity and performance of blastfurnace slags of differing origin, Cem. Concr. Res. 32 (2010) 319–324. [11] Y.L. Qin, X.W. Lv, J. Zhang, J.L. Hao, C.G. Bai, Determination of optimum blast furnace slag cooling rate for slag recycling in cement manufacture, Ironmak. Steelmak. 42 (2015) 395–400. [12] B. Lin, H. Wang, X. Zhu, Q. Liao, B. Ding, Crystallization properties of molten blast furnace slag at different cooling rates, Appl. Therm. Eng. 96 (2016) 432–440. [13] D.J. Min, F. Tsukihashi, Recent advances in understanding physical properties of metallurgical slags, Met. Mater. Int. 23 (2017) 1–19. [14] C.T. Moynihan, A.J. Easteal, J. Wilder, J. Tucker, Dependence of the glass transition temperature on heating and cooling rate, J. Phys. Chem. 78 (1974) 2673–2677. [15] C.T. Moynihan, A.J. Easteal, M.A. DeBolt, J. Tucker, Dependence of the fictive temperature of glass on cooling rate, J. Am. Ceram. Soc. 59 (1976) 12–16. [16] G.W. Scherer, Use of the Adam-Gibbs equation in the analysis of structural relaxation, J. Am. Ceram. Soc. 67 (1984) 504–511. [17] Y.Z. Yue, J.C. Christiansen, S.L. Jensen, Determination of the fictive temperature for a hyperquenched glass, Chem. Phys. Lett. 357 (2002) 20–24. [18] Y. Yue, R. von der Ohe, S.L. Jensen, Fictive temperature, cooling rate, and viscosity of glasses, J. Chem. Phys. 120 (2004) 8053–8059 (Erratum: 121 (2004) 11508.). [19] H. Vogel, Das Temperaturabhängigkeitsgesetz der Viskosität von Flüssigkeiten, Physik. Z. 22 (1921) 645–646. [20] G.S. Fulcher, Analysis of recent measurements of viscosity of glass, J. Amer. Ceram. Soc. 8 (1925) 339–789. [21] G. Tammann, H. Hesse, Die Abhängigkeit der Viskosität von der Temperatur bei unterkühlten Flüssigkeiten, Z. Anorg. Allg. Chem. 156 (1926) 245–257. [22] G. Adam, J.H. Gibbs, On the temperature dependence of cooperative relaxation properties in glass-forming liquids, J. Chem. Phys. 43 (1965) 139–146. [23] I. Avramov, A. Milchev, Effect of disorder on diffusion and viscosity in condensed systems, J. Non-Cryst. Solids 104 (1988) 253–260. [24] J.C. Mauro, Y. Yue, A.J. Ellison, P.K. Gupta, D.C. Allan, Viscosity of glass-forming liquids, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 19780–19784. [25] C.G. Maier, K.K. Kelley, An equation for the representation of high-temperature heat content data, J. Am. Chem. Soc. 54 (1932) 3243–3246. [26] N. Böse, G. Klingenberg, G. Meerlender, Viscosity measurements of glass melts – certification of reference material, Glastechn. Ber. Glas. Sci. Technol. 74 (2001) 115–126. [27] J. Deubener, H. Bornhöft, S. Reinsch, R. Müller, J. Lumeau, L.N. Glebova, L.B. Glebov, Viscosity, relaxation and elastic properties of photo-thermo-refractive glass, J. Non-Cryst. Solids 355 (2009) 126–131. [28] A.A. Cabral Jr., C. Fredericci, E.D. Zanotto, A test of the Hruby parameter to
349
Contents lists available at ScienceDirect
Journal of Non-Crystalline Solids journal homepage: www.elsevier.com/locate/jnoncrysol
Cooling history of a wet-granulated blast furnace slag (GBS) a
N. Pronina , S. Krüger a b
a,1
a
a,⁎
, H. Bornhöft , J. Deubener , A. Ehrenberg
T
b
Institute of Non-Metallic Materials, Clausthal University of Technology, Germany FEhS – Institut für Baustoff-Forschung e.V, Duisburg, Germany
A R T I C LE I N FO
A B S T R A C T
Keywords: Blast furnace slag Cooling rate Calcium aluminosilicate glass Thermal history Fictive temperature Hyperquenched glass
Granulated blast furnace slag (GBS) is a glassy by-product of the steel industry that is formed during quenching of the molten slag after the blast furnace in a water-jet. To elucidate the cooling history of GBS, calorimetric scanning in the glass transition range and viscometric experiments at temperatures above the liquidus were performed. The GBS studied was of industrial origin with a d50 = 700 μm and > 99% glassy. It is found that GBS glass is of high potential energy, showing a fictive temperature that is approx. 160 K higher than the glass transition temperature under standard cooling conditions. Using different viscosity models and the relationship between quench rate and shear viscosity at the fictive temperature it is calculated that GBS was formed at a cooling rate of approx. 2.6 × 105 K s−1 corresponding to a shear viscosity of 105.9 Pa s. Due to the high Tf and high amount of heat that is released during structural relaxation, GBS is assigned to the class of hyperquenched glasses.
1. Introduction Granulated blast furnace slag (GBS) is a non-metallic by-product of the steel industry (annual GBS production is approx. 280 million tons) that is formed when after leaving the blast furnace the molten slag is quenched using powerful water jets or fast rotating drums in air. GBS is a fine, granular, almost fully glassy material of calcium aluminosilicate composition. GBS material exhibits latent hydraulic (cementitious) properties [1–4] and furthermore as cement constituent or concrete addition it serves advantageous technical, environmental and economic benefits [5–8]. The reactivity of ground GBS (so-called GGBS) in cement is found to be influenced by the slag properties such as the glass content and its chemical composition as well as the mineralogical composition of residual crystal fraction [9, 10]. As the residual crystalline material is unreactive, small crystal fractions can be beneficial for blending cement. Thus, temperature-time-transformation (TTT) and continuouscooling-transformation (CCT) diagrams have been determined using the hot thermocouple method in which the tips of a thermocouple are welded in a U-type shape to heat and measure the temperature simultaneously [11, 12]. On the one hand TTT and CCT curves show that GBS glass is formed when the cooling rate exceeds 100–101 K s−1. On the other hand, the cooling rates of industrially manufactured GBS are generally unknown and only estimated to be above 102 K s−1 [11]. However, variation in operation parameter, such as molten slag
temperature, water jet flow and disc rotation speed may cause the formation of GBS glass with different cooling history and therefore different structure and properties [13]. As a consequence of these fluctuations, the performance of GBS from different production campaigns and sites becomes less predictable. Generally, the cooling history of glasses can be estimated through their fictive temperature Tf. Despite the fact that the dependence of Tf on cooling rate has been established for > 40 years [14–17], it has not been applied to GBS. The present work aims therefore to express the cooling history of GBS by Tf and cooling rate. The former is determined by calorimetric scanning, whereas the latter requires information on the viscosity-temperature curve. To provide viscosity data at temperatures above the liquidus, rotational viscometry of the molten granules was utilized. The implementation of Tf and cooling rate as parameters of the GBS production might allow assessment of samples independently of their composition and simplify the control and modification of the quenching process according to desired properties (e.g. reactivity in cementitious systems). 2. Material and methods 2.1. Material
⁎
Corresponding author. E-mail address: [email protected] (J. Deubener). 1 Present address: SCHOTT AG, Mainz, Germany. https://doi.org/10.1016/j.jnoncrysol.2018.07.054 Received 1 June 2018; Received in revised form 20 July 2018; Accepted 23 July 2018 Available online 27 July 2018 0022-3093/ © 2018 Elsevier B.V. All rights reserved.
The wet-granulated blast furnace slag (GBS) under investigation was
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
of industrial origin (moisture 10.6 wt%) and of chemical composition 0.03 P2O5, 33.4 SiO2, 1.98 TiO2, 11.4 Al2O3, 44.9 CaO, 5.7 MgO, 0.75 FeO, 0.23 MnO, 0.36 K2O and 0.17 Na2O (wt%) as analyzed by X-ray fluorescence (Σ = 98.92%; 100-Σ = 1.08% volatile species). Transmission optical microscopy of the immersed granules revealed a glassy volume fraction > 99%. Granulometry (sieve analysis) revealed a monomodal particle size distribution with a mean size d50 = 700 μm (not shown). The GBS glass was analyzed as received without any processing prior to analysis.
lg η = A 4 +
3. Experimental 3.1. Calorimetry
Using logarithmic scales, the cooling rate qc and the shear viscosity at the fictive temperature η(Tf) are related by a dimensionless shift factor K in the form
lg qc = K − lg η (Tf )
To compensate for possible inhomogeneity, three samples of the same lot were used for calorimetry. The first sample was measured in a DSC (404F3 Pegasus; Netzsch, Selb, Germany), whereas for samples 2 and 3 a TGA equipped with DSC sensor (TGA/DSC 3+, Mettler Toledo, Greifensee, Switzerland) was used. Each measurement was carried out with 20–30 mg GBS granules in a PtRh-crucible in a nitrogen environment. Sample 1 was subjected to the sequence of heating to 1095 K, with subsequent cooling to 313 K followed by reheating to 1273 K. Heating (qh1), cooling (qc1) and reheating (qh2) were performed with 10 K min−1. For calculation of heat capacities cp1 and cp2 of the first and second run, the baseline and a sapphire standard were measured as well. Samples 2 and 3 were subjected to a series of down- and upscans with rates 2.5, 5, 10, 20, 25 and 30 K min−1 with preliminary heating of 10 K min−1 for relaxation (heat release) and setting a desired cooling history. The cooling rate was always matching with following heating rate. Each experiment was conducted with upscans up to 1095 K (approx. 80 K above Tg) followed by cooling down to 723 K (approx. 290 K below Tg). All the upscans, except for the first, were used to determine the fictive temperatures Tf as extrapolated onset temperatures. These were computed graphically by determining the crossover temperature of two tangents aligned to the base and the decreasing flank of the endotherm, respectively. Tf of the as received GBS (first run) was also graphically computed but using the method of matching areas, which is described in detail elsewhere [15, 17]. To fit the heat capacity values in the glassy field best, a Maier-Kelley curve equation cpg = a + bT + cT−2 + dT−0.5 [25] was constructed. To obtain the total heat ΔHex that is released by the GBS, the area covered by cp2 and cp1 curves was calculated according to the left-hand side of Eq. (2).
(1)
with K = 11.3 [16] and K = 11.35 [18]. In order to determine the cooling history from Eq. (1), knowledge of both Tf and the viscositytemperature dependence are required. The fictive temperature of a glass of unknown thermal history can be determined from the difference in the heat capacity curves of a first upscan using the heating rate qh1 and a second upscan at the same heating rate (qh2 = qh1) upon cooling the glass with qc1 = qh2. The second upscan represents the thermally equilibrated glass, whereas the first upscan captures the relaxation (heat release) of the initial glass structure towards the liquid state. A hysteresis is evident between the heating curves of the first and second cycle, if the cooling rate of the as received glass and the cooling rate of the second cycle are different. Tf upon cooling can be graphically computed by matching the areas under the curves for which one has [15, 17]: Teq
c
(cp2 − cp1) dT =
∫T
Tf
g
(cpl − cpg ) dT
(2)
where Tc is the onset temperature at which the release of heat starts, Teq is the temperature at which the hysteresis closes (cp2 = cp1), Tg is the glass transition temperature (i.e. the fictive temperature of the standard cooling rate with qh2 = qc1 = 10 K min−1 [17]) and cpg, cpl are heat capacities of the equilibrated glass and liquid state, respectively. As Tf of a highly quenched glass can exceed Tg of the standard cooling by > 100 K (e.g., hyperquenched basalt glass fiber: ΔT = Tf Tg = 204 K; Tf = 1.22Tg [17]), a direct determination of the viscosity at Tf is complicated due to the proneness of GBS to crystallize in this temperature range. For that reason, viscosity data that capture the glass transition range are calculated by Eq. (1) using fictive temperatures which are received from DSC heat flow measurements at different cooling rates. On the other hand, data that capture the range above liquidus temperature TL, (at T > TL the melt is stable against crystallization) are gained directly by conducting rheometric experiments. Both sets of data are used to compute the viscosity-temperature curve of GBS. Frequently used equations representing the temperature dependence of viscosity of glass-forming liquids are: VFTH [19–21]:
lg η = A1 +
B1 T − C1
3.2. Rheometry Due to the proneness of GBS to crystallize the viscosity of GBS was measured by a Searle-type rotational rheometer (RV20, RC20, TP 1700, Haake, Karlsruhe, Germany) at temperatures above the liquidus. Approx. 68 g of GBS was placed into a PtRh10-crucible and heated up to 1730 K until the GBS was completely molten. Subsequently, a Pt-spindle was inserted into the molten GBS with the lower end of the rod about 2 mm above the bottom of the crucible. The spindle was then rotated and the shear stress as well as the shear rate were recorded at constant temperature of 1730 K. After the measurement the temperature was set subsequently to lower values and after temperature equilibrium the next measurement started until crystallization of the GBS was detected by strongly increasing values of the shear stress. The rheometer was calibrated by the standard glass G1 of the PTB (Physikalisch-Technische Bundesanstalt) [26]. For homogenous melts (crystal-free) the error in viscosity is ± 0.02 lg units [27].
(3)
AG [22]:
lg η = A2 +
(6)
where An, Bn and Cn (n = 1, 2, 3 and 4) are adjustable parameters. It is expected that the predicted viscosity at Tf will vary as the accurateness of the above viscosity models is different [24]. To account for the error that arises from the interpolating viscosities by Eqs. (3)–(6) all four models were used in our data analysis.
2.2. Methods
∫T
B4 C exp ⎛ 4 ⎞ T ⎝T ⎠
B2 T ln(T / C2)
(4) 4. Results
AM [23]:
B C3 lg η = A3 + ⎛ 3 ⎞ ⎝T ⎠
The thermal stability of GBS subjected to a first DSC upscan (heating rate 10 K min−1) is illustrated in Fig. 1. The DSC curve consists of three well-defined signals (glass transition, crystallization, melting) that are characteristic for a glass that easily crystallizes during heating. After the
(5)
and MYEGA [24]: 345
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
Fig. 1. DSC upscan (10 K min−1) of GBS. The onsets of the glass transition, crystallization and melting signals are indicated by the temperatures Tg, Tc and Tm, respectively.
Fig. 3. DSC heat flow of the sequence of upscans for qh = qc = 2.5, 5, 10, 20, 25 and 30 K min−1 of sample 3. The first upscan of the as received GBS is not shown.
endothermic glass transition an exothermic signal with a shoulder on the high temperature flank is evident which is assigned to the crystallization of GBS. The crystallization onset is at approx. 1170 K while the peak is centered at 1196 K. The endothermic signal, assigned to melting reactions, comprises 3 peaks at 1509, 1538 and 1606 K. Melting starts at 1466 K and is completed at 1663 K. The heat capacity curve of GBS of the first upscan cp1 reveals a noticeable hysteresis (different curve progression) due to a strong heat release (ΔHex = 54.95 J g−1) as compared to the second upscan cp2 upon cooling with the standard cooling rate (Fig. 2). The fictive temperature of the as received GBS was calculated by the matching-area method based on Eq. (2) to be 1175 K, whereas Tg determined as extrapolated onset temperature of cp2 was 1014 K. Fig. 3 illustrates the DSC signals of series of the upscans of GBS with defined cooling history (qc = qh). Faster cooling rates result in increased Tf of 1023 K at 30 K min−1, whereas at the lowest cooling rate of 2.5 K min−1Tf was 1005.2 K. The fictive temperature of the three samples was found to
Table 1 Fictive temperature of GBS for cycling with different cooling qc and heating rates qh. Standard deviations σ are from replicate analyses of each sample. qc/qh (K min−1)
Fictive temperature (K) Sample 1
2.5/2.5 5/5 10/10 20/20 25/25 30/30 Unknown/10
1014.0
σ
0.9
Sample 2
σ
Sample 3
σ
1004.7 1009.4 1016.0 1021.8 1022.8 1023.7
0.9 1.8 0.9 1.9 0.8 0.9
1005.2 1009.8 1013.6 1020.8 1021.8 1023.0
0.9 0.5 0.9 0.5 1.9 0.8
1175 ± 1⁎
Key: *calculated by area-matching.
Fig. 2. Comparison of the heat capacity traces of the first and second upscan of sample 1 (A) and total heat release during first upscan (B). The cp1 curve (red solid line) is the heat capacity of the as received GBS (unknown cooling history), whereas cp2 (black solid line) is the heat capacity of GBS glass upon standard cooling (qh2 = −qc1 = 10 K min−1). Dotted and dashed lines are constructed to match the areas of the hysteresis and the glass-liquid transition under standard conditions in accordance with Refs. [15, 17]. The temperature dependence of cpg (blue dashed line) is obtained using a Maier-Kelley curve equation cpg = a + bT + cT−2 + dT-0.5 [25] with a = 1.094, b = 3.5 × 10−5, c = −27,913.682 and d = −1.640 to obtain a best fit of cp2 values in the glassy field. The heat capacity of the liquid cpl (black dashed line) was assumed to be temperature-independent, i.e. a horizontal line upon cp1 = cp2 is constructed. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) 346
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
5. Discussion The results show that a direct determination of the viscosity at 1175 K (= Tf of the as received sample) was impeded due to the proneness of GBS to crystallize. Several metrics are proposed for glass stability [28–30] including the most widely used Hrubý parameter TH = (Tc-Tg)/(TmeTc) [31]. Taking the onset temperatures of glass transition, crystallization and melting, which are determined as the intersection of tangents to the curve, traced on the baseline and on the peak side of Fig. 1 we find TH = 0.527, whereas for a good glass former values close to 2 are reported [29, 30]. We note that TH can serve only as a rough measure for the proneness of GBS to crystallize as crystal growth is assumed to proceed from the surface and Tc will depend on the size and surface properties of the particles used in thermal analysis. Further, it should be stressed here that GBS is a product of fast quenching and Tg of the first upscan is affected by thermal history. Therefore, Tg = 1014 K (Table 1) of the second upscan was used for the determination of TH. Besides crystallization issues, standard methods to determine the viscosity of glass-forming liquids at the glass transition range, such as beam bending or micro-sphere indentation [32, 33] were ruled out as the granular GBS requires re-melting to meet the geometric requirements of these experiments. Furthermore, re-melting bears the risk to loose volatiles such as hydrogen, sulfur and carbon containing species that are dissolved in the slag melt under the reducing conditions of the pig iron production in the blast furnace. In particular, traces of dissolved water (as OH-groups) are known to act as a fluxing agent for silicate melts. The effect of reducing viscosity is large at the glass transition range, whereas close to the liquidus temperature it is less pronounced [34]. Further, it has been shown that hydrous melts are relatively stable close to Tg, which allows their repeated cycling through the glass transition range without considerable loss of water [35]. Thus, to bypass a GBS re-melt, the calorimetric approach is used in this study. However, the actual concentration of hydrogen, sulfur and carbon in the GBS glass is not known and their effect on viscosity will be a subject of future studies. In general, different 3-parameter relations can be used to describe the non-Arrhenian viscosity-temperature behavior when the considered temperature range spans the boundary between glass transition and liquidus temperature. To interpolate viscosity between the high and low temperature data Eqs. (3)–(6) are utilized (Fig. 5). Fitting with each
Fig. 4. Viscosity of molten GBS. Vertical dashed line indicates onset temperature of crystallization by visual inspection. The solid line connecting the data points is intended as visual guide. Uncertainty of viscosity of the homogeneous melt is within the symbol size.
vary for each cooling rate only within the error of the measurement, which indicates homogeneous sampling of GBS (Table 1). The low viscosity of the slag melt at high temperatures (> 1730 K) and its crystallization below the liquidus temperature limited the accessible viscosity interval by rotational rheometry to a narrow temperature range (Fig. 4). At T ≤ 1630 K crystallization was observed through a mirror that reflects the image of the melt in the slit between the rotating spindle and the crucible wall. The heterogeneous melt showed sudden, irreproducible breakages of crystal chains and re-agglomeration during spinning accompanied with a strong increase in effective viscosity. Thus, these data were discarded from further evaluation, whereas the three data points of the homogeneous melt at 1730 K (η = 0.40 ± 0.02 Pa s), 1685 K (η = 0.52 ± 0.03 Pa s) and 1638 K (η = 1.00 ± 0.05 Pa s) were used to compute the viscositytemperature dependence, which is described in the following section.
Fig. 5. Viscosity vs. temperature of GBS liquid. Overview (A) and detail close to the fictive temperature (= 1175 K) of the as received GBS (B). Solid and dashed lines are the best fits through the data using Eqs. (3)–(6). The insert of (A) provides details of the dependencies close to the glass transition, where η(Tf)-data was calculated by Eq. (1) using the shift factors K = 11.3 and K = 11.35. 347
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
Table 2 Parameters of Eqs. (3)–(6) which fit the experimental η(T) and η(Tf) data best. Interpolated viscosity at the fictive temperature of the as received GBS and cooling rate during the water-jet quench (Eq. (1)). Viscosity parameters are for η in Pa s and T in K, whereas cooling rates are given in units of K s−1. R2 is the coefficient of determination that measures of how well the regression predictions approximate the real data points. VFTH K A B C R2 lg η(1175 K) lg qc
11.3 −6.11 5804.7 695.8 0.9999 6.01 5.29
AG 11.35 −6.16 5866.1 694.2 0.9999 6.04 5.31
11.3 −4.82 7118.3 670.3 0.9999 5.97 5.33
AM 11.35 −4.86 7205.1 668.4 0.9999 6.01 5.34
11.3 −2.11 1969.8 4.00 0.9999 5.79 5.51
MYEGA 11.35 −2.14 1978.4 3.98 0.9999 5.83 5.52
11.3 −3.45 1067.5 2732.3 0.9999 5.85 5.45
11.35 −3.49 1094.2 2713.4 0.9999 5.89 5.46
glass of the same composition cooled at about 103 K s−1. These reports let us believe that cooling history has to be taken into account in the interpretation of the reactivity of ground GBS in cementitious systems and that its energetic state, expressed by the fictive temperature, should be used to specify these materials in future in addition to glass content and chemical composition.
viscosity model was performed twice to include the small difference in the reported value of the shift factor K (11.3 and 11.35) into the calculations. Inspection of Fig. 5 shows that different viscosity values at 1175 K are received when using Eqs. (3)–(6). The calculated viscosities vary by 0.25 lg units. In particular, the VFTH model (Eq. (6)) results in the highest viscosity (106.04 Pa s), whereas the AM relation (Eq. (5)) led to the lowest viscosity (105.79 Pa s). The quality of the fits is high (coefficient of determination R2 > 99%) and about the same for all of the four viscosity-temperature models (Table 2). Good agreement is also found between the fictive temperature of the standard cooling rate (Tf(10 K min−1) = Tg ≈ 1015 K) and the calculated isokom temperature T12 (temperature at which η = 1012 Pa s ≈ 1016 K) which confirms the established reference point, i.e. T12 = Tg [36, 37]. Further, comparing the extrapolated infinite temperature viscosity, i.e. A (= lgη∞) we find the same order as for the high-temperature limit of the viscosity curves of about 1000 silicate liquids, which show mean values of −3.87 (VFTH), −1.74 (AM) and − 2.93 (MYEGA) [38]. In particular the values −6.11, −6.16 (VFTH), −3.45, −3.49 (MYEGA) and − 2.11, −2.14 (AM) are found to be at the lower boundary of the corresponding distribution. Thus, we believe that deviations in the predicted viscosity are originated by the physical foundation and accurateness of the different viscosity-temperature models. The predicted viscosity at 1175 K was utilized to approximate the cooling rate at which the molten GBS glass was frozen-in during the water-jet quench. Taking both shift factors into account, qc is calculated to be in the range from 105.29–105.52 K s−1 (Table 2). qc of GBS is found to be considerably larger than for an air-quenched alkaline earth aluminosilicate glass of same size. A cooling rate of approx. 101.5 K s−1 was calculated when utilizing data of a commercially drawn flat glass (Corning Jade®: thickness = 700 μm, ΔT = Tf - Tg = 1125 K 1055 K = 70 K [39], lgη∞ = −2.9 (in Pa s) [40] and kinetic fragility mvis = 36.8 [40]). This comparison clearly shows that the use of water as the coolant in combination with the high flow rate of the water-jet has the potential for an improved quenching efficacy. In addition, the total amount of released heat (ΔHex = 54.95 J g−1) together with the large difference between Tf and Tg (ΔT = 160.5 K) qualifies the GBS to be assigned to the class of hyperquenched glasses [17, 41]. It should be stressed here that the GBS processing let to granules of different sizes up to 4 mm. Thus, the fictive temperatures and the cooling rates calculated here are approximates representing mean sizes of samples 1–3. The extreme Tf measured in hyperquenched GBS glass is expected to affect specific volume-dependent glass properties, such as density, enthalpy and refractive index. On the one hand, a dependence of the Na+K+ ion-exchange under hydrothermal conditions on cooling history of aluminosilicate glasses has been reported [42], which can also be used to speeding up the procedures of chemical strengthening by achieving a certain DOL (depth of layer) in a shorter time period [43]. On the other hand, Striepe et al. [44] showed that static fatigue of aluminosilicate glasses and their environmental effects (humidity) decrease with the fictive temperature of the glasses. Moreover, Moesgaard et al. [45] reported the noticeable increase of the reactivity of the aluminosilicate glass that was prepared by about 106 K s−1 cooling if compared to the
6. Conclusions GBS is a by-product of pig iron production in blast furnaces that is manufactured in large quantities by the steel industry using different granulation processes. The GBS of this study was formed by wet-granulation through a water-jet. Under these conditions the glass was hyperquenched. GBS is of high potential energy and shows a fictive temperature that is approx. 160 K higher than Tg of the equilibrated glass. Taking into account the non-Arrhenian dependence of viscosity on temperature and the relationship between quench rate and shear viscosity at the fictive temperature it is calculated that GBS freeze-in at 105.9 ± 0.1 Pa s, which corresponds to a cooling rate of (2.6 ± 0.6) × 105 K s−1. A major topic of further investigations will be to clarify which relevance different cooling rates might have regarding different latent hydraulic reactivities of wet-granulated blast furnace slags. The overall results of this article clearly demonstrate the strong effects of cooling rates on the values of the fictive temperature and the energetic state of the frozen-in glass. Hence, authors are advised to be aware of this dependence and to report the cooling rate and fictive temperature of GBS used in their experiments. Funding The IGF research project 19416 N of the VDEh-Gesellschaft zur Förderung der Eisenforschung mbH was supported by the Bundesministerium für Wirtschaft und Energie based on a decision of the German Bundestag. References [1] F. Schröder, Slags and slag cement, Proceedings 5th Internal Congress Chemistry Cement, Tokyo, vol. IV, 1969, pp. 149–199. [2] H.G. Smolczyk, Slag structure and identification of slags, Proceedings 7th Internal Congress Chemistry Cement, Paris, vol. III, 1980, pp. 1/3–1/17. [3] C. Shi, J. Qian, High performance cementing materials from industrial slags – a review, Resour. Conserv. Recycl. 29 (2000) 195–207. [4] C.J. Shi, Steel slag – its production, processing, characteristics, and cementitious properties, J. Mater. Civ. Eng. 16 (3) (2004) 230–236. [5] A. Ehrenberg, CO2 emissions and energy consumption of granulated blastfurnace slag, Proceedings. 3rd European Slag Conference, 2 Euroslag Publication, Keyworth, 2002, pp. 151–166. [6] R. Rughooputh, J. Rana, Partial replacement of cement by ground granulated blast furnace slag in concrete, J. Emer. Trends Eng. Appl. Sci. 5 (2014) 340–343. [7] D. Suresh, K. Nagaraj, Ground granulated blast slag (GGBS) in concrete – a review, IOSR J. Mech. Civil Eng. 12 (2015) 76–82. [8] S. Samad, A. Shah, Role of binary cement including supplementary cementitious material (SCM) in production of environmentally sustainable concrete: a critical review, Intern. J. Sustain. Built Environ. 6 (2017) 663–674.
348
Journal of Non-Crystalline Solids 499 (2018) 344–349
N. Pronina et al.
estimate glass-forming ability, J. Non-Cryst. Solids 320 (2003) 1–8. [29] M.L.F. Nascimento, L.A. Souza, E.B. Ferreira, E.D. Zanotto, Can glass stability parameters infer glass forming ability? J. Non-Cryst. Solids 351 (2005) 3296–3308. [30] A.F. Kozmidis-Petrović, Theoretical analysis of relative changes of the Hrubý, Weinberg, and Lu–Liu glass stability parameters with application on some oxide and chalcogenide glasses, Thermochim. Acta 499 (2010) 54–60. [31] A. Hrubý, Evaluation of glass-forming tendency by means of DTA, Czech. J. Phys. B 22 (1972) 1187–1193. [32] R. Brückner, G. Demharter, Systematische Untersuchung über die Anwendbarkeit von Penetrationsviskometern, Glastechn. Ber. 48 (1975) 12–18. [33] H.E. Hagy, Experimental evaluation of beam-bending method of determining glass viscosities in the range 108 to 1015 poises, J. Am. Ceram. Soc. 46 (1963) 93–97. [34] P. del Gaudio, H. Behrens, J. Deubener, Viscosity and glass transition temperature of hydrous float glass, J. Non-Cryst. Solids 353 (2007) (2007) 223–236. [35] S. Reinsch, C. Roessler, U. Bauer, R. Müller, J. Deubener, H. Behrens, Water, the other network modifier in borate glasses, J. Non-Cryst. Solids 432 (2016) 208–217. [36] P.K. Gupta, J.C. Mauro, Composition dependence of glass transition temperature and fragility. I. A topological model incorporating temperature-dependent constraints, J. Chem. Phys. 130 (2009) 094503. [37] Y. Yue, The iso-structural viscosity, configurational entropy and fragility of oxide liquids, J. Non-Cryst. Solids 355 (2009) 737–744. [38] Q. Zheng, J.C. Mauro, A.J. Ellison, M. Potuzak, Y. Yue, Universality of the hightemperature viscosity limit of silicate liquids, Phys. Rev. B 83 (2011) 212202. [39] S. Striepe, M. Potuzak, M.M. Smedskjaer, J. Deubener, Relaxation kinetics of the mechanical properties of an aluminosilicate glass, J. Non-Cryst. Solids 362 (2013) 40–46. [40] X. Guo, M.M. Smedskjaer, J.C. Mauro, Linking equilibrium and nonequilibrium dynamics in glass-forming systems, J. Phys. Chem. B 120 (2016) 3226–3231. [41] C.A. Angell, Y. Yue, L.M. Wang, J.R.D. Copley, S. Borick, S. Mossa, Potential energy, relaxation, vibrational dynamics and the boson peak, of hyperquenched glasses, J. Phys. 15 (2003) S1051–S1068. [42] R. Shiraki, J.T. Iiyama, Na-K ion exchange reaction between rhyolitic glass and (Na, K)Cl aqueous solution under hydrothermal conditions, Geochim. Cosmochim. Acta 54 (1990) 2923–2931. [43] M.J. Dejneka, A.J. Ellison, S. Gomez, Ion exchanged, fast cooled glasses, US Patent (2012) 8,232,218 B2. [44] S. Striepe, J. Deubener, M.M. Smedskjaer, M. Potuzak, Environmental effects on fatigue of alkaline earth aluminosilicate glass with varying fictive temperature, J. Non-Cryst. Solids Solids 379 (2013) 161–168. [45] M. Moesgaard, D. Herfort, M. Steenberg, L.F. Kirkegaard, Y. Yue, Physical performances of blended cements containing calcium aluminosilicate glass powder and limestone, Cem. Concr. Res. 41 (2011) 359–364.
[9] S.C. Pal, A. Mukherjee, S.R. Pathak, Investigation of hydraulic activity of ground granulated blast furnace slag in concrete, Cem. Concr. Res. 33 (2003) 1481–1486. [10] A. Bougara, C. Lynsdale, N.B. Milestone, Reactivity and performance of blastfurnace slags of differing origin, Cem. Concr. Res. 32 (2010) 319–324. [11] Y.L. Qin, X.W. Lv, J. Zhang, J.L. Hao, C.G. Bai, Determination of optimum blast furnace slag cooling rate for slag recycling in cement manufacture, Ironmak. Steelmak. 42 (2015) 395–400. [12] B. Lin, H. Wang, X. Zhu, Q. Liao, B. Ding, Crystallization properties of molten blast furnace slag at different cooling rates, Appl. Therm. Eng. 96 (2016) 432–440. [13] D.J. Min, F. Tsukihashi, Recent advances in understanding physical properties of metallurgical slags, Met. Mater. Int. 23 (2017) 1–19. [14] C.T. Moynihan, A.J. Easteal, J. Wilder, J. Tucker, Dependence of the glass transition temperature on heating and cooling rate, J. Phys. Chem. 78 (1974) 2673–2677. [15] C.T. Moynihan, A.J. Easteal, M.A. DeBolt, J. Tucker, Dependence of the fictive temperature of glass on cooling rate, J. Am. Ceram. Soc. 59 (1976) 12–16. [16] G.W. Scherer, Use of the Adam-Gibbs equation in the analysis of structural relaxation, J. Am. Ceram. Soc. 67 (1984) 504–511. [17] Y.Z. Yue, J.C. Christiansen, S.L. Jensen, Determination of the fictive temperature for a hyperquenched glass, Chem. Phys. Lett. 357 (2002) 20–24. [18] Y. Yue, R. von der Ohe, S.L. Jensen, Fictive temperature, cooling rate, and viscosity of glasses, J. Chem. Phys. 120 (2004) 8053–8059 (Erratum: 121 (2004) 11508.). [19] H. Vogel, Das Temperaturabhängigkeitsgesetz der Viskosität von Flüssigkeiten, Physik. Z. 22 (1921) 645–646. [20] G.S. Fulcher, Analysis of recent measurements of viscosity of glass, J. Amer. Ceram. Soc. 8 (1925) 339–789. [21] G. Tammann, H. Hesse, Die Abhängigkeit der Viskosität von der Temperatur bei unterkühlten Flüssigkeiten, Z. Anorg. Allg. Chem. 156 (1926) 245–257. [22] G. Adam, J.H. Gibbs, On the temperature dependence of cooperative relaxation properties in glass-forming liquids, J. Chem. Phys. 43 (1965) 139–146. [23] I. Avramov, A. Milchev, Effect of disorder on diffusion and viscosity in condensed systems, J. Non-Cryst. Solids 104 (1988) 253–260. [24] J.C. Mauro, Y. Yue, A.J. Ellison, P.K. Gupta, D.C. Allan, Viscosity of glass-forming liquids, Proc. Natl. Acad. Sci. U. S. A. 106 (2009) 19780–19784. [25] C.G. Maier, K.K. Kelley, An equation for the representation of high-temperature heat content data, J. Am. Chem. Soc. 54 (1932) 3243–3246. [26] N. Böse, G. Klingenberg, G. Meerlender, Viscosity measurements of glass melts – certification of reference material, Glastechn. Ber. Glas. Sci. Technol. 74 (2001) 115–126. [27] J. Deubener, H. Bornhöft, S. Reinsch, R. Müller, J. Lumeau, L.N. Glebova, L.B. Glebov, Viscosity, relaxation and elastic properties of photo-thermo-refractive glass, J. Non-Cryst. Solids 355 (2009) 126–131. [28] A.A. Cabral Jr., C. Fredericci, E.D. Zanotto, A test of the Hruby parameter to
349