Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement

+Model JECS-9919; No. of Pages 10

ARTICLE IN PRESS Available online at www.sciencedirect.com

ScienceDirect Journal of the European Ceramic Society xxx (2014) xxx–xxx

Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement Mohamed-Ali Maaroufi a , André Lecomte a,∗ , Cécile Diliberto a , Olivier Francy b , Pierre Le Brun c a

Institute Jean Lamour, UMR/CNRS 7198, Lorraine University, Team “Materials for Civil Engineering”, IUTNB, CS 90137, 54600 Villers-les-Nancy, France b Saint-Gobain CREE, 550 Avenue Alphonse Jauffret, BP 20224, 84306 Cavaillon, France c Constellium Technology Center, 725 rue Aristide Bergès, CS 10027, 38341 Voreppe, France Received 2 September 2014; received in revised form 20 November 2014; accepted 21 November 2014

Abstract Refractory castables based on aluminous hydraulic binders are commonly used in aluminium casthouses (furnaces, ducts, etc.). Their selection is based on their good mechanical strength, thermal behavior and compatibility with molten aluminium. However, few studies focus on their hydrous evolution in operation, whereas this property can also have an influence on the produced metal quality. In this article, the internal moisture of twelve hardened cement pastes fired at high temperature, made with four aluminous hydraulic binders and three different Water/Binder ratios was registered under diverse thermo-hydrous conditions, including at high temperature. The water trapped by physisorption and chemisorption can be significant for some products, and it strongly depends on the mineralogy and porosity of the hardened cement paste. The more the binders contain alumina phase, the more the hardened cement pastes mobilize and render moisture. © 2014 Elsevier Ltd. All rights reserved. Keywords: Calcium aluminate cement (CAC); Moisture; Physisorption; Chemisorption; Water vapor adsorption isotherms (WVAIs)

1. Introduction

1.2. Description of the study

1.1. Context

Tests on refractory castables and their components taken separately11 show that the hardened cement paste (HCP) is the main seat of water transfer while the refractory aggregates are quasi-inert in this respect. For this reason the experiments are focused on the cementitious matrix. Several porous structures were studied by varying the amount of added water (water/binder ratio W/B). Here, hydraulic binders are CAC, that also contain various mineral and organic additions in order to increase the paste compactness, to facilitate shaping, to control the curing time or to limit the reactions with the surrounding molten metal (refractory corrosion and inclusion entrainment in the melt).12 Hydration, setting, curing and mechanical behavior of the HCP have been extensively studied.13–17 However, few studies focus on their hygrothermal behavior at different stages of their use while the form of water present in the porous structure of the HCP can evolve (free, adsorbed or chemically bound18–20 ):

Refractory castables based on hydraulic binders or calcium aluminate cement (CAC) are widely used in aluminium casthouses.1 Their selection is based on properties such as corrosion resistance to molten metal, mechanical behavior and insulation.2–5 They also exhibit faster development of the mechanical strength and do not release lime during hydration.6,7 However, these castables contain residual moisture which can affect the safety of the process or even the final quality of the metal produced. This moisture results from water exchange with the surrounding atmosphere, which is controlled by the porosity, the permeability, and the chemical and mineralogical composition of the materials.8,9 The aim of this paper is to present a study of the exchanges in CAC materials as function of both the thermo-hydrous conditions and the temperature.10 ∗

Corresponding author. Tel.: +33 0 383 682 575; fax: +33 0 383 682 532. E-mail address: [email protected] (A. Lecomte).

(1) Free water is condensed liquid or steam in large capillaries.

http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029 0955-2219/© 2014 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10 2

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

Table 1 Density and chemical composition of binders. CAC (g/cm3 )

% weight

Density Spectral particle size (␮m) Median-diameter (␮m) Specific surface (m2 /g) Al2 O3 SiO2 CaO Others (inert) Loss of ignition

(2) Physically adsorbed water covers the inner surface of the pores in several molecular layers which are weakly bound ˚ thick) and have some mobility. Physical adsorp(5 to 10 A tion mainly occurs at low temperature and is a reversible phenomenon. (3) Chemically bound water forms hydrates that are not involved in the transport of moisture. These hydrates can only decompose above their threshold stability temperature21 .

B1

B2

3.198 0.3–250 6 8.7 81.1 <0.01 19.0 0.2 0.3

3.038 0.2–250 4.3 9.3 46.7 23.5 24.1 3.9 2.4

B3

B4

2.771 0.3–250 7 7.4 49.1 14.2 29.3 6.35 2

2.76 0.4–250 9 8.4 32.1 26.6 35.8 4.2 2

delivers controlled humidity. The observed behavior is attributed to the cement hydration reactions, at high temperatures. Finally, since the hygrothermal equilibrium may require a long time before it is reached, an analytical method has been developed to extrapolate measurements and forecast the equilibrium water content. Similarly, WVAIs were modeled. 2. Preliminary investigations on studied materials 2.1. Characterization of binders

The usual approach to study the hydric behavior of porous materials is to determine their water vapor adsorption isotherms (WVAIs) from moisture pickup curves obtained in various thermodynamic conditions. These measurements also provide the moisture pickup kinetics, and the products storage capacity, depending on the test conditions (sample size, etc.). These data may be provided by several methods.22,23 In the current study, the gravimetric method was preferred.24,25 While most tests were made at 20 ◦ C at different relative humidity RH, temperatures as high as 300 ◦ C were also considered. For low temperature tests (20 ◦ C), after shaping, setting and hardening, samples were subjected to curing at high temperature (750 ◦ C) typical for the industrial operation temperature in the aluminium industry. This treatment transforms the products in a totally anhydrous state. They were then weighed hot to determine their dry mass (reference). After cooling, they were immersed and saturated in water to determine their porosity by hydrostatic weighing. Finally, they were dried at 105 ◦ C until a constant weight was reached to eliminate physical water,21 and probably also a part of chemical water (see Rel. (5)). The moisture pick up test consisted to record the samples mass increase over time in a controlled climatic chamber at 20 ◦ C and at different successive RH (drying at 105 ◦ C between each RH level). The equilibrium water content reached for each condition enables to build WVAIs at 20 ◦ C26,27 (adsorption phase). For high temperature tests (T > 100 ◦ C), only the chemical adsorption can theoretically take place.21 It is related to the high-temperature rehydration of some cementitious phases in presence of water vapor, and in relation to the temperature changes which occur within the castable during its use (for example between casting cycles). Tests were performed with a thermogravimetric analysis chamber (Setaram Setsys Evolution TGA) coupled to a steam generator (Setaram Wetsys) which

Four binders based on calcium aluminate cement or CAC and various admixtures were used in this study. They were referenced B1, B2, B3 and B4. These binders are commonly used to manufacture various kinds of refractory castables based respectively on aggregates of tabular alumina, bauxite, fired clay and silica, respectively. The density of the binders was measured with a DKD certified pycnometer (non-reactive liquid). The values are given in Table 1. Spectral particle size, median diameter and specific surface area are also given in Table 1. Their chemical composition, expressed as weight percentage of oxides (X-ray fluorescence, Bruker S4 Explorer AXS apparatus), is given in Table 1. B1 is essentially constituted of alumina. B2 and B3 contain about 50% alumina and variable silica amounts. B4 contains less alumina, but more silica than B2 and B3. These binders also contain other inert elements in variable proportions. The binder mineralogy was measured by XRD analysis (X˚ and Pert Pro diffractometer, copper anticathode K␣ = 1.54506 A) ® EVA-plus software with the JCP2 database. Rietveld quantification was made with Topas® . Results are shown in Table 2.

Table 2 Quantification of aluminous anhydrous cement phases in the four binders (in mass percentage). CAC

B1

B2

B3

B4

A CA CA2 C12 A7 Others (inert)

44 40 9 7 –

30 14 9 0 47

29 22 11 0 38

0 30 20 0 50

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

3

Table 3 W/B (water/binder) ratios required for complete hydration and theoretical hydration rate for different W/B ratios. Binder

B1

B2

B3

B4

Stoichiometric W/B Theoretical hydration rate

0.49 W/B = 0.15 W/B = 0.25 W/B = 0.35

0.26 0.3 0.5 0.7

0.31 0.6 1 1 (1.3)

0.23 0.5 0.8 1 (1.1)

Results are in accordance with the literature,28–31 the main constitutive hydraulic anhydrous phases being alumina A1 , calcium aluminate CA, di-calcium aluminate CA2 , and mayenite (dodecacalcium hepta-aluminate) C12 A7 . This last phase is considered as the most reactive in a CAC. In this case: (1) B1 contains mainly A and CA, as well as CA2 and C12 A7 , with no other constituent. (2) B2 and B3 contain A, CA and CA2 , but no C12 A7. Several other inert crystalline phases are also present. (3) B4 contains only CA and CA2 , and more than 50% of other crystalline phases. It may be noted the absence of silica-containing crystallized phases in B2, B3 and B4, although the chemical analysis indicates a significant amount (Table 1). Silica is actually present as amorphous silica fume, supposed to facilitate the hydration of aluminous cement32 and to improve compactness. 2.2. Hydration mechanisms and evolution of hydrates at high temperature In the presence of water at ambient temperature, CAC initially forms metastable hydrates CAH10 and C2 AH8 , that transform on temperature increase into gibbsite AH3 and stable tricalcium aluminate hexahydrate or hydrogarnet C3 AH6 . The alumina A can also form an amorphous compound AHx . If hydration occurs above 35 ◦ C, the metastable phases hardly appear and stable phases are directly obtained. In such conditions, the hydration reactions are: 3CA + 12H → C3 AH6 + 2AH3

(1)

3CA2 + 21H → C3 AH6 + 5AH3

(2)

C12 A7 + 33H → 4C3 AH6 + 3AH3

(3)

A + 3H → AH3

(4)

When the cured products undergo a significant temperature increase, for example on initial heat-up, sintering process or contact with liquid metal, these hydrates release all or a part of their physically and chemically bound water, and restore the original anhydrous compounds by reverse reactions. Then, the binder phase evolves into a pseudo-clinker of CAC, which porosity depends on the initial W/B ratio. 1

In cement chemist notation, C stands for CaO, A for Al2 O3 and H for H2 O.

0.6 1 (1.1) 1 (1.5)

Dehydration reactions are listed below, with the associated decomposition temperature ranges13 . Anhydrous phases obtained are successively alumina A, lime C, mayenite C12 A7 and calcium aluminate CA. The water amount released by each hydrate was also calculated (weight percentage of water versus of the initial weight of hydrate, %H). Alumina is the phase that releases more water. Temperature range (◦ C)

Dehydratation reaction

%H

Rel.

60–120 110–170

AHx → A + xH 3CAH10 → C3 AH6 + 2AH3 + 18H 3C2 AH8 → 2C3 AH6 + AH3 + 9H AH3 → A + 3H 7C3 AH6 → C12 A7 H + 9CH + 32H CH → C + H C12 A7 H → C12 A7 + H C12 A7 + 5A → 12CA

18x/(102 + 18x) 32

(5) (6)

15.1

(7)

34.6 21.8

(8) (9)

170–240 210–300 240–370 ≈450 ≈750 ≈900

24.3 1.3

(10) (11)

When cooled down to ambient temperature, the anhydrous compounds may at least partially rehydrate in the presence of water. For example, C3 AH6 may appear when the temperature drops below 280 ◦ C19 . It can be noted however that severe and repeated thermal cycles lead to a progressive enrichment of the paste with gibbsite/alumina, and a consequent decreased hydrogarnet (C3 AH6 ) content. Indeed, high temperature decomposition of hydrogarnet, leads to C12 A7 and CA, which rehydrate into hydrogarnet and gibbsite at low temperature. Thus the amount of gibbsite/alumina gradually increases on successive hydration–dehydration cycles. According to XRD results (Table 2) and Eqs. (1)–(4), the stoichiometric amount of water (W) necessary for complete hydration can be estimated. The corresponding values, expressed as weight percentage of the total weight of binder (B), including inert phases, i.e. W/B ratio, are given in Table 3.

2.3. Characterization of hardened cement pastes 2.3.1. Definition and preparation of samples For each of the four binders, three hardened cement pastes or HCP with W/B ratios resp. 0.15, 0.25 and 0.35 were prepared. They are respectively identified HCP B1, HCP B2, HCP B3 and HCP B4, eventually followed by the corresponding W/B ratio. These ratios are similar to those frequently used in industrial

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10 4

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx A CA2

80

100

% of cemenous phases

% of cemenous phases

100

CA C12A7

60

HCP B1

40 20

CA Others

HCP B2

60 40 20

0

0

0,15

0,25 W/B

0,35

100

0,15 100

A CA2

80

CA Others

60

HCP B3

40 20

% of cemenous phases

% of cemenous phases

A CA2

80

0,25 W/B A CA2

80

0,35

CA Others

HCP B4

60 40 20 0

0 0,15

0,25 W/B

0,35

0,15

0,25 W/B

0,35

Fig. 1. Amount of aluminous cement phases after heating up in the four binders according to W/B ratios.

refractory castables based on CAC. They are supposed to lead to different hydration rates (see Table 3) and different porosities. After mixing according to the standard protocol NF EN 1963, pastes were cured in sealed cylindrical molds measuring 4 cm in diameter and 20 cm in height. During curing, molds were slowly rotating to prevent bleeding and segregation. Samples were demolded after 3 days. They were then cut into three parts: two 8-cm-high cylinders for water exchange tests and a central 4-cm-high cylinder for the characterization measurements. Specimens were then subjected to a five days heat treatment up to 750 ◦ C, similarly to in field operations. They were then hot weighed to determine their reference dry mass Ms. 2.3.2. Mineralogy Fig. 1 compares the contents of aluminous anhydrous phases detected in the various HCP (XRD + Topas® ), after the initial firing at 750 ◦ C and subsequent cooling down to ambient temperature in a desiccator. Non-aluminous and/or non-crystalline phases are listed as “Other”: (1) For the four HCPs, the alumina A content is higher than that of the initial binders (Table 2) because the hydration of other aluminous phases led to some gibbsite (Rel. (1)–(3)) which transforms into alumina during the firing (Rel. (8)), confirming the assumption made in the preceding paragraph. The higher the W/B ratio, the higher the alumina content. This remains true for a W/B ratio greater than the stoichiometric value, showing that an excess of water improves hydration. (2) Calcium aluminate CA is still present in the four HCP, although this phase should have disappeared if the binders were fully hydrated. It cannot come from the decomposition of the hydrogarnet because the firing temperature is not high

enough to achieve this transformation, which occurs around 900 ◦ C (Rel. (12)). Thus its presence probably results from an incomplete hydration. The CA content tends to decrease with increasing W/B ratio, and it is more consumed for low alumina content (HCP B4 for example). (3) The amount of di-calcium aluminate CA2 remains almost constant, especially if W/B is low, thus showing poor reactivity with water. (4) C12A7 (mayenite) is detected only in HCP B1, in small quantities. (5) Surprisingly, lime could be detected in no HCP. It could be combined with alumina to give CA, or pozzolanic silica fumes (although this reaction requires the presence of water and generally occurs over long period of time).

These results show that the hydration of the four binders and their constituting phases is not complete, even when W/B is larger than the stoichiometric ratio, since CA and CA2 remain after heating up. A and CA are the most reacting phases, especially at high W/B ratio. HCP B4 not containing alumina, only CA is mostly hydrated. This hardened paste had mechanical properties smaller than other HCP both before and after heating up (measurements not reported here). After firing, a small amount of alumina appears, following the (first) hydrogarnet and gibbsite decomposition. In summary, the CAC hydration rate and the amount/nature of the obtained phases before and after heating up vary from one product to another. The influencing parameters are the initial binder composition, the W/B ratio, the presence of secondary constituents (silica fume . . .). The short curing time (3 days) probably played a significant role in limiting the hydration rate of the less reactive phases like CA2 .

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

5

Table 4 Porosity (%) of the HCP versus W/B ratios, before and after heating up. HCP B1

W/B Before firing After firing Difference

0.15 27.3 33.9 +6.6

HCP B2 0.25 25.4 35.1 +9.7

0.35 33.7 50.4 +16.7

0.15 12.9 30.5 +17.6

HCP B3 0.25 30.6 39.7 +9.1

0.35 43.8 48.8 +5

HCP B4

0.15 16.7 29.4 +12.7

2.3.3. Porosity The porosity of the HCP measured by hydrostatic weighing, before and after heating up at 750 ◦ C is given Table 4. For fired HCP, the immersion into water has certainly led to a partial rehydration of the “pseudo-clinker”, which probably decreased the porosity. But the porosity after firing is generally higher than the porosity after curing, which shows the effect of the constitutive water release and the sintering (see Table 4). Logically enough, the porosity increases steadily with the W/B ratio. Slight variations can be related to

0.25 27.2 38.5 +11.3

2.3.4. Residual water content These measurements were used to define a reference “dry” mass Md, obtained after heating up at 750 ◦ C, cooling down and immersion during 24 h (for hydrostatic weighing), followed by drying to constant weight at 105 ◦ C in order to remove physically bound water (a small part of physical water remains still linked to this temperature). The corresponding values are given in Table 5. There is a good correlation between the residual water content and the binder mineralogy (Table 2): the higher the alumina content and the W/B, the higher the residual water content (12.8% for HCP B1 W/B 0.35). Finally, thermal analysis TGA (Fig. 2) was performed on the four HCP with W/B = 0.25, on powders obtained by crushing above samples. The aim is to assess the amount of chemical water from hydrates, because the XRD analysis on these HCP has proved unable to properly quantify the hydrated phases. Water contents are expressed as weight percentage of the final dried product weight. The temperature was increased up to Table 5 Chemically bound water in HCPs after heating up at 750 ◦ C; immersion and drying at 105 ◦ C. Binder

HCP B1

HCP B2

HCP B3

HCP B4

W/B

0.15 0.25 0.35

5.3 12.6 12.8

2.5 1.6 1.8

2.1 2.7 3.1

0.1 0.1 0.1

0.15 13.8 21.1 +7.3

0.25 25.3 37.2 +11.9

0.35 37.6 45.3 +7.7

Temperature (°C) 0

100 200 300 400 500 600 700 800 900

Water loss (% of Md)

0

10

-5

0

-10 -10

-15 -20

-20 HCP B1 HCP B3

-25

(1) the nature and the content of aluminous and inert phases (Table 2), (2) their hydration degree with respect to the W/B ratio versus stoichiometric requirements (Table 3), (3) the presence of entrained air, which depends on the workability. The stiff pastes prepared with low W/B ratio (e.g. HCP B1 W/B 0.15) contain more air than fluid pastes with high W/B ratio.

0.35 39.6 50.3 +10.7

HCP B2 HCP B4

Heat flow (μV)

HCP

-30

Fig. 2. TGA on the HCP W/B = 0.25 after heating up, rewetting and drying at 105 ◦ C.

900 ◦ C at a 10 ◦ C/min rate. It is worth noting here that the hydrate decomposition temperature is influenced by the conditions of its formation and the thermal gradient applied during the test. As a result, we observed decomposition ranges rather than well-defined temperatures, with overlapping in some cases (see Table 3). This is especially the case for the gibbsite and the hydrogarnet, which cannot be clearly distinguished by thermal analysis. Table 6 shows the released water amounts obtained for different temperature ranges and dehydrated phases. The amount of released water directly depends on the alumina content and on the importance of the inert fraction. Thus HCP B1 (W/B 0.25) is the hardened paste that releases the more water (25% of its dry weight!), and HCP B4 the one showing the lowest water release. However, we find here higher values than those previously obtained on solid samples (Table 5). The differences can be partly explained by the different final temperatures between initial heating up (750 ◦ C) and TGA (900 ◦ C), and probably also by the mass effect between blocks and powder. . .. In conclusion, these tests were used to confirm that the CACbased HCPs have the ability to release significant amounts of chemically bound water amounts above 100 ◦ C, leading to anhydrous phases like A, CA, CA2 , which are likely to recover all Table 6 Water (%) released through decomposition of hydrates under TGA for the four HCP W/B 0.25 rewetted and then dried at 105 ◦ C. Hydrate

C2 AH8 AH3 et C3 AH6 C12 A7 H Total

Range

170–240 210–370 750

W/B 0.25 HCP B1

HCP B2

HCP B3

HCP B4

2.2 16.8 6 25

0.2 2.4 0.8 3.4

0.7 3.7 0.7 5.1

0.2 1.4 0.3 1.9

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10 6

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

3. Moisture pick up experiments The moisture pick up tests were performed on the HCPs in various thermo-hydrous conditions, below or above 100 ◦ C. They consisted in recording the weight evolution of the samples as function of time, for different temperature-humidity conditions kept constant.

12

Water content (%)

or a part of this water on further cooling down, as shown by the following tests.

HCP B3

10

95%

8

80% 60% 40%

6 4

Experimental Model

2 0 5000

0

10000

15000

Time (mn)

3.1. Tests at temperatures lower than 100 ◦ C

Fig. 3. Moisture pick up on HCP B3 W/C 0.35 at 20 ◦ C in climatic chamber for different RH levels. Taking into account the chemical water content at 105 ◦ C.

X − Xo = (Xeq − Xo )

t t + at

(13)

where X0 , X and Xeq are resp. water content at the beginning of the test, at time t and at equilibrium, a and Xeq are the model fitting parameters. 3.1.2. Results and discussion The curves obtained for HCP B3 W/B 0.35 at different RH levels are presented on Fig. 3. Dots are associated to experimental measurements and the continuous curves to the model. We can emphasize that the initial water content (chemical) for this HCP is approximately equal to 3.1%, and that it can pick up a similar additional amount of water, as soon as RH exceeds 80%. Fig. 4 compares the moisture pick up of the four HCP W/B 0.35, at 80%RH. Highly significant differences are noticeable,

not only in terms of chemical water initially present, but also of water adsorbed over time (physically + chemically bound). The equilibrium water contents Xeq obtained for the different RH allow establishing the WVAIs for each HCP. Results are shown on Fig. 5 for three W/B ratios. A model of “GAB”15 was used to fit the experimental data. It is written: Xeq − Xo =

χ × HR + β × HR + γ

(14)

α × HR2

with χ, α, β and γ fitting parameters. The equilibrium water content Xeq depends on the W/B ratio (i.e. porosity). The higher W/B, the higher ability to adsorb moisture. Slight difference for HCP B1 W/B 0.15, which has a higher porosity than expected. It can be assumed that the paste has a “more open porous structure”. More generally, it is always HCP B1 which adsorbs the most moisture and HCP B4 the least. The highest total water content at equilibrium is reached for HCP B1 W/B 0.35 at 95%RH, with about 28%! Obviously, in a refractory castable, the water content is reduced in proportion to the binder fraction in the castable. However, the porosity of the various HCPs does not vary significantly for a given W/B ratio (e.g., from 45 to 50% for W/B 0.35, see Table 4). We can see here that the porous structure controls the physisorption, more than the total porosity volume itself. Pore size distributions obtained by mercury intrusion (Thermo Scientific Pascal 240 device) on sintered refractory castables prepared with these four binders and similar W/B ratios, also show that the pore size distributions are quite 20

Water content (%)

3.1.1. Experimental procedure A first set of experiments was conducted in a climatic chamber at 20 ◦ C and at four relative humidity levels of 40, 60, 80 and 95%, respectively. The test specimens were the same having undergone heating up at 750 ◦ C, followed by immersion into water and drying at 105 ◦ C. Before testing, they were brought to ambient temperature in a desiccator to prevent moisture pick up from the ambient atmosphere. The desiccator was placed in the chamber to reach the test temperature; the samples were then exposed to the humid atmosphere. The weight was either continuously recorded with a servo system, or manually measured at regular intervals. Measurements lasted between 4 and 9 days, depending on the conditions. Results were expressed as water weight gain X, relatively to the dry mass Md measured on fired anhydrous materials. Thus, the residual water content at the beginning of the test (X0 ) corresponds to the chemical water involved in the rehydrated phases after immersion of specimens, followed by drying at 105 ◦ C (Table 4). A simple hyperbolic parametric model was selected to extrapolate the equilibrium water content Xeq from the early water content evolution, which allows the construction of the WVAIs in a limited period of time. This model has been validated on a variety of measurements recorded continuously for several weeks, until effective mass stabilization.11 It shows that specific measurements made over a shorter period are sufficient to obtain good accuracy of the equilibrium water content Xeq . This model is written as:

80 %RH

15

HCP B1 Experimental Model

10

HCP B3 HCP B2

5

HCP B4

0 0

10000 5000 Time (mn)

15000

Fig. 4. Moisture pick up at 80%RH on the four HCP W/B 0.35. Climatic chamber at 20 ◦ C. Taking into account of the chemical water content at 105 ◦ C.

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx 30

30

W/B = 0.15

25

W/B = 0.25

25

Experimental Model

15

Experimental Model

20

HCP B1

Xeq (%)

Xeq (%)

20

HCP B1

15

10

10

HCP B2 HCP B3 HCP B4

5 0 0

20

40

60

80

HCP B3 HCP B2

5

HCP B4

0

100

0

20

Relave Humidity (%RH) 30

7

40

60

80

100

Relave Humidity (%RH)

W/B = 0.35

HCP B1

Experimental

20

Model

Xeq (%)

25

15

HCP B2

10

HCP B3

5

HCP B4

0 0

20

40

60

80

100

Relave Humidity (%RH) Fig. 5. WVAIs at 20 ◦ C for the four binders with the three W/B ratios.

different, especially between Refractory B1 and Refractory B4 (Fig. 6). Refractory B1 has more than 90% of its porosity in pore sizes below 0.5 microns, when Refractory B4 has less than 50%. B2 and B3 have similar and intermediate distributions. For 95%RH, the mesopores have the critical size to be fully saturated by capillary condensation, as expressed by the Kelvin–Laplace law.15 Here we find an explanation for the significant difference in equilibrium water content that appears between the four HCP above 80%RH. Chemisorption also explains the differences between WVAIs, particularly at low RH, since moisture pick up is well correlated to the amount of alumina and calcium aluminate of the HCP (Table 2). HCP dried at 105 ◦ C and then exposed to a humid

Cumulave Intrusion (%)

100

Refractory B1 B2 B3 B4

80 60 40 20 0 100

10

1

0,1

0,01

0,001

Pore size Diameter (μm) Fig. 6. Pore size distribution of refractory castables based on the different CAC.

environment at 20 ◦ C presumably undergo some rehydration of destabilized phases, such CAH10 and AHx .18,19 Therefore, the CAC-based HCPs regain chemical water and physical water at ambient temperature, when the products are pre-dried (at 105 ◦ C or more), the more their alumina and calcium aluminate contents are high and their pore structure is fine. 3.2. Tests at temperatures higher than 100 ◦ C 3.2.1. Description of experiments A second set of experiments was performed with a TGA unit that can inject steam flow into the cell at controlled humidity. Each test consisted at first in heating the sample (100 mg) at 10 ◦ C/min up to 350 ◦ C, so that the temperature would be higher than that of the main signal detected by the preliminary TGA analysis (Fig. 2). At this temperature, hydrates lose most of their water content. A 6-h soak has been observed to ensure an anhydrous state. The reference dry mass Md considered for the calculation of the water content was then recorded. The temperature was then lowered successively to 300, 200 and 100 ◦ C. At each temperature, steam was injected into the cell at four increasing RH levels, each level being applied during 4 h. The generated RH levels were respectively 20, 40, 60 and 80% at 23 ◦ C. Between two temperature soaks, the temperature was raised up to 350 ◦ C for 6 h, to recover the initial dry state. It must be noted that the actual humidity seen by the sample is not equal to that provided by the steam generator at 23 ◦ C, because the temperature in the TGA cell is much higher. Based

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10 8

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

Table 7 Order of magnitude of the “real” relative humidity seen by the samples, depemidity seen by the samples, depending on the TGA test temperature. Temperature (◦ C)

23

100

200

300

Corresponding relative humidity (%)

20 40 60 80

0.55 1.10 1.66 2.21

0.035 0.070 0.105 0.140

0.006 0.013 0.020 0.027

Fig. 8. Thermo-hydrous conditions for moisture pick up tests by TGA.

Fig. 7. Moisture pick up on TGA. Example of HCP B1 W/B 0.35 at 100 ◦ C.

on the Duperray formula (Rel. (15)), we can estimate the saturating vapor pressure in the 90 to 300 ◦ C range, and the actual relative humidity in the TGA cell. Table 7 gives the corresponding values. The range of “actual” RH covered by these tests is not sufficient to establish relevant WVAIs. However, it provides some interesting data on the hydrous behavior of refractory castables at temperature above 100 ◦ C, exposed to a humid ambient atmosphere, which corresponds to the real industrial context.   θ 4 Psat = (15) 100 with θ in ◦ C and Psat in atm. 3.2.2. Results and discussion Fig. 7 shows an example of the moisture pick-up curves obtained for HCP B1 with W/B 0.35 at 100 ◦ C. It appears in these conditions that even a low relative humidity causes a rapid and significant chemical adsorption, which then gradually stabilizes. A higher humidity level causes additional adsorption, with similar kinetics. On the opposite, the temperature increase causes an

(t − t0 ) a + (t − t0 )

HCP B1 HCP B3

0,3

HR% 0.03

0,4 HCP B2 HCP B4

0,2

Xeq

0,4

(16)

where X is the water content at time t, Xeq is the equilibrium water content, t0 is the starting time, a, Xeq and t0 are fitting parameters. The water content evolution over time for the four HCPs W/B 0.25 is given in Fig. 8 (continuous curves). The dotted curve represents the “real” relative humidity sequentially applied on the three imposed temperature soaks. Table 8 shows the equilibrium water content reached for each thermo-hydrous conditions. Values are of course much lower than for moisture pick-up at ambient temperature, due to the low humidity level actually applied. However, they cannot be neglected. They reflect the rehydration of aluminous phases at high temperatures, which occurs when the temperature drops below the hydrate decomposition threshold. The amount of chemically involved water then strongly depends on the nature and the amount of reactive phases present in the HCP, since HCP B1 exhibits, once again, more chemical water than HCP B4 (see Fig. 9a). However, we did not see the influence of the porosity of the product, as no systematic increase of the water content with the W/B ratio has

200 °C

0,5

Xeq

X = Xeq

0,5

0,6

HCP B1 HCP B3

0,3

HCP B2 HCP B4

W/B 0,35

0,2 0,1

0,1

0

0 0

a

almost complete chemical desorption, fairly well synchronized with the temperature evolution. As before, the measures have been fitted with a hyperbolic model to extrapolate the equilibrium water content Xeq :

0,05

0,1 HR %

50

0,15

b

150

250

350

T °C

Fig. 9. Example of the evolution of moisture pick up (chemical water) for the four HCP (a) at 200 ◦ C and various RH, (b) versus temperature at RH 0.03%.

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model

ARTICLE IN PRESS

JECS-9919; No. of Pages 10

M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

9

Table 8 Equilibrium water contents extrapolated for each HCP, temperature and humidity. T (◦ C)

100

W/B

%RH

0.55

1.10

1.66

2.21

0.035

0.070

0.105

0.140

0.006

0.013

0.020

0.027

0.15

HCP B1 HCP B2 HCP B3 HCP B4

0.61 0.06 0.17 0.01

0.76 0.08 0.21 0.01

0.84 0.09 0.23 0.02

0.94 0.11 0.26 0.02

0.537 0.038 0.131 0.005

0.594 0.043 0.158 0.008

0.686 0.049 0.164 0.014

0.738 0.054 0.169 0.018

0.42 0.039 0.144 0.002

0.5 0.042 0.159 0.009

0.541 0.046 0.165 0.013

0.58 0.047 0.169 0.018

0.25

HCP B1 HCP B2 HCP B3 HCP B4

0.46 0.13 0.09 0.04

0.56 0.17 0.11 0.06

0.61 0.19 0.12 0.07

0.67 0.21 0.14 0.08

0.3 0.096 0.058 0.035

0.325 0.103 0.06 0.04

0.37 0.116 0.066 0.047

0.398 0.126 0.073 0.052

0.248 0.108 0.05 0.031

0.277 0.112 0.055 0.038

0.291 0.114 0.059 0.042

0.299 0.114 0.061 0.047

0.35

HCP B1 HCP B2 HCP B3 HCP B4

0.53 0.25 0.23 0.06

0.65 0.31 0.29 0.09

0.72 0.34 0.32 0.10

0.79 0.38 0.36 0.12

0.396 0.118 0.135 0.038

0.438 0.148 0.171 0.045

0.492 0.17 0.197 0.054

0.522 0.188 0.211 0.06

0.361 0.098 0.16 0.028

0.395 0.113 0.18 0.035

0.411 0.123 0.183 0.041

0.418 0.125 0.184 0.044

200

been observed (see Table 8). The very small size of the samples may be a part of the explanation. Since the actual humidity levels vary from one temperature to another, it is difficult to plot the evolution of the water content versus the temperature for a given RH. However, the evolution of the equilibrium water content with the injected humidity follows a rather linear trend, at least on each investigated range, for every HCP (Fig. 9a). We can thus try and extrapolate the obtained values to estimate the water contents at equilibrium for a given RH., Fig. 9b shows the example for RH = 0.03%. We observe that for 200 and 300 ◦ C, there is no significant difference in moisture pick up, indicating only the gibbsite and the hydrogarnet form at this stage (see Table 6). At 100 ◦ C, the weight increases, indicating the AHx formation. Rehydration of these phases is only partial, due to the low humidity, and is related to the initial mineralogy, primarily the amount alumina. Below 100 ◦ C, these phases develop more significantly, as observed in the climatic chamber experiments. 4. Conclusions The quantification of the moisture present in refractory materials based on hydraulic aluminous binder, in various conditions, is a new approach in their characterization. Dehydration and moisture pick up experiments were carried out on hardened cement pastes, considered as the main seat of water exchanges. They showed that these materials can pick up varying amounts of water, by physisorption and chemisorption at ambient temperature, but also by chemisorption at higher temperature. These tests performed in complementary conditions (TGA/DTA, climatic chamber, etc.) lead to the same conclusion: the water exchanges are controlled, not only by the porosity and the pore structure, but also, to a large extent, by their mineralogy. The higher the alumina and calcium aluminate contents, the more water involved during operating cycles (cooling/heating). This is essentially due to the formation of gibbsite and hydrogarnet which dehydrate into alumina and Mayennite, and so on. The binders containing

300

lower amounts of these phases, but alternatively more CA2 or inert compounds, are less prone to water pickup, at least in early operations, since their alumina content will progressively increase upon thermal cycling. However, they usually exhibit lower mechanical properties. The key remains to find the best compromise for each application in aluminium casthouses. Acknowledgment The authors thank the French National Research Agency (NRA10 ) for their financial support. References 1. Aliprandi G. Matériaux réfractaires et céramiques techniques: éléments de céramurgie et de technologie. Paris: Septima; 1989. 2. Wynn A, Coppack J, Steele T, Moody K, Caspersen L. Monolithic material selection for the lining of aluminum holding & melting furnaces. TMS Suppl Proc 2010;3:133–43. 3. Carden Z, Brewster A. Monolithic refractory furnace linings designed for rapid commissioning. TMS Light Met 2008;3:587–92. 4. Richter T, Vezza T, Allaire C, Afshar S. Castable with the improved corrosion resistance against aluminum. Aachen, Germany: Eurogress; 1998. p. 86–90. 5. Quesnel S, Allaire C, Afshar S. Corrosion of refractories at the bellyband of aluminum holding and melting furnaces description. In: Rigaud M, Allaire C, editors. Advances in refractories for the metallurgical industries II, CIM proceedings. 1996. p. 321–7. 6. Duriez M, Arrambide J. Nouveau traité de matériaux de construction—1. Paris: Dunod; 1961. 7. Osborn EF, Muan A. Al2 O3 –CaO–SiO2 system revised and redrawn. In: Hart LD, editor. Phase equilibrium diagrams of oxides system. Plate 1. Westerville: The American Ceramic Society; 1960. 8. Allaire C. Les matériaux réfractaires pour le confinement de l’Aluminium liquide, Symposium sur la coulée de l’aluminium. Revue éditée par le Centre Québécois de Recherche et de Développement de l’Aluminium; 1999. 9. Allaire C. La science des matériaux réfractaires appliquée au secteur de la transformation primaire de l’aluminium. Centre Québécois de Recherche et de Développement de l’Aluminium; 2003. 10. Research National Agency RNA. Matériaux Fonctionnels et Procédés Innovants (MATETPRO): projet, PRINCIPIA., (PRocédés INdustriels de Coulée Innovants Pour l’Industrie, Aéronautique). Research National Agency RNA; 2010. RMNP-007-01.

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029

+Model JECS-9919; No. of Pages 10 10

ARTICLE IN PRESS M.-A. Maaroufi et al. / Journal of the European Ceramic Society xxx (2014) xxx–xxx

11. Maaroufi MA. Etude du comportement hygrothermique de matériaux céramiques réfractaires à liants alumineux pour le confinement de l’aluminium liquide, University of Lorraine, 2014, Nancy (in progress). 12. Oliveira IR. Castable matrix, additives and their role on hydraulic binder hydration. Ceram Int 2009;35:1453–60. 13. Cardoso FA, Innocentini MDM, Akiyoshi MM, Pandolfelli VC. Effect of curing time on the properties of CAC bonded refractory castables. J Eur Ceram Soc 2004;24:2073–8. 14. Das SK. Crystal morphology of calcium aluminates hydrated for 14 days. J Mater Sci Lett 1997;16:735–6. 15. Das SK. Thermal analysis of hydrated calcium aluminates. J Therm Anal 1996;7:765–74. 16. Ukrainczyk N. Thermal properties of hydrating calcium aluminate cement pastes. Cem Concr Res 2010;40:128–36. 17. Talebian N. The hydration products of a refractory calcium aluminate cement at intermediate temperatures. Iran J Chem Eng 2007;6:3. 18. Baroghel-Bouny V. Water vapor sorption experiments on hardened cementitious materials. Cem Concr Res 2007;37(3):414–37. 19. Samiri D. Analyse physique et caractérisation hygrothermique des matériaux de construction: approche expérimentale et modélisation numérique. Lyon: INSA; 2008 (Ph.D. thesis). 20. Drouet E. Impact de la température sur la carbonatation des matériaux cimentaires–prise en compte des transferts hydriques. Cachan: ENS; 2010 (Ph.D. thesis). 21. Baroghel-Bouny V. Caractérisation microstructurale et hydrique des pâtes de ciments et des bétons ordinaires et à très hautes performances. LCPC; 1994 (Ph.D. thesis).

22. Sing KSW, Everett DH, Haul RAW, Moscou L, Pierotti RA, Rouquerol J, Siemieniewska T. Reporting physisorption data for gas/solid systems with special reference to the determination of surface area and porosity. Pure Appl Chem 1985;3:18 (IUPAC). 23. Tada S, Watanabe K. Dynamic determination of sorption isotherm of cement based materials. Cem Concr Res 2005;35(12):2271–7. 24. Iglesias H. Handbook of food isotherms: water sorption parameters for food and food components. New York: Academic Press Inc.; 1982. 25. Lowell S, Shields JE. Powder surface area and porosity. 3rd ed. London: Chapman & Hall Ltd; 1991. 26. Xi Y, Bavzant ZP, Jennings HM. Moisture diffusion in cementitious materials adsorption isotherms. Adv Cem Based Mater 1994;1(6):248–57. 27. Dubina E, Wadsö L, Plank J. A sorption balance study of water vapour sorption on anhydrous cement minerals and cement constituents. Cem Concr Res 2011;41(11):1196–204. 28. Simonin F. Comportement thermomécanique de bétons réfractaires alumineux contenant du spinelle de magnésium. Lyon: INSA; 2000 (Ph.D. thesis). 29. Turriziani R. In: Taylor HFW, editor. The chemistry of cements. 1st ed. London: Academic Press; 1964. 30. Mohmel S, Gebner W, Muller D. The behaviour of CA/CA2 cements during hydration and thermal treatment. In: Proceedings Unitecr 97. 1997. p. 1273–82 (3). 31. De Aza AH. Diseno y desarollo de materiales de alta alumina con matrices de espinela y hexaluminato calcico. Spain: Universidad Autonoma de Madrid; 1997 (Ph.D. thesis). 32. Fumo DA, Segadaes AM. Effect of silica fume additions on the hydration behaviour of calcium aluminates. J Am Ceram Soc 1997;3:1325–33.

Please cite this article in press as: Maaroufi M-A, et al. Thermo-hydrous behavior of hardened cement paste based on calcium aluminate cement. J Eur Ceram Soc (2014), http://dx.doi.org/10.1016/j.jeurceramsoc.2014.11.029