- Email: [email protected]
Impact of drying on pore structures in ettringite-rich cements
Cement and Concrete Research 84 (2016) 85–94
Contents lists available at ScienceDirect
Cement and Concrete Research journal homepage: www.elsevier.com/locate/cemconres
Impact of drying on pore structures in ettringite-rich cements I. Galan a,⁎, H. Beltagui b, M. García-Maté c, F.P. Glasser a, M.S. Imbabi b a b c
Department of Chemistry, Meston Building, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom School of Engineering, Fraser Noble Building, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom Departamento de Química Inorgánica, Cristalografía y Mineralogía, Universidad de Málaga, 29071 Málaga, Spain
a r t i c l e
i n f o
Article history: Received 7 August 2015 Accepted 4 March 2016 Available online xxxx Keywords: Drying (A) Mercury porosimetry (B) Pore size distribution (B) 3CaO·3Al2O3·CaSO4 (D) Ettringite (D)
a b s t r a c t Drying techniques affect the properties of cement pastes to varying extents. The effect of different drying techniques on calcium sulfoaluminate-based (C$A) cements and their constituent phases is reported for a range of simulated and commercial C$A pastes which are benchmarked against an OPC paste. The recommended methodologies used to dry samples were identified from the literature and include D-drying and solvent exchange. These methods were used in conjunction with mercury intrusion porosimetry (MIP) and X-ray powder diffraction (XRPD) measurements to assess the changes in pore structure and the damage to crystalline phases, respectively. D-drying and isopropanol exchange are the most satisfactory and least damaging methods for drying C$A based pastes. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction An important factor influencing the durability of cement systems is permeability. Cement pastes are porous across a wide range of length scales and pore networks often have interconnections which collectively define permeability. However, permeability is difficult to measure directly and it is common to focus instead on pore measurements. The pore structure greatly influences the permeability as well as other properties such as compressive strength and ionic diffusion. Image analysis, BET gas adsorption, mercury intrusion porosimetry (MIP) and scattering techniques (tomographic X-ray, neutron) have been utilised to map the pore structure. Many commonly used methods to determine pore structures require samples to be dried to empty the pores with concomitant danger of irreversible alteration of porosity and pore interconnectivity. The use of MIP to measure the porosity of hydrated cement pastes has been much debated in the literature. MIP has high sensitivity in the right pore size range for cement pastes but, besides being a potentially intrusive and damaging method, it does not necessarily measure the actual pore sizes but instead measures an intruded volume when a given pore entry size is exceeded [1]. Moreover, isolated pores cannot be accessed by mercury. However, these limitations can also be an advantage if the requirement is to obtain information relating to permeability: the pore entry size will give information about the accessibility to the connected pore network. Non-destructive techniques requiring no special specimen preparation are available, such as X-ray computed tomography (CT) and ⁎ Corresponding author. Tel. +44 1224274733. E-mail address: [email protected] (I. Galan).
http://dx.doi.org/10.1016/j.cemconres.2016.03.003 0008-8846/© 2016 Elsevier Ltd. All rights reserved.
neutron scattering. However, these methods are still being developed: data are difficult to obtain and interpret unequivocally. It is also likely that most commercial CT instruments do not have sufficient resolution to map pores smaller than a few microns. While progress has been made in interpreting the data [2–5] very little information relating to permeability has thus far been achieved. Previous studies on the influence of drying methods on the pore structure focus mainly on hydrated OPC pastes, and much conflict is found. There is however general agreement that oven drying at high temperatures (105 °C) is the most damaging method as it expels structural water from C–S–H [6] and it also has a strong impact on ettringite which above approximately 70 °C dissolves and leads to monosulfoaluminate formation [7]. A number of less damaging methods are described in the literature; oven drying (at various temperatures b100 °C), desiccant drying over either silica gel or calcium chloride, D-drying, vacuum drying, solvent replacement and freeze-drying. While a number of studies [8,9] suggest that freeze-drying is the best method to preserve the microstructure, others assert that this method causes irreversible change of the pore structure [10,11] and also results in severe cracking [12]. Collier et al. [12] suggest that acetone quenching, in which water is displaced, thereby stopping hydration reactions but without dissolving significant quantities of cement components, was the better method. According to Konecny and Naqvi [13] solvent replacement best preserves the coarser pore structure, although freeze drying best preserves pores with radius smaller than 5 nm. Zhang and Scherer [14] recommend using isopropanol exchange followed by ambient drying as being the least destructive to the microstructure while also preserving the pore structure of the cement. Snoeck et al. [15] checked the influence of different drying techniques on the water sorption properties
86
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
of Portland cement pastes concluding that isopropanol exchange followed by vacuum drying was the least damaging method. Feldman and Beaudoin [16] and Beaudoin and Tamtsia [17] compared methanol and isopropanol exchange, finding less damage when using isopropanol. It is probable that methanol interacts chemically with cement paste forming artefact solids such as methoxides [18–20]. Calcium sulfo-aluminate (C$A) cements are receiving increasing attention since their manufacture produces less CO2 than ordinary Portland cement (OPC) [21–24] and also due to their special properties which cannot be achieved with Portland cement, such as very high early strength. These binders may have quite variable compositions, but the clinkers normally contain more than 30 wt.% of ye'elimite, also known as Klein's compound or tetracalcium trialuminate sulfate (C4A3$) [25] and, when mixed with calcium sulfate and hydrated, form much AFt (ettringite). Cements with large amounts of ye'elimite may have special applications such as achieving high strength development at early-ages [26,27] and as matrices for radioactive waste encapsulation [28,29]. Whatever the application, durability in the cement service environment is important and it is likely that durability is related to a number of factors, including pore structure and permeability. Not much has been reported about drying damage to C$A based pastes. Typically, the clinker is blended with 6–15% gypsum prior to hydration: AFm, ettringite and aluminium hydroxide are major constituents of hardened C$A pastes. Zhang and Glasser [30] showed that ettringite (AFt) and monosulfate (AFm) phases in cement pastes are particularly sensitive to the drying method used. Both contain much molecular water and can lose crystallinity easily upon heating and/or drying. Not surprisingly, ettringite-rich cement pastes tend to be more sensitive than OPC pastes to the drying method used. While OPC pastes can also undergo drying damage, the problem becomes potentially more severe in the case of C$A cements which contain high amounts of both AFm and AFt. The stability of ettringite in terms of temperature and water vapour pressure has been reported [31]. Formation and decomposition of ettringite have been shown to be reversible with hysteresis. The initial decomposition product, a so-called meta-ettringite, has low crystallinity to XRD and a variable water content of 10–13 molecules per formula unit [32]. AFm can exist in different hydration states but the series based on C4AH13 and including SO4-AFm almost always predominates. Stability of AFm phases as a function of relative humidity (and thus drying technique) has also been reported [33]. The present paper reports new data and the impact of different drying methods in ettringite-based matrices.
2. Experimental procedure 2.1. Materials The experimental work was carried out on hardened cement paste samples, all of which were cured for a minimum of 120 days at ambient and as such can be considered to be ‘mature pastes’. The range of samples used for the study included (i) hydrated ye'elimite mixed with gypsum, (ii) laboratory synthesized C$A cement, (iii) commercial Chinese C$A cement and (iv) an OPC sample which was used as a benchmark. Apart from these multiphase clinkers, phase pure ettringite (AFt) was also synthesized. The compositions of the reactant systems are shown in Table 1. ‘Laboratory synthesized’ is defined as made in temperature controlled furnaces using laboratory-grade reactant chemicals. The specimens intended for measurement were cast in 25 × 25 mm cylinders and cured. Ye'elimite and C$A samples require higher water/binder ratios to completely hydrate and form hydration products with high water content such as ettringite. Specific values for each set of samples are explained in the following sections.
Table 1 Sample information. Sample ID
Cement composition
Water/binder ratio
YE-1
Laboratory synthesized ye'elimite with 33% added gypsum Laboratory synthesized ye'elimite with 9% added gypsum Laboratory synthesized C$A cement with 9% added gypsum Commercial Chinese C$A cement CEM I 52.5R
0.64
YE-2 Lab-C$A Com-C$A OPC
0.85 0.70 0.70 0.32
Note: percentages of gypsum are expressed per sample (YE-1 contains 67% ye'elimite and 33% gypsum).
2.1.1. Ettringite Phase-pure ettringite (C6A$3H32) was prepared according to the method reported by Balonis and Glasser [34]. Ettringite was made by mixing two solutions: the first solution from 6.65 g Al2(SO4)3·18H2O added to 100 ml ultrapure water and the second, by dispersing 4.44 g Ca(OH)2 in 250 ml of ultrapure water. The two solutions were mixed under nitrogen and then diluted to 500 ml with additional reagent water to which 0.5 ml of 1 M NaOH had been added. The mixture was sealed in a high density polyethylene (HDPE) bottle and stirred vigorously on a 60 °C hot plate for 48 h. The purity was checked using XRD. The filtered precipitate was then dried using the range of methods described subsequently. Water content calculation in ettringite samples After drying, the residual water contents of AFt samples were calculated according to Eq. (1) from Zhang and Glasser [30]: nAFt ¼ ð1248:6 31:7 18ÞðW W i Þ=18W i
ð1Þ
where nAFt is the number of water molecules left in AFt after drying (using the different methods explained in Section 2.2), Wi is the weight after ignition at 1050 °C for ~45 min and W is the weight before ignition. The results obtained are shown in Table 2. 2.1.2. Laboratory samples Ye'elimite was synthesized by mixing stoichiometric amounts of previously dried CaCO3, Al2O3 and CaSO4 and heating the mix at 1250 °C for a total of 45 min with intermediate grinding every 15 min. Purity was confirmed by XRD and weight loss measurements to ensure that SO3 was retained. The ye'elimite was mixed with laboratory grade gypsum before adding water to form a paste. Table 1 shows the two gypsum/binder and water/binder ratios chosen. The “water” in the water/binder ratio does not include water contained in gypsum. The first mix (YE-1) had gypsum/binder 0.33 and water/binder 0.64; these values being similar to the theoretical amounts needed for the formation of ettringite (gypsum/ye'elimite 0.56 and water/ye'elimite ~ 1). The composition of the second mix (YE-2), with gypsum/binder 0.09 and water/binder 0.85, was chosen to form a physical mixture of AFm and AFt (see Fig. 5 pattern before drying for details: the AFm formed was sulfate-AFm: C4A$H12). Table 2 Water content of dried AFt. These samples were unsuitable for MIP. Drying method
Drying time
H2O content
Damage (based on XRD)
Saturated solution of CaCl2 D-dried Oven dried at 40 °C Oven dried at 60 °C Oven dried at 60 °C Oven dried at 105 °C
~2 months 15 h 20 h 6h 24 h 72 h
31.8 31.5 29.2 16.6 12.7 12.1
No No No Yes Yes Yes
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Laboratory synthesized C$A (Lab-C$A in Table 1) was obtained by mixing CaCO3, Al2O3, CaSO4, Fe2O3, and SiO2 and submitting the mix to 4 cycles of 15 min each at 1300 °C with intermediate grinding between cycles. The initial mix of raw materials was calculated to form 50% ye'elimite, 25% belite and 25% ferrite. The resulting clinker mineralogy, calculated using the Rietveld method, comprised 53% ye'elimite, 21% belite, 18% ferrite, 4% tri-calcium aluminate and 4% gehlenite (C2AS). The result of the fitting is shown in Supplementary Fig. 1 of the Supplementary information. The fact that the ye'elimite contents are very similar to the expected output means that sulfur losses were negligible during the synthesis. It can also be deduced that part of the SiO2 calculated to form belite has actually formed C2AS and that part of the alumina, instead of forming ferrite, has formed C3A and C2AS. To form workable pastes with set times ~ 60 min, 9% gypsum and water were added to give a water/binder ratio of 0.7. 2.1.3. Commercial C$A samples Commercial C$A cement from Shenzhen Chenggong Building Materials Co., Ltd. PRC was used to prepare paste samples with w/c ratio 0.7. According to our Rietveld analysis, the cement contained 57% ye'elimite, 17% belite, 7% gehlenite (C2AS), 7% calcium aluminate (CA), 6% perovskite (alumina substituted CaTiO3), 3% mayenite (C12A7), 2% tri-calcium aluminate (C3A) and 1% anhydrite (see Supplementary Fig. 2 of the Supplementary information). The w/c ratio was chosen after trials to give a reasonable workability without plasticisers. 2.2. Drying methods After curing for at least 120 days in a sealed desiccator at ~99% RH, the samples were broken into pieces of 1 cm3 maximum and exposed to the different drying methods. The maximum fragment size chosen fulfils the geometry requirements of MIP measurements. The following methods have been used for drying cement and concrete samples intended for pore structural analysis techniques: i. Oven drying at selected temperatures (40 °C, 60 °C and 105 °C) for 24 h. ii. Solvent replacement, by submersion in isopropanol for 3 days at room temperature (~ 20 °C) followed by oven drying at 40 °C for 24 h: isopropanol was used to stop the hydration of the samples, followed by solvent removal in a low temperature oven (~ 40 °C) for 24 h. iii. Drying the samples in a desiccator at ~20 °C over a saturated solution of calcium chloride for sufficient time to achieve a constant weight. iv. D-drying (for varying times): D-drying uses a vacuum desiccator connected to a trap which is kept at −79 °C by a bath of solid carbon dioxide and ethanol. At this temperature the water vapour pressure is ~0.5 μm Hg [35]. This method has been proposed to avoid damage to the microstructure of cements [30]. The recommended duration required for D-drying is variable and is related to the performance of the pump used: “the rate of drying is affected greatly by the pressure in the system” [35]. Zhang and Glasser [30] suggest that Ddrying for 2–3 h is effective in drying a phase pure ettringite sample without damage on the basis of weight change. Korpa and Trettin [8] showed that D-drying for 12 h may be insufficient to dry samples (with particle sizes b1200 μm). However, they did not attempt drying for longer. Copeland and Hayes [35] suggest that 6–7 days of Ddrying may be required to achieve stable weight when the pressure is maintained at around 30 μm Hg. v. Vacuum drying at ~20 °C: the sample is placed in a desiccator and attached to a vacuum. The performance of the pump and the vacuum achieved, not always known to the user, are variables to consider. For the present work a “Speedivac” ES75 pump with no gauge was used, the minimum vacuum achievable being above ~5 μm Hg. The lack of gauge in many vacuum pumps makes it difficult for users to
87
determine exactly the level of vacuum achieved during drying. vi. Solvent replacement in acetone for 3 days at room temperature (~20 °C) followed by oven drying at 40 °C for 24 h: acetone is used to stop the hydration of the sample and replace free water, followed by drying at a low temperature, ~40 °C, to remove the acetone. For all treatments, the loss of water during drying was measured by following the mass of samples, and calculated according to Eq. (2): weight loss ð% Þ ¼
initial mass final mass 100: initial mass
ð2Þ
All samples were analysed by X-ray diffraction (XRD) before and after each drying method in order to assess the degradation of crystallinity. 2.3. Mercury intrusion porosimetry (MIP) MIP measurements were made using a Micrometrics Autopore IV 9500 instrument, which has a pore measuring range between 0.003 and 360 μm diameter. Samples were sealed in a penetrometer which is initially placed in a low pressure chamber and data recorded, followed by a high pressure chamber. The first stage of the operation of the equipment is to lower the pressure within the chamber to ~ 50 μm Hg. This stage of the operation does not involve mercury. In the second stage, the volume of mercury penetrating the sample is measured as a function of increasing pressure (P) and is related to the pore access diameter (D) of the pores penetrated by the Washburn equation (Eq. (3)): D¼
4γ cosθ P
ð3Þ
where γ is the surface tension of the mercury and θ is the contact angle of the mercury with the sample. A constant contact angle θ of 130° was assumed [30]. To check reproducibility, several sets of similar samples were measured under similar conditions. The differences in the total pore volume encountered were ~1–2%. The error of the MIP porosity values reported in this paper is thus taken as ±2%. 2.4. X-ray powder diffraction (XRPD) XRPD measurements were performed on the anhydrous C$A clinkers and on the hydrated pastes. The patterns for the C$A clinkers, both laboratory and commercial, were measured on an Empyrean diffractometer (PANalytical) using strictly monochromatic CuKα1 radiation (λ = 1.54056 Å). In order to determine the composition of the clinkers, these samples were analysed using the Rietveld methodology as implemented in the GSAS software package [36]. Final global optimised parameters were: background coefficients, zero-shift error, cell parameters and peak shape parameters. Peak shapes were fitted using the pseudo-Voigt function [37]. The patterns for the hydrated pastes, before and after drying, were measured on an X'Pert PRO diffractometer (PANalytical) using CuKα radiation. In both data sets the X-ray tube operated at 45 kV and 40 mA. 3. Results and discussion 3.1. AFt (ettringite) XRD patterns for the initially phase pure AFt dried using the various methods are shown in Figs. 1 and 2. Table 2 shows the water content remaining in AFt after drying. Damage is apparent in samples dried at 60 and 105 °C for 6 h, both becoming amorphous to XRD and with the number of water molecules being reduced to 16 and 12 respectively. In the remainder of dried AFt samples, crystallinity is preserved. This
88
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 1. XRD patterns (filtered CuKα radiation) of AFt dried under three different conditions: vacuum, D-drying and over saturated CaCl2 solution at room temperature (~20 °C). None of these methods caused damage to the AFt.
indicates that if the sample loses more than ~3 of the original 32 molecules of water, a measurable degradation of crystallinity occurs. The degradation is often inhomogeneous and it is not uncommon to observe physical mixtures of damaged and undamaged AFt. Attempts to test these samples using MIP were not successful, even in the case of samples that appeared to be damaged, according to both the XRD analysis results and loss of water molecules. The low pressure stage, up to 50 μm Hg, could not be reached, indicating that the samples were insufficiently dried.
3.2. Hydrated cement samples Table 3 shows the drying time, weight loss after drying and volume porosity measured by MIP (cut off limits were 0.003–360 μm). Where no MIP data are recorded the samples could not be measured, i.e. the vacuum necessary to commence measurement (50 μm Hg) could not be attained. “Drying time” normally means that constant weight was achieved. However in a few experiments the drying process was deliberately stopped before stable weight was reached to check if MIP permeation could be achieved. Not all methods were applied to all samples mainly for operation reasons. Also in some samples some of the techniques were not considered relevant.
3.2.1. YE-1 results The mix composition of samples YE-1 used stoichiometric quantities of ye'elimite, gypsum and water to obtain ettringite.
For this sample, all drying methods except oven drying at 105 °C were shown to preserve at least partially the mineralogy of the sample (Fig. 3). However ettringite may have partially decomposed to poorly crystallised products such as AFm, aluminium hydroxide (gibbsite) and an anhydrite-like phase. Decomposition is inhomogeneous, giving mixtures of unaffected ettringite and its decomposition products. Since XRD mainly detects the undamaged portion, crystallinity appears to persist. Not surprisingly the porosities (Table 3) and MIP curves (Fig. 4a and b) differ significantly for each drying method. Oven drying produced the highest porosity, while isopropanol replacement produced the least. Even though the weight losses achieved with acetone and isopropanol were very similar, the sample dried in acetone could not be measured by MIP as the required vacuum, ~50 μm Hg, could not be attained, implying that the sample was insufficiently dry for the test: either water or acetone, or a mixture of both, was retained as pore blockers. The additional volume intrusion spike at ~ 20 μm for samples dried at 105 °C is a noteworthy feature that may also be a proof of damage caused by drying, i.e., pore opening or cracks or both occurred during drying. For this set of samples, a clear correlation exists between weight lost and intruded pore volume (Fig. 12).
3.2.2. YE-2 results The XRD patterns of YE-2 samples are shown in Fig. 5. The only drying treatment by which the sample was not damaged was vacuum treatment but, as can be seen in Table 3, MIP was not possible because drying was incomplete. For YE-2, the higher the weight loss during drying, the higher the porosity measured by MIP (Fig. 12). In the D-dried
Fig. 2. AFt samples dried in an oven at 40, 60 and 105 °C. Note the partial (at 60 °C) and complete loss (at 105 °C) of crystallinity in the latter two samples.
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
89
Table 3 Weight loss after drying and intruded pore volumes. Sample ID
Drying method
Drying time
Weight loss (%)
Mercury porosity (%)
YE-1
Oven at 105 °C D-drying Isopropanol + oven at 40 °C Acetone + oven at 40 °C Oven at 105 °C Oven at 60 °C D-drying Acetone + vacuum Vacuum Oven at 105 °C D-drying Isopropanol + oven at 40 °C Saturated solution of CaCl2 Acetone + oven at 40 °C D-drying Vacuum Oven at 105 °C Oven at 60 °C Isopropanol + oven at 40 °C Saturated solution of CaCl2 D-drying D-drying Acetone + oven at 40 °C Vacuum Oven at 105 °C Isopropanol + oven at 40 °C D-drying Vacuum
24 h ~13 h 24 h 24 h 24 h 24 h ~50 h ~30 h ~60 h 24 h 12 h 24 h 12 days 24 h 6h 6h 24 h 24 h 24 h ~1.5 months 6h 23 h 24 h 45 h 24 h 24 h 18 h 6h
23.9 14.7 2.7 2.9 26.9 23.9 20.9 10.1 8.5 27.5 17.1 7.9 8.1 7.0 9.8 7.7 26.1 23.1 9.3 13.3 9.4 17.1 9.7 14.2 13 6.4 4.3 –
45.3 40.8 26.6 No MIP 40.5 38 35.1 No MIP No MIP 38 32.4 29.9 24.7 24.3 No MIP No MIP 34.1 32.8 38 29.3 29.6 24.5 34.3 No MIP 12 9.5 13.2 8.1
YE-2
Com-C$A
Lab-C$A
CEM I
sample the reflections of ettringite are still visible but are significantly reduced in intensity. In the oven dried (105 °C) sample ettringite has decomposed and only AFm can be detected by XRD. Considering less harsh drying conditions, MIP data (Fig. 6a and b) show that D-dried samples are less affected than after oven drying at 60 °C. Although most of the pore volume is in the same diameter ranges, it is clear that the drier the sample, the greater the accessible porosity. This behaviour was also noted in the sample YE-1 dried at 105 °C, much extra porosity appearing in the range 10–20 μm. The occurrence of the peak at 0.01 μm in the log differential curves of YE-2 (Fig. 6a), which does not show in the YE-1 sample, may be attributed to the different microstructure of the samples, YE-2 forming more AFm due to the higher water and lower gypsum than YE-1. 3.2.3. Commercial C$A From the XRD patterns (Fig. 7), the sample dried at 105 °C is badly damaged after drying: AFt, abundant in the undried sample, has been destroyed. The other methods gave well defined reflections for other
Fig. 4. a. Log differential intrusion curves for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64). The duration of drying treatments is specified in Table 3. b. Cumulative intrusion curves for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64) in ml per gramme of sample after drying treatment.
hydrates: strätlingite, AFt and AFm. It is noticeable that even though the samples were cured for at least 120 days, there was still significant residual unhydrated belite, the persistence of which may have an impact on water losses and pore volumes. The MIP curves (Fig. 8a and b) can be divided into two groups. The first group, which includes the samples dried at 105 °C, D-dried for 12 h and with isopropanol, shows a pronounced intruded volume maximum between 0.6 and 1 μm. The second group, which includes the samples dried with acetone and over CaCl2, shows a broad pore spectrum between ~0.01 and 1 μm. These last two samples needed a much longer vacuum treatment to measure MIP than the samples from the first group. The time taken for the samples to achieve the required vacuum, ~ 50 μm Hg, can in some cases also be damaging. The first group of samples shows a correlation between weight loss and pore
Fig. 3. XRD patterns for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64) before and after drying.
90
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 7. XRD patterns for commercial C$A (w/c = 0.7). St: strätlingite, AFm: monosulfoaluminate, AFt: ettringite, β-C2 S: β-belite. Fig. 5. XRD patterns for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85). AFm stands for sulfate-AFm, monosulfoaluminate or C4A$Hx and OH-AFm stands for hydroxy-AFm or C4AHx.
volume (Fig. 12). The sample dried at 105 °C was the most damaged, according to both XRD and MIP results. D-drying and isopropanol give very similar results. The samples dried over CaCl2 and with acetone are not sufficiently dry to perform reliable MIP measurements. For this type of samples D-drying and isopropanol replacement would be the recommended methods.
3.2.4. Lab-C$A results The XRD patterns of the laboratory C$A samples are shown in Figs. 9 and 10 and the corresponding MIP curves in Fig. 11a and b. The 3 samples shown in Fig. 9, dried with acetone, isopropanol and D-dried for 6 h appeared to be undamaged by XRD and lost only around 9% weight during drying. But these samples show a peak of porosity at ~ 20 μm (Fig. 11a and b) which does not occur for samples that were drier, according to the weight losses, prior to intrusion. For these 3 samples, dried in acetone, isopropanol and D-dried for 6 h, the first
Fig. 6. a. Log differential intrusion curves for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85). b. Cumulative intrusion curves for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85).
Fig. 8. a. Log differential intrusion curves for commercial C$A (w/c = 0.7). b. Cumulative intrusion curves for commercial C$A (w/c = 0.7).
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
91
Fig. 9. XRD patterns of laboratory C$A dried by solvent exchange (isopropanol and acetone) and by D-drying for 6 h.
stage of MIP measurement took significantly longer than for other samples dried with vacuum, or D-drying for 23 h or oven dried at either 60 or 105 °C. The appearance of these larger pores may be due to damage arising from the first stage of the MIP operation where mercury is not involved and where the pressure is lowered to vacuum state (50 μm Hg). A proof of this is the fact that a ‘spike’ appears for all three samples at approximately the same diameter (~20 μm) suggesting that this is a damage-related feature. This does not mean D-drying for 23 h is less damaging than D-drying for 6 h (as shown by the XRD): it shows that 6 h of D-drying is not enough to perform reliable MIP measurements. The MIP vacuum stage will damage the sample more than the drying process if the sample is not dry enough. The sample D-dried for 23 h presents the lowest intruded volume, as measured by MIP, indicating that this drying method was the less aggressive but sufficient to dry the sample. Moreover the vacuum needed
Fig. 10. XRD patterns of laboratory C$A dried under vacuum, D-dried for 23 h, and oven dried at 60 and 105 °C.
Fig. 11. a. Log differential intrusion curves for laboratory C$A cement with 10% gypsum (hydrated at w/c = 0.7). b. Cumulative intrusion curves for laboratory C$A cement with 10% gypsum (w/c = 0.7). As explained in the text, the samples dried in acetone, isopropanol and D-dried for 6 h seem to have been damaged during the first stage of MIP: notice the spike at ~20 μm because they were not sufficiently dry. Samples D-dried for 23 h and oven dried at 60 and 105 °C do not show this spike.
for MIP did not further damage it, as occurred in “wet” samples following either solvent replacement (acetone, isopropanol) or D-drying for 6 h. Excluding the samples which were insufficiently dried or had been damaged during vacuum stage prior to MIP, there is a correlation between weight loss and pore volume (Fig. 12), although the correlation is not as good as in previous examples. 3.2.5. OPC An OPC sample was used as a benchmark; none of the drying methods used appear by XRD to cause damage to the mineralogy of the hydrated cement paste (Fig. 13). However, changes to C–S–H are not well revealed by XRD. Drying at 105 °C almost certainly causes microstructural damage, such as microcracking, and this would be revealed in the course of mercury intrusion as the increased porosity shown in Fig. 14a and b. Oven dried samples show higher porosity at all diameters than the samples dried in vacuum or isopropanol and also a shift in the peak where most pores concentrate towards greater diameters. This agrees with Moukwa and Aitcin [6], who reported pore opening in the range of 0.02 and 0.1 μm after drying cement paste samples for 24 h at 105 °C. According to these authors “oven drying can alter the pore structure mainly by rearranging the hydration products”. Also Galle [9] reported an overestimation of total porosity when drying at 105 °C attributed to significant losses of nonevaporable water from both AFt and C–S–H. Apart from the cement hydrate desiccation and the rearrangements, this later author also attributes the damage to the potential microcrack generation in relation with internal thermally-induced stresses. The D-dried sample of OPC has suffered damage during MIP measurement and the porosity at around 100 μm is tentatively attributed to that damage because it does not show in the other samples. This
92
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 12. Pore volume (ranging pores from 360 to 0.003 μm) versus weight loss after drying in samples of YE-1 (dried at 105 °C, with isopropanol and D-dried, YE-2 (dried in an oven at 60 and 105 °C and D-dried), com-C$A (dried at 105 °C, D-dried and with isopropanol) and lab-C$A (dried with vacuum, D-dried for 23 h and in an oven at 60 and 105 °C).
sample lost less weight during drying but had higher pore volume and increase of porosity at all diameters greater than ~ 0.05 μm after the drying + MIP process, again indicating damage during MIP measurement. According to Korpa and Trettin [8] a short duration D-drying can lead to incomplete water removal, mainly from the very small gel pores, b2 nm, and will not give realistic values for gel nanopores. Samples subjected to vacuum and isopropanol drying show very similar results: total porosity was between 8 and 9% pore volume and most of the pores had diameters in the region 0.01 to 0.1 μm. Again there is a correlation between pore volume and weight loss, excluding the sample damaged during D-drying, where the pore volume is higher due to the pores created during the drying process. Collier et al. [12] used acetone replacement (7 day immersion followed by 10−2 mbar vacuum for 3 days) and vacuum (10−2 mbar vacuum for 7 days) for drying hydrated OPC samples and reported that none of the techniques caused any major deterioration in the composition and microstructure of the samples. Similar conclusions were drawn by Snoeck [15], who reported that the unavoidable changes in the microstructure of hydrated OPC samples due to drying are minimized when using vacuum drying and isopropanol exchange. This later method is also recommended by Zhang and Scherer [14] as the best “for preserving the microstructure with minimal effects on the composition of the cement”. The MIP data reported for OPC in the literature are similar to the ones obtained here. Konecky and Naqvi [13] reported on mortar samples dried in an oven at 105 °C and by solvent replacement. The oven dried ones showed most of the pores having diameters between 0.05
and 0.2 μm whereas the solvent dried ones were in the range 0.02– 0.1 μm. Galle [9] reported that most of the pore diameter concentrated between 0.03 and 0.15 μm in oven dried OPC pastes and between 0.02 and 0.08 μm in vacuum dried OPC pastes. The total porosity in these samples is slightly higher, 14–15%, than in the ones reported in the title study. Absolute comparison of reported values and those of the title study is not possible because the samples and their conditioning are not identical, but it is worth noting that in comparative terms, there is general agreement on the effects of the different drying methods on OPC paste pore volume and pore distribution. Finally, if we compare the MIP values for OPC with those obtained for C$A matrices, a shift towards greater pore diameters is observed in the differential intrusion curves of C$A. In OPC samples the majority of the pores concentrate on diameters between 0.02 and 0.06 μm whereas in C$A, they range from 0.02 to 0.2 μm in the laboratory C$A, and from
Fig. 13. XRD patterns of OPC samples before and after drying using vacuum, isopropanol exchange, oven at 105 °C and D-drying. P: portlandite, A: alite, B: belite.
Fig. 14. a. Log differential intrusion curves for OPC (w/c = 0.32). b. Cumulative intrusion curves for OPC (w/c = 0.32).
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
0.3 to 1 μm in the commercial C$A. The total pore volume is also much smaller in the OPC, 8–9%, than in the C$A samples, 25–30%. This could be in part explained by the different water/cement ratio used, which was much lower for OPC than for C$A. But, given the intrinsic difference in water demand between the two types of cement, the differences in pore sizes and pore volume seem to be more related to the microstructure of the pastes than the amounts of free water. 4. Additional discussion The study confirms that AFt is readily damaged by drying. Zhang and Glasser [30] suggested that the crystallinity of AFt begins to degrade at b30 molecules of water per formula unit. This is in accord with our results (see Section 3.1) obtained on the sample dried at 40 °C; the number of water molecules in AFt was reduced from 32 to ~ 29 and it remained fully crystalline by XRD. However, as more water is lost, the AFt loses water and degrades to a plateau at ~ 12H2O giving rise to a phase termed meta-ettringite [32]. Mixtures of crystalline and metaettringite are commonly encountered: AFt reflections can still be identified at low water contents but the crystalline reflections reduce in intensity as degradation proceeds. AFt also converts in part to AFm (Figs. 5 and 7), going from 32 to ~12H2O molecules per unit formula. With progressive drying, damage to AFm is also noticeable as reduction in the reflection intensity in XRD patterns and the shift of reflections towards increasing 2θ angles, indicating a contraction in basal spacings with formation of lower water SO4-AFm types (Fig. 5), like those reported by [33]. It is clear that drying at ~100 °C severely damages ettringite-based matrices. Drying at ~ 60 °C is also damaging unless the drying time is kept short, less than ~ 5 h, meaning that a stable weight may not be reached: continuing evolution of water vapour protects the crystallinity of at least some of the residual AFt. Drying at 40 °C does not damage ettringite or C$A based materials, ettringite being stable at all but extremely low humidity. Although none of the vacuum dried samples from ye'elimite/C$A based cements used in this study were dry enough for MIP, the OPC sample dried for only 6 h using vacuum was sufficiently dried for MIP. The actual pressures achieved were not recorded but as the same pump was used for all vacuum-treated sample, a relative comparison between results can be made. While vacuum drying has been used in the past for OPC based samples, it does not seem appropriate for drying C$A based cements. As shown in the results, samples left drying under vacuum achieved constant weight but were still insufficiently dry to perform MIP (Table 3). The low pressure evacuation stage needed to perform MIP measurements can in some cases cause more damage than the actual drying process. This step is unavoidable as pores need to be empty enough to perform the high pressure stage where mercury is introduced in the pores. This was proven in samples apparently not damaged by the drying process that then give spikes (Fig. 11) or ‘bumps’ (Fig. 8) in their MIP curves. In these cases, as the samples were not sufficiently dry, the MIP was actually more damaging than the drying. The total values of pore volumes and weight losses of laboratory C$A versus commercial C$A are comparable and in the same range (Table 3). But pore sizes and distributions differ (Figs. 8 and 11). In the ye'elimite samples most pores have diameters of ~0.3–0.4 μm. In the laboratory C$A sample the majority of pores are slightly smaller, ~ 0.1–0.2 μm, but in the commercial C$A the ‘peak’ is shifted to more coarse porosity: 0.7–0.8 μm. These differences can be attributed to the residual unhydrated solid and the different proportion of hydrated phases formed and their densities: strätlingite, Aft, AFm and aluminium hydroxide. In the commercial C$A the amount of strätlingite is greater than in the laboratory C$A, where AFt is the predominant hydrated phase. The use of simulants for C$A such as pure ettringite or mixtures of phases, which were then hydrated, is useful to understand porosity in simpler systems, but simulation of porosity is not always
93
straightforward. The use of simulants as proxies for commercial cements should be done carefully keeping in mind all the extra parameters not taken into account, such as matrix effects and possible solid solutions formed and their effect on the porosity distribution. Finally, we note that the literature contains numerous papers pointing out that the intrusion process is damaging and affects the apparent pore size spectrum. We agree, but also note that the preliminary drying, required to apply many pore-sensitive techniques (MIP, BET), is the source of much damage. Indeed, it is possible that the drying damage may render the solid more liable to further damage in the course of intrusion. And, since drying underlies some of the techniques we use to characterize cement paste (Scanning Electron Microscopy, MIP, thermogravimetry), it is this first step that deserves critical scrutiny. In an attempt to reduce dependency on MIP, we are developing scale-spanning CT methods that are non-intrusive and can in principle be applied to water saturated samples. But the resolution of the refinements and the complex mathematics required to re-image the pore structure and its continuity are as yet insufficiently robust to make a quantitative approach. 5. Conclusions The main conclusions that can be drawn from the results presented are as follows: – An unexpected problem in determining damage is to establish a benchmark state. The hydration state of AFm phase(s) is sensitive to temperature and humidities at or near typical ambient laboratory conditions, 15–30 °C and ranges of relative humidities. Changes in this range affect the intrinsic mineralogy and especially, pore structure. – The extent to which damage is caused by changes in the constitution of C–S–H is difficult to characterize and depends on definitions of “free” and “bound” water in a particular specimen. – AFt retains its crystallinity when dried in an oven at 40 °C. This means that the drying associated with solvent removal, often done at 40 °C, is unlikely to contribute to damage. – C$A cements are more susceptible to drying damage than Portland cements on account of their AFt content relative to Portland cements. Mineralogical damage occurs to C$A samples exposed to prolonged drying times and harsh drying methods. Drying at 100 °C, still common in some engineering proceedings, will always damage cement paste. – D-drying and isopropanol exchange are suggested as conservative methods for drying C$A based samples. Times for D-drying may vary among samples and as a function of the vacuum achieved by the pump, and are not always easy to optimise. – MIP of cement matrices based on ettringite is particularly susceptible to damage. – In samples that are not dry enough for MIP, the first stage of vacuum needed to start the measurement, ~50 μm Hg, can cause substantial damage to the samples, increasing the porosity between 10 and 100 μm. Most previous concerns have been about damages resulting from the physical intrusion but previous drying treatment is partially responsible for artefacts. – None of the available methods to measure porosity and/or permeability are completely satisfactory. However the search for new methods should continue. BET is an obvious target for inclusion as it works well for pores in the size range below 0.1 mm. X-ray tomography is being used on comparable samples but its resolution and discrimination are as yet probably not sufficient to measure sizes and volumes of pores b 1 μm.
Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.cemconres.2016.03.003.
94
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Acknowledgments IG, HB, FPG and MSI gratefully acknowledge the financial support provided by the Gulf Organization for Research and Development (GORD), Qatar, through research grant number ENG016RGG0593. MGM acknowledges the financial support provided by the Junta de Andalucía through research grant number P11-FQM-07517. References [1] S. Diamond, Mercury porosimetry. An inappropriate method for the measurement of pore size distributions in cement-based materials, Cem. Concr. Res. 30 (2000) 1517–1525. [2] S. Diamond, E. Landis, Microstructural features of a mortar as seen by computed microtomography, Mater. Struct. 40 (2007) 989–993. [3] E. Gallucci, K. Scrivener, A. Groso, M. Stampanoni, G. Margaritondo, 3D experimental investigation of the microstructure of cement pastes using synchrotron X-ray microtomography (μCT), Cem. Concr. Res. 37 (2007) 360–368. [4] M. Zhang, Y. He, G. Ye, D.A. Lange, K.V. Breugel, Computational investigation on mass diffusivity in Portland cement paste based on X-ray computed microtomography (μCT) image, Constr. Build. Mater. 27 (2012) 472–481. [5] S. Yoon, I. Galan, K. Celik, F.P. Glasser, M.S. Imbabi, Characterization of micro-pore structure in novel cement matrices, Mater. Res. Soc. Symp. Proc. 2014 (1712). [6] M. Moukwa, P.-. Aitcin, The effect of drying on cement pastes pore structure as determined by mercury porosimetry, Cem. Concr. Res. 18 (1988) 745–752. [7] H. Taylor, C. Famy, K. Scrivener, Delayed ettringite formation, Cem. Concr. Res. 31 (2001) 683–693. [8] A. Korpa, R. Trettin, The influence of different drying methods on cement paste microstructures as reflected by gas adsorption: comparison between freeze-drying (Fdrying), D-drying, P-drying and oven-drying methods, Cem. Concr. Res. 36 (2006) 634–649. [9] C. Gallé, Effect of drying on cement-based materials pore structure as identified by mercury intrusion porosimetry — a comparative study between oven-, vacuum-, and freeze-drying, Cem. Concr. Res. 31 (2001) 1467–1477. [10] R.M. Espinosa, L. Franke, Influence of the age and drying process on pore structure and sorption isotherms of hardened cement paste, Cem. Concr. Res. 36 (2006) 1969–1984. [11] C. Gosselin, Microstructural Development of Calcium Aluminate Cement Based Systems With and Without Supplementary Cementitious MaterialsPhD thesis EPFL, 2009. [12] N.C. Collier, J.H. Sharp, N.B. Milestone, J. Hill, I.H. Godfrey, The influence of water removal techniques on the composition and microstructure of hardened cement pastes, Cem. Concr. Res. 38 (2008) 737–744. [13] L. Konecny, S.J. Naqvi, The effect of different drying techniques on the pore size distribution of blended cement mortars, Cem. Concr. Res. 23 (1993) 1223–1228. [14] J. Zhang, G.W. Scherer, Comparison of methods for arresting hydration of cement, Cem. Concr. Res. 41 (2011) 1024–1036. [15] D. Snoeck, L.F. Velasco, A. Mignon, S. Van Vlierberghe, P. Dubruel, P. Lodewyckx, N. De Belie, The influence of different drying techniques on the water sorption properties of cement-based materials, Cem. Concr. Res. 64 (2014) 54–62.
[16] R.F. Feldman, J.J. Beaudoin, Pretreatment of hardened hydrated cement pastes for mercury intrusion measurements, Cem. Concr. Res. 21 (1991) 297–308. [17] J.J. Beaudoin, B.T. Tamtsia, Effect of drying methods on microstructural changes in hardened cement paste: an A. C. impedance spectroscopy evaluation, J. Adv. Concr. Technol. 2 (2004) 113–120. [18] J.J. Beaudoin, P. Gu, J. Marchand, B. Tamtsia, R.E. Myers, Z. Liu, Solvent replacement studies of hydrated Portland cement systems: the role of calcium hydroxide, Adv. Cem. Based Mater. 8 (1998) 56–65. [19] J.J. Beaudoin, B. Tamtsia, J. Marchand, H.R. Myers, Solvent exchange in partially saturated and saturated microporous systems length change anomalies, Cem. Concr. Res. 30 (2000) 359–370. [20] R.L. Day, Reactions between methanol and Portland cement paste, Cem. Concr. Res. 11 (1981) 341–349. [21] E. Gartner, Industrially interesting approaches to “low-CO2” cements, Cem. Concr. Res. 34 (2004) 1489–1498. [22] J.H. Sharp, C.D. Lawrence, R. Yang, Calcium sulfoaluminate cements — low-energy cements, special cements or what? Adv. Cem. Res. 11 (1999) 3–13. [23] C.D. Popescu, M. Muntean, J.H. Sharp, Industrial trial production of low energy belite cement, Cem. Concr. compos. 25 (2003) 689–693. [24] A. Feraille, A. Alaoui, A. Steckmeyer, R. Le Roy, New cements for sustainable development, Proceedings of the 12th ICCC, Montreal, Canada., 2007. [25] S. Sahu, J. Majling, Phase compatibility in the system CaO·SiO2·Al2O3·Fe2O3·SO3 referred to sulfoaluminate belite cement clinker, Cem. Concr. Res. 23 (1993) 1331–1339. [26] K. Quillin, Performance of belite–sulfoaluminate cements, Cem. Concr. Res. 31 (2001) 1341–1349. [27] F.P. Glasser, L. Zhang, High-performance cement matrices based on calcium sulfoaluminate–belite compositions, Cem. Concr. Res. 31 (2001) 1881–1886. [28] C. Cau Dit Coumes, S. Courtois, S. Peysson, J. Ambroise, J. Pera, Calcium sulfoaluminate cement blended with OPC: a potential binder to encapsulate lowlevel radioactive slurries of complex chemistry, Cem. Concr. Res. 39 (2009) 740–747. [29] Q. Zhou, N.B. Milestone, M. Hayes, An alternative to Portland cement for waste encapsulation—the calcium sulfoaluminate cement system, J. Hazard. Mater. 136 (2006) 120–129. [30] L. Zhang, F.P. Glasser, Critical examination of drying damage to cement pastes, Adv. Cem. Res. 12 (2000) 79–88. [31] Q. Zhou, F.P. Glasser, Thermal stability and decomposition mechanisms of ettringite at b120 °C, Cem. Concr. Res. 31 (2001) 1333–1339. [32] Q. Zhou, E.E. Lachowski, F.P. Glasser, Metaettringite, a decomposition product of ettringite, Cem. Concr. Res. 34 (2004) 703–710. [33] L.G. Baquerizo, T. Matschei, K.L. Scrivener, M. Saeidpour, L. Wadsö, Hydration states of AFm cement phases, Cem. Concr. Res. 73 (2015) 143–157. [34] M. Balonis, F.P. Glasser, The density of cement phases, Cem. Concr. Res. 39 (2009) 733–739. [35] L.E. Copeland, J.C. Hayes, The determination of non-evaporable water in hardened Portland cement pastes, ASTM Bulletin No 194, 1953. [36] A.C. Larson, R.B. Von Dreele, General Structure Analysis System (GSAS), Los Alamos Laboratory, Rep.No.LA-UR, Vol. 862004. [37] P. Thompson, D.E. Cox, J.B. Hastings, Rietveld refinement of Debye–Scherrer synchrotron X-ray data from Al2O3, J. Appl. Crystallogr. 20 (1987) 79–83.
Contents lists available at ScienceDirect
Cement and Concrete Research journal homepage: www.elsevier.com/locate/cemconres
Impact of drying on pore structures in ettringite-rich cements I. Galan a,⁎, H. Beltagui b, M. García-Maté c, F.P. Glasser a, M.S. Imbabi b a b c
Department of Chemistry, Meston Building, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom School of Engineering, Fraser Noble Building, University of Aberdeen, Aberdeen AB24 3UE, United Kingdom Departamento de Química Inorgánica, Cristalografía y Mineralogía, Universidad de Málaga, 29071 Málaga, Spain
a r t i c l e
i n f o
Article history: Received 7 August 2015 Accepted 4 March 2016 Available online xxxx Keywords: Drying (A) Mercury porosimetry (B) Pore size distribution (B) 3CaO·3Al2O3·CaSO4 (D) Ettringite (D)
a b s t r a c t Drying techniques affect the properties of cement pastes to varying extents. The effect of different drying techniques on calcium sulfoaluminate-based (C$A) cements and their constituent phases is reported for a range of simulated and commercial C$A pastes which are benchmarked against an OPC paste. The recommended methodologies used to dry samples were identified from the literature and include D-drying and solvent exchange. These methods were used in conjunction with mercury intrusion porosimetry (MIP) and X-ray powder diffraction (XRPD) measurements to assess the changes in pore structure and the damage to crystalline phases, respectively. D-drying and isopropanol exchange are the most satisfactory and least damaging methods for drying C$A based pastes. © 2016 Elsevier Ltd. All rights reserved.
1. Introduction An important factor influencing the durability of cement systems is permeability. Cement pastes are porous across a wide range of length scales and pore networks often have interconnections which collectively define permeability. However, permeability is difficult to measure directly and it is common to focus instead on pore measurements. The pore structure greatly influences the permeability as well as other properties such as compressive strength and ionic diffusion. Image analysis, BET gas adsorption, mercury intrusion porosimetry (MIP) and scattering techniques (tomographic X-ray, neutron) have been utilised to map the pore structure. Many commonly used methods to determine pore structures require samples to be dried to empty the pores with concomitant danger of irreversible alteration of porosity and pore interconnectivity. The use of MIP to measure the porosity of hydrated cement pastes has been much debated in the literature. MIP has high sensitivity in the right pore size range for cement pastes but, besides being a potentially intrusive and damaging method, it does not necessarily measure the actual pore sizes but instead measures an intruded volume when a given pore entry size is exceeded [1]. Moreover, isolated pores cannot be accessed by mercury. However, these limitations can also be an advantage if the requirement is to obtain information relating to permeability: the pore entry size will give information about the accessibility to the connected pore network. Non-destructive techniques requiring no special specimen preparation are available, such as X-ray computed tomography (CT) and ⁎ Corresponding author. Tel. +44 1224274733. E-mail address: [email protected] (I. Galan).
http://dx.doi.org/10.1016/j.cemconres.2016.03.003 0008-8846/© 2016 Elsevier Ltd. All rights reserved.
neutron scattering. However, these methods are still being developed: data are difficult to obtain and interpret unequivocally. It is also likely that most commercial CT instruments do not have sufficient resolution to map pores smaller than a few microns. While progress has been made in interpreting the data [2–5] very little information relating to permeability has thus far been achieved. Previous studies on the influence of drying methods on the pore structure focus mainly on hydrated OPC pastes, and much conflict is found. There is however general agreement that oven drying at high temperatures (105 °C) is the most damaging method as it expels structural water from C–S–H [6] and it also has a strong impact on ettringite which above approximately 70 °C dissolves and leads to monosulfoaluminate formation [7]. A number of less damaging methods are described in the literature; oven drying (at various temperatures b100 °C), desiccant drying over either silica gel or calcium chloride, D-drying, vacuum drying, solvent replacement and freeze-drying. While a number of studies [8,9] suggest that freeze-drying is the best method to preserve the microstructure, others assert that this method causes irreversible change of the pore structure [10,11] and also results in severe cracking [12]. Collier et al. [12] suggest that acetone quenching, in which water is displaced, thereby stopping hydration reactions but without dissolving significant quantities of cement components, was the better method. According to Konecny and Naqvi [13] solvent replacement best preserves the coarser pore structure, although freeze drying best preserves pores with radius smaller than 5 nm. Zhang and Scherer [14] recommend using isopropanol exchange followed by ambient drying as being the least destructive to the microstructure while also preserving the pore structure of the cement. Snoeck et al. [15] checked the influence of different drying techniques on the water sorption properties
86
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
of Portland cement pastes concluding that isopropanol exchange followed by vacuum drying was the least damaging method. Feldman and Beaudoin [16] and Beaudoin and Tamtsia [17] compared methanol and isopropanol exchange, finding less damage when using isopropanol. It is probable that methanol interacts chemically with cement paste forming artefact solids such as methoxides [18–20]. Calcium sulfo-aluminate (C$A) cements are receiving increasing attention since their manufacture produces less CO2 than ordinary Portland cement (OPC) [21–24] and also due to their special properties which cannot be achieved with Portland cement, such as very high early strength. These binders may have quite variable compositions, but the clinkers normally contain more than 30 wt.% of ye'elimite, also known as Klein's compound or tetracalcium trialuminate sulfate (C4A3$) [25] and, when mixed with calcium sulfate and hydrated, form much AFt (ettringite). Cements with large amounts of ye'elimite may have special applications such as achieving high strength development at early-ages [26,27] and as matrices for radioactive waste encapsulation [28,29]. Whatever the application, durability in the cement service environment is important and it is likely that durability is related to a number of factors, including pore structure and permeability. Not much has been reported about drying damage to C$A based pastes. Typically, the clinker is blended with 6–15% gypsum prior to hydration: AFm, ettringite and aluminium hydroxide are major constituents of hardened C$A pastes. Zhang and Glasser [30] showed that ettringite (AFt) and monosulfate (AFm) phases in cement pastes are particularly sensitive to the drying method used. Both contain much molecular water and can lose crystallinity easily upon heating and/or drying. Not surprisingly, ettringite-rich cement pastes tend to be more sensitive than OPC pastes to the drying method used. While OPC pastes can also undergo drying damage, the problem becomes potentially more severe in the case of C$A cements which contain high amounts of both AFm and AFt. The stability of ettringite in terms of temperature and water vapour pressure has been reported [31]. Formation and decomposition of ettringite have been shown to be reversible with hysteresis. The initial decomposition product, a so-called meta-ettringite, has low crystallinity to XRD and a variable water content of 10–13 molecules per formula unit [32]. AFm can exist in different hydration states but the series based on C4AH13 and including SO4-AFm almost always predominates. Stability of AFm phases as a function of relative humidity (and thus drying technique) has also been reported [33]. The present paper reports new data and the impact of different drying methods in ettringite-based matrices.
2. Experimental procedure 2.1. Materials The experimental work was carried out on hardened cement paste samples, all of which were cured for a minimum of 120 days at ambient and as such can be considered to be ‘mature pastes’. The range of samples used for the study included (i) hydrated ye'elimite mixed with gypsum, (ii) laboratory synthesized C$A cement, (iii) commercial Chinese C$A cement and (iv) an OPC sample which was used as a benchmark. Apart from these multiphase clinkers, phase pure ettringite (AFt) was also synthesized. The compositions of the reactant systems are shown in Table 1. ‘Laboratory synthesized’ is defined as made in temperature controlled furnaces using laboratory-grade reactant chemicals. The specimens intended for measurement were cast in 25 × 25 mm cylinders and cured. Ye'elimite and C$A samples require higher water/binder ratios to completely hydrate and form hydration products with high water content such as ettringite. Specific values for each set of samples are explained in the following sections.
Table 1 Sample information. Sample ID
Cement composition
Water/binder ratio
YE-1
Laboratory synthesized ye'elimite with 33% added gypsum Laboratory synthesized ye'elimite with 9% added gypsum Laboratory synthesized C$A cement with 9% added gypsum Commercial Chinese C$A cement CEM I 52.5R
0.64
YE-2 Lab-C$A Com-C$A OPC
0.85 0.70 0.70 0.32
Note: percentages of gypsum are expressed per sample (YE-1 contains 67% ye'elimite and 33% gypsum).
2.1.1. Ettringite Phase-pure ettringite (C6A$3H32) was prepared according to the method reported by Balonis and Glasser [34]. Ettringite was made by mixing two solutions: the first solution from 6.65 g Al2(SO4)3·18H2O added to 100 ml ultrapure water and the second, by dispersing 4.44 g Ca(OH)2 in 250 ml of ultrapure water. The two solutions were mixed under nitrogen and then diluted to 500 ml with additional reagent water to which 0.5 ml of 1 M NaOH had been added. The mixture was sealed in a high density polyethylene (HDPE) bottle and stirred vigorously on a 60 °C hot plate for 48 h. The purity was checked using XRD. The filtered precipitate was then dried using the range of methods described subsequently. Water content calculation in ettringite samples After drying, the residual water contents of AFt samples were calculated according to Eq. (1) from Zhang and Glasser [30]: nAFt ¼ ð1248:6 31:7 18ÞðW W i Þ=18W i
ð1Þ
where nAFt is the number of water molecules left in AFt after drying (using the different methods explained in Section 2.2), Wi is the weight after ignition at 1050 °C for ~45 min and W is the weight before ignition. The results obtained are shown in Table 2. 2.1.2. Laboratory samples Ye'elimite was synthesized by mixing stoichiometric amounts of previously dried CaCO3, Al2O3 and CaSO4 and heating the mix at 1250 °C for a total of 45 min with intermediate grinding every 15 min. Purity was confirmed by XRD and weight loss measurements to ensure that SO3 was retained. The ye'elimite was mixed with laboratory grade gypsum before adding water to form a paste. Table 1 shows the two gypsum/binder and water/binder ratios chosen. The “water” in the water/binder ratio does not include water contained in gypsum. The first mix (YE-1) had gypsum/binder 0.33 and water/binder 0.64; these values being similar to the theoretical amounts needed for the formation of ettringite (gypsum/ye'elimite 0.56 and water/ye'elimite ~ 1). The composition of the second mix (YE-2), with gypsum/binder 0.09 and water/binder 0.85, was chosen to form a physical mixture of AFm and AFt (see Fig. 5 pattern before drying for details: the AFm formed was sulfate-AFm: C4A$H12). Table 2 Water content of dried AFt. These samples were unsuitable for MIP. Drying method
Drying time
H2O content
Damage (based on XRD)
Saturated solution of CaCl2 D-dried Oven dried at 40 °C Oven dried at 60 °C Oven dried at 60 °C Oven dried at 105 °C
~2 months 15 h 20 h 6h 24 h 72 h
31.8 31.5 29.2 16.6 12.7 12.1
No No No Yes Yes Yes
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Laboratory synthesized C$A (Lab-C$A in Table 1) was obtained by mixing CaCO3, Al2O3, CaSO4, Fe2O3, and SiO2 and submitting the mix to 4 cycles of 15 min each at 1300 °C with intermediate grinding between cycles. The initial mix of raw materials was calculated to form 50% ye'elimite, 25% belite and 25% ferrite. The resulting clinker mineralogy, calculated using the Rietveld method, comprised 53% ye'elimite, 21% belite, 18% ferrite, 4% tri-calcium aluminate and 4% gehlenite (C2AS). The result of the fitting is shown in Supplementary Fig. 1 of the Supplementary information. The fact that the ye'elimite contents are very similar to the expected output means that sulfur losses were negligible during the synthesis. It can also be deduced that part of the SiO2 calculated to form belite has actually formed C2AS and that part of the alumina, instead of forming ferrite, has formed C3A and C2AS. To form workable pastes with set times ~ 60 min, 9% gypsum and water were added to give a water/binder ratio of 0.7. 2.1.3. Commercial C$A samples Commercial C$A cement from Shenzhen Chenggong Building Materials Co., Ltd. PRC was used to prepare paste samples with w/c ratio 0.7. According to our Rietveld analysis, the cement contained 57% ye'elimite, 17% belite, 7% gehlenite (C2AS), 7% calcium aluminate (CA), 6% perovskite (alumina substituted CaTiO3), 3% mayenite (C12A7), 2% tri-calcium aluminate (C3A) and 1% anhydrite (see Supplementary Fig. 2 of the Supplementary information). The w/c ratio was chosen after trials to give a reasonable workability without plasticisers. 2.2. Drying methods After curing for at least 120 days in a sealed desiccator at ~99% RH, the samples were broken into pieces of 1 cm3 maximum and exposed to the different drying methods. The maximum fragment size chosen fulfils the geometry requirements of MIP measurements. The following methods have been used for drying cement and concrete samples intended for pore structural analysis techniques: i. Oven drying at selected temperatures (40 °C, 60 °C and 105 °C) for 24 h. ii. Solvent replacement, by submersion in isopropanol for 3 days at room temperature (~ 20 °C) followed by oven drying at 40 °C for 24 h: isopropanol was used to stop the hydration of the samples, followed by solvent removal in a low temperature oven (~ 40 °C) for 24 h. iii. Drying the samples in a desiccator at ~20 °C over a saturated solution of calcium chloride for sufficient time to achieve a constant weight. iv. D-drying (for varying times): D-drying uses a vacuum desiccator connected to a trap which is kept at −79 °C by a bath of solid carbon dioxide and ethanol. At this temperature the water vapour pressure is ~0.5 μm Hg [35]. This method has been proposed to avoid damage to the microstructure of cements [30]. The recommended duration required for D-drying is variable and is related to the performance of the pump used: “the rate of drying is affected greatly by the pressure in the system” [35]. Zhang and Glasser [30] suggest that Ddrying for 2–3 h is effective in drying a phase pure ettringite sample without damage on the basis of weight change. Korpa and Trettin [8] showed that D-drying for 12 h may be insufficient to dry samples (with particle sizes b1200 μm). However, they did not attempt drying for longer. Copeland and Hayes [35] suggest that 6–7 days of Ddrying may be required to achieve stable weight when the pressure is maintained at around 30 μm Hg. v. Vacuum drying at ~20 °C: the sample is placed in a desiccator and attached to a vacuum. The performance of the pump and the vacuum achieved, not always known to the user, are variables to consider. For the present work a “Speedivac” ES75 pump with no gauge was used, the minimum vacuum achievable being above ~5 μm Hg. The lack of gauge in many vacuum pumps makes it difficult for users to
87
determine exactly the level of vacuum achieved during drying. vi. Solvent replacement in acetone for 3 days at room temperature (~20 °C) followed by oven drying at 40 °C for 24 h: acetone is used to stop the hydration of the sample and replace free water, followed by drying at a low temperature, ~40 °C, to remove the acetone. For all treatments, the loss of water during drying was measured by following the mass of samples, and calculated according to Eq. (2): weight loss ð% Þ ¼
initial mass final mass 100: initial mass
ð2Þ
All samples were analysed by X-ray diffraction (XRD) before and after each drying method in order to assess the degradation of crystallinity. 2.3. Mercury intrusion porosimetry (MIP) MIP measurements were made using a Micrometrics Autopore IV 9500 instrument, which has a pore measuring range between 0.003 and 360 μm diameter. Samples were sealed in a penetrometer which is initially placed in a low pressure chamber and data recorded, followed by a high pressure chamber. The first stage of the operation of the equipment is to lower the pressure within the chamber to ~ 50 μm Hg. This stage of the operation does not involve mercury. In the second stage, the volume of mercury penetrating the sample is measured as a function of increasing pressure (P) and is related to the pore access diameter (D) of the pores penetrated by the Washburn equation (Eq. (3)): D¼
4γ cosθ P
ð3Þ
where γ is the surface tension of the mercury and θ is the contact angle of the mercury with the sample. A constant contact angle θ of 130° was assumed [30]. To check reproducibility, several sets of similar samples were measured under similar conditions. The differences in the total pore volume encountered were ~1–2%. The error of the MIP porosity values reported in this paper is thus taken as ±2%. 2.4. X-ray powder diffraction (XRPD) XRPD measurements were performed on the anhydrous C$A clinkers and on the hydrated pastes. The patterns for the C$A clinkers, both laboratory and commercial, were measured on an Empyrean diffractometer (PANalytical) using strictly monochromatic CuKα1 radiation (λ = 1.54056 Å). In order to determine the composition of the clinkers, these samples were analysed using the Rietveld methodology as implemented in the GSAS software package [36]. Final global optimised parameters were: background coefficients, zero-shift error, cell parameters and peak shape parameters. Peak shapes were fitted using the pseudo-Voigt function [37]. The patterns for the hydrated pastes, before and after drying, were measured on an X'Pert PRO diffractometer (PANalytical) using CuKα radiation. In both data sets the X-ray tube operated at 45 kV and 40 mA. 3. Results and discussion 3.1. AFt (ettringite) XRD patterns for the initially phase pure AFt dried using the various methods are shown in Figs. 1 and 2. Table 2 shows the water content remaining in AFt after drying. Damage is apparent in samples dried at 60 and 105 °C for 6 h, both becoming amorphous to XRD and with the number of water molecules being reduced to 16 and 12 respectively. In the remainder of dried AFt samples, crystallinity is preserved. This
88
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 1. XRD patterns (filtered CuKα radiation) of AFt dried under three different conditions: vacuum, D-drying and over saturated CaCl2 solution at room temperature (~20 °C). None of these methods caused damage to the AFt.
indicates that if the sample loses more than ~3 of the original 32 molecules of water, a measurable degradation of crystallinity occurs. The degradation is often inhomogeneous and it is not uncommon to observe physical mixtures of damaged and undamaged AFt. Attempts to test these samples using MIP were not successful, even in the case of samples that appeared to be damaged, according to both the XRD analysis results and loss of water molecules. The low pressure stage, up to 50 μm Hg, could not be reached, indicating that the samples were insufficiently dried.
3.2. Hydrated cement samples Table 3 shows the drying time, weight loss after drying and volume porosity measured by MIP (cut off limits were 0.003–360 μm). Where no MIP data are recorded the samples could not be measured, i.e. the vacuum necessary to commence measurement (50 μm Hg) could not be attained. “Drying time” normally means that constant weight was achieved. However in a few experiments the drying process was deliberately stopped before stable weight was reached to check if MIP permeation could be achieved. Not all methods were applied to all samples mainly for operation reasons. Also in some samples some of the techniques were not considered relevant.
3.2.1. YE-1 results The mix composition of samples YE-1 used stoichiometric quantities of ye'elimite, gypsum and water to obtain ettringite.
For this sample, all drying methods except oven drying at 105 °C were shown to preserve at least partially the mineralogy of the sample (Fig. 3). However ettringite may have partially decomposed to poorly crystallised products such as AFm, aluminium hydroxide (gibbsite) and an anhydrite-like phase. Decomposition is inhomogeneous, giving mixtures of unaffected ettringite and its decomposition products. Since XRD mainly detects the undamaged portion, crystallinity appears to persist. Not surprisingly the porosities (Table 3) and MIP curves (Fig. 4a and b) differ significantly for each drying method. Oven drying produced the highest porosity, while isopropanol replacement produced the least. Even though the weight losses achieved with acetone and isopropanol were very similar, the sample dried in acetone could not be measured by MIP as the required vacuum, ~50 μm Hg, could not be attained, implying that the sample was insufficiently dry for the test: either water or acetone, or a mixture of both, was retained as pore blockers. The additional volume intrusion spike at ~ 20 μm for samples dried at 105 °C is a noteworthy feature that may also be a proof of damage caused by drying, i.e., pore opening or cracks or both occurred during drying. For this set of samples, a clear correlation exists between weight lost and intruded pore volume (Fig. 12).
3.2.2. YE-2 results The XRD patterns of YE-2 samples are shown in Fig. 5. The only drying treatment by which the sample was not damaged was vacuum treatment but, as can be seen in Table 3, MIP was not possible because drying was incomplete. For YE-2, the higher the weight loss during drying, the higher the porosity measured by MIP (Fig. 12). In the D-dried
Fig. 2. AFt samples dried in an oven at 40, 60 and 105 °C. Note the partial (at 60 °C) and complete loss (at 105 °C) of crystallinity in the latter two samples.
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
89
Table 3 Weight loss after drying and intruded pore volumes. Sample ID
Drying method
Drying time
Weight loss (%)
Mercury porosity (%)
YE-1
Oven at 105 °C D-drying Isopropanol + oven at 40 °C Acetone + oven at 40 °C Oven at 105 °C Oven at 60 °C D-drying Acetone + vacuum Vacuum Oven at 105 °C D-drying Isopropanol + oven at 40 °C Saturated solution of CaCl2 Acetone + oven at 40 °C D-drying Vacuum Oven at 105 °C Oven at 60 °C Isopropanol + oven at 40 °C Saturated solution of CaCl2 D-drying D-drying Acetone + oven at 40 °C Vacuum Oven at 105 °C Isopropanol + oven at 40 °C D-drying Vacuum
24 h ~13 h 24 h 24 h 24 h 24 h ~50 h ~30 h ~60 h 24 h 12 h 24 h 12 days 24 h 6h 6h 24 h 24 h 24 h ~1.5 months 6h 23 h 24 h 45 h 24 h 24 h 18 h 6h
23.9 14.7 2.7 2.9 26.9 23.9 20.9 10.1 8.5 27.5 17.1 7.9 8.1 7.0 9.8 7.7 26.1 23.1 9.3 13.3 9.4 17.1 9.7 14.2 13 6.4 4.3 –
45.3 40.8 26.6 No MIP 40.5 38 35.1 No MIP No MIP 38 32.4 29.9 24.7 24.3 No MIP No MIP 34.1 32.8 38 29.3 29.6 24.5 34.3 No MIP 12 9.5 13.2 8.1
YE-2
Com-C$A
Lab-C$A
CEM I
sample the reflections of ettringite are still visible but are significantly reduced in intensity. In the oven dried (105 °C) sample ettringite has decomposed and only AFm can be detected by XRD. Considering less harsh drying conditions, MIP data (Fig. 6a and b) show that D-dried samples are less affected than after oven drying at 60 °C. Although most of the pore volume is in the same diameter ranges, it is clear that the drier the sample, the greater the accessible porosity. This behaviour was also noted in the sample YE-1 dried at 105 °C, much extra porosity appearing in the range 10–20 μm. The occurrence of the peak at 0.01 μm in the log differential curves of YE-2 (Fig. 6a), which does not show in the YE-1 sample, may be attributed to the different microstructure of the samples, YE-2 forming more AFm due to the higher water and lower gypsum than YE-1. 3.2.3. Commercial C$A From the XRD patterns (Fig. 7), the sample dried at 105 °C is badly damaged after drying: AFt, abundant in the undried sample, has been destroyed. The other methods gave well defined reflections for other
Fig. 4. a. Log differential intrusion curves for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64). The duration of drying treatments is specified in Table 3. b. Cumulative intrusion curves for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64) in ml per gramme of sample after drying treatment.
hydrates: strätlingite, AFt and AFm. It is noticeable that even though the samples were cured for at least 120 days, there was still significant residual unhydrated belite, the persistence of which may have an impact on water losses and pore volumes. The MIP curves (Fig. 8a and b) can be divided into two groups. The first group, which includes the samples dried at 105 °C, D-dried for 12 h and with isopropanol, shows a pronounced intruded volume maximum between 0.6 and 1 μm. The second group, which includes the samples dried with acetone and over CaCl2, shows a broad pore spectrum between ~0.01 and 1 μm. These last two samples needed a much longer vacuum treatment to measure MIP than the samples from the first group. The time taken for the samples to achieve the required vacuum, ~ 50 μm Hg, can in some cases also be damaging. The first group of samples shows a correlation between weight loss and pore
Fig. 3. XRD patterns for YE-1 sample (ye'elimite hydrated with 33% gypsum and w/c = 0.64) before and after drying.
90
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 7. XRD patterns for commercial C$A (w/c = 0.7). St: strätlingite, AFm: monosulfoaluminate, AFt: ettringite, β-C2 S: β-belite. Fig. 5. XRD patterns for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85). AFm stands for sulfate-AFm, monosulfoaluminate or C4A$Hx and OH-AFm stands for hydroxy-AFm or C4AHx.
volume (Fig. 12). The sample dried at 105 °C was the most damaged, according to both XRD and MIP results. D-drying and isopropanol give very similar results. The samples dried over CaCl2 and with acetone are not sufficiently dry to perform reliable MIP measurements. For this type of samples D-drying and isopropanol replacement would be the recommended methods.
3.2.4. Lab-C$A results The XRD patterns of the laboratory C$A samples are shown in Figs. 9 and 10 and the corresponding MIP curves in Fig. 11a and b. The 3 samples shown in Fig. 9, dried with acetone, isopropanol and D-dried for 6 h appeared to be undamaged by XRD and lost only around 9% weight during drying. But these samples show a peak of porosity at ~ 20 μm (Fig. 11a and b) which does not occur for samples that were drier, according to the weight losses, prior to intrusion. For these 3 samples, dried in acetone, isopropanol and D-dried for 6 h, the first
Fig. 6. a. Log differential intrusion curves for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85). b. Cumulative intrusion curves for YE-2 sample (ye'elimite hydrated with 9% gypsum and w/c = 0.85).
Fig. 8. a. Log differential intrusion curves for commercial C$A (w/c = 0.7). b. Cumulative intrusion curves for commercial C$A (w/c = 0.7).
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
91
Fig. 9. XRD patterns of laboratory C$A dried by solvent exchange (isopropanol and acetone) and by D-drying for 6 h.
stage of MIP measurement took significantly longer than for other samples dried with vacuum, or D-drying for 23 h or oven dried at either 60 or 105 °C. The appearance of these larger pores may be due to damage arising from the first stage of the MIP operation where mercury is not involved and where the pressure is lowered to vacuum state (50 μm Hg). A proof of this is the fact that a ‘spike’ appears for all three samples at approximately the same diameter (~20 μm) suggesting that this is a damage-related feature. This does not mean D-drying for 23 h is less damaging than D-drying for 6 h (as shown by the XRD): it shows that 6 h of D-drying is not enough to perform reliable MIP measurements. The MIP vacuum stage will damage the sample more than the drying process if the sample is not dry enough. The sample D-dried for 23 h presents the lowest intruded volume, as measured by MIP, indicating that this drying method was the less aggressive but sufficient to dry the sample. Moreover the vacuum needed
Fig. 10. XRD patterns of laboratory C$A dried under vacuum, D-dried for 23 h, and oven dried at 60 and 105 °C.
Fig. 11. a. Log differential intrusion curves for laboratory C$A cement with 10% gypsum (hydrated at w/c = 0.7). b. Cumulative intrusion curves for laboratory C$A cement with 10% gypsum (w/c = 0.7). As explained in the text, the samples dried in acetone, isopropanol and D-dried for 6 h seem to have been damaged during the first stage of MIP: notice the spike at ~20 μm because they were not sufficiently dry. Samples D-dried for 23 h and oven dried at 60 and 105 °C do not show this spike.
for MIP did not further damage it, as occurred in “wet” samples following either solvent replacement (acetone, isopropanol) or D-drying for 6 h. Excluding the samples which were insufficiently dried or had been damaged during vacuum stage prior to MIP, there is a correlation between weight loss and pore volume (Fig. 12), although the correlation is not as good as in previous examples. 3.2.5. OPC An OPC sample was used as a benchmark; none of the drying methods used appear by XRD to cause damage to the mineralogy of the hydrated cement paste (Fig. 13). However, changes to C–S–H are not well revealed by XRD. Drying at 105 °C almost certainly causes microstructural damage, such as microcracking, and this would be revealed in the course of mercury intrusion as the increased porosity shown in Fig. 14a and b. Oven dried samples show higher porosity at all diameters than the samples dried in vacuum or isopropanol and also a shift in the peak where most pores concentrate towards greater diameters. This agrees with Moukwa and Aitcin [6], who reported pore opening in the range of 0.02 and 0.1 μm after drying cement paste samples for 24 h at 105 °C. According to these authors “oven drying can alter the pore structure mainly by rearranging the hydration products”. Also Galle [9] reported an overestimation of total porosity when drying at 105 °C attributed to significant losses of nonevaporable water from both AFt and C–S–H. Apart from the cement hydrate desiccation and the rearrangements, this later author also attributes the damage to the potential microcrack generation in relation with internal thermally-induced stresses. The D-dried sample of OPC has suffered damage during MIP measurement and the porosity at around 100 μm is tentatively attributed to that damage because it does not show in the other samples. This
92
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Fig. 12. Pore volume (ranging pores from 360 to 0.003 μm) versus weight loss after drying in samples of YE-1 (dried at 105 °C, with isopropanol and D-dried, YE-2 (dried in an oven at 60 and 105 °C and D-dried), com-C$A (dried at 105 °C, D-dried and with isopropanol) and lab-C$A (dried with vacuum, D-dried for 23 h and in an oven at 60 and 105 °C).
sample lost less weight during drying but had higher pore volume and increase of porosity at all diameters greater than ~ 0.05 μm after the drying + MIP process, again indicating damage during MIP measurement. According to Korpa and Trettin [8] a short duration D-drying can lead to incomplete water removal, mainly from the very small gel pores, b2 nm, and will not give realistic values for gel nanopores. Samples subjected to vacuum and isopropanol drying show very similar results: total porosity was between 8 and 9% pore volume and most of the pores had diameters in the region 0.01 to 0.1 μm. Again there is a correlation between pore volume and weight loss, excluding the sample damaged during D-drying, where the pore volume is higher due to the pores created during the drying process. Collier et al. [12] used acetone replacement (7 day immersion followed by 10−2 mbar vacuum for 3 days) and vacuum (10−2 mbar vacuum for 7 days) for drying hydrated OPC samples and reported that none of the techniques caused any major deterioration in the composition and microstructure of the samples. Similar conclusions were drawn by Snoeck [15], who reported that the unavoidable changes in the microstructure of hydrated OPC samples due to drying are minimized when using vacuum drying and isopropanol exchange. This later method is also recommended by Zhang and Scherer [14] as the best “for preserving the microstructure with minimal effects on the composition of the cement”. The MIP data reported for OPC in the literature are similar to the ones obtained here. Konecky and Naqvi [13] reported on mortar samples dried in an oven at 105 °C and by solvent replacement. The oven dried ones showed most of the pores having diameters between 0.05
and 0.2 μm whereas the solvent dried ones were in the range 0.02– 0.1 μm. Galle [9] reported that most of the pore diameter concentrated between 0.03 and 0.15 μm in oven dried OPC pastes and between 0.02 and 0.08 μm in vacuum dried OPC pastes. The total porosity in these samples is slightly higher, 14–15%, than in the ones reported in the title study. Absolute comparison of reported values and those of the title study is not possible because the samples and their conditioning are not identical, but it is worth noting that in comparative terms, there is general agreement on the effects of the different drying methods on OPC paste pore volume and pore distribution. Finally, if we compare the MIP values for OPC with those obtained for C$A matrices, a shift towards greater pore diameters is observed in the differential intrusion curves of C$A. In OPC samples the majority of the pores concentrate on diameters between 0.02 and 0.06 μm whereas in C$A, they range from 0.02 to 0.2 μm in the laboratory C$A, and from
Fig. 13. XRD patterns of OPC samples before and after drying using vacuum, isopropanol exchange, oven at 105 °C and D-drying. P: portlandite, A: alite, B: belite.
Fig. 14. a. Log differential intrusion curves for OPC (w/c = 0.32). b. Cumulative intrusion curves for OPC (w/c = 0.32).
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
0.3 to 1 μm in the commercial C$A. The total pore volume is also much smaller in the OPC, 8–9%, than in the C$A samples, 25–30%. This could be in part explained by the different water/cement ratio used, which was much lower for OPC than for C$A. But, given the intrinsic difference in water demand between the two types of cement, the differences in pore sizes and pore volume seem to be more related to the microstructure of the pastes than the amounts of free water. 4. Additional discussion The study confirms that AFt is readily damaged by drying. Zhang and Glasser [30] suggested that the crystallinity of AFt begins to degrade at b30 molecules of water per formula unit. This is in accord with our results (see Section 3.1) obtained on the sample dried at 40 °C; the number of water molecules in AFt was reduced from 32 to ~ 29 and it remained fully crystalline by XRD. However, as more water is lost, the AFt loses water and degrades to a plateau at ~ 12H2O giving rise to a phase termed meta-ettringite [32]. Mixtures of crystalline and metaettringite are commonly encountered: AFt reflections can still be identified at low water contents but the crystalline reflections reduce in intensity as degradation proceeds. AFt also converts in part to AFm (Figs. 5 and 7), going from 32 to ~12H2O molecules per unit formula. With progressive drying, damage to AFm is also noticeable as reduction in the reflection intensity in XRD patterns and the shift of reflections towards increasing 2θ angles, indicating a contraction in basal spacings with formation of lower water SO4-AFm types (Fig. 5), like those reported by [33]. It is clear that drying at ~100 °C severely damages ettringite-based matrices. Drying at ~ 60 °C is also damaging unless the drying time is kept short, less than ~ 5 h, meaning that a stable weight may not be reached: continuing evolution of water vapour protects the crystallinity of at least some of the residual AFt. Drying at 40 °C does not damage ettringite or C$A based materials, ettringite being stable at all but extremely low humidity. Although none of the vacuum dried samples from ye'elimite/C$A based cements used in this study were dry enough for MIP, the OPC sample dried for only 6 h using vacuum was sufficiently dried for MIP. The actual pressures achieved were not recorded but as the same pump was used for all vacuum-treated sample, a relative comparison between results can be made. While vacuum drying has been used in the past for OPC based samples, it does not seem appropriate for drying C$A based cements. As shown in the results, samples left drying under vacuum achieved constant weight but were still insufficiently dry to perform MIP (Table 3). The low pressure evacuation stage needed to perform MIP measurements can in some cases cause more damage than the actual drying process. This step is unavoidable as pores need to be empty enough to perform the high pressure stage where mercury is introduced in the pores. This was proven in samples apparently not damaged by the drying process that then give spikes (Fig. 11) or ‘bumps’ (Fig. 8) in their MIP curves. In these cases, as the samples were not sufficiently dry, the MIP was actually more damaging than the drying. The total values of pore volumes and weight losses of laboratory C$A versus commercial C$A are comparable and in the same range (Table 3). But pore sizes and distributions differ (Figs. 8 and 11). In the ye'elimite samples most pores have diameters of ~0.3–0.4 μm. In the laboratory C$A sample the majority of pores are slightly smaller, ~ 0.1–0.2 μm, but in the commercial C$A the ‘peak’ is shifted to more coarse porosity: 0.7–0.8 μm. These differences can be attributed to the residual unhydrated solid and the different proportion of hydrated phases formed and their densities: strätlingite, Aft, AFm and aluminium hydroxide. In the commercial C$A the amount of strätlingite is greater than in the laboratory C$A, where AFt is the predominant hydrated phase. The use of simulants for C$A such as pure ettringite or mixtures of phases, which were then hydrated, is useful to understand porosity in simpler systems, but simulation of porosity is not always
93
straightforward. The use of simulants as proxies for commercial cements should be done carefully keeping in mind all the extra parameters not taken into account, such as matrix effects and possible solid solutions formed and their effect on the porosity distribution. Finally, we note that the literature contains numerous papers pointing out that the intrusion process is damaging and affects the apparent pore size spectrum. We agree, but also note that the preliminary drying, required to apply many pore-sensitive techniques (MIP, BET), is the source of much damage. Indeed, it is possible that the drying damage may render the solid more liable to further damage in the course of intrusion. And, since drying underlies some of the techniques we use to characterize cement paste (Scanning Electron Microscopy, MIP, thermogravimetry), it is this first step that deserves critical scrutiny. In an attempt to reduce dependency on MIP, we are developing scale-spanning CT methods that are non-intrusive and can in principle be applied to water saturated samples. But the resolution of the refinements and the complex mathematics required to re-image the pore structure and its continuity are as yet insufficiently robust to make a quantitative approach. 5. Conclusions The main conclusions that can be drawn from the results presented are as follows: – An unexpected problem in determining damage is to establish a benchmark state. The hydration state of AFm phase(s) is sensitive to temperature and humidities at or near typical ambient laboratory conditions, 15–30 °C and ranges of relative humidities. Changes in this range affect the intrinsic mineralogy and especially, pore structure. – The extent to which damage is caused by changes in the constitution of C–S–H is difficult to characterize and depends on definitions of “free” and “bound” water in a particular specimen. – AFt retains its crystallinity when dried in an oven at 40 °C. This means that the drying associated with solvent removal, often done at 40 °C, is unlikely to contribute to damage. – C$A cements are more susceptible to drying damage than Portland cements on account of their AFt content relative to Portland cements. Mineralogical damage occurs to C$A samples exposed to prolonged drying times and harsh drying methods. Drying at 100 °C, still common in some engineering proceedings, will always damage cement paste. – D-drying and isopropanol exchange are suggested as conservative methods for drying C$A based samples. Times for D-drying may vary among samples and as a function of the vacuum achieved by the pump, and are not always easy to optimise. – MIP of cement matrices based on ettringite is particularly susceptible to damage. – In samples that are not dry enough for MIP, the first stage of vacuum needed to start the measurement, ~50 μm Hg, can cause substantial damage to the samples, increasing the porosity between 10 and 100 μm. Most previous concerns have been about damages resulting from the physical intrusion but previous drying treatment is partially responsible for artefacts. – None of the available methods to measure porosity and/or permeability are completely satisfactory. However the search for new methods should continue. BET is an obvious target for inclusion as it works well for pores in the size range below 0.1 mm. X-ray tomography is being used on comparable samples but its resolution and discrimination are as yet probably not sufficient to measure sizes and volumes of pores b 1 μm.
Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.cemconres.2016.03.003.
94
I. Galan et al. / Cement and Concrete Research 84 (2016) 85–94
Acknowledgments IG, HB, FPG and MSI gratefully acknowledge the financial support provided by the Gulf Organization for Research and Development (GORD), Qatar, through research grant number ENG016RGG0593. MGM acknowledges the financial support provided by the Junta de Andalucía through research grant number P11-FQM-07517. References [1] S. Diamond, Mercury porosimetry. An inappropriate method for the measurement of pore size distributions in cement-based materials, Cem. Concr. Res. 30 (2000) 1517–1525. [2] S. Diamond, E. Landis, Microstructural features of a mortar as seen by computed microtomography, Mater. Struct. 40 (2007) 989–993. [3] E. Gallucci, K. Scrivener, A. Groso, M. Stampanoni, G. Margaritondo, 3D experimental investigation of the microstructure of cement pastes using synchrotron X-ray microtomography (μCT), Cem. Concr. Res. 37 (2007) 360–368. [4] M. Zhang, Y. He, G. Ye, D.A. Lange, K.V. Breugel, Computational investigation on mass diffusivity in Portland cement paste based on X-ray computed microtomography (μCT) image, Constr. Build. Mater. 27 (2012) 472–481. [5] S. Yoon, I. Galan, K. Celik, F.P. Glasser, M.S. Imbabi, Characterization of micro-pore structure in novel cement matrices, Mater. Res. Soc. Symp. Proc. 2014 (1712). [6] M. Moukwa, P.-. Aitcin, The effect of drying on cement pastes pore structure as determined by mercury porosimetry, Cem. Concr. Res. 18 (1988) 745–752. [7] H. Taylor, C. Famy, K. Scrivener, Delayed ettringite formation, Cem. Concr. Res. 31 (2001) 683–693. [8] A. Korpa, R. Trettin, The influence of different drying methods on cement paste microstructures as reflected by gas adsorption: comparison between freeze-drying (Fdrying), D-drying, P-drying and oven-drying methods, Cem. Concr. Res. 36 (2006) 634–649. [9] C. Gallé, Effect of drying on cement-based materials pore structure as identified by mercury intrusion porosimetry — a comparative study between oven-, vacuum-, and freeze-drying, Cem. Concr. Res. 31 (2001) 1467–1477. [10] R.M. Espinosa, L. Franke, Influence of the age and drying process on pore structure and sorption isotherms of hardened cement paste, Cem. Concr. Res. 36 (2006) 1969–1984. [11] C. Gosselin, Microstructural Development of Calcium Aluminate Cement Based Systems With and Without Supplementary Cementitious MaterialsPhD thesis EPFL, 2009. [12] N.C. Collier, J.H. Sharp, N.B. Milestone, J. Hill, I.H. Godfrey, The influence of water removal techniques on the composition and microstructure of hardened cement pastes, Cem. Concr. Res. 38 (2008) 737–744. [13] L. Konecny, S.J. Naqvi, The effect of different drying techniques on the pore size distribution of blended cement mortars, Cem. Concr. Res. 23 (1993) 1223–1228. [14] J. Zhang, G.W. Scherer, Comparison of methods for arresting hydration of cement, Cem. Concr. Res. 41 (2011) 1024–1036. [15] D. Snoeck, L.F. Velasco, A. Mignon, S. Van Vlierberghe, P. Dubruel, P. Lodewyckx, N. De Belie, The influence of different drying techniques on the water sorption properties of cement-based materials, Cem. Concr. Res. 64 (2014) 54–62.
[16] R.F. Feldman, J.J. Beaudoin, Pretreatment of hardened hydrated cement pastes for mercury intrusion measurements, Cem. Concr. Res. 21 (1991) 297–308. [17] J.J. Beaudoin, B.T. Tamtsia, Effect of drying methods on microstructural changes in hardened cement paste: an A. C. impedance spectroscopy evaluation, J. Adv. Concr. Technol. 2 (2004) 113–120. [18] J.J. Beaudoin, P. Gu, J. Marchand, B. Tamtsia, R.E. Myers, Z. Liu, Solvent replacement studies of hydrated Portland cement systems: the role of calcium hydroxide, Adv. Cem. Based Mater. 8 (1998) 56–65. [19] J.J. Beaudoin, B. Tamtsia, J. Marchand, H.R. Myers, Solvent exchange in partially saturated and saturated microporous systems length change anomalies, Cem. Concr. Res. 30 (2000) 359–370. [20] R.L. Day, Reactions between methanol and Portland cement paste, Cem. Concr. Res. 11 (1981) 341–349. [21] E. Gartner, Industrially interesting approaches to “low-CO2” cements, Cem. Concr. Res. 34 (2004) 1489–1498. [22] J.H. Sharp, C.D. Lawrence, R. Yang, Calcium sulfoaluminate cements — low-energy cements, special cements or what? Adv. Cem. Res. 11 (1999) 3–13. [23] C.D. Popescu, M. Muntean, J.H. Sharp, Industrial trial production of low energy belite cement, Cem. Concr. compos. 25 (2003) 689–693. [24] A. Feraille, A. Alaoui, A. Steckmeyer, R. Le Roy, New cements for sustainable development, Proceedings of the 12th ICCC, Montreal, Canada., 2007. [25] S. Sahu, J. Majling, Phase compatibility in the system CaO·SiO2·Al2O3·Fe2O3·SO3 referred to sulfoaluminate belite cement clinker, Cem. Concr. Res. 23 (1993) 1331–1339. [26] K. Quillin, Performance of belite–sulfoaluminate cements, Cem. Concr. Res. 31 (2001) 1341–1349. [27] F.P. Glasser, L. Zhang, High-performance cement matrices based on calcium sulfoaluminate–belite compositions, Cem. Concr. Res. 31 (2001) 1881–1886. [28] C. Cau Dit Coumes, S. Courtois, S. Peysson, J. Ambroise, J. Pera, Calcium sulfoaluminate cement blended with OPC: a potential binder to encapsulate lowlevel radioactive slurries of complex chemistry, Cem. Concr. Res. 39 (2009) 740–747. [29] Q. Zhou, N.B. Milestone, M. Hayes, An alternative to Portland cement for waste encapsulation—the calcium sulfoaluminate cement system, J. Hazard. Mater. 136 (2006) 120–129. [30] L. Zhang, F.P. Glasser, Critical examination of drying damage to cement pastes, Adv. Cem. Res. 12 (2000) 79–88. [31] Q. Zhou, F.P. Glasser, Thermal stability and decomposition mechanisms of ettringite at b120 °C, Cem. Concr. Res. 31 (2001) 1333–1339. [32] Q. Zhou, E.E. Lachowski, F.P. Glasser, Metaettringite, a decomposition product of ettringite, Cem. Concr. Res. 34 (2004) 703–710. [33] L.G. Baquerizo, T. Matschei, K.L. Scrivener, M. Saeidpour, L. Wadsö, Hydration states of AFm cement phases, Cem. Concr. Res. 73 (2015) 143–157. [34] M. Balonis, F.P. Glasser, The density of cement phases, Cem. Concr. Res. 39 (2009) 733–739. [35] L.E. Copeland, J.C. Hayes, The determination of non-evaporable water in hardened Portland cement pastes, ASTM Bulletin No 194, 1953. [36] A.C. Larson, R.B. Von Dreele, General Structure Analysis System (GSAS), Los Alamos Laboratory, Rep.No.LA-UR, Vol. 862004. [37] P. Thompson, D.E. Cox, J.B. Hastings, Rietveld refinement of Debye–Scherrer synchrotron X-ray data from Al2O3, J. Appl. Crystallogr. 20 (1987) 79–83.