PCA R&D Serial No.

2925

Volumetric Measurement in Water Bath an Inappropriate Method to Measure Autogenous Strain of Cement Paste
by Pietro Lura and Ole Mejlhede Jensen
BYG DTU - Department Civil Engineering Technical University of Denmark Brovej DTU - Building 118 DK-2800 Kgs. Lyngby Tel: +45 45 25 17 58 Fax: +45 45 88 32 82

Portland Cement Association 2005 All rights reserved

Volumetric measurement in water bath - an inappropriate method to measure autogenous strain of cement paste
Technical report

Pietro Lura and Ole Mejlhede Jensen Technical University of Denmark, Department of Civil Engineering, February 2005

Abstract Volumetric measurements of autogenous strain are frequently performed by placing the fresh cement paste in an elastic rubber membrane immersed in water. The change in volume of the cement paste is measured by the amount of water displaced by the immersed sample, for example by measuring its weight change underwater. Volumetric and linear measurements of autogenous strain should in principle give identical results. However, the measuring results from the volumetric method are typically 3-5 times the results from the linear technique, depending on the type of cement paste and the experimental conditions. Some sources of error in the volumetric method have been identified by different authors and eliminated or compensated for. In this study, absorption of water from the buoyancy bath through the rubber membrane is identified as the principal artifact of the volumetric method and cause of most of the discrepancies between volumetric and linear measurements. Water absorption is driven by a lowering of the water activity in the cement paste due to self desiccation and to dissolved salts in the pore fluid. Very high water absorption in the membrane-enclosed cement paste samples was registered from the moment of casting. The influence of this error on the volumetric measurements was of the same order of magnitude as the autogenous strain itself. By performing the measurements in a paraffin oil bath instead of a water bath, water absorption through the elastic membrane was eliminated. Volumetric measurements performed in paraffin oil were almost identical to linear measurements performed on the same cement pastes. This shows that the principal artifact of the volumetric method was individuated and eliminated. Nevertheless, the presence of many other artifacts makes the volumetric method very difficult to use in practical applications.

Introduction The title of this report makes explicit reference to a notable paper by Sidney Diamond [1], where the author thoroughly discussed the applicability of the mercury intrusion porosimetry technique for cement based materials. In a similar way, the present paper aims at discussing the volumetric technique for autogenous strain measurements in the form it has taken in the recent years: measurements of the underwater weight of an elastic membrane filled with cement paste or mortar. An alternative to this method is to monitor the water level in a capillary tube, which measures the water displaced by the membrane filled with cementitious material. Literature review Measurements of autogenous volumetric strain in the early literature were performed with different principles. In 1928, Neville and Jones [2] described an apparatus for measurement of volumetric strain of cement paste during sealed hardening at constant temperature: the cement paste was cast in a deformable petrolatum mould and topped with paraffin oil, whose change of level in a capillary tube indicated the volume change of the cement paste. In a similar way, Del Campo [3] poured fresh cement paste on an elastic membrane and used mercury as a displacement liquid. Slate and Matheus [4] inserted fresh cement paste or concrete in a container filled with a mineral oil and monitored the change of oil level in a vertical capillary tube. The fundamental elements of the volumetric method as it is used today were already present in the work of Yates in 1941 [5]: here a rubber membrane containing cement paste is suspended in a waterfilled container and the level of water in a capillary tube is measured up to 1 year. A similar method was used for concrete by Wuerpel [6] and for de-aired cement paste by LHermite and Grieu [7]. Also Edmeades et al. [8] used a similar method for cement paste, but readings were taken on a horizontal mercury capillary tube. De Haas et al. [9] measured instead the weight of a beaker that exchanged water with the bath when the volume of the specimen changed. Yamazaky et al. [10] measured the underwater weight of the sample enclosed in a rubber membrane. In 1978, Setter and Roy [11] presented volumetric measurements on cement pastes based on a method similar to de Haas et al. [9]. A very poignant discussion of their paper was written by Baron and Buil [12], based on Buils doctoral thesis work [13], which was perhaps the first international debate on measuring techniques of autogenous strain. Baron and Buil [12] compared a volumetric measurement, by means of the previously described apparatus by Del Campo [3], and a linear one, showing that the strain values in the volumetric technique were one order of magnitude greater. Baron and Buil [12] pointed out two principal artefacts in the volumetric technique: 1) In the first hours after setting, the volumetric method may in reality measure the chemical shrinkage because of reabsorption of a layer of bleeding water. Elimination of the bleeding water with a syringe [11] may not be complete; 2) The pressure exerted by the measuring liquid and by the elastic membrane causes contraction of the paste, especially around setting when it is weakest. The method used in [11] will most likely also be influenced by absorption of water through the membrane. This artefact was not mentioned by Baron and Buil [12], possibly because the apparatus by Del Campo [3] did not suffer from this shortcoming.

In the early 1980ies Ziegeldorf and co-workers [14, 15] cast fresh cement paste into 0.4-mm thick polyethylene bags, which were immersed in a container filled with de-aired water, whose level was monitored. They observed that the measured volumetric strain critically depended on bleeding and varied greatly with specimen size and shape. The volumetric technique for autogenous strain measurements received renewed attention from the mid 1990ies in a series of papers by Justnes, Sellevold and co-workers [16-22], who used measurements of underwater weight of rubber membranes (latex condoms) filled with cement paste. The known artefact due to the reabsorbtion of bleeding water was addressed by rotating the samples. Most of the following papers by different authors on the volumetric technique make explicit reference to this particular version of the volumetric method and use it in a similar form [23-35]. An exception is the apparatus devised by Gagn et al. [36], where the sample is contained in a rubber membrane sustained by a perforated metallic sheet and immersed in a water bath; the volumetric autogenous strain is registered by level changes on a graduated capillary tube. A debate about the appropriateness of the volumetric and of the linear technique started again in the 1990ies [37, 18, 20], but comparison of data on the same materials and systematic analysis of artefacts of both techniques was still missing. Both a critical approach and a deep investigation of the measuring techniques are present in Barcelo et al. [38]. Barcelo et al. [38] performed volumetric measurements of autogenous strain on rotated samples, similar to the procedure by Justnes et al. [17], but monitoring continuously the level of the displacement liquid. Comparison of volumetric and linear measurements on the same paste revealed that, even after 1 d, the volumetric autogenous strain rate was about 3 times the linear. Differences before and around setting could be explained by a number of reasons, including temperature differences, pressure of the membrane in the volumetric method, anisotropy of autogenous strain due to settlement, and friction in the linear method. The discrepancies after a few days were instead much more difficult to explain. Barcelo et al. [38] hypothesized that autogenous strain might not be isotropic even after setting; a second hypothesis was that underpressure in the rubber membrane, which was isolated from external pressure, might increase the shrinkage in comparison with the linear measurements. The hypothesis of anisotropy of autogenous strains after setting was later proven wrong: Charron et al. [39] found that the linear autogenous strain of 2 mortars on 50-mm prisms was identical in 3 different directions. Nevertheless, the linear autogenous shrinkage rate was 1/10 of the volumetric one after setting. As reasons for the discrepancies, they suggested the internal underpressure in the volumetric specimen or the suction of the elastic membrane into the air voids at the surface of the sample. Osmosis through the membrane One fundamental reason for the inconsistency between volumetric and linear autogenous strain measurements after setting is transport of water through the rubber membrane, occurring when the buoyancy liquid used is water [40]. Penetrated water may partially fill the internal voids produced by chemical shrinkage, causing an increase in the submerged weight that is finally interpreted as volumetric shrinkage. On the other hand, water penetration reduces the shrinkage of the sample or in extreme induces swelling. It is important to emphasize that the change in immersed weight due to water absorption overrides the buoyancy change due to shrinkage of the paste.

Rubber membranes used in the volumetric method are not waterproof. For example, condoms filled with water and exposed to air at low relative humidity lost about 0.5% of their water content per day [41, 42]. According to some authors [41, 43], the influence of the permeability of the membrane on the volumetric measurements would be insignificant until self-desiccation occurs in the cement paste, a few days after casting. Self-desiccation would then cause a relative-humidity gradient across the membrane and drive water from the water bath into the sample. However, significant water transport through the condom does not start some days after casting, but instead occurs right from the beginning. A difference in water activity across the membrane exists from the very start of the measurements due to dissolved salts in the pore fluid [40, 44]. This difference in water activity constitutes a driving force for osmosis of water through the membrane into the paste. To show this effect, Marciniak [42] measured the water absorption of condoms filled with synthetic pore solution, immersed in a distilled water bath. The weight of the condoms filled with pore solution increased by 0.1 % per day, while condoms filled with distilled water did not show any appreciable weight change. Beltzung and Wittmann [27] observed a 0.02-0.05% mass increase of cement paste samples immersed in water (w/c 0.29) in the first 3 d and attributed this phenomenon to water diffusion through the latex membrane due to osmosis. Douglas and Hover [45, 46] also measured a substantial weight change of condoms filled with cement paste and mortars immersed in a water bath. They excluded osmosis as a driving mechanism [46], after observing no relevant mass change of elastic membranes filled with a lime-saturated solution immersed in a water bath. However, the composition of the pore fluid in a cement paste is very different from a saturated lime solution [47]. Barcelo [48] tested different latex membranes and observed that all allowed water transfer; as driving force of the water absorption, he suggested the reduced air pressure in the capillary system of the cement paste hydrating in the elastic membrane. Aim of study This study aims at investigating the phenomenon of water absorption into the samples in the volumetric method. In particular, its driving mechanism(s) needs to be ascertained and its influence on the measured strain quantified. The final aim is to compensate for the absorption, if possible, or instead to modify the volumetric measurement as to exclude this artifact. To verify the results of the volumetric measurements, linear autogenous strain was measured on the same cement pastes with a technique based on corrugated moulds [37]. In this research, widely different cement pastes were used, which had previously been investigated with the linear technique [49]. The choice of the pastes was done to cover a range of different deformational behaviors: 2 high performance cement pastes with w/c ratio 0.30 and 20% silica fume addition, one with significant self-desiccation and high initial shrinkage, the other with initial expansion due to the addition of superabsorbent polymers, SAP; finally, a plain cement paste of w/c 0.35, which showed initial expansion followed by moderate shrinkage.

Materials Synthetic pore solution Synthetic pore solutions were produced based on the compositions reported by Page and Vennesland [47]. The composition of the synthetic pore fluid was: 0.4 mol/l Na+, 0.4 mol/l K+, 0.002 mol/l Ca2+, 0.724 mol/l OH- and 0.04 mol/l SO42-. This was obtained by mixing 16.00 g of NaOH, 6.97 g K2SO4, 17.95 g KOH and 0.15 g Ca(OH)2 in 1 l of demineralized water. Another solution with half the ionic strength was obtained by dilution with 1 l of demineralized water. Cement pastes The following cement paste compositions were investigated: A: w/c=0.30, 20% silica fume addition; B: w/c=0.30, 20% silica fume addition, 0.6% superabsorbent polymer, SAP, by weight of cement; C: w/c=0.35, 0% silica fume. The cement used is a low-alkali Danish white Portland cement produced by Aalborg Portland. Blaine fineness is 420 m2/kg and the Bogue-calculated phase composition (in wt.%) is: C3S: 66.1, C2S: 21.2, C3A: 4.3, C4AF: 1.1, C S :: 3.5, free CaO: 1.96, Na2O eq.: 0.17. The silica fume in pastes A and B was added as a dry powder at a rate of 20 wt.% of cement. The specific surface of the silica fume is 17.5 m2/g (BET method). The chemical composition is (in wt.%): SiO2: 94.1, Fe2O3: 1.00, Al2O3: 0.13, MgO: 0.71, SO3: 0.43, and Na2O eq.: 1.09. In mixtures A and B, a naphtalene-based dry powder superplasticizer was added at a rate of 1.0 wt.% of cement+silica fume. The SAP used in mixture B are suspension-polymerized covalently cross-linked acrylamide/acrylic acid copolymers [49]. The spherical particles have diameters about 100-150 m in the dry state. The size of the swollen SAP particles in the cement pastes and mortars is about three times larger due to pore fluid absorption. The SAP were added at a rate of 0.6 wt.% of cement. In mixture B, extra mixing water was added in the amount sufficient to saturate the SAP particles. The amount adsorbed in the SAP corresponds to an entrained w/c of 0.075 [49]. The cement pastes were mixed in a 5-l epicyclic mixer [50]. Cement and all other admixtures, included the SAP in mix B, were put into the bowl. Mixing was at low speed for 1 minute, while gradually adding about 3/4 of the demineralised water. Mixing continued at high speed for 1 minute, after which it was stopped and the paste was scraped off from the blade and the walls of the bowl for 1 minute. Mixing was resumed for 1 minute at low speed while the rest of the water was added; a last minute at high speed followed. The water was added in two steps to assure the homogeneity of the mix and the dispersion of the silica fume. Total mixing time from first water addition was 5 minutes. The temperature of the ingredients was approximately 20C at mixing.

Methods Autogenous relative humidity measurements Autogenous relative humidity (RH) of cement paste was measured by Rotronic Hygroscope DT (Rotronic, Basserdorf, Switzerland) stations equipped with WA-14TH and WA-40TH measuring cells, which were built into a thermostatically controlled box (0.1C). Before and after every experiment, the equipment was calibrated with saturated salt solutions in the range 75100% RH. Autogenous RH change was measured simultaneously on two identical samples. A thorough description of the Rotronic Hygroscope and the sample preparation is published elsewhere [51]. Autogenous relative humidity was measured in this research for two principal reasons: 1) As shown in previous research, the measured RH drop is related to the linear autogenous strain measured [52] and under certain conditions the measured strains can be calculated based on the internal RH development [53]; 2) The autogenous RH is identical to the water activity of the pore fluid in the cement paste [53], which is a driving force of the water transport across the elastic membrane, as will be illustrated in the following. Volumetric measurements of autogenous strain Measurements of volumetric autogenous strain were performed by monitoring the weight of cement paste samples contained in elastic membranes and immersed in a buoyancy liquid, either distilled water or paraffin oil. The experiments were done in a climate room at 200.2C. About 100 to 150 g of fresh cement paste were poured into a membrane, either a latex or polyurethane condom. The latex condoms were of type Plan by RFSU, with thickness 0.06 mm and without gliding crme or reservoir. The polyurethane condoms were of type Avanti by Durex, with thickness 0.04 mm, gliding crme and reservoir. The gliding crme was removed from the external surface of the condom with a paper towel before filling the condom with cement paste. The filled membrane was tightly closed with a knot; attention was paid to avoid entrapment of air bubbles. The excess part was then cut off and a 0.12-mm mono-filament Silicon-PTFE string (fishing line) was tied to the sample by means of a plastic strap. The string, about 400 mm long, was tied to a stainless steel hook at the other end. In the case of measurements in water, the knot of the rubber membrane was sealed with a drop of silicone before immersion in the liquid. Typical samples are shown in Figure 1.

Figure 1. Polyurethane membrane with reservoir (left) and latex membrane without reservoir (right) filled with cement paste. The picture shows also the plastic strap around the knot, to which a thin string was tied, ending with a steel hook that was used to hang the sample under the balance.

During the measurements, the samples hung from a hook beneath the balance plate. The balance was a Sartorious CP 225 D, with sensitivity 0.01 mg for weights below 80 g and 0.1 mg for weights until 7

220 g. All measurements were performed in the 220 g measuring range. The 0.1 mg accuracy in this range was fully sufficient for the measurements; in addition the stability of the readings from the balance was improved. The 0.1 mg weighing accuracy results in a nominal accuracy on the measured strain of 0.4-0.8 m/m, considering typical sample sizes, different buoyancy liquids, and isotropic deformations. The actual uncertainty on the measured strain is much higher, being dominated by a number of factors, included temperature oscillations and drift of the balance. The balance was placed on a 100-kg concrete weighing table to minimize vibrations. The table was provided with a 110-mm circular hole through which the fishing line passed. A view of the test setup is shown in Figure 2. In the case of measurement in water, the sample was submerged in a water bath, consisting of a container of dimensions 0.50.80.4 m, holding about 100 l of distilled water. Water circulation in the bath, which would have helped keeping the sample temperature constant, was avoided to improve the stability of the weight measurements. Most of the surface of the bath was covered with a lid to reduce evaporation; this was done to avoid changes in the water level and thermal gradients. In the case of measurements in paraffin oil, a smaller cylindrical container ( 130 mm, height 200 mm) was filled with paraffin oil; the container was immersed in the larger bathtub filled with water to help keeping the sample temperature constant. Temperature measurements by means of thermocouples showed a maximum temperature increase of 0.5 C in the samples measured in the water bath; the maximum temperature increase of the samples measured in the paraffin oil bath was 1.5 C. In both cases, the temperature peak occurred at 8 h from casting and the temperature equilibrated at 24 h. The immersed weight of the sample was measured and recorded automatically at regular intervals by controlling software. Measurements were generally recorded every 10 minutes from about 30 minutes after casting up to about 2 weeks. Upon immersion of the sample in the buoyancy liquid, air bubbles on the upper outer part of the membrane were gently removed. The first measurement immediately after immersion was discarded because of the influence of the oscillations of the sample on the measurements, though oscillations were quickly damped in the viscous paraffin oil. Another source of error in the very beginning were drops of buoyancy liquid attached to the string, but these fell into the bath in a few seconds. Linear measurements of autogenous strain Linear autogenous strain of cement pastes was measured by a special measuring technique, where the cement paste is encapsulated in thin, corrugated polyethylene moulds with length/diameter approximately 300:30 mm [37]. This ensures insignificant restraint of the hardening cement paste and permits measurements to start at an early age. The cement paste was cast under vibration into the moulds; the specimens were then placed in a dilatometer equipped with automatic data-logging and electronic linear displacement transducers. The dilatometer with samples was immersed into a temperature-controlled glycol bath at 200.1C. Two samples were tested simultaneously in the dilatometer, with a measuring accuracy of 5 m/m. A separate sample with an embedded thermocouple registered the temperature evolution. Measurements were performed every 15 minutes and started 30 minutes after casting. A detailed description of the dilatometric technique can be found in the literature [37]. 8

Absorption measurements To test absorption of buoyancy liquids through the different membranes used, a series of mass measurements on the balance plate was performed on different samples. Membranes filled with either synthetic pore solutions or cement pastes were immersed in the buoyancy liquids and their mass was recorded at regular intervals. Before every measurement, the samples were removed from the bath and the surface of the membranes was gently but accurately wiped with a paper towel to remove the excess water or paraffin oil.

Figure 2. Setup for testing volumetric autogenous strain of cement paste. A Sartorious CP 225 D balance is placed on a massive weighing table. Underneath the balance there is a hole in the table through which the sample is suspended by means of a hook and a string. The sample is immersed in a buoyancy bath, either water or paraffin oil.

Results and discussion Water transport through membranes filled with different solutions To quantify the influence of osmosis on the volumetric measurements, rubber membranes filled with either synthetic pore solution, 50% diluted, synthetic pore solution, or demineralized water, were immersed in a demineralized water bath and their mass was monitored. Both latex and polyurethane membranes were tested. The membranes contained amount of solutions varying from 70 to 140 g. Figures 3 and 4 show the results of the experiments. Water diffusion through the condoms was proportional to the solute concentration. In particular, the mass of condoms filled with synthetic pore solution increased by 0.26% per day, while the mass of the condoms filled with 50% diluted synthetic pore solution increased at about half the rate. The mass of membranes filled with demineralized water remained almost unchanged. No substantial differences in absorption rate were observed between latex and polyurethane membranes. However, after a few days in the water bath, the polyurethane membranes started leaking and lost weight. Examination of the polyurethane membranes revealed extensive degradation.

The scatter between identical membranes filled with similar quantities of synthetic pore solution was quite substantial, as shown by the standard deviation bars in Figures 3 and 4. The main reason is believed to be variability in thickness between the membranes and possibly presence of m-leaks and pinholes in individual membranes; this fact is confirmed by the literature about properties of condoms [54].
15

12

Mass change [%]

9
Synthetic pore solution LATEX

Synthetic pore solution POLYURETHANE

Synthetic pore solution 50% diluted POLYURETHANE

0 0 7 14 21 28

Time [days]

Figure 3. Mass change of membranes containing different solutions immersed in a demineralized water bath at 20.0C. The average and the standard deviation of 3 samples is shown in the case of the latex membranes filled with synthetic pore solution and of 2 samples in the case of the polyurethane membranes with synthetic pore solution. For the polyurethane membranes with diluted (50%) synthetic pore solution, a single measurement is shown.
7 6 5

Mass change [%]

Synthetic pore solution

4 3 2 1 0
Demineralized water Synthetic pore solution 50% diluted

-1 0 500 1000 1500 2000

Concentration of solution [mmol/l]

Figure 4. Mass change of condoms containing synthetic pore solution, diluted (50%) synthetic pore solution, and demineralized water, immersed in a demineralized water bath at 20.0C for 21 days. Each measurement shows the average and the standard deviation based on 3 samples.

10

Osmotic transport is driven by differences of water activity across the membrane. The activity of water in an ideal solution can be calculated as: aW = xW (1) where xW [mol/mol] is the mole fraction of water in the solution. For the pore solution used, the calculated water activity is 0.973, supposing all species are fully dissociated in the solution. The osmotic pressure that can be established at thermodynamic equilibrium across a semi-permeable membrane between a pure liquid and a solution can be calculated by van't Hoffs formula: Posmotic = icRT (2) 3 where c [mol/m ] is the concentration of the solution, i [-] is the ionization factor of the solute, R = 8.314 J/(molK) is the ideal gas constant, and T [K] is the absolute temperature, in this study 293.15 K. For the synthetic pore solution used in this research, Posmotic,PS = 3.8 MPa, considering all ions as dissociated. In the membranes filled with synthetic pore solution, a pressure equal to the osmotic pressure would build up at equilibrium. This equivalent pressure can be used as a measure of the driving force that causes ingress of water into the membranes. As seen from eq.(2), the osmotic pressure is proportional to the solute concentration. The mass gain of the samples in Figure 4 depended linearly on the concentration of the pore solution, an indication that this process is indeed driven by osmosis. Moreover, the linear dependence also on time, as seen in Figure 3, confirms that the driving force remains almost constant through the experiment, as it is expected since almost no dilution of the pore fluid occurred during the test. Figures 1 and 2 are consistent with results obtained by Marciniak [42], who measured mass change of latex condoms filled with a synthetic pore solution with composition very similar to the one used in this research. However, Douglas [46] observed no relevant difference in the mass change of latex condoms filled with a lime-saturated solution immersed in a water bath compared to pure water. As observed before, the composition of the pore fluid in a cement paste is very different from a saturated lime solution [47, 44], especially because of its alkali ions content. In particular, due to the very low solubility of the calcium hydroxide in water, 0.022 mol/l at 20 C [55], the water activity of a saturated solution of calcium hydroxide calculated with eq.(1) is close to unity, or 0.999. The osmotic pressure that can be developed across the membrane has an upper bound at 0.16 MPa, considering the calcium hydroxide as fully dissociated in water. This value is much lower than the value calculated with eq.(2) for the synthetic pore solution used in this research. The conclusion of the experiments on elastic membranes filled with synthetic pore solution and immersed in a water bath is that osmosis through the membrane is a driving force able to account for the mass gain of the samples. By extension, it is a plausible driving force also for absorption of water into elastic membranes filled with cement paste [27, 46]. Moreover, it is a mechanism which is active from the very start of the volumetric measurement, because a high ionic concentration in the pore fluid is established rapidly when the cement is brought into contact with water. Relative humidity measurements on cement pastes Figure 5 shows the autogenous RH in the investigated hardening cement pastes. In paste A, with w/c 0.30 and 20% silica fume addition, a significant autogenous lowering of the RH is observed. After 12 days of sealed hardening, the RH is approximately 85% due to the hydration reactions. Water entrainment based on SAP addition is seen to be efficacious in counteracting autogenous RH 11

change: the autogenous RH change is almost fully counteracted in paste B, with 0.6% SAP addition. The slight autogenous RH change to about 98% RH of paste B can be accounted for by dissolved salts in the pore fluid [52]. Paste C shows a RH drop after 2 days until it stabilizes after about 10 days at 91% RH. The initial internal RH, 98-99% in the 3 mixes, can be attributed to the dissolved salts in the pore fluid of the cement pastes [53]. The equilibrium RH of a solution is described by Raoults law: RH S = xw (3) where xw [mol/mol] is the molar fraction of water in the solution. Since xw corresponds also to the activity of the solution, see eq.(1), the activity value calculated for the synthetic pore fluid in the previous paragraph would correspond to a relative humidity of 97% RH. This value is in quite good agreement with the initial RH in the three pastes, especially if considered that the composition of the pore solution was not measured on the pastes, but instead derived from the literature [47].
100 Paste B: w/c 0.30 + 20% SF + 0.6% SAP

Relative humidity (%)

95

Paste C: w/c 0.35 90

Paste A: w/c 0.30 + 20% SF 85 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Time (days)

Figure 5. Autogenous RH of cement pastes A, B and C. Silica fume, SF, and superabsorbent polymer, SAP, additions are given by weight of cement, where 0.6% corresponds to an entrained w/c of 0.075. Basic w/c is 0.3 for pastes A and B and 0.35 for paste C. Temperature: 20.0C.

More generally, since water vapor is close to an ideal gas, the measured internal RH of a cement paste is equal to the water activity in its pore fluid. Water activity is affected not only by the presence of dissolved salts, but also by the presence of air-water menisci in the pores of the cement paste. The menisci are a consequence of chemical shrinkage and induce a tensile stress in the pore fluid [53]. The influence on the RH from the menisci is described by Kelvins equation. The RH of a cement paste can be described as a product of two terms, one due to dissolved salts in the pore fluid, the other due to the presence of the menisci [53]: M RH = RH S RH K = xw exp cap (4) RT where cap [Pa] is the stress in the pore fluid, M=0.01802 kg/mol is the molar weight of water, and =1000 kg/m3 is the density of water. The derivation of eq.(4) is presented in reference [44]. 12

The stress in the pore fluid may be therefore computed as: RT RH (5) ln cap = RH M S where RH is measured and it is assumed that RHS does not change much during hydration, remaining close to the initial value of about 98% RH. With these assumptions, the calculated cap in the pore fluid after 7 d hydration is negligible for paste B, while it amounts to about -14.6 MPa for paste A and to -8.6 MPa for paste C. These values are of the same order of magnitude of the osmotic pressure calculated in the previous paragraph. According to Barcelo [48], the driving force causing water absorption through the membrane in the volumetric method is the reduced air pressure in the capillary system of a cement paste isolated from the atmospheric pressure. The cement paste in the membrane contains a certain amount of air initially. As further voids are created by chemical shrinkage during hydration, a differential pressure is established between the air voids of the cement paste and the water around the sample: Vair (6) p(t ) = 1 V + V (t ) patm air CS

where Vair (m3/m3) is the initial specific air volume in the paste, VCS (m3/m3) is the specific volume of chemical shrinkage as a function of hydration time, and patm (Pa) is the atmospheric pressure. The maximum differential pressure that can be generated by this mechanism is equal to the atmospheric pressure, about 0.1 MPa. Notice that this pressure is 30 times smaller than the osmotic pressure and about 100 times smaller than the underpressure in the pore fluid due to menisci formation. Considering that the cement paste initially contains a variable content of air (1-2%) and that the chemical shrinkage volume can reach about 7% at full hydration of a cement paste, the maximum underpressure would be reduced by 15-30%. Moreover, the undepressure becomes dependent on the amount of entrapped air, at least until the chemical shrinkage becomes much larger than the initial air content, which does not occur until some hours after setting. Consequently, the initial absorption of the condom would be dependent on the initial air content. However, such dependence has not been observed. Finally, this mechanism does not explain why cement pastes in membranes absorb water from a water bath already during the dormant period and before setting (as will be shown in the following paragraph), when very little chemical shrinkage has occurred and no capillary voids have formed. In conclusion, it seems that the mechanism of underpressure cited by Barcelo [48] is negligible: the main driving forces of water transport through the membrane are osmosis in the early period and self-desiccation after a couple of days of hardening. Water uptake of membranes filled with cement paste Latex membranes filled with cement paste were immersed in a demineralized water bath about 30 minutes after mixing and their mass was measured until about 3 weeks of age. In the first day, the specimens were weighed every few hours, with the first measurement taken at about 2 hours after mixing. Pastes with significant self-desiccation, with low w/c ratio and high silica fume content (A), and pastes with no self-desiccation, containing SAP (B), were tested. Moreover, paste C had a higher w/c ratio and no silica fume, which lead to a much higher permeability than the other pastes. Tests on different pastes aimed at separating the influence of osmosis from the influence of selfdesiccation on the absorption behavior, as discussed in the previous paragraph. 13

Results of the absorption measurement are shown in Figure 6. Pastes A and B are expected to have similar permeability throughout the hydration process, since the presence of SAP should have only limited impact on the transport property of the paste [49]. Therefore it is possible to analyze the water uptake of the two pastes only in terms of the acting driving forces. Long-term water absorption in cement paste B with SAP is lower than in paste A, whereas the absorption in the first hours after mixing is quite similar. According to the analysis presented in the previous paragraph, osmosis dominates the water uptake of the samples in the first hours. Osmotic forces depend on the concentration of the pore solution, which is expected to be quite similar in pastes A and B because the same cement and admixtures have been used. At a later age, self desiccation occurs in paste B and becomes the dominant driving force for water uptake. This explains the higher water absorption of paste B at later ages. It is noticed that water transport into the sample due to osmosis in the first hours occurs at a very high rate compared to later ages. In fact, high porosity and low tortuosity of the cement paste pore system at early age both promote transport of water into the cement paste. Moreover, at early age the transport of water into the sample is entirely liquid based, instead of the slower, combined liquid-gas transport that occurs when pores in the cement paste have been partially emptied by self-desiccation. Whereas at early ages the permeability of the elastic membrane is the limiting factor of the transport process, at later ages the reduced permeability of the cement pastes becomes dominant.
0.8 0.7 Paste C: w/c 0.35 0.6

Mass change (%)

0.5 0.4 0.3 0.2 0.1 0 0 7 14

Paste A: w/c 0.3 + 20%SF

Paste B: w/c 0.3 + 20%SF + 0.6%SAP

21

28

Time (days) Figure 6. Mass change of latex membranes filled with cement pastes immersed in a water bath at 20C. The average and the standard deviation of 5 samples are shown for pastes A and B, while 3 samples were measured for paste C.

Paste C shows a higher absorption in the first hours after casting, but is close to paste A in the following days, confirming that self-desiccation does indeed influence water uptake. Comparison of paste C with the other two pastes is complicated, because the water uptake is influenced also by the transport properties of the paste. In particular, paste C, with higher w/c and no silica fume, should have a much higher permeability than pastes A and B at any time. 14

A simple calculation confirms that the water uptake in the first day might be caused primarily by osmosis. The absorption rate of latex membranes filled with synthetic pore fluid is about 0.4% by weight per day (Figure 3). The cement pastes studied have initial water to solids volume ratio close to 0.5, which corresponds to the volume fraction of pore fluid in the samples. It is therefore possible to estimate the absorption in the first day as 0.4%0.5=0.2% by weight of cement paste. This value is indeed very close to the measured values, around 0.15% by weight of cement paste in the first day of hydration (Figure 6). When water is absorbed from the buoyancy bath into the samples during a volumetric measurement of autogenous strain, the measurements are affected in two ways: 1) the mass of the sample is increased since the water is absorbed into internal voids; this is measured as an apparent shrinkage after setting; 2) the water content of the sample is increased, which potentially increases the w/c before setting but especially reduces self-desiccation and self-desiccation shrinkage of the hardening cement paste.
0

Error on autogenous strain (m/m)

-1000

Paste B: w/c 0.3 + 20%SF + 0.6%SAP

-2000

-3000 Paste A: w/c 0.3 + 20%SF Paste C: w/c 0.35

-4000

-5000 0 7 14 21 28

Time (days)

Figure 7. Error on the measured autogenous strain due to water uptake of cement paste in latex membranes immersed in water bath at 20C. Linear strain was calculated assuming isotropic behavior; shrinkage is plotted as negative. The average and the standard deviation of 5 samples are shown for pastes A and B, while 3 samples were measured for paste C.

In Figure 7, the estimated error on volumetric autogenous strain measurements that would result from the weight change of the immersed samples is shown. The error was calculated based on the assumption that all the water absorbed into the sample filled voids produced by chemical shrinkage. This is a rough assumption, especially before setting, when no voids are present in the pastes and the absorbed water might instead contribute to raise the w/c ratio of the pastes. The error has been recalculated as linear shrinkage, assuming isotropic deformation, in order to facilitate comparisons with linear measurements in the following sections of this report; that means that the calculated error on the volumetric strain has been divided by 3. Absorption of water from the bath produces a 15

fictitious shrinkage after 3 weeks of 2000 m/m for paste B, 4000 m/m for paste A and more than 5000 m/m for paste C. Already at 1 day after casting, the measuring error is substantial, varying from 500 to 1500 m/m for the 3 pastes. Besides the artifacts resulting from the measuring technique, absorption of water from the bath affects the curing conditions of the sample, which will not be autogenous any longer. To perform measurements of autogenous strain, the cementitious system must be: 1) sealed, 2) kept at constant temperature, and 3) not subjected to external forces [52]. Therefore, in the case of water penetrating the membrane the measured strain is not autogenous strain. According to Barcelo [48] the effect of water absorbed into the sample can be described as an increase of the average water/cement ratio of the cement paste; with this assumption, he showed that the effect on self desiccation was negligible in the first week of hydration. However, it is misleading to assimilate the absorbed water to an increase of w/c ratio: the situation is more similar to a cement paste with a given w/c ratio cured with an external supply of water. This external supply of water will have a much greater influence on the autogenous shrinkage of the paste than if it was supplied initially in the form of a higher w/c ratio [49]. Some authors [33, 35, 41, 43] stated that absorption through the elastic membrane is a slow process and its impact on the volumetric measurements can be neglected in short-term measurements; therefore, use of the volumetric technique would be justified for measurements in the first day after casting. However, no data on water uptake of volumetric samples was published by these authors to confirm their hypothesis. On the contrary, Figure 7 shows that the impact of water absorption from the buoyancy bath on the volumetric measurements is great enough to compromise the whole measurement from the time of casting, since an error on the measured strain of 500-1500 m/m in 1 d is of the same order of magnitude as the true strain itself. Having proved that the water uptake of the samples is a significant artifact of the volumetric technique and capable of disrupting its measurement results from the beginning, it is possible to tackle the problem in two different ways: compensate for the absorption or eliminate the absorption. Barcelo [48, 56] compensated for this artifact by measuring the weight increase of immersed samples over a long period, about one month, and using a linear regression to fit a correction to the measured strain. Once this correction was applied, a part of the discrepancy between linear and volumetric measurements was eliminated, according to Barcelo [48]. Besides ignoring the non autogenous conditions of the sample, this type of correction suffers from a number of other drawbacks: 1) According to the measurements in Figure 6, absorption of water into the samples proceeds at a faster rate in the beginning, because of osmosis and high permeability of the paste (Figure 6). A linear correction factor does not take this into account; 2) Absorption shows a very high variability between samples, as shown both in Figure 6 and in reference [46]. The variability in absorption is comparable to the measured strain and questions the use of an average correction factor. To be accurate, one should measure absorption and volumetric strain at the same time on the same sample, which is very problematic and would preclude continuous measurements of volumetric autogenous strain; 16

3) It is uncertain how absorbed water influences the volumetric measurements before and around setting; this is the moment when we are mostly interested in the volumetric measurements, since linear measurements cannot be carried out at this point. In fact, if there are no capillary pores to be filled before setting, the absorbed water might have little influence on the buoyancy of the sample; the weight of the sample would increase, but this would be compensated by an increase of the volume. Since setting time is difficult to determine and it is a gradual transition, it is uncertain when the correction factor should be applied to the measured strain; 4) More generally, it appears unsound to apply a correction factor that is of the same order of magnitude as the measurement results. Another procedure to correct the volumetric measurements was proposed by Douglas [45, 46]. He observed that not only did the samples absorb water from the bath during the volumetric test, but that the latex membranes themselves absorbed water. Finally, by stripping the membrane after the measurement and weighing both membrane and sample, he found that about 70% of the water was absorbed into the cement paste and most of the rest remained as a surface layer, while only a very minor amount was absorbed by the membrane itself. On the basis of these observations, Douglas [45, 46] proposed to correct the volumetric measurements with multiple and nonlinear correction factors, taking into account all this complex behavior. The same considerations advanced in the previous paragraph apply also to this case, with the further remark that the number of corrections needed, their order of magnitude, and their uncertainty makes the method extremely difficult to apply. The other possible solution of this artifact is to eliminate the absorption of water through the elastic membrane. The most obvious way is to choose an impermeable membrane. With this aim, Barcelo [57] measured a number of latex membranes but found that they all were permeable. Douglas [46] tested a number of different latex, polyurethane and neoprene membranes and all were permeable to water. Also in the present research, latex and polyurethane membranes were found to have similar permeability (Figure 4). The lack of tightness of the membranes is due both to their intrinsic permeability and to the presence of defects such as pinholes. Pinholes would be especially significant to increase water uptake at early ages, while their importance at later ages would be lowered by the reduced permeability of the cement paste. Making the membrane thicker, while improving the tightness, would also increases its stiffness and restrain the cement paste, especially around setting time [12]. It is not excluded that a satisfactory balance between stiffness and tightness of the elastic membrane can be achieved; however, this is not the case with the type of membranes used in the volumetric measurements in the last 15 years, i.e. latex condoms. A more radical approach to eliminate the water uptake of the samples in the volumetric test is to substitute the water with a different buoyancy liquid. As documented in the introduction to this report, volumetric measurements with displacement liquids different than water, i.e. oil or mercury, are not a novelty in the literature [2-4]. However, water has been the displacement liquid of choice in the last 25 years [14-36]. Douglas [46] performed one single measurement using latex balloons cured in 100% RH air between the measurements and another measurement using polyurethane condoms cured in vegetable oil. The reason for not using latex membranes in oil is that latex is highly unstable in oils: it swells up and rapidly degrades in a matter of a few hours. 17

Douglas [46] observed first a mass decrease of polyurethane condoms filled with mortar in vegetable oil in the first day, and then a small increase until 7 d. The total variation was about 0.1%, while the water uptake of latex condoms filled with mortar in a water bath was about 0.5% in the same period. The mass decrease of the polyurethane condom filled with mortar was attributed to partial disintegration of the condom as it reacted with the vegetable oil, which afterwards allowed some oil to penetrate into the sample, explaining the later weight increase. In this research, measurements of mass change of polyurethane condoms filled with different cement pastes and immersed in paraffin oil were performed. The mass gain in the first week of immersion was about 0.005%, which is about 1/20 to 1/35 the mass change observed in the same period on latex condoms filled with cement paste and immersed in a water bath. This mass change would lead to an error on the measured autogenous strain of less than 50 m/m in one week. A few samples showed a mass gain much higher, up to 0.03% in 4 days; however, in this case it was easy to verify the presence of leaks in the membranes by observing darker areas on the surface of the cement paste. It was also observed that when the measured volumetric strain curve showed abrupt variations, the samples showed a mass gain and stains on the cement paste due to paraffin oil were also present; in this case, the results were discarded. Volumetric measurements of autogenous strain on cement pastes On the same pastes tested for weight change underwater, the autogenous strain was measured with three different methods: 1) Measurements of immersed weight of cement paste samples cast in latex condoms and immersed in a water bath; 2) Measurements of immersed weight of cement paste samples cast in polyurethane condoms and immersed in a paraffin oil bath; 3) Measurement of length changes of samples cast in corrugated moulds in the dilatometer [37]. Changing the buoyancy liquid from water to paraffin oil eliminates the influence on the volumetric measurements of water absorption from the bath into the sample. Linear measurements were performed in parallel for comparison with the volumetric ones. In all cases, the mass of the samples before and after testing was measured. Mass change of the samples was generally negligible in the case of measurements of type 2) and 3), whereas the mass gain of samples tested in a water bath confirmed the findings in the previous section (Figure 6). In Figure 8-10, measurements of autogenous strain with the three techniques are presented. In addition, another curve is calculated adding to the measurements in paraffin oil the error due to water absorption (Figure 7); this curve shows the effect on the volumetric measurements of the change of buoyancy due to water absorption. In Figure 8, measurements are shown for paste A, with w/c 0.3 and 20% silica fume addition. Volumetric measurements have been converted into linear under the assumption of isotropic behavior, i.e. dividing the volumetric strains by a factor 3. The measurements have been zeroed at 6 h, corresponding to the setting time. Linear measurements and volumetric measurements in paraffin oil run parallel for several days, showing an initial expansion followed by shrinkage. The volumetric measurements in water deviate significantly from the other two from about 12 h, with a 18

shrinkage rate about 5 times greater; they also show a higher experimental scatter than the other types of measurements. If the error due to absorption calculated for samples stored in a water bath (Figure 7) is added to the volumetric measurement in paraffin oil, a shrinkage rate close to the one measured in water is obtained.
1000 500

Autogenous strain (m/m)

0 -500

-1000 -1500 -2000 -2500 -3000 0 1 Volumetric - Paraffin oil + Absorption 2 3 4 5 Volumetric - Water 6 7 Linear Volumetric - Paraffin oil

Time (days)

Figure 8. Measurements of autogenous strain with different techniques at 20C on cement paste with w/c ratio 0.3 and 20% silica fume addition (paste A). Conversion from volumetric to linear strain assumed isotropic behavior; the curves have been zeroed at setting, 6 h after casting; shrinkage is plotted as negative.

In Figure 9, measurements are shown for paste B, with 0.6% superabsorbent polymers by weight of cement. Also in this case, results of volumetric measurements have been converted into linear strain assuming isotropic behavior. The measurements have been zeroed at 6 h, corresponding to the setting time. After an initial shrinkage, the pastes started to expand in all types of measurements. Linear measurements and volumetric measurements in paraffin oil follow a similar pattern after setting, though with a slightly greater initial expansion in the linear measurements. Measurements in water, however, significantly deviate from the other two from about 12 h. They show shrinkage, whereas according to the other measurements the paste continued to expand for a couple of days. Moreover, duplicate measurements in water show a considerable scatter in the expansion phase, probably due to different absorption of the single specimens (compare with the scatter in Figure 6). The subsequent shrinkage is instead quite similar in the two samples. When the error due to absorption for samples stored in a water bath (Figure 7) is added to the measurements in paraffin oil, a shrinkage rate similar to the measurements in water is obtained.

19

1000 Linear 500 Volumetric - Paraffin oil

Autogenous strain (m/m)

0 Volumetric - Water -500 Volumetric - Paraffin oil + Absorption

-1000 -1500 -2000 -2500 -3000

0 1 2

Time (days)

Figure 9. Measurements of autogenous strain with different techniques at 20C on cement paste with w/c ratio 0.3, 20% silica fume addition and 0.6% superabsorbent polymers (SAP) by weight of cement (paste B). Conversion from volumetric to linear strain assumed isotropic behavior; the curves have been zeroed at setting, 6 hours after casting; shrinkage is plotted as negative.
1000 500 Linear

Autogenous strain (m/m)

0 -500 -1000 -1500 Volumetric - Water -2000 -2500 -3000 0 1 2 3 4 5 6 7 Volumetric - Paraffin oil + Absorption

Volumetric - Paraffin oil

Time (days)

Figure 10. Measurements of autogenous strain with different techniques at 20C on cement paste with w/c ratio 0.35 (paste C). Conversion from volumetric to linear strain assumed isotropic behavior; the curves have been zeroed at setting, 3.25 h after casting; shrinkage is plotted as negative.

20

Figure 10 shows measurements of autogenous strain on paste C, with w/c 0.35. Also in this case, volumetric measurements have been converted into linear strain assuming isotropic behavior. The measured strain has been zeroed at setting, corresponding to 3.25 h; notice that in this case setting occurred earlier than in the other two cement pastes because no superplasticizers were used. Linear measurements and volumetric measurements in paraffin oil show some differences in the first day after casting, while the measured shrinkage rate in the following days is almost identical. The measurements in water deviate from the other two from about 12 h; after a first short expansion, shrinkage is measured, whereas according to the other measurements the paste continued to expand for a couple of days. Moreover, the measurements in water show a considerable scatter, probably due to different absorption of the single specimens (Figure 6). Also in this case an additional curve is shows the measurement in paraffin oil plus the error due to absorption in samples stored in water (Figure 7); the calculated curve follows closely the volumetric measurements in water, included the first short expansion and the following shrinkage rate. Measurements in Figures 8-10 show a number of common features. Linear measurements performed with the dilatometer [37] and volumetric measurements in paraffin oil fundamentally agree after some differences in the first few hours. Moreover, their results are in qualitative agreement with RH measurements: for example, paste B shows almost no lowering of the internal RH in the first week (Figure 5) and this agrees with the negligible shrinkage measured both by the linear technique and by the volumetric technique in paraffin oil (Figure 9). The same considerations apply for the start of self-desiccation of paste A and C, which corresponds to the onset of shrinkage according to the linear technique and the volumetric technique in paraffin oil. Moreover, both linear measurements and volumetric measurements in paraffin oil are quite reproducible and show little scatter throughout the measuring period. This good agreement, obtained for 3 cement pastes with very different autogenous strain behavior, confirms the validity of both techniques. On the contrary, the volumetric measurements in water fundamentally disagree with the other two types of strain measurements and with the autogenous RH measurements. The measured shrinkage is systematically higher than for the other two techniques. This agrees with previous data in the literature about the volumetric technique in water bath [38, 39, 48]. As explained in the previous paragraphs, absorption of water through the elastic membrane into the samples is a principal cause of this discrepancy [38, 46]. In Figures 8-10, the estimated error due to water absorption (Figure 6) was added to the volumetric measurements in paraffin oil; this calculated curve followed closely the volumetric measurements in water bath. In particular, a very good agreement is obtained with the shrinkage rate after a couple of days: this shows that almost all the shrinkage measured with the volumetric method in water is an experimental artifact due to water absorption into the sample. Moreover, volumetric measurements in water show a poor reproducibility in the first hours or days, while the strain rate at later ages is more similar between the samples. Similarly, mass gain of the samples stored in water showed a high scatter especially in the first period, which may be due to different permeability of the elastic membranes. The high initial scatter in the volumetric measurements in water is another consequence of the water absorption through the membrane, which varies from sample to sample especially in the first hours. Considering the high variation in the results of the volumetric measurements in water bath, the strategy of correcting those measurements using a coefficient calculated from the average measured absorption [38, 46] does not appear practicable neither convenient. In fact, volumetric measurements in paraffin oil show a better 21

reproducibility, they do not need calculation of any correction factor, and finally they do not alter the curing condition of the sample. Some brief additional comments about the differences between the linear technique and the volumetric technique in paraffin oil in the first hours after setting are listed in the following paragraph. In the expansion phase measured for pastes B and C (Figures 9 and 10), the minimum of the strain is shifted at later ages in the volumetric measurements and the expansion is lower; after the first expansion phase, however, the measured strain is very similar. Causes of these differences are possibly a result of the well-known artifacts of both techniques: - Reabsorption of bleed water may lead to a measured additional shrinkage in the volumetric method that may partially override the initial expansion of the paste. On the other hand, it may lead to an increased expansion in the linear technique. Bleeding might have occurred in paste C (Figure 10), but it is unlikely for paste B, with low w/c, silica fume and SAP (Figure 9); - Absorption of air bubbles and suction of the membrane into the small irregularities on the surface of the volumetric sample would be a relevant mechanism for both pastes B and C. Observation of volumetric samples after the test showed that indeed this had occurred. This artifact would result in apparent extra shrinkage that overrides the initial expansion; - In the volumetric method, the pressure of the elastic membrane [13] acts on the weak interface layer between membrane and pastes, whose deformation should be much greater than the ones of the bulk. All these mechanisms should be especially relevant in the period around and immediately after setting, which corresponds to the period when expansion is measured. Their relevance in the later period would be minor. Conclusions In this report, absorption of water from the buoyancy bath through the elastic membrane is identified as the principal artifact of the volumetric method and cause of most of the discrepancies between volumetric and linear measurements. Water absorption is driven by a lowered water activity in the cement paste, due to dissolved salts and to self-desiccation. The importance of osmosis through the elastic membrane was shown by measuring mass change of membranes filled with synthetic pore solution and immersed in water. Water is absorbed into the cement paste sample at a high rate from the moment of immersion and the error it produces on the measured strain is of the same order of magnitude as the autogenous strain itself. All measurement results obtained with the volumetric technique using latex condoms to encapsulate the sample and a water bath [16-35] should therefore be considered with suspect, due to absorption of water into the sample. Notice that none of these papers reported the mass of the samples before and after the test. Moreover, all papers that make use of these results should be questioned as well. The strategy of limiting the measurements to the first day of hydration [33, 35, 41, 43] does not solve the problem, because absorption in that period is a significant source of error. In all volumetric measurements, the mass of the sample before and after measurement should be measured to verify absorption of the buoyancy liquid into the sample or possibly dissolution of the membrane; if a mass change is observed, the whole measurement should be disregarded.

22

An artifact resulting from water absorption through the membrane is very difficult to compensate [38, 46], especially because of high scatter in the absorption of different samples. Even if correction factors are applied, discrepancies between linear and volumetric measurements results do not appear to be fully eliminated [38]. In this paper, absorption of water through the membrane was avoided by paraffin oil as buoyancy liquid. Samples weighed in a paraffin oil bath did not show any appreciable mass change in the testing period of one week. It was shown that adding the measuring error due to the absorption to the measurement results in paraffin oil, a shrinkage rate much higher and similar to the one measured in water was obtained. This confirmed that, in volumetric measurements in water, the measurement results are dominated by the effect of water absorption through the membrane. The volumetric autogenous strain of three cement paste measured in a paraffin oil bath was in good agreement with measurements performed with a linear technique, the dilatometer [37]. Some minor differences between the two measuring techniques in a couple of hours around setting can be object of further research. Acknowledgements This study is part of a 3-year project dedicated to examine measuring methods for autogenous strain. The funding by the Danish Technical Scientific Research Council (STVF) is gratefully acknowledged. Portland Cement Association was not a sponsor of this research for SN2925. The contents of this report reflect the views of the authors, who are responsible for the facts and accuracy of the data presented. The contents do not necessarily reflect the views of the Portland Cement Association. References
[1] S. Diamond, Mercury porosimetry - An inappropriate method for the measurement of pore size distributions in cement-based materials, Cem. Concr. Res. 30 (10) (2000) 1517-1525. [2] H.A. Neville, H.C. Jones, The study of hydration changes by a volume-change method, Colloid Symp. Monogr. VI, VI Symposium on Colloid Chemistry (1928), pp. 309318. [3] M. Del Campo, A new method to study the early age volume changes on the neat cement paste, Mater. Constr. (Bull RILEM) 4 (1959) 18-23. [4] F.O. Slate, R.E. Matheus, Volume changes on setting and curing of cement paste and concrete from zero to seven days, ACI Journal 1 (1967) 34-39. [5] J.C. Yates, Effect of calcium chloride on readings of a volumeter inclosing Portland cement pastes and on linear changes of concrete, Highway Research Board Proc. 21 (1941) 294-304. [6] C.E. Wuerpel, Laboratory studies of concrete containing air-entraining admixtures, ACI Journal 42 (4) (1946) 305359. [7] R. LHermite, J.J. Grieu, Etudes exprimentales rcentes sur le retrait des ciments et btons, Ann. Inst. Techn. Btiment Travl. Publ., Paris, 52-53 (1952) p. 491. [8] R.M. Edmeades, A.N. James, J. Wheeler, The formation of ettringite on aluminium surfaces, J. Appl. Chem. 16 (1966) 361-368. [9] G.D. de Haas, P.C. Krejger, E.M.M.G. Niel, J.C. Slagter, H.N. Stein, E.M. Theissing, M. van Wallendael, The shrinkage of hardening cement paste and mortar, Cem. Concr. Res. 5 (1975) 295-320. [10] Y. Yamazaki, T. Monji, K. Sugiura, Early age expanding behaviour of mortars and concretes using expansive additives of CaO-CaSO4-4CaO3Al2O3SO3 system, 6th Int. Conf. on the Chemistry of Cement, Moscow, September 1974, Stroyizdat, Moscow III-5 (1976) pp. 192-195. [11] N. Setter, D.M. Roy, Mechanical features of chemical shrinkage of cement paste, Cem. Concr. Res. 8 (5) (1978) 623-634. [12] J. Baron, M. Buil, Comments on the paper Mechanical features of chemical shrinkage of cement pastes by N. Setter & D.M. Roy, Cem. Concr. Res. 9 (4) (1979) 545-547. [13] M. Buil, Studies of the shrinkage of hardening cement paste, D.Eng. thesis, Rapport de recherche LPC No. 92, 23 Laboratoire Central des Ponts et Chausses, Paris (1979). th [14] S. Ziegeldorf, H.K. Hilsdorf, Early autogenous shrinkage of cement pastes, 7 Int. Conf. on the Chemistry of Cement, Vol. 4, Paris 1980, pp. 333-338.

[15] S. Ziegeldorf, H.S. Mller, J. Plhn, H.K. Hilsdorf, Autogenous shrinkage and crack formation in young concrete, Int. Conf. on Concrete of Early Ages, RILEM, 1982, Vol. 1, pp. 83-88. [16] E.J. Sellevold, . Bjntegaard, H. Justnes, P.A. Dahl, High performance concrete: early volume change and cracking tendency, Proc. Int. RILEM Symp. Thermal Cracking in Concrete at Early Ages, Munich 1995, pp. 229-236. [17] H. Justnes, A. Van Gemert, F. Verboven, E. Sellevold, Total and external chemical shrinkage of low w/c ratio cement pastes, Adv. Cem. Res. 8 (31) (1996) 121-126. [18] H. Justnes, A. Van Gemert, F. Verboven, E. Sellevold, D. Van Gemert, Influence of measuring method on bleeding and chemical shrinkage values of cement pastes, 10th Int. Conf. on the Chemistry of Cement, Gothenburg, Sweden, June 1997, Amarkai AB and Congrex Gteborg AB, Gteborg 2 (1997), paper 2ii069. [19] H. Justnes, E.J. Sellevold, B. Reyniers, D. Van Loo, A. Van Gemert, F. Verboven , D. Van Gemert, The influence of cement characteristics on chemical shrinkage, in Autogenous Shrinkage of Concrete, Ed. E. Tazawa, E&FN Spon, 1999, pp. 71-80. [20] H. Justnes, T. Hammer, B. Ardouille, E. Hendrix, D. Van Gemert, K. Overmeer, E.J. Sellevold, Chemical shrinkage of cement paste, mortar and concrete, in Autogenous shrinkage of concrete, Ed. E. Tazawa, E&FN Spon, 1999, pp. 211-220. [21] H. Justnes, F. Clemmens, P. Depuydt, D. Van Gemert, E.J. Sellevold, Correlating the deviation point between external and total chemical shrinkage with setting time and other characteristics of cement pastes, Proc. Shrinkage 2000 Int. RILEM Workshop on Shrinkage of Concrete, Paris 2000, pp. 57-73. [22] H. Justnes, E.J. Sellevold, B. Reyniers, Chemical Shrinkage Of Cement Pastes With Plasticizing Admixtures, Nordic Concr. Res. 24 (4) (2000) 39-44. [23] A. Loukili, D. Chopin, A. Khelidj, J.Y. Le Touzo, A new approach to determine autogenous shrinkage of mortar at an early age considering temperature history, Cem. Concr. Res. 30 (6) (2000) 915-922. [24] V. Morin, F. Cohen Tenoudji, A. Feylessoufi, P. Richard, Superplasticizer effects on setting and structuration mechanisms of ultrahigh-performance concrete, Cem. Concr. Res. 31 (1) (2001) 63-71. [25] A. Feylessoufi, F. Cohen Tenoudji, V. Morin, P. Richard, Early ages shrinkage mechanisms of ultra-highperformance cement-based materials, Cem. Concr. Res. 31 (11) (2001) 1573-1579. [26] J.-P. Charron, J. Marchand, B. Bissonnette, M. Pigeon, B. Zuber, Comparative study of the effects of water/binder ratio and silica fume on the volume instability of hydrating cement pastes at early age, 3rd Int. Res. Sem. on Selfdesiccation and its Importance in Concrete Technology, Lund, Sweden, 2002, pp. 39-50. [27] F. Beltzung, F. Wittmann, Influence of cement composition on endogenous shrinkage, 3rd Int. Res. Sem. on Selfdesiccation and its Importance in Concrete Technology, Lund, Sweden, 2002, pp. 113-125. [28] P. Turcry, A. Loukili, Autogenous and thermal deformations of low water/cement paste at early age, 15th ASCE Engineering Mechanics Conference, June 2002, Columbia University, New York, NY. [29] P. Turcry, A. Loukili, L. Barcelo, J.M. Casabonne, Can the maturity concept be used to separate the autogenous shrinkage and thermal deformation of a cement paste at early age?, Cem. Concr. Res. 32 (9) (2002) 1443-1450. [30] V. Morin, F. Cohen-Tenoudji, A. Feylessoufi, P. Richard, Evolution of the capillary network in a reactive powder concrete during hydration process, Cem. Concr. Res. 32 (12) (2002) 1907-1914. [31] J.L. Plawsky, S. Jovanovic, H. Littman, K.C. Hover, S. Gerolimatos, K. Douglas, Exploring the effect of dry premixing of sand and cement on the mechanical properties of mortar, Cem. Concr. Res. 33 (2) (2003) 255-264. [32] P. Mounanga, Comportement des matrices cimentaires au jeune ge: relation entre volution chimique et dformations chimique et endogne, XXIEMES Rencontres Universitaires de Genie Civil 2003 Prix Rene Houpert, pp. 189-197. [33] V. Baroghel-Bouny, P. Mounanga, A. Loukili, A. Khelidj, From chemical and microstructural evolution of cement pastes, to the Development of Autogenous Deformations, ACI SP 220 Autogenous deformation of concrete, Ed. O.M. Jensen, D.P. Bentz & P. Lura, 2004, pp. 1-22. [34] P. Mounanga, V. Baroghel-Bouny, A. Loukili, A. Khelidj, Autogenous deformations of cement pastes: Part I. Temperature effects at early age and micromacro correlations, Cem. Concr. Res. 2005, in press. [35] V. Baroghel-Bouny, P. Mounanga, A. Khelidj, A. Loukili, N. Rafa, Autogenous deformations of cement pastes: Part II. W/C effects, micromacro correlations, and threshold values, Cem. Concr. Res. 2005, in press. [36] R. Gagn, I. Aouad, J. Shen, C. Poulin, Development of a new experimental technique for the study of the autogenous shrinkage of cement paste, Mater. Struct. 32 (223) (1999) 635-642. [37] O.M. Jensen, P.F. Hansen, A dilatometer for measuring autogenous deformation in hardening Portland cement paste, Mater. Struct. 28 (181) (1995) 406-409.

24

[38] L. Barcelo, S. Boivin, S. Rigaud, P. Acker, B. Clavaud, C. Boulay, Linear vs. volumetric autogenous shrinkage measurement: Material behaviour or experimental artefact?, 2nd Int. Res. Sem. on Self-desiccation and its Importance in Concrete Technology, Lund, Sweden, 1999, pp. 109-125. [39] J.P. Charron, J. Marchand, B. Bissonette, Early-age deformations of hydrating cement systems: comparison of linear and volumetric shrinkage measurements, RILEM Int. Conf. on Early Age Cracking in Cementitious Systems (EAC'01), Haifa, Israel, March 2001, pp. 245-257. [40] P. Lura, O. M. Jensen, A discussion of the paper: On the measurement of free deformation of early age cement paste and concrete by . Bjntegaard, T. A. Hammer and E. J. Sellevold, Cem. Concr. Comp. 27 (7-8) (2005) 854856. [41] T.A. Hammer, . Bjntegaard, E.J. Sellevold, Measurement methods for testing of early age autogenous strain, in: A.Bentur (ed.) Early age cracking in cementitious systems, RILEM TC 181-EAS Committee, RILEM, Cachan (2002) pp. 234-245. [42] S. Marciniak, Autogenous deformations and relative humidity change: experimental studies, M.S. thesis, Department of Building Technology and Structural Engineering, Aalborg University, Aalborg (2002). [43] . Bjntegaard, T.A. Hammer, E.J. Sellevold, On the measurement of free deformation of early age cement paste and concrete, Cem. Concr. Comp. 26 (5) (2004) 427-435. [44] P. Lura, Autogenous deformation and internal curing of concrete, Ph.D. thesis, Delft University of Technology, Delft, The Netherlands, 2003, 180 pp. [45] K.S. Douglas, K.C. Hover, Measuring non-drying bulk shrinkage of cement paste and mortar using Archimedes principle: Part I, ACI SP 220 Autogenous deformation of concrete, Ed. O.M. Jensen, D.P. Bentz & P. Lura, 2004, pp. 39-56. [46] K.S. Douglas, Non-drying Bulk Shrinkage of Cement Paste and Mortar: Application to Pneumatic Transport Mixing of Sand and Cement, M.S. Thesis, Cornell University, Ithaca, NY, 2002. [47] C.L. Page, . Vennesland, Pore solution composition and chloride binding capacity of silica-fume cement pastes, Mater. Struct. 16 (91) (1983) 19-25. [48] L. Barcelo, Influence des caractristiques des ciments sur la structuration et le comportement dimensionnel des matriaux cimentaires au jeune ae, Ph.D thesis, Ecole Normale Suprieure de Cachan, Cachan, France, 2001. [49] O.M. Jensen, P.F. Hansen, Water-entrained cement-based materials II. Implementation and experimental results, Cem. Concr. Res. 32 (6) (2002) 973-978. [50] ISO 679:1989, Methods of testing cements -- Determination of strength, Par. 4.2.3 'Mixer'. [51] O.M. Jensen, P.F. Hansen, Autogenous relative humidity change in silica fume-modified cement paste, Adv. Cem. Res. 7(25) (1995) 3338. [52] O.M. Jensen, P.F. Hansen, Autogenous deformation and change of the relative humidity in silica fume-modified cement paste, ACI Mater. J. 93 (6) (1996) 539-543. [53] P. Lura, O.M. Jensen, K. van Breugel, Autogenous shrinkage in high-performance cement paste: An evaluation of basic mechanisms, Cem. Concr. Res. 33 (2) (2003) 223-232. [54] R.F. Carey, C.D. Lytle, W.H. Cyr, Implications of laboratory tests of condom integrity, Sex. Transm. Dis. 26 (4) (1999) 216-220. [55] Handbook of Chemistry and Physics, R.C. Weast (ed.), Chemical Rubber Company, 1984. [56] L. Barcelo, M. Moranville, B. Clavaud, Autogenous shrinkage of concrete: a balance between autogenous swelling and self-desiccation, Cem. Concr. Res. 35 (1) (2005) 177-183. [57] L. Barcelo, personal communication, January 2005.

25