Mechanical behaviour of polymer modified mortars

We investigate the mechanical behaviour of a mortar and five polymer (latex styrene–butadiene) modified mortars (PMMs) with different polymer contents. The mechanical characterisation of the materials is based on compression and three-point bending tests. As expected, compression tests reveal a decrease of the PMMs rigidity and compressive strength for increasing polymer content. On the contrary, three-point bending tests show an increase of the flexural strength from a polymer-to-cement weight ratio higher or equal to 10 wt.%. After some considerations on material porosity and cement hydration, we establish that the main cause of the PMMs flexural strength increase is the latex percolation into a continuous network. Finally, analysis on damage initiation and material rigidity seems to indicate that latex addition limits skin and inner material micro-cracking due to sample drying out.

Materials Science and Engineering A 380 (2004) 1–8 Mechanical behaviour of polymer modified mortars S Pascal a , A Alliche b,d,∗ , Ph Pilvin c a DEN/DM2S/SEMT/LM2S, CEN Saclay, 91191 Gif-Sur-Yvette Cedex, France b Laboratoire MSS-Mat, École Centrale, Paris, France c LG2M, rue de Saint Maudé, 56321 Lorient Cedex, France d Grande Voie des Vignes, 92295 Chˆ atenay-Malabry, France Received 30 May 2003; received in revised form 22 October 2003 Abstract We investigate the mechanical behaviour of a mortar and five polymer (latex styrene–butadiene) modified mortars (PMMs) with different polymer contents The mechanical characterisation of the materials is based on compression and three-point bending tests As expected, compression tests reveal a decrease of the PMMs rigidity and compressive strength for increasing polymer content On the contrary, three-point bending tests show an increase of the flexural strength from a polymer-to-cement weight ratio higher or equal to 10 wt.% After some considerations on material porosity and cement hydration, we establish that the main cause of the PMMs flexural strength increase is the latex percolation into a continuous network Finally, analysis on damage initiation and material rigidity seems to indicate that latex addition limits skin and inner material micro-cracking due to sample drying out © 2004 Elsevier B.V All rights reserved Keywords: Polymer modified mortar; Latex additive; Mechanical testing; Compressive strength; Flexural strength Introduction Polymer modified mortars (PMMs) were developed to deal with some needs in building and public works These materials are used as tile adhesives, frontage coatings, coverings, and as repair materials in road building The gaol was to obtain materials easy to implement and with a better useful life than usual mortars Besides, they should exhibit good adhesive properties as well as an improved resistance to environmental stress, as wall insulating against humidity in particular Various kinds of polymer [1,2] (vinyl acetate, styrene butadiene rubber for example) are used for these materials Polymer content is usually given according to the dry polymer-to-cement weight ratio (P/C) The presence of polymer in a mortar leads to different microstructures depending on the way the polymer extends turns out into the material, which is also closely related to the polymer content Studies on PMMs are generally based on comparisons between PMMs with various polymer contents and a mortar ∗ Corresponding author Tel.: +33-1-41-13-13-02; fax: +33-1-41-13-14-30 E-mail address: alliche@mssmat.ecp.fr (A Alliche) 0921-5093/$ – see front matter © 2004 Elsevier B.V All rights reserved doi:10.1016/j.msea.2004.03.049 as reference Among these studies, we can distinguish the way in which PMMs have been worked out: (i) by adjusting the water to cement weight ratio (W/C) in order to keep constant the viscosity or the flow of the different materials [3–6], (ii) by keeping constant the W/C ratio, aiming to provide a similar hydration of the cement matrix [2,7–9] In the first case (i), the flowing effects of the polymer due to the surfactants that it contains require to lower the W/C ratio for increasing polymer contents In the second case (ii), the usual air entrainment effect of the polymer increases the porosity of the composite For PMMs made up by keeping constant the viscosity between the different materials, the results indicate that the mechanical strength determined both in tension, compression and three point bending increases for higher polymer contents [1,3,4,6], while elastic modulus decreases [1,5,6] Moreover, the polymer would improve the water retention capability of these materials, and then limit their drying out for the benefit to cement hydration [1,3–5] Accordingly, long-term hydration would be promoted into PMMs [3] These composites also turn out to be more resistant to freeze-thaw cycling and, more generally, would ensure a better protection against environmental stress [1,3,4] However, these studies involve comparisons between cement based S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 Table Mix proportions of materials: polymer (P), super-plasticizer (SP), antifoamer (AF), water (W), sand (S) and silica fume (SF) to cement (C) weight ratios Constituents P/C SP/C AF/C W/C S/C SF/C Mortar 0.05 0.075 0.006 0.4 1.6 0.3 PMMs 0.1 0.125 0.15 0.01 0.01 0.4 1.6 0.3 adverse effects like variations in porosity or cement hydration As a consequence, straight comparisons between the mechanical properties of PMMs of various polymer contents may not only reveal the influence of the polymer phase, but also include the influence of these adverse effects In our previous paper [2], some questions remained according to that In the present study, we still made up mortar and PMM samples of various polymer contents, but special care was taken to these adverse effects, especially by trying to keep constant the porosity over the materials Results about their porosity and other material characteristics [12] are to be subjected to further publications The mechanical properties of the materials are again estimated using three point bend and compression tests After a description of the sample preparation, a summarize of their characteristics in terms of porosity, cement hydration, and other “material properties” is given Mechanical tests are then described and illustrated A discussion based on comparisons between the results got on the different materials lead us to the main conclusions of this study materials in which not only the polymer content is varying but also the W/C ratio In this case, it is not obvious to settle what the real influence of the polymer on the overall properties of PMMs is The compressive strength of PMMs made up with an identical W/C ratio decreases for higher polymer contents while their flexural strength increases [1,7,9] The improvement of the flexural strength may be due to the reinforcement of the Interfacial Transition Zone between the cement matrix and the aggregates [1,10], and to the bridging of the cement matrix micro-cracks by the polymer [1,2,10] In the case of mortars modified by latex rubbers, Justnes and Øye [11] showed that the latex goes into a continuous network beyond a P/C ratio of 10 wt.% This polymer network would be the origin of the improvement of the tensile and/or flexural strength of PMMs In a previous paper [2], we reported about the mechanical behaviour of styrene butadiene modified mortars of various polymer contents These composites were made with an identical W/C ratio Compression and three-point bending tests were carried out Three-point bending test results showed an increase of the flexural strength for higher polymer contents up to 7.5 wt.% P/C ratio, then a decrease beyond 10 wt.% Below 7.5 wt.% P/C ratio, the PMMs remained the relative brittle behaviour of the original mortar Beyond 10 wt.% P/C ratio, the PMMs got more and more ductile, exhibiting a strong non-linearity before peak load on the load-strain curve The compressive strength of the materials remained identical up to 10 wt.% P/C ratio, then decreased for higher polymer contents For both mechanical tests, the strain at peak load increased for higher P/C ratio while the Young’s modulus decreased Nevertheless, latex addition in cement based materials induces some Materials Polymer content is given according to the dry polymer to cement weight ratio (P/C) The considered polymer contents are P/C = 0, 5, 7.5, 10, 12.5, 15 wt.% The water to cement weight ratio (W/C) is identical for all the materials in order to ensure a similar hydration of the cement matrix On top of cement hydration, our goal was also to manage the porosity of the materials to be similar whatever the polymer content To so, we used silica fume to optimise the compactness of the granular mix: silica fume, cement, sand, assuming that a lower porosity would be easier to control A superplasticizer had to be used to ensure a good workability of the pastes Moreover, the air entrainment effect of the latex was offset by the use of an antifoamer Finally, all the samples were prepared with the following components: Portland cement 52.5 (C), fine sand of medium grain size d50 = 330 ␮m (S), silica fume (SF), water (W), latex (P), i.e an aqueous emulsion of styrene butadiene copolymer particles containing 48 wt.% of dry matter, sulfonated polymelamine based synthetic superplasticizer (SP) and an antifoaming compound (AF) The mix proportions of these components are given in Table Table Paste density, dynamic viscosity and slump test results on the fresh materials vs polymer to cement weight ratio (P/C) P/C (wt.%) (g cm−3 ) paste density theoretical density (g cm−3 ) Air content (vol.%) Dynamic viscosity (MPa s) Slump tests (mm) 7.5 10 12.5 15 2.192 ± 0.012 2.275 3.7 ± 0.5 73 ± 11.3 ≈210 2.091 ± 0.003 2.218 5.7 ± 0.1 70 ± 2.3 ≈175 2.068 ± 0.003 2.197 5.9 ± 0.1 71 ± 1.4 ≈175 2.052 ± 0.002 2.177 5.7 ± 0.1 65 ± 1.4 ≈175 2.029 ± 0.005 2.157 6.0 ± 0.2 68 ± 2.8 ≈175 2.010 ± 0.007 2.138 6.0 ± 0.3 67 ± 1.4 ≈ 175 Entrapped air content is calculated from the ratio between the paste density and the “theoretical” density S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 The materials were prepared according to the European standard NF EN 1323 Four batches a material were needed to make all the samples Paste density and dynamic viscosity measurements as well as slump tests were carried out on each batch Results are given in Table The materials have been moulded and cured for 24 h at 90% RH Then, the 40 mm × 40 mm × 160 mm prisms have been removed from mould and stored an another day at 90% RH Samples have been finally stored in a relatively dry atmosphere (50% RH and 23 ◦ C) for all the time Entrapped air content was estimated via the density of fresh pastes compared with the “theoretical” density expected for pastes without any entrapped air (Table 2) The density was determined by weighting a given volume of fresh paste The “theoretical” density was calculated from the mix proportions of the paste constituents and their own density Entrapped air content was nearly the same between the PMM pastes (volume fraction around 6%) but higher than that of the mortar (around 4%) This initial gap clearly highlights our difficulties to offset the air entrainment effect of the latex Fortunately, it has been made up for the sample drying out with time Actually, the porosity of the samples was estimated when the materials were 490 days old The porosity of the mortar samples was about 18% and that of the PMM samples about 20% [12] If the initial gap remained, it had become no more significant with time As a consequence, we assume that this difference will not play a major role in our comparisons between the mechanical properties of the materials and, thus, can be neglected Some other “material properties” were also investigated Scanning electron microscopy (SEM) of the latex distribution within the microstructure of the PMMs showed that the latex phase goes into a continuous network throughout the sample for a P/C ratio higher or equal to 10 wt.% [12] These investigations also revealed that silica fume does not react with other mineral species Fig presents a back scattered electron SEM image taken on a year old PMM Fig Back scattered electron SEM micrograph of a year old PMM with 12.5 wt.% polymer to cement ratio White arrows denote the more visible spherical silica fume particles (P/C = 12.5 wt.%) In this figure, spherical silica fume particles marked with white arrows can be easily distinguished: no signs of degradation due to a possible pozzolanic effect of added silica fume is visible Moreover, all the investigations achieved on the other materials, and especially on the mortar, never revealed any signs of chemical activation of the silica fume particles Therefore, silica fume can be seen only as a filler Setting time measurements showed that the latex delays the setting of the PMMs Electric conductivity measurements highlighted that the main reason of this is a cement hydration kinetic decrease [12] This could be due to latex particle absorption onto the cement grain surface which would reduce the surface of cement in contact with water [1,7,10] Infra-red Fourier transform (IRFT) and X-ray diffraction (XRD) analysis on the hardened materials showed however that their are all made up with the same mineral species Thus the latex would not modify the nature of the cement hydration products even if it slows down the kinetic of the reaction Experimental tests 3.1 Three-point bending tests A sketch of the experimental setup is given in Fig The three-point bending tests were performed on the 40 mm × 40 mm × 160 mm samples, using an INSTRON 4505 type machine equipped with a kN load cell and a PID servo control A strain measure was achieved on the face under tensile stress using an INSTRON extensometer put in the middle of the sample (see Fig 2) Details on the way the extensometer is tied to the sample are reported in Fig The extensometer principle is to measure the relative Fig Experimental set-up for three-point bending tests Three-point bending tests were performed under strain control (˙ε = 10−6 s−1 ) 4 S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 spread of two articulated arms during the test Divided by the initial distance between the two arms (l0 ), the extensometer gives a mean value of the strain between its arms This device is fully fitted into the INSTRON electronic interface and systematically calibrated before any new tests series The PID servo control enabled us to rule the tests under strain control The train rate was set to ε˙ = 10−6 s−1 We systematically proceeded to successive partial unloading at different strain levels to monitor the loss of sample stiffness Loading and unloading were achieved at the same strain rate At least three samples were tested for each polymer content (P/C = 0, 5, 7.5, 10, 12.5, 15 wt.%) The tests were carried out when the materials were 21 months old These tests avoided an unstable fracture of the samples, and allowed to reach the “softening” part of the material response Fig shows a typical load versus tensile strain diagram The ascending part of the curve is quasi-linear However, near the top of the curve, the slope exhibits a slight decrease witch indicates the damage initiation At peak load, the micro-cracks coalesce into an unstable crack The decreasing part of the curve corresponds to the propagation of this crack which induces a loss of sample stiffness Various parameters are used for our comparisons (see Fig 3): the maximum load (Fmax ) applied to the sample over the test, the strain at peak load (εpeak ) associated with Fmax , the initial sample stiffness (k0 ), and the one recorded during the ith unloading–loading cycle (ki ) which allowed us to characterize the damage evolution during the test Note that we not convert the applied load into stresses and the sample stiffness into elastic moduli to avoid the introduction of elastic beam calculation hypothesis which would be false in our sample geometry The non-linear evolution of the load-strain curve indicates the damage initiation and its growth In order to characterize the material strength to damage, we defined a three-point bending damage threshold This threshold (εd ) is a criterion based on a measure (δ) of the gap between the load-strain curve and its initial linear part of slope k0 If Γ is the set of Fig Typical load-strain curve of three-point bending tests (Fmax : maximum load reached over the test, εpeak : strain at Fmax , k0 : initial slope of the curve, ki : slope of the ith reload) Fig Determination of the three-point bending damage threshold δ is the gap between the experimental load-strain curve and the tangent to its initial part The damage threshold (εd , Fd ) is reached for δ = 1% points of the load-strain curve, then: ∀(ε, F) ∈ Γ, (ε, F) = (εd , Fd ) ⇔ δ(ε, F) ε − F/k0 = = 1% k0 (1) In order to determine with a good accuracy the couple of values (εd , Fd ), we used a power function H(δ) = (100.δ)3 which lowers the values of δ before the threshold and increases them beyond Fig gives an example of the (εd , Fd ) determination 3.2 Uniaxial compression tests Compression tests were carried out on 40 mm × 40 mm × 80 mm samples cut from the original 160 mm prisms (two pieces for each bending sample) The loaded faces of compression samples were ground to be parallel and flat The experimental assembly is described in Fig These tests were achieved on a MTS hydraulic/500-kN press During these experiments, we measured and recorded the applied load, the longitudinal strain (εL ) using an MTS extensometer similar to the one used for three-point bending tests, and the transversal strain (εT ) using a set-up especially designed for these tests (see Fig 4) The transversal strain determination is based on the measure of the displacements (U1 and U2 ) of two opposite sample faces This measure is achieved by two proximity sensors (sensors and 2) running on eddy currents Two aluminum targets are then needed to make the sensors working The sum of U1 and U2 divided by the initial width of the sample (a0 ) give a measure of the mean value of εT The sensors are mounted to a support smoothly fixed on the sample Compression tests were carried out under transversal strain control The strain rate was set to: εT = × 10−6 s−1 Two series of tests were achieved when the materials were 13 and 24 months old A typical stress–strain curve for compression test results is shown in Fig The stress (σ) is calculated from the applied S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 Fig Experimental set-up for compression tests The longitudinal strain (εL ) is measured by an extensometer, the transversal strain (εT ) is measured via the relative displacements (U1 , U2 ) of two opposite faces of the sample divided by its initial width (a0 ) U1 and U2 are given by sensors and The applied load is also recorded Compression tests are performed under transversal strain control (εT = × 10−6 s−1 ) load (F) divided by the sample initial section (S0 ) (σ = F /S0 ) An automatic routine of data capture and processing allows to determine the initial Young’s modulus (E0 ), its value at the ith unloading (Ei ), the maximum stress reached during the test (σ max ), the longitudinal and transversal strains peak peak (εL , εT ) associated with σ max As for the three-point bending tests, a compression damage threshold is defined in the following way: ∀(εL , εT ) ∈ Γc , = (εL , εT ) = (εdL , εdT ) ⇔ δ(εL , εT ) εT − υi (εL − εi,ir L ) = 1% εT (2) where Γ c is the set of points of the strain–strain curve, (εdL , εdT ) are the longitudinal and transversal strain at the damage threshold, υi is the Poisson’s ratio at the current loading, εi,ir is the irreversible strain at the current loading As for the three point bending damage threshold, we use the power function H(δ) = (100δ)3to determine the Fig Typical stress–strain diagram of compression tests (σ max : maximum peak peak stress recorded over the test, εL , εT : longitudinal and transversal strains at σ max , E0 , T0 : initial Young’s modulus and transversal stiffness, Ei , Ti : Young’s modulus and transversal stiffness at the ith reload) Fig Determination of the compression damage threshold in the transversal–longitudinal strains diagram νapp is the effective Poisson’s ratio for the reload where a non-linearity occurs in this diagram δ is the gap between the experimental curve and its initial tangent at this reload The damage threshold (εdL , εdT ) is reached for δ = 1% couple (εdL , εdT ) An illustration of the (εdL , εdT ) determination is given in Fig Discussion Result analysis is based on comparisons between the mechanical properties of the materials We successively analyse the variation in Young’s modulus, damage threshold, maximal stresses reached during the tests, their associated strains and, finally, the shape of the stress–strain curves according to the polymer content 4.1 Variation in Young’s modulus and damage threshold Fig presents the variation in Young’s modulus determined by compression tests according to the polymer content of the materials It is obvious that the Young’s modulus decreases almost linearly with the P/C ratio In Fig Initial Young’s modulus vs polymer to cement weight ratio (P/C) for compression and three-point bending tests 6 S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 addition, we note that there is no significant difference between materials about year old and those years old This indicates that the materials were in quite a stable state with regard to cement hydration or samples drying out In three-point bending, the Young’s modulus is identified by finite element computations from the sample initial stiffness k0 However, the value of this module is not exact since it does not take into account the possible difference of stiffness between the part of the beam sample under tensile stress and that under compressive stress in a cross-section The three-point-bending Young’s moduli thus identified are generally lower than those obtained by uniaxial compression tests This difference points out that micro-cracks pre-exist in these materials before any loading However, the difference between the Young’s moduli determined by compression and three-point bending tests, and thus the density of “pre-existing” micro-cracks, decreases for higher P/C ratio and cancels for P/C = 12.5 wt.% Fig presents the variation in damage threshold according to the polymer content For both mechanical tests, the damage threshold increases with the polymer content The compression-damage threshold measured when the materials were years old is systematically lower than that measured year before This result probably points out the ageing of the materials during this period Moreover, the bending damage threshold is lower than that measured in compression It would be possible to straight comparisons between these two damage thresholds by using a criterion on the strain tensor, but it is not obvious whether this comparison has any sense Actually, the compression damage threshold is based on a global measure of the transversal strain over the full width of the sample On the contrary, the tensile strain measure accounting for the three-point bending tests is rather local The sensibility of Fig 10 Mechanical strength (Fmax ) and load at damage threshold (Fd ) vs polymer to cement weight ratio (P/C) for three-point bending tests the strain measurements to damage is then certainly more important in the latter case than in the former Therefore, these data have to be handled with care if ones aims at establishing a damage-threshold criterion 4.2 Load and maximal stress Peak load and maximal stress reached during three-point bending and compression tests are plotted versus the P/C ratio in Figs 10 and 11 Load and stress at the damage threshold are also mentioned in these figures In three-point-bending tests, the maximum load remains relatively constant for all the composites with a P/C ratio lower or equal to 7.5 wt.% For higher polymer contents, i.e P/C ≥ 10 wt.%, the flexural strength increases with the polymer content However this increase of the flexural strength slows down for the highest polymer content For PMMs with a P/C ratio higher than 15 wt.%, a decrease of the mechanical strength can be expected Finally, we notice that the overall damage threshold tends to increase for higher polymer contents peak Fig Transversal trains at damage threshold (εdT ) and at peak load (εT ) for compression tests, and strains at damage threshold (εd ) and at peak load (εpeak ) for three-point bending tests, vs polymer to cement weight ratio (P/C) Fig 11 Mechanical strength (σ max )and stress at damage threshold (σ d ) vs polymer to cement weight ratio (P/C) for compression tests S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 The reasons accounting for the evolution of the bending strength are not obvious First of all, the porosity cannot play a major role, since the volume fraction of pores is similar throughout the materials, and especially in the PMMs The influence of a possible difference in cement hydration is not easy to handle However, it is clear that the cement hydration and the sample drying out hardly evolve between the age of and since the compressive strength of the materials is the same between these two dates (see Fig 11) So, we can consider the materials in a stable state when the three-point bending tests are performed Moreover, we know that cement hydration strengthens the cement-based materials by filling the capillary pores with hydrates, especially during the first month after their making [13] If some differences between the materials in cement hydration were responsible for the bending strength evolution, this would also have a strong influence on the evolution of the compressive strength But this is not the case Fig 11 presents the compressive strength of the materials versus the P/C ratio: it decreases almost linearly with P/C, latex addition acting like an increase of porosity due to its low rigidity Therefore, we assume that cement hydration cannot account for the bending strength evolution presented in Fig 10 Finally, the last reason we consider is the percolation of the polymer phase over the sample This phenomenon has already been evoked to explain the tensile-strength increase of polymer-modified cement-based materials [7,10] In our case, 2x10x20 mm3 pieces of material were put into hydrochloric acid in order to dissolve the mineral phase For and 7.5 wt.% P/C ratio, the pieces were dissolved in a powder probably mainly composed of the fine sand Sometimes, the dissolution of 7.5 wt.% P/C ratio pieces was not complete and sheets of residual material of a few millimetre remained, showing that the polymer starts to get organized in a continuous network at this ratio For a P/C ratio higher or equal to 10 wt.%, the pieces kept the same geometry, only their colour changed from grey to white [12] Some of these pieces were broken and put into a scanning electron microscope for energy-dispersive analysis of X-rays No more calcium was found on the fracture surfaces, hydrates were clearly dissolved [12] In conclusion, it seems clear now that the main reason for an increase of the PMMs flexural strength is polymer continuous network formation Fig 12 Envelope of stress–strain curves in compression for four polymer to cement weight ratios (P/C) ascending part of the curve breaks approximately at 1300 N) and, on the contrary, the linear shape of the ascending part of the other curves almost to the peak: the polymer seems to delay the damage initiation into the PMMs This difference between the mortar and the PMMs could be due to the existence of a film of latex on the PMM sample surfaces This film was revealed by optical microscopy analysis achieved on the PMM sample faces Our hypothesis is that this film limits the PMM sample from drying out and, therefore, from surface micro-cracking On the contrary, micro-cracks on the skin of the mortar sample, due to sample drying out, would be responsible for the early damage initiation This explanation should not be mistaken with the one accounting for the increase of the PMM flexural strength; a thin film of latex on the sample faces cannot improve the overall mechanical strength of the beam It holds for the extremely non-linear shape of the load-strain curve of the mortar which reveals that micro-cracking happens early upon loading Pre-existing skin micro-cracks due to sample drying out may be the cause of such a behaviour On the opposite, the absence of skin micro-cracks on the PMM surfaces would enable their samples to exhibit a very linear behaviour 4.3 Envelope of the stress–strain curves We cleared the stress–strain curves from their loading– unloading parts (see Figs 10 and 11) to get their envelope curve These curves sum up rather well the whole variation in the mechanical behaviour of the materials according to the polymer content For compression tests, Fig 12 clearly shows the decrease in the initial Young’s modulus as in the mechanical strength, and the increase in the longitudinal strain at peak load for higher polymer contents For three-point bending tests, we notice in Fig 13 the early initiation of damage in the mortar sample (the Fig 13 Envelope of load-strain curves in three-point bending for four polymer to cement weight ratios (P/C) 8 S Pascal et al / Materials Science and Engineering A 380 (2004) 1–8 After peak load, the curve of the wt.% P/C ratio sample falls down to the one of the mortar and exhibits the same shape for higher strain levels The curves of the 7.5 wt.% P/C ratio samples, which are not given in Fig 13, showed the same behaviour after peak load The shape of the curve of the 10 wt.% P/C ratio is quite different after peak load, and the one of the 15 wt.% P/C ratio is definitely different Therefore, as for the variation in the flexural strength (see Fig 10), the post-peak behaviour of the materials exhibits a strong change between 7.5 and 10 wt.% P/C ratio This result has to be put in connection with the percolation of latex into a continuous network for P/C ratio higher or equal to 10 wt% Conclusions The mechanical behaviour of a mortar and five PMMs have been investigated using compression and three-point bending tests In compression, the results are the ones we can expect from the addition of a phase with a low rigidity, the latex, to a mortar: the Young’s modulus and the compressive strength decrease for higher polymer contents, while the strain at the maximum stress increases, which can be seen as a gain of ductility Three-point bending test analysis gives less obvious results The flexural strength increases for P/C ratios higher or equal to 10 wt.%, and reaches a maximum for a P/C ratio equal to 15 wt.% The increase on the PMM flexural strength and the latex continuous network formation match Since a difference in cement hydration or in sample porosity cannot explain this evolution, it is suggested that the percolation of the polymer phase into a continuous network sample increases the flexural strength of the polymer modified mortars The analysis of the shape of the load-strain curves showed the early damage initiation in the mortar, unlike the PMMs whose curves exhibit a very linear ascending part We express an hypothesis to explain this change in mechanical behaviour: the presence of a film of latex on the PMM sample faces may prevent these materials from skin micro-cracking which could be the cause of the early damage initiation into the mortar Comparisons between the Young’s modulus obtained from compression and three-point bending tests revealed the existence of micro-cracks in the materials before any loading However, this difference of rigidity decreases for higher polymer content and vanishes for P/C ratios higher than 12.5 wt.% Latex addition seems then to reduce this “intrinsic” damage Acknowledgements This research was carried out in collaboration with the Centre de Recherche d’Aubervilliers de RHODIA The authors are grateful to G Orange and M Sari References [1] J.B Kardon, J Mater Civil Eng (1997) 85 [2] L Bureau, A Alliche, P Pilvin, S Pascal, Mater Sci Eng A 308 (2001) 233 [3] Y Ohama, Classification of concrete-polymer composites, in: S Chandra, Y Ohama (Eds.), Polymers in Concrete, vol 5, CRC Press, London, 1994, pp 81–109 [4] Y Ohama, H Ibe, H Mine, K Kato, Rubber Chem Technol 37 (1964) 758 [5] R.J Folic, V.S Radonjanin, ACI Mater J 95 (1998) 463 [6] F.A Shaker, A.S El Dieb, M.M Reda, Cem Concr Res 25 (1997) 711 [7] E Sakaă, J Sugita, Cem Concr Res 25 (1995) 127 [8] J Shulz, O Killermann, Cem Concr Res 31 (2001) 357 [9] S Sujjanavich, J.R Lundy, ACI Mater J 95 (1998) 131 [10] Y Ohama, Properties of concrete-polymer composites, in: S Chandra, Y Ohama (Eds.), Polymers in Concrete, vol 6, CRC Press, London, 1994, pp 111–145 [11] H Justnes, B.A Øye, Nordic Concr Res (1990) 69 [12] S Pascal, Comportement mécanique de composites mortier polymère, Ph.D Thesis, École Centrale Paris, Paris, 2002 [13] A Folliot, M Buil, La structuration progressive de la Pierre de ciment, in: J Baron, R Sauterey (Eds.), Le béton hydraulique: connaissance et pratique, vol 12, Presses de l’ENPC, Paris, 1982  ... ratio of 10 wt.% This polymer network would be the origin of the improvement of the tensile and/or flexural strength of PMMs In a previous paper [2], we reported about the mechanical behaviour of. .. suggested that the percolation of the polymer phase into a continuous network sample increases the flexural strength of the polymer modified mortars The analysis of the shape of the load-strain curves... comparisons between the mechanical properties of PMMs of various polymer contents may not only reveal the influence of the polymer phase, but also include the influence of these adverse effects