The Effect of Recycled Sand on the Tensile Properties of Engineered Cementitious Composites

: This research aims to investigate the feasibility of replacing natural sand (NS) with recycled sand (RS) to enhance the mechanical property of engineered cementitious composites (ECC). For a comparative study, ECCs incorporating natural sands (NS) and recycled sands with different sieve sizes were taken as experimental subjects. The results demonstrated that RS-ECC possessed better tensile properties featuring saturated cracks and superior strain-hardening behavior than that of NS-ECC. The highest tensile strain capacity of RS-ECC was up to 7%. Meanwhile, the compressive and ﬂexural strengths of RS-ECC were over 50 and 20 MPa. The pseudo-strain-hardening (PSH) index of the RS-ECC-20 grid and RS-ECC-12 grid were 141 and 201, which increased by 46% and 70% than that of NS-ECC. Furthermore, the thicker weak ITZ and comparatively aggregate/ITZ ratio were found in the RE-ECC by a microstructure test, which revealed and explained the mechanism for the lower matrix fracture toughness of RS-ECC.


Introduction
Concrete is the largest artificial material used in civil engineering at present. The production of concrete consumes an enormous amount of fine aggregate, e.g., river sand and manufactured sand. The overexploitation of river sand brings a series of environmental problems. Hence, some governments in the world have issued rules to limit sand mining from rivers, e.g., the No. 320 Decree in China [1]. Manufactured sand is generally made from crushing stone by artificial processes. Besides consuming natural resources, the production of manufactured sand is considered to generate considerably large CO 2 emissions [2,3]. Meanwhile, construction and demolition (C&D) waste, produced by demolishing abandoned structures, takes a relatively large proportion of the total amount of waste in recent years. The most common disposal method of C&D waste in some developing countries is landfilled, inevitably resulting in water, atmospheric and soil contamination as well as landfill saturation [4,5]. Using C&D waste to produce recycled construction materials, e.g., recycled aggregate, is a reasonable method to minimize the negative influence on end-of-life structures, as well as to decrease the demand for extraction of resources. In processing C&D waste, recycled sand (RS) accounts for about 20% of the gross amount of recycled aggregate [6][7][8][9]. Despite RS being abundant in C&D waste, the understanding of concrete incorporated with RS is limited. In recent years, many investigators have attempted to investigate the influence of RS replacement on the mechanical property of concrete. Khatib [10] conducted a series of experiments to investigate the replacement ratios of fine recycled aggregate (FRA) on the mechanical property, which led to 15% and 30% compressive loss at replacement ratios of 25% and 100%, respectively. Evangelista et al. [11] stated that there were no significant reductions with percentages of substitution • Achieve a better understanding of recycled sand applicable to ECC. X-ray fluorescence (XRF), scanning electron microscope (SEM) and particle size analysis were used to obtain chemical compositions, physical properties, and fineness of RS. • Investigate the influence of RS replacement on the basic mechanical properties of ECC. • Find the influence of RS replacement on the meso-scale and micro-scale properties of RS-ECC and explore the connection between the RS-ECCs tensile strain capacity and meso-scale property.

Materials and Mix Proportions
The mixture proportions of the ECC used for the study are presented in Table 1. The mixture proportions differ from each other in RS content or sieve size. Specifically, in two reference mixtures, i.e., NS-ECC-12 grid and NS-ECC-20 grid (NS-ECC), NS with sieve size 12 grid and 20 grid were added as fine aggregate, respectively. In RS-ECC-12 grid and RS-ECC-20 grid, all contents of the NS were replaced by RS with corresponding particle sizes. In all the mixtures, the binder includes P.II 52.5 Portland cement and class F fly ash. Table 2 lists the chemical compositions of cement, fly ash, and RS, which were acquired by XRD analysis. A set of sieves was employed to obtain the particle size distribution of RS and NS used in the mixture. Figure 1 illustrates the analysis results. Due to the relatively high water absorption of RS, a more high-range water reducer (HRWR) was applied in mixing RS-ECC to ensure workability, as shown in Table 1. Ultra-high molecular weight polyethylene (PE) fibers were used as reinforcement in all the mixtures with a constant volume fraction of 2%. The property of PE is presented in Table 3.

Preparation of Specimens
All the mixtures were mixed in a 20 L mixer. The binder and water were continuously mixed for 2 min. HRWR was added into the mixer and mixed to reach proper mortar fluidity. Finally, fibers were added by hand and mixed for 2 min to ensure fiber dispersion well. Afterward, the fresh ECC was cast into molds. All specimens were demolded after 24 h and then cured in the air for another 27 days.

Mechanical Properties Tests
The test scheme is presented in Table 4, listing the number of specimens in each test case. Dog-bone plates ( Figure 2) were adopted as specimens for uniaxial tension tests. A WDW-300Y electro-servo machine (1 N resolution) was employed to apply displacementcontrol loading with a uniform rate of 1.5 mm/min. Two linear variable displacement transducers (LVDTs) were installed on both sides of the specimen to obtain the displacement of specimens, and the gauge length was 80 mm. Six cubic specimens of every mixture were prepared to assess the compressive prop erty of RS-ECC. The size of the geometric dimension was 40 × 40 × 40 mm. The test proce dure is according to the GB/T 50081-2016.
Three-point bending tests were also applied to prism specimens of every mixture in an MTS 244 electro-hydraulic servo machine. The geometric dimension of the prism spec imens was 40 × 40 × 160 mm. The test procedure is according to the GB/T 50081-2016. To Six cubic specimens of every mixture were prepared to assess the compressive property of RS-ECC. The size of the geometric dimension was 40 × 40 × 40 mm. The test procedure is according to the GB/T 50081-2016. Three-point bending tests were also applied to prism specimens of every mixture in an MTS 244 electro-hydraulic servo machine. The geometric dimension of the prism specimens was 40 × 40 × 160 mm. The test procedure is according to the GB/T 50081-2016. To obtain full-field deformation, the digital image correlation (DIC) method was used for acquiring the deformation of specimens of normal composites (with fiber).
Single-crack tensile tests were conducted to examine the interfacial strength between the fiber and matrix. In Figure 3, two symmetrical notches with a width less than 0.6 mm were cut by an ultra-thin saw blade (0.4 mm in thickness), which is conducive for a single crack to form. Figure 4 illustrates the geometric size of the specimen. To avoid excrescent cracks forming inside or outside the notches and to ensure an ideal condition that only a single crack will be generated under tensile load, the adopted notched cross-section had an area of 50% of its original value. In Figure 3, the crack opening displacement was measured by a pair of clip-on gauges. Six cubic specimens of every mixture were prepared to assess the compressive prop erty of RS-ECC. The size of the geometric dimension was 40 × 40 × 40 mm. The test proce dure is according to the GB/T 50081-2016.
Three-point bending tests were also applied to prism specimens of every mixture in an MTS 244 electro-hydraulic servo machine. The geometric dimension of the prism spec imens was 40 × 40 × 160 mm. The test procedure is according to the GB/T 50081-2016. To obtain full-field deformation, the digital image correlation (DIC) method was used fo acquiring the deformation of specimens of normal composites (with fiber).
Single-crack tensile tests were conducted to examine the interfacial strength between the fiber and matrix. In Figure 3, two symmetrical notches with a width less than 0.6 mm were cut by an ultra-thin saw blade (0.4 mm in thickness), which is conducive for a single crack to form. Figure 4 illustrates the geometric size of the specimen. To avoid excrescen cracks forming inside or outside the notches and to ensure an ideal condition that only a single crack will be generated under tensile load, the adopted notched cross-section had an area of 50% of its original value. In Figure 3, the crack opening displacement was meas ured by a pair of clip-on gauges.

Notches
Clip-on gauge  Three-point loading tests were conducted to analyze the influence of RS pos the matrix fracture toughness. The geometry of the notched beam is shown in Fig  As shown in Figure 6, a SANSI electro-hydraulic servo machine was applied in the with a displacement-control loading rate of 1.0 mm/minute. To reveal the format different fracture toughness between RS-ECC and NS-ECC, a 3D-shape measureme and nanoindentation test were performed on the fracture surfaces of the notched after bending tests. The 3D shape measurement instrumentation with grating proje as shown in Figure 7, was used to obtain the photo of the fracture surface. The mech of the instrument is that two symmetrical projectors project the grating on the fr surface, which will be received by the camera after diffuse reflection on the surfac altitude of every point on the surface can be obtained by analyzing the transforma the grating, mainly the change in the width of the stripes. Combining it with the coordinates of points, 3D coordinate matrices of the fracture surface were obtained 354 Three-point loading tests were conducted to analyze the influence of RS posing on the matrix fracture toughness. The geometry of the notched beam is shown in Figure 5. As shown in Figure 6, a SANSI electro-hydraulic servo machine was applied in the tests, with a displacement-control loading rate of 1.0 mm/minute. To reveal the formation of different fracture toughness between RS-ECC and NS-ECC, a 3D-shape measurement test and nanoindentation test were performed on the fracture surfaces of the notched beams after bending tests. The 3D shape measurement instrumentation with grating projection, as shown in Figure 7, was used to obtain the photo of the fracture surface. The mechanism of the instrument is that two symmetrical projectors project the grating on the fracture surface, which will be received by the camera after diffuse reflection on the surface. The altitude of every point on the surface can be obtained by analyzing the transformation of the grating, mainly the change in the width of the stripes. Combining it with the plain coordinates of points, 3D coordinate matrices of the fracture surface were obtained.
after bending tests. The 3D shape measurement instrumentation with grating projection, as shown in Figure 7, was used to obtain the photo of the fracture surface. The mechanism of the instrument is that two symmetrical projectors project the grating on the fracture surface, which will be received by the camera after diffuse reflection on the surface. The altitude of every point on the surface can be obtained by analyzing the transformation of the grating, mainly the change in the width of the stripes. Combining it with the plain coordinates of points, 3D coordinate matrices of the fracture surface were obtained.   of the instrument is that two symmetrical projectors project the grating on the fractur surface, which will be received by the camera after diffuse reflection on the surface. The altitude of every point on the surface can be obtained by analyzing the transformation o the grating, mainly the change in the width of the stripes. Combining it with the plain coordinates of points, 3D coordinate matrices of the fracture surface were obtained.   The nano-indentation tests were conducted by using the Hysitron Triboindenter fitted with a Berkovitch tip. The loading process had a loading rate of 300 μN/s, maintaining a constant force of 1.5 mN for 2 s, and an unloading rate of 300 μN/s. The analytical Oliver and Pharr method were used to calculate the indentation hardness and elastic modulus of sands, interfacial transition zone (ITZ), and cement matrix of ECC.  The nano-indentation tests were conducted by using the Hysitron Triboindenter fitted with a Berkovitch tip. The loading process had a loading rate of 300 µN/s, maintaining a constant force of 1.5 mN for 2 s, and an unloading rate of 300 µN/s. The analytical Oliver and Pharr method were used to calculate the indentation hardness and elastic modulus of sands, interfacial transition zone (ITZ), and cement matrix of ECC. Figure 8 exhibits the residual crack pattern of all four mixtures. The number of cracks in the 80 mm gauge length was counted after the test. Based on the crack numbers and the obtained tensile deformation at peak strength, the average crack width and crack spacing were calculated. Table 5 shows the average values. The crack numbers of RS-ECC, i.e., 44.5 of 20 grid and 37.1 of 12 grid, are larger than those of NS-ECC, i.e., 29.7 of 20 grid and 26.4 of 12 grid, implying the better multi-cracking characteristic of RS-ECC. Meanwhile, the RS-ECC might possess fine penetration resistance and excellent durability even under extreme stretching [32], since the crack widths of RS-ECC of both 20 grid and 12 grid are smaller than those of NS-ECC. When comparing the effect of particle size on crack behavior (Table 5), it is seen that the mixture with 20 grid sand has better performances in both tensile strain capacity and crack distribution than the one with 12 grid, which is compliant with the idea that a larger size of aggregate in ECC leads to lower mechanical property [33].

Tensile Behavior
were calculated. Table 5 shows the average values. The crack numbers of 44.5 of 20 grid and 37.1 of 12 grid, are larger than those of NS-ECC, i.e., 29.7 o 26.4 of 12 grid, implying the better multi-cracking characteristic of RS-ECC the RS-ECC might possess fine penetration resistance and excellent durabilit extreme stretching [32], since the crack widths of RS-ECC of both 20 grid an smaller than those of NS-ECC. When comparing the effect of particle size on ior (Table 5), it is seen that the mixture with 20 grid sand has better perform tensile strain capacity and crack distribution than the one with 12 grid, which with the idea that a larger size of aggregate in ECC leads to lower mechan [33].    Figure 9 and Table 6 present the summaries for the tensile properties of the tested specimens. The average peak tensile strength of the RS-ECC-20 grid, RS-ECC-12 grid, NS-ECC-20 grid, and NS-ECC-12 grid were 4.75, 4.18, 3.96, and 3.86 MPa, respectively. RS-ECC had better tensile strength than NS-ECC. Meanwhile, the average tensile strain at the peak strength of those four mixtures reached 7.00%, 5.52%, 4.64%, and 4.78%, respectively, indicating again the better multi-cracking behavior of RS-ECC. The peak tensile stress over the first cracking stress ratios is presented in Table 6. This ratio of RS-ECC was also a bit larger than that of NS-ECC and they all surpassed two 2 on average. Combining these three aspects, it could be concluded that RS-ECC has a relatively strong strain-hardening characteristic that enables it to have both superior tensile strain capacity and strength.
peak strength of those four mixtures reached 7.00%, 5.52%, 4.64%, and 4.78%, respectively, indicating again the better multi-cracking behavior of RS-ECC. The peak tensile stress over the first cracking stress ratios is presented in Table 6. This ratio of RS-ECC was also a bit larger than that of NS-ECC and they all surpassed two 2 on average. Combining these three aspects, it could be concluded that RS-ECC has a relatively strong strain-hardening characteristic that enables it to have both superior tensile strain capacity and strength.    From the scatter plot in Figure 9, the tensile property of RS-ECC has a larger variance, although it behaves better in all aspects of the tensile test on average. In addition, even using the same kind of sand, the mixture with finer sand had a smaller variance than the one with a coarser one. This might be the result of the effect that the fineness and shape of the aggregate pose on the dispersion of the fiber. Generally, the finer aggregate leads to more even fiber dispersion [33]. Furthermore, the irregular morphology of RS, as shown in Figure 14, may lead to uneven and unstable distribution, which attributes to the large deviation of tensile behavior of RS-ECC. Table 7 summarizes the results of all the uniaxial compression and three-point bending tests. Mixture ID including plain in Table 7 refers to the specimens without fiber. The average peak uniaxial compression stresses were 35.91 MPa for the RS-plain-20 grid and 42.50 MPa for the RS-plain-12 grid, which were similar to their NS counterparts in magnitude. The flexural strength of the RS-plain 20 grid and RS-plain 12 grid was 7.55 and 7.43 MPa, slightly lower than the NS-plain 20 grid and NS-plain 12 grid, which was 8.87 and 8.86 MPa, respectively. Combining the results of both tests, a conclusion can be drawn that RS replacement decreases the mechanical properties of concrete if no fiber reinforcement is incorporated.  Table 7 shows the compressive and flexural strength of materials reinforced with fibers. Comparing the test results of RS-ECC to NS-ECC, the compressive strength of the NS-ECC-20 grid, i.e., 74.73 MPa, is obviously larger than the other composites, not only indicating that both fineness and type of sand will influence the compressive strength but also proving the factor of fineness weights over the type.

Compressive and Bending Properties
Similar to the result of the ECC matrix, the flexural strength of RS-ECC is also slightly smaller than that of NS-ECC. Moreover, the deformations at the midspan of RS-ECC measured by the DIC method are significantly larger than those of NS-ECC, indicating a better multi-cracking behavior and strain-hardening effect that is consistent with the results obtained in the uniaxial tensile test; see Figure 10. To give a clearer view, one typical loaddeformation curve of each mixture case is presented in Figure 11. It is well observed that more cracks were generated throughout the test process on RS-ECC specimens, especially the 20 grid mixture case.
In summary, the mechanical strength of RS-ECC is close to or slightly smaller, but the tensile strain capacity and flexural deformability of RS-ECC are significantly improved as compared with those of NS-ECC. To explore the underlying reason, further analyses were conducted at the meso-scale, as shown in the following section.    In summary, the mechanical strength of RS-ECC is close to or slightly sma the tensile strain capacity and flexural deformability of RS-ECC are significa proved as compared with those of NS-ECC. To explore the underlying reason analyses were conducted at the meso-scale, as shown in the following section.

Discussion on the Effect of RS Replacement at the Meso-Scale
The typical stress vs. crack mouth opening displacement (CMOD) curves o gle crack tensile test are illustrated in Figure 12. In Equation (1), the complemen ergy Jb' is computed from the hatched area, as shown in Figure 12. Jb′ reflects th consumed when fibers are pulled out from the matrix. The CMOD curves of the f tures are shown in Figure 13. The different colored lines in Figure 13 demons different specimens in each groups. Table 8 lists the key parameters of the test.

Discussion on the Effect of RS Replacement at the Meso-Scale
The typical stress vs. crack mouth opening displacement (CMOD) curves of the single crack tensile test are illustrated in Figure 12. In Equation (1), the complementary energy J b is computed from the hatched area, as shown in Figure 12. J b reflects the energy consumed when fibers are pulled out from the matrix. The CMOD curves of the four mixtures are shown in Figure 13. The different colored lines in Figure 13 demonstrate the different specimens in each groups. Table 8 lists the key parameters of the test. (1) inability 2022, 14, x FOR PEER REVIEW    The fiber bridging stress σ OC suffered almost no loss when NS was replaced by RS (in Figure 6), which indicates that RS has little adverse effect on the bond-slip capacity between the fiber and matrix.
The fracture toughness K Q and the fracture energy J tip of the three-point bending tests are calculated by the following equations [34]. (3) The K Q and J tip are listed in Table 9. The average fracture toughness of the RS-plain-20 grid and RS-plain-12 grid was 0.393 and 0.363 MPa·m 1/2 , obviously lower than that of the NS-plain-20 grid and NS-plain-12 grid, which was 0.514 and 0.463 MPa·m 1/2 , respectively. In comparison to NS, RS has the ability to reduce the matrix fracture toughness. When focusing on the effect on fracture energy brought by fineness of the aggregate, it is well known that finer aggregate leads to smaller fracture energy [35]. Consistently, specimens of the 20 grid had smaller fracture energy, as compared to those of the 12 grid. To combine two aspects into one parameter that can measure the strain-hardening effect of a certain kind of composite, according to the pseudo-strain-hardening (PSH) index [27], the tensile strain capacity of cementitious composites is tightly associated with the fiber bridging complementary energy J b and the matrix fracture energy J tip . To be more specific, a relatively lower ratio of J tip /J b tends to easily lead to crack generation and propagation, which finally facilitate the tensile capacity of the composite. As can be seen in Table 10, the replacement of NS by RS efficiently reduces the fracture toughness of the matrix and keeps the fiber bridging capacity J b nearly constant. In addition, it can be concluded that the fracture toughness of the matrix could be reduced by coarse sand (12 grid), not only for RS but also for NS. As a result, the PSH index J b /J tip increases by 46% and 70% after the replacement of NS by RS of the 20 and 12 grids, respectively. There is a similar trend between the J b /J tip and the tensile ductility. The impact of RS replacement is reflected in the enhanced pseudo-strain-hardening (PSH) index J b /J tip , therefore explaining the positive influence of RS on the tensile strain capacity of ECC.   In summary, the enhanced property of RS-ECC is attributed to the reduced fracture toughness of the matrix. According to previous studies, there are two possible explanations for this phenomenon. The first one is due to the different patterns of crack propagation [36,37]. In NS-plain, cracks may bypass the aggregate when it is on the way to crack propagation since the natural sand is too hard and too tough to be torn apart. Therefore, cracks leave a fracture surface with a larger area, which requires more energy. However, it is different in RS-plain because the structure of RS is weaker and thus easier to be torn apart. Thus, the crack can easily go through the aggregate with relatively smaller stress and form a fracture surface with a smaller area. The other reason is the weak nature of the ITZ between the new cement binder and the old mortar on the RS surface [38]. Instead of directly going through sand, the crack needs less stress as well as energy to break the ITZ and tear the sand apart from the cement binder in RS-ECC. Therefore, the fracture toughness and energy of the RS-plain are smaller than those of the NS-plain.
To figure out which reason is dominant, SEM, three-dimensional shape measurement and nanoindentation techniques were used for further study. The SEM photos of RS and NS are presented in Figure 14. RS has a rougher and more porous surface in comparison to NS. The special morphology of RS is the result of the residual old mortar during demolition. In summary, the enhanced property of RS-ECC is attributed to the reduced fracture toughness of the matrix. According to previous studies, there are two possible explanations for this phenomenon. The first one is due to the different patterns of crack propagation [36,37]. In NS-plain, cracks may bypass the aggregate when it is on the way to crack propagation since the natural sand is too hard and too tough to be torn apart. Therefore, cracks leave a fracture surface with a larger area, which requires more energy. However, it is different in RS-plain because the structure of RS is weaker and thus easier to be torn apart. Thus, the crack can easily go through the aggregate with relatively smaller stress and form a fracture surface with a smaller area. The other reason is the weak nature of the ITZ between the new cement binder and the old mortar on the RS surface [38]. Instead of directly going through sand, the crack needs less stress as well as energy to break the ITZ and tear the sand apart from the cement binder in RS-ECC. Therefore, the fracture toughness and energy of the RS-plain are smaller than those of the NS-plain.
To figure out which reason is dominant, SEM, three-dimensional shape measurement and nanoindentation techniques were used for further study. The SEM photos of RS and NS are presented in Figure 14. RS has a rougher and more porous surface in comparison to NS. The special morphology of RS is the result of the residual old mortar during demolition. A three-dimensional shape measurement test was introduced to obtain 3D coordinate matrices of the fracture surface. Every data point in the coordinate matrix is linked to the adjacent points to build a triangle with the area of about 0.03 mm 2 , as shown in Figure 15, which all together form a numerical fracture surface whose area can be easily calculated. Figure 16 shows a contour map of the RS-ECCs fracture surface cut by its base plane. The degree of warmth of color reflects the relative position and distance of a certain datum point to the base plane. As shown in Figure 17, most enhancement rates of the fracture area range from 3 to 7%, which are considered of little influence on the fracture energy at such a large scale. Therefore, the first explanation, i.e., a different pattern of crack propagation, is ruled out. The only explanation for the decrease in fracture toughness brought by RS replacement should be the weakened interface between the new cement binder and RS, which is consistent with the results given by Xiao et al. [38].
To support the above explanation, an insight into the microstructure of ECC was obtained by SEM. Figure 18 shows the microstructure characterization of NS-ECC and RS-ECC at 28 days. An obvious ITZ around the RS is observed, which is much thicker than that of NS-ECC. The weak ITZ makes cracks go through the aggregate with smaller stress and leads to a reduction in fracture toughness of the matrix. Figure 19 shows the average elastic modulus of a different phase of ECC. It can be seen due to the presence of the residual mortar that the gap in elastic modulus between sand and ITZ is much bigger than that of NS. The recycled sands act as "incompatible inclusion" in the matrix, causing serious stress concentration, which to some extent resembles the artificial flaws used to produce ECC [39,40]. Microcracks will be more easily generated and propagated in these regions, thus reducing the fracture toughness of composites.
plane. The degree of warmth of color reflects the relative position and distance of a certain datum point to the base plane. As shown in Figure 17, most enhancement rates of the fracture area range from 3 to 7%, which are considered of little influence on the fracture energy at such a large scale. Therefore, the first explanation, i.e., a different pattern of crack propagation, is ruled out. The only explanation for the decrease in fracture toughness brought by RS replacement should be the weakened interface between the new cement binder and RS, which is consistent with the results given by Xiao et al. [38].   calculated. Figure 16 shows a contour map of the RS-ECCs fracture surface cut by its base plane. The degree of warmth of color reflects the relative position and distance of a certain datum point to the base plane. As shown in Figure 17, most enhancement rates of the fracture area range from 3 to 7%, which are considered of little influence on the fracture energy at such a large scale. Therefore, the first explanation, i.e., a different pattern of crack propagation, is ruled out. The only explanation for the decrease in fracture toughness brought by RS replacement should be the weakened interface between the new cement binder and RS, which is consistent with the results given by Xiao et al. [38].   To support the above explanation, an insight into the microstructure of ECC was obtained by SEM. Figure 18 shows the microstructure characterization of NS-ECC and RS-ECC at 28 days. An obvious ITZ around the RS is observed, which is much thicker than that of NS-ECC. The weak ITZ makes cracks go through the aggregate with smaller stress and leads to a reduction in fracture toughness of the matrix. Figure 19 shows the average Figure 17. Parameters of the fracture surface. elastic modulus of a different phase of ECC. It can be seen due to the presence of the r sidual mortar that the gap in elastic modulus between sand and ITZ is much bigger tha that of NS. The recycled sands act as "incompatible inclusion" in the matrix, causing se ous stress concentration, which to some extent resembles the artificial flaws used to pr duce ECC [39,40]. Microcracks will be more easily generated and propagated in these r gions, thus reducing the fracture toughness of composites.

Conclusions
In the present research, the effect of recycled sand on the mechanical properties of engineered cementitious composites (ECC) was investigated. The strain-hardening mechanism of ECC with recycled sand was explored by the scanning electron microscope (SEM) and nano-indentation tests. The following conclusions could be obtained: (1) The ECC with RS showed excellent tensile ductility with a tensile strain of 7%, which is greater than that of the ECC with NS. Furthermore, the incorporation of RS into ECC maintained the same tensile strength as that of NS-ECC. (2) The compressive and flexural strengths of ECC were over 50 and 20 MPa. The addition of RS did not cause a significant loss of compressive and flexural strength, compared with that of ECC with NS. (3) The addition of RS decreased the matrix toughness and kept the fiber-bridging capacity constant. The pseudo-strain-hardening (PSH) of RS-ECC showed the highest value, which could benefit tensile ductility. The main reason for this was that the old residual mortar around the RS increased the gap of elastic modulus between the sand and ITZ, which decreased the matrix toughness and improved the tensile ductility.

Conclusions
In the present research, the effect of recycled sand on the mechanical properties of engineered cementitious composites (ECC) was investigated. The strain-hardening mechanism of ECC with recycled sand was explored by the scanning electron microscope (SEM) and nano-indentation tests. The following conclusions could be obtained: (1) The ECC with RS showed excellent tensile ductility with a tensile strain of 7%, which is greater than that of the ECC with NS. Furthermore, the incorporation of RS into ECC maintained the same tensile strength as that of NS-ECC. (2) The compressive and flexural strengths of ECC were over 50 and 20 MPa. The addition of RS did not cause a significant loss of compressive and flexural strength, compared with that of ECC with NS.
(3) The addition of RS decreased the matrix toughness and kept the fiber-bridging capacity constant. The pseudo-strain-hardening (PSH) of RS-ECC showed the highest value, which could benefit tensile ductility. The main reason for this was that the old residual mortar around the RS increased the gap of elastic modulus between the sand and ITZ, which decreased the matrix toughness and improved the tensile ductility.