Integrated interface parameters of recycled aggregate concrete

Construction and Building Materials 101 (2015) 861–877

Contents lists available at ScienceDirect

Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat

Integrated interface parameters of recycled aggregate concrete Hongru Zhang, Yuxi Zhao ⇑ Institute of Structural Engineering, Zhejiang University, Hangzhou 310058, Zhejiang, PR China

h i g h l i g h t s
Micro properties of diverse interfaces in recycled concrete (RC) are explored.
Mechanical and durability properties of RC are weakened as RA replacements increase.
An integrated interface parameter of RC is proposed.
The rationality of the proposed interface parameter is discussed and proven.
Mechanisms of the macro inferiority of RC are quantitatively analyzed.

a r t i c l e

i n f o

Article history: Received 13 June 2015 Received in revised form 27 August 2015 Accepted 15 October 2015

Keywords: Recycled aggregate concrete Nanoindentation Integrated interface parameters Correlation analysis

a b s t r a c t An integrated interface parameter of recycled aggregate concrete (RC), n, is proposed, based on experimental study of three RC groups with the recycled coarse aggregate (RA) replacements of 0%, 50% and 100%, respectively. Experimental study includes micro properties of diverse types of aggregate/ mortar interfaces in RC, i.e., the length, width, and elastic modulus, and the macro properties of concrete materials, i.e., the compressive strength, density degree and chloride ion diffusion coefficient. The rationality of the proposed n is proven by the comprehensive relevant interface information and its good correlations with both RA replacement ratios and concrete macro properties, respectively. n-based model prediction of macro concrete properties was also found reliable. The mechanism of RC’s inferiority to natural aggregate concrete can thereby be quantitatively revealed based on n. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Recycled aggregate concrete (RC) using recycled aggregates (RA) has raised worldwide interest because of its great potential benefits in controlling the over-discharge of construction and demolition (C&D) wastes. The use of RC may also reduce the consumption of natural resources, including the landfills used for disposing of these wastes and the natural aggregate (NA) for producing natural aggregate concrete (NC) [1]. Previous research verified that RC is generally inferior to NC in its material properties on the macro scale, including workability [2–6], mechanical properties [7–11], and durability [3,12–14]. To reveal the inferiority mechanism of RC in the macro properties compared to NC, the interface transition zone (ITZ) is the key because ITZs are treated as the primary weak points in cementbased materials [15,16]. With the introduction of RA, more types of ITZs are produced in RC than in NC, as illustrated in Fig. 1. According to Fig. 1, only one type of ITZ exists in NC, lying between ⇑ Corresponding author. E-mail address: [email protected] (Y. Zhao). http://dx.doi.org/10.1016/j.conbuildmat.2015.10.084 0950-0618/Ó 2015 Elsevier Ltd. All rights reserved.

the virgin aggregate and the new cement mortar, which is labeled as ITZ1 in this study. By contrast, in RC, e.g., RC with the 50% replacement of RA, there exist three types of ITZs as follows: the ITZ1s between the NA and the new mortar, the ITZ2s inside the RA between the old virgin aggregate and the surrounding old mortar, and the ITZ3s between the old mortar and the new mortar. It should be noted that RA are usually obtained by crushing and grinding, through which the old cement mortar may be partially or even completely removed from some RA particles, and some old ITZs, i.e., ITZ2s, may have also been damaged or even removed. Therefore, when these RAs with the embedded virgin aggregate partially exposed are cast into concrete mixtures, new ITZs will form between the exposed virgin aggregate and the new cement mortar. This new type of ITZs can be treated as the same with ITZ1s, because both of them are interfaces between virgin aggregate and new cement mortar. In a word, in RC, there do exist three types of ITZs: ITZ1s and the ITZ3s are newly formed interfaces after casting, while ITZ2s are the old ITZs originally existing in RA. The diverse types of ITZs in RC and the porous old mortar adhering to RA are potential weak points in RC, which can weaken the properties of RC on the macro scale. Furthermore, in Fig. 1,

862

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

(a)

(b)

Fig. 1. Schematic diagram of the microstructure of (a) NC and (b) RC with the partial replacement of RA.

certain proportions determined after several trials to ensure that the employed RA and NA were similar in both size ranges and size distributions. The size distributions of NA and RA are illustrated in Fig. 2. According to Fig. 2 the two groups of coarse aggregates employed in this study demonstrated the similar size distribution. The percentage of particles below 16 mm for RA was slightly higher than that for NA, which indicated that smaller particles in RA accounted a larger proportion than in NA. This may be attributed to the crushing operation through the procedure for producing RA, as through crushing the old cement mortar was likely to fracture and be broken into smaller particles. Still the difference between the two curves in Fig. 2 is acceptable, therefore the similar size gradation of the two groups of coarse aggregates on micro properties of concrete could be negligible. The crushing value and the mud content of the employed RA and NA were tested according to The National Specification Pebble and Crushed Stone for Construction (GB/T 14685-2011). It was found that the crushing value of RA was higher than that of NA; the crushing values of RA and NA were 11.3% and 8.0%, respectively. The presence of the weak, porous old mortar adhering to the old virgin aggregate of RA may have affected the surface strength of RA and as a result the anti-crushing property of RA was weakened. Besides, the mud contents were 3.5% and 0.6%, for RA and NA, respectively. Such differences were also mainly induced by crushing through RA’s production, during which the old mortar was partly ground to become fine particles. A natural water absorption of coarse aggregates was defined and measured in this study, which was employed in the mix design method to adjust the water amount for the RC mixture to ensure good RC workability. The natural water absorption in this study, wn , was defined as the water absorption (by weight) of the naturally air-dried (AD) coarse aggregates, rather than the oven-dried (OD) aggregates. wn was calculated as follows:

wn ¼

m1  m0  100% m0

ð1Þ

where m0 was the weight of each of the three air-dried (AD) samples for RA and NA. Each AD sample weighed approximately 5.0 kg. The three samples were soaked in water for 24 h and then removed from the water; excess water was absorbed by a wet cloth until the samples were surface saturated dried (SSD). m1 was the weight of each SSD aggregate sample.

100

80

60 2. Materials 2.1. Concrete materials 2.1.1. Coarse aggregate The RA employed in this study was purchased from a plant in Shanghai, China. It was observed that the old cement mortar had been partially removed from some RA particles through RA’s producing procedure, rendering the old virgin aggregate in these RA particles partially exposed to the atmosphere. New ITZs would form between these exposed old virgin aggregate and the new cement mortar in concrete mixtures, as has been proposed as ITZ1s in the Introduction Section. The NA used in this study was crushed limestone ranging from 5 to 25 mm in size. Since there was no ready-made RA with the same size range as NA produced by this RA manufacturer, the employed RA was obtained by mixing two groups of RAs with different size ranges (5–15 mm and 15–25 mm). The mixture was obtained according to

40

20

NA RA

0 30

25

20

15

10

5

0

Sieve size, mm Fig. 2. The size distribution of NA and RA employed in this study.

Cumulative sieve rate, %

RA-VG represents the old virgin aggregate in RA, while RA-OM represents the old mortar adhering to the old virgin aggregate in RA. Researchers have begun to explore the properties of ITZs in RC and their influence on macro properties. For example, C. S. Poon verified that the interface between RA and cement is relatively loose compared with the interface formed between NA and cement [17]. Ryu found that the compressive strength of RC depends on the relative quality of old and new ITZs in RC. When the watercement ratio is low, the strength characteristic of the concrete is governed by the effect of old ITZs. When the water-cement ratio is high, the predominant governing factors are the new ITZs [18]. Xiao et al. found that the ratio of the old ITZ’s mechanical properties (elastic modulus and strength) to those of the old mortar matrix affects the stress-strain curves and failure processes of modeled recycled aggregate concrete (MRAC). Increasing the ratio resulted in higher strength, but lower ductility. The ratio of new ITZ’s mechanical properties to those of the new mortar matrix has a negligible effect on the compressive strength of MRAC, while the tensile strength of MRAC increases as this ratio increases [19]. However, these studies tend to be limited to just one or two single property parameters of ITZs, which means that they cannot comprehensively and reliably reflect both the geometric and mechanical features of all the diverse types of ITZs in RC, thereby the mechanism of RC inferiority to NC in macro properties could not be sufficiently revealed. Therefore, the focus of this paper is to propose a reasonable and reliable interface parameter, to comprehensively reflect the geometric and mechanical properties of the diverse types of ITZs contained in RC, and furthermore, to quantitatively investigate the influence of multiple ITZ properties on the macro properties of concrete materials. The rationality of the proposed interface parameter of RC is discussed from multiple angles. The proposed interface parameters will reveal the essential differences between RC and NC in interface features. In addition, our findings connect the interface properties and the macro material properties, so that the mechanism of RC’s inferiority to NC can be quantitatively investigated on the macro scale.

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877 The natural water absorptions of RA and NA were 1.2% and 0.2%, respectively. The presence of the porous adhering old mortar and the abundant mud in RA had led to the higher natural water absorption of RA. 2.1.2. Other materials The cement employed in this study was ordinary Portland cement with a grade of 42.5 MPa. The fine aggregate (FA) was natural sand with a fineness modulus of 2.6 and moisture content of 7.2%. Tap water was employed and no superplastizer was added. 2.2. Mixture proportions Three different replacement percentages of RA (0%, 50%, and 100%) were employed; the corresponding concrete groups were labeled as NC, RC-50% and RC-100%, respectively. To ensure good RC workability while adjusting the effective water-to-binder (w/b) ratio in the three concrete groups, more water was added to RC-50% and RC-100%, according to the larger natural water absorption, wn, of RA than that of NA. As introduced in Section 2.1.1, the difference between wn for RA and NA was 1.0%; therefore, the extra water added to RC was 1.0% of the weight of RA. Natural water absorption was used rather than water absorption measured according to the Specification of Pebble and Crushed Stone for Building (GB/T 146852011) to ensure that the coarse aggregates were employed under the AD condition rather than the OD condition. The mixture proportions of the three concrete groups are listed in Table 1; the water added to RC-50% and RC-100% is also presented. It should be noted that in this study the extra water was added through mixing, together with the basic water amount. The coarse aggregates, the fine aggregates and the cement were first added together, while water was the last to be added. The target compressive strength of concrete was 30 MPa. The slump of NC, RC-50%, and RC-100%, measured according to Standard for Test Method of Mechanical Properties on Ordinary Concrete (GB/T 50081-2002), was 64 mm, 95 mm and 85 mm, respectively. The slump of the two RC groups was larger than that of NC, which was because the water added into RC could not be completely absorbed by RA through mixing, whose duration was approximately 3 min. The initial natural water absorption of the employed RA samples (in AD condition) in 3 min was measured as 0.9%, which was lower than the 24-h water absorption, which was 1.2%. Therefore, through concrete mixing, not all the added water was absorbed by the RA particles. Some water, possibly 0.3% of them, was still left in the RC mixture and thereby the free water amount as well as the effective w/c ratio of RC at its fresh state increased. As a result, the measured slump of RC was larger than the target, and was larger than that of NA. Besides, through mixing, cement particles were prone to absorbing most of the water, to first form cement paste, because cement had better hydrophilicity than other materials in concrete, i.e., the coarse aggregates and the fine aggregates. It would be more difficult for RA particles to absorb water when surrounded by the fresh cement paste than in the purewater environment. As a result, the water absorption of RA particles during the 3min mixing may be even less than 0.9%, and more free water was left in the fresh cement mortar. This is another possible reason for the enhanced slump of RC than expected. The effects of such results on the micro and macro properties of concrete are further discussed in Sections 4.1 and 4.2. The friction between the more rough RA particles may weaken the fluidity of fresh concrete, as a result, though more free water may be contained in RC-100% than in RC-50%, the slump of RC-100% was smaller than that of RC-50%. After casting, all concrete specimens were moist cured at a temperature of 20 °C and a relative humidity of more than 95%, for 28 days. Afterwards, the specimens were stored at the room temperature, exposed directly to the atmosphere.

3. Test methods 3.1. Micro properties In this study, both the geometrical and micromechanical properties of the three different types of ITZs in RC were measured at different ages to comprehensively determine the interface properties and their influence on the macro properties of concrete. 3.1.1. Instruments To determine the width and the elastic modulus of the diverse types of ITZs contained in the three concrete groups, respectively, a Table 1 Mixture proportions of concrete produced in this study (units: kg/m3). Concrete type

Cement

RA

NA

FA

Water

NC RC-50% RC-100%

400 400 400

0 563 1125

1125 563 0

648 648 648

177.6 177.6 + 5.6 177.6 + 11.2

863

Hysitron 950 Triboindenter fitted with a Berkovitch tip (tip radius of 0.6 lm and angle of 142.3°) was utilized to administer nanoindentation tests across different ITZs. The high resolution probe microscopy installed in the Triboindenter allowed observation and in-situ image capturing prior to indentation, thereby it was feasible to locate the desired indent areas precisely. A Canon imageCLASS D520 optical scanner was employed to scan sections of a series of concrete slices to obtain highresolution optical images. All of the specimen sections were 30  30 mm squares; the sample processing procedure is discussed in detail in Section 3.1.2. Matlab and various CAD tools were used to obtain the coarse aggregates areas contained in each 30  30 mm concrete section. More importantly, the tools were used to separately obtain the lengths of different types of ITZs in these sections. The testing procedure details are introduced in Section 3.1.3. 3.1.2. Sample processing At the age of 3 days, 28 days and 90 days, for each of the three concrete groups, small concrete slices with dimensions of 10  10  5 mm were cut from the inner part of the 100 mm concrete cubes, to prepare the nanoindentation samples. It was necessary to carefully grind and polish these concrete slices prior to indentation, to obtain adequately even and smooth testing surfaces, thereby to ensure the reliability of nanoindentation results [20]. Since water was used as the cooling agent through cutting, to stop cement hydration at the certain ages, these slices were immersed in pure ethanol immediately after cutting, for 24 h. Afterwards, they were dried in an oven at a temperature of 45 °C till their weight stabilized. The samples were then embedded in epoxy resin and ground on the Buehler-Met paper discs, with gradations of 280 (51.8 lm), 800 (22.1 lm), 1200 (14.5 lm), and 1500 (12.2 lm), successively. Then these samples were polished on a series of Buehler TexMet pads, using diamond suspensions with gradations of 9 lm, 6 lm, and 3 lm, successively. An 2-min ultrasonic bath cleaning in ethanol was then applied on the polished samples to remove the tiny particles adhering to these samples. In this study, the procedure for sample processing mainly referred to Xiao et al.’s work [21]. To obtain the length information of the diverse types of ITZs in the three concrete groups, a series of 30  30 mm concrete slices with a thickness of 8 to 10 mm were cut from a series of 30  30  150 mm concrete blocks. For each of the three concrete groups, 20–30 concrete slices at each of the three testing ages, 3 days, 28 days, and 90 days, were obtained. The prepared and processed samples for nanoindentation and optical scanning are shown in Fig. 3a and b, respectively. Epoxy resin surrounding the concrete specimens, as shown in Fig. 3, was used to protect the relatively fragile structure of the concrete against cutting, grinding and polishing. 3.1.3. Testing procedure 3.1.3.1. Young’s modulus of ITZs, mortar, and virgin aggregate. For the nanoindentation tests, the indent areas across ITZ1s and ITZ2s were 100  100 lm squares, while the areas across ITZ3s were 150  100 lm rectangles to ensure that the indent areas completely covered the ITZ3s. On the 100  100 lm and the 150  100 lm indent areas, a 21  11 matrix and a 31  11 matrix were set for the grid indents, respectively. The indent areas and the corresponding matrices designed for the different ITZs are shown in Fig. 4a and b, respectively. For each type of ITZ at each testing age, 3 randomly located indent areas across the target ITZs were selected. Furthermore, on both the old mortar and the new mortar in the 3 concrete groups, five 30  30 lm areas with a 4  4 matrix were randomly selected for nanoindentation tests, as shown in Fig. 4c.

864

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

3.2. Macro properties 3.2.1. Compressive strength For each of the three concrete groups, three 150-mm cubes were tested to get the compressive strength, f c , at the testing ages of 3 days, 28 days, 90 days, 180 days, 360 days, and 720 days, respectively, based on Standard for Test methods of Mechanical Properties on Ordinary Concrete (GB/T 50081-2002).

(a)

(b)

Fig. 3. Concrete samples employed for micro property tests: (a) the 10  10 mm concrete slices for nanoindentation tests, and (b) the 30  30 mm concrete slices for optical scanning.

The Young’s modulus of each indentation was calculated as follows:


1 E ¼ ð1  v 2 Þ  1=Er  ð1  v 2i Þ=Ei

ð2Þ

where the indenter’s elastic modulus, Ei , is 1140 GPa and its Pisson’s ratio, v i , is 0.07. The Poisson’s ratio of the tested samples, v, is suggested as 0.2 for RAC [22], which is the same of NAC. The reduced elastic modulus, Er , is obtained from nanoindentation tests. 3.1.3.2. Length of ITZs and areas of aggregates. After optical images (see Fig. 5a) of the twenty to thirty 30  30 mm concrete sections were achieved for each concrete group at each of the three testing ages, image binarization was administered with the help of Matlab, according to the different grey scales of the virgin aggregate and mortars (see Fig. 5b). Afterwards, the outline of the virgin aggregates, including both the new virgin aggregate (NA) and the old virgin aggregate inside RA, could be extracted. Therefore, the sum of the length of ITZ1s and ITZ2s, l1 þ l2 , contained in each 30  30 mm concrete section was obtained, and the total areas of the new and old virgin aggregate, ANA + ARA-VG, which were enclosed by ITZ1s and ITZ2s, were also achieved. The limitation of the Matlab image processing was that the grey scales of the old and new mortars were so comparable that it was difficult to separately identify or illustrate each mortar in the binarized images. As a result, the Matlab codes computed in this study could not offer the length of ITZ3s. To solve this problem, CAD software by Autodesk was employed to assist with the profile extraction of the old mortar in RC, so that the length of ITZ2s and ITZ3s contained in the 30  30 mm concrete sections was achieved separately. Meanwhile, the area of the old mortar solely enclosed by ITZ3s or by both ITZ2s and ITZ3s was obtained, as shown in Fig. 5c. In Fig. 5c, the green curves represent the outline of ITZ2, and the yellow curves represent the outline of ITZ3; both curve groups were manually drawn in CAD using the polyline with small segments. Using the processing procedure above, the length of different interfaces, i.e., ITZ1s, ITZ2s, and ITZ3s contained in each 30  30 mm concrete section, were separately obtained and labeled as l1 , l2 , and l3 , respectively. The areas of different components of coarse aggregates contained in these 30  30 mm sections, i.e., the new virgin aggregate, the old virgin aggregate, and the old mortar, were also obtained, which were labeled as ANA , ARAVG , and ARAOM , respectively, in this study.

3.2.2. Density degree It is common knowledge that ultrasonic wave travels faster in a denser media than in a looser one. Therefore, the density degree of concrete can be reflected by the average speed, v of ultrasonic wave travelling through the concrete specimens. Based on the Testing Code for Hydraulic Concrete (SL 352-2006), the ultrasonic tests were carried out for the three groups of concrete, at the ages of 3 days, 28 days, 90 days, 180 days, 360 days, and 720 days. Three 150 mm cubes for each concrete group at each age were employed, and the locations of the five pairs of testing points on the noncasting surfaces are shown in Fig. 6. 3.2.3. Chloride ion diffusion coefficient The rapid chloride migration (RCM) method was used to achieve the Cl diffusion coefficients, D, of the three concrete groups, at the ages of 28 days, 90 days, 180 days, 360 days, and 720 days, respectively, according to Standard for Test Methods of Long-term Performance and Durability of Ordinary Concrete (GB/T 50082-2009). Three concrete cylinders with a diameter of 100 ± 1 mm and a height of 50 ± 2 mm were taken for the RCM tests for each concrete group at each of the five testing ages.

4. Results and discussion 4.1. Micro properties 4.1.1. Widths of ITZs By administering grid indentation across the different types of ITZs, the modulus of each indent of the indent matrix was achieved. Afterwards, the 2-D distribution map of each indent area across the different types of ITZs was plotted, as in the example shown in Fig. 7 where different colors represent the different modulus ranges. The ‘‘denser zone” shown in Fig. 7c between the old and new mortar while surrounding the old mortar was supposed to be a layer of CaCO3 crystals produced through surface carbonation of RA, during the period when RA was directly exposed to the atmosphere, prior to concrete casting [21]; the ‘‘denser zone” was not part of the ITZ3s. Based on the modulus distribution maps, the left and right boundaries of the ITZs were determined, as marked by two red dashed curves in Fig. 7; therefore, the width of these ITZs could be measured. The widths of ITZ2s were found stable, and they were not affected by either the RA replacements or the age; the average width of ITZ2s was approximately 40 lm. Age did not affect the ITZ2s widths because cement hydration in ITZ2s is not significant as the age of RA is usually decades, and the cement particles are typically consumed. By contrast, the widths of ITZ1s and ITZ3s were different in the different concrete groups and changed with concrete age. The average widths of ITZ1s and ITZ3s in the three concrete groups at the ages of 3 days, 28 days, and 90 days are shown in Fig. 8. From Fig. 8, it is clear that the interface width decreased for both ITZ1s and ITZ3s from the age of 3 through 90 days. The constant cement hydration in ITZ1s and ITZ3s could help fill the pores in them, leading to the drop in ITZs’ widths.

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

865

(a)

(b)

(c) Fig. 4. Indent areas and the corresponding indent matrices on different phases: (a) the indent matrix across ITZ1s and ITZ2s, (b) the indent matrix across ITZ3s, and (c) the indent matrix on the old mortar, the new mortar and the virgin aggregate

Generally, at any of the three testing ages, the width of ITZ1s increased as the replacement percentages of RA increased. Meanwhile, the width of ITZ3s in RC-100% was also greater than that in RC-50%. As referred to in Section 2.2, extra water was added to RC-50% and RC-100% according to the larger natural water absorption, wn , of RA than NA. More importantly, the extra water was directly added through mixing. Although some of the extra water may have been absorbed by RA through mixing, some non-absorbed water likely remained in the newly cast cement mortar, as implied by the slump data of the three concrete groups introduced in Section 2.2. Therefore, the effective w/b ratio of the new mortar in RC-50% and RC-100% was larger compared with their NC counterpart. Adverse effects were thereby induced by the larger effective w/b ratio on the formation of denser hydration

products in the new mortar, ITZ1s and ITZ3s; as a result, the widths of the newly formed ITZs, i.e., ITZ1s and ITZ3s, increased as the RA replacement ratios increased. The larger widths of ITZ1s and ITZ3s in RC-100% than in RC-50% also may be related to the larger mud content in RA than in NA, as stated in Section 2.1.1. More mud adhering to RA likely weakened the binding between RA and the new mortar, so that the widths of ITZ1s between the old virgin aggregate and the new mortar, and of ITZ3s around the old mortar were enlarged [23]. 4.1.2. Elastic modulus of ITZs Once the left and right boundaries of the target ITZs were determined in the modulus distribution maps of the indent areas across the ITZs, the elastic modulus data obtained from those indents

866

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

ITZ2 ITZ3

(a)

(b)

(c)

Fig. 5. Processing of the images of the 30  30 mm concrete sections through optical scanning: (a) the optical image of one RC-50% sample, (b) the binarized image of the RC-50% sample, and (c) the CAD-processed image of the RC-50% sample.

Fig. 6. Locations of the five pairs of testing points in the ultrasonic wave tests.

within the boundaries were collected. Statistical analyses were conducted to determine the probability distribution of the modulus of these ITZs. The probability distribution curves of the elastic modulus of ITZ1s and ITZ3s at the ages of 3 days, 28 days, and 90 days are summarized in Fig. 9; the bin-size was set as 5 GPa. Based on Fig. 9, from 3 days through 90 days, mechanical properties of ITZ1s and ITZ3s developed. The peaks of the relative frequency of occurrence curves of the elastic modulus of both the ITZ1s and ITZ3s shifted to the right as time passed, e.g., for ITZ1s in the three concrete groups, all of the peaks at 3 days fell in the interval of 10 to 15 GPa, while at 28 days and 90 days, the peaks fell in the interval of 15 to 20 GPa. Moreover, the probability of smaller modulus decreased over time, e.g., the modulus probabilities below 10 GPa of ITZ1s in NC, RC-50% and RC-100% at 3 days were 0.232, 0.282, and 0.295, respectively, while at 28 days, they were 0.182, 0.208, and 0.250, respectively. The indentation modulus range of the different phases in the ITZ is assumed as follows: porosity (67 GPa), other hydration product mainly including C-SH and ettringite (7–34 GPa), CH (34–50 GPa) and unhydrated cement (P50 GPa) [24–26]. Hence, such results indicated that as time passed, the mechanical properties of C-S-H in ITZ1s and ITZ3s developed; meanwhile, the porosity in the two types of ITZs decreased. Continuous cement hydration in these ITZs was the predominant cause of the enhancement of the newly formed ITZs. It can also be seen in Fig. 9 that at each age, the more NA replaced by RA, the larger the probability of small modulus values (modulus below 10 GPa) obtained in ITZ1s and ITZ3s, indicating a larger porosity of these ITZs. Such differences were also closely related to the extra water added to RC compared with NC, which

may have enlarged the w/b ratio in these newly formed ITZs in RC, as previously explained in Section 4.1.1. The elastic modulus of both ITZ1s and ITZ3s was approximately subjected to the Gaussian distribution, according to Fig. 9. The average elastic modulus, as well as its upper and lower limits of the confidence intervals at a confidence level of 90% were calculated and are shown in Fig. 10 for ITZ1s and ITZ3s at 3 days, 28 days, and 90 days for the different concrete groups. It can also be seen in Fig. 10 that differences between concrete groups with different replacement percentages of RA occurred in the average elastic modulus of their ITZ1s or ITZ3s at an early age (3 days) and decreased with the RA replacement growth. However, at 28 days and 90 days, the difference was not significant. The extra water added as mixing water for RC may have led to a larger w/b ratio in the newly formed ITZs, i.e., the ITZ1s and ITZ3s, which resulted in the inferior mechanical properties of these ITZs at an early age. However, after 28 days, moist curing ceased and all of the specimens were directly exposed to the atmosphere when the beneficial ‘‘reservoir effects” of the old mortar began and the water absorbed by and stored in the old mortar was released to the surrounding ITZs or to the new cement mortar. So, in RC-50% and RC-100%, cement hydration in ITZ1s, ITZ3s, and the new mortar after the 28-day curing period developed faster as the released water from the old mortar guaranteed water supply for cement hydration. Hence, after 28 days when the external water supply was stopped, the faster cement hydration in ITZs of RC than NC caused the adverse effects of the larger w/b ratio induced by the addition of water through mixing, leading to the comparable mechanical properties of ITZs between RC and NC. According to Fig. 10, at the age of 3 days, the average modulus of ITZ3s was smaller than that of ITZ1s; however, as time passed, ITZ3s showed faster modulus gain than ITZ1s. The modulus of ITZ3s surpassed that of ITZ1s at 28 days and 90 days because the beneficial effects of the old mortar by inducing the ‘‘reservoir effects” on ITZ3s were more significant than those on the ITZ1s. The ITZ3s were on the surface of the old mortar, thus cement hydration in ITZ3s was more likely enhanced by the water released from the old mortar. Different from the newly formed ITZs, i.e., ITZ1s and ITZ3s, ITZ2s that originally existed between the old virgin aggregate and the adhering old cement mortar showed time-stability in their mechanical properties, according to the nanoindentation results because the cement hydration in ITZ2s was no longer significant when unhydrated cement particles were consumed during the decades-long age. Fig. 11 gives an elastic modulus probability distribution example of ITZ2s in RC-50% at 3 days, 28 days, and 90 days, respectively. The confidence interval of the elastic modulus of ITZ2s was 17.74–18.80 GPa at a confidence level of 90%; the mean value was 18. 27 GPa.

867

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

100 1.500 19.00

Distance (µm)

80

ITZ1

Natural virgin aggregate

60

36.50 54.00 71.50 89.00

New mortar

106.5 124.0 141.5

40

y 20

x

0 0

20

40

60

80

100

Distance (µm)

(a) 100 0.000 18.11 32.79

Distance (µm)

80

Old mortar

Old virgin aggregate

60

49.18 65.57 81.96

ITZ2

115.5 140.5 153.0

40

y 20

x

0 0

20

40

60

80

100

Distance (µm)

(b) 100 0.5000 18.00

Distance (µm)

80

27.00

Old mortar

35.00 89.75 112.4

ITZ3

60

134.8 157.1 179.5

40

y new mortar

20

0 0

20

40

Denser Zone

60

80

100

120

x

140

Distance (µm)

(c) Fig. 7. The modulus distribution map within the indent area (a) across ITZ1 of one NC sample at the age of 28 days, (b)across ITZ2, and (c) across ITZ3 of one RC-50% sample at the age of 28 days.

868

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

70 65

Width (µm)

age. The heterogeneity of concrete or, more precisely, the nonuniform spatial distribution of coarse aggregate in the concrete P specimen, may have led to the large fluctuation of li . To eliminate the influence of spatial heterogeneity of concrete on the geometric features of ITZs, a new length index, Rl=A is defined as follows:

NC-ITZ1 RC-50%-ITZ1 RC-100%-ITZ1 RC-50%-ITZ3 RC-100%-ITZ3

75

60

Rl=A ¼

55 50 45 40 0

20

40

60

80

100

Age (day) Fig. 8. Widths of ITZ1s and ITZ3s of the three concrete groups at three different ages.

4.1.3. Elastic modulus of new mortar, old mortar and virgin aggregate The elastic modulus of the new mortar in the three groups of concrete developed from 3 days through 90 days, as illustrated by Fig. 12. Similar with the nanoindentation results obtained on ITZ1s and ITZ3s, the elastic modulus of the new mortar decreased as the replacement percentage of RA increased, especially at 3 days; however, after 28 days, the differences among the three concrete groups were not significant. These results were explained in Section 4.1.1. They indicated that even though a larger w/b ratio was induced by adding more water to RC through mixing, a process that led to the inferior mechanical properties of ITZ1s, ITZ3s, and the new mortar at an early age, such adverse effects were gradually overcome with the passage of time. The old mortar in RC-50% and RC-100% did not develop with time. The elastic modulus of the old mortar was also proved to be subject to the Gaussian distribution, and the confidence interval of the average modulus of the old mortar was 23.96–25.51 GPa at a confidence level of 90%.The elastic modulus of the aggregates was also tested in this study. Based on statistical analysis of the nanoindentation results, the confidence intervals of the average elastic modulus were 68.64–71.26 GPa and 70.54–73.82 GPa for the old virgin aggregate and the new virgin aggregate, respectively, at a confidence level of 90%. The slight inferiority of the old virgin aggregate to the NA was likely due to the accumulating injuries of the old virgin aggregate in the parent concrete under longterm service; however, such inferiority was so small that it was ignored in this study. Based on the nanoindentation results discussed above, the average elastic modulus of the old mortar was approximately 35% of that of the old virgin aggregate surrounded by the old mortar. Therefore, the adhering old mortar had indeed lowered the overall mechanical properties of RA, compared with NA. Hence, apart from extra types of ITZs, the old mortar was also a weak point in RC. 4.1.4. Length of ITZs and areas of aggregates By employing the methods introduced in Section 3.1.3, the length of each type of ITZs contained in the 30  30 mm concrete sections was obtained. The average value of the length sum of all of the different types of ITZs contained in these concrete sections for the three groups of concrete at 3 days, 28 days, and 90 days are summarized in Fig. 13. The length data of ITZs shown in Fig. 13 fluctuate widely. No P obvious trends were observed between the value of li and the P age, t, nor between li and the RA replacement ratio at a certain

X

li =ðANA þ ARAVG þ kARAOM Þ

ð3Þ

P where li , ANA , ARAVG , and ARAOM are the sum of the length of the ITZs, the areas of the new virgin aggregate, the old virgin aggregate and the old mortar contained in the tested 30 mm30 mm concrete sections, respectively. In NC there exists only one type of coarse aggregate in NC, i.e., NA, thereby for NC the denominator in Eq. (3) includes only the area of virgin aggregate. By contrast, in RC, apart from the virgin aggregate, i.e., the NA and the old virgin aggregate embedded in RA particles, the other component of RA particles, i.e., the old cement mortar adhering to RA, should also be considered when calculating Rl=A . However, the area of the old cement mortar should be adjusted, to a certain proportion which should be related to the difference between the old cement mortar and the virgin aggregate in mechanical properties, so that the dominators of Eq. (3) can be physically comparable between the NA groups and the RA groups. From this point of view, k, which is defined as the effective area coefficient for the old mortar, is applied to ARAOM , in order to transfer the area of old cement mortar to that of the virgin aggregate. In this study, k is set as the ratio of the elastic modulus of the old mortar to that of the old virgin aggregate in RA, which is 35%, as introduced in Section 3.1.3. k can also reflect the adverse effects of the old adhering cement mortar on the overall properties of RA, compared with NA. Therefore, Rl=A represents the length sum of all types of ITZs by introducing a unit area of virgin aggregate. The calculated Rl=A for each of the three concrete groups at 3 days, 28 days, and 90 days, respectively, are shown in Fig. 14. In Fig. 14, little fluctuations are found in the values of Rl=A among the three different testing ages for one certain concrete group, which has indicated the time-stability of Rl=A . Based on this finding, the Rl=A data obtained at the three different ages was merged and analyzed together, regardless of the age influence. Dashed lines in different colors mark the average Rl=A for each concrete group in Fig. 14. It can be observed that Rl=A increased as the replacement percentage of RA increased, generally indicating that more ITZs were produced in concrete when more NA were replaced by RA. 4.2. Macro properties 4.2.1. Compressive strength Compressive strength of the three concrete groups at different ages is shown in Fig. 15. According to Fig. 15, the compressive strength of the three concrete groups all developed over time. Based on the discussion in Section 4.1, cement hydration in the three concrete groups induced the drop in width and the gain in the mechanical properties of the newly formed ITZs, i.e., the ITZ1s and ITZ3s. The cement hydration also enhanced the new mortar, which contributed to the concrete strength gain over time. Furthermore, the compressive strength decreased as the RA replacement increased. Although after 28 days the average elastic modulus of the newly formed ITZs and the new mortar was comparable among the three concrete groups, the geometric indexes of ITZs measured in this study, i.e., the width of ITZ1s and ITZ3s, and the specific perimeter of all the ITZs, Rl=A , were still larger in RC than in NC. Therefore, the larger amount of ITZs provided more destructive paths against compressive loads, thereby resulting in the lower macro compressive strength of the RC specimens.

869

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877 0.30

0.30

NC-ITZ1(3 d) RC-50%-ITZ1(3 d) RC-100%-ITZ1(3 d)

0.25

0.20

Probability

Probability

0.20

0.15

0.10

0.15

0.10

0.05

0.05

0.00

0.00 0

5

10

15

20

25

30

35

40

45

0

5

10

15

20

25

Modulus (GPa)

Modulus (GPa)

(a)

(b)

0.30

30

35

40

45

0.35

NC-ITZ1(90 d) RC-50%-ITZ1(90 d) RC-100%-ITZ1(90 d)

0.25

RC-50%-ITZ3(3 d) RC-100%-ITZ3(3 d)

0.30 0.25

0.20

0.15

Probability

Probability

NC-ITZ1(28 d) RC-50%-ITZ1(28 d) RC-100%-ITZ1(28 d)

0.25

0.10

0.20 0.15 0.10

0.05 0.05

0.00

0.00

0

5

10

15

20

25

30

35

40

45

0

5

10

15

20

Modulus (GPa)

25

30

35

40

45

50

55

Modulus (GPa)

(c)

(d)

0.35

RC-50%-ITZ3(28 d) RC-100%-ITZ3(28 d)

0.30

0.35

RC-50%-ITZ3(90 d) RC-100%-ITZ3(90 d)

0.30

0.25

Probability

Probability

0.25

0.20 0.15 0.10

0.20 0.15 0.10 0.05

0.05

0.00

0.00 0

5

10

15

20

25

30

35

40

45

50

55

Modulus (GPa)

(e)

-0.05 0

5

10

15

20

25

30

35

40

45

50

55

Modulus (GPa)

(f)

Fig. 9. Probability distribution of the elastic modulus of (a)–(c) ITZ1s and of (d)-(f) ITZ3s at the ages of 3 days, 28 days, and 90 days, respectively.

4.2.2. Density degree As mentioned in Section 3.2.2, the velocity of the ultrasonic wave, v, travelling through the concrete specimens can reflect the density degree of the tested concrete. The average ultrasonic wave velocity traveling through the three groups of concrete specimens is shown in Fig. 16.

As the RA replacement increased, more ITZs and more old mortar were introduced into the corresponding concrete. The ITZs were typically more porous than the cement mortar or the virgin aggregate, so the density degree of RC was affected. This explained why the average ultrasonic wave velocity through RC-100% and RC-50% was smaller than that through NC, as illustrated by

870

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

Fig. 10. The mean value of the elastic modulus of ITZ1s and ITZ3s at the ages of 3 days, 28 days, and 90 days.

750 0.25

RC-50%-ITZ2(3 d) RC-50%-ITZ2(28 d) RC-50%-ITZ2(90 d)

=50%

=100%

mm

650 550

li

Probability

0.20

=0%

0.15

450 0.10

350 0

30

t, d

60

90

0.05 Fig. 13. Total length of all diverse ITZs contained in the 30  30 mm concrete sections for the three concrete groups in this study at the ages of 3 days, 28 days, and 90 days, respectively.

0.00 0

5

10

15

20

25

30

35

40

45

50

1.5

=0%

Modulus (GPa)

=50%

=100%

1.3

Rl/A, mm-1

Fig. 11. Probability distribution of the elastic modulus of ITZ2s in RC-50% and RC100% at the age of 3 days.

1.1 0.9 0.7 0.5 0

30

60

90

t, d Fig. 14. The values of Rl=A in the three concrete groups in this study at the ages of 3 days, 28 days, and 90 days, respectively.

Fig. 12. The mean value of the elastic modulus of the new mortar in the three concrete groups at the ages of 3 days, 28 days, and 90 days, respectively.

Fig. 16. As time passed, the products of cement hydration helped fill the pores in the ITZs and in the old mortar, which contributed to the rise of concrete density; therefore, the ultrasonic wave velocity through the three concrete groups increased with age. 4.2.3. Chloride ion diffusion coefficient The Cl diffusion coefficient, D, reflects the anti-Cl diffusion property of concrete. The larger D is, the easier the Cl diffusion will be. Fig. 17 illustrates the Cl diffusion coefficients of the three concrete groups studied in this research at different ages. Similarly, at any of the three testing ages, the Cl diffusion coefficient of RC-50% and RC-100% was larger than that of NC,

Compressive strength, fc, (MPa)

45 40 35 30 25 20

NC RC-50% RC-100%

15 10 -100

0

100

200

300

400

500

600

700

800

Age (d) Fig. 15. Compressive strength of the three concrete groups at different ages.

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

The target integrated interface parameter in this study should, at a minimum, meet the following three requirements to ensure rationality:

Ultrasonic wave velocity , v, (km/s)

4.8 4.6


The properties of all the diverse types of ITZs should be comprehensively considered in the interface parameter.
The interface parameter should demonstrate a high correlation with the RA replacement ratio so that the proposed interface parameters can be treated as a key to describing the differences between NC and RC, and the differences among RC groups with different RA replacements on the micro scale.
The target interface parameter should be well correlated with the macro properties of concrete. Therefore, the interface parameter can be used to reveal the mechanisms behind the macro inferiority of RC to NC in a quantitative manner.

4.4 4.2 4.0 3.8

NC RC-50% RC-100%

3.6 3.4 -100

0

100

200

300

400

500

600

700

800

Age (d) Fig. 16. Ultrasonic wave velocity of the three concrete groups at different ages.

Cl- diffusion coefficient , D, (10-12m2/s)

871

10

P3

NC RC-50% RC-100%

9

n¼

8 7 6 5 4 0

100

200

300

400

500

600

700

Besides, it would be beneficial if the selected interface parameter were time-dependent one, so that the prediction models of RC’s macro properties over time can be established. According to the requirements listed above, one integrated interface parameter is finally proposed, after several trials on its form design. The proposed interface parameter of RC is expressed as follows:

800

Age (d) Fig. 17. Cl diffusion coefficients of the three concrete groups at different ages.

indicating that it was easier for Cl to penetrate into RC than into NC. This was likely because more ITZs and porous old cement mortar in RC provided more paths for the Cl diffusion. The inferior anti-Cl diffusion property of RC may lead to earlier rebar depassivation and series corrosion-induced injuries of concrete structures, thereby shortening the service life of RC structures when compared with that of NC structures. 4.3. Integrated interface parameters of RC and rationality analysis 4.3.1. The proposed interface parameters of RC Based on the discussion above, given the concert mixture proportions designed in this study, both the geometric indexes, including the length and width of ITZs, and the micromechanical properties of ITZs worked together to determined the inferiority of RC to NC in macro material properties. It was also found that the effects of the geometric and micromechanical properties of ITZs on macro concrete properties fluctuated with age. In order to quantitatively explain how interface properties influence the macro properties of RC, the integrated interface parameter of RC that comprehensively considers the geometric parameters and micromechanical properties of the diverse types of ITZs in RC is proposed in this section, meanwhile the rationality of the proposed interface parameter is discussed from different angles.

1 li di;t ðANA þ ARAVG þ kARAOM Þ

ð4Þ

where li , ANA , ARAVG , ARAOM and k have been introduced in Section 4.1.4, which have constituted the defined length index of interfaces in RC, i.e., Rl=A . Hence, in the expression of n, the different types of RA, i.e., the RA particles with virgin aggregate partially exposed or completely coated by the old cement mortar, and some pure cement mortar particles without embedded virgin aggregate, are all reflected, because the areas of both the two components of RA, i.e., the virgin aggregate and the old cement mortar, are contained in the denominator of Eq. (4). n can be treated as a more advanced interface parameter on the basis of Rl=A , which has introduced an extra interface property index, di;t (unit: mm), as the average width of the Type i ITZ at an age of t (unit: d). The values of di;t can refer to Fig. 8 in Section 4.1.1. It can be seen that n is a non-dimensional parameter. n reflects the sum of all ITZ areas induced by a unit area of virgin coarse aggregates in concrete. The properties of parent concrete can also influence the strength of RA, especially the strength of the old cement mortar adhering to RA, thereby can influence the macro properties of RC. Hence, mechanical properties of the old cement mortar were once considered when determining the integrated interface parameter, which was once labeled as n0 . In the expression of n0 , as introduced in the appendix of this paper, the mechanical properties of both the old cement mortar and the diverse types of ITZs in RC were involved, in order to make the interface parameter physically reliable and more representative. The rationality analysis of n0 , please refer to the appendix. As can be seen in the appendix section, though n0 has contained more property information of the old cement mortar and of the interfaces, the correlations between n0 and the macro material properties of RC were insignificant, at least when compared with the finally chosen n. The reason for such insignificant correlations lies on that the gain in elastic modulus and the loss in width of each type of ITZs typically occur together, which can be seen in Sections 4.1.1 and 4.1.2. As a result, when the elastic modulus and the width of ITZs were both involved, the development of n0 would become disordered, and its effects on macro properties of RC would be altered. Therefore, n0 was not as reliable as n so it was not chosen as the integrated interface parameter of RC. In the following parts of this paper, the rationality of only n will be introduced. However, in further study, we will continue working

872

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

on the optimization of the proposed interface parameter, in order to make it more comprehensive and more reliable.

0.08 0.07

4.3.2. Time-dependence analysis of the proposed interface parameter The values of n obtained at the three testing ages, i.e., 3 days, 28 days, and 90 days, are illustrated in Fig. 18. It was found that a logarithmic fit best represented the correlations between n and the age of concrete, t (unit: d). The fitted curves are plotted in Fig. 18, and the fitted equations are summarized in Eq. (5).

nq¼100% ðtÞ ¼ 7  103 t þ 0:0686;

ð5:1Þ

R2 ¼ 0:9976 R2 ¼ 0:9971

ð5:3Þ

4.3.3. Correlation between the proposed interface parameter and the percentage replacement of RA At each of the three testing ages, the values of n at different RA replacement ratios, are shown in Fig. 19. A linear correlation was found between n and q. The linearly fitted results, as illustrated in Fig. 19, can be expressed as:

R2 ¼ 0:9231

nt¼28 d ðqÞ ¼ 1  10 q þ 0:0470; 4

nt¼90 d ðqÞ ¼ 1  10 q þ 0:0375; 4

3d

28 d

90 d

0.03 0

20

40

60

ð6:1Þ

2

R ¼ 1:0000 R ¼ 0:9758

and ITZ3s, increased as q increased. Since n can be treated as the total area of all diverse types of ITZs in concrete induced by a unit area of virgin coarse aggregate, the positive linear correlation between n and q indicates that the total amount of ITZs would rise, as more NA is replaced by RA. 4.3.4. Correlation between the proposed interface parameter and the macro properties of concrete Because the di;t data involved in n were obtained from the three concrete groups at the three testing ages, 3 days, 28 days and 90 days, n values correspond one to one, to the macro property data obtained from each concrete group at each of the three testing 40 35 30 25 20 15 10 0.03

ð6:3Þ

According to both Fig. 19 and Eq. (6), linear correlations fitted well between n and q at each of the three testing ages, i.e., 3 days, 28 days, and 90 days, as the R2 value approached 1. The good linear fitting results demonstrate that the proposed integrated interface parameter can successfully meet the second requirement listed in Section 4.3.1. Therefore, n is able to directly reflect the essential interface features of RC groups with different RA replacements. Two main factors have contributed to the positive correlation between n and q. First, n was proposed on a basis of the defined length index, Rl=A , in Section 4.1.4, which has been verified to increase linearly as q increased. Second, as verified in Section 4.1.1, despite the ITZ2s inside RA whose widths were not influenced by q, the widths of the newly formed ITZs, i.e., ITZ1s

100

Fig. 19. Linear fitting between n, and the percentage replacement of RA, q.

ð6:2Þ

2

80

,%

ð5:2Þ

where q is the RA replacement ratio in the concrete mixture. According to Fig. 18, n decreased over the age of the concrete. Moreover, n rapidly decreased at an early age, while at higher ages, the decrease slowed and n tended to be stable over time, which was mainly because the cement hydration at higher ages had slowed down. Fig. 18 also illustrates that the higher the RA replacement ratios in a concrete mixture, the larger the values of n. The correlations between n and the RA replacement ratios will be further discussed in Sections 4.3.3.

nt¼3 d ðqÞ ¼ 2  104 q þ 0:0623;

0.04

fc, MPa

nq¼50% ðtÞ ¼ 9  103 t þ 0:0832;

R2 ¼ 0:9907

0.05

0.04

0.05

0.06

0.07

0.08

(a) 5

v, km·s-1

nq¼0% ðtÞ ¼ 8  103 t þ 0:0864;

0.06

4.5

4

3.5 0.03

0.04

0.05

0.06

0.07

0.08

(b) 11

0.09 ( =0%)

( =50%)

( =100%)

10

0.08

D, 10-12m2·s-1

0.07 0.06 0.05 0.04

9 8 7 6 5 0.03

0.04

0.05

0.06

0.07

0.03 0

20

40

60

80

100

t, d Fig. 18. Logarithmic fitting between n and the age of concrete, t (unit of t: d).

(c) Fig. 20. Linear fitting results between (a) f c , (b) v, (c) D, and n.

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877 Table 2 Prediction models of macro properties of RC over RA replacement ratios, q, obtained on a basis of n. Models: MðnÞ ¼ K  q þ b

Macro properties

Fitting parameters K

b

f c (MPa)

3 days 28 days 90 days

11.0  102 5.49  102 5.49  102

22.07 30.47 35.69

v (kms1)

3 days 28 days 90 days

4.69  103 2.35  103 2.35  103

3.91 4.27 4.49

D (1012m2s1)

28 days 90 days

1.88  102 1.88  102

7.51 5.73

ages. The linearly fitted correlations between n and f c , v, and D, respectively, are illustrated in Fig. 20 and are expressed as follows:

f c ðnÞ ¼ 549:13n þ 56:28;

R2 ¼ 0:9141

ð7Þ

873

v ðnÞ ¼ 23:46n þ 5:37;

R2 ¼ 0:9283

ð8Þ

DðnÞ ¼ 187:76n  1:31;

R2 ¼ 0:929

ð9Þ

Fig. 20a–c illustrate that the linear correlations between the three macro properties of concrete and the proposed n are good, as all of the R2 listed in Eqs. (7)–(9) were close to 1. Hence, n can well meet the third requirements stated in Section 4.3.1. According to Fig. 20, as n increased, f c and v increased, while D decreased, indicating that the mechanical properties, density degree and anti-Cl diffusion property of concrete decreased as n increased. Therefore, n could be a key parameter for determining the inferiority of RC to NC and the macro material property differences among RC groups with different RA replacement ratios. 4.3.5. Model prediction of macro properties based on the proposed interface parameter By substituting n in Eqs. (7)–(9) with n taken from Eq. (6), the prediction models for f c , v, and D of concrete groups based on RA replacement ratios, q, can be obtained, which are shown in Table 2. Only ages of 3, 28 and 90 days were illustrated for both groups of

(a)

(b)

(c) Fig. 21. Comparisons between the predicted and the tested values of the compressive strength of concrete with RA replacement ratios of (a) 0%, (b) 50% and (c) 100%, respectively, at ages of 180, 360 and 720 days.

874

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

prediction models because the values of n were obtained from the tested data at these three ages. Because only three percentage replacements of RA were employed in this study, it is difficult to assess the accuracy of the prediction models for concrete groups with other RA replacement ratios, apart from 0%, 50% and 100%. However, still the prediction models obtained using n as an intermediate can be treated as reliable. As seen in Figs. 15–17, the developing speed of f c ; v and D of the three concrete groups was almost stable after reaching an age of 28 or 90 days. This trend is reflected by model groups having the same slope, K, at the ages of 28 and 90 days, as shown in Table 2. Additional RA replacement ratio levels should be analyzed in the future, in order to further compare the accuracy of prediction models of RC’s macro properties over RA replacement ratios. Similarly, by substituting n in Eqs. (7)–(9) with n in Eq. (5), the prediction models for f c , v, and D at age t (unit of t: d) can also be achieved as follows:

f c; q¼0% ðtÞ ¼ 3:84lnðtÞ þ 18:61

ð10:1Þ

f c; q¼50% ðtÞ ¼ 4:94lnðtÞ þ 10:59

ð10:2Þ

f c; q¼100% ðtÞ ¼ 4:39lnðtÞ þ 8:83

ð10:3Þ

v q¼0% ðtÞ ¼ 0:16lnðtÞ þ 3:76

ð11:1Þ

v q¼50% ðtÞ ¼ 0:21lnðtÞ þ 3:42

ð11:2Þ

v q¼100% ðtÞ ¼ 0:19lnðtÞ þ 3:34

ð11:3Þ

Dq¼0% ðtÞ ¼ 1:31lnðtÞ þ 11:57

ð12:1Þ

Dq¼50% ðtÞ ¼ 1:69lnðtÞ þ 14:31

ð12:2Þ

Dq¼100% ðtÞ ¼ 1:50lnðtÞ þ 14:91

ð12:3Þ

(a)

(b)

(c) Fig. 22. Comparisons between the predicted and the tested values of the average ultrasonic wave velocity through concrete specimens with RA replacement ratios of (a) 0%, (b) 50% and (c) 100%, respectively, at ages of 180, 360 and 720 days.

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

Because Eqs. (7)–(9) and Eq. (5) were all achieved based on testing data at the ages of 3, 28 and 90 days, Eqs. (10)–(12) are automatically based on those three ages as well. To verify the accuracy of the prediction models when calculating macro properties at other ages, the predicted values of f c , v and D were compared with the measured values at the corresponding ages of 180, 360 and 720 days, respectively, as shown in Figs. 21–23. In these figures, the black curves represent the prediction models for the macro properties of concrete with

875

a certain percentage replacement of RA. The predicted values of the macro properties at ages of 180, 360 and 720 days are illustrated in these curves, while the measured values are marked with red dots. Meanwhile, the fractional error between each predicted and measured value is also shown by hollow, blue columns. According to Figs. 21 and 23, the prediction models for concrete with three percentage replacements of RA, 0%, 50% and 100%, respectively, correlated well for f c and v. The maximum

(a)

(b)

(c) Fig. 23. Comparisons between the predicted and the tested values of Cl diffusion coefficients of concrete with RA replacement ratios of (a) 0%, (b) 50% and (c) 100%, respectively, at ages of 180, 360 and 720 days.

876

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

fractional errors of the predicted versus measured f c and v values were 6.91% and 6.01%, respectively, at ages of 180, 360 and 720 days. These values are considered accurate and acceptable. By contrast, the predicted models for D exhibited errors, which fluctuated from 4.48% to 27.84%. However, most fractional errors were below 15%. Therefore, the D prediction model is relatively acceptable as well, at least for the concrete groups analyzed in this study. One possible reason for the poorer D predictability may be that the interfaces are not the only factors influencing Cl diffusion in concrete. Other defects in concrete, such as micro cracks and pores in interfaces, and in the old and the new mortar, also influence the diffusion paths for Cl. However, their effects on D are not reflected by the proposed n in this study.

4.3.6. Discussion on rationality of the proposed interface parameter Based on the discussions in previous sections, the rationality of the proposed interface parameter of RC, n, can be assessed here. First, n got the geometric information of all the diverse types of interfaces in RC involved, meanwhile reflected the adverse effects of the adhering old cement mortar in RA on the overall mechanical properties of coarse aggregates. Therefore, it can meet the first requirement proposed in Section 4.3.1, making itself as a comprehensive integrated interface parameter of RC. Second, n is proven to have good correlations with both the RA replacement ratios, q, and the three macro properties of RC materials, i.e., f c , v and D. Hence, n well met the second and the third requirements listed in Section 4.3.1. In this way, n can not only reflect the essential differences between RC and NC, or between RC groups with different q values, in the interface features on a micro scale, but can also determine the material properties of concrete on a macro scale. Furthermore, the RC macro property prediction models over q and t, respectively, which are all obtained by using n as an intermediate, are proven to be reliable, at least for the three concrete groups in this study. As a summary, the proposed interface parameter of RC, n, can be treated as reasonable and reliable, at least for the three concrete groups employed in this study. Given that the rationality of n has been verified, the inferiority mechanism of RC to NC could be quantitatively revealed. n increases with RA replacement growth, indicating that the total areas of all the diverse types of ITZs induced by a unit area of virgin coarse aggregate, becomes larger in RC than in NC. The added areas of the weak ITZs in RC than in NC can provide extra destructive paths or penetrating paths for RC specimens against loads or harmful substance penetration from external environment, respectively. This is why the macro properties of RC specimens, which have a larger n value, are inferior to those of NC specimens. The linear correlations between the target macro property indexes in this study, i.e., f c , v and D, and n, well revealed such inferiority mechanism, in a quantitative manner. Still, according to the discussion above, the optimization of n should be explored in future studies. The prediction model of the durability index of concrete, D, which is based on concrete age, has not yet been well satisfied, as discussed in Section 4.3.5. Hence, more comprehensive interface parameters, or micro parameters, should be considered in the future, taking into account not only the effects of diverse types of ITZs but also those of other concrete defects, such as micro cracks and harmful pores contained in interfaces and the mortars, on macro properties, especially the durability of concrete materials. In addition, the effects of different mixture proportions or aggregate gradations on the stability of the proposed interface parameter also requires further exploration, in order to generalize the models obtained in this study.

5. Conclusions Several conclusions can be drawn based on the experimental study and interface parameter analysis of the three concrete groups employed in this study, which are summarized as follows: 1. The micromechanical properties of the new mortar, the newly formed ITZ1s and ITZ3s decrease as RA replacement increases at an early age of 3 days; the increased effective water to binder ratio caused by adding more water to RC than to NC through mixing may be the predominant reason. However, such differences are no longer significant after 28 days. Therefore, the effects on the long-term mechanical properties of RC materials can be ignored for adding more mixing water to RC on the basis of its natural water absorption. The mechanical properties of ITZ2s are proven to be stable over time or over RA replacement ratios in concrete. 2. The widths of the newly formed ITZ1s and ITZ3s increase as more NA is replaced with RA. Different from the micromechanical properties of these two types of ITZs, such differences among concrete groups with different RA replacements remain significant after 28 days. The widths of ITZ2s were stable. A geometric index, Rl=A , is defined for ITZs contained in concrete, which represents the total length of all types of ITZs induced by introducing a unit area of virgin aggregate. Rl=A is timestable but increases as the RA replacements increase. 3. Macro properties of concrete materials, i.e., the compressive strength, the density degree, and the anti-Cl diffusion property, are found to increase over age, while they decrease as the replacement percentages of RA increase. 4. An integrated interface parameter, n, is proposed, to comprehensively reflect the properties of the diverse types of ITZs contained in RC. n is proven to be reasonable, as it is found to be positively correlated with both the RA replacements and the macro concrete properties. The prediction models of RC’s macro properties over both RA replacement ratios and the age, obtained by using n as an intermediate, is verified as reliable, thereby to further prove the rationality of n. The inferiority mechanism of RC to NC on both the micro and macro scales is quantitatively revealed, with the help of n. 5. Further study is required to optimize the proposed interface parameters by considering the effects of other concrete defects, such as micro cracks and harmful pores, on concrete properties, especially on concrete durability. Meanwhile, the diversity in RA’s properties induced by the uncertainty of parent concrete, i.e., in the strength of the old mortar and thereby the affected properties of old ITZs, as well as in the amount of the old mortar, should also be considered when optimizing the interface parameter, so that it can be more representative, and more reliable. Reliability of the proposed interface parameter on different mixture proportions of concrete should also be explored in the following work.

Acknowledgments Financial support from the National Specialized Research Fund for Doctoral Programs of China (20120101110025), the National Natural Science Foundation of China (5157080183) and the Program for New Century Excellent Talents in University of China (NCET-12-0493) is gratefully acknowledged. The authors also want to thank Prof. Surendra P Shah from Northwestern University for his support and guidance during this study. Nanoindentation tests in this work utilized the NIFTI facility (NUANCE CenterNorthwestern University).

H. Zhang, Y. Zhao / Construction and Building Materials 101 (2015) 861–877

According to Fig. A.1, the square errors of linear fitting between the three macro properties of RC, i.e., f c , v, and D, and n0 were significantly smaller than 1, indicating that linear fitting was not appropriate for those relationships. Moreover, other fits, such as logarithmic or exponential, are unlikely to provide an accurate relationship either, because of the relatively large discreteness of the three macro property indexes over n0 .

40 fc( ')= -2717.1 ' + 62.72, R² = 0.7316

fc, MPa

30

20

References

10 0.01

0.012

0.014

0.016

0.018

'

(a) 5

v, km·s-1

v( ')= -115.44 ' + 5.62, R² = 0.7909 4.5

4

3.5 0.01

0.012

0.014

0.016

0.018

'

(b) 10.5

D, 10-12m2·s-1

9.5 8.5 7.5 6.5 D( ')= 629.45 ' -1.31, R² = 0.6527 5.5 0.01

0.012

0.014

0.016

0.018

'

(c) Fig. A.1. Linear fitting results between (a) fc, (b) v, (c) D, and n’.

Appendix A. Another integrated interface parameter, n0 , which has contained the mechanical properties of both the old cement mortar and the different types of interfaces in RC, can be expressed below:

n0 ¼

877

P3

1 li di;t ei;t ðENA ANA þ ERAVG ARAVG þ ERAOM ARAOM Þ

ðA:1Þ

where ei;t (unit: GPa) is the elastic modulus of the Type i ITZ at an age of t; ENA , ERAVG and ERAOM are the elastic modulus values of the new virgin aggregate, the old virgin aggregate embedded in RA and the old mortar adhering to the old virgin aggregate, respectively. The values of ei;t , ENA , ERAVG and ERAOM refer to Sections 4.1.2 and 4.1.3, respectively. The linear fitting results between n0 and the three macro properties of RC, i.e., f c , v, and D, are shown in Fig. A.1.

[1] S. Ismail, M. Ramli, Mechanical strength and drying shrinkage properties of concrete containing treated coarse recycled concrete aggregates, Constr. Build. Mater. 68 (2014) 726–739. [2] F. Buyle-Bodin, R. Hadjieva-Zaharieva, Influence of industrially produced recycled aggregates on flow properties of concrete, Mater. Struct. 35 (2002) 504–509. [3] N.F. Günçan, Using waste concrete as aggregate, Cem. Concr. Res. 25 (1995) 1385–1390. [4] I.B. Topcu, Physical and mechanical properties of concretes produced with waste concrete, Cem. Concr. Res. 27 (1997) 1817–1823. [5] A.K. Padmini, K. Ramamurthy, M.S. Mathews, Influence of parent concrete on the properties of recycled aggregate concrete, Constr. Build. Mater. 23 (2009) 829–836. [6] S.W. Tabsh, A.S. Abdelfatah, Influence of recycled concrete aggregates on strength properties of concrete, Constr. Build. Mater. 23 (2009) 1163–1167. [7] I.B. Topcu, S. Sß engel, Properties of concretes produced with waste concrete aggregate, Cem. Concr. Res. 34 (2004) 1307–1312. [8] J. Xiao, H. Falkner, Bond behaviour between recycled aggregate concrete and steel rebars, Constr. Build. Mater. 21 (2007) 395–401. [9] S.M. Levy, P. Helene, Durability of recycled aggregates concrete: a safe way to sustainable development, Cem. Concr. Res. 34 (2004) 1975–1980. [10] M. Breccolotti, A.L. Materazzi, Structural reliability of bonding between steel rebars and recycled aggregate concrete, Constr. Build. Mater. 47 (2013) 927– 934. [11] A. Rao, K.N. Jha, S. Misra, Use of aggregates from recycled construction and demolition waste in concrete, Resour. Conserv. Recycl. 50 (2007) 71–81. [12] M. Henry, G. Pardo, T. Nishimura, Y. Kato, Balancing durability and environmental impact in concrete combining low-grade recycled aggregates and mineral admixtures, Resour. Conserv. Recycl. 55 (2011) 1060–1069. [13] F.T. Olorunsogo, N. Padayachee, Performance of recycled aggregate concrete monitored by durability indexes, Cem. Concr. Res. 32 (2002) 179–185. [14] Y.G. Zhu, S.C. Kou, C.S. Poon, J.G. Dai, Q.Y. Li, Influence of silane-based water repellent on the durability properties of recycled aggregate concrete, Cem. Concr. Compos. 35 (2013) 32–38. [15] J.P. Ollivier, J.C. Maso, B. Bourdette, Interfacial transition zone in concrete, Adv. Cem. Based Mater. 2 (1995) 30–38. [16] G. Prokopski, J. Halbiniak, Interfacial transition zone in cementitious materials, Cem. Concr. Res. 30 (2000) 579–583. [17] C.S. Poon, Z.H. Shui, L. Lam, Effect of microstructure of ITZ on compressive strength of concrete prepared with recycled aggregates, Constr. Build. Mater. 18 (2004) 461–468. [18] J.S. Ryu, Improvement on strength and impermeability of recycled concrete made from crushed concrete coarse aggregate, J. Mater. Sci. Lett. 21 (2002) 1565–1567. [19] J. Xiao, W. Li, D.J. Corr, S.P. Shah, Effects of interfacial transition zones on the stress–strain behavior of modeled recycled aggregate concrete, Cem. Concr. Res. 52 (2013) 82–99. [20] M. Miller, C. Bobko, M. Vandamme, F.J. Ulm, Surface roughness criteria for cement paste nanoindentation, Cem. Concr. Res. 38 (2008) 467–476. [21] W. Li, J. Xiao, Z. Sun, S. Kawashima, S.P. Shah, Interfacial transition zones in recycled aggregate concrete with different mixing approaches, Constr. Build. Mater. 35 (2012) 1045–1055. [22] X. Li, Recycling and reuse of waste concrete in China: Part I. Material behaviour of recycled aggregate concrete, Resour. Conserv. Recycl. 53 (2008) 36–44. [23] A. Katz, Treatments for the improvement of recycled aggregate, J. Mater. Civ. Eng. 16 (2004) 597–603. [24] L. Sorelli, G. Constantinides, F.J. Ulm, F. Toutlemonde, The nano-mechanical signature of ultra high performance concrete by statistical nanoindentation techniques, Cem. Concr. Res. 38 (2008) 1447–1456. [25] G. Constantinides, F.J. Ulm, The effect of two types of CSH on the elasticity of cement-based materials: results from nanoindentation and micromechanical modeling, Cem. Concr. Res. 34 (2004) 67–80. [26] J. Neˇmecˇek, V. Šmilauer, L. Kopecky´, Nanoindentation characteristics of alkaliactivated aluminosilicate materials, Cem. Concr. Compos. 33 (2011) 163–170.