The 0 - 4 mm ADR Fines (AF) were produced in a case study of the C2CA project that involved the demolition of a part of a governmental complex constructed from concrete based on CEMIII cement. Recycling of the EOL concrete from the complex was performed applying C2CA technologies and the resulting AF was stored for further studies. The moisture content of AF was 12 wt%.

The research set-up consists of a Heating-Air classification System (HAS) followed by grinding of materials using a Ball Mill (BM). HAS is made of an inclined pipe with a length of 600 mm (see the schematic of HAS in Figure 3). The internal and external diameters of the pipe are 11 mm and 20 mm, respectively, and it is covered by a steel lid to avoid the dispersion of heat. For heating, three burners with a flame temperature of 1410˚C were used. The temperature inside the pipe was measured by three thermocouples connected to a computer. To separate Finer Fraction (FF) which is mostly 0 - 0.25 mm from the Coarser Fraction (CF) which is mostly 0.25 - 4 mm, air classification is applied from the downer opening of the pipe. Thus, FF and light contaminants can be collected in a bin connected to the upper part of HAS. CF is led to a container connected to the down part of the pipe, and the whole system is installed on a vibrator to facilitate the materials movement inside the pipe. After heating the materials, the produced CF is milled using a ball mill. The ball mill used in this study has a mill chamber with a diameter of 600 mm and length of 300 mm. The rotation speed was fixed at 45 rpm. Steel balls were put into the mill chamber to enhance the liberation of sand from cement powder. The balls to materials weight ratio was 2.5 and the diameter of the balls varied from 10 to 22 mm (see Figure 4).

Experimental runs were carried out with varying temperature in HAS and residence time in BM. A variation of ball milling duration (time) and temperature in HAS resulted in a total of 20 different experimental conditions. Figure 5 gives a general overview of the recycling process. In addition, applied temperatures and ball milling durations are indicated in this figure. After each experimental run, particle size distribution of the produced samples was measured applying the sieving method. The sieving process was performed based on the standard EN 933-1. After sieving, the density of each fraction was determined. For measuring the density and avoiding the effects of inaccessible pores on the results, samples were firstly milled to powder smaller than 0.125 mm. Afterwards, the densities were measured using a He pycnometer (model Accupyc 1330 Micromeritics). X-ray fluorescence (XRF) analysis was carried out on the same powders (size smaller than 0.125 mm) to determine the chemical composition of the samples. For this analysis XRF (model PANalytical-Epsilon 3x spectrometers) was applied.

For each experimental run, firstly the desired temperature along the heating pipe in HAS was attained. As it was shown in Figure 3 three thermocouples connected to the burners register the temperature at different points of the pipe. To conduct the experiments, the intensity of the middle burner was varied to reach the target temperature of the experimental design. Residence time of the materials along the HAS pipe was 30 seconds. After processing of the materials at different temperatures, the performance of the HAS was evaluated. Results presented very small deviations in particle size distributions of FF and CF coming from different heating temperature. The particle size distributions of the AF and two outputs (FF and CF) of HAS at 750˚C are shown in Figure 6 as an example. With the multiplication of the cumulative passing percentage of FF and CF to their total mass percentage, following by the summation of the results, it is possible to reconstruct the cumulative passing distribution curve for input materials (shown as the control curve in Figure 7). A comparison between the control and HAS-input curves in Figure 7, shows that a limited amount of liberation (production of new particles) happened during the HAS processing. The recovery of each size fraction in FF and CF and the cut size point is shown in Figure 8. According to the figures, it is clear that while HAS heats the materials, it efficiently uses the air flow to separate FF and light contaminants from CF (see Figure 9). When concrete waste is heated, dehydration of cement accrues which could weaken the bonding strength between cement paste and aggregate. Thus, the possibility of preferential breakage along the boundaries between aggregate and cement paste increases [17] . Although using HAS, the bonding between hydrated cement and sand in CF is weakened; there is still a need to break this bond by means of an extra force which is provided by ball milling in the current study.

To evaluate the effect of heat treatment and ball milling, experimental results were analyzed considering one variable (BM time or HAS temperature) at a time. Figure 10 shows the cumulative size distribution of materials

treated at different temperature by HAS and milled for 300 seconds. It is shown that the products size distributions moved towards the finer size range as the temperature increased. The same trend happens with increasing milling time at fixed heating temperature (for example at 750˚C) (see Figure 11). As mentioned previously, one of the main aims of this study is to investigate the effect of heating and grinding on the production of 0 - 0.250 mm fines which is mostly hardened cement rich. 2D contour and surface plot of weight percentage with respect of milling time and heating temperature in Figure 12 shows the positive influence of both parameters on increasing the mass of this fraction. Using the contour plot one can optimize the heating temperature and milling time for producing a specific amount of 0 - 0.250 mm fines at minimum cost. Ball Milling is usually a costly and an inefficient process in terms of energy requirements. Table 1 presents the costs of ball milling for both laboratory and industrial scale operations. The expected amount of CF for milling in an industrial scale is around 30 tons per hour (if ADR is run at 100 tons per hour). Changing the residence time from 300 to 90 seconds, while the throughput is constant at 30 tons per hour, demands a ball mill with a shorter length. Thus, the total cost of milling drops significantly (see Table 1).

The experimental results in Figure 13 confirm that for production of a constant amount of fines (0 - 0.25 mm), the heating temperature can be increased to achieve reduced milling time accordingly. Decreasing the milling time is advantageous because less unwanted fine silica will be produced. This silica presents a health risk and it limits the hardened cement concentration. In addition, according to Table 1, with reducing the residence time, the cost of milling reduces significantly.

In addition to the cost, the quality of the products is another important factor to ensure an effective recycling process. For the quality aspects, we considered the removal of contaminants, grading of the products, density and water absorption of the produced sand and finally the amount of hardened cement concentrated in FF. Many organic substances such as wood are unstable in concrete when submitted to drying and wetting or freezing and thawing [18] . In order to enhance the quality of the recycled sand for high grade applications, wood and plastic must be removed. Experimental results based on Loss On Ignition (LOI) test show that this removal is complete

^{*Laboratory, **Industry.}

at 500˚C. The LOI test consists of heating of some grams of each fraction at 950˚C and allowing organic substances to escape until its mass stops changing. Figure 14 shows that the difference in LOI between materials recycled at 500˚C and 750˚C is small. It is therefore concluded that for the removal of organic materials, heating for 30 seconds at 500˚C is sufficient. Amount of LOI has an inverse relation with the particle size. In general, the amount of LOI increases with decreasing the particle size. It indicates higher concentration of materials like wood in finer fractions.

Another important quality aspect is the grading of the produced sand. Recycled sand should be well-graded and comply with the standards applied in the market. The grading limit for 0 - 4 mm sand existing in the Dutch market is illustrated in Figure 15 (Max and Min black lines). According to this figure, it is clear that the particle size distribution of the recycled sand heated to 500˚C and milled for 90 seconds, is well within the limiting curves. Considering the current prices in the Dutch market, it is estimated that the selling price for the cementitious part (0 - 0.250 mm) and sandy part (0.250 - 4 mm) would be 6 ?ton and 5 to 7.5 ?ton respectively. The original AF fractions, with limited applicability in the production of new building materials, have the value of ca 1 ?ton.

Density and water absorption of aggregates are usually considered as being inter-related, showing an almost linear inverse relationship [2] . It is well-known that recycled aggregates or sand have lower density and consequently higher water absorption than the natural ones due to the attached hardened cement. Figure 16 shows the density of each particle size resulting from 500˚C and 300 seconds milling, measured using the He pycnometer. As it can be seen particles smaller than 0.125 mm have a density of 2.24 gr/cm³, which is the lowest density among all the particle sizes. The densities have then a slightly increasing trend with particle size to 2.62 gr/cm³ for 2 - 4 mm particles. Previously Florea et al. [20] reported a lowest density of 2.45 gr/cm³ for 0 - 0.63 mm fraction and a density of 2.61 for 2 - 4 mm recycled sand. Using a linear correlation equation between the density and water absorption of recycled aggregate expressed by de Juan et al. [21] the amount of water absorption for the same fractions are estimated and shown in Figure 17. Maximum water absorption for fine recycled aggregates is recommended to be in the range of 3% - 13% in the most and least restrictive standards [2] . Water absorption shown in Figure 17 satisfies well the most restrictive standard of JIS A 5021. In the market, recycled sand with water absorption less than 5% is usually ranked in a good level of quality.

In general, crushed concrete fines contain mostly hardened cement and α-quartz (sand) originated from broken aggregates [20] . XRF analysis was carried out on various size fractions after the recycling procedures. The content of SiO₂ measured by XRF ((SiO₂%) XRF) in fines is the sum of α-quartz (sand) and SiO₂ from the cement hydration products. The content of SiO₂ from hydration products can be traced back from the chemical compo-

sition of cement ((SiO₂%) CEM). Using DSC analysis, it is possible to calculate the amount of α-quartz in the recycled concrete aggregate fractions [19] [20] .

The relationship between the parameters is expressed below [19] :

In the referred research it is reported that the error between XRF results and TG-DSC is relatively low (under 3%). According to the evidence received from the demolished buildings in C2CA project, the type of cement in the utilized old concrete was CEMIII/B. Considering a typical chemical analysis of CEMIII/B received from ENCI, Netherlands, not taking into account possible slight changes in the composition of cement during the time, the amount of (SiO₂%) CEM was considered as 30%. Using the amount of (SiO₂%) XRF from the XRF analysis, it is possible to estimate the amount of hydrated cement and sand in various fractions.

Figure 18 (graphs (A)-(D)) shows the amount of hardened cement and sand in all considered fractions, coming from various heating and milling processes. According to the figure, there are some differences observed between analysed materials. Firstly it is illustrated that the bigger particle sizes are cleaner than the small ones in terms of hardened cement content. Comparing the coarser fractions in all graphs in Figure 18, it is clear that the fractions 1 - 2 mm and 2 - 4 mm processed at 105˚C and no milling (see graph (A)), contain the highest amount of attached hardened cement. The 2 - 4 mm fraction in graph A contains 44% of hardened cement, while this number drops to 20% for the same particle size in graph D (processed at 750˚C and 300 second milling). Considering the size 0 - 0.125 mm in all graphs in Figure 18, the effect of the recycling process on concentration of hardened cement in this fraction is obvious. To understand the effect of the milling (independent from heating) on the hardened cement content of each fraction, graph (A) and graph (B) in Figure 18 can be compared, and it is concluded that milling has an influence on the cleaning of the coarser fraction from hydrated cement and concentrating the cementitious powder into the finer fractions. By introducing the heat during the recycling, much cleaner sand (0.250 - 4 mm) is produced. Increasing the heating temperature to 750˚C may cause the sandy part to get more brittle and results in an increment of the sand percentage in very fine fractions (graph (D)).

Figure 19 presents the cumulative distribution of hardened cement and sand based on the particle size distribution of the final output materials from the recycling process and hardened cement-sand content of each size fraction. According to Figure 19-graph (A) (processed at 105˚C and no milling) just 6% recovery of the total hydrated cement (existing in 0 - 4 mm fraction) can be achieved for particles under 0.250 mm. By adding the milling step (see graph (B)-Figure 19) the recovery rises to 19% for the particles smaller than 0.250 mm size. With heating at 500˚C and 750˚C, the amount of cement recovery in particles under 0.250 mm raises to 34% and 39%, respectively. Considering Figure 19-graph (D) (750˚C heating and 300 sec milling), the cumulative recovery of hardened cement for fractions below 1 mm is 82%, while this number drops to 32% for dry materials processed at 105˚C and no milling (graph (A)). When comparing different graphs of Figure 19, in terms of the efficiency of recovering of hardened cement, it can be noticed that there is a slight difference between the percentage of the produced 0 - 0.250 mm at 500˚C and 750˚C. Therefore, it seems more advantageous to keep the recycling temperature at 500˚C.

Table 2 shows a comparison among the chemical composition of ADR fines and some raw meal materials such as clay and limestone (in two different quality levels). In the same table, the chemical composition of the 0 - 0.125 mm produced within this study is added as well. A comparison between the numbers shows that the amounts of CaO, Al₂O₃ and Fe₂O₃ in the produced 0 - 0.125 mm fraction are well comparable with the amounts in low-quality limestone (around 35 wt% CaO). The required heat energy for drying of one ton ADR fines is estimated to be 300 MJ. Based on the Ecoinvent database, 0.0716 kg CO₂ will be produced for the production of 1 MJ heat (if the fuel is gas) [22] . Therefore, the amount of CO₂ emission resulted from heating of ADR fines for the production of 1 ton of hardened cement is estimated to be 108 kg. On the other hand, the clinker emission factor is the product of the fraction of lime in the clinker multiplied by the ratio of the mass of CO₂ released per unit of lime [23] . Thus, for the production of one ton of clinker, almost 344 kg of chemical CO₂ will be released. A comparison between the aforesaid numbers shows that by reusing the cement paste from old concrete the release of the chemically bound CO₂ will be reduced by a factor of three.