Concrete rubble is produced in Concrete rubble is produced in huge quantities by demolition. Metals and glass are generally recycled industrially, and it would be worthwhile to do the same with concrete. A patient, four-year laboratory project has demonstrated that complete, dry recycling of concrete is possible, based on the finding that hydrated cement can be reclinkered to produce cement of the original quality. Starting with vintage concrete that was crushed and heated to 700°C for four hours, we successfully separated the hydrated cement from the sand and aggregates by attrition, obtaining 60% to 70% pure hydrated cement. Further separation can be obtained by an electrostatic or magnetic process yielding 85% purity. The resulting hydrated cement can be reclinkered with or without additives to produce high-quality Portland or CLC cement (a mixture of Portland cement and blast furnace slag cement). We obtained mortar of excellent quality using the constituents separated in this way. These results plead for ecologically sound management of concrete on a much broader scale.
Statistics show that, after potable water, concrete is the most widely consumed product. The enormous quantities involved call for ecological management, which clearly includes recycling. Of all the CO2 released into the atmosphere in 1991, 70 million tons of carbon were attributed to cement manufacturing throughout the world. Recycling cement, however, does not release CO2 since no limestone is added. Concrete is made from about 1 volume of cement for 2 volumes of sand and 3 volumes of aggregates. Recycling only the aggregates is in fact recycling only half the total concrete volume.
We found no references in the literature discussing complete recycling of concrete; research in this area therefore meant tackling a new problem. Basically, the issue was to determine whether the hydration of pure cement is a reversible process. Despite the lack of published work on the subject, the chemistry involved was promising.
We used normal Portland cement (NPC type 55) for these trials. We first prepared six 4 × 4 × 16 cm pure-paste specimens with a water/cement ratio such that a 10 mm dia. Vicat probe penetrated to a depth of 34 mm, indicating a pure paste of standard consistency. The specimens were hardened either at 60°C and 100% relative humidity, or at room temperature. We measured their mechanical properties after 6 days for the autoclaved specimens, and after 28 days for the room-temperature specimens. The specimens were next ground to the required particle size and cured for 30 minutes at 1450°C using standard cement kiln procedures. The resulting clinker was ground to standard cement particle size and the fabrication cycle was repeated to obtain pure-paste specimens. We compared the properties of the recycled specimens with those of the initial reference materials.
Elemental chemical analysis was performed by atomic absorption flame spectrometry after melting with lithium metaborate and acid dissolution. The results obtained for the cement powder before and after reclinkering show that the oxide chemical composition was only slightly modified. More significant differences were observed for the ignition loss at 1000°C and the for sulfur trioxide content; we attribute the differences to the water content of the hydrated materials (water is eliminated during heat treatment at 1000°C) and to sulfate losses during clinkering.
We recorded diffraction diagrams (using a Siemens diffractometer with a rotating sample holder and linear detector) on unhydrated cement powder samples corresponding to the reference powder and to powder obtained from a reclinkered pure hardened paste to compare their mineralogical structure. The spectra revealed similar mineralogical species in both samples. We identified the usual NPC clinker minerals, notably tricalcium silicate (3CaO·SiO2 or “C3S”), dicalcium silicate (2CaO·SiO2 or “C2S”) and tricalcium aluminate (3CaO·Al2O3 or “C3A”). Calcium sulfate dihydrate (gypsum) was not identified in the reclinkered material, however; sulfates were present in a different form—probably non-crystalline since they were undetectable in the X‑ray diffraction spectrum.
The same unhydrated cement specimens were observed and analyzed by SEM in conjunction with an X‑ray diffraction spectrometer. The unhydrated NPC reference powder was examined to determine the morphological characteristics and composition of its constituents. The powder consisted of angular and subangular fragments, smooth and grainy, with the same overall composition as clinker; many of the fragments were smaller than 10 µm. The fragmentation of the clinker grains revealed isolated phases in the powder that were consistent with the stoichiometry for C3S and C2S, and for C3A and C4AF (tetracalcium aluminoferrite). Energy-dispersive electron microprobe analysis confirmed the presence of often contiguous pseudo-hexagonal alite (C3S) crystals associated with rounded crystals often twinned with belite (C2S) and bonded together by an interstitial phase.
The morphology and composition of the reclinkered cement paste were examined in the same way. The morphological examination revealed a large number of grains, most of which were smaller than 10 µm; some of the grains appeared to be broken. The clinker composition was identical, with C2S and C3S associated with the interstitial phases C3A and C4AF. The crystals were generally smaller, possibly because of the cooling rate.
We compared the mechanical properties of specimens fabricated from reclcinkered NPC with those of specimens produced from the initial NPC. Compressive testing (Table I) show that the reclinkered cement material has hydraulic properties and that its compressive strength (128 MPa) exceeds that of the initial pure paste. In other words, the composition and properties of the material obtained from reclinkering a hardened Type 55 NPC paste were comparable to those of the initial cement, and the original hydraulic properties were restored. These properties are related to the presence of anhydrous mineral species that are specific to cement (C2S, C3S, C3A) and cause it to harden in water. Reactivating NCP paste by reclinkering at 1450°C in cement kiln conditions produces a clinker with equivalent or even improved composition and hydraulic properties when compared with the original cement.
Demonstrating that pure hydrated cement can be reclinkered to restore or even improve its properties opened the way toward complete recycling of concrete. We could then address the difficult task of separating the three major components: aggregates, sand and cement.
Fine particles of hydrated cement adhered to the sand and aggregates. Although this was only to be expected, as it constitutes the strength and cohesion of the concrete, it is also the Gordian knot of the separation process: for adequate separation to take place, it must be “undone”.
Pursuing research done in Dutch universities1, the Dutch company KEMA developed an aggregate separation process known as DECO2. We decided this would be a good starting point, and it turned out to be even better than expected. Our study shows that poor results are obtained by attempting to separate the constituents directly, i.e. without preheating the crushed concrete to 700°C as in the DECO process. Applying electrostatic and magnetic separation techniques directly to the concrete also produces mediocre results. The 700°C preheating step in KEMA’s DECO process thus appears to be a suitable—or even indispensable—means of weakening the adherence of the hydrated cement fines to the sand and aggregates.
Our raw material consisted of two concrete samples over 50 years old, one (CS) from Sellafield (UK) and the other (CM) from Marcoule (France). We found that controlled attrition of the crushed concrete down to 2.5 mm was the best way to break up the constituents; forced sieving at 0.125 mm (or a similar process) actually separates the minute particles (15 µm) of hydrated cement from the much larger sand and crushed aggregate particles. This can be done at laboratory scale by forced sieving at 0.125 mm using a flat spatula, and at industrial scale using Deval, Los Angeles or other ball milling equipment.
The forced sieving or controlled attrition step must produce a hydrated cement phase that is as pure as possible. We obtained 60% to 70% pure hydrated cement, the remainder consisting of a mixture of crushed sand and other aggregates. This level of purity requires additives—notably calcium carbonate—to obtain an 85% pure hydrated cement (crude) that can easily be reclinkered. Depending on cost requirements and on the desired objective, we suggest the use of electrostatic or magnetic separation techniques to avoid the need for additives.
Doping the hydrated cement samples HCM and HCS produced crudes that met the criteria for the coefficients and moduli for the fabrication of the type CLC CEM V/A cement we wanted to obtain. The crudes were cured in a simulation kiln, with blast furnace slag and fly ash added after curing, at the Origny Cement Works (Belgium).
The clinkers were ground in the laboratory to a Blaine specific surface area (BSS) of 4000 cm2·g‑1 for two cement samples consisting of 23% Dunkerque blast furnace slag (indispensable for producing CLC), Hornaing fly ash, 49% clinker (CM or CS) and 5% Taverny gypsum to maintain the SO3 content constant at 2.5%. The cement samples were submitted to elemental analysis, particle size analysis, mineralogical constituent determination, specific weight and BSS measurements. The cements were compared with a reference cement (Lumbres V/A 32.5 PM ES CPI – 1999).
Four mortars were prepared from the reclinkered cement and blast furnace additives: two with sand recovered from the CM and CS samples (sand 1 contained two size fractions: 1 [0.15–0.5 mm] and 2 [0.5–2.8 mm] from crushed CM concrete; sand 2 consisted of a single size fraction [0–2.5 mm] from crushed CS concrete), and two others with standard Fontainebleau sand. The actual mortar compositions, mixing water quantities, workability flow rates and bulk densities of the green mortars are summarized in Table II; the 28‑day bending and compressive strength values for the four mortar compositions are indicated in Table III; mortar samples 645, 646 and 647 are shown in Figure 1 after overcompression testing.
The mechanical strength of the two reconstituted CLC cements after 28 days exceeded 50 MPa, allowing them to be classified easily as 42.5. The average 28‑day compressive strength for the mortar prepared from CS cement and Sand 2 from Sellafield was 37.3 MPa. This relatively low figure can be attributed to the single grain size fraction in Sand 2, which prevented optimization. The mortar prepared from CM cement with optimized 1-2 sand (all from Marcoule) performed very well, with an average compressive strength of 61.1 MPa. The tests show that it is preferable to separate the sand into two size fractions and reconstitute them in order to improve the mechanical properties of the mortars.
This laboratory-scale project involving a few kilograms of material demonstrates the feasibility of completely recycling concrete using standard dry industrial procedures: heating, crushing, controlled attrition, and clinkering. These techniques are well known to concrete professionals, and can thus be applied to all types of concrete rubble from the demolition of civil industrial buildings, at a cost near that of new concrete (only a matter of scale). This in turn would make it unnecessary to open new quarries, limit carbon dioxide emissions, and reduce the volume of ground-sterilizing dumped rubble.
1. Bakiewicz, J.L. & Reymer, A.P.S. Separation of Contaminated Concrete, European Report EUR 12562 EN (1990).
2. Cornelissen, H.A.W. Development of a Process for Separating Radioactive Constituents of Concrete, Including Active Pilot-Scale Testing, European Report EUR 16917 EN (1996).
Total recycling of concrete
J.R. Costes*, C. Majcherczyk† & I. Peeze Binkhorst‡
*Commissariat à l’Énergie Atomique (CEA), Rhône Valley Research Center, BP 17171, 30207 Bagnols-sur-Cèze Cedex, France
†CEBTP, 78470 Saint Rémy les Chevreuse, France
‡KEMA, P.O. Box 9035, 6800 ET Arnhem, The Netherlands
Concrete rubble is produced in huge quantities by demolition. Metals and glass are generally recycled industrially, and it would be worthwhile to do the same with concrete. A patient, four-year laboratory project has demonstrated that complete, dry recycling of concrete is possible, based on the finding that hydrated cement can be reclinkered to produce cement of the original quality. Starting with vintage concrete that was crushed and heated to 700°C for four hours, we successfully separated the hydrated cement from the sand and aggregates by attrition, obtaining 60% to 70% pure hydrated cement. Further separation can be obtained by an electrostatic or magnetic process yielding 85% purity. The resulting hydrated cement can be reclinkered with or without additives to produce high-quality Portland or CLC cement (a mixture of Portland cement and blast furnace slag cement). We obtained mortar of excellent quality using the constituents separated in this way. These results plead for ecologically sound management of concrete on a much broader scale.
Statistics show that, after potable water, concrete is the most widely consumed product. The enormous quantities involved call for ecological management, which clearly includes recycling. Of all the CO2 released into the atmosphere in 1991, 70 million tons of carbon were attributed to cement manufacturing throughout the world. Recycling cement, however, does not release CO2 since no limestone is added. Concrete is made from about 1 volume of cement for 2 volumes of sand and 3 volumes of aggregates. Recycling only the aggregates is in fact recycling only half the total concrete volume.
We found no references in the literature discussing complete recycling of concrete; research in this area therefore meant tackling a new problem. Basically, the issue was to determine whether the hydration of pure cement is a reversible process. Despite the lack of published work on the subject, the chemistry involved was promising.
We used normal Portland cement (NPC type 55) for these trials. We first prepared six 4 × 4 × 16 cm pure-paste specimens with a water/cement ratio such that a 10 mm dia. Vicat probe penetrated to a depth of 34 mm, indicating a pure paste of standard consistency. The specimens were hardened either at 60°C and 100% relative humidity, or at room temperature. We measured their mechanical properties after 6 days for the autoclaved specimens, and after 28 days for the room-temperature specimens. The specimens were next ground to the required particle size and cured for 30 minutes at 1450°C using standard cement kiln procedures. The resulting clinker was ground to standard cement particle size and the fabrication cycle was repeated to obtain pure-paste specimens. We compared the properties of the recycled specimens with those of the initial reference materials.
Elemental chemical analysis was performed by atomic absorption flame spectrometry after melting with lithium metaborate and acid dissolution. The results obtained for the cement powder before and after reclinkering show that the oxide chemical composition was only slightly modified. More significant differences were observed for the ignition loss at 1000°C and the for sulfur trioxide content; we attribute the differences to the water content of the hydrated materials (water is eliminated during heat treatment at 1000°C) and to sulfate losses during clinkering.
We recorded diffraction diagrams (using a Siemens diffractometer with a rotating sample holder and linear detector) on unhydrated cement powder samples corresponding to the reference powder and to powder obtained from a reclinkered pure hardened paste to compare their mineralogical structure. The spectra revealed similar mineralogical species in both samples. We identified the usual NPC clinker minerals, notably tricalcium silicate (3CaO·SiO2 or “C3S”), dicalcium silicate (2CaO·SiO2 or “C2S”) and tricalcium aluminate (3CaO·Al2O3 or “C3A”). Calcium sulfate dihydrate (gypsum) was not identified in the reclinkered material, however; sulfates were present in a different form—probably non-crystalline since they were undetectable in the X‑ray diffraction spectrum.
The same unhydrated cement specimens were observed and analyzed by SEM in conjunction with an X‑ray diffraction spectrometer. The unhydrated NPC reference powder was examined to determine the morphological characteristics and composition of its constituents. The powder consisted of angular and subangular fragments, smooth and grainy, with the same overall composition as clinker; many of the fragments were smaller than 10 µm. The fragmentation of the clinker grains revealed isolated phases in the powder that were consistent with the stoichiometry for C3S and C2S, and for C3A and C4AF (tetracalcium aluminoferrite). Energy-dispersive electron microprobe analysis confirmed the presence of often contiguous pseudo-hexagonal alite (C3S) crystals associated with rounded crystals often twinned with belite (C2S) and bonded together by an interstitial phase.
The morphology and composition of the reclinkered cement paste were examined in the same way. The morphological examination revealed a large number of grains, most of which were smaller than 10 µm; some of the grains appeared to be broken. The clinker composition was identical, with C2S and C3S associated with the interstitial phases C3A and C4AF. The crystals were generally smaller, possibly because of the cooling rate.
We compared the mechanical properties of specimens fabricated from reclcinkered NPC with those of specimens produced from the initial NPC. Compressive testing (Table I) show that the reclinkered cement material has hydraulic properties and that its compressive strength (128 MPa) exceeds that of the initial pure paste. In other words, the composition and properties of the material obtained from reclinkering a hardened Type 55 NPC paste were comparable to those of the initial cement, and the original hydraulic properties were restored. These properties are related to the presence of anhydrous mineral species that are specific to cement (C2S, C3S, C3A) and cause it to harden in water. Reactivating NCP paste by reclinkering at 1450°C in cement kiln conditions produces a clinker with equivalent or even improved composition and hydraulic properties when compared with the original cement.
Demonstrating that pure hydrated cement can be reclinkered to restore or even improve its properties opened the way toward complete recycling of concrete. We could then address the difficult task of separating the three major components: aggregates, sand and cement.
Fine particles of hydrated cement adhered to the sand and aggregates. Although this was only to be expected, as it constitutes the strength and cohesion of the concrete, it is also the Gordian knot of the separation process: for adequate separation to take place, it must be “undone”.
Pursuing research done in Dutch universities1, the Dutch company KEMA developed an aggregate separation process known as DECO2. We decided this would be a good starting point, and it turned out to be even better than expected. Our study shows that poor results are obtained by attempting to separate the constituents directly, i.e. without preheating the crushed concrete to 700°C as in the DECO process. Applying electrostatic and magnetic separation techniques directly to the concrete also produces mediocre results. The 700°C preheating step in KEMA’s DECO process thus appears to be a suitable—or even indispensable—means of weakening the adherence of the hydrated cement fines to the sand and aggregates.
Our raw material consisted of two concrete samples over 50 years old, one (CS) from Sellafield (UK) and the other (CM) from Marcoule (France). We found that controlled attrition of the crushed concrete down to 2.5 mm was the best way to break up the constituents; forced sieving at 0.125 mm (or a similar process) actually separates the minute particles (15 µm) of hydrated cement from the much larger sand and crushed aggregate particles. This can be done at laboratory scale by forced sieving at 0.125 mm using a flat spatula, and at industrial scale using Deval, Los Angeles or other ball milling equipment.
The forced sieving or controlled attrition step must produce a hydrated cement phase that is as pure as possible. We obtained 60% to 70% pure hydrated cement, the remainder consisting of a mixture of crushed sand and other aggregates. This level of purity requires additives—notably calcium carbonate—to obtain an 85% pure hydrated cement (crude) that can easily be reclinkered. Depending on cost requirements and on the desired objective, we suggest the use of electrostatic or magnetic separation techniques to avoid the need for additives.
Doping the hydrated cement samples HCM and HCS produced crudes that met the criteria for the coefficients and moduli for the fabrication of the type CLC CEM V/A cement we wanted to obtain. The crudes were cured in a simulation kiln, with blast furnace slag and fly ash added after curing, at the Origny Cement Works (Belgium).
The clinkers were ground in the laboratory to a Blaine specific surface area (BSS) of 4000 cm2·g‑1 for two cement samples consisting of 23% Dunkerque blast furnace slag (indispensable for producing CLC), Hornaing fly ash, 49% clinker (CM or CS) and 5% Taverny gypsum to maintain the SO3 content constant at 2.5%. The cement samples were submitted to elemental analysis, particle size analysis, mineralogical constituent determination, specific weight and BSS measurements. The cements were compared with a reference cement (Lumbres V/A 32.5 PM ES CPI – 1999).
Four mortars were prepared from the reclinkered cement and blast furnace additives: two with sand recovered from the CM and CS samples (sand 1 contained two size fractions: 1 [0.15–0.5 mm] and 2 [0.5–2.8 mm] from crushed CM concrete; sand 2 consisted of a single size fraction [0–2.5 mm] from crushed CS concrete), and two others with standard Fontainebleau sand. The actual mortar compositions, mixing water quantities, workability flow rates and bulk densities of the green mortars are summarized in Table II; the 28‑day bending and compressive strength values for the four mortar compositions are indicated in Table III; mortar samples 645, 646 and 647 are shown in Figure 1 after overcompression testing.
The mechanical strength of the two reconstituted CLC cements after 28 days exceeded 50 MPa, allowing them to be classified easily as 42.5. The average 28‑day compressive strength for the mortar prepared from CS cement and Sand 2 from Sellafield was 37.3 MPa. This relatively low figure can be attributed to the single grain size fraction in Sand 2, which prevented optimization. The mortar prepared from CM cement with optimized 1-2 sand (all from Marcoule) performed very well, with an average compressive strength of 61.1 MPa. The tests show that it is preferable to separate the sand into two size fractions and reconstitute them in order to improve the mechanical properties of the mortars.
This laboratory-scale project involving a few kilograms of material demonstrates the feasibility of completely recycling concrete using standard dry industrial procedures: heating, crushing, controlled attrition, and clinkering. These techniques are well known to concrete professionals, and can thus be applied to all types of concrete rubble from the demolition of civil industrial buildings, at a cost near that of new concrete (only a matter of scale). This in turn would make it unnecessary to open new quarries, limit carbon dioxide emissions, and reduce the volume of ground-sterilizing dumped rubble.
1. Bakiewicz, J.L. & Reymer, A.P.S. Separation of Contaminated Concrete, European Report EUR 12562 EN (1990).
2. Cornelissen, H.A.W. Development of a Process for Separating Radioactive Constituents of Concrete, Including Active Pilot-Scale Testing, European Report EUR 16917 EN (1996).
<> Style tag for the Acknowledgements.
Correspondence should be addressed to J.R.C. (e-mai
l: mailto:omogine@yahoo.fr.
