Electric recycling of Portland cement at scale

Concrete is the most used material on the planet, after water, and is responsible for approximately 7.5% of total anthropogenic CO₂ emissions^1,2,3.

The two main strategies deployed to date to reduce emissions from Portland cement production are fuel switching (using fossil gas, municipal waste or biofuels instead of coke and petcoke) and increased use of substitution materials, such as fly ash and slag. However, these strategies cannot lead to zero emissions. Moreover, the two most common substitution materials, ground-granulated blast furnace slag and fly ash are themselves by-products of primary steel production and coal power stations, both of which are highly emitting and therefore must be phased out in the transition to zero emissions. There has been a recent surge of interest in ternary blends with calcined clay and limestone (commonly known as LC³)⁴. Nevertheless, it could be difficult to achieve substitution levels greater than 50% as these blends still require Portland clinker for activation⁵. Meanwhile, alternative clinkers or binder systems to replace Portland clinker are under development, but none so far can be made at scale without significant emissions^6,7,8. Carbon capture and storage could potentially be applied to cement making^11,12 but current projects attempt the capture of the process emissions only, without storage¹³.

Most emissions associated with concrete arise in the production of Portland cement, about 40% from fuel combustion and about 60% from limestone decarbonation. Both sources are hard to abate. High-temperature processes are not easily electrified, no alternative cement chemistry can be produced at a comparable scale and no natural source of un-carbonated calcium is available in the required quantities.

A widely deployed high-temperature electric process is the electric-arc furnace (EAF) used for recycling steel. This recycling occurs in two stages. First, scrap is melted and oxidized to remove dirt, carbon and phosphorus. Then, in a ladle furnace, sulfur is removed, and the steel is alloyed. In both steps, a flux is introduced to protect the steel from air, provide the required basicity, protect the lining and the graphite electrodes and increase energy efficiency. This flux is made from limestone–dolomite in a decarbonation process with the same process emissions as Portland cement¹⁴.

A widely available high-volume source of decarbonated calcium is found in the hydrated cement paste embedded in used concrete. It has been known for a long time that it is possible, in principle, to reclinker hydrated cement paste (HCP)¹⁵. However, there are several challenges to applying this process in a conventional kiln when a large fraction of hydrated cement paste is used. The presence of sulfates (added as set retardant during production) increases the belite content of the reclinkered cement at the expense of highly reactive alite¹⁶. Furthermore, sulfates are volatile and will condense in cooler parts of the kiln system, causing operational difficulties¹⁷. Nonetheless, there exist commercial offerings of cement made with partly replaced raw meal (such as Holcim GeoCycle). Separately, cement recovered from used concrete can contain a high fraction of chlorides that, if retained, would exclude the use of reclinkered cement in reinforced applications¹⁸.

Recovered cement paste (RCP) is not commercially available at scale at present. Separating paste from the aggregates in concrete has been a niche but active area of research and development^19,20, driven not by the prospect of using the cement paste but by the possibility of recovering high-quality aggregates that have much higher commercial value than those produced from crushing concrete demolition waste²¹. The value of the improved recovered aggregates is not at present high enough to cover the extra cost of processing, so RCP is currently landfilled. However, the know-how and the technologies required to produce RCP at scale exist²². Recently, interest in RCP has grown because of interest in using it in a carbon mineralization process^23,24, and start-up companies have been launched to sell specialist grinding equipment²⁵. Thus, a nascent market for RCP now exists.

Here we report an innovation that arises from the combination of the above observations: electric cement recycling at scale. In the process shown in Fig. 3e, and described in the patent application²⁶ for Cambridge Electric Cement (CEC), the emitting lime flux used in steel recycling is replaced by recovered cement paste that has already been decarbonated so will release no further process emissions, although it may require small adjustment with lime. The paste is reclinkered as it is fluxed into a slag. The higher temperature of the EAF (compared with cement kilns) both maintains as gas the sulfates and chlorides that held back earlier reclinkering trials and favours the production of alite over belite. The slag is cooled and ground to produce a conventional Portland clinker in an all-electric process, which, with a decarbonized grid, has neither process nor combustion emissions. This approach will in parallel reduce the emissions of steel recycling by reducing the need for flux production. Both steel recycling slags can be made cementitious by this approach, with the ladle slags being closer to cement composition. However, in this paper, we focus on the EAF (oxidizing) slags as their volume is larger.

In the context of steel recycling, fluxes are the minerals added to the steel and slag is the resulting viscous layer floating on top of the molten steel. In this paper, we aim to demonstrate that with the right composition of flux based on hydrated or recovered cement paste, the slag, when cooled rapidly, becomes clinker, the artificial mineral that, after grinding and blending, can be made into cement.

To evaluate the proposed new process route, 28 slags were produced from flux derived from cement paste both prepared for this purpose and recovered from demolition waste. Lime, alumina and silica were added to the fluxes. The composition of a selection of these slags is given in Table 1, with the full set specified in the Supplementary Information. The slags were processed in induction furnaces over clean steel with various crucibles and oxidizing conditions. In an industrial EAF, oxidation and reduction occur in a controlled sequence: these effects are tested separately here. The slags were air-cooled: at this scale, cooling rates of 10–20 K s⁻¹ at least are achieved, which is fast enough to stabilize alite. The slags were then ground and characterized. Some of the slags with compositions matching conventional clinker listed in Table 2 were blended and used to make mortar bars.

The diffractograms of selected slags are shown in Fig. 1. The diffractograms of all slags close to from the alite-forming zone (Fig. 1a, yellow) have the main peaks of cementitious phases typical of Portland cement: alite (C₃S), belite (C₂S) as well as C₄AF, and tricalcium aluminate (C₃A). In general, the phases present are those predicted in the thermodynamic CaO–SiO₂–Al₂O₃–Fe₂O₃ system (Fig. 1a). Rietvield refinement indicates that alite and belite together account for more than 70% of the products in most of the slags made in the zone in which commercial cement kilns operate (indicated by the narrow yellow region in Fig. 1a in which Al₂O₃ is below 6%). For reference, Portland cement should normatively contain 66.7% by mass of alite and belite²⁷. Using this criterion, our process route can make Portland clinkers over a fairly wide zone. In the belite-forming zone (Fig. 1a, orange region), ghelenite is also sometimes observed. Ghelenite (C₂AS) is the dominant species in the ghelenite zone (Fig. 1a, pink region). The effective lime-to-silica ratio (Methods) is an excellent predictor of the silicate phases that will form (Fig. 1b).

a, Ternary diagram pair presenting the oxide composition of the slags studied in the SiO₂–CaO–Al₂O₃ and SiO₂–CaO–Fe₂O₃ systems measured using XRF. Every oxide composition was analysed by X-ray diffraction, and the resulting crystallographic composition is shown as a pie chart. Right, a detail of the SiO₂-CaO-Al₂O₃ ternary diagram on the left. b, Percentage of gehlenite in the slag and the fraction of alite over total alite and belite in the tested systems both as a function of (C/S)*, the available lime-to-silica ratio for the formation of silicate phases. The method for calculating (C/S)* is given in the Methods. The grey shaded region represents the range of C/S for which both Alite and Belite can form. c, Diffractograms and phase compositions of selected slags produced in this work. γ, C₂S-γ; β, C₂S-β; m, C₃S-monoclinic; g, ghelenite; a, C₃A cubic; c, graphite; and q, quartz. d, Comparison between the tested slags and compositions reported in the literature. The literature used to create this figure is given in Supplementary Table 3. Med., medium; Com., commercial; a.u., arbitrary units.

The X-ray fluorescence (XRF) analysis of selected slags is provided in Table 1 (for full results, see Supplementary Table 1). The oxide composition of the resulting slags does not differ markedly when using carbon or magnesium crucibles (Supplementary Fig. 1) indicating that over clean steel, there is no significant exchange of species between the slag and the melt. Early experiments performed with an aluminium oxide lining exhibited significant aluminium leaching (Supplementary Fig. 1) and were therefore abandoned. The contamination inherent to steel scrap used in commercial operations would introduce silica and alumina to the slag: the same final compositions would then be reached by adding more lime (CaO) to the flux.

RCP contains a higher proportion of SiO₂ than HCP, almost certainly derived from aggregates still attached to the paste (Fig. 1a). Assuming that feedstock cement has the composition of the CEM I we used as a control, the RCP would contain 52% HCP and 48% impurities. Once corrected by adding lime to have the right balance of calcium and silica (Table 3), it reclinkers as well as pure hydrated paste.

The metallurgical function of EAF oxidizing slags is to allow the dephosphorization of steel while limiting sulfur in the melt. The slag compositions we tested overlap with those reported in the literature (Fig. 1d) and have the appropriate chemistry. To verify their suitability for EAF operations, slags of typical composition were used, and the resulting steel composition was tested but using a considerably larger flux-to-steel ratio than would be practical in present-day furnace designs (roughly 9:30 compared with 1:30). Despite these unfavourable conditions, only very small amounts of sulfur transferred into the steel (Table 4). In steel-making, following further desulfurization, so far most alloys have a sulfur concentration of less than 0.05% (ref. ²⁸), which can be achieved easily from the values reported in Table 4. This confirms that the basic slags produced from the proposed process can be used in EAFs to de-phosphorize steel.

The cement setting time was measured using calorimetry at 240–280 min (Fig. 2a, ±20 min after²⁹, the middle point between the dormant period trough and the peak of the silicate hydration peak). Mortars made with the cements were fairly easy to place with small additions of superplasticizer required for the slags made in graphite crucibles (GC80) and for LC³ mixes. No early setting or flash setting was observed. No bleeding was observed (Fig. 2d, second from the left). On demoulding, the samples were inspected and no defects were found (Fig. 2d, third from the left). All cements, both pure and those blended with calcined clay and limestone, exhibited similar strength development to those made with commercial clinker despite being undersulfated (Supplementary Fig. 2), and containing a fraction of contaminants probably introduced when the crucible was scraped during tapping (Fig. 1c). The relationship between strength development and heat release (Fig. 2a,c) for both our new clinker and a conventional commercial clinker was similar. Higher belite content is associated with lower early strength, whereas higher alite relates to higher early age strength (Fig. 2a and Supplementary Table 2). The improved performance of commercial cement (compared with that made from commercial clinker) is because of finer grinding³⁰ (Fig. 2b). With commercial grinding equipment the clinker produced with the new process is likely to have the same grindability as commercial clinkers (Fig. 2b).

a, Instantaneous and cumulative heat released by high- and medium-alite cement and high-C₃A cement produced with the new process, commercial clinker ground in the same conditions and commercial cement produced with the commercial clinker shown for reference. b, Cumulative and frequency plots of the particle size distributions of the cements used for the strength tests. c, Strength evolution of cements produced using the new process, both as pure cements and LC³ blends. d, Slag as poured from the furnace, fresh and hardened mortar bars; sample after compression failure. Med., medium; Com., commercial; gyp., gypsum.

Together, these results demonstrate the production of Portland cement by reclinkering HCP or RCP on a bed of molten steel. The slags arising from the new flux have the right composition for their metallurgical function. The mineralogy of the output can be tuned as in any conventional kiln, and high-quality clinker can be produced.

Figure 3a confirms the emissions saving from electric cement recycling and anticipates production costs comparable with those of present-day cement that are dominated by the price of fuel and SCMs³¹. All processes are electrified, and both the extra costs and total power requirements of the new process are dominated by heating old concrete before crushing. This is currently required to separate a clean stream of RCP but will be optimized in future. With a decarbonized grid, the only emissions of co-production arise from the small remaining requirement for lime flux, and these would be allocated between the two materials by negotiation. The amount of lime flux required in EAFs depends on the steel scrap feedstock and the furnace design. Clean steel requires less flux to bind residue from the steel, but more to create enough slag to cover the electrodes. All such flux could be replaced with RCP to produce cement. By contrast, although less clean steel requires less flux to cover the electrodes, more is needed to bind with unwanted silica to ensure high slag basicity required for steel quality. In this case, adding RCP to the flux would not reduce lime use but would still produce additional mass of clinker without increasing the emissions of steel recycling. Therefore, Fig. 3a distinguishes the emissions savings from a typical reduction in lime use and those from avoiding conventional cement production. The fact that operating expenses are much greater than capital or labour costs, suggests that the market viability demonstrated here for the UK will translate to other countries.

LC³-50 is 50% clinker, 30% calcined clay, 15% limestone and 5% gypsum; CEC is the process described in this paper. a, Emissions are estimated based on global data, whereas costs are estimated for the UK. The four recipes for electric cement (CEC) use global average or future zero emissions electricity, with or without blending with calcined clay and are compared with ordinary Portland cement and LC³-50. Full details of the sources and calculation of these estimates are provided in the Supplementary Information. b, Representation of the range of concrete compositions and the outcomes of separation (W is water). c, Historical and projected cement and clinker production worldwide, implied RCP availability assuming a 50-year lifetime for buildings (left); global cement-related emissions under a range of scenarios (right); the potential supply of RCP is greater than that which could be used in EAFs but production could be increased if new dedicated EAFs are built to produce cement. d, The current material flows and industrial operations for the production of bulk construction materials. e, The material flows anticipated if the process described in this paper is deployed at scale in the UK and all arising steel scrap is recycled domestically. cem., cement; decarb, decarbonized.

The overall supply chain reconfiguration is extensive (compare Fig. 3d,e), but every change can be economically viable. Figure 3d draws on current estimates of mass flows in the UK: steel demand is around 15 Mt yr⁻¹, mainly from imports, leading to around 10 Mt yr⁻¹ of scrap, most of which is currently exported³²; cement demand is around 13 Mt yr⁻¹, requiring clinker production of around 9 Mt yr⁻¹ (ref. ³³). Total construction and demolition waste is around 68 Mt yr⁻¹ (refs. ^34,35), most of which is concrete. A conservative estimate is that 4–4.5 Mt of RCP could be produced with a 60% collection rate. The data in Fig. 3e assume that scrap steel volumes approach annual demand to reach 14 Mt yr⁻¹ and that all UK steel scrap is recycled domestically, following the expansion of EAF capacity. This defines the total capacity for electric cement production. In this study, experiments used 5–30% of the steel mass as the input flux mass. EAFs can operate now with up to 20–30% slag. Although this could be increased, higher values are at present considered undesirable as slag is a waste by-product now, so the figure assumes a slag-to-steel ratio of 1:7, or 14%.