Electric recycling of Portland cement at scale

Cement production causes 7.5% of global anthropogenic CO2 emissions, arising from limestone decarbonation and fossil-fuel combustion1–3. Current decarbonation strategies include substituting Portland clinker with supplementary materials, but these mainly arise in emitting processes, developing alternative binders but none yet promises scale, or adopting carbon capture and storage that still releases some emissions4–8. However, used cement is potentially an abundant, decarbonated feedstock. Here we show that recovered cement paste can be reclinkered if used as a partial substitute for the lime–dolomite flux used in steel recycling nowadays. The resulting slag can meet existing specifications for Portland clinker and can be blended effectively with calcined clay and limestone. The process is sensitive to the silica content of the recovered cement paste, and silica and alumina that may come from the scrap, but this can be adjusted easily. We show that the proposed process may be economically competitive, and if powered by emissions-free electricity, can lead to zero emissions cement while also reducing the emissions of steel recycling by reducing lime flux requirements. The global supply of scrap steel for recycling may treble by 2050, and it is likely that more slag can be made per unit of steel recycled. With material efficiency in construction9,10, future global cement requirements could be met by this route.


Material identification
The oxide composition of all slags and raw materials were determined as described in the materials and methods.Supplementary Table 1 gives the resulting oxide composition.A value for Chloride is reported, but it is only indicative due to the calibration of the XRF equipment.
Supplementary Table 1: XRF of all the slags tested as well as the raw materials.The calibration of the XRF does not allow for accurate measure of chloride content.
The crystalline content of all the slags was analysed, following the protocol given in the materials and methods.The results of this analysis are reported in Supplementary Table  2.For most of the materials, the amorphous content was negligible.Except for a few samples (MO2-3 to MO2-7), there is an offset of around 2.65° 2θ due to the loss of alignment of the diffractometer.Nevertheless, apart from that offset, there was no difference in peak intensities.

Supplementary Table 2:
XRD phase composition from Rietveldt analysis.
Determination of the amorphous content in the other slags/clinker is difficult as the amorphous content is lower than 20% so its presence cannot be visually identified.
Quantitative determination of the amorphous content using either the internal or external standard method requires a McCrone mill which was not available.For example, Snellings et al. (2014) showed clinker and alite having mean particle size of around 35μm could have amorphous content of around 20%, while when the same clinkers are ground down to around 3-5 μm, the amorphous content decreases to close to 0%.Therefore, we deduce that the clinkers produced in the experiments reported in this paper have an amorphous content no more than 20%, but this is likely to be a substantial overestimate based on the performance and concordance between XRF and XRD compositions.

Literature survey of EAF slag compositions
An extensive literature survey was conducted to determine the range of EAF slag compositions reported in previous work.These are recorded in Supplementary Table 3.Because of the heterogeneity of the methods used and variations in reporting (sometimes ranges are given) there is some uncertainty about the composition, so they are reported in the main paper as zones, and represented as semi-transparent discs.

Leaching
Early experiments in this study were conducted with an aluminium oxide lining in a 250 kg furnace.As reported in the main paper, this lining proved unsuitable for re-clinkering over molten steel and led to uncontrolled leaching.To illustrate the effect, Supplementary Figure 1 shows the difference in oxide composition between the input flux and the output slag after fluxing over molten steel in three different crucibles.The effect of leaching from the Alumina crucible can be seen from the large increase in Al2O3 in the samples prepared with it.Unlike the results reported in the main paper using MgO or graphite crucibles, the aluminium content of the slag is much larger than that of the flux because of leaching.
Supplementary Figure 1: Oxide composition difference between flux and slag when using three different crucibles.The figure shows the outcomes of six trials in a crucible with aluminium lining, three with a carbon crucible and three with a magnesium crucible.Each trial is represented with a pair of results, of which the upper bar shows the oxide

Sulphur in steel and slag
The major difference between cement paste and the lime flux added in EAF operations, is the inclusion of SiO2, Al2O3, Fe2O3 and SO3.SiO2, Al2O3 and iron oxides are compatible with the slag and are usually present in scrap and slag due to oxidation reactions.Further, Si and Al cannot easily be reduced in EAF operating conditions used to melt steel.Thus Si and Al remain as oxides in the slag and will not affect steel.The effect of having SO3 in flux is unknown to steelmaking, but it is volatile at the operating temperatures of EAFs.
High sulphur content has a negative impact on the mechanical properties of steel, so it is typically limited to less than 0.05%.To test the risk of this with the new process, in some of our experiments, we used a flux to steel ratio at least 7 times higher than a typical flux addition in the EAF process, to increase the potential transfer of sulphur into the system.Nevertheless, the resulting sulphur content of the steel was no more than 0.08% in all cases, so with more conventional flux:steel ratios, we do not anticipate a problem.
In detail, in trial MO2-1, around 150g of flux was added to 500g of steel.The SO3 content in the flux was 2.4%, resulting in 1.24g of sulphur in the flux.Of that, 0.31g stayed in the slag, 0.35g transferred to the steel and the remaining 0.58g was volatile.The resulting sulphur content of steel was therefore around 0.07%.
In the combined double-slagging trials MO2-6, (i.e.MO2-6a and MO2-6b), around 300g of flux was added in total, amounting to 2.4g of sulphur added, of which 0.4g went to steel, 1.1g to slag and 0.9g was volatile, leading to a sulphur content in the steel of 0.08%.
In reducing conditions with a graphite crucible, the SO3 content remained almost unchanged with flux and slag, indicating almost all the S stayed with the slag and hardly any sulphur went to steel.
Nevertheless, the effect of sulphate content in the flux on the partition of sulphur in EAF must be studied further, and the costs of the additional sulphur load must be investigated.

Performance
All the strength tests were performed according to EN 196-1 and can be found in Supplementary Table 4.

Under-sulphation in strength tests
The cements prepared for the strength tests were under-sulphated.As our production volumes are currently constrained by equipment and labour availability, figure 2A in the main paper presents updated calorimetry data (used to anticipate the setting time) for properly sulphated samples, but we have not had the material to repeat the strength tests shown in figure 2C.For completeness, we therefore present here the calorimetric curves, both instant and cumulative, for the samples used to create figure 2C.The earlier results show a large early peak due to aluminate hydration overlayed over the main (silicate) peak.
When, the cement is properly sulphated, as with the samples in figure 2A, this peak occurs after the silicate peak, and is smaller. 2 Calorimetry results for the under-sulfated samples used to create the strength tests of figure 2C in the main paper.

Boundaries and assumptions
To anticipate the cost and emissions impact of making Portland cement by electric-recycling of cement paste, we assume that production occurs in a ~500kt/yr Electric Arc Furnace, creating 150kt/yr of slag, all of which is converted to clinker.We assume that the masses of flux and slag are the same, although generally, the mass of slag will be higher, as it combines the flux with impurities in the steel scrap.
The analysis aims to capture all changes to existing processing.Concrete and Demolition Waste is already collected, transported, processed and sold, so only the additional costs of improved separation are included, and are offset by higher revenues from purer recycled aggregates and a new stream of revenue from sand recycling.
The emissions figures are estimated from global average data, to avoid confusion from regional variations in electricity generation, or transport configuration.However, the cost figures are estimated for the UK in 2023, to ensure consistent comparison based on local energy, labour, cement and raw material prices.
The cost of capital for all additional equipment purchased to enable electric cement recycling is taken as 25%.
Given that cement production is largely localised, while steel recycling entails some trade of scrap, it has been assumed that recycled cement paste (RCP) must be transported 320 km to the electric arc furnace, and 320km from there to a cement distribution centre, with 80% of the transport occurring by ship and 20% by road.These distances would be typical for the UK, which currently has 11 cement plants and 2 large EAFs.
The key proportions, emissions factors and cost basis of the analysis are presented in Supplementary Table 5 Notes to Supplementary Table 5: 1.
The cost and emissions of calcined clay are expected to be around half the cost and emissions of OPC as production requires half the temperature for calcining.However, BGS fact sheets for Kaolin (2009) and brick-clay (2022) suggest the price might be nearer to £100 per tonne, and Alibaba spot prices are around $100-200 per tonne, so the £65/tonne estimate used here may be optimistic.Diaz et al (2017) table 1 give an emissions factor of 0.2 tonnes CO2/tonne clay for flash calcining.Berriel et al. (2016) estimate industrial energy requirements of 23 kg of crude oil and 16kWh of electricity for calcining 300kg of clay while producing one tonne of LC3-50 cement.Using global average emissions factors this leads to the estimate of 0.25 tonnes CO2/tonne clay used in the table.As clay calcination has no process emissions, this emissions factor could be reduced to zero if calcination were powered by emissions-free electricity.Our calculation below for scenario 2b, includes this assumption, leading to the maximum possible abatement from use of LC 3 -CEC.

2.
The emissions factor is taken from various secondary papers citing a 2008 WRAP study by Fisher; the cost estimate is uncertain and based on spot prices on alibaba.com,albeit Hanein et al (2018) estimate £10/tonne for gypsum; Riveiro et al (2016) in Table S6 estimate 474 kt CO2 to mine and process the gypsum required for 10.6 Mt content in plasterboard -giving an emissions factor of 0.05 tonnes CO2/tonne gypsum 3.
Meier et al ( 2005) quote a 1998 source to estimate 0.95-1.20 t CO2/tonne of lime.We used 1.1 as the middle of this range.

5.
Emissions from Proske et al ( 2016), sourced in turn from the GaBi database; price estimated at half the price of lime flux due to absence of processing 6.
Assumed to be half the price of lime flux, with the same emissions -no data found.

Detailed Calculation
Estimates are made of the cost and emissions consequences of four recipes for recycled Cambridge Electric Cement (CEC): (a) a "pure" blend with 95% CEC clinker and 5% gypsum and (b) an LC3-50 blend with 50% clinker, 30% calcined clay, 15% ground limestone and 5% gypsum, powered by (1) a grid having today's global average electricity emissions and (2) assuming complete decarbonisation of the electricity grid.
The detailed calculation for recipe 1b (LC3-50 CEC with today's grid) is provided in Supplementary Table 6.The equivalent calculations for the other recipes follow directly from application of the assumptions in Supplementary Table 5. Notes to Supplementary Table 6: 1.Following the analysis of Prajapati et al (2021), the treatment here assumes that prior to crushing and sieving, the CDW is heated, to enable high quality separation and a new revenue stream for recycled sand.The specific heat capacity of concrete is ~880 J/kgK (so 0.88MJ/tonne/K).We assume a 450K temperature increase (with electrical power) with efficiency of 85% gives ~466MJ per tonne heated.Assuming that concrete feedstock has mass ratios of 4:2:1 for aggregate:sand:cement, and that the resulting RCP has a 75% purity (concentration of cement) then 5.25 tonnes of CDW must be processed to deliver 1 tonne of recycled cement paste (RCP) (Figure 3B).The capital cost of the additional furnace and crushing/separating improvements is estimated at £6m.One additional operator is required.

2.
There could be a price premium for clean recycled aggregate compared with recycled aggregate today of around £10 per tonne, but this would require further processing of the crushed aggregate.It is therefore excluded from this analysis.

3.
New revenue stream for separator estimated as 50% of the price for new sales of sand at £15/tonne (half the price of new sand).This value is uncertain, so Supplementary Table 6 demonstrates a sensitivity analysis, varying the price of recycled sand between 0% and 100% of the price of new sand.

4.
Based on UK Industrial landfill charges

5.
A margin of approximately 25% of costs is added here to ensure a strong business case for commercial production of RCP.

6.
This level of substitution follows the results presented in the paper, although there is likely to be scope for optimisation in future.

7.
The results show that this may be necessary, although in reality it would be delivered through a reduced quality separation of RCP 8.
Costs assume capital purchase of £3m for the press, energy for operation, and one additional operator.Energy estimated as half the energy of crushing taken from Czigler et al (2020)

9.
Assumes that this saving is attributed to CEC where it would more likely be shared with the EAF steel 10.
Assumes £4m cost for the slag line plus one operator plus energy 11.
Assumes £1m cost for new grinding equipment plus one operator plus energy 12.This is the estimate of cost, without any profit margin -which depends on the comparison with existing cement, and the opportunity to attract a price premium arising from market preferences for lower emissions cement.
The results of the emissions calculation are compared with emissions and costs for ordinary Portland cement, and LC3-50 cement.Portland Cement is assumed to cause 0.80 t CO2/tonne cement and cost £100/tonne (Global Cement, 2011).LC3-50 is assumed to combine 50% OPC with 30% calcined clay and 15% ground limestone and 5% gypsum (Bishnoi et al. 2014, Scrivener et al., 2018) which using the data in Supplementary Table 5, leads to emissions of 0.48 t CO2/tonne cement.UK consumers are currently prepared to pay a price premium for lower carbon cement which, based on anecdotal evidence in the industry, we estimate to give a price of around £120/tonne for the 40% reduction in emissions in LC3-50.Potentially, this price premium will be higher for the greater emissions savings of electrically recycled cement and may rise higher as government policies impose stricter obligations on emissions (whether through a carbon tax or other mechanism) giving a greater market attraction to recycled cement.

Results
The four recipes for electrically recycled cement are compared to today's OPC and LC3-50 cement in Supplementary (The negative costs and emissions for producing clinker at the slag line in Supplementary Table 7 arise from the saving from avoiding use of conventional lime flux in the EAF.)Following note 9 above, the emissions calculation assumes that all benefits of substituting RCP for lime flux are credited to the production of cement.Hence, the emissions for "producing clinker at EAF slag line" above are all negative.In reality, some negotiation would take place as to how this benefit should be attributed between steel and cement production.
Figure 3A in the main paper is drawn from the numbers in Supplementary Table 5. N.b. the costs for the three recipes based on the new electric cement do not include a profit margin -where those for OPC and LC3-50 do.Supplementary Table 8 shows the sensitivity of the total costs and emissions from Supplementary Table 7 to the estimated price of recycled sand (note 3 to Supplementary Table 6).The results indicate that this price makes a significant difference to the profit potential of the new recycled cement, but even without any revenue, the new cement is competitive with existing OPC.

REFERENCE PRICES TODAY'S GRID ZERO-EMISSIONS GRID
Cost

Discussion
The energy for making Cambridge Electric Cement (CEC) (over and above that required for electric arc steel making) is dominated by heating prior to crushing for advanced RCP separation.We have assumed the worst case here but anticipate that with development the total energy input will be reduced.
The emissions of pure CEC with today's grid are dominated by electricity generation to power the heating required prior to separation.As the grid is decarbonised these will drop towards zero.The question of how the emissions benefit of using RCP instead of lime-flux are shared between the cement and steel industry is political and will evolve over time.
Once the grid is decarbonised, the emissions of making CEC clinker are almost entirely determined by the emissions associated with any conventional lime-flux blended into the RCP.The results of the paper suggest that this may be reduced to around 5% -although for now the results above assume 25%.
If CEC clinker is blended with calcined clay, while 25% lime is blended with RCP, the emissions of the resulting cement are increased compared with "pure" CEC cement, albeit less in scenario 2b where the clay is calcined in an electric oven powered by emissions-free electricity.However, as the lime-flux content is reduced, this advantage is overcome, and given that total CEC Clinker output will probably be restricted by total EAF capacity, it is likely that blending with calcined clay will remain advantageous.This prioritises further work on reducing the emissions of calcined clay production.
The costs of CEC clinker production are dominated by the costs of heating prior to RCP separation, and this depends on the price of electricity.When blending CEC clinker into a cement with calcined clay, the cost of the required calcined clay and anyhydrous calcium sulphate powder is approximately equal to those of the CEC Clinker but may reduce as the market for calcined clay expands.

Analysis of potential global scale of electric cement recycling
To estimate the global potential scale and abatement that could be delivered by the new process, we have built a simple model.It uses the International Energy Agency (IEA) scenario for future cement consumption as a baseline (IEA, 2009).Historical cement production is obtained from the US geological survey (USGS) with past clinker production from Andrew (2018).The clinker factor for 2020 is estimated to be 0.75, and is estimated to be close to one around 1980.Using the factors used in the previous section of analysis in this supplementary information, the impact of producing all cement as LC3 50 is first assessed.The future availability of calcium-rich RCP is then estimated to be the clinker production delayed by 50 years, assuming average building lives of 50 years (Dunant et al., 2021).Electricity emissions factors are assumed to decrease steadily to 30% of today's value by 2050.Future EAF recycling capacity is calculated using the IEA's iron and steel roadmap estimates (IEA, 2020b).
The resulting material production (of cement and clinker) and emissions arising from two scenarios are shown in Supplementary Table 9 and presented in figure 3C in the main paper: firstly, using only the available EAF capacity to process as much RCP as it can accept and secondly, assuming that additional dedicated EAF capacity is constructed to produce clinker but not steel and ensure that more of the available RCP is processed.In the attached spreadsheet, more ambitious scenarios are also computed for comparison.

Commercial
of the input flux and the lower bar shows the oxide composition of the resulting slag.The effect of leaching in the aluminium crucible is demonstrated by the significant changes in composition between flux and slag in the six pairs of results in the upper part of the figure.
Bond performance of reinforced concrete beams with electric arc furnace slag aggregates.Constr.Build.Mater.244, 118366 (2020).72.Bullerjahn, F. & Bolte, G. Composition of the reactivity of engineered slags from bauxite residue and steel slag smelting and use as SCM for Portland cement.Constr.

Table 4 :
Strength of the samples at 2, 7 and 28 days, tested according to EN 196-1.

Table 5 :
: Proportions, emissions factors and cost basis of the analysis.

Table 6 :
Detailed calculation of emissions and costs for recipe 1b for LC3-50 electric recycled cement made with today's global grid.

Table 7 :
Comparison of electrically recycled cement with current alternatives.

Table 8 :
Sensitivity of CEC cost to the value of recycled sand.