Direct mineralization of atmospheric CO₂ using natural rocks in Japan

Gas-solid mineralization of CO₂ using minerals as a climate change mitigation strategy was first researched in detail in the mid 1990s (Lackner et al 1995). The work failed to produce CO₂ mineralization in industrially relevant time scales due to the lack of humidity and a failure to appreciate the passivating nature of the product layer. Ion diffusion through the product layer, not gas phase diffusion or the mineralization reaction, is typically the rate-limiting step in engineered processes. Product layer passivation is conventionally modeled using the shrinking core model (SCM) per equation (1) (Yagi and Kunii 1955). The time ($t$: s) for a reaction to reach a depth ($\delta$: m) for a particle of radius ($r$: m) is a function of the density of Ca and Mg in the solid (${\rho _{{\text{solid}}}}$; moles/m³-rock), the gas phase concentration of CO₂ (${C_{{\text{gas}}}}$; moles/m³-gas), and the ion diffusivity through the product layer ($D$: m²/s).

$$\begin{equation}t = \frac{{{\rho _{{\text{solid}}}}{r^2}}}{{6D{C_{{\text{gas}}}}}}\left\{ {1 - 3{{\left( {\frac{{r - \delta }}{r}} \right)}^2} + 2{{\left( {\frac{{r - \delta }}{r}} \right)}^3}} \right\}\end{equation} \tag{ 1 }$$

Equation (1) indicates the relative effectiveness of different methods to accelerate CO₂ mineralization. Increasing the CO₂ concentration from atmospheric to pure CO₂ at standard temperature and pressure accelerates mineralization by ∼10^3.4. Grinding from 10 mm to 10 µm accelerates mineralization by 10⁶. In comparison, $D$ varies by a factor of ∼10⁷ for relevant crystalline minerals (Myers et al 2019). Additionally, the $D$ for most amorphous compounds is ∼10³–10⁵ times slower than compositionally equivalent crystalline minerals. Thus, rock selection is the critical step in achieving rapid gas-solid CO₂ mineralization. Further acceleration of CO₂ mineralization can be achieved simply and cheaply by grinding.

An additional consideration for direct air mineralization is the mixed nature of most rock formations. The bulk diffusivity (${D_{{\text{bulk}}}}$) of rock can be calculated from the constituent mineral diffusivities (${D_i}$) and volume concentrations (${\varphi _i}$) by effective medium theory (EMT) as shown in equation (2).

$$\begin{equation}\sum\limits_i {\left[ {{\varphi _i}\frac{{{D_i} - {D_{{\text{bulk}}}}}}{{{D_i} + 2{D_{{\text{bulk}}}}}}} \right]} = 0\end{equation} \tag{ 2 }$$

However, EMT does not account for instances when reactive minerals are completely occluded by neighboring unreactive minerals ('mineral locking'). Since mineral locking emerges from geometric relationships, it can be remedied by extensive grinding. Note that when particles are ground close to the internal crystal grain size, ${D_{{\text{bulk}}}}$ reverts to ${D_i}$. However, as particle diameters approach ∼1 µm, grinding tends to plastically disrupt the crystal structure producing amorphous minerals (Tromans and Meech 2001, Atashin et al 2015). As amorphous compounds exhibit much lower diffusivity, such extensive grinding reduces the CO₂ mineralization rate.

2.1.1. Verification of SCR $+$ EMT method

Gas-solid CO₂ mineralization is mechanistically equivalent for natural minerals and 'artificial' minerals arising from industrial processes. One such artificial mineral is the solid waste from steelmaking called 'slag'. Slag is typically a mixture of calcium silicates and aluminates, with significant quantities of iron, manganese, and magnesium (Piatak et al 2015). The mineralogy and crystallinity of slag varies based on composition and solidification conditions (Gautier et al 2013). Due to slag's substantial variability, it offers a stringent test of the predictive skill of SCM + EMT. The eight slags shown in table 1 were tested for CO₂ mineralization using the methods in Myers et al 2019. Figures 1(a)–(h) demonstrate the accuracy of the SCR + EMT model across the different mineral ensembles, crystallinities, particle sizes, and CO₂ concentrations. As the model is based on empirically determined, mineral-specific $D$, it can be applied to any mineral ensemble for which relevant data is available.

Zoom In Zoom Out Reset image size

Table 1. Elemental composition, mineralogical composition, and crystallinity of slags used to test SCM + EMT method.

	Slag #	1	2	3	4	5	6	7	8
Elemental composition	SiO2	42.1%	13.4%	38.8%	26.6%	32.8%	18.1%	34.1%	35.4%
	CaO	22.5%	44.4%	24.7%	29.6%	24.4%	53.8%	38.8%	37.8%
	MgO	1.4%	2.6%	1.6%	1.7%	1.6%	4.4%	7.1%	6.6%
	FeOa	15.0%	29.0%	15.7%	26.4%	23.4%	15.2%	0.1%	0.3%
	MnOb	8.0%	3.6%	7.6%	7.3%	8.8%	2.6%	0.1%	0.2%
	Al2O3	5.9%	2.6%	7.2%	4.0%	4.5%	2.1%	17.4%	16.4%
	Othersc	5.3%	4.5%	4.6%	4.4%	4.5%	3.7%	2.3%	3.3%
Mineral composition	Ca2SiO4	5.7%	23.4%	10.8%	19.4%	5.5%	24.5%	—	33.2%
	Ca3Al2O6	—	—	—	—	—	8.4%	—	—
	CaSiO3	24.9%	3.9%	17.9%	20.0%	17.3%	—	—	—
	MgSiO3	—	—	—	—	—	4.3%	—	6.0%
	FeSiO3	13.0%	—	0.7%	3.4%	9.7%	—	—	—
	Fe2SiO4	3.3%	—	0.4%	1.9%	2.0%	—	—	—
	Fe3O4	11.7%	20.8%	—	7.5%	7.8%	1.3%	—	0.7%
	FeO	—	—	—	—	—	9.4%	—	37.0%
	SiO2	12.3%	—	12.8%	7.3%	9.5%	—	—	—
	Ca0.82Fe0.18SiO2	20.5%	—	—	—	—	—	—	—
	Fe2O3	0.4%	—	9.7%	6.2%	7.1%	—	—	—
	Ca2Fe2O5	—	25.8%	—	6.0%	5.3%	9.1%	—	—
	(Ca/Fe)SiO3	—	—	15.0%	10.3%	—	—	—	—
	CaO	—	17.5%	—	2.4%	—	12.5%	—	4.9%
	Ca(Fe/Mg)Si2O6	—	—	24.7%	5.4%	9.5%	—	—	—
	Ca3Si2O7	—	—	—	—	6.2%	—	—	—
	CaFeSi2O6	—	—	—	—	9.7%	—	—	—
	Ca3Fe2Si3O12	—	—	—	—	2.7%	—	—	—
	Others	8.1%	8.5%	7.9%	9.9%	7.6%	30.4%	—	18.2%
	Amorphous	34.1%	27.6%	23.3%	17.4%	16.0%	25.2%	100%	25.6%

^aall Fe converted to FeO ^ball Mn converted to MnO ^cincluding phosphates and sulfates

2.1.2. Direct air mineralization of natural rocks

Using SCM + EMT, the CO₂ mineralization rate of several crystalline compounds was modeled to highlight both general trends and the importance of rock selection. Note that these represent a small selection of compounds present in actual rock formations. In figure 2, minerals were assumed to be ground in a vertical roller mill such that small particles (diameter < 3 µm) are elutriated to limit amorphization (CMIC 2016). The particle size number distribution can be approximated by a lognormal distribution characterized by a mean and standard deviation of the ${\text{ln}}$ of particle diameters of 0.35 µm and 0.66 µm, respectively (Kolmogorov 1940). Operating conditions were set to 30 ˚C, a relative humidity of 90%, and 415 ppm CO₂ air. Figure 2 provides the predicted CO₂ mineralization extent over one year. The logic behind targeting the oxide/hydroxide content of mine tailings as suggested by Zarandi et al 2017 is obvious from their rapid CO₂ mineralization rate. However, the scarcity of oxides/hydroxides means they are unable to provide the hundreds of gigatonnes of negative emissions required over the 21st century. Likewise, the rapid CO₂ mineralization displayed by calcium silicates supports the targeting of waste cement, ironmaking and steelmaking slag, and similar industrial wastes. However, the low natural abundance of calcium silicates limits their use for NETs. Also apparent from figure 2 is that the CO₂ mineralization rate of magnesium silicates and calcium aluminosilicates are significantly slower than that of oxides and calcium silicates. Figure 2 highlights that the specific mineral ensemble of a rock formation can dramatically alter not only its total CO₂ mineralization potential but also its CO₂ mineralization rate.

Zoom In Zoom Out Reset image size

The envisioned direct air mineralization system is shown in figure 3. It consists of a mining site and a 23 m-high, 1 km-long greenhouse whose width varies to accommodate the mine size. After clearing vegetation and removing soil, surface rocks are mined, transported, and finely ground. Field studies of mine tailings (Lechat et al 2016) and slag piles (Pullin et al 2019) show limited CO₂ mineralization over the course of years to decades. This relative lack of CO₂ mineralization is largely due to the pressure drop through the bulk solids impeding ingress of fresh air. To ensure adequate supply of CO₂, the direct air mineralization system uses a tiered geometry within the greenhouse to reduce the pressure drop. Additionally, fans provide a slightly negative pressure in the greenhouse to continuously draw in fresh air. A purpose-built solar photovoltaic system sits atop the greenhouse and is tied to a lithium ion battery storage system (PV + LiB) which provides the power for all activities in the direct air mineralization system. The amount of PV + LiB is based on operational energy consumption calculated in this section. Lifetime CO₂ intensity for purpose-built solar PV of 9.8 g-CO₂/kWh (Pehl et al 2017) and LiB of 61 Kg-CO₂/kWh of battery capacity with a 3,000 cycle lifetime (Emilsson and Dahllöf 2019) were assumed. The negative CO₂ emissions for the direct air mineralization system is calculated per the logic in figure 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

2.2.1. Energy for the extraction, handling, and grinding of rocks

2.2.2. Energy for fan operation

2.2.3. Materials production

2.2.4. Biological emissions and losses

The CO₂ mineralization potential of surface rocks in Japan was determined by reviewing geological maps that segregate Japan into 157 915 discrete regions of 127 surface rock types (Geological Survey of Japan/AIST 2020). Rock types lacking large quantities of Mg or Ca were removed. The original maps and literature were then used to manually set the ensemble of mineral compounds and the formation depth of the remaining locations. Areas containing pyrite were removed to avoid potential sulfuric acid dissolution of innate carbonates (Ross et al 2018). Densely populated areas were excluded to avoid public backlash (MLIT 2015). National parks and other areas protected under Japanese law from the erection of structures were removed from consideration (MOE Ministry of the Environment 2009). Areas less than 1 km² in footprint were also removed from consideration due the increased potential for mining large quantities of gangue. Topography was not used to exclude sites from analysis.

All sites with potentially suitable rocks are mapped in figure 5 (i.e. prior to exclusion of populated and protected areas) with quantitative data in table 2. As rocks are at the surface, the CO₂ mineralization potential is defined as a function of the depth of mining per year over the entire site: Gt-CO₂/(m × y). At a mining rate of 1 m y⁻¹, the CO₂ mineralization potential by rock type and site footprint are provided in figure 6 (note that the greenhouses designed in this study can accommodate a mining depth of 2 m/y). Very large sites are primarily volcanic and pyroclastic in origin. These rocks are more likely to have considerable amorphous content, and thus, slower gas-solid CO₂ mineralization rates. Figure 6(b) removes rocks of volcanic and pyroclastic origin.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Consistent with global analyses, the total potential CO₂ mineralization of surface rocks in Japan (17 069 Gt-CO₂) vastly exceeds NETs requirements regardless of long-term climate change mitigation goals. Surprisingly, excluding densely populated and legally protected areas only reduces the CO₂ mineralization potential by ∼9% (15 609 Gt-CO₂). Unsurprisingly for Japan, volcanic and pyroclastic rocks dominate the total CO₂ mineralization potential (∼90%). As such, the practical CO₂ mineralization potential of surface rocks in Japan hinges on the amorphous content of these rocks. Even so, focusing only on crystalline rocks in non-protected, sparsely populated areas still yields a CO₂ mineralization potential for Japanese surface rocks of 1525 Gt-CO₂.

Direct air mineralization applied to crystalline surface rocks in sparsely populated, un-protected areas in Japan yields an average and median net CO₂ removal efficiency of 59.8% and 65.0%, respectively. Figure 7 shows that the low efficiency is due almost entirely to incomplete mineralization of rock after 1 year. Since the CO₂ mineralization rate is controlled by the mineral-specific diffusivity, significant improvements are unlikely. After rocks are removed from the greenhouse, CO₂ mineralization may continue; however, the lack of a controlled gas-solid geometry and humidity will drastically reduce mineralization rates.

Zoom In Zoom Out Reset image size

Carbon dioxide emissions from operations and materials manufacture reduces the net CO₂ removal by ∼1.6%, the bulk of which is attributable to the energy for extraction, movement, and fine grinding of rock. The average and median energy consumption are 1.51 and 1.45 GJ t⁻¹-CO₂, respectively. This is substantially less than the ∼4.7–10 GJ t⁻¹-CO₂ of direct air capture and storage methods (Creutzig et al 2019). Based on current PV power output applied in Japan (GSA 2020), PV panels mounted to the greenhouse roof would supply an average of ∼1.7 times more energy than needed for operations. Higher efficiency PV + LiB with lower CO₂ footprints are likely in the coming years, only increasing the system performance. More efficient fine grinding systems are theoretically possible, but unlikely given the mature nature of the mining industry.

Emissions from removal of vegetation and soil reduce the net CO₂ removal efficiency by only ∼0.34% thanks to the small footprint of the direct air mineralization system. The average and median land requirements are 1.13 km² and 1.08 km² per Mt-CO₂ removed per year, roughly equivalent to direct air capture and storage methods. This area is equally split between the mining site and greenhouse. The average payback period of the combined carbon debt from all sources is 60 days.

In increasing order of size, with a mining depth of 2 m y⁻¹, yields the net CO₂ emissions in Japan as shown in figure 8 (averaged over 50 years of operation). Maximum negative emissions are ∼7.6 Gt-CO₂/y. For allocation of negative emissions quotas to Japan, we use the multi-criteria approach of Pozo et al 2020 to set a lower bound of 11.5 Gt-CO₂ by 2100. We set the upper bound to 98.9 Gt-CO₂ by 2100, equal to all emissions from 1850 to 2100 assuming emissions reductions inline with Japan's mid-century strategy and assuming linear reductions thereafter to a net zero CO₂ society (Government of Japan 2019). This upper bound represents an aspiration to return to the pre-industrial revolution climate. Meeting these negative emissions targets by 2100 requires 0.144–1.236 Gt-CO₂/y of negative emissions if started in 2021. This would require ∼80–619 km² spread across 74–393 sites.

Zoom In Zoom Out Reset image size

The mass of carbonated rock powder produced by direct air mineralization as a function of negative emissions is shown in figure 9(a). This rock powder is similar in size to clays often used in mechanically stabilized earth construction, backfill, and general infrastructure projects. Recent regulations (ASTM A1115/A1115M-19 2019) support such earth-based construction methods in accordance with the United Nations Sustainable Development Goal 12. Japan has a long history of large-scale land reclamation projects using rocks. For example, the building of artificial islands in Osaka Bay from rocks obtained by nearby mountains utilized ∼263 million m³ of fill material to create 15.59 km² of land from 1953–1981 (Tanaka et al 1983). Figure 9(b) shows the potential for land reclamation using carbonated rock powder. Other potential usage pathways include breakwaters, seawalls, foundations for buildings and roadways, and filler in concrete.

Zoom In Zoom Out Reset image size