Olivine armours itself. As iron leaves the mineral in wet, oxidising conditions, iron oxides precipitate back onto the surface and shut down further dissolution [9]. Siderophores are the counter: secreted molecules that chelate and solubilise ferric iron, and the siderophore desferrioxamine had already been shown in batch reactions to speed olivine dissolution by dissolving those passivating oxides [2][10].
The obstacle to running that trick continuously was genetic. In *Alteromonas macleodii*, the genes for the siderophore petrobactin switch on in response to iron limitation, as they do in most siderophore producers [4]. A tank packed with olivine is not iron-limited, so the bacteria stop making the useful molecule at the moment the process most needs it, and the authors confirmed experimentally that this natural regulation rules out continuous production in mineral bioreactors [3]. That is why the answer was a rewired strain rather than a better culture protocol [5].
The reactor format matters as much as the strain. Olivine dissolution generates alkalinity that is mostly stable bicarbonate drawn from atmospheric CO2 [14], which is precisely why the usual industrial way of dissolving minerals is unavailable here: chemical or biological acid leaching precludes alkalinity generation [13]. And in field trials that spread mafic rock on soils or coastlines, removal has been hard to quantify because the weathering is slow [11]. A tank fixes hydrology and mineral surface area, which is what makes the carbon accounting tractable [12].
Here is the arithmetic. At 0.50 g of CO2 per day [8], one of these reactors removes about 183 g in a year [1]. Reaching one tonne per year at that rate takes roughly 5,500 reactors of the same size [2]. Per kilogram of the 4 kg mineral charge, the measured rate is 0.125 g of CO2 per day [3]. So the scaling unit is reactor volume and mineral surface area, and this result describes a measured, engineered rate in a metered format, distinct from a delivered tonne of CO2 removed.
The reported figures leave two things unstated. The abstract gives a 2.6-fold ratio for dissolution rate and an absolute 0.50 g per day for the engineered reactors, so the CO2 removal ratio between engineered and unmodified cells is not stated [5]. And neither quantity is a cost per tonne [4]. The life-cycle analysis points at where that cost will live: renewable feedstock and minimal replenishment of the modified cells are the conditions for net removal at scale [6]. The strain work is the part that has been demonstrated; the acetate supply chain is the part that decides whether the ledger closes. Seawater is the cheap term, since it is abundant and can be returned to the environment carrying alkaline ions [15], and the authors are explicit that silicate weathering on its own is too slow for industrial deployment [1].
