The low oxygen does two jobs at once. The same shortage that keeps carbon in the silicate melt rather than the metal is what leaves the core carbon-poor, and the graphite crystallising out of that melt is too light to sink, so it collects at the top of the magma ocean [5][6]. The coupling matters more than either result on its own, because it means the crust and the core are not independent knobs. You cannot thicken the floating graphite layer without cleaning out the core.
Which is where the arithmetic turns against carbon. MESSENGER's geodesy needs several percent of light elements in a core that is about 70% of Mercury's mass [10], and the experiments allow at most 0.5% carbon there [11]. Half a percent of seventy percent is 0.35% of the whole planet [14]. Read "several" as three and carbon covers at most a sixth of the shortfall [15]. Namur is explicit that the concentration is far too low to account for the deficit [11], and these runs do not nominate a replacement.
The crust figure deserves a more careful read than the core figure. Formation conditions were fixed by comparing the modelled carbon budget across core, mantle, crust and primitive atmosphere against a crust thickness calculated from MESSENGER data [8]. So the 40 to 120 metre range, and its consistency with the 1% to 3% surface carbon [7], partly reflects the observation that pinned the oxygen fugacity rather than an independent hit. The core carbon ceiling is the number that was not tuned to anything, which is what makes it worth testing.
Two things the experiments cannot settle. The primordial graphite crust was later disrupted and redistributed by meteorite impacts and by the volcanism that built the younger crust [9], so nothing on the surface now should be an intact 40 to 120 metre layer, and the measured 1% to 3% is a reworked remnant of one. And there is no Mercury in anyone's collection: no returned sample, no meteorite tied to the planet [12], so bulk composition comes from telescopes plus Mariner 10 in 1973 and MESSENGER in 2011 [13]. The runs covered about 920 degrees, roughly 1,250 C to 2,170 C, at pressures matching planetary interiors [4][16], which brackets metal-silicate separation rather than pinning one instant of it.
BepiColombo is entering the final phase of its journey [1], and this account of the work does not say which of its measurements would catch a carbon-rich core [17]. Under 0.5% is still a specific enough number to be proved wrong, and having it on the record before arrival is worth more than a tidier story about accretion.
