Iron is doing two jobs here, and that is the part worth reading twice. Fe(II) absorbs ultraviolet C light and throws off solvated electrons that reduce the molybdenum catalyst [2]. The Fe(III) left behind absorbs ultraviolet light too, and gets reduced back to Fe(II) by pulling electrons off water [3]. So the oxidative half-reaction is not a separate module bolted on; it is the spent form of the reductant regenerating itself, which is why the authors can describe the system as coupling water oxidation to nitrogen reduction rather than consuming a sacrificial donor [3].
The 7.6% figure in the abstract is a consistency check, not a solar rating. The authors calculate the overall process to be endoergonic by 329.2 kJ per mole and report that this agrees with a light-to-chemical energy conversion efficiency of 7.6% [6]. That is energy stored per photon energy absorbed, under ultraviolet C irradiation [2].
The pressure experiment is the operationally interesting one. Ambient-pressure quantum yield is 5.0%; raising the dinitrogen pressure lifts it to 10.8%, which the authors attribute to nitrogen's low solubility in water limiting catalysis [1][4]. That is a factor of 2.16 [9], and read the other way, the ambient run recovers only 46% of the per-photon yield the same chemistry delivers when the gas is actually in solution [11]. Nothing about the catalyst changed between those two numbers. What changed was how much substrate the liquid was holding.
So the bottleneck has moved from bond activation to mass transfer, and mass transfer is a problem with a well-populated toolbox: membranes, bubble columns, thin films, higher headspace pressure. It is also a problem with a compressor attached to it, which is the cost that reappears the moment you stop using ambient nitrogen.
The feedstock tolerance is the other claim that changes what a flowsheet would look like. The reactivity is reported to run in unpurified seawater, and in ambient air [5]. Purification of the water and separation of the nitrogen are the two front-end unit operations that make conventional routes capital-heavy, and this paper says neither was required to see turnover.
Verification is metered. Data for four of the figures are deposited on Zenodo with source data provided, while data behind twenty-six supplementary figures are available from the authors on request [8]. The full text sits behind a $39.95 single-article charge or a $259 annual subscription [7]. Anyone wanting to check the quantum yields against raw photon counts is buying in first.
