forms during Li deposition in THF most likely
leads to much slower diffusion as compared
to that in aqueous media ( 36 ). Alongside rate-
limiting diffusion, we further consider mass
balance of the incoming protons toward either
H 2 or NH 3 and thus distinguish between two
possible mass-transfer limitations for the nitro-
gen reduction reaction (NRR): a limitation in
nitrogen (Nlim), or a limitation in protons (Hlim).
Within these two regimes, the steady-state rate
of NH 3 production and resulting FE can be
formulated in terms of only the three elemen-
tary (rLi,rH,rN 2 ) diffusion rates. The heatmap
of Fig. 3A shows how the predicted FE changes
as a function of the relative-to-Li rates of in-
coming N 2 (xaxis) and H+(yaxis). A dashed
green line marks the idealrN 2 torHratio that
leads to the highest FE at the boundary be-
tween nitrogen- and proton-limited regimes.
When adding oxygen to the gas feed, it is
intuitive to expect that the NRR will be sup-
pressed because of additional competition
from the ORR. The experimental improve-
ment in FE at low O 2 content, however, is a
counterintuitive result that is much harder to
understand. As a first step toward this under-
standing, we extend our original kinetic model
to incorporate the oxygen electrochemistry
and examine its effect on the resulting FE.
This involves not only the inclusion of the
ORR that consumes protons and electrons at
the catalyst surface, but also the probability
that newly formed H 2 Ocanitselfactasa
proton donor to further drive the catalysis.
Derivation of the enhanced model can be
found in the supplementary materials; we em-
phasize that none of the situations that we
considered can account for an oxygen-induced
improvement in FE as found experimentally.
Indeed, we predict a (likely small) decrease
in FE at low O 2 content because of the ORR
competing as a parasitic side reaction, and
a substantial drop to≈0 when the more reac-
tive oxygen species dominate all surface sites
over nitrogen species at higher O 2 contents.
This result suggests that O 2 must have another
way of influencing the LiNR system.
What can increase the FE is a decrease in
the Li diffusion rate (Fig. 3A). This is simply
because Li electrodeposition also competes
with the NRR for electrons. Ideally, one would
want Li diffusion or deposition to be slow
enough that it consumes a minimal number
of electrons while still providing a full mono-
layer of clean, freshly plated Li that is reac-
tive enough to dissociate nitrogen. Delayed
Li diffusion as a result of oxygen-induced
changes in the SEI layer has, indeed, been ob-
served. Wanget al. recently showed that O 2
can specifically influence the SEI composition
in nonaqueous lithium–air batteries, leading
to increased SEI homogeneity as well as di-
minished Li+diffusivity ( 31 ). Such an effect
on Li+diffusivity should have a substantial
impact on the FE toward forming ammonia.
If the rates of nitrogen and proton diffusion,
rN 2 andrH, remain unaffected by the oxygen
content, a decrease in the Li+diffusion rate will
move the operating point along the dashed
diagonal in Fig. 3A, leading to a substantial
increase in the FE. In reality,rN 2 andrHmay
also change as a result of the oxygen-induced
modification of the SEI layer, but as long as
thedecreaseislessthanthatforrLi,anetin-
crease in the FE is expected, as indicated by
the conical shaded area in Fig. 3A.
The improvement in FE, also shown in
Fig. 3B, is much larger than what can be ex-
pected from moving either horizontally (by,
e.g., changing N 2 pressure) or vertically (by,
e.g., changing proton activity) across the plot.
According to this picture, an optimal O 2 con-
tent should follow as a trade-off between FE
increasing at low values because of restricted
Li+diffusivity and decreasing at high values
when the ORR starts dominating over theNRR. Furthermore, both the optimal O 2 con-
tent and resulting maximum FE should cor-
relate directly with N 2 pressure when the NRR
is in the nitrogen-limited regime. As seen from
Fig. 3A, increasing N 2 pressure (i.e., moving
horizontally to the right on the heatmap) in-
creases FE until the NRR is pushed into the
proton-limited regime. The optimal O 2 content
is defined by the competition between NRR
and ORR and, while increasing N 2 pressure in
the nitrogen-limited regime, the maximum FE
should occur at same absolute O 2 partial pres-
sure or a proportionally lower O 2 content, exact-
ly as observed experimentally (Fig. 1).
Compositional changes in the SEI layer in
thepresenceofO 2 are also consistent with
the observed improvement in stability. For
Li-ion batteries, the addition of O 2 has been
shown to form a more homogeneous and uni-
form SEI layer, leading to increased cyclability
( 31 ). A more homogeneous layer could lead to
more uniform Li plating in the LiNR system,1596 24 DECEMBER 2021•VOL 374 ISSUE 6575 science.orgSCIENCE
Fig. 4. XPS and XRD data of the working electrode after electrochemistry without exposure to air.
Scans of three different WE foils postelectrochemistry with a constant−4 mA/cm^2 applied under 20-bar N 2
with varying amounts of O 2 present in the atmosphere. Foils were transferred without air exposure to
the respective systems. (AtoC) XPS spectra of N 1s scanning window for samples with 0.0 mol % (A),
0.8 mol % (B), and 3.0 mol % O 2 (C) in the reaction atmosphere. More high-resolution scans and full survey
spectra are included in figs. S17 and S18. The N 1s peak at 0.8 mol % O 2 was seen in two independent
measurements. (D) XRD diffractograms of samples with 0.0, 0.8, and 3.0 mol % O 2 in atmosphere and pattern
of the PEEK dome transfer system, which creates reflections in the 15° to 30° region. Reference patterns for
Mo (ICSD: 998-005-2267), Li 2 O (ICSD: 98-017-3206), LiOH (ICSD: 98-002-6892), LiClO 4 (ICSD: 98-016-5579),
LiClO 4 3H 2 O (ICSD: 98-003-2534), and Li 3 N (ICSD: 98-003-4779) are included.RESEARCH | RESEARCH ARTICLES