fig. S5. The electrolyte was 0.3 M LiClO 4 in tetra-
hyrofuran (THF) with 1.0 vol % EtOH, unless
otherwise specified. The working electrode
(WE) was a Mo foil of 1.8 cmgeo^2 area, the CE
was a 1.8 cmgeo^2 Pt mesh, and the pseudo-
reference electrode (RE) was a Pt wire. The
potential scale was determined by calibrating
the Pt pseudo-RE to the well-known Fc/Fc+
couple (fig. S6). A constant-current density of
−4 mA/cm^2 was applied, until the system either
overloaded because of an increased total cell
potential beyond the capacity of the potentio-
stat, or the experiment reached 50 C of total
passed charge. A representative cyclic voltam-
metry is shown in fig. S7. Cycling experiments
according to Andersenetal.( 24 ) were also
conducted, to show the beneficial coupling of
the improved stability due to cycling with the
increased FE from the addition of O 2. The FE
was determined at the end of the experiment
and therefore represents accumulated FE. A
more detailed description of the experimen-
tal procedure and all experimental results can
be found in the supplementary materials.
Figure 1 shows that small amounts of O 2
markedly increase the FE toward ammonia
production at both 10 and 20 bar, showing
peak behavior from 0.5 to 0.8 mol % O 2 in N 2
at 20 bar and 1.2 to 1.6 mol % at 10 bar. The
optimum oxygen content at 10 bar is higher
than at 20 bar by almost exactly a factor of 2,
suggesting that the oxygen partial pressure,
rather than the molar ratio of O 2 in N 2 , is re-
sponsible for the increase in FE (fig. S8). The
efficiency drops rapidly at higher amounts
of O 2 , in accordance with the much higher
oxygen-content experiments (20% O 2 , 50 bar)
of Tsunetoetal.( 23 ). An increase in the H 2 O
content of the electrolyte postelectrochem-
istry is observed with increasing O 2 content
(fig. S9), suggesting that ORR is responsible
for the decrease in FE. The decrease in the FE
at higher O 2 content could also partly be due
to the formation of passivating Li 2 O species on
the surface of the WE as seen later by XRD. As
expected, the EE of the system follows a trend
similartothatoftheFEwithpeaksat8.0±
1.8% and 11.7 ± 0.5% at optimal O 2 content
under 10 bar and 20 bar, respectively (figs.
S10 and S11).
Effect of oxygen on stability
In addition to the notable increase in the FE
of the system because of added O 2 , an increase
in the apparent stability of the total cell poten-
tial was also observed. However, the electro-
chemistry will overload given enough time,
and the only way to ensure stability in this sys-
tem is through potential cycling ( 24 ). Figure 2A
shows representative electrochemistry data of
chronopotentiometry measurements at 20 bar,
where it is clear that the WE potential re-
mains stable for longer times with increas-
ing O 2 content. The CE remains stable, even
for experiments in which the WE is unstable.
This excludes the possibility of oxygen having
an appreciable influence on the CE. To com-
pare the stabilities, we plotted the data against
the starting WE potential of each individual
experiment. The total time it takes for the
WE potential to decrease by 1 V was taken
as a measure of stability (tstable). Because all
experiments exceeding 0.8 mol % O 2 exhibited
a WE potential that did not decrease by more
than 1 V,tstablewas set to 116+ min for these
experiments.tstableis plotted as a function
of the O 2 content in Fig. 2B, and a linear
trend of increasing stability by increasing O 2
is observed. Stability remained high even
forlargerO 2 content when FE diminished.
This relation can be visualized in fig. S10A
and suggests that LiNR stability is qualita-
tively related to O 2 content but not to the FE.
Cycling of the potential to prevent build-up
of passivating lithium species is a previously
proven method to achieve long-term stability
(days) of the potential in the LiNR system
( 24 ). This method was applied for an ex-
periment with optimal O 2 content (20 bar
N 2 , 0.67 mol % O 2 ) and one with suboptimal
O 2 content (20 bar N 2 ,0.26mol%O 2 ), which
would otherwise be unstable under con-
stant Li deposition. We found that cycling the
potential in the presence of O 2 retains the
higher FE (fig. S9) while also stabilizing the
WE during the 0.26 mol % O 2 experiment until
(at least) 50 C of charge has been passed (figs.
S12 and S13). In addition, the water content of
these cycling experiment increased substan-
tially compared to the constant-current exper-iments (table S1), owing to the constant ORR.
However, because the FE seems to be un-
affected by this result, it is believed that the
increase in water on this scale is not detri-
mental to the system and that by applying cy-
cling on the optimum oxygen content of 0.6
to 0.8 mol % at 20 bar, the reaction could be
kept stable for longer (10 hours), as was shown
previously ( 24 ).Mechanistic role of O 2 in LiNR
We recently demonstrated that in the absence
of oxygen, the LiNR system and its FE toward
ammonia are largely controlled by mass trans-
port limitations of the reactant species diffus-
ing from the bulk electrolyte to the catalyst
surface ( 24 ). By developing a kinetic model,
we showed that the experimental FE is a sim-
ple function of the relative diffusion rates be-
tween incoming lithium ions (rLi), protons (rH),
and nitrogen (rN 2 ). Here, we assume a steady-
state FE, although there might be dynamical
changes especially initially where the SEI layer
is formed. Our model considered a series of
elementary steps and relied on the basic, yet
reasonable, assumption that all diffusion steps
are considerably slower than subsequent reac-
tion steps taking place at the surface, i.e., they
are overall rate-limiting. This assumption re-
sults from recognizing that (i) electrochemical
surfacestepsaremostlikelyveryfastand
driven irreversibly forward because of the ex-
treme reducing potentials required to plate Li
( 33 ); (ii) the chemical (dissociative) adsorption
of N 2 on metallic Li has a negligible kinetic
barrier ( 34 , 35 ); and (iii) the SEI layer thatSCIENCEscience.org 24 DECEMBER 2021¥VOL 374 ISSUE 6575 1595
Fig. 3. Graphic illustration of the microkinetic model for Li-mediated ammonia synthesis.(A) Heatmap
of the predicted FE as a function of the ratio of nitrogen to lithium (xaxis) and proton to lithium (yaxis)
diffusion rates. The red star indicates the expected location of the 10-bar experiments without O 2 in the
system. The purple star indicates the improvement in FE ifrLiwere selectively lowered by an order of
magnitude. The cone represents the uncertainty of the location of the purple star. (B) A one-dimensional plot
of FE cut along the optimalrN 2 /rHratio (marked by the green dashed line in Fig. 3A) that clearly shows
the marked FE improvement that can result from loweringrLi. Thexaxis is normalized so that the red star
corresponds to 1/rLi= 1.RESEARCH | RESEARCH ARTICLES